InSAR-Based Tree Height Estimation of Hilly Forest Using Multitemporal Radarsat-1 and Sentinel-1 SAR Data

Abstract

This article presents estimation and validation of tree heights extracted from synthetic aperture radar interferometry (InSAR) digital elevation model (DEM). InSAR technique was used to generate DEM in terms of top of the canopy digital surface model (TCDSM). Multitemporal Radarsat-1 InSAR data pair of February 12, 2004 and March 7, 2004 and Sentinel-1 data pair of December 9, 2017 and December 21, 2017 were used. The TCDSM values at ground truth locations were subtracted from the DEM obtained at forest gap areas (i.e., forest floor elevation). The satellite-derived TCDSM-based tree height was compared with the ground observed data. For this purpose, an experiment in Tundi Reserved Forest, Dhanbad, Jharkhand, India was conducted using global positioning system and altimeter. The study area is a hilly forest tract dominated by Shorea robusta (Sal) with average height of 12 m. Total 275 numbers of trees were selected at 11 locations in different elevation strata. Ground truth locations were also selected on the basis of variability, tree height categories, and accessibility in the area. It was noticed that mean absolute error (MAE) and root-mean-square error (RMSE) between in situ and estimated heights for Radarsat-1 were 1.48 and 1.53 m, respectively, whereas for Sentinel-1 in 2017, MAE and RMSE were 1.3 and 1.34 m, respectively, with strong correlation of >80%.