Integrating meteorological factors for better understanding of the urban form-air quality relationship

Context Understanding the urban form-air quality relationship is essential to mitigate intra-urban air pollution and to improve urban ecology. However, few studies considered urban form and meteorological factors integratively and analyzed their synthetic effects on air pollution. Objectives We investigate how to model the integrated effects on the spatiotemporal distribution of PM(2.5)in the Atlanta metropolitan area to improve the understanding of the urban form-air quality relationship. Methods Two groups of models are developed: one uses urban form only and the other uses wind-direct urban form. Relative humidity, wind speed, and temperature are included as control variables. Both linear (Multiple Linear Regression) and nonlinear models (Random Forest and Artificial Neural Network) are constructed and tested with both tenfold cross-validation and field PM(2.5)data obtained from a portable device, AirBeam2. Results Random Forest overall outperforms other models suggesting that the urban form-air quality relationship is most likely to be nonlinear. Additionally, the group using wind-direct urban form outperforms the other group and the contribution of the same urban form metrics differs in different wind sections proving that meteorological factors and urban form have synthetic effects on PM2.5. Finally, the patch density, dominance, and aggregation of roads and vegetation, demonstrate higher attribute significance than other urban form metrics. Conclusions Urban planners, practitioners, and policymakers need to carefully consider not only the spatial configuration of roads and vegetation but also the local climate patterns to minimize intra-urban air pollution effectively.