High-resolution analysis of seasonal spatial variation for urban air pollution: Implications for exposure assessment

Exposure to air pollutants is associated with significant health effects. Accurate exposure assessment remains a critical yet challenging aspect of environmental health research. Although air quality monitoring station (AQMS) data are commonly used as a surrogate for personal exposure, diverse spatial-temporal variations can lead to significant exposure misclassification. This study aimed to identify the seasonal spatial variation of five air pollutants (PM2.5, PM10, NO2, CO, and O3) at city and small spatial scales (1 km2) in Seoul, Korea, using data from AQMSs and in-situ monitoring sites (IMSs) for high-resolution measurements. Measurements were conducted across four seasons over one year, covering city-, district-, and small-scale spatial units, with a detailed focus on a 1 km2 area within one administrative district. The air pollutant concentrations were obtained from 25 AQMSs in each district. Fine-scale measurements were carried out at eight IMSs within a 1 km2 area surrounding a single AQMS in Guro-gu, Seoul. To enable direct comparison, measurements from AQMS and IMS were simultaneously collected following standard monitoring protocols. Moran's index was used as an indicator of spatial autocorrelation to identify the homogeneity and heterogeneity of air pollutants by spatial units. Concentrations of pollutants at IMSs were overall higher than those at nearby AQMS, except for O3 concentrations in spring and summer. Seasonal spatial autocorrelation patterns in city-scale areas did not reflect variations in small-scale areas. These findings highlight the limitations of relying solely on AQMS data for exposure assessment and underscore the value of integrating high-resolution data to reduce estimation errors. This study provides a framework for enhancing air quality management and exposure assessment strategies by accounting for spatial-temporal variations, especially in areas lacking dense monitoring networks.