A spatial land use clustering framework for investigating the role of land use in mediating the effect of meteorology on urban air quality. (December 2021)

• Record Type: Journal Article
• Authors: Montazeri, Amir; Lilienthal, Achim J.; Albertson, John D.
• Is Part Of: Atmospheric environment. Volume 12(2021)
• Publication Date: 2021-12
• Subjects: Air pollution profiles -- Cluster analysis -- Mobile monitoring -- Land use effects -- K-means -- Exceedance probabilities -- Unsupervised learning -- Machine learning
• DOI: 10.1016/j.aeaoa.2021.100126
• Languages: English
• ISSNs: 2590-1621

Abstract

Abstract: Air pollution in urban areas is driven by emission sources and modulated by local meteorology, including the effects of urban form on wind speed and ventilation, and thus varies markedly in space and time. Recently, mobile measurement campaigns have been conducted in urban areas to measure the spatial distribution of air pollutant concentrations. While the main focus of these studies has been revealing spatial patterns in mean (or median) concentrations, they have mostly ignored the temporal aspects of air pollution. However, assessing the temporal variability of air pollution is essential in understanding the integrated exposure of individuals to pollutants above critical thresholds. Here, we examine the role of urban land use in mediating the effect of regional meteorology on Nitrogen Dioxide (NO2 ) concentrations measured in different regions of Oakland, CA. Inspired by Land Use Regression (LUR) models, we cluster 30-m road segments in the urban area based on their land use. The concentration data from the resulting clusters are stratified based on seasonality and conditionally averaged based on concurrent wind speeds. The clustering analysis yielded 7 clusters, with 4 of them chosen for further statistical analysis due to their large sample sizes. Two of the four clusters demonstrated in winter a strong negative linear relationship between NO2 concentration and wind speed ( R 2 > 0.87) with a slope of approximately 3 ppb/m s -1 . A weaker correlation and flatter slope was found for the cluster representing road segments belonging to interstate highways ( R 2 > 0.73 and slope < 2 ppb/m s -1 ). No significant relationship was found during the summer season. These findings are consistent with the concept of strong vertical mixing due to highway traffic and increased surface heat fluxes during summer weakening the relationship between wind speed and NO2 concentrations. In summary, the clustering analysis framework presented here provides a novel tool for use with large-scale mobile measurements to reveal the effect of urban land form on the temporal dynamics of pollutant concentrations and ultimately human exposure. Highlights: Clustering framework developed for spatio-temporal analysis of mobile measurements. Domain-related procedure for selecting number of clusters in k-means is presented. Effect of meteorology on pollutant levels as mediated by land-use is investigated. Wind speed is only effective in reducing pollutant levels in some urban regions.