Depth derivation using multispectral WorldView-2 satellite imagery

Despite recent efforts to collect high-resolution multibeam bathymetry data across the Pacific Islands Region, significant gaps exist in the 0-30 m depth range. Achieving bathymetric coverage in these areas is critical for assessing the health of coral reef ecosystems that reside there. Here we use WorldView-2 multispectral satellite imagery and two depth derivation methods (Lyzenga, 2006; Stumpf et al., 2003) that relate spectral radiance values to ground truth depth information to derive depths for shallow-water regions in the main Hawaiian Islands. Our results show increased accuracy using the Lyzenga (2006) multiple linear regression method when compared to the Stumpf et al. (2003) ratio method. Furthermore, we achieved improved results by eliminating the linearization process from the Lyzenga (2006) method. This improvement may be related to the lack of large seagrass aggregations within the main Hawaiian Islands because the presence of seagrass has been shown to affect the linear relationship between ground truth depth and spectral radiance values (Doxani et al., 2012). The accuracy of our derived depth product is directly related to the quality of the multispectral satellite images, the availability of ground truth data, and water depth with accuracy decreasing substantially in water depths > 20 m. Our results show that in the absence of shallow (0-20 m) high-resolution bathymetric data, satellite-derived depths are an important resource for assessing, monitoring, and managing shallow coral reef ecosystems. [doi:10.7289/V5668B40 (http://dx.doi.org/10.7289/V5668B40)] 2015 NMFS (National Marine Fisheries Service) PIFSC (Pacific Islands Fisheries Science Center) CoRIS (Coral Reef Information System) Library http://doi.org/10.7289/V5668B40 Public Domain 1860