Artificial neural networks for monthly precipitation prediction in north-west Algeria: a case study in the Oranie-Chott-Chergui basin

Abstract

Abstract The north-west region of Algeria, pivotal for the nation's water resources and agriculture, faces challenges from changing precipitation patterns due to climate change. In response, our study introduces a robust forecasting tool utilizing artificial neural networks (ANNs) to predict monthly precipitation over a 12-month horizon. We meticulously evaluated two normalization methods, ANN-SS and ANN-MM, and assessed four distinct approaches for selecting input variables (no selection, ANN-WO, ANN-CO, and ANN-VE) to optimize model performance. Our research contributes significantly to the field by addressing a critical gap in understanding the impact of evolving precipitation patterns on water resources. Among the innovations, this study uniquely focuses on medium-term precipitation forecasting, an aspect often marginalized in previous research. Noteworthy outcomes include correlation coefficients of 0.48 and 0.49 during the validation phase, particularly with the Endogen variables and correlation-optimized models using Min-Max normalization. Additionally, the Min-Max normalized technique achieves an impressive 67.71% accuracy in predicting the hydrological situation based on the Standard Precipitation Index.