Abstract:The extreme weather forecast products of ECMWF ensemble prediction system (EPS) are analyzed and verified. The products include extreme forecast index (EFI) and “shift of tail” index (SOT) of 08-08 BT mean temperature, maximum temperature, minimum temperature and total precipitation. The results indicate that the forecast skill of temperature EFI is similar to SOT for extreme temperature, but the precipitation SOT is better than EFI for extreme precipitation. Using the last 3 years’ dataset, the thresholds and corresponding measures are estimated for different lead times and extreme events over different percentiles of extreme temperature and precipitation, under the criteria of TS maximization. For the extreme low (high) temperature over 1% (99%) percentile, the thresholds of mean temperature EFI and SOT are about -0.85 (0.75) and 0.38 (0.00), respectively. The thresholds of maximum and minimum temperature EFI and SOT are nearly to that of mean temperature. For the extreme precipitation over 95% and 99% percentiles, the thresholds of EFI are about 0.45 and 0.7, and that of SOT are about -0.6 and 0.4. The longer the lead times, the smaller the thresholds and the lower the forecast skills. And the more extreme the events, the larger the thresholds. Under the criteria of TS maximization, the forecast biases are nearly 1, so these thresholds could be applied in the operation. The thresholds and forecast skill of temperature EFI and SOT show significant seasonal variation, with higher skill and thresholds in the summer half year and lower in the winter half year. The temporal variation of precipitation EFI and SOT is not significant. There are spatial variations of these thresholds and forecast skills and they are different for different indexes.