Comparison of the Warn-on-Forecast System and a High Resolution Rapid Refresh Time-Lagged Ensemble for Forecasting Short-Term Convective Evolution

The Warn-on-Forecast System (WoFS) is an experimental convection-allowing model (CAM) ensemble run at the National Severe Storms Laboratory (NSSL) over targeted daily mesoscale domains with the goal of predicting severe thunderstorm hazards on the 0-6 hour timescale. A key feature of WoFS is multiscale EnKF-based data assimilation with 15-minute cycling, including radar and satellite observations. For spring 2019, WoFS utilized the WRF-ARW core at 3-km grid spacing with 18 mixed-physics forecast members and ran twice hourly preceding and during the diurnal convective peak. WoFS inherits its background from the experimental High Resolution Rapid Refresh Ensemble (HRRRE), a 3-km CAM ensemble run twice daily at NOAA’s Global Systems Division (GSD) over a full-CONUS domain. The HRRRE shares much of its model configuration with the deterministic High Resolution Rapid Refresh (HRRR), a 3-km operational CAM run hourly over a full-CONUS domain.

In this presentation, forecast skill metrics for WoFS and an informal HRRR time-lagged ensemble (HRRR-TLE) are examined at lead times of 0-6 hours. Specifically, we verify neighborhood threshold exceedance probabilities from each ensemble for simulated radar reflectivity using the Multi-Radar Multi-Sensor (MRMS) system. Particular focus will be placed on the comparative performance of WoFS and the HRRR-TLE over a range of different neighborhood sizes, aiming to illuminate whether WoFS offers its greatest advantages at spatial scales smaller than those targeted in current operational post-processing techniques for CAM ensembles. Objective verification is supplemented with subjective ratings from participants in the 2019 Hazardous Weather Testbed (HWT) Spring Forecasting Experiment (SFE).