ASSESSMENT OF INUNDATION RISK FROM SEA LEVEL RISE AND STORM SURGE IN COASTAL NATIONAL PARKS

In coastal ecosystems, sea level rise and an increase in storm frequency a nd intensity are two major impacts expected to result from climate change. C oastal National Parks have many low-lying areas that are at risk from inundati o resulting from these impacts. In order to help park managers meet their goal of preserving valuable resources, I developed a methodology to evaluate risk of inundation from sea level rise and storm surge at sentinel sites, areas of importance for natura l, cultural and infrastructural resources. I performed a literature review on the factors driving sea level rise in the Northeast, and conducted an evaluation of the methods used by scientists and engineers to model sea level rise and storm surge inundation. I selected the mos recent and appropriate geospatial tools, models and datasets to perform a coast al inundation risk assessment in three northeastern coastal National Parks—Boston Harbor Islands National Recreation Area, Cape Cod National Seashore, and Assateague Island National Seashore. I collected elevation data at sentinel sites using real time kinematic g lobal positioning system (RTK GPS) technology and assessed the accuracy of the most recent, readily-available Light Detection and Ranging (LiDAR) deri ved Digital Elevation Models. Because of the poor quality of existing LiDAR data, Bost on Harbor Islands National Recreation Area was excluded from the final assessment . I evaluated risk of inundation at sentinel sites in Cape Cod and Assateague Island using three modeling approaches: bath-tub modeling, Sea Level Affecting Marshes Model (SLAMM), and Sea, Land and Overland Surges from Hurricanes (SLOSH) Model, and developed an overall inundation index, a single measure of inundation likelihood that incorporated output from each modeling approach. I created inundation maps for a range of sea level rise and storm surge scenarios, calculated the pr obability of inundation at each sentinel site given the uncertainty associated with each mode l and dataset, and ranked the relative risk of sentinel sites to inform management and adaptation strategies. Cape Cod’s sentinel sites, which in many cases occ urred in high elevation settings, were found to be less vulnerable to inundation than Assateague Island’s sentinel sites which were distributed in low-lying areas alon g the barrier beach island. This inundation risk assessment methodology can be applied to other coastal parks and to the same coastal parks at different times as more acc urate elevation datasets and updated sea level rise projections become available .


INTRODUCTION
In coastal ecosystems, accelerated sea level rise and an increase in storm frequency and intensity are two major impacts expected to result from climate change (Ashton, Donnelly, and Evans, 2008;Bender et al., 2010;Harvey and Nicholls, 2008).
In the next century, the rate of global sea level rise is anticipated to be several times higher than measured over the past century Overpeck and Weiss, 2009;Pfeffer, Harper, and O'Neel, 2008;Rahmstorf, 2007). The US Northeast coast experiences a rate of relative sea level rise greater than the global average due to substantial regional variations in glacial isostatic adjustment effects and oceanographic processes (Tamisiea and Mitrovica, 2011).
Along the US Atlantic coast, the highest rates of subsidence occur from southern Massachusetts to Virginia  and predicted changes in ocean circulation driven by climate change could potentially add meters of dynamic sea level rise near the Northeast coast ). The frequency and extent of severe coastal storms is expected to increase (Bender et al., 2010), and large surge levels may cause significant damage to coastal infrastructure and alteration of ecosystems (Irish et al., 2010;Kirshen et al., 2008;Lin et al., 2010;McInnes et al., 2003). A discussion of factors driving sea level rise in the region is provided in Appendix 1.
This investigation will assess inundation risk from sea level rise and storm surge at sentinel sites in three coastal, northeastern United States National Parks-Boston Harbor Islands National Recreation Area, Cape Cod National Seashore, and Assateague Island National Seashore. Sentinel sites are locations of natural or cultural resources of special importance to the National Park Service. The term "at risk" is used to indicate that a sentinel site is predicted to be inundated as a result of sea level rise or storm surge. Some sites that are predicted to be at "at risk" are not necessarily threatened or impacted; natural features may be altered as a result of rising water levels yet persist because of their resilience to change. However, habitats, cultural, or infrastructural resources may be severely impacted if they are inundated. Thus, "risk" does not imply "impact" in my study. In either case, an understanding of the relative vulnerability of each sentinel site to inundation will allow park resource managers to develop management and mitigation strategies for these sites.  (Scawthorn et al., 2006a;Scawthorn et al., 2006b). Other GISbased methods have been applied as well (Brown, 2006;Hennecke and Cowell, 2000).
A review of sea level and storm surge inundation models is provided in Appendix 2.
Coarse-scale assessments for sea level rise and storm surge risk have been previously conducted. The Coastal Vulnerability Index (CVI) technique was applied at Cape Cod National Seashore (Hammar-Klose et al., 2003) and at Assateague Island National Seashore (Pendleton, Williams, and Thieler, 2004). The method combines a number of physical variables in order to classify the relative risk of 1.5 km shoreline segments to sea level rise impacts. Gutierrez, Williams, and Thieler (2007) studied potential shoreline changes from sea level rise along the U.S. Mid-Atlantic Coast. The CVI and shoreline change assessments were designed to provide a regional overview of coastal vulnerability and do not have the spatial resolution for site-specific risk assessment.
All vulnerability models and methods rely on elevation data, which are often highly limited in their vertical accuracy and cause large ranges of uncertainty in results

