587 Precipitation Skill from a Convection Allowing Pseudo-Ensemble

Tuesday, 24 January 2017
4E (Washington State Convention Center )
Matthew E. Pyle, NOAA/NWS/NCEP/EMC, College Park, MD; and J. R. Carley, B. T. Blake, and B. Zhou

A set of multi-month experiments have been conducted with the High Resolution Ensemble Forecast (HREF) product generator system, focusing on improving the predictive skill of ensemble mean precipitation at convective (~ 4 km) scales.  The operational HREF currently consists of time-lagged members from several operational convective scale models [High Resolution Window (both WRF-ARW and NMMB dynamical cores) and the North American Model CONUS nest], forming a pseudo-ensemble from which mean, spread, and probability products are generated to 36 hours.  Several probability-matched (PM) mean approaches were tested, and a blend of the arithmetic mean and the PM mean generally performed best in terms of equitable threat score, fraction skill score, and frequency bias within the HREF.  Tests reducing the contribution from HREF constituent models showing less precipitation forecasting skill also were run, leading to some seasonally-dependent conclusions about optimal membership for precipitation forecasts.   Planned changes for an upcoming HREF implementation, including membership changes and enhanced probability guidance, will also be discussed.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner