NASA DEVELOP Summer 2006 Project

Modeling Coastal Inundation Resulting from Tropical Storm Surge

About | Methods | Results | Links

Methods

The first task the DEVELOP team had to complete was processing the LIDAR data. The LIDAR points, originally in text format, were converted to .mdb via Microsoft Access, then converted to a shapefile using the Make X-Y Event Layer in Arc Toolbox. The 121 quarter quad LIDAR tiles were digitally corrected to remove erroneous data. Additional points were added to areas of missing data using 10 foot contour digital raster graphics. Areas of missing data included, water bodies, flight paths, buildings, and some random regions.

Figure 1. A step by step look at filling in the missing LIDAR data using the 10 foot contour digital raster graphic.

The 121 tiles were then merged into five sections with overlapping boundaries.

Figure 2. The five large shapefiles created from the original 121 quater quad shapefiles.

The shapefiles were interpolated to DEMs using the Geostatistical Wizard’s Radial Basis Function and mosaicked together. The Radial Basic Function was set to multiquadric, with a maximum of 8 points and a minimum of 4. The prediction map generated from the RBF was then exported to 100ft, 30ft, and 10ft rasters. An accuracy assessment was preformed on the six different DEM’s generated.(LIDAR 100ft, 30ft, 10ft, SRTM 100ft, NED 30ft, 100ft). Using 34 USGS benchmarks dispersed throughout the state the Develop team calculated the root mean square error (RMSE) for each DEM.

In order to have accurate floods all ocean values had to be set to zero elevation. The CT DEP hydrography layer was edited for greater accuracy using the 2004 digital orthophotos from the CLEAR website. Once the shapefile was edited it was reclassified to 0 for ocean, and 1 for land then multiplied by each LIDAR DEM.

Figure 3. The original LIDAR DEM is on the left, and the image on the right is the result of editing the DEP ocean polygon and using it to set the ocean elevation to 0ft.

Flood heights were generated for Category 1-4 storms occurring at mean tide and high tide using the SLOSH display program. The hurricanes were set to move in a NNE direction at 30mph over the state. These parameters were chosen because they best represented historical storms that have hit the state. Bridges in the LIDAR DEMs created a dam effect, not accounting for the water which can flow underneath. In order to alleviate this problem a shapefile was created to mask out all the bridges using the 2004 orthophotos. This shapefile was then reclassified so that the bridges were 0 and non bridges were 1. This shapefile was then rasterized and multiplied by the DEM so that all the bridges were set to an elevation of zero feet.

Figure 4. The image on the left is the shapefile created to mask out all the bridges in the flood zone, and the image on the right is the resulting DEM after it has been multiplied by the reclassed and rasterised shapefile to the left. Notice how there are now gaps in between the bridges where water can flow.

Each elevation dataset was clipped to the town of Milford. To prevent flooding around the datasets when using a region growing technique in ERDAS IMAGINE Virtual GIS, a border was added at 549.99 feet elevation. The region grower was set to an elevation equal to that given by the SLOSH model and a point was selected within the ocean to grow from. This process provided the extent of the flood but excluded islands. Level slicing was incorporated to create the islands. Using ArcGIS Raster Calculator, the level equal to or below the flood in Virtual GIS was selected for each particular DEM. This procedure created islands as well as unconnected flooded areas. The level sliced region was clipped to the extent of the flood from ERDAS IMAGINE Virtual GIS. This created the final product. This was done for each category hurricane using the NED 30ft, SRTM 100ft, and LIDAR 10ft DEMS.

Figure 5. The three steps to create an accurate flood map.