29th Conference on Hurricanes and Tropical Meteorology, 10–14 May 2010, Tucson, Arizona

P2.96 A SIMPLE COASTLINE STORM SURGE MODEL BASED ON PRE-RUN SLOSH OUTPUTS

Liming Xu*
FM Global Research, 1151 Boston Providence Turnpike, Norwood, MA 02062

1. INTRODUCTION

Storm surge is an abnormal rise of sea level estimate the peak surge heights for the 2005
associated with the passage of hurricanes hurricanes Katrina and Rita. His results were
and other severe storms; the rising water in good agreement with observed values.
often comes with violent force (Harris, 1963).
As Hurricane Katrina struck the U.S. in 2005, Although empirical models can produce
its storm surge was over twenty feet (6 m) reliable estimates for peak surge heights at
high along the coastline, decimated cities, the open coast, they fail to produce accurate
killed more than a thousand people, and estimates for surge distribution along the
caused economic damage of approximately coastline and for inland flooding because
$40 billion, half of the total estimated they are incapable of modeling the complex
hurricane loss (Knabb et al. 2006). To process of surge formation. Storm surge
predict such powerful storm surges and to depends not only on wind speed, pressure
assess their risks, many mathematical deficit, size, translation speed, and moving
models have been developed. They range direction of the storm, but also on coastal
from simple empirical models to bathymetry, topography, barriers, waterways,
sophisticated numerical ones. astronomical tide, and other local
geographical features at the landfall area
Empirical surge models are relatively easy (Harris, 1963). In contrast to empirical
to develop and simple to apply but have models, sophisticated numerical models
limited capability and are no longer widely have succeeded in producing accurate
applied. They generally use empirical estimates for both coastline surge and for
functions to relate peak surge height to the inland flooding. Numerical surge models use
storm pressure deficit, which is the pressure hydro-dynamical governing equations to
difference between the storm center and its represent the surge formation process and
periphery. They tend to calculate only peak run on a grid system that includes most
surges at the open coastline. Conner et al. factors affecting storm surge. One typical
(1957) developed an empirical equation to such numerical model is the SLOSH model,
compute peak surge heights for landfalling developed by the National Weather Service
hurricanes, using observed historical (NWS). SLOSH stands for “Sea, Lake, and
hurricane data and surge heights along Overland Surges from Hurricanes”
open coasts. The equation was obtained as (Jelesnianski et al. 1992). As implied by the
a linear regression to determine peak name, the SLOSH model computes inland
surge as a function of pressure deficit. flood triggered by storm surge. NWS has
Jelesnianski (1972) developed three simulated thousands of hypothetical
nomograms from empirical data and hurricanes using the SLOSH model and
theoretical calculations. These allow rapid loaded the pre-run results into a package
estimation of peak surge heights. The first called SLOSH Display (FEMA et al. 2003).
nomogram uses pressure deficit and radius
to maximum wind to make the initial surge Applications of numerical surge models such
estimation. The other nomograms make as SLOSH require intensive computation
corrections for shoaling effects and for and extensive training for operational
direction and speed of storm motion. Hsu personnel. In many cases, however, simple
(2006) applied these nomograms to models are desirable due to time and
resource constraints. This paper presents a
simple model for computing storm surge
Corresponding author address: Liming Xu, FM along hurricane-prone coastlines of the
Global Research, 1151 Boston Providence United States, which is a hybrid of empirical
Turnpike, Norwood, MA 02062; Email: and numerical surge models. This model
[email protected] uses a parametric wind profile model to

