3rd HyMeX workshop 1-4 June 2009 Heraklion (Gournes), Crete-Greece
Hydrometeorological ensemble forecasting of Mediterranean flash-flood events Béatrice VINCENDON (CNRM/GAME (Météo-France, CNRS)); V. Ducrocq, O. Nuissier, B. Vie, G.M. Saulnier
Several uncertainties affect hydrometeorological prediction especially when one works on Mediterranean intense weather events which often lead to devastating flash-floods. Because of the nature of those precipitating systems but also because of temporal and spatial scales involved, the forecast of those events is difficult even for high-resolution convection-resolved numerical weather prediction systems. Those meteorological uncertainties forecasting strongly influence hydrological modelling.
During the last autumn, the ISBA-TOPMODEL hydrometeorological system, developed within the framework of the GMES/PREVIEW project, was used driven by meteorological forecasts from AROME. Hourly discharges at several outlets of the Cévennes-Vivarais region were predicted every day. Such a hydrometeorological system is useful to evaluate the uncertainties of the high-resolution rainfall forecast and investigate their propagation in the hydrological model.
After evaluating the uncertainty of the rainfall from the deterministic AROME forecast by comparison to observed rainfall (radar and raingauges), two methods are investigated in order to take into account of the uncertainties of the rainfall forecast and produce an ensemble discharge forecast. First, we use the convective-scale ensemble AROME simulation rainfall fields to drive ISBA-TOPMODEL. Then, assuming conservation of some statistical or physical properties of the rainfall forecast some perturbations will be introduced in the AROME rainfall fields. In this way, we hope to take into account location errors, bias in the rainfall intensity distribution and bias in the areal rainfall at the hydrological scales. The first method was carried out on two flash-flood events over the Cévennes-Vivarais region : 21-22 October 2008 and 1-2 November 2008.
In a next step, the second method will be implemented : we will tune the perturbations added to the rainfall forecast and use the obtained fields to drive the ISBA-TOPMODEL system in order to produce enough but relevant ensemble spread to sample the uncertainties in the convective-scale rainfall forecast. Then the hydrological responses obtained with both methods will be compared.
Then, we intend to assess the sensitivity of the hydrological model to soil moisture initial state. Some simulations using several sources of ground water fields will be compared. Observations of soil moisture during HyMeX EOP and LOP should help in validating these products and in specifying the initial moisture field errors.
Finally, uncertainties associated with the hydrological modeling errors will be examined by exploring the space of likely hydrodynamic parameters. The assesment of the sensitivity of the hydrological system to these several sources of uncertainties will allow us to develop an ensemble and probabilistic flash-flood forecast system over the Cévennes-Vivarais region. For both EOP and SOPs, we propose to run it in real-time. The deliverables would be some probabilistic rainfall products as well as ensemble discharge forecasts based on the AROME and ISBA-TOPMODEL systems. Implementation of this ensemble flash-flood forecast to other catchments over the Mediterranean region could be foreseen (over Catalonia for instance).
We are expecting from the HyMeX EOP, observations for the verification of the ensemble hydrological forecasting system, including : discharges at main outlets of Cévennes-Vivarais region (and if possible also for some current ungauged basins); soil moisture contents covering the whole domain of simulation ; high-quality quantitative precipitation estimation.
In addition, we are interested in high-resolution observations from the SOP on a specific watershed of the Cévennes-Vivarais region, to validate and improve the representation of physical processes in the hydrolometeorological model (water and energy budget, soil temperature and moisture, discharges along the river, etc).