*4.3. Inversion Models*

The results of this section show that the conclusions are different for the NDVI correction than for the individual bands. For the red and NIR bands, the Stable model proved to be the best method to use when the data's time series has too much noise to derive the correction parameters from it. This is true especially for AVHRR-pre, for which the Stable method provided a significant improvement in the noise correction. This is, therefore, the recommended model for the derivation of surface albedo using AVHRR-pre and AVHRR data, provided that the NDVI computation is not required. For MODIS data, the use of the MODIS inversion is still preferable, as assumptions made by the models devised introduce uncertainty. In the case of the NDVI, however, the use of the regular VJB model using parameters derived from the same data is preferred. For users with a large computational burden that look for a compromise between computational time and accuracy, the use of the Average method provides very fast correction parameters at the expense of ~10% noise improvement. If the BRDF correction parameters are already available from MODIS data or another sensor with little atmospheric perturbation at the same spatial resolution, the use of these parameters provides the second best correction after the Stable method with virtually no significant computational time required in comparison.