1
29th Conference on Hurricanes and Tropical Meteorology, 10–14 May 2010, Tucson, Arizona

compute hurricane wind speeds along the and flow through barrier gaps (Jelesnianski
coastline and a wind-surge function to derive et al. 1992).
surge heights. The wind-surge function is
based on the pre-run SLOSH outputs by The SLOSH governing equations are driven
NWS. This model does not involve complex by a parametric wind field model that
numerical computation but it uses the requires only a few meteorological inputs.
outputs of the numerical model. This simple They are selected in a way to make it
coastline storm surge model will be possible for them to be forecasted
hereafter referred to as the SCSSM model. operationally so that SLOSH can be used for
The rest of this paper will describe in detail real-time surge forecasting (Jelesnianski et
the development of the SCSSM model. al. 1992). The input parameters include: (1)
Section 2 introduces the SLOSH model and storm positions every 6 hours; (2) central
its pre-run outputs, which are the basis of pressure deficit; and (3) radius to maximum
the SCSSM model. Section 3 describes the wind. The radius to maximum wind is
coastline wind profile model, the wind-surge defined as the distance between the storm
function, and model verification. Section 4 center and the location where the maximum
discusses the accuracy and limitations of the wind occurs. The six-hour storm positions
SCSSM model. Section 5 summarizes the implicitly include direction and speed of the
main features of the model. storm motion. In addition, the initial water
level is also required for surge height
2. SLOSH MODEL AND ITS PRE RUN calculation.
OUTPUTS
The SLOSH model produces two types of
This section describes the characteristics of outputs: surge profile and surge envelope.
SLOSH model and the SLOSH pre-run The surge profile is formed by surge heights
outputs in the SLOSH Display package. at every grid location in a basin at a
particular moment, which produces
2.1 SLOSH model sequential snapshots. The envelope
represents the highest surge level at every
The SLOSH model employs a set of grid location during the passage of the
governing equations, derived from hurricane regardless of occurring time. The
Newtonian fluid motion and continuity laws, maximum surge heights at coastline
to compute the dynamic changes of water locations form the surge envelope along the
level during the passage of a hurricane. It coastline. Hereafter, surge height at a
applies a two-dimensional finite difference location will refer to the maximum surge
scheme to solve the governing equations for height at that location during the passage of
a grid network comprising a basin. SLOSH the storm.
uses 38 basins to cover the Gulf and Atlantic
coast of the U.S.; Oahu, Hawaii; Puerto Rico; SLOSH produces reasonably accurate
and the Virgin Islands. These basins are estimates for maximum surge heights, as
constantly updated or new basins are added supported by many validation cases.
as more accurate basin data become Comparisons with historical hurricanes
available (Glahn et al. 2009). showed that the SLOSH model estimates
were within ±20% of the observed values
The grid system expresses basin physical (Jelesnianski et al. 1992). Good agreement
characteristics such as inland topography, was also reported by Jarvinen and
bathymetry of coastal area, configuration of Lawrence (1985), who used SLOSH to
coastline, and continental shelves. Channels, compute surge heights for 10 historical
waterways, and significant natural and man- hurricanes and compared them with 523
made barriers are included as sub-grid observed values.
elements to increase accuracy. This grid
system enables the SLOSH model to 2.2 SLOSH pre-run output package
generate surge heights along the coastline,
simulate inland flooding, compute routing of NWS has used the SLOSH model to
storm surge into bays, estuaries, or river simulate thousands of hypothetical
basins, and allow the overtopping of barriers hurricanes at each basin along the Gulf and

2
29th Conference on Hurricanes and Tropical Meteorology, 10–14 May 2010, Tucson, Arizona

