Coupled atmosphere-ocean data assimilation experiments with a low-order model and CMIP5 model data

Coupled atmosphere-ocean data assimilation (DA) experiments are performed for estimating the Atlantic meridional overturning circulation (AMOC). Recovery of the AMOC with an ensemble Kalman filter is assessed for a range of experiments over observation availability (atmosphere, upper and deep ocean) and for assimilating high-frequency observations compared to time averages. For an idealised low-order coupled climate model, the traditional DA approach using an ensemble of model trajectories to estimate covariances is compared to a simplified "no-cycling" approach involving climatological covariances derived from a single long model integration. Robustness of the no-cycling method is also tested on data from a millennial-scale simulation of a comprehensive coupled atmosphere-ocean climate model. Results show that the no-cycling approach provides a good approximation to the traditional approach, and that assimilation of time-averaged observations improves AMOC recovery using drastically smaller ensembles than would be required for the case of instantaneous observations. Even in the limit of no ocean observations, the no-cycling approach is capable of recovering the low-frequency AMOC with time-averaged observations; assimilation of noisy instantaneous atmospheric observations fails to recover decadal-scale AMOC variability.