A new parameter for forecasting tornadoes in landfalling tropical cyclones

Tornadoes that occur with land falling tropical cyclones (TCs) are difficult to forecast. We have developed a statistical parameter to assist in forecasting tornadoes in the TC environment. We hypothesized that a composite index tuned specifically to the TC environment would outperform indices developed for the middle latitudes. Our product, called the Tropical Cyclone Tornado Parameter (TCTP), gives the probability that a tornado will occur at locations on a 160 km grid during a 6 h period. Our dataset consisted of 483 TC related tornado reports from 30 storms between the years 2000 to 2008. The locations of these tornadoes were placed onto the 160 km grid, whereas grid locations not experiencing a tornado comprised the null dataset. Geographic parameters described the locations and timing of the tornadoes with respect to the parent TC. In addition, a pool of 28 potential TC environment parameters was obtained from RUC model analyses. These included various measures of temperature, humidity, stability, and wind shear. Stepwise multiple linear regression then was used to isolate six parameters that were best related to tornado occurrence. Results from TCTP have been tested on the dependent data set (2000-2008). Results showed that the spatial probability maps from TCTP exhibited good visual agreement with observed tornado occurrence. There also was good visual agreement with the Storm Prediction Center's (SPC) tornado percentage forecasts. We quantified the degree of agreement using Brier Scores and Brier Skill Scores. Brier Scores for the TCTP (SPC) were 0.00480 (0.00158), compared to 0.015500 for climatology. These values and those from the Brier Skill Scores indicate that the TCTP improves on climatology, but not as much as the SPC forecasts. We hope that further testing on independent data during 2010 will confirm that the TCTP is a useful guidance product for forecasters at the SPC. Our paper will describe the TCTP methodology in detail, provide statistical results for the complete dependent data period, and show examples from specific storms. Finally, if landfalling TCs occur this year prior to the conference, their results also will be shown.