- Email: [email protected]
Representativeness of air quality monitoring networks
Atmospheric Environment 104 (2015) 88e101
Contents lists available at ScienceDirect
Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv
Representativeness of air quality monitoring networks Jan Duyzer*, Dick van den Hout, Peter Zandveld, Sjoerd van Ratingen TNO Urban Environment and Safety, P.O. Box 80015, 3508 TA, Utrecht, The Netherlands
h i g h l i g h t s Air quality networks in cities show important difference in design. Differences in design of air quality networks may lead to exposure assessments that are hard to compare. Differences in design of air quality networks may lead to unbalanced checking of compliance. Models are needed to support data evaluation network evaluation.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 24 July 2014 Received in revised form 24 December 2014 Accepted 31 December 2014 Available online 31 December 2014
The suitability of European networks to check compliance with air quality standards and to assess exposure of the population was investigated. An air quality model (URBIS) was applied to estimate and compare the spatial distribution of the concentration of nitrogen dioxide (NO2) in ambient air in four large cities. The concentrations calculated at the location of the monitoring stations, compared well with the concentrations measured at the stations indicating that the models worked well. Therefore the calculated concentration distributions were used as a proxy for the actual concentration distributions across the cities. The distributions of these proxy concentrations across the city populations was determined and cumulative population distribution curves were estimated. The calculated annual mean values at the monitoring network stations were located on the population distribution curves to estimate the fractions of the populations that the monitoring network stations represent. This macro scale procedure is used to evaluate which subgroups of the monitoring stations can be reliably used to decide on compliance or to estimate the concentration the population is exposed to. In addition, the CAR model and Computational Fluid Dynamics (CFD) models are used to investigate the effect of micro scale siting of the monitoring stations within the streets. The following observations were made: - Berlin and London networks cover the distribution of concentrations to which the population is exposed rather well, while Stuttgart and Barcelona have stations at sites with mainly the higher concentrations and the exposure is covered less well. - The networks in London and Berlin, with a substantial number of urban background stations, seem fit to monitor the average population exposure, contrary to those in Stuttgart and Barcelona with only a limited number of these stations. - The concentrations measured at street stations hardly reflect the calculated differences in street pollution between the cities. In Stuttgart the stations are, in line with the EU directive, placed in the most polluted streets, while in other cities there are no stations in the streets with the highest pollution levels. - The concentrations measured at street stations e particularly where buildings inhibit ventilation e are very sensitive to the exact location within the street. Different siting choices may have an effect that for NO2 could reach up to 10 mg/m3 in realistic conditions. Street stations, representing only a small urban area, are not suitable for characterising the exposure of the general population. It is important to note that epidemiological studies whether investigating short term-effects or those studying long-term effects are potentially affected by the issues raised in the paper. Long-term cumulative exposure estimates that are based rather uncritically on monitoring data may be biased if the stations are not representative. It is recommended to use models to support the interpretation and spatial extrapolation of the results of measurements in existing networks. The use of models also relaxes the need for station relocation in inadequate networks, which often would compromise trend analysis. It
Keywords: Air pollution Monitoring Networks Nitrogen dioxide Exposure
* Corresponding author. E-mail address: [email protected] (J. Duyzer). http://dx.doi.org/10.1016/j.atmosenv.2014.12.067 1352-2310/© 2015 Elsevier Ltd. All rights reserved.
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
89
also relaxes the importance of exact or detailed, comprehensive, station classifications since all stations can be used in exposure assessments. © 2015 Elsevier Ltd. All rights reserved.
1. Introduction 1.1. Background In the course of time EU legislation was developed to monitor air quality throughout the member states. Data quality objectives in this legislation have set requirements for the quality of the measurements and the data coverage in time. These provisions together with legal requirements on station density and siting, have helped to limit the substantial differences that originally existed between the networks in the member states. By now, a generally accepted practice of standardising monitoring equipment including protocols for quality assurance and quality control have become available. There are however still important differences in the spatial coverage of the various networks. In a recent survey of views of member states and other stakeholders on the revision of the EU Ambient Air Quality Directive (AAQD1), further harmonisation of the networks, in particular regarding station siting (and representativeness), was high on the priority list (van den Hout et al., 2012). The foremost purpose of the monitoring networks operating in the states is to assess compliance with the air quality standards of the EU AAQD. The second important purpose is to assess exposure of people, addressing both the highest levels and the levels in other areas where the general population is exposed. In this study we focus on these two objectives, i.e. determination of compliance and exposure assessment and the adequacy of current networks regarding these two objectives. In view of these discussions representativeness could indicate spatial representativeness (how large is the area represented by the measurements at a certain station). This is a very difficult issue. On the other hand it could also indicate how well measurements at a certain station represent the exposure of the general population. Partly this indication overlaps with the previous one. But it is perhaps easier to address and the focus of this study. 1.2. Exposure and network design The actual exposure of a person is determined by the concentration in the breathing zone. It should be noted that the total exposure of (European) individuals to air pollutants is more related to indoor air than ambient air monitored at stations. This is illustrated in Fig. 1 showing an example of how people spend their time (Dons et al., 2011). Exposure as mentioned in the AAQD however relates exclusively to ambient outdoor air.2 In Annex III of the AAQD on siting requirements, exposure it is not further specified; only in the definition of the Average Exposure Indicator some specification is given,
1
The Lancet, Volume 353, Issue 9156, Pages 874e878, 13 March 1999. It is relevant to note here that time activity patterns shown in Fig. 1 may lead to different assessments. In the study by Dons et al., 2011 it is shown that drivers may be exposed to rather significant levels of black carbon (a component emitted by cars) during the period they drive to their work. Exposure at home is much less due to the much lower levels there. This issue, although interesting, is not addressed here at this stage and we follow the directive's approach where exposure relates to ambient outdoor air quality. 2
Fig. 1. An example of the fraction of time that people in Belgium spend on certain activities. Derived from Dons et al. (2011).
relating the average exposure to urban background sites (Article 2#). Although both outdoor and indoor air (including outdoor air penetrated indoors) are important contributors to the actual exposure, the concentration-response functions established in epidemiological studies, on which the air quality standards are based, relate only to outdoor concentrations e usually those measured at urban background locations (Hoek et al., 2002; Boezen et al., 1999). In the study here we assume exposure is directly linked to the home address. This is the place where people spend most of their time as illustrated in Fig. 1. Exposure as described in the AAQD Annex IIIB is derived from measurements at stations, which are always limited in number. Obviously most people do not actually live in the direct vicinity of monitoring stations. In (urban) background conditions concentration gradients are usually small, so measurements at these stations tend to be quite representative for larger areas and are consequently suitable to assess exposure. As outlined above, the foremost aims in air quality monitoring are to assess compliance with air quality standards and population exposure, including trends therein. A network should address both issues. We will study the networks of four large European cities bearing these goals in mind. Two scales are distinguished here: Macro scale siting of a station is the selection of the type of site (urban background, near streets etc.) and the approximate location. In Annex III to the AAQD provisions on macro siting are given: “Sampling points directed at the protection of human health shall be sited in such a way as to provide data on the following:” - the areas within zones and agglomerations where the highest concentrations occur to which the population is likely to be directly or indirectly exposed for a period which is significant in relation to the averaging period of the limit value(s), - levels in other areas within the zones and agglomerations which are representative of the exposure of the general population”3
3 Article 2 of the AAQD: ‘urban background locations’ shall mean places in urban areas where levels are representative of the exposure of the general urban population.
90
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
The micro-scale siting of the station is establishing its exact position with respect to sources and obstacles in the direct vicinity. It is important to note that epidemiological studies whether investigating short term-effects or those studying long-term effects are potentially affected by the issues raised in the paper. Exposure estimates as used in epidemiological studies depend on appropriate time-dependent spatial concentration estimates that can be allocated to the exposed population under study. Thus, also long-term cumulative exposure estimates that are based rather uncritically on monitoring data may be biased if the stations are not representative. 1.3. Other studies Much of the work attributed to air pollution networks does not end up in the peer-reviewed scientific literature but is laid down in (sometimes comprehensive) reports by authorities responsible for air quality or other institutions. Very detailed analyses and interpolation schemes were already developed in the 1990's by German scientists looking into the representativeness of the monitoring network in North Rhine-Westphalia (Beier and Doppelfeld, 1992). More recently Spangl et al. (2007) developed definitions, methods and validation procedures for the classification of air quality monitoring stations. The issue of representativeness is discussed and proposals for station classification and for new concepts of representativeness have been developed and elaborated. The EURAQHEM report (Kuhlbusch et al., 2004) describes an analysis of the EU air quality monitoring network with regard to health assessments. It evaluates the limitations of the current network and makes recommendations for improvement. The study concludes that Member States provide insufficient information on representativeness of measurement sites in relation to populations, that an approach of default areas of representativeness per station site type is inadequate, and that no direct relation of air quality assessment and the exposure of the population could be generated. The SCREAM4 working group of AQUILA (Network of National Air Quality Reference Laboratories) has been developing recommendations on the representativeness of monitoring stations. 2. Methods 2.1. Methodology and models used 2.1.1. Methodology It is difficult to assess how representative a network is for a distribution of concentrations or exposures in a certain area using monitoring data only. There are a limited number of monitoring stations and the true spatial distribution of concentrations across the entire area is not known. Whereas to assess the representativeness of the network, concentration estimates in areas without monitoring stations are also needed. Therefore we took an approach based upon detailed modelling of concentration patterns in cities to study representativeness. Four large European cities (Barcelona, Berlin, London and Stuttgart) are used as examples. The following procedure was followed: An air quality model (URBIS) was applied to calculate the spatial distribution of the concentration of NO2 in ambient air in four cities (resolution: 10 m 10 m). The concentrations calculated at the location of the monitoring stations, compared well with the concentrations measured at the stations indicating that the models worked well. Therefore the calculated concentration distributions were used as a proxy for the actual concentration distributions across the cities.