East coasts of the United States, and surge levels are displayed by colors and the
distributed the results in the SLOSH Display parallel lines with arrows indicate hurricane
package. The composite of SLOSH pre- tracks and their directions.
outputs in this package is the basis for the
development of the simple coastline surge Table 1. Central pressures and pressure
model described in this paper. deficit of simulated hurricanes for MEOW
Actual Modeled Pressure
The SLOSH Display program includes two Category Hurricane Hurricane Deficit
SLOSH composites: (1) Maximum Envelope (mb) (mb) (mb)
of Water (MEOW) elevation and (2) the
1 1010–980 990 Δp=20
Maximum of MEOW (MOM). A large number
of MEOWs were generated in each basin 2 979–965 970 Δp=40
and they are identified by the combination of 3 964-945 950 Δp=60
hurricane category, landfall direction, 4 944-920 930 Δp=80
translational speed, and initial tide level.
5 <920 910 Δp=100
Each MEOW is created with a group of
parallel hurricanes, about 10 miles apart, of
the same category, storm direction,
translational speed, and initial tide level
(Taylor, 2009; FEMA et al. 2003). The
category ranges from 1 to 5, as defined by
the Saffir-Simpson Scale (Simpson, 1974)
using the minimum central pressure. Table 1
shows the central pressures for each
category and corresponding pressure
deficits by using peripheral pressure of 1010
mb (FEMA et al., 2006). Storm directions at
landfall are selected from 16 directions with
0
interval of 22.5 , as shown in the lower left
corner of Fig. 1. The coastline orientation
actually determines the number of storm
directions selected for MEOW generation in
each basin. The selected translational
speeds represent the climatological values
and they are 5 or 10 mph apart depending
on basin latitudes. In addition, each MEOW Figure 1. An image of SLOSH MEOW output
also specifies a tide level, high tide or mean at Pensacola Bay, Florida (Taylor, 2009)
tide. Overall, MEOW represents the highest
surge value for the group of parallel
hurricanes with the specified category, storm 3. MODEL DEVELOPMENT
direction, translation speed, and initial tide
level. The simple coastline surge model includes
two major components: an empirical wind
The number of MEOWs for a typical SLOSH profile model for computing wind speeds
basin can reach hundreds and the number along the coastline, and the coastline wind-
of hypothetical hurricanes to generate them surge function based on SLOSH MEOWs.
may attain thousands. For example, if three This section describes these two
translational speeds, eight storm directions, components and the model verification.
and two tide levels are simulated for each of
the five categories, the total number of 3.1 Coastline wind profile model
MEOWs for the basin will be 240. If each
MEOW is created with a group of 20 parallel The coastline wind profile model is a
hypothetical hurricanes, the number of modification of the Kaplan-DeMaria (1995)
hurricanes to produce these MEOWs will be empirical wind model for coastline locations.
4800. Figure 1 shows a SLOSH MEOW Wind speeds are calculated by
image at Pensacola Bay, Florida, where

3
29th Conference on Hurricanes and Tropical Meteorology, 10–14 May 2010, Tucson, Arizona

