Living near traffic adversely affects health outcomes. Traffic exposure metrics include distance to high-traffic roads, traffic volume on nearby roads, traffic within buffer distances, measured pollutant concentrations, land-use regression estimates of pollution concentrations, and others. We used Geographic Information System software to explore a new approach using traffic count data and a kernel density calculation to generate a traffic density surface with a resolution of 50 m. The density value in each cell reflects all the traffic on all the roads within the distance specified in the kernel density algorithm. The effect of a given roadway on the raster cell value depends on the amount of traffic on the road segment, its distance from the raster cell, and the form of the algorithm. We used a Gaussian algorithm in which traffic influence became insignificant beyond 300 m. This metric integrates the deleterious effects of traffic rather than focusing on one pollutant. The density surface can be used to impute exposure at any point, and it can be used to quantify integrated exposure along a global positioning system route. The traffic density calculation compares favorably with other metrics for assessing traffic exposure and can be used in a variety of applications.
On-road vehicle traffic is a source of air emissions of carbon monoxide, carbon dioxide, oxides of nitrogen, organic gases, metals, crustal materials, diesel particles, and other particles of diverse chemistry ranging in size from nanometer diameter to the coarse fraction. In addition, traffic is a source of noise, heat, and water vapor, and may add stress due to the busyness of the traffic flow. Exposure to traffic has been associated with numerous adverse health outcomes,1 including asthma,2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 low birth weight,13, 14 respiratory disease,4, 8, 10, 15, 16 cardiovascular disease,10, 16, 17, 18, 19 and development deficits,20 among others. This mounting evidence of the unhealthiness of exposure to traffic begs at least two questions: what feature of traffic is responsible for the ill effects, and how can exposure to traffic be quantified?
Many of the individual pollutants emitted from traffic have been shown to cause health effects either alone or in combination with other pollutants.1 However, the complex mixture of traffic-derived pollutants found in ambient air is heterogeneous in space and time and depends upon fleet composition, fuel types, traffic flow, topography, meteorology, and the spatial roughness features of the natural and built environments. Simulating the details of this complexity is beyond the capability of current databases and computer resources, but it is the sum total of this complex and constantly changing mixture, together with the pre-existing vulnerability of the population, that determine the potential for traffic to affect health.
Numerous studies have quantified ambient levels of one or more air pollutants emitted from traffic. Such quantification is possible with monitoring, air dispersion modeling, and combinations of the two. Monitoring studies give the best measures of concentrations of particular pollutants at the monitoring site, but they suffer from several shortcomings. The number of species and monitoring locations are limited, the cost is generally high, and the attribution of the sources of the measured concentrations is uncertain. In contrast, modeling simulations can be performed for many species with relatively small cost at numerous receptor locations, and the sources of the modeled concentrations can be readily determined. However, modeling is subject to errors in the theoretical conceptualization of relevant processes, the algorithms developed to parameterize processes, and the accuracy of the input data. The greatest source of error in air pollution dispersion models is often identified as the input data on emission rates.21, 22, 23, 24, 25, 26
Despite the aforementioned limitations, measurements and/or modeled estimates of air concentrations are used to infer exposure to individual pollutants and groups of pollutants coming from traffic sources.23, 24, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38 These methods are useful but cumbersome because detailed emission inventories and air dispersion modeling are required. Estimating emissions from on-road mobile sources is highly uncertain,36 and modeling spatially resolved emissions requires considerable computer resources.
Another approach that has been used in recent years to spatially resolve air pollution from traffic and other sources is land-use regression (LUR,39, 40, 41, 42, 43, 44, 45, 46, 47). In that approach, measurements of air pollutant concentrations are related to geographic land-use variables. If good relationships can be established, then the land-use variables can be used to predict air concentrations in areas where measurements have not been made. Novotny et al.46 were able to develop estimates of nitrogen dioxide concentrations across the United States and highly spatially resolved estimates for Los Angeles and Minneapolis/St. Paul using land-use regression methodologies.
An alternative to looking at individual pollutants via monitoring or modeling is to look at exposure to traffic itself as a predictor of adverse health outcomes. Using traffic exposure as the metric for evaluating potential health impacts has the advantage of integrating all of the negative effects into one measure rather than looking pollutant by pollutant. One method often used in epidemiological studies is to quantify the traffic occurring around a point of interest, such as the home of a person who has experienced a health event potentially caused by traffic.48, 49 Some of the quantification metrics that have been used include the distance to a major roadway, the amount of roadway within a specified distance, and the sum of vehicle kilometers travelled (VKT) within a buffer around the point of interest.