4 Siting criteria, classification and representativeness of air quality monitoring stations” (SCREAM).
These estimates were used as proxy for the true exposure. The distributions of these proxy concentrations across the city populations were determined and cumulative population distribution curves were estimated. The annual mean calculated concentration values of the monitoring network stations were located on the calculated population distribution curves to estimate the fractions of the populations that the monitoring network stations represent. This macro scale procedure is used to evaluate which subgroups of the monitoring stations (if any) can reliably be used to decide on compliance or to estimate the concentration the population is exposed to. The CAR model is used to investigate the effect of micro scale siting of the monitoring stations within the streets. 2.1.2. Model used: the TNO URBIS model For each of the four cities usually detailed models are available and published by local authorities. In the approach used here, one single model was chosen to calculate the concentration level for all cities in a comparable way and with the same, high, resolution. The high resolution also allows detailed evaluation of concentrations observed at particular monitoring stations which is the purpose of the current study. It is relevant to realize that a higher resolution of the calculations leads to a larger range of concentrations i.e. there are lower and higher concentrations observed in maps based upon high resolution data (such as the 10 m used here) than there are in maps with a lower resolution (See for example for Stuttgart www. lubw.baden-wuerttemberg.de/servlet/is/242644/) Concentrations calculated using a high resolution on sites and monitoring stations near roads with intensive traffic are much higher than those calculated with a resolution of 0.5 or 1 km. The TNO-URBIS model (Beelen et al., 2010) was used to create a concentration maps across the cities. URBIS is an air pollution model designed for cities and regions with a variety of emission sources. Emission rates from all known emission sources in the area of study are input to the model. The dispersion models implemented in URBIS are: the CAR model, a dispersion model for emissions by traffic in streets for distances up to 30 m from the road (Eerens et al., 1993; Wesseling and Sauter, 2007); Motorway Plume model5 model, is a dispersion model for pollution by motorway traffic. This is a Gaussian line source model for dispersion up to 5000 m (van den Hout and Baars 1988; Wesseling and Visser, 2003); and A simplified version of the Dutch standard calculation method for point sources (van Ham et al., 1998). This is a Gaussian plume dispersion model. In this study, concentrations at a 100 100 m grid were calculated. Input to the dispersion models is the emission rate of nitrogen oxides (NOx; the sum of nitric oxide (NO) and nitrogen dioxide (NO2) emissions). The dispersion of NOx through the area is calculated by the dispersion models. The local NO2 concentration is then derived from the calculated NOx concentration using an empirical algorithm. This algorithm uses the local ozone concentrations available for reaction with NO and is based upon long series of measurements of NOx and NO2 and ozone in the air pollution monitoring network in the studied city. 2.1.2.1. Calculations. The annual average of the concentration of nitrogen dioxide in 2010 was calculated for the four cities. The data used to run the models is described in detail in Duyzer et al. (2013).
5
Translated from Dutch: The accredited TNO Pluim-Snelweg model.
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
91
Each calculation is carried out in 4 layers: 1. Regional background: a single number derived from measurements or LOTOS-EUROS calculation. 2. City background: calculation with a Gaussian plume model. The calculation has been carried out at a high resolution for the emissions of all municipal roads. The results of the calculations are averaged to a 1 1 km2 grid. 3. Motorways: calculation with a Gaussian plume model on a 10 10 m2 grid 4. City streets: calculation with the CAR model for street canyons. All contributions are added to calculate the concentration at each grid point. Special attention has been given to the location of the monitoring stations. The CAR model is very sensitive (and so are the concentrations in real life) to the distance between the road and the receptor. At this stage concentrations are calculated every 10 m.6 The locations have been inspected in Google street view, to make sure the distance between the road and the measuring site is correctly estimated. The final accuracy of the coordinates of the monitoring sites in our model calculations is now estimated to be around 1 m. 3. Results 3.1. Results of calculations with the Urbis model Fig. 2, Fig. 3, Fig. 4, and Fig. 5 show the, annually averaged, concentrations of nitrogen dioxide in the four cities. The locations of the monitoring sites are indicated with dots. As expected high concentrations are found near major roads. For Barcelona and London, the calculated concentrations are higher than for Berlin and Stuttgart. In order to optimize the contrast for each of the figures different colour scales (in webversion) have been used. To assess the quality of the calculations, Fig. 6, Fig. 7, Fig. 8 and Fig. 9 show comparisons between measured and calculated concentrations for the year 2010. In general the comparison is good. The correlation coefficients (r2) deviate less than 10% from a value of one. This means that only 10% of the variation in the concentrations measured at the stations is not explained and 90% is explained by local emissions and other parameters whose influence on the concentration is described in these models. Only in Stuttgart the difference is larger. Here the measured concentrations are 10e20 % higher than the calculated ones. This could partly be due to the specific orographic situation resulting in specific meteorological conditions in Stuttgart with for example low wind speeds. These conditions are possibly not correctly accounted for in the URBIS model. Since in the sequel only modelled values are used in the comparisons this (rather small) discrepancy to measured data in Stuttgart has no impact on the outcome of this study on the representativeness of monitoring sites. The good agreement leads us to conclude that the model calculations are suitable for our purpose to study the appropriateness of the siting's of network stations. The calculated concentrations across the cities are used in the rest of this paper as a proxy for real concentrations in these cities. 3.2. Exposure of the general public: macro scale siting 3.2.1. Exposure curves The exposure curves used in this paper present the percentage
6 This high resolution is needed to be able to calculate the concentration levels in streets. Lower resolutions of for example 1 km2 lead to lower concentrations due to spatial averaging of high and low concentrations.
Fig. 2. The calculated, annual averaged, concentration of NO2 calculated for Barcelona for the year 2010. Dots indicate monitoring sites.
of the population exposed to concentrations higher than the values indicated on the y-axis. They are constructed based on modelled concentrations and census data. As was discussed above we use home addresses (from census data, see Annex) as the place where people are exposed. Fig. 10 shows the exposure curves calculated for the four cities in one plot. The curves have a similar shape and all show that many people are exposed to relatively low concentrations and few are exposed to the high concentration ranges. If we take Berlin as an example the curve displays the following going from left to right: - 80% of the population is exposed to levels higher than 15 mg/m3 whereas only 10% of the people is exposed to levels higher than 24 mg/m3 So 70% of the population is exposed to levels between 15 and 24 mg/m3 - Only a very small group, 0.4% of the population, is exposed to levels higher than 40 mg/m3 The figure indicates that, based on our calculations, the general population in Barcelona experiences the highest exposure, followed by London, Stuttgart and Berlin. The median values (50 percentiles), the levels to which 50% of the people are exposed, support this feeling probably best. The median values ranges from around 16 mg/m3 for Berlin, to 23 for Stuttgart, to 34 for London and to 38 mg/m3 for Barcelona.
92
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
Fig. 3. The calculated, annual averaged, concentration of NO2 calculated for Berlin for the year 2010. Dots indicate monitoring sites.
Fig. 4. The calculated, annual averaged, concentration of NO2 calculated for London for the year 2010. Dots indicate monitoring sites.
Fig. 11, Fig. 12, Fig. 13 and Fig. 14 show the same exposure curves extended with the concentrations calculated at the monitoring sites. Note that, to guarantee consistence, modelled (calculated) concentrations are used rather than concentrations measured at the sites. As a result the figures are more or less independent of the quality of the model calculations. In these figures the stations labelled as urban background stations and the traffic stations are indicated. As can be expected, the traffic stations typically represent the higher concentrations. Only in Barcelona some street stations exhibit lower concentrations than background stations. Here the classifications could be flawed. These figures show that the monitoring sites chosen in the cities result in a different coverage of the exposure curves. We judge the distribution of monitoring stations along the x-axis (representing the concentration range) as well as the y-axis (reflecting the population exposure).
Fig. 5. The calculated, annual averaged, concentration of NO2 calculated for Stuttgart for the year 2010. Dots indicate monitoring sites.
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
93
Fig. 6. Annually averaged concentration of NO2 calculated for 2010 using the URBIS model compared to concentration measured at AIRBASE71 stations in Barcelona. The correlation coefficient (squared) and the best fitted (least square method) straight line through the origin is also displayed.
The distribution of sites in Barcelona does not cover either of the two axes well and hence the exposure of the general population nor the occurrence of the highest concentrations is well represented. The stations in Berlin are well distributed across the y-axis, representing the exposure of the general population. The sites cover the x-axis less well, perhaps missing the locations where the highest concentrations are calculated. It should be noted however that only 0.02% of the population lives in areas where the concentrations is higher than 60 mg/m3 (more on this below) In London the sites are fairly evenly distributed along the exposure graphs. The stations are well distributed across the y-
axis, thus representing the exposure of the general population and also across the x-axis, giving a good picture of the highest concentrations. The sites in Stuttgart hardly cover the y-axis and hence the exposure of the general population is not well represented. The x-axis is better covered, but the sites are at locations where high concentrations can be expected. The average urban background calculated from the urban background stations, as well as the population averaged exposure, depends significantly on the siting of the stations that cover this area, particularly when their number is small. Clearly the networks
Fig. 7. See caption Fig. 6 but for Berlin.
94
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
Fig. 8. See caption Fig. 6 but for London.
