Coastal Engineering 54 (2007) 377 391 www.elsevier.

com/locate/coastaleng

The effect of tidal inlets on open coast storm surge hydrographs

Michael B. Salisbury a,1 , Scott C. Hagen b,
b

Water Resources Staff Engineer, Ardaman and Associates, Inc., 8008 South Orange Avenue, Orlando, Florida, 32809, USA University of Central Florida, Department of Civil and Environmental Engineering, 4000 Central Florida Blvd., Orlando, Florida, 328162450, USA Received 25 January 2006; received in revised form 11 September 2006; accepted 12 October 2006 Available online 29 November 2006

Abstract Bridge scour modeling requires storm surge hydrographs as open ocean boundary conditions for coastal waters surrounding tidal inlets. These open coast storm surge hydrographs are used to accurately determine both horizontal and vertical circulation patterns, and thus scour, within the inlet and bay for an extreme event. At present, very little information is available on the effect that tidal inlets have on these open coast storm surge hydrographs. Furthermore, current modeling practice enforces a single design hydrograph along the open coast boundary for bridge scour models. This study expands on these concepts and provides a more fundamental understanding on both of these modeling areas. A numerical parameter study is undertaken to elucidate the influence of tidal inlets on open coast storm surge hydrographs. Four different inlet bay configurations are developed based on a statistical analysis of existing tidal inlets along the Florida coast. The length and depth of the inlet are held constant in each configuration, but the widths are modified to include the following four inlet profiles: 1) average Florida inlet width; 2) 100 m inlet width; 3) 500 m inlet width; and 4) 1000 m inlet width. Results from these domains are compared to a control case that does not include any inletbay system in the computational domain. The Advanced Circulation, Two-Dimensional Depth-Integrated (ADCIRC-2DDI) numerical code is used to obtain water surface elevations for all studies performed herein. The code is driven by astronomic tides at the open ocean boundary, and wind velocities and atmospheric pressure profiles over the surface of the computational domains. Model results clearly indicate that the four inletbay configurations do not have a significant impact on the open coast storm surge hydrographs. Furthermore, a spatial variance amongst the storm surge hydrographs is recognized for open coast boundary locations extending seaward from the mouth of the inlet. In addition, a storm surge study of Hurricane Ivan in the vicinity of Escambia Bay along the Panhandle of Florida is performed to assess the findings of the numerical parameter study in a real-life scenario. The main conclusions from the numerical parameter study are verified in the Hurricane Ivan study: 1) the Pensacola PassEscambia Bay system has a minimal effect on the open coast storm surge hydrographs; and 2) the open coast storm surge hydrographs exhibit spatial dependence along typical open coast boundary locations. The results and conclusions presented herein have implications toward future bridge scour modeling efforts. 2006 Elsevier B.V. All rights reserved.
Keywords: Tidal inlets; Storm surge hydrographs; Bridge scour; Shallow water equations; Hurricane Ivan

1. Introduction The design of coastal bridges is governed by a number of factors: wind, moving (vehicular), and hydrodynamic loads, to name a few. In particular, the coastal circulation patterns are important in determining the amount of scour that occurs during extreme flow events (e.g. hurricane storm surge). At present,
Corresponding author. Fax: +1 407 823 3315. E-mail addresses: msalisbury@ardaman.com (M.B. Salisbury), shagen@mail.ucf.edu (S.C. Hagen). 1 Fax: +1 407 859 8121. 0378-3839/$ - see front matter 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.coastaleng.2006.10.002

local three-dimensional models are used to estimate both horizontal and vertical circulation patterns, and thus bridge scour, in the vicinity of coastal bridges. Typically, these models encompass the bay system where the bridge is located and extend seaward to shallow ocean regions beyond tidal inlets. The magnitude for the design surge event (e.g. 50-yr, 100-yr, or 200-yr surge event) used to force a local bridge scour model (as used herein, the term bridge scour model refers to any high resolution, three-dimensional circulation model that can be applied to compute bridge scour) for a particular area is quite varied between government agencies (Sheppard and Miller, 2003). As a result, it is necessary to elucidate the behavior of

378

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

hurricane storm surge hydrographs along the coast. These open coast storm surge hydrographs will be generated by a largescale ocean circulation model and used as ocean boundary conditions for near-shore, high resolution models that include the inlet and bay system. These local, three-dimensional models will use the boundary conditions to better compute the horizontal and vertical flow patterns within the bay system for prediction of bridge scour. In the future, it is hoped that by generating these open coast storm surge hydrographs, a more accurate depiction of the effect of storm surge on bridge scour can be formulated. Two issues, however, arise when generating these open coast boundary conditions. First, the detail of the coastline resolution in the ocean circulation model becomes an issue. The Florida coast is abundant with tidal inlets that allow for water to continuously circulate through embayment systems. This hydraulic connection provides a conduit for storm surge to enter the bay system during a hurricane event. Incorporating all of the inlet and bay systems along the Florida coast into the model domain is an arduous task at best and it significantly increases the computational nodes included in the model. Therefore, it is necessary to elucidate the effect, if any, that tidal inlets have on the open coast storm surge hydrographs. In order to accomplish this, a numerical parameter study is performed by employing various idealized inlet profiles that are representative of Florida tidal inlets. Each idealized inlet mesh is forced with a synthetic wind field and pressure profile that is representative of Hurricane Ivan (September 2004) and the results of each simulation are compared to one another. Furthermore, the conclusions from the numerical parameter study are verified for Hurricane Ivan near Pensacola, Florida. By doing this, coastal modelers will be able to determine which inlets need to be included in the coastal circulation model, such that accurate open coast storm surge hydrographs can be generated. Second, proper application of the open coast storm surge hydrographs to a local circulation model is critical for the prediction of scour levels within the inlet and bay. Current modeling practice enforces a single design hydrograph along the ocean boundary of local two- or three-dimensional circulation models. This methodology, however, is inherently flawed because the change in bathymetry along the ocean boundary creates varying levels of surge along the coast. The use of a single design hydrograph does not incorporate this effect along the open coast boundary. In lieu of this, the spatial variance of the open coast storm surge hydrographs is examined in both an idealized setting and a Hurricane Ivan storm surge study. The results and conclusions of this study have implications toward more accurate bridge scour modeling in coastal areas. 1.1. Coastal scour modeling Coastal bridges are subjected to foundation scour as a result of circulation patterns found in the bay or estuary. Scour may occur as a result of density stratification, water salinity, freshwater riverine inflow, or astronomic tidal currents (Richardson et al., 1999). Under design conditions, these circulation patterns are often computed based on water elevation data during an extreme

