- Email: [email protected]
A model for evaluating the population exposure to ambient air pollution in an urban area
Atmospheric Environment 36 (2002) 2109–2119
A model for evaluating the population exposure to ambient air pollution in an urban area Anu Kousaa,b,*, Jaakko Kukkonenc, Ari Karppinenc, P.aivi Aarnioa, Tarja Koskentaloa a
b
Helsinki Metropolitan Area Council (YTV), Opastinsilta 6 A, 00520 Helsinki, Finland Department of Environmental Hygiene, National Public Health Institute (KTL), P.O. Box 95, 70701 Kuopio, Finland c Finnish Meteorological Institute (FMI), Air Quality Research, Sahaajankatu 20 E, 00810 Helsinki, Finland Received 21 September 2001; received in revised form 20 February 2002; accepted 27 February 2002
Abstract A mathematical model is presented for the determination of human exposure to ambient air pollution in an urban area. The main objective was to evaluate the spatial and temporal variation of average exposure of the urban population to ambient air pollution in different microenvironments with reasonable accuracy, instead of analysing in detail personal exposures for specific individuals. We have utilised a previously developed modelling system for predicting the traffic flows and emissions, emissions originating from stationary sources, and atmospheric dispersion of pollution in an urban area. A model was developed for combining the predicted concentrations, information on people’s activities (such as the time spent at home, in the workplace and at other places of activity during the day) and location of the population. Time-microenvironment activity data for the working-age population was obtained from the EXPOLIS study (air pollution distributions within adult urban populations in Europe). Information on the location of homes and workplaces was obtained from local municipalities. The concentrations of NO2 were modelled over the Helsinki Metropolitan Area for 1996 and 1997. The computed results were processed and visualised using the geographical information system (GIS) MapInfo. The utilisation of the modelling system has been illustrated by presenting numerical results for the Helsinki Metropolitan Area. The results show the spatial and temporal (diurnal) variation of the ambient air NO2 concentrations, the population density and the corresponding average exposure. The model developed has been designed to be utilised by municipal authorities in urban planning, e.g., for evaluating the impacts of traffic planning and land use scenarios. r 2002 Elsevier Science Ltd. All rights reserved. Keywords: Exposure model; Population exposure; Nitrogen dioxide; EXPOLIS; GIS
1. Introduction Harmful effects have been associated with NO2 exposure levels that are below the ambient air quality guideline levels set by the World Health Organization (WHO, 2000) (e.g., Brunekreef et al., 1989; Schwartz et al., 1991; Mukala et al., 2000). Most of the *Corresponding author. Helsinki Metropolitan Area Council (YTV), Opastinsilta 6 A, 00520 Helsinki, Finland. Fax: +3589-1561-334. E-mail address: anu.kousa@ytv.fi (A. Kousa).
epidemiological studies (exception, Mukala et al., 2000) have been based on nitrogen dioxide concentrations at fixed ambient air quality monitoring sites. However, the measurement data from these stations do not necessarily represent areas beyond their immediate vicinity, as the concentrations of pollutants in urban areas may vary by orders of magnitude on spatial scales varying from tens to hundreds of metres. For example, Alm et al. (1998) addressed the measured personal NO2 exposures of pre-school children in Helsinki in 1991. The concentration data obtained from stationary monitoring sites explained only a minor fraction of the personal
1352-2310/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S 1 3 5 2 - 2 3 1 0 ( 0 2 ) 0 0 2 2 8 - 5
2110
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
weekly NO2 exposure variations of children, as shown by a statistical regression analysis (the value of the corresponding coefficient of determination, r2 ; was 0.37). However, the above-mentioned NO2 exposure variations were statistically strongly associated with the NO2 concentrations that were measured inside and outside the nurseries (r2 ¼ 0:88 and 0.86, respectively). A similar result was obtained in the EXPOLIS study that addressed measured personal exposures of adult inhabitants in the Helsinki Metropolitan Area in 1996 and 1997 (Jantunen et al., 1999; Kousa et al., 2001a). The statistical associations of these exposures and the measured residential outdoor and indoor, and workplace indoor concentrations varied from fairly weak to reasonably strong (r2 ¼ 0:40; 0.45 and 0.55, respectively). In evaluating the exposure of the population to air pollution, there is a need to evaluate spatial concentration distributions produced, for example, by atmospheric dispersion modelling systems or a set of onsite measurements. Clearly, the modelling systems need first to be thoroughly evaluated against experimental data. Models for evaluating exposure to air pollutants have been classified as statistical, mathematical and mathematical-stochastic (modified from Ryan, 1991). The statistical approach involves the statistical determination of the measured exposures in terms of the factors that are assumed to influence these exposures. Mathematical modelling involves application of emission inventories, combined with atmospheric dispersion and population activity modelling. The stochastic approach attempts to include a treatment of the inherent uncertainties in the model, e.g., those caused by the turbulent nature of atmospheric flow. Mathematical modelling utilising emission and dispersion models is also called deterministic modelling. Source apportionment methods can be used in order to analyse the contribution of various emission categories to the total human exposure. Various exposure models have been developed in order to support health risk assessment and management. Examples of such models are ‘‘American national air quality standards exposure model’’ (NEM, Sexton and Ryan, 1988), ‘‘air pollution exposure model’’ (AirPEx, Freijer et al., 1997) and ‘‘simulation of human air pollution exposures’’ (SHAPE, Ott, 1985). All of these models are based on the same concept: the concentrations in various microenvironments are weighted with the fraction of time that people spend in these microenvironments. However, these models do not take advantage of administrative databases containing population statistics or geographical information system’s. Most recently, mathematical exposure models have been presented by Bohler and Riise (1997), Clench-Aas et al. (1999), Jensen (1999), Johansson et al. (1999) and
Dimitroulopoulou et al. (2001). The air quality management system AirQUIS has been extended to include a fairly simple exposure model (Bohler and Riise, 1997; Clench-Aas et al., 1999). The model database contains a building register that includes information on building coordinates, number of floors and number of residents. The coordinates of the buildings can then be defined as receptor points in atmospheric dispersion modelling. The exposure modelling part of the system associates a number of persons with each individual building, and combines this information with the computed ambient air concentrations, utilising a geographical information system (GIS). The model presented by Jensen (1999) is based on the utilisation of traffic flow computations and the operational street pollution model (OSPM) for evaluating outdoor air pollutant concentrations. The activity patterns of the population have been evaluated using various administrative databases and standardised timeactivity profiles. The modelling system utilises a GIS in combining and processing the concentration and population activity data. The model was applied to evaluate population exposure in one specific municipality in Denmark. An advantage of this system is that the application of a street canyon dispersion model facilitates a fine-scale spatial resolution (metres or tens of metres). However, the complexity of the model computations requires a limitation in the spatial domain to be evaluated; the computations for a more extensive area, such as that of a major city, were not possible. Johansson et al. (1999) evaluated population exposure to NO2 and particulate matter (PM) within the SHAPE study (Stockholm study on health effects of air pollution and their economic consequences). They used a multiple-source emission and Gaussian dispersion modelling system to predict the spatial distributions of NO2 and PM10 concentrations in the municipality of Stockholm. The population locations were obtained from official national data sets. However, a treatment of the chemical reactions of nitrogen oxides was not included; instead, a purely empirical relation was utilised in computing the NO2 concentrations. The model applied was not combined with a GIS system. Dimitroulopoulou et al. (2001) have evaluated personal NO2 exposure by using a deterministic model. Their model can distinguish three types of exposures: outdoor air exposure, indoor air exposure resulting from penetration of outdoor air, and indoor air exposure resulting from indoor sources. The model evaluates personal exposures, by combining the information on the movements of typical individuals in various microenvironments and the modelled microenvironment concentrations. The outdoor concentrations of NO2 were evaluated based on the measurement data from urban background sites. However, the model applied was not combined with a GIS system.
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
We have previously developed an integrated modelling system for predicting the traffic flows and emissions, emissions from stationary sources, and atmospheric dispersion of pollution in an urban area; for a description of this system and its evaluation against the data from an urban monitoring network, the reader is referred to Karppinen et al. (2000a,b). The modelling system developed has been utilised to evaluate environmental impacts of traffic planning and land use scenarios in the area. Emissions, NO2 concentrations and preliminary potential NO2 exposures have been assessed in the transportation system plan for the Helsinki Metropolitan Area (Helsinki Metropolitan Area Board, 1999). The main purpose of this paper is to present an extension of the above-mentioned mathematical modelling system to allow for the average exposure of the human population to ambient air pollution. Clearly, the model cannot evaluate the personal exposure of each member of the population. The internal variation of personal exposures is larger (i.e., these can be substantially higher or lower), compared with that of average population exposures. Further, the model presented here considers only hourly averaged values; the short-term peak exposures can be drastically higher. In principle, the model is applicable for any pollutant, if the required spatial concentration distributions are available. The predicted NO2 concentrations utilised in this study, computed for the years 1996 and 1997, have also been compared with the ambient NO2 concentrations measured in an urban air quality monitoring network (Kousa et al., 2001b). These predicted concentrations have also been compared with the measured residential outdoor air concentrations determined in the EXPOLIS study.
