Secondary Logo

Journal Logo

Evaluation of the Danish AirGIS air pollution modeling system against measured concentrations of PM2.5, PM10, and black carbon

Hvidtfeldt, Ulla Arthura; Ketzel, Matthiasb; Sørensen, Mettea; Hertel, Oleb; Khan, Jibranb,,c; Brandt, Jørgenb; Raaschou-Nielsen, Olea,,b

Environmental Epidemiology: June 2018 - Volume 2 - Issue 2 - p e014
doi: 10.1097/EE9.0000000000000014
Original Research
Open
SDC

Background: Adverse health effects of air pollution have been reported in previous studies with varying methodological approaches to the exposure assessment. Measuring individual air pollution exposure for large-scale epidemiological studies is infeasible, calling for refined modeling tools. We evaluated the performance of the Geographical Information System–based air pollution and human exposure modeling system (AirGIS).

Methods: Modeled concentrations were evaluated against measured concentrations of particulate matter (PM) less than 10 and 2.5 μm in aerodynamic diameter (PM10/PM2.5) from two fixed-site monitoring stations (background and street) and from two measurement campaigns in Copenhagen, Denmark. Modeled concentrations of black carbon (BC) were evaluated against measured PM2.5 absorbance and PM10 absorbance.

Results: Mean concentrations measured in the four series were in the range of 10.4–15.3 μg/m3 for PM2.5 and 17.8–25.1 μg/m3 for PM10. The model underestimated by 7%–13% in comparison to the fixed-site monitoring stations. Correlation coefficients of 0.82 and 0.73 were observed for monthly and daily averages of measured and modeled PM2.5 at the background site and, correspondingly, 0.85 and 0.74 at the street site. The spatial variation, as evaluated from the two measurement campaigns, was also well reproduced. Correlation coefficients of 0.77 and 0.79 were observed for BC and PM2.5 absorbance and 0.76 for PM10 absorbance.

Conclusion: The AirGIS framework showed an overall high degree of accuracy and will be applicable to future epidemiological studies on health effects of PM and BC.

aDanish Cancer Society Research Center, Copenhagen, Denmark

bDepartment of Environmental Science, Aarhus University, Roskilde, Denmark

cDepartment of Chemical Engineering & Mary Cay O’Connor Process Safety Center, Texas A&M University at Qatar, Doha, Qatar.

Received: 6 February 2018; Accepted 3 April 2018

Published online 30 May 2018

This work was supported by Nordforsk under the Nordic Programme on Health and Welfare (Project No. 75007: NordicWelfAir - Understanding the link between Air pollution and Distribution of related Health Impacts and Welfare in the Nordic countries).

Access to the data and computing code to replicate the results reported in this article may be obtained by enquiry to the corresponding author.

Supplemental digital content is available through direct URL citations in the HTML and PDF versions of this article (www.epidem.com).

Corresponding author. Address: Diet, Genes and Environment, Danish Cancer Society Research Center, Strandboulevarden 49, 2100 Copenhagen Ø, Denmark. Tel. +45 35257247. E-mail address: ullah@cancer.dk (U.A. Hvidtfeldt)

This is an open-access article distributed under the terms of the Creative Commons Attribution-Non Commercial-No Derivatives License 4.0, where it is permissible to download and share the work provided it is properly cited. The work cannot be changed in any way or used commercially.

Back to Top | Article Outline

What this study adds

AirGIS is widely applied in modeling systems and epidemiological studies on health effects of NO2 and NOx. The newly updated system facilitates calculation of particulate matter and black carbon at the address level for the period from 1979 onwards. The system performs well both in regard to spatial and temporal variation when validated against routine fixed-site monitoring stations and measurement campaigns, representing various levels of traffic density/composition, street geometry, degree of urbanization, as well as time windows. These findings are essential for the application of the modeled exposure data in future epidemiological studies on health effects of particulate matter and black carbon.

Health effects of air pollution have been extensively investigated in epidemiological studies over the past decades. In the early studies, short-term exposures to the gaseous nitrogen oxide pollutants (NO2 and NOx) and particles have been linked to cardiovascular and respiratory outcomes.1–3 In later years, with the increasing availability of exposure data, adverse health effects have also been reported in relation to long-term exposures and mainly ischemic heart disease, stroke, chronic obstructive pulmonary disease, diabetes, and lung cancer.4–19

The biological mechanisms underlying the observed associations between air pollution exposure and health such as systemic inflammation, oxidative stress, blood coagulation, elevated blood pressure, and so forth include both acute (within hours or days) and chronic (months to years) responses.3 Thus, the observed health effects might reflect responses to both short- and long-term exposure. Measuring air pollution exposures at the individual level for large-scale epidemiological studies would be extremely resource demanding and practically impossible; hence, large-scale studies of health effects of air pollution call for the use of refined modeling tools for exposure assessment. A wide range of methods have been applied in previous studies. One simple approach is use of data from nearest monitoring station as a proxy for individual exposure levels,7,20–22 which—depending on location and the number of monitoring sites—is a rather crude measure of exposure. For addressing intra-urban exposure contrasts, simple measures of residential proximity to major roads or residential traffic load have been used,4–7 but more sophisticated methods such as land use regression models10–16 and dispersion models4–6 are also widely applied now.

