Strongly Coupled EnKF Data Assimilation in Coupled Ocean-Atmosphere Models

Current operational coupled ocean-atmosphere models handle data assimilation (DA) separately for the two domains, in what is called weakly coupled DA. Previous studies using simple 1-D coupled models (e.g. Lu et al. 2015, Liu et al. 2013) have shown that improvements are made by assimilating the entire system as a whole using strongly coupled DA. The inclusion of cross-domain background error covariance allows observations in the atmosphere to directly improve the ocean analysis, and vice versa, and a smooth transition of ensemble weights between the atmosphere and ocean results in a more balanced analysis.

Assimilating cross-domain observations adds difficultly to a DA system. However, we show that a system can easily be built using a set of connected Local Ensemble Transform Kalman Filters (LETKF), one for each domain. The nearly "black-box" nature of the LETKF, as well as its existing code layout, make it advantageous in the design of a computationally efficient strongly coupled DA system, and can easily be used for coupled models other than ocean-atmosphere models. Using an intermediate complexity ocean-atmosphere model (SPEEDY-NEMO), observing system simulation experiments show significant improvements when strongly coupled DA is used. We also discuss the development of a strongly coupled EnKF for the Climate Forecasting System v2 (CFS-LETKF).