A subregional shallow water bathymetry derivation method for coral reef using ICESat-2 and Sentinel-2 combined with sediment information

The bathymetric data is of great significance to coral reef protection. Traditional measurement methods have low efficiency, high cost, and are easily affected by natural conditions. Satellite-derived bathymetry (SDB) in the shallow water area has high efficiency. However, the accuracy of conventional single-band, logarithmic ratio, and multi-band model is easily affected by the difference of sediment types, that is, the portability of a single model is poor. It is necessary to take into account the coral reef sediment types and improve the accuracy of SDB. Fine sediment classification relies on field acoustic and spectral measurements, which is costly and subject to many restrictions. In this study, a subregional shallow water satellite-derived bathymetry method in combination with sediment information (SDB_SS) is proposed by integrating various algorithms. Firstly, based on the Sentinel-2 multi-spectral imagery, the study area is roughly divided into six types of sediment samples: Sand, Rubble, Seagrass, Rock, Microalgae mat, and Coral / Algae. Then, the sediment pixel information is counted, and the coral reef area is divided into three sub-regions of high, medium and low reflection by threshold segmentation method. Finally, combined with ICESat-2 and Sentinel-2 data, the improved subregional satellite-derived bathymetry model is constructed to obtain the shallow bathymetry results. Taking the Yongle Islands in the South China Sea as an example, root mean square error (RMSE), mean relative error (MRE), and mean absolute error (MAE) of the retrieval results are 0.97 m, 0.90 m and 21.23 %, respectively, compared with the measured depth. Compared with conventional single-band, logarithmic ratio, multi-band and other models, they are reduced by 0.27 m similar to 1.65 m, 0.11 m similar to 1.76 m and 2.5 % similar to 65.27 %. It is proven that the proposed method can provide a reference for improving the SDB accuracy of coral reefs without measured spectral data.