Complex domain analysis of random volume over ground model for forest height

Abstract

This work focuses on the analysis of X-band Interferometric Synthetic Aperture Radar (InSAR) satellite acquisitions over forested areas in Estonia with the main goal of estimating forest tree height. The objective is to understand the behavior of InSAR coherence better particularly in the forest media and seek for suitable models to derive forest height (can be further linked to forest above ground biomass l) from InSAR measurements. The test area is located in Estonia, in the natural park of Soomaa. A set of 9 InSAR image pairs acquired by the German TanDEM-X satellite mission during 2012 are used in the study. The interferometric SAR processing has been done with SNAP (Sentinel Application Platform), a software designed by ESA (European Space Agency) for Sentinel and third-party missions. The focus is set on Random Volume over Ground (RVoG) and related models to describe the relation between forest parameters and InSAR coherence. In order to derive tree height from the models, complex domain inversion procedure is developed. The inversion results are compared with simplified model inversion and validated with airborne laser scanning data acquired by the Estonian Land Board. The results show robustness of Random Volume over Ground model for modelling the X-band complex coherence over forested areas. Developed tree height retrieval procedure shows an error between 26.7\% (RMSE = 3.8 m) and 30.2 \% (RMSE = 4.3 m) on a pixel-by-pixel basis.