The integration of a hurricane wind hazard model with deep-water and nearshore wave models

Abstract

Along the Atlantic and Gulf of Mexico coastal regions of the United States, the main source of economic losses from natural hazards are produced by hurricanes. As hotels and other structures continue to be built on small islands and along the coastline of the continental USA, the value of exposed property continues to increase as does the need for a risk model to deal with the design and insurance issues for both flood and wind. A hurricane hazard model was developed for modeling the hurricane risk along the US coastline, and included the effects of changing sea surface roughness and the air-sea temperatures difference on the estimated surface-level wind speeds. This model has been extensively validated by comparing time histories of predicted wind speed and direction with the corresponding measured data from over 200 marine and land based anemometers during twenty different hurricanes. The model forms the basis of the design wind speeds given in the US National wind loading standard, ASCE-7. Under the support of National Aeronautics and Space Administration (NASA), a research is undergoing at ARA to develop an advanced severe storm coastal risk assessment methodology in the Federal Emergency Management Agency's HAZUS-MH model. This paper described one part of the research - the integration of the hurricane hazard model designed above with wave models to enhance coastal flood modeling. The model results were compared with real-time buoy data at National Data Buoy Center (NDBC) during several major hurricanes. The evaluation showed that the linked models could provide a good representation of the hurricane wave fields, provided that good estimates of hurricane generated wind fields were available.

Full text

Page 1

1
The Integration of a Hurricane Wind Hazard Model with Deep-Water and
Nearshore Wave Models

Zhanxian Wang
Staff Scientist, Applied Research Associates, Inc.
IntraRisk Division, Raleigh, NC, USA

Peter J. Vickery
Principal Engineer, Applied Research Associates, Inc.
IntraRisk Division, Raleigh, NC, USA



Abstract - Along the Atlantic and Gulf of Mexico coastal
regions of the United States, the main source of economic losses
from natural hazards are produced by hurricanes. As hotels
and other structures continue to be built on small islands and
along the coastline of the continental USA, the value of exposed
property continues to increase as does the need for a risk model
to deal with the design and insurance issues for both flood and
wind. A hurricane hazard model was developed for modeling
the hurricane risk along the US coastline, and included the
effects of changing sea surface roughness and the air-sea
temperatures difference on the estimated surface-level wind
speeds. This model has been extensively validated by comparing
time histories of predicted wind speed and direction with the
corresponding measured data from over 200 marine and land
based anemometers during twenty different hurricanes. The
model forms the basis of the design wind speeds given in the US
National wind loading standard, ASCE-7. Under the support of
National Aeronautics and Space Administration (NASA), a
research is undergoing at ARA to develop an advanced severe
storm coastal risk assessment methodology in the Federal
Emergency Management Agency’s HAZUS-MH model. This
paper described one part of the research - the integration of the
hurricane hazard model designed above with wave models to
enhance coastal flood modeling. The model results were
compared with real-time buoy data at National Data Buoy
Center (NDBC) during several major hurricanes. The
evaluation showed that the linked models could provide a good
representation of the hurricane wave fields, provided that good
estimates of hurricane generated wind fields were available.
I. INTRODUCTION
Along the Atlantic and Gulf of Mexico coastal regions of
the United States, the main source of economic losses from
natural hazards are produced by hurricanes. As hurricanes
make landfall, coastal structures are subject to both direct
wind loads and flooding. The rise in the still water level
associated with storm surge and wave setup combined with
normal astronomical tides may result in water infiltration
should the base of a building become submerged. The rise in
the still water level will also move the coastal wave zone
landward, perhaps leading to waves breaking directly onto
structures and to wave runup washing out foundation
materials. The potential for wave damage is more
pronounced for coastal structures located on small islands,
such as the Caribbean, since the land mass is relatively small
and the surrounding water is deep. As hotels and other
structures continue to be built on small islands and along the
coastline of the continental US, the value of vulnerable
property continues to increase as does the need for a risk
model to deal with building design and insurance issues for
both flood and wind.
Owing to the lack of long-term, full-scale hurricane wind
velocity data, the probabilities associated with hurricane
winds of a specified magnitude at a specified geographic
location is most often assessed using a mathematical
simulation of hurricanes coupled with a hurricane wind field
model [1] [2]. Various national standards for assessing
environmental loads have adopted this approach for
estimating design wind speeds (e.g., [3], [4] and [5]). The
purpose of this work is to extend the accepted approach for
modeling the risk of hurricane-induced winds to also model
the risk of hurricane-induced coastal flooding. The latter of
which, for the USA, has traditionally been modeled
independently from wind risk using the joint probability
method [6]. Since the damage potential to coastal structures
during hurricanes strongly relates to both wind loads and
flooding, it is clearly necessary to model both perils
simultaneously in a risk-consistent manner to begin to gain
an understanding of hurricane-induced economic losses
along coastal locations. HAZUS-MH is a GIS based hazard
model, developed by FEMA that can be used by local
emergency managers and planners to assess the combined
risk of future losses associated with the combined action of
hurricane winds and hurricane induced flood along the US
coastline.
As a part of this effort, this paper describes the
implementation of a hurricane wind model and two third-
generation wave models, WAVEWATCH-III and SWAN
into FEMA’s HAZUS-MH tool. Section II gives an
overview of the model components. Example validation
results are provided throughout Section III, and conclusions
of the study and future plans are presented in Section IV.

