Feature geometry.

ESRI

Hurricane Category (1-4)

ESRI

Feature geometry.

ESRI

Internal feature number.

ESRI

ground condition

ground condition

ground condition

ground condition

ground condition

Obtained SLOSH (Sea, Lake and Overland Surge from Hurricanes) model output from the National Hurricane Center.  The data was provided in ArcGIS shapefile format as two polygon shapefiles.  One of the polygon shapefiles represented the Narragansett Bay / Buzzards Bay SLOSH Basin (PVD), and the other shapefile represented the New York City (2) SLOSH basin (NY2).  For both shapefiles, each polygon contained four attributes.  The attributes represented the water surface elevation (in feet) that would occur from the worst-case hurricane surge within each polygon for hurricane categories 1 through 4 (Mean High Tide MOM’s).  The shapefiles were both in a Geographic NAD 27 horizontal coordinate system and used the NGVD29 vertical datum.

For both SLOSH Basin shapefiles, used XTools to create a point shapefile of the centroids of the SLOSH data polygons.

Overlaid the point and polygon shapefiles showing both the NY2 and PVD SLOSH Basins.  The two SLOSH basins overlapped, and the goal was to determine which values to use from which SLOSH basin in the area of overlap.  The points were labeled with their category 1 through 4 MOM values, and a visual analysis was performed.  It was assumed that the SLOSH values would be most accurate in areas along the coast where the SLOSH model grid had the finest resolution.  The PVD and NY2 SLOSH model grids are approximately the same size along the CT coast at I = 60 for the NY2 grid and I = 59 for the PVD grid.  Those locations are approximately coincident.  The Mean High Tide MOM’s from both grids in those areas were very close in value for category 1 through 4.  That was true even up to a few grid cells east and west of that location.  Therefore, it was decided to use both sets of Mean High Tide MOM’s in the immediate vicinity of NY2 I = 60 and PVD I = 59.  Up to two grid cells away from that location, discretion was used to select the points, erring on the side of higher values for worst case scenario.  West of that location only values from NY2 were used.  Likewise east of that location only values from PVD were used.  This entire process was coordinates with Will Schaffer, Arthur Taylor, and Stephen Baig of the National Hurricane Center in order to ensure the validity of our approach.  Once it was determined which SLOSH points to use, one SLOSH point shapefile was created for action going forward.

Projected the SLOSH point shapefile from Geographic NAD 27 to NAD_1983_StatePlane_Connecticut_FIPS_0600_Feet.  Used Xtools to populate the shapefile with x and y coordinates.

Created four point shapefiles from the previous point shapefile, representing Category 1 through 4 hurricane surge.  Within each of the four shapefiles, deleted records that contained values of "99.9" for the represented Category, which represented areas that were not flooded in the SLOSH model runs.

For each of the four shapefiles, wrote the shapefile horizontal coordinates vertical SLOSH elevations out to a text file, and used CorpsCon to convert the vertical SLOSH elevations from NGVD29 feet to NAVD88 feet.  Then the text files were converted back to shapefiles, but now with SLOSH elevations of NAVD88 feet.

On a county by county basis, interpolated each of the shapefiles to produce interpolated raster surfaces (ArcInfo Grids) representing the hurricane surge water elevation for Category 1 through 4 hurricanes.  Used IDW interpolation, with the following parameters: power = 2, Search radius type: Variable, Number of points: 6.  Used a cell size of 10 feet to facilitate interpolation, as smaller cell sizes proved to be too computationally intensive.  This resulted in raster grids representing Category 1 through 4 hurricane surge for each county.  This completed the preparation of the water surface grids.  The next step was to prepare the land surface grids.

Obtained one LiDAR dataset from FEMA.  This dataset covered all of coastal CT except for the CT River valley.  The data was collected December 16-18, 2006 by Terrapoint USA for FEMA through the Dewberry & Davis Fairfax, VA office.  It was collected in leaf-off conditions at low tide.  The data was in CT Stateplane NAD83 feet, with a vertical datum of NAVD88 and vertical units of feet.  The data was provided in multiple formats, but we elected to use the data in TIN format.  The data documentation certifies: “This LiDAR project covered approximately 40 sq miles along the coastline of Connecticut and was acquired in December of 2006 providing a mass point dataset with an average point spacing of 3 ft.  The LiDAR has been compiled to 3 foot vertical accuracy; tested to 0.23 feet vertical accuracy at 95 percent confidence level”.

