Flash flood modeling with the MARINE hydrological distributed model

Abstract. Flash floods are characterized by their violence and the rapidity of their occurrence. Because these events are rare and unpredictable, but also fast and intense, their anticipation with sufficient lead time for warning and broadcasting is a primary subject of research. Because of the heterogeneities of the rain and of the behavior of the surface, spatially distributed hydrological models can lead to a better understanding of the processes and so on they can contribute to a better forecasting of flash flood. Our main goal here is to develop an operational and robust methodology for flash flood forecasting. This methodology should provide relevant data (information) about flood evolution on short time scales, and should be applicable even in locations where direct observations are sparse (e.g. absence of historical and modern rainfalls and streamflows in small mountainous watersheds). The flash flood forecast is obtained by the physically based, space-time distributed hydrological model "MARINE'' (Model of Anticipation of Runoff and INondations for Extreme events). This model is presented and tested in this paper for a real flash flood event. The model consists in two steps, or two components: the first component is a "basin'' flood module which generates flood runoff in the upstream part of the watershed, and the second component is the "stream network'' module, which propagates the flood in the main river and its subsidiaries. The basin flash flood generation model is a rainfall-runoff model that can integrate remotely sensed data. Surface hydraulics equations are solved with enough simplifying hypotheses to allow real time exploitation. The minimum data required by the model are: (i) the Digital Elevation Model, used to calculate slopes that generate runoff, it can be issued from satellite imagery (SPOT) or from French Geographical Institute (IGN); (ii) the rainfall data from meteorological radar, observed or anticipated by the French Meteorological Service (Météo France); and (iii) the spatially distributed soil and other surface properties viewed from space (land cover map from SPOT or LANDSAT, main rivers, ...). The stream flood propagation model simulates flood propagation in main rivers by solving 1-D Saint Venant equations. The data required for this part of the model are the river morphology, topography and roughness. The MARINE model has already been used previously for real time flash floods forecasting in the frame of the PACTES project on "forecasting and anticipation of floods with spatial techniques'' (funded by the CNES and the French Ministry of Research) concerning the catastrophic 1999 flash flood that occurred in the South of France. The main advantages of MARINE are its ability to run on insufficiently gauged basins (with the help of satellite information) and to run in an operational mode for real-time flood forecasting.