2. Materials and methods 2.1. Emissions and meteorological data We have updated the previously conducted emission inventory of NOx in the Helsinki Metropolitan Area (Karppinen et al., 2000a) for the years 1996 and 1997. The inventory includes the emissions from various mobile sources (road traffic, harbours and marine traffic, and aviation) and stationary sources (power plants, other point sources and residential heating). The traffic flows and average travel speeds were computed using the EMME/2 transportation planning system (INRO, 1994); vehicular emissions were evaluated using the EMME/2 system and emission factors that have been evaluated for this area by the Helsinki Metropolitan Area Council. The model allows for the diurnal and weekly variations both in traffic volumes and speeds, as well as in traffic emissions. Stationary sources
2111
are considered as point or area sources. The computations included approximately 5000 line sources, 169 point sources, area sources and the regional background concentrations. We used the meteorological database of the Finnish Meteorological Institute, which contains weather and sounding observations. A combination of data from the stations at Helsinki–Vantaa airport (about 15 km north of Helsinki town centre) and Helsinki–Isosaari (an island about 20 km south of Helsinki) were employed. The mixing height of the atmospheric boundary layer was evaluated using a meteorological pre-processor, based on the sounding observations made at Jokioinen (90 km northwest of Helsinki) and on routine meteorological observations. 2.2. Atmospheric dispersion modelling The dispersion modelling is based on a combined application of the road network dispersion model CAR. FMI (e.g., H.arkonen et al., 1995), applied for evaluating the dispersion of pollution originating from vehicular traffic, and the urban dispersion modelling system UDM-FMI (Karppinen et al., 2000a), for evaluating the dispersion from stationary sources. The predictions of the CAR-FMI model have been compared with the results of two measurement campaigns conducted near . major roads (i) in a suburban area (e.g., H.arkonen et al., 1997) and (ii) in a rural area (Kukkonen et al., 2001a; . Ottl et al., 2001). However, the CAR-FMI model is a finite line source model that does not explicitly take into account the influence of buildings and other obstacles on dispersion. The relevant meteorological parameters for the models are evaluated using data produced by a meteorological pre-processing model that has been specifically adapted for the urban environment (Karppinen et al., 2000c). The modelling system includes a statistical and graphical analysis of the computed time series of concentrations. The modelling system also comprises a method that allows for the chemical interaction of pollutants (including the basic reactions of nitrogen oxides and ozone), originating from a large number of urban sources (Karppinen et al., 2000a). 2.3. Exposure modelling The main objective was to evaluate the average exposure of the population with reasonable accuracy, instead of the personal exposures of specific individuals. Based on the measured results of the EXPOLIS project, the ratios of the indoor and outdoor concentrations (I/O) for nitrogen dioxide were on the average 0.76. However, the detailed value of this ratio depends on numerous factors, especially on the occurrence of indoor sources, such as tobacco smoke and gas stoves. In this
2112
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
study, we assumed for simplicity that residential and workplace indoor concentrations of nitrogen dioxide were the same as the corresponding outdoor concentrations. The part of the modelling system that evaluates the average exposure of the population to air pollution has been named EXPAND (‘‘EXPosure to Air pollution, especially to Nitrogen Dioxide and particulate matter’’). A schematic diagram of the EXPAND model is presented in Fig. 1. The model utilises as input values (i) data on the spatial location of the population, (ii) time-microenvironment activity data and (iii) computed spatial pollutant concentration distributions. The computational modules contain a separate computer program for combining and processing the data, and the GIS MapInfo. The model yields as output the spatial distribution of the pollutant concentration, the activity (density) of population and the exposure of population to ambient air pollution in the selected grid. 2.3.1. Description of the input data sets 2.3.1.1. Location of the population. The information on the location of the population comprises residential and workplace coordinates, the number of people living or working at these places, and traffic flow information concerning the road and street network. We also evaluated the location and number of people involved in other activities outside their homes, besides working (e.g., recreational activities). We have utilised a dataset that is collected annually by the municipalities in the Helsinki Metropolitan Area and disseminated in compact disk format. This dataset also contains various data on the enterprises
Fig. 1. A schematic presentation of the exposure model.
and agencies located in the area. The dataset provides geographic information on the total number of people living in a particular building or working at a particular workplace. However, this information does not identify individual persons, and there is no information, e.g., on the age of personnel at each specific workplace. The information on other locations, besides residential and workplace sites, was produced using questionnaire information gathered previously by the Helsinki Metropolitan Area Council. We used this information together with information on the amount of journeys attracted by various places (for instance, shops and homes), in order to evaluate other activities. The location of people in vehicular traffic was evaluated using the computed traffic flow information; this information is available separately for buses and cars for each street section on an hourly basis. The computed traffic flow information includes the number of vehicles, their average speeds and passenger mileage.
2.3.1.2. Time-activity data. We utilised the time-microenvironment activity data produced for Helsinki in the EXPOLIS study. EXPOLIS is a European multicentre, multicomponent exposure study; the centres include Athens, Basel, Grenoble, Helsinki, Milan and Prague. The EXPOLIS study has focused on European adult urban populations, from 25 to 55 yr of age, and their personal air pollution exposures to NO2, CO, PM2.5 and 30 volatile organic compounds (VOCs). The EXPOLIS measurements have been conducted during working weeks. In the Helsinki Metropolitan Area, time-microenvironment activity data have been collected for 435 randomly selected working age citizens; for a sub-set of this population group comprising of 201 citizens, microenvironment concentrations and personal exposures have been measured, both of these during the period from October 1996 to December 1997. The timemicroenvironment activity data have been recorded at the intervals of 15 min. For a more detailed discussion, the reader is referred to Jantunen et al. (1998, 1999). The time-activity of the population was divided into four categories: home, workplace, traffic and other activities (e.g., recreational activities). The diurnal variation of population activities in various microenvironments in the Helsinki Metropolitan Area is presented in Fig. 2. In the data presented in this figure, we have combined indoor and outdoor time-activity in each microenvironment. People spend on average 87% of their time indoors in the area considered. As expected, on average people spend most of their time at home and work environments. In the late afternoon and early evening (from approximately 4 p.m.to 8 p.m.), people spend a substantial fraction of their time in traffic and in ‘‘other’’ activities (these include shopping and various recreational activities, and the related travelling).
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
2113
was therefore multiplied by the above-mentioned fraction (0.82). The population activities at other locations (other than for home and workplaces) were also evaluated. The number of persons carried by vehicular traffic was evaluated based on the predicted traffic flow information. In the case of buses, the number of persons and the time they spent in each street section is predicted directly by the EMME/2 model. In the case of private cars, the EMME/2 model predicts the number of cars; we assumed that the number of passengers in each car is equal to the average value in the area i.e., 1.2 (M.akel.a, 1999). Fig. 2. The diurnal variation of the activity of the population in the Helsinki Metropolitan Area during the period from October 1996 to December 1997, evaluated in the EXPOLIS study.
2.3.1.3. Predicted concentrations. We computed the concentrations of nitrogen oxides (NOx) and nitrogen dioxide (NO2) in the Helsinki Metropolitan Area for 1996 and 1997. The hourly concentration time series were computed on a receptor grid containing approximately 10 000 receptor points. The receptor point network covers the whole area, the largest grid intervals being equal to 500 m. A more densely spaced grid was applied in the Helsinki city centre, the grid interval there being 100 m. In the vicinity of major roads in the area, the smallest grid interval was equal to 50 m. A variable receptor grid is necessary in order to evaluate interpolated concentrations from the computed data with adequate accuracy (Karppinen et al., 2000a).
2.3.2. The structure of the EXPAND model 2.3.2.1. Evaluation of the population activity. We combined the home coordinates together with the information on the number of inhabitants at each home location (for the data of 1997) and the time spent at home during the day. Correspondingly, we combined the workplace coordinates, the number of personnel (the data of 1998) and the time spent at the workplace. The resulting data sets correspond to different years, as part of the information was not available; however, considering the whole of the Helsinki Metropolitan Area, only minor changes took place in the workplace and personnel information from 1997 to 1998. We have utilised the time-microenvironment activities produced in the EXPOLIS study, which is applicable for the adult population from 25 to 55 yr of age. According to official statistics, this age interval contains 82% of the working age population of Helsinki (City of Helsinki, 1999). In the numerical computations of this study, we have considered this particular age group of the adult population; the number of persons obtained from the data containing the whole population
2.3.2.2. Numerical interpolation of concentrations and the computation of exposure. The results of the dispersion model computations are produced on a variable receptor grid; the grid interval ranges from 50 to 500 m. However, in order to combine this information with the population activities data, these concentrations have to be interpolated on to a rectangular grid network with a constant spacing. The model performs this interpolation; in the computations for this study, we have selected a grid size of 100 m 100 m. The model also transforms the data regarding population activities (number of persons hour) to the same rectangular square grid, utilising the geographic information on the locations of homes, working places and other activities. The program subsequently combines the interpolated concentration data with the data containing the population activities. For a selected specific time interval, this combination simply means a multiplication of the time-averaged concentration value (mg m3) and the corresponding population activity value, for each grid cell. The resulting value is called the population exposure, corresponding to a specific pollutant, type of activity and integration time. Finally, all the results (concentrations, activities and exposures) are converted into a numerical format that can be transferred directly to the GIS MapInfo system. The GIS system is subsequently utilised in the postprocessing and visualisation of this information.
