Integration of a snow module to a hydrological forecasting model: what contribution to better predict nival floods? ; Prise en compte de la neige dans la prévision hydrologique : quel apport pour mieux prévoir les crues nivales ?

[Departement_IRSTEA]Eaux [TR1_IRSTEA]ARCEAU [Encadrant_IRSTEA]Ramos, M.H. ; Hydrological forecasting presents considerable challenges: protection of people and properties, power production, etc. Good streamflow estimates are essential to make the right decisions. However, certain catchments present particular challenges: this is the case of mountainous areas, where the occurrence of snow and the difficulties of implementing observational networks complicate the modelling of flows. The introduction of a snow model in a hydrological forecasting model should help make improvements on the simulation of flows of snow-affected catchments, while not degrading the prediction on the other basins. This study aims at evaluating the improvements from the integration of the snow module Cemaneige to the hydrological forecasting model GR3P, both tools developed at Cemagref. A comparative analysis of the two models GR3P (without snow modelling) and GR5P (GR3P + snow modelling) at a daily time step was performed on a sample of 176 French catchments located in mountainous areas. The hydrological flow forecasts use as input four years of PEARP meteorological ensemble forecasts from Météo-France (2005-2009). Thus, 11 equally probable scenarios of streamflow are predicted to two forecasting lead times (d+1 and d+2). The study consisted of the introduction of the snow modelling routine within the structure of the GR3P model, and in the evaluation of this new model in ensemble prediction. Following this work, it appeared that the model GR5P was more efficient than the version without treatment of snow. The improvement becomes more significant when the forecasting lead time increases. In fact, the streamflow-based model update (i.e. the assimilation of the last observed flow into the hydrological model) tends to minimize the differences between the two models (GR3P and GR5P) at the first lead time. The classification of catchments according to their hydrological regimes showed that the contribution of the snow routine is significant ...