COUNCIL DIRECTIVE ON AMBIENT AIR QUALITY ASSESSMENT AND MANAGEMENT WORKING GROUP ON BENZENE

POSITION PAPER
September 1998 COUNCIL DIRECTIVE ON AMBIENT AIR QUALITY ASSESSMENT AND MANAGEMENT WORKING GROUP ON BENZENE

1

POSITION PAPER PREFACE Scope Summary

1. INTRODUCTION 1.1 Benzene in the atmosphere 1.2 Emissions of benzene 1.3 Atmospheric chemistry and transport 1.4 Benzene ambient air concentrations 1.5 Current National Standards and Guidelines 1.6 Summary 1.7 References

2. RISK ASSESSMENT Scope 2.1. 2.2 2.3. 2.4. 2.5. 2.6

Human exposure to benzene Health Effects National and other Health-based Air Quality Guidelines Evaluation of human health risks Recommendations for developing limit values Reference

2. ASSESSMENT METHODS Scope 3.1 Introduction 3.2 Description of monitoring methods 3.3 Network design and siting criteria 3.4 Other assessment techniques: air quality modelling 3.5 Data quality objectives 3.6 Random or continuous measurements 3.7 Information on air quality to the public 3.8 Upper and lower assessment thresholds 3.9 Summary 3.10 References

2

4 ECONOMIC ANALYSIS 4.1 Introduction 4.2 Results of economic analysis 4.3 Discussion

5 DISCUSSION AND FINAL RECOMMENDATIONS 5.1 Summary of Chapters 1-4 5.2 Considerations to be taken into account in setting limit values for benzene 5.3 Options for consideration 5.4 Reviewing limit values

3

COMMISSION OF EUROPEAN COMMUNITIES

Council Directive on Ambient Air Quality Assessment and Management Working Group on Benzene

Benzene: Preface

4

Preface
Scope At the meeting of the Steering Group of National Experts on Air Quality in July 1995 it was agreed that woring groups would be established to produce position papers in preparation for development of daughter legislation under the Council Directive 96/62/EC on Ambient Air Quality Assessment and Management. The present position paper is developed for Benzene by a working group with participants from Denmark, Germany, Italy, Joint Research Centre (Ispra), The Netherlands, Spain, Sweden, European Environmental Agency (Topic Centre for Air Quality), from Industry, from the European Environmental bureau and from the European Commission (DG XI). Italy is responsible for the chairmanship and co-ordination of the Working Group. The position paper must be prepared in accordance with the recommendations from the Commission (AAQ/95/1/2) with the amendments from the meeting on 5 July 1995. It will provide a technical basis for establishing limit values and air quality objectives over the EU territory for the protection of human health, ecosystems and the environment and for achieving an effective reduction of Benzene pollution taking into consideration costs and benefits. The position paper will be published as a stand alone technical paper to be used as a technical reference. This paper addresses air quality objectives for the outdoor ambient atmosphere and it does not address indoor air pollution, deposition or synergistic effects between benzene and other pollutants. Even though parts of the population may be exposed to significant indoor and/or workplace benzene pollution, this is outside the scope of the Directive and will not be considered in the position paper. However, as with other common outdoor pollutants, benzene may be present at high concentration levels in indoor environments such as house interiors, inside cars and the working place. In addition, Benzene is a category one carcinogenic compound. It is not expected to present a potential risk of acute effects, and therefore only long term limits air quality objectives are planned.

5

Summary Chapter 1 of the position paper describes the emission sources of Benzene in atmosphere giving some quantitative data for Europe and for other industrialised countries. Emphasis is given to the emission by vehicles, especially in urban locations where it is expected to reach the highest concentration. Emissions are resulting from direct emission from the exhausts and from the evaporation of fuels either by car or from fuel distribution and refuelling. Factors controlling the concentration levels, such as the emission rates and meteorological factors, e.g. mixing properties of the atmosphere, are also presented. The physico-chemical evolution of Benzene after the emission step is also presented. It includes atmospheric chemistry of aromatic compounds related to photooxidant formation and the main processes for transport and deposition. Current air concentrations in several locations are reviewed. They include mainly conurbation, but industrial sites and remote observations are also reported. In order to show a decreasing trend in benzene concentration, some data relevant to past years are included. The impact of Benzene on the environment and especially on human health is reported in Chapter 2. Toxicological data are examined and epidemiological studies on population exposed to Benzene are presented and discussed to give the basis to assess the risk assessment as a function of exposition. Recommendations are made for a basis for developing limit values. Chapter 3 of the position paper addresses methods and tools for the measurements and assessment of Benzene concentrations in environments where a risk of exposure is present. A detailed description of available instruments for the automatic monitoring as well as manual sampling and analytical methods are presented. A network design for Benzene is suggested and the inclusion of Benzene monitors in existing monitoring stations is reviewed. Quality assurance and quality control are presented in order to ensure data of sufficiently high reliability in terms of accuracy and sensitivity. Preliminary assessment techniques are also presented. They are mainly based on the use of proper models which take into account emission rates, traffic distribution and prevalent meteorological conditions. Accuracy and reliability of existing models for Benzene are reviewed. Cost and benefit implications of Benzene reduction are considered in Chapter 4. Finally, Chapter 5 summarises the main points of the previous Chapters and makes final recommendations. 6

Chapter 1

Benzene: Introduction

7

1. INTRODUCTION 1.1 - Benzene in the Atmosphere
Benzene is an aromatic hydrocarbon with molecular formula C6H6 (PM = 78.11). It is a colourless liquid at ambient temperature with a boiling point of 80.1 °C at atmospheric pressure, and a vapour pressure of 10 kPa at 20° C and about 12.27 kPa at 25 °C. Benzene vapour has an "aromatic" odour and an odour threshold of 1.5 ppm (v/v). It is slightly soluble in water (1,8 g/Kg at 25 °C). Benzene is a Volatile Organic Compound (VOC). This is a class of compound, which includes chemical species of organic nature such as alkanes, alkenes, aromatics, ketones, aldehydes, alcohols and others. They are characterised by a vapour pressure at ambient temperature higher than 100 Pa, so that most are found in gas phase. They are commonly present in air at concentrations ranging from a few micrograms to many hundreds of micrograms per cubic meter, depending basically upon the emission sources, meteorological factors, transport and photochemical phenomena. Emissions of benzene to the atmosphere are due mainly combustion processes for energy production (including motor vehicles) and domestic heating. Since fuels are distributed from the producers to the user, evaporative processes are also responsible for the emission of large amounts of aromatic hydrocarbons in the atmosphere. The world-wide industrial production of benzene is in the order of more than 15 million tons per year. In 1991-93 the European production of benzene, excluding benzene in petrol, was 5.5 million tons. The only significant natural sources of benzene are biomass burning, and brush and forest fires. However, these sources do not affect air quality in densely populated areas in the EU.

