Before inputting all variables into the index model, the streams and dams data must be preprocessed. Only the rivers leading into the ocean with dams upstream is selected for this analysis. DEMs of Oahu and Kauai are used in the classification of the dams on these streams through flow accumulation. In order to get the DEMs ready for creating flow accumulation models, a number of ArcGIS tools is applied to prepare the DEMs for calculating flow accumulation. First, the Fill tool is used to get rid of depressions in the DEM to make the slope values more similar to their neighboring values (Planchon & Darboux, 2002). The Flow Direction tool is then used on the corrected DEM to determine which way water will flow towards the coastlines (Tarboton, 1997). Finally, the Flow Accumulation tool is applied to the flow direction raster to determine how much water would accumulate in each raster cell. This is then used to analyze the amount of sediment transfer that would be occurring through rivers (Schauble et al., 2008). This is outlined by Figure 3.

Dams will be reclassified based on how much they intervene with the stream's ability to transport sediment to the coast. The flow accumulation value at the dam is divided by the flow accumulation value at the coast (where the dam’s river leads to) in order to determine the level of interference. This method of using flow accumulation is based on the assumption that more flow would have more sediment. Some streams lie within watersheds that carry a lot more water and sediment than other streams. Therefore, to account for the significance of the watershed that each dam lies within, the flow accumulation ratio is multiplied by the watershed area. The following equation was used for each dam:

The values for the dams are then classified into three categories based on Jenk’s Natural Breaks. These three categories are then placed in the index model on a scale of 1 to 3. From there, the watersheds are then classified based on the sum of indexed dams (values of 1-3) that lie within them (see Figure 4). Table 1 (shown in objective 2) outlines the indexed dams and watersheds. The result of this is a watershed vulnerability map which is then converted to raster format.

The other human variables are processed through the scoring system discussed in objective 2. They are placed into the Coastal Vulnerability Index model where all the scores are summed similar to McLaughlin and Cooper’s (2010) model (see Figure 7). All data layers are shapefiles, so a conversion to raster format is necessary to accrue the values from the index model. An extra step is required, however, with hotel and small harbor data which are represented by point features. Point features needed to be converted to polygon features in order to account for the area that they are taking up on along the coast. The area (in hectares) is provided in the attribute table for small harbors, so a buffer shapefile is then created based on this data to account for the size of the small harbors (see Figure 5). The hotel shapefile did not come with area data. Therefore, a building footprint shapefile is acquired for the island of Oahu which included the area in hectares of each building. Eighty-three hotels within the building footprint shapefile are selected and the average area of those hotels is then calculated and used to create a buffer shapefile for the hotels on both Oahu and Kauai (see Figure 6). Once each human variable are converted to raster format, they were reclassifed according to the index model.

The land use, road, hotel, and harbor raster layers are then added together using the ArcGIS tool ‘Raster Calculator’ to create a vulnerability map. This map is used in conjunction with the watershed vulnerability map to target areas that are considered as high risk to causing coastal erosion due to human activity (see Figure 7). Each cell in the resulting maps represent a value based on the sum of different values from all layers in the index model. The models used to derive this final raster layer in ArcMap is displayed in Figure 4 and 7. The same models is used to create maps for both Oahu and Kauai.

Figure 3. Input variables into the flow accumulation model for the streams and dams layer. After this process is done, it is used in conjunction with the other variables for the coastal vulnerability index. The ‘(A)’ is used here to show that the output from this model is used as an input for the model in Figure 4.

Figure 4. Input variables into Coastal Vulnerability Model to create a vulnerability map based on watershed areas. (A): Output from flow accumulation raster from previous data processing.

Figure 5. Preprocessing harbors point data into polygons to be used in conjunction with the other variables for the CVI. The ‘(B)’ is used here to show that the output from this model is used as an input for the Coastal Vulnerability Index shown in Figure 7.

Figure 6. Preprocessing the hotels point data into polygons to be used in conjunction with the other variables for the CVI. The ‘(C)’ is used here to show that the output from this model is used as an input for the Coastal Vulnerability Index shown in Figure 7.

Figure 7. Input variables into the CVI to create a vulnerability map. The model is repeated for each island. (C): Output hotel buffers and (B): Output harbor buffers from previous data processing.