Inversion of a vegetation reflectance model with NOAA AVHRR data

We invert a bidirectional reflectance model with NOAA AVHRR data collected over a mixed grassland during the First ISLSCP Field Experiment (FIFE). Leaf area index (LAI) and leaf optical properties are accurately retrieved in one-parameter inversions. In two-parameter inversions, the accuracy of the retrieved parameters is coupled: LAI is accurately retrieved only when leaf optical properties are accurately retrieved, and vice-versa. Since inversion accuracy depends on the sampling geometries of the reflectance data, we also develop a “derivative weighting” scheme for the merit function. This scheme causes inversion solutions to be preferentially determined by data containing the most information about the model parameters. We show this scheme increases the gradients of the merit function such that more rapid and accurate inversions are possible. We also use derivative weights to select the most promising data subsets for inversion. This study confirms the operational potential of model inversions with inexpensive, widely available satellite data. Moreover, our methods can be used with future remote sensing systems such as EOS MODIS, MISR, and POLDER.