We used a different approach in which we generated a spatially resolved surface of traffic density from detailed traffic count data. The surface stands alone as a representation of traffic density, as the calculations were independent of any preconceived points of interest such as residences where health outcomes were measured. There are a few reports of work using this general approach.50, 51
The metric that we chose is a kernel density calculation that uses traffic count data to make inferences about the probability density function across an area of interest, in our case the state of Minnesota. Once the density surface is established, the traffic density value at any point can be imputed or other features can be superimposed on the surface. For example, the daily travels of an individual can be recorded using global position system (GPS) hardware, and the traffic exposure over time can be extracted from the surface and used as a metric against which to evaluate health outcomes or other end points of interest. This type of tool has implications for exposure science as suggested in a recent report by the National Research Council52 on exposure science in the 21st Century.
MATERIALS AND METHODS
Four categories of data were used in this study: traffic count data, locational coordinates for homes in a health database, land-use regression estimates of nitrogen dioxide concentrations, and GPS tracks recorded with personal telecommunications devices.
Three files of traffic count data were obtained from the Minnesota Department of Transportation (MNDOT) in the Geographic Information System (GIS) shapefile format:
AADT_1992_2010=Annual Average Daily Traffic=all traffic on all counted roadway segments for years 1992 to 2010;
HCAADT_1994_2010=Heavy Commercial AADT=heavy commercial traffic on all roadway segments where heavy commercial traffic was counted (a subset of the counted roadways from AADT) for years 1994 to 2010; and
RouteSystem_8_23_2010_AADT=estimated traffic counts on non-counted roadway segments.
The traffic count data were used to calculate traffic density across the state of Minnesota. The Hawth’s Analysis Tools for ArcGIS kernel density function in the GIS program ArcMap (version 9.3, ESRI, Redlands, CA) was used to make this calculation. The first step was to merge the traffic count data into a single file of the combined values. The HCAADT counts were subtracted from the AADT counts to give non-diesel counts. Next the HCAADT counts were multiplied by 3.6 to represent the greater emissions mass and toxicity of diesel vehicles53, 54, 55 and added together with the two non-diesel count files to generate toxicity-weighted counts. Straightforward diesel and non-diesel data were also processed and maintained separately.
The next step was to break the road segment traffic count shapefile data into points to use in kernel density estimation. Selecting all roads, the ‘Explode Multipart Features’ tool from the ArcMap advanced editing toolbar was run to separate non-contiguous road arcs so that the following point generation step did not create false pathways by misconnecting arc end points that should not be connected. The ‘Convert Paths to Points’ tool was used to develop a point layer with a resolution of 100 m. Using the point layer, the Hawth's ‘Fixed Kernel Density Estimator’ tool was used to convert the point layer to a raster grid with a scaling factor of 1,000,000, the kernel set to bivariate normal, and the parameter smoothing factor set to 300 m. These parameters resulted in a particular point on a roadway having a zone of influence that was Gaussian with a significant impact extending 300 m from the road. The resulting raster has a 50-m grid cell size, with each grid value depending on all of the nearby traffic and the effect of each roadway segment decreasing with distance from the cell. The calculated traffic density values ranged from 1 to 144, with units of AADT counts/area.
In order to correlate the traffic density metric with nitrogen dioxide concentrations, land-use regression estimates of nitrogen dioxide (LUR-NO2) concentrations from work done by Novotny et al.46 were obtained in GIS format. The traffic density and LUR-NO2 rasters were projected into equivalent coordinate systems with matching grid cell sizes. The ArcMap band collection tool was used to calculate the spatial correlation between traffic density and LUR-NO2.
The density metric was also compared with a more traditional traffic representation, Vehicle Kilometers Travelled (VKT) within buffers around points of interest. We used the locational coordinates of a subset of homes in the Rochester Epidemiology Project (REP) as the points of interest. The data for addresses were obtained from REP for individuals with the conditions of interest. The REP is a medical records linkage database of all residents of Olmsted County, Minnesota. Data from all health-care visits to any health-care facility (office, hospital, outpatient clinic, urgent care, or emergency department) are collected within the REP and linked across facilities for each individual Olmsted County resident.56, 57
Total VKT within buffers around the REP homes was calculated using the ArcMap Intersect utility. The length of each road segment within 250- and 500-m buffers was calculated for each of the MNDOT files. The VKT was calculated by multiplying the traffic count on the segment by the length of the segment within each buffer. Total VKT was calculated as the sum of AADT VMT plus RouteSystem_8_23_2010 VMT in a buffer. Regressions were calculated between the VKT in the buffers and the traffic density value extracted at the point of interest using the SPSS (IBM SPSS Statistics).
