An Improved Methodology for Expressing and Communicating Tropical Cyclone Risk

Tropical cyclone (TC) forecast uncertainty is often estimated via the traditional “cone of uncertainty”. This method only uses historical errors and does not show the actual forecast uncertainty. Preliminary results presented in Ortt et al. (2017) demonstrated that using computer forecast model ensembles provides some promise for depicting the actual forecast uncertainty. That study combined the American GFS, the Canadian CMC, and the European ECMWF ensembles into a single ensemble. It was shown for selected Atlantic and east Pacific hurricanes from 2016 that the center remains within an area bounded by at least a 20 percent chance that the center will pass within 200km of a given point nearly 80 percent of the time. This provides a more accurate representation of the forecast uncertainty than does the “cone of uncertainty” as not only does the center remain within the area more frequently, but it also provides an actual estimate of the forecast uncertainty based upon the atmospheric conditions, instead of merely relying upon historical errors. The current operational “cone of uncertainty” is constructed by connecting the edges of concentric circles at the 12 through 120 hour forecast periods. The circles represent the area in which the TC will remain 2/3 of the time. No adjustments are made based upon the actual uncertainty of the forecast, which can produce unrealistic depictions of uncertainty, especially when there are bifurcated track possibilities.

This study will expand upon the Ortt et al. (2017) study. Instead of using a select number of storms, this study will evaluate all of the storms from the 2017 Atlantic hurricane season. It will seek to validate the results from the previous study. The study will evaluate all 2017 Atlantic tropical cyclone forecasts to determine the percentage of time that the center remains within the area bounded by a 20 percent chance of the TC passing within 200km of a given location. The study will also seek to determine if the 20 percent threshold proposed by Ortt et al. (2017) is appropriate or if a different threshold should be used. Furthermore, the study will also demonstrate how the ensemble forecasts can be used to improve preparation efforts in advance of landfalling TCs. The study will evaluate whether or not there are certain probabilistic thresholds from the ensembles that can be used in advance of a TC arrival that can be used to objectively advance one’s hurricane response plan. The results of the study will be presented at the conference.