1. ^* Earth Systems Group, Department of Applied Science, New York University, New York 10003, USA
2. ^† Lawrence Livermore National Laboratory, Livermore, California 94551, USA

Abstract

To assess the future impact of anthropogenic greenhouse gases on global climate, we need a reliable estimate of the sensitivity of the Earth's climate to changes in radiative forcing. Climate sensitivity is conventionally defined as the equilibrium surface temperature increase for carbon dioxide doubling, T _2x. Uncertainties in cloud processes spread general circulation model (GCM) estimates of this parameter over the range 1.5< T _2x <4.5°C (refs 1, 2). An alternative to model-based estimates is in principle available from the reconstruction of past climates^3–6, which implicitly includes cloud feedback. Here we retrieve the sensitivity of two palaeoclimates, one colder and one warmer than present, by independently reconstructing both the equilibrium surface tem-perature change and the radiative forcing. Our results yield T _2x = 2.3 0.9 °C. This range is comparable with estimates from GCMs and inferences from recent temperature observations and ocean models^7,8. Future application of the method to additional climates in the geological record might constrain climate sensitivity enough to narrow the model uncertainties of global warming predictions.