Estimation of forest cover type and structure from Landsat TM imagery using a canopy reflectance model for biomass mapping in western Newfoundland

Abstract

Geometric optical canopy reflectance models provide an explicit physical-structural basis to the analysis of satellite imagery and represent an alternative approach to existing classification methods for obtaining forest cover type and structural information (density and height) for biomass estimation. The Multiple-Forward-Mode (MFM) approach applied with the GOMS canopy reflectance model (MFM-GOMS) was tested for labeling clusters generated from an unsupervised classification as part of the EOSD project. A reasonable level of correspondence was found between model-based cluster labels and independent descriptors of surface cover, density and height. Errors were found to be less severe in most cases and due in part to the inherent variability of individual clusters comprised of multiple cover types, density ranges and height classes. The next phase of this work involves MFM-GOMS to obtain forest landcover and structural information for direct input to biomass estimation routines, thus not requiring prior cluster analysis and the associated confounding variability.