Bistatic InSAR systems demonstrate significant potential for global and local-scale topographic mapping. However, the accuracy of the InSAR-derived digital elevation models (DEMs) is often compromised by residual errors caused by canopy volume scattering in forested areas. To mitigate these errors and enhance the topographic mapping capability of InSAR in forested regions, this paper proposes a method for retrieving subcanopy topography by integrating spaceborne LiDAR and bistatic InSAR data. The core innovation of this method lies in its modeling of the InSAR DEM residuals, which are decomposed into two components: a deterministic trend that is linearly related to InSAR observables, and a spatially correlated random error. Both components are simultaneously modeled and resolved within the KED interpolation process. Specifically, high-accuracy terrain control points (TCPs) are first extracted from global products of two spaceborne LiDAR missions of ICESat-2 and GEDI, using a rigorous filtering protocol. Building on the confirmed linear relationship between the DEM residuals at the TCP locations and two external covariates of InSAR coherence and terrain slope, KED interpolation is then applied to generate the final subcanopy topography. Evaluated across three representative study areas at different latitudes, the subcanopy topography derived from the proposed KED-based framework showed a significant improvement in accuracy, which exceeds 64%, compared to the original InSAR-derived DEM. Nearly unbiased and refined estimates were achieved in the high-latitude Genhe study area and mid-latitude TALL study area. Even in the challenging equatorial Lope study area, where results were limited by the quantity and accuracy of available TCPs, and although a non-negligible bias remained, the overall topographic accuracy was still improved by 65.4%.
Zhang et al. (Fri,) studied this question.