maps the wind speed V at location X to the

⎛ r ⎞ ⎧⎪1 ⎡ ⎛ r ⎞ ⎤⎫⎪
α

V =Vx⎜⎜ ⎟⎟exp⎨ ⎢1− ⎜⎜ ⎟⎟ ⎥⎬ ± cs r ≥ rmax (1) surge height. The wind speed V can be
⎝ rmax⎠ ⎪⎩α ⎢⎣ ⎝ rmax⎠ ⎥⎦⎪⎭ computed using the wind profile model
equations 1 and 2. The subscript H and X
indicate that function FH,X is defined for a
⎛ r ⎞ given hurricane (H) and for a given location
V = V x ± ⎜⎜ ⎟⎟ cs r < rmax (2) (X). The function FH,X at a location changes
⎝ rmax ⎠ for different hurricanes even if they may
produce the same wind speed at the
where V x = M x − cs is the symmetric location. The function FH,X also changes at
different locations for a given hurricane and
maximum wind, Mx is the maximum wind, cs the changes are due to the variation in wind
is the translational speed, V is the wind speed at different locations. In other words,
speed at a location with a distance r from the function has to be defined for each
the storm track, rmax is the radius of location and for each hurricane. Therefore,
maximum wind, and α is a profile parameter. the function reflects implicitly the parameters
The storm profile parameter determines how that define the hurricane and the local
fast the wind speed decays as the distance geographical conditions that define the
from the storm track increases. The positive location. So the wind-surge function is not
sign is for locations on the right of the just a function that simply relates the wind
hurricane track; the negative sign is for speed to surge height, but it include effects
locations on the left. The wind field is of local geographical features and hurricane
symmetric if the translational speed cs is parameters.
zero. The value of the translational speed
determines the asymmetry of the wind field This function can be extremely complicated
by placing stronger wind on the right. because, as explained before, surge heights
are determined by many factors. However,
3.2 Coastline wind-surge relationship SLOSH MEOWs provide an excellent
opportunity for a simple method of linear
The coastline wind-surge function is interpolation to approximate the function. To
designed to relate wind speed to the surge approximate the function FH.X (V,X) for a
height at a coastline location by assuming hurricane at a coastline location, we choose
that the surge height is a function of wind the basin where the hurricane makes
speed. This assumption seems to consider landfall, the MEOWs that have the storm
only the most important factor of the surge direction and translational speed closest to
formation process—wind; but this function those of the hurricane, and the initial tide
implicitly includes effects of storm direction, level closest to the tide level at the time
translational speed, initial tide level, and when the hurricane approaches the
effects of important local geographical coastline. For the selected storm direction,
features such as bathymetry, topography, translational speed, and initial tide, there are
barriers, and waterways which are modeled five MEOW surge heights at the location,
in the SLOSH grids. To achieve this result, one from each hurricane category. We
the wind-surge function is derived by a linear assume that these five surge heights
interpolation of the surge heights from represent five samples of the wind-surge
SLOSH MEOWs. The details of the function expressed in equation 3. This is the
derivation are given below. fundamental assumption in the development
of the simple coastline surge model.
Let the wind-surge function have the
following form, The five samples constitute the basis for the
linear interpolation to approximate the wind-
G ( X ) = FH , X (V , X ) (3) surge function. If we let index I indicate the
hurricane category, the five samples can be
In this equation, X represents the expressed as
coordinates of a coastline location, latitude
and longitude. G(X) is the surge height at
the location X. The general function FH,X

4
29th Conference on Hurricanes and Tropical Meteorology, 10–14 May 2010, Tucson, Arizona

3.3 Model Verification

G I ( X ) = FH , X (V I , X ) I = 1, 2, 3, 4, 5 (4)
The model has been verified preliminarily
with two hurricanes, 1989 Hugo and 1992
where GI(X) is the MEOW surge height of Andrew, in comparison with the SLOSH
category I at the location X, VI is the model. We selected these two hurricanes
maximum wind speed of category I because their SLOSH surge estimates were
hurricane. The hurricane category is defined included in the SLOSH program provided by
by the central pressure deficit as shown in NWS. At locations on the open coastline, the
Table 1. We use a pressure-wind SCSSM consistently re-produced surge
relationship by Landsea et al. (2004) to height estimates by the SLOSH model.
compute the maximum wind speed from the
pressure deficit, The SCSSM model requires hurricane
parameters and SLOSH MEOWs identified
VI =14.172(Δp I ) 0.4778 (5) by the parameters to compute surge heights
along the open coastline. The computation
where ΔpI is the pressure deficit used for includes five steps. First, determine the
SLOSH MEOW for category I in mb and VI is parameters that describe the hurricane at
1-min surface sustained wind in kt. Using landfall. These parameters include the
the five pairs of wind speed and surge maximum wind speed, central pressure,
height, we can easily build a piecewise storm direction, translational speed, and
linear function by interpolation to initial tide level. The maximum wind speed
approximate the wind-surge function at the can be obtained by equation 5 using the
location. Fig. 2 shows an example of the pressure deficit. In this verification, we
linearly interpolated wind-surge function for selected the same parameters for the two
a coast location using SLOSH MEOWs. The hurricanes as those used in the SLOSH
data points are the MEOW surge heights program in order to have a consistent
and wind speeds corresponding to the five comparison. Second, identify SLOSH
hurricane categories. Extrapolations on both MEOWs for each hurricane category so that
ends of the piecewise linear curve will be the storm direction, translational speed, and
used to calculate surge heights for wind initial tide level associated with the MEOWs
speeds which are lower than category 1 or are closest to the parameters of the land-
higher than category 5. falling hurricane. Third, construct the wind-
surge function by linear interpolation for the
coastline locations of interest. Fourth,
compute wind speeds at the locations using
25 the wind profile model. Finally, use the
piecewise linear wind-surge function at each
location to obtain the estimated surge
20
S to r m S u r g e (F e e t)

heights.
15
Figures 3 and 4 compare the SCSSM and
the SLOSH model for surge height
10 estimates along the open coastline for the
two hurricanes. The data points in these
5 figures represent surge heights at locations
along the open coastline. More locations on
0 the right side of the hurricanes track are
40 60 80 100 120 140 160 selected than those on the left, because we
Wind Speed (mph) are more interested in the high surge
heights that tend to occur on the right side.

