Large-Scale Forest Height Estimation With TanDEM-X InSAR and Sparse LiDAR Based on LSC and Block Adjustment Model

The TanDEM-X bistatic Interferometric Synthetic Aperture Radar (InSAR) system has great potential for forest height estimation. However, single-polarization InSAR data cannot support a uniquely determined inversion of classical physical models due to insufficient observational information. Simplified models make inversion possible using only interferometric coherence amplitude, at the cost of reduced performance by uncertain semi-empirical parameters and a loss of adaptivity to variations in local forest properties. To address this issue, this letter proposes a framework based on a semi-empirical physical model, block adjustment model, and the least-squares collocation (LSC) method. The framework utilizes sparse spaceborne LiDAR canopy height as a reference and selects tie points in overlapping areas to establish constraints, enabling robust estimation of semi-empirical model parameters on a large scale. Subsequently, variance-covariance estimation is employed to develop an LSC-based local calibration method, mitigating residual inversion errors caused by variations in forest attributes and terrain slope within local areas. Validation at two test sites in the USA and Spain shows that, compared to the results obtained from the SINC model and semi-empirical model inversions, the proposed framework improves accuracy by more than 32% and 21%, respectively, highlighting its potential for large-scale forest height mapping with TanDEM-X InSAR and sparse LiDAR data.