3. Model evaluation and numerical results 3.1. Evaluation of the predicted concentrations against experimental data We have performed a statistical analysis to determine the agreement between predicted and measured hourly time series of concentrations of NOx and NO2 at a total of ten urban and suburban monitoring sites in the Helsinki Metropolitan Area in 1996 and 1997. The agreement between the measured and predicted data sets was good, as indicated by the computed statistical
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
2114
parameters. For instance, the index of agreement values of the predicted and measured time series of the NO2 concentrations varied between 0.65 and 0.82. The index of agreement is defined as follows (Willmott, 1981): IA ¼ 1
ðCP CO Þ2 ½jCP CO j þ jCO CO j 2
;
ð1Þ
where CP and CO are the predicted and observed concentrations, respectively. The overbar refers to the average over all values. IA is a measure of the correlation of the predicted and measured time series of concentrations. The values of IA vary from 0.4 (random number distribution) to 1.0 (perfect agreement between the observed and predicted values) (Karppinen et al., 2000b). For a more detailed discussion of these comparisons, the reader is referred to Kousa et al. (2001b). The concentrations of NO2 were also computed in the coordinates in which the residential outdoor concentrations (n ¼ 162) were measured in the EXPOLIS study (in 1996 and 1997). We subsequently evaluated the statistical correlation of two concentration time-series that contained 48 hourly averaged concentrations. The correlation between these predicted and measured concentrations was fairly good; the Pearson correlation coefficient was equal to 0.71 and the p-value was lower than 0.001. 3.2. Numerical results In order to illustrate the diurnal variation of average population exposure, we have classified each day into five time periods; these are called, for simplicity, morning (defined as the period 6 a.m.–9 a.m.), day (9 a.m.–3 p.m.), afternoon (3 p.m.–6 p.m.), evening (6 p.m.–10 p.m.) and night (10 p.m.–6 a.m.). The morning and afternoon time periods correspond to the most busy commuting periods, while the other three time periods have been selected to represent times at which people are mostly in one place, either at home (‘‘evening’’ and ‘‘night’’ periods) or at the workplace (‘‘day’’ period). As an example, we present the results of the afternoon period in March 1996; the measured NO2 concentrations were highest during that particular month in 1996 and 1997 (Kousa et al., 2001b). The computed results of the ambient NO2 concentrations, and the spatial and temporal population density and average exposure have been presented in Figs. 3(a)–(c). The concentration values are monthly averages in the afternoon time-period. The activity and average exposure values have also been presented as average values in this time period; the corresponding numerical values of activity and average exposure in each grid cell therefore refer to the number of people (density of population),
and the concentration times the number of people, respectively. This normalisation in terms of time makes it possible to compare population densities and average exposures during different time periods with each other. The time spent and average exposure in vehicular traffic is also included in the results. The Helsinki Metropolitan Area is located by the Baltic Sea and the centre of Helsinki is on a peninsula that is located approximately in the middle of the southern part of each figure. The Helsinki Metropolitan Area comprises of four cities: Helsinki, Espoo, Vantaa and Kauniainen; the total population is approximately 850 000. The NO2 concentrations are highest in the vicinity of the main roads and streets, and in the centre of Helsinki (Fig. 3a). The figure shows the distinct influence of ring road number 1 (situated at a distance of about 8 km from the city centre), the major roads leading to the Helsinki city centre, and the junctions of major roads and streets. As expected, the population density values are highest in the city centre (Fig. 3b). Results also show that there are elevated levels of population density in the vicinity of the district centres of the other major cities in the area (Espoo and Vantaa), and in the vicinity of major roads and streets; the latter is clearly due to increased vehicular traffic during the afternoon rush hours. As population average exposure is a combination of both the concentration and activity (or population density) values, it exhibits characteristics of both spatial distributions, i.e., elevated values in the Helsinki city centre, along major roads and streets, and in the vicinity of urban district centres (Fig. 3c). The predicted distributions of the population in terms of the NO2 concentrations are presented in Figs. 4(a)–(e) for the five selected diurnal time periods. The total area below each distribution is directly proportional to the number of people times concentration that are located in a specific class of environment (e.g., at home) during a specific time interval (e.g., the morning hour). Clearly, in evaluating the average exposure of the population, one also has to allow for the location of each distribution in terms of the concentration and the exposure time. Such figures can be used to analyse the relative importance of the average exposure in various microenvironments, and the diurnal variation of the population average exposure. During each time-period, the location of the peak of the distributions corresponding to home, work and other activities are on the average of the same order of magnitude. However, the number of people in the home and work environments is substantially larger on the average, compared with those in other activities (the same result can be seen from Fig. 2). It can therefore be concluded that the average exposure to NO2 at home and in the workplace is substantially more important than that in ‘‘other’’ activities.
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
2115
Fig. 3. (a)–(c) The predicted ambient air concentrations of NO2 (mg m3), the density of population (persons) and the average exposure of the population to NO2 concentrations (mg m3 persons), evaluated for the afternoon time period, as an average value in March 1996. The values in brackets in the legend refer to the number of square kilometres with the concentration, population density or average exposure in the selected range. The grid size is 100 m 100 m, the size of the depicted area is 23 km 16 km, and the solid black line is the coastline.
The distributions corresponding to traffic are spread over a wide range of concentrations for all time-periods. Although people are exposed to the highest NO2 concentrations in traffic, the time spent in traffic is relatively short, compared with that spent at home and in work environments. The average exposure in traffic is clearly lower, compared with that at home and in workplace environments. The highest concentrations in traffic occur during the afternoon rush hours, and fairly
high values also occur during the evening and morning periods.
4. Conclusions This paper presents a mathematical exposure model, which combines predicted concentrations, the location of the population, and the time spent at home, in the
2116
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
workplace, in traffic and at other places of activity. The main objective was to evaluate the spatial and temporal variation of average exposure of the whole urban population to ambient air pollution in different microenvironments. The model includes a treatment of all of the most important categories of population activity, including also the exposure in vehicular traffic. Results have been presented to illustrate the application of the model; these include the concentrations, population densities and average exposure in the Helsinki Metropolitan Area. The model presented here evaluates the hourly average population exposure to outdoor air concentrations of NO2. Clearly, personal, and temporally shorter term exposures can be substantially higher or lower, compared with the values computed using this model. We assumed that I/O ratio for NO2 concentrations was equal to unity. In some previous studies, simple empirically determined coefficients have been utilised for estimating indoor air concentrations, based on the corresponding ambient air values (e.g., Jensen, 1999). In the EXPOLIS study, the ratios of the indoor and outdoor concentrations (I/O) for nitrogen dioxide were on the average equal to 0.76. However, the ratio of indoor and outdoor air concentrations is dependent on several factors, including outdoor to indoor transport (ventilation and filtration), and sources and sinks indoors. Clearly, this ratio is dependent on the particular pollutant to be considered; in the case of particulate matter, it is also dependent on the particle size. Various models have been presented for evaluating the ratio of pollutant concentrations in indoor air; Kulmala et al. (1999) and Dimitroulopoulou et al. (2001) have presented such models in the case of particulate matter and nitrogen dioxide, respectively. However, such models require detailed input information concerning, e.g., the building structure, and ventilation and filtration systems. At present, there is insufficient data available for the application of such models over the whole of the metropolitan area. On the average, the indoor air concentrations of NO2 are lower than the corresponding outdoor air concentrations in the Helsinki Metropolitan Area, as there are very few homes and workplaces with gas appliances. We have evaluated the exposure to outdoor concentrations of NO2; however, people spend most of their time indoors (the average value in the area considered is equal to 87%). Despite this, we can conclude that the
2117
average exposure of population to NO2 at home and in the workplace is substantially more important than that in traffic and other activities. People are exposed to the highest NO2 concentrations in traffic, but the corresponding time is relatively short, compared with that spent at home and work environments. Clearly, the dispersion modelling approach has certain inherent limitations: Gaussian dispersion modelling does not allow for the detailed structure of buildings and obstacles. However, the terrain in the area is fairly flat and the average height of the buildings is low (most buildings are lower than 15–20 m). The computations considered here correspond to urban background concentrations, averaged over the length scale of the grid size applied. There is only a moderate number of street canyons in the area. Inside a street canyon, the actual concentrations can vary substantially on a scale of tens of metres; in such locations, nested street canyon dispersion model computations should be used (e.g., Kukkonen et al., 2001b). However, at present it is not feasible to conduct such an analysis for an extensive area including an agglomeration of several cities, due both to a lack of the relevant geometric building data, and to restrictions on computer resources. This model has been designed to be utilised by municipal authorities in evaluating the impacts of traffic planning and land use scenarios. For instance, the transportation system plan (TSP) for the Helsinki Metropolitan Area is revised at four-yearly intervals; the previous version has been described by the Helsinki Metropolitan Area Board (1999). This model will be used to evaluate the impacts of different scenarios in the new revision of the TSP. We also aim to estimate the adverse health effects caused to the population by air pollution and to simulate the burden of disease for each of the TSP scenarios.