Figure 2. Piecewise linear wind-surge

function for a location.

5
29th Conference on Hurricanes and Tropical Meteorology, 10–14 May 2010, Tucson, Arizona

SLOSH MEOW values of five hurricane

25 categories. There are several
Calculated Surge Height (ft)
approximations in this assumption. First, we
20
approximate the complex dynamic
relationship between the surge height and
15
hurricane wind by a simple function of wind
10
speed. Although the function incorporates
many factors considered by SLOSH MEOW,
5 it does not fully account for the effects of
some other important factors. For example,
0 it does not consider the effect of hurricane
0 5 10 15 20 25 size because SLOSH MEOWs were
SLOSH Surge Height (ft) generated with the same radius to maximum
winds. Hurricane size is generally
represented by the radius to maximum wind
Figure 3. Comparison of surge heights by in parametric wind field models. On the
SLOSH and by the SCSSM for 1989 Hugo. other hand, the general SLOSH model
considers the effect of hurricane size by
including the radius to maximum wind as an
input parameter to its wind field model. Irish
15 et al. (2008) showed that hurricane size has
a strong influence on surge height. Secondly,
Calculated Surge Height (ft)

12 the selected SLOSH MEOWs cannot have

the exact same storm direction, translational
9 speed, and initial tide level as those of the
hurricane of interest. The samples derived
from these SLOSH MEOWs are not the true
6
samples of the wind-surge function for the
hurricane but we used them as the true
3
samples. Thirdly, we approximate the wind-
surge function by a linear interpolation even
0
though the relationship between surge and
0 3 6 9 12 15
wind speed is highly non-linear (Irish et al.
SLOSH Surge Height (ft) 2008). However, the model verification
shows that even with these approximations
the model still performs well relative to the
Figure 4. Comparison of surge heights by SLOSH estimates.
SLOSH and by the SCSSM for 1992 Andrew.
The SCSSM model has its limitations in
4. DISCUSSION addition to those of the SLOSH model and
the SLOSH Display package. The model
We have made many assumptions and was developed only for surge estimation
approximations in the development of the along open coastline. It does not compute
SCSSM model to take advantages of the water elevations for inland flooding. It
SLOSH MEOWs. These approximations cannot compute the surge time series, either.
introduce errors in the model and limitations Since it was developed with the SLOSH
to its applications. The model errors Display package, the model covers only
discussed here are relative to the SLOSH regions that the package includes. In
model instead of the observed surge values addition, the model was developed for
of actual hurricanes. landfalling hurricanes, not for by-passing
hurricanes and coastal winter storms like
As the most important assumption, we Nor’easters. Because of these limitations,
assume that there is a wind-surge function the SCSSM model cannot replace powerful
relating wind speed to surge height at an numerical models for applications involving
open coastline location and that this function estimation of inland flooding.
can be obtained by linearly interpolating the

6
29th Conference on Hurricanes and Tropical Meteorology, 10–14 May 2010, Tucson, Arizona

