Development of land-use regression models for metals associated with airborne particulate matter in a North American city

Abstract

h i g h l i g h t s PM 1.0 -components displayed significant intra-urban variability in Calgary, Alberta. Land-use regression models were developed for 30 elements in summer and winter. 12 elements had models with R 2 > 0.7 in both seasons; 24 had R 2 > 0.5 in both seasons. Industrial sources were major predictors, as well as traffic, land-use, and wind. Interspecies dependencies were similar for measured and modeled pollutant data. a b s t r a c tAirborne particulate matter has been associated with cardiovascular and respiratory morbidity and mortality, and there is evidence that metals may contribute to these adverse health effects. However, there are few tools for assessing exposure to airborne metals. Land-use regression modeling has been widely used to estimate exposure to gaseous pollutants. This study developed seasonal land-use regression (LUR) models to characterize the spatial distribution of trace metals and other elements associated with airborne particulate matter in Calgary, Alberta.Two-week integrated measurements of particulate matter with <1.0 mm in aerodynamic diameter (PM 1.0 ) were collected in the City of Calgary at 25 sites in August 2010 and 29 sites in January 2011. PM 1.0 filters were analyzed using inductively-coupled plasma mass spectrometry. Industrial sources were obtained through the National Pollutant Release Inventory and their locations verified using Google Maps. Traffic volume data were obtained from the City of Calgary and zoning data were obtained from Desktop Mapping Technologies Incorporated. Seasonal wind direction was incorporated using wind rose shapes produced by Wind Rose PRO3, and predictor variables were generated using ArcMap-10.1. Summer and winter LUR models for 30 PM 1.0 components were developed using SAS 9.2.We observed significant intra-urban gradients for metals associated with airborne particulate matter in Calgary, Alberta. LUR models explained a high proportion of the spatial variability in those PM 1.0 components. Summer models performed slightly better than winter models. However, 24 of the 30 PM 1.0 related elements had models that were either good (R 2 > 0.70) or acceptable (R 2 > 0.50) in both seasons. Industrial point-sources were the most influential predictor for the majority of PM 1.0 components. Industrial and commercial zoning were also significant predictors, while traffic indicators and population density had a modest but significant contribution for most elements. Variables incorporating wind direction were also significant predictors. These findings contrast with LUR models for PM and gaseous pollutants in which traffic indicators are typically the most important predictors of ambient concentrations.These results suggest that airborne PM components vary spatially with the distribution of local industrial sources and that LUR modeling can be used to predict local concentrations of these airborne BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/ 4.0/). Atmospheric Environment 106 (2015) 165e177 elements. These models will support ...

Exposure assessment models for elemental components of particulate matter in an urban environment: A comparison of regression and random forest approaches

“…The predictors made available for inclusion in the final models for each element were based on previously validated LUR models for elements in PM2.5 (de Hoogh et al, 2013; Zhang et al, 2015) and on a previously validated LUR model for TRAP built using the same ambient sampling data in Cincinnati, Ohio (Ryan et al, 2007). Where applicable, geographic predictors were extracted based on the area within circles centered on the sampling site locations with varying buffer radii.…”

“…C, 2004). Recently, successful LUR models have been developed for PM components in twenty areas in Europe as a part of the ESCAPE study (de Hoogh et al, 2013) and for an urban area in Canada (Zhang et al, 2015). These land use models have allowed for assessment of exposure to individual components of PM and the study of their association with health outcomes (Beelen, 2015; Eeftens et al, 2014; Hampel et al, 2015).…”

Land Use Regression of Particulate Matter in Calgary, Canada

“…Predictor variables were defined on circular buffers from each sampling point. In addition, windrose variables were defined on buffers modified according to the prevailing winds in each season (Zhang et al 2015).…”

“…Land use regression (LUR) yields air pollution estimates at fine spatial resolution based on the relationship between air pollution values and land use variables observed at sampled points (Henderson et al, 2007). Most LUR literature focuses on NO2, with a few studies modelling PM2.5, ultrafine particles, and PM components (e.g., Henderson et al, 2007, Zhang et al, 2015. This paper is the first study comparing models for three PM size fractions.…”

Accounting for spatial effects in land use regression for urban air pollution modeling

“…Wind rose buffer variables were also calculated to examine the influence of prevailing wind directions on potential sources (Zhang et al, 2015). Wind rose buffer shapes representing prevailing wind directions during each seasonal sampling period were generated using WindRose PRO3 (Enviroware, 2014).…”