Fig. 9. See caption Fig. 6 but for Stuttgart.
differ significantly in the extent to which the highest levels are represented and also in the coverage of the levels to which the general population is exposed. 3.2.2. Population weighted concentration The population weighted concentration8 (or Average Exposure
7
http://acm.eionet.europa.eu/databases/airbase/. From the AAQD The Average Exposure Indicator expressed in mg/m3 (AEI) is based upon measurements in urban background locations in zones and agglomerations throughout the territory of a Member State. It should be assessed as a three-calendar year running annual mean concentration averaged over all sampling points established pursuant to Section B of Annex V. The AEI for the reference year 2010 shall be the mean concentration of the years 2008, 2009 and 2010. At this stage it is only mentioned for PM2.5. But it's value may be derived for NO2 as well. 8
Indicator) calculated from the modelled concentration distribution is an interesting measure to compare the exposure in different cities. In Duyzer et al. (2013) it is shown that this parameter compares best with the average concentration measured at the urban background sites. If the results of monitoring at the street sites are included in the average the results differ significantly.
3.2.3. Assessing compliance with limit values in a comparable way Selecting a site with the aim to assess compliance with the limit values and at the same time representative for high exposures is not a simple task. It is interesting to evaluate the locations of the traffic stations designed for this purpose. Such sites are usually selected in busy streets where concentrations are expected to be high. Although such sites aim to fulfil the criteria for assessments of compliance
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
95
Fig. 10. Exposure concentration curves for all four cities. The population weighted average concentration calculated for Barcelona, Berlin, London and Stuttgart is 43.2, 18,7, 37.2 and 26,4 mg/m3 respectively.
Fig. 11. Exposure concentration curves for Barcelona. The symbols indicate the annual average concentrations calculated at the monitoring sites. Urban background stations are distinguished from street stations.
with the limit values, it is often not obvious how well measurements at these stations represent the exposure of the population. 3.2.3.1. How many people are exposed to concentrations observed at the street stations?. From the viewpoint of comparability it is useful to look at the position of these stations on the exposure concentration curves. For each of the four cities, the highest concentration calculated at the monitoring sites is given in Table 1. Also the position this site takes on the exposure curve is presented. Large differences are found. It is interesting to look for the highest concentration calculated for the location of one of the monitoring stations and extract the number of people exposed to such levels from these curves. In Stuttgart only 0.02% of the population is
exposed to concentrations higher than 85 mg/m3 whereas in Barcelona 3% of the population is exposed to concentrations higher than calculated at the monitoring station with the highest concentration of 66 mg/m3. The difference between 0,02% and 3% of the people exposed is especially important if one the associated concentrations. If in Barcelona, London and Berlin the traffic sites would have been located at a site with a concentration to which 0,02% of the population is exposed the mentioned concentration level would have been much higher (like it is now in Stuttgart). Fig. 15 shows that if the sites with the highest concentrations in London, Berlin and Barcelona would also lie at the 0,02% level, the concentration would be higher by 65, 11 and 26 mg/m3 respectively! And the conclusions
96
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
Fig. 12. As Fig. 11 for Berlin.
Fig. 13. As Fig. 11 for London.
on the highest concentrations to which people are exposed would change dramatically. The concentrations would even be higher than those observed in Stuttgart. The example given illustrates the issue that concentrations observed at these traffic stations have little value in comparing exposure to high concentrations in different cities. It also indicates that the network of Stuttgart complies best with the AAQD's provision that the highest concentrations to which the population is exposed should be assessed. The representativeness of street stations for other streets is difficult to judge by itself. It seems that the street stations in the cities studies here represent exposure between 0,5% of the population (Berlin) and 15% of the population (Barcelona). In that respect it seems that these stations serve their purpose in that they represent the exposure in busy streets.
3.3. Micro scale siting The micro scale siting of the stations is described in the AAQD (Annex III). Whereas the location of inlet tubes, distance to walls etc. is described in some detail in the AAQD, not all important parameters, such as the exact location in the area, are treated very extensively. It is clear that urban background stations should not be located in busy roads. For street stations aiming at higher concentrations the traffic intensity is on the other hand crucial. The contribution of local traffic to the air concentration is coupled with local emissions and therefore related to the traffic intensity (and composition etc.). So to find locations for hotspot sites, streets with a high traffic intensity will need attention. Here several factors are relevant and determining the outcome of measurements.
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
97
Fig. 14. As Fig. 11 for Stuttgart.
Table 1 Highest concentration calculated on a monitoring site and the location on the exposure-concentration curve (expressed as percentage of the population that is exposed to concentration higher than this concentration). City
Highest concentration (mg/m3)
Percentage (%)
Barcelona Berlin London Stuttgart
66 50 98 85
3 0,1 0,81 0,02
Fig. 15. Concentrations calculated at the site with the highest concentrations compared to the concentration to which less than of the 0.02% of the population is exposed.
The effect of some of these factors may easily be described using models such as the CAR model described above. This is illustrated in Fig. 16 where the contribution of the emission by road traffic to the concentration of nitrogen dioxide in air, is shown as a function of the distance to the road, the street type and congestion. In this case a typical busy road with 80,000 vehicles per day was used. This graph illustrates in the first place how important the distance of the
sample inlet to the road is and secondly how important, for example, street type is. This means that a concentration measured in one city in a narrow street canyon needs to be judged differently from a concentration measured in another city with similar traffic intensity in a broad street. Table 2 summarizes the results indicating the strong influence of distance to the centre of the road, the difference between a narrow and a wide street and the impact of congestion. In the presented example, whether a monitoring station is 5 or 20 m away from the roadside makes a difference of approximately 16 mg/m3 in the calculated concentration. This indicates how important the distance to the road actually is.9 The street type, describing the type of buildings along the road, may have a large impact as well. Differences in concentrations between two types of streets may be as large as 10 mg/m3 or more. Congestion may have a similarly large impact, in this case of more than 10 mg/m3 All of these calculations show the importance of (knowledge of) the exact location of a monitoring site in a street and the kind of buildings along the road. The differences are large with respect to exceedances of limit values. Measures taken to improve air quality such as a less dense traffic or a reduction of emission factors may lead to improvements in the air concentration. The effect of another location of a station (a site farther away from the road) or a different the street type may be much larger. This makes comparison of monitoring results among different cities more difficult as is shown in Fig. 17 which illustrates different interpretations of the selection criteria for monitoring. In Stuttgart some monitoring stations are within few meters from the first lane whereas in Barcelona the distance of some stations seems larger and even more than 10 m from the first lane with parked cars. Depending on the type of street canyon, differences could be as high as 14 mg/m3 between a station at 10 or one at 20 m away from the centre of the road. This statement is of course only valid for the given example of 80,000
9 For all pollutants, traffic-orientated sampling probes shall be at least 25 m from the edge of major junctions and no more than 10 m from the kerbside (from the AAQD ANNEX III C).
98
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
Fig. 16. The effect of street type and congestion on the contribution of a road to the concentration observed at certain distances from the centre of the road. Concentrations calculated for a distance less then 5 meters are often above the road lane and therefore only here as an indication. The emission source is located in the middle of the road. Traffic intensity of nearly 80,000 vehicles per day with 4% trucks. 10% direct emission of NO2 and the concentration of ozone in the area is 40 ppb.
vehicles per day in this for lane road. Less is known about the impact of the position along the street. In the AAQD it is stated that a station should be representative of at least 100 m along the road. Models such as CAR can be used to investigate the impact of several parameters. It is more difficult to judge the effect of complex, infrastructural conditions. Here, much more complex models are needed. Using Computational Fluid Dynamics (CFD) based models calculations in these complex situations can be made. This would allow estimations of the impact of large infrastructures (buildings, bridges, tunnel exits) on concentrations and measurements thereof on a specific site. The absolute accuracy in the results of CFD calculations may be limited. Still in the absence of detailed local scale measurements the calculations could be helpful to judge the quality of a monitoring station on a local scale. Qualitatively they illustrate where problems might occur or might have occurred. In Duyzer et al. (2013) the siting of a monitoring station (Am Neckartor) in Stuttgart was judged. The calculations showed that in view of some of the definitions in the AAQD (the area where the concentration varies by less than 10%) the representativeness of street stations is only small. Perpendicular to the road the concentration varies by more than 10% within 10 m and in the direction along the road the concentration varies within areas of scale 100 m as well.
4. Discussions and conclusions 4.1. Conclusions We have used model calculations to evaluate the positioning of air pollution monitoring stations in the framework of the AAQD. Basis for our analysis were the calculated concentration fields in four major European cities for nitrogen dioxide (NO2), a pollutant for which limit values exist. The exposure of inhabitants to nitrogen dioxide was calculated and the representativeness of the monitoring network was judged. A distinction was made between
macro-scale and micro-scale siting. Based on the analysis of the concentrations in four cities the following conclusions were drawn: Macro-scale siting. In three cities the exposure of the general population can rather well be assessed using measurements at urban background stations. Averages of concentrations calculated at urban background stations show good agreement with the exposure weighted concentration for a city. Individual urban background stations may however not be representative of the average exposure, as the case of Stuttgart shows. In the smallest of the studied cities there is only one background station. Clearly the uncertainty of only one sample is large. Traffic or hotspot stations are aimed at assessing compliance with the limit values and aim to measure the highest concentrations. The following is noted: The results of measurements at traffic stations are not useful to assess exposure of the general population. They often indicate only the exposure of a (sometimes very) small fraction of the population. The model calculations suggest that in one city (Stuttgart) only 0.02% of the population is exposed to concentrations higher than the concentrations at the most polluted station, while in another city (Barcelona) this is valid for the exposure of 3% of the population. If all cities would have a station at the 0.02% exposure level, like in Stuttgart, the highest observed concentration in those cities would increase by, depending on the city, 10, 20 and even 60 mg/m3 compared to the currently measured highest level. It is therefore difficult to draw conclusions on high exposure when comparing the highest observed concentration in different cities. These depend too much on the location of the monitoring site. Hence the highest concentration at for example the 3% level in Barcelona does not represent the highest level of concentrations as required by the AAQD, in spite of the relatively high proportion of hotspot stations in this city. According to
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101 Table 2 Calculated contribution of emissions of a road to the concentration of NO2 at various distances from the centre of the road. Distance to centre of road C(NO2) in (where emission takes wide street place) (m) canyon
C(NO2) in C(NO2) in narrow narrow street street canyon with canyon congestion
5 10 20
44 36 22
30 24 14
44 35 20
99
CFD calculations may be used to qualitatively judge the quality of monitoring sites on a local scale in more complex situations. Examples indicate that in the direction along the road the concentration may vary within distances of the scale of 100 m. Because of the unclear representativeness for concentrations in the direct surroundings, it is difficult to use differences between concentrations measured at hotspot stations to assess the differences in population exposure between cities.
