Geophysical

MonographSeries
Including
IUGG Volumes

MauriceEwingVolumes
MineralPhysics
 GEOPHYSICAL MONOGRAPH SERIES

GeophysicalMonograph Volumes 21 QuantitativeModelingof Magnetospheric

Processes W. P. Olson (Ed.)
1 Antarcticain the InternationalGeophysical 22 Derivation,Meaning,andUseof Geomagnetic
Year A. P. Crary,L. M. Gould,E. O. Hulburt, Indices P. N. Mayaud
HughOdishaw,andWaldoE. Smith(Eds.) 23 TheTectonicandGeologicEvolutionof
2 Geophysicsand the IGY HughOdishaw and Southeast Asian Seas and Islands Dennis E.
StanleyRuttenberg (Eds.) Hayes(Ed.)
3 AtmosphericChemistry of Chlorine and Sulfur 24 Mechanical Behavior of Crustal Rocks: The
Compounds James P. Lodge,Jr. (Ed.) Handin Volume N. L. Carter,M. Friedman,
4 ContemporaryGeodesy CharlesA. Whittenand J.M. Logan,
andD. W. Stearns
(Eds.)
Kenneth H. Drummond (Eds.) 25 Physics
of AuroralArcFormationS.-I.Akasofu
5 Physicsof Precipitation -HelmutWeickmann andJ. R. Kan (Eds.)
(Ed.) 26 Heterogeneous
Atmospheric
Chemistry David
6 The Crust of the Pacific Basin Gordon A. R. Schryer
(Ed.)
Macdonald and Hisashi Kuno (Eds.) 27 The Tectonicand GeologicEvolutionof
7 Antarctica Research: The Matthew Fontaine Southeast Asian Seas and Islands: Part
Maury Memorial Symposium H. Wexler,M. J. 2 DennisE. Hayes(Ed.)
Rubin,andJ. E. Caskey,
Jr. (Eds.) 28 MagnetosphericCurrents Thomas
A. Potemra
8 Terrestrial Heat Flow William H. K. Lee (Ed.) (E•t.)
9 Gravity Anomalies:UnsurveyedAreas Hyman 29 ClimateProcesses
andClimateSensitivity
Orlin (Ed.) (MauriceEwing Volume5) James
E. Hansenand
10 The Earth Beneath the Continents: A Volume of Taro Takahashi(Eds.)
Geophysical Studies in Honor of Merle A. 30 MagneticReconnection
in SpaceandLaboratory
Tuve JohnS. Steinhartand T. Jefferson
Smith Plasmas EdwardW. Hones,Jr. (Ed.)
(Eds.) 31 PointDefectsin Minerals(MineralPhysics
11 IsotopeTechniques in the Hydrologic Volume 1) RobertN. Schock(Ed.)
Cycle GlennE. Stout(Ed.) 32 The CarbonCycleandAtmospheric
CO2:
12 The Crust and Upper Mantle of the Pacific Natural Variations Archean to Present E.T.
Area LeonKnopoff,CharlesL. Drake,and SundquistandW. S. Broecker
(Eds.)
Pembroke
J. Hart (Eds.) 33 GreenlandIce Core:Geophysics, Geochemistry,
13 The Earth'sCrust and Upper Mantle Pembroke and the Environment C. C. Langway,
Jr., H.
J. Hart (Ed.) Oeschger,andW. Dansgaard(Eds.)
14 The Structureand PhysicalPropertiesof the 34 Collisionless
Shocksin the Hellosphere:A
Earth's Crust JohnG. Heacock(Ed.) Tutorial Review Robert G. Stone and Bruce T.
15 The Use of Artificial Satellites for Tsurutani (Eds.)
Geodesy SorenW. Henricksen,
ArmandoMancini, 35 Collision!essShocksin the Heliosphere:Reviewsof
and Bernard H. Chovitz (Eds.) Current Research Bruce T. Tsurutani and Robert G.
16 Flow and Fracture of Rocks H.C. Heard, I. Y. Stone(Eds.)
Borg,N. L. Carter,andC. B. Raleigh(Eds.) 36 Mineral and RockDeformation:Laboratory
17 Man-Made Lakes: Their Problems and Studies--The Paterson Volume B.E. Hobbs
Environmental Effects William C. Ackermann, and H. C. Heard (Eds.)
GilbertF. White,andE. B. Worthington (Eds.) 37 EarthquakeSourceMechanics(MauriceEwing
18 The Upper Atmospherein Motion: A Selection Volume6) Shamira Das,JohnBoatwright,
and
of PapersWith Annotation C.O. Hinesand Christopher
H. Schotz(Eds.)
Colleagues 38 Ion Acceleration
in the Magnetosphere and
19 The Geophysicsof the PacificOceanBasin and Ionosphere TomChang(Ed.)
Its Margin: A Volume in Honor of George P. 39 High PressureResearch in Mineral Physics
Woollard George H. Sutton,Murti H. Manghnani, (Mineral PhysicsVolume 2) Murli H.
andRalphMoberly(Eds.) Manghnani andYasuhiko Syono(Eds.)
20 The Earth's Crust: Its Nature and Physical 40 Gondwana Six: Structure Tectonics, and
Properties JohnC. Heacock(Ed.) Geophysics GaryD. McKenzie
41 GondwanaSix:Stratigraphy,Sedimentology, 62 ModelingMagnetospheric
PlasmaProcesses
and Paleontology GarryD. McKenzie
(Ed.) GordonK. Wilson (Ed.)
42 Flow and TransportThroughUnsaturated 63 Marine ParticlesAnalysisand Characterization
FracturedRock DanielD. Evansand Thomas
J. DavidC. HurdandDerekW. Spencer (Eds.)
Nicholson(Eds.) 64 Magnetospheric Substorms Joseph R. Kan,
43 Seamounts, Islands, and Atolls BarbaraH. ThomasA. Potemra,SusumuKokubun,and
Keating,
Patricia
Fryer,Rodey
Batiza,
andGeorge Takesi
Iijima(Eds.
)
W. BoehIert(Eds.) 65 ExplosionSourcePhenomenologySteven R. Taylor,
44 Modeling Magnetospheric Plasma T.E. Moore HowardJ. Patton,andPaul G. Richards
(Eds.)
andJ. H. Waite,Jr. (Eds.) 66 Venusand Mars:Atmospheres, Ionospheres,and
45 Perovskite: A Structure of Great Interest to SolarWind Interactions JanetG. Luhmann,Mariella
Geophysics andMaterialsScience Alexandra Tatrallyay,
andRobertO. Pepin(Eds.)
Navrotsky andDonaldJ. Weidner
(Eds.) 67 High-PressureResearch: Applicationto Earthand
46 Structureand Dynamicsof Earth'sDeep PlanetarySciences (Mineral PhysicsVolume3)
Interior (IUGG Volume 1) D.E. Smylie and Yasuhiko
SyonoandMurli H. Manghnani (Eds.)
Raymond Hide(Eds.) 68 MicrowaveRemoteSensingof SeaIce Frank
47 Hydrological RegimesandTheirSubsurface Carsey,
RogerBarry,fosefino
Comiso,
D. Andrew
Thermal Effects (IUGG Volume 2) Alan E. Rothrock,
Robert
Shuchman, W. TerryTucker,
Wilford
Beck,GrantGarven,andLajosStegena(Eds.) Weeks,and Dale Winebrenner
48 Origin and Evolutionof Sedimen• Basins 69 SeaLevelChanges: Determinationand Effects
and Their Energyand Mineral Resources (IUGG Volume11) P. L. Woodworth, D. T. Pugh,
(IUGG Volume3) Raymond A. Price(Ed.) J. G. DeRonde,
R. G. Warrick,and•. Hannah
49 Slow Deformation and Transmission of Stress 70 Synthesisof Resultsfrom ScientificDrillingin the
in the Earth (IUGG Volume 4) StevenC. Cohen Indian Ocean RobertA. Duncan, David K. Rea,Robert
and Petr Vanidek(Eds.) B.Kidd,UlrichvonRad,andJeffrey
K. Weissel
(Eds.)
50 DeepStructure
andPastKinematics
of 71 Mantle Flow and Melt Generation at Mid-Ocean
AccretedTerranes(IUGG Volume 5) JohnW. Ridges Jason
Phipps
Morgan,DonnaK. Blackman,
Hillhouse (Ed.) andJohnM. Sinton(Eds.)
51 Properties
andProcesses
of Earth'sLowerCrust 72 Dynamics
of Earth'sDeepInteriorandEarth
(IUGG Volume6) RobertF. Mereu,Stephan Rotation(IUGG Volume 12) Jean-Louis
LeMou•t,
Mueller,andDavidM. Fountain(Eds.) D.E.Staylie,
andThomas Herring
(Eds.)
52 UnderstandingClimateChange(IUGG Volume 73 Environmental Effectson Spacecraft
Positioning
and
7) AndreL. Berger,
Robert
E. Dickinson,
andJ.Kidson Trajectories
(IUGG Volume13) A. VaIlance Jones
(Eds.) (Ed.)
53 Plasma Waves and Istabilities at Comets and in 74 Evolution of the Earth and Planets (IUGG Volume
Magnetospheres
Bruce
T. Tsurutani
andHiroshi 14) E. Takahashi,
Raymond
JeanIoz,
andDavidRubie
Oy• (Eds.) (Eas.)
54 SolarSystemPlasmaPhysicsJ•.H. Waite,
Jr., 75 InteractionsBetweenGlobal ClimateSubsystems:
J. L. Burch,andR. L. Moore(Eds.) TheLegacy
of Hann(IUGG Volume15) G.A.
55 Aspects
of ClimateVariabilityin thePacific McBean and M. Hahtel (Eds.)
and Western Americas DavidH. Peterson (Ed.) 76 RelatingGeophysical
Struchxres
andProcesses:
The
56 The Brittle-Ductile Transition in Rocks A.G. JeffreysVolume(IUGGVolume16) K.A]a'andR.
Duba,W. B. Durham,J. W. Handin,andH. F. Dmowska(Eds.)
Wang(Eds.) 77 The MesozoicPacific:Geology,Tectonics,and
57 Evolutionof Mid OceanRidges(IUGG Volume Volcanism
M. S. Pringle,W. W. Sager,
W. V. Sliter,
8) JohnM. Sinton(Ed.) and S. Stein (Eds.)
58 Physics
of Magnetic
FluxRopesC.T. Russell, 78 ClimateChangein Continental
IsotopicRecords
E. R. Priest,andL. C. Lee(Eds.) P. K. Swart,K. C. Lohmann,
jr.McKenzie,
andS. Savin
59 Variations in Earth Rotation (IUGG Volume
9) Dennis
D. McCarthy
andWilliams
E. Carter
(Eds.)
60 Quo VadimusGeophysics
for theNext
Generation
(IUGG Volume 10) GeorgeD.
Garland
andJohnR. Apel(Eds.)
61 CometaryPlasmaProcessesAlanD.
Johnstone
(Ed.)
MauriceEwing Volumes 6 Propertiesand Processes
of Earth'sLower
1 IslandArcs,Deep SeaTrenches,and Back-Arc Crust Robert
F. Mereu,Stephan
Mueller,and
Basins Manik Talwani and Walter C. Pitman III David M. Fountain (Eds.)
(Eds.) 7 UnderstandingClimateChange AndreL.
2 DeepDrilling Resultsin the AtlanticOcean: Berger,
RobertE. Dickinson,
andJ. Kidson
(Eds.)
OceanCrust ManikTaIwani,Christopher G. 8 Evolutionof Mid OceanRidges JohnM.
Sinton (Ed.)
Harrison,
andDennisE. Hayes(Eds.)
9 Variations in Earth Rotation Dennis D. Mc-
3 Deep Drilling Resultsin the AtlanticOcean:
ContinentalMarginsand Paleoenvironment CarthyandWilliamE. Carter(Eds.)
ManikTalwani,WilliamHay, andWilliamB. F. 10 Quo VadimusGeophysics for the Next Genera-
Ryan(Eds.) tion GeorgeD. GarlandandJohn R. Apel(Eds.)
11 SeaLevelChanges:Determinations and Effects
4 EarthquakePrediction--AnInternational PhilipL. Woodworth,
DavidT. Pugh,JohnG. DeRonde,
Review DavidW. SimpsonandPaulG. Richards RichardG. Warrick,andJohnHannah(Eds.)
(Eds.)
12 Dynamicsof Earth'sDeepInteriorand Earth
5 Climate Processes and Climate
Rotation Jean-Louis
LeMou•l, D.E. Smylie,and
Sensitivity James
E. HansenandTaroTakahashi Thomas Herring(Eds.
)
(Eds.)
13 EnvironmentalEffectson SpacecraftPositioningand
6 EarthquakeSourceMechanics Shamira
Das,John TrajectoriesA. Vallance
Jones
(Ed.)
Boatwright,
andChristopher
H. Scholz
(Eds.) 14 Evolution of the Earth and Planets E. Takahashi,
Raymond
Jeanloz,
andDavidRubie(Eds.)
15 InteractionsBetweenGlobal Climate Subsystems:
The Legacyof Hann G. A. McBean
andM.
IUGG Volumes Hahtel (Eds.)
16 RelatingGeophysical
Structures
andProcesses:
The
1 Structureand Dynamicsof Earth'sDeep JeffreysVolume K. AkiandR. Dmowska(Eds.)
Interior D. E. Smylie
andRaymond Hide(Eds.)
2 HydrologicalRegimesand Their Subsurface
Thermal Effects Alan E. Beck,Grant Garven,
andLajosStegena
(Eds.
) Mineral PhysicsVolumes
3 Origin and Evolutionof SedimentaryBasins
! Point Defects in Minerals Robert N. Schock
and Their Energyand Mineral
ResourcesRaymond A. Price(Ed.) (Ed.)
4 Slow Deformation and Transmission of Stress 2 High PressureResearch
in MineralPhysics
in the Earth Steven C. Cohen and Petr Vanidek MurliH. Manghnani
andYasuhikoSyona (Eds.)
(Eds.) 3 High PressureResearch:
Application
to Earth
5 Deep Structureand PastKinematicsof andPlanetarySciencesYasuhikoSyono and
AccretedTerrarices JohnW. Hillhouse(Ed.) Murli H. Manghnani
 GeophysicalMonograph79

The Tornado:
Its Structure,Dynamics,
Prediction, and Hazards

C. Church
D. Burgess
C. Doswell
R. Davies-Jones

Editors

American
Geophysical
Publishedunder the aegisof the AGU BooksBoard.

Cover: SpearmanTexas tornado, May 31st, 1990.

Copyright1990Howard B. Bluestein,usedby permission.

Library of CongressCataloging-in-PublicationData
The Tornado: its structure,dynamics,prediction,and hazards/ C.
Church... [et al.], editors.
p. cm. -- (Geophysicalmonograph;79)
Includesbibliographicalreferences.
ISBN 0-87590-038-0
1. Tornadoes---Con••s. L Church,ChristopherR. ELSeries.
QC9,55.T67 1993
55L55'3---dc20 93-42302

ISSN: 0065-8448

ISBN 0-87590438-0

Thisbookisprinted
onacid-free
paper.
Copyright1993by the AmericanGeophysical
Union,2000FloridaAvenue,NW, Washington,
DC 20009, USA.

Figures,tables,and shortexcerptsmay be reprintedin scientificbooksand journalsif the

sourceis properlycited.

Authorization
tophotocopy
itemsforinternal
orpersonal
use,ortheinternal
orpersonal
useof
specificclients,is grantedby the AmericanGeophysical
Union for librariesand other users
registeredwith the CopyrightClearance
Center(CCC)Transactional
Reporting Service,
provided
thatthebasefeeof $1.00percopyplus$0.10perpageis paiddirectlyto CCC,21Congress Street,
Salem, MA 10970. 0065-844•93/•1. +. 10.
Thisconsent doesnot extendto otherkindsof copying,suchas copyingfor creatingnew
collective
worksor forresale.Thereproductionofmultiplecopies
andtheuseof fullarticles
or the
useof extracts,
includingfiguresand tables,for commercial
purposesrequirespermission from
AGU.

Printed in the United States of

 CONTENTS
Preface
Christopher Church, Donald Burgess,CharlesA. Doswell III, and Robert Davies-Jones xi

Plainfield Tornado of August 28, 1990

T. Theodore Fujita 1

Tornado Vortex Theory and Modeling

Tornado Vortex Theory
W. S. Lewellen 19

Numerical Simulationof AxisymmetricTornadogenesis

in ForcedConvection
Brian H. Fiedler 41

Numerical Simulation of Tornadolike Vortices in AsymmetricFlow

R. Jeffrey Trapp and Brian H. Fiedler 49

Discussion 55

Modeling and Theory of SupercellStorms

SupercellThunderstormModelingand Theory
Richard Rotunno 57

NumericalSimulationof Tornadogenesis
Within a SupercellThunderstorm
Louis J. Wicker and Robert B. Wilhelmson 75

Tornado
Spin-Up
Beneath
a Convective
Cell:Required
Basic
Structure
of theNear-Field
Boundary Layer Winds
Robert L. Walko 89

Environmental
Helicity
andtheMaintenance
andEvolution
ofLow-Level
Mesocyclones
HaroldE. Brooks,CharlesA. DoswellIII, andRobertDavies-Jones 97

Mesocyclogenesis
Froma Theoretical
Perspective
Robert Davies-Jonesand Harold Brooks 105

Discussion 115

Observations of Tornadic Thunderstorms

Observations
andSimulations
of Hurricane-Spawned
TornadicStorms
EugeneW. McCad, Jr. 119
TornadicThunderstorm
Characteristics
Determined
WithDopplerRadar
Edward A. Brandes 143

TornadoesandTornadicStorms:A Reviewof Conceptual

Models
CharlesA. DoswellIII andDonaldW. Burgess 161

Lightning
in Tornadic
Storms:
A Review
Donald R. MacGorman 173

Tornadogenesis
viaSquall
Line
andSupercell
Interaction:
TheNovember
15,1989,
Huntsville,
Alabama, Tornado
StevenJ. Goodman
andKevinR. Knupp !83
Discussion
CONTENTS

Tornado Detection and Warning

Tornado Detection and Warning by Radar
Donald W. Burgess,Ralph J. Donaldson,Jr., and Paul R. Desrochers 2O3

Single-DopplerRadar Studyof a Variety of TornadoTypes

Steven V. Vasiloff 223

Radar Signaturesand SevereWeather Forecasting

Paul Joe and Mike Leduc 233

The Use of Volumetric Radar Data to Identify SupercelIs:A Case Study of June 2, 1990
Ron W. Przybylinski,John T. Snow, ErnestM. Agee, and John T. Curran 241

DopplerRadar Identificationof Nonsevere

Thunderstorms
That Have the Potentialof Becoming
Tornadic

Rodger A. Brown 251

An Examinationof a Supercellin MississippiUsinga Tilt Sequence

David A. Imy and Kevin J. Pence 257

Satellite Observations of Tornadic Thunderstorms

James F. W. Purdom 265

Discussion 275

Physical Models and Analogs

Laboratory Models of Tornadoes
Christopher R. Church and John T. Snow 277

Laser Doppler Velocimeter Measurementsin Tornadolike Vortices

Donald E. Lund and John T. Snow 297

Vortex Formation From a Helical Inflow Tornado Vortex Simulator

James G. LaDue 307

Discussion 317

Tornado Observations
A Review of Tornado Observations
Howard B. Bluestein and JosephH. Golden 319

A Comparisonof Surface Observationsand Visual Tornado Characteristicsfor the June 15,

1988, Denver Tornado Outbreak
E. J. Szoke and R. Rotunno 353

On the Use of a Portable FM-CW Doppler Radar for Tornado Research

Howard B. Bluestein and WesleyP. Unruh 367

Discussion
 CONTENTS

Protection of Important or Critical Facilities

Design for Containment of Hazardous Materials
Robert C. Murray and James R. McDonald 379

State-of-the-Art and Current ResearchActivitiesin Extreme Winds Relating to Designand

Evaluation of Nuclear Power Plants
M. K. Ravindra 389

Wind/Tornado
DesignCriteriaDevelopment
to Achieve
RequiredProbabilistic
Performance
Goals
Dorothy S. Ng 399

Discussion 405

Climatology, Hazards, Risk Assessment

Advances
in TornadoClimatology,
Hazards,
andRiskAssessment
SinceTornado
Symposium
II
ThomasP. Grazulis,JosephT. Schaefer,and RobertF. Abbey,Jr. 409

Comparative
Description
of Tornadoes
in FranceandtheUnitedStates
Jean Dessens and John T. Snow 427

Tornadoes of China
Xu Zixiu, WangPengyun,andLin Xuefang 435

SeasonalTornadoClimatology for theSoutheastern

UnitedStates
Linda Pickett Garingerand KevinR. Knupp 445

OregonTornadoes:
More FactThanFiction
GeorgeR. Miller 453

The Stabilityof Climatological

TornadoData
Joseph
T. Schaefer,
Richard
L. Livingston,
Frederick
P. Ostby,
andPreston
W.
Leftwich 459

A 110-YearPerspective
of Significant
Tornadoes
Thomas P. Grazulis 467

Discussion 475

Damage Surveys
AerialSurvey
andPhotography
ofTornado
andMicroburst
Damage
T. T. Fujita andB. E. Smith 479
LessonsLearnedFrom AnalyzingTornadoDamage
TimothyP. Marshall 495

Survey
ofa Violent
Tornado
inFarSouthwestern
Texas:
TheBakersfield
Valley
Storm
ofJune
1, 1990
GaryR. Woodall
andGeorge
N. Mathews 501
An Observational
Studyof the MobaraTornado
H. Niino,
O.Suzuki,
T.Fujitani,
H.Nirasawa,
H. Ohno,
!. Takayabu,
N. Kinoshita,
T. Murota, andN. Yamaguchi 511

Discussion
CONTENTS

Damage Mitigation and OccupantSafety

Damage Mitigation and Occupant Safety
James R. McDonald 523

Tornado Fatalities in Ohio, 1950-1989

Thomas W. Schmidlin 529

Calculation of Wind SpeedsRequired to Damageor DestroyBuildings

Henry Liu 535

Risk Factorsfor Death or Injury in Tornadoes:An EpidemiologicApproach

Sue Anne Brenner and Eric K. Noji 543

Designfor OccupantProtectionin Schools

Harold W. Harris, Kishor C. Mehta, and Ja•nes R. McDonald 545

Discussion 555

Tornado Forecasting
Tornado Forecasting: A Review
Charles A. Doswell IH, Steven J. Weiss, and Robert H. Johns 557

Some Wind and Instability ParametersAssociatedWith Strong and Violent Tornadoes, 1, Wind
Shear and Helicity
Jonathan M. Davies and Robert H. Johns 573

Some Wind and Instability Parameters AssociatedWith Strong and Violent

Tornadoes, 2, Variations in the Combinationsof Wind and Instability Parameters
Robert H. Johns, Jonathan M. Davies, and Preston W. Le.[•wich 583

Diurnal Low-Level Wind Oscillation and Storm-Relative Helicity

Robert A. Maddox 591

Tornadoes: A Broadcaster's Perspective

Tom Konvicka 599

The "Short Fuse" Composite:An Operational AnalysisTechnique for Tornado Forecasting

Jim Johnson 605

The Plainfield, Illinois, Tornado of August28, 1990: The Evolution of Synopticand Mesoscale
Environments
William Korotky, Ron W. Przybylinski, and John A. Hart 611

Characteristics of East Central Florida Tornado Environments

Bartlett C. Hagemeyer and Gary K. Schmocker 625

Discussion 633

Open Discussion
 Preface

During the past two decades, remarkable advances have These analysessustainmy interestand inhabit my memory, for
been made in the understandingof the structureand dynam- they rarely fail to convey, in the words of Napier Shaw, a sense
of "the go of things." They couldeasilybe mistakenfor a mere
ics of tornadoesand tornadic storms.This knowledgehasled summaryof empiricalexperiencebut to the more attentive eye
to improvements in prediction capability, proceduresfor they offera conceptualsetting,a workinghypothesis,or simply
issue and disseminationof warnings, and the practice of invite the reader to considera plausiblespeculation.In short,
hazard mitigation. This progress can be attributed to the Ted's analysescomprisethat elusive combinationof critical
intellect and active mind, which distinguishesoriginality from
development of Doppler radars, wind profilers, lightning just ready accomplishment.
ground-strike location detectors, and automated surface
observing systems;to the application of multispectralsatel- We are deeplyindebtedto the manyanonymousreviewers
lite data; to improvements in numerical simulationof clouds for their careful evaluation of the manuscripts and their
and storms; to the deployment of mobile storm-intercept helpfulcomments.We alsothankKelly Lynn of theNational
teams with means to make quantitative observations; and to Severe Storms Laboratory and Rachel Hill of Miami Uni-
improved understandingof how structuresfail when sub- versity, Ohio, for their administrativeassistance.
jected to tornadoes. We also wish to thank the other members of the program
This volume provides a comprehensiveaccountof recent committeefor their participationin the organizationof the
tornado research, documentingthe advancesmade sincethe symposium:Howard Bluestein, Joseph Golden, Robert
symposiumon tornadoesheld at Texas TechnologicalUni- Johns, James MacDonald, Richard Rotunno, John Snow,
versity in Lubbock, Texas, in 1976 and is basedon work and Roger Wakimoto. Many thanks to William Beasley,
presentedat the Third Tornado Symposium,held in Nor- chair of the local organizing committee, for helping the
man, Oklahoma, April 2-5, 1991. The 53 papers are orga- programcommitteeconducta successful symposium andfor
nized into 11 topical sections,beginningwith the theory and facilitatingpublicationof thisbook.Many organizations and
modelingof tornadovorticesand tornadicstorms.The next institutions contributed funds and human resources that
four sections cover primarily observationalstudies and madethe symposium possibleand successful. Theseinclude
analysisof naturalphenomena,andare followedby a section the Center for Analysis and Prediction of Storms, the
on resultsof laboratory experiments.The practicalaspects Cooperative Institutefor MesoscaleMeteorological Studies,
of tornadoes as natural hazards, includingissuesin building and the School of Meteorology, all at the University of
damagesurveyand mitigationare Oklahoma;the NOAA STORM ProgramOffice,the NOAA
design,risk assessment,
addressedin four sections,and the final groupof papersis Office of Chief Scientist, and the NWS/UCAR Cooperative
devoted to the techniquesof tornado forecasting.Each Programfor OperationalMeteorologicalEducationand
sectionis accompanied by aneditedtranscript of the discus- Training, the Wind EngineeringCouncil, the National
sionperiodthatfollowedeachsession of thesymposium; we Weather Association,the American MeteorologicalSociety,
hopethat the editinghasnot diminished the spiritof the andthe AmericanGeophysicalUnion. Finally, we acknowl-
lively discussions. edge,with great appreciation,supportfrom the National
The Third TornadoSymposium providedan opportunity ScienceFoundation,throughgrantATM-9108873,provided
to celebratethe appointmentof T. TheodoreFujita, a jointly by the PhysicalMeteorology,Mesoscale Dynamic
pioneerin tornadoresearch, asMerriamDistinguished Pro- Meteorology, and NaturalandMan-Made Hazard Mitigation
fessorat the Universityof Chicago.As a measureof our programs. This supportmadepossible the participation
of
esteemfor Ted Fujita, we havechosenhispresentation on manystudents in the symposium andprovidedpartialsup-
the Plainfield tornado to introduce the volume. Ronald C. port for the publicationof this book.
Taylorintroduced the symposium's honored speaker in the
following words: ChristopherChurch,DonaldBurgess
Charles A. Doswell III, Robert Davies-Jones
Ted'spapers,of course,comprise
theusualcombination
of
photographs
andfiguresdepicting
hismeteorological
analyses.
 Plainfield Tornado of August 28, 1990
T. THEODORE FUJITA

The University of Chicago, Chicago, Illinois 60637

Shortly before 5 p.m. CDT on Tuesday, August28, 1990, wind streaks, consistingof standingcrops, are dark when
Chicagoradio stationsbeganannouncingthe occurrenceof viewed from the direction of shadows. On the other hand,
tornado deaths in the Plainfield area, some 60 km (40 miles) high-windstreaks,consistingof damagedcrops, are light
southwest of the University of Chicago. By 5:30 p.m. the coloredbecausethey scatter sunlight,especiallywhen pho-
casualtyfigureswere upgradedto 20 deathsand200 injuries, tographedfrom the favorablescatteringangle.In general,
givingan impressionthat it was a major tornadoevent. The low-wind streaks are seen in the downwind of isolated and
path lengthof the tornado,sketchyin natureat that time, clumpsof treesandthe high-windstreaks,extendingdown-
was givenas 13 km (8 miles).I calledfor a projectmeeting wind from open areas (Plate 1).
before ending the workday, reachingthe conclusionthat As has beenwell known, a slantedroof near the center of
Duane Stieglerof my staff (15 years of surveyexperience) a microburstdeflectsthe descendingairflow, inducinga jet
shouldgo directly to Chicago'sMidway Airport next morn- of highwindswhichextendsdownwindfromthe roof (Plate
ing to fly over the entire path. We assumedthat the survey 2). AlthoughI could not confirmthe numberof the mi-
would take 2 hours. crobursts,a largenumberof themwere involvedin produc-
After completinghis initial flight, Duane informed me ingthe largeareasof wind damagelocatedto the northwest
duringrefuelingthat the pathof the maintornadoextended of Oswegowhere the major tornadotoucheddown.
27 km (17 miles) from Oswego to Joliet. However, an Foundandphotographed in the downburstareasarefour
extensive area of storm damage,possiblyby a series of vortex marks, 1-7 km long and 10-30 m (30-100 ft) wide.
downbursts,extendedfrom Oswegotoward the northwest. Because these vortex marks were located beneath the path
He decidedto fly againon the secondday to completethe of a well-definedwall cloud, they were rated as F1 and F2
damagesurvey and photographyof both downburstand tornadoes(Plate3). The majortornado,afterits touchdown
tornadoareas.I am presenting in thistalk the resultsof my on the Fox River west of Oswego,failedto producecontin-
storminvestigation,makinguse of Duane'ssurveydata, uouswinddamage;instead,it produceda numberof strange
along with satelliteand radar photographs showingthe groundmarksin the cornfields.Theseare identifiedas
nature of the violent (F5) tornado which left behind 29 comma-shaped (Plate 4), swirl-shaped (Plate5), and eye-
deaths,300 injuries,and $160milliondamagein the south- shaped(Plate 6) marks.
west suburbs of Chicago. Thereafter,a numberof isolatedsuctionvorticesformed
whiletheparenttornadointensified to F3 intensitybefore
PHOTOGRAPHIC EVIDENCEOF THE STORM reaching
Wheat Plains. As the vortexdiameter increased,
a
series of suction vortices developed, producingnumerous
Aerialandgroundphotographs takenafterthe Plainfield vortexmarks(Plate7) whichareclearlyvisiblefromtheair.
tornado revealed that the wind effects of the storm system Apparently,
the tornadoreached its peakintensityof F5
wereverycomplicated. In studyingthenatureof thestorm, uponcrossing U.S. Highway30 (Plate8) northwest of
over600colorphotographs
covering
the600km2 areas
of Plainfield.Becausenear-ground
windsweresostrong,bean
DeKalb,Kane,Kendall,andWill counties
wereexamined
in (Plate9) andwheat(Plate10)cropswereliterally
flattened,
detail. and someevenwere pulledout of the ground.
Found in the extensive cornfields in DeKalb and Kane A 20-tontrailer (Plate 11) was blown off U.S. 30 and
countiesare numerousstreaksof high and low windsmade bouncedfive timesbefore reachingthe finalposition350 m
visibleby damaged andundamaged corncrops.The low- (1150ft) fromthehighway. Whiletraveling
overthecorn-
field, the core of the tornado,evidencedby the debris
TheTornado:Its Structure,Dynamics,Prediction,
andHazards.
GeophysicalMonograph79
deposition band (Plate8) shrank
to approximately
10m (30
Copyright1993by the AmericanGeophysicalUnion. ft) in diameter.
Foundnearthepathof thetornado
2 PLAINFIEI.D TORNADO

.... soUtI'H................. S06THWEST ......... WEST

70 ø 180ø 190ø 200 ø 2 I0 ø 220 ø 230 ø 240 ø 250 ø 260 ø 270 ø 280 ø
I I I I I I I I !

:5.0 p,m,

25 ø

20 ø

15 ø

10 ø

5:05 p,m.
•5o

Fig. !. Two compositevideo picturesshowingthe characteristicwall and tail cloudsat 3:04 and 3:05 p.m. CDT. !'he
cloud ba.scwas a.'.,low as 300 m (10½)0ft) above ground level, nutkingthe vis•al identificationo1'the large Plainfield
lt•rnm.lo very difficult, becauseit will not be seenas a typical tornad{•fi•nncl bcnealh a high cl{n•d base.

was a 25 mm (I inch) thick plywood board stuck vertically However, most frame houses were sitting on top of poorly
into the ground (Plate 121where 2 m (6 ft) tall corn crops had anchored foundations.
existed hcl'orc the tornado. The most tragic event occurred at Crest Hi!i, where two
long apartment buildings were damaged by the tornado
F5 PI•AINFIt:I !) I'()RNAD() (Plate 17). The top floors ot' the apartment complex were
sheared ot¾, leaving behind two wedge-shaped structural
I rated the Plainfield tornado }rs F5, based on the damage remnants, with their heights increasing to three stories at the
which was comparable to the worst I have ever seen. 'l'hc t•rthcst distance from the path of the tornado center. An-
damage in the cornfield southeast of U.S. 30 (Plate 8) was
entirely different from the damage adjacent to structures
affected by the F3 or F4 winds. Some corn crops were
stripped ot' leaves and ears and pushed practically down to
the ground. In the worst damage area, corn crops were
blown away entirely, leaving behind the remnants of small
roots connected to the underground root system.
People often use twisted trees as being the evidence of
/[,t•.•-•.
'Z•L-OF'"r'
...NORTH
tornadic winds. What we find in the wake of tornadoes,
however, are the results of windshift, rather than windshear.
Depending upon the position of the damage site relative to
the traveling tornado, a difluence pattern (Plate 13) and a
confluencepattern (Plate 14) are commonly seen in the wake
of tornadoes.Therefore the patterns shownin thesepictures
cannot be used as evidence of a tornadic airflow.
St. Mary Immaculate Church in Plainfield was a stream-
lined quonset structure (Plate 15) affected by F2 to F3 winds.
It lost some stained glass without receiving structural dam-
age visible from the air. Crystal Lawns, southeast of Inter-
state55, was damagedby F3 to F4 v,/inds.Some houseson Fig. 2. Tornadoes i, 2, and 3 in relation to the plan view of the
Byrum Boulevard were blown off foundations (Plate 16). wall cloud shown in Figure
 FUJITA 3

area. In fact, a funnel cloud aloft is seen in the video at the

extrapolatedlocationof tornado2 (Fig. 2).

NEED FOR ELEVATED SCANS AND DOPPLERRADAR

Marseilles(MMO) radar echoeswere depictedby levels

1-6 reflectivity contours (Plate 20) at seven different times,
correspondingto the rapid-scantimes of the GOES satellite.
The result revealed the existence of a hook echo which
beganformingat 1500CDT, shortlybeforethe videopicture
time. Thereafter,the appearanceof the echoturnedinto that
of a supercellwhich could spawn violent tornadoes.The
characteristicsupercellhook echo was most prominent at
3:11 p.m. CDT (2011 UT) when the weak and narrow
tornado 4 was on the ground. When the Plainfield tornado at
its F3-F5 intensitywas on the groundfrom 3:25 to 3:40 p.m.
CDT (2025-2040 UT), however, the identity of the hook
echo was not as clear, in part because of ground clutter.
This evidenceandjudgment might have been more clear if
higher-elevation scans and/or Doppler velocity fields were
available for operational use. We expect that the national
coverage of the NEXRAD radars and intensive training of
Fig. 3. A clo•e-up aerial photographof the eye-shapedbare
ground inside the core of the Plainfieldtornado.(For the location, storm forecasterswill improve ihe capability for revealing
see Plate 23). superell-tornado relationships.

Do TORNADOES GROW UPWARD OR DOWNWARD?

other aerial photographtaken on the followingday with Like the funnel cloud of a waterspout, a tornado funnel
identicalperspective(Plate 18)showstheremovalof cornby descendsfrom the cloud base, reaching ultimately to the
the rescue team in search of missingpersons. ground. On the other hand, the airflow inside a tornado at the
formative stage is predominantly upward. The formation of
MAPPING TIlE STORM DAMAGE
three isolated rotational winds, comma, swirl, and eye (Plate
21), from F2-F3 intensity presents the following questions:
The area of downbursts on the upwind side of the Plain- Did they descend from the rotationg winds aloft? Or, did
field tornado turned out to be much larger than we had they form on or near the ground first and stretch upward'?
estimated. It extended 35 km (22 miles) toward DeKalb, with I begin to think that the eye (Plate 6) formed near the
a maximum width of 19 km (12 miles). The onset time of the ground, because it has no sign of the translational character-
downburst winds was 1445 to 1450 CDT. At 1448 CDT, istics (Plate 3) generated by a fast-moving vortex which
NOAA 11 infrared temperature shows the existenceof two descendedto the ground. Are comma, swirl, and eye circu-
cloud tops or "Twin Peaks" over the onset area of the lations suction vortices? My answer is "yes," because 1 am
downburst winds (Plate 19). hesitant to call them three separate tornadoes. The orbiting
At 1504 CDT, leaving the downburst storm, Paul Sirvatka, suction vortices (Plate 7) probably formed on the ground and
of the College of DuPage, took video photographs,panning stretched upward into the parent tornado, becausethe aerial
his camera through the 160ø-270ø-350 ø azimuths. Composite photographs suggeststhat they were slow moving at forma-
pictures at 1504 and 1505 CDT, 49 s apart (accordingto the tion and gained orbital motion thereafter.
video frame counts), revealed the existence of a well-defined While weakening from F5 to F4, the tornado entered
wall and tail clouds (Fig. 1). These cloudsare characteristic Plainfield, passingdirectly over Plainfield High School (letter
features of the supercell thunderstorm or mesocyclone S in Plate 22) and St. Mary Immaculate School with its
clouds. It is estimated that the wall cloud was traveling church (letter C). After smashing the community of Lily
toward the southeast at 17 m s -• which is within the Cache, the tornado weakened to F2. Then it crossed Inter-
computation errorof the 20 m s-• movement of the hook state 55, injuring six persons in automobiles.
echo determinedby the Marseilles(MMO) radar imagery. The tornado intensified again while passingthrough Crys-
Aerial photographsrevealedthe existenceof four vortex tal Lawns, where the Grand Prairie Elementary School
marks along the path of the wall cloud. In view of the (letter S in Plate 22) was damaged by F3 winds. Thereafter,
commonpracticeof callinga vortex on the groundbeneatha the core diameter shrank while weakening to F2 intensity. In
wall cloud a tornado, these vortices were identified as Crest Hill, however, the small-core tornado intensified into
tornados 1-4, all of which occurred inside the downburst F3 while passingacrossthe Cresthill Lake apartments
4 PLAINFIELD TORNADO

2011 (2012.3) 2026 (2027.3)

2016 (2017.3) 2031 (2033.7)

2021 (2022.3) 2041 (2042.3)

'B A ,' , C
C

Fig. 4. A sequenceof GOES rapid-scanpicturesshowingfingerlikepatternsatop the anvil cloud of the Plainfield

cloud. Radar echoes corresponding to these satellite photos are shown in Plate
 FUJITA 5

Four personswere killed when automobileswere blown

Fujita Tornado Scale
off U.S. 30. A northbound
tractor-trailer
left two separate
,, Fou
IrSrn•11
I Plainfield impactmarks(Plate25)ontheshoulder, indicatingthatthe
'Tornadoes • Tornado tractor and trailer had been separatedbeforethey were
I blown off the highway. The tractor flew 100 m before
, I .. plowingintothedebrisdeposit.Apparently,thetractorleft
I •
i
I .q<'::
:: j the highway moments after the tornado core crossed the
highway.The trailer,after losingmostof its scrapmetal
cargo,bouncedin the cornfieldfive timesbeforerestingin
1920 250 40 50 2000 I0 20 30 40 50 2106•=- the field at positionE (Plate 23).
Anvil-top Finger Index GMT A passengercar, after being blown off the highway,
' I I I II I I I I I I traveledat very low altitude, as seenby the shearingand
, I ', ' clipping offof corncropsalongits720-m(2350-ft)flightpath.
5•Distinct, elevatedfinger ; • ! _
Apparently,the car looped aroundthe core of the tornado
4--Well-defined fin•er I '/•'"/':'':'"'::''"'""'"''':..,'•••....,'.-'•
beforecomingto rest at positionC (Plate26) in the right-
5--Growing fingers- I J .'"':".'"'.",':".".':".'"':".'"'.",'""
[ •i:?:.:.::.:.;:.::..:??:?•!.•.,-
, finger.--.,:'::.!?::,?::!'i?:;??:'::':!!'i side-uppositionwith four passenger
====================================================================== seatsremainingintact.
2.•Short and narrow Onepassenger wasreportedkilledafter beingthrownout of
J--Suspected • ger• •,:-,:!:]'!:i':L'):¾'?!:::i':?::?!:•'!i'!'i'i':::'•'i':,?,'!:.?!?:!'!i....
---jl•-fin ====================================================================
the car. It is remarkablethat a car couldfly sucha long
n - - -I _ distancein a violent tornadowithout either tumblingor
"l-N3[fFnge.r-visi?
e-j- I 'j I.... I'' '1'[' I' '- -.. bouncing.
Fig. 5. Variationof anvil-topfingersratedby index0-5. These
fingersare closely related to the F scale variationof the Plainfield CLOUD-TOP FEATURES BY SATELLITES
storms consisting of tornadoes and downbursts.
Both polar orbiter (NOAA 11) and geostationaryGOES
satelliteswere in operationon August 28. The former, with
A in Plate22), consisting
of twolongbuildingsseparated by 990orbitalinclination,scannedthe stormcloudonly onceat
a 10-m wide pathway. When the tornadocrossedthe 250-unit 2:48.8 p.m. CDT (1948.8 UT) during the northboundorbit
apartment complex, several persons were thrown into the from 851 km altitude over southeast Iowa. The latter ob-
cornfieldto the southeast(Plate 17). By the morningof tained5-min rapid-scandata from 1:00 p.m. to 4 p.m. CDT
August 30 the cornfieldwas cleared(Plate 18), recovering (1300-1600UT) from abovethe equatorat 97.7øWlongitude.
seven bodies, includinga 5-week-oldinfant. Two persons GOES infrared (IR) temperaturesat 30-minintervalswere
perished in cars parked by the apartment.Thereafter, the contouredand superimposedupon MMO radar echoesand
tornado weakened rapidly and disappearedafter crossing the location of downburstsand tornadoes(Plate 27). Mean-
Larkin Road in Joliet. while, NOAA 11 IR data with 1-kmresolutionwere analyzed
with IøC isothermsbelow the -50øC cloud-toptemperature
WHAT HAPPENED IN F5 WINDS.'?
(Plate 28). Because NOAA 11 isotherms are capable of
depictingindividualcold tops,they were superimposed upon
The only location where I estimated F5 winds, based on the isoecho contours of the MMO echo at the same time.
the corn damage, was to the southeastof U.S. 30 (Plate 23). Two significant cold tops, identified herein as the •'Twin
Prior to the highway crossing of the tornado, its core Peaks" (Plate 29), with their coldesttemperatures, -67.6øC
diameter, evidencedby the width of the debrisdeposit(Plate of the West Peak and -61.5øC of the East Peak, coincided
8), shrank from 70 m to 10 m. Meanwhile, the storm with the high-reflectivity cores of the MMO echo. These
intensifiedup to F5 intensity. Six automobileswere blown peaks were located above the onset area of the downburst
into the cornfield.After travelingover the fieldfor about400 winds just to the east of DeKalb, Illinois (Plate 19). The
m, the core diameter beganincreasingagain,reaching100m. high-temperaturering encircling the Twin Peaks is an inter-
A seriesof exposedground, free from corn crops, was left estingfeature, but its physicaland dynamicalmeaningsare
in the field during the increasingstageof the core diameter not known at this time.
(Plate 24). Meanwhile, the corn damageto the southwestof Fortunately, the 5-cm radar of United Airlines northwest
of O'Hare Airport made RHI scans of the Twin Peaks area,
the exposedground was rated as lessthan F2, suggestingthe
obtainingthe RHI cross section at the top of Plate 29. It is
existenceof an extremely largewind shear,characterizedby
suspectedthat the West Peak was in the early sinking stage
stronganticyclonicvorticity. Did it inducea seriesof anti-
cylconic suction vortices on the right-hand edge of thewhile the East Peak was in the downburst-inducingstage
tornado?Another interestingpiece of evidenceis that thewhen its strong echo was already on the ground (Plate 19). It
is premature to make a conclusive interpretation of such
large-core tornado induced a number of orbiting cyclonic
also findingsfrom this case study.
suctionvorticesshortlythereafter.Aerial photographs
showedan eye-shapedvortex mark consistingof exposed In this particular case, several distinct, elevated fingers
ground(Figure 3) at the locationof the eye in Plate 23. extended from above the tornado toward the northwest.
6 PLAINFIELD TORNADO

Plate 1. High- and low-wind streaks seen in the vast area of Plate 4. Comma-shaped damage. The tornado was moving left to
downbursts. right.

Plate 2. Corn crops damaged by a jet of high winds deflected by a Plate 5. A swirling pattern of corn damage. The tornado was
slanted roor. moving right to left.

Plate 3. Path of an F2 tornado located inside the area of Plate 6. Eye-shaped corn damage. The tornado was moving from
downbursts. top to bottom.
 FUJITA 7

Plate 7. Suction-vortex marks near Whisky Road, looking Plate 10. Spring wheat flattened inside the F4-F5 area.
northeast.

Plate 8. The FS damage to the southeast of U.S. Highway 30. Plate II. A 20-ton trailer blown off U.S. 30. It bounced five times.

Plate 12. A plywood board stuck vertically into the ground in the
Plate 9. A close-up view of bean crops in the F4-F5 area. F4 area.
8 PLAI FIELD TORNADO

Plate 13. A difluence pattern produced by the wind shift in the Plate 16. Frame houses with weak foundations blown off by
tornado. F3-F4 winds.

Plate 14. A confluence pattern produced by the wind shift in the Plate 17. The CresthiH Lake apartments on August 29 with dam-
tornado. aged cornfield.

Plate 15. S1. Mary Immaculate church, which lost the cross on the Plate 18. The apartment complex on August 30 with cleancd*up
steeple. cornfield.
 FUJITA 9

PEAKS Plainfield Tornado of 28 Aug. 1990

o T 1.0 2.0 fkm
I ; i : , I i i i i
o 5 10 15 20 mi

Il-38

DE KALB CO

Plano

rrkVill.
KENDALL CO
INO.5
Ploinfield Tornadoj
IL-126 I
Plainfield
WILL CO#
.
.,
leo I I ., "-
{]
LA SALLE I
I ~~~ ~
i ~
t- i
V
, i
...J I
i

Plate 19. Damage maps of the Plainfield storm. Downburst winds are shown by blue arrows and tornadoes by red
areas. The wall cloud depicted in video was located to the northwest of Sugar Grove. The path of the waH cloud center,
with 5-min time markers, was estimated by the shape of the MMO radar echoes.
10 PLAINFIELD TORNADO

S 2021 R 2022
Radar Echoes of Plainfield Cloud C 1223P~p
at Satellite Picture Times
42°00'
I
V
~
S -- Sotellile Piclure Start Time (GMT)
C --
R --
Cloud Scan Time (COT)
Radar Picture Time (GMT)
50' ler ~,

""
,

LEVEL~") 40' r .......

S 2001 C 3:03{ p.m. R2004

rE'OO'

t S 2026
1-1 C 3:27.3 p.m.
20' 88°00'
I

R 2027
40'

I
10' 42°00'

~~l I (i
'p~
\
42°00' 50'

~""
"" ,
\

50'
I
40'

~rI
~
40' ~ I"
69°00' 40' do~ 88°00' 89°00' 40' 2~' e8°00' 40'

S 2011 C 3:12.3 p.m. R 2012 S 2031 C 3:~3.7 p.m. I R 2034

I
~~S
42°00'

\
I -8~':1
42°00'

V&~
/_v ,( L (
~,'
so' I ,.-:',
\~\':h
N=--'
.-J ) L ~~
4°i

89°00' 40'
"\
20' 88000'
L ~~~
do, 40' 2b' 88'O~' 40'

S 2016 R 2016 R 2042

C3:?~. IS20~2"i
I .---"
41'50' ( If

Y
42'OJ ~' l-----J

~ ( 40'

'N\\:l \ ')
50'

•
40'
~"\
30'
"Y't-
89'00' 40' 20' f\ 88'00' 40'
20'
40' 20' 88"00' I 40

Plate 20. Reflectivity contours of the Plainfield storm with superimposed positions of the hook-echo centers and
damage areas. The hook was most pronounced a13:12 p.m., shortly after the video pictures were taken.
 FUJITA 11

."

."
MESOCYCLONE
3'20p.m.COT)' 22
CEN~~'R"'"
V
Wheotlond Plolns .
~.,

_,:vd' VoS'
'"
J ~ }!.A 6-
Suction Vortice 0
FI s..-,j
o~ Twin Vortices
,.

Plate 21. The first half of the damage area of the Plainfield tornado. Variation of the F scale was determined by aerial
photographs of cornfields, trees, and structures.

Plate 22. The second half of the damage area of the Plainfield tornado. The tornado at touchdown was 1.7 km to the
right of the mesocyclone (hook) center. As the F scale increased to F5, both tornado and mesocyclone axes became
close to each other.
12 PLAINFIELD TORNADO

Autom obiles in F-5 Tornad o Winds o

""',m
I !

o 100011

EYE
>
COf. of Tornado
A SER"IESOF EXPosEo·GROUNO- - ••
Cor' HO BOUIICE

F
'" ,. ""

Plate 23. The F5 area of the Plainfield tornado where the core diameter
shrank from 70 m to 10 m and increased again
to 70 m. When the diameter was increasing, a series of suction vortices
formed on the right-han d side of the core, pulling
QuI corn crops and exposing the bare ground.

Plate 24. An aerial photogra ph of a series of exposed ground. Relativel

y light damage on the southwe st side of the
core suggests the existenc e of extremel y large anticyclonic wind
shear at the corc boundar y.

-
 FUJITA 13

Estimated Path of a Tractor-Trailer

~ 10 20 3f' 40 5,0 00 70m
o 50 160 i
150
i
200 feel

:-."
:::."
: ,~"

>'~"~;a~~~ . . .

'"
Plate 25. Various ground marks on U.S. 30 and adjacent fields left behind by the Plainfield tornado at its F5 intensity.

- "'..-
Estimated Path of a Car

,,
, ... - ... ,
,, 1:
o
Ip
I i i II
50
2[) 3,0
iii
100
4,0

150
50 60
i
200 feel
rOm

I '
1
\ ..
lOOP Debris Deposit

Plate 26. The path of a car which circled once around the small core of the tornado. The car, resting at C with fOUf
passenger scats inside, traveled Jess than J m above the ground, probably on the cushion of laminar rising air.
 14 PLAINFIELD TORNADO

'i \
\~:~r"ii.-:;\
I \
, I"I I I I I
I I \
~. "',~
\I"~~
II I I I )
\ I
42 ~ I I \ I
\ \
,
Ir """ 42

.'
I I \
, , "" ~"
".\
T"
I 1
,..A': ~
I
I

II I I
I I I I I I I \ I I I
I: 30 p,m, COT
.\111111 I I I I 3: 00 p. m. COT 4I\-+--\--+-\--!~~~=e::r:.J~\-.
CLOUD-TOP TEMP
'\- I I I 41 CLOUD-TOP TEMP
, HMI
I 57-· .w,u TEWP -:510C
I I III -noe
13·' /IlIN TEMP -6;:1°e
88 8 53" MIN TEMP

8 8'
8r\1111 i
IIIIIII
.... ~ - IIIIII
U:;?';,.,.. :, "', ~ I I I I I I
42,":
~, ,~
':-::' :, ... ~
\/ \~
.. ' _',
IIIII'"X .
\ ~(:.~-~~}~\~. '\IIIIVI'2
'\\ I I I---\-" I
",
'1""'\ I I I
\ I I I I I III
I IIIIIIIII 89
2:00 p,m, COT ,1 IIIIII
...
CLOUD-TOP TEMP \ I I I IT' 3:30 p,m, COT
CLOUD-TOP TEMP l"t~~~~~!~~~~~~~'~
~
loll'
51·' N'N TEMP _nee
-57°(.
'1 J88 I I I I I I' J, 57-· "u.x TUIP ·:H·C
&3·· MIN TEMP -63°C

'1 8f I I I \ '1 8~ I I . r I \ I 1\1

\

I \ \
/:::o,.. _~ I II

11
.;~ I Ii
,L -Il\" ®' '"r "~ , " ----- ,''', ~ I
., ." -\-\--\--\--\-....t-,tIi":'
~>~~(
.,- . : 159,'

1'- . . . . ""..
-\

60
",
", "V,,"
'\

"" "_.
_.J
~
( ~~
~~_~)
" ~\
"
,_.

~
Ill:
IV
'I
--.. .'0""'''' ~ --' ~ I
\ I 'I
\ I I I I
2: 30 p,m, COT I I IIIIIIII 4:00 p.m, COT ,
CLOUD-TOP TEMP 'I I I I r \ I I I I \\41 CLOUO-TOP TEMP

"
n··
lIlAX TEMP -57
NIN TEMP -63'"C I I I ;8 I I I I I .J,li ~1-- MAX TEMP _~7oC

.. --------
8'
I \ ....

Plate 27. Growth of the Plainfield cloud depicted by 1°C isotherms of the cloud-top temperature. II should be noted
that the horseshoe·shaped upwind edge was not characterized by radar echoes of comparable horizontal dimensions.
 FUJITA 15

Plainfield Thunderstorm, NOAA-II IR 88°00'

2:48 p.m. COT 28 August 1990
89°00

42°30'·j--+--t---+"--?-=::~~±:::

41°30'+---I--~~

o 20 40 60 80 km
~I--"'''i .!"'-"'~i _.i"'_~1 ~i--!"'~i -"'-oll
o 10 W ~ ~ ~miles

Plate 28. Infrared temperature of the Plainfield thunderstorm obtained by contouring the NOAA I J data with 1°C
resolution. The boxed area near the west edge of the cloud is enlarged in Plate 29.
 FUJITA 17

the basisof my Lear Jet experiments in the 1970s(Fujita questionsand studiesfor the next generationof tornado
[1974a, b], to be published
in colorin 1992),I amassuming researchers.
that thesefingersconsistof the anvil materialpushedout-
ward from the highly convectiveregion of the cloud. The
Acknowledgments. The researchleadingto this presentation
time-dependent
analysisof the cloudtop features(Fig.4) in beforethe TornadoSymposiumIII Banquethas been sponsored
relation to the temporal F-scale variation resulted in a sincethe 1960sby fourU.S. governmentagencies: NationalScience
positive correlation between the F scale and a so-called Foundation,National Aeronauticsand SpaceAdministration,Na-
tionalOceanicandAtmospheric Administration,andOfficeof Naval
cloud-top finger index (Fig. 5). Research under their successivegrants.
It was an honor to havehad an opportunityto presentmy
researchon the Plainfieldtornadoat the banquetof Tornado REFERENCES
SymposiumIII. As in this researchcase,an investigationof
Fujita,T. T., Overshooting thunderheads observedfrom ATS and
a complicatedphenomenonoften leadsto additionalstudies, Lear Jet, SMRP Res. Pap. 117A, 29 pp., Univ. of Chicago.
makinguse of both new data and advancedtechniques.This Chicago, Ill., 1974a.
particulartornado was, no doubt, a very complicatedone. I Fujita,T. T., Overshooting topsof severethunderstorms revealed
by Lear Jet, satellite,and radar observations,
SMRP Res. Pap.
hope that this presentation will stimulate a number of 117B,48 pp., Univ. of Chicago,Chicago,Ill.,
 Tornado Vortex Theory
W. S. LEWELLEN

Departmentof PhysicsandAtmospheric
Science,Drexel University,Philadelphia,Pennsylvania19104

1. INTRODUCTION does. This common element allows the more benign dust
devils[Idso, 1974]and waterspouts[Golden,1971]to be
A number of reviews of the dynamics of the tornado investigatedassurrogates for someelements
of tornado
flow
[e.g.,Bilbroetal., 1977;Schwiesow,
vortex are available in the literature. The purpose of the 1981;Schwiesowetal.,
presentreview is to updatemy earlierreview [Lewellen, 1981].It alsomakesit difficultto maintaintornadostatistics
1976]. That paper, which will be referred to as L76, at- [Grazulisand Abbey,1983].This reviewis primarilycon-
temptedto provide a critical assessment of the existing cernedwith the well-developed,strongtornado.
theoretical models to see how well they describedthe wind The essential elements of the tornado vortex are illus-
and pressurefield in a tornado,to seewhatunderstandingtrated in the schematicin Figure 1, taken from Whipple
they providedas to the parameters whichgoverntheflow, [1982].The flow spiralsradiallyinwardinto a coreflow
whichremain whichis basicallya swirlingrisingplumebut mayincludea
andto clarify someof the essentialquestions
to be answeredby futureresearch.Althoughit waspossible downward jet alongthe axis. The radial flow is greatly
to piecetogethera relatively
consistent,
qualitative
modelof intensifiedin the surfacelayer. The wholeflow is drivenby
the flow in differentregionsof the tornadoin 1976,it wasnot the thunderstorminto which the vortex feeds through a
possible model.HereI will concen- rotating
to providea definitive wallcloud.Thetornadovortexallowsa significant
trate on the considerable
progresswhichhasbeenmadein fractionof the potentialenergyof the parentstormto be
this directionin the researchpublishedin the last 15years. convertedintowindkineticenergyvery closeto the surface
Again,I will restrictattentionto theimmediate vicinityof whereit can causegreatdamage.Althoughthere is some
the tornado,within a radiusof approximately 1 km. A evidence that the tornado vortex may intensify the storm,
companion reviewon thunderstorm modeling by Rotunno themuchmoreimportant effectof thetornado vortexisthis
[thisvolume]will updatethereviewbyLilly [1976]onthe loweringof maximumwind speedsfromhighin the storm
larger-scale
sources ofvorticityandenergy forthetornado.almostto the surface.Local concentrationof the vorticity
Since a number of relatively recent reviewshave been canleadto windspeeds exceeding 100m/swithina cylinder
published [Bengtssonand Lighthill,1983;Davies-Jones,of approximately
100m radius.As will be discussed in
1986;Deissler,1977;Lugt, 1983;Maxworthy, 1986;Ro- section6, there is even the opportunityfor the unique
tunno,1986;SmithandLeslie,1978;Snow,1982,1984, interactionbetweenthe vortex and the surfaceto permit
1987],I willfeelfreeto incorporate
myownbiases in this maximumlow-levelwind speedsto exceedthat achievable
present review. bya straight
exchange of stormpotential
energy forkinetic
The mostessential
elementof tornadolike,
vortexflowis energy.
theconvergence of ambient, Thisispresent The sharpgradients
axialvorticity. in particulateconcentrations
occur-
in natural
flowsranging
fromfamiliardrainvortices[Lund- ringin tornadoes
makethemveryphotographic. Thesche-
gren,1985] to hurricanes
[Anthes,1982]. Thereareclear maticmaybe compared withthephotograph, Plate1, of a
demarcations betweenthe tornadoand the muchlarger westernKansastornadoonAugust28, 1979[Whipple,1982].
hurricane.The hurricaneis a complete,largestormsystem, It showsa large cloudof dust and debrisengulfing the
while the tornadois an appendage to a thunderstorm.bottomof thefunnel.Thepenetration of theclearregionto
However, the sparseness of cleardistinguishing featuressmallerradiusat the surfaceprovidesevidenceof the strong
between the coredynamics of weaktornadoes andother radial inflowin the surfacelayer. The top of the dense
weaktorna- surface
naturalvorticesmakesit difficultto distinguish cloudissharper thanmightbeimplied bytheupward
verticalvelocitiesinferredfor theseregionsfrom Figure1.
TheTornado: Its Structure,
Dynamics, Prediction,
andHazards. Thisis influenced bothby particlesizeof the debrisandby
GeophysicalMonograph79 unsteady interceptionof appropriate
debrisalongthe
Copyright1993 bytheAmerican Geophysical Union.
20 FORNADO VORTEX THEORY

l:ig. I. Schematic of tornado tlow taken from Whipph, [1982].

do's path. Above this dense cloud it is possible to .seea sharp It is convenient to divide the flow into the sltme four
water cloud funnel boundary surrounded by a faint dust regions ils in L76: region !, the core fiow; region I!, the
sheath. !'he funnel cloud boundary does not denote u st,trace boundary layer flow; region !11, the central corner
streamline, but mt•rc nearly corresponds to a constant pres- flow; and region IV, the top layer. Three of these four
sure surface. It can occur in either the upfiow or the regions are seen in Figure I and P!ale I. The top layer is
downllow region of the schcm•ttic in Figure 1. The percep- ot'thc picture in both cases. I will concentrate on the central
tion of a clear annulus between the cloud funnel and the dust corner flow region, because this is the region where the
sheath suggeststhat the funnel is occurring within the inner maximum velocities {recur and where the dynamical mech-
downflow at this particular time. The sharp boundaries anisms are mosl distinctive.
indicate that the level of turbulence in this core 11ow is low.
l'hc bend in the funnel cloud at height is not unu.sual for
2. Co,•i Ft,ow, RFGI()N I
tornadoes, particularly in the decaying phase. This is evi-
dence that even when the thunderstorm flow pattern forces a In natural vortices the horizontal convergence tends to be
nonvertical core, its interaction with the surface tends to induced by positively buoyant plumes or updrafts. Conser-
rioroea perpendicular intersection. vation of angular momentum with respect to the center of
Perhaps the greatest advances in the last 15 years have these updrafts then amplifies any ambient vertical vorticity.
been in the area of numerical simulations, which were just In terms of the dynamics of vorticity [e.g., Morton, 1969,
beginning to make significant contributions in the mid 1970s. 1984], the existing vortex tubes are stretched, so that the
Thanks in great measure to the roughly 2 orders of magni- cross section becomes smaller and the vorticity amplified to
tude increase in hardware computational capabilities during keep the product of the vorticity and the cross-sectionalarea
this time period, fully three-dimensional, unsteady simula- constant. At the barely discernable level this leads to the
tions are now easier than steady state, axisymmetric calcu- small whirls in the steam fog rising fi-om a warm water pond
lations were at the beginning of the period. 1 will present on a cold morning. Under large air/water temperature differ-
some new "large-eddy simulation" results for turbulence in ences, as may occur after a winter cold front passes over
the tornado vortex made during the preparation of this open water [e.g., Lyons and Pease, 1972], steam devils,
review, which complement the axisymmetric, turbulent which are quite similar to dust devils, can occur. Hess et al.
transport results of Lea'ellen and Sheng [1980]. I think the [1988] have looked at the conditions necessary for such
technology is now at hand, where it should be possibleto boundary layer vortices to occur. They report that the
take boundary conditionsfi-oma numerical storm simulation environment is favorable for their development when -h/L
and obtain valid estimates of the probability of different > 100, where h is the height of the mixed layer and L is the
levels of extreme winds occurring in different storms. Monin-Obukhov length, consistent with the suggestions
 LEWELLEN 21

Deardorff[1972]. Schmidtand Schumann[1989]haveuseda

large-eddysimulationof the convectiveboundarylayer to RAPID
GROWTH
look at theseconditionswherecoherentplumesareinitiated.
Entrainmentinto such a developingplume providesthe
..

convergencewhich can amplify any ambientverticalvortic-

ity. The tiltingand stretchingof horizontalvorticitycanalso
be a major sourceof verticalvorticity for theseboundary
layer vortices [Maxworthy, 1973]. ,,• CU
VORTEX
CORE '"'"
Once a buoyant plume vortex is formed, the strong
stability of the flow allows the flow to persist and even
amplify.This'featurecanallowthe swirlingplumeto be quite
distinctive.As a consequence of their high Reynoldsnum-
co ), vo. t.
ber, mostatmosphericflowsare turbulent,with a cascading
of energyfrom large scalesto smallscales.Althoughtorna- " 3-6 KM •

dolike vorticesoccurin a regionof relativelystrong,three-

Fig. 2. Schematic illustration (not to scale) of cumulus-outflow
dimensional turbulence, the rotational forces occurring interactionsin relation to vorticity generationfor a waterspout
within the vortex coreare ableto impedethe normalcascade according to Simpson et al. [1986].
of turbulent energy and allow the core vortex to dominate.
When the vortex is driven by its core buoyancy, this
dampingof core turbulencealsopreventsthe corebuoyancy [1989] for nonsupercell tornadoes. The same type of
from diffusingas muchas it otherwisewould. This provides close proximity between the cold downdraft and the warm
a natural feedback; the greater the height to which the core updraft appears to be involved in a thunderstorminitiating
buoyancy persists, the stronger the radial pressuredrop the strongtornado [Davies-Jones, 1986], but I will leave that
availableto drive the vortex. It also helpsexplain why fires to Rotunno [this volume] to explain.
may be intensified by the introduction of swirl [Church et al.,
1980]. In the normal inertial range of three-dimensional 2.1. Dimensionless Parameters
cascading turbulence there is an increase of enstrophy
(squared vorticity) and decreasein energy with decreasing Most researchersagree that the primary tornado vortex is
scale. In a tornadolike vortex there appearsto be sufficient essentially axisymmetric within a radius of 1 kin. Although
stability to allow the normal cascade to terminate and the asymmetric effects are important, this axisymmetric nature
coherent vortex to intensify. I will return to this stability of the primary vortex reduces the geometricalvariation of
discussion in section 2.3. the core flow to a single parameter, the geometrical ratio of
A major problem in modelingthe tornado is in determining the domain radius to the domain height, i.e., an aspectratio,
how the thunderstormgeneratesthe high levels of vertical AR = to/h, with r 0 the radius of the domain and h the
vorticity necessaryto supply the tornado vortex. However, height. The aspect ratio has a strong influence on the
this question is predominantly outside the domain of the interaction between the core flow and the top and bottom
current review, since it involves the storm dynamics on a boundary conditions.
scale larger than 1 km. In order to hint at how this generation The core flow through a cylindrical domain with fixed
and concentration may proceed, Figure 2 shows a concep- aspect ratio is of course largely determined by the inflow/
tional schematicof possibleflow in the vicinity of a water- outflow boundary conditions. When this open tornado do-
spout from Simpson et al. [1986]. The convergence in the main is consideredto be drawn from a larger-scale, ambient
immediate vicinity of the waterspout is suppliedby an active flow, one of the most important parameters influencing the
inflow is some measure of the ratio of the ambient vertical
cumulus cloud. The problem (and probably the reason
waterspouts do not occur more frequently) comes in supply- vorticity to the ambient horizontal convergence. Such a
ing sufficientvertical vorticity, that the stretching associated swirl parameter, S, is usually taken as
with the convergence
canincreaseit to order1 s-• within S = (ooro)/(2ahh)= Vo/Wo (1)
the 10 min the cumulus is most active. On the basis of GATE
data and extensive three-dimensional numerical cumulus with •o,the ambient vertical vorticity, and a j,, the ambient
simulations (their grid resolution of 0.6 km is unable to horizontal convergence.This may also be written as the ratio
resolve the waterspout, but they were looking to obtain of the tangentialvelocity,v0, of the inflowat r0, dividedby
sufficiently high vorticity levels on that grid scale), they the averageverticalvelocity, w0, at the top of the domain.
conclude that conditions which bring together cool descend- The value of the swirl ratio has a strong effect on the
ing and warm ascendingair side by side are more likely interaction between the vertical and the horizontal compo-
to increase both low-level convergence and low-level verti- nentsof vorticity. In the work of Lewellen [1962]I showed
cal vorticity sufficientlyto achieve waterspout intensity in that when S is large, any axisymmetriccore flow musthave
the lifetime of the cumulus updraft aloft. A quite similar its axial componentof vorticity essentiallyindependent
of
type of conceptualmodel is given by Wakimotoand Wilson the axial coordinate, z, over large stretchesof z.
22 TORNADO VORTEX THEORY

substantialz variations in the axial component of vorticity 2.2. Simple Models

must occur in narrow layers in the limit of large S.
In the limiting case of incompressible,irrotational flow,
The influenceof physical scale size comes through some
both the circulation and the total pressure are conserved as
Reynolds number, such as
the vortex flow converges radially inward. This implies a
Re = voro/v (2) minimumradius,re, insidewhich the flow cannotpenetrate.
This ideal minimum radius, where the maximum velocity
Here v is the kinematicviscosityof the fluid. This is certainly occurs, is given by
an importantparameter,for laboratoryand numericalsimu-
lations.However, onlythe fact that it is a largevaluemay be r,.= F(2Ap/p) (-•/2) (4)
importantlbr actualtornadoes.With a few exceptions[e.g., where F is the productof the tangentialvelocity times the
Chi and ('ostopotous,1975;Lewellenand Sheng,1979,1980; local radius (circulation/2rr), p is the density, and Ap is the
Lewellen and Teske, 1977; Lewellen et al., 1979], numerical
pressuredrop drivingthe vortex. This providesa simple,
simulationsare at relatively low Reynolds numbers since approximaterelationshipbetween three of the important
sometype of constanteddy viscosityis usedto adjustfor variablesdefiningthe vortex. For flows with roughlyequal
atmospheric turbulence.i believethe dynamicalvariations pressuredrop, the coreradiusis proportionalto circulation,
in the effectiveeddyviscosityin differentpartsof the flow while for flows with similar circulation, the stronger the
are important,but thisis not universallyaccepted.More of pressuredrop, the smallerthe core radius. The pressure
the influence ot' Re will be considered when the corner flow
drop, Ap, is dividedbetweendrivingthe swirl and driving
is discussed.
the throughflow,so (4) assuresthat the swirl parameter,S,
A t'ourthparameterinvolvessomemeasureof the relative will have a stronginfluenceon the flow. In fact, when the
importance ot'buoyancyforceswithintheflow. In laboratory pressure dropdrivingthe vortexflowis heldfixed,(4) canbe
flowsthis is usuallygiven in terms of a Froude number: manipulated to showthat the swirl ratio governsthe mini-
Ft' .•.?:•
(Ap/2#Aph)
•/2 (3) mum radius to which the maximum tangential velocity can
penetrate.However,I expectturbulentdynamicsto playan
where&p is the pressuredrop drivingthe flow, Ap is the importantrole in determining the detailedvelocitydistribu-
densitychangewithinthe tlow, and f•tis the gravitational tions, particularlycloseto the surface.
constant. It is not clear how important this parameter is to Within the core flow the tangential velocity sufficiently
the tornadovortex. Certainly, buoyancydue to latent heat dominatesthe radial velocity so that there is an essentially
releaseby condensation,or absorptionby evaporation, cyclostrophic balancebetweenthe radialpressuregradient
playsa dominantrolein thethunderstorm whichprecipitates and the centrifugal force. Radial integrationof the radial
the tornado. However, much of this driving energy is trans- pressuregradient for anyassumed tangentialvelocitydistri-
ferred to the smallertornado scale by pressureforces. The butionthen providesa relationshipsimilarto (4), with a
Froude numberdependsupon the relative ratio ot' these numericalfactordependingon the assumeddistribution.As
pressureforcesto the buoyancywithin the tornadocore. discussed in L76, a favorite simplemodelof velocitiesin the
The Frotide numberwas varied in the laboratory simulations core flow is the Rankine vortex [Rankine, 1882] which sets
of a dustdevil by Mullen and Maxworthy[1977]and in the the radial and vertical velocities equal to zero and contains
numerical simulationsof Lewellen and Sheng [ 1980]. It is my an innerregionof solid-bodyrotation,v --• r, with an outer
guessthatFr is sufficiently largein thelow-levelcornerflow potential flowregion,v • r -1. Thiscanbereadily
improved
of the tornado that it is of relatively little importance here, uponby solving
for the radialbalancebetweenadvection
but that it becomesquite importantat higherlevels.Its most and diffusionin the presenceof sink flow with a constant
importantlow-levelrole is likely to be associated with viscosity to obtainthe Burgers-Rott
model[Burgers,1948;
stabilizingthe core flow. Rott, 1958]which yieldsa smoothtransition
betweensolid-
A few researchers have considered the influence of com- bodyrotationandpotential flow in the annularregionof
pressibilitythrough theMachnumber[e.g.,Eagleman et al., maximumvelocity.This providesfor somesimplerelation-
1975].However, sinceI will arguein the later sectionsthat shipsbetweenmaximum velocity,coreradius,andtotal
the maximum low-level velocities are limited to less than pressure
dropsimilarto (4) whichare surprisingly
good,
110-130m/s, the Mach numbersquaredshouldbe lessthan considering
that it ignoresturbulence
and all of the tight
0.1, and thus suchcompressibilityeffectsshouldbe rela- coupling
between
thetopandthebottomboundary
condi-
tively small.Also, we mightadd a numberof additional tionswhichmustprevail.The coreradius,definedas the
parameters in attemptsto measurethe influenceof such radius where maximum velocity occurs, implied by the
thingsastheshapeof theradialinflow,theshapeof theaxial Burgers-Rottmodelis approximately
30%largerthanthat
outflowat the top of the domain,andthe buoyancyreleased givenby(4)whenF istakenasthevalueatlargeradius and
due to condensationof water vapor within the domain,but •p is the totalpressure
dropobtainedby integratingthe
suchparameters arenoteasyto formulatein a simpleway. cyclostrophic balance
between r = 0 andinfinity.Thisis
Further, I am willing to speculatethat they are not as truein spiteof thefactthattheBurgers-Rott
modeldeter-
importantasthe fourparameters
just defined. minesthe coreradiusby the radialbalancebetween
 LEWELLEN 23

tion and diffusionof tangentialmomentum,with no direct vorticesappearingwith their axesapproximatelyparallel to

influenceof S onthe velocityfield.However,a strongeffect the rotation axis.
of S is implied when the radial momentumequationis This transitionin the turbulent cascademay be partially
integratedto obtainthe pressuredrop. explainedin terms of helicity, the vector inner productof
It might be argued by some that these simplemodels velocity and vorticity. Lilly [1986a, b] has proposed that
should not even be considered as tornado models because helicitymay be an importantdescriptivevariablein rotating
they leave out so much. However, I think that any model storms. Fundamental research on turbulence [e.g., Andre
whichprovidesusefulrelationships betweenimportantvari- and Lesieur, 1977; Polifke and Shtilman, 1989] has shown
ables within the flow may justifiably be referred to as a that high levels of helicity tend to retard the cascadingof
tornadomodel.If only the completemodeldeservesthe tag, turbulentenergyto smaller scalesand thus reduce dissipa-
then we would still be without any models of a tornado. As tion. The stretchingand tilting termsin the vorticity equation
discussedin more detail in L76, the radial balance between are usually consideredto be most responsiblefor the cas-
advection and diffusion of tangential momentum has been cadingof energyinto the inertial subrange.However, purely
solvedfor a numberof differentassumptionsaboutthe radial helicalflow, when the vorticity vector is a constanttimes the
velocity [Bellamy-Knights,1970, 1971;Dergarabedianand velocityvector(Beltramiflow), allowsthesetwo termsin the
Fendell, 1967; Einstein and Li, 1955;Lewellen, 1962;March- vorticity equation to exactly cancel with the advection
enko, 1961; Sullivan, 1959]. A model which has a stronger terms. Davies-Joneset aI. [1984] has suggestedthat helicity
couplingbetween the swirling velocity and the axial velocity can be used as a predictor of severe storms, with at least
is that by Long [1956, 1958, 1960, 1961] which assumes somepreliminarysuccess[Davies-Jones andBurgess,1990].
conical similarity. Burggraf and Foster [1977] show that for I am sure more will be written about this storm scale helicity
some relatively arbitrary initial velocity conditionsat z = 0, in the rest of this volume. It may bejust as important on the
laminar flows tend to approach one of Long's similarity tornado vortex scale. Certainly, there appears to be strong
solutions asymptotically, as long as the ratio of the axial dampingof turbulent eddies by local rotation in the core
momentum of the jet to the angular momentum is above a from the smoothappearanceof the central funnel in many
critical value (---3.75). For smaller values they surmised that photographs.
some type of vortex breakdown occurred. A number of The rotational damping of turbulence may be even stron-
similarity solutionsalso include buoyancy to generatethe ger in the presenceof a positiveradial gradientin density
updraft[Franz, 1969;Fulks, 1962;Gutman, 1957;Kuo, 1966, which may exist in any buoyantlygeneratedvortex plume.
1967], but, when a two-cell solutionoccursin one of these Thus even when the Froude number of the tornado flow is
buoyant models, the downdraftat the centeris forced by moderatelylargeso that buoyancyeffectsmay be neglected
colder temperatures,rather than beingforcedby the strong in the mean flow dynamics, it may still be important in
low-level swirl as occurs in the tornado. dampingthe turbulencein the tornadocore.The centrifugal
Some new similarity solutionshave been presentedfor accelerationof a 70-m/stangentialvelocity at a radiusof 50
swirlingaxisymmetricflows [Paull and Pillow, 1985;Wang, m is approximately10 timesEarth'sgravitationalaccelera-
1991; Wu, 1985; Yih et al., 1982], but in general these tion. Thus strong damping of radial motions should be
one-dimensionalflows require more artificial restrictions expectedwhenthesewouldcarrywith thema negativeradial
thanappearjustifiednowthathigher-dimensional flowsmay buoyantflux. When the Richardsonnumber criterion of
be readily integratednumerically.The most interesting, stratifiedturbulenceis carried over to axisymmetricswirling
recent core flow solutions are finite difference solutions,but flow, it says that turbulenceis dampedwhen the radial
since these also involve interaction with the top and bottom gradients(denotedby primes) of potentialtemperature,0,
regions,I will deferdiscussing theseuntilsection 5, which axial velocity, w, and circulationsatisfythe relation
deals with the corner flow where much of the strong inter-
actions occur.
0'/0o< 2F'/F - w'2r3/(4F
2) (5)
This oversimplifies
the stabilityproblem[Leibovich,1984],
but it does illustrate that radial buoyancy gradientsdo not
2.3. Stability Considerations haveto be very large to be as importantas incompressible
Althoughthe tornadooccursin a regionof relatively rotational considerations.
strong,three-dimensional
turbulence, forces More details on the stability of tornadolike vortices may
the rotational
occurringwithinthe tornadocoreareableto impede the be found in the literature [Foster and Smith, 1989; Gall,
1983, 1985;Gall and Staley, 1981;Leibovichand Stewart-
normal cascade of turbulent energy and allow the core
vortex to dominate.This phenomenonis apparentin the son, 1983;Rotunno, 1978;Snow, 1978;Staley, 1985;Staley
resultsof therotatingtankexperiments
of Hopfingeret al. and Gall, 1979, 1984; Steffens, 1988; Stewartson,1982;
[1982],whichdemonstrated of the normal Stewartson and Leibovich, 1987; Walko and Gall, 1984].
the blockage
cascade of turbulentenergyin favorof the appearance of When stable conditions exist, inertial waves may travel
coherentvortices. Theyreported a dramatic in the alongthecoreandprovideanimportantpartof anyunsteady
transition
turbulent flow field when the local Rossbynumberwas dynamics [Maxworthyet al., 1985;SnowandLund, 1989].
decreasedbelow about0.2, with a collectionof coherent It is importantto notethat satisfying a stability
24 TORNADO VORTEX THEORY

such as (5) does not assure a laminar vortex, but rather This representsthe balance between the vertical divergence
establishesconditionsfor which we can expect any existing of the turbulent flux of radial momentum and the difference
turbulence to be damped. Fiediet [1989] has arguedthat the between the local centrifugal force, which varies with z, and
central jet eruption from the turbulent boundary layer in a the pressure gradient which is imposed by the centrifugal
full-scale tornado corner flow, to be discussed in section 5, force above the boundary layer. Close to the surface we also
may be laminar on the basis of a local stability criterion. I expect a simple mixing-length, turbulent model to be reason-
think this is unlikely, because of the time scalesrequired to ably valid for the shear stress, so that
damp the turbulent eddies existing in the radial jet which is
converging to form the vertical jet. Although, the details of a u'w' = -c•zOu/Oz (7)
laminar corner flow are not appropriate for flow in a full-
scale tornado, the core flow, which the corner flow trans- If u'w' were constant, (7) would yield the logarithmic result
t3rms into, may have its turbulence sufficiently damped for as it does for the tangential velocity:
the flow to appear laminar, as indicated in the top part of the
funnel in the tornado in Plate 1.
v = c2 In (Z/Zo) (8)

Figure 3 gives the results of integrating (6) and (7) with the
3. SURFACEB½)UNDARYL^¾v:•t, Rv:c•oN Ii
aid of (8) and compares this result with the laboratory
It has long been recognized that the retardation of a observations of Savino and Keshock [1965]. In this compar-
rotating flow at a surfaceperpendicularto the axis of rotation ison, three data points are used to determine the effective
will induce a radial inflow within the boundary layer [Ekman, surface roughness, z0, and the two components of the
19()51. A number of different solutions for the rotating surface shear stress, which are related to the two constants
boundary layer were discussed in L76 [Barelion, 1967; c• and c2. This is the same information required to deter-
Bek'her et al., 1972; Bellamy-Knights, 1974; B6dewadt, mine the usual logarithmic layer. The pressuregradient is
1940;Burggra,/'et a!., 1971; Carrier, 1971; Chaussee, 1972; given by the free stream conditions above the boundary
Chi and Glowacki, 1974; Chi and Jih, 1974; Chi et al., 1969; layer.
Goldshtilc, 1960; Hsu and Te,s]•lmariam, 1975; Jischke and The comparisonin Figure 3 showsthat the addition of the
Parang, 1975; Kidd and Farris, 1968; Kuo, 1971; Rao and pressuregradient influencein the surface layer permits a
Raymond, 1975;Rott, 1962; Rott and Lewellen, 1965, 1966; goodrepresentationof the radial velocity for a much greater
Schwiderski, 1968; Smith and Smith, 1965]. The boundary rangeof z, distancefrom the surface,than is the casefor the
layer beneatha vortex also has continuedto be a popular leadinglogarithmicterm. Naturally, this turbulentvariation
subjectsince 1976[Baker, 1981;Carrier et al., 1988;Chi, producesvelocitieswith muchsharpergradientscloseto the
1977;tlsu and Te,sJ•,nariam,1976;Kuo, 1982;Leslie, 1977; surfacethan are producedin a laminar boundarylayer (the
Monji and Yunkuan,1989;Phillips, 1984, 1985;Phillipsand inherentresultproducedby a constanteddyviscositymod-
Khoo, 1987; Prahlad, 1976; Rostek and Snow, 1985; Ro- el). The sharppeak in the radial velocity very close to the
tunno, 1980;Sychev, 1989; Wilsonand Rotunno, !986]. A surfacein Figure3 indicatesthat if a constanteddy solution
majority of these boundarylayer analyseshave been for is to be used, it is probably better to combine a free-slip
laminar flow. These give a valid qualitative view of the boundaryconditionon the radial velocity, with the no-slip
boundaryinteractionwith vortexflow, but the variationin condition only applied to the tangentialvelocity.
turbulenceclose to the surfaceis important in determining Local conditionsat any radiusprovide an estimateof the
the detailed structure of the flow. I will not attempt an radial volume flow induced into the linear Ekman layer, but
individualcritiqueof all of theseeffortshere, but insteadwill
this is not generallyadequatefor the full nonlinearrotating
onlylookat whatI considersomekey featuresof flowin this boundarylayer. This requiresinformationon the historyof
region. the boundarylayer.A simpleestimateof thiswasreportedin
First, let us considerhow the surfacelayer is affectedby L76 as given by Rott [1962]:
rotation. The presence of a boundary imposesa strong
constrainton the dynamicsof the turbulence.In the absence Qbl= -2.5cfF(ro- r) (9)
of any bodyforceson the flowthisleadsto a logarithmiclaw
of the wall regionin whichthe flux of momentumis constant wherethenegativesignindicatesthattheradialvolumeflow,
while the velocitydecreaseslogarithmicallyas the surfaceis Qbt in the boundarylayer is radiallyinward, cf is the
turbulent skin friction coefficient, F is the circulation in the
approached. This logarithmicvariationis validfor the tan-
gentialvelocityin the boundarylayer undera tornadolike.outer flow, and r 0 is the outer radiusfrom which the
vortex but is only valid for the radial velocity extremely boundarylayer develops.Equation(9) providesa direct
close to the surface. This may be demonstrated by the estimate of the flow which is diverted into the boundary
followingsimpleanalysisfrom LeweIlen[1977]. layerat largeradii and mustbe ejectedverticallyinto the
corner flow to be discussedin section5. It showsthat there
In the surfacelayerthe radialmomentumequationmaybe
approximated as mustbe negativeverticalvelocityinto the boundarylayer
outsidethe radiuswhere the circulationremainsessentially
pOu'w'/Oz
= 9v2/r- Op/Or (6) constant. When the circulation is not constant above
 LEWELLEN 25

1.8 1979;Dergarabedian and Fendell, 1973;LilO,, 1976;Mitsuita

et al., 1987; Walko, 1988; Wylie and Anderson, 1983].
1.6
It is my opinion that most moderate tornadoes are not
associatedwith an eye extendingthrough the thunderstorm,
V/Umoa even when there is a central downdraft at low levels. Rather,
i.4 they are capped by a buoyant plume, a processthat still
remains to be accurately simulated. Such a process may be
1.2
expected to occur when the core radius increaseswith height
sufficientlythat (5) is no longer satisfied;then turbulenceis
allowed to grow and break up the core. The inner eye would
1.0 be capped at this altitude, with angular momentum trans-
ported outward by turbulence. This capping process may
have been almost captured in the laboratory results of
U/Umoa Pautey [1989], where the vertical pressuregradientalongthe
axis was observed up to a region where the strong inner
downflowappearedto be originating.I am willing to specu-
late that the vertical velocity would have been upward along
the axis just above the highestposition reported.
There appearsto be no physicalbarrier to prohibit excep-
tionally strongtornadoesfrom being associatedwith an eye
which penetratesthe entire depth of the storm. It is quite
possible,even probable,that a few F-5 classtornadoes(the
strongesttornadoesin Fujita's [197 la] rating system)have
0
0 2 4 6 8 I0 had such we!l-developedeyes. Indeed, there is some sup-
Z/Z + porting evidencethat the tornado core extendsfar up into
the thunderstorm for such storms. MacGorman [this vol-
Fig. 3. Surface layer, turbulent profiles of tangential, v, and ume] reports that lightningunderwent a sharp transition
radial, u, velocities beneath a strong vortex. Comparison of the from intracloud flashes during a strong tornado to negative
solutionpredictedby Lewellen [1977]with the profilesmeasuredby
Savino and Keshock [1965] is shown.
cloud-to-groundflashesafter the tornado. He also suggests
that the initial domination by intracloud flashes may be
associatedwith higherpenetrationof the warm moistcentral
boundary layer, (9) is stronglymodified,and some of the jet (updraft)from the surface.This suggestion is consistent
radial flow may be ejected from the boundarylayer. with the stabilitydiscussionin section2. The stabilityof the
The primary additional parameter introduced into the vortex core structure would reduce the entrainment between
boundarylayer flow, not discussedin section2.1, is some the rising warm, moist air and the cold downdraftwhich
measureof the surfaceroughness,which may be introduced appearsto be wrapping around it in the thunderstorm
as a secondReynoldsnumberwith r0 in (2) replacedby z0. simulations.
The connection to the thunderstorm makes it difficult to
4. UPPERF•.OW, REASONIV separateoutthe dynamicsin the upperregionof the tornado
The upperflowregion,whichmustnecessarily includethe vortex. Rather than attempt to model this, I prefer to
top boundarycondition,containsa largedegreeof uncer- consider the tornado vortex model as a small inner nest of a
The numericalthunderstormmodel which cannotyet be modeled
tainty,becauseit is buriedin the parentthunderstorm.
mostpopular solutionforthisinlaborfi. isto as an interactivepart of the thunderstormbecauseof insuf-
torysimulations
includea baffle,whichdissipates the angularmomentum,in ficientgrid resolution.Wicker and Wilhelmson[1990]have
theexhausting flowat thetopof thedomain.This,of course, been able to obtain some of the coarse features of a tornado
cannot occur in the real storm. The angular momentum vortex in a simulation which nested down to an inner grid
broughtinto the vortexat low levelsmustbe transferred resolution of 67 m in the horizontal directions. As will be
flowat thetopor by turbulent seenin section7, a resolution approaching10 m appearsto
outward,eitherby diverging
transport.
I arguedinL76thatatthetopofa swirling updraft be neededto incorporatethe full dynamicsof the tornado
wherethe flow is forcedto divergeoutward,thereshouldbe vortex.
If the tornado vortex is considered an inner nest of a
a secondarydowndraftinduced alongtheaxisofrotationby
the low centralpressureat lowerlevels.If thispressure- thunderstorm model,then the upperboundaryconditionsas
induceddowndraftfrom the top penetrates to the surface, well as the outer radial boundary conditions can come
thenthe stormwill havean "eye" like a miniaturehurricane. directlyfrom the thunderstormoutput.Of course,one still
There hasbeenand continuesto be controversyas to when needs to look at the sensitivity of the resulting tornado
or if an eye existsin strongtornadoes
[Anderson,1985; vortex to variations in these boundary conditions which
Bluestein,1985;CarrierandFendell,1985;Carrieret al., would be subgridscale to the thunderstormmodel.
26 TORNADO VORTEX THEORY

can lead to a local separation in the presence of the slight

adverse pressure gradient restilting from the stagnating ra-
dial velocity above the boundary layer. Relatively little swirl
is required to keep the flow attached as in Figure 4b. As S is
increased, the ratio of the volume of the throughflow passing
through the boundary layer to that above the boundary layer
increases, and the center jet is intensified. At a larger value
of S, which depends on both the inflow and the top outflow
conditions, the conditions for Figure 4c are reached. Here
the central erupting jet is stronger than can be supported
throughout the core flow by the top boundary conditions.
The restlit is a sharp transition in the tlow, a type of vortex
breakdown, at some distance above the ground. At still
highcr valucs oI'S, the boundary layer cruption occurs in an
annulus around the center, and the flow pattern of Figure 4d
appears. Whcn instabilities occur in the annular region of

ttttlt

Plate I. l•holograph o1'a tornado l'unncl over the western Kansas

plain,,I Whipple, , takent'rom
1982] a distance _3
5• mi(5.6kin).
o1'_.,--

leaves open the question of how much the thunderstorm

model depends upon its subgrid parameterization. Lilly
[19831has argued that the reluctance of helical turbulence to
cascade in an inertial range makes rotating supercell storms
much less susceptible to turbulent uncertainties and there-
fore more predictable by numerical simulations without
uncertainties due to subgrid turbulent closure assumptions.
The extent to which this is true remains to be determined.

5. Co•4k• FLow, REGION Ill

The most interestinginteraction with the surfaceoccurs in
the neighborhoodof r - 0, where the radial flow inducedby
the boundary layer must turn and produce some type of
verticaljet. The structureof thisjet dependson the relative
fraction of the volume flow which has been diverted to the
boundarylayer. This is qualitativelydeterminedby the swirl Fig. 4. Schematicof differenttypesof cornerflow [Davies-Jones,
1986]for increasingswirl: (a) very weak swirl so that the boundary
ratio. Figure 4, taken fi-om Davies-Jones[1986], shows a layerseparates,(b) one-cellvortex,(c) vortexbreakdownabovethe
schematicplot of the variationof the cornerflow with swirl surface, (d) two-cell vortex with downdraft penetrating to the
ratio. If S = 0, there is no excessflow in the boundary layer. surface,and (e) multiple vorticesrotating about the annulussepa-
Instead, there is a deficit within the boundary layer which rating the two
 LEWELLEN 27

75
Tangential
Velocity
mjs
60

45
Radial
Vertical
Velocity
70 mjs 30

15

Plate 2. Meridional cross section of tangential, radial, and vertical velocities from the LES tornado simulation
described in the text. Pressure isobars are indicated by the light blue lines.

Snapshot of 3-D Unsteady Tornado Simulation

z = 100 m
500 m 1""7,'-:,".'7,.."...,/-/=/"./"'""/'"_"'""_=,.;:",.,...;.::;;~~..,..,.,....,...., 70
°1,.,. t' / / , ; ' / ........ - - - - - - ... - , Vertical
.,1///////------" ",' Velocity
.1 I" t' / / / / / / - - - - .......... , " ... " .. mjs
III//////~-"-"""" 50
/II/////~-"""'"
IIII/////~-" " ,~""~
"Ill///~ ,,,,,.
' I I I I 11/ ::-..\~~ ~ ~ ~ ~ 30
: : \ I/l/~1. '\\~)) ~ ~ Radial
Om \l\~ 1. 1/'/.'\ 1\
Tangential
Velocity
~It 1\11
111\
I I \ \ \ ' \'
. \ \ \ \ '\ ,,\ '\....
III I I 1
/1 /1/1 ~ : :
-.
90 mjs 10

'
, \\ ,""
\ \ \ , , ,~
-...-.:
-..:
...--/// / / 1 I I 1
.-y// / / / / I I I
- " " " - , -~//////IIII ·10
", ,,------_/////1111
"" ,-------//////111
" ...... , - - - - - - _ .... ' / / / / / 1 1 1
.............. .... / / ; / / / / " I I
,500 m ,30
·500m

Plate 3. Horizontal cross section of tangential, radial, and vertical velocities from the LES tornado simulation
described in the lext.
28 TORNADO VORTEX THEORY

Dynamic Pressure Surfaces numerical simulations, but it becomes somewhat ambiguous

in the real tornado vortex because of its dependence on
domain geometry. Results of a laboratory investigation of
the characteristics of tornadolike vortices as a function of
swirl ratio are given by Chl/rch el al. [1979J. A plot of
minimum pressure versus S is given in Figure 5, as given by
Wilkells alld Diamond [1987J. They show that at least the
laboratory vortex goes through a number of transitions as 5
is increased, with both the geometry and the throughflow
held fixed. This figure also shows that variations with 5
depend upon the particular geometry. The Purdue chamber,
with its smaller aspect ratio, shows a much more distinct
minimum in pressure, where there is a transition from the
flow type in Figure 4c to that in Figure 4d [Pauley el al.,
1982]. Numerical simulations [Wilson {Ind ROil/lilia, 1986]
have been able to essentially duplicate the Purdue results at
low Reynolds number.
The flow in the corner flow is sufficiently complex that
solutions have required some degree of numerics even when
part of the analysis is analytical. When attempting finite
difference simulations, the difference between a corner flow
simulation and the full tornado vortex simulation is relatively
arbitrary since it is largely a matter of how the outer radial
and upper vertical boundary conditions are specified. I will
not distinguish between the two, nor will I repeat the
discussion of the pre-1976 finite difference solutions here.
Numerical solutions appearing in the last 15 years include
one-dimensional axisymmetric simulations by Carrier el al.
Plate 4. Three-dimensional perspective of surfaces of constant [1988J, Chen and Watts [1979J, and Gall [1982]. These are
dynamic pressure From the LES tornado simulation described in the similar to the in viscid interaction model of Lilly [1969] in that
text. Tl1e squares on the surface are 100 m on a side. enough assumptions are made about the core flow and the
boundary layer flow to allow the interaction problem to be
reduced to a set of ordinary differential equations. Gall
[1982] is successful in matching the behavior of core radius
concentrated vortiCity, coherent vortices, called suction
as a function of swirl obtained in the laboratory. Goldshlik
vortices [Forbes, 1978; Fujilo, 1970, 1971b; Fujilo et al.,
[1990] has reconsidered his conical similarity solution for a
1976], may rotate around the primary vortex as in Figure 4e.
line vortex perpendicular to a flat nonslip surface [Goldsltlik,
The sharp transition a short distance above the surface at
the center of the sketch in Figure 4c is an example of a type
t960] with some apparent modifications in the critical Rey-
of vortex breakdown. Vortex breakdowns in tubes, in wing nolds number above which no such similarity solution can
tip vortices, and in natural vortices have provided a fasci- exist. This nonexistence of such a similarity solution at even
nating subject for research [Benjamin, 1962; Browll and moderate Reynolds numbers suggests that something more
Lopez, 1990; Burggraf and Fosler, 1977; Escudier, 1986, dramatic such as vortex breakdown may occur.
1988; Hall, 1972; Leibovich, 1978, 1984; Lopez, 1990; Lugl, Laminar, two-dimensional axisymmetric, numerical simu-
1989; Sarpkaya, 1971]. The breakdown provides a jump lations have been presented by Howells and 5m;III [1983],
between two stable states of swirling flow. Upstream of the Howells et al. [1988], Leslie [1977], Leslie {Il1d Smilh [1982J,
jump the axial velocity is supercritical, in the sense that it is McClellall el (II. [1990], Proctor [1979], ROII/II1I0 [1977,
faster than the speed at which inertial waves may travel in 1979], ROIl/II1I0 alld Lilly [1981], Smilh [1987), Smith alld
this vortex environment. The axial velocity is reduced to a Howells [1983], Smith and Leslie [1979], Walko [1988], and
subcritical value across the transition, analogous to a hy· Wilsoll alld ROlI/nllO [1982, 1986]. Turbulent, two-
draulicjump in stratified Row. Even when the classic break· dimen ional axisymmetric simulations have been presented
down does not occur, axial variations in a stro.lgly swirling by Chi [1977], Lewellen alld Shellg [1979, 1980], Lewellell
flow tend to be either quite small or quite abrupt because of alld Teske [1977], and Lewellell el al. [1979]. Three-
the tight coupling between the velocity components forced dimensional simulations have been presented by Rotunno
by axial variation in the tangential velocity. [1982, 1984], Walko [1990], and Wicker alld Wilhelmsoll
The swirl ratio is easy to control in both laboratory and [1990].
 LEWELLEN 29

5.1. Influence of the No-Slip Condition smaller radius in the turbulent transport model than in the
laminarsimulation.Not surprisingly,the reductionin effec-
One of the most interestingfeatures of the corner flow is
that the radial inflow induced in the boundary layer forces
tive eddy viscosity predictedby turbulent transporttheory
close to a surface allows the maximum velocity to occur
the strongest tangential velocities to occur close to the
closer to the surface. The differences occurring at full-scale
surfaceas turningoccurs.The no-slipsurfaceboundarycon-
dition can actually increasethe maximumvelocity a short atmosphericReynoldsnumbersshouldbe even largerthan
the differences evidenced in this simulation of the moderate
distance above the surface. Velocity distributionsobtained
value Reynoldsnumber appropriatefor this laboratory ex-
whenfree-slipconditionsareapplied;for example,the results
of Walko and Gall [1986] show tangentialvelocity distribu- periment.This comparisonwas madefor swirl conditions
tions with a maximum on the surface, with a slow decrease similar to those of Figure 4(,/. Probably an even bigger
difference would have been seen at lower swirl conditions
as the vertical coordinate increases and the core slowly
approximating Figure4c. In a morerecent,high-resolution,
expands.Thismaybe contrasted
withthe no-slipresultsof laminar simulation of conditions more appropriate to this
Howells et al. [1988] exhibited in Figure 6. The no-slip
conditions result in an overshoot in the corner flow which case which incorporatesa vortex breakdown,Wilsonand
Rotunno[1986]have obtaineda maximumvertical velocity
can more than compensate for end-walllosses.Howellset
which is twice the maximum tangential velocity and occurs
al. showthat, at leastfor theirstrongswirlcaseandassumed
between the surface and the breakdown transition. I would
constantviscosity,the no-slipboundaryconditionproduces
not expecta highReynoldsnumber,turbulentsimulationto
a maximumvelocity approximately 25% greaterthan that
exhibit such a strong axial jet.
produced
bya similarsimulation
withfreeslipatthesurface. Howells et al. [1988] show that even in their laminar
At lower values of swirl the no-slip boundaryled to even
simulation,the details of the corner flow are quite sensitive
largerincreases
in the valueof themaximum
velocity. to the simulationviscosity.They concludethat "due to the
extreme sensitivityof the flow to the value of the eddy
5.2. Influence of Turbulence diffusivitycoefficient,any attemptto simulateactualatmo-
Most of the numerical simulationswhich include the sphericvorticesusing a model which incorporates self-
corner flow are for constanteddy viscosity,i.e., laminar regulatingturbulenceclosureshouldincludea thorough
flow. A comparison betweena constant eddyviscosity sensitivitytestof the parametersassumedwithinthe closure
simulationby Rotunno[1979]and a turbulenttransport approximation"
[Ho•,ellset al., 1988,p. 819].I agreewith
simulation and Sheng[1980]for conditionsthis comment and
by œewelten add that modelers who use the simplest
similarto thoseobtainedin thePurduesimulator
[Churchet turbulent closure of constant eddy viscosity should not
al., 1979]showed
thatthemaximum velocitywasa little expecttheirsimulations to provideveryvaliddetailsof the
largerandoccurred
muchcloser
tothewallandatsomewhat flow, particularlyclose to the
30 TORNADO VORTEX THEORY

NUMERICAL TORNADO-VORTEX MODELS

v w

2000

1000
,/

0
o lOOO 2000 0 1000 2000 0 1000 2000
(a) 18 minutes

15oo
t •o
lOOO

tI2
//

o
0 500 1000 0 500 1000 0 500 1000
(b) 21 minutes (note scale change)

Fig. 6. Transient meridional cross sectionsof tangential, radial, and vertical velocities from a simulation with no-slip
lower boundary condition by Howells et al. [19881. Coordinates are in meters, and velocity contours in meters per
second.

The turbulent transport simulations of Lewellen and done with S = 1.3 5, uniform u and v set at 1 km radius, and
Sheng [1980] for low Reynolds number approximating the uniform outflow set at z = 1 km. The value of S was chosen
laboratory conditions of Mullen and Maxworthy [1977] are to correspond to that deduced from Brandes' [1979] obser-
similar to that shown in Figure 6, as seen in Figure 7. The vations of the Harrah, Oklahoma, storm of June 8, 1974. The
simulation which led to Figure 7 not only had a relatively low vertical velocity for the same conditions was quite small on
Reynolds number to simulate the laboratory conditions but the axis with the largest values (0.3Vmax)occurring along a
also was stabilized by relatively high heat flux in this dust cone just inside the cone on which the maximum tangential
devil simulation. The simulated flow was very unsteady, velocity occurs.
with the height of the vortex breakdown oscillating over a The main point I wish to make in the present section is that
large region. A similar oscillation was reported for the the corner flow is quite sensitive to turbulence. Rather than
laboratory results. At the simulated time correspondingto show more results from axisymmetric turbulent transport
Figure 7, the maximum vertical velocity was more than 60% simulations, I will show new results from a more reliable
larger than the maximum tangential velocity. turbulent model in section 7.
The turbulent transport simulations of Lewellen and
Sheng [1980] over rough surfacesat high Reynolds numbers 6. ESTIMATES OF MAXIMUM VELOCITY
were typically more conical at low levels as illustrated in
Figure 8. This figure is also normalized by conditionsat the One of the most important features desiredfrom a tornado
positionof maximum tangentialvelocity. The simulationwas model is an estimate of the maximum wind speed within
 LEWELLEN 31

fewmeters
ofthesurface.
In L76I argued
thatnoexisting
modelwas adequateto providea reliableestimate.Davies-
Jones [1986]providesa review of the resultsfrom several
differenttechniques
whichhaveattempted to measure
this
parameter: photogrammetry,
groundmarks,directpassage
overinstruments,remotesensing,directprobing,
damage
analysis,and funnelcloudanalysis.Each of thesemethods
requires some knowledge of the structureof the tornado
vortex to properly interpret the measurement.Davies-Jones
concludes
thatmostscientists
agreethatwindspeeds
in even
the strongesttornadoes do not exceed 110-130 m/s-! Ad-
ditionalcarefulstatisticalanalysisof the availablemeasure-
ments are required to estimatethe probabilitiesof extreme
winds [Abbey, 1976;Abbeyand Fujita, 1979].
The maximumvelocityto zerothordermaybe determined
by thetotalpressure
dropavailableto theflowasimpliedby
(4). A number of researchers[e.g., Lilly, 1976] have dis-
cussedthe fact that an estimateof thispressure
dropcanbe
madeby takingthe differencebetweenhydrostatically inte-
0 .• 1 I .)$ Z
RADIUS
grating the environmentalthermodynamicsoundingand
integratingthe modelsoundingwhichcouldbe generatedby Fig. 8. Distribution of mean tangential velocity from a turbulent
lifting a surfaceair parcelto its ideal equilibriumlevel. The transport simulation of Lewellen and Sheng [ 1980]for S = 1.35 and
•0 = 0.04 m. Contours are labeled in tenths of the maximum
pressuredeficit (typicallyabout50 mbar)providesa rough tangential velocity, and coordinates are normalized by the radius at
guessbut is subject to considerableuncertainty. If the two- which the maximum velocity occurs.
cell core structurepersistsvertically throughoutthe storm,
i.e., if the storm has an eye, then the resultingsubsidenceof
dry, near-stratosphericair couldincreasethe availablepres- sure drop by up to a factor of 3 [Davies-Jones, 1986]. Also,
when the immediate tornado core region is surrounded by a
cold downdraft created by precipitation falling into dry air in
the midtroposphere,then the contrast between precipitation-
laden downdraft and a central dry subsidence could be even
larger. The other difficulty with this simple estimate is that
the central corner flow region in the tornado is neither
hydrostatic nor cyclostrophic, as seen in section 5.
One of the principal conclusionsfrom my 1976 review was
that "the maximum velocities will occur below'the top of the
ground boundary layer (•--100 m above surface) and may be
significantly higher than that predicted by a cyclostrophic
balance" [Lewellen, 1976, p. 136]. There were many doubt-
ers, since one is conditioned to expect the boundary layer to
reduce wind speeds. In fact, as recently as 1988 Carrier et al.
[1988] regarded this "as unlikely" to be true, although they
did admit to the possibility of a small overshootin the comer
flow region. Nevertheless, numerous numerical simulations
[e.g., Howells et al., 1988; Lewellen and Sheng, 1980] have
confirmed that the overshoot can be quite significant. Lab-
oratory models have also noted that low-level pressureson
the centerline may be significantlylower than that obtained
by cyclostrophicbalance above the boundarylayer [Baker,
1981]. This led Snow and Pauley [1984] also to criticize the
hydrostatic,thermodynamicmethod of estimatingmaximum
tornado velocities.
Fig. 7. Transientmeridional
crosssections
of tangential
velocity Fiedler and Rotunno [1986]provide a physicaltheory for
obtained in the turbulent transportsimulationof Lewellen and why the most intense laboratory vortex velocities occur
Sheng[1980]for a relativelylowReynolds numberapproximatingclose to the surface for the flow type representedin Figure
thelaboratory conditionsof MullenandMaxworthy[1977].Velocity
isnormalized bythemaximum bythevalueof 4c. They argue that the erupting boundary layer flow is
valueandcoordinates
the radius at which the maximum occurs. supercritical and thus permitted to have an
32 T(')RNADO V()RTEX "FItEORY

•owcr prcssurcthan that occurring along the centerline in the

subcritical flow, which is in cyc!ostroph•c balance on the
other s•cle ot' a vortex breakdown transition. Their analysis
shows that the maximum azimuthal velocity at low levels of -'- •+ 2o'v (m/see)
a laminar vortex can be approximately 1.7 times the max,-
mum azimuthal velocity in the subcritical vortex alot•. They
point out that a turbulent boundary may result in a somewhat >,,
06 •\
smaller effect, but demonstrate •'in any event, that the
boundary layer does not simply act as an energy sink•in a 125

much more subtle way it acts to focus the energy of the

larger-scalecycloncti)ward the smaller-scaletornado" [Fie- '*''""•'" 50
diet t•nd R•lun•o, 1986, p. 23391. 25

As staled •n section 2.3, I do not accept the subscqucnt C,•

0
I
.01
I ,,
.02
, I,,,
.0;5
I
.04
I
.05
,I

argumentby b'i•,dlcr[1989]that the full-scaletornado corner ah,Horizontalconvergence

(sec-•)
11owmay bc laminar. However, I do agrec with P'iedlerand
Rolttnno's [19861 basic analysis. Also, I believe the intense Fig. 9. Maximum tangentialvelocity to be expectedin a tornado
velocities occurring •n the corner region can be viewed •n as a I'unction of ambient vertical vorticity (to) and horizontal
convergence(ah) measuredin atcylinderwith 0.8-km radiusfor a
st)mcwhatsimp!crand more general terms. The corner t]ov• relativelysmooth.surface(z0 = 0.05 m, dashedlines) and a rough
velocities •trc allowed to exceed those supported by tint .,uH'ace(:0 = 1.4 m, •solidlines) accordingto the simulationsof
driving prcssurcdeficit availablc aloli, becauselhc expand- l,e•'clh'n and Shetu4119801.
ing, swirling flow d•rcctly above the region of' maximum
vcl()cily'behaves as a (lifl't•ser,providing a mechanismtbr
pressure rccovery. •l'his is truc whether there is a very thunderstorm model. In the study resulting in Figure 9,
ct!icicnt presstire recovery as octtits i•l a sharp, laminar increasingthe ratio of the vertical velocity along the center-
v•)rtcx brcakdown ()r there is ll}c less efficient pressure line at the top ot'the domainto the averageticrossthe top by
recovery which (•ccurs in a region t)i' nearly conical, turbu- a factorot'2 produceda 15% increasein maximumvelocity,
lent 11owcvidenl in the results ot' Lc•'e!len andShcng 119801. while decreasingthe same ratio to zero produceda 15%
II' swirlingtlow is tbrccd througt•a conventionalnozzle, one decreasein maximum velocity. On the basisot' theseresults
is not surprisedthat the maximum velocities occur in the we concludedthat the maximum velocity is less sensitive to
minimum cross scction ol' the nozzle. In this atnalogy the the details of the outflow at 0.8 km than it is to the ambient
central corner 11owis the region ot' minimum cross section vorticity and convergence.
tk)rthe 11ow.The subsequentmaximumvcl()citicswhich occur t:t(/ita[19701hatsshownthat maximumdamageis gener-
ztt this cross sccti(m arc determined by lhc ctlicicncy ()1'the ally associatcdwith unsteadypeak velocitieswhcre the
prcsst•rcrccovcryprocess,as well as by the driving prcssure •suction vortex velocity" adds to the mean tornado veloc-
drop. 'l'husthe uniqueinteractionswhichoccurin the corner ity. In our axisymmetricmodelthese suctionvorticeswere
t]()wR)cusingprovidea subtleIlow arrangementlk)robtain- interpretedasturbulenteddies.Sincethe turbulenttransport
ing tt local maximumin kinetic energy which exceedsthat model provided a predictionof the root-mcan-square(rms)
generallyexpectedfi'om the available stormenergy. fluctuations about the mean velocity, twice the rms tangen-
tial velocitywas addedto the mean tangentialvelocityin
It' the tornado vortex domain is considered as a small
nested region within the thunderstorm,as suggestedin constructingFigure9, whichwas intendedto providemax-
section4, then the maximum velocity shouldbe determined imum wind speedsto be usedfor damageestimates.
by the boundaryconditionsimposedby the thunderstormon The dashedlinesin Figure9 denotethe contoursfor winds
such nests.Two primary inputs requiredfor sucha limited overa relativelysmoothsurfaceandthe solidlinesthosefor
domain simulation are the ambient vertical vorticity and a roughsurface.If a thunderstorm modelor analysisof a
horizontalconvergencefrom the desiredregionof the thun- Dopplerreturnindicates an averagevalueof verticalvortic-
derstorm.Although thunderstormmodelswere not capable ity equalto 0.03 s-• andhorizontal convergence equalto
of supplyingtheseparametersin the late 1970s,Lewellen 0.01 s-• Ibr a cylinderwith a 0.8-km radiusand a 0.8-km
and Sheng [1980] did attempt to see how the maximum heightabovethe surface,then Figure9 would predicta
velocitywoulddependuponthesetwo variables.The results maximumprobablewindof 75 m/s, if the tornadois occur-
from their turbulent transport model, which are shown in ring over smoothprairiegrassland(z0 = 7 cm). This
Figure 9, were intendedto help interpretDoppler radar maximum would be reduced to 60 m/s over rough terrain
results.For pu•oses of thisfigurethe innernestradiusand withz0 = 1.6 m. Thelabeledpointsarevaluessupplied
by
the top of the domainwere both set at 0.8 km, aboutthe Brandesfrom the analysisof dual Doppler radar observa-
limitsof Dopplerradarat the time. The strongestvelocities tions of the Hurrah and Del City tornadoes[Brandes,1979,
requireboth significantvorticity and convergence. These 1981].It is not clear whethersufficientstormshave been
resultsmaybe expectedto vary somewiththedetailsof the capturedinthismannerinthelast10years[e.g.,Hennington
inflow and outflow which would be subgrid scale to any et al., 1982;Uttal et al., 1989;Zrnic et al., 1985]to
 LEWELLEN 33

T•nti•l V•bcity Tsn•nt• Wbcity

CINT. 1.0 Trn• - 7380 C•'T. 1.0 Trne. ;'420
3.4 3.4

0
0 R - Distmoe
InKM 2 0 R-•rtKM 2

Fig. 10. Tangential velocities exhibited by Wicker and Wilhelmson [1990] at different times in their tornado nest
within a thunderstorm simulation. Minimum resolution is 67 m.

a confident bound on the combination of these two variables. the largest eddies responsible for the principle transport
However, it appears that numerical storm simulations such within the flow. It is still not possible to do a direct
as those given by Wicker and Wilhehnson[1990] should now simulation of turbulence at full-scale Reynolds numbers, so
be adequate to supply their own estimates of these limits. some subgridclosure model is still required. However, the
It may be noted that the constant maximum velocity more resolution which can be incorporated into the simula-
curves in Figure 9 have the same general shape as the tion, the less importance the subgrid model assumes.
tornado intensity maps given by Johns et al. [this volume], Rotunno [1984] made a three-dimensional simulation of
where the convergence coordinate is replacedby convective the Purdue simulation chamber under sufficiently large swirl
available potential energy (CAPE) and the vorticity coordi- conditions to exhibit secondary "suction" vortices spinning
nate is replaced by the ambient vertical wind shear. Al- around the central core. He used a constant eddy diffusivity
though theseplots are for quite different domain scales,their with a Reynolds number (uoro/v) = 150. More recently
similarity is much more than just coincidental.CAPE pro- Walko [1990] and Wicker and Wilhelmson [1990] have made
vides a direct measure for possibleconvergenceinto coher- three-dimensional tornado simulations, but both of these
ent plumes, and low-level vertical wind shear, horizontal were emphasizingsomewhat larger flow features. Wicker
vorticity provides a direct source of vertical vorticity via and Wilhelmson were looking at a tornadonestedgrid within
tilting [Davies-Jones,1986].On the otherhand,I do not wish a supercellstorm. Walko was looking at ways of introducing
to imply that Figure 9 providesthe definitiveanswerfor the the vorticity into the tornado. It is quite impressive that
maximumvelocity question.Too many uncertaintiesremain Wicker and Wilhelmson were able to zoom in from the
about the sensitivity.to domainboundaryconditionsand the thunderstorm scale to the tornado scale and see the features
turbulenttransportmodel. If one were to attemptto produce exhibitedin Figure 10. They were ableto seethat the surface
a figure equivalent to Figure 9 appropriatefor laboratory no-slipconditionincreasesthe maximumvelocity and that
vortices,it would have many kinks relatingto the transitions the maximum velocity occurs at low levels. In fact, they
in Figure 5 and a dependenceon bothReynoldsnumberand found that the no-slip surfaceboundaryconditionincreased
geometry. the vorticity at low levels by 50% over that obtainedin a
free-slipsimulation.However, to resolvethe coredynamics
7. LARGE-EDDY SIMULATION TORNADO MODEL we have been discussingin this review, it is necessaryto
As mentionedin the introduction, fully three-dimensional, zoom in to even smaller scale.
unsteady simulationsare now easier, than steady state, While preparingfor this review, I askedmy colleaguesat
axisymmetriccalculationswere 15 yearsago.This presents Aeronautical Research Associates of Princeton to use the
one way of at least partially circumventingthe problems large-eddysimulation(LES) code describedby Sykesand
associatedwith modeling turbulent transport within the Henn [1989]to attemptan LES modelof tornadodynamics.
tornado, that is, to providesufficientresolutionto simulate As mentionedearlier, Schmidt and Schumann [1989]
•4 'I'()RNAD() V()RTEX THEORY

tlscd a large-eddy simulation of the convective boundary lute tangential velocity out of the plane, and the arrows
layer ti• look at the conditions where coherent plumes are representthe radial/vertical velocities in the plane. The solid
initiated. ()ne might look at the conditions which allow these lines represent three pressurecontours when a hydrostatic
t{• becomequite distinct dust devils, but our purposehere is vertical pressure gradient is added to the dynamic pressure
quite dit!krcnt. It is to provide sufficient resolution in the gradientsgenerated in the simulation. One of these pressure
corner ttow region to allow the dominant turbulent features contours might approximate the funnel contour depending
to bc resolved. We set the horizontal boundary conditions to upon the condensation level at the time. Even with the
approximate those expected on a 1-km cube with the tornado translation, the flow looks like an axisymmetric flow with
at the center of this cube. These horizontal boundary con- superimposed turbulence. The nature of the turbulence is
diti•ns wcrc taken as axisymmetric about the center. The clearer in the horizontal slice shown in Plate 3. In this plate
t•tngcnt•al velocity was taken t(• have a vertical variation the color coding represents the vertical velocity, while the
pr{•p•)rt•onal t{) In {::/z{}) as expected in a fully turbulent arrows represent the horizontal velocity vectors. There are
boundary layer. •i'hc radial velocity was given a variation at least three coherent eddies, each with their own up and
•tpproximating that of }'::•gurc3, with a maximum inward down vertical velocities rotating about the primary vortex.
ra.d•al vcl{•c•ty at approximately 3(1 m height. The vertical These three eddies show up even more clearly in the
vcl•}city w•ts set t(• zero. 'l'hc top of the domain was set aI 2 three-dimensional perspective of two contours of the dy-
k•n where the vertical velocity was assumed unilbrm within namic pressure in Plate 4. The lowest dynamic pressures
;• cylinder with a 5(}(}-111 raditlS. ()utfiow boundary conditi½'>ns occur just a little above the surface as illustrated by the dark
•,rc applied to the rcm•tining variables at this top boundary. blue contours. The central dark blue region represents the
'I'hc grid inc[t•dcd 74 / 74 unil'orm horizontal points and 50 center of the main vortex, while the third •'suction vortex"
pt)this siretched in the vertical. Initial conditions were taken is not quite strong enough to show this contour, but the
•:tsappr{)xim•tting a Burgers-Rott vortex above the surf'ace lighter color arm is clearly in evidence.
layer. As was evident in the simulations discussed in section 5,
["t}l' the rcst•lts shown in Plate 2, the horizontal velocities the radial tlow penetrates to a smaller radius near the surface
at the t)utcr bt)tlndary were ct)nsistcnt with axisymmetric and resultsin a higher tangential velocity in this region. The
ct)ndilit)ns witIn b{)th the circulation and the radial volume maximum vertical velocities occur slightly off the center and
flow abovethe surl'acclayer constantat -;,r " 1.5 • 104 are somewhat lower than the maximum tangential velocity
m2/sandllr :::':
2 / I{}'• m2/s.l)t•ringthecourseof themodel which reaches a local maximum of 90 m/s. The overall flow
sin•ulation spin-up, these ht>rizontal boundary conditions pattern is quite similar to the results of the axisymmetric,
wcrc held fixed except a 2{>rJ:
, random variation was permit- turbulent transport model of Le•,e!le• a,'•d Sherig [1980]
ted I,) simulate this level of turbulence in the inflow. The shown in Figure 8. It is interestingto note that although a
st•rl'•ccb{mndary c½•nditions
assumeda surtktcemoving with uniform vertical outflow was imposedwithin a radius of 500
a c½}nst:tnttranslati½•n o!' 15 m/s and with an equivalent m at the top of the computational domain (at 2 km), the
%111'['[tCC
l'(}[lghncs%o[' (}.2 m correspondingto conditionsin vertical velocity tends to be concentrated in an annulus
between the rough and the smooth surlStceconditionscon- about an eyelike central region over most of the lowest
s•dcrcd by Lc•'elle•t a•d She•t.•,[1980]. The vortex spun up kilometer. On the ba:sisof the study of the sensitivity of our
to essentiallya quasi-stationaryturbulent tlow aller about 3 axisymmetricsimulationsto variationsin the top boundary
m•n of simulated time (-•2000 time steps). Atter this quasi- condition [Letvellen and Sheng, 1980], previously discussed
stationarystate was achieved,the velocitieswere savedon a in section6, I expect these low-level results to depend on
vertical slice through the center and on a horizontal s!•ceat details in the prescribeddistribution of the outflow at 2 km
100 m height at t?equcntintervals to gather statisticson the but not be highly sensitiveto them.
turbulence in the flow. Some 26 samples were taken at A snapshotof the helicityin the verticalslice(not shown)
intervals correspondingto approximatelyone third of the indicatedthat the hellcityis primarily concentratedin braids
periodof rotation in the regionof maximumvelocity. within the core. Somewhatsurprisingly,the low-level region
The questionof how successfulwe were at resolvingthe where the maximum value occurs indicated a strong asym-
primary turbulencein the simulationmay be answeredby merry. It is not clear how much of this asymmetrywas
comparingthe turbulenceresolvedby the simulationwith associatedwith the rotating"suctionvortices" evidencedin
thatrepresented
by thesubgridparameterization
of theflow. Plate 3 and how much was associated with the fact the
The maximum rms of the resolved velocity fluctuations was tornado is translating to the right.
approximately
3 timesthe maximumsubgridrmsvelocity.If This simulation raises as many questions as it answers.
Someof thesequestions
this ratio held everywhere in the flow, it would indicatethat areasfollows:What is theunsteady
90% of the turbulent energy had been resolved. We are not behaviorduringstart up and decay?How muchinfluence
claimingto havedonethiswell, but we believethatwe have would buoyancyhave on dampingthe turbulence?How
capturedenoughof theturbulencefor thesimulation to have sensitiveis the flow to input parameterssuch as the swirl
reasonable validity. ratio, surfaceroughness, detailsof the horizontalinflow,and
Plate 2 showsa snapshotof a vertical plane throughthe details of the upperoutflow conditions? The only variation
center of the vortex. The color coding representsthe abso- we triedin thispreliminary investigation wasa reduction
 LEWELLEN 35

theswirlinputratio.Theresultwasa decrease
in thecore achievespeedsthatexceedthosegivenby a straightconver-
radiusanda modestdecrease in themaximum velocities.sionof availablepotentialenergy'
to kineticenergy.
However,
sincethisvariation
alsowascomputed
witha 5. "No existingmodelis adequateto providea reliable
decreasein horizontalresolution, the combination
of de- theoretical estimate of the maximum wind velocities which
creasedcoresizeand decreased resolution
forcedmostof may occurin a tornado"[Le•'ellet•,1976,p. 1361.The key
the turbulenceinto the subgridunresolved
mode. word here is reliable. There have been a numberof ap-
proaches used to estimate maximum velocities, but each is
8. CONCLUSIONS
ANDRECOMMENDATIONS subjectto a numberof questionable assumptions.After
reviewingthe differingapproachesto estimatingtornado
To summarize, I will reviewhowmuchmy conclusionswinds, Davies-Joues [1986] concludes that most scientists
abouttornadovortexdynamics havechanged
in thelast15 agreethat wind speedsin even the strongesttornadoesdo
years and recommendsomedirectionsfor future research. not exceed110-130 m/s. I agreewith this statement.Further,
I think the technologyis now at hand, where it shouldbe
8.1. Updating the 1976 Conclusions possibleto apply boundary conditions from a numerical
storm simulation to a tornado domain simulation and obtain
There were six conclusions
listedin my earlierreview, valid estimates of the probability of different levels of
L76. It is interestingto see how each of these shouldbe extreme winds occurring in different storms.
changed
in lightof research
resultsoverthelast15years.I 6. •'The role of turbulence appears to be critical in
will first quote the earlier conclusionand then discussitsdeterminingboth the maximum velocity and the detailed
present applicability. structureof a tornado" [Le•t'ellen, 1976, p. 136]. I still
1. "A relatively consistentqualitativemodelof the flow believethis is true. I believethat growthand dampingof
in differentregionsof thetornadohasbeenpiecedtogether turbulenceis crucial to the detailsof the overexpansionin
from previoustheoretical work" [Lewellen,1976,p. 136]. the cornerflow, which has a major impacton the maximum
The qualitative view of the tornado vortex flow has not velocity.I alsoexpectturbulenceto becriticalto cappingthe
changed thatmuchin thelast15years.Themostimportant core of the tornado,when an eye doesnot extendthrough
qualitativechangeshavebeenon the slightlylargerscale the storm.
which connectsthe tornadoto the parentstorm.There has
8.2. Recommet•dations
beenconsiderable progress madetowarddefining
thedetails
of the interactionsin the differentregions. The current numerical capabilities should make the next
2. "The main parameter governingthe flow is a swirl few years a fertile time for expandingour knowledgeof
ratio (which should be determinedby some ratio of the tornado dynamics. I offer the following partial list of re-
ambient instability to the ambient rotation) with the detailed searchproblemswhich might very' usefullybe completedin
structuremodifiedby the surfaceroughnessand stabilityof the next few years: (1) turbulent dynamicsin the tornado, (2)
the groundboundarylayer" [Le,•'ellen,1976,p. 136].There mappingof wind field sensitivitiesto variousinputswith a
is perhapsa little morerecognitionnowthata singleparam- tornado model utilized as an inner nest of a numerical storm
eter cannot be isolated from the rest of the flow inputs, but model, (3) unsteady,behaviorduring start-upand decay, (4)
the swirl ratio certainly remainsone of the most important prediction of maximum winds for observableinputs, (5)
parameters. Design Basis Tornado for wind-engineeringstudies to re-
3. "At least three types of wind distributionsare likely to place Fujita's [1977] model, (6) dispersionof debris by
occur immediately above the ground in the center of differ- tornado wind field, •7) statistical model for winds at 10 m
ent tornadoes. One with a central updraft, a secondwith a height exceeding !00 m/s, and (8) interaction of tornado
central downdraft, and a third with a weak central downdraft winds with typical buildings.
and secondary vortices rotating around the primary vortex" This list is not intended to be all-inclusive, but to indicate
[Lewellen, 1976, p. 136]. This shouldprobably be expanded the types of problems which might be solved in the near
to four types by counting two types with a low-level central future given a modest level of effort. The first items in this
updraft, dependingupon whether or not a vortex breakdown list dealwith obtaininga morecompleteunderstanding of the
occurs a short distance above the surface. The fifth type of tornado wind field and what controls it. while the last items
central flow illustrated by Davies-Jones[1986]and included deal with providingusefulwind-engineering
information.
as in Figure 4a cannot be classifiedas a tornado.
4. "The maximum velocities will occur below the top of Acknot,'ledgments. I want to thank R. 1. Sykesand S. F. Parker
the ground boundary layer (--•100 m above surface)and may for the LES calculationsand graphics.The LES code which made
be significantlyhigherthanthat predictedby a cyclostrophic thesesimulationspossiblewas developedunderONR supportwith
balance" [Lewelten, 1976, p. 136]. This conclusiongener- R. F. Abbey' as technical monitor.
ated some controversyin 1976but has been confirmedby
REFERENCES
subsequentnumericalandlaboratorysimulations,alongwith
theoreticalanalyses.As discussed in section6, a very unique Abbey, R. F., Risk probabilities associated with tornado wind-
feature of the corner flow is its tendencyto overexpandand speeds,in Proceedings,S)'tnposittmon T•rmtdocs:Assessmct•t
36 T(•RNAD() VORTEX THEORY

Ktu•wle&,,e
attd Implicatioux.[brMan, editedby R. E. Peterson, tion normal to a solid boundary, Ph.D. dissertation,Iowa State
pp. 177-236, Texas 'I"echUniversity, Lubbock, •976. Univ., Ames, 1972.
Abbey,P,.F., andT. T. Fujita,The DAPPLE Methodforcomputing Chen,J. M., andR. G. Watts,A steadyvortexdrivenby a heatline
tornado hazard probabilitie.s:Refinement and theoretical consid- source, lnt. J. Heat Mass Tranajkr, 22, 77-87, 1979.
erations,in Preprints, 11th Conj?renceon SevereLocal Storms, Chi, j., Numericalanalysisof turbulentend-wallboundarylayersof
pp. 241-248, American MeteorologicalSociety, Boston,Mass., intensevortices, J. Fluid Mech., 82,209-222, 1977.
1979.
Chi, S. W., andT. Costopolous,Windloadingsongroundstructures
Anderson,C. E., The Barneyeldtornado:A new type of tornadic in intenseatmospheric vortices,1, annualreport,grantGK-41469,
stormin the form of a spiralmesolow,in Preprints,14thCon.
I•kr- Natl. Sci. Found., Washington, D.C., 1975.
cute on Severe Local Storms, pp. 289-292, American Meteoro- Chi, S. W., and W. J. Glowacki,Applicabilityof mixinglength
logical Society, Boston, Mass., 1985. theory to turbulent boundary layers beneath intense vortices, J.
Andre, J. C., and M. l•esieur, Influence of helicity on the evolution Appl. Mech., 41, 15-19, 1974.
t•t'isotropicturbulence at high Reynolds number,J. Fluid Met'h., Chi, S. W., and J. Jih, Numericalmodelingof the three-dimensional
81, 1987-2007, 1977. flowsin the groundboundarylayer of a maintainedaxisymmetric
Anthes, R. A., Tropical ('y{'lones: 7'heir Evolution, Structure and vortex, Tellus, 26,444-445, 1974.
I(/.]•'cts,Metcorol. Mono.•r. Set'., vol. 19, American Meteorolog- Chi, S. W., S. J. Ying, and C. C. Chang, The ground turbulent
ical Society, Boston, Mass., 1982. boundary layer of a stationary tornado-like vortex, Tellus, 21,
Baker, (•. I,., Boundary layers in laminar vortex tlows, Ph.l). 693-700, 1969.
disscrtalion, Purdue Univ., West l,afayette, Indiana, 1981. Church, C. R., J. T. Snow, G. L. Baker, and E. M. Agee,
Barcilon, A. 1., Vortex decay above a stationaryboundary,J. Fluid Characteristics of tornado-like vortices as a function of swirl
Mech., 27, 155-175, 1967. ratio: A laboratory investigation, J. Atmos. Sci., 36, 1755-1766,
Belcher, R. J., O. R. Burggraf,and K. Stewartson,On generalized- 1979.
vortex boundary lttyers, J. Fluid Mer'h., 52,753-780, 1972. Church, C. R., J. T. Snow, and J. Dessens,Intense atmospheric
1tcllamy-Knights, P. (j., An unsteady two-cell vortex solution of the vortices associated with a 1000 MW fire, Bull. Am. Meteorol.
Navicr-St()k½,• equations, J. Fhtid Met'It., 41, 673--687, 1970. Sot., 61,682---694, 1980.
llcl!amy-Knights, P. (•., Unsteady multicellular viscous vortices, J. Davies-Jones,R. P., Tornado dynamics,in ThunderstormMorphol-
l"[uid Mc•'h., 50, 1-16, 1971. ogy and Dynatnics, 2nd ed., edited by E. Kessler, pp. 197-236,
!leilamy-Knights, P. (;., An axisymmetric boundary layer solution University of Oklahoma Press, Norman, 1986.
for an unsteady vortex above a phme, Tellu,s,26, 318-324, 1974. Davies-Jones, R., and D. Burgess, Test of helicity as a tornado
1tcngtss•)n,1.......and J. !•ighthill teds.), lntcn.s'eAttnoxpheric Vorti- forecast parameter, in Preprints, 16th Con./i'rence on Severe
•'e.s,Springer-Verlag, New York, 1982. Local Storms, pp. 588-592, American Meteorological Society,
Boston, Mass., 1990.
Benjamin, 'I'. B., Theory ot' the vortex breakdown phenomenon,J.
l:luid Mech., I4, 593-629, 1962. Davies-Jones, R. P., R. Rabin, and K. Brewster, A short-term
Bilbro, J. W., H. B. Jeffreys, J. W. Kaufman, and E. A. Weaver, forecast model for thunderstorm rotation, in Proceedings, Now-
casting-II Symposium, pp. 367-371, European Space Agency,
l,ascr Doppler dust devil measurements, NASA Te('h. Note,
Neuilly, France, 1984.
1)-8429, 1977.
Deardorff, j. W., Numerical investigation of neutral and unstable
Bluestein, H. B., Wall clouds with eyes, Mort. Weather Rev., 113,
planetary boundary layers, J. Atmos. Sci., 29, 91-115, 1972.
1(181-1(185, 1985.
Deissler, R. G., Models for some aspectsof atmospheric vortices, J.
1trdewadt, U. 'F., !)ie l)rchstrrmung fiber restern Grund, Z. Attgew Atmos. Sci., 34, 1502-1517, 1977.
Math. Mc½'h., 20,241-253, 194{).
Dergarabedian, P., and F. E. Fendell, Parameters governing the
Brandes, t5. A., '!'ornadic mesocyclone finestructure and implica- generation of free vortices, Phys. Fluids, I0, 2293-2299, 1967.
tit)ns for t½)rnad()genesis,in Preprints, I lth Con.[•'renceon Severe Dergarabedian, P., and F. E. Fendell, One- and two-cell tornado
Lo('al Stormx, pp. 549-556, American Meteorological Society, structure and funnel-cloud shape, J. Astronaut. Sci., 21, 26-31,
Boston, Mass., 1979. 1973.
Brandes, E. A., Fine structure of the Del City-Edmond tornado Eagleman, J. R., V. U. Muirhead, and N. Williams, Theoretical
mesocirculation, Mon. Weather Rev., 109, 635-647, 1981. Compressible Flow Tornado Vortex Model, in Thunderstorms,
Brown, G. L., and J. M. Lopez, Axisymmetric vortex breakdown, Tornadoes, and Building Damage, pp. 101-131, D.C. Heath,
2, Physical mechanisms, J. Fluid Mech., 22I, 553-576, 1990. Lexington, Mass., 1975.
Burgers, J. M., A mathematical model illustrating the theory of Einstein, H. A., and H. Li, Steady vortex flow in a real fluid, Houille
turbulence, Adv. Appl. Mech., 1, 197-199, 1948. Blanche, 10, 483-496, 1955.
Burggraf, O. R., and M. R. Foster, Continuation or breakdown in Ekman, V. W., On the influence of the Earth's rotation on ocean
tornado-like vortices, J. Fluid Mech., 80, 685-703, 1977. currents, Ark. Math. Astron. Fys., 2, 1-52, 1905.
Burggraf, O. R., K. Stewartson, and R. Belcher, Boundary layer Escudier, M., Vortex breakdown in technology and nature•
induced by a potential vortex, Phys. Fluids, 14, 1821-1833, 1971. Theories, in Introduction to Vortex Dynamics, vol. 2, edited by R.
Carrier, G. F., Swirling flow boundary layers, J. Fluid Mech., 49, Granger and J. F. Wendt, yon Kgrm,4.nInstitute, Rhode Saint
I33, 1971. Gen•se, Belgium, 1986.
Carrier, G., and F. Fendell, Conditions for two-cell structure in Escudier, M.P., Vortex breakdown: Observations and explana-
severe atmospheric vortices, in Preprints, 14th Conference on tions, Prog. Aerosp. Sci., 25, 189-229, 1988.
Severe Local Storms, pp. 220-223, American Meteorological Fiediet, B. H., Conditionsfor laminar flow in geophysicalvortices,
Society, Boston, Mass., 1985. J. Atmos. Sci., 46,252-259, 1989.
Carrier, G. F., P. Dergarabedian, and F. E. Fendell, Analytical Fietiler, B. H., and R. Rotunno, A theory for the maximum
studies on satellite detection of severe two-cell tornadoes, NASA windspeedin tornado-likevortices, J. Atmos. Sci., 43, 2328-2340,
Tech. Note, CR-3127, 1979. 1986.
Can-ier, G. F., F. E. Fendell, and P.S. Feldman, Analysis of the Forbes, G. S., Three scales of motion associated with tornadoes,
near-groundwind field of a tornado vortex, in Preprints, I5th Rep. NUREG/CR-0363, U.S. Nucl. Regul. Comm., Washington,
Conference on Severe Local Storms, pp. 331-334, American D.C., 1978.
MeteorologicalSociety, Boston, Mass., 1988. Foster, M. R., and F. T. Smith, Stability of Long's vortex at large
Chaussee,D. S., Numerical solution of axisymmetric vortex forma- flow force, J. Fluid Mech., 206,405-432,
 LEWELLEN 37

Franz,H. W., Die Zellstruktur

stationarer
Konvektionswirbel,
vortex flow, final report, NOAA grants N22-200-72(G)and 04-4-
Beitr. Phys.Atmos.,42, 36-66, 1969. -22-13, Univ. of Okla., Norman, 1975.
Fujita,T. T., TheLubbock
tornadoes:
A study
of suction
spots, Johns, R. H., P. W. Leftwich, and J. M. Davies, Some wind and
Weatherwise,23, 160-173,1970. instabilityparametersassociatedwith strongand violent torna-
Fujita,T. T., Proposed
characterization
oftornadoes
andhurricanes does, this volume.
by areaandintensity,
SMRPPap.9I, Dep.of Geophys.
Sci., Kidd, G. H., Jr., and G. J. Farris, Potentialflow adjacentto a
Univ. of Chicago,Chicago,I11.,1971a. stationarysurface, J. Appl. Mech., 35,209-215. 1968.
Fujita,T. T., Proposed
mechanismsof suction
spots
accompanied Kuo, H. L., On thedynamicsof convectiveatmospheric
vortices,j.
by tornadoes, SMRPPap.102,Dep.of Geophys.
Sci.,Univ.of Atmos. Sci., 23, 25-42, 1966.
Chicago,Chicago, i11., 1971b. Kuo, H. L., Note on the similaritysolutionsof the vortex equations
Fujita,T. T., Tornadostructure
for engineering
applications
with in an unstably stratified atmosphere,J. Atmos. Sci., 24. 95-97,
design-basis tornado
model(DBT-77),report,contract
AT{49-24)- 1967.
0239, Univ. of Chicago,Chicago,II1., 1977. Kuo, H. L., Axisymmetric flows in the boundary laser of a
Fujita,T. T., G. S. Forbes,T. A. Umenhofer,E. W. Pearl,andT. T. maintained vortex, J. Atmos. Sci., 28, 20-41. 1971.
Tecson,Photogrammetric
analysis
oftornadoes,
in Proceedings,Kuo, H. L., Vortex boundarylayer under quadraticsurfacestress,
Symposiumon Tornadoes:
Assessment
ofKno•vledge
andImpli- Boltndary Layer Meteorol., 22, 151-169,1982.
cations
for Man, editedbyR. E. Peterson,
pp.43-88,TexasTech Leibovich, S., The structureof vortex breakdown,Annu. Rev. Fluid
University, Lubbock, 1976. Mech., 10, 221-246, !978.
Fulks,J. R., Onthemechanics
of thetornado,
NOAATech.Memo, Leibovich,S., Vortex stabilityand breakdown'Survey and exten-
ERLTM-NSSL 4, 33 pp., 1962. sion, AIAA J., 22, 1192-1206, 1984.
Gall, R. L., Internaldynamicsof tornado-like
vortices,J. Atmos. Leibovich, S., and K. Stewartson, A sufficientcondition for insta-
Sci., 39, 2721-2736, 1982. bility of columnar vortices, J. Fluid Mech., 126, 335-356, 1983.
Gall, R. L., A linearanalysis
of themultiplevortexphenomenon
in Leslie, F. W., Surface roughnesseffects on suction vortex forma-
simulatedtornadoes,J. Atmos.Sci., 40, 2010-2024,1983. tions: A laboratory simulation, J. Atmos. Sci., 34, 1022-1027,
Gall, R. L., Lineardynamicsof themultiple-vortexphenomenon
in 1977.
tornadoes,J. Atmos. Sci., 42,761-772, 1985. Leslie, L. M., and R. K. Smith, Numerical studies of tornado
Gall, R. L., andD. O. Staley,Nonlinearbarotropic
instability
in a structureand genesis,in IntenseAtmospheric Vortices, edited by
tornadovortex, in Abstracts,ThirdConference on Atmospheric L. Bengtssonand J. Lighthill, pp. 205-211, Springer-Verlag,New
andOceanicWavesandStability,p. 34,AmericanMeteorological York, 1982.
Society, Boston, Mass., 1981. Lewellen, W. S.. A solution for three-dimensional vortex flows with
Golden, J. H., Waterspoutsandtornadoesover southFlorida, Mon.
strong circulation, J. Fluid Mech., 14,420-432, 1962.
Weather Rev., 99, 146-154, 1971.
Lewel!en, W. S., Theoretical models of the tornado vortex, in
Goldshtik, M. A., A paradoxical solution of the Navier-Stokes
equations, Prikl. Mat. Mekh., 24, 610--621, 1960. Proceedings, Symposium on Tornadoes: Assessmentof Knowl-
Goldshtik, M. A., Viscous-flow paradoxes,Annu. Rev. Fluid edge and Implicationsfor Man, edited by R. E. Peterson,pp.
Mech., 22,441-472, 1990. 107-143, Texas Tech University, Lubbock. 1976.
Grazulis, T. P., and R. F. Abbey, Jr., 103 years of violent torna- Lewellen, W. S., Influence of body forces on turbulent transport
does... patterns of serendipity,population,and mesoscaleto- near a surface, J. Appt. Math. Phys., 28, 825-834, 1977.
pography, in Preprints, I3th Conference on SevereLocal Storms, Lewellen, W. S., and Y. P. Sheng, Influence of surface conditions
pp. 124-127, American Meteorological Society, Boston, Mass., on tornado wind distributions, in Preprints, I!th Conference on
1983. Severe Local Storms, pp. 375-381, American Meteorological
Gutman, L. N., Theoretical modelof a waterspout,Bull. Acad. Sci., Society, Boston, Mass., 1979.
USSR, Geophys. Ser., Engl. Transl., I, 87-103, 1957. Lewellen, W. S., and Y. P. Sheng, Modeling tornado dynamics,
Hall, M. G., Vortex breakdown, Annu. Rev. Fluid Mech., 4, Rep. NUREG/CR-2585, U. S. Nucl. Regul. Comm., Washington,
195-218, 1972. D. C., 1980.
Hennington, L., D. Zrnic, and D. Burgess, Doppler spectra of a Lewellen, W. S., and M. E. Teske, Turbulent transport model of
maxi-tornado, in Preprints, 12th Conference on Severe Local low-level winds in a tornado, in Preprints, loth Con)'•'renceon
Storms, pp. 433-436, American MeteorologicalSociety, Boston, Severe Local Storms, pp. 291-298, American Meteorological
Mass., 1982. Society, Boston, Mass., 1977.
Hess, G. D., K. T. Spillane, and R. S. Lourensz, Atmospheric Lewellen, W. S., M. E. Teske, and Y. P. Sheng,Wind and pressure
vortices in shallow convection, J. Appl. Meteorol., 27, 305-317, distribution in a tornado, in Proceedings, Fifth International
1988. Conferenceot• Wind Engineering, pp. II-3-1-II-3-11, Colorado
Hopfinger, E. J., F. K. Browand, and Y. Gage, Turbulence and State University, Fort Collins, 1979.
waves in a rotating tank, J. Fluid Mech., I25, 505-534, 1982. Lilly, D. K., Tornado dynamics, Manuscript 69-117, Natl. Cent. for
Howells, P., and R. K. Smith, Numerical simulationsof tornado- Atmos. Res., Boulder, Colo., 1969.
like vortices, I, Vortex evolution, Geophys.Astrophys. Fluid Lilly, D. K., Sources of rotation and energy in a tornado, in
Dyn., 27, 253-284, 1983. Proceedings, Symposium on Tornadoes: Assessment of Know!-
Howells, P. A. C., R. Rotunno, and R. K. Smith, A comparative edge and Implications for Man, edited by R. E. Peterson, pp.
studyof atmosphericandlaboratory-analogue numericaltornado- 145-150, Texas Tech University, Lubbock, 1976.
vortexmodels,Q. J. R. Meteorol.Sot., 114,801-822,1988. Lilly, D. K., Dynamics of rotating thunderstorms, in Mesoscale
Hsu, C. T., and H. Tesfamariam,Turbulentmodelingof a tornado Meteorology•Theories, Observations, and Models, edited by
boundarylayer flow, PreprintERI-76126,Eng. Res. Inst., Iowa D. K. Lilly and T. Gal-Chen, pp. 531-544, D. Reidel, Norwell,
State Univ., Ames, 1975. Mass., 1983.
Hsu, C. T., andH. Tesfamariam,Computersimulationof a tornado- Lilly, D. K., The structure, energetics, and propagationof rotating
like vortexboundarylayerflow,PreprintERI-76231,Iowa State convective storms, 1, Energy exchange with the mean flow, J.
Univ., Ames, 1976. Atmos. Sci., 43, 113-125, 1986a.
Idso, S. B., Tornadoor dustdevil:The enigmaof desertwhirlwinds, Lilly, D. K., The structure, energetics,and propagationof rotating
Am. Sci., 62,530-541, 1974. convective storms, II, Helicity and storm stabilization, J. Atmos.
Jischke,M. C., and M. Parang,Fluid dynamicsof a tornado-like Sci., 43, 126-140,
38 Tf)RNAI)() V(}RTEX THEORY

l•ong, R. R., Sot•rccs and sinks at the axis of a rotating liquid, Q. d. Prahlad, T. S., Numerical solutions for boundary layers beneath a
Mech. Appl. Math., 9, 385-393, 1956. potential vortex, Cornput. Fluids, 4, 157-169, 1976.
l,ong, R. R., Vortex motion in a viscous fluid, J. Meteorol., 15, Proctor, F. H., The dynamic and thermal structure of an evolving
108-112, 1958. tornado, in Preprints, 11th ConJi'renceon Severe Local Storms,
Long, R. R., Tornadoes and dust whirls, ONR Set. Tech. Rep. 10, pp. 389-396, American Meteorological Society, Boston, Mass.,
Johns Hopkins Univ., Baltimore, Md., 1960. 1979.
Long, R. R., A vortex in an infinite viscousfluid, J. Fluid Mech., 11, Rankine, W. J., A Manual q['Applied Mechanics, Charles Griffin,
61 I--624, 1961. London, i 882.
Lopez, J. M., Axisymmetric vortex breakdown, 1, Confinedswirl- Rao, G. V., and W. H. Raymond, An investigation of the boundary
ing ltow, J. Fhtid Mech., 221, 533-552, 1990. layer of an atmospheric mesoscale vortex, in Preprints, Ninth
l•ugt, H. J., Vortex P'l•w in Nature and 7"er'hnolo.gy,297 pp., Cot•[•'rence on Severe Local Storms, pp. 234-237, American
Wilcy-lntcrsciencc, New York, 1983. Meteorological Society, Boston, Mass., 1975.
l,ugl, H. J., Vortex breakdown in atmosphericcolumnar vortices, Rostek, W. F., and J. T. Snow, Surface roughness effects on
Btdl. Am. Mctcorol. So('., 70, !52(•-!537, 1989. tornado-like vortices, in Preprint•s. 14th Cot•.l•'rent'eon Severe
l,tlndgrcn,•I'. S., Flow above the drain hole in a rotating vessel, Lo('al Storms, pp. 252-255, American Meteorological Society,
t,7uid Met'h., 155, 381-412, 1985. Boston, Mass., 1985.
l•yons, W. A., and S. R. t)easc,•'Sleanl devils" over l•akc Michigan Roll, N., On the viscous core of a line vortex, Z. Angew Math.
during a January Arctic outbreak, Mon. Weather Rev., 100, Me•'h., 9, 543-553, 1958.
235-237, 1972. Roll, N., Turbulent boundarylayer developmenton the end walls of
Mac(}orlnan, l). R., ()bscrvations of lightning in tornadic a vortex chamber, Rep. ATN-62 (9202)-1, Aerosp. Corp., E!
this v{}lumc. Segundo, CaliF., 1962.
Marchcnko, A. S., I)etcrminalion of maximum wind velocities in Roll, N., arid W. S. Lewellen, Examples of boundary layers in
tt•rnadocsand tr{)picat cyclones(in Russian),Metcorol. Hydrol., rotatingflows, in RecentDevelopmentsin BoundaryLayer Re-
5, i!-i6, 1961. search,A GARl)ograph 97, Advis. Group 'for Aeronaut. Res. and
Maxwt•rthy, T., Vorlicity Sotlrcc l•)r !argo scale dust devils and Dev., NATO, Brussels, 1965.
t)lhcr commenls (•n naturally occurring columnar vortices, ,I. Roll, N., and W. S. I,ewellen, Boundary layers and their interac-
Attno3. ,5't'i., 30, 17!7-1722, 1973. tions in rotating flows, Pros. Aeronat•t. Sci., 7, 11I-!44, 1966.
Maxworthy, '1•.,(;c{}physicalvt}rticcsand vortex rings,in lntrodttr'- Rotunno, R., Numerical simulation of a laboratory vortex, J.
liotl It• Vorlex [)vtlattli('.s, w.}]. l, edited by R. Granger and J. F'. Artnos. S('i., 34, 1942-1956, 1977.
WcndI, v{•n Karman Institute, Rhode Saint Genase, Belgium, Rotunno,R., A note on the stability of a cylindrical vortex sheet,J.
1986. Fluid Mer'h., 87, 761-771, 1978.
Maxworthy, 'F., E. J. Hopfinger, and 1,. t•. Rcdekopp, Wave Rotunno, R., A study in lornado-like vortex dynamics, J. Atmos.
m•ti½•nson vortex c{>rcs,J. l"htid Mct'h., 151, 141-165, 1985. Sr'i., 36, 140-155, 1979.
Mc('lcllan, •I'. M., M. E. Akridgc, and J. T. Snow, Streamlinesand Rotunno,R., Vorticity dynamicsof a convectiveswirlingboundary
vortex lines in numerically-simulatedtornado vortices, in Pre- layer, ,i. Fluid Mer'h., 97, 623--640,1980.
print.s,16th ('Ott.[•'l'•'tl•'('
Oil ,¾t'l'•'!'C
l,o•'al Storm,s,pp. 572-577, Rotunno, R., A numerical simulation of multiple vortices, in It, tense
An•crican Meteorological Society, Boston, Mass., 1990. Atmospheric'Vorlir'ex,editedby L. Bengtssonand J. Lighthill,
Mitsutter,Y., N. Mon.ji, and 1t. Ishikawa, ()n the multiple structure pp. 215-'>" Springer-Verlag,New York 1982
t•f :t!•nosphcricvortices,J. (;cophy,s'.Rc.•., 92, 14,827-14,831, Rotunno,R., An investigationof a three-dimensional
asymmetric
1987.
vortex, .I. Artnos. St'i., 41,283-298, 1984.
M{•n.
ji, N., and W. Y t•nkuan,A iab(•ratoryinvestigation
()f charac- Rotunno, R., Tornadoes and tornadogenesis,in Mesoscale Meteo-
teristics of tornado-like vortices over various rough surfaces,
A•'Itt Mctc•,'ol. Sini('a, 47, 34-42, 1989.
rology and l:orecastin/4,edited by P.S. Ray, pp. 414-436,
A•nericanMeteorologicalSociety, Boston, Mass., 1986.
Mortol:'t,l:}. R., The strengthof w)rtex and swirlingcore flows, J.
t;'hdd Mech., 38, 315-333, 1969. Rotunno,R., Supercellthunderstormmodelingand theory, this
volume.
Morton, B. R., The generationand decay of vorticity, Geophys.
Astrophys.l:'luid Dyn., 28, 277-308, 1984. Rotunno,R., and D. K. Lilly, A numericalmodel pertainingto the
Mullen, J. B., and T. Maxworthy, A laboratorymodelof dustdevil multiplevortexphenomenon,
FinalRep.NUREG/CR-1840,
U.S.
vortices,Dyn. Atmox. Oceans, 1, 181-214, 1977. Nucl. Regul. Comm., Washington,D.C., 1981.
Pauley, R. L., Laboratorymeasurements
of axial pressures
in Sarpkaya,
T., Vortexbreakdown
in swirlingconicalflows,AIAA J.,
two-celled tornado-like vortices, J. Atmos. Sci., 46, 3392-3399, 9, 1792-1799, 1971.
1989. Savino,J. M., and E.G. Keshock,Experimentalprofilesof velocity
Pauley,R. L., C. R. Church,and J. T. Snow,Measurements
of components and radial pressuredistributionsin a vortex con-
maximumsurfacepressuredeficitsin modeledatmosphericvor- tained in a short cylindricalchamber,NASA Tech. Note, TN
tices, J. Attnox. Sci., 39,368-377, 1982. D-3072, 1965.
Paull, R., and A. F. Pillow, Conically similar viscousflows, 3, Schmidt,H., and U. Schumann,Coherentstructureof the convec-
tiveboundary
Characterization of axial causes in swirling flow and the one- layerderivedfromlarge-eddy
simulations,
J. Fluid
parameter flow generated
by a uniformhalf-linesourceof kine- Mech., 200, 511-562, 1989.
matic swirl angularmomentum,J. Fluid Mech., 155, 359-379, Schwiderski,E. W., On the axisymmetricvortex flow over a flat
1985. surface,Rep.TR-2210, Nav. Weapons Lab., Washington,
D.C.,
1968.
Phillips,W. R. C., The effusingcore at the centerof a vortex
boundarylayer, Phys.Fluids,27, 2215-2220,1984. Schwiesow,R. L., Horizontalvelocitystructurein waterspouts,
J.
Phillips,W. R. C., On vortexboundarylayers,Proc. R. Soc. Appl. Meteorol.,20, 349-360, 1981.
London, Ser. A, 400, 253-261, 1985. Schwiesow,R. L., R. E. Cupp, P. C. Sinclair,and R. F. Abbey,
Phillips,W. R. C., andB. C. Khoo,The boundary
layerbeneatha Waterspout
velocitymeasurements
by airborneDopplerlidar,J.
Rankine-likevortex, Proc. R. Soc. London, Ser. A, 411, 177-192, Appl. Meteorol., 20, 341-348, 1981.
!987. Simpson,
J., B. R. Morton,M. C. McCumber,andR. S. Penc,
Polifke,W., and L. Shtilman,The dynamicsof helicaldecaying Observationsand mechanismsof GATE waterspouts,J. Atmos.
turbulence,Phys. FluidsA, 1, 2025-2033,1989. Sci., 43,753-782,
 LEWELLEN 39

Smith,D. R., Effectofboundary

conditions
onnumerically
simu- Prepri•tts, 24th Cono•.)retwetm Radar Meteor•)logy. pp. 6245,
latedtornado-like
vortices,J. Atmos.Sci.,44, 648-656.1987. American Meteorological Society, Boston, Mass., 1989.
Smith,R. C., andP. Smith,Theoretical
flowpatternof a vortexin Wakimoto, R. M., and J. W. Wilson, Non-supercelltornadoes,
theneighborhood
of a solidboundary,
Tellus,I7, 213-219,1965. Mo•. Weather Rev., ! 17, 1113-1140, 1989.
Smith,R. K., and P. Howells,Numericalsimulations
of tornado- Walko, R. L., Plausibilityof substantialdry adiabaticsubsidencein
like vortices,
II, Two-cellvortices,
Geophys.
Astrophys.
Fluid a tornado core, .l. Atmos. Sci., 45,2251-2267, t988.
Dyn., 27, 285-298, 1983. Walko, R. L., Generationof tornado-likevorticesin nonaxisymmet-
Smith,R. K., andL. M. Leslie,
Tornadogenesis,
Q.J. R. Metcorol. ric environments,in Preprit•ts, 16th Cot•rt,•ce ot• Se•'ere Local
Sot'., 104, 189-199, 1978. St(•rms pp. 583-587 American Meteorolo,,ical Society Boston,
Smith,R. K., andL. M. Leslie,A tornadogenesis
in a rotating Mass., 1990.
thunderstorm,Q. d. R. Meteorol.Sot., 105, 107-127,1979. Walko, R., and R. Gall, A two-dimensionallinear stability analysis
Snow,J. T., Oninertialinstability
asrelated
tothemultiple
vortex of the multiple vortex phenomenon, d. Atmos. Sci., 41, 345•
phenomenon,J. Atmos. Sci., 35, 1660-1671, 1978. 3471, 1984.
Snow, J. T., A review of recent advancesin tornado vortex Walko, R. L., and R. L. Gall, Some effects of momentum diffusion
dynamics, Rev. Geophys.,20, 953-964, 1982. on axisymmetric vortices, J. Aimos. Sci., 43, 2137-2148• 1986.
Snow, J. T., The tornado,Sci. Am., 250, 86-97, 1984. Wang, C. J., Exact solutions of the steady-state Navier-Stokes
Snow,J. T., Atmospheric
columnar
vortices,
Rev. Geophy•,.,
25, equations, At•nu. Re•'. Fluid Mech., 23, 159-178, 1991.
371-385, 1987. Whipple. A. B.C., Storm, vol. 4, Platter Earth, Time-LitZ Books,
Snow,J. T., andD. E. Lurid,Inertialmotions
in analytical
vortex New York, 1982.
models, J. Atmos. Sci., 46, 3605-3610, 1989. Wicker, L. W., and R. B. Wilhelmson, Numerical simulation of a
Snow,J. T., andR. L. Pauley,On thethermodynamicmethodfor tornado-like vortex in a high resolution three dimensional cloud
estimating maximumtornadowindspeeds, J. Clim.Appl.Meteo- model, in Prepri•ts, 16th Coqtk're?,ceot• Severe Local Storms, pp.
rol., 23, 1465-1468, 1984. 263-268, American Meteorological Society, Boston, Mass., 1990.
Staley,D. O., Effectof viscosityon inertialinstabilityin a tornado Wilkens, E. M., and C. J. Diamond, Effects of convection cell
vortex, J. Atmos. Sci., 42,293-297, 1985. geometry on simulated tornadoes, d. Atmos. Sci., 44, 14•147,
Staley, D. O., and R. L. Gall, Barotropicinstabilityin a tornado 1987.
vortex, J. Atmos. Sci., 36,973-981, 1979. Wilson, T., and R. Rotunno, Numerical simulation of a laminar
Staley, D. O., and R. L. Gall, Hydrodynamicinstabilityof small vortex flow, in It•tert•atiot•al C(mferet•ce ot• Compttter Methods
eddies in a tornado vortex, J. Atmos. Sci., 41,422-429, 1984. attd Experimet•tal Meast•t'emet•ts,pp. _0.•-_15 Springer-Verlag,
Steffens,J. L., The effectof vorticity-profile
shapeon the instability New York, 1982.
of a two-dimensionalvortex, J. Atmos. Sci., 45, 254-259, 1988. Wilson, T., and R. Rotunno, A numerical simulation of a laminar
Stewartson,K., The stabilityof swirlingflowsat largeReynolds end-wall vortex and boundary layer, Phys. Fhdds, 29, 39934005.
number when subjected to disturbanceswith large azimuthal 1986.
wavenumber, Phys. Fluids, 25, 1953-1957, 1982. Wu, J., Exact vortex solutionsor'the Navier-Stokes equations,Acta
Stewartson, K., and S. Leibovich, On the stability of a columnar Aerodyn. Sinlea, 1, 8•84, 1985.
vortex to disturbances with large azimuthal wavenumber: The Wylie, D. P., and C. E. Anderson, Cloud top anomalies associated
lower neutral points, J. Fhdd Mech., 178, 549, 1987. with the Binget, OK tornado 22 May 1981, in Prepfirsts, 13th
Sullivan, R. D., A two-cell vortex solution of the Navier-Stokes Co•tk,rettt'e ott Se•'ere Local Storms, pp. 15•153, American
equations, ,I. Acrosp. Sci., 26,767-768, 1959. Meteorological Society, Boston, Mass., 1983.
Sychev, V. V., Viscous interaction of an unsteady vortex with a Yih, C.-S., F. Wu, A. K. Garg, and S. Le[bovich, Conical vortices'
rigid surface, Fluid Dyn., Engl. Transl., 24,548-559, 1989. A class of exact solutions of the Naylet-Stokes equations, Plays.
Sykes, R. I., and D. S. Henn, Large eddy simulationof turbulent Fluids, 25, 2147-2158, 1982.
sheared convection, d. Atmos. Sci., 46, 1106-1118, 1989. Zrnic, D. S., D. W. Burgess, and L. Hennington, Doppler spectra
Uttal, T., B. E. Mariner, B. W. Orr, and R. M. Wakimoto, High and estimated windspeed or' a violent tornado, J. Clim. AppI.
resolution dual-Doppler radar measurementsof a tornado, in Mete•rol., 24, 1068-1081,
NumericalSimulation
of Axisymmetric
Tornadogenesis
in ForcedConvection
BRIAN H. FIEDLER

Schoolof Meteorology,Universityof Oklahoma,Norman, Oklahoma73019

1. THE ISSUE OF TORNADO INTENSITY the present model avoids the complications of modeling
storm evolution because the buoyancy field is held fixed in
Intense tornadoesrequire a pressuredeficit in the core each computation. The model therefore does not contain
greater than that which can be provided by latent heat provisionsfor studyingthe processof subsidencewarming in
release alone [Snow and Pauley, 1984]; in other words, the core of tornadoes. However, the results indicate that
tornadowindspeedsexceedthethermodynamic speedlimit. subsidencewarming is not a necessary processfor intense
This paperreportson numericalstudiesof a simpleaxisym- tornadogenesis.
metricconvectionmodelwith a well-definedthermodynamic The experimental procedure will be to compare three
speed limit. A fixed field of buoyancylocalizedabout the integrations, one with no rotation, one with rotation and a
axis drives the convection. The results show that in the free-slip lower boundary, and one with rotation and a no-slip
strongaxial flow of an end-wallvortex,the pressuredeficitcan lower boundary. Although the effect of the slip condition at
be 7 times that providedby the buoyancy,and the thermody- the lower boundary has been investigated before, the closed-
namic speed limit can be exceededby a factor of 2. This domain model used here allows for a more credible interpre-
dynamical enhancementof the pressuredeficit and wind field tation of the results.
could accountfor the intensityof somenaturaltornadoes.
Howells et al. [1988] review numericalmodelingof torna- 2. THE MODEL
does within an open domain with imposed upstream and
downstream conditions. However, neither the open-domain The incompressible, axisymmetric Navier-Stokes equa-
numerical models nor the open-domain laboratory experi- tions are applied in a cylinder rotating at constant rate D:
ments driven by a fan are ideal for addressingquestions of
tornado intensity. Specifically, in the numerical simulations
a certain environment outside the domain is implied by the
du v 2Dr=
dt r at'
+ v - -- •-
r Or
(1)
solution. The question of whether or not the core pressure
deficit at the outflow boundary is consistentwith a realistic dv uv O 1 Orv O-v
convective environment is left unanswered. •+•+2•u= v - + (2)
Recently, there have been several attempts to model
tornadolike vortices in a closed domain. Neitzel [1988]
calculated numerically the flow and concomitant vortex ..... • b + v r • + (3)
breakdowngeneratedby rotationof one of the end walls of
a closed circular cylinder. Walko [1988] modeled torna- 1 Oru Ow
dogenesisin axisymmetric storms with prognosticatmo- - • + • = 0. (4)
r Or Oz
sphericthermodynamicsand thus was able to addressthe
relationshipbetweenthe tornadointensityandthe environ- Here p is the pressure fluctuation from that of the motion-
ment. Solutions were found with hurricanelike tornadoes less,equilibriumstate divided by the constantdensity,b is a
containingcloud-freecoresin whichthesubsidencewarming fixed external force (or buoyancy), and v is the inverse
was augmenting the buoyancyin the core.The modelused Reynolds number. The equations have been made dimen-
hereis alsoa closedsystem,but,unlikethatof Walko[1988], sionlessusinga convective velocity scaleand the depth of
the cylinder. The dimensionlessdepth is 1 and the radiusis
The Tornado: Its Structure,Dynamics,Prediction,and Hazards. r0. In all computationsto be discussedhere
Geophysical Monograph 79
Copyright1993by the AmericanGeophysicalUnion. b(r,z) = 1.264e
-2ø[r2
+(z- o.5)'-1,
 r
42 NUMERICAL SIMULATION OF AXISYMMETRIC TORNADOGENESIS

(0,1) which is plotted in Figure 1. The dimensionless convective

available potential energy is

E b(O, z) dz = 0.5. (6)

The top and radial boundaries are always no-slip. The lower
boundary can be either no-slip or free-slip.
In stretched coordinates R(r) and Z(z), the stream func-
tion vorticity formulation is

ov + R'Z' 00/ orv _ R'Z' 00/ o'v = V(R' ~ R' ov

at ,2 OZ oR r 2 oR oZ oR oR
(0,0) (1,0)
R' av v a ov)
+----+Z' - Z ' - (7)
Fig. I. Contours of the time-independent buoyancy field. View is , aR,2 oZ oZ'
o < r < I, 0 < z < J, as indicated,

2.5

2.0

/1 w
1.5
!;
!,j no-slip

'-'/-"..~.~~~.~'-" ,.,'
10
.'
·-';1
II

05 ~N~----.~"
Fig. 2. The impulsive solution. Acceleration vectors (maximum
0.814) and contours of p. View is 0 < r < I, 0 < z < I.
L:··· ··~· .· :
- -- -- - --- --
,
0.0 o. 10. 20. 30. 40. 50. 60.

2.5

2.0

I.S
free·slip
111ltl l \ \

1.0

r··\~/\ ......... --
. ,:W
0.5: / "'\
.,,----
" u -'.

Fig. 3. Steady solution without rotation. Velocity vectors and

contours ofp; U max = 0.689, W IllaX = 0.849, andp""", = 0.408. View Fig. 4; Time history of "max, v"""" and W max in the free-slip and
is 0 < r < I, 0 < z < I. no-slip integrations with n = 0,2 and l' = 0.0005.
 FIEDLER 43

Fig. 5. Free-slip integrationwith x• = 0.2 and v= 0.0005at t = 10.5' Urnax= 0.395, t,max= 0.815, "'max= 0.495, and
Pmax= -0.975. View is 0 < r < 1, 0 <: < 1.

(12)
•+R'Z' -R' •• - v'
Ot aZ OR OROZ r OZ R'=dR/dr, (13)

+ a Or/
R'ar/
r/ aZ'
and
•R'
OR OR +r OR r2+Z'--OZ Z' =- dZ/dz. (14)
Ob
(8) The following coordinate transformations are used in all the
computations'
and XR + (2- A)R3
r = r0 , (15)
I+R-
0• R' a½
z' a•
0 0
rr/ = R' R' + Z' (9)
OR r OR oz

where
tanh
(yZ-0.Sy) (16)

n ---(ou/az) - taw/or), (1o)

where X and 7 are constants.
. = (]/r)(a½/az), (11) The model is integrated using leapfrog time
44 NUMERICAL SIMULATION OF AXISYMMETRIC T()RNADOGENESIS

...2
1
Fig. 6. Free-slipintegration
with1• = 0.2 and v = 0.0005at t = 40.0' Umax'=0.395,Vm;,
x = 0.682,Wma
x = 0.268,and
Pmax = -0.612. View is 0 < r < 1, 0 < z < 1.

and an Arakawa Jacobianon a grid with constantspacingAR in which the averagingwith the previoustime step was suffi-
and AZ. The time step is maintainedsmall enoughto insure cient to maintain stability. Also, in the advection terms in (8),
both stabilityand sufficientaccuracy.Exceptfor the boundary
rl(R, Z, t) R' (O)
conditions, the additional body force b, and some additional lira = [4r/(AR, Z, t)- r/(2AR, Z, t)].
minor exceptions, the model equations and the numerical •-•0 r 2AR
schemeare those of Wilsonand Rotunno [ 1986].The additional (18)
minor differences are as follows. The model uses Dufort-
Frankel time differencingfor the diffusion terms. When the 3. MODEL SOLUTIONS
no-slipconditionis applied,at the lower boundary,for exam-
3.1. No Rotation
ple, the following modification of Pearson [1965] is used:
A linear, impulsive, solution with fI = 0 and v = 0 (Figure

n(R,
O,
t+zXt)=
0.5{rt(R,
0,t) 2) is obtained by first solving

1 O Op 02p Ob
z'(o) ---r•+ , =• (19)
r Or Or az- Oz
2r(Ag)
2[-7½,(R,
0,t)+ 8½,(R,
AZ,t)
forp and then solvingfor the accelerationsin (1) and (3). The
- ½(R, 2AZ, t] (17) solutionrepresentsthe responseof the fluid at the instant
 FIEDLER 45

is"turned
on"ina motionless
state.Thevertical
symmetrypressure gradient. The solutions at later times, of which t =
andthe weakpressure
fluctuation
of theimpulsive
solution 40 is an example,are more similarto the steady,hydrostatic,
will be in contrastwith the nonlinearsolutions
shownbelow. inviscid solutions(or balanced solutions) obtained from
A nearly steady state, nonrotatingsolutionwith v =
0.0005,r0 = 2, anda no-slip
lowerboundaryisachievedby Op/Or= (z,2/r) + 2D,z, (22)
t = 10.0 (Figure3). The pressure
in thenonlinear
integra- and
tionsis diagnosed by solvinga Poissonequationderived
from the primitive equations,with an additionalconstraint Op/O"= b. {23)
that p(r0, 0.5) = 0. In steady,inviscidflow,
Figure 7 shows a balanced solution for Ft = 0.2 with
w20, z)
2 =p(O,
O)-
p(O,
z)+••b(O,
z')dz'(20)
boundaryconditions•, = 0 and p = 0 at :, = l. Note that the
pressure difference along the axis is 0.500. In contrast, in
Figure 5 the pressure difference between the stagnation
andthepressure difference
betweenthetwoaxialstagnation points is 0.82, with both inertial and viscous forces substan-
points would be tially contributing to the pressure drop. In Figure 6 the
pressure difference between the stagnation points is 0.57,
p(0, 1) - p(O, O) = b(O, z) dz = 0.5. (21) and the weak flow in the core confirms that it is nearly
hydrostatic. Further integration shows that the free-slip case

In the numericalintegrationthe differenceis 0.433, being

slightly less as a result of the viscous terms. The maximum
value of w in the domainis 14,
max: 0.849.The highpressure
of the upper stagnationpoint extendssignificantlydown into
the buoyancyfield and preventsw from achievingthe upper
limit of 1.0 allowed by the buoyancyterm alone. This value
of 1.0 will be referred to as the thermodynamicspeedlimit of
the model. Note that in the formal inviscid limit, the kinetic
energy in the domain could increase without bound and the
thermodynamic speed limit would not be a bound on the v
wind speed. It turns out that the value of viscosity used in
these calculations is enoughto dissipatemost of the kinetic
energy of the fluid before it recycles into the storm so that
the thermodynamic speed limit is especially relevant. Fur-
thermore, we shall see in the rotating solutions that the
kinetic energy productionterm wb is diminishedbecausew
is retarded or even negative near the axis.
Perhapsthe most appropriateanalogto isolatedthunder-
storms would be the case with r0 --->c•. When the above
experiment is repeated with r 0 = 1, the result is Wrnax=
0.851. With both r0 = 1.5 and r 0 = 2.0, the result is Wmax
= 0.849. We will use r 0 = 2 in the remainingintegrationsas
an approximationfor an infinite domain.

3.2. Rotation and a Free-Slip

Lower Boundary
p
In the above steady, nonrotating solution the largest
magnitudes
in velocityandpressureare skewedtowardthe
top of the domain,withoutmuch"severeweather"nearthe
surface.The oppositeis truefor therotatingsolutions.
Time
histories of the maximum wind speedsin the domain in
integrations with fl = 0.2, •, = 0.0005,andfree-slipand
no-sliplowerboundary conditions areshownin Figure4. A
peakin Z•max occursin bothintegrations at t = 10.5.
Thefree-slipsolution isshownat t = 10.5in Figure5 and
at t = 40 in Figure6. At t -- 10.5 the vortexis narrower,
permitting a largeverticalvelocityto occuroutside thecore Fig. 7. Balancedsolutionwith 171= 0.2; Vmax= 0.561andPmax=
wherethe buoyancyis relativelyunimpeded by a vertical -0.500. View is0 < r < 1,0 <z <
46 NtlMERICAI. SIMtJ[.A'I'I(•N ()F AXISYMMETRIC TORNADOGENESIS

Fig. 8. No-slip integrationwith (• ::':0.2 and v = ().()005at t = 10.5' urn,x = -0.568, 'vma
x = !.261, w,nax = 2.042,
and.rJma
x ==-3.617. Close-up,cornerview of rncridionalvelocityis 0 < r < 0.3, 0 < ', < 0.3' otherwise,view is 0
< r < 1, 0 < z < 1.

does not come to steady state, but rather goesthrough cycles the no-slip boundary case [Rotunno, 1986; Howells et al.,
with a period of about 140 in which wind speedsremain less 1988]. Although it has some qualitative similarity with the
than or equal to those shown here after t = 30. Although the solutionat t = 10.5, the corner region of the solutionat t = 40
core pressure deficit, particularly in the initial transient re- would be better described as a case of "overshoot" rather than
sponse,can be greater than that provided by buoyancy, the as a genuinesupercriticalend-wallvortex. As such,the over-
wind speed remains less than the thermodynamicspeedlimit. shoot solutionsprovide only a modestamplificationof the
tYee-slip
intensity,a resultthatwasalsofoundby Howellset al.
3.3. Rotation and a No-Slip [1988].The dynamicsin the solutionat t = 10.5 is something
Lower Boundary fundamentallydifferentfrom overshoot.The collisionof the
When the no-slip boundary condition replacesthe free-slip end-wallboundarylayerresultsin an axialjet that •1owsfor a
boundary condition, significantly larger peak maximum wind corepressure deficitthatis morethan7 timesgreaterthanthat
speedsresult. The peak maximum wind speedsat t = 10.5 in the balancedvortex and more than 5 times greater than that
are associated with an end-wall vortex (Figure 8), the in the overshootsolution.Althoughthistransientsupercritical
dynamics of which are discussed by Wilson and Rotunno vortex exists near its maximum intensity for only a few
[1986] and Fiedter and Rotunno [1986]. The vortex at t = 40 dimensionless time units,thiscorrespondsto about5 min when
(Figure 9) is similar to many previous publishedsolutionsfor scaledto a thunderstorm,and air cyclesthroughthe core
 FIEDLER 47

....
...... • I? I l
........

......... ,•I,tl ; t
........... /,'I// • /
........ ,II/'/' I • r meridional
velocity
......... /I//I'/ /' ! I
......... /I,,'1I/ ! t • .... (close-upview)
......./f//,! ' , i ( ...... ' .
..... //il ,
,,, •//'!,".!,'f I t 1 •, • ', ,, -, ... - - -

•?.li•1•,\,,,,,,.............,
' - _ _ _
//// / / / / i II .......................

Fig. 9. No-slipintegration
with•1 = 0.2 and v = 0.0005at t = 40.0; //max= -0.402, ;',nax= 0.759, wma
x = 0.488,
and Pmax= -0.696. Views as in Figure 8.

20 times duringthis phase.Perhapswith otherparametersthe 4. IMPLICATIONS FOR TORNADOES

vortex breakdown could remain in steady state suspension
away from the lower boundary, which could allow for a more Table I showsthe results for the peak wind speedsin the
impressive,long-lastingsupercriticalvortex.Nevertheless,the free-slip and no-slip experiments and in the three control
dynamics proposedby Fiedler and Rotttnno [1986] to be a experiments. The results are in accord with those of Fiedler
possiblecauseof intensetornadoeshas been captured,albeit and Rotunno [1986]. They deduced from laboratory experi-
only momentarily. ments and theoretical analysis that •'maxin supercritical

TABLE 1. The Peak Maximum Wind Speeds and Pressure Fluctuations

Experiment t ttmax •' max w max P max

Impulse(v = 0, fI = 0, free-slip) 0 0 0 0 0.118

Balanced (v = 0, fI = 0.2, free-slip) steady 0 0.561 0 -0.500
No rotation (v = 0.0005, no-slip) steady 0.689 0 0.849 0.408
Free-slip(v = 0.0005,12= 0.2) 10.5 0.229 0.815 0.495 -0.975
No-slip(v = 0.0005,fI = 0.2) 10.5 -0.568 1.261 2.042
48 NUMERICAl. SIMULATION OF AXISYMMETRIC TORNADOGENESIS

tornadoes could be 1.7 times that in the downstream subcrit- Acknowledgments. This work was supported by the National
ical vortex. Although an exact height of transition from Severe Storms Laboratory and the University of Oklahoma.
supercritical to subcritical flow is difficult to identify in
Figure 8, this ratio can be estimated to be 1.261/0.7 •- 1.7. REFERENCES
They also imply that the dynamics past the breakdown point
Fiediet, B. H., and R. Rotunno, A theory for the maximum
would be essentially that of the free-slip solution, which is windspeedsin tornado-like vortices, J. Atmos. Sci., 43, 2328-
nearly the case here. The ratio of Vmaxin the peak no-slip 2340, 1986.
solution to Vma x in the peak free-slip solution is 1.55. Howells, P. A. C., R. Rotunno, and R. K. Smith, A comparative
Another prediction was that wmax •' 2Vma x in the most studyof atmosphericand laboratory-analoguenumerical tornado-
intense tornadoes' in the most intense vortex found in these vortex models, Q. J. R. Meteorol. Sot'., 114, 801-822, 1988.
Neitzel, G. P., Streak-line motion during steady and unsteady
simulations, •',n:•x = 1.62Vmax- axisymmetricvortex breakdown,Phys. Fluids, 31,958-960, 1988.
These results add credibility to the hypotheses put for- Pearson, C. E., A computationalmethod for viscous flow problems,
ward by Fiediet and Rotunno [ 1986]that someinstancesof ,I. Fluid Mech., 21, 611-4•22, 1965.
the most intense tornadoes could be supercritical end-wall Rotunno, P,., Tornadoes and tornadogenesis, in Mesos•'ale Meteo-
rology and Forecasting, edited by P.S. Ray, pp. 414-436,
vortices with axial jets in the core. Fiediet and Rotunno American Meteorological Society, Boston, Mass., 1986.
[19861 used 3800 m" s ' as an approximate upper limit to Snow, J. T., and R. 1,. Pauley, On the thermodynamic method for
convective available potential energy in the atmospbc're, estimatingtnaximum tornado windspeeds,J. Clim. Appl. Meteo-
which implies a thermodynamic speed limit of 87.2 m 5• • ro!., 23, 1465-1468, 1984.
Waiko, R. 1,., Plausibility ()f substantialdry adiabatic subsidencein
Using this speed limit to dimensionalizethe no-slip result a tornado corc, J. Atmos. Sci., 45, 2251-2267, 1988.
impliesthat the most intensetornado could have ureaX = 5(t Wilson, T., and R. Rotunno, Numerical simulation of a !aminar
ms I :: I10 rn s • : 178 m s • and a core end-wall vortex and boundary layer, Phys. t;'hdds,29, 3993-4005,
pressure deficit of 330 mbar.
 NumericalSimulationof TornadolikeVorticesin AsymmetricFlow
R. JEFFREY TRAPP AND BRIAN H. FIEDLER

Schoolof Meteorology,Universityof Oklahoma,Norman, Oklahoma 73019

1. INTRODUCTION 2. EXPERIMENTAL PROCEDURE

Most current knowledge of the dynamicsof tornadolike

2.1. Physical Model
vortices has been obtained from axisymmetricmodelswith
initial vertical vorticity. One questionwhich continuesto The physicalmodel is meant to be representativeof both
elude researchersis how tornadoes,which appear at least the region beneath a thunderstormand the DS experiments
locally axisymmetric, are born out of nonaxisymmetric (Figure 1). A fixed shear flow (Figure 2) enters the cubic
ambient flow with initial horizontalvorticity only. domain of unit dimension through one wall. The vorticity
A three-dimensional numerical model will be used to associated with this shear is horizontal and transverse to the
investigate this question. The work of De Siervi et al. [1982] inflow velocity.
(hereinafter referred to as DS) on the existence of the inlet Fluid is extracted from the cube through an exhaustregion
(or ground) vortices associatedwith jet enginesservesas the along the top boundary. To establishthe viability of this
template for our numerical experiments.DS showedthat an procedureand for reasonsof computationalefficiency,we
inlet vortex may form as a loop of vorticity transverseto the model only half of a symmetrical domain and place a
advecting flow, is drawn into a horizontally oriented inlet semicircularexhaust region at the edge of the top boundary,
facing the flow, and then is stretchedand deformed. This adjacentto a vertical symmetryplane. The lifting of loopsof
horizontal vorticity yields two symmetrical vortices, only
stretching and deformation processcould result in a pair of
one of which appears within the modeled domain.
counterrotatingvortices, but only a singlevortex is usually
visualized. Indeed, as illustrated by DS, the upper vortex
legsbecomespreadout and diffusein comparisonwith the 2.2. Numerical Model
lower legs which are concentratedabout a line and which
extend downwardtoward a point on the groundbeneath;the The Navier-Stokes equations •k)r compressible flow of
latter is the inlet vortex. The "spin-up" of this vortex might constant viscosity are numerically integrated, using an ex-
be explainedby the secondaryflow attendantwith rotation plicit schemewith fourth-orderaccurateadvectionterms.At
just above a viscous boundary. Becauseonly one vortex the initial time the dependent variables consistingof pertur-
appearseventhoughthe net circulation aroundthe inletis bation pressure (p) and the three momentum components
zero, andsincethe vortexis producedthroughthetiltingand (pu, pv, pw) all have zero value.
stretchingof vortexlines,we believethatthe formationof Boundary conditions are shown in Figure 1. The lower
boundary(z = 0) is either free-slip(w = 0' Ou/Oz= Ov/Oz
inlet vorticesmay be similarto many,but not all, instances
= 0) or no-slip(u = v = ,•, = 0). At the inflow boundary(x
of tornadogenesis.
= 1), v = w = 0, and u = -U(z), where
Here we presentresultsof thefirststageof thistornado
modeling
approach. Our discussionwill focusontheevolu- U( z) = z/0.5 0 <- z < 0.5
tion and structure of a vertical vortex and its image, which
are generatedin laminar,nonaxisymmetric
flow with no U(z) = 1 elsewhere
initial verticalvorticity. The modelemployedrepresentsan
Free-slip conditionsfor u and v are applied at the top
idealized version of the DS experiments.
boundary,while the exhaustconditionon w is specifiedby

The Tornado:Its Structure,Dynamics,Prediction,

andHazards. w = pg(x, y)
GeophysicalMonograph79
Copyright1993by'the AmericanGeophysical Union.
49
50 NUM[';RICAI, SIMUI,A'I'ION ()F T()RNADOI•IKE VOIYFICES

fme-qip.
symmetry convergence and a pressure rise in the domain which, when
/anti- 0w/3x=0) felt at the top boundary, requires fluid to be evacuated. As

outflow:symletry the tluid is then releasedvia the exhaustregion,convergence

w=p
g(x.y)
.......
Ou/i3z:;)v/Oz=O free-
slip. and pressure are reduced, and the volume flow rates of the
•• sy•met•
(0u/0y=v: inflow and exhaust tend toward equality.
[ • /anm 0w/Oy=0)
free-slip.
(3u/0y=v= ]]
symrnet• • •mmet• A finite differencegrid consistingof 30 x 30 x 30 pointsis

Ow/ay=0)
/ann-
•[
symmet. I • inflow:
used in all the cases. For experiments involving higher
Reynoldsnumber flow (Re • 500), the resolution is locally
enhanced near the symmetry wall. Such enhancement is
accomplished by the Rfilowing coordinate transformation in
the y direction:
L•no-•.hp Y= v exp (sy-
3u/0z=3v/)z
where Y replaces y as the new nondimensional coordinate in
[;ig. 1. Boundary condition.sand schematic of physical domain. •hat direction, and s is the stretching constant, assigneda
'v'•]•c of 0.75.
•'hc modeled flow has a Mach number of approximately
t;.':;. The Math number is defined as

M :-: 1!max/½
'.
andx0 :• 0.5, Yo.......
0, and;.....
•'. (0.05)•/2 ()n theremaining where r' is the sound speed, and u maxis the magnitude of the
bt)undarics(x = 0; ) .........
1) and .•.... 1), fi'ec-slip, symmetry/ maximum velocity found in the domain. Also, the Reynolds
antisymmetry conditions tk•ru, -•,, and •' are applied. Hence number hcrc is
at the symmetry plane () ..... 0), tk>rexample, -•, = [)lt/;)y :'•
Ow/ay "= (}, implying that an image solution lies across this Re = VL/p,
boundary (as well as across the other t¾cc-slipboundaries). where V is the refizrence velocity of unit value (the maximum
A t•w aspects of the top boundary condition deserve
inflow velocity), L is the unit length scale, and •, is the
further explanation. First, the exhaust region is positioned
viscosity of air. For scaling purposes a dimensional velocity
midway between the x .... (} a.nd the x I boundaries to
of 1{)ms • a lengtho['l km anda densityof I kgm 3 have
minimize any interiOronce of the boundaries which could be been assumed.
deleterious to the development of the v{>rtcx. Also, the
c-lk)!ding
radiusof r :'=(0.05)•/2 is the smallestpossible
3.
value that afik•rdsnumerical stability yet still provides tbr a
considerable exhaust velocity. The exhaust is driven through In experiment N5(){) we assign Re = 500 and choose a
this condition, then, as tbllows. The entering ll(•w causes no-slip lower boundary. A time series of maximum values
for this integration is given in Figure 3. Here "low-level" is
used to denote those levels below z = 0.5. The evolution of
1.0 .... --• the flow changes significantly at t = 8. Notice the increase
in the maximum low-level convergence out to this time.
initially, fluid moves in the negative x direction toward the x
--- 0 boundary. Fluid which impacts this boundary, or
"stagnation wall," is forced to turn parallel to the wall. A
downward moving branch of the flow spreadshorizontally as
it meets the lower surface and then converges with the
0.• inflow. The zone of maximum convergence propagatesin the
positivex direction out to x ----0.75, then stops. It shouldbe
noted that such a convergencezone is similar to a thunder-
storm outflow boundary, a feature believed by some to be
important, if not essential,for tornadogenesisin the atmo-
sphere.
Referring back to Figure 3, now consider the rise in
0.0 1.0 maximum low-level vertical vorticity, •rt, prior to t = 8.
During this time, horizontal vorticity present in the inflow
U(z) and generatedby pressuregradientsat the no-slipboundary
is tilted into the vertical. Apparently, the horizontal gradi-
Fig. 2. Inflow velocity profile. ents in the vertical velocity, w, responsiblefor such
 TRAPP AND FIEDLER 51

2OO
- 4O
Re=500, areduebothto the risingmotionin the convergence
zone
no-slip andto theexhaust.
Despitetheincreasing valuesof (t, a
well-defined
circulatory
patternin the horizontal
velocity
fieldis not evidentprior to t = 8.

vorticity
30

Rapid changesoccur in the flow after t = 8. Low-level

cycloniccirculation
appearsat t = 9 in an areapreviously
100 - 20
occupied by large horizontal shear. Coincident with this
developing
vortexis an updraftregionof nearlythe same
size and shape, which extends to the exhaust above. The
vortex becomeswell-definedby t --- 10, as can be seen in
Figure4. Localaxisymmetry
isclearlyshown,butalsonote
an intriguingvertical velocity field. Outside of a vertical
velocityminimaare two maxima.Comparison of the veloc-
0 o ity fieldin Figure4 withthethree-dimensional
vorticityfield
o 2o at the sametime and level (Figure 5) showsthat there is a
time
highdegreeof upwardverticaltiltingof horizontal vorticity
wherethe horizontalgradientof w is largeand positive.
Fig. 3. Experiment of low-level Moreover,verticalstretching
N500:timeseriesof peakvalues of tiltedvorticityis occurring
verticalvorticity,•rl, andlow-level
horizontalconvergence.The in positiveverticalgradients
of w. Notefurther,in Figure5,
values are dimensionless.
the convergenceof horizontalvortexlinesinto a regionof

I I I I I ' 1 '1 -

x
Fig. 4. ExperimentN500: horizontalvelocityvectorsandcontoursof verticalvelocityat time t = 10 and level z =
0.03. The magnitudeof the maximumvectoris 1.8, andthe verticalvelocitycontourintervalis
52 NUMERICAl, SIMUI•ATI()N OF TORNADOI,IKE VORTICES

Fig. 5. Experiment N5()0: horizontal vorticity vectorsand contoursot'vertical vorticity at time t = 10 and level z =
0.03. The magnitudeot' the maximum vector is 30.1, and the vertical w)rticity contour interval is 4.0.

maximum vertical vorticity. This reflects well the three- some tornadolike vortices that form in axisymmetric numer-
dimensionalnondivergenceof vorticity just above a no-slip ical models for certain values of swirl ratio [Davies-Jones,
boundary. 1973], radially converging flow in the viscous boundary layer
Horizontal sections of the velocity field taken at t -- 10 erupts to produce an axial jet in the core of the vortex
(Figure 6) show the three-dimensional structure. A slanted [Howells et al., 1988; Wilsonand Rotunno, 1986]. Here Wma x
columnar vortex is quite evident. Over time the vortex • Vmax/3,where U,maxis the maximum vertical velocity
straightens as its lower and mid sections move toward the within the core of the low-level vortex, and Vma x is the
symmetry plane. The low-level vortex reaches its peak effective maximum azimuthal velocity. In intense single-
intensity (•rz= 152 units at t -- 12) while very near the wall celledtornadoes,Wma x •- 2Vrnaxis anticipatedby Fiedler and
(vortex center at x --- 0.75, y --- 0.1). Within the next 1-2 Rotanno [1986]. For such an axial jet to occur in our model,
time units, however, srzdecreasesdramatically (see Figure we believe that the Reynolds number of the flow must be
3), and by t -- 13 the vortex can no longer be visualized. One increased by 1-2 orders of magnitude (which will in turn
reason for such decay could be cross diffusion with the decreasethe depth of the boundary layer), that the exhaust
image vortex across the symmetry plane. radius must be decreased, and that the vortex and boundary
In N500 the maximum low-level convergence rises after layer must be resolved better.
the vortex forms because of radial inflow, beneath z = 0.1, With the preceding discussionin mind, it is now useful to
toward the vortex core. An imbalance between the pressure introduce results of experiments involving lower Reynolds
gradient and centrifugal forces within this viscous boundary numberflow. Though not physically significant,preliminary
layer results in a net inward force which drives the inflow. In caseswith Re < 100 were performed.A vortex did not
 TRAPP AND FIEDLER 53

Fig. 6. Experiment N500: three-dimensional velocity vectors at I = 10. The levels are z= 0.05,0.35,0.65, and
O.~5.

in this flow. However. small amounts of vertical vorticity initially inclined toward the symmetry plane, with the lowest
were located near the top boundary. An experiment with Re position of the vortex near the center of the domain. The
= 200 and a no-slip lower boundary (N200) did yield a vortex did straighten, however, as its lower and mid sections
vortex which in fact was similar to thaI of N500 in terms of moved toward the symmetry wall. In this position and
its local axisymmetry and vertical structure. The evolution location the vortex suffered a rapid drop in intensity, appar-
of N200 progressed very rapidly, with the peak 1:/ of 25 units ently through cross diffusion with the image vortex.
occurring at t - 7. Also, the maximum wind speed of about During the formation of the vortex the low-level conver-
I unit associated with that peak was comparatively weak: in gence increased because of radial inflow into the core. A net
N500 the horizontal velocity at Z = 0.03 was in excess of 3 inward force owing to an imbalance between the pressure
units at the time of the peak 1:/ (of 152 units). gradient and the centrifugal forces within the viscous bound-
ary layer was responsible for this inflow. The converging
4. SUMMARV flow was not strong enough, however, to produce the axial
OUf model results show that a vertical columnar vortex jet seen in some end-wall vortices. The trend shown in our
and its image can arise readily from the low-level horizontal experimentation with the Reynolds number of the flow was
vorticity present in the fixed shear inflow and generated by that of an increase in such converging flow with increasing
pressure gradients at the no-slip boundary. The horizontal He and hence with decreasing boundary layer depth.
vortex lines were tilted into the vertical by sharp gradients The continuation of this research will first involve the
within the updraft at low levels, organized into a locally replacement of the stagnation wall, even though it aids the
axisymmetric state, and stretched. The resultant vortex was formation of the gust front feature, with an open boundary.
54 NLJMERICA[, SIMU[,ATION OF T()RNADOLIKE VORTICES

In hindsight it has been recognized thttt the wall lacks a Acknowledgmettts. The authors wish to thank Robert Davies-
physical analog within the region beneath a thunderstorm (as Jones and the two anonymous reviewers for many helpful comments
well its in the DS model). Future experiments will decide and suggestions.This researchwas fundingin part by the Center for
the Analysis and Prediction of Storms and by the National Science
whether a gusltfront needs to be explicitly modeled. Next, Foundation, grant ATM-9002391.
the symmetry plane will be removed so that, instead of
studying a vertical vortex pair, we may investigate why one
sign ot' vertical vorticity is favored and how it results from REFERENCES
horizontal vorticity. Many studiesconcerningthe origin of
storm rotation [e.g., Davies-Jorte,¾, 1984]have discussedthe Davies-Jones, R. P., The dependence of core radius on swirl ratio in
importanceof storm-relative environmentalwinds, which a tornado simulator, J. Atmos. Sci., 30, 1427-1430, 1973.
I)avies-Jones, R. P., Streamwise vorticity: The origin oI' updraft
vccr with height, and o[ the associated streamwisecompo-
rotation in supercell storms,.I. A tmos. Sci., 41,299 !-3006, 1984.
nent o[ vorticity in the low-levelinflow. This perhapsimplies l)e Siervi, F., tl. C. Viguicr, E. M. Greitzer, and C. S. Tan,
the need here t'or such a variation in the inflow winds. Note, Mechanisms of inlet-vortex l'ormation, J. Fhdd Mech., 124,
however, that the vorticity associated with the upstream 173-21)7, 1982.
flow in thc DS experimentswas entirely crosswi,;c,l-,utthe Fietiler, B. tt., and R. Rotunno, A theory tk•r the maximum
symmetrybreakingmechanismleadingto the t'c)rn'tal•'•o1'a windspeeds in tornado-like vortices, J. Atmox. Sci., 43, 2328-
2340, 1986.
single intensevortex is unclear. Finally, subsequcn::
1towclls, P. A. C., R. Rotunno, and R. K. Smith, A comparative
will involve higher Reynolds number flow on a finer grid.
,dudy of atmosphericand laboratory-analoguenumerical tornado-
This will be accomplishedefficiently through a technique vortex models, Q. J. R. Mctcorol. Sot., !14, 801-822, 1988.
known as dynamic grid adaption in which grid points con- Wilson, T., and R. Rotunno, Numerical simulation of a !aminar
tinually Ilow l'romareas o[ little interest to, in this case,the end-wall vorlcx and boundary layer, Phys. Fluid.Y,29, 3993-4005,
area surrounding the vortex.
 Discussion

Lou WICKER,SESSION CHAIR

University qfilllinois

PAPER A 1 accelerationcertainly is as important as the radial accelera-

Presenter, Steve Lewellen, Drexel University [Lewellen, tion. Turbulence is one of the uncertainties that still remain
this volume, Tornado vortex theory] to be resolved as a goal of research. So I'm not quite sure
where your differences with the model really come in.
(Joe Golden, National Oceanicand AtmosphericAdminis-
tration, Officeof the Chief Scientist)Did you givea number (Fiediet) I'll answer my question during my talk [paper A4].
of 125m s'-• forwhatyoubelieve
themodels
arepresently
predicting for probable wind speedsin tornadoes? PAPER A4

(Lewellen) That is the number I gave. I can't say that I got Presenter, Brian Fiedler, University of Oklahoma [Fiediet,
it from all of the modelers. this volume, Numerical simulation of axisymmetric torna-
dogenesisin forced convection]
(Golden) I will say somethingabout that in my paper [paper
FI]. Second question: That intriguing three-dimensional (Bob Davies-Jones, National Severe Storms Laboratory)
conceptual model of the flow in a tornado indicated subsid- How appropriate is the use of a no-slip boundary condition
ence in the core all the way to the surface. Could you in a numerical model with a turbulent boundary layer? It
comment on the possibility that (a) the axial flow may seemsthat there should be partial slip at the ground.
depend upon the life cycle of the tornado and (b) there may (Fiediet) The real atmosphere does not have partial slip; it
be one or more axial stagnationpoints. has no-slip and varying viscosity.
(Lewellen) The core flow is very unsteady. There are large (Davies-Jones) But it has no-slip only in the lowest millime-
variations from minute to minute and stagnation points at ters.
different points along the axis. That was why I showedthe
(Fiedler) You are throwing out the baby with the bath water
detailed resultsfi'om the large-eddysimulation.The features
don't remain fixed; they changeon a very fast time scale.if you use a surface layer parameterization and think that the
atmosphereworks in this way. There is an important mass
(Brian Fiedler, University of Oklahoma)Your 3-D experi- flux in the boundary layer that is not represented by a
ments, althoughvery impressive,seemto be overly preoc- surface layer parameterization. It might be better to use a
cupiedwith the radial equationof motion,not the vertical sophisticatedturbulence model with the no-slip boundary
equationof motion. The designof the experimentseemsto condition.
imply that large-scaleswirl, radialconvergence, and turbu-
(Davies-Jones) If you did that, would it lower the extreme
lent diffusionin the vicinity of the core are the only factors
wind speedsthat you are getting'?
that limit the wind speedsin tornadoes.Couldyou comment
on how your modelis coupledto the stormthat shouldbe (Fiediet) I find these results to be independent of Reynolds
above it and how the buoyancyin a stormalsois relatedto number above 2000 or so. The curves of wind speed versus
the maximumwind speedsthat couldbe achieved? Re fiatten out, and wind speeds do not exceed twice the
thermodynamic speed limit. So I find that the results are
(Lewellen)I'm notquitesureof thequestion.I agreethatall pretty much independent of diffusion.
of thefactorsyoumentionarein factinvolved. Thevertical
(Bob Walko, Colorado State University) I would like to
The Tornado:Its Structure,Dynamics,Prediction,andHazards.
disagreewith your evaluation of the way that the surface
GeophysicalMonograph79
Thispaperisnotsubjectto U.S.copyright. in 1993bythe layer is related to intensification.I found, in some results
Published
American GeophysicalUnion. that I presentedat the 15th Conference on Severe Local
56 DISCUSSION

Storms, that the intensification factor is very highly depen- environment of Florida Keys waterspouts. Irrespective of
dent on the type of surface layer parameterization that is whether there are many small, weak waterspouts or one very
used. With constant diffusion with height, which is similar to intense one, as on this day, the Key West soundings are all
the low-Re simulations that you are doing and also labora- very similar.They all show very light, east to northeastflow
tory tornado models, the intensification is indeed very and conditionally and convectively unstable stratification.
strong. However, in more realistic models in which the We were able to define the waterspout environment better
diffusion coefficient increases linearly with height in the on the few occasionsthat we were able to get aircraft data in
friction layer, there is much weaker intensification than a the subcloud layer. I found that the subcloud layer was
factor of 2. nearly adiabatic and, in some cases, was superadiabatic in
the immediate vicinity of waterspouts. An environmental
(Fiediet) Right, but again you are using a parameterization soundingneedsto be within a few hundred meters of the
that is more appropriatefor the fair-wind boundarylayer and
portion of the cloud line that producesthe vortex.
does not includethe effects of hellcity on the turbulence.St) •
think that both ofour turbulenceparameterizationsare equal!7 PAI'I..I• A5
as bad. More turbulent laboratory experimentsare needed.
Presenter, Jell' Trapp, University of ()klahoma I?'rapp and
(Walko) I would agree that both models are bad, but my t,'iedh'r, this volume, Numerical simulation of tornadolike
model showed a sensitivity to the parameterization. I think vortices in asymmetric fiowl
that you have to take the factor of 2 intensificationwith a
(Joe Golden, National Oceanic and Atmospheric Adminis-
grain of salt because it is so intense and sensitive to the
tration, ()ffice ofthe Chief Scientist) It appears to me that the
surface layer paramcterization.
nature of the underlying surface is almost incidental to the
(Fiedior) Yes, wc can take it witli z•grain of salt, but ill can modeling results. l)ocs it make any difference whether the
pull a rabbit out Of the hal here, Joe, would you like to lower boundary is solid or water, or whether it is a heat
comment on waterspouts? I don't know what the CAPE sou roe'?
Iconvcctive available potential energyl was in the Florida
(q'rapp} l'm sure that it d•es. We decided to keep our
Keysat thetimeot'this90 m s.....
• waterspout.
I don'tknow
experiment simple, i.e., incompressible, neutral stratifica-
what the thermodynamic speedlimit was on that day, but ..-
tion, constant viscosity, no-slip or free-slip boundaries. As
(Joe Golden, National Oceanic and Atmospheric Adminis- time goes on and we understand the cause and effect in lhe
tration, Office of the Chief' Scientist)The loperationall simple tk)rmulation,we can start becoming more elaborate
soundings do not give much information concerning the with more complicated
 Supercell
Thunderstorm
Modeling
andTheory
RICHARD ROTUNNO

NationalCenterfor Atmospheric
Research,
Boulder,Colorado80307

1. INTRODUCTION updraft. He attributed rotation in the funnel to the rotation of

the Earth.
Tornadoesoccurin thunderstorms.
Ferrel [1889]theo-
Association of rotation with the variation of the horizonto!
rizedthattornadoes
formwhenthethunderstorm
updraft
wind with height(wind shear) was made by a numberof
encounters
a preexisting
"gyratory"windfield.Onlylately
Europeanauthorsin the 1920s,mostnotablyWegener[1928]
hasit beenfoundthat tornadoes/waterspouts
canbe pro-
ducedby nonrotating thunderstorms (Figure2). Noticingthe intenseshear,as evidencedby the
formingin environ-
anvil outflow, he conjectured that there is a "mutterwirbl"
mentswith a preexistinglow-levelgyratorywindfield[Waki-
moro and Wilson, 1989]. However, the most intense,and (mothervortex)that gets bent down at the edges of the
long-lived,tornadoesoccurin a specialtypeof thunderstorm thunderstorm.Figure 2 is from a very thoughtfulessayby
Fulks [1962] on this topic.
knownasthe"supercell,"whichgenerates itsowngyratory
A good deal of the intellectual tension in our subject
wind field. That it doesso is interesting,
but perhapsthe
startedwith this idea sinceby the symmetryof the picture
mostfascinatingaspectof rotationin the supercell,which
(cyclonic vortex on the south, anticyclonic on the north),
becameclear in the pastdecadeor so, is the rotatingwind
onecouldlegitimatelyask, "Why isn't there an anticyclonic
field'svital role in producing
the supercell's
extraordinary
tornado associatedwith the anticyclone on the other side?"
propertiesof longlife and deviatemotion.Thusthe present During and after World War II the advent of radar and
review will focus on what was learnedfrom modelingand
rawinsondesconfirmedthe idea that a specialtype of thun-
theory about the rotation and propagationof, and the rela-
derstormoccurs when the wind shear is large. The radar
tion of tornadoesto, supercellthunderstorms. echoes showed that intense tornadoes were associated with
largecoherentechoesthat traveled to the right of the winds
2. A BRIEF CHRONOLOGY
in the cloud-bearing layer. Rotation was inferred from sur-
In previous tornado symposiathere was no discussionof face observations,and radar signaturessuch as the "hook"
thunderstorm rotation, although the literature extendsback echo. Figure 2 containsa diagram from a lecture given by
to the last century. To understandbetter the significanceof Stout [1957] showing the rotational structure of the tornado-
the discoveriesmade in the last decade,one mustappreciate bearingthunderstormas known at the time. Note the cy-
the context within which those discoveries were made. clone-anticycloneat midlevels, reminiscentof Wegener's
Hence I think it worthwhile to give a brief chronologyhere model.
of the development of ideas of thunderstormrotation.

2.1. Early Ideas 2.2. DiscoveJ3,of the Sztpercell

In the earliest writings, no real distinction was made Continuing studies by radar meteorologistsconcerned
among the various forms of convective phenomena (hurri- with hail formation focused attention on the thunderstorm's
canes, tornadoes, waterspouts, etc.) Figure 1 contains an internalflow structure.Browning [1964] proposeda special
early sketchfrom Espy [ 1841] which showsthat rotationwas airflowpattern for certain especiallysevere,long-livedthun-
not even recognized as an important property of these derstormshe called "supercelIs" [Browning, 1968].His idea
waterspouts. The recognitionthat rotation is importantwas was that longevity is achieved by a flow pattern (Figure 3)
made by Ferrel [1889]. As was mentioned, he made the that allowsthe updraft to unload its rain so it will not disrupt
connection between the tornado and a parent thunderstorm the storm's inflow; the airflow is also consistentwith right-
The Tornado: Its Structure, Dynamics, Prediction,and Hazards.
ward propagation since, from a frame of reference moving
Geophysical Monograph 79 with the storm, there is a component of the airflow from the
Copyright 1993by the AmericanGeophysicalUnion. south at all levels shown. Also note the placement of
57
58 SUPERCEI.1I.. 'FHUNDERS•I'()RM Mr)DEfilING AND

Espy 1841 Ferrel, 1889

[1841] view thai lowered pressurewittiin illtensely rising :tit •cc•lnls lt•r tlic c{•idc•lsi•li•)•l 11•rl•cl;•ict i't'rrt'l's I laagl
recognilionthat the funnels niark r•)tating cc)lt•irlllS•I' air.

tornado at the microfront based on the visual observations of 2.3. Early l)(;l)pler Radar h•,'•'sli,k,ation,v and Three-
Ward [ 1961]. Dimensional ('1o,d Models
in the same period, Browning and Lando' [1963] and
Moving into the 1970s. the technological advance of
Ban•es [1968] developed the idea that the inflow branch of
Doppler radar allowed for a view of the actual three-
the supercellmodelcarried horizontalvorticity that could be
tilted upward and hence account for the rotation in the dimensional flow within the supercell. Figure 5 contains a
updraft (Figure 3). The idea was taken to be incompletein figtire adapted from Ray [•976] showing the cyclone-
some quarters since no accounting for the anticyclonic anticyclone structure of the midlevel flow, with the domi-
vorticity {implicitin any vortex-tiltingtheoryt was made. nance of the cyclonic vortex at low levels, much like
Also, in this same period, a study which I think was the •$'tottt[1957] picture. in addition, the indications were
viewedas a curiosityat the time would later providea vital that the cyclonic rotation was mostly in updraft and the
cluein unraveling the mysteryof the supercell.The paperby anticyclonic rotation was mostly in downdrati. At around
Fujita and Grandoso[1968](Figure4) showedthata growing the same time, computer power had advanced to the point
thunderstorm may split into a rightwardpropagating,
cy- where three-dimensional cloud models could be contem-
clonic thunderstormand a leftward propagating,anticy- plated. Figure 5 shows results from the early simulation
clonicthunderstorm.Browning[1968]also pointedout this by Schlesinger[1975] showing a vortex pair centered on an
possibilityof a leftwardmovingcounterpartto the '•super- updraft at midlevels but little else that resembled the super-
cell" shown in Figure 3.
 ROTUNNO 59

tornado

Wegener, 1928 Stout, 1957

Brooks, 1949
Markgraf, 1928
van Everdingen, 1925 Stout and Huff, 1953

Huff, Hiser, and Bigler,

1954

Fujita, 1950s
Fig. 2. AccordingtoWegetter[•928]thee•ects•fwindshearpr•duceah•riz•nta•y•riented``mutterwirb•thatgets
bent downat the sides[Markgraf, 1928]implyingrotationabouta verticalaxis (cyclonicto the south/anticyclonic
to the
north) [from Fulks, 1962].Weatherradar, radiosondes,and surfaceobservationsallowed Stottt [1957]to deducethe
circulationwithin a tornado-bearing
thunderstorm.Note the cyclonic(anticyclonic)flow on the south(north) side(as
in the Wegenerpicture).(The Fujitapapersarediscussed
by Fujita [1963].)

2.4. Disco•'e•3:of the Nltmerical-Model~Equivalent periodsthan were the Schlesinger[1975]simulations.One of

Supercell the critical insights of Klemp and Wilhelmson was that the
It was not until the three-dimensional modeling papers by supercellular structure shown by Browning [1964] is ac-
Klemp and Wilhehnson[1978a, b], showingthe connec- quired during the act of cell splitting. From the southward
tionsamongthunderstormsplitting,supercell-flow structure, travelingstorm'spoint of view (Figure 6) there is inflow from
the south at all levels. Since the environmental winds are all
andthe rotationalpropertiesof thesestormsdidthevarious
piecesof the puzzlebeginto fall into place.Theselater east-west, this means there is deviate motion, and judging
simulations(e.g., Figure6) werecarriedoutfor muchlonger from the figure, the cell lives a long time. Also, there is
SUPERCELL THUNDERSTORM MODELING AND THEORY

Browning, 1964 Barnes, 1968

Browning and Ludlam,

1962 Browning and Landry,
1963

Browning and Donaldson,

1963
Fig. 3. Browning [1964] proposed that a special three-dimensionalairflow structure allows for a storm cell to be
long-lived.The flow diagramis drawn from the frame of referencemovingwith the storm. The flow arrows indicatethat
the storm motion has a velocity componentacrossthe shear(in addition to the usual component with the shear); in the
parlanceof severe-stormforecasters, it exhibits deviate motion. Barnes [1968] carried out a systematicdata analysis
and showedthat a storm moving acrossthe shearis thus able to tilt up the horizontal vorticity associatedwith the shear
and so accountfor the cyclonic rotation in these severe rightward deviating
 ROTUNNO

STEP
I STEP2 STEP
:3 STEP4 STEP5

i::i"'• .....::.:...:,:.:?":
----

' DEVELOPMENT OF ANT!CYCLONIC STORM ß

! o'

.o Fant
i • .• '

ßV .,•
o•
"•'., "'
ß •, '• I / ..,.

N j ..... . .....

01,.-

'• .. ; , ,./"•..
'0

ß DEVELOPMENT OF CYCLONIC STORM cyc

..... ..... ............. ............

Fujita and Grandoso, 1968

Splitting: Left moving:

Hitschfeld, 1960 Hammond, 1967
Achtemeier, 1969 Browning, 1968
Charba and Sasaki, 1971
Fankhauser, 1971
Fig. 4. Observationsof Fujita and Grandoso[1968]showingthat a stormcell may split into two cells, one rotating
cyclonically and propagatingto the right, the other anticyclonically,propagatingto the left, facing in the general
direction of storm
62 SUPERCEI.I• THUNDERST()RM MODELING AND THEORY

2.45 • 1-12.5• I'tlN Ck$[ P2

Ray, 1976 Schlesinger

1975

Kropfli and Miller, 1976 Wilhelmson, 1974

Brandes, 1978 Miller and Pearce,

1974

Patushkov, 1975
Fig. 5. In the 1970sthe technological advancesof Dopplerradar and more powerfulcomputersprovidedfresh
impetusfor severe-stormresearch.Ray's [1976]deductionof the windsin the supercellshoweda vortex pair at
midlevels(upperpanels;vorticityin dashedlines,divergencein solidlines)anda flow dominatedby cyclonicrotation
a low levels(lower panels).The early simulationby Schlesinger[1975]showsthe developmentof a vortex pair at
midlevels (middle panel) in a three-dimensionalnumerical cloud
 ROTUNNO 63

-6 -6 : X• •
-12 -12

"--'---.--30 ,,,in 2. IS Iom(9

-18 -18 r'

-• .• F---o,,-...... .-- •o•,• •..•.•

-18
-•8 -•2 -• 0 .6 +•2 +•8 -18 -12 • 0 .6 .12 •18
•(m) .(•)

Wilhelmson and Klemp, 1978

Thorpe and Miller, 1978

Klemp and Wilhelmson,1978

Fig. 6. In a seriesof papers,Klemp and Wilhelmson[1978a, b] showedthat althoughthe vortexpair is a signature
of the nascentthermal, mature supercell-likestructureis not obtaineduntil after rainy downdraftsdevelopand the
initial cell splits into two cells propagatingto the left and right, respectively,of the mean shear. (Only the rightward
propagating member is shown
64 SUPERCELL THUNDERSTORM MODELING AND THEORY

(a)

Fig. 7. Illustration [from Klemp, 1987] of the development of rotation within a simulated thunderstorm through
vortex-line tilting. (a) in the early stage a vortex pair forms through tilting of the horizontal vorticity associatedwith
the mean shear. (b) As rainy downdrafts form and the cell splits, vortex lines are tilted downward, and the original
updraft-centeredvortex pair is transformedinto two vortex pairs.The rightward(facingwith the shear)movingmember
propagates towards the positive vorticity on the right flank, and thus a correlation between updraft and positive
vorticity develops.

mirl-or image anticyclonic storm (not shown) as observed by occurred to a few of us in the early 1980sthat even if we did
Fujita and Grandoso [1968]. not understandthe across-shearpropagation, we could ask
how the rotation of the storm develops given the storm
3. MODELING AND THEORY OF THUNDERSTORM ROTATION
motion. Later, it was discovered that rotation in turn induces
3.1. Midlevel Rotation propagation, about which more appears later.
Although these simulations showed across-shearpropaga- Figure 7 is an artistic rendering of the thunderstorm
tion, the reasons for this were not clear at the time. It evolution shownin Figure 4 taken from the review by
 ROTUNNO 65

VORTEX
LINES
• •.•
,. 'X, • I,,'• •
VERTICAL
DISPLACEMENT

STORM-
RELATIVE
MEAN FLOW

STORM-RELATIVE
MEAN FLOW

Fig. 8. Within the context of linear theory, the thunderstormis thoughtof as a "bump" in the isentropic surface (or,
more precisely, moist isentropicsurface)[from Davies-Jones,1984].If one gives the storm motion vector ?, then one
can deduce the phase relation between vertical vorticity and vertical velocity, since the flow lines and vortex lines
follow the isentropic surface.The vortex lines over the bump always imply a vertically oriented vortex pair. In the
upper panel the storm motion is purely in the sheardirection,and so the updraft is colocatedwith the bump and so
straddles(is out of phasewith) the vortex pair. In the lower panelthe stormmotionis acrossthe sheardirection,the
verticalvelocityis up on the right (facingdownshear)anddownon the left; hencethe vertical vorticityis in phasewith
the vertical velocity.

[1987]. As shown in Figure 7a, during the early stage of Barnes diagram, but also an anticyclonically rotating down-
updraft developmentthere is a tilting of the vorticity asso- draft as suggestedin Ray's dual-Doppler analysis.
ciated with the mean shear flow which gives rise to a vortex These ideaswere developed on the basisof a linearization
pair straddlingthe updraft,muchlike theWegener/Markgraf of the vertical vorticity equation about the mean shear flow
picture.As shownin Figure7b, subsequent stormsplitting [Rotunno, 1981; Lilly, 1982]. The most complete develop-
impliesdecreasedvertical velocity,and eventuallydown- ment of this line of reasoning was given by Davies-Jo•zes
draft, developing at the locationof the originalupdraft.This [1984]. Figure 8 clarifies this distinction between stormsthat
behaviorof the verticalvelocityimpliesa downwardturning move with the shear from storms that move across the shear.
of the vortex linesthat in turn impliesthat theoriginalvortex Here the thunderstorm is the "bump" is the isentropic
pair, too, splitsinto two vortex pairs. Next, the two new surface;the top panel shows that with along-shearpropaga-
updraft-downdraft coupletsbeginto propagateacrossthe tion the updraft is maximum on the centerline upshear;
sheartowardthe vorticitycentersof the originalcell. This hence the vortex pair straddles the updraft. The bottom
across-shear propagation of the updraft/downdraft pairto- panelshowsthat with across-shear propagationthe updraft
sectorstendsto correlate is maximum acrossthe shear (on the right); hence it tends to
wards the vorticity-production
verticalvelocityandverticalvorticity.Hencetherightward be colocated with the vertical vorticity. Viewed from the
propagatingcellhasa cyclonically updraft,asin the storm's reference frame the inflow from the right brings
rotating
61• S[JPF•RC[:;•II•L THUN[)F. RSTORM MODELING AND THEORY

Fig. 9. lJl•Irali•)n Jt'romKk,,np, 1987j •1' how lhc Iow-levuJh()r•z()nlal vorlicity i• m(•ditied by lhc lcmpcrature
•radicnl bcncalhlhc lhundcrsl()rm.'l•hcanlbicnl mean shearvorlicity p()inlsacr(•ssI'romsoulh 1()norlh: however, coJd
a•r ()n the n(•rlh •idc meanslha• v()rt•c•ly direclod l'r•)n•east io we•l i• cre•tcd alon• the cold-air bo•ndary. Thi• vorticJly
c•n then I•c tilted tipw•u'das si•)wn and is rcsponsihletk•rlhe strong I()w-levcl lhundcrst()rn•r()tation.

mcan shear vorticity in with it; hence the term "streamwise Figure 9, the relative intlow l'rom the east has its horizontal
vorticily" {this quantity multiplied by the storm-relative vorticity changedby baroclinic production. This baroclini-
wind specd is the now-popular "storm-relative hellcity"). It caily produced vorticity can then be tilted upward to pro-
is important to keep in mind that we have taken storm duce strongpositive vertical vorticity. Hence the dominance
motion as a "given" and without that given one does not of the cyclone at low levels as in the Stout picture and in
know which way "streamwise" is. Ray's Doppler radar study.
Da•'ie,s'-,lon½s[1984] and Rotunno and Klemp [1985] no- At this point we are still stuck with the symmetry of the
ticed that the conservation of equivalent potential vorticity picture, since there is an implied mirror image storm (not
accounts for the fact that these predictions from the linear shown) to the left of the shear vector. One could still
models are much better they should have been since the reasonablyask, "Why isn't there always an anticyclonic
latter are based upon a small-amplitude assumption. supercell(andtornado)when there is a cyclonicsupercel!'?"

3.2. Low-Level Rotation 4. CYCLONIC BIAS AND PROPAGATION

The three-dimensional numerical simulations also re- The second critical insight of the Klemp-Wilhelmson
vealed a previously unsuspectedsource of low-level rota- papersisthedistinctiontheydrewbetweenveeringwindand
tion. The relatively simple picture of the tilting of mean- veering shear. A straight-line hodograph may represent a
shear vorticity by the updraft worked pretty well for veeringwind, but there is completemirror symmetry of a
midlevel rotation but not so for low levels. As noted in the disturbance about the shear vector. They found through
early three-dimensionalsimulations, a feature of the long- numerical experiment (assuming a horizontally homoge-
lived supercellis that the rain-cooledsurfaceoutflowof the neous,nonrotatingenvironment)that the right-movingmem-
thunderstorm stays positioned beneath the main updraft ber of the splitting pair is enhancedwhen the shear vector
[Thorpe and Miller,1978]. As shown in the schematicin veers (when the hodographcurves clockwise, which is
 ROTUNNO 67

Fig. 10. Plan views of cloud-model-produced rainwater fields (at 1.8 km above ground level) for two simulations
using,respectively, a straight-linehodographand one with low-level clockwise curvature. The straight-linehodograph
producesstorms with complete mirror symmetry, while the curved-line hodographenhancesthe rightward moving
storm [from Klemp, 1987].

try in updraft production is related to the vertical pressure

typical case). This is illustrated in Figure 10 [from Klemp,
1987]. gradient. In brief, we found that according to linear theory
Rotunno and Klemp [1982] noticedthat the bias in these the perturbation pressure goes like the dot product of the
simulationsoccursfrom the earliesttime (whenthe updraft's shear and the gradient of vertical velocity w. As Figure 11a
amplitude is small); this suggestedthat a linear analysis shows, this indicates high-pressure upshear and low-
mightshedlighton the matter.Sincethe updraftis arguably pressure downshear of the updraft; in this example of
the definingfeature of a thunderstorm,we looked at the nonveering shear there is complete mirror symmetry about
vertical momentumequationand deducedthat the asymme- the shear
68 SUt•ERCEI,L THUNDERSTORM MODELING AND THEORY

(b)

Fig. 11. Illustration[from Klemp, 1987]of the pressureperturbations arisingas an updraftinteractswith an

environmental wind shearthat (a) doesnotchangedirectionwith heightand(b) turnsclockwisewith height.The high
(H) to low (L) horizontalpressuregradientsparallelto the shearvectors(flatarrows)are labeled.The shadedarrows
indicate the implied vertical pressure gradients.

However, if the shear veers with height, Figure 11b shows Rotunno and Klemp [1982] also found, consistentwith
that the high-low couplet also veers creatingan enhanced Schlesinger[1980],that there are lifting pressuregradients
upwardpressuregradientforce on the right and a negative on the storm flanks at work even in the nonveering shear
such force on the left. Hence the bias. (The basic idea of case.They identifiedthe loweredmidlevelpressureon the
lookingfor preferentialverticalpressuregradientsto explain flankswith the strongrotation there. Weismanand Klemp
deviate motiongoesback to Newton and Newton [1959]). [1982, 1984] carried out detailed sensitivitystudies
 ROTUNNO 69

c) 150 min
,,,

• I I i i i i i Iil I i i t i i i

2' ! .• ... ,-

Z.=q.K m-I
t• !'"'•1I I .t I I t I • I I I I I

L\ % % %"',,"-,'--'-.---,.,--...

I";,\ ,'", z=
,'-,,"-,
,"I, 25k, ,,!
,--,',---,
T",
m__
37 17 X 37 17 X 37

Fig. 12. Plan view oœverticalvelocitycontoursand horizontalflow arrowsfrom a specialno-raincloudsimulation

[fromRotunnoandKlemp,1985].Thetimesequence showsacross-shear propagation
oœtheupdraœt, eventhoughthere
is no rain loading or evaporativelycooledcold air at the surface.

analysis showing that strong shear is a critical factor that 13). On the other hand, the linear models of storm rotation
allows a supercell to form and that these stormsare dynam- discussedabove predict that a steadily propagatingupdraft
ically different than their nonsupercellularcousinsby virtue will be in phasewith the vorticity. It is clear that the linear
of the importanceof the rotationallyinducedlifting pressure theory can tell one only that there is a tendency for the
gradients. In fact, Rotunno and Klemp [1985] (Figure 12), updraft to "catch up with the vorticity" center, since if the
carried out a specialsimulationwith the rain-makingprocess updraft actually did catch the vorticity, the mechanismfor
set to zero and showed that splitting and across-shear propagation is cancelled (like what happens if the dog
propagation can occur without the rain loading and rain- actually catches the car).
cooled outflow. (Also, there was little low-level rotation to
speak of, confirming the importance of the low-level ba- 5. ENERGETICS
roclinic generation mechanism for producinglow-level rota-
tion.) I think most observers would judge the supercell (this
This mechanismof dynamicallyinduceduplift is finding solitary, rotating and propagatingthunderstorm)an extraor-
application in such thermodynamicallyand geographically dinary, possibly singular, meteorologicalphenomenon. Al-
diverse thunderstormsas the low-precipitation(LP) thunder-though the work just described gives, I believe, a good
storms of the high plains discussedby Weisman and picture of how the machine works, one suspectsthat there
Bluestein [1985] and the tornado-producingthunderstorms might be some kind of "super" principle at work. Since
associatedwith land-fallinghurricanesstudiedby McCaul 1986, D. K. Lilly and collaborators have looked for an
[1991]. energy principal.
These simulations show that the essence of across-shear Through the investigation of several simple models of
propagationis that new updraftbe continuously forcedon buoyant thermals in shear, Lilly [1986a] (Figure 14) inferred
the updraft'sflank by the enhancedrotationthere (Figure the very different energetics occurring in
70 SUPERCELL THUNDERSTORM MODELING AND THEORY

versus twt)-dimcnsional fi•)w. A twt)-dimcnsional distur-

bance in shear gives up •ts energy 1o the mc•tn flow, while a
three-dimensionaldisturbance is able t(• derive energy l'rom
Updraft
the mean flow, this effect being tantamount to the production
of rotation about the vertical axis derived through the tilting
of mean flow horizontal vorticity. ()n the basis of a simple
Shear-Induced picture of a rotatingupdraft, l,i!ly furthcr conjccturcdthat
Low Pressure
part of the rotationalkinetic cncrgy could be translbrrcd
the divergent kinetic cncrgy, thc latter ot' which, as mcn-
tiened, goes back to the mcan flow; hence a positivc-
l'kcdbackloop is (lcscribc(l.
These ideas wcrc cvaltiatcd by Wtt[199{}] by •n
two simt•lltti(•ns (straight tl{•cl(>gr;tph versus ctlrvcd
t•oclc)graph,
but also different lhcrmc>dyl•tmics)
Fig. 13, Analysis of the no-rain simulation shown in Figure 12
:s•ng the K!cmp-Wilhciinst•n model. In t•t•lh c:tscs, rt)ta-
This figure shows that rotationally induced low pressure on the
south (across-shearside) of the updraft induces low pre.ssurcthere ....',•al kinetic energy definitely derives t'rt)m lilt tnc•tn
that forces conditionally unstable air upward (shaded arrow) that in ,no divcrgcnl kinetic energy dctin•lcly derives I'r{>iii the
turn regeneratesthe,updraft there [from Rolunno and K!emp, 1985]. pressuretransl•r term. flowever, the clircctit•n of the ex-
change betwccn the (livergent •tncl rc)talit•nal c•lcrgy and

, ,

POTENTIAl_, MEAN FLOW

ENERGY KINETIC ENERGY
-p-z .....

VERTICAL DISTURBANCE DISTURBANC[:.

DIVERGENT ROTATIONAL

KINETIC
ENERGY
bw2/2
.•(•-
........
• •Va.V
•._..___•'
KINETIC

vJ/2
ENERGY
- rCv.vv)/2/
KINETIC

3v/2
ENERGY

...... EST'
"'-1 ENERGY
'•'•........

COUS DISSIPATION

Fig. 14. Lilly's [1986a] proposalfor the energeticsof the supercell.Seetext for
 ROTUNNO 71

Fig. 15. Structureof low-leveltlowandprecipitation

patternof a
•upercell thunderstorm as inferred from surface observers and radar
an•lysis(from Lemonand l)oxwell II979], as adaptedby Davie.s-
,h,nc.s [1986l). FFI) is I2)rward flank downdraft, RFD is rear flank
downdraft, and 'I' is tornado.

Fig. 16. Results from a fine-resolution simulation of the tornado-

bearingregionof a supercell[from Klemp and Rotlmno, 1983].Note
the similaritieswith the observedstructureshownin Figure 15. T is
between the (.livergentenergy and the mean flow is different tornado.
in bothcases.Alsoin bothcasesthedivergentkineticenergy
dicl not recycle energy back to the vertical kinetic energy.
Hence the feedbackaspectsof the Lilly pictureare perhaps
nr•t essential l'or the supercell. I think it would be useful to emphasizedthat the entire gust front undergoesa contortion
pert'otto the same energetics analysis on the Rottmno and reminiscentof the occlusion of a large-scalecold and warm
Kh'mp 119851no-rain case which had an erect-to-downshear front with the tornadoappearingat the tip of the occlusion.
leaning updraft. Klemp et al. [1981] used the Klemp-Wilhelmson three-
Lilly [1986b] proposedthat since supercelIsare, by their dimensional cloud model to simulate the May 20, 1977,
nature, helical (strong alignment between velocity and vor- tornadic thunderstorm observed with dual-Doppler data.
ticity vectors), and since helicity suppressesturbulent en- Although many of the important features of the thunder-
ergy cascade, supercelIs are therefore less prone to dissipa- storm, such as the general flow structure and rotational
tion than ordinary cells. Wu [1990] made a crude estimate of character, were obtained, the fine-scale features associated
the inertial transfer term, which shows that the grid-resolved with tornadogenesiscould not be captured with 1 km hori-
transfer is reduced as the simulated storm cell acquires zontal grid resolution. Klemp and Rotunno [1983] took the
helicity. These results are very interesting, and a more model solution near the time of storm maturity, interpolated
complete spectral energy budget is certainly worth doing. the model data from the vicinity of the mesocycloneto a
nestedfiner grid (250 m), and then continued the integration.
What was found (Figure 16) was a sequenceof developments
6. WHERE THE TORNADO Frrs •N...
that were remarkably in parallel with those described by
When the supercell produces a tornado, there are a Lemon and Doswell [1979]. A rear flank downdraft devel-
number of attendant features of the finer-scale structure of oped in association with an occluded structure to the low-
the supercell that change. These were reviewed and empha- level temperature field and with very high values of vertical
sized in the important essay by Lemon and Doswell [1979] vorticity at the tip of the occlusion.
(Figure ! 5), which was basedon a decadeof their andothers' Although this was encouraging, and allowed Klemp and
personal visual observationsof tornadic thunderstormsin Rotltnno [1983] to come up with a plausible physical model
conjunctionwith concurrentDopplerradar observations.In of the process, the results were less than totally satisfying
particular,they affirmedWard's [1961]observationthat the becausethe nesting is turned on at a time when the storm is
tornado forms between the inflow and outflow, and they fully developed. One could wonder when, or even
72 StJPERCELL THUNDERSTORM MODELING AND THEORY

the same flow scenario would occur if high resolution were storm, in Preprints, loth Weather Radar ConJkrence, pp. 116-
used throughout the simulation. Here the specter of predict- 122, American Meteorological Society, Boston, Mass., 1963.
ability looms since the tornado forms near where the cool air Browning, K. A., and F. H. Ludlam, Air flow in convectivestorms,
Q. J. R. Meteorol. Soc., 88, 117-135, 1962.
and the warm air collide (a region of large wind shear that is Charba, J., and Y. Sasaki, Structure and movement of the severe
almost certainly unstable by any linear stability analysis). thunderstormsof 3 April 1964 as revealed from radar and surface
Wicker and Wilhebnson [1991] have carried out simulations mesonetwork analysis,J. Meteorol. Soc. Jpn., 49, 191-213, 1971.
of a supercell usinginteractive, nestedgrids with a horizon- Davies-Jones, R. P., Streamwise vorticity: The origin of updraft
tal grid interval of 120 m in the innermost grid system. The rotation in supercellstorms,J. Atmos. Sci., 41, 2991-3006, 1984.
Davies-Jones, R. P., Tornado dynamics, in Thunderstorm Morphol-
simulations show vividly that the precise prediction of the ogy and Dynamics, edited by E. Kessler, chap. 10, pp. 197-236,
time and location of the tornado circulation is difficult, but University of Oklahoma Press, Norman, 1986.
that,oncestarted,a very intensevortex(44 m s "•) canbe Espy, J.P., The Philosophy of Storms, 552 pp., Little, Brown,
obtained. Boston, Mass., 1841.
One final observation: Rotunno [1986] pointed out that Fankhauser, J. C., Thunderstorm-environment interactions deter-
mined from aircraft and radar observations, Mon. Weather Rev.,
even with very high resolution,a cloud model usingfree-slip 99, 171-192, 1971.
I{•wcr boundary probably could not simulate the important Ferrel, W., A Popular Treatise on the Winds, 505 pp., John Wiley,
structural features of the tornado. Although the free-slip New York, 1989.
simt,lations by Wk'ker and Wilhelmson [19911 described Fujita, T., Analytical mesometeorology:A review, Meteorol.
above show that tornadic velocities can be achieved, I am Monogr., 5, 77-128, 1963.
Fujita, T., and H. Grandoso,Split of a thunderstorminto anticy-
still impressedby the early photogrammetryof tornadoes clonic and cyclonicstormsand their motionas determinedfrom
showing vcrticalvelocities of 50 m s• • 50 m offtheground numericalmodelexperiments,J. Atmos. Sci., 25,416-439, 1968.
(see Roluttno [19861for references). Moreover, there is good Fulks, J. R., On the mechanicsof the tornado, Rep. 4, 33 pp., Natl.
evidence that horizontal winds in a strong tornado may SevereStormsProj., Dep. of Cornmet.,Washington,D.C., 1962.
Hammond, G. R., Study of a left moving thunderstormof 23 April
cxcccd 1(}()m s •. Such structure certainly cannot be 1964, ESSA Tech. Memo. ERLTM-NSSL 31, 75 pp., Natl. Severe
simulatedby a model with 120m vertical resolution,and on Storms Lab., Norman, Okla., 1967.
the basisof a largebody of theoreticaland laboratorystudies Hitschfeld, W., The motion and erosion of convective storms in
I believe a no-slipcondition is necessaryto producesucha severe vertical wind shear, J. Meteorol., 17, 270-282, 1960.
structure.
Hull', F. A., H. W. Hiser, and S. G. Biglet, Study of an Illinois
tornadousingradar,synopticweatherandfielddata,73 pp., State
Future research will continue on the smaller scale to reach Water Surv., Urbana, I11., 1954.
a better understandingot' how the tornado fits into the Klemp, J. B., Dynamicsof tornadicthunderstorms,
Annu. Rev.
thunderstorm. (1 still do not know where the tornado vortex Fluid Mech., 19, 369--402, 1987.
signature
(MVS) fitsinto the picture[seeRotltnno,19861).I Klemp,J. B., andR. Rotunno,A studyof thetornadicregionwithin
a supercellthunderstorm,
J. Atmos.Sci., 40, 359-377,1983.
believe wc are at the point where the modeler can start Klemp, J. B., and R. B. Wilhelmson,The simulationof three-
looking upscale,to understandhow these thunderstorms dimensionalconvective storm dynamics, .I. Atmos. Sci., 35,
behave in horizontally inhomogeneous,time-de.pendenten- 1070-1096, 1978a.
vironments. Klemp,J. B., and R. B. Wilhelmson,
Simulations
of right-and
left-movingstormsproducedthroughstormsplitting,J. Atmos.
Sci., 35, 1097-1110, 1978b.
Klemp,J. B., R. B. Wilhelmson,and P.S. Ray, Observedand
Acknowledgments. Discussionswith M. L. Weismanare grate- numericallysimulatedstructureof a maturesupercellthunder-
fullyacknowledged.
The NationalCenterfor Atmospheric
Research storm, J. Atmos. Sci., 38, 1558-1580, 1981.
is sponsored
by the NationalScienceFoundation. Kropfli,R. A., and L. J. Miller, Kinematicstructureand flux
quantitiesin a convective
stormfromdual-Doppler radarobser-
REFERENCES vations, J. Atmos. Sci., 33,520-529, 1976.
Lemon,L. R., andC. A. DoswellIlI, Severethunderstorm evolu-
Achtemeier, G. L., Some observationsof splittingthunderstorms tion and mesocyclone structureas relatedto tornadogenesis,
overIowa on August25-26, 1965,in Preprints,6th Conferenceon Mon. Weather Rev., 107, 1184-1197, 1979.
SevereLocal Storms,pp. 89-94, AmericanMeteorologicalSoci- Lilly, D. K., The development and maintenance of rotationin
ety, Boston, Mass., 1969. convectivestorms,in Topicsin Atmosphericand Oceanographic
Barnes,S. L., On the sourceof thunderstorm
rotation,ESSATech. Sciences:IntenseAtmosphericVortices,editedby L. Bengtsson
Memo. ERLTM-NSSL 38, 28 pp., Natl. Severe Storms Lab., and M. J. Lighthill,pp. 149-160,Springer-Verlag,
New York,
1982.
Norman, Okla., 1968.
Brandes,E. A., Mesocyclone Some Lilly, D. K., Thestructure,
evolutionand tornadogenesis: energetics
andpropagation
of rotating
observations,Mon. WeatherRev., 106, 995-1011, 1978. convectivestorms,I, Energyexchangewith the meanflow, J.
Brooks,E. M., The tornadocyclone,Weatherwise,2, 32-33, 1949. Atmos. Sci., 43, 113-125, 1986a.
Browning,K. A., Airflow and precipitation within Lilly,D. K., Thestructure,
trajectories energetics
andpropagation
of rotating
severelocal stormswhich travel to the right of the meanwind, ar. convectivestorms,II, Helicity andstormstabilization,d. Atmos.
Atmos. Sci., 21,634-639, 1964. Sci., 43, 126-140, 1986b.
Browning,
K. A., Theorganization
of severe Weather, Markgraf,
localstorms, H., EinBeitrag
zuWegeners
mechanischer
Trombenthe-
23,429-434, 1968. orie, Meteorol. Z., 45, 385-388, 1928.
Browning,
K. A., andR. J. Donaldson,
Jr.,Airflowandstructure
of McCaul,E. W., Buoyancy andshearcharacteristicsof hurricane-
a tornadicstorm,J. Atmos. Sci., 20, 533-545, 1963. tornado environments,Mon.Weather Rev.,119,1•54-1978,1991.
Browning, K. A., andC. R. Landry,Airflowwithina tornadic Miller,M. J., andR. Pearce,A three-dimensional
primitive
 ROTUNNO 73

model of cumulusconvection,Q. J. R. Meteorol. Soc., 100, Thorpe, A. J., and M. J. Miller, Numerical simulationsshowingthe
133-154, 1974. role of the downdraughtin cumulonimbus motionandsplitting,Q.
Newton, C. W., and H. R. Newton, Dynamical interactionsbe- J. R. Meteorol. Soc., 104, 873-893, 1978.
tween large convective clouds and environment with vertical van Everdingen, E., The cyclone-like whirlwinds of August 10,
shear, J. Meteorol., 16,483-496, 1959. 1925, Proc. R. Acad. Amsterdam Sect. Sci., 25,871-889, 1925.
Patushkov, R. S., The effects of vertical wind shear on the evolution Wakimoto, R. M., and J. W. Wilson, Non-supercelltornadoes,
of convectiveclouds,Q. J. R. Meteorol. Soc., 10I, 281-291, 1975. Mon. Weather Rev., 117, 1113-1140, 1989.
Ray, P.S., Vorticity and divergencewithin tornadic stormsfrom Ward, N. B., Radar and surfaceobservationsof tornadoesof May 4,
dual Doppler radar, J. Appl. Meteorol., 15, 879-890, 1976. 1961, in Preprints, Ninth Weather Radar Conference,pp. 175-
Rotunno, R., On the evolution of thunderstorm rotation, Mort. 180, American Meteorological Society, Boston, Mass., 1961.
Weather Rev., 109, 577-586, 1981. Wegener, A., Beitrage zur Mechanik der Tromben und Tornados,
Rotunno, R., Tornadoes and tornadogenesis,in Mesoscale Meteo- Meteorol. Z., 45, 201-214, 1928.
rology and Forecasting, edited by P.S. Ray, pp. 414-436, Weisman, M. L., and H. B. Bluestein, Dynamics of numerically
American Meteorological Society, Boston, Mass., 1986. simulated LP storms, in Preprints, 14th Conference on Severe
Rotunno, R., and J. B. Klemp, The influence of the shear-induced Local Storms, pp. 167-170, American Meteorological Society,
pressure gradient on thunderstorm motion, Mon. Weather Rev., Boston, Mass., 1985.
I10, 136-151, 1982. Weisman, M. L., and J. B. Klemp, The dependenceof numerically
Rotunno, R., and J. B. Klemp, On the rotation and propagationof simulated convective storms on vertical wind shear and buoy-
simulated supercell thunderstorms, J. Atmos. Sci., 42, 271-292, ancy, Mon. Weather Rev., I10, 504-520, 1982.
1985. Weisman, M. L., and J. B. Klemp, The structure and classification
Schlesinger, R. E., A three-dimensional numerical model of an of numerically simulated convective storms in directionally vary-
isolated deep convective cloud: Preliminary results, J. Atmos. ing wind shears, Mort. Weather Rev., 112, 2479-2498, 1984.
Sci., 32,934-957, 1975. Wicker, L. J., and R. B. Wilhelmson, Numerical simulation of
Schlesinger, R. E., A three-dimensional numerical model of an tornadogenesiswithin a supercell thunderstorm, this volume.
isolated deep thunderstorm, II, Dynamics of updraft splitting and Wilhelmson, R. B., The life cycle of a thunderstorm in three
mesovortex couplet evolution, J. Atmos. Sci., 37, 395-420, 1980. dimensions, ,I. Atmos. Sci., 31, 162%1651, 1974.
Stout, G. E., Mesometeorological systems from dense network Wilhelmson,R. B., and J. B. Klemp, A three-dimensionalnumerical
stations, paper presented at IUGG meeting, Int. Union of Geod. simulation of splitting that leads to long-lived storms, J. Atmos.
and Geophys., Toronto, Ont., Canada, 1957. Sci., 35, 1037-1063, 1978.
Stout, G. 17 , and F. A. Huff, Radar records Illinois tornado genesis, Wu, W.-S., Helical buoyant convection, Ph.D. thesis, 161 pp.,
Bull. Atn. Meteorol. Soc., 34,281-284, 1953. Univ. of Okla., Norman,
 NumericalSimulationof Tornadogenesis
Within a SupercellThunderstorm
Louis J. WICKER AND ROBERT B. WILHELMSON

Departmentof Atmospheric
Sciences
andtheNatioualCenterfi,' Supercompuling
Applicatious,Unive;'sitv
of lllinoi3.
Urbana-Champaigu, Illinois 61801

1. INTRODUCTION which forms in the surface friction run is stronger and has
slightly larger updrafts near the ground. These differences
During the past 15 years, numerical simulations have were due to surface friction, which caused the low-level
captured many features observed in supercell storms wind to be subcyclostrophic,allowing the development of
[Wilhelmson and Klemp, 1978; Weisman and Klemp, 1982, strong low-level inflow into the vortex and enhancingthe
1984]. Encouraged by the successin simulatinglarge-scale updraft within the vortex.
supercell features, Klemp and RotlmnO [1983] (hereafter Snow [1982] divides the study of tornado dynamics into
referred to as KR) used a limited region fine-meshsimulation two categorieson the basis of scale. The larger-scalevortex,
with 250 m horizontal resolution to investigatesmall features called the tornado cyclone, is an intense vortex present
within the Del City supercell [Klemp et al., 1981]. ("Reso- within the mesocyclone circulation of the •upcrcell storm.
lution" refers to the spacing between grid points. The (Note that other researchers define a tornado cyclone as a
smallest scalesadequately resolvedby the grid are roughly 4 mesocyclonethat producesone or more tornadoes. We use
times the grid increment.) The model successfullyrepro- Snow's definition here.) The smaller-scale vortex, the tor-
duced observed features such as the divided mesocyclone nado itself, is the intense columnar vortex which forms
(half updraft, half downdraft), the rear flank downdraft, and within the tornado cyclone. Results from Kle,•p and Ro-
the low-level occlusion process. Further, strong low-level tunno [1983] and Wicker [1990] demonstrate that high-
convergenceconcentratedpreexistingverticalvorticity into resolution numerical simulations of supercell storms can
a ringlikeregionwith maximum
valuesof 0.06s-•. The produceintensevortices having horizontal scalesabout a
results demonstrated that small features often observed with kilometer in diameter within the mesocyclone, even in the
a tornadic storm can be generated with a numerical storm absence of surface friction. We believe that these vortices
model if sufficient resolution is employed. However, the representtornadocyclonesand concludethat surfacefric-
vertical resolution used was rather coarse (• = 500 m), and tion is not crucial in their creation. However, it is well
no effort was madeto studythe impactof surfacefrictionon known that friction at the lower boundary plays a crucial role
storm structure, particularly at low levels. in the dynamicsof tornadoes[Rotunno, 1979;Snow, 1982;
Wicker[1990], in a similarmannerto Klemp and Rotunno Fledlet and Rottmno, 1986]. Therefore it is necessary to
but with a higher resolution(70 m horizontalfine-grid include the effects of surface friction to study the complete
resolution),examinedthe fine-scalefeaturesof a supercell processof tornadogenesis.
stormon April 3, 1964,that was originallysimulatedby The simulationsreported in this paper were made with a
Wilhehnsonand Klemp [1981]. The resolutionnear the new numericalcloud model having a stretchedvertical grid
groundwasincreased to 50m by usinga verticallystretched and a nested horizontal grid in order to resolve storm
grid.The numericalsimulation havingsurface frictionat the features on scales of several hundred meters. It provides
lowerboundary produces thestrongestlow-levelrotation.A higherresolutionthan usedby KR andextendsthe approach
singleconcentrated vortexwitha maximum vorticityvalue of Wicket'[1990]by incorporatinga fully interactivenested
of 0.35 s-• formswithin the mesocyclone.This vortexlasts grid. To shed light about the evolutionof the tornado
for severalminutesbefore dissipating.A comparison be- cyclonewithin the storm, a supercellsimulationwas done
tweena simulation havinga free-sliplowerboundary andthe usinga frictionlesslower boundary.A secondsimulation
simulationhavingsurfacefriction showsthat the vortex was then made with a surface-friction parameterization to
investigatethe role of frictionin creatingthe tornado.These
TheTornado'Its Structure,Dynamics,Prediction,
andHazards. two numerical simulations will be used to discuss the gener-
GeophysicalMonograph79
Copyright
1993by theAmerican
Geophysical
Union. ationof stronglow-level rotationduringtornadogenesis.
76 TORNAD()GENESIS WIrI'HIN A SUPERCELL THUNDERSTORM

In order to clarify the discussion,we adoptdefinitionsof where V is the ground-relativewind speedaveragedto the
tornadic-storm features which are consistent with those used appropriatestaggered positionsof u• (x component)andu2
in observational
studies.Lemonand Dos•vell[1979]givean (y component)windsand AZ is the verticalgrid spacingat
overview of some of these features such as the divided
the lowestmodellevel. The friction term is appliedonly to
mesocycloncandthe rear flankdowndraft.The mesocyclone the horizontalvelocity variablesat the lowestgrid level. For
is associatedwith a strongrotatingupdraft with the vertical the free-slip simulation, Cd = 0.0. The surface-friction
vorticityexceeding
0.01 s--•, basedon Dopplerradaranal- simulationusesa value of C• = 0.04. This value is similar
ysis [Brandes, 1984;Brandes et at., 1988]. A few casesexist
to that usedby Wilhelmsonand Chen [1982] and Droege~
where the velocity field of a large tornado or its parent
meier[1985]. A sensitivitytest showedthat usinga value of
circulation was directly measured by the Doppler radar.
Cd = 0.004 changed the solution very little from the
Maximum shear values from these tornado cyclones are
free-slip solution.
typicallybetween().! and().2s • [Bra•tdes,1981'Ray et at.,
19811. By using these measurements as a basis, tornado
cyclones in the simulations are denoted as shear regions 2.2. Adaptive Grid Methodology
havingverticalvorticitygreaterthan0.125s • In order to capture fine featuresin their supercellsimula-
tion, KR used interpolated coarse grid model data (1000 m in
2. C()M PI.ITA'!'It•NAI.
the horizontal) at the time of peak low-level rotation to
2.1. Numeri•'al Model initialize a small domain with fine resolution (250 m in the
The new cloud model ix similar in design and construction horizontal). The model was then integratedfor a period of 6
t(• the Klcmp-Wilhelmson clot•d model [Kiemp and Wilhelrn- rain using the KW78 open lateral boundary conditions.
.s'•m, 1978a] (hereafter referred to as KW78). A new lkmture Interpreting the initial evolution on their fine grid is difficult
in the cloud model is the capability to integrate horizontally due to flow adjustmentsin responseto decreasesin turbulent
nested grids usingthe adaptive grid techniquesdescribed by mixing coefficients(which are based on the grid size). With
Skatnaro('k and Klemp 11989]. Vertical grid stretching en- the two-way interactive grid system used here, a fine grid
ables increased vertical resolution near the lower surface. mesh can be introduced into the calculation 10-15 min before
Other dill'erencest¾omthe KW78 model are listed briefly the maximum low-level rotation occurs in the simulation,
below. and integration on both the coarse-mesh and fine-mesh
1. The scalar variables are now integrated using a tbr- domains can be carried out indefinitely. Thus adjustments
ward time scheme. Scalar advection is computed using a due to a reduction in grid size occur well before the maxi-
second-order upwind monotonic advection scheme Ivan mum low-level rotation develops. Klemp and Rotunno
l,eer, 1977; Wicker, 1990]. [19831 and Wicker [19901 could not do this because the
2. The turbulent mixing coefficient is computed using a boundary conditions for the high-resolution grid were deter-
diagnostic t2)rmulation [('lark, 1979; l)'ipoli and Cotton, mined solely by flow inside the small domain.
19821. Experiences with nested-grid simulations show that the
3. A geometric mapping is used to stretch the grid solution collapses to the smallest scales which can be
vertically [ Walko, 1988]. representedon the fine grid. This phenomenonis caused by
The upwind monotonic advection scheme is more accu- the use of a grid-dependent mixing length in the calculation
rate than the centered time-space scheme used by KW78 of the turbulent mixing coefficient [Skamarock and Klemp,
[Wicker, !990]. The upwind scheme has smaller phase errors 1989]. The philosophy employed here is that fine features
than the centered scheme, and the monotonicity constraint should be well resolved. Thus a method is needed to prevent
prevents the generation of spurious negative values of non- the solution from collapsing to the smallest resolved scales
negative scalar quantities in regions containing sharp gradi- representedon the grid. We accomplish this by imposing a
ents. The turbulent mixing scheme was changed to the constant (not dependent on grid size) mixing coefficient on
diagnostic formulation because of the difficulty in imple- all grids. Experiences with modeling other types of small-
menting the prognostic approach used by KW78 in a model scale convective flows indicate that a constant mixing coef-
containing nested grids. Geometric vertical grid stretching ficient
oftheorderof 1ms-• •f, whereAfisthehorizontal
method was preferred to the scheme used by Wilhelmson grid spacingof the finest grid, is sufficientto force the
and Chen [ 1982]because it permits higher vertical resolution solution to be well resolved on the finest grid. The simula-
near the ground at a lower computational cost. tionspresentedhere use a constantmixingcoefficientof 100
A simple bulk parameterization for surface friction was m2 s-• onlyon perturbation
of variables,sothattheinitial,
also added to the model physics. This parameterization is horizontally homogeneous base state (winds, temperature,
identical to the formulation used by Wilhelmson and Chen and moisture) is preserved. We did several tests which
[1982] and Droegerneier [1985], i.e., showed that supercell evolution is insensitive to modest
changesin this parameter.The constantmixingcoefficientis
Ou• Ou2 _ CdV used in addition to the grid-dependenteddy viscosity pro-
-•(u•, u2); (1)
Or' •t '/fri
c vided by the turbulence parameterization
 WICKER AND WILHELMSON 77

-70

300

-60
400

/ I I t, zx, 500

/ "• l / / ' 700

ß' " z' ,( ' z• ,• ." • z' • ' 1000

-30 -20 -10 0 10 20 30

Fig. 1. (a) Skew-temperatureLog-pressureplotof theinitialverticalprofileof temperature

andmoisture usedto
initialize
themodel'sdomainforthesupercell
simulation.
Heavylineindicates themoist adiabat
forsurfaceparcellifted
tofree-convectionlevel.Mediumlineisthetemperatureprofile.
Dashed lineisthemoisture profile.
(b) Hodograph of
windsusedto initialize
themodel'sdomain.Axislabelsareinunitsofmeters persecond. Heightof profileisindicated
in decametersnext to profile. Arrow indicatesstorm motion.

