Landfall Prediction of Hurricanes/Typhoons from the FSU Superensemble

T.N. Krishnamurti, T.S.V. Vijaya Kumar, Lydia Stefanova and Jeffrey Wood

Department of Meteorology

Florida State University

Tallahassee, FL 32306-4520

Abstract

The FSU superensemble methodology entails a training phase where forecasts parameters of multimodels are regressed against corresponding observed (or analysis) measures.  This is based on a large sample of forecasts.  That statistics based on the training phase is passed on to a forecast phase where the collective bias of multimodel forecasts are corrected to construct these so called superensemble forecasts.  The landfall of hurricane/typhoon forecasts addresses the following parameters: position, intensity, timing and heading angles at landfall.  These special hurricane/typhoon related landfall superensemble forecasts are constructed at every 12-hour interval between 0 and 72 hrs.  The data sets include all the storms over the Atlantic (including Caribbean and Gulf of Mexico) and the Pacific Ocean for the years 1998, 1999,2000. The member models include all the major operational models that provide operational forecasts on position and intensity over these oceanic basins.

The results of these forecasts show that some major improvements in the position, intensity, heading angle and timing of landfall of hurricanes/typhoon are possible from the superensemble.  Those skills are much higher that those of the ensemble mean or the best models.  The practical utility of this approach is emphasized in this study.