1.2 - Emissions of benzene
Benzene in the atmosphere is due mainly to anthropogenic sources. Natural sources are estimated in the order of 3-5% while more than 90% are estimated to come from anthropogenic sources. The anthropogenic sources in Europe are distributed according to Table 1-I, where an estimation of different contributions to the total emission of benzene is reported.

8

Tab. 1-I - Contribution of main anthropogenic sources of Benzene in Europe Sources Vehicular traffic Petroleum Refineries Fuel Distribution Chemical Industry Domestic Heating Solvent Use (%) 80-85 0.3-1.5 2.6-6 1.3-13 3-7 1-4

Source: Air Quality Report of the Auto Oil Programme (data for 1990) The most important and significant anthropogenic sources are combustion sources, where benzene is emitted as unburned compound. However, benzene is also formed from the thermal degradation of other aromatic compounds. Main contributors are traffic and other mobile sources and energy production (including industrial combustion). Emissions from traffic are the largest source. Therefore, benzene pollution is highest in densely populated areas characterised by high traffic density. Industrial activity can contribute to ambient benzene levels, but its contribution to the total exposure is very low. Unfortunately, national emission data based on emission inventories may not be very accurate, and data from different Member States may not be directly comparable owing to different classification systems and methods of calculation. The results reported here are therefore subject to uncertainty. Since benzene emissions from traffic are not directly regulated (though benzene emissions are reduced by catalytic converters), standard venicle emission tests do not include benzene as such. In research experiments benzene emissions have been determined, and the results have been extrapolated to car fleets, based on an analogy with hydrocarbon emissions, which are measured in standard tests. The uncertainty of benzene emissions from industrial sources depends on the source type. Evaporative emissions from handling and storage of oil and benzene are uncertain and very difficult to measure. Often the benzene content of evaporated solvents is not known and a default percentage is used as an estimate. Clearly, better emission inventories are key elements for an effective benzene reduction program.

9

1.2.1 - National Emission data
Emission data from individual European countries have been gathered from national emission inventories and are reported in the following Tables.

1.2.1.1 - Sweden
In Sweden, the estimated total amount of emitted benzene in 1994 was about 23 KTonnes. The distribution of benzene emissions from different emission categories is given in the Table 1-II

Table 1-II - Sweden, benzene emissions from different source categories given as % of the total benzene emission

Source Category

%

Road traffic, including petrol Evaporation and asphalting Other mobile sources Working machinery Energy production Small scale wood Combustion for heating Industry Domestic use Pesticides

38 3 4 1 (depending on the size and position of residential quarters) No = 1 (supplemented by indicative measurements and/or modelling)

95

Rural Background

Rural sites are of no priority for compliance checking as such. However, to obtain reference data from relatively unpolluted air masses and to assess regional transport phenomena, it is recommended that levels should be monitored at some background sites per member state. Again, the approach chosen by the working group on particulate seems to be applicable also for benzene. Consequently, the following number Ni is recommended for rural areas:
Ni (rural) = A/50.000 (A = the area of the country in square kilometres) With full assessment, this figure may be reduced to No (rural) = 1 per Member State.

Table 3.II summarises the minimum numbers of measurement points proposed by the working group for the different micro-environments:
Table 3-II

Site Urban background, per agglomeration or zone

Ni
4 I , I= number of inhabitants in

No 1 1 1+ P , P = population of Member State (million) 1 (continuous)

millions

Rural background, recommended A x 50,000-1, A = area of per Member States Member State in km2 Roadside, per agglomeratioin or zone Industrial 1 additional site per 1 million inhabitants or more (1-10) > 1 (continuous)

Micro scale considerations for site locations Most criteria for the microscale siting of measurements, which were given by the working groups for other pollutants, apply for benzene as well.

96

- Benzene should be sampled within an unrestricted air flow without any obstructions in the vicinity of the sampler (e.g. a minimum of 1 metre away from buildings, trees, etc.) - Great care should be taken not to sample the exhausts or evaporative losses of measurement vans or from the outlet of stations, pumps, etc. - The inlet sampling point should be between 1.5 m and 4 m above the ground. In the case of background measurements (urban, rural) higher positions up to ten metres may be considered, if long path measurements (e.g. DOAS) are applied. - Roadside sites should be positioned at the building line of residential buildings adjacent to busy streets. - In order to save costs, benzene sampling (also grab sampling) can be colocated with measurement stations for other pollutants, especially for background measurements. - The sampling sites should be fully documented according to the Guidance document of EEA and ERLAP. In addition, interfering sources, security, access, electrical power and telephone, "visibility" of the site - public profile, safety of public and operators, planning requirements, and others, should also taken into account.

3.4 - Other assessment techniques: air quality modelling
Assessment of air quality is an important issue in many areas, especially in the urban environment. The most credible, complete and useful assessment is always obtained by combining measurements and modelling results. The monitoring at chosen sites provides data for the pollution situation and its temporal variability. Dispersion model extends the knowledge by computing concentration fields as a function of time at various spatial scales. Under the Directive, if the air quality is close to the limit value, modelling alone can not and may not be used to assess the air quality situation and exceedances of air quality standards. Models are also useful for the - exploratory assessments in absence of measurements and for planning of measurements and location of monitoring sites; - understanding the pollution situation in relation to factors such as special dispersion conditions and to temporal and spatial emission variability; - relating air pollution to sources and sectors as input information for abatement; - calculating future pollution levels as a consequence of future emission scenarios.

97

A variety of air pollution models have been developed for various spatial and temporal scales of urban air pollution situations, and for the various applications mentioned above. The models can be separated into four classes: - models for pollution near streets and roads; - models for pollution near point sources; - integrated urban scale models, taking account of all source categories, giving concentrations in a grid net and in selected receptor points; - population exposure models. For the evaluation of possible air quality standards exceedances, the integrated urban models are the most useful. Single point source models will only provide information on the contribution from a point source, which is of limited use for the total assessment. Since traffic emissions to a great extent will determine the air quality of benzene, street environment models are useful to assess the zones with risk for exceedances, and for the evaluation and extrapolation of measurement results to larger zones. This type of model is also relatively simple and easy to use for an unskilled user. The results are however, only indications of the real situations and must be evaluated further using monitoring data and more advanced models. There are a number of well-established models in use in urban areas in different parts of Europe. No general recommendation of models can be made here. However, the model must fulfil the requirements and be thoroughly validated against monitoring data. The necessary initial requirements relevant to the successful application of a model focus on : • input data accuracy and its dependency to the model output; • proper model parameterisation and initialisation; • methodologies for judging to what extent model results represent reality. As a consequence of the above, model validation (typically by the aid of available analytical solutions) should be considered as an indispensable part of the model development process, whereas an already validated model should be subject to a genuine evaluation procedure in order to ensure that potential users can assess the degree of reliability and accuracy inherent in the given model (Coseman et al., 1992).

98

An increasing demand of air quality assessment is foreseen and will be the driving force towards further development and improvement of dispersion models. Various models are needed. Operational models are an important tool for city planning. On-line models are used for air pollution forecasting at pollution episodes to show present and nearfuture risks. Models for regulatory purposes are used to demonstrate zones of compliance/non-compliance with air pollution directives. To be powerful tools, the models must fulfil certain quality criteria. Work on quality assurance of models is important and has recently started within different organisations. Accuracy can be documented in terms of uncertainties in input quantities and in model formulation, and could be demonstrated in the validation process. Input, in the form of emissions and dispersion conditions, is necessary at a similar spatial and temporal scale as the output concentrations to be calculated by the model. Data on emissions is required with sufficient spatial and temporal resolution to describe the important gradients occurring in a zone.

Local Scale Models

The following table 3-III gives a rough overview of a number of existing local scale models and their characteristics.

Table 3-III Model name Model category

AUSTAL 86 CAR CAR-Parking PHOENICS CAR-FMI CONTILINK CTDMPLUS HPDM IFDM

Gaussian Semi-empirical Semi-empirical Eulerian Gaussian Gaussian Gaussian Gaussian Gaussian

99

INPUFF ISCST 2 OSPM PLUIMPLUS ROADAIR SCALTURB UK-ADMS

Gaussian Gaussian Street pollution model Gaussian Gaussian Gaussian Gaussian

The state of the art for these models is described in details elsewhere. Among these, is important to cite the CAR-Parking model. It is developed to calculate benzene concentrations in the vicinity of the parking garages. Used data are about usage, number of cars, speed, duration of stay, building dimensions, ventilation and type of surroundings. The model includes a module for calculating the emission from the garage. The model can be used for 70-80% of the configurations of the garages. The model is validated by a measurement campaign around 3 parking garages in the city of Utrecht, Netherlands (Boeft et al.; 1996; Zannetti et al.; 1993). Furthermore, the OSPM is developed to describe the

pollution in street canyon. Input data to this model is meteorological data, emission factors from cars and concentrations above roof tops. The model describes the flow and dispersion in a street canyon (Berkowicz et al. 1997).

3.5 - Data quality objectives
Data quality objectives must be established in order to comply with the monitoring objectives. They will be defined in terms of required precision and accuracy, minimum time coverage and minimum data capture. The precision and accuracy are expressed as maximum uncertainty on the basis of the results of the benzene. The major constituents of a quality assurance programme concern: - network design number of stations, siting criteria; - measurement technique, sampling, analytical and calibration procedure; - equipment evaluation and selection: validation of methods, test of instrument performances;

100

- routine site operation: calibration under field condition, maintenance, management and training.

An acceptable error for mandatory measurements is expected to be about 20% (#this is very low see comments by Ballesta). This figures takes into account various deviations, including: - the losses due to sampling (5%); - the calibration errors (primary calibration, transfer standard, routine calibration under fields conditions): ± 10%; - the analytical errors (noise, non-linearity of the instruments, atmospheric pressure, temperature): ± 5%. Non mandatory or indicative measurements may be carried out with lower accuracy and precision. For example, in the case of passive samplers, typical errors of 20% and even more are easily experienced. However they are still acceptable, providing that the expected concentration levels are sufficiently far below the recommended standard. Another important requirement for benzene analysis resides in the general structure of quality assurance within individual Member States. This requires, at least, that measurements should be carried out by accredited laboratories in accordance with approved methods. The network management structure must be able to fulfil the following requirements: - national focal point; - national co-ordination of data; - harmonised methodology for statistical evaluation; - minimum data capture; - standardised reporting directive; - exchange of information decision.

101

The calibration and quality assurance should be based on common requirements in the Member States. Generic capability transfer and harmonisation must include clear and detailed operational manuals and guide books on: - site operation; - Quality Assurance/Quality Control on samplers and analytical equipment; - data handling - training, workshops, information exchange/technology transfer. As far as the accuracy and precision of the applied method are concerned, the Working Group has agreed that the overall figures should be: Mandatory measurements : 20% of the limit value Indicative measurements : 30% of the limit value Modelling estimation : 50% of the limit value Objective estimation : 100% of limit value

3.6 - Random or continuous measurements
A complete resolution in time of a series of measurements can only be obtained by continuously operating instruments. This is mandatory, when it comes to the monitoring of short time peak levels of the pollution in real time as, for example, for smog alert purposes. Continuous measurements are, however, quite costly. Because of high investment and running costs, networks with only continuously operating instruments with on site analyses tend to have few stations and consequently a rather poor resolution in space. On the other hand, random sampling with subsequent analysis in the laboratory is cheep enough to allow for more sampling points in a certain zone, thus giving a better resolution in space at the expense of an additional uncertainty in time resolution. Especially for compounds like benzene with effects from chronic exposure and hence an assessment on the basis of annual averages, at random sampling and analysis offers a cost effective alternative to continuous measurements, provided that the sample size is big enough to yield an acceptable overall accuracy. The working group thinks that an expanded uncertainty of ± 10 % (on the 95 % confidence level, that is 2 times the standard deviation) compared with continuous measurements is acceptable for compliance checking. It should be borne in mind that this

