GSSHA—Large Watershed Hydrollogy Study

Galveston Island, TX Watershed

1. Modelliing Projject Overviiew

Purpose This project demonstrates how the overland boundary conditions capability in GSSHA is used to model coastal storm surges after a cyclone event. The purpose of this project was to compare GSSHA results with actual inundation depths reported by the National Oceanic and Atmospheric Administration (NOAA) for the same cyclone event in Galveston Island, TX. This model can be used in future studies to show how overland boundary conditions in GSSHA accurately simulate storm surge events.

Model Background The model is located in Galveston Island, Texas as shown in Figure 1. As we are not simulating overland flow, the watershed doesn’t have an outlet location where outflow will be measured; instead, it has boundary conditions. Because the northern section of the island is the most populated, this was the area used for the model, as shown in Figure 2.

Figure 1: Watershed Area

Figure 2: Delineated Watershed

The boundary conditions were added to represent the storm surge caused by Hurricane Ike on Galveston Island in September 2008. Hurricane Ike originated from a tropical wave off the west coast of Africa. Ike made landfall across Great Inagua Island (Bahamas) as a category 4 hurricane and then into the Northeast Coast of Cuba as a category 3 hurricane. Ike crossed Cuba overnight, and emerged into the Caribbean Sea the morning of September 8 th heading to the Gulf of Mexico. By Sept 11, Ike’s tropical storm wind swath was approximately 450 miles wide with a hurricane force wind swath of 180 miles. Ike made landfall on Galveston Island at 2:10 am September 13th as a strong category 2. Ike's large wind swath, along with the fact that it piled water over the shallowest portion of the Gulf, lead to much higher than normal storm surge flooding along the Upper Texas and Louisiana Coasts. (http://www.srh.noaa.gov/lch/ike/ikemain.php ) Ike is directly responsible for 103 deaths across Hispaniola, Cuba, and parts of the United States Gulf Coast. Extensive damage from strong winds, storm surge, and rainfall occurred over Hispaniola, the Turks and Caicos Islands, the southern Bahamas, Cuba, and the U.S. Gulf Coast from Florida to Texas. Additional deaths and significant damage occurred across parts of the Ohio Valley and southeastern Canada after Ike lost tropical characteristics. The Insurance Services Office estimates that the insured damage (not including inland flooding or storm surge) from Ike in Texas, Louisiana, and Arkansas is $9.7 billion dollars. The National Flood Insurance Program estimates that insured losses from inland flooding and storm surge is $489.5 million in the same three states. Using these preliminary figures, total damage is estimated at about $19.3 billion dollars. These estimates suggest that Ike is the fourth costliest hurricane to affect the United States, after Hurricanes Katrina (2005), Andrew (1992), and Wilma (2005). (http://www.nhc.noaa.gov/pdf/TCR-AL092008_Ike.pdf ) To determine the effectiveness of GSSHA in modeling storm surges, we calibrated the model with maximum observed inundation depths on Galveston Island using Hurricane Ike storm surges. The observed depths are shown in Figure 3. This map was generated by the NOAA National Weather Service Forecast Office.

Figure 3: Inundation Depths in Galveston County. http://www.srh.noaa.gov/hgx/projects/ike08/inundation.htm

Data Acquisition and Model Setup We downloaded several datasets to create a GSSHA model of this watershed. The following data were required to develop the model, most of which were obtained using the web service tools in WMS:

1. Land use data: This data was obtained from webgis.com and this particular dataset was placed on the WMS Web Service Catalog site so it could be downloaded from any copy of WMS 8.2 or later. Figure 4 shows Land Use polygons for Galveston Watershed which were used to create land use and combined index maps for determining properties such as overland Manning’s roughness in the watershed.

INDUSTRIAL

CROPLAND AND PASTURE

STREAMS AND CANALS

COMMERCIAL AND NON FORESTED SERVICES

RESIDENTIAL

TRANSPORTATION

LAKE

Figure 4: Land Use Map for Galveston Island Watershed

2. Soil type and texture data: detailed county soil survey geographic (SSURGO) data was obtained from the NRCS. The soil texture for each polygon was obtained from the soil polygons in the SSURGO database. This texture is normally used to determine GSSHA soil parameters. Even though we gathered this data, for this model we did not use Soil type Data to create index maps because in this specific scenario of extreme flood simulation in GSSHA, the infiltration is minimal.

3. Elevation data: 10-meter resolution elevation data was available in this area from the USGS seamless elevation dataset. This data was downloaded directly from WMS. These elevations were used to determine the model extent and to create 2D grid cell elevations in the GSSHA hydrologic model. Figure 5 shows the watershed elevations symbolized by different colors, with red representing the elevation closest to sea level.

Figure 5: Watershed Elevations

4. Topographic and aerial photography data: Digital raster graphic (DRG) topographic maps were obtained from the Microsoft terraserver web service client in WMS. Digital orthophoto quads (DOQ) aerial photographs were also obtained from this same service. The purpose of these maps was to identify outlet locations and other key features within the watershed (Figure 6).

Figure 6: View of delineated watershed over topographic map of Galveston Island 5. Precipitation data: Because we are not simulating the effect of rainfall alone on the watershed and because precipitation is minimal compared to the storm surge caused by Hurricane Ike, a uniform intensity of 10.5 mm/hr was entered for a total of 24 hours of rain. This data was obtained form NOAA National Weather Service Forecast Office, which reported it as a Rainfall Contour Map developed for Harris County, as shown in Figure 7. Galveston Island is located South East of Harris County and these rainfall values were used as a reference for Galveston Island.

Figure 7: 48-Hour Rainfall for Sept. 12-14, 2008 measured by the Harris County Flood Control District, TX http://www.srh.noaa.gov/hgx/projects/ike08/HurricaneIkeRainfall.html

6. Storm Surge Data: Historical storm surge data for the state of Texas was retrieved from the NOAA Tides and Currents website: http://tidesandcurrents.noaa.gov/station_retrieve.shtml?type=Historic+Tide+Data . Figure 8 shows the observation stations along the coast of Texas. We gathered data from two tide gauges, namely: Galveston Pier 21 (29° 18.6' N, 94° 47.6' W) and Galveston Pleasure Pier (29° 17.1' N, 94° 47.3' W) located within the island and shown inside red boxes in Figure 8. In the selection of the tide gauges we had to take into account that some of them, mostly the ones located within our watershed or within Hurricane Ike radius, failed due to strong winds. In addition, some of the gauges did not have data available from Sept 12 to Sept 15, 2008.

Figure 8: Meteorological Stations along the coast of Texas. http://tidesonline.nos.noaa.gov/geographic.html

Figure 9 and Figure 10 show storm surge plots for the two different tide gauges. Note in Figure 9 that peak storm surge values are missing due to gauge failure. These values were estimated using Galveston Pleasure Pier gauge as a reference.

Figure 9: Storm Surge Data for Galveston Pier 21 Station

Figure 10: Storm Surge Data for Galveston Pleasure Pier Station

After obtaining all the data described above, we developed the model using the WMS hydrologic modeling wizard. First, the basin was delineated with an outlet located between Galveston Sewage Disposal Plant and Pier 41 in the west shore of the island as shown in Figure 11. This was just a preliminary delineation to be used to delineate a greater area of the island. With the create arc tool we extended the watershed boundaries to the extent where we needed it, as shown in Figure 12. We did not define stream parameters, redistribute the vertices on the streams or smooth streams in the watershed because the model did not require a stream outlet hydrograph.

Figure 11: Preliminary Delineated Watershed

We created a 60 by 60 meter 2D grid that conformed to the DEM elevations. After the grid was created, we defined the simulation time and the time step, as well as the Land Use index map shown in Figure 13, entered mapping table parameters, and processed the precipitation data described above.