flow event (e.g. hurricane storm surge). Typically, this water elevation data is obtained either through a deterministic hydrodynamic model or a stochastic formulation. Traditionally, hydrodynamics within an estuary are computed using local-scale, two- or three-dimensional models (Zevenbergen et al., 1999). The response of the model near the bridge is a function of the applied boundary conditions, which include open coast storm surge hydrographs. The applied open coast storm surge hydrographs are a function of peak water surface elevation, rising limb time, duration of peak, and falling limb time. The sensitivity of water currents in the model to the change in open coast hydrograph parameters has been well documented by Sheppard and Pritsivelis (1999). The results show that model response is very sensitive to the peak water surface elevation and the duration of peak, but less sensitive to the falling limb time. Hence, the accuracy of the model is heavily dependent on the applied open coast boundary conditions. In the past, peak water surface elevations were obtained from surge models developed by the Federal Emergency Management Agency (FEMA) and the National Oceanic and Atmospheric Administration (NOAA). FEMA employed a two-dimensional, finite difference model (SURGE) to compute maximum surge elevations along the coast. NOAA also employed a two-dimensional, finite difference model (SLOSH) that computes peak surge levels based on each class of hurricane. In both cases, a significant number of storm surge simulations are performed based on varying degrees of hurricane intensity and hurricane track. The results lead to maximum envelopes of water (MEOW) along the coast for use by consultants and emergency managers (Zevenbergen et al., 1999). In view of the fact that only peak values are provided in the MEOW, not the full storm surge hydrographs, Cialone et al. (1993) report a procedure for developing the full hydrograph based on maximum surge levels. The full storm surge hydrograph is computed as follows:   D Stot t Sp 1ejT tj Ht t 1 where, Sp = peak surge height, D = storm duration (defined as the radius to maximum winds divided by the storm's forward speed), T = time to peak surge, t = time, and Ht(t) = the daily tide component. If the daily astronomic tides are excluded from the equation, then the storm surge hydrograph is symmetric about time T. In the absence of more accurate computer generated storm surge hydrographs, this equation can be used to develop the open coast storm surge hydrographs used for boundary conditions in bridge scour modeling. Previous studies have applied this methodology (i.e. applying a single, synthetic hydrograph along the open coast boundary) in studying the influence of boundary conditions on velocities through the inlet and near bridge piers (Edge et al., 1999; Sheppard and Pritsivelis, 1999; Zevenbergen et al., 1999). There are two drawbacks, however, to using this single hydrograph approach. First, Eq. (1) produces a simple hydrograph that does not consider the influence of wind over time. As the hurricane traverses the coastline, the change in wind speed and direction may cause a negative surge that is not predicted by Eq. (1)

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391 Table 1 Average hydrodynamic characteristics of Florida's tidal inlets Parameter Width Depth Length Unit [m] [m] [m] Mean 557 4.7 3142 Standard deviation 822 3.5 6537

379

(Zevenbergen et al., 1999). Second, the spatial variance of the storm surge hydrographs along the open coast boundary is not considered. Previous studies have assumed that applying a single design hydrograph is sufficient, but results presented in this study clearly indicate that this is not a thoroughly accurate procedure. Results presented herein show that the open coast storm surge hydrographs are dependent upon the depth and location of the boundary node. This result potentially has significant implications for future bridge scour modeling studies. 1.2. Hurricane Ivan Hurricane Ivan (September 2, 2004September 24, 2004) was one of four devastating hurricanes that struck the Florida coast in 2004, along with Hurricanes Charley, Frances, and Jeanne. The damage caused by Ivan was estimated to be $14.2 billion (Stewart, 2004), ranking among the costliest hurricanes to ever impact the United States. Damage in Florida alone totaled more than $4 billion, with an estimated $900 million in damage to the Pensacola Naval Air Station. The death toll of Ivan reached 92 fatalities, with 25 deaths in the United States. The maximum sustained winds at landfall reached 210 km/h, with the most intense winds located in the northeast quadrant of the storm. The greatest storm surge occurred along the coast of eastern Alabama and western Florida, with peak surge reaching 34 m along the Panhandle of Florida. In the Escambia Bay region, the storm surge and wind waves were great enough to topple portions of the Interstate 10 Bridge, causing a 400 m section of the roadway to collapse into the bay. 2. Methodology In order to properly describe the physics of meteorological tides (storm surge), a numerical model must resolve coastal features that affect storm surge generation and propagation. Therefore, a model domain must describe complex coastal geometries (bathymetry and topography), large gradients in bathymetry along the continental shelf, and permit reasonable boundary conditions (i.e. tidal harmonics and/or water surface elevations). A finite element based model is ideal for such a task, as its flexibility allows for large spatial scales to be represented in the domain while allowing higher nodal resolution along the coastline (Blain et al., 1994). 2.1. Idealized domain Previous research has shown that an idealized finite element domain provides an excellent means for isolating and examining

inlet parameters. Hench et al. (2002) applied several idealized inlet configurations to examine momentum balances for shallow tidal inlets. Pandoe and Edge (2004) and Kubatko et al. (2005) utilized an idealized domain to study sediment transport through an inlet channel. For all cases, the structure of the domain provided a clear discernment of changes in model results to modifications in model parameters. Furthermore, a planform profile was used to represent the ocean basin and the bay with the inlet and bay dimensions assumed to be realistic (no supporting evidence was provided to verify this claim). Similarly, a planform profile is used in this study; however, the inlet and bay dimensions are modified to represent a range of inletbay configurations found along the Florida coast. 2.1.1. Statistical analysis A thorough review of existing Florida tidal inlets is performed in order to properly design the idealized finite element meshes used in the numerical parameter study. A total of 74 tidal inlets are identified along the Florida coast, 26 on the east coast and 48 on the west coast (Carr de Betts, 1999). A statistical analysis is performed on the hydrodynamic measurements associated with each tidal inlet. The mean and standard deviation of each parameter dataset is determined (Table 1). From this analysis it is noted that the variability amongst Florida tidal inlets is quite recognizable. For example, the average inlet length is 3142 m, while the standard deviation is 6537 m. This implies that a large range of inlet lengths is present in the dataset, which indicates that the mean values listed in Table 1 may not be representative of a typical Florida tidal inlet. In lieu of this, a z-score test is performed to remove the statistical outliers as follows (Mendenhall and Sincich, 1995): z y l r 2

where, y = data point value, = mean, and = standard deviation. The values for and are obtained from Table 1. By definition, the z-score describes the location of the data point relative to the mean value. In order to determine whether a value is considered an outlier, the computed z-score for each data point is compared to a tabulated value of z-scores. If the computed value from the dataset is greater than the tabulated value, |z| N |ztable|, then the data point is considered to be an outlier. For this study, the 0.05 significance level is used for the tabulated z-score values. This process is repeated for each of the datasets, and the outliers are removed. Next, another statistical analysis is performed on the revised dataset to determine mean values that are more representative of a typical Florida tidal inlet. Table 2 presents the results of this analysis.

Table 2 Modified statistics of Florida's tidal inlets with outliers removed Parameter Width Depth Length Unit [m] [m] [m] Mean 388 3.7 2257 Standard deviation 337 2.0 1814

380

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

sectional area of the inlet to the tidal prism for stable tidal inlets. As a comparison, the data of Florida tidal inlets are compared to the solutions obtained from Eq. (3). Overall, Eq. (3) provides an excellent fit to the raw data (Fig. 1). Thus, Jarrett's relationship is deemed appropriate for Florida's tidal inlets. Additionally, the tidal prism is related to the bay surface area through the following relationship (USACE, 2002): X 2ab Ab 4

Fig. 1. Comparison of Jarrett's relationship to Florida tidal inlet data.

