Improving Weather Guidance for Aviation Decision-Making: Enhancements to Rap and HRRR Deterministic and Ensemble Prediction Systems, including Ensemble-Based Data Assimilation, Forecast, and Post-Processing Components

Short-range weather forecasts from frequently updated numerical weather prediction models underpin much of the aviation-related weather guidance. The hourly updated Rapid Refresh (RAP) and High Resolution Rapid Refresh (HRRR) model systems provide a key source of this numerical weather guidance. In this presentation, a summary of the current and planned work to improve the RAP and HRRR system will be presented, with an emphasis on ongoing work to development ensemble-based weather guidance. This work includes ensemble data assimilation (DA), ensemble forecasting, and ensemble post-processing components, which will improve both deterministic and ensemble forecast systems and lead to the generation of improved ensemble guidance information to better convey forecast uncertainty information. The presentation will begin with a summary of the RAPv4 / HRRRv3 upgrade package, for which the testing and development work is complete and for which NCEP operational implementation is planned for early 2018. This package includes a number of model and data assimilation enhancements, resulting in forecast improvements for nearly all fields (surface, upper-air, reflectivity, clouds, etc.), levels, and forecast lengths. Also planned for inclusion in this package is the addition of a 3-km Alaska HRRR domain.

The remainder of the presentation will focus on ongoing ensemble work, including a description of the HRRR Ensemble (HRRRE) system that has been running in a real-time experimental configuration for several months and been evaluated in several different testbed programs. The real-time HRRRE includes a 36-mumber storm-scale ensemble DA system, which is used to initialize a 9-member storm-scale ensemble forecast system and has also been used as part of the initialization for a parallel deterministic HRRR. A number of scientific issues will be discussed, including benefits of the storm-sale ensemble DA to both the ensemble and deterministic forecast systems, use of stochastic physics and other techniques to increase ensemble spread, and work to obtain the optimal blend of storm-scale and larger-scale spread information. Finally, ongoing work to develop an ensemble post-processing system will be described. This system is being designed for use with different models and ensembles (including time-lagged ensembles), and will provide bias correction, member aggregation, and generation of reliable probabilistic information, as well as other ensemble guidance information. This work has also included obtaining feedback from forecasters on use of the ensemble-based products and recommendations for improving them.