Abstract : P.2
Coupled atmospheric-hydrologic ensemble prediction in Alpine catchments

Simon Jaun, André Walser, Massimiliano Zappa, Bodo Ahrens, Joachim Gurtz, Christoph Schär
Institute for Atmospheric and Climate Science, ETH Zurich

Appropriate dispositions in the case of flood occurrences require longer lead times in hydrological forecasting. This in turn implies an increased uncertainty, which cannot be accounted for by a deterministic simulation. A possibility to address this issue is the use of probabilistic forecasts driven by meteorological ensemble prediction systems (EPS), as the meteorological input is considered to represent a main source of uncertainty.

Within the Forecast Demonstration Phase of the Mesoscale Alpine Program (MAP D-PHASE), the provision of the following coupled hydrological forecast chain is planned to provide an end-to-end forecasting system: An operational global atmospheric EPS (ECMWF EPS), dynamically downscaled with a limited-area atmospheric EPS (COSMO-LEPS), is used to drive a semi-distributed hydrological model (PREVAH). Daily forecasts with a range of 120 h, using a subsample of 16 ECMWF ensemble members selected by the COSMO-LEPS cluster analysis will be conducted. The study area will cover the Ticino-Toce catchment, further divided into 6 subcatchments. The hydrological model runs at a spatial resolution of 500 meters and with hourly time steps. The application of this model framework during the MAP D-PHASE should help to explore possibilities, advantages and weaknesses of the hydrological EPS.

Results from a hindcast-experiment with a similar setup in the Rhine catchment till Rheinfelden show potential benefits of the additional probabilistic information. The verification of the model system indicates a very notable performance for selected case studies of extreme flood events (e.g. August 2005) as well as for a continuous long-term hindcast period (i.e. the full year 2005). Although challenging, a similar performance of the model system is targeted for the Ticino-Toce region, running in a quasi-operational mode.