1C.7 Determining predictors for a statistical tropical cyclone genesis tool based on GFS output

Monday, 31 March 2014: 9:45 AM
Regency Ballroom (Town and Country Resort )
Daniel J. Halperin, University at Albany, Albany, NY

Handout (859.5 kB)

Predicting tropical cyclone (TC) genesis has been a vexing problem for forecasters. While the literature describes environmental conditions which are necessary for TC genesis, predicting if and when a specific disturbance will organize and become a TC remains a challenge. As recently as 10 years ago, global models possessed little if any skill in forecasting TC genesis. However, due in part to increased resolution and more advanced model parameterizations, global models now can provide useful TC genesis guidance to operational forecasters.

A recent study evaluated five global models' ability to predict TC genesis out to four days over the North Atlantic basin (Halperin et al. 2013). The results indicate that the models are indeed able to capture the genesis time and location correctly a fair percentage of the time. The study also uncovered model biases. For example, probability of detection and false alarm rate varies spatially within the basin. Also, as expected, the models' performance decreases with increasing lead time. In order to explain these and other biases, it is useful to analyze the model-indicated genesis events further to determine whether or not there are systematic differences between successful forecasts (hits), false alarms, and miss events. This study will examine composites of a number of physically-relevant environmental parameters (e.g., magnitude of vertical wind shear, mid-level relative humidity) and disturbance-based parameters (e.g., 925 hPa maximum wind speed, vertical alignment of relative vorticity) among each TC genesis event classification (i.e., hit, false alarm, miss). We will use standard statistical tests to calculate whether or not any differences are statistically significant. The results may help determine which aspects of the forecast are (in)correct and whether the incorrect aspects can be bias-corrected. This, in turn, may allow us to further enhance probabilistic forecasts of TC genesis.

Supplementary URL: http://moe.met.fsu.edu/modelgen/

- Indicates paper has been withdrawn from meeting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