Thematic mapper analysis of tree cover in semiarid woodlands using a model of canopy shadowing

The effect of variable tree canopy cover, shadows, and topography on Thematic Mapper (TM) data was examined for blue oak (Quercus douglasii) woodland and wooded grassland on rugged terrain in central California. Thematic Mapper data were analyzed from September 1986 (trees in leaf, grass understory senescent) and December 1984 (trees leafless, grasses emerging). Tree cover measured in 84 60 m × 60 m field plots was strongly negatively correlated with TM radiances (excluding Band 6) on both dates, and was positively correlated with the Normalized Difference Vegetation Index (NDVI) in September. TM Band 3 was most highly correlated with canopy cover. The Li-Strahler geometric-optical canopy reflectance model was used to predict the projected cover of shadow cast by the trees based on crown and illumination geometry. The spectral model was aggregated to two components; figure (tree plus shadow) and background (understory), and the pixel radiances modeled as a function of the cover of the figure. Simulated shadow cover was high at intermediate cover values, and was nonlinearly related to canopy cover. Modeled figure cover in the sample plots at the time of the satellite overpasses was more highly correlated with TM Band 3 than measured tree cover in both September (regression r² of 0.76 vs. 0.69) and December (r² of 0.59 vs. 0.53). Slope illumination explained an additional 11% of the variance in the December data. The remaining unexplained variance probably resulted from spatial variation in understory and canopy reflectances.