Multi-model superensemble forecasts: Part II. Hurricane prediction

Multi-model superensemble forecasting technique is applied to hurricane prediction, including track and intensity forecasts. This technique utilizes forecast data sets for all of the named Atlantic tropical systems for the years 1998 and 1999. The superensemble procedure undergoes a training phase based on the forecasts of 11 different models and the best track data. Then the forecast phase takes the available forecasts and creates the superensemble forecast. Experimental results for 1998 and 1999 Atlantic hurricane seasons will be shown, including track and intensity forecasts for Hurricanes Dennis, Floyd, and Irene of 1999. Verification versus the 11 member ensemble will be presented. It is noted that the superensemble forecast is better than the ensemble-averaged forecast and is the best performer in nearly all cases of the multi-model forecasts, especially with respect to the track forecasts.