Hurricane Storm Surge Simulations for Tampa Bay

Hurricane Storm Surge Simulations for Tampa Bay

ROBERT H. WEISBERG* and LIANYUAN ZHENG

College of Marine Science, University of South Florida, 140 7th Avenue South, St. Petersburg,Florida 33701

ABSTRACT: Using a high resolution, three-dimensional, primitive equation, finite volume coastal ocean model withflooding and drying capabilities, supported by a merged bathymetric-topographic data set and driven by prototypicalhurricane winds and atmospheric pressure fields, we investigated the storm surge responses for the Tampa Bay, Florida,vicinity and their sensitivities to point of landfall, direction and speed of approach, and intensity. All of these factors werefound to be important. Flooding potential by wind stress and atmospheric pressure induced surge is significant for a category2 hurricane and catastrophic for a category 4 hurricane. Tide, river, and wave effects are additive, making the potential forflood-induced damage even greater. Since storm surge sets up as a slope to the sea surface, the highest surge tends to occurover the upper reaches of the bay, Old Tampa Bay and Hillsborough Bay in particular. For point of landfall sensitivity, theworst case is when the hurricane center is positioned north of the bay mouth such that the maximum winds associated with theeye wall are at the bay mouth. Northerly (southerly) approaching storms yield larger (smaller) surges since the winds initiallyset up (set down) water level. As a hybrid between the landfall and direction sensitivity experiments, a storm transiting up thebay axis from southwest to northeast yields the smallest surge, debunking a misconception that this is the worst Tampa Bayflooding case. Hurricanes with slow (fast) translation speeds yield larger (smaller) surges within Tampa Bay due to the timerequired to redistribute mass.

Introduction

The recent assaults by major hurricanes on thenorthern Gulf of Mexico, including Ivan, Dennis,Katrina, and Rita in September 2004 and June, July,August, and September 2005, respectively, left nodoubt as to the vulnerability the southeasternUnited States to hurricanes. Among the mostdamaging aspects of hurricanes are the winds andflooding. Coastal flooding includes the effects ofstorm surge, freshwater accumulations, and wavebreaking and run up. Of these flooding effects, thestorm surge is the major culprit. It is also what thewave effects depend on since, without surge, wavesdo not come in contact with coastal structures anymore than normal. So while wave run up adds to theoverall storm surge it would not be a major factorwithout the sea level increase induced by wind stressand atmosphere pressure. Here we examine thepotential for such hurricane storm surges in theTampa Bay, Florida, vicinity.

Tampa Bay, located on the central west Floridacoast (Fig. 1), is comprised of Pinellas, Hillsbor-ough, and Manatee Counties, the first two of whichare among the most populous counties in Florida.Tampa Bay is also the 4th largest of the U.S. ports,when ranked by tonnage. Its potential vulnerabilityto storm surge is manifested in its low-lyingsurroundings. Figure 2 shows the subaerial config-urations under uniform rises of sea level by 1.5 and

6 m above mean water. The 1.5 m level is essentiallya state of no flooding. The 6 m level shows thatmuch of the Tampa Bay surroundings are inundat-ed. A 6 m storm surge is certainly within the rangeof possibilities for the Gulf of Mexico as demon-strated by the hurricanes mentioned above. Fortu-nately, Tampa Bay has not been struck by a majorhurricane since 1921. Historical hurricane stormtracks also show that fewer storms recurve into thecentral west Florida coastline than hit the east coastof the U.S. or points farther west in the Gulf ofMexico. Tampa Bay will inevitably experienceanother hit, and a better understanding on howsurge might evolve is necessary for emergency pre-paredness.

Hurricane storm surge entails a redistribution ofwater mass. Wind stress acting on the sea surface,when confronted by a rigid boundary, causes waterto accumulate along that boundary such that thepressure gradient force associated with the surfaceslope, times the water depth, tends to balance thedifference between the wind and bottom stresses.This tendency results in a spatially inhomogeneoussurge evolution so that the Fig. 2 inundation is notrealizable. Since this evolution entails a complex,three-dimensional, time-dependent, nonlinear pro-cess dependent on local geometry relative to thehurricane winds, it is necessary to model the surgeevolution. The National Oceanic and AtmosphericAdministration (NOAA) uses the Sea, Lake andOverland Surges from Hurricanes (SLOSH) modelof Jelesnianski et al. (1992) to simulate storm surgefor all U.S. coastal regions, and it is the output from

* Corresponding author; tele: 727/553-1568; fax: 727/553-1189; e-mail: [email protected]

Estuaries and Coasts Vol. 29, No. 6A, p. 899–913 December 2006

� 2006 Estuarine Research Federation 899

SLOSH that is routinely used by local emergencymanagers for evacuation guidance. AlthoughSLOSH performs well, its use has drawbacks. Theroutine dissemination of SLOSH results providesonly the worst case of flooding, regardless oflandfall location. Such dissemination may be mis-leading since, depending on landfall location, sealevel can be either set up or set down. SLOSH forTampa Bay has relatively coarse resolution (about2 km) so it cannot accurately resolve the region’sbathymetry and elevations, particularly for thePinellas County barrier islands, intracoastal water-way, and other subtle, but important, features.

Recent advancements in estuarine and coastalocean modeling provide new tools for assessinghurricane storm surge potential. Three elementsare required. The first is a suitable, high-resolutionmodel, including provisions for flooding anddrying. The second is a high-resolution bathymetricand topographic data set, on which to position themodel grid. The third is an accurate specification ofthe surface wind and atmospheric pressure fieldsrequired to drive the model. These are described inthe next section. The third section providesa baseline simulation for Tampa Bay and exploresthe sensitivities of the storm surge responses to the

point of landfall, the direction and speed ofapproach, and the storm intensity. The last sectionsummarizes the paper, provides a set of conclusions,and offers recommendations on future hurricanestorm surge modeling.