Figure 12: Final Watershed Extent

Figure 13: Land Use Index Map

7. Boundary Conditions Setup: We set two boundary conditions in the model, one representing storm surge caused by Hurricane Ike on the gulf side of the island and one boundary condition for the bay side of the island. The boundary condition on the gulf was entered from the cell south of Greens Bayou with the following coordinates: 29 °15’55.29” N and 94°49’46.58” W. The boundary conditions extend to the farthest point east in the jetties with coordinates: 29 °14’48.82” N and 94°42’58.92” W as shown in Figure 14. The boundary conditions were set to variable depth and the storm surge time series from Galveston Pleasure Pier gauge (Station ID 8771510) were used. This storm surge data can be found at http://tidesandcurrents.noaa.gov/station_retrieve.shtml?type=Historic+Tide+Data Boundary conditions on the bay side of Galveston Island started from the end point of the gulf side boundary condition (in the jetties) with coordinates 29 °14’48.82” N and 94°42’58.92” to the Seawall south with coordinates: 29 °17’23.52” N and 94°52’11.98” W as shown in Figure 14. This boundary was broken up into two arcs to include the outlet node in the middle as shown in the same figure. The boundary conditions were set to variable depth and the storm surge time series from Galveston Pier 21 gauge (Station ID 8771450) were used. This storm surge data can be found at http://tidesandcurrents.noaa.gov/station_retrieve.shtml?type=Historic+Tide+Data An embankment arc was added to represent the levee surrounding Old Fort San Jacinto in the northern part of Galveston Island. We added a second embankment arc to represent the section of the seawall (Seawall Boulevard) going inland behind Stewart Beach Park to the levee. Levee height was set to 6 m, or close to that of the neighboring seawall. Bay boundary condition start point

Outlet

Bay boundary condition end point Gulf boundary condition end point

Gulf boundary condition start point Figure 14: Extent of boundary conditions in Galveston Island GSSHA Model

We ran the simulation several times, making boundary condition adjustments each time in an effort to represent as closely as possible the storm surge as it happened in reality. Our first runs were generating inundation depths lower than the depths obtained in the NOAA website. This was because we were only placing boundary conditions on the gulf side of Galveston Island. As we investigated Hurricane Ike more thoroughly, we found out that inundation was caused mostly by storm surge coming from the bay side of the island as well. Thus, to reconcile the difference and to calibrate the model, we added boundary conditions to the northwest side of the island as mentioned previously. After saving and running the model, we converged on a solution that was very close to the inundation depths measured by NOAA. Generally, when the computed results differ from the observed data, hydraulic conductivity values are adjusted to reconcile the difference. We did not adjust hydraulic conductivity values since we were not including infiltration processes in the model nor are we using soil type data.

2. Resullts

Inundation Depths Comparison Figure 15 and Figure 16 show the observed inundation depths and the computed inundation depths respectively. Since this particular model was developed with two boundary conditions that do not include the southwest section of the island, the inundation depths are slightly lower in this area than the observed depths, as would be expected. The rest of the computed values are a good representation of the observed data, especially the northeast end and the residential area (Galveston) in the center of this model. As more processes are added to the model and with more accurate surge information, a better calibration would likely result .

Figure 15: NOAA inundation depths

(m)

(greater than 10 ft) (10ft) (8 ft) (6 ft) (4 ft) (2 ft)

Figure 16: GSSHA inundation depths results

Overland flow depth contours (meters) at the time of maximum flooding are shown on the topographic map in Figure 17. This flooding occurred on September 13, 2008 at 2:00 pm, almost twelve hours after Hurricane Ike hit Galveston Island.

(m)

(greater than 10 ft) (10ft) (8 ft) (6 ft) (4 ft) (2 ft)

Figure 17: Overland flow depths at the time of maximum flooding

Images An animation of the model was exported to Google Earth and the color filled contours representing water depths are shown in Figure 18.

Figure 18: Depth contours representing inundation depths in Google Earth Animations An AVI animation and a KMZ (Google Earth) animation were exported for this model. The files are called galveston.avi and galveston.kmz.