3. SIMULATION RESULTS 3.1. Coarse-ResolutionSimulation(SupercellSimulation)

The model is initialized with a thermodynamicprofile The coarse-grid domain is a 75 x 75 x 16 km box.
(Figure 1a) similarto the one usedby Weismanand Klemp Horizontal resolution is 600 m. Vertical resolution in the
[1982, 1984]. The vertical profile of the horizontal wind lowest grid volume is 120 m and increases to 200 m at 1 km
(Figure lb) is a composite obtained from examining and then to 700 m at 7.5 km. Above this level it remains
hodographs from the Binger, Oklahoma, tornadic storm constant. There are seven grid levels below 1 km in the
[Wicker et al., 1984], the Raleigh, North Carolina,tornadic simulation.The storm is initiated by placinga "standard"
storm [Davies-Jones et al., 1990], and the Davis, Oklahoma, thermal bubble, 4øK in amplitude, at z = 1.5 km in the
tornadic storm on April 19, 1972 [Brown et al., 1973]. The center of the domain. The horizontal "radius" of the bubble
environment has a convective available potential energy of is 10 km, and the vertical "radius" is 1.5 kin. The Coriolis
approximately 2900J kg-• anda bulkRichardson number of force is set to zero for the simulations, since it has little
23 [Weisman and Klemp, 1982;Moncrieff and Green, 1972]. impact on the solution [Klemp and Wilhelmson, 1978b].
Calculationsusing the storm motion observedin the simu- The lower boundary condition is free slip. The model is
lationyielda stormrelative totalhelicityvalueof580m2s-2 integratedon the coarsegrid for 120 min. During the first 30
[Davies-Jones et al., 1990]. Similar values of helicity have min, the initial impulse grows and splits into left- and
been observed in environments of supercell storms that right-movingcells. After the split, the right-movingcell
spawnedsignificanttornadoes[Davies-Jones et al., 1990]. dominates,always containingthe largestvertical
78 'I•()RNAI)()GENES[S WITHIN A SUPERCEl,I_. THUNDERSTORM

Fig. !. (continued)

n•in cch{•, and cyclonic rotati()n. Maxim•rn updrafts during horizontally nested line-mesh grict is introduced at 70 min.
the firsl 90 rain exceed 55 m s •. Al'tcr II0 min the The model is then integrated tk)rward in time on both the
right-moving storm weakens somewhat, and its low-level coarse-mesh and the fine-mesh grids in order to study
rotation decreases. smaller-scale features that are unresolved on the coarse grid.
Horizontal slices of vertical velocity at z. = 1.0 km and the
surface rainwater mixing ratio are shown for 70 (Figure
3.2. b'ine-Resolution Simtdation (Tornado Cyclone
and 90 (Figure 2b) min (for convenience, z. = 60 m, the first
Sin.•lation)
vertical grid level for all variables but vertical velocity, is
referred to as the surface). As a reference, the -1.0øK The fine mesh is 15 km square in horizontal section and is
potential temperature isotherm, representative of the surface centered on the low-level mesocyclone. This domain is
gust-front boundary, is shown as the cold-front symbol. nearly large enoughto contain the convectively active part
Between 70 and 90 rain, updrafts and downdrafts at z, = 1.0 of the storm. Horizontal resolution of the fine domain is 120
km increase. The updraft at z. = 1.0 km develops into a m, while the vertical resolution remains the same as in the
horseshoe shape, and the mesocyclone divides into regions coarsegrid (e.g., zXzat the surfaceis 120 m). The integration
of updraft and downdraft. The surface rainfall field is some- on both the fine-mesh and coarse-mesh domains continues
what disorganized at 70 min and then evolves into a classic from 70 min to 110 min. During this period, two intense
•hook"-echo shape at 90 min. Aloft, a persistent bounded vortices develop within the mesocyclone.Each vortex lasts
weak-echo region, 5 km tall, develops, while the surface for approximately 10 min, is approximately 1 km in diame-
from0.038s-'• to morethan0.057 ter, andhasverticalvorticitygreaterthan0.125s-• fromthe
verticalvorticityincreases
s-•. These vorticityvaluesare indicativeof a stronglow- surfacethrougha depth of severalkilometers. The size and
level mesocyclone.The development of a divided mesocy- intensity of these vortices are similar to tornado cyclones
clone, hook echo, rear flank downdraft, and the increases in and are referred to as such.
low-level rotation are consistent with observational indica- During the first 10 min of integration on the fine grid,
tors of tornadogenesis [Lemon and Doswell, 1979]. To severalsmall-scalehorseshoe-shapedupdraftsdevelop along
capture this evolution using higher spatial resolution, a the western edge of the storm's low-level updraft
 WICKER AND WILHELMSON 79

10 15

0 5 10 15
x (,kin)

Fig. 3. (a) An 80-minhorizontalsectionof verticalvelocityat z =

Fig. 2. (a) Horizontal section of coarse-grid solution at 70 min. 1 km withinfine-griddomain.Contourintervalis 5 m s-t. The zero
Contour plot of vertical velocity at z = 1 km with a contourinterval contour is not plotted. Solid region indicates area where vertical
the0.25# kg-• surface- vorticityat z = I km is greaterthan 0.125 s-•. Peak vertical
of 3 m s-1. Heavysolidlinerepresents
rainwater mixing ratio contour at z = 60 m. Cold-frontalboundary vorticityat this level is 0.125 s-•. "A" is the northernhorseshoe-
locates the position of the -IøK potential temperatureisothermat shaped updraft described in text. (b) An 80-rain horizontal wind
thesurface.
Region
wherevertical
vorticity
greater
than0.03s-t at vectors at surface. Solid region indicates area where vertical vor-
ticityat surfaceis greaterthan0.125s-•. Peakverticalvorticityat
surfaceis indicatedby the soliddot. Peakverticalvorticityis 0.038
s-2. (b) SameasFigure2a exceptat90min.Peakverticalvorticity this levelis 0.125s-1. Heavylineis the0.01s-1 vertical~vorticity
is now 0.06 s -•. contour that delineates the
8t) TORNAI)OGENESIS WITHIN A SUPERCELL THUNDERSTORM

3a). The mtsocyclone, located along the shear zone created

by the forward and rear flank gust fronts, is elliptical (Figure
35). The regionof vorticitygreaterthan 0.125 s-I is a
column which tilts to the north between the surface and z =
I kin. A more detailed examination of the vorticity field (see
Figure 7c) shows that the region of enhanced vertical
vorticity extends upward to only 1 km in height. The shallow
vertical extent of enhanced vorticity suggeststhat this fea-
ture may be analogous to a "gustnado" [Wilson, 1986].
From 80 to 88 min the areal coverage of vorticity greater
than (}.125s • doublesin size, and an intensecirculation
dcvcl{)ps
through
a depth
ofseveral
kilometers.
At 88min
the horseshoe-shaped
updraftat z = I km expandsto 4 km
across and extends from the tip of the occlusion updraft
castward to the gust front updraft (Figure 4a). Just east of
the t{•rnaclocyclonecenter (herealter denotedas TC), an
intensed•wndral'tgreaterthan -25 m s -• is present.Up-
draftsarc 5 m .,, • .justto the southof the vortexcenter.At
the surface :,tnintense circulation is present (Figure 4b). The
'l'( • is ltl the southern end of the mesocyclone. It is defined by
theregiont•l'vorticitygreaterthan0.125s ......
l andis720m in
diamctcr.'l'hc maximumvt)rticityexceeds0.25 s'-•. Hori-
xontalwindsm'egrea. tcr than4(Im s • (ground-relative
flow) 0 5
x (kin)
10 15

It) thewest½>t'thc
'l'(• andexceed53m s -• alongthesouthern
edge el'the •I'C. The central pressureof the TC is -16 mbar,
and horizontalpressuregradientsat the surfitceare approx-
imately14mbarkm }
The first 'I'C decays between 90 and 95 min in the
simulation.The rapid decay occurswhen the first TC moves
s()uthcastward away t¾om the main storm updraft into a
region of strong downdratt. The mesocyclone becomes
somewhatdisorganizedt'ora few minutesand then redevel-
ops in an elliptical shape (Figure 5b). '/'he developmentof
the second 'l"C begins when a downdraft develops on the
westsideof thecirculation
centerandcoolair spreads
toward the east. The out/low increases the low-level conver-
gencealongthe westernsideof the mesocyclone.Horizontal
windsgreater
than55 m s • (ground
relative)
arepresent
there. The increased convergence stretches the vertical
vorticityalongthe westernsideof the mesocyclone, produc-
-1
ing the elongatedregionsof vorticitygreaterthan 0.125 s-1
seen at the surface and at z = 1 km (Figure 5). The -5 rn s
downdraft behind the rear flank gust front in Figure 5a is
associated with the first TC. The developing TC is located
3-4 km northwest
of thisposition
wherethe storm'smain
updraftis reintensifying.
The next 7 min showcontinued
intensification of the circulation at the surface and at higher
Fig. 4. (a) An 88-minhorizontalsectionof verticalvelocityat z =
levels. 1kmwithinfine-grid
domain.
Contour
interval
is5 ms-•. Thezero
The most intense stage of the secondTC (at 103 min) contouris not plotted.Solidregionindicatesarea wherevertical
occurs vorticityisgreaterthan0.3s-• (Figure vorticityat z = 1 km is greaterthan0.125s-•. Peakvertical
whensurface
6). At z = 1 km (Figure6a), theTC hasan averagediameter vorticityat thislevelis 0.20s-• . Strong-downdraft
regioneastof
tornadocycloneisindicated by "D". (b) An 88-minhorizontal
wind
of 840 m and is located in the maximum vertical-velocity vectorsat surface.Solid regionindicatesarea where vertical vor-
gradientbetweenthe updraftanddowndraft.Updraftsalong ticityat surface isgreater
than0.125s-•. Peakverticalvorticity
at
the westernsideof the TC are 5-15 m s-•, anddowndrafts thislevelis 0.25s-•. Heavylineis the0.01s-• vertical-vorticity
along edgeareabout-5 m s-l . Themagnitudecontour that delineates the
theeastern
of the downdrafteast of the cyclonecenter is muchsmaller
for the secondTC than for the first one (see Figure 4a). The
 WICKER AND WILHELMSON 81

15- 15-

\
\

,-- \

lO

/,.-'\ 1! q
t/ \,..,,\ k-5'

0 ,

0 o 5 10 15
X (kin)

15-

ß ,, b

0{::,::,::,::,::,::,::,::,::,',
0 5 10
, • , ,• .,

15 0
..............
5
: ..........
10 15
x (km) X (kin)

Fig. 5. (a) A 96-rainhorizontalsectionof verticalvelocityat z = Fig. 6. (a) A 103-rainhodzontalsectionof verticalvelocityat a =

1kmwithinfine-grid
domain. inte• is5ms-i. Thezero 1kmwithinfine-g•ddomain.
Contour Contour
interval
is5 ms-•. Thezero
areawhereve•ical contouris not plotted. Solid regionindicatesarea whereve•ical
contouris not plotted.Solidregionindicates
than0.125s-]. Peakve•ic• voNicityat z = i km is greaterthan0.125s-l. Peakve•ic•
vo•icityat z = 1 kmis greater
vorticity
atthislevelis0.17s-l. TC1andTC2indicate thepositionsvo•icityat thislevelis 0.24 s-1. (b) A 103-rain horizontalwind
of thedecaying anddeveloping tornado cyclones, (b) vectorsat surface.Solid regionindicatesarea whereveNicalvor-
respectively.
A 96-min hofizont• wind vector at surface.Solidregionindicates ticityat sugace isgreaterthan0.125s- 1. PeakveNic• voNicityat
areawhere verticalvorticity
atsurface is•reater than0.125 thislevelis 0.3 s-•. Heavylineis the0.01s-• ve•ical-vorticity
s-•. contour
Peakve•icalvo•icityat thislevelis0.20s-'. Heavylineisthe0.01 that delineates the
s-] ve•ical-vorticity
contourthatdelineates
themesocyclone.
82
TORNAI)O(iENESIS WITHIN A SUPERCELL THUNDERSTORM

MaximumVerticalVelocity

1_--2
•------: .................
-----:------___x•
s..•....•__•__ ß.'•,-•-•• .•'
' '-'
.'•
'--:.-......'•
."•.. ß
70 75 80 85 90 95 1O0 105 110
Time in Minutes

Minimum Perturbation Pressure

gI .... 2/ • •• _4•• -2• ••

•- '• •-16:.i :, :::i:.-•o::':.•ii

,, .... ,, ,'->,,..
70 75 80 85
,/ 9(1
:%
95 100 105 110
Time in Minutes

MaximumVerticalVorticity

70 75 80 85 }0 95 100 105 110

Time in Minutes

Fig. 7. Time-height
cross
sectionofvariablesfromfine-grid
simulation
using
free-slip
lowerboundary
conditions.
(a)
Maximum updraft.Contourintervalis 5 m s-l. Heavydashedlineindicates
updraftpulsesthatoccurnearthe
mesocyclone.(b) Minimumpressure.Contourintervalis 2 mbar.(c) Maximumverticalvorticity.Contourintervalis
0.03s-t. Regionswhereshearis greater
than0.21s-• are
 WICKER AND WILHELMSON 83

genesisof the second TC is similarto the firstone;the minutesa 2-kin-deepregionof enhancedvorticity develops
surfacemesocyclone contracts
indiameter andchangesfrom aloft. One minuteafter maximumvorticity aloft, the peak
anelliptical
toa circular
shape.TheTChasa vorticity core surfacevorticityof 0.26s-• occurs.Vorticitygreaterthan
with nearlyverticalorientationbetweenthe surfaceand 1 0.2 s-• persistsnear the surfacefrom 88 to 91 min. The
km. Peakground-relativewindspeeds nowexceed55 m s-• . developmentof strong rotation aloft with a subsequent
In orderto examinemorecompletely thechanges in the extensiondownwardto the surfaceis similarto Doppler
storm'sstructurein time and space,a time-height cross radar observationsof the Union City tornadocycloneevo-
sec,Lion
iscreated
ofthemodel
datafromthefinegridduring lution [Bro•,n et al., 1978]. In contrast, the second TC
the 40-minintegrationperiod(Figure7). Maximumor mini- develops strong rotation nearly simultaneouslyfrom the
mum valuesof the verticalvelocity,pressure,andvertical surfaceto 2 km at 99 min. During the period from 99 to 107
vorticity at each horizontal level were used to create the minthedepth,overwhichverticalvorticityexceeds 0.2 s-•
cross sections. The values were checked to make certain increasesfrom 2 km to 4 kin. Peak vorticity at the surfaceis
that their locationsare in or very nearthe mesocyclone0.31s-• at 103rain,andverticalvorticitygreaterthan0.2
circulation.A 1-2-1filter was appliedin bothdirectionson s-• persistsnearthe surfacefrom99 to 110min.
eachvariableto removenoisefromthe fields.Duringthe The secondTC's evolution, with strongrotation develop-
modelintegrationthe domainaverageof the perturbation ing at low levelsand then extendingrapidly upward into the
pressurefieldis forcedto be zeroat eachtimestepbecause storm, is very similar to the Del City mesocyclone and
the perturbationpressurefieldis determined onlyto within tornado signatureobserved on May 20, 1977. Figure 8, from
an arbitrary constant in the cloud model. Examination of the Johnsonet al. [1987], is a time-heightcross sectionobtained
actual pressure fields show that falls (rises) are associated from Doppler observations of the Del City storm. The
with increases(decreases)in the horizontalgradientof the observed maximum vertical-vorticity value at each level
pressurefield. For example, at 100 min the lowest surface within the storm is shown in Figure 8a. At 1830 CST,
pressureis -22 mbar, and the TC hasa horizontalgradient vertical vorticity rapidly increasesfrom the surface through
of 22mbarkm- •' at93rainthecorresponding valuesare-7 8 km. From 1830 CST to 1845 CST the low-level vorticity
mbar and 6 mbar km -• . continues to increase between the surface and z = 4 km.
Severalprominentfeaturesare apparentin the time-height Tornado touchdown is at 1845 CST. Figure 8b shows the
cross section of the maximum vertical velocity within the observed maximum updraft at each level within the Del City
storm (Figure 7a). There are four distinct updraft pulses storm. Beginning at 1830 CST, updrafts greater than 40 m
between 70 and 105 min. These pulses are of two distinct s-• develop
rapidlybetween3 and4 km priorto touchdown
types. The first and third updraftpulsesdevelopabove5 km of the Del City tornado. The developmentof an updraft pulse
in height near the level of maximum buoyancy. The in- prior to TC development in the Del City storm is qualita-
creases in updraft intensity aloft are not reflected at low tively similar to the rapid increase in 2-4 km vertical-velocity
levels. During the development of the second and fourth maxima around 97 min in the model simulation.
pulses, large upper level vertical-velocity increasesare pre-
ceded by increases in low-level vertical velocities. For
3.3. Su•;fhce-Friction Simulation (Tornado Run)
example, during the period between96 and 100min, updrafts
greaterthan 35 m s-1 existwithinthe stormat the 2-km Results from Wicker [1990] indicate that the inclusion of
level,andupdraftsgreaterthan45 m s-1 arepresentat the surface friction increases the low-level convergence and
3-km level. Further examination of data shows that the vertical vorticity within the tornado cyclone, compared to a
second and fourth updraft pulsesare located in the north- free-slip simulation. By using initial conditions from the
western portion of the mesocyclonenear the intersectionof tornado cyclone simulation on the fine grid, surface friction
the forward and rear flank gust fronts. The first and third is added to the model physics. The surface drag increases
pulse are updrafts located southeastof the mesocyclone low-level convergence in the tornado cyclone, resulting in
alongthe southernflank of the storm.At any giventime, the the formation of a concentrated vortex, which we call a
northwestand southeastupdraftsare connectedby a region tornado. To conserve computational resources, the friction
of weakerupdrafts.At midlevels(z = 4-5 km) theyformthe simulation is started at 94 min, well at•ter the start of the
characteristichorseshoepattern describedby others (e.g., nested fine-mesh simulation. At 94 min the first TC has
Del City storm[Ray et al., 1981]andHarrahstorm[Brandes, nearly dissipated, and the intensification of the stronger,
1977]). second TC is 5 rain in the future. The new simulation is
The secondand fourth pulsesdevelopat the onsetof integrated from 94 to 114 min. After 112 min the tornado
pressurefallswithinthe mesocyclone (Figure7b). Minimum weakens considerably from its maximum intensity.
pressuresat low levelsoccurduringthe periodof maximum A time-height cross section is constructedfor the friction
updraft intensity. After the occurrenceof minimumpres- simulation for the period from 94 to 114 min (Figure 9).
sures,peaksurfacevorticityvaluesareattained(Figure7c). Similar structure in the updraft fields is presentbetween the
The developmentof the first and secondTCs are very friction and free-slip simulations (Figures 7a and 9a). Min-
different. Verticalvorticitygreaterthan0.2 s-• developsimum surface pressure during the most intense rotation (102
first at z = 2 km at 87 rain and during the next several min) is -16 mbar, with associatedhorizontal pressure
84 TORNAI)OGENESIS WITHIN A SUPERCELL THUNDERSTORM

MAX VORTICITY (IO•-5) MAY 20, 1977

14,

12IO.,,..
• l0 •OIO

i•30
I IO•!
17• 17•
I.,I•I I 18• 19•
TIME (CST)

Fig. 8a

- M,tLX REFLECTIVITY (dBZ)

5/20/77 --- MAX W
16

12

10

½./ /

1700 1730 1800 i e30 1900

Fig. 8b
Fig. 8. (a) Time-height cross section of maximum vertical vorticity for the Del City storm. Contour values are in units
of 10-3 s-1 . (b) Time-height
sectionof maximum
reflectivity
andmaximum
updraftspeeds
for the Del City storm.
Dashed line is the maximum updraft. Solid line is the maximum reflectivity. Arrows at top of the figure indicate Doppler
analysis times. Figures 8a and 8b were taken from Johnson et al.
 WICKER AND WILHELMSON 85

Maximum
Vertical
Velocity
9-

•35
2- • 35••30
- • 20 •20

0 ß • .... ''1
....".....'....'......
:......
'i'" ':' ' -:'"'-W-
•'-,
94 96 98 100 102 104 106 108 110 112 114
Time in Minutes

Minimum PerturbationPressure

9 b
8 -6 •4--------

ß
• 5 -8•

2 ,,,

-14
1
0 5 10
0 ..-4 x (km)

94 96 98 100 102 104 106 108 110 112 114

Time in Minutes

MaximumVerticalVorticity

?
. I
•4
• 3

0 •, ,

94 96 98 1• 102 1• 1• 108 110 112 114

T•e in Minutes

Fig. 9. Time-heightcross sectionof variablesfrom fine-gridsim-

ulation with surfacefriction. (a) Maximum updraft. Contourinter-
valis5 m s-1. Heavydashed
lineindicates
updraft
pulses
thatoccur
near the mesocyclone.(b) Minimum pressure.Contour interval is 2
-1
mbar. (c) Maximum vertical vorticity. Contour interval is 0.03 s
Regions
whereshearis greaterthan0.21s-] areshaded. 0 5 10 15
X (km)

Fig. 10. (a) A 102-minhorizontal sectionof vertical velocity at z

dientsof about16 mbarkm-• . Duringmostof thetornado = 1 km within fine-grid simulation with surface friction. Contour
event the pressureat z - 250 m is 1-2 mbar lower than the
interval
is 5 m s-1. The zerocontour
is notplotted.Solidregion
indicatesarea where vertical vorticity at z = 1 km is greaterthan
surfacepressure.During the period of peak rotationbetween 0.125s-t Peakverticalvorticityat thislevelis 0.25s-] (b) A
ß ,

101 to 102 min, the pressureis lowest at the surface.This 102-minhorizontal wind vector at surface. Solid region indicates
result is different from the free-slip simulationwhere the areawherevertical
vorticity
at surface
is •reaterthan0,125s-•
pressure is always lowest at the surface during the TC Peak
vertical
vorticity
atthislevel
is0.54sTM.
Heavy
lineisthe0.0i
s-• vertical-vorticity
contour
thatdelineates
the
events. Higher pressuresat the surface for the friction
simulation are the result of significantlylarger values of
low-level convergenceat the baseof the vortex than in the
 WITHIN A SUPERCELL THUNDERSTORM

l."ig. !1. R.adius-hcigl•!

sccli{m(•t'azimutlaally
averagedmtsocyclone
featuresat 103rainin thefree-slipsimulation.(a)
Radialvciocily.
C½>ntr•t•r
intervalis 2 m $ •. {h)Tangential
velocity
withcontour
intervalof 2 m s.......
1. ({.) Vertical
velocitywithct>i•to!,r
interval

free-slipv()rtcx. 'l'hc vcrtical-vorticity evolution (Figure 9(') t'rec-slipand surface-frictionsimulations.The center point
is very different t¾omthat in the free-slip simulation(Figure for the averagingat each level is taken to be the positionof
7•'}. Maximum surface vorticity valH½sarc 70% larger than the maximumvorticity valueat that level. Pressure,vertical-
in the free-slip simulation.However, the vertical extent of velocity, tangential-velocity, and radial-velocity values at
strongvorticity ix much shallower. Surface vorticity values each level are averaged into rings that are 120 m wide and
in the t'rictionrun exceed ().36 s • in the cross section, and extend outward 3 km from the maximum vorticity value.
unsmoothcdvorticityvaluescxcccd0.5s • between101and This averagingprocedureis similar to one usedby Brandes
102min. Verticalvorticitygreatertitan().2 s • duringthe [19771 and Wh'ker [199()}.
periodfrom 98 to 112rain extendst¾omthe surfaceto !.5 km The radial velocity for the free-slip simulationat 103 min
in height, comparedto a depth of 4 km in the free-slipcase. shows inflow below z = 2.5 km and outflow aloft (Figure
Figure 10 depicts the vertical-velocity field at z = I km 11a). Near the surface and between r = 1 km and r = 1.8
and the surface wind vector field at 102 min. Peak vertical km the flow is outward from the axis. This is a manifestation
vorticityis 0.54 s-• and maximumground-relative wind of the downdraft east of the tornado cyclone center (see
speeds are62 m s- • alongthesouthernedgeof thetornado. Figure1lc). The tangential-velocityfield (Figure1lb) hasa
At the surface the area of the mesocyclone is 2-3 times maximum velocityof 34 m s-• aloftat z = I km andhasa
smaller than in the free-slip case (see Figure 6b), and the value of 32 m s-• near the surfaceat r = 0.5 km. The core
wind field has a significantly stronger inward component. radiusof the tornadocycloneis nearlyconstantwith height
The vertical velocity at z, = I km (Figure 10a) has upward andextendsupwardto 4.5 km. The storm'smainupdraftis
velocities of 15 m s-• north of the tornado center. Weaker the dominant feature in the vertical-velocity field (Figure
updrafts of 5 m s-• existto thesouthandeastofthetornado, 11c). Vertical motionsare very weak in the center of the
so the tornado is positionedin the vertical-velocitygradient. tornadocyclonebelowz = 1 km. The azimuthalaverageof
Rather than being located between the mesocyclone'sup- the pressurefield(not shown)showsa downwardpressure
gradientalongthe centralaxis.This causesthe maximum
draft and downdraft as the tornado cyclone is in the free-slip
verticalvelocityto be displacedfrom the axis.
simulationat 103 min (Figure 6a), the tornado is completely
embeddedin updraft. At 1 km, no significantdowndraftsare The radial-velocityfieldfor the surface-friction
simulation
presentwithin the mesocyclone.At z = 180 m, vertical at 102 min indicates stronger low-level inflow than in the
velocitiesalongthe northernsideof the tornadoexceed19 m free-slipsimulation (Figure12a). Betweenthesurfaceandz
S
-1
. = 200m, radialinflowexceeds
14m s-•. Abovetheinflow
To comparethe differencesin the structureof the meso- layerandcloseto theaxis,weakoutflowexistsfromz = 1.5
cyclone,tornadocyclone,andtornadobetweensimulations km to z = 5 km. The tangential-velocity plot (Figure 12b)
an azimuthalaverageof model output is computedfor the showsthat the tornadocyclonehasa largerdiameter
 WICKER AND WILHELMSON 87

f ,..x_•/b
s 4
c
,

0 1 2 3 0 1 2 3
Radius(kin) Radius(krn)

Fig. 12. Radius-heightsection of azimuthally averaged mesocyclonefeatures at 102 min in the surface-friction
simulation.(a) Radialvelocity.Contourintervalis 2 m s-1 . (b) Tangential
velocitywithcontourintervalof 2 m s-1
(c) Verticalvelocitywithcontourintervalof 2 m s-•

z = I km than in the free-slip simulation, while near the The high-resolution storm simulation using free-slip
surface the radius of maximum winds has contracted to a
boundary conditions develops two separate tornado cy-
smaller core radius of 250 m. Maximum tangential velocities clones over a 40-min period. The first tornado cyclone
are 36 m s-1 and are confinedto a shallowlayernearthe develops aloft and then extends to the surf'acein a similar
surface. The vertical-velocity field (Figure 12c) has updrafts manner to the one in the Union City storm [Brown et al.,
of 10 m s-• near the axis belowz = 1 km, witha slight 1978]. Maximum surface wind speeds exceed F2 intensity
downdraft(-2 m s-•) on the centralaxisat z - 600 m. during both events. The second tornado cyclone is stronger
The results described in Figures 11 and 12 are consistent than the first and maintains its intensity over a 10-rain
with the results from Howells et al.'s [1988] axisymmetric period. Unlike the first tornado cyclone, the intensification
vortex simulations. The inclusion of surface friction creates of the second circulation occurs simultaneously from the
significantradial inflow into the baseof the tornadocyclone. surfaceto 2 km in height. This evolution is similar to the Del
At low levels the radial inflow transportsangularmomentum City tornado cyclone [Johnson et al., 1987].
closer to the axis than in the free slip simulation,resultingin Time series analyses of the model output show that the
highertangentialvelocitieswithinthetornadoat the surface. developmentof the two tornado cyclones within the mod-
An increase in low-level convergence •s associatedwith eled storm is preceded5-7 min by intensificationof updrafts
strongerupward motion in the tornado.At higherlevels, to magnitudes
of 30-40 m s-• at an altitudeof 3-4 km. A
radial outflow is presentcloserto the axis in the friction
similarphenomenonis noted in the updraft evolutionof the
simulationthan in the free-slipsimulation.This flow diver-
Del City storm [Johnsonet al., 1987]. Simultaneouslywith
gence causesthe core radius for the tornadocycloneto the updraft intensification, rapid pressure falls develop
increasewith heightand explainsthe reductionin vertical within the mesocyclonethrough a depth of 4 km. After the
vorticity abovez = 1 km in the friction simulation. developmentof these low-pressurecores, maximumvortic-
ity values are observed within the storm, with maximum
4. SUMMARY values at the surface and somewhat smaller values aloft. The

Numerical simulationsof tornadogenesiswithin a super- developmentof these low-pressureregionsand the intensi-

cell thunderstormare carried out usinga three-dimensional fication of updrafts at 3-4 km appear to be linked together.
nested-grid numerical modelwithveryhighhorizontal and Further analysesare being done to understandthis relation-
vertical resolution in the low-level mesocycloneregion. ship.
Simulations lowerbound- To assessthe impact of surfacefriction on the tornado
usingfree-slipandsurface-friction
aryconditions
arecompared theimpactof surface cyclone,the fine-gridsimulationis restarredfrom 94 min
to assess
friction on the fine-scale structure of the storm. with surface friction added. An intense vortex core
88 TORNADOGENESIS WITHIN A SUPERCELL THUNDERSTORM

many characteristics of tornadoes such as strong low-level Klemp, J. B., and R. Rotunno, A study of the tornadic region within
radial inflow, large vorticity, and large low-level updrafts a supercell thunderstorm, J. Atmos. Sci., 40, 359-377, 1983.
develops. At low levels the diameter of the vortex is approx- Klemp, J. B., and R. B. Wilhelmson, The simulation of three-
dimensional convective storm dynamics, J. Atmos. Sci., 35,
imately halt'the size of the vortex in the free-slip simulations. 1070-1096, 1978a.
Surface wind speedsare 10--15%larger, and low-level up- Klemp, J. B., and R. B. Wilhelmson, Simulations of right- and
drafts around the tornado are 5 times greater at low levels left-moving storms produced through storm splitting, J. Atmos.
than in the free-slip simulation. These changesin the low- Sci., 35, 1097-1110, 1978b.
level vortex structure due to the inclusion of surface friction Klemp, J. B., R. B. Wilhelmson, and P.S. Ray, Observed and
are similar to previousstudiesof axisymmetrictornadolike numerically simulated structure of a mature supercell thunder-
storm, J. Atmos. Sci., 38, 1558-1580, 1981.
vortices [Howells et al., 1988]. Further analysis of the Lemon, L. R., and C. A. Doswell lil, Severe thunderstorm evolu-
reportedsimulationsis underway,and new higher-resolution tion and mesocyclone structure as related to tornadogenesis,
simulationsare phmnedto investigatein detail the genera- Mort. Weather Rev., 107, 1184-1197, 1979.
(ion and maintenance of tornadoes. Moncrieff, M. W., and J. S. A. Green, The propagationand transfer
properties of steady convective overturning in shear, Q. J. R.
Meteorol. Soc., 98, 336-352, 1972.
Ackno•'h'dt,tncnts. The authors wish to thank Rich Rotunno, Ray, P.S., B.C. Johnson, K. W. Johnson,J. S. Bradberry, J. J.
Jerry Straka, Bill Skamarock, Brian Fiedior, Harold Brooks, Don Stephens, K. K. Wagner, R. B. Wilhelmson, and J. B. Klemp,
B•rgc•,s, Brian Jewell, and Robert Davies-Jonesfor many useful The morphologyof several tornadic storms on 20 May 1977,
disct•ssionsover ihc pastseveral yearsabout this work. The authors J. Atmos. Sci., 38, 1644-1663. 1981.
also wish It) thank the anonymous reviewers and the editors who Rotunno, R., A study in tornado-like vortex dynamics, J. Atmos.
were very hclpt'ulwith their constructivecriticismsof the paper. Sci., 36, 140-155, 1979.
'!'hiswork was supportedby the National Science Foundationunder Skamarock, W. C., and J. B. Klemp, Adaptive modelsfor 2-D and
the NSt" grant A'I'M-87-()0778 and by the National Center for 3-D nonhydrostaticatmosphericflow, in 6th International Con-
Supercomputing Applicationsat the Universityof Illinois. Jkrenceon Numerical Methods in Laminar and TurbulentFlow,
pp. 1413-1424,Pineridge,Swansea,Wales, 1989.
REI , Snow, J. T., A review of recent advancesin tornadodynamics,Rev.
Geophys.Space Phys., 20, 953-964, 1982.
Brandcs, E. A., Mtsocyclone cw)!uti(m and tornado generation Tripoli, G. J., and W. R. Cotton,The ColoradoState University
within the ttarrah, Oklahoma ,,tOt'ha,NOAA Tech. Memo. ERL three-dimensional cloud/mesoscale model•1982, I, The general
NSSL-S1, 28 pp., Natl. Severe Storms i•ab., Norman, Okla., theoreticalframework and sensitivityexperiments,J. Rech. At-
1977. mos., 16, 185-220, 1982.
Brandes, E. A., Finestructure {)f the Del City-Edmond tornadic van Leer, B., Towards the ultimate conservativedifferencescheme,
mesocirculation, Mon. Weather Rev., 109, 635-647, 1981. IV, A new approachto numericalconvection,J. Cornput.Phys.,
Brandes, E. A., Relationships,between radar derived thermody- 23, 276-299, 1977.
namic variables and tornadogenesis,M•m. Weather Rev., 112, Walk(), R. L., Plausibilityof substantialdry adiabaticsubsidencein
!(133-1052, !984. a tornado core, J. Atmos. Sci., 45, 2251-2267, 1988.
Brandcs,E. A., R. P. Davies-Jones,and B. C. Johnson,Streamwise Weisman,M. L., and J. B. Klemp, The dependenceof numerically
vorticity effects on supercel!morphologyand persistence,J. simulated convective storms on vertical wind shear and buoy-
Atmos. Sci., 45,947-963, 1988. ancy, Mon. WeatherRev., 110, 504-520, 1982.
Brown, R. A., D. W. Burgess,and K. C. Crawford, Twin tornado Weisman,M. L., and J. B. Klemp, The structureand classification
cycloneswithina severethunderstorm,Weatherwise,26, 63-71, of numericallysimulatedconvectivestormsin directionallyvary-
1973.
ing windshears,Mon. WeatherRev., 112,2479-2498,1984.
Brown, R. A., L. R. Lemon, and D. W. Burgess,Tornadodetection Wicker,L. J., A numericalsimulationof a tornado-scale
vortexin a
by pulsedDopplerradar, Mon. WeatherRev., 106, 29-38, 1978. three-dimensionalcloud model, Ph.D. thesis, 264 pp., Univ. of
Clark, T. L., Numerical simulationswith a three-dimensionalcloud I11., Urbana-Champaign, 1990.
model: Lateral boundaryconditionexperimentsand multicellular Wicker, L. J., D. W. Burgess,and H. Bluestein,The pre-storm
severe storm simulations,J. Atmos. Sci., 36, 2191-2215, 1979. environment of the severe weather outbreak in western Oklahoma
Davies-Jones,R. P., D. W. Burgess, and M.P. Foster, Test of
on May 22, 1981,in Preprints,13thConference
onSevereLocal
helicityas a tornadoforecastparameter,in Preprints,16thCon-
Storms,pp. 281-282,AmericanMeteorological Society,Boston,
ference on SevereLocal Storms,pp. 588-592, AmericanMeteo-
Mass., 1984.
rologicalSociety, Boston, Mass., 1990. Wilhelmson,R. B., and C. S. Chen,A simulationof the develop-
Droegemeier,K. K., The numericalsimulationof thunderstorm
ment of successivecells alonga cold outflow boundary,J. Atmos.
outflowdynamics,Ph.D. thesis,695 pp., Univ. of Ill., Urbana-
Sci., 39, 1446-1483, 1982.
Champaign, 1985.
Fiediet, B. H., andR. Rotunno,A theoryfor maximumwindspeeds Wilhelmson,R. B., and J. B. Klemp, A numericalstudyof storm
in tornado-like vortices, J. Atmos. Sci., 43, 2328-2340, 1986. splittingthat leadsto long-livedstorms,J. Atmos.Sci., 35,
1974-1986, 1978.
Howells, P. A. C., R. Rotunno,and R. K. Smith, A comparative
studyof atmospheric andlaboratory-analogue numerical tornado- Wilhelmson, R. B., andJ. B. Klemp,A three-dimensional
analysis
vortex models,Q. J. R. Meteorol. Soc., 114, 801-822, 1988. of splittingseverestormson 3 April 1964,J. Atmos.Sci., 38,
Johnson,K. W., P.S. Ray, B.C. Johnson,andR. P. Davies-Jones, 1581-1600, 1981.
Observations related to the rotational dynamics of the 20 May Wilson,J. W., Tornadogenesis
by nonprecipitation
inducedwind
1977 tornadic storms, Mort. Weather Rev., 115, 2463-2478, 1987. shearlines,Mon. WeatherRev., 114, 270-284,
 Tornado Spin-Up Beneatha ConvectiveCell'
RequiredBasicStructureof the Near-FieldBoundaryLayer Winds
ROBERT L. WALKO

Departmentof Atmospheric
Science,ColoradoStateUniversity,
Fort Collins,Colorado80523

1. INTRODUCTION tion is incompleteunless it also accountsfor the origin of

that weaker vertical vorticity. Wakimoto and Wilson [1989]
Tornadoesoccur in many sizes, shapes,and configura- documented a case where vertical vorticity along a quasi-
tions, but all have in common a central core of highly stationary convergenceline preceded convective develop-
organized, intense vorticity. The variousfundamentalfluid ment, and subsequent convection locally intensified the
dynamicalprocessescapableof producingsucha vorticity convergenceto the point of spinning up a tornado from the
field are well known. How those processesare actually backgroundvorticity field. In other cases it has been argued
brought about by the tornadic storm and its environment that the thunderstorm and its immediate environment inter-
remain topics of study. The present paper addressesthis act to generate a local field of vertical vorticity. For exam-
problem with the aid of three-dimensional (3-D) numerical
ple, Davies-Jones [1982] argues theoretically and cites ob-
simulations.
servational evidence that a rain-cooled downdraft may
The related topic of vertical vorticity in a convective cell
introduce vertical vorticity by tilting environmental horizon-
updraft has been discussed by Lilly [1982], Davies-Jones
tal vorticity, and Rotunno and Klemp [ 1985] describenumer-
[1984], Rotunno and Klemp [1985] and others and has
ical simulations in which solenoids at the edge of a rain-
apparently been adequately explained. Such vorticity is
cooled thunderstorm outflow generate horizontal vorticity,
theorized to arise from ambient horizontal vorticity present
in vertically sheared boundary layer winds, when horizontal which is subsequently tilted into the vertical.
gradients of updraft strength tilt some vorticity into the From the preceding discussion, we may identify several
vertical. Although it has been speculated that the vertical possiblegeneral mechanisms for tornadogenesis.
vorticity in a tornado core may arise from the samemecha- I. Weak to moderate vertical vorticity is present in the
nism, there is a basic weakness to this explanation, as prestorm environment and is concentrated into a tornado by
pointed out by Davies-Jones [1982]. He arguesthat if only horizontal convergence beneath the convective updraft.
updraft and no downdraft is causingthe tilting of vorticity, 2. Horizontal vorticity is present in the boundary layer
vertical vorticity remains zero close to the ground and is winds of the prestorm environment. It gets tilted into the
acquired by a parcel only as it rises away from the surface. vertical by a vertical motion field which contains no subsid-
Thus vorticity tilting by an updraft alone can accountfor ence.

rotation of a thunderstorm convective updraft above the 3. Horizontal vorticity is present in the boundary layer
planetaryboundarylayer but appearsto fall shortof explain- winds of the prestorm environment. It gets tilted into the
ing the vertical vorticity in a tornado very close to the vertical by a vertical motion field which contains some
ground. subsidence.
The more traditional and simpler explanation for the 4. The vorticity responsiblefor the tornado is generated
intense vorticity of a tornado core is that it becomescon- solenoidallyby the effects of the thunderstorm and is sub-
centratedfrom a larger horizontalarea of weakervertical sequentlytilted into the vertical.
vorticity, in some cases identifiableas a tornadocyclone The above list, as specificallyworded, does not describe
[Church et al., 1979], by the field of horizontal convergence all possiblemechanismsof generatingan intensetornadolike
beneaththe thunderstormupdraft. However, this explana- vortex, but it addressesthose mechanismsproposed in the
foregoing discussion. Mechanisms 2-4, which involve vor-
The Tornado:Its Structure,Dynamics,Prediction,andHazards.
GeophysicalMonograph79 ticity tilting, can each be divided into two subcategories.
Copyright1993by the AmericanGeophysical
Union. 1. Vorticity tilting results in only weak-to-moderate
89
 SPIN-UP BENEATH A CONVECTIVE CELL

ticalvorticity,whichmustbe furtherconcentrated
by hori-
zt•tal convcrgcncc to t'orm a tornado.
2. Ct)nccntratit)n
()f verticalvorticityby horizontalcon-
vergeneeis not requiredfor tornadogenesis. A tornadocore
can be producedby tilting horizontalvorticity of equal
intensity.
'l'hc prescnt paper explores possiblemechanismsof
n•c!(•gcncsisthrough a seriesof 3-D numerical simulationsof
•tn idealized storm and its environment. Each simulation
designed to t'()cuson one of the above mechanisms in order
tt• cx•tminc whether it shows any tendency toward tornadic
dcvclt•pmcnt. Although tornadocsoccur t•nctcra wide vari-
ety {'}1'circumstanccs,including apparently some instances
where strong but•yant convection aloJi is absent, wc centinc
:tltcnlit)n hurt tt• It•rnactocswhich form bcncatha. strong 1• •1[l,Xœ11111
i•liLii111Ll•i.
ii.ii, i•Li,iL.i.Li.[i,.LLi21Li
L2.
updraft, as in a supercell sit)I'm. in all simulations the
ct•nvcclivc updrali is generated and maintttincd in the center , (I •' m )

{•1'the ct>mputali{)nal d{)nlain,loc•tlcd•tt (.r, y) = ((), 0), by Fig. I. Initial low-level wind lield containingtwo region,,,,
of uni-
imp{)sing•t c()nst:tnt-in-timc
heat sourceof roughlycylindri- form velocitywith a shearzonein between.Windsdecreaselinearly
c•tl sh•pc it) modellatent heat release.The heatingrate and with heighl, reachingand remainingat zero above 2 kin.
the din•cnsi()ns()l'the heated vt)lumc approximate those
st•pcrccll stt>nn updraft, and Cluasi-stcadyupclral'lspeeds
:tl'()tllld 5()I11 S I arc ()blaincd t}vcr a 5()-k• 2 •trca. •l'hc ticgtco ot' axial symmetry and qualitatively resembles the
cnvirt•nnicntallapscratcis! K km • inlhc previous axisymmetric results.
linc dr)mltin •tnd !{) K km • ab{)vc. •I'hc individual simula- Ncxt, an experiment is initialized with a second type of
tit>ns ditl•r I'rt•m each t)thcr in the initial I()w-lcvcl ambient wind field which containsvertical vorticity: straight-lineflow
wind and temperature ticIris and in the prcscnccor absence with a region el' horizontal shear as shown in Figure 1.
•)l'•t!{)w to middlelevel hcat sink l{• approximatecvaporalivc (Except where noted, this and all other figures are plotted
ct•ling ()1'precipitation and a c{)nscqucnl cool d()wndrati. from fields on the intermediate nested grid. All "low-level"
•!'t•crcst•lts{•}t'these simulationsindicate which categories anct "middlc level" cross sections were taken from the 90-m
•tn•bicnl c{>nditi{•ns will intcracl with the c{•nvcction and 4-kin levels, respectively, but are each representative of
pr•>dt•cc a tt)rnadt)likc vortex and which will nt)t. For each fairly thick layers. )The shear zone contains equal amounts
catcgt)ry, tw{) t•r more distinct sin•t•lali{•nswcrc run. Except {•t'vertical vorticity and detbrmation, but the vorticity alone
l't•r cases which included •t heat sink, simt•lati{)ns within the is the crucial ingredient for vortex spin-up. This category of
samc catcgory yielded the same qualitative result. Ilow is intended to approximate the vertical-vorticity distri-
(.;o•nput•ttions are pertBrined with the Colorado State bution of the observed wind ticld prior to the July 2 Convec-
University Regional Atmospheric Modeling System {RAMS) tive Initiative and Downburst Experiment (CINDE) tornado
using a scries of three telescoping nested grids having lL/ttal et al., 1989]. As in the previous axisymmetric exper-
horizontal grid cell dimensions of 1600 m, 400 m, and 100 m. iment, the convectively driven low-level convergence con-
Vertical spacingon all grids beginsat 20 m near the ground• centrates ambient vertical vorticity in the center of the
stretches gradually to 500 m, and remains at 500 m up to the domain, producing a tornadolike vortex. Figure 2 illustrates
model top 20 km above the ground. All grids contain 50 x 50 the low-level horizontal wind vectors at 36 min into the
/ 54 mcsh points. A slightly coarser mesh with t•wer grid simulation when the strongest wind is obtained. A noticeable
points is used lYr some of the experiments. feature is that the winds 5-10 km away from the vortex are
asymmetric, consisting of two main inflow jets feeding the
2. RESULTS
vortex: an easterly jet to the north and northeast of the
center and a westerly jet to the south and southwest. These
•l•he first numerical experiments to be described belong to grew out of the initial two uniform-wind regions. Due east
the category in which vertical vorticity is contained in the and west of the vortex at the same distance, the winds are
initial low-level wind field. For comparison with previous light. Closer to the vortex center, the winds become much
vortex simulations using axisymmetric models [e.g., Hou'- more axisymmetric, a feature clearly seen in the winds of the
ells and Smith, 1983' Walko, 1988], I first choose as an initial highest-resolutionnested grid (Figure 3). Additional experi-
condition a field of solid rotation (uniform vertical vorticity ments were performed with similar initial wind fields, differ-
of the order of 100 times that of the Earth) centered on a ing in actual wind strength and in the width of the shear
vertical axis in the center of the domain. Horizontal conver- zone; all resulted in the spin-up of a concentrated vortex as
gence at low levels induced by the convection spins up a described above. ! note here that the presence of small
tornadolike vortex within 15 min. The flow retains a high closed circulations, apparently due to
 WALKO 91

10.13-
/ / / / / / ,.., .......
/ / / / ß / /., .......
........
I / / /

/ /' / w' / / ,.,•

.....-. •
i / / I ,/ .,...w....._ ..
. / ,, / / / ,,, ./ •,...........,- •
,. / / /,,,-..-.-._....,/..__.---.-.-,',NN
ß , J t ! d / .i ,,/•t.- ....- .,---' f I t t / / "'"'""'""'--
•_ I / / I / /./ ..-.....
N,,/ -'"'"'""•"'"
•-_rl i I I / /,•f .- . • l/
.... ' * l,/ /•.'•'---•• ,Ell/l/l//,, ,

///Ill/It,,,,

..... ,x ,.X5i,i ' I \1\, x...., , , ,

I'"'"
,,,,,,,, ['
.....
. . .., ,.,.,,.,.,•
..__..•*." / ! r • t , , .

_
•.........,/, 1, / / / / / / .
,!!
.......... .....,., .• / /.,., 1,., ....
_

-lO.O- I I I I II I I I I11I I I I I iii 1[_1

I..r i.i b.i L-I I.•! I•l I.t I.-I I.I !
-lO 13 -8 ½-5 8-4.e•-2.0
x (kin) x tkm)

Fig. 2. Low-level wind field 36 min after the initial condition Fig. 4. Horizontal
velocityvectorsat middlelevelsin experiment
shown in Figure 1. with low-level baroclinic region.

instability, which were reportedalongthe convergenceline Figures4 and 5. The low-levelbaroclinicallygenerated

prior to the July 2 CINDE tornado[Wakimotoand Wilson, horizontalvorticity gets tilted into the vertical along the
1989],is incidentalto tornadogenesis.The tornadooriginates easternand westernedgesof the updraft,resultingin a pair
from a much larger circulation (see section 4) than the of counterrotatingvortices at middle levels. Near the
Kelvin-Helmholtz vortices. ground,however,no significantverticalvorticityis gener-
The next experiment is initialized with no vertical or ated.Thetiltingin thiscaseisdoneby a vertical-motion field
horizontalvorticity but with a pool of air 1 km deep,which whichcontainsno significantdowndraft.Other similarex-
has been cooledby 10 K, in the north half of the domain. periments wereperformedwith differentinitialtemperature
Becausesucha temperaturefield wouldinducea northerly differences
andsoutherlywind speeds,andwith the coldair
densitycurrent,a southerlywindis initializedat all levelsto pooloccupyinga quarteror three-quarter
pie-shaped section
keepthebaroclinic zonesituated
beneath themainupdraftin of the domain. In every case there was no tendencyfor
the center of the domain. This experimentis designedto verticalvorticity develop close to the surface.
of horizontal To investigate
examinethe possiblerole in tornadogenesis the possiblerole of ambientverticalshear
vorticityproduction
by solenoids zone. (horizontalvorticity)in tornadogenesis,
alongthebaroclinic a seriesof simula-
The main resultsof this experimentare summarizedby tionswas performed having no horizontal
gradientsof either
initial wind or temperaturebut containingvariousamounts

13_5
.... x \ ', !,; •, l, • ,, •' 1I / / '"' ' '

/ f / / ! . \ \ \ ,.

ttI tt tt ,tt•
•
x
Fig.3. Asin Figure
2 butplotted
fromthehighest-resolution Fig. 5. As in Figure4 butat low
nested grid.
92 I"ORNAI)() SF'IN-UP BENEATH A CONVECTIVE CELL

100

• 0 0•
• , ......... ,• • x x ........

-2 •- . ....... t I I • 1 I I • , , .
....... ,• • •1

5.0
-4 •- ' ...... ' • • • • t I / / .......
_

-6 0- , ..................

I111 I_1111 Iiiii i iii iiiilllllllllllllllllll iiiii

i i i i i i i i i

-8 g-6.g-4.O-2.O •.• 2,0 4.• 6.• 8.•

x

l:ig. 6. l-lorizontalvelocity vectorsal middle levels in experiment Fig. 8. Horizontal velocity vectorsat low levels in experiment
with low-level westerly shear. with low-level westerly shear and a heat sink southwest of the
convective updraft.

ol'l'wcstcrly) vertical wind shear in the lowest 1-3 kin. Winds

at the lowest levels were easterly and graduated to westerlies to the right of the shear vector, thus encounteringvertically
higherup, so that the vertically averagedwind velocity was veeringwinds. This causedthe southern(cyclonic)vortex of
similar to the {xerol horizontal velocity of the zone of the pair to be stronger, as explained by Davies-Jones [ 1984]
convective forcing. and Rotunno and K!emp [ 1985]. None of these experiments,
Figures 6 and 7 characterize the results of these experi- however, resulted in vertical vorticity production near the
ments. A pair o!' counterrotating vortices forms at middle ground. As in the previous set of experiments, tilting was
levels (Figure 6} within the main updraft. The easterly flow caused by a field of vertical motion in which subsidence was
between the vortices can be explained as resulting from the insignificant.
upward transport of the ambient easterly momentum in the A final set of experiments was run with low-level westerly
lower boundary layer. The surface winds (Figure 7) are shear as in the previous set and, in addition, with a heat sink
strt>nglyconvergent but show no tendency to develop sig- located to the southwest of the main updraft between 2 and
nificant vertical vorticity. Variati{•ns of the experiment 4 km in height. The heat sink representsdiabatic cooling due
shown were perR)rmedwith different strengthsand vertical to evaporation of precipitation and induces a downdraft
limits ot' wind shear. Some experiments contained an ambi- approximately 5 km in diameter.
ent southerly wind, as if the main updraft were propagating Figure 8 shows the low-level horizontal velocity field in
one of these experiments 16 min after initialization. The
winds at this level were easterly initially, and the winds
remain generally easterly at the time shown. However, a
region south of the domain center contains strong westerly,
divergent flow. The temperature field at the same level and
time is shown in Figure 9. A pool of cool air is present south
....... '
of the domain center and coincides with the westerly cur-
......... rent. The cool air is the outflow from the heat-sink-induced
_

downdraft, and the westerly momentum, which has carried

_
the cool air eastward from the heat sink location, results
_
from that downdraft transporting the ambient westerly mo-
_

mentum from aloft downward to the surface. Vertical mo-

mentum transport in a vertically shearedflow was discussed
-5.0-- ......
_

_
ß,, •, ,,., •,. by Wiin-Nielsen [1973, pp. 96-97] and is associated with
tilting of ambient horizontal vorticity into the vertical. The
_

vertical vorticity thus produced resides along the boundary

-1 o. •-,, i_1i~11.1i_]11_1
i_l i.i 1.11.1l-Ji•_1
kl kJLIiLI L_ILI LI LI between the distinct air massesand is responsible in the
-1• • -5.0 a'.0 5.0 1•.• presentexperimentfor the formation of a tornadolikevortex
x (kin)
near the center of the domain.
Fig. 7. As in Pigure 6 but at low levels. Other experiments performed in this set differed in
 WALKO 93

Fig. 11. Three-dimensional perspective of vortex lines present in

the flow from which Figures 8-10 are extracted. The view is from the
southeast, and the vertical scale is exaggerated by a factor of 2.
-4

-6
core, centered at (x, y) = (-500 m, 500 m), are far greater
-8
than elsewhere.
To further illustrate the kinematics of tornadogenesis, I
-8.0-6.0-4.0-2.0 0.0 2.0 4.0 b.O 8.0 construct a field of vortex lines, each of which represents a
x (kin) vortex tube, say, of strength S. Beginning at the horizontal
Fig. 9. Temperature field (in degreesKelvin) at low levels corre-
surface in Figure 10, a starting point for each vortex line is
sponding to Figure 8. selected on the basis of the strength and distribution of
vertical vorticity across the surface. A total of N starting
points are determined, where NS is equal to the area-
strength, size, and location of the heat sink and in the integrated vertical vorticity across the horizontal slice
strength of the westerly shear. Not all of these resulted in a throughthe core. Then, a curve is constructedthrougheach
well-defined tornadolike vortex, but the majority did. A of the starting points, such that the curves are everywhere
critical factor in preventing vortex development in some parallelto the local 3-D vorticity vector. Althoughphysically
a vortex line either is a closed curve or terminates at a
simulations was a rapidly propagatinggust front. This pre-
vented the newly generated vertical vorticity from spending rotating solid surface (which is not applicable to the present
sufficient time beneath the convective updraft where hori- work since Earth rotation is neglected), I have, for clarity,
zontal convergence would concentrate it. constructedonly segmentsof the curves, terminatingthem
when they reach a height of 2 km as they rise through the
tornado core, and when they reach a lateral boundaryof the
3. VORTEX-LINE ANALYSIS
nestedgrid on which they are computed. These vortex line
In this section and the next I investigatein greater detail segmentsare illustratedin 3-D perspectivein Figures 11 and
the mechanismby which the vortex shown in Figure 8 was 12. Note in particular that as the vortex lines are followed
produced. The vertical vorticity 90 m above the ground downward through the vortex core, they turn into the
(outside the friction layer, which is less than 20 m deep) is horizontal within a few meters of the ground and continue
plotted in Figure 10. As can be seen,valuesin the vortex laterally rather than terminating at the ground. This follows
from the facts that velocity and vertical vorticity are zero at
the groundsurface(which was observedin the construction
of the vortex lines) and that the friction layer adjacent to the
ground is a region of strong horizontal vorticity. In the
presentsimulationit can be seen (in Figure 11) that the
vortex lines all eventually lead northward from the vortex

1 0-

--q 1•_-'•'•-.I
I_L!!l•l.•lI I I I IiiI I I I I I I I IllrI I I tI I !111
I I I I I I Il
-2.0 -1 .0 0.0 1 .0
x (kin)
Fig. 12. As in Figure11 but plottedfor vorticityon the highest-
Fig. 10. Verticalvorticity
corresponding
to Figures
8 and9. resolutionnestedgrid and with no vertical
94 '!'()RNAD() SPIN-UP BENEATH A CONVECTIVE CELL

core, in keepingwith the f'actthat the friction layer shearis ß , i , i , i , i , i , i , i , i , i ,

castefly over most of the domain. However, in the immedi- 8

ate vicinity of the core, the vortex lines spiral outward
almost uniformly over all azimuths for a distance of several 6

core radii, before they finally turn toward the north. '['his is
4
consistent with the general feature of tornadoes that just
above the friction layer and out to several core radii, their 2
azimuthal wind component is strong and nearly axisymmet-
ric. Only if a tornado were to translate horizontally at a
speed comparable to its peak azimuthal wind speed would -2
vortex lines spread outward mostly on one side of the core.
13
For example, consider a tornado which is axisymmetric just
above the l'riction layer and has a rotational speed at the edge
ortho tigre (relative to the axis) equal to the translation speed
ot' the axis relative to the ground. For counterclockwise -8
rotation the ground-relative wind at the iet't edge of the core
would be xcro, and the mean radial vorticity component
-"113_8 -6 -4 -2 13 2 4 6 8 1½3
within the i¾ictionlayer on that side must likewise be zero. x (krn)
V{)rtcx lines descending through the core to the friction layer
w{)uld turn outward mostly on the right side o[ this vortex. Fig. 13. Succession of three approximate paths traversed by a
material curve. Curves A, B, and C denote the paths at simulation
Consider allow like that shown in Figure 8 with the same
times oF 5, !1, and 16 min, respectively, the final curve closely
slowly evolving, larger-scale structure but at a shor! time surrounding a tornadolike vortex just above the friction layer.
prior to the {}ccurrenccot'a conccntntted tornadolike vortex. Heavy segment of curve A denotes section which is approximately
Foll{•w the kilometers-wide collection of vortex lines in 3 km above the ground, and the two dots indicate where the curve
l:;'igt•rc II s½)uthwardtoward the future location of the becomes close to the ground. Between the heavy segment and each
dot the curve is sloping, while the remainder of the curve is near the
tornado. '!'hose lines do not yet converge radially toward a ground. Curve B is similarly norated, but the heavy segment denotes
concentrated bundle because the tornado does not yet exist. a height near 2 kin.
Thtls they either continue away t'rt)m the region while
remaining in the [fiction layer or risc into the free atmo-
sphere without being concentrated. Eventually, they must m, 90 m), tracing that path as a material curve backward in
;til rise into the [rce atmosphere if not already there because time, and evaluating the terms in the equation for the
they will soon extend upward throt•gh the tornado core. In circulation around the curve throughout its history, the
the free atmosphere where diffusion is negligible, and in origin of' the concentrated vorticity in Figure 10 can be
cases where s()!cnoidal etli2cts can also bc ignored, vortex determined. This technique was applied by Rotunno and
lines advcct with the llow lBat•'helor, 19671. 1'bus radial Klemp [1985] in a less idealized numerical simulation of a
inttow must advcct the vortex lines from their initial broad tornadic storm. Here, the material curve is approximated by
spread to their location in the tornado core. This argument, a sequenceof 3000 Lagrangiantracer particles. Each particle
although admittedly not a proof, implies that the computed is advected backward in time using a time series of complete
configuration of vortex lines is evidence that flow conver- 3-D velocity fields obtained from the intermediate grid at
gence toward the vortex center from a distance of several 15-sintervals. Advection is performed in 1-s steps, and the
core radii operates to concentrate vorticity into the core as particlevelocity is recomputedeach step by trilinear spatial
the final step in tornadogenesis.Hence, while tilting of interpolation from the gridded velocity components.The
vortex lines is probably responsible for initiating vertical total number of tracer particles is sufficientfor adjacent ones
vorticity beneath the storm, that tilting must take place on a alongthe materialcurve to remain within one grid interval
horizontal scale of at least several tornado core radii, and throughouttheir history, with only minor exceptions.
horizontal convergence (radial inflow) concentrates that Upon trackingthe material curve backwardin time, it is
vorticity as the final stage of tornadogenesis. found to exceed the lateral boundaries of the intermediate
gridfor the first 5 min of the numericalsimulation.The curve
at t = 5 min is depictedas curveA in Figure 13. Becauseof
4. CIRCULATION ANALYSIS
the initial westerly shear, which is still essentiallyintact this
According to Stokes' theorem the circulation along a early in the simulation,it is obviousthat sucha path has
closed curve circling the perimeter of the vortex core is a positivecirculation.The actualnumericalvalueof the cir-
measure of the mean axial vorticity in the core. The circu- culationis 4 x 105m2 s-• , whichis comparable
to probably
lation theorem states that the circulation along any closed the largest documentedcirculationaround a tornado core
material curve in a fluid changesonly as a result of solenoidal [Walko, 1988].As the numericalsimulationadvancesfor the
or viscousterms. By selectinga horizontal circular path of first severalminutes,and the material curve evolvestoward
radius500 m centeredat the point (x, y, •) = (-500 m, 500 its final circular shapeimmediatelyoutsidethe vortex
 WALKO 95

thecirculation
aroundthematerialcurvedecreases
steadily .5. CONCLUSION
due almost entirely to a solenoidalinfluence. The cold air in The experimental results presented here, and several
'the southwestand later southernpart of the domainis others in the samecategoriesbut with differing parameters,
primarily responsiblefor this. The material curve at t = 11 support the argument of Davies-Jones [1982] that tilting of
min (depictedby curve B of Figure 13) has nearly all horizontal vorticity by an updraft alone is not capable of
descended to the surface,suchthat solenoids cannolonger producinga tornado vortex extending to the ground. Either
strongly influence the circulation. The circulation at this time vertical vorticity must be present in the ambient winds, or it
is 1.5x 105 m- s . Thecurvecontracts steadily toward its must be tilted from horizontal vorticity by motions which
final positionunder the stronglyconvergentwind field be- include a downdraft. I have argued here that such tilting
neath the updraft. The circulationdecreasesslowlyat first, must occur at a horizontal scale of at least several tornado
primarily due to surface drag, until the final minute when core radii and that the tornado is then formed by radial
velocities along the curve becomestronger,and the circula- inflow concentrating vorticity •¾om this scale to the core
tion decreases very rapidly. At t = 16 min, the material scale.
curve reachesits designated"final" position(depictedas
curveC in Figure 13)andhasa circulationof 5 x 104 m2 S -• .
Hence in this simulation, virtually all circulation in the Acknowledgmet•ts. This research was supported by Army Re-
search Office contract DAAL03-86-K-0175. Computations were
vortex was derived from the initial low to middle level shear, performed on the Stardent 3040 at the Department of Atmospheric
with solenoidal effects acting to weaken the circulation. This Science at Colorado State University. The Stardent 3040 was
result is oppositeto the resultsof Rotunnoand Klemp [1985] purchasedunder supportof ARO. The model used in this research,
in which solenoids actually generated the circulation within the RAMS, was developed at CSU under support of the ARO and
NSF.
15 min prior to the tornado. Although the solenoidalterm in
the present case was a circulation sink, the downdraft, which
REFERENCES
is instrumental in producing the solenoids,played the crucial
role in tilting horizontal vorticity into the vertical or, equiv- Batchelor, G. K., An Introduction to Fluid Dynamics, 615 pp.,
Cambridge University Press, New York, 1967.
alently, tilting the circulating material curve into the hori- Church, C. R., J. T. Snow, G. L. Baker, and E. M. Agee,
zontal, so that vertical vorticity at the surface then could be Characteristics of tornado-like vortices as a function of swirl
concentrated through horizontal convergence. ratio: A laboratory investigation, J. Atmos. Sci., 36, 1755-1776,
The circulation analysis may appear, at first glance, to be 1979.
Davies-Jones, R. P., A new look at the vorticity equation with
partly in contradiction with the vortex-line analysis of the
applicationto tornadogenesis,in Preprints, 12th Conference on
previous section. In Figure 11 the vortex lines shown that Severe Local Storms, pp. 249-252, American Meteorological
ascend through the vortex core approachthe generalregion Society, Boston, Mass., 1982.
from large distancesto the north, while the vortex lines that Davies-Jones,R. P., Streamwise vorticity: The origin of rotation in
pass through the area enclosed by curve A in Figure 13 supercellstorms, d. Atmos. Sci., 4I, 299!-3006, 1984.
Howells, P., and R. K. Smith, Numerical simulations of tornado-
approach the general region from large distancesto the like vortices, I, Vortex evolution, Geophys. Astrophys. Fluid
south. It seems likely that solenoidal recombinationof Dyn., 27, 253-284, 1983.
vortex lines in baroclinic regions and diffusive recombina- Lilly, D. K., The development and maintenance of rotation in
tion near the ground may accountfor this difference.For convective storms, in Topics in Atmospheric and Oceanographic
Sciences:Intense Atm•spheric Vortices, edited by L. Bengtsson
example, the initial flow consistsof a shallowfrictionlayer and J. Lighthill, pp. 149-160, Springer-Verlag, New York, 1982.
with denselypackedvortex linespointingtowardthe south Rotunno, R., and J. Klemp, On the rotation and propagationof
and a much deepershearlayer abovewith sparselypacked simulatedsupercellthunderstorms,J. Atmos. Sci., 42, 271-292,
vortex linespointingtoward the north.Both solenoidaland 1985.
Uttal, T., B. E. Martnet, B. W. Orr, and R. M. Wakimoto, High
viscouseffectsare capableof combiningsegmentsof oppo-
resolutiondual-Doppler radar measurementsof a tornado, in
sitely orientedvortex lines suchthat a line in the friction Preprints, 24th Con./•'renceon Radar Meteorology, pp. 62-65,
layer approachingfrom the northturnsupwardandreturns American Meteorological Society, Boston, Mass., 1989.
to the north in the deeperoverlyingshearlayer ratherthan Wakimoto, R. M., and J. W. Wilson, Non-supercell tornadoes,
continuingto the southin the frictionlayer asit didinitially. Mort. Weather Rev., 117, 1113-t 140, 1989.
Walko, R. L., Plausibilityof substantialdry adiabaticsubsidence
in
A similarphenomenonoccurssimultaneously with vortex a tornado core, J. Aimos. Sci., 45, 2251-2267, 1988.
linesapproaching
fromthe south.A detailedanalysis
of the Wiin-Nielsen,A., Cotnpendiumof Meteorologyfor ClassI and II
vortex-line evolution would reveal the processesinvolved Personnel.¾'olI, Part I: Dynamic Meteorology,WMO Publ.364,
but has not been performed here. 334 pp., World Meteorol. Organ., Geneva,
 Environmental Helicity and the Maintenance and Evolution of Low-Level
Mesocyclones
HAROLD E. BROOKS,CHARLESA. DOSWELLIII, AND ROBERTDAVIES-JONES

National Severe StormsLaboratory, NOAA, Norman, Oklahoma 73069

1. INTRODUCTION Rasmussen and Wilhehnson [1983] that tornadic environ-

ments are characterized by moderate to high CAPE and
Significant tornadoes are almost always associatedwith shear, they found that they could not discriminatebetween
supercell thunderstorms. SupercelIs are characterizedby tornadic and nontornadic severe thunderstorms (Figure 1).
their long life and the presence of a mesocyclone,a persis- Using a shear/CAPEdiagram, they could discriminatebe-
tent region of rotation several kilometers across through a tween severe and nonseverethunderstormswith a highlevel
deep layer of the storm. Approximately 90% of mesocy- of accuracy.On one hand, this result is extremely encour-
clones observed during the Joint Doppler Operational aging,implyingthatforecasterscanhavesomeconfidence
in
Project (JDOP) were associatedwith severe weather, and a forecast of severe thunderstormsjust by looking at these
50% producedtornadoes[Burgessand Lemon, 1990].As a two parameters.
On theotherhand,it impliesthatsomething
result, Doppler radar identification of a mesocycloneis additional must be done to discriminate tornadic from non-
considereda useful tool for issuingsevereweather warnings. tornadic environments.
However, an important question is why radar-observed Patrick and Keck [1987]found that parametersdescribing
mesocyclonesdo not all producetornadoes.As the WSR- low-level curvature of the hodograph could be useful in
88D Dopplerradarsare installedacrossthe United States, makinga tornado/no-tornado
discrimination.
On thebasisof
our understanding of the connection between radar- recenttheoreticalwork [Davies-Jones,1984;Lilly, 1986]and
observed mesocyclonesand surface severe weather will preliminaryoperationaltestingat the National Weather
become even more important. In particular,establishing Service Forecast Office in Norman, Oklahoma, Davies-
why mesocyclones fail to producesignificant can Jones et al. [1990] proposedthat the most appropriate
tornadoes
of highfalsealarmratesbasedonradar measureof hodographcurvatureis storm-relativehelicity.
reducethepossibility
signaturesof mesocyclones. The environmentalstorm-relativehelicity, •, is definedby
The detectionproblemof distinguishing
betweentornadic
and nontornadicthunderstormsis paralleledby the forecast
problemof distinguishingbetweenenvironments thathave
•(c) = -
z
k-(V-c)
0¾
x•dz (1)

the potentialto supporttornadicand nontornadic severe where h is an assumedinflow depth, c is the stormmotion
storms.Improvingourabilityto solvethisforecastproblem vector, ¾(z) is the environmentalwind profile (from a
hasobviousbenefitsin the publicwarningprocess.Mostof sounding,
profiler,Dopplerradar,or othersensor),
andk is
the researchcommunity'sfocushasbeenontornadicrather the unit vector in the vertical. Helicity has a simplegeomet-
than nontornadicenvironments[e.g., Darkow and Fowler,
ric representation
asminustwicetheareasweptoutby the
1971;Darkow and McCann, 1977].Turcotteand Vigneux storm-relative velocity vector between 0 and h on a
[1987]examined
allofthesevereweather
daysfor 1984-1986 hodograph.Thispropertyallowstheoperationalforecaster
in the QuebecWeatherCentreforecastareaanda large to estimatevalues of helicity from a hodographand use
numberof nonseverethunderstormdays(mostlyfrom 1986) whatever information is available to him or her, such as
and consideredthe amountof convectiveavailablepotential numericalmodel guidanceor asynopticobservations, to
energy(CAPE)andthemeanshearoverthelowest4 kmin update helicity
rapidly.Davies-Jones etal. [1990]
foundthat
themorning
soundings.
Whiletheyagreed
withtheresultof the storm-relativehelicity through3 km altitudeshows
promise asa tornado forecasting tool.Lefiwich[1990]and
The Tornado:Its Structure,Dynamics,Prediction,
andHazards.
Woodall[1990]alsoreportedencouraging
resultsin theiruse
GeophysicalMonograph79
Copyright
1993bytheAmerican
Geophysical
Union. ofhelicity
withforecast
winds.Becauseofthesuccess ofthe
98 I:,.NViR()NMI.•N'I AI, I{I:•[.ICITY ANI) LOW-LEVEl. MESOCYCL()NES

MEAN (0-12000 ft ahear) 2. EXI•ERIMEN'FAI.. DESCRIPTION

4x!O a 8x10 a 8x10 ':• 10x10 '3 12x10'*
4800 ' I•' Tornad•c
1 ........... [ ] [
The simulations that form the basis of this work are part of
From Turcotte and Vigneux, t987
A Sovoro a set of 21 supercell cases described by Brooks [1990]. All 21
4200 ,- 0 Non-sovoro
have the same initial thermodynamic profile, similar to, but
3600 -
slightly drier than, that used by Weisman and Klemp [1982,
1984].Thesurfacemoistureis 15g kg-•, resulting
in a CAPE
3000 of about2100J kg--•. The verticalwind profilehas been
varied to provide a range of helicities while maintaining a
2400
bulk Richardson index [Weisman and Klemp, 1982, 1984] in
•8oo a range associated with numerically simulated supercelIs
{values of 8-40). Since the helicity is a function of the storm
1200
motion, its value cannot be known a priori. After the
600 .......... I......... ,I.__, I I , ,, ! , I completionof the simulationthe helicityas a functionof time
- 5.6 - 5.4 -5.2 -5.0 -4.8 -4.6 -4.4 -4.2
can be computed.
LOG (0•12000 ft shear)
The numerical model used is that of Klemp and Wilheltn-
["lg. 1. Sc;,llcr cli;,gral•l ()f p(}Icntial but)yancy energy and low- xon [19781,as modified by Wilhelm.yonand C/ten [1982] and
level '•l•c;• •:,gn•l,•(Ic' fi)r tt')t]n(tcr•torn•day• and a•sociatcdweather Brookx [19901. !t is three-dimensional, fully compressible,
CVC•)t• •[• (Jutbee We;tiller ('clqIFc'• [i)l'CC;tslill'ca [i)r 1985-1987. and uses a Kessler microphysical parameterization. The
Sl<•l>•t•J,:•
lilac st}()ws;tpprt)xi•atc djv/si()n between severeand non- horizontal domain is 70 x 70 x 16 kin, with I km horizontal
resolution and a vertical resolution that varies as a hyper-
bolic tangentfunction from 200 n:tnear the groundto 600 m
near the top. The horizontal grid spacing is too large to
resolve the tornatilt circulation itself but is sufficient for
t)pcr;tlJ{)n:•llcsls wc h;tvc c()ntint•cda scric.sof numerical resolvinglarge mtsocycloniccirculations.The Coriolis el-
[•()(Icl %lt•ctics[BrooK.v,199()1th•tt I()()k at how thunderstorm l'cot and surface drag arc excluded for simplicity. Each
bcl•tvi()r v;tric% wil}• hclicity. Brookx [19901 t•)cuscd on simulation was carried out until at least 9000 s, with a large
p()%sihlcphysic;tln•cch•tnismslinking the environmental time step of 5 s. Other detailsof the model parametersare
t•clicity t(• the •lcvclt)p•ncnt()t'significantlow-level r()tation given by Brook.s'[ 1990l.
t• tl•tt•tlcrstt)rnts.In this p•tpcrwc extend his work to lo()k at Fhe three hodographsused here and associatedstorm
c:tscswl•crc signilic•tnt,I()ng-l:tsting,It)w-levelr()tationdoes motionsat maturity are shownin Figure 2. All three of the
•()l dcvcl<)p. simulationsproduce supercells,with long-lasting(greater
Wc c()nsidcr;t three-stagequalitative m()dcl {)1'the produc- than2 I/2 hours),intense(greaterthan30 m s.....
•), rotating
ti{)• {}t'I•)•g-livcd,higniticantt()rn•d()csby st•pcrcciis.'l'hc updraftsthatextendI¾omcloudbaseto thetropopause.
Two
li['sl is tt•c dcvc!{)pmcnt()l'str()ng n•idlcvcl( 3-7 kin) rota- hours into the simulations, case A has the largest helicity
titani•thcn•cs()cycl(.)nc(v()rticity
•(}.{)1
$ •).'l'hisproblem (705m2 s o-), withBandC substantially
less(550and120m2
t•;tsbccn treated by l,i!!y [19821and s -, respectivelye.
The hellcities
of A andB are well within
'l't)c scc()•]dstageis the dcvc!()pmcntand maintenanceof the valuesassociatedwith tornadoesby Davies-Joneset al.
%tll'l'aCc
mcsocycl()nic vorticity.The developmentaspectin [1990],but only A developsa significant,long-lasting,
low-
particularis treatedtheoreticallyby Rotunnoand Kle•np level circulation. Simulations A and B had identical strongly
[ 19851andl)t•vic.v-Jo/te.v
andBrook.¾ [thisvolume].The final curvedhodographs below3 km. From3 to 7 km, caseA has
stageis theintensification of thesurfacevorticityto tornadic a constantshearof 10 x 10-3 s-• while the shearfor B is
values( • 1 s I). Given the constraintsot'radar coverage,the half that, 5 x 10-3 s-•. Case C has an environment
resultof the firststageis the only one that canbe observed characterized by a straighthodographwith shearof 10 x
routinelyby Dopplerradar.Sincea criticalwarningissueis 10-3 s-t over the lowest 7 km, which is nearly equal to the
the nature of the surfitcc weather threat, implicit assump- meanshearoverthelowest7 kmin caseA, 9.7 x 10-3 s-• .
tions must be made about the connection between the first We focus on the differences in low-level structure in these
and later stages.Lack of developmentor persistenceat simulations.
either of the first two stageswould result in no significant,
long-livedtornadooccurring.We refer to suchcasesas 3. RESULTS
•'failure modes." Our fBcus here is on the second stage,
examining
whethera low-levelmesocyclone
develops
andis Theearlystages
of development
of the stormsin A andB
maintained. We describe two hilure modes and show howare similar,but after I hourof simulationtime,differencesin
theydifferfromthe developmentof a "classic"supercell the surface pressure field and rain water distribution
have
numericalcloud dramaticimpactson the furtherevolutionof the two storms.
tornadocase, usinga three-dimensional
CaseA provides
model. The failure modes are related to the location of the a classicexampleof an intensifyingand
gustfront relativeto the mainstormupdraft,and to the occluding
low-levelmesocyclone.
The originsof initialver-
sources of low-level rotation. tical vorticitydevelopmentat the surfacein this
 BROOKS ET AL. 99

a 20-

3 7
Case A (4200s)
Velocityat 300 m Pressure at 300 m
,

..
....

2
ß

ß , ' ...::::i:!-:•'½;-":!:::-'•'"--..
"--•

....•f""
-"
.'/'/;/,,,,.,,,.,....
,....
,...,,•i
-30 -20
10 20 30 40
2,.,
:,:.,:=====================
,
: ============================
-10 .' .

D 20 j " ,,
3 7 ß
. .

2 20 m/s -'

I I i
Fig. 3a
-30 -20 -1
• 10 2o 3o 4o
Fig. 3. Velocity (left) and perturbation pressure (right) fields at
300 m in 40 x 40 km section of domain in case A for times (a) 4200
C -10
20- s, (b) 5400 s, (c) 6600 s, and (d) 7800 s in life cycle of low-level
s 1 2 mesocyclone.Contours in left panel show vertical velocity (interval
1 m s-l) andvectorsshowstorm-relative
horizontalvelocityat
every other gridpoint. U indicates main storm updraft, and G
indicatesgust front updraft. Right panel shows pressurewith a
contour interval of 0.5 hPa. M indicates the location of the meso-
I I,
cyclone low pressure region, L other relative lows, and H relative
i o I I .... t I
-3O -20 -10 C 10 20 30 40
highpressure.Features from the pressurefield are indicatedon the
velocity figure for reference. Selected contours are labeled. Dashed
contoursare negative and zero contour is suppressed.
-10-

Fig. 2. Hodographsand storm motionfor three storms(A, B, and continuesto move eastward relative to the main updraft, the
C) discussedin paper. S indicates surface wind, and numbers are circulation becomesmore elongated, and the southerly wind
winds at various heights in kilometers. Individual model levels
below3 km are shownby squares.A and B differonly in A having component west of the gust front weakens. By 9000 s the
10X 10-3 s-I shear t¾om 3 to7 kmandB having 5 x 10-3 s-l over vorticity maximum in the region is now associated with a
that layer. Storm motion duringmaturephaseof stormfor eachcase long region of high shear instead of the more nearly circular
is indicated by stars. flow seen at 6600 s (not shown).
An important 'feature in the surface pressure field is the
inflow low. It is co-located with high wind speeds in the
regionof inflow of warm, moist environmentalair (high 0e)
clone are discussedby Davies-Jones and Brooks [this vol-
to the east of the updraft. The inflow low is a common
ume]. Here we look at the morphologicalcharacteristicsof
feature in observations of tornadic thunderstorms [e.g.,
the storm during the later, mature stageof the storm. A time
Charba and Sasaki, 1971]. Theoretical implications of the
series of horizontal cross sections of the velocity and pres-
inflow low have been discussed by Davies-Jones [1985].
sure fields at the second model level (300 m) shows the
development of long-lived, low-level rotation (Figure 3).
Two separateupdraft maxima are apparentat this level. One
Case A (5400 s)
is at the "nose" of the gust'front,and the other is at the foot Velocityat 300 m Pressure at 300 m
of the main updraft. As the mesocyclonebecomesbetter ..

developed, an extension of the rear flank downdraftwraps ..

aroundthe main updraft as it intensifiesfrom 4200to 6600 s

..
ß .

(Figures3a-3c). Dramatic changesoccurin the horizontal

velocityfield simultaneously. Initia!ly, the flow west of the
gustfront is northerly.In theregionbetweenthetwoupdraft -.========================
..... .........
maxima, westerlywinds soondevelopin the outflowfrom
the rainy regionassociatedwith the rear flankdowndraft.
North of the gust front updraft maximum,the flow has a -2 ' .•..::)fiH'-2..
southerlycomponent,leadingto a closedcirculationcen-
,

: .

tered on the updraft/downdraftinterface [Lemon and 20 m/s

Doswell,1979].By 6600s thisregionof closedcirculation

is
Fig. 3b
large (5-7 km diameter)and intense,with storm-relative
winds of the order of 15-20 m s-1. As the gust front Fig. 3.
I(tt• ENVIR(•NMENTAI, HELICITY AND LOW-LEVEL MESOCYCLONES

Case A (6600 s) a 0.1-1.1 km PressureDifference b 0.9 km VerticalVelocity

..

Pressure at 300 rn
.. .

Velocityat 300 rn ,

,,

..::::::.."....:::::::...
'.,..x.'.>..8
-.
"' .....87':"'• ,'?:
':

.-3
Downward
Directed
L.._.,...,•.•",/.,';
', / _ ,X ......
.'x.

20 •s - Fig. 4. (a) Verticalperturbation

pressuredifferencebetween0.1
and 1.1 km in 40 x 40 km section of domain in case A at 3600 s.
Fig. 3<' Contour interval is 0.1 hPa, and dashedcontours are negative. Sign
conventionsuchthat positive wtlues indicate higher perturbation
Fig. 3. (continued) pressure
at 0.1 km. Selectedcontoursare labeled.(b) Vertical
-1
velocityat (:1.9km in sameregion.Contourintervalis 0.5 m s

l:,"ron•consideration <)t' the Bernoulli relationship [Milne-

'l'hom.¾o•t, 1968lit can bc shownthat the magnitudeof the long-lived tornado.It is interesting
to notethatexperienced
l<)wis proportional t<>the innerproductof thestorm-relative stormchasersfrequentlyhaveobserveddevelopingcumulus
cnvir<)nmcntalwind and the perturbation velocity. The per- in the inflowregionof stormsa shorttime beforethe rapid
tt•r'bati•)npressuregradient is downwardover the lowest movementof the gustfront away from the main updraft.
kill>meterin the regimen just upwindof the surfacelow at 3600 Suchdevelopments are not seenin stormswith long-lived
s. !)t•wnward moti(>nin the intlt•w of the storm is associated mesocyclones.
The downwardmotionaheadof stormA
with this prcsst•rcgradient {Figure 4). would be consistentwith the suppressionof cloud develop-
In c{>ntrast
to the pronouncedinflow low of stormA, the mentin thatregion.On theotherhand,thereis nodowndraft
low in the infl{>wof storm B remains weak {Figure 5). This is in the inflowof B to suppress growingcumulus.Note also
a result {•1'wc•tkcr storm-relative environmental winds, due that masscontinuityimpliesthat there will be an enhance-
!o slower st{•rm motion, which in turn is a consequenceof mentof the component of flowin stormA towardthe main
weaker midlevel winds. With the weaker intlow low, the updraftand northernend of the gustfront beyondthe
pcrturbati{)n
pressure gradientis essentially zerooverthe alreadystrongenvironmentalinflow.
It>westkil½>mctcr,
and there is no regionof downwardmotion Therearealsosignificant
differences
in rain waterdistri-
aheadof the storm. Soon after this time, the gustfront begins butionbetweenA and B. With weaker storm-relativewinds
to moverapidlyaheadof the mainupdraft,cuttingoff its aloft, considerable precipitationis presentin B west and
inltowanddestroying any well-organized circulationat low southwestof the main updraft (Figure 7). The rain water
levels.Onlya regionof shearalongthegustfi'ontis apparent maximumat the tip of the hookat 3.3 km in caseB hascut
later in the storm'slit'etime(Figure 6). The absenceof a into the updraftat this leveland can alsobe seenin a
large,long-lived regionof mesocyclonic rotationin low precipitation
maximum anda larger areaofdowndraft south-
levelsin caseB arguesthatit wouldnotsupporta significant, westof theupdraftat thelowestmodellevel(Figure8). The
associated
stronger,rain-cooled outflow(low 0e) helpsto

CaseA (7800s)
Velocity
at300rn Pressure
at300rn CaseB (4200s)
a Velocityat 300 m b Pressureat 300 rn
...•,..--- . •,•,•,, ,,,..., .... .,,..

ß:......
::'•":,
' '5'
..... •::'7/
ß , ', ............
....
:::::½:::','•
.,".:-'",
::,":2 '"
:::h',:::: ::,..:,.,:
....::::::::::::::::::::::::::
::/....::.::::::.•
......
"œ,,
t:. ........ .'..,.',• .....
',:,.,,...,.'•:::::::::::::::::::::::::::::
....
I •,,:.';:.'...,:'::'.:,,'.:'.'•_':..::•11•
....
I '"'-"',:'::":-""'....."'•.:"""::
:iLk: .......

20 m/s -'-
Fig. 3d

Fig. 3. (continued) Fig. 5. SameasFigure3 for caseB at 4200

 BROOKS ET AL. 101

CaseB (6600s) VerticalVelocity/Rain

Waterat 100 m (4200 s)
Velocityat 300 m a CaseA b Case$

Same as Figure 7 except at 100m and with contourinterval

for rainwaterI g kg-• andfor velocity1 m s-t

20 m/s -•
winds. In relatively weak environmental winds aloft, little
Fig. 6. Velocity field at 300 m in 40 x 40 km sectionof domain in precipitationis thrown far downstreamof the updraft, and
caseB at 6600s. Contoursshowverticalvelocity(interval1 m s-i), muchof the precipitationis carried around the updraftby the
and vectors show storm-relativehorizontalvelocityat every other mesocyclone,leading to the characteristic rain-embedded
gridpoint.U indicatesmain stormupdraft,andG indicatesgustfront circulation of an HP supercell.
updraft.
The results of Turcotte and Vigneux [1987] suggestthat
some storms in high-shear environments may not produce
significantlow-level rotation. An example of sucha storm is
advance the gust front. While the difference in intensity of caseC. Storm C is slower to develop than A but has updrafts
the outflow behind the gust front between A and B probably greaterthan 20 m s-I from 1 hour until the end of the
is the largest factor contributing to the faster gust front simulation (2 hours 50 min). The midlevel vorticity reaches
movement in B, we observe that the difference between the
mesocyclonic
values(0.01 s-l) after only 20 min into the
inflows, discussed above, also contributes to faster gust simulation
andis 0.02 s-1 , similarin magnitudeto A andB,
front movement in B. The precipitation pattern in B, with a from 70 min onward. The low-level vorticity, on the other
large amount of rain in the vicinity of the mesocyclone,on hand,neverexceeds0.005 s-i, while A and B reach0.02
the south side of the updraft, is similar to some high- s-i . We stressthat despitethe similarityin the meanshear
precipitation (HP) supercelIs[Moller et at., 1990].We spec- in A and C, the surface features associated with the storms
ulate that this represents one environmentally controlled
are dramaticallydifferent. The caseC low-level velocityfield
mechanism in the development of HP supercelIs,namely, at 9600 s showsa nearly east-westaligned, weak gust front
the generation of a strong mesocyclonein an environment (Figure9) with the cold air never advancingto the west and
with moderately weak mid to upper level storm-relative southof the updraft. Becauseof the strongenvironmental
storm-relativewinds, parcels of air travelinginto the updraft
from the north side spend only a short time in the strong
Vertical
Velocity/Rain
Waterat3300m (4200s) temperaturegradientgeneratedby the evaporativelycooled
a Case A b Case B air north of the updraft. As a result, there is little baroclinic

Case C (9600s)
a Velocity
,at300m b Pressure at 300 rn

Fig. 7. Rainwater(lightlines)andverticalvelocity(heavylines)
at 3.3 km at 4200 s in 40 x 40 km sectionof domainfor (a) caseA
and(b)case
B.Contour
interval
forrainwater
is2.5gkg-1andfor
velocity
is 4 m s-i. Dashed
contours
indicate
downdraft.
Selected 20 m/s -

contours
are labeled.Updraftsanddowndrafts
areindicated
by U
and D. Fig. 9. Sameas Figure3 exceptfor caseC at 9600
!1}2 ENVIRf)NMENTAI• HEI,I(?I'FY AND LOW-LEVEL MES()CYCLONES

Roth.l'?•sz.and Lilly [19891 to interpret the flow in a helical-

vortex chamber. It can be shown that the basic flow is not a
steady state solution of the Euler equations and that vertical
circulations, which increase in strength with proximity to the
sink, must develop owing to unbalanced pressure forces.
When the environmental vorticity is nearly streamwise, as
in case A, vortex line stretching in the accelerating inflow
amplifies the streamwise vorticity, giving rise to helical flow
with upflow and downflow on the left and right sides of the
inflow, respectively. The surface inflow low, being associ-
ated with the maximum surface wind speed, is located on the
center line oi' the surface inflow. Because the environmental
winds and hence the inflow veer with height, the surface
inflow low is overlain by downdraft, as observed.
In case C the ambient vorticity is nearly crosswise in
storm-relative coordinates. Conlluence o1' the streamlines
Fig. I(). Streamline', tk)r ;t sink (marked t-./ an asterisk) in a
shrinks the vortex lines, i.e., generates perturbation vortic-
unitk>rmcastefly tlow in a horiztmta!cross .cotion. The stagnation
pt}int is labeled S, and X, Y, Z, •tnd W mark the positionof tYur it y in the opposite sense to the ambient vorticity. Because
n•ttlcrialpoinlh;tl tWOditl'crcn•times.The •natcriailine XY, aligned the strength of the circulation increases inward toward the
with the 11ow, is stretched, :t•d the material line ZYW, transverse to sink, the perturbation vorticity must be associated with
ttic tlt'•w, is hhr•nk. 'l'hus, when the environmental winds veer height downward motion on the inflow side.
(hlrcarr•wihcvorticity case), the pcrturbatit•n vt•rticity is westward,
i•nplying upwztl'd(downwin'd}motion c)n the !eti (right) side ot' the
inll(•w. In a westerly sheared cnvironnicnt (crosswise vorticity
4. SUMMARY AND [MPIACATI(')NS
lt)wartl n{•rth), ltic pcrturbatitm vt)rticity is southward, implying
dr•wnwltrd n•c•lit•n {m the east hiclc ot' the sink.
Wc have shown two different evolutions that result in
storms with midlevel mesocyclonesnot producing signifi-
generation()[ h(•rizontalvorticity. Further, the environmen- cant, long-lived low-level rotation, which we refer to as
tal vt•rticityvectortperpendicularto the hod(•graphshownin "t'ailure modes." The intensity ot' the storm-relative winds
lqgurc 2•'), is nearly perpendicular tt• the storm-relative appearsto play a critical role. A comparisonof various
wind,, •tt this level, s{•that tlow has little streamwise vorticity l-'eaturesofthe storms isolates the nature of the failure modes
in this region.Becauseot'thcsetwo effects,air risinginto the (Table I). Storm C Fails because of the absence o'f mecha-
updrat'thas little positivev{•rticity,and a mtsocycloneFails nisms to generate low-level rotation [Davies-Jonesand
to devclop at low levels. Brookx, this volumel. Storm B fails because the outflow
'l'hc low-pressurearea in the inll()w ot' (' is even stronger overwhelms the storm inllow, and the gust Front surges out,
than in A, and associated with the low-pressure area is a undercuttingthe low-level mesocycione.
strongdowndraft I The downward The movementof the gustl¾ont,which affectsthe length
of the orderof 1 m s.......
motion in the inflow of storms A and C can be explained of time that the low-level mesocycloneexists, is a complex
usinga simpleinviscidflow modelconsistingof the environ- problem.We havefocusedon two aspects,the strengthof
mental winds with a sink flow (representingthe effectsof the the storm-generated outflowandthe strengthof the environ-
updratl on the inflow) superimposed,as illustratedby the mental inflow. While the outflow intensity probably is the
schematicshown in Figure 10. A similar model was used by majorfactor,the inflowmay makea significant contribution

TABLE 1. BasicDesciptionof Featuresof Interestin the Three Simulations

Baroclinic
Environmental Streamwise
Storm-Relative InflowLow FlowBehind Low-Level Vorticity
Case InflowWinds Precipitation Character GustFront Vorticity Generation
A strong lessthanin B strong
lowwith parallel mostly
streamwise large
downdraft but less so than
in B
B weakerthan heavyto westand weaklowwithvery perpendicular
and mostly
streamwise lessthanin
in A southof updraft weakupdraft strong caseA
C verystrong moderate
(no verystrong
with parallel
and nearly
crosswise small
hook) strongdowndraft strong

CaseA produces
a large,strong,
long-lived,
low-level
mesocyclone.
In caseB thegustfrontsurges
outahead
of thelow-level
mesocyclone;
no low-levelmesocyclone
formsin case
 BROOKS ET AL. 103

800-
LargeLow-levelMesocyclone nontornadic environments [Turcotte and Vigneux, 1987]
Small Low-levelMesocyclone
700- No Low-levelMesocyclone
may be explainedin part by these results. Our effortsare just
GustFrontSurgesOut beginningto reveal the relationship between the environ-
600- GustFrontEast-WestAligned ment and the storm, and the conditions under which rotating
storms produce or do not produce tornadoes. As Doppler
• 500- radar observations become an important part of the severe
weatherwarning system, these issuesmust be addressedin
E.•4oo-
more detail. The results here reinforce the notion that the
._

questionof whether an environment will support tornadic

. 300--
thunderstormsis not the same as the question of whether it
200-- will support supercelIs.

100--

Acknowledgments. This work was donewhile oneof the authors

0 (H.E.B.) held a National Research Council-NOAA ResearchAsso-
10 15 2O 25 3O 35 ciateship. The computer simulations were carried out on the
SurfaceStorm-Relative
Wind (m/s) CRAY-2 at the National Center for SupercomputerApplications
(NCSA) and softwaredevelopedat NCSA was usedin the analysis
Fig. 11. Classification of low-level flow in storms as a function of of the data. The simulationswere supported,in part, by NSF grant
the environmental helicity and surface storm-relative wind. Letters ATM-87-00778. We thank Doug Lilly, Richard Rotunno, and
indicate storms discussedin detail in paper. Michael Foster for their comments, and Robert Wilhelmson for his
assistance with the simulations. Viateur Turcotte, Denis Vigneux,
andAndre Sevignyof the QuebecWeatherCentregavepermission
to use Figure 1.
to the movement of the gust front. In the case of A and B the
two effects are in the same sense, namely, to acceleratethe REFERENCES
gust front in B more than in A.
While we have discussedonly three simulations,we have Brooks, H. E., Low-level curvature shear and supercellthunder-
characterized the behavior of all of the storms in the larger storm behavior, Ph.D. thesis, 233 pp., Univ. of II1., Urbana-
Champaign, 1990.
data set of Brooks [1990] as a function of helicity and storm Burgess,D. W., and L. R. Lemon, Severethunderstorm
detection
inflow speed (Figure 11). There is a tendencywithin the by radar, in Radar in Meteorology, edited by D. Atlas, pp.
group that producesmesocyclonesto have larger mesocy- 619-647, AmericanMeteorologicalSociety, Boston,Mass., 1990.
clones as helicity increases.Storm C is uniquein the data Charba, J., and Y. Sasaki, Structure and movement of the severe
thunderstorms of 3 April 1964as revealedfrom radarand surface
set, while the failure mode exhibitedby stormB is limited to mesonetworkdata analysis,d. Meteorol. Soc. Jpn., 49, 191-214,
high-helicity,low-inflow speedstormsin the presentstudy. 1971.
This latter point may have relevanceto the observation
of Darkow, G. L., and M. G. Fowler, Tornado proximity wind
Moller et al. [1990] that significanttornadoesin HP super- analysis,in Preprints,7th Co/•f•'renceon SevereLocal Storms,
cells are less common than in "classic" supercelIs. The pp. 148-151,AmericanMeteorologicalSociety, Boston,Mass.,
1971.
results of case B indicate that one reason may be that strong Darkow, G. L., and D. W. McCann, Relative environmentalwinds
outflow associated with HP storms inhibits the maintenance for 121 tornado bearing storms, Preprints, 8th Co/el,renee on
of a long-lived,low-levelmesocyclone.
More studyof ob- Severe Local Storms, pp. 413-417, American Meteorological
Society, Boston, Mass., 1977.
servedand modeledHP supercelIs is necessary
to testthis Davies-Jones,R. P., Streamwise vorticity: The origin of updraft
hypothesis. rotationin supercellstorms,J. Atmos.Sci., 41, 2991-3006,1984.
Our examinationof the spectrumof helicity and storm Davies-Jones,R. P., Dynamical interactionbetweenan isolated
inflowspeedis limitedat thistime.Of particular
interestis convective cell and a veering environmental wind, in Preprints,
the high-inflow-speed,
high-helicity
regionin Figure11for 14thCotd•'renceon SevereLocal Storms,pp. 216-219,American
MeteorologicalSociety, Boston, Mass., 1985.
whichwe haveno data.Rapidlymovingstorms,frequently Davies-Jones, R. P., and H. Brooks, Mesocyclogenesisfrom a
seenin theearlyspring,suchastheonewhichproduced
the theoreticalperspective,Proceedings,Tornado SymposiumHI,
long-track
Selmer,Tennessee,
tornadoin March1991,are edited by C. Church, this volume.
characterized
byinflow
speeds
oftheorder
of30ms-• and Davies-Jones,R. P., D. W. Burgess,and M.P. Foster, Test of
helicityas a tornadotorecastparameter,in Preprints,16thCon-
highhelicity.To date,we havenotmodeled anyof these jkrenceon SevereLocal Storms,pp. 588-592,AmericanMeteo-
stormsandonlycanspeculate asto the relevance of the rologicalSociety, Boston, Mass., 1990.
conceptsdiscussed here.Clearly,somestorms withhigh Klemp, J. B., and R. B. Wilhelmson,The simulationof three-
inflowspeeds produce significant
long-lived,
low-levelrota- dimensionalconvective storm dynamics, J. Atmos. Sci., 35,
1070-1096, 1978.
tion.The abilityto identifyenvironmentalconditionsasso- Leftwich, P. W., Jr., On the use of hellcity in operational assess-
ciatedwith theirproduction wouldbe of benefitto opera- mentof severelocal stormpotential,in Preprints,16thCottœer-
tional meteorologists. ence on Severe Local Storms, pp. 306-310, American Meteoro-
The observation
that shearandbuoyancy,combined
into logicalSociety, Boston, Mass., 1990.
Lemon,L. R., and C. A. Doswell II1, Severethunderstorm
a single
parameter,
cannot
distinguish
between
tornadic
and
104 ENVIIR()NMEN•I'AI, HELICITY AND LOW-LEVEL MESOCYCLONES

titm and mtsocyclone structure as related to tornadogenesis, Rothfusz, L., and D. K. Lilly, Quantitative and theoretical analyses
Mort. Weather Rev., 107, il84-1197, 1979. of an experimental helical vortex, J. Atmos. Sci., 49, 2265-2279,
l,illy, 1). K., The development and maintenance of rotation in 1989.
convective storms, in Topics in Atmospheric and Oceanographic Rotunno, R., and J. B. Klemp, On the rotation and propagation of
.¾ciet•cex:lntenxe Atmospheric Vortices, edited by L. Bengtsson simulated supercell thunderstorms, J. Atmos. Sci., 42, 271-292,
and J. [,ighthi!l, pp. 149-160, Springer-Verlag, New York, 1982. 1985.
I.illy, I). K., The structure, energeticsand propagationof rotating Turcotte, V., and D. Vigneux, Severe thunderstorms and hail
convective storms, II, Helicity and storm stabilization, J. Atmos. forecastingusingderived parametersfrom standardRAOBS data,
S•'i., 43, 126-140, 1986. in Preprints, 2nd Workshop on Operational Meteorology, pp.
Milne-Thomson, 1,. M., Theoretical Hydrodynamics, 743 pp., Mac- 142-153, Atmospheric Environment Service and Canadian Mete-
Millan, New York, 1968. orological and Oceanographic Society, Downsview, Ont., 1987.
Moller, A. R., C. A. Doswell III, and R. Pryzybylinski, High- Weisman, M. L., and J. B. K!emp, The dependence of numerically
precipitation supercells:A conceptual model and documentation, simulated convective storms on vertical wind shear and buoy-
in Prcprint.•, 16th ('<m.[•,renceon Se•,ere Local Storms, pp. 52-57, ancy, Mon. Weather Rev., 110, 504-520, 1982.
American Meteorological Society, Boston, Mass., 1990. Weisman, M. L., and J. B. Klemp, The structure and classification
Patrick, 1.•., and A. J. Kcck, The importance of the lower level of numericallysimulatedconvectivestormsin directionally vary-
windshem'protile in tornado/nontornado discrimination, in Pro- ing wind shears,Mort. WeatherRev., 112, 2479-2498, 1984.
ccedings, Symp(•sium on Mesoscale Analysis and Forecasting, Wilhelmson, R. B., and C. S. Chen, A simulation of the develop-
Eur. ,%'pa•'c A,•4ettt'ySpet'. Plthl. ESA SP-282, 393-397, 1987. ment of successivecells along a cold outflow boundary, J. Atmos.
Rasmt•sscn, I5. N., and R. B. Wilhelmson, Relationships between Sci., 39, 1466-1483, 1982.
slt•rn• characteristicsand 1200GMT hodographs,low level shear, Woodall, G. R., Qualitative forecastingof tornadic activity using
and stability, in Preprints, 13th Con.[•,renceon Severe Lo('al storm-relative environmental helicity, in Preprints, 16th ConJkr-
,%'lort•t.¾,
pp..15-.18, American Meteorological Society, Boston, ence on Severe Local Storms, pp. 311-315, American Meteoro-
Mass., 1983. logical Society, Boston, Mass.,
 Mesocyclogenesis
From a TheoreticalPerspective
ROBERTDAVIES-JONESAND HAROLD BROOKS

NationalSe•,ereStormsLaborator3,,NOAA. Norman. Oklahonla73069

1. INTRODUCTION vation law for moist convection) and may not be useful when
moisture effects are essential to a particular process.
Since the last tornado symposiumin 1976, considerable In this paper we first review the midlevel mesocyclogen-
progresshas been made toward understanding tornadogen- esis mechanism. We then apply vorticity and circulation
esisin supercelIs.The genesisof mesocyclones and associ- theorems to determine how near-ground rotation develops in
ated tornadoesis now known throughnumerousobserva- a three-dimensionalnumerical simulation of a supercell. We
tionsto be a three-stageprocess.The mesocyclone develops conclude with a brief survey of current tornadogenesis
theories.
first at midlevelsthroughtilting by an updraft of streamwise
vorticity presentin low levelsof the environment[Browning
and Landry, 1963; Barnes, 1970; Lilly, 1982, 1986;Davies- 2. THE INITIAL MIDLEVEL MESOCYCLONE
Jones, 1984]. The mesocyclonedoesnot simplybuild down
to low levels; instead, low-level rotation appears to be Linear theory of dry convection in shear virtually explains
caused by a separate, more complicated mechanism [Ro- initial net rotation of the storm's updraft at midlevels as a
tunno and _Klemp, 1985; Johnson et al., 1987] requiring consequenceof tilting toward the vertical of low-level envi-
downward as well as upward motions and also strong ronmental streamwise (horizontal) vorticity by the updraft
horizontal temperature gradients. Finally, in perhaps the [Lilly, 1982, 1986; Davies-Jones, 1984]. Only the flow rela-
tive to the updraft is physically relevant becausethe updraft
least known process, a tornado develops within the meso-
is the tilting agent. TherefOre streamwise refers here to the
cyclone.
direction parallel to the local storm-relative wind. In a
The theoretician has a number of tools that he can use in
horizontally uniform wind that veers (backs) with height
attempting to uncover the origins of rotation. These include
without a change in speed, an air parcel experiences a
vector equationsfor vorticity and entropy gradient[Davies- differential velocity that causes it to spin clockwise (anti-
Jones, 1982a], the streamwise vorticity equation [Scorer, clockwise) about its direction of motion like a propeller or a
1978], vorticity and circulation theorems [Dutton, 1976, pp. spiraling American tootball thrown right- (left-)handedly,
364-374], Beltrami solutions of the Euler equations[Lilly, indicating the vorticity is purely streamwise (antistream-
1982, 1986; Davies-Jones, 1985], integrals of the vorticity wise). When the wind increases with height without turning,
equationfor isentropicflow [Dutton, 1976,pp. 385-390]and the differential velocity is parallel to the wind and the
conservationlaws for entropy, potentialvorticity [Dutton, vorticity is purely crosswise (like a wheel or a ball with top
1976,p. 383], and helicity(the volumeintegralof the scalar spin). These solid body analogiesare limited becauseair
productof velocityandvorticity)[Lilly, 1986].Eventhough parcels change shape as well as spin; strictly speaking,
rotation is a kinematic property ostensiblyunrelatedto vorticity is twice the angular velocity of a sphericalparcel
entropy, the above list implies, rather surprisingly,that that is instantaneouslysolidified while its spin angular mo-
specificentropyS is asimportanta variableasvorticityto.In mentum is conserved. In a horizontally homogeneous,
fact, vorticity and entropyare closelyentwinedin geophys- strongly sheared environment with the Earth's rotation
ical flows [Dutton, 1976,pp. 338, 383]. Unfortunately,sev- "switched off," an incipient, isolated updraft of roughly
eral of the theoremshold only for dry, inviscidconvection circular cross section acquires cyclonic (anticyclonic)rota-
tion in the streamwise (antistreamwise) case and no net
(e.g., thereis no strictequivalentpotentialvorticityconser-
rotation in the crosswise case. Davies-Jones [1984] was the
firstto provethisresultfor generalwindprofiles;further,he
The Tornado:Its Structure,Dynamics,Prediction,
andHazards.
did it from a full set of (linearized) equationsinsteadof from
GeophysicalMonograph79
Copyright1993by theAmericanGeophysical
Union. the vertical vorticity equation alone. A key step in the proof
!()6 MES()('Y('[,OGEN}:.SIS FR()M A THEORETICAl. PERSPECTIVE

wasthe introctt•ction
of verticaldisplacement
asa dependent
variable.Da•'ie,•-,]one,s'[1984,
1986]exp]ainedthe process
physicallyin terms of the updraft upliftingthe initially
horiz(•ntaliscntropicsudSaces and vortex lines, thereby
c•tabIishingthelink betweenthethermalandvorticityfields.
Liftingup loopsof horizontalvorticitynaturallyproduces a
cyclonic-anticyclonicvortex pair. In the streamwise(anti-
slrcamwisc)casethe anticyclonic(cyclonic)vortexis down-
$1Fcamin lesserupdraft or in downdraft, and in the crosswise
casethe vorticesare on the left and right downstreamsides
()t*llrtcupdraft[sccNolttntio,thisvolume,Figure8]. Potential
vt•rlicily (•w- VS (u is specific volume)is conservedeven
w/lht)t•t lincltrization and is zero because of the initial
c{)ntlitit)ns [l),s'ic.v-Jotte,v, 1982a, 1984; Rolunn(• and
Kh,/•q•, 19851.'l'husthe vortex lines remain in their original Fig. !. Schematicdiagramillustratingthe linear theory relation-
iscnlrt)picst.trlhccs.'l'his result has importantimplications shipsbetweenverticaldisplacement,f, verticalvelocity w', vertical
vorticity •", and vertical perturbation pressuregradient (VPPG)
l't}ri{•rn•tdogcncsis (sccbelow). It alsopermitsgeneralization Ovr'/i•Z[k)r a level where the storm-relative winds, v - c, increase
t•!'tlic •aht•vcthct)ry t{>a semilinearone by allowingfinite and veer with height.Environmentalshearand vorticity vectorsare
vertical clisplaccn•cnts.teLvcn simpler explanationstbr up- denoted by S and w. Centers, N, W, Z, and P, of circles indicate the
(tr:tflrc)tlttit)n,will•c)utrccc)t•rsc
to the conceptsot'isentropic relative positionsof the •', w', .C, Orr'/az maxima, respectively.
sllrf•lccs•ttltl v•rlcx lines, •trc given by Rabin ttnd Das'ics- ()vcrlapping,touching,or disjointedcirclessignifypositive,zero, or
negative correlation between the two variables involved. Note,
.1ott•',v[19861 and I)ttl'it',v-./on•',v[1992a, hi. however, that the correlation between w' and Orr'/az. becomes
'!'lie :tb•)vc•sstiincsthat the disttirbanccis rc)tlghlycircular negative for increasing storm-relative wind speed [from Davies-
•nd sc•tili•car. 'l'hcsc asst•mpti{)nsare partly justifiable .Iones, 19841.
[,cc:ttisc(i) stipcrccllsarc qtiitc circular in their initial stages
•t!ltl, ftirlhcrln•rc, the shape dependence elisappearsin the
liniit where the v½>rticityis purely streamwise, and (2) the
c;•vi['t)•lcnt.al winds arc as iitrgc its the storm-generated buoyant production of helicity is observed to be positive
t}ncs for m•ttt•f'c severe st()rms, st) the linear terms arc (instead of zero) in numerical simulations of storms with
ttiw;tys impt•rt•tnt; also, lhc cxcltictcclnt>nlincar terms havemesocyclones, the other correlations shown in Figure 1
lil!!c c!•;•lilativc influence t•n t•pctrttt'lr½•tation. appear to be qualitatively correct. For instance, new up-
Acc•rding I{> the linear thc{•ry the sp•ttial rclationships drafts {due to upward VPPGF) and cyclonic vorticity both
bclwccn imp{•rtant variables at c;tch level arc as depicted in are ()bserved typically on the storms' right sides.
}:igi•rc 1. In the storm's rct•rcncc t'ramc and relative to the Storm motion, which affects the "streamwiseness" of the
l{}cal•½•n{•t'maximum vertical displacement (or pcak buoy- vorticity, thus far hatsbeen treated as given, even though in
•tncy if the lapse rate is unstable at thai level) the maximum the linear theory it is determined by a complex eigenvalue.
vertical velocity and vertical perturbation pressuregradient Rolunno and Klemp [1982] physically and Davies-Jones
t'orcc (VPPGFJ are upwind, while the maximum vertical [1984] mathematically deduced that in linear theory, storm
vorticity is on the right side with respect to the shear vector. motion should lie to the concave side of a curved hodograph
When streamwise environmental vorticity is present, (hori- and on the hodographfor a straight one. Storm splitting and
zontal) environmental helicity is converted into vertical movement off a straight hodograph are nonlinear effects.
hellcity (the volume integral of vertical velocity times verti- Furthermore, boundary layer convergence zones associated
cal vorticity) through a term, which is the integral of vertical with fronts, outflow boundaries of other storms, etc., may
vorlicity times VPPGF. This is analogous to the transforma- affect storm motions considerably, sometimes altering them
tion of horizontal into vertical kinetic energy through a in the right way for significant updraft rotation to develop.
similar pressure term. The buoyant production of helicky Currently, no satisfactory method for quantitatively estimat-
tcsscntiaIly the covariance of buoyancy and ve•ical vortic- ing storm motion has been devised. Existing methods either
ity) is zero in linear theory and in a nonlinear extension are not Galilean-invariant or fail to predict storm splitting.
thereof [Davies-Jones, 1985]. Lil& [1986] suggested that Given observed or crudely estimated storm motions, the
ve•ical helicity receives a contribution from only the buoy- linear theory may be used to nowcast mesocyclonesfrom
ancy term and gives away part of that to horizontal distur- proximity upper air data. Davies-Jones et al. [1984] adapted
bance helicity through the pressureterm, becausebuoyancy a formula for the correlation coefficient between vertical
and the vertical perturbation pressure gradient (VPPG; • - velocity and vertical vorticity derived from linear theory.
VPPGF) are generally positive in peak updraft regions. The correlation coefficient was estimated from soundings as
However, as shown in Figure 1, vertical vorticity and VPPG the relative helicity (cosine of the angle between the envi-
can be negatively correlated, even though both these vari- ronmental vorticity and storm-relative wind vectors) times a
ables are positively correlated with vertical velocity. While factor between zero and one that penalizes for light
 DAVIES-JONES AND BROOKS 107

relativeenvironmental
windscompared
to storm-generated
shear). Thus the streamwise vorticity of inflowing air may
horizontal winds (estimatedfrom the convectiveavailable increaseas it acceleratesand curvesinto a mesocyclone.This
potential
energy).An average
valueovertheinflowlayerof might be a significanteffect after the developmentof intense
the storm(taken to be the lowest3 km) was used.Davies- low-levelrotationand low pressure.At this stage,the stream-
Joneset al. [ 1990]usedthestorm-relative
felicityofthe0- to linesin the warm air may bend tip more abruptly as they enter
3-kmlayerinstead,because it maybecomputed asa simple the updraft becausethe low has descendedfrom midlevels,and
integral without any differentiationand hence is a far more tilting of streamwisevorticity in the warm air might contribute
stable parameter. In the linear theory the environmental now to rotation at fairly low levels.
helicitydensity(streamwisevorticitytimeswindspeedin
storm-relative
coordinates,
i.e., the integrandin thefelicity 3. BAROTROPIC AND BAROCLINIC VORT1CITY
definition) is proportional to the covariance of vertical
Before discussingthe development of low-level rotation, it
velocity and vertical vorticity. Once again, weak storm-
is convenient to introduce the concepts of barotropic and
relativewindsresultin low valuesof the forecastparameter.
baroclinic vorticity. Dutton [1976] shows from an integral of
In additionto computationalstability,felicity has another
the vector vorticityequationfor dry, inviscid,isentropicflow
advantage over other parameters. It is minus twice the
that the vorticity is the sum of barotropic and baroclinic
signedarea swept out by the storm-relativewinds between0
components. This decomposition is dependenton the choiceof
and 3 km (the assumedinflow depth) on a hodograph
anarbitraryinitialreferencetime to.The barotropicvorticityof
diagram.Area is well definedeven for fractal hodographs
a parcel at a later time t is the vorticity that the parcel would
and is easily visualized by forecasters.Also, helicity is a
have from the amplificationand reorientationof initial vortic-
linear function of storm motion. Thus helicity contoursas a
ity, if the baroclinicterm were ignored.Barotropicvortex lines
t•nction of storm motion (for a fixed wind profile) are
are "frozen" into the fluid. The barotropic vorticity field at
parallel straight lines on a hodograph diagram, and the
timet is givenby the Cauchyformulain termsof the initial(i.e.,
effects of deviant storm motion in a given environmentare
reference)vorticity field and the mappingfrom Lagrangianto
easily assessed in advance.
Eulerian coordinates.It is independentof parcel positionsat
In contrastto linear theory, which describessmallgrowing
intermediatetimesr (to < r < t). On the otherhand,baroclinic
disturbances, exact steady state Beltrami solutions of the
vorticity consists of vorticity that has been generated as
Euler equations [Lilly, 1982, 1986; Davies-Jones, 1985] (R.
horizontalvorticity since the initial time by buoyancy torques
P. Davies-Jones and H. E. Brooks, manuscript in prepara-
andsubsequently affectedby vortex tube stretchingandtilting.
tion, 1993) are useful for modeling mature, midlevel meso-
The baroclinicvorticity of a parcel dependson its integrated
cyclones. A Beltrami flow is one where the vorticity and
thermalhistoryfrom to to t. The baroclinicvortex lineslie in
velocity vectors are alignedwith eachother at every point in
isentropicsurfaces;hence the barocliniccomponentdoes not
the fluid. The advantage of having an exact analytical
contributeto potentialvorticity. The generationof vertical
solution is partly offset by omissionof buoyancyforces. vorticitythroughthe tilting of baroclinicvorticity is a nonlinear
Furthermore, practical Beltrami solutionsare intrinsically effect and so has little effect on the initiation of midlevel
steady,so mesocyclonedevelopmentis not described,storm rotation discussed in section 2.
propagationis not modeled,and the coordinatesystemmust Incorporationof latent heat releaseand evaporativecool-
be stormrelative. The pressureandwind fieldsare linkedby ing complicatesthe physical interpretation,but it does not
a universalBernoulli relationship.Beltramiflowsare useful invalidate the vorticity decompositionbecause barotropic
for modelingthe dynamic pressuredistributionaround a vorticity is still well defined. Baroclinic vorticity may be
rotatingupdraftandfor explainingobservedcharacteristicsredefinedsimply as the total vorticity minus its barotropic
of mesocyclonessuch as diameter,circulationvalue, low componentso as to include these effects.For nondiffusive
surfacepressurein the inflowregion[Davies-Jones,
1985] flow with horizontally homogeneous initial conditions,
(R. P. Davies-Jonesand H. E. Brooks, manuscriptin prep- equivalentpotentialvorticity (definedin termsof the entropy
aration, 1993),and anticyclonictrajectoriesin cyclonicup- of moist air) is conserved almost exactly in saturatedair and
drafts(owingto the effectsof environmentalwindveering in air in dry adiabaticascent or descentbut only approxi-
withheightovercoming thoseof cyclonicrotation[Klempet matelyin evaporativelycooled air that has descendedin
al., 1981; Lilly, 1986]). unsaturateddowndrafts [Rotunno and Kletnp, 1985]. This
The streamwisevorticityequationshowsthat a parcel's impliesthat the baroclinicvortex linesno longerlie exactly
streamwisevorticitychanges inthreewaysin steadyinviscid in isentropic surfaces.
flow.Streamwisevorticityisamplified
astheparcelaccelerates
andisstretched
alongthestreamline.
It isgenerated
baroclin- 4. THE DEVELOPMENTOF ROTATION NEXT TO THE GROUND
icallyby buoyancytorquesalongits direction
of motion.
Lastly,exchanges
betweencrosswise
andstreamwisevorticity Davies-Jones[1982a, b] realized that in a shearedenvi-
occur in curved flow. Horizontal flow with cycloniccurvature, ronmentwith negligiblebackgroundvertical vorticity, an
e.g.,inflowto a mesocyclone,issimilar
to theflowarounda "in, up, andout" typecirculationdrivenby forcesprimarily
cyclonic fiverbend[Scorer,1978].
Crosswise (speed aloft would fail to produce vertical vorticity close to the
vorticity
shear)is converted into streamwisevorticity(directionalground.Thisconclusion, whichdepends on eddiesbeing
1/•8 MES()CYCLOGENESIS FROM A THEORETICAL PERSPECTIVE

weakto transport
verticalvorticitydownward
againstthe ters,theoutflowmaycomefroma neighboring
storminstead
tl½•w,
hasbeenverifiedin numerical
modelsbyRotunno
and of the storm's own downdraft.Note the contrastin the
Klemp[1985]and Walko[thisvolume].The Beltramimodel origins of midlevel and low-level rotation. The midlevel
illustratesthe point well. Since the vortex linesare coinci- rotationstemsfrombarotropic vorticitypresentonthewarm
dentwiththestreamlines, parcels flowingintotheupdraftat inflowside of the storm,while the near-ground rotation
verylowlevelsdonothavesignificant verticalvorticityuntil originatesfrom baroclinicvorticityon the coolside.
they have ascendeda few kilometers.Otherwise,there Once it develops, the low-level rotation is often more
wouldhaveto be abruptupwardturningof the streamlines, intensethan the midlevelcirculation.According to the
strong pressuregradients, and large vertical velocities next cyclostrophic
approximation theverticalperturbation pres-
to theground.Thesefeaturesactuallydodevelopin axisym- sure gradient force between low levels and midlevels should
metrictornadomodels,in whichambientverticalvorticityis reversesign,resultingin weakeningupdraftandformationof
concentratedby an updraft[Smith and Lexlie, 19791.In this an "occlusion"downdraftwithinthe mesocyclone.
case the vortex core acts as a pipe with cyclostrophic Mesocyclones propagatediscretelyto the rightby a new
balancepreventinginflowthroughthe sides.The "pipe" core periodicallyformingalongthe bulginggustfront to the
dm.ws air int½>it from below, establishing cyclostrophic south(roughlyevery 40 min) as cold air wrapsaroundand
balanceat successivelylower levels. In this manner, the occludesthe old core [Burgesset al., 1982;Johnsonet al.,
vortex builds d(•wnward to the ground. This mechanism 19871. The firstcoretakesmuchlongerto develop(----2hours)
does not operate in a supercell updraft that derives its than do subsequent onesbecauseof the time that elapses
rotation from ambient streamwisevorticity becausethe beforethe downdratland cooloutflowdevelop.Somenew
t•pdral'tis quiteporou.s
to theambientflowandcyclostrophic cores form alotl first, like the first one. Others form almost
balanceis neverestablishedat any level. This is alsoevident simultaneously over their whole depth, consistentwith new
fi't}mI•jcrkncs' circulationtheoremapplied to a horizontal updrallsseizingsomeof the cyclonically rotatingair flowing
fluid circuit, s•ty, a few kilometers across and 100 m above
outof the downdraft,
spinning it up, andadvecting
it upward.
thegr{)und•tta giventime. In an "in, up, andout" type flow We decidedto investigatethe Rotunnoand Klemp mech-
wilh{mtbackground verticalvorticity,any suchcircuitmust anism further because we still did not understand how
be quasi-h{•riz()ntal at all earlier times and should have little rotation developednext to the ground. It seemedto us that
circulati{>n.'I'herct;t)rea downdratt appearsessentialt•,)rthe even thoughthe horizontalvorticity is greatlyenhancedby
development of rotation in the lowest fi2w hundred meters. baroclinicgeneration,tilting in an updraft still would not
l)aviex-,l(mc.s'
[ !982a, b l neglectedbaroclinicvorticity. produce substantialvertical vorticity in the lowest kilome-
}!c sugge•,ted that the downdrafthad the f()llowingroles in ter. Thereforewe undertooka diagnosticstudyof Brookset
low-level rncsocyclogenesJs. First, tilting of h½)rizontal
vol al.'s [this volume] strong-shearsimulation(their case A) to
ticity by the downdra[lproducesverticalvorticJty;second, determinehow vertical vorticityfirst reachesthe ground.
subsidence transportsair with verticalvorticitycloserto the
.surface; third, this air flows out from the downdraft and
5. CIRCULATION ANALYSIS
enters the updraft where it is stretchedvertically; fourth,
convergencebeneath the updraft is enhanced by the outflow. In Brookset al.'s simulation,the horizontalgrid spacingis
He also showedkinematicallythat the flow responsiblefor I km, 100m is the lowest level at which vertical vorticity is
tiltingand concentrating the vortexlinesalsotilts and packs defined, and 50 min is the first time significantvertical
the isentropicsurt•tces,
thusexplainingobservations of strong vorticity(5 x 10-3 s-•) appears
at this level(Figure2).
entropygradientsacrossmesocyclones at the groundand low AdoptingRotunnoand Klemp's [ 1985]approach,we identify
levels (also see Davies-Jones[1985]). Strong tilting of the at 50 mina square material circuit with 4-km sides around
isentropicsurfacesalsofollowsfrom the zero potentialvortic- the "surface" vorticity maximum and a horizontal material
ity constraint.A favorableconditionfor suchtiltingis midlevel, surface enclosed by this circuit. We traced the circuit
cool, dry air overlyingwarm, moist, surface-based air, long backward in time to 35 min. This was done by computing
known to be a propitiousstratificationfor tornadoes. backward trajectories from model data saved every 2 min
Klemœ)and Rotunno[!983] andRotunnoand Klemp [1985] and by usinglinear spatial and temporal interpolation. The
showed that baroclinic vorticity could not be ignored; in maximumsurfacevorticityat 35 min is 5 x 10-4 S-1 an
fact, it is the main source for low-level rotation in the case order of magnitude smaller than at 50 min. We then used
they simulated. Spiraling rain curtains and downdraft form in Bjerknes' circulation theorem, which states that the rate of
the midlevel mesocyclone. In the lowest kilometer, air that increase of the circulation around the circuit is equal to the
has been moderately cooled by rain acquires a great amount work done by buoyancyforceson a hypotheticalparcel that
of streamwise baroclinic vorticity as it subsidesand travels travels once around the circuit instantaneously [Dutton,
southward parallel to closely packed isotherms just behind 1976]. Note that as time decreases, this circuit cannot stay
the outflow boundary. As this relatively cool air is ingested horizontal because this is inconsistent with its nonzero
into the updraft, its vorticity is tilted upward and amplified circulation (a permanently horizontal circuit encloses no
by stretching (see also Weisman and Klemp [1984], Rotunno barotropic vortex lines and has a zero baroclinic generation
[1986], and Klemp [1987]). In some fortuitous storm encoun- term). Thus the parcels that make up the circuit
 DAVIES-JONES AND BROOKS 109

i
Peak Vorticityat 100 m
,,

• - Full rain

• '--No
evap.
ofrain
;• - No rain
1= 2 ,,,
o

1 /

,,, , ,
• ,
0

o 1o 20 30 40 50

Time (rain)

Fig. 2. Peak vertical vorticity at 100-m level versus time for simulations with full Kessler microphysics, without
evaporation of rain, and without rain.

originate from different levels and hence have different vorticity field (Figure 5) contains high values in the rain-
equivalent potential temperatures(0e). Once again, we see cooledair behindthe gust front. The same field at 100 m (not
that the vorticity maximum must lie in a strong0e gradient shown)is very similar but weaker. The horizontal vorticity
[Davies-Jones, 1982a, b, 1985; Rotunno and Klemp, 1985]. in the cool air has the same general direction as the environ-
The hodograph fbr the simulation is shown in Figure 2 mentalvorticity (unlike Rotunno and Klemp's straightline
(case A) of Brooks et al. [this volume]. The storm motion hodograph,splittingstormcase),but is muchlarger(roughly
vector in metersper secondis (6, 4) at 35 min and (12, 4) at 5 times) due partly to baroclinic generation.The vorticity
50 min. Storm-relative winds are northeasterly at the surface vectorscrossthe 0e contoursat large anglesnear the 0e
and veer to southeasterlyat 3 km. The environmental vorticity minimum,indicatingthat equivalentpotentialvorticity is not
vector(samemagnitudeas and90øto left of theshearvector)is conserved well in air that has been cooled considerably by
southward in the lowest few hundred meters. evaporationof rain (see section 3).
Comparisonrunswere madewith the full Kesslermicro- Figure6 showsthe fluid circuit(ABCD) at 35 and50 min.
physics,withoutevaporation of rain,andwithoutformation Air parcelson the east side(BC) and mostof the northand
of raindropsfrom cloud droplets(Figure 2). The peak south sides (CD and AB) at 50 rain originate from lower
vorticityat 100m is 6 or moretimesgreaterat 50 minin the levels (less than 30 m), while those on the west side (DA)
full microphysicscasethan it is in the otherruns.Rotunno descend in the 15-min interval to 100 m from heights of
and Klemp [1985]founda similarresult.Thusstronglow- 400-650m. The northwestpart of the materialsurfaceat 35
level rotation requiresevaporativecooling. minis tilted up the most, so the environmentalvorticityflux
At 50 min the vorticity maximumat 100 m is in strong is greatestthroughthis part of the surface.The correspond-
gradientsof updraftand temperature with the maximum ingcirculationis positiveso that cyclonicvorticityis gener-
updraftto the northeast,downdrafts to thenorthwest and ated as the surface levels out at 50 min. Note that since the
southwest, andthecoldestair to thewest(Figures 3 and4). surfaceis not folded at 50 min, the elevatednorthwestpart of
Noticethatat thislevelthereispositivevorticityevenwithin the surface must slide down "feet first" rather than "head
the downdrafts. Storm-relative flow converginginto the first" (i.e., the higher points must move southwardmore
vorticitymaximumis northeasterly Most slowlythan the lower points). The area of the matehal
to northwesterly.
ofthepositive
contribution
tothecirculation thefluid surfaceshrinks by a factor of 10 in the 15-min period,
around
circuit is from windson the west (cool)side, indicatingthat implying
a tenfoldincreasein theaveragevorticitynormalto
the sourceof rotation at this level is in the cool air. The the surfaceif only barotropiceffectsare considered.
vorticity
maximum onthegustfrontandleading The circulation around the fluid circuit increasesby 70%
islocated
(Figure5). The300m horizontalfrom 35 to 40 min as the parcelson the side DA sink,
edgeof the 0egradient
 FROM A THEORETICAL PERSPECTIVE

VerticalVelocity@ 100 m @ 50 min (Light) of these descending parcels become more anticyclonic at
VerticalVorticity@ 100 m @ 50 min (Bold) first, then less anticyclonic, and actually cyclonic in the
lowest 50-125 m of their descent.
How is cyclonic vorticity generated in a downdraft? If the
downdraft and outflow are axisymmetric, the horizontal
vorticity (both bareclinic and baretropic) is crosswise and
the vertical vorticity zero because the tilting term vanishes
by symmetry. If purely crosswise horizontal vorticity enters
a downdraft from one direction, tilting produces a vortex
pair with the cyclonic vortex on the left side (looking
downwind), as shown in Figure 9. This case does not apply
here because the vorticity and storm-relative velocity vec-
tors tire generally parallel in the cool air (cf. Figures 4 and 5).
If, on the other hand, purely streamwise vorticity enters a
downdraft and baroclinity is negligible, the vortex lines
(which are frozen into the fluid) and trajectories coincide, so
the downdraft is anticyclonic (Figure 9). We are able to
explain how the vertical spin of parcels reverses during
descent only by including bareclinic generation of stream-
wise vorticity. Consider, for example, a mass ot' evapora-
tivcly cooled air that descends through the mesocyclone on
its north to northwest sides and spreadsout "feet first" (like
it gravity current) toward the south nearly parallel to the
isotherms with the warm air to the east {Figure 9). Such a
flow occurs on the left side of cold-air outflow that advances
cyclonically around a mesocyclone.Normally, a gravity
CI = 0.001/s

!.'ig. t. Vertical volt)city;tnd vcrlicul v()rticily ticidsat 100-mlevel

•,t'IcrS•)n•in. C()nt()t•r inlcrvals are given •t lower lett. Zero-vertical-
HorizontalVelocity@ 100 m @ 50 min
v()r!icity c()nt{)•r is st•pprcsscd.A!s() shown is (>utlinc•)I'4 km/ 4 Perturbation PotentialTemperature
km I•()x ;•rt)t•r'•dv()rticily nutxin•un]. @ 100 m @ 50 min

remains m{)rcor lessconstant dt•ringthe next 10 rain (Figure

7). 'l'hc time intcgrulo1'Bjcrknes' thc(>rcmis
I t, ,',l',l 'l :/ ,,"'/..."/,.?'
, ,

. • •' t ' ' ,' /-" .'

l'(t) = I'(t()) + bk . dl dt (1) t , ', , /, ,/ , .,.. ,'/,' :/
', ', • : •' ,, ,.'

wt•crc l'(t) is the circulation, b is buoyancy, and t 0 is an

initial time (35 min here). We refer to the last term as the
buoyantlygeneratedcirculation.Direct computationo1'this '" I
term shows that it accounts quite well for the increase in
circulation{Figure7). Thus we are confidentthat our trajec-
tory computationsare fairly accurate.

6. VO?,TICITY CHANGES IN INDIVIDUAl. PARCELS

The circulation analysis, although enlightening,does not
portray the acquisitionof verticalvorticity by individual
• • .."':',,'t'..'•..".,"..",
, .........
parcels.Therefore,to gainfurtherinsightsintothedevelop- ?..."
...?':,,,.:':t,'t'..".'.,."'
/' I
ment of low-level rotation, we examined how the vertical
vorticitiesof the labeledparcelsin Figure 6 changewith time
from 35 to 50 min. As expected,the risingparcelson the east
sideacquirelittle verticalvorticity,while the subsiding
ones C I = 0.5 K • 20 m/s
on the west sideshowappreciablegains.Graphsof vertical
vorticityversusheightfor three parcelson the west side •i•. 4. Asin•i•ur• • butforpotential t•m•mtu•e andhorizontal
DHA of the fluid circuit {Figure 8) show that the vorticities w•d •Ids. Scal• fo• vectors •s •iv•n at 1ow•
 DAVIES-JONES AND BROOKS 111

currentmovesnormalto the isotherms;however,the cold Fluid Circuit @ 35 rain and 50 min

air in thiscasehasconsiderable
momentum
alongthe
isotherms,
asit mustin a quasi-steady
storm.It descends
in
a cyclonicspiralaroundthe rearof the mesocyclone
be-
cause,as it falls, it experiences
bothenvironmental
winds Heights in m
thatback(withdecreasing height)andthecyclonicrotation
of the mesocyclone (cf. anticyclonic
trajectoriesin the 21 19

cyclonicupdraft,section2). Strong,storm-relative inflow 132,•,•"•"•24

windsalsoholdtheoutflowbackundertheupdraft[Brooks 57y "'496
et al., thisvolume].Baroclinic
generationproduces a circu- 637,•
r, 115
lationnormalto theisotherms, sotheflowactually is not 11
confinedto two dimensions as depictedin the simplified
schematic.As the air subsides,the vortex linesturn down-
warddueto thebarotropic "frozenvortexlines"effect,but 54 17'

with lessinclinationthan the trajectories.This is because

horizontalsouthward vorticityisbeinggenerated continually •D G c •'• 32
by thebarocliniceffect(whichintroduces slippage between
the fluid and vortex lines). Becausethe gravitycurrent
subsidesfeet first and the vortex lines now crossthe stream-
linesfromlowerto higherones,thebarotropic effectactsto
turn the vortex lines upward even during descent.The
barocliniceffectactsto increasehorizontalvorticityfurther Circuit at 50 Minutes
All at 100 m
but doesnot controlthe signof verticalvorticity.Thusair
with cyclonicvorticityappearscloseto the ground.As this

Fig. 6. Fluid circuit AEBFCGDHA at 35 and 50 min. Numbers

PerturbationEquiv.PotentialTemperature along circuit denote heights in meters. At 50 rain, O is center of
square, A, B, C, and D are corners, and E, F, G, and H are
@ 100 m @ 50 min midpoints of sides.
HorizontalVorticity@ 300 m @ 50 min

air passesfrom the downdraft into the updraft, its cyclonic

spinis greatlyamplifiedby vertical stretching[Da,'ies-Jo•es,
1982a]. This mechanism explains why the horizontal con-
vergenceterm in the vertical vorticity equation is large next
to the ground in the low-level mesocyclone. If the tilting of
vorticity were confined to the updraft, then there would be
no ground-level vertical vorticity to stretch!

7. TORNADOGENESIS

Tornadogenesisis the last of the three stages in the

"spin-up" process (section 1). With the development of
ground-levelrotation in an updraft, it is temptingto explain
mesocyclone tornadoes simply as the result of frictional
interaction of the mesocyclone with the ground, i.e., as
end-wall vortices [Rotunno, 1986]. However, this explains
neither the development of small-scale regions of large
vorticity ("tornado cyclones") within the mesocyclonein
high-resolutionsimulations with a free-slip lower boundary
[Wicker and Wilhehnson, this volume] nor the foreshadow-
ing of some violent tornadoes by tornadic vortex signatures
i (TVSs) that first appear aloft in Doppler radar observations
Environmental
Vorticity 0.025/s [Rotunno, this volume] and are, perhaps, indications of
tornado cyclones rather than tornadoes themselves. How-
C!=0.5K ever, surface friction does play an important role in the
Fig. 5. Sameas Figure4 but for equivalentpotentialtemperature "spin-up" of tornado cyclones into tornadoes [Wicker and
at 100 m and horizontal vorticity at 300 m. Wilhelmson, this volume] and in the flow structure of
!12 MIœS()CY('L()GENESIS FROM A THEORETICAL PERSPECTIVE

Generated/Actual Circulation
0.3

0.3

- Buoyant Circ.
-Circulation

0,2

0.15-q •
35 40 45 50
Time (min)

!:ig. 7. ("irculati(•n(lcl'tscale)and buoym•tlygenerated

circulationsince35 min(left scale,0.15 x 1()•'5m2 s-•) of
fluid circus! ver•,us time.

does. !)avie,,;-,Iottc,,,'[1982a, h] and Walko [this volume] con- why tornadoesare located on the side of the updraft nearestto
elude thtd tornadogenesis is caused by concentration by a the downdraft:and why anticyclonictornadoes,when present,
c{mvcrgentwind tield of vertical vorticity advected at very low are generally in a flanking line updraft further south along the
levels into thc updraft l¾omthe downdraft. This would explain gust front on the right side of the rear flank downdraft (where

Vorticity(t)vs. Height(t)(West Side Parcels)

z (km)

0.1 ø'ø

-2 -1 0 1 2 3 4 5
Vorticityx .001 /s
Fig. 8. Vertical vorticityversusheightof westsideparcelsA, H, andD over 35- to 50-raintime
 DAVIES-JONES AND BROOKS 113

CYCLONIC VORTICITY GENERATION IN DOWNDRAFT updraft. However, this is inconsistentwith the tornado being
CROSSWISE VORTICITY embedded in a low-level mesocyclone, which derives its
vorticity from a cool, rainy downdraft to the north. Walko
[this volume] makes another objection, based on the initial
small scale and weak circulations of the flanking-line vorti-
-•4•
•,, ..•...,,.,'•
' '• ISENTROPIC ces. Note that none of the above processes explains the
•'5,•"" •C SURFACE caseswhere the TVS descends from midlevels to the ground
in 20-30 min, instead of developing near the surface first or
at the same time as at low to mid levels.
At first sight, landspouts and most waterspouts appear to
have completely different origins from mesocyclonetorna-
does.They typicallyoccur in low-shearenvironments,along
STREAMWISE BAROTROPIC VORTICITY
almoststationary,preexisting,frontlike boundariesearly in
a developingstorm's life cycle, and only during afternoons
with steep low-level lapse rates. However, like tornadoes,
neighboringheavy rain showers appear critical to their
formation. Wilczaket al. [1992] suggestthat the dissimilari-
ties with misocyclone tornadoes are not as great as they
seem. The low-level, misosca!e (<4 km diameter) vortex
resultingfrom horizontal shearinginstability in a deep con-
vective boundary layer and the preexistingpseudofrontact
STREAMWISE VORTICITY WITH BAROCLINITY
as surrogatesfor the midlevel mesocycloneand gustfront of
supercelIs,respectively.At leastsomeof the spout'svortic-
----> BAROCLINIC VORTIClTY
GENERATION
ity appears to be derived from tilting of baroclinically
generatedvorticity on the cool-air side of the misocyclone.
Clearer understandingof tornadogenesisand cyclic meso-
cyclogenesisawaits better numericalcloud models. The
waters may be muddied by there being several distinct
mechanisms and combinations thereof with the mixture of
processesvarying t¾omcase to case.

Fig. 9. Schematicdiagramshowinghow cyclonicvorticitymaybe

8. CONCLUSIONS
generatedfrom tilting of barotropicand baroclinichorizontalvor-
ticity in a downdraft.For crosswisebarotropicvorticity(top),tilting
The initial midlevel mesocycloneoriginatesfrom tilting by
producescyclonic(anticyclonic)vorticityon the left (right) side,
lookingdownwindin storm-relativereferenceframe.In the stream- the updraft of environmental streamwisevorticity. Near-
wisebarotropicvorticitywith steadyflow(middle),tiltingproduces groundrotationawaitsdevelopmentof a cooldowndraft.In
a purely anticyclonicdowndraft.In the streamwise vorticitycase thepresentcaseit derivesfrom bothbaroclinicandbarotro-
with flow to the right of the horizontalbuoyancygradientand a picvorticityon the cool-airsideof the storm.In thisregion
southerlyshearcomponentas shown(bottom),a combination of
tiltingand baroclinicgeneration causesthe vorticityof parcelsto thebaroclinicvorticityappearsstrongerthan, but in the same
changefromanticyclonic to cyclonicwhilestilldescending. directionas, the barotropicvorticity becauseof stronglow-
level curvatureof the hodograph.In the straighthodograph
casesof Rotunnoand Klemp [1985]and Walko [thisvolume]
tiltingof horizontalbaroclinicandbarotropic vorticitypro- thetwovorticitycomponents werein opposition in thecoldair,
ducesanticyclonicvorticityin our simulation). with low-levelrotation stemmingfrom baroclinicvorticityin
Anotherpossible mechanism is horizontal
shearing insta- the former and barotropicvorticity in the latter.
bilitycausing the verticalsheetof vorticityat a wind-shift Our mechanism for low-level rotation is quite similar to
vortices,whicharestretched thatdescribedby Rotl•nnoand Kletnp[1985],Weismanand
lineto roll up intoindividual
verticallyby convective updraftsif overhead [Davies-JonesKlemp[1984,p. 2489],Rotunno[1986,p. 425],andKlemp
andKessler,1974;Brandes, 1977].Thisprocess the [1987,p. 396]. There are, however, subtlebut important
explains
formationof multiplevorticesin tornadosimulators (anala- differences.Rotunnoand Klemp' s [ 1985,p. 281] description,
gousto multivortex tornadoesor multipletornadoes around restatedin the other articles, is as follows: "Thus, as the air
therimof a mesocyclone core),gustfronttornadoes, and,at approaches fromthe northeast,it firstacquireshorizontal
leastpartly,landspouts
andwaterspouts and vorticitydirectedtowardsthe southwestfrom baroclinic
[Wakimoto
Wilson,1989].WakimotoandWilsonsuggested thatmeso- generationalongthe cold air boundary.This horizontal
cyclonetornadoesmightoriginate
asvorticesthataregen- vorticityis thentilted upwardto producecyclonicvertical
eratedunderflanking-line
clouds
byshearing and vorticityastheair encounters
instability, theupdraft.Finally,theair is
movewiththeirparent
clouds intothemain subjected
astheymerge to intensestretching..." In contrast,we
114 MES(>C¾('LOGENESIS FROM A TIIE(•RE'I'ICAI, PERSPECTIVE

that lilting of horizontal vorticity (either barotropic or ba- 14th Con.[•,rence

on SevereLocal Storms, pp. 216-219, American
roclinic) toward the vertical by an updraft does not produce Meteorological Society, Boston, Mass., 1985.
rotation very near the ground because vertical vorticity is Davies-Jones,
R. P., Tornadodynamics,in Thunderstorm Morphol-
ogy and Dynamic's,2nd ed., editedby E. Kessler,chap. 10, pp.
generated in rising air. We find in our simulation that the first 197-236, University of Oklahoma Press, Norman, 1986.
cyclonic vorticity at 100 m is generated by tilting of quasi- Davies-Jones,R. P., Thunderstorm,in Encyclopediaof Scienceand
streamwise vorticity in descending, evaporatively cooled Technology, 7th ed., McGraw-Hill, New York, 1992a.
air. This air originates from ---500 m (probably higher at later Davies-Jones,R. P., Tornado, in Encyclopedia of Science and
Technology, 7th ed., McGraw-Hill, New York, 1992b.
times) and is on the east edge of the cool air outflow. Its Davies-Jones,R. P., and E. Kessler, Tornadoes, in Weather and
vorticity changes sign from anticyclonic to cyclonic below (•limate Modt•fication,edited by W. N. Hess, chap. 16, pp.
25t} m, while it is still in the downdraft, owing to simulta- 552-595, John Wiley, New York, 1974.
nc(>usbaroclinic generation and tilting of vorticity. The air is l)avics-Jones,R. P., I). W. Burgess,and M.P. Foster, Test of
then entrained into the southwest side of the main updraft hellcity as a tornado forecastparameter, in Preprints, 16th Con-
fi'ren('c on Severe Local Storms, pp. 588-592, American Meteo-
below 1(}()m. Its cyclonic vorticity is ampitiled by the conver- rological Society, Boston, Mass., 1990.
gence term, with a minor contribution from t'urthertilting. Note l)avics-Joncs, R., R. Rabin, and K. Brewster, A short-term forecast
tha! nothing in this scenario is inconsistentwith the analysisof modcl for thunderstormrotation, in Proceedings,Nowcasting II
numerical results presented by Rotunno and Klcmp. Symposium, Norrk6ping, Sweden, Eur. Space Agency Spec.
t'uhl. ESA SP-208, 367-371, 1984.
'l'hc production and downward transport of cyclonic vor~ I)uttem, J. A., ?'he ('ea.vele,vsWind, 579 pp., McGraw-Hill, New
tictry in the downdraft is important lk)r t{•rnadogcnesisbc- York, 1976.
cat•se of !he supply of cyclonically spinning air practically to Johnson, K. W., P.S. Ray, B.C. Johnson, and R. P. Davies-Jones,
ground level. 'l'ornadoes may result as this air tlows intt• the ½)bservationsrelated to the rotational dynamics of the 20 May
1977 tornadic storms, Mon. Weather Rev., 115, 2463-2478, 1987.
fi•ot of lhc updraft and undergoes considerable vertical
K!cmp, J. B., Dynamics of tornadic thunderstorms, Annt•. Rev.
stretching. Fluid Mech., /9, 369-402, 1987.
Klcn•p, .I.B., and R. Rotunno, A study of the tornadic region within
a supercell thunderstorn•, J. Atmos. Sci., 40, 359-377, 1983.
Acknowit,d/•,menls. We thank ('harlcs A. !)•}swcll 111f{•r several Klcn•p, J. B., R. B. Wilhelmson, and P.S. Ray, Observed and
helpful ctisclissionsduring the course of this w{•rk, which was done numerically .simulated structure of a mature supercell thunder-
while one •ffthe authors {tt.t';.B.) held a National Research C{mncil- storm,,/. Almo.¾. Sci., 38, 1558--1580, 1981.
N()AA Research Associateship. We gratct'ully •tcknowledgc Bt•b l,i!ly, I). K., The development and maintenance of rotation in
Walko and an anonymous reviewer fi•r ptfinting out places where wc convective storms, in 7'opic.¾ in A tmospherir' and Oceanographic
needed t() bolster our argtunenls. ()t•r thanks 1() .loan Kimpcl tk)r ,5'cienr'es:Intense Atmo.v•heric Vortir'es, edited by l.....Bengtsson
drat•ing Figure 9. 'l'he computer simulalions wcrc rnadc ()n the and M..I. l,ighthill, pp. 149-160, Springer-Verlag, New York, 1982.
('ray-2 al the National (•'cn•cr for Supcrc()mptlling Applic;tli()ns l.illy, D. K., The structure, energetics and propagation of rotating
(NCSA), and some were supported by NSf: gran! A•I'M-87-()(1778. convective storms, II, l lelicity and storm stabilization, J. A tmos.
,¾('i., 43, 12(v-140, 1986.
Rabin, R. M., and R. P. Davies-Jones, Atmospheric structure ahead
RI';!'ERt';N('ES ot' thunderstorms, [:'ndea•'otlr, /0(I), 20-27, 1986.
Rotunno, R., Tornadoes and tornadogenesis, in Mesoscale Meteo-
Barnes, S. 1,., Some aspects ot'a severe, right-moving thunderstorm rology and t::orcca.•'ting,edited by P.S. Ray, chap. 18, pp.
deduced from mesonetwork rawinsonde observations, J. 414-436, American Meteorological Society, Boston, Mass., 1986.
Sci., 27, 634-(>48, 11970. Rotunno, R., Supercell thunderstorm modeling and theory, this
Brandes, E. A., Gust front evolution and tornado genesisas viewed volume.
by Doppler radar, J. Appl. Meteorol., 16,333-338, 1977. Rotunno, R., and J. B. Klemp, The influence of the shear-induced
Brooks, H. E., C. A. Doswell I1!, and R. P. Davies-Jones, Envi- pressuregradient on thunderstormmotion, Mort. Weather Rev.,
ronmental hellcity and the maintenance and evolution of low-evel 110, 136-151, 1982.
mesocyclones, this volume. Rotunno, R., and J. B. Klemp, On the rotation and propagation of
Browning, K. A., and C. R. Landry, Airflow within a tornadic simulatedsupercellthunderstorms,J. Atmos. Sci., 42,271-292,
storm, in Preprints, loth Weather Radar ConJkrence, pp. 116- 1985.
122, American Meteorological Society, Boston, Mass., 1963. Scorer, R. S., Environmental Aerodynamics, 488 pp., Ellis-
Burgess, D. W., V. T. Wood, and R. A. Brown, Mesocyclone Horwood, Chichester, England, 1978.
evolution statistics, in Preprints, 12th ConJkrence on Severe Smith, R. K., and L. M. Leslie, A numerical study of tornadogen-
Local Storms, pp. 422-424, American Meteorological Society, esis in a rotating thunderstorm,Q. J. R. Meteorol. Sot., 105,
Boston, Mass., 1982. 107-127, 1979.
Davies-Jones, R. P., Observational and theoretical aspects of tor- Wakimoto, R. M., and J. W. Wilson, Non-supercell tornadoes,
nadogenesis,in Topics in Atmospheric and Oceanographic Sci- Mon. Weather Rev., 117, 11113-1140, 1989.
ences: Intense Atmospheric Vortices, edited by L. Bengtssonand Walko, R. L., Tornado spin-upbeneatha convectivecell: Required
J. Lighthill, pp. 175-189, Springer-Verlag, New York, 1982a. basicstructureof the near-fieldboundarylayer winds, this volume.
Davies-Jones, R. P., A new look at the vorticity equation with Weisman,M. L., and J. B. Klemp, The structureand classification
application to tornadogenesis,in Preprints, 12th Conferenceon of numericallysimulatedconvectivestormsin directionallyvary-
Severe Local Storms, pp. 249-252, American Meteorological ing wind shears,Mon. WeatherRev., 112, 2479-2498, 1984.
Society, Boston, Mass., 1982b. Wicker, L. J., and R. B. Wilhelmson, Numerical simulation of
Davies-Jones, R. P., Streamwise vorticity: The origin of updraft tornadogenesis within a supercellthunderstorm,this volume.
rotation in supercell storms, J. Atmos. Sci., 41, 2991-3006, 1984. Wilczak, J. M., T. W. Christian, D. E. Wolfe, and R. J. Zamora,
Davies-Jones, R. P., Dynamical interaction between an isolated Observations of a Colorado tornado, I, Mesoscale environment
convective cell and a veering environmental wind, in Preprints, and tornadogenesis,
Mon. WeatherRev., 120, 497-520,
 Discussion

MORRIS WEISMAN, SESSIONCHAIR

National Center for Atmospheric Research

PAPER B 1 PAPER B2

Presenter, Rich Rotunno, National Center for Atmo- Presenter, Lou Wicker, University of Illinois [Wicker and
sphericResearch[Rotunno, this volume, Supercellthunder- Wi[hebnson, this volume, Numerical simulation of torna-
storm modeling and theory] dogenesiswithin a supercell thunderstorm]
(Doug Lilly, University of Oklahoma) You still believe in (Don Burgess, National Severe Storms Laboratory) This
[storm] splitting and vortex developmentwithout rain, yet all questionis for both Rich [Rotunno] and Lou. I have always
laboratory experiments in unidirectional shear produce line beenimpressedby the agreementbetween observationsand
convection. To my knowledge, splitting never occurs in numerical modeling for the Del City storm, but there are
either laboratory experiments or numerical simulations of some observational data sets that are somewhat different
them. from Del City. I want to make sure that the numerical
modelingcommunity does not quit with the Del City case.
(Rotunno) The critical factor missing in the laboratory ex- l'd like to see some of these other storms that have different
periments, which are essentially Rayleigh-Benard convec-
features analyzed. It's important that we look at a spectrum
tion with shear, and the distinct feature of the numerical
of stormsbecauseI definitely think that there is a spectrum
cloud simulations are conditional instability. Air has to be of storms out there.
lifted to its level of free convection before updraftsform. In
Figure 13, there is lower pressureon the southflank of the (Wicker) The honest reason why I am showing Del City is
updraft. All the air outsidethe updraftis not goingto rise on becauseother cases didn't work as well. For example, in the
its own without being lifted. It needs a lift [provided by April 3, 1964,casewhich I did for my thesis,thereis similar
upwardpressuregradientforces]to overcomethe inversionevolution, but the mesocyclone is half the diameter of the
one in the Del City case. So I didn't run the April 3 case
that is there, and once the air penetratesthe inversion, it
because1 neededtwice the grid resolution in order to resolve
becomesbuoyant.It risesandregenerates the updrafton the
the same structures and I didn't have the computer time to
south side. Thus the updraft propagatesto the right of the
do this. We are definitely going to look at different cases.
mean shear. You need conditional instability; with absolute
Some of the other cases, which we expected to produce
instability, convectionwould go up everywhere.
intenselow-level mesocyclonesbecausestorm-relativehe-
(Joe Golden, National Oceanicand AtmosphericAdminis- llcity is large, don't work as well as Del City.
tration,Officeof theChiefScientist)
The modelresultsof the
(Brian Fiediet, University of Oklahoma) I noticed that the
last 10 yearsare incrediblyimpressive.In thefuture,canwe vortex that you are quick to call a tornado vortex looks
usea modelto predictthe probability thata givensupercell like...
willproducea tornadoin thenext1to 3 hours,assuming
that
we havedatafromprofilersandNEXRAD radarsto initial- (Wicker) In this particularcase,I wouldn't call it a tornado
ize the model? vortex.

(Rotunno)
I can'tpredictthat[general I thinkthat (Fiediet) Okay, but it looked like the safestplaceto be in a
laughter].
youhaveto try andsee.It's notoutof thequestion. storm. It looks more like a stagnationpoint or an F minus i
tornado. How do you distinguishbetween a tornado vortex
The Tornado:Its Structure,Dynamics,
Prediction,
andHazards. and an inflectionpoint instability?Maybe the vortex has
GeophysicalMonograph79 nothingto do with the tornado that actually occurred.It
Thispaperisnotsubject to U.S.copyright.
Published
in1993
bythe
American Geophysical Union. couldbe a barotropicinstabilityalonga gustfront.
(Wicker) •! c(•uld be. Certainly such a mechanism could relationshipbetween that and your result is subjectto further
bc importat&t at low levels. However, this vortex has study.
time and height continuity. It extends to levels well
ab(•vc the gu•l front. So the low-level vortex may develop
P^P[?:RB4
t'r½•man instability along the boundary. But it must con-
ncct with l•atures aloft that are not connected with the gust
Presenter, Harold Brooks, National Severe Storms Labora-
tyrant.•l•his particular vorticity maximum extended upward
tory [Brooks et al., this volume, Environmental hellcity and
4-5 km int• lhc slorm. In the April 3 simulation, vorticity
the maintenance and evolution of low-level mesocyclones]
greaterthan ().I $ • wa• pre•entup to 3.5 kin. So there
is definitely st)me vertical c(•ntinuity associated with lHowie Bluestein, University of Oklahoma) According to
these v{•rticcs. your argument, if there is a very strong inflow low, then the
gust front won't surge out. But it seems to me that if the gust
!('t•airl I believe that your vortex is within about hall' a
fi'ont has a strong high behind it and if the inflow is very
kil{)•ctcr t•r s{• ot' the observed It)cation ot' many tornadoes,
intense, there will be a strong pressure gradient three which
st) there is •tm•c c•nsixtcncy lhcrc.
will accelerate the gust t¾ontrapidly.
(Rict• Rolt•i•nt•,Nittionltl ('enter tk)r Atmospheric Research)
(Brooks) It dependson the balance between the two effects.
'l'l•c lhrtisl (•t' lhc qt•chtit•n is whether t>r not lhc simulated
If the high behind the gust front is weak, the gust fi-ont won't
vt•rlicity •axi•tim cCttlltlsti•rnad{•. I d{)n'l think that you
advance. If this high is strong, the gust front will move
rapidly.
IWickcr) N{), I don't think s•. l)ctcrmining whether a
(Bluestein) But if the inflow low is very strong and the high
v•rticily •;•xi•nttn• in the lnt)dcl is really a tornado is
behind the gust tYontis very weak, won't the gust front also
surge out?
(Rt)lt•nn<•/('an I a•hwcr [){•n Bt•rgcss' ClUCSti()n?
(Bob Davies-Jones, National Severe Storms Laboratory)
l('h;•irl N•)! [()tit Not if the air parcels ascend at the gust t¾ont.The pressure
gradientforce acts on parcels,which ascendat the gustfront
rather than passingthrough it.
P^Pf.,t• B 3
(Chair) I have a bias against helicity because of its storm-
l'rcscntcr, tl•b Walkt•, (1o!{•rad() State [Jnivcrsity[Waiko, relative nature. Ifone takes the divergence ol'the momentum
!his vt}l•nc, 'l'{)rnad{• spin-t•p bcnc;•.tha convective cell: equations,onecanderiveatdiagnosticpressureequationthat
Rcqt•irctt basic structt•rc ()f the ncar-ticld boundary layer describesthe pressurefield. The tk)rcingfunction in this
wind s ] equationfor pressureis dependenton buoyancyand velocity
gradientsbut is totally independentof any terms that are
litdan l:icdlcr, University of ()klahot•a,• Based on the ther- dependenton referenceframe. Your Bernoullirelationship
modynamicspeedlimit and the depth of the layer, what is [tbr pressurein termsof velocity]is onlycorrectin a steady
the Reynolds nun-lberof your simulations? state reference frame.

(Walko) I haven't computedit, but basedon the numberof (Brooks) We are assumingthat it is a steadystate storm. In
grid points, it is probably around 1000. a ground-relative
frame,thereis an extraterm, O&/Ot[O is
½Ficdler)Wc didn't have any luck making tornadoesin 3-D the velocitypotential],in the Bernoulliequation.In the
until we used a Re above about 2000. Below that, we had to storm-relativeframe, O0/Ot drops out automatically[be-
insert vertical vorticity, but if Re is high enough,we could cause O/at •- 0].
get from horizontalvorticityto shearalonga gustfront to a (Chair) But when you go back to the original diagnostic
tornado.
pressureequation,thereis no storm-relative
term. I seea
contradiction there.
(Walko) That's very interesting. It seemsto me that these
simulationswere generally robust, that I could get a vortex
(Davies-Jones)The diagnostic pressureequationis a Pois-
providedthat I put the downdraftin the correctsector,etc. son'sequationinvolvingsecondderivativesof the pressure
It took sometuning,but thenI couldgeta tornadolikevortex field,notpressure(p) itself.It can'tbe integratedsimplyby
with very low Re, even a few hundred. assuming thatV2p o: -p, as is frequently done,because
(Rich Rotunno, National Center for Atmospheric Re- sometime-dependent termsreappearafter integration.
search)The circulationin the Rotunnoand Klemp simula- (Chair) I think that if you make the same steady state
tion was baroclinicallygenerated. Analysis showed that assumptions, though,thereis still a contradictionthere.
the curve around the vortex at the time of maximum vortic-
ity had circulationproducedby the baroclinicgenera- (RichRotunno,NationalCenterfor Atmospheric
Research)
tion term. This is the mechanismin the full cloudmodel. The The virtue of the diagnosticpressureequationis that
 WEISMAN 117

time dependency,d(V ßv)/dt, drops out. Numerical models PAPER B5

usethis equation[to recover pressureat eachtime step].The
Presenter, Robert Davies-Jones, National Severe Storms
time dependencydoesn't reappear.
Laboratory [Davies-Jonesand Brooks, this volume, Meso-
(Brooks) We'll let the two theoreticiansfight this out! cyclogenesisfrom a theoretical perspective]
(Davies-Jones)
Youarereallyconsidering
V2p,notp itseft;
p is (V. G. Blanchette,retiredprofessional
engineer)Your analyses
found by integrating the forcing function over the whole [of the numericalmodelresults]were for 100-melevation.Do
volume. Admittedly, it is weightedby [inverse]distance,but theyalsoapplyat muchhigherelevations,say,2-4 km?
this integration completely changesthe picture. It gives an
inflow low at the ground which is forced entirely from above (Davies-Jones)
At 2-4 km, there is a lot of vorticitythat has
like Dudhia and Moncrieff [J. Atmos. Sci., 46, 3363-3391, 1989] beengeneratedon the warm sideof the storm.The analyses
said. presentedhereapplyjust to verylow levels,whichare the
importantonesfor investigatinghow rotationgetsdownto
(Doug Lilly, University of Oklahoma) You said that storm the ground.
motion cannot be predicted. I guessyou mean by that from
environmental conditions only or that the evolution as such (RichRotunno,NationalCenterfor AtmosphericResearch)
is maybe too complicatedto predict. But, surely,if you have Where does the TVS fit into all of this?
observationsof the storm at a given time, the equationsof
motion are capable of predictingthe motion. (Davies-Jones)
I don't know. The TVS mystifiesme be-
(Brooks) I'm sayingthat you cannot predict storm motion cause,accordingto Doppler radar observations,
it begins
from the environmental conditions alone. aloft, and I don't understand
 Observations
andSimulations
of Hurricane-Spawned
TornadicStorms
EUGENE W. MCCAUL, JR.

UniversitiesSpace ResearchAssociation,Huntsville, Alabatna 35806

1. INTRODUCTION adiabatic core cooling experienced by air parcels entering

the landfalling hurricane center subsequent to its loss of
The occurrenceof tornadoeswithin landfallingtropical contact with the warm ocean surface. GenttT[1983] argued
cyclones has been recognizedfor a long time, with some
that a more important factor in the generation of the strong
reports extant from as early as 1811 [TannehiIl, 1950].
shearswas the increase in surface drag experienced by the
However, the widespreadrain, low clouds, and generally low level flow as it encounters a coastline.
poor visibility attendingmost landfallingtropicalcyclones
While most of the early research on hurricane-spawned
has hindered direct observations of these tornadoes and
tornadoesfocused on their temporal and spatial patterns of
impeded progress in understanding the character of their
occurrencewithin the parent hurricane, little progresswas
parent convective storms. Still, as more and more reports
made regardingthe specific characteristicsof the tornadoes
have become available, it has becomepossibleto character-
until recently. Stiegler and Fujita [1982] examined the dam-
ize the tornadoes and tornadic storms with growing accu-
age from some of Hurricane Allen's (1980) tornadoes in
racy.
Texas and observed strong evidence of multiple-vortex
in the early studies, limited data forced investigatorsto
structure reminiscent of intense Midwestern tornadoes. The
consider only tornadoes spawned by tropical cyclones of
first direct photographicconfirmation of this finding was
full-fledged hurricane intensity. Malkin and Galway [1953]
given by McCauI [1987], who documentedtwo of the torna-
noted mounting evidence that most hurricane-spawnedtor-
does spawnedby 1985's Hurricane Danny in northern Ala-
nadoeswere weaker than their counterpartsfrom the United
States Great Plains and Midwest. Smith [1965] identified the
bama(Figure I). McCaul [ 1987]alsodiscoveredthat someof
Danny's tornadic storms exhibited persistent distinctive
right-front (RF) quadrant of the hurricane as being the
radar echoes (Figure 2) resembling those of small supercell
t'hvoredarea for tornado development, while Pearson and
thunderstorms.Wind profiles obtained from nearby sound-
Sadowski [1965] found that most of the tornadoesoccurred
outside the area of hurricane-force surface winds. Hill et al.
ings indicated the presence of large amounts of helicity
[Lilkv, 1986]and streamwisevorticity [Davies-Jones,i984],
[1966]suggested that dry intrusionsenteringthe circulations
parameterswhich by that time had been found to be related
of landfallinghurricanesmight contributeto the develop-
to the developmentof rotation in the updraftsof stormsthat
ment of tornadic storms by augmentingthe convective occur in the Midwest and Great Plains of the United States.
instability.The sameauthorsalsonotedthe tendencyfor The accumulation of new observations and severe storm
greaternumbersof tornadoesto occurin the moreintense
conceptualmodelsthat became available by the late 1980s
hurricanes.
prompteda comprehensive new studyof hurricane-spawned
NovlanandGray[1974]employedcompositing techniques tornadoenvironments,along with the first numericalsimu-
on a set of datacomprisinga 25-yearperiodto arriveat the lationsof hurricane-spawnedtornadic convection.This pa-
first representativelook at upperair conditionsaccompany- perreviewsthe resultsof theserecentresearchefforts.
ing hurricane-spawned tornadoes.They documented the With respectto observationsthereremainsa lackof radar
presenceof very stronglow level verticalshearand the data suitable for use in investigating individual hurricane-
relative absenceof thermodynamicinstabilityin the com-
spawned tornadicstorms.As a consequence, recentefforts
positehurricanetornadoenvironment.
The strongvertical have concentrated on extracting more information about
shearswere attributed to thermal wind adjustmentto the
storm environments from already existing standard data
sources.McCaul [ 1991], hereinafter abbreviatedas M91, has
examined39 years of United States soundingdata taken in
TheTornado:Its Structure,Dynamics,
Prediction,
andHazards.
GeophysicalMonograph79
proximity to hurricane-spawnedtornadoesand has con-
Copyright1993by theAmericanGeophysical
Union. structedcompositetemperature,moisture,andwindprofiles
120 HURRICANE-SPAWNED TORNADIC STORMS

Fig. I. Tornado spawned by remnants o! Ilurricane l)anny is photographedat approximately 2039 UTC August 16,
1985, as it moves m)rlhward acro.,,s Redstone Arsenal, near Huntsville, Alabama. Note that tornado is pendent from a
well-defined wall cloud t•)rmation. !'he lornado was one in a series of tornadoes produced by this supercell-like storm,
whmh was •)nc ol scvcr,d spawned by i)anny. For more dr:fails, see McCaul [1987]. Photo is courtesy of A. Junkins,
! !t•ntsville, Alabama.

for a w.tricty o1'stratifications of the data. A major goal of observational work. Section 3 contains results of previous
that work was to find out how well various environmental [McCaul, 1990] and more recent numerical simulation stud-
parameters such as convective available potential energy ies of hurricane-spawnedconvection. Section 4 consists of
(CAPE [Moncrieff and Miller, 1976; Weisman and Klemp, discussionand interpretation of the results of the previous
19821), vertical shear and Bulk Richardson Number (BRN two sections. Section 5 gives a summary and outlook for
[Weisman and Klemp, 1982, 1984]), and helicity [Lilly, 1986; future work.
Davies-Jones et al., 1990] and streamwise vorticity [Davies-
Jones, 1984] describe the tornadic potential of the landfalling 2. OBSERVATIONS
hurricane environment. Detailed comparisons were also
made with the Great Plains tornado proximity environment.
2.1. Background
McCaul [1990] also conducted numerical simulation.,, of
hurricane-spawned severe storms in order to study their McCau!'s [1991] recent study of the characteristics of
structure and dynamics. The simulations were initiated with hurricane-spawned tornado environments employed all
a selection of soundings from his 39-year data sample, along available United States rawinsounding data taken near those
with some of his composite soundings. Comparison of the tornadoesthat were reportedduringthe 39-year period from
results with simulations of Great Plains severe storms 1948through1986. Even thoughonly tornadoesassociated
showed important similarities but also many differences. with Atlantic and Gulf Coasttropical cycloneswere included
These similarities and differences may prove useful in iden- in that study,the data set is still the largestever assembled
tifying the essential kinematic and thermodynamic proper- for the study of hurricane-spawnedtornadoes. Only the
ties of pretornadic environments in general. highlights
of thisstudywill be describedhere;for additional
The remainder of this paper is organized into four major details the reader is referred to the original work. The set of
parts. Section 2 summarizesthe major findingsof the M91 tornado-producing
tropicalcyclonesincludednot only
 McCAUL 121

Fig. 2. Radar echohistoryof four distinctive-echostormsspawnedby HurricaneDanny on August16, 1985,as seen

by NationalWeatherServiceradarat Huntsville,Alabama.Stormlabeled"22" producedtornadoshownin Figure1.
See McCaul [1987, Figure 3] for furtherdiscussionof this event.

hoursof reported hurricane-spawnedtornadoes.These cri-

cyclones of full-fledged hurricane intensity but also some
teria match those establishedby Novlan and Gray [ 1974]in
tropical storms and subtropicalcyclones. Nevertheless,
their earlier compositestudy. An additional 1097raobstaken
throughoutthe remainderof thispaperall thesecycloneswill
within one-day buffer periods before and after those of the
be referred to using the generic term "hurricanes" for
convenience. actualproximity soundingswere includedin the data baseto
The 39-year data sampleyieldedsome 199 rawinsonde help delineate the larger-scale structure of the tornadic
observations("raobs") that weremadewithin 185km and3 hurricanes. All raobs were inspected and erroneous
122 HURRICANE-SPAWNEI) TORNADIC STORMS

HURRIœRNE T•RNRDOE5
intensity associated with each raob. In M91 the term "tor-
nado outbreak" refers simply to the collection of all torna-
does reported with a landfalling hurricane and does not
imply the occurrence of more than some minimum number
of tornadoes. The subjectively assigned F scale values
undoubtedly containnumerouserrors[DoswellandBurgess,
1988], but they are found to correlate significantlywith
various meteorologicalparametersand thus may be useful
indicatorsof the intensity of tornadic activity.
The three measuresof ambient helicity examinedwere the
convective cell-relative total and relative helicity and
streamwisevorticity. Mathematically, the cell-relative total
helicity was defined[Lilly, 1986]as H t = (¾ - ¾c)' (k x
cgV/0z),the relativehe!icityas H r = Ht/[I V - Vcll0W0zl],
andstreamwise
vorticityas •os = Ht/IV - Vol, whereV is
the ground-relativeflow, Vc is cell motion, and k is the unit
vector along the vertical (z) coordinate. Because of the
similaritiesbetween theseparametersthey will be referred to
in the remainder of this paper collectively as "helicity
parameters." M91 presentedestimates of H t, H r, and (os
that were computed by averaging the profiles of point
ALL DAYS
estimatesof those quantities over the lowest 3, 6, or 12 km.
In computations of quantities requiring knowledge of con-
Fig. 3. Locations of all reported hurricane tornadoesin the United vective cell motionsVc, the 0- to 6-km mass-weightedmean
States, 1948-1986. Tornadoes are plotted at azimuthstaken relative winds were used as estimates of the cell motions. Because
to headingsof hurricane centers at tornado times. These headings cell motions were believed to be slower than those mean
have been rotated to coincide with the 360 ø azimuth. Areas of circles
are proportionalto unity plus the F scale rating of the tornado. Note winds (and consequently to be represented by vectors that
clusteringof data points in right-f¾ontquadrant. Range rings, labeled lay farther off the hodographthan the mean wind vector [see
along the 18if' azimuth, are 200 km apart.

F-SCALE SUM PER RROB

deleted. The 199 tornado proximity raobs were close to some 360

366 reported tornadoes. This is more than half of the total of 3 3O •o

626 tornadoes reported in all the Atlantic and Gulf hurri-

canes afl•2ctingthe United States during the 39 years of data
coverage.The hurricane-relativespatial distributionof those
tornadoes,given in Figure 3, confirmsthe findingsof Smith
[1965] and others that the right-front quadrant is the pre-
t•rred sector of a hurricane for tornado formation.
During the analysis of the data all raob winds were
converted from zonal and meridional components to radial
and tangential componentsrelative to the hurricane. Then,
tbr each sounding, bulk parameters such as CAPE, surface
to midtroposphericvertical shear, various measuresof am-
bient helicity, and BRN were computed. The raob location
relative to the hurricane's position and heading were also
recorded. The hurricane-relative spatial distributions of
these quantities were then mapped and cross correlations
computedagainstthe intensityof tornadicactivity reported
in proximity to each raob. Although the per-raobfrequency
distribution of tornadic activity was skewed, M91 found that
the logarithmof the sum of the tornadoes'Fujita scale(F ALL DAYS

scale) numbers [Fujita, 1973, 1981], each augmented by

unityto avoidnullifyingthe F0 events,wassufficientlyclose
Fig. 4. Spatialdistribution,
relativeto hurricanemotion,of objec-
to Gaussianto permit applicationof parametricstatistical tively analyzedvaluesof F scalesumper raob for 1296hurricane
analysistechniques.Thus, in McCaul's study, the log (F- raobs.Objectiveanalysisgrid meshhad spacingof 100 km. As in
sum)per raobis taken as the measureof tornadooutbreak Figure3, hurricanes arecenteredat ori,,•nandaremovingtoward360
 McCAUL 123

McCaul, 1987]),the computed

helicityvaluestendedto 0-6 KH SHERR
360
HRGN.

underestimate
thetruevalues.
M91argued
thatforstrongly
curved
hodographs
theHr estimates
should
beincreased
by
20-50%to obtainvaluescorresponding
to realisticstorm
motions.
Furthermore,
theestimates
of0- 3-kmtotalhelicity ,/
couldbe converted
to theunitsusedby Davies-Jones
et al.
[1990]by multiplying
by thedepthof theaveraging
layer,
3000 m.

Oncethe raobswere analyzed,theywerestratified in a

varietyof ways,thencomposited.
In constructing
thecom-
positeprofilesthe raw temperature,
dewpoint,andwind
component datawere transformedinto a 101-levelsigma
coordinatesystem,then averaged,then transformedback
into pressurecoordinatesusingthe mean of the surface
pressuresasa basevalue.Theroot-mean-square variability
of eachquantitywasalsocomputed duringthecompositing
processto facilitateassessment
of statistical
reliability.

2.2. Hurricane-Relative Spatial Distributions

of Raob Parameters
All 1296raobsin theM91 databasewereusedin preparing
210 150
RLL DRTS

objectivelyanalyzedmaps of the spatialdistributionof the

various raob parameters. The individual raob bulk values
Fig. 6. Spatial distribution of 0-6-km "BRN shear" magnitude
were interpolated to a 100-km Cartesian mesh using a (see text) for the 1296 hurricane raobs. Format is as in Figure 3.
distance weighting function with cutoff at 200 km. The Units of contoursare meters per second.
resultswere then contouredthroughoutthoseregionswhere
data density was sufficientto insurerepresentativeness.
Results of the objective analysesof F-sum, CAPE, BRN
shear is defined as the magnitude of the vector difference of
shear, BRN, 0- to 3-km H t are shownin Figures4-8. BRN
the density-weightedmean winds between 0 and 6 km and
between0 and 500 m (M91). Plots of H r and ms(see M91)
REGULRR CRPE generallyshow patterns resemblingthat of H t and are not
•60
shown here. The pattern of contours seen in Figure 4,
representing the distribution of F-sum, was taken as a
measure of the spatial distribution of tornado outbreak
intensity. Parametersshowing maxima and minima arranged
similarly to those found in Figure 4 may be considered
possiblyuseful as predictors of hurricane tornadoes.
In general, the patterns of BRN shear and the helicity
parameters more closely resemble that of F-sum than do
those of CAPE or BRN. Analyzed CAPEs reach a minimum
of lessthan400 J kg-• near andjust to the left of the
hurricane centers, relative to their motion. CAPE reaches a
maximum well to the right of the hurricane track and is
generally larger at the rear of the hurricane than on the
forwardside.The BRN shearand helicity parameters,on the
other hand, reach distinct maxima in the fight front or fight
quadrants. The occurrence of maximum shear in these areas
2qo
is roughly consistentwith the findingsof Novlan and Gray
[1974, Figures 10 and 12]. Because of the behavior of BRN
shearand CAPE the BRN contour pattern exhibitsits maxi-
mum in the distal portionsof the fight and rear quadrants,with
210 50
RLL DRT$ its minimum in the right-rear (RR) quadrant of the hurricane,
near the center. The BRN minimum is located close to that of
the maximum in F-sum and has a value well within the
Fig. 5. Spatialdistribution
of convective
available
potential
en-
ergyforthe1296hurricane 3. Unitsof supercellrange [Weisman and Klemp, 1982], suggestingthat
raobs.Formatisasin Figure
contours are joules per kilogram. supercellstormscould indeed be responsiblefor some
124 HURRICANE-SPAWNED TORNADIC STORMS

BULK RICHFtRDSON NO.

360 parameter,700-hPawind speed,alsodisplayeda high corre-
lationwithlog(F-sum),0.31.Interestingly,
thehelicityparam-
eters averagedthrough !2 km show the highestcorrelations,
rangingfrom 0.31 to 0.34, even thoughthe mean valuesof the
helicityparametersattain their highestvaluesin the 0- 3-km
layer. This suggeststhat favorable shearsthrough deeper
layersmakefor an environmentespeciallyconduciveto devel-
opment of more numerous and more intense hurricane-
spawnedtornadoes.It shouldbe noted,however,that only for
H r are the 0- to 12-km layer mean values larger than their
0--3-kmcounterpartsin a statisticallysignificantsense.
270 90
Consistentwith Figure 5, CAPE showsa weak negative
correlation of-0.07 with log (F-sum). Such a correlation is
4 0
significantat approximately the 23% level. Because of this
negative correlation between CAPE and log (F-sum), M91
did not automatically exclude from his data base soundings
having zero CAPE. However, exclusion of the zero-CAPE
soundingsfrom the correlation computations was found to
have no major impact on the results.
Although all the highest ranking correlations are statisti-
cally significant at better than the 1% level, their absolute
RLL
magnitudesare not very large. M91 suggestedfour possible
reasons for the smallness or the correlations: (1) mesoscale
Fi•. ?. Spatial distribution o• Bulk Richardson Number variability within the hurricanes, with potentially important
1296 hurricane raobs. Format is as in Figure 3. variations having scales not resolved by the raob network or
amplitudes not much larger than the accuracy of the raob
cane tornadoes. However, the location of the BRN minimum
measurements,(2) errors in F scale assignmentsand proba-
ble underreporting of the tornadoes, (3) for the helicity
and its overallpattern do not correspondas well with F-sum as
parameters, errors in the estimates of storm cell motions,
do the BRN shearand helicity parameter patterns.

2.3. Correlations of Raob Parameters 0-3 KH TOT. HELICIT¾

With Tornado Outbreak Intensity
While the resultsof Figures4-8 depict the spatialrelation-
ships between F-sum and the other raob parameters, linear
cross-correlation coefficients give a concise quantitative
measure of how parameter variations intercompare individ-
ually. In order to examine how the various mob parameters
changed with the reported intensity of hurricane tornado
outbreaks, M91 computed the cross correlationsbetween his
selection of parameters from the 199 proximity raobs and
their respective log (F-sum) data. Data from the other
nonproximity raobs were excluded in order to avoid biasing
the results with an excessivenumber of nonproximity raobs.
270 •'• xO
04 90
Table I contains a listing of the computed correlations,
along with the t values from a Student t test of the signifi-
cance of the correlation's deviation from an assumed zero
hypothetical value. As the table shows, many of the listed
variables exhibit correlations with log (F-sum) which are 0

significantlydifferent from zero. The resultsof the t testsand

some additional correlation simulations show that correla-
tions of approximately 0.14 magnitudeare significantat the
5% level (i.e., have only a 5% chance of occurringas a result ' t•tsø RLL
DRYS
of random effects). 180

As expected from examination of Figures 4-8, the raob Fig. 8. Spatialdistributionof 0-3-km total helicity H t for the 1296
parametersthat correlate best with tornado outbreak sever- hurricane raobs. Format is as in Figure 3. Units of contours are
ity are the helicity parameters and BRN shear. The simple meters per square
 McCAUL 125

TABLE 1. Cross-Correlations
Between Log(F-Sum)andRaob soundings.Theselatter soundingswere also screenedto make
Parameters
UsingOnlyRaobs in General
Proximity surethey were made at greater distancesfrom the hurricane
to Tornadoes
centers than the tornado occurrences. This restriction was
Correlation
designedto maximizethe probabilitythat the soundingswere
With Log
Variable (F-Sum) t Value representative of "inflow" conditionsfeedinginto the tornado-
beatingconvectivestorms.A listingof the specificsoundings in
CAPE(Jkg-1) -0.07 -0.91 the "close proximity" group is given in Table 2.
BRNshear(m S-1) 0.32 4.47
BRN -0.18 -2.50 The general and close proximity compositesare shown in
0- 3-km Hr 0.16 2.18 Figure 9 in skew-T 1og-pand polar hodographformat. Both
0- 6-km Hr 0.22 3.04 profiles feature a conditionally unstable layer below about
0- 12-km H r 0.31 4.34 650 hPa, surmountedby an absolutely stable layer. Above
0- 3-kmHt (ms -2) 0.26 3.58
0- 6-kmH t (m s-2) 0.28 3.99 500 hPa the lapserates are close to moist adiabatic. Relative
0- 12-kmH t (m s-2) 0.32 4.51 humidities are high throughout much of the troposphere,
O-3-km*Os(s- l) 0.26 3.54 especiallyin the close proximity profile, where dew point
O-6-km*Os(s-l) 0.30 4.18 depressionsare often less than 6øC. Maximum buoyanciesof
O-12-kmco s (s-1) 0.34 4.77 boundary layer parcels, approximately 3øC, are achieved
850-hPawindspeed (m s-1) 0.22 3.08
700-hPawindspeed (m s- l) 0.31 4.40 around 650 hPa. The tropopause is found on average near
500-hPawindspeed (m s- l) 0.25 3.46 150 hPa. All these patterns resemblethose found by Sheets
Sfc-850-hPa
shear(m s- 1) O.19 2.54
[1969] in his study of hurricane soundings.
Sfc-700-hPa
shear(m s- I) 0.28 3.92
Sfc-500-hPa
shear(m s- l) 0.20 In both proximity compositesthe winds are characterized
2.69
by inflow toward the hurricane center below approximately
CAPE is convective available potential energy; BRN is Bulk 2-kin altitude, with outflow above. The hodograph has a
Richardson Number; Sfc is surface. distinct "loop" or "horseshoe" shape, with veering winds
and shearsfrom the surface through at least 10-km altitude.
The strongestwinds occur near 2- to 3-km altitude, and very
and (4) inherently limited predictability of detailsof tornado
strongvertical shearsexist in the lowest 1 km. These results
intensity using larger-scale ambient single-parameterdata
are consistent with those of Novlan and Gray [1974] but
derived only from raob profiles. Many of the parameters
have the advantageof showing explicitly how the winds are
examined are designedto assessonly probability of meso-
organized relative to the hurricane center.
cyclone formation, rather than actual tornado formation, and
In both compositesthe altitude of maximum tangential
thus should not be expected to correlate straightforwardly
winds, near 2-3 km, is seen to be almost twice that observed
with tornado activity.
in mature maritime hurricanes [Frank, 1977, Figure 9]. This
suggestsan increase in the depth of the layer containing
2.4. Proximity Composite Soundings positiveshearof the tangential wind during and after land-
fall. Such a redistribution of tangential momentum is consis-
The 199proximity raobswere usedto constructa "general
proximity" hurricanetornado compositesounding.These tent with either the postlandfalladiabaticcore coolingof the
soundingsincludethosemadeundera varietyof conditions, hurricane described by Novlan and Gray [1974] or the
increasinginfluenceof surface friction [Gentry, 1983].
with no stratificationsof the data appliedexceptfor proxim-
ity to tornadoes.A secondaryset of criteria,40 km and 2 Comparisonof the two compositeprofilesin Figure 9 also
shows many important differences. Of special note is the
hours,appliedto the generalproximitysetyieldeda smaller
set of 10 "closeproximity" significantlylarger wind speed and shear in the close prox-
but apparentlyrepresentative

TABLE 2. Soundings
Takenin CloseProximityto HurricaneTornadoes
Raob Site Hurricane Time/Date Azran Maxwnd

Jacksonville,
Florida unnamed 0000UTC June9, 1957 33 at 256 18.0
Charleston,
SouthCarolina Cleo 0600UTC Aug29, 1964 40 at 205 43.8
CapeCanaveral, Florida Isbell 0200UTCOct.15,1964 301at 186 56.6
Miami,Florida Alma 1200UTCJune8, 1966 23 at 412 41.2
Miami,Florida Alma 1800UTCJune8, 1966 52 at 279 41.2
Victoria,Texas Beulah 0000UTCSept.21, 1967 59 at 207 72.1
Chantilly,Virginia David 0000UTCSept.6, 1979 26at 234 43.8
Victoria,Texas Allen 1200UTCAug.10,1980 79 at 265 56.6
Nashville,Tennessee Danny 0000UTCAug.17,1985 344at 305 41.2
Athens,Georgia Danny 1200UTCAug.17,1985 37at298 41.2
Azranismobazimuth,
range(inkilometers)
relative
tohurricane
heading.
Maxwndispeaksurface
windsat landfall (in metersper
126 HURRICANE-SPAWNED TORNADIC STORMS

HURRICANE TORNADO CD#PO$ I TE$

(ALL OURDSo ALL DRYS)
BOLD-CLOSE PillIX. , L IGHT-GENERRL PPIOX.

100

-9O

-8O 2OO

-7O
300

-60
uoo

5oo

600

700

800

900

1 ooo
-20 -lO o lO 20 3O
¾

-2O -15 -10 -5 H/5 5 10 15 20

Fig. 9. Skew-T 1og-pand hodographdiagramsfor the compositesoundingstaken in closeproximity (heavier lines)

and general proximity (lighter lines) to hurricane tornado events. U and V componentsrepresent the radial and
tangentialwinds relative to hurricanecentersat raob time. Small squareson hodographdiagramsmark 0-6-km mean
windsused as estimatesof convectivecell motion. Solid(open)squareappliesto thick- (light-) line
 McCAUL 127

TABLE3. MeanParameters
ofTornado
Proximity
Soundings The mean hurricane tornado proximity sounding and
DataStratifiedby Proximity hodograph were found to differ in many importantrespects
General Close fromtheir tornadoproximitycounterparts on the GreatPlains.
Variable Proximity ProximityThesedifferencesare highlightedin Figure 10,whichshowsthe
Number of temperatureprofiles 180 10 close proximity hurricane tornado profiles along with the
Number of wind profiles 169 10 Oklahomasupercellmean profilesof BluesteinandJain [ 1985].
Rangefrom center 329.3 264.8 The latter profilesclosely approximatethe findingsof other
CAPE(Jkg-l) 760.4 253.4 compositetornado proximity studies[Maddox, 1976;Darkow
BRNshear(ms-•) 9.8 14.0
BRN 66.3 3.3 and McCann, 1979;Schaefer and Livingston, 1988]and were
0- 3-km H r 0.28 0.52 taken to be representativeof tomadic environmentson the
0- 6-km H r 0.22 0.42 Great Plains. Hodographcomparisonsrequired somecaution
0- 12-km H r 0.15 0.39 because,unlike the hurricane composites,the winds in the
0- 3-kmH t (m S-2) 0.035 0.078 Oklahomaprofile were compositedusing true zonal and me-
0- 6-kmH t (m s-2) 0.023 0.055
0- 12-kmH t (m s-2) 0.016 0.046 ridionalcomponents.
0- 3-kmo•s (s- •) 0.0037 0.0068 With regard to the thermodynamic data, M91 found that
0- 6-kmo•s (s-•) 0.0025- 0.0051 the hurricane tornado events displayed much less CAPE
0- 12-km•os (s-1) 0.0017 0.0040 than the Oklahoma events. The mean of the close proximity
CAPE is convectiveavailablepotentialenergy.BRN is Bulk hurricane
tornadoCAPEs(Table3) was253J kg-•, or only
Richardson Number. about10%of the2542J kg-• valuefoundin theOklahoma
supercellcompositeof Bluestein and Jain [1985] (their Table
3, whichgives2490J kg-• , usingothersoftwarethanthat
imity case.In addition,temperatures
in the closeproximity used in M91's calculations). The hurricane CAPEs were,
compositeare slightly warmer in the middle and upper however, comparablein magnitudeto thosefound by Barnes
troposphere and slightly cooler below 700 hPa. The rela- and Stossmeister [1986] near the rainbands of a maritime
tively small difference in the mean radial distancesof the two hurricane.Surfacewet-bulb potentialtemperaturesOwwere
Figure 9 composites from the hurricane centers (264.8 km 22.7øCin the Oklahoma composite and 22.3øC in the hurri-
versus 329.3 km) cannot explain all the differences. The cane composite,but the temperature and dew point profiles
close proximity winds are evidently strongerbecausethey showed significantdifferences in many other respects. Most
are derived from data taken mostly within the RF quadrants prominent among the differences were the extensive dry
of relatively intense hurricanes (mean landfall sustained layer aloft, layer of enhanced static stability below 700 hPa,
windsof 45.6 m s-•), whilethe generalproximity
winds and the steep lapse rates in the Oklahoma composite.
include data collected in a variety of quadrantsin both weak Much of the CAPE in the hurricane tornado cases is
and intensehurricanes(meanwindsof 35.4 m s-I). The realized in the conditionally unstable layer that exists be-
differencesin the temperatureprofilescan alsobe attributed tween the surfaceand about 650 hPa. Boundary layer parcels
at least in part to differencesin hurricaneintensity. ascending through this layer can achieve buoyancies of
Table 3 gives listings of the values of mean CAPE, BRN 2ø-3øC, thanks to a minimum in ambient saturation wet-bulb
shear, BRN, and layer-averagedH r, Ht, and % for 0- to potentialtemperature0* that occursnear 650 hPa. This
w

relatively "cool" layer is apparently created by the hurri-

3-km, 0- to 6-km, and 0- to 12-km layers of the soundings
used in generatingeach composite.CAPE and BRN were cane's own circulation, which draws in unperturbedlower
tropospherictropical air while warming the mid and upper
found to be smaller and the shear and helicity parameters
muchlargerfor the closeproximitydata.Mean totalhelicity troposphere with the anvils from intense hurricane-core
convertedto the form usedby Davies-Joneset al. [1990]wasconvection.Thus it is apparently not the lower troposphere
approximately 300J kg-• forthecloseproximitycomposite.that is being cooled, but rather the mid and upper tropo-
This value is near the Davies-Jonesproposedthresholdfor spherethat are being warmed. This warmingand attendant
"strong" (at least F2 intensity)tornadoes. reduction of upper tropospheric contribution to CAPE is
M91 pointed out that in spite of the simplicityof the mostpronouncedin intensehurricanes,which are just the
compositeprofilesthe largevariabilityof the datamadeit onesmost likely to produce large tornado outbreaks(with
inevitable that some individual hurricane tornado soundings large F sums)during and after landfall.
deviateconsiderablyfrom the compositeprofiles.CAPEs Comparisonof the hurricaneprofile with that of Jordan's
werefoundto range
fromzeroto morethan3000J kg-•, [1958]mean Atlantic hurricane seasonsounding(Figure 11)
while maximumwindswereunder10 m s-• in strengthin reveals how hurricanes modify the mid and upper level
However,it was notedthat the temperatureand moisture profiles. The hurricanetempera-
someof the soundings.
soundings withmore ture profileis slightlycooler than the Jordanprofilebelow
withweakwindswereneverassociated
thanoneor two weak(F0-F1)tornadoes;ontheotherhand, 600 hPa but becomes 2ø-3øC warmer aloft. The dew point
(morethaneighttornadoes profilesare virtually identicalbelow about850 hPa. Citing
all the largetornadooutbreaks
apiece)werefoundto occurin environments wind the absenceof any significantevidenceof midlevel drying,
having
maximain excessof 15 m s-• . both in the compositesand in most individual
H URRICAN E-SPAWN ED TORNADIC STORMS

HURRICRNETORNRDO.õKLR SUPERCELLCOHPO$ITE$
(RLL DR'rS)
BOLD-HURR[ CRNE, L I GHT-OKLR
loo

,-90

8O 2OO

l0
3OO

SO:
•oo

500

800

'7OO

L•0
800

900

1ooo

30 -20 • ]0 0 10 20 30

78

- 20 - 15 - 10 -5 M/S 5 10 15 20

Fig. 10. Skew-T 1og-pand hodographdiagramsfor the hurricanetornadocloseproximitycomposite(heavyline) and

Oklahomasupercellcompositeof Bluesteinand Jain(light line). U and V componentsof the Oklahomacompositeare
relative to true zonal and meridional directions. Squaresmark 0-6-km mean winds as in Figure
 McCAUL 129

HURRZCRNE
TORNADO,
MEANTROPICIlLSUMHER
CõHPõSITES
(ALL DAYS)
BOLD-HURR
! CLINE,L ! GHT-IIEANTROP[ CRL $UHHER
loo

i I !I 1

/, - • • • • 300

-50 / /,•,•
500 600

"7Eld

/ / I I"
looo

-30 -20 -lO 0 10 20 30

Fig. 11. Skew-T log-p diagramshowingthe hurricanetornadocloseproximity compositesounding(heavy line) and

Jordan mean hurricane seasonsounding(light line).

M91 suggestedthat subsidencedoesnot play a majorrole in the small CAPE and large shearsfound in the closeproximity
producingthis warming aloft. Despitethe increasedoverall hurricane tornado cases, the BRN of the hurricane cases(3.3)
moisture in the hurricane sounding, the CAPE decreases is much smaller than that of the Great Plains tornado cases
from1345J kg-• in theJordanprofileto thehurricanevalue (32.8). For both data sets, however, the BRN falls in the
of 253J kg-• . The residuallowertropospheric
minimum in "possiblesupercell"range [Weismanand Klemp, 1982,1984].
0*wandwarmlayeralofthavealsobeenobserved in recon-
naissanceflight soundingstaken in maritime hurricanes 2.5. Variations With Hurricane Intensity and Size
[e.g., Hawkins and Imbembo, 1976,Figures5, 6, and 10], Novlan and Gray [1974], Gentry [ 1983], and Weiss[1985]
other mean hurricanesoundings[Sheets,1969],and in the all pointedout that the likelihoodof hurricanetornadoactivity
hurricanetornadostudyof Novlan and Gray [1974]. increaseswith the "intensity," that is, peak sustainedsurface
With regardto windsthe hurricane
tornadocloseproxim- winds, of the landfalling hurricane. M91's data support this
ity compositedisplaysgreatershearmagnitude, at leastat notion but also suggestthat hurricane "size" [Merrill, 1983],
low levels,than the Oklahomasupercellcomposite. Wind measured,for example, in termsof the radiusof outer closed
vectormagnitude differences
betweenthesurface
and1kmare isobar(ROCI), is important.Hurricanesnotablefor both their
14.8m s-• for thehurricane
profileand9.0m s-] forthe size and intensityhave been amongthe most prolific tornado
Oklahoma
profile.
TheOklahoma
profile
has12.5m s-• of producersin the database:Audrey(1957),Carla(1961),Beulah
BRNshear
andthehurricane
profile14.0m s-•. Because
of (1967), and Allen
130 HURRICANE-SPAWNED TORNADIC STORMS

2.6. Azimuthal Variations Around the Hurricanes

It has already been shown that most hurricane-spawned

tornadoes occur in the RF quadrants of the hurricanes in
Figure 3. The spatial distribution of raobs used in construct-
ing Figure 4 shows a similar pattern (not shown [see M91,
Figure 11]). Insofar as the mean landfall heading of tornadic
hurricanes in this study is 357.5ø, this pattern also indicates
that the northeast quadrant of hurricanes is the favored
geographicquadrant for tornadoes. The preference for tor-
nado activity in the northeastquadrantsof landfalling hurri-
canes has been documented by Hill et al. [1966], Orton
[1970], Novlan and Gray [1974], and Gentry [1983].
To investigatethe strongpreferencefor hurricane tornado
fk>rmation in the RF quadrants of hurricanes, M91 con-
structedcompositeprofilesin each of the quadrantsrelative
to calculated hurricane motion. The composites in the four
quadrantswere alsocombined,usingunit weightingfor each
quadrant'scomposite,to producean azimuthallyaveraged
overall profile for landfalling tornadic hurricanes. In the
latter case the effects of hurricane translation were effec-
tively removedby the averagingprocess,so that the result-
t,'ig. 12. (,)tladranlcompositehodographs(heavy lines)arrangedin ing profile was felt to approximate one that might be ob-
aroundthe compositehurricane,whichis tained around a stationary, symmetric hurricane. The
their respectiveqtllt½.h'anls
asst•medIobc m{)vingt{• top ot'diagramas shown. Mean hodograph averageof the quadrantsformed a standardagainstwhich
{)1'the t'½mrquadrants!light lines) is shown in each quadrant for
rci'crcnce. Bt>xes mark winds at each l-kin increment of altitude each of the quadrantprofilescould be compared.
t'r{.•t the surt'accthrot•gh 1{}kin. 'l'hc labels V r and V o indicate the The temperatureprofiles(not shown)differedin the lower
r;tdi•tlarm tangentialcomponentsof wind, elsewherereferredto as troposphere,where the rear quadrantswere significantly
(I and V. The !•.tbel"SFC" is plotted .just below the surface wind warmer than the front quadrants. In addition, the right
h{•x l'{)r the individual quadrant hodographs.Radii of hodograph
circles
correspond
t•)a windspeed
of 25m s •. quadrantswereslightlywarmerthanthe left quadrantsin the
mid and uppertroposphere,with the mostsignificantdiffer-
encesat upperlevels. Thesetemperaturedifferences,along
M91 surveyed all hurricanes that made landfall in the with enhanced moisture in the front quadrants, suggest
United S!ttles between 1964 and 1986 and that produced considerablequadrant-to-quadrantvariation in thermody-
zcrt), one, or morethan eighttornadoes,and foundthe major namic instability.These variationsin low level temperature
tornado outbreak producers(those with more than eight and moisture are consistent with the patterns of CAPE
tornadoes) were consistently larger and more intense at variation seen earlier in Figure 5.
landt'allthan those producingone or no tornadoes.In fact, The differences between the hodographs were even more
McCaul's data indicate that the hurricanes that produced strikingthan thosebetweenthe thermalpropertiesof the
majortornadooutbreakshadaveragepeaksustained winds quadrants.To highlightthesedifferences,Figure 12 shows
of47.1ms- • atlandfall,compared to only21.9ms-• forthe the four hodographs drawnat the centersof their respective
nontornadic"hurricanes." The major tornadooutbreak pro- quadrants relativeto the meanhurricanecenter,with each
ducersalso had averageROCIs of 2.93ø latitude,more than hodograph rotatedappropriately to reflectthe systematic
twice that for nontornadic hurricanes (1.40ø). These ROCIs changes
in orientationof the local(U, V ) coordinates.
The
are smallerthan those describedby Merrill [1983], probably meanof the four quadrants,which servesas a reference,is
because of differences in the cyclones sampled and, espe- also shown.
cially,in theresolutionof pressurecontouranalyses usedin It is evident from Figure 12 that the hodographin the
measuringROCI. The size and intensityof thosecyclones left-rear(LR) quadranttendsto fold backon itself as winds
that producedonly one tornadoresembledthoseof the firstveer,thenback,with height.Thisis quitedifferentfrom
nontornadic cyclonesmorethantheydidthemajoroutbreak what occursin the RF quadrant,wherewindsveer continu-
producers. ouslywith height.In general,the RF hodographshows
While some of the hurricanes responsiblefor major tor- particularly largeamplitudes of windspeedandshearanda
nado outbreakswere not particularlyintense,they all had more"open" shapethanthe otherquadrants.The compar-
ROCIs largerthan the meanof the nontornadic classof isonbetweenquadrantssuggests that the four hodographs
hurricanes.M91 thus concludedthat in addition to intensity would become nearly i:qual
to the mean if (1) a shearedmean
theareaexposedto favorablewindsandverticalshearswas flow from rear to front of the hurricane and varying in
animportant factorinfluencing
overalltornadoproductivity. magnitude from1.4m s-• at thesurface to 7.6m s-•
 McCAUL 131

TABLE 4. MeanParameters
of TornadoProximitySoundings
Data Stratifiedby Quadrant
Variable LF RF RR LR

Numberof temperature
profiles 36 86 44 21
Numberof windprofiles 32 81 42 20
Range
fromcenter 255.9 330.6 421.1 314.6
CAPE(Jkg-1) 299.8 684.2 1031.4 1212.3
BRNshear(ms-•) 8.4 11.3 8.9 6.5
BRN 32.6 33.5 119.0 190.7
0-3-kmHr 0.30 0.34 0.23 0.07
0-6-kmHr 0.19 0.29 0.18 0.06
0- 12-kin
Hr 0.10 0.21 0.14 0.02
0-3-kmH t (ms-2) 0.031 0.049 0.022 0.012
0-6-kmHt (ms-2) 0.019 0.033 0.013 0.009
0- 12-kmHt (ms-2) 0.011 0.023 0.010 0.002
0-3-kmcos (s-1) 0.0036 0.0048 0.0027 0.0008
0-6-km co
s(s-1) 0.0023 0.0034 0.0018 0.0006
0-12-kincos(s-1) 0.0014 0.0023 0.0013 0.0004
Quadrants
areLF,leftfront'
RF,right
front;
RR,right
rear;
andLR,leftrear.CAPE
isconvective
availablepotentialenergy.BRN is BulkRichardson
Number.

10-kmaltitudewere removedfrom all quadrants;the shearin themostuncertain,

owingto therelativelysmallsample
size
thiscomponent of flow is not uniformwithheight,being (21raobs) inthatquadrant andthepresence ofseveralvery
largerat lower altitudes;and (2) an additional across- unstable
profiles
there.
Examination ofmeanvaluesof 0,,,
in
hurricanemeanflow from right to left, rangingfrom 1.4 m thelowest500m (notlistedin Table4) showsthattheRR
s-1 at thesurface
to roughly 2 ms- • at 1-kmaltitude,were quadrant hasthehighest value,23.2øC,whiletheLF quad-
removedfrom the low-levelwinds;this component of flow ranthasthe lowest,21.5øC.Surfaceandlow levelrelative
becomesnegligibleabove4 km. humidities
aredistinctly
higherin theRF quadrant,
andthe
Thus M91 found that mostof the systematicazimuthal liftingcondensation
levelis at its lowestaltitude(highest
variationin thehodographs
couldapparently
beexplainedin pressure)there.
termsof the superimposition
of an almostunidirectionallyThe reductionof CAPE in the LF quadrantis consistent
sheared steeringcurrent
onthecircularly
symmetricinternal withthelowersurface
temperaturesanalyzedbyNovlanand
circulation of the hurricane itself. Gray[1974] in tornado-producinghurricanes. Mechanisms
To compare thewindprofiles shown in Figure12,a t test thatmight be responsible
for the cooler
boundary layerinthe
wasapplied ateachsigma level,undertheassumption that LF quadrant include
theadiabatic corecooling experienced
the variancesfrom eachprofilewere not necessarily the byinflowing airparcels
thathavebeenseparated fromthe
same [e.g.,Dixon andMassey, 1969,
p. 119].Application of warmseasurfacefor relativelylongtimes[NovlanandGray,
thistechnique to all possible
combinations of hodographs 1974],
raincooling of relativelydry air entrainedintothe
fromthefourquadrants confirmsthatthedifferences be- hurricane circulation
from the fringes
of the storm [McCaul,
tweenthewindprofiles arehighlysignificant statistically,1987],
andtheweakbackground baroclinicity
thatwould be
oftenat betterthanthe 1%level,especially in themidand expectedif shearedwesterlies
were influencingthehurricane.
As Table4 shows,the meanBRN shearsvariedfrom
upper troposphere.
A summary of key sounding for thefour quadrant
parameters seen
toquadrant
withCAPE.
in a pattern
Themean shear
muchdifferent
isgreatest,
thanthat
11.3ms-1, in
quadrantsisgiven
inTable4.Thetableshows
thatthemean theRF quadrant,
andleast,6.5m s-1, in theLR. Mean
distance
fromraobsiteto hurricane
center
rangesfrom256
to 421kmin thevarious
quadrants. mostof the values
However, of BRNrangefrom32.6and33.5in theLF andRF
raobsweremadesufficiently centers quadrants
farfromthehurricane to 119.0and190.7intheRRandLRquadrants.
TheBRNsin thefrontquadrants
arewithinthe"possible
thatthepatterns
inthecompositesandtheircharacteristic
parameters
would notchange greatly
if thedatausedinthe supercell"
1984].
They
range,
are
according
also
to Weisman
somewhat smaller
andKIernp
than
the
[1982,
correspond-
compositing
were more carefully
stratified
toinsure
thatthe ingvaluesinferred
for theRF quadrant fromFigure7,
meandistancesto hurricane
centers
became equal.
AsTable4 indicates,
themeanCAPEs primarily
varyconsiderably becausethe dataused in constructing
that
figure
inthefourquadrants,
with
aminimum
of300 included
Jkg-• inthe for many nontornadic
raobs andBRNtends tobelarger
suchraobs.The meansof eachof thehelicityparameters
leftfront
(LF)quadrant,
and amaximum
of1212
Jkg-• in exhibit largequadrant-to-quadrant
variations,
withvaluesin
theLR.MeanCAPEintherightrear(RR)quadrant
is1031
Jkg-1, while
that
intheRFquadrantJkg-•ßAllthesethe
is684 RFquadrant
being
typically
intheLR quadrant.
about4times
larger
thanthose
Mean0- to 3-kmHr andHt in theRF
values
areconsiderably
smaller
thanthemean
CAPE
ofthe
Great
Plains
tornado
environment.
ThemeanLRCAPEis quadrant
are0.34and0.049
ms-2
132 HURRICANE-SPAWNED TORNADIC STORMS

?..7. Composites Based on Other Data Stratifications movingfasterthanapproximately

15m s-• . The reasons
for
M91 also generated composite profiles for various other this complex relationship require additional research.
Gentry [1983] and Weiss [1985] noted differences in tor-
stratificationsof the data. Because raobs were plentiful in
nado frequency for hurricanes making landfall in different
the RF quadrants of the hurricanes but spotty elsewhere,
areas, with more tornadoes associated with southern land-
these other compositeswere constrained to use RF quadrant
falls. M91 concluded that the most significant differences
data only. Only a general review of the results is possible
were the result of whether the landfall occurred along the
here; for further details the reader is referred to M91.
Gulf of Mexico or Atlantic coasts, with Gulf of Mexico
In order to see how the hurricane thermodynamic and
landfalls showing a pronounced tendency to generate more
kinematic profiles evolved following landfall, M91 stratified
tornadoes. RF quadrant compositesfor both regions showed
and compositedthe RF quadrant raob data with respect to no major differences, prompting M91 to ascribe the differ-
time after hmdfall. The results showed little difference in
ences in tornado frequency to other factors. Among the
temperature and moisture but significant changes in leading candidates were systematic differences in hurricane
hodograph structure. In particular, there was a distinct intensity, intensity tendency, and ability to expose the
tendency for mid and upper level winds to veer with time, favorable RF quadrant winds to !and. In each of these areas,
with only a slight loss of strength relative to the low level Gulf of Mexico hurricanes were found to be more favorable
winds, which weakened substantially. These changes have for tornado formation than their Atlantic counterparts.
the effect of' maintaining, sometimesfor three days or more,
the large valuesof hellcity which prevail in thoseportionsof 3. NUMERICAL SIMULATIONS
the circulationsof landfallinghurricanes which lie downshear
of the centers relative to the steering flow. This pattern of 3.1. Simulation Specifications
evolvinghodographstructurewasevidentto somedegreein all McCaul I'1990] also performed numerical simulationsof
thespecificcasesof individualtornadichurricanes whichcould hurricane-spawnedsevere convection. He used the three-
be traced throughseveral days t•)11owinglandfall. dimensionalcloud model developedby Klemp and Wilhelm-
Becausepreviousinvestigators[Fujita et at., 1972; Weiss, son [1978], with modifications reported by Wilhelmson and
19871noted diurnal variations in hurricane tornado occur- Chen [1982].The model solvesnine prognosticequationsfor
rences,M91 alsocompositedthe RF quadrantraobdata with the basic variables, potential temperature 0, dimensionless
respectto time of day. Only data from the standardtimes pressurerr, the Cartesianvelocity componentsu, v, and w,
0000 and 1200 UTC were available, so that only two com- water substancemixing ratios for vapor q v, cloud water q,.,
positescouldbe constructed.The temperaturesin the 0000 and rainwater q•, and subgridturbulent mixing coefficient
UTC compositewere foundto be slightlywarmerat virtually Km. Ice phasesprocesses are not included.Open lateral
all levels. Thus there was no major difference in CAPE boundaryconditionsare usedat the sidesof the integration
between the two composites.Winds likewise did not show domain,while rigidlid conditionsare specifiedat the top and
majordifferences.Theseresultssuggested a possiblediurnal bottom boundaries.
observationalbias with respect to reporting of hurricane To give satisfactory resolution, the model runs were
tornadoes.However, M91 also pointed out that the observed conductedon a mesh having 500-m horizontal spacingand
frequenciesof occurrenceof the tornadoesare nearly iden- smoothlyvaryingverticalspacingrangingfrom 250 m at the
tical at 0000 and 1200 UTC, so that the results required bottomto 750 m at the top. To capturethe high tropopauses
cautiousinterpretation.Raobswould be neededat approxi- often observed in hurricane environments, a vertical domain
mately0800and2000 UTC to be ableto draw moredefinitive depthof 20 km was specified.Horizontally,the domainwas
conclusionsabout the effectsof diurnal temperature changes a 50-km square.These were approximatelythe largestdo-
on occurrences of tornadic convection. main dimensionsfeasible with the available computing and
Weiss [1987] noted a positive correlation between hurri- dataanalysisresources.Largedomainsizewasfoundto be
cane forward translational speed and tornado occurrences. desirableas a means of insuringthat the developingsimu-
To study how the environmentsof fast- and slow-moving lated storms remained within the simulation area while
hurricanesdiffered,M91 compositedraob data taken during growingthroughatmospheric layerswith considerably dif-
the periodfromthe day of landfallthroughthefollowingtwo feringwinds.For all simulationsof hurricaneconvection the
days, within the RF quadrants of hurricanesthat were y axes of the model grids were alignedparallel to the
observed tobemoving at eitherlessthan4 m s-i orgreater tangentialcomponentof wind flow relative to the parent
than8 m s-• at raobtime. The mid andupperlevel windsin hurricane. For the control Great Plains simulation the y axis
the fast hurricanecompositewere distinctlystrongerthan was oriented toward true north.
those of the slow group, consistent with the notion of In eachsimulation,convectionwastriggeredby releaseof
interactions with sheared steering currents of greater ellipsoidal
thermalbubbles of horizontalradiusof either7.07
strength.The middleandupperportionsof the troposphere or 10.0km, vertical semiaxislengthof 1.4 km, andmaximum
werealsosignificantlydrierin the fast composite.M91 found thermal excess of 2.0øC. A few additional test runs were
a slight tendencyfor faster hurricanesto producemore madeusingeither 1.0ø or 4.0øCbubbles.Unlessotherwise
tornadoes,but the relationshipbroke down for hurricanes noted, resultsdescribedhere representruns using
 McCAUL 133

TABLE5. Characteristics
ofEnvironmental
Soundings
Usedin 3.2. Simulationof Hurricane Danny's Storms
InitializingSimulations
The environmentwhich supportedthe Hurricane Danny
Variable FSI CKL VCT AVG
tornado outbreak of August 16, 1985, in northern Alabama
CAPE(Jkg-1) 2472.8 1833.9 522.0 610.8 apparentlywas well representedby the CKL raob from 0000
BRNshear(ms-1) 12.3 12.9 17.1 12.3 UTC August 17 (see Figure 13). This soundingfeatured
BRN 32.6 22.1 3.6 8.0
0- 3-kmHr 0.48 0.50 0.53 CAPEof 1834J kg-1 anda veeringhodograph
0.86 containing
-2
0- 3-kmH t (mS-2) 0.015 0.077 0.063 0.071 0--3-kintotal helicity of 0.076 m s . The CAPE readingis
0- 3-kmtOs (s-1) 0.0024 0.0073 0.0070 0.0076 large comparedto most other hurricane soundingsbut still
smallerthan the meanof Oklahomasupercellor Great Plains
FSI is FortSill,Oklahoma, composite, 2100UTCMay20,1977.
CKL is Centreville,Alabama, 0000UTC August17,1985.VCT is tornadoenvironments, whichare about2500J kg-1. It is
Victoria,Texas,partiallyinterpolated(above500hPa),1200UTC also larger than the value reported by McCaul [1987], who
September 20,1967.AVGiscloseproximity composite,
modified
to useda preliminary soundingthat underestimatedthe depth
reduce low level dew point depressions.CAPE is convective
availablepotentialenergy. BRN is Bulk RichardsonNumber.
of the very moistboundarylayer. The H t readingis large
enoughto suggestthe possibility of tornadoes,accordingto
the criteria devised by Davies-Jones et al. [1990] for Great
Plains storms.
bubbles.A Coriolisparameterof 0.0001$-1 wasusedin all In the simulationa weak supercelldevelopsduringthe first
simulationsexcept for two testswhere zero and 0.0002 s-1 hour, then strengthens in a pulsating manner during the
were used. All simulations were allowed to run for 7200 s second. By 7200 s the storm still shows a trend of slow
(2 hours) of simulated time, with fields stored each 600 s intensification.
Peakupdraftstrength
of 27m s- • is notedin
(10 rain) for postanalysis. a burst of growth just prior to the end of the simulation
The model storms were initialized in environments se-
period.Classicsupercellcharacteristicsare evidentin the qr
lected from the observationalstudy of M91. Results re- field at 7200 s, with a "weak echo" area embraced by a
viewed here will consistof simulationsinitializedwith (1) a "hook" at z = 0.5 km (Figure 14). A mesocyclonicregion of
tornado proximity soundingat Centreville,Alabama(CKL) enhancedvertical vorticity •r is also apparent, with peak •'
from 1985's Hurricane Danny, where supercelIswere ob- exceeding
0.016s-i . Comparison
of theCKL stormwitha
served, (2) the closeproximity composite(denoted"AVG")
control simulation of the FSI storm [Klemp et al., 1981]
describedin section 2, and (3) a proximity soundingat reveals similarities in structure but differences in horizontal
Victoria, Texas (VCT) from Hurricane Beulah in 1967. These
scale(Figure 15). The FSI storm is noticeably larger in most
three soundingsmay be saidto representmuchof the rangeof
respectsthan the CKL storm.
parcelbuoyancyconditionsseenin the spectrumof hurricane-
Differences in the vertical extent and intensity of the
spawnedtornadoenvironments.Even the high-buoyancy hur-
updrafts of the CKL and FSI storms are also apparent. The
ricane soundings,however, generallycontainlessCAPE than
vertical profilesof updraft intensity along columnsrepresen-
typical Great Plains tornado proximity soundings.
tative of flow in the main updraft cores of the two stormsare
For purposes of comparison, a control simulation of a
shown in Figure 16. The CKL storm achieves peak updraft
Great Plains severe storm (the "Del City Storm [see Klemp
et al., 1981]) was made also, using a compositesounding
amplitude
ofapproximately
22m s- 1nearz = 3.0 km,while
the FSI storm reaches 44 m s-• near 9.0 kin. Because the
from Fort Sill, Oklahoma (FSI) and the same model mesh
used for the hurricane convection simulations. Results were FSI storm has little buoyancy at low levels, the CKL storm
updraft is actually more intense everywhere below 4-kin
essentially similar to those obtainedby Klemp et al. [1981]
altitude. The levels of condensation and free convection are
and are discussedhere only in terms of how they contrast
near the lowest model level in the CKL environment, while
with the resultsof the hurricane-spawnedstormsimulations.
the level of free convection is above 2 km in the FSI storm.
A summaryof the characteristicsof the environmentsused
The CKL storm updraft does not extend above an altitude of
to initialize all the simulationsis given in Table 5. Note that
H and •os are computedusingthe 0- to 6-kinmeanwind(as 10 km, whereas the FSI storm reaches above 14 km.
in section 2) and not the actual storm motion. The perturbation 0 field at z = 0.5 km in the CKL storm
In all the hurricaneconvectionsimulationsmadeby McCaul (Figure 17) reveals only a rather weak pool of rain-cooled
[1990],persistent convective activityoccurred.Theabundance outflow. Minimum O' values are only -2.4øC, only a third of
of low-levelmoisturein the initiatingenvironmentsencouraged the magnitudesseen in the correspondingfield (not shown)
thedevelopment of numerous convective cells,mostof which of the FSI storm. The high relative humidities and relative
wererelativelytransient butsomeofwhichdisplayed supercell lack of dry, potentially cold air at midlevels in the hurricane
characteristics. McCaul'sresultsemphasized onlythe charac- environmentsmakes only weak rain-cooled outflow a com-
teristicsof themostintenseandpersistent
convection
notedin mon feature of the hurricane convection simulations.
eachmodelrun. In this review, only the mostgeneralaspects The CKL mesocyclone has an even stronger, more clearly
of the simulated storms are discussed;further details are defined, closed circulation at 2.0 km (Figure 18) than at 0.5
forthcomingin papersto be published elsewhere. kin. Also seen in Figure 18 is the crescent-shaped
134 HURRICANE*SPAWNED TORNADIC STORMS

ioo

-9O

-80 2OO

-70
300

-60 •00

5OO
-50
600

•00
-•0
800

900

1000
-30 -20 - 10 0 i0 20 9-0

u
-20 -15 -tO -5 H/S 5 10 15 20

Fig. 13. Skew-Tlog-pandhodograph diagramsfor 0000UTC August17, 1985,at Centreville,Alabama(heavyline),

duringpassageof HurricaneDanny'sremnants,andfor 1200UTC September 20, 1967at Victoria,Texas(lightline),
duringpassage of HurricaneBeulah.Dataabove500hPain Beulahareinterpolatedfromadjacent6-hourlyraobs,which
showedlittle changein the upper troposphere.U and V componentsof the winds are radial and tangentialto the
hurricane centers at raob time. Squaresmark 0- to 6-kin mean winds as in Figure 9.

core characteristic of mesocyclonic storms. Peak updraft located on opposite sides of the main updraft core, but
intensityat this altitudeis a respectable18.0m s-•. because the vertical shear in the environment reverses
A vertical cross section of the flow vectors and perturba- direction with height, both minima are located on the down-
tion pressurefield of the CKL storm, taken looking toward shear side of the updraft at their respective levels. The low
negative x (Figure 19), reveals clearly defined S-shaped level downshear mesolow, centered near z = 2.0 km,
streamlines wrapping around and through low level and features a pressure deficit of 2.8 hPa. The flow around these
upper level pressure minima. The pressure minima are mesolows, especially the low level one, actually
 McCAUL 135

manystreamlines
whichappear
toformwell-defined
closed
loops.
Thecentersofthese
loopsarenotstagnant
butrather
contain
strong
flownormaltotheplaneofthecross
section. 25
Thethree-dimensional
character
of thestormflowisthus •'/ //
highly helical.
Theprincipal
features
of theupdraft
dynamicsarewell
described
bytheinviscid
momentum
dV
equation'
20.•
""'"
fl'
!'"/'
•/"/"'•""
•t/,//•.
"*'
/'d ////77'7•
t//////-'-'"-'
.•,/ //,J//. •./•fi t/7/// ) ??: / / / /;: / / / ".' ' ......
"
25•Z•z/',;'_•,.
/,"'•'//•'//•/i,'
•///:// //: / / ,',,,','.......
.• ,/5./2' '/_,r/7 . '.f •/I/I/l//l /• • 2 2 ;, / / / / //////,'.'

dt- -CpOøVrr
+Bk, (1)•
whereV is the velocityvector,•r is the dimensionless
LILI'J• ?'7•.d./l P' I I J I I i l l / l / / /..' .'

perturbation
pressure,
cptheisobaric
heat
capacity
ofair,00 :5 ":: ""0 .'<.•' "NN.x-.L
.....................
theambient potential
temperature,B thebuoyancy,
andk a
unitvectoralong
thez axis.Followingthegeneral
approach
usedbyRotunno andKlemp [1982]andWeisman
andKlemp
[19841,
a diagnostic
expressionfortheperturbation
pressure
'i / l •' / l l / t w ." w t ! i I I I , , .......... %'

may be obtained
by takingthe divergence
of (1). The
diagnosticpressureequationthusobtainedis

0 OB
cpV.
(00V•r)
= -v. (v.vv)- • (V.v)+•,
Oz (2) 5 10 X,KM 15 20
whichisa Poisson
equation
forced
bybuoyancy
divergenceFig. 15. Rainwater
mixingratio(contour
interval0.7g/kg)at z =
(lasttermontheright)anddynamic
effects(othertermson 0.5 kmandt = ?200s in controlsimulation
of Oklahoma
supercell
theright).Thetotalperturbation
pressure
rrmaybeparti- storm.Windvectorsanalogous
to thosein Figure14arealsoshown.
Simulation
wasinitialized
withdatafromFortSill,Oklahoma
(FSI)
raob,asdescribed
by Klempet al. [1981].Note increased
horizontal
sizeof hookechoconfiguration
anditsgreaterdisplacement
from
apparentcenterof mesocyclone,
as comparedto Figure14.

tionedintobuoyancy
anddynamic
contributions
,rb and•rd
by solvingthe Poissonequation(2) individuallyfor the
respective
forcingtermson itsright-hand side.Thepressure
contributions
,r/, and rrd will be calledbuoyancy pressure
and dynamicpressure[Schlesinger,1984].This dynamic
pressureis not to be confusedwith the similarlynamed
pressureassociatedwith parcel energychangesfound in
flowssatisfyinga Bernoulli equation.
Most of the vertical accelerationsresponsible for the
strongupdraftin the CKL stormare causedby gradients of
thedynamic pressure.Thisisstronglysuggestedby compar-
isonof Figures20 and21, whichdisplaycontoured analyses
of the verticalaccelerations
due to dynamicpressureand
totalbuoyancyeffectsrespectively.The dataare displayed
on the samecrosssectionas in Figure19.The totalbuoy-
ancyeffectsincludeverticalpressuregradients
of •rt•aswell
asbuoyancy itself[seeWeismanandKlemp,1984].Detailed
analysesof updraftforcingalongparceltrajectoriesconfirm
15 20 X.KM 25 30 that dynamicpressureforcing in the mature CKL storm is
Fig. 14. Rainwatermixingratio (contourinterval0.8 g/kg)at z -- approximately 3 timesas stronga contributor to netupdraft
0.5 km and t = 7200 s in simulation of supercell-likestorms intensity at the levelof maximumupdraftas is totalbuoy-
spawned by Hurricane Danny on August 16, 1985. Wind vectors ancy forcing.
relative to simulation domain translation velocity are also shown.
In Figure21, strongwavelikepatternsare evidentin the
Simulationwas initialized with soundingfrom 00 UTC August17,
1985,at Centreville, Alabama (CKL). Note warpingof contoursinto buoyancy forcingfield. In particular,a lobeof strongnega-
a distinctive "hook" configuration at lower end of rain area, tive forcingextendsdownwardthroughthe low level down-
adjacentto regionof high cyclonicvorticity. shearmesolow gyre,whereverticalvelocities are
 25

12

FS!

lilttit

I ll'lll
trill
Ilttl II
Illll
If/It

•llll

tlllltt
Illill!Ill i !
illt!Illll ! !
IIIIIIIIif I !
Illlltllll litIll
-20 • 10 0 10 20 30 qO 50 /lit/l/Ill! ! litlit

t;ig. 16. Comparative vertical profiles of maximum tipdraft veloc- 15 20 X,KM 25 30

ity (in meter,, per second) from CKL and FSI simulated storms.
N•te tha! ('KI, st•>rmis actually more intense below z = 4 km than Fig. 18. Verticalvelocity
of CKL stormcontoured
every2 m s-1
is the !;SI storm. I.)epth oi:'('KI .....
storm updraft is much smaller than (negative values dashed) at 7200 s and an altitude z = 2.0 km.
Storm-relative wind vectors are also shown for reference. Note
that o!' I:SI storm (see also Figures 22 and 23).
existence of a strong mesocyclone with closed circulation on con-
cave side of crescent-shapedupdraft. Peak updraft exceeds 18 m
s- • at this altitude.

2O
o .

_ -

2O

tS

l0

15 20 X,KM 25 30
Fig. 19. Verticalcrosssectionof wind flow and perturbation
pressurecontours(contourintervalis 0.3 hPa;negativevaluesare
Fig. 17. Perturbation potentialtemperature fieldfor CKL stormat dashed)of CKL stormat 7200s. Sectionviewedtowardnegativex.
7200 s and an altitude of 0.5 km (contour interval is 0.3øC; negative Note S-shaped streamlinepatternin updraft,with flow wrapping
valuesare dashed).Wind field of Figure 14 is also shownfor aroundtwo closedgyre circulations,one at low levels,the other
reference.Note cold pool at this level has minimumtemperature aloft. Both gyresoccur on downshear sidesof updraftat their
deficitof only2.4øC,whichis onlyaboutonethirdthatin theFSI respective
altitudes,
consistent
withreversalof shearwithheight.
storm simulation. Pressureminimum at z = 2.0 km has amplitude -2.8
 McCAUL 137

negative.
Fartherto theright,buoyant forcingagainbe-
comespositive,
andstreamlinesbegintorecoveranupward
tendency.
Thereason forthispattern
isrevealedin Figure
22, which displaysthe qr fieldfor the CKL stormon the
samecross-section plane.The negatively buoyantlobeis
associated
with a concentrated
falloutof heavyrainin a
narrowsheathjust downshear of the mainupdraft.The
downdraft
accompanyingthisrainhelpsclosetheloopofthe
gyre circulation,then,afterbeingpurgedof rain,becomes
positive
buoyancyasit continues
to subside.
At thetopof
theupdraftthenegative
buoyancy
lobeextends
throughthe
regionwhererainis accumulating
dueto matchingof termi-
nalvelocitywithupdraftspeed.
Updraftairthatmanages to
risethroughthisregionaftersheddingits rainonceagain
becomespositivelybuoyantowingto latentheatingand
absenceof rain loading.
A smallvaultis alsoevidentin the qr fieldof Figure22.
Comparison
of thisfeaturewiththecorresponding onefrom
the FSI storm, Figure 23, againillustratesthe differencesin
scaleof the hurricane-spawned andGreat Plainsstorms.The
FSI stormalsogenerates a sheathof heavyrain,butbecause
the sheathis broader(stormis larger)andsmallerin ampli-
tude(lessambientmoisture),andbecausetheupdraftis less
tilted, it is less effective at generatingthe concentrated Fig. 21. Contours of vertical accelerationsdue to buoyancy ef-
downdraftsat the proper downstreampositionneededto fects(contour intervalis0.01m s-2; negative valuesaredashed),on
produce a closed flow gyre. same cross section as in Figure 18, for CKL storm. Buoyancy
effectsincludethoseof perturbationpressuregradientsproducedby
vertical gradient of actual buoyancy. Storm-relative wind vectors
are also shown for reference. Note smallnessof upward accelera-
tions near base of main updraft and wavelike pattern of accelera-
tions at midlevels of storm.
2o
L. Z

3.3. Simulation of "Close Proximity"

Hurricane-Spawned Storms
Convectionwas also simulatedusingthe "close proximi-
i5
ty" ("AVG") hurricane tornado sounding (see Figure 9),
moistenedat low levels to facilitate storm development.The
simulated storms were weaker than the CKL convection but
still exhibitedmany features suggestiveof incipientsupercell
structure.The q, field at z = 1.0 km at 7200s (Figure24) has
a distinctive curved shape, much like that seen in the CKL
simulation.Enhanced vorticity is also presentin the area of
the mesocyclone.The storm is againquite smallcomparedto
the FSI supercelland appearsnot to be fully mature. Like
the CKL storm, it was continuingto increasein intensity at
the end of the simulation.
The absence of low level rain-cooled outflow is even more
pronouncedin the AVG stormthanit wasin the CKL storm.
The 0.5 km O'field (Figure 25) showsno significantnegative
temperatureperturbationsat all. In fact, the rear flankof the
stormcontainsonly a pool of air warmedas muchas 2.0øC
10 15 7,KM 20 25 by subsidence.However, there is somelow level cool air
beneaththe storm, with a -1.8øC temperatureperturbation
Fi•. 20. Contoursof ve•ic• •cceler•fionsduem e•ects of dy- noted at the lowest model level, 0.125 km.
namically
forced
pressure
(contour
interv• The updraft dynamics of the AVG storm show many
values•e d•shed)on s•me crosssection•s in Figure18for CKL
sto•. Storm-rebfive •ind vectors are •]so sho•n for reference. similaritiesto those of the CKL storm. Dynamic pressure
Note]•Be up•d •ccelemfions
nearb•seof mainupdr•L forcingagainappearsto predominate
overbuoyancy
138 HURRICANE-SPAWNED TORNADIC STORMS

fdoubledto 0.0002s-• fromthevalueusedin theotherruns

began to display some supercell features after t = 6000 s.
The test was intended to serve as a simplified attempt to
mimic conditions associated with the strong background
rotation found in hurricane circulations.
15 '
The "high-f" VCT supercell has many features in com-
mon with the CKL and AVG storms described earlier. The
qr and wind flow fields at z = 0.25 km and t = 7200 s are
given in Figure 27. Noteworthy are the well-defined "hook"
in the rainwater and the highly symmetric mesocyclone flow
10 adjacent to it. Maximum vorticity in the mesocyclonewas
0.044s- i, somewhat
largerthanthatachievedin theCKL or
AVG storms. As before, there is little evidence of a gust
t'ront in association with the rear flank of this mesocyclone.
Many other stormscoexistwith this incipientsupercellin the
model domain, but none of' the others acquired as many
supercell traits.

4. DISCUSSION

The conceptsof he!icity and streamwisevorticity appear

10 15 I',KM 20 25 to offer helpful insights into mechanisms underlying the
genesisof hurricane-spawned tornadoesas well as other
Fig. 22. Vertical cross sectionof wind flow and rainwater con- types of tornadoes.Valuesof ambient
helicityare oftenquite
tours {conttnirinterval is 1.0 g/kg) of CKI• storm at 7200 s. Section high in landfallinghurricanes,suggestiveof considerable
is sameas in Figtires 18--20.Note shieldlikecanopyof heavy rain
formingin updraft,then fallingout on downshearside,producinga
shallowvault under the main updratt and downward motionsin low
levelgyreto its north(right). Fartherto the north,subsiding
air that 2o
has shedits rain load acquirespositivebuoyancy(see Figure 21).

in establishing
total updrattvelocity.A cn)sssectionthrough
the updraft,lookingtowardsnegativex as before, again
showsthe doublegyre flow structure(Figure 26) witnessed
before in the CKL results. This time, however, the storm is
even smaller than the CKL storm. Updrafts in the AVG
storm barely reach 7-km altitude and form a stark contrast
with the massiveupdraftof the FSI storm(Figure23). The q,.
field showsa small vaulted overhangin the lowest 2 km and
a sheathof rain falling into the downwardbranchof the low
level
downshear
gyre.
Theoverall
morphology
and
dynam-
ics of the AVG stormappearto be quite similarto thoseof
the CKL storm.

3.4. Simulation of Hurricane Beulah's Storms

Simulationswere also conductedusingsoundingdata from
1200 UTC, September20, 1967,at VCT duringHurricane
Beulah. The soundingcontained,even for hurricanetornado
10 15 T ,KM 20 25
environments,onlya "moderate"amountof CAPE (seeTable
5), whilethehodograph
againcontained
considerable
he!icity. Fig. 23. Verticalcrosssectionof windflowandrainwatercon-
The environmental conditions are shown in Figure 13. tours(contourintervalis 1.0g/kg)of FSI stormat 7200s. Compar-
Storms simulated in this environment (not shown) are isonwithFigure22showsdifferences
in verticalscalebetween
this
GreatPlainssupercell
andthe simulatedhurricane-spawnedCKL
numerous butgenerallymulticellularin character.Consider- storm. Note also the diffusecharacterof the precipitationshaftto
able vorticityis generatedalongthe elongatedflanks,but therightofthemainupdraft; thisshaftislesseffectiveat producing
clear-cutsupercellfeaturesare not apparent.However, a a concentrated downdraftthanthe onein the corresponding portion
sensitivitytestconductedusinga valueof Coriolisparameter of the CKL storm in Figure
 McCAUL 139

potential for updraft rotation. Correlationsbetween hurri-

canetornadooutbreakintensityand helicityparameters are
positive and rank amongthe leadingindicatorsof tornado
activity. Hurricanetornadoactivityis alsopositivelycorre-
latedwith boththe intensityandsizeof the parenthurricane
circulation,but these two factorsare not independentof
raob-derived wind and shearparameters.
The good spatialagreementbetweenthe patternsof hur-
ricane tornado occurrenceand enhancedambienthelicity
represents additional evidence in favor of the connection
between helicity and tornadoes. The fact that the maximum
ambient helicity occurs in the fight-front quadrant of the
landfalling
hurricane
offersa newexplanation
forwhythat
quadrant bears the brunt of most of the tornado activity.
Examination of the azimuthalvariationof hodographstruc-
ture around the hurricanecenterssuggeststhat it is the
superimpositionof a shearedlarge-scalesteeringflow on the
otherwise circularly symmetric hurricane circulation that
producesthe helicity maximumin the fight-frontquadrant.
This explanation is fundamentallydifferentfrom any pro-
posed
byprevious
investigators
ofhurricane
tornadogenesis.
However, the findingsof M91 do not disprovethe hypothe- 30 35 X,KH L!_0 L!5
sesof shear enhancementby core cooling [Novlan and Gray,
1974] or increased surface drag [Gentry, 1983]as the hurri- Fig. 25. Perturbation
potentialtemperaturefield under AVG
cane encounters land. storm at 7200 s and an altitude of 0.5 km. Wind field of Figure 23 is
also shownfor reference. Note virtual absenceof cold pool usually
The negativecorrelationbetweenhurricanetornadoactiv- seenat thislevelbeneathprecipitating
storms;onlya smallshallow
regionof coolair wasfoundat thelowestmodellevel(z = 0.125
km; not shown).Downdraftshave producedanomalously
warm
5o temperatures
in areato left of mesocyclone.

ity and CAPE is apparentlythe resultof the meagerupper

... • ........ ,,. ,,.. ,,....., ....... ..,•, ..... ,... tropospheric
contributions to CAPE in the large, warm
anvils of the more intense hurricanes. The smallnessof
•5
CAPE in the hurricane environments does not, however,
preventsevere
convection,
because
theleveloffreeconvec-
tion is low and abundantmesoscaleconvergencein the fully
.............. ;-
.... ;•; ........ , developed
hurricaneis alwaysavailableto triggerstorms.
Furthermore,the simulationsindicatethat convective-scale
•0
perturbation-pressure
forcingin the 'hurricane-spawned
stormsis capableof compensating
for the reducedbuoy-
., , . ancy.In fact,theperturbation-pressure
minimain thehur-
ricane-spawned stormsmayachieveamplitudes
comparable
to thosefoundin Great Plainsstorms,owingprimarilyto the
35 favorablecombinationof verticalprofilesof buoyancyand
shearwhichprevailin landfalling
hurricanes.
The factthat
buoyancy isnottheonlyfactorgoverrfing
severe
storminten-
sityisalsoconsistentwiththerecent
findings
ofotherinvesti-
gators[Johnsetat., 1990;Lazarus
andDroegemeier,1990].
Otherfactorsalsoinfluencehowbuoyancyoperatesin the
30 35 X,KH •0 hurricanecases.Becausebuoyancyis smalland environ-
mentalmoistureabundant,the effectsof rainwaterloading
•i8. 24. •i•watermixi• ratio(cornour
imerr•lis0.7•/k•)•t z =
1.0 Ema•d t = 7200sCotstormsimulated
i• "AVG" havea morepronounced
anddirectimpactonstormdynam-
similarto t•t of "o]oseproximity" bu•c•e tornado soundly8.ics than is the casewith Great Plainsstorms.There is less
Wi•d vectorsrelativeto simul•tio•domai•tm•slatio•speed production ofnegativebuoyancy through evaporative
cool-
•lsos•ow•.Asi• CKL simulation,t•r• isw•i• ofcornours imo ing of air by rain thanin the GreatPlainscases,with
• distinctive"•ook" co•fi•umtio••t lower•d o½r•i• •,
attendantdifferences in the spatialdistribution
of
•dj•cemtoreSio•o½e•ced cy01o•ic yogicity.
140 HURRICANE-SPAWNED TORNADIC STORMS

ß $'. J • J . j'. JJ. j. ,2' i i j ,! j f . j . j. j o j . j .

....................................... sphere,with updraftsreachingmaximumintensityin middle
and upperlayers.Thus, the morphologyof convectionin
.
both the Plainsand hurricane-type environments may be
seen to be strongly controlled by not only the bulk or
averagevaluesof characteristic
parametersbut alsoby their
15
actual distributions in the vertical. This conclusion is remi-
niscentof someof the findingsreportedby Simpsonet al.
[ 1985]regarding tropicalwaterspoutsandtheirparentcumu-
lus clouds.It is alsoreminiscentof the findingsof Szokeet
al. [1986]for generalmaritimetropicalconvection,although
10
here the key findingis that most of the buoyancyin low-
buoyancy tropical environments is concentratedin the lower
troposphere,where the vertical shear is strongest.The
presenceof small buoyancy in hurricaneenvironments,in
combinationwith relatively large perturbation-pressure
fields,alsoencouragesthe developmentof convectivestorm
circulationsthat resemblein some respectsthe idealized
Beltrami model of Davies-Jones [1985]. In the Beltrami
model, buoyancyforces are zero, yet a swirling updraft
supportedentirelyby pressureforcescan exist. A summary
listing of typical values of what appear to be the most
35 •r,atq •0 u,5 50 important
buoyancy
andshearparameters,
gradients,
and
scalelengthsgoverningconvective storm dynamicsin Great
Fig. 26. Vertical cross section of wind flow and rainwater con- Plains and hurricaneenvironmentsis given in Table 6.
tours (contour interval is 1.0 g/kg) of AVG storm at 7200 s. Section
is sa•neas in Figure 22. Note shieldlikecanopyof heavy rain
Ibrmingin updraft,then t•tllingout on downshearside,producinga
shallowoverhangunder the main updraft and downward motionsin
low level gym to its north.

buoyancy
regions
in thestorms.
Simulations
of thehurri- 25
cane-spawned storms commonly show the tilted helical
updrafts quickly shedding heavy rain into the downshear
environment. Negative buoyancy associatedwith this rain-
water loading produces a downdraft that creates a helical
gyre circulationin the low-levelstorminflow.Simulation 20
results indicate that this gyre contains at least as much
horizontalvorticityasthemesocyclone doesvertical vortic-•1
ity. The horizontal vorticity arises from baroclinic effects
associatednot only with latent heat release in the updraft but
also with the extra forcing producedby the concentrationof 15
rainwater loadingjust downshearof the updraft. These effects
are also important in Great Plains severe storms,althoughin
those storms the horizont• buoyancy gradientsare usually
associatedwith large temperaturegradientsas well.
An additional factor influencing hurricane-spawnedstorm
dynamics is the existence of a stable layer that lies typically
just above 650 hPa. This stable layer acts to confine the
intense convection in the hurricane environments to the 15 20 X,KH 25; 30 35
lower troposphere.It also hasa strongimpact on the vertical Fig. 27. Rainwater mixing ratio (contour interval is 0.4 g/kg) at z =
changesof buoyancy experienced by storm updraft parcels 0.25 km and t = 7200 s in simulationof tornadic storms spawnedby
and thereby restricts the vertical extent of the low level Hurricane Beulah on September 20, 1967. Wind vectors taken
downshear mesolow, whose shape, intensity, and position relative to simulation domain translation speed are also shown.
play such a pivotal role in the dynamics of the hurricane- Simulation was initialized with soundingfrom 1200 UTC September
20, 1967, at Victoria, Texas (VCT), using a Coriolis parameter
spawnedstorms.By way of contrast, for Great Plainscases, double that usedin the other simulations.Note warping of contours
convection is usually constrained only by the tropopause, into a distinctive "hook" configurationat lower end of rain area,
and the storms there occupy the entire depth of the tropo- adjacentto highly symmetricmesocyclone
 McCAUL 141

TABLE 6. Characteristic
Valuesof Atmospheric
Parameters
in the vertical distributions of ambient vertical shear, buoy-
Hurricaneand Great PlainsTornadoEnvironments
ancy, and moistureplay an important role in governingthe
Variable Hurricane Great Plains morphology of convective storms. Although there is less
buoyancy in the hurricane environments, its concentration
CAPE(J kg- l) 500 2500 in the lower troposphere, along with most of the vertical
BRN shear(m s-•) 12.0 12.0
BRN 7.0 30.0 shear, encouragedthe development of a form of miniature
0- 3-km H r 0.35 0.35 supercell storm not previously seen in numerical simula-
0- 3-kmH t (m S-2) 0.05 0.05 tions. Apparently, large buoyancyis not a prerequisitefor
0-3-kmms(s-1) 0.005 0.005 supercell storms, but what buoyancy there is must be
LFC (km) 0.5 2.0
Maximum O' 3.0 9.0 distributedin a propitious way in the vertical.
Maximumbuoyancy(m s-2) 0.10 0.30 Additional research into the details of the mesoscale
Height of maximumbuoyancy(kin) 3.0 9.0 environments containing hurricane-spawned severe storms
OB/Oz
(10-4 s-2) 0.33 0.33 is also desirable. Although the numerical simulationsindi-
Maximumwindspeed(m s-1) 25.0 35.0 cate that potentiallytornadic convectioncan developfrom
Height of maximum wind speed(kin) 3.0 I0.0
OV/Oz(s-•) 0.010 0.004 warm bubbles alone, it is possible that other dynamical
effectsnot representedin the initial conditionsare playing
CAPE is convectiveavailablepotentialenergy.BRN is Bulk significantroles. Furthermore, becauseof limitationsto the
Richardson Number. LFC is level of free convection. horizontal domain size used, the simulations may not cap-
ture realisticallythe mesoscaleorganizationof the groupsof
tornadic stormsoften observedsimultaneouslyin outbreaks
The simulated storms agree well with thosefew hurricane- of hurricane-spawnedtornadoes.
spawned tornadic storms that have been documentedin the
literature. In both the simulations and the observations it is
seen that their vertical and horizontal scale is reduced Acknowledgments.The observationalanalyseswere done in
compared with typical Great Plains supercells.The hurri- conjunctionwith the author's dissertationat the University of
Oklahoma,underthe auspicesof National ScienceFoundationgrant
cane-spawnedstorms may neverthelessacquire many of the
ATM-850130 and a teaching assistantship.The results were ex-
characteristicsof classic supercellstorms. The simulations tendedandrefinedduringthe author's tenureas anAdvancedStudy
indicate that it is in the lowest 2-3 km where the hurricane- Programpostdoctoral
fellow atthe NationalCenterfor Atmospheric
spawnedsupercelIsgive their strongestand clearestradar Research's Mesoscale and Microscale Meteorology Division. The
numerical simulationsshown here were conducted on Cray X-MP
signatures.Therefore detection of these stormsusing the
supercomputers
at both NCAR andthe NationalCenterfor Super-
existing network of weather radarswill probablybe more computing
Applicationsat the Universityof Illinois.Analysisof the
difficult than detection of Great Plains-type supercelIsbe- simulationresults was done using VAX 11/780, VAX 8550, and
causeof problemsof incompletesurveillanceat low levelsin VAX 6000-410computersat OU's GeosciencesComputingNet-
muchof the regionbetweenradars.Thefinerbeamwidthsof work, MMM, and the Engineeringand Analysis Data Systemat the
National Aeronauticsand Space Administration'sMarshall Space
the new National Weather ServiceDopplerradarswill make FlightCenter,respectively,thelatter undersupportfrom Universi-
it somewhat easier to resolve the circulations of these ties SpaceResearchAssociationthroughgrantNAS8-38769.The
storms,but only a densernetworkof radarscan closethe authorgratefullyacknowledges guidance,support,andconstructive
In the absenceof uniformsurveil- criticismfromDougLilly, HoMe Bluestein,ClaudeDuchon,Kelvin
gapsin the surveillance.
Droegemeier,JohnFiror, Joe Klemp, Rich Rotunno,Morris Weis-
lanceit may still be necessary
to attemptidentification
of man,Gary Barnes,Ed Zipser, Bob Wilhelmson,BobDavies-Jones,
potentially
tornadichurricane-spawned storms
through anal- Mike Kalb, and Jim Arnold. Many usefuldiscussions
were hadwith
ysisof grossfeaturessuchas echointensityandlongevity. otherssuchas Bill Gray, Joe Golden, ChuckDoswell, Don Burgess,
Steve Weiss, Bob Johns, Mark Powell, Frank Marks, and John
Flueck. Many, especially Wes Roberts, Dwight Moore, Pat
5. SUMMARY AND RECOMMENDATIONS Waukau, Ken Hansen, Bill Boyd, Dennis Joseph,Will Spangler,
RameshJayaraman,JayanthiSrikishen,Chris Pizzano,and Bob
McCaul's resultshave offereda new explanationfor why
Wilson,providedassistance
withcomputing
needs.Thanksalsoare
hurricanes
spawntornadoes
preferentially
intheirright-front due to anonymous reviewers,who madethoroughand insightful
quadrants:
helicity
enhancement
caused
bytheinteraction
of commentson earlier versions of this manuscript.
thehurricane'
s swirlingflowwitha steering
currentcontain-
ingshearroughly
parallelto thehurricane
heading.
Ques- REFERENCES
tionsremain,however,regardinghowandwhenrelativeto
!andfallthishelicityenhancement
occurs.Furtherobserva- Barnes,G. M., and G. J. Stossmeister,
The structureanddecayof
tional and simulation work is required before answersto a rainband in Hurricane Irene (1981), Mort. Weather Rev., 114,
2590-2601, 1986.
thesequestions
canbeobtained.Suchinvestigations
could Bluestein,H. B., and M. H. Jain, Formationof mesoscalelinesof
significantly
improveourability
topredict
whichlandfalling precipitation:
Severesqualllinesin Oklahomaduringthe spring,
hurricanes
willbecome
majortornado
producers
andwhich J. Atmos. Sci., 42, 1711-1732, 1985.
will not. Darkow,G. L., and D. W. McCann,Relativeenvironmental winds
The simulation
findings
suggest between for 121tornadobearingstorms,in Preprints, 1ith Conference
thatrelationships
142 HURRICANE-SPAWNED TORNADIC STORMS

Severe!,oc,! Storms,pp. 413-417,AmericanMeteorological environments,in Preprints, 16th Conferenceon Severe Local

Society, Boston, Mass., 1979. Storms,pp. 334---339,
AmericanMeteorological
Society,Boston,
I)avics~Jones,
R. P., Streamwise
vorticity:The originof updraft Mass., 1990.
re,ration
in supercell
storms,J. Atmos.Sci., 41, 2991-3006,1984. McCaul, E. W., Jr., Buoyancyand shear characteristicsof hurri-
l)avies-Jones,R. P., Dynamical interaction betweenan isolated cane-tornado
environments,Mon. WeatherRev., 119, 1954-1978,
convectivecellanda veeringenvironmental
wind,in Preprints, 1991.
!4th ConiC,
reneeonSevereLocalStorms,pp.216-219,American Merrill, R. T., A comparison
of largeandsmalltropicalcyclones,
MeteorologicalSociety, Boston,Mass., 1985. Mon. Weather Rev., 111, 1408-1418, 1983.
Davies-Jones,
R., D. Burgess,
andM. Foster,Testof helicityasa Moncrieff,M. W., andM. J. Miller, The dynamicsand simulationof
tornadoforecastparameter,in Preprints,16th Conference
on tropicalcumulonimbus and squalllines, Q. J. R. Meteorol. Soc.,
SevereLocal Storms,pp. 588-592,AmericanMeteorological 102,373-394, 1976.
Society, Boston, Mass., 1990. Novlan, D. J., and W. M. Gray, Hurricane-spawnedtornadoes,
!)ixon, W. J., and F. J. Massey,Jr., Introductionto Statistical Mon. Weather Rev., 102,476-488, 1974.
Antdy.vi,¾,
638 pp., McGraw-Hill, New York, 1969. Orton,R., Tornadoes associated
with HurricaneBeulahon Septem-
l)oswell, C. A., and D. W. Burgess,On someissuesof United ber 19-23, 1967, Mon. Weather Rev., 98, 541-547, 1970.
Stalestornadoclimatology,Mon. WeatherRev., 116, 495-501, Pearson, A.D., and A. F. Sadowski, Hurricane-induced tornadoes
1988.
and their distribution,Mon. Weather Rev., 93,461-464, 1965.
Frank,W. M., Thestructure
andenergetics
of thetropicalcyclone, Rotunno, R., and J. B. Klemp, The influence of the shear-induced
I, Stormstructure,Mon. WeatherRev., 105, I119-!135, 1977. pressuregradient on thunderstormmotion, Mon. Weather Rev.,
Ftkjita,'I'. 'I'., Tornad{ms
aroundtheworld,Weatherwise,
26, 56-62, I10, 136-15!, 1982.
78-83, 1973. Schaefer,J. T., and R. L. Livingston, The typical structureof
l;ujita,'1'.T., Tornadoes anddownbursts in thecontextof general- tornado proximity soundings,J. Geophys. Res., 93, 5351-5364,
izod planetaryscales,J. Aimos. S½'i.,38, 1511-1534,1981. 1988.
tq[jita,'1'.'l'., K. Watanabe,K. Tsuchiya,and M. Shimada,Ty- Schlesinger,R. E., Effects of the perturbationpressurefield in
phoon-associated tornadoesin Japanand new evidenceof suction numerical models of unidirectionally sheared thunderstorm con-
vt}rticcsin a tornadonearTokyo, J. Meteorol.Soc. Jpn., 50, vection: Two versusthree dimensions,j. Atmos. Sci., 41, 1571-
431-453, 1972. 1587, 1984.
(3entry, R. ('., (}enesis of t(•rnadoesassociatedwith hurricanes, Sheets,R. C., Somemeanhurricanesoundings,J. Appl. Meteorol.,
Mon. Weather Rev., 111, 1793-18(15, 1983. 8, 134-146, 1969.
1Iawkins, H. F., and S. M. Irabembo, The structureof a small, Simpson,J., M. C. McCumber,and J. H. Golden, Tropical water-
intensehurricane--Inez 1966,Mon. WeatherRev., 104, 418-442, spouts and tornadoes, in Preprints, 14th Confi,rence on Severe
1976.
Local Storms, pp. 284-288, American Meteorological Society,
1Ii!1, E. 1......W. Malkin, and W. A. Schulz, Jr., Tornadoesassociated Boston, Mass., 1985.
with cyclonesof tropical {•rigin•Practical features,J. Appl. Smith, J. S., The hurricane-tornado, Mon. Weather Rev., 93,
Meteorol., 5,745-763, !966. 453-459, 1965.
Johns, R. H., J. M. Davies, and P. W. i•eftwich, An examination of Stiegler, D. J., and T. T. Fujita, A detailed analysisof the San
the relationshipof 0-2 km A(•I• "positive" wind shearto poten- Marcos, Texas, tornado induced by Hurricane "Allen" on 10
tial buoyantenergy in strongand violent tornadosituations,in August 1980, in Preprints, I2th Confi,rence on Severe Local
Preprint.•',!6th ConJ?renceon Severe Local Storms, pp. 269-274, Storms, pp. 371-374, American Meteorological Society, Boston,
American Meteorological Society, Boston, Mass., 1990. Mass., 1982.
Jordan, C. 1,., Mean soundings for the West Indies area, d. Szoke, E. J., E. J. Zipser, and D. P. Jorgensen,A radar studyof
Meteorol., 15, 91-97, 1958. convectivecellsin mesoscalesystemsin GATE, I, Vertical profile
Klemp, J. B., and R. B. Wilhelmson, The simulation of three- statistics and comparison with hurricanes, J. Atmos. Sci., 43,
dimensional convective storm dynamics, J. Atmos. Sci., 35, 182-197, 1986.
1070-1(196, 1978. Tannehill, I. R., Hurricanes, 303 pp., Princeton University Press,
Klemp, J. B., R. B. Wilhelmson, and P.S. Ray, Observed and Princeton, N.J., 1950.
numerically simulated structure of a mature supercell tl•under- Weisman, M. L., and J. B. Klemp, The dependenceof numerically
storm, J. Atmos. Sci., 38, 1558-1580, 1981. simulated convective storms on vertical wind shear and buoy-
Lazarus, S. M., and K. K. Droegemeier,The influenceofhelicity on ancy, Mort. Weather Rev., 110, 504-520, 1982.
the stability and morphology of numerically simulated storms, in Weisman, M. L., and J. B. Klemp, The structure and classification
Preprints, I6th Con.l%renceon SevereLocal Storms,pp. 269-274, of numericallysimulatedconvectivestormsin directionally vary-
American Meteorological Society, Boston, Mass., 1990. ing shears, Mon. Weather Rev., 112, 2479-2498, 1984.
Lilly, D. K., The structure, energeticsand propagationof rotating Weiss, S. J., On the operationalforecastingof tornadoesassociated
convective storms, !l, Helicity and storm stabilization, J. Atmos. with tropical cyclones, in Preprints, 14th ConJkrenceon Severe
Sci., 43, 126-140, 1986. Local Storms, pp. 293-296, American Meteorological Society,
Maddox, R. A., An evaluation of tornado proximity wind and Boston, Mass., 1985.
stability data, Mort. Weather Rev., 104, 133-142, 1976. Weiss, S. J., Some climatologicalaspectsof forecastingtornadoes
Malkin, W., and J. G. Galway, Tornadoes associated with hurri- associatedwith tropical cyclones, in Preprints, 17th Conference
canes, Mon. Weather Rev., 81,299-303, 1953. on Hurricanes and Tropical Meteorology, pp. 160-163, American
McCaul, E. W., Jr., Observations of the Hurricane "Danny" Meteorological Society, Boston, Mass., 1987.
tornado outbreak of 16 August 1985, Mon. Weather Rev., 115, Wilhelmson, R. B., and C. S. Chen, A simulation of the develop-
1206-1223, 1987. ment of successivecells along a cold outflow boundary, J. Atmos.
McCau!, E. W., Jr., Simulations of convective storms in hurricane Sci., 39, 1466--1483,
 TornadicThunderstorm
Characteristics
Determined
WithDopplerRadar
EDWARD A. BRANDESl

National SevereStormsLaboratoo',Norman, Oklahoma73069

1. INTRODUCTION of May 20, 1977 [Johnson et al., 1987], and the Lahoma and
Orienta stormsof May 2, 1979 [Brandes et at., 1988].
The absenceof papersconcerned withmultiple-Doppler The evolutionof supercellthunderstormsmay not apply to
radarobservations at the 1976TornadoSymposium points all tornadic storms. Indeed, many storms, for example,
out that much of what is known about tornadic thunder-
those that produce Denver, Colorado, area tornadoes, form
stormsfrom this data sourceis fairly recent.A reviewis in
in environmentsthat are much differentthan thoseof super-
order to summarizethese observations, to interpretthe cells. Storm formation and evolution often are tied to oro-
observationsin view of recentdevelopments in numerically
simulated thunderstorms, and to determine which of the graphic features, and storm lifetimes are relatively short
current tornadogenesis theories are consistent with the ob- compared to supercelIs. Denver area storms offer research
servations.Here the contributionsof dual-Dopplerradar opportunities for studying tornadogenesisnot possible in
observationsare emphasized.A discussionof thunderstorm other regionsor with other storm types.
dynamicsis givenby Klemp [1987].Tornadogenesis theories It is important to note that tornado flow patterns are not
are discussedby Rotunno [1986] and Davies-Jones[1986]. well resolved in Doppler radar measurements.Hence kine-
Field observations of tornadic storms are describedby matic and thermodynamic properties fostering tornado gen-
Davies-Jones [ 1988]. esis and dissipationnecessarily are deduced from observa-
The entire archive of dual-Doppler radar observations tions of the larger-scale storm flow.
from well-sampled supercellstorms,the most violent of the
tornadic storms, probably consistsof less than 10 storms. 2. LIFE CYCLE OF SUPERCELL STORMS
There are a number of reasonsfor this, but basicallyit has
proven difficult to capture a significantportion of a long-lived For the sake of discussionwe define several nonoverlap-
tornadic storm within a relatively small, ground-based,dual- ping stagesof supercellevolution. Development will refer to
Doppler radar network. the entire period from storm initiation up to the first time that
Also, many of the observedsupercellstormsin the archive tornadoesbecome likely. Many of the characteristics asso-
occurred years ago when Doppler radar signalprocessing ciated with supercells (e.g., weak echo regions, hook ech-
was in its infancy. Yet, these stormsconstitutesomeof the oes, and overhangs)develop during this period. In general,
best available data sets. One of these storms, the Harrah early development has received little attention in either
storm of June 8, 1974, remains an outstandingexampleof a observationalor modeling studies. It is impossible when
supercellthunderstormand is usedextensivelyin this re- radar "first echoes" are observed to determine whether or
view. Other accountsof this stormare given by Ray [1976] not a storm will become tornadic. Also, numerically simu-
and Heym•field [1978].Other documented tornadicstorms lated thunderstormsare most influenced by the initialization
include the OklahomaCity storm of April 20, 1974 [Ray, processduring early development. The mature stage will
1975],the Spencer-Luther stormof June8, 1974[Brandes, refer to that critical period in storm evolution at which the
1978],the Del City stormof May 20, 1977[Brandes,1981; basicupdraft and vertical vorticity patterns associatedwith
Ray et al., 1981;Johnsonet al., 1987],theFort Cobbstorm supercelIshave evolved and the storm is primed for torna-
dogenesis.Elevated tornadolike vortices may be observedat
1Nowat theNationalCenterfor Atmospheric
Research,Boulder, this stage. The onset of damaging rotary surface winds
Colorado.
heralds the tornadic stage. We will then review storm
The Tornado:Its Structure,Dynamics,Prediction,andHazards. propertiesassociatedwith the posttornadicstage.This stage
GeophysicalMonograph79
Thispaperisnotsubject toU.S.copyright.
Published in1993bythe representsdissipationin many stormsbut merely a quiescent
American GeophysicalUnion. phasein stormswith cyclic tornado development.
144 TORNADIC THUNDERSTORM CHARACTERISTICS

2.1. Development MULTICELL WESTPLA INS SUPERCELL

Historically, there has been a tendency to categorize MODEL STORM MODEL
severe thunderstormsas either multicell or supercell. Mul- (StrongEvolution)(WeokEvolution)(Quosi-Steody)
ticell storms are composed of several convective cells (the
elemental convection unit [see Byers and Braham, 1949]) at
various stagesof development [e.g., Browning and Ludlam,
1962; Dennis et al., 1970; Chisholm and Renick, 1972]. New
cells fi0rm on a preferred flank, pass through the storm, and
eventually dissipate on the opposite storm flank as they
complete their life cycle. The addition of new cells gives a
propagative component to the storm's motion and causesa
motion that deviates from the mean wind to the side on
which the storm is growing.
Supercell storms are often conceptualizedas being com-
posedof a steady,single, supercel!with right-deviatemotion
[Browning, 1964;Marwitz, 19721.However, manysupercells
on close inspectiondisplay multicellular structure, particu-
larly in their'upperlevels [e.g., Lemon, 1976;Burgesset al.,
1977; Weaver and Nelson, 1982; l•5•ote and Wade, 1982].
Foote arid t'?ank [1983] presented Doppler radar observa-
tions ot'a multicellular Colorado hailstorm that had a number
ot'features normally ascribedto supercelIs.They proposeto
distinguishmulticelland supercellstormsby the ratio of the
distance between successiveupdraft perturbations(L) and
the diameter of the updraft perturbations(D). in multicell
storms, L/D > 1, while in supercelIs the spacingbetween
successiveupdrall perturbationsis much less than their
diameters, that is, L/D << 1. As the ratio L/D decreases, 0 5 i0 15 :•0 0 5 !0 15 :•0 0 5 I0 15 20

persistent
backgroundregionsof updraftandradarreflectiv- HORIZONTAL DISTANCE
ity ½henceR)rth
reflectivity)developand impart a certain
steadinessto the storm (Figure 1). In this interpretation, Fig. 1. Conceptualized updraftmodelsof multicell,hybrid(west
supercelIsformby the overlapof neighboring perturbations plains),andsupercell thunderstorms [FooteandFrank,!983].Areas
of parallellinesrepresent precipitation.
Notethatindividual
rather than the evolutionof a singlelarge, and enduringcell. cells lose their identify in the continuum updraft
from multicell storm to
Vasilq/J'etal. [1986]studiedan Oklahomasupercellstorm supercell.
thatbeganasa clusterof seemingly unrelatedordinarycells.
Althoughindividualcellsandupdraftcentersalwayscould
be identified(e.g., Figure 2a), the increasein size of these
f•aturesresultedin largepersistentregionsof updraftand stormof June8, !974. Figure 4a showsthe storm-relative
horizontalwind flow near cloud base (1.3 km) at 1530. (All
reflectivitywhosemotiondifferedfrom the individualcen-
ters by 65ø (Figure 2b). The inflow air to the persistent heights areabovegroundlevel,andunlessotherwise noted
all times are Central Standard.)The Harrah storm moved
updraft regionhad a significantlylarger storm-relative
from 230øat !8 m s-• . The 40-dBZ reflectivitycontourhas
streamwise component of vorticitythanthe inflowair in the
cell-relative reference frame. Strong vertical vorticity con- beenaccentuated to helporientthe readerwhenotherwind
centratedat the leading edge of the backgroundupdraft field parameters are discussed.
The heavydashedline(near
region(Figure3). A largesupercell withmiddlelevelupdraft x = 11, y = 22 km) marks the locationof the mesocyclone
within which the tornado formed at ---1546. For our pur-
rotation,a hookecho,andradarreflectivityin excessof 60
dBZ wasproduced;however,no tornadoes werereported. poses,a mesocyclone is definedas the regionwherethe
The role of strong vertical vorticity in determiningthe vertical
vorticity
is >-10-2 s-•. The storm's
1
firstechowas
characteristics
of radar-observedsupercelIswasalsostudied detectedat 1406;
thus thestorm was---1•hoursoldwhenthe
by Brandes et al. [1988].Individual andupdraft data in Figure 4 were obtained.
reflectivity
centers were ill-defined and did not propagateacrossthe At thispretornadic stage,thecoverage
of reflectivity>50
storm.Instead,they were sweptby the rotationalflow to dBZ is smallbut growing.An arc-shapedgustfront between
stagnation
pointsupwindof theupdraft. southerlyinflowandrain-cooled air with a westerlycompo-
nentbeginsto the west of the northernend of the mesocy-
2.2. Mature Stage clone,passes throughit, andthenextendsto the southwest.
The discussionof the mature evolutionarystageusesthe A rudimentary hookechoonthewestsideof the circulation
exceptional
radarobservations collectedfromthe Harrah (x = 10, y = 23 km) nodoubtresponds to the
 BRANDES 145

-IO

20••30•
01956
CST
7
CELL2 202:•
-2027
2(•03 ''
2016
•9'56
-20

2057

CELL
7 ß 2(•52
2047
2110
2057
-:30
CELL
6 2&7
2•52
CELL9 2•'16
.,. I

-50

2t53

-60 '
-50

X-DISTANCE FROM NORMAN (kin)

Fig. 2. Motion of (a) individualreflectivitycellsand(b) stormmotionfor the Lindsay,Oklahoma,stormof June19,

1980 [Vasiloff et al., 1986].

