March 13, 2010 10:34 B-936 b936-ch12

PART V

Hydrological Aspects of Tropical Cyclones March 13, 2010 10:34 B-936 b936-ch12 March 13, 2010 10:34 B-936 b936-ch12

Chapter 12

Storm Surge Modeling and Applications in Coastal Areas

Shishir K. Dube Centre for Atmospheric Sciences, Indian Institute of Technology, New Delhi, India [email protected]

Tad S. Murty Department of Civil Engineering, University of Ottawa, Ottawa, Canada

Jesse C. Feyen Office of Coast Survey Development Laboratory, National Ocean Service, National Oceanic and Atmospheric Administration, Silver Spring, Maryland, USA

Reggina Cabrera Eastern Region Headquarters, National Weather Service, National Oceanic and Atmospheric Administration, Bohemia, New York, USA

Bruce A. Harper Systems Engineering Australia Pty Ltd, Brisbane, Australia

Jerad D. Bales North Carolina Water Science Center, U.S. Geological Survey, Raleigh, North Carolina, USA

Saud Amer International Water Resources Branch, U.S. Geological Survey, Reston, Virginia, USA

This chapter introduces the reader to a wide spectrum of storm surge modeling systems used to assess the impact of tropical cyclones, covering a range of numerical methods, model domains, forcing and boundary conditions, and purposes. New technologies to obtain data such as deployment of temporary sensors and remote sensing practices to support modeling are also presented. Extensive storm surge modeling applications have been made with existing modeling systems and some of them are described in this chapter. The authors recognize the importance of evaluating river-ocean interactions in coastal envi- ronments during tropical cyclones. Therefore, the coupling of hydraulic (riverine) and storm surge models is discussed. In addition, results from studies performed in the coast of India are shown which generated maps to help emergency managers and reduce risk due to coastal inundation.

363 March 13, 2010 10:34 B-936 b936-ch12

364 Shishir K. Dube et al.

1. Introduction defined as the rise in water level caused by the surface wind stress, atmospheric pressure deficit, Storm surges resulting from tropical cyclones and momentum transfer from breaking surface expose the population in coastal areas to wind waves (wave setup). The term storm tide flooding and destruction. Accurate and under- is commonly used to describe the water level standable forecasts and flood information change caused by the combined effects of the products play an important role in preparing storm surge and the astronomical tide (Hicks for and mitigating these events, including flood et al., 2000, p. 25). Both the storm surge and protection and evacuation. The purpose of this storm tide refer to a long gravity wave (so called chapter is to discuss theory for modeling tropical because its wavelength is much greater than cyclone storm surge and the application of river depth) that is of similar size to the storm itself. flow and storm surge data for modeling flooding Wind-generated short waves (with wavelengths conditions in coastal areas. smaller than depth), while possibly devastating, Although numerous models are available for are considered a separate phenomenon because forecasting storm surge water levels in coastal of their short time and spatial scale. Figure 1 areas, many of the models still remain in the illustrates the relevant contributions to the total field of research. One of the primary distinc- water level during an extreme storm surge event, tions between research and operational models and further description follows. is that operational models must perform under In deep water, storm surge is minimized all conditions, which constitutes an ongoing tes- because the effects of the wind stress are dis- tament to their robustness. In the operational tributed over depth and counter currents can a framework , there needs to be assurance that be generated in deep water. The primary water forecast products will be available at all times surface elevation response in deep water is for use by emergency managers. Conversely, the atmospheric pressure deficit which causes research applications are able to explore areas the so-called “inverted barometer” effect; the of technological advancement that benefits both rise is approximately 1 cm for each hPa drop future operations (operational models can be in pressure (see, e.g., Pugh, 2004). However, continuously improved by research) as well as as the storm passes over the shelf, reduced other significant needs. These include but are bathymetric depths lead to a set-up of the not limited to calculations of long term risk, water surface due to the currents caused by emergency planning, and coastal management the surface wind stress. A large shallow shelf and engineering. can therefore enhance surge generation. Local Even though the authors recognize the dynamics become increasingly important as the importance of precipitation forecast due to surge approaches the coast. The bathymetry, tropical cyclones, this topic is not discussed in coastline, topography, and inundation barriers this section. This section will address only storm such as levees and dunes are capable of mod- surge models and their coupling with hydraulic ifying the surge. Surges of several meters are models for the application of coastal inundation. commonly generated by land-falling tropical cyclones, and extreme surges approaching 10 2. Storm Surge Simulation meters can occur under certain conditions. The storm’s strength, size, track/angle of attack and As a tropical cyclone approaches land, its impact forward speed effectively determines the surge on the sea leads to a rise in water level at at any one location. the coast that can be extremely severe, inun- Other processes can affect the total water dating vast low-lying areas. This storm surge is level and resulting damage caused by land- falling tropical cyclones. First, wind-generated aThe term “operational” in the context used in this chapter, means executed in real time. waves can be severe and significant. They March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 365

Figure 1. Contributions to the total water level during a severe storm surge event as it reaches land (gray shading). The individual effects vary significantly, depending on a variety of factors and this illustration cannot be construed as specific to any one case.

can contribute to the surge by a transfer of that Hurricane Katrina had on coastal Louisiana momentum, mainly when breaking, which is and Mississippi, and on New Orleans in par- called wave set-up. They also are extremely ticular (Knabb et al., 2006). destructive when impacting structures inun- Considering the severe impact of storm dated by the surge. Second, flooding from surge, it is important to develop effective mit- rainfall can occur in conjunction with a storm igation systems for supporting resilient coastal surge. While there is generally a delay in the communities. This requires an analysis of the flooding caused by heavy tropical cyclone pre- vulnerability due to storm surge activity at the cipitation of a few days after landfall, satu- coast. However, the frequency of severe storm rated conditions and flooding caused by previous activity at any particular location may be lim- events can exacerbate inundation. Third, the ited, hampering statistical analysis because of total water level is also affected by seasonal limited historical observations and precluding and annual variations such as persistent (e.g., adequate in situ data acquisition (and thus trade) winds and solar heating of the large ocean effective data assimilation analysis strategies). basins, which can cause a variation in water Storm surge is so dependent upon the indi- levels compared to the long-term mean. vidual characteristics of each storm and geo- Storm surge and the superimposed wind graphical location that surge heights from a waves cause most of the damage from handful of historical hurricanes may not reli- tropical cyclones at landfall. Extensive low-lying ably predict future inundation risk. Predictive coastal areas (e.g., Louisiana, Bangladesh, and modeling capability is needed to study the Myanmar) can be flooded, destroying commu- potential threat due to storm surge as a func- nities, damaging property and infrastructure, tion of meteorological forcing and geographic and resulting in casualties. Furthermore, coastal conditions. development is rapidly increasing and popu- A model capable of accurate storm surge lation centers are growing (Crossett et al., 2004). simulation is important for minimizing risks These factors are combining to raise the eco- to coastal communities. These simulations are nomic and social risk caused by severe storm needed to forecast potential flooding from an surge. A clear example is the widespread impact incoming hurricane, to develop and implement March 13, 2010 10:34 B-936 b936-ch12

366 Shishir K. Dube et al.

effective evacuation plans, to design flood pro- on the principles of conservation of mass for tection and coastal management projects, and an incompressible fluid, and the conservation of to assess surge risk for flood insurance programs. momentum as described by the Navier-Stokes There are two distinct types of users of storm equations subject to Reynold’s averaging to surge simulations: real-time responders and describe the mean flow. Simplifications are made long-term planners. These users include emer- in the formulation of the SWE which lead to a gency managers; meteorologists and forecasters; more manageable set of equations: the hydro- land-use, coastal zone, floodplain, and natural static approximation makes the vertical accel- resource managers; engineers and public infras- eration and associated viscous terms in the tructure managers; insurance companies; and vertical momentum equation unimportant, the commercial companies and non-governmental Boussinesq approximation allows gradients of organizations. Their needs lead to a wide range fluid density to be ignored, and the shallow of models and applications in use, and the most water assumption leads to a depth-integrated common are discussed below. set of 2-D equations. The diffusion and dis- persion of momentum due to unresolved lateral scales of motion as well as the effects of depth- 2.1. Theory and Methodology averaging are accounted for by a turbulence The need for modeling storm surge and devel- closure model. The resulting SWE are applied to oping surge forecasting systems has led to a simulate ocean and estuarine circulation. When range of numerical modeling approaches uti- these equations are applied on a scale appro- lizing an array of techniques. The predominant priate for geophysical flow problems, the Coriolis approach is to solve the two-dimensional (2-D), term is included to account for the rotational depth integrated form of the shallow water acceleration of the earth, and they can be solved equations, which disregard baroclinic effects in a spherical coordinate system (with horizontal and variations in the velocity field over the dimensions of longitude and latitude) in order vertical direction. The numerical techniques to more accurately account for the earth’s cur- employed primarily include finite difference and vature. For further detail on the derivation of finite element algorithms but other approaches the SWE see, for example, Pedlosky (1987), have been used, such as finite volume schemes. Cushman-Roisin (1994), or Weiyan (1992). The domains chosen tend to be regional, shelf The SWE describe 2-D, depth-integrated cir- based models, although some models use nested culation, which is a valid approach for modeling or large scale unstructured grids in order to storm surge since the response due to hurricane apply far field boundary conditions (albeit forcing is of a relatively short time scale. This with limited, if any, dynamic two-way coupling limits the generation of three-dimensional (3-D) between domains). Bode and Hardy (1997) pro- circulation (namely, a return flow opposite to vides a review of developments in the field the surface gradient). Rather, storm surge prop- of storm surge modeling that identifies many agates as a shallow water wave with height on approaches. the order of 10 meters and length on the scale of hundreds of kilometers. This satisfies the con- dition describing shallow water conditions: hori- 2.1.1. Governing Equations zontal scales of flow are much larger than depth. Coastal ocean models are generally based on When computations of vertical variations in cur- the shallow water equations (SWE). These gov- rents are required a 3-D model can be used. erning equations are appropriate since they However, 2-D models are much more economical describe the dynamics in most free surface geo- and describe the dominant characteristics for physical flow problems. The SWE are based storm surge simulations. March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 367

In order to improve model efficiency, non- Westerink et al., 1994). This allows for high linear terms are omitted within some surge resolution of coastal features such as inlets and models. In particular, the nonlinear advective barrier islands while using much larger node terms can pose a constraint on computa- spacing further away in what may be very tional efficiency (namely the numerical time large grids. The method that proved to be step size) of a surge model algorithm. (The most successful and widely used for FE mod- nonlinear bottom friction and water surface eling of the SWE is the wave equation for- elevation terms are more generally included.) mulation, due to its suppression of spurious It is in principle preferable to implement a oscillations (Lynch and Gray, 1979; Kinnmark, fully nonlinear model, but linearizing the equa- 1986). However, it has been established that tions can be an acceptable choice if the com- for shallow, nonlinearly dominated flows there promise in accuracy is minor compared to the can be mass balance error, as the continuity improvement in simulation cost. In practice the equation is not being satisfied in the wave uncertainties inherent in the forcing, parame- equation, but rather its time derivative (Walters terizations, and grid design (including elevation and Carey, 1984; Kolar et al., 1994). Addi- specification) limit many gains in accuracy tionally, this FE approach has been found to possible through use of a fully nonlinear require very small time steps below the Courant model. condition, leading to costly simulations. When combined with the highly resolved grids that FE models generally apply, the computational 2.1.2. Numerical Methods resources required for FE storm surge sim- Computational models of the SWE have ulations can be significantly more than FD been in existence for over three decades, models, limiting their use in time-constrained employing various numerical methods for simu- applications. lating storm surge. The earliest numerical tech- Recent research has been focused on com- niques employed finite difference (FD) schemes. bining the strengths of FD (mass conserving) The modeling community’s experience with FD and the strengths of FE (high resolution modeling of the Navier-Stokes equations led to unstructured grid) methods. This has led to its early application to the SWE. The resulting various forms of finite volume (FV) techniques, computational codes have proven to be efficient but use of these models has not yet been and robust and are widely applied. However, widespread although results have been reliable. a disadvantage of FD methods comes from The same cost issues regarding the size of highly their use of structured grids. This hinders their resolved grids in FE modeling also apply to FV ability to refine the complex coastal geome- models as well. tries and localized flow patterns near the Considering the cost of storm surge appli- coast and over floodplains. Because these fea- cations, which can rely on large, highly refined tures significantly affect surge height, accuracy grids and time-consuming numerical schemes, can be affected by the grid resolution and state-of-the-art computational techniques are orientation. required to make these simulations feasible. The application of the finite element (FE) For example, parallel processing techniques are method to the SWE has allowed use of unstruc- being used to run models on distributed memory tured meshes able to map complex geometry computers. With these approaches computa- quite accurately. This produces domains that tional platforms among the most advanced minimize computational effort while maxi- available can be utilized, and wall clock times mizing accuracy by varying nodal spacing over are reduced by a factor close to the number of several orders of magnitude (Blain et al., 1998; processors. March 13, 2010 10:34 B-936 b936-ch12

368 Shishir K. Dube et al.

2.1.3. Grid Design around Pensacola Bay. This model domain contains approximately 63,500 cells. Storm surge elevations can be impacted by A storm surge model that uses an unstruc- complex local geometries, as flooding is con- tured grid methodology can apply fine res- trolled by geographic features. Therefore, simu- olution of the coastal region along with lations capturing these features improve storm far-field boundary conditions in the deep surge prediction, and the corresponding model ocean. The paradigm recommended is a large requires fine resolution. However, the size and domain/unstructured grid strategy that couples resolution of the model domain must be bal- the ocean, shelf, and coastal floodplain while anced by consideration of the computational resolving local features. However, this method- requirements. For example, forecasts are con- ology typically leads to costly model grids and it strained by timeliness requirements, but pre- requires sufficient bathymetric and topographic and post-storm investigations for engineering detail to support its high resolution. Figure 3 and risk determination purposes can benefit shows a finite element mesh from a storm surge from more complex approaches that require model of coastal Alabama and the Northwest greater computational resources. Florida Panhandle. Detail of the model grid is FD storm surge models apply structured shown for the inlet to Pensacola Bay, Florida, computational grids with regularly spaced along with the view of the larger model domain discretization points. Curvilinear FD models and the increased resolution for the inundation provide fine resolution near the coastline and region. This model grid has resolution of approx- coarser resolution as the grid extends away from imately 100 m at the coast but expands out to the coast. These grids can be cost efficient tens of kilometers in the deep basins, where the but if the boundary location is not extended open ocean boundary conditions are applied. far enough, these models may not correctly It has a total of approximately 450,000 nodes. capture the dynamics of the basin-shelf inter- Key to any numerical storm surge simu- action. However, nested modeling techniques lation is the representation of the bathymetry, can be employed which couple high resolution topography, and coastal features. The config- coastal grids with coarser larger scale domains. uration of bathymetry (sea floor depth), its These larger basin effects can produce a set-up resolution, and its age/correctness is a critical on the shelf which is not known aprioriand factor in surge generation. Similarly, the topog- therefore cannot be specified as a boundary con- raphy (land elevation) of a coastal region is dition. Additionally, structured grids can have essential to correctly simulating the overland difficulty sufficiently resolving irregular coastal surge. This is because flooding depths over features or may poorly represent them because land are defined in part by the shape of of the regular structure of the grid. Flood topographic contours and features impacting amplitude and propagation are controlled by inundation (e.g., levees, barrier island dunes, details such as the shoreline, inlets, and levees, raised roadways). Waterways can propagate so it is important to ensure the coast is resolved surge well inland, while concave coasts focus as accurately as possible. Therefore, some and amplify surge. Bathymetry and topography models link to subgrid-scale one-dimensional can change quickly due to both natural and (1-D) models to represent flooding barriers, con- anthropogenic causes, so it needs to be regu- veyance, or storage at a resolution higher than larly updated. Considering the accuracy (errors permitted by the native grid. Figure 2 shows an reaching less than one-third of a meter) and example of a structured orthogonal curvilinear resolution (reaching less than 100 meters) of mesh for the Northwest Florida Panhandle, widely used storm surge models (Jarvinen and including a close-up of the model resolution Lawrence, 1985; Westerink et al., 2008), data March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 369

