Spatio-temporal modeling of PM_2.5 exposure in California's San Joaquin Valley using a community sensor network

Persistent disparities in air pollution exposure remain a concern in California's San Joaquin Valley (SJV), yet the spatio-temporal dynamics of these inequities at the community scale remain poorly characterized. In this study, we integrated PM_2.5 measurements from 802 low-cost sensors collected during 2021-2023 into a Bayesian hierarchical spatio-temporal model to quantify how PM_2.5 exposure varies with socioeconomic and demographic composition across census tracts in the SJV. The model identified higher PM_2.5 exposures in census tracts with larger proportions of people of color, particularly Hispanic residents. Socioeconomic disadvantage, reflected by higher financial burden and unemployment, was also associated with elevated PM_2.5 levels. These disparities extended beyond current California Assembly Bill 617 (AB 617) boundaries that are used to guide the allocation of targeted monitoring and emission-reduction programs. By incorporating spatial and temporal random effects, the model also captured substantial residual variability not explained by observed covariates. Spatial random effects identified persistent hotspots along traffic corridors and urban centers, in contrast to cold spots in mountainous areas. This spatial heterogeneity further reveals that disparities among Hispanic-majority communities were not uniform, with urban tracts experiencing compounded burdens from local emission sources. Together, these results show that integrating community-level data with hierarchical modeling enables the identification of fine-scale environmental inequities and supports more equitable air-quality management.