Model Description and Configuration

STORM SURGE MODEL

We use the time-dependent, three-dimensional,primitive equation, finite-volume coastal oceanmodel (FVCOM) of Chen et al. (2003). It incorpo-rates the Mellor and Yamada (1982) level 2Kturbulence closure submodel modified by Galperinet al. (1988) and the Smagorinsky (1963) formula-tion to parameterize flow-dependent vertical andhorizontal mixing coefficients, respectively. A s-coordinate transformation is used in the vertical toaccommodate irregular bathymetry, and a nonover-lapping unstructured triangular grid is used in thehorizontal to accurately fit complex coastlines,headlands, and structures. FVCOM solves the

Fig. 1. The Tampa Bay estuary and the adjacent west Floridashelf. Filled circles denote the locations sampled as discussed inthe text.

Fig. 2. Subaerial configurations for the Tampa Bay regionunder uniform rises of sea level by 1.5 m and 6 m above meanwater (from the merged NOAA-USGS bathymetric-topographicTampa Bay demonstration project data set; Hess 2001).

900 R. H. Weisberg and L. Zheng

primitive equations by using a flux calculationintegrated over each model grid control volume.This ensures mass, momentum, energy, salt, andheat conservation in the individual control volumesand over the entire computational domain. Similarto the Princeton Ocean Model of Blumberg andMellor (1987), a mode-splitting method with exter-nal and internal mode time steps to accommodatethe faster and slower barotropic and baroclinicresponses, respectively, is used for computationalefficiency. FVCOM also includes provision forflooding and drying, a critical element of stormsurge simulation (e.g., Hubbert and McInnes 1999).For storm surge simulations we added the Holland(1980) representations for the wind and pressuredistributions of a prototypical hurricane.

With Boussinesq and hydrostatic approximations,the primitive equations of momentum and (water)mass conservation are:

L(Du)

Ltz

L(Du2)

Lxz

L(Duv)

Lyz

L(uv)

Ls{ fDv

~ {gDL(g { ga)

Lx{

gD2

r0

ð0s

LrLx

{s

D

LD

Lx

LrLs

� �ds

zLLs

Km

D

Lu

Ls

� �z Fu

ð1Þ

L(Dv)

Ltz

L(Duv)

Lxz

L(Dv2)

Lyz

L(vv)

Lsz

fDu ~ {gDL(g { ga)

Ly

{gD2

r0

ð0s

LrLy

{s

D

LD

Ly

LrLs

� �ds z

LLs

Km

D

Lv

Ls

� �z Fv

ð2Þ

LgLt

zL(Du)

Lxz

L(Dv)

Lyz

LvLs

~ 0, ð3Þ

where u, v, and v are the x, y, and s velocitycomponents; f is the Coriolis parameter; g is thegravitational acceleration; Km is the vertical eddyviscosity; r0 is the reference density; r is theperturbation density; D 5 H + g is the total waterdepth where g and H are the surface elevation andreference depth below the mean sea level, re-spectively; ga is the sea level displacement inducedby the atmospheric pressure perturbation’s invertedbarometer effect; and Fu and Fv are the horizontalmomentum diffusion terms.

The surface and bottom boundary conditions formomentum are

r0Km

D

Lu

Ls,Lv

Ls

� �~ tsx,tsy

� �and v ~ 0 at s ~ 0 ð4Þ

r0Km

D

Lu

Ls,Lv

Ls

� �~ tbx,tby

� �and v ~ 0 at s ~ �1, ð5Þ

where (tsx, tsy) and (tbx, tby) are wind stress and

bottom stress components, respectively. At the openboundary, sea surface elevation is calculated byapplying a radiation boundary condition usinga gravity wave propagation speed,

ffiffiffiffiffiffigh

p, where h is

the water depth. The horizontal velocity compo-nents, evaluated at the grid cell centers, arecalculated directly from Eqs. 1 and 2 without thevertical and horizontal diffusion terms.

PROTOTYPICAL HURRICANE MODEL

For the prototypical hurricane storm surgeresponses investigated here, the radial distributionsof wind and atmospheric pressure relative to thestorm center and the maximum wind speed arespecified following Holland (1980), such that:

Vw ~

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiAB(Pn { Pc)exp({ A=rB)

rarBz

r2f2

4

s{

rf

2ð6Þ

P ~ Pc z (Pn { Pc)exp({ A=rB) ð7Þ

Vm ~ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi(B=re)(Pn { Pc)

p, ð8Þ

where r is the radial distance from the hurricanecenter; Vw and P are the wind speed and atmo-spheric pressure as functions of r; ra is the airdensity; Pn and Pc are the ambient atmosphericpressure and hurricane central atmospheric pres-sure, respectively; Vm is the maximum wind speed; eis the natural logarithm base; and A and B are stormscale parameters related by A 5 (Rmax)B, where Rmax

is the radius of maximum winds.Wind stress is computed from:

~tt ~ Cdra~VVw

�� ~VVw, ð9Þ

where Cd, a drag coefficient dependent on windspeed, is given by Large and Pond (1981):

Cd|103~

1:2 ~VVw

�� § 11:0 m s�1

0:49z0:065 ~VVw

�� 11:0 m s�1

ƒ ~VVw

�� ƒ25:0 m s�1

0:49z0:065 | 25~VVw

�� § 25:0 m s�1

8>>>>>>><>>>>>>>:

ð10Þ

Tampa Bay Storm Surge Simulations 901

MODEL CONFIGURATION

While storm surge is manifested locally, conti-nental shelf circulation physics allow for remoteeffects. Consequently the model domain must belarge enough to contain the spatial extent of thestorm and the accumulation of surge response dueto nonlocal excitation. The domain must also belarge enough such that different storm progressionscenarios can be explored.

