Land-cover classification and forest biophysical retrieval from SAR

Orbital SAR data can be used to generate high accuracy classifications of landcover. The existing single-frequency and singly polarized SAR systems provide marginal accuracy when used alone unless a multitemporal data set is acquired. Multi-frequency archival data yield good results that are robust over a number of ecoregions for a very simple land-cover categorization. SAR data complements electro-optical data. Fusion of these two spectral regimes provides superior classification. The next generation of orbital SAR systems now in construction and design will provide multi-polarized and multi-frequency capabilities that will significantly improve land-cover classification. The retrieval of forest biophysical properties (average tree height, basal area, aboveground biomass and timber volume) can be accomplished using orbital SAR data. The best results use a multi-frequency SAR with polarimetric capability at a long wavelength (such as L-band). LightSAR should yield biophysical estimates that are comparable in accuracy to those obtained by traditional field methods.