Development and evaluation of Proxy Surrogate Reconstructions as a tool for characterizing Common Era AMOC variability.

The Atlantic Meridional Overturning Circulation (AMOC) latitudinally redistributes mass and heat, and observed inter- and intra- annual variations in its strength can have widespread climatic impacts. The centennial-scale response of AMOC to external radiative forcing cannot be evaluated within the brief observational record, but can be inferred from the geochemical variations of biogenic material preserved in ocean sediments, and from realistically-forced atmosphere-ocean general circulation model (AOGCM) simulations. Both sources of paleoclimate information are imperfect however, suggesting that applying them in combination may better approximate the true evolution of the climate system, and better characterize the uncertainty in reconstructions. Here we describe how proxy surrogate reconstructions (PSRs) use both paleoclimate data and simulations to estimate Common Era AMOC variability through a simple analog approach. We use Mg/Ca, alkenone unsaturation index, and oxygen isotope records, compiled as part of the PAGES Ocean2K and PAGES2k v2.0 databases, and realistically forced isotope-enabled AOGCM simulations that have been observationally mapped via proxy system modeling (PSM), such that spatiotemporally local misfit to the paleoclimate observations are minimized. As part of this development, we present new, independently-validated, PSMs for Mg/Ca in five foraminifera species based on globally-distributed core-top data. Results exhibit limited validated reconstruction skill, but suggest centennial-averaged ensemble-weighted mean meridional overturning streamfunction anomalies during the Common Era do not exceed 1%. Idealized, but realistically formulated pseudoproxy experiments designed to evaluate these results suggest that simulation-derived best analogs of the available paleoclimate data do not skillfully capture AMOC variability. This result may reflect limitations in the current network of paleoclimate data, the AMOC-related information in that network, and/or model-specific patterns in AMOC's simulated spatial footprint.