Contribution to the prediction of drought in the basin Inaouène using RNA coupled to the stationary wavelet decomposition

Abstract

In this article, based on the coupling of stationary wavelet transform (SWT) models and artificial neural networks (ANN) Multi-layer perceptron (MLP) have been developed for the prediction of drought in terms of SPI values (Standardized Precipitation Index) for various prediction time (1, 3 and 6 months). The relative performance of these models (SWT-ANN-MLP) was compared to the traditional model of artificial neural network (ANN). The variables used to develop and validate the models were, monthly precipitations, the North Atlantic Oscillation Index (NAO) and the SPI values calculated from monthly rainfall. Both meteorological stations Idriss 1 and Bab Marzouka located in the basin of Inaouène in northern Morocco were selected for this study. The SWT-ANN-MLP models that have been proposed to provide drought forecasts of current and long term are more efficient compared to ANN models.
Key Words: wavelet analysis; stationary wavelet transform, artificial neural network; wavelet network model; prediction of drought, SPI index, Inaouène basin.