Recent Improvements of GSI 3DVAR-Ensemble Hybrid Data Assimilation System for Rapid Refresh and High Resolution Rapid Refresh

NOAA's Rapid Refresh forecast system (RAP) is an hourly-updated regional data-assimilation and forecasting system that uses the Weather Research and Forecasting (WRF-ARW) model with 13-km horizontal grid spacing and the Gridpoint Statistical Interpolation (GSI) analysis package. High Resolution Rapid Refresh (HRRR) is a 3-km grid data assimilation and forecast system one-way nested inside RAP. The RAP version 2 has been running operationally at NCEP since February 2013 with enhancements including a 3DVAR-Ensemble hybrid assimilation procedure within GSI with the GFS EnKF global ensemble forecasts. In 2013, we focused on testing and tuning a number of GSI hybrid configurations based on the needs of RAP, including using EnKF ensemble forecasts at one hour intervals, tuning the ratio of static and ensemble background error (BE), allowing the BE ratio vary vertically, and finding better vertical and horizontal localizations for this mesoscale application.

The current application of GSI hybrid data assimilation within the RAPv2 has significantly improved mid- and upper-level wind and moisture forecasts, but we anticipate further improvement replacing GFS-ensemble forecasts with regional RAP ensemble forecasts within the GSI hybrid-ensemble assimilation. In 2014, we continue tuning and testing the configurations for GFS ensemble based GSI-Hybrid analysis but also invested significant effort to build RAP ensembles, which are initialized from GFS EnKF ensemble members, to feed into RAP GSI hybrid system with a goal of improving RAP forecasts of near-surface fields and localized weather phenomena including cloud/hydrometeor fields. In 2014, we also started to apply GSI hybrid analysis on HRRR system for real-time test. Radar reflectivity assimilation is applied in both RAP and HRRR as a key component of the RAP assimilation using specification of latent-heating within a forward-backward digital filter initialization at 13km and in forward mode only at 3km.