Spectral mixture analysis of airborne multispectral video images in mountainous terrain,kananaskis alberta:multi-scale scene fraction validation

摘要

Spectral mixture analysis has been used successfully to improve forest biophysical estimates using remotely sensed optical imagery in areas of flat,boreal terrain.In this paper,we address new issues in mixture analysis os mountainous areas using multi-scale image sets,toward the objective of creating a comprehensive regional scale monitoring strategy encompassing vast areas of both alpine and boreal forest terrain.These ideas have been tested in the Canadian Rockies using multispectral video camera imagery obtained July 1996 over Kananaskis Alberta at altitudes of 500,1000,2000 and 4000 feet.Image endmembers at 30cm spatial resolution were used in mixture analyses of sunlit canopy (aspen,pine) and shadow components at alll altitudes.Scene fraction validation at higher altitudes was performed against a reference classification image derived at the lowest altitude an interpreted to represent sub-pixel fractions at coarser scales.Good correspondence was found between scene fractions and quantitative assess ments.Future work will involve topographic corrections in areas of high relief,and the estimation of alpine forest biophysical variables such as leaf area index for input to models of terrestrial ecosystems and regional scale carbon flux.