Fig. 17. Some street stations in Stuttgart (above) and Barcelona (below). This illustrates the differences in the distances between the station and the road.
our calculations sites may be found in this city with concentration levels several tens of mg/m3 higher. In the interpretation and especially the comparison of results of measurements at traffic stations, models should play a role. The results of model calculations can be used to judge the position of stations on the concentration exposure curves. Without such analysis comparison is very difficult, if not useless at all. Micro-scale siting. Micro-scale siting of street stations is very important as well. Especially the distance to roads is a major parameter; positioning a station closer to the road can lead to substantially higher concentrations. Other important parameters are: the street configuration (width of the road, buildings on one or two sides, height of the buildings), traffic flow (free flow or congestion).
It is relevant to note that the analysis here is made for nitrogen dioxide. This compound is emitted by cars but is also formed locally as a result of the chemical reactions in air between nitric oxide (also emitted by cars but in larger quantities) and ozone. This chemical reaction is relatively fast but slower than the dispersion through the air. As a result the gradient in concentration of nitrogen dioxide downwind of sources is smaller than it is for inert species such as soot or carbon monoxide. Hence the area of representativeness for these compounds is expected to be smaller than for nitrogen dioxide. 4.2. Possible consequences for siting of monitoring stations The most relevant shortcomings that we observed pertained not to individual stations but to the representativeness of the station networks. The four urban monitoring networks that we evaluated and compared do not give a similar representation of the spatial distribution of the concentrations in the cities concerned. The
100
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
Stuttgart network has a strong focus on the highest concentrations in the city, in conformity with the objective to represent the area with the highest concentrations/exposures. The three other networks did not mainly focus on the locations with the highest concentrations. A key question is which of the approaches fulfils the requirements of the directive best. Clearly the Stuttgart network is the most adequate in identifying the highest concentrations in the ambient city air, as also required by the directive. This particular network is however less suited to assess exposure of the general population because of the limited number of background sites. Street stations can of course play a role in assessing the exposure in streets. The results of monitoring here could be coupled with the time spent at such a site. Coupled with the applicable time-activity patterns monitoring at street sites is useful. Using the results to calculate the simple average of exposure for a city may lead to biased results. Station siting issues have been subject to continual debate and fully satisfactory solutions have not been found and may in fact not be existent at all. The need to find a compromise between practicable solutions and requirements has offered network designers a substantial freedom in locating the stations. There is a need for more harmonisation in network design. At the same time it may not be possible or even needed to change current networks and break up important time series. A very important outcome of monitoring is the possibility to analyse trends in concentrations and evaluate for example the effect of (long term) measures aimed at reduction of exposure. Therefore it is recommended to not change current networks drastically but rather harmonise the interpretation of the results. Whereas the results of measurements at urban background stations, if located at representative sites, can usually be used directly to assess the exposure of the general public, the use of concentrations observed at street stations is much more complex. It seems impossible to interpret these data in a harmonised way without using simulation models such as the models used in this study. Model calculations of the concentration distribution in the street network of a city, taking parameters such as traffic intensity, distance to the road, street type into account, can provide insight in how well street stations cover the distribution of the highest concentrations in the city. After more research such model calculations may even be used to compensate for the influence of many important, but well-known, parameters such as distance to the road. This compensation could make it possible to realistically compare the results of concentration measurements in different cities. Models can also be used to estimate the distribution of the exposure of people to concentrations of pollutants. This would allow fair comparison of the exposure between cities. Model calculations would also relief the emphasis on station classifications some off which seem to be flawed at the moment. Apart from stations that are obviously on a wrong location all stations in a network that has a significant spatial coverage can be used and have their value. The cause of higher concentrations at some stations can simply be interpreted since all available knowledge may be incorporated in the model. Acknowledgement The study is funded by EUGT (the Europ€ aische Forschungsvereinigung für Umwelt and Gesundheit im Transportsektor www.eugt.org). Annex. details of input to model calculations For the four cities the following information was used. For all cities the following information was used:
Emission factors: TREMOVE/COPERT, http://www.tremove.org/, http://lat.eng.auth.gr/copert/. Urban background: calculated with Gaussian plume model Regional background: measurement Census data Census data were obtained from the European Union data bases http://www.eea.europa.eu/data-and-maps/data/populationdensity-disaggregated-with-corine-land-cover-2000-2. The follow city specific data were used: Stuttgart Roads (map þ traffic intensities): Lohmeyer, project 60385-050: The basic input data (street types, traffic intensity, meteo€chlin, (2006) rology) are based upon Nagel and Ba The Stuttgart data contained detailed information of the configuration of the street (‘Schluchttype’). This has been converted to the street types of the TNO CAR model. Meteorological data: Lohmeyer, project 60385-05-0. Ten different areas for meteorological conditions have been used in this study.
Berlin Roads (map þ traffic intensities): Email communication. Senatsverwaltung für Gesundheit, Umwelt und Verbraucherschutz - III D 24 Brückenstr. 6, 10173 Berlin The Berlin data contained information about the street type: street canyon or open. Meteorological data: interpolated data from LOTOS-EUROS (Schaap et al., 2008)
London Roads (map þ traffic intensities): London Atmospheric Emission Inventory (LAEI, http://data.london.gov.uk/datastore/package/ london-atmospheric-emissions-inventory-2010 Meteorological data: interpolated data from LOTOS-EUROS
Barcelona Roads (map þ traffic intensities) data collected in INTARESE project (http://intarese.org) Meteorological data: interpolated data from LOTOS-EUROS References Beelen, R., Voogt, M., Duyzer, J., Zandveld, P., Hoek, G., 2010. Comparison of the performances of land use regression modelling and dispersion modelling in estimating small-scale variations in long-term air pollution concentrations in a Dutch urban area. Atmos. Environ. 44 (36), 4614e4621. Beier, Doppelfeld, 1992. LIS-Berichte; Nr. 101 R€ aumliche Übertragbarkeit und €tsdaten im Messnetz TEMES. http://www.zvab. Interpolation von Luftqualita com/LIS-Berichte-R%C3%A4umliche-%C3%9Cbertragbarkeit-InterpolationLuftqualit%C3%A4tsdaten-Messnetz/165294139/buch. Boezen, M., van der Zee, S., Postma, D.S., Vonk, J.M., Gerritsen, J., Hoek, G., Brunekreef, B., Rijcken, B., Schouten, J.P., 1999. Effects of ambient air pollution on upper and lower respiratory symptoms and peak expiratory flow in children. Lancet 353 (9156), 874e878. Dons, E., Int Panis, L., van Poppel, M., Theunis, J., Willems, H., Torfs, R., Wets, G., 2011. Impact of time activity patterns on personal exposure to black carbon. Atmos. Environ. 45, 3594e3602. Duyzer, J., van den Hout, D., Zandveld, P., van Ratingen, S., 2013. Representativeness of Air Quality Monitoring Stations. TNO report 2013 R11055 Available from: EUGT.
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101 Eerens, H., Sliggers, C., van den Hout, K.D., 1993. The CAR model: the Dutch method to determine city air quality. Atmos. Environ. 27B, 389e399. Hoek, G., Brunekreef, B., Goldbohm, S., Fischer, P., van den Brandt, P.A., 2002. Association between mortality and indicators of traffic-related air pollution in the Netherlands: a cohort study. Lancet 360, 1203e1209. Kuhlbusch, Th, John, A., Hugo, A., Peters, A., von Klot, S., Cyrys, J., Wichmann, H.-E., Quass, U., Bruckmann, P., 2004. Analysis and Design of Local Air Quality Measurements. Towards European Air Quality Health Effect Monitoring. Service Contract 070501/2004/389487/MAR/C1, Brussels. Spangl, W., Schneider, J., Moosmann, L., Nagl, C., 2007. Representativeness and Classification of Air Monitoring Stations. UBA, Vienna. van den Hout, Dick, Voogt, Marita, Moosmann, Lorenz, Nagl, Christian,
101
Spangl, Wolfgang, 2012. Survey of Views of Stakeholders, Experts and Citizens on the Review of the EU Air Policy. TNO, Netherlands. TNO report TNO-060-UT2012e00714. Available at: http://ec.europa.eu/environment/air/pdf/Survey% 20AQD%20submitted for publication%20-%20Part%20I%20Main%20results.pdf. van Ham, N.J., Duijm, J., Erbrink, J.J., van Jaarsveld, J.A., Pulles, M.P.J., Schols, E., Verver, G.H.L., 1998. Revision of the Netherlands National Model for short range dispersion of air pollutants. Int. J. Environ. Pollut. 8, 3e6. Wesseling, J., Sauter, F., 2007. Calibration of the Program CAR II Using Measurements of the National Monitoring Network of the RIVM (in Dutch). Bilthoven, RIVM report 680705004.2007. Wesseling, J., Visser, G.Th, 2003. An Inter-comparison of the TNO Traffic Models, Field Data, and Wind Tunnel Measurements. TNO report 2003/207.
