Ensemble Predictability of Convective Weather Hazards

Forecasting convective precipitation remains a considerable challenge for the weather community, but recent progress in storm-scale ensemble forecast system design has enabled considerable progress to be made. A valuable tool in convective weather hazard prediction are ensemble forecasts with sufficiently fine grid spacing to resolve the key attributes of the parent convection, and an appropriate initialization approach that results in ensemble forecast error growth characteristics that approximate the practical predictability of the forecast event. To enable an accurate assessment of the reliability of the forecast system, a long period of evaluation should be considered.

Since early April in 2015, NCAR has produced daily 10-member convection-permitting ensemble forecasts over CONUS out to 48 hours (http://ensemble.ucar.edu). This forecast system draws ensemble initial conditions from a Data Assimilation Research Testbed (DART) toolkit continuously cycled ensemble analysis that uses the same forecast model to advance member states between cycled analyses. This system will continue operating through at least summer 2016. At the conference, we will give a brief description of the basis for the NCAR ensemble system design, followed by a discussion of areas of storm-scale ensemble design that we are actively investigating, including higher resolution ensembles (1-km horizontal grid spacing), more frequent (hourly) cycling of the ensemble analysis, as well as progress in fully cycled convection-permitting ensemble analysis on a CONUS grid. Each of these aspects will be considered with respect making more reliable mesoscale predictions of convective weather hazards.