Forest stand height estimation using inversion of RVoG model over forest of north-eastern India

Abstract

In recent years, multiple studies have been carried out to estimate forest height using Synthetic Aperture Radar (SAR) remote sensing techniques. The forest height is an essential forest resource parameter that is usually used in biomass estimation contributing to carbon sequestration studies. Polarimetric Interferometric SAR (PolInSAR) is a remote sensing technique that combines SAR Polarimetry (Pol- SAR) with SAR Interferometry (InSAR) and has demonstrated tremendous ability for forest height extraction as it is sensitive to the vertical arrangement of the scattering media. SAR Interferometry technique uses two complex SAR images from the same area obtained from two different orbit positions to create an interferogram, whereas PolSAR provides the full scattering matrix at each resolution cell. In this paper, we examine the Random Volume over Ground (RVoG), the polarimetric canopy scattering model, for the forward modeling and the three stage inversion (TSI) for the retrieval of vegetation stand height. This inversion also estimates other forest parameters such as canopy extinction, ground-to-volume amplitude, and ground topographic phase. We investigate the performance of the inversion algorithm for forest height estimation using single baseline L-band ALOS-2 PALSAR data collected on December 17, 2018, and December 31, 2018, over the Saipung Reserve Forest, Meghalaya, over North-Eastern India. The accuracy of the forest height estimate is assessed using extensive ground truth data collected during a week-long expedition to the Saipung reserve forest during December 2018. Correlation between the field measured forest height and the estimated tree height using TSI technique is 0.81 with RMSE of 5.05 m. The study suggests that the PolInSAR technique has significant potential for retrieving forest biophysical parameters such as stand height, and can substantially augment the biomass estimation over areas like reserve forests where accessibility for field data collection remains difficult.