The Geographical Information System–based air pollution and human exposure modeling system (AirGIS) is part of the Danish Air-Quality Monitoring Programme23,24 and the multiscale integrated dispersion modeling system THOR.25–27 The AirGIS and the Operational Street Pollution Model (OSPM), which estimates the local street contribution of the AirGIS, have been applied in modeling systems across the world,28–33 and a vast number of epidemiological studies on the health effects related to NO2 and NOx,4–6,34–48 and recently also to particulate matter (PM),49,50 have been conducted. A previous study addressed the performance of AirGIS in relation to temporal variations in concentrations of NO2, NOx, CO, and O3 by comparing modeled concentrations to measurements at five permanent stations from the Danish Monitoring Network.51 In addition, the OSPM has been evaluated previously.33,52 The performance of the system was also evaluated in relation to geographical variability by applying a Danish sample of measured monthly mean concentrations of NO2.53 The measured and predicted concentrations corresponded well for both long-term averages (annual and monthly) as well as on shorter terms (hourly and daily), and the spatial variability was also reproduced well. The AirGIS is continuously being further developed and improved—this concerns model parameterizations as well as detail and quality of the various input data. One of the more recent updates concerns the Urban Background Model (UBM) for which the model parameterization has been extensively updated, and the model is now run for the entire country on a high-resolution grid as a routine tool in relation to the THOR system.

In this article, we evaluate a recent feature of AirGIS, that is, calculation of PM concentrations. For that purpose, we used measured concentrations of ambient PM less than 10 and 2.5 μm in aerodynamic diameter (PM10 and PM2.5, respectively) from the routine fixed site monitoring stations under the national monitoring program and PM10, PM2.5, and PM absorbance (blackness of the PM filters) from two previous measurement campaigns in the Copenhagen area of Denmark. The measurements represent various levels of traffic density/composition, street geometry, and degree of urbanization as well as time windows.

Back to Top | Article Outline

Methods

The AirGIS dispersion modeling system

The AirGIS contains an Urban Landscape Model based on digital maps of building foot prints with building heights as parameter. It further contains a digital street and traffic map and various national databases and is operated within a GIS. The AirGIS was developed by the Department of Environmental Science at Aarhus University (former Danish National Environmental Research Institute) in the late 1990s.54 The system enables the calculation of ambient air pollution at high temporal (hourly basis) and spatial (individual address) resolutions. The air pollution at a specific location is modeled as the sum of three contributions: (1) Regional background contributions, that is, from sources outside the urban area such as power plants, industry, residential heating, and so forth, modeled with a regional transport model (the Danish Eulerian Hemispheric Model, DEHM)55,56; (2) Urban background contributions in a 1 × 1 km2 grid resolution calculated with the UBM,25,57 and in this case applied for the whole of Denmark, taking into account the emission density originating from all types of emissions estimated with the Spatial high resolution distribution model for emissions to air (SPREAD) methodology58 and average building cover and height;59 and (3) Local air pollution contribution from local traffic computed by OSPM and taking into account data on traffic (intensity, speed, and type), emission factors for the car fleet, street and building geometry, and meteorology.60 The three contributions to the system are illustrated in Figure 1. In this system, air pollution can be calculated for any address location in Denmark. The multiscale integrated model system (DEHM, UBM, and OSPM) is called the THOR system, and when the OSPM is operated at individual addresses in GIS, this part is called AirGIS. An example of the spatial distribution of PM2.5 and PM10 in Denmark for the year 2013 is provided in Figure 2.

Figure 1

Figure 1

Figure 2

Figure 2

Results of previous versions of AirGIS have been evaluated in a Danish context against measured NO2, NOx, CO, and O3.51,54 The system is continuously being refined, for example, with regards to input data on traffic, other air pollution sources, and building configurations, and extended with necessary historical input data on emission factors (e.g., more accurate estimates of domestic wood burning emissions, now based on actual registrations by chimney sweepers) and background concentrations enabling the calculation of particles back in time. Also, the general GIS programming environment has been updated recently.61 The current system allows for the calculation of various pollutants such as the gases NO2, NOx, CO, O3, and SO2. The AirGIS system is a fully deterministic dispersion modeling system, and its predictions are not dependent on any observed concentrations. The latter are only used for validation of the system. The most recent update also facilitates calculation of PM2.5, PM10, and black carbon (BC) in PM for the period from 1979 and onwards. The treatment of chemical reactions and PM formation differs for the different models at the different scales: The (regional) DEHM model contains the most complete chemical scheme including formation of PM components as secondary inorganic aerosol (SIA) and secondary organic aerosol (SOA) from gaseous precursors because these processes are relevant in the long time and spatial scales that DEHM is covering. In contrast, the (urban) UBM and (local) OSPM models cover only short transport times and relatively small spatial scales where only few fast chemical reactions (as, e.g., NO-NO2-O3 conversion) are relevant to consider together with the emissions of primary PM and other pollutants. Further chemical reactions and secondary particle formations are therefore not explicitly treated in UBM and OSPM.

