Convection-Resolving Ensemble-Based Forecasts of Extreme Precipitation Associated With Landfalling Tropical Cyclones: Assessment of Skill and Utility in the Operational Forecasting Process

As a tropical cyclone makes landfall, it is often accompanied by heavy rainfall along its track. Case-to-case variability in tropical cyclone rainfall arises primarily from variability in storm track and motion with secondary influences from environmental conditions and cyclone structure. Further variability in numerical forecasts of tropical cyclone rainfall arises from how the cyclone, its environment, and parameterized physical processes are depicted within the numerical model. While recent work has shown that high resolution convection-resolving numerical models can enhance the skill associated with continental precipitation and thunderstorm forecasts, it is unclear as to whether such enhanced skill translates to tropical cyclone precipitation forecasts and how such skill is manifest.

To evaluate the skill and utility of a convection-resolving ensemble in the operational tropical cyclone precipitation forecast process, a twenty-two member forecast ensemble of a representative high-impact tropical cyclone precipitation event, Tropical Storm Fay of 2008, was created using version 3.3 of the Advanced Research Weather Research and Forecasting model. Simulations were conducted at a horizontal grid spacing of 4 km over a 72-hr period centered on the time period of heaviest rainfall across North Florida and Southwest Georgia associated with Fay. The forecast ensemble was created based upon variations in both meteorological data (initial conditions, sea surface temperature data, and stochastic perturbation application) and physical parameterizations (boundary layer and microphysical processes).

Substantial synoptic-scale and mesoscale variability was observed between individual ensemble members. The greatest synoptic-scale variability was noted among ensemble members featuring different initial boundary conditions. Differences in the location of and accumulated rainfall from the primary feeder bands associated with Fay arose among forecasts in which one or more ensemble parameters varied. Multiple grid point and object-based measures of individual member and collective ensemble forecast system skill in predicting the timing, location, and intensity of rain bands associated with Fay will be presented. Physical understanding as to why such differences between simulations arise will be discussed.

A case-blind event simulation exercise using forecasters from multiple National Weather Service forecast offices was conducted to assess the utility of the ensemble in the operational forecasting process in a subjective, yet applied, manner. The formulation of, and results from, this exercise will be presented. Implications to the operational forecast process, including an assessment of how forecasters evaluate ensemble forecasts of high-impact tropical cyclone precipitation events, will be discussed.