Study Areas
I conducted analyses at three coastal parks ( Figure 1

Data Sources
The study incorporates the most recent, readily-available elevation data and widely-used inundation mapping tools and techniques. Elevation measurements for inundation models were acquired from LiDAR data obtained from aircraft-mounted laser sensors that emit pulses of light energy at the ground and measure the distance based on the time required for the pulses to reflect back to the sensor. LiDAR data are typically accurate to 0.15-1 m

Sea Level Rise Scenarios
Three scenarios were selected to represent the current range of sea level rise predictions for the year 2100: 0.6 m (IPCC, 2007), 1 m , and 2 m (Pfeffer, Harper, and O'Neel, 2008). They were modeled using two methods. The first approach-the bath-tub model-involved creating a planar water surface that represents the sea level rise scenario added to the Mean Higher High Water (MHHW) tidal elevation. Modeling sea level rise or storm surge in addition to the MHHW level represents the worst case inundation scenario. These water surface elevations were calculated using VDatum software (NOAA, 2011), which performs elevation conversions between NAVD88 (an orthometric datum) and tidal datums.
The land surface elevation and modeled water surface were compared and probabilities of inundation at sentinel sites were calculated using the z-score inundation uncertainty technique described by NOAA Coastal Services Center (2010a). Standard scores, or z-scores, were calculated at each sentinel site using the formula: Where the total RMSE (root mean square error) is calculated as: RMSE for LiDAR DEMs is calculated as: Where X LiDAR is the elevation from LiDAR at a single location and X GPS is the elevation as determined by GPS in the same location. RMSEs for the GPS survey elevations were reported by Trimble Software. RMSEs for water surfaces were reported by VDatum. The standard normal cumulative distribution function was used to calculate probabilities of inundation and the certainty of the prediction given errors associated with the data and models (Ott and Longnecker, 2010).  (NOAA, 2006) and National Wetlands Inventory data (USFWS, 2010) and recoding them to SLAMM categories as specified by the User Manual (Warren Pinnacle Consulting, 2010). Some accretion parameters were obtained from Surface Elevation Table (SET) data (Lynch, 2012;NPS, 2009). When parameters were unknown or unavailable, SLAMM's default settings were used. Using linear relationships and decision tree rules, SLAMM calculates water elevation at a particular location, and computes inundation and habitat response over large areas (Mcleod et al., 2010). The output maps showed expected habitat classes and areas of inundation based on the different rates and magnitudes of sea level rise. Sentinel site locations were mapped over the output. Change matrices were created to show changes from initial habitat to the predicted habitat for each sentinel site.

