Sub-canopy Terrain Retrieval from ICESat-2 Data Assisted by Sentinel-2 Optical Imagery

In densely vegetated regions, the spaceborne laser altimeter ICESat-2 often struggles to acquire sufficient ground photon returns, leading to reduced accuracy of sub-canopy terrain retrieval. To address this limitation, a novel inversion approach is proposed based on the extrapolation of average canopy height from sparse vegetation areas. First, photon cloud is filtered to eliminate background noise. Vegetation cover parameters are then derived from Sentinel-2 optical imagery, allowing for classification of the study area into sparse and dense vegetation areas. In dense areas, the average canopy height obtained from sparse areas is utilized to reconstruct the sub-canopy terrain. The method was validated using three ICESat-2 datasets over the HARV forest region in the United States. Comparative analysis with ATL08 demonstrates that the proposed algorithm achieves superior performance, reducing in unusable terrain points ratios (Inv+Res>3) by 24.1%,22.2%,15.1%, respectively. These results confirm the effectiveness of integrating optical imagery and spaceborne LiDAR data for accurate sub-canopy terrain retrieval.