A Bayesian Inverse Modeling Approach to Achieving Triple Wins in Air Quality, Climate, and Equity

With increasing urgency to mitigate air pollution, climate change, and racialized exposure disparities, decision-makers in the United States (US) are faced with three distinct challenges that arise from the same sources but are often managed separately. This is in part because traditional environmental policies are generally designed based on a forward simulation approach: formulating an idea, estimating emission-changes, and modeling the resulting changes to air pollution, climate mitigation, and environmental justice. This process is computationally inefficient for testing multiple strategies and poorly suited for optimizing outcomes that address multiple objectives. Here, we reverse this pipeline to derive emission-reduction pathways that represent the optimal “triple win” strategy for mitigating air pollution exposure, climate change, and exposure inequity across the contiguous US.

To do this, we build upon our novel receptor-oriented, Bayesian optimization method by incorporating an additional cost function that reweights reductions for other priorities. Our approach begins from an atmospheric inverse modeling framework, whereby we set an idealized concentration surface — meeting the US National Ambient Air Quality Standard for particulate matter (PM_2.5) everywhere — as the target variable. Using an alternating gradient descent algorithm, we perturb this optimal solution to find the co- or triple-benefits associated with advancing climate or equity goals. We consider four optimal emission reduction scenarios representing distinct combinations of policy goals:  PM_2.5 Exposure Alone, Climate Priority, Equity Priority, and Triple Win. Our solutions are discretized in space, by precursor pollutant, and by the economic sector of emission.

Although all scenarios meet the PM_2.5 standard, preliminary results suggest that meeting different combinations of goals requires attention to diverse locations, chemical species, and sectors. While the difference in total aggregate emissions reduction is small when comparing the PM_2.5 Exposure Alone case with the other priorities, incorporating additional priorities up front enables the direct identification of distinct mitigation pathways in space and by sector (e.g., marine vessels are important for climate mitigation). We demonstrate how a non-optimal emission reduction pathway results in lesser or neutral air quality and climate benefits; however, the non-optimal reduction pathway can also result in significant harms in terms of environmental injustices. 

This framework could have strong implications for how we think about the challenge of how environmental policy can advance action against compounding risks. Our approach provides a data-driven and scalable strategy for simultaneously achieving a triple win across exposure, climate, and equity goals.