Air pollution exposure monitoring and estimation Part III.‡ Development of new types of air quality indicators† Cristina Guerreiro, Jocelyne Clench-Aas and Alena Bartonova Norwegian Institute for Air Research, PO Box 100, N-2027 Kjeller, Norway. E-mail: cristina.guerreiro@nilu.no; Fax:+47 63 89 80 50; Tel:+47 63 89 80 00 Received 7th April 1999, Accepted 25th June 1999 The temporal pattern of exposure to a specific compound may aVect health in several ways.Exposure to pollution can have short-term eVects or long-term eVects. For some compounds there is a threshold under which there is no presumed measurable eVect, whereas for other compounds, there is no presumed threshold. For short-term eVects, the exposure to a high concentration of a compound one day may either increase or decrease the response if values of the same compound become high again the next day.Adaptation to eVects of short-term exposure to ozone, for example, is reported. Similarly, health response to sudden high peaks of concentration may also possibly diVer in eVect from those to peaks attained more gradually. For long-term eVects of some compounds, the cumulative exposure may be more decisive in influencing health.This paper proposes and describes in detail several air quality indicators that reflect the time variability and the episodic nature of air pollution exposure, as an attempt to represent the temporal aspects of pollution exposure that may have important eVects on health. Mean concentrations, 98th percentile and maximum values are the traditional indicators for estimating exposure.The temporal variability of particulate matter (PM10) and NO2, however, is here described by means of: (1) the rate of change of pollution as the diVerence between two consecutive hourly or daily values, and of (2) episodes, described in terms of number, duration and inter-episode period, maximum concentration in the episode, and integrated episode exposure. example, is reported.5 For long-term eVects of some com- 1.Introduction pounds, the cumulative exposure may be more decisive in Health eVects are increasingly being described in the form of influencing health. the dose-response (or exposure-eVect) relationship. This If there are diVerences in physiological response in the relationship allows quantifying the altered health status of a diVerent exposure situations, then the following characteristics population by quantifying both the current and future health of exposure need to be considered, in addition to average and status based on measured and projected pollution peak exposure: (1) the temporal pattern with which concentrations.exceedances of threshold or AQG occur, (2) how many of Public authorities are currently using as air quality indicators these exceedances are in reality occurring during the same (AQI) for health eVects, average and peak concentrations of episode, (3) how many episodes have in reality occurred, or pollutants, together with the number of exceedances of existing (4) how long the episodes have lasted. air quality guidelines (AQG) (WHO, 1987),1 (WHO, 1998),2 This paper proposes several new types of AQIs, based on (Commission of the European Communities, 1998).3 However, hourly air pollution concentrations calculated with a dispersion it is more and more evident that attempts to characterise model in a 1 km2 grid system covering the city of Oslo for the health eVects using a simple relationship have not been quite winter 1994/95.eVective. The reasons for this are many. For example, it is not precluded that the temporal variation of exposure to a pol- 2. Methods lutant may be influencing the health impact. It may be necessary in the future to account for the temporal pattern of This section describes the criteria and methods used for the exposure in the setting of air quality guidelines.selection of AQIs. What should AQIs represent is discussed. The temporal pattern of exposure to a specific compound Basic concepts such as rate of change of exposure, episode, may aVect health in several ways. Exposure to pollution can and inter-episode