ハイパースペクトルセンサを用いたインバージョン法による水深分布推定

抄録

Water depth information is extremely important for safe marine navigation and offshore security. Currently, it is possible to acquire bathymetric information effectively with multi-beam echo sounder measurements and airborne laser bathymetry scanners. However, Japan is an insular country surrounded by open ocean, and a low-cost technique is needed for the spatial acquisition of shallow-water depths. Experiments with satellite-derived bathymetry using multispectral satellites have been underway since the 1970s. Their disadvantage is that they always require a field survey. However, the inversion method using hyperspectral data can simultaneously estimate bottom reflectance, water optical properties, and depths without a field survey. This study aims to research and develop a spatially derived water-depth method using hyperspectral data. Using airborne hyperspectral data and field survey data from Yamada Bay, Iwate Prefecture, we investigate adding a bottom index algorithm for classification of benthic type classification into the inversion method processing. We then apply the method to airborne hyperspectral data of Aka-jima Island, Okinawa Prefecture, and demonstrate its versatility with different benthic conditions and optical properties. A comparison to field survey data in Yamada Bay revealed a tendency for estimated depths to be deeper than field survey values, causing a slightly higher RMSE of 3.3m. On the other hand, in Aka-jima Island, the range of estimated depths was in good agreement with field survey data, with a reduced RMSE of 1.3m. Also, comparing estimated bathymetry maps to ALB data for Yamada Bay and an M-7000 digital bathymetric chart for Aka-jima Island, we found trends matched except in areas affected by macro alga and rock reefs shadows in Yamada Bay, or by cloud shadows in Aka-jima Island.