Thursday, 2 August 2001: 4:00 PM
Numerical Prediction of Typhoon Tracks and Intensity using a Multimodel Superensemble
Using currently available operational forecast data sets on the tracks and intensity of Pacific typhoons of the years 1998, 1999 and 2000 we have constructed a multimodel superensemble following our earlier work on the Atlantic hurricanes. The models included here comprise forecasts from the ECMWF, EMC/NCEP (AVN and MRF), NOGAPS and JMA. The superensemble methodology includes a bias estimation training phase where a multiple regression based least square minimization principle is employed. This is quite different from a simple bias correction where a mean value is simply shifted. These bias estimates are described by separate weights for each 12-hour of forecasts for each of the member models; these forecasts are carried out to 144-hours. These past weights are used to construct a superensemble forecast for the future. Some 150 past forecasts of typhoon days are used for the training phase. Our findings show a marked improvement for the tracks and intensity of forecasts from the proposed multimodel superensemble as compared to the forecasts from member models and the ensemble mean. The paper will discuss the detailed statistics for the Pacific typhoon forecasts for the three years 1998, 1999 and 2000.