See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/237272997

THE AVIATION FORECAST PREPARATION SYSTEM OF THE NATIONAL WEATHER

SERVICE

Article · January 2004

CITATIONS READS

0 289

5 authors, including:

Matthew R. Peroutka George Trojan

National Oceanic and Atmospheric Administration National Oceanic and Atmospheric Administration
9 PUBLICATIONS   59 CITATIONS    11 PUBLICATIONS   288 CITATIONS   

SEE PROFILE SEE PROFILE

Bailing Li
University of Maryland, College Park
35 PUBLICATIONS   720 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Bailing Li on 16 January 2015.

The user has requested enhancement of the downloaded file.

 THE AVIATION FORECAST PREPARATION SYSTEM
OF THE NATIONAL WEATHER SERVICE

Matthew R. Peroutka and Mark G. Oberfield

Meteorological Development Laboratory
Office of Science and Technology
National Weather Service, NOAA
Silver Spring, Maryland

Michael Graf
Meteorological Services Division
Office of Climate, Water, and Weather Services
National Weather Service, NOAA
Silver Spring, Maryland

George Trojan and Bailing Li

Science Applications International Corporation
Beltsville, Maryland

1. INTRODUCTION nally forecast” (NWS 2004a). The process of

monitoring observations, comparing them to the
Forecasters at National Weather Service current forecast, and judging whether an amended
(NWS) Weather Forecast Offices (WFO) prepare forecast is needed is frequently called a “met-
and maintain a set of weather forecasts specifi- watch.”
cally designed for aviation users. NWS (2004a) NWS (2004b) describes a TWEB as describ-
describes a forecast product named the Terminal ing specific information on sustained surface
Aerodrome Forecast (TAF) and NWS (2004b) de- winds…visibility, weather and obscuration to vi-
scribes a related product named Transcribed sion, sky conditions… mountain obscurement, and
Weather Broadcast (TWEB). Preparing TAF and nonconvective low-level wind shear along a route”
TWEB products and monitoring their verification or near an airport “during a 12-hour period.”
constitutes a major portion of a forecaster’s re- WFO forecasters monitor and amend TWEB
sponsibilities at a WFO. For years, computer forecasts much like they monitor and amend
software has played an important role helping TAFs.
forecasters prepare and monitor TAF and TWEB
forecasts. The Aviation Forecast Preparation Sys- 3. Computer Programs, TAFs, and TWEBs
tem (AvnFPS) is the current computer application
used by NWS forecasters for this important task. Paper and pencil were used for many years
to help forecasters perform their aviation met-
2. TAFs and TWEBs watch. The earliest computer systems used at
NWS WFOs were able to supplement this system.
NWS (2004a) describes a TAF as consisting A number of applications were developed that
“of the expected meteorological conditions signifi- compared observations and forecasts and sent a
cant to aviation at an airport (terminal) for a speci- text alert to the forecaster when discrepancies de-
fied time period.” TAFs prepared by NWS fore- veloped. Other applications performed a Quality
casters follow a modified version of the World Control (QC) check on forecasts before they were
Meteorological Organization’s (WMO) Manual on transmitted or managed a transmission queue.
Codes (WMO 2001). Late in the 1990s, computer software was
WFO forecasters are responsible for compar- introduced into WFOs that combined a number of
ing TAFs to observed weather and issuing these capabilities into a single application. Two of
amended forecasts if “conditions meeting amend- these applications were RAVE (Eme and Spriggs,
ment criteria are imminent or have occurred and private communication) and Aviation Workstation
those conditions will, in the forecaster’s estimation, (Machala, private communication). RAVE was
persist (30 minutes or longer), or new guid- hosted on a personal computer and introduced a
ance/information indicates future conditions are “traffic light” concept that allowed forecasters to
expected to be in a different category than origi- quickly assess the status of their forecasts. The
traffic light was a colored circle, one for each TAF ors are used to indicate potential problems. Like
site. A green circle indicated that observations traffic signals, green, yellow, and red suggest in-
