Very high-resolution satellite-derived bathymetry using panchromatic and multispectral image fusion

Satellite-derived bathymetry can rapidly acquire shallow water bathymetric data with high spatial resolution. However, the high cost of very high-resolution (VHR) imagery limits the acquisition of VHR bathymetric data. Therefore, we focus on VHR bathymetric mapping through images generated by panchromatic and multispectral image fusion, also termed pan-sharpening. In our study, a new, to our knowledge, VHR bathymetric mapping approach [the random forest algorithm incorporated with pan-sharpening (RF-P)] is established by using multispectral and panchromatic images from the GF-2 satellite and ICESat-2 elevation data. RF-P was then applied in Ganquan Island. The experimental results indicate that the RF-P is the most effective method, with mean absolute error (MAE) and root mean square error (RMSE) values of 1.18 and 1.40 m, respectively. Compared with the log-ratio model, which employs fused panchromatic and multispectral images for VHR bathymetry mapping, RF-P reduces the MAE by 0.48 m and the RMSE by 0.60 m. This provides a cost-effective, rapid, and precise method for the mapping of VHR bathymetry data in optically shallow waters.