The model domain extends from the MississippiRiver delta in the north to the Florida Keys in thesouth, with an open boundary arching in between.The grid resolution increases from the open bound-ary toward the west Florida coast, with the highestresolution (of about 100 m) centered on the PinellasCounty Intracoastal Waterway and St. Pete Beach(SPB) to resolve the barrier islands. The resolutionwithin Tampa Bay is less than 300 m, whereas theresolution along the open boundary diminishes toabout 20 km. For the west Florida coastal region themodel domain transitions across the land-sea in-terface and extends to the 8 m elevation contour.This transition allows us to include a flooding-dryingalgorithm in the model, a detailed discussion ofwhich is given by Chen et al. (2004). Here we usea flooding threshold depth of 10 cm.

A total of 88,400 triangular cells with 44,713nodes comprise the horizontal, and 11 uniformlydistributed s-coordinate layers comprise the verti-cal. The model grid is superimposed on a jointNOAA--U.S. Geological Survey (USGS) bathymetric-topographic data set that in the Tampa Bay vicinityhas a 30 m resolution (Hess 2001). Since most ofthe populated regions have seawalls and theseseawalls are at a nominal height of 1.2 m abovemean sea level, we set the minimum land elevationas 1.2 m, which means that a minimum 1.3 m surge(the sum of the seawall height and threshold value)is required to cause flooding in this model.

Based on Courant-Friedrichs-Levy numerical sta-bility condition, the computational time steps of 1and 10 s are used for the external and internalmodes, respectively. Temperature and salinity arespecified to be constant at 20uC and 35 psu,respectively. Sea level elevation (storm surge) inthis model is controlled by wind stress andatmospheric pressure. The effects of tides, rivers,steric (density change) effects, and wave run up arenot modeled. These all occur in nature, and theywould be additive to what we present on the basis ofwinds and atmospheric pressure.

According to Eqs. 6, 7, and 8, Fig. 3 shows thedistributions of air pressure and wind speed asa function of radial distance from the hurricanecenter for category 2 and 4 hurricanes (with centerpressure Pc 5 961 mb for category 2 and Pc 5 935 mb

for category 4). Based on these distributions, Fig. 3also shows the spatial structure, relative to thecomputational domain, for the surface winds of sucha prototypical category 2 hurricane, when thehurricane center is located near the shelf break. Thisdemonstrates the need for a large enough computa-tional domain in order to translate a prototypicalhurricane from the deep ocean to the coast.

To establish the utility of our approach (FVCOMwith flooding and drying driven by prototypicalhurricane wind and pressure distributions) we referto the realistic Hurricane Charley storm surgesimulation for the Charlotte Harbor, Florida, regiongauged against observed data by Weisberg andZheng (2006a). As a further demonstration on theuse of FVCOM for the Tampa Bay circulation inresponse to buoyancy, tides, and winds under morenormal, everyday conditions, also quantitatively

Fig. 3. Profiles of atmospheric pressure and wind speed asfunctions of radial distance from the storm center for a pro-totypical category 2 (thin) and category 4 (thick) hurricane,followed by the spatial structure of the winds for such (Holland1980) prototypical category 2 hurricane positioned within themodel grid. P and Vmax represent the central pressures andmaximum wind speeds, and subscript 2 and 4 are for category 2and 4 hurricanes, respectively.


gauged against observations, we refer to Weisbergand Zheng (2006b).

STORM SURGE SIMULATIONS FOR TAMPA BAY

Hurricane storm surge is sensitive to intensity(Saffir-Simpson scale category), point of landfall,direction of approach, and speed of approach, aswell as to the hurricane’s dimension (eye radius, orradius to maximum winds). To investigate thesesensitivities we ran different scenarios in whichthese factors were varied.

SCENARIOS

Eleven hurricane scenarios are considered (Ta-ble 1). Figure 4 shows seven different storm tracks.Four of these, E1, E2, E3, and E4, are for stormsapproaching from west to east, making landfall atlocations bracketing the Tampa Bay mouth. Threeof these, D2, D4, and D5, are for storms approachingfrom different directions, and D3 is parallel to D2.

For E1, E2, E3, and E4 we consider category 2hurricanes translating at 5 m s21, making landfall atIndian Rocks Beach (IRB), Sarasota, Tampa Baymouth, and Tarpon Springs, respectively. For D2,D3, D4, and D5 we consider category 2 hurricanestranslating at 5 m s21, either traveling parallel to theaxis of the bay from the southwest (D2 and D3) orparallel to the coastline from the northwest (D4) orsoutheast (D5). Table 1 also identifies S2 and S3,which are the same as E1, but with differenttranslation speeds of 10 and 2.5 m s21, respectively.Table 1 identifies I2, which is the same as E1, but fora category 4 hurricane.

For categories 2 and 4, we use nominal centerpressures of 961 and 935 mb, respectively, andmaximum wind speeds and eye radii (Fig. 3) of42.7 m s21 and 48 km for category 2 and 58.3 m s21

and 41.1 km for category 4. For these scenarios, onceset, the storm specifications are held constant. Toapply these as model boundary conditions we updatethe wind and pressure fields with each time step.