Page 2

2
II. MODEL DESCRIPTIONS
A. Hurricane Simulation Model
The mathematical simulation of hurricanes is the most
accepted approach for estimating wind speeds for the
assessment of hurricane risk. The approach was first
described in [7] and [8]. Since that pioneering study, others
have expanded and improved the modeling. The hurricane
simulation model used herein is described in detail in [1].
The storm tracks were modeled using a statistical model
having parameters which vary over the Atlantic Ocean,
Caribbean Sea and Gulf of Mexico. This approach allows the
full track of the hurricane or tropical storm to be modeled
beginning with its initiation over the ocean and ending with
its final dissipation, which can be over land or water.
The model was validated in part by comparing the
statistics of central pressure, translation speed, heading,
minimum approach distance and occurrence rate derived
from the simulated data with those derived from the
historical data given in the HURDAT database [9] during the
period 1886-1996. The comparisons were made along the
Atlantic and Gulf of Mexico coastlines at sites spaced at 92.5
km (50 nautical-miles) for the U.S. and at 185 km (100
nautical-miles) for Mexico. The results showed that the
model was able to reproduce the continuously varying
hurricane climatology along the coastline. The validation
studies have since been updated to include the period 1886-
2004.
B. Hurricane Wind Field Model
Hurricane wind speeds and directions are modeled over
the entire storm domain using the model described in [2].
This model draws from previous research [10] [11] and is an
improvement over the first-generation model [12]. Using a
finite difference scheme on a set of nested rectangular grids,
the model solves the equations of the horizontal motion of
air, vertically averaged over the height of the planetary
boundary layer. The basis of this approach is the assumption
that the large-scale structure of the hurricane wind field
changes relatively slowly over time and therefore the winds
may be considered very near steady-state at any instant.
Wind fields are solved using the finite difference approach
for a wide range of input parameters. The solved wind fields
are then stored to disk using a Fourier fitting approach and
accessed as needed for rapidly estimating winds. The input
parameters are storm heading, translation speed, radius to
maximum winds, central pressure deficit, Holland’s pressure
profile parameter B [13] and the air/sea temperature
difference.
The hurricane wind field model has been extensively
validated by comparing time histories of predicted wind
speed and direction with the corresponding measured data
from over 200 anemometers during twenty different
hurricanes. More than 90 of the validation comparisons are
documented in [2] along with a summary of the error
statistics for both the ten-minute mean and three-second gust
wind speed predictions. Validation studies have been
performed at sites located both over land and over water.
C. Deep Water Wave Model
The modeled wind field is used as an input to
WAVEWATCH-III for predicting
parameters including significant wave height, mean wave
period, dominant wave period, mean wave direction and
surface stress acting at air and
WAVEWATCH-III is a third generation wave model
developed by the National Centers for Environmental
Prediction of the National Weather Service for computing
deep-water wave spectra [14]. It is a further development of
WAVEWATCH-I developed at Delft University of
Technology and WAVEWATCH-II developed at NASA
Goddard Space Flight Center. WAVEWATCH-III accounts
for wave growth and decay due to surface wind stress,
bottom friction, non-linear wave-wave interactions and wave
dissipation. WAVEWATCH-III has been used as a wave
forecast model by National Centers for Environmental
Prediction (NCEP) where the wind fields are obtained from
NCEP’s Global Forecast System (GFS) at three hours
intervals and the Geophysical Fluid Dynamics Laboratory
(GFDL) model for individual hurricanes at hourly high
resolution (e.g. Hurricane Isabel study in [15]).
Reference [16] described the research effort through
which WAVEWATCH-III version 1.18 and the hurricane
wind field model were integrated and tested for Hurricane
Fran (1996) and Hurricane Bertha (1996). In 2002, version
2.22 of WAVEWATCH-III was released [17]. The newer
version includes the following updates: (1) fully recoded in
allocatable FORTRAN 90; (2) improved source term
integration; and (3) new propagation scheme to eliminate the
“garden sprinkler effect” and account for unresolved islands
and sea ice. The new version of WAVEWATCH-III has been
implemented in this study.
deep-water wave
water interface.
D. Coastal Wave Transformation
Characterization of breaking waves near the coast and the
resulting wave forcing, wave setup and runup are very
important to determining hurricane damage to the coastal
structures and the total flooding level. In order to account for
the nearshore wave behavior, especially inside the surf zone,
a suitable coastal wave model is required to simulate the
translation of deep-water waves to the nearshore locations.
Simulating Waves Nearshore (SWAN) is a third-generation,
spectral wave model that describes the evolution of two-
dimensional wave energy spectra under specified conditions
of winds, currents, and bathymetry. SWAN uses the bore-
based model [18] to model the energy dissipation in random
waves due to depth-induced wave breaking. References [19]
and [20] described the formulation and development of the
wave model and its validation with field data. The basic
scientific philosophy of SWAN is similar to that of WAM
[21], another third-generation deep water model, but quite
different than WAVEWATCH-III. The newest version of
SWAN is 40.41 [22]. The model allows the use of nested
grids to successively provide high-resolution results at
desired locations. In addition to spectral wave parameters,
SWAN also provides estimates of wave setup due to
radiation stresses.

