A downscaled bathymetric mapping approach combining multitemporal Landsat-8 and high spatial resolution imagery: Demonstrations from clear to turbid waters

High spatial resolution bathymetric maps of coral reefs can show the details of terrain. However, most satellite-based imagery with a spatial resolution < 10 m has only three visible bands and one near-infrared (NIR) band. When in situ bathymetric data are unavailable, it is difficult to map bathymetry from high spatial resolution imagery with spectral matching or empirical models for clear or turbid waters. In this study, we developed a downscaled bathymetric mapping approach (DBMA) that uses the water depth estimated from multitemporal Landsat-8 data to calibrate the empirical model for high spatial resolution imagery (e.g., Sentinel-2A/B, GaoFen-1/2, ZiYuan-3, and WorldView-2) in the absence of in situ bathymetric data. Our results show that DBMA provides high accuracy for depth ranging from 0 to 12 m for clear waters (0 m to 5 m for turbid waters), with a root mean squared error (RMSE) smaller than 2 m. Relative to empirical models (calibrated with in situ data), DBMA underestimates water depth more for depth >12 m for clear waters (5 m for turbid waters) while slightly overestimates more for depth < 4 m for clear waters (3 m for turbid waters). Nevertheless, DBMA performs better than the empirical models for depth between 4 m and 12 m for clear waters (3 m and 5 m for turbid waters). Furthermore, the good performance of DBMA is also demonstrated by the finding that the scaling effect on the DBMA is limited. DBMA presents a reliable solution to obtain high spatial resolution bathymetric maps without leaving out small regions in the absence of in situ data.