Forecast Skill

Forecast skill is the quantitative accuracy of a forecast measured against a reference such as climatology or persistence, so a positive skill score means the forecast beats the naive baseline. Common metrics are the anomaly correlation coefficient, the root-mean-square error, the Brier and continuous ranked probability scores for probabilistic forecasts, and the skill score that normalizes error against the reference. Skill falls with lead time as predictability decays, and the lead at which skill drops to zero defines the useful forecast horizon. Operational centers track skill to verify models and to set the lead beyond which a wave or current forecast carries little information for routing.