period are defined. have short-term eVects (immediately or in the next few days) or more long-term eVects. For some compounds there is a 2.1 What should the indicator represent threshold under which there is no presumed measurable eVect, The air quality indicator (AQI) should represent both the whereas for other compounds, there is no presumed threshold spatial and temporal aspects of pollution exposure that may (for example PM10) (WHO, 1996).4 For short-term eVects, have important eVects on health. Two indicators are needed, the exposure to a high concentration of a compound one day the population air quality indicator and the individual air may possibly either increase or decrease the response if values quality indicator.of the same compound become high again the next day. An air quality indicator should provide information relevant Adaptation to eVects of short-term exposure to ozone, for in evaluating possible health eVects.It should be applicable for the evaluation of both short-term health eVects and long- †Presented at AIRMON ’99, Geilo, Norway, February 10–14, 1999. ‡ For Part II, see ref. 6. term to chronic health eVects. The indicator should also be J. Environ. Monit., 1999, 1, 327–332 327usable to measure or predict changes in exposure resulting considered one episode, would in this case, because of technical reasons, count as several.from pollution abatement measures. The elements necessary for the proper definition of AQI are, therefore, the spatial The episodes may then be described by (see Fig. 2): (1) peak height (maximum value in episode), (2) duration of episode, distribution of the pollution, the time structure of the exposure, and the magnitude of the exposure.(3) inter-episode period, and (4) integrated episode exposure, episode AOTx (sum of the concentration hours during an The air pollution episode can be the basic entity underlying the air quality indicators. The definition of an episode is the episode that exceeds the threshold value of x). Area over threshold (AOT) values are usually given in ppb h-1 (ozone), period of time that pollution concentrations are above a predefined level such as eVect threshold or air quality guideline.but may also be represented by mg m-3 h-1 as in this paper. Episodes have a time structure that defines when and how often episodes occur, and a magnitude that reflects both the 2.3 Choice of sites duration of the episode and the peak concentration reached Based on the EPISODE model, air quality concentrations during the episode. were calculated for Oslo for the winter 1994/956 at the square Health eVects of pollution are continuously under study.It kilometre level. Three grid squares were chosen in this study remains unknown whether the absolute concentration of polto estimate diVerent air quality indicators using the calculated lution or the rate of change of concentrations has the greatest concentrations of NO2 and PM10.These three grid squares eVect on diVerent health end-points. correspond to the squared kilometres where Carl Berners An air quality indicator may also reflect how rapidly polplass, Majorstua and Lysaker are located in Oslo. The two lution levels change.The health eVect of exposure to first areas (Carl Berners plass and Majorstua) represent areas 100 mg m-3 of a pollutant may diVer if the previous level for containing traYc in the centre of the city, while the third some days has been 20 as opposed to 90 mg m-3. square (Lysaker) represents a major traYc artery around the Exposure may be described on an individual basis, whenever city, and is located on the West border of the city.individual data are available. Population air quality indicators may be obtained from integrated exposure estimates, such as from estimates based on a square kilometre grid, or air quality 3. Results—temporal aspects measurements on a city level. Mean concentrations, 98th percentile and maximum values are the traditional indicators for estimating exposure.The 2.2 Definition of specific air quality indicators to describe temporal variability of PM10 and NO2, however, is