GPS tracks were collected using smart phone hardware and GPS applications including MotionX-GPS, iMapMyRIDE, and Cyclemeter. These applications are designed for specific end users, but all produce GPS tracks that are independent of mode of transportation. A typical GPS track consists of sets of coordinates recorded at equal intervals of time. Using ArcMap, traffic density values were assigned to each coordinate set in a GPS track to generate time-integrated estimates of exposure to traffic over the course of the track.
Figure 1 shows the traffic density surface in the Minneapolis–St. Paul metropolitan area at three scales: the entire metropolitan area (Figure 1a), the central corridor light rail study area in the midway district between Minneapolis and St. Paul (Figure 1b) and a close-up of the Phillips neighborhood just south of downtown Minneapolis (Figure 1c). The major roadways are clearly visible as darker areas of high traffic density. In the overview (Figure 1a), the metropolitan core area appears as a generally high traffic area with resolution around major roadways. The close-up (Figure 1b) allows the finer detail in the raster to be seen in the urban core. This area has been studied by the Minnesota Department of Health58 for potential changes in environmental health stressors because a new light rail transit corridor is being developed in the area.
Figure 1c is a finer detail close-up of the Phillips neighborhood in south Minneapolis overlain on an aerial photograph. Phillips is a diverse inner-city neighborhood facing multiple socio-economic and environmental challenges that has been the subject of several environmental justice-related studies. The northern boundary of the neighborhood is the merged interstate highway commons of I-35W and I-94, which bears the highest traffic load in Minnesota. Within this highly impacted neighborhood, there is a strong gradient in traffic exposure, as shown by the chart inset, in which density values were extracted from the raster at regular intervals on a transect across I-35 W. The locations of the transect points are shown by the hashed line on the left.
Figure 2a shows a close-up view of a section of Olmsted County, MN, with shading according to the traffic density raster. In addition, the 250- and 500-m buffers around the REP data points in the area are overlaid on the density surface. Figure 2b is a graph of the relationship between the raster values and the 250-m buffer VKT values for the REP points. The raster and the 250-m buffer VKT values were well-correlated (r2=0.74 for the log-log regression). The relationship between the raster values and the 500-m buffer VKT values was not as good (r2=0.59 for the log-log regression). Regressions were done in the log scale due to the skewed and approximately log-normal distribution of the VKT and traffic density variables.
The correlation coefficient between traffic density and LUR-NO2 calculated using the ArcMap band collection tool was 0.58 when projected to a contiguous Albers equal area conic coordinate system with a 50-m grid cell size corresponding to the grid cell size of the traffic density raster. The grid cell size of the LUR-NO2 data was originally 100 m. To establish a quantitative relationship between LUR-NO2 and traffic density, the rasters were extracted to points, and a regression was run using a log-transformation due to the skewness of the data. The equation relating the two variables was:
We expect that a stronger relationship could be established if the two data sets were calculated with the same spatial resolution from the original data rather than re-processing afterward as we did.
Figure 3 is a street map of the central Minneapolis-St. Paul metropolitan area overlain with the traffic density surface. The topmost layer is a GPS track recorded with a cellular telephone of a 54 min bicycle ride beginning and ending on the easternmost section of the track. The points recorded by the telephone are color coded according to the traffic density value at that raster point. The track followed relatively lightly travelled roadways, the east and west river roads, for much of the time, but included sections crossing the interstate highway where the highest traffic density values were recorded. The inset graph is a time series of the raster values with each bar representing one point on the GPS track. The height of the bar, as well as its color, indicates the traffic density raster value.
Figure 4 shows the traffic density surface for an area of central St. Paul. GPS tracks for commutes by two modes, bicycle and car, are superimposed on the raster surface. The bicycle commute took nearly twice as long as the car commute (25:28 versus 13:48 min). The bicycle commute was on mostly lower-traffic roadways with an average traffic density raster value of 13.95 versus 16.59 for the car commute; however, because of the longer time on the roadways, the sum of the time-integrated traffic density values by bicycle was 21 334 versus 12 756 by car.
This study addresses the quantification of exposure to traffic. We used traffic count data to calculate a traffic density surface. This surface can be used to impute traffic exposures that can be used as predictive variables in epidemiological studies, and it can also be compared with other geographic variables such as socio-economic status and pollutant concentrations. We suggest that traffic density itself is a useful metric that combines the effects of the multiple stressors, such as gases, particles, and noise, coming from traffic. This approach has advantages and disadvantages that should be taken into consideration in the application and evaluation of the method. One salient disadvantage is that traffic density does not inform us as to which component of traffic is responsible for health impacts. However, we would argue that many of the traffic-related stressors are highly correlated with one another, and that confounding can also affect studies that look at individual pollutants in isolation. A second major limitation is that the pollutants and other stressors from traffic are emitted or occur in the outdoor environment, whereas people spend most of the time indoors and thus exposure can be mischaracterized. However, the tracking methods we describe could be used to characterize time spent in indoor locations. If the indoor environment can be characterized then exposure can be assessed for that microenvironment as well.
