May 22 - July 28
Storm Cell Displacement Errors in the Warn-on-Forcast System
Brina M. Lemke, Derek R. Stratman, and Corey K. Potvin
What is already known:
- The Warn-on-Forecast System (WoFS) uses rapid-update ensemble data assimilation along with ensemble forecasts to provide probabilistic guidance with increased lead times for local severe weather events.
- Displacement errors occur between WoFS-forecast storms and Multi-Radar Multi-Sensor-observed storms, which can decrease forecast accuracy and can self-perpetuate within successive forecasts.
- Giving forecasters a better understanding of these storm displacement errors will provide a basis for manual correction or compensation before the dissemination of information to the public.
What this study adds:
- There is little variation in storm cell displacement errors in WoFS forecasts across the years 2017-2020, forecast ensemble members, and planetary boundary layer schemes.
- Storms forming after initialization have smaller biases than storms that are well-assimilated.
- Storms of higher mean intensity have large northeastward biases while storms of lower mean intensity have northwestward biases.
Abstract:
The Warn-on-Forecast System (WoFS) is a rapidly updating ensemble data assimilation and prediction system that provides probabilistic forecasts of severe weather and stands to increase lead times for localized severe weather forecasts. We use matched object pairs between WoFS-forecast storms and Multi-Radar Multi-Sensor (MRMS)-observed storms, then filter for storms with cell-like characteristics in order to study patterns in storm cell centroid displacements. Identifying patterns and correlations in these storm cell displacement errors could provide insight into their causes and be used to increase forecast accuracy or provide compensation for inaccuracies. This study found that there is little variation in displacement errors across years, ensemble members, or planetary boundary layer schemes. However, error spread does increase as lead time increases and minimum error spread for shorter lead times occurs within 1-3 hours of a storm’s assimilation. Storms forming after the initialization time in a forecast also have the smallest bias. Additionally, higher mean intensity storms were found to have larger northeast biases, a correlation that has been previously observed and documented and which could be indicative that the WoFS is underpredicting deviant motion in rotating storms.