Estimation and Prediction of the AMOC from Multi-timescale Data Assimilation

Estimating space-time variability in the Atlantic meridional overturning circulation (AMOC) is critical both to understanding the dynamical interaction between components of the climate system and to predicting climate change. Addressing these issues is challenged by the fact that the AMOC is poorly observed on decadal and centennial timescales, limiting our ability to document and understand the variability, and to assess the accuracy of forecasts over a wide range of historical conditions. The key to both the variability and the prediction problems lies in objective estimates of the AMOC over long periods of time from sparse and noisy measurements. Here we present a novel data assimilation technique that operates across multiple time scales to robustly estimate and predict the AMOC. We find that the AMOC is most skillfully reconstructed by incorporating observations of the climate system across multiple time scales compared to using observations at short (monthly to annual) or long (decadal) time scales alone. Moreover, we find that the AMOC may be estimated from atmospheric observations or proxies alone. Finally, the recovered states allow for estimates of the AMOC predictability timescale.