Results

SENSITIVITY TO LANDFALL LOCATION

The baseline run to which all other simulationsare compared is E1, a category 2 hurricane,approaching from the west at 5 m s21, makinglandfall at IRB and proceeding east across the bayuntil it exits the computational domain. Planar viewelevation snapshots taken at different times in thesimulation are given in Fig. 5. These frames showthe elevations at hours 24, 28, 29, 30, 32, and 36from the initiation of the simulation, and show thesurge in relation to the storm location as itapproaches and passes over the bay. The asterisk isthe IRB landfall site and the closed circles denotethe storm center. As the hurricane moves eastwardfrom its initial location (about 520 km west of IRB)we initially see a small sea level rise at the stormcenter due to the inverted barometer effect. Byhour 24, when the storm center is located about90 km west of IRB, the surge in the vicinity of

Fig. 4. The eight different storm tracks considered. E1, E2, E3,and E4 are for eastward moving hurricanes making landfall atIndian Rocks Beach, Sarasota, Tampa Bay mouth, and TarponSprings, respectively. Tracks D2, D3, D4, and D5 are for hurricanesparalleling the axis of the bay or paralleling the coastline from thenorthwest or the southeast, respectively.

TABLE 1. Summary of 11 proposed hurricane experiment scenarios.

Track Intensity Landfall Location Approaching Direction Moving Speed (m s21)

E1 Category 2 Indian Rocks Beach From west to east 5E2 Category 2 Sarasota From west to east 5E3 Category 2 Bay mouth From west to east 5E4 Category 2 Tarpon Springs From west to east 5D2 Category 2 Bay mouth From southwest to northeast 5D3 Category 2 Tarpon Springs From southwest to northeast 5D4 Category 2 Parallel coastline From northwest to southeast 5D5 Category 2 Parallel coastline From southeast to northwest 5S2 Category 2 Indian Rocks Beach From west to east 10S3 Category 2 Indian Rocks Beach From west to east 2.5I2 Category 4 Indian Rocks Beach From west to east 5


Tampa Bay is more than 1 m, since a southerly,alongshore-directed, downwelling-favorable windstress begins to affect the coast, with its associatedonshore Ekman transport. By hour 28, one hourbefore the storm makes landfall, the winds shift toonshore (offshore) south (north) of the hurricanecenter, and the direct effect of these winds pileswater up (sets water down) against a windward(leeward) shoreline. Sea levels south of hurricanecenter, including those of Tampa Bay, SPB, andSarasota Bay, continue to rise as the sea surfaceslopes up with the tendency for the pressuregradients by the surface slope to balance the stressof the wind on the sea surface. Maximum elevationsof 3.5 m occur over large areas on the eastern sideand the upstream ends of Old Tampa Bay andHillsborough Bay, leading to flooding there. Re-gions north of the storm center, such as ClearwaterBeach, experience a sea level set down of up to 1 m(Fig. 5) due to offshore-directed winds.

Sampling the model at the Courtney CampbellCauseway (CCC), the Howard Frankland Bridge(HFB), and the Gandy Bridge (GB) shows in-undation there. Similarly parts of SPB are flooded.By hour 30, one hour after the storm makeslandfall, the center is located on the Pinellas Countypeninsula west of Old Tampa Bay, but east of thecoast. The winds within the bay continue to pilewater up, with surge heights as high as 4 m at theupper reaches of Hillsborough Bay. The winds overthe entire coastal ocean are now northerly andupwelling-favorable, with offshore Ekman transportcausing a pronounced sea level set down along thecoast, which dries the barrier island beaches.

As the storm continues eastward and eventuallypasses Tampa Bay the winds over the bay reversedirection. By hour 32 we see a rapid sea level setdown inside the bay (by as much as 2.5 m) overthose regions that had been flooded. With the windsnow directed toward the east side of the bay, theregion of flooding shifts to Manatee County. Byhour 36, when the storm center is about 130 kmeast of the landfall location, the winds over the bayare much weaker and sea levels are returning tonormal. In particular, the three bridge causeways,which had previously been flooded, reappear.

Time series of the model-simulated sea levelsampled at IRB, SPB, Egmont Key (EK), PortManatee, St. Petersburg (SP), Port of Tampa,CCC, HFB, GB, and the Sunshine Skyway Bridge(SSB) are shown in Fig. 6, the locations for which

are shown in Fig. 1. At the gulf coast locations (IRB,SPB, and EK) the sea levels begin to rise aroundhour 15. The subsequent storm-induced variationproceeds rapidly with the movement of the stormitself such that by hours 29.5, 31, and 38 sea leveldisplacements become negative at IRB, SPB, andEK, respectively. In between this 14.5 to 23 hinterval the maximum surge amplitudes are about0.7, 1.9, and 1.2 m at IRB, SPB, and EK, respectively.The smallest surge occurs at IRB since the windspeed is weakest there when the hurricane makeslandfall. Despite the wind speed at EK being largerthan at SPB, the surge at SPB exceeds that at EKbecause of the geometry. SPB is along a continuousbarrier island, while EK is at the bay mouth so watermust accumulate at the former, whereas it can flowpast the latter. Surge is a very localized phenome-non on the basis of the storm attributes relative tothe regional geometry, a finding that will bea repeated theme of these simulations.