and forecasts agreed well for that site. A yellow creasing problem severity. Orange and purple
circle indicated some problems, and a red circle indicate additional levels of severity.
indicated serious problems. Additional colors pro-
Table 1: Monitoring Codes
vided additional information. RAVE also included
an editor to help forecasters compose TAFs as
well as a verification feature. Code Meaning
Machala’s Aviation Workstation was imple-
mented on the NWS’ Advanced Weather Interac- tpo If current forecast includes a forecast
tive Processing System (AWIPS; Seguin 2002). It of intermittent (TEMPO) or probabilis-
displayed forecast and observational data in a tic (PROB30) conditions, tracks
tabular format that allowed a forecaster to readily whether those conditions have been
compare them. Color coding highlighted areas observed during the past 2 h.
where forecasts and observations did not agree. wnd Wind direction and speed
Aviation Workstation included a TAF/TWEB editor
as well as a forecast quality control feature. vsb Visibility
Early in the 2000s, Kirkwood and Hotz (2002)
introduced the AWIPS Aviation Workstation wx Weather
(AAW). AAW adopted RAVE’s traffic light monitor- cig Ceiling
ing technique, and included several useful fea-
tures including editing, quality control, and guid- ts Thunderstorm
ance display.
sky Sky cover
AvnFPS began its development in earnest
late in 2002 with the intent of integrating many of
the useful features found in previous aviation soft-
ware into an application that was part of the The indicators in the section labeled “persis-
AWIPS operational baseline. Since that time, the tence 4hr” compare the forecast for 4 h beyond
application has evolved through three major re- the current time with the current observation.
leases. This paper describes version 3.0 of The indicators in the section labeled “ltg”
AvnFPS. compare the TAF forecast with a 0-3h lightning
nowcast. This nowcast is derived from radar ob-
4. Forecast Monitoring servations, satellite cloud top temperatures, and a
forecast 700 mb wind vector. Kitzmiller, et al.
Figure 1 shows the AvnFPS interface for fore- (1998) describe the statistical techniques used to
cast monitoring. Across the top of the Graphic generate this product.
User Interface (GUI) are three buttons that can be The indicators in the section labeled “grid”
used to launch TAF and TWEB editors as well as compare the forecast with data extracted from
configure AvnFPS to monitor forecasts from a dif- grids generated by the Interactive Forecast Prepa-
ferent WFO (Backup). The next row shows the ration System (IFPS; Ruth et al. 1998; Peroutka et
status of various background processes that sup- al. 1998). IFPS grids are not generally used to
port AvnFPS. Most of the GUI is dedicated to generate aviation forecasts. For consistency’s
monitoring TAFs, observations, and various guid- sake, data from the IFPS grids should agree with
ance sources. data in the TAFs at some nominal level.
The left column of buttons lists the location The rules used to compare TAFs with obser-
identifiers of the stations for which TAFs are being vations and guidance are controlled, in large part,
monitored. Timestamps indicate the valid times of by the local WFO. These alerting rules can be
the most recently processed TAFs and observa- different among TAF sites, allowing WFOs to sup-
tions. Four sections of color-coded indicators fol- port variations in runway alignments and opera-
low. The buttons labeled “Editor Shortcuts” allow tional procedures at the airports they serve. For
forecasters to quickly begin an editing session to each alerting rule, the WFO can configure thresh-
amend, correct, or issue a routinely delayed TAF. old settings, a severity level (which, in turn, deter-
The indicators in the section labeled mines alert colors in the GUI), and a message that
“METAR” compare the most recent set of observa- will be displayed to the forecaster. AvnFPS is de-
tions with the current TAF. Table 1 provides a key livered with a set of rules that support all the
for the contractions used in all four sections. Col- amendment criteria required by NWS policy. The
rules that compare observations and forecasts are (MOS) (Glahn and Lowry 1972). The third tab pro-
generally more straightforward than rules that vides numerical guidance taken from the Eta