here temporal variability described by means of: (1) the rate of change of pollution as the diVerence between two consecutive hourly or daily values, Both the temporal and spatial aspects of air pollution concentrations need to be described.To obtain relevant data, the and of (2) episodes, described in terms of number, duration and inter-episode period, maximum concentration in the epi- AirQUIS/Episode model was run hourly for 6 months. The assumptions of the model, the emissions used and type of sode, and integrated episode exposure (episode AOT35/100). model, are described in Walker et al.6 and Grønskei et al.7 The pollution concentrations (hourly and daily) were given for each selected square kilometre grid, and for the components NO2 and PM10.The statistical parameters of the time series were given as mean (mg m-3), maximum hourly and daily concentration, and 98th percentile of hourly values. In addition to these descriptors, rate of change of the time series, and description of episodes, may prove useful.The rate of change for hourly and daily data may be described as the diVerence between the two consecutive values in the time series (‘delta concentrations’) (see Fig. 1). As pollution concentrations change, episodes occur. An episode is in this study defined as the period when concentrations of pollutants exceed a threshold, here set as 100 mg m-3 for hourly NO2 and 35 mg m-3 for daily PM10.Should the values consistently lie around the threshold for several hours, a series of concentrations which is generally Fig. 2 Definition of ‘episode length’, ‘episode peak’ and ‘area over Fig. 1 Defining the rate of change (delta conc. in legend) of exposure to pollution. threshold’ (AOT). 328 J. Environ. Monit., 1999, 1, 327–332The general features of pollution exposure can be described The calculated rate of change of hourly concentrations of NO2 and daily concentrations of PM10 in the three selected grid by the more traditional measures as seen in Table 1.It is evident in this table that Lysaker has the highest concentration squares in Oslo are presented in Table 2. especially for PM10. However, from the point of view of health, this information may not be suYcient.Measures to 3.2 Episodes—time pattern protect health may need to account for the pattern in exposure Exposure to air pollution occurs as a series of episodes. people endure. The severity of pollution in an area is not Episode statistics, characterising the time pattern of exposure completely indicated by a simple 98th percentile, since health in terms of duration of episodes and length of periods between eVects may be worse if high pollution concentrations come episodes (inter-episode periods), have been computed for the sporadically, not allowing the body to adapt to them as may 3 grid squares in Oslo and are given in Table 3.be the case in one or two long episodes. Therefore, it may be Only Lysaker had episodes of NO2 that lasted longer than necessary in the future to specify, in addition to concentrations, 8 h (1% of total time), whereas for 4% of the total time they acceptable patterns of exposure.It may be necessary to control lasted only 1 to 2 h. Lysaker had only 5 periods of 5 days or the number of episodes, and the severity of the episodes, which more without an episode.For PM10, the episodes had a longer are described by the integrated exposure. duration, so that all three sites experienced episodes lasting longer than 8 h, and the inter-episode periods were shorter. 