102

additional uncertainty refers to the poorer time resolution, but may be out-balanced by a better representativeness in space, if several sampling points are chosen. Consequently, at random measurements for benzene are equivalent to continuous measurements, provided that • the sample size is big enough to yield an expanded uncertainty compared with continuous measurements ≤ ± 10 % (95 % confidence level) • the sampling dates be evenly spread during the measurement period (normally one year), which can be easily realised if they are chosen at random. The necessary sample size to keep the expanded uncertainty below ± 10 % can be calculated by resampling of stochastic samples from a continuous time series of monitoring data, in this case for benzene. Three time series from urban background sites (Eschweiler, Brühl and Köln-Eifelwall) and two series from industrial hot spots (DuisburgBruckhausen and Köln-Godorf) were chosen for this exercise. In the cases of Köln-Eifelwall and Köln-Godorf, time series from three years (1994 - 1996) were considered. Resampling was performed by means of a computer program with replacement and without stratification. From each time series, a total of 100 stochastic "bootstrap" samples of the considered size N were taken on the basis of daily means (24-values) and 1hvalues. For each bootstrap sample, the statistic of interest (arithmetic mean) was calculated. From the resulting bootstrap distribution of the mean value, the overall uncertainty including the uncertainty due to stochastic sampling (standard deviation) can be calculated (compare Table 3-IV). It can be concluded from the table that samples of 104 daily means or more from urban background sites have standard deviations < 5 % and thus expanded uncertainties ≤ 10 %, whereas a sample size of 360 1/2h-values is necessary to reach this accuracy. In addition, Table 3 - V shows that in the case of the time series from industrial hot spot with strong concentration fluctuations in time, more than 200 daily means are necessary, so that random measurements do not offer an advantage any more. The industrial hot spots represent a cokery (Duisburg-Bruckhausen) as well as a distribution terminal in the harbour of big petrochemical plants.

103

Figure 3.2 - Results from passive sampling of benzene (Sept. 96 - June 97; numbers in circles) and from diffusion and modelling (isolines) of cokery emissions for the same period (µg/m3). Sampling was performed in a 250 x 250 m grid in a residential quarter east of the cokery. In addition, some more distant sites representing the urban background were measured. Continuous active monitoring was performed at the site represented by a rhomb.

104

Table 3-IV: Results of resampling of random samples of benzene from continuous measured time series. Daily and ½-hourly means

Data: Cologne - Eifelwall; 1996; benzene; urban background 24-hourly values Mean value Standard deviation 3 [µg/m3] [µg/m ] 2.65 0.40 2.70 0.32 2.71 0.21 2.70 0.13 2.69 0.09 2.69 0.06 2.69 0.00 ½-hourly values Mean value Standard deviation 3 [µg/m ] [µg/m3] 2.62 0.54 2.76 0.43 2.75 0.33 2.74 0.21 2.69 0.21 2.69 0.15 2.73 0.12

Sample size N 13 26 52 104 156 208 322 365

Standard deviation (%) 15.11 11.80 7.67 4.81 3.20 2.23 0.00 -

Standard deviation (%) 20.50 15.62 12.04 7.60 7.61 5.68 4.50

(Resampling from 322 valid daily mean values) hourly values)

(Resampling from 3134 valid 0.5

Data: Hannover-Linden; 1996; benzene; road side 24-hourly values Mean value [µg/m3] 9.53 9.28 9.24 9.27 9.26 9.27 ½-hourly values Mean value [µg/m3] 9.45 9.18 9.23 9.20 9.26 9.26

Sample size N 13 26 52 104 156 208 365

Standard deviation [µg/m3] 0.872 0.611 0.440 0.276 0.150 0.000

Standard deviation (%) 9.20 6.60 4.80 3.00 1.60 0.00

Standard deviation [µg/m3] 1.676 1.087 0.832 0.565 0.412 0.296

Standard deviation (%) 17.70 11.80 9.00 6.10 4.40 3.20

(Resampling from 365 valid daily mean values)

(Resampling from 17569 valid 1/2 hourly values)

105

Data: Cologne-Gordof; 1996; benzene; industrial site 24-hourly values Mean value [µg/m3] 4.50 4.57 5.38 4.79 4.75 4.81 4.82 ½-hourly values Mean value [µg/m3] 4.24 4.42 4.15 4.46 4.41 4.36 4.47

Sample size N 13 26 52 104 156 208 318 365

Standard deviation [µg/m3] 2.54 2.13 1.99 1.00 0.74 0.54 0.00 -

Standard deviation (%) 56.34 46.46 37.01 20.80 15.55 11.16 0.00 -

Standard deviation [µg/m3] 2.12 2.28 1.24 1.33 0.88 0.73 0.57

Standard deviation (%) 50.11 51.59 29.90 29.76 19.93 16.76 12.72

(Resampling from 318 valid daily mean values) hourly values)

(Resampling from 3135 valid 1/2

Table 3-V:

Results of resampling of random samples of benzene from continuous measured time series. Weekly means

Hannover, 1990, road side Mean values Standard deviation [µg/m3] 1.10 0.57 0.00

Sample size N 13 26 52

Mean value [µg/m3] 17.67 17.65 17.65

Standard deviation (%) 6.20 3.30 0.00

Hannover, 1993, road side Mean values Standard deviation [µg/m3] 0.98 0.49 0.00

Sample size N 13 26 52

Mean value [µg/m3] 12.43 12.37 12.40

Standard deviation (%) 7.90 3.90 0.00

Hannover, 1994, road side Mean values Standard deviation [µg/m3] 0.55 0.32 0.00

Sample size N 13 26 52

Mean value [µg/m3] 11.81 11.82 11.83

Standard deviation (%) 4.60 2.70 0.00

106

Hannover, 1996, road side Mean values Standard deviation [µg/m3] 0.55 0.27 0.00

Sample size N 13 26 52

Mean value [µg/m3] 9.90 9.96 9.92

Standard deviation (%) 5.50 2.70 0.00

Cologne-Eifelwall, 1996, urban background Mean values Standard deviation [µg/m3] 0.24 0.15 0.00