compare forecasts with the lightning nowcast or model. The fourth tab includes forecast data
IFPS grids. taken from IFPS grids.
Guidance data can be displayed in the
5. Forecast Preparation AvnFPS TAF Editor in two ways, tabular and for-
matted. The tabular format shows the guidance
The editing interfaces of AvnFPS are de- arranged in rows and columns with little interpreta-
signed to give forecasters ready access to obser- tion. The formatted display attempts to render the
vations and guidance as they prepare their fore- guidance in a format that is ready to “cut and
casts. An emphasis is placed on presenting these paste” into a TAF forecast. Figures 5 and 6 show
data in ways that will contribute the most to the AVN MOS guidance in these two formats. The
forecast generation process. software that generates the formatted displays
Figure 2 shows the TAF Editor, configured to generally adds a considerable amount of informa-
prepare a TAF for station KIAD The text of the tion. These routines must infer specific cloud
TAF can be edited in the upper portion of the GUI, heights from categorical forecasts, so they can
while a set of observations is displayed in the create TAF-ready statements about precipitation
lower portion of the GUI. Figure 2 shows the and obstructions to vision.
METAR observations in their original, coded form.
Figure 3 shows the METARs reformatted in a way 8. Local Tools
that makes them easier to compare to each other.
Notice how background shading is used to convey Near the center of the AvnFPS TAF Editor
flight category information. The TAF editor allows GUI is a menu labeled “Tools.” The Tools Menu
forecasters to use cut and paste tools to copy text allows forecasters to use a number of editing tools
between the observation and forecast sections of which are then applied to all forecasts in the TAF
the GUI as well as among multiple instances of the Editor. These tools are implemented in a scripting
GUI. The tab in the lower section of the TAF Edi- language (Python) which allows WFOs to modify
tor shows observations from multiple locations. existing tools and/or create new tools. Four tools
The rest of the tabs will be addressed below under are provided with AvnFPS. “AdjustTimes” modi-
“Guidance Display.” fies each TAF by removing periods that are older
than the current system time and adjusting issue
6. Forecast Quality Control and valid times. “CopyForecasts” lets the fore-
caster use an interactive menu to copy forecast
Near the top of the TAF Editor GUI in Fig- data from one TAF to another. “UseMetarForPre-
ures 2 and 3 is a button labeled “QC.” This button vailing” updates the earliest hours of each TAF
activates the AvnFPS’ QC routine. The QC func- with data taken from the relevant observations.
tion validates the syntactic correctness of all TAFs “WestFlow” is an example of more sophisticated
in the editor. If the QC routines identify any is- tools that could be built. It generates forecasts for
sues, the TAF text in question is highlighted. The multiple TAFs, based on two human-generated
forecaster can point to any of the highlighted text TAFs. Adjustments are made to account for tim-
and receive a description of the problem. Figure 4 ing and station elevation.
contains a portion of the TAF Editor display that
shows two problems. The first problem is an in- 9. Support for TWEB
appropriate cloud base value and the second
show misuse of the code FG. Many of the features that AvnFPS provides
for the generation of TAFs are also provided to
7. Guidance Display support the generation of TWEBs. Figure 7 shows
the AvnFPS TWEB Editor. Like the TAF Editor,
AvnFPS attempts to make relevant objective the upper portion of the GUI provides an editing
forecast guidance available to the forecaster while interface for one or more TWEBs. The lower por-
TAFs are being generated. Four of the tabs tion of the interface allows the forecaster to display
shown in the lower portion of the TAF Editor in relevant supporting data, such as TAFs and ob-
Figures 2 and 3 provide the forecaster with nu- servations. A TWEB Quality Control feature can
merical guidance. Two of the tabs, labeled “AVN- validate the syntactic correctness of all TWEBs in
MOS” and “NGM-MOS” provide statistical fore- the editor.
casts generated from Model Output Statistics
10. Implementation a. Monitor Low-level Wind Shear (LLWS)

