Continental Estimates of Canopy Gap Fraction by Active Remote Sensing

Abstract

This study estimates canopy gap fraction (GF) from Geoscience Laser Altimeter System (GLAS) waveform data retrieved from a cross section of the Australian ecosystem, represented by 4 sites. However, as the literature supports, waveform profiles from systems such as GLAS require scaling in order to retrieve more accurate estimates of GF; the degree of scaling required is not a priori known, which has hindered large-scale estimates of GF by such systems to date. We employ spatially coincident GLAS footprints and airborne laser scanning (ALS) point cloud data from each site to scale initial estimates of GLAS GF to match ALS estimates, which allows the inference of scale factors to all footprints across the Australian forest landscape via the Random Forest (RF) technique. These refined estimates of GF are submitted to a second RF model to yield regional predictions of GF across the forest landscape. These predictions are validated against ALS data from 20 sites throughout Australia (R2 = 0.56, RMSE = 25.95%) and compared to MODIS, and also compared to ALS data (R2 = 0.13, RMSE = 66.46%). The developed method for producing large-scale estimates of GF shows promise for future studies and refinement, and is particularly pertinent as future spaceborne waveform systems come online (ICESat-2 and the Global Ecosystem Dynamics Investigation LiDAR).