Estimating Conifer Forest Cover with Thematic Mapper Data Using Reflectance Model Inversion and Two Spectral Indices in a Site with Variable Background Characteristics

This investigation tests the inversion of a geometric–optical forest reflectance model and the utility of two spectral indices (NDVI and SAVI) for estimating crown cover in a conifer forest site, using Landsat Thematic Mapper (TM) spectral data. Emphasis was placed on incorporating the effects of variations in the background signatures into the inversion process. The inversion methodology and the sources of error were tested using simulated reflectance data. One-parameter inversions using constant background signatures could not retrieve stand cover because a significant proportion of the total variance in the spectral data resulted from background variations. The spectral indices performed better than the one-parameter inversions because the indices partially reduced background effects, but could not provide estimates of cover. When the background signatures were varied in three- and four-parameter inversions, inversion performed better than the spectral indices, giving estimates of cover for the test stands. The biophysical basis of the results are examined. The results indicate that variations in the background signatures can be included in the inversion in certain situations, and should be included when these effects are important. Inversion may provide a more robust method for estimating forest characteristics from multispectral satellite data; however, further testing is needed.