Back to Top | Article Outline

Measured air pollution data

The modeled temporal variation was evaluated against two permanent monitoring stations in Copenhagen representing urban background (HCOE, roof station at the H.C. Ørsted Institute) and street (JGTV, Jagtvej kerbside street canyon).24 In recent years, the Danish air quality monitoring network has measured PM10 and PM2.5 by Low Volume Sampling (LVS) at these two sites. At HCOE, PM10 was measured from June 2013, and PM2.5 was measured from August 2012 and onwards, whereas the JGTV monitoring station has PM data from January 2014 and onwards. The LVS method collects particles on filters for 24-hour intervals with a flow of 2.3 m3/h and subsequent gravimetric determination of the sampled dust mass in the laboratory.24

In addition, two data sets with front door measurements were applied to evaluate the performance of the modeling system with regards to spatial variation (Figure 3). The first set of measurements (“Measurement campaign 1”) was carried out during the time period November 1999 to September 2000 as part of a Danish research project and followed 30 subjects, in up to four measurement campaigns of NO2, PM2.5, and the light absorption coefficient of the PM2.5 measurement filters (denoted as PM2.5 absorbance). Participants were recruited among voluntary students living in central parts of Copenhagen with no inclusion criteria relating to amount of traffic or other sources of air pollution near the residence. New subjects were recruited in situations where subjects left the measurement campaign, resulting in a total of 98 front-door measurements. The measurements were distributed evenly over the four seasons and were performed for 48-hour sampling periods for each study subject, with half the subjects being monitored from Monday to Wednesday and the other half from Wednesday to Friday. The PM2.5 samples were collected using a BGI400 pump operating at 4 l/min, a KTL PM2.5 cyclone, and a 37-mm Teflon filter. Further details on sampling methods and laboratory analyses are given in Sørensen et al (2005).62

Figure 3

Figure 3

The second data set of measurements (“Measurement campaign 2”) was conducted as part of the European Study of Cohorts for Air Pollution Effects (ESCAPE), with focus on the spatial variation of particles and nitrogen oxide pollutants within many different study areas in Europe. The Danish measurement campaigns took place from November 2009 to October 2010 and covered 20 sites for the measurements of PM2.5 and PM10, as well as the PM2.5 absorbance and PM10 absorbance. The sites were located in the Greater Copenhagen area and selected to represent three different levels of air pollution: (1) rural areas not directly influenced by traffic, (2) urban areas at least 50 m away from traffic, and (3) areas in streets with a high traffic intensity (>10,000 vehicles per day). Street sites were over-represented and sampled to represent different traffic intensities, distances to the road, and different street configurations. A 14-day measurement was conducted three times for each site during different seasons. Failed measurements were repeated in a later round to ensure three valid measurements for each site. Thus, the data set included 60 measurements of particles. The PM2.5 and PM10 samples were collected using Harvard impactors with a sampling flow rate of 10 l/min on 37 mm Teflon filters. Further details are given in Eeftens et al.63

Back to Top | Article Outline

Statistical methods

Measured concentrations of air pollution were compared with concentrations modeled at the geo-coded address location. The receptor point for modeling was at the façade of the building at 2 m height. All AirGIS calculations were performed on an hourly basis, and thus modeled concentrations were averaged over the time period corresponding to that of the measured values. Modeled values were plotted against measured concentrations for each single measured value. Statistical procedures for comparing modeled to measured concentrations included the following:

The mean bias, which is a simple measure of the extent to which the model over or under estimates, given by:

The root mean squared error, which is the standard deviation of the error term, that is, a measure of unexplained variation, as a measure of how close modeled values are to the measured given as:

The coefficient of variation given by the ratio of the standard deviation to the mean, which is the standard deviation of the error term as a measure of the relative variability of the sample comparable across modeled and measured means:

Pearson’s correlation coefficients (r) were computed to evaluate the strength of the linear relation between measured and modeled values:

Also, the Spearman’s rank order coefficient was computed to evaluate monotonic relations in the data:

The percentage of explained variability was expressed as the coefficient of determination, given by the relation between the deviations of observations from their predicted values and deviations of observations from their mean:

The statistical analyses were carried out in SAS version 9.464 and R version 3.1.3.65

Back to Top | Article Outline

Results

Descriptive statistics for modeled and measured concentration

The four series of measurements in the Greater Copenhagen area showed measured PM2.5 average concentrations of 11.8, 15.3, 10.4, and 11.3 μg/m3, and the three series of PM10 measurements showed measured average concentrations of 17.8, 25.1, and 17.8 μg/m3 (Table 1). The measured concentrations were 7%–13% higher than the corresponding modeled concentrations for the measurement series at the fixed site monitoring stations and 24%–43% lower than the corresponding modeled concentrations in measurement campaign 1 and 2.

Table 1

Table 1

The coefficient of variation for PM2.5 concentration was generally higher for measured than for modeled concentrations, whereas those for the PM10 series were quite similar for measured and modeled concentration.

Back to Top | Article Outline

Correlation between time-series of modeled and measured concentrations at two fixed-site monitoring stations

Figure 4 shows that the model reproduced the monthly averages of PM2.5 at a background (roof top) site (“HCOE”) in Copenhagen well, and this was confirmed by a Pearson’s correlation coefficient of 0.82 for the series of 53 monthly observations (Table 2). Daily averages of PM2.5 for the same time period at the same location produced a Pearson’s correlation coefficient of 0.73 (N = 1,517) (results not shown). As shown in Figure 5, the model also reproduced measured PM2.5 concentrations well in a street with dense traffic (“JGTV”). Here, the Pearson’s correlation coefficient was 0.85 when based on monthly averages (N = 36; Table 2) and 0.74 when based on daily averages (N = 1,071; results not shown). Supplemental digital content 1 (Figure s1; http://links.lww.com/EE/A8) shows the variation in daily means of measured and modeled PM2.5 at the HCOE station and monthly means at the JGTV station. The temporal variation in measured PM10 was also reproduced well by the model (see Figure s2 and s3, Supplemental digital content 2–3; http://links.lww.com/EE/A8), although the correlation between measured and modeled concentration was slightly lower than that for PM2.5 (Table 2).