Figure 2. Finite difference storm surge model grid: (a) orthogonal curvilinear grid domain and (b) grid detail around Pensacola Bay. Figures provided by the US National Oceanic and Atmospheric Administration National Weather Service.

collection and processing need to meet or exceed It is important that expensive data collection these constraints. Many new remote sensing efforts include: necessary ties to vertical datums; technologies are being employed for bathy- accurate post-processing to produce bare-earth metric and topographic data collection (e.g., data sets; and appropriate collation into a sidescan and multibeam sonar, Light Detection useful Digital Elevation Model (DEM) or similar and Ranging (LiDAR), interferometric radar). product. March 13, 2010 10:34 B-936 b936-ch12

370 Shishir K. Dube et al.

Figure 3. Finite element storm surge model grid: (a) unstructured grid domain, (b) high resolution grid in project region, and (c) model grid and bathymetry for the inlet to Pensacola Bay (m NAVD 88). Figure provided by the US National Oceanic and Atmospheric Administration National Ocean Service.

There are often a number of vertical datums datums also include 3-D ellipsoidal datums used used as a reference for elevation measurements. by global navigation satellite systems. Storm Bathymetry is generally referenced to a tidal surge simulations require consistent referencing datum, which is determined by computing the of all data at the coast, where tidal, ortho- mean of a stage of an observed tide. For metric, and ellipsoidal datums are all used. example, in the US, bathymetry is generally Therefore, adjusting all datasets (bathymetry, provided relative to Mean Lower Low Water topography, and water level observations) to (MLLW). Topography is usually referenced to a common vertical datum is required when a land-based geodetic datum, which is deter- building a surge model which describes prop- mined by surveying techniques from a national agation of water level from offshore to the set of established benchmarks. In the US this coast and inland. This can be done by careful geodetic datum standard is the North American examination of data at locations where ties Vertical Datum of 1988 (NAVD 88). While to multiple datums are provided, commonly NAVD 88 is constrained to Mean Sea Level at tidal gauges which have geodetic bench- (MSL) at one location along the US/Canadian marks. For example, the US National Oceanic border, the difference between NAVD 88 and and Atmospheric Administration (NOAA) is MSL varies along the US coast. Water level developing a vertical datum transformation tool observations are generally collected relative to called VDatum (Myers 2005) based upon these a tidal datum but peak surge values can be datasets to provide these adjustments for all of referenced to another datum. Finally, vertical the US coastline (see http://vdatum.noaa.gov). March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 371

2.1.4. Boundary Condition (hundreds to thousands of meters) is coarse Specification compared to features which locally impact flood- ing. Generally, the inundation front is described The boundaries to a storm surge model have by allowing a region to flood if a sufficient vol- a significant influence on model response. For ume of water can enter it from neighboring example, forcing at lateral boundaries (e.g., grid cells. astronomic tides, river inflow, and inverted During hurricane flooding conditions, storm barometer pressure signal) requires accurate surge may overtop levees, dunes, road systems boundary conditions in order to generate the and similar features which are local barriers to proper surge. Many models also employ spe- flood propagation. However, these barriers gen- cialized boundary conditions to characterize erally fall below grid scale and represent a non- geographic features that can not be adequately hydrostatic flow scenario. Therefore, it may be modeled at the mesh resolution (e.g., levees, effective to treat these structures as subgrid streams). Finally, storm surge simulation has scale objects within the domain, generally using the special case of a moving boundary at the boundary conditions which apply weir formulae. wet/dry interface which must be handled in Similar techniques have also been applied for a way that correctly models inundation and subgrid scale hydraulic features such as small recession. streams and channels. Initial conditions specify the water level before the storm arrival. Models generally apply an elevation-specified boundary condition at 2.1.5. Surface and Bottom Boundary the open ocean to define water level vari- Specification ations occurring outside the domain. Tidal constituent data from global ocean models The specification of boundary conditions defines or satellite altimetry can be applied if tidal the forcing entering a storm surge model forcing is desired. However, care must be taken domain. However, the simulation also requires to ensure that the constituents provided at the stresses at its surface and bottom boundaries the open ocean boundary are sufficient and that guide the circulation. Due to the signifi- accurate, since global databases are often not cance of the storm as a forcing mechanism, mete- accurate in shallow water and many areas of orological conditions provide the most signif- the world lack adequate shallow water observa- icant contribution to model accuracy. Tropical tions. Storm-induced set-up in the water level cyclone wind and pressure fields input to the surface can be applied as well, if known. Also, hydrodynamic model essentially control the gen- the inverted barometer effect of the hurricane’s eration of storm surge. A hurricane’s track, low pressure system can be readily applied strength, and size are critical factors which if known. determine storm surge height. A shift in track A model can define river inflows or a no- location to either side of a river, for example, normal flow condition for land boundaries. will significantly alter the impact that the storm However, a simple specified flux at river inflow has on that waterway. boundaries can reflect surface gravity waves The dominant forcing mechanism for storm back into the domain, so use can be made of a surge is the surface wind stress (the inverted wave radiation boundary condition that allows barometer and wave set-up effects generally con- such waves to pass out of the domain. tribute less than a meter of what may be a Modeling inundation requires simulation of several meter surge). Therefore, accurate defi- wetting and drying fronts. However, this is a nition of the wind field (including its strength, challenge because the size of grid resolution size, and shape) as well as parameterization March 13, 2010 10:34 B-936 b936-ch12

372 Shishir K. Dube et al.

of the wind stress at the water surface criti- GFDL and HWRF models both use high reso- cally affect accuracy. While an analysis of closely lution nested grids (smaller than 10 km) to rep- spaced observations can produce an accurate, resent the storm’s dynamics and can predict the objectively analyzed wind field (Powell et al., storm’s track and intensity variation. 1998), this approach is not viable for forecast The second class of atmospheric tropical or hypothetical storm simulations, but only cyclone models develops an idealized represen- for hindcast scenarios. Therefore, meteorological tation of the storm based on a few key input tropical cyclone models are applied for these parameters to characterize its location, size, scenarios. Considering the importance of the and intensity; these are referred to as para- wind field, the quality of the tropical cyclone metric models. This type of model is effective model is very important to the surge prediction. when estimated storm conditions are desired Therefore, validation of the wind fields should despite a lack of specific data about the event. be undertaken when possible, and probabilistic This makes them good candidates for simu- approaches may be employed during forecast lating hypothetical storms or when dealing with scenarios to account for uncertainty in tropical forecast uncertainty (e.g., ensembles of predic- cyclone predictions. tions). Parametric wind models include the one There are two primary classes of atmo- incorporated with the Sea, Lake, and Overland spheric models used to generate tropical cyclone Surges for Hurricanes (SLOSH) storm surge wind and pressure fields. The first class solves model (Jelesnianski et al., 1992), the tropical the 3-D mathematical equations describing the cyclone model of the planetary boundary layer dynamics of the atmosphere, and they are from Thompson and Cardone (1996), and the therefore called dynamical models. Dynamical Holland model of the gradient winds in a models benefit from their fully 3-D solution of storm (Holland, 1980), among others. Typical the governing equations but are costly to run input parameters are the track (to define and require extensive initial and boundary con- location and translational speed), peak wind ditions. This type of model is used to guide speed or minimum central pressure (to charac- forecasts of a tropical cyclone’s track (and terize intensity), and the radius to maximum also intensity from some models); the wind winds (to define size). While each parametric and pressure fields computed during these pre- model approaches the calculation of the tropical dictions can also be used to drive a storm cyclone wind and pressure fields from a different surge model. Three predominant dynamical basis, the resulting wind and pressure fields models used by US National Weather Service tend to be simplified, fairly uniform approxima- (NWS) are the Global Forecast System (GFS; tions that do not account for the variations in Global Climate and Weather Modeling Branch structure that occur in individual storms. Para- 2003; see http://wwwt.emc.ncep.noaa.gov/gmb/ metric models also lack meteorological condi- STATS/html/model changes.html for the latest tions outside the storm. However, they solve version), the Geophysical Fluid Dynamics Lab- simplified equations and therefore are very effi- oratory hurricane model (GFDL; Bender et al., cient and are often used to run large numbers of 2007), and the Hurricane Weather Research hypothetical storms (e.g., for planning purposes and Forecasting model (HWRF; see http:// or as ensembles of possible land-falling tracks). www.emc.ncep.noaa.gov/HWRF/index.html). Air-sea interaction governs storm surge gen- The GFS is a suite of global model products eration caused by the wind field. The surface produced for numerical weather prediction at shear stress parameterization is particularly about 35 km resolution. It has an incomplete important since this is the predominant forcing description of the tropical cyclone and therefore mechanism for surge generation. Wind shear is only relied upon for track predictions. The stress at the water surface causes momentum March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 373

transfer from the hurricane wind field to the the drag coefficient. Modeling efforts which vary water column, causing a gradient in the ele- the drag coefficient dependent upon wave con- vation of sea surface (set-up). Use of a wind ditions improve momentum transfer from wind drag law dictates that shear stress is gov- to water (e.g., Mastenbroek et al., 1993). Third, erned by the wind velocity and the wind drag waves affect local bottom friction effects, which coefficient. The drag coefficient is defined by impacts hydrodynamic circulation. Therefore, empirical data collected to define the rela- provided that good quality nearshore land and tionship between wind velocity and surface seabed elevation data is available, surge simu- roughness. Examples of drag coefficient rela- lations might be improved when combined with tionships include Garratt (1977) and Large and wave models that describe the wave field. Pond (1981), although recent research (Powell Accurate representation of bottom shear et al., 2003; Makin, 2004) indicates that the stress is also important in governing the prop- generally linear correlation these laws provide agation of the surge inland. In 2D models for drag as a function of wind speed may not the bottom shear stress is generally defined apply as winds reach hurricane strength, where as a quadratic function of the depth-averaged drag may level off. Finally, overland regions have current and a coefficient of friction. Bottom increased roughness in comparison to marine shear stress parameterization is especially sig- conditions, reducing the wind speed and the nificant during flooding processes due to shallow surface shear stress. Overland roughness can water depths and complex variations in topo- vary from relatively low values for open terrain graphic roughness (e.g., from sandy beaches to high values in heavily vegetated and urban and parking lots to grassy fields and wooded regions. However, drag coefficient relationships areas). While there can be a lack of empirical are developed based upon data collected under data for characterizing bottom friction under open marine conditions. Therefore, surge models this range of conditions, many models do allow which are applying parametric-type tropical variable friction coefficients spatially to account cyclone models which assume open marine con- for land cover type, such as by applying varying ditions often reduce wind speeds according to Manning’s n values (e.g., see Chow, 1959). land type classification (Jelesnianski et al., 1992; One advantage of 3D models is their ability to Westerink et al., 2008). describe the bottom boundary layer according In addition to wind shear stress, the action to local roughness heights, which can be more of wind-generated short gravity waves can affect accurate than parameterizations used in the 2D the surge through several mechanisms and, in equations. consequence, the total water level. First, waves directly impart moment to the water column. 2.1.6. Calibration and Validation This action can be described by radiation stresses which are applied at the water surface, Hindcasts of recorded historical tropical leading to a localized wave-generated set-up cyclones are used to validate and calibrate surge in areas where waves are breaking sharply. model implementations. Ideally large datasets However, when wave breaking occurs gradually from multiple storms are used so that calibra- and energy is dissipated by bottom friction tion and validation can be conducted indepen- it does not produce a significant increase in dently. Considering the model grid’s important water elevation. Second, the sea state is char- role in representing coastal geometry, this pro- acterized by the condition of the wind waves cess needs to be done separately for each model (e.g., height and steepness), which depends on domain. their age. Changes in sea state therefore affect In order to validate storm surge models, the roughness of the ocean surface and thus observations of water level variations caused by March 13, 2010 10:34 B-936 b936-ch12

