Modelisation de l'humidite du sol du bassin versant de la riviere des anglais avec donnees du reseau Mesonet-Montreal.

摘要

Even if evaluating the potential hydrological impacts due to the apprehended climate change still is unsure, most scientists agree that the latter may have significant consequences regarding to water resources management or even flooding. Therefore, improving rainfall-runoff models' accuracy and in particular their performance in flood prediction represents today a major key point for most hydrologists. Thus, this work, investigating more specifically the Chateauguay river basin, aims first to present different approaches consisting on assimilating soil moisture data throughout the hydrological simulation process and then to point out whether those methods are advantageous enough to be worth handling. Indeed, using such a strategy enables to explicitly take into account the soil's hydric state as well as the whole dominating factors in the processes of the transformation of rainfall into flow arguing that it could significantly improve the predictions' accuracy. In this way, the assimilation of the soil moisture data was made into the model "des Anglais" built by the DRAME laboratory while the entire data were provided by the Mesonet-Montreal network.; Four different events and four infiltration methods as well were used in that study. The first approach used the Antecedent precipitation index (API) in order to describe the soil moisture conditions over time during and after a storm event. Then, the transition to the second method showed significant improvements regarding the 1st and 4th event and according to the Nash criterion, revealing all the benefits brought by integrating soil moisture data throughout the hydrological simulation process. The transition to the 3rd method consisted on pointing out all the advantages related to the use of a distributed model and showed meaningful improvements regarding the 1st event but also deteriorations regarding the others. Finally, the last method consisted on an updating process of the soil parameters which led to the deterioration of all events. The analysis of those results demonstrate how important it is to pay attention to the observed data quality and to the way to define the events' inputs while incorporating soils' data over larger depths should be considered as a major key to refine the model's accuracy.