Improving Background Error Covariance for Data Assimilation in the Rapid Refresh Forecast System (RRFS)
Dr. Sho Yokota
JMA Visiting Scientist at EMC
20 Mar, 2pm
Abstract:
The Rapid Refresh Forecast System (RRFS) is the next-generation regional ensemble forecast system consisting of hourly data assimilation and forecasts with the FV3 limited area model covering the North America region. To improve the ensemble and static background error covariances (BECs) in the variational data assimilation in the RRFS, the following were implemented in a prototype of RRFSv1 and tested.
1. Scale-dependent localization (SDL, Buehner and Shlyaeva 2015, Tellus) to apply the larger localization length for ensembles of large-scale waveband.
2. Variable-dependent localization (VDL, Wang and Wang 2023, MWR) for simultaneous assimilation of conventional and radar reflectivity data.
3. Ensemble-based tangent linear model (ETLM) to evolve the static BEC in time for the long assimilation window.
4. Convective-scale static BEC including hydrometeors and cross-variable covariance between control variables (Wang and Wang 2021, MWR) for radar reflectivity assimilation.
5. The Multigrid Beta Filter (MGBF, Purser et al. 2022, MWR) for more efficient parallel calculation of localization.
In this seminar the implementation and impacts of these methods are introduced and discussed.