Impact of spatial aggregation of inputs and parameters on the efficiency of rainfall-runoff models: A theoretical study using chimera watersheds

Abstract

This paper examines the respective merits of aggregative and disaggregative approaches in hydrological modeling and aims at determining how to identify the most appropriate level of spatial distribution in rainfall-runoff modeling. The lumped approach is compared with two types of semidistributed approaches to assess the relative importance of rainfall and parameter distribution on modeling results. In order to base these comparisons on a large number of definitely heterogeneous basins, we introduce chimera watersheds. Chimera watersheds associate two actual watersheds of similar size to constitute a third, highly heterogeneous virtual basin, where the knowledge of spatialized flows makes it possible to test disaggregated approaches and to compare them to the lumped approach. We show that the greatest portion (two thirds) of improvement contributed by spatial distribution comes from accounting for rainfall variability. If spatial distribution is considered to be a useful direction leading to improving the reliability of hydrological models, we believe that efforts should be directed first and foremost toward the use of spatially distributed rainfall data and only secondary to the disaggregation of watershed (land-surface) parameters.