2.1.2. Inletbay configuration Four different inlet width profiles are used in the numerical parameter study: an average inlet width (corresponding to the average width identified in Table 2), a 100 m inlet width, a 500 m inlet width, and a 1000 m inlet width. This range of inlet widths encompasses the majority of Florida's tidal inlets. For each inlet width configuration, the inlet length and depth are held constant according to the average values determined in Table 2. In contrast, the bay surface area varies between each inlet width configuration according to a tidal prisminlet crosssectional area relationship developed by Jarrett (1976). Converted to the System International unit system, the relationship is expressed as follows: Ac 2:09 10 X
5 0:95

where, ab = the bay tide amplitude, and Ab = the bay surface area. The bay tide amplitude is defined as one half the tidal range; thus, Eq. (4) relates the tidal prism to a product of the tidal range (2ab) and the bay surface area. Substituting this relationship into Eq. (3) yields the following solution: Ac 2:09 105 2ab Ab 0:95 : 5

Furthermore, it's assumed that the tidal range (2ab) is 1 m, which is archetypical of the tidal range along the coast of Florida [see Kojima (2005) for a detailed tidal analysis along the United States coast]. Substituting this assumption into Eq. (5), and rearranging for the bay surface area, leads to the following equation:  1=0:95 Ac Ab : 6 2:09 105 This expression is used to develop the inletbay configurations in the finite element meshes of the numerical parameter study. A rectangular cross-sectional profile is assumed for the entire length of the inlet; thus, the inlet cross-sectional area (Ac) can easily be determined based on the desired inlet width

where, Ac = minimum cross-sectional area of the inlet (m2) and = tidal prism (m3). This equation relates the minimum cross-

Fig. 2. Inletbay configurations for the idealized finite element meshes: a) average inlet width; b) 100 m inlet width; c) 500 m inlet width; and d) 1000 m inlet width.

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

381

Fig. 3. a) Finite element mesh for the idealized domain; b) bathymetry for the idealized domain. Thick dashed lines represent locations of open boundaries.

(average, 100 m, 500 m, and 1000 m) and the constant inlet depth (3.7 m). Following this procedure for each inlet width yields a unique bay surface area for each inlet configuration. The result of this process leads to the following inletbay configurations (Fig. 2).

The node spacing ranges from 50 m in the inlet to 100 m in the bay for the average, 500 m, and 1000 m inlet width configurations, and 25 m in the inlet to 50 m in the bay for the 100 m inlet width configuration. This level of resolution ensures that the inlet is represented by at least 4 elements across its width for

Fig. 4. a) Finite element discretization for the ocean-based domain; b) bathymetry for the ocean-based domain; c) finite element discretization near Escambia Bay; and d) bathymetry for coastal areas near Escambia Bay.

382

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

all inletbay configurations. Furthermore, the domain measures 150 km alongshore with the ocean boundary placed 100 km seaward from the coastline. A depth of 73 m is set at the ocean boundary and a linear transition is applied to the bathymetry profile between the ocean boundary and the coastline. Fig. 3a shows the finite element mesh for the idealized domain, and Fig. 3b displays the associated bathymetry profile. 2.2. Escambia Bay domain 2.2.1. Study location For the Hurricane Ivan analysis, a large-scale, ocean-based domain is employed with higher mesh resolution in the Escambia Bay region. The study area is located along the Panhandle of Florida in the northwestern portion of the state of Florida. The main focus of this study is an inletbay system that was devastated by storm surge from Hurricane Ivan in September 2004. The system includes the Pensacola Pass, Pensacola Bay, Escambia Bay, Blackwater Bay and East Bay (Fig. 4). The system is connected to two adjacent bay systems via a Gulf Intracoastal Waterway: Perdido Bay to the west and Choctawatchee Bay to the east. Pensacola Pass is a relatively deep inlet that connects the Pensacola Bay to the Gulf of Mexico. The inlet is part of Escambia County, and is bordered by two low-lying, sandy barrier islands: Perdido Key to the west and Santa Rosa Island to the east. The inlet became a federal navigation project in 1881 and is regularly dredged to a depth of approximately 15 m to accommodate U.S. Navy aircraft carriers (Carr de Betts, 1999). 2.2.2. Model domain An ocean-based domain is used to simulate storm surge due to Hurricane Ivan in the vicinity of Escambia Bay. The domain encapsulates portions of the Atlantic Ocean found west of the 60 West Meridian, the Gulf of Mexico, and the Caribbean Sea. The mesh was developed by incorporating the Escambia Bay coastline features into an existing 52,774 node mesh for the Western North Atlantic Tidal model domain [see Hagen et al. (in press) for verification of the existing mesh]. The new mesh includes 88,318 nodes and 165,137 elements, and covers a horizontal surface area of approximately 8.347 106 km2. Bathymetry for the Escambia Bay region was obtained via the National Geophysical Data Center (NGDC) Coastal Relief Model, volume 3. The database consists of a combination of 3arc second digital elevation maps (via the United States Geological Survey) and hydrographic soundings (via the National Ocean Service). For regions extending beyond the Escambia Bay and Pensacola Pass, the bathymetry was interpolated from an existing high resolution Western North Atlantic mesh. Fig. 4a through d presents the finite element discretization and bathymetry for the ocean-based domain with zoomed in views of the Escambia Bay and Pensacola Pass. In addition, a second computational domain is used that includes the low-lying barrier islands as inundation areas. The same ocean and Escambia Bay coastline features are used from the previously described mesh, but Santa Rosa Island to the east and Perdido Key to the west of Pensacola Pass are included as

inundation areas. Including the barrier islands into the model domain allows the storm surge to inundate the barrier islands, creating higher surge levels within the bay. Model results are compared to historical data at a National Ocean Service (NOS) tide gauge located in Pensacola Bay (Fig. 4d). 2.3. Storm surge model The water surface elevations and circulation patterns are computed by the Advanced Circulation model (ADCIRC) for shelves, coasts, and estuaries (Luettich et al., 1992). In this study, barotropic dynamics are predominant and density gradients are assumed to be relatively small, as these conditions are common near tidal inlets (Hench et al., 2002). These assumptions permit the use of the fully nonlinear two-dimensional, depth-integrated option of the model (ADCIRC-2DDI). Westerink et al. (1994) present the basic governing continuity Eq. (7) and momentum Eqs. (8) and (9) that are used in the model: Af AUH AVH 0 At Ax Ay   AU AU AU A ps U V f V gfag At Ax Ay Ax q0 1 sSX MX s U H q0 H   AV AV AV A ps U V fU gfag At Ax Ay Ay q0 1 sSY MY s V H q0 H 7

where, t = time; x,y = horizontal coordinates, aligned in the East and North directions respectively; = free surface elevation, relative to the geoid; U = depth-averaged horizontal velocity, x direction; V = depth-averaged horizontal velocity, y direction; H = total water column depth, h + ; h = bathymetric depth relative to the geoid; f = 2sin = Coriolis parameter; = angular speed of the earth; = degrees latitude; ps = atmospheric pressure at the free surface; g = acceleration due to gravity; = Newtonian equilibrium tide potential; = earth elasticity factor; 0 = reference density of water; SX = applied free surface stress, x direction; SY = applied free surface stress, y direction; p U 2 V 2 s C f h H = bottom i stress; Cf = bottom friction coefficient; M X = E h2
A VH sion, x direction; MY = Eh2 AAVH x2 Ay2 = depth-integrated momentum dispersion, y direction; and Eh2 = horizontal eddy viscosity. The governing equations are discretized in space by linear finite elements and in time by a finite difference scheme (Luettich et al., 1992). The finite element solution to the shallow water equations gives rise to spurious modes and numerical instabilities. Hence, it becomes necessary to reformulate the equations into a form that provides a stable solution in its finite element representation. Therefore, ADCIRC-2DDI employs the Generalized Wave Continuity Equation (GWCE), together with
2 2

