Introduction

Currently, a European Air Quality Directive on Polycyclic Aromatic Hydrocarbons (PAHs), as a daughter directive under the Air Quality Framework Directive (96/62/EC), is in preparation. The group of PAHs comprises hundreds of individual chemicals. In environmental compartments, usually only a few compounds are monitored because of practical and financial reasons. The criteria for the selection and the number of compounds measured at air monitoring sites are often not obvious. In a review on particulate matter air pollution, the particle-bound PAHs have been pointed out as more important in determining the effects of air pollution than particle mass (Lighty et al., 2000). With a mass balance model source, transport and fate of seven PAH compounds have been described in water, sediments and in the atmosphere, giving a better understanding of the distribution of PAHs in environmental media (Mackay and Hickie, 2000).

Some PAHs or the ratio of individual PAHs give hints at the source of pollution. The benzo(a)pyrene (BaP)/COR ratio, for instance, is about 1:2 if traffic is the major origin of pollution, whereas coal combustion causes ratios between 5and 10 (Grimmer, 1980). The temporal and spatial dependence of PAH profiles of 14 PAH compounds were investigated by deRaat and de Meijere (1991), revealing volatility, reactivity and sources of PAH as the most important factors influencing the profile. A key question in the administrative context is whether air quality with respect to the PAH can be adequately assessed by a single compound or at least by a selected set of single compounds.

Table 1 shows 25 relevant compounds. Until now, BaP has frequently been used as an indicator, as it is one of the most potent carcinogens among the PAHs. However, there are other PAHs that are relevant for human health (IPCS, 1998). In Table1, the compounds that are carcinogenic in animals are noted (IARC, 1987). Since the pattern of PAHs in ambient air varies and the individual compounds have different potency, a single indicator might not adequately reflect the total carcinogenic risk from exposure in ambient air. Several attempts have been made to classify the relative carcinogenic potency of individual PAHs. In Table 1, the toxic equivalency factors derived by Nisbet and LaGoy (1992) and the WHO (1998) for some compounds — relative to BaP fixed at 1 — are listed. Because of many uncertainties, so far, the European administrative agencies have not introduced these factors in applied risk assessment to define threshold values for indoor and ambient air, water or soil. In view of these aspects, it might be more appropriate to monitor and regulate a greater number of PAHs. In the US, USEPA has recommended to determine the concentrations of 16 PAHs (see Table 1; USEPA, 1993).

Table 1 - 25 PAH compounds.

Full table (78K)

The question, "Which or how many PAHs are representative for PAH exposure in the air?", strongly depends on the correlation between the levels of single compounds. If it turns out that all PAHs or at least all health relevant PAHs are strongly correlated, a representative compound might be used to predict levels of other PAHs. The concentrations of other PAHs could be estimated with sufficient confidence. Thus, regulation of a single marker compound might be adequate. If correlations are weak, levels of single compounds cannot be used to predict the concentrations of other PAHs. The determination of several marker compounds would be necessary. Then, exposure and impact on human health had to be estimated on the basis of multiple measurements.

The PAH concentration data obtained in Germany between 1990 and 1998 are subjected to descriptive analysis with special attention to their intercorrelation. An appropriate statistical procedure to determine how many "dimensions" of independent information are contained in a set of variables is the factor analysis. This method has been developed in social sciences. Factor analysis is aimed to differentiate subsets of variables with high intercorrelation. These subsets of variables are considered to represent a single factor. Different factors provide different and independent information. Thus, the statistical procedure of factor analysis suits the questions involved in the topic of PAH air pollution.

Here, the number of factors derived from the correlation matrix of PAH concentration data should give an impression on how many dimensions have to be taken into account. This would offer an answer to the question on how many and which marker compounds should be selected for regulative purpose in order to characterize air quality with respect to the PAHs. In an early study (Prinz and Stratmann, 1968), a small sample of PAH concentrations had been submitted to factor analysis.

Methods

In Germany, PAH concentrations in ambient air are ascertained under the authorship of the federal states. In 1999, the actual collection of emission data was kindly submitted by the emission authorities of the federal states to facilitate this joint analysis. Figure 1 gives a map of the sampling locations.

Figure 1.

Map of Germany with the 159 PAH monitoring sites (1990–1998) included in this investigation.

