Balance theories, such as the mid-latitude quasi-geostrophic theory, are crucial to operational data assimilation systems. As large-scale motions in the mid-latitude atmosphere are approximately in geostrophic balance and largely devoid of fast unbalanced gravity wave motions, the analyses produced by data assimilation systems must respect these important balances. Balance theory aid in the construction of the background error covariances, which controls the level of imbalance in analysis. Moreover, a temperature observation provides information to both temperature and wind field through balance relations. In contrast, due to the lack of valid balance models in the tropics, assimilation is currently carried out in a univariate fashion, i.e. a temperature observation only results in an update to the temperature field.
In this talk, I will describe how an equatorial balance theory can be used to improve data assimilation in the tropics in the context of Ensemble Kalman filters (EnKF). In EnKF the background error statistics is constructed by sampling from a small ensemble, but the method requires localization to suppress spurious long-range and cross-variable correlations. Previous research have shown that localization can lead to significantly unbalanced analyses. Through a series of identical twin experiments, I will show how the equatorial balance theory can be used to reduce spurious imbalance and improve accuracy of the analyses.