A novel bathymetric mapping framework integrating indirect inversion of ICESat-2 and multi-source remote sensing data

Satellite-derived bathymetry (SDB) plays a critical role in coastal zone management, navigation safety, and marine environmental monitoring. However, conventional SDB methods are constrained by limited depth penetration and reduced accuracy, largely driven by water optical properties, environmental variability, and sensor limitations. To address these challenges, this study proposes a novel bathymetric mapping framework that integrates wave-based indirect depth inversion from photon-counting LiDAR data with multi-source remote sensing data. Specifically, a novel Progressive Adaptive Window for Local Period (PAWLP) algorithm is developed to derive water depth from ICESat-2 surface waves. By dynamically adjusting the analysis window to local wave variations, PAWLP enhances inversion robustness based on linear wave theory. In addition, we construct a multi-source SDB random forest inversion model by fusing multispectral imagery, synthetic aperture radar (SAR), tidal height, and tidal velocity. To further improve model generalizability and reduce scene-specific noise, a temporal sample transfer strategy is applied. In this study, the proposed methods are validated using in situ bathymetry data from the U.S. Virgin Islands (clear waters) and from Bar Harbor (turbid waters). Results show that PAWLP adaptively captures local wave characteristics to retrieve water depth, achieving an average root mean square error (RMSE) of 1.56 m and weighted mean absolute percentage error (WMAPE) of 10.01 %, with reductions of approximately 17.89 % in RMSE and 19.46 % in WMAPE compared to the fixed-period method. The proposed multi-source bathymetric inversion framework further improves prediction accuracy, achieving RMSEs of 1.64 m in clear waters and 2.32 m in turbid areas, outperforming traditional methods across diverse conditions. The integration of SAR data and tidal features substantially enhances prediction stability, particularly under optically complex waters. Overall, this study highlights the potential of wave-based indirect depth inversion to extend the effective depth range for SDB. By integrating ICESat-2 bathymetric measurements with multi-source remote sensing data and temporal sample transfer strategy, our method enhances mapping accuracy and spatial coverage, mitigates optical saturation effects, and provides a scalable solution for reliable bathymetric mapping across diverse coastal environments.