The evolution of sea levels inside the estuaryexhibit similar behavior, with sea levels rising afterhour 15, reaching maxima around hour 30 (onehour after IRB landfall), and becoming negative byaround hour 38. Storm surge generally increasesfrom the bay mouth to the upper reaches of the bay,with maximum values exceeding 3.2 m at CCC(Fig. 6) and beyond, versus about 1.5 m at theSSB. Rather than a wall of water, as often describedby news reporters, storm surge entails the tendencyfor the pressure gradient force by the sea surfaceslope to balance the stress of the wind on the seasurface. The faster the storm moves the quicker thesurge evolves, but since a finite amount of time isrequired to transport water in order to build theslope, faster moving storms can result in smallersurges. Given that storm surge equates with surfaceslope the farther upslope the location the larger thesurge; the upper reaches of the bay show largersurge heights than either the barrier islands or thebay mouth. It is also noted that the largest rate ofrise of sea level (south of the storm) coincides withthe strongest winds at the time of landfall.

We next consider the results from E2, for whichall specifications are the same as for E1, except thehurricane makes landfall at Sarasota to the south ofTampa Bay. We again refer to Fig. 6 for time seriesrepresentations of the model-simulated sea levelssampled at selected locations on the barrierbeaches, within the bay, and at the bridge cause-ways. The surge magnitudes for the Tampa Bay

R

Fig. 5. Planar view of model-simulated elevation snapshots at hours 24, 28, 29, 30, 32, and 36 for a prototypical category 2 hurricane,approaching from the west at 5 m s21 and making landfall at Indian Rocks Beach. The asterisk denotes the landfall location. The filledcircles denote the location of storm center. The bold lines are elevation contours at 1-m intervals.


vicinity are much less for E2 than for E1 since thewinds are generally directed offshore to the north ofthe landfall point. With the cyclonic rotation of thewinds around the storm center, the west side of thebay is still susceptible to onshore winds so we seesurge heights of about 2.2 m at CCC and 0.7 m atSSB. The differences in surge times are due to thedifferences in time for which the winds are directedin a surge-favorable manner. For the PinellasCounty beaches, very little surge occurs becausethe winds are directed offshore after the initiallydownwelling-favorable winds in advance of landfall.At SPB, after the small Ekman transport-induced sealevel rise we see a set down of sea level sufficient tocompletely dry the point (slightly offshore) at whichthe model is sampled. Similarly all of the other

points sampled within the bay eventually show a setdown of sea level through the course of thissimulation. While the Sarasota vicinity is outsidethe scope of the present paper, the model alsoprovides (lower resolution) simulations for thatregion.

E3 is similar to E1 and E2 except that for the pointof landfall being at the Tampa Bay mouth midwayin between the points of landfall for E1 and E2.Referring to Fig. 6 we see that the surge time serieslie in between those of E1 and E2, as expected.

E4 addresses the question of what happens in theTampa Bay vicinity as the point of landfall movesfarther north, in this case to Tarpon Springs,located about 70 km north of the bay mouth.Maximum surge heights increase slightly at IRBand SPB, but they are slightly smaller inside the bay(Fig. 6). These differences are explainable on thebasis of the wind distribution relative to the pointssampled. Surge heights with the bay are smallersince the winds at the bay mouth are smaller.

When the hurricane approaches from west to eastwith constant translation speed, the storm surges inthe Tampa Bay vicinity are dependent upon thepoint of landfall. Landfall to the north of the baymouth results in larger surge than landfall at or tothe south of the bay mouth. Maximum surges withinthe bay correspond to situations for which landfalloccurs sufficiently north of the bay mouth such thatthe maximum winds at the eye wall are situated atthe bay mouth. In this set of examples the IRB pointof landfall generally results in the largest surgeheights. Surge heights generally begin to diminishas the landfall point proceeds farther north. FromFig. 3 we see that large winds for a prototypicalhurricane can persist for large distances beyond theeye radius. So, for instance, while a hurricane thatmakes landfall 150 km north at Cedar Keys mightnot produce hurricane force winds over Tampa Bay,the potentially tropical storm force winds would stillresult in elevated sea levels.

SENSITIVITY TO DIRECTION OF APPROACH

We now consider the direction of approach forprototypical category 2 hurricanes identical to E1

(with translation speed of 5 m s21), but instead ofapproaching IRB from the west, we direct onehurricane up the Tampa Bay estuary axis fromsouthwest to northeast (D2), direct a second oneparallel to D2, but displaced by a radius tomaximum winds to the northwest (D3), and directtwo others parallel to the coastline, either fromnorthwest to southeast (D4) or from southeast tonorthwest (D5). For D4 and D5, the hurricane centeris located about 3 km offshore as it passes the baymouth at hour 20.

Fig. 6. Time series of model-simulated elevations sampled atIndian Rocks Beach, St. Pete Beach, Egmont Key, Port Manatee,St. Petersburg, Port of Tampa, Courtney Campbell Causeway,Howard Frankland Bridge, Gandy Bridge, and Sunshine SkywayBridge (from top left to bottom right) for category 2 hurricanes,approaching from the west at 5 m s21 and making landfall asdescribed in Fig. 4 for E1 (solid), E2 (bold solid), E3 (dashed),and E4 (bold dashed).


Fig. 7. Same as Fig. 5, except for a category 2 hurricane traveling up the axis of the bay at 5 m s21.


