Agricultural drought forecasting using satellite images, climate indices and artificial neural network

Abstract

A new model is proposed for agricultural drought forecasting based on normalized difference vegetation index (NDVI), which is based on satellite data, using effective climatic signals and artificial neural network (ANN). The applied ANN is a feedforward multiple neural network. The inputs of the model are the climatic signals Southern Oscillation Index (SOI) and North Atlantic Oscillation (NAO). In order to forecast NDVI with ANN, the normal method (NM) was used for the recent period, and for evaluation, the moving window method (MWM) was used for a longer (18 years) period. This model was applied to Ahar-chay Basin in Azerbaijan Province, which is located in the northwest of Iran. The results show that in spring (May, June and July (MJJ)) synthetic NDVI can be predicted using ANN, with the input of SOI and NAO indices of the preceding (1 year) spring period. The determinant coefficient (R ²) between observed and predicted NDVI is 0.79, the root mean square error (RMSE) is 0.011 and the discrepancies are less than 1 SD.

Acknowledgements

The data used by the authors in this study include the data produced through funding from the Earth Observing System Pathfinder Program of NASA's Mission to Planet Earth in cooperation with NOAA. The data were provided by the Earth Observing System Data and Information System, Distributed Active Archive Center at Goddard Space Flight Center, which archives, manages and distributes this dataset. We thank an anonymous reviewer and the editor for their critical and constructive remarks.