A2 UH Ax 2

AAUH y2 = depth-integrated h i momentum disper2

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

383

model. In order to convert wind speed to wind stress, ADCIRC employs a relationship developed by Garrett (1977). Furthermore, the wind model also provides a pressure gradient to ADCIRC. A conversion is then performed within ADCIRC to convert the pressure gradient to an equivalent water column height through the transformation P / wg (Blain et al, 1994). In the end, the wind stress and the equivalent water column height are linearly interpolated to each computational node of the finite element mesh used by ADCIRC. 2.3.2. Synthetic wind field model For the numerical parameter study, a synthetic hurricane wind field model is used to force the ADCIRC-2DDI model. The wind field model is parameterized according to the forward velocity and pressure field of Hurricane Ivan, and simulates a simplified, circularly symmetric version of the storm. Compared to the wind field model used in the Hurricane Ivan study, the synthetic wind field model is much less sophisticated. However, the model does allow the user to specify the hurricane parameters and path of the storm, which makes it ideal for a numerical study such as this. In all cases presented herein, the wind field model is initialized to represent a typical worst-case scenario for coastal storm surge in the vicinity of the tidal inlet. First, the path of the storm is setup to traverse the coastline in a perpendicular manner. Additionally, the eye of the storm makes landfall at a point 25 km left of the tidal inlet, placing the inlet in the upper right quadrant of the storm (Fig. 6). 2.3.3. Model parameters for idealized domains For the idealized study, the inletbay configurations are isolated to examine their influence on the model results. Consequently, the model parameters are held constant for each simulation and are specified according to the following: The equations are solved in the Cartesian coordinate system. Simulations are spun up from rest over a 2-day period via a hyperbolic ramp function. Total simulation time is 4 days, with a 0.25 second time step used to ensure model stability. A hybrid bottom friction formulation is employed with the following specifications: Cfmin = 0.0025, Hbreak = 10 m, = 10, and = 1 / 3.

Fig. 5. Hurricane Ivan wind field just before landfall.

the momentum conservation equations, to eliminate this problem (Westerink et al., 1994). The result is a noise-free solution that is used to solve for the deviation from the geoid () and the velocities in the x and y directions. 2.3.1. Hurricane Ivan wind field model The state-of-the-art wind data (Fig. 5) for the hurricane storm surge simulations was provided using a tropical wind model developed by Cox and Cardone (2000). The model is a derivative of the TC96 model that was first implemented in the Ocean Data Gathering Program. The TC96 model provides snapshots in time that are blended into a synoptic-scale wind field using the Interactive Objective Kinematic Analysis algorithm (Cox et al., 1995) of the Wind Workstation. Using a numerical integration technique, the model solves the vertically averaged equations of motion for a boundary layer under horizontal and vertical stresses. Based on this principle, the numerical model provides a fairly thorough description of the timespace evolution of the wind speeds within the planetary boundary layer (PBL) during a tropical cyclone event (Thompson and Cardone, 1996). The wind field model is driven by snapshots in time of the storm's intensity and is based on the assumption that the structure of the hurricane changes relatively slowly (Cox and Cardone, 2000). In addition to the TC96 model, these snapshots are also obtained from the Hurricane Research Division Wind Analysis System, a distributed system that uses real-time tropical cyclone observations as input (Powell et al., 1998). The slow evolution of the storm's intensity is then interpolated from these snapshots. In addition, the model is also controlled by several input parameters: the storm's speed and direction, the geostrophic flow of the ambient PBL pressure field, a pressure profile parameter, and a scaling factor for the exponential radial pressure profile. The wind model is capable of simulating any type of storm for any particular region, and has been verified and validated for a number of test cases (Cox and Cardone, 2002). The wind speeds that are computed by the wind field model are used as surface stress conditions in the ADCIRC-2DDI

Fig. 6. Synthetic wind field used in the numerical parameter study.

384

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

Fig. 7. Semi-circular arcs with output locations for idealized domain.

and are ramped over a 20-day period. The hybrid bottom friction formulation is employed with the following settings: Cfmin = 0.0025, Hbreak = 1 m, = 10, and = 1 / 3. The horizontal eddy viscosity coefficient is set at 5 m2/sec, and a time step of 5 s is used to ensure model stability. Similarly, the model parameters for the storm surge simulations are set as follows: A spherical coordinate system is applied. Simulations are begun from a cold start, and are run over a ten day period (September 7, 2004 at 6 p.m. to September 17, 2004 at 6 p.m.) with a 2.5-day ramp time. Meteorological forcings (wind stress and pressure) are read into the model every 30 min. The wetting and drying of elements is employed with the minimum depth set to 0.1 m. A 2.5 second time step is used for each model domain to ensure stability. A hybrid bottom friction formulation is used with the following settings: Cfmin = 0.001, Hbreak = 10 m, = 10, and = 1 / 3, and the horizontal eddy viscosity coefficient is set to 5 m 2/sec. 2.3.6. Model output locations for the Escambia Bay domain In a similar manner to the idealized domain, semi-circular arcs of varying radii are extended seaward from the mouth of Pensacola Pass. The arcs range from 1 km to 15 km in radial distance. Fig. 8 shows the observation points along each semicircular arc extending seaward from Pensacola Pass. Observation points for locations due east and west from the mouth of the inlet are not included as some of the points would either be outside of the model boundaries or within the bay system. In this case, seven output stations are located on each semi-circular arc: eastsoutheast (1), southeast (2), south southeast (3), south (4), southsouthwest (5), southwest (6), and westsouthwest (7). These output locations are used to compare the effect that the Pensacola Pass has on the open coast storm surge hydrographs and to examine the spatial variance of the hydrographs along each arc. 3. Simulation results Primary focus is given to the effect that inletbay configurations have on the open coast storm surge hydrographs. Two

The horizontal eddy viscosity coefficient is set to 5 m2/sec. A constant Coriolis parameter corresponding to 27.5N (approximately the middle latitude of Florida) is used. One harmonic forcing frequency corresponding to the M2 (principal lunar) constituent is applied to the system with an amplitude of 0.5 m and 0 phasing. This produces a tidal range of approximately 1 m near the coastline, which is representative of the tidal range characteristics along the Florida coast. Meteorological forcings (wind stress and surface pressure) are read into the simulation every 30 min, with the eye of the synthetic hurricane making landfall approximately 3 days into the simulation. 2.3.4. Model output locations for idealized domains In lieu of the fact that previous bridge scour modeling efforts have incorporated open coast boundaries at varying distances from the mouth of the inlet, a wide range of semi-circular arcs are created with output stations located along each arc. These arcs have radii ranging from 1 km to 15 km from the mouth of the inlet, and are representative of typical open coast boundary locations used in bridge scour modeling. Water surface elevations are recorded at each of these stations over the entire duration of the simulation at six minute intervals. Fig. 7 displays the semi-circular arcs with the output locations for the different radii. 2.3.5. Model parameters for the Escambia Bay domain Astronomic tides are included in the Escambia Bay domain simulations; however, the astronomic tides and storm surge are computed separately due to the differences in model parameterization and simulation length. The total hydrograph is computed by superimposing the storm surge output with a resynthesis of the astronomic tides. The results are then compared to historical National Ocean Service (NOS) gauge data located within Pensacola Bay to verify the model. The model parameters for the astronomic tidal simulations are set as follows: The coordinate system is set to spherical. Total simulation length is 90 days with all runs begun from a cold start. Seven harmonic forcings are applied simultaneously along the ocean boundary (M2, K1, O1, N2, K2, Q1, and S2)

Fig. 8. Semi-circular arcs with output locations for Escambia Bay domain.