Obtained another LiDAR dataset from FEMA.  This dataset covered the CT River valley.  That dataset was collected from May 8, 2004 through June 16, 2004 by Spectrum Mapping under contract to ENSR International for FEMA.  The data was in CT Stateplane NAD83 feet, with a vertical datum of NAVD88 and vertical units of feet.  The data was provided in ASCII x,y,z format.  The data documentation certifies: “Spectrum Mapping was tasked by ENSR International to collect LIDAR data and digital ortho imagery to generate breaklines and bare earth DEM to support 2-foot contour intervals in accordance with FEMA Appendix A.  The data meets contract specifications for 2-foot contours of 0.50ft (15 cm). The combined DEM accuracy for CT and MA project areas was .483 feet RMSE.  The terrain data match the orthophoto imagery and are consistent with the elevation dataset.  The data conforms to the Guidelines and Specifications for Flood Hazard Mapping Partners, Appendix L”.  We converted this data to TIN format, and then converted the TINs to ArcInfo Grids.

Obtained one LiDAR dataset from the CT Department of Environmental Protection (DEP).  This dataset was collected for FEMA in April 2000 by TerraPoint, LLC.  The data was in CT Stateplane NAD83 feet, with a vertical datum of NAVD88 and vertical units of feet.  The data had a 20 foot post spacing, and was in ASCII x,y,z format.  Coverage for that dataset was statewide.  The data documentation certifies: 1) The LiDAR data acquisition and data processing efforts were conducted using state-of-the-art commercial means to achieve dual objectives of providing a bare earth digital elevation model (DEM) capable of supporting orthorectification and for the future development of 5-foot contour maps meeting National Map Accuracy Standards (NMAS). 2) Compliant with existing and recommended test methodologies for LiDAR data, developed by FEMA (see FEMA 37, Appendix 4b) and others, and the requirements of NMAS for 5-foot contour interval maps, LiDAR data will be accurate to within 1.25 feet rmse in open flat areas; and 3) Compliant with NMAS requirements, 90 percent or more of the LiDAR-derived bare earth DEM will be within one-half of the 5-foot contour interval in open flat areas.  The data points are nominally spaced at 1.5 foot intervals with an approximate 3 foot horizontal accuracy.  Surface elevation accuracy is better than 1 foot”.  We converted this data to TIN format, and then converted the TINs to ArcInfo Grids.

On a county by county basis, developed bare earth ground surface elevation grids using the three LiDAR datasets, in the following order of priority: FEMA coastal LiDAR, FEMA CT River LiDAR, State of CT statewide LiDAR.   Data was closely analyzed for areas of erroneous data.   If found, data from one of the other datasets was used in those areas, and the resulting surface was closely scrutinized to ensure that elevations were consistent across both data sources.  The resulting bare earth ground surface elevation grids were used for inundation calculations, and for extracting a zero NAVD88 shoreline.

Created a zero NAVD88 shoreline using both the bare earth ground surface elevation grids and the hydro shapefile for the state of Connecticut esi_2002_hydro_line.shp.  The shoreline was then converted to a “land” polygon shapefile which was used for clipping the inundation data in later stages of the process.  A “water” polygon shapefile was also created to represent the adjacent ocean/water areas.  That shapefile was used in subsequent steps.

The next step was to subtract the land surface grids from the water surface grids to create “inundation” rasters depicting which areas would and would not be flooded (inundated) by worst-case hurricane surge for the mean high tide scenario.  On a county by county basis, ran the following aml script from GRID.  This script subtracts the bare earth land surface grid from each category 1 though 4 slosh water surface elevation grid.  Positive values indicate areas that are “inundated”, and zero or negative values indicate areas that are not inundated.    The script codes areas that are inundated only by category 1 with a “1”, areas inundated by category 1 & 2 with a “2”, areas inundated by category 1, 2 and 3 with a “3”, and areas inundated by category 1 through 4 with a “4”.  The paraphrased aml script is: “inundation = con(slosh_cat1 – bare_earth > 0, 1, ~  slosh_cat2 – bare_earth > 0, 2, ~  slosh_cat3 – bare_earth > 0, 3, ~  slosh_cat4 – bare_earth > 0, 4)”

