Issuing alerts for extreme events is the outcome of a tedious appraisal between avoiding misses and limiting the number of false alarms. In operational practice, many non-meteorological effects are taken into account. However, as demonstrated by Murphy, the most valuable basis for general decision strategies are probabilistic forecasts reflecting the ‘true’ meteorological skill of the forecasting system. Similarly, the intrinsic value of extreme weather forecasts can only be demonstrated by probabilistic verification.