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

385

Fig. 9. Storm surge hydrographs at point 3 on the 1 km radius observation arc for each of the inletbay configurations.

Fig. 11. Storm surge hydrographs at point 3 on the 15 km radius observation arc for each of the inletbay configurations.

different studies are presented in this section to illustrate these results: 1) a numerical parameter study, where the effect of four different inlet widths (average, 100 m, 500 m, and 1000 m widths) on the open coast storm surge hydrographs are isolated and compared; and 2) a storm surge study that focuses on the effect of the Pensacola Pass and Escambia Bay system on the open coast storm surge hydrographs produced from Hurricane Ivan. Additional discussion is provided on the spatial variance of the open coast storm surge along the observation arcs. 3.1. Numerical parameter study results 3.1.1. Inlet comparison Ultimately, the results and conclusions from this study will be used to determine whether an inlet and bay system (e.g. the Pensacola Pass and Escambia Bay system) should be included in a large-scale ocean circulation model (e.g. the Western North Atlantic Tidal model domain) for generating open coast boundary conditions for local bridge scour models. In order to accomplish this, four inletbay configurations and a control case are developed in an idealized setting. The control mesh does not include an inletbay system along the coastline and

provides a test case for better determining the influence of the inletbay system on the open coast storm surge hydrographs. Model output is compared at each of the observation points for all of the inletbay configurations. Figs. 911, show model results for each of the inletbay configurations at point 3 (perpendicular to the coast) on the 1 km, 5 km, and 15 km radii arcs, respectively. It is evident from the plots that no appreciable difference exists in the storm surge hydrographs between the inletbay configurations. A difference of 5 cm is noticed in Fig. 9 (1 km radius) for the 1000 m inlet width, but the difference is practically negligible at the 5 km radius (Fig. 10) and the 15 km radius (Fig. 11). Similar results are obtained at all of the other model output locations, but are not included in the interest of brevity. It is apparent from this analysis that the flow through each of the inlets (velocity = 3.1 m/s) does not produce a large enough velocity head to draw down the water surface elevations along the open coast when compared to one another. Proper modeling practice dictates that the open coast boundary be placed far enough from the inlet to allow for appropriate circulation patterns to be generated within the model (Militello et al., 2000). Thus, bridge scour modeling studies would place the boundary farther than 1 km from the inlet to avoid momentum flux problems. At locations further

Fig. 10. Storm surge hydrographs at point 3 on the 5 km radius observation arc for each of the inletbay configurations.

Fig. 12. Storm surge hydrographs along the 1 km radius observation arc.

386

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

Fig. 13. Storm surge hydrographs along the 5 km radius observation arc.

Fig. 15. Astronomic tide comparison at the NOS station in Pensacola Bay.

from the inlet, the open coast storm surge hydrographs are nearly identical. Hence, the inletbay configurations do not appear to create a significant difference in the open coast storm surge hydrographs. Consequently, a large-scale model (e.g. the WNAT domain) would not need to include an inletbay system along the coastline when generating boundary conditions for a local bridge scour model. 3.1.2. Spatial variance Previous research has applied a single design hydrograph along the open coast boundary for bridge scour modeling studies. No consideration was given in these studies to the spatial variance of the open coast storm surge hydrographs along the boundary location. This was most likely due to the unavailability of a large-scale model to produce boundary conditions for a local, high resolution circulation model. Hence, stochastic formulations were developed for generating open coast boundary conditions. The use of a large-scale model for generating coastal boundary conditions allows for the spatial variance of storm surge hydrographs to be incorporated into the local circulation model. Two different results are presented in this section: 1) hydrographs along both the 1 km radius and 5 km radius arcs; and 2) hydrographs for point 3 (perpendicular to the coast) on

each arc ranging from 1 km to 5 km (model results at the other output locations show similar patterns, but are not included in the interest of brevity). Water surface elevations are compared at the same open coast boundary locations as previously described for the average inlet width. Analysis in the previous section indicates that all inlet profiles produce nearly identical results as the average inlet width; therefore, the results and conclusions presented in this section can be extended to other inlet profiles as well. Fig. 12 shows computed sea stages along the 1 km radius observation arc for the average inlet profile. The highest peak (recorded at point 5) is 2.16 m above mean sea level (MSL), whereas the lowest (recorded at point 3) is 2.04 m above MSL, a difference of 0.12 m. The peaks reveal that the points (4 and 5) closest to the eye of the storm produce the greatest surge. Overall, the behavior of the storm surge hydrographs shows that a spatial variance exists along this observation arc during the surge event. However, the hydrographs also show that the astronomic tide signals are identical prior to hurricane landfall (i.e. before day 2.75 of the simulation), thus indicating that the spatial variance is limited to the surge event. Similarly, Fig. 13 presents computed water elevations along the 5 km radius observation arc. The highest peak (recorded at point 5) is 2.18 m above MSL, whereas the lowest (recorded at

Fig. 14. Storm surge hydrographs at point #3 on radii extending from 1 km to 5 km.

Fig. 16. Astronomic tide comparison along the 5 km radius observation arc extending seaward from the mouth of Pensacola Bay.

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

387

Fig. 17. Hurricane Ivan water elevations at the NOS station in Pensacola Bay for the domain that treats the barrier islands as model boundaries.