It is likely that the effects of different stressors caused by traffic may manifest themselves over different spatial scales. The grid cell size should have a resolution that captures the spatial scale of the feature under consideration. We chose the 50 m size based on previous work suggesting that concentrations of a number of air pollutants vary from peak to background values over distances of a few hundred meters from highways. However, many studies (see the review inHealth Effects Institute.1) show that gas to particle transformations, as well as changes in particle size from nano- and ultrafine to accumulation mode, occur over shorter distances. Therefore, a 50-m grid might not capture exposure to very small particles. Similarly, the spatical scale for noise effects may be different than for air pollution. However, if one assumes that the grid cell size captures the spatial scale of an effect, then imputation of a value from the surface can be useful. We also note that there is a balance between the scale of resolution and the amount of computer resources required for the calculation of a density surface. The 50-m resolution over a regional scale requires significant computation time.
The traffic density surface (Figure 1) clearly shows high traffic corridors; however, the detailed spatial scale shows that there is considerable heterogeneity in traffic density, even in the center of the urban core. This result is not surprising, as we started with traffic counts on roadway segments. The urban core in our study (and most modern cities) has a dense network of residential streets with interspersed arterial roads and major highways in high traffic corridors. The intersecting scales of these different types of roadways determine the pattern of traffic density, which comprise an underlying background with corridors of greater density on the more highly trafficked roads. Many studies1, 39, 59, 60 have shown that air pollution levels disperse from peak levels near the centerline of busy roadways to background values over a distance of about 300 m. For that reason, we chose an algorithm that allowed the density impact to decay to insignificant levels over a distance of 300 m. This pattern is clearly evident in the inset to Figure 1c. Traffic density values across the transect over the interstate highway follow a Gaussian pattern with respect to the interstate highway because the traffic counts in that corridor are much higher than on nearby residential and arterial streets.
A traffic density surface could be created using other metrics and other algorithms. For example, a grid could be created using a metric such as a 250-m buffer VKT; however, that metric would result in a non-smooth surface with disjointed, step-like features at specific distances from major roadways unless the resulting surface were later smoothed. The Gaussian formulation in the kernel density algorithm, on the other hand, was used with parameter values that result in a gradual decay of effect with distance from a roadway. We believe this approach gives a result that is more in keeping with the way pollutants disperse (and potential detrimental effects propagate) with distance from a roadway. If evidence warranted, the form and/or parameter values of the density algorithm could be changed to modify the distance and influence pattern in space over which the roadway traffic count values are considered.
The strong relationship (Figure 2) between VKT in 250-m buffers around points of interest and the traffic density at those points is not surprising. Either method could be used to characterize traffic exposure. The advantages of our traffic density approach are that the effect of traffic is a smoothed rather than a step function of distance, and the traffic density surface can be re-used in other analyses for different sets of points (or GPS tracks).
We took the relative influence of heavy duty diesel vehicles as 3.6 times the impact of light duty gasoline vehicles. This is also based on the weight of evidence from the HEI review,1 and takes into account increased emissions and toxicity from heavy duty diesel vehicles. However, other values or approaches could be justified, especially considering that the toxicity of diesel emissions is decreasing with the use of ultra-low sulfur diesel fuel and new generation of diesel engines with particle and nitrogen oxide emission controls.53, 61
Another component of exposure mischaracterization is temporal mismatch. We used annual average daily traffic volume data as one of the inputs to the traffic density algorithm. These data do not capture diurnal, weekly, seasonal, or other temporal fluctuations in traffic or in meteorological conditions that may be important factors affecting exposure. The surface does not correspond to exposure at a specific point in time nor to a specific bicycle ride on a given day. Rather it can be thought of as the average exposure over a large number of repetitions of a bicycle ride taken at various times. Despite this caveat, the relative differences in space may persist during different times in the temporal cycles, and thus comparisons of locations could be considered valid as long as the time frames correspond. Some air quality modeling applications employ temporal (day of week, hour of day) scaling factors to capture temporal variability, and it is possible to use such factors to generate temporally varying traffic density rasters. In addition, many metropolitan areas have traffic management organizations that maintain sophisticated models to predict and manage daily rush hour traffic demand. Data from such models could be used to develop time-resolved traffic density calculations.
