UTILISATION DES MULTI-MODELES POUR AMELIORER LA MODELISATION HYDROLOGIQUE REGIONALISEE ET APPLICATION AU BASSIN DE CHELIFF

Abstract

Hydrologists around the world face many difficulties that complicate the task of hydrological modeling for water resources management. Our main objectif through this work is to explore and propose solutions to improve regionalized hydrological modeling in the Cheliff basin by comparing two widely used models, HEC-HMS (Hydrologic Engineering Center-Hydrologic Modeling System) and SWAT (Soil and Water Assessment Tool), studying their sensitive parameters and studying the regionalization of ungauged basin parameters. For this, we have chosen the Upper and Middle Cheliff basin as a study area, using twenty-six rainfall stations in winch eight hydrometric stations covering a period of six years from 2007 to 2012. For the exploration of regionalization methods, the calibration and validation of the HEC-HMS model for all sub-basins showed a satisfactory performance slightly better in the validation phase. NSE varies between 0.61 and 0.79. R 2 is between 0.62 and 0.79. RMSE is stable around 0.50 to 0.60 in the calibration phase. The regionalization results show that the spatial proximity method offers slightly better performance than the physical similarity, with NSE values exceeding 0.75 during validation. The arithmetic mean of the parameters was found to be more effective than the inverse distance weighting. For the model comparison and model sensitivity part, three rainfall stations and one hydrometric station were selected for the Oued Tikzal sub-basin over a ten-year period from 2003 to 2012. For HECHMS, the parameters related to basin response time (Lag time), base flow (Recession Constant) and canopy storage capacity (Max storage) were found to be the most sensitive. For SWAT, the parameters related to runoff (CN2), hydraulic conductivity (CH_K2), average slope length, (SLSUBBSN), depth of water threshold in the shallow aquifer required for water return (GWQMN), soil evaporation compensation factor (ESCO) and base flow alpha factor (ALPHA_BF) showed significant influence on the results. HEC-HMS offers better performance in terms of accuracy and reliability, with NSE values of 0.68 for HEC-HMS and 0.51 for SWAT in the calibration phase, and 0.71 for HEC-HMS and 0.60 for SWAT in the validation phase. This work demonstrated that the use of multi-models and regionalization techniques can significantly improve hydrological modeling in semi-arid regions, particularly in the Cheliff basin.