The Probability of Precipitation (PoP) Index measures the accuracy of our ensemble forecast system rainfall events for UK Met Office observing sites.
The Index is calculated for two precipitation amount thresholds, greater than 4.9 mm and greater than 11.9 mm and for two 24 hour periods, Day 1 and Day 2 (00 UTC to 24 UTC for each day).
Forecast probabilities are issued for each site for both days for each of the two thresholds such that in effect 4 sets of forecast data are issued.
A rolling 36-month score is produced using the combined available data from the quality controlled site list for each threshold and both Day 1 and Day 2 forecasts.
PoP is verified using the Brier Skill Score, which is one minus the ratio of the Brier Score for the model forecasts and the Brier Score for a reference forecast based on the long-term frequency of such rainfall accumulations at each site.
The Brier Score and Brier Skill Scores are well-known verification measures used for probabilistic forecasts. The Brier Score is a mean-squared difference between forecast probability and the binary observation (1 if an event occurred, 0 if not).
The Brier Skill Score ranges from 1 (a perfect score), through 0 (indicating no skill over the reference) to minus infinity. A negative score would indicate that forecasting the long-term frequency of such rainfall accumulations is better than the forecasts.
The nature of probabilistic forecasts means that the forecasts can never be completely wrong or right. This absolute dichotomous behaviour is only available if a "yes/no" forecast is issued.
Last updated: 12 May 2011