Acknowledgements This study has been part of the National Research Programme on Environmental Health (SYTTY), the Health Promotion Research Programme (TERVE) and the research programme EUROTRAC-2, SATURN, ‘‘Studying atmospheric pollution in urban areas’’, 1998– 2002. The funding from the Academy of Finland for all of these studies is gratefully acknowledged. In Helsinki, the EXPOLIS study has been supported by EU Contract
Fig. 4. (a)–(e) Predicted distribution of the number of people (N) against the concentration of NO2 during the morning, day, afternoon, evening and night periods in the Helsinki Metropolitan Area in March, 1996. The vertical scale is different for the number of people at home and in the workplace (left-hand side scales), compared with the number of people in traffic and other activities (righthand side scales). The vertical scale for the number of people at home and in the workplace is also different in the night period, compared with those in the other panels.
2118
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
N ENV4-CT96-0202 (DG 12 -DTEE), Academy of Finland Contract N 36586 and intramural funding by the National Public Health Institute of Finland.
References Alm, S., Mukala, K., Pasanen, P., Tiittanen, P., Ruuskanen, J., Tuomisto, J., Jantunen, M.J., 1998. Personal NO2 exposures of preschool children in Helsinki. Journal of Exposure Analysis and Environmental Epidemiology 8, 79–100. Bohler, T., Riise, A., 1997. Using the air quality assessment system AirQUIS in modelling the population’s exposure to traffic induced air pollution. Fourth International Scientific Symposium on Transport and Air Pollution, 9–13 June 1997, Avignon, France. Brunekreef, B., Lumens, M., Hoek, G., Hofschreuder, P., Biersteker, K., 1989. Pulmonary function changes associated with an air pollution episode in January 1987. Journal of Air Pollution Control Association 39, 1444–1447. City of Helsinki, Urban Facts, 1999. Statistical yearbook of the City of Helsinki. Gummerus Kirjapaino OY, Jyv.askyl.a. Clench-Aas, J., Bartanova, A., Bohler, T., Gronskei, K.E., Sivertsen, B., Larsssen, S., 1999. Air pollution exposure monitoring and estimating part I: integrated air quality monitoring system. Journal of Environmental Monitoring 1, 313–319. Dimitroulopoulou, C., Ashmore, M.R., Byrne, M.A., Kinnersley, R.P., 2001. Modelling of indoor exposure to nitrogen dioxide in the UK. Atmospheric Environment 35, 269–279. Freijer, J.I., Bloemen, H.J.Th., De Loos, S., Marra, M., Rombout, P.J.A., Steentjes, G.M., Van Veen, M.P., 1997. AirPEx: Air Pollution Exposure Model. RIVM, The Netherlands. Report No. 650010 005. . H.arkonen, J., Valkonen, E., Kukkonen, J., Rantakrans, E., Jalkanen, L., Lahtinen, K., 1995. An operational dispersion model for predicting pollution from a road. International Journal of Environment and Pollution 5 (4–6), 602–610. . H.arkonen, J., Walden, J., Kukkonen, J., 1997. Comparison of model predictions and measurements near a major road in an urban area. International Journal of Environment and Pollution 8 (3–6), 761–768. Helsinki Metropolitan Area Board, 1999. Helsinki Metropolitan area transportation system plan, PLJ 1998. Helsinki Metropolitan Area Series A 1999:4. YTV Helsinki Metropolitan Area Council, Helsinki. INRO, 1994. EMME/2 User’s manual. INRO Consultants Inc., Montreal, Canada. . Jantunen, M.J., H.anninen, O., Katsoyanni, K., Knoppel, H., Kunzli, . N., Lebret, E., Maroni, M., Saarela, K., Sram, R., Zmirou, D., 1998. Air pollution exposure in European Cities: the Expolis study. Journal of Exposure Analysis and Environmental Epidemiology 8, 495–518. . Jantunen, M.J., Katsoyanni, K., Knoppel, H., Kunzli, . N., Lebret, E., Maroni, M., Saarela, K., Sram, R., Zmirou, D. 1999. Final report: Air Pollution Exposure in European Cities: the Expolis study, Publications of the National Public Health Insititute, KTL B16/1999, Kuopio.
Jensen, S.S., 1999. A geographic approach to modelling human exposure to traffic air pollution using GIS. Ph. D. Thesis. National Environmental Research Institute, Denmark. . Jonson, T., Johansson, C., Hadenius, A., Johansson, P.A., 1999. Shape. The Stockholm study on health effects of air pollution and their economic consequences Part I: NO2 and particulate matter in Stockholm—concentrations and population exposure, Swedish National Road Administration No. 1999:41, Stockholm, Sweden. Karppinen, A., Kukkonen, J., Elol.ahde, T., Konttinen, M., Koskentalo, T., Rantakrans, E., 2000a. A modelling system for predicting urban air pollution, model description and applications in the Helsinki metropolitan area. Atmospheric Environment 34, 3723–3733. Karppinen, A., Kukkonen, J., Elol.ahde, T., Konttinen, M., Koskentalo, T., 2000b. A modelling system for predicting urban air pollution, comparison of model predictions with the data of an urban measurement network. Atmospheric Environment 34, 3735–3743. Karppinen, A., Joffre, S.M., Kukkonen, J., 2000c. The refinement of a meteorological preprocessor for the urban environment. International Journal of Environment and Pollution 14 (1-6), 565–572. Kousa, A., Monn, C., Rotko, T., Alm, S., Jantunen, M.J., 2001a. Personal exposures to NO2 in the EXPOLIS-study: relation to residential indoor, outdoor and workplace concentrations in Basel, Helsinki and Prague. Atmospheric Environment 35, 3405–3412. Kousa, A., Kukkonen, J., Karppinen, A., Aarnio, P., Koskentalo, T., 2001b. Statistical and diagnostic evaluation of a new-generation urban dispersion modelling system against an extensive dataset in the Helsinki Area. Atmospheric Environment 35 (27), 4617–4628. . Kukkonen, J., H.arkonen, J., Walden, J., Karppinen, A., Lusa, K., 2001a. Evaluation of the dispersion model CAR-FMI against data from a measurement campaign near a major road. Atmospheric Environment 35, 949–960. Kukkonen, J., Valkonen, E., Walden, J., Koskentalo, T., Aarnio, P., Karppinen, A., Berkowicz, R., Kartastenp.aa. , R., 2001b. A measurement campaign in a street canyon in Helsinki and comparison of results with predictions of the OSPM model. Atmospheric Environment 35, 231–243. Kulmala, M., Asmi, A., Pirjola, L., 1999. Indoor air aerosol model: the effect of outdoor air, filtration and ventilation on indoor concentrations. Atmospheric Environment 33, 2133–2144. M.akel.a, K., 1999. LIPASTO, Calculation system for traffic emissions and energy consumption. http://www.vtt.fi/rte/ lipastoe/index.htm. VTT Communities and Infrastructure, Espoo. Mukala, K., Alm, S., Tiittanen, P., Salonen, R.O., Jantunen, M.J., Pekkanen, J., 2000. Nitrogen dioxide exposure assessment and cough among preschool children. Archives of Environmental Health 55, 431–438. Ott, W.R., 1985. Total human exposure. Environmental Science and Technology 19 (10), 880–886. . Ottl, D., Kukkonen, J., Almbauer, R.A., Sturm, P.J., Pohjola, M., H.ark.onen, J., 2001. Evaluation of a Gaussian and a Lagrangian model against a roadside dataset, with focus on low wind speed conditions. Atmospheric Environment 35, 2123–2132.
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119 Ryan, P.B., 1991. An overview of human exposure modeling. Journal of Exposure Assessment and Environmental Epidemiology 1, 453–473. Schwartz, J., Spix, C., Wichmann, H.E., Malin, E., 1991. Air pollution and acute respiratory illness in five German communities. Environmental Respiratory 56, 1–14. Sexton, K., Ryan, P.B., 1988. Assessment of human exposure to air pollution: methods, measurements, and models. In:
2119
Watson, A.Y., Bates, R.R., Kennedy, D. (Eds.), Air Pollution, the Automobile, and Public Health. National Academic Press, Washington, D.C., pp. 207–238. Willmott, C.J., 1981. On the validation of models. Physical Geography 2, 184–194. World Health Organization (WHO), 2000. Air quality guidelines for Europe, 2nd Edition. WHO regional publications, European Series, 91, Copenhagen, Denmark.