374 Shishir K. Dube et al.

tropical cyclones are required. Most valuable produce mean error measures of one half of a are time series observations of water level, gen- meter and less (Westerink et al., 2008). erally from a preexisting water level gauge. It is important that the datums for these gauges 3. Storm Surge Systems be accurately tied into a common, documented datum such as a national network of bench- The storm surge community continues to marks so that adjustments can be made to develop and apply systems for a variety of pur- match the model’s datum. The design and poses. These include providing local high res- location of a water level gauge is normally olution predictions, improving modeling skill, intended to filter out short gravity wind waves, examining location-specific environmental and leaving the remaining storm tide signal. Unfor- management issues, and more. Descriptions of tunately, water level time series observations some prominent storm surge modeling systems are not widely available due to the cost of are provided in section 3.1 below to illustrate installing and maintaining gauge sites. Addi- the types of efforts underway. tionally, many gauges subject to the strong storms are destroyed, eliminating data from 3.1. Models in Use severe events. There is a wide range of storm surge mod- In order to expand the suite of data available eling systems in use for predicting the impact of to characterize the impact of the storm, many tropical cyclones, covering a range of numerical times a post-event survey of peak storm tide methods, model domains, forcing and boundary levels is undertaken. These surveys collect High conditions, and purposes. However, in the Water Marks (HWMs) which are generally US there are two primary surge models: the waterordebrislinesonstructures or similar Sea, Lake, and Overland Surges for Hurri- features that represent the peak storm tide. canes (SLOSH) model and the Advanced Cir- However, these observations often include not culation (ADCIRC) model. While sometimes only the storm tide but also wave run-up, used for similar tasks, they are quite different flooding from precipitation, and other local vari- in approach, have different strengths and weak- ability. Therefore confidence is increased by col- nesses, and each is likely more well-suited for lecting a HWM from a building interior and by a particular application. Of course, there are ensuring consistency with marks in close prox- many other models in use, and a survey of some imity. Still, their values may vary by 20 percent notable approaches is provided. or more against nearby marks (Jelesnianski et al., 1984; Harper et al., 2001b), and one 3.1.1. SLOSH study has found that only 15 percent of marks were high quality indicators of peak storm tide The SLOSH storm surge model was developed (Ebersole et al., 2007). by the US NWS (Jarvinen and Lawrence, Once a storm surge model has been 1985; Jelesnianski et al., 1992). It is based developed for a particular geographical region on a linearized form of the governing SWE; and a hindcast has been made of a historical the advective terms in the equations are dis- storm which has good observations of the storm carded. The SLOSH model is a FD model tide, a skill assessment of the model can be per- that employs an orthogonal curvilinear grid. formed. This is necessary to validate the model SLOSH grids for all vulnerable areas of the for use. The correlation between the modeled US coast have been constructed by the NWS and observed peak surge can be calculated for use in emergency management (http:// to evaluate model skill. State-of-the-art storm www.weather.gov/tdl/marine/Basin.htm). The surge model simulations have been found to structure of a SLOSH grid results in fine March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 375

resolution near the pole of the grid (which is in regions outside of the area of interest (e.g., placed near the area of interest) and larger ele- inland beyond the floodplain). Second, the shelf- ments at its outer boundary. These grids are based, regional nature of a SLOSH domain limited in domain size and have their open ocean limits accurate specification of boundary condi- boundaries located on the shelf. SLOSH has the tions during storm surge events because of lack capability to simulate wetting and drying as well of knowledge of set-up at the open boundary, as parameterize sub-grid scale features such as and prevents dynamic coupling to larger basins. 1-D channel flow with contractions and expan- In the US, flood-prone areas are usually sions, vertical obstructions to flow with over- determined by using the SLOSH model and topping (levees, roads, and banks that include input parameters from thousands of hypo- cuts) and increased friction drag in heavily veg- thetical hurricanes. These model runs are used etated areas. However, it is not able to account to create atlases of potential surge, which guides for inflow, rainfall, or tides. emergency managers in creating evacuation The SLOSH model contains an internal para- plans (Jarvinen and Lawrence, 1985). The US metric tropical cyclone wind model used for NWS’s National Hurricane Center (NHC) also forcing both forecast and hypothetical storm runs SLOSH for prediction of storm surge from simulations. The tropical cyclone track, central potentially land-falling tropical cyclones. The pressure deficit, and radius of maximum winds storm’s parameters (track, size, and intensity) are used as input to create wind and pressure are provided by the official hurricane forecast. fields which are interpolated onto the grid. These forecasts are distributed in real time A wind drag coefficient that is constant with to the forecast and emergency management respect to wind speed is applied, although it community as Geographic Information System changes according to vegetation over land. (GIS) products. They begin when a storm is SLOSH has an extensive history of use within 24 hours of landfall due to the limita- by the U.S. National Weather Service (NWS) tions of storm forecast accuracy; therefore they over nearly four decades. An extensive skill are only used to augment the pre-computed assessment found peak surge errors to be less atlases of surge potential, which is the basis than 0.6 m for nearly 80 percent of evaluations for emergency management activities. A prob- (Jarvinen and Lawrence, 1985). However, while abilistic component based on the official NHC it is a capable model, it does have limitations. track forecast has been developed which fore- First, at present it cannot provide astronomical casts the likelihood of surge as a function tide forcing or river inflow. Second, it is a lin- of the historical range of track error; see earized model subject to error in locations where http://www.weather.gov/mdl/psurge/. advection is important, such as in the vicinity of tidal inlets and similar constrictions. Finally, 3.1.2. ADCIRC a SLOSH grid is generally limited in size to the coastal shelf surrounding the study area. Two The ADCIRC model was originally developed issues arise from this limited domain strategy. for the US Army Corps of Engineers for high First, the use of a structured grid limits the resolution coastal ocean modeling (Luettich capability to provide localized refinement. While et al., 1992). It has since been used by federal SLOSH grids do have increased refinement at agencies, academic researchers, and private com- their center, it is not always possible to resolve panies for a wide range of modeling appli- additional features that may be elsewhere (e.g., cations, include storm surge, basin-scale tidal an inlet along the coastline that is away from the modeling, and tidal inlet circulation studies. center of the grid). Conversely, the semi-annular ADCIRC solves the SWE discretized in space structure of the grid also leads to over-resolution using the FE method, which allows for highly March 13, 2010 10:34 B-936 b936-ch12

376 Shishir K. Dube et al.

flexible, unstructured grids. All non-linear terms institutions to extend an existing ADCIRC have been retained in these equations. ADCIRC storm surge application for Southern Louisiana can be run either as a 2-D depth-integrated (Westerink et al., 2008). The extension of the (2DDI) model or as a 3-D model. In either case, model built for studying the hurricane pro- elevation is obtained from the solution of the tection system for Louisiana was to conduct depth-integrated continuity equation in Gen- an extremely large and highly refined forensic eralized Wave-Continuity Equation (GWCE) study of Katrina’s disastrous storm surge. As form. The GWCE is implemented to prevent part of the Interagency Performance Evalu- generation of spurious oscillations; however, ation Task Force (https://ipet.wes.army.mil/), it replaces the continuity equation with its this project applies very high resolutions of 50 time derivative, meaning continuity is not sat- to 100 m across coastal Louisiana and Missis- isfied and mass balance errors exist. Velocity sippi, resulting in grids exceeding one million is obtained from the solution of either the nodes. This model was adopted by FEMA to 2DDI or 3D momentum equations. ADCIRC produce flood insurance rate maps for the Gulf requires input of wind and pressure fields to Coast. Furthermore, both NOAA and the US consider the effect of storm surge. These wind Navy have been applying ADCIRC as a storm and pressure fields are developed by meteoro- surge model for specific research and devel- logical models independent of ADCIRC. For opment projects. further description of the model implementation see the ADCIRC theory report (Luettich and 3.1.3. Survey of Other Approaches Westerink, 2004). In order to accurately simulate storm surge, In addition to the two predominant storm surge several features have been included within models in the US (SLOSH and ADCIRC), a ADCIRC. It can model the wetting and drying number of other modeling systems have been of inundated areas, it can represent subgrid scale developed and applied worldwide. A few of the obstructions to flow as weirs, and it can apply predominant systems are presented here as an the transfer of momentum from breaking wind overview in order to illustrate how storm surge waves as surface radiation stresses. Most impor- simulations are being improved and the range of tantly, its unstructured grid methodology allows issues being addressed. for very high refinement of coastal regions (suc- One of the predominant FD coastal ocean cessfully modeled at scales less than 50 meters). models is the Princeton Ocean Model (POM; These highly refined coastal regions are built Blumberg and Mellor, 1987). POM is a 3-D, into basin scale domains by smoothly varying fully nonlinear hydrodynamic model that has element size. With the high resolution of inun- been in development and use by a wide range dation regions and the small time steps required of user over the past three decades. It has been by the semi-explicit solution scheme comes high applied for use in tropical storm surge modeling computational cost, however. The model can be in North Carolina (Xie et al., 2003; Peng et al., run on a single processor, but most often is 2004, and Pietrefesa et al., 2007). In this appli- run in parallel on high performance computing cation it has been coupled with meteorological systems with hundreds of processors using the (wind, pressure, and precipitation), wave, and Message Passing Interface (MPI). river discharge models in order to provide a com- ADCIRC is presently in wide use in the prehensive description of a storm’s impacts. US by federal agencies for storm surge mod- Another well-developed FD model that has eling. Following 2005’s Hurricane Katrina, the been applied to storm surge is the Curvilinear- US Army Corps of Engineers partnered with grid Hydrodynamics model in 3D (CH3D; other federal agencies, academic, and private Sheng, 1986; Sheng and Alymov, 2002). It is March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 377

a parallelized model capable of running large the storm surge model with a high resolution grids using MPI on high performance computing (4 km) atmospheric prediction system and river systems. CH3D uses a structured curvilinear discharge forecasts from NOAA in order to grid and has been applied on domains with res- provide a comprehensive characterization of the olution reaching less than 20 meters. It is a fully flooding threat for a coastal riverine system. nonlinear model that has been coupled with Model output is provided in GIS formats that wave models and has been coupled with larger allow for mapping and visualization applications scale circulation models. Various wind models in conjunction with other datasets. have been used to provide input. Two additional FV coastal ocean models In Australia, the FD MMUSURGE model have been used for tropical cyclone surge mod- developed at James Cook University has been eling applications. The fully nonlinear 3-D widely applied along the north-eastern coastline model FVCOM has been applied to storm surge and especially optimized for the complex simulations in Florida (Weisberg and Zheng Great Barrier Reef environment using sub- 2006). This modeling application demonstrates grid boundary controls (e.g., Bode and Mason, a mass-conserving 3-D simulation on an unstruc- 1994; Bode et al., 1997; Harper et al., 2001b tured grid extending past the continental shelf. (Appendix D), Harper et al., 2007). Also, Several projects have focused on approaches the commercially available TUFLOW hydrody- that leverage models available in the com- namic model (www.tuflow.com) is well suited to munity. The Southeastern Universities Research linked 2D and 1D applications and has been suc- Association’s Coastal Ocean Observing and cessfully used for many storm surge modeling Prediction (SCOOP) project is a community studies. approach employing Grid computing techniques Another FD SWE model has been developed that leverage resources over a wide range of in Australia for tropical storm surge modeling. locales (Bogden et al., 2007). By combining The Global Environmental Modeling Systems a suite of models available in the community 2D Coastal Ocean Model (GCOM2D; McInnes (including ADCIRC, ELCIRC, CH3D, and wave et al., 2002) solves for depth-averaged flow and models) the system is not tied to one model includes overland flooding and runoff as well as but rather is an infrastructure for enabling wave setup effects through coupling to atmo- model predictions. It does this by providing spheric and wave models. In applications, the streamlined access to model inputs and outputs researchers have employed a coarse (1 km) far- over a high speed network in standardized field model and coupled that to a finely resolved formats. Similarly, a project funded by the (100 m) local model using a nested grid strategy. National Oceanographic Partnership Program The Chesapeake Inundation Prediction (NOPP) combines readily available models for System (CIPS) focuses on the forecast needs surge (ADCIRC), waves (the Wave Analysis of emergency managers and first responders Model (WAM)), and tropical cyclones (NOAA’s (Stamey et al., 2007). CIPS is a combined surge- real-time analysis system H∗WIND and Ocean- tide-river discharge flood prediction system weather’s Interactive Objective Kinetic Analysis for the Potomac River near Washington, DC. (IOKA)) for simulations of events (Graber et al., It applies the mass-conserving unstructured 2006). By constructing an ensemble of forecast model ELCIRC (Eulerian-Lagrangian Circu- tracks this networked system of models is used lation; Zhang et al., 2004), which is based upon as a probabilistic forecast tool. a low order finite volume technique. While the Finally, commercial hydrodynamic modeling domain is of limited extent (the Potomac River), systems by the Danish Hydraulic Institute it is highly resolved and includes inland regions (DHI) and Deltares (Delft Hydraulics) are at scales of 50 to 100 meters. CIPS combines widely used in storm surge risk assessment March 13, 2010 10:34 B-936 b936-ch12

378 Shishir K. Dube et al.

studies around the world. For example, DHI’s so that barometric pressure can be measured MIKE 21 model has a FV version that has been and used to correct the measurements from the used to simulate storm surges for the coast of unvented transducers used to measure water- Bangladesh (Paudyal, 2002). surface elevation. These transducers then record the timing, spatial extent, and magnitude of 3.2. Storm Surge Measurements storm surge over the affected region. The first deployment was along the coasts of southwest 3.2.1. Storm Surge Measurement Louisiana and southeast Texas, USA (Fig. 4) on Land in anticipation of the landfall of Hurricane Rita Storm surge height typically is based on high in September 2005 (Fig. 5). Subsequent deploy- water marks (see section 2.1.6) or eye-witness ments were made along the east coast of Hur- accounts. This type of information can be useful, ricane Eduardo, but storm surge was minimal and combined with information such as debris during that event. piles, and structural and vegetation damage A total of 34 storm surge sites were deployed can provide estimates of the inland extent of for Hurricane Rita. Sensor locations ranged the storm surge. These somewhat qualitative from a few meters from the coast to about records, however, do not provide information 50 km inland. An additional 13 sensors were about the timing of the surge and can be deployed to measure barometric pressure. Data subject to major errors. In 2005, the US Geo- were recorded at 30-second intervals and are logical Survey (USGS) initiated a program to available from McGee et al. (2006b). This collect more detailed storm surge information. unique network provides some of the most The program consists of deploying a network comprehensive storm surge data yet available. of unvented pressure transducers in advance of Time-series of storm surge elevations are cyclone landfall along the predicted path of available for the almost 2-day period when the the storm. Some transducers are positioned coast was inundated, and maps showing inun- well above the expected storm surge height dation at 3-hour intervals have been prepared

Figure 4. Path of Hurricane Rita, September 22–24, 2005; satellite imagery obtained from National Aeronautics and Space Administration (2006); from McGee et al., 2006a. March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 379

Figure 5. Example of Hurricane Rita storm surge data, along with a well-established high-water mark (from McGee et al., 2006b).

