12.4 Experimental Forecast Evolution using the Warn-on-Forecast System during the 2019 HWT Spring Forecasting Experiment

Thursday, 16 January 2020: 9:15 AM
252A (Boston Convention and Exhibition Center)
Burkely T. Gallo, CIMMS/Univ. of Oklahoma and NOAA/NWS/SPC, Norman, OK; and K. A. Wilson, J. J. Choate, K. H. Knopfmeier, P. S. Skinner, B. Roberts, P. L. Heinselman, and A. J. Clark

The 2019 Spring Forecasting Experiment (SFE) in NOAA’s Hazardous Weather Testbed included an evening activity focused solely on the use of the National Severe Storms Laboratory’s experimental ensemble: the Warn-on-Forecast System (WoFS). WoFS ensemble forecasts contain 18 forecast members, provide guidance out to 6 h, and are initialized every 30 minutes from analyses produced from rapidly cycled (15-min) assimilation of Doppler radar and satellite cloud water path observations. Throughout the 5-week SFE, the first and last initializations began at 1900 UTC and 0300 UTC, respectively.

During the evening activity, two National Weather Service (NWS) forecasters issued a series of probabilistic outlooks covering all types of severe convective thunderstorm hazards (wind, hail, and tornadoes) from 4–8 PM Monday through Thursday. Three types of outlooks were issued each hour: 1) 1-h outlooks covering the hour after forecast issuance, 2) 4-h outlooks covering the time period 1–5 h after forecast issuance, and 3) 1-h targeted outlooks issued for the same valid hour (8–9 PM CDT). This targeted outlook resulted in a series of forecasts that provided steadily decreasing amounts of lead time.

While issuing outlooks, forecasters were surveyed on trends in storm attribute and environmental information that they observed in WoFS forecasts. Additionally, forecasters’ product usage was tracked. This work combines these three data sources (the outlooks, the survey responses, and the product usage data) to assess forecasters’ processes throughout the evolving WoFS initializations. Specifically, we address topics including the skill of outlooks issued by forecasters at varying lead times, the relationships between outlook skill and factors such as event type and forecaster-perceived challenges, and whether specific product usage can be tied to forecast outcomes.

- Indicates paper has been withdrawn from meeting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