HYDROLOGICAL MODELING IN SEMI-ARID ZONE: APPLICATION OF ARTIFICIAL NEURAL NETWORKS AND GR2M MODEL TO WADI ALLALA’S BASIN

Abstract

River flow results from the interplay of numerous variables for which quantitative information is not easily available. The study described in this paper has consisted of building models that are less demanding in terms of the number of explanatory variables and therefore more economical to simulate the flow rates of rivers in semi-arid zones. The proposed rainfall-runoff models belong to two different categories: black box models “artificial neural networks”, and conceptual models “the GR2M model”. In a practical way, this study will indeed make it possible to optimize and compare neuronal models and a global conceptual model (GR2M) for the simulation of the monthly flows of wadi Allala’s basin. The neural model was optimized with the Levenberg Marquarld (LM) with early stopping, while the GR2M model was optimized with method “step by step”. The Nash criterion (%) and Pearson correlation coefficient (R) were used to assess the performance of these models. For the neural models and GR2M model, the Pearson correlation coefficients (R) are higher than 0.70 at all stations. Nash criterion, it is generally above 66% for the two models (Neural Networks and GR2M). However, the neural models appear more efficient than the GR2M model.