The temperature of the deep ocean is a robust proxy for global mean surface temperature during the Cenozoic

Reconstructions of global mean surface temperature (GMST) are one of the key contributions that palaeoclimate science can make to societally-relevant questions, for example, by providing the information required to derive equilibrium climate sensitivity from the geologic record and as a means of testing climate model performance under warmer-than-present conditions. One relatively simple method of doing so is to parameterize GMST as a function of the temperature of the deep ocean, which has the advantage that deep ocean temperature is relatively well constrained for much of the Cenozoic. A commonly-used transformation approach is based on a 1:1 deep ocean-GMST scaling factor prior to the Pliocene, which is a simple assumption, but to our knowledge, without a firm mechanistic basis. Here, we test the reliability of this assumption using output from a suite of climate model simulations, including those from the DeepMIP project, as well as curated data compilations for well-studied intervals throughout the Cenozoic. Our analysis demonstrates that a simple 1:1 scaling factor is likely to be a good approximation for much of the Cenozoic, possibly mechanistically rooted in an increasing winter bias in deep water formation offsetting an increase in polar amplification/stratification during intervals of global warmth. Building on this, we reevaluate the Cenozoic records of deep ocean temperature and derive a new, continuous record of GMST. Our record is substantially warmer than the most common previous approach for much of the Cenozoic, from which we derive GMST during the early Eocene Climatic Optimum of 31.3±1.3°C, supporting the notion of a greater-than-modern ECS in this past warm climate state.