Full figure and legend (50K)

Additional information was gathered to characterize the monitoring site. The rough environmental classification (rural or urban environment, traffic or industry dominated) had no general and precise definition. For the purpose of this investigation, the classification was made according to information delivered by the emission authorities or their documentation of monitoring sites. The selection of monitoring sites and their description reflect the purpose of the measuring activities. Further details such as traffic density, specific industrial sources or population size affected are not available.

There are concentration data of 25 PAHs, characterized by their location, the environmental classification, the year of measurement and the number of measurements. The concentration is always given in nanograms per cubic meter of air. There is no information about the laboratories involved in chemical analysis; yet, PAH analysis in Germany follows a national guideline (Gladtke, 1998), the "VDI Richtlinie" (VDI Guideline, 1996). The method of sampling and detection is described in the VDI Guideline, according to which mostly only the particle phase is sampled. Detection and quantification are carried out with gas chromatography and GC-FID as analytical method.

The available quantitative and qualitative data were summarized and described by indices of central tendency and variability. The original, as well as the ln-transformed, values were examined using the Kolmogorv–Smirnov–Lilliefors test for deviance from normal distribution. Various subgroups related to background information were compared on condition that the compound was represented with a sufficient sample size of measurements. Here, exemplary for BaP, the t-test for differences of the mean (with unequal variances) was applied. Correlations between the PAH compounds are described with Pearson's correlation coefficient and completed with the calculation of the confidence interval with Fisher's z-transformation.

Multiple regression analyses were performed on these data to investigate if, and to which proportion variability of single PAH compounds might be explained by the combined effects of the background variables. The dependent variables — the PAHs — were included as ln-transformed values. All independent variables supposed to influence the distributive patterns of PAHs — environmental classification, East–West classification and time — were included in one step. The independent variables were checked for intercorrelation to preclude collinearity, residual analysis and the examination of the Durbin–Watson coefficient for autocorrelation. Stepwise regression would not improve the final model.

The matrix of PAH concentrations gives rise to two types of question: Can a smaller set of measurements be used to replace the original set? What are the underlying dimensions being measured by the entire data set? The explorative factor analysis performed in this investigation sets out to answer these questions with the method of principal components analysis. Yet, a prerequisite for any kind of factor analysis is a correlation matrix with sufficient value pairs. Therefore, only six PAH compounds could be included in factor analysis. Principal component analysis with varimax rotation was applied to explain as much of the total variation in the data as possible with as few factors (i.e., principal components) as possible (Kleinbaum et al., 1988). Split samples techniques were applied to prove the results. All statistical analyses were performed with SPSS (Brosius and Brosius, 1998).

Results

Eleven federal states transmitted their values (N=526). Six spot checks from Bavaria, a single measurement of benzo[c]phenanthrene (0.43) and an average value based on 5 years were excluded from analysis. The 518 values included in the analysis are predominantly annual means (98%), which mostly (84%) rely on six measurements per year. The sum of PAHs could not be included in the statistical analysis because number and selection of included compounds differed between the federal states.

The great majority (75%) of measurements were ascertained in North Rhine-Westphalia, especially in the Rhine-Ruhr Area (see Figure 1). More than half of the remaining measurements (80 of 130) were ascertained in the new federal states. One third of the measurements was taken in a rural (34%) and urban environment (32%), respectively; the rest is equally assigned to industrial region (17%) or traffic-dominated environment (17%).

With respect to the time span involved, the first and last years included the in analysis, 1990 and 1998, offer only two to six measurements; the years 1991–1997 are represented by more than 50 measurements each.

The measurements differed in terms of the combination of single compounds included in analysis. Thus, we find different frequencies for 25 PAH compounds included in the PAH monitoring of ambient air in Germany. Descriptive characteristics of the 16 PAH compounds with sample size <70 are available in the technical report (Fertmann et al., 1999). The following description is confined to the single compounds with more than 70 measurements as prerequisite for statistical examination: BaA, BbF, BkF, BeP, BaP, INP, DBA, BghiP and COR (see Table 2).

Table 2 - Nine PAH compounds and their descriptives.

Full table (51K)

Comparing the statistics of central tendency and percentiles reveals left skewed distributions. Neither the BaP concentration in ambient air nor any of the other compounds is normally distributed, as most values are located in the lower range of the distribution (Kolmogorov–Smirnov goodness of fit, P0.0001). For statistical analysis, the data were submitted to logarithmic transformation; as ln-transformed values, these data are asymptotic normally distributed. Mainly, the deviations occur in the lower range, probably because of the heterogeneous origin of the measurements. Only the ln-transformed values of BaA and BeP reach normal distribution as confirmed by statistical testing (Kolmogorov–Smirnov goodness of fit, P0.20 and P0.07; more details in the technical report of Fertmann et al., 1999).