Similar to Fig. 5, Fig. 7 shows planar views of theD2 model-simulated elevations for hours 24, 28, 29,30, 32, and 36. Prior to hour 24, the winds along thecoast were downwelling-favorable and sea level hadgenerally increased due to onshore-directed Ekmantransport. By hour 24, when the storm center islocated about 100 km southwest of the bay mouth,offshore directed winds some distance to the northof the bay mouth are beginning to set sea leveldown, whereas onshore winds to the south areadding to the sea level set up there. Within the baywe see sea level beginning to set down on theeastern shore while setting up on the western shore.By hour 28 a substantial sea level depression (setup) is seen to the north (south) of the bay mouth.Inside the estuary there now exists a substantialacross-bay sea level slope with some drying occur-ring along the eastern shore driven by an across bay-directed wind stress. Local surge maxima arebuilding both in Sarasota Bay and Old Tampa Baywhere the winds are directed onshore. Theseconditions continue to build through the time inwhich the storm center enters the bay (betweenhours 29 and 30). Once the storm center is in thebay, the winds along the coast are directed eitheroffshore (north of the bay) or along shore towardthe southeast (near the bay mouth) such that wateris transported offshore, lowering sea level outsideTampa Bay and in Sarasota Bay. By hour 32, with thestorm center in Hillsborough Bay, water is beingdriven out of Old Tampa Bay and onto the easternshore of Tampa Bay. By hour 36 these across baytransports result in flooding along the eastern shoreand drying in Old Tampa Bay and also in SarasotaBay.

Time series representations of these directionalsensitivity experiments are provided in Fig. 8, usingthe same format as in Fig. 6. When compared withE1 we see that the positive portions of the stormsurge inside the bay simulated for D2 are much less,whereas the negative portions are much larger,particularly along the coast at IRB, SPB, and EK.The D2 surge heights at the bridge causeways aremuch less than those for E1. The folklore oftenexpressed by media reporters that a hurricanepassing directly up the bay is the worst case isshown to be false. To the contrary, at least withregard to storm surge, three factors tend to mitigatelarge flooding. In advance of the storm arriving atthe bay mouth, while the winds are directedalongshore and are downwelling-favorable, the Ek-man transport related sea level set up is much lessthan the set up due to onshore directed wind stress.Once the storm center is in the bay, the strongestwinds are directed across the bay, so instead ofdriving coastal ocean water into the bay these windsredistribute water across the bay. Once the storm

center is in the bay, the sea levels along the coastalocean are being set down, forcing water out of thebay.

By translating the D2 track line northwestward,the winds at the mouth of the bay become orientedup the axis of the bay throughout the stormevolution, and the D3 surge is much larger thanthat for D2. Relative to E1, D3 results in a comparablesurge height, either higher on the Hillsborough Bayside or slightly lower on the Old Tampa Bay owingto the different orientations of these Tampa Baysegments. Because of these orientation differencesit is difficult to define a worst case direction sincewhat may be the worst case for one segment of thebay may not be for another.

The comparisons between storm surge simula-tions for hurricanes paralleling the coastline fromnorthwest to southeast (D4) and from southeast tonorthwest (D5) are also revealing. Relative to E1, D4

and D5 result in larger and smaller positive surgeheights, respectively. For locations both inside andoutside Tampa Bay we see opposite evolutions insurge height for these two different directions ofshore parallel storm movement. One causes a sealevel set up, while the other causes a set down. Theexplanation lies with the initial set up or set down asthese storms approach the bay mouth. For D4 thewinds are directed onshore in advance of the stormso sea level is continually rising as the stormapproaches. For D5 the winds are directed offshorein advance of the storm so sea level is continuallyfalling as the storm approaches. The initial condi-tions for the largest part of the surge responses aredifferent among these two cases. At CCC the D4

(D5) generated surge is about 0.4 m higher (1.5 mlower) than that of E1 (see Fig. 8). Also along thePinellas County beaches the D4 generated surgeheights tend to be slightly larger than those of E1.

Tampa Bay vicinity storm surges are sensitive tothe direction of storm approach. While no stormscenario is a good one, the translation of a hurricanedirectly up the axis of the bay is the best of the worstcases with regard to storm surge. This stormscenario results in a redistribution of water fromwithin the bay, as contrasted with additional coastalocean water being transported into the bay. D2 isa hybrid case between point of landfall and di-rection of approach. From the point of view ofapproach direction, all other factors being equal,the worst case for the Tampa Bay region as a wholeis that of shore parallel from the northwest forreasons given above. Storm approach angles shiftingfrom shore parallel from the southeast to shoreparallel from the northwest result in increasingoverall storm surge (albeit specific locations in thebay may have larger surge for other cases dependingon the individual compartment orientations, e.g.,


Hillsborough Bay for D3). This finding would be justthe opposite for the east coast of Florida. There,southerly approaching storms would yield worsesurges than northerly approaching storms for thesame reason given here, namely that the initialconditions would be one of sea level increase versusdecrease.

SENSITIVITY TO APPROACH SPEED

So far we considered an approach speed of5 m s21. What happens if we either double or halvethis value? Consider S2 and S3, which are identicalto E1 with the exception of the approach speedbeing 10 m s21 for S2 and 2.5 m s21 for S3. Becauseof different approach speeds the landfall times aredifferent, as marked on Fig. 9 for the selectedlocations and bridge causeways as in Fig. 6.

Since storm surge entails a redistribution of mass(in order to generate a sea surface slope) the timefor which a given wind acts on a given coastal oceanor estuary geometry is relevant. For the Pinellas

County beaches, even though the approach speedmay vary, the surge height responses are similarbecause the winds are acting over a sufficient timeto raise sea level there. Once we begin to considerheight variations within Tampa Bay we see a dramat-ic effect of storm translation speed. While somewhatoversimplified, the leading edge of the storm raisessea level in the bay. The trailing edge then lowerssea level. If the trailing edge arrives before theleading edge can fully effect the mass redistributionthen the surge height will be suppressed. The slowerthe translation speed the larger the surge within thebay as seen at the various bridge causeways (Fig. 9).At CCC, the surge height for the slower movingstorm (4.2 m) exceeds twice that of the fastermoving storm (2.0 m).

