An evaluation of a real-time statistical-dynamical tropical cyclone genesis guidance tool based on global model output

The National Hurricane Center (NHC) has stated that guidance on tropical cyclone (TC) genesis is an operational forecast improvement need. This research seeks to help address that need by bias correcting TC genesis forecasts from three global models (GFS, UKMET, CMC) using multiple logistic regression. The derived regression equations provide a 48 h and 120 h probabilistic genesis forecast for each TC genesis event that is forecast by these global models. The regression equations ran quasi-operationally during 2014 and 2015 over the North Atlantic and eastern North Pacific basins. Reliability diagrams indicate that the regression-based forecasts can provide value as a guidance product for the NHC's Tropical Weather Outlook product.

One limitation to this technique is that the regression equations may not be well-calibrated immediately following an upgrade to the global models. TC genesis forecast verification statistics from the 2011-2014 operational GFS and 2015 GFS reforecasts of 2011-2014 are presented to quantify the impact of a change in model configuration. Results show that the probability of detection is reduced for the reforecast GFS configuration compared to the operational GFS configuration over the North Atlantic and eastern North Pacific basins. The false alarm ratio is greater (reduced) for the reforecast GFS configuration compared to the operational GFS configuration for the North Atlantic (eastern North Pacific) basin.