The distribution of vertical velocity at 1.3 km, with the propagation.Individual centers do not appear to move along
40-dBZ reflectivity contour and the mesocycloneposition the arc; rather, the characteristic pattern periodically rees-
superimposed, is shown in Figure 4b. The long arclike tablishes itself as new centers intensify, deform the gust
region of updrafts, which extends from the mesocyclone front between outflow and inflow, and, ultimately, choke the
southwestward along the gust front and northwestwardto flow to older centers. This evolution, described conceptually
by Burgesset al. [1982] and illustratedwith observationsby
the small region of updrafts near x = 7, y = 25 km, is typical
of tornadic storms. (Vertical velocity contours are broken inJohnson et al. [1987], may cause tornado families [Agee et
the vicinity of the mesocyclone because of low signal al., 1976].
strength near ground (upward integration of the continuity An anomalous shear zone in the radial velocity measure-
equationis usedhere). Horizontal wind vectorsto the eastof mentsfrom both radars at 1530 indicatesthe presenceof an
the mesocyclone(Figure 4a) have been computedby assum- elevated tornadolike circulation [Brown et aI., 1978]. The
vortex signature,between 1.5 and 3 km, was located within
ing that the vertical wind is zero. The error in the horizontal
wind introducedby this assumption,approximatelyequalto the northernmostvorticity maximum. Note that the circula-
the sine of the radar antenna elevation angle times the tion is located on the storm's left rear flank. (A similarly
verticalvelocity,isestimated tobe<1 ms-1.) Theintrudingbehavedand placed circulation appearedabove 3 km in 1826
area of weak reflectivity (x = 12, y = 23 km), dubbedthe measurementsfrom the Del City storm [Brandes, 1981]). As
weak-echoregionor WER by Marwitz [1972]and vault by is often observed with tornadoes [e.g., Brandes, 1978], the
Browning [1964], marks the primary inflow and updraft elevated vortex lies in the vertical velocity gradient between
regionof the storm.Downdraftswithinthereflectivitycore the weak downdraft behind the updraft arc and the small
arerelativelyweak(>-3 m s-1) compared to thosethat updraft at the left end of the arc. Photographs,taken at
subsequentlydevelop. •-1500, when the storm passed to the west of Norman,
The distributionof verticalvorticityat 1.3 km is displayed Oklahoma, show a small funnellike appendage that pro-
in Figure5. Thearresting featureisthelargearcof truded from the storm's cloud base. No surface damage was
structural
with the reportedwith the elevatedcirculationsat either 1500or 1530.
positiveverticalvorticitythat roughlycoincides
updraft arc. Curiously,intenseelevated vorticeshave not beenreported
for simulatedsupercellstorms. Perhapsthis is due to poor
The severalvorticitycenters(nearx = 8, y = 24 km;x =
11.5, y = 22 km; andperhapsat x = 10, y = 17 km) are spatialresolution.
apparentlyrelatedto development cells(up- In the Harrah storm, negativevertical vorticity concen-
of elementary
drafts)withintheHarrahstorm.Newcenters to trates on the left flank of the storm, for example, from x =
arethought
develop infight-hand
portions
ofthearc(when inthe 5.5, y = 30 to x = 5, y = 23 km and beneath a small
viewed
onthe adjoining cell to the southwest (x = 4, y = 19 km).
directionof stormmotion),whileoldcentersdissipate
left. Clearly, the evolutionof centersis tied to storm Anticyclonic vorticity is also found behind the
146 TORNAI.)IC TItUNDERSTORM CHARACTERISTICS

6;,/19/80
2203 CST 6 KM
2 RFtOBR5
i t I 1 I I I i i 1
o o

22

.0

<o
_

•-_•--0 -
_

0-
• 2 -

-2 o 2 ,4 6

VORTICITY(10-3 $ -1 )
Fig. 3. Vertical vorticityat 6-km elevationin the Lindsay,Oklahoma,storm [VasilqlTetal., 1986].The vorticity
contour
intervalis2 x I0 ......
:•s.....
I. Theregionof updrafts
greater
than5 m s-1 is shown
bytheheavyline.

vorticity arc. The basic configurationof positive vorticity on theless,the profiles, except for the twisting term, are readily
the upwind sideof updraftsand negative vorticity downwind reproduced for other storms. In the absence of strong
agrees with the simple twisting of ambient horizontal vortic- vertical gradients of vorticity, the turbulent diffusion of
ity by the storm updraft [see Davies-Jones, 1984, Figure 8]. vorticity is usually small. This mechanismwill be ignored in
The vertical-vorticity maximum that spawns the Harrah the discussion which follows.
tornado residesat the nosesof the vorticity and updraft arcs Especially at low levels, the convergenceterm dominates
(x = 11.5, y = 22 km). Maximum vorticity increases the twisting term. This is primarily due to the small horizon-
slightly with height at 1530 (Figure 6). At each radar analysis tal gradientsof vertical velocity at low levels. The maximum
level, the distribution of divergence was determined and the vorticity production by the twisting of horizontal vorticity
mean value for all grid points within the arbitrarily defined occurs at the leading edge and just upwind of the mesocy-
mesocyclonewas computed. This analysis(Figure 7) reveals clone. Most often the contribution of twisting to the vorticity
that the low-level mesocyclone flow is wholly convergent budget increasestoward middle storm levels.
below 2 km and convergent in the mean up to 3.8 km At 1543, approximately 3 min before tornado touchdown,
elevation. the Harrah storm was sampledagain. Prominent changesin
Figure 8 presents the mean vertical-vorticity tendencies storm structure included an increase in the areal coverage of
-1
within the mesocyclone due to convergence, twisting, and radar reflectivity >50 dBZ and the development of 3 m s
turbulent mixing. Results for the twisting and turbulent rainy downdrafts within the enhanced reflectivity core.
vorticity terms are suspect due to the absenceof scatterers Vertical vorticity increased slightly at all levels (Figure 6).
at low levels and the subsequent loss of the vertical wind Two small-scale shear anomalies were present in the raw
component in regions adjacent to the mesocyclone. None- radial-velocity measurements.The stronger of the two
 BRANDES 147

32
Horizon•tl
wind (a) 1.3km Vertical
wind 0D) 1.3km
...........
• '/' I I [ , , t I '" ' ' I ' ' [ [ I [ ' , 'r 32

28

F tit / 4 --'
b / 4 '"
2 6 1• 14 18 22 2 6 lfl 14 18 22

x-disttnce (Am) x-distance(•)

Fig. 4. The wind field at 1.3 km in the Ha•ah prior to tornadogenesis
(1530CST). (a) Horizontal wind vectorsrelative
to thestorm.Thelongest
vectorrepresents
24m s-] . Radarreflectivity
contours
(in dBZ)•e superimposed,
andthe
40-dBZ contouris accentuated.(b) Ve•ical velocity(in metersper second)with the 40-dBZ contoursuperimposed.The
mesocyclone
thatproduced
theHa•ahtornado,
defined
astheregion
withve•ic• vonicity•10 -2 s-1, isdashed
in
Figures4a and 4b. The tornadodamagepath (beginningnearx = 22, y = 28 km) is shownby stipplingin Figure4a.

natures persistedat the extremeleft end of the updraft and strongestupdraft, resides in the ve•ical-velocity gradient
vertical-vorticity arcs. The secondcirculation,thoughtto be betweenthe updraft and a downdraftthat is formingbehind
the incipient Harrah tornado,was locatedwithin the meso- the occludedgustfront. This configurationof vertical drafts
cyclone at the nose of the vorticity arc. seemstypical of tornadicstorms,for example,the Spencer-
Luther storm[seeBrandes, 1978,Figure 13]andthe Del City
2.3. Tornadic Stage storm [Brandes, 1984b, Figure 4]. Desiccation of cloud
material within the downdraft causes a "clear slot" often
Figure9 depictsthe radarreflectivityandwindpatternsin seen to the right of tornadoes [e.g., Lemon and Doswell,
the Harrah storm 7 min after damagingwinds began. Al-
1979].Suchdowndraftsare thoughtto be inducedby down-
though not readily apparentat 1.3 km, the tornadohas
ward pressuregradientswhich respondto the suddenlow-
formed along a rolled-upor occludedsectionof the gust
front. Details of the surface wind field at this stage are
levelbuildupof vertical vorticity and associatedpressurefall
[Klemp and Rotunno, 1983;Brandes, 1984a].
discussedby Brandes[1978];storm-scale flow patternsare
At 1.3 km a ringlikeregionof strongvertical.vorticitywith
describedby Ray [1976] and Heymsfield[1978].High-
severalmaxima>150 x 10-4 s-1 has developed(Figure
resolutionradar radial-velocitymeasurements
during the
10a). Thisshapeis alsoseenin a fine-scalesimulationof the
tornadicstagein the Del City storm[Brandes,
1981]suggest
thatthe mesocyclone flowconsistsof windbandsthatspiral Del City storm [see Klemp and Rotunno, 1983, Figure 7]
inward about the tornado. and, at least in the Harrah storm, seems linked to the
intensificationof the rainy outflow and its influenceon the
A prominentstructural changein theHarrahstormisthe
development of a strongrainydowndraft withinthereflec- distributionof updraftsand vorticity.The verticalprofileof
tivitycore.Outflowfromthedowndraft caused themesocy- maximumvertical vorticity (Figure 6) revealsthat a rapid
cloneto elongatein the north-to-south Themore increasein vorticity has occurredbelow 4 kin. Little change
direction.
easterlytornadopathafter1553mayalsorespond to the is evident above that level. The observations do not show
intenseoutflow.Although usually rainy the descentof a strongmesocyclone
notquitethisstrong, from the uppermiddle
at thisstage. troposphere
downdraftsare salientfeaturesin all supercelIs [e.g.,Burgesset al., 1977],butrathertheyshow
Maximumupdrafts, ex- an intensificationof the mesocyclonein the lower tropo-
locatedwithinthemesocyclone,
ceed12m s-• (Figure9b). Thetornado, fromthe sphere.Althoughthe transitionto the tornadicstagein
displaced
148 TORNAI)IC TIqUNDERSTORM CHARACTERISTICS

Del City storm was marked by increasedvertical vorticity

throughout a deep layer, vorticity amplification on the
mesocyclone scale was most pronounced below 3 km. A
factor of 4 increase occurred at 0.8 km between 1826 and
1847[Brandes, 1984a, Figure 3]. The radar vortex signature
of the incipient Del City tornado, a maximum at 0.6 km
elevation (1838), could not be distinguishedfrom mesocy-
clone flow above 3 km [Brandes, 1981].
Numerical simulations have shown that the vertical wind
shear(horizontalvorticity) of the ambientair upon which the
storm feeds is a primary source of rotation in supercelIs
[Schlesinger, 1975, 1978; Klemp and Wilhelmson, 1978a, b;
Wilhelmson and Klemp, 1978]. Vertical vorticity is gener-
ated when the horizontal vorticity is tipped into the vertical
by the storm updraft. This rotation source was first sug-
gested in the observational studies of Browning and Landry
[19631, Browning [1968], and Barnes [1970]. However, if'
twisting of horizontal vorticity by updrafts were the only
mechanism operating, only elevated vertical vorticity max-
ima should result [Davies-Jones, 1982]. Obviously, other
mechanismsare required to produce the buildup of vertical
vorticity necessary for tornadoes to develop at ground.
x-distance (km) Rotunno and Klemp [1985] argue that the origin of low-level
vorticity lies with baroclinically generated horizontal vortic-
t;ig. 5. Vertical vorticity at 1.3 km in the Harrah thunderstorm ity that is produced locally as environmental air from the
priorto tornadogenesis (1530CST).Contourintervals are50 x 10-4
s •. The40-dBZcontourandmesocyclone aresuperimposed as in storm's right flank is cooled differentially by the evaporation
}:igure 4b. of precipitationfalling from the storm. Temperature gradi-
ents are produced that create strong horizontal streamwise
vorticity near ground in air that flows parallel to the iso-
therms. Again, the generated vorticity must be twisted into

6 -,,

5 .-

- ;
16'11.. I ]
:>

I;'.....
• ; 1543 '•... I
0 ............
• ........ •....... •, i I
0 1 2 3 4 5

MAXIMUM VERTICAL VORTICITY o

-2( )0 -I00 0 I00 200
Fig. 6. Vertical distributionof maximumverticalvorticityin the
tornado-spawning mesocyclone of the Harrah storm.The 1530and DIVERGENCE (10-4s -•)
1543 CST profilesrepresentthe pretornadic(mature) stage, the
profileat 1553CST is nearlymidwayduringthe life of the tornado, Fig. 7. Vertical distributionof mean divergence(central curve)
and 1603 and 1611 CST are early and late posttornadic stages, and rangein grid-pointvaluesfor the pretornadicstageof mesocy-
respectively. clone development(1530
 BRANDES 149

., the vertical by updrafts to produce tornadoes; however, in

this case the vertical vorticity maximum is close to the
ground. The twisted vorticity and any preexistingvertical
vorticityare then amplifiedby convergencebetweenoutflow
and inflow air masses. The final intensification of the wind to
l !!
tornado strengthoccurs as the circulation interacts with the
ground [Rotunno, 1986].
Figure 10b shows the strong horizontal vorticity, associ-
ated with a veering wind, that exists in the storm's environ-
ment. Some air parcels that enter the updraft and mesocy-
clone at low levels (below 0.5 km) have, at times, horizontal
vorticityin excessof 200 x 10-4 s-1 (e.g., Figure11).
Vorticity maxima at x = 25, y = 34 and x = 29, y = 36 km
lie near strongradar-reflectivity gradients at the edge of the
storm, between the axis of the weak-echo intrusion and
heavier precipitationon the right (when facing in the direc-
tion of the storm-relative wind). Baroclinic generation of

0 •,,,
-40
I I,,,I• • I
0 40 80
horizontal vorticity is likely in this region as precipitation
falls into the inflow air and evaporative cooling takes place.
A componentof vorticity that aligns with the flow results.
With increasingaltitude the source of the inflow air shifts to
VORTICITY GENERATION(10-6s-2)
the storm's right rear [e.g., Brandes, 1984b, Figure 14]. The
c CONVERGENCE TERM implication is that the baroclinic generation of horizontal
..... TWISTING TERM vorticity was confined only to a shallow layer near ground.
..... TURBULENCE TERM In the Harrah storm, rapid generation of positive vertical
vorticity takes place when the inflow air encounters the
Fig. 8. Vertical distributionof meanvertical-vorticityproduction intenseupdraft in northern sectionsof the mesocycloneand
by convergence, twisting, and turbulent diffusion in the tornado-
spawning mesocyclone prior to tornadogenesis(1530 CST).
the strong horizontal vorticity is tipped into the vertical
(Figure 10c). Twisted vorticity and preexisting vertical

Horizontal
wind (a) 1.3km Ver!•ieal
wind (b) 1.3km
421 i • i i I • • i • It • ' • • I • ,/i • I i , , •/ 42
!--.... : .... '9 ......
i.../,:w--. ~ _ ....... .y, ,, ', ', ,, , ,
• ......... ...:,•... , ,,., x/•-, ,, ,, ,, ', ,•

- " / Ii3• ' ' ' \ I ! '" ß

, , • ..... • •, • ,,

t::/ ":"'
26 26 --

22
22 30 34

x-distance
(km) x-distance
(km)
Fig.9. Windfields,
asinFigure
4,except
forthetornadic
stage
(1553
CST).
Thetornado
location
isindicated
bya
dot.Thelongest
vector
represents
36ms
15t} T()RNADIC THUNDERSTORM CHARACTERISTICS

Horizontal vorticity
42
42

38
38

26

14

X......
dist.ance (km) x-distance (km)

Twfsting
term (C) 1.3km Convergence
term (d) 1.3km
42 -'"T---T---r"• I

38 38

"'-'
34 •'""'"
" -
• -
>,
•.--
_
_
,..
",, x, ? 0
',,
-,./
i,/
:
-
cY '0 1
22-
•
' ' 1B 22 26 •0 18 22 26 30 34

z-distance (kin) x-distance (km)

Fig. 10. Distributionof (a) verticalvorticity,(b) horizontalvorticity,(c) twisting-termvorticitygeneration,and(d)

convergence-termvorticity amplificationduring the tornadic stage of the Harrah storm (1553 CST). The contour
interval
inFigure10ais50x 10-4 s-1. The200x 10-4 s-1 horizontal-vorticity
contour
hasbeenadded
toFigure10b.
Thecontourintervalin Figures10cand10dis 50 x 10-6 s-2. The40-dBZcontourandmesocyclone
location
are
superimposedas in Figure
 BRANDES 151

vorticity are then amplified by convergence(Figure 10d). 6

Mesocyclone-meanconvergenceshows little changefrom
the pretornadicstage(Figures7 and 12). The largerrangein
individual values within the mesocyclonestems from the 5
intrusion of the induced downdraft and from strong local
convergence as the rainy outflow collides with inflow air.
The maximum vertical-vorticity production by the conver-
genceterm occurswhere the rainy outflow interactswith the
low-level inflow. Mesocyclone flow is convergent in the
mean throughout the layer of marked vorticity increase.
Mean vorticity amplification by convergencebelow 2 km
shows a correspondingincrease (Figure 13). At this stage,
the mean twisting-term contribution to the vorticity ten-
dency showsthe customaryincreasewith height.

2.4. Posttornadic Stage

The Harrah storm was next sampledat 1603. The areas
with echo coverage>50 dBZ, and regionswith rainy down-
drafts<-4 m s-• had shrunkenconsiderably.
An intense o
!

-?_00 -IOO o IOO 2OO

occlusiondowndraft had developedto the southeastof the
principalupdraft.The downdraftprincipallywasfed by air DIVERGENCE(10-4s-•)
parcelsthat overtookthe stormfrom the rear [Brandes,
1984b, Figure 18]. Fig. 12. Verticaldistribution
of meandivergence
andrangein
Althoughtornado-intensity damageendedat --•1559,me- grid-point values,
mesocyclone
asin Figure7, except
development (1553CST).
forthetornadic
stage
of
socycloneverticalvorticitynearthe groundcontinuedto
increase(Figure 6). Further, severalshearanomaliesper-
sistedalongthe elongated(horizontal)axis of the mesocy-
6

Streamlines 0.3km
42 ] [ • ! ' "' 5

26 .
'-40 0 40 80 120
VORTICITY
GENERATION
[10
2 ......
0 CONVERGENCE TERM
• •a • 2• 3• •4
TWISTING TERM
x-distance (km)
TURBULENCE TERM

Fig.11. Streamlines
showing
origin
of airparcels
entering
the Fig. 13. Verticaldistribution
of meanverticalvorticity
tendency
mesocyclone
at0.3-kin
elevation.
Tickmarksgiven
parcel
locations
duetoconvergence, twisting,
andturbulent
diffusion,
asinFigure8,
at2-rain
intervals.
The40-dBZ
contour
andmesocyclone
locationexceptfor the tornadicstaget,1553
aresuperimposed
asin Figure4b.
152 TORNADIC THUNDERSTORM CHARACTERISTICS

Horizontal
wind (a) 1.3krn Vertical wind
(b)
,,,,,' ...... (- ........
• %%%x ß ./ _ .. -. .. ß x ., X
I , ,.,.... "%
46 t , .... .-. -.
ß &t "" %'
. . .. ,, • • -- '-'--'-• \ \ \ \ •

' .. • ,,,,x_
\ \ \ \'•\\ \ \2
--- . .......
.......... .... t
....... • .... • _.'_.-g.i...., _ . s! • • • • • _

•, .................. /! t t t?'t.
,,,,,, ,, ,,.. xl

::::::::::::::::::::::::::
} XX >} ! ! . -
3•'::'"
"+•_d'd_.• ' J .J J I , , • ,
26 30 34 38 42 46 26 30 34 38 42 46

x-distance (krn) x-distance (kin)

Fig. 14. Windfields,asin Figure4, except(brtheposttornadic

stage(1611CST).Thelongest
vectorrepresents
38m s-1

clone. Low-level mesocycloneintensification causesa ver- 2.5. Summary

tical lapseof vorticitybetweengroundand 3 km. The growth Low-level features observed during the tornadic stage of
of vortici!y toward ground is associated with a downward
supercell development are summarized in Figure 16. The
pressure-gradientforce that is thought to have induced the
primary region of tornadogenesis(the uncircled T) is along
occlusion downdraft [Brandes, 1984a !. Radar observations
an occluded or rolled-up section of the gust front that
from the Del City storm [Brandes, 1984a: Hane and Ray,
separates rain-cooled outflow from ambient air. Low-level
1985]also supportthis conclusion.The range of divergence
inflow to the mesocyclonic circulation, in which the tornado
values, the mean divergence, and the vorticity tendencies
is embedded, travels parallel to the isotherms and through a
due to convergenceand twisting were not much changed
region where the temperature gradient, created by the evap-
from the tornadic stage.
oration of precipitation, enhances the ambient horizontal
When last sampled(1611), the gust front had moved well
streamwise vorticity (in the vicinity of the stationary frontal
to the eastof the mesocyclone(Figure 14a), and the rotational
boundary). The horizontal vorticity is then twisted into the
flow had begun to decrease. The principal updraft had
slowedto 8 m s-• (Figure14b), andthe reardowndrafthad vertical when the flow encountersthe primary storm udpraft.
declined to 8 m s-•. Updraftswereconfined to thewestern The tipped vorticity and preexisting vertical vorticity are
third of the mesocyclone,while downdrafts in excessof 4 m amplified further as the flow continues to converge in the
s- • filledtheeastern half.All airparcels
passing through the updraft. The tornado lies on the updraft side of the vertical
mesocycloneat 1.3 km originated at higher levels on the velocity gradient between the principal storm updraft and
storm's rear [Brandes, 1984a, Figure 2 !]. the rotation-induced downdraft. Many tornadoes are con-
Mesocyclone vertical vorticity was still strong near fined to the lowest 3 or 4 km of the troposphere, but
ground but rapidly diminished with height. The vortex core tornadolike vortices have been observed to form at great
could no longer be detected above 2 km (Figure 6). The heights [Burgess et al., 1977; Lemon et al., 1982]. The
rangeof divergencevaluesand the mean divergencewith the tornadic stage often is preceded (sometimes by tens of
mesocycloneare shownin Figure 15. The flow, exceptfor a minutes) by public reports of funnels aloft and/or radar
neutral layer near ground, was divergent in the mean; the signatures of tornadolike vortices.
convergence term of the vorticity tendency equation had Specific events probably trigger tornadoes in storms
become negative. No new mesocyclonicvorticity centers primed for development.Tornadogenesismay be tied to the
appeared at the gust front nose or along its southern exten- life cycle of individual elemental cells that compose the
sion, and no additionalsevereweather was reportedwith the parentthunderstorm.Spreadingoutflow from elementalcells
storm. in their dissipating stages would alter the low-level
 BRANDES 153

properties
in several
ways.Wheretheoutflowmostinteracts
withstorminflow,localtemperature
gradientswouldbe
enhancedandthebaroclinic
generation
of horizontal
vortic-
ity increased.
Low-levelconvergence
andupdrafts
would
alsostrengthen.
Thetwistingofhorizontal
vorticity
andthe
convergentamplification
ofvertical
vorticity
would
bepro-
moted.The sizeof the interactionzonewheretornadoes
formwouldbe determined by the scaleof the individual
convectiveelements,that is, wouldhave dimensions of
several
kilometers.
Tornadoes wouldsimply
develop
from
the enhanced
vorticity.Or, possibly,
tornadoes
beginas
shearinginstabilities
alonglocallyperturbed
sectionsof the
gustfrontthatseparatestheinflowandoutflowair[Brandes,
1978].Development wouldbe muchlike vortices along
disturbedvortexsheets. Some tornadoesareprobablytrig-
geredbyoutflow boundaries fromneighboring
storms which STORM MOTION
interfere
withthelow-levelcirculation
oftheparental
storm.
Theveryprocesses whichleadto thebuildup
oflow-level 0 5 !0 km
vorticityandtornadogenesis
promote stormdissipation.
The
rotation-induced
downdraftincreasingly
infiltrates
themeso-
cyclonewith time. Radarobservations[Brandes,1984a] Fig. 16. Schematicdiagramof prominentlow-levelfeaturesob-
show that the air that enters the downdraftovertakesthe servedin supercellthunderstorms
duringtornadogenesis.
Arrows
coldandhaslittle showstorm-relativeflow. Rainy downdraftsand rotation-induced
stormfromtherear.Thisair is potentially
downdragt (behindthegustfront)areshownby parallellinesthatrun
streamwisevorticity. Eventually, this air mixes with and
fromtheupperlefttothelowerright.Updraftsareshownby parallel
reducesthe intensityof the updraftandthemesocyclone. linesthatrun fromthe lowerleft to the upperright.The primary
Not all supercelltornadoesfit the patterndescribedin tornadogenesisis shown by an uncircledT. The locationsof less
previoussections. Brandes[1978]showsdual-Doppler wind frequentlyobservedgust-frontcyclonictornadoes(circledT) and
anticyclonictornadoes(circled A) are also indicated.
fieldsfor a multipletornadoproducingthunderstormwhilea
"gust front" tornadowas in progress.The tornadoformed
near the noseof the gust front, well outsidethe mesocy- clone. Its relative position is shownby the encircledT in
Figure 16. Rarer still are anticyclonic tornadoes which
appearto favorgustfrontlocationsto the rightof the strong
• windsthatgirdthe mesocyclone (theencircledA; seeFujita
and Wakimoto[1982]for an example).
Statisticsshow [Burgesset al., 1977]that more than 50%
of the stormswith middle level mesocyclones,
including
many with prominenthook echoes,do not producetorna-
does. Clearly, mesocyclones are neither a sufficient nor
4 necessaryconditionfor tornadoes.Why then are they im-
portantand so often associatedwith tornadoes?Perhaps
theirprimary
roleisin thesculpturing
of theiow-level
wind
flowandtemperaturefieldssothat air parceltrajectoriesand
the isothermsbecomeparallel and the horizontalvorticity
generationprocessdescribedby Rotunnoand Klemp [1985]
can take place.
An importantmotivationfor deployinga nationalnetwork
of Doppler radars is the prospectof improvedtornado
warnings. The purported lead time of tens of minutes is
based primarily on the detection of the middle level meso-
cyclonic circulation and the average lag for the onset of
damagingwindsat ground. If tornadoesare largelylow-level
0 .....I I • phenomena, warning schemes based on the detection of
-200 -I00 0 I00 200 middlelevel rotation are goingto have a highfailure rate.
DIVERGENCE (iO-4s -•)
3. NONSUPERcELL TORNADOES
Fig. 15. Vertical distribution of mean divergenceand range in
grid-pointvalues, as in Figure 7, exceptfor the posttornadicstage Sometornadolikevortices are not associatedwith super-
(1611 CST). cells. Waterspouts are a common phenotype.
154 TORNAI)!C THUNDERSTORM CHARACTERISTICS

observedwith thunderstormsalong cold fronts [Carbone, location above 2 km. A correspondingarclike region of
1983] and with cold upper level synoptic circulations strongvertical vorticity is not present (Figure 21 c); rather,
[Cooley, 1978] are other examples. Some tornadoes that there are severalmaxima, the largestof which coincideswith
occurin the Denver area appearto be more closelyrelatedto the tornado. The distributionof strong and weak updrafts
waterspouts than to supercell tornadoes. Environmental aboutthe tornadois similarto the updraft/downdraftpattern
forcing is weak. The initial circulation often forms near in supercelIs.
ground along convergenceboundaries,at the intersectionof Scale and magnitude in the Colorado storm are small, but
two boundaries, or when boundaries collide. (There are the arclike region of vorticity amplificationby stretching
exceptions, of course; some vortices form aloft and later (convergence)is like that in supercelIs(compareFigures2!d
descendto ground [e.g., Wakimoto and Wilson, !989, Table and 10d). Curiously, at this time, when the tornado is just
I].) A possible sequence of events, taken from the study of becoming visible, mesocyclone vertical vorticity is being
Brady and Szoke [19891, is shown in Figure 17. The initial decreasedby both the stretchingand tilting (twisting)mech-
circulation precedes cloud formation and develops upward anisms(see also Figure 21e). Radar reflectivity values near
from ground as convection builds. the mesocycloneare fairly high (> 10 dBZ), suggestingthat
Figure 18, t¾omthe study of Wilson and Roberts [19901, deducedflow characteristicsare based on strong meteoro-
shows a radar-derived horizontal wind field shortly after two logicalsignals.At earlier times, prior to the developmentof
outflow boundaries had collided on July 15, 1988. The wind the updraft minimum within the mesocyclone, vorticity
field has been reconstructed from essentially "clear air" tendenciesdue to stretchingand tilting were both positive
radar returns since precipitation had not yet developed. [ Wilczak and Christian, 1990, Figure 2].
Strong convergenceand positive vertical vorticity marks the The origin of the cyclonic rotation in Denver tornadoes is
resultant boundary. Labeled vorticity centers T1, T2, and T3 the subject of speculation. Wakimoto and Wilson [1989] and
subsequentlyproduced tornadoes. The evolution of center Brady and Szoke [1989] attribute the ultimate source of the
TI, which produceda tornado between 2204 and 2210 GMT, rotation to horizontal wind shear across preexisting conver-
is presented in Figure 19. Low-level rotation was detected gent boundaries. Shearing instabilities are thought to pro-
•--20 rain before tornadogenesisand developed upward with duce small vortices which amplily to tornadic intensity when
time. As in supercelIs, the vorticity tendency within the they encounter updrafts. Becausetornado proximity sound-
mesocyclone,
thatis, the regionof verticalvorticity>10 -'2 ings may show only weak winds and little wind shear [e.g.,
s •, wasdominatedby convergence termamplification. Brady and Szoke, 1989, Figure 2], possible influences of
Note that if the pretornadic (1530) profile of the Harrah vertical wind shear are discounted.
supercell storm in Figure 6 were subtracted from the remain- An important mesoscale feature on many Colorado tor-
ing profiles, the resulting positive perturbations could be nado days is the orographically induced Denver Cyclone
plotted to yield a vorticity distribution similar to that in which forms under southeasterly wind conditions in the lee
Figure 19. The 1530 profile then approximatesthe influence of an east-west ridge to the south of Denver. Szoke and
of the supercell environment, particularly the low-level Augustine [19901note that tornadoes are 3 times more likely
veering of the shear vector, while the perturbations repre- on cyclone days.
sent changes associated with tornadogenesis. Numerical simulations of the Denver Cyclone [Crook et
Details of low-level storm structure during tornadogenesis al., 1990]indicatethat the cycloneforms only after nocturnal
in the Colorado storm of July 2, 1987, have been described in inversions are removed by daytime heating. Rotation origi-
a preliminary study by Wilczak and Christian [!990]. The nates with baroclinically produced horizontal vorticity that
storm formed when a preexisting stationary convergence developswhen isentropicsurfacesare deformedby the flow
boundary (pseudofront),representedby a radar "fine line" about the obstructingridge. In a subsequentstudy, Crook et
of weak reflectivity, was overtaken by an outflow boundary al. [1991] show that continued heating leads to the develop-
from a nearby storm (Figure 20). ment of boundary layer rolls that may intersect the conver-
Rapid convective growth ensuedalong a bend that devel- gence zone. Concentratedvortices (mesocyclones),thought
oped in the stationary boundary so that when the tornado capable of producingtornadoes,develop at the intersection
was just becoming visible, maximum reflectivity exceeded points.
50 dBZ (Figure 21a). Although the intruding reflectivity Wilczak et al. [1992] found a temperature gradient of
minimum near x = 1 1 and y = 15.5 km and the reflectivity 2.5øC/10
kmandstronghorizontal
vorticityof 100x 10-4
maximum centered at x = 11.5 and y -- 13.5 km resemble s-• with the boundarythat ultimatelyspawnedthe July 2
the WER and hook echo in supercelIs, in this case they tornado. They suppose that the horizontal vorticity was
apparentlyresult in part from the juxtapositionof two cells. baroclinical!ygenerated and that this vorticity played a
(Often, pronouncedhook echoes are observed[e.g., Waki- significantrole in tornado formation. They assert that the
moro and Wilson, 1989, Figure 2d].) The vertical-velocity surface convergenceboundary, baroclinicity, related hori-
-1
field (Figure 21b) shows several updraft centers >3 m s zontal vorticity, and the parent vortex in which the tornado
that are distributed about a mesocycloneand in a trailing line formed are surrogatesfor similar features in supercell
of updrafts.A velocity minimumresideswithin the mesocy- storms.

clone (x = 8.2, y = 13.6 km). A downdraft existed at this Thus the observational studies (see also Carbone
 BRANDES 155

4 km--

t = 0 rain
3 km- No clouds

Outflow
boundary"
-• • Convergence
zone
1kin
-- ,..... ••fUpward
motion
Weak,
low-level
circulation
atintersection
• ?/x x/x//?/'?/'?/?

t = 15 min

New convective development

4 km--

3 km--

2 km--

Ikm-- ICirculation
advected
vertically
//????////////////////•////"?////////////////////'2

t = 30 rain

Rapidconvectivegrowth

4 km --

3 km m

IStrong
convective
updrafts
stretch
and
2 km •

intensifyvortexintoF1 tornado

I km --

/xx/////x/,////??///?////2/?/?//////x
Fig.17.Schematic
diagram
ofa nonsupercell
tornado
forming
attheintersection
oftwoboundaries
[œrady
and
Szoke,
156 TORNADIC THUNI)ERSTORM CHARACTERISTICS

2101 UTC WI2qDS/REFLECTIVITY 0.5/2.0 km AGL

2200
////1
////
/// t I /// / / / / /
??.

20. t / I /I // I/ ////• I / / I/ /I I/ /I I 1//

I I I I /////// /////."'•8,{// - .

111
I II II //,,//',/'/
1//'/'// // /I IJ ////'//
/

V' /
"/l I

l.O 3.0 5.0 7.0 9.0 11.0 13.0 15.0

0!3T/• •T Or F12 6er0 70. H/5 --..-4,, X KM 20. M/S ' )

Fig. 18. Radar~derivedwind field at 0.2-km elevation shortly after Fig. 20. Horizontal wind field at 0.5-km elevation and reflectivity
the collision of two boundaries(2200 GMT). The contoursdepict (in dBZ) at 2 km prior to tornadogenesisin a Colorado thunderstorm
vertical-vorticity
maxima.The contourintervalis 5 x 10-3 s-•. observedon July 2, 1987,at 2101 UTC [Wilczak and Christian, 1990].
Labeled centers spawned tornadoes JWilson and Roberts, 1990].
a) 2124 [ITC WINDS/REFLECTIVI• 0.5/2.0 km AGL
20.

suggestthat local concentrationsof baroclinicallygenerated

vorticity, likely in all casesinvolving outflow boundariesand
boundary layer rolls, also may be the primary source of
rotation for nonsupercelltornadoes. Supercell tornadoes, •// / / I //
waterspouts,and nonsupercelltornadoes may not be sepa- ' //// 1/
•' ///11!
rate genre.The differencesin intensityamongthesevortices ///1//
may lie with the strengthof the updraftsthat tilt and amplify
the horizontal vorticity.
;;; ,,,,
111 1/111
/111 /1111

re. ,/
I0
/ '1 / / ,,,,,,'
_ Vorticity _ /
/ I '•"
_ (XlO-3 s-• ) _
8
- !0 -
10.0 ß- "'•
///1!

-
-

/•iO -.

t.0 3,0 5.0 7.0 9.0 11.0 13.0 15.0

_ 1,50 _ X gM 20. H/$ •

Fig. 21. Horizontal wind field and kinematic properties of the

2140 2150 2200 2210 2220
Colorado tornadic thunderstorm of July 2, 1987, at 2124 UTC
TIME (GMT) [Wilczak and Christian, 1990]. The horizontal wind field and reflec-
tivity (Figure21a) are as in Figure20. Figures21b-21 e are for 1-km
Fig. 19. Time versus height plot of maximum vertical vorticity, elevation. Contour intervals
are1m s-• in Figure2lb, 2.5 x 10-3
estimated from radial wind measurements, in circulation T1 of s- • in Figure21c, and5 x 10-6 s-6 in Figures21d and21e. The
Figure 18. The tornado occurred between 2204 and 2210 GMT mesocy½lone, definedby vorticity->10 x 10-3, is indicated
by an
[Wilson and Roberts, 1990]. open
 BRANDES 157

20.0 b) 2124
UTC VERTICAL
VELOCITY 1.0kmAGL c) 2124 UTC VERTICAL VORTICITY 1.o km AGL

18.0
L
-1.36

16.0

L:
-2.53

12.0'

lO.O

-2.75
L
-2. •
'L
6.0
[.0 3.0 5.0 7.0 9.0 11.0 13.0 15.0 1.0
3.0 5,0 7.0 9.0 11.0 13.0 15.0
X KM

d) 2124 UTC STRETCB/NG 1.0 km AGL e) 2124UTC TILTING 1.0km AGL

/' . .-40_3.

l.O
......... ....... .... .....
3.0 5.0 7.0 9.0 11,0 13.0 15,0 1.0 3.0 5.0 7.0 9.0 11.0 13.0 15.0
X KM

Fig. 21. (continued)

4. SUMMARY AND RECOMMENDATIONS

radarobservations
willremaintheprincipal
basisforissuing
severeweather warnings in the foreseeablefuture.
Dopplerradar hasbecomethe primarysensorfor observ- Althoughtornadolike vorticescandevelopat anyheight
ing tornadic storms. The observationsare an important withinthunderstorms, mostsupercellandnonsupercell
tor-
complement to numerical simulations as we seek to com- nadoesdevelopin the lower troposphere.
Thesecirculations
plete our knowledgeof tornadogenesis.Moreover, Doppler apparently
tap the rich supplyof baroclinica!ly
158 TORNADIC THUNDERSTORM CHARACTERISTICS

horizontal vorticity that exists near ground in regions of Browning, K. A., Airflow and precipitation trajectories within
strong temperature gradients. This vorticity is tipped and severe local storms which travel to the right of the winds, d.
Atmos. Sci., 21,634-639, 1964.
amplified within the storm updraft. Paradoxically, surface
Browning,K. A., The organizationof severelocal storms,Weather,
friction further enhances the intensity of the surface winds 23, 429-434, 1968.
by allowing low-level air to approach closer to the axis of Browning,K. A., and C. R. Landry, Airflow in convectivestorms,
rotation. Q. d. R. Meteorol. Soc., 88, 117-135, 1963.
Why some tornadolike vortices form and stay aloft is not Browning,K. A., and F. H. Ludlam, Airflow in convectivestorms,
clear. They develop along vertical extensions of the same Q. d. R. Meteorol.Soc., 88, 117-!35, 1962.
Burgess,D. W., R. A. Brown, L. R. Lemon, and C. R. Safford,
boundariesthat produce tornadoes, suggestingthat at least Evolution of a tornadic thunderstorm, in Preprints, loth Confer-
initially they respondto the same mechanismsthat act near ence on Severe Local Storms, pp. 84-89, American Meteorolog-
ground. Becausethey do not make contact with the ground, ical Society, Boston, Mass., 1977.
these circulations are probably weaker than tornadoes. Burgess,D. W., V. T. Wood, and R. A. Brown, Mesocyclone
While much has been learned about supercell and nonsu- evolution statistics, in Preprints, 12th Confi'rence on Severe
Local Storms, pp. 422-424, American MeteorologicalSociety,
percell tornadic storms, some important questionsremain. Boston, Mass., 1982.
We still cannot predict,when tornadoeswill form within Byers,H. R., andR. R. Braham,TheThunderstorm,
287pp., U.S.
storms, preciselywhere they will form, and what their size Department of Commerce,Washington,
D.C., 1949.
and strength will be. Carbone, R. E., A severe frontal rainband, II, Tornado parent
vortex circulation, J. Atmos. Sci., 40, 2639-2654, 1983.
'l'() complete our understandingof tornadic storms, we Chisholm,A. J., and J. H. Renick, The kinematicsof multicell and
need new high-resolution data sets t'rom state-o/:the-art supercellAlbertahailstormsAlbertaHail Stud. Rep. 72-2, pp.
radar systemsthat permit determinationof the fine-scale 24-31, Res. Counc. of Alberta, Edmonton, 1972.
structt,re of storms.()f particularinterestis the sequenceof Cooley,J. R., Cold air funnelclouds,Mon. WeatherRev., 106,
events that lead to tornadogenesis.Coordinated photogram- 1368-1372, 1978.
Crook, N. A., T. L. Clark, and M. W. Moncrieff, The Denver
metric observations&om the field are necessaryto relate the cyclone,1, Generationin low Froudenumberflow, J. Atmos.
visual characteristicsand kinematic properties o!' tornadic Sci., 47, 2725-2742, 1990.
storms.Finally, more studiescomparingradar-observedand Crook, N. A., T. L. Clark, and M. W. Moncrieff, The Denver
numericallysimulatedstormsare needed. Understanding cyclone,lI, Interactionwith the convectiveboundarylayer, J.
Atmos. Sci., 48, 2109-2126, 1991.
situations where models Ifil to reproduce observed storms Davies-Jones,R. P., Observationaland theoreticalaspectsof tor-
are just as important as the successes. nadogenesis, in Intense AtmosphericVortices,edited by L.
Bengtsson andJ. Lighthill,pp. 175-189,Springer-Verlag,
New
York, 1982.
Davies-Jones,R. P., Streamwisevorticity: The origin of updraft
A•'know!edgment.This studybenefittedfrom indispensable
sug- rotationin supercellstorms,J. Atmos.Sci., 41, 2991-3006,1984.
gestions andcomments
providedby RobertP. I)avies-Jones.
Bob's l)avies-Jones, R. P., Tornado dynamics, in 7'htmderstortns:A
in-depthand thought-provoking
reviewsare alwaysa sourceof Social,Scientific,and Technological
Documentary,vol. 2, Thun-
inspiration. derstormMorphologyand Dynamics,2nd ed., edited by E.
Kessler,pp. 197-236, Universityof OklahomaPress,Norman,
1986.
REFERENCES Davies-Jones,R. P., Tornadointerceptionwith mobileteams, in
Instrumentsand Techniques fi•r Thunderstorm
Observationand
Agee,E. M., J. T. Snow,andP. R. Clare,Multiplevortexfeatures Analysis,2nd ed., pp. 23-32, Universityof OklahomaPress,
in the tornadocycloneand the occurrenceof tornadothmilies, Norman, 1988.
Mort. Weather Rev., l(bl, 552-563, 1976. Dennis,A. S., C. A. Schock,andA. Kosceilski,Characteristics
of
Barnes,S. L., Someaspectsof a severe,right-movingthunderstorm hailstormsof western South Dakota, J. Appl. Meteorol., 9,
deduced from mesonetwork rawinsonde observations, J. A tmos. 127-135, 1970.
Sci., 27, 634-648, 1970. Foote, G. B., and H. W. Frank, Case studyof a hailstormin
Brady,R. H., andE. J. Szoke,A casestudyof nonmesocyclone Colorado,III, Airflow from triple-Dopplermeasurements,J.
tornadodevelopmentin northeastColorado:Similaritiesto wa- Atmos. Sci., 40, 686-707, !983.
terspoutfbrmation,Mort. WeatherRev., 117,843-856,1989. Foote, G. B., and C. G. Wade, Case studyof a hailstormin
Brandes,E. A., Mesocycloneevolutionand tornadogenesis:
Some Colorado,I, Radarechostructureandevolution,J. Atmos.Sci.,
observations,Mort. Weather Rev., 106,995-1011, 1978. 39, 2828-2846, 1982.
Brandes,E. A., Finestructureof the Del City-Edmondtornadic Fujita,T. T., andR. M. Wakimoto,
Anticyclonic
tornadoes in 1980
mesocirculation,Mort. WeatherRev., 109, 635--647,1981. and 1981,in Preprints,12thConJkrence
on SevereLocal Storms,
Brandes,E. A., Relationshipsbetweenradar-derivedthermody- pp. 401-404,AmericanMeteorological Society,Boston,Mass.,
namic variablesand tornadogenesis,Mon. WeatherRev., 112, 1982.
1033-1052, 1984a. Hane,C. E., andP.S. Ray, Pressure
andbuoyancy fieldsderived
Brandes,E. A., Vertical vorticity generationand mesocyclone from Dopplerradardatain a tornadicthunderstorm,J. Atmos.
sustenancein tornadic thunderstorms:The observational evi- Sci., 42, 18-35, 1985.
dence,Mort. WeatherRev., 112, 2253-2269,1984b. Heymsfield,
G. M., Kinematic
anddynamicaspectsof theHarrah
Brandes,E. A., R. P. Davies-Jones,
andB.C. Johnson,Streamwise tornadicstormanalyzedfrom dual-Dopplerradar data, Mon.
vorticityeffectson supercellmorphologyand persistence,
J. Weather Rev., 106,233-254, 1978.
Atmos. Sci., 45,947-953, 1988. Johnson,K. W., P.S. Ray, B.C. Johnson,andR. P. Davies-Jones,
Brown,R. A., L. R. Lemon,andD. W. Burgess,Tornadodetection Observationsrelated to the rotationaldynamicsof the 20 May
by pulsedDopplerradar,Mon. WeatherRev.,106,29-38,1978. 1977tornadicstorms,Mon. WeatherRev., 115,2463-2478,
 BRANDES 159

Klemp, J. B., Dynamics of tornadic thunderstorms, Annu. Rev. simulatedsupercell

thunderstorms,
J. Atmos.Sci., 42, 271-292,
Fluid Mech., 19, 369-402, 1987. 1985.
Klemp, J. B., and R. Rotunno, A study of the tornadicregionwithin Schlesinger,
R. E., A three-dimensional
numericalmodelof an
a supercellthunderstorm,J. Atmos. Sci., 40, 359-377, 1983. isolateddeep convectivecloud: Preliminaryresults,J. Atmos.
Klemp, J. B., and R. B. Wilhelmson, The simulation of three- Sci., 32,934-957, 1975.
dimensional convective storm dynamics, J. Atmos. Sci., 3.5, Schlesinger,
R. E., A three-dimensional
numericalmodelof an
1070-1096, 1978a. isolatedthunderstorm,I, Comparative
experiments
for variable
Klemp, J. B., and R. B. Wilhelmson, Simulations of right- and ambientwind shear,J. Atmos. Sci., 35, 690-713, 1978.
left-moving storms produced through storm splitting, J. Atmos. Szoke,E. J., andJ. A. Augustine,A decadeof tornadooccurrence
associatedwith a surface mesoscale flow feature--The Denver
Sci., 35, 1097-1110, 1978b.
Lemon, L. R., The flankingline, a severethunderstormintensifica- Cyclone,in Preprints,16th Conferenceon SevereStorms,pp.
tion source, J. Atmos. Sci., 33, 686-694, 1976.
554-559,AmericanMeteorological Society,Boston,Mass.,1990.
Vasiloff,S. V., E. A. Brandes,R. P. Davies-Jones,
andP.S. Ray,
Lemon, L. R., and C. A. Doswell, III, Severe thunderstorm
evolutionand mesocyclonestructureasrelatedto tornadogenesis,
An investigationof the transitionfrommulticellto supercell
storms,J. Clim.Appl. Meteorol., 1022-1036,1986.
Mon. Weather Rev., 107, 1184-1197, 1979. Wakimoto,R. M., and J. W. Wilson, Non-supercelltornadoes,
Lemon, L. R., D. W. Burgess,and L. D. Hennington,A tornado Mon. Weather Rev., 117, 1113-1140, 1989.
extendingto extreme heightsas revealedby Doppler radar, in Weaver,J. F., andS. P. Nelson,Multiscale
aspects
of thunderstorm
Preprints,12thConferenceon SevereLocal Storms,pp. 430-432, gustfrontsandtheireffectson subsequent stormdevelopment,
American MeteorologicalSociety, Boston,Mass., 1982. Mon. Weather Rev., I10,707-718, 1982.
Marwitz, J. D., The structureand motion of severehailstorms,I, Wilczak,J., and T. Christian,A vorticityanalysisof the non-
Supercellstorms,J. Atmos. Sci., II, 166-179,1972. supercell,
2 July1987tornado,in Preprints,
16thConference
on
Ray, P.S., Dual-Dopplerobservation
of a tornadicstorm,J. Appl. SevereLocal Storms,pp. 560-565,AmericanMeteorological
Meteorol., 14, !521-!530, 1975. Society, Boston,Mass., !990.
Ray, P.S., Vorticityanddivergence fieldswithintornadicstorms Wilczak,J. M., T. W. Christian,
D. E. Wolfe,R. J. Zamora,andB.
fromdual-Doppler observations,J. Appl.Meteorol.,I5,879-890, Stankov,Observations of a Coloradotornado,I, Mesoscale
1976. environment andtornadogenesis, Mon. WeatherRev., 120,497-
Ray, P.S., B.C. Johnson, K. W. Johnson, J. S. Bradberry,
J. J. 520, 1992.
Stephens, K. K. Wagner,R. B. Wilhelmson, andJ. B. Klemp, Wilhelmson, R. B., andJ. B. Klemp,A numerical studyof storm
The morphology of severaltornadicstormson 20 May 1977,J. splitting that leadsto long-livedstorms,J. Atmos.Sci., 35,
Atmos. Sci., 38, 1643-1663, 1981. 1974-1986, 1978.
Rotunno,R., Tornadoesandtornadogenesis,
in Mesoscale
Meteo- Wilson,J. W., andR. D. Roberts,Vorticityevolutionof a non-super
rologyand Forecasting,editedby P.S. Ray, pp. 414-436, ceil tornadoon 15 June 1988 near Denver, in Preprints,16th
AmericanMeteorologicalSociety,Boston,Mass., 1986. Conference on SevereLocal Storms,pp. 479-484,American
Rotunno,R., andJ. B. Klemp,Ontherotationandpropagation
of Meteorological Society,Boston,Mass.,
 Tornadoes
andTornadicStorms'A Reviewof Conceptual
Models
CHARLESA. DOSWELLIII ANDDONALDW. BURGESS

NationalSevereStortnsLaboratory,Norman, Oklahoma73069

1. INTRODUCTION with the involvementof the Departmentof Meteorologyat

the Universityof Oklahoma[seeGoldenand Morgan, 1972;
Thedefinitionof a tornadoin theGlossary
of Meteorology Moller et al., 1974].Stormchasing(whichhasbecomea part
[Huschke,1959]beginswith the following:"A violently of many scientificprogramsoutsideof Oklahoma) has pro-
rotatingcolumnof air, pendantfroma cumulonimbuscloud, vided an opportunity for meteorologiststo observe the
and nearly always observable as a 'funnel cloud' or tuba." visually recognizable characteristics of both tornadoes and
This definitionseemsrelatively straightforward, but in real- the storms associated with them.
ity distinctionsimplied by the words we use tend to blur. Finally, the developmentof computermodelscapableof
Our ability to make distinctionsamongtornadicstorm realistic simulationsof tornadic storms [Schlesinger, 1975;
types and among tornadoes has been affected by three Klemp and Wilhelmson, 1978] and tornadoes [Rotunno,
importantdevelopmentssincethe last tornadosymposium: 1979]provideda meansfor careful quantitative evaluationof
(1) high-resolutionDoppler radar observations,(2) extensive the physicalprocesseswhich yield tornadoes.With time, the
visual observationsby storm interceptteams, and (3) de- sophisticationof these models has continued to grow, and it
tailed three-dimensional numerical cloud models. The defi-
is now possibleto produce "tornadolike vortices" within a
nition and classification of convective vortices and the
numerical model of the entire storm_[Wicker, 1990].
storms which produce them have become more complex One thing has become quite dear during this evolution:
and, in many ways, more troublesomethan ever before. On the storm type Browning called a "supercell" producesby
the other hand, the complexity we are encounteringis really far the most intense convective vortices and certainly is the
a positive sign that our understandingis growing. type of storm most likely to produce them. Indeed, most of
While it was well known long before the early 1960sthat the radar reflectivity structuresthat were associatedempir-
tornadoeswere associatedwith deep, moist convection,the ically with tornadoes have come to be recognized as char-
relationship between these intensevortices and the convec- acteristicof supercelIs.
tive storms with which they are associatedwas not well However, the quantitative evaluation of physical pro-
understood. Beginningin the 1950s,with the deploymentof cessesmade possibleby numerical models has made it clear
weather radar, it became clear that at least some tornadic that the morphologyof radar reflectivity alone is inadequate
stormsexhibitedspecialcharacteristics[seeStoutand Huff, for categorizingconvective storms. Instead, a persistent
1953; Garrat and Rockney, 1962]. Careful, systematicexam- correlation (positive or negative) between vertical velocity
ination of suchstormsbeganin the early 1960swith the work and vertical vorticity has come to be recognizedas the most
of Browning [Browning and Donaldson, 1963; Browning, usefuldefinitionof a supercell [Weisman and Klemp, 1984].
1964] using reflectivity radar as the primary observational That is, the airflow pattern most characteristicof a supercell
tool. Research Doppler radars began to make detailed, is a deep and persistent cyclone, called a "mesocyclone"
systematicobservationsof airflow in tornadic stormsin the [see Burgess and Lernon, 1990].
early 1970s[e.g., Burgessand Brown, 1973],just priorto the Many convective vortices are associatedwith nonsuper-
last symposium. cellular convection, however. The ways in which these
In 1972, buildingon early pioneeringwork like that of nonsupercell-relatedevents arise remain quite inadequately
Ward [1961], an organizedprogramto intercepttornadic understood, with some efforts to deal systematically with
storms began at the National Severe Storms Laboratory, one type having only just begun [Brady and Szoke, 1988;
Wakimoto and Wilson, 1989].
The Tornado:Its Structure,Dynamics,Prediction,andHazards. As we study the tornadic vortices themselves, it is becom-
GeophysicalMonograph '79
Thispaperis notsubjectto U.S. copyright.
Published
in 1993by the ing evident that not all tornadoesare the same, and a proper
American GeophysicalUnion. classificationof them is not simple. Understanding that a
162 CONCEFFUAL MODELS OF TORNADOES AND TORNADIC STORMS

vortex is a kinematically defined processrather than an cipitation, radar reflectivity may not reveal the circulation
object, with different air parcels participatingin the flow adequately,if at all. LP stormsfrequentlyare nontornadic,
frommomentto moment,turnsout to be an importantnotion and many are nonseveredespiteexhibitingpersistentrota-
in trying to classify observedevents. This, in turn, affects tion.
our climatological record of events, as we shall discuss At the other end of the supercell spectrum are the so-
below. calledhigh-(or heavy-)precipitation(HP) supercelIs(Figure
The dynamics of vortices, such as those simulated numer- 2). WhereasLP stormshave little or no precipitation(and
ically [e.g., Rotunno, 1979;Gall, 1983]and in the laboratory hence low reflectivity) within their mesocyclones, HP
[e.g., Ward, 1972; Church et al., 1979], certainly are perti- stormsare characterizedby substantialprecipitation within
nent to tornadoes. Such studies have improved our under- their mesocyclonic circulations. When HP storms have a
standing of phenomena like multiple vortices and vortex recognizable hook echo (many do not), reflectivities in the
breakdown,but they concernpropertiesof vorticesin gen- hook will be comparableto those in the precipitationcore.
eral, not necessarily those of tornadoes. From a purely HP supercelIs are probably the most common form of
dynamical viewpoint, tornadoes arise from amplification of supercell, occurring not only in the humid half of the United
either existing or locally created vorticity [Rotunno, 1986; Stateseastof the Mississippibut also westward into the high
Davies-Jones, 19861. However, this is a somewhat abstract plains. They produce severeweather of all types (including
tYamework tbr understanding tornadoes. This paper at- tornadoes)and, unlike other types of supercelIs, also may
tempts to review tornadoes in the context of the convective produce torrential, flash flood-producingrainfalls [Moller et
events giving rise to them. In particular, we shall distinguish al., 19901. Some of the distinctive radar echoes [Forbes,
between cvenls associated with supercell storms and those 1981] traditionally associated with tornadic storms, like the
produced in association with nonsupercellular convection. so-called bow echoes, comma echoes, and line-echo wave
Not everything we present will have been thoroughly inves- patterns (or LEWPs [Nolen, 1959]) can be associated with
tigated in the scientificliterature, especially nonsupercellular HP supercell storms. The rationale for including these forms
events; in l'act, we wish to mention some of these lesser in the HP supercell class is that they result from persistent
known phenomena in hopes of stimulating their systematic mesocyclonesembedded within precipitation-filled regions
study. of the storm.
Because HP supercelIs often occur in humid, cloud-filled
environments, visible signs of rotation may difficult to de-
2. S u I't';R('},;i,I• ST½)RMS
tect. In contrast, since the circulation in HP storms is
embedded within precipitation, radar reflectivity usually
2.1. The Supercell Sper'trum
depictsthe HP storm circulationsreadily, sometimeseven as
We have chosenthe presenceof a deep, persistentmeso- curved bands apparently aligned with the flow.
cyclone to be the single distinguishing characteristic of Finally, between these two extremes is the classic super-
supercelIs.(By "deep," we mean a significantfraction of the cell, which exhibits moderate precipitation production (Fig-
depth of the cumulonimbuscloud in which the circulation is ure 3). Such storms typically match the traditional supercell
embedded (several kilometers). By "persistent," we mean conceptual models [e.g., Browning, 1964; Lemon and
in comparisonto a convective time scale, defined by the time Doswell, 1979] and are most common in the transitional
it takes for air parcels to rise from within the inflow layer of environments of the Great Plains. Many of the tornadic
the updraft to the anvil outflow (a few tens of minutes).) storms in major tornado outbreaks east of the Mississippi
Even within the category of supercell storms, however, it River are of the classic variety, however.
turns out that distinctions that appear to have significance Althoughthere may be someprecipitationwithin a classic
can be made [see Doswell et al., 1990]. While we concur with supercell'smesocyclone,it typically is not heavy. Note that
the kinematic-dynamic approach for defining a supercell, because a radar (even at its lowest elevations) scans a storm
first advocated by Browning [1977] and recently reempha- above the surface, a region with little or no surface precip-
sized by Weisman and Klemp [1984], it appears that the itation may still be within radar-detectable precipitation
amount and spatial distribution of precipitation with the aloft. If such a storm has a hook echo (and many do), the
convection are important indicators of the weather phenom- hook reflectivities will be less than those of the precipitation
ena associated with a particular storm. core. Late in a classic supercell'slife cycle, during collapse
Some supercellstorms produce relatively little precipita- of its updraft[see Lemon, 1977],the mesocyclonemay fill
tion and yet show clear visual signs of rotation (Figure 1). with precipitation, but this should not be considereda
Such storms have come to be called low-precipitation (LP) transition to an HP supercell unless the mesocyclone per-
supercelIs[Bluestein and Parks, 1983]. LP supercelIsoccur sists well after the collapse phase.
most often near the surface dryline and, owing to the sparse ClassicsupercelIsare readily detectableboth visuallyand
precipitationand relatively dry environmentswith little or via radar reflectivity and produce a full range of severe
no interveningcloudiness, cloud structuresshowingrotation weather, but only rarely are they associatedwith flash
are visible readily to a suitably positioned observer. flooding.ClassicsupercelIsprobablyaccountfor the major-
On the other hand, precisely because of the sparse pre- ity of violent (F4-F5)
 DOSWELL AND BURGESS 163

LOW PRECIPITATION SUPERCELL (a)

Small Hail

ModerateRainand/or
Large Hall

Anvil Eclß
Flanking Line

0 5 0 km

0 5 mt

LOW PRECIPITATION SUPERCELL (b)

Overshooting Top

Anvil

Precipitation
(Possible Hall)
I

Fig. 1. Low-precipitationsupercellschematics for(a) low-levelradarstructureandcloudfeatureslookingdownfrom

above and (b) visual structuresfrom the viewpointof a nearbyobserveron the ground.

2.2. Hybrid Events ble, naturally. The tornadic storms reported on by Burgess
and Davies-Jones [ 1979] and Burgess and Donaldson [1979]
Since class distinctions are much less obvious in the real
producedintensetornadoes, and yet, as LP storms,they had
atmospherethanthey are in the abstract,it is quitecommon little or no distinctive radar structure.
to see events that do not fit the preceding prototypes
Supercell storms seem capable of evolving from LP to
precisely.For example,it is likelythat mostLP stormsdo
not becometornadicunlessthey evolvealongthe supercell classical, from classical to HP, and so on. As noted by
spectrum towardthe classical structure.TheJune5, 1982, Dos•,ell et al. [1990], the variety of radar reflectivity mor-
Borger,Texas,tornadicstorm(Figure4) hada visualap- phologies, especially within the HP group, can be quite
pearancethat mightsuggest confusing (see also lmy and Burgess [1991]). Nonsupercell
it to be an LP storm,but its
appearance on radarwas morelike a classicsupercell, convection can evolve into supercelIs [Bltrge$s and Curran,
exhibiting
a substantial
hookecho.Exceptionsareinevita- 1985]and vice versa. The common factor in all supercelIs
164 CONCI,•P
I'UAL MODELSOF TORNADOES
AND TORNADICSTORMS

HEAVY PRECIPITATION SUPERCELL (a)

• Light Rain
and Outflow

Inflow Band
Heavy Rain
and Hall -
Updraft

0 10 20 km

0 10 mi

HEAVY PRECIPITATION SUPERCELL (b)

r o Tp

Anvil

Fig. 2. High-precipitation supercell schematicsas in Figure 1.

the deep, persistent mesocyclone, regardless of the storm's single, persistent "cell" is arguably the most commonly
precipitation characteristics. However, the observed varia- accepted radar characteristicassociatedwith supercelIs.The
tions in precipitationamount and distributionmake supercell difficulty with this as a defining characteristic is that when
recognition a challenge, especially when that recognition observingconvective storms visually or with especially high
depends mostly on non-Doppler radar observations; this resolution radar, it turns out that a multicellular structure
situation that will be remedied in time with the deployment can be observed to be superimposed on most convective
of the operational Doppler radar (the WSR-88D network). storms, including supercelIs. Although the Byers-Braham
prototype convective "cell" typically is depicted as
2.3. Supercell Identification Criteria "plume"-like [e.g., Weisman and Klemp, 1986, Figure 15.1],
Having focused on the mesocyclone as the criterion for such cells really are more "bubble'Mike, even in supercelIs
identifying supercelIs, we wish to review some of the tradi- (compare Figure 12b with Figure 12a of Newton [ 1963] and
tionally acceptedsupercell characteristics.The presenceof a see Hane and Ray [1985], especially their Figure
 DOSWELL AND BURGESS 165

CLASSIC SUPERCELL (a)

aft

Flanking
Line• 20
krn
0 10 ml

CLASSIC SUPERCELL (b)

• Overshooting
Top
Anvil

Fig. 3. Classic supercell schematicsas in Figure I.

Associated with the single-cell notion is another com- ground" processthat evolves only slowly over periodsof a
monly employed yardstick to identify supercelIs: their ten- few hours, the characteristic lifetime of the constantly re-
dency for "steady state" character. As with the single-cell generating supercell structure [Foote and Frank, 1983]. In
criterion, this simply does not hold up to detailed observa- extreme cases, supercelIsevolve very slowly indeed and may
tions. SupercelIsproducing "tornado families" (cyclic tor- have tornadoeson the ground for periods approaching(and
nado-producing storms) undergo an evolution over time occasionallyexceeding) an hour. Events of this character are
scales of the order of several tens of minutes (as described quite rare, and it is not yet understood how such steadiness
by Lemon and Doswell [1979] and Burgesset al. [1982]). arises. If such steadinessis a defining characteristic,then the
There also is the constantly evolving cellular structure supercellclass is quite sparsely populated. In our opinion,
superimposed
on the overall stormevolutionthat hasa time there has been so much published emphasison the steadiness
scale of several minutes. These subprocessesmake even an and unicellularity criteria for supercelIsthat operational iden-
approximatelysteadystateunlikely.Nevertheless,it canbe tificationof supercelIsusing these characteristicsoften misses
argued that supercelIsdo exhibit a long-lasting"back- lessprototypical(but still clearly supercellular)
166 CONCEPTUAL MODELS OF TORNADOES AND TORNAD1C STORMS

Fig. 4. A tornadic supercell in the vicinity of Borger, Texas, on June 5, 1982. Photograph copyrighted 1982 by A.
Moller, used by permission.

On occasion, supercells arise in environments with rela- of storms: it is with such structural knowledge that a radar
tively modest instability, as in tropical cyclones [McCaul, observer can separate bona fide supercell structures from
1991] and in strongly baroclinic systems [e.g., GOllski et al., "imposters" created by particular echo juxtapositions.
1989]. Although the instability may be weak, there can be Since LP storms typically exhibit few, if any. distinctive
lillie doubt of the supercellular character of storms in such reflectivity features, and since detection of the classic echo
events. Although there is always a question about the features is so range- and resolution-dependent, these distinc-
existence of a small-scale, unobserved region of strong tive structures do not seem appropriate criteria for supercell
instability. it seems unnecessary to postulate some mecha- identification.
nism not supported by the existing data. The evidence is Thus on the basis of the above arguments, we advocate a
substantial that supercells do not require an environment deemphasis for many of the traditional supercell identifica-
with strong instability. tion criteria. With Doppler radar data, a time- and space-
Finally, there often has been considerable emphasis on the continuous mesocyclone is the best way for identifying such
deviate motion (from the direction of the mean flow in which storms. With reflectivity alone, the three- and four-
the storm is embedded) of the supercell, even to the point of dimensional echo structure can be used to infer the presence
suggesting it as a defining characteristic. Not all supercells of mesocyclones in many cases. The optimum situation for
deviate significantly from the mean wind direction (see interpretation is when combining velocity and reflectivity
Davies and Johns, this volume]; therefore deviate motion is information with a knowledge of characteristic storm struc-
not required for development of a deep, persistent mesocy- tures. LP storms present a problem to any purely radar-
clone. This is especially so when hodographs are curved [see based identification process, so visual recognition (spotters)
Doswell, 199Ib]. still has an important role to play, even in the era of
Hook echoes and other "distinctive" structures (as dis- operational Doppler radars.
cussed by Forbes [1981]) are the direct resull of the meso-
cyclone circulations of a supercell. The distribution of pre-
2.4. Tornadoes Within Supercells
cipitation quite clearly depends on airflow within the storm,
which we have agreed is the most characteristic feature of a The common association between mesocyclones and tor-
supercell, so such struclUres certainly are potentially useful nadoes in supe:-cells hints that conservation of angular
in identifying supercells. Sometimes, though, such features momentum may explain tornadoes associated with mesocy-
as hook-shaped reflectivity structure can arise as a result of clones. However, even superceU tornadogenesis may be
"configuration" instead of circulation. Thus Lemon's [1977J more complicated than that. Tornado development in the
emphasis on the three- and four-dimensional echo structure vicinity of the so-called wall cloud [Fujita, 1960] suggests
 DOSWELL AND BURGESS 167

Fig. 5. An exampleof a nonsupercell tornadoevent(sometimescalleda "landspout")near Sublette,Kansas,on May

15, 1991.Note that thisis a relativelyhighcloudbase,estimated
at about5000feet (---1525m). Photograph
copyrighted
1991 by C. Doswell.

that nearby downdrafts play an important role in getting teristic of terminology debates in general.) While we use
tornadic/mesocyclonic vorticity to low levels in the storm certain terms that have been common in the vernacular
[Lemon and Dos,vell, 1979; Davies-Jones, 1982; Davies- and/or the literature, we do not necessarily endorse those
Jones and Brooks, this volume]. terms. These events comprise several categories, and we
Intense vortices associated with supercelIs do not neces- will attempt to give a brief descriptionof each.
sarilyall developvia identicalprocesses.
(Recallthe discus-
sion in the introduction, distinguishingbetweenthe limited 3.1. Landspouts
number of abstract mechanismsfor creating intense vortices
In an analogywith the common waterspout[Bluestein,
and the processesoperatingat storm scalesto allow the
vortex dynamicsto operate.)There may well be morethan 1985], most of which develop from nonsupercellstorms,
one mechanismoperatingfor any givenvortexassociated manynonsupercell tornadicevents(e.g., Figure5) arisevia
intensification
of preexisting,shallowvertical vorticesnear
with a supercell,or within differentregionsof the same
the surface,throughsimplevortex stretchingwhena devel-
supercell.
Moreover,thoseeventsleadingto tornadoinitia-
tion may not be the sameas thosemaintaining the large opingconvectiveupdraftmovesover them[seeBradyand
Szo!•e, 1988]. (Of course, some waterspoutsdo arise from
vorticity.Molleret al. [1974]havedescribed
funnelclouds
on the northwestsideof the Union City tornadicstorm,with supercelIs.
They have beenc•dledtornadicwaterspouts
by
cool outflow at the surface,even as the primarytornadowas
Golden [1971] and appear to be virtually identicalto torna-
doesassociatedwith supercelIsover land. The distinction
developing on the southwestern updraftflank,near the betweena tornado and a waterspoutis basicallyof little or
inflow/outflow interface of the same storm. It is hard to
no scientific value.) Doppler radar evidence shows the
imaginethe samestorm-scale processesoperating in these
pretornadicexistence of these vortices on convergence
areas,althoughbothwereintensevorticesin the abstract boundaries[Wilczal,et al., 1991].The detailsof the originof
sense.
these"misoscale"[Fujita, 1981]vorticesare as yet unclear,
but suchpreexistingvorticesmayexplainthe "dark spots"
3. NONSUPERCELLSTORMS seenon the seasurfaceprior to the developmentof common
A varietyof intenseatmospheric
vortices in waterspouts[Golden, 1974],as noted by Wakimotoand
candevelop
association
withnonsupercellstorms. canbea Wilson [ 1989].
Terminology
controversial
topic,butwe donotwishto getinvolved
in Perhapsa related phenomenon is the weakly unstable,
terminologydebates.(Seethediscussion
among linearly
Fanld•auser convectivetornadiceventfirst documentedby Car-
et al. [1983a,b], Doswell[1983],andMoller[1983]for bone [1983]. As with landspouts,a frontal boundarymay
somesenseof theterminologyissues;
although thatdebate developlocallyenhancedcirculationce'nters,whichsubse-
the namesfor cloudfeatures,its flavoris charac- quentlycanattaintornadicproportions.
concerns Whatmakes
168 CONCEPTUAL M()DELS OF TORNADOES AND TORNADIC STORMS

events distinctive is the weak buoyancy in their environ- of cold pool vortices to which we refer only arise more or
ment; the updraftsare forced along the frontal zone [see less directly under the upper circulation center (where the
('arbo•e, 1982],andthe tornadiccirculationsare compara- vertical shear usually is weak) and well poleward of the
ble in depthto the updraft{whichwasshallowto beginwith, surface cold front. Systematic investigation of such events
only a few kilometers). Again, such events have not been has not been done, to our knowledge.
observedoftenenoughto havebeensubjectedto systematic
study.
3.3. Gustnadoes
Preexistingvorticesat low levels also may be associated
with tornadoesarising as convergenceboundariescollide Very smallscale, shallowvortices(Figure 6) may develop
[e.g., 1tollc and Maier, 1980]. Such events are associated near the surface along outflow boundaries and/or cold fronts,
with multicellularlines and clusters,and the resultingflows with or without deep convection overhead [see Idso, 1974,
can bc quite complex. Although multicell stormshave been 1975;Meaden, 1981;Doswell, 1985].The boundarydevelops
lhc st•jcct of many observationalstudies [e.g., Maru itz, "lobes" and bulges, with cyclonic circulationsat the cusps
19721, they have not yet been given the attention they created by those lobes. Sometimes, for reasons that essen-
deservein three-dimensional numericalmodeling.Therefore tially are not known, those circulations become quite in-
the dynamics•)1'interactingconvectivecells are as yet poorly tense; at least as intense as weak tornadoes. If they are
undcrsttmcl. Tornadogenesis under such circumstances is associated with a damaging outflow, they may create short,
thetel'ore correspondingly poorly understood. narrow zones of even more intense damage than is common
along the rest of the outflow. They also can produce damage
3.2. ('old Pool Vortir'es
swaths along an otherwise nondamagingoutflow.
Although we have no documentation for making this
'1'• our knowledge, the sole reference to these is that by distinction, we propose that they are distinguishable from
('ooh, v 119781.'Fhcy seem to be associated with cold pools "landspouts" by remaining quite shallow. Virtually no cir-
aloft, which frequently passoverhead with clear skies in the culations can be seen at cloud base, visually or on Doppler
wake ot' cold fronts. Such cold pools aloft (not necessarily radar. Such events seem not to depend on the superposition-
coincident with the upper circulation center) may be associ- ing of a developing updraft above them. If such a vortex is,
ated with high !apse rates if skies are clear and surface indeed, deepened and intensified by an overriding updraft,
heating couples the boundary layer with the cold air a!ot't. In we believe it will undergo a transition to a landspout.
such cases there is enough residual moisture in the postfron- Obviously, considerablygreater documentation and study of
tal environment that deep convection ensues. In most cases these events are needed.
the cloud base is high owing to lack of abundant moisture,
while cloud tops are low because of a cool troposphere, 3.4. Fair Weather Vortices
giving risc to a low tropopause.
'I'he mechanism by which these cold pool vortices form is There is a substantial variety of distinct fair weather
quite tinclear, because of a lack of quantitative observational convective vortices, even ignoring the "dust devil" phenom-
studies. Since the midtropospheric environment in which enon [see ldso, 1974]. Dust devils arise in association with
they occur may be rich in vertical vorticity, they might result dry rather than moist convection, of course. (Interestingly,
from simple vertical vortex tube stretching. The rarity of some citizens observing the deadly Cheyenne, Wyoming,
tornado touchdowns from these cold pool vortices may be tornado of July 16, 1979, thought they were seeing a dust
associated with the relative weakness of the initial vorticity devil; this confusionmay have arisen becauseof the relative
at low levels (as discussed by Smith and Leslie [1978]). rarity of tornadoesin Wyoming, along with the absenceof a
These events are distinct from those along and ahead of visible condensation funnel for the early part of the torna-
cold fronts. If a cold pool aloft is situated over a front rather do's life.)
than behind it, low-topped storms can develop in such an Meteorologists operating on storm intercept teams have
environment. Whereas wind shears usually are weak be- observed relatively long-lived funnel clouds in association
neath cold pools in the postfrontalregion, stormsalong and with quite ordinary cumulus clouds (Figure 7). A rather
near fronts often arise in relatively highly sheared environ- different phenomenonhas been observed on fair weather
ments; funnel clouds and tornadoes developing in these days, the so-called"horseshoevortex" (Figure 8). These
conditionssimply are shallowversionsof supercelIs.(At the may arise in much the same way as "mountainadoes"
risk of being repetitious, it is the presence of a deep, [Bergen, 1976]: tilting and the associatedstretchingof an
persistentmesocyclonewhich defines a supercell,not the enhancedregion of horizontal vorticity over some upward
depth of convection. When the mesocycloniccirculation protrudingobject,or perhapsby an isolatedupdraft(a small
exists through a substantial fraction of the depth of the cumulus-scale version of the process depicted by Klemp
storm, it doesnot matter if the storm is relatively shallow;it [ 1987, Figure 3a].
is a supercell.Storms poleward of, say, 45ø latitude often With most of these fair weather vortices, it seems unlikely
have low tops becausethe environmentis relativelycold, they ever wouldreachdamagingproportionsat the surface,
with a correspondinglylow tropopause.)The specificclass and so it is improbable that they would (or should)
 DOSWELL AND BURGESS 169

Fig. 6. An example of a circulation alonga gust front (sometimescalled "gustnadoes") near Welch, Texas. on May
23, 1982. In contrast to Figure 5, this cloudbaseis quite low, around500 feet (150 m) or less. Photographcopyrighted
1982 by C. Doswell.

classified as tornadoes. Knowledge that they exist may be ties arise from improvements in observations and under-
important in respondingappropriately to citizen reports of standing,so the problems are really the sign of progress.
such events, however. Many supercelIsproducetornadoesfrom and/or near the
so-called wall cloud (see section 2.1). It has been shown
4. CLASSIFICATION OF VORTICES observationallyand numerically [see Rotunno and K!emp,
1985] that the wall cloud arises from the admixture of
At present,our perspectivehascometo paraphraseRich- outflow and inflow within the mesocyclone. Now, suppose
ardson's famous limerick about vortices: the extratropical
the mesocycloniccirculation becomes so intense that it
cyclonecontainsmesolows,the convectionwithinthe vicin- reachesdamagingproportions, with a wall cloud reaching
ity of mesolowsdevelops mesocyclones,the tornado cy- near, or perhapsdown to, the surface.Is sucha damaging
clonedevelopswithin the mesocyclone, the tornadowithin circulationa tornado? It certainly meets the definitiongiven
the tornadocyclone, the subvortex(or suctionvortex [see in the introduction, as well as that advocated by Forbes and
Fujita, 1971]within the tornado,and soon (presumably, to Wakimoto. Both storm intercepts and eyewitness accounts
viscosity).Thus the processesassociatedwith tornadoes(at
suggest thatmesocyclonic vorticescanbe damaging whether
least those developingfrom supercelIs)can be seen in a
or not they ever produce a visible cloud to the ground.Are
contextof a largervortex and containssmallersubprocesses
such damaging events "straight line" winds? How large
within. In such a hierarchy of processesthe boundaries
does the radius of curvature have to be to call an event
between events can become blurry when observedin the
natural world.
"straight winds" as opposed to a tornado?
Moreover, what about a large wall cloud that spins out
Forbes and Wakimoto [1983] have presenteda quite
visible funnels that develop damagingground circulations
insightfuldiscussion
on classification
of tornadoes.We are
basicallyin agreementwith theirconclusions, whichadvo- every few minutes(either one at ,t time or severalat once)
catea morepragmatic approach to defininga tornadothan over a periodof a few tens of minutes?Are we seeingone
impliedby the glossary namely,anydamagingtornado with many subvortices,or are we seeingseveral
definition,
vortex associated with a convective storm, including its differenttornadoes?Again, sucheventshavebeenobserved
accompanying
windfield,should Forbes and recorded, but how one classtriessuch an event seems
becalleda tornado.
unclear to us.
and Wakimoto[1983,p. 233]alsosuggested, andwe agree,
As visual observations of tornadoes accumulate, it is clear
that "Damagingvorticesnotassociatedwiththunderstorms
[oughtto be] considered tornadic
vortices
of a particular that tornadoesvirtually never "skip" in the senseof the
type." However,we believethe issuescanbe evenmore circulation"lifting and descending";instead,the circulation
Thesedifficul- at the surfacemay strengthenandweakenon time scales
difficultto resolvethantheyhavedescribed.
170 CONCFPTUAL MODEI. S OF TORNADOES AND TORNADIC STORMS

}:ig. 7. An example of a lair weather vortex with an ordinary Fig. 8. An example of a vortex associated with a dissipating
cumuluscloud near Sayre, Oklahoma,on May 17, 1983.Photograph cumulus cloud (sometimes referred to as a "horseshoe vortex")
copyrighted 1983 by C. Doswell. near Shamrock, Texas, on April 13, 1976. Photographcopyrighted
1976 by C. Doswell.

a few secondsor more, but a significantcirculationtypically does?This issueis complicated by the existence of multiple-
remains on the ground for the lifetime of the event. The vortex phases interspersed with single-vortex modes.
funnel cloud aloft associatedwith the event may be contin-
uous during such a weakening and strengtheningcycle, or it,
5. DISCUSSION
too, may dissipate and redevelop. If the winds cease to be
damaging as a result of a weakening circulation and then A tornado, no matter how one chooses to define it, is a
redevelop, is this redevelopment a new tornado or should we kinematic structure that renews itself from instant to instant
say that the gap is a "skip" in the path of a continuous via one or more dynamic processes. It is not a "thing" in the
tornado? If the answer depends on the distance and/or time sensethat a table or a book (neglectingatomic or molecular
between damage, is there a nonarbitrary way to establish fluctuations) is the same from one moment to the next. Much
criteria for making such classifications? confusion about tornadoes comes from thinking of tornadoes
There are numerous movies and videos showing quite as objects rather than as the kinematic manifestation of
clearly the dissipation of one damagingfunnel cloud/tornado dynamic processes. The actual physical processes are not
with the nearly simultaneous development of another within heedful of our somewhat arbitrary classification schemes
close proximity. A ground survey of the track would prob- and, as scientists, we need constantly to remind ourselves
ably reveal a continuousdamage swath, perhaps with a small that our understanding of tornadoes and tornadic storms can
offset. Are these subvortices within a larger, more or less be clouded by an inappropriate classification scheme (see the
continuous tornado, or are we seeing two different torna- discussion by Doswell [ 199 l
 DOSWELL AND BURGESS 171

The only scientificjustification for a classificationscheme Burgess,D. W., and E. B. Curran, The relationshipof stormtype to
is if that scheme proves to be useful in developingour environment in Oklahoma on 26 April 1984, in Preprints, 14th
understandingand/or in application of that understanding. Conference on Severe Local Storms, pp. 208-211, American
Meteorological Society, Boston, Mass., 1985.
While we probablyhave muddiedthe waters by mentioning Burgess,D. W., and R. P. Davies-Jones, Unusual tornadic stormsin
additional difficulties with event classification, we believe eastern Oklahoma on 5 December 1975, Mort. Weather Rev., 107,
that an appreciationfor classificationproblems is needed in 451-457, 1979.
any proper use of the data derived from classification. Burgess, D. W., and R. J. Donaldson, Jr., Contrastingtornadic
The more we learn about tornadoes and tornadic storms, storm types, in Preprints, 11th Conference on Severe Local
Stortns, pp. 189-!92, American Meteorological Society, Boston,
the more they seem to be terribly complicatedprocesses.It Mass., 1979.
is possiblethat some insight we have yet to find will simplify Burgess,D. W., and L. R. Lemon, Severe thunderstormdetection
our understandingof tornadoes and tornadic storms. On the by radar, in Radar in Meteorology, edited by D. Atlas, pp.
other hand, new observationsmay not result in some simple 619-647, American Meteorological Society, Boston, Mass., 1990.
reconciliation but may raise new and even more confusing Burgess, D. W., V. T. Wood, and R. A. Brown, Mesocyclone
evolution statistics, in Preprints, 12th Conference on Severe
issueswith which to deal. There is nothing that guarantees
Local Storms, pp. 422-424, American Meteorological Society,
simplicity in nature. Boston, Mass., 1982.
Despite the confusion it has caused, however, our new Carbone, R. E., A severe frontal rainband, I, Stormwide hydrody-
understanding developed since the last symposium (as a namic structure, J. Atmos. Sci., 39, 258-279, 1982.
result of radar, storm chasing, and numerical and laboratory Carbone, R. E., A severe frontal rainband, II, Tornado parent
vortex circulation, J. Attnos. Sci., 40, 2639-2654, 1983.
modeling) has been applicable in both a research and an
Church, C. R., J. T. Snow, G. L. Baker, and E. M. Agee,
operational sense.The recognitionof a range of processesat Characteristics of tornado-like vortices as a function of swirl
the scale of the convective storm and at the tornado scale ratio: A laboratoryinvestigation,d. Atmos. Sci., 36, 1755-1766,
has been valuable to our scienceand to society as a whole. 1979.

It is likely that numerical cloud models soon will be able to Cooley, J. R., Cold air funnel clouds, Mort. Weather Rev., 106,
1368-1372, 1978.
resolve tornadic flows, offering the chance for new insights
Davies, J. M., and R. H. Johns, Some wind and instability param-
into tornadoes. As new operational and research observing eters associatedwith strong and violent tornadoes, 1, wind shear
systemsare implemented,it is virtually certain that we shall and helicity, this volume.
come to know much more about nonsupercellevents than at Davies-Jones, R. P., A new look at the vorticity equation with
present. We close by noting that a considerablechallenge applicationto tornadogenesis,
in Preprints,12th Conferenceon
confrontsus in applyingany new understandingof tornadoes Severe Local Storms, pp. 249-252, American Meteorological
Society, Boston, Mass., 1982.
and tornadic storms to benefit society; efforts to do so have Davies-Jones, R. P., Tornado dynamics, in Thunderstorms: A
been painfully slow, up to the present. We hope that Social, Scientific,and TechnologicalDocumentary,vol. 2, Thun-
operationaldeploymentof new technologies will be associ- derstorm Mot•vhologyand Dynamics, 2nd ed., edited by E.
ated with concomitant accelerations in the application of Kessler, pp. 197-236, University of Oklahoma Press,Norman,
1986.
scientific understandingto serve society.
Davies-Jones,R. P., and H. E. Brooks, Mesocyclogenesisfrom a
theoreticalperspective,this volume.
REFERENCES Doswell,C. A., III, Commentson "Photographicdocumentation of
some distinctive cloud forms observed beneath a large cumulon-
Bergen,W. R., Mountainadoes:A significantcontributionto moun- imbus", Bull. Am. Meteorol. Soc., 64, 1389-1390, 1983.
tain windstorm damage?, Weatherwise,29, 64-69, 1976. Doswell, C. A., II!, The operationalmeteorologyof convective
Bluestein, H. B., The formation of a "landspout" in a "broken- weather,vol. !I, Storm-scaleanalysis,NOAA Tech. Metno. ERL
line" squallline in Oklahoma,in Preprints,14th Conferenceon ESG-15. (Available from Natl. Severe Storms Lab., Norman,
Severe Local Storms, pp. 267-270, American Meteorological Okla.)
Society, Boston, Mass., 1985. Doswell, C. A., III, Comments on "Mesoscale convective patterns
Bluestein, H. B., and C. R. Parks, A synopticand photographic of the southernHigh Plains", Bull. Am. Meteorol. Soc., 72,
climatologyof low-precipitationsevere thunderstormsin the 389-390, 1991 a.
southernplains,Mon. WeatherRev., 111, 2034-2046,1983. Doswell,C. A., III, A review for forecasterson the applicationof
Brady,R. H., andE. J. Szoke,The landspout--Acommontypeof hodographs
to forecastingseverethunderstorms,
Natl. Weather
northeast Colorado tornado, in Preprints, 15th Conferenceon Dig., 16, 2-16, 1991b.
Severe Local Storms, pp. 312-315, AmericanMeteorological Doswell,C. A., II[, A. R. Moller, andR. Przybylinski,A unifiedset
Society, Boston, Mass., 1988. of conceptual modelsfor variationson the supercelltheme,in
Browning,K. A., Airflow and precipitationtrajectorieswithin Preprints,16thConferenceon SevereLocalStorms,pp. 40-45,
severe local storms which travel to the fight of the winds, J. AmericanMeteorologicalSociety, Boston,Mass., 1990.
Atrnos. Sci., 21,634--639, 1964. Fankhauser,J. C., G. M. Barnes, L. J. Miller, and P.M. Rost-
Browning,K. A., The structureand mechanismof hailstorms, kowski,Photographic
documentation of somedistinctivecloud
Meteorol. Monogr., 38, 1-39, 1977. formsobservedbeneatha large cumulonimbus,
Bull. Am. Mete-
Browning,
K. A., andR. J. Donaldson,
Jr., Airflowandstructure
of orol. Soc., 64, 45{)-462, 1983a.
a tornadic storm, J. Atmos. Sci., 20, 533-545, 1963. Fankhauser,J. C., G. M. Barnes, and L. J. Miller, Responseto
Burgess, of a severe-fight- commentson "Photographicdocumentation
D. W., andR. A. Brown,The structure of somedistinctive
movingthunderstorm:New singleDopplerradarevidence,in cloudformsobservedbeneatha largecumulonimbus",Bull. Am.
Preprints,8th Conference
on SevereLocal Storms,pp. 40-43, Meteorol. Soc., 64, 1391-1392, 1983b.
AmericanMeteorologicalSociety,Boston,Mass., 1973. Foote, G. B., and H. W. Frank, Case study of a hailstorm
172 CONCEPTUAL MODELS OF TORNADOES AND TORNADIC STORMS

Colorado, IlI, Airflow from triple-Doppler measurements, J. tion and mesocyclone structure as related to tornadogenesis,
Atmos. Sci., 40, 686-707, 1983. Mon. Weather Rev., 107, 1184-1197, 1979.
Forbes, G. S., On the reliability of hook echoes as tornado indica- Marwitz, J. D., The structure and motion of severe hailstorms, II,
tors, Mon. Weather Rev., 109, 1457-1466, 1981. Multicell storms, J. Appl. Meteorol., I1, 180-188, 1972.
Forbes, G. S., and R. M. Wakimoto: A concentrated outbreak of McCaul, E. W., Jr., Buoyancy and shear characteristicsof hurri-
tornadoes, downbursts and microbursts, and implications regard- cane-tornado environments, Mort. Weather Rev., 119, !954-1978,
ing vortex classification,Mon. Weather Rev., 111,220-235, 1983. 1991.
Fujita, T., Detailed analysisof the Fargo tornadoesof June20, 1957, Meaden, G. T., Whirlwind formation at a seabreeze front, Weather,
Weather Bur. Res. Pap. 42, 67 pp., U.S. Dept. of Commerce, 36, 47-48, 1981.
U.S. Govt. Printing Office, Washington, D.C., 1960. Moller, A. R., Commentson "Photographicdocumentationof some
Fujitat,T. T., Proposedmechanismof suctionspotsaccompaniedby distinctive cloud forms observed beneath a large cumulonimbus",
tornadoes,in Preprints, 7th Conference on Severe Local Storms, Bull. Am. Meteorol. Soc., 64, 1390-1391, 1983.
pp. 208-213, American Meteorological Society, Boston, Mass., Moller, A., C. Doswell, J. McGinley, S. Tegtmeier, and R. Zipser,
1971. Field observations of the Union City tornado in Oklahoma,
Ft[iita, T. T., Tornadoesand downburstsin the context of general- Weatherwise, 27, 68--77, 1974.
ized planetary scales,J. Atmos. Sci., 38, 1511-1534, 1981. Moller, A. R., C. A. Doswell Ill, and R. Przybylinski, High-
Gall, R. 1.... A lineal'analysisof the multiple vortex phenomenonin precipitationsupercelIs:
A conceptualmodelanddocumentation,
simulated tornadoes, J. Atmos. Sci., 40, 2010-2024, 1983. in Preprints,16thConferenceon SevereLocalStorms,pp. 52-57,
Garrett, R. A., and V. D. Rockney, Tornadoes in northeastern American MeteorologicalSociety, Boston, Mass., 1990.
Kansas,May 19, 1960,Mon. WeatherRev., I10, 118-!35, 1962. Newton, C. W., Dynamicsof severeconvectivestorms,Meteorol.
Golden,J. H., Waterspoutsand tornadoesover southFlorida, Mon. Monogr., 5, 33-58, 1963.
Weather Re•'., 99, 146-154, 1971. Nolen, R. H., A radarpatternassociatedwith tornadoes,Bull. Am.
Golden,J. H., The life cycle of Florida Keys' waterspouts,J. Appl. Meteorol. Soc., 40, 277-279, 1959.
Meteorol., 13, 676-692, 1974. Rotunno, R., A studyof tornado-likevortex dynamics,J. Atmos.
Golden, J. H., and B. J. Morgan, The NSSL-Notre Dame tornado Sci., 36, 140-155, 1979.
interceptprogram,spring 1972, Bull. Am. Meteorol. Soc., 53, Rotunno, R., Tornadoesand tornadogenesis,in MesoscaleMeteo-
1178-I 180, 1972. rology and Forecasting,edited by P.S. Ray, pp. 414-436,
Gonski, R. F., B. P. Woods, and W. D. Korotky, The Raleigh AmericanMeteorologicalSociety, Boston,Mass., 1986.
tornados-28 November 1988: An operational perspective, in Rotunno,R., and J. B. Klemp, On the rotationand propagationof
Preprints,12thConfi,renceon WeatherAnalysisandForecasting, simulatedsupercellthunderstorms,J. Atmos. Sci., 42, 271-292,
pp. 173-178,AmericanMeteorologicalSociety, Boston,Mass., 1985.
1989. Schlesinger,R. E., A three-dimensional numericalmodel of an
Hane, C. E., and P.S. Ray, Pressureandbuoyancyfieldsderived isolatedthunderstorm:Preliminary results, J. Atmos. Sci., 32,
from Doppler radar data in a tornadicthunderstorm,J. Atmos. 835-850, 1975.
St'i., 42, 18-35, 1985. Smith,R. K., and L. M. Leslie,Tornadogenesis, Q. J. R. Meteorol.
Holle, R. L., and M. W. Maier, Tornadofbrmationfrom downdraft Soc., 104, 189-198, 1978.
interaction in the FACE network, Mon. Weather Rev., !08, Stout,G. E., andF. A. Huff,Radarrecordsan Illinoistornado,Bull.
1010-1028, 1980. Ant. Meteorol. Soc., 34, 281-284, 1953.
Huschke,R. E., Glossaryof Meteorology,p. 505, AmericanMete- Wakimoto, R. M., and J. W. Wilson, Non-supercelltornadoes,
orologicalSociety, Boston,Mass., 1959. Mon. Weather Rev., 117, 1113-1140, 1989.
Idso,S. B., Tornadoor dustdevil:The enigmaof desertwhirlwinds, Ward,N. B., Radarandsurfaceobservations of tornadoes on May
Am. Sci., 62,530-541, 1974. 4, 1961, in Proceedings,9th WeatherRadar Conference,pp.
ldso, S. B., Whirlwinds,densitycurrents,and topographicdistur- 175-180,AmericanMeteorological Society,Boston,Mass., 1961.
bances: A meteorologicalmelange of intriguing interactions, Ward, N. B., The explorationof certainfeaturesof tornadodynam-
Weatherwise, 28, 61-65, 1975. icsusinga laboratory
model,J. Atmos.Sci.,29, 1194-1204,1972.
Imy, D. A., and D. W. Burgess,The structuralevolutionof a Weisman,M. L., andJ. B. Klemp,The structure andclassification
tornadicsupercellwith a persistentmesocyclone,in Preprints, of numericallysimulatedconvectivestormsin directionally
vary-
25thConference on RadarMeteorology,pp. 408-411,American ingwindshears, Mon. WeatherRev.,112,2479-2498, 1984.
MeteorologicalSociety,Boston,Mass., 1991. Weisman,M. L., and J. B. Klemp, Characteristics of isolated
Klemp,J. B., Dynamicsof tornadicthunderstorms, Annu.Rev. convectivestorms,in Mesoscale Meteorology and Forecasting,
Fluid Mech., 19, 369-402, !987. editedby P.S. Ray, pp. 331-358,AmericanMeteorological
Klemp,J. B., and R. B. Wilhelmson, The simulation of three- Society, Boston,Mass., 1986.
dimensionalconvective storm dynamics, J. Atmos. Sci., 35, Wicker,L. J., A numerical simulation
of a tornado-scale
vortexin a
1070-1096, 1978. three-dimensional
cloud model, Ph.D. thesis,264 pp., Univ. of
Lemon,L. R., New severethunderstorm
radaridentification
tech- I11., Urbana, 1990.
niquesandwarning report,NOAATech. Wilczak,J. M., T. W. Christian,
criteria:A preliminary D. E. Wolfe,R. J. Zamora,andB.
Memo. NWS NSSFC-1, 60 pp. (Availableas PB-273049from Stankov,Observations of a Colorado tornado,PartI, Mesoscale
Natl. Tech. Inf. Serv., Springfield, Va.) environment
andtornadogenesis,
Mon. WeatherRev., 120, 497-
Lemon,L. R., andC. A. DoswellIII, Severethunderstorm
evolu- 520,
 Lightning in Tornadic Storms' A Review
DONALD R. MACGORMAN

National Severe Storms Laboratory, NOAA, Norman, Oklahoma 73069

1. INTRODUCTION has been learned from analysis of data from lightning map-
ping systems.
There have been many reports of unusual lightning char-
acteristics in tornadic storms. For example, eyewitnesses 2. LIGHTNING TERMINOLOGY
have reported scorchingbeneath tornado funnels, a steady This sectionexplains lightningterminology and concepts
or rapidly oscillatingglow inside funnels, rapidly recurring that will be used in the rest of the paper. Lightningcan be
small patches of light on the side of the thunderstorm,or classifiedinto two types: (1) Cloud-to-groundlightninghasat
unusuallyhigh or low flashrates [e.g., ChurchandBarnhart, leastone channelspanningfrom the cloud to the ground.(2)
1979; Vaughan and Vonnegut, 1976; Vonnegut and Weyer, Intracloud lightning does not have a channel to ground.
1966]. It is difficult, however, to quantify relationships Becausethe locationsof lightningchannelsrelativeto visible
between lightningand tornadic stormsfrom these eyewit- cloud boundaries are difficult to detect remotely, scientists
ness reports. who studylightningand storm electrificationnormallygroup
Prior to 1975most quantitativemeasurementsof lightning in-cloud, cloud-to-air, and cloud-to-cloud lightning together
in tornadic storms examined the rates and characteristics of
in the intracloud lightning classification. Cloud-to-ground
sferics, the electromagneticnoise radiated by lightning. lightningcan be further classifiedby the polarity of charge
These measurements were difficult to interpret, because thatit effectivelylowersto ground:Positivecloud-to-ground
resultsdependedpartly on the characteristics
of the sferics lightninglowers positive charge;negativecloud-to-ground
receiver, and becausesferics instrumentationprovided at lightninglowers negative charge. Most cloud-to-ground
best only the bearingto the lightningflashgeneratingthe flashesare negative.
sfericsand often provided no locationinformationat all. A cloud-to-ground lightningflashusuallybeginsinsidethe
Furthermore,manyof the sfericsstudieslackedradardata, cloudand is first apparentwhen a faint channel,calledthe
andonlyonehadanyDopplerradardata.Equallyimportant, steppedleader,movesfrom the cloudto the groundinjumps
there was not an adequateappreciationof the variety of roughly50-100 m long. When the steppedleaderconnects
storms that produce tornadoes. with the ground,a brightpulsemovesback up the lightning
Some of these limitations were overcome when systems channel in a processcalleda returnstroke.After a pauseof
weredevelopedto mapthe locationof lightningflashes.For roughly20-150ms,anotherleadercantravelbackdownthe
example,beginning in the late 1970s,variousorganizationsalreadyestablishedlightningchannel,followed by another
begandeploying newlydeveloped systems for automatically return stroke. The combination of a leader and return stroke
mappingwherelightningchannels strikegroundoverranges is called a stroke. All strokesgoingthroughessentiallythe
of a few hundredkilometers[e.g., Krideret al., 1980;Orville samechannelto ground make up a singlecloud-to-ground
et al., 1983;BentandLyons,1984;MacGorman andTaylor, flash. There can be anywhere from one to a few tens of
1989].Now coverageby thesesystemsis continualacross strokes in a flash.
thecontiguous UnitedStates,witherrorsin strikelocations Sfericsare generatedby changesin the vector electric
typicallylessthan10km [e.g.,Machet al., 1986;MacGor- current,thefrequencyof the radiationdepending onthetime
man and Rust, 1989]. scaleof the changes.Since currentsin a lightningchannel
In this paper we first review what was learnedabout increasefaster than they decay, a lightningchannelwill
tornadic stormsfrom sfericsstudiesand then considerwhat generally radiatehigherfrequencies duringaninitialcurrent
surgethanduringits decay.The timerequiredfor a current
The Tornado:Its Structure,Dynamics,Prediction,andHazards. pulseto propagate througha channelsegment depends on
GeophysicalMonograph79
Thispaperisnotsubject to U.S.copyright.
Publishedin 1993bythe the length of the segment, so longer channels radiate more
American GeophysicalUnion. energyat lowerfrequencies,if all elseis equal.
!74 REVIEW OF LIGHTNING IN TORNADIC STORMS

The spectrumof sfericsfrom a typical return strokepeaks BURSTRATEversus FREQUENCY

at roughly 5 kHz. Most strong signalsat frequenciesbelow 4O

about30 kHz are from cloud-to-groundreturn strokesand so :I i : ::::l : :l ' : :'Jill :
occur in discrete bursts, separated by tens of milliseconds.
At frequencies near or above 1 MHz, the amplitudesof
sferics from return strokes, from other processesof a .• 30 Tornadic
Storms
--.-.•

cloud-to-ground flash, and from intracloud flashes are all

comparable. Furthermore, at these higher frequencies, sfer-
©X Severe
Storms
z 20
ics are more numerous and more continual throughout a
lightning flash, and relatively few are associatedwith return
strokes.
'"-' 10
The vertical componentof sfericsmeasuredat the stirface
of the conducting Earth is generally much larger than the H ...... e-.---.......
©....../Thunderstorms
horizontal, even if the two components are comparable at
the source. Any horizontal component that exists is reduced 31.6 kHz 100 kHz 316 kHz 1 MHz 3.16 MHz

by currents that are induced in the ground by the stZrics. Frequency

'l"hcreforea sferics receiver at the ground receives little of

Fig. 1. Average sferics burst rate as a function of receiver fre-
the energy radiated by horizontal components of current quency for various storm classifications[from MacGorman et al.,
change. Airborne receivers are less affected by currents 1989]. Data are from four tornadic storms, three nontornadic severe
induced in the ground and so are better able to detect the storms, and seven nonsevere thunderstorms. Sferics burst rates
horizontal componentsof sferics. were measuredat 10 receiver frequencies between 10 kHz and 137
MHz, indicated by dots. There was little change above 3 MHz, so
higher frequencies are not shown.
3. SFERICS STUDIES

Initial studies of sferics from tornadic storms examined

radio bands f¾om 10 to 500 kHz. (Sferics normally are increasingseverity, until for tornadic storms, the energy is
classified by the frequency of the receiver used to detect independent of frequency.
them.) Dickson and McConahy [1956] found that sferics There were exceptions to all of these findings. In several
rates detected by a 10-kHz receiver often increased as cases, high sferics rates did not occur in tornadic storms
stormsgrew rapidly taller, but ratespeakedearlier in storms [Ward et al., 1965; Stan3'brdet al., 1971; Lind et al., 1972;
that had more violent severe weather. Rates peaked about Taylor, 1973; Greneker et al., 1976; Johnson et al., 1977;
1.5 hour before tornadoes and decreased during tornadoes to MacGorman et al., 1989]. Taylor, who examinedthe largest
about 40% of peak value. At frequenciesof 150 kHz and number of storms, found that roughly 20% of tornadic
higher, sfericsrateswere found to increaseuntil tornadoes storms did not have high sferics rates [MacGorman et al.,
were produced,and rates becameexceptionallyhighduring 1989].It also was possiblefor high sfericsrates to occur in
periodswhen tornadoesoccurred[jones, 1951, I958; Jones stormswithout tornadoes [Ward et al., 1965; StanJbrdet al.,
and Hess, 1952; Kohi, 1962; Kohl and Miller, 1963]. Kohl 1971;MacGorman et al., 1989]. Taylor found that although
only 23% of nontornadicseverestormsand 1% of nonsevere
and Miller [1963] observed that sferics rates at 150 kHz
usuallypeakedduringsevereweatherand begandecreasing storms had sferics rates comparableto those of tornadic
prior to the end of severeweather.As might be expected storms, 70% of the storms with high sferics rates were not
from the above observations, Jones and Hess [1952] found tornadic, because the number of nonsevere storms was
muchlarger than the numberof tornadicstorms[MacGor-
that the ratio of the numbers of higher-frequencyto lower-
frequencysfericsincreasedfor sfericsof large amplitude man et al., 1989]. Sometornadic stormswithout high sferics
when storms were more severe; the ratio was 1:20 before rateshad only weak tornadoes[e.g., Johnsonet al., !977],
and after tornadoes and in nonsevere thunderstorms versus and the combined effect of several simultaneous storms
1:1 during tornadoes. explainssomenonseverestormswith highsfericsrates[e.g.,
Later studiesat frequencybandsup to 150 MHz [Silberg, MacGorman et al., 1989]. However, these situationsdid not
1965; Taylor, 1973;Stanford et al., 1971;Johnsonet al., accountfor all exceptionsto the rule that 3-MHz sferics
1977]found that the increasein sfericsrates during severe ratesare muchhigherin tornadicstormsthan in nonsevere
weatherwas greatestat frequenciesabove 1 MHz. This is storms.

shownin Figure 1, which presentsa summaryof Taylor's The mostobviousinferencefrom the increasingnumberof
dataconcerningthe dependence of sfericsrateson receiver sferics at high frequencies and the decreasingor constant
frequency.Silberg[1965]and Taylor[1973]alsofoundthat number at low frequencieswas that intracloudflash rates
theenergyradiatedassferics(notshown)increases at higher usually increase while cloud-to-groundlightning return
frequenciesas storm severity increases;radiated energy strokes decrease or remain fairly constant during the tor-
peaksat roughly5 kHz for nonsevere storms[Taylor,1972], nadicstageof storms.This interpretationwas further sup-
buttheenergyat higherfrequencies graduallyincreases with portedby the theoretical
work of Stanford[1971]and
 MAcGORMAN 175

airborneobservations of Shanmugatn and Pybus[1971]. originatedalongazimuthsto low-levelreflectivitycoresand

Both reported that horizontalpolarizationof sfericsin- that sfericsrateswere largerfrom coreswith higherreflec-
creasedand verticalpolarizationdecreased,indicatinga tivity. Furthermore, Brown and Hughes [1978]found that
greaterpreponderanceof horizontallightningchannels,as the pattern in the number of VLF sferics(which are pro-
the sfericsrate and stormseverityincreased. ducedpredominantlyby return strokes)from the Union City
Other observationsalso supportedthis interpretation. tornadicstormwassimilarbothin timeandspaceto patterns
Jones[ 1958]reportedwatchinga stormwhenit wasproduc- in reflectivity at low levels of the storm.
ing high sfericsrates about 1 hour before a tornado;no Although the sferics studies clearly demonstratedthat
groundflashescould be seen, but relatively dim circular lightningtended to evolve in characteristicpatternsin tor-
patches,about0.5 km in diameter,wereilluminatedin rapid nadic storms, there were a number of uncertainties. (1)
successionon the sideof the storm.He suggested
that these There were uncertaintiesin extracting lightningflash type
were an unusualform of intracloudlightning.Electric field and rates from sferics data. (2) No sferics study examined
recordsfor thissamestormindicatedthat the lightningflash the evolution of mesocyclones or considered differencesin
ratewas 12 min-1 duringthe tornado[Gunn,1956].Von- the types of storms that produced tornadoes in order to
negut and Moore [1957] inferred from recordingsof the examinewhy sfericsrates evolved differently. (3) No sferics
closest electric field sensor to the tornado that there were no study examined the three-dimensional vector wind field for
cloud-to-ground
flashesnear the tornado. Unusuallyhigh tornadic storms.
lightningflash rates were also observedin two other tornadic
storm systemsby Orville and Vonnegut [1977] and Turrnan
4. LIGHTNING MAPPING STUDIES
and Tettelbach[1980], who usedsatellite-borneopticalde-
tectors. Films of 17 tornadoes and one funnel cloud docu- Studies using lightning mapping systems have begun to
mented low ground flash rates near tornadoes; Davies-Jones address some of the shortcomingsof the sferics studies.
and Golden [1975a] (see also Davies-Jones and Golden However, only two of the mapping studies thus far have
[1975b, c], Colgate [1975], and Vonnegut [1975] for a includedintracloudlightning; all others have examinedonly
discussion)found cloud-to-groundlightningflashesnear only cloud-to-groundlightning. Furthermore, mapping studies
two of the 18 filmed vortices. One of these two was a tornado have found considerablevariability in the relationship of
with 12 ground flashes;the other was a funnel cloudwith two tornadoes to cloud-to-ground flash rates. In two tornadic
ground flashes. storms, ground flash rates were small before and during
From studies of sferics, however, it appears that the tornadoes [Orville et al., 1982; MacGorman et al., 1989]. In
tornado itself is not likely to be the cause of unusual sferics three, ground flash rates peaked during tornadoes [Kane,
rates. For example, Jones [1965] noted that sfericsrates can 1991;MacGorman and Nielsen, 1991;Keighton et al., 1991].
be unusuallyhigh 60-90 min before tornadoes,althoughthey In two, tornadoes occurred after ground flash rates had
may increase further when tornadoes occur. Scouten et al. increased but when ground flash rates were not at their
[1972] further analyzed the data of Jones [1958] at 150 kHz maximum [Kane, 1991; MacGorman et al., 1985].
and concluded that touchdown of the tornado did not affect Results from two of the studies suggestsome reasonsfbr
the high sferics rate. Although larger-amplitudesfericsclus- this variability. In the Binger tornadic storm studied by
tered more closely about a central core near the azimuth of MacGorman et al. [1989], negative ground flash rates within
the tornado, sfericsas a whole were spreadamongazimuths 10km of the mesocyclone
centerwere lessthan1 min-•
spanning most of the storm. A similar lack of correlation until the tornado began dissipating. Negative cloud-to-
with the time and location of tornado touchdown was groundflash rates reached a relative maximum after the last
reported for sferics in three bands between 300 kHz and 3 tornado, as its mesocyclone core dissipated, and reached an
MHz [Taylor, 1973] and for sferics at 1-50 kHz [Brown and absolute
maximum(about4 min-•) as the lastmesocyclone
Hughes, 1978]. Brown and Hughes [1978] also noted a lack core in the storm dissipated (see Figure 2). Intracloud flash
of obvious relationships with the tornadic vortex signature rates reached an absolute maximum of approximately 14
[Brown et al., 1978]. In a study of sferics burst rates at high min-• duringthemostviolenttornadoandwerewellcorre-
frequencies (30-300 MHz) in two tornadic storms,Johnson lated with low-level cyclonic shear when the shear was
[1980]found that the tornadoesbegannear the time of peaks greaterthan1 x 10-2 s- • duringthe tornado.(Cyclonic
in the sferics burst rate but were not coincident with anyshear is one half vertical vorticity for ideal solid-body
peaks. rotation. In calculating this shear, tornadic winds were
Several studiesconsideredchangesin tornadicstormsthat ignored;shear was measuredacrossthe diameterof maxi-
were correlated with the high sferics rates. For example, mumtangentialwind speedfor the mesocyclone,typically a
severalinvestigatorsfound periodicitiesof 10-60 min in thedistance of a few kilometers.) After the last tornado, when
sferics data and suggestedthat these resulted from the therewas a largedecreasein low-levelcyclonicshearbut the
cyclical intensificationand decay of thunderstormcells mesocyclonewas still strong at middle levels, intracloud
[Kohl and Miller, 1963; Stanford et al., 1971; Shantnugam flashrateswereonly0-2 min-• .
and Pybus, 197!; Lind et al., 1972; Trost and Nornikos, In the Edmond storm studiedby MacGorman and Nielsen
1975].Taylor [1973]reportedthat sfericsat highfrequencies [1991] there were no intracloud data, but negative
176 REVIEW OF LIGHTNING IN TORNADIC STORMS

12 5 KM
30 _

20

1.5 KM
30
1.5 KM

o 20
,.•i':::

=: 10 r,;':

<• 40

'i I ! I i
3o
1.2

0.8

0.4

12

0.0 i i I i

z 10
_ 10 KM

z
S• o

z
2 I i i i .... i"
1803 1815 1830 1845 1900 1920
TIME (CST)
0, , • , , • • • • , ,

Fig. 3. Time series plots of cyclonic shear at the 1.5- and 5-km
levels and of ground flash rates within 10 km of the mesocyclone
center in the Edmond storm of May 8, 1986 [from MacGorman and
z
Nielsen, 1991]. The bars on the bottom indicate when tornadoes
occurred.
o

1833 1900 1930 20'00 ' 20'30 ' 20•54

TIM E (CST)
TORNADO
[ ' ..' • .'..':'iI [] flash rates within 10 km of the mesocyclonecenter evolved
differently than in the Binger storm. Rates increased to a
Fig. 2. Time series plots of cyclonic shear at the 1.5- and 6-km
levels and of ground and intracloud flash rates within 10 km of the peakof 11min-• during
tornadic
activityandappeared
to be
mesocyclone core in the Binger storm of May 22, 1981 [from correlated with cyclonic shear at the 5 km level (Figure 3).
MacGorman et al., 1989]. The bars on the bottom indicate when Ground strike locationstended to cluster in the vicinity of a
tornadoes occurred.
reflectivity core north of the mesocyclone during
 MAcGORMAN 177

activity and to be more scattered before and after the

(a)
tornadic stageof the storm (Figure 4). Keighton et al. [1991] 70
observed a similar increase in scattering of lightning strikes
after tornadic activity on another day.
MacGorman and Nielsen [1991] suggestedthat observed
• 6o
differences between the Binget and Edmond storms might
account for the differences in the evolution of their lightning
flash rates. The Binget storm was a classic supercell storm,
• so
with a prominentweak-echoregion, a strong,deep updraft
core, and a long-lasting mesocyclonehaving a family of 40
mesocyclonecores [Burgesset al., 1982]. The Edmond
storm also was a supercell storm but was weaker than the
Binger storm by almost any measure:the duration of its 30

supercellstagewas only 30% that of the Binget storm;its

mesocyclonewas weaker, shallower,and shorterlived; and
features such as the weak-echo vault and tornadic vortex
signaturewere not as pronounced.In fact, the Edmond
storm did not form a tornado until the outflow boundary of (b)
anotherstormbeganto overtakeit from the west. MacGor-
man and Nielsen [1991] suggested
that the increasein flash
rates in both storms when the mesocyclone was strongest
probablywasa resultof thestrongupdraftthatexistedthen.
They alsosuggested,
however,that the deeperand stronger
updraftcore and weak echoregionof the Bingetstorm
probablydelayed
groundflashactivityandcaused
intracloud
activity to dominate.
To understandthe hypothesisthat MacGorman and
Nielsen [1991] offeredto explainwhy groundflash rates 35
behaveddifferentlyin thesetwo storms,it is necessary
first
to understanda little aboutthunderstormelectrification.One -15 -10 -5 0 5 10
important
classof thunderstorm
electrification
mechanisms EAST (kin)
relieson microphysical
particleinteractionsto placecharge
of one signon graupelandchargeof the opposite signon 30 20
cloudice. Suchmechanisms are affected(1) by factorsthat (c)
7O
control the sizesand numbersof graupeland ice particles
that collide, and (2) by factorsthat controlsubsequent ' •(... 40,,.-50--.I. /
transport
of thecharged
particles
intodifferent
regions
of a ,.-, 60
E
thunderstorm.Factors affecting the number of relevant
collisions
includethe strength
anddepthof theupdraftand
I-.. 50
the microphysics
of the storm.Subsequent
transport
is
dominated
initiallyby thesedimentationof graupel butmay o

becomedominated by the relativevelocityof the windsin 4O

,,..F"20

different
regions asthegraupel
andicecrystals movefarther .. '•4.0
apart.Thunderstorm measurements [e.g.,KrehbieI,1981;
Byrnne et al., 1983;Chauzy
et aI., !985;Dyeet al., I986; 30
,I i 1__
Koshakand Krider, 1989]suggest
that mostof the main
I ! I .......I ! I I
-20 -lO o lO 20
negative
charge
at middle levels
ofthestormisongraupel EAST (kin)
andprecipitation,
whilethemainpositive
chargeisgenerally
onicecrystals
atcolder temperatures
(thenextsection
will
discussonepossibleexception). Fig.4. Lightning
ground strikes
superimposed
onradarreflectiv-
Severalinvestigators
haveobserved
thata strengthening ityatthe3-kmleveloftheMay8, 1986,
Edmondstorm
duringthree
updraft
ataltitudes
colderthan-20øCincreasesflashrates periods:
(a) 1802-1806
CST,(b) 1814-1817
CST,
and(c) 1834-1838
(seediscussion
byMacGorman ThisprobablyCST
etal.[!989]). [fromMacGorman
onlyduring
andNielsen,199!].Therewasa tornado
theperiodin Figure
4b. Minuses indicate
thestrike
explains
theobserved
increases
in flashratesduring
the pointofflashes
thateffectively
lowerednegativecharge
toground;
tornadic
stage
oftheBinget
andEdmond storms.MacGor-pluses, flashes
thateffectively
loweredpositivecharge.
Thelarge
manet al. [1989]alsoobserved areasof large dotmarksthecenterof themesocyclone
increasing core.Radarreflectivity
is
reflectivity
atheights
between beforelabeledin dBZ.NotethatFigure4b hasa differentdistance
6kmand8kmshortly
178 REVIEW OF LIGHTNING IN TORNADIC STORMS

and duringperiodsof high flashrates.They suggested that TOTALCHARGEDENSITY (NC/M**3)

the increasedflashratesprobablywere causedby increased 2{• , i i ' i i i i' i i i i ....
particleinteractionsleadingto reflectivitygrowthat 7-9 km
in and near the strong updraft. Similar correlationsof
lightningrateswith the horizontalareaof largereflectivityat
midlevelsof stormshave been observed,for example,by _

Larson and Stansbury [1974], Lhermitte and Krehbiel

[1979], and Keighton et al. [1991].
One of the primary problemsposed by the Binger and
Edmondstormsis explainingwhy groundflashrateswere so
small near and during the time when total and intracloud
flash rates were large in the BingeTstorm and why this
apparently was not true for the Edmond storm. MacGorman
et al. [1989] suggestedthat the strong,deep updraftand the
resultingpronouncedweak-echoregionof the Bingerstorm
enhancedproductionof intracloud flashesand delayed or 2
suppressedproduction of ground flashes because of effects
0.
on the thunderstorm charge distribution. 1 0. 2{3. 30. 4•. 50. 60.
For example, one primary effect on the chargedistribution
is to keep negative charge higher than in most storms. This HORIZONTAL DISTANCE
would increasethe energyrequiredfor lightningto spanthe
Fig. 5. Total space charge density in a southwest-northeastverti-
distance to ground and would decrease the electric field at cal cross section through the main updraft, from a numerical
most heightsbelow the negative charge. Negative charge is simulation of the Binger storm of May 22, 1981. Solid contours
higherfor three reasons.(1) The strongupdraft core rapidly indicate zero or positive net spacechargedensity; dashedcontours,
lifts all but the largestparticles to upper levels of the storm. negativenet chargedensity.The contourintervalis 0.5 nC m-3.
Stippling
delineatesthe regionof updraftexceeding5 m s-1. The
(2) Large updraft speedsdo not allow most hydrometeorsto thin horizontal curve near a height of 4 km indicates the melting
remain long in a given layer. The short residence time level. The bold curve outlines the cloud and precipitation boundary.
prevents hydrometeors from acquiring much charge and
prevents positive and negative particles from moving far
apart after becoming charged. Therefore there is little net negative charges move farther apart (thereby increasing the
negative or positive charge in a given layer of a strong amount of net negative charge that can be separated before
updraft. (3) Temperatures in stronger updraft cores are causing intracloud lightning).
generally warmer, and this causesthe ice processesthought Since MacGorman et al. [ 1989] first suggestedthis hypoth-
to be responsiblefor charging (see discussionof II!ingworth esis, there has been support from two sources. First, the
[1985]) to occur somewhat higher. lowest charge detected by the only published balloon-borne
A secondimportant effect of a strong, deep updraft is that electric field soundingthrough a mesocycloneupdraft was a
regions of net positive and negative charge would be rela- region of negative charge starting at 9.5 km above mean sea
tively close together near the top of the storm updraft. This level, where the temperature was approximately -37øC in
occurs for two reasons. (1) As noted above, sedimentation the storm's environment and -31øC in the updraft [Rust et
and differing relative velocities of positively and negatively al., 1990]. This lower boundary is 3-5 km higher, and at
charged particles would not have time to separate the environmental temperatures 20ø-30øCcolder, than the lower
charges very far in a fast updraft. (2) The vertical shear of boundary normally reported for the main negative charge in
horizontal winds in a classic supercell storm causesgraupel continental thunderstorms. Second, Ziegler and MacGor-
to fall to the side of the updraft core, so much of the man [1990] used a kinematic retrieval that included electri-
storm-scaleseparationof charge occurs outside the updraft. ficationprocesses[Ziegler et al., 1991]to model the electric
When two charged regions are closer together, less net field and charge structure of the Binger storm. In agreement
charge is needed in each region to create electric field with the balloon-borne measurements, the lower boundary
magnitudessufficientto initiate lightningbetween them. This of negative charge inside the updraft core in the retrieved
enhances intracloud flash rates. It also reduces the negative charge distribution was elevated to a higher altitude than
charge available for cloud-to-ground lightning near the up- normal (Figure 5).
MacGorman and Nielsen [1991] suggestedthat the charge
draft, because the threshold for initiating lightning regulates
the net negative(and positive) chargethat can be separated. distribution near the updraft of the Edmond storm was more
This, combined with the unusually large height of negative like that of an ordinary thunderstorm and that cloud-to-
charge discussedabove, makes cloud-to-groundlightning groundflashesthereforewere able to occur frequently when
much less likely to occur near a very strong, deep updraft the mesocyclonewas strong.They concludedthat negative
core. Ground flash rates increase later, as the negative charge was lower, a larger fraction of oppositelycharged
charge moves closer to ground and as the positive and particleshad time to separate,and positive and
 MAcGORMAN 179

charge were farther apart because(1) the updraft core was

weaker, shallower, and shorter lived, (2) the resultingweak- 30
echo region was shallower and contained larger reflectivi-
ties, and (3) the 55-dBZ reflectivity cores extended to a w

height of only 5-6 km in the Edmond storm versus over 10 • 20

km in the Binger storm. Similarly, Keighton et al. [1991]
observed another storm in which cloud-to-ground flash rates
did not decreaseduring the tornadic stageand noted that it
also had a weaker mesocyclone than the Binger storm.

FO F1 F4 F1 FO
5. POSITIVE CLOUD-TO-GROUND Ill I I I I
LIGHTNING IN TORNADIC STORMS

Data on the less common flashes that effectively lower

positivechargeto ground(+CG flashes)havebeenstudied 30

in detail for only one tornadic storm, the Edmond storm

investigatedby MacGorman and Nielsen [1991]. (Reliable
positive cloud-to-grounddata were not availablefor the % 20
Bingerstormor for othertornadicstormsbefore1983.)In
the Edmond storm, positive cloud-to-groundflashesbegan
to occur a few minutes before the tornado touched down,
andpositivegroundflashratesincreased to theirmaximum
value during the tornado, near the time when negative 15()0 1600 1700 1800 1900 2000 2100
groundflashratesalsopeaked(seeFigure3). MacGorman
TIME (CS'I)
and Nielsen [1991] noted from watchinggroundflash dis-
playsfor severaltornadicstormsin realtimethatpositive
groundflashesoftenoccurredduringtornadoes. Sometimes Fig.6. Timeseries plotsof thenumberof positive
andnegative
cloud-to-ground
flashesin each15-min
intervalduringa stormthat
a fewpositivegroundflasheswereclusteredonthesouthern produced a violenttornado nearHutchinson,
Kansas, onMarch26,
flank of stormsin which negativecloud-to-ground
lightning 1991. Bars indicatethe time of tornadoes,labeledwith F scale
still dominatedgroundflashactivity (as in the Edmond ratings. Theapproximate timeoflargehailreportsareindicated
by
storm),and sometimes
mostof the groundflashesin the the leadingedgeof Hs.
stormwere positivegroundflashes.
Althoughmanysevereor tornadicstormsdo not have
largenumbers of positivegroundflashes,
preliminaryevi-
dencesuggests that whenrelativelyhighflashratesand tornadonearHutchinson,Kansas,on March26, 1991.This
densities
of positivecloud-to-ground occur,they storm
lightning producedsevereweatherthroughoutmuchofitslife.
are associatedwith severeweather. Reap and MacGorman In the secondmode the dominant polarity of ground
[1989]reportedin a climatological
studythattherewas a flashesin the stormswitchesfrompositiveto negative,with
betweena highdensityof positivegroundflashes thepeaknegative
correlation groundflashratescomparable to orlarger
andsevereweather.Furthermore,in August1990andspring thantheearlierpeakpositiveground flashrates.An example
1991,MacGorman and Burgess [1991]observed several ofthesecond modeisgivenin Figure7, whichshows ground
stormsthat wereunusualbecause(1) mostcloud-to-ground flashrates for the storm that produced an F5 tornadoin
flasheswerepositive groundflashes and(2)ground strike Plainfield,
Illinois,on August 28, 1990. Prior to the F5
pointsof positive cloud-to-groundlightning occurred in tornadomostgroundflashesin the stormwerepositive.
denseclusters,muchlike the clustersof negativeground Duringthisperiod,largehailandfourtornadoes
ratedupto
flashesobserved
in mostelectrically
activestorms.(Positive F2were produced.
Threeofthesetornadoesoccurredduring
ground flashes
aregenerallyinfrequent anddiffuse.)
Allof theperiodof maximumpositive
ground flash
rates.
thestorms withfrequentpositiveground produced The dominantpolarity of cloud-to-ground
flashes lightning
largehail,andmanyproduced tornadoes. observa-switchedas the Plainfieldtornadobegan.Most subsequent
Similar
tionswerereportedby Rustet al. [1985],CurranandRust groundflashes werenegative, asshown clearly
in thetime
[1992],andBranickandDoswell[1992]. series
plotsin Figure 7. Also notethattherewas a decrease
These stormswith unusualpositivelightningactivity inoverallground flashactivityshortly
beforeandduring the
appear
to occurin at leasttwomodes.In one,positiveF5 tornado,muchas described earlierfor negative
ground
flashesin sometornadicstorrns.Negativegroundflashrates
ground
flashes
dominate cloud-to-ground
activity
throughout increased to theirlargestvaluesat theendof thetimeseries
the lifetimeof the storm,including
all tornadicactivity.An
example
ofa storminthismode isshown 6,which plot,whenthestorm
inFigure wasmerging withotherstorms
toform
presents
groundflashrates
forthestorm anF4 a squallline (similarto lightning
thatproduced behaviorobserved
180 REVIEW OF LIGHTNING IN TORNADIC STORMS

400
high-precipitationsupercell was related to the transition in
350

• 200
•,1
?
.....
,i,i i groundflashpolarity. A similartransitionwas observedby
Curran and Rust [1992].
MacGorman and Nielsen [1991] suggestedtwo reasons
why positive ground flashes might occur in severe storms.
:5 •5o First, the structureof supercellstorms,in whichprecipita-
tion at middleto low levelsis displacedhorizontallyfrom the
o, •oo
updraft core, might lead to a tilted bipolar chargedistribu-
5O
tion, with a lower negativechargeoccurringin the vicinity of
precipitationandpositivechargeoccurringon ice crystalsat
the top of the updraftcore, abovethe weak echoregion.The
F1 FO-2 F5 tilt might be sufficientfor the upper positive chargeto be
I IIIII
effectively exposed to the ground, as for winter storms
H HH HHH
400
observedby Brook et al. [1982]. Branick and Doswell [1992]
have suggesteda variation of this: not only does the hori-
350 zontal displacementof the updraft and precipitationtilt the
/
dipole, but the lack of low-to-middlelevel precipitationin
m 200 low-precipitationstormsinhibitsformation of a large region
of negative charge.
5 •5o The second mechanism discussedby MacGorman and
Nielsen [1991] involved a possible region of significant
+ lOO
positive charge beneath the main negative charge of the
50 thunderstorm. Such a region could be formed in different
ways. For example, several investigators have suggested
o that a lower positive charge could be formed in thunder-
1200 1300 1400 1500 1600
storms by a noninductive mechanism studied in the labora-
TIME (CST) tory by Takahashi [1978], Jayaratne et al. [1983], and
others. In most regions in which this mechanism for micro-
Fig. 7. Same as Figure 6 but for the Plainfieldstormon August28, physical charge separation might occur in thunderstorms,
1990. l,arge hail reportswhose H symbolswould overlap earlier Hs collisions between graupel and ice crystals place negative
are omitted.
charge on graupel and positive charge on ice crystals. These
collisionsresult in the normal tendencyfor the main positive
charge to be generally above the main negative charge. At
Goodman and MacGorman [1986] when storms merged to temperatures warmer than some threshold between -10 ø
form a mesoscaleconvective complex). and -20øC, however, this is reversed: positive charge is
Visual observations indicate that the character of the placed on graupel; negative charge, on ice crystals. If
Plainfield storm changedat approximately the time when the enough collisions occur between 0øC and this reversal tem-
last and most violent tornado beganand groundflashpolarity perature, they could create a region of positive charge on
switched. Initially, the storm grew rapidly into a supercell, graupel beneath the main negative charge of thunderstorms.
and it remained a supercell for much of its life. During this
period the storm had a prominent, deep weak-echo region 6. CONCLUSION
and produced large hail. At about the time that the Plainfield
tornado began, heavy precipitation began to wrap around the Obviously, the studiesreviewed in this paper fall far short
mesocyclone, filling in the weak-echo region. Observers of of fully characterizingthe diversity of lightning evolution in
the Plainfield storm noted that the tornado was obscured by tornadic storms. Most studies examined lightning behavior
heavy precipitation. relative just to the existenceof tornadoes. Only a few studies
Curran and Rust [1992] and Branick and Doswell [1992] examinedlightningevolution relative to the evolution of the
have observed that low-precipitation supercell storms often reflectivity structure, mesocyclone,or wind field of tornadic
produce high densities of positive ground flashesand that storms. To test adequately the hypothesized relationship
nearby supercell storms with heavy precipitation produce between updraft characteristicsand suppressionof ground
the normal preponderance of negative ground flashes. At flashesrequiresa more systematicstudy of large data setsof
least some severe storms (including the earlier stagesof the lightning and storm characteristics,possibly supplemented
Plainfield storm) that have high numbers and densities of by numerical simulations.Such studiesalso may be able to
positive groundflashesare not low-precipitationstorms,but determine better the underlying relationship that caused the
there are no reportsof this positivecloud-to-groundlightning observed close correlation of low-level cyclonic shear and
signaturein high-precipitationsupercellstorms.Therefore it intracloud flash rates in the Binger storm.
appears that the transition of the Plainfield storm to a If the hypothesesdiscussedin this paper are correct,
 MAcGORMAN 181

expect that more comprehensive studies of tornadic storms relationship between supercell structure and lightning ground
will findthefollowing.Almostall stormsthathavemesocy- strike polarity, Weather Forecasting, 7, 143-149, 1992.
clones will have an increase in intracloud and total flash rates Brook, M., M. Nakano, P. Krehbiel, and T. Takeuti, The electrical
structure of the Hokuriko winter thunderstorms, J. Geophys.
as the updraft increasesin depth, horizontalextent, and Res., 87, 1207-1215, 1982.
speedat altitudesabovethe -20øC level,therebyincreasing Brown, R. A., and H. G. Hughes, Directional VLF sfericsfrom the
the numberof hydrometeorcollisionsthat generatecharge. Union City, Oklahoma, tornadic storm, J. Geophys. Res., 83,
Intracloud flash rates will be enhanced even more in classic 3571-3574, 1978.
,supercellstorms,which tendto have very largeand strong Brown, R. A., L. R. Lemon, and D. W. Burgess,Tornado detection
by pulsed Doppler radar, Mon. Weather Rev., 106, 29-38, 1978.
updrafts that are quasi-steady,in contrastto updraftsin Burgess, D. W., V. T. Wood, and R. A. Brown, Mesocyclone
lesser thunderstorms.In some tornadic storms, cloud-to- evolution statistics, in Preprints, 12th Conference on Severe
groundflasheswill begin and peak shortly after intracloud Local Storms, pp. 422-424, American MeteorologicalSociety,
lightning (roughly 10-15 rain later), as in most nonsevere Boston, Mass., 1982.
storms. However, in storms that have updraftslarge and Byrnne, G. J., A. A. Few, and M. E. Weber, Altitude, thickness,
and charge concentrationsof charged regions of four thunder-
strongenoughto create prominent, deep weak-echoregions
stormsduringTRIP 1981 basedupon in situ balloonelectricfield
(with high lower boundariesof negativecharge)we expect measurements,Geophys. Res. Lett., I0, 39-42, 1983.
cloud-to-groundlightningnear the updraft and mesocyclone Chauzy, S., M. Chong, A. Delannoy, and S. Despiau,The 22 June
to be suppressed,so that few ground flashesoccur when the tropical squallline observedduringCOPT 81 experiment:Elec-
updraft is strong and ground flash rates peak when the trical signatureassociatedwith dynamicalstructureand precipi-
updraft finally weakens. tation, J. Geophys.Res., 90, 6091-6098, 1985.
Church, C. R., and B. J. Barnhart, A review of electrical phenom-
There are far fewer data on positive cloud-to-ground ena associatedwith tornadoes,in Preprints, 1Ith Conferenceon
lightning in severe storms, so observed relationships and Severe Local Storms, pp. 337-342, American Meteorological
suggestedhypothesesshould be regarded as more tentative. Society, Boston, Mass., 1979.
If a tilted chargedistributioncausespositivecloud-to-ground Colgate,S. A., Commenton "On therelationof electricalactivityto
tornadoes," by R. P. Davies-Jonesand J. H. Golden, J. Geophys.
lightning, then we would expect to see positive ground
Res., 80, 4556, 1975.
flashes in storms in which the top of the main updraft is Curran, E. B., and W. D. Rust, Positivegroundflashesproducedby
displaced sufficientlyfar horizontally from the reflectivity low-precipitationthunderstormsin Oklahomaon 26 April 1984,
core. If the described mechanism for the lower positive Mon. Weather Rev., I20, 544-553, 1992.
charge causespositive cloud-to-ground(+CG) flashes,then Davies-Jones,R. P., and J. H. Golden, On the relation of electrical
we would expect to see + CG lightningin stormsin which the activity to tornadoes,J. Geophys.Res., 80, 1614-1616,1975a.
Davies-Jones,R. P., and J. H. Golden, Reply, J. Geophys.Res., 80,
number of collisions between graupel and ice crystals is 4557--4558, 1975b.
enhanced at temperatures between freezing and roughly Davies-Jones,R. P., and J. H. Golden,Reply, J. Geophys.Res., 80,
-15øC. Such a situation could occur, for example, if there is 4561--4562, 1975c.
significantrecirculationof graupel,as hasbeenreportedfor Dickson, E. B., and R. J. McConahy, Sferics readingson wind-
some hail storms. storms and tornadoes, Bull. Am. Meteorol. Soc., 37, 410--412,
1956.
It may be difficult to determinewhich of theseor other Dye, J. E., J. J. Jones,W. P. Winn, T. A. Cerni, B. Gardiner,D.
mechanismsare responsiblefor positive cloud-to-ground Lamb, R. L. Pitter, J. Hallet, and C. P. R. Saunders,Early
lightning.For example, the large separationreportedbe- electrificationand precipitationdevelopmentin a small isolated
tweenthe mesocyclone andprecipitationfor somestormsin Montanacumulonimbus,J. Geophys.Res., 9I, 1231-1247,1986.
which positive ground flashesdominate [Branick and Goodman,S. J., and D. R. MacGorman, Cloud-to-groundlightning
activityin mesoscaleconvectivecomplexes,Mon. WeatherRev.,
Doswell, 1992] could result from either scenario. There 114, 2320-2328, 1986.
probablywouldbe significanttilt in the chargedistribution, Greneker, E. F., C. S. Wilson, and J. I. Metcalf, The Atlanta
and a significantnumberof recirculatinggraupelparticles tornado of 1975, Mort. Weather Rev., 104, 1052-1057, 1976.
mightpassthroughthe layerbetween0øand-15øC.Distin- Gunn, R., Electric field intensityat the groundunderactive thun-
derstorms and tornadoes, J. Meteorol., 13, 269-273, 1956.
guishingwhich mechanismdominateswill requireat least
Illingworth,A. J., Chargeseparation in thunderstorms: Small-scale
someof the following:simultaneous electricfield measure- processes, J. Geophys.Res., 90, 6026-6032,1985.
mentsin upperregionsof the updraftandin heavyprecipi- Jayaratne,E. R., C. P. R. Saunders,and J. Hallett, Laboratory
tation,improvedDopplerradardata, simulations of severe studiesof the chargingof softhail duringice crystalinteractions,
storms,and three-dimensionallightningmappingsystems. Q. J. R. Meteorol.Soc., I09, 609-630, 1983.
Johnson,H. L., R. D. Hart, M. A. Lind, R. E. Powell, and J. L.
Stanford,Measurementsof radiofrequencynoisefrom severeand
REFERENCES nonsevere thunderstorms, Mort. Weather Rev., 105, 734-747,
1977.
Bent, R. B., andW. A. Lyons,Theoreticalevaluationsandinitial Johnson,R. L., Bimodal distributionof atmosphericsassociated
operational experiences
of LPATS(Lightning Position
andTrack- with tornadicevents,J. Geophys.Res., 85, 5519-5522,1980.
ing System)to monitorlightninggroundstrikesusinga time-of- Jones,H. L., A sfericmethodof tornadoidentification
andtracking,
arrival(TOA)technique, Confer- Bull. Am. Meteorol. Soc., 32,380-385, 1951.
in Preprints,7thInternational
ence on AtmosphericElectricity, pp. 317-324, American Jones,H. L., The identificationof lightningdischargesby sferic
MeteorologicalSociety,Boston,Mass., 1984. characteristics,in Recent Advancesin AtmosphericElectricity,
Branick,M. L., and C. A. DoswellIII, An observation of the editedby L. G. Smith,pp. 543-556,Pergamon, New York,
182 REVIEW OF LIGHTNING IN TORNADIC STORMS

Jones, H. L., The tornado pulse generator, Weatherwise, 18, 78-79, pp. 448-451, American Meteorological Society, Boston, Mass.,
85, 1965. 1982.
Jones, H. L., and P. N. Hess, Identification of tornadoes by Orville, R. E., R. W. Henderson, and L. F. Bosart, An east coast
observation of waveform atmospherics, Proc. IRE, 40, 1049- lightning detection network, Bull. Am. Meteorol. Soc., 64, 1029-
1052, 1952. 1037, 1983.
Kane, R. J., Correlating lightning to severe local storms in the Reap, R. M., and D. R. MacGorman, Cloud-to-ground lightning:
northeastern United States, Weather Forecasting, 6, 3-12, 1991. Climatological characteristics and relationships to model fields,
Keighton, S. J., H. B. Bluestein, and D. R. MacGorman, The radar observations, and severe local storms, Mon. Weather Rev.,
evolution of a severe mesoscale convective system: Cloud-to- 117, 518-535, 1989.
ground lightning location and storm structure, Mon. Weather Rust, W. D., D. R. MacGorman, and S. J. Goodman, Unusual
Rev., 119, 1533-1556, 1991. positive cloud-to-groundlightningin Oklahoma stormson 13 May
Kohl, D. A., Sfericsamplitudedistributionjump identificationof a 1983, in Preprints, 14th Conferenceon Severe Local Storms, pp.
tornado event, Mon. Weather Rev., 90, 451-456, 1962. 372-375, American Meteorological Society, Boston, Mass., 1985.
Kohl, D. A., and J. E. Miller, 500 kc/sec sferics analysisof severe Rust, W. D., R. Davies-Jones,D. W. Burgess,R. A. Maddox, L. C.
weather events, Mort. Weather Rev., 91,207-214, 1963. Showell, T. C. Marshall, and D. K. Lauritsen, Testing a mobile
Koshak, W. J., and E. P. Krider, Analysisof lightningfield changes version of a cross-chain Loran atmospheric sounding system
duringactive Floridathunderstorms,
J. Geophys.Res., 94, 1165- (MCLASS), Bull. Am. Meteorol. Soc., 71, 173-180, 627, 1990.
1186, 1989. Scouten, D.C., D. T. Stephenson,and W. G. Biggs, A sferic rate
Krehbiel, P. R., An analysisof the electric field changeproducedby azimuth-profileof the 1955 Blackwell, Oklahoma, tornado, J.
lightning,Ph.D. dissertation,400 pp., Inst. of Sci. and Technol., Atmos. Sci., 29, 929-936, 1972.
Univ. o1•Manchester, Manchester, England, 1981. (Available as
Shanmugam,K., and E. J. Pybus,A note on the electricalcharac-
Rep. T-11, Geophys.Res. Center, N. Mex. Inst. of Mining and teristicsof locally severe storms, in Preprints, 7th Conferenceon
Technol., Socorro.)
SevereLocal Storms,pp. 86-90, AmericanMeteorologicalSoci-
Krider, E. P., A. E. Pifer, and D. L. Vance, Lightning direction-
ety, Boston, Mass., 1971.
findingsystemsfor forestfire detection,Bull. Am. Meteorol.Soc.,
61, 980-986, 1980.
Silberg, P. A., Passiveelectricalmeasurements
from three Okla-
homa tornadoes, Proc. IEEE, 53, 1197-1204, 1965.
Larson, H. R., and E. J. Stansbury, Associationof lightningflashes
Stanford,J. L., Polarizationof 500 kHz electromagneticnoisefrom
with precipitationcoresextendingto height7 kin, J. Atmos.Terr.
Phys., 36, 1547-1553, 1974. thunderstorms:A new interpretationof existing data, J. Atmos.
Lhermitte, R., and P. R. Krehbiel, Doppler radar and radio obser- Sci., 28, 116-119, 1971.
vations of thunderstorms, IEEE Trans. Geosci. Electron., GE-17, Stanford,J. L., M. A. Lind, and G. S. Takle, Electromagneticnoise
162-171, 1979.
studies of severe convective storms in Iowa: The 1970 storm
Lind, M. A., J. S. Hartman, E. S. Takle, and J. L. Stanford,Radio season,J. Atmos. Sci., 28, 436-448, 1971.
noise studies of several severe weather events in Iowa in 1971, J. Takahashi,T., Rimingelectrificationas a chargegenerationmech-
Atmos. Sci., 29, 1220-1223, 1972. anism in thunderstorms, J. Atmos. Sci., 35, 1536-1548, 1978.
MacGorman, D. R., and D. W. Burgess,Positive cloud-to-ground Taylor, W. L., Atmospherics andseverestorms,in RemoteSensing
flashesduring strongconvectionin severe storms,Eos Trans. of the Troposphere,
editedby V. E. Derr, pp. 17-1-17-17,U.S.
AGU, 72, Fall Meeting Suppl., 93, 199!. GovernmentPrintingOffice, Washington,D.C., 1972.
MacGorman,D. R., andK. E. Nielsen,Cloud-to-groundlightningin Taylor, W. L., Electromagneticradiationfrom severestormsin
a tornadic storm on 8 May 1986, Mon. Weather Rev., 119, OklahomaduringApril 29-30, 1970,J. Geophys.Res., 78, 8761-
I557-1574, 1991. 8777, 1973.
MacGorman,D. R., and W. D. Rust, An evaluationof the LLP and Trost, T. F., andC. E. Nomikos,VHF radioemissions associated
LPATSlightning groundstrikemappingsystems,in Preprints,5th with tornadoes,J. Geophys.Res., 80, 4117-4118, 1975.
on Interactive Information and Pro- Turman, B. N., and R. J. Tettelbach, Synoptic-scalesatellite
International Con3rkrence
cessingSystems, pp.249-254,AmericanMeteorological Society, lightningobservationsin conjunctionwith tornadoes,Mort.
Boston, Mass., 1989. Weather Rev., 108, 1878-1882, 1980.
MacGorman, D. R., and W. L. Taylor, Positive cloud-to-ground Vaughan,O. H., Jr., and B. Vonnegut,Luminouselectricalphe-
lightningdetectionby a direction-finder
network,J. Geophys. nomenaassociatedwith nocturnal tornadoesin Huntsville, Ala., 3
Res., 94, 13,313-13,318, 1989. April 1974,Bull.Am. Meteorol.Soc.,57, 1220-1224,1976.
MacGorman,D. R., W. D. Rust, andV. Mazur, Lightningactivity Vonnegut,B., Comment on "On therelationof electricalactivityto
and mesocyclone evolution,17 May 1981, in Preprints,I4th tornadoes,"by R. P. Davies-Jones
andJ. H. Golden,J. Geophys.
Conferenceon SevereLocal Storms,pp. 355-358,American Res., 80, 4559-4560, 1975.
MeteorologicalSociety,Boston,Mass., 1985. Vonnegut,
B., andC. B. Moore,Electricalactivityassociated
with
MacGorman,12).R., D. W. Burgess,V. Mazur, W. 12).Rust,W. L. the Blackwell-Udall tornado, J. Meteorol., 14, 284-285, 1957.
Taylor,andB.C. Johnson, ratesrelativeto tornadic Vonnegut,
Lightning B., andJ. R. Weyer,Luminous
phenomena
in nocturnal
storm evolution on 22 May 1981, J. Atmos. Sci., 46, 221-250, tornadoes,Science,153, 1213-1220, 1966.
1989. Ward,N. B., C. H. Meeks,andE. Kessler,Sfericsreceptionat 500
Mach, D. M., D. R. MacGorman,W. D. Rust, andR. T. Arnold, kc/sec,radarechoes,and severeweather,in Paperson Weather
Site errors and detectionefficiencyin a magneticdirection-finder Radar,Atmospheric Turbulence,Sferics,and Data Processing,
networkfor locatinglightningstrikesto ground,J. Atmos.Oce- Tech. Note 3-NSSL-24,pp. 39-71, Natl. SevereStormsLab.,
anic Technol., 3, 67-74, 1986. Norman, Okla., 1965.
Orville,R. E., andB. Vonnegut,
Lightning fromsatellites, Ziegler,C. L., andD. R. MacGorman,
detection Observed lightning
flash
in ElectricalProcessesin Atmospheres,editedby H. Dolezalek ratesrelativeto modeledspacechargeandelectricfielddistribu-
and R. Reiter, pp. 750-753,DietrichSteinkopffVerlag,Darm- tions in a tornadicstorm (abstract),Eos Trans.AGU, 71, 1238,
1990.
stadt, Germany, 1977.
Orville,R. E., M. W. Maier,F. R. Mosher,D. P. Wylie,andW. D. Ziegler,C. L., D. R. MacGorman, P.S. Ray,andJ. E. Dye, A
Rust,The simultaneous displayin a severestormof lightning modelevaluationof noninductive graupel-icecharging
in the early
ground strikelocationsontosatellite
imagesandradarreflectivity electrification
of a mountainthunderstorm,J. Geophys.Res., 96,
patterns, in Preprints,
I2th ConferenceonSevere LocalStorms, 12,833-12,855,
 Tornadogenesisvia Squall Line and Supercell Interaction'
The November 15, 1989, Huntsville, Alabama, Tornado
STEVEN J. GOODMAN

NASA Marshall Space Flight Center, Huntsville, Alabama 35812

KEVIN R. KNUPP

AtmosphericScienceProgram, Universityof Alabama, Huntsville 35899

1. INTRODUCTION the use of WSR88-D radars by operationalmeteorologists.

Unfortunately, this ability degradeswith range since, for
Even though the motion of intense long-lived thunder- example,the heightof the radar beamabovethe surfaceis
storms (e.g., supercelIs) can be extrapolated reasonably nearly 1 km at a distance of 100 km. In the present case we
well, nowcastingthe occurrenceof tornadogenesisis difficult attempt to assessthe important low-level processes,espe-
and generally beyond present capabilities. The tornadic cially the interaction between a vigorous squall-linegust
storm that struck the city of Huntsville (HSV) in Madison front and a preexistingmesocyclone,that appear to have
County, Alabama, at 2230 UTC on November 15, 1989, can intensifiedlow-level vorticity, thereby strengtheningan ex-
be described as the end result of a scenario where an isolated istingtornadoor providingan impetusfor tornadogenesis.
supercel! storm developed ahead of and, subsequently, This case study analysis servesto reemphasizethe exist-
interacted and mergedwith a squallline. The supercellwas enceof a highconditionalprobability of tornado occurrence,
identifiedfirst by National Weather Serviceradarsin central given the mergerof a gust front (or storm outflow)with an
Mississippi5 hoursbeforereachingHSV, whereit produced adjacentmoderateto strongthunderstorm (firstsuggestedby
an F4 intensitytornadothat killed 22 people, injurednearly Cook [ 1961]).Mesoscale observationsof the mergerprocess
are used herein to supportthis hypothesis.Observational
500, and causedan estimated$100 million in damage.
data sets include National Weather Service (NWS) compos-
Considerableprogressin our understanding
of tornadic
ite radarproducts,localmesonetdata,MarshallSpaceFlight
storms has been made over the last decade. Doppler radar
Center (MSFC) cloud-to-ground lightning network data
observations[Brandes,1984],laboratorysimulations[Ward,
[Goodman et al., 1988a], and visual observations.
1962],and numericalmodeling[Klempand Rotunno,1983] Section2 documentsthe large-scaleenvironment.Section
have indicatedthe importanceof low-leveldowndraftand 3 providesan overviewmorphologyof the squallline and
outflowin generationand/oramplificationof low-levelvor- supercellstorm. Section4 presentsa detailedmesoscale
ticity within supercellstorms.Satelliteobservations
have analysiscenteredaroundthe time of tornadogenesis.
Section
shown that thunderstormsusually tend to intensifywhen 5 discussesother cases where similar squall line and super-
interactingwith an outflowboundaryproducedby adjacent cell interactions have been observed, and section 6 summa-
deepconvection [Purdom,1976].Low-levelboundariesalso rizes the findings.
playimportantrolesin nonsupercell tornadicstorms[Waki-
moto and Wilson, 1989]. Thus vorticity generationand
2. ENVIRONMENTAL SETTING
amplification withinthe lowest1 km of the atmosphereare
crucial to tornadogenesis. The observed synoptic environmentwas of the type
The abilityto infer the generation
of low-levelvorticity commonlyobservedfor widespreadsevereweather out-
priorto tornadogenesis mayhavepracticalsignificancefor breaks.A deepeningtrough,locatedoverwesternOklahoma
at 1200 UTC November 15, moved eastward to western
The Tornado:Its Structure,Dynamics,Prediction,andHazards.
GeophysicalMonograph79
Arkansasby 0000 UTC. The 1200 UTC lifted indicesat
Copyright1993by the AmericanGeophysical
Union. Nashville, Tennessee (BNA) and Centreville, Alabama
184 TORNADOGENESIS VIA SQUALL LiNE AND SUPERCELL INTERACTION

Severe Weather Events,24 h ending 0600 UTC 16 November1989

IL IN +o L

+ 12 o ',

o O0 +

• + BNA + • +
+ +
o • + • + Huntsville
_ • + _• +•. NC

o
o • ,+ + ' Wind
o Hail>19mm

• o• .• Hail>38mm

Fi•. 1. Distribution of sev'ercweather events over the 24-hour period from 0600 UTC November 15 to 0600 UTC
•ovember ]6. Tornado locations are indicated by inverted trianglesin which the embedded digits indicate the F scale.
Straight line (SL) wind d•ma•e is indicated by plus signs.Soundin• sites are indicated by three-characteridentifiers
defined in the text. The source is the •tion•l Severe Storm Forecast Center severe-weather data base t•pe.

fCKL) were between -5 ø and -6øC. Warm advection and contains significantly more shear than the 1200 UTC LIT
positive vorticity advection occurred east of the trough sounding, consistent with the fact that a jet streak had
within a moist, unstable air mass throughout the day. The entered the synoptic-scaletrough during the day. The upper
unstable air mass was widespread over the southeastern level jet and low-level wind shearare also evident in the 0000
United States, as indicated by relatively large values ('1000- UTC (1900 EST) sounding at Athens, Georgia (AHN) and
2000 J kg l) of convectiveavailable potentialenergy thus corroborates some salient features of the composite
(CAPE) in the !200 UTC soundingsat Little Rock, Arkansas sounding. TheAHN sounding displayed a 64 m s-1 south-
(LIT), zmdJackson, Mississippi (JAN). Combining the LIT westerly wind maximum at 20 kPa and a changein low-level
upper air data with the observed surface conditionsover wind direction from 170ø to 235ø with a corresponding speed
northern Mississippiat 1800 UTC yielded a CAPE of 2800 J increasefrom 4 to 22 m s-• in the 0- 3-km AGL (above
lifted index of-?øC. A large groundlevel) layer. The NMC analysisof the geopotential
kg-1 and a corresponding
number of severe weather events including 16 tornadoes, 199 heightfield over northern Alabama at 0000 UTC displayed
damagingwind, and 63 large hail events were reported over even strongergradientsat all levelsthan that over AHN. The
a 24-hour period. A map of these events is shownin Figure Richardsonnumber (Ri = 12) of the compositesoundingis
1. quite low by virtue of the appreciablewind shearover the
Figure 2 showsan estimatedsoundingfor northern Ala- lower troposphere.Sucha low numberis indicativeof high
bamaat 2200UTC. Input to thiscompositewas derivedfrom potentialfor steady(supercell)storms[WeismanandKlemp,
four soundings,two 1200 UTC upstream NWS soundings 1982].
(LIT and JAN) and two 0000 UTC soundingsat Centreville,
Alabama (CKL) and Nashville, Tennessee (BNA), which 3. STORM SYSTEM MORPHOLOGY
were launched within 1 hour of the tornado occurrence but
were both greater than 180 km distant (see Figure 1 for The primaryfocusof this paperto identifythe mesoscale
tocations). Although the composite soundingis relatively events associatedwith the development of the Huntsville
unstablefor the time of year (LI of -4.9øC and CAPE of tornado.The mostprominenteventinvolvedthe mergerof
1829J kg-•), theinstabilityislessthanthatof the1200UTC an active squall-linesystemwith a supercellstormnear the
LIT soundingmodifiedwith observedsurfaceconditions time of tornadogenesis. This mergeroccurredas a resultof
over northern Mississippi. The composite soundingalso the different motions of the two systems, as seen in
 GOODMAN AND KNUPP 185

2O

5O

6O

7O

8O

9O

PN=31 LI--q.9 CFIPE= 1829 INHIB--23 RI== 12 )

THETR:297.1
RVBRR=13.3
TLCL:290.
I PLCL=919.0
ZLCL=8,•3. 20 pl/S

Fig. 2. Composite sounding estimated for the Huntsville region at 2200 UTC November 15, 1989, plotted on a
skew-T, In p diagram. Refer to section2 of the text for details. The reference dry adiabat (O = 297 K) representsaverage
subcloud values, and a saturated adiabat (0 e = 336 K) is drawn through the lifting condensationlevel (LCL).
Parameters plotted at the bottom include precipitable water (PW in millimeters), lifted index (LI, in degrees Celsius),
convective available potential energy (CAPE in joules per kilogram), lower negative area (INHIB in joules per
kilogram), bulk convectiveRichardsonnumber(RI), mean valuesof boundarypotential temperature and mixing ratio
(theta in degreesKelvin, and RVBAR in gramsper kilogram), and values of temperature, pressure,and height of the
LCL (TLCL in degreesKelvin, PLCL in millibars, and ZLCL in meters).

sequenceof composite radar plots in Plate 1. Two isolated segments,one extending from northeast Louisiana to the
VIP-5 intensity supercell storms (SC1 and SC2) located Texas gulf coast, and the other from southeast Arkansas to
ahead (east-southeast) of the line at 2000 UTC were over- the Indiana-Ohio border. After 1800 UTC, isolated cells and
taken as the line moved eastward more rapidly than the small mesoscalepatches of precipitation formed in advance
of the squallline. Such features are visible in the sequenceof
isolated cells. The motion vector of cells embedded in the
squallline averaged25 m s-• from 235ø, whilethe line composite radar images from 2000 to 2300 UTC (Plate 1).
moveddueeastat 22 m s-•. In contrast,the averagemotion During this time, individual cells within the squall line
vectorof thesupercell
was243øat 18.3ms-•, 8øtotheright produced strong winds and large hail over Mississippi,
of the meanlowertroposphericwind basedon the 1200UTC Tennessee, and Alabama (Figure 1).
soundingsfrom BNA and CKL. Thus the squallline even-
3.2. Mature Phase of the Squall Line
tually overtook the slower supercellstorms(Plate 1). The
severeweatherreportedover northernAlabamafrom each At 2000 UTC (Plate l a) the squall line consisted of a
systemis shownin Figure 3. Both the squallline and thebroken configurationof intense cells. From 2100 to 2200
supercellstormSC1 exhibiteda historyof severeweather UTC the squall system evolved to a more uniform and linear
shape(Plates l b-ld). Although located both in advanceof
prior to their intersection,somedetailsof which are pro-
vided in the following sections. and behind the convective cores, stratiform precipitation
always was more extensive to the rear of the squall line.
3.1. Early Developmentof the Squall Line During the latter stagesof the line's mature phase(Plates 3e
The squalllineformedaround1500UTC. It becamequite and3f), isolatedsupercellstormsformed well in advanceof
extensiveby 1800UTC, whenit consistedof two very long the line, over extreme eastern Alabama and western
186 TORNADOGENESIS VIA SQUALL LINE AND SUPERCELL INTERACTION

-.
.'

-- •
•

2000

Plate I. Composite base-level radar reflectivity maps (NOWrad Composite Radar produced by WSI Corporation,
Billerica, Massachusetts) oblained from regional NWS radars al (a) 2000 UTC, (b) 2100 UTC, (e) 2130 UTC, (d) 2200
UTC, <e) 2230 UTC, and (f) 2300 UTe. The radar reneelivily is conloured al 18,30,40,45,50, and 55 dBZ (VIP levels
1-6).
 GOODMAN AND KNUPP 187

. J
C ,-, -'",-I
,J'-"- - - - - - - '
---J',
,~
r

...
... 4-

~~ 2130

2200

Plate I. (continued)
 188 TORNADOGENESIS VIA SQUALL LINE AND SUPERCELL INTERACTION

••

2230

Plate I. (continued)

-- -- -- -

•
 GOODMAN AND KNUPP 189

northernGeorgia,and producedstrongF2-F3 tornadoes 4OO

overnorthernGeorgia(Figure1)withoutinteracting
withthe
squall line.
Figure4 showsthe 9-hour(1600-0100UTC) timeseriesof
cloud-to-ground lightningproducedby the portionof the
mesoscale convectivesystem(thesqualllineplusthe super-
300
cell) within 200 km radius of HSV. For reference,the
domainof thisanalysisis approximatelythat shownin Plate
1. The mostactivehourof cloud-to-ground lightningactivity
occurredbetween2000and2100UTC. The secondary peaks
in flashratesduringthe declinein electricalactivity(from
2100to 2200UTC) were apparentlycausedby cell mergers 200
and cell interactions.withthunderstormoutflows,both of
whichhelpedreinvigorate the thunderstorms.The lightning
frequency decreasedrapidly after 2200 UTC.
A particularlynoteworthyfeatureof the squallline at 2100
' UTC is the bulge (bow echo) appearingover northwest
Alabama(Plate1b). Very strongsurfacewindsproducingF1 100
damage (at 2122 and 2130 UTC at Florence and Killen,
Alabama, Figure 3, located north of Muscle Shoals(MSL))
were associated with this bow echo, consistent with the
conceptualmodel of a collapsingstorm with surgingoutflow
winds (downburst)proposedby Fujita [1981]. Strongwinds
appear to have been associated with this bow echo for the
1600 1800 2000 2200 0000 02O0
next hour, up to the time of intersection with the supercell 15 November 16 November
storm over Huntsville, as indicated in Figure 3. Time (UTC)
The spatial distribution of cloud-to-groundlightningactiv-
ity during a 1.5-hour period prior to the tornado is shown in Fig. 4. Cloud-to-ground lightning time series (discharges per 10
Figure 5. The storm tracks from southwest to northeast are rain) for the mesoscale convective system observed on November
i5, i989, in the Tennessee Valley.
clearly delineated. Isolated lightning discharges of both
positive (plus sign) and negative (dot) polarity occurred
ahead of the line from thunderstorm anvils, and from the
cloud-to-grounddischargesoccurred with the cell in south-
trailing stratiform rain region behind the main line of storms ern Tennessee throughout the period shown by Figure 5.
(see Plate 1). Numerous, tightly clustered,positivepolarity, Extensive straight-linewind damage was reported from this
cell (see Figure 1). Buechler and Goodman [1988] have
previously documented an increase in the frequency and
density of positive polarity cloud-to-grounddischargesin
Severe weatherassociatedwithsupercelland squallline association with the arrival of downdrafts at the surface
emanatingfrom long-lived microburst-producingstorms.
Squall line darnagingwind path In summary, the squall line was associatedwith initially
intense convection that decreased in strength (but was still
---..
•"• ,,[• + SL
wind
(2215) severe) after 2200 UTC, prior to intersection with the
supercellstorm. Strong surfacewindsand an associatedbow
J • hail
(2200)
• • • Fo•at•n echo apparently existed continuously over a 120-km path
/ Funnel
(2055) _ • • D•catur• between northwest Alabama and Huntsville.
/ k • • / WallC•d
*•P / • • •2•0 Funnel
(2205)
3.3. The Tornadic Supercell Storm
The tornadic supercell storm, identifiedas SC1 in Plate 1,
initially formedprior to 1800UTC, approximately150km in
Path of advance of the squall line in central Mississippi,and main-
Su•ellSC1 / taineda quasi-steadystate after 2000 UTC. A trailingsuper-
cell (SC2), which formed at the same time, followed a path
Fig. 3. Severeweatherevents,asin Figure1, butplottedonlyfor nearly identicalof that of the leadingsupercelland dissipated
the northernAlabamaregion.Additionalobservationsare included just ahead of the squallline over westernAlabamabetween
in the Huntsville region. The two dashedlines connectsevere
weatherpointsproducedby eachsystemand representan average 2130 and 2200 UTC. The only known severe weather pro-
linear motion of each system. duced by SC2 was large hail over eastern
190 TORNADOGENESIS VIA SQUALL LINE AND SUPERCELL INTERACTION

ß o
o

Fig. 5. Cloud-to-groundlightning activity from 2105 to 2236 UTC on November 15, 1989. The polarity of the
dischargesis denoted by a plus sign for positive and a dot for negative.

Several reports of large hail (up to golf ball size) and strong Observations around Huntsville between 2210 and 2230
winds and two reports of funnel clouds near 2100 and 2200 UTC indicate that the largest hail was only pea size, signif-
UTC were associated with the leading supercell (SC1), icantly smaller than that observed in the previous 30 min-
which eventually produced the F4 tornado at Huntsville. utes. A wall cloud was observed by NWS personnel and
Thus supercellSC1 was relatively steadyand long-livedbut NASA scientists between 2215 and 2230 UTC (Plate 2). No
apparentlywas unableto generatetornadoesduringthe first rotation was evident as the wall cloud passedNWS observ-
5 hours of its lifetime. ers at ---2215UTC. A photographof the wall cloud5 minutes
During the 1-hour period prior to tornadogenesis,the before tornado touchdown (Plate 2) also reveals the dense
NWS local warning radar at HSV revealed that supercell precipitationcurtainassociatedwith the approachingsquall
SC1 had high echo tops (--•17 km) but no well-definedhook line. For the next 5 min, ropelike funnel-shapedcloudswere
echosignatures.At 2200UTC the supercellachieveda peak observed beneath the wall cloud and were associated with
in vertically integratedliquid water (VIL), as measuredby sporadicsurface damage. Cloud-to-groundlightning was
the Nashville Radar Data Processor (RADAP) II systeminfrequentin the vicinity of the wall cloud, but intracloud
located 190 km to the north. Reports of golf ball sized hail lightningflashes,calculatedfrom two separatevideo tape
were common about 30 km west-southwest of HSV around recordingsbetween2230and 2235UTC, were producedat a
this time. VIL decreased significantly after 2200 UTC, rate in excessof 40 per minute. Large ratios of intracloudto
indicatinga weakeningin storm vigor. A funnel cloudwas grounddischargeshave been previouslydocumentedin
observedjust southof Decatur, Alabama(see Figure 3 for Oklahomasupercells[MacGorman et al., 1989], Alabama
location)duringthisdecline.The impendingmergerbetween air-massstormsproducingmicrobursts[Goodman et al.,
the squallline andsupercel!is clearlyindicatedby the5-min 1988b], and in deep tropical stormsduringthe premonsoon
totals of cloud-to-groundlightningshownin Figure 6. seasonnear Darwin, Australia [Williams et al.,
 GOODMAN AND KNUPP 191

]. ci 17

o o o O O D O

D D D O

o o 17 o 17 o o o o o o O O O D

O D D D D O ff D

D O D O O O O D D D O I O D O D D O O O O D O O

O O O D D O O D O O D O O O O O O

O O O O O O O O O O O 17 D O O O

O O O

O O O D O O O O O O O

O O 11'•DI o OOOD O 0 O OO OOOO •

D

!1 fi 1 n II n II n n n n n II fi n

Fig. 6. Contoured lightningdensity map during the period 2215-2220 UTC. Contour interval is 0.01 dischargesper
squarekilometer, beginningwith the value 0.02 dischargesper squarekilometer.

Observations from UAH (see Figure 7a for location) were high-precipitation(HP) supercel1,examined by Moller et al.
made by one of the authors (K.R.K.) as the storm passedto [1990] and thought to be common in the moist environment
the south. At this location there was an absence of outflow typical of the SoutheasternUnited States. Rather, this super-
air with low values of 0e. Winds were generally light and cell may be betterclassifiedas a cyclic supercell,a term coined
variable as the supercell passed from SW to SSE. The to describe the periodic genesis of new mesocyclonecores,
estimated
rainfallratewasmoderately
low (10-15mmh-•), wall clouds, funnel clouds, and tornadoes [Burgesset al., 1982;
showers of pea-sized hail were observed, and visibility Jensen et al., 1983; Johnson et al., 1987].
through the precipitation core was not severely restricted. The radar summaries in Plates 1d and 1e show merger of
Cloud-to-ground lightning and low cloud formations were the supercell and squall line near the time of tornado
visible in the supercellinflow. As the supercellstorm core formation. As indicated in the previous discussion,both the
passedSSE of UAH near 2230 UTC, the squallline reached squallline and supercell were declining in intensity near the
UAH. Estimated wind gusts immediatelybehind the gust time of merger. The squall-line passagewas a highly impres-
frontwere ---30m s-• , visibilitywasreducedto ---300m in sive event throughout much of the Huntsville region. Con-
heavy rain, and the temperaturedroppednearly 6øC. sequently, many eyewitnesses were able to relate the loca-
Other eyewitnessobservationsand a video recordingof tions of the tornado to the squall line. The next section
the squall-lineshelfcloud (but not the tornado)suggestthat provides finer-scale details of the interaction between the
the gustfront exhibitedappreciabledistortionandcurvature squall-line gust front and supercell storm.
by ---2237UTC (Plate 3). When viewed from the southeast,
the leadingedgeof the gust-frontsegment(labeled"squall 4. MESOSCALE ASPECTS OF THE MERGER PROCESS
line shelfcloud" in Plate 3) southof the tornadois well to the
right (east) of the tornado.The easternmost locationof the 4.1. Characteristics of the Tornado
gust-frontalsurgeis estimatedto be midwaybetweenthe
observerand the tornado. This panoramicview also indi- Tornado characteristics and associated cloud patterns
catesa secondary wall-cloudformation(in theforeground of were determined using information from a detailed ground
Plate3) locatedalongthe leadingedgeof the gustfront. Thus survey, aerial survey photographs, and eyewitness reports
the supercelldoesnot appearto be characteristic of the and videos. The tornado path length was ---30 km,
 As Seen From 84481
looking Southwest

Plate 3. Panorama composite image constructed from 11 video frames laken around 2237 UTe, showing cloud
formations associated with the squall line. The view is north through southwest (counter·c!ockwise) from point GT in
Figure 7a. The tornado is located -6 km north allhis lime. Video was taken by Greg Talley.

- - - - ------
 GOODMAN AND KNUPP 193

Jones Valley
Element;

UAH

TR

MS2
MS1 [] 3

O8
MSFC
O
HSV WSO O4 ARF Path Site
Photo
Site
StationSymbol: o5
T Gust
Tdß Rain
o9

ol 1

OArmy Mesonet Site TopographicContours:

OAnalog RecordingSite River
Interval: 100 ft
[] EyewitnessObservationSite Beginning:700 ft

2215 UTC
(b)

StationSymbol:
T Gust
Tdß Rain
23e½
18
G5 1.0
RFD Air
G12 Calm
o
TRW-A
>'
211•
16
G6(•
0.5 /2230

22•G5
16 22•
eG7
17
Observed Wall
18 e•l•G7
22•G7
Cloud at 2215

16

OArmy MesonetSite
22
%G7
2O12•;G9
OAnalog RecordingSite
[] EyewitnessObservationSite
10 m/s
8i
Fig. 7. (a) Station locationsand tornadoF scale intensityat points alongthe path of the tornado and analysisof
mesonetdata for (b) 2215 UTC, (c) 2230 UTC, and (d) 2245 UTC. The thin vectors in Figures7b-7d represent15-min
wind averagesfrom the Army mesonetsites, while the thick vectorsdenotemeasuredor estimated(E) instantaneous
winds. The dashedline intersectingwith the tornadopath in Figure7 c marksthe range of the time-to-spaceconversion
of 1-minwindspresented
inFigure10.Vectormagnitudes
aresuchthat1kmrepresents
5 m s-2 forthe15-min
average
and 10 m s-2 for the instantaneous
value.Plottedstationdataat the Army sitesincludetemperatureT anddew point
Ta (bothin degreesCelsius),peakwindgust(in metersper second)overthe 15-minperiod,and 15-rainrainfall(,in
millimeters).In Figures7c and 7d the estimatedlocationof the squall-linegustfront is shown.The tornadotrack is
indicatedby the stipplingand extends20 km further to the
194
TORNADOGENESIS
VIA SQUALLLINEANDSUPERCELL
INTERACTION

(c) 2230 UTC

/

VSBY
<1/....•
/• • /2245
Squall
Line
StationSymbol: Gust
F•nt
/• and
,nt•nVVlif%•
s •"•0
T
Td ß Rain
Gust

21 G9 L
•' 223•
Peak Gust 21
O
TRW+A
• ' I
i,

22•G21/ •G15 22•G•0

/ 18• Gll
/

// 22
16 •G9

/ 22 G9 G9
20

ß Army Mesonet Site

OAnalogRecordingSite
[] EyewitnessObservationSite
>

012345km 10 m/s

(d)
2245_UTC
/
[] ~ 25 m/s lt
EB • •' ,

-- 15m/s
• 2240
StationSymbol: •-,••• Tornadic
T Gust •ec • • •Mesocyclone
Td ß Raill
17Gll /• Gusts
/•
HSV 13•9.7 • • • • 25-• .•
TRW+
_/ / Video
•..... • • • • • //•S•- ff •,econdary
•:•X • • 21/// -'• • W
VallCloud
'½ 19G2•••, '•el 5
16•••'•' 61 • •

G151•20' ' •
21 2.8 •
16•'
21•16 ;' ?
16 1.5
©Army Mesonet Site
oAnalog RecordingSite
[] EyewitnessObservation Site
(•'i ,• • •-•km 10 m/$

Fig. 7 (continued)

maximumpath width was ---800m, and the peak intensity Atmospheric Administration, 1989]. An exhaustive search
was marginally F4 over a small portion of the path. A failed to uncover photographyof the tornado itself. Infor-
detailed damage survey analysisby the University of Chi- mationon the tornadoappearanceis basedsolelyon eyewit-
cago can be found in Storm Data [National Oceanic and ness observations. The initial 60% of the smoothed
 GOODMAN AND KNUPP 195

26
Surface
- ,
mesonetdata timeseries-sites$ (solid)and8 (dashed) observedthroughoutthe entire path length. There is little
_.a
.... -1:(8) sq'uall
line
passage
- evidence in the surface damage patterns that the tornado
_

evolvedto a thin ropelike vortex before dissipating.

18 i. - .....
•\•
T(3)• Td(3) pas,sage
-
cold
front
'
4.2. Mesoscale Analysis: General Patterns
14
Becausethe supercell passedover a mesonetduring the
10
tornadogenesis phase, some surface characteristics of the
007
supercell storm, and of the interaction between the evolving
...• , • , [ ",...• , [ , 3.(- , • • /
• 1005 supercell thunderstorm (mesocyclone/tornado) and the
squall line, can be defined. Mesonet data consisting of

%,missing
dat•
•/
'- 10O3 15-min averages of all basic meteorologicalparameters,
including peak wind gust within the 15-min period, were
•_.1001 • /
recordedat 12 sites operated by the U.S. Army on Redstone
999 '. t • • I • .• ..• . ..•./ , •... , I Arsenal. These data are supplemented with two additional
24 surface stations(from which analog recordingswere avail-

••C ½Gust
(3)
2O able), one at the HSV WSO and the other at the MSFC
Atmospheric Research Facility (ARF), as well as with eye-

Mean
(3)•f
Gust
(8)
E 16
witness observations at sites shown in Figure 7a. Because
-• 12
the Army data are 15-min averages, spatial scales less than
• 8 .
---10 km are unresolvable. Finer-scale structures were
4
gleaned from 1-min wind data available from the HSV and
MSFC ARF sites.
340 Time series of 15-min data from two key mesonetstations
300 (3 and 8) are presented in Figure 8. These data clearly show
260 the passage of the squall line near 29'0 .,.• UTC and the
22O synoptic-scalecold front near 0100 UTC. Important differ-
180 ences in wind characteristics, discussed in further detail
140 below, are apparent at each site.
100 Merger of the 30-dBZ echoes started at ---2145UTC, and
•" i , i '", i" • i ' , .
by 2230 UTC the supercell had become absorbed by the
e squall line (Plate l e). At the surface the mesonet station
12 3 -
_
plots displayed in Figures 7b-7d for 2215, 2230, and 2245
Total Precipitation _

UTC depict the passage of the system over the mesonet.

8

8f '\ . Prior to the arrival of the system, the surfaceanalysisat 2215

3 - 32.5 mm
_

8 - 34.5 mm
4 UTC (Figure 7b) reveals uniform conditions. Averaged
~

.
.
windswere light southerlyand temperature/dewpoint values
• I,/' ,, ! I I I •7'"'•1 ,
1800 2000 2200 0000 02(10 were near 22øand 17øC,respectively. The 2230 UTC analysis
Time (UTC) (Figure 7c) is more complicatedand ambiguousbecausethe
15-minaverages(2215-2230 UTC) have smoothedthe gradi-
ents associated with the mesocyclone and approaching
Fig. 8. Time seriesof surface parametersfrom mesonetsites 3 squall-linegust front. The flow is, however, generally cy-
(solid lines) and 8 (dashed lines) (mesonetlocationsare shown in
Figure7): (a) temperatureand dew point, (b) sealevelpressure,(c) clonic and convergent. The location of the squall-linegust
15-rainaveragedwindsand peakgustwithin the 15-raininterval, (d) front, depictedin Figure7c, was determinedprimarilyfrom
wind direction, and (e) precipitationover each 15-mininterval. 1-min time series data from HSV and MSFC ARF. Rela-
tively warm and moist air remainsover the easternpart of
the mesonet,but average winds have increasedconsiderably
line of the tornado path, along with estimatesof F scale from 2215 UTC, and wind gusts within the southeasterly
intensity, are shown in Figure 7a. flowarerelatively
uniformin therange9-11m s-•. These
Prior to tornadogenesis,a well-definedwall cloud was gustsmost likely occurredin responseto the increasing
observedfrom the photo site location at 4481, shown in pressuregradient around the mesocyclone.One notable
Figure7a. Intensification to F3 severityoccurrednear2235 exception
isthe15ms- • gustrecorded
atsite3 at2230UTC
UTC, at aboutthe time the squall-linegustfront intersected (Figures7c and8). This gustmayhavebeenassociated with
the tornado/mesocyclone circulation(Figure 7d). At this the recent arrival of the squall-linegust front, which dis-
point, the tornadowidened,and a rightwardturn of---22ø playedan eastwardbulgesouthof the mesocyclone center,
occurred in the direction of movement. A strong southerly as shown in Figure 7c. To the north of the mesocyclone
inflow of F0 intensity,inferredfrom sporadictree falls, was track, gustsaheadof the squallline were producedby
196 TORNADOGENESIS VIA SQUALL LiNE AND SUPERCELL INTERACTION

JHSV
I...."J ['-50
.....
Windspeed(kts)
/. !]•SquallLiie•ust
i I /' ,ii'i 40.._2L_7
mr
n • • . •. SupercellRFD

'/' ,/ • /
o
••
r/"•330
t
[
. . i•.2%.
•supe,rcellRFD
½•]••J•
...•..•Squall LineGust

%/30 t 4,/min• , I , , •.

/i I/7/i
/ •/Microbarograp
////
MSFC ARF /7//////
Wind
Direct0n ••• ©'•'cqo • • .•-04o
Time (CST)

Time (CST)

Fig.9. Analog recordingsfrom2200to2250UTC(1600--1650

CST)ofwindspeed
(inknots)
attheHSVWSO(top),
windspeed (inknots)attheARF(middle),andwinddirection
(indegrees)
attheARF(bottom).Theplottotheright
isa microbarograph
trace(inHg)attheARFfrom0000to2400CST.Thetimelagbetweenthesquall
line'sgust
front
andrearflankdowndraft (RFD)isindicated
inthetopandmiddlepanels.
(10knots
= 5.1ms-1, 1inHg= 33.9mb.)

supercell's with its rear flankdown- (11 m s-•) with the squall-line
gustfront associated passage
than siteslocated
draft, as happenedat HSV and MSFC ARF. southand northof site8 (Figures7c and7d). For example,
By 2245UTC (Figure7d), windsoverthe regionshowa a wind estimateof---25 m s-• occurredat UAH, and 22 m
westerlycomponent, andtheflowis againbroadlycyclonic. s-• was measuredat site 3. This observationsuggeststhat
The reconstructed wind flow (consistent with the video the pressure gradientswereweakerwithinthe supercell's
scenein Plate 3) indicatesthe squall-linegust front had core than at other locations behind the squall line's gust
front.
passed
overtheentirenetwork.Themaximum
gusts(indi-
catedin bothFigures7 and 8) weremostlikelyassociated In summary, thereis strongevidencethatthe supercell's
withthe squallline'sgustfront.Of particularinterestis the mesohighblocked the eastward advance of the squallline
variation in mean wind direction and speed of the maximum gustfront.Thesquall-line
gustfrontsurged eastwardsouth
gustoverthenetwork. Fromthetimeseries datainFigure8, ofthemesocyclone center(Figures
7c and7d), thusforming
wenotedpreviously thatsite8 recorded weakerwindgusts a significant perturbationin the gustfront. These
 GOODMAN AND KNUPP I97

MSFCTimeto Space,2220-2232UTC The gustfrontassociated with the RFD of the supercell

was in closeproximity(<1.5 kin) to the mesocycloneat
location 4481 (shown in Figure 7a) at 2227 UTC. The
sequenceof events observedthere by one of the authors
(S.J.G.)wasas follows:a wall cloudwasobserved moving
Full
Barb
=5m/s FFE).
•
• 2220 fromsouthwestto the east,andwhenit wasto thesoutheast,
there was a gustfront passageassociatedwith the RFD. This

G9
RFiI6.
•,•224••
2222 wasfollowed---2min later by the morevigoroussquallgust
andassociatedheavy rain. Thus there were two gustfronts
G18G•..••
'•2226'• in closeproximity to the mesocyclone.
To quantifythe relative locationsof the mesocyclone
and
•2228 Wind
Shift
gustfronts(bothfromthe supercellandfromthe squallline)
2232 '1
I
Arrivalof SquallLine
as a function of time, time series data from the HSV and
MSFC locations(depictedin Figure9) werefurtheranalyzed
and compared,notingthe relative time separationof wind
Fig. 10. Time-to-space conversion of theanalog winddatafrom patternsassociatedwith the supercelland squallline. This
locationMSFC,alongthedashed lineintersectingthetornado path analysiswas supplementedwith key eyewitnessobserva-
(a andb denotetheendpoints of theline)in Figure7c. An average tions(locationsMS1, MS2, and4481in Figure7a) wherethe
storm speedof20ms- • hasbeenassumed. Windgusts areinmeters time of the squall-linegust frontal passagecould be deter-
per second.The locationsof the data relative to stormfeaturesat
2230 UTC are shownin Figure 7c.
mined from the relative locationsof the gust front and
tornado. All time difference observations were converted to
distancesassuming a translationalspeedof 20 m s-] The
resultinganalysispresentedin Figure 11 includesplots of
suggest that tornado formation and/or intensification were
associatedwith the interactionof the squallline with the (Figure1la) the distancebetweenthe supercell'sRFD gust
supercell'smesocyclone. However,the 15-minaveragesare frontandthe squall-linegustfront, (Figure lib) the distance
insufficient Analy- betweenthe mesocyclone/tornado
to uncoverthe detailsof thisinteraction. and the squallline's gust
sisof the 1-mindata from the HSV and MSFC ARF sites, front, and (Figure 11c) one point showingthe distance
augmentedby key eyewitnessobservations,provide some betweenthe supercell'sRFD gust front and the mesocy-
clarificationof the merger. clone. All four points on curve b form a nearly straightline
and suggest that the fastersquall-linegustfront apparently
intersectedthe tornadoat the golf course(seeFigure7a for
4.3. SurfaceFeaturesand Timingof the Merger
location)near2235UTC, just prior to the maximumintensity
The time series wind data from HSV and MSFC ARF both of the tornado. This intersection was viewed from the NW
show two wind maxima whose time separationdecreases sideof the tornadoby a knowledgeableobserver(MS2), who
from ---7 min at HSV to ---4 min at MSFC ARF (top and watched the tornado pass just to the south, and then ob-
middle panels in Figure 9). The initial wind maximum was servedstrongwinds and very heavy rain associatedwith the
apparently associated with the rear flank downdraft of the squall line. By the time the tornado reached F4 intensity
supercellstorm, while the secondpeak wasproducedby the (Figure 7a) a common eyewitness observation was that
squall-line gust front. The 1-min data at MSFC, shown in weak winds precededthe damagingtornadic winds and that
Figure 10 have been convertedto a distancescaleusingthe strong winds and heavy rain occurred for several minutes
observed storm-motion vector. These data reveal that the after the tornadopassed.These observationssuggestthat
supercell exhibits the classical structural details at the sur- the tornado, while at maximum F4 intensity, was indeed
face, includinga forward flank downdraft (FFD that gives locatedat the leading edge of the (distorted)gustfront and
rise to northeasterly flow) and the rear flank downdraft apparently remained there for the remaining 14 min of its
(RFD, the westerly flow). As noted previously, the squall- lifetime.
line gust front appears to have exhibited appreciablevari- Details of the squall line-supercell merger south of the
ability over the analysis region. At HSV the supercelland tornado are less certain becauseof the lack of good eyewit-
squall-linemaximumwind gustswere 14 and 23 m s-• nessreports. The surface data suggestthat, to the south, the
respectively,
while
atMSFC
they
were
nearly
thesame
at1• squall-linegustfront infiltrated the mesocycloneat an earlier
and18m s-• (thewinddirection associated withthesegusts time. There are observations (e.g., Plate 3 and Figures 7c
was ---270ø).Over the mesonet the peak squall-linewind gust and 7d) that the squall-line gust front to the south of the
variedappreciably from11m s-• at site8 to 21m s-• at site tornado was moving faster. Several eyewitnessespositioned
3 (Figure 8). The strongestwind gust was observed10 km to a few kilometers south of the tornado path noted that the
-1
the north at UAH (Figure 7b), where a value of 25-30 m s strongest winds associated with the gust frontal passage
(275ø) was estimated. No outflow winds were observed in were fi'om the south to southwest. During the period of
association with the supercell at this location as the storm tornado intensificationfrom F1 to F4, the eyewitnessrefer-
passed to the south. enced in the previous paragraph (MS2 in Figure 7a)
198 TORNADOGENESIS VIA SQUALL LINE AND SUPERCELL INTERACTION

servedthat the tornadowidth increasedrapidlyfrom ---100m 10 5

at 2233 to >500 m by 2236 UTC. Although such a rapid • Estimated

Tomado
lntensity
•
increaseis consistentwith a pressurereduction associated
with an intensifyingvortex, the condensationfunnel expan- a)oist•nco
equivalent
ofL
sionmay have been accentuatedby the arrival of outflow air • 6 - gust fronttimeseparation
(either from the RFD or squall line) that the surfacedata in
Figure 83 show was closer to saturation than that of the air
4- b) Messyclone
•M•FC2230 2 5 0 0
mass preceding the squall line.
center
to•.•11.... • MSFC
5. O'r•4}•m C^sv:s
2 c)Mesocyc!one
center
'- to supercel•ust front • •o• •. •Golf
What is the significance of the interaction of the type 0 ................... , .... ,.......
L • . ,• .... •uøurse
......... I
documented here and how frequently do such interactions 22'5 2220 2225 2230 2235 2240 2245

occur? This case has served to reemphasize an important Time (UTC)

scenario, defined previously by Ward [1962, 1968] and
ttamfiton[1969], involving the merger of a squall line with an Fig. 11. The interaction between the squall line and the supercell
isolated vigorous storm ahead of the line. Hamilton summa- mesocyclone(tornado) is depicted in terms of the time dependence
of (a) the distance separation of peak wind gusts associatedwith the
rized work conducted by Cook [1961], regarding observed
rear flank downdraft outflow of the supercell and the squall-line gust
cases in which cells that developed ahead of a squall line front, as measured at HSV and MSFC (the time differences have
were subsequently overtaken by the squall line. When the been converted to space, and data points are marked by solid
cells were intense at the time of the merger, a tornado was squares), (b) the distance between the squall-line gust front at a
reported in eight of I! cases. In two of the remaining three point and the estimated location of the mesocycloneor tornado at
that time (the line represents the best linear fit to the four data
case:s, t'unnel clouds were observed. A similar interaction
points, denoted by the plus signs),and (c) the distancebetween the
has been noted in association with the F4 Edmonton, Al- supercell's rear flank downdraft (RFD) gust front and the mesocy-
berta, tornado that caused 27 deaths and more than 300 clone (shown by the solid dot with error bar). These data show that
injuries on July 31, 1987 [Bullas and Wallace, 1988]. The the time of the squall-line gust front merger with the tornado
coincided with the time of tornado intensification around 2235 UTC,
Huntsville case can be classified similarly and provides a
while the supercellRFD gustfront was closeto the tornado near the
quantification of the process involved in such a merger. time of tornado formation. The location of the golf course is shown
For the November 15, 1989, event it was noted that the in Figure 73.
active squall line also merged at an earlier time (•--2145UTC,
Plate I) with supercell SC2. In this merger there were no
reportsof tornadoes,and the supercellappearedto weaken analysisof surfacemesonetdata, that the radar-echomerger
prior to becoming engult'ed by the squall line. One may was accompaniedby the interactionof a vigoroussquall-line
speculate that this supercell, which trailed the Huntsville gust t¾ontwith the mesocyclonecirculation. As the squall
supercell, may have ingestedoutflow air left in the wake of !ine's gust front intersectedthe supercellmesocyclone,the
the t'ormer. The echo structures associated with the 15 other gustfront becamedistortedandappearedto wrap aroundthe
tornadoes reported on this day were also examined to south side of the mesocyclone.In addition, this interaction
determinepotentialinteractionsbetweenthe parent cell and was associatedwith a rapid intensificationof the existing
the squall line. Interestingly,eight tornadoesoccurrednear tornado.The potentialimportanceof this interactionon the
the squallline, eitherfrom new cellsformingimmediatelyin initiation of the tornado is less clear. A less vigorous gust
advanceof the primary line or from existing cells (not of front associated with the rear flank downdraft of the super-
prolongedsupercellstature,as wasthe casefor SC1)within cell storm was in the vicinity of the tornado at the time of
the line. Tornadoesnear the squallline were generallyweak tornado formation. Thus one unresolved aspect of the tor-
(two F0, four F1, and one F2 intensity; see Figure 1), with nadogenesis is whetherit was triggeredby the squallgust
the exception of the Huntsville storm. Generally stronger front or by the supercell'srear flank downdraft.
tornadoes (F2-F3) were produced by isolated supercell The importanceof the rear flankdowndraftin intensifying
storms over eastern Alabama and Georgia. low-level vorticity within the mesocyclone,prior to the
tornadogenesis
stage,was suggested by LetnonandDoswell
[1979]and has sincebeen further quantifiedwith Doppler
6. DISCUSSION AND SUMMARY
radar data [Brandes, 1984] and with cloud model results
This paperhasdescribedsomeof the featuresassociated [Klemp and Rotunno, 1983]. These studieshave demon-
with the merger of a vigorous squall line with a strong stratedthe potentialimportanceof the gustfront in increas-
supercellstorm,whichformed5 hoursearlierin advanceof ing verticalvorticityvia additionaltilting of low-levelvor-
the squallline. At the time of the merger,both systemswere ticity, generationvia the solenoidalterm of the vorticity
decliningin strength.To our knowledge,the supercelldid equation, andadditional amplificationvia convergence along
not generatetornadoesprior to this merger,althoughtwo thegustfront.The importantrole of vorticityamplification
funnel clouds were observed earlier. We have shown, using byconvergence alsohasbeensuggested in theDoppler
 GOODMAN AND KNUPP 199

analysis of Johnson et al. [1987]. Our observationsare Fujita, T. T., Tornadoes and downbursts in the context of general-
consistent
with thesestudiesbut bringan additional
compli- ized planetary scales, J. Atmos. Sci., 38, 1511-1534, 198l.
cating factor into the model, that is, the introduction of a Goodman, S. J., D. E. Buechler, and P. J. Meyer, Convective
tendency images derived from a combination of lightning and
morevigoroussquall-linegustfrontgeneratedexternally,its satellite data, Weather Forecasting, 3, 173-!88, 1988a.
interactionwith the RFD gustfront, andits apparentimpor- Goodman, S. J., D. E. Buechler, P. D. Wright, and W. D. Rust,
tance in intensifyingan existingtornado. Lightning and precipitation history of a microburst-producing
It is possiblethatthisproblemmaybe investigated further storm, Geophys. Res. Lett., 15, 1185-1188, 1988b.
with three-dimensional numerical models in order to define Hamilton, R. E., A review of use of radar in detection of tornadoes
and hail, Tech. Metno. WBTM-ER-34, 64 pp., Weather Bur. East.
the detailedphysicsof the mergerand the precursorcondi- Reg. Headquarters,Garden City, N.Y., 1969.
tions needed for tornadogenesis. Jensen,B., E. Rasmussen,T. P. Marshall, and M. A. Mabey, Storm
scale structure of the Pampa storm, in Preprints, 13th Conference
on Severe Local Storms, pp. 85-88, American Meteorological
Acknowledgments. Irv Watsonof NOAA kindlyprovidedNWS Society, Boston, Mass., 1983.
compositeradar informationwhich made a significantcontribution Johnson,K. W., P.S. Ray, B.C. Johnson,and R. P. Davies-Jones,
to this study. Mike Botts expertly completedthe imageanalysisof Observations related to the rotational dynamics of the 20 May
Plate 3. Discussionswith Don Burgessand R. Davies-Jones,and a 1977 tornadic storms, Mort. Weather Rev., 115, 2463-2478, 1987.
thorough review by R. Davies-Jones,improved the content of this Klemp, J. B., and R. Rotunno, A studyof the tornadicregionin a
paper. Steve Williams assistedin reductionand interpretationof the supercellthunderstorm,J. Atmos. Sci., 40, 359-377, 1983.
mesonet and soundingdata, which were furnished by the meteoro- Lemon, L. R., and C. A. Doswell III, Severe thunderstorm evolu-
logical team on RedstoneArsenal. Tim Rushingassistedin analysis tion and mesocyclonestructure as related to tornadogenesis,
and reduction of the mesonet data. We also thank the many Mon. Weather Rev., 107, 1184-1197, 1979.
individuals who kindly provided videos and descriptiveinformation MacGorman, D. R., W. D. Rust, D. W. Burgess, V. Mazur, W. L.
of particular events. Partial funding supportwas providedby NASA
Taylor, and B.C. Johnson, Lightning rates relative to mesocy-
under grant NAG8-654. Reference herein to any specificcommercial clone evolution in tornadic storms on 22 May 1981, J. Atmos.
products, process, or service by trade name, trade mark, manufac-
Sci., 46, 221-250, 1989.
turer, or otherwise, does not constitute or imply its endorsement,
Moller, A. R., C. A. Doswell III, and R. Przybylinski, High-
recommendation, or favoring by the United States Government or
precipitationsupercells:A conceptualmodel and documentation,
the University of Alabama in Huntsville.
in Preprints,16th Conferenceon SevereLocal Storms,pp. 52-57,
American Meteorological Society, Boston, Mass., 1990.
REFERENCES National Oceanic and Atmospheric Administration, Storm Data,
vol. 31, no. 11, 6-17. (Available from the Natl. Clim. Data Center,
Brandes, E., Vertical vorticity generation and mesocyclone suste- Natl. Oceanic and Atmos. Admin., Asheville, N. C.)
nance in tornadic thunderstorms: The observational evidence,
Putdom, J. F. W., Some uses of high resolutionGOES imagery in
Mort. Weather Rev., 112, 2253-2269, 1984.
the mesoscaleforecastingof convectionand its behavior, Mort.
Buechler, D. E., and S. J. Goodman, Initial observations of cloud-
Weather Rev., 104, 1474-1483, 1976.
to-groundlightningactivity in microburst-producingstorms, in Wakimoto, R. M., and J. W. Wilson, Non-supercell tornadoes,
Proceedings,8th International Cot!ferenceon AtmosphericElec-
Mort. Weather Rev., 117, 1113-1140, 1989.
tricity, pp. 842-848, SwedishNatural ScienceResearchCouncil
Ward, N. B., The effectof low level wind shearon the formationof
and UppsalaUniversity, Uppsala,Sweden.(Availableas S-755
92 from Univ. Inst. of High Voltage Res., Husbyborg,Uppsala, atmosphericvortices,in Proceedings,2nd Confi'renceon Severe
Sweden.) Local Storms, pp. 1-4, American MeteorologicalSociety, Bos-
Bullas, J. M., and A. F. Wallace, The Edmontontornado, July 31, ton, Mass., 1962.
1987,in Preprints,15th Conferenceon SevereLocal Storms,pp. Ward, N. B., Rotational characteristicsof a tornado cyclone, in
438-443, AmericanMeteorologicalSociety, Boston,Mass., 1988. Proceedings,13th Weather Radar Cot•œerence, pp. I83-186,
Burgess,D. W., V. T. Wood, and R. A. Brown, Mesocyclone AmericanMeteorologicalSociety, Boston, Mass., 1968.
evolution statistics, in Preprints, 12th Conference on Severe Weisman,M. A., and J. B. Klemp, The dependenceof numerically
Local Storms, pp. 422-424, AmericanMeteorologicalSociety, simulated convective storms on vertical wind shear and buoy-
Boston, Mass., 1982. ancy, Mon. Weather Rev., I10, 504-520, 1982.
Cook, B. J., Someradar LEWP observationsandassociatedsevere Williams, E. R., S. A. Rutledge, S. G. Geotis, N. Renno, E.
weather, in Proceedingsof the 9th Radar Conference,pp. 181- Rasmussen,and T. Rickenback, A radar and electrical study of
185, AmeNcanMeteorologicalSociety,Boston,Mass., 1961. tropical"hot towers," J. Atmos. Sci., 49, 1386-1395,
 Discussion

ALAN MOLLER, SESSIONCHAIR

National Weather Service

PAPER C 1 (Joe Golden, National Oceanic and AtmosphericAdminis-

tration, Office of the Chief Scientist) Could you say a little
Presenter, Bill McCaul, Universities Space ResearchAsso- moreaboutthe detectionproblem?It is a vexing operational
ciation [McCaul, this volume, Observations and simulations problem. What percentage of the time do you expect a
of hurricane-spawnedtornadic storms] mesocyclone that is detectableby NEXRAD radars,and is
(Bob Davies-Jones, National Severe Storms Laboratory) the supercel1the dominant[tornado-producing] mechanism
You showed that the updrafts are weak in these storms.Are in general,as it was in Hurricane Danny?
the downdrafts also weak? There is not much evaporative
(McCaul)I havelookedat severalotherlandfallingtropical
cooling in hurricanes.
cyclones,and Danny is not the only one that produced
(McCaul) No, there is not much. In fact, in a lot of the supercelIs.I have seenevidencein HurricanesBeulahand
simulations, there is little evidence of a cold pool. Allenof long-livedstormsthat producedmultipletornadoes.
Concerningdetectability,I think that it will be a problemif
(Davies-Jones)So how do you explain the wall cloud [in
the stormsare not close enoughto the radars for the lowest
Hurricane Danny]?
3 km of the stormsto be sampled. If the lowest 3 km cannot
(McCaul) That casemay have beena little atypical.It had be sampled,thenthe only hopefor detectionis to examine
more CAPE and somewhat more potential for producinga grossfeaturessuchasstormlongevity.I findthatsupercelIs
downdraft.It would be nice to get moreobservationsof such in hurricanes last at least 2 to 3 hours, sometimes even
cases.
longer.
(Howie Bluestein,University of Oklahoma)Do you think (Chair)I guessthatthe storm-relative
mid-levelwindsdeter-
that there are other instances in nature where there are weak mine the usual location of the tornado in the right rear
CAPE andhodographs like thosein tropicalcyclones?
Very quadrantof the individual storm.
weak CAPE is quitecommon,but whereelsewouldsimi-
larly shapedhodographsoccur? (McCaul) There can be some variability from hurricaneto
hurricane,dependingon the directionof the shearof the
(McCaul)I personallyhaven't seenany otherhodographs ambientsteeringcurrent.Generally,it is almostfromrear to
with the same shapeas the tropicalcycloneones.These front of the hurricane, becauseit is guidingthe hurricane's
have a low-level or 700-mbarjet, with weak windsaloft. In
motion.So the rightfront quadrantis favored,butthisis not
mid-latitude storm environments, there is generally a jet
always the case.
aloft, as in the Bluesteinand Jain [J. Atmos. Sci., 42,
1711-1732,1985]compositesounding.Frequently,thereare (Chair)I am talkingabout an individualstormcell, not an
thermodynamic profilesthatarequitesimilarto theonesin individualtropicalcyclone.I want informationfor tornado
hurricaneenvironments, almostmoistadiabaticwitha small spotters.
I sawsomefootagefromnearAustin,Texas,of
positiveareaat low levels.Thisis seenin a lot of mid- sometornadoes thatappearedto be onthe northeastsideof
latitudeandtropicalconvectivesystems.But I don'tknow a stormcomplexin a westwardmovinghurricanethat had
of other environmentswith semicircularor nearly circular comeinland. Would you expect the tornadoesto be in the
hodographs. fightrear quadrantof their parentcell?
The Tornado:Its Structure,Dynamics,Prediction,andHazards. (McCaul)I thinkthat'scorrect.It's the sameas Hurricane
GeophysicalMonograph79 Danny,exceptrotated90ø becauseof the difference in the
Thispaper
isnotsubject
toU.S.copyright.
Published
in 1993
bythe motions of the hurricanes.
AmericanGeophysicalUnion.
202 DISCUSSION

PAPER C2 PAPER C3

Presenter,Ed Brandes,NationalSevereStormsLaboratory Presenter, Chuck Doswell, National Severe Storms Labora-

[Brandes, this volume, Tornadic thunderstorm characteris-
tory [Doswell and Burgess, this volume, Tornadoes and
tics determinedwith Doppler radar] tornadic storms:A review of conceptualmodels]
(Chair) If we in the National Weather Service detect a
(Keith Brewster, University of Oklahoma) You discussed
mesocyclone,shouldwe issue a tornado warning basedon
that alone'?
the different classificationsof storms. Do you think that
tornadoesshouldbe classifiedaccordingto whetherthey are
(Brandes)I think that I would, but Don Burgessis a better producedby mesocyc!ones or whetherthey are landspouts?
expert on that than ! am. Given a mesocyclone,there is a
50% probability of a tornado. That number may be biased (Doswell) From a practical viewpoint, it is difficultto tell the
more to mid-level circulations than low-level circulations. If type of tornado at present. When there is a network of
there were strong rotation at low levels, I would certainly Doppler radars acrossthe country, I would advocate classi-
issue a warning. fying tornadoes into mesocyclone and nonmesocyclone
types.
tChair) But we should wait for the low-level circulation'?

(Brandcs) Yes. PAPER C4

(V. G. Blanchette, retired professionalengineer) The data

Presenter, Don MacGorman, National Severe Storms Lab-
you show are at some distance from the tornado. Have you
oratory [MacGorman, this volume, Lightning in tornadic
considered what is going on inside the tornado?
storms: A review]
(Brandes) The tornado is not well resolved in the radar data.
We need really dense measurementsto determine exactly (Tom Grazulis, Environmental Films) In the [August 28,
what is going on. The observations are spaced several 19901 Plainfield [Illinois] tornadic storm, was there also a
hundred meters apart. In this particular case [the Harrah ratio of 50 cloud-to-cloudflashesto every 1 cloud-to-ground
flash?
storm], I think that the [low-level] mesocyclone flow was
excited by rainy downdraft development. The trigger for the
(MacGorman) We don't have data on cloud-to-cloud flashes
tornado itself may have been some type of shearing instabil-
ity at the side of strong mesocyclone winds. in the Plainfield case. [The cloud-to-cloud data for the Binger
storm were acquired from a special L band radar.] The
(t•ou Wicker, University of Illinois) From your observa- [cloud-to-ground] data that I presented are from the State
tions, does the tornado-scale vortex develop nearly simulta- University of New York system that operates continuously.
neously through a fairly deep layer, say, the lowest 2-3 km, So we were able to get the [cloud-to-ground] data, much like
or does it develop first at a particular height'? we are hopingto be able to do with Doppler radar data when
NEXRAD radars are installed.
(Brandes) I suspect that it is the latter. The data that were
used here were roughly at 10-minute intervals, and it is far
too coarse to really see what is going on. We need rapidly PAPER C5
spinning radars that get 2-minute data, and even they may
not provide the answer. In the Del City storm where the Presenter, Steve Goodman, NASA Marshall Space Flight
sampling was roughly every 5 minutes, it appears that the Center [Goodman and Knupp, this volume, Tornadogenesis
initial formation was at about 2-kin elevation. The initial
via squall line and supercell interaction: The November 15,
radar signature of the vortex was more intense at 2 km than 1989, Huntsville, Alabama, Tornado]
it was above or below this height. In the Harrah storm, the
vortex did not build up to 5 km until later. No time for questionsbecause sessionran
 Tornado Detection and Warning by Radar
DONALD W. BURGESS 1

National SevereStormsLaboratory, NOAA, Norman, Oklahoma73069

RALPH J. DONALDSON, J•.

HughesSTX Corporation,Lexington,Massachusetts
02173

PAUL R. DESROCHERS

Geophysics
Directorate,
PhillipsLaborato•3•,
Hanscom
Air ForceBase,Massachusetts
01776

1. INTRODUCTION mated tornado detectionalgorithms(section 4), and con-

cludewith thoughtson future radar-warningimprovements
Traditionally,tornadoeshave been difficultto detectby (section $).
radar. Conventionalradar signatures,althoughavailablefor
over 30 years, havebeenonly qualitativelyassociated
with 2. LIMITATIONSOF DOPPLERWEATHER RADAR
actual tornado existence. As recently as the !980s, it has
been suggestedthat most successful tornadowarningsare A Dopplerradarmeasures theradialcomponent of target
still based on visual sightings,even when the warning motion.Radartargets,or scatterers,are typicallyprecipita-
message mentions radarsignatures [seeKellyandSchaefer, tionparticles(rainandhail), althoughsensitiveradarsmay
1982]. detectreturnfrom particulatematter(insectsand airborne
The adventof meteorological Dopplerradarwithvelocity debris)and refractiveindex gradients.Clouddroplets,such
informationhashelpeddefinetornadolocationconsiderably, as the condensationfunnel of a tornado, are too small to be
but the small size of the tornado(with respectto radar pulse detected by allexceptveryspecialized radars.Combinations
volume)makesunambiguous identificationdifficult.How- of two or more radars (multiple Doppler networks)can be
ever,evidenceof tornadopotentialis readilyavailablefrom used to extract full, three-dimensional, air motion from
individualradar velocity components.The high cost of
Dopplerwiththedetection of thelarger,parentcirculation
surrounding the tornado(i.e., the mesocyclone or misocy- maintainingseveraloperationalradars in the same area
clone).The impending upgradeof the UnitedStatesopera- precludes useof real-time,multipleDopplernetworks for
tional weather radar network to Doppler (the NEXRAD warning purposes.
Much useful information for tornado detection and warn-
Program) warrantsfocusof attentiononDopplersignatures
insteadof thoseassociatedwith conventional(non-Doppler) ingcanbe obtained
from single-Doppler,
radial-component
radar.
information. An area of rotation stands out in a single-
of problemsasso- Dopplerfield of mean velocityas a coupletof strong,
This reviewwill beginwith discussion
ciatedwith radar detectionof tornadoes(section2), continue localizedflow, towardand awayfromthe radar(Figure1). If
the tornadois extremelylarge and closeto the radar, the
witha lookat theprogress
madein tornadodetectionand
warningsincethelastsymposium
(section auto- rotationmay be attributableto the tornadoitself.Often,
3), discuss
however,aspectratio problems(seebelow)precludetor-
1Nowat WSR-88DOperational Facility,Norman,Okla- nadoobservation,
Support at leastfor a well-definedcouplet,andthe
homa. detectedsignature is the largerareaof swirlthat surrounds
the tornado,definedas the mesocyclone or misocyclone,
TheTornado:Its Structure,
Dynamics,
Prediction,
andHazards.
depending on the sizeof the swirl.
GeophysicalMonograph79
Copyfight
1993by theAmerican
Geophysical
Union. Thefirstsuggested Dopplersignature for a tornado[Atlas,
 ~

d
'"Z
1;
o
o

~
i5
z
~
~
Z
Z
a
-<
'"
~
~

Plale 1. Reflectivity (left) and velocity (right) for misocyclone signatures of June 15, 1988 (from Wakim% and Wilrotl, 19891. Color codes
are in the middle; misocyclone and tornado locations are marked by arrows on velocity image. Range marks are in kilometers. Data are
courtesy of Lincoln Laboratories and the National Severe Storms Laboratory, NSSL.
a

Plate 2. (0) Relectivity and (b) velocity displays from the Oklahoma City. Oklahoma, WSR·88D for the tornado outbreak
of April 26, 1991. Mesocyclone algorithm output has been overlaid in Plate 2b; past (future) tornado tracks are asterisks (solid
lines). The Wichita tornado is beyond quantitative velocity range (230 km, 125 om). Violent (F4·F5) tornadoes are underway
with each of the three large Slorms. Future WSR-88D sites (KEND and KICT) are indicated in Plate 2a. Displays are couI1esy
of WSR-88D Operational Support Facility.
206 TORNADO DETECTION AND WARNING BY RADAR

4
- .

1. Radar Horizon Problem

•-1
-2

-3

-3 -2 -1 0 1 2 3 4
DISTRNC[
2. Aspect Ratio Problem
Fig. 1. Single-Doppler signature of a vortex [from Wood and
Brown, 1983]. Heavy lines are true airflow; light lines are radial Fig. 2. Schematic depicting limitations of radar.
velocity contourswith dashesfor inboundradial component;short
arrows mark radial velocity maxima.
Also, lowest-elevation-angle measurements for ranges
greater than 80 km (40 nm) are frequently above cloud base.
1963]was not the horizontal pattern of the coupletbut a very Because a tornado and funnel are defined only below cloud
large spectrum width (the spread of velocities about the base, interpretation of longer range observations must be
mean value). This is reasonable because return from a limited to findingrotation in the cloud, without knowing for
sample volume containing a tornado would feature both sure that the rotation extends to the ground. This means that
inbound and outbound radial velocities and thus high spec- radar observationsmust be supplementedby other informa-
trum width. Radar observations have shown that spectrum tion (i.e., storm spotter reports) to be completely definitive.
width signatures, by themselves, are not good tornado Similarly, boundary layer convergence lines, possible
indicators. This occurs because there are several causes of tornado formation areas, cannot be detected at longer ranges
broad spectrum widths, both meteorological and hardware unless they extend well above cloud base. Typically, these
associated(see Doviak and Zrni• [1984] for a more thorough boundaries are hard to detect beyond 100 km (50 nm) range.
discussion).Some smaller Doppler weather radars currently
available commercially provide only spectrum-width-related 2.2. Aspect Ratio Problems
outputs. These radars are thought to have only low skill in
tornado detection. During operational Doppler tests, spec- The radar recognition of atmospheric vortices depends on
trum width, by itself, has never been demonstrated as a the aspect ratio between the radar illumination volume
useful tornado signature. (beam width) and the vortex size (core radius). For more
All radars, including Doppler, suffer from two main limi- details, see Burgess and Lemon, [1990]. Aspect ratios for a
tations in observing weather echoes. These limitations affectRankine vortex model are graphed in Figure 3. When the
rotation signatures and make them less useful. A short radar beam width exceeds the core radius by a factor of 3,
discussion of each follows. less than one half of the true maximum velocity is detected,
and the vortex will probably not be recognized.
This is illustrated by the lower portion of Figure 2. The
2.1. Radar Horizon Problems
vortex on the left (aspect ratio < 1) is at a range where it is
Refraction of microwave energy by the atmosphere pro- definitely large enough to be detected. The vortex on the
duces radar beams that do not bend at the same rate as the right (aspectratio of 2) is at a range where it will be detected
Earth curves. For this reason, radar beams, even at 0ø only for a favorable combination of radar viewing angles.
elevation angle, will increase in height as they travel away The center vortex of the lower part of Figure 2 is at a range
from the radar. This processis illustratedby the schematicin where it is much smaller than the beam width (aspect ratio
the top part of Figure 2. Since tornadoesand mesocyclones >3) and therefore will not be detected by the radar. The
are best defined in the lower and middle portions of storms, center vortex approximates the typical sampling condition
a radar will detect a rotation signature from a nearby storm for an operational radar scanningan average-sizedtornado
but will overshoot a signaturefrom a storm at longer range. at moderate
 BURGESS ET AL. 207

J.O- BEAMWIDTH .
MEAN DOPPLER VELOCITY PROFILES • R'-•"•
S'0'0
THROUGH A RANKINE COMBINEDVORTEX
FOR VARIOUS BEAMWIDTH /CORE RADIUS RATIOS

io

io

0-2'•0.0
-!.0 4 II • •
i I I I I Jr- I I I ,,I ,, I I I

NORMALIZED AZIMUTHAL DISTANCE (DISTANCE/CORE RADIUS)

Fig. 3. Continuousvelocity profile acrossa Rankine combinedvortex for a number of beam width to core size ratios
[from Burgessand Lemon, 1990].

In real radar observationsthe recognitionproblem is turbance, and descent of its base, as tornado touchdown
further complicatedby discreteazimuthalsampling.A peak time approached.
in velocity will not be optimally resolved unlessone of the The Doppler observations of the Union City tornado
discretesamplevolumesalongan azimuth is centeredon the alternated between PSI mode and digital recordings of the
velocity peak. velocity data. Analysis of the digital data by Burgess et al.
[1975] revealed a unique and characteristic signature of
3. RADAR SIGNATURES FOR TORNADO DETECTION extremeazimuthalshear, in excessof 0.05 s-j, that first
AND WARNING appeared aloft 33 min before the earliest tornado damage.
The base of this pattern, now called a tornadic vortex
This section will document the development of Doppler signature(TVS) because of the large magnitudeand vertical
signatures during the 1970s and 1980s, the time since the structureof the shear, descendedto the groundcoincidentin
Lubbock, Texas, tornado symposium(1976). It will begin by both time and space to tornado touchdown.
reviewing research that was in progressat symposiumtime This discovery stimulated a search for more TVS exam-
and continue through the rapid developmentsthat began plesby Brown and Lemon [ 1976]and was formally discussed
soon thereafter. Real-time Doppler applications were, in by Brown et al. [1978]. These authorsfound 10TVS appear-
part, made possibleby significantimprovementsin computer ances in the years subsequent to and including the 1973
processors, velocity estimators, and display technology. Union City tornado. In all but two cases a tornado or funnel
Rapid data processingand real-time, color display made it cloud was observed to accompany the TVS. The two unver-
possibleto learn about tornadicstormevolutionand to apply ified cases occurred in sparsely populated areas with no
the results in public warnings. report of whether or not a tornado occurred, so they cannot
fairly be considered as false alarms. Although TVS detect-
3.1. Tornadic Vortex Signatures
ability is severely limited by resolution, when a TVS does
An analogdisplay,the plan shearindicator(PSI) devel- appear, it is prudent to assume that a tornado is on the
oped by Armstrong and Donaldson [1969], was employedto ground or soon will be.
record the Doppler velocity field in the stormthat spawned An example of a very strong TVS is shown in Figures 4a
the devastatingUnion City, Oklahoma,tornadoof May 24, and 4b. The displayed velocities for adjacent radar-viewing
1973. Analysis of these observationsby Donaldson[1978] azimuths (Figure 4a) indicate a maximum velocity difference
revealed a disturbedpattern of velocitiesat midlevels(5-9 of 114m s-• acrossthe azimuthchangeincrement(0.5ø),
km) in the storm 45 min before the earliest tornado damage. leadingto a shearvalueof 1.9 x 10-• s-•. At observation
of the dis- time, the tornado was 1 km wide and was producing
Subsequentobservationsshowedintensification
 208 TORNADO DETECTION AND WARNING BY RADAR

•o/ I 286. 0 I0 ratedat F4 onthe Fujitascale(seeVasiloff[thisvolume]for

1909 CST a pictureof thetornadoanda plotof theTVS signaturefrom
0,4 ø ELEV anotherNSSL radar). Doppler velocity data collectedin a
speciallyadaptedmode,whereinvery largewind speed
could be more easily resolved, indicated maximum tornado
windspeeds of at least100m s-• [Zrni•et al., 1985a].The
TVS andtornadolocationwithinthe mesocyclone maybe
estimatedfromFigure4b, although the mesocyclone signa-
o
tureis somewhat obscured by the strongTVS. The implied
mesocyclonediameteris 4 km (seeWoodandBrown[1983]
o
for a simulationof the two-signaturecombination).The
hook-shapedechoof thereflectivity
field(stippledin Figure
Lfl
4b) is relatedto the mesocyclone,not the TVS.
It is importantto point out that not all tornadoes,even at
• -20 m closerange, will have TVSs. Assuminga 1ø beam with
samplesevery 1ø (adjacentbut not overlapping),the maxi-
,
mum detection range for TVSs is estimated at not much
morethan 100km (50 nm), evenfor largetornadoes(1-2 km
diameter).Narrow tornadoes(10-50 m diameter)may es-
22 MAY 1981
-IO
cape detectionat only 20 km (10 nm) range.Thus whole
i i I I i classesof tornadoesmay be excludedfrom observationfor
-73 X-DISTANCE FROM NORMAN (km) -65
most of the radar coverageinterval. For example,many
Fig. 4a
nonsupercelltornadoes such as those along gust fronts
(weak, narrow,and confinedto the boundarylayer) do not
havedetectable TVSs(seeVasiloff[thisvolume]for a plotof
one that was detected).Other nonsupercelltornadoessuch
-2O
as landspoutsand waterspoutsare undetectablebeyond
0.4 ø ELEVATION close range, and supercell tornadoes cannot be detected at
moderateto longrange.However,the presenceof supercell
andnonsupercelltornadoesmay be inferredby the strength
of the mesocyclonesor misocyclonessurroundingthem.
TVS observationcan lead to warninglead times of 10 min or
more, particularly for supercell tornadoes.
Although there have been a considerable number of TVS
observationsat National SevereStormsLaboratory(NSSL)
[Vasiloff;this volume]and at other places(Air Force Geo-
physics Laboratory [Kraus, 1973], National Center for At-
mospheric Research [Wilson et al., 1980], the National
Z ..:.:.:,:,:.:,:,:,:.:,:
Weather Servicein Alabama [Petit, 1990], Colorado [Dunn,
1990],andOklahoma[Burgesset al., 1991]),no completeset
of climatologicalstatisticshas been compiled.Until that is
i
accomplished, conclusions related to the TVS must be
termed tentative and subject to change. The development
and interpretationof a large data set will probablyawait the
deployment of the NEXRAD network.
-1o -!o
-12 t I ,,
-73 X-DISTANCE FROM NORMAN (kin) -65
3.2. Mesocyclone Signatures
Fig. 4b Mesocyclonesserve as temporal bridgesbetween super-
Fig. 4. Plot of (a) single-Doppler velocities and (b) overlaid cell tornadoesand effectiveforewarningof their destructive
reflectivitiesfor Binger, Oklahoma, tornado of May 22, 1981 [from power. The initial observationby Doppler radar of a meso-
Burgess and Lemon, 1990]. Heavy lines are single Doppler con- cyclone was presented by Donaldson et al. [1969]. They
tours; light lines in Figure 4a are radial centers; reflectivity factor observed a pattern suggestiveof a mesoscalevortex in a
>30 dBZ is shadedin Figure 4b. Selectedvelocitiesalongthe vortex heavily damaging storm with a suspected tornado. The
centerline are shown for every other gate location in Figure 4a. An
example of radar sample volume size (150 m by 0.8ø) is shaded in mesocyclonewas confirmed by visual observations of cloud
Figure 4a. Peak mesocyclone velocities are marked with "A" and base motions offered by an eyewitness.
"B" in Figure 4b. A singleDoppler radar providesonly thosecomponents
 BURGESS ET AL. 209

RANGE (kin)
70 65 60 55 50 47 50
8

1 o.

o
1700 1730 FI FO 1800 F41830 1900
TIME(CST) F1F2

Fig. 5. Time-heightplot of shearfor the mesocyclone of April 30, 1978(adaptedfrom Burgessand Donaldson,
[1979]).Shadingindicatesthe tornadicvortex signaturepresent;solidbarsat bottommark tornadotimes.The last
tornado is at Piedmont, Oklahoma.

the velocity field directedalongthe radar beam. A pattern pier-aidedtornadowarningsachievedan averagelead time
suggestiveof rotation therefore is subject to ambiguity. before tornado touchdown of 21 min, comparedwith the
Consequently,Donaldson[1970]proposedcriteriaof shear, average 2-min lead time for warningsprepared without the
persistence, and vertical extent to establish confidence in the benefit of Doppler radar information. This superiorJDOP
validity of a mesocyclonesignature,by showingthat alter- detectionand warningperformancevalidatedthe incorpora-
native interpretationsof this distinctive Doppler velocity tion of Doppler capability into NEXRAD.
pattern are unlikely when the criteria are met. A time-height plot of one of the mesocyclonesdetected
Shortly thereafter, Burgess [1976] initiated a search for during JDOP (Figure 5) indicatesthe relationshipsbetween
mesocyclonesin Oklahoma, and during a 5-year period he the mesocyclone, the TVS, and the accompanyingtorna-
identified 37 of them using Donaldson's criteria. Tornadoes does. The mesocyclone forms well before the first tornado
were associatedwith 23 of these mesocyclones,with a mean but is defined only at storm midlevels. The first tornado (F1)
lead time of 36 min from mesocyclonedetection to tornado occurs as the mesocyclone becomes defined in the boundary
touchdown. In the same study, Burgess found that no layer. The second and third tornadoes form as the mesocy-
verified tornado occurred duringradar observationswithout clone shear intensifies (at low levels and midlevels), followed
a preceding mesocyclone signature. Furthermore, all but soon by the fourth tornado (F2), accompanied by a brief,
two of the 37 mesocyclonesproduced some form of severe midlevel TVS. The fifth and strongest tornado (F4) forms
weather events at the surface. Recent observations with about 15 min after the rapid intensificationof a new meso-
more sensitive radars and different scanning techniques cyclone core and the developmentof a new TVS. Note that
indicate a somewhat less but still significant association of the intense phase of the mesocyclonecore and TVS occur-
tornadoes with mesocyclones. rence are almost coincident and that the mesocycloneweak-
The close association between mesocyclones and torna- ens rapidly as the violent tornado and TVS dissipate.This
does revealed by the Burgess [1976] study, and the promise suggeststhat vortex stretching and at least partial conserva-
of warnings offered by the precedence of mesocyclones, tion of angularmomentum are associatedwith the formation
provided crucial encouragementfor initiation of the Joint of the violent tornado. After the last tornado, the mesocy-
Doppler OperationalProject (JDOP). The JDOP experiment clone weakens, and its top lowers rapidly.
was designed to determine the improvements, if any, that The caseillustrated in Figure 5 is a classicsupercellstorm.
could be achieved in warning of tornadoesand other severe For these storms it can be seen that the mesocyclone
storm hazardsby the introductionof Doppler radar data into signatureprecedestornado occurrenceby several tens of
an operational scenario. The results reported by the JDOP minutes and that the time of maximum mesocyclonestrength
staff [Burgesset al., 1979] were very impressive,showing is the most likely time for strong and violent tornadoes.
improved detection for tornadoes and a remarkably small Also, the mesocyclone signature extends through a deep
number of false alarms when compared with using conven- layer and therefore is visible at extended ranges from the
tional forecasting techniques. Most significantly, the Dop- radar. These characteristics make mesocyclone
210 TORNADO DETECTION AND WARNING BY RADAR

16

MS

OS - ORGANIZING STAGE
MESOCYCLONE
MS- MATURE STAGE
CORE #1
DS - DISSIPATING STAGE

MESOCYCLONE
CORE # 2
4--

0,,

MESOCYCLONE
CORE
4-.

0 40 80 120 160 200 240 280

TIME (rain) • 40 40

Fig. 6. Time-height evolution of (top) radar echo and (bottom) multiple mesocyclonecores [from Burgess et al.,
1982]. Solid horizontal bars are tornado occurrence; tornado formation interval is indicated along bottom.

from supercellstormsvery useful in the radar contribution to the vortex core exists only over a small, shallow depth and
the tornado warning process. is associated with divergence.
With single-Doppler radar, characteristics of the mesocy- Some mesocyclones produce just one core, but others
clone inner core are readily measured (see Figure 1). The produce a seriesof cores. These cores occur in a predictable
characteristicsand evolution of supercell mesocyclonecores way, acting as the mesocyclone propagation mechanism.
have been studied at NSSL [Burgess et al., 1982] and at The relationship between multiple mesocyclone cores is
other research facilities. Core life is composed of three conceptualizedin Figure 6. The mesocycloneforms after the
stages [Burgess and Lemon, 1990]: storm echo is mature and near its peak height. The first (and
Organizing stage. This is a period of growth, both up- sometimes only) mesocyclone core has a relatively long
ward and downward, from midlevel beginnings (5 km organizing and mature stage. The second core organizes as
height). It is common for convergenceto exist below the the first begins dissipating. Second and succeeding cores
mesocyclone base. The organizing stage ends when the have extremely short organizingstagesas they quickly form
mesocyclonebase extends below cloud base. over a large depth and have relatively short mature stagesas
Mature stage. This is the period of maximum strength evolutio• proceeds rapidly. Some supercell storms have a
when velocity parameters have their highest values and long successionof cores, and the mesocyclonepersistsfor
tornado formation potential is highest. The mesocyclone severalhours. These stormscan produce "tornado families"
signatureextends through a deep layer, perhapstwo thirds with a tornado recurrence interval of about 40 min.
of the storm height. A conceptualmodel of the evolution of a supercell meso-
Dissipating stage. This stage begins with a rapid de- cyclonein horizontalsection(Figure 7) revealsthe existence
creasein mesocycloneheight and is generally characterized of a gust front that wraps cyclonically about the mesocy-
by weakeningvelocities. At the end of the dissipatingstage, clone core. The core evolution closely resembles
 BURGESS ET AL. 211

localizes at the point of the occlusion, and a secondvortex

core organizesrapidly becauseof the vorticity-rich environ-
MESOVORTEX CORE ment in which it forms.
EVOLUTION
Not all mesacyclones produce tornadoes. In fact, the
converse is true; most mesacyclones do not produce torna-
does. Statistics available from 20 years of Doppler radar
CORE NO, 2 operationin Oklahomaindicatethat only in yearswith large
numbers of significant supercelIs do 50% of the mesacy-
clones produce tornadoes. In other years, tornadoesare
producedby as few as 30% of the mesacyclones.The
long-termaverageis somewherebetween30% and 50%.
The job of trying to categorizemesacyclonesby their
tornado potential has not been easy. Data from 45 well-
observedsignatures(Figure 8) indicate few well-defined
•2 boundariesbetween mesacyclonesthat produced no torna-
NO. I does, weak or strong tornadoes, and violent tornadoes.
There seemsto be a tendency for tornadic mesacyclonesto
have higher rotational velocity and stronger shear.
Donaldson and Desrochers [1985], in a search for an
indicatorof mesacycloneintensitythat might correlatewith
the severity of subsequenttornadoes,proposedcalculation
Fig. 7. Conceptual model of mesacyclone core evolution [from
Burgesset al., 1982].Heavy lines are low-level wind discontinuities; of excess rotational kinetic energy (ERKE) of the mesacy-
tornado tracks (exaggeratedfor size indication) are shaded. Insert clone core. ERKE is the rotational kinetic energy remaining
showstornadofamily tracks, and the squareis the regionexpanded after subtractionof the productof core radiusanda selected
in the figure. value of shear from the observed rotational velocity. The
resultant diminished velocity can be regarded as excess to
cyclonedevelopment:duringthe first core maturestage,the the minimumrequirementfor maintenanceof mesacyclonic
gust front acceleratesaround the right flank; occlusion shear.
occurs, and as the mesacyclonewarm sectorseparatesfrom Continuationof ERKE study by Desrocherset al. [1986]
the first core, its dissipation begins; strong convergence showedthat a notable quality of tornadic mesacyclonesis

MESOCYCLONES

Core Diameter Shear/Vortlcity

LLI 15
Z

• i
I I i
0
I krn 5 i0
o o

bJ Rotational Velocity

lxlO"$s 'l 5

Fig.8. Characteristic
parameters
of mature
supercell
mesocyclones
fortornadic
andnontornadic
storms
[from
Burgess
etal., 1982].
Tornado
implies
F0-F3andmaxi-tornado
implies
212 TORNADO DETECTION AND WARNING BY RADAR

De! City (20 May 1977) Ada (29April 1978)

• , • ', 6 .--. • . • 6

FZ F3
'--"' [ ,..
'• , ] , 12 , . , 12

6
'• ...
3

0
18oo 1900 2000 1900 2000

TIME (CST) TLM• (CST)

Fig. 9. Examples of mesocyclone excess rotational kinetic energy (ERKE) evolution for a tornadic storm (Del City)
and a nontornadic storm (Ada) [from Desrochers et al., 1986]. ERKE is normalized by the value of a climatological
mature mesocyclone (CMM). Tornado durations are indicated by the solid bars.

their intensification at low levels prior to tornado formation. portion near the ground. Third, beam averaging will reduce
The evolution disparity between tornadic and nontornadic the mesocyclone peak rotational velocity and, perhaps,
storms is borne out well by the Del City (May 20, 1977) and causeunderestimatesof mesocycloneparameters. For all of
Ada (April 29, 1978) storms (Figure 9). Although the Ada these reasons, it is generally best to correlate Doppler
mesocyclone was at times as intense as Del City, this energy signatureswith spotter reports whenever possible.
was confined to midlevels, precluding tornado formation. In
contrast, Del City generated two strong tornadoes.
3.3. Misocyclone Signatures
Recently Donaldson and Desrochers [1990], with an ex-
panded sample of 17 Oklahoma mesocyclonic storms, im- Beginningwith the late 1970s, there was realization that
proved the ERKE technique by integrating the values from Dopplerdetectablecirculationsandtornadoesoccurredwith
surface up to several kilometers in height. They also per~ other storm types besides supercelIs. Fujita [1979] docu-
R)rmed similar integrationsfor other mesocyclonicfeatures, mented a tornado along the leading edge of a squall line with
including velocity and shear. They found (Figure I0) that a bow-shapedecho. Burgess and Donaldson [1979] com-
ERKE predictedthe violent (F4) tornadoessuperbly,with a pared a supercellmesocyclonewith a smaller,weaker cir-
minimum lead time of 28 min and no false alarms. For the culationin a newly developing,multicellstormandfoundthe
strong tornadoes(F2 and F3), velocity as well as ERKE two signaturesdifferent. Forbes and Wakimoto [1983],
provideda morethanadequatewarningleadtime to the first studyingconventionalradar data, suggestedthat tornadoes
occurrenceof a strongtornado in a storm. Little or no skill observedwith a squall line came from small-scalecircula-
was achieved for identification of the so-called weak (F0 and tions, perhapsformedin associationwith interactingdown-
FI) tornadoes. bursts and gust fronts.
Recently, National Weather Service officeswith accessto Bluestein[1985]observeda developingsquallline and saw
Doppler radars have based accurate tornado warningson a waterspoutliketornado that he dubbeda "landspout."
detectingmesocyclones[Dunn, 1990; Burgesset al., 1991, Dopplerstudiesof thesephenomena by Wilson[ 1986],Brady
Burgessand Letnon, 1991]. Mesocyclonesignaturesare the and Szoke [1989], and Wakirnotoand Wilson [1989] identi-
most often used Doppler radar inputs to tornado warnings fied small-scale,single-Dopplersignaturespreceding and
becausethey are larger and deeper than TVSs or misocy- surroundingthesenonsupercell tornadoes(seePlate 1 for an
clones and therefore can be seen throughout all of the example). Since the signaturesare smaller than average
velocity coveragerangeof modernDoppler radars(230 km mesocyclones with supercelIs,they have been called miso-
or 125 nm). cyclones(lessthan4 km sizein the scaleclassification
of
There are limitations in usingmesocycloneinformationin Fujita [ 1981]).
tornadowarnings.First, it is indirectevidenceof a tornado. Misocyc!ones and associatedtornadoesoccurwith multi-
As already mentioned,less than 50% of all mesocyclones cell storms, with ordinary cells in squall lines, and with
produceverifiedtornadoes.Second,rememberthat Earth newly developingstorms(sometimesjust cumulusconges-
curvature (horizon limitation) placesthe center of the low- tus)alonglow-level,wind shearboundaries [Wilson,1986].
est-elevationradar beam at greater than 4 km height when A time-height plotof a misocyclone
(Figure11)indicatesthat
the rangeis 200km (100nm).No informationis availableon the circulationfirst developsbefore there is appreciable
the low-level character of the mesocyclone, the important radar echo, is mainly confinedto the boundarylayer
 BURGESS ET AL. 213

Piedmont(F4)

./

/
- F2 & F3 Tornadoes

• r'" .-':--"F0&F1Tornadoes

0 I

!
0% 20% 40% 6O% 80%

Combined Error Rate: False Alarms and Failures to Detect

Fig. 10. Performance of excessrotational kinetic energy in warning of three intensity classesof tornadoes: violent
(solid line), strong (long-dashedline), and weak (short-dashedline) [from Donaldson and Destochefs, 1990]. The
ordinate gives median lead time, consideringonly the earliesttornadoof that intensity in a storm. The abscissashows
combined errors expressed as the sum of false alarms and failed detections.

its lifetime, and has a relatively short lifetime. In general, time to provide a positive warning. In fact, in one case a
these characteristics are much different from the supercell tornado developed 5 rain, before detection of a first radar
mesocyclone of Figure 5. echo! Furthermore, the small size of a misocycloneimposes
On the basis of Doppler radar data from Colorado, Brady a severe limit to the range at which the pattern of Doppler
and Szoke [1989] proposed a conceptual model of misocy- velocitiesmight indicate the possibilityof a tornado. Finally,
clone tornado formation (Figure 12). Small vortices develop there are indications that the most intense rotation of miso-
in the boundary layer along convergent wind-shift lines. cyclones is generally confined to heights just above the
New convective updrafts stretch and intensify the misocy- ground, and these most active parts of the phenomenon
clone. A tornado can result if the updrafts and the interme- would be obscured by the Earth's curvature except at
diate rotation are strong. As a strong thunderstormis pro- nearby ranges.
duced, the misocycloneweakens, and its diameterincreases The possibility exists that certain large-scalefeatures of
when precipitation-induceddowndraft replacesupdraft in the storm environment may provide some basisfor a gener-
the vortex vicinity. alized warning of misocyclonic tornadoes. Some of the
Warnings of tornadoesassociatedwith misocyclonesare tornadoesoccur in conditionsof largethermal instabilityand
limited for several reasons. Burgess and Donaldson [1979] near boundariesdetected on synoptic surface maps. The
called attention to several instances of tornadoes associated Newkirk, Oklahoma, tornadoesof April 17, 1978,occurred
with Oklahoma misocyclones. In each case the tornadoes in a storm developingrapidly at the intersectionof a dry line
formed during the early developmentphase of the storm, and an overtaking cold front. Photographspresented by
offering no opportunity,even under the best of viewing Donaldson and Burgess [1982] show that both boundaries
conditions,to verify rotationin the incipientmisocyclonein were detected as clear-air echoes (Figure 13). The
214 TORNADO DETECTION AND WARNING BY RADAR

JUNE 30 3.4. Conventional Radar

Range We offer a short treatment of tornado warningswith

from CP-..• conventionalradar (reflectivity) signatures.This is justified
because conventional radar is still the standard for most of
the operationalcommunityand becauseproperuse of Dop-
pler radar involvescombinedinterpretationof reflectivityas
well as velocity signatures.
Sometimes the shape of reflectivity contours of radar
echoes may provide a distinctive clue to the dynamics of
processesoccurringwithin storms (see Figure 4b). A hook-
shaped appendage to a convective echo, the so-called tor-
nado hook, has attracted the attention of forecasters since
the early reports by Stout and Huff[1953] and Fujita [1958].
A number of years later, Forbes [1981] provided a defin-
itive study of the reliability of the hook echo as a tornado
•_,•OdBz , I I TORNAO0 signature. He found a moderate false alarm ratio but only a
,•1,, , .... ,
I ' ,J,,,.,,d,,.._ low probability of detection of a hook echo in a tornadic
1420 MST 1440
storm. Furthermore, the hook was absent three quarters of
Fig. !1. Time-height plot of misocyclone shear on June 30, 1987. the lifetime of a tornadic storm and often made its first
Gray lines are retlectivity; dashed line is threshold shear. The appearance after a tornado had already started. Forbes
tornado is indicated by the solid bar at bottom. (From Wakimoto found somewhat better results with so-called "distinctive
and Wilson [1989].
echoes." Clearly, hook echoes are not very helpful as a
tornado warning indicator, though their presence may serve
tion of the boundariesserved as a preferred location for the to reinforce the more accurate and timely predictors achiev-
strongest convective activity. able through Doppler radar.
Wakimoto and Wilson Jl989] reported on 17 misocy- The bounded weak echo region (BWER) is another dis-
tinctive reflectivity feature that can be readily detected with
clones, none of them related to a supercell, during a Colo-
conventional radar. The BWER, when observed for more
rado field prqject in the summer of 1987. All of the vortices
than a brief interval, is a reliable indicator of a supercell, a
were located on radar-detected convergence boundaries,
and eight of them occurred netJrthe intersection of two or
storm type that produces the overwhelming majority of
more boundaries, similar to the situation of the Newkirk,
strong tornadoes and nearly all of the violent ones. In
horizontal section the BWER is characterized by a region of
Oklahoma, storm. The misocyclone signatures were de-
diminished reflectivity entirely surrounded and capped by
tected at ranges averaging 25 km, with a maximum range
t'rom the radar of 45 km (22 rim). Interestingly, there was
higher reflectivity. In vertical section, the highest echo top
above the BWER marks the location of the intense, sus-
some lead-time as the radar signatures preceded the obser-
tained updraft required for a supercell [Browningand Lud-
vation of a visual vortex by an average of 14 min. The nearby
lam, 1962].
locations and a radar capable of detecting weaker echoes
The value of the B WER as a forecasting aid in warning of
may explain the lead-time differencesfrom the Oklahoma
observations.
severe and tornadic storms was fully appreciated by Lemon
[1980] after his study of many Oklahoma storms.He devised
Although misocycIonetornadoesare generally weak and
an observingschemeto aid operationalforecastersto rapidly
short-lived, they are capable of inflictinginjuries or fatalities,
identify storm types incorporatingthe BWER. His criteria,
and deserve our best efforts to warn of their appearance.
or variations thereof, based on three-dimensional reflectivity
Unfortunately, the study by Wakimoto and Wilson suggests
structure, are considered to be the most reliable indicators of
that a large minority of nonsupercelltornadoes may be
tornado likelihood during the pre-Doppler operationalera, at
undetectableby radar, with the detectable ones limited to a
least in environments supportive of supercell formation.
maximum rangeof around 50 km (25 nm), and with marginal
warning times in many cases. However, these authors re-
mind us that convergenceboundaries,the spawninggrounds 3.5. Estimated Skill
of nonsupercelltornadoes, are often detected by radar to
ranges of 100 km (50 nm) or more, and they suggest An estimationof radar's potential skill in tornado warn-
monitoring boundaries displaying strong horizontal shear, ingsis given.Thesesuggestionsfor operationalperformance
of Dopplerweatherradaremanatefrom pseudo-operational
those with wavelike inflections, and intersections of bound-
aries as locationsof possibletornadogenesis.We agree and tests performedover the last decadeand includeJDOP,
offer the further suggestionthat regionsin boundarieswith Doplight(Dopplerandlightningdataapplication[Forsythet
the greatestreflectivitymay offeran alternativeandperhaps al., 1989]), CINDE (convective initiation and downburst
more detectable means for indicating strong convergence. experiment[Wakimotoand Wilson 1989]),and
 4 km-- t = 0 rain

3 km-- No clouds

Weak,
low-level
circulation
atintersection
////?//////////?/////////////////////////?////,//

t = 15 min

New convective development

4 km•

3 km--

Ikm-- I Circulation
advected
vertically
///////?'/////•////////////////////////?///////'?/

t = 30 rain

Rapid convectivegrowth

4 km --

3 km --

2 km --
Strong convectiveupdrafts stretch and
intensifyvortex into F1 tornado

1 km --

2'////•///////////////'/'•//X/,?;?////:.,'>////'2'//////?/'•///

Fig. 12. Conceptual

modelof misocyclone
andlandspout
tornadoevolution.
[fromBrad),andSzoke,
216 TORNADODETECTION AND WARNING BY RADAR

Fig. 13. Dopplerrelectivityat 1357CST on April 17, 1978[fromDonald,son

and Burgess,1982].One thin line (dry
line) oriented north-south and a second thin line (cold front) oriented northeast-southwestare clear-air returns.
Thunderstormechoesare •een north of dry line/front intersectionpoint. Rangemark is 115km.

IOT&E (initial operational test and evaluation, 1989; Wey- the entire radar coverage interval, but it is low for TVS and
man and Clancy [1989; also unpublished results, 1991]). misocyclonesignaturesbecausethey can only be seenin less
Radar of course is just one input into warning decision than a quarter of the radar coverage area. POD is moderate
making, but the skill of that part of the warning processcan for the hook echo because of its occasional, relatively long
be estimated. Also, warning skill will be discussed for range of observation and close association with mesocy-
operational radars (such as the WSR-88D), and skill scores clones. The hook echo, however, has a high false alarm ratio
will reflect values for the entire coverage range of'the radars
(assumed to be 230 km (125 nm)). Further, it is assumed that
the radars meet network specifications (i.e., 10-cm wave- TABLE 1. Estimated Tornado Warning Skill Scores (Network
length, narrow beam width, high sensitivity, elaborate real- Coverage of 125 nm)
time processing, etc.). Finally, it should be known that Predictor POD FAR CSI
better performance (higher skill) will be achieved for local
use of the radars, such as protecting a nearby metropolitan Hook echo M H L
area.
TVS L L L
Mesocyclone H M M
Table I lists estimated skill for the most commonly used Misocyclone L M L
radar signatures.The probability of detection (POD) is high
for mesocyclone signatures because they can be seen over L, low; M, medium; H,
 BURGESS ET AL. 217

(FAR) owingto the high degreeof subjectivity in correct

identification.Mesocycloneand misocyclone signatures
havemoderate FARsbecause of the significant
percentage
of parentvorticesthat do not spawntornadoes. Onlythe
TVS, as the most direct indicatorof a tornado,has a low
FAR.

The resultantcritical successindex (CSI [Donaldsonet

al., 1975]) for radar warning of tornadoes is low for all
signaturesexcept mesocyclones,where a moderateCSI is
estimated.We feel that with today'sradartechnology and
our current level of understanding a high CSI is unlikely
exceptfor violenttornadoes.Higherlevelsof success may
be achievedthroughthe combinationof radar signatures
with accurateand timely spotterreports.
Of course,it is disappointing
that manytornadoes will go
undetected, even with Doppler radar, but all missesshould
not be treatedthe same.Most tornadoesmissedby Doppler
radar are likely to be weak, small,and short-lived(F0-F1 on
the Fujita scale).Strongtornadoes(F2-F3) and,particularly,
violenttornadoes(F4-FS) seemto requirea Dopplerdetect-
able signaturefor their existence.Also, a potentialbenefitof
widespread Doppler use will be a reduction in FARs for
tornadowarnings.Some measuresof pre-Dopplerskill [e.g.,
Kelly and Schaefer, 1982] indicate FARs as high as 80%.

4. AUTOMATED RADAR ALGORITHMS

Fig. 14. (a) Single-Dopplermesocyclonecoupletand (b) compan-
Algorithmsfor automatedanalysisand detectionof severe ion pattern vectors.
weather phenomena such as TVSs and mesocyclonescan
significantly assist the forecaster if they are reliable, but
these algorithms are likely to be dismissed if they are The basic steps of the mesocyclone and TVS algorithms
generally perceived to be ineffective [Walker and Heideman, are the identification of pattern vectors, the association of
1989]. Reliability is utmost. Technically, an algorithm must vectors from the same elevation scan into two-dimensional
be sufficiently robust to addressthe problem at hand while features, the association of features from differing eleva-
its computation load and memory requirements are kept to a tions, and the evaluation of features for meteorological
minimum. An algorithm must have predictable response to significance.Although the general frameworks of the algo-
permit operation in real time. rithms to be discussedare similar, the intermediate steps are
Hennington and Burgess [1981] developed a viable ap- quite distinct. The following subsectionswill addresssome
proach for automatedvortex detection that hasbeen adopted of the primary considerations and differences among the
in general form in most subsequentTVS and mesocyclone approaches.
algorithms. In the Rankine combinedvortex usedto model a
mesocyclone, the relative velocity along the radar beam
4.1. Mesocyclone Algorithtns
changesfrom one sign to the other across the mesocyclone
core, and therefore the sign of the velocity gradient remains The ideas presented by Hennington and Burgess [1981]
constant across the core. Segmentswhere the sign of the were implementedinto a working mesocyclonealgorithmby
velocity gradient is constant are referred to as pattern Zrni• et al. [1985b], and the technique was provisionally
vectors (Figure 14). Since the mesocycloneis essentiallya selectedfor use in NEXRAD. At the heart of this algorithm
rotational phenomenon,pattern vectors are best definedin is the establishmentof criteria to help restrict the detection
the azimuthal direction. of pattern vectors to mesocyclones,which has positive
The beginning and ending points of a pattern vector implications for memory requirements and processing
representrelative maxima of incomingand outgoingveloci- speed. Thresholds of shear and momentum(length times
ties. For a mesocycloneor TVS the velocity maxima are velocity, which is more precisely momentumper unit mass)
located at the core radius. Since we are interested only in are used to identify likely mesocyclone-associated pattern
defining the core, it is sufficientto retain just the pattern vectors.

vector endpointsand velocity maxima. Patternvector anal- Pattern vectors from the same elevation scan are associ-
ysis techniquesthereforerepresentan efficientand concise ated into features based on a general proximity condition.
methodologyfor extraction of well-behavedphenomena. Feature parameters like size, shear, and momentum
218 TORNADO DETECTION AND WARNING BY RADAR

determined from the collective values of all the pattern Given this, Canadian mesocyclonesappear to provide simi-
vectors that compose a feature. Values are momentum lar lead times to their American counterparts, 20 or more
weighted to bias toward the greatest values of a feature. minutes from the time of first mesocyclone detection to
Features satisfyinga shape and size requirementare identi- incipient tornado formation.
fied as mesocyclonelike, having already satisfied pattern
vector criteria 'for mesocyclones. Features from different
4.2. TVS Algorithms
elevations are associatedtogether into the same feature if
they overlap. Finally, a mesocycloneis identified when two The provisionalTVS algorithmfor NEXRAD [Forsyth,
or more elevations of a vertical feature contain a mesocy- 1984] was designed to operate in conjunction with the
clonelike feature. NEXRAD mesocyclone algorithm. Any detected mesocy-
In a new algorithm developed by Destochefs [199!], clones are reexamined for regions of large, TVS-like shear
thresholds of shear and velocity difference are used for (0.05 s-i). When strong,rotationallyindicativeshear is
pattern vector retention.This is done to overcomethe size detected at multiple elevations of a mesocyclonefeature, a
restrictionimposedby momentumthresholding,permitting TVS is declared. Currently, the technique detects only
concurrent detection of TVSs. cyclonically rotating features.
Following the lead of Wieler [1986] in an earlier mesocy- A shortcomingof the currentNEXRAD TVS algorithmis
clone algorithm, resolutioncorrectionsare applied to the its restrictionof TVS detection to supercell-typetornadoes.
pattern vector thresholdsin order to range normalizethe Nonsupercelltornadoes,like thosecommonlyobservedin
detectioncriteria. This approachis basedon work by Brown Colorado [Wilson, 1986; Wakimoto and Wilson, 1989], will
and Lemon [1976.1illustrating how a Rankine combined escapedetectionwith sucha technique.Albers [1991] has
vortex is viewed with varying radar resolution. Resolution subsequentlymodified the NEXRAD mesocyclonealgo-
corrections for vortex size and rotational velocity can be rithm to detect these phenomena. For example, pattern
applied if the radar beam width is no greater than the vector thresholds of shear have been increased while those
mesocyclone coreradius.As resolutiondegradesfurtherthe of momentum have been reduced, and an additional require-
necessarycorrectionbecomesambiguous.For an average ment for velocity differenceacrossthe pattern vector has
size mesocyclone(core diameterof 5 km), rangenormaliza- beenaddedto acceptthe patternvectorsof weakTVSs that
tion is possibleto rangesof about 150km (75 nm) assuming momentum thresholdingwould eliminate. These changes
illumination by a 1ø beam. have resulted in detection of some Colorado tornadoes,
Mesocycloneparameterssuchas size and rotationalve- particularly
at closerange.Suchresultsattestto theinherent
locity are determinedfrom a weightedaverageof pattern tunablenessof the original NEXRAD design.
vectorsby usingonly selectedpattern vectorscentered A new TVS algorithmis under developmentat NSSL
aroundthe peakvelocitiesof a feature.The techniqueallows [Vasi!off, 1991;this volume]for possibleinclusionin the
the degreeof rotationand divergenceassociatedwith any TerminalDopplerWeatherRadarProgram(TDWR). Pattern
particularmesocyclone signatureto be determined. vectorsare utilized,but only adjacentazimuthsare consid-
Three-dimensional(3-D) featuresare constructedby find- ered. Range-consecutive pairsof gatesare requiredto meet
ingthe "best" combination of featuresfromdifferingeleva- certainvelocitydifferencethresholds,and three elevation
tions.This processconsidersthe distancebetweenfeatures anglesmustbe verticallyassociated to verifya TVS. No
aswell astheirenergyin ERKE. The objectiveis to identify alignmentwith a mesocyclone signatureis necessary for
the mostenergetic3-D features.A 3-D featureis allowedto identification.
slopeasmuchas45øto account for actualtilt andto account The Vasiloffalgorithmhasbeentestedon tornadoes from
for feature movement during the volume scan period. A differenttype storms, occurring in differentparts of the
trackingroutineassociates featuresfrom consecutive vol- country. Results thusfar are somewhat encouraging in that
ume scans. Mesocyclonesare identifiedby applyingthe both nonsupercell tornadoes(gustfront tornadoes,land-
classificationcriteria first proposedduring JDOP. spouts,etc.)andsupercell tornadoes havebeendetected.
One significant task that remainsfor NEXRAD is the However, as expected, detectability
decreases asa function
regionalization of algorithms. This is especially true for of reduced tornado size and range from the radar. Also,
detection.The NEXRAD provisionalalgo- rapidscanning
mesocyclone is necessary
to detectshort-lived
tornadoes,
rithm and the Desrochers'algorithmwere designedentirely and even then, some are missed.
aroundOklahoma-type
supercellstorms.Characteristics
of
shearand verticalextentare likely to vary significantly
from 4.3. Tornado Prediction Algorithms
region to region.
One exampleof site sensitivity
is Ontario,Canada.Joe Algorithms canreducethetediumof radarmonitoring and
[1989]hasfoundthatthe NEXRAD mesocyclone algorithm serve to call attention to developingsituations.They can
withtheprescribed
Oklahoma thresholdsis notsensitive
to alsoprovide useful quantitative
assessments as well. One
thegeneralCanadian mesocyclone. Patternvectorthresh- suchapplication
is the use of numerical
outputof the
olds of shear and momentum were reduced by as much as mesocyclonealgorithmfor tornadoprediction.ERKE, a
75% of their Oklahoma value to detect these phenomena. mostpromising
identifier
of tornadic
mesocyclones,
can
 BURGESS ET AL. 219

50 it was with the beginningof the Dopplerradarresearchera at

the time of Tornado Symposium II. As described in the
preceding sections, the implementation of the NEXRAD
network will bring vast changesto the use of operational
radar in tornado warnings. The new WSR-88D radars will
feature rapid, three-dimensionalscanningand will bring
powerful new signatures and algorithms [AlberO' et al.,
•o 30
1991]. Plate 2 gives examplesof useful WSR-88D products
from a recent outbreak of violent tornadoes in Oklahoma and
/%
Kansas (April 26, 1991).
• 2o The advent of the NEXRAD network will aid warning
forecastersgreatly, but certain problem areas must be over-
,•',1
Momentum come to achieve maximum potential. Algorithm improve-
o _1o ments are needed in several areas. Adaptation of algorithms
to regional and seasonal differences in storm and tornado
climatologies is needed. Overall algorithm immaturity, in-
cluding the inability to detect consistentlythe existence of all
hazards, will guide the near-future direction of applied
research and development.
radius (kin)
In the longer term, an increase in our understandingof
Fig. 15. The relationshipsamong some mesocycloneevaluators
thunderstorm morphology and tornado formation processes
usingmean valuesof mature mesocycloneintensity. are essentialto improved data interpretation and new algo-
rithm creation. New conceptualmodels of tornadic weather
systems are expected from analysis of WSR-88D data.
readily obtainedfrom the standardoutput of the mesocy- NEXRAD will revolutionize our nowcasting capability,
clone algorithm. but this technologyis not singularly sufficientto provide for
Besides the initial performance capability for tornado all our detection and warning needs. Warning/forecasting
prediction demonstrated thus far [Donaldson and Dearo- workstationsthat combine radar outputs with other weather
chers, 1990], ERKE has certain design attributes that make data are needed. Because we know that the surrounding
it attractive for algorithmic use. ERKE is conceived as a environment is related to storm type and tornado potential,
single-parameter function that relates tornado likelihood to diagnosis of a storm's mesoscale environment will lead to
one measure of energy. It is a continuous function that earlier and more reliable warnings.
evolves gradually in magnitude between tornadic and non- Improvements in interaction between warning forecasters
tornadic mesocyclones.It is readily tunable to varying storm and the preparedness community (emergency management
structure. A comparison of ERKE with velocity, shear, and officials, storm spotters, the media, etc.) will result in better
momentum is shown in Figure 15 for a typical mature warnings. For example, there is no plan for Civil Defense
mesocyclone as defined by Burgess et al. [1982].
Recently, the ERKE evaluation technique has been auto-
mated in the Dearochers [1991] mesocyclone algorithm. Median Lead Time to First Strong Tornado

Dearochers and Donaldson [1992] tested it as a tornado provided by ERKE

probability algorithm on 23 Oklahoma mesocyclonicstorms.
They found the ERKE magnitude of tornadic mesocyclones
distinctive from other mesocyclonesat the 99.9% level of • 4o

significance. In this limited sample, ERKE provided a me-

dian 27 min lead time for the first strong tornado (F2-F3) of
a storm with no errors. ERKE performancestatistics(Figure
16) show that there is a trade off between lead times above 27 • 20
min and false alarms. What is not shown in the figure is that
the thresholdrange of ERKE necessaryto providepositive
lead times for tornadoes with small errors is large enough so
that this algorithmshouldnot be overly sensitiveto viewing
angle variations. 0.0 0.1 0.2 0.3 0.4 0.5

Combined Error Rate (H + FA)

5. FUTURE DIRECTIONS
Fig. 16. Median lead time to the first strongtornado provided by
The timing of Tornado SymposiumIII in 1991finds the excessrotationalkinetic energy comparedto errors composedof
weather radar community on the verge of a new era, just as misses {M) and false alarms
220 TORNADO DETECTION AND WARNING BY RADAR

directorsto view the sameradardisplaysthat forecasters NSSL-86, 84 pp., Natl. Oceanicand Atmos. Admin., Boulder,
use, and forecasterscan neither see spotterlocationson Colo., 1979, (Available as NTIS PB80-lO7/88/AS from Natl. Tech.
theirradardisplaysnor participatein directingthe spotters Inf. Serv., Springfield,Va.)
to neededareas. Similar problemsrelate to interactionswith Burgess,D. W., V. T. Wood, and R. A. Brown, Mesocyclone
evolution statistics,in Preprints, 12th Conferenceon Severe
thebroadcastcommunity.Radarisjust a portionof thetotal Local Storms,pp. 422-424, AmericanMeteorologicalSociety,
warningsystemnecessary to adequatelyprotectthe public; Boston, Mass., 1982.
all portionsof the systemmustfunctiontogetherif lives are Burgess,D. W., D. L. Andra, Jr., and W. F. Bunting, The
to be saved. Stillwater,Oklahomatornadoof May 15, 1990:A casestudyof
the use of radar in the modernizedNational Weather Service, in
Preprints,25th InternationalConferenceon Radar Meteorology,
pp. 51-54, American MeteorologicalSociety, Boston, Mass.,
Acknowledgments.Contributionby Ralph Donaldsonto this 1991.
manuscriptwas supportedby the United StatesAir Force,Phillips Desrochers,P. R., Automatedmesocyclonedetectionand tornado
Laboratory GeophysicsDirectorate, under contract F19628-90-C-
0088. We thank Jim Wilson, Arthur Witt, and Steve Vasiloff for forecasting,Tech.Rep. 91-2051,168pp., PhillipsLab., Hanscom
Air Force Base, Md., 1991.
manuscriptreview and .loan Kimpel for graphicspreparation.
Desrochers, P. R., and R. J. Donaldson, Jr., Automatic tornado
predictionwith an improvedmesocyclonedetectionalgorithm,
REFERENCES Weather Forecasting, in press, 1992.
Desrochers,P. R., R. J. Donaldson, Jr., and D. W. Burgess,
Albers, S.C., The detectionof isolatedtornadic vortex signatures Mesocyclone rotational kinetic energy as a discriminator for
with NEXRAD algorithms,in Preprints, 25th International Con- tornadicand non-tornadictypes, in Preprints,23rd Conferenceon
.[['t'•'ttr't'oft Radar Meteorolog,y, pp. 123-126, American Meteoro- Radar Meteorology,pp. 1-4, American MeteorologicalSociety,
logical Society, Boston, Mass., 1991. Boston, Mass., 1986.
Alberty, R. A., T. Crum, and F. Toepfer,The NEXRAD program: Donaldson,R. J., Jr., Vortex signaturerecognitionby a Doppler
Past, presentand future; a 1991perspective,in Preprints,25th radar, J. Appl. Meteorol., 9, 661-670, 1970.
lt•tcrnational Con./•'renceon Radar Meteorology, pp. 1-8, Amer- Donaldson, R. J., Jr., Observations of the Union City tornadic
ican Meteorological Society, Boston, Mass., 1991. storm by plan shear indicator, Mon. Weather Rev., 106, 39-47,
Armstrong, G. M., and R. J. Donaldson, Jr., Plan shear indicator for 1978.
real-time Doppler radar identification of hazardous storm winds, Donaldson, R. J., Jr., and D. W. Burgess, Results of the Joint
J. Appl. Meteorol., 8, 376-383, 1969. Doppler Operational Project, in Proceedings of the NEXRAD
Atlas, D., Radar analysis of severe storms, Meteorol. Monogr., 5, Doppler Radar Symposium, CIMMS, University of Oklahoma,
177-220, 1963. Norman, pp. 102-!23, 1982.
Bluestein, H. B., The formation of a "landspout" in a "broken line" Donaldson, R. J., Jr., and P. R. Desrochers, Doppler radar esti-
squallline in Oklahoma, in Preprints, 14th Con.[krenceon Severe mates of the rotational kinetic energy of mesocyclones,in Pre-
Local Storms, pp. 267-270, American Meteorological Society, prints, 14th Conference on Severe Local Storms, pp. 52-55,
Boston, Mass., 1985. American Meteorological Society, Boston, Mass., 1985.
!3rady, R. H., and E. J. Szoke, A case study of nonmesocyclone Donaldson, R. J., Jr., and P. R. Desrochers, Improvement of
tornado development in northeast Colorado: Similarities to wa- tornado warnings by Doppler radar measurementof' mesocyclone
terspout formation, Mon. Weather Rev., 117, 843-856, 1989. rotational kinetic energy, Weather Forecasting, 5,247-258, 1990.
Brown, R. A., and L. R. Lemon, Single-Doppler radar vortex Donaldson, R. J., Jr., G. M. Armstrong, A. C. Chmela, and M. J.
recognition. Part II: Tornadic vortex signatures,in Preprints, 17th Kraus, Doppler radar investigation of air flow and shear within
Radar Meteorological Con.l•krence,pp. 104-109, American Mete- severe thunderstorms, in Preprints, 6th Conference on Severe
orological Society, Boston, Mass., 1976. Local Storms, pp. 146-154, American Meteorological Society,
Brown R. A., L. R. Lemon, and D. W. Burgess, Tornado detection Boston, Mass., 1969.
by pulsed Doppler radar, Mon. Weather Rev., 106, 29-38, 1978. Donaldson, R. J., Jr., R. M. Dyer, and M. J. Kraus, An objective
Browning, K. A., and F. H. Ludlam, Airflow in convective storms, evaluator of techniquesfor predicting severe weather events, in
Q. J. R. Meteorol. Soc., 88, 117-135, 1962. Preprints, 9th Conference on Severe Local Storms, pp. 321-326,
Burgess, D. W., Single-Doppler radar vortex recognition: Part I: American Meteorological Society, Boston, Mass., 1975.
Mesocyclone signatures, in Preprints, 17th Conference on Radar Doviak, R. J., and D. S. Zrnid, Doppler Radar and Weather
Meteorology, pp. 97-103, American Meteorological Society, Bos- Observations, 458 pp., Academic, San Diego, Calif., 1984.
ton, Mass., 1976. Dunn, L. B., Two examples of operational tornado warnings using
Burgess, D. W., and R. J. Donaldson, jr., Contrasting tornadic Doppler radar data, Bull. Am. Meteorol. Soc., 71,145-153, 1990.
storm types, in Preprints, 11th Con.[krence on Severe Local Forbes, G. S., On the reliability of hook echoes as tornado indica-
Storms, pp. 189-192, American Meteorological Society, Boston, tors, Mon. Weather Rev., 109, 1457-1466, 1981.
Mass., 1979. Forbes, G. S., and R. M. Wakimoto, A concentrated outbreak of
Burgess,D. W., and L. R. Lemon, Severe thunderstormdetection tornadoes, downbursts and microbursts, and implications regard-
by radar, in Radar in Meteorology, edited by D. Atlas, pp. ing vortex classification,Mort. WeatherRev., 111,220-235, 1983.
619-647, American Meteorological Society, Boston, Mass., 1990. Forsyth, D. E., TVS detection algorithm description,NEXRAD
Burgess,D. W., and L. R. Lemon, Characteristicsof mesocyclones Algorithm Report, NEXRAD Joint Sys. Prog. Office, Washing-
detectedduring a NEXRAD test, in Preprints, 25th International ton, D. C., Dec. 1984.
Conference on Radar Meteorology, pp. 39-42, American Meteo- Forsyth, D. E., D. W. Burgess,M. H. Jain, and L. E. Mooney,
rological Society, Boston, Mass., 1991. DOPLIGHT '87: Application of Doppler radar technologyin a
Burgess,D. W., L. R. Lemon, and R. A. Brown, Tornado charac- National Weather Service Office, in Preprints, 24th Conference
teristics revealed by Doppler radar, Geophys. Res. Lett., 2, on Radar Meteorology, pp. 198-202, American Meteorological
183-184, 1975. Society, Boston, Mass., 1989.
Burgess,D. W., R. J. Donaldson,Jr., T. Sieland,andJ. Hinkelman, Fujita, T. T., Mesoanalysisof the Illinois tornadoesof 9 April 1953,
Final report on the Joint Doppler Operational Project (JDOP). J. Meteorol., I5, 288-296, 1958.
Part I: MeteorologicalApplications.NOAA Tech. Memo. ERL Fujita, T. T., Objectives,operations,and resultsof Project
 BURGESS ET AL. 221

ROD, in Preprints,11th Conference

on SevereLocalStorms,pp. Wakimoto, R. M., and J. W. Wilson, Non-supercelltornadoes,
259-266. AmericanMeteorologicalSociety, Boston,Mass., 1979. Mon. Weather Rev., 117, 1113-1140, 1989.
Fujita, T. T., Tornadoesand downbursts in the contextof general- Walker, D.C., and K. F. Heidman, DARE-I evaluation: Forecast-
ized planetary scales,J. Atmos. Sci., 38, 1511-1534, 1981. er's assessmentand use of the NEXRAD algorithm products
Hennington,L. D., and D. W. Burgess,Automatedrecognitionof during the 1987and 1988 warm seasons,NOAA Tech. Rep. ERL
mesocyclones from single-Doppler data, in Preprints,20th Con- FSL 3, 22 pp., Natl. Oceanic and Atmos. Admin., Boulder, Colo.,
ference on Radar Meteorology,pp. 704-706, AmericanMeteoro- 1989.
logical Society, Boston, Mass., 1981.
Weyman, J. C., and K. I. Clancy, NEXRAD initial operationaltest
Joe, P. I., The King City mesocyclone detectionalgorithm,report, and evaluation phase II (IOT&E2), final report, AFOTEC Rep.
46 pp., Atmos. Environ. Serv. of Can., ARPP, Toronto, Ont.,
1989. 0167, 68 pp., Air Force Test and Evaluation Center, Kirtland Air
Force Base, N. M., 1989.
Kelly, D. L., and J. T. Schaefer,Implicationsof severelocal storm
warning and verification,in Preprints, 12th Conferenceon Severe Wieler, J. G., Real-time automated detection of' mesocyclonesand
Local Storms, pp. 459-462, American Meteorological Society, tornadic vortex signatures, d. Atmos. Oceanic Technol., 3, 98-
Boston, Mass., 1982. 113, 1986.
Kraus, M. J., Doppler radar observations of the Brookline, MA Wilson, J. W., Tornadogenesis by nonprecipitation-inducedwind
tornado of 9 August 1972, Bull. Am. Meteorol. Soc., 54, 519-524, shear lines, Mon. Weather Rev., 114, 270-284, 1986.
1973. Wilson, J. W., R. E. Carbone, H. Baynton, and R. J. Serafin,
Lemon, L. R., Severe thunderstormradar identificationtechniques Operational applications of meteorologicalDoppler radar, Bull.
and warning criteria, NOAA Tech. Memo. NWS NSSFC-3, 60 Am. Meteorol. Soc., 61, 1154-1168, 1980.
pp., Natl. Oceanic and Atmos. Admin., Boulder, Colo., 1980. Wood, V. T., and R. A. Brown, Single Doppler velocity signatures:
(Available as NTIS PB-273049 from Natl. Tech. Inf. Serv., An atlas of patterns in clear air, widespread, precipitation, and
Springfield, Va.) convective storms, NOAA Tech. Memo. ERL NSSL-95, 71 pp.,
Petit, P. E., Doppler radar detection capabilitiesat Montgomery, Natl. Oceanic and Atmos. Admin., Boulder, Colo., 1983. (Avail-
Alabama, 1982-1988, Natl. Weather Dig., 15, 23-28, 1990. able as NTIS PB84155779from Natl. Tech. Inf. Serv., Springfield,
Stout, G. E., and F. A. Huff, Radar recordsIllinois tornadogenesis, Va.)
Bull. Am. Meteorol. Soc., 34, 28!-284, 1953. Zrni(•, D. S., D. W. Burgess, and L. D. Hennington, Doppler
Vasiloff, S. V., The TDWR tornadi½ vortex signature detection spectra and estimated windspeedof a violent tornado, J. CIim.
algorithm, in Preprints, 4th International Conferenceon Aviation
Appl. Meteorol., 24, 1068-1081, 1985a.
and Weather Systems, pp. J43-J48, American Meteorological
Zrnid, D. S., D. W. Burgess, and L. D. Hennington, Automatic
Society, Boston, Mass., 1991.
detection of mesocyclone shear with Doppler radar, J. Atmos
Vasiloff, S. V., Single-Doppler radar study of a variety of tornado
Ocean. Technol., 2,425-438,
types, this volume.
 Single-DopplerRadar Studyof a Variety of Tornado Types
STEVEN V. VASILOFF

National SevereStormsLaboratory, Norman, Oklahoma 73069

1. INTRODUCTION landspouts,some of which were rated F2. They showed that

these vortices form when preexisting vorticity along surface
Tornadoes occur in a spectrum of sizes and intensities boundaries becomes entrained into and stretched by thun-
[e.g., Forbes and Wakimoto, 1983;Dosa,ell and Burgess, derstorm updrafts.
this volume]. While the largest and most violent garner the The largest and most violent tornadoes are produced by
most attention, the predominance of tornadoes are weak to supercell thunderstorms [Bro•vning, 1964]. Strong midlevel
moderate in intensity. Herein, Doppler radar velocity signa- rotation in the updraft regions of these thunderstorms(i.e.,
tures of tornadoes having a wide variety of sizes and mesocyclone) often precedes tornado formation by more
intensities are examined. Attention is focused on radar than 30 rain [Burgess, 1976]. Brown et al. [1978] found that
characteristicsof each tornado and their relationshipto the the tornadic circulation for the Union City, Oklahoma,
tornado's visual appearance and associateddamage. tornado formed at midlevels and lowered with time.
Background material is provided which discussesforma- When a tornado is probed by a scanningpulsed Doppler
tion mechanismsfor various tornado types. Following pre- radar, a tornadic vortex signature (TVS) may be revealed. A
sentation of the observations, attributes of the Doppler TVS is characterized by velocity extrema of opposite signs
signatures will be related to their detectability by a TVS (after removal of a motion vector) at adjacent azimuths with
detection algorithm being developed at the National Severe the resolution volumes at the same range as the tornado. In
Storms Laboratory [Vasilo•; 1991]. studiesof supercell tornadoes, Bro•vn et aI. [ 1978]found that
the theoretical peaks in a TVS were one beamwidth apart.
This is primarily the result of the tornado diameter being
2. BACKGROUND
smaller than the radar beamwidth (see Burgess et al. [1991]
Tornado characteristics can be attributed largely to how for a review of Doppler limitations). Burgess and Lemon
they are formed. Near the low end of the spectrumare small [1990] required that the TVS shear exceed the surrounding
spin-ups along thunderstorm outflow boundaries.These mesocyclonic shear by a factor of 5. For Colorado land-
spin-ups are often referred to as gust front tornadoes(or spouts, Wakimoto and Wilson [1989] defined a TVS as
"gustnadoes") and can cause F1 damage [Fujita, 1979; approximately 6 times the surrounding "misocyclonic"
1981]. Their vertical extent is limited to the lowest 1-2 km. shear. For the purposesof this study, we follow the Bro•vn et
In a dual-Doppler radar study of the Del City, Oklahoma al. [1978] definition of a TVS, which includes time and height
tornado, Brandes [1978] showed that the signature of a continuity, but do not require that the shear be of any certain
gustnadowas limited to the lowest 2 km and had no closed value.
circulation field.
3. OBSERVATIONS
Higher up on the spectrum are waterspoutsand water-
spoutliketornadoesoccurringon theHigh Plains,sometimes The observationspresented in this section are representa-
referredto as "landspouts." Wakimotoand Wilson[1989] tive of a larger set of data that has been studied. Following
andBrady and Szoke[ 1989]documenta numberof Colorado the discussionin section 1, the data have been grouped into
three categories: gust front tornadoes (or "gustnadoes"),
The Tornado: Its Structure,Dynamics,Prediction,and Hazards.
landspouts, and supercell tornadoes. For reasons to be
GeophysicalMonograph 79
in 1993by the shownlater, the supercelltornadoesare further divided into
Thispaperis notsubjectto U.S. copyright.Published
American GeophysicalUnion. brief supercelland long-lived supercelltornadoes.
224 SINGLE-DOPPLER RADAR STUDY OF TORNADO TYPES

l'•ig. !. Photographof two gustfront tornadoesalonga thunderstormoutflowboundaryin westOklahomaon August

21, 1979 Icourtcsy of G. Moore).

