A NEW CONCEPT OF SKILL SCORE FOR RATING QUANTITATIVE FORECASTS

View More View Less
  • 1 Weather Bureau Forecast Center, San Bruno Calif.
© Get Permissions
Full access

Abstract

Skill scores for rating quantitative forecasts are proposed to take into account the deviations occurring between forecast and observed values. One score, the “deviation” skill score, weights the forecasts linearly according to the deviation; a second score, the “quadratic” skill score weights them according to the square of the deviation. These two scores are compared with the conventional skill score for two sets of forecasts, and for the same forecasts with bias introduced. It is concluded that use of either the deviation or the quadratic skill score is preferable to use of the conventional skill score in rating quantitative forecasts. Examples of the step-by-step computations of the two new scores are given.

Abstract

Skill scores for rating quantitative forecasts are proposed to take into account the deviations occurring between forecast and observed values. One score, the “deviation” skill score, weights the forecasts linearly according to the deviation; a second score, the “quadratic” skill score weights them according to the square of the deviation. These two scores are compared with the conventional skill score for two sets of forecasts, and for the same forecasts with bias introduced. It is concluded that use of either the deviation or the quadratic skill score is preferable to use of the conventional skill score in rating quantitative forecasts. Examples of the step-by-step computations of the two new scores are given.

Save