Bivariate intercorrelation within the matrix of the nine selected compounds (after logarithmic transformation) yields Pearson's correlation coefficients r_P within the range of 0.85–0.95, thus demonstrating a close statistical association. Table 3 shows the bivariate intercorrelation with respect to the marker compound, BaP, and the corresponding confidence interval.

Table 3 - Eight selected compounds in correlation with BaP.

Full table (37K)

Comparison of Subgroups: East–West Germany

The different political, economical and ecological history over the last 40 years in East and West Germany has influenced the environmental situation, e.g., coal heating remained an important component of air pollution in the East even during the last 10 years. Therefore, PAH compounds were analyzed after aggregation of the corresponding measurements from the East and the West. Table 4 gives an overview of East–West-related differences for the nine PAH compounds, Figure 2 by way of example a closer view on East–West-related differences for BaP.

Figure 2.

Distribution of BaP concentration in ng/m³ comparing East and West Germany. (*) Outliers with a distance of 1.5–3 times the height of the box to the 75th percentile. () Extremes with a distance of more than three times the height of the box to the 75th percentile.

Full figure and legend (14K)

Table 4 - Nine PAH compounds and their descriptives comparing East and West Germany.

Full table (45K)

Range and median of the observed eastern BaP measurements are higher (2.24 compared to 1.11 ng/m³; P0.0001), but the distribution of measurements taken in the West is characterized by many extreme values. Significant differences in central tendency apply as well to BaA, BbF, BkF, BeP, INP, BghiP (P0.001). The distributions of DBA and COR do not differ in the comparison of East and West. Altogether, the measurements' origin seems to affect the distributive pattern of the PAH compounds.

Comparison of Subgroups: Site Specificity

The classification for monitoring sites yields PAH measurements from rural, urban, traffic- and industry-dominated environments. Table 5 demonstrates the distributive patterns of the nine PAH compounds related to the location of the measurements. Additionally, Figure 3 shows by way of example the BaP concentrations dependent on the site specific classification.

Figure 3.

Distributions of BaP concentrations in ng/m³ comparing industrial, rural, urban and traffic-dominated sites. (*) Outliers with a distance of 1.5–3 times the height of the box to the 75th percentile. () Extremes with a distance of more than three times the height of the box to the 75th percentile.

Full figure and legend (17K)

Table 5 - Nine PAH compounds and their descriptives comparing the environmental classes — rural, urban, traffic- and industry-dominated sites — in Germany.

Full table (57K)

The lowest BaP concentrations are measured in rural environments; the median is 0.92 ng/m³. In comparison, the median values from urban (1.29) as well as traffic- (1.77) and industry (1.33)-dominated environments are higher (P0.0001). Variability of BaP concentrations seems especially high in the traffic-dominated environment. A similar pattern of differences in the distribution depending on site specificity is to be found for BghiP and COR; the other compounds show different patterns.

Comparison of Subgroups: Time

Because of the above-mentioned site- and region-determined heterogeneity, the descriptive analysis of time-dependent changes in the PAH concentration in Germany is restricted to the data of North Rhine-Westphalia; each year from 1991 to 1997 is represented by 4–5 measurements of traffic-dominated sites, 10–13 industrial sites, 17–22 rural sites and 19–21 urban sites (except in 1997, the measurements include only eight urban sites). Figure 4 shows the yearly median of the six single compounds BeP, BghiP, BaP, BaA, COR and DBA. BbF, BkF and INP could not be included as 65–85% of the measurements stem from 1997. Figure 5 gives a closer look on the time-related changes of the BaP concentrations.

Figure 4.

Median concentrations of BeP, BghiP, BaP, BaA, COR and DBA in ng/m³ from North Rhine-Westphalia, 1991–1997. Each substance with 45N122/year.

Full figure and legend (15K)

Figure 5.

Distribution of BaP concentrations in ng/m³ in North Rhine-Westphalia from 1991 to 1997. (*) Outliers with a distance of 1.5–3 times the height of the box to the 75th percentile. () Extremes with a distance of more than three times the height of the box to the 75th percentile.