3. !. Guxtnadoes into a wind speedestimateof 33-50 m s-•. Unfortunately,

Three tornadoes were associated with a brief supercell timing on the damage for this event could not be ascertained
storm that occurred on May 8, 1986. The first one was a during the damage survey. Placement of the damage path
small spin-up along a gust front on the right tlank of the arises from when the signature was near the affected area
storm. A brief touchdown occurred near 2301 UTC in The and is in keeping with the timing observed for the gustnado
Village near Oklahoma City, 25 km from the radar. The
debris cloud and visual funnel appearance were very similar _

to the vortices shown in Figure 1 (D. W. Burgess, personal VILLAGE

communication, 1991). Only minor tree damage resulted, May 8, 1986
thus the tornado's F0 rating. The time-height profile of the
Doppler signature is shown in Figure 2. Data are from the
Cimarron Doppler radar, one of two research Dopplers
operated by the National Severe Storms Laboratory 1
(NSSL). The TVS was very short-lived and shallow, extend-
ing to only 1.5 km above ground level (AGL). The maximum
Av (velocity difference between two adjacent gates at the //20,\/
samerange)is 30 m s-•, nearthe time of the observed
funnel. A constant-elevation angle scan (also referred to as a [ l/!
plan position indicator (PPI)) of Doppler velocities at l.l ø 5
through the signature at 2303 UTC is shown in Plate 1. The I: \
Doppler velocity signature of the tornado can be seen at
about 70ø azimuth and 24 km range. Although the velocities II\q[ I--
are relatively weak, the pattern is readily discerniblein the
velocity field. I I !
Two other gustnadoes occurred in central Oklahoma on 22,5`5 2300 230,5
May 3, 1991.The evolution of the Doppler velocity signature TIME (UTG)
of one of them is shown in Figure 3. The data are from
NSSL's Norman Doppler. This particular gustnadoformed Fig. 2. Time-heightplot of Avs (in metersper second)from a gust
front tornado that occurred in The Village on May 8, 1986. Contour
near Blanchard, Oklahoma, about 20 km SW of Norman. intervalsare every5 m s-I. The hatchedbar alongthe bottom
MaximumAvs werelessthan25 m s-• althougha damage indicates times that the tornado was observed and contains damage
survey revealed Fl damage, which Fujita [1981] translates intensities based on the Fujita
 VASILOFF 225

2 ,-
6 -
BLANCHARD
STAPLETON
May 3, 1991
June 15, 1988
5- /
I 2O
4
v

20 15 10

0
OO55 0100 0105 0110 0115 0120
TIME (UTC) 2200 2205 2210 2215 2220 2225 2230 2235 2240
TIME (IJTC)

Fig. 3. Asin Figure2 exceptfortheBlanchard

gustnado
of May3,
1990.The placementof the hatchedbar is approximate. Fig. 4. As in Figure 2 except for the Stapleton tornadoof June 15,
1988. Fifteen and 45 m s -1 contourshave been added. The dashed
hatchedbar indicatestimes when only a funnel cloud was observed.
Damage intensities as a function of time are not available.
that occurredin The Village(i.e., nearthelarge/Xvgradient
at the beginningof the TVS). Since the low-level scanswere
separated by several minutes, the maximum winds in the tornado; i.e., visual appearances are not indicative of dam-
gustnadomay not havebeensampled.DopplervelocityPPIs age intensity levels.
(not shown) of the May 3 events reveal signaturesvery The secondtornado occurredin the southpart of Denver,
similarin horizontalsize and intensityto the signaturein The about 22 km from FL-2. This tornado was rated F3, unusu-
Village. ally highfor a landspout.The TVS time-heightprofile for this
tornado (Figure 6) is very similar to the Stapleton tornado's
3.2. Landspouts with a double maximum, one near the surface and another
directlyabove,althoughzXvsaloft were 5-10 m s-t less.
At least four landspoutsformed along colliding outflow However, in contrastto the Stapletontornado signature,the
boundaries near Denver, Colorado, on June 15, 1988. Data
strongerAvs at low levels appear to build upward with time.
are from the MassachusettsInstitute of TechnologyLincoln Plate 2 showsa Doppler velocity display at 2223 UTC. The
Laboratory's FL~2 Doppler radar, which was part of the Stapleton signature is WNW of the radar, while the Denver
Terminal Doppler Weather Radar (TDWR) 1988 real-time signatureis nearly due west of the radar. Both signaturesare
demonstration [Turnbull et al., 1989].
easily discerned and similar in size and intensity to the
The Stapleton tornado (rated F2) causedtemporary evac- signaturein The Village.
uation of nonessentialpersonnelfrom the control tower at
Stapleton international airport. The tornado touched down
about 20 km from FL-2 and moved slowly to within 17 km of 3.3. Brief-Supercell Tornadoes
the radar before it dissipated. Maximum Avs for this tornado The May 8, 1986, Edmond, Oklahoma, tornado (rated F3)
(Figure4) exceeded
45 m s-] at thelowestlevels,anda 60 began just alter the gustnado described previously dissi-
m s-] Av was observedat midlevels.A circulationaloft was pated. The tornado formed about 30 km from the NSSL
detected by the radar nearly 20 min before the tornado Cimarron radar and moved toward the ENE. Unlike the
began. Weaker rotation near the surface was also detected gustnadoesand landspouts, the Edmond tornado was asso-
(as shownby the 15 m s-] contour).Note that the Av ciated with a midlevel mesocyclone.The parent stormwas a
gradient near the surface increased when the funnel came in short-lived supercell which had a well-definedhook echo in
contact with the ground. Furthermore, it appears that the the reflectivity field. The tornado formed along a low-level
stronger Avs develop downward with time. This is contrary convergence boundary, away from the mesocyclone core.
to most of the landspoutsstudiedby Wakimotoand Wilson With time, the mesocyclone intensified while the tornado
[1989], where the vorticity developed from the ground up- seemedto migrate toward its center, with the TVS blending
ward. in with the mesocyclonecirculation; this is not a typical (at
A photographtaken at 2223 UTC (Figure 5), when the fi,vs leastfor the small samplein the literature) supercell-tornado
were weaker, shows a funnel whose appearance did not sequence of events.
changemuch over the lifetime of the event. When a video The time-heightprofile of the Edmond TVS (Figure 7) is
tape of the entire event was viewed, it often was difficultto very similar to the landspoutprofiles. There is a simulta-
distinguish between what was a funnel and what was a neous occurrence of a fi,v maximum at midlevels above
226 SINGLE-I)OP!'LER RADAR STUDY OF TORNADO TYPES

Fig. 5. Photograph of the Stapleton tornado taken at 2223 UTC (courtesy of B. Gallus).

maximum at lower heights. In this case, with the aid of a (Plate I). The larger-scale velocity field appears to be more
detailed damage survey, we can :,,co that the low-level convergent than rotational at this height. Note the anticy-
maximum corresponds to the time of the most intense clonic signature to the south which was associated with a
damage. brief funnel but no damage.
A photograph of the Edmond tornado near its most intense The Stillwater, Oklahoma, tornado of May 15, 1990, was
phase is shown in Figure 8. The vortex is very narrow but also of the brief supercell cldss. Like the Edmond storm the
intense. The corresponding PPI (Plate 3) shows the velocity Stillwater storm had a well-defined hook echo and a midlevel
'•ignaturc near 55ø azimuth and 33 km range. The TVS is mesocyclone. The time-height profile of the TVS is shown in
much smaller than the gustnado signature in The Village Figure 9. The tornado touched down about 105 km from the
Norman radar so the lowest data are above 1 km AGL. The

June
15,1988 / ,"; ' EDMOND

/! ,.,' 5 2020 May

8,1986
20 25 2O

•,,/.30x.....•
, l? ,i ;!

2O

.'/ ..... / ',,. (i \

; ,• -,. ',, 4O
O -
I I .
40
I i , i 0 •[• ..................
22O5 2210 2215 2220 2225 2230 2235 2305 2310 2315 2320 2325 2330
TIME (UTC) TIME (UTC)

Fig. 6. As in Figure 2 except for the Denver tornadoof June 15, Fig. 7. As in Figure2 exceptfor the Edmondtornadoof May 8,
1988
 VASILOFF 227

Fig. 8. Photographof the Edmond tornado taken near 2317 UTC (courtesy of M. Pennington).

Stillwater TVS appearsto form and intensifybetween 2 and [ZrniE et al., 1985]. The parent storm was long-lived, pro-
3 km before descending.Two separate .A• maxima can be ducing at least five tornadoes. Touching down about 55 km
seen although the maximum aloft is not as pronounced.Note from the Cimarron radar, the Binger tornado eventually
that the most intense damage was associated with the moved to within 35 km of the radar before dissipating.
strongest velocities. Noteworthy in this case is the fact that for most of its lifetime
the tornado's diameter was larger than the radar beam's
3.4. Long-Li•'ed Supercell Tornadoes width. Combined with 0.6 ø azimuthal sampling, compared to
1ø for all of the other cases, this resulted in the tornado's
The Binget tornado of May 22, 1981 (Figure 10), was rated
velocity signature being spread out over three azimuths, not
F4 and produced a damage path that was 2 km wide at times
just two as for the classical definition of a TVS [see Burgess
eta!., 1991]. To be consistent with the other cases. the
maximum ZXz, s in the Binger TVS have been determined
9
using every other azimuth.
STILLWATER
The time-height profile for the Binget TVS (Figure 11)
8 May
15,1990 10 showsa rapidly developing signature which began aloft and
/

7
lowered rapidly. Its formation was associated with strong
convergenceat low levels (letter C in Figure 11). The TVS at
times extended nearly to the top of the storm and was
6
colocated with a pronounced bounded weak echo region
[Lemonet al., 1982].The _Xz,of 133m s-• was measured
near the time the tornado was producing F4 damage and is
the largest AT, on record.
The Doppler velocity display at 0002 UTC (Plate 4) shows
a massive signature with velocities on both sides of the
tornado aliased: the red inside the blue region represents
2 aliased velocities moving toward the radar, and the green
50 40 30 inside the red on the other side indicates aliased velocities
1 movingaway. The largest A•. in this velocity field is over 100
-i
ms

The evolution of the April 30, 1978, Piedmont TVS is

oo'oooo,ooo'o odo 0040ooo o,oo shownin Figure 12. The data are from Burgessand Donald-
TIME (UIC) son [1979]. The tornado, rated F4, formed about 50 km from
Fig. 9. As in Figure2 exceptfor the Stillwatertornadoof May 15, the Norman Doppler radar and was associatedwith a long-
1990. lived supercellthat producedseveral tornadoes.Much
228 SINGLE-DOPPLER RADAR STUDY OF TORNADO TYPES

Fig. 10. Phott•graph •I' the Bingcr tornado taken near 0003 U FC on May 22, 1981 (courtesy of H. Bluestein).

lIingcr, the lbrmation of the rI'VS was associated with strong Binger TVS values. Note the large Av gradients near the
low-level to midlevel convergence. The beginning of the beginning and end of the tornado.
tornado ct•incidcd with rapid intensification of the TVS over
the lowest 2-3 km. Maximum values were 30 m s -• less than 4. IMPACT ON TVS DETECTION ALGORI FHMS

This section addresses the detectability of the above

signaturesby a TVS detection algorithm being developed at
lo

/ 30 40 50 5.040 qn,,,.

i 6
5
60
,_,4 ,o
80
•.3
2

' 2305 2310 2315 2320 2325 2330 2335 2340 2345
2340 2345 2350 2355 0000 0005 0010 00•5 0020 0025 0030
TIME(IJTC) TIME (UTC)
Fi•. il. AsinFigure
2except
fortheBinder
tornado
oœ
May22, Fi•. 12. AsinFigure
2except
forthePiedmont
tornado
oœApril
98]. 30,
 VASILOFF 229

Plate 1. Constant-elevation scan at 1.1° of Doppler velocities from the Cimarron radar showing the May 8, 1986,
gustnado signature. The color scale for the magnitudes is shown at the bottom. Standard convention is used where
positive values indicate motion away from the radar and negative values indicate motion toward. Range rings are every
10 km. The time ploued at the top is CST. Arrow indicates Doppler signature referred to in text.

Plate 2. As in Plate I except for the landspout signatures at 2223

UTe. Elevation angle is o.r and range rings are every 5 km.
230 SINGLE-DOPPLER RADAR STUDY OF TORNADO TYPES

Plate 3. As in Plate I except for the Edmond tornado. Elevation angle is I 1°,

Plate 4. As in Plale I except for the Binger signature. Elevation angle is 0.3C>.
 VASILOFF 231

theNSSL[Vasiloff, 1991].Asdiscussed indetailbyBurgessthe size of the tornado. Work on the TVS algorithmcontin-
et al. [thisvolume],resolution
of a TVS by a Dopplerradar ues, and techniquesfor better and more timely detectionof
isprimarily limited
bythetornado' saspectratiowithrespect weak tornadoesare being investigated.
to the radarandhorizontal andverticalbeamspacing. The
TVS algorithm requires
at leastthreeconsecutive(inrange)
pairsof velocityextrema(so-called"patternvectors")in Acknowledgments. The authoris gratefulto Don Burgess,who
reviewedthispaperand providedinvaluableconsultation duringthe
orderto ensurethata coherent featureis beingdetected.In progress of this study. Rodger Brown, Chuck Doswell, and Mike
addition,three differentelevationanglesare used in the Eilts also reviewed the manuscriptand provided helpful sugges-
vertical associationof the individualfeatures,where each tions. Joan Kimpel drafted the figures. Doppler radar data used in
feature must be within 2.5 km of the one below. this study are from the National Severe StormsLaboratory and the
The TVS algorithm was tested on all of the above cases. MassachusettsInstitute of Technology Lincoln Laboratory under
sponsorshipfrom the Federal Aviation Administration.
Scan-by-scan resultsof severalof themaregivenby Vasi-
loff [1991].The algorithmhad no problemdetectingthe REFERENCES
largestsignatures.
However,as expected,detectability de-
creasedasa functionof sizeof thetornado.For example,the Brady, R. H., and E. J. Szoke, A case study of nonmesocyclone
Edmondtornadohad strongavs, but the numberof pattern tornado development in northeast Colorado: Similarities to wa-
terspout formation, Mon. Weather Rev., 117, 843-856, 1989.
vectors was below algorithmthresholdsand was not consis-
Brandes, E. A., Mesocyclone evolution and tornadogenesis:Some
tently detected. In contrast,the gustnadothat occurredin observations, Mort. Weather Rev., I06,995-1011, 1978.
The Village was detected, mostlikely becauseit was closer Brown, R. A., L. R. Lemon, and D. W. Burgess, Tornado detection
to the radar. The algorithm also performedwell on the by pulsed Doppler radar, Mort. Weather Rev., 106, 29-38, 1978.
landspoutsignatures.However, the ZXvsprobablywould not Browning, K. A., Airflow and precipitation trajectories within
severe local storms which travel to the right of the winds, J.
havemetthealgorithmthreshold
(nominally17m s-1) if the Atmos. Sci., 21,634-639, 1964.
tornadoeshad been much farther away. Burgess, D. W., Single-Doppler radar vortex recognition: Part I:
Another aspectof tornadodetectabilityis the longevityof Mesocyclonesignatures,in Preprints, 17th Conferenceon Radar
the tornado. Short-lived events can be missed if time inter- Meteorology, pp. 97-103, American MeteorologicalSociety, Bos-
ton, Mass., 1976.
vals between volume scansare large. Recall that the gust- Burgess, D. W., and R. J. Donaldson, Jr., Contrasting tornadic
nado signaturein The Village lasted for only 5 min. The storm types, in Preprints, Ilth Conference on Severe Local
signature would have been missed if a 10-min interval Storms, pp. 189-192, AmeNcan Meteorological Society, Boston,
between scans had been used. Mass., 1979.
Burgess, D. W., and L. R. Lemon, Severe thunderstorm detection
by radar, in Radar in Meteorology, edited by D. Atlas, chap. 30a,
5. SUMMARY AND DISCUSSION pp. 619-647, American Meteorological Society, Boston, Mass.,
1990.
Single-Doppler radar signaturesfrom a wide variety of Burgess, D. W., R. J. Donaldson, Jr., and P. R. Desrochers,
tornadoes have been examined. The tornadoesrange from Tornado detection and warning by radar, this volume.
brief F0 gustnadoesto High Plains nonsupercelltornadoesto Doswell, C. A., and D. W. Burgess, Tornadoes and tornadic storms:
violent F4 tornadoes associated with long-lived supercell A review of conceptual models, this volume.
Forbes, G. S., and R. M. Wakimoto, A concentrated outbreak of
storms. Only the gustnadoes were clearly confined to the tornadoes, downbursts and microbursts, and implicationsregard-
boundary layer. At the other end of the spectrum the ing vortex classification, Mort. Weather Rev., I I1,220-235, 1983.
long-lived supercell tornadoes had signatureswhich first Fujita, T. T., Objectives, operations, and results of Project NIM-
appearedthroughout a 3- to 4-km depth above the boundary ROD, in Preprints, 11th Conference on Severe Local Storms, pp.
259-266, American Meteorological Society, Boston, Mass., 1979.
layer before the tornadoes begandoing damage.In addition,
Fujita, T. T., Tornadoes and downbursts in the context of general-
low-level to midlevel convergence was observed in the ized planetary scales, J. Atmos. Sci., 38, 1511-1534, 1981.
mesocycloneregionjust before the formation of the TVS. Lemon, L. R., D. W. Burgess, and L. D. Hennington, A tornado
Differences between brief supercelland nonsupercelltor- extending to extreme heights as revealed by Doppler radar, in
nadoeswere not as apparent. Both types had a circulation Preprints, 12th Conference on SevereLocal Storms, pp. 430-432,
American Meteorological Society, Boston, Mass., 1982.
maximum at low levels concurrent with a separatemaximum Turnbull, D., J. McCarthy, J. Evans, and D. Zrni•, The FAA
aloft. While it is apparentthat a continuumof tornadosizes Terminal Doppler Weather Radar (TDWR) program, in Preprints,
and intensitiesexists, relationshipsamongmechanismsthat 3rd International Conference on the Aviation Weather System,
produce them are poorly understood. pp. 414-419, American Meteorological Society, Boston, Mass.,
1989.
Characteristics of the TVSs were used to assess their
Vasiloff, S. V., The TDWR tornadic vortex signature detection
detectabilityby a TVS detectionalgorithmbeingdeveloped algorithm,in Preprints, 4th International Conferenceon Aviation
at the NSSL. It was found that small tornadoes, even though WeatherSystems,pp. J43-J48, AmericanMeteorologicalSociety,
they may produceup to F3 damage,may not be detected Boston, Mass., 1991.
Wakimoto, R. M., and J. W. Wilson, Non-supercell tornadoes,
even at relatively close ranges (30 km or so), while large
Mon. Weather Rev., 117, 1113-1140, 1989.
tornadoes can be detected at much longer ranges. Thus Zrnid, D. S., D. W. Burgess, and L. D. Hennington, Doppler
tornado detection by the algorithm is dependentlargely on spectraand estimated windspeedof a violent tornado, J. Clim.
the distance between the radar and the tornado, as well as Appl. Meteorol., 24, 1068-1081,
 Radar Signaturesand Severe Weather Forecasting
PAUL JOE

King WeatherRadarResearchStation,Atmospheric
Environment
Serviceof Canada,Downsview,
Ontario,CanadaM3H 5T4

MIKE LEDUC

Ontario Weather Centre, Toronto AMF, Ontario, Canada L5P lB1

1. INTRODUCTION to the outputsof the Dopplerradar.The implementation

of
Dopplercapabilityhas led to a significant
improvementin
The central United States is well known throughout the the processof producingthe severeforecastand warning.
world for the extreme intensity and frequency of occurrence This involves a feedback loop.
of severe thunderstorms. This makes it an ideal location forThe forecastprocessbeginswith an analysisof the syn-
the studyof the most severeof stormswith hail, tornadoes, opticdata and leadsto an operativeconceptual modelor
gust fronts, and microbursts.Many insightshave been hypothesis. Dopplerradial velocitydata presented in real
gained and conceptualmodelshave been derived from time allow the forecasterto verify or discountthe hypothe-
studiesin these regions [e.g., Donaldson, 1965;Browning, sis. In either case, the radar data reinforce or negate the
1982; Ludlam, 1980; Burgesset al., 1982; Burgessand operativeconceptual model,whichmayleadto a changein
Lemon, 1990; Fujita, 1981]. the warningor forecast.Effectiveuseof the Dopplerradar
The developmentof forecasttechniquesin other severe forces the forecaster to think to terms of wind fields on
weatherregimeshas largely been basedon theseresults variousspatialscales.Wind informationis importantto the
[AtmosphericEnvironment Service of Canada, 1982; forecasterbecauseit is fundamental to the understandingof
Doswell,1982,1985].Whileit is expectedthatall regionswill weather systems.
have severe weather, these other localescannotmatch the Now that the forecasteris thinkingalongtheselines, there
climatologyof the centralUnited Statesin intensitynor arebenefitsin the warningof stormsthat maybe outsidethe
frequency[Newark, 1984]. Differentwatch and warning effective
Dopplercoverage area.Theforecasterissensitized
requirementsandconstraintsareimposed on weatherfore- to variouspotentialsevereweatherscenariosbeyondDop-
castingprograms.Therefore,whilethe physicsand detec-
plerrangebasedon observations withinthe coverage
area.
tiontechniquesmaybe similar,the applicationof the tech-
The informationgleamedfrom a limited regionhas been
niques will vary.
Since the last tornado symposium,Doppler radarshave extrapolatedto a larger area.
A review of the climatology of tornadoes and severe
beenimplemented
in severaloperational
environments
for
stormswithinrangeof the Dopplerradarcoverageof south-
theforecastingof tornadoesandothersevereweather.This
ern Ontariowill emphasizethe variety of stormstructures
paperreviewsthe progress madeat the OntarioWeather
Centre(OWC) in Canadain the forecastingof theseevents thatproduce tornadoes in this area.If we are to do an
afterthe implementationof Dopplercapability at the King effectiveforecastingjob for this type of environment,we
Radar Station[Crozieret al., 1991].Improvementsin the need to understand this variety.
forecastprogramwereexpectedaftersuccessful demonstra- Two examplesof severestormswill be presentedto
tionsin theJointDopplerOperational Project(JDOP)(1979) illustratethe interactiveprocessof severeweatherforecast-
program.Suchwasthe caseat the OWC. However,the ingdescribed above.
In onecase,a weaktornado
developed
reasons
for theimprovement
cannotbe singularly
attributed in a situation
wheretheprestorm
conceptualmodelindicated
tornadoes
wereunlikely.RealtimeDopplerdataallowedfor
revision of the model in time to affect the forecast process.
The Tornado:Its Structure,Dynamics,Prediction,andHazards. The secondcase describesa "classic" tornadooutbreak
GeophysicalMonograph79
Published
in 1993bythe AmericanGeophysical Union. situationwhereradardata reinforcedthe expectedscenario.
234 RADAR SIGNATURESAND SEVERE WEATHER FORECASTING

thoughseveraldid displaymesoscalerotation. Four different

types of Doppler signaturesare indicated.
Though the incidence and intensities are lower than that in
Oklahoma,thereare over six millionpeoplewho live under
the Dopplerradarcoverageareaof the Kingradar.Thusany
severe weather event can potentially have grave conse-
quences. In this environment the focus for the severe
weather program at the Ontario Weather Centre, which has
the watchandwarningresponsibility
for thisregion,is on the
"marginal" and strongerevents.

3. ROLE OF ALGORITHMS

The complexityof the radial velocityimageshashastened

the developmentof automatedalgorithmsfor the detection
of severe weather. Algorithms participate in the forecast
processas a tool for the analysisand interpretationof the
Doppler radar data.
Initial testswith the mesocyclonealgorithmusingthresh-
old values appropriatefor Oklahoma storms, suggestedby
Zrni• et al. [1985], filtered out the weak tornadic mesocy-
Fig. 1. Tornado incidence map for North America, 1953-1979 clone signatures that could be visually identified in the
lat'terNewark, 19841.Thick solid line represents4.0, dashedline is imagery. This shouldnot be too surprisingin that the data
['or 2.0, medium solid line is for 0.8, dotted line is for 0.4, and the
sets used to tune the algorithms were biased by research
thinlinesarefor lessthanor equalto 0.2 incidences per 1002km2
per year. projectslooking at the severeststorms. At this stage, tuning
the algorithm is subjective.
The algorithm has been detunedfor operational use at the
2. CLIMATOLOGY OWC to achieve high probability of detection for the "mar-
ginal" tornadic storms. Mesocyclones have been detected
Figure 1 shows the climatology of tornado incidence per for over an hour preceding tornado touchdown [Joe and
1002 km2 peryearfor NorthAmerica[afterNewark,1984]. Crozier, 1988]. Meaningful statistics have not been estab-
In southern Ontario the incidence is about one quarter of lished due to the rarity of the tornadic events.
that in Oklahoma. Generally, the intensities are much lower.
Tornado incidence is shown since it is a phenomenonthat is
4. EXAMPLES OF THE FORECAST
rare but memorable. Table 1 lists the severe weather events
FEEDBACK PROCESS
of a typical year (1986) within the Doppler coveragearea of
the King radar (1 10-km radius) [Leduc and Joe, 1987]. In this section, two cases are presented to illustrate the
Notice that only one event is listed as a confirmed tornado feedback processin severe weather forecasting.

TABLE 1. Storms in 1986 Within Doppler Range (110 kin) of King City Radar
Lead Time of
Lead Time of Doppler Signature
Warning Before Before First
First Occurrence Occurrence of
Time, Reported of Severe Severe Weather, Doppler Radar
Date UT Damage Weather, rain min Signature

May 6 1930 wind 0 20 descendingjet

June 16 1830 two wind events 60 40 rotation couplet
descendingjet
June 29 1830 wind 0 20-30 jet
July 13 2300 wind and hail 0 0 convergence
July 18 2100 wind and hail 30 40 jet
Aug. 1 2030 hail 120 90 convergence
Aug. 1 2400 hail 0 30 convergence
Aug. 2 1820 hail 70 60 convergence
Aug. 8 2400 wind 0 0 rotation
Aug. 15 2000 floods 40 70 rotation
Sept. 29 2330 F0 tornado 0 20
 JOE AND LEDUC 235

4.1. September 17, 1988, FO Tornado

A weak tornado briefly touched down in a highly popu-

,/ • )J• Severe
/' 17September
198i••'
Weather
C.
omp•si•e' lated area northeast of the Pearson International Airport in
/....---l..... ! + .•,00Z Toronto on the Afternoon of September 17, 1988. Twelve
hours before the event, a moderately active warm front
/' •
'i '"-'"% _,..C'-..•i 500 mb '1
"-.--.,,__•.•-• \ I I moved through Southern Ontario.
The severe weather assessment(Figure 2) indicated (1) a
low-level moisture axis pushed into the region with dew
•. • ,/ '-, '•,• ,, pointsof 19øC,(2) a 85-knot (42 m/s) high-levelwesterlyjet,
(3) a 30-40 knot (15-20 m/s) curving low-level west to
southwesterlyjet, and (4) a sharp vorticity trough at 500
mbar.
The modified 1200 UT Flint soundingindicated a potential
buoyant energy of 1700 J/kg, and the average wind shear
/ • ', I-, •"' /" , between the surface and 4 km was 9 m/s/km (Figure 3).
Observed surface temperaturesof 24øC and dew points of
Fig. 2. Composite map for 1200 UT September 17, 1988, summa- 19øCwould break the inversion and produce clouds to 11
rizing the features conductive for severe weather. km. According the Rasmussenand Wilhemson [1983], this

72637 FNT Flint, Michigan SEP 17 1988 1200Z

9O

Fig.3. Sounding
fromFlint,1200UTSeptember
17,1988.
Plottwig
convention
isstandard.
Conditions
aremarginal
for severe
236 RADAR SIGNATURES AND SEVERE WEATHER FORECASTING

would place the environmentalconditionsas marginalfor September 17 1988

severe storm development.
PMAX m/s krn
The conventional radar observations indicated a few small
isolatedcells, none of which appearedsevere.Echo tops
were mostly in the 4- to 6-km range. However, the tornado-
600
producingstorm showeda small area of 50-dBZ echo, and
the echo tops showedrapid developmentto 8-10 km in a 5O0

10-rain period between 2020 and 2030 UT which was still 40O

below the height of the tropopause.Only three cloud to 300

ground lightningstrikes were detected. At this point, the
20O
storm track veered to the right, but this motion was attrib-
uted to steeringof the storm by the winds at the 8- to 10-kin 100

height rather than to propagationeffects[Klemp, 1977].The 0

strongest reflectivities were observed at a height of 3 km. 2010 2020 2030 2040 2050 2100 2110 2120 2130 2140
Deep vertical development and high reflectivities aloft (40
dBZ at a height of >5.5 km), consistentwith severe storms,
TIME f
• Elev - 0.5 • Elev - 1.4 •E• Elev - 3.5
were never observed [Donaldson, 1961]. Given this assess-
ment, the storms observed on radar were assumed to be Fig. 5. Time evolution of momentumcorrespondingto Figure 4.
nonsevere.

The automated detection algorithm identified a tornadic

mesocyclone beginning at 2030 and at 2040 UT, based on hodograph, the storm relative helicity values were around
shear (>10 m/s) and momentum (>100 km m/s) thresholds 150m2 s2, whichis marginalfor tornadodevelopment.
(Plate 1 and Figures 4 and 5). At 2050 UT the classificationof However, beginning at 2030 UT, a strong low level jet of
the mesocylone was degraded to nontornadic status due to 20-25 m/s was detected on Doppler radar. Using this wind
low momentum. The tornado touched down at 2120 UT. The
fieldimpliesa stormrelativehelicityof around300m2 s2,
classificationis tentative and not statistically defensible due well above the threshold for tornadoes. Other cases in
to a low number of samples. southern Ontario [Murphy, 1991] have produced weak tor-
This appears to be a situation where a tornado warning nadoes with marginal dynamics but with storm relative
would never be issued. However, other data were available helicityvaluesjustabove150m2 s2.
to modify this assumption. Storm relative hellcity is a The three pieces of radar information (cell motion, auto-
relatively new concept that has shown promise in discrimi- mated mesocyclone identification, and wind profiling) sup-
nating tornadoes from nontornadoes [Davies-Jones, 1984; port and are consistent with each other in providing cre-
Davies-jones et al., 1990]. A storm motion of 290ø at 10 m/s dence for tornado potential. This reassessment could not
was deduced from radar data. Using the 1200 UT Flint have been done without the Doppler radar information. This
evidence is sufficientto alter the hypothesized conceptual
model. The forecaster can take advantage of this to reeval-
uate the severe weather situation. A new model of a "weak
September 17 1988
tornado" case has been established. The type of warning
<S> m/s/km messageto issue in this case remains a problem. Further
discussionis beyond the scope of this paper.

4.2. May 3I, 1985, F2-F4 Tornadoes

lO
On the afternoon of May 31, 1985, a cold front moved
through southern Ontario triggeringa series of strong to
violent tornadoes. On the morning of May 31, a 98.5-kPa
low-pressurecenter tracked across upper Michigan and
moved to the north of Lake Huron by evening. A very sharp
cold front trailed southward from the low-pressure center.
Soundingseast of the cold front showed warm moist air
overlain by cool dry air, a classic "loaded gun" sounding.
2010
2020
2030
2040
2050
2100
2110
2120
2j•0
2140 Time
Potential buoyant energy was 3000 J/kg, and the 0- to 4-km
• Elev = 0.5 • Elev ß 1.4 • Elev = 3.5 shear was 13 m/s/km (Figure 6). The wind profile showed
strongshearwith southerlywinds(180øand 10knots(5 m/s))
Fig. 4. Time evolutionof shearfor the September17, 1988,F0 at the surfaceand a strongwesterlyjet (200 knots(100 m/s))
tornadic mesocyclone.Arrow indicatestornado dissipationtime.
The algorithmhasbeentunedto providehighprobabilityof detec- at 10 km. The cold front and a sharpupper trough provided
tion. the triggeringmechanismto releasethe instability.In
 JOE AND LEDUC 237

Plate 1. Radial velocity image (0.5ø plan position indicator (PPI)) for 2120 UT September 17, 1988. shox•inga
rotationalsignatureat range 30 km and azimuth 190ø and the output from the automatedmesoscyclonedetection
algorithmfor 2040 UT showinga detectedfeaturethat was classifiedas tornadic(indicatedby the asterisk)

RA IN RATE
MH,"HR

64

16, "--'
8

4.

SO KM
320 6,
772 •2
896 16
1458 8
1800 4
1680 '2.
.- ,' 100.25
34088
40 .-%.
KM R IML. 48 KPl R I

Plate2. 1.5-kmconstant
altitudeplanposition
indicator
(CAPPI)anda MAX R chart(maximum
reflectivity
above2
km)showing
hookechoes
andweakechoregions
for2020UTMay31,1985.
AreaswheretheMAXR shows
intense
echoaboveweakerechoes
onthe1.5-km
CAPPIareindicative
ofweakechoregions.
TheFujitascaleisusedtoindicate
the intensityof the threetornadoesat this
238 RADAR SIGNATURES AND SEVERE WEATHER FORECASTING

72429 DAY Dayton,OH MAY 31 1985 1200Z

\

90

Fig. 6. Sounding from Dayton May 31, 1985, a classic "loaded gun" soundingconducive for severe weather.

all the necessary classic factors [_Fawbush et al., 1951;

Fawbush and Miller, 1953] were present for tornadic severe
storms (Figure 7). /
The first thunderstorm cells developed at 1740 UT, and by
1820 UT, severe thunderstorm warnings were issued that
included the statement "remember that some thunderstorms '"-. /
/.... '/'
/ 31May
•evere 1985
Weather
Compostt3
produce tornadoes." In this case, the radar data were / ,, .-_.._.,_. + ,, '

consistent with the forecast scenario and confirmed the

conceptual model. Table 2 showsthe sequenceof warnings
and severe weather events. Each tornado was preceded by a
warning of its imminence.
As the events unfolded, the conventional radar imagery
exhibited all the classic reflectivity signatures of tornadic
severe storms [Donaldson, 1965; Forbes, 1981; Burgess and
Lemon, 1990] (high reflectivity throughout the core of the
storm, large reflectivity gradients, hook echoes, high echo
..... c./ ,•/•-'=•
tops, weak echo region, large areal extent); this feedbackto ' '-'N

the forecaster continually reinforced the conceptual model

(Plate 2). Fig. 7. Severe weather compositefor 1200 UT May 31, 1985,
In somerespects,this casewas "easy"; everythingfit into showing the relevant synoptic
 JOE AND LEDUC 239

TABLE 2. Time of Warningsand Tornado Touchdown

Issue Time of Issue Time Time of
Severe of Tornado Actual Event
Thunderstorm Warning, (Tornado),
County Warning, UT UT UT Town

Northern Bruce 1825 1900 Rush Cove

Northern Wellington 1915 2015 Arthur
Dufferin 1954 2028 Grand Valley
2045 Orangeville
Southern Grey 1915 2017 Corbetton
Southern Simcoe 1954 2100 2118 Holland Landing
Northern Simcoe 1954 2100 Barrie
Northern York 2053 2100 2125 Holt
Northern Durham 2053 2120 2140 Wagner Lake
Southern Victoria 2120 2120 2205 Reaboro
SouthernPeterborough 2205 2205 2220 Cavan
2225 Birdsall
Southern Hastings 2225 2225 2235 Minto

a consistentpicture, and all the signspointed to tornadic radar does not play a role here. There is a consistent20%
severe weather. All the various severe weather indicators improvement in these measures for warnings during the
were at the extreme end of the spectrum. Doppler era. It is estimated that 10-15% of the cases were
related to tornadoes. Surveys have not been consistently
5. DISCUSSION done to distinguishbetween tornado and straight-linewind
damage.
This paper discussed and illustrated the change in the What is interesting in these results is that improvements
severe weather forecasting process at the Ontario Weather have been noted even in geographical locations outside
Centre/King Radar that has come with the introduction of Doppler coverage. The consensusis that exposure to Dop-
Doppler radar technology. There has been a positive influ- pler data has allowed the forecasterto practically use his or
ence, and this is reflected in the improved warning perfor- her knowledge of storm structure. This has led to better use
mance in the past few years. Figure 8 shows the results for of non-Doppler data that in turn leads to better warnings.
verifications of 8 years of watches and warnings. Doppler For example, hook or pendant echoes, inflow notches, and
capability was included in the last 4 years of the verification rear inflow notches are readily observed, analyzed, and
assessment.
interpreted when combined with the velocity information.
The probability of detection (POD) and reliability (com- The Doppler data have reinforced the reflectivity observa-
plement of the false alarm ratio) for watches over the two tions.
periods has not changed. This is synoptic in nature, and An attempthasbeen made to explain the improvementsin
terms of an enhancement in the forecast process due to a
feedback loop. The radial velocity information is central to
Watches/Warnings the process.At this time, the direct outputsof the Doppler
radar, the algorithms and imagery, play an initiating or
4 year averages supportingrole to the warning process.This information
0.8 must be combined with other observations and conceptual
models for effective use as illustrated in the examples. The
implicationis that successfulfully automatedwarning sys-
0.6
tem must consider all the available information.
While the focus of this paper has been the feedback
0.4 processin the forecastingof tornadoes,the commentsapply
to other weather situations on a variety of scalesas well.
0.2

6. CONCLUSION
o
Watches POD 1 - FAR Warnings POD I - FAR Doppler radar has had a positiveimpact on the severe
weatherprogramat the OntarioWeatherCentre.It provides
I 1983-1986 • 1987-1990 wind information that is directly applicable by the severe
Fig. 8. Watch and warning verification statisticsfor 8 years weather forecasterfor validation of hypothesis.A feedback
(1983-1989). loopis established
that allowstheforecasterto iterateto
240 RADAR SIGNATURES AND SEVERE WEATHER FORECASTING

correct forecastor warning. No other sensorprovidesinfor- as a tornado forecasting parameter, in Preprints, 16th Conference
mation as effectively on this spatialand temporalscale. The on Severe Local Storms, pp. 588-592, American Meteorological
Society, Boston, Mass., 1990.
feedback is most important in climatic regimes and popula-
Donaldson, R. J., Radar reflectivity profiles in thunderstorms, J.
tion centers where severe weather of any intensity can have Meteorol., 18, 292-305, 1961.
grave consequences. Donaldson, R. J., Jr., Methods for identifying severe thunderstorms
Though the images can be difficult to interpret and the by radar: A guide and bibliography, Bull. Am. Meteorol. Soc., 46,
automated detection algorithms have yet to reach maturity, 174-193, 1965.
there is sufficient information to allow the forecaster to make Doswell, C. A., III, The operational meteorology of convective
weather, vol. 1, Operational mesoanalysis, NOAA Tech. Memo.,
creative use of his knowledge of storm structure to provide NWS NSSFC-5, 158 pp., Natl. Oceanic and Atmos. Admin.,
more accurate and timely warnings. Automated systems Boulder, Colo., 1982.
must consider all data sources and not just radar data to Doswell, C. A., III, The operational meteorology of convective
provide the most effective warnings. weather, vol. 2, Storm scale analysis, NOAA Tech. Memo., ERL
ESG-15, 240 pp., Natl. Oceanic and Atmos. Admin., Boulder
Though there may be differencesfrom one radar systemto Colo., 1985.
another, the major impact of these systems will be in the Fawbush, E. J., and R. C. Miller, Forecasting tornados, Air Univ.
enhancement in knowledge of the user. The process of Q. Rev., pp. 108-117, Maxwell Air Force Base, Ala., 1953.
interpreting and using the Doppler imagery has led to a Fawbush, E. J., R. C. Miller, and L. G. Starrett, An empirical
methodfor forecastingtornadodevelopment,Bull. Am. Meteorol.
better understandingof mesoscalemeteorology. This paper Soc., 32, 1-9, 1951.
described the process by which this knowledge was and Forbes, G. S., On the reliability of hook echoesas tornado indica-
continues to be gained at the Ontario Weather Centre. tors, Mon. Weather Rev., 109, 1457-1466, 1981.
Similar results can be attained elsewhere provided that the Fujita, T. T., Tornadoesand downburstsin the contextof general-
necessarytraining and education programs are in place. A ized planetary scales,J. Atmos. Sci., 38, 1511-1534, 1981.
Joe, P., and C. L. Crozier, Evolution of mesocycloniccirculationin
developmentprogram would be a great benefitas well. severestorms,paper presentedat 10thInternationalCloud Phys-
ics Conference, International Association of Meteorology and
REFERENCES AtmosphericPhysics, Bad Hornborg, Fed. Rep. of Germany,
Aug. 15-20, 1988.
Atmospheric Environment Service of Canada, Summer severe Klemp, J., Dynamicsof tornadicthunderstorms,
Annu. Rev. Fluid
weathercorrespondence
course,Training Branch,Toronto, Ont., Mech., 19, 369-402, 1987.
1982. Leduc, M. J., and P. Joe, Use of a 5 cm Doppler radar to detectand
Browning, K. A., General circulationof middle-latitudethunder- forecast severe thunderstorms in a real time forecast operation,
storms, in Thunderstorms, vol. 2, edited by E. Kessler, pp.Proceedings Symposiumon MesoscaleAnalysisand Forecasting,
211-248, National Oceanic and Atmospheric Administration, Eur. SpaceAgencySpec. Publ., ESA SP-282,49-54, 1987.
Boulder, Colo., 1982. Ludlam, F., Cloudsand Storms: The Behavior and Effect of Water
Burgess,D. W., and L. R. Lemon, Severethunderstormdetection in theAtmosphere,405 pp., PennsylvaniaStateUniversityPress,
by radar, in Radar in Meteorology,edited by D. Atlas, pp. State College, 1980.
619-647, AmericanMeteorologicalSociety, Boston,Mass., 1990. Murphy,B., Tornadoandflashfloodin southwestern Ontarioon 31
Burgess,D. W., V. T. Wood, and R. A. Brown, Mesocyclone May 1991,OntarioReg. Tech.Note, ORTN-91-3,8 pp., Atmos.
evolution statistics, in Preprints, I2th ConJkrenceon Severe Environ. Serv. of Canada, Toronto, Ont., 1991.
Local Storms,pp. 422-424, American MeteorologicalSociety, Newark, M. J., Canadiantornadoes,1953-1979,Atmos. Ocean, 22,
Boston, Mass., 1982. 343-353, 1984.
Crozier, C. L., P. I. Joe, J. W. Scott, H. N. Herscovich,and T. R. Rasmussen,E., and R. B. Wilhelmson,Relationshipbetween storm
Nichols,The KingCity operational Dopplerradar:Development, all characteristicsand 1200 GMT hodographs,low-level shear, and
seasonapplications
and forecasting,
Atmos.Ocean,29, 479-516, stability,inPreprints,13thConference onSevereLocalStorms,pp.
1991. 55-58, AmericanMeteorological Society,Boston,Mass., 1983.
Davies-Jones,R. P., Streamwisevorticity: The origin of updraft Zrnid, D. S., D. W. Burgess,and L. D. Hennington,Automatic
rotationin supercellstorms,J. Atmos.Sci., 41, 2991-3006,1984. detectionof mesocyclonicshearwith Doppler radar, J. Atmos.
Davies-Jones,R. P., D. W. Burgess,and M. Foster,Testof helicity Oceanic Technol., 24, 425--438,
 The Use of Volumetric Radar Data to Identify SupercelIs'
A Case Study of June 2, 1990
RON W. PRZYBYLINSKI

National Weather Service, NOAA, Saint Charles, Missouri 63304

JOHN T. SNOW ^ND ERNEST M. AGEE

Department of Earth and AtmosphericScience, Purdue University, West Lafayette, Indiana 47907

JOHN T. CUtLRAN

National Weather Service, NOAA, Indianapolis, Indiana 46241

1. INTRODUCTION 1988 to the present across the mid-Mississippiand Ohio

valley regions.
Conventional radar is used in National Weather Service
offices to help issue timely warnings of severe weather
2. PRESENT KNOWLEDGE
(tornadoes, severe thunderstorms,and flash floods) to the
public. Monitoring potentially severeor tornadicthunder- Duringthe early and middle 1960s,Browningand Ludtam
storms can become a significantchallenge,particularly dur- [1962]and Browning[1964] introducedthe term "supercell"
ing a major tornadicoutbreak.This was the situationduring to describethe quasisteadystate structureattainedby some
the eveningof June2, 1990,acrossmuchof Indianawhen38 storms during their intense phases. Supercell storms pro-
tornadoes occurred over a period of 6 hours. Early in the duce most, if not all, violent tornadoes and account for a
outbreak,as many as sevensupercelIswere identifiedsimul- highpercentageof extremely large hail events. One impor-
taneouslyacrossparts of west central Indiana and extreme tant feature of the radar echo from this type of storm is a
southeastIllinois. The rapid evolutionand movement(great- distinctly shapedregion void of detectable echo which
er than 25 m s-1) of the individualradarechostructures penetratesupwardinto the core of the stormbeneaththe
furthercomplicatedan alreadydifficultwarni'ng
situation. storm'shighesttop. Browningand Ludlam[ 1962]referredto
The purposesof this paper are twofold. The first part this regionas an "echo free vault." Chisholm[1973]rede-
focusesuponthe variationof radarechostructures associ- fined this region as the "bounded weak echo region"
ated with supercelIswhich occurredduringthe eveningof (BWER) which is often observed with the main storm
June2, 1990.Both low-levelandelevatedpresentations will updraftnearthe rightrear flankof the supercell.
Depending
be shownat selecttimesduringthe evolutionof the tornadic uponradarresolutionandrange,theBWER maynotalways
outbreak.Many of the examplesare maturesupercelIs. be detected;however, the weak echo region(WER) is often
The secondpart of this paper builds upon the current more identifiable.Other equally important radar echo fea-
conceptualhigh-precipitation
(HP) supercellmodelsintro- tures associatedwith the supercellinclude(1) the highest
ducedby MollerandDoswell[1988]andMolleret al. [1990]. stormtop is positionedover the BWER (or WER) and(2) a
Two crude three-dimensionalconceptualmodels of high- notchof weakerechosurroundedby a hook(pendant)echo
precipitationsupercelIsare introduced.Thesemodelswere signifying
thelocationof possibleupdraftrotation.Research
derived from low-level and elevatedplan positionindicator effortsconductedduringthe 1970susingDopplerradar have
(PPI) radar observationscollectedduring the period from successfully identifiedthe BWER as [Burgesset al., 1977]a
centerof significantupdraft and cyclonicrotation.
The Tornado: Its Structure,Dynamics,Prediction,and Hazards.
GeophysicalMonograph 79
Lemon [1977, 1980]introduceda volumetricradar scan-
Copyright1993by the AmericanGeophysicalUnion. ningmethodto aid operationalforecasters
in identifying
242 IDENTIFYING SUPERCELLS WITH VOLUMETRIC RADAR DATA

5a 6a 7a

FFN

cL

• front
flank 5b 6b 7b
notch
rear
Inflow
notch

SCHEMATIC EVOLUTION OF t-LP SUPERCELL COMPOSITE LIFE CYCLES OF I4P STORMS THAT
SHOWING DEVELOPMENT OF BOW' ECHO HAVE BEEN IDENTIFIED.

STRUCTURE (DOSWELL, 1985).

Fig. I. Conceptual modelsand compositeradar structures(life cycles) of high-precipitationsupercelIs(adaptedafter

Doswell [ 1985]and Moller et al. [ 1990]). M signifiesprobablelocation of mesocirculation.Arrows on right side of figure
denote location of rear inflow jet.

classic supercelIs and multicell thunderstorms. The tech- istics of HP supercelIsinclude (1) a weak reflectivity notch
nique relies upon the identification of indicators of updraft located along the forward flank (signifying the probable
strength, since the strength of the storm's updraft greatly location of a rotating updraft or mesocyclone),(2) extensive
influencesthe shape and appearance of the radar echo and precipitation(includingtorrential rainfall and hail) along the
type of severe weather events. Forbes [1981], usingconven- right rear flank, and (3) the mesocycloneembeddedwithin
tional radar reflectivity data from the April 3, !974, super significant precipitation.
tornado outbreak over the Ohio valley region, examined the Doswell [1985] (referred to hereinafter as D85) M90, D90,
shape and reflectivity intensity of several tornadic and Przybylinski [1989] (referred to hereinafter as P89), and
nontornadic storms. He classified a group of unique echoes Przybylinskiet al. [1990] (referred to hereinafteras P90) all
as "distinctive echoes," suggestingthat this group signified have shown that HP storms can exhibit a number of echo
strong, persistent rotation and had the highest tornadic shapes.Examplesof radar imagesand echoevolutionasso-
probability. ciated with HP stormsare shownin Figure 1 (left side). Echo
Recent studies completed by Doswell et al. [1990] (re- characteristicsinclude (1) an overall echo that may take the
ferred to hereinafter as D90) have shown that supercelIscan form of a kidney-beanshapedconfiguration,(2) a hook echo
take on a variety of echo shapesand differing evolutions. that is exceptionallylarge (a broad hook echo filled with
They pointedout that there is a "spectrumof supercelIs"in significantprecipitation,(3) weak echo notchesalong the
which all supercelIsare neither completely unique nor forward and rear flanks of the storm, (4) persistentlow-level
completelyidentical. However, D90 have shownthat the reflectivitygradientsadjacentto thefrontflanknotch(FFN),
common denominator of all supercelIs is that they have and (5) a spiralor "S" shapedechostructure.Weaverand
updraftswith persistentrotation. Nelson [1982], Foote and Frank [1983], and Nelson [1987]
The spectrumof superceliscan be subdividedinto three have indicated that one variation of HP storms may contain
basic categories:(1) classic(C) [Browning, 1964],(2) low- multiplemesocyclonesand BWERs. Their data set suggests
(LP) [BluesteinandParks, 1983],and(3) high- that "hybrid" supercelIs(apparentlythe sameas HP) may
precipitation
(HP) [Molleret al., 1990](referredto hereinaf- exhibit multicell traits (a storm having several high-
precipitation
ter as M90). The conceptual model of HP supercelIswas reflectivitycores)and are oftenassociated with widespread
recentlyintroducedby M90 on the basisof radarandvisual damaging hail events.
observations. M90 have shown that the key echo character- Studiesby M90 also suggested that supercelIsfollow
 PRZYBYLINSKI ET AL. 243

LOWER OHIO VALLEY TORNADO OUTBREAK OF dUNE

I
2-3, 1990
I
!

CHAMPAIGN
DANVl i

SPRINGFIELD OF.CATL•
RICHMOND
1
• ! oAY
TON
OHIO
I

KENTUCKY

I•ILEXINGTON

Fig. 2. Overall mappingof the damagecausedby the June 2, 1990, storm. Dark solid contours represent tornado
tracks. The damage survey was conductedby National Weather Service personnel from Indianapolis, Louisville,
Evansville, Cincinnati, and Springfield. Data source is Storm Data.

variety of life cycles(right sideof Figure 1). SomesupercelIs life cycle. This type of cyclic mesocyclone development is
follow the LP-C-HP (low-precipitation-classic-high- similar to cyclic mesocyc!one evolution observed in classic
precipitation) lifecycle; a more common transition is the C to supercelIs[Burgess et al., 1982].
HP (classic to high-precipitation). Additionally, it is quite
3. R^D^R ECHO P^TTERNS IDENTIFIED
common for HP storms to evolve into a bow echo (BE)
storm (HP-BE) with a rotating comma head. The rotating ON JU•qE2, 1990: ILLUSTRATIONS
OF CLASSIC AND HP SuPœRCELt.S
comma head echo documented by Wakimoto [1983], P89,
and P90 can be a reflection of a strong mesocyclone. The June 2, 1990, tornado outbreak over central and
However, it is often noted during the dissipatingstagesof southernIndiana presents an excellent opportunity to show
the HP storm. One of the unique characteristics of HP variations in the radar echo patterns of the supercelIs sur-
supercelIsis that the mesocycloneis embeddedin significant veyed and their transitions within the classic phase and
precipitation for an extended period of time. This character- classic to HP storm phase. A map of the tornado tracks
istic separatesHP storms from classicsupercelIsin which across central and southern Indiana is shown in Figure 2.
mesocyclonesare wrappedup in significantprecipitationfor Detailed surveys of the damagetracks and times of tornado
only a brief period during the dissipatingstages. occurrencewere conducted by National Weather Service
The right-handpart of Figure 1 showstwo differentlife personnel (including two of the authors). The strongest
cycles from the C-HP transition. The C-HP transition noted tornadoes(F4 intensity) occurredacrosssouthernIndiana,
(2-4) may evolve either in a bowing echo structure(C-HP- where several supercelIs spawned families of tornadoes.
BE, 2-8A life cycle) or, alternatively,as a new mesocyclone Additional tornadoes (F2-F3 intensity) traveled across cen-
is forming on the storm's right flank (2-7B). M90 have noted tral sectionsof the state. Total damageexceeded75 million
that cyclic mesocyclonedevelopmentcan occur with either
244 IDENTIFYING SUPERCELLS WITH VOLUMETRIC RADAR DATA

200 km 200 km

150 km i50 km

km

50 ken 50 km
Frotit flank
notch

Flanking line echo

• iND (DiN
D
Pendant
Echo
Pendant
Echo

Pendant
Pendant
Echo
Echo

Fig. 4. Sameas Figure 3 except for 2346 UTC.

Fig. 3. Plan position indicator radar display from Indianapolis
radar at 2328 UTC, June 2, 1990. Antenna elevation angle is 0.5 ø.
Reflectivitycontoursare 18, 30, 41, and 46 dBZ. Shadedregions flank to the forward flank of the storm (Figures 3 and 4). A
representreflectivity values greater than 50 dBZ. front flank notch evolved, and the mesocirculation was
embeddedin significantprecipitation.A secondstrongF2
intensity formed near the front flank notch at 2346 and
Reflectivity data were collectedfrom the WSR-74C (5.4 produceda continuousdamagetrack of 18 km.
cm wavelength,1.5øbeamwidth) National WeatherService In contrast, storm F continuedto show classicsupercell
radar at Indianapolis,Indiana (IND). Both 16- and 35-mm characteristicsin which the pendantecho, located near the
photographic film recordsof PPI displayswerecollectedfor rightrearflank,wasfree of surroundingprecipitation(Figure
the study. Volumetric reflectivity data were collected at 3). Storm F spawneda seriesof weak intensity(F0-F1)
tornadoes from 2315 through 2345.
discrete times to determine storm structure. Owing to the
large number of supercelIspresent and rapid movement, Storms G, H, and I, also shown in Figure 3, were more
only partial tilt sequences
were accomplished.Lower and isolated and exhibited variations of supercell structures.
middle levels of storms were frequently examined to deter- StormG spawnedits secondF4 intensitytornadoat 2307
mine radar echo characteristics.When possible,the location beforeweakeningconsiderablyafter 2328as a weak convec-
of the maximumecho top was documentedand comparedto tive cell near the southeastflank appearedto mergewithin
low-level and midlevel echo features. theupdraftregionof thetornadiccell. Supercell
H spawned
Figure3 showsa detailedview of the radarreflectivities an F4 intensitytornadoat 2246 and a secondtornado(F2
from north central Indiana through extreme southeastIlli- intensity)at 2345,whilesupercell
I generated
itsfirsttornado
nois at 2328 (all times UTC). A nearly continuousline of after 0019.
thunderstormswith embedded supercelIs extended across Detailed low-level and elevated PPI scansof supercelIsD
Indiana. Storm D spawneda strongF2 intensitytornado and F over west central Indiana at 2346 are shown in Figure
nearthe rightrear flankof the cell at 2318andproduceda 5. Storm D continued to exhibit HP storm characteristics
damagetrackof over 12 km in length.Between2328and from2346throughits finalstages.Characteristics includeda
2346, stormD graduallyfollowedthe classic-HP(C-HP) front flank notch,indicatingthe probableupdraft regionof
transition in which the mesocirculation moved from the rear the cell, a mesocirculationembedded in significant
 PRZYBYLINSKI ET AL. 245

updraftcenters
neartherightrearflanksof eachreflectivity
core.The intense
low-levelreflectivity
gradients,
weakecho
notches(WEN), andpendantechoesidentified alongthe
rightflanks
ofeach50-dBZcoresuggest
theprobable
updraft
region of each supercell.
If PPI reflectivity
datawereexamined
solelyat 0.5øfor
stormF, the attentionwouldhavelikely beenfocusedon
bothpendantechostructures adjacentto each50-dBZre-
IND flectivitycorefor possiblesevereweatheror tomadicactiv-
ity. However,examination
of PPI reflectivity
dataat subse-
quent elevation angles (1.5ø and 2.5ø) showed that the
100km (' 50km
downwind(frontflank)50-dBZreflectivitycorehadlimited
verticaldepthcompared
to itsupwind
neighbor.
Theupwind
50-dBZ core extendedwell into the middle levelsof the cell
(2.5øscan)withwell-defined
echooverhangalongtheright
flank.Suchlow-levelandelevated
reflectivity
features
sug-
flanking gestthat the strongerof the two updraftcenterswaslocated
line
echo
nearthe rightrear of the system.Havingtheseadditional
dataavailable,the warningforecasterwouldfocushis atten-
tion on the rightrear flankof stormF for possibletorna-
dogenesis.StormF spawned a strongF2 intensity
tornadoat
IND 0030 approximately50 km west-northwestof IND.
EL1.5 ø ExaminingPPI reflectivitydata at two or more elevation
100km
anglesto determinestormseverityhas significant
advan-
tages comparedto viewing stormsat only one elevation
angle. This type of samplingformat needsto be exercised
IN /• (T) frequentlyto observechangesin the stormevolution.
Detailedlow-levelandelevatedscansof supercelIs
H and
I (Figures6 and7, respectively)illustratesomeaspectsof
theverticalstructuresof thesestormsat 2346.Bothsuper-
/ • ? •J• • flanking cellswereisolatedcompared to theirnorthernneighbors
(D
I )•"• • ,,,,•"•,J-•[ /• -line
andF). Data sampledat 1.5øand 2.5ø at a rangeof 150km
(stormpositionfor cell H) representheightsof 5.0 and 7.5
C•••••Pe•ant echo
• km, respectively.The overallstructureof supercell H sug-
Pe•antecho
•••• • % IND geststhatthemesocirculation andtornadowerewrappedin
• 2346UTC EL05ø
l•km WEN 50km ' precipitation(nearthe southernflankof the VIP 3 echo).The
RIN along the trailing edge of storm H at low-level and
Fig. 5. Detailedlow-leveland elevatedplan positionindicator elevatedscansindicatesthe probablelocationof descending
radarrefiectivities
from Indianapolis
radarat 2346UTC, June2, flow similarto observationsnotedby D90, M90, and P89.
1990,ofconvective storms D andF. Reflectivity
contoursare18,30, The frontflanknotches,indicatingpotentialupdraftcenters
41,and46 dBZ. Shaded regionsrepresentreflectivity
valuesgreater
than50dBZ.T signifieslocationoftornado.EL represents elevation alongthe leadingedge,are not well defined.The persistent
anglesduringtilt sequence.R!N is rear inflow notch. WEN is weak verticaldepthof the RIN and overallcomma-shaped echo
echo notch. FFN is front flank notch. patternnotedat 2.5ø suggested
the presenceof a mesocircu-
lation which acted to redistributethe precipitationfield
(D90).At 2346,stormH waslikelyin theocclusion phaseof
and multiple high-reflectivitycores (multicelltraits) down- a classicsupercellevolution since a strongF2 intensity
windfromtheparentmesocirculation. Additionalimportant tornadohavinga 25-km damagetrack wasoccurringat this
echofeaturesassociated with stormD includeda flanking time. SupercellH later spawneda secondF2 intensity
line echo southof the parentcirculationanda well-defined tornado approximately 50 km southeastof Terre Haute,
rear inflownotch(RIN) immediatelyupwindof the flanking Indiana (HUF).
line echo.Suchechofeaturesindicatethe probablelocation Furthersouth,supercellI appearedto exhibitHP super-
of descendingflow (rear flank downdraft)similarto radar cell characteristics in which a front flank notch was observed
signaturesidentifiedby D90 and M90. along the leading edge of the convective echo, and the
In contrast, storm F continuedto show nearly classic
mesocirculation outlinedby thependantechoappearedto be
supercell characteristics. The two 50-dBZ cores identified
wrapped in significantprecipitation(Figure 7). The RIN,
within the larger echo mass suggestthat storm F was signifyingthe locationof descendingflow, was not evident
composedof two supercelIshaving distinct and separate alongthe trailing edge as comparedto stormH.
246 IDENTIFYINGSUPERCELLS
WITH VOLUMETRICRADARDATA

EL2.5 ø
RiN•
ß
• 150km
EL
2.5
ø echo overhang

'""•'•150km
"• FFN
EL1.5 ø

rear
Inflow

notch
r'•'••18

•
Pendant echo
•, front
flank
150
km
EL
0.5
ø
notch 2346 UTC

Pendant
echo
EL0.5 ø
Fig. 7. Sameas Figure 5 exceptfor convectivestorm I.
2346 UTC

Fig. 6. Sameas Figure 5 except for convectivestorm H.

Nelson [1982]in which a hybrid supercellintersectedand
movedalonga boundarygeneratedby a downwindsuper-
cell. These boundariesact as a sourceof baroclinically
the front flank notchdownwindfrom the pendantechowas generated horizontal vorticity, similar to numerical simula-
observedat low levels and cappedby midlevelecho over- tions shown by Klemp [1987] and observationsdocumented
hang. This echo feature appeared to be a reflection of the by Maddox et al. [1980] and Purdom and Sinclair [1988].
updraft region and is consistent with observations recorded The alignmentof all three supercelIsand their boundaries
by D85, D90, and M90. At 0019, approximately35 min after also served to inhibit the transport of very unstable air
this image,supercellI spawnedits first of a family of eight northward across parts of central Indiana after 0100. The
tornadoesnearMontgomery,Indiana, approximately100km mappingof tornadicdamagetracks also supportsthis claim
north-northeast of Evansville, Indiana (EVV). The first since tornadic activity was absent east and just south of
tornado reached F3 intensity, while subsequenttornadoes Indianapolis after 0115.
ranged from F2 to F4. Detailed low-level and elevated reflectivity scans of
A broad view of radar reflectivitiesof three large super- storms J and K at 0235 illustrate some aspects of their
cells (I, J, and K) at 0235 is shownin Figure 8. During the vertical structure(Figure 9). Both stormsrevealed a general
previous two hours each cell producedfamilies of tornadoes elongatedecho structure with multiple 50-dBZ reflectivity
varying from F1 through F4 intensity and continued to cores similar to storms D and F and observations docu-
spawn tornadoes through 0500. The radar presentation mented by Nelson and Knight [1987]. Notches linked to
showedthat each storm exhibited multiple high-reflectivity 50-dBZ coresand boundedby echo aloft alongthe southern
cores, notches along the southern and eastern flanks, and an and easternflanks of each cell suggestedthe possibleexist-
elongatedecho structure. The storm locationsaligned in a ence of multiple B WERs and WERs. At this time, an F3
northeast-southwest orientation, suggesting each storm intensitytornado(T) was locatednear the fight rear flank of
moved along the outflow boundary of its downwind neigh- stormJ's 50-dBZ core. The tornadoproduceda damagepath
bor. Essentially, storm K traveled along the outflow bound- of 54 km and had a life spanof 42 min. It is noteworthy that
ary producedby storm J, and storm J moved northeastalong storm J had two probable updraft centers, one near the
the outflow boundary generated by storm I. This type of forward flank capped by 50-dBZ core aloft and another near
storm movement is similar to that observed by Weaver and the fight rear flank of the storm (near location of
 PRZYBYLINSKI ET AL. 247

//

150 km
1)0km 50km 1• IblD
O CommaEcho

Fig. 8. Same as Figure 3 except for 0235 UTC, June 3, 1990.

Further upwind, storm K exhibited a weak echo notch Figure 10 shows low-level and elevated scansof storm I at
along the forward flank of the southernmost50-dBZ core, 0235. A small comma-shapedecho structureand pronounced
signifying a probable updraft center. The adjacent notch RIN observed through 2.5ø near the northern flank of the
downwind of the first was probably linked to the downwind storm are reflectionsof a weakeningcirculationand probable
50-dBZ core. Supercell K spawned a strong F2 intensity descendingflow. An F2 intensity tornado lifted near the
tornado at 0300 having a damagetrack of 9.3 km (approxi- comma head at approximately 0225. Near the right rear flank
mately 50 km north northwest of Louisville) and a F3 of storm I, the reflectivity structure at elevated scans sug-
intensitytornadoat 0340 approximately100km northeastof gests the probable presence of a BWER. A strong F2
Louisville. intensity tornado (T) was located within the BWER at this
Both convective storms J and K were variations of HP time. The high reflectivities and reflectivity gradients sur-
supercelIsin which each cell exhibited front flank notches rounding the weaker echo within the center of the overall
(updraft centers), heavy precipitation along the right rear circular echo structure at all three scans appear to suggest
flank, and multiple 50-dBZ reflectivitycoreswithin an elon- the BWER evolving through a period of collapse. The
gated echo structure. reflectivity patterns suggestthat the mesocyclonewas
248 IDENTIFYING SUPERCELLS
WITH VOLUMETRICRADARDATA

ß
fB-•100km
100kin.
•
/ 150km
30

B R

echo overhang
K
150 km FFN
echo
overhang 100km
Comma
sh.ape .,1• !50km
ecno /,•
3

FFN

100kin
notch Comma
shape • \ 150kin

Fig. 10. Same as Figure 9 except for convective storm I.

notc•
one elevation angle. In an attempt to show a crude three-
Fig. 9. Detailed low-level and elevated plan position indicator dimensionalview of the HP supercelIs,conceptualHP storm
radar refiectivitiesfrom Indianapolis radar at 0235 UTC, June 3, modelsare introduced in Figure 11. The modelsare basedon
1990,of convectivestormsJ and K. Reflectivity contoursare 18, 30, a limited sample of 15 HP storms collected over the lower
41, and 46 dBZ. Shadedregionsrepresentreflectivityvaluesgreater Ohio and mid-Mississippivalley regionsduring the period
than 50 dBZ. T signifieslocationof tornado. EL representselevation
angles during tilt sequence.
1988-1991. The conceptual models are preliminary and con-
tinue to be tested and modifiedwith updatedradar observa-
tions.
rounded by significantprecipitation. This claim was further
supported by several witnesses who experienced heavy
rainfall and hail with the tornado. The total damagetrack of TABLE 1. Storm Identification Letters and Type of Supercell
this tornado was 47 km. Additionally, storm I spawned a Structure Observed During the June 2, 1990, Tornado Outbreak
strongerF4 intensity tornado at 0310. The reflectivity struc-
Storm Type of Supercell
tures at low-level and elevated scans suggestthat this storm
was another variation of an HP supercell. D C-HP
Table 1 reveals the phase of the identified tornadic storms. F C
G C
It is noteworthy that many of the tornadic storms evolved
H C-HP
into the classic supercell category during the early part of I C-HP-BE
their life cycle and gradually evolved into the HP storm J C-HP
phase during the later part (evening hours). K C-HP

4. NEW CONCEPTUAL HP STORM MODELS C representsclassicsupercellstructures.C-HP signifiestransition

from classicto high-precipitationsupercell. C-HP-BE signifiestran-
Conceptualmodels of HP storms presentedby M90, D85, sition from classic to high-precipitation supercell to bow echo
and D90 show evolutionary changesof the PPI radar echo at
 PRZYBYLINSKI ET AL. 249

5. SUMMARY
Inflow
Inflow
notch

notch A major
tornado
outbreak
occurred
across
much
ofcentral
L and
southern
Indiana
during
the
late
afternoon
and
evening
of June 2, 1990. Thirty-eight tornadoes were identified over
a 6-hour period [Storm Data, 1990]. During the early part of
2•s•otheoutbreak,
STRUCTURE
asmany
asseven
supercelIs
wereidentified,
echo overhang some showing variations of classic supercell structures.
echo overhang However, during the latter stagesof the outbreak, many of
rear
the storms evolved to HP supercell structures. The align-
ment of three supercelIs across southern Indiana suggested
that each storm traveled along the outflow boundary of its
downwind neighbor. This kind of storm configurationserved
Inflow
notch
to inhibit the transport of very unstableair northward across
(M) parts of central Indiana duringthe later part of the evening.
Two high-precipitation(HP) supercellconceptualmodels
front weak LOW-LEVELwere derived from 15 cases. Several cases included volumet-
fi•k ec• •DAR ECHO
noah, notc•s ric reflectivity data. These models are additions to the
current conceptual models presented by M88, M90, and
Fig. 11. Conceptualmodelsof plan positionindicatorradar echo
D90. Both models revealed several important echo charac-
structures(at low and middle levels) of high-precipitationsupercelIs
during their mature phase. Contours are schematicreflectivities, teristicsincluding(1) a broad or large hook echo at low levels
with the shading showing high-reflectivity regions. M signifies along the right flank of the storm, (2) one or severalweak
probable location of mesocirculation. Storm motion is assumedto echo notches,signifyingan updraft center, boundedby high
be to the right.
reflectivitiesaloft along the forward flank of the storm, (3) a
singularrear inflow notch (RIN) havingvertical depth along
the trailing edge of the storm suggestingthe presence of
One model in Figure 11 (left side) shows that the parent descendingflow, and (4) a spiral echo structure at low and
mesocycloneand tornado are surroundedby highreflectivity sometimesmiddle levels indicating a mesocirculation.
(wrapped in significantrainfall and hail). During this phase, The conceptualmodels presentedhere and the models
the overall echo pattern often takes the form of a large introducedby M88, M90, and D90 are being examinedand
comma-shaped echo configuration. The notch observed testedwith casesfrom other parts of the country. The testing
along the forward flank of the storm is often boundedabove includes the collection of new radar observations (including
by high reflectivities at midlevels. Such echo features sug- Dopplerdata)to determineif the modelsare consistent with
data or if there is a need to alter the models in light of new
gest the presenceof an updraft center along the forward
flank of the storm. A singular weak echo notch along the evidence.The supportingevidencecollectedthusfar hasnot
trailing flank (RIN) is often observedwithin the lower and been sufficientto conclusively validate the conceptualmod-
els.
sometimes middle levels of the storm. This echo feature
The radar observations shown during the June 2, 1990,
suggeststhat strongdescendingflow may be present.M90,
P89, and P90 have documentedthe occurrenceof damaging outbreaksupportthe hypothesisthat there are many varia-
winds in the vicinity of this echofeature. Reflectionsof the tionswithin the classicand HP supercellthemes.Identifying
parentcirculation(mesocyclone) mayat timesbe outlinedby severalsupercelIsin real time may often be difficultdue to
a spiral echo configuration.Additionally, more than one the variety of echo structuresand storm evolutions.How-
high-reflectivitycore may be presentwithin the overallever,persistent
researchutilizingvolumetricreflectivityand
stormstructuresuggesting
thepossiblepresenceof multiple Doppler velocity data will hopefullyprovidesolutionsfor
updraftcenters.A data set of 10 stormshas beendocu- understandingthe various supercellevolutions.The new
WSR-88D radar systemswill allow warningforecastersto
mented to support this model.
Figure 11 (fight side) showsthe conceptualmodelof a surveythe low-level and elevatedreflectivityand velocity
structures of several storms more efficiently than they can
secondtypeof HP storm.Thistypeof stormis composed of
several high-reflectivitycores within an elongatedecho with today'smethods.However, researchers and forecast-
structure. One or several weak echo notches,signifyingthe ers will face the challengeof properlyinterpretingthesenew
presenceof updraftcenters,are notedalongthe right side data setsto producethe mostaccurateandtimelywarnings
and forecasts.
and forward flank of the storm; these notches are often
cappedby highreflectivitesaloft. This type of stormstruc-
ture is similar to structures documentedby Nelson and
Acknowledgrnents.The authorsare gratefulto JosephT.
Knight[1987]in whichoneor severalweakechoregionsare Schaefer,DirectorNWSTC, RichardLivingstonof ScientificSer-
present.Only a smalldataset, comprising
five storms,has vicesDivision(Central Region),Don Burgessof the Operational
been documentedto supportthis model. SupportFacility,DavidAndra,WSFO Norman,andFred
250 IDENTIFYING SUPERCELLS WITH VOLUMETRIC RADAR DATA

WSFO Saint Louis, for their many beneficialsuggestions on improv- technique for radar warnings, in Preprints, loth Conference on
ing the manuscript. The authors would also like to thank Al Moller Severe Local Storms, pp. 77-80, American Meteorological Soci-
for his review of the paper and acknowledge Steve Thomas ety, Boston, Mass., 1977.
(MIC/AM WSFO Saint Louis) for his encouragementin this study. Lemon, L. R., Severe thunderstormradar identificationtechniques
Additional thanks goes to Jim Allsopp (WPM) WSFO Chicagoand and warning criteria, NOAA Tech. Memo. NWS NSSFC-3, 60
Phil Manuel and Jim Keene of WSO Cincinnati for their contribu- pp., Natl. Oceanic and Atmos. Admin., Boulder, Colo., 1980.
tions. This paper is part of a COMET Partner's project between (Available as NTIS PB-273049, Natl. Tech. Inf. Serv., Spring-
Purdue University and WSFO Indianapolis.This paperwas partially field, Va.)
funded by UCAR/COMET. Maddox, R. A., L. R. Hoxit, and C. F. Chappell, A study of
tornadic thunderstorm interactions with thermal boundaries,
REFERENCES Mon. Weather Rev., 108, 322-336, 1980.
Moller, A. R., and C. A. Doswell III, A proposed advanced storm
Bluestein, H. B., and C. R. Parks, A synoptic and photographic spotter's training program, in Preprints, 15th Conference on
climatology of low-precipitation severe thunderstorms in the Severe Local Storms, pp. 173-177, American Meteorological
southern plains, Mon. Weather Rev., 111, 2034-2046, 1983. Society, Boston, Mass., 1988.
Browning, K. A., Airflow and precipitation trajectories within Moller, A. R., C A. Doswell, and R. W. Przybylinski, High-
severe local storms which travel to the right of the winds, J. precipitationsupercelIs:A conceptualmodel and documentation,
Atmos. Sci., 21,634--639, 1964. in Preprints, 16th Conferenceon Severe Local Storms, pp. 52-57,
Browning, K. A., and F. H. Ludlam, Airflow in convective storms, American Meteorological Society, Boston, Mass., 1990.
Q. J. R. Meleorol. Sot., 88, 117-135, 1962. Nelson, S. P., The hybrid multicell-supercellstorm---An efficient
Burgess, D. W., R. A. Brown, L. R. Lemon, and C. R. Safford, hail producer,II, General characteristicsand implicationsfor hail
Evolution of a tornadic thunderstorm, in Preprints, loth Confer- growth, J. Atmos. Sci., 44, 2060-2073, 1987.
ent'e on Severe Local Storms, pp. 84-89, American Meteorolog- Nelson, S. P., and N. C. Knight, The hybrid multicellular-supercell
ical Society, Boston, Mass., 1977. storm•An efficient hail producer, !, An archetypal example, J.
Burgess, l). W., V. T. Wood, and R. A. Brown, Mesocyclone Atmos. Sci., 44, 2042-2059, 1987.
evolution statistics, in Preprints, 12th Conference on Severe Przybylinski,R. W., The Raleightornado--28 November, 1988,A
Local Storms, pp. 422-424, American Meteorological Society, radar overview, in Preprints, I2th Conference on Weather Fore-
Boston, Mass., 1982. casting and Analysis, pp. 186-191, American Meteorological
Chisholm, A. J., Alberta hailstorms, 1, Radar case studies and Society, Boston, Mass., 1989.
airflow models, Meteorol. Monogr., 14, 1-36, 1973. Przybylinski, R. W., S. Runnels,P. Spoden, and S. Summy, The
l)oswell, C. A., Ill, The operational meteorology of convective Allendale, Illinois tornado--7 January, 1989: One type of an HP
weather, vol. 2, Storm scale analysis, NOAA Tech. Memo. ERL supercell,in Preprints, 16th Con)•rence on SevereLocal Storms,
ESG-15, 240 pp., Natl. Oceanic and Atmos. Admin., Boulder, pp. 516-521, American MeteorologicalSociety, Boston, Mass.,
Colo., 1985. 1990.
Doswell, C. A., A. R. Moller, and R. W. Przybylinski, A unified set Purdom,J. F. W., and P. C. Sinclair, Dynamicsof convectivescale
of conceptualmodelsfor variationson the supercelltheme, in interactions,in Preprints, 15th Con)'?renceon Severe Local
Preprints, 16th Confi,rence on Severe Local Storms, pp. 40-45, Storms,pp. 354-359, AmericanMeteorologicalSociety, Boston,
American MeteorologicalSociety, Boston, Mass., 1990. Mass., 1988.
t;oote, G. 13., and H. W. Frank, Case study of a hailstorm in Storm Data, Natl. Clim. Data Center, Natl. Oceanic and Atmos.
Colorado, I11, Airltow from triple-Doppler measurements, J. Admin., Asheville, N. C., June 1990.
Atmos. Sci., 40,686-707, 1983. Wakimoto, R. M., The West Bend, Wisconsin storm of 4 April,
Forbes, G. S., On the reliability of hook echoesas tornado indica- 1981: A problem in operationalmeteorology,J. Clim. Appl.
tors, Mon. Weather Rev., 109, 1457-1466, 1981. Meteorol., 22, 181-189, 1983.
Klemp, J. B., Dynamics of tornadic thunderstorms,Annu. Rev. Weaver, J. F., and S. P. Nelson, Multiscale aspectsof thunderstorm
b7uid Mech., 19, 369-402, 1987. gustfrontsand their effectson subsequent
stormdevelopment,
Lemon, L. R., Severe thunderstormevolution: Its use in a new Mon. Weather Rev., 110, 707-718,
 Doppler Radar Identification of Nonsevere Thunderstorms That Have the
Potential of Becoming Tornadic
RODGER A. BROWN

National SevereStormsLaborato•3',NOAA, Norman. Oklahoma 73069

1. INTRODUCTION On a day with severethunderstormand tornadopotential

the forecaster ideally would use the WSR-88D to identify
Doppler weather radar has been an important research candidate severe storms before the formation of mesocy-
tool during the past 25 years in our efforts to understandthe clones. In some instances,a precursorfeature is evident in
evolution of severe thunderstorms. Research starting in the the Doppler radar data 30 rain or more beforethe appearance
late 1960s at the Air Force Cambridge Research Laborato- of the mesocyclone/supercell storm. This feature is a vortex
ries (now the GeophysicsDirectorate of U.S. Air Force's pair occurringat middlealtitudeson the downstream
flanks
Phillips Laboratory) and in the early 1970sat the National of a strong updraft, with cyclonic vorticity on the right
Severe Storms Laboratory (NSSL) led to the identification forward flank and anticyclonicvorticity on the left forward
of the Dopplervelocitysignaturefor a rotatingupdraftcalled flank. In this paper a controversialmechanismis proposed
a mesocyclone[e.g., Donaldson, 1970;Brown et al., 1973]. for the evolution of the vortex pair into a mesocyclone.Data
Becausethe researchresultssuggestedthat a mesocycloneis then are presentedin supportof the proposedevolution
a precursorfor tornadoformation[e.g., Burgess,1976],the process.
JointDopplerOperationalProject(JDOP)wasconductedby
the National Weather Service, U.S. Air Force, and Federal
2. PROPOSED MECHANISM FOR
Aviation Administration during 1977-1979to test the opera-
MESOCYCLONE iNITIATION
tional significance
of the researchfindings.The JDOP oper-
ation indicated that nearly all the stormswith mesocyclone Giventhe presenceof a middle-altitudevortexpair on the
signaturesproducedsurfacedamage,aboutonehalf of them leftandrightforwardflanksof a strongupdraft,a mechanism
producedtornadoes,and the existenceof a mesocyclone is proposedto explainthe transitionof thecyclonicportion
signaturepermittedthe leadtimefor tornadowarningsto be of the vortex pair into a mesocyclone.This mechanism,
increasedby an averageof 20 min [Joint Doppler Opera- whichincorporates portionsof a hypothesisproposedby
tional ProjectStaff, 1979].Sincethe JDOP test alsoindi- Fujita and Grandoso[1968],differsfrom the one that is
cated other operationalbenefitsfor Doppler radar, the currentlyacceptedin the meteorological literature.Empiri-
decisionwas made to procurea nationalnetworkof Next cal evidence for the mechanism is based on the sole dual-
Generation Weather Radars (NEXRAD) incorporatingDop- Dopplerradarstudythat providesdetailedtemporalresolu-
pler capability(now calledWSR-88D). tion of the transitionprocess[Brow'n,1989].
In a single-Dopplerradar study of 41 mesocyclones, The Fujita-Grandoso hypothesisstartswith a developing
Burgesset al. [1982]foundthat the mesocyclone withina storm that has a strong nonrotatingupdraft. A middle-
supercellstorm undergoesa distinctevolution.When a altitudepairof counter-rotating vorticesformsontheflanks
thunderstormfirst becomesorganizedinto the supercell of the updraft.As the expandingsurfacegustfront moves
stage,updraftrotationexistsonlyat middlealtitudes. How- beneaththe vortex pair, convergence alongthe gustfront
ever, as the stormreachesmaturity,rotationis evidentfrom forcesa newupdraftto form beneathbothmembers of the
nearthe groundto the upperportionsof the storm.Torna- pair. Rotunnoand Klemp [1982]proposed that middle-
doeswereproduced duringthematurestageofabout40%of altitudelow pressureassociatedwith eachvorticitycenter
the mesocyclonesstudied. produces
anupwarddirectedperturbation
pressure
gradient
force that further enhances vertical motion at those loca-
The Tornado:Its Structure,Dynamics,Prediction,and Hazards. tions.
GeophysicalMonograph79
Thispaperisnotsubject
to U.S. copyright.
Published
in 1993bythe In the more typical severethunderstorm
situationthe
American GeophysicalUnion. significant
updraftformsonlyontherightflank,in response
252 IDENTIFICATION OF POTENTIALLY TORNADIC STORMS

4O
to storm-relative, low-altitude flow that approaches the
storm from the right. As the new updraft developsabovethe STORM-RELATIVE HODOGRAPHS
right flank gust front, it entrains the ambient vertical vortic-
ity and starts to rotate cyclonically at middle altitudes as 13 12
vorticity is stretched in the vertical. The processof discrete
updraft propagation therefore is responsiblefor the storm's
new updraft starting to rotate. Subsequent right flank up- •
20-- 14
km• 6 7._a• •o
1
.v....... 2 37
drafts likewise develop rotation through the entrainment of ..... ....
ambient vertical vorticity. Increasing precipitation produced
by each new updraft blends with the decreasingprecipitation ...........
2• >0 1•'0 [ 10
a20 30 40 510
Ill 8'1
from the previous dying updraft, resulting in a low- and
+A
middle-altitude radar reflectivity pattern that appears to + 10
.... Agawam Storm (6-6-79)
B,G
move to the right in a continuous manner. ......... Billings and Guthrie
Storms (4-26-84)
This hypothesis is distinctly different from the currently
20
popular hypothesis that states that any strong updraft in an
environment with low-altitude streamwise horizontal vortic-
Fig. 1. Storm-relative hodographsfor June 6, 1979 (solid) and
ity (that is, strong clockwise curvature in a storm-relative April 26, 1984 (dashed). Plus signs indicate the origins of the
hodograph) should rotate becausethe horizontal vorticity is respective ground-relative hodographs. The June 6, 1979,
drawn into the updraft and tilted into the vertical [e.g., hodographis a compositeof the 1500CST rawinsondeobservations
Browning,,and Landry, 1963; Barnes, 1970; Rotunno, 1981; from Fort Sill and Norman, Oklahoma. The April 26, 1984,
hodographis a compositeof the 1556 and 1806 CST rawinsonde
l)avies-Jones, 1984]. Data discussed in the next section
observations from Chickasha, Oklahoma.
indicate, however, that a strong nonrotating updraft can
exist within an environment having low-altitude streamwise
horizontal vorticity.
dual-Doppler area covered by the Norman radar and the
Cimarron radar located 41 km to the northwest of Norman.
3. OBSERVATI()NAI• EVIDENCE Dual-Doppler data collectioncommencedat 1434 and con-
tinuedthrough1550.A storminterceptteam reportedhail 26
Dual-Dopplerradar measurementspermit the identifica- mm in diameter beneath the storm at 1535.
tion of updraft locations and strengths and of vorticity At 1434, a new nonrotatingupdraft (called U l) was just
regions within a storm. Doppler radar evidencefor the developingwithin the storm. A middle-altitudepair of ver-
middle-altitude vortex pair being a precursor signaturefor tical vorticity maxima was found on the right and left
mesocycloneinitiation is presented for three Oklahoma forward flanks of the updraft (Figure 2). As the updraft
supercellthunderstorms,one of which did not evolve be-
yondthe organizingstage;dual-Dopplerdatawerecollected
only during portionsof the storms' life cycles. The storms
-20:
are the Agawamhailstormof June6, 1979,[Brown,1989]and
the Billingshailstormand Guthrietornadicstormof April 26,
Vorticity
:1:5
x I0'* s-•
1984 [Burgessand Curran, 1985;Bluesteinand Woodall,
<t -3o
Nonrotating
Updraft
UI
19901.
Storm-relativehodographsfor the environmentswithir• o
which the storms formed are shown in Figure 1. The z
- 1502 • 1510
environmenton both daysexhibiteddirectionalshearof over o -40
90øin the lowest 3 km, minimal shear in the 4- to 6-km layer, 1502

and primarilyspeedshearabove 6 km. With the storms

movingfrom the southwesttowardthe northeast,the mag- z

nitude of the low-altitude (0-2 km), storm-relative flow !-- -50

approachingthe rightflanksof the stormswasnearlytwice 1434

{ AGAWAM
6 JUNE STORM
1979
as greaton April 26, 1984,as on June6, 1979.Thus one /4,:34
mightexpectstrongerstormson April 26, 1984,owingto -60 • • I • ! •, ! •
-60 -50 -40 -30 -20
greaterconvergence alongthe storms'gustfrontsandstron- DISTANCE FROM NORMAN (krn)
ger perturbationpressureforces.
Fig.2. Composite
of vertical
vorticity
(contours
of +-5x 10-3
s-1) relativeto updraftU1 at a heightof 5.5 km in the Agawam
3.1. Agawam,Oklahoma,Hailstormof June6, 1979 storm[afterBrown, 1989].Counterclockwise arrowsindicatecy-
clonicvorticity,andclockwise
arrowsindicateanticyclonic
vortic-
The Agawamstormfirstappeared
on radarat 1325CST ity.Vertical
velocities
m15ms-• within
theupdraft
areshaded.
The
nearFort Sill asthe right-handmemberof a pairof adjacent timeintervalbetweensuccessive
updraftandvorticitypositionsis 4
storms.The stormmoved to the northeasttowardthe NSSL
 BROWN 253

_ AGAWAM
STORM
-45 6 JUNE 1979
1434-1550 CST

-50

-55

C) Updraft
Initiation
(10ms
• Right-Flank
GustFront
Fig. 3. (a) Storm-relative
wind streamlines
and (b) ground- ,:¾11'•'•!!i
",',•
Middle-Altitude
Vorficity
relativewindspeedsat 5.5-kinheightin theAgawam stormat 1446
CST [fromBrown,1989].Dashedreflectivity
Updraftregionsare stippled,anddowndraft
outlineis 20 dBZ.
regions are hatched.
• , • • • ,,•(5,
10x10'•s'•)
.... I , 1.... I . •_

-60 -50 -40 -30 -20

UpdraftU1 (labeled1) is thematurenonrotating updraft,U0 is a DISTANCE
FROMNORMAN
DOPPLER
RADAR(kin)
dyingleftflankupdraft,
andUX isa newrotating rightflankupdraft.
Fig. 4. Juxtapositionof the right edge of the surfacegust front
boundary(heavy curves), location of updraft initiation (circled
numbersor letter), and time-spacepresentationof maximumcy-
moved to the northeast,the vortex pair remainedon the clonic vertical vorticity at 5.5-km height (shadedregion) in the
forwardflankof the updraft.UpdraftU 1 reachedmaximum Agawamstormbetween1434and 1550CST. The spatialorientation
strength(44ms-1) at 1446,afterwhich timethepairmoved of the verticalvorticitydata wasnormalto the meanwind passing
slowly downstreamrelative to the updraft. By 1510 the throughthe locationof the strongestupdraftat that time. Owingto
the lack of dual-Doppler radar data at 1530 and 1534, outflow
updraft'smaximumverticalvelocitywaslessthan10m s-1 , boundariesare missingand vorticity is interpolatedfor thosetwo
and the magnitudeof vorticity within both membersof the times. Times on the right are for updraftinitiation, and times on the
pair wasdroppingbelow5 x 10-3 s-•. left representthe gust front locationsat the times of updraft
initiation.
The middle-altitude flow field associated with the vortex
pair at 1446is shown in Figure 3. The basicstorm-relative
flow in Figure 3a divergedaround updraft U1. The air
the cyclonic member of the middle-altitudevorticity pair.
closestto the updraftconvergedinto the wakeregionbehind
Thus it appearsthat the interactionbetweenthe gustfront
the updraft where the associateddowndraftwas located. The
and storm-relative inflow produced a zone of favorable
ground-relativewind speedsin Figure 3b were a minimumin
convergence,whereas the upward directed pressuregradient
the wake region,beingup to 12 m s-• weakerthan the
force inducedby low pressureat the vorticity center dictated
undisturbed environmental flow of 19 m s-• at the same
where the updraft would form along the gust front [e.g.,
height.The wind speedson the lateralflanksof U 1 were up Fujita and Grandoso, 1968;Rotunno and Klemp, 1982]. As
to 7-18 m s-• stronger thantheundisturbed environmentalthe updraft grew vertically, it entrained and stretched the
flow. The wind speedpatternin Figure 3b is the sameas the
surroundingvertical vorticity and started to rotate cycloni-
single Doppler velocity signaturefor a vortex pair when cally at middle altitudes.
viewed from the southwestor northeastby a Doppler radar
One can speculatethat the Agawamstormdid not develop
[e.g., Brown and Wood, 1991].
beyond the organizing supercell stage becauseof the rela-
As indicated in Figure 3, both wind direction curvature
tively rapid formation of new updrafts at 10- to 15-min
and lateral speedshear contributedto anticyclonicvorticity
intervals (compared to 40-50 min for the mature stage)that
in the vicinity of preexistingupdraft U0 on the left flank of did not permit time for rotation to develop in the lower
U 1 and to cyclonic shear in the vicinity of new updraft UX portions of the updraft. However, little is known about the
on the right flank. The new updraft was a short-lived and organizing stage of supercell storms owing to the lack of
relativelyweak(maximumof 22 m s-]) secondary
updraft, other dual-Doppler radar case studies.
but it had the distinction of being the first in a seriesof fight
flank updraftsto develop middle-altituderotation.
The factors contributing to the initiation of updraft rota- 3.2. Billings and Guthrie, Oklahoma, Storms
of April 26, 1984
tion are presentedin compositeform in Figure4. Shownare
the initiation locations for updrafts U l-U5 and UX relative At 1555CST on April 26, 1984,four radar echoesappeared
to the fight side of the surfacegust front and to the pattern of along the dry line in west central Oklahoma; these storms
cyclonic vertical vorticity at middle altitudes. Each updraft have been studied by Burgess and Curran [1985] and
formed along the right side of the gust front where the Bluestein and Woodall [1990], and additional information
low-altitude, storm-relative inflow (Figure 1) was essentially about them was obtained from photographsof the Norman
normal to the frontal surface, producing convergence. The Doppler velocity and reflectivity displays. Initially, these
locationof updraftinitiationalongthe gustfront wasbeneath storms were low-precipitation (LP) storms because
254 IDENTIFICATION OF POTENTIALLY TORNADIC STORMS

was little visual precipitation below cloud base [Bluestein

6 June 1979 • Agawam, Okla. Storm
and Woodall, 1990]. The middle two storms were the stron-
ger, and the single-Doppler velocity signature of a vortex ?Vortex• OS DS
pair (similar to the pattern in Figure 3b) quickly appeared at
First UX U2 U3 U4 U5
middle altitudes in both of them. At 1636, dual-Doppler Echo
• movement to right
analyses of the southern storm by Bluestein and Woodall
[1990] revealed the presence of new right and left flank 26 April 1984 Billings, Okla. Storm
updrafts, which indicated that the storm was ready to split
Vortex o$
into right- and left-moving portions (beginningof the orga- ;•( .............. ?
1• Pair
nizing supercellstage). The right flank updraft was rotating First
right-movingmember
of splittingstorm
cyclonically at middle altitudes, and the left flank updraft Echo

was rotating anticyclonically. By that time, the updraft i

responsiblefor producingthe vortex pair had decayed,and 26 April 1984 • Guthrie, Okra. Storm
only the associatedprecipitationdowndraftremained.Radar
i

['effectivitypatternsaloft in the northernstormindicatedthat

Vortex
•
;!.' OS ; (, ,, MS ?+D_S.
it also was startingto split at that time. Duringand following 1'" pair •right-moving
member
First
Echo
• of splittingstorm
the splittingprocess,the northernright-movingstorm, called • intermittent tornadoes
the Billingsstorm, produceddamagingsurfacewinds and
......................... I ....... I ...... I, , I,

-90 -60 -30 0 30 60 90 120

hail as largeasgolfballs [National Oceanic'andAtmospheric TIME RELATIVE TO INITIATION OF UPDRAFT ROTATION (rain)
Admi/#stration (NOAA), !984]; after 1730, it moved beyond
the first trip range from the Norman Doppler radar. The Fig. 5. Temporalvariationof basicfeatureswithin the Agawam,
southern right-moving storm, called the Guthrie storm, Billings,andGuthriesupercellthunderstorms. Whenknown(solid
produceddamagingsurfacewinds and hail up to baseballs lines) or estimated(dashedlines), the durationsof the organizing
and grapet'ruitin size. Intermittentdamagel¾omtwo torna- (OS), mature (MS), and dissipatingstages(DS) of the storms'
mesocyclonesare indicated.
doeswas reportedfrom Guthrienortheastwardto Stillwater
between 1753 and 1906 [NOAA, 19841. The important tran-
sition between the organizing and mature supercell stages
tatkingplacewithinthe Guthriestormbetween1640and 1720 ity that wasentrainedandstretchedby the growingupdraft
could not be documented with dual-Doppler measurements to producemiddle-altitude
rotation[e.g., Fujita and Gran-
becausethe stormwas crossingthe baselinepassingthrough doso, 1968].
the Norman and Cimarron Doppler radar sites.Owing to the Thisexplanation for theinitiationof updraftrotationis in
availabilityof dual-Dopplerradardata,it wasnot contrast with the popular explanationthat employs the
int'requent
possible
to document of updrafts verticaltiltingof low-altitude
theinitiationandevolution streamwise
horizontalvorticity
and their rotational characteristics within the Billings and in the inflowair as it risesin the updraft[e.g., Browningand
Guthrie storms. Landry, 1963;Barnes,1970;Rotunno,1981;Davies-Jones,
1984].it can be noted in Figure 1 that the storm-relative
winds in the lowest 3 km of the environment on both storm
4. CONCLUDING DISCUSSION
days were nearly parallelto horizontalvorticity vectors
On the basis of this limited survey of only three severe (normalto the hodograph curve,pointingto the left), in
thunderstorms, it appearsthatthe presenceof a vortexpair agreementwith the popularexplanation.However, not
at middlealtitudesholdspromiseas an early indicatorthat a everystrongupdraftin thesestormsrotated;for example,
stormhasthepotentialof becoming severeor tornadic.The theupdraftsproducing thevortexpairsdidnotrotate.The
sequence
of eventstakingplacewithinthethreestormsis firstupdrafts
thatdeveloped rotationappearto bethosethat
summarized in Figure5. A commonfeatureis the middle- were in the vorticityregionof a memberof the middle-
altitude vortex pair that appearedearly in each storm's altitudevortex pair on the lateral flanksof an existing
lifetime.Thereprobablywheresequential updraftsduring nonrotatingupdraft.Thus verticaltilting of low-altitude
the vortexpair stage,but neitherdual-Doppler radardata streamwisehorizontalvorticity does not appearto be the
(June6, 1979)nor analysesfrom the literature(April 26, sourceof initialupdraftrotationin thesestorms.However,
1984)occurredat suitabletimesto documentsuchan evo- streamwisevorticitymay play somerole duringthe meso-
lution. cyclone's
mature stage,whenupdraftrotation extends
from
The vortexpairappearsto haveplayedtworolesin the nearthe groundto the upperportionsof the storm.
development asso- The summary
of a rotatingupdraft.First,lowpressure in Figure5 indicatesthatthe portionsof
ciatedwith eachmiddle-altitudevorticitycenterproducedan mesocyclone/supercell
evolutionthatcouldbedocumented
upwarddirected pressuregradient forcethatenhancedup- fit the conceptual
modelof Burgesset al. [1982].Initial
draft formationalongthe surfacegustfront beneaththe updraft rotation
occurred onlyat middlealtitudes
(organiz-
verticalvorticitycenter[e.g.,RotunnoandKlemp,1982]. ingstage).
The Guthrie stormwas the only oneto havea
Second, theverticalvorticityfielditselfprovided
thevortic- documented mature stagewith rotationextendingto
 BROWN 255

ground.
Although
all threestorms
produced
damaging
hail Bluestein,H. B., and G. R. Woodall,Doppler-radaranalysisof a
andthetwoonApril26, 1984,produced
damaging
winds, low-precipitationsevere storm, Mon. WeatherRev., 118, 1640-
only the Guthriestormproducedtornadoes. 1664, 1990.

To properlyinterpret
verticalvorticitypatterns
withina Brown, R. A., Initiationandpropagation
of thunderstorm
mesocy-
clones, Ph.D. dissertation, 321 pp., Univ. of Okla., Norman,
thunderstorm,
it is importantto knowthe relativelocations 1989.
of updrafts.Single-Doppler
radardataprovide
two indica- Brown, R. A., and V. T. Wood, On the interpretationof single-
tionsof updraftlocation,andbotharesituatedintheupper Doppler velocity patterns within severe thunderstorms,Weather
portionsof the storm.Oneis the single-Doppler velocity Forecasting, 6, 32-48, 1991.
signature [e.g.,BrownandWood,1991]near Brown, R. A., D. W. Burgess, and K. C. Crawford, Twin tornado
of divergence
cycloneswithin a severe thunderstorm:SingleDoppler radar
the top of the updraft.The otheris the radarreflectivity observations, Weatherwise, 26, 63-69, 71, 1973.
maximum(upwardbulgeat stormtop) that marksthe Browning, K. A., and C. R. Landry, Airflow within a tornadic
presenceof precipitation
particles
at thetopofthe updraft. storm, in Preprints, loth WeatherRadar Conference,pp. 116-
Duringthe vortexpair stage,the updraftis centeredat the 122, American Meteorological Soc., Boston, Mass., 1963.
Burgess,D. W., Single-Dopplerradar vortex recognition:Part I:
upwindendof thepair(Figure2). Duringthemesocyclone/ Mesocyclonesignatures,in Preprints, 17th Conferenceon Radar
supercellstage,the updraftis coincident witha vorticity Meteorology,pp. 97-103, AmericanMeteorologicalSociety,Bos-
maximum,markinga rotatingupdraft. ton, Mass., I976.
The few single-or multiple-Dopplerradarstudiesof su- Burgess,D. W., and E. B. Curran,The relationshipof stormtypeto
percellstormsthatexistprovideonlyfragmentarydocumen- environmentin Oklahoma on April 26, 1984, in Preprints, 14th
Conference on Severe Local Storms, pp. 208-211, American
tationof the initiationandevolutionof the supercell
stage. Meteorological Society, Boston, Mass., 1985.
The subsetpresentedhere suggests that the presenceof a Burgess, D. W., V. T. Wood, and R. A. Brown, Mesocyclone
middle-altitudevortex pair is a necessaryconditionfor the evolution statistics, in Preprints, 12th Conference on Severe
formation of updraft rotation and the transitionof a storm Local Storms, pp. 422-424, American MeteorologicalSociety,
Boston, Mass., 1982.
intoits supercell
stage.However,a moresystematic
studyof Davies-Jones,R. P., Streamwisevorticity: The origin of updraft
hundreds of severeandnonseverethunderstorms
is required rotation in supercell storms, J. Atmos. Sci., 41, 2991-3006, 1984.
before we will know how often a severe/tornadic storm is Donaldson, R. J., Jr., Vortex signaturerecognitionby Doppler
preceded by a vortex pair and how often a storm with a radar, J. Appl. Meteorol., 9, 661-670, 1970.
vortex pair evolves into a severe/tornadic storm. Fujita, T., and H. Grandoso, Split of a thunderstorminto anticy-
clonic and cyclonic storms and their motion as determined from
numerical model experiments,J. Atmos. Sci., 25,416-439, 1968.
Joint Doppler Operational Project Staff, Final report on the Joint
Acknowledgments.Useful commentson the manuscriptby Doppler Operational Project {JDOP) 1976-1978, NOAA Tech.
Ralph Donaldson, Jr., Michael Eilts, David Stensrud, and an Memo. ERL NSSL-86, 84 pp., Natl. Severe Storms Lab., Nor-
anonymousreviewer are most appreciated.Valuableeditorial sug- man, Okla., 1979. (Available as NTIS PB80-lO7188/AS from Natl.
gestionswere providedby ChristinaThomas.The figureswere ably Tech. Inf. Serv., Springfield, Va.)
prepared by Joan Kimpel. National Oceanic and Atmospheric Administration {,NOAA), Storm
Data, vol. 26, no. 4, pp. 50-51, Asheville, N. C., 1984.
REFERENCES Rotunno, R., On the evolution of thunderstorm rotation, Mon.
Weather Rev., 109, 577-586, 1981.
Barnes,S. L., Someaspectsof a severe,right-moving
thunderstorm Rotunno, R., and J. B. Klemp, The influence of the shear-induced
deduced from mesonetwork rawinsonde observations, J. Atmos. pressure gradient on thunderstorm motion, Mon. Weather Rev.,
Sci., 27, 634-648, 1970. 110, 136-151,
 An Examination
of a Supercell
in Mississippi
Usinga Tilt Sequence
D^VlD A. IM¾

National WeatherServiceForecastOffie, NOAA, Denver, Colorado80239

KEVIN J. PENCE

NationalWeatherServiceForecastOffie, NOAA, Jackson,Mississippi

39288

1. INTRODUCTION Using the low-level reflectivity field alone to determine if

low-precipitationsupercelIs(LP) [Bluesteinand Parks, 1983]
Most of the scientificcommunitynow recognizethat a or high-precipitation supercelIs (HP) [MoIler et al., 1990]
supercellis a stormthat containsa deeppersistentrotating contain circulations is even less reliable than with the classic
updraft or a mesocyclone(i.e., vertical velocity correlated supercell.Hook echoestypically are not observedwith these
with vertical vorticity [Doswell et al., 1990]).Most forecast-
types of supercelIs.The LP supercell often exhibits rela-
ers do not, however, have accessto Doppler data to verify
tively low reflectivities and usually lacks indicationsof a
the existence of a mesocyclonein a storm. Operational
low-level circulationin the reflectivity field. Przybylinskiet
forecastershave relied primarily on low-level reflectivity
al. [1991] and Doswell et al. [1990] noted several variations
signatures such as hook echoes, "distinctive" echoes
in the low-level reflectivity echo shape associatedwith HP
[Forbes, 1981], and notches [Przybylinskiet al., 1990] to
supercelIs. MoIler et al. [1990] noted that the most com-
infer the existence of a mesocyclonein a storm.
monly identified echo configurationsassociatedwith the HP
Relying on low-level reflectivity signaturesfor issuing
tornado warnings, without investigatingthe vertical struc-
supercellincluded the "comma," "kidney bean," spiral,
and "fat hook" echoes.
ture of a storm, many times results in missed tornadic
events. The classic supercell is often identified on radar by National Weather Service forecastersin Mississippi[i.e.,
the existence of a hook or distinctive echo in the lower Imy, 1986] noted differences in the low-level reflectivity field
portions of the storm. Browning [1964] found that the associated with supercelIs in Mississippi from those de-
three-dimensional structure of the classic supercell was scribed by Browning. In summary, hook echoeswere rarely
characterizedwith an extensive slopingoverhang, a region observed with these supercelIs. The storms usually were
of weak low-level reflectivity capped by the storm top distinguishedfrom other stormsby their unusuallylarge size
(vault), and an intense hook-shaped echo surroundingthe (greater than 35 km in diameter). They contained an exten-
vault. The vault is now referred to as the bounded weak echo sive area of high reftectivities (video integrator and proces-
region (BWER) [Chisholm and Renick, 1972]. These features sor (VIP 5), 50-56 dBZ). Most of the storms were not
were noted with supercelIsin England as well as Oklahoma isolated but surrounded by light to moderate precipitation.
[Browning, 1963, 1965]. Lemon [1980] used the work of Although most of these supercelIsin Mississippigenerated
Browning and others to develop the Lemon Technique to hail, the hail was not much larger than 2.5 cm in diameter
examine not only the three-dimensional storm structure of and normally affected only small areas. One supercell(Feb-
supercelIs but all severe thunderstorms. The Lemon Tech- mary 28, 1987) spawned an F4 tornado, but a post storm
nique uses the structural relationshipsamongthe low-level survey found no one near or in the path of the tornado who
reflectivity gradient, midlevel overhang, and the location of observedhail prior to or after the tornadic event. SupercelIs
the storm top as warning criteria to determine the storms in Mississippiseemto be more typically of the HP supercell
potential for severe weather and/or tornadoes. variety, except they are not usuallyprolific hailers.
Although the low-level reflectivity field observed with
The Tornado: Its Structure, Dynamics, Prediction,and Hazards.
thesestormsin Mississippidifferedfrom thosedescribedby
GeophysicalMonograph79
This paper is not subjectto U.S. copyright. Publishedin 1993by the Browning[1963, 1965] and Lemon [1980], most of the storms
American GeophysicalUnion. unfortunately were not examined three-dimensionally.On
258 MISSISSIPPI SUPERCELL

March 12, 1986, however, a supercell in Mississippi was

evaluated using a variation of the Lemon Technique. Al-
though the low-level reflectivity characteristics resembled
the HP supercell, the three-dimensional structure was simi-
lar to the classic supercell, with minor deviations. The need
to use volumetric radar data to evaluate the potential sever-
ity of a storm is illustrated with this particular supercell.
Also, an alternative tilt scan strategy is presented,since the
full utilization of the Lemon Technique may at times be too
time-consuming during the warning process. ø I

2. THE SYNOPTIC/MESOSCALE
ENVIRONMENT

At 1200 (all times UTC) on March 12, 1986, a long-wave

midtropospherictrough over the central United States re-
sultedin a strongsouthwesterlyflow alott over Mississippi.
A series of short-wave troughs were rotating through the
base of the long-wave trough and approachedMississippi
from the southwest. The important low-level mesoscale
feature was an outflow boundary/trough that was moving 1400Z 1800Z
slowly castwardacross Mississippi.The outflow boundary
was generatedby a line of thunderstormsthat moved east-
Fig. 1. Outflowboundarylocationsfrom 1000to 1800Z (UT),
ward from Louisiana and Arkansas during the night. This which served as the focus for the tornadic storms on March 12, 1986.
line contained numerous severe thunderstorms that yielded The circle surroundingthe radar representsthe 92-km circle for
hailandwinddamagealongwith a few tornadoes.The lineof references to Figures 3-8.
thunderstormsdissipated around 1200, but the outflow
boundarypersistedand becamequasi-stationarynortheast
to southwest across Mississippi (Figure 1). The 1200 Jackson,Mississippi,hodographin the lower
Surface observationsin Mississippiand southernLouisi- levelswas unrepresentativeof the environmentalwind pro-
anatogetherwith radardata(that includeda line-echowave file east of the boundary. The 1500 wind observation at
pattern,or LEWP), depicteda seriesof mesoscale low- McComb,Mississippi,wasusedto adjustthe hodograph in
pressuresystemsthat movedrapidlynortheastward along the lower 2 km to better reflect the wind conditions on the
thisboundary.As thesesurfaceperturbations
nearedreport- east sideof the boundary(Figure2). The stormmotionof
ing stations,the pressuredroppedand the windsbacked 225øat 23 m s- 1 resultedin a strongstorm-relativeflow from
t¾omthe south to the southeast. Six tornadoes that devel- the northeast in the lower 2 km. Southwest flow aloft,
opedalongtheboundarythisdayappearedto haveoccurred combinedwith the strongnortheaststorm-relativeflow in
nearor slightlynorthof thesesmallareasof low pressure. the lower levels of the environment, indicated that the
Moist low-level south to southeast winds were advecting storm'supdraftflanklikelywouldbe locatedon the south-
air having 15ø-18øC dew pointsnorthwardaheadof the eastportionof the low-levelprecipitationecho.
boundary,while dew points west of the surfacetrough Thepotentialfor mesocyclonesexistedasthe helicityin
lowered to around 12øC. The 1200 Jackson, Mississippi, thelowest
3 kmwas173m2s-2, whichexceeded
the150m2
sounding (notshown),withtheboundary justeastof the site, s-2 thresholdgiven by Davies-Joneset al. [1990]. The
indicatedthat the atmosphere wasmoderatelyunstablewith positive shearwas9.3x 10-3 s-• [Davies,1989].TheJohns
lifted indicesaround-6 and a convectiveavailablepotential et al. [1990]relationship
of positiveshearand CAPE sug-
energy
(CAPE)near1500m2 s-2 (or 1500J kg-1).The gestedthatthe potentialexistedfor the development
of
CAPE, or positivebuoyantenergy,is a measureof the strongto violentmesocyclone-related
tornadoes.
positive
energy
locatedbetween thelevelof freeconvection
(LFC) andequilibrium
level(EL) onthe sounding[Rasmus- 3. THE SUPERCELLEVOLUTION
sen and Wilhelmson, 1983]. In contrast to the Jackson
sounding,
a significant
low-levelinversionwaspresentin The stormof interest,on March 12, 1986,developedin
southernLouisianaat Boothville(or -120 km southeastof south central Louisiana and moved northeastwardinto
New Orleans).The low-levelcapwas5.7øC,andwithoutany southwest Mississippi
shortlyafter1430.Twotornadoes (F2
otherlifting,a surfacetemperature
of 30øCwasneededto and F1 intensity)
were spawned with this storm between
breakthiscap;maximum temperatures only26øC 1500and 1700in southcentralMississippi,alongwith some
reached
priorto thetimeof convection. thatforced hail and wind damage.The stormwas examinedalmost
Thisindicated
suchasalongtheboundary, continuously
liftingvia strongconvergence, with a low-elevation
angle(0.5ø) usingthe
was necessaryto break the cap. WSR-57locatedat Jackson,Mississippi.
Occasionally,
 IMY AND PENCE 259

.1.80 ø was not systematically located as described by Lemon

[1980], since only one upper level tilt was used to examine
5 8
the midlevelsof the storm. We used the center point of the
midlevel echo as the approximate location of the top. In
Lemon' s Technique it is not the heightbut the location of the
top relative to the low-level reflectivity gradient that is
important. If the midlevel overhang extends more than 6 km
beyond the low-level reflectivity (as described by Lemon for
severe thunderstorms), then it would indicatethat the top is

225 / 23m/s
likely located on the updraft flank of the storm and not over
the storm's core.
The low elevation angle slice of the storm, prior to the

•FB
ø I
• I t'
20
I
4O
I I
60
ø time it became severe, is illustrated in Figure 3 at 1455. A
small appendage or pendant located on the storm's rear
flank, which we will see later, was not associated with the
updraft flank of the storm and therefore was not dynamically
Fig. 2. Jackson 1200 hodograph adjusted in the lower 2 km to
better reflectthe wind profile alongand east of the boundary.The significantaccording to Letnon [1980]. Note also the band of
helicityis 173m2 s2. precipitation to the southeast of this storm. The relationship
between this band of precipitation and the development of
the tornadic storm are unclear, but the band is notable in all
operator elevated the antenna to examine the upper portions of the low elevation slices used in this paper. The precipita-
of the storm. Since this was performed during a warning tion bands southeast and southwest of the echo resemble an
situation, time did not allow for complete tilts. With the "arch echo" found by Smith and McKee [1988].
higher elevation tilt, the VIP 3 (41--46 dBZ) core (bright The first elevated tilt of 8.3ø at 1455 is presented in Figure
white) was used as the midlevel echo since the storm was 4 (all upper level tilts bisected the storm at 10 km above
surroundedby light to moderate precipitation. The storm top ground level, assumingstandardatmosphericrefraction con-

1455

46 km

92 km

138 km

Fig. 3. Low level sliceof stormlocatedwestof arrow at

260 MISSISSIPPI SUPERCELL

1455

46 km

92 kmi

138 kin!

Fig. 4. Upper level slice of storm (8.3 ø) at 1455.

ditions). The midlevel core was almost vertical over the low issued for the entire precipitation echo, several counties
level core. Both VIP 5 and 6 (>50 dBZ) cores were evident would have been overwarned. Locating the most dangerous
even at this height. The VIP 3 midlevel echo (41-46 dBZ) portion of the storm is essential both for aiding spotters and
fitlied to display a significant echo overhang. The height of warning those communities most threatened [Mo!!er eta!.,
the VIP 5 (and VIP 6) above 10 km, however, would warrant 1990]. The combination of the low-level and high-level tilts
the issuance of a severe thunderstorm warning using Lem- indicated that the most dangerousportion of this storm was
on's criteria. on the east side of the storm near the low-level concavity.
In Figure 5, at 1519, the low-level echo configuration had At 1525 an F2 tornado developed on the east side of the
evolved into a kidney bean shape, which, as mentioned storm, just south of the concavity previously noted. The
earlier, may be suggestiveof an HP supercell. A concavity tornado was on the ground for approximately 15 min with a
was evident on the east side of the storm along with a tight 10-km path length. The largest hail with the storm, 2 cm in
reflectivity gradient, both suggestiveof the storm's updraft diameter, occurred in the VIP 5 core north of the concavity.
flank. The kidney bean shapepersistedfor only about 5 min. At 1542 the low-level echo in Figure 7 had become
An elevated cut was not available at 1519, but 2 min earlier, elongatedfrom north to south, but there were no hints of
the higher tilt revealed an extensive midlevel overhang rotation. A channel of weak echo [Przybylinski and DeCaire,
above the concavity on the east side of the storm (Figure 6). 1985]had developedon the back side of the storm, possibly
The development of the massive midlevel overhang since indicative of the rear flank downdraft (RFD) [Mo!ler eta!.,
1455 indicatedthat the updraft strengthhad increasedsignif- 1974]. The VIP 3 midlevel echo had a circular appearance at
icantly. 1538(Figure 8, closestavailableelevatedtilt to the low tilt).
The combination of the tight low-level reflectivity gradient No tornadoes ensued with the circular midlevel echo, but
in the concavity, the expanse of the midlevel overhang, the surface
windsgreaterthan50 knots(26 m s-1) occurred
at
mesoscale environment (storm moving along a preexisting the surface near this midlevel feature at 1545.
boundary), moderate CAPE, and favorable helicity sug- A small concavitywas noted on the northeastside of the
gestedthat this stormhad the potentialto becometornadic. storm on the low elevation slice at 1605 (not shown). The
Once it appearsthat a storm is (or is becoming)tornadic,it high level tilt at 1604, however, revealedthat the midlevel
is critical to pinpointthe tornadicregionof the storm. With overhangwaslocatedon the eastsideof the storm.Thusthis
a large storm, such as this one, if a tornado warning was concavityprobablywas not dynamicallysignificant,since
 0.5 ø 1519

46 km

92 km

,138 km

Fig. 5. Low-level slice of the storm at 1519. The midlevel overhang is depicted by the dotted line. The overhang
extends approximately 10 nm east and northeast of the low-level reflectivity gradient in the concavity. A tornado
initially touched down 6 min later. The arrow depicts the location of the tornado touchdown relative to the precipitation
echo at this time.

1517
5.0¸

46 km

92 km•

38 km

Fig. 6. Upper level sliceof storm(5.0ø) at

262 MISSISSIPPI SUPERCELL

6.5¸
.,
, ,
1542

46 km

) ' 92 km

1138 kin

Fig. 7. Low-lcvcl slice of slorm with a weak echo channel at the,rear of storm (southeastof arrow) at 1542.

wit.,;not c½•upledwith the .,,Iorm'stipdraft. The overhang was such as the hook or distinctive echoes for issuing severe
associated with an intense updraft and low-level circulation thunderstorm or tornado warnings. SupercelIs in Missis-
located on the extreme southern portion of the precipitation sippi, similar to the HP supercell, are characterized by
echo. The overhang extended about 8 km northeast of the continuouslychanginglow-level echo configuration.,,..
Broad
low-level reflectivity gradient. Also, the midlevel echo was hooks may be observed, but waiting for these features to
in the shapeof an "S," which Doswell et al. [1990] state can appearbeforeissuinga tornadowarninglikely will resultin
be an indication of circulation. small lead times or missed events.
Without the use of an elevated tilt and a knowledge of the The three-dimensional structural evolution of this super-
environmental wind profile, once again it would have been cell was similar to that described by Browning [1964] and
extremely difficult to determine the most menacingportion Lernon and Doswell [1979] for supercelIs in the central
of this storm. A weak tornado (FI) developed on the plains,or the classicsupercell.Althoughthe low-level echo
southernportion of the storm with a 2-km path length. The configurations doesnot resembleBrowning'smodelfor the
low-level echo associated with this second tornado never classicsupercell,this storm still displayeda vertical struc-
assumedthe kidney bean shapeillustratingthe variability of ture similar to the classicsupercell.With or without Doppler
low-level echo configurationsassociatedwith supercelIsin data, only through a tilt sequencecan most severe and
Mississippi.Also, while the first tornado developedon the tornadic thunderstorms be warned for effectively, regardless
east side of the precipitation echo, the second tornado of the low-level echo shape. Any tilt technique, however,
developedon the southernportion of the echo. may becomeineffectiveduringthe storm'scollapsein the
later stagesof tornadicdevelopment,but tornadowarnings
shouldbe consideredthroughoutthe life of a storm with a
4. CONCLUSIONS
past historyof tornadoeswhenDopplerinformationis not
Strongstormsin shearedenvironmentsmustbe analyzed available [lmy and Burgess, 1991].
frequentl.
y with at least one higherelevationangleto deter- However, radar operatorsmay feel that the Lemon Tech-
mine storm structural features suggestive of severe or tor- nique is too time-consumingfor real-time warning opera-
nadic storms. As illustrated in this case, forecasters should tions.We proposethe followingmethodology (as performed
not rely solely on characteristiclow-levelconfigurationsby theradaroperatorin thiscase)for analyzingstorms
 IMY AND PENCE 263

4-5ø 1538

46 km

92 km

138 km

Fig. 8. Upper level slice of storm (4.5ø)at 1538. Note the semicircle-shapedechojust south ofthe main midlevel echo.

conventional radar systems (with the antenna rotating con- the other screen, velocity products at the same elevation
tinuously through both steps). angles as the reflectivity slices also could be displayed to
1. On the lowest elevation (usually 0.5ø), contour the help validateor nullify the existenceof a mesocyclone.All of
VIP 2 (30-41 dBZ) echo on the PPI scope, or the VIP 3 this can be done easily and repeatedly and in a few seconds
(41-46 dBZ) echo if VIP 2 (30-41 dBZ) precipitation sur- through the use of a user function [Lemon et al., 1992]. The
roundsthe storm. Mark or highlightthe stronges t low-level radar does the tedious work of collecting and displaying the
reflectivity gradient if present. data at various elevation angles, while the operator inter-
2. Quickly elevate the antenna so the radar beam slices prets the three-dimensionalstructure of stormsto determine
through the midlevel portion of the storm (5-12 km) [Lemon, when and where warnings should be issued.
1980] using the VIP 3 (41-46 dBZ) contour.
After performing these steps, use Letooh's [1980] guide-
lines for determining if a warning is required and, if so, AcLnowledgments. A special thanks to Le,Lemon for his
whether it should be for a severe thunderstorm or a tornado.
exceptionally helpful comments that significantly improved the
This technique can be executed quickly (in less than a content in this manuscript. We thank Bob Ponds, retired Senior
Meteorological Technician, who examined this storm three-
minute) yet provide critical information necessaryto warn dimensionally.We thank Dan Purcell for preparingthe radar pic-
effectivelyfor most severestorms,includingthe wide range tures and RichardArmstrong for preparingone of the figures.Also,
of low-level echo configurationsassociatedwith supercelIs. we thank the OSF staff and the anonymousreviewersfor ofi•ring
The Lemon Technique will be much easierto perform on helpful comments that improved this manuscript.
the Weather Surveillance Radar-1988 Doppler (WSR-88D)
system.The WSR-88D automaticallymakesone low-level REFERENCES
slice (0.5ø) and 8 or 13 elevated slicesevery 5 or 6 min. Up
to four productscan be displayedon one screenat the Alberty, R. A., T. Crum, and F. Toepfer, The NEXRAD program:
principaluserprocessor(PUP) workstationat any one time Past, present, and future; a 1991 perspective, in Preprints, 25th
International Conference on Radar Meteorology, pp. 1-8, Amer-
[Albertyet al., 1991]. Dependingon the distanceof the storm ican Meteorological Society, Boston, Mass., 1991.
from the radar, the operatorcould selectthree reflectivity Bluestein, H. B., and C. R. Parks, A synoptic and photographic
slicesdepictingthe lower, middle,andupperportionsof the climatology of low-precipitation severe thunderstormsin the
stormand an echotopsproductin the fourthquadrant.On southernplains, Mon. Weather Rev., 111, 2034-2046,
264 MISSISSIPPI SUPERCELL

Browning, K. A., Airflow and structure of a tornadic storm, J. and warning criteria, NOAA Tech. Memo. NWS NSSFC-3, 60
Atmos. Sci., 20, 533-545, 1963. pp., Natl. Oceanic and Atmos. Admin., 1980. (Available as NTIS
Browning, K. A., Airflow and precipitation trajectories within PB-273049, Natl. Tech. Inf. Serv., Springfield, Va.)
severe local storms which travel to the right of the winds, J. Lemon, L. R., and C. A. Doswell I!I, Severe thunderstorm evolu-
Atmos. Sci., 21,634-639, 1964. tion and mesocyclone structure as related to tornadogenesis,
Browning, K. A., Some inferencesabout the updraft within a severe Mon. Weather Rev., 107, 1184-1197, 1979.
local storm, J. Atmos. Sci., 22,669-677, 1965. Lemon, L. R., E. M. Quoetone, and L. J. Ruthi, WSR-88D:
Chisholm, A. J., and J. H. Renick, The kinematics of multicell and Effective operationalapplicationsof a high data rate, in Preprints,
supercell Alberta hailstorms, Alberta Hail Studies, 1972, Hail Symposium on Weather Forecasting, pp. !73-180, American
Stud., Rep. 72-2,24-31, Res. Counc. of Alberta, Edmonton, 1972. Meteorological Society, Boston, Mass., 1992.
Davies, J. M., On the use of shear magnitudesand hodographsin Moller, A. R., C. A. Doswell III, J. McGinley, S. Tegtmeier, and R.
tornado forecasting, in Preprints, 12th Conferenceon Weather Zipser, Field observationsof the Union City tornado in Okla-
Analysis and Forecasting, pp. 219-224, American Meteorological homa, Weatherwise, 27, 68-77, 1974.
Society, Boston, Mass., 1989. Moller, A. R., C. A. Doswell III, and R. W. Przybylinski, High-
Davies-Jones, R. P., D. W. Burgess, and M.P. Foster, Test of precipitationsupercelIs:A conceptualmodel and documentation,
hellcity as a tornado forecast parameter, in Preprints, 16th Con- in Preprints, 16th Conferenceon Severe Local Storms, pp. 52-57,
Jkrenceon SevereLocal Storms, pp. 588-592, American Meteo- American Meteorological Society, Boston, Mass., 1990.
rological Society, Boston, Mass., 1990. Przybylinski,R. W., and D. M. DeCaire, Radar signaturesassoci-
Doswell, C. A., IIi, A. R. Moller, and R. W. Przybylinski,A unified ated with the derecho,a type of mesoscaleconvectivesystem,in
set of conceptualmodels for variations on the supercel!theme, in Preprints,14th Conferenceon SevereLocal Storms,pp. 228-231,
Preprints, 16th Conferenceon Severe Local Storms, pp. 40-45, American Meteorological Society, Boston, Mass., 1985.
American Meteorological Society, Boston, Mass., 1990. Przybylinski, R. W., S. Runnels, P. Soden, and S. Summy, The
Forbes, G. S., Relationshipbetweentornadoesand hook echoeson Allendale, Illinois tornado--January 7, 1989: One type of an HP
April 3, !974, Mon. Weather Rev., 109, 1457-1466, 1981. supercell,in Preprints,16thConferenceon SevereLocal Storms,
Imy, D. A., A modifiedsupercellof November 31st-December1st, pp. 516-521, AmericanMeteorologicalSociety, Boston, Mass.,
S. Reg. Headquarters Tech. Attachment, Mar. 25, 1986, 6 pp., 1990.
Natl. Weather Serv. S. Reg. Headquarters, Fort Worth, Tex., Przybylinski,R. W., J. T. Snow,E. M. Agee,andJ. T. Curran,The
1986. useof volumetricradar datato identify supercelIs:A casestudyof
Imy, D. A., and D. W. Burgess,The structural evolution of a June 2, 1990, this volume.
tornadic supercel!with a persistentmesocyc!one,in Preprints, Rasmussen,E. N., and R. B. Wilhelmson, Relationshipsbetween
25th International Conferenceon Radar Meteorology, pp. 408- stormcharacteristicsand 1200GMT hodographs,low-level shear,
411, American MeteorologicalSociety, Boston, Mass., 1991. and instability,in Preprints, 13th Conferenceon SevereLocal
Johns,R. H., J. M. Davies, and P. W. Lef•wich, An examinationof Storms, pp. J5-J8, American MeteorologicalSociety, Boston,
the relationshipof 0-2 km AGL "positive" wind shearto poten- Mass., 1983.
tial buoyantenergy in strongand violent tornado situations,in Smith, C., and R. McKee, Arch echoes and maxi tornadoes,
Preprints,16thConfi,renceon SevereLocal Storms,pp. 593-598, S. Reg. HeadquartersTech. Attachment,Apr. 12, 1988, 5 pp.,
American MeteorologicalSociety, Boston, Mass., 1990. Natl. Weather Serv. S. Reg. Headquarters, Fort Worth, Tex.,
Lemon, L. R., Severethunderstormradar identificationtechniques
 Satellite Observations of Tornadic Thunderstorms

JAMES F. W. PURDOM

Regional and Mesoscale MeteorologyBranch, NOAA/NESDIS, CooperativeInstitutefor Research in the Atmosphere

Colorado State University, Fort Collins, 80523

1. INTRODUCTION Lemon, 1979]. Satellite data, particularly from geostationary

satellites, can play an important role in monitoring the
Satellite data provide valuable information concerning horizontalvariability of that constituent.That capabilitywas
tornadic storms and the mesoscale environment in which
first demonstratedby Hillger and VonderHaar [1981]using
they form. Below, selectedaspectsof the tornadic storm's polar orbiting satellite soundingdata. They were able to
mesoscale environment as observed using satellite data are extract moisture information at a much finer horizontal
discussed. That is followed by a discussion of features resolution than was possible using conventional data
evident in satellite imagery that accompany tornadic storm sources. Similar results have been demonstrated from geo-
development.Thoughtsconcerningthe dynamicand ther- stationaryaltitudesusing data from GOES-VAS [Smith et
modynamicimportanceof some of those featuresare dis- al., 1984; Hillger and Purdom, 1990]. (GOES-VAS is a
cussed.In a few examples, other data are presentedalong doubleacronymfor GeostationaryOperationalEnvironmen-
with satellite data. However, while the author realizes the
tal Satellite-VISSR AtmosphericSounder,with VISSR be-
importanceof other data sets in tornadicstorm investiga- ing visible and infrared spin scanradiometer.)
tions, this paperfocuseson usesof satellitedata. The major The characteristics of GOES-VAS sounding channels are
emphasis is howto usesatellitedatato helpthereaderinterpret shown in Table 1. For VAS the energy received in each
why stormsform, evolve, and take on characteristic appear- channel originatesin the region labeled "representative
ances. A satellite image or a series of images represents thickness." The amount of energy emitted in that regionis a
on-goingdynamicandthermodynamic processesin the atmo- functionof the concentrationandtemperatureof the absorb-
sphere;what is importantis recognizing thoseprocesses. ing/emittingconstituent,with cloudsactingas a contami-
nant. For the window channelsmost of the energyoriginates
2. MESOSCALE ENVIRONMENT INFORMATION at the Earth' s surface,with a small amountof absorptionand
FROM SATELLITE DATA reemission by water vapor.
While the studiesof Hillger and Smithcitedaboveusedall
Satellitedata provide both quantitativeand qualitative availabledata,simplerapproaches usinga limitednumberof
informationconcerning the mesoscale environment in which VAS channels have been demonstrated [Chesters et al.,
tornadic storms form. In the subsectionsbelow, both quan-
1982;Petersenand Mostek, 1982]. For example,a product
titativeandqualitative(imageinterpretation) usesof satellite knownas "split window" may be derivedusingthe differ-
data are discussed.
encein energybetweenthe 12.7-tzmand 11.2-•minfrared
(IR) channels.Accordingto Chesterset al. [1983p. 742]:
2.1. Moisture and Water Vapor Data
"The VAS splitwindowclearlydifferentiates thoseareasin
The horizontal and vertical structure of moisture in the whichthewatervaporextendsovera deeplayerandis more
atmosphere,an importantfactor in stormdevelopment, ableto supportconvectivecellsfromthoseareasin which
exhibitssignificantmesoscalevariability[Doswelland the water vapor is confinedto a shallowlayer and is
thereforeless able to supportconvection."Split window
The Tornado:Its Structure,Dynamics,Prediction,and Hazards. products are usefulover land duringthe afternoon when
GeophysicalMonograph79
Thispaperisnotsubject toU.S. copyright.
Publishedin 1993bythe there are large differences
in the signalbetween the two
American GeophysicalUnion. channels. This occurs with a high surface temperature, a
266
SATELLITEOBSERVATIONS
OF TORNADIC
THUNDERSTORMS

TABLE 1. GOES-VASChannelCharacteristics 2.3. ImageInterpretation:

Daytime
Channel
Squall Line Development
Central Peak Representative
Wavelength, Absorbing Level, Thickness, Generally, organized convergence zonesthattrigger
strong
microns Constituent mbar mbar convection aredetectablein satellite
imagerypriortothunder-
stomformation. Thoseconvergence zonesmaybeassociated
14.7 CO2 40 150-10
14.5 CO2 70 200-30 with dry lines,frontalzones,or areasof prefrontal conver-
14.3 CO2 300 500-10 gence.The abilityto locateareasof incipientsquallline
14.0 CO2 450 800-300 development wasoneof theearliest usesof geostationary
13.3 CO2 950 sfc-500 satellite
imagery
in stormforecasting
[Purdom, 1971].
4.5 CO2 850 sfc-500
12.7 H20 sfc sfc-700 Figure1 presents a typicalexampleof how squallline
11.2 window sfc development appears in geostationary
satellite
imagery.
In
7.3 H2O 600 800-400 Figure1 a well-defined convective
lineis locatedalonga
6.7 H 2O 450 700-250
4.4 CO2 500 800-100 surfaceconvergence zone.Theconvergence zone,analyzed
3.9 window sfc in Figure 2, is movinginto warm and moist surfaceair that
extendsfrom easternKansasinto portionsof Missouriand
The abbreviation s'fcindicates surface.
Iowa. Later,a squalllinehasdeveloped
alongthe northern
portion of the convergencezone.
For the caseshownin Figure 1, upperlevel dynamic
steep lapse rate, and significantlow-level moisture:ideal support was strongestacross the northern portion of the
low-levelthermodynamicconditionsfor intenseconvection. region.When 15-minintervalsatelliteimagerywasanimated
In the eveningwhenthe landcools(or overtheocean)the relativeto the developingconvectiveline, warm sectorcumu-
signaldifferential
is small,and a meaningful
split-windowlus mergedwith the line to the northbut not with the line to the
product is difficult to derive. south.Thisimpliesthat strongerlow-levelconvergence to the
The 6.7-tzmchannel'sdatadepictregionsof middlelevel northaidedearlierthunderstorm formationin thatregion.
moistureand clouds.Distinct patternsof cool moist areas Anotherexampleof squallline developmentis shownin
and warm dry areasare readily detected. Suchfeaturesare Figure3. In Figure3a a brightconvectiveline,locatedalong
related to areas of horizontal advection and vertical motion a surface convergence zone, extends from north central
at both synoptic scales and mesoscales. When viewed in Texasinto southernOklahqma.Smokefromrangefiresin
timelapse,theyexhibitexcellentspatialandtemporalcon- west Texas extends into the rear of the convective line. To
tinuity. Strongbaroclinicregionssuchas jet streamsand the eastof the line skiesare clear,becomingcloudyagainin
vorticitymaximacanoftenbeidentifiedin cloudfreeregions the vicinity of Tyler and Longview, Texas. Most of the
by the sharpmoisturegradientdetectedin the 6.7-tzmimage cloudiness
aroundTyler is cumuliformwith a distinctrope
[Weldon and Holmes, 1991]. appearance. Such cloudinessis indicative of cumulus devel-
oping under a strong low-level inversion. The mechanism
2.2. Wind Field lnj•brmation responsiblefor the capping inversion may be a "lid" of
Tornadic storm outbreaksare characterizedby rapidly midlevelhotdry air that hasbeenadvectedacrossthe region
evolving storms embeddedin an environmentof strong [Carlson et ai., 1983]. Figure 3b, taken a few hours after
vertical wind shear. Because of this, during outbreaksof Figure 3a, shows a squall line acrossTexas and Oklahoma.
severeweather the GOES satelliteis placed in a rapid scan About 100 miles(--•161km) to the east of that squallline the
imagingmode(RISOP) to supportoperationsat the National air is very stable.Abundantrope-appearingcumulusand the
Severe Storm Forecast Center. During RISOP, 5-min inter- Longview, Texas, sounding(Figure 4) confirmthat stability.
val imagery is scheduledfor a portion of selectedhours. Is the Longview sounding representative of low levels
Althoughnot doneoperationally,with RISOP imageryit is immediatelyahead of the squall line? If so, a very strong
possible to compute high-resolution cloud motion vectors for capping inversion must be removed for thunderstorms to
both cumulus and cirrus [Lubich and Putdom, 1992]. When form. Such stability may be overcome by (1) broad-scale
compared with winds from ground-based profilers, such vertical motion, (2) precipitation into the stable layer de-
cloud motion vectors were found to match within the noise stroying the inversion through evaporative cooling, (3) dif-
level of the two systems[Dills and Smith, 1992]. Recently, ferential advection at various levels, or (4) as described by
Putdom and Weaver [ 1992] performed a detailed analysisof Shapiro et al. [1985], stronglocalized vertical forcing along
cloud motions for the April 26, 1991, Kansas/Oklahoma the surfaceconvergencezone. Becauseof the proximity of
tornado outbreak. They found that (1) cirrus winds exhibited the stable air to the squall line and the lack of midlevel
mesoscale detail as a jet streak moved into the outbreak cloudsover the region, the first two mechanismsmost likely
area, (2) cirrus to cumulus level shear was greatest in the did not play a significantrole. Most likely, a combination of
region of the tornadic storms, and (3) individual supercelIs the latter two mechanismscame into play in the develop-
blocked cirrus level flow. ment of this squall
 PURDOM 267

$D

Fig. 1. Montageof GOES-East 1-kinresolutionvisibleimagesfrom left to right at 1300, 1400, 1600,and 1700CST for
June 14, 1976. Arrows point to organizedcloudinessalong the surface convergenceline as well as the fully developed
squall line.

2.4. Image Interpretation: Nighttime lated with severe weather [Fujita, 1978: McCann, 1981;
Squall Line Development Heymsfield and Blackruer, 1988].
The air mass over northwest Texas and New Mexico is
Squall lines may form at nighttime over the Great Plains of
dry. In Figure 5a, relatively cold surface temperatures, with
the United States. In many cases that development occurs
with the northward advection of low-level moist air ahead of
subtle variations due to terrain, are apparent. From the Rio
Grande river into central Texas, temperatures appear much
an approaching synoptic scale system. Becauseof a small
amount of absorption and reemissionby water vapor in the
VAS 11.2-tzm "window," such developmentmay be moni-
tored using that infrared channel's imagery. The following
generalizationsmay be made concerninginterpretationof
11.2-tzmdata: (1) for a moist low-level environmentduring
the daytime, after the land surface has heated, temperatures
will appear cooler than the actual surfacetemperaturedue to
someabsorptionand emissionby the overlyingcoolermoist
air, (2) at nighttime, low-level moisture within a nocturnal
inversion will causetemperaturesto appearwarmer than the
colder underlyingsurface, (3) when the atmosphereis dry,
day or night, 11.2-tzmtemperatureswill be closeto the true
surfacetemperature.In addition,the surfacewill heat(cool)
more rapidly during the day (night) in dry versusmoist
environments.
On the eveningof May 13, 1985, strongnighttimethun-
derstormactivitydevelopedin southwestTexas.That activ-
ity was triggeredas a vorticity center and surfacefront
moved out of New Mexico into a region of low-levelmoist
air. In Figure5a, thunderstorm
activityextendsfromsouth-
ern Kansas into central Texas. The black area at the south-
west end of the anvil in north Texas is associated with the
colder overshootingtops of a severe thunderstorm.That Fig. 2. Selected surface observationsat noon CST on June 14,
overshootingtop regionappearsat the vertexof two cold 1976, over the region shown in Figure 1. Analyzed are pressure
plumes.That signature,an "enhancedV," hasbeencorre- (solid lines) and dew point temperature (dashed
268 SATELLITE OBSERVATIONS OF FORNADIC THUNDERSTORMS

TEXAS

Fig. 3b. GOES-East l-km resolutionvisible image at 1731CST on

!:ig. 3a. G()ES-East l-km resolution visible image at 1431 CST on
March I I, 1988. Arrows point to the squall line in eastern Oklahoma
March II, 1988. Arrows point to initial cumulus formation in
and Texas. Blowing smoke is still evident to the rear of the squall
()klahoma itlld north Texas where a squall line is beginningto form.
line, as is rope cumulus at Longview (G).
N{•Ic the blowing smoke at A and the "rope" cumulus around Tyler
('1')and l•ongview (G). b. GOES-East I-kin resolution visible image
at 1731 CST on March 1!, 1988. Arrows point to the squall line in
eastern ()klahoma :•nd Texas. Illowing smoke is still •zvident to the lines that develop under strong synoptic forcing where winds
rear of the squall line, as is rope cumulus at l•ongview (G). veer and increase to strong westerly with height. Another
important type of highly organized mesoscale convective
system, the mesoscaleconvective complex (MCC), forms with
warmer (darker) than they do to the north and west. That relatively weak synoptic scale forcing [Maddox, 1980]. MCCs
dark region is a reflection of low-level moisture and perhaps appear to be a convectively driven weather system. With
low-level cloudiness at the height of the low-level inversion.
l)uring the evening this low-level warm and moist air ad-
vccted north into the Midland, Texas, area. Another rela-
tively warm region is apparent between the storms in north
central Texas and the cooler ground in northwest Texas. The
appearance of that region is the result ofboth warmer ground
and low-level air that has been moistened by the earlier
thunderstorm activity. The westward extent of that more
moist air is determined by local terrain: the Palo Duro
Canyon and region to the east is around 1000 feet (300 m)
lower than the higher ground to the west (the Cap Rock
escarpment).
Over the next several hours the low-level moisture's
northward advance into the Midland area could be followed
using animated GOES infrared imagery. In Figure 5b, upper
level cloudiness associated with the vorticity center is evi-
dent in New Mexico. The crescent-shaped white area in
southwest Texas locates the surface front. Figure 5c shows
the squall line that formed as the front moved into the warm
and moist environment to its east. Special ground-based
radiometer data confirmed the rapid increase in moisture at
Midland [Parsons et al., 1986] during the time the dark
region shown in Figure 5a moved acrossthat area.

2.5. Image Interpretation: Mesoscale -30 -20 -10 0 10 20 30

Convective Complexes
Fig. 4. Soundingfor Longview, Texas, that would have been
In the previouscases,anvilcirruswas seenblowingoff to released near the time of Figure 3b. Note the strong capping
the east ahead of the squall line. This is typical for squall inversion near 800
 PURDOM 269

TEXAS

P
I,

Fig. 5a. EnhancedGOES infrared image taken at 2144 CST on Fig. 5c. As in Figure 5a, 6 hours later. Arrows point to a well-
May 13, 1985. Dark to white represents warm to cold (-32øC); developed squall line.
temperaturesthen becomecolder in 6øCincrements.Arrows point
to the low-level moist region referred to in the text. M, P, and C
locate Midland, the Palo Duro Canyon, and Cap Rock escarpment,
respectively. b. Enhanced imagery as in Figure 5a, 4 hours later. weather system occurs most often over the central United
Arrows point to the surfacefront. c. Enhanced imageryas in Figure States during the late evening and nighttime hours With one
5a, 6 hours later. Arrows point to a well-developedsquall line. type convective system that is similar to the MCC, the
Derecho [Johns and Hirt. 1983], strong surface winds and
gust front tornadoes are the primary type of severe activity.
mature MCCs, similar to tropical convection,there is often a MCCs have been observed to interact with and modify the
reversalof vertical wind shearin the upper troposphere(rela- larger scale environment in which they are embedded. By
tive to the moving MCC). The resulting appearance in a influencing the larger-scale environment they affect down-
satellite image is active convection along the leading edge of stream weather long after their demise.
the MCC with a large stratiform anvil area trailing behind.
This important spring and summertime convective
3. STORM SCALE INFORMATION DETECTABLE
IN SATELLITE DATA

In the previous section, aspects of the tornadic storm's

mesoscale environment observed using satellite data were
discussed. In the following section, features evident in
satellite imagery that accompany individual tornadic storm
development are discussed. At the end of the section,
thoughts concerning those features are presented.

3.1. Isolation of Preferred Areas for Tornadic Storm

Development: Interaction With Boundaries
Under proper dynamic forcing, tornadoes are spawned
when squall line thunderstorms interact with an organized
synoptic scale boundary, such as a warm front [Millet',
1972]. Another favored region for tornado activity exists
where a strong thunderstorm and a low-level boundary due
to previous convective activity interact [Purdom, 1976,
1990].Usually, both synoptic and stormscaleboundariesare
detectablein satellite imagery. Such detection is especially
aided by analysisof animated RISOP imagery [Purdotnand
Weaver, 1992]. When surface data are combined with satel-
Fig. 5b. As in Figure5a, 4 hourslater.Arrowspointto the surface lite imagery, the ability to isolate boundaries is further
front. enhanced [Purdom,
270 SATELLITE OBSERVATIONS OF TORNADIC THUNDERSTORMS

ILLINOIS

Fig. 7. GOES-East l-km resolution visible image at 1645 CST on

June 13, 1976. Arrows at A point to the "well-defined" arc cloud
line south of Chicago. The thunderstorm at B is tornadic. Another
set of arrows points to arc cloud lines in western Indiana; they are
a product of the storm to their east.
Fig. 6. GOES-East 1-km resolution visible image at 1513CST on
May 2, 1979. Wave clouds and low-level stratus predominate in the
stable air mass (B), with cumulus filling the potentially unstable air
(A). Arrows point to the convergence line referred to in the text.
storm and an arc cloud line interact locates a favored region
for tornado activity [Purdom, 1976, 1990]. On June 16, 1976,
As mentioned above, prior convective activity may pro- the stage for tornado activity around Chicago was set after
duce a low-level boundary that will focus tornado activity. thunderstorm activity moved through that area leaving be-
Most often, that boundary appears as an arc cloud line hind a well-defined arc cloud line (Figure 7). Tornado
[Purdorn, 1976]; however, on occasion the boundary is activity was triggered where that arc cloud line merged with
located by a transition from low-level wave to cumulus the thunderstorm in northeast Illinois. The satellite image in
cloudiness [Beckman, 1982]. In both cases, the boundary is Figure 7 was taken while tornado activity was in progress
located at the transition between warm sector air and bound- near Chicago [Fujita and Hjelmfelt, 1977].
ary layer air that has been cooled and stabilized by the
evaporation of min. The reason for the difference in cloud
appearance (wave/cumulus versus arc) has not been ex-
plained; however, the most likely cause is the difference in
thermodynamic structure of the boundary layer air into
which the rain fell and evaporated. Arc cloud lines prevail
during afternoon hours, while wave/cumulus transition re-
gions are mainly the result of nighttime and early morning
thunderstorm activity.
On May 2, 1979, early morning thunderstorms helped
stabilize the boundary layer in north central Oklahoma but
not in southwest Oklahoma. By afternoon (Figure 6), wave
clouds and low-level stratus predominate in the stable air
mass, with cumulus filling the potentially unstable air in
southwest Oklahoma. The intense thunderstorm activity in
western Oklahoma developed where an organized conver-
gence line merged with the boundary. Those thunderstorms
produced tornadoes at Orienta and Lahoma, Oklahoma;
downburst activity was also reported. Radar reflectivity and
Doppler velocity images of the tornadic storms showed well
defined hook echoes associated with strong cyclonic wind Fig. 8. GOES-West l-km resolution visible image at 1816 CST on
shears [Purdorn et al., 1982]. April 10, 1979. The view is into the back side of the Wichita Falls,
Arc cloud lines and their convective scale interactions are
Texas, tornadic storm. White wavy arrows point to the low-level arc
cloud line, while the bold white arrow points to the location of the
a natural part of the thunderstorm development and evolu- tower associated with the Wichita Falls tornado. The black wavy
tion process [Purdorn, 1986]. Where an intense thunder- arrow points to the overshooting top above the tornadic
 PURDOM 2'71

ho surface
wind
north
of arc cloud line

stable cold air edge

left behind storm

• • • a.tte
cloud

cold storm f•ow w•h •a•v•y

i•e•e low lev• conv•gence so••y • •
lo• lev• flow

Fig. 9. Schematicof severethunderstorminteractingwith an outflow boundaryfrom another storm, resultingin

tornado development.

3.2. Satellite Observationsof SupercelIs dome of cold air, subsiding in time, leads to clear skies
behindthe storm. New convective towers grow alongthe arc
Isolated supercelis,apparently unassociatedwith a bound-
ary produced by other convective activity, may produce cloud line. They are especially intense where they join the
strong tornadoes. The Wichita Falls, Texas, tornadic storm right rear flank of the storm: at that location the supercell's
provides a good illustrative example of how a supercell updraft is regenerated. At that location in Figure 8, convec-
tive towers extend from near the surface into the overshoot-
thunderstormappearsin satellite imagery (Figure 8). On that
day, anvil cirrus expanded rapidly eastward. Typical for ing top region at the rear of the anvil. The Wichita Fails
such cases, masking by anvil cirrus and satellite viewing tornado is on the ground beneath the eastern most of those
perspective combine to make it difficult to detect low-level towers. Evidence of rain-cooled air left behind by the
storm features with GOES-East imagery. (GOES-East loca- supercelIshas been noted with other tornadic thunderstorms
tion is nominally over 75øWat the equator.)However, as can [Purdom, 1990; Purdom and Weaver, 1992].
be seen in Figure 8, low-level features are easily detected In Figure 8 the towers at the rear of the storm extend
with GOES-West imagery because of that satel!ite's im- upward into a well-defined region of overshooting tops.
proved viewingperspective.(GOES-Westlocationis nomi- Using special 3-min-interval GOES-East imagery, Anderson
nally over 135øW at the equator.) [1982] found the overshooting tops associated with this
The following conceptualmodelis given to help interpret storm were in cyclonic rotation. The National Severe Storm
Figure 8. Initial supercell development occurs in a well- Laboratory's Doppler radar detected a well-defined cyclonic
mixed boundary layer. As the storm matures,a cold air mass circulationwith the Wichita Falls storm even thoughit was
is generated beneath its echo due to evaporation. As the nearly 200 km from the radar site: this indicates the storm
storm evolves and moves away from its origin it leaves was in rotation to great depths. This author found evidence
behind that stable dome of cold air: a new local boundary of cyclonicrotation at cloud top when viewing stormrelative
layer. In Figure 8 the thin curved arc cloud line behindthe RISOP satelliteimagery for the Plainfield, Illinois, tornadoes
supercellenclosesthe new low-level boundarylayer. That of August 28,
272 SATELLITE OBSERVATIONS OF TORNADIC THUNDERSTORMS

byStorm

!palraft old storm motion

Mid-level
flow
//
n•o storm motion

flank downdraft
surface to near •oud base
region/covered
by
cold dome left
by low level inflow
r•a• flank
low leu• downdraft
flow
Strongly growing cumulustowe•s further b•ock rear flow

Fig. 10. Schematicillustrationof (a) boundarylayer modificationdue to colddomeleft behindby thunderstormand

(b) supercellenteringits tornadic phase. Both modifiedand unmodifiedsoundingsare similarabove cold dome in
middle and upper troposphere.

3.3. A Conceptual Model of Tornadic Storms On the basis of these observations the following concep-
Severe thunderstorms interaction with a boundary: In tual modelis presentedfor tornadic storm developmentafter
case studies of tornadic storm development along larger a severe thunderstorminteracts with a boundary (Figure 9).
mesoscale boundaries, Maddox et al. [1979] found that Along the boundary, strong vertical wind shear (3v/3z) is
mesoscale convergence and cyclonic vorticity tended to localized; this is due to the evolving solenoidal circulation
maximizealongthe boundary. Using aircraft measurements, and interaction of low-level southerlyflow with the bound-
Sinclair and Putdom [1983] found low-level convergence, ary. It is well known that severethunderstormshave intense
vertical wind shear, and an evolving solenoida! circulation updraftsand downdrafts;acrossa very shortdistancestrong
concentratedalong arc cloud lines. convergenceleads to intensevertical motion (a
 PURDOM 273

•x). Thus, when such a storm moves from the west into a 4. CONCLUSIONS
preexistinglow-level boundary, vorticity associatedwith the In the previous sections it was shown how satellite data
boundaryis tilted in a cyclonicsenseinto the storm'supdraft. may be used to help diagnose tornadic storms and their
Further, where the severe storm's downdraft interacts with the mesoscale environments. The data can be used to monitor
boundary, cyclonicboundarylayer flow experiencesintense the spatial variability of clouds and moisturefrom synoptic
convergence.Thusboth convergenceon existingvorticityand to thunderstorm scales; geostationary satellites add a vital
tilting of vorticity from one fluid planeto anotherresultsin a temporal capability.
rapid increasein verticalvorticity and tornadogenesis. In discussingthe mesoscaleenvironment in which tor-
Is there a common link between the mechanismcausing nadic storms form, both quantitative and qualitative appli-
supercelIsto produce tornadoes and those describedimme- cations of satellite data were addressed. However, a number
diately above? Before answeringthat question, considerthe of issues remain (not solely satellite analysis issues). For
development of the midlevel rear inflow jet, a critical com- example, how does instability combine with the vertical
ponent in supercell tornadic storm evolution. In their con- forcingwithin a convergencezone to producethunderstorm
ceptual model of a supercell, Lemon and Doswell [1979] activity? With satellite imagery, cloud developmentalong a
showupper level blockingby a supercellas necessaryfor the convergencezone may be monitored and cumulus flow
developmentof the midlevel rear inflowjet. As pointedout relative to it may be measured. However, we do not know
previously, such blocking has been observed using RISOP the convergence zone's width or depth (clear air returns
imagery. Furthermore, in Figure 10a the new boundary from NEXRAD may allow analysis on a very restricted
layer immediately to the rear of the storm, being shallow, scale). That information is basic for performing the simplest
of calculations for vertical motion at the scale on which
cool, and stable, inhibits mixing between air near the surface
with higher-momentumair aloft. Thus any forces acting to triggeringoften occurs.Supposewe had the aboveinforma-
accelerate the air above the cold dome may do so without tion. What of the stability that must be overcome for storms
surface frictional and mixing effects. In laboratory experi- to form? It is impracticalto routinely releaserawinsondesat
ments, Simpsonand Britter [ 1980]found that when a density frequent intervals and mesoscale spacings. Frequently,
available information from satellite soundingsmay be able to
current moves into a fluid of lesser density the less dense
fluid's flow accelerates and increasesin velocity as it moves
provide suchinformation. Even with suchir•ormation, how
do we use it?
acrossthe top of the density current. Upper level blocking
Two typesof tornadicstormswere discussed:(1) intense
coupled with the characteristicsof the cool and stable
thunderstorms that become tornadic after interacting with a
low-level air to the rear of the storm should aid the devel-
low-level boundary due to another source and (2) supercelIs.
opment of a rear, midlevel inflow jet. In the former case, it was shown how satelliteimagerycould
As shown in Figure 10b, the mature supercellcontinues
be used to help isolate the low-level boundary. For super-
generatinga cold air massbeneathits echo,a forwardflank cells, satellite viewing perspective was discussedwith
downdraft. Along the boundaryof the forward flank down- GOES-West often being preferred. A conceptualmodelwas
draft, the thermodynamicand dynamic characteristicsof an providedto help interpret supercellappearancein satellite
arc cloud line exist: only amplified by proximity to the imagery. That model explored the role of the modified
supercell.(For informationconcerningarc cloudline char- boundarylayer left in the wake of the storm. The conceptual
acteristicsin relationshipto the arc cloud line's source,see model needs to be verified.
Purdom [1986].)The supercell,rotatingcyclonicallyowing A common link was proposed for thunderstormsthat
to updraft and environmentinteraction [Rotunno and become tornadic after interaction with a boundary due to
Klemp, 1985],beginsmovinganomalously to theright.From another source and tornadic supercells. In both instances,
early on, the domeof cold air left behindby the stormhas storm/boundaryinteractionwas proposedto be the key to
beensinkingand spreadingout. Now, however,insteadof tornadogenesis.In the case of the supercell,the boundary
beingleft behindby the supercell,it and the forwardflank was due to its forward flank downdraft. For both cases,
downdraftmerge.At the mergerlocation,very stronglow- convergenceon preexistingvorticity and tilting of vorticity
level convergencedevelopsand intenseconvectivetowers from one fluid plane to anotherwas proposedto resultin a
grow. With the developmentof these new towers, the rapid increasein vertical vorticity and tornadogenesis.Is
midlevel rear inflow jet is further blocked. This blocking this common link correct? If so, how do we characterize
leadsto strongevaporation whichaidsin thedevelopment of boundaries and then use that information?
an intenseoutflowalongthe back side of the flankingline.
(Accordingto Moller et al. [1974], groundinterceptteams Acknowledgments.Portions of this work were supportedby
have observedcloud decks "apparently evaporated"and NOAA grant NA85RAH05045. My thanksto the reviewersfor
clearingof midlevelcloudinessat the rear of the flanking pointingout areasfor clarification.
line.)That rearflankdowndraftsurgesoutward,intersecting
REFERENCES
the forward flank downdraft, where the same mechanisms
discussedfor a severestorm's interactionwith a preexisting Anderson, C. E., Dramatic developmentof thunderstormcircula-
boundarylead to tornadogenesis. tion associated with the Wichita Falls tornado as revealed
274 SATELLITE OBSERVATIONS OF TORNADIC THUNDERSTORMS

satellite imagery, in Proceedings of the 12th Conference on storm signature,NOAA Tech. Memo. NWS NSSFC-4, 31 pp.,
Severe Local Storms, pp. 493-498, American Meteorological Natl. Severe Storm Forecast Center, Kansas City, Mo., 1981.
Society, Boston, Mass., 1982. Miller, R. C., Notes on analysis and severe-storm forecasting
Beckman, S. K., Relationship between cloud bands in satellite procedures of the Air ForceGlobalWeatherCentral,A WS Tech.
imagery and severe weather, in Proceedings of the I2th Confer- Rep. 200 (rev), 190 pp., Air WeatherServ., Offutt, Air Force
ence on Severe Local Storms, pp. 483-486, American Meteoro- Base, Nebr., 1972.
logical Society, Boston, Mass., 1982. Moller, A., C. Doswell III, J. McGinley, S. Tegtmeier, and
Carlson, T. N., S. G. Benjamin, G. S. Forbes, and Y. F. Li, R. Zipset, Field observationsof the Union City tornado in
Elevated mixed layers in the regional severe storm environment: Oklahoma, Weatherwise, 27, 68-77, 1974.
Conceptual model and case studies, Mon. Weather Rev., 111, Parsons,D. B., R. A. Hardesty, and M. A. Shapiro, The mesoscale
1453-1473, 1983. structureof the dryline in the formation of deep convection,in
Chesters, D., L. W. Uccellini, and A. Mostek, VISSR Atmospheric Proceedingsof the International Conferenceon Monsoon and
MesoscaleMeteorology, pp. 117-122, American Meteorological
Sounder (VAS) simulation experiment for a severe storm envi-
ronment, Mon. Weather Rev., 110, 198-216, 1982.
Society, Boston, Mass., 1986.
Petersen,R. A., and A. Mostek, The use of VAS moisturechannels
Chesters, D., L. W. Uccellini, and W. D. Robinson, Low-level
in delineatingregionsof potentialconvectiveinstability,in Pro-
water vapor fieldsfrom the VISSR atmosphericsounder(VAS) ceedingsof the 12th Conferenceon Severe Local Stortns, pp.
"split window" channels,J. Clin. Appl. Meteorol., 22,725-743, 168-171, American MeteorologicalSociety, Boston, Mass., 1982.
1983.
Putdom,J. F. W., Satellite imageryand severeweatherwarnings,in
Dills, P. N., and S. B. Smith, Comparisonof profiler and satellite Proceedingsof the 7th Conferenceon SevereLocal Storms, pp.
cloud-tracked winds, in Proceedings of the 6th Conference on 120-127, American MeteorologicalSociety, Boston, Mass., 1971.
Satellite Meteorologyand Oceanography,pp. 155-158,American Purdom,J. F. W., Some usesof high resolutionGOES imagery in
MeteorologicalSociety, Boston, Mass., 1992. the mesoscaleforecastingof convection and its behavior, Mon.
Doswell, C. A., III, and L. R. Lemon, An operationalevaluationof Weather Rev., 104, 1474-1483, 1976.
certain kinematicand thermodynamicparametersassociatedwith Purdom,J. F. W., Convectivescaleinteraction:Arc cloudlinesand
severe thunderstorm environments, in Proceedings of the 1Ith the developmentand evolutionof deep convection,Atmos. Sci.
Conferenceon Severe Local Storms, pp. 397-402, American Pap. 408, 197pp., Dep. of Atmos.Sci., Colo. StateUniv., Fort
MeteorologicalSociety, Boston, Mass., 1979. Collins, 1986.
Fujita, T. T., Manual of downburstidentificationfor ProjectNIM- Purdom,J. F. W., Convectivescaleweather analysisand forecast-
ROD, SMRP Res. Pap. 156, 103pp., Dep. Geophys.Sci., Univ. ing, in WeatherSatellites:Systems,Data, and Environmental
of Chicago, Chicago, Ill., !978. Applications, editedby P. KrishnaRao, chapVII-8, pp. 285-304,
Fujita, T. T., andM. R. Hjelmfelt, Mesoanalysisof recordChicago American Meteorological Society, Boston, Mass., 1990.
rainstormusingradar, satellite,and raingaugedata, in Proceed- Purdom,J. F. W., andJ. F. Weaver, Analysisof rapidscansatellite
ingsof the loth Conferenceon SevereLocal Storms,pp. 65-72, imageryto diagnosetornadicstormsand the environmentin
American Meteorological Society, Boston, Mass., 1977. which they form, in Proceedingsof the 6th Conferenceon
Heymsfield,G. M., and R. H. Blackmer, Jr., Satellite-observed SatelliteMeteorologyand Oceanography,pp. J25-J30,American
characteristics of midwest severe thunderstorm anvils, Mon. MeteorologicalSociety, Boston, Mass., 1992.
Weather Rev., 116, 2200-2224, 1988. Purdom,J. F. W., R. N. Green, and H. A. Parker, Integrationof
Hillger, D. W., and J. F. W. Purdom, Clusteringof satellite satellite and radar data for short range forecastingand storm
radiances to enhance mesoscale meteorological retrievals, J. diagnosticstudies,in Proceedingsof the 9th Conferenceon
Appl. Meteorol., 29, 1344--1351,1990. WeatherForecastingand Analysis,pp. 51-55, American Meteo-
Hillger, D. W., and T. H. Vonder Haar, Retrievaland use of rologicalSociety, Boston, Mass., 1982.
high-resolutionmoistureand stabilityfields from NIMBUS 6 Rotunno,R., and J. Klemp, On the rotationand propagationof
HIRS radiancesin pre-convectivesituations,Mon. WeatherRev., simulatedsupercellthunderstorms,
J. Artnos.Sci., 42, 271-292,
109, 1788-1806, 1981. 1985.
Johns, R. H., and W. D. Hirt, The derecho---Asevereweather Shapiro,M. A., T. Hampel,D. Rotzoll,andF. Mosher,Thefrontal
producing convectivesystem,in Proceedings of the 13thConfer- hydraulic head:A micro-cscale(- 1kin)triggering
mechanism
for
ence on Severe Local Storms, pp. 178-181, American Meteoro- mesoconvectiveweather systems, Mon. Weather Rev., 113,
logical Society, Boston, Mass., 1983. 1166-1183, 1985.
Lemon, L. R., andC. A. DoswellIII, Severethunderstorm evolu- Simpson,J. E., and R. E. Britter, A laboratorymodelof an
tion and mesoscalestructure as related to tornadogenesis,Mon. atmospheric
mesofront,Q. J. R. Meteorol.Soc., 106, 485-500,
Weather Rev., 107, 1184-1197, 1979. 1980.
Lubich, D. A., and J. F. W. Purdom,The use of cloudrelative Sinclair,P. C., andJ. F. W. Purdom,The genesisand development
animationin the analysisof satellitedata, in Proceedingsof the of deep convectivestorms,CIRA Pap. 1, 56 pp., 0737-5352,
6th Conferenceon SatelliteMeteorologyand Oceanography, pp. CooperativeInst.for Res.in theAtmos.,Colo.StateUniv., Fort
118-119,AmericanMeteorologicalSociety,Boston,Mass., 1992. Collins.
Maddox, R. A., Mesoscaleconvectivecomplexes,Bull. Am. Mete- Smith,W. L., et al., Nowcasting•Advanceswith MCIDAS III, in
orol. Soc., 61, 1374-1387, 1980. Proceedings of the NowcastingH Symposium,pp. 433-438,
Maddox,R. A., L. R. Hoxit, andC. F. Chappell,A studyof thermal EuropeanSpaceAgency,Kerplerlaan,Netherlands,1984.
boundary/severe thunderstorm interactions, of the Weldon,R. B., and S. J. Holmes,Water vaporimagery:Interpre-
in Proceedings
11thConference onSevereLocalStorms,pp. 403-410,American tation and applicationsto weatheranalysisand forecasting,
MeteorologicalSociety,Boston,Mass., 1979. NOAA Tech. Rep. NESDIS 57, 213 pp., Natl. Oceanicand
McCann, D. W., The enhancedV, a satellite observablesevere Atmos. Admin., Boulder, Colo.,
 Discussion

RON ALBERTY, SESSIONCHAIR

National Weather Service

PAPER D 1 the Red River approximately 100 nm from Norman. I believe

one or more of the mesocyclonesproduced tornadoes,but I
Presenter, Don Burgess, National Severe Storms Labora-
don't know if they all did.
tory [Burgess et al., this volume, Tornado detection and
warning by radar] (Ian Harris, STX Corporation) You briefly mentioned the
importance of a parameter called ERKE. I wonder if you or
(Jim Wilson, National Center for AtmosphericResearch)
Ralph Donaldson could explain a little more about that?
Don, I heard you say somethingabout the percentageof
mesocyclonesthat produces tornadoes. I've heard totally (Donaldson) We like ERKE because it works very well for
confusingnumbersfrom severalsources.Would you clarify the most intense tornadoes. It works quite well as Don
it for us? What is the number? showedfor the F2 and stronger tornadoes. Why it works, we
are not too sure. Part of it is becausewe are in the proper
(Burgess)It's a changingnumber from the sampleswe've
domain of where observations have shown that mesocy-
seen. It changeswith location and even with time period at
clones later produce tornadoes. What it represents is the
the same location. The first work of NSSL was analyzing
energy left after you subtract the velocity required to main-
data after the fact, no real-time displays. We used the
tain an average mesocyclone. Therefore ERKE represents
WSR-$7 radar to search for big, ugly-looking storms, and
some kind of a surplus available for providing a tornado.
then we pointed the Doppler radar at them. We found a
strong relationship between mesocyclones and tornadoes
PAPER D2
because we were looking at the most classic supercelIs.
After we began getting real-time displays, we found that Presenter, Steve Vasiloff, National Severe Storms Labora-
there were a lot more weaker circulations out there that
tory [VasiloJ•; this volume, Single-Doppler radar study of a
weren't classical features. Therefore the percentageof tor-
variety of tornado types]
nadic mesocyclonesdropped. During the JDOP Experiment,
where we were trying to look at all storms, we got something (HoMe Bluestein, University of Oklahoma) In the case in
close to 50% of the mesocyclones producing tornadoes. which it appears as if there is a strong TVS aloft and maybe
Those were years with many significant supercelIs. Since one near the ground, is it possible that the vortex it tilted?
that time, some years have had fewer supercelIs, and the
(Vasiloff) In the Stillwater case, there was no TVS aloft. I
percentagesfor certain periods have been as low as 30%. So
found the strongest shear at the lowest levels and high, but
a lot of mesocyclones don't produce tornadoes. It also
not maximum, shear above that. But by no means did the
dependson your threshold:How low a magnitudeof shearor
TVS extend to higher height than the 3.2øelevation you saw.
velocity, or whatever threshold you use, are you willing to
The circulation aloft was more likely associated with the
call a circulation that you think is important?I am going to
mesocyclone, and I'm not sure about the tilt.
say, with the thresholds that I am familiar with and advo-
cate, somewhere between 30 and 50% of the mesocyclones (Jim Wilson, National Center for Atmospheric Research)In
will produce tornadoes. Colorado tornadoes, there is no evidence that we ever see
anything build down. It's always ground up. You were
(Chair) We were operatingthe WSR-88D duringlast night's
showing something a little bit different.
storms, and we saw several mesocyclones,one down along
(Vasiloff) There were four different signaturesfor that June
The Tornado: Its Structure, Dynamics, Prediction, and Hazards.
day in Denver. I picked out the one that was most different
GeophysicalMonograph79
This paperis not subjectto U.S. copyright.Publishedin 1993by the to causediscussion,but it did appearto build down. We can
American GeophysicalUnion. compare data later if you like.
276 DISCUSSION

PAPER D3 don't remove storm motion, you need to look for change in
velocity, not necessarily a change from outbound to in-
Presenter,Mike Leduc, AtmosphericEnvironmentService
bound. Our algorithm uses velocity change, not necessarily
of Canada[Joe and Leduc, this volume, Radar signatures
crossingO radial velocity.
and severeweather forecasting]
(ArnoldCourt, California)I was wonderingwhy oneof your
PAPER D6
algorithmswasn't in terms of radiansper secondinsteadof
shear.
Presenter, Dave Imy, National Weather Service, Opera-
(Leduc) Paul, can you answer that? tional Support Facility [Imy and Pence, this volume, An
examination of a supercell in Mississippi using a tilt se-
(Paul Joe, Atmospheric Environment Service of Canada)
quence]
The nature of the Doppler estimates does not lend them-
selves to the use of that measure. (Keith Brewster, University of Oklahoma) You mentioned
doing RHIs or vertical cross sectionswhich you can do on
(Jeff Waldstreicher, National Weather Service, Eastern Re-
NEXRAD. Do you have any recommendations on the ori-
gion) I was wondering if you looked at the storm north of
entation of cross sections that would be the best?
Lake Ontario and if you subtractedout storm motion? You
had a storm moving away from the radars, and movement (Imy) If you know where your overhang is, the extent of
may have prevented seeing flow back toward the radar. overhang and the direction, you should set the cross section
to go right through them. In fact, we have done just that,
(Leduc) I haven't looked at that. Have you, Paul?
made cuts where we believe there to be an overhang, and we
(Joe) Yes, I have, and when you remove storm motion, you have seen bounded-weak-echo regions on NEXRAD cross-
do get a symmetricpattern, flow away and toward the radar. section products. For our study case with the WSR-57,
It is important to remember in real-time detection, if you looking for such features is nearly
 Laboratory Models of Tornadoes
CHRISTOPHER R. CHURCH

Departmentof Aeronautics,Miami University, Oxford, Ohio 45056

JOHN T. SNOW

Departmentof Earth andAtmospheric

Sciences,PurdueUniversity,WestLafayette, Indiana47907

1. INTRODUCTION attributesof experimentsin physicalscience,namely, pre-

cision,control, and repeatability, to a complexand transient
Nature providesmany examplesof intensebut small-scale atmospheric phenomenon.
atmosphericvortices,the mostdevastatingbeingtornadoes. At the preceding Tornado Symposium, Davies-Jones
Other small vorticesincludewaterspouts,fire whirls, dust [1976] gave a critical review of various TVCs that had been
devils, and steam devils. Several aspects of small-scale developed up to that time. His discussionof vortex core
atmosphericvortexflowsare of concernto the atmospheric dynamicswas based on available experimentalresults. The
scientist,namely: determinationof their kinematicstructure, general purpose of the present paper is to review the
understandingof their formation and dynamics,identifica- progressthat has been made since the 1976 symposium.It
tion of the factorsthat controltheir intensities,andapplica- dealswith severalaspectsof tornado vortex simulationand,
tion of new knowledgeand insightsin waysthat will provide in greater detail, with the contributionsto the understanding
greater protection for society from the hazards of these of vortex flows that have resulted from the more recent
phenomena.Although someof the vortex types listedabove laboratory experiments. The reader should be aware of other
occur more frequently and are more readily available for recent review papers on this subject: Maxworthy [1982],
observation than tornadoes, all small-scale vortices are
Snow [1982], and Wilkins [1988].
inherently infrequent, short-lived phenomena;it has been
expedient for some scientiststo simulatetornadolikeflows in
2. TORNADO VORTEX SIMULATION
the laboratory. This laboratory work constitutesa small but
significantpart of the overall tornado research effort. The purposeof the laboratory apparatusis to createflows
Swirling flows are also studied in several fields of engi- that, to the extent that it is possible,are geometricallyand
neering. The resulting literature on experimental work is dynamically similar to natural flows, in order that meaningful
voluminous, and although only a small fraction is pertinent observationsand physical measurementscan be made. By
to the meteorological phenomenon, it should not be over- injecting visualizing matedhal(for example, smoke and fog)
looked. An example of this material is the investigationof into the surface boundary layer, one is able to observe
vortex breakdown (a phenomenon described below) by details of the inflow and the core flow. The most important
Escudier [1982]. Maxworthy [1982] provides an entry bibli- measurementis that of velocity, although measurementsof
ography to some of this work. staticpressurehave also proven usefulin developinginsights
Tornadolike vortices are created in a tornado vortex
into vortex core dynamics. The accuracy of the laboratory
chamber (TVC). From necessity the backgroundflow con- data depends upon factors pertaining to the measurement
ditions are considerably simplified, as it does not seem technique and also upon the validity of the modeling tech-
feasible to incorporatein a laboratory model all the factors nique, that is, on details in the designof the TVC. All TVCs
(dynamic, thermodynamic,and microphysical)that exist in incorporate boundaries and walls that have no counterparts
the thunderstorm environment. A compelling reason for in nature. These cause the developmentof secondaryflows
taking the laboratory approach is that it brings the classic that may alter the vortex flows in some important ways.
The Tornado: Its Structure, Dynamics, Prediction, and Hazards. Davies-Jones'[1976]review of TVCs describesapparatusof
GeophysicalMonograph79 many different types and sizes, some with air, others with
Copyright 1993by the American GeophysicalUnion. water as the working fluid. He concluded that the most
278 LABORATORY MODELS OF TORNADOES

1 that form are scaledby the diameter of the exhaust hole (see,
e.g., Wan and Chang [1972]). The flow-straighteningbaffle
Exhaust also ensuresthat vorticity generatedby the exhaustfan does
not feed back into the swirling flow.)
The resultant axisymmetric flow is a representation of a
tornadoforming at the center of a mesocyclone.The updraft
diameter is identified with that of the tornado mesocyclone.
Examination of the streamline sketches in Figure 1 shows
that a secondarycirculatingflow existsin the lower corner of
the convectionregion, but this isjudged to exert a negligible
influence on the flow near the centerline.
In their analysisof swirling flows of this type, Lewellen
[1962] and Davies-Jones[1973] showed that a complete set
of governingequationsin terms of the dimensionlesscircu-
I I l'! "l'""li'"'"']
I" 1""'1I1"I I I I I i I t.[ I l I }' "i'""'i
1I L
lation and stream function involve three key characteristic
parameters'

Q
Rer = --
• Region
roF
S-
2Qh

i• . •1• Convergence where Rer is the radial Reynoldsnumber,S is the swirl

[ "'"" '• Region Rotating ratio, and a is the internal aspectratio. Here Q is the volume
Screen flow rate per axial length; •, the kinematicviscosity;r0, the
radiusof the updraft;h, the depth of the inflow; and F, the
circulation at r 0.
Swirl ratio is a particularlysignificantparameterin that it
has been found to determine the particular vortex configu-
ration, that is, one-cellversustwo-cell, singleversusmulti-
ple vortex flow.
Fig. 1. Conceptualsketchof the Ward-typeTVC. External geometricparametersformed from dimension-
less combinations of characteristic radial and axial dimen-
sions are significantin characterizingthe flow. These are
appropriateapparatusfor tornadovortex modelingwas the rs/to, r,•/ro, andl/h, wherers is the radiusof the conflu-
TVC developedby Ward [1972]. ence/convergence region,r,• is the radiusof the convection
region,and I is the depth of the convectionregion.
2.1. The Ward-Type TVC For all these dimensionlessgroupingsexcept one, a labo-
In this apparatusa swirlingflow is createdby injecting ratory TVC can match the expected atmosphericvalues
vorticity into a sinkflow, so that an intensevortex forms on [Church and Snow, 1979]. The exceptionis the radial Rey-
the centerline of a rotating upflow. Angular momentum is nolds number, which, when computedusing the molecular
continuouslysuppliedto maintainthe vortex. A schematic kinematic viscosityfor air, is severalorders of magnitude
form of the apparatusis shownin Figure 1. Its principal greaterfor atmosphericflows. Fortunately,for flows over
elements are as follows: (1) a means of supplyingstarting smoothsurfaces,the dependenceof flow characteristicson
vorticity and then circulation to a radial inflow; in the Reynoldsnumberappearsto be weak. For flowsover rough
original Ward apparatusthis was provided by a rotating surfaces,the dependenceon Reynoldsnumberhas not been
cylindricalmeshwire,a featurethat was alsoadoptedin the clearly established.
Purdue TVC I [Church et al., 1979]; (2) convergenceand
upflowcreatedby an exhaustfan mountedat thetop; and (3) 2.2. Current Status of Working TVCs
a flow-straighteningbaffle placed upstream of the fan to
provide a "free" downstreamboundaryconditionon the In comparingTVCs nowbeingusedto exploretornadolike
swirlingflow. (If a flow-straighteningbaffle followed by a flows, we find some common characteristics:(1) all are
largeplenumis not used, the core diameterof the vortices variantsof the originalWard apparatus,and (2) all
 CHURCH AND SNOW 279

TABLE 1. Tornado Vortex Chamber Characteristics

Principal Maximum Updraft Maximum Flow No. of Vane Spacing/Chord

Investigator/Location Radius,
m Rate,m3 m-• Vanes LengthRatio
Snow/PU 0.76 2.00 112 0.26
Wilkins/O U 0.61 1.56 NA '"
Monji/KU 0.38 0.55 48 0.5
Church/MU 0.26 0.57 18 0.92

PU, Purdue University;OU, University of Oklahoma;KU, Kyoto University; MU, Miami

University. NA, not applicable.

systemsof vanesin place of the rotating meshwireto provide movable flow-straighteningbaffle in order to vary the aspect
circulation. (Not all TVCs have been built as research tools. ratio of the convection zone, and (3) a laser Doppler veloci-
During the early 1980s, Kimpel [1981] developeda Ward- meter (LDV) for making nonintrusive velocity measure-
type TVC for education and demonstrationpurposes.This ments.

TVC was housed in the OMNIPLEX, a science and art The originalWard TVC at the University of Oklahoma
museum in Oklahoma City. It was viewed by a large number wasmodifiedby Rothfitsz[1986]in orderto producevortices
of people during its years of operation, and it probably from tilting of the horizontalvorticity presentin a helical
representedthe finest visual display of tornado vortex char- flow. This was accomplishedby replacingthe whole of the
acteristics ever made available to the general public.) Table confluence/convergence regionwith a three-tiersystemof
1 comparessomeof the principaldimensionalparametersof vanes. All the vanes in each tier were mutually parallel, but
various chambers. Maximum updraft radius is particularly the vane anglefrom one tier to the next veeredwith height.
significantsincethe radius of the vortex core scalesdirectly This flow had zero circulation about the vertical, a feature
with it. Typically, vortex core radii vary from 3% to 50% of that distinguishes
this physicalarrangementfrom all others.
updraftradius,dependingon swirl ratio. Tangentialveloci- This resulted in a complex flow pattern {,seeFigure 3).
ties in the vortex core scale directly with updraft velocity, Alternatequadrants had cyclonicandanticyclonicinflow.A
which in turn is a function of the flow rate. mesocyclonelikezone of rotationdevelopedwith a smaller-
A second-generationTVC at Purdue University is de- scale tornadolike vortex at its center. This vortex appeared
scribedby Snowand Lurid [ 1988]and illustratedin Figure2. more turbulent than those produced in other simulators.
In addition to the use of vanes and larger dimensions,the The TVC at Kyoto University [Monji, 1985]is illustrated
main differences from TVC I are (1) an antiturbulence screen in Figure4. An earlierversionof thisapparatus
[Mitstttaand
surroundingthe inflow and a confluencezone of varying Monji, 1984]incorporateda systemof four fansoperatingin
depth(the purposeof thesewill be madeclear later), (2) a an annularregion surroundingthe inflow. These supplied

•___•1
(•r•> EXHAUST
VFR TM
0.24 to 2.00 rn3s-I

,
.....
':"
....
" :
• ..t----A
.••'on••on
t!low
straigh
U ••jusabe
flow
straightener

!ower•
,..
:..... • II..Adjustab
vanes
Adjustable
surface

,-/ ,/ / / / /
Fig.2. Cross
section
ofTVCII atPurdue
University
illustrating
major
components.
Overall
horizontal
dimension
is
approximately
5 m. FromSnowandLund
280 LABORATORY MODELS OF TORNADOES

RV

5Ocrn

Convection region

• D (= :2R)---,

,
region
Guide vanes

Fig. 3. Plan view of the Ward TVC at Oklahomaas modifiedby

Rothfuszshowingone tier of vanes.Air parcelsenteringparticular
quadrantsfollow trajectoriesthat are eithercyclonicor anticyclonic,
as indicatedby the arrows. From Rothfusz[1986].

circulation but producedturbulent vortices at all swirl ratios.

It was foundpreferablefor later experimentsto configurethe
apparatusas is now illustrated. As in the PurdueTVC II, this
employs a relatively large number of flat vanes that can be
set at different angles. The annular vertical inflow feature of
the originalversionis retained;this may help to reduce the Fig. 4. Crosssectionand plan view of the TVC at Kyoto Univer-
disturbingeffectsof turbulentmotionsin the laboratoryair. sity. From Monji [1985].
Distinguishingfeatures of the apparatusdevelopedby
Churchat Miami University in Ohio are (1) aerodynamically
shaped vanes that resemble turbine blades, (2) a large 2.3. Practical Aspects
continuously variable iris diaphragm, and (3) termination of We considerbriefly some of the experimental problems
the vortex in a horizontally divergent flow. Flow is induced that confront investigators of laboratory vortex flows and
by 12 fans arranged around the top in a quasicylindrical some design aspects of tornado vortex simulators that have
configuration.The iris allows swirl ratio to be changedby proven useful in overcomingthese problems. These remarks
varyingr 0, eliminatinglaboriousand time-consuming adjust- should be helpful to potential designersand users of future
ment of the vanes. By this meansa wide range of vortex flow TVCs.
configurationscan be displayed in a matter of seconds.This Flow Stabilization. In fluid flow measurements,probes
investigator paid particular attention to producing stable, are placed at fixed locations, thus providing an Eulerian
clearly defined vortices in a chamber of compact dimen- description of the flow. The amplitude of "vortex wander"
sions. In order to accomplishthis, quite elaboratemodifica- (seeminglyrandom motions about the mean position of the
tions to the lower working surface were required, as illus- point of contact of the vortex core with the lower surface)
trated in Figure 5. The effectivenessof these featureswill be can easily be on the same order of magnitude as the core
discussed below. radius. Given the large gradientsof velocity and
 CHURCH AND SNOW 281

Perf. Board

To i let

'1'oSu• on Source

To 9uctlon Source
Smoke

Fig. 5. Cross sectionof a portion of the inflow region of the TVC at Miami University. Distance from centerlineto
screen is 1.2 m.

that are present, measurementsin such unsteadyflows are S. It has been found for the vane-type systemthat variations
difficult to interpret. While some degree of vortex wander in circulation largely compensate for variations in flow rate,
may be an inherent property of a vortex core normal to a flat keeping the ratio F/Q very nearly constant and so minimiz-
surface, two external causes of vortex wander have been ing variations in S.
identified: (1) turbulence in the air entering the TVC and (2) Church attempted to determine the optimum number,
flow separationat the lower edge of the inflow. shape,and spacingof vanes requiredto producea high sw,:rl
To address the first, an antiturbulence screenconsistingof flow without high turbulence and concludedthat (1) curved
a loosely woven fabric material surroundingthe inflow has vanes are preferable to flat ones, (2) curved platesarejust as
proven effective in reducing the strength of eddies drawn effective as thick airfoil shapes, (3) it is preferable to use
into the apparatus. Various schemeshave been tried to fewer vanes with longer chord length, and (4) the optimum
addressthe secondproblem. The Purdue TVC II employsa vane spacingto chord length (s/c) is about 1.
variable depth inflow geometry to keep the boundarylayer The size of the chamber tums out to be a limiting factor in
attached, although this feature also reducesthe effective the quality of the flow, particularly at the highest swirl
swirl angle in the chamber. The Miami TVC incorporatesa configurations.It has proven muchmore difficultto produce
circular gap placed close to the antiturbulencescreenand stablemultiple-vortexpatternsin the smallerMiami TVC (r 0
connected to a source of suction; however, this has not = 0.26 m) than in earlier experimentswith the PurdueTVC
producedany noticeabledifferencein the qualityof the flow. I (r0 = 0.61 m). In both TVCs, somemeansof extractingthe
If circulation and flow rate are not steady, the swirl ratio turbulent inflowing air close to the surface of the conver-
will vary. The consequentvariationsin flow configuration gence zone has proven very effective in improving the
will be reflected in measurements of velocity and pressure. stability of multiple-vortex flows.
This problemcan be greatlyreducedby usinga systemof Instrumentation for Veloci•, Measurements. It is re-
vanes in place of a rotating meshwire.In the latter case, quired to make measurementson flows that are inherently
variationsin circulationand flow rate are independentof one three dimensional and to derive information on mean veloc-
anotherand may therebycauserelativelylargevariationsin ities, turbulence intensities, and shear stresses.Many of
282 LABORATORY MODELS OF TORNADOES

TABLE 2. RecommendedDimensionalRelationshipsfor a Ward observations and to numerical simulations of laboratory

TVC vortices.

Dimensional Quantity Functional Relationship 3.1. Visual Observations

Overall radius of inflow, m 3.5 r0 Over the years there has developed a greater general
Depthof inflow,m 0.7r• '• awarenessof the complexity of tornadic flows, owing in part
Depthof convection
region,m 3r•-ø'5 to the results of laboratory simulation but also because of
Diameter of convectionregion, m 4r0
Volumeflowrate,m3 s-• r• increased skill in intercepting and photographing them.
Number of curved vanes 12-18 Ward [1972] was the first to observe the wide range of vortex
Vane turning angle, deg 60 flow phenomenato be found in axisymmetric swirling flows
Spacingbetween vanes, m r0 in the TVC, and by now many other experimentalists have
Vane chord length, m r0
documented them, the most recent being Monji [1985].
Church and Snow [1979] described the kinematics of the
different vortex flow configurations and presented photo-
graphsthat showed striking visual similaritiesbetween lab-
flow features of interest occupy very small volumes and oratory vortex configurationsand actual tornadoes.Various
contain large gradientsof velocity. Until now most velocity investigatorshave now concluded that the swift ratio is the
measurementshave been made using intrusive probes. How- dominantparameter that governsthe structure of the vortex
ever, experience has shown that these can cause problems. core. The point of transition from one flow configurationto
For example, the probeand its supportingstructure(perhaps another dependsnot only on S but also on radial Reynolds
large in size relative to the flow feature of interest) can number; however, the Reynolds number dependence has
perturb the flow. Because of the spiraling streamlinesnear been found to be slight at the higher flow rates. A summary
the core, it is easy for the probe to be in the spreadingwake descriptionof the evolution of the vortex core and the
of its own support.Further, when placedin a regionthat has developmentof thesevortex configurationsas a function of
a large velocity gradient, the sensingelement on the probe increasingS now follows.
may cut across a range of velocities, thereby resultingin a Vortex Core Formation. At very low values of swirl (S
loss of spatial resolution.The magnitudesof the resulting • 0.1), no concentrated vortex core exists at the surface.
errors are generally not known, and so the validity of the This is because of a ringlike zone of separated flow that
data becomesa subjectfor speculation.Such uncertainties forms in the inflow on the lower surface. It prevents low-
can be avoided or greatly reducedby resortingto a nonin- level angular momentum from approachingthe region close
trusive technique such as laser Doppler velocimetry. to the centerline. As S increases,a core zone developsaloft
Recommended Design Dimensions for a TVC. For and builds downward. For S -• 0.1, a concentrated core
greater clarity and measurement accuracy the vortex core makes contact with the surface.
shouldbe as largeas possible.As mentionedabove, the core intensificationand Breakdown. For S • 0.1, the flow is
radius is proportionalto the size of the updraft hole, so the characterizedby the following: (1) a thin inflow boundary
radius of the updraft (r0) is the parameter of greatest layer in which the depth decreaseswith decreasingradius,
importancein establishingthe dimensionsof the vortex. It (2) a cornerregionwhere the horizontalinflow turns into the
also turns out that all of the significant dimensions of the vertical direction, and (3) a cylindrical column of upflow
TVC can be expressedas functionsof r0. These empirically containing a rotational core and extending upward for a
derived relationshipsare presentedin Table 2. The func- considerable distance.
tional relationshipsshown were derived with economy in The core radiusincreasesgraduallywith height;its initial
mind, that is, producingthe largestvorticesin relationto the responseto an increase in swift is to contract. Vertical
overall size of TVC, and therefore these yield the minimum velocities (inside and outside the core) are everywhere
recommended dimensions for a given updraft radius and upward with a vertical velocity maximumalongthe center-
vortex core size. Experience has shown that it is very line; the vortex is thus said to exhibit a one-celledstructure,
difficult to develop the wide range of tornadolike vortex and this generalconfigurationis termedan end-wallvortex.
configurationswith updraft radii smaller than about 0.2 m. Exceptfor the very smallvaluesof S, at somelevel in the
The generalrecommendation is that r 0 be of the orderof 1 convection zone the core structure undergoes a dramatic
m. changedue to vortex breakdown.An adverseaxial pressure
gradientin the core culminatesin a free stagnationpoint
which becomesthe leading point of the vortex breakdown
3. EXPERIMENTAL RESULTS
(VBD). There is an abruptincreasein core radiusas upward
Investigatorshave developedinsightsinto vortexdynam- movingfluid divergesand flowsaroundan inner circulation
ics from visual observations of laboratory vortices and have termed the breakdown "bubble," which forms the boundary
made measurementsof velocity and pressure. In this section betweena supercriticalflow regimeupstreamand a subcrit-
we review the principal aspectsof such experimentsand ical flow regimedownstreamof that level. The inflow, the
discuss them where appropriate with reference to field flow in the corner, and the supercriticalcore are
 CHURCH AND SNOW 283

z
actualtornadoes. However,Lugt [1989]hassincepresented
severalexamplesof VBDs in atmospheric flows.Fromvideo
tapeof the funnelanddebriscloudproduced by the Minne-
apolistornadoof July 18, 1986,PauleyandSnow[1989]infer
the presence of a vortex breakdown.
As the inner core flow continues to decelerate in the axial
direction, another on-axis stagnationpoint occursat some
centralcore with \
higherlevel. Downstreamof this secondstagnation point,
verticalvelocitiesare small.In somesituations,
an organized
decelerating
upflow3• [ / /)l•11/
/ downflowis observed;in others, the inner core appears
stagnant.Mullen and Maxworthy [ 1977]found a core struc-
shear
zone
with ture similar to this in their dust devil model. On the other
accelerating
upflo•- hand,Pauley [1989]did not find evidenceof a meanupfiow
• ',
/
downstream of VBD for a wide range of flow conditions,
suggesting
that the total decelerationof the upflowand the
secondstagnationpoint had been incorporatedin the struc-
partially cl
ture of the breakdown bubble.
breakdown bubble
The far downstreamcore now appearsbroad and turbulent
and superficially resembles the turbulent wake downstream
breakdown of a bluff body placed in a straight-line flow. Vertical
vorticity is concentratedwithin an annularregionof upfiow
that surroundsa central downflow that contains compara-
tively little rotation. The concentriccombinationof upfiow
and downflow defines a two-celled vortex structure.
As the swirl ratio increases,the VBD moves upstream,
that is, towardthe lower surface,therebyreducingthe depth

boundary
hayer'
of the supercriticalregion and causingthe one-celledvortex
to intensify. With sufficient swirl a condition is reached
.... CSP
' "• r wherethe greatestverticalaccelerationsandlargestpressure
deficits are experienced in the surface layer (Figure 7).
Fig. 6. Schematicof the flow at low swirl, showingthe narrow Maxworthy [1972] has termed this condition a "drowned
iaminar end-wall vortex, breakdown bubble, and downstream the vortex jump" (DVJ). In the laboratory the DVJ occurs at S
enlarged turbulent core. From Snow [1982]. = 0.45. It representsone of the most significantfeatures of
a tornadicflow becausea tornado in this configurationmay
have its greatest impact on the surface.
laminar in appearance,though this is not always the case. Two-Celled Structure. As S is progressivelyincreased,
The subcritical flow downstream of the breakdown is almost the leading breakdown point penetrates to the surface,
always turbulent. followedby a generalradial expansionof the core regionand
The significanceof the criticality of the flow is that inertial penetration of the downflow to the surface. The vortex has
waves are able to propagateupstreamin a subcriticalregime now become two celled over its entire length. Helical waves
but not in a supercriticalone. In a low-swirl one-cellvortex, develop and propagatedownstream in the annular cylindri-
this prevents filling in of the supercriticalflow from above, cal shear zone that surroundsa nearly quiescentinner core.
and low central pressuresand high velocities are sustained As S increases further, the core radius increases, the shear
(otherwise, the pressure gradient force within the core, zone thins out, and higher-order helical modes are excited. A
which is directed upstream, would promote a centerline more complex vortex configuration containing two inter-
downflow from aloft). The flow is turbulent immediately twining spiral vortices emerges. At first these appear close
downstream of the breakdown bubble with a diminished and together at the surface, as if forming a node, but additional
decelerating upflow near the axis and maximum vertical swirl causes them to move apart and take up "antinodal"
velocities in an annular region surroundingit. positionsat the surface. They then resemble two separate
The flow immediately downstreamof the VBD is unsteady tornadolike vortices, each spinningabout its own axis and
and has many variants, so verbal descriptionsof it as well as circulating about the centerline at the mean radius of the
physical measurementsare challenging. The VBD features turbulent column. This occurs for S = 1.0. Further swirl
shownin Figure 6 are therefore a simplifiedrepresentationof produces higher-order multiple-vortex patterns; vortex
the flow. Various types of VBD have been photographedand cores containing up to six subsidiary vortices have been
discussedby Sarpkaya [1971], Faler and Leibovich [1977], documentedin the laboratory. These higher-order multiple-
and Phillips [1985]. At the time when VBDs were first being vortex flows are very difficult to maintain, and only at the
observed in TVCs, it was not clear whether they occurred in lowest Reynolds numbers, because in the TVC they are
284 LABORATORY MODELS OF TORNADOES

\ \

\ \
\ \
\ \

centrifugal

.:

"drowned"

ø•CS
P
Fig. 7. Schematicof a one-celledflow showingthe "drownedvortexjump" configuration,in which the highest
velocitiesand greatestpressurefalls are experiencedat the surface.From Snow [1982].

closely spaced that they mutually interact and generate they will not be produced in a conventional (axisymmetric)
turbulence that causes them to break up. Tornadoes (and TVC.
dust devils) have often displayedthis type of multiple-vortex
behavior (see, e.g., Fujita [1976]).
3.2. Velocity Fields
The complex nature of tornado damage can often be
attributed to these types of flows. The combination of the, Measurements With Hot Fihn Probes. The three-
translation of the parent thunderstorm and the rotation of dimensional and turbulent nature of vortex flows makes
subsidiary vortices about their own centers and about the velocity measurementsdifficult. In regions where the mean
centerline of the parent vortex leads to an inhomogeneous flow can be assumed to be two dimensional, effective mea-
pattern of damage with a maximum impact at points where surementshave been made using single hot film probes. By
the velocity componentsadd together most effectively and a this means, Baker and Church [1979] measured the core
reduced impact in nearby locations where the velocity radius and the average maximum core velocity (Vm) as a
componentstend to cancel one another. At the present time, function of S for various flow rates. The mean updraft
for a tornado formed in a flow of a given sink strength, it is
velocity (w) could be used as a scaling parameter, and
not clear whether the DVJ configuration or the multiple-
resultsshowedthat the ratio Vm/w remainednearlyconstant
vortex configuration would produce the most violent effect
at about 2.6 through a wide range of meteorologically
on surface structures. This merits further investigation.
important swift ratios (including the VBD and multiple-
It would be incorrect to think that laboratory TVCs have
vortex modes).
been able to reproduce all of the vortex flow features that
have been seen in nature. Over the past several years there Baker [1981] further developed the technique of velocity
hasbeen a proliferation of automaticstill and video cameras measurementwith hot film probes. By carefully combining
that has made it possible for members of the general public, single hot film measurementswith a smoke filament tracing
perhaps sometimesinjudiciously, to obtain excellent docu- method, he was able to resolve three componentsof velocity
mentation of tornadoes. Some of the observed features serve at selected points. A detailed set of measurements was
to remind us that nature continues to hand out surprises in obtained for a vortex with a swift ratio S = 0.28. This flow
terms of new and unusual vortex features. It may be that the was one celledthroughoutthe depth of the inflow layer. The
formation of some of these features depends on an absence VBD was located above the level of the updraft hole, in the
of axial symmetry in the atmospheric flows, in which case lower part of the convection region. Some of the results
 CHURCH AND SNOW 285

Fig. 8. Velocity profilesobtainedby Baker [1981]for a vortex of S = 0.28 {,datapoints)and the numericalsimulation
results (solid lines) of Wilsonand Rotunno [1986]at four dimensionless radial positions:(a) r* = 0.0475, (b) r* =
0.1025, (c) r* = 0.2125, and (d) r* = 0.75. From Wilson and Rotunno [1986].

shown in Figure 8. (The points are Baker's data; the signif- creased, though they are still smaller than the tangential
icance of the lines will be discussedunder the next heading.) velocities.Further, a significant"nose" hasappearedin the
Figure 8d shows data for a radial position far from the vertical profile of radial velocities, with a peak at z* •
vortex core (r* = 0.75), one where the flow is almost 0.025. Now the inflow boundary layer has a two-layer
entirely horizontal (the vertical velocity increasesfrom zero structure.A theoretical treatment by Burggraf et al. [ 1971]
at the surface to some very small value at z* = 1). describedthis type of structure. They suggestedthat the
Tangential velocities are small; radial velocities are even boundarylayer comprisesan inner viscoussublayerand an
smaller. A weak maximum in the u profile can just be outer inviscid layer.
discerned at z* • 0.03. Figure 8b showsa continuationof the trend toward higher
Moving nearer to the centerline, the velocity components valuesof u, v, and w. The thicknessof the viscoussublayer
in Figure 8c can be characterizedthus: vertical velocities has now diminished;the peak in the u profile now occursat
increasewith height but are still small; tangentialvelocities z* -• 0.02. At this location (about three core radii from the
have substantially increased; radial velocities have in- centerline)the strongradialjet achievesits maximum
286 LABORATORY MODELS OF TORNADOES

Closeto the core, in the turningregion(Figure 8a) the

radial velocitiesbecome reduced. Over a wide range of
heightsthe vertical and tangentialvelocitieshave quite
similar magnitudes. Baker also measured vertical velocities
along the centerline and showed that w attained its maximum
value at z* -• 0.25 and diminishedgraduallywith height
tii!l
•.Viscous
i Outer
Flow
I!•!•Subcore
\ uz=wr
toward the level of the VBD.
In reviewing the maximum values of u, v, and w that
occurred in the flow, the following conclusionsare drawn:
(1) Urnaxwas found at z* • 0.02, r* • 0.1; (2) Vmaxwas o.5o ', % -
foundat z* -• 0.25, r* • 0.05; (3) Wmax wasfoundat z* •
0.25, r* = 0; and (4) the quantitiesUmax
:Vmax
:Wmaxwerein
the approximateratio 1:2: 4, indicatingthat one-celledvor-
tex flows contain comparablylarge values of velocity in all I•l EffectivelyInviscid

0.25
• Rotational
Region
"',.,,,,
.,.
three coordinate directions.
Numerical Simulations. In view of the difficulties in-
volved in trying to simulatetornadoesnumerically,a number • r• =r2-dH d (F• "--•
r •'-•'•
\T! /Viscous
Sublaye
of investigators have instead addressed numerical simula-
tions of the flows in the Ward TVC. For a more complete
review of these studies, see Wilkins [1988]. In this paper we 0 02.5 0.50 0.75 i.0
focus on the results of Wilson and Rotunno [1986]. Of
particular significanceis their determination of the velocity
field in a vortex flow that was identical to the one studied by Fig. 9. Cross section of the convergence region showing the four
Baker. The solid lines in Figure 8 show the results of their principal vortex flow domains deduced by Wilson and Rotunno
numerical simulation. The main differences are found in the [1986]: irrotational outer flow, effectively inviscid rotational inner
flow, viscous surface sublayer, and viscous subcore around the
vertical velocity field, where it appears that the numerical centerline. From Wilson and Rotunno [1986].
simulation may have predicted a somewhat larger core
radius than was found experimentally. However, in general
the numerical results match the laboratory data remarkably surface static port in a radial track. A sensitive transducer
well. This independent validation of the experimental work measured the static pressure deficit, that is, the difference
tends to undermine the reservations expressed earlier about between the surface static pressure at the various radial
the validity of measurements made with intrusive velocity positions and the surface static pressure at the edge of the
probes. It would be useful to extend the numerical simula- TVC. A sample of their results is shown in Figure 10.
tion technique to include vortex flows at other swirl ratios, in Principal conclusionswere as follows'
order to see if the velocity components continue to be 1. In one-celled vortices the radial pressure gradients
related in the same way. were generally of larger magnitude, and the central region of
From an analysis of the terms in the momentum equations, greatest pressure deficit, the "pressure well," was more
Wilson and Rotunno categorized the flow in the convergence closely confined to the centerline than in two-celled flows.
zone as shown in Figure 9. The hyperbolic dashed line 2. The magnitude of the central pressure deficit was
divides an outer irrotational flow from the inner rotational sometimes larger in one-celled vortices than in the two-
region. Thus we see the concentrationof horizontal vorticity celled flows.
in a viscous surface sublayer and in an effectively inviscid 3. Radial variations of surface static pressure inside the
layer above it. Vertical vorticity is concentrated in a cylin- cores of fully developed two-celled vortices were generally
drical viscous subcore that is concentric with the centerline small in comparison with the magnitudes of the maximum
and in the effectively inviscid region that surrounds the core pressuredeficits.
viscous subcore. Pautey et al. [1982] examined how the central surface
pressuredeficitsvary as a function of S and also presented
3.3. Pressure Fields
measurementsof the steady state surface pressurefields in
multiple-vortex flows. Figure 11 is an example of how the
Measurements With Pressure Probes. As static pressure central surface pressure deficit respondsto changesin swirl
is omnidirectional, it can be measuredwith comparativeease ratio. The pressure values have been nondimensionalized
using a singlepressure probe, although an investigatorstill with pw2 (the flow forceper unit area).The symbolsin
has to deal with the problem of fluctuating signals. Useful Figure 11 identify a vortex configuration for each corre-
insightsinto vortex core dynamicshave been obtainedfrom spondingrange of swirl ratios as one-celled laminar (L) or
staticpressuremeasurementsmadeon the surfaceand along two-celled turbulent (T) flow containing two fully developed
the central axis. Surface pressurefields as a functionof swirl multiple vortices (2), transitions from two to three multiple
ratio were mappedby Snow et al. [1980]. They installeda vortices (2-3), and three multiple vortices (3). Figure
 CHURCH AND SNOW 287

•g
x,d

R• (NON-OIM RROIU

Fig. 10. Radial profiles of time-averagedsurfacestatic pressuredeficit for swirl ratios ranging from zero (curve a) to
1.79 (curve g). From Snow et al. [1980].

presents the data in two ways: the time-dependentextrema increasedrapidly with S, and an approximately cubic depen-
of the fluctuating signal (solid line) and the time-averaged dence on S was found. The general form of this relationship
pressurevalues (dash-dotline). There are clearly substantial could be accounted for in terms of the similarity theory of
differences between the two quantities. The results show Burggraf et al. [ 1971].
that vortex flow fields contain very large central pressure 2. The largest value of pressure deficit anywhere in the
deficits. For example, they are larger by an order of magni- flow was Ap* --• 60 (curve e); this occurredjust upstreamof
tude or more than the pressurecoefficientsthat are tradition- the VBD with the VBD located at z* • 0.3. The difference
ally found in wind tunnel flows. The most prominentfeature in pressure between this point and the central surface point
of these data is closeto the point of transitionfrom a laminar can be used to derive a value for the maximum axial velocity
to a turbulent vortex and is attributedto the developmentof (Wmax)in terms of the mean updraft velocity (w). Thus Wma x
the DVJ. The very large values of central pressuredeficit = 10.5w at z* -• 0.3 for S = 0.31.
associatedwith this configuration(AP* = 40) were also 3. The largest value of pressuredeficit at the surfacewas
found for a variety of flow rates and inflow geometries. AP* - 38 (curve g), closely matching the corresponding
Church and Snow [ 1985]examinedthe axial distributionof value of Pauley et al. [ 1982] for the DVJ configurationunder
pressurealong the centerlinein a seriesof nine one-celled identical flow conditions.
and two-celled vortices over swirl ratios rangingfrom S = 4. Downstream of the VBD in the two-celled region of
0.18 to S = 0.70. The results are presentedin Figures 12 the vortex the pressuredeficitsbecamerelatively small, with
and 13 and are summarized thus: ZXP* • 10 typically and diminishinggradually with height.
1. In one-celled vortices the central pressure deficit The static pressureprofiles above z* • 0.3 did not contain
increased from a small value at the surface to a maximum as much detail as they did below that level.
(i.e., a pressureminimum) at z* -• 0.3 and diminished Pauley [1989] made additional axial pressure measure-
graduallywith height above this level. At this level, AP* ments, focusingon two-celled vortices. He arguedthat in
288 LABORATORY MODELS OF TORNADOES

•o

,,
2 J 2-• Ii :5 ii 3-4

Re,.= 2.90 ß 10•'

h = .305 m

r. , ,.•08 m

Fig. 11. Dimensionless

centralsurface
staticpressure
deficit
versus
swirlratio,showing
time-averaged
data(dash-dot
curve)andtime-dependent
maximumvalues(solidline).FromPauleyet al. [1982].

turbulent flow downstream of the VBD, his time-averaged spherethe effectiveviscosityis likely to be someordersof
data capturedthe characterof the axial pressuregradients magnitudegreaterthanthe kinematicviscosityand alsomay
betterthanthetime-dependent measurements of Churchand vary in a complexmanner. Church and Snow [1985] con-
Snow [1985]. Pauley's valuesof AP* were typically about cludedthat in spiteof the relativelylargepressuredeficits
one half of those of Church and Snow, with significantly that were measuredin laboratory vortices, surface pressure
better spatialresolution.Figure14 showsan exampleof his deficitsin tornadoesare expectedto be small. So far there is
data. From thisit was possibleto deducethe characterof the no evidenceto the contrary. It seemsthat the most effective
downflow and to see that the strongestdownflowsoccurred way of inferringpressuresinside tornadoeswould be by
at middle levels in the two-celled flows. Pauley also con- carryingout integrationson the derivedvelocityfields.
cluded that (I) turbulent stressesare an importantfactor in The experimental pressure(andvelocity)datawere used
helpingto maintainlow pressuresand high velocitiesnear by Fiedler and Rotunno[1986]in their developmentof a
the surfaceand(2) numericalsimulationsof tornadoeswould theoryfor maximumwind speedsin tornadoes.The theory
be improvedff they were to includeturbulencemodeling. addressedfour physicalquantities:(!) the maximumtangen-
Applicationsto Tornadoes. It is temptingto try to use tial velocity(Vmax) in the supercritical
flow, (2) the maximum
the laboratorypressuremeasurements to infer directlythe
axialvelocity(Wmax) in the supercriticalflow, (3) the maxi-
minimum pressuresin tornadoes,but there are serious
difficultieswith this. It is probablethat the floorof the TVC mum tangentialvelocity (Vsub) in the subcritical flow (i.e.,
downstreamof the VBD), and (4) the maximum swirl ratio
(smoothin these experiments)was not as "rough" as the
thatroughening possible
Earth's surface.Pauley [1980]demonstrated (S*) to maintainan end-wallvortex(i.e., the DVJ
the flow with a coveting of carpet causeda substantial configuration).
The followingrelationships
werefoundfrom
reduction in the measuredvalues of pressure.The effective theory:
viscosityof the fluidis alsoa problem.Theoryimpliesthat
Vma
x • 1.7Vsu
pressuredeficitsvary inverselywith viscosity.In the atmo-
 CHURCH AND SNOW 289

.[4
Curve S Curve S
O.18
f 0.37
O20
g 0.45
022 h O.55
02_6 i O.7O
0.$1

'LO

(•=0.6 .O.8
a=0.6 0.8
Rer=2_.88x
I04 Rer=
2.88xlO
4

O.6
O6
Vortex
Bre(•kdown

.0.4 0.4

O2

0.0
60 50 40 ,•p•O 2.0 10 0 60 50 40 •.p.•30 20 10 0
Fig. 12. Dimensionlesscentral pressuredeficit versusdimension- Fig. 13. Dimensionlesscentral pressuredeficit versus dimension-
lessheight for five end-wall vorticesfor the rangeof swirl ratio 0.18 lessheight for four vortices for the range of swirl ratio 0.37 < S <
<- S -< 0.31. (As was noted by Pau!ey [1989], owing to an error in 0.70. Curve g (S = 0.45) is for the DVJ configuration. From
measuring volume flow rate, the Ap* values here and in Figure 13 Church and Snow [1985].
shouldbe increasedby about 30%.) From Churchand Snow [1985].

3.4. Surface Roughness

w max• 2.1 v max The issue of surface roughnessadds one more dimension
S* =0.43.
of complexity to the investigation of tornado dynamics. It is
clear from the results of several investigatorsthat the nature
It was demonstrated that the velocities that had been of the surface over which vortices form has a significant
derived theoretically were in very good agreementwith the impact on their flow structure and intensity. Basic questions
corresponding experimental values. The maximum swirl that have been addressed concern how the swirl ratio, core
ratio was also close to the experimental value of 0.45. radius, velocities, and pressuresare affected by changesin
Fiedler and Rotunno applied their resultsto the problemof the roughnessof the surface. To a certain extent, laboratory
the "thermodynamicspeedlimit," that is, that the maximum experiments, field observations, and numerical simulations
possible wind speeds in a saturated atmosphericcolumn have provided answers, but there is still scope for much
containinga Rankine vortex have been shownby thermody- more work. The results of previous efforts are summarized
namicargument to be •65 ms-]. Thisismuchsmaller than
-1
in Table 3. In some cases there appearsto be good agree-
the wind speedsin some tornadoes.By taking the 65 m s ment; in others there appear to be conflictingconclusions.
limit as the valueof Vsub,they thusobtainedvaluesfor •,max Effect of Roughness on Swirl Ratios. There seemsto be
• 110m s-• andWmax '• 220ms-]. Whilethelatterfigure a consensusamong investigators as to the effect on swirl
appears to be somewhathigh, these are more in line with ratio of increasingsurface roughness'higher swirl vortices
what damageanalysessuggestfor maximumwind speedsin tend to form more readily over smooth surfaces.Increasing
tornadoes. the roughnesscauses an increased frictional dissipation
290 LABORATORY MODELS OF TORNADOES

the surface layer, causing transition to a lower swirl config-

uration. Alternatively, a greater angular momentum input is
needed to maintain the same vortex configuration over a
rough surface. The experimental work of Desserts [1972],
Wilk#•s et al. [1975], Leslie [1977], and Rostek and Snow
[1985], as well as the numerical study of Ayad and Cermak
[1980], all support this conclusion.
The preceding sounds straightforward, but the following
discussionillustrates the need to delve more deeply in order
to provide substantive answers to practical questions. A
typical question might be: if a tornado moves from smooth
ground to rough ground, is it going to produce more or less
damage?Field observationsare conflicting. For a tornado in
Indiana, Baker et al. [1982] found that after the tornado had
entered a grove of trees the damage increasedwith depth of
penetration. For tornadoes in France, Dessens and Snow
[1989] also concluded that those occurring over forested
areas had a greater than average intensity, path width, and
path length. On the other hand, Monji and Wang [1989]
showed that a tornado in Japan after having caused light
Fig. 14. Dimensionless central pressure deficit versus dimension- damage in an urban (rough) area moved over smoother
less height. Time-averaged data for a two-celled vortex (S = 0.85). terrain and caused much heavier damage. There is no simple
Strongest downflow occurs in region II. From Pauley [1989]. way to resolvethis issue. Considerthen the following points:

TABLE 3. Summary of Effects of Increasing Surface Roughness

Reference Comments z0 X 104m Results
Dessens [1972] laboratory measurement 0, 20 Smaller azimuthal, larger vertical velocities; transition
from 2-cell to 1-cell state; core radius larger and more
turbulent.
Bode et al. [1975] numerical study Smaller azimuthal, larger vertical velocities; increased
core radius.
Blechman [ 1975] field observa!ion Transition to lower swirl state; increased inflow
velocity.
Wilkins et al. [1975] water tank Increased convergence and vertical velocity; larger core
radius; less intense vortices.
Leslie [ 1977] laboratory measurement 0, 8.5 Flow more turbulent; effective swirl reduction of---30%.
Leslie [ 1979] laboratory measurement Core radius decreasedat small aspectratios, increased
at large aspect ratios.
Lewellen and Sheng [1979] numerical study Increased radial and vertical velocities.
Ayad and Cermak [ 1980] numerical study Transition from 2-cell to 1-cell state.
Nakamura and Nakama [1980] laboratory measurement Core radius increased.
Baker [1981] field observation Damageincreasingwith penetrationafter enteringgrove
of trees.
Diamond and Wilkins [1984] laboratory measurement O, 34 Core radius decreased; effect of translation was to
increase effective swirl and core radius.
Rostek and Snow [1985] laboratory measurement 0, 0.5,2.4, Effective swirl reduced; thin ropelike vortex for medium
3.6 roughnessat S = 0.1; VBD over roughestsurfaces
did not penetrateto surface, owing to separatedflow
region;rough surfaceenhanceshigh-pressurering;
little variation in core radius.
Monji and Wang [1989] laboratory measurement 1.5, 5.6, 15 Smoke-filled core radius increases; thin ropelike vortex
extends to surface in roughestcase; location of
maximum winds moves upward; surface winds
decrease;zone of high radial winds broadens and
deepens;zone of high tangentialwinds lifts and
shrinks.
field observation Considerableincreasein damagedensity and path width
upon leaving urban (rough) area.
Dessens and Snow [1989] field observation Transition to lower swirl state. Tornadoes over wooded
areas are more intense and have greater than average
path widths and
 CHURCH AND SNOW 291

z= 20(½m) z:

SMOOTH 'SMOOTH
50
i ßROUGH
5O
ROUGHI ..,,,•
ROUGH
1I .,.
,,•,,>-
3O
30 ßROUGH
10 10

i i , I, _ I . ! I .. I ,_ L JL ....

0.s .0 o.i o.5 1 .o

Fig. 15. Diameter

ofsmoke-filled
vortex
coreversus
swirlratioforsurfaces
ofdifferent
roughness
andforheights
of
5 and20 cm. Inflowdepth:47 cm. FromMonjiand Wang[1989].

1. If a vortex is one celled, rougheningthe surfaceis datashownin Figure 16 provideanotherway of determining

goingto causeit to movefurtheraway(deintensify)fromthe coreradius.Here the trend is not soclear:It appearsthat the
DVJ configuration,thereby reducing near-surfacewind core radius for the rough I surfaceis larger than either the
speed. smoothor roughII surface, which is puzzling.
2. If a vortex is two celled, rougheningthe surfaceis Effect of Roughnesson Velocities. In a numericalstudy,
goingto causeit to move closerto the DVJ configuration,
thereby increasingnear-surface winds.
3. It is to be expectedthata roughsurfacecreateshigher
radial velocities and lower tangential velocities than a v Smooth
1 60
smooth surface.
20 80 ' ' 12010080 60
Does increasedroughnesscausethe total velocity vector
in the surfacelayer to increaseor decrease?The answer is
180:•
.-..'-
not clear. To summarize,the answerto an apparentlysimple : ::;i
questionis evidentlycomplexand may ultimatelyrequirea
knowledge of how much swirl there is and how much the
surfaceroughnesschangesin a given situation.Thesefactors
require more careful scrutiny in future laboratorywork. v Rough I
Effect of Roughnesson Core Radius. A majorityof the
laboratory experiments and the numerical studiesof Bode et
2 1 20•00
!20 80 60
al. [1975]andLewellenand Sheng[ 1979]showedan increase
in core radius with increasing surface roughness.The fact 10
that some investigatorsfound the oppositeeffect indicates
that there is somedegree of complexityto this issuealso; the ! (b) 40
result may depend not only on how core radius is measured, i ! .............. i. ' ß ! .... l

but where it is measured. The most recent data are those of

Monji and Wang [1989]. They used small wooden cubes as o Rough
roughness elements and worked with three surfaces:
smooth, rough I, and rough II (the latter beingthe roughest).
20 ,//}} 12010080 60
They measured the diameters of smoke-filled cores at two I t
levels as a function of swirl ratio. The results are shown in 2
Figure 15. Both levels show a systematic increase in the 40
diameter of the smoke-filled core with increasingroughness (c)...,?
for all values of S, although the effect is less pronounced at
lo is 2o ts io
the higher values (i.e., S > 1). Monji and Wang also r (cm)
measured horizontal velocities using a hot wire velocity Fig. 16. Tangential velocity fields for vortices formed under the
probe and a minivane which enabled them to resolve the same background flow conditions over surfaces of different rough-
tangential and radial components. The tangential velocity ness. From Monji and Wang
292 LABORATORY MODELS OF TORNADOES

an increasedconvergenceaccompaniedby higher radial and

vertical velocities and reduced tangential velocities.
Effect of Roughness on Pressures. As mentioned above,
Pauley [1980] briefly examined the effect of surface rough-
ness on the magnitudesof surface pressures. A more sys-
tematic series of measurementswas performed by Rostek
and Snow [1985]. The entire flow of the TVC was covered
with a pegboard in which up to 5500 roughnesselements
(wooden pegs) were installed. The densities of the pegs
determined the roughness characteristics of the surface.
Roughness lengths were determined using separate peg
board test surfacesplaced in a wind tunnel.
Experiments were performed using one smooth surface
and four of different roughness. The general characters of
the swirling flows made visible by smoke injection were
observed as a function of increasing roughness. Radial
profiles of surface static pressure deficit were obtained for
u Rough
each of the surfaces. The general conclusions were as
140. 80 60 follows:

1 1. For a given input of angular momentum, a vortex over

a rough surface has a structure and an associatedpressure
deficit profile similar to that of a vortex over a smooth
surface with a smaller input of angular momentum.
(½) •0
, ,,
2. Surface roughness has a significant impact on one
'fo ' i'S i• .30 stagein the evolution of the vortex, namely, the extension of
r (era) the concentrated core downward to make contact with the
surface. The evolution of the concentrated core and the
Fig. 17. Same as Figure 16, except radial velocity fields. From descent of the breakdown bubble as seen over smooth
Monji and Wang [1989]. surfaces becomes less distinct when observed over rough
surfaces. Instead, a smaller subcore reached down through
the thicker boundary layer and made contact with a localized
surface inflow.
Lewellen and Sheng [1979] showed that the following oc- 3. In order to establishmultiple-vortex flows over rough
curred with increasing roughness: (1) an increase in the surfaces,a greater angular momentum input is required.
maximum radial and vertical velocities (nondimensionalized 4. In comparing pressure profiles of smooth and rough
by the maximum tangential velocity), (2) an increase in the surfaces, a region of nearly zero radial pressure gradient
ratio of the height to radius at which the maximum tangential within the core and surroundingthe central pressure well
velocity occurred, (3) a decreasein the maximum tangential was observed for rough surfaces.
velocity, and (4) an increase in the radius of maximum This last may be seen in Figures 18 and 19 by comparing
tangential velocity. The experimental results of Monji and the pressureprofilesfor the smoothestand the least rough of
Wang (Figures 15, 16, and 17) are generally consistentwith the surfaces; the feature of interest is located between 1- and
the conclusions of Lewellen and Sheng. The increase in 7-cm radius. This feature may be compared with the notch or
radial velocities with roughnessis striking; the details of the step observed within the core region of the 1962 Newton,
trends in the tangential velocity field appear less systematic. Kansas, tornado [Ward, 1964]. It has also been seen in other
Monji and Wang also showedthat there was little variation in barograph records of tornadoes. No generally accepted
the magnitude of the horizontal velocity vector (sum of physicalexplanationfor its appearancehas yet been devel-
radial and tangentialcomponents)with roughness.The main oped. (This feature shouldnot be confusedwith the ring of
effect was that the maximum horizontal velocity was found higherpressuresurroundingthe centralcore that occursfor
close to the surface in the smooth case, and with increasing low swirl vortices.)
roughnessthe positionsof the velocity maximumwere found Several workers have now investigatedthe effect of vary-
at progressivelyhigher levels. The radial positionsof the ing the surface roughnessof the lower boundary and have
velocity maximum were relatively unchanged. It is not found that this exerts a significantimpact on vortex dynam-
known at what swirl ratios these measurements were made, ics. It is necessaryto ask whether the roughenedlaboratory
and they do not seemto addressthe complex natureof the surfaces were valid representations of the Earth's surface
issue as it was represented earlier in section 3.4. To con- characteristics.Table 3 lists the roughnesslengths(z0) that
clude, the effect of increasing roughnessis to increase the were determinedfor the experimental studies. It has been
frictional dissipationof angularmomentum.This results in estimated that the ratio of the depth of the
 CHURCH AND SNOW 293

Radial Location (cm) suitedto studiesof swirling flows. For example, the appli-
5 10 15 20 •5 cationof variousgradesof sandpaperis an idea which merits
further consideration in future laboratory experiments.
From the foregoing discussionit is evident that further work
is needed to clarify the issues pertaining to surface rough-
-2 ness.

4. SUMMARY AND SUGGESTIONS FOR FUTURE WORK

Surf'ace The past decade and a half have seen significantprogress

a=O. 50 in the developmentof our understandingof the dynamicsof
S=0•54 tornadic flows. Laboratory modeling has played an impor-
Re?4x!O tant role in this progress. Tornado vortex chambers have
demonstrated the range and complexity of atmospheric
-8
vortex configurations.Greater skill in documentingtornado
Fig. 18. Radial profile of time-averagedsurfacestaticpressurefor events has led to a general confirmation of the similarity
a vortex flow (S = 0.54) over a smooth surface (surface A). From between vortex flows observed in nature and those observed
Rostek and Snow [1985]. in the laboratory. Physical measurementsmade in the labo-
ratory have provideda significantamountof informationon
velocity and pressurefields. Numerical simulationsof TVC
boundary layer to the TVC boundary layer falls in the range flows using the latest generation of computershave demon-
3000 to 8000. An upper bound for the roughnesslength of the strated the feasibility of developing computer models that
Earth's surface is about 3 m [Arya, 1988], z0 for most capturethe principalfeaturesof the laboratoryvorticesand
"rough" terrain being of the order of 0.1-1 m. From an which lead to consideration of issuesof practical concern,
examination of the experimental roughnesslengthsin Table such as maximum intensities in actual tornadoes. Tornado
3 it would appear that a number of the experiments were interceptteamsnow seemto be on the thresholdof providing
performed over surfaces of extreme roughness,and conse- physical measurementson tornadoes, which could then be
quently the effect of roughnessmay have been overstatedin used to help validate numerical models of the full-scale
some of the results. flows.
Another question concerns the means of preparing the Concerning the future of laboratory measurements, we
roughsurface. Someinvestigatorsusedcarpetof convenient offer the following suggestions:
thicknesses,but carpet is a nonstandardworking material. 1. The LDV technique, althoughmore difficult to imple-
Others followed the acceptedwind tunnel practiceof using ment than other technologies,is the only one capableof
regularlyspacedarraysof roughnesselements.Thisplacesa measuring the flow without perturbing it. LDV measure-
constrainton the spatial resolutionof the measurements.It ments on vortices will provide measurementsof the mean
is concludedthat other types of rough surfacemay be better values of three componentsof velocity at different swirl
ratios. These data shouldoverlap and extend thosethat have
beenobtainedpreviously by other methods.From measure-
Radial Location (cm) mentsof the fluctuatingcomponentsit will be possibleto
determine turbulence intensities and Reynolds stresses,
0 5 10 15 20 •5 :30
0 • "'• • • ...... • '" which will lead to a more completeempiricalevaluationof
the relative importance of the terms in the momentum
equations.
2. Questionsremain concerningsurfaceroughnessef-
fects. Although much of the data obtained so far seem
consistent with the basic physics, there is still no clear
understandingof how the maximum near-surfacewinds
+++ + • Surf'ace B dependon roughness.It is thereforedesirableto continue
• 4.4.
++4.
• '• ß ß a=0.50 the type of measurementsoutlinedin suggestion1 above
-G +• ß ß ' S=0 54' using surfacesof suitable roughness.
,, . . . Re•4x184 3. As the dependenceof flow characteristicson Rey-
nolds number, especially for roughenedsurfaces,has not
been established,it would be interestingto vary the kine-
Fig. 19. S•e as Figure 18 exceptfor a surfaceof roughness maticviscosityof the workingfluid.This wouldallowoneto
length5 x 10-5 m. SurfaceB wastheleastroughof theprepared determinehow the thicknessof the boundarylayers, magni-
surfaces.•ofile is virtuflly identic• to that of Figure18for radial
positionsgreaterthan 7 cm. High-pressure featureof interestis tudesof velocityand pressure,and vortex transitionpoints
located between I and 7 cm. From Rostek and Snow [1985]. vary with viscosity. This could be done by settingup
294 LABORATORY MODELS OF TORNADOES

closedcirculationsystemand, in place of air, establishinga Church, C. R., J. T. Snow, G. L. Baker, and E. M. Agee,
Characteristics of tornado-like vortices as a function of swirl
flow of a differentgas, for example, carbon dioxide. (Carbon
ratio: A laboratory investigation, J. Atmos. Sci., 36, 1755-1776,
dioxide has approximately one-half the kinematic viscosity 1979.
of air.) Such an undertaking would lend itself best to the Davies-Jones,R. P., The dependenceof core radius on swirl ratio in
smaller TVC which would require less elaborate modifica- a tornado simulator, J. Atmos. Sci., 30, 1427-1430, 1973.
tions to develop a closed system. Davies-Jones, R. P., Laboratory simulations of tornadoes, in Pro-
4. A question which has never been addresseddirectly ceedingsof the Symposiumon Tornadoes:Assessmentof Knowl-
through experiment is how does a tornado interact with the edge and Implications of Man, pp. 151-173, American Meteoro-
logical Society, Boston, Mass., 1976.
electrical environment of a thunderstorm. Although infre- Dessens, J., Influence of ground roughnesson tornadoes: A labo-
quent, numerous documented cases of luminous glows and ratory simulation, J. Appl. Meteorol., 11, 72-75, 1972.
other evidence of electrical dischargesassociatedwith tor- Dessens, J., and J. T. Snow, Tornadoes in France, Weather
nado funnels have now been compiled. Somehow,electric Forecasting, 4, 110-132, 1989.
Diamond, C. J., and E. M. Wilkins, Translation effects on simulated
fields develop that are close to breakdown strength:orders
tornadoes, J. Atmos. Sci., 41, 2574-2580, 1984.
of magnitude greater than what is normally found near the Escudier, M.P., Vortex breakdown and the criterion for its occur-
surface under a thunderstorm. The surface of the Earth is a
rence, in Intense Atmospheric Vortices, edited by L. Bengtsson
plentiful sourceof ions. One suggestedexperimentwould be and J. Lighthill, pp. 247-258, Springer-Verlag,New York, 1982.
to determine how a vortex flow can influence an ionized Faler, J. H., and S. Leibovich, Disrupted states of vortex flow and
surface layer and whether it could locally causea sufficient vortex breakdown, Phys. Fluids, 20, 1385-1400, 1977.
Fiedler, B. H., and R. Rotunno, A theory for the maximum
amplification of the background electric field to produce
windspeedsin tornado-like vortices, J. Atmos. Sci., 43, 2328-
glowlike discharges. 2340, 1986.
Finally, it should be borne in mind that in a laboratory Fujita, T. T., Graphicexamplesof tornadoes,Bull. Am. Meteorol.
experiment it takes an investigator of the order of 1000 s to Soc., 57, 401-412, 1976.
make a single measurement. A supercomputer capable of Kimpel, J. F., An interactiveweatherexhibitat OMNIPLEX, Bull.
Am. Meteorol. Soc., 62, 1219-1223, 1981.
one billion operationsper secondis able to generatevelocity
Leslie, F. W., Surface roughnesseffects on suction vortex forma-
profilesfar more rapidly than the experimentalist.The main tion: A laboratorysimulation,J. Atmos. Sci., 34, 1022-1027, 1977.
question is the validity of the model. It is evident that for Leslie, F. W., The dependenceof the maximum tangentialvelocity
maximum economy, laboratory measurements must be on swirl ratio in a tornado simulator,in Preprints, 11th Confer-
made sparingly, an essential role of the laboratory work ence on Severe Local Storms, pp. 361-366, American Meteoro-
being to provide information that can be used to fine tune logical Society, Boston, Mass., 1979.
Lewellen, W. S., A solutionfor three-dimensionalvortex flows with
numerical models of swirling flows. strongcirculation,J. Fluid Mech., 14, 420-432, 1962.
Lewellen, W. S., and Y. P. Sheng, Influence of surface conditions
REFERENCES on tornado wind distributions, in Preprints, 11th Conference on
Severe Local Storms, pp. 375-381, American Meteorological
Arya, S. P., Introductionto Micrometeorology,307 pp., Academic,
Society, Boston, Mass., 1979.
San Diego, Calif., 1988.
Ayad, S.S., andJ. E. Cermak, A turbulencemodelfor tornado-like Lugt, H. J., Vortexbreakdownin atmospheric columnarvortices,
Bull. Am. Meteorol. Soc., 70, 1526-1537, 1989.
swirling flows, in Vortex Flows, edited by W. L. Swift, P.S.
Barna, and C. Dalton, 171 pp., American Society of Mechanical Maxworthy, T., On the structureof concentrated,columnarvorti-
ces, Astronaut. Acta, 17, 363-374, 1972.
Engineers, New York, 1980.
Baker, D., E. Agee, G. Baker, and R. Pauley, The Rush County, Maxworthy,T., The laboratorymodelingof atmosphericvortices:A
Indiana, tornadoof 9 July 1980,in Preprints, 12th Conferenceon critical review, in Intense Atmospheric Vortices, edited by L.
Severe Local Storms, pp. 379-392, American Meteorological Bengtssonand J. Lighthill, pp. 229-246, Springer-Verlag, New
Society, Boston, Mass., 1982. York, 1982.
Baker, G. L., Boundarylayersin laminarvortex flows,Ph.D. thesis, Mitsuta, Y., and N. Monji, Developmentof a laboratorysimulator
143 pp., Purdue Univ., West Lafayette, Ind., 1981. for small-scaleatmosphericvortices,Nat. Disaster Sci., 6, 43-54,
Baker, G. L., and C. R. Church, Measurementsof core radii and 1984.
peak velocitiesin modeledatmosphericvortices,J. Atmos. Sci., Monji, N., A laboratoryinvestigation
of the structureof multiple
36, 2413-2424, 1979. vortices, J. Meteorol. Soc. Jpn., 63,703-713, 1985.
Blechman, J. B., The Wisconsin tornado event of April 21, 1974: Monji, N., and Y. Wang, A laboratoryinvestigationof the charac-
Observationsand theory of secondaryvortices, in Preprints, 9th teristics of tornado-like vortices over various rough surfaces,
Conference on Severe Local Storms, pp. 344-349, American Acta Meteorol. Sin., 3,506-515, 1989.
MeteorologicalSociety, Boston, Mass., 1975. Mu!len, J. B., andT. Maxworthy,A laboratorymodelof dustdevil
Bode, L., L. M. Leslie, and R. K. Smith, Numerical study of vortices, Dyn. Atmos. Oceans, I, 181-214, 1977.
boundary effects on concentratedvortices with applicationto Nakamura, I., and N. Nakama, Some features of a tornado-like
tornadoesand waterspouts,J. R. Meteorol. Soc., 101, 313-324, vortex core (in Japanese),Sutnmary29, pp. 39-47, Philos. Dep.,
1975. Univ. of the Ryukyus, Okinawa, Japan, 1980.
Burggraf,O. R., K. Stewartson,and R. Belcher,Boundarylayer Pauley,R. L., Laboratorymeasurements of surfacepressuremin-
inducedby a potentialvortex, Phys.Fluids, 14, 1821-1833,1971. ima in simulator tornado-like vortices, M.S. thesis, 91 pp.,
Church, C. R., and J. T. Snow, The dynamicsof natural tornadoes Purdue Univ., West Lafayette, Ind., 1980.
as inferred from laboratory simulations, J. Rech. Atmos., 12, Pauley, R. L., Laboratory measurements
of axial pressuresin
11!-133, 1979. two-celled tornado-like vortices, J. Atmos. Sci., 46, 3392-3399,
Church, C. R., and J. T. Snow, Measurementsof axial pressuresin 1989.
tornado-like vortices, J. Atmos. Sci., 42, 576-582, 1985. Pauley,R. L., andJ. T. Snow,On the kinematicsanddynamics
 CHURCH AND SNOW 295

the 18 July 1986Minneapolistornado,Mon. WeatherRev., 116, Severe Local Storms, pp. 323-326, American Meteorological
2731-2736, 1989. Society, Boston, Mass., 1988.
Pauley, R. L., C. R. Church, and J. T. Snow, Measurementsof Snow,J. T., C. R. Church,and B. J. Barnhart,An investigation
of
maximumsurfacepressuredeficitsin modeledatmospheric
vor- the surfacepressurefieldsbeneathsimulatedtornadocyclones,J.
tices, J. Atmos. Sci., 39, 368-377, 1982. Atmos. Sci., 37, 1013-1026, 1980.
Phillips, W. R. C., On vortex boundarylayers, Proc. R. Soc. Wan, C. A., andC. C. Chang,Measurements of the velocityfieldin
London, Set. A, 400, 253-261, 1985. a simulatedtornado-likevortex usinga three-dimensional velocity
Rostek, W. R., and J. T. Snow, Surface roughnesseffects on probe, d. Atmos. Sci., 29, 116-127, 1972.
tornado-likevortices, in Preprints, 14th Conferenceon Severe Ward, N. B., The Newton, Kansas tornado cyclone of May 24,
1962,in Preprints, 11th Weather Radar Conference,pp. 410-415,
Local Storms, pp. 252-255, AmericanMeteorologicalSociety,
Boston, Mass., 1985.
American Meteorological Society, Boston, Mass., 1964.
Ward, N. B., The exploration of certain features of tornadodynam-
Rothfusz,L. P., A mesocyclone andtornado-likevortexgenerated ics using a laboratory model, J. Atmos. Sci., 29, 1194-1204, 1972.
by the tilting of horizontal vorticity: Preliminary results of a Wilkins, E. M., Influence of Neil Ward's simulator on tornado
laboratory simulation,J. Atmos. Sci., 43, 2677-2682, 1986. research, CIMMS Rep. 87, 64 pp., Univ. of Okla., Norman, 1988.
Sarpkaya, T., On stationaryand traveling vortex breakdowns,J. Wilkins, E. M., Y. Sasaki, and H. L. Johnson, Surface friction
Fluid Mech., 45,545-559, 1971. effects on thermal convection in a rotating fluid: A laboratory
Snow, J. T., A review of recent advances in tornado vortex simulation, Mon. Weather Rev., 103,305-317, 1975.
dynamics, Rex,. Geophys., 20, 953-964, !982. Wilson, T., and R. Rotunno, A numerical simulation of a laminar
Snow, J. T., and D. E. Lund, A secondgenerationtornadovortex end-wall vortex and boundary layer, Phys. Fhdds, 24, 3993-4005,
chamberat Purdue University, in Preprints, 15th Con•krenceon
 Laser Doppler Velocimeter Measurements
in Tornadolike Vortices

DONALD E. LUND AND JOHN T. SNOW

Department of Earth and AtmosphericSciences,Purdue University, West Lafayette, Indiana 47907

1. INTRODUCTION cell. A Doppler shift arises as a result of the relative motion

of a scatteringparticle and two plane waves (one from each
Using data recently gatheredwithin the PurdueUniversity beam) that approach the beam crossing (the "measuring
tornado vortex chamberII (PU TVC II) usinga state-of-the- volume") from different directions. The velocity component
art laser Doppler velocimeter (LDV), we presentand discuss in the plane of the beams and normal to the optical axis is
radial and vertical profiles of measuredradial and tangential detectedas a beat frequency, or frequency difference, pro-
velocity components and derived vertical velocity compo- duced by the heterodyne mixing on a photodetector of
nent. A mean offset of the vortex from the centerline of the
backscattered light from particle interactions with each
tornado vortex chamber (TVC) required a coordinate trans- plane wave in the measuring volume [Adrian, 1983]. The
formation of the measured data from a "chamber" reference
opticsare shownin Figure I configuredto measurethe radial
frame to a "vortex" reference frame. Further, the impact of velocity component. The entire assembly is rotated 90ø to
small vortex translations (vortex wander) on the data set is measure the tangential velocity component. A precision
examined through the use of a simplemathematicalmodel of traversing mechanism (r _- 0.0005 cm, z --+ 0.0025 cm)
the data collection process. allows positioningof the measuringvolume anywhere in an
rz planecuttingthe convergenceand confluencezonesand
2. LABORATORY APPARATUS the lower portion of the convection zone.
Following Ward [1972], the PU TVC II is configuredto
give geometric and dynamic similarity to tornadolikeflows 3. MEASUREMENT PROGRAM
(Figure 1). The portion of the lower chamber beneaththe
dividing surface is referred to as the confluencezone, and 3.1. Configuration of the Apparatus
the region below the updraft hole is called the convergence The measurementregion of the TVC was configuredfor
zone. The upper chamberis termedthe convectionzone. In the data set presentedhere with an updraft hole of radius
the PU TVC II, a mechanically driven throughflowis de- r 0 = 20.395 cmandan inflowdepthof h = 30.383 cm.The
flectedby an assemblyof vanesat the outerperimeterof the swirl ratio, the dimensionlessparameter which characterizes
lower chamber. The vortex initially forms along the center- the nature of the flow in a TVC [Davies-Jones, 1973] by the
line throughthe convergenceof vorticity and is maintained ratio of the mean swirl velocity v 0 to the mean throughflow
by a steady inflow of angular momentum.See Snow and velocity u0, both measuredat r 0, was S = rovo/2huo =
Lund [1987] for more details concerningthe PU TVC II. 0.222, a moderatevalue. The volume flow rate per unit axial
Figure 1 depictsschematically the majorcomponents of depthwasQ = 2rrr0u 0 = 2.35 m2 s-l. Thusthe radial
the LDV as positionedin the PU TVC II. In this systemthe Reynolds number, Re,. = roUo/V,was2.46x 104,wherev
outputof a He-Ne laseris transmittedvia a fiberopticcable - 1.52 x 10-5 m2 s-1 is kinematic viscosity. These condi-
to an opticalsystemthat splitsthe singleinputbeam.The tions resulted in a turbulent single-celled vortex in the
two resultant beams are focused to cross in the working convergencezoneand lower one half of the convectionzone
volume of the TVC. To facilitate detection of flow reversal, (the only portions of the vortex that were visually exam-
one of the beamsis frequencyshiftedby an acousto-optic ined). No vortex breakdown was observed; however, past
experiencesuggeststhat for conditionssuch as these a
breakdownwas likely presentin the upper one half of the
The Tornado'Its Structure,Dynamics,Prediction,andHazards. convection zone.
Geophysical Monograph 79
Copyright1993by the AmericanGeophysical
Union. Radialand tangentialvelocity componentswere measured
298 LASER DOPPLER VELOCIMETER MEASUREMENTS

COMPUTER

LASER
, .

•--SlGNAL
PROCESSOR
Fig. 1. Cross-sectionalview of the Purdue University tornado vortex chamber II showing the arrangement of the
major componentsof the laser Doppler velocimeter in the work spacebeneaththe "floor" of the chamber.

in the convergence zone along zenith lines at radii of 5, 10, values of 0• and 02, while correspondingpositionsalong the
15, and r 0 = 20.395 cm and along a chamber radius at LOM yield r• and r 2. These four parameters,together with
elevations of 1, 3, 6, 15, and h = 30.383 cm. Measurements the assumed circular symmetry, uniquely determine a, /3,
consisted of a minimum of 384 individual realizations accu- dr, l, and R. The measured velocity components may then
mulated according to particle arrival rates within the mea- be transformedto an orthogonalpair in the vortex reference
suring volume. Velocities presentedhere are mean values of frame throughIVhl and a and the positionreassignedthe
histograms constructed in this fashion. value R.
Typical results of the transformation are also depicted in
3.2. Radial Profiles of Velocity Components Figures 2 and 3, where dr = 0.369 cm and l = -0.115 cm.
In Figure 2 the transformation has the effect of moving each
Figures 2 and 3 show typical radial profiles of mean radial
data point upward and to the left, shifting the points toward
and tangential velocity components, in this case for an
what might be expected for corner flow, while in Figure 3,
elevation of 15 cm • h/2, i.e., at midheight. In these figures
little change is noted. Careful inspection suggeststhat away
and subsequent discussion, lowercase variables indicate a from the axis the effect of this transformation is minimal and
vortex chamber or absolute reference frame centered in the
that near the axis, for R < 0, the radial component data are
updraft hole, while uppercase variables denote a reference
undercorrected, while for R > 0, they are overcorrected.
frame centered on the vortex axis. The peculiar pattern in
The coordinate transformation has little impact on v since a
the measured
dataof Figure2 for Irl < 4 cm (asopposed
to
what might be anticipatedfor a corner flow) is to be noted.
meanoffsetalongthe LOM simplyresultsin a shift of origin;
a shift normal to the LOM (but within the core) results in a
One possible cause for this peculiarity is an offset in the
long-time mean position of the vortex with respect to the greatervortex tangentialvelocity, V(R), at the location of
center of the updraft hole. This would place the line of the measurement,but measuring only the component of
measurement (LOM) along a chord of the vortex polar V(R) normal to the LOM has a compensatingeffect.
This transformation fails near the core since it assumes the
reference frame. Consequently, near the axis of the vortex,
the velocity component along the LOM would be highly vortex is steady. In practice, the vortex constantly wanders
influenced by the tangential componentof the vortex veloc- about the mean offset position. We discuss the effect of
wander on the near-axis measurements in section 4.
ity field.
To determine and correct for this offset, a coordinate At first glance,the profile of Figure 3 appearssimilar to a
transformation technique was applied that assumesthe mag- classical Rankine-combined vortex (an irrotational outer
nitudeof the meanhorizontalvelocityvector,[Vhl, was flow surroundinga core in solid rotation). However, a least
constant on a circle of fixed radius R with respect to the squaresfit to the circulationfield impliedby thesemidheight
mean vortex position. For each elevation, all measured data indicatesthat for R > 1 or 2R c (where Rc is the core
(u, v) pairsalongcorresponding radial profileswere usedto radius, here estimated to be about !.08 cm), the outer flow is
constructa plot of Ivt versusposition,r, alongthe LOM, proportional
to I'-0'63 ratherthanr -•. This reflectsthe
where u and v are the velocity components measured along three-dimensional,viscousnature of the underlyingswirling
and normal to the LOM, respectively. Figure 4 shows the corner flow. (We will adopt this midheight value for the core
coordinatetransformationgeometry; (u, v) pairs determine radius as a reference length scale in what
 LUND AND SNOW 299

Radial Profileof RadialVelocity

z = 15.000 cm

r,R (cm)

* Measured
• Transformed
,[
Fig. 2. Radialprofileat 15-cmelevationof measured
radialvelocitycomponent
(pluses,lowercase
variables)and
results of coordinate transformation(asterisks,uppercasevariables).

Radial Profileof Tangential Velocity

z = 15.000 crn

10

8-

6-

4-

2-

0-

'2'

-4'

'6-

-8-

-10
.- 1'0 25
r,R (cm)

Measured
• Transformed
Best
Fit .I
Fig.3. Radial
profile
at 15-cm
elevation
of measured
tangential
velocity
componentandresults of coordinate
transformation.
Symbols andnotation
asin Figure
2. Solidlinerepresents
theradialdistribution
of tangential
component,
V = 27.1F•R(R
-<Rc),V = 4F•(R•/R)n(Rc
- R -<R•.),resulting
from
a "best
fit"tothecirculation
fieldof thetransformed data.Forouterradius
R:•takentobe25cm,thebestfit givesRc = 1.08cm,n = 0.63, and
far-fieldcirculation
F:, • 0.329m2 S -1 (Weadoptthismidheight
valueof Rc asourreferencelengthscale.)
300 LASER DOPPLER VELOCIMETER MEASUREMENTS

Coordinate Transformation Geometry

iVl

Fig.4. Schematic
depiction
ofcoordinate
transformation
geometry.
Measured (u,v)pairs
provide
01,02,rl, andr2.
Theseparameters
andcircular
symmetry
uniquely
determine
a, 13,dr, l, andR. Measuredpairscanthenbe
transformed
throughIV;z[anda andreassigned
to the position
R.

3.3. VerticalProfilesof VelocityComponentsin the surfaceis prominent

at a radiusof 20 cm andstillevidentat
Outer Flow a radius of 10 cm.
Verticalprofilesof tangentialvelocitycomponent
(Figure
Figure5 showsverticalprofilesof radialvelocitycompo- 6) showan overallincreasein V with decreasing R at all
nent at nominalradii of 5, 10, and 20 cm; a profile at nominal
levels.The no-slipboundaryconditionassociated with the
15-cm radius was similar to the 10-cm profile and has been
undersideof the upper dividingsurfaceis manifestin the
omittedfor clarity. The overallflow field is one of acceler-
ation below 5 cm and deceleration above 5 cm. Estimatesof profileat R ---20 cm.Thatthelargestvaluesof tangential
velocitycomponent
radial convective accelerationsat the level of the "nose" of appearnearz '-- 30 cm for R = 5, 10,
and 15 cm reflectsthe convergence of the mainbody of the
the developing surfaceinflowjet (z -'- 1 cm)givevaluesof
flowintotheupdrafthole.Fluidelements"spinup" asthey
4.7.qat R ---7.5 cm,anddecelerations (wheretheflowturns
upward)wereabout8.59 at R ---2.3 cm.Theflowin thisjet approach
tum.
theaxisowingto conservation of angularmomen-
attains
a peakmeasured
speed
of about2.75m s-•. The
locationof thismaximumis in agreementwith Baker's[1981] The oscillatorypattern in the profile at R --- 5 cm seems
results, which indicate
a peakvelocity
ofabout 0.86ms-• at too systematic to attributeto normalscatter.Thereis an
indicationof oscillationsin the profileat R --- 10 cm as well.
(r, z) "' (4, 1) cm.Thehighervelocityin thepresentcaseis
consistent withthe largerReynolds number (2.46x 104 A similarpatternis evidentnearerthe axisin the Table 1
versusBaker's1.22x 104)andthesmallervalueof h (30cm dataextractedfromradialprofiles.The V' in Table 1 are the
versus Baker's 41 cm). measuredmaxima at the indicated levels. These observa-
The near-constantinflow speedin the vicinity of z "- 5 cm tionspossibly
reflectthe presence
of an inertialstanding
marks the level where frictional effects have become small wavein the core.Similarphenomena
havebeenobservedby
enough
to allowcyclostrophic
balanceto establish
itself.A Church et al. [1979] and modelednumerically by Rotunno
fluid elementcan approachthe axis with nearlyconstant [1979].SnowandLurid[1989]describe inertialmotions
with
speedwhiletheopposing centrifugalandpressure gradient a simpledynamicalmodel.
forces continueto build. This is consistentwith the radial A plot of a verticalprofileof dimensionlesstangential
velocitycomponent
momentumbalancesof Wilson and Rotunno's [1986]numer- at r ---20 cmis shownin Figure7. The
icalexperiment.
Above5 cmtheradialmomentum to friction
begins velocity,
v* = (,w/p)2/2, isusedto normalizez and
decreasewith decreasingradius,graduallyat lower eleva- v to produce
the"innervariables"
z + andv+. Theprofile
tions, then more rapidly as the flow turns to exit the exhibits a logarithmiccharacterfor 100< z + < 300anda
convergence zonethrough theupdrafthole.Thepresence of linear behaviorfor z + < 10. Phillipsand Khoo [1987]also
withtheupperdividing observedthis behavior in their LDV measurementsof
a strongradialinflowjet associated
 LUND AND SNOW 301

Vertical Profiles of U

30-

25-
• '"',.•.

O
-3.o -2.5 -2.o -1.5 -1.0 -o.5 0.0
u (m/s)

--*--
R=4.68
cm--•--
R=9.68
cm,xR=20.1
cmI
Fig.5. Vertical
profiles
ofradial
velocity
component
showing
thedevelopment
ofa low-level
inflow
jet between
surface
and5 cm.Notethatradialcomponent
magnitude
increases
totheleft;thenegative
signdenotes
inflow.

