A Note on the Relationship between Relative Precipitation Frequency and Percent of Correct Forecasts

View More View Less
  • 1 Department of Atmospheric Sciences, University of Washington, Seattle, Wash. 98195
© Get Permissions
Full access

Ramage (1982) has presented evidence, based on extensive verification data for precipitation forecasts, that the percent of correct forecasts decreases approximately linearly with increasing precipitation frequency. However, only a limited range of precipitation frequency was examined in his study. Here the theoretical reltionship between these quantities is derived for the full range of precipitation frequency with use of a model that assumes the forecasts to be unbiased and the Heidke Skill Score to be independent of precipitation frequency. It is shown that for this simple model the relationship is quadratic, the percent correct dropping from 100% for a relative precipitation frequency of zero to a minimum value for a relative frequency of 0.5 and rising again to 100% for a relative frequency of 1.0. The data presented by Ramage appear to fit the quadratic relationship well, within the limited range shown in Fig. 4 of his article.

1 Authors should submit manuscripts for this section directly to Dr. Robert W. Burpee, Editor, Focus on Forecasting, Hurricane Research Division, Atlantic Oceanographic and Meteorological Laboratory, NOAA, Gables One Tower, 1320 S. Dixie Highway, Coral Gables, Fla. 33146. Three copies of each manuscript (text and illustrations), prepared in accordance with “Information for Contributors” on the inside covers of a recent issue of an AMS research journal, are required.

Ramage (1982) has presented evidence, based on extensive verification data for precipitation forecasts, that the percent of correct forecasts decreases approximately linearly with increasing precipitation frequency. However, only a limited range of precipitation frequency was examined in his study. Here the theoretical reltionship between these quantities is derived for the full range of precipitation frequency with use of a model that assumes the forecasts to be unbiased and the Heidke Skill Score to be independent of precipitation frequency. It is shown that for this simple model the relationship is quadratic, the percent correct dropping from 100% for a relative precipitation frequency of zero to a minimum value for a relative frequency of 0.5 and rising again to 100% for a relative frequency of 1.0. The data presented by Ramage appear to fit the quadratic relationship well, within the limited range shown in Fig. 4 of his article.

1 Authors should submit manuscripts for this section directly to Dr. Robert W. Burpee, Editor, Focus on Forecasting, Hurricane Research Division, Atlantic Oceanographic and Meteorological Laboratory, NOAA, Gables One Tower, 1320 S. Dixie Highway, Coral Gables, Fla. 33146. Three copies of each manuscript (text and illustrations), prepared in accordance with “Information for Contributors” on the inside covers of a recent issue of an AMS research journal, are required.

Save