EXPERIMENTS OF SHORT-RANGE ENSEMBLE PRECIPITATION PROBABILITY FORECASTS

In order to obtain useful information and create probability forecasts from ensemble, experiments of short-range ensemble precipitation probability forecasts are made for 22 precipitation cases from November 2001 to May 2002. The ensemble is created by using MM5 model configuration with different model physical process parameterization schemes and identical initial conditions. There are 8 ensemble members. Precipitation probability forecasts are created from the ensemble by using the methods of "Average", "Correlation" and "Rank". Calculations of ranked probability score (RPS), Brier score (BS) and relative operating characteristic (ROC) indicate that, for the synthetic effect of all precipitation categories' probability forecasts, "Rank" is much better than "Correlation" and "Average", and "Correlation" is almost same as "Average". For the average effect of every precipitation category's probability forecasts, "Rank" is also the best among the three methods, and "Correlation" is same as "Average". The average BS difference between "Rank" and "Average" is large and the average ROC square difference between the two methods is small. Averagely, the three methods' probability forecasts are all skillful for all precipitation categories. The skill of probability forecast for small precipitation category is higher than the skill large precipitation category.