(Berenbrock and Mason, 2008). One key finding tropical cyclones in 1998. At this time, wave from the data network was that high-water heights, wave variations and direction of motion marks may underestimate storm surge heights around the tropical cyclones were measured by as much as 0.5 m, perhaps because the and provided near real-time estimates of wave high-water marks are established as water is height and storm surge at landfall (http://www. retreating to the ocean rather than during the aoml.noaa.gov/hrd/hrd sub/milestones.html). initial surge. Researchers from both the hydrodynamic and the remote sensing communities should 3.2.2. Remote Sensing Techniques work closely to have remotely sensed derived data not only for wave characteristics as an For more than a decade, radar technology on board orbital satellites have been utilized to input into coupled model systems but also in track, measure and monitor the ocean surface, support of determination for storm surge. The including the wind, waves, and eyes of dam- remotely sensed data can be used to fill gaps aging hurricanes. Satellite altimeters provide in the absence of sufficient in-situ data while indirect measures of ocean surface currents hydrodynamic models outputs could be used to through monitoring sea level anomalies. The validate remotely sensed derived storm surge currents are then interpreted to generate a products. The new satellites missions, such as topographic map of the sea surface, which the Advanced Land Observing Satellite (ALOS), relates to wave height and direction. Ocean Sentinels, and the RADARSAT, in addition to surfacewavemeasurementsweremadefrom their primary missions, are designed to address airborne Scanning Radar Altimeter (SRA) in ocean tidal characteristics based on some of the March 13, 2010 10:34 B-936 b936-ch12

380 Shishir K. Dube et al.

concerns from researches (Johannessen et al., implementations provide guidance to forecasters 2006). and emergency managers about possible flood- ing conditions. For example, when a hurricane 4. Storm Surge System Applications is threatening the US coast, NOAA’s National Hurricane Center initiates storm surge predic- A properly implemented, well-validated storm tions based upon forecast advisories it provides surge model can be used to simulate surge from (Glahn et al., in preparation). The following historical, predicted, or hypothetical tropical information is used to drive a parametric tropi- cyclones. Each application has value in a cal cyclone wind model: (1) storm track to char- different context, and the constraints of an acterize location and forward speed, (2) radius of individual project can define which modeling maximum winds to define the storm’s size, and approach will be the most appropriate. (3) the central pressure deficit, which defines the storm’s intensity. 4.1. Review and Status of Storm However, considering the importance of the Surge Modeling Applications tropical cyclone wind field in producing an in the US accurate surge simulation, these storm surge forecasts are only valuable if the tropical 4.1.1. Forecasts cyclone prediction is accurate. The error in It is common for predictions of a tropical track prediction leads to a wide variation in cyclone to be made in order to provide infor- distribution of wind speeds and thus possible mation needed to implement emergency man- surge predictions depending on where a storm agement measures such as evacuations. These makes landfall (Ginis, 1995). Similarly a tropical predictions can be combined with hydrody- cyclone forecast of landfall 24 hours in advance namic models to forecast storm surge. However, has an error in track location that exceeds tropical cyclone storm surge is very sensitive to 113 km two-thirds of the time (National Hur- the bathymetry and topography of the ultimate ricane Center, 2007). Considering a typical landfall location. Equally important are the storm forward speed of 25 km/h, the time of internal characteristics of the storms them- landfall has an uncertainty of 4.5 h. Astro- selves — especially intensity and structure. nomical tidal conditions cannot be accurately Regardless of the model(s) chosen for cal- predicted for landfall because of this uncertainty culating storm surge, their operational use in timing. This makes it difficult to sequence presently is limited by the quality of the mete- surge simulations with astronomical tide forcing orological input. Better meteorological forecasts (Jelesnianski et al., 1992). Therefore, it is dif- will result in better storm surge calculations. For ficult to combine surge and astronomical tide example, cyclone landfall accuracy better than predictions in a forecast simulation. Instead, the 20 miles at 36 to 48 hours in advance, along tide level can be added linearly to the peak surge with intensity forecasts to within 10 mb central prediction once more confidence in land-falling pressure deficit are needed to provide accu- time is gained; this approach is taken by US rately useful storm surge calculations for evac- National Hurricane Center forecasts (Jelesniaski uation purposes. Therefore, the most promising et al., 1992). Finally, predictions of tropical avenue for operational storm surge calculations cyclone intensity can be less reliable than useful for evacuations lies in improvements in track predictions, also enhancing surge forecast meteorology. errors. Many national meteorological service agen- Emergency managers and first-responders cies have implemented operational surge predic- still find forecasts of storm surge valuable, tion systems for vulnerable regions. These model despite the uncertainty. One strategy under March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 381

development for dealing with the inherent errors By generating a suite of possible tropical in tropical cyclone predictions is to implement cyclone parameters that could affect a region, an a probabilistic forecast of surge based upon an atlas of possible storm surges can be produced ensemble of simulations (e.g., Graber et al., from a storm surge model to guide planning. 2006). A suite of possible tracks can be con- By varying important parameters such as track, structed based upon knowledge of the range of size, and intensity, a complete set of hypo- tropical cyclone prediction errors. These hypo- thetical storms can be created. The output from thetical tracks can vary the location, speed, simulations of these storms can then be com- size, and intensity of the predicted storm. This piled into useful tools that describe the distri- suite of possible storm conditions can then bution of risk across a region. In the US, NOAA be fed into a storm surge model which pro- produces maps of maximum surges generated by duces the corresponding suite of results. From a specific class of tropical cyclones (e.g., Cat- this output the distribution of risk can be egory 3 storms approaching from the south) for inferred along the coast. One difficulty in this a specific impact zone. This map is called the approach is that now instead of needing suf- Maximum Envelope of Waters (MEOW). A set ficient resources to complete one surge sim- of maps can then be referred to when developing ulation; computational requirements increase evacuation plans or when a storm of similar several-fold so that all runs can be finished characteristics is approaching land. For more in a timely manner. However, an Australian detailed discussion of the implementation of this forecasting system has implemented a hybrid methodology, see Jelesnianski et al., (1992). approach that uses dynamic models to inform While an atlas of possible storm surges parametric probabilistic models that allow for is a useful planning tool for emergency man- variation of a wide variety of variables, such agement, it does not provide information about as track, size, and intensity (Systems Engi- the probability of surge for each region. By com- neering Australia, 2005; Harper et al., 2007). bining a suite of potential storm conditions with This method provides rapid computation and statistical information about the likelihood of is well suited to regions where more sophisti- occurrence for these events, an understanding cated dynamical modeling is either unaffordable, of flooding risk can be determined for coastal impractical or the lack of accurate data (atmo- regions. This information can then be used to spheric, bathymetric or topographic) provides define insurance rates and building codes. In little improvement. Another challenge is com- the US, the Federal Emergency Management municating information about the resulting dis- Agency (FEMA) applies storm surge models to tribution of risk in a manner that can be construct flood insurance rate maps according easily understood and acted upon by emergency to the recurrence interval of potential tropical managers. cyclones (Massey et al., 2007).

4.1.2. Planning and Management 4.1.3. Coastal Engineering Although forecasts of potential surge heights Storm surge simulations are needed for engi- may be valuable in real time, emergency man- neering design projects to predict the impacts agement plans such as evacuation routes must of future storms. One significant example is be in place well before the storm approaches in the design of hurricane protection systems, in order to be effective. Therefore, characteri- which need to simulate not only the potential zation of the vulnerability from future storms surge height but also how it will be changed is needed to guide these types of planning and by proposed design projects. Along the US management activities. Gulf Coast many design projects are underway March 13, 2010 10:34 B-936 b936-ch12

382 Shishir K. Dube et al.

to restore barrier islands, dunes, and beaches in the world’s largest river system (Ganga- as well as improve levee systems in order to Brahmaputra-Meghna). Detailed reviews of the reduce storm surge impacts in highly populated problem of storm surges in the Bay of Bengal regions (e.g., see http://lacpr.usace.army.mil/). are given by Ali (1979), Rao (1982), Roy Considering the detail involved in designing (1984), Murty (1984), Murty et al. (1986), Das these public works, it is common for very (1994a,b), Dube et al. (1997, 1999, 2000a, 2006, high resolution models to be utilized. Simi- 2009), Rao et al. (1997), Chittibabu (1999), and larly, building standards are generally defined G¨onnert et al. (2001), by the risk that exists from natural hazards Although storm surges are less frequent within local areas. By computing the surge in the Arabian Sea than in the Bay of risk from a stochastically chosen set of Bengal, major destructive surges occasionally storms, the potential likelihood of surge occur- have occurred along the Gujarat Coast of India rence can guide implementation of building and Pakistan. Storm surge only infrequently standards. affects the Gulf of Oman, but there have been some cases when storm surge caused destruction along the coasts of Oman. The most recent 4.2. Review of Recent Developments example was the super cyclonic storm Gonu that in Predicting the Storm Surges struck Oman and caused about $4 billion in in the Bay of Bengal and damages and 49 deaths (JTWC, 2007). Arabian Sea Storm surges are extremely serious hazards 4.2.1. Real Time Storm Surge Predictions along the east coast of India, Bangladesh, System for the Bay of Bengal Myanmar, and Sri Lanka. Although Sri Lanka and the Arabian Sea is affected only occasionally by storm surge, tropical cyclones of November 1964, November In India, the study of numerical storm surge 1978 and November 1992 caused extensive loss prediction was pioneered by Das (1972). Sub- of life and property damage in the region. Storm sequently several workers attempted the pre- surges affecting Myanmar appear to be much diction of storm surges in the Bay of Bengal less intense than those occurring in India and (Das et al., 1974; Ghosh, 1977; Johna and Ali, Bangladesh. Notable storm surges, which have 1980; Johns et al., 1981; Murty and Henry, 1983; affected Myanmar, occurred during May 1967, Dube et al., 1985a; etc.). Dube et al. (1994), 1968, 1970, 1975, 1982, 1992, 1994, and 2008. Of Dube and Gaur (1995) and Chittibabu et al. these, Nargis 2008 was the worst cyclone. Nargis (2000) developed a real-time storm surge pre- generated storm surge in excess of 4 m near diction system for the coastal regions of India. Ayeyarwady deltaic region. The entire deltaic Real-time storm surge prediction systems have coast of Myanmar was flooded with surges also been developed for Bangladesh, Myanmar, ranging from 1.5–4.5 m. Pakistan, Sri Lanka, and Oman (Dube et al., Of all the countries surrounding the Bay 2004; Jain et al., 2006 a, b; Rao et al., 1994; and of Bengal, Bangladesh suffers most from Chittibabu et al., 2002). storm surges. The main factors contribut- The national meteorological and national ing to disastrous surges in Bangladesh are: hydrological services of many countries sur- (a) shallow coastal water, (b) geographical con- rounding the North Indian Ocean have achieved vergence of the bay, (c) large astronomical some success in providing storm surge warnings tides, (d) densely-populated lowlying islands, and implementing improved models through (e) favorable cyclone track, and (f) innu- cooperative and coordinated sharing of respon- merable number of inlets including those sibilities. Much of this work is done within the March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 383

framework and overall guidance and supervision 4.2.2. Location-Specific Models of the Tropical Cyclone Programme (TCP) The evolution of storm surges near the coast of the World Meteorological Organization is known to be very sensitive to the coastal (WMO). The TCP of WMO supported tech- geometry and offshore bathymetry at the nology transfer to run and operationalize location of the landfall of the cyclone. As a storm surge models for Bangladesh, Myanmar, result, operational models should include these Oman, Pakistan, and Sri Lanka from the factors as accurately as possible. This means Indian Institute of Technology (IIT)-Delhi/ that in addition to the large-scale storm surge Kharagpur. prediction models, operational offices should The forecasting system developed at IIT also use high-resolution location-specific models Delhi is based on the vertically-integrated for accurate prediction of the surges. Based numerical storm surge models that were on this need for location-specific models, Chit- developed by the group (Johns et al., 1981, tibabu (1999), Chittibabu et al. (2000, 2002); 1983; Das et al., 1983; Dube et al., 1985a,b). Dube et al. (2000b, 2000c; 2004) and Jain et al. The model is fully nonlinear and is forced (2006a, b) have developed location specific high- by wind stress and uses a quadratic bottom resolution models for the Andhra, Orissa, Tamil friction formulation. The nonlinear advection Nadu, and Gujarat coasts of India, and for terms in the model were found to have a sig- Bangladesh, Pakistan, Myanmar, Oman and Sri nificant effect on the simulations, especially in Lanka, following an approach similar to that of the shallow coastal waters of the head Bay Dube et al. (1994). An important feature of each of Bengal. Therefore, although the numerical of these location-specific models is that they solution of these terms affects the computational use accurate and highly-detailed bathymetry for speed of the model, the nonlinear terms must the offshore waters. A simple wetting -drying be included in the operational model in order scheme also has been included in the models in to provide the best forecasts. The details of order to avoid the exposure of land near the the model and the numerical solution procedure coast to strong negative surges. The reliability are described in Das et al. (1983) and Dube of these models has been tested using data from et al. (1985b). several severe cyclones, which struck the coastal Surface winds associated with a tropical regions of the countries in the Bay of Bengal and cyclone are derived from a dynamic storm model the Arabian Sea. The following sections describe of Jelesnianski and Taylor (1973). The only results of selected numerical experiments carried meteorological inputs required for the model out using location-specific models to simulate are the positions of the cyclone, pressure drop the surges. and radii of maximum winds at any fixed interval of time, and the model can be run 4.2.2.1. Myanmar: Mala cyclone in a few minutes on a PC in an operational (May 2006) office. The system is operated via a terminal menu and the output consists of the peak sea Peak storm surge of about 4 m was reported for surface elevations and currents. One of the the May 2006 Mala cyclone on the Myanmar significant features of this storm surge pre- deltaic coast near Gwa by the Department of diction system is that the meteorological input Meteorology and Hydrology, Yangon. The sim- needed for surge prediction can be periodically ulated surge contours (Fig. 6) show a maximum updated with the latest observations and fore- surge of 3.6 m to the right of the landfall point casts of national weather services. The model near Gwa, and the peak surge at Pathein is has extensively been tested with severe cyclonic about 2 m. The coast from Ye to Moulmein also storms. was affected by surges of 1 to 2 m. March 13, 2010 10:34 B-936 b936-ch12

384 Shishir K. Dube et al.

Figure 6. Simulated peak surge envelope for the May 2006 Mala cyclone (after Dube et al., 2009).

4.2.2.2. Sri Lanka: Rameswaram radius of maximum wind of 40 km (Fig. 8). A cyclone (December 1964) maximum surge of 2.6 m was simulated to have occurred near the eastern tip of Oman, which The December 1964 Rameswaram cyclone was is where the storm was nearest the shore. Peak one of the most severe storms to have affected surges of about 2.6 m and 2.5 m were simulated Sri Lanka and the southern Indian peninsula. at Sur and Nazwa Fins, respectively. The peak The simulated maximum surge of about 5.4 m surge value at Muscat was around 2 m. About occurred near Tondi (Fig. 7). Heavy inland 100 km of the coast in northeastern Oman was flooding in the region, with an estimated peak flooded with surges of about 1 m. The computed elevation of about 5 m was reported by Rao and maximum surge value of 2.6 m is comparable Mazumdar (1966), which is in good agreement with the estimated range of surge heights. with the simulations. It can also be seen from the figure that the coastal region to the north of Mannar Island along the west coast of Sri Lanka 4.3. Review of Recent was flooded with a surge of more than 4 m. Developments in Predicting Storm Surge in Australia Australia has an extensive tropical coastline 4.2.2.3. Oman: Gonu cyclone that is regularly impacted by severe tropical (June 2007) cyclones (TCs) and, in spite of its sparsely Simulations for the June 2007 Gonu cyclone settled population, has recorded significant were made using a pressure drop of 98 hPa and storm tide impacts dating from the late 1800s. March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 385

Figure 7. Simulated peak surge envelope associated with the December 1964 Rameswaram cyclone (after Dube et al., 2009).