Sample size N 13 26 52

Mean value [µg/m3] 2.68 2.64 2.67

Standard deviation (%) 8.80 5.60 0.00

Cologne-Eifelwall, 1996, industrial site Mean values Standard deviation [µg/m3] 0.62 0.36 0.00

Sample size N 13 26 52

Mean value [µg/m3] 4.29 4.27 4.24

Standard deviation (%) 14.40 8.30 0.00

The results shown in the Table above allow the following statements: 1. If the number of 24-hourly benzene measurements taken in a measurement period (1 year) is reduced from 365 to N = 104 randomly distributed daily samples, the resulting standard deviation of the yearly mean value due to random sampling would be less than 5 % at sites in urban background. Close to heavy traffic carrying roads, the number of random 24-hourly measurement may even be reduced to N = 52 with the same effect. If the number of ½-hourly benzene measurement taken in a measurement period (1 year) is reduced from 17520 to N = 208 randomly distributed samples, the resulting standard deviation of the yearly mean value due to random sampling would be less than 5 % at sites close to heavy traffic carrying roads. In the urban background, the resulting standard deviation of the yearly mean value due to random sampling would be 5.7 % at the monitoring site considered.

2.

107

3.

4.

5.

6.

If the number of weekly means of benzene measurement in a measurement period (1 year) is reduced from 52 to N = 26 randomly distributed samples, the resulting standard deviation of the yearly mean value due to random sampling would be less than 5 % at road sides or slightly above 5 % (5.6 %) at urban background sites. Close to industrial hot spots with strong fluctuation of benzene concentration in time random measurements should not be recommended for compliance checking. The industrial hot spot considered represents a distribution terminal in the harbour of big petrochemical plants. Similar results were obtained by an analysis of a short time series measured in the vicinity of an old cokery. For indicative measurements in urban background and close to traffic hot spots, a minimum of N = 52 randomly distributed measurements of ½ or 1 hour sampling period is recommended in the measurement period (1 year). This would lead to a standard deviation of resulting yearly mean values due to random sampling of less than 12.5 %. Considering 24-hourly measurements, minimum of N = 26 randomly distributed samples is recommended for indicative measurements in urban background and close to heavy traffic hot spots. This would lead to a standard deviation of resulting yearly mean values due to random sampling of less than 12.5 %. For this exercise, the following recommendations of the working group can be

extracted: Urban sites, road side 104 (24 h); 208 (½ h or 1 h), 26 (weekly samples) 26 (24 h); 52 (½ h or 1 h), 13 (weekly samples) Industrial hot spots continuous or quasicontinuous monitoring 208 (24 h); 365 (½ h or 1 h), 13 (weekly samples)

For compliance checking Indicative measurements

Reliable results for indicative measurements (expanded uncertainty ≤ 25 %) can already be obtained with a small sample size of ≥ 26 daily means or ≥ 52 1h-averages for urban sites. From this exercise, it can be concluded that the following sample sizes N are necessary: urban sites for compliance checking indicative measurements 104 (24 h); 360 (1 h) 26 (24 h); 52 (1 h) industrial hot spots continuous or quasi-continuous monitoring 104 (24 h); 250 (1 h)

108

In conclusion, the working group also agrees that the minimum time coverage of the measurements should be: - almost continuous measurements: 100%; - indicative measurements: 50% (every second day, or at random, or 6 months per year by example). A lower coverage can be discussed according to the concentration levels and possible time trends of atmospheric pollution and other factors.

3.7 - Information on air quality to the public
All information on air quality shall be available to the public and mass media. As means to involve the public in the efforts to fulfil the aims of the Framework Directive, responsible authorities shall actively and regularly disseminate information on the air quality. Appropriate ways can include e.g. summarising papers but also day-by-day reports in newspapers, local radio and television or permanent presentations screens at public locations etc. To improve the understanding and concern for the air quality the development of simple methods for school children, citizen groups etc. to study and monitor the situation shall be encouraged (Nilsson, 1995). However, it should be taken into account that for benzene, unlike other atmospheric pollutants, no short-term limit values are recommended. Thus information to the public will need to be related to long term limit values. Clear information about the significance of short-term concentrations (for instance daily or hourly means) in relation to longer term exposure should be given to the public. It is suggested that benzene is an important topic to present to public in coincidence or before the Directive implementation. Review articles on popular press and TV spots are the ideal tools to inform the public about the risks related to benzene exposure and for suggesting practical behaviour in urban areas (Public environmental hygiene). A great deal of information should be addressed to benzene health effects in order to show the relative risks at different concentration levels. It is also suggested to elaborate sufficiently accurate emission inventories at national, regional and local levels in order to clear the sources of benzene and how these sources are supposed to be controlled in the development of directive.

3.8 - Upper and lower assessment thresholds

109

Art. 6 of the Framework directive sets out the requirements for measurements and other assessment techniques depending on the pollution burden as follows: Zone and/or pollution level
• above the limit value • above x % of limit value (upper assessment threshold – UAT) • agglomerations in the case of pollutants with alert thresholds(not relevant for benzene)

Assessment technique

Measurement mandatory but other techniques may also be used Combination of modelling and lower intensity of measurement Sole use of modelling or objective estimation techniques possible

Below x % but above y % of the limit (lower assessment threshold – LAT) Below lower assessment threshold

The interannual variability Sv of the annual averages and/or percentiles was used as a basis by the previous working groups for SO2, NO2, Lead and Particulate to arrive at numbers for the UAT and LAT. The rationale behind this approach is that assessment regimes with higher levels of uncertainties should be allowed only for zones where the limit values are definitely met, irrespective of variations of the pollution levels from year to year. An analysis of the time series (annual averages from 1989 to 1995) from 7 urban background stations and 2 traffic exposed stations in Germany was performed and the interannual variations, corrected for the long term trends and the additional variations introduced by grab sampling (15 % in these cases), were calculated. The interannual variability Sv (over the time period considered (7 years) was 20 % or less (for details refer to Table 3-VI). If a 95 % confidence interval for the definition of the UAT is chosen (that is twice the standard deviation), a value of 60 % for the UAT related to the limit value can be inferred. A similar approach using 3 times the standard deviation yields 40 % for the LAT. Consequently, the proposed values for the upper and lower assessment thresholds are as follows:

110

Assessment threshold

% of limit value

Comment

Upper Assessment Threshold Lower Assessment Threshold

60 % 40 %

2 x standard deviation 3 x standard deviation

Table 3-VI: Analysis of benzene trends at selected sites in Germany and calculation of the interannual variability Sy (in µg/m³)

Year Essen 1989 1990 1991 1992 1993 1994 1995 Sy 3,02 3,93 5,14 3,71 3,25 3,81 2,20 1 2 3 4 5 6 7 0,92

Stations (urban background) Bonn 4,25 3,25 4,10 2,56 2,26 1,97 2,85 1 2 3 4 5 6 7 0,63 9,74 5,60 4,74 3,61 3,75 2,38 3,83

Bottrop 1 2 3 4 5 6 7 1,53

Year 1989 1990 1991 1992 1993 1994 1995 Sy 3,62 3,32 3,90 2,54 2,74 2,00 3,68

Stations (urban background) Datteln Düsseldorf 1 2 3 4 5 6 7 0,70 4,64 3,70 5,78 3,74 3,09 4,21 3,81 1 2 3 4 5 6 7 0,88

111

Year

Stations (urban background) Schwerte 5,74 2,83 6,66 3,18 2,94 2,51 3,60 1 2 3 4 5 6 7 1,51

traffic exposed VDDF 1 22,75 21,34 17,67 15,49 16,30 11,10 2 3 4 5 6 7 1,34 VESN 19,73 12,66 11,84 11,19 8,85 9,99 7,15 1 2 3 4 5 6 7 2,08

urban background Wesseling 8,35 9,88 6,87 4,93 5,06 4,15 4,28 1 2 3 4 5 6 7 1,12

1989 1990 1991 1992 1993 1994 1995 Sy

The relative standard deviation S % was calculated from the absolute variance Sy by relating Sy to the average of the annual means of each station. As the benzene trends were measured by grab sampling, S % was corrected for the additional uncertainty Sg introduced by grab sampling (15 %) as follows:
2 2 Sv (%) = S % − S g

112

Benzene Calibration Method

Method
Static volumetric method

Description

Reference

Advantage

Disadvantage

A known liquid ISO 6144, volume of benzene to VDI 3490 a known volume of air.

Primary calibration method. Control of the Cost effective method. purity of Also suited for other benzene. pollutants. Precise and accurate (uncertainty < %). Primary calibration method. Control of the Diffusion rate determined by purity of weighing. benzene. Continuous production of calibration gas. Cost effective method. Precise and accurate.

Permeation method Benzene permeation ISO 6349 through a membrane into a flow of carrier gas at a constant temperature. The loss of benzene is determined by weighing. Dynamic Introduction of a ISO 6145 volumetric method given rate of a gas into a constant flow rate of a complementary gas. The gas mixture is in general a high concentration gas mixture.

Gas cylinders are commercially available. Easy to handle. Certified standards are available from NIST (USA).

Unknown accuracy. No primary standard. The certified standard is expensive.

113

Benzene Sampling Methods

Method

Description

Reference Advantage

1

Laminar flow manifold Flow 1501 min-1, tube diameter 15 cm EPA Inlet material: glass, stainless steel, teflon Turbulent flow manifold Modular sugar cane design. Inlet material: glass, stainless steel, Teflon

Isokinetic sampling, sample unaffected

2

Low cost, modular construction Low cost, efficient sampling without losses

3

Direct connection of analyser inlet to station sampling head

114

Review of existing measurement methods

Method

Average time Detection limit

Reference

Advantage

Disadvantage

Pumped sampling 0.5-24 hour on tubes with thermal desorption-GC

0.14 0.4

prEN 1076, VDI 3482-4, ISO/TC 146/Sc 6 N 19-20

- Low cost - Many sampling sites with one GC analytical instrument - Work with toxic species is minimised - Calibration with spiked tubes - Low cost - Many sampling sites with one analytical instrument - Analysis is made on a cheap GC - No electrical power is needed for sampling - Same comments as in No. 2 - Toxic solvents are needed

Pumped sampling 3-12 hour on tubes with solvent 24 hour extraction-GC

0.14 0.4

prEN 1076, VDI 3482-4, ISO/TC 146/Sc 6 N 19-20

Diffusive sampling 1-4 weeks with solvent extraction-GC Diffusive sampling with thermal desorptionGC Canisters Thermal Desorption 1-4 weeks

0.5

CEN/TC 264/WG 11/N 15-17

- Long sampling time needed - Same comments as in No. 2 - Differences to active sampling of about 20% - Long sampling time needed - Differences to active sampling of about 20% may occur - Only grab samples (with short average times) can be taken (discontinuous measurements)

0.5

CEN/TC 264/WG 11/N 15-17

- No electrical power is needed - Same comments as in No. 1

seconds to minutes

0.3

VDI 3482-2,3

- No electrical power is needed - Same comments as in No. 1 and 2 depending on the analysis technique - High time resolution - Real time data - Analysis of BTX simultaneously

BTX monitors

1/2 to 1 hour

0.16

DIN 33961-1,2

- High cost - Need of a monitoring hut with air condition - Need of carrier gas supply and burner gas (for FID)

115

On line GC instruments

1/2 to 1 hour

0.1

ISO

- High time resolution - Real time data - Analysis of C2-C12 simultaneously

- High cost - Need of a monitoring hut with air condition - Need of carrier gas supply - Instruments from some companies need - High cost - Measurements over a long path - Zero point offset problems -Not sufficient documentation available - Difficulties with measurements at low visibility - Problems with calibration

DOAS

30 seconds

1.5

Validation report under preparation by UBA, Germany

- High time resolution - Real time data - Measurements over a long path - No chemical interference - Multicomponent measurements