3.1 Rate of change of pollution exposure These statistics would obviously change if the threshold was The temporal pattern of exposure may aVect health in several changed.For daily PM10 over the new Norwegian guideline9 ways. An impact on health by exposure to high concentrations of 35 mg m-3, all three sites experienced episodes of 1 day can lead to a direct linear response, a potentiation (sensitis- duration. Lysaker had 2 over 6 daylong episodes. Lysaker had ation) or a reduction (adaptation) of response. The potenti- episodes 34% of the total time.ation or reduction can last for a period of time after exposure. In the future, information of this type may be used to Although the phenomenon of sensitisation is not described, specify control measures, which: (1) will not allow more than adaptation to ozone is reported.5 Similarly, peaks attained a certain number of episodes, (2) will restrict the allowable through rapid increase in pollution concentrations may also duration of the episode and (3) will not for example allow diVer in eVect from those attained more gradually.The pos- more than two episodes with an inter-episode period of 7 days sibility exists that occasional high concentrations of air pol- or more, as better information concerning the nature of the lution may lead to larger health eVects than more continuously health eVects becomes available.rising concentrations. The phenomenon of harvesting may also occur, where immediately after exposure to a peak concen- 3.3 Episodes—peak concentrations tration, most members of the potential population of sensitive individuals either enter the hospitals or die. Continued expo- Air pollution situations can be characterised by average concentrations.However, high mean values can be caused by sure to high concentrations during a latency period does not lead to further morbidity or mortality.8 occasional extra high concentrations or by more frequent, notso- high episodes. The 98th percentile of short-term concen- Urban air pollution concentrations vary periodically with time as shown in the time series plot in Fig. 3 in Walker et al.6 trations does not diVerentiate between the two situations. In judging the potential health eVects of diVerent air pollution Urban air pollution concentrations vary periodically with time. Table 1 Mean, maximum and 98th percentile of hourly concentrations of NO2 and daily concentrations of PM10 in three selected grid squares in Oslo, winter 1994/95 N Mean Maximum Standard 98th percentile Grid square cases /mg m-3 /mg m-3 deviation /mg m-3 Hourly NO2 Carl Berners plass 4368 53 139 25.8 101 Majorstua 4368 47 127 25.2 94 Lysaker 4368 63 169 28.0 117 Daily PM10 Carl Berners plass 182 17 81 13.6 60 Majorstua 182 12 47 8.6 37 Lysaker 182 30 112 23.0 63 Table 2 Calculated rate of change of hourly concentrations of NO2 and daily concentrations of PM10 in three selected grid squares in Oslo, winter 1994/95.Absolute values of both positive and negative changes in concentrations are included Maximum rate 98th percentile of N of change rate of change Standard Grid square cases /mg m-3 h-1 /mg m-3 h-1 deviation Hourly NO2 Carl Berners plass 4367 82.2 39.6 15.2 Majorstua 4367 80.7 38.8 13.9 Lysaker 4367 88.7 40.6 16.0 Daily PM10 Carl Berners plass 181 55.4 36.4 15.7 Majorstua 181 33.7 21.9 10.2 Lysaker 181 77.3 62.8 26.4 J.Environ. Monit., 1999, 1, 327–332 329Table 3 Duration of episodes and periods between episodes for NO2 and PM10 at three selected sites in Oslo. Percentages are presented both as a function of episode time and of total time Carl Berners plass Majorstua Lysaker Duration of episode Duration of episode Duration of episode % of % of % of Episode Total Episode Total Episode Total No.time time No. time time No. time time Hourly NO2>100 mg m-3 Episode length/h 1–2 51 85 1.33 18 90 0.46 135 79.4 3.98 3–7 9 15 0.82 2 10 0.14 30 17.6 2.68 8+ 0 0 0.00 0 0 0.00 5 2.9 1.01 Inter-episode period 1–2 h 7 11.5 0.25 2 9.5 0.07 35 20.4 1.12 3–8 h 5 8.2 0.73 1 4.8 0.09 34 19.9 4.12 9–24 h 10 16.4 4.42 2 9.5 0.73 60 35.1 23.53 2 days 10 16.4 7.85 2 9.5 1.81 22 12.9 19.55 3 days 8 13.1 12.09 1 4.8 1.19 9 5.3 13.76 4 days 4 6.6 7.49 1 4.8 1.81 6 3.5 12.52 5 days 7 11.5 18.20 3 14.3 7.90 2 1.2 5.43 <5 days 10 16.4 46.82 9 42.9 85.81 3 1.8 12.29 Daily PM10>35 mg m-3 Episode length/days 1 8 40 4.40 5 100 2.75 14 23 7.69 2 3 30 3.30 0 0 0.00 5 16 5.49 3 2 30 3.30 0 0 0.00 3 15 4.95 4–5 0 0 0.00 0 0 0.00 3 23 7.69 6+ 0 0 0.00 0 0 0.00 2 23 7.69 Inter-episode period/days 1–3 6 7 6.59 1 2 1.65 18 26 17.03 4–10 3 17 14.84 1 5 4.40 9 48 31.87 11–20 3 26 23.08 1 10 9.34 0 0 0.00 21–50 2 50 44.51 1 19 18.13 1 26 17.58 51+ 0 0 0.00 2 66 63.74 