A modeling approach for studying forest chlorophyll content

Imaging spectroscopy from space is a potentially powerful tool for assessing vegetation chemistry with approaches that rely either on empirical relationships or on the inversion of reflectance models. However, this assessment can be erroneous if the 3-D spatial distribution of the vegetation is neglected. Sophisticated radiative transfer models are often required to account for the 3-D canopy architecture. Due to long computation times, however, these models are not well adapted to sensitivity analyses and numerical inversions that require hundred of calls of the merit function. This paper presents a methodology developed to stimulate vegetation reflectance spectra quickly and accurately (i.e., without neglecting the 3-D canopy architecture). Canopy reflectance spectra are calculated by linearly interpolating spectra pre-computed with a coupled model: a 3-D canopy model (DART) and a leaf optical properties model (PROSPECT). This approach was successfully tested by studying the influence of forest architecture on the determination of leaf chlorophyll concentration was characterized by the position of the inflection point of the red edge (lambda(1)). Apart from Chl(f), we considered four other influential factors on lambda(1): the LAI (leaf area index), the viewing direction, the understory reflectance, and the canopy architecture (i.e., a theoretical turbid medium, a pole stand, and a mature stand). Results demonstrated the strong influence of canopy architecture. For example, the lambda(1) has larger values for mature stands for pole stands (delta lambda(1)>10 nm), whatever the LAI and the viewing directions. Thus, errors on Chl(f) can be larger than 23 mu g/cm(2) if canopy architecture is neglected. (C) Elsevier Science Inc., 2000.