1Institut National Polytechnique Félix Houphouët-Boigny (INP-HB), Department of Earth Sciences and Mineral Resources (STeRMi), Po Box 1093, Yamoussoukro (Côte d'Ivoire); Home phone: (225) 30 644 897, Mobile phone: (+225) 07 492 712
2University of Cocody-Abidjan, Laboratory of Science and Technology of Water and Environment (LSTEE), Department of Earth Sciences and Mineral Resources (STRM), 22 Po Box 582, Abidjan-22 (Côte d'Ivoire); Work phone: (225) 22 483 803, Mobile phone: (+225) 07 271 713
*Email: michel.a_kouassi@yahoo.fr
Online published on 26 November, 2013.
This study presents a comparison between two models of the rainfall-runoff transformation on an annual basis: a conceptual model and an artificial neural network (ANN). Both models are applied to three watersheds of the N'zi River (Bandama) in Côte d'Ivoire. The comparative analysis is based on the performances of simulation in terms of Nash-Sutcliffe criterion. The models have been tested on two periods, a dry (1973–1997) and a wet one (1961–1972). The input data of the two models are the rain and the potential evapotranspiration to annual time step. The main results of this work show that the performances of both models (conceptual and neuronal) remain satisfactory in general with Nash-Sutcliffe criterion higher than 60%. These models appeared also robust and suitable for the simulation of the annual flows of rivers. The comparison of the two models has showed that the neural network performed significantly better than the conceptual model.
Rainfall-runoff modeling, conceptual model, artificial neural network, N'zi-Bandama, Côte d'Ivoire
