Combining Precipitation Probabilities

Abstract

Two methods are discussed for combining the routine forecasts of the 12 h probability of precipitation made by the National Weather Service, for use when longer period probabilities are desired but cannot be created independently. Both apply a year's forecasts from 28 forecast offices to basic equations of probability to adjust for the obvious dependence of the precipitation events among the forecast periods. Both methods suggest that warm season precipitation events are more independent than cold season ones, as would be expected. One method gave unrealistic results for probability combinations outside the range of those actually used. The other method applied realistic constraints to eliminate this undesirable feature. The largest deviations from probabilities for independent events occurred when combining probabilities of 60%, but the deviations wore only about 5% in the warm season and 10% in the cold season. Tables and an equation for combining probabilities are given.