Forecasting drought in Europe with the standardized precipitation index : an assessment of the performance of the European Centre for Medium Range Weather Forecasts Variable Resolution Ensemble Prediction System

This report describes an assessment of the performance of the European Centre for Medium Range Weather Forecasts variable resolution ensemble prediction system as a tool for forecasting drought using the Standardized Precipitation Index (SPI) with one month lead time. The model is verified using standard verification measures of the Brier Score, the Brier Skill Score, reliability and relative operating characteristics. It is found that for the 1-month SPI, the model has little skill in forecasting drought events and the forecast is generally unreliable. For the 3-month SPI the model has more skill, but this skill comes from the use of 2 months of reanalysis precipitation in and 1-month of forecast precipitation in building in the 3-month SPI. Calibration of the forecasts through adjusting the forecast probabilities to observed frequencies improved the verification statistics. Two case studies using the model were analysed and it was found that the model did not give useful guidance, and in fact calibration had the effect of underestimating the probability of extreme events where the model had some skill. It is recommended that ensemble probabilistic forecasts not be used as a tool for decision making with regard to drought without further improvement in the model performance.