Ensemble forecasts of severe convective storms

Deterministic high-resolution numerical weather prediction models that explicitly produce convective storms have been run in real-time demonstrations during the last several years. These forecasts have provided valuable guidance about convective storm timing, location, and mode. An alternative to deterministic forecasting is ensemble forecasting. Ensemble forecasting has the potential to convey probabilistic information about the forecast and to identify storm scenarios that are relatively difficult to anticipate.

As part of the NOAA Hazardous Weather Testbed 2007 Spring Forecasting Experiment, a 30-member mesoscale short-range ensemble forecasting system using the Weather Research and Forecasting (WRF) model was developed for the contiguous 48 states at a horizontal grid spacing of 30 km (described in a companion abstract). The current study explores the possibility of using such mesoscale ensembles to initialize high-resolution ensemble convective forecasts. Specifically, the mesoscale ensemble analyses and forecasts provide initial and boundary conditions for ensemble forecasts employing a 3-km horizontal grid spacing.

Retrospective 3-km WRF ensemble forecasts with explicit convection have been produced for several severe weather cases in spring 2007, including 28 March 2007 (high plains tornado outbreak) and 4 May 2007 (Greensburg, Kansas tornadic storm). Ensemble forecasts have been produced out to about 6 hours, or roughly the same lead time as for severe weather watches. Some of the issues being explored are ensemble design (multi-parameterization approaches, initial condition perturbations, and boundary condition perturbations) and the influences that assimilating surface observations have on the ensemble forecasts. Initial results suggest that these convective-scale ensembles provide reasonable probabilistic information that could be used by forecasters for purposes such as issuing watches.