The traffic density surface assumes that all of the density occurs at ground level, and it does not account for bridges, overpasses, and other elevational variations that can affect exposure. Overlaying a GPS track therefore does not capture the possible variations in traffic exposure that could occur with height. These sources of error and uncertainty can be reduced with better data on roadway elevations. Similarly, the analysis shown in Figure 4 comparing commute modes does not account for factors such as whether windows or vents are open for the car commute. With bicycle commuting the precise location of the cyclist with respect to nearby traffic can significantly affect exposure, and this effect is not fully captured in the traffic density metric.
The fact that the LUR-NO2 results correlate well with traffic density is not surprising, as the LUR used data on the locations of roadways as an input and traffic is a major source of NO2 emissions. If LUR is done on a resolved spatial scale, then the same exposure assessment methods we used here for traffic exposure could also be used to estimate exposure from the LUR pollutant surface. It is also possible to use traffic density as a predictor variable in the LUR analysis in place of other variables like distances to major and minor roadways.
One of the major sources of mischaracterization of air pollution exposure lies in the fact that people do not stay in a specific location where air pollution measurements or model results are obtained. Our approach allows us to track movements over time and integrate the exposure over the course of an individual’s travel. We believe this methodology can be useful for better characterizing exposure in future studies relating health effects to air pollution. Modern telecommunications technology allows travel recording to be done easily and inexpensively. Studies using such technology are appearing in the literature,62, 63, 64 and public and private entities are using telecommunications to track activities and travel patterns. For example, the Metropolitan Council, the regional governing body for the Minneapolis-St. Paul metropolitan area, developed a smart phone application to track bicycle usage in the area.
The ability to track travel and compare routes as is shown in Figures 3 and 4 raises questions about assessing time-activity data for apportioning exposure to micro-environments. Much past work assigns exposure to micro-environment categories in which the pollutant concentration is assumed to be constant or at a fixed ratio to central monitoring site data. Our analysis shows that these assumptions are likely to be overly simplified, and exposure in a micro-environment such as ‘in transit’ constantly varies depending upon the route travelled, the mode of travel, the ventilation characteristics of the vehicle, and other factors. In the basic comparison shown in Figure 4, the bicycle track has lower average exposure to traffic but higher total exposure because the bicycle route takes longer. If the extra 11 min and 40 s saved by the driver were spent at the western terminus of the route (i.e., at home), the total traffic exposure would be lower for the car driver. If that time were spent at the eastern terminus (i.e. at work), the total exposures for car and bicycle would be approximately the same. If, however, the car driver were to be stuck in highway traffic for those additional minutes, then the exposure of the car driver would be much higher than the bicyclist.
Starting with traffic count data, we used a kernel density algorithm in a geographic information system to calculate a raster of traffic density over the state of Minnesota. The density value in each cell reflects all the traffic on all the roads within the 300 m distance specified in the kernel density algorithm. The effect of a given road on the raster cell value depends on the amount of traffic on the road segment, its distance from the raster cell, and the form of the algorithm. This traffic density calculation compares favorably with other metrics for assessing traffic exposure. The method integrates the deleterious effects of traffic rather than focusing on one specific pollutant at a time. The traffic exposure at any point on the surface (in our case the state of Minnesota) can be compared with that at any other point, a comparison that can be used in evaluating the effects of traffic on health. The density surface can also be used to quantify the integrated exposure along a travelled route recorded with a global positioning device, something that enhances the capability to characterize exposure. The general methodology lends itself to a variety of other potential applications in exposure science and other fields.
Health Effects Institute. Traffic-related air pollution: A critical review of the literature on emissions, exposure, and health effects 2010 Retrieved 29 July 2012, from http://pubs.healtheffects.org/view.php?id=334.
Chen E, Schreier HMC, Strunk RC, Brauer M . Chronic traffic-related air pollution and stress interact to predict biologic and clinical outcomes in asthma. Environ Health Perspect 2008; 116: 970–975.
Gent JF, Koutrakis P, Belanger K, Triche E, Holford TR, Bracken MB et al. Symptoms and medication use in children with asthma and traffic-related sources of fine particle pollution. Environ Health Perspect 2009; 117: 1168–1174.
Holguin F, Flores S, Ross Z, Cortez M, Molina M, Molina L et al. Traffic-related exposures, airway function, inflammation, and respiratory symptoms in children. Am J Respir Crit Care Med 2007; 176: 1236–1242.
Juhn YJ, Qin R, Urm S, Katusic S, Vargas-Chanes D . The influence of neighborhood environment on the incidence of childhood asthma: a propensity score approach. J Allergy Clinl Immunol 2010; 4: e2.
Juhn YJ, Sauver JS, Katusic S, Vargas D, Weaver A, Yunginger J . The influence of neighborhood environment on the incidence of childhood asthma: a multilevel approach. Soc Sci Med 2005; 60: 2453–2464.
