Ensemble-Based Tangent Linear Model to Extend Assimilation Window in the Rapid Refresh Forecast System (RRFS)

The Rapid Refresh Forecast System (RRFS) is the next-generation regional ensemble forecast system being developed by NOAA and the Unified Forecast System (UFS) community. As an option of the RRFS development, it is considered to use four-dimensional ensemble-variational (4DEnVar) data assimilation with the overlapping window (Slivinki et al. 2022), where observations not used on time are assimilated in the subsequent analysis. In 4DEnVar, however, static background error covariance is not evolved in the whole assimilation window. Although En4DVar evolves it with tangent linear model (TLM), development of TLM requires high cost. Therefore, ensemble-based TLM (ETLM) suggested by previous studies (e.g., Frolov et al. 2018) may be beneficial for RRFS. This study suggests simplified ETLM that can be implemented in RRFS, and introduces results of data assimilation experiments with the two data assimilation systems based on the toy model (Lorenz 1996) and Grid-Point Statistical Interpolation (GSI). They indicated that this ETLM could make better analyses than those with 3DVar.