AvnFPS is implemented with a flexible, dis- LLWS is a major peril to aviation operations.
tributed processing model. This takes advantage Wind profile data can be retrieved from the NOAA
of the distributed design of the AWIPS platform. Profiler Network (Beran and Wilfong; 1998) as well
Processes that capture data for AvnFPS can be as WSR-88D radar information. Software has
located on hosts where these data are readily been developed for AvnFPS that can process wind
available. Moreover, as AWIPS evolves and vari- profile data looking for indications of LLWS, and
ous data ingest processes are rehosted, AvnFPS compare the observations to the current forecast,
can readily adapt. Interprocess communications alerting the forecaster as necessary.
are implemented using Python Remote Objects
(PYRO; de Jong 2004). PYRO uses a reserved b. Climate-based Quality Control
range of Internet Protocol (IP) addresses as its
data transport layer. The QC techniques described above validate
Figure 8 illustrates the processes that support the syntax of the text forecasts. No attempt is
AvnFPS and many of the data flows. AvnFPS is made, however, to assess the meteorological con-
designed to ingest data from multiple sources. tent. Algorithms under development can assess
The AWIPS Text Database supplies TAF, TWEB, the climatological frequency of the weather ele-
and observations. Lightning guidance must be ment combinations found in the TAF. To support
queried from netCDF files that contain gridded this capability, observational data for 1259 US sta-
data. MOS guidance comes from netCDF files tions were amassed. These data included all
that are organized by station. In the current hourly and “special” observations since 1973.
AWIPS configuration, these data reside on differ- Aviation-relevant weather elements were extracted
ent hosts. For illustration, Figure 8 shows two from these observations and stored in a format
data feeds on two hosts. AvnFPS can be readily that would support rapid access.
configured to support more hosts or fewer hosts. When a forecaster invokes Climatological
A key element in the implementation of QC, AvnFPS categorizes each group of weather
AvnFPS is the Name Server process. The Name elements found in the TAF. These combinations
Server maintains information on all other AvnFPS of categories, along with the time of day and the
processes. All other processes are directed to the time of year are then compared with all the obser-
Name Server to determine the configuration of the vations that are available for that station. If the
rest of the modules of AvnFPS. Thus, an instance combination forecast has a low climatological fre-
of the AvnFPS GUI running on a workstation quency, subsets of weather elements are com-
needs only to be configured to communicate with pared in an attempt to identify an “outlier.”
the Name Server. After contacting the Name Equation 1 shows the basic mathematical
Server, the GUI process will obtain the information expressions used to flag unlikely combinations of
needed to contact any other processes it needs to weather elements.
contact. This implementation is flexible enough to
support a GUI process running outside the WFO P(C | V ∩ O ∩ P ∩ W) << P(C)
Local Area Network (LAN). P(V | O ∩ P ∩ W ∩ C) << P(V)
P(O | P ∩ W ∩ C ∩ V) << P(O) (1)
When data arrive, an instance of the Data
P(P | W ∩ C ∩ V ∩ O) << P(P)
Ingest Server captures the data, performs what- P(W | C ∩ V ∩ O ∩ P) << P(W)
ever decoding is needed, and stores the data in a
format that is readily readable by the Data Re- In these expressions, P(C) represents the
quest Server. The Data Request Servers respond probability of finding the forecast category for ceil-
to requests from any active GUIs. The event ing in the climatological record. Probabilities are
server buffers data arrival events and provides represented in a similar way for visibility (V), ob-
notification to the GUI. The Transmission Server structions to vision (O), precipitation (P), and wind
manages the details of transmitting TAFs and (W). The conditional expression P(C | V ∩ O ∩ P
TWEBs on communications circuits, including de- ∩ W) represents the probability of finding the fore-
layed transmission. cast category for ceiling, given the occurrence of
the other weather elements. When one or more of
11. Future Plans the inequalities shown in (1) is true, the weather
element involved is flagged to the forecaster as a
MDL continues to improve AvnFPS. Below potential problem.
are some areas that are under development.
c. Climatology Viewer ordinator for Meteorological Services and
Supporting Research (OFCM) Rep. FCM-
In conjunction with the Climatological QC, a R14-1998, 56 pp. [Available from OFCM,
Climatology Viewer has been developed to help 8455 Colesville Road, Suite 1500, Silver
forecasters visualize climatological frequencies in Spring, MD 20910.]
the observational dataset. Figure 9 shows an in- de Jong, I., 2004: PYRO—Python Remote Ob-
