ON THE POTENTIAL OF SATELLITE DERIVED VEGETATION PHENOLOGY FOR WATERSHED NUTRIENT MODELING: A NEURAL NETWORK APPROACH

To our knowledge, this study is the first to attempt to build a model that can rely on a time series of remotely sensed vegetation indices (VIs) for predicting the dynamics of water-phase total phosphorus (TP) concentration. We examined the possibility of using four literature based VIs; enhanced vegetation index (EVI), normalized difference vegetation index (NDVI), greenness fraction vegetation index (GFVI), in addition to two indices proposed in this study (SRVIm and GFVIm); to provide sufficient landscape information for water-phase TP modeling. While EVI-based indices were seen to successfully represent landscape phosphorus dynamics, NDVI-based indices were slightly lower in their performance. EVI and its normalized transform, GFVIm were considered as excellent inputs to an artificial neural network (ANN) model for TP predictions in a primarily forested landscape. However, their corresponding NDVI-based indices—NDVI and its normalized transform, GFVI—were considered as adequate inputs for TP modeling, especially in times when EVI values are not available since NDVI has more than 20 years of historic data. Such ANN models can potentially serve as a valuable tool for simulating the impact of different watershed harvesting activities on water quality parameters.