Fig. 19. Comparison between model results for the domain that treats the barrier islands as inundation areas versus treating the barrier islands as model boundaries.

point 3) is 1.66 m above MSL, a difference of 0.52 m. Compared to the 1 km radius, the difference between the highest and the lowest peak values is much greater for the 5 km radius (0.52 m for the 5 km radius compared to 0.12 m for the 1 km radius). Furthermore, a difference in peak elevations is also recognized between points that have the same depth. Comparing locations 2 and 4 (depth of 6.2 m at each location) in Fig. 13 indicates a peak surge of 1.73 m at point 2 and 1.81 m at point 4 (a difference of 8 cm). A similar difference (10 cm) is also recognized between points 1 and 5 (at a depth of 3.7 m). Therefore, the variability in the peak storm surge elevations is, in part, attributed to the horizontal location along the open coast boundary. Moreover, differences in peak surge levels are also recognized at the same point on different observation arcs. Results show that the peak storm surge increases as the surge wave propagates into shallower waters. Fig. 14 displays recorded water elevations at point 3 (perpendicular to the coast) on the 1 km and 5 km radius observation arcs. The highest peak water elevation is 2.05 m at the 1 km radius arc, and the lowest peak

elevation is 1.66 m at the 5 km radius arc. The results clearly indicate that the storm surge is dependent on the depth of the location. The results presented thus far potentially have significant implications for local bridge scour modeling. Previous research (Edge et al., 1999; Zevenbergen et al., 1999; Sheppard and Pritsivelis, 1999) has applied a single design hydrograph at the open coast boundary for bridge scour studies (distance of the ocean boundaries from the inlet was not specified in each study, but they appeared to be within a few kilometers of the coast). Results presented herein show that the hydrographs vary spatially over this open coast boundary, thus indicating that individual hydrographs should be applied at each open boundary node for proper bridge scour modeling. 3.2. Hurricane Ivan storm surge results A Hurricane Ivan storm surge study is undertaken to determine if conclusions from the numerical parameter study can be extrapolated to a real-life scenario. In contrast to the numerical

Fig. 18. Maximum storm surge elevations within Pensacola Bay and Escambia Bay for the domain that treats the barrier islands as model boundaries.

Fig. 20. Maximum storm surge elevations within Pensacola Bay and Escambia Bay for the domain that treats the barrier islands as inundation areas.

388

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

hydrographs in the following sections is a result of the storm surge phenomenon rather than the astronomic tides. 3.2.2. Historical comparison Both model domains (barrier islands as model boundaries and barrier islands as inundation areas) are used to simulate storm surge within Escambia Bay for a ten day period: September 7, 2004 at 6 p.m. through September 17, 2004 at 6 p.m. Fig. 17 compares model results to historical gauge data at the NOS station for the domain that treats the barrier islands as model boundaries. Two noticeable results are obtained from this figure. First, the NOS tide gauge appears to have stopped recording close to the time of hurricane landfall. As a result, the peak surge level and the set-down behavior at this location are not shown in the historical data, making a comparison throughout the entire hurricane event impossible. In lieu of this, it is difficult to accurately compare model results to historical data; however, high water marks in Escambia Bay suggest a peak water level of 34.5 m near the Interstate 10 Bridge in Escambia Bay (Douglass et al., 2004). A plot of the maximum surge elevations within the bay (Fig. 18) for the model results that treat the barrier islands as model boundaries indicates that the model produces peak surge levels that are slightly less (2.53 m) than the high water marks near the bridge. Second, a difference in the rising limb behavior is noticed between the model results and the historical data (between 9/14 and 9/15). This difference can be attributed to several factors that are not included in this study. First, short-wave (wind waves) processes are not included in the modeling process. Typically, wind-induced waves travel ahead of the storm surge, creating water elevations to rise before the hurricane makes landfall. In order to perform a true storm surge hindcast (which is not the objective of this study), wind waves would have to be included in the modeling process. Additionally, the difference may also be attributed to the model parameter (bottom friction coefficient, wind drag coefficient, and horizontal eddy viscosity) values used in this study. However, the model parameter values used are considered standard values and a sensitivity analysis is outside the

Fig. 21. Storm surge hydrographs at point #4 along the 5 km radius observation arc for the model domain that does not include the Pensacola Pass and the model domain that does include the Pensacola Pass.

parameter study, historical water elevation data is available to compare model results as a means of verification. First, an astronomic tide verification and a storm surge study of Hurricane Ivan are performed to validate the model domain in the coastal region of Pensacola Pass and Escambia Bay. Next, the effect of the Pensacola PassEscambia Bay system on the open coast storm surge hydrographs and the spatial variance of the storm surge hydrographs along the open coast are evaluated. The width (980 m) of Pensacola Pass and the size (6.8 108 m2) of the adjoining bay indicate that the domain is similar to the inletbay configuration with the 1000 m inlet width used in the numerical parameter study. 3.2.1. Astronomic tide verification First the model domain is verified through an astronomic tide comparison. A National Ocean Service tide gauge located within Pensacola Bay (see Fig. 4) provides historical tidal constituent data. A 14-day resynthesis of model results is compared to a resynthesis of historical constituents at the NOS tide gauge in Pensacola Bay (Fig. 15). The 14-day resynthesis period is chosen to allow for a complete springneap tide cycle. The results indicate that the model performs reasonably well at this location. A slight discrepancy is recognized in phasing between days 8 and 14, as well as the troughs through most of the resynthesis. Overall, however, the model domain faithfully simulates the tidal dynamics within the Pensacola PassEscambia Bay system. In addition, the astronomic tide variance is examined along the open coast observation arcs extending seaward from the mouth of Pensacola Pass. Similar to the tide verification at the NOS station in Pensacola Bay, a 14-day resynthesis is performed at each of the model output locations along the observation arcs. Fig. 16 shows a resynthesis of the astronomic tide signal along the 5 km radius observation arc. The results indicate that the tide signal is nearly identical at each of these locations. Similar results are obtained along the other semi-circular observation arcs, indicating that the astronomic tides do not display a spatial variance near the inlet. As a result, the spatial variance recognized amongst the storm surge

Fig. 22. Storm surge hydrographs along the 1 km radius observation arc extending seaward from Pensacola Pass.

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

389

Fig. 23. Storm surge hydrographs along the 5 km radius observation arc extending seaward from Pensacola Pass.

realm of this study. Lastly, inaccuracies in the wind field and pressure field may have also led to errors in the storm surge results. Furthermore, when the barrier islands are treated as inundation areas, the surge levels within the bay increase by almost half a meter compared to treating the barrier islands as model boundaries. Fig. 19 compares the storm surge hydrographs between the two model domains (barrier islands as model boundaries and barrier islands as inundation areas) to historical data at the NOS tide gauge. The results clearly indicate that allowing the storm surge to flood over the barrier islands creates greater water elevations within the bay. In this case, the highest surge value for treating the barrier islands as inundation areas is 2.12 m, whereas the highest surge value for treating the barrier islands as model boundaries is 1.66 m, a difference of 0.46 m. Similarly, comparing the maximum storm surge contours within Escambia Bay to high water marks (34.5 m) recorded near the Interstate 10 Bridge indicates that including the barrier islands as inundation areas creates maximum surge levels that are more consistent with historical records (Fig. 20). The results shown in this figure (34 m near the bridge) compare more favorably with the high water marks than the results shown in Fig. 18 (barrier islands treated as model boundaries). Thus, including the barrier islands as inundation areas creates more realistic surge levels within the bay than treating them as model boundaries. Model results still indicate a discrepancy when compared to the historical data during the rising limb period. It's important to note, however, that this is not a true storm surge hindcast study for two reasons: 1) windwave processes are not included in the modeling process; and 2) standard, not necessarily optimal, values for model parameters (i.e., bottom friction coefficient, wind drag coefficient, and horizontal eddy viscosity) were used. In order to identify optimal model parameter values a sensitivity study would be required, but that is outside the realm of this study. The important conclusion from the results presented thus far is the model reproduces the overall long-wave behavior reasonably well, which is paramount to the open coast storm surge hydrograph analysis presented in the following sections.

