Mapping Tree Canopy Cover and Canopy Height with L-Band SAR Using LiDAR Data and Random Forests

Abstract

The aim of this paper is to systematically combine complementary LiDAR and synthetic aperture radar (SAR) observations to map tree canopy cover in a boreal forest. LiDAR data can provide direct measurements of vegetation structures but are limited by the sparse spatial coverage of observations. SAR systems can perform wall-to-wall high-resolution mapping without weather constraints but the information about vegetation and ground subsurface are mixed in the backscatter data. In this paper, we adopted the Random Forests algorithm to train an upscaling function using tree canopy cover (TCC) and canopy height model (CHM) derived from Goddard's LiDAR, Hyperspectral and Thermal Imager (G-LiHT) point cloud data. The regression model was then applied to the L-band Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR) data acquired during the 2017 Arctic-Boreal Vulnerability Experiment (ABoVE) airborne campaign to map the TCC and CHM over the Delta Junction area in interior Alaska.