Page 3

3
The coastal wave transformation model is nested with
WAVEWATCH-III, which provides the wave spectra at the
outer boundaries. The input wind fields are obtained from the
hurricane wind field model.
III. MODEL VALIDATIONS
The predicted wave parameters from the integrated model
package are compared with data buoys from National Data
Buoy Center (NDBC). The bathymetry data used in
generating the computational grids are NGDC Coastal Relief
Model and 2-mimutes Gridded Global Relief (ETOP02) data
from the National Geophysical Data Center (NGDC)
(http://www.ngdc.noaa.gov/ngdc.html). The grid size in
WAVEWATCH-III is taken as 20km, while four different
nested grid sizes are used for SWAN calculation: 8km, 2km,
500m and 100m. Several hurricanes occurred at North
Carolina and Florida coasts have been investigated (Fig. 1).
Due to space limitations, only results from two hurricanes are
discussed: Hurricane Ivan (2004) and Hurricane Fran (1996).
A. Hurricane Fran (1996)
Hurricane Fran (1996) formed from a tropical wave that
emerged from the west coast of Africa on 22 August and
moved across the Atlantic during the peak of the hurricane
season. It made landfall on the North Carolina coast as a
Category 3 hurricane while moving northward near 17 mph.

The center moved over the Cape Fear area around 0030
UTC 6 September, but the circulation and radius of
maximum winds were large and hurricane force winds likely
extended over much of the North Carolina coastal area of
Brunswick, New Hanover, Pender, Onslow and Carteret
counties. At landfall, the minimum central pressure is
estimated at 954 mb and the maximum sustained surface
winds are estimated at 115 mph.
Fig. 2 presents the comparison of the calculated wind
speed and direction at NDBC stations with the recorded data.
All wind speeds are 10-minute means adjusted to 10m above
sea level. The wind field model gives a very good prediction
of hurricane winds. Fig. 3 presents the WAVEWATCH-III
model results and recorded data from 4 NDBC C-MAN and
Buoy stations. The results show good agreements except at
C-MAN station FPSN7, where the model overpredicts the
wave height near landfall. This overestimate can be attributed
to inadequate modeling of shallow water processes such as
surf zone physics (which
WAVEWATCH-III). At station FPSN7 (water depth =
14.3m) during Hurricane Fran, the observed maximum wave
height to depth ratio Hs/d ≈ 0.6 indicates wave breaking
within the surf zone, whereas the predicted maximum wave
height to depth ratio is Hs/d > 0.6. Fig. 4 shows the results at
station FPSN7 from the SWAN model, which includes the
dissipation by wave breaking and provides better agreement
with the recorded data than does the deep-water wave model.
are not included in

Fig. 1. Hurricane tracks used in the model validation.

Page 4

4
Hurricane Fran, FPSN 7
0
12:00
5
10
15
20
25
30
35
40
18:000:006:0012:00
Time UTC, 9/5/1996-9/6/1996
W ind Speed (m/s)
Data
Model

Hurricane Fran, FPSN 7
0
12:00
60
120
180
240
300
360
18:000:006:0012:00
Tim e UTC, 9/5/1996-9/6/1996
W ind Direction (degree)
Data
Model
H urricane Fran, B uoy 41002
0
5
10
15
20
25
30
6:0012:0018:000:00
Tim e UTC, 9/5/1996-9/6/1996
W ind Speed (m /s)
Data
Model
H urricane Fran, B uoy 41002
0
60
120
180
240
300
360
6:0012:0018:000:00
Tim e UTC, 9/5/1996-9/6/1996
W ind Direction (degree)
Data
Model

