Storm-scale data assimilation and ensemble forecasting for Warn-on-Forecast

The NOAA Warn-on-Forecast research project is tasked with the development of very short-range (0-1 h) probabilistic forecasts that accurately predict severe convective storms. A component considered essential for a Warn-on-Forecast system is a regional convection-allowing (grid spacing ~3 km) ensemble, which could serve as the background for nested very-high-resolution, event-dependent grids.

Convection-allowing ensemble analyses and forecasts will be produced for recent severe weather events over a South Plains domain with a horizontal grid point spacing of 3 km. The ~50-member ensemble is constructed from initial and boundary conditions provided by members of the Global Ensemble Forecast System (GEFS) forecast cycle starting at 0000 UTC. The WRF physics options are also varied amongst the ensemble members to further populate the ensemble, and to address deficiencies in model physics. The same observations that are assimilated into NOAA's Rapid Refresh model will be assimilated hourly into the convection-allowing ensemble starting at 0000 UTC. The data will be assimilated with an ensemble Kalman filter (DART software) coupled with the Gridpoint Statistical Interpolation (GSI) software for preprocessing observations and computing prior ensemble estimates. The ensemble also employs radar reflectivity observations from the NSSL National Mosaic to initiate observed storms that are absent in the model forecasts and to suppress unobserved regions of simulated storms.

This ongoing research is currently focused on the tornadic periods spanning 22-24 May 2011 and 15-20 May 2013. Preliminary results from one or two days will be presented. Of particular interest is the ability of the ensemble to spin-up realistic mesoscale structures in the near-storm environment, as well as producing realistic convective structures in proper locations.