Using Late Pleistocene SST reconstructions to constrain future greenhouse warming

Global warming projections for a given anthropogenic greenhouse gas concentration scenario still exhibit a substantial spread.  In order to constrain this spread and to provide robust warming projections, our understanding of Earth's climate sensitivity needs to be further improved. Here, we employ a global network of 64 paleo-proxies of SST to reconstruct global-mean SST variations during the Last Glacial Cycle. This temperature reconstruction is then used as a target function for 25 transient model simulations conducted for the same period with 25 different climate sensitivities. Our combined proxy/model approach allows us to determine an optimal range of model climate sensitivities corresponding to a minimum of the weighted mean squared error calculated for reconstructed and simulated global-mean SSTs. Based on our best estimate, Earth's averaged Late Pleistocene equilibrium climate sensitivity is in the order of ~4.2 K per CO₂ doubling with an associated transient climate response of ~2.4 K/2xCO₂. The latter value translates into a global-mean surface warming of about 5.0 K by the year 2100 (relative to pre-industrial levels) based on the Representative Concentration Pathway 8.5. This warming estimate is in excellent agreement with the ensemble-mean projection of climate simulations conducted as part of the Coupled Model Intercomparison Project Phase 5 (CMIP5). Error bars resulting from uncertainties in aerosol and ice-sheet forcing as well as in temperature reconstruction clearly document the current limitations for paleo-based constrains of both climate sensitivity and future greenhouse warming and demonstrate the need for more robust forcing and temperature reconstructions that can be utilized to narrow down the spread in global warming projections.