Full figure and legend (28K)

All substances, except DBA, display a decrease during the observed time span. Based on the data from North Rhine-Westphalia from 1991 to 1997, a significant decrease in BaP can be determined (1991, N=51, median 1.6 ng/m³; 1997, N=45, median 0.7 ng/m³; P0.0001). The exemplary graph of BaP concentrations indicates a decrease from 1993 to 1994 compared to the years 1991/1992. No further time dependence can be seen in the following years. There are only slightly higher concentrations in 1996 compared to the previous and the following year. These data give us the impression that time has to be accounted for in multivariate analysis.

For the representative compound BaP, there are enough measurements from other federal states to have a look at time-dependent changes in BaP concentrations comparing East and West. Gradually, the BaP concentrations in the East approach the range of the levels in the West. An overlap of BaP concentrations in East and West Germany should be reached by now (graph omitted here).

Multiple Regression Analysis

The above comparison of subgroups verifies subtle differences in PAH concentrations with respect to environmental classification, the East–West division and time of observation. With simultaneous inclusion of all independent variables in statistical analysis, the results of the multiple regression analysis should yield the proportionate influence on the variability in PAH concentrations.

The dependent variables are the natural logarithms of the selected nine single compounds; the independent variables are site specificity, the East–West division and the year of measurement. To account for possible associations between the independent variables, additional tests (test for collinearity, residual analysis for normal distribution, Durban–Watson statistics for autocorrelation of the error terms) were applied. None of the resulting statistics warranted for restrictions in the interpretation. Measurements from the rural environment represent the background concentrations.

Table 6 shows that overall, 20–62% of the variability in the concentrations of the PAH compounds can be explained by the simultaneous impact of the described background information on the origin of the measurements. For BaP as the marker compound of PAH levels, 34% of the variability is to be delineated by the East–West affiliation, the site specificity and the year of measurement.

Table 6 - Results of the multiple regression analysis with nine PAH compounds as dependent variables and five independent variables.

Full table (59K)

The determination coefficients of BeP, BaP, INP, BghiP and COR are in a similar scale, BbF and BkF rank higher, DBA and BaA rank less. All dependent variables show a significant decrease over time. Also, all show significantly higher values in the urban environment and in East Germany. The site specification of traffic and industry is rather inconsistent in their predictive contribution for PAH concentrations. Traffic is the most important predictor for BghiP and COR, industry for BaA. Altogether, the available background information is an important aspect for the overall variability in PAH concentrations in Germany.

Thus, simultaneously, the five aspects influence background BaP concentrations — as observed at rural sites — to the following estimated proportions: a decrease of about 10% per year, an increase of about 31% for the urban environment, an increase of about 50% for the industrial site, an increase of about 40% for the traffic-dominated site and an increase of more than 50% for the East.

These multiple regression results underline the above-shown results from the comparison of subgroups. Especially in the western federal states and during the recent years, we observe lower BaP concentrations. In East Germany, they are more than twice as high compared to the West. Urban environment, traffic and industry contribute significantly to higher levels of PAH.

Factor Analysis

Factor analysis of the correlation matrix of the selected PAH compounds is aimed to define one or several latent background variables. The variables and the correlations on the surface are reduced to virtual factors. The pertaining factor analysis concentrates on examination of the correlation matrices and factor extraction.

Table 7 shows high correlations within the matrix of the nine single compounds; only 3 of 36 bivariate combinations rank below 0.5. The bivariate combinations have different sample sizes. We take into account only those correlation coefficients based on more than 390 value pairs and omit all other combinations. The single compounds BbF, BkF and INP have to be dropped from analysis. The resulting correlation matrix is shown in Table 8.

Table 7 - Correlations within the ln values of the nine PAH compounds; each cell with Pearson's correlation coefficient.

Full table (29K)

Table 8 - Correlation matrix of the ln-transformed values of six PAH compounds with 399 value pairs included in factor analysis.

Full table (25K)

The correlation matrix of these six compounds was subjected to factor analysis with the principal component method. The series of eigenvalues is 5.5, 0.3, 0.1, 0.04, 0.03 and 0.01, with 92% explained variance by the first factor.

Factor loadings describe the correlation between the factor(s) emerging from the factor analysis and the original variables used in the construction of the factor. As to be seen in Table 9, here the result of analysis is a one-factor model. This one factor shows high loadings on all variables included (0.898–0.988). These results were examined with split sample techniques and proved to be consistent.

Table 9 - Results of the factor analysis.