Vertical Profiles of V
35,

30-

25-

N 15

10

0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

v (m/s)

--,,.-R=20.1cm
• R=14.7cm
= R=9.68cm
: R=4.68cm
I
Fig. 6. Corresponding
verticalprofilesof tangentialvelocitycomponentfor the situationin Figure
302 LASER DOPPLER VELOCIMETER MEASUREMENTS

TABLE 1. Measured Maxima of Velocity ComponentNormal near the core where an axial jet "erupts" from the boundary
to LOM layer. Vertical velocities nearly quadruple between 1 and 3
-1
Z, cm V', m s-1 R, cm cm near the axis and reach a maximum of about 10.5 m s
at z = h. These findings are in qualitative agreement with
1.00 9.49 0.626 Baker's [1981] measured vertical velocities. In both cases
3.00 8.53 0.141
6.00 8.84 0.300
thejet appearsto erupt inside a radius of about 4 cm at 3-cm
15.00 7.16 0.642 elevation. Baker's peak velocity approaches the somewhat
30.38 7.58 1.122 lowervalueofabout4 m s-• whereas
in thepresent
casethe
vertical velocity at the same dimensionlessheight is at least
5.5m s-• . Thisisconsistent
withthehigherRer andsmaller
r0 in the presentcase. While direct comparisonswith the
vortex generatedin a rotating tank of water. White [1974, p.
LDV measurementsof Phillips and Khoo [1987] are more
473] describes this structure as a universal characteristic of
difficult owing to differences in experimental apparatus,
all turbulent boundary layer data. This figure shows that
similarprofile shapesexist at about 3-cm elevation, and the
credible near-wall velocity measurements with the LDV
axial jet erupts from the boundary layer inside a radius of
wereattainedwell into the linearsublayer,downto z + "- about 6 cm.
2.5 (here equal to 1.65 ram).
Rough estimates of vertical convective accelerations in
the eruptingjet yield values around 34g. To put this in a
3.4. Estimates of the Vertical Component of Velocity different perspective, a fluid element moving through a
Radial profilesof vertical velocity component(Figure 8) divergence
fieldof 1 m s-• cm-• at 3.3m s-• experiences
a
were derived from the measured radial velocity component 34g acceleration.This occursat a relatively low elevationof
via an integratedcontinuity equation.In recognitionof the z •" 2Rc, just abovethe entrance(at z '• R c) to the annular
difficulty in resolvingthe profilesnear the core, these data volumecontainingthe greatesttangentialspeeds;see below.
were selectivelyreduced, eliminating"bad" pointsby in- The vertical velocity profile for z = h reflectsthe presence
spectionprior to being used for computations.Advantage of the upper boundaryat outer radii, where the flow turns
wasagaintakenof circularsymmetryby reflectingpointsfor abruptly as it passesthe edge of the updraft hole.
R < 0 about both coordinate axes, thus enhancingthe data
set.
4. ASSESSMENTOF THE IMPACT OF VORTEX WANDER
The weak vertical velocities near the surface at outer radii To better assessthe impactof vortex wanderon near-axis
firstincreasegraduallywith decreasingR, then morerapidly measurements, a numerical simulation of the measurement

End-WallBoundary
LayerProfile
r = 20.125 crn

25

l White
Eqn.
(6-52a)
ßv+:(1/0.41)1nz
++5.0I
2O-

15-

/ •

+> lO-
/

o
"" Linear
Sublayer:
v+=z+I
1 ' " ....... lO "' ...... -i6o .... ""'"boo
1
Z+= ZV*/"d
v* = 0.0314 m/s

Fig.7. Boundary
layerprofiles
plotted
interms
of"inner
variables"
[White,
1974],
dimensionless
height
z+ and
dimensionless
tangential
velocitycomponent
v+, showing
the linearandlogarithmic
layersthat are universally
characteristic of turbulent boundary layer
 LUND AND SNOW 303

Radial Profiles of W
11

9-

-- 30 . 383 cm

z -'- 14.986 cm

z = 6.000 cm

z = 3.000 cm

z = 1.000 cm

0 :• 4 6 8 10 12 14 16 18 20
R (ore)

Fig. 8. Radialprofilesof derivedverticalvelocitycomponent.

An edgeeffectassociated
with the uppersurfaceis
evident in the profile for z = h.

processhas beendeveloped.In this model,lengthmeasures The first numericalexperiment considereda stationary

are defined as in Figure 9. All quantities are taken to be vortex with the offset normal to the LOM having a typical
dimensionless,with the core radius as the length scale and value of I = -0.500. The resultingradial velocity profiles
the actualtangentialvelocitymaximumasthe velocityscale. are shownin Figure 10. The model-measuredu(r) closely
Using prescribedfunctionsfor the radial distributionof resemblethe profile of measureddata in Figure 2. The
radialand tangentialvelocitycomponents, the velocitycom-
ponentsalongandnormalto the LOM (seeFigure9) canbe
expressed as Geometryof Instantaneous
MeasuredVelocityComponents
/i(r) -- 1•'(r) + t7(r) (1)

v(r) = v'(r) + •(r), (2)

where u(r) and v(r) are measuredradial and tangential
velocity components,respectively.Quantitiesu'(r) and
v' (r) arethe contributions
fromactualradialandtangential
velocities,U(R) and V(R), respectively,and are givenby

lit(F)---IU(R)I cos(•) 1,3)

= IV(g)l sin(/3). (4)
Finally,t?(r),thecontribution
to themeasuredradialveloc-
ity component fromV(R), and/•(r),thecontribution
to the
tangentialcomponent from U(R), aregivenby
a(r) = Iv(g)l cos
•(r) = IU(g)l sin(a). {,6)
Fig. 9. Schematic
depictionof thegeometry
for an instantaneous
For thisanalysis,
radialvelocities
weretakento bepropor- measurementof the velocity componentalong or normal to the
tionalto the squareroot of the radiusand normalized
to LOM. The measuredvelocity componentsu and v are each com-
preserve
theratioU(rO)/Vma
x inFigures
2 and3.Thetangentialprisedofcontributions
frombothactualvelocitycomponents
U and
V. LDV MV denotes the location of the measuring
velocity
fieldwastakento be a Rankine-combined
vortex.
304 LASER DOPPLER VELOCIMETER MEASUREMENTS

Modelfor RadialVelocityProfile
[ = .o.$0o

0.2

_>, 0.1-
o
'• 0.0-

-•- -0.1-

rr -0.2-
'• -0.3-
o

c- -0.4-
E
• -0.5-

-0.6
-20 -'•0 0 1'0 2'0 3'0 40
DimensionlessRadius,r

Fig. 10. Mathematical model for the profile of radial velocity componentthrough a stationary vortex with a mean
offset normal to the LOM of I = -0.500.

component velocity u' is virtually identical to U(r), suggest- componentat a time, the model is run once for each velocity
ing from (1) that a reasonable estimate of U can be obtained profile. The sample size of 384 representsthe minimum used
from u' = u - /7. for each laboratory measurement. Since LDV measurements
Previous comments with regard to the general insensitivity are made on the basis of particle arrival rates at the measur-
of the tangential velocity profile to a small offset normal to ing volume, this was considereda reasonable simulation of
the LOM are borne out in Figure 11. Tangential velocities V, the data collection process.
v, and v' were so nearly identical that they have been Mean values of the histograms are plotted in Figure 12,
presentedin terms of absoluteerror [v(r) - V(r)] and [v'(r) along with the prescribedfunctionsfor U and V. The most
- V(r)]. Inside the core, for both r and R < 1, [v'(r) - striking feature is the gross underestimate of Vmax (here
V(r)] = 0. From (6) and Figure 9 one sees that z7is small more than 50%) that results from the inherent averaging
everywheresincefor larger, sin(a) -->0 andwhenIsin(a)l inducedby vortex wander. Companionexperimentsindicate
--> 1 as r --->0, U --> 0. Hence •7 is plotted as a velocity that this is somewhatdependenton profile shape and ap-
component in Figure 11. The sign of •7 remains the same pearsto be around 25% for more realistic profiles.
along the entire profile (compareFigure 9) and hencein this In spite of wander-inducedaveraging,high velocitieswere
case increases sensor error for r > 0 and decreases it for r < still observedin the vortex in which the data of Figures 2 and
0. Errors are everywhere below 5% except in the range 1 < 3 were gathered. This flow had an observed maximum
Irl < 2. With the exceptionof an underestimate
of Vrnaxby tangential
velocitycomponent
of 9.49 m s-• at (R, z) =
about 20%, v(r) is a good estimate of V(r). To summarize, u' (0.626, 1.000) cm. The above argumentswould suggestthat
is a good approximationof U, which implies that we must the actual speedwas somewhathigher than this, perhapsas
subtract the contribution /7 from the measured velocity u, highas 11.9m s-• .
and the measured velocity v is a good estimate of V with the
exception of Vmax. 5. SUMMARY AND CONCLUSIONS
In a secondexperiment, the vortex was allowedto "wan-
der" in a random fashion within ranges of offsets dr and I. The juxtapositionof a tornadolikevortex with the axis of a
Theseranges
werechosen
to yieldprofiles
thatresembledTVC affectsprofilesof velocity componentsgatheredin a
thoseof Figures2 and 3. At eachof 17 radii corresponding "chamber" reference frame. A straightforwardcoordinate
roughlyto the locationsof actualmeasurements, histograms transformationto recover actual tangentialand radial profiles
of u and v were constructed by generating sequencesof 384 meets with limited success. For the vortex considered here,
vortex positions.Radial profileswere then plottedusingthe radial velocity component data collected outside of
mean values. Since the LDV measures only one velocity Irl• 5 cmwe'reaffected
littlebythedegreeof offsetand
 LUND AND SNOW 305

TangentialVelocityError
1 = -OSO0

0.20
0.20[ , -0.15
0.151
-O.
lO.•
-0.05..•
0.00" 0.00
-0.05-
• -0.10- --0.05
.•
--0.10
•
-0.15- --0.15

-0.20- --0.20

• , . ' -0.25
-0'25-20 -iO 0 10 •0 3'0 40
DimensionlessRadius,r

v-V(r) .... v'-V(r)........•

Fig. 11. Model tangential velocity componentsfor a stationaryvortex with a mean offset were virtually the same
exceptfor valuesof Vmaxandare presentedin termsof errorfor clarity.Actualvalueswere retainedfor v owingto their
small magnitude. Here I = -0.500.

of the vortex with respect to the TVC centerline, while componentsreveal characteristicboundary layer and vortex
tangential component data were only significantlyaffected flow features. A series of vertical profiles of radial velocity
(albeitdramatically)for 1 cm -< Ir[-< 2 cm. show the rapid development of a near-surfaceinflow jet.
Vertical and radial profilesof radial and tangentialvelocity Vertical profilesof tangentialvelocity suggestinertial oscil-

Vortex Wander Model

(dr= 0.500, [= -0.700)

0.6-
0.4-
0.2-

-0.2:
•E -0.4-
• -o.6-

-1.0_20 -iO 0 1'0 ' 2'0 30 40

DimensionlessRadius, r

* u U • v ............
VI
Fig. 12. Comparison
ofprescribed
velocity
profiles
withmodel-generated
measured
velocities
fora vortex
inrandom
motionwith a specified
degreeof wander.Heredr = 0.500 andl =
306 LASER DOPPLER VELOCIMETER MEASUREMENTS

lations in the core, while radial profiles exhibit the irrota- that LDV measurements can be used to obtain quantitative
tional nature of the outer flow and the solid-bodyrotation of information about critical flow structure such as the near-
surfacejet in the radial componentof velocity in the vortex
the core. Vertical profiles in dimensionlessinner variables
substantiatethe capability of near-wall measurementsevenboundarylayer. Measurementshave been made deeper into
the near-surface core region than ever before. However,
into the linear sublayer.Radial profilesof verticalvelocities
derived through continuity compare qualitatively to mea-vortex wander precludes accurate measurementsof the
surementsby Baker [1981] and Phillips and Khoo [1987]. innermost core. It appears that a separate, simultaneous
Rough calculations at the location of the maximum in measurementof the wander may be necessary.Possibilities
tangentialvelocity(9.49m s-I at (R, z) = (0.626, 1.000) includepressuremeasurementsor video imagery. Parallel
cm) give a centrifugalaccelerationof about 1470# with a mathematicalmodeling of the data acquisitionprocessalso
correspondingpressure gradient, assumingcyclostrophic appearsto be a necessitybothto clearlyunderstandwhat is
balance,of about 170 mbar m-I . The regionof greatest beingmeasuredandto developtechniquesto compensatefor
tangential velocity component(here taken to be speeds wander.
greaterthan7.5m s-•) extended
upwardfromthelocationof
the maximum of 9.49 m s-t. As is shown in Table 1, a
Acknowledgments. A specialthanks to P. J. Smith and W. L.
secondary
maximum
of 8.84m s-• occursat (0.300,6.000) Wood for their support while one of us (J.T.S.) participated in
cm; this maximumis accompaniedby a centrifugalacceler- OperationsDesert Shield/DesertStorm. R. L. Walko provideda
ation of about26559'and a cyclostrophicpressuregradient numberof usefulideasin supportof this work. C. R. Churchand R.
of 310 mbar m-1 . Maxima in centrifugalaccelerationand Rotunno provideduseful remarkson the initial manuscript.This
research was supported by the National Science Foundation
pressuregradientfrom observeddata occur at a heightof through grant ATM 8703846.
3.00 cm where these quantitieshave values of about5260#
and614mbarm-1 respectively. REFERENCES
The greatesttangentialspeedswere found in an annular
Adrian, R. J., Laser velocimetry, in Fluid Mechanics Measure-
volume of small radius and small radial width but significant ments, edited by R. J. Goldstein, chap. 5, pp. 155-244, Hemi-
vertical extent (and coincidentwith the axial jet), with large sphere, Washington, D.C., 1983.
centrifugalaccelerationsand largestpressuregradientsoc- Baker, G. L., Boundary layers in laminar vortex flows, Ph.D.
curring well aloft (up to a height of about 5.6R•.). This dissertation,Purdue Univ., 143 pp., West Lafayette, Ind., 1981.
Church, C. R., J. T. Snow, G. L. Baker, and E. M. Agee,
findingsuggeststhat a single-cell,F1 or F2 intensitytornado Characteristics of tornado-like vortices as a function of swirl
(the type modeledhere) may contain a larger region of ratio: A laboratoryinvestigation,
J. Atmos. Sci., 36, 1755-1776,
potentiallydamaging
wind thanpreviouslythought.(Recall 1979.
that intensitiesand hence wind speedsin suchtornadoesare Davies-Jones,R. P., The dependenceof core radiuson swirl ratio in
inferred from surface structural damage.) Two-cell torna- a tornado simulator, J. Atmos. Sci., 30, 1427-1430, 1973.
Fledlet, B. H., and R. Rotunno,A theory for the maximumwind
doesmay attainF4 or F5 ratingssimplybecausethisregion speedsin tornado-likevortices,J. Atmos. Sci., 43, 2328-2340,
is broader and comes closer to the surface, exposing more 1986.
structuresto highwind speeds.The laboratoryobservations Phillips,W. R. C., andB.C. Khoo,The boundarylayerbeneatha
of Church et al. [1979] also seem to show such structure. Rankine-like vortex, Proc. R. Soc. London, Ser. A, 411, 177-192,
1987.
The occurrence of 34# convective accelerationsat low
Rotunno, R., A study in tornado-like vortex dynamics,J. Atmos.
levels(hereat 2R•.) in the eruptingjet suggests
(in a respect Sci., 36, 140-155, 1979.
after Fiedler and Rotunno [1986]) that some of the destruc- Snow, J. T., and D. E. Lund, A secondgenerationtornadovortex
tive force in tornado vortices of the type modeledhere lies chamberat PurdueUniversity, in Preprints, 15th Conferenceon
within the near-surface axial jet. SevereLocal Storms, pp. 323-326, AmericanMeteorological
Society, Boston, Mass., 1987.
A simplenumericalinvestigation
of the impactof wander Snow,J. T., andD. E. Lund, Inertial motionsin analyticalvortex
on the measured speeds has shown that u' is a good models,J. Atmos. Sci., 46, 3605-3610, 1989.
of U, implyingthat the contribution17must Ward, N. B., The explanationof certainfeaturesof tornadodynam-
approximation
be subtracted from the measured velocity u and that the icsusinga laboratory
model,J. Atmos.Sci.,49, 1194--1204,
1972.
White, F. M., ViscousFluid Flow, 725 pp., McGraw-Hill, New
measured velocity v is a good estimate of V with the York, 1974.
exception of Vmax. Wilson, T., and R. Rotunno,Numericalsimulationof a laminar
In conclusion, the feasibility of LDV measurementsin end-wallvortex andboundarylayer, Phys. Fluids,29, 3993-4005,
tornadolike flows has been demonstrated. We have shown
 Vortex Formation From a Helical Inflow Tornado Vortex Simulator

JAMES G. LADUE

CooperativeInstitutefor MesoscaleMeteorologicalStudies,Normatt,Oklahoma73019

1. INTRODUCTION lation into a vortex. Either a rotating screen or a set of vanes

around the perimeter of the inflow layer provided the source
The University of Oklahoma tornado vortex chamber of vertical circulation. A baffle was placed below the updraft
(TVC) was recently reconfiguredfrom the Roth•sz [1985] fan assembly to distribute the pressure minimum over a
versionwith the goal of improvingthe horizontalhomoge- wider area than the vortex as is the casein the upper regions
neity of the helical inflow. The three tiers of guide vanes of a thunderstorm. These simulator characteristics allowed
usedin the RothfuszTVC were replacedwith a set of flat the Ward-type TVC to produce vortices that exhibited many
boxes stackedon top of each other. Each box was able to be similarities to actual tornadoes [Church et al., 1979].
oriented in any direction allowing for a variety of shear For investigatingthe structure of the tornadolike vortex,
profiles to be created. the classicalWard-type TVC is satisfactory. However, most
Using the new configurationof the TVC inflow layer, the major tornadoes are produced by supercell thunderstorms
objectivesof this paper are to (1) establishthe flow param- which derive their vertical vorticity from the horizontal
eters that produce the most intense tornadolike vortex, (2) helicity in the environment [Lilly, 1986;Davies-Jones, 1984].
report on significantchangesto vortex morphologydue to To simulate tornadoes in such highly helical environ-
changing inflow shear profiles, (3) show similarities and ments, Rotttfi•sz [1985] used the inflow layer design men-
differencesof the strongestvortex morphologyto the vortex tioned previously to produce a positive helical inflow with no
produced by the Rothfusz TVC and to that of observed and net vertical circulation. Any vertical circulation that was
modeled supercell tornadic thunderstorms, and (4) offer created within the TVC was found to come from the tilting of
suggestionsfor an improved helical inflow TVC based on the ambient horizontal vorticity in the flow. Rothfusz was
current observations. successfulin creating a small-scale version of a mesocyclone
The goal of tornado simulation by physical means is to with an embedded tornadolike vortex. Even though Roth-
produce a vortex in a laboratory environment and to do so fi•sz [1985] was able to create a strong vortex connected to
t¾oma flow field with characteristicsof the real atmosphere. the groundwith the new inflow configuration, there were still
Most varieties of vortex simulators generate vorticity in a inherent problemswith the design. The configurationdid not
fluid and then concentrate that vorticity by some means to allow the inflow layer to have horizontally homogeneous
produce a vortex [Davies-Jones, 1976]. The Ward-type helicity, and thus a new configuration was designed.
tornado simulator, known as the tornado vortex chamber,
which is to be discussedhere, more closely represents the
2. METHODOLOGY
formation and stretching of vorticity within the tornadic
region of a thunderstorm [Ward, 1972]. Presently, the configuration of the inflow layer has similar
Based on observations of tornado-producing thunder- dimensionalmeasurementsas that of the originalWard TVC
storms,Ward [1972]produceda TVC with a shallowinflow as shown by a side view schematic in Figure 1 and listed in
layer depth (h) comparedwith the radius(r 0) of the updraft Table 1, although the meanings of some nondimensional
regionsuchthat the ratio h/ro, is of the order of unity. The numbers [Davies-Jones, 1973] are different. For instance,
inflow layer allowed air to flow into the interior regionof the the formulation of the swirl ratio becomes unclear because
TVC with a symmetricalvertical circulationsituatedin such there is no symmetrical vertical circulation being introduced
a way that the updraftfan assemblyconcentrated
the circu- into the TVC. The relative changes in the swirl ratio can be
broughtabout in the current TVC configurationby adding or
The Tornado: Its Structure, Dynamics, Prediction,and Hazards.
subtractingtiers.
GeophysicalMonograph 79
This paperis not subjectto U.S. copyright.Publishedin 1993by the Three types of airflow through each tier are possiblein the
American Geophysical Union. new inflow configuration shown in Figure 2. The first is a
308 TORNADO VORTEX SIMULATOR

Honeycomb
tubes .•

Honeycomb
• Baffle
' illiiii!111
]l,,llll
[[i,,,,11i!!1111111
!111111111i'111111111

Inflow

Baffle•
3 [ co
Interaction
zone

Fig. 1. A side view schematicof the latest University of Oklahomahelical flow TVC. The accompanyingdimensions
of the variables portrayed are listed in Table 1.

full-sectional airflow, where air is allowed in from both ends fans mounted at the "downstream" end of each tier. Each
of each tier. The second configuration is a half-sectional tier has the option of being pivoted about a vertical axis such
inflow where one end of each tier is blocked. The third that vertically veering inflow can be achieved. The angle
occurs with the ability to enhance the inflow with blower between successivetier orientations is defined as the angle
/3. As one final option, the addition of filters at one or both
TABLE 1. Dimensional Measurements for the Current
ends of each tier can induce many permutations of vertical
University of Oklahoma Tornado Vortex Chamber speed shear.
When the tier boxes veer with height, helicity is created as
Variable Value
the vorticity along the top and bottom boundaries between
r0 0.6096 m
R 0.9100 m
Z 0.9100 m
Inflow tier dimensions
For each tier
hi 0.1780 m
w 2.2900 m
Four-tier system
h 0.7100 m
Areaof entry(=4. hi ßw) 1.625m2
Three-tier system
h 0.5300 m
Areaof entry(=3 ßhi ' w) 1.218m2
Areaof updraft,rrr0 1.167m2
Re (=1.4 x 104at inflow) 0.6 m s-i -1
Laminarviscosity,
Q/Reh 1.45x 10-5 m2 s
Aspect ratio a (=h/r o)
Four-tier system 1.16
Three-tier system 0.87

The measurementsfor the four- and three-tier system correspond Fig. 2. A three-dimensionalperspective schematic of the inflow
to the number of tiers usedin the inflow layer. The variable Q stands layer accompanyingthe helical flow TVC at the University of
for the volume flow rate, and Re is the Reynolds number. All other Oklahoma oriented unidirectionally (most experimentation was
variables are measurements which are labeled in Figure 1. done with four
 LADUE 309

the tiers reaches a common edge at the beginningof the Inflow Velocity vs Tier SectionNo,
interaction zone. An example is illustrated by Figure 3:
horizontal
vorticity
}! iscreated
near
theceiling
oftierone 0.8

by velocity3 •, while •r2 is createdby 32 alongthefloorof tier

two. At the edge of the interaction zone (labeled "a" in
Figure 1) the two vorticity vectorsinteract constructivelyto
create •rt in a narrow vertical strip parallel to the mean
0.6
velocity vector betweenthe two tiers. That narrow strip of
vorticity eventually diffusesvertically within the interaction
zone to fill in the layer above and below the tier one/tier two ß
. ,.m.--- tier 1
interface. The width of the interaction zone determines to ..........• ........ tier 2
0.2
what degree the vorticity diffuses into the spacesbetween ........ • ....... tier 3
the tier interfaces. ----O•--- tier 4

The purposeof the experimentationusingthe current TVC

0.0 ......
1...... • ....... • ' "
configurationsis to determine whether a mesocyclonewith a 1 2 3 5 6
concentratedvortex can be produced. Experimentationus-
ing combinations of varying updraft strength, inflow layer left tier section# right
fans, and even mechanical downdrafts is carried out to Fig. 4. Inflow velocity profile acrossthe entrance region of each
determine the configuration that produces strong cyclonic tier correspondingto half-sectional inflow used in the experimenta-
vortices. Once realistic results are obtained, observations of tion. The width of each tier was divided into six sections labeled
across the bottom.
the internal flow are documented and compared with the
flow observedin the Rothfusz TVC. Comparisonsof the flow
structure are also made with observational and numerical
produce multiple vortices within the rotation [Davies-Jones,
studies of tornadic supercelIs. 1973' Church et al., 1979].
Using the TVC shear layer configurationthat producesthe
strongest cyclonic vortex as a standard, additional experi-
3. RESULTS
ments were conductedby employingvariations in the shear
layer profile while keepingthe/3 angle constant.These shear Experimentation has shown that a single-cell, end-wall
tests explored the changes of the vortex character and vortex formed with the current configuration of the inflow
strengthto changesin the inflow layer velocity in individual layer. In each experiment conducted where rotation oc-
layers. Seven experimentsof significancewill be reportedin curred, a counterrotating couplet was observed. Cyclonic
this paper. A seventhexperimentwill be discussedas part of rotation was observed to the right of the anticyclonic rota-
the sensitivity studies with one of the tiers removed to make tion with respectto the inflow layer shear vector in all cases
a three-tier inflow system and thus lowering the aspectratio. where rotation was observed.
It was hoped that lowering the aspectratio (three-tier aspect The major problemdiscoveredin many of the experiments
ratio of 0.87) would increase the swirl ratio enough to has been that at times, the anticyclonic rotation was ob-
servedto be strongerthan expected. Anticyclonicrotation
dominated the cyclonic rotation in all of the TVC inflow
configurationswhen the TVC inflow was facing the labora-
tory room entrance. Even when the TVC inflow layer was
oriented to minimize the anticyclonic bias caused by the
room, there were still some unexpectedly strong anticy-
clonic vortices with some of the shear layer profile experi-
ments.The additional anticyclonic bias was found to be the
fault of the TVC itself when a series of hot film anemometer
measurementswere taken in equidistantintervalsalong the
entrance of each tier to measure the consistency of the
Interface between two fiers
inflow velocities. Tiers two and three had considerably
higherrelativeinflowvelocitiesneartheirleft edgeshownin
Figure4. A visualflow analysisusingsmoketracersseemed
Fig. 3. Schematicof the edge of the interfacebetweentwo tiers
to indicate that the airflow enteringthe inflow regionof the
showingthe generationof helicitybetweentwo tiers. The subscript middle two tiers was blocked and channeled around the
1 (2) refersto the quantityin the lower(upper)tier. The variabler, midsection of each tier. The channelingobserved in the
is thevelocityin themiddleof eachtier, while} is thevorticity middletwo tierswas causedby the partialobstructionby the
generate• alongthesurface of a tier(seetextfor moredetails). cornersof the surroundingadjacenttiers. This effect only
Variable(t corresponds to the total vorticityvectorproducedby
addingthe othertwo vorticitycomponents (the lengthof the vectors occurredwith a veeringshearlayerprofile.Under unidirec-
is not to scale). tional shear flow, no significantnet anticyclonicshear
310 TORNADO VORTEX SIMULATOR

TABLE 2. ExperimentsCompletedas Part of the Vortex SensitivityExperimentsWhose

HodographsAppear in Figures 5 Through 7
Total
Filter No. of Helicity, Figure
Experiment Arrangement Tiers m2 s-2 No. Remarks
1 no filters 4 0.42 5a dominant cyclonic vortex
(strongest case)
2 3, 4 4 O.50 5b dominent cyclonic rotation
(no concentrated vortex)
3 1 4 0.49 6a dominant anticyclonic vortex
4 !, 2, 4 4 0.45 6b dominent anticyclonic rotation
(no concentrated vortex)
5 1, 4 4 0.63 6c dominant anticyclonic vortex
(weaker than experiment 3)
6 2, 3 4 0.58 7a no observed rotation
7 2, 3, 4 4 0.63 7b no observed rotation
8 no filters 3 0.70 5c dominant cyclonic vortex
(wider vortex than experiment 1)

The/3 angle is the same for all experiments.The column under the filter arrangementlists the tiers
filtered, 1 (4) being the lowest (highest) tier.

introduced into the TVC as evidenced by the lack of rotation inflow layer shows no significant trend in determining the
observed in the interior. Even with the anticyclonic bias, strength or type of rotation observed in the TVC.
further experimentation was consideredpossible. Another part to the experiments listed in Table 2 was the
Table 2 provides a description of the shear profile exper- lowering of the aspect ratio of the TVC in the hopes that
iments and the experiment with the lower aspect ratio. The there would be a transition from single cell to multiple
hodographs of the experiments which produced dominant vortex. The same unfiltered, half-sectional inflow that pro-
cyclonic rotation are shown in Figure 5. Dominant anticy- duced the strongest cyclonic vortex (Figure 5a) was also
clonic vortices occur with the hodographsshownin Figure 6, used in the lower aspect ratio test. Lowering the aspect ratio
while the hodographsin Figure 7 occurred with no observed also meant lowering the volume of the inflow area since one
rotation. For each of the experiments, helicity is calculated tier was removed. As a result, inflow velocities increased
for comparative reasons by taking 2 times the area under- along with the observedhelicity (Figure 5b). Two effects on
neath the hodographfor the entire inflow layer. the vortex were noted when the aspect ratio was lowered.
In the set of shear tests completed, concentrated cyclonic First the vortex diameter appeared larger in the low aspect
vortices were produced when the flow in the lowest two tiers ratio case. Second, the vortex exhibited many disruptions,
was left unobstructed. Even with the negative velocity shear turbulent breakdowns (not related to the breakdown bub-
profile from tiers two to three shown in Figure 5b, cyclonic ble), and more rapid cycloidal motions. Davies-Jones [1973]
vortices formed. However, when the inflow velocities of tier observed that increases in vortex diameter occurred with
one relative to tiers two or three were dropped, allowing the increasing swirl ratios. A similar effect may be happening
strongesthorizontal anticyclonic shear to dominatethe flow here. As for the second effect, the sudden increase of the
inside the TVC, a strong anticyclonic vortex was created. inflowshearfrom1.3s- • to 2.2 s-• whileholdingtheupdraft
Lowering the negative shear in the TVC to the lower half of strengthconstant may have led to a more turbulent inflow
the shear layer (Figure 7) resultedin no rotation of the flow. leading to a less organized vortex. No sign of multiple
In summary, a slowly changing profile of velocity and vortices was observed.
horizontal vorticity throughout the inflow layer (Figures 5a The aspect ratio threshold needed for vortex breakdown
and 5b) produces the strongestcyclonic vortex. However, may need to be much smaller in a helical TVC than in the
strong velocity and vorticity in the lowest two tiers is more standard Ward TVC. This is an expected departure from
important than in any other set of tiers in determiningthe axisymmetric simulators where radially consistentangular
flow morphologyin the TVC. Lowering the inflow velocity momentum produces circulations occupyingthe entire area
in the lowest two tiers as shown in Figure 6 enhances the of the updraft core. A helical flow TVC, where no net
strengthof the anticyclonic vortex since the available hori- vertical angular momentum is introduced into the inflow,
zontal helicity in the lower half of the inflow layer fails to impliesthat there are equal amountsof positiveand negative
match the horizontal anticyclonic shear in the middle of the angularmomentumflux. As a result, in the present experi-
inflow layer. Lowering the positive horizontal vorticity in mentation, the radius of the circulation is only 1/3 that of r 0,
the lowest two tiers as shown in Figure 7 results in no which means the aspect ratio of the disturbanceis 3 times
organized rotation. Note that the helicity for the entire higher than the TVC aspect
 LADUE 311

180 180

0.42 m 2 s -2
h = 0.49 m 2s-2

90 270 9O 270

180
b) 3
0.50 rn 2 s -2 0.45 m 2 s -2

90 270 90

1 2 2
c) c)

0.70m2s-2 •'= 0.63

m2s-2

/ /
Fig. 5. Hodographscorresponding to (a) half-sectional unfiltered Fig. 6. Similarto Figure 5 except(a) singlefilter on tier l,
inflow, (b) half-sectionalinflow singleair filterson tiers3 and 4, and singlefilterson tiers 1, 2, and 4, and (c) singlefilterson tiers 1 and
(c) half-sectional unfiltered inflow with three tiers. The values h 4.
correspondto the helicity over the inflow layer depth. Labels upon
the hodographcorrespondto the tier number. The directionallabels
on the major axis of the hodographare merely usedas a convenient
frame
-1
of reference.The maximumradiusof eachhodograph
is 1.0m Figure8b showsthe cyclonicvortexto the left and a deep
S .
layer of smoke on the right associatedwith the lifted
anticyclonic flow (labeled A).
Significant differences in the internal flow field between
With both the high and low aspect ratio flow fields within the present version and the Rothfusz version of the TVC also
the TVC, there are many similaritiesbetweenthe present exist. Since the most productive cyclonic vortices come
TVC inflow versionand that of RothJ}tsz[ 1985]andRothjS•sz from half-sectional inflow (Rothfusz used a full-sectional
and Lilly [1989]. The flow on the right side of the vortex as inflow),thereis a singleregionof anticyclonicflow (Figure
seenin Figures 8a and 9a reacts in a similar way to the flow 9a), whereasthe Rothfusz versionhas two regions,one on
on the right side of the Rothfusz vortex by becomingdrawn eachsideof the cyclonicvortex (Figure9b). The sourceof
into the vortex at groundlevel. On the left sideof the vortex, the surfaceconvergenceline in the present version is the
the flow turns anticyclonically. However, as in the Rothfusz highstagnationpressureas the flow meetsthe oppositewall
case, the surfaceair flow is forcefully lifted by a convergence while the opposingairstreamsin the Rothfusz versionforce
boundary formed by the opposingflow and never becomes the surfaceconvergence.One uniquefeatureof the present
entrained into the cyclonic circulation. The photographin TVC is the existenceof a closedanticycloniccirculation
312 TORNADO VORTEX SIMULATOR

1
180 vortex. Several similarities are apparent upon inspection
(the surface inflow vector is used as a frame of reference
0.58 rn2 s-2 unlessotherwise noted), including (1) the presence of a rear
flank gust front left of the surface cyclonic vortex labeled
RF, (2) a region of positive (negative) angular momentum to
the right (left) of the vortex where only the flow on the fight
side is entrained into the vortex, (3) anticyclonic curvature
left of the vortex, maximized on the left side of the rear flank
27O gust front (labeled A), and (4) kidney-shapedupdraft con-
forming to the shape of the rear flank gust front.
The presence of the anticyclonic circulation in Figure 8a
would seem to be a manifestation of the anticyclonic bias
1
discussed earlier. However, there were observations that
180 indicatedmore than the anticyclonicbias was responsiblefor
b) this circulation. First, the experimentation showed that no
•= 0.63 m2s -2 rotation was observed in the unidirectional flow control test.
Such evidence suggeststhat under directional shear condi-
tions, an underlying region of anticyclonic flow would be
necessary in order for the anticyclonic bias to produce an
observable circulation center. Second, the cyclonic and
anticyclonic circulation centersalways remained in the same
9O relative positions to each other. This fact held true even
_

when the anticyclonic was strongerthan the cyclonic vortex.

The anticyclonic rotation always remained left of the cy-
Fig. 7. Similar to Figure 5 except (a) single filters on tiers 2 and 3 clonic vortex. Then, as mentioned before, anticyclonic flow
and (b) single filters on tiers 2, 3, and 4. was observed by dual-Doppler radar (labeled A in Figure
10b) by Brandes [1984] in this same region.
Both visual observations (right side of the photograph in
of the cyclonic vortex with respect to the tier one inflow Figure 8b) inside the TVC and inspection of Figure 10b
vector. This anticyclonic circulation may be enhancedby the indicate that this anticyclonic flow is located on the edge of
anticyclonic bias of the TVC. Another feature unique to the a region of positive vertical velocity. This is the region
present TVC is the evidence of a downdraft on the rear side relative to the cyclonic tornado where anticyclonic torna-
of the cyclonic vortex revealed by the observation of smoke does have also been observed to form by Brown and Knupp
tracers. The downward moving air is observed to reach the [1980]. It is possible that the anticyclonic rotation observed
floor to the left of the vortex and then become part of the is due to the stretching of the anticyclonic vorticity by a
opposing flow which forms the convergence boundary. strong gradient of positive vertical velocity found along the
There are no such observations of downward air motions tail end of the rear flank gust front. The anticyclonic bias
near the vortex in the Rothfusz version of the TVC [Roth- inherent in the present TVC acts to enhancethe strength of
fusz, 1985]. the circulation.
In the relatively steady state of the vortex (well after
vortex initiation), the source of the vertical vorticity at the 4. DISCUSSION
lowest levels also appears similar to the scenarioproposed
by Rothfusz and Lilly [1989]. The nearly linear flow with A number of performance tests were made upon a newly
positive angular momentum is suppressedand allowed to reconfiguredhelical flow TVC consistingof three and four
feed rotation to the vortex while the negative angular mo- independentlymovableinflow boxes.This reconfigurationof
mentum is forcefully lifted out and away from the vortex by the inflow layer was considered to be an improvement over
the strong convergence line. Vertical vorticity is also sup- the circular guide vane inflow of the Rothfusz TVC version
plied by the tilting of boundarylayer vortex lines contained in that no negative helical flow regionswould be introduced
in the flow reaching the vortex. into the TVC from the inflow layer.
Now the question remains whether the simulatedvortex The results were rather mixed because of the inadvertent
flow structure is analogous to established observations and horizontal anticyclonic shear that was introduced into the
numerical simulations of the flow morphology in a typical TVC inflow box construction and from the laboratory room.
tornadic supercell. A numerical simulationschematicfrom Even with the anticyclonic bias, meaningful results have
Klemp and Rotunno [1982] and a dual-Doppler analysisby been achieved with the present TVC. A few important
Brandes [1984] of the low-level flow around a tornadic results are summarized below:
supercellare shown in Figures 10a and 10b, respectively, 1. Strong vorticesdevelop only when there exists a con-
for a basisof comparisonof the observed surface-simulated vergencelayer as deep as the inflow layer. The
 LADUE 313

TIER 1
INFLOW

UPDRAFT
CORE

INFLOW

• UPDRAFT
PERIMETER
CENTERLINE
CORE

Photograph

Fig. 8. (a) Surface streamline analysis of the flow structure correspondingto the hodograph in Figure 5a. The thick
line corresponds to the convergenceboundary and the small stippled circles represent the average location of the
rotation centers. The label RF marks the rear flank boundary, while A marks the anticyclonic rotation. lb)
Accompanyingphotographtaken from the TVC exterior at a locationlabeled "photograph" in Figure 8a. The end-wall
vortex can be seen to the left. The label A marks the region over the anticyclonic rotation shown in Figure 8a.

inflow layer allowed the opposingwall to the inflow layer to vortex morphology more than any other layer. Additionally,
create a stagnationregion and a deep convergenceboundary. a slowly varying profile of velocity and vorticity produced
2. The shear profile experiments suggestedthat the ve- the strongest cyclonic vortex, not just high helicity.
locity and vorticity of the lowest two tiers influenced the 3. The TVC vortex morphology showed many
314 TORNADO VORTEX SIMULATOR

Updraft

TVC floor

Updr•
b)

':

TVC floor

Fig. 9. (a) A three-dimensionalperspective schematic of the current flow morphology correspondingto the
hodographin Figure 5a. The thickest line indicatesthe primary vortex location, while the line bridgingthe two vortex
centersmarkedby the two stippledcirclesindicatesa convergenceboundary.The dotted arrowsindicatestrongupdraft
regionsoutside vortex centers. (b) Similar to Figure 9a except valid for the Rothfusz full-sectionalinflow case.

ties to observational and numerical studies of classical Second, the width of the interaction zone labeled in Figure 1
tornadic supercelIsincluding a rear flank gust front and the will have to be increased such that the diffusion of intertier
presence of anticyclonic curvature in a region known to boundary layer vorticity results in more complete smoothing
produce real anticyclonic tornadoes. of the inflow layer vorticity profile at the edge of the updraft
4. The dynamics maintainingthe vortex are similar to the hole. Observations now indicate that there are regions of
dynamics of the Rothfusz TVC. Flow on the right side of the high and low horizontal vorticity as the flow passesinside of
vortex containing positive angular momentum was allowed the interaction zone.
to reach the vortex while the negative angular momentum A future helical flow TVC will also need the capability of
flow was removed vertically by the rear flank gustfront. The having independently controllable inflow and outflow flow
flow reaching the vortex also allows a continuoussupply of velocities without changing the aspect ratio or the use of
boundary layer vorticity to be tilted up into the updraft, as is extra fan assemblies in the inflow layer. Increasing the
the case with axisymmetric Ward-type simulators. updraft in the present inflow layer configuration leads to
The capability of the present TVC to produce flow struc- increasing the shear layer velocity when the aspect ratio is
ture similar to that of a tornadic supercell indicates that held constant. Similarly, decreasingthe aspectratio reduces
further investigations will be useful; however, the inflow the inflow volume and thereby increases the inflow velocity
layer of a future helical flow TVC will have to be redesigned given the same updraft velocity. The simplest way of keep-
to improve the homogeneity of the vertical shear profile. ing the updraft velocity, the aspect ratio, and the inflow
First, the entrance to each inflow box must be unobstructed, velocity independentfrom each other will be to have inflow
either from the exterior room geometry or from adjacent layer tiers with adjustable widths (therefore an adjustable
boxes. Currently, the corners of the tiers obstructthe inflow inflow volume). In such a configuration,the aspect ratio will
to the middle of adjacent tiers creatingthe anticyclonicbias. be able to be lowered without a correspondingchangein
 LADUE 315

HEIGHT !.3 KH 20HPS --

Fig. 10. (a) A schematic of the surface flow in a tornadic supercell from Klemp and Rotunno [1982] based on
numerical simulations. The solid (dashed) lines represent positive {negative) vertical velocity. Flow artoxusrepresent
storm relative streamlines,and the shadedregion representsthe surface rainwater field. A vertical vorticity maximum
is labeled with a small T. (b) Low-level wind field of a tornadic thunderstorm near Del City, Oklahoma, from a
dual-Doppleranalysis.Left panel showsstormrelative horizontalwind with contours of radar reflectivity (in dBz).
Right panel showsvertical velocity {in metersper second).The heavy dashedline representsthe mesocyclone[from
Brandes, 1984]. The labels RF and A mark similar regionsdescribedin Figure
316 TORNADO VORTEX SIMULATOR

ratio of the inflow layer to the updraft velocity. That means Characteristics of tornado-like vortices as a function of swirl
the inflow layer helicity could be held constant in an exper- ratio: A laboratory investigation, J. Atmos. Sci., 36, 1755-1776,
1979.
iment of changingupdraft velocity or aspect ratio. Such an
Davies-Jones,R. P., The dependenceof core radius on swirl ratio in
increasein flow control could help determine if a helical flow a tornado simulator, J. Atmos. Sci., 30, 1427-1430, 1973.
TVC can produce multiple vortices. Davies-Jones, R. P., Laboratory simulation of tornadoes, in Pro-
ceedingsof the Symposiumon Tornadoes:Assessmentof Knowl-
edgeand Implicationsfor Man, pp. 151-174, American Meteoro-
Acknowledgments. Throughout this work, I greatly appreciate logical Society, Boston, Mass., 1976.
the patienceand help of my immediateadviser, EugeneWilkins, for Davies-Jones,R. P., Streamwise vorticity: The origin of rotation in
his insightand support.Lans Rothfuszhelpedme in gettingfamiliar supercellstorms,J. Atmos. Sci., 41, 2991-3006, 1984.
with the laboratory equipment, and Meta Sienkiewicz helped with Klemp, J. B., andR. Rotunno,A studyof the tornadicregionwithin
the myriad of computer-relatedproblems.I am also gratefulto Sue a supercellthunderstorm,J. Atmos. Sci., 40, 359-377, 1982.
Weygandt for helping with the figures.This work was conducted Lilly, D. K., On the structure, energeticsand propagationof
under National Science Foundation grant ATM 8914434. rotating storms, II, Helicity and storm stabilization, d. Atmos.
Sci., 43, 113-125, 1986.
REFERENCES Rothfusz, L. P., A mesocycloneand tornado-likevortex generated
by the tilting of horizontal vorticity: A laboratory simulation,
Brandes, E. A., Vertical vorticity generation and mesocyclone M.S. thesis, 106pp., Sch. of Meteorol., Univ. of Okla., Norman,
sustenance in tornadic thunderstorms: The observational evi- 1985.
dence, Mon. Weather Rev., 112, 2253-2269, 1984. Rothfusz, L. P., and D. K. Lilly, Quantitative and theoretical
Brown, J. M., and K. R. Knupp, The Iowa cyc!onic-anticyclonic analysisof an experimentalhelical vortex, J. Atmos. Sci., 46,
tornado and its parent thunderstorm,Mon. WeatherRev., 108, 2265-2279, 1989.
1626--1646, 1980. Ward, N. B., The explorationof certainfeaturesof tornado dynam-
Church, C. R., J. T. Snow, G. L. Baker, and E. M. Agee, ics usinga laboratorymodel, J. Atmos. Sci., 29, 1194-1204,
 Discussion

GENE WILKINS, SESSIONCHAIR

UniversiO,of Oklahoma

PAPER E 1 vorticity on the inner side, at equal rates. So the circulation

Presenter, Chris Church, University of Miami, Ohio [Church around the whole set of vanes is zero, as it ought to be
and Snow, this volume, Laboratory models of tornadoes] becausethere is negative circulation around each vane and
positivecirculationat the center. But the circulationaround
(Brian Fiedler, University of Oklahoma) When Rich [Ro- a circuit inside the vanes is positive, of course; otherwise,
tunno] and I estimated that the maximum vertical velocity there wouldn't be a vortex.
would be roughly ten times the average external [vertical]
velocity [J. Atmos. Sci., 43, p. 2337, 1986], we did not intend (Church) In the Rothfusz experimentsthat I was referringto,
that to be applied to thunderstormsbecauseof the different the vanesare setup in a very differentway. The inflow layer
outflow conditions; that really only applies to the Purdue was divided into three layers. In the lowest layer, the vanes
simulator. We think that ratio is too high in the simulator. In are all parallel, e.g., oriented in the north-southdirection.
my paper [A4], that ratio was 2, not 10. Could you comment The middle and highestlayers are similar, but the direction
on your new simulator, where there is divergence aloft of the vanes veers with height between the layers, e.g.,
insteadof the fan suckingthe air out vertically? What is the NE-SW in the middle layer and E-W in the top layer. Each
ratio of maximum vertical velocity to mean updraft velocity layer producesno circulation.Nevertheless,vorticesform.
in that simulator? Is it the same as in the Purdue chamber? Basically, it is helical flow, but I don't know why concen-
trated vortices are produced.
(Church) I haven't measured it.

(Fiedler) Do you have the same•p* [nondimensional

central PAPER E2
pressuredeficits]as in the Purduechamber?
Presenter, Don Lund, Purdue University [Lund and Snow,
(Church)They are very similar. But you must appreciate
this volume, Laser Doppler velocimeter measurementsin
that it takes a long time to make these measurements,
and
tornadolike vortices]
one does not routinely compare one's own chamberwith
other ones by repeatingexperiments.So I don't have very (Brian Fiedler, University of Oklahoma)The questionkeeps
many substantivemeasurements for comparison. comingup of whetheror not the ratio of maximumtornado
wind speedto thermodynamicspeedlimit still applieswhen
(Bruce Morton, Monash University) I couldn't really pass
the lower boundarylayer is turbulent. Could an experiment
over that challengeto me. It is quitesimple.The setof vanes
beperformed,usinga roughlower surface,to seeif axialjets
are oriented so that the vanes are at a constant angle to the
in the core still occur with a turbulent boundary layer?
radials. When the flow is turned on, each vane sheds a
startingvortex, which has positivevorticity, and there is (Lurid) Snow did some work with Wayne Rostek [see
negativecirculationaroundthe vane. Thereafter,the nega- Preprints, 14th Conf. Severe Local Storms, 1985]using
tive circulation continues around the vane and the contribu- roughness elementsin the chamber.Are you familiarwith
tion to positivecirculationcontinues in the center.Positive this work?
vorticityis generated on theoutersideof thevane,negative
(Fiedler) Did he find axial jets in the core?
The Tornado:Its Structure,Dynamics,Prediction,andHazards.
GeophysicalMonograph79
(Lund) Unfortunately,I am not familiarwith it. LDV [laser
Copyright1993by the AmericanGeophysical
Union. Dopplervelocimeter] work with roughness elementsin the
318 DISCUSSION

chamberwould be problematicbecauseof havingto see (Chair) Bob Walko, by the way, is responsiblefor much of
through the roughnesselements. the constructionof this apparatusduring the time he was at
(RichRotunno,NationalCenterfor Atmospheric Research) the University of Oklahoma.
Aren't the vortexwanderandthe helicoidalwavesa disap- (Rich Rotunno,National Centerfor AtmosphericResearch)
pointment? I thoughtthat the motivationfor the new appa- I have a comment. It seems to me that it is more difficult to
ratus was to get a vortex that was stable and vertical. Are the
explain what's goingon in that experimentthan to explain
vortex wander and the helicoidal wave on the core different
the tornado.
or the same phenomena?
(LaDue) That's true. There are too many boundaries and
(Lund)I wouldlike to seeif the apparenthelicoidalstructure
is repeatable.Regardingthe wander, the data in Figure 2, walls. But there is some significanceto our experiments
whichare averagesof lots of samples,look very symmetric because there was no organized rotation when all the inflow
aboutthe origin. So there is still a questionwhetherwander was from one direction or when the inflow speedsdecreased
is really presentor whether there is a steady stateoffset in markedly with height. Organized rotation occurred only for
the vortex position[from the axisof the apparatus].I haven't favorable shears.
determined that yet.
(Chair) Jim has tried different speed and directional shears,
(Rotunno) If the bottom part of the vortex wanderedand the and has drawn hodographs,so that the simulatorresults may
top part didn't, there would be a helical structure. So the be interpreted in terms of hodograph shape. He has much
wander and the helical structure would be connected. more information.

PAPER E3
(Bruce Morton, Monash University) I don't yet fully under-
stand what is happening. Is the mean inflow velocity the
Presenter, Jim LaDue, University of Oklahoma [LaDue, this same or different in the successivelayers?
volume, Vortex formation from a helical inflow tornado
vortex simulator] (LaDue) It dependson what experiment we're doing, but...

(Bob Walko, Colorado State University) In your successful (Morton) But you should always measure that so that you
experiment, for each tier of ducts,did you let air flow in from know what is happening. You can try to explain the vortex
opposite sides of the apparatus, or did you let air in on one without worrying about all the boundaries until later. But I
side and draw air out on the oppositeside? believe that it is essentialto know how the inflow profile
varies from layer to layer.
(LaDue) No, we didn't draw air out. We createda stagnation
pressure on the opposite side from the inflow, so that we (LaDue) That's what the hodographstry to show, but they
produced a convergencelayer that was as deep as the inflow are not totally accurate. The velocity varies from left to right
layer and that was steady underneaththe updraft. in each tier, and sometimes this leads to vorticity of the
(Walko) What was the condition at the opposite side from wrong sign or unwanted vorticity. The tiers have to be
the side where most of the flow was coming in? constructed very carefully to ensure that the velocity is more
or less uniform in each tier.
(LaDue) At each tier level, flow would come in from one
side, the oppositesidewas blocked.There was no air passing (Chair) Jim has tried many different vertical profiles of
in or out from the oppositeside, the air exited upwards. velocity. He doesn't have time to present all of
 A Review of Tornado Observations

HOWARD B. BLUESTEIN

Schoolof Meteorology,Universityof Oklahoma,Norman, Oklahotna73019

JOSEPH H. GOLDEN •

NationalOceanicandAtmospheric
Administration,
Officeof theChiefScientist,Washington,
D.C. 20235

1. INTRODUCTION 2. VISUAL OBSERVATIONS

Duringthe past15yearsfollowingTornadoSymposiumII, Many of the early visual eyewitness accounts of tornadoes

our understandingof the basic morphologyof tornadoesand were well summarized by Flora [1953]. In fact, the difficulty
their parentstormshasadvancedsignificantly.The fragmen- of Flora's accounts is that some of them were obviously
tary observations obtained by the National Severe Storms made under great psychological stress by observers and
Laboratory (NSSL)-organized Tornado Intercept Project must, therefore, be viewed with skepticism. One is struck by
[Golden and Morgan, 1972; Davies-Jones, 1988] have the vivid account of one eyewitness who survived the
evolved to a three-dimensional, time-variant conceptual overhead passageof a large tornado and was able to peer
upward into the heart of the funnel cloud from his basement
model. There has been a synthesisof more sophisticated
as the remainder of the house was disintegrated and blown
(i.e., well trained) eyewitness and storm-chaser observa-
away. This observer, and several others in similar encoun-
tions, including mobile soundingsand various new portable
ters, reported seeing intense electrical discharges inside the
instrumentation with fixed Doppler radar data.
tunnel walls. These accounts, while fascinating, have not
Our ideas concerning tornadogenesis mechanismshave
been replicated by trained storm chase teams over the past
tended to focus on the wealth of observations taken over the
two decades; moreover, they fed controversy regarding the
southern Great Plains but have also evolved with increasing possible role of magnetic anomalies and/or lightning in
field experiments in diverse locations, such as Alabama, tornadogenesisand waterspouts which persists to this day
Florida, Colorado, and the U.S. West Coast. Indeed, there [e.g., Davies-Jones and Golden, 1975a, b, c; Colgate,
has been a reexamination of how we should define a tornado 1975; Rossow, 1969, 1970; Vonnegut, 1960]. These eyewit-
[Forbes and Wakimoto, 1983] in relation to the primary ness credibility problems are also addressed by Grazulis
source of vorticity in the parent storm system. Mechanisms [this volume] in assessingthe 100-year climatological record
involving features other than the mesocyclone(visualizedby of tornadoes in this country. Nevertheless, technological
the presence of a "wall cloud" or other rotating cloud base) advances have occurred since Tornado Symposium II that
appear to be operative in certain situations,e.g., alonggust have permitted more comprehensive studies of intracloud
fronts and other types of convergencezones or boundaries. and cloud-to-ground lightning in tornadic storms [MacGor-
We shall attempt here to summarizethese new findings man, this volume; MacGorman and Nielsen, 199I]. Argu-
regardingtornado observationsbut shall excludeclimatol- ably, the single most comprehensive set of eyewitness
ogy and fixed Doppler radar observationsof storm-scale photographsand accounts were those assembledand diag-
features, which are addressed elsewhere in this volume. nosedin a study by Fujita [ 1960] on the June 20, 1957, Fargo,
North Dakota, tornadoes.

•Now at NOAA, Oceanicand AtmosphericResearch,Silver 2.1. NSSL Tornado Intercept Project

Spring, Maryland.
The advent of the NSSL Tornado Intercept Project in the
The Tornado: Its Structure,Dynamics, Prediction,and Hazards.
early 1970s[Golden and Morgan, 1972] led to confirmation
Geophysical Monograph 79
of Fujita's earlier Fargo model in a number of cases. One of
Copyright
1993bytheAmerican
Geophysical
Union.
the best photographedtornadic storms in the early days of
320 REVIEW OF TORNADO OBSERVATIONS

Fig. 1. Same rotating storm as Plate 1, looking NW-NNE from light aircraft. Note large tilt of be!l-shaped cloud
toward the right and tornado pendant from flanking line merging, with relatively shallow cloud tops, into main updraft.
(Courtesy of D. Ray Booker.)

the Intercept Project occurred 1 hour before sunset on April 2.2. Tornadoes and Rough Terrain
30, 1972, in extreme western Oklahoma. The small tornado
Plates 2a, 2b, and 2c show typical examples of tornadoes
that it produced lasted only a few minutes (Plate 1) and
observed in very high mountainous terrain, elevated moun-
remained over open farm country [Golden and Morgan,
tain valleys, and at the foot of Pikes Peak. The latter
1972]. However, several weeks later it was discovered that
produced FI/F2 damage in Manitou Springs, Colorado, even
the same visually rotating, bell-shaped cumulonimbus was
though the visible funnel never extended more than halfway
photographed by two aircraft flying at different altitudes (just
to the ground. This is a typical feature of tornadoes in
below cloud base (Figure 1) and by a U-2 flying at 18 km
mountainous or elevated terrain, and we believe that it is
(Figure 2)). Unfortunately, NSSL Doppler data were not
primarily due to somewhat drier environments in the low
available that day, and the storm would have been beyond
levels [Szoke et al., 1984].
the effective range anyway. Conventional WSR-57 radar
reflectivities depicted a very large, intense (radar reflectivity While destructive tornadoes appear to be still rare in
in excess of 60 dBz) storm, with a possible hook echo at the mountainous terrain, Fujita [1989] documented a very large
time and location of the tornado. and > 100-km path length tornado that devastated a heavily
The Union City, Oklahoma, tornado of May 24, 1973, has forested area at over 3 km above mean sea level (MSL) from
NE Utah and climbed the Continental Divide into the Grand
to date provided the best overall data set for synthesizing
Doppler radar observations with the evolving three- Teton National Park, Wyoming, before weakening. Fujita
dimensional visual storm and tornado structures. The storm [1974] found that a few of the tornadoes in the 1974 Jumbo
morphology and strategy used to intercept it on the ground Outbreak developed in valleys and climbed substantial
are given by Golden [1976]. The Union City case led to the mountain ridges of up to 0.5-km elevation, maintaining
development of a life cycle hypothesis for tornadoes and is continuous destructive paths on both sides of the mountain.
illustrated in Figure 3, after Golden and Purcell [1978a]. There continues to be some controversy regarding the
Finally, Golden and Purcell [1978b] presented photogram- possibility of strong tornadoes affecting urban areas. Des-
metric wind speed estimates derived from high-quality films sens [1972] was the first to suggestthat tornado formation
of the Union City tornado. These results clearly indicated could be affected by surface roughness, and he based this
that even though in later stagesof its life cycle the tornado inference on laboratory simulations. Blechman [1975] stud-
shrinks,it retainsdestructive
windspeeds(upto 65 m s-•). ied the formation of a multivortex tornado and found
 BLUESTEIN AND GOLDEN 321

Fig. 2. Photo of same rotating storm as Plate 1 and Figure I taken from overflying U-2 jet about 1 hour earlier from
18-km altitude. Note overshooting cloud tops, striated midlevel cloud bands, and flanking line entering main
cumulonimbus from center left. View NE-E. (Courtesy of J. T. Lee.)

correlations of suction vortex formation with increasing There may also be a "tail cloud" and a "feeder band"
surface roughness. Later, Elsore and Meaden [1982] used spiraling into the wall cloud from the northeast. Doppler
long-term storm records over the greater London area to analysespresentedelsewhere in this volume indicate that the
infer that the urban "heat island effect," coupled with "wall cloud" is usually the cloud base reflection of an
greater surface roughness, accounted for the minimum of organizing mesocyclone aloft within the supercell storm.
"weak" tornadoes in the climatic record over London.
Rarely, the mesocyclone become• so intense that the wall
Nevertheless, significant tornadoes have been documented cloud itself lowers to the ground, although this was the case
in the Lubbock, Texas, Kalamazoo, Michigan, and Denver, in the Binger, Oklahoma, tornado studied by Zrnic et al.
Colorado, urban areas. (See Szoke et al. [1984] and Szoke [•985].
and Rotunno [this volume] for the Denver area cases.)
Variants of the conceptual model in Plate 4 exist for other
partsof the country but have not yet been documentedin the
2.3. A Conceptual Model Jbr Great Plains literature. In fact, there is still •ome lively debate about the
Supercell Tornadoes conjecturedpreeminent role of supercelIsin tornadogenesis
We summarize in Plate 3 the overall structure of a torna- (see Bltrgessand Lemon [1990] and Golden et al. [1990] for
do-spawningsupercellover the southernGreat Plains, as detailed arguments). More recently, there have been low-
seenby an observerpositionedto the southeastof the storm. precipitation (LP) and high-precipitation (HP) supercell
This is meant to be a conceptualmodel and will not apply to storms documented by chase teams (see, for example,
most tornadic storms in the High Plains of Colorado and Bluestein and Woodall [1990]); generally only the latter are
Wyoming and in the southeastU.S. Note that many of the important tornado producers, although several casesof LP
cloud features first documented by Fujita in the Fargo storms with tornadoes have been documented by cha•e
tornadoappearin this model, suchas the rotating,precipi- teams recently over eastern Colorado, Wyoming, and Texas
tation-free base of the wall cloud, which may be associated (E. W. McCaul and D. Blanchard, personal communication,
with one or morependanttornadofunnelsduringits lifetime. 1990). On the other hand, tornadoes in the southeast U.S.
In addition, slnaller,short-livedtornadoesmay form beneath often occur late at night and are frequently ensheathedby
the flankingline to the left of the wall cloud in Plate 4. heavy rain curtains (e.g., Huntsville, Alabama
322 REVIEW OF TORNADO OBSERVATIONS

UNION CITY, OKLAHOMA

24 MAY 1973 TORNADO

US
1540 1550
A B C
•5•8
OrganizingStage

$H/•

1600

Duration 26 min
Path Length 17 km
Max. Width 0.Skm

FILMING
SITE 1604 CST I ....
o 0.5 I.O km

$H

! I ,, ,I
I 2. 3 km US

Fig. 3. Synthesis of the May 24, 1973, tornado damage track through Union City, Oklahoma, with insets depicting
tornado morphologicalchangeswith life cycle stagesalong track {,After Golden and Purcell [ 1978b].)

and Knupp, 1990;McCaul, 1987]). Wall clouds are occasion- that waterspoutsand their land-based cousins, tornadoes,
ally observed with these tornadoes but may exhibit little or are electrically driven. However, he found from several sets
no rotation in early stages. of measurementsthat "atmospheric vortices over water can
exist without an electric field or current of appreciable
3. WATERSPOUTS AND THEIR LINK TO TORNADOES magnitude." Lightning strokes were indicated by the instru-
mentswhile near waterspoutson four occasions,but in only
Our understandingof waterspout structure, wind speeds,
one instance could the strokes have been in the parent cloud
and behavior prior to the late 1960swas based largely upon
mariner's tales of chance encounters at sea [e.g., Hurd,
containing the waterspout.
1950; Gordon, 1951]. That waterspouts pose a threat to
Golden [1974a, b] followed up on Rossow's pioneering
structuressituatednear their pathsis beyond doubt [Golden, studiesand was able to documenta repetitive waterspout life
1973; Fujita et al., 1972; Macky, 1953]. Many of the damag- cycle, from close-rangeaircraft observations of over 100
events in the 1969 summer season over the lower Florida
ing tornadoes affecting the central and eastern Gulf Coast
during the late fall and early spring originate over the Keys. Church et al. [1973] developed a trailing-wire probe
northern Gulf of Mexico as intense waterspouts. Some of which they towed through Florida Keys waterspoutsfrom a
these do not have hook echoes or other classic, distinguish- light aircraft. However, the complicatedinteractionsbetween
ing characteristicson radar. Moreover, many of the torna- the probe itselfand the flow field in and aroundthe waterspout
does along the U.S. West Coast originate offshoreas water- funnel made reduction of the pressure and temperature data
spoutsin the presenceof an intenseupper cold low. Golden extremelydifficult.The double-peakedtemperatureanomalies
[1973] noted that these events tend to be small and short aroundthe waterspoutcore appearto be systematicallywarm,
lived, with minor damage. as depictedin Figure4, a three-dimensional conceptualmodel
based on 3 years of aircraft observationsand tracer experi-
3.1. The Waterspout Life Cycle ments [Golden, 1971, 1974a].
During the late 1960s,Rossow [1970] usedan instrumented During the latter part of the 1970s, further advances were
Navy Grumman S2E aircraft to study waterspout develop- made on our knowledge of the internal flow and thermal
ments in the Florida Keys. He wanted to test the hypothesis featuresof waterspouts.Leversonet al. [1977] usedan
 BLUESTEIN AND GOLDEN 323

acrobaticaircraftequippedwith a gustprobeto penetrate

waterspoutsjust belowcloudbase.Their dataandfindings
(basedon a limited sampleof small-to medium-sizedfunnels
in the FloridaKeys) showthat in the waterspout
funnelthe
following features exist:
"COLLAR CLOUD"
Waterspout Vortex Structure Near Cloud Base

1. A coreof risingmotionof --•5-10m s-1

CLOSE
2. A horizontal circular flow field (helical upward) that is
SHELL
RANGE broaderin weaker waterspoutsandspatiallyconcentratedin
!LENT WALL
SECTION VERT. EDDIES more intensewaterspouts.
VIEW 3. A central core region of--•0.3øK warmer than ambient
temperatures.
4. A core pressuredeficit of the orderof---1-10 mbar, depend-
ing on waterspout intensity.
TENTATIVE
TEMPERATURE
ANOMALY One example of the data Leverson et al. [1977] obtained on
A.T,,O tangentialand vertical wind components,temperature,and
COMPOSITE
pressureanomaliesis given in Figure 5. Note that all of the
MODEL cases sampled by the AT-6 were relatively small water-
Of
&•ShOWER
MATURE spouts.No data were obtained above parent cloud base or
STAGE
within the waterspout's spray vortex. An important new
findingwas the asymmetric structure of the waterspout's
circulationaroundthe funnel (notedin the sprayvortex of an
intenseanticyclonicwaterspout by Golden [!974a]) and its
large lateral radius of influence. Large upward vertical
velocitiespeak just outside the visible funnel, with likely
weak descent in the core.
Schwiesow [1981] and Schwiesow et al. [1981] used an
.TRAIN•/mz airborneinfraredDoppler lidar to probewaterspoutfeatures
Fig. 4. Compositedconceptualmodelof maturestageof water- in the subcloudlayer. Two of the more interestingradial
spout life cycle [after Golden, 1971, 1974a, c]. Vertical scale velocity power spectraobtained through waterspoutfunnels
greatly exaggerated. are shownin Figures 6a and 6b. We note the relatively close

I0 298

..... .5

I0 TEMPERATURE,
(øK)

15

7.5

1 o
50 o 50 •oo •5o 2oo

DISTANCE(m)FROM THE VORTEX CENTER(v=o)

Fig. 5. Profilesof temperatureanomaly(dashedline), verticalvelocity (solidline), core pressuredeficit(dashedline),

and tangentialvelocity Vr (solid line) for AT-6 penetrationthroughan anticyclonicwaterspout. Aircraft altitude was
about 645 m and funnel diameter at penetrationtime was about 15-20 m (midsizedcase). (After Leversonet al.
324 REVIEW OF TORNADO OBSERVATIONS

30
tures in common.The severemidwesternevents have pre-
conditioning processes that are mechanistically different,
25 --
stronger,and larger in scaleand extend throughthe depth of
•' 20 ---
the troposphere,as commonly do the parent clouds. Simp-
son et al. [1991] extended this work to a well-documented
>, 1.5 -- anticyclonic waterspout over the Great Salt Lake. The
cloud-scale modeling study confirmed earlier GATE results
that a superadiabaticlapse rate, together with moderate
shear, needs to be collocated in the layer in which the
updraft increasesupward, in order for strongcumulus-scale
vortices to have maximum intensities at the lowest levels.
15 10 5 0 5 10 However, these ingredients had different root causes in the
Distance (m) Great Salt Lake case and GATE waterspouts. A high-
resolution axisymmetric vortex model [Howells and Smith,
1983] was used to investigate whether an intense vortex
could develop within a 5- to 15-min interval, when the model
cumulus vortices were used as initial conditions [Dietachm-
eyer, 1987].Simpsonet al. [1991]found that with the strong
25f Salt Lake cumulus vortex, an intense vortex with maximum
tangential
velocityof 35 m s-• developed
in nearlyall the
-• 20-- experimentsthey ran.
E

• 15--
o
3.2. Waterspout Linkages to Tornadoes
> 10--
Golden [1971, 1973, 1974b, c] has presentedevidence to
show that larger waterspouts have many similarities to
weak-to-moderate tornadoes. Golden [1971] documented a
0 large tornadic waterspout which was spawned by a line of
15 10 5 0 5 10 15
thunderstorms and pendant from a "wall cloud." In addi-
Distance (rn)
tion, several intense waterspouts have been observed that
moved onshore, or crossedover islands, inflicting consider-
able damage [Maier and Brandli, 1973; Golden and Sabones,
Fig. 6. (a) Waterspout velocity spectra measured by airborne
1991]. Both types of convective vortices have preexisting
Doppler lidar at midlevel (460 m), after Schwiesow [1981]. The cloud-scale circulations, which if large enough and close
simple, early solid rotation velocity structure shows the highest enough to a weather radar may appear as a hook echo or
tangential velocity observed in the experiment and the usual transi- "spike" in the case of large waterspouts.
tion to reduced angular velocity near the edge of the condensation The first caseof a Doppler-observedmesocycloneassoci-
funnel. The funnel core diameterwas approximatelyhalf that near
cloudbase.Translation inflowwas4 m s-1. (b) SameasFigure6a ated with a large, damaging waterspout in the Cape
except for a different waterspout at an altitude of 675 m. This was a Kennedy, Florida, area was documented by Golden and
double-walled funnel with sufficient condensation to mark the flow Sabones [1991]. In that case, the mesocyclone was only
between and outside the two major shells. detected in the lower levels of the parent thunderstorm, at
the southwestend of a line. The case of preexisting meso-
cyclones for at least the supercell-spawned class of torna-
fit of the first, most intense example to a classicalRankine- does is compelling; Golden [1974c] has used ground-based
combined profile and the double peak in velocity of the time lapse photography to document precedent circulations
second (a rare, double-walled funnel). at cloud base which subsequently were associated with
Holle and Maier [1980] reported a tornado over the waterspout formation.
Florida peninsula which was associated with invigorated During the period since the last Tornado Symposium,
convection as two gust fronts intersected. Simpson et al. there have been increasing numbers of tornadoes docu-
[1985] reported on an investigationof six waterspoutevents mented near and east of the Rocky Mountains. In fact,
documented during the 1974 GATE project. Triggering of tornado incidence over the state of Colorado has increased
intensification to funnel stage appears in some cases to be more than 100% in the decadeof the 1980sas compared with
effected by intersecting convergent features, with an invig- the 1970s (F. Ostby, personal communication, 1990). Many
orated cumulus tower overhead. of the increased sightings have occurred along the Front
Comparingtropical with severe continentalfunnel events, Range corridor of the Rockies from Denver north and
Simpson et al. [1985, 1986] noted that there is now some northeast in conjunction with field experiments conducted
evidence that the parent clouds have some important fea- by Programs for Regional Observing and Forecasting
 BLUESTEIN AND GOLDEN 325

t = 0
vices (PROFS) and the National Center for Atmospheric
Research (NCAR).
Zipset and Golden [1979] studied a mini-outbreak of three
4
tornadoes northeast of Denver, Colorado, on August 14,
1977. They documented several atypical features of the
parent cloud systems;indeed, the tornadoeswere pendant 3 km • No clouds

from rapidly growing cumulus congestus clouds on the

northwest flank of strongthunderstormsbut 5-10 km distant
from any significantprecipitation. Two of the three torna-
doeswere documentedphotographically,as shownin Plates (•Ut!!owbounda"ry
"'-"•/j• Convergence
zone
4a and 4b and were of large size and long duration. A , \._j.! o,on
pronounceddustband was a conspicuous,althoughtransient, Weak,
low-level
c•rculatlon
atintersectran
/ J ///////?/j/l//,.. j/.///////. //////..j/j./r/ /.//./.j/j/j/. }'//
featureduringthe maturestageof the largesttornado(Plate2b)
and is reminiscent of a similar feature documented in the Great t = 15 rain

Bend, Kansas, tornadoby Golden and Purcell [1977].Aspects

New convective development
of the tornadoes which could cause public confusionwere
noted, such as the disproportionatelyshortcondensationfun- 4 km m

nel from high-basedcumulusclouds(Plate 4a).

Indeed, recent field studiesof tornadoesin the High Plains 3 km •

of eastern Colorado have highlighted the fact that many

tornadoes develop without an associated mesocyclone
[Wakirnotoand Wilson, 1989]. Wilson [1986] describedthe

Ikm-- • ICirculation
advected
vert,cally
formation of several small tornadoes (sometimes termed
"gustnadoes"), not associated with mesocyclones,along
boundarylayer convergencelinesin northeastColorado.On
,,•'/•//////J////////•'•////'JJ////'•lt//r//•'r//"/'////'/'/'/'•"•'/"//
the basis of a study of the evolution of the July 26, 1985,
Erie, Colorado, tornado, Brady and Szoke [1989] developed
the schematicmodel shown in Figure 7. They believe that
this representsthe sequenceof formationof a class of
tornadoes which they term "landspouts," after Bluestein
[1985],who observeda similarvortexdevelopment in Okla-
homa during benign synoptic conditions.Landspoutsare
weaker,boundary-layer-forced tornadoeswhoseinitial cir- t = 30 rain
culationsform along or near the intersectionof mesoscale
boundariesand then develop up to cloud base with time. Rapid convective growth•:::

BradyandSzoke[1989]assertthatthe processes
depictedin 4 km•

Figure 7 "may accountfor a significantpercentof all

confirmedF0 and F1 tornadoes,and that processesinvolved 3 km •
in misocyclonetornadogenesis maybe crucialto the forma-
tion of mesocyclonetornadoes." (According to Fujita
2 km •
[1981],misocyclones rangefrom40 m to 4 km in diameter.) Strongconvectiveul;x:lrafts stretchand
intensifyvortex Into F1 tornado
Hess and Spillane[1990]presentedstrikingphotographs
of two maturewaterspouts in theAustralianGulfof Carpen- 1 km •

tariaandnotedthatthey exhibitedmirrorimagesof mostof

the key featuresdocumented by Golden[1974a, b, c] ,.///>,/////.,r////ii//////),////'//i//./

duringthe five stagesof the FloridaKeyswaterspout

life
Fig. 7. Schematicconceptualmodelillustratingthe development
cycle.TheseAustralian waterspouts wereproduced
froma of the Erie, Colorado,tornadostudiedby Brad),and S•oke[1989].
warmtopcumulus cloudline,withrainshowers
present,
and Theyhypothesized thatthisbasicscenarioappears to bevalidfor
converging gustfrontsmay haveplayeda role in vortex most nonmesocyclonetornadoes.
formation.
Finally, we concludeby notingthe stronglinkagesbe-
tweenwaterspouts and tornadoes foundby Goldenand c] for FloridaKeyswaterspouts.The two vorticesappearto
Purcell[1978b], after their detailedphotogrammetric and be similarin mostrespects,with differences
primarilyrelat-
damageanalyses of the UnionCity, Oklahoma, tornado. ing to parentclouddevelopment (bothappearto require
Theywereablefor the firsttimeto definea tornadolife vigorous convection)
andthe nearenvironment (tornadoes
cycle,depicted for theUnionCitycasein Figure3, which seemto requirea stronger
shearprofileandattendant helicity;
closelyparallelsthatdocumented earlierby Golden[ 1974a, both have high low-levellapserates).Wind profilers
326 REVIEW OF TORNADO OBSERVATIONS

NEXRAD Doppler radars may offer the prospectof adequate measured photogrammetrically may be somewhat lower
samplingof a variety of tornadic storm environments. than actual maximum wind speeds because (1) the velocity
component along the camera axis is not measured, (2)
differences may exist between tracer motions (size of several
4. PHOTOGRAMMETRIC WIND SPEED ESTIMATES
centimeters to a few meters over time periods of 2-10 s) and
There have been only a few new casesof photogrammetry actual air motions, (3) the locus of maximum winds may lie
applied to eyewitness tornado films since the last Tornado within an opaque dust column or in a region devoid of
Symposium. This appears to be largely due to the sudden tracers, and (4) the tangential wind velocity around suction
shifting of interest on the tornado wind speed problem by vortex axes often cannot be resolved.
previously supportive federal agencies to seismic and other Pauley and Snow [1988] photogrammetrically analyzed
issues. Golden [1976] summarized the spectrum of photo- the slowly translating Minneapolis, Minnesota, tornado and
grammetric analyses of tornado films up to that time and documented vortex breakdown, a precessing vortex core,
noted that this technique gives conservative estimates of a and a suction vortex. Their conjectures for the structure of
tornado's wind componentsin the plane normal to the line of this tornado, which was videotapedin great detail from a TV
sight of the camera. news helicopter, are given in Figure 9. It illustrates a
The assignment of radius is difficult because of typical supercriticalflow at the surface,a vortex breakdown a short
asymmetric flow fields in the tornado's debris cloud. These distance aloft, and a region of subcritical flow containing
problems are discussed in detail by Golden and Purcell axial downflow just above the breakdown.
[1978a]. These results taken together imply that highest Forbes [1979] also made some observations of relation-
wind speedsare in the tornado's boundary layer, below 50 m ships between tornado structure, underlying surface, and
above ground level (AGL), and generally in the range 75-95 tornadoappearancefor the Parker, Indiana, tornado of April
m s-•. However,we shouldemphasize thatthisis precisely 3, 1974, and the Cabot, Arizona, tornado of March 29, 1976.
where there are the fewest data points in previous photo- He proposedthat a "poor man's swirl ratio"
grammetric tornado studies, owing to the frequent obscura-
tion by intervening shrubs and/or buildings. Figures 8a and S -- Vmax/ Wmax
8b show velocity estimates and boundary layer streamlines may be useful in classifying single-vortex and multivortex
obtained by Golden and Purcell [1978a] for the Union City tornadoes. Multivortex tornadoes were observed for a range
tornado, by photogrammetric analysis of cloud tags and of S approximately 1.2 to 2.5. Forbes [1978] found that the
debris elements, in the tornado's decay stage. Parker, Indiana, tornado probably represents the extreme
(after the Xenia, Ohio, case) in development of suction
vortices. He was able to photogrammetrically analyze mo-
4.1. Flow Asymmetries tions in the suction vortices themselves and found maximum
It became increasinglyapparent during the mid-1970sthat netvelocities
of upto 125m s-• aloftandmaximum vertical
a significantclassof tornadoesexists that exhibit "suction velocityof upto 36 m s-• (at 324-melevation)in a suction
vortices" at some stage in their life cycle [Fujita, 1970]. vortex. It should be noted that there are different suction
Moreover, there are other forms of asymmetric flow struc- vortex configurations.Agee et al. [1977] documented a
tures in waterspoutsand tornadoes, including orbiting up- tornado near West Lafayette, Indiana, in which suction
ward jets at low levels [Golden, !974a, c] and accelerating vortices were confined to the surface boundary layer be-
radial inflow alongtrailing dust bands [Golden and Purcell, neath a single condensationfunnel; on the other hand,
1977]. The latter authors documented for the first time suction vortices in a tornado at Parker, Indiana, extended to
quantitativelythe existenceof such a dust band, as well as cloud base.
important upward nonhydrostaticaccelerationsin the dense Finally, Rasmussen[1982] and Rasmussenand Peterson
debris cloud of the Great Bend, Kansas, tornado, as illus- [1982] performeda digitally basedphotogrammetricanaly-
trated in Plate 5. Lee [1981] also presented new results on sis, includingan assessmentof errors, on the Tulia, Texas,
two tornadoes that exhibited suction vortices during a sig- outbreak tornado of May 28, !980. Campbell et al. [1983]
nificant portion of their lifetimes: Seymour, Texas, and alsoappliedthe samemethodologyto the Lakeview, Texas,
Orienta, Oklahoma. Highest net velocities measured photo- tornado of May 28, 1980 (in the same cyclic outbreak);
grammetricallyfrom the movies(tracking dust parcels and radial-vertical cross sections of the three wind components
cloudtags)were about90 m s-I at 30 m AGL for the for that tornado at its most intense stage are shown in
Seymourcaseand76m s-2 at 50m AGL forthetranslation Figures 10a, 10b, and 10c. Principalfindingswere (1) the
speedof a suctionvortex in the Orienta case. Lakeview tornado had a highly asymmetricvelocity distri-
The highest wind speed measured photogrammetrically bution like that of an occluded mesoscalefrontal system; (2)
thusfar by NSSLis the 95 m s-• netvelocitymeasured
at inflow was confined below about 50 m AGL; (3) core flow
200 m AGL in the Xenia, Ohio, tornado [Golden, 1976]. oscillationspossiblyindicatethe up-and-downtranslationof
Speedsalmostaslarge were measuredat muchlower heights an upperstagnation pointin the lowest300m (comparewith
(15-50 m) in movies taken by the Tornado Intercept Project Figure4), andcoreflow reversalandtransitionto a two-cell
(see Table 1, after Lee et al. [1981]). Note that wind speeds structureapparentlyoccurredabove 300 m; and (4)
 t•o

SMALL
43 WALL
CLOUD
4• R- 97m
33

b VELOCITIES
INm'c-t FUNNEL
(LOOKING
NE) /

PLAN VIEW OF INFERRED STREAMLINES

SURFACE
VORTEX

uP

Fig. 8. (a) Photogrammetricwind speedsfrom Union City, Oklahoma, tornado in meters per secondshowingcloud
tag velocity estimatesaround upper portion of funnel during tornado'sdecay stage. (b) Inferred horizontal streamflow
and vertical velocity distribution around tornado's base circulation in decay stage. (,By J. H. Golden and D. Purcell; see
Lee et al. [1981] for other
328 REVIEW OF TORNADO OBSERVATIONS

TABLE I. TornadoMoviesAnalyzedby NSSL, AfterLee et al. [1981]

Maximum Official F
Measured ScaleDamage
Wind Speed, Occurredat Rating of
Tornado Date TakenBy m s-1 Height,m Tornado Reference

KailuaKona,Hawaii Jan.28, 1971 eyewitness 56 4O Zipser [ 1976]

(tornadicwaterspout)
Union City, Oklahoma May 24, 1973 NSSL 80 90 F4* Golden and Purcell [1978a]
Salina,Kansas Sept.25, 1973 eyewitness 59 45 F5 Zipser [ 1976]
Xenia, Ohio April 3, 1974 eyewitness 95 200 F5 J. Golden and D. Purcell
(unpublished)
GreatBend,Kansas Aug. 30, 1974 eyewitness 85 80 F2 Golden and Purcell [1977]
Alva, Oklahoma June6, 1975 privatechase <60? * F1 unpublished
(anticyclonic team
tornado)
Seymour,Texas April 10, 1979 NSSL 77 15 F2 R. P. Davies-Jones[Lee et al., 1981]
Seymour,Texas April I0, 1979 G. Moore 90 30 F2 R. P. Davies-Jones[Lee et al., 1981]
Orienta, Oklahoma May 2, 1979 OU 76 50 F2 R. P. Davies-Jones [Lee et al., 1981]

*Shouldhave beenF5 accordingto Davies-Joneset al. [1978].

translationspeeddecreasedmarkedly as the tornadointen- speedsfrom surface and aerial damage surveys gained
sified and increased as it weakened. momentum from the maxitornado at Lubbock, Texas, in
May 1970.Golden[1976]givesdetailson methodsdeveloped
4.2. Probable Maximum Wind Speedsand Uncertainties by the wind engineeringcommunity to study modes of
failure to engineeredstructuresfor objective estimatesof
The highestwind speedsobtainedso far by photogram-
metric studiesare in the tornado'sboundarylayer, below z
= 50 m AGL, andgenerallywithintherangeof 75-95m s-1 .
(However, most data points in these studies lie above this
layer.) The highestwind speedstendto be closelyassociated
with transientfeatureslike suctionvortices, dustbands, etc.
There are, at times, large upward nonhydrostaticaccelera-
tions, as well as horizontal (inflow) accelerations.Highest
windspeedestimates
sofar (about125m s-1) arewell above
the surface[e.g., Forbes, 1978]and somewhathigher than
fixed-site
Dopplerestimates
(90m s-l for the 1981Binger,
Oklahoma, case [Zrnic et al., 1985]). Unresolved issues 2nd BREAKDOWN

requiringfurther researchincludethe frequency,role, and

causes of suction vortices and the role of horizontal and
vertical wind componentsin producingtornadodamage. Futwnl

5. WIND ESTIMATES FROM TORNADO DAMAGE SURVEYS

Someof the earliestassessments

of tornadoseverityfrom
aerial photographsof damagepatternswere madeby Van
Tassel [1955] and Prosser [1964]. Moreover, Van Tassel
[1955] publisheda remarkablephotographof cycloidalpat-
terns (loops) in the damage path of a tornado, which he
attributed to scratch marks on the ground made by large
debris elements.Fujita et al. [ 1970],however, suggestedthat
similar cycloidal patterns in the damage to grain fields from
one of the 1965 Palm Sunday tornadoes was due to the
presence of one or more "suction spots" which rotated
aroundthe tornado'score. By using aerial photographsto
make estimates of damage-loop parameters, Fujita et al. Fig. 9. Summarycross-sectionalsketch (scale exaggerated)show-
ing conjecturedmeridional flow structure of Minneapolis, Minne-
[1970]calculatedgroundspeedsof 72 to 78 m s-1 for one sota, tornado [after Pauley and Snow, 1988]. This is a compositeand
tornado.
does not necessarilyrepresentthe tornado' s flow or other features at
The systematic, quantitative assessmentof tornado wind any single
 BLUESTEIN AND GOLDEN 329

wind speedsthat causedthe failure (damage).Both wind

FILM SEGMENT8 engineersand Fujita and colleagues
independently
studied
the Lubbock damage. From the Lubbock case and subse-
quenttornadoes,Fujita [ 1971] andFujita andPearson[ 1973]
developedthe F scale, which subjectivelyranks tornadoes
accordingto their damage severity (discretewind speed
rangefrom each of six categories),path length,and path
width. Concernsdevelopedregardingthe applicationof the
F scale by untrained personnel and especiallyits use for
residentialdamagewhere there may be great local inhomo-
geneity in the quality of construction and materials used.

2.%
From all tornadoessurveyed up to 1977, Texas Tech wind
engineers claim [Minor et al., 1977] that "no conclusive
evidencecan be found that groundlevel wind speedsexceed
112m s-1, andmostbuilding
damage is caused
bywindsin
!
' the34-56m s-• range."Theyalsofoundthatatmospheric
pressurechangesin tornadoes play only a minor role in the
damagingmechanism.In particular, they suggestedthat the
0"
discretewind speedranges assignedto the F3-F4-F5 cate-
gories should be lowered somewhat and should overlap
[Minor et al., 1977; Marshall et al., 1983, 1985; Marshall,
1985].
The F scale is now a routine part of local National
Weather Service office tornado reporting to the national
storm data archives; Fujita has cautioned, however, that the
assignmentof an F scale rating to a given tornado event is
dependent on observer training and experience and that
theseratings have a possible range of error of -+1 F number.
This is especially true when the tornado in question has
movedover open terrain; for example, the Seymour, Texas,
tornado had photogrammetric wind speed estimates of 77
ms-• at 15mAGL, at a pointin itstrackoveropencountry
where the damage rating was F0 [Lee, 1981].
It is imperative that damage surveys be continued, not
only for verification with research programs on severe
storms, but for operational verification as we enter the
NEXRAD era. There will be many hybrid severe weather
phenomena,including convective vortex phenomenaover a
wide spectrum, which require surface "ground truth."
Doswell and Burgess [1988] argue convincingly for the
continued need for quick-response damage surveys follow-
ing severe storm events. Ideally, teams shouldbe composed
of at least one meteorologistand a wind engineerand should
be dispatched within 24 hours after the event if at all
possible,owing to the rapid cleanupof damage.The process
needsto be both interdisciplinary and institutionalized;we
should be prepared for many surprisesin the NEXRAD/
wind profiler era. Moreover, there is an urgentresearchneed
for independentwind estimatesat various points and times

Fig. 10. (Opposite) Vertical-radial cross sections, relative to

Lakeview, Texas, tornado's center at lower left of (a) tangential
velocitycomponent,(b) radial component,and (c) verticalcompo-
nent, all in meters per second. Obtained by photogrammetric
analysesof film segments,after Campbelleta!. [1983].Outline of
main tornado debris cloud is
330 REVIEW OF TORNADO OBSERVATIONS

along the same tornado's path, using remote sensors,

quickly deployed in situ sensors, photogrammetry, and
damage surveys. Fujita et al. [1976] made this point elo-
quently at the last Tornado Symposium, and progressin this
area has been slow up to recent years.

6. IN SITU MEASUREMENTS

Although the wind field may be mapped above the ground

remotely with Doppler radars (see section 7), pressure,
temperature, and water vapor content cannot be similarly
measured with ease. Furthermore, even Doppler radar can-
not measurethe wind field directly at the surface. During the
late 1970s, several instruments were developed which could
be brought to bear on tornadic storms by intercept crews, so
that the number of direct measurements could be increased
substantiallyover that obtained by the chance passageof a
tornado or its parent wall cloud over a fixed instrument.

6.1. Serendipitous Slt•J•lr'e Measurements

Occasionally, a tornado may pass over or close by a
weather station and produce a pressureor wind trace, if the
instruments survive and can be calibrated. Flora [1953] and
Davies-Jones and Kessler [1974] have presented most of
these records, which include eyewitness observations of
wind or pressureextrema on anemometeror aneroidbarom-
eter dials. While pressure drops of up to about 200 mbar
have been observed on aneroid barometers by private citi-
zens, these pressure drop measurements are suspect. A
34-mbar pressuredeficit was recorded in the 1962Newton,
Kansas, mesocyclone[Ward, 1964]. Some measurementsof
Fig. 11. TOTO being testedin a wind tunnelat Texas A&M
small pressuredeficitsmay be underestimatedowing to the Universityin March 1983.The stripchartsand electroniccircuitry
slow responsetime and dampingof the instruments. are contained inside the cylinder; the wind, pressure. temperature,
The bulk of the data, in our opinion, indicates that the and coronadischargeinstrumentsare mountedon the boom at the
top. (Photographcopyrightby H. Bluestein.)
largest, crediblepressuredrop measurementto date in a
tornado'$ core is about 100 mbar [Davies-Jones and Kessler.
1974].There have been fewer reliable wind recordsthat have 6.2. TOTO
surviveddirect hits by tornadoes.Fujita et al. [1970] found
one anemometertrace that clearly shows the effects of two In 1980,Bedard and Ramzy [1983] developeda 181.6-kg
(400 lb) portable instrumentpackage (Figure 11) called
tornado
passages'
thefirstcaused
peakgusts
to65m s-• as TOTO (TotableTornado Observatory),whichwas appropri-
the tornado center passedjust north of the station' the
ately namedafter Dorothy'sdog, who, alongwith Dorothy,
second
hadgusts
onlyto nearhurricane
force(>•30m s-•) was carriedaway by a tornadoin C. F. L. Baum'sbook The
about I hour later.
WizardoJ'Oz.TOTO wasdevelopedfurtherby J. Carter, S.
Tornadoeshave passedclose to or directly over NSSL Frederickson, and E. Kelly at NSSL. The device used
mesonet surface sites on more than one occasion, but the "hardened" sensors that had been used to measure wind
instruments have usually been destroyed or severely dam-
shear and severe downslopewinds. It was designedto be
aged.Barnes[1978]studiedan extensiveoutbreakof severe mountedin a pickuptruck anddeployedin 30 s or less.Wind
stormsthat passeddirectlythroughthe NSSL mesonetwork speed,wind direction,temperature,static pressure,and
on April 30, 1970;one of the tornadoespassedcloseby a coronawere recordedon stripcharts. TOTO was first tested
mesonet
siteandproduced
peakguststo 46ms-• . Another in Colorado during the summer of 1980.
NSSL mesonetsite was completely destroyedby a tornado Interceptteamsfrom the Universityof Oklahoma(OU)
near Fort Cobb, Oklahoma, on May 20, 1977. However, it traveled in two vehicles, the parent vehicle and TOTO's
has never been possible to determine whether or not a pickuptruck,to stormsin Oklahoma•tndnearbypartsof
tornado'score region passeddirectly over the recording Texas duringthe springseasonsof 1981-1983,in an attempt
instrument in these serendipitous cases. to placeTOTO directlyin thepathof tornadoes (Figure
 BLUESTEIN AND GOLDEN 331

Fig. 12. Photograph of a tornado on May 11, 1982, over Altus Air Force Base, Oklahoma, to the southwest, another
funnel cloud developing to the west, and TOTO deployed in the foreground. The tornado dissipatedand did not pass
over TOTO; the funnel cloud to the right developed into a large tornado that moved to the north-northwest for over an
hour and also never passedover TOTO. (Photographby K. Hodson.)

and then to recover the instrument packagefor analysisback onlyapproximately

50 m s-l, whichis lessthanthe maxi-
home [Bluestein, 1983a, b]. Teams from NSSL attempted mum wind speed in some tornadoes. If TOTO is ever used
to do the same during 1984 and 1985 [Burgesset al., 1985]. again, the tip-over speed will have to be increased by rapid
Pressure perturbation deficits of 2-5 mbar were found anchoring of TOTO or by increasing TOTO's diameter.
within 1.5 km of the path of tornadoesand underwall clouds.
Thehighest (guststo 36m s-l) weremeasured 6.3.
windspeeds Turtles

on May 22, 1981,near Cordell, Oklahoma, in the rear flank Ten portable instrument packages, which are much
downdraft, 2.3 km from a tornado. The traces of pressure, smaller and lighter than TOTO, were developed at OU and
wind speed, wind direction, and temperaturefor a case in first tested in the spring of 1986 [Brock et al., 1987] in
which TOTO was located in the path of a tornadoare shown Oklahoma and Texas. Owing to their appearance, the instru-
in Figure 13. The maximum instantaneousrate of pressure ments were named '•Turtles" (Figure 14). Only pressure and
fall was 16 mbar min-1; the maximumwind gust,however, temperature can be measured' they are recorded, and the
wasonly30 m s-l. Turtles are retrieved and read later, back home. The prob-
TOTO, which is no longer being used, is currently at ability of making a direct measurement in a tornado is
National Oceanic and Atmospheric AdministrationHead- increased by deploying a number of instruments, such as the
quartersin Washington,D.C. It turnedout to be extremely Turtle, in series along ro'ad•vays rather than by deploying
difficult to place TOTO directly in the path of tornadoes, only one, such as TOTO. Furthermore, a small-scale net-
owingto the short-livednatureof most tornadoesand the work can be deployed in the vicinity of the tornado by
coarsenessof the network of goodroads.Furthermore,wind several storm-intercept vehicles, with the objective of map-
tunnel tests indicated that TOTO's tip-over wind speed is ping out the larger-scalefeatures around the
332 REVIEW OF TORNADO OBSERVATIONS

29 APRIL 1985 ARDMORE 6.4. Portable Rawinsondes

8 '-

With the advent of portable radiosonde units in the early

1980s, efforts were undertaken to release soundingsinto the
wall cloud of tornadic and potentially tornadic storms and,
just outside these storms, to determine the "environmental"
0 C- profile of temperature and water vapor content. Such sound-
ings are useful to numerical modelers who try to simulate

T
E
27 I tornadic storms and to theoreticians who try to understand
why certain environments are conducive to tornadogenesis.
Mobile ballooning increases substantially the likelihood of
obtaining an in-cloud or tornado proximity sounding, be-
LU 22
cause the probability that a tornadic wall cloud will pass near
a fixed sounding site at launch time is extremely low.
Z 16 The first attempted sounding release occurred in a tor-
nadic storm on April 26, 1984, in central Oklahoma
0 [Bluestein and Woodall, 1990]' however, owing to underin-
flation, the sounding skimmed across a wheat field toward
o the wall cloud of the storm. The first successful release of
soundings both into the wall cloud of a tornadic supercell
180 and in the storm's environment (Figure 15) occurred in the
Canadian, Texas, storm of May 7, 1986 [Bluestein et al.,
270 1988]. Winds were computed using an optical theodolite; an
360 updraftof almost50 m s-] wasestimated
andcomparedto
• 4
parcel theory calculations, which were in near agreement.
Other soundings have been successfully released into wall
clouds [Bluestein et al., 1989, 1990a] and close to growing
•-•- 2 storms [Bluestein et al., 1990b].
The first attempts to measure both the thermodynamic
structure and the wind profile inside severe storms began in
0
1987, when NSSL had NCAR's Cross-Chain LORAN At-
1 2 3 4 5 6 7 8
mospheric Sounding System (CLASS) [Lauritsen et al.,
TIME (min)
1987] installed in one of their storm intercept vehicles [Rust
Fig. 13. Relative pressure, wind speed, relative wind direction, et al., 1990; Rust and Marshall, 1989]. They named their
and relative temperature as a function of time on April 29, 1985, as mobile version of CLASS "M-CLASS." (NCAR now has its
recorded by TOTO almost directly in the path of an F2 tornado near own M-CLASS unit, which is used in field experiments.) On
Ardmore, Oklahoma. (Courtesy of R. P. Davies-Jones.)
May 31, 1987,NSSL successfullylaunched a CLASS sound-
ing in eastern Oklahoma into the wall cloud of a storm after
a tornado had dissipated [Rust, 1990]. The advantage of
CLASS is that the winds inside the storm can be determined

On May 2, 1988, a Turtle was placed under a dissipating

tornado by OU students near Reagan, Oklahoma. Unfortu-
nately, the Turtle was apparently picked up and tampered temperoture
outer shell sensor
with; no useful data were recovered (J. LaDue, personal
communication, 1988). There were a number of Turtle
deployments in 1989 and 1991 (J. LaDue and M. Shafer,
personal communication, 1991). On May 26, 1991, four
! ""t'"""' '""'"""' /
i......o...
oo
Turtles were deployed around a tornadic storm near Moore-
land, Oklahoma. One recorded a pressure perturbation def-
icit of 4 mbar, approximately 1.6 km from a tornado. This is
consistent with TOTO's measurements. Based on the TOTO lead rinõ
and Turtle measurements to date, it seems that one must be
much closer than I km from a tornado to observe pressure 10 œm

deficits in excessof 5 mbar. Temperature anomalieswithin a

Fig. 14. Schematiccross section,drawn to scale, of the Turtle.
tornado's core region at the ground have not yet been The plastic container that contain,the electronic circuits and
measured. It is anticipated that the Turtles will continue to batteriescan be removed quickly from the rest of the protective pod.
be used at OU. (From Brock et al. [
 BLUESTEIN AND GOLDEN 333

from trackingusingLORAN. The main problemsinvolve

highnoiselevelsinsideelectricallyactivestormsand "lock-
ing on" distant LORAN stations.

6.5. Instrumented Rockets

Colgate [1982] designedand constructedsmall, light-

weight(lessthan 1 kg)rocketswhichhe proposed
to launch
from a light aircraft into tornado funnels. Miniaturized
electronicswere packagedto measurepressure,tempera-
ture, ionizationand electricfield variationsalonga rocket
trajectory penetratinga tornadofunnel. Colgatebelieved
that sucha tornadorocket probe shouldfly closeto Mach I
in order to penetrate the tornado funnel and not have its
trajectory severely perturbed by the tornado'sflow field.
Figure 16 showsthe rocketsmountedon the wingrack of
the private aircraft. Colgate flew for three spring storm
seasonsin Oklahomaand adjacentstates,guidedby fore-
casts and radar information relayed in real time from NSSL.
Colgatewas able after severalunsuccessful
flightsto inter-
cept tornadoes and launch his rockets into a few of them; Fig. 16. S. A. Colgate inspecting small tornado rocket probes
mountedon wing rack of his Cessnalight aircraft, prior to a mission.

unfortunately, the light weight and fragility of the rockets

themselves, mandated by Federal Aviation Administration
regulations,proved to be an insurmountableobstacle. Many
of the rockets became soaked with water during the air-
craft's maneuversaround the tornado in heavy precipitation
and misfired, missing their intended target. Moreover, in
-1
1982 Colgate's aircraft was caught in an estimated 63 m s
inflow into a tornado near cloud base, experienced extreme
turbulence, and was forced to make an emergency landing in
an open field.
One of the tornadoes probed by Colgate's rockets is
shown in Plate 6; the path of the rockets was documented b)
three different photographic systems mounted on the air-
craft. Colgate noted that the "pilot load with this much
equipment has proved excessive." A potentially more prom-
isingtechnique for launching instrumented, constant-volume
balloons into a tornado was proposed by Grant [1971], using
a research aircraft to drop the self-inflating balloon package
ahead of the tornado's path. In the light of hindsight, this
approach deserves teevaluation, perhaps with field trials on
waterspouts.

7. DOPPLER RADAR MEASUREMENTS

The safest way to obtain wind measurements in tornadoes

is to probe them remotely by Doppler radar. The Doppler
radar measures the line-of-sight (radial) component of the
wind in the presence of precipitation or cloud droplets.
The main problem in obtaining good Doppler data sets is
that the tornadoes must be relatively close to the radar, so
that the spreading of the radar beam with range does not
Fig. 15. Photograph of an OU facultymember {G. Lesins)an•t compromi.,,ethe •patial resolution of the measurements and
graduate student (E. W. McCaul, Jr.) releasinga radiosondeinto a
developingsupercellin western Oklahomaon May 20, 1985.{Pho- so that measurements can be made relatively close to the
tograph copyright by H. Bluestein.) ground, where we lack data in other techniques. For
334 REVIEW OF TORNADO OBSERVATIONS

pie, a 1ø half-power beamwidth at 20 km correspondsto a folding is not a problem in a CW radar; however, no range
cross section of 350 m, which is wider than most tornadoes. information is obtained.)
Thus the tornado must be very close to the radar, an event NSSL in the 1970s and early 1980s made a systematic
which is unlikely. However, even if the tornado is very close attemptto obtaintornadicwind spectrawith a fixed 10-cm
to the radar, scans at low levels might be contaminated by pulsedDopplerradarlocatedin Norman,Oklahoma.In all of
ground clutter. their successful measurements, the parent storm of the
When the radar field of view encompassesall or most of tornado had supercell characteristics.Spectra were first
obtained on May 24, 1973, in a tornado that struck Union
the tornado, spectra (backscattered power versus wind
City, Oklahoma,at a rangeof approximately50 km [Zrnic
speed) of the radial wind component are computed. The
and Doviak, 1975]. For the first time the sense of wind
spectra allow one to estimate the maximum wind speed
directionalongradialswas determined.With correctionsfor
within the tornado and, by examining the shape of the
velocityfolding(aliasing)and simulatedspectrafitted to the
spectra,learn, with some assumptions,somethingabout the observations, maximumwind speedsof 72 m s-1 were
nature of the radial profile of the wind [Zrnic and Doviak, estimated[Zrnic et al., 1977].In a tornadothat hit Stillwater,
1975].Dopplerradar wind estimatesof the maximumwinds Oklahoma,on June 13, 1975,at a rangeof approximately 100
are likely to be underestimatesbecause the highest wind km,maximum windspeeds of 92m s-• wereestimated using
speedsmay be confinedto very small volumes, so that the the sametechniquesusedto obtainthe wind speedestimates
energy backscatteredto the radar is below the noise level in the Union City tornado.
and therefore cannot be detected. When the radar field of NSSL obtained the first tornadic wind spectra uncontam-
view encompassesa small portion of the tornado, the mean inated by velocity folding in the Del City, Oklahoma, tor-
velocities within azimuth and range bins are computed. nadoon May 20, 1977,at a rangeof 35-40 km, with the radar
These Doppler radar wind estimatesare also likely to be operatingat a highpulserepetitionfrequency(PRF). Com-
underestimatesowing to averagingwithin the radar volume. bined with simulated spectra, Zrnic and Istok [1980] re-
The ideal situation is one in which the tornado is close portedmaximum
windspeeds
of 65 m s-•. Unaliased
enoughto two radarsto obtain a dual-Doppleranalysison a spectrawereobtainedin the high-PRFmodein the Binger,
scale small enoughto resolve the wind speedswithin the Oklahoma, tornado on May 22, 1981, at a range of 30 km
tornado vortex, an event which is not likely. [Zrnicetal., 1985].
Maximum
windspeeds
of90m s-• were
Unfortunately, the maximum unambiguousvelocity of estimated (Figure 17).
typicalpulsedDopplerradarsoperatingat wavelengths
of 3,
5, and 10cm, with pulserepetitionfrequenciesshortenough 7.2. Fixed-Site Doppler Radar Observations
to obtain reasonablylong unambiguousranges,is small in of NonsupercellTornadoes
comparisonwith the maximumwind speedsexpectedin In eastern Colorado during the late spring and early
tornadoes.For example,the maximumunambiguous veloc- summer,tornadoespendantfrom parentthunderstorms
that
ity for a maximumunambiguous rangeof 100km at 10 cm (3 are not supercelIs("gustnadoes"and "landspouts")are
cm)is only37.5m s-• (11.3m s-j) (seeequation (7.4)of morecommonthan supercelltornadoesin Oklahoma.Con-
Doviak and Zrnic [1984]). To attain maximum unambiguous sequently, between1984and1988therehavebeenmorethan
velocitiesin excessof 100 m s-• at 10 cm, the maximum two dozenserendipitous measurements at rangesof 15-45
unambiguous rangemust be lessthan 37.5 km; sincethun- km in thesetornadoesduringvariousfieldexperimentsusing
derstormsare typically of the order of 30 km or more in 3-cm,5-cm,and10-cmDopplerradars,whichwere operated
width, the signalmay be contaminatedwith rangefolding. by NCAR andNOAA. Sincethe tornadoeswere relatively
small in diameter, most of the observationswere of the
tornadoes'parentvortex; consequently
most of the wind
7.1. Fixed-Site Doppler Radar Observations measurementsare probablyunderestimates.
of Supercell Tornadoes Maximum single-Dopplerwind speed estimates (from
meandata,notfromspectra)
rangefrom16to 57 m s-•
J. Q. Brantleyof CornellAeronauticalLaboratoryfirst [Wilson, 1986;Wakimotoand Martnet, 1989;Brady and
suggestedin 1957thatDopplerradarcouldbeusedto probe Szoke,1989;Wakimotoand Wilson,1989].Photographsof a
tornadoes[Brant!ey, 1957].The first Doppler radarmeasure- tornadonear Denver, Colorado,on July 2, 1987, superim-
mentsof wind speedsin a funnelcloudwere madeby the posed uponanalyses ofradialwinddata(Figure18),havefor
U.S. Weather Bureau on June 10, 1958, at a range of 41 km the firsttimerevealedthe relationship betweenthe wind field
in a supercellstorm that shortlythereafterproduceda in a tornadoand the appearance of its condensationfunnel
tornado that hit E! Dorado, Kansas [Smith and Holmes, [Wakimoto and Martnet, 1989].
1961].The 3-cm, continuouswave (CW) radar, obtained The first dual-Dopplermeasurementsof a tornado were
from the U.S. Navy, yielded wind spectrathat did not also madein the tornadonear Denver on July 2, 1987[Uttal
discriminate
betweenapproaching and recedingvelocities. et al., 1989], using two pulsed 3-cm radars operatedby
The maximumvelocitiesdetectedwere 92 m s-• . (Range NOAA's WavePropagation Laboratory(WPL) at a range
 I'' ' I •l , I ' ' ' I'' ' I' ' ' I •" ' l' ' ' I • I'll ' ' ' • ••'• • • 1• I•., • • • I • '•1' '• I ' •-"1 c• •-I • ' ' 1'•"-' I • t t-'•,-t ß] , , • i •'"' 1

•.. •-'•.•.• • .• ... ......:.%

.... • . •:.- ..
"•ß"•'
•,'.'
' ':"•""•"• '•
. q..:,..• •".
RANGE
30.tt3KMGATE
03
.... :• '•:.'.... . ,.•••

• ,.,..,.• ,

RANGE
29.•13
KMGATE
01
o-- I' ' ' I ' '"1 ' ' "1"• I "' 'l•' ' I ' ' ' I '½' I "' ' I ' ' , •

L ,½ ' '"- XS•. . •' ' .. ½

.": . .: .;.,..
, • RLTITUDE TI•E
190512
.251 RRNGERZI•UTH
281.6
ELEVRTIBN
.•
28.•1• K• 6RTE
05 S/N 17.•Cf
RLT[TUDE ,251
TI• 130512 RZI•TH282.1 ELE•TI•
R• 28.•1• K• GRTE • 5/N 15, RECT
0,52281 T[HE 19•S13 RZ[MUTH 281.1 ELEVRT[BN.3
RLT[TU• .2•3 RRN• 28.91• K• GRTEOS 5/N 15. RECT

Fig. 17. Spectrafor three azimuthsand three rangegatesin the Binger, Oklahoma,tornado on May 22, 1981.The
solid line is a runningaverageof data points(crosses).The rangesof the spectraare for volumes28-30 km from the
Norman Doppler radar The altitude of the volumesis about 200 m AGE Power in decibelsbelow the peak power is
plotted on the ordinate; the radial wind velocity in meters per secondis plotted on the abscissa.(From Zrnic et al.
[1985]; courtesy of D. Zrnic.)

NOAA-C •
1535:14- 1537:09 •

, 4

Fig. 18. Photograph

of a tornadoonJuly2, 1987,nearDenver,Colorado,
superimposed
on a verticalcrosssection
of radialwind velocitiesin metersper second,with .•olid(dashed)contoursrepresenting flow out from (into) the plane
of the cros•section,whichpasses throughthecenterof a hookechoassociated with the storm.Elevationangleplotted
on the ordinate;azimuthfrom the radar plottedon the absci,a; the heightaboveground(kilometers)indicatedon the
right.(Courtesy
of R. Wakimoto;
fromWa]•imoto
and
336 REVIEW OF TORNADO OBSERVATIONS

87/ 7/ 2 [5 35 3q-I5 35 3q COMBIN Z= O.qO KN

(RS OF 2/1.3/89) ORIGIN--( 0.00, 0.00] KM X-AXiS= 90.00EG
NOIqt:::I/WPLRROAR OI::;TPi- VOLUME 7

•.o ,., •, •, •, •, •, •, •, •, •, •, •,,,,,/,/,/,l, [, •, •.,•.,•.,\, •.,•.,•

,_,,,, , , .... I t: ix
._,,,,,,..,. ../ t l ix
,.,, ://1 1. II

-22.0 -21.0 -20.0 -19.0 -18.0 -17.0

X KM UFt .Vt::) EVERY 1 PT. 20. HIS .... >

Fig. 19. Dual-Doppler analysisof the wind field in the tornadonear Denver, Colorado, on July 2, 1987,at 1535MDT.
Abscissa(ordinate)is distancein kilometerseast (north) of the NOAA-C radar. (Courtesyof B. Martner.)

approximately 20 km. The vortex is clearly resolved at a more, the sensitivity to the highest wind speedscan also be
height of 400 m above the ground (Figure 19). Although the increased by getting closer to the tornado. The portable
parent vortex is well resolved, details about the vortex are Doppler radar can also probe volumes closer to the ground
limited by the 200-m resolution. Dual-Doppler analyses than distant radars. Simultaneous visual documentation is
(Figure 20) were also obtained in other tornadoes near also possible with a portable radar; videotapes can be
Denver on June 15, !988, based upon data from the Massa- photogrammetrically analyzed to obtain complementary az-
chusetts Institute of Technology's Lincoln Laboratory imuthal wind speed estimates.
10-cm pulsed radar and the University of North Dakota's In 1986, T. Morton of Texas Instruments contacted H.
5-cm pulsedradar at rangesof approximately 20 km [Roberts Bluestein and suggestedthat a portable radar he had been
and Wilson, 1989; Wilson and Roberts, 1990].
working on at the Los Alamos National Laboratory might be
useful for tornado research. The portable, low-power CW
7.3. Portable Doppler Radar Observations Doppler radar (Figure 21) was modified in 1987 for meteo-
of Supercell Tornadoes rological use [Bluestein and Unruh, 1989]. The radar was
Zrnic et al. [1985] suggestedthat higher-resolutionwind similar to the Smith and Holmes [1961] radar, except that
data in tornadoes could be obtained with a portable radar, approachingand receding velocities could be discriminated.
which could be transported by storm intercept teams to During the summer of 1988 the radar was modified so that
locations close to tornadoes. A portable Doppler radar is range information could be obtained using an FM-CW pro-
useful also because the number of data sets obtained can be cessor [Strauch, 1976]. Details on the technical aspects of
significantlyincreased by going to the tornadoes,rather than the radar and its signal processingand results are given by
by waiting for the tornadoes to come to the radar; further- Bluestein and Unruh [1991, this
 BLUESTEIN AND GOLDEN 337

•CAITAL Wt,DS 4AD VERTICALVCR'FEiTY

22OO
22.

20.

18.

16.

!
!
12.
t
10.

8.

6.

,%%,,,,

\,,,

2.

,,,
-28. -26. -2q. -22. -20. -!8. -16. -tel. -12. -10. -8. -6. -q.

DISTAINEI•
EASTOF Fl_2 Oreo 20. M/$ ----->
Fig. 20. Dual-Doppleranalysisof the wind fieldin tornadoesnearDenver, Colorado,on June 15, 1988,at 2200UTC.
T1, T2, and T3 are locationsof three tornadoesthat occurred;notethat not all vorticesproducedtornadoes.Vertical
component
of vorticitycontoured
at intervals
of5 x 10--• s-l. Abscissa
(,ordinate)
distance
in kilometers
east(north)
of the MassachusettsInstitute of TechnologyLincoln Laboratory radar. (Courtesy of R. Roberts.)

The portable radar was first used in CW modeto probe FM-CW processing.An FM-CW data set was first recorded
wall clouds(Figure22). On May 2, 1987,a distanthigh-based in a wall cloud on June 6, 1989, near Floydada, Texas.
funnelcloudwasprobed;however,its fieldof viewwasonly Subsequent CW data sets, some FM-CW data sets, and
a tiny fraction of the volume illuminatedby the radar. The some simultaneousvideotapes were recorded for many wall
first tornado to appear in the field of view of the radar clouds and for one tornado in 1990 and five tornadoes in
occurredbriefly,withoutadequatevideodocumentation,
on 1991. A preliminary inspection of the data indicates wind
windspeeds speeds
May 25, 1987,in the northernTexasPanhandle; in theCW modeashighas120-125m s-1 [Bluestein
as highas 60 m s-• or higherwererecorded.On May 13, and Unruh, this volume] in a tornado on April 26, 1991, in
1989, the completelife history of a tornadoin Texas was north central Oklahoma; the radar was 1.2 km away from the
capturedin the field of view of the radar (Figure 23). tornado's damage path, which was 0.8 km wide. This might
Unfortunately,both the CW and FM-CW data sets were be the first recorded data set in an F5 tornado [Fuji(a, 1981]
contaminated as a result of an incorrectly set bias in a by a Doppler radar. The results from the 1991 data are
preamplifier,
whichhad beenrecentlyaddedto allowfor
338 REVIEW OF TORNADO OBSERVATIONS

1988). The objective of the RPV is to fly around a tornado

and telemeter back thermodynamic measurements.
To date, two planes, with 1.5- and 2.4-m wing spans, have
be•n developedOne is equippedwith a real-timevideo
camera and can be piloted remotely. It takes 2 min to launch
the RPV from an intercept vehicle. Experience needs to be
acquired in launching the RPV near storms. The RPV should
be developed further to have a navigation aid location
system and meteorological instrumentation. It has the po-
tential to provide valuable thermodynamic and wind infor-
mation near tornadoes.

8.3. Portable Doppler Radar

Improvements are suggested for the existing portable
CW/FM-CW radar by Bluestein and Snrtt]t [this volume]. In
addition, technology has advanced to the point at which it
may be possible to develop a portable, low-power, pulsed
Doppler radar.
In order to improve the sensitivity of a portable radar to a
tornado composed of cloud droplets and small precipitation
particles, millimeter-wavelength radars need to be devel-
oped [Pasqualucci et al., 1983; Hobbs et ai., 1985: Lher-
mitre, 1987]. A portable 35-GHz CW Doppler radar already
exists at the Los Alamos National Laboratory (W. Unruh,
personal communication, 1991) and needs to be tested: if
tests are successful, the radar should be modified to have
FM-CW capability.

8.4. Portable and Airborne Doppler Lidar

Schwiesou' et al. [1981 ] and Schu'iesou' [1981] used an
airborne 10.6-/zmCW Doppler lida.r from WPL in the Florida
Fig. 21. Photograph of University of Oklahoma graduate student Keys in August and September 1976 to obtain radial wind
G. Martin operating the first version of the Los Alamos National spectra, with 0.75-m azimuthal resolution, of waterspout
Laboratory portable Doppler radar in 1987. (From Bhtestein and
Um'uh [ 19891.)
winds integrated over range. M('Caltl et al. [1987] used
NASA's airborne, lore-and-aft scanning, pulsed Doppler
lidar to obtain analyses of the horizontal wind field in gust
8. THE FUTURE OUTLOOK
fronts and around growing cumulus congestus clouds.
Pulsed Doppler !idars from WPL and NASA were used in
An impressive array of mobile instruments was developed Colorado in 1982 to obtain dual-Doppler analyses of gust
in the 1980s. Instruments now in existence need to be fronts, with 320-m range resolution [Rothermei et al., 1985].
improved. Other devices need to be developed. Based on the aforementioned experiments, it appears that a
Doppler lidar mounted on an aircraft or made portable
8.1. Turtles
enough to be mounted in an intercept vehicle can be used to
The Turtles need to be developed further, with the addi- obtain wind measurements, with fine azimuthal resolution,
tion of water vapor sensors and telemetering capability back in the clear air regions of tornadoes and in condensation
to the intercept vehicle. The locations of deployed Turtles funnels. NCAR is developing an airborne infrared !idar
need to be determined from navigation aids, and the most system (NAILS) that will have 100-m range resolution and a
promising technology for doing this accurately is with the range of 7.5-10 km [Schn'ieso•v, 1987; Schn'ieson' et al,
GPS satellite system. 1989; R. L. Schwiesow, personal communication, 1990].
NASA Marshall Space Flight Center, in cooperation with
8.2. The Remotely Piloted Vehicle
WPL and the Jet Propulsion Laboratory, is also tentatively
The development of a remotely piloted vehicle (RPV) planningto develop a new airborne Doppler lidar system (J.
began in the fall of 1987 by K. Bergey and his aerospaceand Rothermel, personal communication, 1991).
mechanical engineering and electrical engineering students The problems in designinga lidar system that is suitable
at OU (K. Bergey and J. LaDue, personal communication, for probing the wind field of a tornado are
BLUESTEIN a. ND GOLDEN
340 REVIEW OF TORNADO OBSERVATIONS

Plate 2a. Tornado in mountainous terrain: over Basalt Mountain (10,000 feet MSL) in Colorado at 1300 MDT on July
12, 1982. (Photo by S. Jones from 16 km away.)
 BLUESTEIN AND GOLDEN 341

Plate 2b. Tornado in mountainous terrain 4 km south, over Alamosa Valley in Colorado (near the New Mexico
border) at 1115 MDT on July 6, 1979, looking SSE. (Photo by R. Alexander and courtesy of R. F. Abbey.)
342 REVIEW OF TORNADO OBSERVATIONS

Plate 2c. Destructivetornadoover ManitouSprings,Colorado,at the foot of PikesPeak (elevation2 km) on June24,
1979 Funnelnever extentledmore than 30-40% downwardto surfacefi'omparentcloud
 BLUESTEI A D GOLDE 343

Plate 3. Composite view of typical Great Plains tornado-spawning 5upercell storm, from 8 years of NSSL tornado
intercept observations (produced circa 1980 by J. H. Golden). Horizontal scale is compressed, and all features shown
cannot necessarily be seen simultaneously from a single location.

j
344 REVIEW OF TORNADO OBSERVATIONS

Plate 4a. Early stage of intensifying tornado near Bennett, Colorado, on August 14. 1977, part of a minioutbreak
studied by Zipser and Golden (19791. ote short condensation funnel, narrow debris cloud and lill of vortex toward the
east.
 BLUESTEIN AND GOLDEN 345

Plate 4b. Same tornado, viewed at close range on opposite side of Plate 4a, 8 min later, tov,ard the SS¾•. Note lack
of visible funnel and trailing dust band. (Courtesy of M. White, Colorado State Patrol, and Ms. Copeland and Ms.
Saylot of Bennett,
346 REVIEW OF TORNADO OBSERVATIONS

Fig. 22. Photographof W. Unruh operatingthe radar and probingthe wind field in a wall cloud west of Morton,
Texas, at 2357 UTC on May 30, 15188.Looking to the northwest. {From Bluestein and Unruh [1989].)

Although the azimuthal resolution of a !idar is excellent, the the "updraft and downdraft mechanism" with the more
range resolution is limited by the pulse, length. The !idar predominant updraft-driven mechanism found in the work of
system must be sensitive enough to yield a strong enough Brady and Szoke [1989] and Wakimoto and Wilaon [1989].
backscattered signal from aerosol and other tiny scatterers The highest wind speed estimates in tornadoes are around
that an aircraft or storm intercept vehicle can be positioned 90 m s-I with some preliminaryestimatesas hi.•h as
at a safe distance. Furthermore, in an airborne system, the 120-130m s-• Technologyhasadvancedto the pointat
location and attitude of the aircraft must be known with high which we can further refine these estimates using radars and
accuracy.
lidars mounted on mobile platforms and using advanced
photogrammetric techniques that make use of more than one
9. SUMMARY camera. The thermodynamic structure of tornadoes may be
Tornadoes have been observed since the last Tornado mapped with sensors aboard RPVs.
Symposium in seasons and locales where once they were New instruments need to be tested on vortex phenomena
considered extremely unlikely (e.g., the destructive Teton- such as dust devils and waterspouts, which are much more
Yellowstone case); indeed, there has been a major shift in common and easy to intercept than tornadoes. The instru-
ments can then be used to make measurements in and near
tornado incidence to states farther west of the Mississippi
River from the decade of the 1970s to the 1980s(F. Ostby, tornadoes' data from them can be integrated with informa-
personal communication, 1990). Many larger tornadoes un- tion from damage surveys and photogrammetric analyses to
dergo a life cycle similar to the waterspout life cycle; yield a clearer picture of the dynamics of the tornado vortex.
however, there are notable differences between the classic Data will provide better boundary conditions for high-
supercell tornadic thunderstorms, which contain mesocy- resolution vortex models.

clones, and the nonmesocyclone tornadic thunderstorms The resolution of current wind speed and minimum pres-
that produce '•landspouts" and •gustnadoes." There is sure controversies will have not only a scientific payoff, but
some difference of opinion among researchers concerning an economic payoff as well. Damage to structures can be
the resemblance of the latter class of tornadoes to the minimized by constructing new buildings that are strong
supercell-spawned variety. For example, we have noted the enough to withstand tornadic, but not any greater, wind
contrast of the Simpson eta[. [1991] waterspout results and
 347

Vz 7 Vn
1712;:
44 Vh

n50 Vn 20

Plate 5. (0) Photograph of Great Bend, Kansas, tornado's dense, turbulent debris cloud with trailing dust band. (b)
Scaled outline of same tornado's dust column with representative net horizontal and vertical velocity vectors (units of
meters per second) superimposed. Velocities derived by photogrammetric tracking and analysis as described by Goldell
IlIld PlIrceJlI1977]. Note accelerations implied vertically and along trailing dust band. from the right.
348 REVIEW OF TORNADO OBSERVATIONS

Plate6. Example oftornado

investigated
byS. A. Colgate
overwestern Oklahoma.
Viewtakenbyhimfi'omaircraft
preparing
to launch
instrumented
rocket.
Notenearby precipitation,
whichimpaired
rocket's
 BLUESTEIN AND GOLDEN 349

thunderstorms,Part II. Field experiment• x•ith TOTO. J. Appl.

Mcteor{•l., • 919-930. 1983a.
Bluestein. H. B.. Measurementsin the vicinity of severe thunder-
stormsand tornadoeswith TOTO: 1982-1983results.in Preprints.
13th Con.It,renee on Severe Local Storms, pp. 89-92, American
Meteorological Societx, Boston, Mas.,,., 1983b.
Bluestein. H. B.. The formation of a "landspout" in a "broken-
line" squall line in Oklahoma. in Preprints', 14th Con.t•'rence
Severe Local Storms. pp. 267-270. American Meteorological
Society. Boston, Mass., 1985.
Bluestein. H. B.. and W. P. Unruh. Observations of the v, ind field
in tornadoes. funnel clouds, and •,•ali cloud.,, with a portable
Doppler radar, Btdl. Am. Meteorol. Sot.. 70, cover, 1514-1525.
1989.
Bluestein, H. B.. and W. P. Unruh, On the measurement of v, ind
speed• in tornadoes with a portable CW/FM-CX• Doppler radar.
in Prepfiats. 25th International Con.[•'renceon Radar %h'teorol-
ogy, pp. 848-851, American Meteorological Society, Boston.
Ma•., 1991.
Bluestein, H. B.. and V•. P. Unruh. On the use of a portable FM-CX•,
Doppler radar for tornado research. this volume.
Bluestein, H. B., and G. R. Woodall, Doppler-radar anals..sisof a
1ou-precipitation severe storm, Mort. Weather Rev., 118, 1640-
1664, 1990.
Bluestein. H. B., E. W. McCaul, Jr.. G. P. Bxrd. and G. R.
Woodall, Mobile sounding observations of a tornadic storm near
the dryline: The Canadian. Texas storm of 7 May 1986. Mort.
Weather Rev.. l l6, 1790-1804. 1988.
Bluestein. H. B.. E. W. McCaui, Jr.. G. P. B5rd. G. R. Woodall. G.
Martin, S. Keighton. and L. Showeli. Mobile ,unding ob,rva-
tions of a thunderstorm near the dryline: The Gruver, Texas storm
complex of 25 May 1987, Mou. Weather Rev., 117, 244-250, 1989.
Bluestein, H. B., E. W. McCaui, Jr., G. P. B5 rd, and R. L. Waiko,
Thermodynamic measurements under a wall cloud.
Weather Rev., 11(5', 794-799, ! 990a.
Bluestein, H. B., E. W. McCaui, Jr., G. P. Byrd, R. L. Walko, and
R. Davies-Jones, An observational stud5 of splitting convective
clouds, Mort. Weather Rev., 118, 1359-1370, 1990b.
Brady, R. H.. and E. J. Szoke, A case study of nonmesocyclone
tornado development in northeast Colorado: Similaritie_,,to wa-
terspout formation, 3Ion. Weather Rev., 117, 843-856, 1989.
Fig. 23. Photograph of University of Oklahoma graduate •tudents
BrantIcy, J. Q., Some weather observations x•ith a continuous-v•ave
S. Hrebenach, G. Martin. and S. Contorno, attemptingto probe a
Doppler radar, in Proceedings of the Oth Weather Radar ConiC'r-
tornado to the west at 2329 UTC on May 13, 1989, west of Hodges,
ent'c, pp. 297-306, American Meteorological Society, Blue Hill
Texa.,,. (Photograph copyright by H. Bluestein.)
Meteorological Ob.,,ervatorx, Geoph.xsics Research Directorate of
the Air Force Cambridge Research Center..Massachusetts Insti-
tute of Technologs, Boston and Cambridge, Mass., 1957.
Brock, F. V., G. Lesins. and R. L. Waiko. Mea.,,urementof pre•,re
Ackm•wledgment. Part of this paper was supportedby National and air temperature near severe thunderstorms: An inexpensive
Science Foundation grant ATM-8902594. and portable instrument. in lz:rtendedAbstrat'ts. 6th Symposiutn
on Meteorological Observations and Inslrumcntation, pp. 320-
323, American Meteorological Sotlets, Boston, Mass., 1987.
REFFRENCE3 Burgosh, D. W.. and L. R. Lemon, Severe thunderstorm detection
b5 radar, in Radar in ,llete•m•logy, edited by D. Atlas, pp.
Agee, E. M., J. T. Snow, F. S. Nickerson, P. R. Clare, C. R. 619-647. ,\meric:m Meteorological Societs. Boston. Mass., 1990.
Church, and L. A. Schaai, An observational stud> of the Best Barges,. D. W.. S. \. Vasiloff, R. P. Davieh-Jone,, D. S. Zrnic. and
Lafayette, Indiana, tornado of 20 March 1976, Mort. I•eathcr S. E. Frederickson, Recent NSSL work on windspeed measure-
Rev., 105,893-907, 1977. ment in tornadoes. in Pr(•cet'dingx. 5th U.S. •¾ationalConj•'rcnt'e
Barnes, S. L., Oklahoma thunderstorm• on 29-30 April 1970, Part I, on i,•ind Engittcering, pp. I.k-53-1A-o0, Texas Tech Universits',
Morphology of a tornadic storm, Mort. Weather Rev., 106, Lubbock, 1985.
673-684, 1978. Campbell. B. D., E. N. Rasmus.,,en,and R. E. Peterson, Kinematic
Bedard, A. J.. Jr.. and C. Ramzy. Surface meteorologicalobserva- analysis of the Lake,• ie•,, TX tornado, in Preprints. 13th C•mfi'r-
tions in se•,erethunderstorms.Part 1, Designdetailsof TOTO, J. et!ct' ½}1•
Severt' L(•t'½tlStorms, pp. 62-65. American Meteorolog-
Appl. Meteorol., 22, 911-918, 1983. ical Society, Boston, Mass., 1983.
Blechman, J. B., The Wi.,,consintornado event of April 21, 1974: Church, C. R., C. M. Ehresman, and J. H. Golden, Instrumentation
Observationsand theory of secondaryvortices,in Preprints,9th for probing waterspt.•ut.,,,in Preprints, 8th Conference on Severe
Conferenceon Severe L•wai Storms. pp. 267-270, American Local Storms, pp. 1o9-172, American Meteorological Society,
MeteorølogicalSociety, Boston. Mass., 1975. Boston. Mass.. 1973.
Bluestein, H. B., Surface meteorological observations in .,,½•,ere Colgate, S., Comment on "On the relation of electrical activit3
350 REVIEW OF TORNADO OBSERVATIONS

tornadoes"by R. P. Davies-Jonesand J. H. Golden,J. Geophys. Typhoon-associated tornadoes in Japan and new evidence of
Res., 80, 4556, 1975. suctionvorticesin a tornadonear Tokyo, J. Meteorol. Soc. Jpn.,
Colgate,S., Smallrockettornadoprobe,in Preprints,12thConfer- 50, 431-453, 1972.
ence on Severe Local Storms, pp. 396-400, American Meteoro- Fujita, T. T., A.D. Pearson, G. S. Forbes, T. A. Umenhofer, E. W.
logical Society, Boston, Mass., 1982. Pearl, and J. J. Tecson, Photogrammetric analyses of tornadoes,
Davies-Jones,R. P., Tornado dynamics, in Thunderstorms--A in Proceedings, Symposium on Tornadoes, edited by R. E.
Social, Scientificand TechnologicalDocumentary, vol. 2, Thun- Peterson, pp. 43-88, Institute for Disaster Research, Texas Tech
derstorm Morphology and Dynamics, 3rd ed., edited by E. University, Lubbock, 1976.
Kessler, 412 pp., University of Oklahoma Press, Norman, 1988. Golden, J. H., Waterspouts and tornadoes over south Florida, Mort.
Davies-Jones, R. P., and J. H. Golden, On the relation of electrical Weather Rev., 99, 146-154, 1971.
activity to tornadoes,J. Geophys. Res., 80, 1614--1616,1975a. Golden, J. H., Some statistical aspects of waterspout formation,
Davies-Jones,R. P., and J. H. Golden, Reply, J. Geophys.Res., 80, Weatherwise, 26, 108-117, 1973.
4557-4558, 1975b. Golden, J. H., The life-cycle of Florida Keys' waterspouts, I, J.
Davies-Jones,R. P., and J. H. Golden, Reply, J. Geophys.Res., 80, Appl. Meteorol., 13, 676-692, 1974a.
4561-4562, 1975c. Golden, J. H., The life-cycle of Florida Keys' waterspouts, II, J.
Davies-Jones, R. P., and E. Kessler, Tornadoes, in Weather and Appl. Meteorol., 13,693-709, 1974b.
Climate Modification, edited by W. N. Hess, pp. 552-595, John Golden, J. H., Life-cycle of Florida Keys' waterspouts, NOAA
Wiley, New York, 1974. Tech. Memo., ERL-NSSL-70, 147 pp., 1974c.
Davies-Jones, R. P., D. W. Burgess, L. R. Lemon, and D. Purcell, Golden, J. H., An assessmentof wind speeds in tornadoes, in
Interpretation of surface marks and debris patterns from the 24 Proceedings, Symposium on Tornadoes, edited by R. E. Peter-
May 1973 Union City, Oklahoma, tornado, Mon. Weather Rev., son, pp. 5-42, Institute for Disaster Research, Texas Tech Uni-
106, 12-21, 1978. versity, Lubbock, 1976.
Dessens, J., Influence of ground roughnesson tornadoes: A labo- Golden, J. H., and B. J. Morgan, The NSSL/Notre Dame tornado
ratory simulation, J. Appl. Meteorol., 11, 72-75, 1972. intercept program, spring 1972, Bull. Am. Meteorol. Soc., 53,
Dietachmeyer, G., Numerical modelling of intense atmospheric 1178-1179, 1972.
vortices, Ph.D. dissertation, Dep. of Math., Monash Univ., Golden, J. H., and D. Purcell, Photogrammetric velocities for the
Clayton, Victoria, Australia, 1987. Great Bend, Kansas tornado of 30 August 1974:Accelerations and
Doswell, C. A., and D. W. Burgess,On someissuesof U.S. tornado asymmetries, Mort. Weather Rev., 105,485-492, 1977.
climatology, Mort. Weather Rev., 116,495-501, 1988. Golden, J. H., and D. Purcell, Airflow characteristics around the
Doviak, R. J., and D. S. Zrnic, Doppler Radar and Weather Union City tornado, Mot,. Weather Rev., 106, 22-28, 1978a.
Observations, 458 pp., Academic, San Diego, Calif., 1984. Golden, J. H., and D. Purcell, Life-cycle of the Union City, OK
Elsom, D. M., and G. T. Meaden, Suppression and dissipation of tornado and comparison with waterspouts, Mon. Weather Rev.,
weak tornadoes in metropolitan areas: A case study of greater 106, 3-11, 1978b.
London, Mon. Weather Rev., 110, 745-756, 1982. Golden, J. H., and M. E. Sabones, Tornadic-waterspout formation
Flora, S. R., Tornadoes of the United States, 194 pp., University of near intersecting boundaries, in Preprints, 25th International
Oklahoma Press, Norman, 1953. Conference on Radar Meteorology, pp. 420-423, American Me-
Forbes, G. S., Three scales of motion associated with tornadoes, teorologicalSociety, Boston, Mass., 1991.
NUREG/CR-0363, final report, 359 pp., U.S. Nucl. Reg. Comm., Golden, J. H., et al., Severe storm detection: Panel report, in Radar
Washington, D.C., 1978. (Available as NTIS PB-288 291 from in Meteorology, edited by D. Atlas, pp. 648-656, American
Natl. Tech. Inf. Serv., Springfield,Va.) Meteorological Society, Boston, Mass., 1990.
Forbes, G. S., Observations of relationships between tornado Goodman, S. J., and K, R. Knupp, Tornadogenesisvia squall line
structure, underlying surface, and tornado-appearance, in Pre- and supercell interaction revisited: The 15 November, 1989
prints, 11th Conference on Severe Local Storms, pp. 351-356, Huntsville tornado, in Preprints, 16th Conference on Severe
American Meteorological Society, Boston, Mass., 1979. Local Storms, pp. 566-571, American Meteorological Society,
Forbes, G. S., and R. M. Wakimoto, A concentrated outbreak of Boston, Mass., 1990.
tornadoes, downbursts, and microbursts, and implications regard- Gordon, A. H., Waterspouts, Weather, 6, 364-371, 1951.
ing vortex classification,Mort. Weather Rev., 111,220-235, 1983. Grant, F. C., Proposed technique for launching instrumented bal-
Fujita, T. T., Mother cloud of the Fargo tornadoesof 20 June 1957, loons into tornadoes, NASA Tech. Note, TN D-6503, 21 pp., 1971.
in CumulusDynatnics, pp. 175-177, Pergamon,New York, 1960. Grazulis, T. P., J. T. Schaefer, and R. F. Abbey, Jr., Advances in
Fujita, T. T., The Lubbock tornadoes: A study of suction spots, tornado climatology, hazards, and risk assessmentsince Tornado
Weatherwise, 23, 160-173, 1970. Symposium II, this volume.
Fujita, T. T., Proposedmechanismof suctionspotsaccompaniedby Hess, G. D., and K. T. Spillane, Waterspouts in the Gulf of
tornadoes, in Preprints, 7th Conference on Severe Local Storms, Carpentaria, Aust. Meteorol. Mag., 38, 173-179, 1990.
pp. 208-213, American Meteorological Society, Boston, Mass., Hobbs, P. V., N. T. Funk, R. R. Weiss, St., J. D. LocateIll, and
1971. K. R. Biswas, Evaluation of a 35 GHz radar for cloud physics
Fujita, T. T., Jumbo outbreak of 3 April 1974, Weatherwise,27, research, J. Atrnos. Oceanic Technol., 2, 35-48, 1985.
116-126, 1974. Holle, R. L., and M. W. Maier, Tornado formation from downdraft
Fujita, T. T., Tornadoesand downburstin the contextof general- interaction in the FACE mesonetwork, Mort. Weather Rev., I08,
ized planetaryscales,J. Atmos. Sci., 38, 1511-1534,1981. 991-1009, 1980.
Fujita, T. T., The Teton-Yellowstonetornadoof 21 July 1987,Mon. Howells, P., and R. K. Smith, Numerical simulationof tornado-like
Weather Rev., 117, 1913-1940, 1989. vortices, Part I, Vortex evolution, Geophys. Astrophys. Fluid
Fujita, T. T., and A.D. Pearson,Resultsof FPP classificationof Dyn., 27, 253-284, 1983.
1971 and 1972 tornadoes, in Preprints, 8th Conferenceon Severe Hurd, W. E., Somephasesof waterspoutbehavior, Weatherwise,3,
Local Storms, pp. 142-!45, American Meteorological Society, 75-82, 1950.
Boston, Mass., 1973. Lauritsen, D., Z. Malekmadani, C. Morel, and R. McBeth, The
Fujita, T. T., D. L. Bradbury, and C. F. Van Thullenar, Palm Cross-ChainLORAN AtmosphericSoundingSystem (CLASS),
Sunday tornadoesof April 11, 1965, Mon. Weather Rev., 98, in Preprints,6th Symposiumon MeteorologicalObservationsand
29-69, 1970. Instrumentation,pp. 340-343, American MeteorologicalSociety,
Fujita, T. T., K. Watanabe, K. Tsuchiya, and M. Schimada, Boston, Mass.,
 BLUESTEIN AND GOLDEN 351

Lee, J. T., D. S. Zrnie, R. P. Davies-Jones,

and J. H. Golden, (.-mSevere Local Storms, pp. J31-J35, American Meteorological
Summaryof AEC-ERDA-NRC supportedresearchat NSSL Society, Boston, Mass., 1990.
1973-1979, NOAA Tech.Memo.,ERL NSSL-90,93pp., 1981. Rust, W. D., and T. C. Marshall, Mobile, high-wind, balloon-
Leverson,V. H., P. C. Sinclair,andJ. H. Golden,Waterspout launching apparatus, J. Atmos. Oceattic Techt•o[., 6, 215-217,
wind, temperatureand pressurestructurededucedfi'om aircraft 1989.
measurements, Mort. Weather Rev., 105, 725-733, 1977. Rust, W. D., R. Davies-Jones, D. W. Burgess,R. A. Maddox, L. C.
Lhermitte, R., A 94-GHz Dopplerradarfor cloudobservations,
J. Showell, T. C. Marshall, and D. K. Lauritsen, Testing of a mobile
Atmos. Oceanic Techno!., 4, 36-48. 1987. version of a cross-chain LORAN atmospheric soundingsystem
MacGorman, D. R., Lightning in tornadic storms:A review. this (M-CLASSI, Bull. Am. Meteorol. Sot., 71, 173-180, 1990.
volume.
Sch,,•iesow, R. L., Horizontal velocity structure in waterspouts,J.
MacGorman,
D. R., andK. E. Nielsen,Cloud-to-ground
lightning
in Appl. Meteoro!., 20, 349-360, 1981.
a tornadic storm on 8 May 1986, Mon. Weather Rev., 119, Schwiesow, R. L., The NCAR airborne intYared lidar system
1557-1574, 1991. (NAILS) design and operation, NCAR Tech. Note NCAR/TN-
Macky, W. A., The Eastertornadoesat Bermuda,Wcatt,erwise,6, 29I+L4, 38 pp., Atmos. Tech. Div. Natl. Center for Atmos. Res.,
7475, 1953. Boulder, Colo., 1987.
Maier, M. W., and H. W. Brandli, Simultaneous observations of a Schwiesow, R. L., R. E. Cupp, P. C. Sinclair, and R. F. Abbey, Jr.,
tornado, waterspout, and funnel cloud, Weather, 28, 322-327, Waterspout velocity measurements by airborne Doppler lidar, J.
1973. Appl. Meteorol., 20, 341-348, 1981.
Marshall,T. P., Damageanalysisof theMesquite,Texastornado,in Schwiesow, R. L., A. J. Weinheimer, and V. M. Glover, Applica-
Proceedings,5th U.S. National Conferenceon Wind Engineer- tions and development of a compact, airborne Doppler lidar for
ing, pp. 4B-11-4B-18, Texas Tech University, Lubbock, 1985. atmospheric measurements, in Proceedings, 5th Cot½•,renceon
Marshall, T. P., J. R. McDonald, and K. C. Mehta, Utilization of Coherent Laser Radar: Technology and Applications, pp. 20-24,
load and resistancestatisticsin a wind speedassessment,
Rep. Society of Photo Optical InstrumentationEngineers,Bellingham,
670, 91 pp., Inst. for Disaster Res., Tex. Tech Univ., Lubbock, Wash., 1989.
1983.
Simpson,J., M. C. McCumber, and J. H. Golden, Tropical water-
Marshall, T. P., J. R. McDonald, and K. C. Mehta, Performance of spouts and tornadoes, in Preprints, 14th Conference on Severe
structures during Hurricane Alicia and the Athens tornado, in Local Storms, pp. 284-288, American Meteorological Society,
Proceedings,DOE Natural PhenomenaHazards Mitigation Con- Boston, Mass., 1985.
ference, pp. 140-149, U.S. Departmentof Energy,Washington, Simpson, J., B. R. Morton, M. C. McCumber, and R. S. Penc,
D.C., 1985.
Observationsand mechanismsof GATE waterspouts,J. Atmos.
McCaul, E., Observations of the Hurricane •'Danny" tornado Sci., 43, 753-782, 1986.
outbreak of 16 August 1985, Mon. Weather Rev., 115, 1206-1223,
1987.
Simpson,J., G. Roff, B. R. Morton, K. Labas, G. Dietachmeyer, M.
McCumber, and R. Penc, The Great Salt Lake waterspout, Mort.
McCaul, E. W., Jr., H. B. Bluestein, and R. J. Doviak, Airborne
Weather Rev., 119, 2741-2770, 1991.
Doppler lidar observations of convective phenomena in Okla-
homa, J. Atmos. Oceanic Technol., 4,479-497, 1987.
Smith, R. L., and D. W. Holmes, Use of Doppler radar in meteo-
rological observations, Mon. Weather Rev., 89, 1-7, 1961.
Minor, J. E., J. R. McDonald, and K. C. Mehta, The tornado: An
Strauch, R. G., Theory and application of the FM-CW Doppler
engineering-oriented perspective, NSSL Tech. Memo. ERL-
NSSL-82, Natl. Severe Storms Lab., Natl. Oceanic and Atmos. radar, Ph.D. thesis, 97 pp., Dep. of Electr. Eng., Univ. of Colo.,
Boulder, 1976.
Admin., Norman, Okla., 1977.
Pasqualucci, F., B. W. Bartram, R. A. Kropfii, and W. R. Mon- Szoke, E. J., and R. Rotunno, A comparison of surface observations
and visual tornado characteristics for the June 15, 1988, Denver
inger, A millimeter-wavelength dual-polarization Doppler radar
tornado outbreak, this volume.
for cloud and precipitation studies,J. Clim. Appl. Meteorol., 22,
758-765, 1983. Szoke, E. J., M. L. Weisman, J. M. Brown, F. Caracena, and T. W.
Pauley, R. L., and J. T. Snow, On the kinematicsand dynamicsof Schlatter, A subsynopticanalysis or'the Denver tornado of 3 June
the 18 July 1986 Minneapolis tornado, Mon. Weather Rev., 116, 1981, Mrm. Weather Rev., 112, 790-808, 1984.
2731-2736, 1988. Uttal, T., B. E. Martner, B. W. Orr, and R. M. Wakimoto, High
Prosset, N. E., Aerial photographsof a tornado path in Nebraska, resolution dual-Doppler radar measurements of a tornado, in
May 5, 1964, Mon. Weather Rev., 92, 593-598, 1964. Preprints, 24th Confi'rence on Radar Meteorology, pp. 62-65,
Rasmussen,E. N., The Tulia outbreak storm: Mesoscaleevolution American Meteorological Society, Boston, Mass., 1989.
and photogrammaticanalysis, M.S. thesis, 180 pp., Dep. of Van Tassel, E. L., The North Platte Valley tornado outbreak of
Geosci., Tex. Tech Univ., Lubbock, 1982. June 27, 1955, Mort. Weather Rev., 83, 255-264, 1955.
Rasmussen, E. N., and R. E. Peterson, Tornado photogrammetry: Vonnegut, B., Electrical theory of tornadoes, J. Geophys. Res., 65,
Determinationof windspeedsand assessment of errors,report, 56 '•3 '•1 '• 1960.
pp., Inst. for DisasterRes., Tex. Tech Univ., Lubbock,1982. Wakimoto, R. M. , and B. E. Martner, Photogrammetric,•radar
Roberts, R. D., and J. W. Wilson, Multiple Doppler radar analyses analysis of the 2 July tornado during CINDE, in Preprints, 24th
of the 15 June 1988Denver tornadoes,in Preprints, 24th Con.l'•'r- Cot½•renceon Radar Meteorology• pp. 58-61, American Meteo-
ence on Radar Meteorology, pp. 142-145, AmericanMeteorolog- rological Society, Boston, Mass., 1989.
ical Society, Boston, Mass., 1989. Wakimoto, R. M., and J. W. Wilson, Non-supercell tornadoes,
Rossow,V. J., On the electrical natureof waterspouts,in Preprints, Mon. Weather Rev., 117, 1113-1140, 1989.
6th Conferenceon SevereLocal Storms,pp. 182-187.American Ward, N. B., The Newton, Kansas tornado cyclone of May 24,
MeteorologicalSociety, Boston, Mass., 1969. 1962, in Preprints, 11th Weather Radar Confi,rence, pp. 410-4!5,
Rossow,V. J., Observationsof waterspoutsandtheir parentcloud•, American Meteorological Society, Boston, Mass., 1964.
NASA Tech. Note, D-5854, 63 pp., 1970. Wilson, J. W., Tornadogenesisby non-precipitation-inducedwind
Rothermel,J., C. Kessinger,and D. L. Davis, Dual-Dopplerlidar shear lines, Mon. Weather Rev., 114, 270-284, 1986.
measurement of winds in the JAWS experiment, J. Atmos. Wilson, J. W., and R. D. Roberts, Vorticity evolution of a non-
Oceanic Technol., 2, 138-147, 1985. supercelltornado on 15 June 1988near Denver, in Preprints, 16th
Rust, W. D., Severe storm electricity researchwith a mobile Con.t•'rence (m Severe Local Storms, pp. 479-484, American
laboratoryandmobileballooning, in Preprints,16thCon./i,
rence Meteorological Society, Boston, Mass.,
352 REVIEW OF TORNADO OBSERVATIONS

Zipser, R. A., Photogrammetricstudies of a Kansas tornado and a Zrnic, D. S., and M. Istok, Wind speedsin two tornadic stormsand
Hawaiian tornadic-waterspout,
M.S. thesis, 75 pp., Dep. of a tornado, deduced from Doppler spectra, J. Appl. Meteorol., 19,
Meteorol., Univ. of Okla., Norman, 1976. 1405-1415, 1980.
Zipser, E. J., and J. H. Golden, A summertime tornado outbreak in Zrnic, D. S., R. J. Doviak, and D. W. Burgess, Probing tornadoes
Colorado: Mesoscale environment and structuralfeatures, Mort. with a pulseDoppler radar, Q. J. R. Meteorol. Soc., 103,707-720,
Weather Rev., 108, 1328-1342, 1979. 1977.
Zrnic, D. S., andR. J. Doviak, Velocity spectraof vorticesscanned Zrnic, D. S., D. W. Burgess, and L. Hennington, Doppler spectra
with a pulse-Dopplerradar, J. Appl. Meteorol., 14, 1531-1539, and estimated windspeed of a violent tornado, J. Clim. Appl.
1975. Meteorol., 24, 1068-1081,
 A Comparisonof SurfaceObservationsand Visual Tornado Characteristics
for the June 15, 1988, Denver Tornado Outbreak
E. J. SZOKE

ForecastSystems
Laborator3,,
NOAA, andNationalCenterfor Attnospheric
Research,Botdder,Colorado80307

R. ROTUNNO

National Centerfor AtmosphericResearch,Boulder,Colorado80307

1. INTRODUCTION work of up to 22 automatedsurfacestations(the current

configuration
is shownin Figure 1). The networkwas put
On June 15, 1988, four tornadoes developed in and near intoplacein 1980by the Programfor RegionalObserving
Denver, Colorado, alongthe Denver Convergence-Vorticity andForecasting Services(PROFS) of the NationalOceanic
Zone (DCVZ) (.orDenver Cyclone [Szokeet al., 1984]),after and AtmosphericAdministration(NOAA) (now part of
its intersection with two thunderstorm outflow boundaries. NOAA's ForecastSystemsLaboratory(FSL)) to test the
Althoughthe tornadoes
formedin proximityto eachotherin effectsuchdatamighthave on improvingshort-range fore-
time and space, the two observed by the authors were casts.The mesonetis noteworthy in that it has provided a
distinctlydifferentin appearance.
One wasquitetypicalof long-term setofdenseobservations.
Additionalobservations
tornadoes in this area, with the vortex mainly defined by fromDopplerradar(s),storminterceptteams,specialsound-
dust,whereasthe otherhad an unusuallylargecondensation ings,aircraft,andfiner-scale
mesonets havebeenobtained
funnel.In thispaperthe questionof why thetornadoes were duringvariousPROFSforecastexercises anda numberof
so visibly differentis examinedusing two networksof experiments including
the JointAirportWeatherStudies
nearbyautomatedsurfacestations.Our hypothesis is that (JAWS)[McCarthyet al., 1983]in 1982andthe Convective
the visual differenceswere primarily the result of local Initiation and Downburst Experiment (CINDE) in 1987
variations in low-level moisture. Before examiningthe June [Wilson et al., 1988].
15case,however,in keepingwith the themeof theTornado One of the first discoveries noted when the PROFS
Symposium, tornadoobservational
andresearch highlights mesonetdata became available was the formation of a zone
from northeasternColoradoduringthe past decadewill be of convergence, the DCVZ (see Figure 1), that typically
summarized. developednearthe Denverarea whenthe ambientflow
farthereaston the plainswas from the southor southeast.
There had been speculationthat such a featureexisted
2. REVIEW OF COLORADOTORNADOOBSERVATIONS
AND beforethe adventof the mesonetobservations.For example,
RESEARCHDURING THE PAST DECADE ZipserandGolden[ 1979]notedmesoscaleconvergence
and
cyclonicvorticitybasedon SurfaceAviationObservation
In this brief review we focus on observationalresearch,
(SAO) stationsDenver(DEN) and Limon (LiC) beforean
concentrating
oneastern Colorado. A numberoffieldexper- August1977tornadooutbreakeastof Denver.Noteworthy
imentsconductedin northeasternColoradoin the 1980shave
among theirobservations wasthatthestormsproducing the
provided thedatato increaseourunderstanding of convec- tornadoeswere of the ordinarycell type and did not display
tionin generalandtornadogenesis in particular. thehookecho(fromthe WSR-57 10-cmwavelengthNational
Onedatasetthathasprovidedvaluableinformation about WeatherService(NWS) radarat LIC) characteristic of more
the environmentof stormsis that generatedby a mesonet-
organizedsupercellularstorms.Thesestorms weregrowing
"explosively,"withnewdevelopment onthewestern edges,
The Tornado:Its Structure,Dynamics,Prediction,andHazards.
GeophysicalMonograph79
justbeforethetornadoes
wereobserved.
Thetornado
for-
Copyright
1993by theAmerican
Geophysical
Union. mationduringthe rapid verticaldevelopment
phaseof the
354 TORNADO CHARACTERISTICS

LAR
© Cheyenne
Ridge
2100 ß / SNY
Nebraska
41øN

• 1800n•n
150(
bgd
gly
ft

40øN
i
/aur bye
i 0

elb I
LIC

39øN (• ImerRidge 500

cos

3000 ©PUB I
0
.....
!
50km

LHX

38øN
105ow 104øW 103øW
Fig. 1. Elevation contours (meters) of northeastern Colorado, shaded above 3000 m, with the NOAA/FSL mesonet
stations(open circles), SAO stations(closedcircles), schematicflow for the DCVZ (with "mean" location shown by
the dashed line), analysis areas for Figures 10a-10f (two solid boxes), and area in Figure 4 (dashed box).

parent thunderstorm was also noted by Golden [1978] for the sets, Brady and Szoke [1989] and Wakimoto and Wilson
May 30, 1976, tornado just east of Denver. [1989] documentedtornado development and proposed sim-
The importance of the DCVZ for the initiation of convec- ilar tornado mechanisms. Brady and Szoke studied the July
tion and its possible relationship to tornadic development 26, 1985, tornado that developed near Erie ("eri" in Figure
was documented by Szoke et aI. [1984] in their study of the 1), Colorado, within 20 km of a Doppler radar. They found
June 3, 1981, Denver area tornadoes, which caused about 15 thatlow-levelcirculations
of >5 m s-• differentialvelocity
million dollars in damage. They speculated that the Denver over an -•500-m diameter were evident along the DCVZ
area thunderstorm became tornadic at least in part because before any clouds had formed along the zone. Extremely
of low-level vertical vorticity along the DCVZ. rapid thunderstorm development occurred along the DCVZ
Subsequent tornado events have been observed with after a thunderstorm outflow boundary from weak thunder-
Doppler radar and other supplementaryinstruments,allow- storms over the foothills to the west collided with the DCVZ.
ing for a mo]:e complete picture of the tornadogenesis Doppler data showed that the tornado formed in association
processfor these types of storms. Using independentdata with an intensified low-level circulation, presumably
 SZOKE AND ROTUNNO 355

to stretching of air columns in the vertical beneath the t 0 rain

vigorous convection that developedafter the intersection.

The tornado began before any precipitation had fallen from
the storm, and no midlevel rotation (down to the maximum 4 kin--

radar observable scale for their case, an ---300-mby 350-m

radar gate) in the cloud preceded the tornado. Brady and 3 km• No clouds

Szoke made a detailed comparisonwith the life cycle of a

waterspout, as documented by Golden [ 1974a, b]. Because
of the similarities in the formation mechanismsthey used the
name "landspout" to describe this type of tornado. A
Outflow• •/• Convergence
zone
schematicfrom Brady and Szoke of the typical life cycle of
the landspoutis shownin Figure 2. The landspoutterminol- lkm--boundary
•&/
Upward
motion
Weak,
low-level
c•rculat•on
at•ntersect•on
ogy was first used by Bluestein [1985] to document an
atypicalOklahomatornadothat formed underconditionsof
weak vertical wind shear. Bluestein observed the landspout t •--15 min

under "a line of rapidly growingtowers alonga cold front."

Wakimoto and Wilson [1989] used data collected in 1987 New convective development

during CINDE to documentthe single-Dopplerradar and 4 km--

visual characteristics of 27 vortices, many of which were

tornadoes.Their schematicfor the nonsupercelltornadois
shown in Figure 3. Thus two independentstudiesdocu-
mentedtornadogenesis in associationwith the stretchingof
3 km--

ß I
low-levelverticallyorientedcyclonicvorticitylocatedalong 2 km•

a boundary(or intersectingboundaries)beneatha rapidly

•'-•
C•rculat•on
advected
vertical
developing
parentcloud.In bothstudiesa midlevelmesocy- 1 km--

clonewas not presentbeforetornadoformation.

Evidence that tornadoesdevelop elsewherefrom storms
without mesocyclones is found not only from Bluestein's
observations,but also from the Oklahomaobservations of
Burgessand Donaldson[1979].Additionally,Brady and
Szokespeculated
thatareaslikeFloridamightbeespecially
susceptible
to nonmesocycloneorlandspout
tornadogenesis,
because of the common occurrence of suitable low-level
boundaries(the sea breeze and Lake Okeechobeebreeze
fronts).
t -- 30 min
A vorticityanalysis
usingmultiple-Doppler
dataof oneof
the CINDE casesby Wilczaket al. [1992] suggests
a some-
what more complexpicturefor the July2, 1987,tornado.
Theyfoundthatthe tornadodeveloped frombothvertical 4 krn •

stretchingandthetiltingof environmental horizontal

vortic-
ity afterthe collision
of a thunderstorm outflowboundary 3 km --

witha stationary boundary. The environmental horizontal

vorticitywas apparently enhanced in the vicinityof the 2 km-- Strongconvect,
veupdrafts
stretch
and
collision.Thisstormappears similarto twoof thetornadoes •ntenstfyvortex •ntoF1 tornado
in theMay 1984casestudied by Wilson[1986],wherethe
tornadoes wereproduced frommoreorganized stormsthan 1 krn --

thoseassociatedwith the nonmesocyclone

type tornadoes
notedearlier,but with the boundaries
neededfor torna-
dogenesis
alreadypresent, ratherthanproduced by the
stormitself (as in the typicalevolutionof a supercellFig. 2. Schematic
[ 19891.
of thetypicallandspout,
fromBradyandSzoke
tornado).
Wilsonalsostudied
threesmaller
tornadoes
thatoccurred
during
theMay1984case,of thetypeoftenreferred
to as
"gustnadoes."
His studymaybethefirstDoppler radar of tornadothat forms alongthe gustfront from a thunder-
storm and is removed from the parent storm and any
documentation
of the gustnado
phenomenon,
whichis a
relatively
common
typeof weaktornado in associatedmesocyclone,if present.Forbesand Wakimoto
thatoccurs
Coloradoandelsewhere. refersto a type [1983]discuss
The termgustnado therangeof vorticesthattheybelievefit
356 TORNADO CHARACTERISTICS

Fig. 3. Schematicof tornadogenesis

from nonsupercellstorms,from Wakimotoand Wilson [1989].

classification
of a tornadoandsuggestincludingthegustnado two tornadoes that touched down in Denver itself were
becauseit is associatedwith a thunderstorm.The stretching classifiedas F2, with one having a small area of F3 damage
mechanismresponsiblefor the landspout or nonmesocy- (Table 1).
clone tornado also occurs with the gustnado, but the lack of The day was characterizedby a DCVZ, and althoughthe
a vigorousupdraftover the low-level vortex (in many cases) tornadoes eventually developed in the area where the con-
preventsthe gustnadofrom becomingvery strongand could vergence zone was located, they formed only after two
be considereda distinguishingfactor when comparedwith thunderstorm outflow boundaries intersected at the location
the landspoutor nonsupercelltornado. Whatever the termi- of the DCVZ. A detailedmultiple-Dopplerstudyof this case
nology used, the important point is that these tornadoesare is under way [Roberts and Wilson, 1989; Wilson and Rob-
distinguishedfrom supercell tornadoes by the lack of a erts, 1990], and preliminary results indicate that the storms
mesocyclonein the parent thunderstorm. did not possessa midlevel mesocyclonebefore the tornadoes
The discussionabove has emphasizedthe occurrenceof formed. Instead, the rotation associated with the tornadoes
tornadoes from storms that do not possessthe typical beganat the surface,intensified,and built upward with time
characteristicsof a tornadicsupercell.Althoughthesetypes beneath the storm updrafts, similar to the casesdocumented
of tornadoes account for many of those observed in Colo- earlier. In light of this Doppler study and its implicationsfor
rado, especiallycloser to the Denver area, long-livedsuper- tornadogenesiswith the circulation intensifying from the
cell storms do occur in Colorado, as documented for the surface upward, our assumption is that the condensation
1972 Fleming storm [Browning and Foote, 1976] and the funnels are composedof low-level air rising upward. Thus
Limon tornado of June 6, 1990. No detailed study has been we feel that surface-based values of moisture are most
done to indicate the frequency of such supercelIsin Colo- relevant for this case when discussingthe funnel character-
rado or the percentage that produce tornadoes. The above istics.
studiessuggestthat a variety of tornadic stormtypes exist in In the remainder of this section we concentrate on the
Colorado, from the supercellto the nonsupercellor land- visual contrast between the tornadoes and examine mesonet
spout type, with a type of "hybrid" storm between these data to discernthe reason for the variation in appearance.
extremes that might possessa midlevel mesocyclone but
would not develop a tornado without the proper preexisting Visual Observations
boundary configuration and interaction.
When the authors arrived in Denver (at location L1 in
Figure 4) at about 2200 UTC (1600 MDT; all subsequent
3. THE JUNE 15, 1988, TORNADOES
times will be UTC), a north-northeast to south-southwest
The four tornadoes that developed in and near the Denver expanse of relatively uniform cloud basesextended to either
area on June 15, 1988, were the most damaging in Denver side of our location. Initially, a few large raindrops fell at
since the June 3, 1981, tornadoes [Szoke et al., 1984]. The location L1, but these soon ended, and we saw the first

TABLE 1. Summary of the June 15, 1988, Denver Area Tornadoes

Time,
Tornado UTC Fujita Rating Damage Visual Characteristics

T1 2205-2213 F1 open terrain dust column, tiny condensation funnel

T2 2215-2236 F2 old neighborhood with tall trees huge condensationfunnel
T3 2218-2235 F2-F3 commercial area moderate size condensation funnel
T4 2220-2233 F1 open terrain huge dust
 SZOKE AND ROTUNNO 357

in Figure 7. Although the condensation funnel associated

_CP•eri
©] ©bri
/
with T3 is much larger than that associated with T1, it is
considerablysmallerthan T2's funnel. Occasionally,a very
narrow additional condensation funnel was visible in a
portionof the area beneaththe mainfunnel, faintly visiblein
the original photograph. A debris cloud is apparentin the
video of this tornado as it passedover a commercialarea of
southDenver, causingF2 and someF3 damage.In the video
for T2 and T3, it is clear that the rotation in the condensation
funnel of T3 is considerably more rapid than the slowly

(3rv ]
turning rotation in the larger funnel with T2 (at a compara-
tive vertical location midway between the groundand cloud
base).
/ / STPL. Finally, a fourth tornado (T4, Figure 8) formed minutes

L3x•
T2/•'aur
after T2 and T3 had developed, just to the southeastof the
lak
ß
town of Brighton ("bri"). It touched down in an open field
and was marked by an impressive cloud of dust that ex-
tended to cloud base. Little if any condensation cloud was
evident with T4, so visually it resembled T1.
In summary, the appearance of the two northern torna-
0 5 km does (T1 and T4) was similar to the "typical" nonmesocy-
I•1
clone tornado of this area; they had very small (if any)
condensationfunnels and were made visible by dust clouds.
In contrast, T2 had one of the largest condensation funnels
observed in this area, whereas the southernmost tornado,
Fig. 4. Map of the Denver area showingmain highways,locations T3, had a somewhat smaller condensation funnel.
of tornadoes(T1-T4), photographerlocations(L1-L3), and mesonet
stations (large circles for the larger-scale mesonet and smaller
circles for the more dense mesonet). Overview of Conditions

The overall synoptic scenario was like that described by

tornado form to the north (T1 in Figure 4). Figure 5 shows Doswell [1980] for multiday severe weather events over the
this tornado at ---2208 from L1. The characteristics of the High Plains. The event was preceded by a cold front passage
tornado are fairly classic for this area: a tiny condensation from the north early on June 13 that began several days of
funnel extends from a flat cloud base, and the tornado itself thunderstorms and severe weather in eastern Colorado.
is defined by a column of dust extendingt¾omthe ground to The Denver sounding at 1200 had a conditionally unstable
-1
cloud base. As with most of the tornadoes in the Zipser and lapserate, with a potential buoyancy (CAPE) of 1154J kg
Golden case, the tornado was at the extreme west edge of and a Lifted Index of -5.5øC for predicted afternoon condi-
the storm. T1 passedmainly over an open area with very few tions (the actual prestorm dewpoint in the Denver area was
structures. higher than the value used for this calculation). An estimate
Within minutes after the first tornado had ended, another from profiler winds of the hodographfor Denver at the time
funnel cloud appeareda few kilometersto the south.Unlike of the tornadoes is shown in Figure 9. The combination of
the first tornado, the condensation funnel associated with increased southerly flow, following the passage of the out-
this tornado (T2) continued to grow, and by ---2220it had flow boundary from the south, and midlevel west to north-
become very large with condensationnearly to the ground. west flow created ---20 m s -• of shear in the lowest 5 km.
A photographof this tornadoat ---2223from L2 is shownin Wind profiler observations from sites northeast of DEN
Figure 6. During the time of this photograph,occasional indicated stronger midlevel northwest winds that increased
wisps of condensationcloud were seen at and near the duringthe day and, coupled with increasinglow-level south-
ground, with an area of clear air remainingbetweenthe easterly flow, might explain the existence of a long-lived
groundand the lowest extent of the funnel. In this case, no supercellular storm in extreme northeastern Colorado at the
dust cloud was observed because the tornado was passing time of the Denver area tornadoes.
over an older residential area noted for its boulevards with With ambient southeasterly flow, a DCVZ had developed
very large trees. The increasedroughnessin this area may by late morning, approximately north-south acrossthe east-
have reduced the actual surface damage, in the manner ern portion of Denver. Convection began by late morning
discussedby Elsom and Meaden [1982]. over the higher terrain along the Front Range. By noon i 1800
While T2 was on the ground,a third tornado(T3) devel- UTC), storms had moved out onto the plains and developed
oped7 km to the southwest. Thistornadowasnot visible eastward along outflow boundaries. These early storms
fromL2, but a photograph takenfromL3 at ---2225is shown tended to weaken as they moved east away from the
358 TORNADO CHARACTERISTICS

Fig. 5. Photographof TI at •-2208. looking north from location L!. (Photographby E. Szoke.)

terrain because temperatures had not yet warmed to convec- ment on this day was rather complex. By 1900 some cells
tive levels over the plains. had begun to form along the DCVZ boundary, but the
As noted by Pltrdom and Weat'er [1990], who studied this intersection of an outflow boundary with the DCVZ failed to
event using visible satellite data, the evolution of develop- initiate further growth. This has been found in other cases

Fig. 6. Photograph
of T2 at ---2223,lookingsouthfromlocationL2. (Photograph
by E.
 SZOKE •ND ROTUNNO 359

Fig. 7. Photographof T3 at -2225, looking •outh-southeastfrom location L3. (Photograph by S. McGinniss of the
Colorado State Patrol.)

(see, for example, Szoke and Brady [1989]) and reflects Northward moving outflow from storms that had developed
surface temperatures too low to overcome the capping over the western portion of the higher terrain of the Palmer
inversion even with the added lift provided by a boundary Divide intersected the southxvard moving outfiox• and the
collision. Purdom and Weaver indicated that by 2000, one southernportion of the DCVZ to initiate a nex• thunderstorm
main outflow boundary had become organized from about just southeast of Denver around 2000. This storm grex•
Platteville ("ptl") south into the Denver area, with the stronglyafter 2030 and produced golf ball size hail east of
strongest push to the south toward the Palmer Divide. Littleton ("ltn") at 2115. The storm developed a midlevel

Fig.8. Photograph
ofT4at --,2225,
takenoffof video
froma television
station
helicopter.
(Photograph
courtesy
of
KMGH-TV, Channel 7,
360 TORNADO CHARACTERISTICS

15

10 __ Sfc
0.5
() _.
i

--- 1.0
•
1.5 •
9

.......... 0 ,,

( 2.0

• 34 6 ,--
-5

-10

-15 I .,,I I
-5 0 5 10 15 2(

u (ms'•)
Fig. 9. Hodograph (heights in kilometers AGL) for DEN at 2200, using the DEN surface observation, DEN profiler
winds through 4 km, and PTL profiler winds from 5 to I0 km.

mesocyclone (as observed by the Mile High 10-cm wave- Detailed Surface Observations
length Doppler radar [Pratte et al., 1991], located 15 km In addition to the 5-min data from the mesonetwork shown
northeast of DEN) and visually had a well-defined lowered
in Figure 1, a much smaller-scale mesonetwork of surface
base (wall cloud) by 2100. However, the storm quickly died
as it moved east into the cold air behind the northward
stationswas in place around Stapleton Airport, as shown in
Figure 4. This network was established to make detailed
moving outflow.
1-min observations of surface parameters for an experiment
By contrast, a strong thunderstorm that had formed well
west of Fort Collins (FCL) over the mountains maintained to measure and predict downburst/microburst winds. Al-
its intensity as it moved eastward. This storm had a midlevel though individual pressure observations from the mesonet
mesocyclone as early as 1945 as it was just coming off the were somewhat suspect,the temperature and moisture mea-
foothills. Of importance to events in the Denver area was a surements were accurate. Therefore temperature and rela-
very organized strong outflow produced by the storm that tive humidity are presented in the following figures.
moved southward. In terms of the tornado intensities, it is difficult to be
What followed between 2130 and 2200 is an ideal type of certain of the true maximum tornado winds. Although Rob-
boundary collision that could generate low-level cyclonic erts and Wilson [1989] and Wilson and Roberts [1990] have
vortices and lead to a tornado outbreak of this type. The examined the tornadoes using several Doppler radars at
detailed observations follow in the next section, but in close range, the maximum resolution of the data is ---250by
essence the two outflow boundaries intersected at the loca- 250 m, so that the actual intensity of the tornadic winds
tion of what remained of the DCVZ, in an area of full cannot be determined. Their vorticity estimates indicated
sunshine and possibly deeper moisture (owing to the maximum
valuesof 200(x 10-3 s-1) for T1,250for T2, and
DCVZ). Additionally, the boundariesintersectedin such a 125 for T3 (T4 was not studied). It is possiblethat the smaller
manner (not head on) as to maximize the cyclonic shear in values for T3, which actually had the highest F rating, were
and north o[ the Denver area along the intersection. caused by a narrower vortex not being suitably
 SZOKE AND ROTUNNO 361

Spectradata, if available,wouldprovidea morecertain active updraft in this area. Similar to the earlier tornado, T1,
estimateof the actualwindsin the tornadoes.
Photogram- tornado T4 also occurred at the western edge of the echo and
metricestimatesmightalsobepossible, although
thesehave the northern end of the line (Plate lb). From Figure 10c it is
not been attempted. clear that dry northerly flow was presentjust to the west of
The F ratingsin Table 1 wouldindicatethatT2 andT3, the T4, from the observation at "bri" located 3 km to the
tornadoeswith the larger condensationfunnels,were stron- west-northwest, which reported a northerly wind and an RH
gerthanT1 andT4, whichhadfunnelscomposed mainlyof of only 40% at 2220. We hypothesizethat the visual appear-
dust. If the F ratingswere in fact representative of the ance of T4 (Figure 8), composed mainly of a dust cloud,
tornadointensities,thenthedifferencesin thevisualappear- resulted from the tornado occurring at the western edge of
ancecould,at leastin part, be accountedfor by T2 and T3 the echo, in close proximity to the dry, low-level northerly
beingstrongertornadoes,giventhat strongerrotationwould flow.
leadto a lowerpressure
deficitin thetornadoanda compar- By contrast, tornado T2 with the huge condensation
atively longer condensationfunnel, other factors being funnel is seen from Figure 10c to be located closer to more
equal. However, as noted in the earlier discussion and moist air. T2 was ---3 km west of a broad area of ->60% RH,
summarizedin Table 1, T1 and T4 passedover vastly while the earlier T1 had been surroundedby drier low-level
different terrain, with virtually no structuresthat could be air (see Figures 10a and ! 0b). Analysis of the more detailed
damaged.This makesthe F scalea less than satisfactory mesonetin Figure 10d indicates two separate areasof higher
meansof comparingintensitiesfor this case(a moregeneral (->60%) RH in the larger area shown in Figure 10c to the east
problemdiscussedby Gra,•ulis[thisvolume]).The Doppler of T2. The area of higher RH northeast of STPL was the
studies cited above suggest that the tornadoes are more result of rain from the cell associated with the first tornado,
comparable in intensity than the F scale ratings would T1. The higher RH just to the east of T2 was from the echo
indicate, but given the maximum resolution of the radar in the line of cells that was directly associated with T2, and
data, one cannotreally determineby what degreethe actual this area of moisture increased during the lifetime of the
tornadicwinds differ. We believethereis uncertaintyregard- tornado, as seen in Figure 1Of at 2230. Note also from Figure
ing the comparative strengthsof the tornadoes,but there is 10f the strong easterly flow from the very moist air toward
insufficientevidenceto be certain of the differences.Regard- T2, with the RH at the station just to the east of T2 having
lessof the possiblevariationsin intensity, the appearanceof increased
to 96%,with an easterlywindof 11m s-1.
the funnels is consistent with the variations in the low-level The detailed mesonet's western edge is just near the
moisture that will now be presented and therefore consistent location of T2, so it does not resolve the drier northwesterly
with our hypothesis that such moisture variations can ac- flow to the west side of the intersecting boundaries where the
count for the funnel appearances. tornado had formed. Single-Doppler velocities though indi-
Figures 10a through 10f present the mesonet data from cate that the strongernorthwesterly flow, which was also dry
the two networks for three representative times during the as indicated from Figures 10c and 10e, was at least 5-10 km
tornado outbreak. Images for the same times from the Mile west of T2 between 2215 and _-30. The low-level radar
High Doppler radar, located ---15km northeastof DEN, are reflectivity did indicate that T2 also formed at the western
displayedin Plates 1a and 1b (0.7øelevationanglescans)and edge of the echo (Plate lb), although not as much at the
Plate 1c (4.00 scan). The first time shown in Figures 10a and extreme western edge as for T1 and T4. Our hypothesis for
10b is 2205, when the first tornado, T1, had touched down. the appearanceof T2 is that the higher low-level moisture
It is apparentfrom the analysesthat at the time of the first valuesjust to the east of T2 were in large part responsiblefor
tornado the air was rather dry at low levels, with relative the unusually large condensation funnel observed.
humidity (RH) valueslessthan 50% north of Denver. Addi- The conditions associated with T3 were somewhat more
tionally, T1 occursat the very western edgeof the parent complex. The condensation funnel associatedwith T3 was
cloud, away from any moist low-level air. The more detailed relatively large for this area and considerably different in
mesonet data for 2205 (Figure 10b) show the highest RH appearancethan T1 and T4. However, except duringa short
well to the southeast of T1. This moisture had increased periodnear the end of its lifetime, the condensationfunnel of
rapidlyfrom valuesmore uniformlybetween45% and 50% T3 was much narrower and did not extend as near to the
before2200, as rain beganto fall from the easternportionof ground as for T2. Like T2, T3 formed in an area of much
the developing line. higher low-level moisture than T1 and T4. Unfortunately,
Duringthe 2218-2220time period(Figures10c and 10d), the detailed mesonet is more than 6 km to the northeast of
threetornadoeswere in progress(T1 had endedat 2213).By T3. However, examining the less dense mesonet data (Fig-
this time the strongestlow-level reflectivity(Plate lb) con- ure 10e) together with the Doppler radar data indicatesthat
sistedof a line of several cells extendingfrom east of "bri" three different airflows were probably involved in the low-
southward to eastern Denver county, consistentwith the level air feeding the tornadic storm: (1) moist air to the east
higherRH locatedin the sameareathat wouldbe theresult and northeast of the storm, some of which was also involved
of downdraftsand rainingout of the initial cell in the line in T2; (2) drier northerly flow to the west that had surged
associatedwith T1. At higher levels, greater reflectivity southward' and (3) a surge of relatively warm and dry air
extended into southern Denver and over T3, indicating from the southwest. Thus T3 essentially evolved at a
362 TORNADO CHARACTERISTICS

5km
i

2205 UTC

Fig. 10a Fig. 10b

c)

\
!z•5 \

•5

o 5km
I I

15km _.•elb
, 2 220 UTC
T3
22.18 UTC

Fig. 10c Fig. 10d

Fig. 10. Analyses of relative humidity (percent) for the two mesonets. Also shown for the larger-scale mesonet
(Figures 10a, 10c, and 10e) is a temperature analysis, with values reduced to the elevation of DEN (degreesCelsius,
shortdashes)
windbarbs(fullwindbarb,5 m s-l; shortbarb,2.5m s-l), andboundaries.
Arrowsaredisplayed
for
the winds with the smaller-scale mesonet (Figures 10b, 10d, and
 SZOKE AND ROTUNNO 363

Fig. 10e Fig. 10f

Fig. 10. (continued)

point intersection,which might explain why, at least from The differingappearancesof the tornadoesfor the june 15
the video, it appears to have the most intense vortex. casewere related to the low-level moisture present near the
Interestingly, this triple-point configurationresemblesthe tornado circulations using detailed mesonet observations.
schematic for the low-level boundaries with a supercell The observationsare consistentwith the appearanceof the
tornadic storm [Lemon and Doswell, 1979], althoughin our tornadoesand supportthe idea that variationsin the low-
case the boundaries evolved from the interaction of nearby level moisture could be responsible for both the enormous
events, rather than an evolution of the storm itself. The funnel that occurred with T2 and the lack of condensation
source of the dry air from the southwestwas a weak funnels with T1 and T4. The wide variations in conditions
downburstfrom a separateechothat had movedeastfrom over sucha smallarea and the apparenteffectson the visual
the mountains.From the video and Doppler radar data it appearance of thetornadoesraisesomeinterestingquestions
appearsthat the increasein the sizeof the visibleconden- regardingspotterobservations of tornadoes.The sizeof the
sation funnel toward the end of the lifetime of T3 occurred as visiblefunnel cloud is quite often used as an indicationof
moreprecipitationbeganfallingfrom the cloudline. tornadointensitywhen observingtornadoesfrom a distance.
Our observationssuggestthat this may not always be a
4. SUMMARY AND CONCLUSIONS reliable indicator. Whether our results for this type of
Studies of the June 15, 1988, tornado outbreak in the tornadogenesis applyto more supercellularstormsis opento
Denver area indicatethis is a goodexampleof tornadogen- question, but certainly variations in the boundarylayer
moistureoccur. Additionally, visual observationsof torna-
esis resultingfrom the intersectionof two thunderstorm
outflow boundarieswith a stationaryboundary. The inter- does associatedwith heavy-precipitationsupercelIsalone
section occurred in such a manner as to maximize the might be considerablydifferent than those of the same
cyclonicshearacross theintersection,yielding
a number of strengthtornadoassociatedwith a low-precipitationstorm
small-scalecentersof vertical vorticity [Wilsonand Roberts, [Doswell et al., 1990].
1990].Someof thesewereassociated withtornadogenesisas Considerable progresshas been madein the observation
theywerestretched in theverticalbeneathvigorouslygrow- and understandingof tornadoesof this type over the past
ing stormsinitiatedby the collidingboundaries.Thissce- decade,primarily from observationsobtainedthrough a
nariofor nonsupercell
tornadogenesis is consistent
withthat numberof field programs.A goodconceptualmodelof this
documented by Wakimoto andWilson[1989]andBradyand typeof tornadogenesis
and the antecedentconditions
to be
Szoke[1989]from studiesof otherColoradoevents. observednow exists [Brady and Szoke, 1989; Wakimoto
364 TORNADO CHARACTERISTICS

Plate 10
Plate 1. Doppler radar reflectivity images from Mile High radar for the same times as in Figure 10. Elevation angles
arc 1.2° for Plates 10 and Ib and 4.00 for Plate Ie. The white arrow marks the location of (ll) Tl, (b) T4, and (e) T3.
The county outlines are shown for reference 10 the Denver county outline in Figures lOa. IOc, and 10e.
E_ _ ~

Plate lb

Plate Ie
366 TORNADO CHARACTERISTICS

Wilson, 1989]. Nonetheless, observation of these events tornadoes, downburstsand microbursts, and implications regard-
over the yearsfrom the forecastingstandpointindicatesthat ing vortex classification,Mon. Weather Rev., 111,220-235, 1983.
there is often a fine distinctionbetween the productionof Golden, J. H., The life cycle of Florida Keys' waterspouts,J. Appl.
Meteorol., 13, 676-692, 1974a.
tornadoesfrom the type of interaction documentedhere and Golden, J. H., Scale-interactions for the waterspout life cycle, J.
just the formation of thunderstorms. A number of the Appl. Meteorol., 13,693-709, 1974b.
thunderstormsformingalongthe DCVZ will developfunnel Golden, J. H., Picture of the month: Jet aircraft flying through a
clouds that do not become tornadoes. Further documenta- Denver tornado?, Mon. Weather Rev., 106,575-578, 1978.
tion of a numberof casesover a long periodwouldbe useful Grazulis, T. P., A 110-year perspective of significanttornadoes, this
volume.
in determining the distinctions between the three outcomes: Lemon, L. R., and C. A. Doswell, Severe thunderstorm evolution
ordinary thunderstorms,funnelsonly, and tornadoes.It is an and mesocyclone structure as related to tornadogenesis, Mon.
important distinction from the public's perspective, since Weather Rev., 107, 1184-1197, 1979.
unlike conditionsassociatedwith supercellstorms, in many McCarthy, J., R. Roberts, and W. Schreiber, JAWS data collection,
cases the tornado is the only form of severe weather that analysis highlights and microburst statistics, in Preprints, 21st
Conference on Radar Meteorology, pp. 596-601, American Me-
occurs with these nonmesocyclone type storms. teorological Society, Boston, Mass., 1983.
Pratte, J. F., J. H. Van Andel, D. G. Ferraro, R. W. Gagnon, S. M.
Maher, and G. L. Blair, NCAR's Mile High meteorologicalradar,
Acknowledgments. Thanks to Rita Roberts (NCAR) for useful in Preprints, 25th Conference on Radar Meteorology, pp. 863-
discussionsregardingRoberts and Wilson's Doppler analysesof the 866, American Meteorological Society, Boston, Mass., 1991.
June 15 tornadoes and for help in getting the smaller-scalemesonet Purdom, J. F. W., and J. F. Weaver, A satellite perspective of the
data. The authors appreciate helpful reviews of this manuscriptby 15 June, 1988tornado outbreak in Denver, Colorado, in Preprints,
Morris Weisman (NCAR), Christopher Church, and an anonymous 16th Conference on Severe Local Storms, pp. 167-170, American
reviewer. The National Center for Atmospheric Research is spon- Meteorological Society, Boston, Mass., 1990.
sored by the National Science Foundation. Roberts, R. D., and J. W. Wilson, Multiple Doppler radar analysis
of the 15 June 1988Denver tornado, in Preprints, 24th Conference
REFERENCES on Radar Meteorology, pp. 142-145, American Meteorological
Society, Boston, Mass., 1989.
Bluestein, H. B., The formation of a "landspout" in a "broken- Szoke, E. J., and R. H. Brady, Forecasting implications of the 26
line" squall line in Oklahoma, in Preprints, 14th Conference on July 1985 northeastern Colorado tornadic thunderstorm case,
Severe Local Storms, pp. 267-270, American Meteorological Mon. Weather Rev., 117, 1834-1860, 1989.
Society, Boston, Mass., 1985. Szoke, E. J., M. L. Weisman, J. M. Brown, F. Caracena, and T. W.
Brady, R. H., and E. J. Szoke, A case study of nonmesocyclone Schlatter, A subsynopticanalysis of the Denver tornadoes of 3
tornado development in northeast Colorado: Similarities to wa- June 1981, Mon. Weather Rev., 112,790-808, 1984.
terspout formation, Mort. Weather Rev., 117, 843-856, 1989. Wakimoto, R. M., and J. W. Wilson, Non-supercell tornadoes,
Browning, K. A., and G. B. Foote, Airflow and hail growth in Mon. Weather Rev., 117, ! 113-1140, 1989.
supercell stormsand some implications for hail suppression,Q. J. Wilczak, J. M., T. W. Christian, D. E. Wolfe, R. J. Zamora, and
R. Meteorol. Soc., 102, 499-533, 1976. B. Stankov, Observations of a Colorado tornado, Part 1, Meso-
Burgess, D. W., and R. J. Donaldson, Contrasting tornadic storm scale environment and tornadogenesis,Mon. Weather Rev., I20,
types, in Preprints, 1Ith Conferenceon SevereLocal Storms, pp. 497-520, 1992.
189-192, American Meteorological Society, Boston, Mass., 1979. Wilson, J. W., Tornadogenesisby nonprecipitation induced wind
Doswell, C. A., Synoptic-scaleenvironmentsassociatedwith High shear lines, Mon. Weather Rev., 114, 270--284, 1986.
Plains severe thunderstorms, Bull. Am. Meteorol. Soc., 61, 1388- Wilson, J. W., and R. D. Roberts, Vorticity evolution of a non-
1400, 1980. supercelltornadoon 15 June 1988near Denver, in Preprints, 16th
Doswell, C. A., A. R. Moller, and R. Przybylinski, A unified set of Conference on Severe Local Storms, pp. 479-484, American
conceptual models for variations of the supercell theme, in Meteorological Society, Boston, Mass., 1990.
Preprints, 16th Conference on Severe Local Storms, pp. 40-45, Wilson, J. W., J. A. Moore, G. B. Foote, B. Martner, A. R. Rodi,
American Meteorological Society, Boston, Mass., 1990. T. Uttal, and J. M. Wilczak, Convection Initiation and Downburst
Elsom, D. M., and G. T. Meaden, Suppressionand dissipationof Experiment,Bull. Am. Meteorol. Soc., 69, 1328-1348,1988.
weak tornadoes in metropolitan areas: A case study of greater Zipser, E. J., andJ. H. Golden, A summertimetornadooutbreakin
London, Mort. Weather Rev., I!0, 745-756, 1982. Colorado: Mesoscale environment and structural features, Mort.
Forbes, G. S., and R. M. Wakimoto, A concentrated outbreak of Weather Rev., 107, 1328-1342,
 On the Use of a Portable FM-CW Doppler Radar for Tornado Research
HOWARD B. BLUESTEIN

Schoolqf Meteorology,EnergyCenter, UniversityqfiOklahoma,Norman, Oklahoma 73019

WESLEY P. UNRUH

Mechanicaland ElectronicEngineeringDivision,Los AlamosNationalLaboratoo',Los Alamos,Nen' Mexico87545

1. INTRODUCTION et al., 1985].The maximumwind speedsfound were around

90 m s-•. Volume-averaged
wind speedshavebeenesti-
Although much has been learned about tornadoesin the mated in severaldozen nonsupercelltornadoesin Colorado
last 15 years, we still do not know their maximumwind [e.g., Wakimoto and Wilson, 1989]' however, volume-
speeds,three-dimensional wind field, and sourceof vortic- averagedwind speedsare less than the wind speedsdeter-
ity, especiallyin supercellstorms.Photogrammetric analysis mined from spectral shapes.
of debris movies [e.g., Golden and Purcell, 1978] does not The useof a portableDoppler radarto make wind speed
reveal the line-of-sightwind componentand doesnot work measurementsin tornadoeswas suggestedby Zrnic et al.
insideopaquecondensation funnels,where debrismotionis [1985]' they noted that higher spatialresolutioncould be
hidden,or elsewherewherethereis no debris.Serendipitous obtainedby transporting a radarto a locationneara tornado.
direct measurementsby instrumentsare rare, and the instru- There are additionalreasonsfor usinga portableradar. Since
mentsare often destroyed[Davies-Jones,1986],whiledirect the likelihooda tornado will pass near a fixed-siteradar is
measurementsmade with instrumentsdeliberately placed in very small,a portableradar can increasesubstantially the
the path of tornadoesare also rare and difficultto obtain number of data sets collected. The strategy used to get near
[e.g., Bhtestein, 1983]. Analysis of condensation funnel tornadoes in an intercept vehicle is given by Bluestein
photographs [Davies-Jones,1986]and nearbysoundings [1980]. When a radar is brought to within 3-5 km of a
[Snowand Pauley, 1984]basedon the hydrostaticassump- tornado,it is usuallyalsopossibleto confirmthe detectionof
tionhasyieldedwindspeedestimates thoughtto be dubious the tornadoand to map out photogrammetrically the azi-
[Fiedlerand Rotunno,1986],owingto strongverticalaccel- muthal(crossline of sight)componentof the wind if debris
erations in the tornado boundary layer. Damage analysis and cloudtagsare trackableand to relate the windsto the
[Fujita, 1981]requiresthat objectsexistandare damaged condensation funnel, if it is present.The photogrammetri-
andthat we understandtheir responseto wind [Dos,,'elland callydetermined windanalysescomplement the radial{,line
Burgess, 1988]. of sight)measurements made by the radar. Nearby radar
Previousstudieswith fixed Dopplerradarsare reviewed observationscan also determine the wind field below cloud
by Bluesteinand Golden[thisvolume].Dopplerradarmea- base,something that distantradarscannotdo. Finally,we
surementsof tornadoesin supercelIsare not commonbe- note that since the highest wind speedsin tornadoesare
causeit is rare that tornadoespassby closeto them (i.e., probablylocatedin relativelysmallvolumes,the radar's
within20-50 km). The firstmeasurements of windspeedsin abilityto measure the speedsof thefastest-moving scatterers
a tornado with a radar were made by Smith and Holmes is improved by bringingthe radarascloseaspossible to the
[1961].The NationalSevereStormsLaboratorymeasured tornado.
wind(power)spectra(backscattered
powerasa functionof
radialwindspeed)withtheir10-cmpulsedDopplerradarin 2. PRINCIPLESOF FM-CW DOPPLERRADAR
onlyfoursupercell
tornadoesfrom1973to 1981[e.g.,Zrnic
The principlesof FM-continuouswave (CW) Doppler
radarare reviewedby Skolnik[1980],Strauch[1976],and
The Tornado:its Structure,Dynamics,Prediction,andHazards.
Do•,iakand Zrnic [ 1984].One way to modulatethe transmit-
GeophysicalMonograph79
Copyright1993by theAmerican
Geophysical
Union. tedfrequency
asa functionof timeisshownin Figure1.The
368 PORTABLE FM~CW DOPPLER RADAR

the targetis obtainedfrom the Doppler shiftfrequencyof the

impulse function having the greatest amplitude (i.e., from
fo+B
fT,•/,'1/I/,
fR the beat frequency).
If the target is moving, then the phase of each wave
segmentis uniformlyadvancedor retardedfrom the phaseof
the previous one (Figure 3a). (Only for certain speedswill
there be a phasechangefrom segmentto segmentof integer
multiples 2rr.) It is this change in phase from segmentto
segmentthat allows one to estimate the Doppler velocity.
The spectrum of the moving target (Figure 3b) is similar to
the spectrumof the stationary target; it is equivalent to the
stationary spectrum but shifted in frequency from the sta-
TIME
tionary spectrumdue to the progressivephase shift.
Fig. 1. Transmittedfrequencyfr (solid line) and received fre- In practice, one computesthe spectrumfrom a finite series
quencyfR (dashedline) as a function of time. The sweepwidth is B; of ramps, not from an infinite series. The spectrum of a
the sweep repetition frequency is 1/T. (Adapted from Strauch
[1976].)
moving target computedfrom a seriesof N ramps is similar
to the spectrumshownin Figure 3b, except that the impulse
functions are broadenedand there are some low-amplitude
frequency is linearly ramped by B Hz, the sweep width, power fluctuationsat frequenciesin between the frequencies
periodically every T s, the sweep repetition time; 1/T is the at which there are larger-amplitudepeaks (not shown).
sweep repetition frequency. The latter is analogous to the Like the pulsed Doppler radar, the FM-CW radar is
pulse repetition frequency in pulsed radars. FM-CW Dop- susceptible to range folding and velocity aliasing. After
pler radars are sometimesreferred to as "chirped" radars, Strauch [1976], we list the following important characteris-
becausebirds make chirping soundsas the pitch they emit tics of a linearly modulated radar system:
changes. The backscattered signal is mixed with the trans- 1. The maximum unambiguousrange is about cT/4.
mitted signal; only that part of the signal representing the 2. The maximum unambiguous velocity is A/4T, where A
difference frequency is retained. is the wavelength.
If the radar target is stationary, then the signalis a periodic 3. The range resolution is c/2B.
series of wave segmentsof identical phase (Figure 2a); the 4. The velocity resolution is X/2NT.
correspondingspectrum is composed of a series of equally For equivalent data record lengths, the sensitivity of an
spaced impulse functions, whose spacing is constrained to FM-CW Doppler radar is lower than that of a CW Doppler
be the sweep repetition frequency (Figure 2b). The range to radar because the effective bandwidth required to recover

(a) -t
Z • ' I I

---IJlAAjlAAkAA,\AA k TIME

Mvv v,,v v vi,,vv v v v vjv

STATIONARY TARGET

n"•
wr•
(b)

1 2 3 4 5
T T T T T
FREQUENCY

Fig. 2. (a) Receivedsignalas a functionof time for a stationarytarget. (b) Powerspectrumfor Figure2a. (Adapted
from Strauch
 BLUESTEIN AND UNRUH 369

(a) -t

MOVING TARGET

(b)
I Doppler
Shift
1 2 3 4 5
T T T T T
FREQUENCY

Fig. 3. As in Figure 2, but for a movingtarget.The spectrumis identicalto that in Figure 2b, but for a movingtarget.
(Adapted from Strauch [1976].)

both range and Doppler velocity is larger than that necessary return for which the scatterers are relatively small and/or
to recover Doppler motion alone. distant, yet not be saturated by ground clutter and strong
scatteringfrom nearby targets.
3. DESIGN CONSTRAINTS
4. DESIGN OF THE Los ALAMOS NATIONAL LABOIL&TORY
The size of the radar places constraintson the power and
PORTABLE RADAR
wavelength. In order that the radar be portable, it must be
small, the power requirementsrelatively low, and the anten- The Los Alamos National Laboratory (LANL) radar is a
nas relatively small. Although higherpower output is attain- solid-state,bistatic (i.e., having separatereceiving and trans-
able using solid-state devices, it is prohibitively costly. The mitting antennas)device [Bluestein and Unruh, 1989]. De-
antennas cannot be too small, however, or the half-power tailed specificationsof the radar are summarizedin Table 1.
beamwidth would be too large and the spatial resolution A functional diagram of the radar is given in Figure 4.
compromised. For a given size, an antenna for a smaller- Originally a CW radar, it was modified in 1988 to have
wavelength radar can have a smaller beamwidth and in- FM-CW capability so that range informationcould also be
creased sensitivity for small scatterers. However, if the obtained.In FM-CW mode it operateswith a sweeprepeti-
wavelengthis too short, attenuationby hydrometeorscould tion frequency derived from the boresightedvideo camera.
be a problem. If the wavelength is relatively long, then The FM-CW signalsare recordedin VHS video format; it is
attenuationis muchlessof a problem,but the antennawould therefore convenient to derive the FM modulation from the
have to be large to get a reasonablysmall beamwidthand horizontal sync signalsof the video field. FM-CW Doppler
high spatialresolution,and sensitivityto smallscatterersis radarshave been usedbefore in hydrometeorstudies[Ligth-
reduced. art and Nieuwkerk, 1989]; however, the LANL radar is the
The rangefoldingand velocity aliasingcharacteristics of first FM-CW radar to be used for tornado research.
the radar also impose serious constraints. If the sweep The bistaticradar has two parabolicantennaswith linearly
repetitionfrequencyis too low, the maximumunambiguous polarizedfeeds(Figure5), separatedby a septumto provide
velocity is too low; if the sweeprepetitionfrequencyis too increasedisolation; one is for transmitting,and the other is
high, the maximum unambiguousrange is too low. for receiving.Although individualantennapolarizationscan
Other constraintsare imposedby the sweepwidth. If it is be set to be horizontal or vertical, they cannotbe adjustedin
too low, the range resolutionis poor; if it is too high, the the field. We have used horizontal polarization for both
detected bandwidth is high, and it becomesvery expensive antennas.Taperedfeedsto the reflectorsreducesidelobesto
to record the data with a high-bandwidthrecordingdeviceor over 20 dB down below the main beam and thus help to
to process the data in real time. reduce near-field clutter. At 1.6 km (3 km) the azimuthal
The radar and its recording system must have a wide resolution is 140 m (262 m). For comparisonwe note that
enoughdynamicrangeto make measurements of weak radar some fixed-site radars have beamwidths as low as 0.8 ø,
370 PORTABLE FM-CW DOPPLER RADAR

TABLE 1. Specificationsof the LANL Portable Doppler Radar (1991)

Specifications
General CW FM-CW

Wavelength/frequency 3 cm/10.5 GHz

Power output 1 W maximum
Dynamic range (before 70 dB
recording)
Antenna type 2 parabolic dishes
Antenna half-power beamwidth 5ø
Weight (approximate)
Radar-antenna package <50 lb
Case containing 50 lb
batteries/recorders
Setup time (approximate) 2-3 min
Power requirements two 12-V rechargeable
batteries
Modulation
Sweep repetition frequency NA* 15.750 kHz
Sweep width NA 1.92 MHz
Maximum unambiquous range NA 5 km
Maximum unambiguousvelocity -+292m s-• +-115 m s -•
Range resolution NA 78 m
-1
Velocity resolution NA 1.8 m s
Data recording Stereo audio HD channels of video video channel of video
recorder; stereo channels of hi-fi recorder
audio recorder
Signal monitoring earphones video monitor
Boresighted video monitoring video monitor, video channel of video none
and recording recorder

*Not applicable.

which the azimuthal resolution is 140 m (262 m) at 10 km us to separate,in real time, the CW Doppler signalsfrom the
(18.8 km) range. I/Q mixers into two broadband(0-20 kHz) audio signals,one
Backscatteredreturn is amplified directly at 10 GHz by a of which containsthe approaching(negative) and the other
low-noise microwave preamplifier. This device sets the the receding (positive) Doppler shifts. These two audio
overall noise figure of the radar. The microwave detectors signalsare recorded directly. The effective channel separa-
are conventionalhomodynein-phase/quadrature(I/Q) mix- tion is greater than 50 dB, which providesDoppler signals
ers. The low-level signals are amplified at baseband (0-5 uncontaminatedby opposingmotion in the same field of
MHz) by matched operational amplifiers. High-pass filters view. These audio signals are monitored in a set of stereo
further reduce the effects of ground clutter. headphonesworn by the operator; approachingtargets are
The radar and antennas are built into one package, which heard in one ear only, receding in the other. With no radar
fits into an altitude/azimuth mounting tripod and is pointed return, only the steadyhissdue to the noisefloor of the radar
by hand. Batteriesprovide over 2 hours of operationin the is heard on both channels. The pitch of the sound from an
field. The radar can be instantlyswitchedbetween CW mode actual Doppler-shiftedreturn is usually proportionalto the
and FM-CW mode with a switch accessible to the operator. line-of-sightwind speed.Groundclutter from waving wheat,
A boresightedvideo camera for visual documentationis grass,and tree leaveshas a low-frequency,randomsound;
mounted on the radar. An umbilical cord of several cables birds and passingcars that appearin the antennasidelobes
from the radar is connected to terminals on a separate case have nearly pure tones that come and go rather quickly.
(Figure 6), which contains the batteries, a videotape re- Althoughit is somewhatof an art to distinguishandinterpret
corder and monitor, an audiotape recorder, and meters that in the field the meteorologicallysignificantsoundsfrom the
monitor the state of the batteries. The whole system is easilyextraneous ones, one soon learns to make good use of this
set up and dismantledby three people. One sets up the real-timeaudiooutput of the radar as an aid in pointingthe
tripod,whilethe othertwo mountthe radar-antenna package device in CW mode. One can, for example, locate the center
onto the tripod. One then preparesthe package,while the of a tornado vortex and the region of highest Doppler
other two set out the case and connect the case to the radar. velocitiesby scanningacrossthe funnel cloud or debris
Care must be taken to deploy the radar away from structures cloud and listeningto the stereo audio signal.
(e.g., treesandhouses)that mightproduceexcessiveground The poweroutputof the radaris adjustedmanuallyby the
clutter. operatorso that the recordedsignalis not saturated.The
A dedicatedbroadband 90øaudio phase shift circuit allows poweris notrecorded,however,sothatquantitative
 BLUESTEIN AND UNRUH 371

Boresight
Video Video Picture Data
Camera

700 Hz
Ramp
C•o Generator
0?c
H. •5750
RampH"Z'
Generator

Amp.
' I c%d' I o ,|
-'--] sy:o. I

4 dB
N.F.
II!1
,F:F
IImFramer
ve•'rt.
Picmff.
I
I PhaseI ..__L__,
I icture

I Audiø
CW and700 Hz +
ø
Doppler
Data
•] VHS
VCR
with
Stereo
Fig. 4. Functionalblock diagramof the LANL portableDoppler radar. (The 700-Hz modulationcapabilityis for
another application.)

tivity measurements are not possible. In CW mode, bore- resulting visual pattern on the monitor appears as a pattern
sighted video of the tornado is simultaneouslyrecorded on of vertical lines or bars whose separationon the screen is
the video channel of the videotape, with the analog I and Q inverselyproportionalto range. If the target is moving, the
signalsstored on the so-calledaudio "HD" channelsof the progressivephaseshift resultingfrom the motion leadsto a
VHS format tape. A separate, very high quality audio tilt of this pattern of lines on the screen, the directionof the
recorderis used,as well, to provideredundancy(in casethe tilt being determined by the sense of the motion. Range
video recorder or tape fails) and improved signal-to-noise foldingis seenas an abrupt changein the periodicityof this
recovery of the data. pattern. Ground clutter, resultingfrom scatteringclose to
In the FM-CW mode the wideband radar signal is stored the radar, is seen as an overall smooth variation in pattern
on the video channel of the videotape. Coincidentalvoice brightnessacrossthe video field. The radar power level is
documentationfrom a microphonemountedon the radar is adjustedto keep this strong signalcomponentfrom saturat-
recorded on the audio channelsof the videotapeand backed ingthe overallrecordedsignal.Complextargets,in whichall
up on the audiotape.The qualityof the video signalrepre- typesof motionare present, result in complexvisualvideo
sentingthe FM-CW return is viewedon the videomonitor. patterns. Such complexity appearsin video data from a
Since the dynamic range of the video cassetterecorder tornado(Figure7). The real-timeappearanceof thesevideo
(VCR) is only about 30-40 dB, which is at least30 dB less patternsis thusa powerfuldiagnosticof the actualscattering
than that of the radar, one must be very careful not to characteristics being measured.
producea saturatedsignal,which appearsas an unstable, Our scan strategy is to scan back and forth across the
unsynchronizedpicture. Groundclutter can easilysaturate tornado(Figure 5) as near to the groundas we can without
the recordingsystem,even with the high-passfiltersin the saturatingthe radarwith groundclutter. We scanonly in the
radar. CW mode, so that we have a record on videotape of where
Scatteringreturnin the FM-CW modeis formattedto be the radar is viewing. We switch into the FM-CW mode only
when the radar is stationary. Usually we have to turn the
recorded as VHS video fields. Each horizontal scan line of
the videofieldrepresentsthe recoveredradarsignalfrom a radar power down to prevent saturatingdue to ground
singlerampof the periodicfrequencysweep.For a single, clutter and then turn it back up when in CW mode. We
stationarytarget,the return from each ramp is a single frequentlyalternatebetweenCW mode and FM-CW mode.
sinusoidwhose frequencyis proportionalto range. The In CW mode we have a nearly unlimited
372 PORTABLE FM-CW DOPPLER RADAR

5. DATA PROCESSING

The recorded video data from the FM-CW mode is visu-

ally examined, frame by frame, to identify segmentsof data
judged to be the best available. The desired field (one half of
the interleaved frame) from this videotape record is captured
for analysis by computer hardware and software that digi-
tizes a single field of the record. The result consists of
concatenated time series of the 252 horizontal lines of the
video field, each digitized at 128 points, which forms the
primary data set of 32,256 points used by the computer.
From this stored data set, a segment of at least 16,384 points
(128 x N, where N is usually around 200), which is judged
to have the least contamination, is chosen for analysis.
Stationary ground clutter is removed from this time series
by a time averagethat leavesonly time-varyingsignalsin the
record. This is accomplished by averaging together the 128
separatereturns from individual ramps and subtractingthis
composite average from each of the ramp returns. The
subtracted signal represents the return from targets that are
essentially stationary over the elapsed time of 128 contigu-
ous ramps (--•8 ms). We have found this procedure to be
highly effective.
Spectral estimationof this corrected time series is ob-
tained by usinga slight modificationof the Welch procedure
[Welch, 1967]to computethe discreteFourier transform and
at the same time obtain optimum noise reduction. The total
time series is divided into three overlapping segments of
16,000 points each. Each of these uniformly overlapping
time seriessegmentshas a Hanning window applied before
the fast Fourier transformis computed.Rangenormalization
by a factorof 1/R2 is appliedin thetransform
process.
The
resultingspectraldensitiesare averagedtogetherto provide
a spectralestimationrecordof 8192points.This 8000record
represents64 rangebins, eachwith 128unambiguous Dop-
Fig. 5. Photographof University of Oklahoma (OU) graduate
student J LaDue operating the LANL radar, with OU graduate plervelocity points ranging from- 115ms-• to + 115ms-•
studentD. Speheger(left) and O U undergraduatestudentH. Stein with
velocity
resolution
of1.8ms-• Data
inthefirst
hal•
(right)attendingto the radarbox, whilecollectingdatafor a tornado and last half of the first and last range bins, respectively, are
in north central Oklahoma on April 12, 1991. (Photographby H. B. contaminatedby folding and are not used.
Bluestein.)

6. PRELIMINARY RESULTS

The CW wind spectrum(averageof four obtainedover a

velocity(292m s-•) but get no rangeinformation; in the 0.5- to l-s period) measuredlooking toward a tornado
FM-CW mode we do get range information but are suscep- (Figure8) in the northernTexasPanhandleon May 31, 1990,
tible to rangefolding,velocityaliasing,andmoresusceptible is shownin Figure9. The highestwind velocitiesestimated
to ground clutter contamination.When the tornadois be- appear tobebetween +_80and_+90ms-l; theyareindicated
yondthe maximumunambiguous rangeor doesnotproduce at the pointswherethe spectrumfalls off to the noiselevel.
a satisfactorysignalin the FM-CW mode, we record data These wind velocities are consistent with a National
only in the CW mode. We scanat higher-elevation angles Weather Service (NWS) damage survey estimate (courtesy
only afterwe haveobtainedwhatwe judgeto be a gooddata of S. Cooper,NWS, Amarillo,Texas) of F3 damage The
set at as low an elevation angle as possible. Typically, the width of the tornado was estimatedphotogrammetricallyto
elevation angle is around 10ø-20 ø. At these low-elevation be approximately200-300 m in diameter;this comparesto
anglesthe error added to the horizontalDoppler velocity the antennahalf-powerbeamwidthof slightlyunder 1000m.
estimatesis lessthan5 m s-•. If there is no tornadobut there The tornadoshownin Figure 8 was too far away to obtain
is a rotatingwall cloud,then we scanat high-elevation angle acceptable FM-CW data.However,whenthe rotatingwall
near the base of the wall cloud to search for signs of an cloud associatedwith the remains of the storm's circulation
incipient tornado circulation. camecloser,we were able to obtaina data set lookingat
 BLUESTEIN AND UNRUH 373

Fig. 6. Photographof the electronicequipmentinsidethe •adar box: audiorecorder (rear left), batteries(rear center),
video recorder (front left), meters (center front), and video monitor (far right). (Photograph by H. B. Bluestein.)

southern portion, where the rear flank downdraft is located. speedsaround35 m s-• were indicated.This estimateis
The FM-CW spectra (Figure 10) indicated return from ap- consistentwith gustsexperiencedfrom the southwestas the
proximately 2.3-3 km in range; this suggeststhat the wall wall cloud passedby overhead and 6-cm hail fell.
cloud was about 700 m in diameter. Approaching wind During the spring of 1991, five tornado data sets were

Fig.7. Photograph ofthevideomonitorofoneframeofFM-CWdataforthetornado showninFigure4. Theimage

isdominated bylinesthatsopeupwardtotheright,which
indicate
motionawayfromtheradar.Thespacing between
thelinesisinversely
proportional
to therange
ofthetargets.
Thedarkerareaontheleft-hand
sideofthevideoframe
is due to nearby ground
374 PORTABLE FM-CW DOPPLER RADAR

obtained. Of these, two included FM-CW data and excellent

video documentation. A table of all tornado data sets col-
lected is shown in Table 2. Wind speeds at or in excess of 70
m s-• werecommon.
The Dopplerwindspeeds
in a tornado
that passed by at 1.6-km range were estimated as high as
120-125m s-• Theseare the highestwind speedsever
measured in a tornado by Doppler radar and verify the
existence of F5 wind speeds. Detailed results of analyses of
the 1991 CW and FM-CW data are forthcoming.

7. SUGGESTED FUTURE IMPROVEMENTS

The greatest problem associatedwith the LANL portable

FM-CW Doppler radar is ground clutter contamination. It
could be reduced by spectrally analyzing the data in real
time, to make use of the tull 70-dB dynamic range of the
radar and then recording the spectra. Currently, the system
is limited by the 30- to 40-dB dynamic range of the video
recorder and the lack of real-time signal processing.
The radar could be made more versatile if the operator had
the ability to change sweep repetition frequency and sweep
width at the flick of a switch, so that the resolution and the
folding and aliasingcharacteristicscould be adapted to each
specificcase. The problem of how to record the data and
displaythem on a video monitor would have to be solved.
The "housekeeping" data would also have to be recorded.
Finally, if the beamwidthof the antennacould be reduced,
without increasing the size of the antenna to the point at
which the radar system is no longer portable, the spatial
resolution of the radar could be much improved. A shorter-
wavelength radar with an antenna having the same dimen-
sions as that of the LANL radar would have increased
Fig. 8. Photographlooking west at a tornado located east of
Spearman,Texas, at 1909CDT on May 3 I, 1990.(Photographby H. resolution; it would also be more sensitive to smaller targets
B. Bluestein.) such as cloud droplets in a tornado condensationfunnel.

RELATIVE SPECTRAL DENSITY

10 db/div.

SPEARMAN
31
TORNADO T

' I .... I .... I .... I ' ' ' ' I '

- 100 -50 0 +50 + 100
RADIAL
VELOCITY
(ms
-1)
Fig.9. CW radialwind•power) spectrum of thetornadoshown in Figure8, several
minutes later.Noiselevelis
indicated
by horizontal
lines;lineardropoffin logarithm
of spectral
density
indicatedby a slopingsolid
 BLUESTEIN AND UNRUH 375

/ '

• @$o
-100 -50 0 +50 +100

(a) DOPPLER
VELOCITY
(ms"•)

• 78.

2574•
• •- -100 -50
, I .... ,,, ,,...,.,,,
0 +50
,, ,,,
+100

(b) DOPPLER
VELOCITY
(ms
-1)
Fig. 10. FM-CW radial wind (power) spectraof the southernportion of the wall cloud associatedwith the tornado in
Figure 8, which had dissipated10-20 rain earlier. Each spectrumis plotted as a function of range every 78 m from (a)
1950-2496 m and (b) 2574-3120 m. The ordinate is the relative spectral density plotted on a logarithmic scale.

TABLE 2. Tornado Data Sets Obtained With the LANL Portable Doppler Radar
Maximum
Approximate Wind Speed,
Date Location m s- 1 CW FM-CW

May 25, 1987 Gruver, Tex. 60 yes no

May 13, 1989 Hodges, Tex. "' EM EM
May 31, 1990 Spearman,Tex. 80-90 yes WC
April 12, 1991 Enid, Okla. 75-80 yes yes
April 12, 1991 PondCreek, Okla. 55-60 yes no
April 26, 1991 Red Rock, Okla. 120-125 yes* no
May 16, 1991 Clearwater, Kans. 70-75 yes yes
May 26, 1991 Mooreland,Okla. 100-105 yes* no

EM, equipmentmalfunction;no usefuldata. WC, for wall cloudonly.

*Receding spectrum
376 PORTABLE FM-CW DOPPLER RADAR

Since the numberof tornadowind speedmeasurements Bluestein, H. B., and J. H. Golden, A review of tornado observa-
madewith the LANL portableradar over severalyears has tions, this volume.
far exceededthe numbermadewith a fixed radarover many Bluestein, H. B., and W. P. Unruh, Observations of the wind field
in tornadoes, funnel clouds, and wall clouds with a portable
years, we believe that the portableradar shouldcontinueto Doppler radar, Bull. Am. Meteorol. Soc., 70, 1514-1525, 1989.
be improved and used in future field experiments.An Davies-Jones,R., Tornado dynamics,in ThunderstormMorphology
improvedradar couldprovide the high resolutionand sensi- and Dynamics, edited by E. Kessler, pp. 197-236, University of
tivity needed to map out the wind field in tornadoes. A Oklahoma Press, Norman, 1986.
Doppler lidar, which can senseaerosolmotion in "clear air" Doswell, C. A., and D. W. Burgess,On some issuesof U.S. tornado
outside a tornado debris cloud or condensation funnel could climatology, Mon. Weather Rev., 116, 495-501, 1988.
Doviak, R. J., and D. S. Zrnic, Doppler Radar and Weather
augmentthe portableDoppler radar, which can estimatethe Observations,458 pp., Academic, San Diego, Calif., 1984.
motion inside the debris cloud or condensation funnel. Fiedler, B. H., and R. Rotunno, A theory for the maximum
Finally, several portable Doppler radars and lidars could windspeeds in tornado-like vortices, J. Atmos. Sci., 43, 2328-
probe tornadoes from different viewing angles and thereby 2340, 1986.
Fujita, T. T., Tornadoes and downburstsin the context of general-
provide detailed information on the three-dimensional wind
ized planetary scales,J. Atmos. Sci., 38, 1511-1534, 1981.
field.
Golden, J. H., and D. Purcell, Airflow characteristics around the
Union City tornado, Mon. Weather Rev., 106, 22-28, !978.
Ligthart, L. P., and L. R. Nieuwkerk, Studies of precipitation
Acknowledgments. This project has been funded by National processesin the troposphere using an FM-CW radar, J. Atmos.
Science Foundation grant ATM-8902594. M. Wolf and R. Bracht at Oceanic Technol., 6, 798-808, 1989.
LANL contributed to the design and modification of the radar. C. Skolnik, M. I., introduction to Radar Systems, 581 pp., McGraw-
Doniec, one of LANL's summer students, was especially helpful in Hill, New York, 1980.
making the FM-CW modification. Recent storm intercept crews Smith, R. L., and D. W. Holmes, Use of Doppler radar in meteo-
have included J. LaDue, H. Stein, D. Speheger, G. Martin, E. W. rological observations, Mon. Weather Rev., 89, 1-7, 1961.
McCaul, Jr., S. Hrebenach, and S. Contorno. The National Severe Snow, J. T., and R. L. Pauley, On the thermodynamic method for
Storms Laboratory, the National Weather Service in Norman, and estimating maximum tornado windspeeds,J. Clim. Appl. Meteo-
students at the University of Oklahoma contributed nowcasting rol., 23, 1465-1468, 1984.
support. C. Church and two anonymousreviewers provided helpful Strauch, R. S., Theory and application of the FM-CW Doppler
comments on the manuscript. radar, Ph.D. thesis, 97 pp., Dep. of Electr. Eng., Univ. of Colo.,
Boulder, 1976.
REFERENCES Wakimoto, R. M., and J. W. Wilson, Non-supercell tornadoes,
Mon. Weather Rev., 117, 1113-1140, 1989.
Bluestein, H. B., The University of Oklahoma Severe Storms Welch, P. D., The use of the fast Fourier transform for the
Intercept Project--1979, Bull. Am. Meteorol. Soc., 61,560-567, estimation of power spectra, IEEE Trans. Audio Electroacoust.,
1980. AU15, 70-73, 1967.
Bluestein, H. B., Surface meteorological observations in severe Zrnic, D. S., D. W. Burgess, and L. Hennington, Doppler spectra
thunderstorms,Part II, Field experiments with TOTO, J. Clim. and estimated windspeed of a violent tornado, J. Clitn. Appl.
Appl. Meteorol., 22,919-930, 1983. Meteorol., 24, 1068-1081,
 Discussion

RON TAYLOR, SESSIONCHAIR

National Science Foundation

PAPER F1 PAPER F2

Presenters, Joe Golden, National Oceanic and Atmospheric Presenter, Ed Szoke, Forecast Systems Laboratory and
Administration, Office of the Chief Scientist, and Howie National Center for Atmospheric Research [Szoke and
Bluestein, University of Oklahoma [Bluestein and Golden, tunno, this volume, A comparisonof surface observations
this volume, A review of tornado observations] and visual tornado characteristics for the June 15, 1988,
Denver tornado outbreak]
(Arnold Court, California State University) One lessonfrom
the last symposium,15 years ago, has not been learnt. Then (Jim Putdom, NOAA/NESDIS CIRA, Colorado State Uni-
we spent half a morningarguingabout the meaningof wind versity) You said that the Denver ConvergenceZone was
speeds. The point was that a wind speed statement must associated 'with all of these tornadoes. It seemed to me that
include the volume and the time interval of the estimate. For the Denver ConvergenceZone did not play a role in the
example, an estimatemight be for a cube with 10-m sides formation of the Denver tornadoes on this day. Also, we get
over a 1-min period. Before you comparewind speeds,you other types of tornadoesin Colorado,not just onesin the
must decide what they represent.Joe, what do your esti- narrow Denver Convergence Zone. One inflicted heavy
mates represent? damageon Limon, Colorado[on June6, 1990];thisformed
(Golden) The answer dependson the measuringtechnique. where a supercellintersecteda boundary.
With a lidar, only the wind componentalong the beam is (Szoke)I meantto mentionthe Limontornado.Yes, different
sampled.Photogrammetric estimates
areconservative because typesof tornadoes
do occurin Colorado.I readyourworkwith
thetracersrangein sizefrom dustparticlesto largechunksof John Weaver on satellite analysis of the day of the Denver
rooftops.The sameappliesto Dopplerradarmeasurements. tornadoes. I believe that there was still a weak Denver Con-
(Court) Yes, but you shouldat least indicatewhat volume vergenceZonepresentin midafternoon just prior to tornado
and time interval is involved. Otherwise, people compare development and that two outflows collided right on the
lidar measurementswith the velocity of barn doors,which, zone. Obviously,the presence of the outflows increased the
threat of severe weather considerably. We have found that
incidentally,fly about25% lessthan wind speedbecause
the probabilityof tornadoesformingon the zoneincreases
they dependon relativeair motionto keepthemaloft.
when the zone is intersected by another boundary.
(Golden)For the photogrammetric
estimates,the sampling
time is less than 10-20 seconds.
PAPER F3

(Bluestein)
For Dopplerradardata,thevolumes
aretypically Presenter, AI Bedard, Wave Propagation Laboratory,
several hundred meters on a side. The Doppler lidar has a
NOAA (A. Bedard, Tornadoes and the hierarchyof atmo-
very narrow, collimatedbeam, perhapsa few centimeters
wide. The azimuthal resolution is tremendous, but the range sphericvortices,not in this volume)
resolutionis not verygood,around150m. In thefuture,the (JoeGolden,NationalOceanicand AtmosphericAdministra-
rangeresolution maybe reducedby an orderof magnitude. tion,Officeof the Chief Scientist)I agreewith yourconclusions
exceptthe first one ["For a rangeof atmospheric vortices,
The Tornado:Its Structure,Dynamics,Prediction,andHazards.
GeophysicalMonograph79
tangential
windspeedmeasurements indicatesolid-bodyrota-
Copyright1993by the AmericanGeophysical
Union. tion within the core, l/r variationoutsidethe core"]. Although
378 DISCUSSION

someof my own data on large waterspoutsdo indicatethis (Ed Szoke, Forecast Systems Laboratory and National
profile, I believe that we have to worry about asymmetries, Center for Atmospheric Research) How difficult would it be
suchas suctionvorticesand bandsof strongerinflow, which to save some velocity spectra when there is a tornado close
make the flow noncyclostrophic and depart from the Rankine to a fixed-base Doppler radar?
combined profile. Many of the most destructive tornadoes
have these asymmetries, so we need better models. (Bluestein) Ed Brandes or Don Burgess should answer that.

(Bedard) I agree, but I think that most vortices are feeding (Don Burgess, National Severe Storms Laboratory) The
off a quasi-axisymmetriclarger-scalecirculation. The com- WSR-88D is not a radar that collects time series.
plexities that you mention are small-scalefeatures. It estimates the mean velocity in the sampling volume
and the spectral width in real time, and that is all the
PAPER F4 information you get. A different radar configuration is
Presenter, Howie Bluestein, University of Oklahoma needed to collect time series. The National Severe

[Bluestein and Unrltb, this volume, On the use of a portable Storms Laboratory has a research radar with this capa-
FM-CW Doppler radar for tornado research]
 Design for Containment of Hazardous Materials
ROBERT C. MURRAY

Lawrence Livermore National Laboratory, Livermore, Cali•brnia 94550

JAMES R. MCDONALD

Texas Tech University, Lubbock. Texas 79409

1. INTRODUCTION enclosed buildings and other structures by extreme, hurri-

cane, and tornado winds are so similar that it is almost
Use of deterministic design and evaluation criteria to meet impossibleto look at damage to an individual structure and
probabilistic performance goals is the approach used by tell which type of wind produced it. Thus the approach for
Department of Energy (DOE) facilities across the United determining wind pressures on buildings and other struc-
States as given in UCRL-15910, "Design and Evaluation tures proposedin this paper is consideredto be independent
Guidelinesfor Department of Energy Facilities Subjectedto of the type of wind storm. The recommendedprocedure is
Natural Phenomena Hazards" [Kennedy et al., 1990]. An essentially the same for straight, hurricane, and tornado
overview of the approach is presented in this paper. winds.
In this paper, three types of winds are discussed:extreme UCRL-15910 wind and tornado provisions use determin-
(straight), hurricane, and tornado. Extreme (straight)winds istic evaluationcriteria with the hazard annual probability of
are nonrotating such as those found in thunderstorm gust exceedance specified to obtain design/evaluation wind
fronts. Wind circulating around high- or low-pressure sys- speeds.The evaluation of responseand structureor equip-
tems are rotational in a global sense,but they are considered ment capacitiesis performed using deterministic methods
"straight" winds in the context of this paper. Tornadoesand which are familiar to engineers. Wind speedsare developed
hurricanes both have rotating winds. The diameter of rotat- from probabilistic hazard curves and are used to meet
ing winds in a smallhurricaneis considerablylargerthan the performancegoalsfor various building occupancies.
diameter of a large tornado. However, most tornado diam-
eters are relatively large in comparisonwith the dimensions
2. PERFORMANCE GOALS
of typicalbuildings.It is estimatedthat the diameterof 80%
of all tornadoes is greater than 300 ft. Performancegoals may be expressedin terms of annual
(Metric equivalents, in general, will not be provided. probabilityof exceedanceof somelevel of damage.Levels
However, the conversion factors are as follows: 1 in. = ---2.5 of damagecan include that damagebeyond which occupants
cm; 1 ft = ---30.5cm; 1 mi = --•1.6 km; and 1 lb = ---0.5 kg.) are endangered,beyond which hazardous materials cannot
Wind pressuresproducedby extremewindsare studiedin be confined, beyond which a facility cannot safely shut
boundary layer wind tunnels. The results are generally down, or beyond which a facility cannot continueits mis-
considered reliable because they have been verified by sion.The performancegoal for the safety of advancedlight
selectedfull-scale measurements.Investigationsof damage waterreactorsis of the orderof 1.0 x 10-5 annualproba-
producedby extremewindstendto supportthe wind tunnel bility of exceedancefor externaleventsinducedcore dam-
findings.Althoughthe rotatingnature of hurricaneand age [ElectricPower ResearchInstitute, 1990].Performance
tornadowinds cannoteasily be duplicatedin a wind tunnel, goalsuponwhichdesign/evaluation guidelinesare basedfor
damageinvestigationssuggestthat pressuresproducedon DOE facilities in UCRL-15910 are presentedin Table 1.
For each performancegoal there are separatewind and
tornado design and evaluation criteria. DOE management
The Tornado: Its Structure, Dynamics, Prediction, and Hazards.
categorize facilities or individual structures, systems, or
GeophysicalMonograph79 componentsdependingon the cost, missionimportance,or
Copyright1993by the AmericanGeophysicalUnion. hazardto peopleor the environment.For UCRL-15910,a
380 HAZARDOUS MATERIALS CONTAINMENT DESIGN

TABLE 1. PerformanceGoals for Each Usage Category

Performance Goal
Usage (Annual Probability of
Category Performance Goal Description Exceedance)*

General use maintain occupant safety ---10- 3 of theonsetof major

component damage to the extent
that occupants are endangered
Important or occupant safety, continued •-5 x 10-4 of component
damage
low hazard operation with minimal to the extent that the component
interruption cannot perform its function
Moderate occupant safety, continuedfunction, --10-4 of component
damage
to the
hazard confinement extent that the component cannot
perform its function
High hazard occupant safety, continued function, --10-5 of component
damage
to the
very high confidence of extent that the component cannot
confinement perform its function
Commercial
reactorS'

*Component refers to structure, equipment, or distributionsystem.

•-Currently beyond the scope of UCRL-15910.

DOE natural phenomena hazards panel selected reasonable suchas 1 x 10-3, or by (2) a median-centered
evaluation/
and achievable performance goals, bounded by current acceptanceapproach coupledwitha lessfrequent1 x 10-4
practice: hazard probability. UCRL-15910 uses the former approach
1. For ordinary facilities the performance goals are con- becauseconservative evaluation/acceptanceapproachesare
sistent with design according to conventional building code well established, extensively documented, and commonly
provisions. practiced.
2. For high-hazard facilities the performance goals are
comparableto performance reached by nuclear power plants 4. DESIGN AND EVALUATION CRITERIA FOR WIND LOAD
as measuredby probabilisticrisk assessments(PRAs). (Note
that the UCRL-15910 performance goals are related to A uniform approach to wind load determination that is
individual component behavior, while PRA results are in applicable to the design of new facilities and the evaluation
terms of core damage.) of existingfacilities has been developedfor DOE. A uniform
treatment of wind loads is recommended to accommodate
extreme, hurricane, and tornado winds. Buildings or facili-
3. PERFORMANCE GOAL ACHIEVEMENT
ties are first assignedappropriate usage categories. Criteria
Structure/equipment performance is a function of (1) the are recommended such that the performance goals for each
likelihood of hazard occurrence and (2) the strength of the category can be achieved. Proceduresaccordingto Ameri-
structure or equipment item. Therefore design and evalua- can Society of Civil Engineers(ASCE) 7-88 (formerly ANSI
tion criteria have been developed to attain performance A58.1) [American Society of Civil Engineers, 1990] are
goals by (1) specificationof hazard probability for definition recommended for determining wind loads produced by
of wind and tornado loadings and (2) specification of re- straight,hurricane,and tornado winds. The extreme wind/
sponse evaluation methods, acceptancecriteria, and design tornado hazard models developed for DOE sites were used
detailing requirements with controlled levels of conserva- to establishsite-specificcriteria for each of some 25 DOE
tism. Acceptable performance can only be achieved by sites [Coats and Murray, 1985].
consistent specification of all design or evaluation criteria The performancegoals establishedfor general use and
elements. important or low-hazard usagecategoriesare met by con-
UCRL-15910 wind and tornado design and evaluation ventional building codes or standards [see International
guidelinescontain the following provisions:(1) lateral force Conferenceof Building Officials, 1988]. These criteria do not
provisions, (2) damage control provisions, (3) detailing pro- account for the possibility of tornado winds, becausewind
visions, and (4) quality assuranceand peer review provi- speedsassociatedwith extreme winds typically are greater
sions. than thosefor tornadoesat exceedanceprobabilitiesgreater
Wind/tornado performance depends on the level of the thanapproximately
1 x 10-4. For this reason,tornado
hazard and on conservatism in the responseevaluation and designcriteriaarespecifiedonlyfor buildingsandfacilitiesin
acceptancecriteria. For example,a performancegoalof 1 x moderateand high-hazardcategories,where hazard exceed-
evaluation/ anceprobabilities
10-4 canbeachievedeitherby (1) a conservative arelessthan1 x 10-4.
acceptanceapproachfor a morefrequenthazardprobability, The traditionalapproachfor establishingtornado
 MURRAY AND MCDONALD 381

lO7 -7
! 10
I ! I I

Tornado Region

6
10
Wind -Tornado ORNL
Transition

Tornado Region

SLAC

Straight Wind Region

1½

2
-2
10
10

ORNL Design Criteria Considers Effects of Tornadoes -

SLAC Design Criteria Does Not _

1
-1
lO
lo

5O lOO 150 200 250 300 350

WindSpeed (mph)

Fig. 1. Straightwind and tornado regionsof wind hazard curves.

criteria is to selectextremely low exceedanceprobabilities. significantlygreater than the uncertaintiesin earthquake

For example, the exceedance probability for design of probabilityassessment.The strongestargumentagainstus-
commercial nuclear powerplantsis 1 x 10-7 [U.S.Nuclear ing low-probabilitycriteria is that a relatively shortperiodof
Regulatory Cornmission, 1974; Kimura and Budnitz, 1987; record (40 years) must be extrapolatedto extremely small
Ravindra and Nafday, 1990]. There are reasonsfor depart- exceedance probabilities. For these reasons, an alternative
ing from this traditional approach. The low exceedance approachwas used in the UCRL-15910 guidelines.
value for commercialnuclearpower plantswas established
circa 1960, when very little was known abouttornadoesfrom
Establishmentof Wind and Tornado Hazard Annual
an engineering perspective. Much has been learned about Probabilities
tornadoessincethat time. Use of a low-hazardprobabilityis
inconsistentwith the practice relating to other natural haz- The rationale for establishingtornado criteria is described
ards, such as earthquakes. There are many uncertaintiesin below. Figure 1 showsthe tornadoand straightwind hazard
tornado hazard probability assessment,but they are not curvesfor two DOE sites(SLAC and ORNL). The wind
382 HAZARDOUS MATERIALS CONTAINMENT DESIGN

at the intersection
of the tornadoand straightwind curvesis the same usagecategory, but at different geographicalloca-
definedfor purposesof this discussionas the transitionwind tions, will have different input loading specifiedto achieve
speed. An exceedanceprobability is associatedwith each the same performance goal.
transitionwind speed.If the exceedanceprobabilityof the The minimum wind design criteria for each of the four
transition
windspeed
islessthan1 x 10-5 peryear,tornadoesusage categories are summarized in Table 2. The recom-
are not a viablethreatto the site,becausestraightwind speeds mended basic wind speedsfor extreme wind, hurricanes,
are higherthantornadoesfor a givenreturnperiod.Thusfrom and tornadoesare containedin Table 3. All wind speedsare
Figure 1, tornadoesneed not be consideredat SLAC, but they fastestmile. Minimum recommendedbasicwind speedsare
should be considered at ORNL. also noted in the table.
Mean wind and tornado hazard curves were developedfor
each DOE site, along with the transition wind speed.Those
General Use Facilities
sites with transition wind speed exceedance probabilities
greaterthan 10-5 shouldbe designed
for tornadoes;
others The performance goals for general use facilities are con-
should be designedfor extreme winds or hurricanes.
sistent with objectives of ASCE 7-88 building class I, ordi-
The tornado wind speed is obtained by selectingthe wind
nary structures.The wind force resisting structural system
speedassociated
withanexceedance probability
of2 x 10-5 should not collapse under design load. Survival without
peryear.Thevalueof 2 x 10-5 isthelargestonethatcanbe collapseimpliesthat occupantsshouldbe able to find an area
used and still represent a point on the tornado hazard curve. of relative safety inside the building. Breach of the building
For example, the tornado wind speed for the ORNL is 130 envelope is acceptable, since confinement is not essential.
mph (peak gust at 10 m). Flow of air through the building and water damage are
A comparison of the slopes of the tornado hazard curves
acceptable. Severe damage, including total loss, is accept-
for the DOE sites reveals that the slopes are essentially the able, so long as the structure does not collapse.
same even though the transition wind speeds are different.
The ASCE 7-88 standardcalls for the basic wind speedto
The criteria required to meet the performance goals of
be basedon an exceedanceprobability of 0.02 per year. The
moderate and high-hazard facilities can be met by using importance factor for this class of building is 1.0. For those
multipliers that are equivalent to an importance factor in the sites within 100 miles (160 km) of the Gulf of Mexico or
ASCE 7-88 designprocedure. The multipliers are specifiedin Atlantic coastlines, a slightly higher importance factor is
lieu of two different exceedance probabilities for moderate recommended to account for hurricanes.
and high-hazardfacilities. The value of the importancefactor Terrain surrounding the facilities should be classified as
is selected to achieve lower probability of tornado damage exposure B, C, or D, as appropriate. Gust responsefactors
for high-hazard facilities compared with moderate hazard and velocity pressure exposure factors should be used
facilities. The importance factors are then chosen to meet according to rules of the ASCE 7-88 procedures.
the performance goals. Wind pressuresare calculated on the walls and roofs of
In general, designcriteria for each usage category include enclosedbuildingsby appropriate pressurecoefficients spec-
the following: (1) annual hazard exceedance probability, (2) ified in the ASCE 7-88 standard. Openings, either of neces-
importance factor, (3) wind-generated missile parametersfor sity or created by wind forces or missiles, result in internal
moderate and high-hazard facilities, and (4) tornado param- pressuresthat can increasewind forces on componentsand
eters for moderate and high-hazard facilities, if applicable. cladding. The worst cases of combined internal and external
The criteria are formulated in such a way that a uniform pressuresshouldbe considered,as required by the standard.
approachfor determiningdesignwind loads, per ASCE 7-88, Structures in the general use category may be designedby
can be used for extreme, hurricane, and tornado winds. either allowable stressdesign (ASD) or strength design (SD)
In order to apply the ASCE 7-88 procedure, tornado gust as appropriatefor the material used in construction.Except
wind speeds must be converted to fastest-mile speeds. when applicable codes provide otherwise, plausible load
Appropriate gust response factors and velocity pressure combinations shall be considered to determine the most
exposurecoefficientsare utilized in the processof determin- unfavorable effect on the building, foundation, or structural
ing wind loads. Appropriate exposure categoriesalso are memberbeingconsidered.When usingASD methods,allow-
considered in the wind load calculations. Open terrain able stressesappropriate for the building material shall be
(exposure C) should always be assumedfor tornado winds, used with the following combinations that involve wind'
regardlessof the actual terrain conditions.
DL + WL (1)

Criteria Jbr Design of Facilities 0.75(DL + WL + LL) (2)

The criteria presented herein are consistentwith the where DL is dead load, LL is live load, and WL is wind load.
performancegoalsdescribedearlier. Buildingsor facilitiesin When using SD methods, the following load combinations
each category have a different role and represent different that involve wind are recommended:
levels of hazard to people and the environment.In addition,
the degreeof wind hazard varies geographically.Facilities in 0.9DL + 1.3WL
 MURRAY AND MCDONALD 383

TABLE 2. Summaryof Minimum Wind DesignCriteria

Important or
Building Category General Use Low Hazard Moderate Hazard High Hazard

Wind
Annual probability of 2 x 10-2 2 x 10-2 1 x 10-3 1 x 10-4
exceedance
Importance factor* 1.0 1.07 1.0 1.0
(1.05) (I.11) tl.05) (1.05)
Missile criteria 2 x 4 timber plank 15 lb at 50 2 x 4 timber plank 15 lb at 50
mph (horiz.); maximum height mph (horiz.); maximum height
30 ft 50 ft

Tornado
Annual hazard probability 2 x 10-5 2x 10-5
of exceedance
Importance factor I= 1.0 I = 1.35
APC? 40 psf at 20 psffs 125 psf at 50 psffs
Missile criteria 2 x 4 timber plank 15 lb at 100 2 x 4 timber plank 15 lb at 150
mph (horiz.); maximum height mph (horiz.); maximum height
150 ft; 70 mph (vert.) 200 ft; 100 mph
3-in. diameter standard steel pipe, 3-in. diameter standard steel pipe,
75 lb at 50 mph (horiz.); 75 lb at 75 mph (horiz.);
maximum height 75 ft; 35 mph maximum height 100 ft; 50 mph
(vert.) (vert.)
3000-lb automobile at 25 mph,
rolls and tumbles

*The first value representsthe importancefactorfor sites 100miles (160 km) or more inland. The value in parenthesesshouldbe usedat
the coastline. Linear interpolation between the two values shouldbe usedfor sites within 100 miles t 160 km) of the coastlineto account for
hurricanes.
?Atmosphericpressurechange.

1.2DL + 0.5LL + 1.3WL importancefactors applied, the determination of wind loads

(4)
on important or low-hazard category structuresis identical
The SD method requires that the strength provided be to that describedfor general use category structures. Facil-
greater than or equal to the strength required to carry the ities in this category may be designed by ASD or SD
factored loads. Appropriate strengthreductionfactors must methods,as appropriate, for the constructionmaterial. The
be applied to the nominal strengthof the materialbeingused. load combinations described for general use structures are
the same for important or low-hazard structures. However,
Important or Low-Hazard Facilities greater attention should be paid to connectionsand anchor-
Important or low-hazardfacilitiesare equivalentto essen- ages for main members and components, such that the
tial facilities (class II), as defined in ASCE 7-88. The integrity of the structure is maintained [see McDonald,
structure'smain wind force resistingstructuralsystemsshall 1988].
not collapseat designwind speeds.Completeintegrityof the
building envelope is not required becauseno significant Moderate Hazard Facilities
quantities of toxic or radioactive materials are present.
However, breach of the building envelope may not be The performancegoal for moderate hazard facilities re-
acceptableif wind or water interfereswith the facility quires more rigorous criteria than is provided by standard
function. If loss of facility function is caused by water building codes. For some DOE sites, tornadoes must be
damageto sensitiveequipment,collapsedinteriorpartitions, considered.
or excessivedamageto HVAC ductsand equipment,then Extreme Winds and Hurricanes. For those sites where
lossof claddingand missileperforationat the designwind tornadoes are not a viable threat, the recommended basic
speeds must be prevented. wind speedis based on an annualexceedanceprobability of
An annualwind speedexceedanceprobabilityof 0.02 is I x 10-3. The importance
factoris 1.0. For thosesites
specified,but the importancefactor for importantor low- located within 100 miles (160 kin) of the Gulf of Mexico or
hazardcategorystructuresis 1.07. For thosesiteslocated Atlantic coastlines, a slightly higher importance factor is
within 100miles(160kin) of the Gulf of Mexicoor Atlantic specifiedto account for hurricanes.
coastlines,a slightlyhigherimportancefactor is usedto Once the basic wind speeds are established and the
account for hurricane winds. importancefactors applied, determinationof moderatehaz-
Once the designwind speedsare establishedand the ard categorywind loads is identicalto that describedfor
384 HAZARDOUS MATERIALS CONTAINMENT DESIGN

TABLE 3. RecommendedBasic Wind Speeds(in Miles per Hour) for DOE Sites
BuildingCategory/Fastest-Mile
Wind Speedsat 10-mHeight
General Important or Moderate Hazard/ High Hazard/
Use/Wind Low
DOE Project Sites Hazard/Wind Wind Tornado Wind Tornado

Bendix Plant, Mo. 72 72 "' 144 -" 144

Los AlamosNational Laboratory, N.M. 77 77 93 "- 107 -"
Mound Laboratory, Ohio 73 73 "' 136 --- 136
Panrex Plant, Tex. 78 78 "' 132 -" 132
Rocky Flats Plant, Colo. 109 109 138 -"? 161 -"?
SandiaNational Laboratories,Albuquerque,N.M. 78 78 93 "- 107 "-
Sandia National Laboratories, Livermore, Calif. 72 72 96 --- 113 -'-
Pinel!as Plant, Fla. 93 93 130 '-- 150 "-
Argonne National Laboratory--East, II1. 70* 70* "' 142 --- 142
Argonne National Laboratory--West, Idaho 70* 70* 83 "- 95 '"
Brookhaven National Laboratory, N.Y. 70* 70* "' 955 '" 95•
Princeton Plasma Physics Laboratory, N.J. 70* 70* "' 103 --- 103
Idaho National EngineeringLaboratory, Idaho 70* 70* 84 '-- 95 '-'
Feed Materials Production Center, Ohio 70* 70* "' 139 --' 139
Oak Ridge National Laboratory, X-10, K-25, and 70* 70* "' 113 -" 113
Y-12, Tenn.
Paduach Gaseous Diffusion Plant, K.Y. 70* 70* "' 144 --' 144
Portsmouth Gaseous Diffusion Plant, Ohio 70* 70* "' 110 --- 110
Nevada Test Site, Nev. 72 72 87 "- 100 ---
Hanford Project Site, Wash. 70* 70* 80* "- 90* '"
Lawrence Berkeley Laboratory, Calif. 72 72 95 "' 111 '-'
Lawrence Livermore National Laboratory, Calif. 72 72 96 "' 113 '"
Lawrence Livermore National Laboratory, Site 300, 80 80 104 --- 125 '--
Calif.
Energy Technology and Engineering Center, Calif. 70* 70* "' 955 '" 955
Stanford Linear Accelerator Center, Calif. 72 72 95 "' 112 "'
Savannah River Plant, S.C. 78 78 "' 137 "' 137

*Minimum extreme wind speed.

?Although extreme winds govern at Rocky Flats, it is recommendedthat facilities be designedfor the tornado missile criteria. APC need
not be considered.
5Minimum tornado speed.

general use category. Facilities in this category may be hurricane winds, or weak tornadoes. A 2 x 4-in. timber
designedby ASD or SD methods, as appropriate, for the plank weighing 15 lb is the specifiedmissile. Its impact speed
material being used in construction. Plausible load combina- is 50 miles per hour (mph) at a maximum height of 30 ft
tions shall be considered to determine the most unfavorable above ground level. The missile will break glass; it will
effect on the building, foundation, or structural member perforate sheet metal siding, wood sidingup to 3/4 in. thick,
being considered. When using ASD, allowable stresses or form board. The missile could pass through a window or
appropriate for the building material shall be used with the a weak exterior wall and causepersonal injury or damage to
following load combinations: interior contentsof a building. The specified missile will not
0.9(DL + WL) (5) perforate unreinforced masonry or brick veneer walls or
other more substantial walls.
0.67(DL + WL + LL) (6) Tornadoes. For those sites requiring design for torna-
The SD load combinations recommended for the moderate does, the criteria are based on site-specificstudies.The basic
hazard category are wind speedis associatedwith an annual hazard probability
of exceedance
of 2 x 10-5. The windspeedobtainedfrom
J)L + 1.3WL (7) the tornado hazard model is converted to fastest mile. The
importancefactor for the moderatehazard categoryis 1.0.
1.1DL + 0.SLL + 1.2WL (8)
With the wind speedconvertedto fastest-milewind and an
Greater attention should be paid to connectionsand anchor- importancefactor of 1.0, equationsin the ASCE standard
agesfor main membersand components,suchthat the integrity should be used to obtain design wind pressures on the
of the structure is maintained [see McDonald, 1988]. structure.Exposurecategory C shouldalways be used with
A minimum missile criterion is specified to account for tornado winds regardless of the actual terrain roughness.
objects or debris that could be picked up by extreme winds, The velocity pressure exposure coefficient and the
 MURRAY AND MCDONALD 385

responsefactorare obtainedfrom appropriatetablesin the horizontal speedcould be effective to heightsof 75 ft above

ASCE standard.External pressurecoefficientsare used to ground level. The missile will perforate conventionalmetal
obtain tornado wind pressures on various surfacesof the siding,sandwichpanels, wood and metal deckingon roofs,
structure. A distinction is made between the main wind force and gypsum panels. In addition, it will perforate unrein-
resistingsystemand components and cladding. forced concrete masonry and brick veneer walls, reinforced
If the building is not specificallysealed to maintain an concrete masonry walls less than 8 in. thick, and reinforced
internal negative pressure for confinement of hazardous concrete walls less than 6 in. thick. Although wind pressure,
materials, orif openingsgreaterthan1ft2(•--0.1m2)per1000 APC, and missile impact loads can act simultaneouslyin a
ft3(-•28m3)ofvolume arepresent, orif openingsofthissize tornado, the missile impact loads can be treated indepen-
can be created by missile perforation, then the effects of dently for designand evaluation purposes.
internal pressureshouldbe consideredaccordingto ASCE
procedures. If the building is sealed, then atmospheric High-Hazard Facilities
pressure change (APC) pressures associatedwith the tor- The performancegoalcan be achievedfor thiscategoryif
nado should be considered. the main wind force resisting members do not collapse,
APC pressureis half its maximumvalue at the radius of structuralcomponents do not fail, andthe buildingenvelope
maximum wind speed in a tornado. Thus critical tornado is not breachedat the designwind loads. Loss of cladding,
loadingwill be one-halfthe maximumAPC pressureplusthe broken windows, collapsed doors, or significantmissile
maximum tornado wind pressure. A loading condition of perforations mustbe prevented.Strongair flow throughthe
APC alone can occur on the roof of a buried tank or sand buildingor water damagecannotbe tolerated.
filter, if the roof is exposed at the ground surface. APC Extreme Winds and Hurricanes. For those sites that do
pressurealways acts outward. The effect of rate of pressure not require specificdesignfor tornado resistance,the rec-
changeon ventilation systemsshouldbe analyzed to assure ommendedbasicwind speedis basedon an annualhazard
that it does not interrupt any function or processescarried exceedance probabilityof 1 x 10-4. Theimportance factor
out in the facility. Procedures and computer codes are is 1.0. The wind speed is fastest mile at an anemometer
available for such analyses. heightof 10 m above groundlevel. Once the basic wind
Plausible load combinations shall be considered to deter- speedsare establishedand the importancetactors are ap-
mine the most unfavorable effect on the building, founda- plied, determinationof high-hazardfacility wind loads is
tion, or structural member being considered.When using identical to that described for the general use category.
ASD methods,allowable stressesappropriatefor the build- Facilitiesin this categorymay be designedby ASD or SD
ing materialsshallbe usedwith the followingloadcombina- methods,as appropriate,for the materialbeing used in
tions that involve tornado loading: construction. Recommended wind load combinations are the
same as for moderate hazard facilities. Greater attention
0.75(DL + Wt) (9) shouldbe paid to connectionsand anchoragesfor main
membersand components,such that the integrity of the
0.63(DL + Wt + LL) (10)
structureis maintained [see McDonald, 1988].
The SD load combinations recommended for the moderate The missile criteria are the same as for the moderate
hazard category are hazardcategory,exceptthat the maximumheightachieved
by the missileis 50 ft insteadof 30 ft.
DL + Wt (11) Tornadoes. For those sites requiring designfor tornado
resistance,the criteria are basedon site-specificstudies.The
DL + LL + Wt (12)
recommended
basicwind speedis associated
with an annual
whereWt is tornadoloading,including APC,asappropriate. hazard probabilityof exceedance of 2 x 10-5 (thesameas
Two missilesare specifiedas minimumcriteriafor this the moderatehazard category). The wind speedobtained
usagecategory. The2 x 4 in. timberplankweighing 15lb is from the tornado hazard model is converted to fastest mile.
assumed to travel in a horizontal direction at a speedup to The importance factorfor the high-hazardcategoryis 1.35.
100mph.Thehorizontal speedis effective upto a heightof With the wind speed expressedas tastestmile and an
150ft abovegroundlevel.If carriedto a greatheightby the importance factorof 1.35,ASCE 7-88equations shouldbe
tornadowinds,the timberplankcouldachievea terminal used to obtain design wind pressures on the structure.
verticalspeedof 70 mph in fallingto the ground.The ExposurecategoryC shouldalwaysbe usedwith tornado
horizontalandverticalspeeds areassumed to beuncoupled, windsregardlessof actualterrainroughness. The velocity
andtheyshould notbecombined. Themissile willperforate pressureexposurecoefficientandthe gustresponsefactor
mostconventionalwall and roof claddingexceptreinforced areobtainedfromappropriatetablesin the ASCE standard.
masonry or concrete. The cellsof concrete masonry walls Externalpressurecoefficientsare usedto obtaintornado
mustbefilledwithgroutto preventperforation bythetimber wind pressureson varioussurfacesof the structure.A
missile.The secondmissileis a 3-in.-diameterstandardsteel distinction is made between the main wind force resisting
pipe,whichweighs
75lb. It canachieve impact systemandcomponents
a horizontal andcladdingin determining
wind
speedof 50 mphand a verticalspeedof 35 mph.Its
386 HAZARDOUS MATERIALS CONTAINMENT DESIGN

TABLE 4. Ratio of Hazard Probability to Performance Goal

Ratio of Hazard
Probability to
Wind/Tornado Performance
Usage Category Performance Goals Hazard Probability Goal

Extreme Winds
General use 10-3 2 x 10-2 20
Important or low hazard 5x 10-4 10-2* 20

Moderate hazard 10-4 10-3 10

Highhazard 10-5 10-4 10
Tornadoes
Moderate hazard 10-4 2 x 10-5 <i
Highhazard I0-5 3 x 10-67 <1
*Here 2 x 10-2 with I = 1.07--- 10-2
?Here2 x 10-5 with I = 1.35--- 3 x 10-6.

If the buildingis sealedto confinehazardousmaterials,the Wall barriers tested to date include reinforced concrete
wind and APC load combinationsspecified for the moderate walls from 4 in. to 10 in. thick; 8-in. and 12-in. walls of
hazard usagecategory also shouldbe used for this category. reinforced concrete masonry units (CMU); two other ma-
The effectsof rate of pressurechangeon ventilating systems sonry wall configurationsconsistingof an 8-in. CMU and a
shouldbe analyzed. Recommendedtornado wind load com- 4-in. clay brick veneer; and a 10-in. composite wall having
binations for moderate hazard facilities also apply to high- two wythes of 4-in. clay brick.
hazard facilities. The impact tests series was designed to determine the
Three missilesare specified as minimum criteria for this impact speedrequired to producebackface spall of each wall
usagecategory.The 2 x 4- in. timber plank weighing15 lb is barrier. A set of 15 wall sections have been constructed and
assumed to travel in a horizontal direction at speeds up to tested at this time. Preliminary findings suggestthat all cells
150 mph. The horizontal missile is effective to a maximum of CMU walls must be grouted to prevent missile penetra-
height of 200 ft above ground level. If carried to a great tion. Results of these impact tests as well as previous
height by the tornado winds, it could achieve a terminal available test data will be compared with existing impact
speed in the vertical direction of 100 mph. The horizontal formulas. In particular, the Rotz Equation and the Modified
and vertical speedsare uncoupled and should not be com- NDRC formula are being examined. A new impact formula
bined. The missilewill perforate most conventionalwall and for CMU walls is being developed.
roof claddingexcept reinforced masonry and concrete. Each
celloftheconcrete
masonry
shall
contain
a •-in.-diameter
Comments on Load Combinations
rebar and be groutedto prevent perforation by the missile. The ratios of hazard probabilities to performance goal
The second missile is a 3-in.-diameter standard steel pipe, probabilitiesfor the usagecategoriesas shownin Table 4 are
which weighs75 lb. It can achievea horizontalimpactspeed an approximatemeasureof the conservatismrequiredin the
of 75 mph and a vertical speedof 50 mph. The horizontal designto achieve the performancegoals. The most conser-
speedcouldbe effectiveat heightsup to 100ft aboveground vatism is needed in the responseevaluation and acceptance
level. This missile will perforate unreinforced concrete ma- criteria for design of general use and important or low-
sonry and brick veneer walls, reinforced concrete masonry hazard facilities. Somewhat less conservatism is needed for
walls less than 12 in. thick, and reinforced concrete walls moderateandhigh-hazardfacilities.The hazardprobabilities
less than 8 in. thick. The third missile is a 3000-lb automobile
specifiedfor tornadoesare lessthan the performancegoal
that is assumedto roll and tumble on the ground and achieve probabilities.Hencethe performancegoalsare theoretically
an impact speedof 25 mph. Impact of an automobilecan met with no added conservatism in the design.
cause excessivestructural responseto columns, walls, and Conservatism can be achieved in designs by specifying
frames. Impact analysesshouldbe performedto determine factorsof safetyfor allowablestressdesign(ASD) and load
specificeffects. Collapse of columns,walls, or frames may factorsfor strengthdesign(SD). Consistentwith the ratiosin
lead to further progressivecollapse. Table 4, the loadingconditionsrecommendedfor designfor
DOE facilities are summarized in Table 5.
Tornado Missile Impact Tests Since the ratio of extreme wind hazard probability to
Wall barrier specimenshave been tested at the Tornado performancegoalprobabilityfor generaluseand important
Missile Impact Facility at Texas Tech University. The or low-hazardfacilities are the largest, 20, designsfor these
facility hasan air-activatedtornadomissilecannoncapable categoriesshouldbe the most conservativein terms of
of firing2 x 4 timer planksweighing12lb up to 150mphand factorsof safetyfor ASD and load factorsSD. The recom-
3-in.-diametersteelpipesweighing75 lb at speedsto 75 mph. mendedcombinationsare essentiallythose given in
 MURRAY AND MCDONALD 387

TABLE 5. Summary of RecommendedWind and Tornado Load Combinations

Facility GeneralUse Importantor Low Hazard ModerateHazard High Hazard
ASD
Extreme winds DL + WL DL + WL 0.9tDL + WL) 0.9(DL + WL)
0.75(DL + LL + WL) 0.75(DL + LL + WL) 0.67(DL + LL + WL) 0.67(DL + LL + WL)
Tornadoes 0.75(DL + W t) 0.75(DL + Wt)
0.63(DL + LL + Wt) 0.63(DL + LL + Wt)
SD
Extreme winds 0.9DL + 1.3WL 0.9DL + 1.3WL DL + 1.3WL DL + 1.3WL
1.2DL + 0.5LL + 1.3WL 1.2DL + 0.5LL + 1.3WL I.IDL + 0.5LL + 1.2WL 1.1DL + 0.5LL + 1.2WL
Tornadoes DL+ Wt DL+ Wt
DL + LL + Wt DL + LL + W t

ASD, allowablestressdesign;useallowablestressappropriate
for buildingmaterialSD, strengthdesign;useqbfactorsappropriate
for
buildingmaterial.DL, deadload;LL, live load;WL, extremewind load;and Wt, tornadoload, includingAPC if appropriate.

