Hydrometeorological Ensemble Prediction System for flash flood forecasting within HYMEX program

Mediterranean flash floods (FF) are quite devastating. The last example over the south-east France was the fall 2014. The watersheds of the area that are prone to FF have experienced outstanding rainfall intensities and repetitive rainy events that generated several successive floods. The impacts have been dramatic as casualties and heavy infrastructure damage were deplored.

Increasing the lead time of FF forecasts is crucial to better anticipate their impact. FF forecasting is one of the issues addressed within HYMEX (Hydrological cycle in the Mediterranean Experiment http://www.cnrm.meteo.fr/hymex/ ). Uncertainty in Mediterranean flash-floods forecasting. is particularly difficult to handle with. Hydrological Ensemble Prediction Systems (HEPS) are becoming more and more popular methods to deal with the uncertainty that affects discharge forecasts. An HEPS based on both the AROME convective-scale atmospheric model and the ISBA-TOP coupled system has been built within HYMEX first special observing period (SOP1) and is being completed. ISBA-TOP is a coupling between the surface scheme ISBA and a version of TOPMODEL dedicated to Mediterranean fast responding rivers. It was developed as part of the HYMEX program.

The uncertainty on QPF (quantitative precipitation forecast) is sampled by two distinctive methods. The first one classically is the use of an atmospheric EPS. A prototype of a meteorological ensemble at convective scale based on AROME model running on a limited area has been built at CNRM. ISBA-TOP driven with the 12 members of AROME EPS produces a 12 members discharge ensemble.

The second method consists in introducing perturbations in AROME deterministic outputs. Those perturbations are based on rainfall forecast errors and PDFs of errors in term of amounts of rain and location of the heaviest rains previously established. 50 scenarii of precipitation are obtained, they permit to produce an ensemble of 50 scenarii of hourly streamflow time series up to a 30h range on several watersheds of the French Mediterranean regions.

Both methods are then combined so as to manage the uncertainty on QPF at several scales and with a number of member large enough.

Another step toward a complete hydro-meteorological ensemble forecasting system is to take into account the hydrological modelling uncertainty. A sensitivity study has established the hydrodynamical parameters that greatly impact ISBA-TOP simulations. This preliminary step is needed in order to built a hydrological ensemble prediction system in an efficient way. This will be the next challenge.