Canadian air zones represent a complex mixture of urban and rural land-use impacted by diverse emissions sources. The Calgary Spatial and Temporal Exposure Modeling (CSTEM) Study was designed to provide spatial and temporal air quality information for Calgary and surrounding areas to support local air zone management strategies and air pollution health studies. CSTEM collected two-week integrated measurements of nitrogen dioxide (NO2), volatile organic compounds, particulate matter (PM10, PM2.5), black carbon (BC), and PM-components at 125 sites in summer (August 5-19, 2015) and winter (January 20-February 3, 2016). Seasonal trends were assessed by collecting two-week integrated data every two weeks at four temporal sites across the city from March 25, 2015-April 27, 2016. NO2 and VOCs were measured using Ogawa Passive Samplers and Organic Vapor Passive Samplers. Gravimetric PM2.5 and PM1.0 measurements were collected using Harvard Cascade Impactors with 37 mm Teflon filters. PM2.5 samples were analyzed for elemental composition using HF-nitric acid digested inductively coupled plasma mass spectrometry. BC was measured via optical scanning of gravimetric PM2.5 samples using a SootScan Model OT21 Transmissometer. Continuous BC was collected at 40 sites using microAethalometers. Analysis of collocated BC samples showed good agreement (R2>0.70) between the methods. Air pollution data were combined with land-use information to develop land-use regression (LUR) models. Stepwise selection and regression tree methods were used to identify best predictors. The Getis-Ord Gi statistic and global Moran's I were applied to assess local variation of pollutants. Land use regressionLUR ordinary least squares (OLS) regression and geographically weighted regression (GWR) techniques. Summer results follow. NO2 displayed greater local variation compared with PM2.5. Therefore, GWR and regional OLS models were developed for NO2. Global OLS models performed poorly, predicting only 56% of the variability in NO2. Regional OLS models performed slightly better, with R2 ranging from 0.56-0.60. GWR models performed best, explaining > 80% of the global variability in NO2, with local R2 ranging from 0.56-0.87 (Q25=0.72, Q75=0.84). ANOVA tests confirmed that GWR provided a statistically significant improvement over OLS. Industrial zoning, infrastructure and major roads were significant predictors of NO2. Industrial zoning, PM emitting facilities, and local roads were major predictors of PM2.5. CSTEM results provide insight into best approaches for characterizing air pollution in a large, diverse air zone. Future analyses will focus on seasonal and temporal modeling; modeling BC, VOCs, and metals; and integrating data from chemical transport models, satellite remote sensing, and continuous regulatory monitoring.
展开▼