Analysis of Bidirectional Reflectance Effects on 20-years of AVHRR NDVI Data, for Agricultural Regions in

The Normalized Difference Vegetation Index (NDVI) provides synoptic information on environmental conditions in agricultural regions such as in Ukraine and Russia. However, past studies have raised questions whether the Maximum Value Compositing (MVC) method, used to produce time-series of NDVI, is affected by Bidirectional Reflectance (BR), as a result of view geometry, possibly leading to erroneous NDVI values. This study uses Angular Dependence Models (ADM), derived from the NOAA/NASA Pathfinder AVHRR Land (PAL) data set, to correct PAL Channels 1 & 2 reflectance for BR. MVC is used to produce ten day NDVI composites, from both the pre-BR correct channels and the BR corrected channels. Comparisons of the two NDVI time-series show a slight increase in NDVI for the BR corrected product, but not a significant difference. Further analysis indicates most of the BR affects on NDVI values are a result of the atmosphere, this being an artifact of the PAL data set not being corrected for aerosols or water vapor. The atmosphere causes lower NDVI values off-nadir and as a result the MVC tends to compensate for it, as demonstrated in some past model studies. However, results also suggest that MVC, for pre-BR corrected NDVI, is susceptible to selecting off nadir pixels in regions with high cloud cover, resulting in lower NDVI values for those regions