0 0 0.00 Table 4 Maximum hourly and daily concentrations in episodes for NO2 and PM10 at three selected grid squares in Oslo Carl Berners plass Majorstua Lysaker Episode peak/mg m-3 Frequency % Frequency % Frequency % Hourly NO2>100 mg m-3 100–110 36 60.0 15 75.0 103 60.6 111–130 23 38.3 5 25.0 49 28.8 131–150 1 1.7 0 0 15 8.8 151–170 0 0 0 0 3 1.8 >171 0 0 0 0 0 0 Daily PM10>35 mg m-3 35–40 3 23.1 4 80.0 6 22.2 41–50 5 38.5 1 20.0 4 14.8 51–70 3 23.1 0 0.0 6 22.2 71–90 2 15.4 0 0.0 7 25.9 91+ 0 0.0 0 0.0 4 14.8 situations, it is of interest to know whether the 2% of time Table 5 DiVerent expressions for the cumulative dose of hourly NO2 and daily PM10 for diVerent thresholds in three selected grid squares that values exceed the 98th percentile all occur during the same episode, as opposed to occasional but not consecutive Threshold=100 mg m-3 high values.This diVerentiating can only be carried out by No. of (hourly NO2) examining the number of episodes and the episode peak height. hours/days and 35 mg m-3 Episodes can be characterised by the peak height, the highest Grid square over threshold (daily PM10) concentration reached during the episode (Table 4). At Hourly NO2 AOT100/mg m-3 h-1 Lysaker, for NO2, as many as 10.6% of episodes had peak Carl Berners plass 94 870 values over 130 mg m-3, whereas Carl Berners plass had only Majorstua 26 174 1.7% of such episodes.The episode concentrations of PM10 Lysaker 335 4001 were higher. At Lysaker 15% of episodes exceeded 91 mg m-3 Daily PM10 AOT35/mg m-3 day-1 daily average. Carl Berners plass 20 258 In the future, control measures may be formulated as a Majorstua 5 24 restriction in the number of episode peak values over a given Lysaker 61 1290 concentration instead of the 98th percentile. 330 J. Environ. Monit., 1999, 1, 327–332Table 6 Distribution of episode AOT for NO2 and PM10 at three selected sites in Oslo. (Threshold=35 mg m-3 for daily values of PM10 and 100 mg m-3 for hourly NO2) Carl Berners plass Majorstua Lysaker Episode AOT S mg m-3 Frequency % Frequency % Frequency % NO2 AOT100/mg m-3 h-1 0–50 33 55.0 14 70.0 92 54.1 51–80 17 28.3 5 25.0 43 25.3 81–150 8 13.3 1 5.0 15 8.8 151–300 1 1.7 0 0 6 3.5 301–600 1 1.7 0 0 6 3.5 601–1000 0 0 0 0 5 2.9 1001–2000 0 0 0 0 3 1.8 Daily PM10 AOT35/mg m-3 day-1 0–10 9 45 4 80 7 11 11–30 5 25 1 20 14 23 31–60 6 30 0 0 2 3 61–100 0 0 0 0 19 31 101–150 0 0 0 0 15 25 151+ 0 0 0 0 4 7 3.4 Episodes—integrated exposure percentile).For ozone, it has been suggested to use AOT60= 0 (ppb h-1) as an approximation for health guidelines.10 For some short-term health eVects it is the concentration that As this paper illustrates, however, these air quality indicators is decisive in initiating a health eVect. However, for other do not account for the time variability of the exposure, and short-term and for long-term eVects, the cumulative or inte- do not account properly for the cumulative exposure to grated exposure may be the determinant in causing an eVect.pollution. As knowledge of the health eVects of pollution It may be of importance whether an episode having a ‘total increases, there is a basis for specifying limits based on both exposure’ of 120 mg m-3 h-1 occurs as 60 mg m-3 over 2 h or temporal patterns and cumulative exposure.as 120 mg m-3 over 1 h. In the future, it may be necessary to The following set of indicators seems useful in evaluating specify in addition to an AQG, a limited total integrated the health eVects of air pollution based on the time structure exposure, or a maximum acceptable integrated exposure over of the exposure.one single episode. (1) Annual mean, daily/hourly mean, 98th percentile, 99.9th Another and more usual way is to define the accumulated percentile (or other high percentiles). exposure over a given threshold (AOT exposure for episodes (2) Total semi-annual AOT35/AOT100, etc. (threshold or episode AOT35/100).Based on the example given above chosen separately for each component). this would mean that with a threshold of 50 mg m-3 h-1, the (3) Total semi-annual number of episodes. 