481 Short and extended range probabilistic tropical cyclone genesis forecasts based on a climatology of global numerical model output

Tuesday, 8 January 2013
Exhibit Hall 3 (Austin Convention Center)
Daniel J. Halperin, Florida State University, Tallahassee, FL; and J. Cossuth, H. E. Fuelberg, and R. E. Hart

The Graphical Tropical Weather Outlook (GTWO) is a highly visible forecast product that is routinely issued by the National Hurricane Center (NHC) to highlight areas of interest in the tropical basin. It includes a probabilistic forecast of tropical cyclone (TC) genesis within the next 48 hours as subjectively determined by the hurricane specialist. These forecasts partly rely on output from global numerical models. Prior research has shown that global models have some skill in predicting TC genesis, and that these forecasts have a higher success rate at shorter forecast lead-times.

To provide operational forecasters with an additional tool for diagnosing the probability of TC genesis, a climatology of 6-48 h model-indicated TC genesis forecasts has been developed for five global models (CMC, ECMWF, GFS, NOGAPS, UKMET) over several hurricane seasons. All TC genesis forecasts between 6 and 48 h were compiled for each model. We then determined how many model-indicated TC genesis forecasts corresponded to actual genesis events within 48 h of the model initialization time. Results are divided by region and time of year. A similar analysis is performed on 6-120 h model-indicated TC genesis forecasts. While the 6-120 h temporal window for TC genesis is longer than 6-48 h, the models tend to produce more false alarms at longer forecast-lead times. Therefore, the higher probabilities one might expect from a longer verification window are offset by the increase in false alarms. Nevertheless, we believe a climatology for extended range TC genesis forecasts will be useful since it provides an explicit probability of genesis for disturbances that may not yet currently exist.

Results show noticeable differences in the success rates among various regions. They also depict the regions where models are more or less likely to forecast TC genesis (e.g., a large number of genesis events over the main development region, but fewer events over the Gulf of Mexico). The models' 6-48 h genesis forecasts generally have a higher success rate in regions closer to land (i.e., Gulf of Mexico, Caribbean Sea). Our results suggest such a probabilistic genesis tool will be useful to operational forecasters. We are testing the application of this tool in real time at http://moe.met.fsu.edu/modelgen/.

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