1.3 Explicit Forecasts of Low-Level Rotation with Convection-Allowing Models Using 1-Km Horizontal Grid Spacing

Monday, 22 October 2018: 9:30 AM
Pinnacle room (Stoweflake Mountain Resort )
Ryan A. Sobash, NCAR, Boulder, CO; and C. S. Schwartz, G. S. Romine, and M. L. Weisman

Explicit attributes of convective storms within convection-allowing model (CAM) forecasts are routinely used as surrogates for the occurrence of severe convective weather events. In previous work, probabilistic tornado forecasts using low-level rotation surrogates (e.g., 0 – 3 km AGL updraft helicity or 1 km AGL vertical vorticity) were more skillful than forecasts using mid-level rotation surrogates (e.g., 2 – 5 km updraft helicity; Sobash et al. 2016), and environmental information (e.g. LCL height, significant tornado parameter) combined with mid-level rotation surrogates improved CAM forecasts of tornadoes (e.g., Gallo et al. 2016). These studies used CAM forecasts with horizontal grid spacing >= 3 km, which under-resolves processes relevant for the development of low-level rotation. In fact, 1-km grid spacing was necessary in previous idealized studies to produce significant near-ground rotation, justifying moving toward 1-km horizontal grid spacing in future operational storm-scale prediction efforts (e.g., NOAA Warn-on-Forecast).

To examine the skill of 1-km horizontal grid spacing for next-day tornado prediction, we produced a large set (N=497) of 36-hour, CONUS-wide, CAM forecasts using 3-km and 1-km horizontal grid spacing and extracted a variety of low-level surrogate diagnostics. Forecasts of tornadoes were produced by applying a set of thresholds to the surrogate fields and smoothing the forecasts; forecasts were verified against SPC tornado reports. The 1-km grid spacing forecasts were more skillful than the 3-km grid spacing forecasts, producing larger fractions skill scores (differences > 0.1) and larger Brier skill scores over all smoothing length scales. Consistent with prior work, shallower integration depths closer to the surface were more skillful (i.e., 0 – 1 km AGL UH) than deeper integration depths (i.e., 0-3 km AGL or 2-5 km AGL UH). The most skillful next-day tornado forecasts were produced using 0 – 1 km AGL UH from the 1-km forecasts, although smoothing was still necessary to improve the reliability at the expense of sharpness. Reasons for the differences in skill between the 3-km and 1-km grid spacing forecasts will be discussed.

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