McConnell R, Islam T, Shankardass K, Jerrett M, Lurmann F, Gilliland F et al. Childhood incident asthma and traffic-related air pollution at home and school. Environ Health Perspect 2010; 118: 1021–1026.
McCreanor J, Cullinan P, Nieuwenhuijsen MJ, Stewart-Evans J, Malliarou E, Jarup L et al. Respiratory effects of exposure to diesel traffic in persons with asthma. N Engl J Med 2007; 357: 2348–2358.
Meng Y-Y, Wilhelm M, Rull RP, English P, Nathan S, Ritz B . Are frequent asthma symptoms among low-income individuals related to heavy traffic near homes, vulnerabilities, or both? Ann Epidemiol 2008; 18: 343–350.
Peretz A, Sullivan JH, Leotta DF, Trenga CA, Sands FN, Allen J et al. Diesel exhaust inhalation elicits acute vasoconstriction in vivo. Environ Health Perspect 2008; 116: 937–942.
Shankardass K, McConnell R, Jerrett M, Milam J, Richardson J, Berhane K . Parental stress increases the effect of traffic-related air pollution on childhood asthma incidence. Proc Natl Acad Sci USA 2009; 106: 12406–12411.
Wilhelm M, Meng Y-Y, Rull RP, English P, Balmes J, Ritz B . Environmental public health tracking of childhood asthma using California health interview survey, traffic, and outdoor air pollution data. Environl Health Perspect 2008; 116: 1254–1260.
Brauer M, Lencar C, Tamburic L, Koehoorn M, Demers P, Karr C . A cohort study of traffic-related air pollution impacts on birth outcomes. Environ Health Perspect 2008; 116: 680–686.
Pereira G, Nassar N, Cook A, Bower C . Traffic emissions are associated with reduced fetal growth in areas of Perth, Western Australia: an application of the AusRoads dispersion model. Australian and New Zealand J Public Health 2011; 35: 451–458.
Kim JJ, Huen K, Adams S, Smorodinsky S, Hoats A, Malig B et al. Residential traffic and children’s respiratory health. Environ Health Perspect 2008; 116: 1274–1279.
Künzli N, Kaiser R, Medina S, Studnicka M, Chanel O, Filliger P et al. Public-health impact of outdoor and traffic-related air pollution: a European assessment. Lancet 2000; 356: 795–801.
McDonald JD, Campen MJ, Harrod KS, Seagrave J, Seilkop SK, Mauderly JL . Engine-operating load influences diesel exhaust composition and cardiopulmonary and immune responses. Environ Health Perspect 2011; 119: 1136–1141.
Tonne C, Melly S, Mittleman M, Coull B, Goldberg R, Schwartz J . A case–control analysis of exposure to traffic and acute myocardial infarction. Environ Health Perspect 2006; 115: 53–57.
Weichenthal S, Kulka R, Dubeau A, Martin C, Wang D, Dales R . Traffic-related air pollution and acute changes in heart rate variability and respiratory function in urban cyclists. Environ Health Perspect 2011; 119: 1373–1378.
Gauderman WJ, Vora H, McConnell R, Berhane K, Gilliland F, Thomas D et al. Effect of exposure to traffic on lung development from 10 to 18 years of age: a cohort study. Lancet 2007; 369: 571–577.
Armitage JM, Cousins IT, Hauck M, Harbers JV, Huijbregts MAJ . Empirical evaluation of spatial and non-spatial European-scale multimedia fate models: results and implications for chemical risk assessment. J Environ Monit 2007; 9: 572–581.
Holmes N, Morawska L . A review of dispersion modelling and its application to the dispersion of particles: An overview of different dispersion models available. Atmos Environ 2006; 40: 5902–5928.
Kumar P, Ketzel M, Vardoulakis S, Pirjola L, Britter R . Dynamics and dispersion modelling of nanoparticles from road traffic in the urban atmospheric environment—a review. J Aerosol Sci 2011; 42: 580–603.
Pratt GC, Dymond M, Ellickson K, Thé J . Validation of a novel air toxic risk model with air monitoring. Risk Anal 2012; 32: 96–112.
Pratt GC, Wu CY, Bock D, Adgate JL, Ramachandran G, Stock TH et al. Comparing air dispersion model predictions with measured concentrations of VOCs in urban communities. Environ Sci Techno 2004; 38: 1949–1959.
Smit R., Ntziachristos L., Boulter P. . Validation of road vehicle and traffic emission models—a review and meta-analysis. Atmos Environ 2010; 44: 2943–2953.
Beckerman B, Jerrett M, Brook JR, Verma DK, Arain MA, Finkelstein MM . Correlation of nitrogen dioxide with other traffic pollutants near a major expressway. Atmos Environ 2008; 42: 275–290.
