Skill in Precipitation Forecasting in the National Weather Service

Abstract

All known long-term records of forecasting performance for different types of precipitation forecasts in the National Weather Service were examined for relative skill and secular trends in skill. The largest upward trends were achieved by local probability of precipitation (PoP) forecasts for the periods 24?36 h and 36?48 h after 0000 and 1200 GMT. Over the last 13 years, the skill of these forecasts has improved at an average rate of 7.2% per 10-year interval. Over the same period, improvement has been smaller in local PoP skill in the 12?24 h range (2.0% per 10 years) and in the accuracy of ?Yea/No? forecasts of measurable precipitation. The overall trend in accuracy of centralized quantitative precipitation forecasts of ?0.5 in and ?1.0 in has been slightly upward at the 0?24 h range and strongly upward at the 24?48 h range. Most of the improvement in these forecasts has been achieved from the early 1970s to the present. Strong upward accuracy trends in all types of precipitation forecasts within the past eight years are attributed primarily to improvements in numerical and statistical centralized guidance forecasts. The skill and accuracy of both measurable and quantitative precipitation forecasts is 35?55% greater during the cool season than during the warm season. Also, the secular rate of improvement of the cool season precipitation forecasts is 50?110% greater than that of the warm season. This seasonal difference in performance reflects the relative difficulty of forecasting predominantly stratiform precipitation of the cool season and convective precipitation of the warm season.