A Cross-City Analysis of Mobility Responses to Air Pollution in China

Understanding how air quality influences mobility patterns is crucial for effective transport and environmental management. While existing studies have explored the mobility impacts of air pollution, their findings remain inconsistent, partly due to the dualistic mobility responses (increased population redistribution within the city and reduced mobility for pollution avoidance) during high-pollution episodes. Here, we examine the causal effect of PM_2.5 levels on within-city mobility and explore driving factors of the spatial heterogeneity underlying this effect across 318 Chinese cities from May to November 2023, employing econometric models and explainable machine learning techniques. We find that, on average, a one-standard-deviation (16.99 μg/m3) increase in PM_2.5 level reduces daily trips by approximately 0.28-0.41%, reflecting widespread travel avoidance on heavily polluted days. However, mobility responses exhibit significant spatial heterogeneity. Among sample cities, a negative association between air pollution and mobility has been observed in 68% of them, with the rest suggesting the other way around. The explainable machine learning model results further reveal that this disparity is mainly driven by transboundary air pollution and long-term mobility levels, sociodemographic factors (e.g., age, population, and tertiary industry GDP share), as well as land use features (e.g., population density, geographical area of the city). Our findings offer a scientific basis for continued air pollution mitigation efforts and support the design of targeted land use policies tailored to local conditions.