Using seasonal forecasts in a drought forecasting system for water management : case-study of the Arzal dam in Brittany

This study focuses on an integrated hydro-meteorological forecasting system developed to forecast low flows upstream the Arzal dam and based on a lumped hydrological model. Medium-range meteorological forecasts from the ECMWF ensemble prediction system (51 scenarios up to 9 days ahead) are combined with seasonal meteorological forecasts also from ECMWF to provide extended streamflow forecasts for the summer period. The performance of the forecasts obtained by this method is compared with the performance of two benchmarks: (i) flow forecasts obtained using an ensemble of past observed precipitation series as precipitation scenarios, i.e. without any use of forecasts from meteorological models and (ii) flow forecasts obtained using the seasonal forecasts only, i.e. without medium-term information. First, the performance of ensemble forecasts is evaluated and compared by means of probabilistic scores. Then, a risk-based visualisation tool was set up to assess whether the different ensembles can forecast past drought risks. The tool is designed to support decision-making in situations that may potentially be a source of conflict between stakeholders and users by characterising future events in relation to past observed extremes. Results are presented and the capacity of the system to provide useful information in an operational context is discussed.