Hybrid ML Model Improves Forest Height Estimation from TanDEM-X and Landsat Data

A hybrid machine learning model integrating physical models and optical Landsat data has been developed to estimate forest height from TanDEM-X interferometric coherence measurements. The approach extends a previously proposed ML model by incorporating Landsat-derived features that provide complementary information on forest type or structure, addressing ambiguities related to height/structure and baseline/terrain slope. Validated on multiple TanDEM-X acquisitions over Gabon's Lopé National Park against airborne LiDAR measurements, the extended model achieved a 13.5% reduction in RMSE and a 16.6% reduction in error metrics. The research, published on arXiv (2605.20997), demonstrates the potential of combining radar and optical remote sensing with machine learning for improved geophysical parameter retrieval.