Optimizing PlanetScope-Based Satellite-Derived Bathymetry: A Single-Band Approach with 3D Spatial and Statistical Filtering

Satellite-Derived Bathymetry (SDB) is an efficient method for mapping shallow water depths, particularly in large tropical areas that are difficult to survey using conventional techniques. This study optimises the single-band approach on PlanetScope imagery by integrating statistical filtering with 3D geospatial interpolation to improve accuracy and reduce prediction errors. Of the three spectral channels used (red, green, blue), the green band provided the best initial performance (R^2 = 0.211; RMSE = 2.080 m) and was selected for the basis predictive model advanced. Bathymetry reference data from Single Beam Echosounder System (SBES) surveys, reduced to Mean Sea Level (MSL), were used as reference data for model calibration and validation, with 30 points as controls. Optimisation involved applying Cloth Simulation Filtering (CSF) to the estimated depth point cloud, followed by spatial interpolation using Kriging, Inverse Distance Weighting (IDW), Spline, and Natural Neighbour methods. The results show that the combination of CSF and IDW provides the best performance with an increase in the correlation value of 0.462 (from R = 0.211 to R = 0.673), an increase in determination R^2 of 0.251, and a decrease in RMSE to 1.619 m. Additionally, using a smaller cloth resolution resulted in a more accurate model, indicating the high sensitivity of the CSF parameter to the quality of the final results. This study demonstrates that the integration of statistical filtering and geospatial interpolation can significantly improve the performance of single-band SDB method, offering an accurate and cost-effective solution for large-scale bathymetry mapping with improved accuracy.