Fruin S, Westerdahl D, Sax T, Sioutas C, Fine PM . Measurements and predictors of on-road ultrafine particle concentrations and associated pollutants in Los Angeles. Atmos Environ 2008; 42: 207–219.
He M, Dhaniyala S . Vertical and horizontal concentration distributions of ultrafine particles near a highway. Atmos Environ 2012; 46: 225–236.
Kaur S, Nieuwenhuijsen MJ . Determinants of personal exposure to PM2.5, ultrafine particle counts, and CO in a transport microenvironment. Environl Sci Technol 2009; 43: 4737–4743.
Kittelson DB, Watts WF, Johnson JP, Remerowki ML, Ische EE, Oberdörster G et al. On-road exposure to highway aerosols. 1. Aerosol and gas measurements. Inhal Toxicol 2004; 16 (Suppl 1): 31–39.
Löndahl J, Massling A, Swietlicki E, Bräuner EV, Ketzel M, Pagels J et al. Experimentally determined human respiratory tract deposition of airborne particles at a busy street. Environ Sci Technol 2009; 43: 4659–4664.
McAdam K, Steer P, Perrotta K . Using continuous sampling to examine the distribution of traffic related air pollution in proximity to a major road. Atmos Environ 2011; 45: 2080–2086.
McNabola A, Broderick BM, Gill LW 2008 Relative exposure to fine particulate matter and VOCs between transport microenvironments in Dublin: Personal exposure and uptake. Atmos Environ 42: 6496–6512.
US EPA Region 6 Compliance Assurance and Enforcement Division. RAIMI—Regional Air Impact Modeling Initiative, Region 6, US EPA. Retrieved 16 February 2012, from http://www.epa.gov/region6/6en/raimi/index.htm.
US EPA. Office of Transportation and Air Quality. MOVES2010b: Additional Toxics Added to MOVES, EPA-420-B-12-029 2012b.
US EPA. Office of Transportation and Air Quality. Kansas City PM Characterization Study Final Report Kansas City PM Characterization Study Final Report, rEPA420-R-08-009 2006.
US EPA. NATA | National-Scale Air Toxics Assessments | Technology Transfer Network Air Technical Web Site | US EPA. Retrieved 16 February 2012, from http://www.epa.gov/ttn/atw/natamain/index.html.
Venkatram A, Isakov V, Seila R, Baldauf R . Modeling the impacts of traffic emissions on air toxics concentrations near roadways. Atmos Environ 2009; 43: 3191–3199.
Aggarwal S, Jain R, Marshall JD . Real-time prediction of size-resolved ultrafine particulate matter on freeways. Environl Sci Technol 2012; 46: 2234–2241.
Bechle MJ, Millet DB, Marshall JD . Effects of income and urban form on urban NO2: global evidence from satellites. Environ Sci Technol 2011; 45: 4914–4919.
Briggs DJ, Collins S, Elliot P, Fischer P, Kingham S, Lebret E et al. Mapping urban air pollution using GIS: a regression- based approach. Int J Geogr Inf Sci 1997; 11: 699–718.
Liu L-JS, Tsai M-Y, Keidel D, Gemperli A, Ineichen A, Hazenkamp-von Arx M et al. Long-term exposure models for traffic related NO2 across geographically diverse areas over separate years. Atmos Environ 2012; 46: 460–471.
Marshall J, Mckone T, Deakin E, Nazaroff W . Inhalation of motor vehicle emissions: effects of urban population and land area. Atmos Environ 2005; 39: 283–295.
Marshall JD, Nethery E, Brauer M . Within-urban variability in ambient air pollution: comparison of estimation methods. Atmos Environ 2008; 42: 1359–1369.
Novotny EV, Bechle MJ, Millet DB, Marshall JD . National satellite-based land-use regression: NO2 in the United States. Environ Sci Technol 2011; 45: 4407–4414.
Ross Z, Jerrett M, Ito K, Tempalski B, Thurston G . A land use regression for predicting fine particulate matter concentrations in the New York City region. Atmos Environ 2007; 41: 2255–2269.
Skene KJ, Gent JF, McKay LA, Belanger K, Leaderer BP, Holford TR . Modeling effects of traffic and landscape characteristics on ambient nitrogen dioxide levels in Connecticut. Atmos Environ 2010; 44: 5156–5164.
Barzyk TM, George BJ, Vette AF, Williams RW, Croghan CW, Stevens CD . Development of a distance-to-roadway proximity metric to compare near-road pollutant levels to a central site monitor. Atmos Environ 2009; 43: 787–797.
Rose N, Cowie C, Gillett R, Marks GB . Weighted road density: a simple way of assigning traffic-related air pollution exposure. Atmos Environ 2009; 43: 5009–5014.
