Yet, how important the model choice is in the assessment of pollution and distributional outcomes remains unclear. This study compares three air quality models with varying complexity and examines the effects of model choice on key outcomes of policy interest. With a focus on the US, we consider one baseline scenario for 2017 and one future climate policy scenario that assumes the country reaches net-zero emissions by 2050. We use three models to simulate ambient particulate matter (PM2.5) and then assess county-level PM2.5-attributable deaths: (i) a fully-coupled atmospheric chemistry and transport model (WRF-Chem), (ii) a reduced-form air quality model (InMAP), and (iii) a source-receptor matrix (ISRM).
We find that all three models generate nationwide net health co-benefits, but they vary substantially in the distributional effects. Nationally, the three models estimate annual average pollution-weighted PM2.5 exposure reductions of 13% (WRF-Chem), 21% (ISRM), 29% (InMAP) in the net-zero 2050 scenario compared to 2017, lowering the age-specific mortality risk from PM2.5 exposure. However, different models find different geographic distributions of the impacts: WRF-Chem finds the greatest pollution reduction in the Eastern region, whereas InMAP and ISRM models find larger reductions in the Central region. These differences are driven by model-varying representations of the chemical and physical processes. They are also affected by different assumptions for the meteorological inputs: WRF-Chem uses 2017 meteorological conditions, whereas InMAP and ISRM were developed using 2005 meteorology. Such discrepancies in pollution distributions lead to substantially different equity conclusions related to the racial and income disparities of pollution exposure and mortality risks. A careful consideration of the model choice is thus essentially for assessing the policy effectiveness to advance equity goals.