Storm Surge Scenarios
The SLOSH (Sea, Land and Overland Surges from Hurricanes) model, a forecast model for hurricane-induced water levels for the Gulf and Atlantic Coasts (Jelesnianski, Chen, and Shaffer, 1992) was used to model expected surge heights along park coasts from Saffir-Simpson Category 1-4 hurricanes. Surge heights are not uniform along the coastline and depend on the hurricane track, wind speed, and topography and bathymetry at the point where the storm makes landfall (FEMA, 2003). Storm surge heights were derived from the Providence/Boston and Ocean City storm basins in SLOSH and used as input in the Applied Science Associates, Inc.
(ASA) Inundation Toolbox -Interpolation tool (Isaji and Knee, 2009). The tool interpolated the point heights to a raster surface of the same extent and resolution as the DEM. Elevations from the DEM were compared to the elevations from the storm surge surfaces and probabilities of inundation at sentinel sites were calculated using the z-score uncertainty technique described above. The uncertainty technique incorporated known sources of error unique to the water surface modeled by SLOSH.
For each of the four storm scenarios, two probabilities were calculated: one given the elevation of a sentinel site from the DEM and another given the elevation of a sentinel site from the RTK GPS survey.
ArcGIS 10 software was used for all geospatial data processing (ESRI, 2011).
A summary of the models used and sources of data are provided in Table 1.

Statistical Procedures
At each sentinel site, probabilities of inundation were calculated for three sea level rise scenarios (0.6 m, 1 m, 2 m), four storm surge scenarios (Category 1-4) and two sources of elevation data. Descriptive statistics were calculated for each risk estimate variable and the variables were tested for normality using the Shapiro-Wilk normality test. Because data were usually non-normally distributed, I performed pairwise comparisons using the non-parametric Wilcoxon signed rank test, and where there were 3 or more groups in a comparison, I used the Kruskal-Wallis rank sum test.
When data were normally distributed, I used the paired t test for pairwise comparisons.
A principal components analysis (PCA) was used to reduce the large number of risk measures to a smaller number of variables in order to develop a composite measure of inundation risk at sentinel sites. All analyses were conducted using the statistical software package R (R Development Core Team, 2011).

Elevations of Sentinel Sites
Sentinel sites are locations of natural, cultural and infrastructural resources of special importance to the National Park Service and were provided by park managers.

Quality of the LiDAR Data
The metadata for each of the three LiDAR-derived DEMs reported a 0.15 m vertical root mean square error (RMSE). To validate the accuracy estimates, I calculated the vertical RMSE using ground control points of high quality and accuracy (< 2 cm vertical and horizontal accuracy) collected using survey-grade GPS (Table 3). Using these 20 additional control points, the vertical RMSE was found to be 1.53 m.
This result meant that the BOHA LiDAR could not be used for modeling sea level rise and storm surge on the scales proposed (see discussion of elevation in Appendix 1).
For elevation data of this quality, NOAA (2010)  Category 1-4 storms were modeled at CACO and ASIS (Table 4). The bath-tub and storm surge models were used to map areas at risk from inundation. An area was considered at risk from inundation if it had an elevation less than or equal to the water surface elevation that was expected in any given location (

Probabilities of Inundation
Each sentinel site was intersected onto the scenario's modeled water surface and probabilities of inundation were calculated using equations 1 -3. Probabilities were determined using the RTK GPS and LiDAR elevations. Mean probabilities from RTK GPS elevations are reported (Table 6 and Table 7) due to their higher accuracy.
The complete list of sentinel sites and probabilities of inundation can be found in Appendix 4 (Table A4.1).
The mean probabilities of inundation at CACO were significantly different under the two sea level rise scenarios (Wilcoxon signed rank V=0, p < 0.001). At ASIS, the mean probabilities of inundation were significantly different under the three sea level rise scenarios (H=66.02, df=2, p < 0.0001). The mean probabilities of inundation for four storm surge scenarios were significantly different in both CACO (H=44.51, df=3, p < 0.0001) and ASIS (H=96.52, df=3, p < 0.0001).

