Prediction of Tropical Cyclone Genesis

Predicting tropical cyclogenesis is one of the major challenges in numerical weather prediction. Tropical disturbances with different origins and characteristics serve as tropical cyclone (TC) precursors exist all the time, but only a small percentage of them became TCs. Using NOGAPS daily analysis and TRMM rainfall rates, Peng et al. (2012) and Fu et al. (2012) identified key meteorological parameters that distinguish developing from non-developing disturbances for tropical cyclone formation. These studies indicate that TC genesis is controlled more by thermodynamic variables in the North Atlantic while it is more controlled by dynamic variables in the western North Pacific. Based on these early findings, a TC genesis potential index (GPI) is constructed. A nonlinear regression formulation was used to combine optimally key meteorological variables that control the TC formation. Basin-dependent relationships between large-scale forcing and TC genesis require that separate formulations be constructed for different basins. The Bayesian Information Criterion is applied to further refine the formula. The TC genesis index based on 2004-2008 data was applied to global analysis fields in hindcast mode to provide both the probability and deterministic forecast of TC genesis events for individual disturbances.

The GPI for the western Pacific was tested in the real-time environment since 2011 at the Joint Typhoon Warning Center (JTWC). For the typhoon season of 2013 from June to Sep, the genesis index formula successfully predicted all 20 genesis events in the western Pacific, with a false alarm rate as low as 22.6%. Distinct trends in GPI for developers and non-developers were also noted. Most developing systems tended to show either an increasing trend of the GPI, particularly between 48 and 24 hours prior to formation, or a steady upward trend over several days at values exceeding the development threshold. The study indicates that GPI trends, in combing with individual values, may have greater utility as an operation forecasting tool.