7-88 [American Society of Civil Engineers, 1990]. The rec- probabilisticbasis, but responseevaluationmethodsand
ommendedload combinationsfor moderateand high-hazard acceptable behaviorlimitsare deterministicapproaches with
facilities are slightly less conservative than for the general which designengineersare familiar.
use and important or low-hazard categoriesbecausethe ratio
of extreme wind hazard probability to performance goal REFERENCES
probability is less. The load factor coefficientshave been AmericanConcreteInstitute, Code requirementsfor nuclearsafety-
reduced by approximately 10%. related concrete structures (ACI 349-85) and commentary (ACI
The tornado hazard probabilities for both moderate and 349R-85), Detroit, Mich., 1985.
high-hazardfacilitiesare lessthan the performancegoal prob- American Institute of Steel Construction, Nuclear facilities4teel
abilities. The tornado load combinations for both ASD and safety-relatedstructuringfor design,lhbrication,and erection,
ANSI/AISC N690, Chicago, II1., 1984.
SD recognize that the performancegoals are theoretically AmericanSocietyof Civil Engineers,Minimum designloadsfor
met and no added conservatism in the load combination is buildingsand other structures,ASCE 7-88 (formerly ANSI
required. This approachis consistentwith criteria for com- A58.1), New York, 1990.
Coats, D. W., and R. C. Murray, Natural phenomenahazards
mercial nuclear power plants as given in ACI 349 [American
modelingproject:Extremewind/tornadohazardmodelsfor De-
Concrete Institute, 1985] for concrete and ANSI/AISC N690 partmentof Energysites,Rep. UCRL-53526Rev. 1, Lawrence
[American Institute of Steel Construction,1984]for steel. Livermore Natl. Lab., Livermore, Calif., 1985.
Additional background material on wind and tornado Electric Power Research Institute, Advanced Light Water Reactor
effects can be found in the work of McDonald [1985a, b]. A Requirements Document,vol. '[, ALWRPolicyandSutnrna•'of
Top-TierRequirements,Palo Alto, Calif., 1990.
commentaryon the designcriteriasummarized thereis also InternationalConference of Building Officials, Uniform Building
includedin AppendixB of Kennedyet al. [1990]. Code, 1988 ed., Whittier, Calif., 1988.
Kennedy,R. P., S. A. Short, J. R. McDonald,M. W. McCann,
R. C. Murray,andJ. R. Hill, Designandevaluation
guidelines
for
5. SUMMARY AND CONCLUSIONS
Departmentof Energyfacilitiessubjectedto naturalphenomena
UCRL-15910 is an example of deterministicdesign/ hazards,Rep. UCRL-i5910, Lawrence Livermore Natl. Lab.,
Livermore, Calif., 1990.
evaluationcriteriadevelopedto achieveprobabilistic
perfor- Kimura, C. Y., and R. J. Budnitz, Evaluationof externalhazardsto
mancegoals.UCRL-15910alsocoversseismic
andfloodas nuclearpowerplantsin the United States,NUREG/CR-5042,
well as wind and tornado criteria as discussedin this paper. Rep. UC1D-21223,LawrenceLivermoreNatl. Lab., Livermore,
The criteria developedare consistentwith the consensus Calif., 1987.
McDonald,J. R., Extreme windsand tornadoes:An overview,Rep.
standardASCE 7-88 and requiredby DOE GeneralDesign UCRL-15745, Lawrence Livermore Natl. Lab., Livermore, Ca-
Criteria[U.S. Departmentof Energy,1989].Conservatism
is lif., 1985a.
specifiedwhich is sufficientto achievethe performanceMcDonald, J. R., Extreme winds and tornadoes:Designand eval-
goals.This conservatismincreasesfromthe generaluseto uationof buildingsand structures,Rep. UCRL-15747,Lawrence
Livermore Natl. Lab., Livermore, Calif., 1985b.
the high-hazard usage category. McDonald, J. R., Structural details for wind design,Rep. UCRL-
The UCRL-15910 wind/tornado design and evaluation 21131, LawrenceLivermore Natl. Lab., Livermore,Calif., 1988.
guidelines
followthephilosophy
of (1)gradualreduction in Ravindra,M. K., and A.M. Nafday, State-of-the-artand current
hazardannualexceedanceprobabilityand (2) gradualin- researchactivitiesin extreme windsrelatingto designand evalu-
crease in conservatism of evaluation procedureas one goes ationof nuclearpowerplants,Rep.NUREG/CR-5497,
U.S. Nucl.
Reg. Comm., Washington,D.C., 1990.
from a generaluseto a high-hazardfacility.Fourseparate U.S. Departmentof Energy,Generaldesigncriteria,DOE Order
setsof design/evaluationcriteriahavebeenpresented in 6430.1A, Washington, D.C., 1989.
UCRL-15910,eachwitha different performance goal.In all U.S. Nuclear Regulatory Commission,Regulatoryguide 1,76,
thesecriteria, loadingis selectedfrom hazardcurveson a Washington,D.C.,
 State-of-the-ArtandCurrentResearchActivitiesin ExtremeWindsRelating
to Design and Evaluation of Nuclear Power Plants
M. K. RAVINDRA

EQE International, Irvine, Calijbrnia 92715

INTRODUCTION modelingof tornadoes, tornado missile trajectory, and im-

pact effects on structural barriers was sponsored by the
The objective of this paper is to review resultsfrom recent
Nuclear Regulatory Commission and the Electric Power
and ongoing research projects on extreme winds and sum-
Research Institute (EPRI) in the 1970s in order to develop
marize the current state of the art as a backgroundfor the
the design procedures that are included in the regulatory
developmentof criteria for individual plant examinationof
criteria. The wind speed data collected at over 130 weather
nuclear power plants for external events (IPEEE). Past
stations were examined systematically to develop mean
studies have shown that extreme winds from tornadoes,
annual recurrence intervals for different wind speeds near
hurricanes, and extratropical stormsmay be an important
nuclear power plant sites [Simiu et at., 1979].
external event for some nuclear power plants in the United
In the last 10 years, probabilistic risk assessment(PRA)
States. The evolution of tornado design criteria in the
studies of nuclear power plants have considered extreme
nuclearindustry is traced, startingwith U.S. Nuclear Reg- winds as external event initiators of accidents and estimated
ulatory Commission (NRC) Regulatory Guide 1.76 and the
the frequencies of such accidents. These studies included the
StandardReview Plan (SRP) and concludingwith American
wind hazard analysis, wind fragility evaluation of structures
Nuclear Society (ANS) Standard 2.3. The review covers the
and equipment, and systemsanalysisof accident sequences.
hazard analysis for tornadoes includingtornado missilesand It was found that the contribution of extreme winds to the
straight winds. Fragility analysis of structuresfor extreme
overall risk of nuclear power plants is highly plant specific
wind loading is discussed. The techniques used in the
and may not be insignificant [Kimura and Budnitz, 1987].
systemsanalysis and quantification are describedalong with Similar conclusionswere drawn from the studies performed
a summary of the results from different probabilistic risk by Sandia National Laboratories as part of Task Action Plan
assessments. A study was performed for those nuclear
(TAP) A-45, "Decay Heat Removal Requirements."
power plants that conform to the current licensingcriteria In order to provide information that would be useful in
specified in the Standard Review Plan to show that these
preparingguidelinesfor extreme wind analyseswithin IP-
plants are likely to have a mean core damagefrequencyof EEE, current information on wind design procedures, wind
lessthan 10-6/yr. Finally, the proceduredeveloped
for hazard, fragility, and systems analysis is needed. For this
extreme wind analysisin IPEEE is described. purpose,the NRC funded researchat Lawrence Livermore
Code of Federal Regulations (CFR) Title 10 part 50 National Laboratory (LLNL) to survey the state of the art in
appendixA criterion 2 requiresthat nuclearpower plantsbe the treatment of extreme winds in nuclear power plant
designedto withstand the effects of extreme winds such as designand risk assessment[Ravindra and Na./day, 1990].
tornadoes, hurricanes, and extratropical winds. These ef-
fects include wind pressures,pressuredrop, and missile
CURRENTNRC REQUIREMENTS
impacts.The proceduresfor designagainstextremewinds
havebeendevelopedover the last30yearsandaredescribed Title !0 of the Code of Federal Regulations (10 CFR)
in regulatoryguides(RG), the StandardReview Plan, and specifies,in general terms, the conditionsand factors that
industry standards. mustbe consideredin constructing,licensing,and operating
Extensive analytical and experimentalresearchin the a nuclear power plant and the regulatory processthat must
be followed in performing this function. The regulation of
nuclearpower plantsfor protectionfrom highwinds/tornado
The Tornado: Its Structure, Dynamics,Prediction,and Hazards.
GeophysicalMonograph79 hazards is given in 10 CFR 50 (design) and 10 CFR 100
Copyright1993by the AmericanGeophysicalUnion. (siting).
390 EXTREME WINDS AND NUCLEAR POWER PLANTS

NORTH
DAKOTA
MINNESOTA

IDAHO

SOUTH
DAKOTA• WISCONSIN

NEWYORK

IOWA PENNSYLVAN IA
NEVADA NEBRASKA
•

J II/
INOIS

KANSAS
MISSOURI
KENTUCKY

NO. CAROLINA

ARIZONA OKLAHOMA

TEXAS . .__.._....

Fig. 1. Tornado regionalizationmap of the United States.

The scopeof regulation of nuclear power plant designand areas, etc., to be considered in the study of tornado risk to
requirementsfor acceptability have evolved over the years. ensure (1) the integrity of the reactor coolant pressure
Initially, the General Design Criteria were adopted as the boundary, (2) the capability to shut down the reactor and
minimum requirementsin developing the standardsfor de- maintain it in a safe shutdown condition, and (3) the capa-
tailed design.Safety guidesissuedin 1970becamepart of the bility to prevent accidents that could result in potential
Regulatory Guide series in 1972. The specific regulatory off-siteexposuresthat are a significantfraction of the guide-
documents of interest are the NRC Regulatory Guides 1.76 line exposure of 10 CFR 100. The portions of structures,
and 1.117. RG 1.76 defines three tornado regions in the systems,or componentswhose continuedfunction is not
United States based on the regional variations of tornado required but whose failure could reduce the functional
hazard adopted in WASH-1300 [Markee et al., 1974] and capabilityof any plant feature includedin the specifiedlist
specifies a design basis tornado (DBT) for each of these are also required to be designedfor the designbasistornado.
regions(Figure 1). DBT for each region is describedin terms Detailedrequirementsfor tornadoloadingsare coveredin
of the maximum wind speed, rotational wind speed, trans- NRC Standard Review Plan section 3.3.2 and are essentially
lational wind speed, radius of maximum rotational speed, based on RG 1.76. Procedures based on American National
and the rate of pressuredrop. DBT for regionI, coveringall Standard ANSI A 58.1 [American National Standards Insti-
states east of the Rocky Mountains, has a maximum wind tute, 1982]are acceptedfor transformationof DBT to actual
speedof 360miles per hour (mph) (1 mi. = ---1.6km). Region loadings:the appropriateload combinationsare also speci-
II covers the coastal western United States, and its DBT has fied in the SRP. SRP section 3.3.1 gives the acceptance
a maximum wind speed of 300 mph. Region III covers the criteriafor straightwinds (i.e., hurricanesand extratropical
remaining western states. Ramsdell and Andrews [1986] storms)as speedswith 100-yearmeanreturn period,but this
have collected tornado strike data since the publication of value is much less than the DBT wind speed. Therefore the
WASH-1300 and updated the tornado regionalizationmap designbasisfor the plantsis invariablycontrolledby torna-
for the United States. doesfor all regionsin the United States.The frequencyof
RG 1.117 lists the plant systems, structures, components, hurricanesand extratropicalstormsexceedingDBT
 RAVINDRA 391

TABLE 1. NRC SpectrumII Missiles jectiles. The beam controls the designof wall thicknessto
HorizontalStrikeSpeed,*mph preventperforationand scabbing{i.e., back face damageand
disintegration of concrete), whereas the automobile deter-
RegionI RegionII Region mines the design for structural responseof wall panels,
Wood plank, 4 in. x 12 in. 185 157 130 columns, or frames.
x 12 ft, weight 115 lb
Steel pipe, 6-in. diameter, 117 94 23
TORNADO MISSILE RISK
schedule40, 15 ft long,
weight 285 lb Missiles generated by tornadoesmay causedamageif they
Steel rod, 1-in. diameter x 114 89 86
3 ft, weight 8 lb
impact safety-related structures or exposed equipment.
Steel pipe, 12-in. diameter, 105 63 16 Equipment located within the structures is vulnerable to
schedule40, 15 ft long, damage from missile penetration of protective structures,
weight 750 lb secondaryimpact tffom spalied concrete, or missilesentering
Utility pole, 13.5-in. 123 107 58
through openings. Damage depends on the mass, terminal
diameter, 35 ft long,
weight 1500 lb velocity, deformation characteristics, and angle of impact of
Automobile, frontal area 132 116 91 the missiles. Historically, the licensing criteria specify suf-
28 ft 2 4000 lb ficient thicknessof concrete to prevent perforation, spalling
(i.e., front face damage and disintegration of concrete), or
Metric conversions:1 in. = ---2.5cm; 1 ft = --•0.3m; lib = ---0.45 scabbing of the barriers in the event of missile impact.
kg.
*Vertical speedsshouldbe takenas 70% of horizontalspeed. Several formulas have been suggestedfor determining wall
thicknessesrequired to prevent scabbing on the basis of
missile impact tests. From the tests conducted by Bechtel
speeds is usuallymuchlessthan10-7/yr. Section 3.5.1.4of [Vassallo, 1975], EPRI [Stephenson, 1977], and Stone and
SRP describesthe acceptancecriteriafor missilesgenerated Webster [Jankov et al., 1976] and using different types of
by tornadoes, their impact speeds, and transformationto missiles, it was concluded that the concrete thickness re-
static loads on the structures. In 1977 the NRC adopted quired to prevent damage is much less than 12 in. for all
alternative missile criteria consistingof two setsof missiles missilesexcept for the 12-in.-diameter steel pipe missile (a
denoted spectrum I and spectrum II. SpectrumI missiles barrier thickness of 18 in. is required). However, examples
consisted of a rigid slug and a 2 in. x 4 in. timber plank. of 12-in.-diameterpipe missileshave never been observed as
Spectrum II missilesare more conservativeand are listed in documented in several damage studies [Abbey and Fttjita,
Table 1. (Metric conversion rates for inches, feet, and 1975]. SRP section 3.5.3 gives the minimum acceptable wall
pounds are provided in a footnote to Table 1.) The total and roof thicknessesto protect against local damage from
tornado missiles (Table 2).
tornado load is calculated using a number of load combina-
tions which include tornado wind load, differential pressure Damage from tornado missilesresults only as the culmi-
load, and missile load. Load combinations (for wind and nation of a sequence of low-probability events. This se-
tornado loads with other loads) and structuraldesigncriteria quencestartswith the tornado strike in the plant vicinity and
includes missile injection (from available objects) and trans-
are specified in SRP section 3.8.
An industry standard,ANSI/ANS 2.3 [AmericanNational port, missileimpact, and barrier damage. Owing to analyti-
Standards Institute/American Nuclear Society, 1983], has cal complexity of the tornado missile injection, transport,
and impact analyses,the probability of strike is computedby
been developedwhich specifiesguidelinesto determinethe
simulation as described by T,'isdale tt•d Dunn [1981], John-
wind velocity, atmosphericpressurechanges,missiletype,
son et al. [1977], and Reinhold and Ellingwood [1982]. These
and impact speedthat resultfrom tornadoes,hurricanes,and
other extreme winds to be used in nuclear plant design.This
standardprovidesmapsof tornadowind speedscorrespond- TABLE 2. Minimum Acceptable Barrier Thickness for Local
ingto annualexceedance probabilities
of 10-7, 10-6, and Damage Protection Against Tornado-Generated Missiles
10-5 . Each map definesthreetornadointensityregions. Concrete Wall Roof
Records from 1916-1975 are used in the tornado hazard
Strength, Thickness, Thickness,
assessments.The methodologyby Fujita [1978]which ac- Region psi in. in.
countsfor gradationsof damageacrossand alongthe path
3000 23 18
features,andbiasesin reportingoccur- Region I
length,topographical
4000 20 16
rencesand intensityof tornadoeswas usedin developing 5000 18 14
these maps. A spectrumof missileswas selectedby the Region II 3000 16 13
AmericanNuclear SocietyWorkingGroupANS 2.3 to cover 4000 14 11
the range of characteristics
associatedwith all potential 5000 13 10
3000 <6 <6
objectsthatcanbe propelledby a tornado.Thetwo missiles, Region Ill 4000 <6 <6
a 750-1b-wideflangesteelbeamand a 4000-lbautomobile, 5000 <6
have been documentedto be actual tornado-generatedpro-
392 EXTREME WINDS AND NUCLEAR POWER PLANTS

TABLE 3. ExtremeWindCoreDamageFrequencies
FromPlant-Specific
PRA
Tornado Straight Wind Tornado Core Total Core
Frequency Core Damage Damage Damage
Plant Name (AnySize),
(mi2/yr)
-l Frequency,
yr-1 Frequency,
yr-1 Freque•tcy,
yr-
Indian Point 2 1.00E - 04* 3,60E - 05 <E - 07 3.60E - 05
Indian Point 3 1.00E - 04 1.30E - 06 <E - 07 1.30E - 06
Limerick 1 and 2 1.13E - 04 9.00E - 09 <E - 08 9.00E - 09
Millstone 3 1.87E - 04 low <E - 07 <E - 07
Oconee 3 2.50E - 04 low <E - 09 <E - 09
Seabrook 1 and 2 1.26E - 03 <3.89E - 08 2.06E - 09 2.06E - 09
Zoin I and 2 1.00E - 03 NA <E - 08 NA

NA, not available.

*Read,for example,1.00E- 04 as 1.00 x 10-4.

studieswere basedon computersimulationsinvolvingmath- EXTREME WIND RISK

ematicalmodelingof the geometricaldistributionof potential
Winds due to tornadoes,hurricanes,and straightwinds
missiles, tornado movement, tornado wind field, missile
cancausefailureof structuresdueto directpressureloading.
injectioninto the wind field, aerodynamicflightsof missiles,
Failure modesmay be local (e.g., wall panel or roof) or
and the missileinteraction with the plant. The resultsfrom
global (e.g., failure of building shear walls or buckling
these studiesshow that the three parametersimportantfor
the analysis of tornado missile risk are (I) tornado charac- collapse of tanks and stacks). The probability of failure
induced by wind loading for any structure is evaluated by
teristicssuchas wind speed,radius, and pressuredrop; (2)
integrating (the procedure is called "convolution") the ex-
number and locations of potential tornado missiles;and (3)
resistanceof plant barriers to missile strikes. treme wind hazard probabilitydistributionwith the fragility
The study by Twisdale and Dunn [1981] simulatedtornado of the structurefor the specifiedfailure mode. Wind hazard
missile historiesfor two hypothetical plants utilizing the denotes the annual probability of exceedance of a hazard
standardspectrumof missiles.Their results indicate that the parameter,suchas wind velocity. Wind fragility is definedas
annualprobabilityof missilestrikeis lessthan10-7 in NRC the conditionalprobabilityof failure of a structurefor a given
RG region I. The probability of back face scabbingwas wind velocity. In casethe wind hazard and the wind fragility
found to be several orders of magnitude lower. Conserva- are fully known, there will be a single hazard curve and a
tively assumingthat scabbingdamageresultsin core dam- singlefragility curve; the result of the convolutionis a single
age, the thicknessof reinforced concrete walls required is value of the annual frequency of failure. If there is uncer-
obtainedas lessthan 8 in. for region A, the regionof highest tainty in the hazard, that is representedby a discrete set of
intensity accordingto the tornado region classificationby hazard curves with subjective weights attached to them.
Twisdale. This region is entirely contained in region I of Similarly, a family of fragility curves may be specified. The
NRC RG 1.76. Barrier thicknesses(Table 2) for all nuclear convolution is done by taking one hazard curve and one
plantsin the United Statesdesignedto SRP criteria are much fragility curve at a time, and the resulting frequency of
larger than 8 in. Therefore it is concludedthat the plants failure carriesa subjectiveweight equal to the product of the
designedto SRP criteria for missile strikes have a mean core subjective weights assigned to these specific hazard and
damagefrequencyof substantially lessthan 10-7/yr. In a fragility curves. The result of convolving the family of
recent paper, Twisdale [ !988] has provided further examples hazard curves with the family of fragility curves is a subjec-
on tornado missile risk analysis. tive probability distribution of the annual frequency of
Componentsinside the reinforced concrete buildings still failure. This convolution could be accomplishedby numer-
may be vulnerablefrom missilesenteringthroughopenings ical integration or by simulation techniques.
such as vents, ducts, doorways, and roll-up doors. The Extreme wind PRAs have been conducted for seven
probability of damage to these componentsdependson the nuclear power plants. Table 3, extracted from a report by
number and size of these openingsand the location of critical Kimura and Budnitz [1987], summarizes the tornado fre-
componentsinside the building. In some nuclear power quencies, the tornado strike frequencies, the straight wind
plants constructedin the 1960sand early 1970s,it is possible core damage frequencies, the tornado wind/missile core
to find safety-related components outdoors or in non- damagefrequencies, and the total straightwind/tornado core
category I structures. Potential for missile generationfrom damagefrequencies for these sites. A study of these previ-
failure of any nonsafety structures such as metal sided ous PRAs shows that core damage frequency from straight
structures, masonry walls, and steel and concrete stacksalso winds/tornadoes
is muchlower than 10-6/yr exceptfor
needs examination. Indian Point 2. The frequency at this plant was estimated
 RAVINDRA 393

be higher because of a unique plant feature; that is, the Development of the fragilityis donein termsof thefactor
postulated collapse of the unit 1 superheaterstack was of s•ety, definedas the capacitydivided by the response
judged to affect the unit 2 dieselgeneratorbuildingand/or associatedwith designbasisloadsfrom extremewinds. The
control building. Abbreviated PRAs have been conducted structure's capacityto withstandwindloadeffectsis formu-
for five nuclear power plants under the NRC's programfor lated in terms of the variablesinfluencingthe wind velocity
the resolution of generic safety issue A-45 (decay heat necessary to causefailure.For example,if thedesignwind
removal) [Reed and Ferrell, 1987]. The core damage fre- speed is Va, thestructure capacity C canbe expressed as
quencies due to high winds alone, tornado winds, and/or
missiles and their combination were estimated. The core C = VaFwFs (3)
damagefrequency computationswere basedon the assump-
whereF w is the safetyfactorrelatingthe designwind
tion of loss of off-sitepower. All these core damagefrequen-
pressuretotheactualwindpressure onthestructure
andFs
cies are generally higher than those reported in Table 1
is the safetyfactor relatingthe actual capacityof the
because in almost every plant, the risk was dominatedby structure to the calculated capacity.
non-category I structures interacting with safety-related Eachof thesesafetyfactors,sayF, can be modeledas
systems.It is to be noted that in the abovestudies,resis-
tance to wind loadingfor reinforcedconcretestructureswas F = FroeRe
w (4)
consideredto be adequate,andwind-loading-induced failure
of these structureswas eliminatedon the basisof its negli- whereF m is the mediansafetyfactor,eR is a random
gible contributionto the frequencyof core damage. variablereflectingthe inherentrandomness in the safety
It is shownin the following that the structuresdesignedto factor,andet: isa random variable
reflecting
theuncertainty
SRPrequirements
havenegligible
contribution
(<10-6/yr)to in thecalculationof Fro-BotheRande c7areassumed to be
the core damagefrequency.Coredamagefrequencycompu- lognormally distributedwithlogarithmicstandarddeviations
tationswere performedfor two sitesin tornadoregionI /3R and/3t;.Usingthemedian safetyfactors
FwmandFsm,
(regionof highesttornadohazard),representing extremely logarithmic standarddeviations due to randomness
high tornadohazard(plant A, Cooper,Nebraska)and an- and/3R, s, andlogarithmicstandarddeviationsdueto uncer-
contributionto the tainty/3
other site in the regionwith substantial t•,•' and/3
t•,s,themedian
structuralcapacity Cm,its
hazard curves from the hurricanes(plant B, St. Lucie, variability dueto randomness /•R,canduncertainty in the
Florida). The plant A site is exposedto potentialwinds median/3•,•,c canbe computed.
generated by tornadoes andextratropicalstorms,whereas The variablesinfluencingthe factor of safetyfor external
the plant B site experiences hurricanes, tornadoes, and windpressure are structuralstrength, structuralmodeling,
extratropicalstorms.The hazarddata specified for these and determinationof the wind pressure distributioninclud-
sitesfor both tornadoand straightwindsare availablein the ingeffects of adjacentstructures andterrain.Thefactorof
TAP A-45 reports[ReedandFerrell, 1987]. safetydueto structuralstrength refersto theactualstrength
For the structures designed to SRPrequirements, fragility compared with the stresslimitationsimposed bytheoriginal
curvescanbe developed on the basisof the plantdesign designacceptance criteria.The factor of safetydue to
criteria.Thepurpose of windpressure fragilityanalysisisto modeling refersto the accuracy of analyticalprediction of
develop a relationshipbetween theconditional probability of the loadsand stresses.The factor of safetydue to determi-
failureof a structureandthe windvelocity.For example,if nationof windpressure distributionincludesthe variables
thewindvelocityis V, andtheresponse (e.g.,force)dueto associated with wind pressurecoefficients, gustfactors,etc.
thiswindvelocityat a specified locationisR, thestructure's This factor also includesthe considerationof shieldingand
capacity to withstand windforceis a random variable,C. channeling effectsof otherstructures at the sitetogether
The fragilityof the structureis thendefinedas withlocalsitegradechanges in determining median-centered
wind loads on the structure.
f = Probability
{C -<R} (1) Lognormal
distribution
is usuallyadopted
for modeling
structural
fragilities.From the knowledge
of medianwind
If thestructure
capacityismodeledasa lognormally distrib- capacityCmandvariability dueto randomness /3•,c and
uted random variable with medianCm and logarithmic variability dueto uncertainty/3v,c in termsoftheirlogarith-
standard deviation/3,
thenthefragilityis givenby mic standarddeviations,fragilityof the structure,j', at a
windspeed17foranynonexceedance probabilitylevelQ can
(2) be derivedby using the formulationgiven in the PRA
Procedures Guide [1983]:

wherecI>(
ß) isthestandard
function;
fordifferent
calculated,
Gaussian
values
anda curvecalled
cumulative
ofwindspeed
"fragility
distribution
thefragility
curve"
f is
isobtained
In(V/Cm)
+13v,cCg-•tQ
(5)

byplottingf versusv (ora factorof safety

withrespect to whereQ = P[f -< f'tVl is the subjective
probability
DBT) (Figure 2). ("degree
ofbelief")thattheconditional
probability.{'
is
394 EXTREME WINDS AND NUCLEAR POWER PLANTS

lad

95% Confidence Median 5% Confidence

.0 I

FACTOR OF SAFETY WITH RESPECTTO DESIGN BASIS TORNADO

Fig. 2. Fragility curves for structuresdesignedto SRP criteria.

thanf' givena wind speedV and wherecI)-•( - ) is the strength,pressurecoefficient,and modelingof structuresfor
inverse of the standard Gaussian cumulative distribution wind load analysis. The coefficient of variation in material
function. Note that Q is the sumof the probabilitiesassigned yieldstrength,/3my,
wasestimated
to be0.15[Galambos
and
to all the fragilitieslessthanf'. In this formulation, both the Rayindra, 1978; Mirza and MacGregor, 1979; Mirza et al.,
inherent randomnessand the uncertainty are explicitly rep- 1979].
resented, and this leads to a family of fragility curves for the The pressure coefficient relates the induced pressure on
specifiedfailure mode. wall panelsto maximum wind velocity. Induced pressureon
A structurewhich has been designedfor a region I design a wall panel is a function of structural shape as well as the
basis tornado is expectedto have a wind-resistingcapacity locationon a wall panel. Therefore there is someuncertainty
larger than 360 mph becauseof conservatismsinherent in associatedwith the pressure coefficient. There are no data
designcodes,material strengthspecifications,assumedfail- on tornado wind pressuresto derive the uncertainties in
ure modes, etc. The median capacity and the variability pressurecoefficientand modeling. Since the physical phe-
associatedwith the calculatedmedian capacity are estimated nomenonof induced pressuredue to tornadoesand straight
by accountingfor these conservatismsas explained below. winds can realistically be consideredto be similar, the values
The main sourceof conservatismin the capacity is due to estimatedby Ellingwood [1978]for straightwinds are used,
the fact that the nominal steel yield strengths and the i.e., uncertainty
in pressurecoefficient
(/3pc)of 0.15 and
nominal concrete strengths specified by the designer are uncertaintyin wind modeling(/3,,,m)of 0.05.
much lower than the median values. This conservatism Next, it is assumed that the variability in wind pressure
factor in nominal yield strengthto median strengthhas been (ep) canbe modeledas the productof randomvariables
estimatedto be 1.2 [Ravindra and Banon, 1988]. Variability representingvariabilityin pressurecoefficient(epc.),wind
in wind pressureis a functionof variabilitiesin materialyield modeling(e,,,,), andmaterialyieldstrength
 RAVINDRA 395

TABLE 4. AnnualExtremeWind Core DamageFrequencies

for
Two NuclearPowerPlantsin RG 1.76RegionI (1) REVIEW PLANT-SPECIFICHAZARD
DATA AND LICENSINGBASES(FSAR)
Combined

Tornado
Straight
Wind
Tornado and
StraightWind HI
II11
I I I IIII I 11
Plant A 3.20E - 07* 8.88E - 14 3.20E - 07 (2) IDENTIFY SIGNIFICANT CHANGES,
Plant B 1.39E - 08 2.65E - 09 1.65E - 08 IF ANY, SINCE OL ISSUANCE

*Read, for example,3.20E - 07 as 3.20 x 10-7.

NO -- YES
ep = epce.,me.,y (6)
Since the uncertainties associated with these variables are • •1 • i • m Hill

essentiallyindependentvariables,they may be combinedto

findtheoverallor combined
uncertainty/3p
by

......
•-YES
(4) !S THE HAZARD FREQUENCY
'• '• 2 2
/3; = /3jc +/3,,,m+/3,,,v (7) ACCEPTABLY LOW?

where /3pc is the lognormalstandarddeviationfor the

pressurecoefficient,/3,,,m
is the lognormalstandarddeviation
iiiiiii
iiiiiiiiiiii
iiiiiiiiii
iiI i i i
•YES
(5) BOUNDING, ANALYSIS
for wind modeling,and 13myis the lognormalstandard OR • (RESPONSE/CONSEQUENCE)
deviation associatedwith material yield strength.
The logarithmicstandarddeviationfor the wind pressure
(/3p)is calculated
as0.21.Sincethecalculated
windpressure
is proportional to the square of wind velocity, logarithmic
OR •
standarddeviation (/3R) of wind velocity is 0.1. The uncer-
tainty in structural modeling (/3u) is estimated as 0.15
[Rayindra and Banon, 1988]. Thus the wind capacity of
reinforced concrete structures defined in terms of design
basistornadohas a mediancapacity(Vm) of 1.2,/3R of 0.1,
and/3t• of 0.15. Using these computedvalues, a family of
fragility curves can be obtained. Figure 2 shows fragility
curves for reinforced concrete structures designedto SRP Fig. 3. NRC-recommended IPEEE approach for extreme winds.
requirements.
It is conservatively assumedthat the failure of a structure
results in failure of all safety-related equipment inside the
structure and leads to subsequent core damage. The core
damage frequency is evaluated by the convolution of the vere accidentsand cost-effective safety improvementsthat
family of fragility curves for the structureand hazard curves reduce or eliminate the important vulnerabilities. Extreme
for tornado and straight winds for the two sites, A and B. winds (called high winds) are identified as an external event
The mean frequencies obtained are shown in Table 4. The that needs to be included specifically in the IPEEE. A
mean frequencies for combined tornado and straight winds progressivescreeningapproach has been developed [Nucle-
are approximated by adding their individual frequencies. ar Regulator3'Commission, 1991; Prassinos eta!., 1989] to
The values of core damagefrequenciesfrom extreme winds identity potential vulnerabilities from extreme winds at
for boththe sitesare substantially
lowerthan10-6/yr. nuclear power plants. Figure 3 shows a flow chart of this
To summarize, the analysis describedabove has indicated approach.The steps shown in the figure representa seriesof
that the frequency of failure under straight wind and tornado analyses in increasing level of detail, efl•)rt, and resolution.
loading for the structures designed to SRP criteria for However, the licensee may chooseto bypassone or more of
structures housingthe safety-related equipment is less than the optional steps so long as the 1975 Standard Review Plan
10-6/yr. criteria are met and the potential vulnerabilities are either
identified or demonstrated to be insignificant. This proce-
dure is based on the study results discussedabove: current
INDIVIDUAL PLANT EXAMINATION OF EXTERNAL. EVENTS
SRP requirements assure that the frequency of tornado and
Based on the Policy Statement on Severe Accidents, the other extreme wind damage from pressure loading and
NRC has requestedthat the licensee of each nuclear power missile
impacts
isacceptably
low(i.e., <10-6/yr).Themajor
plant perfbrm an individual plant examination.This plant steps are as follows:
examination systematically looks for vulnerabilitiesto se- 1. Review the site-specificwind hazard data and
396 EXTREME WINDS AND NUCLEAR POWER PLANTS

ing bases, including how the specific hazard issues were tion are describedalong with a summary of the results from
resolved at the licensing stage. different PRAs.
2. Identify significantchanges since the issuanceof the The review showedthat the risk of tornadomissiledamage
operating license; in the case of an extreme wind event, for nuclearpower plantsdesignedaccordingto the Standard
these changescould be new sources of tornado missiles or Review Plan requirementsis acceptably low (i.e., less than
construction of structures vulnerable to extreme winds. 10-6/yr).In addition,
casestudies
described
in thepaper
3. Determine if the plant meets the Standard Review haveindicatedthat the frequencyof wind-induceddamageof
Plan criteria. As part of this evaluation, a confirmatory structures designed to the Standard Review Plan criteria is
walkdown of the plant should be performed. This walkdown alsolessthan10-6/yr.Results
frompastPRAshaveindi-
would concentrate on outdoor facilities (steel structures, cated that the major contributionto core damagefrequency
unreinforced masonry walls, steel tanks, etc.) that could be comesfrom failures of non-safety-relateditems which may
affected by extreme winds and missiles. If this evaluation endangerthe safety-relatedstructuresby collapsingon them
indicatesthat the plant conformsto the SRP criteria and the or by missileimpact. Therefore the proceduresdeveloped
walkdown reveals no potential vulnerabilities, it is judged for the wind portion of IPEEE require a confirmatory
that the contribution from the extreme wind hazard to core walkdown of the plant to search for vulnerabilities to ex-
damage
frequency
is lessthan10-6/yrandthattheIPEEE treme wind even when the plant designis consideredto have
screeningcriterion is met. met the SRP requirements.
If the SRP criteria are not met, the licensee should take
one or more of the optional steps given below to further REFERENCES
evaluate the extreme wind event.
4. Determine if the extreme wind hazard frequency is Abbey, R. F., and T. T. Fujita, Use of tornado path lengths and
acceptablylow. If the original design basisdoes not meet the gradationsof damageto assesstornadointensity probabilities,in
Preprints of the Ninth Conference on Severe Local Storms,
current regulatory requirements, the licensee may chooseto
American Meteorological Society, Boston, Oct. 1975.
demonstrate that the frequency of exceeding the original American National StandardsInstitute, Minimum designloadsfor
design
basisis sufficiently
low(thatis, lessthan10-5/yr)and building and other structures, ANSI A58.1-1982, New York,
that the conditional core damage frequency from winds March 1982.

exceedingthe original designbasis is judged to be less than American National StandardsInstitute/American Nuclear Society,
Standardfor estimating tornado and extreme wind characteristics
10-•/yr.If theoriginaldesign
basishazardmultiplied
bythe at nuclear power sites, ANSI/ANS 2.3, Am. Nucl. Soc., La
conditional core damage frequency is not sufficiently low Grange Park, II1., 1983.
(thatis, lessthanthe screening
criterionof 10-6/yr), addi- Ellingwood, B., Reliability basis of load and resistance factors for
tional analysis may be needed. reinforcedconcrete design,Natl. Bur. of Stand., Washington,D.
C., Feb. 1978.
5. Performa boundinganalysis:this analysisis intended
Fujita, T. T., Workbook of tornadoesand high winds for engineering
to providea conservativecalculationshowingthat eitherthe applications, SMRP Pap. 165, Univ. of Chicago, Chicago, I11.,
hazardwould not resultin core damageor the core damage 1978.
frequency
is belowthe reportingcriterionof 10-6/yr.The Galambos,T. V., and M. K. Ravindra, Propertiesof steel for use in
level of detail is that level neededto demonstratethis point. LRFD, J. Struct. Div. Am. Soc. Civ. Eng., 104, 1459-1468, 1978.
Jankov, Z. D., J. A. Shanahan, and M.P. White, Missile tests of
6. Performa probabilisticrisk assessment: if options4 quarter-scalereinforced concrete barriers, paper presentedat the
and 5 are not able to screenout the extreme wind event, a Symposium on Tornadoes, Assessment of Knowledge and Impli-
PRA consisting of hazardanalysis,fragilityevaluation,plant cations for Man, Tex. Tech Univ., Lubbock, June 1976.
systemsand accident sequenceanalysis, and core damage Johnson, B., et al., Simulation of tornado missile hazards to the
frequencyestimationshouldbe performed.If the core dam- Pilgrim 2 nuclear thermal generatingstation, draft report, Science
Applications, Inc., Sunnyvale, Calif., Nov. 1977.
agefrequency
is lessthan10-6/yr,the eventneednotbe Kimura, C. Y., and R. J. Budnitz, Evaluation of external hazards to
considered further. nuclear power plants in the United States, NUREG/CR-5042,
Lawrence Livermore Nat!. Lab., Livermore, Calif., Dec. 1987.
Markee, E. H., J. G. Beckerley, and K. E. Sanders,Technical basis
SUMMARY AND CONCLUSIONS for interim regional tornado criteria, WASH-1300, U.S. Gov.
Print. Off., Washington, D.C., May 1974.
This paper has reviewed the results from recent and Mirza, S. A., and J. G. MacGregor, Variability of mechanical
ongoingresearchprojectson extremewindswith the objec- propertiesof reinforcingbars, J. Struct. Div. Am. Soc. Civ. Eng.,
tive of summarizingthe current state of the art as a back- 105,921-937, 1979.
ground for the development of criteria for extreme wind Mirza, S. A., M. Hatzinikolas, and J. G. MacGregor, Statistical
descriptionsof strengthof concrete, J. Struct. Div. Am. Soc. Civ.
analysisin IPEEE. The evolutionof tornadodesigncriteria Eng., 105, 1021-1037, 1979.
in the nuclear industry is traced startingwith RG 1.76 and Nuclear Regulatory Commission, Procedural and submittal guid-
the StandardReview Plan and concludingwith ANS 2.3. The ance for the individual plant examination of external events
reviewhascoveredthe hazardanalysisfor tornadoincluding (IPEEE) for severe accident vulnerabilities, Final Rep. NUREG-
1407, June 1991.
tornado missiles and straight winds. Fragility analysisof PRA ProceduresGuide, Analysis of external events, chap. 10,
structures subjected to extreme wind effects is discussed. NUREG/CR-2300, vol. 2, AmericanNuclear Societyand Nuclear
The techniquesused in the systemsanalysisand quantifica- RegulatoryCommission,Washington,D.C., Jan.
 1LhVINDRA 397

Prassinos,P. G., et al., Individual plant examinationfor external 4767, 4458, 4448 and 4762, Sandia National Laboratories, Albu-
events: Guidance and procedures, draft, NUREG/CR-5259, querque, N.M., 1987.
Lawrence Livermore Natl. Lab., Livermore, California, March Reinhold, T. A., and B. Ellingwood, Tornado damage risk assess-
1989. ment, NUREG/CR-2944, Nucl. Reg. Comm., Washington,D.C..
1982.
Ramsdell, J. V., and G. L. Andrews, Tornado climatologyof the
contiguous United States, NUREG/CR-4461, PNL-5697, Pac. Simiu, E., M. J. Changery, and J. E. Filliben, Extreme wind speeds
Northwest Lab., Richland, Wash., March 1986. at 129 stations in the contiguousU.S., NBS Build. Sci. Set. I [8.
Natl. Bur. of Stand., Washington, D.C., 1979.
Ravinalta, M. K., and H. Barton, External event scopingquantifi-
cation of LaSalle unit 2 nuclear power plant: Risk methods
Stephenson, A. E., Full scaletornadomissileimpacttests,Fina!
Rep. NP-440, Electric Power Res. Inst., Palo Alto, Calif., July
integration and evaluation program (RMIEP), NUREG/CR-4832, 1977.
Sandia Natl. Lab., Albuquerque, N. M., Jan. 1988. Twisdale, L. A., Probability of facility damagefrom extreme wind
Ravindra, M. K., and A.M. Nafday, State-of-the-art and cun'ent effects,J. Struct. Div. Am. Soc. Civ. Eng., 114, 2190-2209, 1988.
research activities in extreme winds relating to designand evalu- Twisdale, L. A., and W. L. Dunn, Tornado missilesimulationand
ation of nuclear power plants, NUREG/CR-5497, UCID-21933, designmethodology,NP-2005, vol. 1, ResearchTriangle Inst.,
Lawrence Livermore Natl. Lab., Livermore, Calif., May 1990. ResearchTriangle Park, N. C., Aug. 1981.
Reed, J. W., and W. L. Ferrell, Extreme wind analyses for five Vassallo, F. A., Missile impact testing of reinforced concrete
nuclear power plants, Appendix G, in Shutdown Decay Heat panels,HC-5609-D-1,reportpreparedfor BechtelCorp.,Calspan
Retnoval Analysis of Nuclear Power Plants, NUREG/CR-4713, Corp., Buffalo, N.Y., Jan.
 Wind/TornadoDesignCriteriaDevelopmentto Achieve Required
Probabilistic Performance Goals

DOROTHY S. N G

Lawrence LivermoreNational Laborator),, Livertnore, California 94550

1. INTRODUCTION The pertinent wind/tornado design requirements are sum-

marized below:
The U.S. Department of Energy (DOE) has established 1. The designof facilities shall encompassannual safety
specificrequirementsfor the designof structures,systems, performance
goalsof (1) 5 x 10-7/yrprobability
foranearly
and components(SSCs) of critical facilities for resistingall fatality to an average individual assumedto be located within
types of internal and external natural hazard events. The 1 mile (---1.6 km) of the reactor facility control perimeter; (2)
Lawrence Livermore National Laboratory (LLNL) was 2 x 10-6/yrprobabilityfor a long-termfatalityto plant
chartered to develop design guidelinesfor natural hazards workers or a member of the general public located within 10
utilizing the best technical knowledge and engineeringjudg- miles (---16 km) of the reactor facility, but outside of the
ment to assure that the design of the SSCs conform to controlperimeter;
(3) 1 x 10-6/yrprobability
for a large
performance goals set forth by DOE. release of radioactive materials from a reactor accident; and
This report documents the strategy employedto develop (4) 1 x 10-5/yrprobability
for coredamage.
wind/tornado hazard design criteria for a critical facility to 2. DOE encourages the contractors to incorporate the
withstand loading induced by the wind/tornado hazard. knowledge and experience over many years in nuclear
These guidelineswere developedby a Wind/TornadoWork- reactor design accumulated.
ing Group (WTWG) at LLNL composedof six expertswho 3. DOE requires the contractors to make use of past
are knowledgeable in the fields of structural engineering, regulatorycompliancein commercial reactor designor dem-
wind/tornadoengineering,and meteorology.Utilizing their onstrateimprovementover regulatoryprovisions.
best technical knowledge and judgment in the wind/tornado The design requirements document stated that external
field, they met and discussed the methodologiesand re- events shall be addressedproactively in the design process
viewed available data. A review of the available wind/ in the context of the large-release goal. To investigate the
tornado hazard model for the site, structuralresponseeval- achievement of the large-release goal requires additional
uation methods, and conservative acceptancecriteria led to considerationof containment failure modes, namely, direct,
a proposeddesigncriteria that has a high probabilityof bypass,and penetrationfailures. It is more realisticto show
achieving the required performancegoals. the achievement of the core damage goal as an interim step,
by wind/tornadorisk assessment.The WTWG expectsthat
2. DESIGN REQUIREMENTS the mean large-release risk frequency will be at least an
order of magnitudelower than the mean core damagerisk
The primary considerations of the designof the critical frequencybecauseit is unlikely that wind and missileeffects
facility are publicsafety,worker safety,andenvironmental will significantly contribute to breaching containment.
protection. The DOE designrequirementsdocumentserves Therefore achieving the core damage goal is believed to
as the basisfor designof the facilities. The contractorsmay ensure achievement of the large-release goal.
proposemethodswhich offerimprovements overthe speci-
fiedrequirements.However, if suchmethodsare proposed, 3. DESIGN GUIDEI.INES DEVELOPMENT STRATEGY
the contractoris required to demonstratethe validity of
resulting improvements. To ensure the safety of the plants, DOE requires that
facility designmeet stringentsafety goalsand shouldincor-
porate reactor experienceand compliancewith regulatory
The Tornado:Its Structure,Dynamics,Prediction,andHazards.
GeophysicalMonograph 79 requirementsin facility design. An effective approachto
Copyright1993by the AmericanGeophysical
Union. satisfythese requirementsis to employ a panel of experts
400 WIND/TORNADO DESIGN CRITERIA

who havetechnicalknowledge,experience,andjudgmentas on DOE guidelines and the Electric Power Research Insti-
a result of years of research and service in this area. Thus the tute (EPRI) NP-2005 report; (4) to define conservative struc-
WTWG was establishedfor the developmentof wind/ tural responseevaluation and acceptance criteria to substan-
tornadodesignguidelines.The WTWG consistsof a groupof tiate the design guidelines; (5) to assure the conformance of
six nationallyrecognizedtechnicalexperts;they are briefly designguidelineswith safety goals by performingprelimi-
describedbelow in alphabeticalorder: nary probabilistic risk assessments;and (6) to make use of
Robert F. Abbey, Jr., is a well-known meteorologistwho nuclear industry standards, NRC provisions, and national
is currently serving as the director of Marine Meteorology codes and standards as benchmarks and references.
Research,Office of Naval Research.He specializesin prob-
abilistic and statisticalanalysesof the occurrenceof natural
3.1. Developing Wind Load Specifications Based
phenomena.
on Site-SpecificHazard Frequency Curves
W. Lynn Beason is associateprofessorin the Civil Engi-
neering Department of Texas A&M University. His exper- In the 1970sthe NRC divided the U.S. into three geo-
tise is in wind-borne missile research, tornado risk analysis, graphical regions and recommended design basis tornado
and structural design. (DBT) wind speeds for each region in NRC Regulatory
T. Theodore Fujita is a world-known tornado assessment Guide 1.76 [U.S. Nuclear Regulatory Commission, 1974].
expert. Fujita developed the Fujita scale (F scale) for esti- These DBT wind speedsare believed to be quite conserva-
mating the relative intensity of tornadoes and the DAPPLE tive, resultingfrom coarseassumptionsand problemswith
methodfor estimatingthe probability of tornado occurrence. the regional data sets employed.
Dale C. Perry, chairman of the WTWG, is the head of the More recently, the Office of Nuclear Reactor Regulation
Department of Construction Science of Texas A&M Univer- [1988] positionpaper divided the U.S. into four regionsand
sity. His expertise is in structural designand wind engineer- recommendedDBT wind speedsbasedon regionalanalyses.
ing. He has over 20 years of experience in postdisaster Recommendationsbased on regional analyses cannot pro-
investigations. vide a realisticportrayal of the risk associatedwith specific
John W. Reed is an associate at Jack R. Benjamin & sites within the region.
Associates,Inc., California. He has 25 years of experience In 1979,two leadingwind engineeringexperts, Fujita and
in nuclear reactor design and probabilistic risk assessment McDonald, performed independent site-specific wind/
(PRA) for seismic and wind hazards. Reed performed the tornado hazard assessments for 26 DOE facilities. The site of
preliminary wind speed PRA for this project. interest was among them. These analysesutilized advanced
Lawrence A. Twisdale, Jr., is the senior vice presidentof methodologies.Two of the most important improvements
Applied Research Associates, Inc., Southeast Division in involved the use of wind speed gradation, both across the
North Carolina. He specializesin tornado missile simulation width and along the length of tornado damage paths. In
and design methodology for tornado hazard assessment. addition, techniqueswere introduced to allow adjustments
Twisdaleperformedthe preliminary missilecriteria PRA for to be made to tornado data sets to account for unreported
this project. tornado events. Based on these hazard assessment results,
Additionally, James McDonald, a professor and re- Kennedy et al. [1989], in their DOE-UCRL-15910 report,
searcherat Texas Tech University, is a member of the senior provided design and evaluation guidelines for hazard fre-
review groupfor this project; he providesguidanceto the quencies
upto 2 x 10-5/yr.
WTWG andreviewsreportsproducedby thegroup.McDon- In 1985, a third site-specifichazard assessmentwas per-
ald specializesin wind hazardassessment andmissileimpact formed by Twisdale for the site of interest. This analysis
research. used computerized simulationtechniquesto estimate hazard
Shortly after formation of the WTWG and prior to their probabilities and accounted for building size effects.
first meeting, the existing wind/tornado hazard assessment In summary, wind assessment technology has evolved
reports,Nuclear RegulatoryCommission(NRC) provisions, from overly conservative regional analyses to realistic site-
design standards, and other references available for the site specific assessments.In this process, coarse assumptions
of interest were distributed to the WTWG. The methodolo- have been refined by application of accumulated knowledge
gies and data utilized in existingregionaland site-specific appropriate to tornado hazard assessment.In addition, dur-
hazard assessmentswere comparedand discussedby the ing the ensuing years since the introduction of NRC Regu-
WTWG in a series of meetings.Then, usingtheir cumulative latory Guide 1.76, more complete tornado occurrence data
technicalknowledgeandjudgment in the wind/tornadofield, have been recorded, allowing more accurate assessments.
the WTWG plannedthe developmentof wind/tornadodesign The resulting site-specific analyses, which take these ad-
guidelines. vancements into account, are improvements over earlier
The six basic goalsof the WTWG strategyare as follows: regional analyses. Therefore the WTWG opted to incorpo-
(1) to build in conservatismby reducingwind/tornadogoals rate site-specific hazard curves in the design guideline de-
to 1/10of total project goals;(2) to specifydesignwind load velopment.
specificationsbased on site-specific hazard frequency The use of site-specifichazard frequency curves in design
curves;(3) to specifydesignmissileload specificationsbased guidelinesdevelopment reflects the accumulated
 NG 401

and experience in wind engineering.The recommended 4. DESIGN GUIDELINES TECHNICAL SUPPORTINGBASES

designguidance
presented in thispaperis a majorimprove-
The objectiveof the designguidelinesdevelopmentis to
mentto NRC provisions. The conformance of thesepro- recommendguidancefor the designof the facility to meet the
poseddesignguidelines
for designwasconfirmed by prelim- safety goals set forth by DOE. The basic philosophyfor
inary PRAs describedlater in this paper.
design guidelines development is to select low wind load
It must be noted in wind hazard assessmentthat limita-
specifications.The designgoal is met by applyingconserva-
tionsof tornadointensityandoccurrencedata,F scalewind
tive responseevaluation and acceptancecriteria.
speedassumptions, and analysisprocesses incorporated in Core damage guidelines were developed for the struc-
the hazard models result in assessment uncertainties.
tures, systems,and componentsrequired to bring a plant to
a safe shutdown and to maintain this conditionfollowing a
3.2. DevelopingMissile Load Specifications Based wind or tornado event. Production guidelines were devel-
on DOE Guidelines and EPRI Tornado Missile oped for SSCs which cannot be cost effectively repaired or
Simulation Report replaced within 90 days following a wind or tornado event.
Ordinary SSCs, which are classified as low-hazard SSCs,
The Offie of NuclearReactorRegulation[1981]Standard
must satisfy only UCRL-15910 wind/tornado design guide-
Review Plan (SRP), NUREG-0800, recommends a conser- lines.
vative spectrum of tornado-generatedmissilesand corre-
spondingvelocity fractions.There is no regionaldistinction
in the SRP missile types and sizes. However, the missile 4.1. Wind Load Specification
impact velocities vary on the basis of regionalDBT wind At the work group meetings the WTWG held long discus-
speeds.DOE-UCRL-15910 [Kennedy et al., 1989]also rec- sions to establish the proper direction of wind/tornado
ommendedmissilespectrafor variouscategoriesof facilities design guidelines. Based upon references provided to the
based in part on results of missile researchperformed at working group and the combined judgment of the hazard
Texas Tech University. The missilespectrawere developed assessmentand the risk assessmentanalysts, the following
for riskfrequencies up to 2 x 10-5/yr. major factors were taken into account: (1) conservatism
The Electric Power Research Institute sponsoreddevel- embeddedin methodologiesof assessment,(2) uncertainties
opment of a probabilistictornado missilerisk methodology identified in the assessment,(3) site-specificwind speedsin
to assesspotential missile hazard for nuclearpower plants. the geographical areas of interest, (4) site-specific wind
Results of this effort are presented in the EPRI NP-2005 speeds which achieve risk goals, and (5) acceptable wind
report [Twisdale and Dunn, 1981]. Missile impact velocities speedsby the consensusof WTWG members.
were estimated on the basis of computer simulations of After careful study of the hazard frequency curves and the
potential missile populations at nuclear power plants and examination of preliminary risk calculations, through an
computer-generated missile trajectories. The velocity frac- iterative process,the WTWG selected a tornado fastest-mile
tions for steel pipes generated by EPRI are slightly larger wind speedof !80 miles per hour (mph) (1 mi. = ---1.6 km),
than SRP velocity fractions. which corresponds to a probability of occurrence of 3 x
In summary, state-of-the-art missile simulationtechnology 106/yr,andit achieves thecoredamage goalof 1 x 10-6/yr.
incorporating relevant data bases and newly developed A preliminary risk analysis on the production event indi-
experimental results is an improvementover the technology cates that straight wind controls the design. Using the
used for NRC missile impact estimatesin the 1970s.In the straightwind model, a fastest-milestraightwind speedof 130
judgment of the WTWG, the recommendedmissilespectrum mph was selected. This wind speed corresponds to an
and the maximum impact velocities guidelines should be occurrence probabilityof 2 x 10-4/yr.
formulated by modificationof DOE UCRL-15910 guidelines.
The modification consists of incorporation of additional
4.2. Missile Load Specification
missilesand extrapolationfrom the EPRI tornado missile
simulation results. After evaluatingthe SRP provisions, EPRI missile simu-
lation results, and DOE-UCRL-15910 missile spectra, the
WTWG formulated missile guidelinesbasedon the following
3.3. AssuringConformanceof Design Guidelinesby
supporting bases: (1) selection of missile types that have
Performing Preliminary Probabilistic Risk Assessment
been observed in past tornado damage investigations, (2)
Probabilisticrisk assessmentswere performedto estimate modification of the DOE-UCRL-15910 to extend the risk
annualdamagerisk frequenciesfor the facilitiesbasedon the frequency from2 x 10-5/yrto 1 x 10-6/yr,and(3) extrap-
recommendeddesignguidelines.These wind damagerisk olation of the EPRI tornado missile simulation results.
frequenciesare measuresof achievementof theperformance As a result of this effort, the WTWG added an 8-in.-
goals. To ensurethe conformanceof the designwith the diameter steel pipe to the DOE UCRL-15910 missile spec-
requiredperformancegoals,independentpreliminaryPRAs trum to extendthe risk frequencyto 1 x 10-6/yr. The
were performedto evaluateboth wind speedand missile horizontal impact velocity for the 8-in. steel pipe was ob-
impact. tained by extrapolating EPRI results for a
402 WIND/TORNADO DESIGN CRITERIA

TABLE 1. Wind-Borne Missiles Criteria

Missile Missile Criteria Safety Production

horizontalimpactspeed,mph 150 115

2 in. x 4 in. timber plank
effectiveheight,ft 200 200
12 ft long, 15 lb
verticalimpactspeed,mph 100 75

horizontalimpactspeed,mph 120 90
3-in.-diameter standard steel pipe,
effectiveheight,ft 150 150
10 ft long, 75 lb
verticalimpactspeed,mph 80 60

8-in.-diameter standard steel pipe, horizontalimpact speed,mph 100 75

effective height, ft 75 30
15 ft long, 430 lb
verticalimpactspeed,mph 70 50

Automobile, 3000 lb rolls/tumblesalongground, mph 35 25

effectiveheight,ft 30 30
verticalimpactspeed,mph 20 15

Metric conversions:1 in. = ---2.5 cm; 1 ft = ---0.3 m; 1 lb = ---0.45 kg.

Horizontal
andverticalmissilespeedsareuncoupledandshouldnotbecombined.
Specific
missiles
represent
a classof debris.Effective
heights
areelevations
abovegroundlevel.

included rier design,the loadfactorof 1.3 doesnot applyto missile

steelpipeat the 90 percentile.This extrapolation
conversionfrom a maximumvelocity of 300mph to 200mph. loads.

The verticalimpactvelocitywas set at 2/3 of the horizontal The SSCsshallbe designedfor the mostseveretotal wind
impactvelocity by the WTWG. load(Wt) weight.This is a combinationof wind pressure
Table I presentsspectraof potentialmissileswith recom- (W•,,),atmospheric pressure change(APC) (Wp), andmis-
mended impact velocities and correspondingeffective sileimpactloads(Win). Theloadfactorappliesonlyto wind
heights(metricconversions are providedin a footnote).The pressure
WwandAPCpressure W•,,butnotto missileloads.
barrier thicknessshall be determinedby equationsgiven by In addition,the APC pressureload is half of its maximum
Twisdaleand Dunn [1981].The minimumthicknessof rein- value at the radius of maximum wind speed in a tornado;
forced concrete barriers is 12 in. for all elevations below 75 therefore,only 0.5W•,is combined
with wind pressure
feet, and8 in. for all elevationsabove75 feet. The minimum and/or missile impact, as given below:
thickness of steel barriers is 1/2 in. for all elevations.
Wt = 1.3 W,,,

4.3. Structural ResponseEvaluation Wt = 1.3Wp

and Acceptance Levels
Wt =Wm
The WTWG incorporatedconservativestructural re-
sponse
evaluation
methods
andacceptance
levelsasfollows: Wt = 1.3W,•, +Wm
1. Account for the fact that the critical facility is a
high-hazard
facility by increasingthe importancefactor to Wt = 1.3W,,,+ 0.5 x 1.3Wp
1.35.
2. Use exposureC for tornado-controlled designsby Wt = 1.3Ww+ 0.5 x 1.3W•,+Wm
assumingthat the facilitiesare locatedin openterrainwith In the caseof eventscontrolledby straightwind, the APC
obstruction
havingheightsgenerallylessthan 30 feet. This
assumptionaccountsfor uncertainties
pressure
in the characteristicsas follows:
loaddoesnotexist,andtheloadcombinationsare
of tornadoes.
3. Account for the variability in structural resistance Wt = 1.3 Ww
propertiesby usinga load factor of 1.3 on designwind
pressures. Wt -- Wm
4. Select the velocity pressureexposurecoefficientand
the gustfactorcorresponding
to the heightof structures
for Wt = 1.3 Ww + Wm
both tornado and straightwind controlleddesigns.
5. Account for the uncertainty in missile impacts by
4.4. Risk Assessmentfor Wind Speeds
followingthe American Concrete Institute (ACI) 349-85 and Missile Criteria
recommended 20% increase in calculated barrier thickness.
6. Since a conservative missile spectrum and maximum Thepreliminary PRAswereperformedduringtheconcep-
tornado-generated
impactvelocitieswere selectedfor bar- tual designstage;hence,component
andplantdetails
 NG 403

notavailable.
Thislimitstheapplicability
of thepreliminary lessthan the scabbingdamagerisk frequency. Therefore the
PRAs to the final design. In addition, the results must be WTWG believes that the recommended missile criteria con-
considered to bepreliminary
owingto thefollowing
simpli- form to the core damagegoal by a significantmargin.
fications
whichwerebuiltintothewindspeedriskanalyses: At an impact velocity of 35 mph, which is the core damage
1. Thepreliminary PRAanalysis
wassimplified
by using designguideline,the exceedanceprobability for automobile
theresultsfrompreviousPRAsto obtainanapproximate
but impactis 3.7 x 10-7/yr;thus,the automobile resultsmeet
usefulestimateof the expecteddamagefrequency. the core damage performance goal.
2. The preliminarymeandamagefrequency,whichwas
5. SUMMARY OF DESIGN GUIDELINES DEVELOPMENT
used, representsdamagewithin the 60-90% probability
range. LLNL evaluated NRC provisions and acknowledgedthe
3. Mean damagefragility curvesused in the analysis excessconservatismembedded in their supportinganalysis.
were inferredon the basisof knowledgeof the rules and Therefore designguidelines which deviate from NRC provi-
proceduresusedin typical plant designprocesses.A more sions are developed on the basis of site-specific hazard
rigorousanalysiswouldincludeperforming detailedfragility assessmentsand EPRI report results for the project. These
analysisfor each componentand combiningthe resulting credible design guidelines were developed on the basis of
meancomponentfragility curvesthroughBooleanlogic. technological advancements and past experience in wind
The tornadowindspeedsof 180mphanda straightwindof engineering. The load specifications reflect updates and
130mph,corresponding to 3 x 10-6 and2 x 10-4 annual enhancements to NRC regulations. The justification and
hazard probabilities of occurrence, produce damage fre- supporting bases for these updates and enhancements are
quencies
of 1 x 10-6 and5 x 10--sfor coredamage
and provided in this paper. The proposed structural response
production events, respectively. evaluation and acceptancecriteria follow industry codes and
For missilecriteria risk analysesthe simplifications
are as standards with minor modifications. The general design
follows: guidelinesrecommendedin this paper are believed to closely
1. Evaluation of missile damage is limited to exterior comply with DOE orders.
balTiersfor core damageand productionevents.
2. Estimation of impact probabilities for vulnerable ar-
Acknowledgment. This work was performed under the auspices
eas, such as doorways and vent openings,is not includedin of the U.S. Department of Energy by the Lawrence Livermore
the analysis. National Laboratory under contract W-7405-Eng-48.
3. Failures of plant stacks due to high winds are not
considered. REFERENCES
4. Missile generation contributions resulting from the
Kennedy, R. P., S. A. Short, J. R. McDonald, M. W. McCann, and
collapse and/or failure of tall structuresare not included. R. C. Murray, Design and evaluation guidelines for DOE facilities
5. Nonlinear relationshipsbetween the numbersof mis- subjectedto natural phenomena hazards, UCRL-159iO, Lawrence
siles, wind speed, facility layout, and damageprobabilities Livermore Natl. Lab., Livermore, Calif., Oct. 1989.
are not considered. Office of Nuclear Reactor Regulation, Standard Review Plan for the
review of safety analysis reports for nuclear power plants,
For core damage events, reinforced concrete barriers NUREG-0800 (formerly NUREG-75/087), rev. 2, U.S. Nucl. Reg.
were assumed to be damaged if the barriers suffered back- Comm., Washington, D.C., Jul,x 1981.
face scabbingor worse from any singlemissileimpact. As a Office of Nuclear Reactor Regulation, Safety evaluation by the
first approximation, it was assumedthat these minimum Office of Nuclear Reactor Regulation of recommended modifica-
tion to the R.G. 1.76 tornado design basis for the ALWR, NRR
barrier thicknessescontrolled plant design.
position paper, U.S. Nucl. Reg. Comm., Washington, D.C.,
The sum of scabbing damage risk frequenciesfor mini- March 1988.
mumdesignrequirements is 1.1 x 10-6/yr.Thisvalueis Twisdale, L. A., and W. L. Dunn, 7•rna&• Missile Simulation and
slightlyhigherthanthe coredamage goalof I x 10-6/yr. Design Methodology, vol. I, Simulati•n Methodology, Design
However, this damage criterion is very conservative,be- Applications, and TORMIS Computer Code, EPRI NP-2005,
Electric Power Research Institute, Palo Alto, Calif., Aug. 198I.
causeeven if backface scabbingoccurs, a core damageevent U.S. Nuclear Regulatory Commission, Design basis tornado
is not very likely to occur.The WTWG believesthat the core nuclearpower plants, NRC Reg. Guide i.76, Washington,D.C.,
damagerisk frequencyis at least 1 to 2 ordersof magnitude April
 Discussion

BOB MURRAY, SESSIONCHAIR

Lawrence Livermore National Laborato•'

PAPER G 1 (Golden) My second question concerns looking at the tor-

nado incidence in the west over the last two decades. Ostby
Presenter, Bob Murray, Lawrence Livermore National Lab-
has shown that there has been a significantincreasein western
oratory [Murray and McDonald, this volume, Design for
states'tornadoesduring that period, increasesamountingup
containment of hazardous materials]
to 300% in some cases. You seem to be ignoring these
(Unknown) I assumefor straight-linewinds you take the gust tornadoesand worrying only about the straight-linewinds.
response factor from the UBC or ASCE manual?
(Murray) You are correct. As a matter of fact, the next
(Murray) No, use the ASCE standard. The conceptis built speaker lives in a California city that just recently had a
around the ANCI 58.1, now ASCE 788. You use the rules in tornado.It's part of the elementof continuallyreviewingthis
there to go from the wind speedsto the pressurecoefficients. information and learning from new storm data that we are
We've imposed the importance factors to use, and we've doing now. One of our southern California sites, Santa
varied the importance factors if you're at a coastlinewhere Suzanna, has tornadoes as a designcriterion.
a hurricane could be the threat, and let that slide as you go
inland 100 miles. PAPER G2
Presenter,Ravi Ravindra, EQE Engineering[Rayindra, this
(Unknown) Do you take the samegust responsefactor for a
volume, State-of-the-art and current research activities in
tornado ?
extreme winds relating to design and evaluation of nuclear
(Murray) No gust responsefactors are usedfor tornadoes. power plants]

(Joe Golden, National Oceanic and AtmosphericAdminis- (Arnold Court, California) Any probabilistic risk assessment
tration, Office of the Chief Scientist) I have two questions. must involve first and foremost the expected or desired
First, is the infamous document, the so-called New-Reg lifetime of the structure. And becausethe longer you want it
1200, still enforced for nuclear power plants? to last, the greater the risk that it will be damaged, so you
start first with the lifetime. Do you assumethat these plants
(Murray) I thinkthe informationfromthat is incorporated
in are built to last 10 years, 500 years, or what?
Regulatory Guide 176, which providesthe three tornado
(Ravindra)The nuclear power plants are licensedto operate
zonesfor the country,and a standardreviewplan,whichis
for 40 years. There is currently a program to extend the
what the NRC uses to review a design.
license period by another 20 years, so you can assume a
(Golden)My urgentrecommendationis thatthe community lifetime of about 60 years. The probabilisticrisk assessment
represented take a carefullook at that that is done today is being done about 20 years into the
at this symposium
document to see if the criteria in the document are still valid. operationof the plant. So the risk numbersthat you see are
Based on what we've heard so far and will hear tomorrow, I typicallyover the life of the plant, or you cantalk aboutper
don't believetheyare. They mayhavebeenthebestresults year values.
of their time, but that was !5 or 20 years ago. (Court)Sowhatisthecalculated
risk'?
10-5 offailurewithin
the next 40 years?
(Murray) The wholepanelagreeswith that also.
iRavindra)Yes, a very low risk. Becausemeteorologists
have
The Tornado:Its Structure,Dynamics,
Prediction,
andHazards. donea goodjob in sayinghow muchwind to expect,to design
GeophysicalMonograph 79
Thispaperisnotsubject
to U.S. copyright. in 1993bythe for, the probabilistic
Published risk assessmentsshowthat the risk in
American GeophysicalUnion. somewhere
between10-4 and10"6, depending
uponthesite.
406 DISCUSSION

(Court) I have another questionabout missiles.Are they (Golden) You should know that Storm Data is in a very
assumed to fall end on, or broadside? fragile state. There are some important considerationsthat
everyone in this room needs to understandwhen they use
(Ravindra) In the design, it is assumedto fall end on. In the Storm Data. I think we need to get the community together
risk analysis,like Twisdaleis goingto talk about,the impact to supportStorm Data and to improve it.
is randomly oriented. Sometimes it could be end on; other
times it could be broadside. (Ravindra) I fully concur with you. Perhaps the engineers
and meteorologistshave not had enough interaction.
(Roger Tanner, National Climatic Data Center) Are there
studies that take into consideration, particularly in the PAPER G4
fragility area, the temperature in any of the three tornado
Presenter, Jim Hill, Department of Energy (J. Hill, not in
regions? It could affect tensile strength of the reinforced
this volume, Natural hazard losses: A Department of Energy
steel or concrete. Also, what about soil conditions.
Perspectivemlnjury and property damage experience from
(Ravindra) The temperature effects are considered mainly natural phenomenahazards)
for the internalsof the nuclear power plant. There is concern (Don Burgess, National Severe Storms Laboratory) Did I
about how the construction would perform over the years. understandyou correctly that DOE has a Doppler radar at
The embrittlement is considered in terms of the structure, Amarillo, Texas?
but the effect is minimal.
(Hill) No, not a Doppler, lightningdetectionequipment. Also
(Tanner) What about moisture? we have equipmentto detectpotentialelectricalgradient.We
(Ravindra) It is minimal in the sense of the aging we are
find high gradientsfrom wind and dust as well as thunder-
storms.They have sensitiveexplosivesat that plant they do not
involved in. We are looking at ways to inspect the structures
to see if there is some deterioration as a result of moisture or
like to operateat times of high potentialgradient.They have
had problemswith farmersplowingtheir fields duringstrong
temperature. But from what we've seen the effect is mini-
windsandproducingdangerous,highpotentialgradients.They
mal. This is because the structures are pretty beefy and the
shut the plant down when the potential gradient exceedsa
power plants are well maintained by law. certain amount, no matter what the cause. I could add a
(Joe Golden, National Oceanic and Atmospheric Adminis- commentaboutweatherpredictionat our sites.At leastfive of
tration, Office of the Chief Scientist) Missiles in tornadoes the siteshave very sophisticatedand complex weather fore-
are becomingmore complicatedthan we used to think. The casting for local weather. For example, when a tornado
touched down at the Savannah River Plant last March,
slide I'm having projected shows the aftermath of the
Huntsville, Alabama, tornado. What you see there are about people were sheltered. There was some damage but no
serious structural damage. Also, if there is any kind of
eight automobilesthat have been wrapped together by the
release of hazardous materials, then we can do tracking and
tornado. How would your model handle somethinglike that?
postulatepotentialtrajectoriesfor the releaseof materials.
(Ravindra)The plants are designedto withstandautomobile
impact, based on speed and weight. The estimates are PAPER G5
conservative,so they could withstand additionsfrom multi- Presenter, Larry Twisdale, Applied Research Associates
ple automobiles.Also, the plants are typically locatedin (L. A. Twisdale, Jr., not in this volume, Safety assessment
isolated areas, where you wouldn't expect so many cars to of critical facilities for tornado effects)
impact on them at one time.
(Arnold Court, California) The winds that you are using, are
(Golden) That's not true. In Amarillo, Texas, they have a lot those fastest mile?
of automobilesand there' s a plant just to the north. Nor is it
true at Rocky Flats, Colorado, where there is a main (Twisdale) No, those are gusts.
highwaythat goes right in front of the plant. (Court) One second, 10 seconds, what?
(Ravindra) When doing a risk analysisfor a nuclear power (Twisdale) Damage-producing gusts averaged over, we
plant, one would considertwo stages.Where the plant is would guess, several seconds.
beingbuilt, and a secondstagewhen it is operating.Then
(Court) Two-second gusts?
there is the strengthof constructionmaterialto consider.We
do consider all of those things in the analysis. (Twisdale) No, I wouldn't say two seconds.! would just call
them damage-producinggusts. We don't know what the
(Golden) I would like to ask a second question. In your
averagingtimes are, sincetornado wind speedsare basedon
hazard analysis,do you use Storm Data? i mean, do you
look at the storm statistics?
damage produced at the ground. So it would vary, structure
by structure. Most engineerswould say it's just damage-
(Ravindra) Yes. producing
 MURRAY 407

(Jim Hill, Departmentof Energy) The tornadomodel that (Rudy Engleman) I'm a little surprisedthat you have so little
you used, was that the NRC model or was that a site- risk involved with operator error. In view of Three Mile
developed model? Island and Chernobyl and other incidents, I've sort of come
to the conclusionthat operator error is the biggestproblem.
(Twisdale)That was a modelwe developedas part of an
EPRI program. We reviewed three or four models that (Twisdale) That's an interesting comment. We used inputs
containedthe basicphysics,and we addeda tornadobound- from the plant on operator error. However, we did a study
ary layer and inflow. We essentiallysynthesizedthe model where we assumedperfectly reliable and perfectly unreliable
from Fujita and a few others. We developedprobability operators. There was little difference for the sequenceof
distributionsfor four or five key parametersthat were best tornado-initiated events.
guessesat the time. We also put in suctionvortices.They
were added in 1979 or 1980 because we were concerned that (Engleman)This is just for tornado-initiatedevents?
they could lead to more missile effects. What we found was
(Twisdale) Yes, any damage that shows up is just because
that if you hada mini vortex with the peakspeedthe sameas
you've got the tornado.
in the parentvortex, you coulduse the parentvortex model
and get essentially the same results. (Ravi Ravindra, EQE Engineering) I don't want to leave the
(Hill) What was the maximum wind speedthat you used? impressionthat the older plants will go without evaluation
and identification.The main purposeof the IPEEE is to look
(Twisdale) We tracked the hazard curve, so it's a risk for vulnerabilities, and it calls for a detailed walk-down
assessment.
throughthe plant. Some of the thingswe have learned over
(Hill) I understand, but what would be the upper limit on the last 20 to 30 years of tornado design and that we've seen
your curve? in real tornado damage can be applied. If vulnerabilities
exist, the IPEEE will correct them in older operatingplants.
(Twisdale) The curve goes to infinity. The curve that I In new plants, obviously they don't exist becausewe have
developed goes asymptotic at about 300 mph. And it has a taken care of them in the design.
slightly downward slope, which means that as you get to
higher and higher risk, you would expect less incremental (Twisdale) I agree with you except that I think that identi-
increase in wind speeds. fying these vulnerabilities,when we've done so few analy-
ses,is heavilydependenton the individual.There are a lot of
(Hill) For the example you showed, what was the return
caseswhere, at least at plants we've looked at, somethings
period of the wind that causedthe damageto the facility.
have cropped up that don't necessarilyshow up from the
(Twisdale) All of them contributed, and the range from F2 to walk-down. Sometimes it's good to run through the num-
F4 dominatedthe prediction of failure frequencies.
 Advances in Tornado Climatology, Hazards, and Risk
AssessmentSince Tornado Symposium II
THOMAS P. GRAZULIS

St. Johnsbury, Vermont 05819

JOSEPH T. SCHAEFER

ScientificServices
Division,NationalWeatherServiceCentralRegion,KansasCity, Missouri64106

ROBERT F. ABBEY, JR.

Officeof NavaI ResearcT•,

Arlington,Virginia22217

1. RISK ANALYSIS In addition,the NRC establishedspecificquantitativegoals

for tornadoclimatology.Prior to this time, as illustratedby
1.1. Introduction
Court [1970],most climatologicalstudiesinvolvedthe cre-
ation of tornado occurrence maps, with little or no separa-
At Tornado SymposiumII, R. F. Abbey, Jr., of the
tionof eventsby destructivepotential.The NRC requiredan
Nuclear RegulatoryCommission(NRC), presentedan out-
estimate of the actual maximum wind speed values that
line of the history of effortsto advancethe scienceof
couldbe expectedfrom tornadoesfor differentlevels of
tornadoclimatologyand risk assessment. This presentation
summarizedthe relatively slow progressin the field prior to probability,particularly10-". Thefirstnumerical models
that could be used to create wind speed probability maps
1971.However, a few advancements in the yearsjust prior
from climatologicaldata were developedprior to Tornado
to that 1976Symposium drasticallychanged
the courseof
Symposium II. This was followedby yearsof refinement,
tornadoresearchandpermanently alteredourdiscussions
of
tornadoes in general.Onewasthedevelopment
analysis,discussion of limitations,
andthedevelopment of
oftheFujita alternativemodels.With NRC support,tornadoclimatology
scaleof tornadointensity[Fujita, 1971].Of equalimportance
andrisk analysisadvancedsteadilyfor a decade.However,
was the decisionby the NationalWeatherService(NWS)
and National Severe Storms Forecast Center (NSSFC) Di- by 1984onlya few scatteredprojectsin risk analysiswere
stillunderway,andthosewerelargelyeffortsto solidifythe
rector,A. Pearson[FujitaandPearson,1973],to embrace it
gainsof the previousdecade.
as the officialgovernment-certified
intensityclassification
systemfor tornadoes.
Despiteits shortcomings
the Fujita
scalebecamea muchneededpointof focusfor discussions 1.2. Probabili• Map Sets
andanintegralpartof nearlyall moderntornado climatology The first full set of wind speed probability maps was
and risk assessment.
presentedbyAbbey[1976]at Tornado Symposium II. Since
The introduction of the Fujitascalewasalmostcoinciden- then,threeseparate setsof suchmapshavebeencreated.
tal withtheentryof theNRC intotornadohazardresearch. Figure1reproduces twomapsfromTornado Symposium I!,
The NRC, underthe guidance of R. Abbey,Jr., provideda basedon a model by Abbey and Fujita [1975].These
sourceof fundingfor projectsthat rangedfromhistorical eventually evolvedintothemorerefinedsetshown in Figure
climatology to computer simulation of tornado
windfields. 2 [fromTecson and Fujita, 1985],whichusedthe Fujita/
Universityof Chicago(UC) tornadodatabase.A model
(usingthesameUC database)wasdeveloped byJ. McDon-
The Tornado:Its Structure,Dynamics,Prediction,andHazards. ald at Texas Tech. Charts are shownin Figure 3 as they
GeophysicalMonograph79
Copyright
1993by theAmerican
Geophysical
Union. appeared
in theworkof McDonaMandAllen[1981].
410 ADVANCES SINCE TORNADO SYMPOSIUM !I

reproduce diagrams from Fujita [1978] dealing with wind

distributionin suctionvortices. They illustrate a small part
of the Fujita-Abbey attempt to deal with the horizontal and
vertical winds fields in tornadoes.
At the heart of this model is the damage area per path
length (DAPPL) method for risk computation. This method
is centeredaround the conceptthat each tornado producesa
damagearea that includes a portion of all F scale levels from
the estimatedmaximum intensity down. The F scale velocity
lines, or isovels,for a modelF-3 tornado are shownin Figure
! 1 [from Fufita, 1978].
A difficultyin implementingthis model is the assignment
of appropriateareasfor eachF scaledamage.Sincefull-path
photographicaerial surveysrarely are performed for torna-
does, it was necessary to develop a model using data
collectedfrom the most thoroughly surveyed population of
tornadoesthat were available. That was, and is, the spec-
tacular outbreak of April 3-4, 1974.
From extensiveanalysisof the 148tornado paths on April
3-4, 1974, empirical formulas were developed that related
the width of the tornado track and its maximum intensity to
the distribution of damage by F scale within the track. For
instance, a 400-yd-wide (366.2 m wide) F-5 tornado might
containa 5-yd-wide (4.6 m wide) swath of F-5 intensity. The
remaining395 yd (361.6 m) would be distributed as 7 yd (6.4
m) ofF-4, 17 yd (15.6 m) of F-3, 41 yd (37.5 m) of F-2, 97 yd
(88.8 m) of F-I, and 233 yd (213.3 m) of F-0. In reality, of
course, the F-5 damage swaths would be larger and scattered
along the track. There has been some discussion as to
whether the tornado population of April 3-4, 1974, actually
Fig. 1. Wind speed probability charts based on the model by is typical of all tornadoes in all areas of the country. It has
Abbey and Fujita [1975] and presentedat Tornado SymposiumII: been suggestedthat perhaps these tornadoes have a higher
(a) occurrence
probability of 10-6 and(b) occurrenceprobability
of percentage of F-3, F-4, and F-5 damage areas than the
10-5 '
general population of tornadoes. Since only a few of the
most destructive tornadoes each year ever get a full photo-
graphic survey, there is no alternative to using this outbreak,
An alternate model, using the N SSFC data base, was and no conclusive data exist to resolve any debate on the
developedby J. Schaeferand,colleagues at the NWS offices subject.
in Kansas City, Missouri. It does not deal directly with Path length is an important parameter in the DAPPL
specificwind speeds,but rather gives the probabilitythat method. Extensive efforts were made to produce models to
tornadoesgreaterthan each individual F scalecategorywill correct path lengthsfor various topographicand population-
occur. Some maps are shown in Figures 4-7 as they ap- based factors. Indices for six parameters that might affect
peared in the work of Schaefer et al. [1986]. path length reports were created. The United States was
There were four major efforts in tornado risk analysis divided into 13,690 subboxes, each 15 min by 15 min. A
duringthisperiod, all fostered(at leastin their early stages) value for each parameter was assignedto each box from 1970
by the NRC. These will be analyzed in a bit more detail. censusdata and U.S. Geological Survey (USGS) 1:250,000
scalemaps. This was a tediousprocess, becausecensusdata
are organized by county, not !5 ø box. There were some
1.3. University of Chicago Model
efforts at UC to adjust tornado paths for the percentageof
The most comprehensive study on tornado risk was water or forest area, slope, road separation, number of
headedby T. T. Fujita at the University of Chicago.That communities,and population. Tecsonet al. [!983] presented
work culminated in that of Tecson and Fujita [!985] and maps for population and topographyindices, and these are
Fujita [1987]. Much of the basis of the model for the reproduced in Figure 12.
FujitaJUCeffort was detailedby the Fujita [1978] and by In the end the area in each subbox affected by tornado
Abbey and Fujita [!979]. This rigorous treatment of the damageof a given F categorywas computedby taking the
"design tornado" addressessuch conceptsas the wind sum of the products obtained by multiplying the adjusted
speed distribution in suction vortices. Figures 8, 9, and 10 path length for each F scalegreater than or equal to the
GiL•,,.ZULIS ET AL.
412 ADVANCES SINCE TORNADO SYMPOSIUM
 GRAZULIS ET AL 413

Fig. 4. Probable annual tornado hazard. Contours are labeled in

negativepowersof 10per year(i.e., 4 indicates10-4). Maximaare Fig. 6. Probableannual tornado hazard from F-3 or greatertorna-
denoted by "X," minima by "N." Stippled area has a hazard does; legend as in Figure 4.
greaterthan4 x 10-3 peryear.Dottedlineis 10-8 or less.

Similarly, if the path lengthor path width is missingfrom the

of interest by the appropriate mean path width, and the report, then the tornado had minimal ground contact or was
DAPPL ratio. This approximationrequiresthat path length extremely narrow and impacted very little area. Caution
not be correlated with path width, and that the DAPPL must be exercisedwhen interpretingthis chart in areas such
numbersobtainedfrom the outbreak of April 3-4, 1974,be as the Dakotas or Colorado, where many tornado records
universally true. contain no path width estimates and the observed damage
area data may tend to underestimate the true hazard.
1.4. NWS/NSSFC The NWS office in Kansas City used the same empirical
methodsto produceother chartswhich can be termed risk or
The research effort at the NWS office in Kansas City,
hazard maps. Figure 13, reprinted in many different govern-
Missouri, analyzed tornadoes documented in the National
ment publications, shows the frequency of occurrence of all
Severe Storms Forecast Center (NSSFC) data base. Statis-
tornadoes
withina 10,000-mi
2 (25,900km2) box. Figure14
tics consideredwere occurrence,affectedarea, and specific
[from SchaeJ•ret al., 1980] showsthe isoplethsfor the area
F scale category rather than specific wind speeds. These
thatwill contain1 mi2 (2.59km2) of tornadodamage
in 100
efforts were headed largely by J. T. Schaefer. The resulting
years.
probability maps were shown in previous figures as pub-
lished by Schae•'•r et al. [1986]. Here the hazard probability
is obtained by summingthe individual damageareasof each 1.5. Texas Tech UniversiO'
tornado of intensity greater than the desired F scale thresh- The Texas Tech University group, headed by J. McDon-
old that was reported within localized quasi-homogeneous ald, divided the United States into the 14 regions. Empirical
regions. This tornado-affected area is then normalized by the tornado area-intensity relationshipswere determinedin each
size of the region and the duration of the data base. of these "global" regions. This was an effort to addressthe
Implicit in this analysis is the assumptionthat all signifi- concept that tornado characteristicsare not geographically
cant information about a tornado is reported. If no F scaleis uniform. For instance, tornadoes in Florida have different
reported, then the tornado produced negligible damage.

Fig. 5. Probable annualtornadohazardfromF-2orgreatertorna- Fig. 7. Probable annualtornadohazardfromF-4 andF-5 torna-

does;legendas in Figure4. does;legendas in Figure
414 ADVANCES SINCE TORNADO SYMPOSIUM II

Vertical Distributionof Max.Windspeeds

of F• Tornado ft
3OO

'--1200

..,,Mr= 210 rnph

ß , -- I00

49m
50 I00 150 200 250mph

Fig. 8. Themaximum
windspeeds
insidean F-4 tornado
withan embedded
suction
vortexasmodeled
by Fujita
[1978].

area-intensity characteristics than those in Texas and Okla-

homa.
The Texas Tech methodologyused empirically derived
20%
occurrence-intensity relationships for local regions (3ø
squares) containing the 1ø square of interest. Thus the
probability of experiencingor exceedingwinds in any F
scaleinterval could be determinedfor any point.
Texas Tech determined that at least 6500 tornadoes were
probably missing from the earlier years of the NSSFC
tornadodata set (1950-1972).These 6500 "missing" events
were distributed across the United States and included in the
+Ro
probability calculations.The distributionis shown in Figure
I":" I00% 15. The placementof theseunreportedtornadoeswas based
I HitProbability
on studiesof reportingefficiencyas determinedby LaGreca
Fig. 9. The "hit probability"acrossthe path of a tornadowith an []980].
embeddedsuctionvortex as modeledby Fujita [1978].

SCHEMATIC ISOVELS OF AN F3 TORNADO

A2(.•r-"Area inside I2 isovel of F3 tornado

VELOCITY PROFILE at Hi
-'-'" -= : • '...•
(SUCTION
VORTEX)
-Y... .•T+• v.+,+•
Beg
•ng
-•'• '-.'••3.:
'::'"':"':
"':":'
'..... ' '

LeftSide /T• '•7•0• VORTgXRight

Side L
,,
Lo

A21•3)
.....
Mean
width
oftheareainside
the
W2(r3)
= L2(.•) I2 isovel
ofF3tornado
Fig. 10. Velocity profile across a tornado with an embedded Fig. 11. A schematicof the F scaledistributionof damageacross
suctionvortex as modeledby Fujita [1978]. the swathof an F-3 tornadoas modeledby Fujita
 GRAZULIS ET AL. 415

Population Tndex
0 1 2 • 4 5 6 7 8 9

Topography Zndex •
- • & , ß i ii i O

Fig. 12. Chartsof Fujita'sindices

for(top)population
and(.bottom)
topography
aspresented
by Tecson
et al.
416 ADVANCES SINCE TORNADO SYMPOSIUM II

Fig.13. Average
annual
tornado
incidence
per10,000
mi2 (25,900
km2),aspublished
by NationalOceanic
and
AtmosphericAdministration[1982].

1.6. Twisdale/Electric Power sourcesof error he identified were unreported events, unre-
Research Institute ported F scale, direct misclassificationof F scale, and
misclassification of F scale due to the randomness of dam-
Twisdale [1978, 1982, 1983], of the Electric Power Re-
searchInstitute (EPRI), used a rigorous statisticalapproach age. He synthesizedhis data into four broad regions and
to identifyprobablesourcesof error in the NSSFC tornado givesan analysisof occurrencerate, intensity,path length,
data set and to deal with them on a regional basis. The path width, and directionof tornado movementfor each.

Fig. 14. isopleths

of the areathatwill contain! mi2 (2.59km2) of tornadodamage
in 100yearsaspublished
by
Schaefer et al.
 GRAZULIS ET AL. 417

Fig. 15. The numberof unreported

tornadoes,1950-1979,per 1ølatitude/longitude
squareasestimated
by McDonald
and Allen [1981].

Twisdale [1978] suggestedan alternative set of wind differencebetween the wind speedestimatesproducedby
speedsto be associatedwith the Fujita scale.Twisdale'sand the two models. The exact reasons for the differences are
Fujita'svaluesare comparedin Table 1. The debateas to just complexanda productof the assumptions
in the modelsand
whatwindspeedsareneededto causejust whatkindof damage the way the data are handled.
also continues.Any adjustmentsto the Fujita scale, if and The largestdifferencesare in the Ohio Valley area of the
when needed,wouldcomeby eitheralteringdescriptions of the April 3-4, 1974,outbreak.This maybe becausethe Fujita/
kindsof damageascribedto eachwindspeedrangeor changing DAPPL model uses data extendingback to 1916, while the
the wind speedsassignedto eachof the six levels. McDonald/Texas Tech model uses data beginning in 1950.
Grazulis [1991] noted that the 1916-1949 period was less
1.7. Fujita-McDonald Comparison active for violent tornadoes in the Ohio Valley. Grazulis also
the FujitaandMcDonald suggested
Despitedifferingmethodologies, that the wind speedmay be exaggerated
at the
site-specific agreefairlywell Pinelias,Florida, site. The data there are heavily influenced
10-7 windspeedprobabilities
by the violent, 135-mi-long(217-kinlong), Gulf-to-Atlantic
across much of the United States. Differences are usually
lessthan50mi/hr(mph)(22.4m s-]) at anysite.Coatsand tornadotrackof April 4, 1966.This entiretrack wasgivenan
Murray [1985],in a reportby LawrenceLivermoreNational F-4 ratingbasedon the destructionof a singlehomein a rural
Laboratory (LLNL) for the Departmentof Energy, com- communityin central Florida. The event was probablya
pared theFujitaandMcDonald for10-7 windspeed tornadofamily which was not brokeninto individualfamily
models
probability at 25 sites. members.
These sites are listed in Table 2 and ordered by the Perhapsthe westcoastof Florida,the FloridaPanhandle,
central Florida, and the hurricane-proneareasof southeast
Florida should all be treated as independentsubregions.
TABLE 1. Comparisonof Wind Speeds While other statesmay need to be treated in this way, the
Fujita Range, TwisdaleRange, meteorological factors that underliethe long-termtornado
Scale mph mph risk are not as readily identifiableas they are in Florida.
Difficulties associated with analysis of long-track torna-
0 40--72 40--73 does and tornado families were addressedby Doswell and
I 73-112 73-103
2 113-157 103-135 Burgess[1988].Their reanalysis of the historicWoodward
3 158--206 135-168 Tornadoof April 9, 1947,is shownin Figure16.Thepaperis
4 207-260 168--209 a summaryand discussion of someof the many difficult
5 261-318 209-277
issuesin tornadoclimatologythat remain
418 ADVANCES SINCE TORNADO SYMPOSIUM II

TABLE 2. Site-Specific
10-7 WindSpeedProbabilities
Fujita (A),* McDonald B-A,. +
Site State mph (B),? mph mph

Mound Ohio 283 364 81

Oak Ridge Tennessee 261 340 79
FMPC Ohio 287 364 77
Portsmouth Ohio 257 330 73
Paducah Kentucky 289 340 51
Argonne, east Illinois 276 318 42
Pantex Texas 271 297 26
Rocky Flats Colorado 204 228 24
Pinelias Florida 268 244 -24
Princeton New Jersey 229 210 -19
Site 300 (LLNL) California 164 182 18
Bendix/Kansas City Missouri 292 310 18
Stanford California 182 165 -17
Sandia, Albuquerque New Mexico 175 191 16
Nevada Test Site Nevada 152 136 -16
Broo 'khaven New York 224 215 -9
Los Alamos New Mexico 183 190 7
ETEC California 179 174 -5
Berkeley California 168 165 -3
Hanford Washington 179 177 -2
Sandia, Livermore California 164 165 1
Argonne, west California 185 184 -1
Idaho (inel) Idaho 185 184 -1
Livermore (LLNL) California 164 165 1
Savannah River South Carolina 283 283 0

*The 10-7 probability

maximum
windaccording
to theFujitamodel.
'•The10-7 probability
maximumwindaccording
to theMcDonaldmodel.
*+The difference between the two models.

1.8. Engineering Studies Grazulis [1991] also noted difficulty in locating newspaper
descriptionsfor tornadoes that occurred on Saturday.
There have been many important wind-engineeringstudies
It seems likely that these unreported tornadoes would be
which potentially can impact tornado climatology but were
weak. However, the Saturday minimum is present at even
not, in themselves, climatological studies. The most signifi-
the higher F scale levels. This may be due to a lack of
cant of these was probably by Minor et al. [1977], which
broadened the discussionof what kinds of damage occur at impressive newspaper damage photographs, resulting in a
various wind speeds. The concept that structural failure of tendency for Saturday tornadoes to be rated lower. This
hebdomadal distribution does not hold for tornado out-
homes and other buildings (resulting in complete "destruc-
tion") can take place under both lower F scale wind (F-2) breaks. Two of the 10 largest 1-day tornado outbreaks
and higher F scale winds (F-4) continues to evolve. The occurred on Saturday (the lowest tornado day). In contrast,
irregular means by which meteorologists become aware of Thursday and Friday (the two most likely tornado days)
engineeringconcept topics may affect climatology by creat- accounted for zero and one of the largest outbreaks, respec-
ing state-to-state inconsistenciesin the application of the tively. Tecson et at. [1979] noted a similar day-of-the-week
Fujita scale. distribution but found it less significant. Statistical studies on
tornado path length and path width were done on the UC
data set by Tecsonet al. [1979], and on the NSSFC tape by
2. CLIMATOLOGY
Kelly et al. [1978] and by Schaefer et al. [1986].
The paper by Kelly et al. [1978] contains detailed analysis
2.1. General Climatology
of tornado occurrences for each F scale category. Among
Wilson and Kelly [1977] examined the distribution of the findings were that violent (F-4-F-5) tornadoes occurred
tornadoes by day of the week and noted a statistically at all times of the day or night. Weak (F-O-F-1) and strong
significant (10-20%) lower number of tornadoes on Satur- (F-2-F-3) tornadoes showed more diurnal trends. Studies
day. This studyconcludedthat Saturday tornadoeswere less were done of tornadoesin three geographicalregions. Figure
likely to be reportedby newspapers.Newspapersalert NWS 17 shows clearly that the southeast has the greatest fre-
personnel to the occurrence of about 10% of tornadoes, quency of morning tornadoes.
which probably would not have been reported otherwise. Table 3 provides some comparativepath length and
 GRAZULIS ET AL. 419

500

WICHITA []

100
_ OK •o4•
/
TX .•
_.X wooo,•A.D
CANADIAN 50

AMARILLO
II

10

Fig. !7. The diurnaldistributionof tornadooccurrence time (nor-

malizedwith respectto sunrise/sunset) in the Great Plains(GP,
dottedline), midwest(MW, solidline), and southeast (SE, dashed
WICHITA [] line) [after Kelly et al., 1978].

KS ,/!
l_ OK /
....

TX
•,•"••,• CA
•ADIAN
width valuesfor UC data [from Tecsonet al., 1979]and for
NSSFC data [from Schaeferet al., 1986].A comparisonof
the N SSFC meanshowsa shorterpath lengththan for UC
AMARILLO
II .J'
ßPAMPA
•LAH•A CITY
meanF-2-F-5 tornadoes.This is possiblydue to the lower
path length for many tornadoesin the 1916-1949,pre-
NSSFC period,coveredby UC data. More completesur-

b
WICHITA
FALL•• veys in "modem" times (i.e., since 1953) tend to find that
actualpathlengthsare often longerthan initiallyreportedin
the press.The 1916-1949shorterpath lengthsmay have
been, in some cases, only the length of the path that
Fig. 16. Path of the April 9, 1947, tornado which struck Wood- producedobviousdamage.The UC path lengthsfor F-0--F-1
ward, Oklahoma, (a) as originally classifiedand (b) as reclassified, are longerthanthe NSSFC path lengths.This may be due to
on the basis of a review of the data. Tornado occurrence times
shown in Figure 16a are in central standard time. The dashed
the nonlistingof minor, short path length tornado in the
segmentin Figure 16b correspondsto what was in 1947,and still is, 1916-1949period. Another factor may be that the UC data
a very sparsely populated region within which path continuity set (column 1) gives poorly documented tornadoesa default
cannotbe determineddefinitely[after Doswelland Burgess,1988]. path length of 0.5 mi (0.8 km). The calculations from the

TABLE 3. Comparative Path Length and Path Width Data

1950-1982 1950-1982 1950-1978 1950-1978
1916-1978 NSSFC NSSFC NSSFC NSSFC
UC Mean Mean PL, Median Mean Median
F Scale PL, mi mi PL, mi PW, yd PW, yd
F-0 1.41 1.11 0.30 46 17
F-I 2.94 2.59 0.98 93 47
F-2 5.56 5.66 2.19 167 99
F-3 10.25 12.08 6.76 290 180
F-4 18.66 22.42 ! 3.80 432 297
F-5 28.55 34.17 23.44 616 496
All 4.67 4.40 0.98 128 48

Abbreviationsare as follows' PL, path length' PW, path width. One mile equals 1.6 km, and I yd
equals. 0.9
420 ADVANCES SINCE TORNADO SYMPOSIUM II

Some conversion from kilometers to miles and miles to yards

was done from the original papers.
Gal•vav and Pearson [1979] studied winter tornado out-
_

breaks, finding they were often accompanied by heavy snow

or ice conditionsto the northwest of the track of the synoptic
low. Winter outbreaks were comparable in violence to spring
outbreaks but caused an even higher number of deaths than
spring outbreaks, despite average or above average watch
and warning times. This is possibly due to the unexpected
nature of the events at that time of year. Winter outbreak
tornadoes are mapped in Figure 18.
Gal•vay's [1983] climatology of killer tornadoesfound total
death statistics dominated by F-4/F-5 tornadoes: however,
half of all killer tornadoes caused only a single death. He also
found a clear relationship between long-track tornadoes and
high death totals for individual tornadoes.
An apparent natural division of the United States into
eastern and western groups of tornadoes was revealed by
Tecson et al. [1982]. The two large groups of tornadoes are
Fig. 18. The distribution of 372 tornadoes associated with major
•inter (December-February) tornado outbreaks between 1950 and separated by a slanted axis (along a relative minimum of
1977 [after Gal•vay and Pearson, 1979]. activity) that extends from upper Michigan and northern
Wisconsin through the Ozark mountains to southeast Texas.
A plot of total tornado path length, from west to east, across
NSSFC tape used only tornadoes that had a path length and the United States is shown in Figure 19. The central mini-
a path width recorded. The NSSFC data set for the 1950- mum is apparent. This central minimum was also located by
1982 period had 22,840 tornadoes, of which 14,243 had both Grt•zulis and Abbey [ 1983] in a documentation of F-4 and F-5
a path length and a path width. tornadoes, as seen in Figure 20. It was concluded in both
It is worth noting the difference between the mean and the papers that population may not be the sole reason for the
median values. Relatively few very long track tornadoes existence of the relative minimum along the dividing axis.
skew the distribution and boost the mean value so much that Changnon [1982] looked at trends in tornado days, rural
the median is a more representative figure for tornadoes, in population, and the number of days with tornado deaths in
general. Path lengths are in miles, path widths are in yards. Illinois since 1916. He noted that the increase in tornado

•, Eastern
:..•-

Fig. 19. Summedtornadopath length(1916-1980)by F scalewithin slightlyslantednarrowbands(15 min of longitude

wide) stretch from 30øN to 42øN [after Tecson et al.,
 GRAZULIS ET AL. 421

the many potential pitfalls that await the tornado climatolo-

gistare quite valuable.He alsonotedhow easilytornado
statisticscan be twisted and misinterpreted.The "user
beware"noticeheadvocated, albeittonguein cheek,isvery
applicableto most tornado statistics.

2.2. The Annual Number of Tornadoes

Eshehnanand Stanford [1977], in a study of Iowa torna-
does for the year 1974, reported that funnel touchdowns
couldbe confirmed81 times in Iowa during 1974.This is 3
times as many tornadoes as the officially recorded 27.
Similardrasticdifferences
in the numberof actual,compared
to the number of recorded, tornadoes were found in the
statisticscompiledby Snider [1977]on Michigantornadoes.
Schaefer and Galway [1982] divided counties into four
groups based on population density. They noted that the
tornadodistributionwas rather uniform acrossthe catego-
Fig. 20. A map of violent (F-4 and F-5) tornadoesfrom Grazulis ries and that on a nationwide basis there was little statistical
andAbbey[ 1983].Isoplethsaredrawnfor40and80pathmiper 1000
mi2 (1mi- 1.6km).Twenty
subregional
maxima by evidencefor populationbias, except in countieswith ex-
areenclosed
80-mi isopleths.A central minimum is so labeled, and a relative tremely high populations.This lack of a strong statistical
minimumin the sparselypopulatedFlint Hills area of Kansasand biasfor most countiesruns contraryto the standardlogic
Oklahoma is marked with an "F."
that few tornadoreportswould comefrom countieswith few
people.This lackof populationbiasis mostevidentalongthe
western plains from Oklahoma through Kansas, to the
numberssince 1916took place while rural populationwas Dakotas (see Figure 21).
droppingandthat the increasewasdueto a gradualincrease Grazulis [1991]found that in the moderatelypopulated
in awareness. It wasconcludedthatat leastsomeof thelarge counties of eastern Kansas, about 50% of tornadoes hit
increase in tornado numbers from 1971 to 1980 was due to buildings.Oddly enough, he found that same percentage
the reporting of various nontornadic windstorms as torna- (50%) of the tornadoesin the sparselypopulatedrural
does. Many of the trends arise from changesin society, countiesof westernKansasalso hit buildings.Becauseof the
rather than changesin the meteorology.His cautionsabout variancein buildingdensity, this stronglyhints that there is

Fig. 21. Apparent population bias by county in the number of reported tornadoes, 1950-1979. Top number is for
countieswitha 1970population
densitygreater
thanorequalto250mi2;placeisforpopulationdensity50to249mi2;
penultimate
number isforpopulation
density
of 10to49mi2;bottomnumber isforcountieswithpopulation
density
of
lessthan10mi2. (Onemileequals1.6km.Asteriskindicates
nocounties
inthatcategory;
a dot'afternumberindicates
less than two counties in category.) No bias indicated by 1.00; numbersgreater than 1.00 indicate overreportingof
tornadoes;numbersless than 1.00 indicate underreporting(see Schaefer and Galway [1982] for
422 ADVANCES SINCE TORNADO SYMPOSIUM II

TABLE4. Comparison
of Numberof Tornadoes
Reported
in UrbanCounties
Containing
StateCapital
to Statewide
Tornado
Counts
Tornadoes Tonadoes Tornadoes Tornadoes
Recorded Recorded Estimated Recorded
City, State (State Area)* County (Area)* in County? in State? in State:• in State,õ%
Little Rock, Arkansas (52,078) Pulaski(767) 30 597 2037 29
Indianapolis, Indiana (35,932) Marion(396) 19 557 1724 32
Oklahoma City, Oklahoma (68,655) Oklahoma (708) 48 1515 4655 33
Des Moines, Iowa (55,965) Polk(582) 26 934 2501 37
Topeka, Kansas (81,778) Shawnee (549) !8 1105 2681 41
Springfield,Illinois (55,645) Sangamon (785) 29 785 1863 42
Jackson, Mississippi (47,233) Hinds(875) 29 686 1565 44
Columbus, Ohio (41,004) Franklin(982) 13 441 982 45
Austin, Texas (262,017) Travis(989) 31 3861 8213 47
Madison, Wisconsin (54,426) Dane(1205) 17 599 768 78
*Areasgivenin square
miles;1 mi2 = 2.6km2.
•-NSSFC data for 1959-1989.
•:Based
ona simple
proportion,
thenumber
of tornadoes
thatwouldhavebeenreported
in thestateif thecapitalcountywere
representative.
õThepercentage
of tornadoes
actuallyreported,
compared
to theestimate.

a higher percentageof unreportedtornadoesin western tornadoesacross corn fields near Springfield, Illinois, on
Kansas than in eastern Kansas. He thus concluded that August6, 1977 (see Figure 22), and discussedthe larger
tornadonumbersactuallyincreasedas onemovedwestward problemof whichvorticesshouldbe officiallycalledtorna-
across Kansas. This increase in tornado numbers compen- does. Many of these August 6, 1977, tornadoeswere gust
satedfor the dropin population, thusyieldingthe statistical front or other shear zone vortices. That study hints at the
lack of a populationbiasin reportedtornadoes. possibilitythat somelarge areas of thunderstorm wind
This same situation occursin Wisconsin. A cursory glance damagemay be fiddled with small tornadolikevortices.
at the damagecodeson the NSSFC data baseshowsthat Category3 containsmostlyweak tornadoesand probably
virtuallyall of the countiesin the heavilyforestedareasof accounts for most of the 2000 "missing" events.
northeastWisconsinreporthigh damagefiguresfor all tor- Fujita [1987]presentsa wide variety of climatological
nadoes. A check of Storm Data and newspapersindicates maps,as derivedfrom the UC data tape, through1985.
that this is becausethe tornadoeshit one or more buildings. Figures23, 24, 25, and26 area few samples of the over 120
It is highlyunlikelythat 100%of all tornadoes wouldhit a figures in thatvolume,whichdisplays thegreatvariabilityin
buildingin the sparselypopulatedcountiesof northern tornado activity.
Wisconsin. It should be noted that Grazulis [1991] studied
only Wisconsinand Kansas,and only to the extentthat it 2.3. State CIimatologies
provideda hintat explaining
thelackof a strongpopulation
bias found by Schaeferand Galway [1982]. Tornado climatologieshave been compiled for many
Grazulis [1991] comparedthe number of tornadoesre- states. Because of the number of them, only a few of the
ported in urban countiescontainingthe state capitalto more notable are cited here. In California, Hales [1983]
statewidetornado counts(Table 4). Other capitals, suchas found several areas of concentrated tornado activity. He
Nashville, St. Paul, or Lansing, and large cities, such as suggested
that the relative maximumin the numberof
Birmingham,Minneapolis,St. Louis, Chicago,Omaha,and tornadoesoccurringin the Los Angelesarea is causedby
Dallas, could be used to generateeven more drastic num- enhanced
convergence
inducedby localtopography.There
bers. On the basis of these results, Grazulis [1991] suggested is a southeastto northwest curvature in both the coastline
that about 2000 tornadoes go unreported each year. He and an adjacentmountainbarrier that is positionedsome-
noted that most of these "missing" tornadoesprobablyfall what inland, rather than right along the coast.
intothreecategories:(1) tornadoeson thewesternplainsand Speheger et al. [1990]notedsignificant
differences
in the
in other very rural areas that were unseenand thus unre- numberof reportsof Indianatornadoes, funnelclouds,and
ported;(2) poorlyformedtornadoesthatwerenot recogniz- thunderstormwinds between the 1980s and previous de-
able as such from a distance and hit nothingexcept perhaps cades. The number of funnel clouds and tornado reports
tree topsor openfields;(3) gustfront andothershearzone were decreasing while the numberof downburst/microburst
tornadoesthat have a convergentrotation but may not be wind reports were increasing.Thesetrendswerethoughtto
connected to the base of a thunderstorm, and yet they are be, at leastin part, dueto changes
in reportingprocedures
typicallyreportedas tornadoesby the generalpublicin andto changing attitudestowardeventclassification
within
urban areas. the publication Storm Data.
Forbes and Wakimoto [1983] studied an outbreak of 18 Someaspectsof the uniqueand little studied
 SPRINGFIELD TORNADOES and DOWNBURSTS
August 6. /977
-::::::- M ~
9~ 1lI7~ '-.Q.J' No.5 ~ ---- ....
OOlIr_uttt.T ..lellte...",n ...... ~,. TaUIMIO . . . . .,.... It,.•••
- ...
t I ' i! 1 t J \ i' I' i '1-:.....
r ~
__, .....
I\ ~ SANG ACOUNTY
MON.

/
--- .~
6-., ~1~Urg
- '~"C~t~~;;:~ __ \'I
_ _ ....
~
~~ ~~.'~=
---
---- ./ - I
~
3~
-----
,,~'" ~~:~:::--':_-=-_";.
~.
~ . ~~~
~-- ~- ~-
4~ _.__ __ _
...--: ---- -------- - - --- l'f!.
".orr:,
./ I ~ t:t)Kincoid
M, _----- .,'" - _,/ , ~ ... '
I
~~"
- ~ '·o~~ -====---=--,g~~
~ _
~,. ===--~-----_• ~.
__
..- \\ \'I
.'-., ill t.// I:
' , CHRISTIAN COUNTY
MORGA N / \ "'
...!JI'- "______
I -,__ "'.._ . . I
-------"------ ,-
~i
I' --- ----------0------: I MONTGOMERY ,!, _
---- ,I
111.0lIl
MACOUPIN
111.0lIl
I
- .
1 0

Fig. 22. Mapping of damage from tornadoes (identified by "No."), downbursts (identified by large numeral), and microbursts (identified
- ~
t:
by "m") on August 6, 1977, in central Il1inois. "Streamlines" of damage and F scale contours are mapped [after Forbes and Wakimoto, Vl
1983].
~
~
~
""
424 ADVANCES SINCE TORNADO SYMPOSIUM II

Fig. 23. Chartof the summedpathlengthper 15-minbox of Decembertornadoes,1916-1985,constructed

from the
UC data tape [after Fujita. 1987].

climatology of Florida were noted by Anthony [1988]. ever, the next "great leap," similar to those in the 1970s, is
Among the findings were that an unusually high proportion probably in the distant future. Such a leap could be a
1,42%)of severe reports in that state were tornadic. national mesocyclone climatology, based on NEXRAD Dop-
pler radar data. The mesocyclonedata might be used as a
basis for adjusting traditional climatological data for less
3. THE FUTURE OF TORNADO CLIMATOLOGY
populated areas. Another significant advancement may
AND RISK ASSESSMENT
come from the use of portable Doppler radar. Such units can
For the foreseeable future, studies in the tornado clima- be brought to within I mi (1.6 km) of even the most intense
tology of individual states and regions will continue. How- tornadoes. These instruments may drastically change our

Fig. 24. Chartof the summed

pathlengthper 15-minboxof tornadoes,
1916-1985,
thatoccurredbetween0600and
0800CST constructedfrom the UC datatape [afterFujita, 1987].The anomalyin Coloradoarisesfrom the tornadoes
of November 4, 1922, while the one in western Nebraska is a reflectionof the events of May 8,
 GRAZULIS ET AL. 425

Fig. 25. Chartof thesummed tornadoes

per15-minboxthatoccurred duringthe1930s
constructedfromtheUC data
tape[afterFujita, 1987].Note thattherewasmoretornadoactivityin IowathanIndianaduringthisdecade.

perceptions of tornado wind speeds and thus require a REFERENCES

rethinking of current risk analysis models. In the meantime it Abbey, R. F., Jr., Risk probabilities associatedu ith tornado wind
is likely that traditional methods of tornado climatology will speeds, paper presented at Symposium on Tornadoes: Assess-
produce only minor refinements in our present overall pic- ment of Knowledge and Implications for Man, Tex. Tech Unix.,
Lubbock, Tex., June 22-24, 1976.
ture. This period would also be an ideal time to addresssome
Abbey, R. F., Jr., and T. T. Fujita, Use of tornado path lengthsand
of the questions concerning tornado documentationand use gradations of damage to assess tornado intensit.• probabilities, in
of the Fujita scale that were brought forth by Doswell and Preprints, Ninth Conference on Severe Local Storms, pp. 286-
Burgess [1988]. 293, American Meteorological Society, Boston, Mass., 1975.

Fig. 26. Chartofthesummed tornadoes

per15-minboxthatoccurred
during the1950s
constructed
fromtheUCdata
tape[afterFujita,19157].
Notethecontrast
intherelative
amount
of activity
in IowaandIndiana
betx•een
thisdecade
and the 1930• (Figure
426 ADVANCES SINCE TORNADO SYMPOSIUM II

Abbey, R. F., Jr., and T. T. Fujita, The DAPPL method for R. F. Abbey, Jr., An augmented tornado climatology. Mon.
computingtornado hazard probabilities: Refinementsand theoret- Weather Rev., 106, 1172-1183, 1978.
ical considerations, in Preprints, 11th Conference on Severe LaGreca, K. W., Factors affecting tornado hazard probabilities,
Local Storms, pp. 241-248, American MeteorologicalSociety, Master's thesis, Tex. Tech Univ., Lubbock, 1980.
Boston, Mass., 1979. McDonald, J. R., and B. S. Allen, Regionalizationof tornado hazard
Anthony, R., Tornado/severethunderstormclimatologyfor the probabilities,Preprint81-540,16 pp., Am. Soc. of Civ. Eng., St.
southeasternUnited States, in Preprints, I5th Conferenceon Louis, Mo., 1981.
Severe Local Storms, pp. 511-516, American Meteorological Minor, J. E., J. R. McDonald, and K. C. Mehta, The tornado; an
Society, Boston, Mass., 1988. engineering-orientedperspective,Tech. Memo. ERL-NSSL-82,
Changnon,
S. A., Trendsin tornadofrequencies:
Factor fallacy,in 192pp., Natl. OceanicandAtmos. Admin., Boulder,Colo., 1977.
National Oceanic and Atmospheric Administration, Tornado safety,
Preprints,12th Conferenceon Severe Local Storms, pp. 42-44,
survivingnature's most violent storms, Rep. NOAA/PA 82001,
AmericanMeteorologicalSociety, Boston,Mass., 1982.
Boulder, Colo., Jan. 1982.
Coats, D. W., and R. C. Murray, Natural phenomenahazards
Schaefer, J. T., and J. G. Galway, Population biases in the tornado
modelingproject;extremewind/tornadohazardmodelsfor de- climatology, in Preprints, 12th Conference on Severe Local
partmentof energysites,Rep. UCRL-535266,LawrenceLiver- Storms, pp. 51-54, American MeteorologicalSociety, Boston,
more Natl. Lab., Livermore, Calif., 1985. Mass., 1982.
Court, A., Tornadoincidencemaps, Tech. Memo. ERLTM-NSSL Schaefer,J. T., D. L. Kelly, and R. F. Abbey, Jr., Tornado track
49, 76pp.,Natl. SevereStormsLab., Environ.Sci.Serv.Admin.,
Norman, Okla., !970.
characteristicsand hazard probabilities, in Wind Engineering,
edited by J. E. Cermak, pp. 95-110, Pergamon, New York, 1980.
Doswell, C. A., III, and D. W. Burgess,Some issuesof United
Schaefer, J. T., D. L. Kelly, and R. F. Abbey, A minimum
Statestornadoclimatology,Mon. Weather Rev., 116, 495-50I, assumptiontornado hazard probability model, J. Clim. Appl.
1988.
Meteorol., 25, 1934-1945, 1986.
Eshelman,S., and J. L. Stanford, Tornadoes,funnel cloudsand Snider, C. R., A look at Michigan tornado statistics, Mort. Weather
thunderstorm damagein Iowa during 1974,Iowa State J. Res., 51, Rev., 105, 1341-1342, I977.
327-361, 1977. Speheger,D. A., D. J. Shel!berg, J. R. Gibbons, J. A. DeToro, J. T.
Forbes, G. S., and R. M. Wakimoto, A concentrated outbreak of Snow, and T. P. Grazulis, A climatology of severe thunderstorm
tornadoes,downbursts,and microbursts,and implicationsregard- events in Indiana, in Preprints, 16th Conference on Severe Local
ing vortex classification,Mon. Weather Rev., 111,220-235, 1983. Storms, pp. 18-23, American Meteorological Society, Boston,
Fujita, T. T., Proposedcharacterizationof tornadoesand hurricanes Mass., 1990.
by area and intensity, Res. Pap. 91, Satell. and Mesometeorol. Teeson, J. J., and T. T. Fujita, Automated mapping of maximum
R•es.Proj., Univ. of Chicago,Chicago,II1., 1971. tornado wind speeds in the United States as a function of
Fujita, T. T., Workbook of tornadoes and high windsfor engineering occurrenceprobabilities, in Preprints, 14th Conference on Severe
application,Res. Pap. 165, Satell. and Mesometeorol. Res. Proj., Local Storms, pp. 21-24, American Meteorological Society, Bos-
Univ. of Chicago, Chicago, III., 1978 ton, Mass., 1985.
Fujita, T. T., U.S. tornadoes, part 1, 70 year statistics, report, 122 Tecson, J. J., T. T. Fujita, and R. F. Abbey, Jr., Statistics of the
U.S. tornadoesbased on the DAPPL tornado tape, in Preprints,
pp., Smell. and Mesometeorol. Res. Proj., Univ. of Chicago,
l 1th Conference on Severe Local Storms, pp. 227-230, American
Chicago,II1., 1987.
Meteorological Society, Boston, Mass., 1979.
Fujita, T. T., and A.D. Pearson, Results of FPP classification of
Tecson, J. J., T. T. Fujita, and R. F. Abbey, Jr., Climatologic
1971 and 1972 tornadoes, in Preprints, Eighth Conference on mapping of U.S. tornadoes during 1916-1980, in Preprints, 12th
Severe Local Storms, pp. 142-145, American Meteorological Conference on Severe Local Storms, pp. 38-41, American Mete-
Society, Boston, Mass., 1973.
orologicalSociety, Boston, Mass., 1982.
Galway, J. G., and A.D. Pearson, Winter tornado outbreaks, in Tecson, J. J., T. T. Fujita, and R. F. Abbey, Jr., Statistical analyses
Preprints, 1Ith Conferenceon Severe Local Storms, pp. 1-6, of U.S. tornadoes based on the geographicdistribution of popu-
American Meteorological Society, Boston, Mass., 1979. lation, community, and other parameters, in Preprints, 13th
Grazulis, T. P., Significant Tornadoes, 1880-I989, vol. I and II, 970 Conference on Severe Local Storms, pp. 120-123, American
pp., Tornado Project, St. Johnsbury, Vt., 1991. Meteorological Society, Boston, Mass., 1983.
Grazulis, T. P., and R. F. Abbey, Jr., 103 years of violent Twisdale, L. A., Tornado data characterization and wind speedrisk,
tornadoes... patternsof serendipity,population,and mesoscale J. Struct. Div., 104(STlO), 1978.
topography, in Preprints, I3th Conference on Severe Local Twisdale, L. A., Regional tornado data base and error analysis, in
Storms, pp. 124-127, AmericanMeteorologicalSociety, Boston, Preprints, 12th Conferenceon Severe Local Storms, pp. 45-50,
Mass., 1983. American Meteorological Society, Boston, Mass., 1982.
Hales,J. E., Synopticfeaturesassociated
with LosAngelestornado Twisdale, L. A., Probabilistic analysis of tornado wind risks, J.
occurrences, in Preprints, 13th Conference on Severe Local Struct. Eng., 109(2), 1983.
Storms,pp. 132-135, AmericanMeteorologicalSociety, Boston, Wilson, L. F., and D. Kelly, Tornado climatology by day of the
Mass., 1983. week, in Preprints, loth Conferenceon Severe Local Storms, pp.
Kelly, D. L., J. T. Schaefer,R. P. McNulty, C. A. DoswellIII, and 194-198, American Meteorological Society, Boston, Mass.,
 ComparativeDescriptionof Tornadoesin France and the United States
JEAN DESSENS

Laboratoired'Adrologie,UniversitdPaul Sabatier,65300Campistrous,France

Jo•4 T. S•40w

DepartmentqœEarth and AtmosphericSciences,PurdueUniversity.WestLafayette, Indiana 47907

l. INTRODUCTION frequenciesfor tornadoes moves from the lower Mississippi

in winter to the central part of the United States in summer
In comparison with the Great Plains of the United States, [see Fujita, 1978, Figures 5.7-5.12]. In France the winter-
severe tornadoes are rare in France. Statistics based on
time concentration in the coastal zones probably reflects
significant(strong (F-2 and F-3) and violent (F-4 and F-5) greater availability of both latent heat energy and (to a lesser
[see Hales, 1988]) tornadoes that have occurred in France in
extent) kinetic energy (winds are strongeralong the coastsin
the past three decades, however, show that this country winter) in these regions, while the summertime events in the
averages two such tornadoes a year. By combining these interior reflect the stronger surface heating that takes place
observations with information from the scientific literature there.
about events that occurred before the modern period, it has
been possible to establish a tornado climatology for France
2.2. Diurnal and Seasonal Distributions
based on 107 significant tornadoes. The main data of this
climatology, recently published by Dessens and Sno•t, Figure 3 showshourly and monthly distributionsof the 103
[1989], are summarized here and compared with American significanttornadoesfor which time of touchdown is known.
tornadoes.
The central portion of this figure is a plot of day and time of
occurrence of each of the events. The histogram along the
2. SUMMARIZED CLIMATOLOGY ordinategives frequency of occurrence by month, while that
alongthe abscissagives frequency of occurrenceby hour of
2.1. Geographical Distributions the day.
Figure 1 shows the geographical location of the 107 Juneand August are the months with the greatestnumber
significanttornadoes. These are distributed in intensity as of tornadoes. The interval 1600 to 1700 UTC is the one in
follows: 50 F-2, 44 F-3, 11 F-4, and two F-5. These events which occurrencehas been most frequent, with a secondary
are concentrated in the northwestern part of the country, maximum between 1800 and 1900 UTC. The mean time for
with a small secondaryconcentrationin the south, around all tornadoes is 1512 UTC (UTC and Sun time are not very
the Mediterranean shore. different, since the Greenwich meridian crosses over
It is clear from the mappingin Figure 2 that the distribu- France). The mean time for "winter" tornadoes is 1322
tion in winter is different from that in summer; during the UTC, while the mean time for "summer" tornadoesis 1546
cold months of November through March, tornadoesoccur UTC, a difference of 144 min.
mainly in the two coastal zones, while during the warm These results can be compared with the diurnal distribu-
monthsof April throughOctober,they occurin the interior. tion of U.S. tornadoes, as summarized in by Abbey [1976,
This is very similarto the distributionsobserved(on a larger Table 2]. While the monthly distributionis different(in the
scale) in the United States, where the region of maximum United States, May and June are the months with the
greatestactivity),the sametrend is presentfor the seasonal
mean times of occurrence: summer tornadoes occur later in
the afternoon, with a difference of 51 min between the mean
The Tornado:Its Structure,Dynamics,Prediction,and Hazards.
times of occurrence for winter and summer events.
GeophysicalMonograph79
Copyright1993by the AmericanGeophysicalUnion. Winter events are more scattered throughout the day than
428 TORNADOES IN FRANCE AND THE UNITED STATES

o
o

ß.....o oß oo o

'e e"

ß0 ß
• ß ATLANTIC

CLIMATE

F5
F4
.%
ß F2

0 November-March MEDITERRANEAN

ß Apri 1-October CLIMATE

Fig. 1. Geographical locations of 107 significant tornadoes that Fig. 2. Geographicallocations with correspondingseason of oc-
occurred in France in the period 1680-1988. Greatest reported currence.The two curved lines divide France into an interior region
intensity is indicated by the size of the plotted circle. Light lines and the Atlantic and Mediterranean coastal regions. Topography is
outline the 500-m elevation contour and so show the Pyrenees as in Figure 1.
Mountains in the south, the Alpes Mountains to the east, and the
Massif Central in the southeast.

L = 0.25Wø'57,

the summer events, which are concentrated in the afternoon. with L in kilometers and W in meters. The mean path length
For the 24 tornadoes in the November through March for significantevents is 8.5 km, while their mean path width
period, only nine (37%) occurred between solar noon and is 300 m (see Dessens and Snow [1989, Table 3] for more
sunset. Of the 79 tornadoes in the April through October detail). (For a few "weak" (F-0 and F-l) tornadoes, path
period, 65 (82%) occurred between solar noon and sunset. length (14) and path width (12) data are also available. If
While the size of the winter sampleis small, these percent-
ages suggest that winter events are less dependent on
destabilizationdue to solar heating. Related findingsin the
United Statesare the bimodaldaily distributionin tornado
occurrencesin the southeasternstates with a secondary
peak near sunrise[see Kelly et al., 1978, Figure 12] and
findings
by Moller [1979]that "cool" season
eventsin the
southern
GreatPlains
aremorediurnally
scattered
thanlate
springevents.The datafor Franceshownin Figure3 support
Moller's conclusion that cool/winter season events are
largely dependenton dynamicforcing.

2.3. Path Dimensions

Figure 4 is a scatter plot of path length L and maximum 0 2 4 6 8 10 12 14 16 18 20 22 24

path width W for the 78 significanttornadoesfor which both HOUROF THE DAY (UTC)

L and W were available. The regression between L and W

Fig. 3. Hourly and monthly distributionsof tornadoes.Numbers
(with a correlation coefficient of 0.53) is given by the refer to F scale intensities. The dashed curves give the times of
following equation: sunrise (SR), solar noon (SN), and sunset (SS) at Paris (48ø52'N,
2ø20'E). The histogramsbordering the ordinate and the abscissa
log•0L = -0.60 + 0.57 logl0TM, show frequencyof occurrencesby month and time of day, respec-
tively. (This figurecorrectsa smallerror in the correspondingfigure
or of Dessens and Snow
 DESSENS AND SNOW 429

10.000
path statisticswould probably shift toward those for Amer-
ican events: the limited sampleof data for F-0 and F-1 events
supports this. All French tornadoes (winter and summer)
have mean path sizes very similar to the tornadoes of
1,000
3
3 3
43 Arkansas-Tennessee[see Howe, 1974, Figure 2].

3. RISK PROBABILITY

Path area A was computed for the 78 significantevents for

which both L and W were available. The mean value of this
100
area,4.0 km2, allowsan estimation
of themeanriskproba-
bility of tornado occurrence for France. During the well-
2 '•2• surveyedmodern period (1960 through 1988), there were 58
tornadoes of intensity F-2 or greater (about two per year).
Therefore on a national basis the mean area exposed to
lO
O.1 1 lO lOO significant
tornadoes
isabout8 km2peryear.Fromtheratio
PATH LENGTH(KM) of thisareawiththatfor France{549,000km2),theproba-
bility of a given "point" being struck by a significanttornado
Fig. 4. A plot of the maximum reported path width versus the isabout1.5 x 10-5 peryear,andthereturnperiodis about
reported path length for each significanttornado for which both 70,000 years. (Inclusion of the 10 weak events for which both
parameters are known. Numbers refer to F scale intensities. The
path length and width data are available modifies the mean
solid line is the regressionline (see text).
pathareavalueto 3.58km2butdoesnotsignificantly
change
the overall conclusion.)
When compared with the United States, the mean annual
these data are merged with thosefor significantevents, the frequency of tornadoes per area of 1ø squares [Thom, 1963,
mean path length becomes 7.8 km (based on 100events) and Figure 4], together with the mean tornado path area by
the mean path width becomes 268 m (94 events).) region
[Howe,1974,Figure1]orper10,000mi2(25,900km2)
The mean path length and path width are different for [Reed, 1971, Figure 6], shows that the risk is about 1t30times
winter and summer tornadoes: smaller in France than in Oklahoma, or about the same as in
the northeast of the United States.
Winter tornadoes
The geographicaldistributionsof Figure 1 showsthat the
level of risk is not the same in all regions of France. Along
L = 5.4 km, W = 190 m;
the Atlantic the tornado frequency is relatively high, but the
Summer tornadoes meanpath area struck by a tornado in this regionis low. For
example,a computationmade for a region near Nantes gives
L = 9.5 km, W = 330 m. a meanpathareaof 0.6 km2pertornado.Theriskprobabil-
Another stratification can be made between "coastal" ity for thisareais 4 x 10-6 peryear.
tornadoes,ones occurringless than 50 km inland from the
Atlantic or from the Mediterranean, and those occurring in 4. PARTICULAR FEATURES RELATED
the "interior" of France: TO GROUND CONFIGURATION

Coastal tornadoes In France, topography and surface characteristicschange

markedly from one region to another, often over short
17= 5.5 km, 1• = 154 m; distances. This allows computation of mean values of F
scale,path length, and maximum path width for significant
Interior tornadoes
events by landscape type. A very marked result appears
/7,= 10.3 km, 1• = 350 m. concerningmean path dimensionsof tornadoesover wooded
areas as contrasted to those over more open areas:
Using data for all tornadoes(weak, strong,and violent),
"Forest" tornadoes
Thom [1963],Reed [1971],and Howe [1974]havecomputed
tornadopath sizesfor U.S. tornadoes,findinga meanpath L= 17.1 km, W=850m;
width of around 150 m. An analysisby Fujita and Pearson
that Other tornadoes
[1973,Figure 2] of path statisticsby intensitysuggests
U.S. significant(strongand violent)tornadoeshavea mean L=6.8 km, W= 180m.
path width of around 300 m.
Frenchtornadoes We suggestthat this may reflect an effect of surface
It appearsthat in the mean,significant
havepathwidthssimilarto theirAmericancounterparts.If roughnesson tornadoes. Several laboratory simulations
dataon all tornadoesoccurringin Francewere available,the [e.g., Wilkinset al., 1975; Desserts,1972;Diamond,
430 TORNADOES IN FRANCE AND THE UNITED STATES

UAI ==]
VALENC
IENNES

F-'lARRAS 20.15

/•P 11
uel
./•XXSaudem•at
kkRiencourt
l'sC
1q.•5
• Err i 11 ers
'Ecoust-S•-M. CAMBRAI

'•PERONNE

Fig. 5. Paths of the June 27, 1967, tornadoesacrossthe departmentsof Nord and Pas-de-Calais.Locations of villages
where significant structural damage occurred and times of touchdown and termination are given. The dashed lines
outline densely forested areas. The stippledrectangledenotesthe broad path of the tornado through partly forested
regions.

Diamond and Wilkins, 1984] and numerical models [e.g., correlateswith the nature of the ground, which is completely
Bode eta!., 1975]suggestthat an increaseof core radius with bare for the western path but partly wooded for the eastern
increasingsurface roughnessis to be expected. The increase one.

of mean path length for tornadoes over a forest may be a

consequenceof the increaseof the mean path width, as there 4.2. September 20, 1973, Sancy-les-Provins
is a moderately strong direct relationship between the two (Seine-et-Marne)
dimensions (see section 2).
Another particular feature related to surface characteris- This tornado system occurred about 80 km east of Paris. It
tics is that severalreportssupportthe conceptof occurrence initially was a large (=2000 m diameter) F-2 tornado, moving
of suctionvortices when a tornado goes from a rough area through a forested area from Moret-sur-Loing to Rampilion
(forest) to a smooth one (e.g., bare ground). A few well- (Figure 6). After coming out of the forest, it continued for a
documented cases illustrating these observations will be few kilometers to near Vanville and Maison-Rouge as two
briefly related. vortices moving along parallel tracks separated by about 1
km. As it continued on across the plain for 10 km toward
Neuvy, it took the form of an F-3 tornado, leaving a
4.1. June 27, 1967, PaHuel (Pas-de-Calais)
200-m-wide damage path across bare ground.
and Pommereuil (Nord)
Two companionviolent tornadoeskilled eightpeople and
4.3. A Tornado Alley in the Jura Mountains?
injured 80 in the north of France on June 27, 1967. The
nearlyparalleltornadopathswere separatedby about35 km A local "tornado alley" [Ga!limore and Lettau, 1970]
(Figure 5). The east tornado touched down while the west appears to exist north of Geneva, along the France-
tornado was in action. The eastern tornado was rated F-4, Switzerland border. In this region the southwest-northeast
and the western tornado probably reachedF-5 intensity. foldings of the Jura mountains provide a mesoscalepattern
Seventeenhouseswere razed to the ground, cars were lifted of significantrelief, in some places with as much as 500-m
up and hurled in fields or over houses,grassblades and tree differencesin elevation between adjacent crests and valleys.
brancheswere stuck into tree trunks, large piecesof wood On August 19, 1890, an F-4 tornado touched down be-
and tree trunks were stuck into the ground, and tombs were tween Oyonnax and Saint-Claude (Figure 7) and then re-
opened at Riencourt, and so on by the F-5 event. mainedon the groundfor 1 hour to Croy (Switzerland). This
The western tornado path was only 250 m wide, while the tornado was well documented by Bourgeat [1890] for the
eastern tornado was nearly 10 times larger. This difference French portion of its track and by Gauthier [1890] for
 DESSENS AND SNOW 431

K1
ROMILLY-s-SFINE

I 18 km I

Fig. 6. Damagetracksof the tornadoesof September20, 1973,acrossthe departmentof Seine-et-Marne.Locations

of villageswhere significantstructuraldamageoccurredand reportedtimesof tornadopassageat a few locationsare
given. The dashedlinesagainoutlinedenselyforestedareas.Here the stippledrectangleshowsthe broad path of the
tornado through a wholly forested region.

Swiss portion. Reexamination of their descriptionsis of strong gales or collateral and secondary tornadoes, only on
interest now that we understand better the physicsof torna- the right-hand side. The longest of these ramifications (Bois
does. For example, Gauthier [1890, p. 419] reports, "The d'Amont-Carroz) was 3 km long and went I km eastwards."
ground configuration transformed the main vortex into On August 26, 1971, an F-4 ton.ado retraced almost

L ' •b•na2e

• I0 km t

Fig. 7. Mapping
of selected
damage
tracksinthe'•tornado
alley"of theJura
432 TORNADOES IN FRANCE AND THE UNITED STATES

exactly the Swissportion of this 1890tornado[Piaget, 1976]. 18 •]ULY1983 12 km

The replicationwas so perfect that a reorganizationof the

1971 tornado into several vortices was observed along the 0

samepart of the path where this had beenobservedin the

1890event, abovethe flat, bare groundof the valley of Joux,
between the wooded areas of Bois d'Amont and Le Brassus
[see Piaget, 1976, Figure 27].
The tornadoes of 1890 and 1971 both traveled at 60 km
hr-• , indicative
thewindvelocity
of strongmidandupperlevelwinds.In fact,
was30m s-• at 300hPaand55ms-l at /
the tropopausein the jet stream over the tornadoarea on /
August 16, 1971 [Piaget, 1976].
Three other violent storms,in the years 1624, 1768, and
1842,have devastatedthe samevalley. All are suspectedto
have been tornadoes.The event in the year 1624 is suffi- WIelDSPEED(m.sec'• )
cientlywell documented to allowa detailedmappingof the 1•0 20
stormpath (Figure7). In !967 an F-3 tornadostruckthe
villageof Lajoux,6 km to the rightof this"tornadoalley"
17 AUGUST 1986
(alsoshownin Figure7). On severalotheroccasions,narrow 10 12 km

swathshave been reportedin the forestsof the same area; 8 •-•- *'
they are alsothoughtto be due to tornadoes.Theseobser- 7

vationssuggestthat low-levelwind flow modifiedthrough 6

channeling by the mountainsmightprovidea locallyfavor- 3

able wind shear for tornadogenesis[Nuss, 1986].

5. METEOROLOGICALCONDITIONS

Synopticsituationson daysof significanttornadicevents

have been classifiedaccordingto surfaceand 500-hPaflow
0•• SiORM
MOTION
regimes.Generally,thesedays may be identifiedas falling
into either a summer or a winter pattern. WINDSPEED(m.sec-])

For tornado-producing configurations

in summer(roughly Fig. 8. Hodographsfor (top) July 18, 1983,and (bottom)August
June-August),the 500-hPa-heightfield typically showsa 17, 1986.Bothtakenat Bordeauxin southwestFranceat 1200UTC
closedlow either over the Bay of Biscay (to the west and onthedaysindicated.For theJuly18, 1983,eventthislocationis 70
km to the southeast of and 6 hours before the tornado. For the
south, betweenFrance and Spain) or farther west in the
August17, 1986,eventthislocationis 350km to the southwest
of
Atlantic and high pressurecenteredover the Sahara. As a and 6 hoursbefore the tornado;it is almost exactly upwind of the
consequenceof the circulationaround these two centers, touchdown point. Storm motionis shownfor August17, 1986;
thereis midlevelflowfromthe southwest of about20 m s-• unfortunately,
thesedata are not availablefor July 18, 1983.
over France. In the surfacepressurefield, there is a weak
low over France. A weak, slow moving frontal systemis
sometimespresentjust west of the tornadoarea. Surface the west over France. This is associatedwith a strongtrough
winds are usuallyweak. These conditionsare illustratedin containing a deepprimarylowto thenorthwest of theBritish
the hodographsin Figure 8. Isles.In the surfacepressurefield, thereis a deeplow in the
Back trajectoriesover isentropicsurfacescomputedfor samearea,with a well-definedcoldfrontwestof the tornado
two casesillustratethe summertornado-producingconfigu- region.Strongwesterlywinds extenddown to near the
ration(Figure9). Thesetrajectoriesshowthat the low-level surface,so that there is little directionalshearin the lower
inflow was Mediterranean air moving very slowly into cen- troposphere.
tral France from the south. Trajectories at high levels show
a much faster flow over the Iberian Peninsula coming from
6. SUMMARY AND CONCLUSIONS
the mid-Atlantic. This turns to the northeast to flow over
France. This situationis roughly similar to that observedin An examination of historical and modern accounts span-
the Great Plains of the United States (see, for example, the ning more than 300 years has shownthat a few square
tornado-producingsynoptic pattern shown by Pearson kilometers of France are struck by tornadoes each year.
[1976, Figure 8]). While the effects of these occurrences have not usually been
In winter (roughlythe periodOctober-March)the 500-hPa sufficientto cause a national calamity, lossesin life and
patternfor a tornado-producing situationtypicallyshowsa propertyandthe numberof injurieshaveoccasionally been
strong
midlevel of20-40m s-•) from substantial.
flow(withwindspeeds This explainsthe interestof several
 DESSENS AND SNOW 433

!,•00 UTC 18 JULY

!983 300h'•a
40N

3O N

20 E 10E 0 20E 10E 0

1800 UTC 17 AUGUST1986 900 hPa 1800 UTC i7 'AUG'USl:'

1986 "'200'
hPa•
4O N
,/

30 N

20 E 10 E 0 20 E 10E 0

Fig. 9. (Top) Air parceltrajectoryarrivingat the (left) 900-hPaand (fight) 300-hPalevels over Bordeauxduringthe
F-3 tornadooutbreakof July 18, 1983,and(bottom)air parceltrajectoryarrivingat the (left) 900-hPaand (fight} 200-hPa
levels over La Charit6 during the F-3 tornadooutbreak of August 17, 1986. Spacingbetween two crossesalong a
trajectory represents the distance traveled in 6 hours.

agencies, such as D•l•gation aux Risques Majeurs and States analog to French conditions is found in the New
Electricit• de France, and of many of the French people. England area, which does indeed have recurrence rates and
To summarize, France is somewhat like a miniature of the intensity statisticssimilar to those observed in France.
United States. This is particularlytrue when consideringthe
synoptic conditions supporting tornadogenesis:in France,
Acknowh,dgments. The authors wish to acknowledge the efforts
tornadoes result from the interaction of slow moving, low-
of an anonymousreviewer and of Alan Moller. Moller in particular
level Mediterranean air with an upper level cold Atlantic made numerous suggestionsthat led to several significantimprove-
flow, whereas in the United States, the interaction occurs ments in the content of the manuscript. His observation of the
between upper level are massesof Pacific origin and low- greaterscatterto be found in the winter events displayedin Figure
3 was especially important.
level northward incursionsof tropical air from the Gulf of
Mexico. This situationprobablyexplainswhy Franceis one
REFERENCES
of the very few European countrieswhere "strong" and
"violent" tornadoes are observed from time to time. Abbey, R. F., Jr., Risk probabilities associatedwith tornado wind-
Why are tornadoessome 15 times lessfrequentin France speeds,paper presented at Symposium on Tornadoes, Tex. Tech.
Univ., Lubbock, Tex., June 22-24, 1976.
than in the United States?A possibleexplanationis that
Bode, L., L. M. Leslie, and R. K. Smith, A numerical study of
there is a requisite"gestationtime" for tornadogeneration boundary effects on concentrated vortices with application to
in a thunderstorm. In the central United States the relatively tornadoesand waterspouts, Q. J. R. Meteorol. Soc., 101, 313-
uniformextensivetopographyof the Great Plainsfrequently 324, 1975.
allowsquasi-steadyatmosphericconditions.Strongthunder- Bourgeat, M., Premieres observations sur le cyclone du 19 arut
dans le Jura, C. R. Hebd. Seances Acad. Sci., 11l, 385-389, 1890.
stormscan often maintain their intensityfor periodsof time
Dessens,J., Influence of ground roughnesson tornadoes:A labo-
sufficient for tornadoes to form. In France, with diverse ratory simulation, J. Appl. Meteorol., I1, 72-75, 1972.
topographyand mesoclimatesin a relatively smallregion, Dessens, J., and J. T. Snow, Tornadoes in France, Weather
suchconditionsoccuronly rarely. Probablythe bestUnited Forecasting, 4, 110-132,
434 TORNADOES iN FRANCE AND THE UNITED STATES

Diamond, C. J., Laboratory simulation of tornado-like vortices Kelly, D. L., J. T. Schaefer, R. P. McNulty, C. A. Doswell III, and
under the effects of translation, M.S. thesis, 75 pp., Univ. of R. F. Abbey, An augmentedtornado climatology, Mort. Weather
Okla., Norman, 1982. Rev., 106, 1172-1183, 1978.
Diamond, C. J., and E. M. Wilkins, Translation effects on simulated Moller, A., Climatology and synoptic meteorology of southern
tornadoes, J. Atenos. Sci., 41, 2574-2580, 1984. plains tornado outbreaks, Masters thesis, 70 pp., Univ. of Okla.,
Fujita, T. T., Workbook of tornadoesand highwindsfor engineering Norman, 1979.
applications,Res. Pap. 165, 142 pp., Satell. and Mesometeorol. Nuss, W. A., Observations of a mountain tornado, Mort. Weather
Res. Proj., Univ. of Chicago, Chicago, Ill., 1978. Rev., 114, 233-237, 1986.
Fujita, T. T., and A. Pearson, Results of FPP classificationof 1971 Pearson,A., Tornado prediction,paper presentedat Symposiumon
and 1972 tornadoes, in Preprints, Eighth Conferenceon Severe Tornadoes, Tex. Tech. Univ., Lubbock, Tex., June 22-24, 1976.
Local Storms, pp. 142-145, American Meteorological Society,
Piaget, A., L'6volution orageuseau nord des Alpes et la tornade du
Boston, Mass., 1973.
Jura Vaudois du 26 a6ut 1971, Publ. 35, SZ ISSN 0080-7346, 102
Gallimore, R. G., and H. H. Lettau, Topographic influence on
tornado tracks and frequenciesin Wisconsinand Arkansas, Wis. pp., L'Inst. Suissede M6t6orol., Payerne, Switzerland, 1976.
Acad. Sci. Arts Lett., 58, 101-127, 1970. Reed, J. W., Someaveragedmeasuresof tornado intensity basedon
Gauthier, L., La trombe-cyclone du 19 a6ut 1890, C. R. Hebd. fatality and damagereports, in Preprints, SeventhConferenceon
Seances Acad. Sci., 111,417-421, 1890. Severe Local Storms, pp. 187-193, American Meteorological
Hales, J., Improvingthe watch warning programthroughthe useof Society, Boston, Mass., 1971.
significantevent data, in Preprints, 15th Conference on Severe Thom, H. C. S., Tornado probabilities, Mon. Weather Rev., 9I,
Local Storms, pp. 165-168, American Meteorological Society, 730-736, 1963.
Boston, Mass., 1988. Wil'kins, E. M., Y. Sasaki, and H. L. Johnson, Surface friction
Howe, G. M., Tornado path sizes, J. App!. Meteorol., 13,343-347, effects on thermal convection in a rotating fluid; a laboratory
1974. simulation, Mort. Weather Rev., 103, 305-317,
 Tornadoes of China

XU ZlxIU, WANG PENGYUN, AND L!N XUEFANG

Academyof Meteorological
Science,StateMeteorological
Administration,Beijing, China

1. INTRODUCTION elevations of 1000-2000 m). Near the eastern coasts, there

are several major plain regions, including (from north to
While the most important severe convective weather south) the Northeast China Plain, the North China Plain, and
phenomenonin China is hail, the occasionaltornadoesthat the plains of the middle and lower reaches of the Yangtze
do occur can inflict considerabledamage.For example,on River. All of these plains are typified by msl elevations of
March 26, 1967, a family of 13 tornadoes occurred near less than 50 m.
Shanghaiand North Zhejiang province. Along the path of During the 1l-year period of record, from 1971 to 1981, it
this tornado outbreak, more than 10,000 homes were dam-
was noted that the region receiving hail most frequently was
aged. Several steel high-voltagepower transmissiontowers, not coincident with the tornado frequency maximum (see
including some built to withstand Beaufort force 12 winds, Figure 1). Hail is more common in the western part of China,
were blown down or severely damaged.
with its high plateaus, than in the eastern plains (see Figure
Tornadoes have been recorded in China as far back as
2). We note that in the United States, lane-hail (->2 cm)
1000 years, but the total number of such reported events is
occurrencesare most frequent in about the same area as the
rather small, indicating that many events have gone unre-
tornado frequency maxima [e.g., Kessler and White, 1981;
ported. In 1949 many new weather stationswere established Kelly et al., 1985]. However, considerablehail that is below
all over China, and tornado reporting in considerablygreater the arbitrary sizethresholdfor "severe" hail occursover the
detail hasresulted. In this paper we shall presentinformation highplainsof the United States,accountingfor the resultsof
basedon the recordsfrom 26 provincesand municipalitiesof Changnonet al. [1977], which show peak hail fall frequency
China (Xinjing, Tibet, Guangdong, and Taiwan provinces over the high plains of Colorado, Wyoming, and Nebraska.
are not included owing to lack of data) for the period Tornadoes in China seem to occur most frequently over
1971-1981. This data set provides an introduction to the the plainsof the middle and lower Yangtze River (Figure 3),
climatologicaltornado distributionover China. We alsoshall especiallyin Jiangsu,Anhui, and Hunan provinces, with
present some brief case examples to illustrate the general Jiangsuprovincehavingthe highestfrequencyat 30 tornado
synoptic and mesoscale regimes under which tornadoes days during the l 1-year period. This averages out to 2.7
occur in China. Finally, some radar and visual characteris- tornadodays per year. In Jiangsuprovince, there was a total
tics of tornadoes in China will be presented. of 217 hail days during the same period, so hailstorm days
were about 7 times more frequent than tornado days.
2. GEOGRAPHICAL DISTRIBUTION AND CLIMATIC For the same period in Anhui province, there were 12
CHARACTERISTICS tornadodays and 218 hailstorm days, making hail daysabout
18 timesmore frequent. In Hunan province the ratio reached
China is characterizedby diverseand complexterrain. In about 30. Elsewhere in China the ratio is even larger. Some
broad terms, high plateaus are characteristicof the west, of the reasonsfor this might include low populationdensities
whereas low plains are typical in the east. The western in some of the provinces.
plateausinclude the Qinghai-Xizhangplateau(with typical Generallyspeaking,the region of maximumtornadofre-
mean sea level (msl) elevations of over 4000 m, and its quencyin China is associatedwith the commonpresenceof
Mount Qomulangmais knownas the "roof of the world"), frontalboundariesandextratropicalstormsduringthe spring
the Inner Mongoliaplateau,the Yellow Earth (or "1oess") and summer. Moreover, the coastalplains have bountiful
plateau, and the Yunnan-Guizhouplateau (with typical msl low-level moisture and a lack of terrain features to disrupt
the flow of moisture and instability.
The Tornado: Its Structure,Dynamics, Prediction,and Hazards.
GeophysicalMonograph79 As can be seen in Table 1, tornadoes occur from April
Copyright1993by the AmericanGeophysicalUnion. throughAugust, with a peak frequencyin July. Tornadoes
436 TORNADOES OF CHINA

\ 1
! I

Fig. 1. Annual average number of hailstorm days in the period 1951-1980.

I
•-. I g yg go -""
. " •' '""•"• (139
) •'..."- 'ø

-. ". ' •' i 0. 2 . '•Hebei

• 8
• •-. .... •, •.•. • . •- • •(680)•
( (1196) • t ' / • , "

• •bet A•nomus Region ' •', •" '•"•.. • ' 8-"k• _•?•.o• ..... • Sh•ghai

•• • ,_•% . •'%•; •4 )½56) ..>t

•• t • (•;;•(sos) •, H?? ; ,•1•)

' • • •uang• ng
YunnanGu•gx•/ • /

Fig. 2. Totalnumberof tornado

daysandhailstorm
days(inparentheses)
in eachprovince
in theperiod
 XU ET AL. 437

Fig. 4. Tornado tracks (small arrows) on August 26, 1987, in the

North China Plain region.

country, and the typical synoptic situationsdiffer from place

to place. Some basic patterns are as follows.

3.1. Middle and Lower Reaches

Fig. 3. Geographical
distributionof tornadoesduringthe period of the Yangtze Valley
1971-1981.
3.1.1. Southerlyflow around a subtropicalhigh. When
a low-pressuretrough is present with southerly flow around
are virtually nonexistent during the winter months. Similar the western side of a subtropicalhigh-pressurecenter, there
to tornadoes in the United States [Kelly et al., 1978] and often is a mesoscalesystem (either a closed low-pressure
elsewhere, most tornadoes in China occur between 1200 and system or a convergence boundary) at the surface, with
1800 LT. abundant moisture. Tornadic storms develop in association
with such systems.
3. METEOROLOGICAL CONDITIONS WITH 3.1.2. Stationary fronts. When a frontal boundary be-
TORNADO OUTBI•EAKS comes stationary over the plains of the middle and lower
Yangtze River, it may be associated with strong southerly
By analyzing the synoptic charts for 94 tornado days in flow (jetlike streams)at 850 and 700 hPa. Cyclonic waves can
China, we determined the following factors were commonto develop on the quasi-stationary front, leading to the devel-
the events: (1) potential instability, (2) moistureflux conver- opment of tornadic storms.
gence at low levels, (3) a localized lifting mechanism(front, 3.1.3. Typhoons. When a weakening typhoon moves
terrain, etc.), (4) upper and lower leveljet streams,(5) strong from Fujian province northward, the northern part of the
vertical wind shear, and (6) potentially cold air in midlevels. circulation moves over the plains of the middle and lower
These conditionsare quite similarto thosefoundfor torna- Yangtze River. This puts a low-level cyclonic shearline over
does elsewhere in the world. In China, these factors are the area, having northeasterly flow converging with south-
brought together in different ways in different parts of the easterly winds, in the presence of considerable low-level
moisture. When cold dry air at 700 hPa is coming from
northern China at the same time, the stratification can
TABLE 1. Monthly Distributionof Tornado Days in China become quite potentially unstable. In such cases, tornadoes
During the Period 1971-1981
of strong intensity can develop.
Month Tornado Days In the southernpart of sucha typhoon a convergenceline
having northwesterly and southwesterly winds can be
January 0
February 1 present. If tornadoesoccur with this feature, they tend to be
March 3 weak. Tornadoes are quite uncommonin other quadrantsof
April 14 typhoons.
May 14 3.1.4. Extratropical cyclones. Some tornadoes occur
June 16
within the warm sector of extratropical cyclones. In such
July 27
situations it is common for the warm sector to have above-
August !4
September 3 normal moisture content, with the moisture axis at 850 hPa
October 2 ahead of the cold front, as in the work of Miller [1972]. In
November 0
fact, these situationslook very much like those described in
December 0
"classical" tornado outbreaks in the United States,
438 TORNADOES OF CHINA

1•' 45 -49

12'53 12'57

Fig. 5. A seriesof photographsof the tornado near Laishui, Hebei province, on May 15, 1981, with times indicated
in local standard
----------

XU ET AL. 439

Fig. 5. (continued)
440 TORNADOES OF CHINA

vortex, it may move across the North China Plain. This

troughoften is associatedwith severethunderstorms,espe-
cially when the trough interactswith a preexisting, low-level
mesoscale boundary [Xu et al., 1977].

4. VISUAL ASPECTS OF TORNADOES

Tornadoes in China present a visual appearance that is

quite comparableto that of their counterpartsin the United
Statesand elsewherearound the world. Figure 5 illustratesa
tornado that developed near Laishui in Hebei province on
May 15, 1981 [Chang, 1981]. A large thunderstorm was ob-
servedto be movingtowardthe northwesternpart of Laishuiat
1220(all times local standard time). At 1245a funnel cloud was
observed.This funnellengthenedand then dissipated,but by
1249, dust and debris were observed below the funnel cloud.
Four minuteslater the funnelbecamemoreclearly visible,and
the damage track width increased as the tornado intensified.
From 1253to 1257the tornadoattained its greatestintensity,
with a damagetrack 50 m wide. By 1301the tornadobeganto
weaken,andthefunneldiametershrankinto a ropelikeappear-
ance,with the tornadodissipatingby 1310.Total track length
was about 5 kin. This visual evolution is quite similar to that
describedby Golden and Purcell [1978], althoughthe tornado
showndid not becomeas largeduringits maturestageas some
tornadoes do in the United States.

Fig. 6. Waterspoutover the Yangtze River, photographedin 1933

near Shanghai.

dry, potentiallycold air in midlevels,jet streamsat both low

and high levels, etc.
3.1.5. Prefi'ontalsquall lines. When strongcold fronts
come from northern China into the Yangtze River plains,
tornadoesmay occur with squalllines. Tornadoesin suchcases
arise mostfrequentlywhen a preexistingmesoscaleconver-
genceboundaryis foundaheadof the approachingfront.

3.2. North China Plain

The frequency of tornadoes in the North China Plain is

much lower than in the middle and lower reaches of the
YangtzeRiver, but whentornadoesdo occur,they cancause
serious damage. As before, the conditions for tornadoes
arise in several different synoptic situations.
3.2.1. Extratropical cyclones. On August 26, 1987,
nine tornadoesoccurredin Hebei and Shandongprovinces
of the North China Plain (Figure4). They developedin the
warm sector of an extratropical cyclone associatedwith a
strong closed low at 700 and 500 hPa [Zhao, 1990]. The
situationwas characterizedby strong moistureflux conver-
gence at low levels, substantial horizontal and vertical wind
shears, and considerable instability.
3.2.2. Mongolian cold vortices. When a cold vortex
aloft is present over Mongolia, daily thunderstormsoften
developin the North China Plain. When a troughis foundin
the westerlyjet stream along the southernpart of the cold Fig. 7. Funnel cloud near Haikou, Guangdong
 X U ET AL. 441

Fig. 8. Funnel cloudsoffshorenear Qingdao, Shandongprovince.

Waterspouts also are observed in China. A waterspout is the two funnelspersistedfor about 10 min. The lell funnel
essentiallya type of tornado, only over water. Shownin Figure touched down briefly as a waterspout.
6 is a waterspout on the Yangtze River not far from Shanghai
in 1933.Noteworthy is the heightto which the water spraywas 5. RADAR ASPECTS OF TORNADOES
observed;one explanationfor thiscanbe foundin the work of
Nalivikin [1982], although there is no way to verify this Radar observations in China indicate that most tornadoes
explanationfrom the informationavailable. develop at the right rear portion of the radar-observed
Of course, funnel clouds are observed in China. In Figure thunderstorm echoes. as in the United States. Those radar
7 an example from Haikou, Guangdongprovince,is shown, echoesmay take on the "distinctixe" forms [Forbes. 1981]
while twin funnels over the East China Sea (,nearQingdao, of hooks, pendants, and S shapes.
Shandongprovince) are shown in Figure 8. In the latter As an example,on June8, 1977.at 1520(all timesare local
examplethe funnelsare comingfrom a cumulonimbus
cloud; standardtime), sexere thunderstorms were observed by the
Yongdengcountyradar in Gan.•upro,,ince [Zhengand Liu,
1982].The positionof the radar echoand tornadoat various
times relative to the radar is shown in Figure 9. The storm
developedover relatively flat terrain in a valley; at 1530it
wasa part of an east-we.,,torientedsquallline about200 km
long, moving toward the southeast.
A hook-shapedecho with an apparent "eye" bas ob-
served at 1530 (all times are local standard) from the right
rearquadrantof the parentecho(Figures10a and 10b); such
a configuration
is obviouslya clearindicationof a aupercell
storm.The ese'sdiameter•as le•s than I km andv,'asin the
' '. center of the circular end of the hook. similar to Fufita's
[1965]observations. The eye was not visibleat 1532or 1535
withoutattenuation,but appearswith 22-dB attenuation.Al-
thoughits appearance was quite variablewith time, the eye
0 1Okra
wasmaintained until 1542.Figure10crevealsthepresence of a
bounded weakechoregion(or vault)thatextendsto a heightof
, severe center of PPI echo
i
about6 km. The tornadobeganat about 1530and dis.-,ipated
around1551,havingtraveledroughly10km. The tornadowas
Fig. 9. Sketchofterrain,showing thetrackofthemainradarecho onthegroundduringthe timewhentheradarwasre•etding,at
thatproduceda tornadoat theindicated localstandardtimesnear
Yongdeng,Gansuprovince,on June8, 1977. leastintermittently,the eyelike featureat low
442 TORNADOES OF CHINA

•0

15'30

15'32

Fig. 10a. Plan positionindicator(PPI) photographstaken from the 3-cm radar in Yongdengcounty on June 8, 1977.
The radar was operating at 1.6ø elevation, and range circles are at 10-km intervals.

6. DISCUSSION Wang [1989], showing that surface roughnesscan cause the

wind speed maximum in a circulation to be higher above
The geographicaldistribution of tornado reports in China, the surface than when the surface is smooth. In western
showing most events on the relatively flat plains may be China, where the surface roughness is large, the height
explained, at least in part, by the experiments of Monji and of the maximum wind is higher than over smooth
 XU ET AL. 443

15'35

,,• 10
i 15'42
I
•'1 •,

15-54

Fig. 10b. PPI photographsas in Figure 10a, except including22-dB

444 TORNADOES OF CHINA

On the other hand, in eastern China, the smooth terrain

encouragesthe development of strong low-level circulations
from which tornadoesmay develop. It also appears that the
tiltingterm in the vorticity equationmay be particularlyimpor-
tant for tornadogenesisin associationwith tropical cyclones
[Shen, 1990].
However, owing to the lack of detailed information for
tornadoesin China, such hypothesescan be offered only as
speculativeideas. Many more observationsand studies are
needed to confirm them.
l

l't•n Acknowledgments. The authors are grateful to Charles A.

Doswell III for his help in editing and revising this paper, as well as
to the anonymous reviewers for their valuable comments. Joan
Kimpel (Cooperative Institute for Mesoscale Meteorological Stud-
ies, Norman, Oklahoma) helped with some of the figures, for which
we are grateful.

REFERENCES

Chang, J. A casestudy of tornado in Laishui city, Hebei province (in

Chinese), Meteorol. Mottogr., 11, 32, 1981.
Changnon, S. A., Jr., et al., Hail suppression:Impacts and issues,
Final Rep. ERP75-00980, Ill. State Water Surv., Urbana, 1977.
Forbes, G., On the reliability of hook echoes as tornado indicators,
Mon. Weather Rev., 109, 1457-1466, 1981.
Fujita, T. T., Formation and steering mechanism of tornado cy-
clones and associated hook echoes, Mort. Weather Rev., 93,
67-78, 1965.
!sr Golden, J. H., and D. Purcell, Life cycle of the Union City,
Oklahoma tornado and comparison with waterspouts, Mon.
Weather Rev., 106, 3-11, 1978.
Kelly, D. L., J. T. Schaefer, R. P. McNulty, C. A. Doswell Ill, and
15 R. F. Abbey, Jr., An augmented tornado climatology, Mort.
Odb Weather Rev., 106, 1172-1183, 1978.
Kelly, D. L., J. T. Schaefer, and C. A. Doswell III, Climatology of
nontornadic severe thunderstorm events in the United States.
Mon. Weather Rev., 113, 1997-2014, 1985.
Kessler, E., and G. F. White, Thunderstorms in a social context, in
Thunderstorms: A Social, Scientific, and Technological Docu-
mentary, vol. l, The Thunderstorm in Human AjJ•irs, edited by
E. Kessler, pp. 1-22, U.S. Department of Commerce, Washing-
ton, D.C., 1981.
Miller, R. C., Notes on analysis and severe-storm forecasting
procedures of the Air Force global weather central (revised),
Tech. Rep. 200, 190 pp., Air Weather Serv., Scott Air Force Base,
I11., 1972.
Monji, N., and Y. K. Wang, A laboratory investigation of charac-
teristics of tornado-like vortices over various rough surfaces,
Acta Meteorol. Sinica, 3, 506-513, 1989.
Nalivikin, D. V., Hurricanes, Storms, and Tornadoes, pp. 282-284,
Amerind, New Delhi, India, 1982.
Shen, S., An analysis of general characteristicsand genesiscondi-
tions of tornado in front of typhoon (in Chinese), Meteorol.
Monogr., 16, 11-16, 1990.
Xu, Z., et al., The analyses of mesoscale weather processes in the
summer in the Beijing-Tianji-Tangshan areas (in Chinese), in
Collected Papers in Radar Meteorology, pp. 17-39, Institute of
Central Meteorological Bureau, 1977.
Fig. 10c. Range height indicator photographs for the events Zhao, Y., An analysisof environmental conditionsfor generation of
shownin Figures10a and 10b acrossthe hook-shaped echo, at (top) tornadoes over North China (in Chinese), Meteorol. Monogr., 16,
1533, with no attenuation and azimuth angle of 329ø, (middle) 1539, 36-38, 1990.
with no attentuation and azimuth of 332% and (bottom) 1539, with Zhen, C., and D. Liu, An analysis of radar echoes of tornado,
22-dB attenuation and azimuth of 332 ø. plateau (in Chinese), Meteorol. Mottogr., 9, 95-98,
 SeasonalTornado Climatologyfor the SoutheasternUnited States
LINDA PICKETT GARINGER AND KEVIN R. KNUPP

Atmospheric Science Program, University of Alabama. Huntsville, Alabama 35899

1. INTRODUCTION appreciable inhomogeneity in both monthly and hourly tor-

nado frequency. Anthony's analysis indicates that individual
Although the climate of the southeastern United States seasonsdisplay contrastingdiurnal patterns in tornado fre-
(referred to hereinafter as Southeast) is classified as humid quency. Thus potentially confusingresults may be obtained
(continental) subtropical, inspectionof average annual and when hourly tornado frequency is composited over the
seasonal precipitation totals [Shea, 1984], thunderstorm entire year. This study extends the work of Anthony by
frequency [Court and Griffiths, 1982], and the diurnal nature examining and quantifying the diurnal distribution of torna-
of precipitation and thunderstorms [Wallace, 1975] shows does by season, over four relatively homogeneoussubre-
appreciable spatial variability within the region. For exam- gions of the Southeast.
ple, Wallace [1975] found a strong diurnal tendency in
thunderstormfrequency during summer. Within the interior
2. DATA AND ANALYSES PROCEDURES
Southeasta prominent peak in thunderstormfrequency was
found near 1500 local solar time (LST), while along the Gulf This analysisutilizes the National Severe Storms Forecast
Coast and Florida Peninsula, thunderstorm frequency Center's (NSSFC) updated tornado data base covering the
peaked near 1300-1400 LST. Such summertime variability 39-year period 1950-1988. Data used include tornado loca-
can be related to mesoscale influencesproduced by land- tion information, time, and intensity estimates. The limita-
ocean contrasts (e.g., the sea breeze). tions of these data have been summarized by Kelly et al.
Patterns of severe weather, including tornadoes, also are [1978, 1985] and Schaefer et al. [1986]. Some of the more
inhomogeneousin frequency and diurnal distribution, over noteworthyproblemsinclude populationbias and geograph-
the United States in general and the Southeastin particular. ical (state) variations in severe weather recognition and
Kelly et al. [1978] conducteda comprehensiveclimatological reporting procedure. For large samples we assume that
survey for the period 1950-1976, in which the diurnalbehav- random time errors will cancel. In addition, analysis biases
ior of tornadoesfor the three distinct regions (Great Plains, may have been introduced by variations in reporting accu-
Midwest, and Southeast) was examined. (They defined the racy as a function of time of day (e.g., day versusnight), at
Southeastas consistingof the major portionsof Mississippi, least for weak tornadoes. The analysis in subsequentsec-
Alabama, and Georgia.) The diurnal behavior of tornadoes tions was conducted on five subregions shown in Figure 1.
over the Southeast was found to differ from that of the other The two Atlantic Coast subregions are combined in the
two regionsin havinga weak maximumin frequencyin the analysisof monthlyand hourly frequencyin sections3 and4.
early morning.All three regionsshoweda prominentmiddle Subregionboundary selection was basedon severalfactors,
to late afternoon maximum. House [1963] found relative including (1) the results of previous climatologies[e.g.,
patternssimilar to those of Kelly et al. [1978]for the period Anthony, 1988], (2) geographical breakdown, and (3)an
1916-1955. In an earlier study, Skaggs [1969] performed attemptto equalizethe area within each region. However,
harmonicanalysisof hourly tornadofrequency,which dis- sincethe areaswithin each subregionare not equal,the data
playedappreciableinhomogeneityin diurnalbehaviorover werenormalized to a standardunitareaof 104 square statute
the Southeast. miles(25,900km2).
In a preliminary tornado climatologywhich examined All times have been converted from the standard time
tornado and severe weather statistics for individual states (EST or CST) recordedin the data log to local mean solar
within the Southeast region, Anthony [1988] also found time (LST). All recorded times were discretized into 1-hour
intervalsbeginningon the hour. Thus a given point in the
The Tornado: Its Structure, Dynamics, Prediction, and Hazards.
GeophysicalMonograph79 time seriesrepresents30 min pastthe hour. In the seasonal
Copyright 1993by the American GeophysicalUnion. analyses,winter is defined as December, January, and
446 SEASONAL TORNADO CLIMATOLOGY

Lat
=36.5"mr
INTER/OR

Fig. 1. Definition of subregionalboundaries.

February; spring as March, April, and May; summer as

June, July, and August; and fall as September, October, and
November.
In section4, harmonic analyseshave been completedfor Fig. 2. F scale values and tracks (lines connectingbeginningand
hourly distributions (24 samples) of all tornadoes, subdi- ending points) of very strong to violent (F-3 and greater) tornadoes
reported during the 39-year period 1950-1988.
vided by subregion and by season. In each case the data
were first smoothedby an equal-weighted,three-point mov-
ing average to reduce high-frequency noise. A slow Fourier
over the Interior Southeast and Gulf Coast subregions. A
transform then was appliedto obtain frequency components.
relatively large number of long-track tornadoes occur over
The resulting equation,
Mississippi,northern Alabama, and eastern North Carolina,
as identified previously by Kelly et al. [1978].
Y= Z Apcos(C•p
+ Wpt) (1)
The average (normalized) number of all reported torna-
p=1,2,3,''', 10, does,andof F-3 andgreatertornadoes,
per 104 mi2 (25,900
km2)peryearforeachsubregion isdepicted
in Figure3. The
(where A is amplitude, p is the order of the harmonic, q•is Florida Peninsula has by far the greatest number of torna-
phase angle, w is frequency, and t is time) describesthe total does, with an averageof 12.1 per unit area per year, but very
distribution.The variance of the pth harmonicis given by few (1.1%) are F-3 or stronger. By contrast, 11.6% of the

2= Ap2/2
Fp . (2)
The variance fraction, which indicates the importance of a '--'-' 14 3.5 ,-',
single harmonic relative to the total, is defined as
-g"12 3.0 g-

VF = r rpi, (3)
% 2.0 %
where the summation is again over all 10 harmonics. 6 1.5

3. GENERAL PATTERNS

We first describesomeof the generalspatialand temporal

inhomogeneitiesof tornadofrequencythat are apparentover o o.o
GC FP IS AS AC u_
the Southeast.Tracks of very strong to violent (F-3, F-4, and Subregion
F-5) tornadoes during the 39-year period 1950-1988 are
Fig. 3. Average annual number of all tornadoes (open histogram)
shown in Figure 2. This provides a good indication of the
and very strongto violent tornadoes (F-3 and greater; solid histo-
inhomogeneous nature of tornado distribution over the gram)per10,000
mi2 (25,900km2),during20 years(1969-1988)
for
Southeastand shows a prevalence of very strongtornadoes each of the five subregionsdefined in Figure
 GARINGER AND K.NUPP 447

INTERIOR SOUTHEAST ATLANTIC COAST

50 5O

40 40

30 3O

E 20 20

10
o.

0
• 2 4 6 8 10 12 2 4 6 8 10 12
Month Month

GULF COAST FLORIDA PENINSULA

5O 50

4O 40

30 30

20 20

2 4 6 8 1'0 12 2 4 6 8 lo 12
Month Month

Fig. 4. Monthly distributionof area-normalizedtornadofrequencyin each subregionfor the period !950-1988. The
South and Central Atlantic Coast subregionsare combined. Open columnsrepresent all tornadoes, and solid columns
represent F-3 and greater.

tornadoes in the Interior Southeast subregionare classified activity somewhat later, from spring to early summer. The
as F-3 or greater. The Gulf Coast is secondin total number secondarypeak during November-December is less signifi-
of occurrences, with an average of 6.3 tornadoesper unit cant both here and in the Florida Peninsula. The Florida
area, 7.9% of which are classified as F-3 and greater. The Peninsula shows a maximum in July, although May, June,
Central and South Atlantic Coast subregionsdisplay similar and August also display significantly more tornadoes than
numbers of total tornadoes, 3.64 and 2.42 events respec- any individual monthly maximum in the other subregions.
tively, with F-3 and greater tornadoesaccountingfor 4.5% The monthly distribution of very strong to violent (F-3 and
and 5.7% of the total. Because the Central and South greater)tornadoes,also shown in Figure 4, indicatesthat the
Atlantic Coast areas are similar in number and relative Gulf Coast regionhas a significantnumber of F-3 + events in
distribution (monthly and hourly) of tornadic occurrences, December, in addition to the spring.
they have been combinedinto a singleAtlantic Coast sub- The generalpatterns described above require physicalex-
region in the following sections. planation.Assumingthat F-3 and greatertornadoesoriginate
Monthly distributionsof tornadoesdisplaydissimilarities from supercellstorms, -we suggestthat Florida experiences
for each subregionand are generallysimilarto those de- only a few such occurrencesduring the cold seasonwhen
scribed on a statewide basis by Anthony [1988]. Spring upper level winds are strong. We hypothesizethat further
producesthe greatestnumberof tornadoes in theGulf Coast north and west, over the Interior Southeast and Gulf Coast
and Interior Southeastsubregions,with significantsecond- subregions,environmentalconditionsfor long-livedsupercell
ary peaks during November-December(Figure 4). The storms,such as strong vertical shear of environmentalwind
November-December peak in the Gulf Coastand Interior [Weismanand Klemp, 1982] or high values of storm-relative
Southeastdistributions
is likely dueto southwardadvanceof helicity [Davies-Joneset al., 1990],are more favorable.
the jet streamover moist air massesproducedby the The preponderance of tornadoes during summer over
relatively warm adjacentGulf waters. Florida (and to some extent over the Atlantic Coast regions)
The Atlantic Coast subregionsdisplay a maximumof is likely associatedwith strongmesoscaleforcing
448 SEASONAL TORNADO CLIMATOLOGY

INTERIOR SOUTHEAST
WINTER SPRING
7 , ,

' 1st:7'9

.b 7-
6 2rid: 12
5

' //
_•, 4

O 3
o 2

• 1

rr 0

0 1'6 20 24 10 • '8 12 16 20 24
SUMMER FALL
7 ! , , , , • , ,, [ , ,

•.. 6
C 1st:
7'4 1st: 70
2rid: 22 2nd: 24
z
• 5
O
4

-1
o 0 4 8 12 16 20 24
TIME (LST) TIME (LST)

Fig. 5. Smoothedhourly distributionof all tornadoes(heavy curves)in the Interior Southeastsubregionfor each
season.Dashed curvesrepresentthe first harmonic, and the light solid curvesrepresentthe secondharmonic. The
variancefractionsfrom significantharmonicsare indicatedin the top right-handcornerof eachpanel. Vertical dotted
lines show the approximatetimes of local maxima in the smootheddata.

with sea breeze circulationsthat frequently interact with subregionsexhibit the oppositebehavior. As expected from
other boundariessuch as thunderstorm outflows [Purdom, the results of Skaggs [1969], Wallace [1975], and Anthony
1982]. Such tornadoesoriginatefrom nonsupercellstorms [1988], the diurnal cycle (first harmonic) is dominant (52% to
and are henceweak and short-lived[e.g., Holle and Maier, 95%) in all cases,but a large amount of variability exists on
1980]. It remains to be determined whether the mechanisms both a seasonaland an interregional basis.
of tornadogenesisresemble those within nonsupercell With one exception (the winter distribution over the
storms which generate tornadoes over eastern Colorado Florida Peninsula)all distributionspeak during the middle to
[e.g., Wakimoto and Wilson, 1989]. late afternoon hours. During spring and fall the southern
subregions generally peak at times similar to a summer
4. DIURNAL BEHAVIOR OF TORNADOES distribution, while maximum value times in the northern
subregionsmore closely resemble a winter distribution (Ta-
The seasonal(smoothed) hourly distributionsof tornado ble 1). The Gulf Coast exhibitssignificantsecondarymaxima
frequencyfor each subregionare shownin Figures5-8. In in the morning(0700 to 0900) during all seasonsand has the
addition, those harmonic components,generallythe first lowest interseasonal variability. The secondary morning
two, that have a variance fraction (VF) exceeding5% are maximum also was noted by Kelly et al. [1978] for a portion
superimposed.In many of the distributionsthe peak in one of the southeastUnited States consistingprimarily of Mis-
or more harmonics coincides very closely with the data sissippi,Alabama, and Georgia.This studyindicatesthat the
peak, while in others there is a significant offset due to geographicalarea primarily responsible for the secondary
skewed distributionsand/or secondarymaxima. The Interior morning maxima is enclosed by the Gulf Coast subregion.
Southeastregionis the only one in which the time seriesdata Some additional seasonal characteristics are described in the
peak before the peak in the first harmonic. The other three following
 GARINGER AND KNUPP 449

ATLANTICCOAST (SOUTHANDCENTRAL)
WINTER SPRING
, , , , , .... •. .., ..... • .c, . .•... •

/ :15!
1st' 66 b ; st: 80
ß 2nd: 33

0 4 8 12 16 20 24 0 4 8 12 1• 20 24
SUMMER FALL
, , ,

c I st: 74 1st: 66-

• 4 2nd: 30

<F--2
,._..
2rid:
22
1
.

_
ß .

TIME (LST) TIME (LST)

Fig. 6. As in Figure5, for the combined(SouthandCentral)AtlanticCoastsubregions.

4.1. Winter Patterns secondarymaximumat ---0530LST. In this case, however,

the first-harmonic contribution is stronger (79%), and that
The winter season shows the most variability in the time
from the second (and third) is much less. Although the
series distributionsamong subregions.The Florida Penin-
Atlantic Coastsubregions(Figure 6a) show relatively little
sula distribution(Figure 8a) displaysa very strong(95%) tornadicactivity during the winter, a secondarymorning
contribution from the first harmonic which reaches its max-
peak is suggested.
imum at -• 1100.This late morningpeak in the first harmonic
is unique to the winter Florida Peninsula,as noted by
Anthony[1988].Thisprominentmorningtornadomaximum 4.2. Spring Patterns
contrasts with the late afternoon (1700-1800 LST) cold-
seasonthunderstormfrequencymaximumin the first har- In relativeterms,springis the mostactiveseasonfor all of
monicdetermined explanation the studyareaexceptthe FloridaPeninsula.All subregions,
by Wallace[1975].A physical
for this time difference between thunderstormand tornado againwith theexceptionof the FloridaPeninsula,showdata
frequency is not obvious. maxima at 1600-1700LST. All subregionsdisplay skewed
The Gulf Coast distribution (Figure 7a) has a primary distributionsin which tornado frequency increasesslowly
maximumat -1700 LST and a significant secondary maxi- duringthe early morninghours •0600-0800)and then de-
mum at -•0900. The contributionfrom the first harmonicis creasesmore rapidly '•ter the afternoonpeak. An earlier
veryweak(52%)whencompared from maximum at -1430 is seen in the Florida Peninsula distri-
to thecontributions
the first harmonicin other seasonsand subregions.
This is bution(Figure 8b). The Florida distributionseemsto con-
the only casein whichthe contributionfromthe third tain characteristics of both the summer and the winter
harmonic (>5%). The peakin thethirdhar- distributions. The contribution from the first harmonic is
is significant
monic occurs at -0900 and coincides with the secondary relativelyhigh(80-90%)in all subregions
excepttheFlorida
morningmaximum
in thesmoothed
timeseries. Peninsula,whichdisplaysa moderatecontributiont¾omthe
The Interior Southeastdistribution(Figure5a) also dis- first harmonic(74%) and a moderatelystrongcontribution
a primarypeakat -•1700anda weak from the secondharmonic(22%). The Gulf Coast(Figure
playstwo maxima,
450 SEASONAL TORNADO CLIMATOLOGY

GULF COAST
WINTER SPRING

b ! // x 2nd:
15

m -1 I
4 8 12 16 20 24 0 4 8 12 16 20 24

SUMMER FALL

C 1st: 86
2rid: 12

4
3 2nd:
8t
2

-1 i i I i l
0 4 • '12 1•.....2'0 24
Time (LST) Yimo (LSI)

Fig. 7. As in Figure 5, for the Gulf Coastsubregion,exceptthat the third harmonicis shownas a dotted-dashedcurve
in Figure 7a.

again shows a prominent secondary maximum centered at as synopticscale perturbations migrate northward, the Flor-
-0800. ida time series does not exhibit relative peaks at the times
(--•0600 and ---1800 LST) found by Brier and Simpson.
4.3. Summer Patterns
4.4. Fall Patterns
The summer distributions are generally symmetric, with
the exception of the Interior Southeast, which is skewed into Fall generally is the least active time of year for the
the evening. The Florida Peninsula (Figure 8c) records a Interior Southeast, Atlantic Coast, and Florida Peninsula
maximum in activity during summer, when the Gulf Coast subregions(Figures 4, 5d, 6d, and 8d). As indicated in
(Figure 7c) subregiondisplaysa relative minimum (Figure section 3, the Interior Southeast and Gulf Coast subregions
4). In comparison to the other seasons, the secondary display a significant secondary peak in monthly tornado
morning maximum of the Gulf Coast subregionis relatively distribution from late fall to early winter (Figure 4). The
weak. The time series for the Florida Peninsula and Atlantic secondarymorning maximum in the Gulf Coast subregionis
Coast subregionsare similarly symmetricin shape,although again evident (Figure 7d). Contributions from the first
the Atlantic Coast exhibits lower frequency. The variance harmonic vary from a weak 66% in the Atlantic Coast
fractions and the harmonic phases are almost identical for subregion to a strong84% in the Gulf Coastsubregion,while
the first two harmonicsin these subregions.The increasing contributionsfrom the secondharmonic range from a strong
significance(from winter to summer)of the secondharmonic 30% in the Atlantic Coast subregionto a very weak 8% in the
in Florida is consistentwith the work of Brier and Simpson Gulf Coast subregion.
[1969], who demonstrated a close relationship between
cloud/rainfallpatterns and the semidiurnalpressureoscilla-
5. Su•^•¾
tion (which has a very strong second harmonic) within
tropical oceanic regimes. Although the relative influence of In general,the resultsemphasizethe variability in tornado
the semidiurnal tide would increase from winter to summer characteristicsand temporal patterns from both regional
 GARINGER AND KNUPP 451

FLORIDA PENINSULA
WINTER SPRING

20
ta .... i.... lS't:
•54 •'0
t"b..... "isi:
7242
16I ' 2nd:
1I 16:f 2nd:
• -4L , , , ,/ '
• 0' •t' "• 1• 1'6 •0 24 0 4 8 12 16 20 24
"7, SUMMER FALL
o 20 2O

• C ' ' ' 'A 'llt:321 1st: 76

2rid: 19

• 16
z
O•12 /•i•2nd:
24
o;
-4
0 0 4 8 12 16 20 24
TIME (LST) TIME (LST)

Fig. 8. As in Figure 5, for the Florida Peninsula subregion.

seasonalperspectives. Some of the specificresults of this and Gulf Coast) show significant secondary peaks in late fall
study are as follows: to early winter.
1. Tornadoes occur most frequently per unit area over 3. The hourly tornado frequency distributions show sev-
the Florida Peninsulasubregion,followedby the Gulf Coast, eral significantseasonaland interregional differences.Win-
Interior Southeast, South Atlantic, and Central Atlantic. ter, in particular, reveals that each subregion has unique
Very strong to violent tornadoes (F-3 and stronger)occur distributioncharacteristics.The Florida Peninsuladisplaysa
most frequently in the Gulf Coast and Interior Southeast, morning(1100 LST) maximum in frequency, which is repre-
followed by South Atlantic, North Atlantic, and finally the sentedalmost entirely by the first harmonic. In contrast, the
Florida Peninsula subregions. Gulf Coast shows a significant contribution from the second
2. The monthly distributionsfor the Florida Peninsula (26%) and third (19%) harmonicsduringthe winter. Only the
showpeak tornadofrequencyduringthe summer.The other Gulf Coast has a significant secondary peak during the
subregionsdisplay a maximum monthly frequencyin the morning hours (0800-0900 LST) in all seasons.
spring.Only two of the four subregions(Interior Southeast
The observed seasonal and regional differences likely
point to a multitude of environmental(mesoscale)controls
on storm structure which are not understood fully. Further
TABLE 1. ApproximateTime of MaximumNumberof physicalinterpretationof thesefindingsis clearlyneeded.A
Tornadoesby Seasonand Subregion compositeanalysisof proximity soundings,combinedwith
interior Atlantic Gulf Florida hourly wind profiler data or NEXRAD VAD winds, would
Southeast Coast Coast Peninsula provide additionalinsight on the physicalmechanismsre-
Winter 1645 1600 1630 1100
sponsiblefor the observed patterns. In addition, synoptic
Spring 1700 1615 1545 1430 scale compositingand analysis of selected case studies
Summer 1445 1530 1445 1530 would likely shed insight on the physicalprocessesrespon-
Fall 1615 !615 1430 1430
sible for the variation in observed hourly
452 SEASONAL TORNADO CLIMATOLOGY

Acknowledgments. This research was sponsoredby the Na- House, D.C., Forecasting tornadoes and severe thunderstorms.
tional Oceanicand AtmosphericAdministrationthroughthe South- Meteorol. Monogr., 5(27), 141-156, 1963.
east Regional Climate Center. Tornado frequency data were pro- Kelly, D. L., J. T. Schaefer,R. P. McNulty, C. A. Doswell III, and
vided by the National SevereStormsForecastCenter(NSSFC). R. F. Abbey, Jr., An augmented tornado climatology, Mon.
The NationalCenter for AtmosphericResearch•,NCAR)provided Weather Rev., 106, 1172-1183, 1978.
the graphicspackage used in this study. Kelly, D. L., J. T. Schaefer,and C. A. Doswell III, Climatologyof
nontornadic severe thunderstorm events in the United States,
Mon. Weather Rev., 113, 1997-2014, 1985.
REFERENCES
Purdom, J. F. W., Subjective interpretation of geostationary satel-
Anthony, R., Tornado/severethunderstormclimatologyfor the lite data for nowcasting,in Nowcasting, edited by K. A. Brown-
southeasternUnited States, in Preprints, 15th Conferenceon ing, pp. 149-166,Academic, San Diego, Calif., 1982.
Severe Local Storms, pp. 511-516, American Meteorological Schaefer, J. T., D. L. Kelly, and R. F. Abbey, A minimum
Society, Boston, Mass., 1988. assumptiontornado-hazardprobability model, J. Appl. Clim.
Brier, G. W., and J. Simpson,Tropical cloudinessand rainfall Meteorol.• 25, 1934-1945, 1986.
relatedto pressureand tidal variations,Q. J. R. Meteorol. Soc., Shea, D. J., The annual variation of precipitation over the United
95, 120-147, 1969. States and Canada, Rep. NCAR/TN-243+STR, 66 pp., Natl.
Court, A., and J. F. Griffiths, Thunderstorm climatology, in Thun- Cent. for Atmos. Res., Boulder, Colo., 1984.
derstorms:A Social, Scientific and TechnologicalDocumentary, Skaggs,R. H., Analysisand regionalizationof the diurnal distribu-
vol. 2, ThunderstormMorphologyand Dynamics,edited by E. tion of tornadoes in the United States, Mon. Weather Rev., 97,
Kessler,pp. 11-77, U.S. Departmentof Commerce,Washington, 103-115, 1969.
D.C., 1982. Wakimoto, R. M., and J. W. Wilson, Non-supercell tornadoes,
Davies-Jones,R., D. Burgess,and M. Foster, Test of helicity as a Mon. Weather Rev., 116, 1113-1140, 1989.
tornado forecast parameter, in Preprints, 16th Conference on Wallace, J. M., Diurnal variations in precipitation and thunderstorm
Severe Local Storms, pp. 588-592, American Meteorological frequencyover the conterminousUnited States, Mon. Weather
Society, Boston, Mass., 1990. Rev., 103, 406-419, 1975.
Holle, R. L., and M. W. Maier, Tornado formation from downdraft Weisman, M. L., and J. B. Klemp, The dependenceof numerically
interaction in the FACE mesonetwork, Mon. Weather Rev., I08, simulated convective storms on vertical wind shear and buoy-
1010-1028, 1980. ancy, Mon. Weather Rev., 110, 504-520,
 Oregon Tornadoes' More Fact Than Fiction
GEORGE R. MILLER

National WeatherServiceForecastOffice,Portland, Oregon97218

1. INTRODUCTION in the small farming community of Lexington in north-

central Oregon on June 14, 1888.
Factual accounts of tornadoes in Oregon are rare. The Tornadoes have been reported in Oregon in every month
U.S. Department of Commerce's Weather Bureau Annual except January (Figure 1). But even in this month an
Meteorological Summary With ComparativeData for Port- unconfirmed report was received in 1989 from the coast.
!and, Oregon, for the years 1922 and 1944 seemsto verify There are two distinct maxima, one in the springand one in
this with two very bold statements.The 1922publicationon the fall. April, May, and June have the greatest number,
page 3 states, "Tornadoes are unknown." The 1944publi- which is similar to other sectionsof the country [Kelly et at.,
cation on page 8 modifies this slightly by saying, "Well- 1978]. The incidence drops off markedly in July, the driest
developed tornadoes are unknown." month in Oregon. The fall maximum is quite likely due to the
These accountsfit well with the opinion of the majority of migration southward of the jet stream and storm track. The
Oregonians. They will tell you that they are happy to be total number, excluding the January 1989 report, is 44.
living in this state, where storms like the ones in the Midwest Tornadoes that occur in Oregon during the cool season
do not occur. This paper will show that tornadoes do occur (October through March) are confined to the area west of the
in Oregon, under atmospheric conditions that are similar to Cascade Mountain Range which bisects the state north to
those that produce tornadoes elsewhere. A few of the south. During these months, near-surface air east of the
atmospheric characteristicsassociatedwith Oregon torna- Cascadesis quite cool. Thus the instability that is necessary
does will be examined. The paper also offersthe hypothesis for convective activity is almost nonexistent. All reports of
that tornadoesoccur in Oregon at a frequency greater than tornadoes in December and January have been from the
that perceived by the populace. coastal strip.

2. TORNADO CLIMATOLOGY 3. TORNADO INTENSITY

Oregon is definitely out of the tornado belt [Flora, 1953]. Fujita classifiestornadoesas F-0 to F-5 [Fujita, 1987].His
One study [Hollifield, 1990]showsthat Oregonaveragedone scheme is abbreviated here for those unfamiliar with it: F-0,
tornado and one tornado day per year for the period 1953- winds40-72 mph (18-32 m/s); F-I, winds 73-112 mph (33-50
1989.That studyindicatesthat the greatestnumberreported m/s); F-2, winds 113-157 mph (51-70 m/s); F-3, winds
for 1 year was four in 1984. 158-206 mph (71-92 m/s); F-4, winds 207-260 mph (93-116
There have been no recorded deaths from tornadoes this m/s); and F-5, winds 261-318 mph (117-142 m/s). Using this
century in Oregon. (On April 5, 1972, a tornadotouched classification,Oregon tornadoes are ranked in the F-0 or F-I
down on the Oregon side of the Columbia River, moved range.
acrossthe river, and struck a shoppingcenter and school, Oregontornadoesinvestigatedover the last5 yearsby the
killing six peopleand injuring300 in Vancouver,Washing- author all have been less than F-2. The tornado that touched
ton.) Newspaperaccounts,however,indicatethatfour per- down briefly in Oregon before crossingthe ColumbiaRiver
sons were killed from tornadoes in eastern Oregon prior to in 1972 was classified F-2 or F-3 in Washington, but no
1900. Three of those deaths occurred in the eastern Oregon evidenceexists that it reached this intensity in Oregon.
communityof Long Creek on June3, 1894.The fourthwas The Long Creek and Lexington tornadoesprobablycome
the closestto F-2 strength in Oregon. A vivid, dramatic
The Tornado: Its Structure, Dynamics, Prediction, and Hazards. account of the former storm can be found in an article in The
Geophysical Monograph 79
This paperis not subjectto U.S. copyfight.Published
in 1993by the Long CreekEagle [Patterson,1894].A similaraccountof
American GeophysicalUnion. the Lexingtontornadocan be foundin a historicaldocument
454 OREGON TORNADOES

METEOROLOGICALFACTORS(SYNOPTIC
SCALE)AND EXAMPLES

As of this writing, there have been no detailed case studies

of tornado occurrencesin Oregon. No singleweather pattern
stands out as the most favorable for tornado formation in
Oregon. An unstable air mass with a strong jet and a
vigorousshortwave trough are items for which to look.
However, these features occur with regularity in Oregon as
elsewhere. An important difference between these features

-1 associated with tornado events in Oregon and those in the

Central Plains appears to be in the degree of instability.
Large, negativelifted index valuesare virtually unheardof in
Fig. I. Plot of tornadoesreportedin Oregonsince 1884. The Oregon, even with tornadoes.
Januaryreport was unconfirmed.
The jet streamappearsto be a factor for tornadoformation
in Oregon. Tornadoes have been reported with a strong
northwesterlyjet, a westerly jet, a south to southwesterly
jet, and a cyclonically curved jet. However, despite the
about Umatilla and Morrow counties [Parsons and Shiach,
orientation of the jet, most tornadoes cases that were exam-
1902]. It shouldbe taken into consideration,when reading
ined occurred in the left front quadrant of the jet maximum
historical accountsof tornadoes from small newspapers, that
(jet streak). This coincides with the area where upward
there may be a bit of dramatizationto the reports. Never-
vertical motion would be expected with a jet streak circula-
theless,they underline the fact that tornadoeshave caused
tion [Uccellini and Johnson, 1979].
destructionand killed people in Oregon.

11 7E76
7277

12 72•S

-11 579 ?_ _ee'•

-11 583

Fig. 2. The 500-mbar geopotentialheight analysis(60-m contours), 1200 UTC, August 22,
 MILLER 455

-lOO -90 -80 -GO -S0 -40

/!

// / /

/,
/
/ •3e//•,•/ 230/ 43

/ SHOWALTERINDEX = 03
+ ESO/ SS
L•FTED INDEX " -01
ß K INDEX =
23S/ 82
./ PRECfPtJATER(IN')=' 0.95

E00
/ '•/ --40

/ , . CROSS S =
/ ,
/ VERTICALTOTALS'='
TOTAL TOTALS '
> MOD•W•ATINDEX='018S 830/ 93

230/104
?
30(
300
/ /
/ /
,/
/ /
/
. / / 230/
/
400
400
/
,/
/ ./
/ . --20 2IS/ S5
500
S00
ß/ /' / . --16 22S/ 47
g00
/. / . .,/ ,--'14 •/+ 225/39
•00
/ ?. /', --12 22S/ 32
700 700 •10
--9
•+ 23S/
240/
13
11
/, , / ', /,. './ ,. . --B •+ ESO/
!1
800 800
. ,./ ', .
'/ , y. .
ß/ '
900 .900
' 3 + 330/ 10
eee / ' "/" /" "/ ' / --1 190/ 03
10s0 -30 -ZO -!0 0 le ZO 3o 1050 DIR/KTS DIR/KTS

Fig. 3. Upper air soundingin Salem, Oregon,at 1200UTC, August 22, 1989.

Bend, Oregon, Tornado, August 22, 1989 PacificNorthwestwith relatively cool air aloft. The trough
On August 22, 1989, at around 1 P.M. PDT (2000 UTC) a moves slowly through the area.
No soundingsare available from east of the Cascades in
tornado was reported just east of Bend, Oregon, in the
central portion of the state. This area is east of the Cascade Oregon. The winds at Salem (Figure 3) show backingwith
Mountains and at an elevation of 3000-4000 feet (915-1220 heightbelow about 700 mbar, indicatingcold air advection.
m). An active shortwave trough (Figure 2) is entering the Above 700 mbar, winds show considerablespeedshearbut
little directional change with height. (The portion of the
soundingbelowabout700 mbar is probablyunrepresentative
for the central part of the state, but the moderatelapserate
and considerablespeed shear above 700 mbar suggestthat
the soundingin central Oregon may have been quite compa-
rable to tornado soundings elsewhere.) The subsequent
soundingfrom Salem at 0000 UTC on August 23, 1989(not
shown), revealed cooling of the order of 2ø--5øCat most
levels of the atmosphere,especiallyabove 550 mbar.

Eugene, Oregon, Tornado, November 24, 1989

On November 24, 1989, a tornado was sightedjust south
of Eugene, Oregon, in the southernWillamette Valley, west
Fig. 4. The 300-mbarisotachs(knots' 1 knot = 0.5 m/s)at 1200 of the Cascade Mountain Range. A jet maximum of 130
UTC, November 24, 198.9. knots (65 m/s) (Figure 4) was just west of the
456 OREGON TORNADOES

-40
100' lOO --53

--S0
/ ?'
/, /, / /
/
13•
/ SHOW•LTE•INDEX =
L•FY•D INDEX =" 00
,/ K INDEX. = 19
/
20 PRECfP•AT%IN)=
,
i 047
CROSS TOYALS =
87
V•T I CAL TOTALS' =' 88
>/ ß
•OTAL'TOTALS
=' SS ß 0366

,/ ¾-,, •k\\\ 270/ B1

/
300 --30
300

/ / ß
•/ /-
? /
/ /
•00
/- / '/
ß / /
./

S00 •18 +2•S/

S2
S0(
/, / ,
/
/', '. , './ 16 / •0/ S2
/,
.

/ , /, ,
•oo ,/
+ 2GS/ 27
800 800 2S0/ 24
,/ , ' / /' . / '/
230/ 24
900
/,
ß

ß / ,. / . ,,/ / ß /
•00 • 215/ 2•

looo • 1Bo/
11
1000
10S0 lOSO •o
-30 -80 -10 0 10 20 •0

Fig. 5. Upper air soundingin Salem,Oregon, at 1200UTC, November 24, 1989.

Californiaborder,andOregonwasin the left frontquadrant.

The lifted index, as shownin Figure 5, is near zero with a
strongjet above the tropopause.Although it is relatively
shallow,a layer of steeplapserate is presentin midtropo-
sphere(about600to 450 mbar),andthe windsshowveering
off the surface.The entire soundingis cool relative to its
midwestern counterparts,but it is not inconsistentwith a
shallow supercell.As Braun and Monteverdi [1991] have
shown, supercelisare not unheard of on the West Coast. The
subsequent soundingtaken at 0000 UTC on November25,
1989 (not shown), revealed cooling above 600 mbar and a
slightdecreasein wind with little changein direction.

Vancouver, Washington, Tornado,

April 5, 1972

The pattern that spawnedthe destructiveVancouver,

Washington,tornado on April 5, 1972, consistedof a south-
southwesterly
jet in excessof 110knots(55 m/s)(Figure6). Fig. 6. The 300-mbar isotachs(knots' 1 knot = 0.5 m/s) for 1200
The tornadooccurredaround1 P.M. localtime (2000UTC) UTC, April 5,
 MILLER 457

in postfrontalconditions
and was associated
with an active uncommon, it is typical that tornadoes often go unreported
squall line. [e.g., D•swell, 1980]. Oregon appears to be t.vpicalin this
regard, but the author is trying to change this perception.
5. SUMMARYANDCONCLUSIONS More occurrences should lead to a better understanding of
the meteorologicalfactors associatedwith tornado develop-
There have been documented accountsof tornadoesin
ment in Oregon.
Oregon with recorded deaths. Some have causedconsider-
able damage, others slight damage.In the tornadocases
considered,soundings
taken beforeand after generally Acknowledgments. The author is grateful for the numerous
showednegativeto only slightlypositiveliftedindices.For suggestions
and helpful comments provided by the reviex•ers,the
the May throughSeptembertornadoes
the mid and upper staff in Scientific Services Division, Western Region, National
atmosphericflow was generally from the southwestwith a Weather Service, and especially to C. A. Doswell lII, National
SevereStorms Laboratory, for his immeasurableassistance.
tropopauselevel of between30,000and 40,000feet (9 to 11
km). OctoberthroughMarch tornadocasesshowmostlya
REFERENCES
west to northwesterly flow.
Much work remainsto be done regardingtornadoesin Braun, S. A., and J.P. Monteverdi, An analysis of a mesocyclone-
Oregon.There is a need for detailedstudiesaboutspecific induced tornado occurrence in northern California, Weather
occurrences,includingsoundingandhodographanalyses,as Forecasting, 6, 13-31, 1991.
well as constant level charts. Perhapsafter this is accom- Doswell, C. A., III, Synoptic scale environmentsassociatedwith
High Plainsseverethunderstorms,Bull. Am. MeteoroL Soc., 60,
plished,somedefiniteideascanbe identifiedfor anticipating 1388-1400, 1980.
tornado formation in Oregon. Flora, S. D., Tornadoesof the United States, 194 pp., Universityof
Currently, the only facts about Oregontornadoesare as Oklahoma Press, Norman, 1953.
follows: (1) They reach a maximum in June. (2) Tornadoesin Fujita, T. T., U.S. tornadoes,part i, 70 year statistics,report, 120
pp., Satell. and Mesometeorol.Res. Proj., Dep. of the Geophys.
the colder months of the year (October-March) are confined Sci., Univ. of Chicago, Chicago, Ill., 1977.
to the western sections of Oregon. (3) In the few casesthat Hollifield, J., National summary of tornadoes, 1989, Storm Data
were superficially studied, Oregon was located in the left Bull., 31, 1-10, Dec. 1990.
front quadrant of the jet stream. Kelly, D. L., J. T. Schaefer,R. P. McNulty, C. A. Doswell III, and
R. F. Abbey, Jr., An augmented tornado climatology, Mort.
It seems likely that reported occurrencesof tornadoesin Weather Rev., 106, 1172-1183, 1978.
Oregon will increase. This is due, in part, to the publicity Parsons,W., and W. S. Shiach, An Illustrated Histo•' of UmatilIa
these storms are receiving in other parts of the United County and Morrow Counr3.',W. H. Lever, t902.
States. Another contributing factor is the increase in the Patterson,O. L., in The Long Creek Eagle, p. 1, Friday, June 8,
1894.
number of residents from areas outside Oregon where tor-
Uccellini, L. W., and D. R. Johnson, The coupling of upper and
nadoes are more common.
lower troposphericjet streaks and implicationsfor the develop-
As has been noted elsewhere, tornado reporting is always ment of severe convective storms, Mon. Weaiher Rev., I07,
uneven. In stateswhere the perceptionis that tornadoesare 682-703,
 The Stability of ClimatologicalTornado Data
JOSEPH T. SCHAEFER 1 AND RICHARD L. LIVINGSTON

CentralRegionScientific
Ser•,ices
Division,
NationalWeatherService,KansasCity,Missouri64106

FREDERICK P. OSTBY AND PRESTON W. LEFTWICH 2

National SevereStormsForecastCenter,National WeatherService,KansasCity, Missouri64106

1. INTRODUCTION ters and from station to station [Court, 1968]. Hence an

empiricalexploratory
dataanalysisapproach
is neededto
The National Severe Storms Forecast Center's (NSSFC) describethe climateadequately[Guttman, !989].
log of all tornadoesreportedin the United Statessince In this paper, tornadooccurrencestatisticsover the co-
January 1, 1950, containsthe latitude/longitude of each terminousUnited Statesfor the 40-year period 1950-1989are
tornado'sbeginningand endingpoints, its observedtrack examined.First, variationsin the annualnumber of torna-
lengthandaveragepathwidth,theFPPindicesasdefinedby doesreported duringthestudyperiodareanalyzed.Thenthe
Fujita and Pearson[1973],and an estimateof the dollar spatialdistributionof tornadooccurrence frequency is com-
amountof damagecaused.Many climatological-type studies putedforvariousaveraging periods.Finally,it is shownthat
[e.g.,McNultyet al., 1979;Schaeferet al., 1980,1986]have the tornadooccurrencepattern doesnot stabilizeuntil 35
usedthis historicaldata set. However, none have questioned yearsof dataare considered. It is arguedthat this pattern
whether these studiesfulfill the basicpurposeof a climatol- corresponds to the U.S. tornadoclimatology.
ogy,thepresentation of thecharacteristicstateandbehavior
of the atmosphere[Jagannathanet al., 1967]. 2. True SER1ES ANALYSIS
To seethe dependence of the apparenttornadoclimatol-
ogyuponthe periodoverwhichdatawereanalyzed,one In the40 years 1950-1989,confirmedtornadoreportsover
needonlycompare state-by-state tornadototalsforthe1970s the coterminousUnited States totaled 28,384. The minimum
and 1980s(Figure1). The markeddifferences betweenthe was201in 1950,whilethe maximumwas 1102in 1973.A graph
state totals for the two 10-yearperiodsindicatethat as a of the annualtornadocount(Figure 2) showsmarkedyear-to-
general rule,tornadoactivitydecreased eastof theMissis- yearfluctuations.Thehighest countin thefirstthreeyearsof
sippiRiverandincreased westof it. Thusa 10-year tornado the databaset260in 1951)is only62% of the lowestcountof
summary couldbeusedto indicate eitherthatIllinoisranks theremaining 37years(421in 1953.). Thusa sizablenumber of
fifth amongthe statesin the annualnumberof tornadoes tornadoeswere mostlikely unreportedbefore 1953.
(fromthe 1970-1979 data)or fifteenth(1980-1989 data). While the 1953 count is also low, since it is comparableto
Climatologicaldataare usuallypresented in termsof the 463 reportsreceivedin 1963,it cannotbe so easily
"normals"which,in accordance withWorldMeteorologicalflagged asbeingerroneous. The increase in the numberof
Organization (WMO)procedures, areaverages overspeci- tornadoes reported intheearly1950scanberelatedto public
fied30-yearperiods. However,thelengthof a statisticallyawareness.The U.S. Air Force startedalertingtheir facili-
meaningful averagingperiodvariesamong weather parame- tiesof potentialtornadicweatherin 1948,andthe Weather
Bureaustartedreleasingpublic tornadoforecastsin 1952
•Now at NationalWeatherServiceTrainingCenter,Kansas
City,[Galway,1993].Further,tornadoesdevastated Waco,Tex-
Missouri 64124.
as; Flint, Michigan:andWorcester,Massachusetts,
in 1953,
2Nowat CentralRegionScientificServices Division,
National causing 114, 116,
and 90fatalities,
respectively.
(Thereissome
WeatherService,KansasCity, Missouri64106.
disagreement
as to the numberof fatalities
withthe Flint,
TheTornado:Its Structure,
Dynamics,
Prediction,
andHazards. Michigan,
tornado.TheNSSFCdatabase,Flora[1954], and
GeophysicalMonograph79 Snider[1975]eachattribute116•htalities
to thisstorm.How-
Thispaper
isnotsubjecttoU.S.copyright.
Published
in1993
bythe ever,Linehah[1957]onlylists115deaths.Graz.
ulis[19901
puts
AmericanGeophysicalUnion.
460 TORNADO DATA STABILITY

1970s TO 1980s TORNADO TREND: -383 ,,

Fig. 1. The changein decadal tornado count by state between the 1970sand 1980s. A decreasein the number of
reportedtornadoeswas recordedin the 1980sas comparedto the 1970s.The national changedoes not equal the sum
of the changesfor the individual statesbecauseof tornadoesthat crossedfrom one state into another.

the death toll for this storm at 1 !5 and attributes another death The vagaries of the annual tornado count can be seen in
to a secondstormthat occurredin the sameareathat night. The accumulated tornado occurrences for local areas (Figure 3).
NSSFC data base shows no deaths with this second storm.) A rapid increase in population typically corresponds to a
To distinguishsubtle features in the pattern underlying consistentgrowth in the rate that tornadoesare reported. An
thesedata, a shape-preserving smootherdevelopedby Tukey exampleof this is HillsboroughCounty in west central Florida.
[1977] was applied to it. The resulting curve (superposedon This county, which containsthe city of Tampa, had a popula-
Figure 2) shows that the increase in the annual number of tion increaseof 156,695 between 1970 and 1980. The cumula-
tornadoes through the 1950s and 1960s did not persist tive numberof tornadoesin Hillsborough County since 1950is
through the 1970sand 1980s.In fact, quite possibly tornado concavedownward throughthe mid-1980s,indicatingthat, in
activity has decreasedsince 1973. Perhaps this decreasing general,more tornadoeswere reported each subsequentyear.
trend in tornado occurrences is related to procedural In the greaterHouston area, Harris County, Texas, the 1970-
changes in the National Weather Service (NWS), which !980 populationincreasewas 667,632. The accumulatedtor-
abolishedits state climatologistpositionsduring 1972 and nado curve generallyparallelsthat of HillsboroughCounty,
1973. After the demiseof this programthere were no longer Florida, except for a distortionin 1983 when 22 tornadoes
individuals who were dedicated to the collection of informa- occurredas hurricaneAlicia approachedthe Gulf Coast.
tion on the occurrence of severe thunderstorms. In contrast, the population of Kay County, Oklahoma,
One way to work around the apparent inflation in the which contains Ponca City and is in "tornado alley," has
annual number of reported tornadoesis to consider"tornado been rather sluggish.It had a population decreaseof 2251
days" rather than actual tornadoes. Court and Griffiths during the !960s and gained only 1065 people in the 1970s.
[1985] showed that since 1916, the annual number of tornado The slope of the accumulated tornado curve was rather
days has increased much less than the number of tornadoes. constant,indicating that the rate of reported tornado occur-
Part of thisis a reflectionthat there is a physicalupperbound rence did not change substantially over the past four de-
to the possiblenumber of annual tornado days. Further, as cades.Weld County, Colorado, around Greeley in the north
noted by Court [1970, p. 39], "the nonmeteorologistis central part of the state, has also remained rather unpopu-
interestedin the one tornado that may effect him, and hence lated. Even thoughits populationgrew by 34,141 duringthe
the public is better served by tabulations and maps of 1970s, the 1980 population density was still less than 3!
tornado incidence rather than of the number of days on people per squaremile. Through 1978 most of the tornadoes
which they occur." reportedin this county occurredin a few active years
 SCHAEFER ET AL. 461

FO and F1
6O

i,:.,t
1200
!1oo

h i ooo
5O
F2 and great.er
9o0

$oo

: 700
6OO P 3o
5OO r

4OO c 20

300 n

t lO io
, 200

100
o
1
5O 55 60 65 70 75 80 85 50 55 60 65 70 75 8.0 85
YEAR '• EAR

Fig. 4. The ratioof strongand violenttornadoes(F-2 andgreater)

Fig. 2. The number of reportedtornadoesin the United Statesby to the total number of' reported tornadoes.
year. The solid lines superimposedover the bars indicate data
smoothedby the Tukey filter.

on the Fujita scale)was constructed(Figure 2). While the

1957, 1958, and 1976). Then tornadoes started occurring fluctuationsare smaller, the Tukey-smoothed annual count
extremely often but at a nearly constantrate. One possible of these more intense tornadoes has a shape similar to the
explanationfor this is the advent of the PrototypeRegional one for the annualnumberof all reportedtornadoes,regard-
Observing and Forecast Service (PROFS) programwith its lessof intensity.However, the percentageof tornadoesthat
tornado "chasers" and its emphasis on providing ground are F-2 or greater(Figure4) has decreasedwith time. It has
truth over northeastern Colorado [Beran and Little, 1979]. beenproposedthat the decreasein the annualnumberof
It has been noted that populationbiasesin tornadoreport- tornadoes since 1973 was a result of the recognition of the
ing are reducedwhen only strongertornadoesare considered microburstphenomena[Fujita, 1976] causing storms that
[Kelly et al., 1978]. A time seriesfor thosetornadoesthat are hadoncebeenreportedas weak tornadoesto be classifiedas
rated as strongandviolent in the NSSFC log (F-2 andgreater straight-lineseverethunderstormwinds. However, the con-

c n ,, WELDco

C
12.5
......................
i'i
i 12,5
U • HILLSBOROUGH
FL
m

u • KAYOK

O HHAEEI$
..•
...............
'.'.'.'
d 7s ........................................ 7s

T
o

d
o 25 25
S

o • •" ' I" ' ' ' I ' ' ' ' i ' '"""I ' '"' ' I '"' '" ! ' ' '" ! ' ' '""'
1950 1955 1960 1965 1970 1g75 1980 1985 1990
Year

Fig.3. Theaccumulated
number
ofreported
tornadoes
since
January
1,1950,
lbrfourselected
462 TORNADO DATA STABILITY

Fig. 5. Tornadofrequencycomputedfor four differentbut overlappingtime periods:(a) 5 years(1960-1964),(b) 10

years(1960-1969),(c) 15 years(1955-1969),and (d) 20 years(1955-1974).Stippledareasrepresentfour to six, eight to
10,and12to 14tornadoes
per 10,000mi2 (25,900km2) peryear.Crosses
markthelocations
of relativemaxima,and
N marks the locations of relative minima.

tinuedincreasein the percentageof weak tornadoes(F-0 and the past half-century," Court [1970, p. 40] recommended
F-l) through the 1980s counters this notion. that tornado frequency charts for 10-year periods be pro-
ducedevery 5 years. While this is a reasonableprocedurefor
3. TIME DEPENDENCE OF SPATIAL DISTRIBUTION many meteorologicalvariables,Figure 1 demonstratesthat it
is not valid for tornado data.
To evaluate the geographicdistribution of tornadoes, the
To illustrate the effect of the averaging period on apparent
annual frequency of tornado occurrence per unit .•rea is
needed. This parameter is computedby tabulating the num- tornado frequency, first consider the chart produced from
ber of tornado touchdownsin 2ølatitude-longitudequadran- the tornado record from the 5-year period !960-1964 (Figure
gles and then normalizingby both area and time. Calculating 5a). This short record yields the rather classicpicture of a
the 2ø "square" values at 1ø latitude and longitude incre- "tornado alley" running north-northeastward from near
ments acrossthe coterminousUnited States yields a light Lubbock, Texas, through Oklahoma City, Oklahoma, and
smoothing,equivalent to that achieved by passinga boxcar immediatelywest of Kansas City, Missouri, to the Mason
average over 1ødata [Kelly et al., 1978]. City, Iowa, area. An area of enhanced tornado frequency
However, as noted in the introduction, the period of the (greater
thansixtornadoes
per10,000mi2 (25,900km2)per
record, or "averaging period," plays a significantrole in year) is also found in north central Indiana.
determining the apparent tornado climatology. Court [1970] The averaging time is then increased by 5-year incre-
published a compilation of I09 different maps and/or ments, still includingthe initial data. Adding the second 5
pseudomapsof tornado occurrence acrossthe United States years, 1965-!969, expandedthe data baseto includethe 1965
dating back to 1884. In his discussionof these charts he Palm Sunday tornado outbreak across Iowa, Wisconsin,
noted that the true distribution of tornadoes cannot be Illinois, Michigan, and Indiana and the 118 tornadoes re-
establishedsincethe variousanalyseswere basedon reports portedin southTexasbetweenSeptember18and September
for differing periods. On the basis of the premise that "the 23, 1967, while the circulation of Hurricane Beulah was
most recent decade is a better estimator of the next one than affecting the area. (Estimates of the number of
 SCHAEFER ET AL. 463

Fig. 6. Tornado frequencycomputedfor four different25-year periods:(a) 1950-1974,(b) 1955-1979, (c) 1960-1984,
and (d) 1965-1989. Legend is as in Figure 5.

associatedwith Hurricane Beulah range from 47 [Dye, 1968]can argue that this pattern Jsthe "real" tornado climatology
to 141 [Novlan and Gray, 1974]. Both of these events are for the coterminous United States, it is necessary to deter-
reflected in the decadal frequency chart (Figure 5b). mine if the 25-year pattern is stable. Accordingly, four
The 15-year presentation created by including 1955 25-year charts were constructed from the 40-year data set
through 1959 in the analysis (Figure 5c) gives less promi-
(Figure 6).
nence to Iowa and Minnesota. However, largely becauseof Marked interchart differences, over and above the popu-
February outbreaksin both 1956and 1959, the area from St.lation density increases in northeast Colorado and central
Louis, Missouri, across Illinois to north of Indianapolis,
Florida, appear. The 1950-1974 chart indicates a corridor of
Indiana, now has an apparently higherfrequency. minimal activity running northward from Louisiana to north-
The addition of a fourth 5-year period still doesnot yield a
west Iowa. In the 1955-1979 presentation this corridor be-
stable pattern. The tornado distributioncomputedover the comesbroken up and remains so in the later two charts.The
20 years 1955-1974(Figure 5d) is markedlydifferentfrom
zone of enhancedactivity that extended northward from cen-
the one obtained from the 15 years 1955-1969. Large por-
tral Oklahoma through central Kansas in the 1955-1979map is
tions of Mississippi,Alabama, and Georgia now have an
replacedby a local minimum in the 1965-1989analysis.
apparent annual tornado frequencyof greater than four
Since the 25-year pattern is not stable, a longer data span
tornadoes per 10,000mi2 (25,900km2).Thisarosebecause
is neededto define the tornado climate. Three presentations,
the early 1970swere extremelyactivein the southeast,with
based on a 30-year average time, were constructed (Figure
Georgia and Mississippiranking in the top six statesin
tornado occurrences in both 1971 and 1972.
7). Significant changes between the charts are found in
Kansas, Illinois, and Louisiana. Tornado alley, which is a
definitefeature in the 1950-1979 chart, does not appearin the
4. CONSISTENCYOF SPATIAL PATTERNS
1960-1989 analysis. These charts illustrate that the usual
The 25-year tornadooccurrencechart computedfrom climatological averaging period of 30 years is not long
1950through1974(Figure6a) is quitesimilarto the20-year enoughto yield a meaningful representationof the tornado
onecomputed
from1955through1964.However,beforeone occurrence
464 TORNADO DATA STABILITY

Fig. 7. Tornado frequency computed for three different 30-year periods: (a) 1950-1979, (b) 1955-1984, and (c)
1960-1989. Legend is as in Figure 5.

enough to yield a meaningful representation of the tornado

occurrence pattern.
In contrast, the two 35-year climatologies (Figure 8)
constructed from these data are quite consistent. Further,
they are compatible with the 40-year one constructed using
the entire data set (Figure 9). Obviously, part of the consis-
tency can be attributed to the fact that all three of these
presentations contain data from the same 30-year span
(1955-1984). However, the first two 30-year displays (Fig-

Fig. 8. Tornado frequency computedfor two different 35-year Fig. 9. Tornado frequencycomputedover 40 years, 1950-1989.
periods:(a) 1950-1984and (b) 1955-1989.Legendis as in Figure5. Legendis as in Figure
 SCHAEFER ET AL. 465

used to depict the tornado occurrence pattern across the

contiguousUnited States.

5. U.S. TORNADO CLIMATOLOGY

The salient feature of the tornado climatology of the

United States is a general tornado-prone region between the
Rockies and the Appalachians, generally south of 44øN,
which contains embedded areas of enhanced activity from
northern Texas through Oklahoma, and along a west-to-east
corridor from northeastern Colorado through eastern Ne-
braska and central Iowa to north central Indiana. Areas of
decreasedactivity in this area are zones from northeastern
Missouri through central Arkansas, eastern Mississippi
through central Alabama, and a small pocket in east central
Texas.
The above procedures also were applied to the data set
containing only strong and violent tornadoes (F-2 and great-
er). The geographicpattern again stabilized with the 35-year
averaging time (Figure 10). Again, the 35-year charts are
similar to the one constructed from the entire 40 years of the
data set (Figure 11). The area most prone to strong and
violent tornadoesis central Oklahoma, but small pocketsof
enhanced occurrence are also found in eastern Kansas,
eastern Missouri, central Arkansas, southeastern Missis-
Fig. 10. Frequency of strong and violent tornadoes (F-2 and sippi, northern Alabama, western Kentucky, north central
greater) computed over two 35-year periods:(a) 1950-1984and (b) Indiana, and north central Iowa.
1955-1989. Stippled areas represent two to three and four to five
tornadoes
per 10,000
mi2 (25,900km2) peryear.Crosses
markthe However, since many of these areas are located around
locations of relative maxima, and N marks the locations of relative cities, these may not be real zones of enhanced strong and
minima. violent tornado activity, but rather reflectionsof population
biasesin the data and the fact that probability of obtaininga
high rating in a damage-basedtornado classificationsystem
ures 7a and 7b) cover a common 25-year span (1955-1979) increaseswith the number of damageable structures as well
and do not show a similar consistency. It is hard to argue as with the strength of the storm. Indeed, the listing of
that having86% of the time spanin common(30 of 35 years) individual "significant" tornadoes over the last century
is significantlydifferentthan having83% (25 of 30 years) in compiled by Grazulis [1990] indicates the many areas that
common. Thus it can be stated heuristically that a 35-year historically have been susceptible to strong and violent
averagingtime is necessaryto obtain statisticsthat can be tornadoes are not enhanced zones in the 40-year climatol-
ogy. It is quite likely that the relative infrequency of strong
and violent tornadoes (the NSSFC log only records 8271 of
those storms in its 40 years of record) precludesdetermining
a meaningful occurrence pattern of F-2 and greater torna-
does over the United States at the present time.

Acknowledgments. We owe specialgratitudeto Leo Grenierand

John Halmstad, who maintain the NSSFC tornado data base and
who providedus with much detailed informationaboutparticular
items discussedin this paper. The graphicswere generatedby
MichaelManker, and the wordswere expertlyeditedandprocessed
by Beverly Lambert. Joe Galway provideda very insightfulreview
which markedly clarified the presentation.

REFERENCES

Beran, D. W., and C. G. Little, PrototypeRegionalObservingand

Fig. 11. Frequencyof strongand violent tornadoes
iF-2 and Forecast Service: From concept to implementation, Natl.
greater)computed
over40 years,1950-1989.
Legendis asin Figure Weather Dig., 4, 2-5, 1979.
10. Court, A., Climatic normals as predictors, 5, Conclusions,
466 TORNADO DATA STABILITY

AFCRL 69-0003,74 pp., Air Force CambridgeRes. Lab., Bed- Guttman, N. B., Statistical descriptors of climate, Bull. Am. Mete-
ford, Mass., 1968. orol. Sot., 70, 602-607, 1989.
Court, A., Tornado incidencemaps, Tech. Metno. ERLTM-NSSL Jagannathan,P., R. Arley, H. ten Kate, and M. V. Savarina, A note
49, 76 pp., Natl. Severe Storms Lab., Environ. Sci. Serv. Admin., on climatological normals, Tech. Note 84, World Meteorol. Or-
Norman, Okla., 1970. gan., Geneva, Switzerland, 1967.(Also availab!eas Rep. WMO-
Court, A., and J. F. Griffiths, Thunderstormclimatology,in Thun- No. 208.TP.198.)
derstorms:A Social, Scientific,and TechnologicalDocumentaty, Kelly, D. L., J. T. Schaefer,R. P. McNulty, C. A. Doswell III, and
vol. 2, ThunderstormMo•ohologyand Dynamics,2nd ed., edited R. F. Abbey, Jr., An augmentedtornado climatology, Mon.
Weather Rev., 106, 1172-1183, !978.
by E. Kessler, pp. 9-40, University of OklahomaPress,Norman,
Okla., !985. Lineran, U. J., Tornadodeathsin the United States,Tech. Pap. 30,
48 pp., Natl. Weather Serv., Silver Spring, Md., 1957.
Dye, L. W., Tornado summary for 1967, Weatherwise,21, 22-25,
1968. McNulty, R. P., D. L. Kelly, and J. T. Schaefer, Frequencyof
tornado occurrence, in Preprints, 1Ith Conference on Severe
Flora, S. D., Tornadoesof the United States, reviseded., 206 pp.,
Local Stortns,pp. 222-226, American MeteorologicalSociety,
University of Oklahoma Press, Norman, Okla., 1954. Boston, Mass., 1979.
Fujita, T. T., Spearheadecho and downburstnear the approachend Novlan, D. J., and W. M. Gray, Hurricane-spawned tornadoes,
of a JohnF. Kennedy Airport runway, New York City, Res. Pap. Mon. Weather Rev., 102,476-488, 1974.
137, 51 pp., Satell. and MesometeoroI. Res. Proj., Univ. of Schaefer, J. T., D. L. Kelly, and R. F. Abbey, Tornado track
Chicago, Chicago,I11., 1976. characteristicsand hazard probabilities,in Wind Engineering,
Fujita, T. T., and A.D. Pearson, Results of FPP classificationof edited by J. E. Cermak, pp. 95-110, Pergammon,New York,
1971 and 1972 tornadoes, in Preprints, Eighth Conference on 1980.
Severe Local Storms, pp. 142-145, American Meteorological Schaefer,J. T., D. L. Kelly, and R. F. Abbey, Developmentand
Society, Boston, Mass., 1973. applicationof a minimumassumptiontornadohazardprobability
Galway, J. G., Early severe thunderstormresearchby the U.S. model, J. Clim. Appl. Meteorol., 25, 1934-1945, 1986.
Weather Bureau, Weather Forecasting, in press, 1993. Snider, C. R., Michigan tornadoes1834-1975,internal report, 37
Grazulis, T. P., Significant Tornadoes1880-1989, vol. II, A Chro- pp., Natl. Weather Serv. Office, Ann Arbor, Mich., 1975.
nology of Events, 368 pp., Environmental Films, St. Johnsbury, Tukey, J. W., Explorato•3,Data Anal, sis, 535 pp., Addison-Wesley,
Vt., 1990. Reading, Mass.,
 A 110-YearPerspectiveof SignificantTornadoes
THOMAS P. GRAZULIS

St. Johnsbur3,,Vermont 05819

1. EVOLUTION OF THE PROJECT oratories,and resultedin the publicationof Grazulis [1984].

The two data setswere made compatiblefor violent events
A tornado documentationproject began in 1980 as an
(F-4-F-5) andbroughtto within one F scaleratingpointfor
effortto resolvedifferences
betweenthe two independently all others. A best possible F-4-F-5 data set was crucial to
designed tornadodatabasesusedby theNuclearRegulatory NRC assessment of 10-7 maximum windspeedprobabili-
Commission(NRC). One was establishedfor the NRC at the ties. Schaefer et al. [1986] noted that 4% (about 1300
Universityof Chicago(UC) by T. T. Fujita for the years tornadoes)of the NSSFC data base had a rating change.
1916-1985. In addition, the NRC assisted the National These were based on the recommendations of the author, in
SevereStormsForecastCenter (NSSFC) in addingF scale this first phaseof the project. A similarnumberof changes
ratingsto the existingNSSFC data base in KansasCity, were made at UC. No other changeswere made after 1983.
from 1950 to the present. The final task of this effort There remained,however, about5000 eventsdifferingby
(hereinafter called the project) was to locate and list all one F scale rating, many of them across the important
significanttornadoesfrom 1880through1989.Significantis dividingpoint betweenweak (F-0-F-1) and strong(F-2-F-3)
defined here as all tornadoes doing confirmable F-2 or events. These differences, and all F-2-F-5 events since 1880,
greater damage or causinga death. were explored in the secondphaseof the project, fundedby
The inherent subjectivity of the Fujita scalegave rise to the National Science Foundation (NSF). It culminated in the
thousandsof F scalerating differences,as the two databases work of Grazulis [1991], which describes the 12,209 torna-
were compiled independentlyby teams of students.For the does found to be "significant" from 1880 to 1989.
period 1950001970 (the years before the introductionand use The entire effort required about 500 days of full-time
on-the-roadlibrary work, spreadover nine years.The search
of the Fujita scale by Fujita [1971]), nearly half of all
for information required the use of about 35,000 reels of
tornadoes were assigneda different rating. In a few drastic
newspaper microfilm. Most of this microfilm is archived in
casesthe ratings differed by as much as four F scalerating
centralstate historicallibraries, exceptin Texas, New York,
points, as UC graduatestudentsworked from StormData (in and Massachusetts, where visits to local libraries and news-
Chicago)and undergraduateNSSFC studentsworked from paper offices were required.
both Storm Data and newspaper microfilm (in their home
states during the summer of 1975). (Storm Data is published
by the U.S. Department of Commerce, National Climatic 2. PRE-1915

Data Center, in Asheville, North Carolina.)

The original purposefor attempting to expand the useful
The discrepancieswere found originallyby McDonald and
tornadodata basefrom 1916 back to 1880was to provide the
Abbey [1979]. Researchby this author, aided by computer- NRC with a long violent tornado data set. Since there are
assistedcomparisonsat Texas Tech, revealedthe extent and only 100020 F-4/F-5 tornadoes per year, it was assumedthat
nature of the differences. For instance, both data sets the relatively few documented tornadoes for that period
initially had 40 violent (F-4-F-5) tornadoesin Oklahoma,but would include the violent events. In addition, it was thought
only seven were in common. There were about 2000 events that virtually all killer tornadoes could be located and that
that differed by two or more F scalerating points. such a data set would provide at least some insight into
The NRC phase of the project was completed under the long-term trends. The project documented2432 significant
guidance of R. Hadlock at Battelle, Pacific Northwest Lab- tornadoes (including 1060 killer tornadoes) for the years
The Tornado: Its Structure, Dynamics, Prediction, and Hazards. 1880-1915.These paths are plotted in Figure 1. Tornado and
GeophysicalMonograph79 severe weather occurrence dates were found in the Monthly
Copyright 1993by the American GeophysicalUnion. Weather Review,',Month(v Climate and Crop Report (the
468 SIGNIFICANT TORNADOES OVER 110 YEARS

! _..

Fig. 1. A subjectivecomparisonof 1880-1915tornadopathswith 1916-1985risk from Grazulis[1991].The tornado

pathsarefor 2432significant
tornadoes,
1880-1915.
Isopleths
aredrawnfor a 120-mph(53.7m s-•) maximumwind
speed,
at 10-4 probability,
asderived
fromFujita[1987].
Areas
withrelatively
lowriskfor1916-1985
andhighactivity
for 1880-1915includetheupperMissouriRiverValley fromOmahato southwest
MinnesotaandeasternSouthDakota,
theOzarkPlateauofsoutheast
Missouriandnorthwest
Arkansas,
tworelativelysmallareasofcentralKansasandalong
the Kansas-Nebraska border, and northwest Illinois.

precursorof ClimatologicalData), the AnnualReportof the areas of high pre-1916 activity that were not active in more
Chief of the U.S. WeatherBureau, 19th-centuryreports recent periods. Figure 1 identifies five such areas. This adds
suchas thoseof Finley [1884, 1889],the author'sfiles, and a cautionarynote to the use of current risk analysismaps.
much personalcorrespondence.For every known or sus- The long-term risk from strong and violent tornadoes for
pectedtornadoin this period,an attemptwas made to seek some areas of the United States may be underestimatedby
out local newspaperdescriptions.No previouslyunlisted Fujita [ 1987],as well as by other recent risk analysismodels.
tornado was added unless a sufficiently detailed funnel This pre-1916 period has, unfortunately, additional vari-
descriptioncouldbe locatedin a localnewspaper. ables and uncertaintiesthat may render the data little more
Attemptingto usethisrathersparsedatasetfor quantita- than a curiosity. For instance, the greater tornado concen-
tive estimatesshouldbe donewith greatcaution.However, tration in Georgia may be influencedby the distribution of
someusefulqualitativeobservations canbe made.Figure1 the post-ReconstructionArmy Signal Corps in Georgia. This
alsocontains theoutlineof the10-4 maximum windspeed 1880-1915 period was largely before the development of
probability for 140mihr-• (mph)(62.6m s-•) fromFujita poured concretefoundations,thus addingmore uncertainty
[1987]. This isopleth is the midpoint of the nominal F-2 to Fujita scaleF-3 or greaterrating standards.In addition,
windspeedrangeand enclosesthe major tornadorisk areas preservationof local newspapersin the south, especially
for the period1916-1985.A roughcomparison of the 1916- Mississippi,was very poor until about 1910. The role of
1985 risk can now be made with the distribution of the worst racismin the documentationof deathsalso may be a more
pre-19!6activity.The purposeof thiscomparison
is to locate importantvariablethan was originally
 GRAZULIS 469

3. 1916-1949 5. Oms•oss

For thisdataperiodthe UC database,derivedaccording The following statementsare opinions, developed during

to Fujita [1987],wasthe startingplace.A localnewspaper the NRC work and the author's previous historical research,
description of everyeventwassoughtout. Thehighsuccess andwere presentat the start of the NSF phaseof the project.
ratein locatingthesedescriptions(over90%)maybedueto They might be consideredby the reader as biasesaffecting
the possibilitythat officiallylisted eventswere thosetorna- the final results of the project.
doeswhich were widely reportedin newspapers.
As with 1. Large numbersof F-2-rankedtornadoesin the N SSFC
1880-1915,tedioussearchingand the inadvertentlocationof and UC data bases(especially for the years 1950-1976)are
new events (while searchingfor informationabout known overrated. Hundreds of tornadoes, having only partially
tornadoes)expandedthe data basefor this period.Fujita unroofedhomes,collapsedold barns,andoverturnedmobile
[1987] lists 3465 F-2 or greatertornadoesfor this period. homeswere incorrectly assignedan F-2 rating.
Grazulis [1991] lists 4028 such events for the 1916-1949 2. Some F scale ratings after 1976 have state-to-state
period.About 200 of thosetornadoesare previouslyunlisted inconsistencies and some lack of appreciationfor engineer-
events. Over 2000 previously unlisted tornadoeswere lo- ing principles.
cated, the majority of which were F-0/F-1 events. 3. The entire research and rating effort should be done
by a singleperson,to maximize the conceptof consistency.
4. 1950-1989 This applies not only to the project but also to current
documentation at NSSFC.
The project's philosophy in reviewing official NSSFC 4. The year 1880is the best startingpoint for the docu-
ratings for this period was somewhatdifferent for the 1950- mentation,as it was Finley's [ 1884]first year of documenting
1970 period than for !971-1989. Virtually all official F scale over 100 tornadoes.
ratings prior to 1970 (pre-Fujita scale) were based on old 5. The turn-of-the-centurydistributionpattern of farms
newspapers and Storm Data descriptions.The ratings for provideda bettertornadointensitydetectiongridthanfarms
that period were made without benefit of on-site surveysor do today. In 1900,there were tens of thousandsof 160-acre
publishedstudies,suchas that of Minor et al. [1977],which (0.6km2) homesteads,
eachwith a barnandhouse.Today,
brought engineeringconcernsthe forefront. mosthavebeenreplaced
by2000-acre
(8.1km2)agribusiness
No official NSSFC ratings were acceptedoutright for the or expandedfamily farms, sometimeswith no houseand an
1950-1970 period. The original studentrating sheetswere absenteeoperator.Thus the distributionof tornadoesduring
obtained from NSSFC and studied. These contained some
the first half of the century may show less population bias
notesfrom newspapersearches.More credencewas placed than that during the later half.
in the officialratingfrom 1971to 1976,but therewas still the 6. Urbanization tends to increase both the number and
obvious lack of concern for engineering principals. Local intensityof significanttornadoesand the numberof reported
newspaperdescriptionswere soughtout for about 7000 weak tornadoes.Short-pathsignificanttornadoesthat would
events from 1950 to 1976.
hit nothingin ruralareascan hit buildingsin urbanizedareas
For the years 1971 to !976 the philosophyof National andbe givena higherF scalerating. Urbanizationincreases
Weather Service (NWS) personnel,or the state climatolo- the number of buildingshit by longer-track significanttorna-
gists,apparentlywasto defaultto F-2 on borderline F-l-F-2 does.The potentialfor longer-tracktornadoesto revealtheir
damage.This wasthe casefor tornadoeswith oddtypesof F-3 or greaterwindsincreasesas moretargetsare placedin
damagethat were not describedin the officialFujita scale theirpath. Weak or poorly formedshearzonetornadoesthat
descriptions.Most of the largelyoverrated1974events(of mighthave goneundocumented in rural areaswill produce
whichthe projectlowered113ratings)standin boldcontrast urban damage and be recorded.
to the tornadoesof April 3-4, 1974,whereFujita'spersonal 7. The farther back in time the work extended, the
ratingswerequitethe opposite.
His F scaleratings
for this greaterthe likelihoodof F scaleratingerrors.Thisis dueto
so-called
superoutbreakareamong themostconservative in great differencesin building practicesand construction
the entire data base. (Both data setsusedFujita'spersonal methodsin times past. A stressof consistencywill result,
ratingsfor thatoutbreak.)Manyof his F-1 ratingsfor that however, in data that is useful in making qualitativejudg-
daywouldprobably begivenanF-2in moststates today.It ments of risks and trends.
is Fujita'sstandardsfor April 3, 1974,the insights
from
Minor etal. [1977],and discussionswith othermeteorolo- 6. FUJITA SCALE STANDARDS
gistsandengineersthatguidedthecreation of theproject's
strict standards. It was apparent,very early in the project,that the Fujita
In the 1976-1989period, engineeringconsiderationsscalestandardsare inadequatefor creatinga consistentdata
slowlyfilteredintotheF scaleratingprocess, anda whole- set. The current standards, as listed in each issue of Storm
salereviewof all ratingswasnot necessary. Problems did Data, focus on house damage,but with lessthan adequate
arisehere,astheywill in thefuture,withthedifficult-to-rate,detail.There are no specificson how to rate differenttypes
short-path,
borderline
F-l/F-2events of house construction. There
thathitnonengineered is no mention of what percent-
structures. ageof a roofneededto be "removed"for an F-2 rating
470 SIGNIFICANT TORNADOES OVER 110 YEARS

40[3
a trivial matter when newspapersmay refer to "unroofed
homes" that only lost shingles).However, lessthan half of
all F-2 tornadoes are rated as such on the basis of house
damage. There is no standardthat clarifiesjust what degree SO0

of damageconstitutesa "destroyed" barn or a "destroyed"

mobile home, so an expanded set of rules of thumb have
been developed to enhance consistency. 2OO

Rather than try to expand the list of the F-2 damage

descriptions,it is useful to expand and clarify the types of
damagethat are not consideredto be F-2 damageand must !oo
be assignedan F-1 rating. The project became largely an
effort to eliminate the F-0/F-1 events. A list of these guide-
lines is presentedbelow.
The destructionof the following items is consideredF-l, 1950 1960 1970 1980 1989
unless an engineeringsite survey is completed:(1) boat
docks, marinas, and boat houses; (2) sheds, even if called Fig. 2. A comparisonof the number ofNSSFC F-2-F-5 tornadoes
"barns"; (3) barns with two or more walls remainingintact; (top hatched area) with project F-2-F-5 tornadoes (bottom hatched
area), 1950-1989.
(4) mobile homes which still have a recognizable shape; (5)
single or isolated trees; (6) chimneys; (7) sheet metal com-
mercial buildings;(8) homes or other buildings swept away
8. KILLER TORNADOES AND DEATHS
while in highway transit; (9) attached garagesand carports;
(10) small chicken or brooder houses; (11) private single- Figures 4, 5, and 6 show that killer tornadoes in most
engine airplanes; (12) athletic field grandstands;(13) tele- categories peaked in the 1920s. Figure 7 shows a gradual
phone poles; (14) Quonset-type storage buildings; (15) decline in the height of the peak killer tornado years since
drive-in movie screens; (16) windmills and oil derricks; (17) 1909. This trend was interrupted by 1974, which was heavily
homes destroyed by falling trees; and (18) conveyor belts or influenced by the remarkable outbreak on April 3 of that
irrigation systems. year. Figures 8 and 9 show the trend in total deaths to be
The following conditions are also considered as F-l: (1) similar to that in killer tornadoes. The last four decades
homes with over 20% of roof remaining; (2) homes "clean- cover the period of NWS watches and warnings. In Figure 4
ly" unroofed (as if not nailed down); (3) cattle lifted or rolled the trend for all killer tornadoes shows no decline in these
along the ground; (4) boats carried into trees; (5) objects last four columns. Figures 5 and 6, for tornadoes killing at
driven into wood; (6) collapsed barns; (7) blown-down least 10 and 50 people respectively, show a definite decline.
unreinforced concrete block walls; (8) unroofed tourist cab- In Figure 6, both single events, during 1960-1979, were in
ins; (9) a house rotated on, or shifted off a foundation, but Mississippi,and both (February 21, 1971:58 killed, 160 path
not unroofed; (10) unroofed fast-food restaurants; and (11) miles (257 km); March 3, 1966:57 killed, 70 path miles (113
barns blown over but not swept away. km)) were probably tornado families with fewer than 50
When this list was applied to NSSFC data (1950-1989) deaths for any single member. The combination of NSSFC
some2567(out of 8273) F-2-F-5 ratingswere foundto belong watches and warnings (since the early 1950s) and the dis-
in the F-1 category, a larger-than-expectednumber.

7. NSSFC-PRoJECT DIFFERENCES

Figure 2 graphs the comparison of NSSFC data and

projectdatafor F-2-F-5tornadoes
for 1950-1989.
Figure3
displays the differences in total events for each year. The
NSSFC curve (top) in Figure 2 shows a steady, rather
dramatic decline in F-2-F-5 tornadoes beginningin 1974,
aftersteadily
highnumbers
since
1953.
Theproject
data 100
(bottom) showsan irregular seriesof peaksand valleysfrom
1950 through 1984. It is concluded here that the NSSFC
decline
isa falseone,having
nothing
whatsoever
todowith 50.....
climate or meteorology. It is the product of early-year
inexperience, lack of engineering considerationsin the F
scaleratingof tornadoes,
anddefaultphilosophy
in oddand 1950 1960 1970 1980 1989
borderline situations. The short-term decline in both data
sets, during the late 1980s,is apparently valid. (That decline Fig. 3. The annual number of differences in F-2-F-5 tornadoes
apparently was reversed in 1990.) between NSSFC and project records,
 GRAZULIS 471

5OO

432

357
400
,...'4•_

.380

20O

1880- 1890- 1900 1910- 1920 1930 1940 1950 1960 1970 19S0
1889 1899 1909 1919 1929 1939 1949 1959 1969 1979 1989
1889 1899 909 919 1929 193'9 •'}4q 959 1969 1979 1989

Fig. 4. Tornadoeswith one or more deathsby decade.

Fig. 6. Tornadoeswith 50 or more deaths by decade.

semination of information through schools and the media chance, of course, that the violence of tornadoes themselves
apparently have done a remarkable job of reducing the peaked in the 1920s and that an undetected decline in
number of multiple-deathkiller tornadoes,despite an in- intensity has resulted in a decline in all death categories.
creasingpopulation. The totals deathsby decade(Figure 8) Another possiblemeteorologicalreasonis a shift of the most
is rather stable, except for the peak in the 1920s,from 1880 violent activity away from vulnerable populated areas.
to 1949. Maintaining the current downward trend through While neither of thesereasonsseemlikely, there is a shredof
another decade may be a difficult challengefor the NSSFC. evidencethat the intensity of major outbreakspeaked in the
Figure 9 shows the annual death tolls. The total has ex- 1916-1949 period. The project uncovered 33 tornado out-
ceeded 300 deaths in a single year 16 times since 1880, but breaksduring !880-1989, with 20 or more significanttorna-
only once since the NSSFC was establishedin Kansas City does each. The count for each period is as follows: 1880-
in 1954. The 1980s were the first decade in over a century to 1915, six outbreaks; 1916-1949, 16 outbreaks: and 1950-
have no single year with 150 or more deaths. 1989,11 outbreaks.The 1916-1949period includedsevenof
The decline in multiple-deathkiller tornadoessince the the 10 worst outbreaks.This 1916-1949peak may be a result
1920smay reflect both a declinein rural populationfollowing of the author's misjudgmentof buildingconstruction(with a
the Great Depressionand an increasein awarenesslevels correspondingincrease in F scale ratings! and/or a more
brought about by the spreadof radio, television,broader favorable distribution of homes and barns in rural areas. This
public education,spotter groups,and finally the watch/ is a highly subjectiveand speculativefield.
warning system.One can also speculatethat the declinein As noted by Doswel! and Burgess[1988], the problem of
Figure8 beginsin 1940sand coincideswith the formationof properanalysisof tornadofamiliesis unresolved.If all of the
organized spotter groups after World War II. There is a pre-1950long-tracktornadoeslisted by the projectas single

65 i

5o
41
759
36
33 ,34
28
26

•8

18[K)-1890
1883 1899 1900- 1919
t909 1910- 1929
1920 19•9
1930 1940
•,•49 1950- 1901•
1959 t960 1•1/0,
lq79 198d
1,98'1
Fig. 7. Annual killer tornadoes,188(Y-1989.
The start of each
Fig. 5. •ornadoes
withtOor moredeaths
bydecade. decade is marked by a heavy
472 SIGNIFICANT TORNADOES OVER 110 YEARS

qooo ooo

3220
15101515
sooo ,

5o0 1428

2o0o 1954
1•?tq
1•I?R
184S 1992
,,1Ft2FI
1412
ooo 1205
1188
1i52 1255
980
i ooo
5oo
509

1880-
1889 1189o
899 1900 ,9,0.
1909 1919 ,920
1929 ,9ao•
1939 1940. 11•50.
1949 959 1960 ,970.
1969 1979 1980.
989

1880-
1889 1890
1899 1900
1909 195•
191 1920
1929 1930
1939 1940
1949 1950
1959 •960 1970
1909 1979 1960
1989

Fig. 8. Total tornado deaths by decade.

Fig. 10. Significanttornadoesby decade.

events were broken properly into families, the number of

killer tornadoesin the three listed categories (Figures 4, 5, (or vice versa) that are interesting. The most intriguing
and 6) would probably increaseor stay the same. This would tornadominimum is that over Fort Worth, Tarrant County,
make the declinein the past 80 years even more pronounced. Texas (and possibly some counties to the west), when
Most killer tornadoesproduceddeaths in small concentrated compared to neighboring Dallas County. The count of all
areas, such as residential areas or clusters or rural homes. NSSFC tornadoes does not indicate a tornado minimum at
For example, the April 24, 1908, tornado (with 143 deaths in Fort Worth. Tarrant County (home of the National Weather
Louisiana and Mississippi)actually may be a family of four Service Southern Region Office) leads Dallas County in all
killer tornadoes, with 24, 64, 33, and 22 deaths each. A tornadoes, 55-53. Prior to 1976, tornadoes were overrated in
similar breakdown of post-1950 events would have little Tarrant County, distorting the NSSFC statistics for F-2-F-5
impact,exceptfor the two eventsin Mississippilisted above. tornadoes. However, project statistics show Dallas County
Some of the increasein the number of significanttornadoes, with a 23 to seven lead in significant tornadoes. For killer
as shownin Figures 10 and 11, may be due to better analysis tornadoes, Dallas leads eight to one and is ahead in deaths 34
of tornado families. to one. Counties to the north and northeast, with a total
population equal to Tarrant County, outnumber it about 50
9. A TALE OF TWO CITIES to one in killer tornadoes. The city of Fort Worth, sitting in
If long-term,small-scalemaxima and minima exist, they what is often called "tornado alley," has had a sizable
are effectivelymaskedby an endlessvariety of population population for over a century without a killer tornado. As
variations and human activities. However, there are a few seenin Figure 12, the one death was in a rural area southeast
areasthat combinetornadominima and populationmaxima of the city. Since 1976, when NSSFC and project data are in

6OO

300

0
18s0 1900 920 1940 1960 1980 880 1900 1•920 1940 1960 1980

Fig. 9. Annual tornado deaths, 1880-1989. The start of each Fig. 11. Annual significant tornadoes, 1880-1989, project count.
decadeis marked by a heavy line. The start of each decade is marked by a heavy
 GRAZULIS 473

Wichita
1
2 7 1
42 o
Archer
Henrietta Montague River
Grayson
1
18 1 1 1 14 3 1 1

o
17 1 73
I 7 1
1 I 110 1
1

Archer 2 1 2
City 9 2
Jack
Wise Denton Coliin Titus

on Yoq•g 1 10 13 I 1
31 o
o
Decatur
11 1
Jacksboro 1 3 2
o
Graham
4 2 2
2 I 19
Camp
Polo Parker TarranT Dallas115 Wood
Stephen•4 Pinto UpshL
24

o F•'•orth
I % Kaufman
ol
Breckenridge
Palø
o Min•
o :ral
Pinto Well Weatherford 2
"---/ I 1
Hood Johnson Ellis 1 17
Eastland Erath
o 1 7
23 o
Eastland ;ranbury
16 Cleburne
101 Henderson
4
4
Stephenville
O2 INavarro

Ch•okee
Fig. 12. A mapof thedeaths
fromeachkillertornado
in north-central
Texas,1880-1989.
Somelargertowns,westof
Fort Worth, have their names listed.

agreement,the significanttotal is Dallas six, Tarrant zero. to tornadoes that did no damage, if they were visually
There is somethingpeculiar here. Either it is a classic impressive.Kansasdoesnot, despitehavingas manyoppor-
example of blind chance sparinga town, or there is a tunities as Colorado to do so. This makes it difficult to
permanent(semipermanent?)
minimumfor strongand vio- comparethe two states and distortsrisk analysismaps.
lent tornadoes west of Dallas. No speculationis madePerhapstwo F scale ratings shouldbe archived,one for
concerning
the meteorology if indeed damageandonet•)r estimatedwindspeed,howevercumber-
behindthisdifference,
there is an identifiable reason. With an atmosphere in some.

chaotic,turbulentmotion,something suchastheprincipleof This matter and others were raised by Doswell and Bur-
strangeattractors[Gieick, 1987]may lead to long-lasting gess[1988]in their examinationof severalunresolvedissues
maxima and minima without a directly identifiablecause. in tornadoclimatology.Some of the concernsdiscussedin
that paper were major obstaclesto the completionof the
10. RECOMMENDATIONS project.Oneof the conclusions reachedby thisauthorwasat
The NSSFC shoulddevelopa comprehensive setof guide- odds with a conclusion reached by Doswell and Burgess.
linesfor the useof the Fujita scale.This, fortunately,seems Theysuggest thattheremay be an advantageto havingmany
to beunderway.Preliminary workby BuntingandMarshall peopleinvolvedin the ratingseffort, assumingthat errors
[1991]on sucha guideusessomeof the basicphilosophy and biases would be more likely to cancel out for the
outlinedin section6. A similarlist was suppliedto themfrom aggregate database.This would producenationalaverages
ourunpublished reportto NSF in 1987.Of equalimportance that are less prone to individualbias. However, national
is that the NSSFC decide whether it has a damage-based averagesmay be little more than a curiosityand of no
data base or an estimatedwind speedbaseddatabase.A practicalusein regionalrisk assessment. The authormain-
clearstatement ofpolicy,backed byinsistence onconsistent tains that biased or not, consistency is the only means by
procedures, should bemadebyNSSFC.Somestates assign which progress is possible.Thus this author also recom-
F-2ratings to nondamaging (orF-i damaging) tornadoes, on mendsthat all ratingsbe reviewedat NSSFC by a single
thebasisof spotter/chaser intuitiveestimates of rotationalindividualwith the experienceandauthorityto overrideany
speed. Forinstance, Colorado routinely applies F-2ratings ratingsmadein an NWS forecast
474 SIGNIFICANT TORNADOES OVER 1I0 YEARS

11. A FINAL NOTE orologist Short Course, Natl. Oceanic and Atmos. Admin./Natl.
Weather Serv., Norman, Okla., Feb. 26-28, 1991.
The large numberof significantN SSFC tornadoesfound Doswell, C. A., III, and D. W. Burgess, Some issues of United
by thisprojectto belessthansignificant
does,at firstglance, States tornado climatology, Mort. Weather Rev., 116, 495-501,
seemexcessive.Sincethe UC data tape hasonly300fewer 1988.
F-2-F-5 tornadoes thanthat of the NSSFC (1950-1979), this
Finley, J.P., Report on the character of six hundred tornadoes,
must mean that the UC data are almost equallyoverrated. Pt'of Pap. VII, Signal Serv., Washington, D.C., I884.
Finley, J.P., State tornado charts, Am. Meteorol. J., 5, 446-476,
That is, indeed,the conclusionreachedby thisauthor.From 501-507,545-551, 1889.
1950-1970,Storm Data usedwords like "destroyed"and Fujita, T. T., Proposedcharacterization of tornadoes and hurricanes
"unroofed" withoutregardto their eventualuse in quanti- by area and intensity, Res. Pap. 91, Satell. Mesometeorol. Res.
tative judgments.The Fujita-trainedstudentsat UC had no Proj., Univ. of Chicago, Chicago, Ill., 1971.
choice but to take Storm Data at its word. Both data bases Fujita, T. T., U.S. Tornadoes,part 1, 70 Year Statistics,report, 122
have interestingand convolutedhistories. The author used pp., Satell. and Mesometeorol.Res. Proj., Univ. of Chicago,
Chicago, Ill., 1987.
every means practicalto maintainconsistency.While hun- Gleick, J., Chaos, Making a New Science, 354 pp., Penguin,
dreds of the NSSFC ratings are obviously overrated,it Middlesex, England, 1987.
would be difficultto make a powerfulobjectivedefensefor Grazulis,T. P., Violent tornadoclimatography,1880-1982,Rep.
most of the 2567 rating changes.It would also be equally NUREG/CR-3670,165pp., U.S. Nucl. Regul. Comm., Washing-
ton, D.C., 1984.
hard to makea strongcasefor restoringany particularevent
Grazulis, T. P., Significant Tornadoes, 1880-1989, vol. I and II, 970
to its originalF-2 rating. There is a huge "gray area" here. pp., Tornado Project, St. Johnsbury,Vt., 1991.
The objectivefrom the beginningwas not the unachievable McDonald, J. R., and R. F. Abbey, Jr., Comparisonof the NSSFC
goal of "correctness."It was the more achievablegoal of and DAPPL tornado data tapes, in Preprints, Eleventh Confer-
"consistency," which would reveal trends and uncover enceon SevereLocal Storms, American MeteorologicalSociety,
Boston, Mass., 1979.
possibilities for future research.
Minor, J. E., J. R. McDonald, and K. C. Mehta, The tornado: an
engineering-orientedperspective, Tech. Memo. ERL NSSL-82,
REFERENCES 192pp., Natl. Oceanicand Atmos. Admin., Boulder, Colo., 1977.
Schaefer, J. T., D. L. Kelly, and R. F. Abbey, A minimum
Bunting,W., and T. Marshall, A resourceguide for conducting assumptiontornado hazard probability model, J. Cl#n. Appl.
damagesurveys,paperpresentedat WarningCoordinationMete- Meteorol., 25, 1934-1945,
 Discussion

K. CRAWFORD, SESSION CHAIR

Oklahoma Climatological Survey

PAPER H 1 (Grazu!is) Yes, I understand that, and everyone should

appreciatethat distinction.There really is no effort to project
Presenter, T. P. Grazulis, private consultant[Grazulis et al., 10 million years. However, it effectively presents the absur-
this volume, advancesin tornado climatology,hazards,and dity of the problem better. There isn't time for me to try to
risk assessmentsince Tornado SymposiumII] explain it. That wind speed probability may be off in some
placesby as much as 80-100 mph from what I believe to be
(C. Anderson, North Carolina State University) At one time,
the true meteorological wind speed, because it's being
I lived in Wisconsin and used to consider tornado climatol-
handled purely statistically.
ogy there. I noticed some very strangereporting, also. There
were a number of counties along the Lake Michigan shore (Twisdale) We recognize that the wind speed estimates are
that showed no reported tornadoes from 1915 to 1950. Did not inherently correct, and that's why we put adjustmentson
you find something like this as well'? those probabilities.

(Grazulis) That's a curious thing, becauseif you look at risk (Grazulis) You need enormous analysis and caution on this.
analysis in Wisconsin, there is a clear minimum along the The details of how to estimate the true wind speed are really
eastern side of the state, except for the last 15 years. There out of my league. I simply look at distributionsand what has
are major shifts in the climatology, filling up these apparent happened.
minima;there are other examplesI can't think of specifically
at the moment. I don't think it's reporting; it appears that {C. Doswell, National Severe Storms Laboratory) It's fasci-
there really were no tornadoes in that area of Wisconsinfor natingto work with this data base, and I think you and I both
that time. We have to be cautious with short data bases. agree that it's a mess. You've talked about Ted Fujita's
Given a map of tornado occurrence, is that a prediction'?I monumental effort to account for a variety of factors in
have yet to see a map that I feel comfortable with as a adjustingthe data, and offered some interestingsuggestions
prediction for the next 20 years. of your own. Do you have any plans to create an adjusted
data base that's consistent with your view of the true hazard
(,L. Twisdale, AppliedResearchAssociates)I'd like to make distribution ?
one commenton sayingthat risk assessments are 10-million-
year projections.That's an incorrectinterpretationof prob- (Grazulis) If three or four very small adjustmentscould be
ability of exceedance;what you're trying to do is estimate made to the Kansas City data base, you could begin to create
the probabilityof wind speedexceedance overthe designlife a data basethat would really work. The problem is that Allan
of a plant, whichmightbe anywherefrom 1 to 30 years.It's Pearson created that data base without realizing what the
like throwing darts at a board; you're consideringa popula- problemswould be 20 and 30 years later. It's a fossilof the
tion of rare events over an area and trying to estimatethe early ideas. There are several things I'd change. Number
chances of a certain target being hit. There's a lot of one, I think there need to be two F-scale estimates; one for
randomness
and spatialvariabilityin that. The projectionis the actual damage done, and the other is the subjective
not whether or not a tornado is goingto occur in the next 10 adjustmentfor the intensity. The secondadjustment:did it
million years, it's the chancesof a structurebeinghit. hit a building'?Another: was it or was it not spawnedby a
supercell'?We've discussed the famous Sunray, Texas,
The Tornado:Its Structure,Dynamics,Prediction,and Hazards. eventand it wasgiven an F-4 ratingeven thoughit neverhit
Geophysical Monograph 79
Thispaperis not subjectto U.S. copyright.Published
in 1993by the anything.It's very hard to compareColoradoand Kansas
American GeophysicalUnion. statistics,becausein Colorado, the rating is very subjective,
476 DISCUSSION

while in Kansas, the rating is strictly based on damage.The (Xu) Yes, often tornadoes occur in front of typhoons. This is
data base needsadjustment. discussedin my paper.

(Allan Pearson,private consultant)Having gonethroughthe (Petersen) Are most of the tornadoes you showed associated
North Dakota data base, if it weren't for outdoorprivies, no with typhoons, or just many of them?
one would know what events occurred there. [laughter]
There is absolutely nothing in those publications;it's the (Xu) From July to September, typhoons often occur and
best we could do. I've told you this before, but I don't think tornadoes often are related to those typhoons. However,
there's any way to go back and adjust those past events. tornadoes also occur with convergence lines, squall lines,
How people use the data is the problem. and in the warm sector of extratropical cyclones.

(Grazulis) I'm not saying there's any way to go back and (D. Burgess,National Severe Storms Laboratory) The radar
fix it up. The very sophisticatedclimatologicaldata adjust- echoesyou showed seemed to be a supercell thunderstorm.
ment modelsbuilt by Ted [Fujita] and Joe [Schaefer]and Jim Was there large hail associated with that storm?
[MacDonald] are like building Rolls-Royces to carry manure (Xu) When a tornado occurs, the associated hailstorms often
by the teaspoonful;my work has been confinedto compost- are very strong.More often, the storm is a hailstorm without
ing [lauughter]. You just can't go back and fix it; you can a tornado. But if a tornado occurs, usually the hail event is
make it more useful, but you can't improve it very much and significant.
you've got to be very careful how you use the data.
(A. Court, California State University, Northridge) In this
country and in western Europe, the tornado has been under-
PAPER H2
stoodto be a rotatingstormfor only about 150years. How old
is the understandingof the tornadoin China?Is that knowledge
Presenter, J. Dessens, Universit0, Paul Sabatier [Desseus relativelyold, or has it been effectivelyintroducedby Western
and Sno,,', this volume, A comparativedescriptionof torna- meteorologistsas a separatekind of storm?
does in France and the United States]
(Xu) In our ancient times, we have recorded the events for a
(T. Grazulis, private consultant) Did you only plot the F1 long time, but only reported the situation for damage, and
through F5s? not very accurately.

(Dessens) Yes, only F2 through F5s. (Court) It was recognized that it turned?

(Grazulis) But there were some F0s and Fls? (Xu) Yes, it was known to be rotating.

(Dessens) Yes, certainly, we have many more F0s. We have PAPER H4

perhaps 100 per year, but F0s cause little or no damage.
Presenter, M. Leduc, (M. J. Newark, Tornado incidence in
(Grazulis) The tornado distribution in France appearsto be
Canadian cities, not in this volume)
similar to New York and Pennsylvania.
(T. Grazulis, private consultant) It's so satisfyingto have
(Dessens) Yes, I agree.
someoneagree with me. [laughter]Did you do any break-
(R. Petersen, Texas Tech University) Are there any attempts down or distributions by F scales?
in France or adjacent countries to warn people for these
(Leduc) i know there is, but I don't have the figures.
tornadoes, or do they come as a complete surprise to
people? (Grazulis) Well, Edmonton comes to mind. Are there any
other examples out there?
(Dessens) No attempt is made to make tornado warnings in
France. For hail, yes, but not for tornadoes. After a tornado (Leduc) For example, in Ontario, we've had six F3 or
the Meteorological Offices study the situation, but only greater outbreaks since 1979. In southern Ontario, we've
afterward. beentalkingaboutone F3 or greaterevent every 4 years, but
the statisticsthe last 10 years now indicate it might be more
like every 1 or 2 years. Certainly, the numbersare goingup
PAPER H3
as we learn more about it.

Presenter, Z. Xu, Chinese Academy of Meteorological Sci-

(J. Golden, National Oceanic and Atmospheric Administra-
ences [Xu et al., this volume, Tornadoes of China]
tion, Officeof the Chief Scientist)There' s a host of intriguing
thingsin your talk. Would you say a little more about that
(R. Petersen, Texas Tech University) China has many ty- frequencymaximum map. I gather that the Alberta maxi-
phoons. It there any evidence that any of these tornadoes mum is fairly new, but what about that "intermountain"
have occurredwith typhoons coming on land'? maximum? ! know that area is pretty
 CRAWFORD 477

{Leduc) I'm afraid I can't help you very much on that one. (J. Golden, National Oceanic and Atmospheric Administra-
This isn't my paper, after all. tion, Office of the Chief Scientist) You were joking that it
took days to confirm that the Vancouver event was a
(Golden) Governments tend to react to big events, some- tornado...
times overreacting. I understand your government is con-
{Miller) No, I wasn't! [laughter]It took them 2 days before
templating the deployment of Doppler radars to protect your
biggest cities. In fact, I understand that you've recently they officially came out and said it was a tornado.
Dopplerized the C-band radar in Edmonton. What are your (Golden)Where I live, in the Washington,D.C., metroarea,
plans in this regard? we had an outbreak of "something" this year, during the
(Leduc) Well, of course, the plans are subject to budget winter. At the time, the damage reports indicated downburst
constraints. In fact, there just was a fairly significantbudget damage, although people reported possible funnels and
cut in March. There was a plan last year that we would have touchdowns. It was at least 2 days before the staff at the
four new Doppler radars by 1996, with eventually the forecastofficegotout and did their surveysandconfirmedat
network of 20 radars being Dopplerized. The Edmonton least two of them were, indeed, tornadoes. Small ones, and
Doppler will be up and running this summer, and we have in fact one of them touched down closer to my house than
the Doppler radar north of Toronto. The radar at McGill is, I've ever experiencedafter 6 years of living here in Okla-
I believe, about to be Dopplerized. homa...

