Artificial neural network and physical based models for water-level forecasts ofinner Niger Delta in Mali

Abstract

The Niger Inner Delta (NID) is a wetland that was selected as an International Important Wetland under the Ramsar Convention (on February 1st, 2004) and can still be considered a hotspot of biodiversity in the Sahel. The Niger River is the main water source for the NID and is also used for urban life and irrigation. Therefore, the sustainable use of water to ensure environmental flow in the NID is under discussion. In this paper, the performance of different models established with empirical approaches (Artificial Neural Network and Regressions) or Conceptual Variable Source Area (Water Balance Method WBM) approaches were evaluated. The results of evaluation and validation based on determination coefficient (R2), Root Mean Squared Error (RMSE) and Nash-Sutcliffe Efficiency (NSE) show that all the models gave good results, however, the Levenberg Marquardt Artificial Neural Network (with 20 hidden neurons) was the best fit for the validation and testing periods.