Abstract: The bathymetry information is fundamental to characterizing the coral reef environment and natural resources of the South China Sea. However, the data on bathymetry is incomplete and spatially limited due to the difficulty of obtaining accurate and well-distributed soundings in remote oceans. In light of the inherent optical properties (IOP) of the coral reefs dominated by case I water, we used Sentinel-2 images and MODIS data acquired on the same date to derive bathymetry, taking Kugui Reef as a case study. First, we constructed reflectance spectra for different benthic substrates in optical shallow waters, and the remote sensing reflectance and chlorophyll (Chl-a) concentrations were estimated by a semi-analytical model. Then we adopted the ratio algorithm model to carry out the water inversion. Finally, the water depths derived from multitemporal images were averaged to reduce noise signals and to improve the result accuracy. The overall root-mean-square error and mean relative error of retrieved water depths were 2.68 m and 9.99%, respectively, by validating with soundings. This approach used Chl-a concentration as a proxy to deduce IOP parameters and rapidly obtained the preliminary water depths in optically shallow waters of coral reefs from multispectral satellite images in the South China Sea. The depth estimation method can be applied to similar marine environments for related studies.

Key words: water depth inversion, coral reef, multispectral imagery, semi-analytical model, log-ratio model