Figure 8. Simulated peak surge envelope associated with June 2007 Gonu Cyclone (after Dube et al., 2009).

Accordingly there has been a history of devel- which is bounded by the Great Barrier Reef opment of storm surge modeling, risk assessment (GBR). The most intense TCs in Australia and associated atmospheric modeling capabil- are observed in the northwest of Western Aus- ities since the availability of significant com- tralia, although the population there is rela- putational power in the 1970s. This work tively sparse and storm tide studies have been has focused on the east coast of Queensland typically limited to specific towns or industrial (Harper, 1999) where TCs occur along a settled facilities. This brief summary follows Harper and rapidly developing coastal margin, much of et al. (2007). March 13, 2010 10:34 B-936 b936-ch12

386 Shishir K. Dube et al.

4.3.1. Queensland Climate Change using the JCU-MMU storm surge, wave, and Study (QCCS) TC wind field models that had been exten- sively calibrated over many years. The extent This refers to a number of investigations under- of the numerical ‘A’ and ‘B’ modeling domains taken over the period 2000 to 2004 that is shown in Fig. 9, covering some 3500 km of were funded primarily through Government coastline and completely enclosing the Great enhanced-greenhouse research allocations. The Barrier Reef at spatial resolutions of 13.9 km initial work (Harper, 2001b) was conducted and 2.78 km respectively. A further 20 ‘C’ model by Systems Engineering Australia (SEA) in domains were nested within the ‘B’ domains at association with the James Cook University 0.55 km resolution. A series of demonstration Marine Modeling Unit (JCU-MMU) and others. hindcasts of significant historical storms indi- This initial reference sets out recommended cated that peak surge levels were likely repro- methodologies for storm tide studies that would ducible within 5% of measured values provided also be capable of addressing climate change that adequate meteorological and bathymetric (enhanced-greenhouse) issues within the context data was available. An example of one such of tropical cyclones. Importantly, a holistic event, Severe Tropical Cyclone Althea in 1971, approach was advocated for identifying the is shown in Fig. 10. physical forcing mechanisms, ocean responses, vulnerabilities and impacts that would lead to 4.3.2. Local Coastal Inundation Studies informed decision making and the long term mitigation of storm tide threats. Subsequent In addition to the State-wide QCCS described studies conducted principally by JCU-MMU above, coastal Local Government regions in provided updated TC-induced surge plus tide Queensland have been active in assessing updates for selected east coast sites, all with storm tide risks for the purposes of updating and without allowance for predicted enhanced- local planning levels in response to enhanced- greenhouse effects by 2050. This work was done Greenhouse and for emergency response. Local

Figure 9. Queensland climate change ‘A’ and ‘B’ numerical domains (after Harper et al., 2001b). March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 387

3.5

With Tide 3.0 Without Tide 2.5 Recorded

2.0

1.5

1.0 Surge HeightSurge (m) 0.5

0.0

-0.5 23/00h 23/06h 23/12h 23/18h 24/00h 24/06h 24/12h 24/18h 25/00h

Time UTC

Figure 10. Measured and modeled surge component at Townsville Harbour for Althea (1971) (after Harper et al., 2001b).

inundation studies are typically conducted at require consideration of concurrent non-linear spatial resolutions of 50 m or finer, depending tide, surge and wave interactions. The principal on the available topographic data and the threat of inundation in such low lying environ- need for representation (e.g., Fryar et al., ments is due to extreme water levels on the outer 2004). Many such studies have adopted a reefs due to the combined effects of the pressure methodology similar to that described as the deficit component of a storm surge acting coin- “hybrid” approach recommended in Harper cidentally with a high tidal level and high wave (2001b). To achieve this, a parametric storm tide setup, caused by wave breaking on the outer model (surge and wave) is developed based on reefs and reef entrances. Breaking wave setup comprehensive hydrodynamic model parameter can be expected to be the dominant water level response mapping of the study region. The controller with magnitudes as high as several parametric model then provides an economical meters. Wind-stress induced storm surge setup kernel for Monte Carlo simulation and a flexible is small because of the surrounding deep ocean test bed for sensitivity testing of a range of environment but potentially more significant in model assumptions, El Ni˜no Southern Oscil- some parts of shallow lagoons. lation (ENSO) or climate change influences. Because of the complex non-linear interac- Water level persistence and joint probabilities tions of storm surge, tide and breaking wave of surge, tide, wave height and period are also setup, a statistical simulation methodology is available from these types of studies. again advisable and a combination of parametric surge, wave and wave setup models have been used successfully within a Monte Carlo driven 4.3.3. Coral Atoll and Reef Environments TC climatology (e.g., Harper et al., 2001a). The Coral atoll and fringing reef environ- method has been validated against historical ments can present with significantly different events and delivers sequences of time-varying vulnerabilities from the typical continental water level components that would be generated margin environments described previously and during a severe tropical cyclone event, with March 13, 2010 10:34 B-936 b936-ch12

388 Shishir K. Dube et al.

breaking wave setup on the outer reef flats and coast that would augment their existing access local wind stress influencing the shallow lagoon to real time hydrodynamic modeling. This pro- regions. Both these effects are critically mod- vided a significant benefit in operational flex- ulated in a non-linear fashion by the stage of ibility and also in training of forecasters, who the tide. could quickly examine and test a range of forecast scenarios. Importantly, this approach recognizes that the greatest uncertainty in the 4.3.4. Storm Tide Warning Systems storm tide forecast relates to the variability One of the aims of the QCCS was to develop of the forecast meteorological parameters. This self-consistent methodologies for both miti- is especially relevant to the vast majority of gation/planning and forecast/warning appli- tropical cyclone forecast centers around the cations that could be efficiently developed world that do not have aerial reconnaissance and deployed to provide coverage over large capability, where the Dvorak satellite image sections of coastline. The proposed “hybrid” interpretation method is the primary intensity approach, a combination of detailed deter- forecast tool. Even with the rapid development ministic hydrodynamic modeling and derived of advanced atmospheric models this situation is parametric models, satisfied these criteria by unlikely to change significantly within the next providing a self-consistent framework for long few decades. term climatology modeling and short term The Northern Territory BoM engaged SEA warning needs. This method differs from the in 2005 to provide a turnkey parametric storm static Maximum Envelope of Waters (MEOW) tide forecasting system that incorporates tidal approach in that it provides a rapid response prediction, storm surge, waves and wave setup tool that can incorporate tidal variation and for the entire northern Australian coast from storm parameter uncertainty. the Gulf of Carpentaria west to the Kimberley Based on these recommendations, the region in Western Australia. Figure 11 shows the Bureau of Meteorology (BoM) in Queensland area that was comprehensively modeled using subsequently adopted this approach in devel- seven along-coast domains, each of about 400 km oping a parametric surge model for the east extent, to provide the basis for the necessary

Figure 11. Extent of the Northern Territory storm tide warning modeling system (after Harper et al., 2007). March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 389

parametric models. Approximately 25,000 indi- the quality of the geographic data. Coordinated vidual TCs were modeled as part of this devel- collection of high-resolution topographic and opment and storm tide predictions are available bathymetric data, especially LiDAR datasets, for over 1,000 named localities. However, the is needed to accurately model coastal inun- resulting parametric prediction model SEAtide dation. Furthermore, vertical datum transfor- can generate more than 100 probabilistic storm mations are essential for combining coastal tide forecasts within one minute on a typical geospatial data. Elevations that are referenced desktop computer. to inconsistent vertical datums can cause arti- ficial discontinuities in model grids, especially 5. Future Development at the coast. Therefore, the capability to transform coastal elevations to a common datum Extensive storm surge modeling applications is crucial for developing the geospatial infras- have been made with existing modeling systems. tructure for the coastal zone. Finally, the cre- These efforts have been extremely valuable ation of seamless elevation datasets (such as in reducing the damages caused by extreme Digital Elevation Models) for storm surge model tropical cyclones by provided emergency man- development would enable accurate grid con- agers with reliable planning tools. However, struction. there are many issues where future development is needed to further decrease the potential for 5.3. Air-Sea Interaction loss. A brief discussion follows, including some of the main issues that are outstanding or under The interaction between the wind, sea surface development by the research community. and water column is not well understood, par- ticularly at hurricane wind speeds. What is 5.1. Improved Hurricane Forecasts required is an accurate description of the physics of momentum transfer from wind to waves to The most important factor affecting the water column. However, data used to parame- accuracy of a storm surge model is the forcing; terize wind drag coefficient relationships is dif- that is, the ability of a tropical cyclone model ficult to collect and is nearly non-existent from to construct realistic wind and pressure fields. tropical cyclones. Research indicates that the Since the most significant improvement in storm assumption of a linear increase in surface stress surge forecasting depends on improvements in may not apply at hurricane strength winds predicting tropical cyclones, particularly for (Powell et al., 2003). Furthermore, wind drag track and intensity, surge simulations are criti- laws are developed under the assumption of cally dependent on the accuracy of these storm open marine conditions, but increased roughness characteristics. Without confidence in tropical is likely to exceed these conditions in near- cyclone forecasts, storm surge models have shore and overland conditions. Surface waves limited use for prediction. also affect air-sea interaction; this is especially true as waves break and inundation occurs, since 5.2. High Quality Bathymetry roughness increases and becomes more localized. and Topography Further research is needed to characterize air-sea The most significant factor affecting model skill interaction under tropical cyclone conditions. besides meteorological forcing is the accurate description of geographic features in the model 5.4. Observations of Surge grid. These errors can be minimized by careful grid construction and inclusion of sub-grid scale There is a shortage of comprehensive data features in the model, but this depends upon available to evaluate storm surge simulations. March 13, 2010 10:34 B-936 b936-ch12

390 Shishir K. Dube et al.

Observations of water levels during hurricanes models comprehensively consider. Conversely, it are not extensive. Gauges are not widespread is important that the high-resolution models and are often destroyed during severe events. being implemented run quickly in order to Post-storm HWM surveys are unreliable because satisfy the large suites of runs required for of the difficulty in collecting accurate data and probabilistic forecasts, creation of storm surge are not done for all storms. Therefore, there atlases, and calculation of historical flooding is an important need for expanded and coordi- probability. nated collection of water level data from tropical cyclone systems. A network of gauges that are 5.7. Wave Modeling hardened to survive strong storms and collect data throughout is needed. Furthermore, this One of the predominant effects of a tropical permanent network should be augmented by a cyclone is the large wind waves generated by the suite of temporary pressure gauges installed in storm, which affect the storm surge in a number the path of a storm, as done in the US for Hur- of ways. While research and development is ricane Rita in 2005 (McGee et al., 2006b). Third, ongoing into coupling surge and wave models post-storm collection of HWMs should be sys- (see section 2.3.3), many systems are lacking a tematically completed and archived. Fourth, all full description of the physics that occurs. A water level observations need to be accurately wind wave model will provide wave radiation tied to a geodetic network so that a common ver- stresses which drive wave-current interaction tical datum can be used for both model output and increase storm surge height. Furthermore, and observations. wave-current interaction modifies the bottom friction stresses in shallow areas, increasing the bottom friction as the wave height increases. 5.5. Communication of Uncertainty Finally, the wave model describes the conditions There is inherent uncertainty in tropical cyclone of the sea state as the hurricane travels from off- forecasting and corresponding predictions of shore to land, enabling improved description of surge. However, users of such products need the drag coefficient based upon sea state. estimates of this uncertainty in order to guide decision-making processes. Development 5.8. River Forecasting of probability distributions of storm forecasts There are important interactions between rivers will be increasingly required, created by an and the coastal environment during tropical ensemble of simulations whose results need to be cyclones. While the flooding caused by rainfall combined in a fashion that communicates risk generally occurs after the surge conditions have appropriately. passed, the occurrence of multiple storms in succession requires the accurate knowledge of 5.6. Improved Model Accuracy river inflow and water levels to accurately model and Efficiency surge throughout an estuarine system. Inclusion of river inflow and precipitation can address Although there is a large number of storm these issues. surge models available in the community, there are still outstanding needs to improve model 5.9. Coastal Geomorphology accuracy by including additional physics and increasing resolution. For example, surge water Future focus is needed on study of geomor- levels are affected not only by wind and pressure phological changes that can occur during surge fields, but also by tides, waves, river inflow, pre- events, as these can be important in defining cipitation, and erosion, factors that few if any the volume of water that passes over an eroding March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 391

barrier island, for example. The present state by sharing common technology. It also promotes of sediment transport models is generally not uniformity of techniques which enhances relia- sufficiently development to consider these con- bility, and it provides the best leverage of limited ditions. Sediment transport models are needed funding. Open sharing of data, tools, models, that accurately predict the time rate of change and approaches increases the value and reli- of the coast, and especially dune systems, under ability of the system as a whole. Leadership the effects of waves and surge from a tropical from the public sector is necessary to support cyclone. a community approach as they are the major provider of storm surge mitigation strategies and they develop and implement relevant tech- 5.10. Decision-Support Tools nical standards. Increased utilization of surge model output One major benefit of a community approach can be made by providing it in GIS formats is a standardization of terminology that and combining it with other data layers. This increases understanding of storm surge threat leads to easily utilized maps and visualizations, in the user community. It avoids confusion and enables incorporation of socioeconomic and caused when federal agencies provide infor- demographic data with model output. This mation described in a multitude of ways; uncer- enables decision makers to evaluate risks in a tainty about differences in technical language more thorough manner. has confused the public in the past. A community approach does not mean that a single model is selected for use by the 5.11. Community Approach community. Different models will likely con- Improvements in storm surge modeling benefit tinue to be advantageous for specific applica- both the public and private sector. When there tions. Rather, it establishes an infrastructure are advancements in the knowledge and pre- that enables developments to be shared across diction of storm surge impacts, appropriate mit- models. Tools such as standards and documen- igation steps can be taken to minimize losses. tation will enable technology advancements to A community approach to model development benefit the community at large. that shares the funding burden as well as the Finally, an open, coordinated forum for a benefits of research and development advances community approach to modeling can provide all interests. A community approach is one standards and regional test beds can be used where both resources and developments are for model comparisons which will guide future shared across all model users, whether gov- applications. This will enable the community to ernment, academic, or private. It allows all select an appropriate model implementation for members of the community to benefit from specific applications. the advancements that are made so that the state of the technology is continually moving 6. Integrated Modeling in Coastal forward. This strategy enhances both research Areas and applications by allowing researchers to focus on making scientific improvements without Generally, it takes a long time for riverine having to recreate the larger set of technology flooding to reach the coast, whereas storm needed to support such work, and facilitating surges are roughly coincident with landfall. operational application of developments which However, high river flows occurring simultane- improve skill and further the use and value ously with storm surge likely will have adverse of a model. Therefore, a community approach impacts on coastal inundation. The discussion streamlines the research to operations pathway below will refer to coupling of river flow March 13, 2010 10:34 B-936 b936-ch12

