Hyperspectral Bathymetry Retrieval using a Newly Developed Normalized Algorithm in Shallow Water

Shallow water bathymetry is highly significant to regional and national economic development. It is also fundamentally important to coastal benthic environments. Thus, numerous different types of algorithms have been explored to identify one that accurately determines shallow water depth. Among the algorithms that employ data from passive sensors, Ma Sheng's algorithm, which uses the Pearson correlation coefficient (CC) and the similarity coefficient (SC), shows good performance. After analyzing the bathymetry retrieval theory and algorithms included in Ma Sheng's model, we determined that [ln(nSC) - ln(nCC)/ [In(nSC) + ln(nCC) has a good relationship with the field data. Based on that finding, we established a normalized algorithm for estimating shallow water depth from hyperspectral data. Finally, the Ma Sheng's algorithm and our new, normalized, algorithm were compared through analysis of Hyperion satellite imagery against in situ, measured bathymetry over the coastal regions of Saipan and Landfall Islands. Bathymetry retrieved using our method showed a root-mean-square error of 1.937 for Saipan Island and 2.37 for Landfall Island, both of which are comparable to results from Ma Sheng's method. Moreover, the validation results demonstrate that our method exhibits good performance and is an acceptable alternative algorithm for bathymetry retrieval.