Contents lists available at ScienceDirect
Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv
Representativeness of air quality monitoring networks Jan Duyzer*, Dick van den Hout, Peter Zandveld, Sjoerd van Ratingen TNO Urban Environment and Safety, P.O. Box 80015, 3508 TA, Utrecht, The Netherlands
h i g h l i g h t s Air quality networks in cities show important difference in design. Differences in design of air quality networks may lead to exposure assessments that are hard to compare. Differences in design of air quality networks may lead to unbalanced checking of compliance. Models are needed to support data evaluation network evaluation.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 24 July 2014 Received in revised form 24 December 2014 Accepted 31 December 2014 Available online 31 December 2014
The suitability of European networks to check compliance with air quality standards and to assess exposure of the population was investigated. An air quality model (URBIS) was applied to estimate and compare the spatial distribution of the concentration of nitrogen dioxide (NO2) in ambient air in four large cities. The concentrations calculated at the location of the monitoring stations, compared well with the concentrations measured at the stations indicating that the models worked well. Therefore the calculated concentration distributions were used as a proxy for the actual concentration distributions across the cities. The distributions of these proxy concentrations across the city populations was determined and cumulative population distribution curves were estimated. The calculated annual mean values at the monitoring network stations were located on the population distribution curves to estimate the fractions of the populations that the monitoring network stations represent. This macro scale procedure is used to evaluate which subgroups of the monitoring stations can be reliably used to decide on compliance or to estimate the concentration the population is exposed to. In addition, the CAR model and Computational Fluid Dynamics (CFD) models are used to investigate the effect of micro scale siting of the monitoring stations within the streets. The following observations were made: - Berlin and London networks cover the distribution of concentrations to which the population is exposed rather well, while Stuttgart and Barcelona have stations at sites with mainly the higher concentrations and the exposure is covered less well. - The networks in London and Berlin, with a substantial number of urban background stations, seem fit to monitor the average population exposure, contrary to those in Stuttgart and Barcelona with only a limited number of these stations. - The concentrations measured at street stations hardly reflect the calculated differences in street pollution between the cities. In Stuttgart the stations are, in line with the EU directive, placed in the most polluted streets, while in other cities there are no stations in the streets with the highest pollution levels. - The concentrations measured at street stations e particularly where buildings inhibit ventilation e are very sensitive to the exact location within the street. Different siting choices may have an effect that for NO2 could reach up to 10 mg/m3 in realistic conditions. Street stations, representing only a small urban area, are not suitable for characterising the exposure of the general population. It is important to note that epidemiological studies whether investigating short term-effects or those studying long-term effects are potentially affected by the issues raised in the paper. Long-term cumulative exposure estimates that are based rather uncritically on monitoring data may be biased if the stations are not representative. It is recommended to use models to support the interpretation and spatial extrapolation of the results of measurements in existing networks. The use of models also relaxes the need for station relocation in inadequate networks, which often would compromise trend analysis. It
Keywords: Air pollution Monitoring Networks Nitrogen dioxide Exposure
* Corresponding author. E-mail address: [email protected] (J. Duyzer). http://dx.doi.org/10.1016/j.atmosenv.2014.12.067 1352-2310/© 2015 Elsevier Ltd. All rights reserved.
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
89
also relaxes the importance of exact or detailed, comprehensive, station classifications since all stations can be used in exposure assessments. © 2015 Elsevier Ltd. All rights reserved.
1. Introduction 1.1. Background In the course of time EU legislation was developed to monitor air quality throughout the member states. Data quality objectives in this legislation have set requirements for the quality of the measurements and the data coverage in time. These provisions together with legal requirements on station density and siting, have helped to limit the substantial differences that originally existed between the networks in the member states. By now, a generally accepted practice of standardising monitoring equipment including protocols for quality assurance and quality control have become available. There are however still important differences in the spatial coverage of the various networks. In a recent survey of views of member states and other stakeholders on the revision of the EU Ambient Air Quality Directive (AAQD1), further harmonisation of the networks, in particular regarding station siting (and representativeness), was high on the priority list (van den Hout et al., 2012). The foremost purpose of the monitoring networks operating in the states is to assess compliance with the air quality standards of the EU AAQD. The second important purpose is to assess exposure of people, addressing both the highest levels and the levels in other areas where the general population is exposed. In this study we focus on these two objectives, i.e. determination of compliance and exposure assessment and the adequacy of current networks regarding these two objectives. In view of these discussions representativeness could indicate spatial representativeness (how large is the area represented by the measurements at a certain station). This is a very difficult issue. On the other hand it could also indicate how well measurements at a certain station represent the exposure of the general population. Partly this indication overlaps with the previous one. But it is perhaps easier to address and the focus of this study. 1.2. Exposure and network design The actual exposure of a person is determined by the concentration in the breathing zone. It should be noted that the total exposure of (European) individuals to air pollutants is more related to indoor air than ambient air monitored at stations. This is illustrated in Fig. 1 showing an example of how people spend their time (Dons et al., 2011). Exposure as mentioned in the AAQD however relates exclusively to ambient outdoor air.2 In Annex III of the AAQD on siting requirements, exposure it is not further specified; only in the definition of the Average Exposure Indicator some specification is given,
1
The Lancet, Volume 353, Issue 9156, Pages 874e878, 13 March 1999. It is relevant to note here that time activity patterns shown in Fig. 1 may lead to different assessments. In the study by Dons et al., 2011 it is shown that drivers may be exposed to rather significant levels of black carbon (a component emitted by cars) during the period they drive to their work. Exposure at home is much less due to the much lower levels there. This issue, although interesting, is not addressed here at this stage and we follow the directive's approach where exposure relates to ambient outdoor air quality. 2
Fig. 1. An example of the fraction of time that people in Belgium spend on certain activities. Derived from Dons et al. (2011).
relating the average exposure to urban background sites (Article 2#). Although both outdoor and indoor air (including outdoor air penetrated indoors) are important contributors to the actual exposure, the concentration-response functions established in epidemiological studies, on which the air quality standards are based, relate only to outdoor concentrations e usually those measured at urban background locations (Hoek et al., 2002; Boezen et al., 1999). In the study here we assume exposure is directly linked to the home address. This is the place where people spend most of their time as illustrated in Fig. 1. Exposure as described in the AAQD Annex IIIB is derived from measurements at stations, which are always limited in number. Obviously most people do not actually live in the direct vicinity of monitoring stations. In (urban) background conditions concentration gradients are usually small, so measurements at these stations tend to be quite representative for larger areas and are consequently suitable to assess exposure. As outlined above, the foremost aims in air quality monitoring are to assess compliance with air quality standards and population exposure, including trends therein. A network should address both issues. We will study the networks of four large European cities bearing these goals in mind. Two scales are distinguished here: Macro scale siting of a station is the selection of the type of site (urban background, near streets etc.) and the approximate location. In Annex III to the AAQD provisions on macro siting are given: “Sampling points directed at the protection of human health shall be sited in such a way as to provide data on the following:” - the areas within zones and agglomerations where the highest concentrations occur to which the population is likely to be directly or indirectly exposed for a period which is significant in relation to the averaging period of the limit value(s), - levels in other areas within the zones and agglomerations which are representative of the exposure of the general population”3
3 Article 2 of the AAQD: ‘urban background locations’ shall mean places in urban areas where levels are representative of the exposure of the general urban population.
90
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
The micro-scale siting of the station is establishing its exact position with respect to sources and obstacles in the direct vicinity. It is important to note that epidemiological studies whether investigating short term-effects or those studying long-term effects are potentially affected by the issues raised in the paper. Exposure estimates as used in epidemiological studies depend on appropriate time-dependent spatial concentration estimates that can be allocated to the exposed population under study. Thus, also long-term cumulative exposure estimates that are based rather uncritically on monitoring data may be biased if the stations are not representative. 1.3. Other studies Much of the work attributed to air pollution networks does not end up in the peer-reviewed scientific literature but is laid down in (sometimes comprehensive) reports by authorities responsible for air quality or other institutions. Very detailed analyses and interpolation schemes were already developed in the 1990's by German scientists looking into the representativeness of the monitoring network in North Rhine-Westphalia (Beier and Doppelfeld, 1992). More recently Spangl et al. (2007) developed definitions, methods and validation procedures for the classification of air quality monitoring stations. The issue of representativeness is discussed and proposals for station classification and for new concepts of representativeness have been developed and elaborated. The EURAQHEM report (Kuhlbusch et al., 2004) describes an analysis of the EU air quality monitoring network with regard to health assessments. It evaluates the limitations of the current network and makes recommendations for improvement. The study concludes that Member States provide insufficient information on representativeness of measurement sites in relation to populations, that an approach of default areas of representativeness per station site type is inadequate, and that no direct relation of air quality assessment and the exposure of the population could be generated. The SCREAM4 working group of AQUILA (Network of National Air Quality Reference Laboratories) has been developing recommendations on the representativeness of monitoring stations. 2. Methods 2.1. Methodology and models used 2.1.1. Methodology It is difficult to assess how representative a network is for a distribution of concentrations or exposures in a certain area using monitoring data only. There are a limited number of monitoring stations and the true spatial distribution of concentrations across the entire area is not known. Whereas to assess the representativeness of the network, concentration estimates in areas without monitoring stations are also needed. Therefore we took an approach based upon detailed modelling of concentration patterns in cities to study representativeness. Four large European cities (Barcelona, Berlin, London and Stuttgart) are used as examples. The following procedure was followed: An air quality model (URBIS) was applied to calculate the spatial distribution of the concentration of NO2 in ambient air in four cities (resolution: 10 m 10 m). The concentrations calculated at the location of the monitoring stations, compared well with the concentrations measured at the stations indicating that the models worked well. Therefore the calculated concentration distributions were used as a proxy for the actual concentration distributions across the cities.
