Does Model Calibration Reduce Uncertainty in Climate Projections?

Abstract Uncertainty in climate projections is large as shown by the likely uncertainty ranges equilibrium sensitivity (ECS) of 2.5–4 K and transient response (TCR) 1.4–2.2 K. model could arise from way which unresolved processes are represented, parameter values used, or targets for calibration. We show that, two ensembles that were objectively calibrated to minimize differences observed large-scale atmospheric climatology, uncertainties ECS TCR about 2–6 times smaller than CMIP5 CMIP6 multimodel ensemble. also find projected surface temperature, precipitation, annual extremes relatively small. Residual largely arises unconstrained sea ice feedbacks. The more 20-year-old HadAM3 standard configuration simulates hemispheric-scale observations preindustrial temperatures well median while optimized configurations simulate these better almost all models. Hemispheric-scale not systematically simulated although ensemble seems patterns configurations. Our results suggest most CMIP models be improved their simulation systematic However, (for a given scenario) choice parameterization schemes (“structural uncertainty”), with different tuning another possible contributor. Significance Statement Climate represent phenomena controlled uncertain parameters. Changes parameters impact how current its projections. Multiple calibrations single model, using an objective method, performed. These produce very similar at both global regional scales. An analysis combines has small showing this constrain Recently developed have broad range abilities only some improvement ability despite decade development.