A model for evaluating the population exposure to ambient air pollution in an urban area Anu Kousaa,b,*, Jaakko Kukkonenc, Ari Karppinenc, P.aivi Aarnioa, Tarja Koskentaloa a
b
Helsinki Metropolitan Area Council (YTV), Opastinsilta 6 A, 00520 Helsinki, Finland Department of Environmental Hygiene, National Public Health Institute (KTL), P.O. Box 95, 70701 Kuopio, Finland c Finnish Meteorological Institute (FMI), Air Quality Research, Sahaajankatu 20 E, 00810 Helsinki, Finland Received 21 September 2001; received in revised form 20 February 2002; accepted 27 February 2002
Abstract A mathematical model is presented for the determination of human exposure to ambient air pollution in an urban area. The main objective was to evaluate the spatial and temporal variation of average exposure of the urban population to ambient air pollution in different microenvironments with reasonable accuracy, instead of analysing in detail personal exposures for specific individuals. We have utilised a previously developed modelling system for predicting the traffic flows and emissions, emissions originating from stationary sources, and atmospheric dispersion of pollution in an urban area. A model was developed for combining the predicted concentrations, information on people’s activities (such as the time spent at home, in the workplace and at other places of activity during the day) and location of the population. Time-microenvironment activity data for the working-age population was obtained from the EXPOLIS study (air pollution distributions within adult urban populations in Europe). Information on the location of homes and workplaces was obtained from local municipalities. The concentrations of NO2 were modelled over the Helsinki Metropolitan Area for 1996 and 1997. The computed results were processed and visualised using the geographical information system (GIS) MapInfo. The utilisation of the modelling system has been illustrated by presenting numerical results for the Helsinki Metropolitan Area. The results show the spatial and temporal (diurnal) variation of the ambient air NO2 concentrations, the population density and the corresponding average exposure. The model developed has been designed to be utilised by municipal authorities in urban planning, e.g., for evaluating the impacts of traffic planning and land use scenarios. r 2002 Elsevier Science Ltd. All rights reserved. Keywords: Exposure model; Population exposure; Nitrogen dioxide; EXPOLIS; GIS
1. Introduction Harmful effects have been associated with NO2 exposure levels that are below the ambient air quality guideline levels set by the World Health Organization (WHO, 2000) (e.g., Brunekreef et al., 1989; Schwartz et al., 1991; Mukala et al., 2000). Most of the *Corresponding author. Helsinki Metropolitan Area Council (YTV), Opastinsilta 6 A, 00520 Helsinki, Finland. Fax: +3589-1561-334. E-mail address: anu.kousa@ytv.fi (A. Kousa).
epidemiological studies (exception, Mukala et al., 2000) have been based on nitrogen dioxide concentrations at fixed ambient air quality monitoring sites. However, the measurement data from these stations do not necessarily represent areas beyond their immediate vicinity, as the concentrations of pollutants in urban areas may vary by orders of magnitude on spatial scales varying from tens to hundreds of metres. For example, Alm et al. (1998) addressed the measured personal NO2 exposures of pre-school children in Helsinki in 1991. The concentration data obtained from stationary monitoring sites explained only a minor fraction of the personal
1352-2310/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S 1 3 5 2 - 2 3 1 0 ( 0 2 ) 0 0 2 2 8 - 5
2110
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
weekly NO2 exposure variations of children, as shown by a statistical regression analysis (the value of the corresponding coefficient of determination, r2 ; was 0.37). However, the above-mentioned NO2 exposure variations were statistically strongly associated with the NO2 concentrations that were measured inside and outside the nurseries (r2 ¼ 0:88 and 0.86, respectively). A similar result was obtained in the EXPOLIS study that addressed measured personal exposures of adult inhabitants in the Helsinki Metropolitan Area in 1996 and 1997 (Jantunen et al., 1999; Kousa et al., 2001a). The statistical associations of these exposures and the measured residential outdoor and indoor, and workplace indoor concentrations varied from fairly weak to reasonably strong (r2 ¼ 0:40; 0.45 and 0.55, respectively). In evaluating the exposure of the population to air pollution, there is a need to evaluate spatial concentration distributions produced, for example, by atmospheric dispersion modelling systems or a set of onsite measurements. Clearly, the modelling systems need first to be thoroughly evaluated against experimental data. Models for evaluating exposure to air pollutants have been classified as statistical, mathematical and mathematical-stochastic (modified from Ryan, 1991). The statistical approach involves the statistical determination of the measured exposures in terms of the factors that are assumed to influence these exposures. Mathematical modelling involves application of emission inventories, combined with atmospheric dispersion and population activity modelling. The stochastic approach attempts to include a treatment of the inherent uncertainties in the model, e.g., those caused by the turbulent nature of atmospheric flow. Mathematical modelling utilising emission and dispersion models is also called deterministic modelling. Source apportionment methods can be used in order to analyse the contribution of various emission categories to the total human exposure. Various exposure models have been developed in order to support health risk assessment and management. Examples of such models are ‘‘American national air quality standards exposure model’’ (NEM, Sexton and Ryan, 1988), ‘‘air pollution exposure model’’ (AirPEx, Freijer et al., 1997) and ‘‘simulation of human air pollution exposures’’ (SHAPE, Ott, 1985). All of these models are based on the same concept: the concentrations in various microenvironments are weighted with the fraction of time that people spend in these microenvironments. However, these models do not take advantage of administrative databases containing population statistics or geographical information system’s. Most recently, mathematical exposure models have been presented by Bohler and Riise (1997), Clench-Aas et al. (1999), Jensen (1999), Johansson et al. (1999) and
Dimitroulopoulou et al. (2001). The air quality management system AirQUIS has been extended to include a fairly simple exposure model (Bohler and Riise, 1997; Clench-Aas et al., 1999). The model database contains a building register that includes information on building coordinates, number of floors and number of residents. The coordinates of the buildings can then be defined as receptor points in atmospheric dispersion modelling. The exposure modelling part of the system associates a number of persons with each individual building, and combines this information with the computed ambient air concentrations, utilising a geographical information system (GIS). The model presented by Jensen (1999) is based on the utilisation of traffic flow computations and the operational street pollution model (OSPM) for evaluating outdoor air pollutant concentrations. The activity patterns of the population have been evaluated using various administrative databases and standardised timeactivity profiles. The modelling system utilises a GIS in combining and processing the concentration and population activity data. The model was applied to evaluate population exposure in one specific municipality in Denmark. An advantage of this system is that the application of a street canyon dispersion model facilitates a fine-scale spatial resolution (metres or tens of metres). However, the complexity of the model computations requires a limitation in the spatial domain to be evaluated; the computations for a more extensive area, such as that of a major city, were not possible. Johansson et al. (1999) evaluated population exposure to NO2 and particulate matter (PM) within the SHAPE study (Stockholm study on health effects of air pollution and their economic consequences). They used a multiple-source emission and Gaussian dispersion modelling system to predict the spatial distributions of NO2 and PM10 concentrations in the municipality of Stockholm. The population locations were obtained from official national data sets. However, a treatment of the chemical reactions of nitrogen oxides was not included; instead, a purely empirical relation was utilised in computing the NO2 concentrations. The model applied was not combined with a GIS system. Dimitroulopoulou et al. (2001) have evaluated personal NO2 exposure by using a deterministic model. Their model can distinguish three types of exposures: outdoor air exposure, indoor air exposure resulting from penetration of outdoor air, and indoor air exposure resulting from indoor sources. The model evaluates personal exposures, by combining the information on the movements of typical individuals in various microenvironments and the modelled microenvironment concentrations. The outdoor concentrations of NO2 were evaluated based on the measurement data from urban background sites. However, the model applied was not combined with a GIS system.
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
We have previously developed an integrated modelling system for predicting the traffic flows and emissions, emissions from stationary sources, and atmospheric dispersion of pollution in an urban area; for a description of this system and its evaluation against the data from an urban monitoring network, the reader is referred to Karppinen et al. (2000a,b). The modelling system developed has been utilised to evaluate environmental impacts of traffic planning and land use scenarios in the area. Emissions, NO2 concentrations and preliminary potential NO2 exposures have been assessed in the transportation system plan for the Helsinki Metropolitan Area (Helsinki Metropolitan Area Board, 1999). The main purpose of this paper is to present an extension of the above-mentioned mathematical modelling system to allow for the average exposure of the human population to ambient air pollution. Clearly, the model cannot evaluate the personal exposure of each member of the population. The internal variation of personal exposures is larger (i.e., these can be substantially higher or lower), compared with that of average population exposures. Further, the model presented here considers only hourly averaged values; the short-term peak exposures can be drastically higher. In principle, the model is applicable for any pollutant, if the required spatial concentration distributions are available. The predicted NO2 concentrations utilised in this study, computed for the years 1996 and 1997, have also been compared with the ambient NO2 concentrations measured in an urban air quality monitoring network (Kousa et al., 2001b). These predicted concentrations have also been compared with the measured residential outdoor air concentrations determined in the EXPOLIS study.
