The Importance of Process-oriented Evaluation to Reduce Aerosol-cloud Radiative Forcing Uncertainty

Abstract:

Measurements of atmospheric and climate state variables provide the ultimate constraint on highly complex and uncertain model simulations. It is assumed that extensive and accurate measurements will enable a reduction in model uncertainty, although this is rarely demonstrated rigorously. Here we attempt to reduce the uncertainty in the aerosol-cloud forcing, one of the largest uncertainties in historical forcing of climate, by constraining a global model using idealised aerosol measurements. We show that the modelled forcing is poorly constrained compared to modelled aerosol because the aerosol measurements constrain only the relationships between multiple compensating parameters rather than individual parameter values. Although aerosol concentrations and aerosol-cloud forcing are controlled by the same parameters, the relationships between these parameters can be very different for the two quantities. Thus, much of the multi-dimensional aerosol-cloud forcing parameter space remains unconstrained. Our results show that while matching model state to observations is a necessary condition of model evaluation, it is not sufficient to constrain aerosol cloud forcing and process-oriented evaluation is also needed to constrain the model parameter values. We demonstrate how a sensitivity analysis of the model being constrained can identify the key compensating processes that require more detailed evaluation if uncertainty in aerosol-cloud forcing is to be further reduced.