Résumé
The objective of this study is to apply the rainfall-runoff modeling, for the reconstruction of flows and the forecasting of hydrological risks in two sub-catchments (Bounamoussa and Kebir-Est) located in El-Taref, North-East of Algeria. Considering the geomorphological and climatic characteristics of the study region, redundant and catastrophic floods have been recorded. The available data cover a period of 39 years (1981–2019). The GR2M model (GR at monthly time step) is implemented by using two R packages; airGR and airGRteaching. These tools enable the automatic affectation of the parameters that control the two reservoirs of the GR2M model, X1 (production function) and X2 (routing function). This modeling is optimized by the parsimonious character of the model; requiring few monthly input data: rainfall, temperature and potential evapotranspiration. As for its robustness, it is provides by Michel's algorithm. The evaluation of its efficiency is determined, for the calibration and validation periods, by the Nash–Sutcliffe (NSE), Killing-Gupta (KGE) and modified Killing-Gupta (KGE′) criteria. The airGR package gives NSE (Q) and KGE (Q) values greater than 80% during the calibration period and greater than 86% during the validation period. For the airGRteaching package, the KGE′(Q) values vary from 89 to 92% for both periods. Production and routing reach values successively: (99.50 mm, 36.22 mm) in the Bounamoussa sub-catchment and (125.12 mm, 33.70 mm) in the Kebir-Est sub-catchment. Thus, this study shows the capacity of this model to correctly reproduce the flows of this basin, although heterogeneous.