Dependence of Vortex Characteristics on Grid Resolution in Simulated Supercells

The National Oceanic and Atmospheric Administration (NOAA) Warn-on-Forecast (WoF) research project seeks to increase the lead time of warnings for hazardous weather, such as tornadoes, through the use of convective-scale model ensembles that provide decision support for National Weather Service (NWS) forecasters (Stensrud et al. 2009). For a short-term forecast model to be useful to tornado warning operations, it must be run at a resolution that both adequately resolves low-level rotation associated with tornado development and fits within the constraints of available computational resources. Current projections estimate that early WoF systems will run at Δx=1 km. Low-level mesocyclones, which are substantially more likely than mid-level mesocyclones to be associated with tornadoes, require Δx~100 m to be well resolved. It is unclear how coarse the model grid can be while still usefully representing the frequency, intensity, and longevity of low-level mesocyclones when compared to the circulations generated on a Δx~100 m grid.

In an attempt to answer this question, numerical simulations based on the sounding from the May 29, 2004, tornadic supercell event in Geary, OK, are generated at 500 m, 250 m, 100 m, and 50 m horizontal grid resolutions using the CM1 cloud model (Bryan and Fritsch 2002). The vortex detection and classification (VDAC) algorithm described in Potvin (2013) is applied to gather data on the physical characteristics, longevity, and frequency of vortices produced at each resolution. A statistical analysis of the results compares the ability of each resolution to produce coherent low-level circulations that comport with the 50 m solution and, by extension, offers an estimate of the resolution that would be required for a WoF ensemble to successfully anticipate tornado development.