The magnitude of aerosol effective radiative forcing is the dominant source of uncertainty in net anthropogenic forcing over the industrial period. Aerosol effective radiative forcing is also one of the largest sources of uncertainty in recent decadal and near-future anthropogenic forcing. Knowledge about the main sources of aerosol forcing uncertainty can be used to guide the development of models and ultimately reduce forcing uncertainty. The research in this thesis identifies important parametric sources of aerosol radiative forcing uncertainty in global models
using perturbed parameter ensembles, statistical emulation and variance-based sensitivity analyses.

Industrial and recent decadal anthropogenic emission periods are used to quantify the sources of aerosol forcing uncertainty over different timescales. Natural aerosol parameters dominate the uncertainty in aerosol forcing over the century-scale industrial period. However, anthropogenic and model process parameters are dominant over recent decades. In each case specific parameters have been identified as priorities for model development that targets aerosol forcing uncertainty reduction.

At the regional scale changes in climatic effects over recent decades may be partly attributable to anthropogenic aerosol forcing. The credible ranges of aerosol radiative forcing, quantified in 11 climatically important regions, support some hypotheses about the role of aerosols in regional climate forcing and call others into question. Reducing uncertainty in the identified parameters would further clarify the role of anthropogenic aerosols in influencing large-scale climate effects.

Physical atmosphere model parameters are found to be far more important than aerosol parameters as sources of top-of-the-atmosphere radiative flux uncertainty. However, aerosols are the dominant source of uncertainty in how the radiative flux changes in response to aerosol emissions (the aerosol radiative forcing). Observations of present-day radiative fluxes provide only a weak constraint of aerosol radiative forcing.

These results provide insight into, and motivation for, model development that focusses on uncertainty reduction rather than quantification.