On geological time-scales, the production and degree of recycling of biogenic carbon in the marine realm and ultimately its removal to sediments, exerts a dominant control on atmospheric CO2 and hence variability in climate. This is a highly complex system involving a myriad of inter-connected biological, chemical, and physical processes. For this reason alone, linking observations, often highly abstracted in the form of proxies, to the primary processes involved and ultimately to explanatory hypotheses for specific geological events and transitions, is challenging. The past few decades have seen a progressive improvement in theoretical and process-based understanding of the various components that make up the marine carbon cycle and, hand-in-hand with this, the development of numerical model representations of the complete system. Models have also been designed and/or adapted with paleoclimate questions in mind and applied to quantitatively explore the role of the marine carbon cycle in both perturbations and long-term geologic evolutionary trends in global climate, and possible feedbacks between them. However, we must ask whether paleoclimate models incorporate sufficiently appropriate representations of the dynamics and sensitivities of the marine carbon cycle, and indeed, whether in the geological context, we really know what these dynamics are.
Here we provide a comprehensive overview of how marine carbon cycling and the biological carbon pump is treated in available paleoclimate models, with the aim of critically evaluating their ability to help interpret past marine carbon cycle and climate dynamics. To this end, we first provide an overview of commonly used paleoclimate models and some of their associated paleo-applications, drawing from a wide range of global carbon cycle box models and Earth system Models of Intermediate Complexity (EMICs). Secondly, we review and evaluate the three dominant processes involved in the cycling of organic and inorganic carbon in the marine system and how they are represented in models, namely: biological productivity at the ocean surface, remineralisation/dissolution of particulate carbon within the water column, and the benthic-pelagic coupling at the seafloor. We generate and employ illustrative examples using the model GENIE to show how different parameterisations of water-column and sediment processes can lead to significantly different model projections. Our compilation reveals the prevalence of static parameterisations of marine carbon cycling among existing paleoclimate models, which are commonly empirically derived from present-day observations. Although such approaches tend to represent carbon transfer in the modern ocean well, they are potentially compromised in their ability to reflect the true degree of freedom and strength of feedbacks with respect to past climate events, particularly those characterised by environmental boundary conditions that differ fundamentally from today. Finally, we discuss the importance of using models of different complexities and how questions of model uncertainty may start to be addressed.