Forecasting Major European Droughts using North-American Multi-Model Ensemble (NMME)

Tuesday, 16 December 2014
Rohini Kumar1, Stephan Thober1, Luis E Samaniego1, David Schaefer2 and Juliane Mai1, (1)Helmholtz Centre for Environmental Research UFZ Leipzig, Leipzig, Germany, (2)Helmholtz Centre for Environmental Research - UFZ, Leipzig, Germany
Agricultural droughts are defined as water storage deficit in the soil. They have the potential to diminish crop yields causing economic damage or even threatening the livelihood of societies. State-of-the-art drought forecasting systems incorporate seasonal meteorological forecasts to estimate future drought conditions. Meteorological forecasting skill, however, in particular that of precipitation, is limited to a few weeks because of the chaotic behaviour of the atmosphere. One of the most important challenges in drought forecasting is to understand how the uncertainty in the atmospheric forcings (e.g., precipitation and temperature) is further propagated into hydrologic variables such as soil moisture. In this study, we analyse the skill of the North-American Multi-Model Ensemble (NMME), which provides the latest collection of a multi-institutional seasonal forecasting ensemble, for correctly estimating the major drought events (in a retrospective mode) over Europe.

The monthly NMME forecasts are downscaled to daily values to force the mesoscale hydrological model (mHM). The forecasts provided by these dynamical models using mHM are then compared to that of a simple statistical forecasting model. This statistical model uses geo-potential height and antecedent soil moisture conditions to provide future soil moisture estimates and is used here as a benchmark. Both forecasts are compared against reference soil moisture conditions obtained by observation based meteorological forcings for the period from 1981 to 2010. The study area ranges roughly from 10°W to 40°E and 35°N to 55°N, and covers large parts of the Pan-European domain.

Results indicate that the drought forecast by most dynamic models is underestimating the observed drought extent for the most severe events (e.g., 2003), but are outperforming the statistical model. The spatial pattern is in general well represented by the NMME models with anomaly correlations higher than 0.6 for lead times up to seven months. The forecast issued by the NCEP-CFSv2 model shows the highest skill among all models for lead times up to five months. In summary, this study shows the potential of NMME seasonal forecasts for skilful predictions of drought events over Europe.