Quantifying decision-relevant climate uncertainties in climate model ensembles across multiple spatial and temporal scales

Thursday, 18 December 2014
Ryan L Sriver, University of Illinois, Urbana, IL, United States, Chris E Forest, Earth and Environmental Systems Institute, The Pennsylvania State University, University Park, PA, United States and Klaus Keller, Pennsylvania State University Main Campus, University Park, PA, United States
Climate models are valuable tools for understanding how Earth's climate system is changing, yet they are inherently uncertain. Two key sources of uncertainty include: (1) internal variability affecting simulations from a single model initialized from different states, and (2) structural uncertainties affecting multi-model intercomparisons. These uncertainties can significantly increase the spread of climate model projections and interpretations of potential future change. Here we compare results from a new 50-member climate change ensemble experiment, utilizing the low-resolution configuration of the fully-coupled Community Earth System Model (CESM), with results from the CMIP5 models to diagnose the skill of climate model ensembles that predominantly sample structural uncertainties versus internal variability. The CESM ensemble is comprised of transient hindcasts and projections (1850-2100) using historical forcings and the RCP8.5 projection scenario. The transient simulations are initialized from different initial model states, sampled from a ~10,000 year fully-coupled unforced equilibrium simulation that captures internal unforced variability of the coupled ocean-atmosphere system. We find that both the CESM and CMIP5 ensemble strategies have a significant effect on projections of key climate change metrics, and the projected ranges increase with decreasing spatial scale. Further, the CESM ensemble demonstrates considerable skill in simulating key regional climate processes relevant to decision-makers, such as seasonal temperature variability and extremes. When combined with statistical estimation methods, the CESM ensemble provides a useful framework for analyzing projection uncertainties surrounding regional extremes on decadal timescales. Given the tradeoffs between model resolution and computational cost, our results indicate that the ensemble/statistical methods presented here are a useful resource for quantifying decision-relevant uncertainties and analyzing climate change impacts.