(D. Burgess,National Severe Storms Laboratory) When I {Chair) Joe, do you have a point? [laughter]
was in Kananaskis, I remember seeinga videotape of a very
impressive tornado, but on the audio, there was radio !,Golden)There are sometools you can use. The peopleat
communicationsuggesting an amateurradio network, storm SELS [Severe Local Storms] have done a remarkablejob;
spotters,and a successful
warningfor that storm.Couldyou Jack Hales has been very successfulwith forecastingtorna-
comment on your warning system?Was that just an excep- does on the west coast, particularly California. I'd also
tion, or was it organizedin many areas? suggestthat rapid scansatellitedata will be of significant
help to you.
(Leduc) Well, I'm the severe weather meteorologistat the
OntarioWeather Centre, so our programis excellent.[laugh-
PAPER H7
ter]. Speakingfor Ontarioespecially,in the past2 or 3 years,
we've been organizingthe ham radio people into spotter
networks. We had a terrific successon August 28, 1990. We Presenter, R. L. Livingston, National Weather Service.
were fortunateto have a spotterseethe first funneldevelop KansasCity [SchaeJ•'r et al., this volume,The stabilityof
and had tornadowarningsout for over an hour beforethat climatologicaltornado data]
storm did the worst of its damage.
(D. Burgess,NationalSevereStormsLaboratory)I can't
(J. Anderson,Prairie Weather Centre) I think I can help out helpbut comment,sinceyou didn't, aboutthe trend.You
with the tornadohe's referringto: that was the Saskatoon talked about the recent decreasingtrend in tornadoes,but
tornado.The warningwent out just as it was reachingthe you didn't showlast year, which I believehad the greatest
rope stage.We don't really have a well-established
ham number of tornadoes in the data base.
radio weather network. I went chasingtwo nightsagohere in
Oklahoma,and I was really impressedby that network. I (Livingston)That's probably true. I think there ,,*,'asa
thinkmaybewe cantry andencourage someof ourpeopleto numberof yearsin thetimetrendthatshowedspikes.I don't
do it for us. The year beforelast was the first time we got think that's unusual at all. I think you need to look at a
someof our youngermeteorologists to go out and chase longertimeperiodbefiore
youinferthetrend.Yearslike 1973
storms;we haven't seenany tornadoesyet. Rightnow we showedsuchspikesand ! believethat's a normalcharacter-
relyona pre-established
networkof watchers,
usuallyfi'om istic.
government and provincialagencies,and we can phone
them.It's actuallymostunhelpfulin mostcases;we typi- {C. Anderson,North Carolina State University)Do you
callyhaveaboutthreepeoplein a 50 km squareareaand haveany explanationfor the Missouriminimum?It appears
theyreportlooking outthewindow andseeingit'sdarkand over and over gain.
stormy outside. (Livingston)
We'vebeenscratching ourheadsfor a number
of years over that. Tom [Grazulis]talked aboutit this
PAPER H6 morning;
he'slookedat it andcan'tcomeupwithanything.
The Ozark mountainsmay divert the winds. Population
Presenter,G. R. Miller, NationalWeatherServiceForecast density
is low,butthepopulation
in otherpartsof Missouri
Office[Millet',this volume,Oregontornadoes: More fact islowandyettheydon'tshowthatpronounced a minimum.
than fiction] It's a real
478 DISCUSSION

PAPER H8 through a valley and hits a house on a cliff, the "obser-

vation point" moves from a flat surface into a higher
Presenter,T. P. Grazulis,privateconsultant[Grazldis,this wind speed regime in the tornado. Can you comment
volume,A 110-yearperspectiveof significant tornadoes] on this?
(S. Vasiloff,NationalSevereStormsLaboratory)I'd like (Grazulis)Yes, I'd like to avoidthe idea! [laughter]Topog-
to offer the prospectof terrain normalization.In photo- raphy doesn't alter the frequency of tornadoes, but wind
grammetry, the maximum windspeeds in tornadoes are speedmeasurements certainlyareinfluencedby topography.
tens of meters above the surface. If a tornado moves It's another complication,
 Aerial Survey and Photographyof Tornado and Microburst Damage
T. T. FUJITA

The Universityof Chicago, Chicago, Illinois 60637

B. E. SMITH

National Severe Storms Forecast Center, Kansas City, Missouri 64106

1. INTRODUCTION circular marks, left behind by the North Platte Valley

tornado of June 27, 1955, were reported by Van Tassel
The tornado as defined in the Glossa•3, of Meteorology
[1955]. He assumedthat the grey circles on plowed fields
[Huschke, 1959] is "a violent rotating column of air, pendant
from a cumulonimbuscloud, and nearly always observable (Figure 2) were producedby a singleobject caughtin the
tornado funnel, computing216 m/s rotational speed of the
as a funnel cloud or tuba." In reality, however, no funnel
cloud can be confirmed in blinding rain or during a darkobject.Similargroundmarkswere photographed in the wake
night.Furthermore,a well-definedfunnelon thegrounddoes of the Shelby,Iowa, tornadoof May 5, 1964,and reportedby
not always leave behind a continuous damage swath pro- Prosset [1964]. These marks gave him an impressionthat an
ducedby a singlevortex travelingon the ground. enormousvacuum cleaner had swept the ground clean of
In explaining the break in a vortex swath, the terms vegetation,loosesoil, and other movableobjects.
skippingand lifting were used frequently,implyingthat a A major advancein the interpretationof the circular/
tornado funnel intensifies or weakens within a very shortcycloidalmarkswas madeby takingzoomphotosfrom 200
distance.Duringposttornadointerviewswe oftenhear"the m abovegroundlevel (AGL) and visitingthe siteson the
tornadoleveled my neighbor'shousebut it skippedover my pathsof the Palm Sundaytornadoesof April 11, 1965,and
house." In the wake of the Palm Sundaytornadoesof April the Barrington,Illinois, tornadoof April 21, 1967.As evi-
11, 1965,Fujita and his associatesconductedtheir coordi- dencedin Figure 3, a circular mark was neither a scratch
natedaerialphotographyover the vastareasof the northern mark nor a band of cleaned-upbare ground. Instead, it was
U.S. Midwest, becomingsuspiciousthat a tornadodoesnot a band of debrisdepositconsistingof short piecesof corn
skip or lift within short distances,but rather its wind crops, dry leaves, chicken feathers, etc. The maximum
structure is very complicated. heightof the depositwaslessthan5 to 10cm. In explaining
The objectiveof the aerialsurvey/photography
in 1965by the mechanismof the debris band, Ft(/ita [1971] proposed
the Fujitagroupat the Universityof Chicagowasto deter- the conceptof a suctionvortex in the tornado(Figure4).
mine multiscale airflows in and around tornado funnels and The diameter of a suction vortex is at least I order of
to identifynontornadicdamagingwindsinducedby severe magnitude smallerthanthat of the parenttornado.By virtue
thunderstorms.During the 27 years sincethen, over 300 of itsspinningmotionandsmalldiameter,thevortexgathers
damageswathshavebeenflownandmappedphotogrammet- up near-ground debristowardits rotationaxis,but it failsto
rically (Figure 1). A total numberof 30,000aerialphoto- pickup the debrison the groundat the centerof rotation,
graphsweretakenfromlow-flyingaircraft,mostlyCessna. leavingbehinda narrowband of debrisdepositalongthe
path of the vortex center.
2. DETERMINATIONOF MULTISCALEAIRFLOWSOF Becausethe shapeof the cycloidalmark is a simple
TORNADOES
functionof the velocityratio, rotationalvelocity V divided
Althoughthe newsmediahad taken numerous aerial by thetranslationalvelocityU, FujitaetaI. [ 1970]generated
the firstaerialphotosof well-definedtheshapes
photosof structures, of thegroundmarkby changing thevelocityratio
from I to 10(Figure5). No loopwill formwhenthe velocity
The Tornado:Its Structure,Dynamics,Prediction,and Hazards.
ratio is 1.00, but a suctionvortex staysmomentarilyat one
GeophysicalMonograph79
Copyright1993by theAmericanGeophysical
Union. spot,creatinga steppingspot(Figure6). ,ks the velocity
480 TORNADO AND MICROBURST DAMAGE SURVEYS

IiO IO5 IOO

Edmonton,
1•7-
7-31 Aerial Survey of Tornadoes by the Fujita Group

' CANADA
/ During the 27-year Period,

Ted Fujita, Duane Stiegler, Greg Forbes, Roger Wakirnoto, Brian Smith,
,Jim Partacz, Ed Pearl, Tom Umenhofer, ,Jaime Teeson, Brian Waranauskas,
Pete McGurk, Peter Black, Bernie Ginsberg, Cindy Pealea
1965-1991

95 90 85 80 70

St Molo, 77-6-18

CANADA

7-18 ,
Teton-Yellowstone, Ei7-7- 20 polls. - ,•,•'Bloomer

77-6-5
Horlcon
Windso?Locks

76-6 - 13
Central Iowa
80-6-3
Grand Island ß

•92-6-12 %* / 65- I,PALM

•,Fujlta'$:irstTornado

74 - 8 - 30
GreatBe•nd
9O -3-13
Hesston
88- I I- 28

84 -3-28

79-4-10
•Red
70-5-11
Lubbock

77-4 -4

Atlantic

83-5-20
BO- B- I0 " Houston
SonMarcos,,. .•. =

73-

MEXICO
Gulf of Mexico

I
Fig. I. Tracks of tornadoes surveyed by the Fujita group during the 27-year period 1965-1991. The first aerial
photography was conducted immediately after the Palm Sunday tornadoes of April II, 1965.

ratio increases, the size of the loop increases (Figure 7),

reaching a near-circular loop when the ratio approaches 10
(Figure 8).
The maximum horizontal wind speed inside an orbiting
suction vortex is the sum of U, V, and S, the spinning
velocity of the suction vortex. A strong suction vortex in a
residential area could induce a one- to two-house-wide swath
in which housescouldbe wiped off their foundations(Figure ~ •

9). On the contrary, several "lucky" houses located be-

tween intersecting paths of multiple suction vortices could
be left untouched(Figure 10). These damagepatternscannot
be explained by the so-called skipping phenomenon of a
tornado. Threatened by such a tornado, one shouldnot open
windows because there is no way of guessingthe direction of
oncoming suction vortices. At this point, the evidence of
aerial surveys did alter one of the traditional tornado safety
rules.
Fig. 2. An aerial photo of the circular ground mark assumedto be
A large number of aerial photos showed the existence of the scratch mark by a single object caught in the tornado funnel.
cycloidal marks in the swaths of many large-core tornddoes. From Van Tassel
 FUJITA AND SMITH 481

the location where the vortex center had existed momen-

tarily.
We also observed the path of an isolated vortex mark
suggestinga single-loopmotion of the suction vortex (Figure
15). An interesting vortex signature is the path of twin
vortices which traveled side by side while rotating slowly
around their common center (.Figure 16). Another remark-
able aerial photo shows a curved path v, ith five intensifica-
tion spots along the centerline {Figure 17). The picture also
shows that the initial vortex disappeared, being taken over
by the new vortex which flattened the corn crop along its
path. This picture evidences the rapidly changing nature of
an orbiting vortex which could cause unexpected damage.
Apparently, the smaller the vortex, the stronger the ver-
tical winds relative to the horizontal winds around a small
vortex. Figure 18 shows the corn crop pushed over by the
Hobart, Indiana, tornado of June 30, 1977. A telephoto view
of the strong shear zone reveals the existence of several tin5
vortices, 1 m to 2 m in the core diameter (Figure 19), in
which several corn plants were pulled off the ground. Con-
vergence inside the core of an axisymmetric vortex is
approximated by
Conv= u/D

where tt denotes the inflow' velocity and D denotes the core

diameter.Whenconvergence
is 2 s-•, a 2 m/sverticalwind
is expected at I m AGL, and a 4 m/s vertical wind is
expected at 2 m AGL. Vertical winds of these magnitudes
would be able to pull loosened 5oung plants out of the
ground.
One of the best examples of a small tornado with domi-
Fig. 3. (Top) Aerial photo of the cycloidal mark of the Barrington,
nantly vertical winds just above the ground is seen in a video
Illinois, tornado of April 21, 1967. (Bottom) The cycloidal mark sequence of the Minneapolis tornado of July 18, 1986, which
photographed on the ground. Both photos by Ted Fujita. was taken from a low-flying helicopter. Fujita and Stiegler
[1986] pointed out that a tree in the field caught by a small
tornado funnel which was 3 m in diameter on the ground
tilted for about I s before it was blown down when the funnel
Their frequencies far exceeded our initial expectation. New
ertheless, pictures of tornadoes showing suction vortex moved away from the tree (Figure 20).
A small-core tornado less than 10 m in diameter, east of
funnels had been very rare until the Jumbo Tornado Out-
break of April 3-4, 1974. Since then, a large number of Denver, Colorado, on June 30, 1987, was investigated by
multiple-vortex (suction vortex) pictures (Figure 11) have Wakimoto and Wilson [1989] using both aerial and ground
become available from various parts of the United States. photographs. In spite of the herringbone pattern of damage
These pictures, along with cycloidal marks, were analyzed due to the storm motion, their photographs suggest the
by Fujita et al. [1976], Agee et al. [1975, 1977], and many existence of an appreciable inflow into the small core,
others. implying a strong rising motion inside the small core.
The library of the tornado data collected by the Fujita
group during the past 30 years indicates that tornadoes in
3. EVIDENCE OF SUCTION VORTICES
general are more complicated than earlier conceived. It is
Wind effects of suction vortices on the ground can be often very difficult to distinguish a suction vortex from its
photographedfrom a low-flying aircraft. Their appearances parent tornado (Figure 21). Furthermore, their appearance
vary with the scattering angle of the sunlight, appearing and structure keep changing very rapidly within a matter of
either relatively darker or lighter as an aircraft circlesaround seconds (Figure 22).
the target. As has been well known, tracks of orbiting During the 1976 Symposium on Tornadoes at Texas Tech
suction vortices appear as a group of cycloidal curves University, it was noticed that there was a basic disagree-
(Figure 12). On the other hand, a stationary suction vortex ment between meteorologists and structural engineers on the
leaves behind a pattern of high winds indicatingthe exist- mechanism of structural damage by tornado winds. Most
ence of either a small (Figure 13) or a large •,Figure14) eye at engineers, at that time, approximated tornado winds
482 TORNADO AND MICROBURST DAMAGE SURVEYS

MODEL OF TORNADO WITH MULTIPLE SUCTION VORTICES

co. .:'..' ß /

BO0•- -•e•'-•i.-
. ß • • ::-'.
• ':'' '-"•'-':'"-.:E.'
•" •-
K '.•.....
.../ /• ".••••;./ / • .• • • T • .•:• ,7 '"•••"•77 VORTEX

•. - .• • ""•""

-"•'.....
'.........
,.•,•,•.,...•.••"'•' '• .... V+ T

FUJITA, 1971

Fig. 4. A model of a tornado with three suction vortices orbiting around the core of the parent tornado. From Fufita
[1971].

straight-line winds. This assumption is valid when vertical write a joint paper [Fuji/a and Byers, 1977]. Shortly there-
winds are negligibly smaller than horizontal winds, such as after, Fernando Caracena applied the downburst concept to
in the case of hurricanes and downbursts. Since the near- the probable cause of the Continental Airlines Flight 426
ground convergence is approximated by the inflow velocity accident (takeoff) on August 7, 1975, at the Stapleton,
divided by the vortex diameter, the straight-line wind as- Denver Airport. The results of the joint research on three
sumption becomes invalid for most small vortices such as aircraft accidents is found in the work of Fujita and Cat-
small tornado, suction vortex, dust devil, etc. A structure in acena [ 19771.
such a small but intense vortex could be rapidly torn apart Fujita had great confidence in the downburst concept,
vertically under high-speed vertical winds just above the backed by numerous aerial photos of the starburst damage;
ground (Figure 23). nevertheless, a large number of nonbelievers expressed their
controversial views, summarized by West [1979]. Most
4. MICROBURST,INDUCEROF NONTORNADICDAMAGING meteorologists who expressed strong opposition probably
WINDS did not have the opportunity to fly over areas of tornado
Fujita's aerial survey and photography of the Jumbo damage. Since then, Fujita has trained his group and initi-
Outbreak Tornadoes of April 3-4, 1974, played an important ated extensive downburst-hunting flights.
role in developing his concept of the downburst. After the The training and flights have been very successful, estab-
tornadoes, when Fujita was circling over an area of reported lishingthe existence of multiscale airflows [Fujita and Waki-
tornado damage, he found a diverging pattern of uprooted moto, 1981]. By the end of the 1970s, solid evidence of
downburst winds and their horizontal scales was estab-
trees (Figure 24); thereupon he reached the conclusion that
the damage was caused by a strong downdraft as it impacted lished, on the basis of aerial photos and the National Center
on the tree-covered ground. for Atmospheric Research's Doppler radars operated during
While investigating the Eastern Airlines Flight 66 accident the Northern Illinois Meteorological Research on Down-
(landing) on June 24, 1975, at John F. Kennedy Airport in burst (NIMROD), the landmark experiment which ended
New York, Fujita [1976] attempted to apply his downburst most controversies that prevailed at that time. At the termi-
concept in explaining the strong tailwind and downwind nation of the experiment, the downburst concept was sub-
shears encountered simultaneously by the accident aircraft. divided into microburst (<4-km horizontal size) and mac-
Horace R. Byers, Fujita's mentor professor, was the first roburst (>4 kin).
person who supported his downburst concept, agreeing to An aerial photo of a large microburst (Figure 25) shows
 FUJITA AND SMITH 483

Fig. 7. Cycloidal ground marks left behind by suction vortices

with velocity ratio between 2 and 3. Photo by Ted Fujita after the
West Lafayette, Indiana, tornado of March 20, 1976.

extensive area of diverging xvindswhich blew down numer-

ous corn plants. Frequently, a small microburst touches
down with a sharp boundary of the wind speed increase from
lessthan 25 m/s to 40 m/s within a 5- to 10-m distance (Figure
26). The estimated divergence at the boundary should reach
1.5 to 3.0 s-l, suggesting
that the parentdowndraftde-
scended to the treetop height without weakening signifi-
cantly. A swath of high winds which deflected off of the
sloping roof of a farm building in Indiana (Figure 27) also
suggeststhat a downdraft descended to the rooftop height.
Fig. 5. Geometric path of the center of a suction vortex computed An extremely small microburst may be only 30 to 50 m wide
bv changing the velocity ratio from I to 10. From Fujita et al. [ 1970].
and 100 to 300 m long, with an appearance of a rush of
diverging jet (Figure 28). The parent downward current

Fig. 6. Steppingspots(West Lafayette, Indiana, tornddOof March Fig. 8. Near-circular cycloidal marks of the Goessel, Kansas,
20, 1976) where orbiting suction vortices pause momentarilywhen tornado (F5) of March 13, 1960, as the tornado was traveling at 20
the velocity ratio is 1.0. Refer to Figure 5. Photo by Ted Fujita. m/s. Aerial photo by Duane
484 TORNADO AND MICROBURST DAMAGE SURVEYS

'

Fig. !I. Six suction vortices inside the Wichita Falls, Texas,
tornado of April 10, 1979. Three vortices at the far side are those
forming in the inflow region of the tornado airflow Copyrighted
photo by Floyd Styles.

Fig. 9. (Top) An arc of a suctionvortex track left in the residential

section by the Wichita Falls, Texas, tornado of April 10, 1979.
(Bottom) An enlargement of the boxed area. Photo by Ted Fujita.
This damage is similar to the Lubbock, Texas, tornado case re-
ported by Fltjita 11970].

Fig. 12. Typical cycioidal ground marks of the suction vortices

orbiting around the core of a traveling tornado. Magnet, Nebraska,
tornado of May 6, 1975. From F•dita [1981].

Fig. 10. One-, two-, and three-housegroupsleft untouchedby the

Wichita Falls, Texas, tornado of April 10, 1979, becausethey were Fig. 13. Small eye of,t stationary suction vortex inside the Mat-
located between a number of intersecting tracks of suction vortices toon Lake, Illinois, tornado of August 21, 1977, made visible by the
Photo by Ted Fujita. pattern of blown down corn crops. From Fujila
 FUJITA AND SMITH 485

Fig. 17. Paths of two suction vortices which were joined smoothly
Fig. 14. Large eye of a stationary suction vortex inside the
as the first vortex weakened and w•ts taken over by the second
Bloomer, Wisconsin,tornadoof July 30, 1977. From Fujita [1978]. vortex. Bright dots along the centerline of the first vortex denote
successiveintensifications. Photo by Ted Fujita after the Bloomer
Wisconsin, tornado of July 30, 1977.
, . ß •'•,_' 'z..•t.•-,-• •_..:.• -...••d. '• '-' • •

•, 5:.•;-•"•;•,' ';•'•E'
;""•......
':'•'•'
•' '•";'•t•"'
•'•"•'•
ß' .•. ,. • .. •,,,• ,.,-• &,•',.% • '. x, - . •'.% ,

%x';•'•' .•'•,.,, .,, ,,-,•..... . ,. ,r• • .,.•.• .•. •., .•..

....... _.•...:•-• ......... •:-,--..•
ß •, •t.• M

Fig. 15. A single-loop motion of a suction vortex in the Bloomer Fig. 18. Corn crop• pushed over by the Hoba•, Indiana, tornado
Wisconsin, tornado of July 30, 1977. Photo by Greg Forbes. of June ]0, 1977, which moved •rom right (west) to left (east) across
the picture. Photo by Ted •ujita.

..

• ,

,,,,

Fig. 19. Telephoto view of the strong shear zone in Figure 18,
Fig. 16. Two parallel tracks of twin vortices inside the Sandwich, showinga tiny suction vortex which pulled several corn crops off the
Illinois, tornado of June 30, 1977. Photo by Ted Fujita. ground. Photo by Ted
486 TORNADO AND MICROBURST DAMAGE SURVEYS

O.Osec

0.3sec
Fig. 21. Suction vortices and their parent tornado amalgamated
into a complex system of vortices.

O.Osec

0.6sec

2.0sec

0.9sec

Fig. 20. A tree trying to stand up in the small funnel of the Fig. 22. Two frames of the movie in Figure 21. These frames are
Minneapolis. Minnesota, tornado of July 18, 1986. Four selected 2 s apart, showingthe rapid changeof the vortex system. Figures 21
frames of the video taken from the helicopter of KARE TV. and 22 were enlargedfrom the movie of the first Lornira, Wisconsin,
Courtesy of Paul Douglas, Chief Meteorologist, News !l. tornado of April 21, 1974. Courtesy of Larry
 FUJITA AND SMITH 487

Fig. 23. At 2:05 P.M. PST on February 19, 1980, a small funnel cloud moved over the Fresno, California, Airport.
Although the funnel was not on the ground, the roof of an airport building was blown upward and broken into pieces,
suggestingthe existenceof strong vertical winds beneaththe funnel cloud. The damagepath extended fi-om the airport
into the residential areas of Fresno, California. Courtesy of Peter Stommel.

Fig. 24. The diverging pattern of uprooted trees photographed

near Beckley, West Virginia, where tornadodamageby oneof the
JumboOutbreak Tornadoes of April 3, 1974, had been reported.We
believe that this starburst damage was located at the root of the Fig. 25. The Danville, Illinois, microburst of September 30, 1977,
downdraft which induced a microburst. Photo by Ted Fujita. which blew down corn plants in a large area. From Fujita
488 TORNADO AND MICROBURST DAMAGE SURVEYS

Fig. 26. A small but intensemicroburstfoundnear the Cornell, Wisconsin,tornadoof July 30, 1977.The 100-m-wide
and 130-m-longdamagearea was located on Brunet Island in the Chippewa River. Photo by Ted Fujita.

Fig. 27. A microburst airflow deflected by a tin roof during the Fig. 28. A forest in northern Wisconsinblown down by a rush of
downburst storm of September 30, 1977. Photo by Ted Fujita in diverging winds embedded inside the large downburst of July 4,
Kingman, Indiana. From Fujita [1978]. 1977. Photo by Ted
 FUJITA AND SMITH 489

Fig. 29. A photographof a microburst outflow with a ring vortex along the spreading edge. Photo taken on July 1,
1978, from near Wichita, Kansas. Courtesy of Mike Smith of Weather Data Inc

Fig. 30. A traceof mesocyclone

windswithlargeradiiof curvaturephotographed
nearthe touchdown
locationof the
Sadorus,Illinois, tornado of March 20, 1976,investigatedby Fujita et al. [1976]. Are such winds the precursorof a
tornado touchdown'?Photo by Ted
490 TORNADO AND MICROBURST DAMAGE SURVEYS

I I Helena..
I 95•00, WINDSTORMS ON 2 MAY 1979 km
36030'-
0 5 I0 27 30 4•0 5?
',¾', .'"'..."i .... • ....... , ....
Cleo
Springs
Tornado(FO)
Lahorna
Tornado(F4) o 5 I0 2•0 30miles

•
15õ•C$T
•
0....
to
•••••'•Mesocycione
•632
'.• v•:,ENI I 97g00
,
- --' (F2
•'75 wau•o•,•Waukomis
Downburst
Or/onto
Tornado
•F•
....
: '••o• Ames
36ø15L

First
Marshall
Tornado
(F•)• •Second
Marshall
Tornado
(F2)
1745CST
Morsholl Orlondo STILLWATER
Longdole Okeene
Hermessoy
I•r SecondMuirall Downburst

-36ø00 '
First
Muirall
D....
Downburst
"••
5 own
--
c......t • •o.•,,o.
Coyle
Twisting
Dowo•u,rst
Guthrie

Watonga Kinghsher
98o00, 97000
'

Fig.31. TheCoyle,
Oklahoma,
twisting
downburst
ofMay2,1979,
extending
from
Mullhall
toCoyle
tothesouth
of
Stillwater,
Oklahoma.
Thecenter
ofthemesocyclone,
whichdidnotspawn
atornado,
moved
along
thenorth
edge
of
the twisting downburst.

shouldhavereached verycloseto thetopof theforestbefore photography,

and mapping. Documented tornadoes with
divergingviolently. nearby
microburstswere the Canton, Illinois,
tornadoes of
Microburstsin progresshave been photographed by July23, 1975;theEarlville,Illinois,tornadoes of June30,
many,but themostdramaticsequence of photographs was 1977;andtheMattoonLake, Illinois,tornadoof August21,
takenby Mike Smithon July 1, 1978(Figure29), and 1977.Thereafter,Forbesand Wakimoto[1983]investigated
reportedby Fujita [ 1985]. theSpringfield,
Illinois,areatornadoes
of August6, 1977,
revealing
that7 out of 18 tornadoes
mapped fromthe air
5. MESOCYCLONE
ANDTWISTINGDOWNBURST
WINDSON werelocatedon the left (cyclonicshear)sideof microbursts.
THE GROUND
The WindsorLocks tornadoof October3, 1979,surveyed
by RogerWakimoto, DuaneStiegler,andPeteMcGurkof
Duringtheoverflightsof anareaof winddamage,weoften theFujitagroupwasassociated witheightmicrobursts,
all
foundtracesof twistingdownburst windswithlargeradiiof locatedontheright-hand
sideof the30-km-long,
F4 tornado
curvature. These winds are often located in the areas of which moved from south to north acrossthe Massachusetts-
eitherpretouchdown or postliftoff
of tornadoes,suggestingConnecticutstateline (Figure 32). The Teton-Yellowstone
thattheyareprobablyinducedby theparentmesocycloncstornadoof July 21, 1987,surveyedby BrianSmith,Jim
of tornadoes (Figure 30). Partacz,and BradleyChurchillwas analyzedin detailby
One of the mostsignificanttwistingdownbursts was the
Fujita [1989],revealingthat therewere 72 microbursts
Coyle,Oklahoma, twistingdownburst of May 2, 1979.Its locatedmostlyontheright-hand sideof the40-km-longpath
areawasapproximately 10kmwideand30kmlong(Figure of the F4 tornado.
31), indicatingthat the surfacewindswere inducedby a The interactionbetween a tornado and a nearby mi-
travelingmesocyclone whichdidnotspawna tornado.So croburstwasfirstevidencedwhiletakingaerialphotosof the
far, we havenotresearched themesocyclone windson the
ground.However,an intensive
pathof the Rainsville,
effortin searchof large Figure33shows
Indiana,tornadoof April3, 1974.
radiusof curvaturewindson the groundwill be importantfor thatthecourse of thetornadodeviatedby
30øtoward the east as microburstwinds from the northwest
a betterunderstanding
of the velocitydatafrom the future
blew towardthe tornado(Figure33). Sincethen, no action
NEXRAD coveringthe United States.
picturesof a tornadointeractingwith a microburst
were
obtaineduntil April 12, 1991,whenKAKE TV obtained
6. TORNADO-M1CROBURST
INTERACTION video scenesfrom Lincoln, Kansas (Figure 34). Another
Prior to the NIMROD experimentin 1978,Fujita [1978] interactionwas confirmedby the aerial photosby Duane
documenteda numberof microburstsin the proximityof Stiegler.In this case,the Hesston,Kansas,tornadoof
tornado tracks flown over for the purposeof aerial survey, March 13, 1990, deviated its track, being pushedby
 WINDSOR LOCKS TORNZ•DO OF OCT .3, 1979
i. 3o T T FUJITA Um.ersff.• of Chlcogo
DarnageSur.e): Roger ;,ak,moto
Duane Shegler
WESTFI ELD " Peter Mc Gurk

,b -'-

WEST
SPRINGFIELD

42ø05 '

Fig. 33. Swirl marks of the Rainsville, Indiana, tornado of April 3,

1974. Undisturbed ground marks extend until crossing U.S. High-
way 41. Thereafter, a microburst airflow from the left (northwest) of
+ the picture pushed the track toward the right. From Fujita [1978].
I/ASSACHUSETTS
.... •bNNECT/CUT

+ 42ø•

CONN - 190

c)

(86mph a! I 58pro)
NWS WIND TOWERe

NWS BAROGRAPH.
(Press Drop, 8 I rnb)

WINDSOR LOCKS

LOCAL DAMAGE MAP

72" 42' 30" Fig. 34. Two video scenes of the April 12, 1991, tornado and a
nearby microburst near Lincoln, Nebraska. Courtesy of Mike
Fig. 32. The Windsor Locks tornado of October 3, 1979, which Phelps. (Top) Formative stage of a tornado and a microburst on the
traveled from south to north. Seven microbursts were mapped on right side. {Bottom) The tornado after touchdown was pushed
the right-hand side of the tornado track. toward the left by microburst
492 TORNADO AND MICROBURST DAMAGE SURVEYS

7. CONCLUSIONS
0 500m IOOOmt
/ ' '
An organized effort of fact-finding aerial survey, photog-
raphy, and mappingof selectedU.S. tornadoes by the Fujita
group gave rise to the identification of the multiscale surface
winds associatedwith tornadoes. These wind systems are
the mesocyclone, tornado, and suction vortex which are
blended into a complicated system of vortices.
N Also identified and clarified were downbursts which are
subclassified into microbursts and macrobursts based on
•.,.' .: their horizontal dimensions.It has been recognizedthat the
microburst is the inducer of the intense wind shear which
endangersaircraft during takeoff and landing operations. It is
likely that the microburst would not have been identified and
confirmed in the 1970s had there been no aerial photos of
strange starburst damage found in the vicinity of tornado
tracks.

'..v...•;•_",. 250.8
m/s The renewed aerial survey and photography of the damage
in storm-affectedareas in future years will be important in
evaluating the NEXRAD data, velocity in particular, in
1915o'7;.•/' relation to the estimated surface winds in different scales.

•'• OpOwer
offat 1915
JST
Acknowledgments. Long-term, continuoussupport of a specific
subject is very difficult to obtain. Fortunately, the aerial survey of
tornadoes and microbursts has been supported for 27 years, since
1965, without interruption by five agencies as partial objectives of
their grants/contractsto T.T.F. as principal investigator at the

1914
o •boraJR
Station
University of Chicago. These objectives are "interpretation of
meteorological satellite data," "design of satellite sensors for se-
vere-storms detection," "understanding of tornado winds for nu-
clear safety," and "ground-truth and detection of microburst for air
safety." The principal investigator wishes to express his sincere
appreciation to those agencies which rendered their long-term
c•29
ai61912
rn/s
JST '"'""
'•' supportin achievingthe research presentedin this paper. Specific
agencieswhich contributedin part to the ground-truthaerial surveys
35o25'N ..........
are as follows: National Science Foundation under grants ATM
140ø18'!E 140Ol9'
............
78-01074, 79-21260, 81-09828, and 85-16705 (1978-1990); Nuclear
Fig. 35. The Mobara, Japan, tornado of December 11, 1990. The Regulatory Commission under contracts NRC-04-74-239 and 04-82-
tornado was pushedtoward the left by a microburstwhich touched 004 (1974-1987); Office of Naval Research under Storm Data con-
down on the right-hand side of the tornado track. The interactive tract N00014-86-K-0374 (1986-1989); NOAA/NESDIS under grants
04-4-158-1, NA80-AAD0001, NA85-AADRA064, and NA90-
processis similar to that shown in Figure 34.
AADRA511 (1973-1991); and National Aeronautics and Space Ad-
ministration under grant NGR 14-001-008(1962-1990). This paper is
funded by grant NA90AADRA511 from the National Oceanic and
Atmospheric Administration.
microburst located on the right side of the tornado track.
Such interactions will alter or contaminate the velocity REFERENCES
pattern of tornadoes.
Agee, E. M., C. R. Church, C. M. Morris, and J. T. Snow, Some
In analyzing the Mobara, Japan, tornado of December 11, synoptic aspectsand dynamic features of vortices associatedwith
1990, it was found that the tornado was pushed off the the tornado outbreak of 3 April 1974, Mon. Weather Rev., 103,
straight-linetrack by an F1 microburston the right-handside 318-333, 1975.
of the tornado (Figure 35). This tornado was rated by the Agee, E. M., J. T. Snow, F. S. Nickerson, P. R. Clare, C. R.
Japan Meteorological Agency as the worst tornado in 50 Church, and L. A. Schaal, An observational study of the West
Lafayette, Indiana, tornado of 20 March 1976, Mon. Weather
years. A research Doppler radar of the Meteorological Rev., 105,893-907, 1977.
ResearchInstitute of Japan indicatedthat the direction of the Forbes, G. S., and R. M. Wakimoto, A concentrated outbreak of
positive-negativevelocity coupletrotated 45øwhile the large tornadoes, downbursts,and implicationsregardingvortex classi-
tornado was being pushedoff the track by a microburst.It is fication, Mon. Weather Rev., Ill, 220-235, 1983.
Fujita, T. T., The Lubbock tornadoes: A study of suction spots,
obvious that the existence of a nearby microburst does alter Weatherwise, 23, 160-173, 1970.
the Doppler-velocity field, making the NEXRAD data inter- Fujita, T. T., Proposedmechanismof suctionspotsaccompaniedby
pretation difficult for tornado warnings. tornadoes, in Preprints, Seventh Conference on Severe
 FUJITA AND SMITH 493

Storms, Kansas Cio', pp. 208-213, American Meteorological Fujita, T. T., and D. J. Stiegler, Tornado of Minneapolis, Minnesota
Society, Boston, Mass., 197I. on July 18, 1986, NOAA Storm Data, 28, 10--13, 1986.
Fujita,T. T., Jumbotornadooutbreakof 3 April 1974,Weather,vise, Fujita, T. T., and R. M. Wakimoto, Five scalesof airflow associated
27, 116-126, 1974. with a series of downbursts on 16 July 1980, Mon. Weather Rev.,
Fujita,T. T., Spearhead
echoanddownburst
neartheapproach end I09, 1438-1456, 1981.
of a John F. Kennedyairportrunway,New York City, SMRP Fujita, T. T., D. L. Bradbury, and C. F. Van Thullenar, Palm
Res. Pap. 137, 51 pp., Univ. of Chicago,Chicago,II1., 1976. Sunday tornadoes of April 11, 1965, Mort. Weather Rev., 98,
29-69, 1970.
Fujita,T. T., Manualof downburstidentificationfor ProjectNIM-
Fujita, T. T., G. S. Forbes, and T. A. U menhofer, Close-upview of
ROD, SLURPRes.Pap. 156, 104pp., Univ. of Chicago,Chicago,
II1., 1978. 20 March 1976 tornadoes: Sinking cloud tops to suction vortices,
Weatherwise, 29, 116-131, 1976.
Fujita,T. T., Tornadoesanddownbursts
in the contextof general- Huschke, R. E. (Ed.), Glossal.'yof Meteorology, p. 585, American
ized planetary scales,,/. Atmos. Sci., 38, 1511-1534, 1981.
MeteorologicalSociety, Boston, Mass., 1959.
Fujita, T. T., The downburst-microburst and macroburst, SMRP
Prosset, N. E., Aerial photographs of a tornado path in Nebraska,
Res. Pap. 210, 122pp., Univ. of Chicago,Chicago,II1., 1985. May 5, 1964, Mon. Weather Rev., 92,593-598, 1964.
Fujita, T. T., The Teton-Yellowstonetornadoof 21 July 1987,Mon. Van Tassel, E. L., The North Platte Valley tornado outbreak of
Weather Rev., 117, 1913-1940, 1989. June 27, 1955, Mon. Weather Rev., 83,255-264, 1955.
Fujita, T. T., and H. R. Byers, Spearheadecho and downburst in Wakimoto, R. M., and J. W. Wilson, Non-supercell tornadoes,
the crash of an airliner, Mon. WeatherRev., 105, 129-146, 1977. Mon. Weather Rev., 117, 1113-1140, 1989.
Fujita, T. T., and F. Caracena, An analysisof three weather-related West, S., Are downbursts just a lot of hot air?, Sci. News, 115,
aircraft accidents,Bull. Am. Meteorol. Soc., 58, 1164-1181, 1977. 170-171,
 LessonsLearned From Analyzing Tornado Damage
TIMOTHY P. MARSHALL

Haag Engineering Company, Dallas, Texas 75381

1. INTRODUCTION 2. WIND Is WINO

In the past two decades, much has been learned by Aerodynamic forces are induced as air flows over and
studying building damage in the wake of tornadoes. Still, around buildings. As a result, the greatest outward (or uplift)
there are problemsat the basiclevel in decidingwhetherthe wind pressures occur around windward walls, roof comers,
eaves, and ridges. The damage due to wind typically in-
damagewas causedby tornadoesor straight-linewinds.Past
volves the removal of wall cladding and roof coverings at
studiesof wind damagehavefrequentlyrevealedthatobjects
these locations. Damage surveys by McDonald and Mar-
were transported along straight paths in tornadoes and
shall [1983] after tornadoes and Savage [1984] after hurri-
followedcurvedtrajectoriesin straight-linewinds.The flight
canes have revealed the same types of building response
characteristicsof objectsand the variability of the terrain are
regardlessof the phenomenon creating the wind.
only two of the many factors which causethis uncertainty. Mehta et al. [1975] and Abernathy [1976] determined that
Thus damageinvestigatorsmustexercisecautionwhenusing large-span structures, such as auditoriums and gymnasiums,
a singlepoint of damageto determine the type of wind field. are quite vulnerable in high winds owing to their large
Assigning F scale numbers to structures based on the surface areas which induce large loads. Such buildings have
degreeof damageis a subjectivevisual procedure.However, been just as susceptible to wind damage in hurricanes as in
when trying to derive the intensity of the winds, it is tornadoes.The general consensusnow is that people should
important to considerhow well the buildingsare constructed avoid shelter in auditoriums and gymnasiums during any
and to recognize weak links or flaws within such structures. type of windstorm.
Large variabilities in the strength of wood-framed buildings
will yield an F scale number with no greater confidencethan 3. BUILDINGS DO Nox EXPLODE
plus or minus one F scale.
Studies by Minor et al. [1977a] and Minor and Mehta It was once thought that the low pressurewithin tornadoes
[1979] have dispelled some of the myths associated with caused buildings to explode. This theory was based on the
erroneous assumption that a building somehow remains
tornadoes and the damage they cause. Cooperative efforts
structurally intact after passing the radius of maximum
between engineers and meteorologists have continued to
winds on the periphery of the tornado. Furthermore, the
yield a better understandingof tornado/structureinteraction.
theory assumesthat the building remained sealed such that
Still, the processof disseminatingthis knowledgeis slow and
the barometric pressure inside the building can become
confusing. Many of the popular beliefs about tornadoes
significantly greater than outside.
conflict with what we know today. We still read about
Studiesof tornado damage presentedby Mehta [1976] and
opening windows as a tornado approaches,and yet people Minor [1976] indicated that building damage initiates from
are told to board up their windows when a hurricane wind pressurebreaching the building, not from low baromet-
threatens. Tornado safety rules tell us to stay away from ric pressure. The wind typically enters the building through
auditoriums, but in hurricane situations, officials still place broken windows or doors. Evidence of mud, insulation,
people in them. This confusion stemsfrom a perceptionthat glassshards, and wood missiles inside buildingsthat remain
wind damagefrom tornadoesis somehowdifferentthan wind partially intact indicate wind had entered the buildings.
damage from hurricanes. Some of the lessons learned in Openings on the windward side of a building actually in-
analyzing tornado damage are the subject of this paper. crease the internal wind pressures, resulting in additional
The Tornado: Its Structure, Dynamics, Prediction, and Hazards.
uplift on the roof {,Figure 1). Thus persons are no longer
Geophysical Monograph 79 advised to open their windows in advance of a tornado.
Copyright 1993 by the American Geophysical Union. Another reason is that flying debris will likely break the
496 LESSONS FROM TORNADO DAMAGE

In addition, the variations of building strength, orienta-

tion, number and type of openings, roof type, degree of
shielding, and impact by neighboring objects are known to
influence trajectories in the damage path. Thus it is impor-
tant to study a large area of damage with several points of
reference before drawing definitive conclusions about
whether a tornado caused the damage. Furthermore, these
ideas should be kept in mind when assigning an F scale
rating.

5. TYPICAL WOOD FRAME BUILDING FAILURES

Damage investigations by Marshall [1985b] and Liu et al.

[1989] have identified certain building connections that have
failed in windstorms, particularly at the following locations:
(1) wall/foundation, (2) wall stud/bottom plate, (3) roof
Fig. 1. This simplifiedschematicshows the airflow (long arrows)
joist/top plate, and (4) rafter/top plate. Uplift forces are often
not considered when the connections are utilized. Each
extending over and into a building with a windward wall opening.
The pressureon the building's interior actually increases,resulting member of a structure should be thought of as a link in a
in additional uplift or outward forces (small arrows). Thus opening chain; the weakest link usually initiates failure.
windows when a tornado approaches could actually be detrimental
Inadequate wall/foundation anchorage has meant failure of
to the structure. Figure courtesy of the Institute for Disaster
Research at Texas Tech University. large portions of structures. Conventional wood structures,
especially rural dwellings, sometimes have little to no an-
chorage to their foundations. Minor [1981] has shown exam-
ples of unanchored buildings that have moved laterally off
windows anyway; thus people should use any advance
their foundations in strong winds. See Figure 3. Properly
warning time to seek appropriate shelter rather than opening
windows.
installed anchor bolts in slabs would help secure walls to the
foundation and provide greater resistance against lateral
movement.
4. OBJECTS TWIST ACCORDING TO THEIR OWN PROPERTIES Wooden wall framing usually is straight nailed to the
The tact that a tree, house, or object is twisted during a bottom plates. As a result, laterally applied forces distrib-
tornado does not indicate that the varying direction of the uted over the height of the wall cause rotation of this
wind causedthe damage. Although the primary wind flow in
a tornado at the ground is rotational, the rotating wind field
extends over a diameter much larger than the dimension of
most objects. The width of an average house is much smaller
than the diameter of an average tornado. Thus at any given
instant, a building in the tornado path would receive winds
that are approximately unidirectional. Tornado damage stud-
ies by Minor et al. [ 1977b] and Minor [1982] have indicated
that twisted buildings are usually the result of variations in
the strengthof foundation anchorage,not the rotating winds.
Often the bathroom plumbing provides the greatest anchor-
age of a house to the foundation, and the house will pivot
around this point. McDonald [1971] concluded that a twisted
house was more likely the result of different resistances in
foundation anchoragerather than the spiralingwinds (Figure
2). Such damage has been known to occur even in straight-
line winds fi-om severe thunderstorms.
In the northern hemisphere the greatest wind velocities
Fig. 2. This premanufactured home was rotated off its concrete
typically occur on the right sides of cyclonically rotating block foundation during the Hereford, Texas, tornado. The house
tornadoes as the effects of translation are added to the actually pivoted about bathroom plumbing which was the only
rotation. Using computer simulations, Metcalf [1978] has significant point of floor-to-ground anchorage. The rotation or
shown that fast translating, weak tornadoes can leave "twist" of this object was best explained by the variations inherent
within the structure rather than rotating wind currents. Note the
straight-line damage paths. Marshall [1985a] found such lack of damage to roof shingles and felt paper underlayment along
straight-line damage trajectories in the debris left behind in walls indicative of relatively low wind velocities. Photograph cour-
the Mesquite, Texas, tornado. tesy of the Institute for Disaster Researchat Texas Tech
 MARSHALL 497

connection,and the nail endspull out. A strongerconnection

would be to install strapsand bracesto put nails or bolts in
shear. Significant resistance to racking failure can be
achieved by installing plywood sheets in wall corners of
wood-framed structures.
Wooden roof joists/top plate and rafter/top plate connec-
tions are usuallytoe nailed. Suchconnectionstypically fail
in tension, causing large sections of the roof to become
displaced(Figure 4). Properly installedstrapsor bracesare
needed to place fasteners in shear, not tension, in order to
provide greater resistanceagainstuplift (Figure 5). As wind
velocities increasewith height above the ground, roof sys-
tems usually experience strong wind uplift pressures.Con-
her et al. [1987] showed variousillustration• in usingstraps
and tie-downs in securingroofs to perimeter walls. Pull tests
conducted by Canfield et al. [1991] have shown a dramatic
increase in the strength of the rafter-top plate connection Fig. 4. The roof on this house was not anchored very well and was
when metal rafter ties were used instead of simple toe uplifted and displaced during peripheral winds in the Mesquite,
Texas, tornado on December 13, 1984. Wind velocities of less than
nailing.
100 mph {161 km/h) can cause such residential damage.
The best time to install connectionsto resist uplift forces
is during construction. Proper placement of anchors, braces,
and connections is essential to anchor floors to foundations,
cantly. An extensive study by Mehta et al. [1971] on metal
walls to floors, and roofs to walls. Increasingthe wind uplift
building performance after the Lubbock, Texas, tornado
resistance of a building after it is constructed is more
found that inward buckling of overhead doors frequently led
expensive, more difficult, and often less effective. Sherwood
to loss of roof and wall corner cladding. Openings in the
[1972] presents several construction details showing how to
install anchors and braces to resist wind forces. windward side of a metal building resulted in increased
interior wind pressures, especially when there were no
openings on the remaining building t;?tces.Similarly, open
6. METAL BUILDING PROBLEMS
bays "catch" the wind, causing increased wind pressureson
Studies of metal building performance after windstorms cladding. Similar failure initiation points in metal structures
have revealed several weak links within such structures. have been documented in hurricane damage by Perry et al.
When failure of a weak link leads to breach of the building [1989] and Ellifi-itt [1984].
containment, the damage to the building increases signifi-
7. UNREINFORCED MASONRY PROBLEMS

The absence of steel reinforcement in concrete block

masonry makes such a system vulnerable to lateral wind
loads. Even load-bearing block masonry walls have often
collapsed owing to the lack of steel reinforcement and cell
grouting. Examples of such failures have been shown by
Sparks et al. [1989] and Hogan and Karwoski [1990] in both
tornado and hurricane damage. Bond beams atop masonry
walls have performed poorly in past windstorms.McDonald
and Marshall [1983] had attributed the failure of an entire
roof system to a failed bond beam system on a masonry
structure. Failure initiated as the mortar joint below the
bond beam failed in tension.
Absence of brick ties to •ecure masonry veneers has led to
wall failures at relatively low wind velocities (Figure 6).
Even when brick ties have been properly installed,corrosion
and fatigue over time can reduce the performance of such
ties.
Fig. 3. Lateral displacementof this unanchoredhouseduringthe
Grand Island, Nebraska, tornado caused the wooden floor structure
to lose support and fall into the basement. Lack of damage to 8. ROOF SYSTEM PROBLEMS
surrounding trees and roof shingles indicated house movement
occurred at wind velocities of probably less than 100 mph (161 A number of damage studies ha,,e been conductedi.nvolv-
km/h). ing the removal of various roof coverings.A common
498 LESSONS FROM TORNADO DAMAGE

WALL
STUD-
•JI!F•
16
GA.
PLATE

RAFTER

16 GA. METAL PLATE

ANC•R BOLT

(a) (b)
Fig. 5. Connectiondetailsto resistwind uplift' (a) wall-foundationdetail and (b) roof-wall detail.

initiation point on roof systems occurs where the roof was used to secure the roof membrane to the insulation
membranes are attached to edges and corners. McDonald board. As the membrane was uplifted, failure initiated in the
and Smith [1990] attributed several roof system failures to layer of insulation board just beneath the location where the
the lifting and peeling of metal edge flashings. Uplift of the adhesive was absorbed. In this construction the insulation
roof edgesallows the wind to penetrate underneaththe roof board was inadvertently used to resist some of the wind
membrane, resulting in pressurerise beneath the membrane uplift forces. Similar roof system failures have been docu-
and removal of the roof covering. mented by Kramer [1985].
Another failure initiation point has been traced to the lack
of attachment between insulation board and the roof deck. In
9. FLYING DEBRIS PROBLEMS
instancesinvolving the application of hot bitumen on a metal
roof deck, it is important that insulation board be installed The presence of flying debris or missiles in the wind field
immediately after application of the bitumen in order to can greatly increase the damage to structures by creating
develop a significantbond. A number of roof failures have openings in the building where the wind can enter. McDon-
been documented by the author where the applied asphalt ald [1976] has documented a wide range of missiles found in
had cooled prior to the installation of the insulation board. In the damage paths of tornadoes. Among the most common
other instances, an insufficient amount of bitumen had been missiles are wooden boards, sheet metal, and roof gravel.
applied to the metal deck. This led to virtually no bond Missile drop tests conducted by Thompson [1973] showed
between the insulation board and the deck. that small missiles could penetrate walls with single sheath-
Failure of roof systems has occurred when an adhesive ing at speedsas little as 32 mph 1:51km/h).

10. SUMMARY

It is important to consider how well buildings are con-

structed when interpreting the damage after a windstorm.
Variations in damage along the path are not always ex-
plainedby a changein wind velocity and could be attributed
to weak links in building construction. Engineering studies
over the past two decades have yielded increasingevidence
that the extent and type of anchors, braces, and connections
usedin buildingscorrelate well with how buildingsbehave in
the wind. A more accurate F scale rating can be obtained by
consideringthe structuralstrengthsand weaknessesof dam-
aged buildings.

REFERENCES

Abernathy, J. J., Protection of people and essentialfacilities, in

Proceedingsof thc Symposiumon Tornadoes,pp. 407-418, Texas
Fig. 6. Damage to a two-story brick veneer wall in the Mesquite, Tech University, Lubbock, 1976.
Texas tornado. There were no brick ties to anchor the wall to the Canfield, L., S. Niu, and H. Liu, Uplift resistance of wtrious
wooden frame. The wall toppled on the leeward or "suction" side. rafter-wall connections, For. Prod. J., 41(7/8), 27-34,
 MARSHALL 499

Conner, H., et al., Roof connections in houses: Key to wind Abernathy, and U. Koehler, Engineering Aspects of the Torna-
resistance, J. Struct. Div. Am. Soc. Civ. Eng., 113, 2459-2473, does of April 3-4, 1974, 110 pp., National Academy of Sciences,
1987. Washington, D.C., 1975.
E'llifritt, D. S., Performance of metal buildings in Houston- Metcalf, D. R., Simulated tornado damage windfieldsand damage
Galveston area, in Hurricane Alicia One Year Later, pp. 117-123, patterns,M.S. thesis,67 pp., Tex. Tech Univ., Lubbock,I978.
American Society of Civil Engineers, New York, 1984. Minor, J. E., Applications of tornado technologyin professional
Hogan, M., and A. Karwoski, Masonry performancein the coastal practice, in Proceedingsof the Symposiun•on Tornadoes,pp.
zone, in Hurricane Hugo One Year Later, pp. 195-206, American 375-392, Texas Tech University, Lubbock, 1976.
Society of Civil Engineers, New York, 1990. Minor, J. E., Effects of wind on buildings,in The Thunderstormin
Kramer, C., Wind effects on roofs and roof coverings,in Proceed- Human Affairs, pp. 89-109, Universityof OklahomaPress,Nor-
ings qf the Fifth U.S. National Conferenceon Wind Engineering, man, 1981.
pp. 17-34, 1985. Minor, J. E., Advancementsin the perceptionsof tornadoeffects
Liu, H., H. Saffir, and P. Sparks, Wind damage to wood frame (1960-1980), in Preprints of the Twelfth Conferenceon Severe
houses: Problems, solutions, and research needs, J. Aerosp. Local Storms,pp. 280-288, AmericanMeteorologicalSociety,
Transp. Div. Am. Soc. Civ. Eng., 2, 57-69, 1989. Boston, Mass., 1982.
Marshall, T. P., Damageanalysisof the Mesquite,Texastornado,in Minor, J. E., and K. C. Mehta, Wind damageobservationsand
Fifth Conferenceon Wind Engineering, session4B, pp. 11-18, implications,
J. Struct.Div. Atn. Sot'.Civ. Eng., 105,2279-2291,
American Society of Civil Engineers, New York, 1985a. 1979.
Marshall, T. P., Performance of structuresduring Hurricane Alicia Minor, J. E., K. C. Mehta, and J. R. McDonald, 1977a:The
and the Altus tornado, in Proceedings of the Department of tornado: An engineeringoriented perspective,NOAA Tech.
Energy Natural PhenomenaHazards Mitigation Conference,pp. Memo., ERL-NSSL-82, 196 pp., 1977a.
140-150, Departmentof Energy, Washington,D.C., 1985b. Minor, J. E., K. C. Mehta, and J. R. McDonald, Engineering
McDonald, J. R., The Hereford tornado: April 19, 1971, report, 15 orientedexaminationsof the tornadophenomenon,in Preprintsof
pp., Tex. Tech Univ., Lubbock, 1971. the Tenth Conferenceon Severe Local Storms, pp. 438--445,
McDonald, J. R., Tornado generatedmissilesand their effects,in AmericanMeteorologicalSociety,Boston,Mass., 1977b.
Proceedingsof the Symposiumon Tornadoes,pp. 331-348,Texas Perry,D.C., J. R. McDonald,andH. Saffir,Strategies
for mitigat-
Tech University, Lubbock, 1976. ing damageto metal buildingsystems,J. Aerosp.Transp.Div.
McDonald, J. R., and T. P. Marshall, Damage survey of the Am. Soc. Civ. Eng., 2, 71-87, 1989.
tornadoes near Altus, Oklahoma, on May 11, 1982, Publ. 68D, Savage,R. P., Hurricane Alicia: Gah,estonand Houston, Texas,
Tex. Tech Univ., Lubbock, 1983. August17and18, t983, 158pp., CommitteeonNaturalDisasters,
McDonald, J. R., and T. L. Smith, Performance of roofing systems National AcademyPress,Washington,D.C., 1984.
in Hurricane Hugo, report, 42 pp., Inst. for DisasterRes., Tex. Sherwood, G., Wood structures can resist hurricanes, in Civil
Tech Univ., Lubbock, 1990. Engineering,pp. 91-94, AmericanSociety of Civil Engineers,
Mehta, K. C., Windspeedestimates:Engineeringanalyses,in New York, 1972.
Proceedingsof the Symposiumon Tornadoes,pp. 89-103, Texas Sparks,P. R., H. Liu, and H. Saffir, Wind damageto masonry
Tech University, Lubbock, 1976. buildings,J. Aerosp. Transp.Div. Am. Soc. Civ. Eng., 2, 186-
Mehta, K. C., J. R. McDonald, J. E. Minor, and A. J. Sanger, 198, 1989.
Responseof StructuralSystemsto the LubbockStorm, Storm Thompson,R. G., The responseof residentialwall construction
Res. Rep. 3,427 pp., Texas Tech University,Lubbock, 1971. concepts to missileimpact,report,59pp., Dep.of Civ. Eng.,Tex.
Mehta, K. C., J. E. Minor, J. R. McDonald, B. R. Manning,J. J. Tech Univ., Lubbock,
 Survey of a Violent Tornado in Far Southwestern Texas'
The Bakersfield Valley Storm of June 1, 1990
GARY R. WOODALL AND GEORGE N. MATHEWS

National Oceanic and AtmosphericAdnzinistration/NationalWeather Service, Lubbock, Texas 79401

1. INTRODUCTION the area from the southwest. A major disturbance at 500

mbar was located over Idaho, with several shortwaves
During the late afternoon of June 1, 1990, a violent tornado rotatingaroundthe large 500-mbarlow. An area of difiuence
struck northern and eastern Pecos County, Texas. This was presentover western Texas at 200 mbar, with the left
tornado killed two people, injured 21, and causedover $5 front quadrant of a 200-mbarjet streak [UcceIlini, 1990]
million in damage. The tornado passedover the unincorpo- moving toward the area.
rated ranching community of Bakersfield Valley, so the Area soundingstaken at 1200 (Figure 2) also suggested
tornado will be referred to as the Bakersfield Valley (BV) severe weather potential [Doswell, 1982]. The soundingat
tornado. MAF had a low-level moist layer with a layer of drier air at
On June 4 and 5, 1990, personnel from the National midlevels (above 750 mbar). Although the winds in the
Weather Service office in Midland, Texas (MAF), and the lowest3 km of theMAF sounding
werelessthan15m s-• ,
National Weather Service forecast office in Lubbock, Texas the winds veered more than 90 ø in the lowest 3 km. The
(LBB), performed a ground and aerial'surveyof the tornado soundingfrom E1 Paso (ELP) indicateddry air and westerly
damage.The survey revealed a damagetrack 22 miles (•-35 to southwesterly winds at all levels.
km) long and an average of 0.7 mile (---1.1 km) wide. There During the midmorning, a mesoscaleconvective system
were several isolated occurrences of F4 damage on the (MCS) developed near Childress, in the southeasternpan-
Fujita damage scale [Fufita, 1981]. Results of the survey handle.The MCS propagatedslowly southeastward,leaving
suggestedthat this was probably the largest and strongest much of northwest Texas in cool, stable outflow. Mean-
tornado to strike western Texas since the Lubbock tornado while, the dryline became very well defined in southwest
of 1970. Texas, near and to the south of MAF, as low-level moisture
Section 2 of this work reviews the synoptic and mesoscale was mixed out west of the dryline. See Figure 3 for a surface
environment of the BV tornado. Section 3 outlines the life analysisat 2100, shortly after the time thunderstormsbroke
historyof the stormas viewed by satelliteand radar. Section out southwest of MAF.
4 describes the results of our damage survey in detail, and Had the MCS's cool outflow not affected the region, much
and concludingremarks. of West Texas would likely have seen significant severe
section5 presentssomediscussion
weather. As it was, though, the main severe weather threat
2. SYNOPTIC AND MESOSCALE ENVIRONMENT area was shifted to the south of MAF, where strongsurface
heating and the pronounced dryline acted to focus the
At 1200UTC (all times are givenin UTC) on June 1, 1990, development of strong convection.
a number of parametersconduciveto severeconvection
were assemblingover West Texas [Miller, 1972;National 3. MORPHOLOGY OF THE STORM
Severe Storms Forecast Center, 1991]. The 1200 composite
-1
analysis(Figure 1) illustratestheseparameters.A 20 m s The first low-level radar echoesappearedto the southwest
southerlylow-leveljet was centeredover the region.Dew of MAF around 1900. By 2015 (Figure 4), a supercellhad
pointsat 850 mbar were greaterthan 15øCover all but the developedin southwesternCrane County, with a tornado
Texas Panhandle.A 700-mbardry punchwasmovingtoward reported8 miles (•-13 km) west of Crane at aboutthis time.
New cells had developed south of the supercell, between
The Tornado: Its Structure, Dynamics,Prediction,and Hazards. McCamey and Grandfalls. The cell which went on to become
GeophysicalMonograph 79
This paperis not subjectto U.S. copyright.Publishedin 1993by the the BV supercellwas one of thesecellsand developedon the
American GeophysicalUnion. Pecos River west of McCamey.
502 BAKERSFIELD VALLEY STORM

Fig. 1. Upper air compositeanalysisfor 1200UTC on June 1, 1990. Thin arrow marks 850-mbarjet. Scallopedline
denotes850-mbar dew point temperatures of 15øCor greater. Ticked line marks 700-mbar dry intrusion. Broad arrow
shows500-mbarjet. Broad L denotes 500-mbar low center, and dashedlines denote 500-mbar shortwaves.

By 2130 (Figure 5) the Crane County supercell had lost BV supercell by 2330 (Figure 7). The BV tornado lifted at
someof its radar structure. The low-level reflectivity gradi- about the time this radar observation was made. The BV
ent relaxed, and the curvature earlier seen on the echo's supercell had lost much of its structure by this time and was
southernflank was no longervisible.The BV stormacquired propagating roughly parallel to the mean wind. The storm
radar characteristicsassociatedwith a supercell, namely, a produced large hail after this time, but no further tornadoes
tight low-level reflectivitygradientand a broad concavity on were reported.
the echo's south side. The cells were moving from 270ø at
about 8 m s-]. A hook echo was evident at the southwest
4. DAMAGE SURVEY RESULTS
flank of the BV supercell, near the town of Girvin.
The BV supercell had taken on the appearance of a Personnel at the National Weather Service (NWS) fore-
high-precipitation(HP) supercell [Moller et al., 1990] by cast office LB B were notified of the casualties in Pecos
2230 (Figure 6). The echo had become large (70 km x 60 km County late in the evening of June 1, 1990. Because both of
along its long and short axes). The radar echo had taken on the fatalities and most of the injuries were in vehicles and
a kidney bean-shapedappearance,with a broad hook echo because little additional information was available to the
and a tight reflectivity gradient on the southeasternflank of NWS, it was concluded that the B V tornado was probably
the cell. Unfortunately,radar data at higher-elevationangles not a large, violent tornado. Thus the NWS concluded that
were not available. The BV tornado had touched down 10 delaying a damagesurvey until Monday, June 4, 1990, would
min prior to this radar observation. As the tornado moved be acceptable.
toward Iraan, eyewitnesses noted that the tornado was at The damage survey team was shocked at what it discov-
least partially embedded in precipitation. The storm was ered when it arrived in Pecos County on June 4, 1990. A map
now propagatingto the right of the mean wind, movingstill of the tornado track, with documented damage and esti-
at 8 m s-• but from about 300ø. mated F scale damage intensities, is shown in Figure 8. The
Another supercelldeveloped to the west-southwestof the tornado track was 22 miles (-•35 km) in length, with
 "•-•"i
.......
:' ,,'....
,-T"•'"
' " ' ',",
...., ,'.....
•.......
:.......
, '• :• , , , , , , • ,, ., , , , , , ,

-, • • • ', : ', :", : '.: '. s•• •x': -•

'''' ' • '.,' ',, , ' ' : • CROSS

TOT•L•
=
•ee:.•_._'_.•..•,•q
, -.,,
.... T•..•...z,,•;
,,-. ,, , ,, -.
..... •......
...... •_.,,...,,.•6•, suz,r
;
•R=, es•
,
......
. '•..
L .. ,•-.....•• . •...:
:, .;. ,:,, ,.., . :.
,:, :,, ,:
',-- •' ..•i
--' ,•:•,
"• ":•-. -.•L.'
,, "' .......
.... ' ' ,.' ', ',
:". : ".: ', : ¾. : ' "•.'. :k '-. :', '. :
: '. : :- ß , , -. , ', , •. , , -. ; ',: , , ,
-. , -., , '. : : , -• . : ,'. : ', :'. .....
See ' ' " ' • , .... -• ........ •:---': .......................
.... -. , , , ,-.: . , •. : , •.. ,, •
,, -• , .. , ,, ., , , -• , , ,-., , ;, -. ;; ,

. , -., , -. ,, ,, .. ,: , ,-. , : ,' ..... .,,

k ,. : ,-.,
see • ....

,]:..... ,
see ,,,•...... , ,,,: •, •;,, .... , ,..••..
...... .
' ', , , '- ' : ,'- ', • ', ''-, ', ', ' ', ' ' ' '-A:
' . .... -'• ...... -' , .:..2.•_5 .... j...•..:' , ..•...:: , ,, _.... •,_
-70 -60 -•0 -•0 -30 -•0 -•e ee +re +2e +3• +•0

Fig.2. Areasoundings
for 1200UTConJune1, 1990.
Soundings
plotted
inpseudoadiabatic
diagram
format.
MAF,
Midland, Texas' ELP, El Paso,
504 BAKERSFIELD VALLEY STORM

29.4 24

-3. 1.7 989

42
13

25.0(•)033

I
I
31.7•033
• " •h-. 30.3•
1• 20.0• 31.1 9
•
,,• 21.1

.....
•"• ...........
• .................. 20.
//.

Fig. 3. Surfaceanalysisat 2100UTC on June1, 1990.Dry bulbanddew pointtemperaturesin degreescentigrade.Sea

levelpressures
given
intenths
ofmillibars
withleading
9 or!0omitted.
Fullwind
barbs
indicate
5 ms-l; halfbarbs
denote2.5 m s-'. Drylineandoutflowboundaryplottedin standardnotation.Asteriskindicateslocationof BV tornado.

average width of 0.7 mile (•-1.1 km). The maximum path assigningdamageratings. As with all damagesurveys, this
width was 1.3 miles (---2 km). There were several isolated fact shouldbe kept in mind duringthe following paragraphs.
instances of F4 damage. The times indicated in the following summary were ac-
Bunting and Marshall [1991] and Minor et al. [1977] note quired as follows. Reliable eyewitnesstimes were obtained
that numerous variables exist when attempting to assign a at the Co-Op building, the McKenzie residence,the Kim-
rating to damagedstructures. The designand constructionof broughresidence,and the locationsof the vehicle fatalities.
a buildingare perhapsthe two biggestfactors contributingto The radar-derived cell motion of 8 m s -• was used to
overall building strength. Orientation of the buildingto the estimate the remaining times along the damage path.
wind and variations in the strength of construction materials The tornado first touched down at 2220, approximately 9
also affect the building's ability to withstand strong wind. miles(--•14.5km) southof McCamey. The first 2 miles(---3km)
When all of thesefactors are combined,it becomesapparent of the tornado track were marked by intermittent F0 damage,
that a variation of +--1 F rating can be expected when inflictedmainly on telephonepolesand lightweight
 WOODALL AND MATHEWS 505

C;ty

tO'

NG

A?no
.

..

OToyan

REEVES o _

Ker•
. ,.s.•..o,.. : : , ,
.• : iraan

• : PEC . Ozor
Fig. 4. Midland, Texas, radar overlaytaken at 2030 UTC on June 1, 1990.D/VIP levels 2, 3, 4, 5, and 6 shown.

SLASSCOCK
•!,Mentone

II•Toy,•rl

REEVES

.........
...

..
.eSarago•sa

e•ort

Fig. 5. As in Figure4, but for 2130

506 BAKERSFIELD VALLEY STORM

ING 0• essa
O i : City
'MIDLAND
--
.

SCOCK
.

Am,
•on•l

CRANE

:IraIls

e•igLa•e
REEVES

,•Sarago•sa
I

CROCKETT
•C Ozoma,

Fig. 6. As in Figure 4, but for 2230 UTC.

Mon

Pyo:ee
CR.•NE
UPT .

%
.-

REEVES

,•Sara•o•
sa

-
ROCKETT
. PEC Ozona,
-

•l•or•Dayas :

Fig. 7. As in Figure 4, but for 2330

 WOODALL AND MATHEWS 507

AAcCamey

Co-Op Road

Iraan

. Empty Oil Tanks Moved

2. Co-Op Building Destroyed
3. Foster Residence Damaged
4. McKenzie Residence Destroyed
5. 300-ft Asphalt Section Removed
6. Vehicle Fatality
7. Vehicle Fatality
8. Oil Tanks Hoved 3 Miles
9. Kimbrough Residence Destroyed
10. 57 Pump Jacks Damaged/Dear. toyed

6.5 13
, .. , ..... Ikm

Fig. 8. Damage track of the BV tornado.Numbers insidetrack indicatedocumenteddamagelocations.Numbers

abovetrack denoteestimatedF scalerating,and numbersbelow track indicateapproximatetime (CDT).

At 2225 the tornado began a period of dramatic increase in mesquitetrees, etc.) was completely removedby the torna-
size and intensity. Two miles west of Farm to Market Road do's wind and debris. Becausethe cooperativebuildingwas
(FM) 1901, a pair of empty 500-barre!oil tanks were moved of fairly lightweightconstruction,we rated the damageas
by the tornado. One tank was moved about 50 feet (---!5 m) marginal F3.
to the northeast, while the second tank was carried nearly After leaving the farm cooperativebuilding,the tornado
0.5 mile (---0.8 kin) to the north. Roughly 0.25 mile (--0.4 kin) moved along a secondary highway called Co-Op Road.
north of the tanks' starting point, several ranch outbuildings Between 2230 and 2240, 4 miles (---6.5 km) southeast of the
were destroyed. cooperative building, the tornado struck a residenceowned
Between 2225 and 2230 the tornado strucka farm cooper- by the Foster family (Figure 10). The Foster home was a
ative building (Figure 9). The building was of corrugated 5-year-old brick building with an adjacent barn and work-
metal and I-beam construction. The building was heavily shop constructed of corrugated metal. The home and the
damaged, and there was evidence of vegetation scouringin workshopwere heavily damaged,but the barn sufferedonly
this area. This location marked the startingpoint of a 0.5 x minor damage.Extensive vegetationscouringwas evident in
8 mile (0.8 x 14 km) swath in which the vegetation (grass, this area. Damage to the Foster home was rated as
508 BAKERSFIELD VALLEY STORM

Fig. 9. Farm Co-Op building damage (F3); view to west. Fig. 11. McKenzie residence damage {F4): view to east.

The Fosters were home when the tornado struck. They The McKenzies were also at home when the tornado
had no warning of the tornado, however, until it appearedto struck. Mrs. McKenzie, who taught science in high school,
their west (electric power had already been lost). The immediately recognized the large tornado when it appeared
Fosters quickly took shelter into the interior portion of the to their west. Like the Fosters, the McKenzies quickly
building and escaped with only minor injuries. moved into the interior part of the ground floor of their home
A few moments later, the tornado struck the McKenzie and, amazingly, escaped with only minor injuries.
residence, located 1 mile (---1.6 km) southeast of the Foster Almost 2 miles (---3 km) east of the McKenzie residence,
residence(Figure 11). The McKenzie home was a 3-year-old, the tornado removed a 300-foot (---90 m) section of asphalt
two-story, well-engineeredhome of solid adobe brick construc- from Co-Op Road (Figure 12). The removal of asphalt by a
tion. The home was destroyed by the tornado. The survey tornado is a fi•nction not only of the tornado's strength, but
team ratedthis damageas F4. All that was left wasthe staircase also of the age of the asphalt and the quality of road
and some cabinets at the northeast corner of the building. engineering. While asphalt removal is not unique, the re-
Papersfrom the McKenzie home were found in Big Lake, 55 moval of a large area of roadway is suggestiveof a violent
miles (---88 km) east-northeast of Bakersfield Valley. The tornado (A. R. Moller, personal communication, 1990).
damage track at this time was over I mile (---1.6 km) in width.

Fig. 12. Sectionof asphaltremovedfrom Co-Op Road (F4); view

Fig. 10. Foster residencedamage{F3); view to northeast. to
 WOODALL AND MATHEWS 509

Fig. 13. Pickup truck in which fatality occurred, near Co-Op Road and FM 305 (F2).

Extensive vegetation scouring continued into this area. it approachedthe intersection of Co-Op Road and FM 305.
Suction marks, suggestinga multiple-vortex structure, were At roughly 2250 a man driving a pickup truck on Co-Op Road
also evident. The damage to the road and vegetation, which was caught by the tornado. The truck was blown 150 feet
was inflicted at approximately 2245, rated as F4. (-45 m) off the road, and the driver was killed (Figure 13).
After the asphalt removal, the BV tornado showed some Several minutes later, a family traveling along FM 305
signsof weakening. However, the tornado became a killer as apparently drove into the tornado. The tornado had possibly
become embedded in precipitation by this time. The family's
pickup truck was blown 500 feet t-150 m) off the road by the
tornado, and the father was killed. Damage in thi• area was
estimated as only F2.
The tornado crossed FM 305 at about 2300. The tornado
seemed to increase in intensity after crossing FM 305. The
tornado entered an oil field, hit a tank battery, broke three of
the 500-barrel oil tanks loose from the battery, and rolled
them 3 miles (-•5 kin) across a field. Consultation with local
authorities revealed that the tanks were full when they were
moved and that they weighed approximately 70 tons each.
Two of the tanks were pushedonto a hillside, while one tank
was pushedover the hill. Weak suction marks were evident
near the tank battery. The survey team rated this damage a•
marginal F4.
From 2315 to 2330 the BV tornado moved through the
Yates Oil Field and damaged or destroyed 57 oil pumps.
Numerous other pieces of oil field equipment were damaged.
The tornado also struck a residence owned by the Kim-
broughfamily (Figure 14). The Kimbroughhousewas picked
Fig. 14. Kimbrough residencedamage(F3)' view to north. up, carried roughly 30 feet (-•9 m), and dropped by
510 BAKERSFIELD VALLEY STORM

tornado.The buildingwas of fairly lightweightconstruction The NWS must continue to stresstornado safety proce-
andwas not well engineered,sothe damagein thisareawas dures, especially tornado safety in vehicles. Admittedly,
rated as only F2-F3. The Kimbroughswere not homewhen vehicle safety will probably remain a major problem in
the tornado struck. A laborer in their workshopescaped tornado safety, even with the addition of National Oceanic
injury by hiding under a drill press when the tornado and Atmospheric Administration weather radios in some
approached. vehicles. People outside of the main tornado threat areas
The damage to the oil field was the last documented must continue to be reminded that they are indeed vulnera-
damagecausedby the BV tornado. The BV tornadolifted at ble to the type of destruction which occurred in Bakersfield
2330, 5 miles (---8 kin) west-southwest of Iraan. Valley on June 1, 1990.
Eyewitnessesnear the farm cooperativebuildingand at
the Foster/McKenzie residences described the tornado as a
Acknowledgments. John Wright of NWS Midland, Texas, led
turbulent, debris-filled cylinder extending up to the cloud the damage survey of June 4-5, 1990, and provided the radar data.
base. The tornado possibly appearedsimilar to the Wichita The Pecos County SheriWs Department gave us the initial tour of
Falls, Texas, tornado of April 10, 1979 [ U.S. Department of the damage area. The Texas Department of Public Safety provided
Commerce, 1980]. By the time the tornado reachedUS 190 an aerial tour of the damage track. Alan Moller of NWS Fort Worth,
west of Iraan, eyewitnesses noted that the vortex had Texas, and Tim Marshall of Haag Engineering in Dallas, Texas,
assistedwith the F scale estimates.A special thanks to the employ-
become embeddedin precipitation and that it was difficultto ees of Bakersfield Co-Op, the Foster family, the McKenzie family,
discern. Unfortunately, no pictures or video were taken of the Kimbrough family, the Iraan News, the Fort Stockton Pioneer,
the tornado. and the OdessaAmerican for their eyewitness and news accounts of
the tornado. Melody Woodall drafted the figures.
5. DISCUSSION AND CONCLUSIONS
REFERENCES
On June 1, 1990, a violent multiple-vortextornado struck
Bunting, W., and T. Marshall, A resource guide for conducting
northern and eastern Pecos County, in far southwestern
damage surveys, in WCM Manual, National Weather Service,
Texas. This tornado, named the Bakersfield Valley (BV) Southern Region, Fort Worth, Tex., 1991.
tornado, had a 22-mile (---35 km) path length and an average Doswell, C. A., III, The Operational Meteorology of Convective
path width of 0.7 mile (--1.1 kin). There was evidence of Weather, vol. I, Operational Mesoanalysis, NOAA Tech. Memo.
multiple-vortex structure in the damage path. Several in- NWS NSSFC-5, 160 pp., National Severe Storms Forecast Cen-
ter, Kansas City, Mo., 1982.
stances of F4 damage were noted. Such occurrences are Fujita, T. T., Tornadoes and downbursts in the context of general-
rare; during the preceding 40-year period, the Saragosa ized planetary scales, J. Atmos. Sci., 38, 1511-1532, 1981.
tornado of 1987 was the only violent tornado documented in Miller, R. C., Notes on analysis and severe storms forecasting
this region [National Severe Storms Forecast Center, 1990]. proceduresof the Air Force Global Weather Central, Tech. Rep.
200, revised, Air Weather Service, U.S. Air Force, Washington,
Damage was observedand documentedin the BV tornado D.C., 1972.
which was not seenin the Saragosatornado. The removal of Minor, J. E., J. R. McDonald, and K. C. Mehta, The tornado: An
asphalt from Co-Op Road and the extensive scouring of engineeringperspective, NOAA Tech. Memo. ERL NSSL-82, 196
vegetation are examples of damage which was documented pp., Natl. Severe Storms Lab., Norman, Okla., 1977.
in the BV tornado but not in the Saragosa tornado. On the Moller, A. R., C. A. Doswell III, and R. Przybylinski, High-
precipitationsupercelIs:A conceptualmodel and documentation,
basis of this additional damage, we suspect that the BV in Preprints, 16th Conference on Severe Local Storms, pp. 52-57,
tornado was at times stronger than the Saragosatornado. American Meteorological Society, Boston, Mass., 1990.
However, insufficient conclusive evidence existed to war- National Severe Storms Forecast Center, Tornado breakdown by
rant an F5 rating for B V. counties: 1950-1989, report, 108 pp., Kansas City, Mo., 1990.
National Severe Storms Forecast Center, Storm scale processes:
Spotter operations were hampered by the remote location
What to look for when forecasting severe storms, report, 40 pp.,
and rugged geography in the vicinity of BV. However, Kansas City, Mo., 1991.
emergency management operations in and around Iraan and Uccellini, L. W., The relationship between jet streaks and severe
McCaracy were handled exceptionally well. People who convective storm systems, in Preprints, 16th Conference on
found themselves in or near the tornado's path generally Severe Local Storms, pp. 121-130, American Meteorological
Society, Boston, Mass., 1990.
employed proper protective measures. These measures, U.S. Department of Commerce, Red River Valley tornadoes of
repeatedlystressedby the NWS's public preparednesspro- April 10, 1979, Natur. Disaster Surv. Rep. 80-1, 60 pp., Natl.
grams, undoubtedly contributed to the low fatality toll. Oceanic and Atmos. Admin., Rockville, Md.,
 An Observational Study of the Mobara Tornado
H. NIINO, O. SUZUKI, T. FUJITANI,H. NIRASAWA,H. OHr40, I. TAKAYABU,AND N. KINOSHITA

Meteorological Research Institute, Tsukuba 305, Japan

T. MUROTA AND N. YAMAGUCHI

Building Research Institute, Tsukuba 305, Japan

1. INTRODUCTION structure. Thus it seems of interest to examine the differ-

encesof tornadic storms in Japan and the United States.
According to statisticsfrom 1961 to 1982 [Mitsuta, 1983], Since the number of tornadoesper year is about 40 times
an average of 18 tornadoes per year occur in Japan. This smallerin Japanthan in the United States,not manydetailed
number is much less than 771, which is the 30-year annual studies of tornadoes and tornadic storms have been made so
mean in the United States [Fergussonet al., 1989]. Since the far. One exceptionis the study by Fujita et al. [1972], who
United States is about 25 times as wide as Japan, this makes madea damagesurveyand mesoscaleanalysisof the Omiya
the number density about 1.7 times greater in the United tornadoon July 7, 1971. In order to examine differencesof
States than in Japan. tornadic stormsin Japan and the United States, it is neces-
Tornadoes in Japan are generally believed to be weaker sary to increaseour knowledgeabout tornadic stormsin
than those in the United States. In fact, there have been no Japan.
tornadoes ranked F4 or more since 1950 [Fujita, 1971; On the eveningof December I 1, 1990, one of the severest
Mitsuta, 1983]. However, the relative frequenciesof F0, F1, tornadoesafter the World War II era hit Mobara City, 55 km
F2, and F3 tornadoes in Japan and the United States are southeastof Tokyo and 75 km southof the Meteorological
quite similar.One possibleexplanationfor the absenceof F4 Research Institute (MRI) in Tsukuba (see Figure 1). One
tornadoesin Japan may be as follows: While most of the persondied, 73 peoplewere injured.82 houseswere totally
violent tornadoes in the United States are known to be destroyed, 161 houseswere severely damaged,and 1504
spawnedby supercellstorms[Browning,1964],thistype of houseswere lightly damaged. On December 12, scientists
stormhasrarely beenobservedin Japan.This maybe mainly from MRI and the Building Research Institute (BRI) started
due to the difference in the abundance of moisture. In the to make detailed damage surveys and to interview the
United Statesthe atmosphereis usuallydrier than in Japan, occupants.
which is surroundedby oceans.Consequently,cloudbases During the evening of December 11, the singleDoppler
are generallyhigherover the United States(exceptfor some radarat MRI was operatingand detecteda mesocyclone.As
eastern states), and storms can have strongerdensity con- will be shown in the following, this mesocycloneproduced
trast between the ambient air and the downdraft-induced the Mobara tornado. Therefore it will hereinafter be called

cold air. This density contrasttogetherwith southeasterly tornadocyclone(TC) [Brooks, 1949].A considerable amount
flow at low levels is important for producing a strong of meteorologicaldata was collectedfrom the JapanMeteo-
convergence on the fight flankof the storms.Variousstudies rologicalAgency, local governments,schools,fire depart-
ments, and elsewhere to analyze the mesoscalefeatures
on the supercellstorms[e.g., Browning, 1964]have also
which led to the tornadogenesis.
revealedthat the existenceof a dry southwesterlyflow in the
In the followingthe characteristicsof the tornadoderived
middle troposphereis essentialto maintaintheir typical
from the damagesurvey and the results of the mesoscale
analysiswill be described.This is the firststudyin Japanin
The Tornado: Its Structure, Dynamics,Prediction,and Hazards.
GeophysicalMonograph 79
which a TC detected by a Doppler radar was analyzed
Copyright 1993by the American GeophysicalUnion. togetherwith surfacemeteorologicaldata.
512 OBSERVATIONS OF THE MOBARA TORNADO

,Hokka
Tsuchiura

O
Tsukuba

_Korean Pen. / •'

¸ Nari ta Choshi
Tokyo
Noto Pen.-'--f• )/ o Chiba
o

Haneda
Mobara
Yokohama
Kisarazu
Boso Pen.
Futtsu

Pac f c Ocean Kamogawa

Tateyama
Maruyama
Okinawa 100 km
5O
!

Fig. 1. Maps of Japanand the Kanto District.

2. CHARACTERISTICS OF THE TORNADO Careful examination of Figure 2 shows that the damage
distribution has an eastward protuberance of about 500 m
2.1. Direct Observation
near the Mobara City Gymnasium north of the Takashi area
The tornado started at about 1913 JST (see section 2.6) (see Figure 4). It is also noteworthy that several housesin
when it was already dark. There were frequent lightning the eastpart of the protuberancewere completelydestroyed
flashes, and the tornado itself seemed to emit a blue light or severely damaged.This point will be discussedlater.
occasionally. The funnel cloud illuminated by lightningwas
recorded on video by Kazumi Sasaki from a distance of 1
2.3. Meteorological Records
km. The video showed that the tornado was rotating anti-
clockwise and moved approximately northward. Meteorologicalrecords near the tornado path were taken
at several schools and at other observation points where
2.2. Damage Characteristics there was no means to accurately register the times. The
Figure 2 showsthe distributionof damagedhousesbased times described in this subsectionare mostly based on those
on the data provided by the Mobara City governmentoffice. recorded on barographsand anemographsand will involve
The first damage to a house occurred 500 m south of the some uncertainty.
intersection of Route 128 and Route 409. The width of the Wind speed and direction. A maximum wind speed of
damage path increased as the tornado moved northward. 30.8 m/s was recorded at 1910 JST by an anemometer at a
The damage path extended to the area around the Japan heightof 10 m on the farm of the Mobara AgriculturalHigh
Railways (JR) Shin-Mobara Station and further to the north. School (point D in Figure 2). The wind direction was SE
The length of the damage path was about 6.5 km, and its before 1850, then changedto NE, and was SE just before the
average width was about 500 m (the maximum width was tornado cut the electric power supply. The anemometer at
1200 m). the Mobara Transforming Station of the Tokyo Electric
Most of the severe damage to houseswas concentrated in Power Company (point F) recorded a maximum speed of
the Takashi area near the corner where the JR Sotobou Line 29.5 m/s at about 1920 JST. The variation of the wind
changesdirectionfrom northwest to north northeast.Figure directionat point F was quite similar to that at point D. The
3 shows the damagein the Takashi area as photographed anemometer of the Environmental Division of the Chiba
from a helicopteron the morningof December 12. Prefecture (point B) showed that the wind direction
 NIINO ET AL. 513

before the interruptionof electricalpower supplywas NE.

The wind directionsat thesethree pointsare consistentwith
the anticlockwise rotation of the tornado observed in the
video.
Pressure. A suddenpressuredrop of about 9 hPa was
recordedat about 1900JST at the ChoseiHigh School(point
C), about400m eastof the tornadopath.This pressuredrop
is considered to be due to both the tornado and the TC. At
pointD, 900 m eastof the path, a pressuredropof 2 hPawas
recordedat about 1900JST, and at the Fujimi Junior High
School(pointE), about2000m westof the path,a dropof 2.5
hPa was recorded. These pressuredrops are consideredto
be due to the TC.

0 500 1000m •9
%ø
o co

$hin-Mobara
St. •
o
Oo o
o
•o Fig. 3. Damage in the Takashi area (courtesy of Kyodo News
% o o
Service).

2.4. Tornado Path

Figure 4 shows the spatial distribution of directions of the

strongestwinds derived from the damage survey. The wind
directions were estimated from the directions in which road
signs, poles, trees, and farm crops were leaning or had
fallen. The wind directions on the east side of the dashed
Bx xC lines are S---SE and those on the west side W---NE. Thus the
M0bara St. wind direction changes markedly across the dashed lines,
which are essentially convergence lines. These lines are
JR Sotobou Line located on the northwest side of the severely damaged area
in Figure 2. These results suggest that the center of the
tornado vortex moved along the dashed lines and the tor-
Route •09 o Fx
nado had a cyclonic rotation, also in accordance with the
xA video images and the records of wind directions.
The tornado moved directly north northeastward from 500
m south of the intersection of Route 128 and Route 409,
where the first damage was noted, to the south of the Mobara
City Gymnasium. The tornado path, as determinedfrom the
distributions of wind directions and damage, became ob-
Fig. 2. Distributionof the damagedhousesand the points where
meteorological data were recorded. The open circles show lightly scure near the gymnasium. The path became evident again
damagedhouses;the solid triangles, severely damagedhouses:and near Route 128 to the north of the gymnasium and extended
the solid circles, completely destroyedhouses. north northeastward. The tornado crossed the JR
 •ind Direction

Shin?•lo__bara•
Station
I•.llI

ymnasium'

DamagedArea

5O

Fig. 4. Distributionof wind directionsobtainedfromthe damagesurvey.The dashedlinesshowconvergence

lines.
The damaged area is enclosed by solid
 NIINO ET AL. 515

Line just southof the Shin-Mobara Station and continuedon

to the north.
The tornadopath seemedto jump near the gymnasium
wherethe eastwardprotuberanceof the damagedistribution
is seen.Two possiblereasonsfor this may be considered:
First, the gymnasium, dimensions of which are 77 m in
width,94 m in length,and21 m in height,couldhaveblocked
theinflownearthegroundto thetornadovortex.Thismight
have resulted in the generationof two vortices on the east
and northwestsidesof the gymnasium.Second,either an
intensificationof the mesocyclonecirculationor an increase High-Vol tage
of surfaceroughness may haveled to the structuralchange
of the tornadovortex.Laboratoryexperiments showthat in
eithercasea transitionamonga thin laminarvortex,a thick
turbulentvortex, and multiplevorticescan occur [e.g., Sotobou-Line
Church et al., 1979; Leslie, 1977]. Further investigation
concerningthe causeof this structuralchangeis needed.

2.5. Maximum Wind Speed

Maximum wind speedswere estimatedfrom damageto
structure. The implicit assumptionshere are that the struc-
tures were damagedby pressureforces due to winds which
were steady and uniform in the vertical direction. Since most
of the structuresusedfor the estimationof the wind speed
were no taller than 3 m, the estimatedwind speedsare likely
to have been affected by local conditions such as the
proximity of other buildings. 0 lOOm 200m
I I ,, ,i
The locationsof the structurestogetherwith the estimated
wind directionare indicatedby solidarrowsin Figure5, and
the estimated wind speedsare given near the arrows. The
largestwind speednear the groundwas determinedfrom the
twisting of a road sign at point H and is 78 m/s; hence the Fig. 5. Distribution of the wind speedsand wind directions (solid
Fujita scale of the Mobara tornado was at least F3. arrows) estimatedfrom damage to structures.The open arrows are
data from Figure 4. The buildings and houses which were com-
pletely destroyed or severely damagedare shaded.
2.6. Translational Velocity
One of the uniquefeaturesof the presentstudyis that the
time data recordedby a singlecomputerat the control center
of a building security company was used to determinethe
translationalvelocity of the tornado. The computerrecords onset of damage to a house near the Shin-Mobara Station.
times and places at which alarm signalsdue to damage to According to the Tokyo Electric Power Company, the
housesor interruptionof power are reported. The accuracy stoppage occurred between 19!5 and 1916 JST, when the
of the time data is 1 s. high-voltagecable was cut. The time lag was estimatedto be
The earliest alarm signal, which signifiedan interruptionof 90-100 s from an account provided by an occupantof the
electrical power supply, came from the Hayano area, 1 km houseand a repetition of her movementsduringthe event.
south of the intersection of Route 128 and Route 409, 2 s The distanceof 1600 m between the high-voltagecable and
before 1913JST (see Figure 6). This time was adoptedas the her housegave the translational velocity of 16.2-18.0 m/s,
time of the tornadogenesis in the presentstudy. which agreesreasonablywell with the previousestimation.
Four signalswhich signified direct damage to buildings Thus one may conclude that the tornado moved north-
came from the Takashi area at 1, 7, 10, and 37 s after 1915 northeastwardat the speed of 16 m/s.
JST. The distancesbetweenthe four pointsin the Takashi The translationalvelocity of the tornado is closelyrelated
area and the point in the Hayano area togetherwith the time to the movementof the TC which spawnedthe tornado. As
differences gaveestimatesfor the translationalspeedof 13.4, will be describedin section 3, the translationalvelocity of
14.8, 16.5, and 17.8 m/s, the averagebeing 15.6 m/s. the TC obtainedby the singleDoppler radar at MRI between
A further estimate of the translationalspeedwas made 1909and 1917JST was 14.9 m/s, which again agreeswell
from the time lag between the stoppageof power and the with that of the
516 OBSERVATIONS OF THE MOBARA TORNADO

3.2. Parent Storms

Sh i n-Moba ra St,
During the evening of December 11, two isolated strong
cells moved northeastward off the southern coast of Honshu
accordingto the Haneda radar (Figure 8). The eastern cell,
hereinafterreferred to as cell A, became detectableby the
Tokyo radar at 1322 JST, when it was 80 km south of
Hamamatsu. It lasted for more than 6 hours and maintained
an isolated structure throughout its lifetime. After 1730,
when it reached the Boso Peninsula, a vault-shapedecho
region was clearly identified on both the Haneda and Narita
19' 15'01' radars (see Figure 1), and a mesocyclone circulation was
found from an analysis of the surface wind and pressure
data. Thus cell A showed many interestingfeatures that are
19:15'37' characteristicof a supercell,which is very rare in Japan. It
19:15'07' produced at least one tornado in Kamogawa and heavy
winds in Maruyama and Choshi. The analysisof the phe-
:15'10'
nomenaassociatedwith cell A is reported elsewhere [Niino
et al., 1991].
19:14' 43'
The western cell, hereinafter referred to as cell B, became
Mobara-St, detectable by the Tokyo radar by 1500 JST, when it was 70
km south of Hamamatsu. It crossed the Izu Peninsula at
around 1700 JST. It merged with a cell to the north and
exhibited a fairly complex structure after it reachedthe Boso
Route Peninsulaat around 1815 JST. Cell B produced heavy winds
N at Futtsu and Kimitsu from 1830 to 1850 JST and the tornado
in Mobara City from 1913 to 1920 JST.

•-"19' 12'58' 3.3. Tornado Cyclone

0 lkm
The singleDoppler radar at MRI was operating after 1840
!. I,, , I
JST. Its resolution is 2 km in both the vertical and crossbeam

Fig. 6. Locations of buildings and houses where the damage or

stoppageof electricitywas monitoredby a computerof a building
security company. The southernmoststar showsthe location where
the high-voltage cable was cut, and the northernmost star shows the Surface map
location of the housewhere the eyewitnessaccountwasgiven by its
130E
occupant. 1020

3. MESOSCALE ANALYSIS

L
nr
3.1. Synoptic Conditions

Figure 7 showsthe surfaceweather map about 1 hour prior

to the tornado in Mobara City. It is seen that the tornado (':1'000
occurred in the warm sector of a developing extratropical 002'
cyclone (cf. Figure 1). At 0900 JST a cyclone with central
pressure 1004 hPa was located east of the Korean Peninsula
(see Figure 1). It moved eastward while continuing to
develop and reached the Noto Peninsula by 1500 JST. At
1500 JST a new cyclone was generated on the cold front 140g
extending southwestward from the original cyclone. It
moved eastward along the southern coast of the island of 1800JST, Dec. 11,1990
Honshu. The new cyclone reached 20 km north of Tokyo at
2100 !ST. Fig. 7. Surfaceweather map at 1800JST, December ! 1,
 NIINO ET AL. 517

directionsand0.5 km in the beamwisedirection.A regionof

highvorticity which correspondsto the TC wastrackedfrom
1840to 1924 at 1 km above groundlevel (AGL). The core
diametersof the tornadocyclone,asdefinedby thedistance
E]MRI
betweenwind maxima,are shownin Figure9 by the bars.

single
Doppler
RadarX
The center of each bar indicates the center of the TC at
successivetimes.The numberto the rightof the centergives
the time in JST, and the numberto the left gives the
maximumvorticityin unitsof 10-3 s-•.
The black regionsshow the damagepathsdue to heavy
winds in Futtsu and Kimitsu and due to the tornado in
Mobara. These areas of damage occurred when the TC
passed close to those regions. When the tornado or the
damage due to heavy winds occurred, the diameter of the TC
was relatively small. The open circles show the locations of
barographs. Each barograph recorded a suddenpressure
drop of about 2 hPa at about the time when the TC passed
near the barographs.
Figure 10 showsthe surfacewind flow and the tempera-
ture distributionderivedfrom surfacemeteorologicaldata at
1900 JST. In addition to the wind data at 1900 JST shown in
the figure, wind data generatedby the time-spaceconversion
technique [Fujita, 1963] from 1850 to 1910JST were used to
draw the flow pattern, where the storm systemwas assumed
to move from 9 ..36.5o at a speed of 16 m/s.

0 20
I !

Haneda
Radar :..

• Mobara
/ ,•":'-,
Choshi
Fig. 9. The movement of the TC accompanied by the Mobara
(19.13-1920)
:. (-•1940) tornadoas revealedby the MRI singleDoppler radar. The thick line

Futtsu.
Kimitsu /
showsthe path of the center of the TC.

(1820-1850)
/ 1930 "• The temperature contour lines (dashedlines) were drawn
.. 1930 on the basis of data from the Automated Meteorological
Data Acquisition System (AMeDAS), whose stations are

...'
i830
•""••L••-1900
.............
'-...'"'-.-..
distributedfor each 20-kin by 20-kin square on average. The
contour interval is 1ø, and the numbers on the contour lines
.. are temperaturesin degreesCelsius.The data showa fairly
steeptemperaturegradient over the southernpart of the
Boso Peninsula.
The wind field at 1900 JST clearly shows a convergence
1737..=
• • (1750-1800)line
.?'O,,00
1700 1035 (----1800)
(indicated by the double line) where the warm southerly
wind meets the cold northerly wind. The southwestend of
the convergenceline nearly coincideswith the location of
Ce11.BCel1.A the TC at 1902 in Figure 9.
I.... t ........ I, , ! !
Figure I1 shows the areal distribution of precipitation
o 50 100km intensityderived from the Tokyo radar divided into 2.5-km
Fig. 8. Evolutionof two tornadicstormsas seenby the Haneda by 2.5-kin grid spaces.The positionof the convergenceline
radar. The contourlinesare drawnfor precipitationintensitiesof 4, in Figure 10 coincideswith the southeastedgeof the region
16, and 64 mm/h. The radar echoof cell A at 1737JST is enlarged wherethe precipitationrate is greaterthan 16 mm/h. This
and shownin the upperleft comer, where the contourlinesare
drawnfor 1, 4, 16, and64 mm/h.The timesandlocationsof damage suggests that the convergenceline wasproducedby precip-
due to tornadoes and heavy winds are also shown. itation-induced cold moist air spreading
518 OBSERVATIONS OF THE MOBARA TORNADO

; /

/ '-,,,
//
f t
(
/
t

' ,, ' / !
/

/
//
/ /
/ /
!
//
/

/%-

1900JST, Dec. 11, 1990

50km

Fig. 10. The wind flows and the temperature distribution near the surfaceat 1900 JST, December 11, 1990. The thick
solid lines with arrows show the wind flows. The surface winds at certain AMeDAS stations and at several stations of
local governmentsat 1900 are shown. One barb correspondsto 1 m/s.

againstrelatively warm southerly wind. It is also interesting If we compare the characteristic features of this storm
to note that there is a region of relatively low intensity with the model for tornadic thunderstormsby Lemon and
precipitationat the southwestend of the convergenceline, Doswell [1979], the main difference is that no evidence of a
which correspondsto the location of the TC. rear flank downdraft is found in Figure 10. Several reasons

Fig. 11. The radarechoon the Tokyo radarat 1900JST, December11, 1990.The weak echoregionis located7.5 km
(threegrid squares)north of the tip of the arrow labeled"tornado
 NIINO ET AL. 519

for this can be considered:

First, cell B mergedwith a over Japan and over the United States are generally quite
northern cell before it reached the Boso Peninsula and different, it is of interest to examine what percentage of
exhibited a complexstructure (seeFigure11).Thusit may tornadoes are generated by superceLIsor by shear lines in
not have been a typical supercelland doesnot needto have Japan. It is also of interest to examine how the structure of
showna structuresimilarto Lemon and Doswell'smodel, a supercell storm in Japan, if a unique structure exists,
although a TC had been detected for more than 40 min. compares with the classical model for those in the United
Second,giventhat the atmosphere overJapanis generally States. We hope to answer these questionsin the near future
more humid than that over the United States,therewas no when we have accumulated several more case studies like
dry southwesterlyflow in the middletroposphere,and a the present one.
significant
downdraft
andgustfrontcouldnotdevelop.This
may then have resulted in a different structure for the
tornadic storms.To arrive at any more definiteconclusions, Acknowledgments. The authors are grateful to the Prefectural
Officeof Chiba, the City Office of Mobara, the Japan Meteorological
further researchon tornadic stormsin Japanwill be re- Agency, the Mobara Agricultural High School, the Chosei High
quired. School, the Fujimi Junior High School, the Tokyo Electric Power
Company,and the information media for providingthe variouskinds
of data concerningthe tornado event. The authors are also grateful
4. SUMMARY AND CONCLUSIONS
to H. Nagamatsu(BRI) for cooperation in the damagesurvey, to K.
The characteristics of the Mobara tornado on December Sasaki for providing a copy of his video tape, and to anonymous
referees for useful comments. The present investigation was made
11, 1990,were derived from a damagesurvey;the mesoscale as part of a cooperative project between MRI and BRI entitled "An
featureswhich led to the tornadogenesiswere analyzedfrom Investigation on the Characteristics of the Air Flows Associated
the surfacemeteorologicaldata, a singleDopplerradar, and With the Tornado in Mobara City."
three conventional radars.
The cyclonically rotating tornado moved north-northeast- P•EFERENCES
ward at the speedof 16 m/s andproduceda damagepath 6.5 Brooks, E. M., The tornado cyclone, Weatherwise, 2, 32-33, 1949.
km in length and 1.2 km in maximum width. The maximum Browning, K. A., Air flow and precipitation trajectories within
wind speed near the ground was estimatedto be at least 78 severe local storms which travel to the right of winds, J. Atenos.
m/s. A suddenwideningof the damagepath and a disconti- Sci., 21,634-639, 1964.
nuity in the tornadopath were found near the Mobara City Church, C. R., J. T. Snow, G. L. Baker, and E. M. Agee,
Characteristics of tornado-like vortices as a function of swirl
Gymnasium;these developmentsseem to suggestthat the ratio: A laboratory investigation, J. Atmos. Sci., 36, 1755-1776,
tornado vortex experienced some structural change when 1979.
the vortex approached the gymnasium. Fergusson, E. W., F. P. Ostby, and P. W. Leftwich, Jr., The
The single Doppler radar at MRI detectedthe TC at 1 km tornado season of 1986, Mort. Weather Rev., 117, 221-230, 1989.
AGL at least 33 min before the tornado formed and was able
Fujita, T. T., Analytical Mesometeorology: A Review, Meteorol.
Monogr. 27, pp. 77-125, American Meteorological Society, Bos-
to track it for 44 min. In the cities of Futtsu and Kimitsu, ton, Mass., 1963.
damage to several houses was produced by heavy winds Fujita, T. T., Proposedcharacterization of tornadoes and hurricanes
when the TC passed nearby. The degree of coincidence by area and intensity, SMRP Res. Pap. 91, 42 pp., Dep. of
Geophys. Sci., Univ. of Chicago, Chicago, Ill., 1971.
between the location and damage path of the Mobara tor-
Fujita, T. T., K. Watanabe, K. Tsuchiya, and M. Shimada, Ty-
nado and the location and path of the TC was excellent. phoon-associatedtornadoes in Japan and new evidence of suction
Suddenpressuredrops were observedalong the path of the vortices in a tornado near Tokyo, J. Meteorol. Soc. Jpn., 50,
TC when it passednear barographs.Analysis of the surface 431-453, 1972.
wind fields showed that the TC was located at the southwest Lemon, L. R., and C. A. Doswell III, Severe thunderstorm evolu-
tion and mesocyclone structure as related to tornado genesis,
end of a convergenceline which formed betweenthe warm Mon. Weather Rev., 107, 1184-1197, 1979.
southerlywind and the cold northwesterlyair flowingout of Leslie, F. M., Surface roughness effects on suction vortex forma-
the region of intense precipitation. A region of weak radar tion: A laboratory simulation, J. Atmos. Sci., 34, 1022-1027, 1977.
reflectivity also seemedto correspondto the locationof the Mitsuta, Y. (Ed.), Studies on wind disasterscaused by tatsumaki
TC. (tornadoesand waterspouts) and severe local stormsin Japan (in
Japanesewith English abstract), final report, Spec. Res. Proj. for
This is the first study in Japanin which a TC was detected Nat. Disaster, 124 pp., Minist. of Educ., Tokyo, 1983.
by a singleDopplerradar and at the sametime the flow field Niino, H., Y. Ogura, I. Takayabu, and H. Nirasawa, A mesoscale
aroundthe TC was analyzed on the basisof surfacemeteo- analysisof the tornadoes and heavy winds in the Chiba Prefecture
rological data. In the United States it is generallybelieved on December 1I, 1990 (in Japanese),in Proceedingsof the 1991
Spring Meeting of the Japan Meteorological Society, Japan
that there are two kinds of tornadogenesis:one is associated Meteorological Society, Tokyo, 1991.
with supercellstorms,and the other is associatedwith shear Wakimoto, R. G., and J. W. Wilson, Non-supercelltornadoes,Mon.
lines [Wakimoto and Wilson, 1989]. Since moisture content Weather Rev., 117, 1113-1140,
 Discussion

GREG FORBES, SESSIONCHAIR

Pennsylvania State Uni•,ersiLv

PAPER I 1 for doingthe damage.Thus we may be underestimatingthe

Presenter, Ted Fujita, University of Chicago [Fujita and actual wind speed.
Smith, this volume, Aerial survey and photographyof tor- (Marshall) That is true. We may also make a totally errone-
nado and microburst damage] ous F scalerating on the basis of the location of the building
(Dave Keller, National Severe Storms Laboratory) All of [in the tornado path].
your damageslideslooked like they were taken under sunny (Joe Golden, National Oceanic and Atmospheric Adminis-
and calm conditions. I assume that your surveys were tration, Office of the Chief Scientist) Why do you think that
conductedon the day after the tornado. What is the range of
it has taken so long to get some of the simple and inexpen-
weather conditions on the day after the event, and how do
sive procedures, that you described and that have been
the conditionsaffect your damage surveys?
around for 20 years, into common construction practice? Is
(Fujita) We like to fly on the next morning, but getting an it because building codes do not require them? Is the
aircraft that soon is difficult. To survey the Plainfield [Illi- insurance industry to blame?
nois] tornado [August 28, 1990] damage, we took off at 1
(Marshall) It is all of the above. The insurance industry does
P.M. A few days ago, we did a survey of the damagefrom
not promote this becausethey offer replacement cost cover-
the March 27th [1991] Lamont [Illinois] tornado, which was
age. This does not encourage better construction. Building
on the ground for 16 miles. The next day was clear, but the
turbulence was terrible. We had to learn how to hold a codes are great, but cities and municipalities are hurting
financially, so they do not have the staff to enforce them. So.
camera steady while the aircraft was shaking. Quite often,
the weather the next day is very good, but windy and I am probably going to have to preach the same messagefor
turbulent. Maybe you have to withstand 1.5 #'s all the time. the next 20 years. This is symptomatic of the punic's
When we came to Oklahoma to survey the Lahoma and conceptionsof tornadoesand hurricanes. They first learned
Orienta tornadoes [of May 2, 1979], we had to wait in about tornadoes exploding buildings and opening windows,
Oklahoma City for 2 days because of dust being blown from and we spend the next 20 years trying to correct that. But
[New] Mexico. We couldn't fly because visibility was less people are very reluctant to give up the old beliefs.
than 1 mile. So dust may be a problem the next day in (J. T. Lee, Cooperative Institute for Mesoscale Meteorolog-
Oklahoma. But east of Oklahoma, the next day is best. It's ical Studies)JohnWeaver wrote a paper [on anchoringroofs
not cloudy, but very turbulent. to walls] some 12 years ago [available from the National
Severe Storms Laboratory]. The original instruction on
PAPER 12 windows was to open the windows on the oppositeside to
the direction from which the tornado was approaching,
Presenter,Tim Marshall, Haag EngineeringCo. [Marshall, because if the windows [on the windward side] blew in, the
this volume, Lessons learned from analyzing tornado dam- wind would blow through the building and the structure
age] would be saved [by reducing the pressure force acting to
(Don Burgess, National Severe Storms Laboratory) One blow out the leeward wall]. Somehow, a misconception
concern that I have about engineeringanalysis is that very arose that the windows on the windward side, not on the
often the estimate is for the minimum wind speedrequired leeward side, were the ones supposedto be opened.

The Tornado: Its Structure, Dynamics, Prediction, and Hazards. (Marshall) I think that it was just an oversimplification.
GeophysicalMonograph79 People just said, "Well, heck, in order to make my house
Copyright1993by the AmericanGeophysicalUnion. stay there, I will open all the windows."
522 DISCUSSION

PAPER 14 (Chris Church, Miami University of Ohio) Do you have

Presenter, George Marshall, National Weather Service, any explanation for why the structure of the tornado
Lubbock, Texas (for Gary Woodall of same affiliation) changed?
[Woodall and Mathews, this volume, Survey of a violent
tornado in far southwestern Texas: The Bakersfield Valley (Niino) Part of the reason why the structure changed may
storm of June 1, 1990] be that the tornado was passing the last building. But the
vortex structuremay have changedanyway becausethe path
(Joe Golden, National Oceanic and AtmosphericAdminis- width was increasing continuously. i don't have a definite
tration, Office of the Chief Scientist) After the Saragosa answer.
[southwestTexas] tornado [of May 22, 1987], people noted
that violent tornadoes are extremely rare in this part of the (Church) I would like to suggestthat there was an increase in
country. Do you think that it was just coincidence that swirl at that particular point. Upstream the tornado had a
another violent tornado occurredjust 4 years later, or have narrow core. The increase in swirl caused a vortex break-
we undersampled tornadoes in this part of Texas? Also, down to descend very close to the surface, producing a
could you elaborate on the poisongas that you referred to? maximum in intensity ...
(Marshall) I don't doubt that we miss some small tornadoes, (Niino) That is one possibility.
but I hope that we haven't missedany of this magnitude.It's
just coincidencethat two violent tornadoesoccurredso close (Church)... and that might have been triggered by a change
togetherin time. Concerningthe poisonousgas,on the one side in surface roughness.A transition from a very rough surface
of the ridge, wherethe 57 pumpjacks were knockeddown, the with a lot of housing to one that is smoother seems to be a
well was not under pressure.On the other side of the ridge, dangerous place.
that oil was under pressureand had hydrogensulfide, which
(Don Burgess, National Severe Storms Laboratory) Was the
is very potent. I don't know if the gas would have travelled
antenna tilted to measure the vertical extent of the high-
to Iran. The tornado decayed about 4 miles southwest of
vorticity region?
Iran, but part of the storm passedover Iran. People at Iran
told me that only two whiffs of that gas and you are dead. So (Niino) Unfortunately, Mobara City is about 75 km south of
the tornado could have been much more disastrous. our institute, and the beam width there is about 2 km. The
results that I presented were for 0 to 2 km. Since the tornado
PAPER 15 occurred last December, we haven't carefully looked at all
Presenter, H. Niino, Meteorological Research Institute, the data yet. But at some time the circulation extended to at
Tsukuba, Japan [Niino et al., this volume, An Observational least 2 to 4 km. We are planning to make a detailed
study of the Mobara tornado] investigation into this
 Damage Mitigation and Occupant Safety
JAMES R. MCDONALD

Institutefor DisasterResearch,TexasTech University,Lubbock, Texas 79409

INTRODUCTION large segment of the farm population moved to the city,

cellars were constructed in backyards but were used mostly
Each year tornadoeskill and injurea significantnumberof for shelter.
people and, at the same time, cause millions of dollars in When shelters that are specifically designedto withstand
property damage.Althoughtimely watchesand warnings tornado-inducedforces are not available, people must rely
issued by the National Severe Storms Forecast Center and on those areas in a building that are least likely to be
local NationalWeatherServiceofficehave resultedin sig- damaged,where the chances of survival are relatively large.
nificantreductionsin deathsandinjuriesin the last 10years Such areas are designated"protective areas." When desig-
[National Oceanic and AtmosphericAdministration, !991] nated by a knowledgeablestructural engineer, a protective
the needfor readilyavailablesheltersremainsan important area is the best available space for occupantprotection in a
issue.The NEXRAD Doppler radarsthat are scheduledfor particular building.
installationacrossthe country beginningin 1993have the
potentialfor increasingthe lead time for tornadowarnings,
which should allow people to take precautionsand seek Tornado Shelters

shelter when tornadoes threaten (Golden, 1989]. However, Two forms of shelters are described herein: an in-
early warnings are of little value if there are no places
residence shelter and a community shelter. The in-residence
available for protectionagainstthe storms.
shelter is an abovegroundcloset, bathroom, or hallway that
Property damagefrom severe windstorms, includingtor-
has been strengthened to resist the wind forces and missile
nadoes, continuesto escalate each year. The magnitudeof
impacts produced by very intense tornadoes. The commu-
the losses has reached a level where insurance alone is no
nity sheltercanbe freestandingor incorporatedin a building.
longer able to cover the private losses. The costs to local,
In-residence shelters. The concept of an in-residence
state, and federal governmentsare having detrimental effects
shelter was developed when engineers at Texas Tech Uni-
on the economicsof the affected regions. The potential for
versity frequently observed that a small interior room or
damage in the future is even more ominous.
closetnearly always remained standing, even thoughthe rest
The objective of this paper is to review the stateof the art
of the house was completely destroyed [Kiesling et al.,
for providing occupantprotectionand for mitigatingtornado
1977]. Engineering attention to the room could assure that
damage. Both the state of knowledge and the extent of
the room would remain intact againstvirtually all tornadoes.
implementation of this knowledge into practice will be
Resisting the wind-induced forces is relatively simple.
examined. Future research needs are also indicated.
Specialconsiderationhad to be given to the impact of flying
debris. The design criteria for in-residence shelters is based
OCCUPANT PROTECTION on the wind speedsassociated with a tornado whose inten-
sity correspondsto an F-4 Fujita scaletornado (F-5 beingthe
The term "occupant protection" appliesin this contextto most intense) [Defense Civil Preparedness Agency, 1975].
anyone needing protection when severe weather threatens. Tornadostatisticsfrom 1916to 1979suggestthat at least99%
The first sheltersagainsttornadoesin the United Stateswere of all recorded tornadoes are less than or equal to F-4
root cellars found on almost every farm in the midwest. intensity. The criteria translates into a tornado with a 260
These undergrounddugouts served as cool damp placesfor mi/hr (mph) (1 mph = 1.6 krn/hr) wind speed. The atmo-
the storage of foodstuffs, as well as tornado shelters. As a sphericpressureinducedforces can be negatedby providing
The Tornado: Its Structure, Dynamics, Prediction, and Hazards.
sufficient venting area in the wall or roof of the shelter.
GeophysicalMonograph79 Adequateventing area allows the pressureto equalize as the
Copyfight 1993 by the American GeophysicalUnion. tornado passesover the shelter.
524 MITIGATION AND SAFETY

Fig. I a

Fig. lb

Fig. I. Two design concepts for in-residence shelters: (a) concrete masonry construction, recommended for new
construction, and !,b) plywood construction on walls and ceiling, recommended for existing construction.

Missile impact resistancepresenteda more complex prob- most common missiles were pieces of timber from damaged
lem. A wide range of missiles, including roof gravel, insula- or destroyed roof structures. For these reasons the missile
tion, sheet metal, pieces of timber, pipes, automobiles, and criteria for shelterdesignis a 2 x 4 timber plank weighing 15
storagetanks, have been observedin tornado damagepaths. lb (1 lb - 0.45 kg) impactingon end at 100 mph. On the basis
Ordinary residential walls are capable of resisting roof of extensive tests at Texas Tech University [McDonald and
gravel, insulation, and sheet metal. Walls of heavily rein- Bailey, 1985; McDonald and Kiesling, 1988], two shelter
forced masonry or concrete are required to resist the pipe, designs have emerged: one for installation in an existing
automobile, and larger missiles. Furthermore, damage inves- home and one more suitable for new construction [Institute
tigations through typical residential areas revealed that the for Disaster Research, 1989]. (Plans for construction
 MCDONALD 525

in-residence shelter can be requested from the Institute for

Disaster Research, Texas Tech University. Lubbock. Texas
79409-1023.)Figure 1 illustratesthe two concepts.
3
The shelterfor an existinghomehasfour layersof • inch
(1 inch = 2.54 cm) plywood attached to the ceiling and walls
from inside the room. These four layers of ply•,•ood are
capableof stoppingthe designmissile.The advantageof this
concept is that all construction related to the shelter can be
done from inside the room or closet.
With construction of a nexv house, 8-inch concrete ma-
sonry,groutedand reintbrcedwith one :•-inch-diameter
BI ¾. K. (number 3) reinforcing bar in each vertical cell, is used for
the walls. A 4-inch reinforced concrete slab is normally used
fortheroofof theshelter,although
thefourlayersof :]-inch
plywood would be acceptable, if properly anchored.
Common features of the two shelter concepts include a
sliding door capable of resisting the missile impact and vent
pipes to allow equalization of pressure. The shelter must be
securely anchored to a concrete floor slab. The in-residence
shelter is not recommended with a pier and beam foundation
unlessa special foundation is designedfor the shelter itself.
Statistics are not available on how many in-residence
shelters have been built. The Institute for Disaster Research
at Texas Tech University distributes more than 100 sets of
plans each year. There are many advantages to in-residence
shelter. the main one being convenient access from any
location in the house without having to go outdoors. The
spacecan be utilized for other purposes•such as a closet or
bathroom.Costsof an in-residenceshelter range from $1500
to $3000.
Fig.2a Several commercial ventures to build and market propri-
Fig. 2. Two examples of proprietary tornado shelter concepts: (a)
etary shelters have been attempted. Aboveground commer-
conical reinforced-concrete shelter, where door as shown did not cial conceptsthat have been evaluated at Texas Tech include
meet tornado missile impact criteria, and (b) shelter bed consisting a conically shaped concrete structure, a reinforced bed
of a heavy steel box that is bolted to the concrete foundation. frame (shelter bed), and a plate steel structure. Figure 2
Neither concept has achieved financial success. shows the concrete shelter and the shelter bed. Neither of
these has been economically successful.

Fig.
526 MITIGATION AND SAFETY

Fig. 3. Freestandingabovegroundcommunityshelterconstructedof reinforcedmasonrywalls and a roof with 4-inch

concrete slab supported on closely spaced open web steel joists.

Comrnunity shelters. Many situations arise where there Community-type shelters should be readily accessible.
simply are no safe places for people to seek protection from Although somewhat more expensive, two smaller shelters
a severe windstorm. Manufactured home (mobile home) placed in close proximity to the potential occupants are
communities are the first that come to mind. Because of the better than one large shelter located some distance away.
uncertainty of the effectiveness of ground anchors and the Community sheltersare often incorporated into larger build-
large area to weight ratio of manufactured homes, occupants ings. One architectural engineeringfirm in Lubbock, Texas,
are advised to evacuate when severe thunderstorm or tor- has successfullyincorporatedcommunity sheltersin a number
nado warnings are issued. Underground or above ground of school constructionprojects [Harris and Mehta, 1991].
shelters with a capacity of up to 200 people can be built to
the same criteria stated for the in-residence shelters. Allow-
Protective Areas
ing approximately
6 squarefeet (1 ft2 = 0.096 m2) per
person, a shelter 30 feet x 40 feet (9.1 m x 12.2 m) in plan The identification of protective areas in existing buildings
will hold approximately 200 people. grew from experiencegainedin wind damageinvestigations.
The reinforced and grouted masonry walls of an aboveg- By combining the experience of many damage studies,
round shelter are capable of resisting impact of the 2 x 4 performancesof certain classesof buildings are reasonably
timber missile, as well as the 260-mph wind-induced forces. well understood. Factors that contribute to adequate perfor-
Venting is provided to equalize the atmospheric pressure mance of structures in windstorms are (1) degree of engi-
change forces. A concept illustrated in Figure 3 uses open neering attention to the design, (2) construction matedhal,(3)
web steel joists to support a 4-inch-thick concrete roof slab. reserve strength, (4) ductility, and (5) redundancy.
By supporting the slab on the joists, it does not have to be The engineer identifying protective areas first gains an
designed as a long-span structural element. Slab weight and understandingof the building construction and the structural
thickness offset uplift forces and missile impacts. Careful system. Possible failure modes are identified. With the
attention must be paid to the structural integrity of the failure modes in mind, areas of relative safety for building
building, so that no weakness exists in the load paths. This occupantsare identified on building floor plans. Consider-
type of shelter has been constructed in manufactured home ation is given to accessibility, as well as safety from hazards
communities, at schools where protective areas in the school that may exist in the building. In buildings without base-
building are not adequate to hold the students and staff, and ments, protective areas are designated in the interior of
at manufacturingplants where processinghazards preclude buildings with at least two walls between the protective
the use of the building itself for protection of people. A large areas and the outside. Small rooms are preferred, because
number of peopleworking in a light-frame steelbuildingmay the partition walls are less likely to collapse and are more
require construction of a community shelter to provide for likely to support the roof elements, should they collapse.
their safety. Protective areas can be designated in schools,
 MCDONALD 527

homes,officebuildings,shoppingmalls, and other public Peer reviews of tornado-resistantdesignsare recom-

places. An elementary school teacher in Huntsville, Ala- mendedfor thosefacilitiesthat havehighpotentialimpacton
bama (1989), movedher classfrom a gymnasiumwhich had the environment(for example,nuclearpower plants,certain
a large, open span to a designatedprotective area and critical DOE facilities). The designsmust then be executed
probably savedthe lives of her 13 students.The gym roof using good construction practices that also involve strict
collapsed to the floor under the tornadic forces. In the quality control.
designatedprotectiveareasthe studentsreceivedonly a few
minor scratches,eventhoughthe schoolbuildingwas almost
Evaluation of Existing Structures
totally destroyed.
Efforts shouldbe put forth to get protectiveareasidenti- Years of wind damage documentation have produced a
fied in all publicbuildings.Floor plans and signsare needed body of knowledge of building performance which allows
to quickly directbuildingoccupantsto the protectiveareas engineersto evaluate the potential performance of existing
when severe weather threatens. buildingsin windstorms [Menta et al., 1981]. Two levels of
assessmentare possible: one qualitative (level I) and the
DAMAGE MITIGATION other quantitative (level II). The level I procedure is based
on a walk-down of the facility. Depending on the structural
Twenty or more years of research and experience in system, components,and cladding, potential failure modes
dealingwith tornadodamagehasnot resultedin a significant are postulated, on the basis of performance of similar
reductionof tornadodamage.In fact, becauseof escalating systemsin windstorms. The level I procedure is ideal for
building costs and urban growth, the damage continues to identifying protective areas in buildings, for developing
increase. This section addressesthe question;What is re- emergency plans, or as a preliminary assessmentprior to
quired to mitigate tornado damage? conductinga level II procedure. With the level II procedure,
failure modes are postulated, and by performing engineering
Design of Structures
calculations, the wind speeds to produce the postulated
Knowledgeablearchitectsand engineersare needed, who damage can be calculated. If a tornado hazard assessment
will insist on constructionpractices that will lead to wind- has been performed for the site, the probabilisticrisks of the
resistantconstruction.Wind designin the pasthasreceived damagecan be determined. Knowing the risks is important,
very little attention in the engineeringor architecturalclass- if decisions on extensive retrofit must be made when a
room, becausethe problem is not perceived asbeing serious. facility does not meet acceptable criteria.
More emphasis on wind loading is needed in undergraduate Strengtheningand retrofit of existing structureshave not
and graduate courses. Continuing education courses for been used to any great extent to improve wind resistivity.
practicing engineersand architects should be offered. Repair of wind-damaged structures seldom incorporate im-
In the last 10 years, more and more companiesare coming provementsin the designs. The buildings are simply rebuilt
to realize that tornado-resistant design is necessaryto pro- exactly as they were before the storm. Retrofit projects for
tect facilities that are critical to the company's operation. earthquake resistance in California are becoming popular.
NCR Corporation designedits corporate computerfacility in Efforts are needed in the future to encourage improved
Dayton, Ohio, to be tornado-resistant. The American Air- building designs following major windstorm events. The
lines computer facility in Tulsa, Oklahoma, and Indiana Bell problem is particularly difficult in the case of tornadoes.
in Indianapolis, Indiana, along with other high-valuefacili- Once hit by a tornado, people tend to think their chancesof
ties, have been designedto resist the effectsof tornadoes. being struck again are extremely small.
The Department of Energy is engagedin a naturalphenom-
ena hazards project to evaluate and upgrade all of their Reluctance to Invest in Tornado Protection
facilities. Those facilities, judged to be moderate or high
hazard, relative to injury of people or damageto the envi- The major reason progress has been slow in tornado
ronment, considertornadoesin the designcriteria, if torna- damage mitigation is the reluctance of owners to invest in
does are a viable threat at the site. tornado protection. From a layman's view the risks of
The processof tornado-resistantdesigninvolvesclassify- tornado damage seem very low. Given the low risks, there
ing the facility with regard to importance or hazard threat, are very few significantfinancial incentives, becauseinsur-
performing a site-specific hazard probability assessment, ance or federal aid minimizes financial losses. Building code
and selectinga designwind speedand designprocedure(for requirementsdo not include considerationof tornadoesin
example, allowable stressdesignor strengthdesign)that are design. Furthermore, there is a general perception that
consistentwith establishedperformance goals.Connections added tornado protection is very expensiveand disruptive.
and anchorage details must assure continuity in the load Insurance rates for extended coveragedo not reward owners
paths. Finally, considerationmust be given to impact resis- for investing in more than the minimum wind resistivity in
tance of tornado-generatedmissiles.This processis usedby buildingsrequired by building codes.
the Department of Energy (DOE) for designand evaluation Another factor that contributes to reluctance of an owner
of its facilities [Kennedy et al., 1990]. to design a new building or retrofit an older one is
:528 MITIGATION AND SAFETY

excessiveconservatismthat sometimescreeps into the de- loads. Use of this knowledge would provide better protec-
signprocess.Pointsof conservatismcan be heapedone upon tion of occupantsand would mitigate property damage.
the other, especially in evaluating the resistanceof a struc- While there are still a number of technical questions to be
ture. Conservativevalues of matedhalproperties, conserva- answered, the major problem is in implementation of the
tive load factors (or factors of safety), and conservative knowledge into practice. Until this is done on a broad scale,
limits on structural response all tend to add costs to a the damage from tornadoes each year will continue to be
project. Reluctance to utilize inelastic responsefor wind very high.
design also adds to costs.
REFERENCES

ADDITIONAL RESEARCH NEEDS Defense Civil PreparednessAgency, Interim guidelinesfor building

occupant protection from tornadoes and extreme winds, Rep.
Convincingthe professionalcommunity, as well as the TR-83A, U.S. Dep. of Del., Washington, D.C., 1975.
Golden, J. H., The coming deluge of wind data for wind engineers,
owners of the built environment, to construct more wind-
in Proceedings,6th U.S. National Conference on Wind Engineer-
resistantnewbuildingsandto strengthenandretrofitexisting ing, Houston, Tex., 1989.
ones is a societal problem. Solution to this problem is Harris, H. W., and K. C. Mehta, Design of tornado protective areas
long-termand, in fact, will never be complete. in schools,paper presentedat Ninta, StructuresCongress 1991,
There are several areas where additional technical re- Am. Soc. Civ. Eng., Indianapolis, Indiana, April 29 to May 1,
1991.
search,from an engineeringperspective,is needed. Highly Institute for Disaster Research, Inresidence shelter for protection
resistant, but economical, roofing systemsare needed. Cur- from extreme winds, pamphlet, Tex. Tech Univ., Lubbock, 1989.
rent roof compliancetestsdo not correctly representthe true Kennedy, R. P., S. A. Short, J. R. McDonald, M. W. McCann, Jr.,
effects of the wind on roofs. R. C. Murray, and J. R. Hill, Designand evaluationguidelinesfor
department of energy facilities subjected to natural phenomena
Designinga buildingto resisttornado-inducedloads tends hazards, Rep. UCRL 15910, U.S. Dep. of Energy, Washington,
to preclude the use of windows in exterior walls. Flying D. C., 1990.
debris tends to break the windows, allowing wind and water Kiesling, E. W., K. C. Mehta, and J. E. Minor, Protection of
to circulate through the building. Research is needed to property and occupants in windstorms, Rep. 27D, Inst. for
Disaster Res., Tex. Tech Univ., Lubbock, 1977.
develop missile-resistantshutters or grills that could be McDonald, J. R., and J. R. Bailey, Impact resistance of masonry
closed when tornadoes threaten. The devices need to be
walls to tornado-generatedmissiles, paper presented at Third
capableof stoppingthe 2 x 4 timber plank missile.Heavier North American Masonry Conference, The Masonry Soc. and
missiles,comparableto a 3-inch-diametersteel pipe, are not Univ. of Tex., Arlington, Tex., June 3-5, 1985.
as common as the timber plank and do not fly as readily in McDonald, J. R., and E. W. Kiesling, Impact resistance of wood
the wind.
and wood productssubjectedto simulatedtornadomissiles,paper
presentedat InternationalConferenceon Timber Engineering,
Wash. State Univ., Seattle, Wash., Sept. 19-22, 1988.
Mehta, K. C., J. R. McDonald, and D. A. Smith, Procedures for
CONCLUSIONS
predictingwind damageto buildings,J. Struct. Div. Am. Soc.
Eng., 107 (STll), 2089-2096, 198!.
Twenty years of tornado research have produced much National Oceanic and Atmospheric Administration, Climatological
knowledge than can be used to design new buildings and Data, National Summary, Environmental Data Service, Wash-
retrofit existingonesto be more resistantto tornado-induced ington, D.C.,
 Tornado Fatalities in Ohio, 1950-1989
THOMAS W. SCHMIDLIN

Departtnentof Geography,'WaterResourcesResearchInstitute, Kent State University, Kent, Ohio 44242

1. INTRODUCTION death, and location when the storm struck, (3) identify risk
groupsin the population,and (4) makerecommendations for
Ohio is at the eastern edge of the "tornado alley" of the improvingtornadopreparedness and warningin Ohio. This
United States, yet the risk of killer tornadoes is greater in paperreportson the resultsof the research.
Ohio than some states within the core of the tornado alley
[Abbey, 1976].This is due to the higherpopulationdensityin
2. REVIEW OF LITERATURE
Ohio and the relatively high frequency of violent tornadoes
in portions of Ohio [Allen, 1981; Schaefer et al., 1986]. Previous research on mortality due to tornadoes focused
These factors, along with recent Ohio tornado disasters, on specificstorms[Mandelbaumet al., 1966;Beelman,1967;
spurredme to intensify researchinto minimizingthe risk of Abbey and Fujita, 1981; Glass et al., 1980; Centersfor
death from tornadoes. White and Haas [1975, p. 280] sug- Disease Control (CDC), 1985, 1986, 1988, 1991; Topp and
gestedthat researchon the tornado hazard shouldemphasize Sauve, 1988; Carter et al., 1989]. While useful, the studies
social ramifications of the hazard. Riebsame et al. [1986] that describemortality from only one tornado event suffer
called on the atmospheric sciencescommunity to be more from studying only a small impacted population in a rela-
effective in reducinghazard vulnerability, in part by encour- tively small geographicregion at one point in time. The
aging historical hazards analysis. This statewideexamina- attributesof personskilled in particular stormsmay reflect
tion of tornado mortality patterns over four decades ad- unique circumstancesof the community or impacts on a
dresses these issues in tornado hazards, incorporating the specialpopulation.Conclusionsconcerningrisk factorsare
literature and methods of atmospheric sciences, natural difficult to draw from a review of these studies because the
hazards, and epidemiology. reportedsummarystatisticson deathswere generallybrief,
The goal of community tornado preparednessprograms incomplete,and inconsistentamong studies [Sanderson,
and tornado watches and warnings issued by the National 1989]. A study of tornado fatalities on a statewide basis,
Weather Service is the reduction of personal injury and samplingmany disastersthat occurred in various circum-
death from tornadoes. An axiom of epidemiology is that stancesover several decades, has not been published. The
adverse health effects of natural disasters do not occur National Severe Storms Forecast Center (NSSFC) began
randomly within a population but occur in a somewhat keepingdata on the personalattributesof tornadofatalities
predictablepatternclusteredin time, in space,or in certain in 1985 [Ferguson et al., 1987]. The CDC has examined
groupsof persons[Binderand Sanderson,1987].Knowledge tornado mortality of specific events but maintains no na-
of the attributes of per'sonskilled by tornadoesand the tional archive of tornado fatalities (L. M. Sanderson, per-
circumstances of the deaths may be useful to evaluate sonal communication, 1988).
preparedness programsand tornado-warning methods[Fer- White and Haas [1975, p. 276] observed that tornado
gusonet al., 1987;Sanderson, 1989]. This knowledgemay fatality rates were decreasing in the United States but
identifyrisk groupsof personswho shouldbe targetedin geographic differencesin the deathrateswere notexplained
tornadopreparedness and warningand who need special by differencesin tornado occurrence.They suggestedthat
attention during tornado disasters. regionaldifferencesin the death rates could be causedby
The goalsof thisresearchwereto (1) establisha database differencesin tornado severity, urbanization, building con-
of the personskilledby tornadoesin Ohio duringthe period struction,preparedness,hospitalfacilities, warningsystems,
1950-1989, (2) summarizetheir age, sex, race, cause of and the distinctive behavioral characteristics of individuals.
Sims and Baumann [1972] concluded that cultural differ-
The Tornado: Its Structure,Dynamics,Prediction,and Hazards.
Geophysical Monograph 79
encesbetween the northern and the southern regions of the
Copyright1993by the AmericanGeophysicalUnion. United States were partly responsiblefor the higher death
530 TORNADO FATALITIES IN OHIO

rate from tornadoes in the south. Their work suffered from a on individual tornado events where local circumstances mav
small and narrowly stratified sample [Davies-Joneset al., have affected mortality patterns. Different methods of inves-
1973],but it is reasonableto expect that culturaldifferences tigation and reporting of results among the published litera-
within society will affect risk of death from tornadoes and ture preclude simply combining those studiesinto a conclu-
other hazards. Casualty rates in disasters have been shown sive regional or national summary of tornado mortality risk
to differ between the sexes, among age groups,and among factors. This research addressesthat dilemma by examining
race or ethnic groups. Females have been shown to have mortality patterns in many killer tornadoes that occurred in
higher death rates than males in some tornado disasters Ohio over a 40-year period.
[Glasset al., 1980;Abbey and Fujita, 1981;Fergusonet al.,
1987] but lower than males in other tornadoes [Beelman,
3. METHODS
1967; Topp and Sauve, 1988; Carter et al., 1989]. This
variety of results relating sex and tornado mortality rein- A list of the dates, locations, and strengthsof all tornadoes
forces the earlier contention that results from individual in Ohio during the period 1950-1989 was provided by NS-
tornado events cannot be used to represent general patterns SFC. This list included the number of fatalities caused by
of risk and mortality. each tornado. A total of 172 fatalities were indicated in this
The youngandold sufferhigherratesof disastercasualties NSSFC list. Newspapers from the days after the storm were
than other agegroups[Friedsam, 1962].While few studiesof examined initially to obtain the name, age, and sex of
tornado mortality have identified children as a high-risk victims. This intk)rmation was found to be unreliable. There-
group, young children are especially vulnerable to death fore county coroner records and death certificates on file at
from head trauma, a major cause of seriousinjury in torna- county health departments and the Ohio Department of
does. The elderly have been shown to be at greater risk of Health were examined. The personal attributes of age, sex,
injury or death in weather disasters [Moore, 1958; Abbey race, location when the storm struck, and cause of death
and Fl•jita, 1981;CDC, 1985;Sanderson, 1989;Carter et al., were obtainedfrom the death certificates, when possible, for
1989]. This may be due to their high incidence of mobility- each tornado fatality. Deaths that were not directly related
limiting chronic disease, lower likelihood of receiving warn- to a tornado or that resulted from injuries after the storm,
ing due to isolated living arrangementsor disability, and such as heart attacks, traffic accidents, cleanup injuries, or
reluctance to evacuate [Friedsam, 1962]. Risk-taking behav- fires, were not considered as tornado fatalities in this study.
ior and responsesduring a disasterare a function of person- This research procedure revealed several errors in the
ality [White and Haas, 1975, p. 101], so it is reasonableto NSSFC count of fatalities and tornadoes. When the adjust-
expect that in addition to age and sex, culture and ethnicity ments were tallied, the number of persons.killed by torna-
may be factors in tornado mortality. Several studies have does in Ohio during the period 1950-1989 was determined to
shown that ethnicity affects accessand responseto warnings be 155, not the 172 reported in the NSSFC listing. All
and mortality in weather disasters[Moore, 1958;Pero' et al., subsequentfigures quoted in this paper utilize the adjusted
1982; Aguirre, 1988]. fatality counts. These 155 deaths were caused by 33 torna-
The location of an individual at the time the tornado does on 16 days.
strikes can affect the extent of injuries and risk of death. The
percentage of tornado fatalities occurring in a house has
4. RESULTS
been variously reported to range from 14% [CDC, 1991] to
56% [CDC, 1986] and in motor vehicles from 10% [CDC,
4.1. Characteristics of Killer Tornadoes
1986]to 60% [Glass et al., 1980].Ed,vards [1985]reported
that only 5% of the U.S. populationlived in mobile homes, Figure 1 shows the paths of the 33 killer tornadoes. The
yet the proportion of tornado fatalities that occurred in unglaciated Appalachian plateau of southeastern Ohio is
mobile homes nationwide was 46% in 1984. The cause of hilly and wooded with a low population density compared to
death from tornadoes is most often head and chest trauma the rest of the state. The reported tornado frequency is
[Mandelbaum et al., 1966; Beetman, 1967; Glass et al., 1980; lowest in this portion of Ohio [Schmidlin, 1988], and rela-
Carter et al., 1989]. These injuries result from impact with tively few tornadoes caused fatalities in southeastern Ohio.
objects that were made airborne by the wind, commonly The medianlengthof the tornado pathsfrom the first damage
called missiles or projectiles, from the person becoming point to the last damage point was 40 km, much longer than
airborne and striking the ground or a fixed object, or from the typical Ohio tornado path of about 2 km [Schmidlin,
crushing injuries in collapsed buildings or vehicles. Addi- 1988]. Most (80%) tornado fatalities were caused by violent
tional knowledgeof the causesof death in tornadoesmay tornadoes (F4-F5), yet these storms comprised only 3% of
lead to improved safety and shelter procedures. reported Ohio tornadoes [Schmidlin, 1988]. The greatest
It is clear from the literature that personal attributes, such number of people killed by a single tornado was 32 by
as age, sex, race or ethnicity, accessto warnings, location NSSFC tornado 3 on April 3, 1974. The deadliest day was
when the tornado strikes, and others, may affect tornado April 11, 1965, when 55 were killed by seventornadoes.This
mortality rates. Numerous inconsistenciesappear in the represented35% of total fatalities in the 40-year period. The
literature becausethe previously published studiesfocused median number of deaths per killer tornado was two.
 SCHMIDLIN 531

3O

2O

10

o -2,
18 24 06

HOUR

Fig. 2. Number of fatal tornadoes (heavy stippling)and tornado

fatalities(light stippling)by hour (EST) of tornado touchdown.

Resultsare given in Figure 3. The very young and the elderly

die at a higher rate than teenagers and middle-aged adults.
The death rate for children under 5 years old was twice the
Fig. 1. Paths of Ohio tornadoesthat causedfatalities during 1950-
1989. Paths are shown from first touchdown to last lift-off as rate for children 5 to 9 years old and 3 times the death rate
provided by NSSFC, althoughthe paths were often not continuous for children 10 to 19 years old. The death rate for children
on the ground. Some long-path tornadoes may actually represent under 1 year old (not shown) was about 10, higher than any
more than one tornado.
other age group except those over 80 years old. Forty
percent of the fatalities under 5 years old died along with a
sibling, and at least 55% were with a parent or care-giver
median number of deaths per day when tornado fatalities who also died in the storm. There is no evidence that any of
occurred was 3.5.
those under 5 years old were alone at the time the storm
Tornado fatalities occurred in 14 (35%) of the 40 years struck. Age-specificmortality rates fell to 1.7 for ages 10 to
from 1950 to 1989. Nearly 93% of deaths and 78% of fatal 19 years, the lowest of any age group, then rose and leveled
tornadoes occurred during April, May, and June. April had out at 3.0 to 3.5 for ages 20 to 59 years. Mortality rates
45% of the fatal tornadoes and 65% of the deaths. There
doubledto over 6.0 for the ages 60 to 79 years. Personsaged
were no tornado fatalities during January, February, July, 65 years and older comprised 9.8% of Ohio's population but
October, or December. During the late autumn and winter, accounted for 23.9% of tornado fatalities. These high tor-
only 4% of the strong or violent (F2-F5) tornadoescaused nado mortality rates among the elderly in Ohio are in
fatalities, compared to 20% during other months. This is in agreement with previous studies of disaster mortality. A
contrast to previous studies for Indiana [Keyset et al., 1977] portion of these excess deaths may be due to the generally
and the entire United States [Galway and Pearson, 1981] poorer health of the aged which can affect their reaction to
that showed a high proportion of cold season tornadoes warnings and ability to survive injuries.
causedfatalities. The pattern of hourly occurrencesof fatal The age-specificmortality rate for persons aged 25 to 34
tornadoes(Figure 2) is similar to the general diurnal cycle of years was 3.7. This was 50% higher than the mortality rate in
tornado occurrence in Ohio [Schmidlin, 1988]. Nearly all the 20- to 24-year age group and about 10% higher than
(98.7%) of the fatalities resulted from tornadoes that touched
down between 1400and 2400 EST. Sixty-three percentof the
fatalities occurred with tornadoesthat struck duringthe dark
hours between sunset and sunrise. Risk of death from
,< • Female
tornadoes may be related to the diurnal cycle of society, "10
includingsleep, work, school, transportation,and attention •- [• Male
to radio or television [CDC, 1991].

4.2. Age of Victims

The age-specific mortality rate [Friedman, 1987, p. 11]
was calculated to determine whether some age groups were Z .

at higher risk of tornado death than others. This was

calculated as

(average
annual
fatalities
pergroup/group
population)
x 107 Fig. 3. Ohio tornado mortality rai½ By age group and
532 TORNADO FATALITIES IN OHIO

mortality rates in the 35- to 44- and 45- to 54-year agegroups. Other

Since personsaged 25 to 34 years are more likely to have Apartment 2%

small children, their higher mortality rate may result from
putting themselvesat risk to gather and shelter their chil- Commercial Bldg. 9%
dren. Of the 21 fatalities in the 25- to 34-year agegroup, eight
(38%) died along with one of their children. It is not known
how many other persons in this age group died while Mobile Home 9%
successfullyprotecting a child's life. The sacrifice of par-
ents' lives in protecting their children was evident in the
Saragosatornado[CDC, 1988].This aspectshouldbe inves-
tigated in future tornado disasters. House 62%
Motor Vehicle 13%

4.3. Sex of Victims

Mortality rates were nearly the same for males and fe-
males in the agegroupsof 10 to 19 years, 20 to 29 years, and
30 to 39 years (Figure 3). At ages less than 10 years, males
had a mortality rate 57% higher than females. At ages of 40 Fig. 4. Location of Ohio tornado fatalities.
years and older, females generallyhad a progressivelyhigher
mortality rate than males. The higher mortality rates among
young male children than among female children was not (Figure 4). Most (62%) of the victims were in a house (not
expected from reading the disasterliterature. The sex-based including mobile homes) when the tornado struck. Informa-
difference in mortality rates among young children may be tion on the precise location of victims within structures was
caused by a lower ability of boys to survive trauma, less not available for most victims in this study but should be
efficient evasive behavior amongboys than girls in disasters, collected in future disasters. Twelve of the fatalities in
more aggressiveprotective measurestaken by adultstoward vehicles were in cars, five died on a commercial bus, and one
girls than toward boys, or other factors. Further research is fatality was noted in both a van and a small truck. The
needed to identify the causes of higher tornado mortality relative risk of persons in vehicles cannot be determined
rates among boys than girls. The higher mortality rates for without knowingthe percentageof the populationthat was in
females than males over age 50 was expected. The general a vehicle in the storm's path when the tornado struck. Of the
reason given for the higher mortality rates among older Ohio tornado fatalities that occurred at home, 13% occurred
females has been the brittle nature of bones in postmeno- in mobile homes. This was 9.2% of all Ohio tornado fatali-
pausal women [Glass et al., 1980] resulting in more severe ties, a relatively low rate of death in mobile homescompared
traumatic injuries. to results from other studies. However, the proportion of
housingunits in Ohio that are mobile homes is only 2.5%, so
4.4. Race of Victims the proportion of Ohio fatalities that occurred in mobile
homes affirms the well-known risk of those locations. The
Most of the Ohio tornado victims were white. Only 3.2%
relatively low percentage (9.2%) of Ohio tornado fatalities
were black, well below the state's9.1% proportionof blacks,
that occurred in mobile homes compared to percentages
weighted over the 1960, 1970, and 1980 census data. All of
reported from some southern states may result from the
the five black victims were killed on April 3, 1974, and four
relatively low number of Ohio housingunits that are mobile
were employees, students, or visitors at Central State Uni-
homes. Only 4.6% of the victims were struck at their place of
versity, a predominantly black institution struck by a tor-
employment, away from the home, perhaps because most
nado. The low proportion of blacks among Ohio tornado
fatal tornadoes struck after traditional working hours. The
fatalities may be explained by the generally segregated
greatestnumber of fatalities in one location was five. Logue
populationpatternsin the state. The 1980 Ohio population
et al. [1981] suggestedthat higher mortality of women in
was over 99.0% white in 22 counties comprising 25% of the
tornado disasters may be due to women being in more
state's area. Among the rural regions of Ohio, only 0.9% of
vulnerable locations when the storm struck, but there was no
the population was black in 1980, while 27.3% of the
statistical difference in these Ohio data between male and
central-city urban population was black. Therefore it was
female fatalities in their location when the storm struck.
unlikely that large numbers of blacks would have been
tornado victims in Ohio unless a violent tornado struck a
black urban neighborhood or predominantly black institu- 4.6. Cause of Death
tion, such as Central State University. Head injuries were the primary cause of death for 49% of
the Ohio victims (Figure 5). Crushing chest injuries and
4.5. Locations of Victims When Struck trauma and multiple injuries, each likely to include some
The general location of the victim at the time the tornado head and chest injuries, were also common causes of death.
struck was determined for 152 (98%) of the Ohio fatalities Most victims of asphyxia suffocated under the debris
 SCHMIDLIN 533

Other 6% should be made. The National Oceanic and Atmospheric

Hemorrhage 2ø/,
Administration (NOAA) Weather Radio should continue to
Shock 3%
be a priority for the National Weather Service and disaster
planners.
FracturedSpine 7%
2. Laws requiring underground community shelters in
mobile home communities should be enacted and enforced,
since tie-down anchors are ineffective in winds over 22 m/s
Asphyxia 8% [Glass et al., 1980].
Head 3. The unsafe nature of motor vehicles in a tornado
Injury 49% should be stressed more in our mobile society.
4. Tornado safety procedures for public transportation
should be formulated and practiced, especially for school
Chest Injury 12%
buses. Students should not board school buses during a
tornado warning, and students already boarded shouldevac-
uate and enter a sturdy building.
5. The vulnerable nature of small children should be
Trauma/Multiple
stressedto parents and care-givers. Family tornado drills
Injuries 13%
shouldbe practiced to familiarize children with procedures
and locations of safe shelter. The differences between fire
Fig. 5. Primary cause of death from Ohio tornadoes. and tornado safety procedures in the home should be
stressed to children.
6. More attention should be given among the elderly to
collapsed buildings, but two young brothers drowned after their education in tornado hazards, access to warnings, and
beingblown from their parents' car into a ditch. One person availability of shelter. This will become increasingly impor-
was electrocutedwhile extricating herself from a destroyed tant as the proportion of elderly persons grows in our
house. society.
7. While tornado fatalities did not occur during the study
period in an Ohio school, daycare center, nursing home,
5. CONCLUSIONS AND RECOMMENDATIONS
office building, or factory, special attention should be given
This examination of 155 tornado fatalities that occurred in
to warning and sheltering persons in those facilities. Each
Ohio during a 40~yearperiod is the first large-scalesummary facility should have a designated "storm watcher." This
of the personal attributes of tornado victims. Newspapers personwould be trained by the National Weather Service as
were found to be an unreliable source of data on tornado a storm spotter, monitor threatening weather through visual
fatalities, so death certificates were used. observation and NOAA Weather Radio, and issue a tornado
Tornado fatalities have occurred over a wide area of Ohio, warning for the building if a tornado is observed. Regular
in rural and urban settings, but were least common in the tornado drills should be held, regardless of the disruption
hilly southeast. Most fatalities occurred during April, May, caused by drills in these facilities.
and June, and nearly all occurred between the hours of 1500 We are now extending this data base of Ohio tornado
and 2400 EST. The age groups at highest risk of tornado fatalities back through the nineteenth century to examine
death were the very young (<5 years), especiallymales, and changesin tornado mortality patterns as Ohio evolved from
the elderly (>65), especially females. There is some indica- a frontier state into the industrial era. Additional research is
tion that parents of small children die at higher rates than plannedto compare the historical tornado fatality patternsin
expected, perhaps becauseof placing themselvesin danger Ohio to those in other regions of the United States where the
while attempting to protect their children. Most fatalities population, culture, and tornado history differ from those of
occurred in houses and motor vehicles. Mobile homes were Ohio.
the location for 9.2% of Ohio fatalities. This is much less Since historical research on tornado fatalities is limited to
than in most other studies, but only 2.5% of Ohio housing data on death certificates, it is important that additional
units are mobile homes, so this still representsa higherdeath relevant data on modern victims be gathered soon after each
rate in mobile homes than other forms of housing. Head fatal storm. Such data should include income, education,
injuries accounted for nearly half of the fatalities, with awareness of storm warnings, exact location when struck,
trauma, chest injuries, and asphyxiaaccountingfor signifi- responsibility for care of family members, and previous
cant portions of the tornado deaths. serious health problems.
The following recommendationsare made to reducemor- The disparity in mortality rates between young boys and
tality from tornadoesin Ohio. They arise from the resultsof young girls and between men and women over 50 years old
this research and other referenced literature. needs further research. The efficiency of various electronic
1. Continued efforts toward providing efficient warning mediain conveyingwarnings needsattention. More research
through an unmistakable and audible community siren on the effects of culture, education, and income may
534 TORNADO FATALITIES IN OHIO

focusedtargetingof populationsfor tornado preparedness Davies-Jones, R., J. Golden, and J. Schaefer, Psychological re-
and warning. A determination of the percentageof the sponseto tornadoes: Letter, Science, I80, 544, 1973.
affectedlocal populationin houses,vehicles, mobilehomes, Edwards, C. M., Statewide tornado drills--Have we gone far
enough?,in Preprints, I4th Conference on Severe Local Storms,
or officesduringa tornadowould allow mortalityratesto be pp. JS-J6, American Meteorological Society, Boston, Mass.,
determinedfor thoselocations.Finally, the continuedefforts 1985.
of meteorologistsand disasterplannersto improvetornado Ferguson, E. W., F. P. Ostby, and P. W. Leftwich, Jr., Annual
warningsis an importantmeansof reducingtornadomortal- summary: The tornado season of 1985, Mort. Weather Rev., 115,
ity. 1437-1448, 1987.
Friedman, G. D., Primer of Epidemiology, 305 pp., McGraw-Hill,
New York, 1987.
Acknowledgments.Appreciationis extendedto the National Friedsam, H. J., Older persons in disaster, in Man and Society in
Severe Storms Forecast Center, Herman Butler of the Ohio Depart- Disaster, edited by G. W. Baker and D. W. Chapman, pp.
ment of Health, the staff of the Ohio Historical Society, county 151-182, Basic Books, New York, 1962.
coroner offices,county health departments,Daniel Rooney, and Galway, J. G., and A. Pearson, Winter tornado outbreaks, Mon.
JeanneAppelhansSchmidlin.A portionof thisresearchwasfunded Weather Rev., 109, 1072-1080, 1981.
with a SummerResearchAppointmentfrom the ResearchCouncil Glass, R. I., R. B. Craven, D. J. Bregman, B. J. Stoll, N. Horowitz,
of Kent State University. Travel to Tornado SymposiumIII was P. Kerndt, and J. Winkle, Injuries from the Wichita Falls tornado:
supportedby the College of Arts and Sciencesat Kent State Implications for prevention, Science, 207, 734-738, 1980.
University. Keyset, D. A., E. M. Agee, and C. R. Church, The modern
climatology of Indiana tornadoes, Proc. Indiana Acad. Sci., 86,
REFERENCES 380-390, 1977.
Logue, J. N., M. E. Melick, and H. Hansen, Research issues and
Abbey, R. F., Jr., Risk probabilitiesassociatedwith tornadowind- directions in the epidemiology of health effects of disasters,
speeds,in Proceedingsof the Symposiutnon Tornadoes:Assess- Epidemiol. Rev., 3, 140-162, 1981.
merit of Knowledgeand Implicationsfor Man, editedby Richard Mandelbaum, I., D. Nahrwold, and D. W. Buyer, Management of
E. Peterson,pp. 177-236,TexasTech University, Lubbock,Tex., tornado casualties, J. Trauma, 6, 353-361, 1966.
1976.
Moore, H. E., Tornadoes Over Texas: A Study of Waco and San
Abbey, R. F., Jr., and T. T. Fujita, Tornadoesas representedby the Angelo in Disaster, 334 pp., University of Texas Press, Austin,
tornado outbreakof 3-4 April 1974,in Thunderstortns: A Social, 1958.
Scientific,and TechnologicalDocumentary,vol. 1, The Thunder- Perry, R. W., M. K. Lindell, and M. R. Greene, Crisis communi-
storm in Human Affairs, edited by E. Kessler, pp. 47-88, National cations: Ethnic differentials in interpreting and acting on disaster
Oceanic and Atmospheric Administration, Washington, D.C., warnings, Social Behav. Personality, 10, 97-104, 1982.
1981.
Riebsame, W. E., H. F. Diaz, T. Moses, and M. Price, The social
Aguirre, B. E., The lack of warningsbefore the SaragosaTornado, burden of weather and climate hazards, Bull. Am. Meteorol. Soc.,
Int. J. Mass Emerg. Disasters, 6, 65-74, 1988.
67, 1378-1388, 1986.
Allen, B. S., Regiona!izationof tornado hazard probability, M.S.
Sanderson, L. M., Tornadoes, in The Public Health Consequences
thesis, 170 pp., Tex. Tech Univ., Lubbock, 1981.
Beelman, F. C., Disaster planning: Report of tornado casualtiesin of Natural Disasters 1989, pp. 39-49, Centers for Disease Con-
Topeka, J. Kansas Med. Soc., 68, 153-161, 1967. trol, U.S. Department of Health and Human Services, Atlanta,
Ga., 1989.
Binder, S., and L. M. Sanderson,The role of the epidemiologistin
natural disasters,Ann. Emerg. Med., 16, 1081-1084, 1987. Schaefer, J. T., D. L. Kelly, and R. F. Abbey, A minimum
Carter, A. O., M. E. Millson, and D. E. Allen, Epidemiologicstudy assumption tornado-hazard probability model, J. Clim. Appl.
of deaths and injuries due to tornadoes,Am. J. Epidemiol., 130, Meteorol., 25, 1934-1945, 1986.
1209-1218, 1989. Schmidlin, T. W., Ohio tornado climatology, 1950-85, in Preprints,
Centers for Disease Control, Tornado disaster--North Carolina, 15th Conferenceon Severe Local Stortns, pp. 523-524, American
South Carolina, March 28, 1984, Morbidity Mortality Weekly Meteorological Society, Boston, Mass., 1988.
Rep., 34, 205-206, 211-213, 1985. Sims, J. H., and D. D. Baumann, The tornado threat: Coping styles
Centersfor DiseaseControl, Tornadodisaster--Pennsylvania,
Mor- of the north and south, Science, I76, 1386-1392, 1972.
bidity Mortalit), Weekly Rep., 35, 233-235, 1986. Topp, F. M., and A. G. Sauve, The Edmonton Tornado, 31 July
Centers for DiseaseControl, Tornado disaster--Texas, Morbidity I987: The Police Perspective, 220 pp., Edmonton Police Depart-
and Mortality Weekly Rep., 37, 454-456, 461, 1988. ment, Edmonton, Canada, 1988.
Centers for DiseaseControl, Tornado disaster--Illinois, 1990,Mor- White, G. F., and J. E. Haas, Assessmentof Research on Natural
bidity Mortality Weekly Rep., 40, 33-36, 199I. Hazards, 485 pp., MIT Press, Cambridge, Mass.,
 Calculationof Wind SpeedsRequiredto Damage or Destroy Buildings
HENRY LIU

Departtnentof Civil Engineering,Universityof Missouri-Colurnbia,Colltmbia, Missottri 652I 1

1. INTRODUCTION tornadoes. Such a clause releases the builder, the architect,

and the engineerfrom any liability for any buildingsthey
Determination of wind speeds required to damage or have designedor constructed,in the event that the building
destroy a building is important not only for the improvement fails in a tornado or a reported tornado.
of building design and construction but also for the estima- Becauseof the foregoing reasons, it is essential that the
tion of wind speedsin tornadoesand other damagingstorms. maximumwind speedsof tornadoesand other severestorms
For instance, since 1973 the U.S. National Weather Service be neither overestimated nor underestimated. This points to
has been using the well-known Fujita scale (F scale) to the importanceof realisticdeterminationof wind speedsof
estimate the maximum wind speeds of tornadoes [Fujita, severe storms.

1981]. The F scale classifiestornadoes into 13 numbers, F-0 The state of the art in determining wind speedsof torna-
through F-12. The wind speed (maximum gust speed)asso- doesfrom damagedstructuresis describedby Mehta [1976]
ciated with each F number is given in Table 1. Note that F-6 and Minor et al. [1977]. These publications indicate that
through F-12 are for wind speedsbetween 319 mi/hr (mph) ordinary engineeringestimatesof tornado wind speed are
and the sonicvelocity (approximately760 mph; 1 mph = 1.6 based on the analysis of simple structures such as traffic
km/kr). However, since no tornadoes have been classifiedto signs,or engineeredbuildings, structureswhose behaviors
exceed F-5, the F-6 through F-12 categorieshave no practi- are more predictable and hence given higher "credence
cal meaning [Fujita, 1981]. levels." These studies have found that the wind speedsfor
While underestimatingthe maximum wind speedsin tor- high F scale tornadoes(F-3 and above) given in Table 1
nadoes and other severe storms can lead to unsafe designof appearto be grosslyoverstated.Additional studyis needed
buildings,overestimatingsuchwind speedsleadsto wasteful to improve the accuracy of engineeringestimatesof wind
constructions.For instance, prior to 1970, nuclear power speed from damagedbuildings, not only the engineered
plants in the United Stateswere designedto resisttornado buildingsbut also the nonengineeredand marginally engi-
wind speedshigherthan 600 mph. Suchdesignshave been neered buildings which constitute the bulk of buildings
excessivelycostlyto the publicand havecontributedto the damaged in wind storms.
diminution of the cost effectiveness of nuclear power. An- The purposeof this paper is to explore how to predict as
other harmful effect causedby exaggeratedwind speedsof accuratelyas possiblethe minimum wind speedrequired to
tornadoesis the hesitanceof the publicto accepttornado- damageor destroya building, and to make suchpredictions
resistantdesignand construction.The generalpublic still more a sciencethan an art. Improved accuracyof suchwind
believes that tornado wind speeds are so high that any speedpredictionswill benefit not only the engineerwho
attemptsto improvebuildingperformancein tornadoesare must design and construct safe buildings and structures at
futile. This attitudehaseven permeatedinto buildingcodes. reasonablecosts, but also the meteorologist who must report
Most buildingcodesin the United Statescontainan escape to his (her) organizationand to the public the estimated
clauseto the effect that buildingsare not designedto resist maximumwind speedassociatedwith eachdamagingstorm.
Understanding the basic characteristics of tornadoes and
The Tornado:Its Structure,Dynamics,Prediction,and Hazards.
other wind storms will also be enhanced by improved
GeophysicalMonograph 79
Copyright1993by the AmericanGeophysicalUnion. accuracy in such wind speed estimates.
536 CALCULATION OF DESTRUCTIVE WIND SPEEDS

TABLE 1. F ScaleClassificationAccordingto Wind Speed

F Number

F-0 F-1 F-2 F-3 F-4 F-5 F-6 to F-12

Maximum gust 40-72 73-112 113-157 158-206 207-260 261-318 319 to sonic
speed, mph* velocity

Note: 1 mph = 1.6 km/hr. ,

2. WIND LOADING AND BUILDING RESPONSE where A is the claddingarea, P e is the averageexternal
pressureon the part of claddingfor whichF is calculated,
Accurate determination of the wind speed from building
andP i is the internalpressure.WhenF is positive,the net
damages observed, and the converse task of predicting wind load F is directed toward the building, and when F is
buildingdamagesfrom known wind speeds,requirea good
negative,the windload actsawayfrom the building.
knowledgeof wind engineering,structuralengineering,and
It is importantto realize that the external pressureP e
constructionpractice. Some knowledgeof meteorologyis
dependson buildinggeometry,not on claddedopenings.In
also helpful in accomplishingthis goal. contrast,the internal pressureP i dependsstronglyon clad-
Wind-engineering knowledgeis neededfor the determina-
dingopenings.For instance,with a window or door open on
tion of the wind loads(pressureandforces)on buildings,and the windward wall, the internal pressure rises (becomes
a knowledge of structural engineeringand construction positive).However, when claddingopeningsare on leeward
practiceis essentialto predictingthe buildingresponse:how and side walls, the internal pressure drops (becomes nega-
buildingsreact to known wind loadsand how buildingsget tive). Figure2 showsthe variationof internalpressurewith
damagedor fail in high winds. Both the loading and the claddingopenings.Very high positive internal pressurecan
response must be accurately determined before one can
be generatedby openinga door or a windowon the wind-
accuratelydeterminethe extentof damageassociated with a ward wall in high winds. Such high internal pressureoften
given wind speed. contributesto the failure of lightweight roofs.
In general,the wind pressureP on a building, for both
2.1. Wind Loading internal or external presssuresand for both pressure and
suction,is proportionalto the square of the wind velocity
Wind createspressureon the surfaceof buildings,and the
(speed)V in the following manner:
pressuregeneratesforces(loads).Differentpartsof a build-
ing encounterdifferentwind pressuresand loads.For in- P = CppV2/2 (2)
stance,when wind direction is perpendicularto a building,
the windward wall experiencesan external pressurethat is where 9 is the density of the air and the proportionality
higherthan the ambientatmosphericpressureat theground constantCp is calledthe "dimensionless
pressure"or
level, whereas the leeward and the side walls experience an "pressurecoefficient" Substituting(2) into (1) yields
external pressure lower than ambient. Thus the relative
pressure (i.e., wind pressureon the building minus the F = A(Cpe
- Cpi)
9V2/2 (3)

ambientpressure)on the windwardwall is positive,while it whereCpe and Cpi are the externalor internalpressure
is negative on leeward and sidewalls. The relative pressure coefficients,respectively.
on ordinary roofsis negativeexcepton the windwardpart of Equation(3) showsthat accuratedeterminationof wind
a steeproof. Figure 1 depictsthe pressuredistributionover
a typical rectangularbuilding with a flat roof. Note that
arrowspointingtowardthe buildingrepresentpositivepres-
sure (simply "pressure"), whereas arrows pointingaway
from the buildingrepresent "suction" (negativepressure).
The pressurediscussedin the previousparagraphis that
generatedon the external surfacesof buildings.It is called
the "external pressure." Wind also generatesa pressure Wind

inside buildings termed the "internal pressure." The wind

loads on building cladding (i.e., walls, roofs, and ceilings)
depend on the difference between the external and the
internal pressuresacting on oppositesidesof the cladding.
Generally,the net wind load, F, on any part of claddingis , //, / / ///////,

F = A(Pe- Pi) (1) Fig.1. Wind-generated

pressure
around
a block-type
 LIU 537

p.L
= +_ 0

'/? ? / / /"'2"' / / ? '?// // /' /" :

(a) HermeticBuilding (b) Windward

Opening

(c) SuctionOpening (d) Openingson More

Than One Side

Fig. 2. Variation of equilibrium internal pressurewith opening. (Note' Wind is from left to right. Arrows show
directions of pressure.Arrows outsidethe buildingindicate external pressure,whereas arrows inside indicate internal
pressure.)

loads on buildings requires an accurate knowledge of the homes), ordinary wood frame houses, ordinary buildings
externaland internalpressurecoefficients,
Cpe and Cpi, with sheet metal roofs, and ordinary buildings that use
respectively.
Valuesof Cpeare normallydetermined from unreinforced masonry walls are known to be particularly
wind tunnel tests and are given in handbooksand standards sensitive to wind. They are often damaged or destroyed by
such as those of the American Society of Civil Engineers winds at less than 100 mph (gust speed at roof height). In
(ASCE)[1988].Valuesof Cpi for a givenbuilding,on the contrast, structurally engineered buildings (i.e., buildings
other hand, can be calculated in a simple manner if the that have been designedthrough proper structural analysis),
claddingopeningconditionsare known, as discussedby Liu such as reinforced concrete buildings or steel-framed build-
[ 1991] and Liu and Darkow [ 1989]. ings(even for skyscrapers)are seldom destroyedor severely
Accurate determination of the wind loads on a building in
damagedby wind.
high wind requires the use of (3) on various parts of the The wind resistanceof buildings depends not only on the
building,usingdifferentvaluesof Ct,e for differentbuilding type of buildingsbut also on constructiondetails. A good
parts.Thevalueof Ct,i, in contrast, is uniformthroughoutknowledge of various ways buildings are constructed is
the interiorsof a buildingexcept in caseswhere the building
essentialto the determination of building responsesto high
hastightly sealedrooms,which give rise to differentinternal
winds. Because building codes govern the construction
pressures in different rooms.
practice of buildings, the kind of building code used in a
particular city or county has a profound effect on the wind
2.2. Building Response resistance of buildings. For instance, some wood frame
Different types of buildingsresponddifferentlyto high houses built in rural counties in the nation where no building
winds. Some buildingsare wind-sensitive;othersare wind- codes exist may use nails to anchor walls to foundations, a
resistant. For instance, manufactured housing (mobile totally inadequate practice. Such houses can be blown
538 CALCULATION OF DESTRUCTIVE WIND SPEEDS

3. DETERMINING WIND SPEED REQUIRED

TO DAMAGE BUILDINGS

3.1. Weak Links and Failure Modes

Wind damage investigations conducted by various engi-

neering groups have establishedthat most wind-sensitive
buildings have particular weak points (the "weak links")
that are responsiblefor the failure of such buildings in high
winds. For instance, it is known that the most common
failure of wood [¾amehouses is roof failure, or the collapse
of the entire housetriggeredby roof failure. Once the roof of
a house is lifted off by winds, the walls of the house are no
longer tied together on the top and they may be blown over
Fig. 3. Ordinaryrafter-to-top-plate
joint. (Note that two toenails by the wind, resultingin total buildingfailure. Roof failure of
are used on one side of the rafter, and one toenail on the opposite suchhouses,on the other hand, is often causedby the failure
side not shown).
of connections,especiallythe connectionsbetween the roof
and the walls (the rafter-to-top-plate joint; see Figure 3).
This particular joint normally consists of three toenails; it
cannot resist the large uplift forces generatedby high winds.
foundationsand be destroyed in an 80-mph wind. (All wind The first step toward determiningthe wind speedthat has
speeds referred to in this section and hereafter, unless damaged a given building is to conduct a detailed wind
otherwise specified, are the gust speedsat the mean roof damage investigationto determine how the building failed
height of a house.) No building codes would allow such (the failure mode). Normally, the failure mode is related to
dangerousconstruction. one of the weakest links of the building. Once the weakest
On the other hand, houses in southern Florida built link responsiblefor building failure is identified, an engineer-
accordingto the South Florida Building Code [Saffir, 1987] ing analysiscan be conductedto calculatethe corresponding
minimum wind speed that can cause the damage. The weak
and houses in the coastal regions of North Carolina built
links of wood frame houses, masonry buildings, and metal
according to the 1986 North Carolina Uniform Residential
buildingsare discussedin detail by Lilt et al. [1989], Sparks
Building Code [Sparks, 1987] (two of the best codes in
et al. [1989], and Pertly et al. [1989], respectively.
providing wind-resistanthouses)shouldbe able to resisthigh
winds of the order of 140 mph without substantialdamage.
This is due to the stringentrequirementsof thesecodeswith 3.2. Equationsfor Calcldating
respect to roof tie-down and other constructiondetails. In Wind Speed From Damage
most cities in the United States governed by "model codes" Different failure modes require different equations for
such as the Uniform Building Codes, the Standard Code determining the failure wind speed. The equations for only
(Southern Code), and the Basic Code (BOCA), a wood frame one mode of failure, roof blow-off due to roof-to-wall-joint
house will lose its roof in 80- to 120-mph winds becauseof failure, will be described here. More equations for other
the toenailed rafter-to-top-plate connection, and the roof modes of failures are given by Liu [1990].
failure often causeswalls to fail as well. Note that toenailing For wind perpendicularto the ridge of a gable roof house
is a common practice in roof connectionsanctionedby most (Figure 4), the vertical pullout force on the windward joints
building codes. Each toenailedjoint containsthree toenails per tinear length of the roof along the eave is
which are nails driven at a 45ø angle to lumber surface near
the bottom edge of the lumber (see Figure 3). As analyzed by
Conner et al. [1987] and Canfield eta!. [1991], the use of
rafter ties instead of toenails greatly increases the wind
resistance of wood frame roofs.
The foregoing discussionshowsthat the wind speedsthat
damage or destroy a building depend to a large extent on
constructiondeails which in turn depend on buildingcodes.
Trying to determine wind speed from building damages
without considerationof the particular buildingcodeused in
the damage area, or constructiondetails of individual dam-
aged houses, as normally done in F scale classificationof
tornadoes, can result in grossly inaccurate wind speed
estimates. Fig. 4. Typical constructionof the wood frame of a gable
 LIU 539

Fin = khq(Cplw- Cpi) (8)

Fit,
=Bqcos
0 1-B2
• 4cos
20 - 1+ GCv•For the leeward joints,
+ '•- GCpsl
+ 4-•2G(Cpsl
- Cps2)
+ GiCpi ] F2H= khq(Cpi- Cp2w)
In the aboveequations,Cpl,,,andCp2•,.are the external
pressurecoefficientsfor the windward and leeward walls,
(9)

-Bcos
0(u'+,•,0•-•
ø) (4) respectively;h is the wall height;and k is the fractionof the
out-of-planeforce distributed to the top of the wall or the
roof-to-walljoints. Note that k equals0.5 if we assumethe
where B is the length of the r•ter or top chord of the truss horizontal wind load on each wall to be equally sharedby the
measuredfrom the ridgeto the wall; B0 is the lengthof the supportson the top and the bottom of the walls. The
overhangingpart of the rafter or top chord; q is the stagna- horizontalforces Fl•q and F2H are in the wind direction
tionpressure 0.5pV2 wherep istheairdensity andV is the when positive and in the oppositedirection of wind when
fastest-mile wind speedat the mean roof height; 0 is the roof negative.From (8) and (9) the horizontalforcesonwindward
slope (assuminga gable roo•; G is the gust responsefactor andleewardjoints, F1H and F2H, dependto a largeextenton
for external pressure; G i is the gust responsefactor for the internal pressureof the building:larger internal pressure
internalpressure;Cp• is the averageexternalpressure causes a larger force on the leeward joints, F2H, and a
coefficienton the windwardpart of the roof; Cp2 is the smallerforce on the windward joints, Fin.
average external pressurecoefficienton the leeward part of Once the vertical forces and horizontal forces on the joints
the roof; Cpsiand Cps2are the meanpressure coefficients are determined from the foregoing equations, the resultant
for the windwardsoffitandleewardsoffit,respectively; force is
is the internal pressurecoefficient;w is the weight per unit
area of the roof in the regionbetweenthe walls; and w0 is the F = (F} + F•)1/2 (10)
weight per unit area of the roof overhangs.Likewise, the
and the direction of this force is
pullout force on the leeward joints per linear length of the
roof along the eave is
4>= tan-1 F u/F•, ( 11)
where & is the angle measured from the vertical. The
F2v=Bq
cos
0 [B2- 1 4cos
20 quantitiesF, F H, and F•, in (10) and (11) can be for either
windward or leeward joints.
Oncethe force per length, F, is foundfrom (10), the force
- 1+ GCp2
- 4• (Cps•
- Cps2)G on eachjoint, be it the three-toenailjoint or any other joint,
can be found from

B
+•GCps,+ - Gic,'-B cos 0 w+w 0 Fj = F/n (12)
(5) where n is the number of joints on each side of the roof per
linearwidthof theroofalongthe eaveor ridge.TheforceFj
For a flat roof without overhang, (4) and (5) reduce to can be compared to the strength of the joint in order to
determine whether the joint is adequate to resist the wind
Fv = Fir = F2;,= B[(GiCpi- GCpe)q- w] (6) load. Joint strength can be found from published indepen-
dent sources such as Canfield [1989].
whereCpe= Cpi = Cp2andB ishalfof therooflengthin
the wind direction. From (6), F•, is positivewhen w is less
3.3. Procedure and Example
than(GiCpi- GCpe)qandviceversa.Thismeans
theroof
pullsthejoints upwardonly whenthe wind-generated uplift Equations(4) through (12) can be usedfor calculatingthe
on the roof is greaterthan the weightof the roof. force on each roof-to-wall joint once the wind speed V is
For a flat roof with overhangs,(4) and (5) yield known. The converse problem of determining the wind
speedV from observedroof damagecan be solvedby using

2+;_0)
(7) the same equationsin the same order through a trial-and-
Fly
- F2v
=GqB•B (C•,sl
- C•,s2) error approach. The approach assumesvarious wind speeds
to determine the various correspondingforces on the joint.
Because
(Cpsl- Cps2)is positive,
(7) shows
that(Fl•, - By comparingthe calculatedjoint forceswith the anticipated
F2v)ispositive(orFi•, isgreaterthanF2v)forflatroofswith or tested pullout strength of the joint, the minimum wind
overhang,andFiG,= F2v whenno overhangexists. speedthat can causeroof failure can be ascertained.This is
For the horizontal componentof force on joints, the illustrated in the following example.
followingequationsmaybe used:For the windwardjoints, Supposea wood frame househas lost its gableroof in
540 CALCULATION OF DESTRUCTIVE WIND SPEEDS

wind storm. An examinationof the damageshowsthat the ordinary (nonengineeredand marginally engineered) build-
whole roof was lifted off the walls with the latter left ings. If most suchbuildingscan be destroyedby a 120-mph
standing.Failure was causedby toenailsbeingpulledout of wind, then F-3 and higher-number tornadoes cannot be
the top plate, a common mode of failure. From the damagejustified on the basis of wind damageto ordinary houses, no
pattern, debris scatter, and bendingof nearby vegetation,it
matter how severe the damage appears to be. Furthermore,
was determinedthat the damagingwind was from southwest Golden [1976, p. 39] conducted a detailed analysis of torna-
does and concluded: "Even allowing for errors and conser-
to northeast, which is in the direction prependicularto the
ridge/caveof the building.Thus the equationsderivedin thevative nature of the tracers used in photogrammetricstudies
previous section hold. of tornadoes,I shouldconcludethat maximumwind speeds
From damageinvestigationit was found that B0 = 2.0 feet
in tornadoesare no more than 110m/s (240 to 250 mph) in the
(1 foot = 0.3 m), B = 12 feet, L = 20.8 feet, 0 =30 ø, h (wall
lowest 10-20 m above ground." All these confirm a growing
height) = 15 feet, •, = 5 psf (1 psf = 48 Pa), and •'0 = 4 psf.
consensusin the wind-engineeringprofession and by many
FromASCE[1988], Cpl = -0.2, Cp2= -0.7, Cpl•,,= 0.9, meteorologists,includingT. T. Fujita, that the wind speeds
andCv2,. = -0.5 for thisbuilding.The pressure
coefficients associatedwith F-3 and higher-scaletornadoesappear over-
undertheroofoverhangs
areapproximatelyCpsl= 0.9 and rated.

C•,s2= -0.5. Becausea large windwardwindowfailed

during the storm, high internal pressureis anticipated, and 4. CONCLUDING REMARKS
hencewe choose
Cpi = +0.7. ThegustfactorsG andGi are
assumed to be unity because we are interested in the gust The foregoingprocedurefor determiningwind speedmust
speed rather than the fastest-mile speed used in ordinary be basedon a detailed wind damageinvestigationto provide
engineering designs. the basic data needed such as the geometry and the type of
Using the aforementioned quantities, (4), (5), (8), and (9) building; the mode of failure; the construction details of the
yield, respectively, failed parts; the conditions of windows, doors, and other
exterior openings immediately before the occurrence of the
Fl•,- 13.04q- 58.9= 0.0143V2 - 58.9 (13) wind damage;and the wind direction at the building site. In
addition, if the strengthof the parts or the joints causing the
F2•,= 13.1lq - 58.9= 0.0144V2 - 58.9 (14) failure is uncertain, tests must be performed to ascertain
such strength. The damage investigation should be con-
FiH = 1.5q= 0.00165V
2 (15) ducted professionallyand scientifically in a manner recom-
mended by Liu [1990]. For instance, the direction of the
F2H= 9.0q = 0.0099V2 (16)
wind that has damageda building should not be determined
Let us first try the value of V = 100 feet/s (70 mph). simply by debris scatter. The damage pattern, rather than
Equations (13)-(16) yield Fly = 84.1 lb, F2v = 85.1 lb, the debris scatter pattern, is often a more reliable indication
Fin = 16.5 lb, and F2n = 99 lb (1 lb = 0.45 kg). From (10) of the damaging-wind direction. Whether a window or door
the resultant forces on the windward and leeward joints are was opened or damaged before or after a building was
F• = 85.7 lb and F 2 = 130.5 lb, respectively. This shows damaged cannot be determined without careful consider-
that the critical joint for this case is the leeward joint. ation of evidence and without interviewing the surviving
Supposethe roof frame consistsof wood trussesspaced24 occupantsof the building about what happenedjust before
inches(61 cm) apart; the number of joints per linear width of the building failed.
the roof is n = 0.5. Then, from (12) the uplift on each Even after taking all the precautionsand cares required of
leewardjoint is Fj = 130.5/0.5= 261lb. Suppose
eachjoint a well-conducted engineering investigation of wind damage
consistsof three 8d toenailsas requiredby the local building to buildings and after performing all the engineering calcu-
code. From tests reported by Cranfield [1989], the rafter-to- lations illustrated here, it should be realized that the wind
toe-platejoint using3-8d toenailshas average pullout resis- speed calculated is still an approximation which can easily
tance strength of 264 lb, which is only slightly higher than be in error by more than 20%, mainly because of the
that calculated from the 100-feet/swind. Using the same uncertainties in some of the data collected and because of
procedure, it can be shown that the average failure force of the simplifyingassumptionsused. The margin of error can be
264 lb is reached when the gust speedat the roof level of this reduced by calculating the wind speedsfrom several dam-
house reaches 100.5 feeifs. aged neighboring buildings and other structures, and then
The foregoing example, plus numerous other examples taking the average of the speedsfound.
studied by the writer, indicate that it takes no more than a Furthermore, one should realize that the wind speed
120-mphwind, often much less speed,to severelydamageor calculatedfrom a failed structureor structural componentis
destroy an ordinary wood frame house, a building with a the minimum, rather than the actual, wind speed.Any wind
sheetmetal roof, or a building with nonreinforcedmasonry speed higher than that calculated could have caused the
walls, constructedaccording to existing building codes.This structure to fail. On the other hand, the wind speed calcu-
raisesseriousdoubt about the high wind speedsgivenfor F-3 lated from a structure that has survived a wind storm without
and higher-scale tornadoes on the basis of damages to major damage gives the maximum wind speed; any
 LIU 541

speedlowerthanthatwouldhaveproducedthe sameresult. M.S. degree, 48 pp., Dep. of Civ. Eng., Univ. of Missouri,
Thereforedeterminationof windspeedfromfailureanalysis Columbia, 1989.
of bothfailedandunfailedneighboringCanfield, L.
requiresan analysis R., S. H. Niu, and H. Liu, Uplift resistanceof various
rafter-wall connections,For. Prod. J., 41(7/8), 27-34, 1991.
structuresor partsof structures.
If the failurewindspeeds
Conner, H. W., D. S. Gromala, and D. W. Burgess,Roof connec-
fora failedandanunfailedstructure(orpart)exposed to the tions in houses:Key to wind resistance,J. Struct. Eng., I13(12),
samewind are V1 and V2, respectively,the actualwind 2459-2473, 1987.
speed must be between these two values. Fujita, T. T., Tornadoesand downburstsin the context of general-
Engineering calculationof wind speedbasedon structural ized planetary scales, J. Attnos. Sci., 38(8), 1511-1534, 1981.
damageand analysis,suchas describedhere, is inexactand Golden, J. H., An assessment of windspeeds in tornadoes, in
crude. This is so becauseof possibleuncertaintiesinvolved Proceedings of the Symposium on Tornadoes: Assessmentof
Knowledge and Implications for Man, pp. 5-42, Texas Tech
in the determinationof wind directionandopeningcondi- University, Lubbock, 1976.
tionsat thetimeof damage,anduncertainties withregardto Liu, H., How to conduct scientific investigationof wind damageto
materials'propertiesand construction quality.In spiteof non-engineeredbuildings(2nd draft), report preparedfor the Task
that, such calculationstill providesthe most accurateand Committee on Wind Damage Investigation, 65 pp., Am. Soc. of
Civ. Eng., New York, 1990.
reliable wind speedestimateof any damagingstorm in
Liu, H., Wind Engineering--A Handbook for Structural Engineers,
situationswhereno directmeasurements
of windspeedare 209 pp., Prentice Hall, Englewood Cliffs, N.J., 1991.
takenor availableandwherephotogrammetric analyses
of Liu, H., and G. L. Darkow, Wind effect on measuredatmospheric
wind speedare absent.Using suchestimatesin the case of pressure, J. Atmos. Oceanic Technol., 6(1), 5-12, 1989.
tornadoes can provide the meteorologistwith a tool to Liu, H., H. S. Saffir,and P. R. Sparks,Wind damageto wood-frame
recalibratethe wind speedsin the F scaleclassifications,
a houses:Problems,solutionsand researchneeds,J. Aerosp. Eng.,
2(2), 57-70, !989.
task that the meteorologicalprofessionshouldundertake
Mehta, K. C., Windspeedestimates:Engineeringanalysis,in Pro-
with the help of wind engineers. ceedingsof the Symposiumon Tornadoes:Assessmentof Knowl-
More researchis neededto improvethe methodreported edge and Implicationsfor Men, pp. 89-103, Texas Tech Univer-
and illustratedherein,and applicationof this methodto wind sity, Lubbock, 1976.
damageinvestigationsis recommendedfor the future. Mis- Minor, J. E., J. R. McDonald, and K. C. Mehta, The tornado: An
engineering-orientedperspective, Tech. Metno. ERL NSSL-82,
useof this methodcanbe avoidedor kept to a minimumby 196 pp., Natl. Severe Storms Lab., Natl. Oceanic and Atmos.
taking proper precautions discussed herein and in more Admin., Norman, Okla., 1977.
detail by Liu [1990]. Perry, D.C., J. R. McDonald, and H. S. Saffir, Strategies for
mitigating damage to metal building systems, J. Aerosp. Eng.,
2(2), 71--87, 1989.
Acknowledgment.This study resultedfrom a researchproject Saffir,H. S., State of Florida and South Florida buildingcodesto
entitled"Responseof Wood-FrameHousesto High Winds," sup- prevent hurricane damage, in Proceedings of the WERC/NSF
ported by the Natural and Man-Made Hazard MitigationProgram, Symposiumon High Winds and Building Codes, pp. 115-128,
National ScienceFoundation,undergrant 88-08425. EngineeringExtension, University of Missouri, Columbia, 1987.
Sparks,P. R., The North Carolina residentialbuildingcode and its
REFERENCES wind load requirements,in Proceedingsof the -WERC/NSF Sym-
posium on Higtz Winds and Building Codes, pp. 107-114, Engi-
American Society of Civil Engineers, Minimum designloads for neering Extension, University of Missouri, Columbia, 1987.
buildingsand other structures,Stand. 7, 94 pp., New York, 1988. Sparks,P. R., H. Liu, and H. S. Saffir, Wind damageto masonry
Canfield, L. R., Ultimate strengthof variousrafter ties, report for buildings,J. Aerosp. Eng., 2(4), 186-199,
 Risk Factors for Death or Injury in Tornadoes'
An Epidemiologic Approach
SUE ANNE BRENNER AND ERIC K. NoJI

Centers•>r Disease Control, Atlanta, Georgia 30333

INTRODUCTION RESULTS

On August28, 1990,between 3:15 and 3:45 P.M., a tornado The majority of persons whom we interviewed reported
beat a path of destructionthroughthe Will County,Illinois, having had little or no time to reach protective areas.
towns of Plainfield, Crest Hill, and Joliet [National Weather Of persons injured during the impact phase, 301 sought
Service, 1991]. The parent severe thunderstormformed on the medical attention at hospitals within 48 hours of the disaster,
Illinois-Wisconsin border and moved southeastward across 193 (64%) were treated in an emergency department and
northeastIllinois. Over its 4-hourlifetimeit producedseveral released,80 (27%) were admitted to a hospitaland survived,
otherlessdamagingtornadoes,as well aslargehail andstrong five (2%) were hospitalizedand later died, and 23 (8%) were
straightline winds. The Plainfieldtornadowas rated as an F-5 killed instantly.
tornado,witha pathlengthof 16.5mi (0.62mi = 1km)andpath Although most impact-related deaths occurred instanta-
width of 700 yd (1.09 yd = 1 m). No tornadowarningwas neously, four impact-related injuries resulted in death 2-8
issued by the National Weather Service. It was the worst weeks after the tornado, including the death of one man who
tornado in Illinois in more than 20 years and one of the most died 8 weeks later because of complications of a chest
violentin U.S. history.The tornadoseveredelectricalpowerto contusion suffered during the impact phase. At least 44
65,000 homesand businesses,cut off phoneserviceto 10,000 injuries and one death were due to postimpact events. The
residences, and caused more than $200 million worth of dam- postimpact death involved a man who put a gas-powered
age.When the tornadohit, few peoplewere in a protectedarea generator into his garage after electricity was lost. The
such as the basement or inner area of a house. As a result of the
automatic garage door was open due to loss of power;
storm'simpact, 302 peoplewere injured,including80 persons however, when power was restored, the garagedoor closed,
who were hospitalizedand survived and 28 who died. resulting in the man's asphyxiation from carbon monoxide
produced by the generator.
METHODS
Of the 28 personswho died from impact-relatedinjuries,
To preventsuchcasualtiesduringfuturetornadoes,the Cen- eightwere youngerthan 20 years of age(range: 1 monthto 69
ters for DiseaseControl conductedan investigationto assess years; mean:34 years); 14 were male. Nine people died in a
the risk factorsfor injury or death. Researchersabstracted350 large apartment complex, eight in vehicles, five in schools,
emergencyroom and inpatient medical records from eight four in houses, and two outside. We could not determine
hospitals.We obtainedfurtherinformationfor thisreportfrom whethermost of thosewho died in the apartmentcomplex
the AmericanRed Cross,coroners,and newspapers. were inside or outside when the tornado hit, since several of
An impact-relatedinjury or death was definedas an injury them were reported to have been blown from the second
or death caused by the direct mechanical effects of the floor. Three peopledied at one high school,where at least 10
tornado. Postimpactinjuries were definedas injuriesoccur- studentsavoideddeathor severeinjury by crouchingagainst
ring within 48 hours of the tornado that would not have the only hallway wall that did not collapse.
occurred in the absence of the tornado (such as injuries Most victimswere treated at one of eight local hospitals,
sustainedby walking throughthe debris, during cleanup,or with severalof the more severelyinjuredtransferredto tertia-
as a result of the loss of electrical power). ry-carefacilitiesin Chicago.Of thoseinjuredin the tornado,
221 patientswere treatedin one local emergencyroom, with
The Tornado:Its Structure,Dynamics,Prediction,and Hazards.
GeophysicalMonograph79 the next highesttotal being 38 at a secondhospital.This
This paperis not subjectto U.S. copyright.Publishedin 1993by the unequal distributionof patients occurred because many
American GeophysicalUnion. victimsreachedhospitalsby their own devices,effectively
 544 EPIDEMIOLOGIC APPROACH TO RISK FACTORS

bypassinglocal ambulance services. Fortunately, since the of the difficulty in locating apartment occupants who have
disaster occurred during the first hospital's change of shift, lost their homes, previous disaster researchers have not
twice as many health care personnel were available to care comparedthe risk of being in an apartmentwith that of being
for the onslaughtof patients. In addition, numerousphysi- in a single-familyhome. Glass et al. [ 1980] found that people
cians and nurses volunteered their services. inside brick homes suffered serious injuries less frequently
than those inside wood frame homes. On the other hand,
DISCUSSION Duclos and Ing [1989] found that people were more likely to
Previousstudiesof tornadoeshave shownthat personsolder be injured if they were occupants of a house with walls
than 60 years of age are more likely to be injured than are constructed of wood (see also Glass et al. [1980]). In these
peopleyoungerthen 20, presumablyas a resultof preexistent two studies, information on housing materials was obtained
medical illnesses, decreased mobility, decreasedability to from the building occupants themselves; in neither study
comprehendand respondrapidly to tornado warnings,and were other sources of information about housing structure
greatersusceptibility to injury from comparableamountsof used (for example, from architects, wind engineers, local
mechanicalenergy[Glasset al., 1980].In the Plainfielddisaster land use offices, tax assessors, etc.).
the relatively high proportion of deaths and injuriesamong The findingsin this report are consistentwith the results of
personsyoungerthan 20 years of age (37.9%) comparedwith previous investigations. These include the following obser-
those older than 65 years (9.7%) may reflect the population vations:(1) automobilesare particularly lethal; (2) most deaths
affected (primarily a suburban, family-oriented community, occur during the impact phase of a tornado disaster; (3)
with a median age of 27.5 years), the time of day (3:15 P.M., tornadoescan result in large economic losses; and (4) timely
when homemakersand young children are at home), and the warningsand appropriateprotective actions may significantly
lack of warning,makingit almostequallylikely that thosewith reducethe number of deaths and seriousinjuries.
and without disabilitieswould be in a dangerousarea. We make the following recommendations to minimize
The injuries responsible for most of the deaths in this tornado-related morbidity: (1) institution of adequate early
tornado includedhead injuries, multiple fractures, and arte- warning systems in regions at risk of tornado disasters
rial lacerations. In one study [Glass et al., 1980] the most (improved methods of detecting tornadoes, such as "next
common primary diagnosesnoted in the hospital records of generation radar," in all highly tornado-prone areas), (2)
seriously injured patients included trauma to the head, construction of tornado-resistant shelters or structurally
extremities, and thorax and severe lacerations and abra- sound areas within buildings, (3) development of disaster
sions. As in our study, they were unable to determine preparednessand response plans, (4) identification of occu-
whether injuries were due to high-velocity projectilesor to pant behaviors that maximize the possibility of survival
the collapseof structures.In the studynoted, Glass[1980, p. when a tornado strikes, and (5) institution of public educa-
736] states, "most patients who received major abrasions tion programs that teach appropriate protective actions
- and lacerations had not covered themselves with blankets, duringtornadoes.We are currently studyingthe contribution
pillows, or mattresses"; in our study the victims did not of specifichousingcharacteristicsand occupantbehaviorsto
have sufficienttime to take suchpreventivemeasures. the risk of death or injury during a tornado disaster.
The results of epidemiologicstudies of tornadoes have
shownthat peoplewho attemptto drive their carsaway from
Acknowledgments.The authorsexpressgratitudeto the following
theirhomesto escapea tornadohavean increasedrisk of death organizationsfor their assistancein carrying out this study: the
and severeinjury, as do occupantsof mobilehomes[Glass et AmericanRed Cross,the IllinoisDepartmentof PublicHealth, the Will
al., 1980].Thereare anecdotalaccountsof peoplesurvivinga County Health Department, and the Illinois EmergencyServices
tornado's impact by leaving their cars for ditches or over- DisasterAgency.The staffsof the followinghospitalsare alsothanked:
St. JosephMedical Center, Silver Cross Hospital, Joliet; Loyola
passes.Unfortunately,no systematicdata analysishas been UniversityMedicalCenter,May; ChristHospitaland Medical Center,
undertakento determinewhether one should stay inside a Oak Lawn; Copley Memorial Hospital,Aurora; Edward Hospital,
vehicleor leave the vehiclefor a ditch or flat groundff no Naperville; Palos CommunityHospital, Palos Heights; and Mort'is
buildingis nearbyfor sheltering.A currentNational Weather Hospital, Morris.
Servicerecommendation suggestsleavingone'scarfor a ditch
REFERENCES
if caughtin the pathof a tornado.No recommendations exist,
however,for occupants of vehiclesin areaslackingditchesor DefenseCivil Prepardness Agency,Wind-resistantdesignconcepts
other protection. for residences: Guidelinesfor homeownersand builders,Rep.
The degreeof buildingfailure is highly correlatedwith the TR-83, p. 17, U.S. Gov. Print. Office, Washington,D.C., 1979.
Duclos,P. J., and R. T. Ing, Injuriesand risk factorsfor injuries
risk of deathor injury [Glasset al., 1980;Duclosand Ing, from the May 29, 1982tornado,Marion, Illinois, Int. J. Epide-
1989].Buildingsmay fail in severalways: translation(lateral miol., 18, 213-219, 1989.
movementof the entire structure),racking(lateralcollapse Glass,R. I., R. B. Craven,D. j. Bregman,B. J. Stoll, N. Horowitz,
of the structure), overturning, material failure, and connec- P. Kerndt, andJ. Winkle, Injuriesfrom the Wichita Falls tornado,
Science, 207, 734-738, 1980.
tion separation[De.l•knse
Civil Preparedness
Agency,1979].
NationalWeatherService,The Plainfield/Crest
Hill tornado,report,
Apartmentsoften sustainheavy damage,probablybecause Natl. OceanicandAtmos.Admin., U.S. Dep. of Commer.,Silver
of their long-spanroofs and other structuralfactors.Because Spring,Md., May
 Design for OccupantProtection in Schools
HAROLD W. HARRIS

BGR Architects/Engineers,Lubbock, Texas 79411

KISHOR C. MEHTA AND JAMES R. MCDONALD

Texas Tech University, Lubbock, Texas 79409

INTRODUCTION possiblefor us to makerationalchoicesof designparameters

for buildingswith differing occupanciesand economic re-
The authors of this paper live in Lubbock, Texas. Lub- straints.
bock is situated on the High Plains of west Texas at the The authorsof this paper have been involved in the design
southwest end of the so-called "tornado alley" which and construction of at least seven schools and various other
stretchesfrom Texas through Oklahoma and Kansas into the buildingswith occupantprotective areas. These buildings
midwest. Schools with severe storm protection capabilities, have been constructed over a 20-year period.
as illustrated in the examples, were designed by BGR
Architects/Engineers(A/E).
RATIONALE FOR SCHOOL BOARDS
People everywhere, and particularly residents of "tornado
alley," are very concerned about protection from tornadoes. We have had the designguidelines that would allow us to
School administrators often ask for information about the
build protectionfor the occupantsof buildingsfor more than
location of relatively safe areas in their existing buildingsor 10 years. However, there have been relatively few public
about the possibility of providing protective areas in new buildings constructed with substantial protection for occu-
construction.In the past they have often been told that there pants from severe storms and tornadoes. This protection can
was no economical way to construct this protection. only be affordedwhen owners become willing to build it into
We now know that economical protection for occupants their buildings. In most cases this would not be a problem if
from severe storms can be incorporated into buildingsduring the potential owners were aware that the building could and
the designprocess.There have been many buildings,suchas would meet the following criteria.
nuclear power plants, atomic weapons assemblyfacilities, Cost of construction must be acceptable. Costs can be
and disaster control centers, which have been constructed kept relatively low if construction for occupant protection is
with nearly 100% tornado protection. This protection has properly planned. The protective areas would be built of the
been achieved at a very high cost. Although this cost is same materials as the remainder of the building. A prolifer-
acceptablefor these facilities becauseof their high potential ation of differing materials and building trades almost always
for causinga widespread disaster, it cannot be justified for increases the construction costs. Protective areas should be
other buildings. However, there is a great need for protec- built using techniques well known in the building industry.
tion for the occupants of schools, public buildings, nursing Builders become accustomed to certain details and construc-
homes, and residences. A fairly high degree of protection tion types and can build more economically using them.
can be achieved in designated shelter areas within these Adequate strength can usually be achieved without going to
buildingsfor a cost that is not prohibitive. more exotic materials and methods of construction. The
We already have enough knowledge about design of protectivearea shouldbe no larger than necessaryto provide
buildings for forces that tornadoes impose on structures. quick and easy accessfor all the occupantsof the building.
Research has established a reasonable forecast of the recur-
Costs can be reduced by using as much inherent protection
rence interval for tornadic wind speeds. This makes it in the building as is available. Speed of construction is

The Tornado: Its Structure, Dynamics, Prediction, and Hazards. always a large factor in costs; therefore the protective area
GeophysicalMonograph79 should be designedso that its constructioncan proceed at
Copyright 1993by the American GeophysicalUnion. the same rate as the rest of the building.
546 PROTECTION DESIGN IN SCHOOLS

Fig. 1. High School building in Xenia, Ohio, destroyed by tornado on April 4. 1974.

It is also important that the design of the protective area School administrators and the public may lear that a
doe• not interfere v•ith the normal function of the building. It protective are,t will look like a concrete bunker. In fact,
might be desirable from a protection standpoint to provide protective areas that have been built are constructed in such
heav• doors and massive screens in corridors, but if these a way that they can barely be distinguished from the rest of
items present a major inconvenience or hazard to daily use. the building.
it is likely that they will provide a major objection to the For economy, protective areas should be formed by
construction of the protective area. strengthening spacesthat have normal functions to provide.

. o

Fig. 2. Undamagedinterior hallway of the high schoolin Xenia,

 HARRIS ET AL. 547

ß 1-

t 1

all 00,00
• •-- O,,..,
•000000•0 E-
purpose
. . .
¸ø•øo • o [] o [
'

I
'

stage
o ¸ o
%o

music , , !

! i

000
0 4 8 11• 32

I
I

-i
I ll__l• I I l= ß

section

FLOOR PLAN

PRO
TECTIVE
AREA
::iiiii?i?iiiii!iiii•iiiiiiiiii?iiiiiiii!ii{ili!11iii?ii?iiii
'
i 1

SECTION

Fig. 3. Hereford Elementary Schoolwhere mediacenter located in the center of the building was strengthenedfor
occupant protection.

If strengtheningof'protectiveareas is kept as an alternatebid on-site windstorm damage documentation at more than 60

in construction documents, the owner is assured of the locations around the country since 1970. More than half of
actual cost of the protection. these windstorms were tornadoes. These documentation
efforts have been published as papers in technical journals or
BACKGROUND reports [Mehta etal., 1971, 1981; Minor and Mehta, 1979;
Personnel of the Institute for Disaster Research at Texas F,jita and McDo/mhl, 1978; McDonald and Marshall, 1982].
Tech University, including the authors, have conducted These documentation experiences, in aggregate, have
548 PROTECTION DESIGN IN SCHOOLS

Fig. 4. Architectural rendering of junior/senior high school in Sudan. Texas.

cated that custom-engineered frames of buildings do not = 2.54 cm) long piece of lumber moving parallel to its long
collapse, even in the most intense tornadoes. In addition, axis at I00 mph. These design force criteria are based on
interiors of the buildings have provided good protection to categorization of tornadoes and data obtained from field
occupants even when no consideration of tornado-resistant research in the aftermath of a large number of tornadoes. It
design was given to construction of the building. As an is believed that they exceed the forces encountered in 98% of
example. a high school in Xenia. Ohio, was virtually de- the recorded tornadoes.
stroyed in the tornado of April 4, 1974 [Mehta et al., 1976], Experience gained from the construction of each building
as indicated in Figure I. The first-floor interior hallway of has been used to refine subsequent designs. The cost of
that school, as shown in Figure 2. sustained no damage and construction of these protective areas has decreased as
could provide good occupant protection. These observations refinements are introduced.
suggestthat occupant protective areas should be located in It is important that tornado protective areas be easily
the interior as far as possible. accessible. The accuracy of weather warning continues to
Fujita and Pearson [1973] developedthe FPP scale rating improve: however. the time available to move to a protective
to categorize intensity and size of tornadoes. Teeson et al.
area after a warning has been issued varies from seconds to
[1979] categorizedevery recordedtornado for the period a few minutes. There are occasions when a tornado strikes
between 1916and 1978. This categorizationof intensity of
inhabited areas before a warning can be issued. In cases of
tornadoesshowsthat most of the tornadoesare of intensity
very short warning times, it is extremely important to have
of F-4 or less. It would be reasonableto designoccupant
an easily accessible protective area. Schools often conduct
protective areas that would resist tornado intensity of F-4
tornado drills for the schoolpopulation. In one of the schools
and could po.•,siblyresist even more intense tornadoes with
some damage.
described in this paper (Littlefield Elementary), all the
students can be moved to the protective area in 1-2 min.
Protective areas are usually constructed on the same floor
DESIGN CON SIDEtL•.TIONS
as the occupancy they serve. Moving several hundred stu-
The buildings that are described in this paper were all dents up or down stairs is time consuming. Ease of access
designedusingthe "Interim guidelinesfor buildingoccupant often dictates a protective area near the center of the
protection" published in 1975 [Mehta et al., 1975]. Wind occupiedspace,fed directly by corridors. If heavily popu-
pressures, rapid internal pressure changes, and wind-borne lated spaces, such as school classrooms or media centers,
missile criteria are specified in these guidelines. Recommen- are constructedas protective areas, many occupants may
dations include a design wind speedof 260 mi/hr (mph) (1 already be shelteredwhen a warning is issued.This could be
mph = 1.6 km/hr), an internaldesignpressureof 205 psf (1 even more significantin a case where no warning is received
psf - 47.9 Pa) actingoutward unlessventingis provided,and before a tornado strikes. Obviously, from a protection
a designmissileconsistingof a 2 inch x 4 inch x 12fi (I inch standpointthe ideal case would be for the entire building
 HARRIS ET AL. 549

PROTECTIVE AREA

NEW CONSTRUCTION

RENOVATION

o lO 20 40 lgd
1

FLOOR PLAN

Fig. 5. Protective area in junior/senior high school in Sudan, Texas, is surrounded by corridors on all sides.

be strengthened, but this is seldom economically feasible. pants anywhere else in the building. It should prevent injury
Difficult or time-consuming movements may discouragethe to nearly all of those it shelters, except from those rare
use of sufficient drills, and in the case of an actual storm may storms whose intensity exceeds the design criteria. A prop-
cause a delay in getting occupants to safety. erly designedprotective area can provide substantialsafety
The most important function of the protective area is to from most severe tornadoes.
provide dependable protection from tornadoes. This does One damage mitigation strategy for the protective area is
not necessarilymean absolute protection, but it shouldbe to locateit as far from the exterior walls as possible.Walls of
much better protectionthan would be available to its occu- conventionalconstructionprovide someshieldingfrom
550 PROTECTION DESIGN IN SCHOOLS

1•'•-•.>'LIITr"TLEFIELD
PRIMARY
SCHO.OL
• .,..,

Fig. 6. Exterior view of the primary school in Littlefield, Texas.

aging forces. Physical distance from the exterior allows strengthof conventional roof systemsto resistthese upward
dissipation of missile energy before the missiles strike the loads makesthis a critical condition. Resistanceto uplift can
walls of the protective area. Some of the protective areas be providedby adequately bracing the bottom portion of the
shown in this paper utilize a method of connecting the roof beamsandjoists, by providing extra stiff members, and
conventionalroof deck in areas outside the protective area by providingextra weight in the roof by providing thick roof
to the roof of the protective area with a very stronghinge- decks of normal-weight concrete. The heavy roof decks
type connection. We believe that forcing the debris of the provide extra missile protection from above. It is also
collapsed structure to remain in contact with the walls of the essential to provide positive connection between the roof
protectivearea will provideadditionalbracingand protec- memberand the wall structureto resist high localized uplift
tion from missiles. If the collapsingroof of corridors is pressures.
forced to drape over shelter walls, it can be effective in Another very important function of the protective area is
preventing the entry of missilesinto the protective area to provide protection from airborne missiles. Much research
through doors. has been conductedat Texas Tech University, by one of this
The mostobviousstrengthrequirementof the protective paper's authors and by others, on the resistance of various
area is to prevent its lateral collapsedue to horizontalwind wall types to missile penetration [McDonald and Bailey,
forces.Two force-resisting systemshavebeenemployedby 1985].Onlytwo wall typeswere deemedsafeand practicalto
the authorsin the designof protectiveareas.Early designs resist airborne missile penetration. A 6-inch-thick moder-
utilized heavy steel frames to support the walls and roof. ately reinforced concrete wall provides resistance to the
This was a conventionalapproachto the problem,but the designmissile,but in the area where the examplebuildings
largestructuralelementssometimesprojectedinto corridors were built, this wall was consideredto be more expensive
and interfered•,•ith the movementof people.Later designs thanmasonryconstruction. All shieldingwallsin the protec-
employeda seriesof room-sizedboxeswith heavilyrein- tive areas designedby BGR A/E were constructed of 8-inch
forced walls and roofs. Calculationsindicatethat great thick concrete masonry with all cells filled with normal-
strength can be attained in this way. weight concrete and each cell reinforced.
An importantproblemin tornado-resistant
designis to A possiblesourceof injury to occupantsof a shelter is
provide adequate resistance to uplill of the roof structure. entry of missilesthroughits door openings(we do not put
Roofsof conventional construction are designed to mostly windows in protective areas). There are several methods of
resist downwardgravity-inducedloads. With high wind providingprotectionfrom this hazard. We could provide
velocitiesthe net upward loads become great, and the low massivedoorsfor each room of the protectivearea.
 HARRIS ET AL. 551

FLOOR PLAN
0' 10' 2 5' 50' 100'
GRAPHIC SCALE • • '•''•=•

Fig. 7. Floor planof primaryschoolin Littlefield,Texas,w herestrengthening

of protectivearea costlessthan4% of
the total building cost.

would provide protectionbut would be fairly costly. Another disruption to the normal movement of people. This method
more important objection to this method of protection may also will require the widening of corridors, resulting in a
be the difficulty in managing these doors during normal larger building and increased cost. These two methods
operation of the facility. An alternate method of protecting should provide adequate protection, and each should be
openingswould be to construct screeningwalls to block considered. The buildings designed by BGR A/E used con-
direct paths of missiles. The screen walls could be con- ventional doors without screening walls. This method was
structedwith strengthequal to the walls of the protective economical and avoided the disadvantages of restricted
area. They would offer the advantage of eliminatingthe movement and difficult operation of the doors. It resulted in
problem of operatingthe heavy doors but would provide some added exposure to injury for the occupants. We
552 PROTECTION DESIGN IN SCHOOLS

Fig. 8. Classroomin primary school of Littlefield, Texas, is strengthenedfor occupant protection.

that the technique of causing the roof of the collapsed bents support a 6-inch-thick concrete roof deck and are
structure to drape over the walls and doors of the protective enveloped with heavily reinforced masonry walls. The shel-
area will provide a degree of protection. as will the place- ter contains
4300squarefeet(1 ft2 = 0.093m2) of space.
ment of the protective area awa) from the exterior of the The junior/senior high school in Sudan, Texas •,seearchi-
building. An evaluation of the possible solutions to this and tectural rendering in Figure 4), was constructed in 1981-
other design problems must balance the degree of protection 1982. This school has a capacity of 400 students and serves
against the possibility that a shelter of any kind might not be as a focus for community activity with its large auditorium
built if economic or convenience criteria are not met.
and competitive gymnasium. Occasionally, twice the student
population assemblesin the school for a community activity.
EXAMPLES OF SCHOOLS INCORPORATING OCCUPANT Ten classroomsin the middle of the building (shown in floor
PROTECTIVE AREAS plan in Figure $) were strengthened to provide a protective
Each of the example schools was designed by BGR A/E area of 7500 square feet. Each classroom has an 8-inch-thick
•ith the senior author as structural engineer. In each case concrete-filled masonry wall on all four sides and has a
the protective area was presented to the school board as a heavy concrete roof supported on steel joists. These "box-
refuge from severe storms, rather than as a total protection es" are all linked together. The protective area is surrounded
from tornadoes. There has been no active marketing for this by corridors that extend to all parts of the building. The
type of construction, but there have been many inquiries conventional roof decks over the adjoining corridors are
from schooldistricts becauseof their knowledgethat neigh- attached to the protective area in such a way that a collapse
boring districts have constructed protective areas. The of the surrounding structure will leave the corridor roof
schools presented here are typical of the schools that have draped over the walls of the protective area.
been constructed. The primary school of Littlefield Independent School
Hereford Elementary School in Hereford, Texas, was District was completed in August 1990 at a total cost of
constructedin 1974. This schoolhas a maximumcapacityof $2,133,900and has a capacity of 550 students(see photo-
600. The media center indicated in the floor plan (see Figure graphof exterior in Figure 6). Five adjoiningclassroomsand
3) was strengthenedto provide storm protection.This pro- a resource room in the interior (see floor plan in Figure 7)
tective area is located near the center of the buildingand has were strengthenedto provide more than 5500 square feet of
corridors approaching it t¾omall four sides. Wind loads are protective space. The cost of strengtheningthe protective
resisted by heavy steel bents spanningthe media center. The area was determined by an additive alternate at the time
 HARRIS ET AL. 553

biddingto be $79,000.This onlyadded3.8%to thetotalcost Fujita, T. T., and A.D. Pearson, Results of FPP classificationof
of thebuilding.Theprotective
areais locatedawayfromthe 1971and 1972tornadoes, Preprints, Eighth Conference on Severe
exteriorof thebuildingandhaseasyaccess fromallpartsof Local Storms, pp. 142-145, American Meteorological Society,
Boston, Mass., 1973.
the school. The doors are of conventional classroom con-
McDonald, J. R., and J. R. Bailey, Impact resistance of masonry
struction without screens or massive storm doors. The walls to tornado-generated missiles, in Proceedings, at Third
adjoiningconventional construction is designed to drape North American Masomy Conference, University of Texas, Ar-
over the protectivearea in the event of a collapseof the lington, Tex., 1985.
McDonald, J. R., and T. P. Marshall, Damage surveys of tornadoes
remainderof the building.As illustratedin Figure8, the near Altus, Oklahoma on May 11, 1982, report, Inst. for Disaster
protectiveareadoesnotlookanydifferentthananordinary Res., Tex. Tech Univ., Lubbock, 1982.
classroom. Mehta, K. C., J. R. McDonald, J. E. Minor, and A. J. Sanger,
Responseof structural systems to the Lubbock storm, Storm Res.
Rep. 03, 428 pp., Tex. Tech Univ., Lubbock, 1971.
SUMMARY
Mehta, K. C., J. E. Minor, J. R. McDonald, and D. B. Ward,
It is possibleto providea high level of protectionfrom Interim guidelines for building occupant protection from torna-
does and extreme winds, Rep 83-A, 24 pp., Def. Civ. Prep.
tornadoesfor occupantsof schoolsand by extensionfor Agency, Baltimore, Md., 1975.
manyothertypesof buildings.The designshouldprovidefor Mehta, K. C., J. E. Minor, and J. R. McDonald, Wind speed
accessible,usable, and affordableshelter space.An addi- analysis of April 3-4, 1974 tornadoes, J. Struct. Div., 102(ST9),
tional cost of less than 4% seems to be attainable for this 1709-1724, 1976.
Mehta, K. C., J. R. McDonald, R. D. Marshall, J. J. Abernetny, and
type of construction.
D. Boggs, The Kalamazoo Tornado of May 13, 1980, 54 pp.,
National Academy Press, Washington, D.C., 1981.
REFERENCES Minor, J. E., and K. C. Mehta, Wind damage observations and
implications, J. Struct. Div., 105(ST11), 2279-2291, 1979.
Fujita, T. T., and J. R. McDonald, Tornadodamageat the Grand Tecson, J. J., T. T. Fujita, and R. F. Abbey, Jr., Statistics of U.S.
Gulf, Mississippinuclear power plant site: Aerial and ground tornadoes based on the DAPPLE tornado tape, in Preprints, 11th
surveys, Rep. NUREG/CR-0383, U.S. Nucl. Regul. Comm., Conference on Severe Local Storms, pp. 227-234, American
Washington, D.C., 1978. Meteorological Society, Boston, Mass.,
 Discussion

TIM MARSHALL, SESSIONCHAIR

Haag Engineering Company

PAPER J 1 contractor.He does what people are willing to pay for. It

Presenter,Jim McDonald, Texas Tech University [McDon- goes back to the education process.
ald, this volume, Damage mitigation and occupantsafety]
(Arnold Court, California State University) Can you tell me PAPER J2
how to park my car to mitigate tornado damage? Presenter, Tom Schmidlin, Kent State University [Schmid-
(McDonald) I don't subscribe to the theory that all tornado fin, this volume, Tornado fatalities in Ohio, 1950-1989]
winds come from the southwest, so putting your car on the (Mat Biddle, Oklahoma University School of Geography) I
northeast side of a structure wouldn't help. We know that wanted to ask if you considered ethnicity instead of race, or
winds in a tornado can blow from any direction.
if not, why not. I would guess that ethnicity would be
(Court) Should I park against a fence or againsta building? important,especiallygoingback to Grazulis' study.What is
an F2 barn? If your parents were German Mennonite and
(McDonald) The building is probably better than the fence, you livedin Fulton County, you've got an F4 barn, not an F2
but I'm not going to say that's going to help you a whole lot. barn. Thesetypes of things could also be applied to housing
If you were at home, I would recommend that you put your and fatalities.
car in your garage. This is because we have seen many cases
where the automobile parked at the curb is blown into the (Schmidlin) I'm sure you're right. Unfortunately, the
house and collapsesa wall or a chimney. death certificates do not give that information. All you can
go by is the last name, and of course,that gets too confusing;
(Rudy Engleman, private citizen) Would you comment on so you cannot separate ethnicity from a historical stand-
the recent constructionpractice of using foam material or point. That kind of informationwould needto be gatheredin
wafer board for siding, rather than plywood? a quick response survey, perhaps within a week of the
(McDonald) If a weaker material is being used as opposedto disaster.
plywood, which normally gives additional shear resistance (Larry Twisdale, Applied Research Associates) You
to a building,I would not think that suchconstructionwould certainly derived many statistics from your data. Did
be goodpractice. If you're not makingother provisionsfor you do any significancetests? Looking at your chart on age
shear resistance,then you're creatinga problem for your- of victims, the only thing that looked significant to me,
self.
given the small sample of your data, was the older age
(Chair)There is generalcomplacencyin buildingpracticeout group.

there. We see more wafer board going up rather than (Schmidlin) You make a good point. One thing that lent
plywood. We see a tendency toward "economic construc- credibility to the age conclusionsis that they were consistent
tion" and complacency.Look at the graphsTom Grazulis over time, not just from one event. Also, remember this is
showedthis morningaboutthe numberof 50 or morefatality not a sample;this is the population. That was the death rate
tornadoes. There has been a large reduction and people are for that 40-year period.
no longeras interestedin tornado safety.Don't blamethe
(Twisdale) Yes, but it is a sample of larger nationwide
TheTornado:
ItsStructure,
Dynamics,
Prediction,
andHazards. population.I'm not sure you can make some of the conclu-
GeophysicalMonograph79
Thispaperis notsubjectto U.S. copyright.Published
in 1993by the sionsyou have made. I think you have overworked the data
American GeophysicalUnion. and border on speculation.
556 DISCUSSION

(Schmidlin)The only thingI can sayis that the datasupport (Court) Over a square foot?
the results. Things weren't different in different areas of the
state, and they didn't vary from time to time. (Liu) No. I mean the fastest speed at roof height. The
engineer uses the mean roof height wind speed to get his
pressurecoefficientand so forth.
PAPER J3
Presenter, Henry Liu, University of Missouri [Liu, this (Court) This assumesthe same wind speed over the entire
volume, Calculation of wind speedsrequired to damageor length of the roof?
destroy buildings]
(Liu) Yes, it is the same speed. In most of the storms I have
(Joe Golden, National Oceanic and AtmosphericAdminis- investigatedthere is fairly widespreaddamage.The width of
tration, Office of the Chief Scientist) I want you to deliver the tornado is wide compared to the size of the roof.
somethingyou promised in your title, Henry. What are the
highest wind speeds so far in your investigationsfrom
damage analysis of tornadoes'? PAPER J4

(Liu) We just startedthe programlast year and investigated

Presenter, Sue Brenner, Centers for Disease Control [Bren-
three stormsin Missouri, only one labeled a tornado. I found
ner and Noji, this volume, Risk factorsfor deathor injury in
the highestwind speedto be approximately100mph. In the
tornadoes: An epidemiologic approach]
case of the Columbia, Missouri, tornado in November, I was
shocked, and so was my colleague, Dr. Darkow, a senior (Tom Schmidlin,Kent State University) Did you look at any
meteorologist, that it was classified as an F3 tornado. That interactionbetween age of house and age of people in the
placesthe wind speedmuch, muchhigherthan it appearedto house?
be from our investigation.
(Brenner) No, ! did not look at that. We also did not look at
(Arnold Court, California State University) Over what time interaction of socioeconomicfactors or other things that
periodand how big a volumeare you talking aboutwhen you
could have factored into the building constructionof the
mentionwind speedfailure?Are you talkinggustsor average
house. We just used tax assessordata, which I realize is not
wind speed?
particularlyaccurate, but it was a good source of a large
(Liu) I'm talking about gust speedand at roof level. amountof data for our statistical
 Tornado Forecasting' A Review
CHARLES A. DOSWELL III

National SevereStormsLaborarGo,,Norman, Oklahoma,73069

STEVEN J. WEISS AND ROBERTH. JOHNS

NationalSevereStormsForecastCenter,KansasCity, Missouri64106

1. INTRODUCTION cause of the perception that tornado forecasts would cause

public panic. It is clear that modern tornado forecasters owe
Present-day operational tornado forecasting can be a great deal to the pioneering efforts of Fawbush and Miller
thoughtof in two parts:anticipationof tornadicpotentialin [1952, 1954], two Air Force officers who had some early
the storm environment and recognition of tornadic storms
tornado forecasting successesat Tinker Air Force Base in
oncethey develop.The formeris a forecasting issue,while
Oklahoma in the late 1940s. On the civilian side, work was
the latter is associatedwith warnings(or so-callednowcast-
proceeding[e.g., Showalter and Fulks, 1943; Lloyd, 1942],
ing). This paper focuseson the forecastingaspectof torna-
but until 1952 the civilian weather service (then called the
doesby dealingprimarilywith the relationshipbetweenthe
Weather Bureau) still was reluctant to use the word "torna-
tornadicstorm and its environment(Recognitionand detec-
tion issuesare treated by Burgesset al. [this volume]).We do" in any forecast. The successes of Fawbush and Miller
beginwith a shorthistoryof tornadoforecastingandrelated clearly paved the way for a civilian tornado forecasting
program.
researchin section2; in section3 we providean overviewof
current tornado forecasting procedures within the Severe The first civilian tornado forecastsbeganwith the forma-
Local Storms (SELS) Unit at the National Severe Storms tion of a specialized unit as part of the Weather Bureau
Forecast Center (NSSFC). In section4 we give a short Analysis (WBAN) Center in Washington, D.C., during
summary of 35 years of SELS tornado and severe thunder- March of 1952 (see Galway [1973, 1989] for more details).
storm forecast verification. In section 5 we describe our This unit becamethe Severe Local Storms(SELS) Center in
current understandingof the connectionbetween tornadoes early 1953and movedto Kansas City, Missouri, in August
and their environment. We conclude in section 6 with our 1954, eventuallyforming part of the National Severe Storms
thoughtsabout the future of tornadoforecasting. Forecast Center (NSSFC) in 1966.
When SELS first came into being, the relationshipbe-
2. SHORT HISTORY OF TORNADO FORECASTING tween the synoptic scale environment and the tornado was
Our historicalreview necessarilymustbe brief; interested not well understood;forecastingwas essentiallyempirical.
Various forecasters and researchers observed that certain
readers can consult Schaefer [1986] for additional details
about the history of severeweatherforecastingin general; meteorologicalelements, detectable within the large-scale
another review by House [1963] is somewhat dated but data networks(surfaceand aloft), tendedto be presentin
provides excellent backgroundmaterial. Although tornado many tornado events [e.g., Fawbush et al., 1951;Beebe and
forecastinghasits rootsin the nineteenthcentury,stemming Bates, 1955].These "features" includedstatic instability,
mostlyfrom the work of J.P. Finley (see Galway [1985]for significantextratropical cyclones, abundantlow-level mois-
more on Finley), it was not until the early 1950sthat serious ture, jet streams, surface convergenceboundaries, and so
tornado forecastingbegan.Before then, the use of the word forth.
"tornado" in public forecastswas prohibited, largely be- Early on, however, it became clear that no singleset of
such featureswas present with each and every tornado
The Tornado:Its Structure,Dynamics,Prediction,and Hazards. event; rather, particular collections of elements were asso-
GeophysicalMonograph79
Thispaperis not subjectto U.S. copyright.Published
in 1993by the ciated with particulargroupsof cases.In effect, pattern
American GeophysicalUnion. recognitionbecamethe basisfor forecasting.
This approach
558 TORNADO FORECASTING

Plate I. An example of a composite chart; in this case il is a composite prognosis depicting l2·hour forecast positions
of surface and upper air features al 0000 UTe Apri127, 1991, based on initial data from 1200 UTe April 26. 1991. Solid
symbols are conventional surface frontal features, green Jines denote 50 and 70% mean (surface to 500 mbar) relative
humidity, red streamlines are at 850 mbar with a 50-knot (25 m S-I) maximum indicated, blue siream line is the 500 mbar
jet stream axis. with a 76·knot (38 m 5- 1) maximum indicated, blue dashed frontal symbols depict 500 mbar thermal
trough axes, and orange lines as isopleths of the forecast lifted index.

- :~
 DOSWELL ET AL. 559

Plate 2. Example oran "enhanced V" signature from a satellite image [from Set\lok and Doswell, 19911. showing the
visible appearance (left) and the enhanced thermal infrared appearance (right), the latter of which emphasizes the
signature.

Plate 3. An example of a lime-height cross section of wind from a steerable Doppler radar using the velocity-azimuth
display algorithm.
560 TORNADO FORECASTING

SEVERE THUNDERSTORM AND TORNADO PA.R•.METER WORKSHEET

nontornadic situations?For forecasting purposes, the con-
,

nection between the storm and its environment has been and
LA-MS-AL 86 I FEB 7l 21 18 z ,. 22 06
largelyremainsby meansof a synopticclimatology,gener-
SW•T 500-550: S+ 5 66 S CNTRL& NRNMS
ally derived quite subjectively.
When the techniquesof the 1950swere being developed,
•o•.Ls 58 S+58-60
•S therewaslittle comprehension of the structureandevolution
of tornadicstormsand what relationshipexistedbetweenthe
PV,
........... 30' M+40
• { •S
-•ooM s -•. ..... tornadic storm and the tornado. Weather radars were a
•oo•e
•E•• 95K S-- 90K S brand-newtechnology,and no scientificbasisexistedto use
•so
•s.o•sTu•
8S0
TEaP
RIDGE
.......
wOF•OIS• S WOF
RID,qE
MOIST
I I• M
RIDGE S
I3 ' S ,, _

a radar for understandingtornadicstorms,muchlessdetect-

'45- 55K S 50K S
....

ingthemwhenpresent.Real-timeradarin a nationaltornado
LO-LEvE
L JET
................ •- •-I--• ---; •....s forecastingunit was a distantdream, as were real-time
700
M6NO.CHANGE
TE•P -- CROSS
Z20ø W+ CR•S• ACTUALLY
STRONG
satellite views of storms and computer-basedanalysis and
forecasting.
....
The radar observationsof the 1950screated a great deal of
laTERSECTING
UPAND
LO
JETS YES • • YES interestin learningmoreabouttornadicstormstructureand
SFC
OEw
Pam• 62 • M+ 66 • ' S
. .
evolution. Research surface mesonetworks had been estab-
lishedin the early 1950sto pursuesome ideas about torna-
F*LLIHG
PRESSURE YES -- YES i -- _____
INCEEA51NGSFCTEMP YES • -- YES
does[e.g., Tepper,1950,1959],but radarrapidlybecamethe
INCREASING
DEWPOINT
YES -- YES
•
--
, .
primarymeansfor observingtornadicstorms.Thus, while
forecasters concentrated on empirical methods based on
,

THICKNESS
RIDGE YE S • Y ES
• .

THICKNESSNO.
CHANGE YES • Y ES , _•_ synoptic-scale
surfaceand upper air data, the focus of
MESO
OR
SYNOP
PATTERN
FAVORABLE
• FAVORABLE , ,
tornado-related research dealt with storm-scale processes as
MARKED DIFLUENCE OVER THREAT AREA AT 120OZ AND PROGGED FOR OOOOZ. revealedby radar. This schismbetweenresearchand oper-
NUMEROUS TORNADOES OCCURRED FROM N.E. LA INTO MS AFTN AND EVNG.
ationalgoalsgrew with time; by the early 1960san institu-
tionalized fission of the tornado research and forecasting
........ , .....

communities was created with the establishment of the

Fig. 1. Exampleof a checklistfor severeweather. National Severe Storms Laboratory, created from the
Weather Bureau's National Severe Storms Project [see
National Severe Storms Project Staff, 1963] in 1964.
was appliedto synopticmaps and vertical soundingsto The researchproduceda picture of the tornadic storm as
developwhat Schaefer[1986]calls a "forecastrote." The a "supercell"(asdetailedby BrowningandFujita [1965]),a
essentialreferenceon this forecastingapproachis by Miller type of convectivestorm that differedsignificantlyfrom
[1972],who describesmaptypescommonlyassociated with other, nontornadicstormsin its radar structureand evolu-
majorsevereweatheroutbreaks.The notionof indicatingthe tion. Althoughit became apparentthat not all supercelIs
location and orientation of the various features of interest on producedtornadoesand that not all tornadoescamefrom
a singlemap, the so-calledcompositechart (Plate 1), is the supercelIs[seeDoswell and Burgess,this volume], super-
cornerstone of the forecast rote. It is noteworthy that the cells clearly were prolific tornadoproducersin comparison
compositechart,as employedin severeweatherforecasting, to other convective storms.
specifically attempts to establishthe interaction between In thesepioneeringstudiesit also was found that super-
features aloft and at the surface. Thus it is a product with a cells favored certain environments, although the reason for
long history of addressingwhat Mass [1991] considersa this association remained somewhat unclear. In spite of
common deficiencyin synopticanalysis, namely, the failure these gains, the knowledge of tornadic storms developed
to depict three-dimensional
relationshipsamongfeaturesat during this researchwas not readily incorporatedin the
operationalenvironment(someof the researchresultshave
different levels. Another form of this method is the checklist
(also describedby Miller [1972]), as shown in Figure 1. had an impactin some operationaldetectionand warning
Unfortunately, there was little information in this ap- programs,althougheven there, progresshas been slow);
proachthat allowedforecastersto makea directconnection research and operations seemed unable to communicate
between what they saw in their analyzed weather data and effectively. Becausethey concentrated on different scales
the stormsresponsiblefor producingthe tornadoes.On the and data streams, most forecasters and researchers no
scale of a weather map, a tornado is a microscopic dot. longer spoke the same language.
There is no information on a weather map distinguishinga By the mid-1970s, numerical cloud modeling had become
capable of fully three-dimensional,time-dependent storm
tornadic from a nontornadic storm, except insofar as there is
some association between the storm and its environment. simulations. We think the flowering of these models has
What weather map typing gives one is an association;it signaledthe beginningof the end to the barrier separating
doeslittle to explaincauseand effect. What wasthe basisfor basic storm-scale research from operational forecasting,
making the operationaldistinctionbetween tornadic and although this was not widely recognized at the time.
 DOSWELL ET AL. 561

believethisbecausethecloudmodelscanbe usedto explore tornado forecasting have begun to collaborate once again.
how the characteristic
featuresof a simulatedstormdepend For example, recent research into streamwise vorticity
on the largerenvironmentin whichit develops.Subsequent [Davies-Jones, 1984] is being applied directly in assessing
cloudmodel-basedresearchindeedhasbeen quite success- tornado potential [e.g., Johns et al., !990; Davies-Jones et
ful in developing the storm-environmentconnectionfor the al., 1990; Davies and Johns, this volume] operationally.
first time [e.g., Weismanand Klemp, 1982, 1984].
Another critical sourceof insightinto convectivestorms 3. CURRENT SELS TORNADO FORECASTING
has been research Doppler radar observations. While PROCEDURES
Browning'swork [e.g., Browning, 1964] made innovative
use of reflectivity information to infer storm flow, the Present-day SELS tornado forecasting comprises three
detailed velocity field information has confirmed the basic steps: the "Second Day Severe Thunderstorm Outlook"
supercellstormstructuresdeducedfrom non-Dopplerradar (hereinafter referred to as the DY2 AC), the "First Day
studies[e.g., Brandes,1977]andhasbeenquiteimportantin Convective Outlook" (hereafter referred to as the DY ! AC),
validatingconceptsdevelopedfrom numericalcloudmodels and severe thunderstorm/tornado watches. This suite of
[e.g., Weisman and Klemp, 984]. SELS products has evolved over time. The products are
A third important researchdevelopmentof relevancehas partially described by Weiss et al. [1980], but for a full
been the deploymentof "storm chase" teams: groupsof descriptionthe interestedreader shouldconsult the National
meteorologistsattempting to observe tornadoesand tornadic Weather Service Operations Manual, chapter C-40.
stormsfirsthand.This hasproducedan unprecedented num- The Convective Outlooks (or ACs) are regularly issued
ber of detailedvisualobservations,includingmanystormsof (and updated) general forecasts of severe thunderstorm
the nontornadic variety. For the first time, scientistshave potential for relatively large areas. (For official purposes, a
been able to relate events (tornadic and nontornadic) ob- severe thunderstorm is defined as one which producesone or
serveddirectly in the field to structuresseenin large-scale more of the following: hail ->3/4 inch (2 cm) in diameter,
weather maps. It shouldbe obviousthat tornadoforecasting measured
winds->50knots(25 m s-•), "damaging"winds
is an essentialpart of a storm chase; thus storm chasershave (involving some subjective judgment of effects required to
become contributors to forecastingresearch [e.g., Weaver meet the threshold), a tornado. Heavy rain, large quantities
and Doesken, 1991; Davies and Johns, 1991; Brady and of subthresholdhail, funnel clouds, frequent lightning, etc.
Szoke, 1988]. are not considered to meet the official criteria (see discussion
In spite of the proliferation of new technologiesin the by Doswell [1985].) Watches, on the other hand, are issued
workplace and the burgeoningresearchdevelopments,the only as needed (in the judgment of the SELS lead forecaster)
decadesfollowing the 1950shave not seen much changein and are more specific in terms of timing, location, and
operational tornado forecasting techniques. Rather than expected types of severe weather. The basic premise is that
supportingqualitative changesin the way tornado forecast- as the time of the event approaches, it is possible to refine
ing is done, new observingand analysistools have beenused the forecasts of severe thunderstorm type, timing, and
to increase the precision and timeliness of the forecasting location. While this premise seems logical, it is not neces-
approaches primarily developed in the 1950s. The new sarily valid; the relevant scales decrease as the event devel-
observations most often have been used to identify new ops, first shrinking from synoptic scale to mesoscaleand
ways to detect severe storms (e.g., the satellite-observed then on to the convective storm scale. However, the data
"enhanced V" signaturenoted by McCann [1983]; see Plate available to the forecaster do not undergo an increase
2) as well as to enhance recognition of previously known commensurate with this scale decrease. It is not uncommon
elements (e.g., intersectingthunderstorm-generatedoutflow for forecasting to become more difficult as the time of the
boundaries, as described by Purdom [1982]). These new event approaches(as discussed by Doswell et al. [1986]).
observations have improved severe storm detection and Generally, it is during the watch phase that SELS attempts
recognition, but they have not been very useful in forecast- to distinguish between tornadic and nontornadic storms.
ing. Their value in forecasting is compromisedby the fact
that most studies of their use have focused on cases where
3.1. Forecasting Procedures: Convective Outlooks
storm events actually happened;caseswhere the features
(e.g., outflow boundaries) were present but nothing hap- OperationalSELS tornado forecastingemploysthree gen-
pened have been studied only occasionally[e.g., Stensrud eral approaches:synoptic pattern recognition, meteorologi-
and Maddos, 1988]. cal parameter assessment (checklists), and climatology.
Moreover, computershave beenusedmostlyto speedand These are the tools that developed historically as noted in
enhance subjective analysistechniquesdevelopeddecades section 1. Specialized, synoptic pattern-specific,or geo-
earlier (mostly by automated data plotting) rather than to graphicallylocalizedforecastingtechniques(for someexam-
createnew techniques.In effect, for operationalforecasting, ples, see Doswell [1980], Hales [1985], Johns [1984], Weiss
the computer often has been asked to duplicate electroni- [1985], Hirt [1985], and Weiss [1987]) also contribute to the
cally what used to be done manually. ACs.
In recent years the research and operational sides of As more is learnedabout the physicalprocesses
562 TORNADO FORECASTING

in tornadoesand/or severe thunderstorms,parameterscon-

sideredoperationallyrelevant have been changingto reflect
that new understanding.Thus, for example, vorticity advec-
tion, emphasizedby Miller [1972] and questionedby Mad-
dox and Doswell [1982], is giving way to helicity-related
parameters, as discussed by Davies-Jones et al. [1990].
Continuingefforts to refine the climatologicalinformation
about tornadoes[Kelly et al., 1979] and nontornadicsevere
thunderstorms[Kelly et al., 1985] are aimed in part at
improving operationalforecasting;recall that climatology,
modifiedby knowledgeof the synopticpattern, is a tradi-
tional basis for distinguishingtornadic from nontornadic
situations.
For the long lead times of the ACs (up to 52 hoursin the
caseof the DY2 AC), the primary input to theseproductsis
the numericalweather predictionmodel guidancefrom the Fig. 2. Plot of tornado watch 183 on April 26, 1991 (compare
National MeteorologicalCenter (NMC), of which NSSFC Figure 3).
(and henceSELS) is a part. With diminishingleadtimes,and
especiallywith regardto the watches,diagnosticevaluation
of surfaceand upper air data becomesdominantover model sphericvorticesare not associatedwith deep, moist convec-
prognosis.In conjunctionwith the analysisof surfaceand tion. These are not considered to be tornadoes. In tornado
soundingdata, the remotely senseddata (suchas satellite, forecasting, vertical wind shear structure is becoming the
radar, and lightninggroundstrikelocation)that havebecome key factor in distinguishingtornadic from nontornadic
available in ever-growingamounts are increasinglyimpor- events, so this becomes a critical limiting factor in delineat-
tant. These data, especiallysatellite imagery, are useful in ing tornado threat areas.
assessing the numericalmodelinitial conditions[e.g., Hales, In a few cases (typically fewer than 10 days per year),
1979] and in data voids (most oceanic regions and some tornado "outbreaks" are forecast in the ACs. Such forecasts
sparselypopulatedland areas,as well as whenconventional began in the mid-1970s, following the April 3-4, 1974,
data are missingor contaminatedwith convection). outbreak. A separatepublic version of the AC is issuedin
SELS continuesto employ many parametersdesignedto such situations. Although it is impossible to be completely
summarize information contained in the data; such parame- general regardingtornado outbreak conditionsthat might
ters often are called "indices" (see for example, Miller result in suchan AC, they typically are associatedwith what
[1972], Galway [1956], and Showalter [1953]for some of the we call "synoptically evident" patterns. Tornadoes may be
myriadthermodynamicindicesmeasuringstaticinstability). mentioned in the outlooks when moderate or greater thermal
Recently, a more comprehensiveparameter than the tradi- instabilityis likely to be coupledwith favorablevertical wind
tional indices for static instability is coming into use in shear structures [Davies-Jones et al., 1990; Johns et al.,
SELS: the potential buoyant energy (PBE, also called the 1990; Leftwich, 1990].
convective available potential energy, or CAPE; [see Mon-
crief and Miller, 1976] is the "positive" area on a sounding
3.2. Forecasting Procedures: Watches
associated with the buoyant part of a lifted parcel asceot
between the level of free convection (LFC) and the equilib- The foremost SELS public forecast products are the
rium level [see Doswell et al., 1982]). tornado and severe thunderstorm watches (an example of
Nearly all the parameters (past and present) used on which is shownin Figure 2). Theseforecastsusually take the
composite charts and/or checklists can be shown to be form of quadrilateralscoveringon the order of 20,000 miles
associatedwith (1) synopticand mesoscaleupward motion, (about52,000km2) andarevalidfor timeperiodsof several
(2) sufficient moisture and lapse rate for a parcel to be hours. Some statistical information about watches in 1990 is
positivelybuoyant, and (3) vertical wind shearstructure. shown in Table 1.
Therefore the scientific connection between the parameters In order to convey information about SELS forecaster
used and the physical processescan be made, even if interpretationsto the forecastingcommunity (not to the
forecastershave not always recognized that connection. public), SELS MesoscaleDiscussions(MDs) are issuedas
A key notion employedin tornadoforecastingis that of needed.The MDs provide a narrative of probableweather
"limiting factors." Once a preliminary generalthreat area developmentsand possible forthcoming watch issuance;
has been defined, it is refined by consideringwhat factors MDs began in 1986. Another information-conveying, non-
make it unlikely that some parts of the originalthreat area public product is the watch-related Status Report, begun in
actually will experiencesevere weather. Obviously, any- the 1950s,which has several aims: to keep the field offices
thingprecludingthunderstormswill precludetornadoes.(As informed about severe weather conditions in and near an
noted by Doswell and Burgess [this volume], some atmo- issuedwatch area, to clear those parts of the watch
 DOSWELL ET AL. 563

the severeweatherthreat has ended,and to provideinfor- BULLETIN

TORNADO
-
WATCH
IMMEDIATE
NIIMBER 183
BROADCAST REQUESTED

mation about additional, follow-on watches. NATIONAL

1210 PM
WEATHER
CDT FRI
SERVICE
APR 26
KANSAS
1991
CITY MO

In general, SELS forecastersmust deducethe characterof

A..THE NATIONAL SEVERE STORMS FORECAST CENTER HAS ISSUED A TORNADO
the subsynopticscale processesrelevant to severe weather WATCH FOR

watchesfrom limitedoperationaldata:hourlysurfaceobser- PARTS OF CENTRAL AND EASTERN KANSAS

vations,satelliteimages,and radar displays.Thesecurrent EFFECTIVE THIS FRIDAY AFTERNOON AND EVENING UNTIL 800 PM CDT.

sourcesprovide the highestoperationallyavailablespace THIS IS A PARTICULkRLY DANGEROUS SITUATION WITH THE POSSIBILITY OF
and time resolutionfor the task at hand. Subjectivesurface VERY DAMAGING TORNADOES. ALSO..LARGEHAIL...DANGEROUS
DAMAGING THUNDERSTORM WINDS
LIGHTNING
CAN BE EXPECTED.
AND

analyseslocate and track features believed relevant [see THE TORNADO WATCH AREA IS ALONG AND 65 STATUTE MILES EAST AND WEST
Miller, 1972; Doswell, 1982] to tornado and severe thunder- OF A LINE FROM 45 MILES EAST SOUTHEAST OF MEDICINE LODGE KANSAS TO
45 MILES NORTHEAST OF CONCORDIA KANSAS.
storm forecasting. These features then are related to the
REMEMBER...A TORNADO WATCH MEANS CONDITIONS ARE FAVORABLE FOR
satelliteimagery usingthe advancedinteractivecomputer TORNADOES AND SEVERE THUNDERSTORMS IN AND CLOSE TO THE WATCH AREA.
PERSONS IN THESE AREAS SHOULD BE ON THE LOOKOUT FOR THREATENING
system called NSSFC's VAS Data Utilization Center WEATHER CONDITIONS AND LISTEN FOR LATER STATEMENTS AND POSSIBLE
(VDUC [see Browning, 1991]). Recently, lightningground WARNINGS.

strike data have become available [Mosher and Lewis, 1990] B..OTHER WATCH INFORMATION..THIS
THUNDERSTORM WATCH NUMBER 181.
TORNADO WATCH REPLACES SEVERE
WATCH NUMBER 181 WILL NOT BE IN
and are usedin SELS for defining,locating,and monitoring EFFECT AFTER 100 PM CDT.

convection, as a supplement to satellite and radar data. C...TORNADOES AND A FEW $VR TSTMS WITH HAIL SFC AND ALF TO 3 IN.
EXTRMTURBC AND SFC WND GUSTS TO 75 KNOTS. k FEW CBS WITH MAX TOPS
In the past it was common to differentiate tornadic from TO 600. MEAN WIND VECTOR 23040.

nontornadic situations using parameters related to the D...LNTCU DVLPG FM SW OF CNK TO BTWNRSLAND SLNATTM. EXPCTRAPID
TSTM DVLPMT WITHIN NEXT HR ALG DRYLN/CDFNT WITH SUPERCELLS AND
strength of the winds aloft [e.g., Miller, 1972]. This created TORNADO DVLPMT LKLY.

a seasonalbias to the frequency of issuingtornadowatches, E...OTR TSTM$...CONT WW NR 182. WW LKLY TO BE RQRD WITHIN NEXT HR
with increased likelihood of a tornado watch in the winter OR TWO OVR PTNS WRN AND CNTRL OK. WW LKLY TO BE RQRD LATER THIS
AFTN OVR PTNS ERN KS AND WRN MO.
and spring (and to a lesser extent in the fall) when strongly
... JOHNS
baroclinic disturbancesare present in severe weather situa-
tions. During the summer, with weaker synoptic scale dis- Fig. 3. Tornado watch message 183, issued on April 26, 1991,
turbances, the tendency was toward severe thunderstorm usingthe "enhanced" wording. Parts A and B are transmittedto the
(instead of tornado) watches. Although this method matches public and so are in plain text, whereas parts C-E are not made
climatology reasonably well [see Kelly et al., 1979], there public and are written in contractions to save characters.
was no understanding of the processesrelevant to torna-
dogenesisbeing employed, largely becausethat understand-
ing was not available. It was not easy to justify the choicein capableof the highestdamage potential associatedwith any
any given situation, except by experience and climatology. tornado. These cases,not surprisingly, are exemplified best
It is only within the last several years that this situation by "classical" severe weather patterns, that is, those that
has begunto change.At present, recognitionof the supercell we have called synopticallyevident. (Note that a situation
environment is becomingthe cornerstoneof SELS tornadic we label as synoptically"evident" shouldnot be automati-
versus nontornadic decision-making. This philosophy also cally equatedwith an "easy" forecast. No real-worldfore-
reflects a lack of knowledge about nonsupercelltornadoesat cast situationis ever easy, except in retrospect!)Thus SELS
present,at least comparedto supercelltornadoes;as noted has had the option (since the early 1980s) to issue tornado
by Doswell and Burgess [this volume], the study of the watches that have "enhanced wording" to highlight the
nonsupercelltornado has only just begun. tornado threat (see Figure 3). Let us define a tornado day as
On the other hand, the high degreeof associationbetween a day with one or more tornadoes and a "big" tornado day
supercelIsand tornadoeshas made it possibleto identify as a day with two or more violent (F4-F5) tornadoes. The
situations that are likely to produce tornado outbreaks. interananual variation in tornado days is rather small, aver-
Outbreak-relatedtornadoesusually are producedby super- aging about 175 days per year (see Figure 4), whereas big
cells, and the tornadoes that form from supercelIs are tornado days fluctuate considerably from year to year (Fig-
ure 5), with an average frequency of about 5 such days per
year. Enhancedwording is used when tornado outbreaksare
expected, and outbreaks typically have two or more violent
TABLE 1. SELS Watches in 1990
tornadoes, meeting our criterion for a big tornado day. Thus
Value the enhanced wording in tornado watches is not commonly
employed.
Number of tornado watches 249
Number of severe thunderstorm watches 496
The role of vertical wind shear-related parameters in
Average watch duration 4.86 hours tornado forecasting has made the development and opera-
Average watch area 76,519km2 = 29,544 tional availability of additional sources for vertical wind
(statute
mi)2 structure quite critical. Therefore the demonstration net-
Median lead time (begin time minus 31 min
work of vertical wind profilers [Gage and Ba!sley, 1978]
issue time)
presently being implemented [National Weather
 r
564 TORNAOO FORECASTING

250

··" 200

.·
Q
0
150

~
·
....
c
0

0 100

"·
•
E
z
, 50

O.OM56566062646668~72H76788082~66M90
Year
Fig. 4. Plot of yearly totals of tornado days (i.e., days with one or
more reported tornadoes).

Fig. 6. Illustrating the manually digitized radar (MDR) grid with

those blocks considered in the verification shaded gray.
1987] is of great interest to operational forecasters. Also, the
wind profiling capability [see Rabin and Zrnic', 1980] of the
WSR-80 (NEXRAD) Doppler radars in the prestorm, clear
air environment may be extremely valuable. a tornado watch. However, for the purposes of this paper we
have not accounted for this fact in the verification; that work
4. 35 YEARS OF TORNADO FORECAST VERIFICATION is not yet complete, but it will be reported upon in the
aforementioned future publication.
Since the essential aspect of tornado forecasting is to Another important aspect of tornado forecast verification
distinguish between tornadic and nontornadic situations, this is that tornado watches are area forecasts that typically
will be the primary issue discussed here. There are numer- cover several tens of thousands of square kilometers,
ous other aspects of tornado and severe thunderstorm watch whereas tornadoes affect only a few square kilometers even
verification that we cannot dwell on here; some will be in major events. This disparity in coverage means that
presented in a future publication. Our data consist of the successful tornado watches (i.e., those with tornadoes in
final SELS log of severe weather reports and records of them) are mostly "false alarms" in the sense that the vast
tornado and severe thunderstorm watches, covering the majority of the forecast area is unaffected. The original
35-year period 1955-1989, inclusive. watch verification schemes considered them as area fore-
It should be noted that a tornado watch/event is, in a casts, so that a single tornado effectively verified the entire
sense also a severe thunderstorm watch/event. That is, a area of the watch. Recently, as described by Weiss ef al.
tornado-producing storm is, by definition, a severe thunder· [1980], watch verification has been changed such thai a
storm. Therefore a tornado in a severe thunderstorm watch single report verifies only a portion of the total watch
verifies that watch, but a nontornadic event does not verify area/time, However, this new scheme still does not incorpo·
rate information about areas outside the watch.
The verification scheme used herein (first described by
12
Doswell el til. [1990b]) is based on the so-called manually
··
"
"o"
10
digitized radar (MDR) grid shown in Figure 6. Each MDR
box is roughly 40 km (25 nautical miles) on a side. (Since the

·
""
E
....o
• grid is defined on a polar stereographic map projection, the
grid boxes vary in size across the map by as much as about
10%. The nominal size applies only at 600 N latitude, whert~
.~

F
• the map scale factor is unity,) Every watch has been broken
~ down into MDR grid boxes, using the convention that if the
o

·,•
~

E 2
centroid of an MDR box is within the watch, that MDR box
is considered to be within the watch. The valid time of the
z
00 ~ ~
M56~OO62646668W72~767880~84868890
~ watch is broken down by hours; if a watch begins on the
hour or within the first 29 min of the hour, the watch is
considered valid for the whole hour, whereas watches be-
Year
ginning 30 min or more after the hour apply to the next whole
Fig. 5. Plot of yearly totals of ,'big" tornado days (i.e., those days hour. The MDR box hour is the basic unit of the verification,
with at least two or more violent (F4-F5) tornadoes). and it naturally gives a somewhat "grainy" picture. We have
 DOSWELL ET AL. 565

TABLE 2. ContingencyTable for Severe Thunderstormand skill(see

Doswell
etal. [1990a,b]fordetails).
Asindicated
Tornado Watch Box-Hours, as Described in the Text in Table3 (andFigure10),skillscores
haveincreasedby
Observed nearlyan orderof magnitude overthe 35 years.If we
Tornado Severe Nothing
compare
thefirstdecade(1955-1964)
withthelastdecade
Thunderstorm Total (1980-1989)
ofourrecord(Figure
7),it canbeseen
thatthe
primary
"tornadoalley"skillmaximum haspersisted,
but
Forecast additional
centers
of relativeskillhavedeveloped
in North
tornado n 11 rt12 t• 13 tl 1. Carolina,
New York state,Montana,andIdaho.It is at
severe thunderstorm n 21 rt22 n23 n 2.
nothing n 31 n32 n 33 n 3. present
difficult
to knowhowto interprettheseresults;
Total n.• n.2 n.3 n.. however,
spatial
distribution
ofverification
scores
clearly
is
influenced
by the distribution
of severeweatherreports
(compare
Figure7 andFigure8).
Thereporting
of nontornadic
severe
weather
hasin-
tested the effect of increasingthe resolutionboth in space creased markedly
withtime(Figure9), whereas tornado
and time and found that for our verificationpurposes,it is
reporting
has
detectablebut doesnot affect the overall patterns.There are hasclimbed
remainedmore nearlyconstant.
Whiletheskill
moreor lesssteadilyduringthe35yearsunder
4533 MDR boxes over the United States (boxes over water
consideration
(Figure10),howmuchofthisskillisattribut-
are not counted),and each nonleapyear has 39,709,080 ableto enhanced
reporting?
To attempt
to account
forthis
MDR box hours.
"inflation,"
wedidthefollowing.
If onedoesnotdistinguish
For verificationpurposes,a severeweatherreportis for themomentbetweentornadoes andnontornadic severe
consideredto verify an entire MDR box hour if it occurs
thunderstorms,
thenthecontingency
table(Table2) reduces
anytime withinthathour.Reports otherthanthefirstinthat to the 2 x 2 table shown in Table 4.
MDR boxhourareignoredunlesstheyareof a differenttype We take the numberof severereportsin 1955as the
(thetwotypesof reports are"tornado" and"nontornadicstandard, denoted by (x + Y)55-(Asis thecasein account-
severe thunderstorm"). If one or more tornadoesoccur
withina givenboxhour,it is counted asa "tornado hit" ingforcurrency inflation,
thisdoes notimplythatthereis
irrespectiveof anyconcurrent
nontornadicsevere anything
special
thunder- reference about 1955.
It simplyrepresents
a base,or
state.Wecould justaseasilyhaveadjustedtoward
stormreports.If oneor morenontornadic severeevents
occurwithina givenbox hour, it is countedas a "severe a 1989standard,with no material
difference
in our conclu-
thunderstorm hit." sions.)
Subtractthisfromthenumber ofseverereportsinthe
Our basic tool for verificationis the 3 x 3 contingency ithyear,(x + Y)i,toobtain
thedifference
di.Thisdifference
for is an estimateof the numberof eventsfor the ith yearthat
tableshownin Table2, derivedvia the aboveprocess
everyMDRbox.Fromtheinformation contained this would
within havegone unreportedin 1955,sotheseareallputinto
basictable,a widevarietyof summary
measures,
histo- thez.box (referto Table 4) in the contingency
table.We
assume thattheratioof thex boxto they boxremains the
grams,maps,etc.canbeconstructed,
ofwhich
weobviously
haveroomfor only a smallfraction.Althougha single same,andweredotheskillscoreverificationontherevised
numbercannotexpress
allofthecontent implied
inTable2, table.
Of course,
some of the di might
actually
belongin the
weshallusetheHeidkeskillscoreasa summarymeasureof w box, butsince
w is typically muchlarger
than theother

TABLE3. Actual
3 x 3Contingency
Tables
forYears
1955
and1989
Showing
theNumber
of
Watch-BoxHoursas Describedin the Text
Observed

Tornado Severe Nothing

Thunderstorm Total
1955
Forecast
tornado 85 62 39,696 39,843
severethunderstorm 40 56 78,580 78,676
604 39,589,357 39,590,561
nothing 420 722 39,707,813 39,709,080
Total 545
1989
tornado 237 1372 61,036 62,645
severethunderstorm 102 1965 96,325 98,392
4255 39,543,395 39,548,043
nothing 393 7592 39,700,756 39,709,080
Total 732

Heidke
skillscore
is0.00317
for1955
dataand0.0348
for1989
$66 TORNADO FORECASTING

Table 3, the occurrence of tornadoes in severe thunderstorm

watcheshasincreasedwith time, but only by abouta factor
,"-... ". , / ,•o•f' '-•'," of 2. Thisrelativelysmallincreaseappearsin part to be a
,, ,:',..,._..;
....., ,. :,, consequenceof the relatively modest rate of increase in
tornadoreportingoverall.However,thenumberof correctly
forecasttornadoeventshasincreasedby abouta factorof 3,
suggestingsome increase in the skill of discrimination. The
occurrenceof tornadoeswithoutany watchof eithertype
actuallydecreasedduringthe period, althoughsuchtotal
missesstill constitutethe majority of observedtornadic
events. Since most tornadoesare weak, and becausethe
probability of detection for tornadoes in watches is lowest
Heidke
SkillScore
f• SE• •atchBoxes.
!955•1964
(smoothed)
.
for weaktornadoes[LeftwichandAnthony,199!], mostof
thesemissedeventsare weaktornadoesandprobablyare
,
.

not associatedwith supercelIs.Galway [1975]has observed

thatthe majorityof tornadodeathsoccurin watches,which

Heidke
Skill
S.... •ro;
SEI•
•tstch
.
e..... 1980'-1989
(.... thed)

Fig. 7. Mapsof smoothed

Heidkeskillscorefor(a) 1955-1964
and
(b) 1980-1989.