Habitat Changes Predicted by SLAMM
The This may not be an appropriate assumption for some National Park study sites. A few sentinel sites in the "Undeveloped Dry Land" category experienced conversions to "Transitional Marsh," "Estuarine Beach," and in one case, to "Open Ocean" after 2 m of sea level rise (Table 8). In ASIS, many sentinel sites were in the "Undeveloped Dry Land" class (24 out of 34) and experienced conversions to "Transitional Marsh," "Estuarine Beach," "Ocean Beach," and "Open Ocean". Points that started out in the "Irregularly Flooded Marsh" converted to "Salt Marsh" after 2 m of sea level rise (Table 9). According to the recommendation of Scarborough (2009), similar classes were aggregated for ease of interpretability (Table 10 and Table 11).
Similarly, mapping the SLAMM output with the aggregated classes aided in interpretability. Figure

Overall Inundation Index
The three modeling methods yielded several measures of inundation risk. I  (Table 12). Nearly all variables had similar loading values on PC1, thus PC1 represents a "size" effect (August, 1983) and is an excellent index of overall inundation likelihood. The other principal components had high loadings on only one or two variables and reflected specific risk factors. Furthermore, they did not predict a large amount of overall variation, thus are not candidates for an overall risk index. The range of PC1 values for each park was separated into five quintiles. Large positive PC1 scores for sentinel sites indicate that inundation is very unlikely; large negative PC1 scores indicate that inundation is very likely (Table 13). Sentinel sites' raw PC1 scores are given in Table A4.1. Overall inundation index classification at sentinel sites is mapped for CACO ( Figure 8) and ASIS ( Figure 9).

Quality of Elevation Data
The were calculated with greater certainty. The application of RTK GPS is a promising solution to elevation uncertainty issues in sea level rise inundation risk assessments. It is important to note, however, that RTK GPS protocols require operating a GPS base station at a location that has been surveyed to within a few millimeters. Thus, a network of accurate geodetic control sites within 5 km of potential sentinel sites is an essential requirement for RTK GPS measurement (Murdukhayeva et al., 2012).

Inundation Models
Bath-tub modeling is a technique that tends to overestimate inundation extents and calculate uncertain predictions (Mcleod et al., 2010;Poulter and Halpin, 2008).
The use of high accuracy RTK GPS equipment helped minimize the error in estimating inundation risk. Because the error associated with each sentinel site elevation was low, there was less uncertainty associated with each modeled scenario, i.e. many of the sentinel sites had probabilities of inundation of either 0 (very unlikely) or 1 (very likely). This was not the case with probabilities of inundation calculated using elevations at sentinel sites based on LiDAR data, where many probabilities of inundation were included in the range of 0.25 to 0.75. Therefore, the LiDAR-derived DEMs were used only for mapping areas at risk and providing a map assessment of the extent of inundation (Figures 2-5 and A5.1-A5.13). Inundation probabilities at sentinel sites using LiDAR-derived elevations were not used for developing the overall inundation index.
One of the most valuable products of this assessment is the range of storm surge heights modeled by SLOSH (Table 4) Thus, it is important that the National Park Service evaluate risk for each site and develop mitigation plans accordingly.

SLAMM
The SLAMM model was limited by the quality of the data driving the model (LiDAR-derived DEMs). Using the RTK GPS elevations to enhance the results of the model was not possible. An important baseline dataset for the SLAMM model was initial land use classification. In this study, initial land use conditions were obtained from National Wetlands Inventory maps created using aerial photography from 1988 (ASIS) and 1993 (CACO), and as a result these maps did not reflect changes that might have occurred over the past two decades. Furthermore, there is no way to quantitatively determine the uncertainty associated with a SLAMM prediction. These factors and others limit the output results (Kirwan and Guntenspergen, 2009;Scarborough, 2009).
For the risk assessment, I was most interested in identifying predicted land cover changes to open water, i.e. inundation. At CACO, one site (1 m scenario) and three sites (2 m scenario) are expected to experience inundation. At ASIS, four sites (1 m scenario) and 12 sites (2 m scenario) are expected to be inundated. The bath-tub model predicts greater amounts of inundation (Table 6). At CACO, three (1 m scenario) and 11 sites (2 m scenario) have probabilities of inundation greater than 75%, and at ASIS, 11 (1 m scenario) and 32 (2 m scenario) sites have probabilities of inundation greater than 75%. This difference confirms the notion that bath-tub models over estimate inundation (Mcleod et al., 2010;NOAA, 2010), but it may also suggest that SLAMM under estimates inundation.
In presenting the SLAMM output in maps, it was important to stress the uncertainty of habitat predictions due to errors in input elevation and land cover maps.
To increase the interpretability of the SLAMM results, I aggregated the 11 possible SLAMM classes into five land cover categories for mapping applications: upland, forested wetland, marsh, beach, and open water (Figures 6 and A5.14-A5.20). These maps will be a useful tool for managers interested in wetland-specific predictions and comparing modeling approaches.

