Lagged-Average Predictions of Tropical Cyclone Tracks

Abstract

The lagged-average technique of combining a set of forecasts verifying at the same time is applied to tropical cyclone track prediction. Only a 3% improvement is achieved when the 24-b one-way tropical cyclone model (OTCM) forecasts are combined with unmodified track predictions verifying at t + 24 h. After statistically modifying the previous OTCM forecasts to take into account more recent information, the mean errors of the 24-h lagged-average forecast are reduced by 15% compared to the original 24-h OTCM. The departure of the 24-h OTCM forecast from the modified lagged-average forecast appears to be a useful predictor of whether the 24-h forecast errors will be below average, average, or above average. Thus, the lagged-average forecast approach appears to be useful for tropical cyclone track prediction.