3. Diiscussiion

Procedure We followed the following procedure to setup and create the Galveston GSSHA model:

1. Open the hydrologic modeling wizard, use the UTM Zone 15 (metric) coordinate system, and define the model boundary. 2. Open the DEM (36991726.hdr), the land use data (Houston.shp), and the image (galveston2m.tstopo.web.jpg) (optional). 3. Run Topaz to compute the flow direction and flow accumulation grids. 4. Delineate the watershed with the outlet located between Galveston Sewage Disposal Plant and Pier 41 in the west shore of the island as shown in Figure 14. Use a stream threshold value of 0.05 square miles. The preliminary watershed area should be approximately 1.21 square miles. 5. Starting from the outlet point, create an arc that follows the DEM contour lines to extend boundary of watershed. Once you have outlined your new watershed extent, define your basin, and then convert from basin to polygon. The new watershed area should be approximately 13.07 square miles. 6. Select GSSHA as your model and initialize the model data. 7. You do not need to set stream arc attributes as you will not use the stream in your model. 8. Create a 2D grid with a resolution of 60 m. 9. Run a 2-day simulation, starting at 6 PM on 9/12/2008 and ending at 6 PM on 9/14/2008. 10. Define your land use coverage by selecting the “Create Coverages” button in the Define Land Use and Soil Data step of the wizard. 11. Create land use index maps. Use the land use index map for your roughness values. 12. Read the gssha.cmt file to get initial values for overland Manning’s roughness. In your job control, turn off the Green and Ampt method of infiltration, select no routing, and set your output to write a hydrograph value every 10 minutes and to write in English units. Turn on Channel depth and Channel flow in the Link/Node data set output. 13. Set the precipitation method to Uniform and enter an intensity of 10.5 mm/hr for 24 hours. 14. Create a node in the start point and end point of your east boundary arc as shown in Figure 14. Set your arc attributes to variable depth and enter the Galveston Pleasure Pier storm surge time series. GSSHA automatically named the time series as ts_26. 15. Create a node in the start point and end point of your west boundary arc as shown in Figure 14. Include a node in the middle where the watershed outlet is located. Set your arc attributes to variable depth and enter the Galveston Pier 21 storm surge time series. GSSHA automatically named the time series as ts_27. 16. Using the topo map as a guide, create an embankment arc where the levee surrounding Old Fort San Jacinto is located (north of Galveston Island). In the embankment dialog box, set embankment height to 6 m. 17. Using the topo map as a guide, create an embankment arc where the seawall goes inland (behind Stewart Beach Park in the northeast section of the island). In the embankment dialog box, set embankment height to 6 m. Make sure that the two embankments don’t overlap. 18. Save your model. 19. Select the option to clean up your model and make sure there are no errors in your model. 20. Run the GSSHA simulation (using GSSHA 5.0). Calibrate your inundation depths to the observed depths using the values shown in this report or using some other calibration method. 21. Create an animation filmloop using this model and open the KMZ animation file in Google Earth.

Uses This model has several potential uses. Using this model as a base model, you could change the storm surge data and determine what areas would be flooded, the changes in inundation depths as well as time of maximum flooding. You could use this model to simulate how would inundation depths change if the seawall of the island would be extended, as well as determine the minimum storm surge that would cause inundation problems.

Possible Errors There are also several possible errors in this model. One error might be that not all the watershed processes present in this watershed, such as infiltration, and storm surge boundary conditions along the southwest part of the island, were considered . Another possible error is that precipitation was only entered for two days, when in reality it rained almost twice as that. The inches of rain were divided into 24 hours instead of 48 hours, thus increasing the intensity. Nevertheless, this was done because precipitation is minimal when compared to storm surge values. This is the same reason why infiltration processes were not included. The model could also be improved by increasing the resolution of the elevation grid. This particular model uses a 60 m grid resolution. This resolution was chosen to optimize runtime. A higher resolution grid would certainly improve the accuracy of the model, especially when viewing flood depths on a more local scale. 4. Concllusiion The calibrated GSSHA model is a good representation of the response of this watershed to a cyclone event. Other storm surges could be run through this model to compute the inundation depths and to determine the effects of changes in the watershed such as the addition of boundary conditions or changes in land use. Further calibration might be necessary as more calibration data becomes available.