Table 2

Table 2

Figure 4

Figure 4

Figure 5

Figure 5

Back to Top | Article Outline

Correlation between modeled and measured concentrations for two campaigns at various locations

Figure 6 shows that the AirGIS model reproduced 2-week average concentrations of PM2.5 and PM10 in Measurement campaign 2 with a high degree of accuracy, which was reflected in Pearson’s correlation coefficients of 0.80 and 0.74 (Table 2). The figure shows more outliers for PM2.5 in Measurement campaign 1, which was reflected in a lower Pearson’s correlation coefficient of 0.67 (Table 2). A sensitivity analysis showed a seasonal difference in the correlation, with high correlation (r = 0.81) in the warmer season (>8°C) and lower correlation (r = 0.60) in the colder season (see Figure s4, Supplemental digital content 4; http://links.lww.com/EE/A8).

Figure 6

Figure 6

The results of the comparisons between modeled concentrations of BC and measured PM2.5 absorbance and PM10 absorbance from the two campaigns are presented in Figure 7. Pearson’s correlation coefficients of 0.79 and 0.77 were observed for PM2.5 absorbance in Measurement campaigns 1 and 2, respectively, and 0.76 for PM10 absorbance in Measurement campaign 2.

Figure 7

Figure 7

Back to Top | Article Outline

Discussion

In this article, we evaluated calculations of PM2.5 and PM10 by AirGIS against concentrations measured at two fixed-site monitoring stations representing urban background and street, respectively, and various address points in the Copenhagen area from two measurement campaigns. BC was evaluated against measured PM2.5 absorbance and PM10 absorbance from the two campaigns. Overall the concentrations modeled by AirGIS correlated well with the measured concentrations in regard to reproducing both temporal and spatial variation. In absolute terms, the modeled PM concentrations were similar to those measured at the fixed-site monitoring stations, whereas the comparison with data from measurement campaigns using low-cost, portable monitoring equipment showed substantially lower measured concentrations. Although the mean bias varied across the measurement campaigns and fixed-site monitor measurements, the correlation coefficients were similarly high.

The AirGIS model has previously been validated in relation to NO2, NOx, CO, and O3 by comparing modeled concentrations to measurements at five permanent stations from the Danish Monitoring Network.51 The performance of the AirGIS model in relation to PM2.5, PM10, and BC can be considered similar for both the long- and short term averages. Likewise, models comparable to the AirGIS have been developed and evaluated, for example, the Swedish national SIMAIR Dispersion modeling system including long-range transport, local traffic exhaust, and road dust.66 Modeled outdoor PM1 and PM10 levels have been evaluated against annual averages and daily time series in urban background settings (monitoring station data from three large cities) and heavily trafficked settings (four cities). Differences between annual measured and modeled PM10 urban background concentrations were below 20% with correlation coefficients for the time series of daily average PM10 varying between 0.57 and 0.76. For the traffic sites, corresponding differences for the annual means were below 10%, with correlations for the daily averages ranging between 0.56 and 0.71. Thus, the performance of AirGIS may be considered similar. The relatively large differences between average concentrations of measured and modeled PM in the two measurement campaigns could be due to overprediction by the AirGIS model or due to the monitoring equipment systematically measuring too low concentrations. The PM concentrations calculated by AirGIS were on average very similar to those measured by the fixed-site monitoring stations applying LVS gravimetric determination. Since LVS is known to measure PM at high precision and according to the EU-reference method,24 we believe that the differences in PM mean concentrations are due to the samplers of the monitoring campaigns measuring systematically too low.67 This notion is supported by a comparison between the Harvard impactors used in Measuring campaign 2 with a high-quality fixed-site monitoring equipment, the SM200 (OPSIS, Sweden), comparable to the LVS.24 Comparing the Harvard Impactor measurements to SM200 β-gauges measurements from identical time periods and placements showed a high correlation (rp = 0.93) but a deviation in absolute terms, in which the Harvard Impactor systematically measured lower values of PM2.5 (31% lower on average; see Figure s5, Supplemental digital content 5; http://links.lww.com/EE/A8). Likewise, the correlation between the Harvard Impactors and SM200 measurements of PM10 were high (rp = 0.77), with Harvard Impactor measurements being 25% lower on average.

Differences were also observed between the two measurement campaigns. For PM2.5, the performance of the model in terms of correlation was considerably higher when evaluated against measurements from Campaign 2 than Campaign 1. One potential explanation for this difference may be the length of measuring time per sample. Our results showed a better correlation between PM2.5 modeled and measured at monitoring station when based on monthly averages compared to daily averages. The reason for a better correlation for longer term averages might be that modeled short-term concentrations are more sensitive to unresolved exceptional conditions or uncertainties in various model parameters or inputs (e.g., meteorology or emissions),68 and the measurement uncertainty might be higher at shorter sampling times. In Measurement campaign 1, data were collected for 48-hour periods, whereas Measurement campaign 2 included 14-day averages. Another explanation could be differences in equipment applied. A previous study comparing the monitoring equipment used in Measurement campaigns 1 and 2 showed no sign of systematic differences.69 However, the equipment in Measurement campaign 1 was smaller and mounted on a bike parked at the front door, whereas Measurement campaign 2 was a stationary setup using a larger pump and sampling head—which may imply better precision. As indicated by previous analyses of data from Measurement campaign 1,62 low temperature may have affected the measurement precision negatively, and restricting the data from Measurement campaign 1 to measurements in the warmer months of April through September revealed results comparable to those of Measurement campaign 2.

The PM2.5 absorbance and PM10 absorbance measurements applied represent the blackness of the PM2.5 and PM10 filters. BC is considered the dominant light-absorbing substance in ambient air,70 justifying the comparison in the present study between modeled BC and measured PM absorbance. A previous study showed a high correlation between measurements of elemental carbon and PM2.5 absorbance (r = 0.93),71 and we found high correlations between modeled BC concentrations and measured levels of both PM2.5 absorbance and PM10 absorbance. Since most of the BC is found in the fine fraction, similar results for PM10 absorbance and PM2.5 absorbance were expected.72 However, other light-absorption substances than BC may influence the PM absorbance measures, and thus a higher correlation would be expected if BC could be perfectly singled out from these light absorbance measurements.

The perspectives in using the AirGIS model for exposure assessment in future epidemiological studies rely on the ability of the model to calculate air pollution concentrations correctly. The model depends on comprehensive and detailed input data, and uncertainties in the input data—which are inevitable when modeling exposure for a large number of locations—will affect the model output. Another point to consider is the excessive computational requirements of the model which has, however, improved considerably in the latest version of the AirGIS.

Systematic over- or under-prediction by the model would likely affect the magnitude of the estimated exposure–response association, and a poor correlation between modeled and true concentrations could make it difficult to detect true exposure–response associations. The present study showed that average PM concentrations calculated by AirGIS were in good compliance with measured concentrations obtained from the most reliable measurement devices (i.e., low-volume samplers at the fixed-site monitoring stations). Further, the modeled concentrations correlated well with series of measurements characterized by both temporal and geographical variation. However, although we included all available data from measurement campaigns and the Danish air quality monitoring network, the comparisons were limited by the geographical coverage of measurements (the larger Copenhagen area) and the limited time frame of the measurements (monitoring stations, 2012–2017; measurement campaigns, 1999–2000 and 2009–2010). Nevertheless, even the best air pollution model will entail error. The consequence of the error for the exposure–response function estimated in an epidemiological study depends both on the amount of error and the underlying structure of the error. Two dimensions of errors are typically dominating: (1) The “classical” measurement error type, in which the exposure error is independent of the true exposure, and (2) the “Berkson” error type, in which the error is independent of the modeled exposure—e.g., in relation to aggregation of data to a group level. When estimating air pollution concentrations with a model, we would expect a Berkson type error; however, model-based exposure assessment can also imply an additional element of classical error, for example, deriving from uncertainty of the input data for the exposure prediction model. Classical error is expected to bias the effect estimates toward the null, whereas the Berkson-type error does not affect the size of the risk estimate. Berkson error does, however, reduce the precision of the risk estimate, leading to wider confidence intervals than would have been observed without error.73,74 Several studies have addressed differences in air pollution modeling approaches, related exposure measurement error, and methods for correction of risk estimates affected by such error,75–77 but there is a need for future studies addressing the amount and type of error associated with different air pollution exposure assessment methods and for the development of methods to correct the estimated exposure–response associations for such errors.

In conclusion, the results of the study imply that the AirGIS modeling system performs well both in regard to spatial and temporal variation, with minor absolute differences between measured and modeled concentrations and high correlation coefficients, and will be applicable to future epidemiological studies on health effects of PM and BC.

Back to Top | Article Outline

Conflict of interest statement

The authors declare no conflicts of interest in relation to the work described.

Back to Top | Article Outline

Acknowledgments

The research was funded by Nordforsk under the Nordic Programme on Health and Welfare (Project No. 75007: NordicWelfAir - Understanding the link between Air pollution and Distribution of related Health Impacts and Welfare in the Nordic countries).

