Probing Uncertainty in Aerosol Radiative Forcing: Sensitivity Analysis with a Single Climate Model

Total uncertainty in direct and indirect aerosol radiative forcing (ARF) arises through empirical uncertainty in aerosol emissions, chemical composition (and corresponding radiative properties) and microphysical interactions with clouds, along with the imperfect representation of all of these aspects in climate models. Recent estimates suggest the range of ARF may lie between -1.6Wm-2 and +0.6Wm-2, which translates into significant uncertainty in model-derived estimates of climate sensitivity to increasing GHG concentrations; i.e., ARF and climate sensitivity are correlated. Given recent and projected trends for the dramatic reduction of global aerosol emissions in the 21st century (almost independent of GHG emissions; e.g. in the RCP scenarios), better understanding of the climate and policy implications of this uncertainty is necessary.

Climate model intercomparison projects (MIPs), such as CMIP3 and CMIP5, are the current standard method for quantifying uncertain climate parameters like ARF. However, using MIPs to identify specific climatic behaviours attributable to specific ARF values is challenging, because the models differ in much more than just their representation of ARF. Moreover, given that current models potentially suffer similar biases in aerosol representation, conventional MIP methods could be limited in their ability to sample the full empirical range of ARF.

In this study, we present a new sensitivity analysis protocol for individual coupled ocean-atmosphere climate models (in this case the NCAR Community Earth System Model v1) to sample the full range of empirical ARF uncertainty. In particular, the model's parameterized radiative properties of sulfate and black carbon aerosols are varied to capture the full range of ARF for both species. Versions of the model with ARF values spanning the range of empirical uncertainty are then retuned to reproduce the observed historical evolution of 20th Century global mean temperatures. Through analyses of Committed Warming and RCP scenario projections for the models realisations with different ARF values, we aim to determine the sensitivity of global and regional climate patterns to empirical ARF uncertainty within a single climate model.