Since 2004, ESRL-GSD has been producing real-time probabilistic forecasts of convective from spatially filtered, time-lagged ensembles of deterministic Rapid Update Cycle (RUC) model forecasts. Known as the RUC Convective Probability Forecast (RCPF), these forecasts have been made available in real-time to the Aviation Weather Center (AWC) since 2005. Forecasters at the AWC use the RCPF as guidance in the generation of a strategic thunderstorm likelihood forecasts for use by the U.S. aviation industry. In late 2005, an ensemble-based potential echo top forecast product was added to the RCPF to help provide guidance on when aircraft may be able to fly over low-topped thunderstorm areas.

This presentation will focus on enhancements to the RCPF associated with a specific improvement to the underlying RUC deterministic model, the addition of a radar reflectivity data assimilation procedure. This new procedure for assimilating radar reflectivity data within the hourly updated Rapid Update Cycle (RUC) system has been developed and tested and is producing significant improvements for short-range forecasts of precipitation systems. The new procedure ingests the National Severe Storms Laboratory (NSSL) national radar reflectivity mosaic data and forces convection in echo coverage regions and suppresses convection in echo-free regions. Forcing of convection is accomplished by applying latent heating in radar echo coverage regions during a pre-forecast diabatic initialization procedure. Application of this latent heat derived temperature tendency induces an associated vertical circulation, with low-level convergence and upper-level divergence. Suppression of convection is accomplished by inhibiting the cumulus parameterization during the first 30 min of the model forecast within an echo-free area diagnosed from the reflectivity data.

Because the new reflectivity assimilation procedure projects onto both parameterized and explicit precipitation within the RUC model, the original RCPF algorithm (which relied only on parameterized precipitation) does not fully capture the improvement in the RUC deterministic forecasts. We are testing and evaluating two approaches to make the RCPF more fully reflect the RUC improvements. The first is to include both parameterized and grid-scale precipitation (appropriately thresholded) as predictors for the RCPF and the second is to use model simulated reflectivity as the predictor. When either grid-scale precipitation or simulated reflectivity are used, an instability mask (from the model predicted fields) must also be used to differentiate between convective and stratiform precipitation regions.

At the conference I will presents results from the modified RCPF algorithm and discuss a number of planned enhancements. These enhancements include verifying and calibrating the potential echo-top forecasts, improving the statistical reliability of the RCPF, and further improvements to the RCPF be making use of sub-hourly RUC model output and 3-km deterministic model forecasts from the High Resolution Rapid Refresh.