This study investigates how the balance between wind and mass is treated in data assimilation, and how it affects the quality of the model states in an analysis–forecast cycle. This is done in terms of the dependence of balance on latitude and on the type of variable. The impact of the nonlinear balance equation is compared with regressed wind–mass balance in the 3‐Dimensional Variational data assimilation (3D‐Var) system of the Korea Institute of Atmospheric Prediction Systems (KIAPS). Its impact is significantly positive in temperature rather than in wind, despite the two‐way influence of the cross‐correlation between wind and mass, in terms of the root‐mean‐square difference (RMSD) of 6 h forecasts against the ERA‐Interim reanalysis data. This temperature effect is observed in the southern hemispheric (SH) polar jet of the mid‐troposphere, the SH midlatitudinal jet, and the mid/lower stratosphere in the Tropics, where there is strong zonal mean flow. Although the zonal wind forecast was harmed by application of the nonlinear balance, the temporal consistency of the damage is relatively weak compared to the improvement by the nonlinear balance in the temperature forecasts. In the SH midlatitudinal jet and the mid/lower stratosphere in the Tropics, the nonlinear balance equation, including the advection term, improves the quality of temperature RMSDs in the analysis–forecast cycle by imposing the proper balance in the initial conditions. However, in the SH polar jet of the mid‐troposphere, where the observation density is relatively low, the nonlinear balance equation achieves the same effect by reducing the analysis error (i.e. generating initial conditions more accurately). The nonlinear balance equation contributes to robustly improving the model states of the analysis–forecast cycles, depending on the dynamical activity and the observation density of the corresponding regions.

No abstract

No abstract