Fig. 2. Wind validation for Hurricane Fran (1996).
B Hurricane Ivan (2004)
Hurricane Ivan (2004) was a classical, long-lived
hurricane that reached Category 5 strength three times on the
Saffir/Simpson Hurricane Scale. It was also the strongest
hurricane on record at the far south east portion of the Lesser
Antilles. Ivan caused considerable damage and loss of life as
it passed through the Caribbean Sea. Ivan made landfall in
the US as a 121 mph hurricane (Category 3) at approximately
0650 UTC 16 September, just west of Gulf Shores, Alabama.
By this time, the eye diameter increased to 40-50 nautical
miles, which resulted in some of the strongest winds
occurring over a narrow area near the southern Alabama-
western Florida border.
The wind model and WAVEWATCH-III model results
along with the recorded data for NDBC Buoys are shown in
Fig. 5 and 6, respectively. The model results have a good
agreement with the wind data and a fairly good agreement
with the wave data at Buoys 42003, 42036 and 42039.
However, the model under-predicts the wave heights by large
margins at Buoy 42001 which is located on the left side of
the hurricane track. Because the wind field inputs are in good
agreement with the observations, there must be other factors
contributing to this discrepancy. Fig. 7 shows a comparison
of the observed (left) and modeled (right) wave spectra at
6:00am, September 15, 2004. The recorded data indicate that
there are multiple wave components traveling to this
location. These may have come from background swells
generated during the early stage of the hurricane, not
modeled here. Comparisons
WAVEWATCH-III and SWAN are similar to shown in Fig.
6. These observations are all in deep water but cannot be
shown due to space limitations.
of the combined
IV. CONCLUSIONS AND FUTURE WORKS
A hurricane wind field model and two third-generation
spectral wave models have been successfully implemented in
this study. The accuracy of the combined model is validated
against real time-series of NDBC buoy data during several
hurricanes. The evaluation results show that the model can
provide a good representation of the hurricane wave fields in
both deep-water and shallow-water conditions, provided that
good estimates of hurricane generated wind field are
available.
The deficiencies of the wave model results can be
attributed to several factors: (1) the most important
deficiency is the wind fields used in the forcing in the wave
Models, however; comparison with measured wind speeds
suggest that this is not the case; (2) the accuracy of
underlying bathymetry data controls the wave shoaling,
refraction, and depth-induced breaking processes; (3) the
model only computes the waves generated by the hurricane
winds, while waves (such as tidal waves) already existed
before the hurricane arrives and those generated by local
winds are not included; and (4) uncertainties and
complexities of the natural environment and the simplifying
assumptions in the numerical techniques will always
introduce some errors in the simulation results.

Page 5

5
Hurricane Fran, FPSN7
0
3
6
9
12
15
0:00 12:000:0012:000:0012:000:00
Time UTC, 9/4/1996-9/7/1996
Significant W ave Height (m)
Data
M odel
Hurricane Fran, D SLN7
0
2
4
6
8
0:0012:00 0:0012:00 0:0012:00 0:00
Time UTC, 9/4/1996-9/7/1996
Significant Wave Height (m)
Data
Model


Hurricane Fran, FPSN7
0
4
8
12
16
20
0:0012:000:00 12:00 0:0012:00 0:00
Tim e UTC, 9/4/1996-9/7/1996
Dom inant Wave Period (s)
Data
M odel
H urricane Fran, D SLN 7
0
4
8
12
16
20
0:0012:000:00 12:000:00 12:000:00
Tim e UTC, 9/4/1996-9/7/1996
Dom inant W ave Period (s)
Data
Model


Hurricane Fran, B uoy 41002
0
2
4
6
8
10
0:00 12:000:0012:000:0012:000:00
Time UTC, 9/4/1996-9/7/1996
Significant Wave Height (m)
Data
Model
Hurricane Fran, B uoy 41010
0
2
4
6
8
10
0:0012:000:00 12:000:00 12:000:00
Time UTC, 9/4/1996-9/7/1996
Significant Wave Height (m)
Data
M odel


Hurricane Fran, B uoy 41002
0
4
8
12
16
20
0:0012:000:0012:000:0012:000:00
Tim e UTC, 9/4/1996-9/7/1996
Dom inant Wave Period (s)
Data
M odel
H urricane Fran, Buoy 41010
0
4
8
12
16
20
0:0012:000:00 12:000:0012:000:00
Tim e UTC, 9/4/1996-9/7/1996
Dom inant W ave Period (s)
Data
Model


Fig. 3. Wave validation for Hurricane Fran (1996) using only WAVEWATCH-III.