The SCSSM model, however, has its own REFERENCES

advantages: simple to understand, easy to
use, and efficient to apply. It can be used to Conner, W. C., R. H. Kraft, and D. L. Harris,
make quick and approximate predictions of 1957: Empirical methods for forecasting the
surge distribution along a coastline for a maximum storm tide due to hurricanes and
hurricane approaching landfall. It can be other tropical storms. Mon. Wea. Rev., 85,
applied to effectively and efficiently assess 113–116.
coastal surge risks involving computations
for tens of thousands of synthetic hurricanes. FEMA, URS and the U.S. Army Corps of
The model provides an alternative for Engineers, 2003: SLOSH Display Training,
applications where simple empirical models 2003.
fail to meet the needs and where the use of
complex numerical models is restricted by FEMA and the U.S. Army Corps of
time and resource constraints. Engineers, 2006: Hurricane Evacuation
Study Technical Data Report. Available
5. CONCLUSIONS at:http://chps.sam.usace.army.mil/USHESda
ta/HESHOME.htm
A simple coastline storm surge model
(SCSSM) has been developed to make Glahn, B., A. Taylor, N. Kurkowski, and W. A.
estimates of surge heights in hurricane- Shaffer, 2009: The Role of the SLOSH
prone sections of the coastline of the United Model in National Weather Service Storm
States. The model uses an empirical wind Surge Forecasting. National Weather
profile model to compute wind speeds at Digest, Volume 33, Number 1, 3-14.
coastline locations and derives surge
estimates by relating the wind speeds to Harris, D. L., 1963: Characteristics of the
surge heights through wind-surge functions. hurricane storm surge: U.S. Weather Tech.
For every coastline location where the surge Paper 48, 139 p.
height is estimated, a wind-surge function is
obtained by linearly interpolating the surge Hsu, S.A., D. Braud, and B. Blanchard, 2006:
values of SLOSH MEOWs at the location. Rapid Estimation of Maximum Storm Surges
These SLOSH MEOWs are composites of Induced by Hurricanes Katrina and Rita in
pre-run SLOSH surge estimates contained 2005. NWA Electronic Journal of
in the SLOSH Display package produced by Operational Meteorology EJ9.
the National Weather Service. The SCSSM
model is basically a combination of empirical Irish, J. L., D. T. Resio, and J. J. Ratcliffe,
and numerical models. 2008: The influence of storm size on
hurricane surge. J. Phys. Oceanogr., 38,
This combination enables the SCSSM 2003–2013.
model to be applied as an alternative
between the traditional empirical models and Jarvinen, B. R., and M. B. Lawrence, 1985:
the sophisticated numerical models. SCSSM An evaluation of the SLOSH storm surge
enhances traditional empirical models by a model. Bull. Amer. Meteor. Soc., 66, 1408–
unique way of taking advantage of the 1411.
results from a complex numerical model. It
expands the capability of traditional Jelesnianski, C.P., 1972: SPLASH (Special
empirical models with increased accuracy. Program to List Amplitudes of Surges From
Its efficiency and effectiveness make it a Hurricanes): 1. Landfall Storms. NOAA
useful tool for assessing the storm surge risk Technical Memorandum NWS TDL-46.
involving computation of tens of thousands NWS Systems Development Office, Silver
of synthetic hurricanes. Spring, MD, 56 pp.

Jelesnianski, C. P., J. Chen, and W. A.

Shaffer, 1992: SLOSH: Sea, lake, and
overland surges from hurricanes. NOAA
Tech. Report NWS 48, 71 pp.

7
29th Conference on Hurricanes and Tropical Meteorology, 10–14 May 2010, Tucson, Arizona

Kaplan, J., and M. DeMaria, 1995: A simple

empirical model for predicting the decay of
tropical cyclone winds after landfall. J. Appl.
Meteor., 34, 2499–2512.

Knabb, R., J. Rhome and D. Brown, 2006:

Tropical Cyclone Report: Hurricane Katrina
23-30 August 2005. National Hurricane
Center, 20 December 2005, updated. Aug
2006.

Landsea, C. W, and Coauthors, 2004: The

Atlantic hurricane database reanalysis
project: Documentation for 1851–1910
alterations and additions to the HURDAT
database.

Simpson, R. H., 1974: The hurricane

disaster potential scale. Weatherwise, 27,
169 and 186.

Taylor, A., 2009: SLOSH Display Program

(1.61e). National Weather Service, National
Oceanic and Atmospheric Administration.