Utility of reflectance model inversion versus two spectral indices for estimating biophysical characteristics in a boreal forest test site

In this investigation, the inversion of a forest reflectance model (FLIM) was tested, along with two spectral vegetation indices (NDVI ans SAVI) for a boreal forest site, using Landsat Thematic Mapper (TM) data. At the measurement plot scale, model inversions could not provide accurate information on cover or crown transmittance (a function of leaf area index, LAI) using only single-date TM data. A primary limitation was due to variations in other stand characteristics that could not be taken into account in the inversion process; including the background spectral signatures, stand structure, and spatial distribution of trees. A single value of a given parameter derived from inversion was associated with multiple values of that parameter as measured on the ground. The remote sensing derived parameters were grouped into classes based on homogeneous plot characteristics. From analysis of variance of the grouped data, the coverages derived by inversion, the NDVI, and the SAVI were found to be related to measured coverage at the 0.01 level. Crown transmittances derived from inversion were related to measured transmittances at the 0.01 level, although the indices were not found to be related to measured transmittance. The results indicate that the FLIM performs in a similar way to a physically based, nonlinear spectral index. Further work is needed to develop methods for incorporating additional inventory and remote sensing information into the inversion process. Work is also needed to investigate methods to relate the biophysical information derived from remote sensing to routine forest inventory variables, which are required in forest management. (C) Elsevier Science Inc., 1999.