Back to Top | Article Outline

References

1. Mills IC, Atkinson RW, Kang S, Walton H, Anderson HR. Quantitative systematic review of the associations between short-term exposure to nitrogen dioxide and mortality and hospital admissions. BMJ Open 2015; 5e006946
2. Shah AS V, Lee KK, McAllister DA, Hunter A, Nair H, Whiteley W, et al. Short term exposure to air pollution and stroke: systematic review and meta-analysis. BMJ 2015; 350h1295
3. Brook RD, Rajagopalan S, Pope CA, Brook JR, Bhatnagar A, Diez-Roux A V., et al. Particulate matter air pollution and cardiovascular disease. Circulation 2010; 1212331–2378
4. Andersen ZJ, Kristiansen LC, Andersen KK, Olsen TS, Hvidberg M, Jensen SS, et al. Stroke and long-term exposure to outdoor air pollution from nitrogen dioxide: a cohort study. Stroke 2012; 43320–325
5. Andersen ZJ, Hvidberg M, Jensen SS, Ketzel M, Loft S, Sørensen M, et al. Chronic obstructive pulmonary disease and long-term exposure to traffic-related air pollution: a cohort study. Am J Respir Crit Care Med 2011; 183455–461
6. Raaschou-Nielsen O, Andersen ZJ, Jensen SS, Ketzel M, Sørensen M, Hansen J, et al. Traffic air pollution and mortality from cardiovascular disease and all causes: a Danish cohort study. Environ Health 2012; 1160
7. Heinrich J, Thiering E, Rzehak P, Krämer U, Hochadel M, Rauchfuss KM, et al. Long-term exposure to NO2 and PM10 and all-cause and cause-specific mortality in a prospective cohort of women. Occup Environ Med 2013; 70179–86
8. Chen G, Wan X, Yang G, Zou X. Traffic-related air pollution and lung cancer: a meta-analysis. Thorac cancer 2015; 6307–318
9. Hamra GB, Laden F, Cohen AJ, Raaschou-Nielsen O, Brauer M, Loomis D. Lung cancer and exposure to nitrogen dioxide and traffic: a systematic review and meta-analysis. Environ Health Perspect 2015; 1231107–1112
10. Dijkema MBA, van Strien RT, van der Zee SC, Mallant SF, Fischer P, Hoek G, et al. Spatial variation in nitrogen dioxide concentrations and cardiopulmonary hospital admissions. Environ Res 2016; 151721–727
11. Beelen R, Hoek G, van den Brandt PA, Goldbohm RA, Fischer P, Schouten LJ, et al. Long-term effects of traffic-related air pollution on mortality in a Dutch cohort (NLCS-AIR study). Environ Health Perspect 2008; 116196–202
12. Cesaroni G, Badaloni C, Gariazzo C, Stafoggia M, Sozzi R, Davoli M, et al. Long-term exposure to urban air pollution and mortality in a cohort of more than a million adults in rome. Environ Health Perspect 2013; 121324–331
13. Beelen R, Raaschou-Nielsen O, Stafoggia M, Andersen ZJ, Weinmayr G, Hoffmann B, et al. Effects of long-term exposure to air pollution on natural-cause mortality: an analysis of 22 European cohorts within the multicentre ESCAPE project. Lancet 2014; 383785–795
14. Stafoggia M, Cesaroni G, Peters A, Andersen ZJ, Badaloni C, Beelen R, et al. Long-term exposure to ambient air pollution and incidence of cerebrovascular events: results from 11 European cohorts within the ESCAPE project. Environ Health Perspect 2014; 122919–925
15. Cesaroni G, Forastiere F, Stafoggia M, Andersen ZJ, Badaloni C, Beelen R, et al. Long term exposure to ambient air pollution and incidence of acute coronary events: prospective cohort study and meta-analysis in 11 European cohorts from the ESCAPE Project. BMJ 2014; 348f7412
16. Raaschou-Nielsen O, Beelen R, Wang M, Hoek G, Andersen ZJ, Hoffmann B, et al. Particulate matter air pollution components and risk for lung cancer. Environ Int 2016; 8766–73
17. Hamra GB, Guha N, Cohen A, Laden F, Raaschou-Nielsen O, Samet JM, et al. Outdoor particulate matter exposure and lung cancer: a systematic review and meta-analysis. Environ Health Perspect 2014; 122906–911
18. Shanley RP, Hayes RB, Cromar KR, Ito K, Gordon T, Ahn J. Particulate air pollution and clinical cardiovascular disease risk factors. Epidemiology 2016; 27291–298
19. Eze IC, Hemkens LG, Bucher HC, Hoffmann B, Schindler C, Künzli N, et al. Association between ambient air pollution and diabetes mellitus in Europe and North America: systematic review and meta-analysis. Environ Health Perspect 2015; 123381–389
20. Dockery DW, Pope CA, Xu X, Spengler JD, Ware JH, Fay ME, et al. An association between air pollution and mortality in six U.S. cities. N Engl J Med 1993; 3291753–1759
21. Chen X, Zhang L, Huang J, Song F, Zhang L, Qian Z, et al. Long-term exposure to urban air pollution and lung cancer mortality: a 12-year cohort study in Northern China. Sci Total Environ 2016; 571855–861
22. Lin H, Liu T, Xiao J, Zeng W, Li X, Guo L, et al. Mortality burden of ambient fine particulate air pollution in six Chinese cities: results from the Pearl River Delta study. Environ Int 2016; 9691–97
23. Hertel O, Ellermann T, Palmgren F, Berkowicz R, Løfstrøm P, Frohn LM, et al. Integrated air-quality monitoring—combined use of measurements and models in monitoring programmes. Environ Chem 2007; 465–74
24. Ellermann T, Nygaard J, Nøjgaard JK, Nordstrøm C, Brandt J, Christensen J, et al. The Danish Air Quality Monitoring Programme—Annual summary for 2015. Aarhus University, DCE—Danish Centre for Environment and Energy ©: Aarhus.2015. http://dce.au.dk/en
25. Brandt J, Christensen JH, Frohn LM, Palmgren F, Berkowicz R, Zlatev Z. Operational air pollution forecasts from European to local scale. PERGAMON Atmos Environ 2001; 3591–98
