Evaluating Glacial CO₂ System Reconstructions From Benthic Proxies in Earth System Simulations

Reconstructing marine dissolved inorganic carbon (DIC) across glacial cycles is critical for understanding the sensitivity of the marine carbon sink to natural climatic change. Published estimates of DIC invoke linear relationships between DIC and CO^2/₃^-, apparent oxygen utilization (AOU), or δ¹³C. These relationships are based on conceptual models and correlations from modern spatial tracer gradients. However, it remains unclear whether the spatial correlations also hold for temporal change. Here, we apply these empirical methods to Earth system model results to test their applicability to transient glacial-interglacial changes. The model uses established, experimentally-constrained carbonate system solvers and explicitly tracks the various components of the carbon cycle (e.g., DIC, alkalinity, temperature, CO₂). Predicting simulated DIC from simulated CO^2/₃^-, AOU, orδ¹³C often results in large prediction errors. The interplay of the carbonate system, ocean circulation, and biologically-mediated DIC and alkalinity re-distributions creates a system too complex to be captured by existing empirical methods. Specifically, large local alkalinity changes can arise due to circulation and export production changes even without substantial changes in global mean alkalinity. Consequently, reconstructed CO^2/₃^- constrains DIC changes insufficiently. Similarly, marine AOU or δ¹³C are not reliable proxies of remineralized DIC. Furthermore, DIC changes are not a direct metric for atmospheric CO₂ drawdown even without considering changes in global mean alkalinity because of net carbon exchange with sediments and the land biosphere. We suggest that spatially-resolved, transient Earth system simulations may provide a more reliable means of estimating carbon cycle shifts observed in proxy data than current empirical methods.

Plain Language Summary

Atmospheric  CO₂ varied substantially over glacial-interglacial cycles as shown by ice core data. The exact causes of these past CO₂  variations remain unknown but are shown to be related to large-scale reorganizations of the ocean carbon cycle. A difficulty is that past ocean carbon cannot be directly measured but is estimated from proxy data. We tested three published methods of calculating how much carbon was dissolved in the ocean during ice age cycles. For this, a model of the Earth system is applied to represent plausible real-world processes of the ocean carbon cycle and ocean-sediment interactions. The results show that none of the methods is generally correct because the way that carbon and proxy signals in the ocean change over time is more complex than the carbon-proxy correlations inferred from modern ocean observations. We discuss the processes that create this complexity and suggest the best way forward is to estimate carbon changes in the ocean from proxy data by using spatially resolved Earth system models rather than simple correlations.