Answers are grounded in this paper and cite their sources.
Photon-counting light detection and ranging (LiDAR) observations from Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) enable along-track bathymetry in remote reefs and data-sparse coastal waters, but solar background radiation, water-column scattering, and depth-dependent attenuation make underwater terrain photons sparse, unevenly distributed, and mixed with nonterrain returns. Fixed-orientation or single-scale filters may therefore fragment terrain-photon sequences over reef slopes and locally undulating seabeds. This study proposes MSAE-Ph, a local slope-guided terrain-photon extraction method based on multiscale seed points. Following water-surface detection using kernel density estimation and the subsequent separation of candidate underwater photons, principal component analysis is used to estimate local terrain orientation for slope-aligned photon transformation. Small- and large-window density peaks are fused using a signal significance index to form a scale-stabilized terrain seed skeleton, which guides density-adaptive elliptic extraction under vertical geometric constraints. Experiments in four island and reef study areas showed that MSAE-Ph maintained terrain tracking across complex slopes, variable candidate-photon densities, and weak bottom-return conditions. Independent validation of the MSAE-Ph-derived bathymetry from five Oahu tracks against the NOAA airborne LiDAR DEM yielded an R2 of 0.995, an RMSE of 0.504 m, an MAE of 0.386 m, and a bias of 0.304 m over the 0–30 m depth range. These results indicate that MSAE-Ph can improve the continuity of ICESat-2 underwater terrain-photon extraction under the examined shallow-water conditions.
No takes yet. Share an insight, caveat, or question.
Luan et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: