J13.1
Ensemble data assimilation for the short-term prediction of severe convection

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Wednesday, 26 January 2011: 10:30 AM
Ensemble data assimilation for the short-term prediction of severe convection
2A (Washington State Convention Center)
Glen Romine, NCAR, Boulder, CO; and D. Dowell and C. Snyder

Considerable challenges remain in understanding and predicting the initiation and subsequent evolution of high impact convective weather events, particularly in the vicinity of complex terrain. Forecasters often rely on trends in observations and personal experience for short-term prediction since model forecast guidance at the temporal and spatial scales relevant to convective forecasting is often poor and/or cumbersome to interpret. As such, significant opportunities for improving guidance from storm-scale ensemble forecasts should exist where atmospheric conditions are rapidly evolving such as during convective initiation. To address this a WRF-DART based 50-member ensemble assimilation system is demonstrated with both meso-scale and storm-scale probabilistic analyses and forecasts. A retrospective period of 4-17 June 2009 is examined with 3-hourly meso-scale ensemble analysis (15-km horizontal grid spacing) on a CONUS domain with continuous cycling, which provides initial and boundary conditions for regional storm-scale analyses (3-km horizontal grid spacing) centered near the Colorado Front Range. On the storm-scale domain, continuous assimilation of conventional Doppler radar observations for one hour precedes ensemble forecasts extending out six hours, with ensemble forecasts every three hours from 15-00Z daily. Control deterministic forecasts, drawn from the meso-scale analysis member with a ‘best fit' to the ensemble mean at the start of the forecast period, are also spawned during each ensemble forecast period. We will demonstrate recently developed tools and methods for probabilistic guidance from convection resolving ensemble forecasts.