2. Materials and methods 2.1. Emissions and meteorological data We have updated the previously conducted emission inventory of NOx in the Helsinki Metropolitan Area (Karppinen et al., 2000a) for the years 1996 and 1997. The inventory includes the emissions from various mobile sources (road traffic, harbours and marine traffic, and aviation) and stationary sources (power plants, other point sources and residential heating). The traffic flows and average travel speeds were computed using the EMME/2 transportation planning system (INRO, 1994); vehicular emissions were evaluated using the EMME/2 system and emission factors that have been evaluated for this area by the Helsinki Metropolitan Area Council. The model allows for the diurnal and weekly variations both in traffic volumes and speeds, as well as in traffic emissions. Stationary sources
2111
are considered as point or area sources. The computations included approximately 5000 line sources, 169 point sources, area sources and the regional background concentrations. We used the meteorological database of the Finnish Meteorological Institute, which contains weather and sounding observations. A combination of data from the stations at Helsinki–Vantaa airport (about 15 km north of Helsinki town centre) and Helsinki–Isosaari (an island about 20 km south of Helsinki) were employed. The mixing height of the atmospheric boundary layer was evaluated using a meteorological pre-processor, based on the sounding observations made at Jokioinen (90 km northwest of Helsinki) and on routine meteorological observations. 2.2. Atmospheric dispersion modelling The dispersion modelling is based on a combined application of the road network dispersion model CAR. FMI (e.g., H.arkonen et al., 1995), applied for evaluating the dispersion of pollution originating from vehicular traffic, and the urban dispersion modelling system UDM-FMI (Karppinen et al., 2000a), for evaluating the dispersion from stationary sources. The predictions of the CAR-FMI model have been compared with the results of two measurement campaigns conducted near . major roads (i) in a suburban area (e.g., H.arkonen et al., 1997) and (ii) in a rural area (Kukkonen et al., 2001a; . Ottl et al., 2001). However, the CAR-FMI model is a finite line source model that does not explicitly take into account the influence of buildings and other obstacles on dispersion. The relevant meteorological parameters for the models are evaluated using data produced by a meteorological pre-processing model that has been specifically adapted for the urban environment (Karppinen et al., 2000c). The modelling system includes a statistical and graphical analysis of the computed time series of concentrations. The modelling system also comprises a method that allows for the chemical interaction of pollutants (including the basic reactions of nitrogen oxides and ozone), originating from a large number of urban sources (Karppinen et al., 2000a). 2.3. Exposure modelling The main objective was to evaluate the average exposure of the population with reasonable accuracy, instead of the personal exposures of specific individuals. Based on the measured results of the EXPOLIS project, the ratios of the indoor and outdoor concentrations (I/O) for nitrogen dioxide were on the average 0.76. However, the detailed value of this ratio depends on numerous factors, especially on the occurrence of indoor sources, such as tobacco smoke and gas stoves. In this
2112
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
study, we assumed for simplicity that residential and workplace indoor concentrations of nitrogen dioxide were the same as the corresponding outdoor concentrations. The part of the modelling system that evaluates the average exposure of the population to air pollution has been named EXPAND (‘‘EXPosure to Air pollution, especially to Nitrogen Dioxide and particulate matter’’). A schematic diagram of the EXPAND model is presented in Fig. 1. The model utilises as input values (i) data on the spatial location of the population, (ii) time-microenvironment activity data and (iii) computed spatial pollutant concentration distributions. The computational modules contain a separate computer program for combining and processing the data, and the GIS MapInfo. The model yields as output the spatial distribution of the pollutant concentration, the activity (density) of population and the exposure of population to ambient air pollution in the selected grid. 2.3.1. Description of the input data sets 2.3.1.1. Location of the population. The information on the location of the population comprises residential and workplace coordinates, the number of people living or working at these places, and traffic flow information concerning the road and street network. We also evaluated the location and number of people involved in other activities outside their homes, besides working (e.g., recreational activities). We have utilised a dataset that is collected annually by the municipalities in the Helsinki Metropolitan Area and disseminated in compact disk format. This dataset also contains various data on the enterprises
Fig. 1. A schematic presentation of the exposure model.
and agencies located in the area. The dataset provides geographic information on the total number of people living in a particular building or working at a particular workplace. However, this information does not identify individual persons, and there is no information, e.g., on the age of personnel at each specific workplace. The information on other locations, besides residential and workplace sites, was produced using questionnaire information gathered previously by the Helsinki Metropolitan Area Council. We used this information together with information on the amount of journeys attracted by various places (for instance, shops and homes), in order to evaluate other activities. The location of people in vehicular traffic was evaluated using the computed traffic flow information; this information is available separately for buses and cars for each street section on an hourly basis. The computed traffic flow information includes the number of vehicles, their average speeds and passenger mileage.
2.3.1.2. Time-activity data. We utilised the time-microenvironment activity data produced for Helsinki in the EXPOLIS study. EXPOLIS is a European multicentre, multicomponent exposure study; the centres include Athens, Basel, Grenoble, Helsinki, Milan and Prague. The EXPOLIS study has focused on European adult urban populations, from 25 to 55 yr of age, and their personal air pollution exposures to NO2, CO, PM2.5 and 30 volatile organic compounds (VOCs). The EXPOLIS measurements have been conducted during working weeks. In the Helsinki Metropolitan Area, time-microenvironment activity data have been collected for 435 randomly selected working age citizens; for a sub-set of this population group comprising of 201 citizens, microenvironment concentrations and personal exposures have been measured, both of these during the period from October 1996 to December 1997. The timemicroenvironment activity data have been recorded at the intervals of 15 min. For a more detailed discussion, the reader is referred to Jantunen et al. (1998, 1999). The time-activity of the population was divided into four categories: home, workplace, traffic and other activities (e.g., recreational activities). The diurnal variation of population activities in various microenvironments in the Helsinki Metropolitan Area is presented in Fig. 2. In the data presented in this figure, we have combined indoor and outdoor time-activity in each microenvironment. People spend on average 87% of their time indoors in the area considered. As expected, on average people spend most of their time at home and work environments. In the late afternoon and early evening (from approximately 4 p.m.to 8 p.m.), people spend a substantial fraction of their time in traffic and in ‘‘other’’ activities (these include shopping and various recreational activities, and the related travelling).
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
2113
was therefore multiplied by the above-mentioned fraction (0.82). The population activities at other locations (other than for home and workplaces) were also evaluated. The number of persons carried by vehicular traffic was evaluated based on the predicted traffic flow information. In the case of buses, the number of persons and the time they spent in each street section is predicted directly by the EMME/2 model. In the case of private cars, the EMME/2 model predicts the number of cars; we assumed that the number of passengers in each car is equal to the average value in the area i.e., 1.2 (M.akel.a, 1999). Fig. 2. The diurnal variation of the activity of the population in the Helsinki Metropolitan Area during the period from October 1996 to December 1997, evaluated in the EXPOLIS study.
2.3.1.3. Predicted concentrations. We computed the concentrations of nitrogen oxides (NOx) and nitrogen dioxide (NO2) in the Helsinki Metropolitan Area for 1996 and 1997. The hourly concentration time series were computed on a receptor grid containing approximately 10 000 receptor points. The receptor point network covers the whole area, the largest grid intervals being equal to 500 m. A more densely spaced grid was applied in the Helsinki city centre, the grid interval there being 100 m. In the vicinity of major roads in the area, the smallest grid interval was equal to 50 m. A variable receptor grid is necessary in order to evaluate interpolated concentrations from the computed data with adequate accuracy (Karppinen et al., 2000a).
2.3.2. The structure of the EXPAND model 2.3.2.1. Evaluation of the population activity. We combined the home coordinates together with the information on the number of inhabitants at each home location (for the data of 1997) and the time spent at home during the day. Correspondingly, we combined the workplace coordinates, the number of personnel (the data of 1998) and the time spent at the workplace. The resulting data sets correspond to different years, as part of the information was not available; however, considering the whole of the Helsinki Metropolitan Area, only minor changes took place in the workplace and personnel information from 1997 to 1998. We have utilised the time-microenvironment activities produced in the EXPOLIS study, which is applicable for the adult population from 25 to 55 yr of age. According to official statistics, this age interval contains 82% of the working age population of Helsinki (City of Helsinki, 1999). In the numerical computations of this study, we have considered this particular age group of the adult population; the number of persons obtained from the data containing the whole population
2.3.2.2. Numerical interpolation of concentrations and the computation of exposure. The results of the dispersion model computations are produced on a variable receptor grid; the grid interval ranges from 50 to 500 m. However, in order to combine this information with the population activities data, these concentrations have to be interpolated on to a rectangular grid network with a constant spacing. The model performs this interpolation; in the computations for this study, we have selected a grid size of 100 m 100 m. The model also transforms the data regarding population activities (number of persons hour) to the same rectangular square grid, utilising the geographic information on the locations of homes, working places and other activities. The program subsequently combines the interpolated concentration data with the data containing the population activities. For a selected specific time interval, this combination simply means a multiplication of the time-averaged concentration value (mg m3) and the corresponding population activity value, for each grid cell. The resulting value is called the population exposure, corresponding to a specific pollutant, type of activity and integration time. Finally, all the results (concentrations, activities and exposures) are converted into a numerical format that can be transferred directly to the GIS MapInfo system. The GIS system is subsequently utilised in the postprocessing and visualisation of this information.
3. Model evaluation and numerical results 3.1. Evaluation of the predicted concentrations against experimental data We have performed a statistical analysis to determine the agreement between predicted and measured hourly time series of concentrations of NOx and NO2 at a total of ten urban and suburban monitoring sites in the Helsinki Metropolitan Area in 1996 and 1997. The agreement between the measured and predicted data sets was good, as indicated by the computed statistical
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
2114
parameters. For instance, the index of agreement values of the predicted and measured time series of the NO2 concentrations varied between 0.65 and 0.82. The index of agreement is defined as follows (Willmott, 1981): IA ¼ 1
ðCP CO Þ2 ½jCP CO j þ jCO CO j 2
;
ð1Þ
where CP and CO are the predicted and observed concentrations, respectively. The overbar refers to the average over all values. IA is a measure of the correlation of the predicted and measured time series of concentrations. The values of IA vary from 0.4 (random number distribution) to 1.0 (perfect agreement between the observed and predicted values) (Karppinen et al., 2000b). For a more detailed discussion of these comparisons, the reader is referred to Kousa et al. (2001b). The concentrations of NO2 were also computed in the coordinates in which the residential outdoor concentrations (n ¼ 162) were measured in the EXPOLIS study (in 1996 and 1997). We subsequently evaluated the statistical correlation of two concentration time-series that contained 48 hourly averaged concentrations. The correlation between these predicted and measured concentrations was fairly good; the Pearson correlation coefficient was equal to 0.71 and the p-value was lower than 0.001. 3.2. Numerical results In order to illustrate the diurnal variation of average population exposure, we have classified each day into five time periods; these are called, for simplicity, morning (defined as the period 6 a.m.–9 a.m.), day (9 a.m.–3 p.m.), afternoon (3 p.m.–6 p.m.), evening (6 p.m.–10 p.m.) and night (10 p.m.–6 a.m.). The morning and afternoon time periods correspond to the most busy commuting periods, while the other three time periods have been selected to represent times at which people are mostly in one place, either at home (‘‘evening’’ and ‘‘night’’ periods) or at the workplace (‘‘day’’ period). As an example, we present the results of the afternoon period in March 1996; the measured NO2 concentrations were highest during that particular month in 1996 and 1997 (Kousa et al., 2001b). The computed results of the ambient NO2 concentrations, and the spatial and temporal population density and average exposure have been presented in Figs. 3(a)–(c). The concentration values are monthly averages in the afternoon time-period. The activity and average exposure values have also been presented as average values in this time period; the corresponding numerical values of activity and average exposure in each grid cell therefore refer to the number of people (density of population),
and the concentration times the number of people, respectively. This normalisation in terms of time makes it possible to compare population densities and average exposures during different time periods with each other. The time spent and average exposure in vehicular traffic is also included in the results. The Helsinki Metropolitan Area is located by the Baltic Sea and the centre of Helsinki is on a peninsula that is located approximately in the middle of the southern part of each figure. The Helsinki Metropolitan Area comprises of four cities: Helsinki, Espoo, Vantaa and Kauniainen; the total population is approximately 850 000. The NO2 concentrations are highest in the vicinity of the main roads and streets, and in the centre of Helsinki (Fig. 3a). The figure shows the distinct influence of ring road number 1 (situated at a distance of about 8 km from the city centre), the major roads leading to the Helsinki city centre, and the junctions of major roads and streets. As expected, the population density values are highest in the city centre (Fig. 3b). Results also show that there are elevated levels of population density in the vicinity of the district centres of the other major cities in the area (Espoo and Vantaa), and in the vicinity of major roads and streets; the latter is clearly due to increased vehicular traffic during the afternoon rush hours. As population average exposure is a combination of both the concentration and activity (or population density) values, it exhibits characteristics of both spatial distributions, i.e., elevated values in the Helsinki city centre, along major roads and streets, and in the vicinity of urban district centres (Fig. 3c). The predicted distributions of the population in terms of the NO2 concentrations are presented in Figs. 4(a)–(e) for the five selected diurnal time periods. The total area below each distribution is directly proportional to the number of people times concentration that are located in a specific class of environment (e.g., at home) during a specific time interval (e.g., the morning hour). Clearly, in evaluating the average exposure of the population, one also has to allow for the location of each distribution in terms of the concentration and the exposure time. Such figures can be used to analyse the relative importance of the average exposure in various microenvironments, and the diurnal variation of the population average exposure. During each time-period, the location of the peak of the distributions corresponding to home, work and other activities are on the average of the same order of magnitude. However, the number of people in the home and work environments is substantially larger on the average, compared with those in other activities (the same result can be seen from Fig. 2). It can therefore be concluded that the average exposure to NO2 at home and in the workplace is substantially more important than that in ‘‘other’’ activities.
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
2115
Fig. 3. (a)–(c) The predicted ambient air concentrations of NO2 (mg m3), the density of population (persons) and the average exposure of the population to NO2 concentrations (mg m3 persons), evaluated for the afternoon time period, as an average value in March 1996. The values in brackets in the legend refer to the number of square kilometres with the concentration, population density or average exposure in the selected range. The grid size is 100 m 100 m, the size of the depicted area is 23 km 16 km, and the solid black line is the coastline.
The distributions corresponding to traffic are spread over a wide range of concentrations for all time-periods. Although people are exposed to the highest NO2 concentrations in traffic, the time spent in traffic is relatively short, compared with that spent at home and in work environments. The average exposure in traffic is clearly lower, compared with that at home and in workplace environments. The highest concentrations in traffic occur during the afternoon rush hours, and fairly
high values also occur during the evening and morning periods.
4. Conclusions This paper presents a mathematical exposure model, which combines predicted concentrations, the location of the population, and the time spent at home, in the
2116
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
workplace, in traffic and at other places of activity. The main objective was to evaluate the spatial and temporal variation of average exposure of the whole urban population to ambient air pollution in different microenvironments. The model includes a treatment of all of the most important categories of population activity, including also the exposure in vehicular traffic. Results have been presented to illustrate the application of the model; these include the concentrations, population densities and average exposure in the Helsinki Metropolitan Area. The model presented here evaluates the hourly average population exposure to outdoor air concentrations of NO2. Clearly, personal, and temporally shorter term exposures can be substantially higher or lower, compared with the values computed using this model. We assumed that I/O ratio for NO2 concentrations was equal to unity. In some previous studies, simple empirically determined coefficients have been utilised for estimating indoor air concentrations, based on the corresponding ambient air values (e.g., Jensen, 1999). In the EXPOLIS study, the ratios of the indoor and outdoor concentrations (I/O) for nitrogen dioxide were on the average equal to 0.76. However, the ratio of indoor and outdoor air concentrations is dependent on several factors, including outdoor to indoor transport (ventilation and filtration), and sources and sinks indoors. Clearly, this ratio is dependent on the particular pollutant to be considered; in the case of particulate matter, it is also dependent on the particle size. Various models have been presented for evaluating the ratio of pollutant concentrations in indoor air; Kulmala et al. (1999) and Dimitroulopoulou et al. (2001) have presented such models in the case of particulate matter and nitrogen dioxide, respectively. However, such models require detailed input information concerning, e.g., the building structure, and ventilation and filtration systems. At present, there is insufficient data available for the application of such models over the whole of the metropolitan area. On the average, the indoor air concentrations of NO2 are lower than the corresponding outdoor air concentrations in the Helsinki Metropolitan Area, as there are very few homes and workplaces with gas appliances. We have evaluated the exposure to outdoor concentrations of NO2; however, people spend most of their time indoors (the average value in the area considered is equal to 87%). Despite this, we can conclude that the
2117
average exposure of population to NO2 at home and in the workplace is substantially more important than that in traffic and other activities. People are exposed to the highest NO2 concentrations in traffic, but the corresponding time is relatively short, compared with that spent at home and work environments. Clearly, the dispersion modelling approach has certain inherent limitations: Gaussian dispersion modelling does not allow for the detailed structure of buildings and obstacles. However, the terrain in the area is fairly flat and the average height of the buildings is low (most buildings are lower than 15–20 m). The computations considered here correspond to urban background concentrations, averaged over the length scale of the grid size applied. There is only a moderate number of street canyons in the area. Inside a street canyon, the actual concentrations can vary substantially on a scale of tens of metres; in such locations, nested street canyon dispersion model computations should be used (e.g., Kukkonen et al., 2001b). However, at present it is not feasible to conduct such an analysis for an extensive area including an agglomeration of several cities, due both to a lack of the relevant geometric building data, and to restrictions on computer resources. This model has been designed to be utilised by municipal authorities in evaluating the impacts of traffic planning and land use scenarios. For instance, the transportation system plan (TSP) for the Helsinki Metropolitan Area is revised at four-yearly intervals; the previous version has been described by the Helsinki Metropolitan Area Board (1999). This model will be used to evaluate the impacts of different scenarios in the new revision of the TSP. We also aim to estimate the adverse health effects caused to the population by air pollution and to simulate the burden of disease for each of the TSP scenarios.