2 h episode would have a value of 20 mg m-3 h-1, whereas (4) Episode AOT35/100 etc. the 1 h episode would have a value of 70 mg m-3 h-1. The These indicators will allow a more complete description of the AOT calculated over a nonzero threshold would give more air pollution situation, so that regulations may be developed weight to the higher peaks.In Table 5, the summed results of in the future that will more completely protect the population. the two ways of calculating the cumulative dose are presented, This paper attempts to show how current modelling tools together with the total number of hours of ‘episode time’. The may be used to develop and present diVerent air quality frequency distribution of the individual episode integrals is indicators.It will be the role of health authorities to specify, presented in Table 6. if necessary, what regulations need to be imposed. They will set maximum allowable integrated episode exposures, maximum number of allowable episodes, and maximum allowable 4.General discussion peak episode concentrations. Guidelines should set the limits of population exposure to air pollution to protect health and the environment. Currently, guidelines exist that use standard statistical concepts such as 5. Acknowledgements mean and maximum concentrations, or percentiles. However, This study was funded by the Norwegian Pollution Control maximum values are very diYcult to use, since errors in Authority.measurement or estimation can produce false maxima. Means may not reflect suYciently the short-term peaks in exposure. The 98th percentile (or other high percentile) is a more References stable measure. 1 WHO Air Quality Guidelines for Europe, WHO Regional The European Commission3 has recently proposed guide- Publications, European Series No. 23, Copenhagen, 1987.lines with percentiles diVerent from the 98th percentile. For 2 WHO Air Quality Guidelines for Europe, WHO Regional example for sulfur dioxide, hourly maxima cannot be exceeded Publications, European Series, Copenhagen, 1998, Internet address: http://www.who.dk/eh/airqual.htm more than 24 times per year and daily maxima, 3 times per 3 Commission of the European Communities, Common position year (equivalent to a 99.97th percentile) whereas for NO2, (EC) No 57/98, OV.J. Eur. Com. C360, 1998, pp. 99–122. hourly maximum concentrations cannot be exceeded more 4 WHO Quantification of health eVects related to SO2, NO2 , O3 than 18 times per year (equivalent to a 99.8th percentile). For and particulate matter exposure.Report from the Nordic Expert PM10 the daily maximum values cannot be exceeded more Meeting, Oslo, 15–17 October, 1995, ed. J. Clench-Aas and than 35 times per year in 2005 (equivalent to a 90.4th percen- M. Krzyzanowski, Kjeller (NILU OR63/96) (EUR/ICP/EHAZ 94 04/DT01), 1996. tile) and 7 times per year in 2010 (equivalent to a 98th J. Environ. Monit., 1999, 1, 327–332 3315 M. J. Hazucha, D. V. Bates and P. A. Bromberg J. Appl. Physiol, base, part of the follow-up process for the Helsinki Declaration, 1989, 67, 1535. Phase 2 (in Norwegian), National institute for public health, 6 S. E. Walker, L. H. Slørdal, C. Guerreiro, F. Gram and K. E. Oslo, 1998. Grønskei, J. Environ. Monit., 1999, 1, 321. 10 United Nations Economic Commission for Europe/World Health 7 K. E. Grønskei, S. E.Walker and F. Gram, Atmos. Environ., 1993, Organization, Health eVects of ozone and nitrogen oxides in an 27B, 105. integrated assessment of air pollution, convention on Long-Range 8 J. Samet, S. L. Zeger and K. Berhane, Particulate air pollution and Transboundary Air Pollution, the proceedings of an International daily mortality: replication and validation of selected studies, the Workshop, Eastbourne, UK, 10–12 June, 1996, Institute for phase I report of the particle epidemiology evaluation project, Environment and Health, University of Leicester. Health EVects Institute, Cambridge, MA, 1995. 9 National Institute for Public Health, Norms/threshold values/standards. I: Environment and health—a research based knowledge Paper 9/02778E 332 J. Environ. Monit., 1999, 1, 327–332