Developing a Tropical Cyclone Genesis Forecast Tool using Global Model Output

Recent research has evaluated global numerical model forecasts of tropical cyclone (TC) genesis out to 5 days over the North Atlantic and eastern North Pacific basins. Results show that the models regularly predict genesis correctly. However, false alarms are still a problem. An operational forecaster has great interest in knowing whether a particular model-indicated genesis forecast is more likely to be correct or a false alarm. Thus, as a part of the Joint Hurricane Testbed, the authors have developed a real-time statistical guidance product that identifies TCs in the model forecast fields and provides a probability of genesis within 2 days and within 5 days. The genesis probabilities are calculated using multiple logistic regression equations developed from a 10 year archive of output from the CMC, GFS, and UKMET models. Separate regression equations were developed for each global model and for each basin. Since research has shown that the probability of genesis increases when more than one model predicts the same genesis event, a consensus-based regression equation also was developed for each basin. This paper will provide details about the development of the guidance product, show preliminary results from quasi-operational testing during the 2014 hurricane season, and discuss some of the operational requirements that must be met for the product to be a useful guidance tool for the National Hurricane Center's Tropical Weather Outlook.