Sensitivity of PolSAR decomposition to forest disturbance and regrowth dynamics in a managed forest

Abstract

Managed forests, dynamic with activities of harvest and afforestation, have become an area of imminent research opportunity to understand the effects of these disturbances and regrowth dynamics. As low frequency polarized radar signals have the ability to penetrate forest canopies, the fully polarized Synthetic Aperture Radar (SAR) data gives an ideal representation of conditions in the forest. However, this valuable information has not been optimally utilized in understanding the forest disturbances and regrowth dynamics. Several forest change monitoring studies have been performed using single or dual polarization SAR data, whereas limited studies utilized full polarimetric information. In this study, L-band ALOS/PALSAR and ALOS-2/PALSAR-2 fully polarimetric data is used to study and understand the behavioural dynamics of homogeneous and heterogeneous forest compartments having single and multiple species composition, respectively. Radar scattering pattern of different species, and its variation with respect to phenology, disturbances, harvesting and plantation regrowth are analysed to understand the trend in scattering components with respect to above said forest changes. Based on the scattering properties and their analyses, a two stage rule based classifier has been developed to categorize the forest based on its disturbance and regrowth status into disturbed, re-growth, stable/no change and non forest classes. The classifier performed with, considerably, high accuracy having an overall accuracy of 85.39% and a kappa coefficient of 0.8. The classifier developed in this work presents the potential of fully polarimetric data in understanding the level and extent of disturbances and regrowth dynamics of a forest.