Evidence for higher Earth-system sensitivity from long-term carbon-cycle observations

Projecting global temperature changes and developing sound strategies to address challenges posed by climate change depend strongly on Earth system sensitivity (ESS). ESS, the long-term (e.g., millennial-scale and longer) global warming due to a doubling of atmospheric CO₂ levels relative to pre-industrial conditions, is an important Earth system property and is highly uncertain. Much of the Phanerozoic (541 Ma to present) was much warmer than today, therefore information on deep-time ESS can be particularly useful in projecting long-term climate change in the future. Previous work, however, has been largely silent on the effects of key uncertainties surrounding model forcing, correlated model parameters and the asymmetric nature of observational CO₂ constraints. Here, we improve on ESS estimates by assimilating deep time paleoclimate information with a long-term carbon cycle model to address these questions. We assimilate multiple long-term CO₂ proxy data sets to constrain Earth-system sensitivity over the past 420 million years. Using a sophisticated model parameter calibration approach and a detailed sensitivity analysis for the model parameters, we find the updated ESS estimate shows a narrower range, but higher median, than previous assessments (we raise the median from 3.8^oC to 4.3 ^oC per doubling of CO₂). These findings imply increased future temperatures and risks in the presence of ice sheets on millennial timescale. We find that the chemical and plant-assisted weathering parameters interact strongly with ESS in affecting the simulated atmospheric CO2. Research into improving the understanding these weathering processes hence provides potentially powerful avenues for further constraining this fundamental and policy relevant Earth system property.