Overall Inundation Index and Implications for Sentinel Sites
For the most part, Cape Cod's sentinel sites were located in high elevation  Figure   A6.1). In their study, segments of shoreline were ranked low to high vulnerability using an index that combined geological and physical variables. The locations of high vulnerability sentinel sites as ranked by the Inundation Index tend to appear immediately inland of those shoreline segments with a high CVI vulnerability rank.
At Assateague Island, all of the sentinel sites were in low elevations below 2.6 m. Many of them are at risk from sea level rise and storm surge from large hurricanes.
The CVI assessment (Pendleton, Williams, and Thieler, 2004) at ASIS ( Figure A6. 2) and spatial pattern of the overall Inundation Index corroborate each other also.
Disagreement occurs at points near the Chincoteague Inlet. The CVI rates that shoreline as low vulnerability because of high accretion rates, and the Inundation Index rates sentinel sites in that area as high vulnerability because of the low elevations.

Directions for Future Work
The study presented here assessed inundation risk at 97 sentinel sites located in two northeastern U.S. coastal National Parks. The methodology I used can be applied at other coastal parks and at the same parks at future dates and with future datasets.
Models of sea level rise and storm surge are continually being refined and inundation probabilities can be recomputed as new models are developed. Estimates of sea level rise are evolving as new data from satellite altimetry and ice melt studies are acquired.

Introduction
An increase in the rate of sea level rise is one of the most serious potential impacts of climate change (IPCC, 2007). Global (or eustatic) sea level rise is caused by the thermal expansion of ocean water due to rising global temperatures, and an increased output of water from land-based sources, such as melting glaciers. The rate of eustatic sea level rise has accelerated since the 19 th century (Donnelly et al., 2004;Kemp et al., 2011). However, the rate of relative ( Other local effects contributing to relative sea level include soil compaction, fluid withdrawal and shallow subsidence in marshes . On the Northeast Atlantic Coast, many regions are experiencing subsidence from glacial isostatic adjustment effects  with rates (as measured by mean sea level trends at tide gauges) generally increasing towards the south (Table A1.1).

Uncertainty in Projections
The largest challenge in performing sea level rise inundation risk assessments is the great uncertainty regarding future expected sea level. Determining the rate and acceleration of local sea level rise is complicated by the small number of long-term tide gauges (Houston and Dean, 2011), strong spatial variation in the distribution of melting ocean waters , and seasonal-to-decadal temporal variation (Church, White, and Arblaster, 2005).
However, many climate scientists agree that the Northeast coast is particularly vulnerable to higher rates of relative sea level rise (Frumhoff et al., 2007). Along with high rates of subsidence, the northeast North American coast faces a predicted increase in dynamic sea level due to Atlantic meridional overturning circulation slowdown in the 21 st century . Global climate models predict ocean surface warming would shut down deep convection in the Labrador Sea and slow the sub-polar gyre, and these impacts would result in dynamic, or ocean circulation driven, sea level rise in the Northeast region .
Given this combination of influences and the numerous estimates of future global sea level rise rates, downscaling global projections to a local level is challenging. To address this difficulty, risk assessments must model multiple scenarios in order to gain an understanding of the range of potential impacts. In this assessment, I chose 0.6 m, 1 m and 2 m of relative sea level rise as plausible scenarios for 2100 in the Northeast.

