Using storm surrogates to produce probabilistic severe weather forecasts from an EnKF-based ensemble forecasting system

Convection-allowing model forecasts provide information about simulated convective-storm properties (e.g. updraft rotation, updraft speed, surface winds, etc.). As described in Sobash et al. (2011), storm surrogates can be derived from these fields to produce surrogate severe probabilistic forecasts (SSPFs). Here, we investigate the skill of SSPFs generated from an EnKF-based ensemble forecasting system run with 30-members, at 3-km resolution, initialized at both 00 UTC and 12 UTC for 32 days during Spring 2013. These forecasts extend through 48-hours, providing an opportunity to document the skill of SSPFs during the first and second day diurnal convective cycle (Day 1 and Day 2). Hourly and daily SSPFs generated from ensemble output were compared to SSPFs generated from a deterministic forecast initialized with GFS initial conditions.

The hourly ensemble SSPFs were consistently more skillful than their deterministic counterparts, as judged by fractions skill scores. Interestingly, the 12 UTC ensemble SSPFs valid on Day 2 possessed similar skill to the 12 UTC deterministic SSPFs valid on Day 1. Also, the 12 UTC ensemble SSPFs valid on Day 2 possessed similar skill to the 00 UTC ensemble SSPFs valid on Day 1. These results support running 00 UTC, and especially 12 UTC, convection-allowing ensemble forecasts to provide useful guidance for the Day 2 convective period.

Finally, to investigate the sensitivity of model resolution on the behavior of the storm surrogate fields, 10 members of each day's 00 UTC ensemble were re-run at 1-km resolution. The skill of the 3-km forecasts will be compared to the 1-km probabilistic severe weather forecasts to determine if the added resolution provides additional benefit for forecasts of severe weather using storm surrogates.