Abstract
Digital Compact Airborne Spectrographic Imager (CASI) data and SPOT satellite HRV multi-spectral imagery were compared in their relationship to individual tree and forest parameters in fourteen lodgepole pine (Pinus contorta) stands in western Canada. Volume estimates derived from DBH and height observations in field plots were clustered to produce natural stands which were studied in the remote sensing data. Spatial analysis of the aerial imagery was undertaken with a combination of filtering and first-order texture derivatives. The range, measured by image semivariograms calculated over the stand, was used to 'customize' window sizes. The unaltered red CASI band was found to be useful for extracting parameters which are expressed at the individual tree level; texture variables were more important in explaining stand-level variations over the spatial domain. Mean filters were found to be of little use except in the prediction of canopy coverage. Models based on the SPOT HRV variables were found to be poorer estimators of forest stand parameters than similar spectral resolution but higher spatial resolution CASI data. Discrimination of seven volume classes of lodgepole pine was accomplished with approximately 75 per cent accuracy overall.