4 Siting criteria, classification and representativeness of air quality monitoring stations” (SCREAM).
These estimates were used as proxy for the true exposure. The distributions of these proxy concentrations across the city populations were determined and cumulative population distribution curves were estimated. The annual mean calculated concentration values of the monitoring network stations were located on the calculated population distribution curves to estimate the fractions of the populations that the monitoring network stations represent. This macro scale procedure is used to evaluate which subgroups of the monitoring stations (if any) can reliably be used to decide on compliance or to estimate the concentration the population is exposed to. The CAR model is used to investigate the effect of micro scale siting of the monitoring stations within the streets. 2.1.2. Model used: the TNO URBIS model For each of the four cities usually detailed models are available and published by local authorities. In the approach used here, one single model was chosen to calculate the concentration level for all cities in a comparable way and with the same, high, resolution. The high resolution also allows detailed evaluation of concentrations observed at particular monitoring stations which is the purpose of the current study. It is relevant to realize that a higher resolution of the calculations leads to a larger range of concentrations i.e. there are lower and higher concentrations observed in maps based upon high resolution data (such as the 10 m used here) than there are in maps with a lower resolution (See for example for Stuttgart www. lubw.baden-wuerttemberg.de/servlet/is/242644/) Concentrations calculated using a high resolution on sites and monitoring stations near roads with intensive traffic are much higher than those calculated with a resolution of 0.5 or 1 km. The TNO-URBIS model (Beelen et al., 2010) was used to create a concentration maps across the cities. URBIS is an air pollution model designed for cities and regions with a variety of emission sources. Emission rates from all known emission sources in the area of study are input to the model. The dispersion models implemented in URBIS are: the CAR model, a dispersion model for emissions by traffic in streets for distances up to 30 m from the road (Eerens et al., 1993; Wesseling and Sauter, 2007); Motorway Plume model5 model, is a dispersion model for pollution by motorway traffic. This is a Gaussian line source model for dispersion up to 5000 m (van den Hout and Baars 1988; Wesseling and Visser, 2003); and A simplified version of the Dutch standard calculation method for point sources (van Ham et al., 1998). This is a Gaussian plume dispersion model. In this study, concentrations at a 100 100 m grid were calculated. Input to the dispersion models is the emission rate of nitrogen oxides (NOx; the sum of nitric oxide (NO) and nitrogen dioxide (NO2) emissions). The dispersion of NOx through the area is calculated by the dispersion models. The local NO2 concentration is then derived from the calculated NOx concentration using an empirical algorithm. This algorithm uses the local ozone concentrations available for reaction with NO and is based upon long series of measurements of NOx and NO2 and ozone in the air pollution monitoring network in the studied city. 2.1.2.1. Calculations. The annual average of the concentration of nitrogen dioxide in 2010 was calculated for the four cities. The data used to run the models is described in detail in Duyzer et al. (2013).
5
Translated from Dutch: The accredited TNO Pluim-Snelweg model.
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
91
Each calculation is carried out in 4 layers: 1. Regional background: a single number derived from measurements or LOTOS-EUROS calculation. 2. City background: calculation with a Gaussian plume model. The calculation has been carried out at a high resolution for the emissions of all municipal roads. The results of the calculations are averaged to a 1 1 km2 grid. 3. Motorways: calculation with a Gaussian plume model on a 10 10 m2 grid 4. City streets: calculation with the CAR model for street canyons. All contributions are added to calculate the concentration at each grid point. Special attention has been given to the location of the monitoring stations. The CAR model is very sensitive (and so are the concentrations in real life) to the distance between the road and the receptor. At this stage concentrations are calculated every 10 m.6 The locations have been inspected in Google street view, to make sure the distance between the road and the measuring site is correctly estimated. The final accuracy of the coordinates of the monitoring sites in our model calculations is now estimated to be around 1 m. 3. Results 3.1. Results of calculations with the Urbis model Fig. 2, Fig. 3, Fig. 4, and Fig. 5 show the, annually averaged, concentrations of nitrogen dioxide in the four cities. The locations of the monitoring sites are indicated with dots. As expected high concentrations are found near major roads. For Barcelona and London, the calculated concentrations are higher than for Berlin and Stuttgart. In order to optimize the contrast for each of the figures different colour scales (in webversion) have been used. To assess the quality of the calculations, Fig. 6, Fig. 7, Fig. 8 and Fig. 9 show comparisons between measured and calculated concentrations for the year 2010. In general the comparison is good. The correlation coefficients (r2) deviate less than 10% from a value of one. This means that only 10% of the variation in the concentrations measured at the stations is not explained and 90% is explained by local emissions and other parameters whose influence on the concentration is described in these models. Only in Stuttgart the difference is larger. Here the measured concentrations are 10e20 % higher than the calculated ones. This could partly be due to the specific orographic situation resulting in specific meteorological conditions in Stuttgart with for example low wind speeds. These conditions are possibly not correctly accounted for in the URBIS model. Since in the sequel only modelled values are used in the comparisons this (rather small) discrepancy to measured data in Stuttgart has no impact on the outcome of this study on the representativeness of monitoring sites. The good agreement leads us to conclude that the model calculations are suitable for our purpose to study the appropriateness of the siting's of network stations. The calculated concentrations across the cities are used in the rest of this paper as a proxy for real concentrations in these cities. 3.2. Exposure of the general public: macro scale siting 3.2.1. Exposure curves The exposure curves used in this paper present the percentage
6 This high resolution is needed to be able to calculate the concentration levels in streets. Lower resolutions of for example 1 km2 lead to lower concentrations due to spatial averaging of high and low concentrations.
Fig. 2. The calculated, annual averaged, concentration of NO2 calculated for Barcelona for the year 2010. Dots indicate monitoring sites.
of the population exposed to concentrations higher than the values indicated on the y-axis. They are constructed based on modelled concentrations and census data. As was discussed above we use home addresses (from census data, see Annex) as the place where people are exposed. Fig. 10 shows the exposure curves calculated for the four cities in one plot. The curves have a similar shape and all show that many people are exposed to relatively low concentrations and few are exposed to the high concentration ranges. If we take Berlin as an example the curve displays the following going from left to right: - 80% of the population is exposed to levels higher than 15 mg/m3 whereas only 10% of the people is exposed to levels higher than 24 mg/m3 So 70% of the population is exposed to levels between 15 and 24 mg/m3 - Only a very small group, 0.4% of the population, is exposed to levels higher than 40 mg/m3 The figure indicates that, based on our calculations, the general population in Barcelona experiences the highest exposure, followed by London, Stuttgart and Berlin. The median values (50 percentiles), the levels to which 50% of the people are exposed, support this feeling probably best. The median values ranges from around 16 mg/m3 for Berlin, to 23 for Stuttgart, to 34 for London and to 38 mg/m3 for Barcelona.
92
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
Fig. 3. The calculated, annual averaged, concentration of NO2 calculated for Berlin for the year 2010. Dots indicate monitoring sites.
Fig. 4. The calculated, annual averaged, concentration of NO2 calculated for London for the year 2010. Dots indicate monitoring sites.
Fig. 11, Fig. 12, Fig. 13 and Fig. 14 show the same exposure curves extended with the concentrations calculated at the monitoring sites. Note that, to guarantee consistence, modelled (calculated) concentrations are used rather than concentrations measured at the sites. As a result the figures are more or less independent of the quality of the model calculations. In these figures the stations labelled as urban background stations and the traffic stations are indicated. As can be expected, the traffic stations typically represent the higher concentrations. Only in Barcelona some street stations exhibit lower concentrations than background stations. Here the classifications could be flawed. These figures show that the monitoring sites chosen in the cities result in a different coverage of the exposure curves. We judge the distribution of monitoring stations along the x-axis (representing the concentration range) as well as the y-axis (reflecting the population exposure).
Fig. 5. The calculated, annual averaged, concentration of NO2 calculated for Stuttgart for the year 2010. Dots indicate monitoring sites.
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
93
Fig. 6. Annually averaged concentration of NO2 calculated for 2010 using the URBIS model compared to concentration measured at AIRBASE71 stations in Barcelona. The correlation coefficient (squared) and the best fitted (least square method) straight line through the origin is also displayed.
The distribution of sites in Barcelona does not cover either of the two axes well and hence the exposure of the general population nor the occurrence of the highest concentrations is well represented. The stations in Berlin are well distributed across the y-axis, representing the exposure of the general population. The sites cover the x-axis less well, perhaps missing the locations where the highest concentrations are calculated. It should be noted however that only 0.02% of the population lives in areas where the concentrations is higher than 60 mg/m3 (more on this below) In London the sites are fairly evenly distributed along the exposure graphs. The stations are well distributed across the y-
axis, thus representing the exposure of the general population and also across the x-axis, giving a good picture of the highest concentrations. The sites in Stuttgart hardly cover the y-axis and hence the exposure of the general population is not well represented. The x-axis is better covered, but the sites are at locations where high concentrations can be expected. The average urban background calculated from the urban background stations, as well as the population averaged exposure, depends significantly on the siting of the stations that cover this area, particularly when their number is small. Clearly the networks
Fig. 7. See caption Fig. 6 but for Berlin.
94
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
Fig. 8. See caption Fig. 6 but for London.
