Estimating Uncertainty in Global Climate Projections using Gaussian Error Propagation

Current techniques to estimate uncertainty in projections of future climate emphasize multi-model or single model ensembles, which require huge computational resources and are impractical except for large integrated efforts like the IPCC Assessment Reports. We developed techniques to estimate uncertainty for a single climate projection by comparison with observations and Gaussian error propagation. We demonstrate the technique for a climate projection from 1850 to 2300 for the A1B scenario using Community Earth System Model (CESM). We estimated overall model bias and uncertainty by comparing simulated and observed global climate sensitivity (change in temperature per change in atmospheric CO2 concentration). We estimated the relative contribution of uncertainty in fossil fuel emissions and the Permafrost Carbon Feedback to total uncertainty in simulated global surface air temperature. We verified our results by comparing to a more traditional ensemble of CESM perturbation simulations. Our results indicate that uncertainty in the reported fossil fuel emissions dominate total uncertainty in simulated surface air temperature. These results demonstrate the feasibility of estimating uncertainty for a single climate projection without multi-model or single model ensembles.