On a county by county basis, the resulting inundation raster was then converted to a polygon shapefile and clipped using the coastline/land area polygon created above.  The reason for this step is that the inundation rasters sometimes extend beyond the zero NAVD88 coastline that we will be suing for the mapping, because the LiDAR may contain elevation data seaward of zero NAVD88 if collected at low tide.  If this step were not taken, the mapping would show inundation seaward of the zero NAVD88 shoreline, which would look poor cartographically.  The resulting shapefile shows the inundation over land areas only.

On a county by county basis, converted the inundation polygon shapefile back to a raster (grids).  Also converted the “land” and “water” polygon shapefiles to rasters, setting the cells of the land raster to a constant value of "10", and the cells of the water raster to a constant value of "0".  The reason for this will be explained in the next steps.

On a county by county basis, merged the inundation raster with the “land” and “water” rasters.  The result of the merge is an integer raster containing values of 1, 2, 3, 4, and 10.

Ran the ArcInfo Grid "Fill" command to fill any "sinks" or depressions in the inundation grids. This runs much much faster than filling DEM’s, primarily because the inundation grids are integer grids and DEM’s are floating point grids.  The results of the fill process are the same.  It removes areas that are low, but which have higher ground between their location and the coast which would prevent coastal surge from reaching the area.  In previous process steps, the land areas were set to "10", and ocean and river areas were set to "0". In between, the inundation was coded as 1 thorough 4 according to hurricane category. Therefore, the "Fill" command ensures that all flooding "flows" from high ground to low ground, or from 10 to 4, 3, 2, 1, and 0. The paraphrased aml script is: “fill prefillgrid filledgrid sink”

Converted the resulting filled inundation rasters to shapefiles.

Performed Quality Control on the resulting inundation shapefiles.  After running the fill command on the inundation raster, all isolated flooded areas, or "sinks" will have been removed. However, there is one special circumstance that needs additional inspection.  This occurs when a flooded area is divided by a thin, man made raised surface, such as a highway.  On the DEM,  this raised surface will appear as high ground and the surrounding area will be lower, which created a few areas of “artificially isolated sinks”, since flooding can actually occur underneath the road and progress beyond it.  But the fill command will remove these areas because they are isolated. These areas must be correctly represented as flooded areas, as they were originally.  This was performed by copying the inundated area from the pre-filled inundation shapefile pasting it into the post-filled layer.

Two additional areas required additional scrutiny.  One was Stamford Hurricane Barrier in Stamford (Fairfield County).  The initial inundation raster showed that the area behind the Barrier would be flooded by category 1 through 4 hurricanes.  Closer scrutiny showed that the barrier protects the areas behind it against all but rare Category 3 and Category 4 storms moving in a NW or WNW direction.  Therefore, the areas flooded by category 1 & 2 hurricanes was changed to show flooding only by category 3 hurricanes, and the following note was added to the map: “Surge flooding generated by Category 3 & 4 hurricanes with westerly track directions can exceed the design height standards of the Stamford Hurricane Barrier. Although hurricanes of this nature are considered rare events, their occurrence is possible."  The second area that required additional scrutiny was the Pawcatuck Hurricane Local Protection Project in Stonington (New London County).  The initial inundation raster showed that the area behind the Barrier would be flooded by category 3 & 4 hurricanes.  However, closer scrutiny showed that the barrier protects the areas behind it against all SLOSH surges.  Therefore, the area behind the barrier was shown as being protected (not flooded), and the following note was added to the map: “Worst-case hurricane surge inundation is not expected to overtop the Pawcatuck Hurricane Local Protection Project”.

Using ArcGIS 9.3, created a polygon feature class named CT_Hurricane_Surge_Inundation and populated it with the Worst case Hurricane Surge Inundation shapefile data for the four counties along the Connecticut coastline - Fairfield, Middlesex, New Haven, and New London - to create a statewide feature class. Subsequently imported the metadata from one of the county shapefiles. (Metadata for all 4 counties were identical).