' ,' • '.Tornoclo

Reports
,•
¾e•r•= 1955-1964 ,
entriesin the table,ignoring
thiscreatesonlya negligible (contou¾
•,2,5,•0,20,50)
'.
error. While we temporarilyhave lumpedtornadoesand ß

nontornadicseverethunderstorm eventstogetherfor this

purpose, the Heidke skill scores for the 2 x 2 and
3 x 3 versions of the tablearenot markedlydifferent[see
Doswellet al., 1990b].The resultsof thisprocedureshow
,

that overall, when report "inflation" has been accounted

for, the skillof severeweather
watches
hasimproved
by
about 50% rather than the order of magnitudeincrease
without correctingfor this inflation.
Turning to the specific issue of tornado versus severe
thunderstorm
forecasts,the asymmetryof the forecastsis
important.That is, aslongasoneor moretornadoesoccurin
a tornadowatch,thereis no problemwith havingmany
reports of nontornadic severe thunderstormevents in that
tornadowatch(elementn22in Table2). Ontheotherhand,
havingtornadoesoccur in severethunderstormwatchesis ,

an undesirableevent(elementn21in Table2). Of course, ' - • ,. ',tornodo

Reports
¾e•r•= 1980-1989
thereis an easyway to preventthisfromeveroccurring: (contours.
•,2,5.•0,20,50)',
alwaysissueonlytornadowatches. Doingsowouldrepre- ß

sent no attempt to discriminate between tornadic and non-

.
.

tornadicseverethunderstorms;
thisclearlyis notdesirable, Fig. 8. Mapsof smoothed
tornadooccurrence
for (a) 1955-1964
nor does it reflect what has been done. As can be seen in and (b)
 DOSWELL ET AL. 567

8000NON_-TORNADIC
SEVERE,
FREQUENCY
BYYEAR TABLE 4. Table I Contingency Table With No Distinction
Between Tornadoes and Severe Thunderstoms

Observed
6000 _
Severe Nothing Total

Forecast
4000 severe x y x + y
nothing z w z + w
Total x + z y + w x + y + z + •'
eooo
Since no distinction is made, all forecasts and observations are
combined in the generic "severe weather" category.
0.0
54 56 5õ 60 62 64 66 6õ 70 72 74 76 78 80 82 84 86 88 90
YEAR they are characterized by large-scale environments readily
identifiedas beingfavorable for supercellswith today's data.
Fig. 9. Plot of yearly totals of nontornadicsevereweather. Such synopticallyevident events are not frequent during the
tornado season, however (see also Maddox and Doswell
[19821).
remainstrue to thisday, suggesting
thatwatchescapture On some tornado days (typically those with large CAPE
most of the significanttornadoes. but weak to marginal shear), while severe storms are likely,
the apparenttornado potential is not high. A notable recent
5. CURRENTUNDERSTANDING
OF TORNADICSTORMS example is August 28, 1990, on which the violent Plainfield,
Illinois, event occurred. In such environments (many of
Our understanding of tornadicstormsandhowtheyinter-
which produce no significant tornadoes) some mesoscale
act with their environmentis in an excitingstateof transi-
process creates a local environment such that isolated
tion. As Doswell and Burgess [this volume] indicate, we
storms become tornadic in ideal conditions, but those con-
havecometo realizethattornadoescanoccurin manyways
and are not limited to supercell events. Not all tornado ditions are uncharacteristic of a large area. In contrast,
major outbreak days (such as the Kansas-Oklahoma out-
reportsrepresentthe samemeteorologicalphenomenon,and
break of April 26, 1991, that produced the Andover, Kansas,
we are just beginningto understandhow nonsupercell
tor-
killer tornado, among others) have widespread supercell-
nadoesmight arise [e.g., Wakimoto and Wilson, 1989].
favorableenvironmentsand many supercelIs.
Even for supercellevents,however,mesoscale
variability
Saidanotherway, more than 95% of the tornadodaysper
in atmosphericstructurecan be crucialin estimatingthe
year are not synoptically "evident"; most tornado events
chancesfor tornadoes(see Burgessand Curran [1985]for a
involvemesoscaleprocessesthat are difficultto anticipateat
casestudyexample).Mesoscaledetails,often slippingmore
present [see Doswell, 1987; Rockwood and Maddox, I988].
or lessunnoticedthroughthe present-dayobservingsystem,
can be the difference between correct and incorrect fore- The forecastermustformulate a correctprior assessment of
tornado potential, then carefully monitor the observations
casts.The sortsof eventsusedby Miller [1972]to exemplify
andwatchfor crucialdevelopmentsthat mightescapenotice
tornadicenvironmentsare not as difficultto predict,because
even when they occur.
Recent research has suggestedthat the vertical wind shear
0.05
SELS SKILL, WITH AND WITHOUT REPORT INFLATION structureis the most crucial elementin supercelIs.Thus the
0,045 sourceof rotation in supercell-relatedtornadoesseemsto be
the verticalwind shearin the environment.Staticinstability

0.04
•-I
doesnot seemparticularlyusefulin distinguishing supercell
0.035 d from nonsupercellevents, since supercelIsoccur within a
0,03 broadrangeof instabilities[Johnset al., 1990,this volume],
althoughit maybe importantin determining otheraspectsof
severe thunderstormpotential.
To the extent that supercelIs are responsiblefor most
strong and violent tornadoes, the current task of tornado
forecastinghingeson predictingenvironmentsfavorablefor
supercelIs.
Thisbeginswith the requirement that deep,moist
0.005

0.0
54 56 58 60 62 64 66 68 70 convectionbe possible;in turn, this dependson having
72 74 76 78

sufficient
moisture,instability,andlift suchthat potentially
YEAR

Fig. 10. Plot of yearly values in Heidke skill score, both unad- buoyantparcelsreach their LFCs. A large group of the
justed and adjusted(as describedin the text) for "inflation" of empirical forecasting"rulesof thumb"havephysicalexpla-
nontornadicsevere weather reports. nationsrooted in the requiredpresenceof deep,
568 TORNADO FORECASTING

convection.Beyondthis, for supercelIsit appearsthat the proveour capacityto diagnoseandanticipatethe wind shear
developmentof a supercell'sdeep, persistentmesocyclone structure.
takes longer than the 20- 40-rain lifetime of an ordinary Widespread deployment of automated surface observa-
convective cell. Long lifetimes for convective eventsare the tions should enhance the resolution of our surface data. The
result of propagation,with new convectivecellsdeveloping new technologiesthat make automated surface observations
in preferredlocationsrelativeto existingcells.Preferential feasible have made it realistic to propose the operation of
convectivedevelopmentis now known to be relatedto the what we now view as research-density mesonetworks over
verticalwind shear[seeWeismanand Klemp, 1982,1984].It much of the country as we enter the new century. In the
is also importantthat the combinationof vertical wind shear past,processesobservedwith high-resolutionnetworks [see
and storm motion producesenoughstorm-relativehelicity Fujita, 1963] were not resolved in operational networks.
[e.g., Davies-Joneset al., 1990]to allow the mesocycloneto With the proliferation of such networks in operations, it
reach down to the surface [Brooks et al., this volume]. shouldbe possibleto put into practice the conceptsderived
Despite some recent progress,these ideas have yet to be from the researchnetwork observationsof the past, as well
validatedin forecastingpractice,in part owingto a lack of as to do new researchon a wider range of situationsthan the
mesoscale detail in the observations. research networks, with their limited area and time of
Although nonsupercellevents have not yet received the operations,could sample.
researchattentiongiven to supercelIs,it appearsthat at least As of this writing, we are on the verge of having a
some of them may be related to mesoscaleprocessesasso- nationwide network of lightning ground strike detectors and
ciatedwith terrainfeatures[e.g., Brady and Szoke,1988].To a space-basedlightning mapper [Turman and Tettelbach,
whatever extent nonsupercelltornadoes are associatedwith 1980] that will allow observations of intracloud and inter-
topography, their prediction may be correspondingly cloud lightningas well as ground strikes. The ultimate value
straightforward [see Doswell, 1980; Weaver and Doesken, of such information in tornado forecasting remains un-
1991]. As of this writing, it is not known what fraction of all known. If such data are to be of value in forecasting, we
tornadoesfall into this categoryof being terrain related. must integrate the lightning data with the rest of the obser-
vations in determininghow lightningground strike informa-
tion relates to storm severity [see MacGorman, this vol-
6. FUTURE DIRECTIONS
ume].
We have indicatedthat tornadoforecastingis presentlyin Multispectral satellite observationsof an unprecedented
a state of rapid change.We have emphasizedhow important scope are promised for the near future, also. We believe the
scientificunderstandinghas becomein the tornadopredic- real value in such observationsis not in trying to emulate
tion problem. It is the new technologiesthat will make new direct measurement(e.g., rawinsonde)data but in usingthe
sciencepossible,but breakthroughs will not arisesimplyby new data in ways that are consistent with their character
switching on new systems. Systematic research will be (e.g., layer-averaged variables [see Fuelberg and Olson,
neededto achievetechnology'spromise, and a commitment 1991]). The most complete and useful observations will
to transferringnew understandingto operationsis required. involve the union of all observingtechnologies(e.g., Earth-
As new ideas are developed, their forecast value must be and space-basedremote sensingtechnologiessuch as wind
verified rigorously and the ideas must be modified on the and thermodynamicprofilersand Doppler radars, as well as
basis of the results of the verification. various direct measurements such as rawinsondes and air-
craft measurements),a task easier said than done.
6. I. Technological Tools Anothertechnologicaltool is the operationalmeteorolog-
ical workstation.Although new observingtechnologiesare
There are severalnew observingtechnologies to whichthe about to unleasha torrent of new data on the operational
operationalweatherservicesin the UnitedStatesalreadyare work environment,the samebasictechnologiesalso give us
committed.The WSR-88D radarnetworkis the keystonefor the capacityto absorbandintegrateit. If a futureoperational
the future health of tornado warnings, but what about workstationis to have a positive impact on forecasting, it
tornado forecasting?With the current emphasison the must meet two requirements. First, the workstation hard-
importanceof vertical wind shearin supercellstorms,the ware must have sufficientdata processingresourcesto deal
clearair windprofilingcapabilityof the WSR-88Ds(e.g., see effectively in real time with the data transfer rates associated
Plate3) becomes potentiallyvaluable.Sincethiscapabilityis with the new observations. Second, the workstation soft-
in generallimited to low levels,it nicely complements the ware must make this torrent of data availablein operation-
vertical wind profilers, which do not provide low-level ally useful ways.
winds. An opportunity for increasingour wind observations The last technologicaltool for the future we shall discuss
is in the use of automated instruments on commercial is the numericalprediction model. There can be no doubt
aircraft,whichgatherhigh-resolution dataduringascentand that numerical prediction models will continue to assume
descent(in effectcreatingwind and temperaturesoundings ever greaterrolesin the tornadoforecastproblem.Clearly,
near major air terminals). Altogether, these enhancementsto the tornadois manyordersof magnitudesmallerthan the
observations of the mesoscale flow variations should im- environmentalprocessesthat give it birth, so scale
 DOSWELL ET AL. 569

an outcome depends on (1) successfulresearch aimed at

increasing understanding of storm-environment relation-
ships, (2) a meaningfuleducation and training program for
operationalforecasters, and (3) a collective will to overcome
a decades-longschismbetween research and operations.
An issueof considerable concern is the forecasting of the
detailed thermodynamic structure of the storm environment,
be it tornadic or not. Present models and observations give
only a synoptic scale picture that often is inadequate to
resolve the important mesoscaledetails. Of course, on some
days it is possibleto do an adequatejob with the subjective
and objective forecasting tools at hand. The processesby
which moisture, momentum, and static stability change are
well known in principle, but those processesare not neces-
sarily well handled even in research-level modeling, much
less in operationalpractice. Some of this stems from lack of
resolution, both in our models and in our observations; some
of the inadequacies arise from deficiencies in the physical
parameterizationsused in numerical models. The parameter-
Fig. 11. Output of radar reflectivity (contouredat 10 dBz inter- ization of convection is a crucial issue in the large-scale
vals)andvorticity(contouredat 0.0025s-• interval,with the zero
contour suppressed)from a numericalcloud model; tic marks are forecast evolution of static stability; release the convection
5.45 km apart. The modelbeganwith horizontallyhomogeneous too soon and the model never develops realistic convective
initial conditions,usinga forecastthermodynamic soundingand instability, whereas if the parameterized convection is
hodographon May 26, 1991.The forecastinput is valid for north- "turned on" too late, the forecast instability can reach
westernOklahoma,anda supercelltornadicstormdeveloped that unrealistically high levels.
afternoon near Woodward, Oklahoma.
If we continueto dependon physical parameterizationsin
our numerical models, as we almost certainly shall, it is
tion is a crucialquestion.However, it will be manyyears(if obvious that improved parameterizations are needed. We
ever) before a singlenumericalpredictionmodelwill encom- believe that research aimed at improving physical under-
passall these scalesexplicitly and simultaneously. standingis preferable to improving parameterizationsin the
The real impactof mesoscalenumericalpredictionmodels long run; parameterization is, in eft•ct, a concessionmade in
in operationsremainsto be seen,but experimentalreal-time ignorance, even when it is a necessary concession. Of
predictions with mesoscale models are underway [see course,improvedphysicalunderstandingcan have a positive
Warner and Seaman, 1990]. Even numerical cloud models impact on parameterization schemes, as well.
Sensible and latent heat fluxes can have an enormous
may have a role in operationalforecasting(see Figure 11).
See Brooks et al. [1992] for an extended discussionon this impact on static stability and moisture availability. The
topic. The capacityfor suchmodelingcan only increase,so operationalmodel "boundary layer physics" parameteriza-
suchmodelscould cometo be viewed as essentialdiagnostic tions leave much to be desired from the viewpoint of a
tools [see Keyser and Uccellini, 1987] of the forecaster. The tornadoforecaster,especiallyin data-void areas like the Gulf
forecasterof the future must be much more than a passive of Mexico [see Lewis et al., 1989].
recipient of model "guidance" [see Snellman, 1977]; the In addition to the thermodynamic fields, it is clear that
forecaster should be using numerical models in a way with the apparentimportance of vertical wind structurefor
analogousto using pencils and paper weather charts. The supercelIs,wind forecastingis a key problem in numerical
computer should not be used to do what can be done with a prediction.Owing to their poor resolutionusingpresent
pencil on paper, however. Rather, the numerical models operational data, mesoscale structures in the wind field are
shouldform the basisfor operationallyusefulexplorationof createdby, rather than included initially in, the model. Our
the intricate,nonlinearrelationshipsassociatedwith weath- experiencesuggests that currentoperationalweatherpredic-
er-making processes. tion models are rather unsuccessfulin predictingthose
importantdetailsreliably. The modelshave their successes,
but theredoesnot seemto be any consistencyin them; the
6.2. ProductiveAreasfor Research predicted details are not reliable, at least for the needs of a
We believe that tornado forecastingeventuallywill out- tornado forecaster.
grow its purely associative, empirical roots. The new ob- As the new, more detailed observationsaccumulate,we
serving systems mean that forecasters and researcherswill may be able to developnew understandingof thesemeso-
be sharingthe samedata sets in the near future and that the scaleprocesses, heretoforeinadequatelysampled.This un-
applicationof meteorologicalscienceto tornado forecasting derstandingmay lead to improvementsin mesoscalenumer-
probablywill be a more natural processthan it is now. Such ical prediction that will be essential to the tornado
570 TORNADO FORECASTING

problem. However, an additionalcomplicationto forecast- Davies-Jones, R. P., Streamwise vorticity: The origin of updraft
ing detailsof the wind structureis the impactfrom convec- rotation in supercell storms, J. Atmos. Sci., 41, 2991-3006, 1984.
tion. Maddox [1983], Ninomiya [1971], and others have Davies-Jones, R. P., D. W. Burgess, and M. Foster, Test of helicity
as a tornado forecast parameter, in Preprints, 16th Conference on
shown that persistent deep convection can alter the sur- Severe Local Storms, pp. 588-592, American Meteorological
rounding environment. This means that new convection Society, Boston, Mass., 1990.
developingnear preexistingstormswill encountera different Doswell, C. A., III, Synoptic scale environments associated with
wind and thermodynamic environment than the convection High Plains severe thunderstorms,Bull. Am. Meteorol. Soc., 60,
1388-1400, 1980.
which preceded it.
Doswell, C. A., III, The operational meteorology of convective
Given all of the new technologicaland associatedscientific weather, vol. I, Operational mesoanalysis, NOAA Tech. Metno.
developmentsthat are likely to take place in the next decade, NWS NSSFC-5, 172 pp., Natl. Severe Storms Forecast Center,
it is quite plausible to be optimistic about the future of Kansas City, Mo., 1982.
tornado forecasting. There can be no doubt that considerable Doswell, C. A., III, The operational meteorology of convective
progresswill be made during the next 10 years, and we look weather, vol. II, Storm scale analysis, NOAA Tech. Metno. ERL
ESG-15,240 pp., Natl. Severe Storms Forecaster Center, Kansas
forward to thosedevelopments.
City, Mo., 1982.
Doswell, C. A., III, The distinction between large-scaleand meso-
scale contributionsto severe convection: A case study example,
Acknowledgments. The authors wish to thank Robert A. Mad- Weather Forecasting, 2, 3-16, 1987.
dox (NSSL) and the anonymous reviewers for their helpful com- Doswell, C. A., III, and D. W. Burgess, Tornadoes and tornadic
ments on previousversionsof this paper. Thanks are alsoextended storms: A review of conceptual models, this volume.
to Dave Keller (NSSL) for his efforts in the verification study and Doswell, C. A., III, J. T. Schaefer, D. W. McCann, T. W. Schlatter,
Joan Kimpel (Cooperative Institute for Mesoscale Meteorological and H. B. Wobus, Thermodynamic analysis procedures at the
Studies)for her help in drafting, as well as Louis Wicker (University National Severe Storms Forecast Center, in Preprints, 9th Con-
of Illinois and National Center for SupercomputingApplications) ference Weather Forecasting and Analysis, pp. 304-309, Ameri-
and Harold Brooks (National Research Council) for their help with can Meteorological Society, Boston, Mass., 1982.
Figure 11. Doswell, C. A., III, R. A. Maddox, and C. F. Chappell, Fundamen-
tal considerationsin forecasting for field experiments, in Pre-
prints, I lth Conferenceon Weather Forecasting and Analysis,
REFERENCES
pp. 353-358, American Meteorological Society, Boston, Mass.,
1986.
Beebe, R. G., and F. C. Bates, A mechanismfor assistingin the
Doswell, C. A., III, R. Davies-Jones,and D. L. Keller, On summary
release of convective instability, Mort. Weather Rev., 83, 1-10,
1955. measuresof skill in rare event forecastingbased on contingency
tables, Weather Forecasting, 5, 576-585, 1990a.
Brady, R. H., and E. Szoke, The landspout--A common type of
Doswell, C. A., III, D. L. Keller, and S. J. Weiss, An analysisof the
Northeast Colorado tornado, in Preprints, i5th Conferenceon
temporal and spatial variation of tornado and severe thunderstorm
Severe Local Storms, pp. 312-315, American Meteorological
Society, Boston, Mass., 1988. watch verification, in Preprints, 16th Conference on Severe Local
Brandes, E. A., Flow in severe thunderstormsobservedby dual- Stortns, pp. 294-299, American Meteorological Society, Boston,
Doppler radar, Mort. Weather Rev., 105, 113-120, 1977. Mass., 1990b.
Brooks, H. E., C. A. Doswell IIi, and R. P. Davies-Jones, Envi- Fawbush, E. J., and R. C. Miller, A mean soundingrepresentative
ronmental hellcity and the maintenanceand evolution of low-level of the tornadic airmass environment, Bull. Am. Meteorol. Soc.,
mesocyclones, this volume. 33, 303-307, 1952.
Brooks, H. E., C. A. Doswell III, and R. A. Maddox, On the use of Fawbush, E. J., and R. C. Miller, The types of air massesin which
mesoscale and cloud-scale models in operational forecasting, North American tornadoes form, Bull. Am. Meteorol. Soc., 35,
Weather Forecasting, 7, 120-132, 1992. 154-165, 1954.
Browning, K. A., Airflow and precipitation trajectories within Fawbush, E. J., R. C. Miller, and L. G. Starrett, An empirical
severe local storms which travel to the right of the winds, J. method of forecastingtornado development, Bull. Am. Meteorol.
Atmos. Sci., 22,664-668, 1964. Soc., 32, 1-9, 1951.
Browning, K. A., and T. Fujita, A family outbreak of severe local Fuelberg, H. E., and S. R. Olson, An assessmentof VAS-derived
stormstoA comprehensivestudy of the storms in Oklahoma on 26 retrievals and parameters using in thunderstorm forecasting,
May 1963, Part 1, U.S. Air Force Spec. Rep. 32, AFCRL-65- Mon. Weather Rev., 119, 795-814, 1991.
695(1), 346 pp., L. G. HanscornField, Bedford, Mass., (Available Fujita, T., Analytical mesometeorology: A review, Meteorol.
as NTIS AD 623787from Natl. Tech. Inf. Serv., Springfield,Va.) Monogr., 5(27), 77-125, 1963.
Browning, P. A., The VDUC interactive computer system at the Gage, K. S., and B. B. Balsley, Dopplerradar probingof the clear
National Severe Storms Forecast Center, in Preprints, 7th Inter- atmosphere,Bull. Am. Meteorol. $oc., 58, 1074-1093, 1978.
national Conferenceon Interactive Information and Processing Galway, J. G., The lifted index as a predictor of latent instability,
Systemsfor Meteorology, Oceanography,and Hydrology, pp. Bull. Am. Meteorol. Soc., 37, 528-529, 1956.
204-207, American Meteorological Society, Boston, Mass., 1991. Galway,J. G., Severelocal stormforecastingin the United States,
Burgess,D. W., and E. B. Curran, The relationshipof stormtype to SELS workshopnotes, 13 pp., Natl. Severe Storms Forecast
environment in Oklahoma on 26 April 1984, in Preprints, I4th Center, KansasCity, Mo., 1973.
Conference on Severe Local Storms, pp. 208-211, American Galway, J. G., Relationshipof tornado deaths to severe weather
MeteorologicalSociety, Boston, Mass., 1985. watch areas, Mort. Weather Rev., 103, 737-741, 1975.
Burgess, D. W., R. J. Donaldson, Jr., and P. R. Desrochers, Galway, J. G., The evolution of severe thunderstorm criteria within
Tornado detection and warning by radar, this volume. the WeatherService,WeatherForecasting,4, 585-592, 1989.
Davies, J. M., and R. H. Johns, Some wind and instabilityparam- Galway,J. G., J.P. Finley: The first severestormsforecaster,Bull.
eters associatedwith strong and violent tornadoes, 1, Wind shear Atn. Meteorol. Soc., 66, !389-1395, 1985.
and helicity, this volume. Hales, J. E., Jr., A subjectiveassessmentof model initial
 DOSWELL ET AL. 571

usingsatelliteimagery,
Bull.Am. Meteorol.Soc.,60, 206-211, sphere Electricity, pp. 692-697, American Meteorological Soci-
1979.
ety, Boston, Mass., 1990.
Hales,J. E., Jr., Synoptic
features
associated
withLosAngeles National Severe Storms Project Staff, Environmental and thunder-
tornadooccurrences,
Bull. Am. Meteorol.Soc., 66, 657-662, storm structure as shown by National Severe Storms Project
1985. observationsin spring1960-61, Mort. Weather Rev., 91,271-292,
1963.
Hirt, W. D., Forecasting
severeweatherin North Dakota,in
Preprints,
14thConferenceonSevereLocalStorms, pp.328-331, National Weather Service, 1987: Assessmentplan for the uses of
AmericanMeteorologicalSociety,Boston,Mass.,1985. Wind Profiler technology in NWS operations, internal planning
House,D.C., Forecastingtornadoesand severethunderstorms, document,70 pp., Office of Meteorol., Silver Springs, Md., 1987.
Meteorol. Monogr., 5(27), 141-155, 1963. Ninomiya, K., Mesoscale modification of synoptic situations from
Johns,R. H., A synopticclimatologyof northwestflow severe thunderstormdevelopmentas revealed by ATS III and aerological
data, J. Appl. Meteorol., 10, ! 103-1121, 1971.
weatheroutbreaks, II, Meteorologicalparametersandsynoptic Purdom, J. F. W., Subjective interpretation of geostationary satel-
patterns,Mon. WeatherRev., 112,449-464, 1984.
lite data for nowcasting, in Nowcasting, edited by K. Browning,
Johns,R. H., J. M. Davies,and P. W. Leftwich,An examinationof pp. 149-166, Academic, San Diego, Calif., 1982.
therelationshipof 0-2 kmAGL "positive"windshearto poten- Rabin, R. M., and D. S. Zrnic', Subsynoptic-scale vertical wind
tial buoyantenergyin strongandviolenttornadosituations, in revealedby dual Doppler-radar and VAD analysis,J. Atmos. Sci.,
Preprints,16thConference
onSevereLocalStorms,pp.593-598, 37, 644-654, 1980.
AmericanMeteorologicalSociety,Boston,Mass., 1990. Rockwood, A. A., and R. A. Maddox, Mesoscale and synoptic scale
Johns,R. H., J. M. Davies, and P. W. Leftwich, Somewind and interactions leading to intense convection: The case of 7 June
instabilityparameters associated with strongand violenttorna- 1982, Weather Forecasting, 3, 51--68, 1988.
does,2, Variationsin the combinations of wind andinstability Schaefer, J. T., Severe thunderstorm forecasting: A historical
parameters, this volume. perspective, Weather Forecasting, 1, 164--189, 1986.
Kelly, D. L., j. T. Schaefer,R. P. McNulty,C. A. DoswellIII, and Setv•ik,M., and C. A. Doswell III, The AVHRR channel3 cloudtop
R. F. Abbey, Jr., An augmentedtornadoclimatology, Mon. reflectivity of convective storms, Mon. Weather Rev., 119, 841-
Weather Rev., 106, 1172-1183, 1979. 847, 1991.
Kelly,D. L., J. T. Schaefer,andC. A. DoswellIII, Theclimatology Showalter, A. K., A stability index for thunderstorm forecasting,
of non-tornadic severethunderstorm eventsin the UnitedStates, Bull. Am. Meteorol. Soc., 6, 250-252, 1953.
Mon. Weather Rev., 113, 1997-2014, 1985. Showalter,A. K., andJ. R. Fulks, Preliminaryreport on tornadoes,
162pp., U.S. Weather Bur., Washington, D.C., 1943.
Keyser, D., and L. W. Uccellini, Regionalmodels:Emerging
researchtools for synopticmeteorologists,
Bull. Ant. Meteorol. Snellman, L. W., Operational forecastingusing automatedguid-
ance, Bull. Am. Meteorol. Soc., 58, 1036-1044, 1977.
Soc., 68, 306-320, 1987.
Stensrud,D. J., and R. A. Maddox, Opposingmesoscalecircula-
Leftwich, P. W., Jr., On the use of helicityin operationalassess- tions: A case study, Weather Forecasting, 3, 189-204, 1988.
ment of severelocal stormpotential,in Preprints,16thConfer- Tepper, M., A proposedmechanismof squall lines: The pressure
ence on Severe Local Storms, pp. 306-310, American Meteoro- jump line, J. Meteorol., 7, 21-29, 1950.
logical Society, Boston, Mass., 1990. Tepper, M., MesometeorologymThe link between macroscale at-
Leftwich, P. W., Jr., and R. W. Anthony, Verificationof Severe mosphericmotions and local weather, Bull. Am. Meteorol. Soc.,
Local Storms ForecastIssued by the National Severe Storms 40, 56-72, 1959.
ForecastCenter: !990, NOAA Tech. Mento. NWS NSSFC-31, 9 Turman, B. N., and R. J. Tettelbach, Synoptic-scale satellite
pp., 1991. (Available at Natl. Severe Storms ForecastCenter, lightning observation in conjunction with tornadoes, Mon.
Kansas City, Mo.) Weather Rev., 108, 1878-1882, 1980.
Lewis, J. M., C. M. Hayden, R. T. Merrill, and J. M. Schneider, Wakimoto, R. M., and j. W. Wilson, Non-supercelltornadoes,
GUFMEX: A study of return flow in the Gulf of Mexico, Bull. Mort. Weather Rev., 117, ! 113-1140, 1989.
Am. Meteorol. Soc., 70, 24-29, 1989. Warner, T. T., and N. L. Seaman, A real-time, mesoscalenumerical
Lloyd, J. R., The developmentand trajectoriesof tornadoes,Mort. weather-predictionsystemusedfor research,teaching,andpublic
Weather Rev., 70, 65-75, 1942. serviceat the PennsylvaniaState University, Bull. Am. Meteorol.
MacGorman, D. R., Lightningin tornadic storms:A review, this Soc., 71,792-805, 1990.
volume. Weaver,J. F., and N.J. Doesken, High PlainssevereweathermTen
Maddox, R. A., Large-scalemeteorologicalconditionsassociated yearsafter, WeatherForecasting, 6, 411-414, 1991.
with midlatitude, mesoscale convective complexes, Mon. Weisman,M. L., andJ. B. Klemp, The dependence of numerically
Weather Rev., 111, 1475-1493, 1983. simulatedconvectivestormson vertical wind shearand buoy-
Maddox,R. A., andC. A. DoswellIII, An examination ofjet stream ancy, Mon. Weather Rev., I10, 504-520, 1982.
configurations,500 mb vorticity advectionand low level thermal Weisman, M. L., and J. B. Klemp, The structure and classification
advectionpatternsduringextendedperiodsof intenseconvection, of numericallysimulatedconvectivestormsin directionallyvary-
Mort. Weather Rev., I10, 184-197, 1982. ing wind shears,Mon. Weather Rev., 112, 2479-2498, !984.
Mass, C. F., Synopticfrontal analysis:Time for a reassessment?, Weiss,S. J., On the operationalforecastingof tornadoesassociated
Bull. Am. Meteorol. Soc., 72, 348-363, 1991. with tropicalcyclones,in Preprints,14th Conferenceon Severe
McCann, D. W., The enhanced-V: A satellite observable severe Local Storms,pp. 293-296, AmericanMeteorologicalSociety,
storm signature,Mort. Weather Rev., III, 887-894, 1983. Boston, Mass., 1985.
Miller, R. C., Notes on the analysisand severe-stormforecasting Weiss, S. J., On the relationship between NGM mean relative
proceduresof the Air Force Global Weather Central, Tech. Rep. humidityandtheoccurrence of severelocalstorms,in Preprints,
200 (revision), 190pp., Air Weather Serv., Scott Air Force Base, 15thConferenceon SevereLocal Storms,pp. Jlll-Jl14, Ameri-
Ill., 1972. can MeteorologicalSociety, Boston, Mass., 1987.
Moncrieff,M. W., and M. J. Miller, The dynamicsand simulationof Weiss, S. J., D. L. Kelly, and J. T. Schaefer,New objective
tropical cumulonimbusand squalllines, Q. J. R. Meteorol. $oc., verificationtechniquesat the National Severe Storms Forecast
102,373-394, 1976. Center,in Preprints,8th Conference
on WeatherForecasting
and
Mosher, F. R., and J. S. Lewis, Use of lightninglocationdata in Analysis,pp. 412-419,AmericanMeteorological
Society,Boston,
severe storm forecasting, in Preprints, Conference on Atmo- Mass.,
 SomeWind and InstabilityParametersAssociated
With Strong and Violent Tornadoes
1. Wind Shear and Helicity
JONATHAN M. DAVIES

Pratt, Kansas 67124

ROBERT H. JOHNS

.NationalSevereStormsForecastCenter, KansasCity, Missouri64106

1. INTRODUCTION demonstrate the development of a mesolow in the inflow

region of tornadic supercelIs, initiated by the dynamical
Althoughthe verticalwind profilethroughthe troposphere interaction of the updraft with shear-induced vertical pres-
hasbeenrecognizedto be importantin tornadodevelopment sure forces [Rotunno and Klemp, 1982] that are largely a
since the beginning of tornado forecasting efforts in the result of low-level curvature shear and helicity. This feature
1950s, only recently have researchersbegun to investigate serves to intensify the vertical velocity of the growing
more detailed characteristicsof wind profilescontributingto updraft, as well as to strengthen inflow into the storm, an
low-level mesocycloneformation and tornadoproductionin important requirement noted by Lazarus and Droegemeier
supercellthunderstorms.Davies-Joneset al. [1990] provide [1990].
an overview of recent work completed in this area. From The strength of the wind fields and shear through a layer
modeling results and storm observations,it appears that deeper than the low levels is also significant in supercell
both (1) the wind profile in the low levels (i.e., the storm development. While low-level curvature shear is an impor-
inflow layer) and (2) the strengthof the wind field and shear tant factor for producing the vertical pressure gradient
extending through a deeper layer of the troposphere (i.e., mentioned in the prior paragraph, Weisman and K!emp
through middle levels) are important to supercell-induced [1986] note that it is also necessary for the vertical wind
tornado development. shear to extend through middle levels of the troposphereto
Regarding the low-level wind profile, mean shear [Ras- sustaina rotating updraft via shear-inducedpressureforces.
mussen and Wilhelmson, 1983; Davies, !989] is one param- One measureof this deeper ambient shear is the U param-
eter that has been used for assessingthe veering of winds eter in the denominator of the Bulk Richardson Number
with height and turning of the hodographassociatedwith [Weisman and Klemp, 1982, 1986]. Doswell [1991] also notes
supercell thunderstorms. More recently, storm-relative he-
that middle-levelwinds of sufficientstrengthare a necessary
licity [Lilly, 1986; Davies-Jones et al., 1990], which ad-
componentof hodographsthat can support supercell devel-
dressesthe importance of the wind fields viewed from a
opment. A similar requirement is mentioned by McCaut
storm's frame of reference, has become recognized as an
[1991], who notes that 700 mbar wind speeds show good
effectiveparameterfor measuringrotationalpotentialin the
correlationwith tropical cyclone tornado outbreaks. Strong
low-level wind field. Hellcity in the storm inflow layer is
middle-levelwinds may move precipitationdownwindout of
associatedwith streamwise vorticity [Davies-Jones, 1984]
the upper portion of the updraft (or in some cases tilt the
which, when "ingested" and tilted into the stormupdraft,
updraft so that precipitation falls out downwind), thereby
inducesrotation. Also, numericalsimulationsby Brooksand
Wilhelmson[1990], Brooks et al. [1992], and McCaul [1990] eliminatinga potentialimpedimentto the developmentof a
strong and sustainedupdraft.
The Tornado: Its Structure, Dynamics, Prediction,and Hazards.
On the basis of the above discussion,the following wind
GeophysicalMonograph79 parameters were chosenfor examination in this paper: (1)
Copyright1993by the AmericanGeophysicalUnion. low-level mean shear(0-2 kin/0-3 km/0-4 km above ground
574 PARAMETERS ASSOCIATED WITH STRONG AND VIOLENT TORNADOES, 1

level (AGL)), (2) storm-relative low-level helicity (0-2 km/ In spiteof this effortto incorporatelocal data, the authors
0-3 km/0--4 km AGL), (3) Bulk Richardson Number shear recognizethat mesoscalevariationsnot detected by the
(U) (boundary layer through 6 km AGL), and (4) mean wind soundingnetwork or surfaceobservationscan have major
speedin the middle levels (3-6 km AGL). Johns et al. [ 1990] impacton thunderstorm behavior.Thereforeparameterval-
(henceforth referred to as JDL) assembled a large and ues obtained in this study can be viewed as only roughly
comprehensivedata set of strong and violent mesocyclone- representative
of the actualprestormwindenvironmentsfor
inducedtornadoes(242 cases)for the purpose of examining eachcase.By limitingthe studyto casesfrom nonlocalized
the associatedmean shear and buoyancy values. The same tornado outbreaks, the probability is increased that rawin-
data set is utilized in this study to examine the aforemen- sonde network observations sampled at least some aspects
tioned wind parameters, with suggestionsfor possibleappli- of the wind environmentssupportingthe observedtornadic
cation to operational forecasting. supercellthunderstorms.

4. MEAN SHEAR RESULTS

2. CASE SELECTION METHODOLOGY
From Rasmussen and Wilhehnson [1983], mean shear (S)
As describedby JDL, Storm Data was examinedsystem-
is essentially
atically for the 10-year period April 1980 through March
1990. All nonlocalized tornado outbreaks that involved six or
S = hodograph
length(m s-i)/depthof layer(m)
more tornadoes were selected, with the added requirement
that two or more of the tornadoes be F2 or greater in "Positive" mean shear [Davies, 1989], as amended by JDL,
intensity and separatedby at least 60 nautical miles(111 km). modifies this definition by setting the shear magnitude to
Also, any F3 or greater intensity tornado accompaniedby an zero for those hodograph segments where the ground-
uncontaminated proximity sounding was selected. These relative winds back significantly with height. Unlike storm-
criteria were used in an attempt to eliminate most tornadoes relative helicity, mean shear does not consider storm motion
of nonsupercellorigin [Wakimotoand Wilson, 1989].In large and is computationallyvery sensitiveto the fine-scalewind
outbreaks, more than one case was selected if the constitu- structure. Recognizing these limitations, positive mean
ent tornadoes were separated by >200 nautical miles (370 shear can, in many cases, provide an estimated assessment
km) in distance or -> 10 hours in time. of rotational potential when a storm motion is not available.
Averagepositivemeanshearmagnitudes
(x10-3 s-1)
3. DETERMINATION OF PARAMETER VALUES
were computedfor the 242 casesin the JDL data set using
.
three atmosphericdepths (AGL) for comparison:
As discussedin detail by JDL, parameterscomputedfrom
0- 2-km average positive shear 13.6
rawinsonde observationswere either used directly or inter-
0- 3-km average positive shear 10.7
polated, dependingon tornado case time and distancerela- 0-4-km average positive shear 9.1
tive to soundings. In general, if a tornado case occurred
within 3 hours and 75 statute miles (121 km) of an uncon- The fact that the 0- 2-km averagepositive shear is larger than
taminated soundingin the warm sector, parameters com- both the 0- 3-kin and 0- 4-km averages suggeststhat in most
puted from the observation were used directly. If the tor- casesa majority of the low-level wind shear is below 2 km
nado case was more removed in distance, yet within 3 hours AGL, similar to results from Davies [1989] using a smaller
of soundingtime, an interpolation was performed between data set.
parametervaluesfrom two or more soundings,dependingon Grouping casesby tornado intensity (205 strong and 37
the availability and location of adjacent warm sectorrawin- violent tornado cases), average positive mean shear values
sonde observations. If the tornado case was more removed were

in time, an interpolation of soundingparameter values in

time and distancewas performedby observingthe evolution Strong(F2/F3) Violent(F4/F5)
0- 2-km average positive shear 13.4 14.7
of wind featuresbetween soundingtimes, providedthat wind
0- 3-km average positive shear 10.5 11.7
fields were well defined on the synoptic scale. Casesthat did 0- 4-km averagepositive shear 9.0 10.0
not meet these criteria or that involved significant amounts
of missing data were not accepted into the data set. These values suggestthat, on an average, violent tornadoes
The limitations of the coarse rawinsonde network are well occur with higher values of low-level positive mean wind
known. In an attempt to incorporate localized data, surface shear than do tornadoes of lesser intensity.
wind observations were also examined for each case. If the
wind direction from a surface observation nearby in the 5. ESTIMATION OF STORM MOTION
warm sector exhibited significant backing (i.e., >20 ø) when FOR STORM-RELATIVE HELICITY
comparedto the surface wind of the sounding(s)associated COMPUTATIONS
with the case, the wind from the surface observation was
blended into the sounding wind profile(s) for computing In order to calculatestorm-relative helicity a storm motion
parameter values such as helicity and mean low-level shear. is required. Storm motionsoriginallywere not obtained
 DAVIES AND JOHNS 575

the JDL data set. For this study the authorshave examined rangingfrom77 to 122%(threeof 11casesmovedat 100%or
a subsetof casesfor whichradar imagerywas availableat more of the mean wind speed).This differencesuggeststhat
the National Severe Storms Forecast Center. This subset of the strengthof the wind fields (which varies by seasonand
31 casescontainstornadoes that occurredduringtheperiod locale) may play a significantpart in storm motion.
1988-1990. Two of the cases from 1990 were not in the
originalJDL data set. Mean environmentalwinds (0- 6-km 5.2. Storm Motion Related to Wind Field Strength
AGL [after Weismanand Klemp, 1986])were computed In the 31 case subset, extreme right-moving storms (i.e.,
from the soundingdata for these cases,and comparisons >25ø right) are found to be associatedwith significantly
were made with the storm motions obtained from the radar weaker wind fields (0- 6-km mean wind speedsaveraging 32
data. Though the storm motion sampleis small, the cases knots(16.5m s-i)) thanthe othercases(0- 6-kmmeanwind
appearto representall seasonsandmostareasof the country speeds averaging46knots(23.7m s-l)). Thistendency for
east of the Rocky Mountains. It should be noted that the tornadic storms in relatively weak wind fields (i.e., 0-6-km
stormmotiondeviationshere are not exactlycomparableto meanwindmagnitudes <30-35knots(15.5-18.0 m s-l)) to
Maddox [1976], Darkow and McCann [1977], and Bluestein deviate increasinglyright of the mean wind also can be seen
and Parks [1983], who computedthe mean wind through in a studyby Darko•vand McCann [ 1977].If their Figures3a
much deeperlayers that generallyencompassed the tropo- and 1a are compared,it becomesapparentthat their extreme
sphere. right-movingstormsoccurred in significantlyweaker wind
fields when comparedto their data set as a whole.
5.1. Observed Storm Motions From the Data Subset
The weakest 0- 6-km mean wind speedsfound in the 31
The directionaldeviationswithin the 31 case subsetvary casesubset werenear20 knots(10.3m s-l), suggesting that
from 2ø left to 42ø right of the 0- 6-km mean wind. Speed mean wind speedsof lesser magnitudemay be too weak to
deviations vary from 58 to 158% of the 0- 6-km mean wind support strong or violent tornadoes in supercell thunder-
speed. It is notable that five of the casesmoved faster than storms.For purposesof this discussion,relatively "weak"
the mean wind speed; four of these were associatedwith wind fields that can support significant supercell-induced
bow echo structures [see Doswell et al., 1990]. From these tornadoes will be defined as those with 0- 6-km mean wind
results it is apparent that for any given sampleof tornado- speeds
-<30knots(15.5m s-1) and>20 knots(10.3m sl).
producing supercell storms there can be a wide range of
deviant motions when compared to the 0- 6-km mean envi- 5.3. Suggested Storm Motion Assumptions
ronmental wind. This fact illustrates that without a better On the basis of the above discussion, some rough first
understandingof the complex problem of storm motion it guessassumptionscan be suggestedregarding storm move-
will be difficult to implement a predictive cell movement ment for right-moving cells in a given wind regime that is
procedure that will be successfulin all cases. potentially tornadic. For environments characterized by
Despite this problem a closer examination of the storm relatively strong wind fields (i.e., 0- 6-km mean wind speeds
motion subset reveals some useful general information. >30 kt) a reasonable first guess movement would be a
Mean deviant storm motion for the 31 casesis 20øright of the motion such as 20R85, similar to the 31-case subset mean
0- 6-km mean wind at 89% of the mean wind speed (or deviant motion of 20R89. The storms in the subset that are
20R89, as deviations will henceforth be annotated). Roughly not extreme right movers yield a mean deviant motion of
half of the subset (15 cases) exhibits a directional deviation 16R86, also close to 20R85. When weaker wind fields are
within 5ø of this mean. This suggeststhat when considering involved (i.e., 0- 6-km mean wind speeds -<30 knots (15.5 m
supercell development, a motion deviation assumptionof sl)), a motioncloserto 30R75,asusedby Maddox[1976],is
20R85 or 20R90 would be a reasonably accurate forecast probably more appropriate. This is suggestedby the fact that
motion estimate roughly 50% of the time. the storms in the 31-case subset that were extreme right
Some storm motion differences are apparent when group- movers traveling at less than the mean wind speed yielded a
ing casesby both seasonand geographicallocale. From the mean deviant motion of 34R76, close to a 30R75 assumption.
data subset, tornadic storms in the Great Plains during It should be strongly emphasized that these estimates are
March through June (eight cases)display a mean motion of only general reference points and do not addressmany other
27R80, while tornadic storms east of the Mississippi during factors that affect storm motion, such as the orientation of
November through February (11 cases) exhibit a mean boundaries and low-level forcing.
motion of 18R97. The Great Plains storms have a directional For this study, these rough assumptions will provide
deviationspreadrangingfrom 13ø to 37ø right of the mean starting points, in lieu of actual observed storm motions, for
wind (with only one of eight caseslessthan 20øright) and a examination of storm-relative helicity utilizing the JDL data
speeddeviationspreadrangingfrom 61 to 122%of the mean set.
wind (only one case exceeded 100% of the mean wind
6. STORM-RELATIVE HELlCITY RESULTS
speed). In contrast, the cool season storms east of the
Mississippihave a smaller directional deviation spread, For this study, the authorshave computedintegratedtotal
rangingfrom 12øto 24øright of the meanwind (six of 11cases storm-relativehelicity (H) for specific layers. The computa-
moved less than 20ø right) and a speeddeviation spread tion used is that of Davies-Jones et al.
576 PARAMETERS ASSOCIATED WITH STRONG AND VIOLENT TORNADOES, 1

0- 2-km mean assumed helicity 332

H = -2 x the signedarea sweptout by the storm-relative 0- 3-km mean assumed helicity 356
wind vector while ascendingthrougha layer 0-4-km mean assumed helicity 375

Unitsarem2 s-2 (identical

to J kg-•). Helicityisusefulin As with the 31-case subset, it is notable that so much of the
measuringrotational potential becauseit has a soundphys-
low-level helicity is below 2 km AGL.
ical link to overallupdraftrotationand is sensitiveto storm
Groupingcasesby tornado intensity (205 strong and 37
motion while remainingcomputationallyinsensitiveto the
fine-scale wind structure.
violent tornado cases), helicity values were computed using
the same motion assumptions:
First, observedhelicity valuesobtainedfrom the 31-case
storm motion subset will be examined. Then, utilizing the Strong(F2/F3) Violent(F4/F5)
storm motion refinementassumptionsfrom section5, helic- 0- 2-kin mean assumedhelicity 317 415
ity values will be examinedusingthe entire 242-caseJDL 0- 3-kin mean assumedhelicity 339 452
data set. 0- 4-kin mean assumed helicity 357 478

As with observedhelicity from the 31 case subset,the larger

6.1. ObservedHelicity From the Storm Motion Subset data set yields similar results, showing violent tornadoes
occur, on an average, in environments with larger helicity
Helicity magnitudeswere computed for the observed values than do strong tornadoes. However, notable excep-
storm motions of the 31 subset cases and the same atmo-
tions certainly do occur, as will be seen in section 10.
spheric layers (AGL) used previously in computingmean For the 29 overlapping cases between the storm motion
shear.Mean helicitymagnitudes (m2 s-2) for the three subset and the JDL data set, over half the mean hellcity
layers were as follows. valuescomputedusingthe motion assumptionswere within
0- 2-kin mean observed hellcity 392
15% of the observed values, and all but eight cases were
0- 3-kin mean observed helicity 416 within 20% of the observedvalue. It therefore appearsthat
0-4-kin mean observed helicity 428 the 20R85/30R75assumptionsbased on general strength of
the wind fields are useful starting points for generating
Note that most of the low-level helicity residesbelow 2 km helicity output.
AGL, similar to the mean shear results in section 4. An Using this approach,Figure 1 illustrates a helicity forecast
averageof 94% of the 0- 3-km helicity is locatedwithin the 0- derived from the National Meteorological Center's Limited
2-km layer, and an average of 92% of the 0- 4-km helicity is Fine Mesh (LFM) model, blending winds aloft forecast data
located within the 0- 2-km layer. with surface winds from Model Output Statistics (MOS)
The 31 cases(21 strongand 10 violent tornado cases)were forecastguidance.For each output station the 0- 6-km mean
also examinedfrom the standpointof helicity and tornado wind direction and speed are printed along with helicity
intensity: values computed using storm motion deviations of 20R85
and 30R75 relative to the 0- 6-km mean wind. This would
Strong(F2/F3) Violent(F4/F5)
0- 2-km mean observedhelicity 359 460
allow a forecaster to see how the helicity changes with
0- 3-km mean observed helicity 369 519 increasingstorm deviation to the right at different locations
0-4-kin mean observed helicity 378 539 plottedon the samemap. On the basisof the strengthof the
0- 6-km mean wind or other factors the forecaster then could
Resultssuggestthat violenttornadoes,on an average,occur selectfor analysisthe deviationdeemedmostappropriatefor
with larger helicity magnitudesthan do strongtornadoes. the situation.This output format would also help the fore-
This agreeswith the generalhelicityguidelinespresented by caster assesswhether the risk is widespread or localized. If
Davies-Jones et al. [1990] for strong and violent tornado helicityvaluesare largeregardlessof deviation,the potential
development. for low-level rotation is more extensive than if helicity
valuesbecomesignificantonly with extremeright deviation.
This or a similarformat couldbe used to generatehelicity
6.2. HeliciO, Using the JDL Data Set and outputmapsfrom sourcessuchas rawinsondedata, profiler
an Assumed Storm Motion
and WSR-88D wind data, and numerical model forecasts.

For computing
helicity,stormmotionswereestimated
by
applying a 20R85 assumption when 0- 6-km mean wind BULK RICHARDSONNUMBER
speeds
were>30knots(15ms-l), anda 30R75
assumption SHEAR(U) RESULTS
when0- 6-kmmeanwindspeeds
were-<30knots(15m s-•)
(seesection5). Helicity magnitudeswere computedfor the As discussedin section 1, the Bulk Richardson Number
samethreelowertropospheric layersexaminedusingthe 31 (BRN) shear, U [Weismanand Klemp, 1982, 1986], is a
case storm motion subset, yielding the following mean parameter that measures the ambient environmental shear
values for the 242 cases.
thatis important
for supercell
 DAVIES AND JOHNS 577

0-6 KM WIND SHEAR (U)

O-3km
Hellcity
(m2s
-z)
12 hr LFM winds aloft forecast
valid OOUTC5/17/91
180/18 24 30
210/12 19 23
(240 CASES)

NR OF CASES
•i 160/16
ß
3749 •: 60

ß ,,-"'"'x
50/05'
• 5 190/24
I 141
ß 164
220/22 48 64 50 48

[ k :•220/12
6274 44

200/30 !61 182--,•

40 38

.... / 21e37 ......'

35
33

220/33 171 204 30

2.:•0/29
6999
.•
....'-•X,,,,......,.•J
. .

220/29.98
112
220/28
1011•"•........
'....'.......................
:::.- ß

20 17
14

Fig. 1. Example of 12-hourforecast of 0- 3-km AGL integrated

storm-relative helicity, derived from LFM forecast winds aloft data 10
and surfacedata from MOS guidancevalid at 0000 UTC May 17,
1991,for sitesat approximatedot locations.Three computations are
indicated for each site: the first number/groupis the 0- 6-km AGL
meanwind directionand speedin degreesandknots' the secondand i i

third numbersare helicity values derived using storm motionsof 10-12 13-15 16-18 19-2122-2425-2728-3031-3334-3637-394,0-42
20R85 and 30R75 (see text for explanationof annotation),respec-
tively. The 30R75 helicity (third number) is analyzed for values WIND SHEAR (M/S)
>-150m2 s -. In thisparticular
casethe LFM guidance
verified
reasonably well, and damaging supercell-inducedtornadoes oc-
curred near Wichita and Tulsa. Fig. 2. Distribution of casesfrom JDL data set, grouped according
to U magnitude(Bulk RichardsonNumber shear)(in unitsof meters
per second).

U = the straight vector difference between the 0- 6-km

AGL mean wind and the boundary layer wind The averagemean 3- 6-km wind speedfor the JDL data set
(i.e., 0- 500-m AGL mean wind) is49knots(24m s-•). Theaveragemean3- 6-kmwindspeed
for relatively weak wind cases (0- 6-km mean wind speed
Units are m s-t . It is importantto notethat although
the -<30knots(15 m s-t)) is 34 knots(17.5 m s-•). The lowest
BRN correlates reasonably well with observed storm type
mean 3- 6-km wind speed encountered in the data set is 20
(e.g., supercell versus multicell), it is a poor predictor of
storm rotation in low levels because it does not account for
knots(11m s-•), whichagreeswith Doswell'sobservation
that middle-level winds should generally be 20 knots (11 m
someaspectsof the wind profile, suchas low-level curvature
s-•) or greaterregarding
hodographs
thatindicatepotential
shear [Lazarus and Droegemeier, 1990].
for supercelldevelopment.
Figure 2 shows the distribution of U for the JDL data set.
The majority of cases (70%) are associatedwith U values
9. CASE STUDIES
greater than 18, with one case exceeding 40. None of the
caseshas a U value less than 12. This suggeststhat U values Horizontal fields generatedfrom soundingdata for param-
muchlessthan 12 m s-• may not be able to support eters selectedfor this study will be examined briefly here for
supercell-inducedtornadoes. two casesfrom the 31-case storm motion subset, one involv-
ing strong wind fields and one involving less intense wind
8. MIDDLE-LEVEL WIND SPEED RESULTS
fields. These serve as examples of how wind parameters
Doswell [1991] specifiesmiddle-levelwinds as those in might come together diagnostically to indicate potential for
roughly the 700- to 500-mbar layer. For consistencywhen tornadic supercell development. However, the forecaster
dealingwith surfaceelevationsrangingfrom the high plains needsto remember that atmospheric wind profiles are chang-
to ocean coastal areas, this study considersthe mean mea- ing constantly. The reader is referred to Doswell [1991] for a
suredwind speedin the 3- 6-kin AGL layerasrepresentative discussionconcerningthe problems of forecasting changes
of middle levels. to the hodograph/wind
578 PARAMETERSASSOCIATEDWITH STRONGAND VIOLENT TORNADOES,1

9.1. Raleigh, North Carolina, November 28, 1988

Davies[1989]examinedthiscasefrom the perspective of
positivemeanshear,contrastingthe shearin severaldiffer-
ent low-levellayers.For comparison,
the samecasewill be
examinedhere usingthe wind parametersselectedfor this
study.
Using 0000 UTC soundingdata (6 hoursprior to the
Raleightornado),hellcity fields were generatedusing a
deviantstormmotionassumption of 20R85(,the0- 6-kmwind
fieldswerequitestrong).Thesefieldswerecomputed for the
0- 2-km, 0- 3-km, and 0- 4-km AGL layers(only the 0- 2-km
and0-4-km fieldsare shown;seeFigures3a and3b). While
the 0- 2-km field appearsmorefocusedthan patternsfrom
deeperlayers,all threeanalysesplacea maximumoverthe
Greensboro-Raleigh area,asdidsimilaranalyses
for positive
meanshearby Davies[ 1989].Tornadoeslateroccurredwith (ass as)
stormsin North Carolinaand southernVirginia wherehelic-
ity valueswereverylarge(600m2 s-2 or greater)
butnot ::::•

with storms that occurred earlier in Georgia. The lack of

tornadoesfurther southwhere helicity magnitudeswere less
butstillsignificant (values
around 300m2 s-2) was probably
due to the fact that instability (not shown) was weak or
limited mainly to coastalareas,whereasweak to moderate
instabilityextendedinlandfurthernorth acrossNorth Caro-
lina. Part 2 of thispaper[Johnset al., thisvolume]examines
combinations of instability and helicity associatedwith
strong and violent tornadoes.
Helicity fields were also computed using the observed
motion of the Raleigh storm (24R86). The resultsare very
similar to the fieldsderived usingan assumedmotion. This
can be seen by comparingthe 0- 2-kin fields in Figures 3a
and 4 (the 0- 3-kin and 0- 4-km observed helicity fields are
not shown). In this case the use of an assumedstorm motion
of 20R85 relative to the 0- 6-kin mean wind worked quite well
in offering a useful diagnosticdepiction of the helicity.
Bulk Richardson Number shear U (not shown) was strong
with a maximum of more than 28 m s -• over central North
Carolina. Middle-level winds (not shown) were also strong,
ranging
from40tomorethan80knots(20-40ms-i ) overthe Fig. 3. Horizontal distribution of storm-relative helicity subjec-
eastern United States. Consideringjust the wind criteria in tively analyzedfrom 0000 UTC November 28, 1988, soundingdata,
this case (apart from other factors such as instability•, all using an assumed storm motion of 20R85. Helicity is integrated
through(a) 0-2 km AGL and (b) 0-4 km AGL. Units of contours are
parameters are more than adequatefor supercell de,•elop- m2 s-2 Heavylinesrepresent
tracksofsignificant
tornadoes
which
ment, most over a large area. It is the helicity pattern that is occurred 6-8 hours after 0000 UTC.
most helpful in narrowing the threat area, particularly the 0-
2-km field. This is typical of tornado casesinvolving strong
wind fields.
applied. Note that even with the coarse resolution of the
rawinsondenetwork, two generalareas of maximum helicity
9.2. Central Kansas/Southeast Wyoming,
are evident, eachroughly correspondingto the locationsof
May 24, 1990
tornado occurrences.The assumedhelicity depictions com-
Figure 5 showsthe 0- 2-km assumedhelicity field (the 0- pare well with analyses generated using observed storm
3-km and 0-4-km fields are not shown)from 0000UTC May motions(the observedhelicity field for 0-2 km is shown in
25, 1990, soundingdata, when significantsupercell-induced Figure 6).
tornadoes were occurring over central Kansas and southeast An analysis of BRN shear U (Figure 7) indicates two
Wyoming. The wind fields are weaker than in the Raleigh maximumareas/ridgesof strongshearthroughmiddlelevels,
case(0- 6-km meanwind -<30 knots(15 m s-1) over a one over Kansasand another extendingfrom western Colo-
majority of the area), so a 30R75 motion assumptionwas radointo the northernhighplains.Middle-level wind
 DAVIES AND JOHNS 579

, ................
• i!
'i'-F'l•k•
H--•:lc'•
(m2s
OOUTC
-•)
5/25t90
I (observed
cellmotion
31R94)

..

Fig. 4. Horizontal distribution of storm-relative helicity as in

Figure 3a, except observed motion of Raleigh tornadic storm Fig. 6. Horizontal distributionof storm-relativehelicity as in
(24R86) is used. Tornado tracks are as in Figure 3. Figure 5, except observedmotion of tornadic storm in central
Kansas(3IR94) is used.Tornado occurrencesare as in Figure 5.

(3- 6-km AGL averages) in Figure 8 reveal two streams of

maximum winds, each correlating well with the U parameter
10. EXTREMES OF HODOGRAPHS ASSOCIATED
shear maxima in Figure 7. These wind maxima are moving
WITH SUPERCELL-INDUCED TORNADOES
downstream across the areas of tornado occurrence, provid-
ing shear through middle levels and additional supportfor Examinationof the largeJDL data set exposedthe authors
supercell thunderstorms. to a wide range of wind profiles that were associatedwith
strongand violent tornadoes.While many hodographswere

....................
• 0-2km
Helicity
(m2s
'•)
'" 'fi' • (assumed
cellmotion
30R75)
..............
./ I ! U(ms") II
a 00UTC
5/•5/90
,. / I BulkRichardson
Number
shearII

t[ / I i:
..............

18
II•1
!",........
." '!i:..
' i '"':'"':i

Fig. 5. Horizontaldistribution
of storm-relative
helicityintegrated
through0-2 km AGL, subjectivelyanalyzedfrom 0000UTC May
25, 1990,rawinsondenetworkdata usingan assumedstormmotion Fig. 7. Horizontal distribution of U (Bulk Richardson Number
of 30R75.Unitsof contoursarem2 s-2. Squaredotsandheavylines shear) subjectively analyzed from 0000 UTC May 25, 1990, rawin-
represent approximatelocations
of tornadoesoccurring within2-3 sonde network data. Units of contours are meters per second.
hours of 0000 UTC. Tornado occurrences are as in Figure
580 PARAMETERS ASSOCIATED WITH STRONG AND VIOLENT TORNADOES, 1

nificant or optimum parameter values as suggestedin the

literaturefrom theoreticaldeduction,numericalmodeling, or
.........
,.,•.. ] 3-6km
mean
wind
speed
(kts)
\ I (middle
levelmeasured
winds) storm observationsappear to be in reasonable agreement
-- = I OOUTC
5/25/90 with the results from the large data set assembledby the
authors. These wind parameters address the contributions of
wind fields to several aspects of supercell and tornado
development (e.g., updraft rotation, shear-induced vertical
pressuregradient, and storm inflow). However, there are
many other parametersapart from the wind profile that are
important for development of supercelIs and supercell-
induced tornadoes. These include thermodynamic factors
40
such as atmosphericinstability and the potential buoyant
energy available to an updraft (discussedin part 2 of this
paper), as well as the humidity of middle-level air entrained
into downdrafts. A forecaster also must address issues such
as dynamic forcing and capping inversions to determine
whether thunderstormswill develop in the first place. Be-
.........................
i/'" / / :•o'i
cause of these factors one cannot produce a forecast con-
cerning supercell-induced tornadoes from examination of
wind parametersalone. However, it does appear that suffi-
Fig. 8. Horizontaldistributionof mean3- 6-km AGL wind speed cient critical or "optimum" values of the wind parameters
subjectivelyanalyzedfrom 0000 UTC May 25, 1990, rawinsonde mustbe presentfor thunderstorms to developinto supercelIs
network data. Units of contoursare knots. Heavy dashedlinesare that produce tornadoes.
axesof maximumwindspeed.Tornadooccurrences areasin Figure While statisticalcorrelationsof the selectedwind param-
5.
eters to tornado occurrenceand likelihood are beyond the
scope of this study, some subjective comments can be
similar in scopeand shape,the extremesof hodographs offered regardingtheir relative usefulnessin forecasting.
encounteredare significantand are worth discussing briefly.
Figure 9 showstwo low-level hodographsfrom the JDL
data set that are quite differentfrom each other in size and
scope. Both hodographswere associatedwith tornadoesof
F4 intensity. While not obviousin a ground-relativeframe,
the directionalturning for both hodographsis nearly the O-3km
Hodograph
t...i
.........
'.....
samewhenshiftedto a storm-relative frame(roughly100øfor
Figure 9a, and 90øfor Figure 9b). The mostnotablecontrast
betweenthe two casesis in the magnitudes
versus140m2 s-2, basedonobserved
of helicity(745
stormmotions)
and ......
/.....
:.....
:....
....!....•io..;-..•o"4O-.-;-:':'•a....:'::,:;:::•:
.........
:::;;::::i'"!i.
•.....................
i.........
i.....
i.....:•.....i.....i.....
buoyancy (convectiveavailablepotential energy, CAPE)
: ' : :, ..:.
.......
::.....
:: .....
"'..:•,':"::...
(.......
.-'
......
!. .: :' :i ::.
(300versus4500J kg-l). In theBNA example
theextreme
wind shearandhelicityappearto havebeenresponsible for ...!;
........
i}
....
a..,,!•;
......
..... Observed
:,:...:-.:,..-
CAPE helicity=745
=300j kg-' m2s -2'? ....
?i.......
.'........
:..-
L'"?
;;
-"' '.:1
producingupdraftrotation,in spiteof meagerinstability.At ...... : "' .......... I.••l• fi.. ' . ..l'l • .• I;'.....
i• i I
the oppositeendof the spectrum the extremeinstabilityin
the OMA example may have somehowplayed a role in oo,o '::'
.....
enhancing the rotational potential of a wind environment
exhibitingweakhelicityyet significant
directional
turningin
0-3km
Hodograph
t.-!.
.....
;:.....
::"'"';.':;:
.....
?':...
low levels.While undetectedlocalchangesfavoringrotation
[see Doswell, 1991] also could have been involved in the
/......
/..-::.
..... .....
::::::>'::.:
....
latter case, the fact remainsthat thesewidelydiffering
hodographs were associated with violent tornadoes. These :.....
..... .....
::::).:
.......
two casesillustratethat combinationsof helicityand insta-
.......
i.... .....
..........
!..........
i........
!........
i.......
.
bility are an importantconsideration
in assessing
thepoten- .....
:•:
......
:!'
....
:!'
.....
:ob•"erved'
hel'ici•Y
'L'i4'0
"r•2•
tialfor supercell-induced
tornadoes(seepart2 ofthispaper). b :.'•:
.....
L:
...... :.-
_.';¾','"':.:,
", CAPE-
.......4500 j kg-1
.-......... ß ....
.•....
........
..:' .r .......'
1 I. DISCUSSION Fig. 9. Ground-relative
low-levelhodographs (0-3 km AGL) for
The wind parametersexaminedin this studyappearto (a) Nashville,1200UTC December 24, 1988,and(b) Omaha0000
UTC July29, 1986.Eachringincrement represents10knots(5
relatewell to supercell-induced
tornadodevelopment.
Sig-
 DAVIES AND JOHNS 581

Helicity is probablythe mostcrucialwind parameterfor sections5 and 6 and in Figure 1 is one simple way of
indicatingtornadicsupercellpotential.The numericalsimu- addressingthis.
lations of Brooks and Wilhelrason[1990] tend to reinforce The depthof the layerthat is usedto computelow-level
this impression. Their results show that stormswith similar helicity dependslargely on what constitutesthe primary
BRN valuesbut difefing shearcurvatureprofilesin thelow inflow layer for a given thunderstorm.This is becausethe
levels often develop differingcirculations.Their storm sim- streamwisevorticity [Davies-Jones, 1984] associatedwith
ulationswith low-levelcurvatureshear(significant low-level helicityin the inflowlayer is tilted into an updraftto produce
hellcity)developedlow-levelmesocyclones typicalof ob- rotation.The definitionof inflow layer is probablyrelatedto
servedtornadicsupercelis, whilesimulations involvingrec- the level of free convection (LFC) in a thunderstorm's near
tilinearlow-levelshear(weakor negligible low-levelhelicity) environment,amongotherfactors. Using a mobile sounding
did not producelow-level mesocyclones. Both types of system,Bluesteinet al. [1989]foundthe lapserate withinthe
simulations hadsimilarvaluesof U (vertical"straight-vector updraftandwall cloudof a tornadicthunderstormto be "wet
difference"shearthroughmiddlelevels),andbothproduced adiabatic" above the LFC, which was measuredbetween 1.5
supercell storms with identifiablemiddle-levelmesocy- and 2 km AGL. Because the LFC varies considerably
clones.AlthoughBrookset al. [this volume]did encounter between "wet" environments and somewhat "drier" envi-
somenumericallymodelledstormsin high-helicityenviron- ronments(e.g., alongthe dry line), the depth of the relevant
ments that failed to produce low-level mesocyclones,it inflow layerfor computinghelicity probablyalso variesfrom
nevertheless appears that low-level curvature shear and caseto case.This suggests usingthe layer below the LFC for
helicity are crucial factorsfor the developmentof mesocy- computationof helicity, which may be a useful subjectfor
clones at low levels. future study.
Positive mean shearcorrelateswell with helicity in most Operationally, different studies have used different layers
casesinvolving moderateto strongwind fieldsand is a useful for computing helicity. Davies-Jones et al. [1990] use 0-
alternatecalculationin suchsituations.As the positivemean 3-km AGL for the generalinflow layer, while Woodall [ 1990]
shearbecomesgreater,the inferenceis that a largerrangeof computeshelicity for a deeper layer (0- 4-km AGL). From
stormmotionscan supporttornadicsupercelldevelopment. this study, the vertical distribution of helicity (section 6) and
However, when wind fields are relatively weak (yet ade- the case studies (section 9) suggest that useful results can
quate to support supercell development), helicity is the often be obtained usingthe 0- 2-km AGL layer. On the basis
preferableparameterbecauseit can provide a usefuldepic- of the examination of data set and recent operational expe-
tion of the enhancement of rotational potential due to a hence the following general comments are offered.
storm's local motion. 1. Helicity derived from the 0- 2-km layer often tends to
Becausedeeper vertical shear(through middle levels)is an focus the horizontal pattern into a more useful area diagnos-
important ingredient for supporting supercelIs, the BRN tically (see Figure 3), while helicity patterns from deeper
shear and the strength of the middle-level winds also have layers tend to spreadout. This is particularly true in situa-
importance in a forecast setting. Although most casesexhib- tions involving strong wind fields. In weak wind fields,
iting significantlow-level helicity also will be associatedwith helicity patterns derived from different layers (particularly
significant wind fields through a deeper layer (due to the 0-2 km and 0-3 km) often are nearly identical.
baroclinity and dynamics associated with weather systems), 2. In high plains environments that tend to be "drier,"
it is possiblethat significantlow-level helicity can be present resulting in higher LFCs, the 0- 3-km or 0- 4-km layer works
without deeper shear through middle levels when weaker better in capturing the inflow layer and relevant areas of
weather disturbancesare involved. Hence it is important to helicity.
examine the middle-level wind fields in addition to the 3. In hurricane/tropical cyclone tornado environments
low-level helicity. (which are associatedwith low LFCs), the largest low-level
Returningto storm-relativehelicity as a crucialparameter, shear and helicity is often in the bottom 1-1.5 km [McCaul,
key questions are (1) how to estimate in advance of storm 1991], suggestingthat a relatively shallow layer be used for
developmenta reasonablestorm motion for the more right- helicity computation.
moving storms that may occur and (2) what layer is most The merits of one layer versus another for computing
appropriatefor computinglow-level helicity. helicity are not dealt with here' the important point is that
Regarding storm motion estimation, section 5 and recent the inflow layer is not fixed from situation to situation. This
operationalexperiencesuggestthat applyingone exclusive suggeststhat operational helicity programs might be im-
"preset" storm deviation in all situationsis inappropriatefor proved somewhatby including output information about the
estimatinghellcity operationally.From examinationof the vertical distributionof helicity, as well as offering computa-
31-case storm motion subset the idea that supercelIsin tion options involving different layers.
weaker wind environments tend to deviate more from the As noted earlier, the potential for tornadic supercell
mean wind has some validity. Therefore it would seem developmentdoes not depend on wind parameters alone.
appropriatefor a forecasterto be able to "phase in" larger Part 2 of this paper [Johns et al., this volume] will examine
deviations as wind fields become weaker for purposesof combinationsof instability and shear/helicity that are asso-
computinghelicity. The 20R85/30R75schemesuggested
in ciated with strong and violent supercell-induced
582 PARAMETERS ASSOCIATED WITH STRONG AND VIOLENT TORNADOES, 1

Acknowledgments. The authorswish to thank Grant Bean, Dave the relationshipof 0-2 km AGL "positive" wind shear to poten-
Higginbotham,and other NSSFC computer staff in helpingto tial bouyant energy in strong and violent tornado situations, in
assemblethe largedata set usedfor this study.Ken Howard, NSSL, Preprints, 16th Conferenceon SevereLocal Storms, pp. 593-598,
ERL; Mike Ryba, WSO DDC; and Joe Schaefer, SSD, Central American Meteorological Society, Boston, Mass., 1990.
RegionNWS, arealsoacknowledged for theirhelpwiththisproject. Johns, R. H., J. M. Davies, and P. W. Leftwich, Some wind and
instability parametersassociatedwith strong and violent torna-
REFERENCES does, 2, Variations in the combinationsof wind and instability
parameters,this volume.
Bluestein,H. B., and C. R. Parks, A synopticand photographic Lazurus, S. M., and K. K. Droegemeier,The influenceof helicity on
climatology of low-precipitation severe thunderstormsin the the stability and morphology of numerically simulated storms, in
southernplains, Mon. Weather Rex,., 111, 2034-2046, 1983. Preprints,I6th Conferenceon SevereLocal Storms, pp. 269-274,
Bluestein,H. B., E. W. McCaul, Jr., G. P. Byrd, G. R. Woodall,G. American Meteorological Society, Boston, Mass., 1990.
Martin, S. Keighton, and L. C. Showell, Mobile soundingobser- Lilly, D. K., The structure, energetics,and propagationof rotating
vations of a tornadicthunderstormnear the dryline: The Gruver, convective storms, II, Helicity and storm stabilization, J. Atmos.
Texas storm complex of 25 May, 1987, Mon. WeatherRev., 117, Sci., 43, 126-140, 1986.
244-250, 1989. Maddox, R. A., An evaluation of tornado proximity wind and
Brooks, H. E., and R. B. Wilhelmson, The effects of low-level stability data, Mort. Weather Rev., 104, 133-142, 1976.
hodographcurvature on supercellstructure, in Preprints, I6th McCaul, E. W., Jr., Simulations of convective storms in hurricane
Conferenceon SevereLocal Storms, pp. 34-39, AmericanMete- environments, in Preprints, 16th Conference on Severe Local
orological Society, Boston, Mass., 1990. Storms, pp. 334-339, American Meteorological Society, Boston,
Brooks, H. E., C. A. Doswell III, and R. P. Davies-Jones,Envi- Mass., 1990.
ronmentalhelicity and the maintenanceand evolutionof low-level McCaul, E. W., Jr., Buoyancy and shear characteristics of hurri-
mesocyclones,this volume. cane-tornadoenvironments, Mon. Weather Rev., 119, 1954-1978,
Darkow, G. L., and D. W. McCann, Relative environmental winds 1991.
for 121tornadobearingstorms,in Preprints, 11th Conferenceon Rasmussen,E. N., and R. B. Wilhelmson, Relationshipsbetween
Severe Local Storms, pp. 413-417, American Meteorological stormcharacteristics and 1200GMT hodographs,low-level shear,
Society, Boston, Mass., 1977. and stability, in Preprints, I3th Conference on Severe Local
Davies, J. M., On the use of shearmagnitudesand hodographsin Storms, pp. J5-J8, American Meteorological Society, Boston,
tornado forecasting,in Preprints, 12th Conferenceon Weather Mass., 1983.
Forecastingand Analysis, pp. 219--224,AmericanMeteorological Rotunno, R., and J. B. Klemp, The influence of the shear-induced
Society, Boston, Mass., 1989. vertical pressuregradient on thunderstorm motion, Mon. Weather
Davies-Jones,R. P., Streamwise vorticity: The origin of updraft Rev., 110, 136-151, 1982.
rotation in supercellstorms, J. Atmos. Sci., 41, 2991-3006, 1984. Wakimoto, R. M., and J. W. Wilson, Non-supercell tornadoes,
Davies-Jones,R. P., D. W. Burgess,and M. Foster, Test of helicity Mon. Weather Rev., 117, 1113-1140, 1989.
as a tornadoforecastparameter,in Preprints, 16thConferenceon Weisman,M. L., and J. B. Klemp, The dependenceof numerically
Severe Local Storms, pp. 588-592, American Meteorological simulated convective storms on vertical wind shear and buoy-
Society, Boston, Mass., 1990. ancy, Mort. Weather Rev., 110, 504-520, 1982.
Doswell, C. A., III, A review for forecasterson the applicationof Weisman, M. L., and J. B. Klemp, Characteristicsof isolated
hodographsto forecasting severe thunderstorms,Natl. Weather convectivestorms,in Mesoscale Meteorology and Forecasting,
Dig., 16, 2-16, 1991. edited by P.S. Ray, pp. 331-358, American Meteorological
Doswell, C. A., III, A. R. Moller, andR. W. Przybylinski,A unified Society, Boston, Mass., 1986.
setof conceptualmodelsfor variationson the supercelltheme,in Woodall, G. R., Qualitative forecastingof tornadic activity using
Preprints,16th Conferenceon SevereLocal Stortns,pp. 40-45, storm-relativeenvironmentalhelicity, in Preprints, 16th Confer-
American MeteorologicalSociety, Boston, Mass., 1990. ence on SevereLocal Storms, pp. 311-315, American Meteoro-
Johns, R. H., J. M. Davies, and P. W. Leftwich, An examinationof logical Society, Boston, Mass.,
 SomeWind and InstabilityParametersAssociated
With Strong and Violent Tornadoes
e Variations in the Combinationsof Wind and Instability Parameters
ROBERTH. JOHNS,JONATHAN
M. DAVIES,l ANDPRESTON
W. LEFTWICH

National WeatherService,NOAA, NationalSevereStormsForecastCenter,KansasCity, Missouri64106

1. INTRODUCTION spheric mean wind shear (U) [see Weisman and Klemp,
1982]are the three wind parameters with which the relation-
Meteorologistshave longknownthat bothpotentialbuoy- ship with instability is examined. The wind parameters are
ant energy and the strength and vertical profile of the defined and the method of calculation is discussed in part 1
troposphericwind fields are important in the processof [Davies and Johns, this volume]. The primary instability
tornado development[e.g., Miller, 1972]. Further, opera- parameter examined in this study is CAPE. Surface parcel
tional experience suggeststhat the combinations of wind lifted index (SPLI) values based on lifted surface parcel
parameters and instability parameters vary considerably temperatures [Hales and Doswell, 1982; Bothwell, 1988] are
from one tornado situation to another. Numerical models also examined since the SPLI is widely utilized operation-
suggestthat the type of storm that develops in a given ally.
situation (e.g., an isolated supercell) is related to both the
vertical wind profile and the potentialbuoyantenergyof the 2. DETERMINATION OF BUOYANT ENERGY PARAMETER
air in the updraft entrainmentlayer [Weismanand Klemp, VALUES
1982, 1986]. Observational studies by Rasmussen and
Wilhelmson [ 1983] and Leftwich and Wu [1988] have exam- 2.1. Calculation of CAPE
inedthe wind shear/potentialbuoyant energyrelationshipin Values of potential buoyant energy for soundingsutilized
association with tornado development. These studies in- in this study have been determined by an algorithm de-
volved limited data sets and were concerned with the wind scribed by Doswell et al. [1982]. The values are essentially
shear in a relatively deep layer of the troposphere(0- 4-km equivalentto CAPE with the lifted layer being the lowest !00
above ground level (AGL)). Recently, interest has focused mbar AGL. To make representative estimates of CAPE in
on the nature of the wind fields in shallower layers of the the vicinity of events consideredfor inclusion in the data set,
lower troposphere,layers that more nearly correspondto the these guidelineswere followed:
updraft entrainment region [Davies, 1989; Davies-Jones et 1. If a tornadocaseoccurredwithin1•_hoursof the
al., 1990]. Utilizing a large comprehensivedata set, Johnset sounding time, within 40 nautical miles (74 kin) of the
al. [1990] (hereafter referred to as JDL) examined the soundingsite, and in the same air mass as the soundingsite,
relationshipbetween 0- 2-km AGL "positive" wind shear the CAPE value from the soundingwas used directly.
(PWS) and convective available potential energy (CAPE) 2. For tornado cases not satisfying the conditions in
[Moncrieff and Green, 1972]in associationwith strongand guideline 1, the thermodynamic profiles of the surrounding
violent tornado development. In this paper (part 2) the soundingswere examined. Interpolation of a CAPE value for
authors review the initial work of JDL and examine addi- both the time and location of the tornado event from the
tionalwind andpotentialbuoyantenergyparameterrelation- CAPE values computed for the regularly scheduled sound-
shipsassociatedwith the data set compiled by JDL. ings was attempted only if (1) the horizontal temperature
Positive wind shear (PWS), helicity, and deep tropo- gradientsat the standard levels above the boundary layer
appeared to be relatively weak and (2) surface data and the
1Alsoat Davies,Incorporated,
Pratt,Kansas
67124. temperatureand moisturestratificationin the boundaryla-yer
The Tornado:Its Structure,Dynamics,Prediction,and Hazards. of the soundingsindicated that any instability maximum
GeophysicalMonograph79 (center or axis) associatedwith the case was adequately
Copyright1993by the AmericanGeophysicalUnion. sampled by at least two nearby stations.
584 PARAMETERS ASSOCIATED WITH STRONG AND VIOLENT TORNADOES, 2

CONVECTIVE AVAILABLE POTENTIAL ENERGY placeof eventoccurrence

was obtainedby interpolation
242 TORNADO CASES from radiosonde constant level data.
3. By usinga skew-Tdiagramthe surfacevaluesfor each
NO. OF CASES
event were combined to produce a measurement of the
pseudo-adiabat
whichthe surfaceparcelwouldfollow if
lifted to saturation. The difference between the temperature
at which this adiabat crosses 500 mbar and the environmen-
tal 500-mbartemperatureis the SPLI value.

3. RESULTS

3.1. Potential Buoyant Energy Distribution

Figure 1 illustratesthat the strongand violent tornadoes
are associated with an extremely wide range of CAPE
VALUES IN HUNDREDS
values. Values in the JDL data set range from 200 to 5300 J
Fig. 1. Distribution of the convective available potential energy
kg-] with abouttwo thirds(64%) of the casesexhibiting
(CAPE) values associatedwith the 242 strong and violent tornado valuesbetween1000and3000J kg-] . The datasetdisplays
cases in the data set. Number above each bar is the number of cases seasonaland geographicaldifferencesin the values of CAPE
in that particular range of CAPE values. associated with strong and violent tornado development.
The cold season (November 1 to March 31) cases occur
mostly in the eastern portions of the southernPlains and
Gulf coastalregion(see JDL) and exhibit CAPE values that
3. For most tornado casesnot satisfyingguideline 1 or 2,
are mostly weak to moderate (Figure 2). Ninety-five percent
proximity soundingswere constructed by using proximity
surfacedata representativeof the inflow air and interpolating
(for time and/or location) from the surrounding soundings
both the boundary layer thermodynamic profiles and the SHEAR
standard level data above the boundary layer. The CAPE 26
values for these tornado cases were derived from the con-
structed soundings. Soundings were constructed for three 24

fourths (181 cases) of the cases utilized in the data set.

4. In a few instances, missing data or ditficulties in 22

interpolationresulted in potential casesbeing excludedfrom

2O
the data set.
The procedurefor determiningthe CAPE for a casediffers 18
from that utilized to determine the mean shear and helicity
values (see part 1). One of the reasonsfor the differenceis 16

that in determiningCAPE, it was often necessaryto consider

the temperature profile through the depth of the entire 14

troposphere.Also, the thermodynamicpatterns(particularly

12
for moisture) can be quite complicated in the boundary
layer. These complicationsrequired that proximity sound- 10
ings be constructed in most cases in order to arrive at a
representativeCAPE value. 8

6
2.2. Surface Parcel Lifted Index (SPLI) Calculations
An estimate of the instability may be obtainedby comput- 4

ing SPLI values.For all of the tornadocasesin which CAPE 1000 2000 3000 4000 5000
was successfullycomputed, SPLI values were also com-
CAPE
puted. The following method was employedto obtain SPLI
values. Fig. 2. Scatterdiagram showingthe relationshipbetween convec-
1. Surfacetemperature, dew point, and pressurevalues tive availablepotentialenergy(CAPE) in joules per kilogramand 0-
near the time and place of the event and representativeof the 2-kmAGL positivewindshear(x 10-3 s-1) for the75 coldseason
(November 1 to March 31) tornado cases. Solid curved line is a
air mass from which storm inflow was occurring were suggestedlower limit of combined CAPE/low-level shear value that
obtained. would supportthe developmentof strong or violent mesocyclone-
2. A 500-mbar temperature for the time and over the induced tornadoes[after Johns et al.,
 JOHNS ET AL. 585

of the cold seasoncasesare associatedwith CAPE values of HELlCITY

lessthan2500J kg-1.
Casesduringthe warm season(May 15 to August31)
generallyoccur farther west and north than the cold season
cases(seeJDL) and exhibita wide rangeof CAPE values,
froma weak500J kg-1 to a strong
5300J kg-l (Figure
3).
lOI)
900 ,,i I .......
The two cases exhibiting CAPE values of less than 1000 J
kg-• wereassociated
withtropicalcyclones.
ThelowCAPE 800 .... .
values in these cases agree with the findingsof McCaul 70( .
[199!]. The two cases associatedwith warm seasonderechos
[Johnsand Hirt, 1987]exhibitcharacteristically
highCAPE
: ß
values. Despite the wide rangeof values,note that over two ß = ß

thirds (68%) of the warm season cases are associatedwith 5oo- ' -. ." ß
.
CAPEvalues
of2500J kg-l orgreater.
Thisagrees
withthe
4oo• • i'"-"-"•'' . ß '--'-•
e

findingsof Rasmussenand Wilhelmson[1983].

Surfaceparcellifted index(SPLI) valuesalsosuggest
that
strong and violent tornadoes are associated with a wide
300.......
rangeof instability.Values in the JDL data set rangefrom 200 i : , ,A ß
- 1 to - 14 (Figure 4). However, a largemajorityof the cases e

(72%) are associated with values from -5 to -10. 100 .......

e

o
3.2. Bulk Richardson Numbers
-1 -2 -3 -4 -5 -6 -7 -8 -9 -lO -11 -12 -13 -14
Weisman and Kle/np [1982, 1986] have shown that the
SPL!
type of storm that developsin a given environmentis at least
partially related to a Bulk Richardson Number (BRN) de- Fig. 4. Scatter diagram showing the relationship between surface
fined as
parcel lifted index (SPLI) in degrees Celsius and 0- 2-km AGL
helicityin m2 s -2 utilizing the 20R85/30R75 storm motion assump-
tion method for all 242 cases in the data set.
SHEAR
I' I 1

24
o-TROPICAL
"
CYCLONE
TORNADO
i
WARM SEASON DERECHO TORNADO
BRN = •,
}u 2
B

22 "

20
where B is the buoyant energy (CAPE) for a lifted parcel in
e the storm's environment, and U is a measure of the vertical
18I wind shear through a relatively deep layer (0- 6-km AGL).
Results from numerical modeling experiments and a limited
16 ..........
e ee ß
number of storm observationshave suggestedthat growth of
supercelIsis confined to values of BRN between 10 and 40
ß
ß ß

14 " ' ß ' '.........'

o ß oe ß ß [Weisman and Klemp, 1986].
Figure 5 illustrates the range of BRN values associated
with the strong and violent tornadoes in the JDL data set.
Almost one half (47%) of the BRN values are less than 8. The
10 ß •..' . ß .--•-- =• number of cases diminishes as the B RN values become
larger, with most casesassociated with values of less than 40
I (94%). Theseresultsagree well with the findingsof Riley and
Colquhoun [1990], who also examined a large number of
cases.

The questionarisesas to why there are so many caseswith

' 1000 2000 3000 ' ' 4000 ' 5000
very low BRN values. The prevailing theory has been that in
CAPE such an environment the very strong shear would inhibit the
growth of the updraft, thus not allowing for a deep rotating
Fig. 3. As in Figure 2 exceptfor the 69 warm season(May 15 to convectivecolumn to develop and be sustained. One poten-
August31) tornadocases.The open circlesrepresenttornadocases
tial explanationfor this apparent contradiction may lie with
associatedwith tropical cyclones. The open triangles represent
tornado cases associated with warm season derechos [after Johns et the nature of the BRN. As pointed out by Weisman and
al., 1990]. Klemp [1982] and Lazarus and Droegemeier [1990],
586 PARAMETERS ASSOCIATED WITH STRONG AND VIOLENT TORNADOES, 2

RANGE OF BULK RICHARDSON NUMBERS supercellcirculations.Examination of the radar imagery

(low-elevationreflectivity data) associatedwith the 31 strong
NR OF CASES and violent tornadocasesin the data subsetdescribedin part
140
1 supportsthe hypothesisof Doswell et al. Only 10 cases
(32%) appearto be associatedwith •'classic" isolated super-
cells, while the large majority of cases (68%) appear to be
120 113
associatedwith a variety of multicellular systemsincluding
lines, spiral bands, clusters, and bow echoes. These results
suggestthat in strongand violent tornado situations,super-
100
-'i•o
cell circulationsassociatedwith complex multicellular con-
vective structures are quite common.
Recall that the B RN is related to the type of storm
structure that develops in a given situation [Weisman and
•o Klemp, 1986].The 10-40 range of BRN values that Weisman
and Klemp have associated with supercell development
applies to isolated convection (i.e., the "classic" isolated
<o
supercell).While an isolated supercell is not likely to de-
velop in an environment with a very low BRN, the larger
2o multicellular convective structuresdescribed in the previous
paragraphcan be initiated and sustainedin suchan environ-
ment by synoptic scale forcing (e.g., a squall line ahead of a
cold front). Once developed, irregularities and differential
0-7 8-19 20-29 30-39 40-49 50-59 60-69
movementsin the storm outflow patterns [seeDoswell et al.,
BRN VALUES
1990, Figure 4] may allow some updrafts to encounter
Fig. 5. Distributionof Bulk Richardsonnumbers[Weismanand enhanced inflow and develop rotation. Once rotation is
Klemp, 1982] for all 242 strong and violent tornado cases. established, the effect of shear-induced pressure forces
(discussedearlier in this section) can help to sustain this
rotation.
BRN is a "bulk" measure of the ambient shear and does not
accountfor detailed aspectsof the wind profile, particularly 3.3. CAPE and O- 2-kin AGL Positive Wind Shear
low-level curvature. Modeling experiments [Weisman and
Relationship
Klemp, 1984;Brooks and Wilhelmson, 1990; McCaul, 1990]
indicate that low-level curvature shear and storm-relative While the BRN calculations involve the wind shear
helicity interacting with the deeper troposphericshear can through a deep layer (0- 6-km AGL), recent interest has
enhance strongly the intensity of an updraft. McCaul's focused on the importance of both the shear and nature of
simulationsof storms in the low-buoyancy environmentsof the hodographin the updraft inflow layer [e.g., Davies, 1989;
tropical cyclonesindicatethat shear-inducedpressureforces Davies-Jones eta!., 1990; JDL]. JDL calculated 0- 2-km
[Rotunno and Klemp, 1982], related at least in part to the AGL positive wind shear for the purpose of indirectly
low-level curvature shear, can be up to 3 times as important estimatingthe rotational potential of the environmentalwind
as buoyancy in controlling updraft strength. Thereœoreit field (i.e., the strengthof the 0- 2-km AGL positive shear is
appears likely that in many situations where the BRN is a related to the range of storm motions that would support
very low value, the low-level curvature shearplays a crucial supercell development).
role in helpingto sustaina rotating convective column. To investigate the relationship of the low-level environ-
Althoughthe above simulationssuggesthow a supercell mental shear to potential buoyant energy in strong and
might be maintained in a low BRN environment, there is still violent tornado situations, JDL constructed scatter dia-
the questionof initial development,or how an updraft can grams. From Figures 2, 3, and 6 it is evident that there is a
sustainitself in its earliest stages.Operationalforecasting pattern to the combinations of 0- 2-km AGL PWS and CAPE
experiencesuggestsa possibleexplanationfor this question. associatedwith strongand violent tornadoes(Figure 6) and
The authors have observed cases where tornado-bearing that there are seasonalvariationsin this pattern (Figures 2
supercelIsare associatedwith (or evolve from) complex and 3). For example,the cold seasoncases(Figure 2), which
convective structures, sometimeson a scale much larger occur primarily from the easternportion of the southern
than the supercellitself. Othershave observedthis type of Plainsinto the Gulf coastalregion, are generallyassociated
storm structure (or evolution) also [e.g., Nolen, 1959;Bur- with the combinationof strong0- 2-km AGL PWS values
gess and Curran, 1985; Jldy, 1990; Moller et al., 1990; and weak to moderate CAPE values. Therefore the data
Przyby!inskiet al., 1990]. Further, Doswell et al. [1990] have pointsfor cold seasoncasesare concentratedin the upper
proposedthat the "classic" isolated supercellis just one left portionof the scatterdiagram.
storm type in a broad range of storm types associatedwith Whenconsidering all cases(Figure6), the pattern of
 JOHNS ET AL. 587

pointssuggeststhat progressively
stronger0- 2-km AGL 0-2 KM
PWS valuesare associated
with strongandviolenttornado HELlCITY ......

developmentas the CAPE decreases.In a few caseswhere

i - ! i ! -

11oo o - VIOLENT- SLOWER THAN MEAN WIND OF >30 KT. -

thePWS wasvery strong,greaterthan 16 x 10-3 s-• CAPE ß - STRONG - SLOWER THAN MEAN WIND OF >,30 KT.
valuesaslowas200-300J kg-• wereobserved.
At theother lOOO
ß
z
- STRONG
- STRONG
-
-
FASTER THAN MEAN WIND OF > 30 KT.
FASTER THAN MEAN WIND OF • 30 KT.
extreme,in situations
wherethe CAPE wasgreaterthan a - STRONG - SLOWER THAN MEAN WIND OF •_ 30 KT,
4000J kg-• , strong
andviolenttornadoes
havebeenasso- 900
.... eI
ciatedwithPWSvalues aslowas6 x 10-3 s-• . Despite
the
80(3
widerangeof valuesthereappearsto beanoptimum range
,,

of 0- 2-kmAGL PWS valuesassociated

with strongand 700
violenttornadodevelopment
for any particularCAPE value.
Further,JDL haveshownthatfor violenttornadoes (F4 and 60O

F5 intensity)shearvaluesare generallyin the highertwo o

thirdsof any givenoptimumrange. 500

•11 o

400
3.4. CAPE and Helicity Relationship A

• o o
Helicity hasbecomean importantparameterfor evaluat- 300

ingthe rotational
potentialof air in the storminflowlayer
[e.g., Davies-Jones et al., 1990]. To calculate storm-relative
200 .....

• • ........ • --
helicity a storm motion vector is needed. On the basisof the
actual storm motions and the associated environmental
100 ............ • .................
conditionsfrom a 31-casedata subset,a methodfor assum- 0

ing the storm motionsfrom precursorconditionshas been ....

1000
•
2000
.....
3000 4000 5000
proposed(see part 1). This method assumesa storm motion
of 20R85 (see part 1 for notation) for caseswhere the 0-6-km CAPE

AGL meanwindspeeds
aregreaterthan30knots(15m s-1) Fig. 7. Scatter diagram showingthe combinationsof convective
and a motion of 30R75 for cases where the 0- 6-km AGL availablepotentialenergy_
(CAPE) injoules per kilogramand 0- 2-km
mean wind speedsare of lesserintensity.Althoughthis AGL helicityin m2 s-2 utilizingobservedstormmotionsfor a
31-casedata subset.Open circles represent violent tornadoes(F4-
F5). All squares(solid and hatched) representcaseswhere the 0-
SHEAR 6-km AGL meanwind speedis equal to or lessthan 30 knots (15 m
s-I). Triangles
andhatched
squares
represent
casesin whichthe
associatedsupercell moves faster than the 0- 6-kin AGL mean wind
speed.

,, e!
22 ß '

ß
method for assumingstorm motions appears to be rather
20 ø
ß
ß ß
ß
"crude," it does representa refinementover applyinga
18 .... ß
singleassumeddeviation,suchas 30R75 [Maddox, 1976].
On the basis of the findings in part 1, the authors have
I t I i
1• ..•. . '- . _ I. . chosento constructandcomparescatterdiagramsdisplaying
ß. ...-.. -.• •.,•-
,,.. , .':- • ß e ß
ß . .
combinationsof CAPE and helicity using the following
criteria:(1) 0- 2-km AGL helicity usingthe observedstorm
12 •'
.....
• "'
='' . .'- . •--
ß'- '
motionsfor the 3l-casedata subset,(2) 0- 3-kmAGL helicity
ß. .,... '• 1[. ß using the observed storm motions for the 31-case data

lO...... ., ' •,F";--- 1.... . : ." '--

subset, (3) 0- 2-km AGL helicity using the 20R85/30R75
storm motion assumptionmethod for the 242-caseJDL data
7• ß '- .'. set, and (4) 0- 3-km AGL helicity using the 20R85/30R75
storm motion assumptionmethod for the 242-caseJDL data
set.

Althoughthe datasampleis not large,the scatterdiagram

4
(Figure 7) depicting the combinations of 0- 2-km AGL
helicity using observed storm motion and CAPE for the 31
, , , ._ . ._ .... •. _ -
1OOO 2000 3000 4000 5000
subsetcasesdisplaysa patternthat is qualitativelysimilarto
CAPE
the pattern for combinations of 0- 2-km AGL PWS and
Fig. 6. As in Figure2 exceptfor all 242casesin the dataset[after CAPEfor the242JDL cases(Figure6). Thereis a tendency
Johnset al., 1990]. for casesexhibitingstrongvaluesof low-levelhelicityto
588 PARAMETERS ASSOCIATED WITH STRONG AND VIOLENT TORNADOES, 2

O-2KM Since a number of researchers have used 0- 3-km AGL as

HELlCITY
the approximateinflow layer, a scatter diagram similar to
' ! I i
1100[
' 'o '•-VIOLENT
-30R75
A _ STRONG - 30R75
Figure 8 has been prepared (not shown) with the only
differencebeingthat helicity has been computedusingthe 0-

1000
/
* o - VIOLENT - 20R85 3-km AGL layer. The scatter diagram patterns for both
e - STRONG - 20R85
assumed inflow layers (0- 2-km and 0- 3-km AGL) are
' ! * - STRONG - 20R85..
900[,
800
TROPICAL CYCLONE
similar.
The four casesassociatedwith tropical cyclones (indicat-
o
ed by crossesin Figure 8) tend to have relatively low values
7OO of both CAPE and helicity, resultingin the points residingin
the lower left portion of the scatter diagram (one case in
600
particular is far below the other points). The fact that
500--
tropicalcyclonetornadocasesnormally are associatedwith
.' ß * ø o low CAPE values is well documented [McCaul, 1991].
However, the reason for the accompanying low helicity
ø --i.......................
•
400--
. .!.....o.•.. - o
valuesin Figure 8 is unclear. It may be related to the vertical
0 e• : 0 I
300 distributionof helicity (helicity density) and the depth of the
' l'. ' '?' "::',
ß I'**'"'*•. "1• .'* ß,," effectiveinflow layer [McCaul, 1991] (also see part 1). Also,
/. ' T.,t ,-,,'ø1
20(
A . .' lo o there is a possibilitythat the motions of tornadic storms in

,",'
......
tropicalcyclonesituationsare estimatedpoorly by the storm
IO0
motion assumptionsutilized to compute helicity.

1000 2000 3000 4000 5000 3.5. SPLI and Helicity

Figure 4 is a scatterdiagramdepictingthe combinationsof

CAPE
SPLI and 0- 2-km AGL helicity utilizing the 20R85/30R75
Fig. 8. Scatter diagram showing the combinations of convective storm motion assumptionsfor all 242 cases in the JDL data
available potentialenergy (CAPE) injoules per kilogramand 0- 2-kin set. The pattern is very similar to that for when CAPE is
AGL hellcityin m2 s-2 utilizingthe 20R85/30R75 stormmotion
assumptionsfor the 242 casesin the JDL data set. All triangles(open used as the instability parameter (Figure 8). The slightly
and solid) represent cases in which the assumed storm motion is greaterscatteringof points in Figure 4 is probably a result of
30R75, while the assumed storm motion for the remainder of the the SPLI being a less precise method of estimating instabil-
casesis 20R85. The open circlesand open trianglesrepresentviolent ity. Nevertheless, Figure 4 appears to have value for the
tornadoes (F4-F5). The crosses represent cases associated with
operationalforecaster who is computing SPLi values on an
tropical cyclones.
hourly basis and does not have access to the finer-
incremented model forecasts of CAPE.

associatedwith environmentscharacterizedby weak buoy- 4. DISCUSSION

ancy, and vice versa. A scatterdiagram(not shown)depict-
ing 0- 3-km AGL helicity for the 31 subsetcasesis very From the JDL data set it has been determined that strong
similar to Figure 7. and violent tornadoesare associatedwith an extremely wide
In Figure 7, note that those casesin the subsetwhich are range of potential buoyant energy (CAPE) values. Most of
associatedwith the weakest0- 6-km AGL meanwinds(equal the caseswith low valuesof CAPE occur duringthe cooler
to or lessthan30knots(15m s-l) anddepicted assolidand monthsof the year and almost always are associatedwith
hatchedsquares)are alsoassociated with very highvaluesof dynamic weather systemsand strong troposphericwind
CAPE. This impliesthat supercell-induced strongand vio- fields (and wind shear). A few others with low CAPE values
lent tornadoesoccurringin a weakwindenvironment usually are associatedwith tropical cyclones.Both of these environ-
are associatedwith very strongto extremeinstability. ments result in very low Bulk Richardson Number (BRN)
When the 20R85/30R75storm motion assumptions for values.As a result,whenall strongand violenttornadoesare
computinghelicity are applied to all 242 casesin the JDL considered, almost half are associated with a BRN of less
data set, the resultantscatterdiagramdepictingcombina- than8. Until recently,the generalperceptionhasbeen that
tions of 0- 2-km AGL helicity and CAPE (Figure8) has a the BRN associated with supercelldevelopmentshouldfit
strongsimilarity to the one for 0- 2-km AGL PWS and CAPE into a certainrange,generallybetween10 and40. However,
(Figure 6). A progressivelyhigherrange of valuesfor both theresultsof thisstudy(andtheoneby RileyandColquhoun
helicity and PWS is associatedwith strong and violent [1990])suggest thatwhilethe BRN hasutility in determining
tornadodevelopmentas the CAPE decreases.However, the whentheinitiationof isolatedsupercelIs is possible,it is not
rate of changeappearsto be more gradualin the case of a gooddiscriminator for supercelldevelopment in general.
helicity and CAPE. Thisis particularlysofor strongshear/lowbuoyancy
 JOHNS ET AL. 589

tions(BRNfrom0 to 8), whichneedto berecognized

more others of the NSSFC computer staff in assemblingthe massivedata
widely as havingpotentialfor supercell-induced
tornadoes. set for this study. The authors also wish to thank Bill Henry,
The importanceof low-levelpositiveshear(PWS)and NWSTC; Ken Howard, NSSL, ERL; Mike Ryba, WSO DDC; and
Joe Schaefer, SSD, Central Region, NWS, for their help with this
helicityin strongandviolenttornadodevelopment
hasbeen project and Deborah Haynes for helping with manuscript prepara-
discussedin part 1. When the combinationsof these wind tions.
parameterswith CAPE are considered,
a definitepattern
emerges.
In caseswherethevaluesof CAPEareveryhigh,
the associated
PWS valuesare usuallyrelativelylow. Gen- REFERENCES

erallyhighervaluesof PWSare associatedwith strongand

Bothwell, P. D., Forecasting convection with the AFOS data
violent tornado development as the CAPE decreases. At
analysis programs (ADAP-version 2.0), NOAA Tech. Memo.
verylow CAPE values,casesare associated
withrelatively NWS SR-122, 91 pp., Natl. Weather Serv. S. Reg., Fort Worth,
high values of PWS. For any particularCAPE value there Tex., 1988.
appearsto be a rangeof PWS valuesthat is optimumfor Brooks, H. E., and R. B. Wilhelmson, The effects of low-level
strongand violent tornado development. hodographcurvature on supercell structure, in Preprints, 16th
Conferenceon Severe Local Storms, pp. 34-39, American Mete-
The relationship
betweenCAPE andhelicityin strongand orologicalSociety, Boston, Mass., 1990.
violent tornado developmentis similar to that for CAPE and Burgess,D. W., and E. B. Curran, The relationshipof stormtype to
PWS. That is, generallyhighervaluesof helicityare associ- environmentin Oklahoma on 26 April 1984, in Preprints, 14th
ated with strong and violent tornadoes as the CAPE de- Conference on Severe Local Storms, pp. 208-211, American
creases. If it is assumed that most tornadoes in the data set Meteorological Society, Boston, Mass., 1985.
Davies, J. M., On the use of shear magnitudesand hodographsin
are associatedwith supercelIs,then it can be statedthat for tornado forecasting, in Preprints, 12th Conference on Weather
any particular CAPE value there appearsto be a rangeof Forecastingand Analysis,pp. 219-224, AmericanMeteorological
hellcity valuesthat is optimumfor inducingstrongrotation Society, Boston, Mass., 1989.
Davies, J. M., and R. H. Johns, Some wind and instability param-
and low-level mesocyclonesin supercellstorms (as sug-
eters associatedwith strongand violent tornadoes, 1, Wind shear
gested by Lazarus and Droegemeier [1990]). and helicity, this volume.
The fact that there is considerable scatter on the PWS/ Davies-Jones,R. P., D. W. Burgess, and M. Foster, Test of helicity
CAPE and helicity/CAPE diagramsemphasizesthat there as a tornado forecast parameter, in Preprints, 16th Conference on
are other factors involved in supercell-inducedtornado Severe Local Storms, pp. 588-592, American Meteorological
Society, Boston, Mass., 1990.
development. One of these factors is the downdraft circula- Doswell, C. A., III, J. T. Schaefer, D. W. McCann, T. W. Schlatter,
tions affectingthe supercell.It generallyis acceptedthat ff a and H. B. Wobus, Thermodynamic analysis procedures at the
supercell is to produce a mesocyclone-associatedtornado, National Severe Storms Forecast Center, in Preprints, 9th Con-
the development of a sufficiently strongrear flank downdraft ference on Weather Forecasting and Analysis, pp. 304-309,
American Meteorological Society, Boston, Mass., 1982.
(RFD) is necessary.The strengthof the rear flank downdraft
Doswell, C. A., III, A. R. Moller, and R. W. Przybylinski, A unified
is dependent on conditions, particularly relative humidity, in set of conceptualmodelsfor variations on the supercell theme, in
its middle level source region. Preprints, 16th Conference on Severe Local Storms, pp. 40-45,
Another factor also involves downdraft outflows, but American Meteorological Society, Boston, Mass., 1990.
usually on a larger scale. The intense downdraft and outflow Hales, J. E., Jr., and C. A. Doswell III, High resolution diagnosisof
instability using hourly surface lifted parcel temperatures, in
that develops with some bow echoes affects storm motion, Preprints, 12th Conference on Severe Local Storms, pp. 172-175,
acceleratingthe storm structure. In somecasesthis resultsin American Meteorological Society, Boston, Mass., 1982.
the bow echo "experiencing" increasedhellcity, which may Johns, R. H., and W. D. Hirt, Derechos: Widespread convectively
induce a supercell that is associated with the bow echo induced windstorms, Weather Forecasting, 2, 32-49, 1987.
structure. Recall that in the 31-case data subset there were Johns, R. H., J. M. Davies, and P. W. Leftwich, An examination of
the relationshipof 0-2 km AGL "positive" wind shear to potential
four cases associated with bow echoes that moved faster
buoyant energy in strong and violent tornado situations, in
than the mean wind speed (from part 1). Note that the Preprints,16th Conferenceon Severe Local Storms, pp. 593-598,
20R85/30R75storm motionassumptionsusedin constructing American Meteorological Society, Boston, Mass., 1990.
the hellcity/CAPE diagram do not take this type of storm July, M. J., Forcing factors in the violent tornado outbreak of May
5, 1989: A study in scale interaction, in Preprints, 16th Confer-
motion into account.
ence on Severe Local Storms, pp. 72-77, AmeNcan Meteorolog-
Thesetwo examplesinvolvesomeeffectsof stormdown- ical Society, Boston, Mass., 1990.
drafts. There are probably many more factors involvedin Lazarus, S. M., and K. K. Droegemeier, The influenceof helicity on
supercell-inducedtornado development(someof which are the stability and morphology of numerically simulated storms, in
Preprints,I6th Conferenceon Severe Local Storms, pp. 269-274,
mentionedin part 1). This suggeststhat much additional American Meteorological Society, Boston, Mass., 1990.
work needs to be done toward understandingthe processes Leftwich, P. W., Jr., and Wu X., An operational index of the
that initiate and supportstorm rotation and mesocyclone- potential for violent tornado development, in Preprints, 15th
induced tornado development. Conference on Severe Local Storms, pp. 472-475, AmeNcan
Meteorological Society, Boston, Mass., 1988.
Maddox, R. A., An evaluation of tornado proximity wind and
stability data, Mort. Weather Rev., 104, 133-142, 1976.
Acknowledgments.The authorsespeciallyappreciatethe many McCaul, E. W., Jr., Simulations of convective storms in hurricane
hoursof work volunteeredby GrantBean,DaveHigginbotham, and environments, in Preprints, 16th Conference on Severe
590 PARAMETERS ASSOCIATED WITH STRONG AND VIOLENT TORNADOES, 2

Storms,pp. 334-339,American MeteorologicalSociety,Boston, storm characteristicsand 1200GMT hodographs,low level shear,

Mass., 1990. and stability, in Preprints, 13th Conference on Severe Local
McCaul, E. W., Jr., Buoyancyand shearcharacteristics
of hurri- Storms, pp. J5-J8, American Meteorological Society, Boston,
cane-tornadoenvironments,Mort. Weather Rev., 119, 1954-1978, Mass., 1983.
1991. Riley, P. A., and J. R. Colquhoun,Thermodynamicand wind
Miller, R. C., Notes on analysis and severe stormsforecasting related variables in the environment of United States tornadoes
procedures of the Air ForceGlobalWeatherCentral,Tech.Rep. and their relationship to tornado intensity, in Preprints, 16th
200, (rev.), 1972. Conference on Severe Local Storms, pp. 599--602, American
Moller, A. R., C. A. Doswell III, and R. Przybylinski, High- MeteorologicalSociety, Boston, Mass., 1990.
precipitation supercelIs:A perceptualmodelanddocumentation, Rotunno, R., and J. B. Klemp, The influence of the shear-induced
in Preprints,I6th Conference onSevereLocal Storms,pp. 52-57, pressuregradienton thunderstormmotion, Mort. Weather Rev.,
AmericanMeteorologicalSociety,Boston,Mass., 1990. 110, 136-151, 1982.
Moncrieff,M. W., and J. S. A. Green,The propagationandtransfer
Weisman, M. L., and J. B. Klemp, The dependenceof numerically
propertiesof steadyconvectiveoverturningin shear,Q. J. R.
simulated convective storms on vertical wind shear and buoy-
Meteorol. Soc., 98, 336-352, 1972.
Nolen, R. H., A radarpatternassociatedwith tornadoes,Bull. A•n. ancy, Mon. WeatherRev., 110, 504--520,1982.
Meteorol. Soc., 40, 277-279, 1959. Weisman, M. L., and J. B. Klemp, The structure and classification
Przybylinski,R. W., S. Runnels,P. Spoden,and S. Summy,The of numericallysimulatedconvective storms in directionally vary-
Allendale,Illinoistomado--January7, 1989--Onetype of an HP ing wind shears,Mon. Weather Rev., I12, 2479-2498, 1984.
supercell,in Preprints,16thConferenceon SevereLocal Stor•ns, Weisman, M. L., and J. B. Klemp, Characteristics of isolated
pp. 331-357, AmericanMeteorologicalSociety, Boston,Mass., convective storms, in Mesoscale Meteorology and Forecasting,
1990. edited by P.S. Ray, pp. 331-358, American Meteorological
Rasmussen, E. N., and R. B. Wilhelmson, Relationshipsbetween Society, Boston, Mass.,
 Diurnal Low-Level Wind Oscillationand Storm-Relative Helicity
ROBERT A. MADDOX

EnvironmentalResearchLaboratories,NOAA, National SevereStormsLaborator3.',Norman, Oklahoma 73069

1. INTRODUCTION shift are not clear. However, it is known that all diurnal wind
circulations(e.g., seabreeze, nocturnaljet, mountain-valley
Thunderstormscharacterizedby significantmesocyclones
breeze) are expected to be most pronouncedat about 30ø
occur within local environmentsthat exhibit strong,storm-
latitude north or south, where the inertial period is nearly
relative low-level winds which veer substantiallywith height
equal to the forcing period [Rotunno, 1983]. It is likely that
[e.g., Weisman and Klemp, 1984;Burgessand Curran, 1985;
the well-known diurnal oscillation of the low-level jet [e.g.,
Davies-Jones et al., 1990]. Such an environment, i.e., one
Blackadar, 1957; Wexler, 1961; Bonner, 1968] can, under
with high storm-relative helicity [Lilly, 1983, 1986; Davies-
some conditions, act to increase storm-relative helicity after
Jones, 1984;Brooks et al., this volume; Davies-Jones,1991],
is favorable for the occurrence of intense tornadoes when
dark and help support the nighttime occurrence of intense
tornadoes.
and if deep moist convection occurs. All atmosphericpro-
This possibility is explored first by considering typical
cesses that act to modify the character of the low-level,
afternoon conditions that favor supercell thunderstorms.
vertical wind profile can lead to enhanced or diminished
The vertical wind profiles for the afternoon setting then are
storm-relative helicity and thereby affect the structure and
compared with profiles modified in ways that typify the
character of thunderstorms, should they occur. Since
development of the nocturnal low-level jet. It is shown that
changes in storm motion also affect relative helicity, it is
storm-relative helicity can be enhanced during the evening
assumedfor simplicity that changesin the lower troposphere
and nighttime hours because of the development of the
winds have little affect upon the movement of thunder-
nocturnaljet. The relevance of the diurnal cycle for under-
storms.
standing observed storm behavior is discussed.
Thunderstorms that produce intense tornadoes of F3
rating or higher [Fujita, 1971]tend to occur during middle to
late afternoon. For example, F3 or greater tornadoes occur 2. TYPICAL SUPERCELL ENVIRONMENTS
most frequently in the central United States between 1700
and 1900local standardtime [Fujita, 1987].The geographical The early empirical models of severe thunderstorm envi-
locations of F3 and greater tornadoes that have occurred ronments [e.g., Fawbush and Millet', 1954; Newton, 1963]
between 1200 and 1800 central standard time (CST) are highlightedthe importance of the low-level jetstream for the
shown in Figure 1. During the afternoon, winds near the occurrenceof intense storms. The role of veering winds with
surfaceare strongand often backedrelative to geostrophic. height also was noted and related to differential advections
If the winds between 2 and 3 km above the ground are strong of temperature and moisture, which serve to destabilize the
and highly veered relative to the surfacewind duringthe late atmosphereand support the development of deep convective
afternoon, the environment may be characterizedby high storms [Beebe and Bates, 1955]. Studies of proximity sound-
storm-relative helicity. ings taken within several tens of kilometers of tornadic
However, not all major supercellthunderstormsand tor- thunderstorms [Beebe, 1958; Darkow and Fowler, 1971;
nadoesoccur during the afternoon (refer to Figure 2). As Maddox, 1976] clearly showed the presence of strong and
illustratedin Figure 2, many fewer intensetornadoesoccur highly veered low-level winds (refer to Figure 3). More
betweenmidnightand sunrisethan occurduringthe after- recent work [Lilly, 1983, 1986; Davies-Jones, 1984] estab-
noon, and the geographical preferenceis clearly shiftedto lished the importance of storm-relative helicity for the de-
the south central states. The reasons for the geographical velopment of supercell storms. McCaul [1991] has shown
that proximity hodographsfor tornadoes occurring within
The Tornado:Its Structure,Dynamics,Prediction,andHazards. hurricanes also are characterized by strong veering and
Geophysical Monograph79
Thispaperis notsubjectto U.S. copyright.Published
in 1993by the helicity (see Figure 4).
AmericanGeophysicalUnion. Storm-relative hellcity is proportional to the area of the
592 DIURNAL LOW-LEVEL WIND OSCILLATION

' ' Npon-6PM i

F,3+
"'
TORNADO',S,',l•-18
I "\
CST,':1955-1989
/-. •' !

• t
', , L288 ornados , - -

',. •j• I '

• "-' ") i
•/
•' /'• ,,
""l

_ •

i
-- -- -i
i •.\ i
i i .-
i I
"\ i \1
'•, I

Fig. 1. Locations of occurrenceof F3 and greaterintensitytornadoesbetween 1200and 1800CST. Period of record

is 1955-1989.

k - -_

I
.

•1 L
!

Fig. 2. SameasFigure1 butfor period0000to 0600

 MADDOX 593

560 ø

SFC160/20 13/5 •N
5• 225/

200 •45/

Tornad
-"."•• '- "2•kts
I .........
• : '-.....
•'•
Xx• xxx
•
/ StormReBtive
/ {......)//
Winds
Assumi• StormMohon 245ø/38 knots
X• SFC
090/4•
• • 850 I 30/36 •

180

Fig. 3. Compositehodographfor 23 tornado outbreaks.Estimated storm-relativewind vectors are shown below the
hodograph[from Maddox, 1976].

low-level wind hodograph that lies between the head of a

v(m
•/a)3;•3
4
storm motion vector and the environmental wind hodograph.
This area is computed from the surface to some arbitrarily
determined level chosen to capture the inflow layer for

' 576
convective storms, typically 3 km [Davies-Jones et al., 1990;
Doswell, 1991]. The storm-relative wind hodograph for a
composite of 28 supercell environments [Brown, 1990] is
contrasted with three similar hodographs for long-track,
damagingderechos [Johns and Hirt, 1987] in Figure 5. The
Sfc supercellenvironment exhibits large helicity, while the dere-
cho straight-line wind storm environment exhibits almost
I I ! ', none.
-20 -10 0 10 20 Thus the typical environments of tornadic storms possess
u (m g'b substantial storm-relative helicity. A pronounced low-level
jet also typifies these environments [Newton, 1963; Miller,
Fig. 4. Mean hodograph for hurricanetornadoproximitysound-
ings.SeeMcCaul [ 1991]for details.The height(in kilometers)above 1967] and can contribute to the helicity, especially when it is
surface is indicated. stronglyveered relative to the surface wind. The
594 DIURNAL LOW-LEVEL WIND OSCILLATION

T poorly organized storms occasionally change markedly

around sunset,with supercell events developing during the
eveningto early nighttime hours (C. A. Doswell and R. P.
Davies-Jones, personal communication, 1992). However,
evolution from supercell storms into multicell storms or
850 ! Sfc m t mesoscalesystems(e.g., derechos [Johns and Hirt, 1987]) at
or after dusk is more typically observed. These contrasting

-30 -

Fig. 5. Composite hodograph for 28 supercell environments

(dashed) presented in the storm-relative reference frame [from ovofo
Brown, 1990].Also shownare system-relativehodographs for three Thundorstorm
long-lived derecho events (solid lines; refer to Johns et al. [1990]). and Tornado

ment of the low-level jet in severe thunderstorm situations

characterizedby strong large-scalebaroclinicityand highly
ageostrophic winds has been discussed by UcceIlini and
Johnson [1979]. In these situations, intense isallobaric accel-
erations of the low-level xvindsand the veeringof the wind i

with heightbeneaththe exit regionof the upperleveljet can b

AIIobaric
result in a high-helicity environment.
During late spring and early summer, when large-scale
storm systems are not intense, late afternoon conditions still •, • Acceleration
lead frequently to environmentscharacterizedby high helic-
ity. An example of this type of situation is illustrated in ',,, % ^
Figure 6. A weak subsynopticsurfacelow often developsat
the northern extent of the dry line (Figure 6a) with severe
storms and tornadoes occurring to the north and east of this -20 / /x IsaJIobaric
/• -2(• .Acceleration
feature [Tegtmeier, 1974]. The local isallobaricfield acts to
increasethe meanlow-level winds as the southerlypressure
gradient increasesand to back the winds (Figure 6b). These
--Vector
changesoften occur in the presence of substantiallow-level
warm temperature advection and its associatedveering of
the geostrophicwind with height (Figure 6c). The net result
can be a hodograph characterized by high storm-relative
helicity over a distinctly mesoscaleregionduringthe :niddle
to late afternoon.
The morning and afternoon hodographsfrom the Binger
tornado(F4) day in centralOklahomaillustratewell thistype
of diurnal evolution of the hodograph(Figure 7). As Doswell
[1987] discussed,these types of mesoscalestormepisodes
are naturally more difficult to anticipate than are severe
thunderstorm and tornado outbreaks that usually occur
within highly baroclinic, synopticweather systems.Indeed,
Anthony [1990] has shown that the forecasters at the Na-
Fig. 6. (a) Typical late springsynopticpatternassociatedwith
tional Severe Storms Forecast Center have considerably supercell storms in the southern Plains. Fronts, isobars, an iso-
higher verificationscoresfor tornado outbreakdays. therm,andthesevereweatherthreatareaare shown.Approximate
timeof mapis 1500CST [afterTegtmeier,1974].(b) Typicalthree
hourlypressurefall field associated
with this pattern[after Tegt-
3. THE NOCTURNAL JF,T AND
meier, 1974].(c) Typical meanisothermsfor the 1.5- 3.0-km layer
STORM-RELATIVE HELlCITY associatedwith this pattern. Broad arrow indicates mean wind
vector in the layer, while inset shows the thermal wind vector and
It is observed by storm chasers, whose activities tend to
associated
veeringthroughthe layer due to geostrophicwarm
be confined to the open spaces of the Plains states, that
 MADDOX 595

observationsleadto an obvious question: whatrole,if any,

doesthe diurnalwind cycleplay in determining
storm-relative
the local,
hellcityandtherebystormorganization?
The typicalscenariofor low-leveljet development
absenceof any evolvinglarge-scale
(in the
baroclinicforcing)be-
v (mi'•) 2ø
l o3c1650•C / ,
gins as radiationalcoolingdecouplesthe surfacecontact s fc
layer from the well-mixedafternoonboundarylayer. This
resultsin an inertialoscillation,becauseof the largeimbal-
-10 0
• 2•Z•• •3650'
ancebetweenfrictional and Coriolisforces,that accelerates
andveersthe windsthougha layerthat mayextendup to 3 u (m
km abovethe surface.The flow becomessupergeostrophic
and the low-leveljet reachesits maximumintensityin the Fig. 8. Average diurnal variation from the mean wind at Okla-
homa City, Oklahoma, and Wichita, Kansas, for 29 quiescent
early morninghours [Blackadarand Buajitti, 1957;Hoxit, summerdays[afterBlackadarand Buajitti, 1957].The windswere
1975]. Uccellini [1980] has shown that many of the cases averagedrelativeto the700-mbarwind, whichfirstwas rotatedto be
used to study the diurnalcycle of the low-leveljet under from the west in all cases.
"quiescent"conditionsalsohavebeencharacterized by an
evolving baroclinic environment that influenced the devel-
opment of the jet. This complicatesinterpretationof the
cooling/decouplingaspects of the diurnal circulation fea- and cloud cover, leadingto very complex and dramatically
tures. However, he found that even in the face of compli- varying evolutionsof the low-level jet and the hodographon
cating baroclinicprocesses,the maximumwind speedsin a case-by-casebasis;again, see the case analysesby Hoe-
low-leveljet streaksusuallyoccurredduringthe earlymorn- cker [1963] and Bonner [1966]. The special observations
ing hours. reported by Stensrud et al. [1990] also appear to have
Althoughthe broad characterand climatologicalaspects capturedexamplesof such local variability.
of the low-level jet are well known [Bonner, 1968; Reiter, The work of Blackadar and Buajitti [1957] can be used to
1969], particularly under synopticallystagnantsummercon- illustrate that the diurnal wind cycle leads also to a cycle in
ditions,the detailedtemporaland spatialstructuresof thejet storm relative helicity. This is apparent from examinationof
have not been well observed. Hoecker [1963] and Bonner Figure 8, where the amplitude of the averagediurnal cycle is
[1966] used serial observations from a special two- largestfrom 1000to 3000 feet (300-900 m) above the surface.
dimensionalpibal line operatedduringspringof 1961to show This diurnal cycle can lead to a rapid increase in storm-
the complicated structures and evolution of the southern relative helicity after 1500 local time that maximizes in the
Plains low-level jet for several case events. Maddox [1985] early morning, if it is assumed that middle tropospheric
suggestedthat the mesoscale structures and timing of the "steering" winds, and thus the general movement of thun-
diurnal jet may be influenced strongly by local radiative derstorms, remain relatively unaffected by the diurnal cycle.
conditions,particularly the effectsof mean relative humidity It must be remembered that these changes are "average"
summertime features and that their relative importance and
magnitudevaries markedly from case to case, as shown by
Uccellini [ 1980].
For example, consider the four hypothetical scenarios

30
1 22
MAY
1981
shown in Figure 9. In each of these four cases an initial
v (m•l) 1 afternoon storm-relative hodograph is modified by the same
representativediurnal changes(as indicated by Figure 8) to
produce a 0300 local time hodograph. The initial profiles

01
z ••,
(23) range from a straight-line hodograph (i.e., one which is
characterized by no turning with height of the wind shear
vectors) to highly curved ones. In the first case (Figure 9a),
Sfc 2t
changesin the wind profile after dark lead to little change in
oKc • J
12 Z $fc • storm-relative hellcity and a trend toward a straight-line

(22) I .•-•Storm
Motion
hodograph below 750 m. However, the other three cases
I ...... I , i ....... I (Figures 9b-9d) lead to substantial increases in storm-
-10 0 10 20 30 relative helicity. Thus the meteorological situation that de-
U (m g•) velopsduringthe afternoonand the resultinglocal wind profile
relate directly to how the environment will evolve, in a storm-
Fig. 7. Morning and evening hodographsassociatedwith the relative helicity sense, as radiational heating decreasesrap-
Binger, Oklahoma,tornadicsupercellof May 22, 1981.The morning
idly at the end of the day. It appears that in most situations
soundingis from OklahomaCity, while the afternoonsoundingwas
taken at Tuttle, Oklahoma(approximately15 miles southwestof the componentof the diurnal wind cycle driven by frictional
OklahomaCity) (hodographs providedby D. Burgess). decouplingalone acts to increase low-level
596 DIURNAL LOW-LEVEL WIND OSCILLATION

a 30 b 80
v (m •'•)
T v(m
•1)l
20 //•
: ! i' !
-20 -10 0 10 20 -20 -10 0 10 20
u (m s•1) u (m •)

cv(ms
3o
"•)T.4;:•5AM
dv(m AM
.... 00m .....

-20 -10 0 10 20 -20 -10 0 10 20

u (m •'•) u (m •'•)
Fig. 9. •odo•mphs showJn•hypotheticalchan•esin storm-relative
low-levelwinds betweeni500 and 0300 local
time.Thediurnalc•an•esimposed
aresimilarto thoseindicated
in •isure8 andarethe samein eachpanel.

Although we have preciousfew observationsof the evo- modifiedoperationalprocedureswill be requiredto capture

lutionof the verticalwind profilewith hightimeandspace importantdetailsof' low-level jet structuresand evolution
resolution,the capabilitiesof radar wind profilersand the when the noseof the jet occursvery near the surface.
new National Weather Service (NWS) Doppler radars The clearair wind data shownin Figure 10were gathered
(WSR-88D)bringpromisethat, at least in the time domain, by the first NWS WSR-88D radar operatingat Twin Lakes,
we will soonbe ableto observethe diurnalwindcyclewith Oklahoma,thispastspring.The resultinghodographs
(Fig-
more detail thanever in the past.Stensrudet al. [1990]have ure 11) showthat storm-relativehelicity in the environment
noted that the radar wind profilersdeployedin the central over central Oklahomaessentiallydoubled between sunset
United Statesdo not samplethe windswell from the surface and midnight. There were no thunderstorms in the state on
to 500-750 m. This means that supplementalsystemsor this particularnight, so the possiblethunderstorm-related

•o '• '",,x• "•'- ",,½"- '•--'- ........ • "½,"• '",:• "'• .......
• ...... '.................
• • • -•-•' '• • • ,•-',,"• '"• "• "½•--?",•.-','-"-,,•
•,':• ,,,.• .,..,:.• •4• ..............

I .......
1..... 1 I I t, t 1 I I i I i I I I I I I I I I
TIME 2347 23560006 0018 0026 003G0046 00550105 0115 01250445 0454 05040514 0524 05330543 0553 0603 0613 0622
........
,,,

Fig. 10. Clearair,velocity

azimuth
display
(VAD)vertical
windprofiles
fromtheNWSWSR-88D
radarat Twin
Lakes,
Oklahoma
(approximately
25miles
east-southeast
ofOklahoma
City).Period
extends
from2347UTC(1747
CST)onMarch
25to0622
UTC(0022
CST)onMarch26,1991.
Winds
areinmeterspersecond;
fullbarbindicates
5
ms-1, andpennant
indicates
25m s-•. Notethetimebreak
at0125
 MADDOX 597

WSR-88D occur during the night may be decoupled from the near-
Hodographs Evening surfacelayer, if it has stabilized. The greateststorm-relative

v(m
9!)40
1 of
25/26
March
1991
helicity often resides in this layer, further complicating the
understandingof nighttime thunderstorm behavior. It is not
surprising that detailed prediction of the character and
130 4 Midnight evolution of individual thunderstorms remains elusive.

Acknowledgments. Discussions with David Stensrud and

Harold Brooks helped the author to improve this presentation.
Suggestionsprovided in reviews by Doug Lilly and Lou Uccellini
have helpedthe author clarify the central thesisof the paper. David
Keller providedthe mapsof tornado occurrences.The figureswere
6 PM
carefully preparedby Joan Kimpel. The hodographsfrom the Binger

Stor•9000'
tornado day were provided by Don Burgess.The WSR-88D wind
observationswere made available by Ron Alberty of the NEXRAD
Motion OperationalSupport Facility in Norman. The author thanks all of
I I "i I these personsfor their help and support.
-10 0 10 20 30
u (m •) REFERENCES

derivedfrom the velocityazimuth Anthony, R., Trends in severe local storm watch forecasting per-
Fig. 11. Three hodographs
display (VAD) winds show a marked increase in storm-relative formance at the National Severe Storms Forecast Center, in
helicitybetween1800and 0000 CST. Middle tropospheric winds Preprints 16th Conference on Severe Local Storms, pp. 281-286,
wereusedto estimatea likelystormvelocity(noconvection actually American Meteorological Society, Boston, Mass., 1990.
occurred).Profiletimesare (circles)0006UT, (triangles)0306UT, Beebe, R. G., Tornado proximity soundings,Bull. Am. Meteorol.
and(squares)0603UT. Stippledareashowsincreasein helicityfrom Soc., 39, 195-201, 1958.
1800 to midnight. Beebe, R. G., and F. C. Bates, A mechanismfor assistingin the
release of convective instability, Mort. Weather Rev., 83, 1-10,
1955.
Blackadar, A. K., Boundary layer wind maxima and their signifi-
effectsof thesechangesin the hodograph
asthe low-leveljet cance for the growth of nocturnal inversions, Bull. Am. Meteorol.
developed remain a matter of speculation. Soc., 38, 283-290, 1957.
Blackadar, A. K., and K. Buajitti, Theoretical studies of diurnal
wind structure variations in the planetary boundary layer, in
4. DISCUSSION Studies of Wind Structure in the Lower Atmosphere, edited by
J. E. Miller, pp. 45-79, Dep. Meteorol. and Ocean., New York
This brief presentation illustrates that at least in certain Univ., 1957.
situationsthe developmentof nocturnallow-leveljet streaks Bonner, W. D., Case study of thunderstorm activity in relation to
can have dramatic effectsupon the storm-relativehelicity of the low-level jet, Mon. Weather Rev., 94, 167-178, 1966.
the environment within which storms may be occurring. Bonner, W. D., Climatology of the low-level jet, Mon. Weather
Rev., 96, 833--850, 1968.
These changescan, dependingupon the specificmeteorolog- Brooks, H. E., C. A. Doswell III, and R. P. Davies-Jones, Envi-
ical details of any particular situation,act either to increase ronmental helicity and the maintenance evolution of low-level
or to diminish the likelihood of highly organized, supercell mesocyclones, this volume.
thunderstorms. Becauseof the crude time and spaceresolu- Brown, R. A., Characteristicsof supercellhodographs,in Preprints,
tion of current observing systems it is not possible at this 16th Conferenceon Severe Local Storms, pp. 30-33, American
Meteorological Society, Boston, Mass., 1990.
time to quantify these effects; indeed, because of observa- Burgess,D. W., and E. B. Curran, The relationshipof stormtype to
tional limitations we do not yet have fine-resolutiondocu- environmentin Oklahoma on 26 April 1984, in Preprints, 14th
mentation of the behavior and structuresof the low-level jet Conferenceon Severe Local Storms, pp. 208-211, American
over a wide variety of geographicaland meteorological Meteorological Society, Boston, Mass., 1985.
Darkow, G. L., and M. G. Fowler, Tornado proximity sounding
settings.However, this paper illustratesthat the well-known wind analysis, in Preprints, 7th Conference on Severe Local
diurnal cycle of the low-level jet may be influencing the Storms, pp. 148-151, American MeteorologicalSociety, Boston,
severe storm environment in many and more complex ways Mass., 1971.
than have been considered in the past. Davies-Jones,R. P., Streamwisevorticity: The origin of updraft
rotation in supercellstorms,J. Atmos. Sci., 41, 2991-3006, 1984.
Naturally, the effects discussed in this paper occur in
Davies-Jones,R. P., and H. Brooks, Mesocyclogenesis from a
concert with a wide range of other atmospheric processes theoretical perspective, this volume.
that also act to modify the low-level winds. These include Davies-Jones,R. P., D. W. Burgess,and M. Foster,Test of helicity
slopingterrain, spatial variationsin surfacecharacter,spa- as a tornadoforecastparameter, in Preprints, 16th Conferenceon
tial and temporal variations in cloudiness, synoptically Severe Local Storms, pp. 588-592, American Meteorological
Society, Boston, Mass., 1990.
evolvingpressureand heightfields, etc. Further, the diurnal Doswell,C. A., III, The distinctionbetweenlarge-scaleand mesos-
wind cycle is usually out of phasewith the cycle in convec- cale contributions to severeconvection:A case studyexample,
tive instability.This meansthat the updraftsof stormswhich Weather Forecasting, 2, 3-16,
598 DIURNAL LOW-LEVEL WIND OSCILLATION

Doswell,C. A., III, A review for forecasterson the applicationof on SevereLocal Storms, pp. 202-207, American Meteorological
hodographsto forecastingseverethunderstorms,Natl. Weather Society, Boston, Mass., 1985.
Dig., 16, 2-16, 1991. McCaul, E. W., Jr., Bouyancy and shear characteristics of hurri-
Fawbush,E. J., andR. C. Miller, The typesof air massesin which cane-tornadoenvironments,Mon. Weather Rev., 119, 1954-!978,
North American tornadoesform, Bull. Am. Meteorol. Soc., 35, 1991.
154--165, 1954. Miller, R. C., Notes on analysisand severe-stormforecasting
Fujita, T. T., Proposedcharacterizationof tornadoesandhurricanes proceduresof the Military WeatherWarningCenter, Tech. Rep.
by area and intensity, SMRP Res. Pap. 91, 42 pp., Univ. of 200, Air Weather Serv., 94 pp., Scott Air Force Base, II1., 1967.
Chicago,Chicago,Ili., 1971. Newton, C. W., Dynamics of severeconvective storms, Meteorol.
Fujita, T. T., U.S. tornadoes.Part One: 70-year statistics,SMRP Monogr., 5(27), 33-58, 1963.
Res. Pap. 218, 122pp., Univ. of Chicago,Chicago,I11.,1987. Reiter, E. R., Tropopausecirculation and jetstreams, in World
Hoecker, W. H., Three southerlylow-leveljet streamsdelineated Surveyof Climatology,vol. 4, Climate of the Free Atmosphere,
by the Weather Bureau special pibal network of 1961, Mon. edited by D. F. Rex, pp. 85-193, Elsevier, Amsterdam, 1969.
Weather Re,,., 9, 573-582, 1963. Rotunno, R., On the linear theory of the land and sea breeze, J.
Hoxit, L. R., Diurnalvariationsin planetaryboundary-layer winds Atmos. Sci., 40, 1999-2009, 1983.
over land, Boundat3,Layer Meteorol., 8, 21-38, 1975. Stensrud,D. J., M. H. Jain, K. W. Howard, and R. A. Maddox,
Johns,R. H., and W. D. Hirt, Derechos:Widespreadconvectively Operational systems for observingthe loweratmosphere: Impor-
inducedwindstorms,WeatherForecasting,2, 32-49, 1987. tance of data samplingand analysisprocedures,J. Atmos. Oce-
Johns,R. H., K. W. Howard, and R. A. Maddox, 1990:Conditions anic Tech., 7, 930-937, 1990.
associatedwith long-lived DERECHOS--An examination of the Tegtmeier,S. A., Theroleof thesurface,sub-synoptic
low pressure
large-scaleenvironment,in Preprints,16th Conferenceon Severe systemin severeweatherforecasting,M.S. thesis,66 pp., School
Local Storms, pp. 408-412, American MeteorologicalSociety, of Meteorol., Univ. of Okla., Norman, 1974.
Boston, Mass., 1990. Uccellini,L. W., On the role of uppertroposphericjet streaksand
Lilly, D. K., Dynamicsof rotatingthunderstorms,in Mesoscale leesidecyclogenesis in the development of low-leveljets in the
Meteorology--Theories,Observationsand Models, edited by Great Plains,Mon. WeatherRev., 108, 1689-1696,1980.
D. K. Lilly andT. Gal-Chen,pp. 531-543,D. Reidel,Hingham, Uccellini,L. W., and D. R. Johnson,The couplingof upperand
Mass., 1983. lowertropospheric jet streaksand implicationsfor the develop-
Lilly, D. K., The structure,energetics
and propagation
of rotating ment of severe convectivestorms, Mort. Weather Rev., 107,
convectivestorms,II, Helicity and storm stabilization,J. Atmos. 682-703, 1979.
Sci., 43, 126-140, I986. Weisman,M. L., and J. B. Klemp, The structureand classification
Maddox, R. A., An evaluationof tornado proximity wind and of numerically
simulatedconvectivestormsin directionally
vary-
stabilitydata, Mort. WeatherRev., 104, 133-142, 1976. ing wind shears,Mon. WeatherRev., 112, 2479-2498, 1984.
Maddox, R. A., The relationof diurnal, low-level wind variationsto Wexler,H., A boundarylayer interpretationof the low-leveljet,
summertime severethunderstorms, in Preprints,14thConference Tellus, 13, 368-378,
 Tornadoes' A Broadcaster's Perspective
TOM KONVICKA

Lanford Companies, KALB-TV, Alexandria, Louisiana 71309

1. INTRODUCTION fatality statisticsare analyzed.It is generallyacceptedthat a

ratherdramaticdropin the tornadodeathtoll has occurred
Tornadoes are quite common in the United States. In 1990 duringthe past 20 yearsand that propereducationof the
the Severe Local Storms Unit (SELS) of the National Severe public is at least partially responsible.For these reasons,
Storms Forecast Center (NSSFC) tallied 1132 tornadoes; broadcastmeteorologists must be aggressiveleadersin the
this is the highesttotal on record (SELS smoothlog, internal attemptto educatethe publicabouttornadoes(andweather
report, 1990). This number represents only tornadoes that in general).
were properly documented. Thus it is likely the actual One primary vehicle for accomplishingthe goal of a
number of tornadoes in the United States for 1990 is some-
successful
campaignon tornadoeducationis the schoolvisit.
what (perhaps even considerably)higher. This visit can include preschool through college age groups.
Research, operational, academic, and various "private" In the schoolsetting,many eager and impressionableminds
entities each contribute to addressing hazards associated await the meteorologist'sprogram.The agendabeginswith
with tornadoes. The purposeof this paper is to present some
an explanationof howthunderstormsdevelop.Then, a video
tornado-related concerns from the author's viewpoint as a
potpourriof tornadoesin actionstimulatesaudienceinterest.
broadcast meteorologist. Section 2 discussesthe essential A demonstration model such as a "tornado in a bottle"
role a broadcast meteorologistmust play in educatingthe enlightensstudentsby showing them a simple physical
public about tornadoes.Section3 coversthe functionof the analoguewhich behaves, albeit crudely, like a natural tor-
broadcastmeteorologistbefore, during, and after a tornadic nado. The presentationconcludesby answeringquestions,
event. Section 4 analyzes some issuesregardinginteraction sharingexperiences,and reviewing safety rules.
between broadcast meteorologists and their peers in the A second avenue the broadcast meteorologist must ex-
operational sector in the context of severe local storms. ploreto accomplishhis/hergoal of publiceducationis a visit
Section 5 highlightstwo effects the ongoingtechnological to local civic organizations.It is commonfor these meetings
boom could cause in the broadcastingarena. to include local decision makers. Some of these people may
even be involved with handling weather-related disasters.A
2. ROLE OF BROADCAST METEOROLOGISTS comprehensiveprogram is appropriate. As in the school
IN PUBLIC EDUCATION presentation,the tornadovideo is the mostpopular compo-
nent of the agenda.
Meteorologists,be they in the operational,research,aca- Another favorite aspect of the meteorologist'svisit is the
demic,or broadcasting areas,can play an importantrolein questionand answer session. Many tornado-relatedanec-
helping to increasepublic knowledgeand awarenessof dotes are told, and it is up to the meteorologist to find
tornadoes. The broadcaster's function, however, may be quickly a plausible explanation for these experiences.Yet,
especiallyvital since in most areas the local broadcast following the meeting, more serious discussionmay ensue,
meteorologistis the mostvisiblerepresentativeof the pro- and the broadcaster may learn that some members of the
fession. In smaller markets (where no National Weather civic organization are involved with civil defense, the 911
Service(NWS) officeexistsand only one or two television program,or the NWS. This example of "networking" with
stationsare present)the broadcastmeteorologist maywell people enables the broadcast meteorologistto be more in
bethesoleexampleof his/herprofession.
Theimportanceof touch with those involved with the meteorological profes-
properpublic educationis evidentwhen tornado-relatedsion as well as those exterior to the meteorological commu-
nity who must also deal with tornado hazards.
The Tornado:Its Structure,Dynamics,Prediction,and Hazards.
GeophysicalMonograph 79 A third aspect of public education is the production of
Copyright1993by the AmericanGeophysical Union. written materials. This could include almost anything rang-
600 TORNADOES: A BROADCASTER'S PERSPECTIVE

ing from a one-pagepamphleton tornado safetyrules to a broadcast meteorologists can, like NWS forecasters, be-
full-lengthbookdescribingseveraltornado-related
topicsin come aware of the salient features which may lead to an
detail. The task of writing some of these items can be outbreak of tornadoes. Some broadcasters are tempted to
time-consuming but is quitebeneficialto both the publicand consideronly the available graphicalguidance.Although this
the meteorologist.The pamphleton tornadosafetyrulescan graphicalproducteasily delineatesthe severe local storm
be locatedin the home where it is easily accessibleduringa threat area, it does not provide insight into the conditions
tornado warning. The full-length book can provide hoursof operatingto bringaboutpotentialfor tornadoes.
interestingand enjoyablereading.By authoringsuchmate- As time passesand it becomespossiblefor SELS fore-
rial the broadcastmeteorologistenlists the power of the casters at NSSFC to refine some of the details of the
written word to help carry out tornado education. Hence evolving situation, a Mesoscale Discussion is usually dis-
television meteorologistsneed not be limited to the usual seminated. This product, like the earlier convective out-
verbal method of conveying information. In addition to
looks, is easilyreceived via alphanumericdata services. The
employing the written word as a communicationmethod, the
MesoscaleDiscussionis recommendedreading for the seri-
television meteorologistis increasinghis/her knowledgeof
ous broadcastmeteorologistbecause it gives details, based
tornadoes by researching the topic. An important end result
on the interpretation of SELS forecasters, concerning the
is to solidify credibility with viewers.
evolutionof factors which may be responsiblefor producing
tornadoes in the near future. At this point the broadcast
3. TORNADIC EVENTS:
meteorologistshould already have performed (1) sounding
THE BROADCAST METEOROLOGIST AT WORK
analysis and interpretation and (2) subsynopticanalyses of
Besides fostering public education, another critical re- available surface data. A helpful resource for the author on
sponsibility the broadcast meteorologist must accept is the topics of sounding analysis and surface subsynoptic
proper coverage of tornadic events. For the discussion analysisis given by Doswell [1982]. By performing these two
presentedin this section, the processa broadcastmeteorol- fundamental tasks, broadcast meteorologists exercise their
ogist engagesin when covering tornadic episodescan be skills as practitionersof meteorology. Also, the telecaster's
divided into three phases: anticipation, recognition, and ability to understandthe suite of valuable convective prod-
postevent.The anticipationphasemay be definedas the time ucts from SELS will be augmented. Thus the individual feels
when dynamic/thermodynamicconditionssuggestpotential that he/she is becoming more mature and competent as a
for tornado development. As such, then, the anticipation
professional,and his/her level of on-air confidence rises.
phase is a forecastingstage. The recognitionphasebegins
Viewers will notice the difference in knowledge and confi-
when echoes appear on the radar display and must be dence.
interpreted as either severe or nonsevere. The postevent
As parameterscontinue to mesh, a tornado watch is issued
phase occursafter tornadoeshave dissipatedor have moved
out of the viewing area. by SELS. During a tornado watch it is important to convey
to viewers the ideas containedin the following statements.
3.1. The Anticipation Phase 1. Continue normal activities but be aware of nearby
thunderstormsand the weather changesthey bring.
Successfulperformanceduringtornadic episodesbegins
with proper anticipationof the event. It is essentialthat both 2. It is commonto have little or no warningof a tornado.
the operationalpersonnel and the weather staff at the local Be prepared to move quickly to a place of safety if you
television station correctly diagnose tornadic potential perceive danger is imminent.
present in the atmosphere.This is one processwhosevalueThe conceptscontainedin thesetwo statementsproperly
will not be rendered obsoleteby ongoingtechnological inform the publicon the situationwithout creatingunneces-
advances. Foster [ 1990] comments that even when the NEXt sary alarm.
Generationof WeatherRADars(NEXRAD) becomesfully At this point the qualified [American Meteorological So-
ciety, 1991a] broadcast meteorologist, who is also well
operational,successfulanticipationof an event will allow the
full potentialof the N EXRAD systemto be realizedduring trained in the field of severe storms meteorology, must
warning situations. analyzecritically the reasoninggiven by SELS forecasters
As onemightexpect,muchof the informationusedby the as the basis for issuingthe watch. It is vital that broadcast
broadcastmeteorologist duringthe anticipationphaseorigi- meteorologists be able to think for themselves. Viewers are
nates at the NSSFC. The first inkling of a tornado threat is best served when independentthoughtful synthesisand
often discussedin either the First Day SevereThunderstorm uniqueknowledgeof local meteorological effectsprovided
Outlook or the Second Day Convective Outlook. These by the local broadcastmeteorologistare combinedwith
outlooks are designed to assist NWS field offices with the expertisefromSELS forecasters. Even if futuregovernment
task of forecastingconvectiveweather.Theseproductsare plansto transferwatch responsibilityfrom SELS to NWS
transmittedfrom the NSSFC and are easily accessibleto all field offices[Friday, 1988]becomereality, the broadcaster
broadcastmeteorologists.By evaluatingtheseproductsand will still needto be capableof followingthe anticipation
by interpreting computer model output for themselves, phase
 KONVICKA 601

3.2. The Recognition Phase ous, a broadcast meteorologist may actually be the one to
For the discussionpresented in thissection,assumethat survey and report on how' the viewing area was affected by
thunderstormshave developedand must be monitoredfor the tornadoes. Subsequently, the broadcaster should work
indicationsof tornadicactivity.Again,it is desirablefor the with NWS personnel to assure that the event is properly
broadcast meteorologistto follow storm motion and evolu- documentedand takes its place on the climatological record.
tionindependently, butonlyif he/sheisqualified
andhasthe The motivated broadcastmeteorologist may considerwrit-
appropriate equipment (aconventional ora Dopplerweather ing a case study review of the event. The case study could be
radar) to use. presented, for example, at a conference on severe local
Offundamental importance duringtherecognition phaseis storms or at a weather analysis and forecasting conference.
the association of rotation in a storm with certain radar The published version of the paper would appear in the
signatures. The pendantand hookecho[Stoutand Huff, respective conference preprint volume. As he/she writes
1953],the line-echowavepattern,or LEWP [Nolen,1959], about the episode, the process of fusing observations with
the bow echo [Fujita, 1978],and the derecho[Johnsand concepts and reading the literature allows the individual to
Hirt, 1987]shouldbefamiliarto the broadcaster.In addition, become more knowledgeable about severe local storms.
the review by Letnon[1980]is highlyrecommended to the Results of the case study can be applied to future severe
seriousmeteorologistwho desiresto understandmethodsof storm episodeswith the outcome being an enhanced perfor-
recognizing radar signaturesassociatedwith severe storms. mance level during the anticipation and recognition phases.
One importantfacet of the recognitionphasefrom the
broadcasting viewpointconcernshowto relayinformation
to 4. BROADCAST METEOROLOGISTS
the viewer. The choiceinvolvesusinga "crawl" statement AND NWS
(a message moved across the bottom of the television
interruptionof programming. It is appropriateto discussthe sometimesthorny question
screen)or a live, unscheduled
A "crawl" offers the advantageof allowingthe viewer to of interplay between the NWS and broadcast meteorologists
continueto watchthe programwithoutinterruption.How- in the context of severe local storms. Each entity plays an
ever, in rapidlychangingsituationswith multiplewarnings, essentialrole in ensuring that timely warning of impending
the "crawl" is not an efficient method of communication. severe weather is received; these respective functions have
The live on-airappearance is preferredfor tornadowarnings. already been clearly defined [American Meteorological So-
In this instance, the implication to viewers is that the ciety, 1991b]. It is precisely for these reasons that the
weather situation is becomingincreasinglydangerousand broadcastmeteorologistshould seek to develop a positive
that it is appropriate for them to react by being more relationship with the local NWS office and should not feel
conscious of weather conditions. Most viewers will not react obligedto criticize the NWS as an inept competitor for the
to a severeweather threat unlessthey believethere is danger public's attention.
or someone they trust tells them about it in a personal It has already been stated that the roles of the broadcast
manner. The on-air appearanceoffers this personaltouch. meteorologist and the NWS are to remain distinct during a
The maximum amount of information about the weather severelocal storm threat. It should be pointed out, however,
situationis includedduringan on-air update. It is crucialfor that there are circumstances which are not so easily re-
the public to receive the maximum amount of information solved. Specifically, the author has been faced with the
possible, in the most personal manner, from someonewho following instances:(1) a person calls the television station
has followed the recognition phase (and the anticipation and reports sightinga tornado, (2) a member of a spotter's
phase) independently. The viewer's reaction to the informa- network calls and reports sightinga tornado. (3) there is no
tion given may mean the differencebetween life and death. NWS warningin a situationthat appearsto deserveone, and
(4) NWS issuesa warning that appears unwarranted.
Several factors enter into the decision on whether or not a
3.3. The Postevent Phase
tornadosightingby a caller is legitimate. First, the sighting
Once tornadic thunderstorms have moved through the must be reported in a competent manner. One method of
area and the Watch StatusReport from SELS has lifted the assessing competenceof the report is to askthe caller (1) Did
watch, the chore of locating and assessingdamage begins. you actually see the tornado? and (2) How do you know it
The broadcastmeteorologistcan contributevaluableservice was a tornadoand not somethingelse?Generally, thesetwo
during the posteventphase. In workingdirectly with news questionsprovide enough information to decide whether or
reporters (who have little or no meteorological training) the not to pursuethe matterany further. If the descriptionseems
meteorologistcan speculateon the forces which causedthe to hold some credibility, a fresh took at the radar is neces-
visible damageand can provide expertiseon the phenome- sary. This is anotherreasonwhy it is recommendedthat the
non deemed responsible.Also, it is best to have the meteo- qualified broadcast meteorologistfollow the recognition
rologist critically review a reporter's packagefor accuracy stage independently. If he/she has no skill in weather radar
before it airs on a newscast. In this way, the viewer has a interpretation,or no radar to interpret, then ignorance
better chanceof gettingcorrect information.In larger tele- becomes the basis for the decision and the viewer's best
vision markets, where weather personnelare more numer- interestis not well served.While questioningthe caller
602 TORNADOES: A BROADCASTER'S PERSPECTIVE

5. Two POSSIBLE FUTURE TRENDS

lookingat weatherradar,themeteorologist shouldwaitfor a
IN BROADCAST WEATHER
second reportof a tornado.Theconcurrent testimony oftwo
dependable witnesses is considered to be sufficient for Continuing technological advances will bringprofoundyet
placing a callto theNWSofficewithwarningresponsibilitybeneficialchangesto the entire meteorological profession.
andreporting theinformationto them.An on-airbulletinis The NEXRAD system, wind profilers, a series of new
deliveredafterthe reporthasbeenrelayedto the NWS. weather satellites,and the continuedgrowth of the computer
It is not uncommon for broadcast weather departmentsto industryconstitute themajorityof thetechnological revolu-
havean arrayof severestormspotters.Theseindividuals, tion. The broadcastingsector must directly addressthe
whenproperlytrained,are of greatvaluebecausetheyare challenges presentedby the presentwave of technology.
ableto seedetailwhichcan be missedby weatherradarand Oneway to accomplish thiswouldbefor a policyon hiring
satellitephotographs. Propertrainingincludesa working onlyqualifiedmeteorologists to staffthe broadcastweather
knowledge of the materialcoveredin the NationalOceanic center.A similarplanhasalreadybeenimplementedby the
andAtmospheric Administration seriesof spottertapesand NWS in an attemptto developa more professionalwork
pamphlets. A reportof a tornadofrom a memberof the force[Friday, 1988].It is bothlogicaland necessaryfor the
spotter'snetworkcan be considered reliableenoughto televisionindustryto follow the NWS lead in this matter.
immediatelyrelayto the NWS. In fact, the broadcastmete- The hazards associated with tornadoes demand maximum
orologist shouldencourage theseindividuals to placecallsto use of all available data in order to warn the public accu-
the NWS first and the television station second. rately. Thosebroadcasters not acquaintedwith all the new
Anotherpoint of interactionbetweenthe NWS and the technology(and its properuses)will not be able to deliver
broadcast meteorologistconcernsthosetimeswhennoNWS the qualityof informationthat will be expectedby viewers.
warningexistsin a situationthat appears,at least to the The competitivenatureof the broadcasting businesswill
broadcaster,to meritone. It is bestto be directwith viewers. dictatethe employment of qualifiedmeteorologiststo deliver
Therefore the author endorses a live on-air bulletin that weatherinformation,especiallyduringinclementconditions.
would convey ideasincludedin the following statement. In the future, in weather consciousbroadcastingmarkets
across the United States, the need for individuals with
KALB-TV 5 Dopplerweatherradar is monitoringa thunder- specialized knowledgeof severelocal stormswill increase.
stormwith the POTENTIAL to producea tornado.This storm In this setting, the viewer is assuredthat the broadcaster
is located at X POSITION and is expectedto affect CITIES
AND TOWNS duringthe next NUMBER of minutes.If youare
deliveringcriticalinformationis notonlya meteorologist but
in the expectedpath of this stormpleasebe keenlyawareof a specialistin the field of severelocal storms.Perhapseven
suddenchangesin the wind. If you feel dangeris imminent, a "severe stormsspecialist"certificationprogramcould be
then move quicklyto your basement,an interior hallway,or designedby inputfrom the National SevereStormsLabora-
closet. Do not wait for us to broadcastan official warning from
the National Weather Service. We will continue to follow this
tory, the NSSFC, andbroadcastmeteorologists. It is evident
situationcloselyand will inform you on any new developments.
that with the increasein technologycomes more responsi-
bility, not less, for the broadcastmeteorologist.

In effect, this approachrepresentsa stepbetweenthe watch 6. CONCLUSION

and warning[Friday, 1988].Also, this type of bulletinmakes
viewers aware of a potential threat but allows the broad- Thispaperhasaddressed severalfacetsof tornadoesfrom
caster to stop short of issuinghis/herown warning. the perspectiveof a broadcastmeteorologist.The role of the
A fourth concern occurs when it appears the NWS has broadcast meteorologistin fostering public education will
issueda warning that is not necessary.Again, honestywith remain vital for some time to come. Telecasters must not
the viewer is best. They shouldbe informedthat the warning forget that their highly visible position can enable them to
exists but that questions concerning the validity of the hold considerable sway over the opinion viewers have
warning are present. A typical exampleis a F0-FI tornado concerningthe tornadothreat. A crucial element in whether
spawnedby a pulsethunderstorm.In this circumstance,the or not the best information possibleis being conveyed to the
authorbelievesproperprocedureis to broadcastthe warning publicduringa tornadothreat is the individual's independent
but communicate that this type of tornado is usually weak performance.Clearly, proper coordination with the NWS
and short-lived. Thus a distinction is made between the reduces the chance for misunderstanding during warning
F0-F1 tornado and the F3-F5 variety. It is an opportune situations,assuresthat public/private sector roles remain
time to remind viewers that all tornadoesdeserverespectbut distinctand, when appropriate, permits the broadcastmete-
some command more concern than others. In this way, the orologistvaluableinput to the warning/nowarning decision.
broadcastmeteorologistfulfills his/her role as a partner in Also, documentation of tornadic events can be increased
the public/privaterelationshipby broadcastingthe tornado with better teamwork between the NWS and media. With
warning. In addition, the broadcastermaintainshis/her in- the presentboostof technologyin its infancy, it is hoped that
tegrity with viewers by being honest and keeps his/her television stationswill keep pace with the NWS by raising
credibility intact by not falling victim to the "media hype" standardsfor employment in the broadcast weather depart-
temptation.
 KONVICKA 603

The author's experience indicates that there is a gap Doswell,C. A., III, The operational meteorology of convective
separatingthe broadcasting,operational, and researchcom- weather,vol.I, Operational
mesoanalysis, NOAATech.Memo.,
NWS NSSFC-5,158pp., Natl. SevereStormsForecastCenter,
munities. However, the author believes technological im-
Kansas City, Mo., 1982.
provementswill help result in a fusion of these splintered Foster,M.P., NEXRADoperational issues:Meteorological
consid-
groupsby increasingtheir interdependence.Motivated and erationsinconfiguring
theradar,in Preprints,16thConference
on
qualifiedbroadcastmeteorologistsshouldwork with peersin SevereLocal Storms,pp. 189-192, AmericanMeteorological
the operational, research, and academic sectors to assure Society,Boston,Mass., 1990.
that future growth of the meteorological profession is har- Friday,E. W., Jr.,TheNationalWeatherService SevereStorms
monious.
Program:Year2000,in Preprints,15thConferenceon Severe
LocalStorms,
pp.Jl-J8, AmericanMeteorological
Society,Bos-
ton, Mass., 1988.
Fujita,T. T., Manualof downburst
identification
forProject
NIM-
Acknowledgments. The author wishes to thank cognizant editor ROD, SMRPRes.Pap. 156,42 pp., Univ.of Chicago,Chicago,
Charles A. Doswell III and two anonymous reviewers for their II1., 1978.
patience and for the professionalmanner in which they accom- Johns,R. H., andW. D. Hirt, Derechoes:
Widespread
convectively
plished their duties. This paper would not have been published inducedwindstorms,WeatherForecasting,2, 32-49, 1987.
without their dedicated effort. Lemon,L. R., Severethunderstorm
radaridentification
techniques
andwarningcriteria,NOAA Tech.Memo.,NWS NSSFC-3,60
REFERENCES pp., Natl. SevereStormsForecastCenter,KansasCity, Mo.,
1980.(Available
asNTIS PB81-23809
fromNatl. Tech.Inf. Serv.,
American Meteorological Society, What is a meteorologist?,Bull. Springfield, Va.)
Am. Meteorol. Soc., 72, 61, 1991a. Nolen,R. H., A radarpatternassociated
with tornadoes,
Bull.Am.
American Meteorological Society, Policy statementon the Weather Meteorol. Soc., 40, 277-279, 1959.
Service/privatesector roles, Bull. Am. Meteorol. Soc., 72,393- Stout,G. E., andF. A. Huff,RadarrecordsanIllinoistornado,Bull.
397, 1991b. Am. Meteorol. Soc., 34, 281-284,
 The "Short Fuse" Composite'An OperationalAnalysis Technique for
Tornado Forecasting
JIM JOHNSON

Weather Service Office, Dodge City, Kansas 67801

1. INTRODUCTION convection.The "short fuse" compositeuses only five of

the ADAP productsin a fast graphicalapproachto forecast
The conceptof compositeanalysisfor usein predictionof only tornadoes.It is meant to be used with, not to replace,
severestormsis far from new andprobablyreachedmaturity the ADAP decision tree.
in the late 1960sto early 1970sthrough the work of Miller
[ 1972]. A real value of the compositeanalysisis that it leads
the meteorologist through a careful analysis of the initial 2. THE "SHORT FUSE" COMPOSITE
atmospheric conditions [Doswell, 1982]. The "short fuse"
The processbeginswith a typical Miller [1972]analysisof
compositeis an attempt to extend the compositemethod to
synopticscalefeaturesat 0000UTC and 1200UTC, to which
hourly surfacedata setswith the explicit purposeof making it may be helpfulto add an analysisof storm-relativehelicity.
a tornado forecast.
Davies-Jones [1983, 1984], Davies-Jones et aI. [1990], and
In forecasting the tornado environment it is clear that no
Woodall [1990] have shown the value of helicity as a forecast
"mandatory" or "magic" numbers will suffice. Rather, it is
tool. Areas identified as favorable for significant rotating
through the interpretation and understandingof the analyzed
convection by this composite analysis are tracked hourly
fields that one comes to recognize those mesoscalefeatures
with the "short fuse" composite technique.
which are most closely associatedwith an incipienttornado. By usingthe analyzed parametersfrom ADAP, the follow-
This technique is designed for use by an operational
ing statement was seen to define a relatively small geo-
forecaster, who must function under the constraints of data
graphic area that contained a large proportion of the tornado
sets that are often incomplete due to observationand com- occurrences.
munication failures, as well as computer run-stream priori-
ties. It is possible that with the tremendous volume of new That area where the surface moisture flux convergence 1.5 g
data setsbecomingavailableto the forecaster,the saturation kg-l h-1 isopleth overlapsthe downwindsideof the axisof
point cannot be far away. Therefore any method which potential temperature advection within the plateau (axis) of
highestinstability and where the cap inversion is less than 2øC.
allows a large set of atmosphericinformationto be digested
simultaneously,especially on the mesoscalewhere the data Perhaps the most significantportion of the threat area so
density only compoundsthe problem, is worthy of consid- defined is with the overlapping of the downwind side of the
eration. axisof potential
temperature(0) advectionby the 1.5g kg-•
The idea of the "short fuse" composite grew from an h-• surfacemoistureflux convergence isopleth.It is this
attempt to refine the use of the Automation of Field Opera- region where one can anticipate the greatest low-level direc-
tions and Services (AFOS) Data Analysis Program (ADAP) tional shear or the greatest positive subcloud layer shear
to forecastonly tornadoes.Bothwell [1988]hasprovidedthe [Davies, 1989]to coincide with the strongestinferred upward
forecaster with an excellent forecast tool which takes advan- motion and sufficient moisture for significant convection.
tageof the densityof surfaceobservations,thusgivingmuch Removal of the convective cap and inclusion of the most
finer resolution in a shorter time frame (hourly). He outlines thermodynamically unstable air may add more to the chance
a systematicapproachto analysisof the 15 ADAP products of significant convection than it does to the likelihood of
that he presentsin the form of a decisiontree for forecasting tornadoes. The implication is that storms developing in or
moving into this environment would have the best chance of
The Tornado: Its Structure, Dynamics, Prediction,and Hazards.
producing tornadoes.
GeophysicalMonograph79
Thispaperis notsubjectto U.S. copyright.Published
in 1993by the The five charts for the composite and the parameters
American Geophysical Union. analyzed are as follows. (The AFOS graphics selected for
606 "SHORT FUSE" COMPOSITE

THET•
•iVECT•ON
(DEG
FHR-1*10)
0602• WT01

Theta Advection
Axes
Fig. 1. The analysisof a typical 0 advection(STA) graphicchart. The dashesshow ridges(axes) of 0 advection. It is
also helpful to mark the centroid of maximum 0 advection.

the composite will be identified according to the last three of them provided an opportunity to use the technique in
characters of the nine character group with the first six real-time forecasting.
(NMCGPH) assumed.) (1) Surface cap inversion (SSC): This technique is empirical; the derived fields of ADAP
select the 2øC isopleth, (2) Surface moisture flux conver- products were compared to a number of actual tornado
gence(SMC): selectthe 1.5 g kg-i h-1 isopleth,(3) 0 events for which the data were available, and an isopleth
advection (STA) (Figure 1): analyze the axes of maximum 0 which containedthe majority of the events was selected.
advection values greater than zero and, using the surface The results were as follows.
wind vector streamlines (SSW), locate the downwind side of
Cap inversion. Of 66 tornadoes (27 cases), all but two
the axes, (4) Surface lifted index to 500 mbar (SSL) (Figure
tornadoeswere first reportedwhere the cap was weaker than
2): selecttwo or more isoplethsthat identify the "plateaus"
2øCduring the precedinghour.
(axes) of highest instability (greatest negative values). Fi-
Surface moistureflux convergence. Of 66 tornadoes(27
nally, the parameters are transferred onto the composite
cases),all but four tornadoeswere first reported where the
map and the threat area is drawn in.
surfacemoistureflux convergence
was 1.5 g kg-• h-• or
greater during the precedinghour.
3. RESULTS OF THE CASE STUDIES
Do•vnwindsideof the 0 advectionridge. Of 66 tornadoes
Since ADAP is a locally run computerapplicationspro- (27 cases),all but two tornadoeswere first reportedon or
gram and is not archived (except occasionally on the local just downwindof the axesof maximumadvectionof poten-
level), finding the data sets collected during actual tornado tial temperatureduringthe precedinghour.
events proved to be a challenge. Twenty-seven caseswere "Plateau" (axis)of greatestinstability. Of 66 tornadoes
found with sufficientADAP data to permit analysis.Several (27cases),all butfourtornadoes werefirstreportedupon
 JOHNSON 607

($FC) LI (I)K_G C) AT 500 M• 0•0290 •Z

Fig. 2. The analysis of a typical surfacelifted index (SSL) graphic chart. Heavier lines are the isopleths selected to
define the "plateau" of greatest instability.

plateau of highest instability, with all but 14 first reported indicated at 2200 UTC May 5, 1989 (Figure 4). Two F4
inside the isopleth of the highestinstability. Again, this was tornadoes occurred during the hour (2200-2300 UTC). It is
from the hour immediately preceding the event. not known if the F4 occurring at 0001 UTC would have been
4. THREE SPECIFIC CASES forecastor not since no ADAP data were available past 2200
UTC.
An exampleof a very smallthreat area was found Novem-
ber 28, 1988, at 0600 UTC (Figure 3), the day of the Composite charts from 2100 UTC and 2300 UTC on the
Raleigh-Durham tornado. Moisture was pooling along the day of the major tornado outbreak of June 2-3, 1990, are
east side of the mountains as south-southeast wind advected given in Figures5 and 6. The initial tornadoeswere reported
the 0 axis into the moisture pool. Instability was not as on the downwind side of the strong 0 advection and on the
extreme as in other cases. This storm likely would not have downwind side of the surface moisture flux convergence
been forecastby the threat area on a purely subjectiveusage gradient [Waldstreicher, 1989] during the hour following the
of the composite,sincethe threat area did not materialize 2100 UTC composite threat area. Two hours later, the threat
until 0600 UTC, the time of the first tornado report. How- area had become elongated by southerly surface winds as
ever, the chart-to-chart changefrom the precedingseveral moistureflux convergencecontinuedto increase.During the
hoursdid show the developingand nearly stationarymois- next 2 hours, 10 tornadoes were reported. The threat area
ture pool and the 0 ridge moving northwestwardinto it. identified all but three of them, and the three misses were
In a second case, a relatively large threat area was within a few kilometers of the threat
 /

/ \ • S.F. Composite
•.• •z• valid- 11/28/88
0600 z

Fig. 3. A "threat area" is hatchedin, whilethearrow indicatesthe reportedtornado(with time andF scalestrength).
Light dashesare the plateauof instability.The heavy solidline is the 2øCcap. The heavy brokenline is the 0 axis and
thelightsolidlineisthe1.59 kg-l h-1 surface
moisture
fluxconvergence
isopleth.
The"threatarea"materialized
at
0600 UTC, the time of the first tornado report.

CAP •-r •
AVAt
I.,AE:,LE-
•

F-4 2254
/

• F-4 0001 Z

S. F. COMPOSITE
• VALID - 5/5/8•

F-4
2220
Z •
Fig. 4. Anotheroutbreakin the Carolinas,thistime witha large"threat area", The F4 tornadothat occurredat 0001
UTC onMay6, 1989,wasjustoutsidethe2-hour-old
"threatarea."Hadthecapstrength
(SSC)beenavailable,perhaps
the area would have been
 JOHNSON 609

\ )

c•

S. F. COMPOSITE
VAœID - 6/2/90
2100Z
•D
i ..

I
!

Fig. 5. Two "threat areas"are analyzed,with the northernmostbeingon the edgeof the data fieldswherethe
objectiveanalysismaybe in somedoubt.For thisreason,the tornadothat occurredjust northof Chicagoandinside
this small "threat area" was not included in the verification.

I \ 2

S. F. COMPOS I TE
VA
ZID- 6/2/90 \
2300Z
x

-.io

Fig. 6. A strongaxis of 0 advectionwas over southernIllinois at 2300UTC (seeFigure 1). The initial tornadoesfrom
2300 UTC beganalmostdirectly atop the strongestpart of the 0 axis. Analysisis as in Figure
610 "SHORT FUSE" COMPOSITE

TABLE 1. Results of Simple Verification encouraging.Spring 1991 was a particularly busy severe
1 Hour After 2 Hours After
weather season, and the technique was at least partly
CompositeTime CompositeTime responsible
for the timely issuanceof severalverifiedtor-
nadowarnings.Most importantly,the warningswere issued
Total cases 27 9
beforethe event with a much greater degreeof confidence,
Total tornadoes 66 26
Total within threat areas 56 17 anduseof the techniqueallowedprobabletornadosituations
Probabilityof detection 85% 65% to be separatedfrom probableseverethunderstormevents.

Acknowledgments.The author wishesto thank the following:

5. VERIFICATION OF THE RESULTS For assistancein collection of data, Ron Przybylinski, Jeff Wald-
Only simpleverificationwasattemptedon the composite streicher,and Dick Livingston;for encouragement and enthusiasm,
forecasts from the 27 cases studied. To relate the size of the SteveLetro, Ed Berry, andJoe Schaefer;for trying the techniquein
operations,
the staffat WSO DodgeCity, Kansas,andDon Bakerat
threat area versus the number of tornadoes within the area is WSFO Lubbock, Texas; for piquing my interest in severe weather,
difficult sincethe threat areas were usually mesosynopticto Bob Miller, Bob Maddox, and Chuck Doswell; and finally, for
subsynoptic in scaleandthe tornadoitselfmuchsmaller.For introducingme to the wonders of dynamic meteorology, Doug
thisreason,the onlycomputation performedwasprobability Sinton and Pete Lester.

of detection(POD). The POD is simply the numberof

correctlyforecasteventsdividedby the total numberof REFERENCES
events.The POD was computedusingthe numberof torna-
does that occurred both in and out of the threat area for the Bothwell, P. D., Forecastingconvection with the AFOS data
analysisprogram(ADAP-version2.0), NOAA Tech.Memo. NWS
hourimmediatelyfollowingthe compositetimeandagainfor SR-122,91 pp., Sci. Serv. Div., Natl. Weather Serv. S. Reg., Fort
the hourbeginningonehour afterthe compositetime (Table Worth, Tex., 1988.
1). Davies, J. M., On the use of shear magnitudesand hodographsin
As a control, two of the 27 caseshad no severe thunder- tornado forecasting,in Preprints, 12th Conference on Weather
storms and no tornadoes. A further six cases had severe Forecastingand Analysis,pp. 219-224, AmericanMeteorological
thunderstormsbut no tornadoes. The "short fuse" compos- Society, Boston, Mass., 1989.
Davies-Jones,R., The onset of rotation in thunderstorms, in Pre-
ite correctly indicated no tornadic activity for all of these
prints, 13th Conferenceon SevereLocal Storms, pp. 215-218,
null cases.In fact, every time a threat area developed,there American Meteorological Society, Boston, Mass., 1983.
was at least one tornado in it. Davies-Jones,R., Streamwise vorticity: The origin of updraft rota-
tion in supercellstorms,J. Atmos. Sci., 41, 2991-3006,1984.
6. SUMMARY Davies-Jones,R., D. W. Burgess,and M. Foster, Test of helicity as
a tornado forecast parameter, in Preprints, 16th Conference on
A POD of 85% for the initial hour after the compositetime Severe Local Storms, pp. 588-592, American Meteorological
is quite good, even if one considersthat ADAP requires15 Society, Boston, Mass., 1990.
min to produce the analysis. Additional cases are being Doswell, C. A., III, The operational meteorologyof convective
weather, vol. I, Operational mesoanalysis,NOAA Tech. Memo.
collected and analyzed to add more credenceto the compos-
NWS NSSFC-5, Natl. Severe Storms Forecast Center, Kansas
ite scheme presented here, but it bodes well for the tech- City, Mo., 1982.
nique that in all caseswhere the threat area did materialize, Miller, R. C., Notes on analysis and severe-storm forecasting
at least one tornado did occur within the threat area. proceduresof the Air Force Global Weather Central, Tech. Rep.
Further, the technique often identified the initial convection 200 (revision), 190pp., Air Weather Serv., Scott Air Force Base,
that subsequently became tornadic and in several cases II1., 1972.
identified very small threat areas (of the order of 2-4 Waldstreicher,J. S., A guide to utilizing moisture flux convergence
asa predictorof convection,Natl. WeatherDig., 14, 20-35, 1989.
counties) in which a lone tornado was observed.
Woodall, G. R., Qualitative forecastingof tornadic activity using
The technique presented has been in use at the weather storm-relative environmental helicity, NOAA Tech. Metno. NWS
service office in Dodge City, Kansas, for nearly six months SR-127, Sci. Serv. Div., Natl. Weather Serv. S. Reg., Fort Worth,
as of this writing, and the results have continued to be Tex.,
 The Plainfield, Illinois, Tornado of August 28, 1990'
The Evolution of Synoptic and Mesoscale Environments
WILLIAM KOROTKY

National Severe Storms Forecast Center, Kansas City, Missouri 64106

RON W. PRZYBYLINSK!

National Weather Service Forecast Offie, St. Louis, Missouri 63304

JOHN A. HART

National Weather Service Forecast Office, Charleston, West Virginia 25311

1. INTRODUCTION cyclones. Storms that do not develop this dynamical struc-

ture are referred to as nonsupercell storms.)
During the afternoon of August 28, 1990, a devastating This paper will investigate the synoptic and mesoscale
tornado ripped through the southwestern suburbs of Chi- conditions associated with the Plainfield tornadic event.
cago. The F5 intensity tornado was responsiblefor 29 deaths Thermodynamic and vertical wind shear characteristics of
and over 300 injuries and causedproperty damageexceeding the evolving tornadic environment will be examined with the
160 million dollars (with much of the destruction occurring aid of the SHARP (Skew T/Hodograph Analysis and Re-
between Plainfield and Crest Hill, Illinois). The tornadic search Program) Workstation [Hart and Korotky, 1991], a
thunderstorm produced a nearly continuous path of severe softwarepackagethat allows interactive analysisof upper air
weather for more than 4 hours as it tracked across northern soundingdata. This paper will focus on those aspects of the
Illinois into central Indiana. changing storm environment that appeared to highlight the
The environment over northern Illinois exhibited limited growing tornadic potential, with a special emphasis placed
tornadic potential at 1200UTC August 28. A combinationof on low-level storm-relative characteristics (e.g., low-level
large-scale features, including an approaching midlevel storm relative flow, magnitude and orientation of the low-
trough and a southward moving cold front (both forecast to level shear, vorticity, and helicity characteristics of the
interact with an extremely unstable airmass as strong mid storm inflow layer). Forecast implications will be discussed
and upper level winds moved over the region), indicated in an effort to increase the state of readiness for similar rare
increasingpotential for severe thunderstormdevelopment event occurrences. Finally, the study will underscore the
across northern Illinois during the afternoon. Although the important role of interactive software (e.g., the SHARP
overall setting (i.e., strong instability and relatively weak Workstation) for diagnosingenhanced convective potential
"bulk" shear through the lowest 6 km) appearedto favor in a changingenvironment.
nonsupercellstorms (with large hail and damagingwind as
the primary severe weather threat), poststormanalysisof 2. DATA
conventional radar images indicated supercellstorm signa-
tures along much of the path traversed by the tornadic The SHARP Workstation is a soundinganalysis program
storm. (Supercell stormsare used here as a generalterm to that generatesSkew T plots and hodographsfrom observed
include all convective storms that develop persistentmeso- upper air sounding data. The software also produces a
comprehensiveinventory of sounding-derived parameters
The Tornado: Its Structure, Dynamics, Prediction, and Hazards.
and allows all aspects(thermodynamic and kinematic) of the
GeophysicalMonograph79
This paperis not subjectto U.S. copyright.Publishedin 1993by the sounding to be modified in an interactive manner. SHARP
American GeophysicalUnion. was used in poststormanalysisto examine upper air sound-
612 SYNOPTIC AND MESOSCALE ENVIRONMENTS

576 • 5680564-•, -2q558 552 558

5590
5510
576
-12
578•
_26
ø
-16
5700
-21
'570
-I4

14 5790

•?•o
576

-8

-12

5910
582
5860

588 ##--1
5890 5950 l
! 4 5950 .,

•8_5931
1-4 5940
588

5900 -5 59 0
,••"• _•%• ,.
-20 -21
7 0

-1;

Fig. 1. Analysis of 500-mbarheights(,solidcontoursevery 60 m) and temperature(dashedcontoursevery 4øC) for

1200UTC August28, 1990.Stationreportsshowwindvelocity(1/2barb= 2.5m s-i, barb= 5 m s-i , triangular
barb
= 26ms- ]), temperature,
anddewpoint(degrees Celsius).
Axesof coldtroughsindicatedwithsolidtriangles.Shaded
arearepresents
windspeeds
equalto or greaterthan50 knots(25m s-i).

ingsfor stationsacrossthe Great Lakes region at 1200UTC air analyses were produced by ADAP (Automation of Field
August 28. Plainfield, Illinois (PLF), proximity soundings Operations and Services (AFOS) Data Analysis Program
were created for 1200UTC and 2000 UTC by modifyingthe [Bothwell, 1988]). The output was used in conjunctionwith
1200UTC Peoria,Illinois (PIA), Skew T and hodographwith radar, satellite, and surface analyses to examine the chang-
data representingthe estimatedthermodynamicand vertical ing thermal, moisture, stability, and wind patterns at the
wind profilesover PLF at 1200UTC and 2000 UTC. Data for surfaceprior to the tornadic event.
the PLF proximity soundingswere obtained by linearly
interpolatingin spaceand time betweenGreen Bay, Wiscon-
3. METEOROLOGICAL SETTING
sin(GRB) andPIA, using1200UTC and 0000UTC sounding
data. Surfacevaluesof temperature,dew point, and wind for The 500-mbar pattern at 1200 UTC August 28 was char-
the PLF proximity soundingswere estimated from surface acterizedby a low-amplituderidgeextendingfrom California
observations.SHARP also was usedto producean extensive to the Carolinas(Figure 1). The northernextent of the ridge
summary of thermodynamic variables, convective indices, had been erodedby a seriesof short-wavetroughsmoving
and wind-shear-relatedparametersfor selectedatmospheric eastwardacrosswestern and central Canada, with another in
layers(for boththe 1200UTC and2000UTC PLF proximity this progressionof short waves approachingthe western
soundings). GreatLakesat !200 UTC. A bandof strongwesterlywinds
Several derived fields incorporatinghourly surfaceobser- associated
with the approachingtroughstretchedfrom Mon-
vations and the 1200UTC August28, 1990, regionalupper tanato Minnesota,
witha 60-knot(31m s-2)jet core
 KOROTKY ET AL. 613

153 144 14' 138

i!
,J i

/ I 1$4
1

428

16 15;

/
1457 .4

1,•55

I ',:'2 17
150 147
o?16

150
/

20--
20 1528 i
% " ,58
-1 1557
i
18 t
80=,, , 22 1548 153

.,%

Fig. 2. Analysis of 850 mbar heights(solid contoursevery 30 m) and temperature(dashed contours every 4øC) for
1200UTC August28, 1990.Stationreportsshowwindvelocity(1/2barb= 2.5 m s-], barb= 5 m s-I), temperature,
and dew point (degreesCelsius). Shadedarea representsdew points equal to or greater than 14øC.

into Minnesota from the Dakotas. A difiuent pattern was lower Great Lakes regionto the southernPlains, with highest
present over the western Great Lakes region at 250 mbar dew points in the vicinity of the front near GRB.
(notshown),with strongwindsaround80-knots
(41m s-•) Soundingsfor stationsacrossthe Great Lakes region were
over northern Illinois. The approaching trough caused a characterizedby very steep lapse rates (especially from
northwesterly component to develop in the upper flow over ---800 to -600 mbar) and considerable convective available
the Great Lakes region and allowed a surface cold front to potential energy (CAPE). A thermal cap near 850 mbar
move southeastward across northern Illinois during the (evident in the 1200UTC August 28 PLF sounding,Figure 3)
afternoon of August 28. For the 12 hours ending at 1200 was strongenough to prevent large-scale overturning of the
UTC, height falls and cold advection associatedwith the atmosphereby widespreadbut ordinary (nonsevere)convec-
changingupper flow pattern caused 500-mbar temperatures tion, therebyallowingheat and moistureto increasesteadily
to decrease from -4øC to -7øC at PIA and from -8øC to acrossnorthern Illinois prior to the arrival of the cold front.
-12øC at G RB. However, the area of negative buoyancy below the level of
The 850-mbar chart for 1200 UTC August 28 revealed a free convection(LFC) was not impenetrable;the projected
significantthermalgradientacrossthe upperMississippiValley effectsof frontalforcingand diurnalheatingappearedstrong
and western Great Lakes, associatedwith a cold front moving enoughto allow isolated deep convection during the after-
throughthe region(Figure2). Northwesterlywindsat GRB and noon. The 1200UTC PLF soundingindicated a "best" lifted
a decreasein temperature/dewpoint duringthe previous 12- index (liftinga parcelcontainingthe highestwet bulb poten-
hour periodindicatethat the front had passedGRB just priorto tial temperature within the lowest 150 mbar; i.e., from
1200 UTC. An axis of higherdew pointswas noted from the approximately1200feet (365 m) above groundlevel
614 SYNOPTIC AND MESOSCALE ENVIRONMENTS

...... / •/28/98 /
LI ......
TT......
-8
59 X•ODIFIEB..
TEl ..... 3•,?
1(....... 32 ,<
SXEtlT... 234 /-

ClIP
.....
PtlRCEL!•TA

B+
.... 3559
B-.... 33

LPL,. 958W-,,
' EL,,. 46
MPL.. -999ft

LCL,. 4225œ
FZL., 14168œt
I•B2,, 13•74{t
I

1•i... !.56in

58

?6

•5

198
185
-18

Fig. 3. PLF proximity soundingfor 1200UTC August 28, 1990,as displayedby the SHARP Workstation. CAPE is
representedby B + and depictedas the area betweenthe parcelascentcurve (dottedcontour) and the temperaturecurve
(from the level of free convectionto the equilibriumlevel (EL)). See Hart and Korotky [ 1991] for a thorough discussion
of display characteristics,terms, and units associatedwith the Skew T presentation.

of -8 at 500mbarwithCAPEevaluated
near3550J kg-] . It 4. ANALYSIS
is noteworthy that the sounding indicated even greater Visible satellite imagery (Figure 5) and surface analysis
instability above 500 mbar, with a best lifted index of - 12 at
(Figure 6) revealed a complex pattern of features across
300 mbar and -10 at 200 mbar. Thus the 1200 UTC PLF
northern Illinois at 1900 UTC. The most important features
soundingindicatedextreme instability, even without the aid
included a cold front across northern Illinois (Figure 6), a
of diurnalheating.With projectedafternoontemperaturesin
the mid 90s (degreesFahrenheit) (mid 30s degreesCelsius) gradual wind shift line south of the cold front (confirmed
and projecteddewpointsin the mid 70s (degreesFahrenheit) with time sections of surface data, not shown), and a
(mid 20s degreesCelsius) the thermodynamicstructure of moisture convergence gradient across north central Illinois
the 1200 UTC PLF soundingdisplayed a potential for (Figure 7). Surface dew points ranged from the mid 70s
uncommonly large instability acrossnorthern Illinois before (degrees Fahrenheit) (mid 20s degrees Celsius) to around
the forecastarrival of the cold front (especiallyfor a surface- 80øF (27øC) across much of north central Illinois. However,
based parcel). Kinematic analysis of the PLF sounding dew points were around 70øF (20øC) even north of the cold
indicated considerable speed shear in the vertical wind front, with the true moisture gradient still in Wisconsin.
profile between 600 and 500 mbar, but the strongestwinds More subtle features revealed by satellite imagery included
were near and above 400 mbar (approximately 7 kin). several converging lines of cumulus across north central
Moderate speed and curvature shear was noted within the Illinois (Figure 5), but these features were difficult to resolve
first kilometer, but the wind shear pattern was rather weak in the surface analysis. The ADAP lifted index pattern
abovethis layer to around 600 mbar (Figures 3 and 4). A bulk (Figure 7) depicted extreme instability across the lower
measure of theshearthrough6 km (19 m2 s-2, usedin the Great Lakes region, with surface-basedlifted indices be-
calculation of the Bulk Richardson Number (BRN) [Weis- tween -12 and -15 across much of Wisconsin, Illinois,
man and Klemp, !982]) was rather weak relative to the large Indiana,andMichigan.Operationalexperiencesuggests that
instability, yielding a large BRN, as we shall discussbelow. instabilityof this magnituderarely occursover sucha
 KOROTKY ET AL. 615

PLF I 1•t28199 I 12Z '

l•an Hin•l •I)IFII•
9-3•,, 272/15
36

Positive SM•
8-2•,. 4,8
•3•,, 3.2

SR
•elicit•57
•-2•.,
,19

[HSL(•)
l)ii,/kt
2{tl
59{t
1868 259/
220/ 9
239/ 29
22
I
I

15• 278/14
2{t8{t 289/16
259• 39• 1•
3• 3•/19
35• 3• 19
• 298/13
45• 3• 12
5• 29• 13
55• 283/22
6• 282/32
65• 287/36 36
7• 288/•

- •

SFC- 70•1 •
' 59

Fig. 4. PLF proximity hodographfor 1200UTC August28, 1990, as displayedby the SHARP Workstation. Units
alongaxes in knots. Data pointson hodographare every 500 m. See Hart and Korott•7 [ 1991]for a thoroughdiscussion
of display characteristics,terms, and units associatedwith the hodographpresentation.

geographicalareal The incipient Plainfield storm intensified measured cloud bases (3000 feet (920 m) AGL) and the lifted
rapidly in the vicinity of the cold front near Rockford, condensation level (LCL, 3100 feet (945 m) AGL) calculated
Illinois (RFD) between 1800 and 1900 UTC. One or two by lifting a surfaceparcel associatedwith the 2000 UTC PLF
small F1 tornadoes were reportedjust west of RFD at 1842 sounding.
UTC, with damagingwind and large hail reported at RFD The PLF proximity hodograph for 2000 UTC (Figure 9)
between 1900 and 1945 UTC [National Oceanic and Atmo- reflected the estimated vertical wind profile, representative
spheric Administration, 1991]. surfacewind (300ø/5knots(2 m s-l)) and observedstorm
The PLF proximity soundingfor 2000 UTC (Figure 8) motion(fromradarobservations, 308ø/32knots(16 m
reflected an estimatedsurfacetemperature(92øF)(33øC)and near Plainfield just prior to the tornadic phase of the storm.
dew point (78øF) (25øC) consistent with the low-level air In addition to producing thermodynamic parameters,
most likely entrained into the storm as it approachedPlain- SHARP was used to calculate the storm-relative, vertically
field. The soundingrevealedthat importantthermodynamic integrated helicity (hereafter referred to as helicity) and the
changeshad occurred acrossnorthern Illinois between 1200 low-level storm-relative flow structure of a storm moving
UTC and 2000 UTC, with CAPE (now representing the through the estimated PLF environment with the observed
surfaceparcel)increasingfromunder4000J kg-I (1200 Plainfield stormmotion.A computed helicityof 165m2 s-2
UTC) to almost7000J kg-1 (2000UTC). It shouldbenoted through the lowest kilometer (AGL) of the PLF sounding
that we could have determined CAPE'for a parcel represent- was within the 0- to 3-km threshold for weak mesocyclone
ing the mean thermal and moisture propertiesof the lowest development suggestedby Davies-Jones et al. [1990] but less
100 mbar of the sounding(e.g., the lifted index usedby the than the 0- to 3-km threshold established by Lazarus and
Severe Local Storms Unit of the National Severe Storms Droegetneier [1990]. However, the storm-relative inflow
Forecast Center), in which case CAPE would have been (resultinglargely from storm motion) was quite significant
around4200J kg-l at 2000UTC. However,we feltthatthe (28knots(14m s-•)) through
thislayer.
air being lifted into the updraft by a fast-movinggustfront By 2100 UTC the Plainfield tornado had just lifted, but the
was probably related to properties much closer to the parent thunderstormwas still producing damagingwind and
surface.This is supportedby the near equivalencebetween large hail as it continued across northeastern Illinois.
616 SYNOPTIC ,•qD MESOSCALE
 KOROTKY ET AL. 617

12 10 08
80 131

_•o•
TM
•50

90•1'18
59 •
)6

83

6,? 080 89

12-- •o•,,.,, ss/ 10

)7

87 111

95
92 12

59 $20

99 09g 112
%%

99 109 89 136

85•47/
131

85 '•56-""'
12

Fig. 6. Subjective surfaceanalysisfor 1900UTC August28, 1990. Solid contoursare isobars(12 = 1012 mbar). Fronts
indicated with the usual symbols. Large dashedcontoursare boundaries.Station reports show wind velocity (1/2 barb
= 2.5m s-1 , fullbarb= 5 m s-l), temperature,
anddewpoint(degrees
Fahrenheit).
Smallerdashedcontours
represent
76øF and 80øF isodrosotherms. Shaded area representsdew points equal to or greater than 80øF.

face analysis(Figure 10) and PPI radar film (not shown)used cated reflectivity features commonly associatedwith super-
in conjunction with a sequenceof radar tracings(Figure 11) cell storms between 1900 UTC and 2100 UTC (e.g., fight
indicated a convectively generated mesohighwith rapidly rear appendage with associated high reflectivity, inflow
expanding outflow along its leading edge. The storm- notch characterized by a strong low-level reflectivity gradi-
produced mesohighand gust front are further supportedby ent on the updraft flank of the storm, pronouncedweak echo
satelliteimagery(Figure 5) and by the surfacemoistureflux region with a periodic bounded weak echo region, etc.).
convergence/divergencepattern at 2100 UTC (Figure 12). Although the tornadic storm appeared to have interacted
Time sections of surface observations (not shown) indicated with several converging lines of cumulus (noted earlier) as it
a gradual northwesterlyshift in the surfacewinds aheadof approachedPlainfield, the operational surface observational
the fast-movinggustfront (Figure 11). This suggeststhat the network is not sufficiently dense to resolve the details.
ground-relative surface wind field of the mesoscaleenviron-
ment was not enhancingconvergenceand inflow. Rather, the 5. DISCUSSION
movement of the storm and its attendant outflow created
significantlow-level convergenceand inflow despite a north- Observations and numerical simulations [Weisman and
erly componentto the surfacewinds just ahead of the gust Klemp, 1982] indicate that the dynamic organization of
front! storms is strongly influenced by instability and by the
Analysis of conventionalradar images (Figure 11) [Na- vertical structure of the horizontal wind. Weak vertical wind
tional Oceanic and AttnosphericAdministration, 1991]indi- shear supports limited convective scale organization
618 SYNOPTIC AND MESOSCALE ENVIRONMENTS

-12 -14

I
I
I

-14

lO
15
•'25

-12
15
10

Fig. 7. Automationof Field Operationsand ServicesData AnalysisProgram(ADAP) surfacemoistureflux

convergence
(solidcontours
depictg kg-• h-• x 10),ADAPliftedindexto 500mbar(dashed
contours
in degrees
Celsius),and surfacecold front (with usualsymbols)for 1900UTC August28, 1990.Large dashedcontoursrepresent
surface boundaries. Shaded area indicates active convection.

generally leads to loosely organized nonsupercellstorms. but a bulk measureof the shearthrough 6 km (i.e., the vector
However, strongershearfavors progressivelybetter organi- difference between the 0- to 6-km AGL density-weighted
zation (well-organized nonsupercellsand supercellstorms) meanwind and the 0- to 500-m density-weightedmean wind)
by maximizing boundary layer convergenceand by promot- wasweak(19m2 s-2 [Weisman
andK!emp,1982])relative
ing dynamicalprocessesthat result in stormrotation. to the large CAPE. The resulting high BRN for PLF (189)
The BRN [Weisman and Klemp, 1982] has been used to suggesteda tendencyfor nonsupercellstormsand indicated
measure the relative importance of instability and vertical a severe weather threat primarily involving damaging
wind shear for a particular environment and generally is straight-linewind and large hail. While the BRN apparently
associated closely with observed storm type (nonsupercell failed to forecast the potential for supercell growth within
or supercellstorms)for modeled values of CAPE between this environment, recall that Weisman and Klemp did not
1000and 3500J kg-•. For this rangeof CAPE, numerical modelenvironments
whereCAPEexceeded
3500J kg-• . In
simulations and a limited number of storm observations fact, high-CAPE environments have yet to be modeled
suggesta tendencyfor supercellstormsto form when the successfully(H. E. Brooks, personalcommunication, 1991).
BRN is less than 45. On the other hand, BRN values By 2000 UTC the PLF soundingindicated CAPE approach-
exceeding 45 are generally associatedwith nonsupercell ing7000J kg-• (Figure8), withan associated
BRN of 234.
convection. Recently, much interest has focused on the vertical wind
The Plainfield environment exhibited substantial instabil- shearstructureof the low-level storm environment, specifi-
ity at 1200UTC(i.e.,CAPEaround
3550J kg-•, Figure3), cally on the region associated with entrainment into
 KOROTKY ET AL. 619

19 / 98/28/9g /
LI ...... -12 \
T! ...... 59
TEl ..... 47.8 .,,,
R....... 32
$14•T,, ,239

,,%
C•P..... 1.I

B+.... 6989
B-.,.. 18

LPL,. 989.b
EL,.. 494•f t
NPL,, -999ft

LCL,. 3231[t
FZL..14160[t
•B'Z,. 12241f t
/

.f

/ : ,/
?0 '

Fig. 8. Sameas Figure 3 for 2000 UTC August28, 1990.

updraft [Brooksand Wilhelmson,1990;Davies, 1989;Davies- inflow and playsan importantrole in determiningthe poten-
Jones et al., 1990; Johns et al., this volume; Lazarus and tial for the updraftto rotate [Da•,ies-Jones,1984].Helicity is
Droegemeier,1990].Shearcharacteristics withinthis inflow related to streamwisevorticity [Doswell, 1991] and measures
region appearto play a major role in determiningboth the the rotationpotentialthat can be realized by a storm moving
potential for a storm to develop supercellcharacteristics throughthe verticallyshearedenvironment(i.e., streamwise
(i.e., persistentmesocyclone)
and the potentialfor a partic- vorticity associatedwith the inflow layer can be drawn into
ular supercellto developa tornadiclow-level circulation. and correlatedwith the updraft core to produce a persistent
Numerical simulations[Brooksand Wilhelmson, 1990] indi- mesocyclone).Since the storm referenceframe is physically
cate that midlevel rotation does not guarantee a tornadic important, the actual movement of a storm can exert a
low-level mesocyclone.Lazarus and Droegemeier[1990] significantinfluenceon these characteristics.Consequently,
found that sufficienthelicityin the low-levelenvironmentof storms that "experience" helical low-level environments
simulatedstormsresultsin strongstormrotationbut suggested favor the persistentrotating updrafts characteristicof super-
that rotation and storm longevityalso dependon sufficient cell storms and represent the greatest tornadic threat.
low-level storm inflow(i.e.,greater than19knots(10m s-1) Helicity values between 150 [Davies-Jones et al., 1990]
[Davies-Jones,1984].The helicityand inflow structureof the and250m2 s-2 [LazarusandDroegemeier, 1990]withinthe
estimated Plainfield storm environment will be discussed in 0- 3-km layer appear to rnark a transition region between
greaterdetail;however,it canbe notedthat the flow into the nonsupercell and supercell storms (given that adequate
developing tornadicstorm(resulting largelyfromstormmotion CAPE and forcing are also present). However, Davies and
characteristics)
was rathersignificant(around30 knots(15 m Johns [this volume] found that much of the 0- to 3-km
s-l)) withinthelowestkilometer(AGL). helicity associatedwith an extensive set of modified tornado
proximity soundingsoccurred within the lowest 2 km. We
suggestthat the relevantlayer for determiningthe significance
5.1. Characteristicsof the Storm-Relative Environment
of inflow propertiesshouldbe the layer associatedwith entrain-
For a givenstormmotion,streamwisevorticityrefersto ment into the updraft, which is probably related to the LFC for
the componentof horizontalvorticity parallelto the storm a given storm environment[Bluestein et al., 1989].
620 SYNOPTICAND MESOSCALEENVIRONMENTS

?LF ! 98/28/98 ! 2•
WODIFIED
O-3k•.. 270/18
•-6k•.. 283/23 24 898--+--278{

Positive She• 368 ,

0-21•.. 4.1
9-3•.. 4,4 16

gR {{elici
8-2k•.,
tt47
8-3k•.. 158

Sto• • ti'on
388/ 32

INFLO•

•L(•) Di•/kt
8 129/27
299 165/31
799 174/26
1299 165/21
1799 168/18
2299 169/14
2799 171/18

W•an Inœ1ow
8-2k•.,,159/24 24
8-3k•...161/21
St•ea•ise
•-2k•,,,159/14
8-3k•,.,16!/18 32

Fig. 9. SameasFigure4 for 2000UTC, August28, 1990.

ingly,estimated
inflowcharacteristics the motionat 75% of the meanwind speedand 30øto the right of
for thelayerbeneath
LFC were determinedfrom the 2000 UTC PLF sounding. themeanwinddirection,
318ø/34
knots(17m s- •)).
For this case, the movement of the developing storm For the Plainfield case it appearsthat the 0- 6-km mean
appearedto exert the strongestinfluenceon the low-level wind was unduly affectedby the lower troposphere,where
storm environment. While the topic of storm motion is wind fields were relatively light. Given the extraordinary
beyondthe scopeof this paper, somegeneralcommentscan heightattainedby the Plainfieldstorm(over65,000ft, or 19.8
be made. The Plainfieldstormappearedto deviatefrom the km) and the presenceof strongerwindsabove400 mbar, it
mean steeringflow. However, the amountof this deviation seemsplausible thata deeper(cloudlayer)meanwindwould
(both direction and magnitude)dependson how "steering t•etterrepresentthe steeringflow for this environment.In
flow" is defined. Steeringflow has been representedin the any case,stormmotioncan be influencedby a numberof
research community and operationally by the mean flow additionalfactors, including the dynamical structure (and
throughthe approximatecloud-bearinglayer but alsoby the evolution)of the storm and external features (e.g., bound-
0- to 6-km density-weightedmeanwind. Since the observed aries, etc.) with which the storm may interact [Doswell et
stormmotion(308ø/32 knots(16 m s-i)) was significantlyat., 1990]. Since the relative importanceof factors affecting
faster than the estimated 0- to 6-km mean wind over PLF at stormmotionfor a particularenvironmentare still unclear,
2000UTC (283ø/23 knots(12m s-•)), it mightseemreason- we suggestthat the important issue (from an operational
able to suggestthat the actual vertical wind profile in the point of view) is not the deviation of a storm from some
environment of the tornadic storm was stronger than the perceivedsteeringflow (aside,perhapsfrom the association
interpolatedvertical wind profile for 2000 UTC. However, of supercell characteristicswith apparent deviant storm
anotherpossibilityis that the 0- 6-km layer wastoo shallowto motion). Rather, we believe it is more important operation-
representa true steeringflow for the Plainfieldstorm.A deep ally to recognizethe impactthat an observedstorm motion
layer meanwind (approximatingthe depthof the cloudlayer) can have on the evolution of possible storm type. Since
overPLF at 2000UTC (288ø/45 knots(23m s-I)) wasmuch storminflow characteristicsfor a given environmentcan be
closerto the reportedstorm motion, especiallyafter applica- maximizedwithin a certain rangeof possiblestorm motions
tion of an adjustmentfactor [Leftwich, 1990] representing (as they were with the Plainfield storm), it is important to
observed mean deviant supercellstorm motions (i.e., storm recognizethe limits of this range. Given a reasonably
 KOROTKY ET AL. 621

6,•.G080124
12 08

2 125

83•'126 62
117 loo

70 81

90Li
885
82

87J•
o
09,½
-I
•o 1 81

70 04 6
057
12

66• 72 ,,

O6
10
76 --
91
87 083 87 08U

lO
98
97
90 105 89 102

93 104
1 oo 089 102

7 '"" 90 9 3
'" 12
108
97 111 96 1 ;0 6

8 92 123
98
10
,g6 114 91 118
99 112
89 134

121 97 125

101 91

lol 114
•"•-• 12

Fig. 10. Sameas Figure6 for 2100UTC August28, 1990.Largedashedcontours

alsoshowthe outflowboundary
(northeastern Illinois) associated with the Plainfield storm.

mated vertical wind profile, the interactive features of the Of course,the helicity and shear values presentedhere
SHARP Workstationcan be used to establisha range of represent only an approximation, using data from widely
storm motionsrepresentingthe greatestthreat for supercell spacedupperair sites.The potentialfor supercelldevelop-
development. If the observed storm motion associatedwith mentwithin the environment"experienced" by the moving
developingconvectionfalls within this "higher threat" range Plainfieldstormmay have been enhancedconsiderablyby
of motions,and additionalanalysessupporta growingsevere storm scale processes involving intense convection and/or
weather threat, the storm environment may be changingto by the interaction of intense convection with external fea-
(or may have changedto) a statefavoring supercellgrowth. tures (e.g., boundaries,etc.). Still, it is noteworthy that
The 2000 UTC August 28 PLF soundingindicated mar- computedhelicityandinflowthroughthe layer representing
ginalhelicity(165 m2 S-2) from the surfaceto the LFC the updraft entrainmentregion of the Plainfield storm were
(approximately0-1 km). However, we are comparing0- to within a range compatible with storm rotation.
1-km helicity values with 0- to 3-km rotation thresholds.
Since we are measuringthese "marginal" values over a
relatively shallow inflow layer, it is possible that they 5.2. CAPE and Storm-RelativeHelicity
represent more than marginal potential. Lazarus and Although strong mesocyclonesoften are associatedwith
Droegemeier [1990] suggestedthat storm rotation and lon- largevaluesof helicity,there appearsto be a relationship
gevity depend not only on he!icity but also on the specific between helicity and CAPE that contributes to the formation
combinationof storm inflow and shearvorticity producingthe of stronglyrotatingupdraftsfor any particularenvironment
hellcity.Accordingly, the meaninflow(around15 m s-l) [Lazarus and Droegemeier, 1990]. As a result, intense
within the first kilometerappearsto be highlysignificant. rotatingupdraftscanformwith relativelyweakinstability
622 SYNOPTIC AND MESOSCALE ENVIRONMENTS

1948

• 2002

CGX ccx

Jo
I

Fig.11. Tracings
ofthereflectivity
patterns
fromtheMarseilles,
Illinois
(MMO)PPIfilmfor(a) 1948UTC,(b)2002
UTC,(c)2021UTC,and(d)2034UTC,August 28,1990.Allpanels represent
VIPlevels 2-6atanelevationangle
of
0.5ø[after
National
OceanicandAtmosphericAdministration,
1990].Boldcontour
indicates
tornadotrack.Stations
in
Figure1l a show
windvelocity
(1/2barb= 2.5ms-l, fullbarb= 5 ms-l). Gustfrontassociated
withthetornadic
Plainfieldstormis depictedby contouralternating
betweendashesandtwo dots.

by stronginflow.On the low-level, storminflow may have been sufficientto foster

highvaluesofhelicityare supported
other hand, marginalhelicity still appearsto supportstorm supercell development
withinthe marginally
shearedPlain-
rotation if CAPE and inflow characteristicsare sufficiently field environment.
large.Johnset al. [thisvolume]examinedtornadoproximity
soundingsfor an extensivedata set involvingstrongand 5.3. The Supercell Environment
violent tornado occurrences to observe the relationship
between low-level helicity and CAPE. The distributionof TornadicsupercelIscan develop in a wide variety of
data presentedin that study indicatesthat progressively environments,some of which do not exhibit "clear-cut"
stronger(weaker)0- to 2-km helicity valuesare associated supercellpotential(i.e., highlybaroclinicsystemswith as-
with strongand violent tornadodevelopmentas the CAPE sociatedstrongwind shear;seeDoswell et al. [1990] for a
decreases(increases).Further, for any particularvalue of discussion of supercellenvironments).Of course,the deter-
CAPE there appearsto be a rangeof helicity valuescom- mination of low-level inflow characteristics is more reliable
patible with the initiation of strongand violenttornadoes. when significantbaroclinicitycontributesstrongvertical
Accordingly,the Johns et al. [this volume]data supporta wind shear through a deep atmosphericlayer (because
possiblerelationshipbetweenCAPE and helicity.The pre- strongsynopticscale features can often be observedby
tornadic environment in the vicinity of PLF exhibited mar- operational surface and upper air networks). However,
ginalhelicity through thestorminflowlayer(165m2 s-2), tornadicsupercelIsalsocandevelopwhenthe synopticscale
but the thermodynamiccontributionof instabilitywas ex- indicatesweak or marginalshear, especiallyif the environ-
traordinary
(possibly
near7000Jkg-•). Onthebasis
ofthese ment is characterizedby anomalouslyhigh CAPE. Under
findings,it appearsthat high values of CAPE and strong such conditions,dynamical processesassociatedwith
 KOROTKY ET AL 623

+
RFD

5
10
15

10
10
20
5 30

20 15

-10 -15 -15 0

Fig. 12. Same as Figure 7 for 2100 UTC August 28, 1990. Lifted index is not included.

tense convection (e.g., storm-scale rotation, vertical pres- instability. Consequently, the "measurable" environment
sure gradient accelerations, etc.) and external features (e.g., favored nonsupercell storms, and the main threat appeared
surface pressure troughs, thermal/moisture boundaries, to be from large hail and damaging straight-line wind.
wind shift/convergence lines, fronts, etc.) with which a However, the environment changed significantly after 1200
developingstorm may interact can greatly enhancetornadic UTC, withCAPEincreasing
to almost7000J kg-! priorto
potential on the storm scale (see Doswell et al. [1990] for a the tornadic event. The vertical wind shear also increased
discussionof internal and external storm processes;also see during this period, although the amount of increase remains
Maddox et al. [1980] for a discussion involving thunder- uncertain. Of course, VAD (velocity azimuth display) wind
storm/thermal boundary interactions), where critical infor- profiles and wind profilers would have provided critical
mation is often missing.Consequently,mesoscaleenviron- prestorm information regarding important changes in the
ments that can support supercell characteristics will vertical wind profile across the region affected by the Plain-
continue to escape detection (especiallywhen significant field storm.
low-level shear is localized).
Satellite imagery, surface observations, and ADAP fields
revealed a complex pattern of surface features across north-
5.4. Summary of the Plainfield Enviromnent ern illinois prior to the tornado. The intensifying tornadic
The prestormenvironmentnear Plainfieldsuggested lim- storm appeared to interact with several converging lines of
ited tornadic potential at 1200 UTC on August 28. The cumulusas it approachedPlainfield, but the observing net-
large-scalepatternindicateda potentialfor severethunder- work was not sufficiently dense to resolve any specific
storm development,but a "bulk" measureof the 0- 6-km cell/boundaryinteractions. The most significant "measur-
bulk was rather weak relative to the large thermodynamic able" influence on the evolution of the low-level
624 SYNOPTIC AND MESOSCALE ENVIRONMENTS

environment was related to characteristics of the observed and to Brian Smith, NSSFC, for providing satellite and radar
storm motion (and associatedgust front), which resultedin picturesof the event. Additionalthanks to the editorsfor patience
and to Deborah Haynes for initial manuscriptpreparations.
strong low-level inflow and significant low-level vorticity
(despitea northerly componentto the surfacewind field of
REFERENCES
the mesoscaleenvironment ahead of the gust front). Al-
though0- to 3-km helicity was only marginallycompatible Bluestein,H. B., E. W. McCaul, Jr., G. P. Byrd, G. R. Woodall, G.
with stormrotationprior to the tornado,it appearsthat high Martin, S. Keighton,and L. C. Showell, Mobile soundingobser-
valuesof CAPE and stronginflow characteristicswithin the vations of a thunderstorm near the dryline: The Gruver, Texas
stormcomplexof 25 May 1987,Mort. Weather Rev., 117, 244-250,
lowestkilometermay have been sufficientto promotesuper- 1989.
cell development. Finally, the tornadic potential of the Bothwell, P. D., Forecasting convection with the AFOS data
developing storm may have been enhanced by storm scale analysisprogram(ADAP-version 2.0), NOAA Tech. Metno. NWS
processesinvolving the internal properties associatedwith SR-I22, 91 pp., Sci. Serv. Div., Natl. Weather Serv., S. Reg.,
Fort Worth, Tex., 1988.
intense convection and/or by the interaction of intense Brooks, H. E., and R. B. Wilhelmson, The effects of low-level
convection with external environmental features.
hodographcurvature on supercell structure, in Preprints, 16th
Conferenceon Severe Local Storms, pp. 34-39, American Mete-
6. FORECAST IMPLICATIONS orological Society, Boston, Mass., 1990.
Davies, J. M., On the use of shear magnitudesand hodographsin
The central theme of this paper emphasizesthe impor- tornado forecasting, in Preprints, 12th Conference on Weather
tance of storm-relative inflow characteristics,especiallyfor Forecastingand Analysis,pp. 219-224, American Meteorological
Society, Boston, Mass., 1989.
large-scaleenvironments not clearly indicatingtornadic po- Davies, J. M., and R. H. Johns,Some wind and instability param-
tential. Sincepropertiesof the inflow are largely determined eters associatedwith strongand violent tornadoes, 1, Wind shear
by the movement of a storm through the vertically sheared and helicity, this volume.
environment,it is important to evaluate the range of storm Davies-Jones,R. P., Streamwisevorticity: The origin of updraft
motions that will maximize low-level inflow characteristics rotation in supercell storms, J. Atmos. Sci., 41, 2991-3006, 1984.
Davies-Jones, R. P., and D. W. Burgess, and M. Foster, Test of
for a given environmentand to recognizethe impact of a helicity as a tornado forecast parameter, in Preprints, 16th Con-
given stormmotion on the evolutionof possiblestormtype ference on Severe Local Storms, pp. 588-592, American Meteo-
for that environment. rological Society, Boston, Mass., 1990.
Supercellstormscan developin a wide variety of environ- Doswell,C. A., II1, A reviewfor forecasterson the applicationof
hodographsto forecasting severe thunderstorms, Natl. Weather
ments(someof which are not obviously supercellsituations). Dig., 16, 2-16, 1991.
As a result,tornadicsupercelldevelopmentmay still be possi- Doswell,C. A., III, A. R. Moller, and R. W. Przybylinski, A unified
ble when large-scaleconditionsindicateweak or marginal set of conceptualmodelsfor variationson the supercelltheme, in
shear,especiallyif the environmentis characterizedby anom- Preprints, 16th Conferenceon Severe Local Storms, pp. 40-45,
alouslyhigh CAPE. Since the processescontributingto the American MeteorologicalSociety, Boston, Mass., 1990.
Hart, J. A., and W. D. Korotky, The SHARP Workstation version
developmentof tornadic supercelIsin high-CAPE environ- 1.5,A Skew-t/Hodograph Analysisand ResearchProgramfor the
ments are not well understood, it would seem reasonable to IBM and compatiblePC, 58 pp., NOAA Eastern Reg. Cornput.
view suchenvironmentswith an extra measureof caution, Programs, Natl. Weather Serv., Bohemia, NY, N.Y., 1991.
especiallyif cel//boundaryinteractionsare possible. Johns, R. H., J. M. Davies, and P. W. Leftwich, Some wind and
Finally, if the large-scalepatternindicatesa potentialfor instabilityparametersassociatedwith strongand violent torna-
does, 2, Variationsin the combinationsof wind and instability
severethunderstormdevelopmentin a high-CAPE environ- parameters, this volume.
ment,the mesoscale andstormscalepossibilitiesbecomevery Lazarus,S. M., andK. K. Droegemeier,The influenceof helicityon
important.Althoughstormscaleprocesses andinteractionsare the stabilityand morphologyof numericallysimulatedstorms,in
not resolvedby conventionalobservingnetworks,forecasters Preprints,I6th Conferenceon SevereLocal Storms,pp. 269-274,
AmericanMeteorologicalSociety, Boston, Mass., 1990.
must recognizethe potentialwithin a particularenvironment Leftwich,P. W., Jr., On the useof hellcityin operationalassess-
for thesestorm scaleprocesses to occur.If the large-scale mentof severelocal stormpotential,in Preprints,I6th Confer-
environment shows less than favorable conditions for the ence on Severe Local Storms, pp. 306-310, American Meteoro-
developmentof tornadic supercelIs,an awarenessof storm logicalSociety, Boston, Mass., 1990.
for important Maddox,R. A., L. R. Hoxit, and C. F. Chappell,A studyof
scalepotentialcanincreasethe stateof readiness
tornadic thunderstorminteractions with thermal boundaries,
radar signatures,spotterreports,etc. Mon. WeatherRev., I09, 171-180, 1980.
NationalOceanicand AtmosphericAdministration,The Plainfield/
Crest Hill Tornado,NorthernIllinois, August28, 1990,Nat.
Acknowledgments. The authors wish to thank Bob Johns and DisasterSurv.Rep., 119pp., Natl. WeatherServ., Silver Spring,
SteveWeiss,NSSFC, for carefullyreviewingthismanuscript
and Md., 1991.
offering many insightful suggestions.Thanks also to CharlesA. Weisman,M. L., andJ. B. Klemp,The dependenceof numerically
Doswell III for guidance; to Jeff Waldstreicher, NWS Eastern simulatedconvectivestormson verticalwind shearand buoy-
RegionScientificServicesDivision,for reviewingour presentation; ancy, Mon. WeatherRev., 110, 504-520,
 Characteristics of East Central Florida Tornado Environments
BARTLETT C. HAGEMEYER AND GARY K. SCHMOCKER

National WeatherServiceOffice,Melbourne,Florida 32935

1. INTRODUCTION storms in the wet season. Wakimoto and Wilson [1989] have
proposeda model to explain nonsupercelltornado develop-
The diagnosisof the dynamicand thermodynamic struc- ment in which low-level boundaries play a crucial role in
ture of the environment in which tornadic thunderstorms tornadogenesis.This theory may have important applica-
develop, as well as potential mechanismsto initiate intense tions in central Florida.
convection,are crucialto severestormforecasting.Overthe The tracking of boundaries with Doppler radar, satellite,
years many researchers such as Bebbe [1958], Darkow and mesonetdata aids in the short-term prediction of poten-
[ 1969],and TaylorandDarkow[1982]haveusedmeanupper tially tornadic thunderstorms, but few boundaries, or bound-
air sounding data to investigate the structure of the atmo- ary intersections, actually result in tornadic development
sphere in proximity to tornadic thunderstorms.Recent stud- [Holle and Maier, 1980]! More information concerning the
ies indicaterenewedinterestin the useof tornadoproximity characteristicsof the overlying dynamic and thermodynamic
soundingsto relate thermodynamicand dynamicvariablesto structure in these wet season situations is needed if forecast-
tornado intensity [Riley and Colquhoun,1990;Johnset al., ers are to have much successin assessingthe tornado threat.
1990] and the examination of the structural characteristics Over all of central Florida most of the strong and violent
and evolutionof differenttypes of tornadoproximitysound- tornadoes (F2-F5 [Fujita, 1981]) occur in the dry season,
ings [Schaefer and Livingston, 1988, 1990]. and a majority of these occur in the morning [Schmocker et
Characteristically, these studies combine data collected al., 1990]. A more complete understanding of the environ-
from a large area of the country and contain little data from ment of these tornadoes, which are responsible for most of
peninsular Florida. Byers and Rodebush[1948] recognized the deaths and injuries in central Florida, is also needed.
the uniquenessof the peninsular Florida environmentand The current situation is one where tornado forecast tech-
the need for investigationof dynamic mechanismsthat result niques and conceptual models developed from mean upper
in a United States maximum of thunderstorms in central air data over the Great Plains and Midwest are applied in
Florida. Recently, Golden and Sabones [1991], usingDop- central Florida with limited success.This investigationcon-
pler radar and mesonet data, investigated two tornadic sistsof the determination of the mean atmospheric structure
waterspouts near Cape Canaveral in east central Florida. of two significantcentral Florida tornado environments. The
However, a specific,systematicstudyof the environmentof results presented should lead to improved severe weather
central Florida tornadoes has not been done despite their forecasts for central Florida.
significanceand relatively frequent occurrence[Kelly et al.,
1978]. Past work hasbeenprimarily confinedto casestudies
2. METHODOLOGY OF CASE SELECTION
of tornadoes in south Florida in the wet season. Gerrish
[ 1967]produceda meantornadosoundingfor Miami for the This study is unique in that the area of investigation was
wet season, and case studies of wet seasontornadoesin restricted to the 10 county warning area (CWA) of future
south Florida were investigatedby Hiser [1968], Gerrish National Weather Service (NWS) weather forecast office
[1969], Golden [1971], and Holle and Maier [1980]. (WFO) Melbourne in east central Florida. The location of
These researchers have documented that low-level con- the MelbourneCWA and upper air stationsusedin this study
vergence boundaries, particularly intersectingoutflow are shown in Figure 1. On the basis of a climatological
boundaries,are a triggeringmechanism
for tornadicthunder- investigationof this area by Schmocker et al. [ 1990], tornado
characteristics were divided into two seasons: dry season
The Tornado: Its Structure,Dynamics,Prediction,and Hazards.
(November through April) and wet season (May through
Geophysical Monograph 79
Thispaperis notsubjectto U.S. copyright.
Published
in 1993by the October). The hourly distribution of central Florida torna-
American GeophysicalUnion. does by season (Figure 2) clearly shows a diurnal afternoon
626 EAST CENTRAL FLORIDA TORNADO ENVIRONMENTS

HOURLY DISTRIBUTIONOF TORNADOES

WITHIN 125 NM OF WFO MELBOURNE
BY SEASON(1950 - 1988)

lO

6E•C•
o

Sl

PBI
• ,
1 2 3 4 5 6 7 8 9 10 !1 12 131416 16 1718 192021222324
T • ? NOUR
(EST) T • ?
Includee All Torn•doee

Fig. 2. Hourly distributionof dry seasonand wet seasontorna-

does within 125 nautical miles of future WFO Melbourne. The time
intervals of this studies cases (1200 UTC (0700 EST) and 0000 UTC
(1900 EST) _+2hours) are shownby the arrows along the x axes.
..,...• . . •.
- .•. '•. •- consideredby looking at first spatial, then temporal, selec-
tion criteria. Printouts of all tornadoes reported in Florida
supplied by the Verification Section, National Severe
Storms Forecast Center (NSSFC), were reviewed for cases
84' 83' 82" 81' 80" of tornado touchdowns in the Melbourne CWA (Figure 1)
that occurred within +_2 hours of standard observation times
Fig. 1. Peninsular
Florida.The locationof thefutureWFO Mel-
bourne I0 countywarming area is enclosedby a bold line in east (0000 UTC (1900 EST) and 1200 UTC (0700 EST)) from 1980
central Florida. Upper air stations are indicated by bold arrows: to 1988. This resulted in 16 0000 UTC cases and seven 1200
Cape Canaveral (XMR), Miami (MIA, moved to PBI in 1977), UTC cases and reflects the fact that fewer tornadoes occur at
Tampa Bay (TBW), and West Palm Beach (PBI). 1200 UTC than at 0000 UTC (see Figure 2). Dry season
morning tornadoes often are stronger than wet season tor-
nadoes, so five additional 1200 UTC cases were added by
maximum in the wet seasonwhile the dry season is charac- searchingthe tornado records back to 1975 to get a larger
terized by fewer tornadoesbut a more even hourly distribu- sample.Twenty-eightcandidatecaseswere thus identified
tion with significant morning activity. (16 0000UTC and 12 1200UTC) and soundingsobtained. Six
The distributions of F2-F5 tornadoes and those that have soundings
wereremovedbecauseof poordata quality and/or
caused injury and death specifically in future WFO Mel- contaminationby deep convective. Four tropical cyclone
bourne's CWA are shown in Figure 3. After peaking in cases were also removed from this data set. This selection
March and April, strongtornado activity dropsoff sharplyin criterion yielded nine 1200UTC dry seasoncases and nine
May, as the influence of vigorous mid-latitudedisturbances 0000 UTC wet season cases.
diminishes and the transition from dry to wet season takes The upper air station nearest to a reported tornado was
place. The increase in August and September is due to designatedthe proximity soundingfor each season. Since
tornadoesassociatedwith tropical cyclones. Climatological 0000UTC wet seasonproximity data were not availablefor
analysesindicate that an attack on the problem by seasonal Cape Canaveral, but 1200UTC data were, a 1200 UTC XMR
and diurnal divisions is necessary. wet seasonprecedentsounding(12 hours prior to tornado
Upper air data were availablefor 0000 and 1200UTC for touchdown, +-2 hours) was examined to provide useful
TBW, PBI, and MIA (note that the MIA site was deactivated information on tornado precursor conditions. Soundings
and moved 100 km north to PBI in 1977). Only 1200 UTC were then pressure averaged at 50-mbar intervals to 200
soundingswere available for XMR since 1980. With these mbar. Mean soundingsand diagnosticparameters for the
data limitations in mind, tornado candidate cases were seasonalatmosphereswere computed,and vertical
 HAGEMEYER AND SCHMOCKER 627

STRONG ANDVIOLENTTORNADOES any mean capping inversion or steep lapse rate overlying a
IN WFOMELBOURNE ½WA well-mixed moist layer. An inspection of all individual
(1950- 1988) soundingsin the data set revealed the presence of a few
minor inversion layers, but none of the low-level moist
layers were capped by an inversion.
/ TORNADOES
• KILLER:".'.'"!•
INJURY Dry season stability indices are less than for typical
Midwestern tornado cases [Miller, 1972]. This is probably
becausemidtropospherictemperaturesare generally warmer
14 over Florida; also, because these are morning tornado cases,
there is not much contribution to destabilization from diurnal
12
heating.
!0
The dry seasonhodographshowsstrongshearin the lower
8 levels and winds veering with height. This is similar to the
6 mean F1 and F2 tornado hodographs documented by Riley
and Colquhoun [1990]. A comparison of the wind profile
4
with Miller's [!972] key tornado forecast parameters reveals
2
that the low, middle, and upper level jets of the mean dry
0 seasonsoundingsare all in the strong category.
JAN FEBMARAPRMAYJUN JULAUG$EP OCTNOVDEC The wet seasonmean proximity sounding and hodograph
I•10NTH
(Figure 4) are unlike any of the classic tornado environments
F2 to F$ TornadoesOnly
[Miller, 1972] and represent a regional hybrid. There is a
general similarity to the mean dry season sounding in that a
Fig. 3. Monthly distributionof strongand violent(F2-F5) torna-
does,and killer and injury tornadoesreportedwithinfuture WFO distinct dry layer overlies the moist layer; thus it is most
Melbourne's CWA. unlike Newton's [1980] type C and Miller's [1972] type II for
the Gulf Coast and southeastern regions which are typified
by high moisture through the troposphere.
of temperature(T), dew point (Ta), and wind components The wet season proximity hodograph is quite different
(U and V) were constructedfor dry and wet seasontornado from the dry season.It exhibits very weak shear in the lower
proximity and wet seasontornadoprecedent,atmospheres. levels, and winds are nearly unidirectional from the west.
These seasonal profiles were then compared to seasonal Tornadic thunderstormsforming in this type of an environ-
mean atmospheres. ment would likely be of the nonsupercellvariety [Wakimoto
and Wilson, 1989].

3. RESULTS
3.3. Relative Attnospheric Profiles
3.1. Case Tornado Characteristics
To determine, what, if anything, is unusual about the dry
Of the nine dry seasoncases, five had F2 tornadoes,two and wet seasonproximity and wet seasonprecedent tornado
had F 1 tornadoes, and two had F0 tornadoes. Two caseshad atmospheresprofiles of potential temperature 0, wet-bulb
multiple tornadoes. Of the nine wet seasoncases,one had potential temperature 0w, U, and V were computed and
multiple tornadoes,four had F1 tornadoes,and five had F0 comparedto seasonalmeans [U.S. Department of Defense,
tornadoes. The mean time of occurrence was 1220 UTC 1983].
(0720 EST) and 2240 UTC (1740 EST) for the dry and wet 3.3.1. Dry season. Comparisons of mean dry season
seasoncases, respectively. profiles to adjusted seasonal means (average for all months
with cases, January through April) are shown in Figures
3.2. Mean Soundingsand Diagnostic Parameters 5a-5d. Potential temperature values are significantlyhigher
Skew-T, Log-P plotsand hodographs of the dry and wet than mean values below 800 mbar and above 400 mbar but
seasonmean tornado proximity soundingsare shown in are only slightly above mean between about 700 and 450
Figure4. A summaryof diagnosticparameters derivedfrom mbar indicating less static stability below 700 mbar when
the mean soundingsare shownin Table 1. comparedto normal. More notable is Owgreatly exceeding
The generalthermodynamic structureof the meandry normal values below 650 mbar (Figure 5b). The mean
seasonsounding(Figure4) hasboth similarities atmospherehas a degree of convective instability with a 0w
and differ-
enceswhencompared with theclassicMidwesterntornadic minimum around 800 mbar, but the depth of the moist layer
environment[Doswell,1982].Both environments displaya in the mean tornado atmosphere is about twice as deep, and
pronounced deep,dry layerin themiddleandupperlevels convective instability is much greater.
overlayinga moistlayer. However,the Floridadry season The high U values of the mean tornado profile, reachinga
a deepermoistlayerthanisthecasewiththe maximumof 45 m s-• at 200 mbar, are not significantly
profileexhibits
"classic"profile.Themostnotable isthelackof higherthan seasonalmeans(Figure5c). Indeed, there are
difference
628 EAST CENTRAL FLORIDA TORNADO ENVIRONMENTS

DRY --- WET - -

• DRY WE1
200 • '
',./
/,,,\ '",/'' . . , ,, ,•. "•/ " .
290
/.
.
'x / 'k
'\ .,
,4., ß Y '.• )< '. /•". "/ • '- ,'½' _

'. /
/., './ ., /.. , V ", ½.. \'./ '-\. /
300 /

..
q00 ß ,/. ".
,/ ., /.
½.o. ../ ...
5oo
../. ...'./
/,. /- .
/ . /..2'"•,. '- "./ \",,• /•'. , /.• -'7
600 •--
/ , ../ ß '...,,,.,,,
... ,/-%.,,,,
'. ;z,, \,,',/..,, ....V'r.../j --/•.__
.' '• 'A" \, % .•" ". • ß
.. /,
......,./. ....
',
..•../ ß..,/ ' \ /
•.,.:• \ \% ! , i
,,••. )/.".•, '1 '
.•...., •o,';•'..,½/
/ '. ,/'.. ,/'.. ,/-•. ,,'•'_\,•,¾,',•--,/'I'-..• •,./
700

800 "../
'/ '../ f '.',, ' ',:.-
/ / \ 'F':.- &. ,-<'
9OO
• '..ß/'/' '.' / '
.' '/ • "/ \ :., /,ff½--t.¾.
I /. ,
1000
1050 '" '
½;'"/' ..- "• .....
ß /
/' 'v"" * •- ?Y' •' %? •
-.
'/'•"
/
";/•;"

Fig. 4. Skew-T,Log-Pthermodynamic profilesfor meandry seasonproximity(solidlines)andmeanwet season

proximity(dashed
lines)tornadoenvironments.
Hodographs andverticalwindprofiles(halfwindbarb2.5m s-•, a full
windbarb5 m s-l anda pendant25 m s-• are alsoshown.

minordifferences (<5 m s-•) from the surfaceto 250 mbar. potentialtemperaturevaluesare 2øhigherat the surfaceand
This is notthe casewith the southerlywind(¾) components the same at 650 mbar when compared to the tornado
(Figure 5d), where mean seasonalV is nearly zero and is precedent soundingtaken 12 hours earlier. This indicates a
greatly exceeded by the mean tornado environment at all much greaterdegreeof convectiveinstabilityand illustrates
levels.Very strongshearof V is foundin the lowest100mbar, how diurnalheatingcan nearly double CAPE (see Table 1)
andthereis an indicationof a midlevelsouthwest jet between between 1200UTC (904 J/kg) and 0000 UTC (1683 J/kg).
600 and 400 mbar and an upperjet at 200 mbar. The most outstanding feature of the wet season tornado
3.3.2. Wet season. Comparisons of the mean wet sea- kinematic environment was found to be the existence of
son precedent and proximity profiles to adjusted seasonal significantlyincreasingshear, and westerly winds greatly
means (averagefor monthswith cases, May throughJuly) exceedingseasonalmeans,in the mid and uppertroposphere
are shown as Figures 6a-6d. Except for differencesat the (Figure6c). The trend and magnitudeof the V componentis
surface due to diurnal heating the potential temperature generallyvery close to seasonalmeans(Figure 6d).
valuesof the precedentand proximity atmospheresare very All nine wet seasoncaseshad westerly winds in the mid
close to the mean seasonalvalues (Figure 6a) indicating and upper troposphere,and eight caseswere westerly from
there is little differencein the vertical temperaturestructure the surfaceto 200 mbar. This dominanceof westerly flow
between a nontornado and tornado day in the wet season. cases has several causes. Hagemeyer [1991] found that
Comparisonsof 0w profiles(Figure 6b) showthe depthsof lower troposphericflow is westerly over central Florida into
the moistlayersare nearly the samewith Owminimaaround Juneand that persistenteasterlyflow doesnot appearuntil
650 mbar in all three atmospheres. Proximity wet-bulb well into the wet season.Most casespresentedhere
 HAGEMEYER AND SCHMOCKER 629

TABLE 1. MeanDiagnostic
Parameters tion in the central peninsula interact with the ECSB and
Dry Wet Wet other outflow boundaries to enhance convection before
Prx Pre Prx dissipationduringthe evening. Type III days exhibit greater
echo coverage,and dissipationtakes place much later in the
Freezinglevel (m AGL) 4104 4351 4477
Wet-bulbzero (m AGL) 3360 3410 3419 day, whereas on type I and II days, convective activity
Showalterindex(øC) -0.8 - 1.1 1.5 develops earlier and moves through east central Florida
Liftedindex(øC) -1.8 -3.5 -3.4 before the 0000 UTC (_*2 hours) selection criteria.
Totalsindex(øC) 48 48 43
Crosstotalsindex(øC) 23 22 19 The wet season tornado cases presented here are clearly
Verticaltotalsindex(øC) 25 26 24 westerly flow type III cases, but it is important to note that
K Index (øC) 33 31 26 the wind speedsbetween 500 and 200 mbar on the mean wet
SWEAT Index 325 194 168 seasontornado precedent sounding are twice as high as for
Precipitablewater (cm) 3.8 4.2 3.9
LCL height(m) 386 596 882 the mean type III day soundingfor MIA and PBI produced
LCL mixingratio (g/kg) 13.9 15.8 16.1 by Blanchard and Lopez [1985].
LFC height(m) 2584 2109 1484 Stronger upper level winds and shear appear to be an
CCL height(m) 631 1187 1236 important factor in significant wet seasontornado cases. The
EL temperature(øC) -26.2 -51.1 -55.0
EL level height (m) 8150 11727 12346 strong, damaging, tornado that struck Miami on June 17,
Convective temperature (øC) 24.1 30.6 31.2 1959, developed in the evening in westerly flow, and Hiser
CAPE (J/kg) 164 904 1683 [1968] found that the most outstanding synoptic weather
Bulk Richardson Number 1.5 166 576
Mean 0- 6-kin wind (m/s) 215/19 260/06 267/04
featureat the timeof the tornadowasa 47 knot (24 m s-1)
Assumed storm motion 245/14 290/04 westerly wind speedmaximum at 12 kin.
297/03
Absolute
helicity(m2/s2) 250 10 06
Although the wet seasondata sample is small (five F0 and
Mean 3-10 km wind (m/s) 230/28 268/12 271/09 four F1 cases)there doesappearto be a positiverelationship
Mean shear 850-200 mbar (m/s) 34 14 13
between higher middle and upper tropospheric wind speeds
Parameters were computed from mean dry and wet season and shear and tornado strength. The maximum upper level
proximityandfrommeanwet seasonprecedent,tornadosoundings. winds associated with F1 tornadoes were not lower than 20
AGL is above ground level. m s-• while valuesas low as 10 m s-• were associatedwith
F0 tornadoes.More researchis plannedon this aspectof the
wet season environment.
Tornadoes and waterspouts also occur in easterly flow,
but no east flow wet season cases are included in this study,
curred early in the wet seasonwhen westerly disturbances because, as stated earlier, they would tend to occur in east
are more likely compared to late in the wet seasonwhen
central Florida well before 2 hours prior to 0000 UTC.
easterly flow dominates.Additionally, undisturbedeasterly
Research is planned on these east flow events in the future.
flow early in the wet seasontends to be drier and to have a
shallower moist layer than later in the wet season[Hagem-
eyer, 1991]and is thus lesslikely to producestrongthunder- 4. CONCLUDING REMARKS
storms and tornadoes.
There is also a bias toward west flow cases on the east
We have examined dry and wet season environmental
coast in the -*2 hours from 0000 UTC selection criterion
profiles relative to seasonal means. However, to put this
usedin this study that can be explainedby reviewinga study information to good use operationally, forecasters must
of spatial patterns of south Florida convection, without diagnose the whole picture in detail and understand the
regard to severity, done by Blanchard and Lopez [1985]. interactions and physical processesinvolved to the degree
They identifiedthree basicpatternsof convectionover south
that technology allows. Key factors that effect whether
Florida duringthe summer.Type I exhibitsweak southeast
thunderstormson one day may be tornadic or not depend on
flow with early developmentof convectionalongthe East
the characteristics of the overlying airmass and on the
Coast Sea Breeze convergencezone (ECSB) which moves
probable existence, and strength, of low-level triggering
inland and mergeswith the West Coast Sea Breezeconver-
boundaries. While low-level boundaries are necessary for
gencezone (WCSB) west of the centralpeninsula.Type II
convective initiation, the overlying atmosphere plays an
exhibitsstronger,dry, east flow with early passageof the
ECSB with limited convectionwhich is quickly advectedto important role in the development of tornadic thunder-
the west coastto mergewith the WCSB andmoveout into storms.NEXRAD/WSRo88D with its ability to detect, track,
Gulf. Type III exhibitssouthwestto northwestflow over and possibly quantify the strength of low-level boundaries,
centralFlorida with the westerlyflow advectingthe WCSB as well as diagnosethe dynamic environment at much higher
inland while the ECSB remains anchored along the east spatial and temporal resolutions than the current upper air
coastby ambientflow, causingstrongercirculations
and network, offers the promise of improving short-term fore-
convergence.
Outflowboundaries from the WCSBconvec- castingof tornadic thunderstorms in east central
 CENTRAL
FLORIDA
MEANTORNADO
[XMR) CENTRAL
FLORIDA
MEANTORNADO
(XMR)
16OO0

14OOO

12OO0

10000

80OO

6000

4000
4000

2000
2000

A B
o
10 20 30 40 50 60 70 80 90 100 110 120 12 14. 16 18 20 22 24 26 28
POTENTIAL
TEMPERATURE WET-BULB
POTENTIkL
TEMP
El DRY SEASONPROXIIQITY A DR Y SEASONI•EAN [:3DRY SEASON
PROXIMITY A DRY SEASON
MEAN

CENTRAL
FLORIDA
MEAN
TORNADO
(XMR) CENTRAL
FLORID&
MEAN
TORNADO
(XMR)
160O0

14O00

120O0

100oo

4000

200O

c
-5 0 10 20 30 40 -5 0 10 20
U COMPONENT
(m/s) V COMPONENT
(m/s)
D DRY SEASONPROXIidlTY A DRY SEASON
klEAN [3DRY SEASONPROXIMITY ADRY SEASONMEAN

Fig. 5. Mean verticalprofilesof (a) 0, (b) 0},,, (c) U, and (d) V for meandry seasontornadoproximitycasesand
the mean dry season
 CENTRAL
FLORIDA
MEAN
TORNADO
(XMR) CENTRAL
FLORIDA
MEAN
TORNADO
(XMR)
16000 ,,,

14000

14OOO

120O0

12000

10000
10000

8001)
8000

6000
6ODD

4000
4000

2000 2000

A B
i [ i i [ i " o
20 $0 40 50 60 70 BO go 100 !10 120 !6 18 20 22 24 26 28
POTENTIAL
TEMPERATURE WET-BULB
POTENTIAL
TEMP(øC)
EIWET SEASONPRECEDENTAWET SEASON
MEAN DWET SEASON
PRECEDENT
&WET SEASON
MEAN

CENTRAL
FLORIDA
MEAN
TORNADO
(XMR) CENTRAL
FLORIDA
MEAN
TORNADO
(XMR)
16000 16000

14000 14000

12ODD 12000

10000 10000

•ooo 8ODD

6000 6000

4000 4000

2000 2000

0 0
-5 0 5 10 15 20 -5 0 5 10 15 20
U COMPONENT
(m/s) v co•Po.œ.• (u/s)
DWET SEASON
PRECEDENT
AWET SEASON
MEAN D W[T S[ASOIq PRECEDEN A WET SEASONMEAN

Fig.6. Mean
profiles
of(a)0,(b)0,,,,(c)U,and
(d)Vformean
wetseason
tornado
precedent
andproximity
cases
and the mean wet season
632 EAST CENTRAL FLORIDA TORNADO ENVIRONMENTS

Acknowledgments.Cape Canaveraldata were providedby Hal Hiser, H. W., Radar and synoptic analysis of the Miami tornado of
Herring, ComputerSciencesRaytheonCorporation.Daniel Smith 17 June 1959, J. Appl. Meteorol., 7, 892-900, 1968.
providedTampaBay, Miami, andPalmBeachdata.BarrySchwartz Holle, R. L., and M. W. Maier, Tornado formation from downdraft
providedSkew-T,Log-Pplotsof meansoundings andcomputations interactionin the FACE mesonetwork,Mon. Weather Rev., 108,
of diagnosticparameters.PrestonW. Leftwich,Jr., providedcalcu- 1010-1028, 1980.
lations of helicity. Jo Ann Carney and Karen Hileman assistedin Johns, R. H., J. M. Davies, and P. W. Leftwich, An examination of
data tabulationandfigurepreparation.Thanks to JoeGoldenfor his the relationshipof 0-2 km AGL "positive" wind shearto potential
advice and encouragement to continuethe study. Paul Hebert and buoyant energy in strong and violent tornado situations, in
Mike Sabones providedhelpfulreviewsof the paper.Specialthanks Preprints,16thConferenceon SevereLocal Storms,pp. 593-598,
to Ron Holle, Irv Watson, and Raul Lopez for providing an American Meteorological Society, Boston, Mass., 1990.
extensivecollectionof papersrelatingto Florida convectionfor our Kelly, D. L., J. T. Schaefer,R. P. McNulty, C. A. Doswell III, and
use. Financialsupportfor this paperwas providedby the National R. F. Abbey, Jr., An augmentedtornado climatology, Mort.
WeatherServiceSouthernRegion,Harry Hassel,RegionalDirec- Weather Rev., 106, 1172-1183, 1978.
tor.
Miller, R. C., Notes on analysisand severe storm forecasting
procedures of the Air Force GlobalWeatherCentral, Tech. Rep.
200 (revision),190pp., Air Weather Serv., Scott Air Force Base,
REFERENCES II1., 1972.
Newton, C. W., Overview on convectivestorm systems,in Pro-
Beebe, R. G., Tornadoproximity soundings,Bull. Am. Meteorol. ceedingsof CIMMS Symposium,edited by Y. K. Sasaki et al.,
Soc., 39, 195-201, 1958. pp. 3-107, University of Oklahoma Press, Norman, 1980.
Blanchard,D. O., andR. E. Lopez, Spatialpatternsof convection Riley, P. A., and J. R. Colquhoun, Thermodynamic and wind
in southFlorida, Mon. WeatherRev., 113, 1282-1299,1985. related variables in the environment of United States tornadoes
Byers, H. R., and H. R. Rodebush,Causesof thunderstormsof the and their relationshipto tornado intensity, in Preprints, I6th
Florida Peninsula,J. Meteorol., 5,275-280, 1948. Conferenceon Severe Local Storms, pp. 599-602, American
Darkow,G. L., An analysisof over60tornadoproximitysoundings, MeteorologicalSociety, Boston, Mass., 1990.
in Preprints,6th Conferenceon SevereLocal Storms,pp. 218- Schaefer,J. T., and R. L. Livingston, Structural characteristicsof
221, AmericanMeteorologicalSociety, Boston, Mass., 1969. tornadoproximitysoundings, in Preprints,15th Conferenceon
Doswell, C. A., III, The operationalmeteorologyof convective SevereLocal Storms, pp. 537-540, AmericanMeteorological
weather, vol. 1, Operationalmesoanalysis, NOAA Tech. Memo. Society, Boston, Mass., 1988.
NWS NSSFC-5, Nail. Severe Storms Forecast Center, Kansas Schaefer,J. T., and R. L. Livingston, The evolution of tornado
City, Mo., 1982. proximitysoundings,in Preprints, I6th Conferenceon Severe
Fujita, T., Tornadoesanddownbursts in the contextof generalized Local Storms,pp. 96-101, AmericanMeteorologicalSociety,
planetary scales,J. Atmos. Sci., 38, 1511-1534, 1981. Boston, Mass., 1990.
Gerrish,H. P., Tornadoesand waterspouts in the southFlorida Schmocker, G. K., D. W. Sharp,and B.C. Hagemeyer,Three
area, paperpresentedat Army Conferenceon TropicalMeteorol- initialclimatological studiesfor WFO Melbourne,Florida: A first
ogy, U.S. Army, Coral Gables, Fla., June 8-9, 1967. stepin the preparationfor future operations,NOAA Tech. Memo.
Gerrish,H. P., Intersectingfinelinesanda southFloridatornado,in NWS SR-132,52 pp., Natl. Weather Serv., Fort Worth, Tex.,
Preprints,6th Conferenceon SevereLocal Storms,pp.188-191, 1990.
AmericanMeteorologicalSociety,Boston,Mass., 1969. Taylor,G. E., andG. L. Darkow,Atmosphericstructureprior to
Golden,J. H., Waterspouts andtornadoes oversouthFlorida,Mon. tornadoes derivedfrom proximityandprecedentupperair sound-
Weather Rev., 99, 146-154, 1971. ings,U.S. Nucl.Regul.Comm.Rep., NUREG/CR-2359,95 pp.,
Golden,J. H., andM. E. Sabones, Tornadicwaterspoutformation 1982.
near intersectingboundaries,in Preprints,25th Conferenceon U.S. Departmentof Defense,CapeCanaveral,FloridaRangeRef-
RadarMeteorology, pp. 178-181,AmericanMeteorologicalSoci- erenceAtmosphere 0-70 km altitude,203pp., Secretariat,Range
ety, Boston, Mass., 1991. Commanders Council,White SandsMissileRange,N. Mex.,
Hagemeyer, B. C., A lower-tropospheric
thermodynamic climatol- 1983.
ogyfor MarchthroughSeptember: Someimplications for thun- Wakimoto.,
R. M., and J. W. Wilson,Non-supercell tornadoes,
derstormforecasting, WeatherForecasting,6, 254-270,1991. Mon. WeatherRev., I17, 1113-1140,
 Discussion

D. MCCARTHY, SESSIONCHAIR

National WeatherServiceForecast Office

PAPER L 1 restructuringdemonstrationtakes place, this will be the test

period. The issue probably will be resolved gradually.
Presenter, C. A. Doswell III, National Severe Storms Labora- (Doswell).My responseto Bill's [Bonner's]commentsis that
tory [Doswell,this volume,Tornadoforecasting:
A review] the actualresolutionof suchquestionsoften haslittle to do with
the outcomeof a test; they tend to be dominatedby politicsand
(H. Volkman, private consultant)Who's leadingthe effortto
economics. Further, I don't believe you can extrapolate the
get SELS [SevereLocal Storms;seesessionH, paperH6]
out of the tornado and severe thunderstorm watches? We resultsof a test at a singlelocation around the entire country.
need to know in case there's some influence we can have to If you watch the Weather Channel, you see flash flood and
divert this change? winter storm watches issuedby the local offices, the result of
which I find to be a national disgrace. They're a patchwork
(Doswell)How honestdo I want to be? [laughter]The effortis quilt, wherethe forecastfor a weather event walks up to a state
essentiallyone of politicsand economics,the originsof which boundary,stops,and then picks up againsomewhereelse. This
havelittleto do with the scienceof meteorology or thepractical is the classicwarning and coordination problem. The severe
issuesof forecastingthe weather.Basically,it's a questionof thunderstormand tornado watches, along with the hurricane
where we are to invest our limited resources.If you want to watches,are issuedfrom a central location. Whether they're
have an influence,write lettersto your congressman. right or not (and as we've seen, that's an interesting and
complex issue), at least we're not sendingconflictingsignalsto
(W. Bonner, COMET) I'd like to make it clear that there is
the public. I think there's a good argument in favor of the watch
no intent to destroy or do away with SELS. The intent is to
responsibilitystaying in the hands of SELS.
take advantage of the Doppler radars and so forth to blur the
distinctionbetween watchesand warnings.There shouldbe PAPER L2
an element of prediction with warnings, i.e., those with lead
time. You often end up trying to tell simple stories, perhaps Presenter, J. M. Davies, private consultant [Davies and
too simply in this case; the lead time with a warningoverlaps Johns, this volume, Some wind and instability parameters
with the forecast aspects of a watch. The National Severe associatedwith strongand violent tornadoes, 1, Helicity and
Storms Forecast Center [NSSFC] in this story becomes mean shear magnitudes]
much more of a mesoscaleguidancecenter. The questionis,
(R. Davies-Jones,National Severe StormsLaboratory) Helic-
do products go directly from NSSFC to the public, or do
ity seemsto work well in the wintertime but not quite so well
they go out like NMC [National Meteorological Center]
in the summertime.You get tornadoesin the summertimewith
products as guidance to the local offices, who then issue
high CAPE [convective available potential energy] and low
whatever statements are required? My own feeling was that
helicity. There may be a samplingproblem, but such events
this is an issue you could argue philosophically...
with low helicitydo seemto occur. Have you found this?
(Doswell) And I have.
(Davies) Yes, I've found that, so far. That's been my general
(Bonner) Yes, you and I have discussedit. I have been experience. We have to get better with our observation
beaten up by SELS lead forecasters on this, trying to networks for nowcastingpurposes.
representWeather Servicepolicy. As the modernizationand
(Davies-Jones) Maybe the storms are different from the
The Tornado: Its Structure, Dynamics, Prediction, and Hazards. wintertime storms.
GeophysicalMonograph79
This paperis not subjectto U.S. copyright.Publishedin 1993by the (Davies) That's a very good point. I think that there are
American Geophysical Union. storms which occur with relatively low helicity. The
634 DISCUSSION

hodographsare still in a nice, curved shape,but it's a small parametersassociatedwith strongand violent tornadoes, 2,
curve, and you have strongbuoyancy. I'd like to have some Variations in the combinationsof wind and instability pa-
of the modelers look at that situation and see what you get rameters]
withlowhelicity(say,150m2 s-2) andhighCAPE(say,5000
(J. Jarboe, National Weather Service Training Center) Since
J kg-•). That'sthe oppositeof what you get in the cool
helicity is so dependent on the storm motion vector, I was
season,with very low CAPE and strong helicity.
wondering if you could integrate your data with the obser-
(Davies-Jones)One more comment: the layer over which vations of Ted Fujita. He's flown down many tornadic storm
you compute helicity doesn't really matter very much, as paths, and it would seem you could get an accurate storm
long as it's deep enough. motion from his data.

(Davies) I was wondering if there was much importanceto (Johns) Part of the problem with that is not many storms in
the vertical distribution of the helicity (or the shear). this data set involve long-track tornadoes, and so most of
them have not been surveyed directly. Big outbreaks are
(Davies-Jones) The lower it is, the better it is.
surveyed, of course, but I'm not certain if we could ever
get much more path information for tornadoes that way.
PAPER L3
What we can do, now that we in SELS are getting radar
data, is to collect the radar data and track the cells directly.
Presenter, R. H. Johns, National Severe Storms Forecast We eventually shouldbe able to build up to a good data set
Center [Johns et al., this volume, Some wind and instability that
 Open Discussion
C. DOSWELL, SESSIONCHAIR

National Severe Storms Laboratory

PRESENTATION 1 (Anderson)I'm interestedin your resultsand I'd like to see

a lot more data. You've got maybe 20 cases;I'd like to see
100 cases.
Presenter, P. Desrochers(Air Force GeophysicsLaborato-
ry): Determination of which mesocycloneswill be tornadic, (Desrochers)Yes, we would, too. That's what we're work-
using singleDoppler radar observationsof a new parameter: ing on.
Excess rotational kinetic energy (ERKE)
(D. Burgess,National Severe Storms Laboratory) I have a
(C. Anderson, North Carolina State University) Don Bur- concernabout the algorithmsat their current level of devel-
gess showed yesterday that as range from the radar in- opment.There are thingsthat we mightnot expectandmight
creases, the radar seeshigher and higher into the storm. You be out there when we collect a lot of data. One of these is
seem to be saying that mesocycloneshigh in the storm will cyclonicallyrotatingdowndrafts. We haven't really consid-
determine whether or not that storm will produce a tornado. ered that in the past in algorithm development. They have
In the Huntsville [Alabama] tornado [November 15, 1989] been observed, a few in Oklahoma, many more in Colorado.
A human observer, as he looks to identify the updraft region
the storm produced a mesocyclonefor 2 hours before the
in a storm, isn't going to be confused by a cyclonic down-
Huntsville tornado; it produced funnel clouds and hail, but it
draft, but an automated algorithm, without developing some
certainly didn't produce a tornado within 20 minutes. I don't
code to approximatewhat a human does, may not recognize
believe your predictions can be as accurate as you have
that. There are potentially a lot of such problems, and we
shown, especiallywith respect to false alarms. need to remember that in this stage of algorithm develop-
ment.
(Desrochers) We're keying in here on mesocyclonesat low
levels...
(P. Joe, Atmospheric Environment Service of Canada) I
(Anderson) How are you goingto see at low levels at ranges don't have a problem with ERKE; I think it points out that
over 100 km? a change in the algorithm is needed. I do echo Charlie
Anderson's comment. Thirty mesocyclones out of 8000
(Desrochers) With violent tornadoes, we see that low-level tornadoes per year is a pretty limited data set, especially
intensification is reflected at mid-levels, also. when it comes from Oklahoma, where it's highly biased.
(Anderson)I don't seethat intensificationat mid-levelsneces- (Desrochers) I would agree. What can you say about a small
sarily meansyou're gettingintensificationat low levels. data set'?When you have consistency, though, you can get a
good feeling for it. We haven't been let down yet, and we're
(Desrochers) No, that's not what we're saying. It's not at hoping it will continue. Maybe we'll be surprised.
mid-levelsthat it's important,but at low levels. Ed Brandes
showeda good figureyesterdayof a tornadicstormwhere
PRESENTATION 2
the mesocyclone
at low levelsintensifiedup to about4 km.
The low-level intensificationis reflected at moderate heights
within the storm and that's what we're detecting, and why Presenter, M. Weisman (National Center for Atmospheric
Research): Some cautionary comments on the use of helicity
we get the lead times we do at these ranges.
to assesssupercellpotential, and the use of the time changes
The Tornado:Its Structure,Dynamics,Prediction,and Hazards. in storm-relative helicity
Geophysical Monograph 79
Thispaperis notsubjectto U.S. copyright.Published
in 1993by the (R. Davies-Jones, National Severe Storms Laboratory) I
American GeophysicalUnion. agree with Morris that helicity is intended as a nowcast
636 OPEN DISCUSSION

parameter.If you have a lot of helicityto beginwith, using who had to put out the warning. Someoneasked Chuck
the initial storm motion, the potential is high for a severe Doswell on the first day "If 50% of all mesocyclonesput out
mesocyclone.Most mesocyclonesdon't producea tornado tornadoes,do you put out tornadowarningswith detected
for over an hour. Once the mesocyclonegets cranking, the mesocyclones?" It appearsthatthe publicrespondsmoreto
more the storm deviates in its motion. Storm rotation and tornado than to severe thunderstorm warnings, so if you
motion are interlocked, as is well known. have indicationsfrom the time rate of change in the data that
tornadoes are more likely, then that would be a useful
(Weisman)I think my concernwith this is with a few of the nowcasting tool.
formal NEXRAD products,where the only indicationof
helicityfor nowcasting
is a numberin a cornerof thedisplay, PRESENTATION 3
asopposedto consideration of the processes
thatgointothe
productionof severeweather.I'm advocatingdisplayinga
hodograph,not just a number. Presenters,R. Wilhelmson and L. Wicker (National Center
for SupercomputingApplicationsand Universityof Illinois):
(Davies-Jones)At the Weather Service in Norman, now, we Modelling of long-lived, low-level mesocyclonicand tor-
havean interactiveprogramwherethe forecastercanupdate nadic vortices in supercelIs
the storm motion and see a revised estimate of helicity. We
planto be ableto changethehodograph windsaswell, sothe (C. Anderson,North CarolinaState University)I wantedto
hodograph canbe amendedwithnewwindinformation, such ask Bill McCaul about this earlier. This pressure-perturbed
as profilerwinds,VAD [velocityazimuthdisplay]winds, updraftgoesagainstwhat we learnedearlierin numerical
new surfacewind observations,and model forecastwinds. modelling,wherethe verticalperturbationpressuregradient
opposes thebuoyancyandreducesthe verticalvelocity.I'm
(Chair)I agreewith Morris, in the sensethat we shouldnot not sure how this comes out of the models.
expecta singlenumbercalculatedfrom the sounding to give
usanythingbuta poorrepresentation of thetotalinformation (E. W. McCaul, UniversitiesSpace ResearchAssociation)
content of the data. I certainly hope that helicity will not The pressureperturbationswould tend to counteractthe
become the next Lifted Index, which everybody uses mind- buoyancyif the environmenthad no shear. In a sheared
lessly. [applause] environment,air parcelscan be acceleratedupwardsby this
perturbation
pressureif the patternsof buoyancyand shear
(J. Davies,private consultant)I think helicitycanbe useful are related in the vertical in just the right way. It's most
in forecastingas well as nowcasting.Gary Woodallhasbeen noticeable when there's not that much environmental buoy-
circulatinga programwith the NationalWeatherServicethat ancy available.Lou Wicker hasbeen usingCAPEs [convec-
calculateshelicity density using LFM [limited-areafine- tiveavailable
potential
energy]
of around2900J kg-l' I've
mesh model] winds, that has shown some utility. Of course, done a series of simulations with CAPEs of around 800 J
the LFM is questionablein a lot of cases.I've adaptedhis kg-•. I also see dramaticdifferences
in stormbehavior,
programto my systemat homeand I've foundsomeuseful- dependingon the verticaldistributionsof the buoyancyand
nessfor the helicityforecastsbasedon the modelwindsand shear.Youcangetsteadyupdraftsasweakas6 m s-I and
some rough estimatesof storm motion. If helicity values as strongas25 m s-• depending
on howshearandbuoy-
usingmost storm motionsare high, this suggestsa stronger ancy are related in the vertical.
probability of tornadoes. You're trying to anticipatethe
helicity as the situationchangeswith time. (Anderson) Is that set of conditions rare, or is it common?
How doesthat fit in the spectrumof mesocyclones?
(Weisman) I agree and I want to stress again that it is the
change in helicity over the life of the storm that is most (McCaul) Convection in shear is extremely common. Meso-
important.It's when the stormchangesits motionthat you cyclonesonly form when there's pretty strongshear avail-
get really concerned. able. Vertical distribution variations need a lot more study.
I think there are a lot of caseswhere buoyancy is weak, but
(M. Leduc, AtmosphericEnvironment Serviceof Canada)I
it's locatedin the lower part of the tropospherewhere the
agreevery muchwith the previouscomment.It's a matterof vertical shears are strongest. That can promote strong
continuous testing, using hourly data, pireps, or whatever.
storms in situationswhere it might not be expected.
You're sort of pretunedto watchfor certainthingsto happen
on radar; when it happens, then you're ready to jump. (Chair) I'm delightedto see this result, since it suggeststhat
You're goingto be testingall the time for every differentarea CAPE as a number is not so important as the vertical
and cell. With the King Doppler radar, we'll have an distributionof buoyancy. It pounds another nail in the
interactiveprogramto look at the hodographanda varietyof particular coffin I'm trying to build. [laughter]
shear parameters.
(M. Leduc, AtmosphericEnvironment Service of Canada)
(S. Goodman, NASA) The Huntsville, Alabama, tornado I've seen that, specifically, with instability. Last summer,
case [November 15, 1989] pointed out that there was a lot of we saw three squalllines with Lifted Indices of 0 or -1, but
conflictingor imperfectinformationgettingto the forecasters virtually dry adiabatic below about 750 mbar and lots
 DOSWELL 637

shear. They all three produced long-lived squall lines, and neous, and perhaps every storm initialized that way would
we kept thinking that this shouldn'thappenwith this little behave the same way. But out there where the outbreak is
instability. We keep seeingthis happen. taking place, that atmosphere is not homogeneous,and
that's got to be the answer to your question.
(H. Brooks, National Severe StormsLaboratory). Charlie, I
think I have an answerto your question.We've beenlooking (Chair) Furthermore, I'd argue that when we see big out-
at the accelerationson parcelsthat end up as the maximum breaks, most of the cells are supercelIsand most do produce
updraft, and also the maximumbuoyancyparcel (sincethe tornadoes, at least in the big ones. "Most" does not mean
maximum buoyancy ends up displacedfrom the maximum "all," of course.
updraft),in a shearedenvironment.One of the thingswe find
is that for the maximum updraft parcels, at low levels (below PRESENTATION 4
4 km), the majority of the accelerationis due to the vertical
pressuregradient. The parcelsthat go throughthe maximum Presenter, A. Siemon (State University of New York at
buoyancy location, on the other hand, encounter an adverse Albany): Unusual aspectsof the Plainfield, Illinois, tornadic
pressuregradient over most of their trajectory. Helicity in storm [August28, 1990], includingits cloud-to-groundlight-
this case is a measure of the amount of curvature in the
ning characteristics
hodograph. For 21 simulations using the same thermody-
namic profile, considering the maximum updraft in each (J. Jarboe, National Weather Service Training Center) I've
simulation, in those storms with more curvature in the done some helicity calculationsfor this storm. You said you
hodograph, the stronger the updraft. It's not just the shear, had a very low shearof + 4. Is that the environmentalshear
because we see different behavior with storms that have the or the storm-relative shear? Since the storm was moving
same amount of linear shear but different curvature. In the rapidlyfrom the northwest, the numberI foundwas more on
Beltrami model, the amplification due to the circular the order of 170. You also said the storm wasn't capable of
hodograph over a kilometer depth is about 2.2 times the beingwarnedfor. Had you lookedat the storm-relativeflow,
squareroot of your initial updraft perturbation, from what- it would have indicated that the hook should be on the
ever means. In a shearedenvironment, the vertical pressure leading edge.
gradientaccelerationcan contribute50% or more of the total (Siemon)How commonlyis sucha feature observed[direct-
updraft. ed to the audience]?You guys have looked at a lot of hook
(R. Rotunno, National Center for Atmospheric Research) echoes. How often have you seen major tornadogenesis
Charlie, this is what I was emphasizingin my review paper. alongthe leadingedgeof a storm?I have an eyewitnesswho,
In a shearedenvironment,there are lifting pressuregradients 3 minutes before the tornado, was in sunshine.
on the storm flanks. When observed, it is found that the
(D. Burgess,National Severe Storms Laboratory) It's very
buoyancyis actually negativeat cloud base. This requires common in storms moving from the northwest.
low pressuresomewhereat mid-levelsto lift the air to its
level of free convection. In some situations, this lifting (Siemon) It is? That's very interesting.
pressuregradientcan be quite substantial. (K. Brewster, University of Oklahoma)I think someof the
(Anderson)Everyone is jumping on me for making that thingsthat occurredin this storm show how long-lived
comment.[laughter]If you havean outbreakof tornadoes stormsreact to the inflow of moisture. This was moving from
with, say, five tornadicells, that outbreakmay be accompa- the northwestwith goodmoistureinflow from the southwest.
nied by 20 or more nontornadiccells.If you wantto make (Siemon)The environmentalwind flow was west northwest
thatargumentaboutthe hodograph, why aren'ttheseothers at low levels.
beingaffectedby the sameconditions?
(Brewster) But relative to the storm, the flow at low levels
(Rotunno)I don't know, but theycan't all makeit. was from the southeast,since it was moving so quickly f¾om
(D. Burgess,NationalSevereStormsLaboratory)Charlie, the northwest. Some modeling results don't take into ac-
is countthe interactionwith nearby storms, so the near-storm
I've thoughtaboutthat.My answeris thattheatmosphere
not homogeneous. In the modelthe atmosphereis homoge- environmentmay be quite