Diffuse attenuation coefficient and bathymetry retrieval in shallow water environments by integrating satellite laser altimetry with optical remote sensing

Shallow water environmental information is crucial for the study of marine ecosystems and human activities. There have been numerous satellite remote sensing studies focused on this area. However, accurate information acquisition from remote sensing data remains difficult in this region due to the complexity of the environment and the coupling between benthic reflectance and water column scattering. In this study, we developed a method to retrieve the diffuse attenuation coefficient (Kd), seafloor classification, and bathymetric maps by combining satellite laser altimetry and optical remote sensing imagery in shallow water areas. Firstly, the relationships between remote sensing reflectance (Rrs), water depth, and Kd were established based on radiative transfer theory. This method allows for the retrieval of Kd in shallow water regions, overcoming the limitations present in previous studies. Secondly, we eliminated the water column attenuation and obtained the bottom reflectance index (BRI). The BRI allowed us to determine the bottom reflectance and classify the seafloor using the Gaussian mixture model clustering method. This approach can effectively reduce the error in bathymetric inversion caused by variations in bottom reflectance. Finally, we developed a neural network model for bathymetric inversion. The model inputs consist of Rrs data and spectral shape data containing physical constraint information, aiming to achieve a robust estimation performance. We conducted the study in two experimental areas (the Bimini Islands and the Yongle Atoll) and compared the results with validation data to evaluate the algorithm performance. The results indicated an agreement between the estimated Kd and the validation data (inferred Kd490 values of 0.062m−1 and 0.058m−1, compared to a validation data range of 0.055–0.087m−1 and 0.059–0.070m−1, respectively). In addition, the seafloor classification accuracy was 86.74 % for the Yongle Atoll area. Finally, the neural network model accurately predicted the bathymetry in the two regions. The accuracy of the bathymetric maps improved significantly with seafloor classification, as indicated by reductions in root mean square error (RMSE) of 0.12 m and 0.15 m, and in mean absolute percentage error (MAPE) by 2.24 % and 5.87 %, respectively. Overall, the proposed method can be used to effectively decouple benthic and water column signals and accurately obtain Kd, bottom reflectance, and bathymetric information for shallow water environments, providing unprecedented information for assessing and monitoring ecosystems and facilitating further research.