Verification to Determine and Measure Forecasting Skill

Abstract

Three distinct purposes for the verification and scoring of forecasts have been generally recognized: determination of the accuracy of the forecasts, their operational value, and, lastly, the skill of the forecaster. Although the question most frequently asked is ?How accurate?,? the answer, ?usually about 85%,? is the most trivial. The operational value of forecasts is much more interesting and significant, but difficult or impossible to determine. The skill of the forecaster, however, is a tractable subject. It is defined as ?The ability of the forecaster to sort or group the weather situations so that within any group the probability of one out of several mutually exclusive subsequent events is increased above its climatic frequency.? A set of scores can be designed to reward the forecaster for skilled grouping or sorting of weather patterns and to permit no advantage to an unskilled strategy. Such a scoring system was described more than 10 years ago, but it is not popularly accepted, partly because there never has been a set of uniform goals for verification.