392 Shishir K. Dube et al.

and coastal hydrodynamic models to address model capabilities. Low-lying coastal estuaries potential coastal inundation. are complex hydrodynamic systems that often require 2-D models for accurate simulation, and the 1-D approach may not be appro- 6.1. Hydraulic and Storm Surge priate. Issues of computational grid resolution Model Coupling must sometimes be addressed during the cou- Most of the studies reviewed address models pling process. For example, if the downstream that forecast either storm surge or river flow, boundary of the 1-D hydraulic model includes but few studies have described the coupling of multiple surge model computational cells, then both inland flood and storm surge models for averaging of surge elevation across the hydraulic forecasting purposes. Even though in most cases, model boundary may be required. Likewise, surge and river flooding occur separately, with in the case of a 2-D depth-average hydraulic storm surge coming at time of land-fall and river model, the downstream boundary will river flooding occurring later. There are situa- likely include multiple river cells which must be tions when both effects might co-exist and setup linked with one or more surge model computa- a severe flooding situation. This was experi- tional cells, again requiring some averaging or enced in 1999, when Hurricane Floyd affected smoothing across the boundary. The Tar River the East Coast of the United States. Not only example discussed below illustrates this issue. Floyd’s storm surge was affecting coastal areas The Tar River, North Carolina, USA, was of North Carolina but also torrential rainfall modeled using the 1-D NWS hydraulic model occurred in an area hit by Hurricane Dennis just called Flood Wave (FLDWAV) and coupled weeks earlier. Rivers were already in flood by the with a tidal model (Riverside Technology, Inc., time Floyd made land-fall. In the USA, coupled 2006). The lower river reach of the Tar River is systems using 1-D models are operational for described as follows (p. 8): several coastal rivers and some of them were “The Tar River drains into the Pamlico utilized in pilot projects for generation of inun- Sound, which also receives flow from several dation mapping in real-time. Reed and Stucky other rivers in North Carolina. The Sound is (2005) described the application of the NWS protected by a series of barrier islands with Dynamic Wave Operational Model (DWOPER) a number of inlets through which tidal influ- ences are transmitted and hydrologic inflows on the lower Ohio/Mississippi Rivers to the Gulf drain to the Atlantic Ocean. Tidal influences of Mexico. This model uses the Sea, Lake, and predominate in the determination of water Overland Surges for Hurricanes (SLOSH) water levels in the Pamlico Sound and have a strong level forecasts as the downstream boundary. influence on stages in the Tar River as far Brown et al. (2007) recently described “the first upstream as Greenville under normal flow con- attempt to model storm surge flooding of an ditions and as far as Rock Springs under very low flow conditions.” urban area with 2-D hydraulic model and to explore the uncertainties associated with model This is an example in which the use of a predictions” for the Canvey Island, UK. 1-D hydraulic model might be inadequate for In order to accurately reproduce the passage simulating conditions near the Sound; a 2-D of a flood wave through a river reach, the model might be more appropriate for simu- responses (water level, flow, velocity) are sim- lating flood and storm surge in the estuary. ulated using unsteady-state hydraulic models. However, because of operational constraints it Many hydraulic models are available to sim- was modeled using a 1-D model. ulate water levels and river flow, but the Following selection of the hydraulic and tidal local site characteristics might require specific models, the general approach for simulating the March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 393

coupled surge-inland flooding response of the for time increments on the order of a river system includes: (i) selection of river-reach few minutes. However, because the forecast boundaries, type and location; (ii) collection of system architecture might be too rigid and cross-section geometry to build the channel and because of operational limitations for exe- flood plain geometry for the hydraulic model, cution of the hydraulic models, the coupling and (iii) determination of local flows using data is performed in hourly increments. or previously-developed hydrologic models. The 4. Consistency of reference datums is essential upstream boundary of the model reach is usually for the accuracy of results. This includes ver- defined by a time-series of flow; the water-level tical and horizontal datums for cross sections, time series for the downstream boundary will water levels, and tide gages. be provided by the surge model. This rather 5. Consistency of boundary conditions must simplistic approach might become cumbersome be maintained for situations in which when considering operational implementation. multiple river computational cells (2-D Data demand for research and model devel- hydraulic model) coincide with a single surge opment is usually confined to historical data for model cell. Multiple iterations between the calibration. Real-time operations require either hydraulic and estuarine models might be observed or forecast information on rainfall, a required. forecast inflow hydrograph, and forecast surge 6. Continuous execution of the coupled system elevations derived from a surge model driven by ensures that the models are initialized cor- meteorological forecast. rectly, but this might not be possible for a Issues to be considered for the hydraulic/ situation in which a surge model is not run tidal coupling include: continuously and no downstream real-time tidal data are available. 1. The calibration period should be selected to ensure there is enough data for all the A coupled system for St. Johns River, time series required by the model system. In Florida, USA has been in operation since addition, the period selected should include 2001 (Fig. 12). The US NWS Southeast River a wide range of river flows (low, medium and Forecast Center (SERFC) is currently fore- high flows) and minor, medium and major casting water levels in the lower end of the surge effects. St. Johns River near Jacksonville using the 2. The forecast window for individual models NWS hydraulic model FLDWAV. The model (including the hydrologic and meteorological is executed for a 300-mile reach; the upstream models) must be examined to define the boundary (flow hydrograph) as well as the local window for the coupling system and to ensure inflows from tributaries is derived from the that data from each model is available as hydrological model Sacramento Soil Moisture needed by the dependent models. Accounting. The downstream boundary selected 3.Thetimeintervalforthesystemcoupling is the tide gage located at Mayport at the mouth must be selected based on the minimum of the river near Jacksonville, Florida. A forecast allowed to describe the process while main- window of 5 days is used. taining feasibility of model execution for The time series for the water level at the forecasting purposes. For example, in the downstream boundary are composed of the USA most of the operational models used to observed water level up to the forecast time. forecast river flow are performed for incre- For the forecast period, the astronomical tide mental periods between 1-hour and 6-hours. forecast provided by the National Ocean Service Storm surge models can produce output (NOS) is combined with the output from the March 13, 2010 10:34 B-936 b936-ch12

394 Shishir K. Dube et al.

Figure 12. St. Johns River, Florida, USA. The river flows from South to North.

Charley France Jeanne

3.5 3 2.5 2 1.5 1 0.5 Stages (meters) 0 7/4/2004 8/5/2004 9/6/2004 6/10/2004 6/18/2004 6/26/2004 7/12/2004 7/20/2004 7/28/2004 8/13/2004 8/21/2004 8/29/2004 9/14/2004 9/22/2004 9/30/2004 10/8/2004 Date GENF1 SNFF1 DLAF1

Figure 13. Water levels in St. Johns River during 2004 hurricane season (Charley in August, Frances and Jeanne in September).

Sea, Lake and Overland Surges from Hurricanes with Hurricane Charley produced a significant (SLOSH) model. The NWS developed blending rise, but it wasn’t until the passage of Hurricane techniques to correct the downstream water Frances that the rises reached flood stage and level once observed data becomes available remained above flood stage until the beginning The 2004 hurricane season provided insight of November 2004. on the performance of the hydraulic model. During the operational performance of the Three hurricanes affected this area: Charley, system, the wind acting on the river reach was Frances, and Jeanne (Fig. 13). Rainfall produced observed to greatly influence the propagation by these systems affected the water levels in the of the surge and forecasts of water levels. The St. Johns River. Water levels were low at the jump in water levels due to wind effect (Fig. 13) beginning of August, and the rainfall associated was more noticeable at the station of Deland March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 395

(DLAF1) when Charley affected the area and steady-state models for the generation of static- the river was at low levels. The hydraulic model, map libraries of inundated areas at different used by the NWS at that time, (FLDWAV) had water levels. Maps from these libraries are then the capability to incorporate wind direction and called up for application based on water-level magnitude as a single value for the whole reach, forecasts. but this is not practical for the entire 300-mile reach. As a result, wind effects on the river are 6.2.1. Development in the USA not included in the coupled system. The NWS Inundation maps are based on predictions of is in the process of incorporating this variable water levels along the river reach and the corre- into the hydraulic modeling. sponding terrain information. The whole process to determine inundation mapping in coastal 6.2. Inundation Mapping areas due to topical cyclonesinvolvesmeteo- Several countries are working on projects rological forecasting (storm tracking, intensity, involving potential inundation mapping. In and precipitation totals), oceanographic, estu- China, the production of flood maps has been arine, and riverine hydrodynamic modeling in practice since 1986. Today all available (including wave effects), watershed modeling of Flood Hazard Mapping are being posted on the storm runoff, and spatial mapping of inundation internet (WMO Draft Report, Typhoon Com- (Fig. 14). mittee, Macao, Republic of China, 2006). A pilot In addition to forecasts of surge timing project in the Philippines produced a prelim- and height, flood inundation maps are increas- inary flood hazard map of flood-prone areas in ingly being requested by emergency managers the San Juan River Basin. Flood maps were and decision makers. Maps are needed not produced for Kuala Lumpur in Malaysia, as only of coastal storm surge, but also inland well as maps showing minimum, moderate, and flooding resulting from high rainfall associated severe flooding for the Gomback basin. Most of with cyclones. In the USA, since the NWS the mapping projects mentioned above involved instituted a program to model tropical cyclone

Figure 14. Suite of models required to simulate inundation from storm tides and upland flooding. March 13, 2010 10:34 B-936 b936-ch12

396 Shishir K. Dube et al.

storm surge, more fatalities occur from inland In the USA, after Hurricane Floyd (1999) flooding resulting from tropical cyclones than the State of North Carolina began a project to from storm surge (not including statistics from generate maps depicting inundated areas. This Hurricane Katrina), although this is not the case was the first NWS project involving the gen- in all areas of the world. Flood inundation maps eration of inundation map libraries for a river are useful in advance planning, as well as for reach (Fig. 15). The graph below is a represen- response during events and for post-assessment tation of flood inundation for NWS flood cat- after events. egories. These maps are based on steady-state

Figure 15. Inundation map for the Neuse River near Clayton, North Carolina, USA. (http://newweb.erh.noaa.gov/ ahps2/inundation/inundation.php?wfo=rah&gage=clyn7) March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 397

hydraulic modeling of water surface elevations estimated using a spline interpolation. Inun- for incremented discharges. dated areas were identified by subtracting the The alternative to the use of the steady- water-surface elevation in each grid cell from the flow assumption and development of inundation land-surface elevation in the cell. An automated map libraries is to estimate inundated areas in procedure was developed to identify all inun- real time during, or immediately prior to, a dated cells that were hydraulically connected to flood so that the particular characteristics of the cell at the downstream-most gauge in the the rainfall and flood hydrograph are well rep- model domain. This process resulted in a set resented in the hydraulic modeling. The lim- of inundation map libraries for each modeled itation of this approach is that the models reach. Inundation polygons were merged with must be executed in real time for each event a variety of other geospatial data to provide and that results must be distributed quickly to information for flood mitigation and emergency emergency management officials and all other response. interested parties. Moreover, there will be some At this time, the NWS does not use 2-D uncertainty in the forecast flows for which hydraulic models for operational purposes, so the inundation modeling is to be conducted. tests and pilot projects developed for dynamic A recent pilot project in the USA to test and maps have mainly focused on areas for which the evaluate dynamic mapping concluded that on- 1-D approach is valid or can be approximated. going efforts in static mapping should be fully However, in one of the pilot projects, St. Johns tested and evaluated before embracing imple- River, Florida there was an opportunity to test mentation of operational, real-time (dynamic) the coupling by using the 1-D hydraulic model inundation mapping (NWS, Office of Hydro- to generate flow outputs which in turn were used logic Development, R-Time Report, 2007, in as inputs for a 2-D estuarine model. The estu- review). arine model was also used to forecast salinity Several approaches have been taken to and temperature. develop inundation maps and include: (i) assum- Currently, the USA’s strategy is to develop ing that a storm surge of a given height will static maps for flood-prone areas and grad- inundate or impact up to that height of the ually develop hydrodynamic models for estu- land contour, (ii) pre-generating a library of aries to provide real-time flood maps. The maps for a range of water levels (static maps), increased demand of probabilistic inundation and (iii) generating inundation maps from real- maps by emergency managers is being recog- time forecasts of water level (dynamic maps) nized, but the need for developing operational based on unique features of a given event. procedures remains, including priorities for Although dynamic mapping might seem the addressing mapping uncertainty. best option, implementation can be difficult and costly. 6.2.2. Case Study: Andhra Pradesh An example of static maps is presented Coast of India by Bales et al., 2007. He generated a set of water-surface profiles at 0.305 m (1 ft) incre- Rao (1968) classified the Indian coastline into ments for a reach of the Tar River. Based on three categories based on combined storm surges the water-surface profile, a water-surface ele- and wind waves. According to this classification, vation was assigned to each cross section in the Andhra Coast of India from 14◦ N to 16.5◦ the reach; the water surface was assumed to N falls in the B-category (2 to 5 m surge) with be level across the cross section, which is con- a short C-type belt (> 5 m surge) near Nizam- sistent with the 1-D modeling approach. Water- patnam bay. According to an analysis of his- surface elevations between cross sections were torical records by Jayanthi (1999), the Andhra March 13, 2010 10:34 B-936 b936-ch12