The mass redistributions by winds depend onboth the along and across shore components ofwind stress. Surge due to the along shore compo-nent sets up within a fraction of a pendulum day

Fig. 8. Same as Fig. 6, except for the D2 (bold solid), D3

(dashed), D4 (bold dashed), and D5 (dot dashed), respectively.

Fig. 9. Same as Fig. 6, except for a category 2 hurricane,making landfall at Indian Rocks Beach and approaching from thewest at 5 m s21 (solid), 10 m s21 (bold solid), and 2.5 m s21

(dashed), respectively. Arrows denote the landfall time.


through the Coriolis acceleration. Friction limits themagnitude of the resulting along shore current andthe across shore sea surface slope in geostrophicbalance with it. In shallow enough water the acrossshore component of wind stress effectively trans-ports water downwind resulting in larger surfaceslope and larger surge. The time scale for this is onthe order of hours. Mass redistributions within theshallow, geometrically complex estuary are some-what longer than along the coast. When acted on bya steady, axially directed wind stress the timerequired to effect the mass redistribution to steadystate within a rectangular bay can be estimated frommass and momentum conservation arguments as:T ~ L2b pgh= , where L is length of the bay, b isturbulent resistance coefficient, g is the accelerationof gravity, and h is the mean depth (Feng 1982). ForTampa Bay with L 5 50 km, b 5 1.5 3 1023 s21, andh 5 5 m, T is approximately 6 h, implying that thesurge in Tampa Bay can fully develop if the leadingedge of the storm takes longer than 6 h to transitthe bay. For the storm translation speeds of 2.5, 5,and 10 m s21 used here, it takes roughly 5, 2.5, and1.25 h, respectively, for the storm to transit the bay.Consequently the faster storms do not have suffi-cient time to fully affect the storm surge. In contrastwith this inside the bay finding, outside the bay andalong the coastline the set up time is more rapid(since the water is deeper) and the surge is not assensitive to storm translation speed (Fig. 9).

SENSITIVITY TO HURRICANE INTENSITY

Our last sensitivity analysis compares the effects ofprototypical category 2 and 4 hurricanes, approach-ing from the west at 5 m s21 and making landfall atIRB. The simulation for the category 4, I2 storm iscompared with the simulation for E1. Planar viewelevation snapshots for I2, similar to those of Fig. 5for E1, are shown in Fig. 10. The evolutions aresimilar except that the surge heights and the areasflooded are larger for I2. At hour 28 the PinellasCounty beaches to the south of IRB have surgeheights of up to 3.5 m above mean water. Fromhours 29 through 32 we see a merging of waterbetween the intracoastal waterway and Tampa Bay.As evidenced by the sea level gradients, this occursfirst from the intracoastal waterway side when thebeaches are flooded and then from the bay sidewhen the west side of Old Tampa Bay is flooded. Forthis brief period of time SP is an island, and thenortheast portions of SP are flooded with surgeheights of about 5 m above mean water. Also, for I2

we see that large portions of Oldsmar at the head ofOld Tampa Bay, South Tampa between Old TampaBay and Hillsborough Bay, and downtown Tampa atthe head of Hillsborough Bay are flooded with surgeheights up to 6 m above mean water. After hour 32,

when the storm has crossed over the bay, surgeheights begin to abate over the northern portions ofthe bay while building on the eastern side of thebay. Flooding occurs along the Manatee Countyregions of the bay and in particular within theManatee River, where surge heights are againapproaching 6 m above mean water.

Time series comparisons sampled at specificpoints are given in Fig. 11. Once the surge begins,it advances and declines quickly with the stormtranslation across the Tampa Bay region. By thetime of IRB landfall, the surge in the bay is alreadynear its peak value, with water beginning to risesome 10–15 h prior to IRB landfall. Depending onposition the peak surge lasts for some 3–6 h (forthis translation speed; slower moving storms havelonger lasting surges, and conversely for fastermoving storms). The main difference between theE1 and I2 results is the surge magnitude andconsequently the areas flooded. The surge heights,in response to wind stress, scale approximately withthe square of the wind speed.

Conclusions and Recommendations

Our use of the finite volume coastal ocean modelof Chen et al. (2003) to study the hurricane stormsurge potential for the Tampa Bay region hasallowed us to investigate the sensitivity of hurricanestorm surge to point of landfall, direction and speedof approach, and intensity. For landfall (all otherfactors being equal) the worst case is when thestorm center is positioned a radius to maximumwinds north of the bay mouth such that themaximum winds are at the bay mouth. Stormsmaking landfall farther north result in lesser surges.Storms making landfall to the south of the baygenerally depress sea level along the Pinellas Countybeaches, although localized surge can still occur inthe bay by a combination of onshore Ekmantransport in advance of landfall and localizedonshore winds as the storm translates past the bay.Surge is also sensitive to the approach direction withstorms approaching from the south have lessersurge than those approaching from the north. Theworst approach direction depends on the specificbay location. Speed of approach is also important. Ifa storm moves quickly enough, such that thetranslational time scale is shorter than the surgeset up time scale, the full storm surge potential isnot realized. Since surge is in response to windstress, surge height tends to scale with the square ofthe wind speed (equations 9 and 10). Storms withstronger winds result in quadratically larger surgeheights; a category 2 storm experiment suggestedlimited flooding, while the flooding potential fora category 4 storm may be catastrophic. Hurricane


Fig. 10. Same as Fig. 5, except for a category 4 hurricane.


storm surge is also sensitive to the storm size (i.e.,radius to maximum winds).

