A new strategy to simulate daily spatial distributions of urban air pollutants based on land use regression in Wuhan, China

Abstract

This study develops daily-scale land-use regression (LUR) models to simulate high-resolution spatiotemporal maps of air pollutant concentrations in Wuhan from November 2015 to October 2019. Eight buffer scales ranging from 100 m to 5000 m are established around the monitoring sites and incorporate geographic prediction variables, including land use, elevation, nighttime light, road length, and population density. By analyzing the temporal variation in six air pollutants (PM_2.5, PM₁₀, NO₂, O₃, CO, and SO₂), this study further evaluates how predictor effects vary with spatial scale. Results show that the relationships between geographic prediction variables and pollutant concentration varies across different scales, with the most frequently selected scales in the models being 100 m and 500 m. Moreover, this study finds that built-up land within 3000m contributes 22.92 % to overall pollutant levels and 43.54 % to PM₁₀ concentration. Therefore, this study not only overcomes the limitations of traditional LUR models in capturing fine-scale temporal resolution but also provides a basis for health impact assessment of air pollution.