A framework for understanding and narrowing modeling uncertainties for Stratospheric Aerosol Injection (SAI)

In this work, we propose to separate the concept of uncertainty for the purpose of SRM evaluation into the following two macro-categories: the first is storyline uncertainty, which includes the sub-categories of scenario uncertainty (related to emissions and societal response); target uncertainty (what is the specific target that SRM tries to achieve) and strategy uncertainty (how is the SRM deployed to achieve the target). The second is physical uncertainty, which can be investigated through the exploration of both inter-model SRM uncertainty (related to the Earth system processes specific to SRM) and inter-model climate uncertainty (related to underlying uncertainty in the representation of climatic processes). Finally, there is a component of internal variability that can be treated similarly to climate change.

We apply our framework to a large and diverse set of climate model simulations of stratospheric aerosol injections (SAI) and find four important results using our framework analyzing surface temperature and precipitation data: 1) in the same set of models, inter-model uncertainties due to climate change are larger than inter-model SRM uncertainties independently of spatial aggregation considered, (68% larger for surface temperatures, 23% larger for annual precipitation); 2) inter-model SRM uncertainties in SAI simulations are strongly driven by uncertainties in the representation of aerosols, including their impacts on composition, transport and secondary effects, leading us to conclude that resolving such uncertainties across models has the potential to reduce overall uncertainties by 33% and 39%, respectively, for surface temperatures and precipitation; 3) uncertainties related to the deployment of different kind of aerosols in a single model are small compared to inter-model SRM uncertainties, constituting only a 10% and 16% fraction of overall uncertainties, respectively, for surface temperatures and precipitation, with some important regional differences for the latter; 4) by looking at results of different strategies and scenarios with one single model, we conclude that, while at the global level temperature and precipitation are overwhelmingly driven by target uncertainty, at the regional level the strategy uncertainty can be a relevant portion of the overall uncertainty.