Hurricane Charley provides an example for whichall of these factors apply. Hurricane Charleyapproached the Charlotte Harbor estuary rapidlyfrom the south, made landfall just south of BocaGrande Pass, and then translated up the estuaryaxis, with a relatively small radius to maximumwinds. Despite its category 4 status and its resultantwind damage, the storm surge was small (Weisbergand Zheng 2006a).

Two strategies may be considered for futurehurricane storm surge estimation. The first entailsreal time predictions in advance of an actual storm.The second entails scenario developments as pro-vided here. Recognizing that storm propertiescannot be predicted accurately enough in advanceof the time required to evacuate, hurricane stormsurge scenario developments are necessary foremergency preparedness. Improving the accuracy

of coastal ocean weather forecasts is necessary steptoward improving hurricane property specifications.This requires data for assimilation, and hencecoastal ocean observing systems (cf. Bortone 2006).

Additive to wind stress and pressure inducedsurges are the effects of tides, rivers, and waves. Tideranges for Tampa Bay of about 0.5 m to 1 m aresmall relative to the surge potential by winds andpressure. Flooding by fresh water may be importantlocally, and additional flooding and damage bywaves can be severe. Wave models need to becoupled with storm surge models to analyze thecombined effects, as was demonstrated clearly byhurricanes Ivan in 2004 and Katrina in 2005. Withsurge heights at bridge causeway levels, pounding bywaves destroyed sections of bridges. From thisexample, we can speculate that the Tampa Baybridges would be in jeopardy under severe hurri-cane conditions.

ACKNOWLEDGMENTS

Support was provided by the Office of Naval Research grantsN00014-05-1-0483 and N00014-02-1-0972. The second of these isfor the Southeast Atlantic Coastal Ocean Observing System(SEACOOS). Changsheng Chen, University of MassachusettsDartmouth, kindly provided the FVCOM code along with helpfuldiscussions.

LITERATURE CITED

BLUMBERG, A. F. AND G. L. MELLOR. 1987. A description of a three-dimensional coastal ocean circulation model, p. 1–16. In N.Heaps (ed.), Three-dimensional Coastal Ocean Models, Coastaland Estuarine Series Volume 4. American Geophysical Union,Washington, D.C.

BORTONE, S. A. 2006. Recommendations on Establishing a Re-search Strategy in the Gulf of Mexico to Assess the Effects ofHurricanes on Coastal Ecosystems. Estuaries and Coasts 29:1062–1066.

CHEN, C. S., H. D. LIU, AND R. C. BEARDSLEY. 2003. Anunstructured, finite-volume, three-dimensional, primitive equa-tion ocean model: Application to coastal ocean and estuaries.Journal of Atmospheric and Oceanic Technology 20:159–186.

CHEN, C. S., R. C. BEARDSLEY, AND G. COWLES. 2004. Anunstructured grid, finite-volume coastal ocean model: FVCOMuser manual. SMAST/UMASSD University of Massachusetts-Dartmouth, Technical Report-04-0601. New Bedford, Massa-chusetts.

FENG, S. Z. 1982. An Introduction to Storm Surge, 1st edition.Science Press, Beijing, China.

GALPERIN, B., L. H. KANTHA, S. HASSID, AND A. ROSATI. 1988. Aquasi-equilibrium turbulent energy model for geophysicalflows. Journal of Atmospheric Science 45:55–62.

HESS, K. 2001. Generation of tidal datum fields for Tampa Bayand the New York Bight. National Oceanic and AtmosphericAdministration, National Ocean Service, Technical Report:NOS CS 11. Silver Spring, Maryland.

HOLLAND, G. J. 1980. An analytic model of the wind and pressureprofiles in hurricanes. Monthly Weather Review 108:1212–1218.

HUBBERT, G. D. AND K. L. MCINNES. 1999. A storm surgeinundation model for coastal planning and impact studies.Journal of Coastal Research 15:168–185.

JELESNIANSKI, C. P., J. CHEN, AND W. A. SHAFFER. 1992. SLOSH: Sea,lake, and overland surges from hurricanes. National Oceanic

Fig. 11. Same as Fig. 6, except for category 2 (solid) andcategory 4 (bold solid) hurricanes approaching from the west at5 m s21 and making landfall at Indian Rocks Beach.


and Atmospheric Administration, Technical Report NWS 48.Silver Spring, Maryland.

LARGE, W. G. AND S. POND. 1981. Open ocean momentum fluxesin moderate to strong winds. Journal of Physical Oceanography 11:324–336.

MELLOR, G. L. AND T. YAMADA. 1982. Development of a turbulenceclosure model for geophysical fluid problems. Review ofGeophysics 20:851–875.

SMAGORINSKY, J. 1963. General circulation experiments with theprimitive equations, I: The basic experiment. Monthly WeatherReview 91:99–164.

WEISBERG, R. H. AND L. ZHENG. 2006a. A simulation of theHurricane Charley storm surge and its breach of North CaptivaIsland. Florida Scientist 69:152–165.

WEISBERG, R. H. AND L. ZHENG. 2006b. A finite volume coastalocean model simulation of the tide, buoyancy, and wind-drivencirculation of Tampa Bay. Journal of Geophysical Research 111:C01005, doi:10.1029/2005JC003067.

Received, January 4, 2006Revised, May 24, 2006Accepted, June 5, 2006


Hurricane Storm Surge Simulations for Tampa Bay

Documents

Transcript of Hurricane Storm Surge Simulations for Tampa Bay