How to Use Low-Resolution Ensemble in Predicting Severe Local Storm?

Storm scale weather system is extremely sensitive to uncertainties in both initial condition and model physics. It is, therefore, quantifying uncertainty in NWP products using ensemble approach is very much desired in mesoscale and storm scale (even more so than for synoptic scale). However, due to limited computing power, regional ensemble forecasting system has to be run at reduced model resolution, while severe local storm is small in spatial scale such as convective mode. How to fill this gap?

An ensemble downscaling and calibration technique called "Hybrid Ensembling" has been developed and tested at NCEP. The method unifies lower-res ensemble system with high-res single deterministic model runs. As expected, the Hybrid Ensembling could recover spatially detailed structures of flow and weather phenonema which were lost due to reduced-resolution in raw ensemble. After one month of testing, it shows that this technique could result in about 12hr improvement in forecast lead time over the operational NCEP Short-Range Ensemble Forecasting (SREF) system (http://wwwt.emc.ncep.noaa.gov/mmb/SREF/SREF.html). Hybrid ensemble mean also outperforms the operational 12km Eta. Due to reduced forecast error, hybrid ensemble has better spread too by reducing outliers and being closer to ensemble mean rms error. Finally, potential impact of this new technique on future mesoscale ensembling will be discussed.