Contribution à la modélisation pluie-débit dans les bassins versants de l’Est de l’Algérie.

Abstract

During the second part of the 20th century, Algeria experienced notable water stress, as well ascatastrophic floods, highlighting the urgency of more efficient water resource management,inparticular through the use of rainfall-runoff models to perform simulations and forecasts. In fact, dams' real-time operating instructions depend primarily on reservoir inflows and storage capacity. Forecasting these flows using models is the most efficient operating method, which makes it possible to anticipate which flows must be evacuated from the respective reservoir before a flood occurs. The objective of this PhD Thesis is to develop an event-based model that can accurately reproduce flood hydrographs in semi-arid climatic conditions where continuous data are not available in all regional basins. To this end, six potentially applicable rainfall-runoff models structures were tested and compared on an hourly time step, for five catchments in northern Algeria with different geomorphological characteristics. Two distinct approaches were used for the models' development. On the one hand, five production models were selected, namely the SCS-CN method (M1) associated to three levels of antecedent moisture conditions (AMC) and four methods derived from the SCS-CN method (M2, M3, M4, and M5), which are based on the incorporation of antecedent moisture amount (AMA). These production functions were coupled to both a base flow separation model, namely the digital filter model, and the NASH transfer model. On the other hand, a watershed model structure (MBV), from the SUPERFLEX framework, was adapted for an hourly time step and for event-based modeling. The models' parameters were calibrated using the same multi-criteria optimization procedure, the 'NSGA-II' method, which was combined with the TOPSIS method to select the best solution among the proposed alternatives. Model M1 was the poorest model in terms of its ability to predict flood volumes for all of the studied catchments, due to the discontinuous mathematical relationship linking antecedent moisture (P5) to AMC levels that induces sudden jumps in CN values. This led us to conclude that the concept of AMC, as formulated, is unsuitable for the Algerian semi-arid hydroclimatic context. Moreover, the MBV was considered inefficient due to its performance,and it needs further refinement in the integration of antecedent moisture before being considered a reliable model. The modified SCS-CN models produced satisfactory results. However, with the exception of the M3 model, each model was efficient for a specific criterion in a particular catchment. As a result, no ideal model fulfilling all criteria, for all basins, has been identified. Thus, the outputs of models M2, M4, and M5 were combined using the weighted average (WA) combination method, which reduced the weakness of each model and refined the results by achieving a reasonable balance between the performance criteria.