Uncertainty within human system, Earth system and coupled models is often represented as the ranges resulting from model inter-comparison projects focused on a standardized set of scenarios (e.g. SSPs, RCPs). While these exercises are useful in many ways, they limit the uncertainty space explored and provide no probabilistic interpretation. There is also often a lack of consistency in assumptions between human and earth system components. In this work, we provide a consistent framework for probabilistic uncertainty quantification in coupled human-Earth system models, using the MIT Integrated Global System Model (IGSM) to demonstrate. The IGSM links the Economic Projection and Policy Analysis (EPPA) model, a multi-sector, multi-region, computable general equilibrium (CGE) model of the world economy to the MIT Earth System Model (MESM) of intermediate complexity. Our approach involves formal uncertainty quantification of key parameters in both the human and Earth system components of the coupled model, sampling from the probability distributions to explore the uncertainty space, and developing integrated, probabilistic socio-economic and climate projections. These projections provide insight into the probability of outcomes of interest, including emissions, concentrations, temperature, economy-wide welfare and energy.

This probabilistic uncertainty quantification approach provides several benefits. It can help inform decision-making and risk-based discussions about mitigation and adaptation. It can identify key components and assumptions in models, as well as uncertainties of greatest importance or least understanding. This in turn can guide research and model development efforts. This approach can also provide insight into how multiple uncertainties can compound to either amplify or dampen uncertainty in outcomes. It can also be used to explore the propagation of uncertainty through various connected sectors and systems, for example the contribution of human system uncertainty to uncertainty in environmental outcomes. Here we provide an overview of our approach, example results and a discussion of some insights gained.