3.2.3. Pensacola Pass The main conclusion from the numerical parameter study is that the effect of tidal inlets on open coast storm surge hydrographs is negligible. The implication of this conclusion is that a large-scale ocean circulation model does not need to include an inletbay system in the computational domain when generating open coast boundary conditions for a local bridge scour model. This section verifies this conclusion for the case of Pensacola Pass during Hurricane Ivan. Two computational domains are used in this analysis: 1) a finite element mesh of the Western North Atlantic Tidal model domain that does not include the Pensacola PassEscambia Bay system; and 2) a finite element mesh of the same region that does include the Pensacola PassEscambia Bay system (previously described). The domain that does not include the Pensacola Pass Escambia Bay system has been optimized in previous research for storm surge applications (Hagen et al., in press; Kojima, 2005), and is representative of a model domain that would be used to generate open coast boundary conditions for a local bridge scour model if the inletbay system is not included in the domain. In contrast, the domain that does include the Pensacola Pass Escambia Bay system is typical of a model domain that would be used to generate the open coast boundary conditions if the inlet bay system is included in the domain. Model output is generated along the semi-circular observation arcs extending seaward from the mouth of the inlet. Fig. 21 shows sea stages at point 4 (due south) on the 5 km radius observation arc. Model results for the domain that does not include the Pensacola Pass indicate a peak surge level of 1.79 m; whereas model results for the domain that does include the Pensacola Pass indicate a peak surge of 1.71 m (a difference of 8 cm). Similar results are obtained at the other open coast observation locations. At this point, it is impossible to state that this difference would or would not have an appreciable effect on the performance of a bridge scour model. However, clearly it is unlikely that a difference of 8 cm (4.5% relative to the peak value) would significantly alter the model's performance, especially since the rising and falling limbs of the hydrographs are nearly identical. These results are consistent with the conclusions from the numerical parameter

Fig. 24. Storm surge hydrographs at the same location (point 4, due south) on each observation arc from 1 km to 5 km in radius.

390

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

study, i.e. tidal inlets do not appear to have a significant influence on the open coast storm surge hydrographs. 3.2.4. Spatial variance This section verifies the spatial variance results shown in the numerical parameter study with a real-life scenario (e.g. Pensacola Pass). The 1 km radius and 5 km radius observation arcs are used to verify these findings. Fig. 22 displays the storm surge hydrographs along the 1 km radius extending seaward from the mouth of Pensacola Pass. The plots indicate that a minimal difference exists between the hydrographs along this observation arc. The highest peak surge value is 1.87 m at point 7 (westsouthwest), whereas the lowest peak value is 1.83 m at point 3 (southsoutheast), a difference of only 4 cm. Fig. 23 shows the water elevation recordings along the 5 km arc radius. In contrast to the 1 km radius, a noticeable spatial variance is evident in the plots. The greatest peak surge value is 1.94 m (point 7) and the lowest peak surge value is 1.65 m (point 3), a difference of 0.29 m. The spatial variance along the 5 km radius is much greater than the spatial variance along the 1 km radius, which is consistent with the results from the numerical parameter study. In addition, spatial variance is also recognized as the radius of the observation arc is increased. Fig. 24 shows hydrographs at the same location (point 4, due south) along the 1 km and 5 km observation arcs. Similar to the numerical parameter study results, the peak storm surge increases nearer to the coastline. 4. Conclusions Presented herein is a numerical parameter study focusing on the effect that tidal inlets have on open coast storm surge hydrographs. Four different inletbay configurations are constructed based on a statistical analysis of existing Florida tidal inlets. The inletbay configurations allow for a number of test cases to elucidate the influence that tidal inlets have on open coast boundary conditions. A secondary focus of this study is the spatial variance of the storm surge hydrographs along nearinlet boundary locations. Results from the numerical parameter study are compared with results from a Hurricane Ivan storm surge study in the vicinity of Escambia Bay. The conclusions from this study have implications toward future bridge scour modeling efforts. 4.1. Numerical parameter study conclusions Two primary conclusions are drawn from the numerical parameter study: 1) the effect of tidal inlets on open coast storm surge hydrographs is minimal; and 2) a noticeable spatial variance of storm surge hydrographs exists along the open coast boundary locations. The effect of including inletbay systems on the open coast storm surge hydrographs is negligible. Comparing model output for all four inletbay configurations to the control mesh indicates that including an inletbay system in the model domain does not significantly alter the open coast hydrographs. Taking this into account, a large-scale model domain (e.g. the WNAT

model domain) would not need to include the inletbay system in the computational domain to generate open coast boundary conditions for a local inlet-based model (a significant timesaving benefit for coastal modelers). Another significant finding in this study is the spatial variance of the storm surge hydrographs along the open coast boundary locations. Nearer to the inlet, the spatial variance of the hydrographs is minimal; however, the spatial variance increases further from the inlet (i.e. going from the 1 km radius to the 15 km radius). This suggests that it is more appropriate to apply spatially varying hydrographs along the open coast boundary, instead of the single design hydrograph method used in previous bridge scour modeling studies (as a point of reference, the open coast boundary in these studies appeared to be several kilometers offshore). If the open ocean boundary is placed close to the inlet so that a single hydrograph can be used at each node, then momentum flux problems may arise from the boundary being placed near the area of interest. By placing the boundary further from the inlet and applying individual hydrographs at each node, the cross-basin hydrodynamics of storm surge propagation can be artificially incorporated into the inlet-based model. The results from this study imply that current bridge scour modeling practices can be improved upon by incorporating spatially varying hydrographs along the open coast boundary. 4.2. Hurricane Ivan conclusions The primary focus of the Hurricane Ivan storm surge study is the verification of the numerical parameter study conclusions in a real-life scenario. The effect that the Pensacola PassEscambia Bay system has on the open coast storm surge hydrographs was examined. The results are consistent with the conclusions from the numerical parameter study, such that including the Pensacola PassEscambia Bay system in the model domain does not have a significant impact on the open coast storm surge hydrographs. Furthermore, a spatial variance is also recognized along observation arcs extending seaward from the mouth of Pensacola Pass. Similar to the behavior recognized in the numerical parameter study, the magnitude of the spatial variance increases as the open coast boundary locations are extended farther from the mouth of the inlet. Thus, it is apparent that applying a single design hydrograph along the open coast boundary is inappropriate and may lead to erroneous results. A corollary to this study is the significance that the low-lying barrier islands had on the storm surge within Escambia Bay during Hurricane Ivan. In this case, the storm surge increased by half a meter in the bay when the barrier islands were treated as inundation areas compared to treating the barrier islands as model boundaries. This is an important conclusion for future storm surge studies in areas that have barrier islands prevalent in the coastal system. Acknowledgments This study was funded in part by Award UFEIES0404029UCF as a subcontract from the Florida Department