Uncertainty in Elevation Mapping
Sea level rise inundation risk assessments are further complicated by the lack of high resolution topographic data. Detailed maps of elevation are necessary to determine which areas fall within elevations that may be inundated under various sea level rise scenarios. The U.S. Geological Survey (USGS) National Elevation Dataset contains the most accurate readily available digital elevation models for the United States. For much of the Northeast coast, the highest resolution data available are derived from 5 or 10 foot contour USGS topographic maps and are accurate to ± 2.4 m . A more recent source of elevation data is from LiDAR (Light Detection and Ranging) acquired from a plane-mounted laser sensor that emits pulses of light energy at the ground, and is accurate to 0.15-1 m . The accuracy associated with elevation data limits the sea level rise increment that can be modeled, and determines the range of uncertainty associated with inundation predictions ( Figure   A1.1) The National Park Service Inventory and Monitoring Program maintains and manages LiDAR data for coastal parks in the Northeast. At present, there are some LiDAR elevation data for every park in the URI-NPS Monumentation study (Murdukhayeva et al., 2012) except Acadia. However, the coverage of these data is sometimes incomplete or in need of updating (Skidds, 2011 . This conversion has associated errors which have been calculated on a regional basis (Table A1.2)

Conclusion
Given these limitations and uncertainties, it is important that inundation modelers and mappers use their tools and models properly, select sea level rise scenarios appropriate to the available data accuracy, and most importantly, interpret resulting maps and products with an understanding of the possible inaccuracies involved. In the future, collection of higher accuracy regional elevation data and installation of more coastal tide gauges would be great assets to assessing coastal risk from sea level rise and helping in resource management prioritization.

Mapping 1 m of sea level rise on land, adapted from Gesch (2009) Digital elevation models with different vertical RMSEs result in inundation zones with
95% confidence intervals and estimates of uncertainty.

Introduction
There are many methods for modeling sea level rise and storm surge impacts, ranging from simple estimates of inundation based on available elevation data and

Zero-Dimensional
There are varying levels of complexity for flood inundation models, ranging from zero-dimensional (0D) to three-dimensional (3D). The most basic models are 0D and do not include any physical laws. 0D models overlay the determined high water level with a digital elevation model to create a water surface. This is a popular method and is also referred to as "linear superposition" or "bath-tub modeling." It is relatively inexpensive to run and can coarsely approximate coastal vulnerability for a variety of scenarios. The results can be incorporated into a Geographic Information System (GIS) to calculate potentially inundated areas.
The bath-tub approach provides quick analyses of vulnerability at regional and global scales. The analysis is done in a raster environment. The approach identifies all cells with elevations lower than the projected sea level rise as vulnerable to inundation. A more sophisticated approach incorporates connectivity and requires that the cells identified as vulnerable have elevations lower than the projected sea level rise and are adjacent to the ocean or to other inundated cells (Poulter and Halpin, 2008).
The approach is limited: it does not incorporate flood defenses, or recognize coastal processes such as wetland accretion. It tends to result in large overestimations of flood extents. Like all modeling approaches, it is limited by uncertainties in sea level projections and elevation data, and lack of data on the feedback of physical and social systems (e.g. sediment transport regimes and human adaptation responses). However, due to its simplicity and efficiency, the approach has been applied in a variety of studies. Weiss, Overpeck, and Strauss (2011) used it to identify areas at risk from sea level rise of 1 to 6 m in 20 coastal cities in the United States, and Demirkesen, Evrendilek, and Berberoglu (2008) used it to identify vulnerable low-lying coastal areas in Turkey.

ANUGA Hydrodynamic Model
The ANUGA hydrodynamic inundation modeling tool was first developed by The user inputs the bathymetry and topography for the study area, the initial water level and boundary conditions such as tide or any forcing terms that may drive the system such as wind stress or atmospheric pressure (Nielsen et al., 2005). The study area is represented by a mesh of triangular cells. The model was developed primarily for modeling effects of tsunami and was validated with a wave tank simulation of the 1993 Okushiri Island Tsunami (Nielsen et al., 2005). Developers suggest using the model for detailed inundation modeling of small sections . ANUGA is being developed to incorporate riverine flooding and storm surge flooding scenarios. The model has only been tested at a few locations and the current release (Version 1.2.1) is still in the development and debugging process. I downloaded and installed the software from the website and found that the documentation and forums were unhelpful in understanding implementation. I also found that I did not have enough technical knowledge of Python to debug some of the issues.

MIKE by DHI Software Series
The ArcGIS and was used in Sweden to simulate riverine flooding and obtain flood information for emergency planning (Yang and Rystedt, 2002).The program is available for $57,400 and there is a university discount for student researchers (DHI, 2011). The prohibitive nature of the price also led me to remove it from consideration for the NPS study. However, the model seems to be a suitable tool for engineers and could be appropriate for other research endeavors.

