On the Verification of Seasonal Climate Forecasts

Abstract

Readily available forecasts of winter season temperature anomalies for the continental United States are analyzed and compared to the observed anomalies for each of the past five winter seasons. Forcast skill is evaluated by different verification methodologies, and it is shown that a judgment of skill can be dependent on the particular verification technique employed. Verification in terms of principal components is shown to be a useful diagnostic aid, in that it allows for the recognition of naturally occurring temperature anomaly patterns in the atmosphere. Other general issues concerning the current state of seasonal climate forecasting also are discussed as they relate to the question of verification strategies.