Quels apports hydrologiques pour les modèles hydrauliques? Vers ...

5forecasting (Smith et al., 2009, this issue). Within a flood forecasting system,such models can be connected to a hydraulic model in order to calculatethe extension of the inundation area and facilitate the dissemination of floodwarnings.Although distributed models are more and more widely used, they are still10subject to problems such as data requirements or parameter identifiability,tel-00392240, version 1 - 5 Jun 2009152025which may limit their robustness and their adequacy for operational purposes.Besides, it has not been clearly demonstrated that their greater complexity(in comparison to simpler approaches) contributes additional modelefficiency (Michaud and Sorooshian, 1994; Refsgaard and Knudsen, 1996;Savenije, 2001; Smith et al., 2004; Te Linde et al., 2008). Therefore the useof lumped RR models still remains a valuable alternative in the analysis ofrainfall-runoff transformation at the catchment scale (Sivapalan et al., 2003;Young, 2003) and for proposing robust methods to couple these models withhydraulic models.Lumped RR models can be considered conceptual, whereas hydraulic modelsare often viewed as physically based. Connecting these two kinds of modelraises a number of scientific and technical issues, as will be discussed hereafter.However, the scientific literature is surprisingly lacking in studies focusingon model coupling, although several authors (Klemes, 1983; Robinson andSivapalan, 1995) have pointed out the importance of doing so. As indicatedby Robinson and Sivapalan (1995): “the work towards finding connectionsbetween the physically based and conceptual models has been inadequate,although this is precisely what is required for the advancement of hydrologicalmodeling for predictive purposes.”3