Gunier RB, Hertz A, Von Behren J, Reynolds P . Traffic density in California: socioeconomic and ethnic differences among potentially exposed children. J Expoe Anal Environ Epidemiol 2003; 13: 240–246.
National Research Council. Exposure Science in the 21st Century: A Vision and a Strategy. The National Academies Press: 195pp, 2012.
Eiguren-Fernandez A, Miguel AH . Size-resolved polycyclic aromatic hydrocarbon emission factors from on-road gasoline and diesel vehicles: temperature effect on the nuclei-mode. Environ Sci Technol 2012; 46: 2607–2615.
Johnson J, Kittelson D, Watts W . Source apportionment of diesel and spark ignition exhaust aerosol using on-road data from the Minneapolis metropolitan area. Atmos Environ 2005; 39: 2111–2121.
Pang X, Mu Y, Yuan J, He H . Carbonyls emission from ethanol-blended gasoline and biodiesel-ethanol-diesel used in engines. Atmos Environ 2008; 42: 1349–1358.
St Sauver JL, Grossardt BR, Leibson CL, Yawn BP, Melton LJ, Rocca WA . Generalizability of epidemiological findings and public health decisions: an illustration from the Rochester Epidemiology Project. Mayo Clinic Proc Mayo Clin 2012; 87: 151–160.
St Sauver JL, Grossardt BR, Yawn BP, Melton LJ, Rocca WA . Use of a medical records linkage system to enumerate a dynamic population over time: the Rochester epidemiology project. Am J Epidemiol 2011; 173: 1059–1068.
Minnesota Department of Health. Healthy Communities Count! Indicators of Community Health along the Central Corridor Light Rail Transit (LRT) Route, 2010. Retrieved 29 July 2012 from http://www.health.state.mn.us/divs/eh/hazardous/lightrail/report.html.
Schneider J, Kirchner U, Borrmann S, Vogt R, Scheer V . In situ measurements of particle number concentration, chemically resolved size distributions and black carbon content of traffic-related emissions on German motorways, rural roads and in city traffic. Atmos Environ 2008; 42: 4257–4268.
Gordon M, Staebler RM, Liggio J, Shao-Meng L, Wentzell J, Lu G et al. Measured and modeled variation in pollutant concentration near roadways. Atmos Environ 2012; 57: 138–145.
Hesterberg TW, Long CM, Sax SN, Lapin CA, McClellan RO, Bunn WB et al. Particulate Matter in New Technology Diesel Exhaust (NTDE) is Quantitatively and Qualitatively Very Different from that Found in Traditional Diesel Exhaust (TDE). J Air Waste Manag Assoc 2011; 61: 894–913.
Boruff BJ, Nathan A, Nijënstein S . Using GPS technology to re-examine operational definitions of “neighbourhood” in place-based health research. Int J Health Geographics 2012; 11: 22.
Boulos MNK, Berry G . Real-time locating systems (RTLS) in healthcare: a condensed primer. Int J Health Geographics 2012; 11: 25.
Wiehe SE, Carroll AE, Liu GC, Haberkorn KL, Hoch SC, Wilson JS et al. Using GPS-enabled cell phones to track the travel patterns of adolescents. Int JHealth Geographics 2008; 7: 22.
We thank Megan Forbes of the Minnesota Department of Transportation for her help in obtaining and understanding traffic count data and Shawn Nelson of the Minnesota Pollution Control Agency for assistance with geoprocessing. We also thank Julian Marshall and Matthew Bechle of the Civil Engineering Department at the University of Minnesota for allowing us to use their land-use regression data. This work was supported in part by grant #R833627010 (‘Measuring the Impacts of Particulate Matter Reductions by Environmental Health Outcome Indicators’) from the US Environmental Protection Agency. This study was reviewed and approved by the Olmsted Medical Center IRB for use of REP data.
About this article
Cite this article
Pratt, G., Parson, K., Shinoda, N. et al. Quantifying traffic exposure. J Expo Sci Environ Epidemiol 24, 290–296 (2014). https://doi.org/10.1038/jes.2013.51
- land-use regression
- geographic information system
- global positioning system
Prenatal air pollution exposure and neurodevelopment: A review and blueprint for a harmonized approach within ECHO
Environmental Research (2021)
The Journal of Allergy and Clinical Immunology: In Practice (2018)
Pediatric Allergy, Immunology, and Pulmonology (2016)
Aggregated GPS tracking of vehicles and its use as a proxy of traffic-related air pollution emissions
Atmospheric Environment (2016)
Application of plume analysis to build land use regression models from mobile sampling to improve model transferability
Atmospheric Environment (2016)