6B.6 Multi-model Statistical-Dynamical Climate Forecasts of Tropical Cyclone Landfall

Tuesday, 29 April 2008: 11:30 AM
Palms E (Wyndham Orlando Resort)
Samson K.S. Chiu, City University of Hong Kong, Hong Kong, Hong Kong; and J. C. L. Chan

A multi-model statistical dynamical technique for the seasonal prediction of tropical cyclone (TC) landfall is developed by relating the number of TC landfall with monthly hindcasts. The 1980 – 2001 European Centre for Medium-Range Weather Forecasts monthly forecasts of the European Union's DEMETER (Development of a European Multimodel Ensemble system for seasonal to inTERannual prediction) dataset is chosen to correlate with the number of TC landfall. The sum of TC landfall in August and September is correlated with the averaged two-month forecast fields of the same months. Area of strong correlations (confident level ≥ 95%) in the correlation maps are selected as predictors. Simple least-square linear regression is performed to compute the correlations. A stepwise linear regression is performed to build a regression model to forecast TC landfall. To improve the regression model, the residual is correlated with the DEMETER forecasts again to find out whether other large-scale areas have parameters that strongly correlate with residual of the regression model. A Jackknife regression is performed to test the robustness of the forecast equation. The predictors may affect the TC landfall through the steering flow and other atmospheric oscillations such as the El Nino-Southern Oscillation and the Quasi-Biennial Oscillation.

Predictions are first tested for the Atlantic region and the results are promising. The next step is to test this concept for the western North Pacific. A multi-model technique using hindcasts from other models in the DEMETER dataset will be developed. These results will be presented at the conference.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner