From a global to a semi-distributed approach in rainfall-runoff modelling

The question whether a distributed approach is more relevant than a lumped one for rainfall-runoff modelling has not yet been answered clearly. Most of the time, the problem is solved rather intuitively on a case by case basis. In this research, we attempted to give an objective answer to this problem, in order to progress in catchment modelling. Two complementary approaches were used. In the first approach, we used a ample of 15 gauged catchment triplets. We tried to assess whether it was interesting or not to partution a catchment in two subunits during the modelling process. The goodness of fit, as expressed by the Nash and Sutcliffe criterion, did not show a significant difference between bot modelling approaches. In the second approach, we used the same catchment sample and measured, for each subunit, physical characteristics that we assumed relevant for its hydrological behaviour. Differences observed in the subunits characteristics were quantified to obtain an index measuring the potential interest of distributed modelling. Both approaches gave indications about the interest of subcatchment. Therefore a comparison of results has been achieved in order to determine if it was possible to establish a link between the degree of heterogeneity and the success of a modelling type. This comparison has showed a relation between the studied features and results of the two types of modelling, global and semi-distributed. The choice of a modelling type is linked to the size of basins when these are classified by family size and to the degree of heterogenity. However, the reduced size of the sample doesn't allow us to bring about a definitive answer to this question. Other researches must be undertaken to extend the study to more catchments.