A modeling approach for studying forest chlorophyll content

Imaging spectroscopy from space is a potentially powerful tool for assessin g vegetation chemistry with approaches that rely either on empirical relati onships or on the inversion of reflectance models. However, this assessment can be erroneous if the 3-D spatial distribution of the vegetation is negl ected. Sophisticated radiative transfer models are often required to accoun t for the 3-D canopy architecture. Due to long computation times, however, these models are not well adapted to sensitivity analyses and numerical inv ersions that require hundred of calls of the merit function. This paper pre sents a methodology developed to stimulate vegetation reflectance spectra q uickly and accurately (i.e., without neglecting the 3-D canopy architecture ). Canopy reflectance spectra are calculated by linearly interpolating spec tra 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 l eaf chlorophyll concentration was characterized by the position of the infl ection 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). Res ults demonstrated the strong influence of canopy architecture. For example, the lambda(1) has larger values for mature stands for pole stands (delta l ambda(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 neglec ted. (C) Elsevier Science Inc., 2000.