M.B. Salisbury, S.C. Hagen / Coastal Engineering 54 (2007) 377391

391

of Transportation (FDOT) via the University of Florida (UF); Award N00014021-0150 from the National Oceanic and Atmospheric Administration (NOAA), U.S. Department of Commerce; and by National Oceanographic Partnership Program (NOPP) Award No. N00014021-0150 administered by the Office of Naval Research (ONR). The views expressed herein are those of the authors and do not necessarily reflect those of the FDOT, UF, NOAA, Department of Commerce, ONR, or NOPP and its affiliates. The authors also wish to thank Vince Cardone and Andrew Cox of Oceanweather Inc. for providing the wind field used in the Hurricane Ivan study, and Don Slinn at the University of Florida for providing the synthetic wind field code used in the numerical parameter study. References
Blain, C.A., Westerink, J.J., Luettich, R.A., Scheffner, N.W., 1994. ADCIRC: an advanced three-dimensional circulation model for shelves, coasts, and estuaries report no. 4. Hurricane storm surge modeling using large domains. Technical Report DRP-926, August 1994, U.S. Army Engineer Waterways Experiment Station, Vicksburg, Mississippi. Carr de Betts, Erica Eva, 1999. An examination of flood deltas at Florida's tidal inlets. Master's thesis, Coastal and Oceanographic Engineering Department, University of Florida, Gainesville, Florida. Cialone, M.A., Butler, L., Amein, M., 1993. DYNLET1 application to Federal Highway Administration projects. Technical Report CERC-936, U.S. Army Corps of Engineers. CERC, Vicksburg, Mississippi. 93 pages. Cox, A.T., Cardone, V.J., 2000. Operational systems for the prediction of tropical cyclone generated winds and waves. Proceeding of 6th International Workshop on Wave Hindcasting and Forecasting. Monterey, California. Cox, A.T., Cardone, V.J., 2002. 20 years of operational forecasting at Oceanweather. Proceeding of 7th International Workshop on Wave Hindcasting and Forecasting. Banff, Alberta, Canada. Cox, A.T., Greenwood, J.A., Cardone, V.J., Swail, V.R., 1995. An interactive objective kinematic analysis system. Proceeding of 4th International Workshop on Wave Hindcasting and Forecasting. Banff, Alberta, Canada. Douglass, S.L., Hughes, S.A., Rogers, S., Chen, Q., 2004. The impact of Hurricane Ivan on the coastal roads of Florida and Alabama: a preliminary report. Coastal Transportation Engineering Research & Education Center, University of South Alabama. Edge, B.L., Vignet, S.N., Fisher, J.S., 1999. Determination of bridge scour velocity in an estuary. In: Richardson, E.V., Lagasse, P.F. (Eds.), Stream Stability and Scour at Highway Bridges, Compendium of Papers ASCE Water Resources Engineering Conferences 19911998. ASCE, Reston, Virginia, pp. 738747. Garrett, J.R., 1977. Review of the drag coefficients over oceans and continents. Monthly Weather Review 105, 915929. Hagen, S.C., Zundel, A.K., Kojima, S., in press. Automatic, unstructured mesh generation for tidal calculations in a large domain. International Journal for Computational Fluid Dynamics. Hench, J.L., Blanton, B.O., Luettich, R.A., 2002. Lateral dynamic analysis and classification of barotropic tidal inlets. Continental Shelf Research 22, 26152631. Jarrett, J.T., 1976. Tidal prisminlet area relationships. General Investigation of Tidal Inlets Report 3, U.S. Army Engineer Waterways Experiment Station, Vicksburg, Mississippi.

Kojima, S., 2005. Optimization of an unstructured finite element mesh for tide and storm surge modeling applications in the Western North Atlantic Ocean. Master's thesis, Department of Civil and Environmental Engineering, University of Central Florida, Orlando, Florida. Kubatko, E.J., Westerink, J.J., Dawson, C.N., 2005. An unstructured grid morphodynamic model with a discontinuous Galerkin method for bed evolution. The 3rd International Workshop on Unstructured Grid Numerical Modelling of Coastal Shelf and Ocean Flows, Toulouse, France. Luettich, R.A., Westerink, J.J., Scheffner, N.W., 1992. ADCIRC: an advanced three-dimensional circulation model for shelves, coasts, and estuaries report no. 1. Theory and Methodology of ADCIRC-2DDI and ADCIRC-3DL, Technical Report DRP-926, November 1992, U.S. Army Engineer Waterways Experiment Station, Vicksburg, Mississippi. Mendenhall, W., Sincich, T., 1995. Statistics for Engineering and the Sciences, fourth ed. Prentice Hall, New Jersey. Militello, A., Kraus, N.C., Brown, M.E., 2000. Regional coastal and inlet circulation modeling with application to Long Island, New York. Proceedings 13th Annual National Beach Preservation Technology Conference, Florida Shore and Beach Preservation Association, Tallahassee, Florida, pp. 191201. Pandoe, W.W., Edge, B.L., 2004. Cohesive sediment transport in the 3Dhydrodynamicbaroclinic circulation model, study case for idealized tidal inlet. Ocean Engineering 31, 22272252. Powell, M.D., Houston, S.H., Amat, L.R., Morisseau-Leroy, N., 1998. The HRD real-time hurricane wind analysis system. Journal of Wind Engineering and Industrial Aerodynamics 77 & 78, 5364. Richardson, J.R., Richardson, E.V., Edge, B.L., 1999. Bridge scour analysis in tidal waterways. In: Richardson, E.V., Lagasse, P.F. (Eds.), Stream Stability and Scour at Highway Bridges, Compendium of Papers ASCE Water Resources Engineering Conferences 19911998. ASCE, Reston, Virginia, pp. 748766. Sheppard, D.M., Miller, W., 2003. Design storm surge hydrographs for the Florida coast. Florida Department of Transportation, Final Report, FDOT Contract Number BC-354 RWPO 70, Tallahassee, Florida. 126 pages. Sheppard, D.M., Pritsivelis, A., 1999. Sensitivity of currents and water elevations in tidal waters to storm surge parameters. In: Richardson, E.V., Lagasse, P.F. (Eds.), Stream Stability and Scour at Highway Bridges, Compendium of Papers ASCE Water Resources Engineering Conferences 19911998. ASCE, Reston, Virginia, pp. 774779. Stewart, Stacy R., 2004. Hurricane Ivan: September 224, 2004. Tropical Cyclone Report, National Oceanic and Atmospheric Administration, National Weather Service, Tropical Prediction Center, Miami, Florida. Thompson, E.F., Cardone, V.J., 1996. Practical modeling of hurricane surface wind fields. ASCE Journal of Waterway, Port, Coastal, and Ocean Engineering 122 (4), 195205. United States Army Corps of Engineers, 2002. Hydrodynamics of tidal inlets. Coastal Engineering Manual, Part 2, Chapter 6. Coastal and Hydraulics Laboratory, Vicksburg, Mississippi. Westerink, J.J., Blain, C.A., Luettich, R.A., Scheffner, N.W., 1994. ADCIRC: an advanced three-dimensional circulation model for shelves, coasts, and estuaries report no. 2. User's Manual for ADCIRC-2DDI, Technical Report DRP-926, January 1994, U.S. Army Engineer Waterways Experiment Station, Vicksburg, Mississippi. Zevenbergen, L.W., Richardson, E.V., Edge, B.L., 1999. Computer models for tidal hydraulic analysis at highway structures. In: Richardson, E.V., Lagasse, P.F. (Eds.), Stream Stability and Scour at Highway Bridges, Compendium of Papers ASCE Water Resources Engineering Conferences 19911998. ASCE, Reston, Virginia, pp. 807814.