FOREST TEXTURAL PROPERTIES FROM SIMULATED MICROWAVE BACKSCATTER - THEINFLUENCE OF SPATIAL-RESOLUTION

To explore the effects of radar scene degradation on the interpretatio n of forest pattern, we used a model to simulate L-band (23 cm) backsc atter (VV, HV, and HH, polarizations) for a range of resolutions (5 m, 10 m, 12.1 m, 15 m, and 24.2 m). Simulated pixels were parameterized with tree density and height for each tree (> 3 cm dbh) covering a 200 m x 150 m area from an immature, conifer-dominated stand in the North ern Experimental Forest located near Howland, Maine. Airborne syntheti c aperture radar (AIRSAR) imagery from the stand was used to evaluate the ability of the model to predict single pixel and textural patterns at 12.1 m resolution. Although the model was unable to accurately pre dict pixel-by-pixel responses, probably due to registration error, tex tural patterns from the HH polarized image were correlated to the simu lated patterns. The model's assumption Of randomly distributed sterns within each pixel was another probable source of error as trees in thi s stand were aggregated at scales from 1 m to 50 m. Despite model inad equacies, significant pattern was found at all simulated backscatter r esolutions for all polarizations for most quadrat sizes (< 10,000 m2), but only for smaller quadrat sizes (< 6,000 m2) for within-pixel stru ctural properties (i.e., stem, branch, and leaf biomasses, stem densit y, and average height). This suggests that radar textural information resulting from forest heterogeneity should be available at scales coar ser than 24.2 m for forests of similar structure.