Fig. 9. See caption Fig. 6 but for Stuttgart.
differ significantly in the extent to which the highest levels are represented and also in the coverage of the levels to which the general population is exposed. 3.2.2. Population weighted concentration The population weighted concentration8 (or Average Exposure
7
http://acm.eionet.europa.eu/databases/airbase/. From the AAQD The Average Exposure Indicator expressed in mg/m3 (AEI) is based upon measurements in urban background locations in zones and agglomerations throughout the territory of a Member State. It should be assessed as a three-calendar year running annual mean concentration averaged over all sampling points established pursuant to Section B of Annex V. The AEI for the reference year 2010 shall be the mean concentration of the years 2008, 2009 and 2010. At this stage it is only mentioned for PM2.5. But it's value may be derived for NO2 as well. 8
Indicator) calculated from the modelled concentration distribution is an interesting measure to compare the exposure in different cities. In Duyzer et al. (2013) it is shown that this parameter compares best with the average concentration measured at the urban background sites. If the results of monitoring at the street sites are included in the average the results differ significantly.
3.2.3. Assessing compliance with limit values in a comparable way Selecting a site with the aim to assess compliance with the limit values and at the same time representative for high exposures is not a simple task. It is interesting to evaluate the locations of the traffic stations designed for this purpose. Such sites are usually selected in busy streets where concentrations are expected to be high. Although such sites aim to fulfil the criteria for assessments of compliance
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
95
Fig. 10. Exposure concentration curves for all four cities. The population weighted average concentration calculated for Barcelona, Berlin, London and Stuttgart is 43.2, 18,7, 37.2 and 26,4 mg/m3 respectively.
Fig. 11. Exposure concentration curves for Barcelona. The symbols indicate the annual average concentrations calculated at the monitoring sites. Urban background stations are distinguished from street stations.
with the limit values, it is often not obvious how well measurements at these stations represent the exposure of the population. 3.2.3.1. How many people are exposed to concentrations observed at the street stations?. From the viewpoint of comparability it is useful to look at the position of these stations on the exposure concentration curves. For each of the four cities, the highest concentration calculated at the monitoring sites is given in Table 1. Also the position this site takes on the exposure curve is presented. Large differences are found. It is interesting to look for the highest concentration calculated for the location of one of the monitoring stations and extract the number of people exposed to such levels from these curves. In Stuttgart only 0.02% of the population is
exposed to concentrations higher than 85 mg/m3 whereas in Barcelona 3% of the population is exposed to concentrations higher than calculated at the monitoring station with the highest concentration of 66 mg/m3. The difference between 0,02% and 3% of the people exposed is especially important if one the associated concentrations. If in Barcelona, London and Berlin the traffic sites would have been located at a site with a concentration to which 0,02% of the population is exposed the mentioned concentration level would have been much higher (like it is now in Stuttgart). Fig. 15 shows that if the sites with the highest concentrations in London, Berlin and Barcelona would also lie at the 0,02% level, the concentration would be higher by 65, 11 and 26 mg/m3 respectively! And the conclusions
96
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
Fig. 12. As Fig. 11 for Berlin.
Fig. 13. As Fig. 11 for London.
on the highest concentrations to which people are exposed would change dramatically. The concentrations would even be higher than those observed in Stuttgart. The example given illustrates the issue that concentrations observed at these traffic stations have little value in comparing exposure to high concentrations in different cities. It also indicates that the network of Stuttgart complies best with the AAQD's provision that the highest concentrations to which the population is exposed should be assessed. The representativeness of street stations for other streets is difficult to judge by itself. It seems that the street stations in the cities studies here represent exposure between 0,5% of the population (Berlin) and 15% of the population (Barcelona). In that respect it seems that these stations serve their purpose in that they represent the exposure in busy streets.
3.3. Micro scale siting The micro scale siting of the stations is described in the AAQD (Annex III). Whereas the location of inlet tubes, distance to walls etc. is described in some detail in the AAQD, not all important parameters, such as the exact location in the area, are treated very extensively. It is clear that urban background stations should not be located in busy roads. For street stations aiming at higher concentrations the traffic intensity is on the other hand crucial. The contribution of local traffic to the air concentration is coupled with local emissions and therefore related to the traffic intensity (and composition etc.). So to find locations for hotspot sites, streets with a high traffic intensity will need attention. Here several factors are relevant and determining the outcome of measurements.
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
97
Fig. 14. As Fig. 11 for Stuttgart.
Table 1 Highest concentration calculated on a monitoring site and the location on the exposure-concentration curve (expressed as percentage of the population that is exposed to concentration higher than this concentration). City
Highest concentration (mg/m3)
Percentage (%)
Barcelona Berlin London Stuttgart
66 50 98 85
3 0,1 0,81 0,02
Fig. 15. Concentrations calculated at the site with the highest concentrations compared to the concentration to which less than of the 0.02% of the population is exposed.
The effect of some of these factors may easily be described using models such as the CAR model described above. This is illustrated in Fig. 16 where the contribution of the emission by road traffic to the concentration of nitrogen dioxide in air, is shown as a function of the distance to the road, the street type and congestion. In this case a typical busy road with 80,000 vehicles per day was used. This graph illustrates in the first place how important the distance of the
sample inlet to the road is and secondly how important, for example, street type is. This means that a concentration measured in one city in a narrow street canyon needs to be judged differently from a concentration measured in another city with similar traffic intensity in a broad street. Table 2 summarizes the results indicating the strong influence of distance to the centre of the road, the difference between a narrow and a wide street and the impact of congestion. In the presented example, whether a monitoring station is 5 or 20 m away from the roadside makes a difference of approximately 16 mg/m3 in the calculated concentration. This indicates how important the distance to the road actually is.9 The street type, describing the type of buildings along the road, may have a large impact as well. Differences in concentrations between two types of streets may be as large as 10 mg/m3 or more. Congestion may have a similarly large impact, in this case of more than 10 mg/m3 All of these calculations show the importance of (knowledge of) the exact location of a monitoring site in a street and the kind of buildings along the road. The differences are large with respect to exceedances of limit values. Measures taken to improve air quality such as a less dense traffic or a reduction of emission factors may lead to improvements in the air concentration. The effect of another location of a station (a site farther away from the road) or a different the street type may be much larger. This makes comparison of monitoring results among different cities more difficult as is shown in Fig. 17 which illustrates different interpretations of the selection criteria for monitoring. In Stuttgart some monitoring stations are within few meters from the first lane whereas in Barcelona the distance of some stations seems larger and even more than 10 m from the first lane with parked cars. Depending on the type of street canyon, differences could be as high as 14 mg/m3 between a station at 10 or one at 20 m away from the centre of the road. This statement is of course only valid for the given example of 80,000
9 For all pollutants, traffic-orientated sampling probes shall be at least 25 m from the edge of major junctions and no more than 10 m from the kerbside (from the AAQD ANNEX III C).
98
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
Fig. 16. The effect of street type and congestion on the contribution of a road to the concentration observed at certain distances from the centre of the road. Concentrations calculated for a distance less then 5 meters are often above the road lane and therefore only here as an indication. The emission source is located in the middle of the road. Traffic intensity of nearly 80,000 vehicles per day with 4% trucks. 10% direct emission of NO2 and the concentration of ozone in the area is 40 ppb.
vehicles per day in this for lane road. Less is known about the impact of the position along the street. In the AAQD it is stated that a station should be representative of at least 100 m along the road. Models such as CAR can be used to investigate the impact of several parameters. It is more difficult to judge the effect of complex, infrastructural conditions. Here, much more complex models are needed. Using Computational Fluid Dynamics (CFD) based models calculations in these complex situations can be made. This would allow estimations of the impact of large infrastructures (buildings, bridges, tunnel exits) on concentrations and measurements thereof on a specific site. The absolute accuracy in the results of CFD calculations may be limited. Still in the absence of detailed local scale measurements the calculations could be helpful to judge the quality of a monitoring station on a local scale. Qualitatively they illustrate where problems might occur or might have occurred. In Duyzer et al. (2013) the siting of a monitoring station (Am Neckartor) in Stuttgart was judged. The calculations showed that in view of some of the definitions in the AAQD (the area where the concentration varies by less than 10%) the representativeness of street stations is only small. Perpendicular to the road the concentration varies by more than 10% within 10 m and in the direction along the road the concentration varies within areas of scale 100 m as well.
4. Discussions and conclusions 4.1. Conclusions We have used model calculations to evaluate the positioning of air pollution monitoring stations in the framework of the AAQD. Basis for our analysis were the calculated concentration fields in four major European cities for nitrogen dioxide (NO2), a pollutant for which limit values exist. The exposure of inhabitants to nitrogen dioxide was calculated and the representativeness of the monitoring network was judged. A distinction was made between
macro-scale and micro-scale siting. Based on the analysis of the concentrations in four cities the following conclusions were drawn: Macro-scale siting. In three cities the exposure of the general population can rather well be assessed using measurements at urban background stations. Averages of concentrations calculated at urban background stations show good agreement with the exposure weighted concentration for a city. Individual urban background stations may however not be representative of the average exposure, as the case of Stuttgart shows. In the smallest of the studied cities there is only one background station. Clearly the uncertainty of only one sample is large. Traffic or hotspot stations are aimed at assessing compliance with the limit values and aim to measure the highest concentrations. The following is noted: The results of measurements at traffic stations are not useful to assess exposure of the general population. They often indicate only the exposure of a (sometimes very) small fraction of the population. The model calculations suggest that in one city (Stuttgart) only 0.02% of the population is exposed to concentrations higher than the concentrations at the most polluted station, while in another city (Barcelona) this is valid for the exposure of 3% of the population. If all cities would have a station at the 0.02% exposure level, like in Stuttgart, the highest observed concentration in those cities would increase by, depending on the city, 10, 20 and even 60 mg/m3 compared to the currently measured highest level. It is therefore difficult to draw conclusions on high exposure when comparing the highest observed concentration in different cities. These depend too much on the location of the monitoring site. Hence the highest concentration at for example the 3% level in Barcelona does not represent the highest level of concentrations as required by the AAQD, in spite of the relatively high proportion of hotspot stations in this city. According to
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101 Table 2 Calculated contribution of emissions of a road to the concentration of NO2 at various distances from the centre of the road. Distance to centre of road C(NO2) in (where emission takes wide street place) (m) canyon
C(NO2) in C(NO2) in narrow narrow street street canyon with canyon congestion
5 10 20
44 36 22
30 24 14
44 35 20
99
CFD calculations may be used to qualitatively judge the quality of monitoring sites on a local scale in more complex situations. Examples indicate that in the direction along the road the concentration may vary within distances of the scale of 100 m. Because of the unclear representativeness for concentrations in the direct surroundings, it is difficult to use differences between concentrations measured at hotspot stations to assess the differences in population exposure between cities.
