De l'incertitude dans un système de prévision d'ensemble des crues rapides méditerranéennes

Mediterranean flash floods can have dramatic consequences and it is crucial to increase the lead time for those flash floods to better anticipate their impact. But the hydrometeorological forecast is affected by several sources of uncertainty. The major uncertainty comes from the precipitation forecasts used to drive the hydrological models. The knowledge of the initial soil moisture and the hydrological model itself are also affected by uncertainty. In this work, several hydrometeorological ensemble prediction systems are designed and compared to sample these uncertainties. They are based on the AROME kilometer-scale atmospheric model and the ISBA-TOP coupled system, which is dedicated to Mediterranean fast responding catchments. Uncertainty in rainfall prediction is sampled using the AROME-EPS, ensemble prediction with AROME, to drive ISBA-TOP, or perturbing the AROME deterministic outputs. Those perturbations come from a climatology of rainfall prediction errors. To consider the uncertainties of hydrological modeling and initial soil moisture conditions, a perturbation method of the most sensitive parameters of ISBA-TOP and of initial soil moisture has been designed. The precipitation scenarios issued by AROME-EPS are then used to drive a version of ISBA-TOP where these parameter and initial conditions perturbation methods are applied. Evaluated on several basins and case studies, the so-built prediction system properly encompasses observed discharges. A reasonable number of members allows better skill for discharge prediction than a deterministic forecasting system.