398 Shishir K. Dube et al.

coast is high-risk prone with a small very high- In summary, the approach for determi- risk prone zone near Nizampatnam Bay. The nation of the Physical Vulnerability (PV) is as storm surges occurred that during 1977 and follows: 1990 near Machilipatnam further support the (a) A database of Tropical Cyclone (TC) gen- vulnerability of Andhra coast for disastrous erated Storm Surges (SS) impacting the surges. In recent years, there has been consid- AP Coast was drawn from the India Mete- erable concern regarding the vulnerability of orological Department (IMD) and from coasts due to cyclones and associated surges in several other national and international view of projected global warming and sea level sources. rise. In this section, we have undertaken as a (b) Because of climate change, projections into case study the development of Disaster Man- the future were limited to 50 years all the agement Plan (DMP) for cyclones and asso- available cyclone tracks for AP were synthe- ciated storm surges for mitigation in the nine sized into composite tracks to cover each of coastal districts of the State of Andhra Pradesh the coastal districts of AP. (AP), India. (c) Making use of the projected pressure Based on historical cyclone data, through a drop, the IIT-Delhi Storm Surge Model simple statistical analysis, Delta P (atmospheric was applied using the synthetic tracks to pressure deficit) was determined for cyclones determine the maximum possible storm making landfall on the AP coast, for return surge amplitude (during a 50 year period) periods of 2, 5, 10, 25 and 50 years. The at various locations along the AP Coast. Storm Surge Model developed by IIT-Delhi was (d) The Total Water Level Envelope (TWLE) applied with the 50-year Delta P value for a was determined by superimposing the tidal set of synthetic tracks, which were developed by amplitudes and wind wave setup on the compositing actual tracks, ensuring that each surge amplitudes. coastal district was covered. The results of the (e) These water levels were then projected onto computer simulations, calibrated with observed the coastal land using onshore topography surge data for each region of the coast, pro- data to demarcate the horizontal extent of vided maximum probable surge amplitudes at inundation. the mandal level, which is the geographical (f) This conservative approach may slightly unit immediately below the district level, and over-estimate the extent of inundation, but is made up of several villages and maybe is desirable for Hazard Mitigation and for towns. Coastal Zone Management, and is widely A generally accepted procedure in deter- used around the world. mining the extent of land inundation by a (g) Maps of regions subjected to possible wind storm surge is to assume that, a water level of damage from cyclones also were prepared. 5-m at the coastline would have an impact up to the 5-m land contour, a 10-m water level 6.2.2.1. Physical Vulnerability (PV) would impact up to the 10-m contour, and so Maps for the Coastal Districts on. This is a standard approach when very detailed orographic information is not available Inundation by storm surge and regions sub- and might somewhat over-estimate the extent jected to wind damage were mapped for the of inundation, but is an acceptable approach districts of Prakasham (Fig. 16) and Guntur for coastal zone storm mitigation planning (Fig. 17) of coastal AP. The PV maps were pre- purposes. pared for four scenarios: (a) frequent (10-percent March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 399

(a)

16

Latitude 15.5

15 79 79.5 80 80.5 Longitude (b)

16

Latitude 15.5

15 79 79.5 80 80.5 Longitude

Figure 16. (a) Land inundation map affected by storm surge, (b) regions affected by cyclonic winds for Prakasham District, Andhra Pradesh, India (after Dube et al., 2009).

annual recurrence interval), (b) infrequent The three large rivers in AP, Godavari, (2-percent annual recurrence interval), (c) a Krishna and Pennar, are subject to storm surge future climate scenario resulting in an intensi- penetration. The storm surge penetration into fication of the pressure field by 5 percent, and these rivers was determined by projecting the (d) a more extreme case of intensification of surge water levels into the rivers. It was assumed 7 percent. that for a river with many meanders, the March 13, 2010 10:34 B-936 b936-ch12

400 Shishir K. Dube et al.

(a) 16.8

16.6

16.4

Latitude 16.2

16

15.8 79.5 80 80.5 81 Longitude

(b) 16.8

16.6

16.4

Latitude 16.2

16

15.8 79.5 80 80.5 81 Longitude

Figure 17. (a) Land inundation map affected by storm surge, (b) regions affected by cyclonic winds for Guntur District, Andhra Pradesh, India (after Dube et al., 2009).

storm surge would penetrate 10 percent farther 6.2.2.2. Social Vulnerability (SV) than on land. If the river had few meanders, the increased penetration was 15 percent. Social Vulnerability was developed for physically The 10–15 % numbers are arrived at based vulnerable mandals. By using available popu- upon actual observations of storm surge pene- lation and other data, along with the PV maps, tration through these rivers (Murty, 1984). PV overall cyclone vulnerability index maps were maps for storm surge penetration up the rivers developed. Figure 19 shows the map for one of were then prepared (Fig. 18). the districts of coastal Andhra Pradesh. March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 401

17

16.5 Latitude

16

80 80.5 81 81.5 Longitude

Figure 18. Storm surge penetration through the Krishna River System (after Dube et al., 2009).

Overal Cyclone Vulenerability Index Map of Prakasham District

16

Latitude 15.5

15 79 79.5 80 80.5 Longitude

Figure 19. Overall Cyclone Vulnerability map for Prakasham District, Andhra Pradesh, India (after Dube et al., 2009). March 13, 2010 10:34 B-936 b936-ch12

402 Shishir K. Dube et al.

References Bogden, P., T. Gale, G. Allen, J. MacLaren, G. Almes, G. Creager, J. Bintz, L. D. Wright, H. Aijun, Z., E. Myers, and F. Aikman III, 2006: The Graber, N. Williams, S. Graves, H. Conover, St. Johns River Operational Forecast System K. Galluppi, R. Luettich, W. Perrie, B. Toulany, (SJROFS) and Its Skill Assessment. Coastal Y.P.Sheng,J.R.Davis,H.Wang,and Survey Development Laboratory, National Ocean D. Forrest, 2007: Architecture of a com- Service, NOAA, Silver Spring, Maryland 20910. munity infrastructure for predicting and ana- February 10, 2006. lyzing coastal inundation, Marine Technology Ali, A., 1979: Storm surges in the Bay of Bengal and Society Journal, 41(1), 53–61. some related problems, Ph.D. Thesis, University Brown, J. D., T. Spencer, and I. Moeller, 2007: Mod- of Reading, England, 227 pp. eling storm surge flooding of an urban area with Atlantic Oceanographic and meteorological labo- particular reference to modeling uncertainties: ratory, Hurricane Research Division, (http:// A case study of Canvey Island, UK: Water www.aoml.noaa.gov/hrd/hrd sub/milestones. Resources Research, 43(6):W06402, 22 p. html Cheung,K.F.,A.C.Phadke,Y.Wei,R.Rojas, Bales,J.D.,C.R.Wagner,K.C.Tighe, Y. J.-M. Douyere, N. Dodd, S. Liao, and S. Terziotti 2007: LiDAR-derived flood- and E. Nakazaki, 2003: Modeling of storm- inundation maps for real-time flood-mapping induced coastal flooding for emergency man- applications, Tar River basin, North Carolina: agement. Ocean Engineering 30, 1353–1386. US Geological Survey Scientific Investigations www.sciencedirect.com Report 2007–5032m, 42 p. Chittibabu, P. 1999: Development of storm surge Bender,M.A.,I.Ginis,R.Tuleya,B.Thomas, prediction models for the Bay of Bengal and and T. Marchok, 2007: The operational GFDL the Arabian Sea, Ph.D. Thesis, IIT Delhi, India, coupled hurricane-ocean prediction system and a 262 pp. summary of its performance. Mon. Wea. Rev., Chittibabu, P., S. K. Dube, A. D. Rao, P. C. Sinha, 135(12), 3965–3989. and T. S. Murty, 2000: Numerical simulation of Berenbrock, C. and R. R. Mason, 2008: Mapping extreme sea levels using location specific high res- Hurricane Rita inland storm-tide: Toronto, Pro- olution model for Gujarat coast of India, Marine ceedings of the 4th International Symposium on Geodesy 23, 133–142. Flood Defence, Institute for Catastrophic Loss Chittibabu, P., S. K. Dube, P. C. Sinha, A. D. Reduction, 8 p. Rao, and T. S. Murty, 2002: Numerical simu- Blain, C. A., J. J. Westerink, and R. A. Luettich, Jr., lation of extreme sea levels for the Tamil Nadu 1998: Grid convergence studies for the prediction (India) and Sri Lanka coasts, Marine Geodesy 25, of hurricane storm surge. Int. J. Numer. Methods 235–244. in Fluids, 26, 369–401. Chow, V. T., 1959: Open Channel Hydraulics, Blumberg, A. F. and G. L. Mellor, 1987: A McGraw-Hill, New York. description of a three-dimensional coastal ocean Crossett, K. M., T. J. Culliton, P. C. Wiley, and circulation model, in Three-Dimensional Coastal T. R. Goodspeed, 2004: Population trends along Ocean Models,N.S.Heaps,(ed.),AmericanGeo- the coastal United States: 1980–2008. Coastal physical Union, Washington, DC, pp. 1–16. Trends Report Series, National Oceanic and Bode, L. and T. Hardy, 1997: Progress and Atmospheric Administration/National Ocean recent developments in storm surge modeling. Service/Management and Budget Office/Special J. Hydraulic Engr., 123(4), 315–331. Projects, Silver Spring, Maryland. Bode, L., and L. B. Mason, 1994: Application Cushman-Roisin, B. 1994: Introduction to Geo- of an implicit hydrodynamic model over a physical Fluid Dynamics. Prentice-Hall, 320 p. range of spatial scales, Proceedings of 6th Com- Das, P. K., 1972: A prediction model for storm putational Techniques and Applications Con- surges in the Bay of Bengal, Nature 239, 211– ference (CTAC93),D.Stewart,H.Gardnerand 213. D. Singleton, (eds.), World Scientific, Singapore Das, P. K., 1994a: Prediction of storm surges in the pp. 112–121. Bay of Bengal, Proc.Indian.Natn.Sci.Acad.60, Bode, L., L. B. Mason and J. H. Middleton, 1997: 513–533. Reef parameterisation schemes for long wave Das, P. K., 1994b: On the prediction of storm surges, models, Progress in Oceanography, 40, 285–324. Sadhana, 19, 583–595. March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 403

Das, P. K., M. C. Sinha, and V. Balasubramanyam, Dube, S. K., I. Jain, and A. D. Rao, 2006: Numerical 1974: Storm surges in the Bay of Bengal, Quart. storm surge prediction model for the North J. Roy. Meteor. Soc. 100, 437–449. Indian Ocean and the South China Sea, Disaster Das, P. K., S. K. Dube, U. C. Mohanty, P. C. Sinha, and Development 1, 47–63. and A. D. Rao, 1983: Numerical simulation of the Dube,S.K.,I.Jain,A.D.RaoandT.S. surge generated by the June 1982 Orissa cyclone, Murty, 2009: Storm surge modeling for the Bay Mausam 34, 359–366. of Bengal and Arabian Sea, Natural Hazards, Dube, S. K. and V. K. Gaur, 1995: Real time storm 51, 3–27. surge prediction system for the Bay of Bengal, Ebersole,B.A.,D.T.Resio,andJ.J.West- Current Science 68, 103–113. erink, 2007: A community approach to improved Dube, S. K., P. C. Sinha, and G. D. Roy, The prediction and characterization of coastal storm numerical simulation of storm surges along hazards, Marine Technology Society Journal, the Bangladesh coast, Dyn. Atmos. Oceans 9, 40(4), 56–68. 121–133. Fread, D. L., NWS Operational Dynamic Wave Dube,S.K.,P.C.Sinha,A.D.Rao,andG.S.Rao, Model, verification of mathematical and physi- 1985b: Numerical modeling of storm surges in the cal models, in Hydraulic Engineering, Proceed- Arabian Sea, Appl. Math. Modeling 9, 289–294. ings of 26th Annual Hydraulic Division Specialty Dube, S. K., P. C. Sinha, and G. D. Roy, 1986: Conference; 04/16/2002 II.4-DWOPER-13 rfs: The effect of continuously deforming coastline 24dwoper.wpd, College Park, MD, August 1978, on the numerical simulation of storm surges in pp. 455–464. Bangladesh, Math. Comput. Simul. 28, 41–56. Fryar, R. M, B.A. Harper and I. Botev, 2004: Storm Dube, S. K., A. D. Rao, P. C. Sinha, and tide modeling of the Whitsunday coast and resort P. Chittibabu, 1994: A real time storm surge pre- islands, Proc Conference Hydraulics, Engineers diction system: An application to East coast of Australia, July, 2004. India, Proc. Indian Natn. Sci. Acad. 60, 157–170. Ghosh, S. K., 1977: Prediction of storm surges on Dube, S. K., A. D. Rao, P. C. Sinha, T. S. Murty, the east coast of India. Ind. J. Meteo. Geophys, and N. Bahulayan, 1997: Storm surge in the Bay 28, 157–168. of Bengal and Arabian Sea: The problem and its Glahn, B., A. Taylor, N. Kurkowski, and W. A. prediction, Mausam 48, 283–304. Shaffer. In preparation 2003: SLOSH’s role in Dube, S. K., A. D. Rao, P. C. Sinha, and National Weather Service Storm Surge Fore- P. Chittibabu, 2000a: Storm surge modeling and casting. prediction, in Mathematical Analysis and Appli- Global Climate and Weather Modeling Branch, cations (ed. Dwivedi), Narosa Publishing House, 2003: The GFS Atmospheric Model, NCEP New Delhi, India, pp. 109–124. Office Note 442, National Oceanic and Dube, S. K., P. Chittibabu, A. D. Rao, P. C. Sinha, Atmospheric Administration/National Weather and T. S. Murty, 2000b: Extreme sea levels asso- Service/National Centers for Environmental Pre- ciated with severe tropical cyclones hitting Orissa diction/Environmental Modeling Center, Camp coast of India, Marine Geodesy 23, 75–90. Springs, Maryland, 14 p. Dube, S. K., P. Chittibabu, A. D. Rao, P. C. Sinha, G¨onnert, G., S. K. Dube, T. Murty, and W. Siefert, and T. S. Murty, 2000c: Sea levels and coastal 2001: Global Storm Surges: Theory, Observations inundation due to tropical cyclones in Indian and Applications,DieKueste,623pp. coastal regions of Andhra and Orissa, Marine Graber, H. C., V. J. Cardone, R. E. Jensen, Geodesy 23, 65–73. D. N. Slinn, S. C. Hagen, A. T. Cox, Dube, S. K., P. Chittibabu, P. C. Sinha, A. D. Rao, M. D. Powell, and C. Grassl, 2006: Coastal fore- and T. S. Murty, 2004: Numerical modeling of casts and storm surge predictions for tropical storm surges in the head Bay of Bengal using cyclones, Oceanography, 19(1), 130–141. location specific model, Natural Hazards 31, 437– Harper, B. A., Storm tide threat in Queensland: 453. History, prediction and relative risks, Conser- Dube,S.K.,P.C.Sinha,A.D.Rao,I.Jainand vation Technical Report No. 10, Queensland N. Agnihotri, 2005: Effect of Mahanadi river on Government, Dept. of Environment and Her- the development of storm surge along the Orissa itage, January 1999. coast of India: A numerical study, Pure Appl. Harper, B.A, M.R. Gourlay and C.M. Jones, Geophys. 162, 1673–1688. 2001a: Storm tide risk assessment for the Cocos March 13, 2010 10:34 B-936 b936-ch12