stance of the AvnFPS Climatology Viewer display- jects, Version 3.4. Irmen de Jong, released
ing data from Great Falls International Airport under MIT License. [Available online at
(KGTF) in Great Falls, Montana. The controls in http://pyro.sourceforge.net/.]
the top portion of the viewer allow the forecaster to Glahn, H. R., and D. A. Lowry, 1972: The use of
filter the observational data for date, time, and model output statistics (MOS) in objective
various weather elements. The lower portions of weather forecasting. J. Appl. Meteor., 11,
the viewer contain histograms that display the fre- 1203-1211.
quency of various events in the climatological re- Kirkwood, P. and D. Hotz, 2002: AWIPS Aviation
cord. Workstation Program Documentation (Ver-
The data shown in Figure 9 have been filtered sion 3.2). 18th Int. Conf. on Interactive Infor-
with the following criteria: 1) observation time be- mation and Processing Systems (IIPS) for
tween 1700 and 1900 UTC, 2) observation date Meteor., Oceanography, and Hydrology, Or-
within a 20-day window around the first of June, lando, FL, Amer. Meteor. Soc. [Extended
3) southwest wind direction, and 4) wind speed Abstract available online at http://ams.confex.
between 10 and 20 knots. These criteria were com/ams/pdfpapers/26349.pdf.]
selected to find events influenced by a strong Kitzmiller, D. H., M. A.. R. Lilly, and S. D. Vibert,
downslope effect at KGTF. Note that higher val- 1998: The SCAN 0-3 hour rainfall and light-
ues for ceiling and visibility dominate, the probabil- ning forecast algorithms. [Available online at
ity of precipitation occurring in an observation is http://www.nws.noaa.gov/tdl/radar/
small (~4%), and that fog and haze are virtually 03h_doc.htm.]
unknown with these conditions. NWS, 2004: Terminal Aerodrome Forecasts,
NWS Instruction, 10-813, NWS, 57 pp.
d. Improvements to Formatted Guidance [Available online at http://www.nws.noaa.gov/
directives/010/pd01008013c.pdf.]
Generating “TAF-ready” text from forecast _____, 2004: Transcribed Weather Broadcasts,
guidance presents a complex set of challenges. NWS Instruction, 10-805, NWS, 25 pp.
Specific values must be determined where guid- [Available online at http://www.nws.noaa.gov/
ance supplies only categories. Moreover, very few directives/010/pd01008005c.pdf.]
guidance sources supply all the weather elements Peroutka, M. R., R. K. Meiggs, and M. B. Rom-
needed to produce a TAF. Techniques are under berg, 1998: The generation of products in in-
development to combine various guidance teractive forecast preparation. Preprints, 14th
sources and “fill in” data where needed. Condi- Int. Conf. on Interactive Information and
tional climatology may prove useful in choosing Processing Systems (IIPS) for Meteor.,
deterministic values. Oceanography, and Hydrology, Phoenix, AZ,
Amer. Meteor. Soc., 350-354.
12. Conclusion Ruth, D. P., M. A. Mathewson, T. J. LeFebvre, and
P. K. Wu, 1998: Interpretation and editing
AvnFPS is a powerful application that helps techniques for interactive forecast prepara-
forecasters monitor weather conditions and pre- tion. Preprints, 14th Int. Conf. on Interactive
pare their aviation forecasts. This application in- Information and Processing Systems (IIPS)
cludes tools that help forecasters assess the qual- for Meteor., Oceanography, and Hydrology,
ity of the forecast before it is issued as well as Phoenix, AZ, Amer. Meteor. Soc., 321-326.
maintain a weather watch. Tools are available in Seguin, W. R., 2002: AWIPS—An end-to-end
AvnFPS to help forecasters view guidance prod- look. Preprints, 18th Int. Conf. on Interactive
ucts and incorporate them into their forecasts. Information and Processing Systems (IIPS)
for Meteor., Oceanography, and Hydrology,
13. References Orlando, FL, Amer. Meteor. Soc., J47-J51.
WMO, 2001: Manual on Codes, FM 51–XII TAF
Beran, D.W., and T.L. Wilfong, 1998: U.S. wind (Aerodrome forecast), Vol. 1.1, WMO No. 306,
profilers: A review. Office of the Federal Co- 492 pp.
Figure 1: AvnFPS Monitor
Figure 2: AvnFPS TAF Editor with coded observations
Figure 3: AvnFPS TAF Editor with reformatted observations
 Figure 4: Detail of Quality Control in AvnFPS Text Editor

Figure 5: AVN MOS Guidance in tabular form in AvnFPS TAF Editor

Figure 6: AVN MOS Guidance in formatted form in AvnFPS TAF Editor
Figure 7: AvnFPS TWEB Editor
Figure 8: AvnFPS Implementation

Key

ns Name Server Allows various processes to locate each other. All

processes connect to ns. Lines to/from this process
are omitted for simplicity.
es Event Server Notifies GUI of asynchronous events such as data
arrival. Instances of dis and xs report events to es.
dis Data Ingest Server Makes AWIPS data available for use in AvnFPS.
Performs any decoding/reformatting needed.
drs Data Request Server Provides AWIPS data to GUI as requested.
xs Transmission Server Manages product transmission.
 Figure 9: AvnFPS Climatology Viewer

View publication stats