Full table (11K)

Discussion

This investigation was aimed to explain the variability of data on PAH concentrations in ambient air gathered under fieldwork-like circumstances in Germany over almost the past decade. The explorative descriptive analysis of 25 PAH compounds offers an informative overview and allows a comparison of the results of PAH monitoring.

In addition to the classical statistical analysis of exposure data by various descriptive procedures, factor analysis was applied here in an exploratory sense. It proved to be a useful tool; yet, the reliability of these results has to be tested with a comparable data set. Thus, the following conclusion is clear, but tentative: The high intercorrelations within the subset of BaA, BbF, BkF, BeP, BaP, INP, DBA, BghiP and COR and the one-factor solution for the subset of BaA, BeP, BaP, DBA, BghiP and COR lead to a straightforward interpretation: PAH patterns in ambient airare rather homogenous and can be represented by a single compound, as far as relevant sites are adequately represented.

With respect to the typical emission situation for ambient air, this investigation relies on rural, urban, traffic- and industry-dominated sites. Obviously, these different emission sources do not result in profound differences in the concentration pattern in the subset of compounds considered in analysis. PAH patterns in other European countries might show distinguishable features, e.g., as a consequence of climate influences or differences in technology induced by specific national legislation in the past. Furthermore, the results of this study cannot be applied to PAH emission situations other than ambient air as there are indoor sources like aging building materials (e.g., parquet glue; Heudorf and Schubert, 2000).

With the six single compounds included in factor analysis — BaA, BeP, BaP, DBA, BghiP and COR — also the most relevant compounds with respect to effects on human health are covered. Except COR, all these compounds are carcinogenic in animal experiments (IARC, 1987), with BaP and DBA being the most potent compounds (Nisbet and LaGoy, 1992; Petry et al., 1996). BaP has been used before as an indicator for the mixture of PAH exposure assessing the cancer risk for humans (e.g., coal oven workers; Pott, 1985; WHO, 1987; WHO, 1995).

However, the results of this investigation cannot be interpreted without reservations: The observations and concentration data included in the statistical analysis do not rely on a systematic, representative investigation. The values stem from rather nonsystematically taken measurements, triggered by actual circumstances, not standardized by nationwide routines. As the data were delivered as annual means, differences between seasons could not be detected. Especially the impact of coal heating — predominating in East Germany — would be interesting to test for seasonal aspects. Yet, the data include the most important emission sources — traffic, industry and urban environment, including coal heating. The collection of all nationwide available PAH concentrations supplied by the federal states covering the past time span of about 10 years offers the best possible information to develop a useful concept for PAH monitoring in the near future. This may be of particular interest connected with the installation of a European directive on air quality regarding PAHs, as the fourth daughter directive will be set in force presumably in 2002.

Based on these results, it seems justified to rely on BaP measurements as representative indicator of the PAH contamination in ambient air. Because of the high intercorrelations, the concentrations of the other carcinogenic PAHs included in routine monitoring can be estimated with sufficient confidence.

Finally, this analysis offers some interesting insights on the origin of the heterogeneity of the PAH patterns. For most single compounds, the variability in concentrations could be attributed to time- and site-specific characteristics, thereby explaining the variability to a surprisingly high degree. Higher concentrations in the East German federal states, the decrease in concentrations over the past 8 years and subtle differences in the concentrations of BaA, BeP, BaP, DBA, BghiP, COR, BbF, BkF and INP could be related to the respective dominant emission source.

Although we had to rely on rather rough and not standardized site characterization, the environmental site specification gave valuable insights to the causes of variability. Urban environment is characterized by higher BbF and INP levels, traffic-dominated sites by higher BghiP and COR levels, and industry by rather higher DBA levels. These observations are in line with earlier detailed investigations on PAH patterns and their sources by Menichini (1992). The results of multiple regression analysis underline the simultaneous effects of time, East–West differences and site specificity, and explain a considerable proportion of the variability in the patterns of PAH concentrations.

In order to further confirm the results and conclusions from this study, it may be helpful to carry out similar investigations in other European countries. Possibly, improved qualitative and eventually quantitative site specification of actual measurements, as well as the exploitation of nonaggregated data, would allow a more sophisticated analysis of the imminent effects. Pooling of measurements might be useful to enlarge the data base. This might also provide an opportunity to include more PAH compounds in factor analysis than in the present investigation.