ADCIRC
The ADCIRC (Advanced Circulation) Model was developed as a joint project between the University of Notre Dame and University of North Carolina Chapel Hill (Luettich, Westernick, and Scheffner, 1992;Westerink et al., 1992). The 2D and 3D components of the program solve equations for motion for a moving fluid on a rotating earth using traditional hydrostatic pressure and Boussinesq approximations.
The software is used to model tides and wind driven circulation, analyze hurricane storm surge and flooding, and conduct dredging and material disposal studies and larval transport studies. FEMA uses ADCIRC to develop FIRMs (Flood Insurance Rate Maps). ADCIRC has been used to model storm surge for New Orleans, Louisiana (Westerink et al., 2008), the Chesapeake Bay (Shen, Gong, and Wang, 2005;Shen, Gong, and Wang, 2006) and the northeastern Gulf of Mexico (Chen, Wang, and Tawes, 2008). These studies have demonstrated the high accuracy of the ADCIRC model in simulating coastal storm surge (Lin et al., 2010). The program is extremely complex, run on over 100 computers simultaneously at the UNC campus and cannot be used by me for the NPS study. While the software is too computer intensive for the NPS study, it could be an appropriate approach for other projects.

HAZUS-MH
The HAZUS-MH (Multi-Hazards) software was developed by ABS Consulting as a hazard model to be used with ArcGIS. It is the modeling program used by FEMA to estimate potential losses from natural disasters. The program models earthquakes, hurricane winds and floods, and estimates the physical, economic and social impacts of disasters. It identifies high-risk areas, and illustrates spatial relationships between populations and permanently fixed geographic assets and resources (FEMA, 2010).
The model's flood component uses a flood loss estimation methodology consisting of two modules. The first module is a hazard analysis module that uses characteristics such as frequency, discharge and ground elevation to estimate flood depth, flood elevation and flow velocity. The second module is the loss estimation module that calculates the physical damage and economic loss using census block and default building inventory data (Scawthorn et al., 2006a;Scawthorn et al., 2006b). The user can edit this database or incorporate locally available flood information using the Flood Information Tool (FIT) in the software.
I decided not to use HAZUS-MH for the NPS study. Damon (2010)  Another concern is that SLAMM does not consider increased rates of accretion due to sea level rise, a feedback mechanism that can be incorporated into numerical models.
Many wetland scientists recommend use of the elevation capital technique  and the use of numerical coastal models that predict the response of sea level through non-linear feedback mechanisms (Kirwan and Guntenspergen, 2009) for local studies.

SLOSH
The  (Jarvinen and Lawrence, 1985;Jelesnianski, Chen, and Shaffer, 1992). The output can be used to estimate potential surge and flooding for a given hurricane category, forward speed and direction (FEMA, 2003 MOMs represent a worst case scenario of surge inundation, and therefore should not be used for emergency planning (Marcy, 2011). The reported accuracy for the SLOSH model is plus or minus 20% of the peak storm surge. If the model calculates a peak storm surge of 10 feet for the event, the observed peak ranges from 8 to 12 feet. The accuracy was accessed by looking at surge measurements (primarily high water marks) from past hurricanes. The model does not account for rainfall amounts, river flow or wind-driven waves (FEMA, 2003). SLOSH has been utilized in many inundation vulnerability assessments (Lin et al., 2010;Stockdon and Thompson, 2007).

Coastal Vulnerability Indices
The Coastal Vulnerability Index (CVI) was developed by the USGS Coastal and Marine Geology Program to determine the relative risk that physical changes will occur as sea level rises (Thieler, Williams, and Hammar-Klose, 1999). The index is a simple classification of the relative vulnerability of shoreline segments. The index is based on six criteria: tidal range, wave height, coastal slope, geomorphology, shoreline erosion rate and historical relative sea level rise rate, gathered from a variety of sources. The approach yields a relative ranking of possibility that future physical change will occur.   There is an error of 0.02 m associated with those points that are "fixed," so 0.02 m was added to each error value to represent cumulative error at a point.