26. Brandt J, Christensen JH, Frohn LM, Berkowicz R. Operational air pollution forecasts from regional scale to urban street scale. Part 1: system description. Pergamon Phys Chem Earth (B) 2001; 26781–786
27. Brandt J, Christensen JH, Frohn LM, Berkowicz R. Operational air pollution forecasts from regional scale to urban street scale. Part 2: performance evaluation. Phys Chem Earth (B) 2001; 26825–830
28. Kousa A, Kukkonen J, Karppinen A, Aarnio P, Koskentalo T. Statistical and diagnostic evaluation of a new-generation urban dispersion modelling system against an extensive dataset in the Helsinki area. Atmos Environ 2001; 354617–4628
29. Jensen SS, Larson T, Deepti KC, Kaufman JD. Modeling traffic air pollution in street canyons in New York City for intra-urban exposure assessment in the US Multi-Ethnic Study of atherosclerosis and air pollution. Atmos Environ 2009; 434544–4556
30. Mensink C, Cosemans G. From traffic flow simulations to pollutant concentrations in street canyons and backyards. Environ Model Softw 2008; 23288–295
31. Lazi L, Ani Ci M, Sevi C U, Miji Z, Vukovi G, Ili L. Traffic contribution to air pollution in urban street canyons: integrated application of the OSPM, moss biomonitoring and spectral analysis. Atmos Environ 2016; 141347–360
32. Righi S, Lucialli P, Pollini E. Statistical and diagnostic evaluation of the ADMS-Urban model compared with an urban air quality monitoring network. Atmos Environ433850–3857
33. Kakosimos KE, Hertel O, Ketzel M, Berkowicz R. Operational Street Pollution Model (OSPM)—a review of performed application and validation studies, and future prospects. Environ Chem 2010; 7485
34. Andersen ZJ, Raaschou-Nielsen O, Ketzel M, Jensen SS, Hvidberg M, Loft S, et al. Diabetes incidence and long-term exposure to air pollution: a cohort study. Diabetes Care 2012; 3592–98
35. Andersen ZJ, de Nazelle A, Mendez MA, Garcia-Aymerich J, Hertel O, Tjønneland A, et al. A study of the combined effects of physical activity and air pollution on mortality in elderly urban residents: the Danish Diet, Cancer, and Health Cohort. Environ Health Perspect 2015; 123557–563
36. Raaschou-Nielsen O, Bak H, Sørensen M, Jensen SS, Ketzel M, Hvidberg M, et al. Air pollution from traffic and risk for lung cancer in three Danish cohorts. Cancer Epidemiol Biomarkers Prev 2010; 191284–1291
37. Raaschou-Nielsen O, Ketzel M, Harbo Poulsen A, Sørensen M. Traffic-related air pollution and risk for leukaemia of an adult population. Int J Cancer 2016; 1381111–1117
38. Lee P-C, Raaschou-Nielsen O, Lill CM, Bertram L, Sinsheimer JS, Hansen J, et al. Gene-environment interactions linking air pollution and inflammation in Parkinson’s disease. Environ Res 2016; 151713–720
39. Fisher JE, Loft S, Ulrik CS, Raaschou-Nielsen O, Hertel O, Tjønneland A, et al. Physical activity, air pollution and the risk of asthma and chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2016; 194855–865
40. Hjortebjerg D, Andersen AMN, Ketzel M, Pedersen M, Raaschou-Nielsen O, Sørensen M. Associations between maternal exposure to air pollution and traffic noise and newborn’s size at birth: A cohort study. Environ Int 2016; 951–7
41. Monrad M, Sajadieh A, Christensen JS, Ketzel M, Raaschou-Nielsen O, Tjønneland A, et al. Long-term exposure to traffic-related air pollution and risk of incident atrial fibrillation: a cohort study. Environ Health Perspect 2017; 125422–427
42. Poulsen AH, Sørensen M, Andersen ZJ, Ketzel M, Raaschou-Nielsen O. Air pollution from traffic and risk for brain tumors: a nationwide study in Denmark. Cancer Causes Control 2016; 27473–480
43. Sørensen M, Hjortebjerg D, Eriksen KT, Ketzel M, Tjønneland A, Overvad K, et al. Exposure to long-term air pollution and road traffic noise in relation to cholesterol: a cross-sectional study. Environ Int 2015; 85238–243
44. Ritz B, Lee P-C, Hansen J, Lassen CF, Ketzel M, Sørensen M, et al. Traffic-related air pollution and parkinson’s disease in denmark: a case–control study. Environ Health Perspect 2016; 124351–356
45. Sørensen M, Lühdorf P, Ketzel M, Andersen ZJ, Tjønneland A, Overvad K, et al. Combined effects of road traffic noise and ambient air pollution in relation to risk for stroke? Environ Res 2014; 13349–55
46. Pedersen M, Halldorsson TI, Olsen SF, Hjortebjerg D, Ketzel M, Grandström C, et al. Impact of road traffic pollution on pre-eclampsia and pregnancy-induced hypertensive disorders. Epidemiology 2017; 2899–106
47. Raaschou-Nielsen O, Sørensen M, Ketzel M, Hertel O, Loft S, Tjønneland A, et al. Long-term exposure to traffic-related air pollution and diabetes-associated mortality: a cohort study. Diabetologia 2012; 5636–46
48. Huynh S, von Euler-Chelpin M, Raaschou-Nielsen O, Hertel O, Tjønneland A, Lynge E, et al. Long-term exposure to air pollution and mammographic density in the Danish Diet, Cancer and Health cohort. Environ Health 2015; 1431
49. Jørgensen JT, Johansen MS, Ravnskjær L, Andersen KK, Bräuner EV, Loft S, et al. Long-term exposure to ambient air pollution and incidence of brain tumours: The Danish Nurse Cohort. Neurotoxicology 2016; 55122–130
50. Stockfelt L, Andersson EM, Molnár P, Gidhagen L, Segersson D, Rosengren A, et al. Long-term effects of total and source-specific particulate air pollution on incident cardiovascular disease in Gothenburg, Sweden. Environ Res 2017; 15861–71
