TOWARD AN OMNI-PROXY RECONSTRUCTION OF CENOZOIC CO2 (Invited Presentation)

Atmospheric CO₂ links Earth’s carbon cycle, climate, and biotic systems. As such, knowing how CO₂ concentrations have varied over Earth history is fundamental to understanding and learning from the geologic record. In the absence of direct evidence prior to the ice core record, numerous paleo-CO₂ proxies have been developed and used to reconstruct paleo-CO₂ on multi-million-year time scales. Each of these proxy systems involves complex, interactive sensitivities to multiple environmental and biological factors. As a result, quantitative interpretation and robust error propagation are challenging for individual proxies, let alone multi-proxy reconstructions.

As a component of the community-engaged CO₂ Proxy Integration Project (CO2PIP, https://paleo-co2.org/) we are working toward comprehensive, multi-proxy reconstructions of paleo-CO₂ change within the Bayesian Joint Proxy Inversion (JPI) framework. The core of these reconstructions is a set of Proxy System Models (PSMs) that describe the response of four widely used classes of paleo-CO₂ proxies to physical and biological forcing. Development of the PSMs has involved re-framing and integrating existing perspectives on proxy interpretation and highlights opportunities for future refinement of our understanding of these systems. Our Bayesian analysis consists of PSMs coupled with time- and/or space-explicit models of forcing variables, including atmospheric CO₂ concentration, and a model for the age and sampling uncertainty of proxy observations. This system is inverted using Bayes’ Theorem to produce a posterior sample of the model state-space conditioned on compiled proxy data.

I will illustrate results from early applications of JPI to 1) reconstruct the Cenozoic CO₂ history using reduced-order PSMs, and 2) reconstruct early Paleogene CO₂ timeseries using a multi-proxy marine dataset. These examples show the power of Bayesian proxy inversion to facilitate propagation of uncertainty within and across these complex proxy systems and develop quantitative reconstructions. The results suggest strong and consistent Earth System Sensitivity to CO₂ forcing across a range of Cenozoic timescales and show that it is likely that modern levels of CO₂ (~420 ppm) have not occurred on Earth since the mid-Miocene.