J13.9
The Gerrity equitable skill score as a near optimal alternative to the modified Heidke score for verification of categorical seasonal forecasts
Robert E. Livezey, NOAA/NWS, Silver Spring, MD
The Heidke Skill Score as modified by the NWS Climate Prediction Center and commonly used by other practitioners of climate forecasts has virtually all of the undesirable attributes usually associated with a skill score for ordinal categorical forecasts. These include the ability of forecasters to artificially manipulate the score in several ways to their advantage, inconsistency of the score with other related scores, and no dependence of the degree of forecast error or difficulty on forecaster rewards and penalties. There is an alternative called the Gerrity score that for practical purposes has none of these shortcomings and is easy to compute. The Gerrity score is a particular subset of the Gandin and Murphy family of equitable skill scores for ordinal multi-categorical forecasts. Both the Heidke and Gerrity scores will be described, placed in context with the Gandin and Murphy and other scores available in the literature, and their attributes contrasted. CPC and others have argued that the Heidke Score should be retained because it has been used for a long time and because partners and customers are familiar with it. Because digital monthly and seasonal forecast histories exist the Gerrity score can easily be computed for the entire record of these forecasts. Generally CPC partners are knowledgable professionals who can easily comprehend the Gerrity score, while CPC customers generally don't understand the Heidke score anyway. Thus, a recommendation is that CPC adopt the Gerrity score as the official measure of skill for categorical forecasts, thereby encouraging others to do so as well.
.
Joint Session 13, Seasonal to interannual climate prediction with emphasis on the 2002 El Nino (Joint with 15th Symp. on Global Change and Climate Variations and the Symp. on Forecasting Weather and Climate of the Atmosphere and Ocean (Room 6C)
Thursday, 15 January 2004, 8:30 AM-4:30 PM, Room 6C
Previous paper Next paper