References

1. Brosius G Brosius F, 1995 SPSS Base System and Professional Statistics. International Thomson Publishing, Bonn 1995
2. deRaat WK de Meijere FA Polycyclic aromatic hydrocarbon (PAH) concentrations in ambient airborne particles from local traffic and distant sources; variation of the PAH profile, Sci Total Environ (1995) 103(1): 1–17
3. Fertmann R Tesseraux I Schümann M Neus H Prinz B, 1999. Auswertung der in der BRD vorliegenden Immissionsmessdaten von polyzyklischen aromatischen Kohlenwasserstoffen (1990–1998), Bericht im Auftrag des Unterausschusses Wirkungsfragen des Länderausschusses für Immissionsschutz (LAI), Hamburg (1999)
4. Gladtke D Air pollution in the Rhine-Ruhr area, Toxicol Lett (1998) 96: 277–283
5. Grimmer G Vergleich der PAH-Profile aus Umweltproben, VDI Report 358. VDI-Verlag, Düsseldorf 1980 39–50
6. Heudorf U Schubert W Innenraumbelastung mit polycyclischen aromatischen Kohlenwasserstoffen (PAK) durch PAK-haltige Parkettklebstoffe-Hinweise der Projektgruppe Schadstoffe der Bauministerkonferenz, Umweltmed Forsch Prax (2000) 5: 341–344
7. International Agency for Research on Cancer (IARC) 1987. Overall evaluation of carcinogenicity. An updating of IARC Monographs Volumes 1 to 42, IARC Monogr Eval Carcinog Risk Man Suppl 7. Lyon, France 1987
8. IPCS International Program on Chemical Safety Selected Non-Heterocyclic Polycyclic Aromatic Hydrocarbons, Environmental Health Criteria 202. WHO (World Health Organization), Geneva 1998
9. Kleinbaum DG Kupper LL Muller KM Applied Regression Analysis and Other Multivariable Methods 2nd EdPWS-KENT Publishing, Boston 1988
10. Lighty JS Veranth JM Sarofim AF Combustion aerosols: factors governing their size and composition and implications to human health, J Air Waste Manage Assoc (2000) 50(9): 1565–1622
11. Mackay D Hickie B Mass balance model of source apportionment, transport and fate of PAHs in Lac Saint Louis, Quebec, Chemosphere (2000) 41(5): 681–692 | Article | PubMed |
12. Menichini E Urban air pollution by polycyclic aromatic hydrocarbons: levels and sources of variability, Sci Total Environ (1992) 116: 109–135 | PubMed |
13. Nisbet ICT LaGoy PK Toxic equivalency factors (TEFs) for polycyclic aromatic hydrocarbons (PAH), Regul Toxicol Pharmacol (1992) 16: 290–300 | PubMed |
14. Petry T Schmid P Schlatter C The use of toxic equivalency factors in assessing occupational and environmental health risks associated with exposure to airborne mixtures of polycyclic aromatic hydrocarbons (PAH), Chemosphere (1996) 32: 639–684 | Article | PubMed | ISI | ChemPort |
15. Pott F Pyrolyseabgase, Profile von polycyclischen aromatischen Kohlenwasserstoffen und Lungenkrebsrisiko-Daten und Bewertung, Staub-Reinhalt Luft (1985) 45: 369–379
16. Prinz B Stratmann H Anwendungsmöglichkeiten der Faktorenanalyse bei Immissionsuntersuchungen, Staub-Reinhalt Luft (1968) 28: 25–30
17. United States Environmental Protection Agency (USEPA). Provisional Guidance for Quantitative Risk Assessment for Polycyclic Aromatic Hydrocarbons (1993) EPA/600/R-93/089
18. VDI Guideline 3875 Outdoor-air pollution measurement. Indoor air pollution measurement. Measurement of polycyclic aromatic hydrocarbons (PAH). Gas chromatographic determination. Sheet 1, VDI (1996)
19. World Health Organization (WHO) Air quality guidelines for Europe. WHO regional publications European series, no. 23WHO Regional Office for Europe, Copenhagen 1987
20. World Health Organization (WHO) Updating and Revision of the Air Quality Guidelines for Europe, Brussels, Belgium, October 2–6 (1995)
21. World Health Organization (WHO) Guidelines for drinking-water quality, Addendum to Volume 2 "Health criteria and other supporting information" 1998 WHO/EOS/98.1

Acknowledgements

The authors thank H. Blunck and M. Wessel for their technical assistance and the anonymous reviewers for their valuable suggestions.