Using high-resolution bottom albedo images and object-based classification to improve benthic habitat maps.

The complexity of coral reefs and the heterogeneity of shallow coastal waters over small spatial scales provide a challenging environment for mapping and monitoring these benthic habitats using remote sensing imagery. High-resolution benthic habitat maps were developed from AVIRIS- and WV2-modeled bottom albedo products from pre-processed imagery (atmospheric and water column corrected) for La Parguera Reserve, southwestern Puerto Rico. An ISODATA classification was performed with an initial high clustering that was gradually reduced to 10 clusters. The segmented images were converted to polygons and exported to ESRI ArcMap 10.3 where Spatial Join was performed with ground validation points to classify the polygons. The classes were: coral reefs, seagrass, hardbottom, mixed sand/hardbottom/coral, mud, sand, and sand with benthic algae. An accuracy assessment was performed where the overall accuracy (AVIRIS = 63.55%, WV2 = 64.81%), kappa coefficient (AVIRIS = 55 %, WV2 = 57%), and tau coefficient (AVIRIS = 59%, WV2 = 60%) were evaluated. No major class differences were found between the AVIRIS and WV2 classification totals, except for coral reefs and sand classes’ totals. The reduction in coral reefs class totals could be attributed to temporal differences of the images depicting changes in the coral reef distribution within the reserve. The overall accuracies were lower when compared with other studies using similar object-based methods. However, these studies used areas that were relatively small, in shallow clear waters, and not as optically complex as our study area. A major contribution of this study was the creation of the first benthic habitat map for La Parguera Reserve that: 1) provided multi/hyperspectral information at a spatial resolution (4 meters), 2) covered the extent of the reserve to depths of 30 meters, and 3) provided a baseline for future development of benthic habitat studies using an objective classification scheme.