Eighty-Five Percent and Holding—A Limit to Forecast Accuracy?

Abstract

The significance of Percent Correct Scores for National Weather Service (NWS) probability of precipitation (PoP) forecasts is examined. It is shown that the areal variability of rainfall and the nature of PoP forecasts preclude the achievement of a score of 100%—even for the best possible forecasts. A maximum possible percent correct is defined and radar estimates of rainfall coverage are combined with actual forecasts to determine how closely NWS forecasters approached this limit. Day- and nighttime percent correct scores were 75% and 85%, respectively, for the data examined. These values were close to the respective maximum possible scores of 83% and 90%. Relatively small changes in forecasters' percent correct scores are considered in light of these findings.