A Simultaneous Multiscale Data Assimilation Using Scale-Dependent Localization in GSI-Based Hybrid 4DEnVar for NCEP FV3-Based GFS

Abstract A scale-dependent localization (SDL) method was formulated and implemented in the Gridpoint Statistical Interpolation (GSI)-based four-dimensional ensemble-variational (4DEnVar) system for NCEP FV3-based Global Forecast System (GFS). SDL applies different to scales of ensemble covariances, while performing a single-step simultaneous assimilation all available observations. Two variants with (SDL-Cross) without (SDL-NoCross) considering cross-wave-band covariances were examined. The performance two- three-wave-band experiments (W2 W3, respectively) evaluated through 1-month cycled data experiments. improves global forecasts 5 days over scale-invariant including operationally tuned level-dependent (W1-Ope). W3 SDL-Cross experiment shows more accurate tropical storm–track at shorter lead times than W1-Ope. Compared W2 experiments, counterparts applying tighter horizontal medium-scale wave band generally show improved below 100 hPa, but degraded above 50 hPa. While outperformance SDL-NoCross versus hPa lasts days, that 3 days. Due local spatial averaging may alleviate sampling error, slightly better times. However, outperform longer times, likely from retention heterogeneity resultant analyses balance. Relative are consistent forecasts.