51. Ketzel M, Berkowicz R, Hvidberg M, Jensen SS, Nielsen OR. Evaluation of AirGIS: a GIS-based air pollution and human exposure modelling system. Int J Environ Pollut 2011; 47226
52. Ketzel M, Jensen SS, Brandt J, Ellermann T, Olesen HR, Berkowicz R, et al. Evaluation of the street pollution model OSPM for measurements at 12 streets stations using a newly developed and freely available evaluation tool. J Civ Environ Eng 2012. 1–11
53. Raaschou-Nielsen O, Hertel O, Vignati E, Berkowicz R, Jensen SS, Larsen VB, et al. An air pollution model for use in epidemiological studies: evaluation with measured levels of nitrogen dioxide and benzene. J Expo Anal Environ Epidemiol 2000; 104–14
54. Jensen SS, Berkowicz R, Sten Hansen H, Hertel O. A Danish decision-support GIS tool for management of urban air quality and human exposures. Transp Res Part D Transp Environ 2001; 6229–241
55. Brandt J, Silver JD, Frohn LM, Geels C, Gross A, Hansen AB, et al. An integrated model study for Europe and North America using the Danish Eulerian Hemispheric Model with focus on intercontinental transport of air pollution. Atmos Environ 2012; 53156–176
56. Frohn LM, Christensen JH, Brandt J. Development of a high-resolution nested air pollution model the numerical approach. J Comput Phys 2002; 17968–94
57. Brandt J, Christensen JH, Frohn LM, Berkowicz R. Air pollution forecasting from regional to urban street scale––implementation and validation for two cities in Denmark. Phys Chem Earth, Parts A/B/C 2003; 28335–344
58. Plejdrup MS, Gyldenkærne S. Spatial distribution of emissions to air - the SPREAD model.NERI Technical Report no. 823. Aarhus, 2011. Available at: http://www.dmu.dk/Pub/FR823.pdf. Accessed March 14, 2017
59. Berkowicz R. A simple model for urban background pollution. Environ Monit Assess 2000; 65259–267
60. Berkowicz R. OSPM—a parameterised street pollution model. Environ Monit Assess 2000; 65323–331
61. Khan J, Ketzel M, Kakosimos KE, Hvidtfeldt UA, Sørensen M, Raaschou-Nielsen O, et al. Towards AirGIS re-development—a GIS based air pollution and human exposure modelling system.18th International Conference on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes2017Bologna, Italy
62. Sørensen M, Loft S, Andersen HV, Raaschou-Nielsen O, Skovgaard LT, Knudsen LE, et al. Personal exposure to PM2.5, black smoke and NO2 in Copenhagen: relationship to bedroom and outdoor concentrations covering seasonal variation. J Expo Anal Environ Epidemiol 2005; 15413–422
63. Eeftens M, Tsai M-Y, Ampe C, Anwander B, Beelen R, Bellander T, et al. Spatial variation of PM2.5, PM10, PM2.5 absorbance and PMcoarse concentrations between and within 20 European study areas and the relationship with NO2—Results of the ESCAPE project. Atmos Environ 2012; 62303–317
64. SAS Institute Inc.. 2016. Cary, NC: SAS Institute Inc
65. R version 3.1.3. 2015. Available at: http://cran.rproject.org
66. Gidhagen L, Omstedt G, Ran Pershagen G, Willers S, Bellander T. High-resolution modeling of residential outdoor particulate levels in Sweden. J Expo Sci Environ Epidemiol 2013; 23306–314
67. EN 2014. EN 12341:2014Ambient air - Standard gravimetric measurement method for the determination of the PM10 or PM2,5 mass concentration of suspended particulate matter.2014. BelgiumCEN, European Committee for Standardization
68. Berkowicz R, Ketzel M, Jensen SS, Hvidberg M, Raaschou-Nielsen O. Evaluation and application of OSPM for traffic pollution assessment for a large number of street locations. Environ Model Softw 2008; 23296–303
69. Yanosky JD, MacIntosh DL. A comparison of four gravimetric fine particle sampling methods. J Air Waste Manag Assoc 2001; 51878–884
70. Olstrup H, Johansson C, Forsberg B. The use of carbonaceous particle exposure metrics in health impact calculations. Int J Environ Res Public Health 2016; 13249
71. Cyrys J, Heinrich J, Hoek G, Meliefste K, Lewné M, Gehring U, et al. Comparison between different traffic-related particle indicators: elemental carbon (EC), PM2.5 mass, and absorbance. J Expo Anal Environ Epidemiol 2003; 13134–143
72. Putaud J-P, Raes F, Van Dingenen R, Brüggemann E, Facchini MC, Decesari S, et al. A European aerosol phenomenology—2: chemical characteristics of particulate matter at kerbside, urban, rural and background sites in Europe. Atmospheric Environment 2004; 382579
73. Zeger SL, Thomas D, Dominici F, Samet JM, Schwartz J, Dockery D, et al. Exposure measurement error in time-series studies of air pollution: concepts and consequences. Environ Health Perspect 2000; 108419–426
74. Sheppard L, Burnett RT, Szpiro AA, Kim S-Y, Jerrett M, Pope CA, et al. Confounding and exposure measurement error in air pollution epidemiology. Air Qual Atmos Health 2012; 5203–216
75. Kim S, Sheppard L, Kim H. Health effects of long-term air pollution: influence of exposure prediction methods. Epidemiology 2009; 20442–450
76. Bergen S, Sheppard L, Sampson PD, Kim S-Y, Richards M, Vedal S, et al. A national prediction model for PM2.5 component exposures and measurement error-corrected health effect inference. Environ Health Perspect 2013; 1211017–1025
77. Kaufman JD, Spalt EW, Curl CL, Hajat A, Jones MR, Kim S-Y, et al. Advances in Understanding Air Pollution and CVD. Glob Heart 2016; 11343–352
Keywords:

Particulate matter; Exposure modeling; Epidemiology

Supplemental Digital Content

Back to Top | Article Outline
Copyright © 2018 The Authors. Published by Wolters Kluwer Health, Inc.