Fig. 17. Some street stations in Stuttgart (above) and Barcelona (below). This illustrates the differences in the distances between the station and the road.
our calculations sites may be found in this city with concentration levels several tens of mg/m3 higher. In the interpretation and especially the comparison of results of measurements at traffic stations, models should play a role. The results of model calculations can be used to judge the position of stations on the concentration exposure curves. Without such analysis comparison is very difficult, if not useless at all. Micro-scale siting. Micro-scale siting of street stations is very important as well. Especially the distance to roads is a major parameter; positioning a station closer to the road can lead to substantially higher concentrations. Other important parameters are: the street configuration (width of the road, buildings on one or two sides, height of the buildings), traffic flow (free flow or congestion).
It is relevant to note that the analysis here is made for nitrogen dioxide. This compound is emitted by cars but is also formed locally as a result of the chemical reactions in air between nitric oxide (also emitted by cars but in larger quantities) and ozone. This chemical reaction is relatively fast but slower than the dispersion through the air. As a result the gradient in concentration of nitrogen dioxide downwind of sources is smaller than it is for inert species such as soot or carbon monoxide. Hence the area of representativeness for these compounds is expected to be smaller than for nitrogen dioxide. 4.2. Possible consequences for siting of monitoring stations The most relevant shortcomings that we observed pertained not to individual stations but to the representativeness of the station networks. The four urban monitoring networks that we evaluated and compared do not give a similar representation of the spatial distribution of the concentrations in the cities concerned. The
100
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101
Stuttgart network has a strong focus on the highest concentrations in the city, in conformity with the objective to represent the area with the highest concentrations/exposures. The three other networks did not mainly focus on the locations with the highest concentrations. A key question is which of the approaches fulfils the requirements of the directive best. Clearly the Stuttgart network is the most adequate in identifying the highest concentrations in the ambient city air, as also required by the directive. This particular network is however less suited to assess exposure of the general population because of the limited number of background sites. Street stations can of course play a role in assessing the exposure in streets. The results of monitoring here could be coupled with the time spent at such a site. Coupled with the applicable time-activity patterns monitoring at street sites is useful. Using the results to calculate the simple average of exposure for a city may lead to biased results. Station siting issues have been subject to continual debate and fully satisfactory solutions have not been found and may in fact not be existent at all. The need to find a compromise between practicable solutions and requirements has offered network designers a substantial freedom in locating the stations. There is a need for more harmonisation in network design. At the same time it may not be possible or even needed to change current networks and break up important time series. A very important outcome of monitoring is the possibility to analyse trends in concentrations and evaluate for example the effect of (long term) measures aimed at reduction of exposure. Therefore it is recommended to not change current networks drastically but rather harmonise the interpretation of the results. Whereas the results of measurements at urban background stations, if located at representative sites, can usually be used directly to assess the exposure of the general public, the use of concentrations observed at street stations is much more complex. It seems impossible to interpret these data in a harmonised way without using simulation models such as the models used in this study. Model calculations of the concentration distribution in the street network of a city, taking parameters such as traffic intensity, distance to the road, street type into account, can provide insight in how well street stations cover the distribution of the highest concentrations in the city. After more research such model calculations may even be used to compensate for the influence of many important, but well-known, parameters such as distance to the road. This compensation could make it possible to realistically compare the results of concentration measurements in different cities. Models can also be used to estimate the distribution of the exposure of people to concentrations of pollutants. This would allow fair comparison of the exposure between cities. Model calculations would also relief the emphasis on station classifications some off which seem to be flawed at the moment. Apart from stations that are obviously on a wrong location all stations in a network that has a significant spatial coverage can be used and have their value. The cause of higher concentrations at some stations can simply be interpreted since all available knowledge may be incorporated in the model. Acknowledgement The study is funded by EUGT (the Europ€ aische Forschungsvereinigung für Umwelt and Gesundheit im Transportsektor www.eugt.org). Annex. details of input to model calculations For the four cities the following information was used. For all cities the following information was used:
Emission factors: TREMOVE/COPERT, http://www.tremove.org/, http://lat.eng.auth.gr/copert/. Urban background: calculated with Gaussian plume model Regional background: measurement Census data Census data were obtained from the European Union data bases http://www.eea.europa.eu/data-and-maps/data/populationdensity-disaggregated-with-corine-land-cover-2000-2. The follow city specific data were used: Stuttgart Roads (map þ traffic intensities): Lohmeyer, project 60385-050: The basic input data (street types, traffic intensity, meteo€chlin, (2006) rology) are based upon Nagel and Ba The Stuttgart data contained detailed information of the configuration of the street (‘Schluchttype’). This has been converted to the street types of the TNO CAR model. Meteorological data: Lohmeyer, project 60385-05-0. Ten different areas for meteorological conditions have been used in this study.
Berlin Roads (map þ traffic intensities): Email communication. Senatsverwaltung für Gesundheit, Umwelt und Verbraucherschutz - III D 24 Brückenstr. 6, 10173 Berlin The Berlin data contained information about the street type: street canyon or open. Meteorological data: interpolated data from LOTOS-EUROS (Schaap et al., 2008)
London Roads (map þ traffic intensities): London Atmospheric Emission Inventory (LAEI, http://data.london.gov.uk/datastore/package/ london-atmospheric-emissions-inventory-2010 Meteorological data: interpolated data from LOTOS-EUROS
Barcelona Roads (map þ traffic intensities) data collected in INTARESE project (http://intarese.org) Meteorological data: interpolated data from LOTOS-EUROS References Beelen, R., Voogt, M., Duyzer, J., Zandveld, P., Hoek, G., 2010. Comparison of the performances of land use regression modelling and dispersion modelling in estimating small-scale variations in long-term air pollution concentrations in a Dutch urban area. Atmos. Environ. 44 (36), 4614e4621. Beier, Doppelfeld, 1992. LIS-Berichte; Nr. 101 R€ aumliche Übertragbarkeit und €tsdaten im Messnetz TEMES. http://www.zvab. Interpolation von Luftqualita com/LIS-Berichte-R%C3%A4umliche-%C3%9Cbertragbarkeit-InterpolationLuftqualit%C3%A4tsdaten-Messnetz/165294139/buch. Boezen, M., van der Zee, S., Postma, D.S., Vonk, J.M., Gerritsen, J., Hoek, G., Brunekreef, B., Rijcken, B., Schouten, J.P., 1999. Effects of ambient air pollution on upper and lower respiratory symptoms and peak expiratory flow in children. Lancet 353 (9156), 874e878. Dons, E., Int Panis, L., van Poppel, M., Theunis, J., Willems, H., Torfs, R., Wets, G., 2011. Impact of time activity patterns on personal exposure to black carbon. Atmos. Environ. 45, 3594e3602. Duyzer, J., van den Hout, D., Zandveld, P., van Ratingen, S., 2013. Representativeness of Air Quality Monitoring Stations. TNO report 2013 R11055 Available from: EUGT.
J. Duyzer et al. / Atmospheric Environment 104 (2015) 88e101 Eerens, H., Sliggers, C., van den Hout, K.D., 1993. The CAR model: the Dutch method to determine city air quality. Atmos. Environ. 27B, 389e399. Hoek, G., Brunekreef, B., Goldbohm, S., Fischer, P., van den Brandt, P.A., 2002. Association between mortality and indicators of traffic-related air pollution in the Netherlands: a cohort study. Lancet 360, 1203e1209. Kuhlbusch, Th, John, A., Hugo, A., Peters, A., von Klot, S., Cyrys, J., Wichmann, H.-E., Quass, U., Bruckmann, P., 2004. Analysis and Design of Local Air Quality Measurements. Towards European Air Quality Health Effect Monitoring. Service Contract 070501/2004/389487/MAR/C1, Brussels. Spangl, W., Schneider, J., Moosmann, L., Nagl, C., 2007. Representativeness and Classification of Air Monitoring Stations. UBA, Vienna. van den Hout, Dick, Voogt, Marita, Moosmann, Lorenz, Nagl, Christian,
101
Spangl, Wolfgang, 2012. Survey of Views of Stakeholders, Experts and Citizens on the Review of the EU Air Policy. TNO, Netherlands. TNO report TNO-060-UT2012e00714. Available at: http://ec.europa.eu/environment/air/pdf/Survey% 20AQD%20submitted for publication%20-%20Part%20I%20Main%20results.pdf. van Ham, N.J., Duijm, J., Erbrink, J.J., van Jaarsveld, J.A., Pulles, M.P.J., Schols, E., Verver, G.H.L., 1998. Revision of the Netherlands National Model for short range dispersion of air pollutants. Int. J. Environ. Pollut. 8, 3e6. Wesseling, J., Sauter, F., 2007. Calibration of the Program CAR II Using Measurements of the National Monitoring Network of the RIVM (in Dutch). Bilthoven, RIVM report 680705004.2007. Wesseling, J., Visser, G.Th, 2003. An Inter-comparison of the TNO Traffic Models, Field Data, and Wind Tunnel Measurements. TNO report 2003/207.