Late Ordovician Climate Reconstruction Based on State-Dependent Climate Sensitivity

The Hirnantian glacial maximum was a brief but intense glacial event that occurred during the latest Ordovician (~445-443 million years ago). It was characterized by global cooling, major ice-sheet expansion over Gondwana, and substantial perturbations to the carbon cycle. Previous studies have combined Earth system models with proxy records to investigate the magnitude of the cooling and to explore the mechanisms linking ocean deoxygenation to the Late Ordovician mass extinction. However, the results of these reconstructions exhibit considerable discrepancies, primarily due to the increasing uncertainty of proxy data with geological age and the difficulty of constraining boundary conditions required by models in deep time. Here we introduce state-dependent climate sensitivity, in which the radiative forcing of atmospheric CO₂ increases with its concentration, to improve the Earth system modelling. We then perform a series of simulations with varying levels of greenhouse gases and nutrients to identify the climate-productivity conditions that plausibly drove the cooling during the Hirnantian glacial maximum. Applying rigorously screened Late Ordovician sea-surface temperature estimates derived from oxygen isotope studies as constraints, alongside a semi-quantitative constraint based on a new compilation of local redox proxies, we identify a plausible scenario of Hirnantian climate and redox changes. Our results show that deep-ocean deoxygenation during the Hirnantian was driven by a combination of cooling and changes in ocean nutrient inventory, and that temperature-driven microbial respiration can reconcile the spatial distribution of seafloor anoxia as reconstructed, providing new insights into the decoupling of redox conditions between the surface and deep waters. In addition, our simulations suggest that Late Ordovician atmospheric CO₂ levels before cooling may have been substantially overestimated (up to 6,720 ppm according to previous studies), likely due to a fixed climate sensitivity assumed in previous modelling studies. This overestimation may not be limited to this event, but could also affect climate simulations of other periods.