404 Shishir K. Dube et al.

(Keeling) Islands, Proc. 15th Australasian Con- Arabian Sea. Presented at National Conference ference on Coastal and Ocean Engineering, on Tropical Meteorology (Tropmet 99), Chennai, Coasts and Ports 2001, Engineers Australia, Gold India. 16–19, February 1999. Coast, September, pp. 75–80. Jelesnianski, C., J. Chen and W. Shaffer, SLOSH: Harper,B.A.(ed.),T.Hardy,L.Mason,L.Bode, Sea, lake, and overland surges from hurricanes. I. Young, and P. Nielson, 2001b: Queensland Technical Report NWS 48, National Oceanic Climate Change and Community Vulnerability to and Atmospheric Administration, Silver Spring, Tropical Cyclones: Ocean Hazards Assessment — Maryland (1992). At URL http://www.csc. Stage 1, State of Queensland Department of noaa.gov/hes/images/pdf/SLOSH TR48.pdf Natural Resources and Mines, 383 pp. Jelesnianski, C., J. Chen, W. A. Shaffer, and Harper, B., T. Hardy, L. Mason, and R. Fryar, 2007: A. J. Gilad, 1984: SLOSH — a hurricane Developments in storm tide modeling and risk storm surge forecasting model. Preprints, Oceans assessment in the Australian region, Proceedings 84, Washington, D. C., Marine Technology of JCOMM Scientific and Technical Symposium on Society and IEEE/Oceanic Engineering Society, Storm Surges, Seoul, Korea, 2–6 October 2007. pp. 314–317. Heaps, N. S., 1973: Three-dimensional numerical Jelesnianski, C. P. and A. D. Taylor, 1973: NOAA model for the Irish Sea, Geophys., J. R. Astron. Technical Memorandum,ERL,WMPO-3,33pp. Soc. 35, 99–120. Johannessen, J., Lofti, A., and Harald J, 2006: ESA, Hicks, S. D., R. L. Sillcox, C. R. Nichols, B. Via, Benefiting our Economy, Observing the Earth, andE.C.McCray,Tide and Current Glossary. http://www.esa.int/esaEO/SEMKZSMVGJE National Oceanic and Atmospheric Admin- economy 0.html istration/National Ocean Service/Center for Johns, B. and A. Ali, 1980: The numerical modelling Operational Oceanographic Products and Ser- of storm surges in the Bay of Bengal, Quart. J. vices, 2000, Silver Spring, Maryland, http://co- Roy. Meteor. Soc. 106, 1–8. ops.nos.noaa.gov/publications/glossary2.pdf Johns, B., S. K. Dube, U. C. Mohanty, and P. C. (accessed 22 April 2008). Sinha, 1981: Numerical simulation of the surge Holland, G. J. 1980: An analytic model of the wind generated by the 1977 Andhra Cyclone, Quart. and pressure profiles in hurricanes, Mon. Wea. J. Roy. Meteor. Soc. 107, 915–934. Rev., 108, 1212–1218. Johns, B., P. C. Sinha, S. K. Dube, U. C. Mohanty, IMD Report, Report on cyclonic disturbances over and A. D. Rao, 1983: Simulation of storm surges North Indian Ocean during 1998, Abridged using a three-dimensional numerical model: an report for circulation during the meeting of application to the 1977 Andhra Cyclone, Quart. WMO/ESCAP panel on Tropical cyclones. J. Roy. Meteor. Soc. 109, 211–224. RSMC-Tropical cyclones. New Delhi, 1999, Johns, B., A. D. Rao, S. K. Dube, and P. C. Sinha, 71 pp. 1985: Numerical Modeling of the tide surge inter- IOC-UNESCO, Project proposal on storm surges action in the Bay of Bengal, Phil. Trans. Roy Soc. disaster for the northern part of the Indian London A 313, 507–535. Ocean, Project no. IOC/INF-1131, Paris, 1999. Joint Typhoon Warning Center (JTWC), Northern Jain, Indu, P. Chittibabu, Neetu Agnihotri, S. K. Indian Ocean Tropical Cyclone Best Track Data. Dube, P. C. Sinha, and A. D. Rao, 2006a: Sim- Retrieved on 2007-06–04, 2007. ulation of storm surges along Myanmar coast Kinnmark, I., The Shallow Water Wave Equations: using a location specific numerical model, Natural Formulation, Analysis, and Application. Lecture Hazards 39, 71–82. Notes in Engineering No. 15, Springer-Verlag, Jain, Indu, P. Chittibabu, Neetu Agnihotri, S. 1986. K. Dube, P. C. Sinha, and A. D. Rao 2006b: Knabb, R. D., J. R. Rhome, and D. P. Brown, Numerical storm surge prediction model for Tropical Cyclone Report, Hurricane Katrina, Gujarat coast of India and adjoining Pakistan 23–30 August 2005, National Oceanic and coast, Natural Hazards 42, 67–73. Atmospheric Administration/National Weather Jarvinen, B. and M. Lawrence, 1985: An evaluation Service/National Hurricane Center, 2006, http:// of the SLOSH storm-surge model. Bulletin of the www.nhc.noaa.gov/pdf/TCR-AL122005 Katrina. American Meteorological Society, 66(11), 1408– pdf (accessed 23 April 2008). 1411. Kolar,R.,J.J.Westerink,M.Cantekin,and Jayanthi, N., Storm surges and its risk assessments C. A. Blain, 1994: Aspects of nonlinear sim- over the coastal areas of Bay of Bengal and ulations using shallow models based on the March 13, 2010 10:34 B-936 b936-ch12

Storm Surge Modeling and Applications in Coastal Areas 405

wave continuity equation. Computers and Fluids, May 16, 2006, at URL http://www.nasa.gov/ 23(3), 523–538. vision/earth/lookingatearth/h2005 rita.html Luettich, R. and J. J. Westerink, 2004: Formulation National Hurricane Center 2007: Average NHC and Numerical Implementation of the 2D/3D 5-Year Average Forecast Errors (2002–2006), ADCIRC Finite Element Model Version 44.XX, All Tropical Cyclones. http://www.nhc.noaa. http://adcric.org/adcirc theory 2004 12 08.pdf gov/verification/pdfs/OFCL 5-yr averages.pdf, (access 8 May 2008). accessed June 3, 2008. Luettich, Jr., R., J. J. Westerink and N. Scheffner, Office of Hydrologic Development (OHD), Real- ADCIRC: An advanced three-dimensional circu- Time (Dynamic) Inundation Mapping Evalu- lation model for shelves, coasts, and estuaries. ation (R-Time) Team Report, December 2007 (in Technical Report DRP-92–6, US Army Corps of review). Engineers, Washington, D.C. (1992). Office of Hydrologic Development (OHD), Evalu- Lwin, T., Review of methods of storm surge pre- ation of Hydraulic Models, June 2007. diction currently used in Burma, Report to the Paudyal, G. N. 2002: Forecasting and warning of WMO Workshop on Storm Surges, Nov. 10–15, water-related disasters in a complex hydraulic Rangoon, Burma, 1980, 8p. setting—the case of Bangladesh. Hydrological Lynch, D. and W. Gray, 1979: A wave equation Sciences, 47(S), S5–S18. model for finite element tidal computations. Pedlosky, J. 1987: Geophysical Fluid Dynamics,2nd Computers and Fluids, 7(3), 207–228. edition, Springer, New York, 728 pp. Massey, W. G., J. W. Gangai, E. Drei-Horgan, Peng, M., L. Xie and L. Pietrafesa, 2004: A and K. J. Slover, 2007: History of Coastal numerical study of storm surge and inundation Inundation Models, Marine Technology Society in the Croatan-Albemarle-Pamlico estuary sys- Journal, 41(1), 7–17. tem.” Estuarine, Coastal and Shelf Science, 59, Mastenbroek,C.,G.Burgers,andP.A.E.M. 121–137. Janssen, 1993: The dynamical coupling of a wave Pietrafesa, L. J., K. Kelleher, T. Karl, M. Davidson, model and a storm surge model through the M.Peng,S.Bao,D.Dickey,L.Xie,H.Liu, atmospheric boundary layer. J.Phys.Oceanogr., and M. Xia, 2007: A new architecture for coastal 23, 1856–1867. inundation and flood warning prediction, Marine McGee,B.D.,R.W.Tollett,andR.R.Mason, Technology Society Journal, 40(4), 71–77. 2006a: Monitoring inland storm surge and Powell, M., S. Houston, L. Amat and N. Morrisseau- flooding from Hurricane Rita: US Geological Leroy, 1998: The HRD real-time hurricane wind Survey Fact Sheet 2006–3136, 4 p. analysis system. Journal of Wind Engineering McGee, B. D., B. B. Goree, R. W. Tollett, and Industrial Aerodynamics, 77–78, 53–64. B. K. Woodward, and W. H. Kress, 2006b: Hur- Powell, M., P. Vickery and T. Reinhold, 2003: ricane Rita Surge Data, Southwestern Louisiana Reduced drag coefficient for high wind speeds and Southeastern Texas, September to November in tropical cyclones. Nature, 422: 279–283. 2005. US Geological Survey Data Series 220. Pugh, D. T., 2004: Changing Sea Levels: Effects of http://pubs.usgs.gov/ds/2006/220/ Tides, Weather, and Climate. Cambridge Uni- McInnes, K., G. Hubbert, D. Abbs and S. Oliver, versity Press, Cambridge, 280 p. 2002: A numerical modelling study of coastal Rao, A. D., 1982: Numerical storm surge prediction flooding. Meteorology and Atmospheric Physics, in India, Ph.D. thesis, IIT Delhi, New Delhi, 80: 217–233. 211 pp. Murty, T. S., Storm Surges: Meteorological Ocean Rao, A. D., S. K. Dube, and P. Chittibabu, 1994: Tides, Department of Fisheries and Oceans, Finite difference techniques applied to the sim- Ottawa, Canada, 1984. ulation of surges and currents around Sri Lanka Murty, T. S. and R. F. Henry, 1983: Tides in the and Southern Indian Peninsula, Computational Bay of Bengal. Journal of Geophysical Research, Fluid Dynamics 3, 71–77. 88(c-10), 6069–6076. Rao, N. S. B., 1968: On some aspects of local and Murty, T. S., R. A. Flather, and R. F. Henry, 1986: tropical storms in the Indian area. Ph.D. Thesis, The storm surge problem in the Bay of Bengal, University of Jadavapur, India, 262 pp. Prog. Oceanog. 16, 195–233. Rao, N. S. B. and S. Mazumdar, 1966: A technique National Aeronautics and Space Administration, for forecasting storm waves, Indian Journal of 2006: Life on Earth — Looking at Earth: accessed Meteorology and Geophysics 17, 333–346. March 13, 2010 10:34 B-936 b936-ch12

406 Shishir K. Dube et al.

Rao, Y. R., P. Chittibabu, S. K. Dube , A. D. Rao, Unisys, http://weather.unisys.com/hurricane/ and P. C. Sinha, 1997: Storm surge prediction indian oc/index.html. and frequency analysis for Andhra coast of India, Walters, R. and G. Carey, 1984: Numerical noise in Mausam 48, 555–566. ocean and estuarine models. Advances in Water Reed, D. B. and B. E. Stucky, 2005: Fore- Resources, 7, 15–20. casting Hurricane Storm Surge on the Mississippi Weisberg, R. and L. Zheng, 2006: Hurricane storm River. Proceedings of ASCE Conference “Solu- surge simulations for Tampa Bay. Estuaries and tions to Coastal Disasters 2005”, Charleston, SC, Coasts, 29(6A), 899–913. May 8–11. Weiyan, T., 1992: Shallow Water Hydrodynamics. Riverside Technology, Inc., Tar River Calibration, Elsevier Oceanography Series, V. 55, Ams- 2006. terdam, 434 p. Riverside Technology, Inc., FLDVIEW Evaluation, Westerink,J.J.,R.A.Luettich,J.C.Feyen, 2007. J.H.Atkinson,C.N.Dawson,M.D.Powell, Roy, G. D., 1984: Numerical storm surge prediction J. P. Dunion, H. J. Roberts, E. J. Kubatko, and in Bangladesh. Ph.D. Thesis, Indian Institute of H. Pourtaheri, 2008: A Basin to Channel Scale Technology, Delhi, 188 pp. Unstructured Grid Hurricane Storm Surge Model Roy,G.D.,A.B.M.HumayuanKabir,M.M. Applied to Southern Louisiana. Mon. Wea. Rev., Mandal, and M .Z. Haque, 1999: Polar coordinates 136, 833–864. shallow water storm surge model for the coast of Westerink, J. J, R. A. Luettich Jr., and J. A. Bangladesh, Dyn. Atmos. Oceans 29, 397–413. Muccino, 1994: Modeling tides in the western Schuler, D. L. and Lee, J. S, 2004: Mea- North Atlantic using unstructured graded grids. surement of Ocean Wave Spectra and Surface Tel lus, 46A, 178–199. . Slopes by Polarimetric SAR, Remote Sensing WMO Draft Report. Reports of the Workshop on Division, http://www.nrl.navy.mil/content.php? Integrated Activities of the Hydrology, Meteo- P=04REVIEW186 rology and DPP Components of the Typhoon Systems Engineering Australia Pty Ltd, 2005: Committee into Related International Frame- Darwin TCWC Northern Region Storm Tide Pre- works for Disaster Risk Management for Better diction System — System Development Technical Impacts and Visibility: Hydrology Component. Report, Bureau of Meteorology Darwin, 208 pp. Economical and Social Commission for Asia and Sheng, Y. P., 1986: A Three-Dimensional Mathe- the Pacific and World Meteorological Organi- matical Model of Coastal, Estuarine and Lake zation, Macao, Republic of China 4–9 September Currents Using Boundary-Fitted Grid. Technical 2006. Report No. 585, Aeronautical Research Asso- Wright, C.W., E. J. Walsh, W. B. Krabill, ciates of Princeton, Princeton, New Jersey. D. Vandemark, A. W. Garcia, P. G. Black, Sheng, Y. P. and V. Alymov, 2002: Coastal F. D. Marks, Jr., R. A. Luettich, Jr., 2002: Real- Flooding Analysis of Pinellas County Using time storm surge measurement with a scanning ALSM Data: A Comparison between UF’s 2-D radar altimeter. Geoscience and Remote Sensing Method and Results vs. FEMA’s Method and Symposium, 2002. IGARSS apos; 02. 2002 IEEE Results. Report, Department of Civil and Coastal International, 3, pp. 1492–1495. Engineering, University of Florida. Xie, L., L. Pietrafesa and K. Wu, 2003: A numerical Sielecki, A., 1968: An energy-conserving difference study of wave-current interaction through surface scheme for the storm surge equations, Mon. and bottom stresses: Coastal ocean response to Weather Rev. 96, 150–156. Hurricane Fran of 1996. Journal of Geophysical Stamey, B., H. Wang, and M. Koterba, 2007: Pre- Research, 108(C2). dicting the next storm surge flood, Sea Tech- Zhang, Y.-L., A. M. Baptista, and E. P. Myers, nology, 48(8). 2004: A cross-scale model for 3D baroclinic cir- Thompson, E. and V. Cardone, 1996: Practical mod- culation in estuary-plume-shelf systems: I. For- eling of hurricane surface wind fields. Journal mulation and skill assessment. Continental Shelf of Waterway, Port, Coastal, and Ocean Engi- Research, 24, 2187–2214. neering, 122(4), 195–205.