Acknowledgements This study has been part of the National Research Programme on Environmental Health (SYTTY), the Health Promotion Research Programme (TERVE) and the research programme EUROTRAC-2, SATURN, ‘‘Studying atmospheric pollution in urban areas’’, 1998– 2002. The funding from the Academy of Finland for all of these studies is gratefully acknowledged. In Helsinki, the EXPOLIS study has been supported by EU Contract
Fig. 4. (a)–(e) Predicted distribution of the number of people (N) against the concentration of NO2 during the morning, day, afternoon, evening and night periods in the Helsinki Metropolitan Area in March, 1996. The vertical scale is different for the number of people at home and in the workplace (left-hand side scales), compared with the number of people in traffic and other activities (righthand side scales). The vertical scale for the number of people at home and in the workplace is also different in the night period, compared with those in the other panels.
2118
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119
N ENV4-CT96-0202 (DG 12 -DTEE), Academy of Finland Contract N 36586 and intramural funding by the National Public Health Institute of Finland.
References Alm, S., Mukala, K., Pasanen, P., Tiittanen, P., Ruuskanen, J., Tuomisto, J., Jantunen, M.J., 1998. Personal NO2 exposures of preschool children in Helsinki. Journal of Exposure Analysis and Environmental Epidemiology 8, 79–100. Bohler, T., Riise, A., 1997. Using the air quality assessment system AirQUIS in modelling the population’s exposure to traffic induced air pollution. Fourth International Scientific Symposium on Transport and Air Pollution, 9–13 June 1997, Avignon, France. Brunekreef, B., Lumens, M., Hoek, G., Hofschreuder, P., Biersteker, K., 1989. Pulmonary function changes associated with an air pollution episode in January 1987. Journal of Air Pollution Control Association 39, 1444–1447. City of Helsinki, Urban Facts, 1999. Statistical yearbook of the City of Helsinki. Gummerus Kirjapaino OY, Jyv.askyl.a. Clench-Aas, J., Bartanova, A., Bohler, T., Gronskei, K.E., Sivertsen, B., Larsssen, S., 1999. Air pollution exposure monitoring and estimating part I: integrated air quality monitoring system. Journal of Environmental Monitoring 1, 313–319. Dimitroulopoulou, C., Ashmore, M.R., Byrne, M.A., Kinnersley, R.P., 2001. Modelling of indoor exposure to nitrogen dioxide in the UK. Atmospheric Environment 35, 269–279. Freijer, J.I., Bloemen, H.J.Th., De Loos, S., Marra, M., Rombout, P.J.A., Steentjes, G.M., Van Veen, M.P., 1997. AirPEx: Air Pollution Exposure Model. RIVM, The Netherlands. Report No. 650010 005. . H.arkonen, J., Valkonen, E., Kukkonen, J., Rantakrans, E., Jalkanen, L., Lahtinen, K., 1995. An operational dispersion model for predicting pollution from a road. International Journal of Environment and Pollution 5 (4–6), 602–610. . H.arkonen, J., Walden, J., Kukkonen, J., 1997. Comparison of model predictions and measurements near a major road in an urban area. International Journal of Environment and Pollution 8 (3–6), 761–768. Helsinki Metropolitan Area Board, 1999. Helsinki Metropolitan area transportation system plan, PLJ 1998. Helsinki Metropolitan Area Series A 1999:4. YTV Helsinki Metropolitan Area Council, Helsinki. INRO, 1994. EMME/2 User’s manual. INRO Consultants Inc., Montreal, Canada. . Jantunen, M.J., H.anninen, O., Katsoyanni, K., Knoppel, H., Kunzli, . N., Lebret, E., Maroni, M., Saarela, K., Sram, R., Zmirou, D., 1998. Air pollution exposure in European Cities: the Expolis study. Journal of Exposure Analysis and Environmental Epidemiology 8, 495–518. . Jantunen, M.J., Katsoyanni, K., Knoppel, H., Kunzli, . N., Lebret, E., Maroni, M., Saarela, K., Sram, R., Zmirou, D. 1999. Final report: Air Pollution Exposure in European Cities: the Expolis study, Publications of the National Public Health Insititute, KTL B16/1999, Kuopio.
Jensen, S.S., 1999. A geographic approach to modelling human exposure to traffic air pollution using GIS. Ph. D. Thesis. National Environmental Research Institute, Denmark. . Jonson, T., Johansson, C., Hadenius, A., Johansson, P.A., 1999. Shape. The Stockholm study on health effects of air pollution and their economic consequences Part I: NO2 and particulate matter in Stockholm—concentrations and population exposure, Swedish National Road Administration No. 1999:41, Stockholm, Sweden. Karppinen, A., Kukkonen, J., Elol.ahde, T., Konttinen, M., Koskentalo, T., Rantakrans, E., 2000a. A modelling system for predicting urban air pollution, model description and applications in the Helsinki metropolitan area. Atmospheric Environment 34, 3723–3733. Karppinen, A., Kukkonen, J., Elol.ahde, T., Konttinen, M., Koskentalo, T., 2000b. A modelling system for predicting urban air pollution, comparison of model predictions with the data of an urban measurement network. Atmospheric Environment 34, 3735–3743. Karppinen, A., Joffre, S.M., Kukkonen, J., 2000c. The refinement of a meteorological preprocessor for the urban environment. International Journal of Environment and Pollution 14 (1-6), 565–572. Kousa, A., Monn, C., Rotko, T., Alm, S., Jantunen, M.J., 2001a. Personal exposures to NO2 in the EXPOLIS-study: relation to residential indoor, outdoor and workplace concentrations in Basel, Helsinki and Prague. Atmospheric Environment 35, 3405–3412. Kousa, A., Kukkonen, J., Karppinen, A., Aarnio, P., Koskentalo, T., 2001b. Statistical and diagnostic evaluation of a new-generation urban dispersion modelling system against an extensive dataset in the Helsinki Area. Atmospheric Environment 35 (27), 4617–4628. . Kukkonen, J., H.arkonen, J., Walden, J., Karppinen, A., Lusa, K., 2001a. Evaluation of the dispersion model CAR-FMI against data from a measurement campaign near a major road. Atmospheric Environment 35, 949–960. Kukkonen, J., Valkonen, E., Walden, J., Koskentalo, T., Aarnio, P., Karppinen, A., Berkowicz, R., Kartastenp.aa. , R., 2001b. A measurement campaign in a street canyon in Helsinki and comparison of results with predictions of the OSPM model. Atmospheric Environment 35, 231–243. Kulmala, M., Asmi, A., Pirjola, L., 1999. Indoor air aerosol model: the effect of outdoor air, filtration and ventilation on indoor concentrations. Atmospheric Environment 33, 2133–2144. M.akel.a, K., 1999. LIPASTO, Calculation system for traffic emissions and energy consumption. http://www.vtt.fi/rte/ lipastoe/index.htm. VTT Communities and Infrastructure, Espoo. Mukala, K., Alm, S., Tiittanen, P., Salonen, R.O., Jantunen, M.J., Pekkanen, J., 2000. Nitrogen dioxide exposure assessment and cough among preschool children. Archives of Environmental Health 55, 431–438. Ott, W.R., 1985. Total human exposure. Environmental Science and Technology 19 (10), 880–886. . Ottl, D., Kukkonen, J., Almbauer, R.A., Sturm, P.J., Pohjola, M., H.ark.onen, J., 2001. Evaluation of a Gaussian and a Lagrangian model against a roadside dataset, with focus on low wind speed conditions. Atmospheric Environment 35, 2123–2132.
A. Kousa et al. / Atmospheric Environment 36 (2002) 2109–2119 Ryan, P.B., 1991. An overview of human exposure modeling. Journal of Exposure Assessment and Environmental Epidemiology 1, 453–473. Schwartz, J., Spix, C., Wichmann, H.E., Malin, E., 1991. Air pollution and acute respiratory illness in five German communities. Environmental Respiratory 56, 1–14. Sexton, K., Ryan, P.B., 1988. Assessment of human exposure to air pollution: methods, measurements, and models. In:
2119
Watson, A.Y., Bates, R.R., Kennedy, D. (Eds.), Air Pollution, the Automobile, and Public Health. National Academic Press, Washington, D.C., pp. 207–238. Willmott, C.J., 1981. On the validation of models. Physical Geography 2, 184–194. World Health Organization (WHO), 2000. Air quality guidelines for Europe, 2nd Edition. WHO regional publications, European Series, 91, Copenhagen, Denmark.