A logistic regression model for WNP tropical cyclone formation forecast

Tropical cyclone (TC) formation forecast is one of the most difficult tasks in numerical weather prediction. Previous studies show that numerical models with grid size near 10 km and explicitly resolved cloud schemes are required to have the potential to simulate the emergent evolution of TC precursors and predict the occurrence of TC genesis events (Jin et al. 2008). Nevertheless, recent studies by the authors (Peng et al. 2009, Fu et al. 2009) suggest that some key meteorological parameters can distinguish between developing and non-developing disturbances for TC formation. Using NOGAPS model outputs, those key parameters are calculated and serve as predictors of a logistic regression model for the western North Pacific (WNP) TC formation. Six years (2003-2008) data are used to develop the WNP TC prediction model. A genesis index is defined from the model to predict future TC formation. This model have been validated with the 2009 WNP TC events, the results show encouraging model performance.