116

3.9 Summary
Monitoring methods are generally based on gas chromatography, with either active (pumped) or passive (diffusive) sampling and thermal or solvent desorption prior to analysis. Automated instruments with active (pumped) sampling and thermal desorption are available, which allow for quasi continuous monitoring. Advantages and disadvantages as well as typical uncertainties of the different methods are given. Pumped sampling on tubes followed by GC analysis including calibration is recommended as basis for a reference method. A detailed reference method will be worked out by CEN. Any network design for compliance checking must take the requirements of the Framework directive as starting points. In addition, it has to be borne in mind that for benzene long term exposure with possible adverse effects on human health in areas representative for the majority of the population as well as at hot spots has to be assessed. Consequently, monitoring should cover the following microenvironments: urban background, road side, residential quarters near industrial sites and rural background (for reference data only). The following minimum numbers of monitoring stations N per zone or agglomeration are proposed: Site urban background Ni (without full assessment) 4 √ ¬I I = number of inhabitants in million 1 per zone; 1 additional station per 1 million inhabitants ≥ 1 (tailor made) 1 per 50.000 km² 1 N0 (with full assessment)

road side

1 + √ ¬P P = Population of Member State (million) 1 1

industrial rural background

A full assessment is only possible by combining modelling with monitoring. This holds especially true for road side locations. For this purpose, a variety of air pollution models for various spatial and temporal scales exist. An overview of existing local scale models and their characteristics is presented, but no general recommendations are given by the working group.

117

The considerations for microscale siting are very similar to those laid down in the first daughter directive. Principles of Quality assurance are described, and the working group recommends the following acceptable overall accuracies (trueness and precision) for assessments: mandatory 20 %, indicative 30 %, modelling 50 %, objective estimation 100 % of limit values. The working group feels that random and continuous measurements can be regarded as equivalent, provided that the sample size of the random measurements is big enough to yield expanded uncertainties ≤ 10 %, and that the sampling dates be evenly spread during the measurement period. Necessary sample sizes can be calculated by resampling of stochastic samples from continuous time series. Examples for urban background, road side and industrial locations are presented. Upper (UAT) and lower assessment thresholds (LAT) were derived from the interannual variability of annual averages from 9 stations. The working group recommends 60 % as UAT and 40 % as LAT. Finally, the working group stresses the importance of informing the public on long term benzene levels and explaining the significance of short term concentrations in relation to longer term exposure.

118

3.10 References
Allegrini I., Febo A., Giliberti C., Perrino C.: Intercomparison of DOAS and conventional analysers in the measurement of atmospheric pollutants in an urban background monitoring site of Rome. European Symposium on Environmental Sensing III Fairground, Munich FR Germany; 1997. Allegrini I., Febo A., Giliberti C., Giusto M.: Measurement of photochemical pollution in the urban environment of Rome by means of a proper combination of DOAS and atmospheric stability monitor. Asger B. Hansen, Palmgren, F.: VOC Air Pollutants in Copenhagen, The Science of the Total Environment, 189-190 (1996) 451-457. Barrefors, G.: Monitoring of Benzene, Toluene and P-xylene in Urban Air with differential optical Absorption Spectroscopy Technique, The Science of the Total Environment, 189190 (1996) 287-292. Bartley D.L., Doemeny L.J., Taylor D.J.: Diffusive monitoring fluttuating concentrations. Am. Ind. Hyg. Assoc. J. 1983; 44: 241-247. Bartley D.L., Woebkenberg M.L, Posner J.C.: Performance of the thick-sorbent diffusive samplers. Ann. occup. Hyg. 1988; 32: 333-343. Berckowicz R., Hertel O., Larsen S.E., Sørensen N.N., Nielsen M.: Modelling traffic pollution in streets. Ministry of Environment and Energy - National Environmental Research Institute. January 1997, p52. ISBN 87-7772-307-4. Berkovicz R., Palgrem F., Hertel O., Vignati E.: Using measurements of air pollution in streets for the evaluation of urban air quality - meterological analysis and model calculations. The Science of Total Environment 1996; 189/190: 259-265 Bertoni G., Bruner F., Liberti A., Perrino C.: Some Critical Parameters in Collection, Recovery and Gas Chomatographic Analysis of Organic Pollutants in Ambient Air Using Light Adsorbents. J. Chromatogr. 1981; 203: 263-270.

119

Bertoni G., Canepari S., Rotatori M, Fratarcangeli R., Liberti A.,: Laboratory Validation and Field Application of Thermally desorbable tube-type Passive Samplers. L.J. Brasses and W.C. Mulder (Eds), Man and his Ecosystem, Proceedings of the 8th World Clean Air Congress !989 - The Hague - The Nederlands 1989; 5: 207-212 Bertoni G., Canepari S., Rotatori M, Fratarcangeli R., Liberti A.,: Evaluation tests and applications of a double-layer tube-type passive sampler. J. Chromatogr. 1990; 522: 285-294. Bertoni G., Perrino C., Fratarcangeli R., Liberti A.,: Limites dans l’adsorbtion del polluants gazeux sul les echatillonneurs passifs qui utilisent supports a faible surface. Vie Congres Mondial pour la Qualitè de l’ Air - Paris 1983 Bertoni G., Perrino C., Fratarcangeli R., Liberti A.,: Critical parameters for the adsorption of gaseus pollutants on passive samplers made of low specific area adsorbents. Anal. Lett., 1985; 18 (A4): 429-438 Bertoni G., Perrino C., Liberti A.,: A Graphitized Carbon Black Diffusive Sampler for the Monitoring of Organic Vapours in the Environment. Anal. Lett., 1982; 15 (A12): 1039-1050. Blakley K.C.: Atmospheric monitoring of aromatic hydrocarbons in the community around BP Chemicals at Baglan Bay from Jannuary 1991 to December 1991. British Petroleum Branch Report No. 124 380 1992. Boeft J., Thijsse Th.R.: Model voor het berekenen van de jaargemiddelde benzeenconcentratieebijdragen van parkeergarages en parkeerterreinen. CAR Parking (prototype) - TNO-MEP report R 95/222, Delft, The Netherlands (1996) [in Dutch]. Brocco, D. et al., Valutazione dei Composti Organici Volatili (VOC) nel Centro Storico di Roma, Acqua-Aria, 5, (1993). Brown R.H.: The use of diffusive samplers to assess local air pollution problems. Presented at workshop on Air Pollution Problems in Rapidly Developing Countries, Cheng De, China 1994. Brown R.H.: The use of diffusive samplers for monitoring of ambient air. Pure and Appl. Chem. 1993; 65, No. 8: 1859-1874

120

Brown R.H., Charlton J., Saunders K.J.: The development of an improved diffusive sampler. Am. Ind. Hyg. Assoc. J. 1981; 42: 865-869. Busch K. A.: SCP Statistical Protocol in Documentation of the NIOSH Validation Tests, Taylor D.G., Kuppel R.E., Bryant J.M. eds., Pub. No.77-185, DHEW (NIOSH). Cassinelli M.E., Hull R.D., Crable J.V., Teass A.W.: Protocol for the evaluation of passive monitors, in Diffusive Sampling. Royal Society of Chemistry - London - England 1987; 190202. Clarke, A. G., Ko, Y.-H., The Relative Significance of Vehicular Emissions and Other Emissions of Volatile Organic Compounds in the Urban Area of Leeds, UK, The Science of the Total Environment, 189-190 (1996) 401-407. Choo-Yin C., Layton Matthews G.: Diffusive Sampling - An Alternative Approach to Workplace Air Monitoring. Development of a diffusive sampler for monitoring ambient levels of organic vapours particularly benzene. CEC Publ. No 10555EN, ISBN 0-85186-343-4 1987. Coseman G., Kretzschmar J., Maes G.: The Belgian Immission Frequency Distribution Model IFDM. Proc. of the DCAR Workshop on objectives for next generation of practical short-range atmospheric dispersion models Ed. by H. Olesen, T. Mikkelsen, Riso, Denmark 1992; 149-150. De Saeger E., Gerboles M., Perez Ballesta P., Amantini L., Payrissat M.: Air Quality measurement in Bruxels (1993-94) NO2 and BTX monitoring campaings by diffusive samplers. EUR 16310 1995. Donati, J., Teneurs en Hydrocarbures Aromatiques Monocycliques Dans l'Air Ambiant à Paris, Pollution Atmospherique, Janvier-Mars 1995. EEA: Guidance Report on Supplementary Assessment under EC Air Quality Directives. European Commission: Database, monitoring and modelling of urban air pollution Inventory of models and data sets, 1997. Gelencser A., Kiss Gy, Hlavay J., Hafkenscheid Th. L, Peters RJB, de Leer EWB: The evaluation of a tenax GR diffusive sampler for the determination of benzene and other volatile aromatics in outdoor air. Talanta 1994; 41, No. 7: 1095-1100

121

Gunnar, Barrefors: Monitoring of benzene, toluene and p-xylene in urban air with differential optical absorption spectroscopy technique. The Science of the Total Environment, 1996; 189/190: 287-292 Kennedy E:R. Fischbach T.J., Song R., Eller P.M., Shulman S.A.: Guidelines for air sampling and analitycal method development and evaluation. DHHS (NOSH) Publication No. 95-117 Cincinnati 1995. Hampton, C. V. et al., Hydrocarbon Gases Emitted from Vehicles on the Road. 1. A Qualitative Gas Chromatography/Mass Spectrometry Survey, Environ. Sci. Technol. (1982), 16, 287-298. Hastenteufel S., Betz W.R.: Comparison of multi-bed adsorbent tubes to trap toxic organic compounds. Clean Air at work. Ed. Brown R.H., Curtis M., Saunders K.J., Van den Driessche 1992: 190-194. HSE/CAR Working Group 5: The diffusive monitor. Issue 6 1993. Maltoni C., Conti B., Cotti G., Belpoggi F.: Experimental studies on benzene carcinogenity at the Bologna institute of Oncology: current results and ongoing research. Am. Ind. Med. 1985; 7: 415-446. MHDS 66.: Laboratory method using porous polymer diffusion samplers, thermal desorption and gas cromatography. Mixed hydrocarbons (C5 to C10) in air 1989. Namiesnik J.: Preconcentration of gasseous organic pollutants in the atmosphere. Talanta 1988; 7: 567-587. NIOSH: Manual of Analytical Methods, 4th Ed., Eller, P.M., U.S. DHHS Publication No. 95117, Cincinnati 1995. NIOSH: Manual of Analytical Methods, 2nd Ed., vol 1 Washington D.C. 1977: 127-1, 127-7 OSHA: Development of a protocol for laboratory testing of diffusive samplers. OSHA Salt Lake Technical Center Report 1996.

122

Perez Ballesta P., Gonzalez Ferradas E., Miñana Aznar A.: Muestro de contaminantes en aire por captacion pasiva. Servicio de publicaciones de la Universitad de Murcia 1993. Mucci N. e Camoni I.: Raccolta dei pareri espressi dalla CCTN Committee nel 1994The National Advisory Toxicology Committee, (a cura di)Roma 1995. Taylor J.K.: Quality Assurance of Chemical Measurements. Lewis Publishers, 1987; 251Appendix A. Underhill D.W., Feigley C.F.: Boundary layer effect in diffusive monitoring. Annal. Chem. 1991; 63: 1011-1013. Zannetti P.: Numerical simulation modeling of air pollution: an overview, in Air Pollution (P. Zannetti et al., eds), Computational Mechanics Publications, Southampton 1993; 3-14.

123

Chapter 4

Benzene: Economic Analysis

124

4.1 Introduction
All Working Groups preparing technical position papers on ambient air quality are asked by the Commission to develop their recommendations for air quality objectives on the basis of their assessment of the risks of the pollutant with which they are dealing. Legislation setting ambient air quality limit values does not however merely list objectives. It sets binding targets and binding dates by which they must be met. Working Groups are therefore asked also to take into account practical considerations in framing their final detailed recommendations. In particular they are asked to identify cost-effective means by which recommended objectives can be met, to recommend timetables and to identify the benefits of meeting the targets and the consequences of not doing so. If it is not possible to identify reasonable means of reaching a desired target within a reasonable timescale, then economic analysis can play an extra role in helping identify interim targets. Economic analysis is a specialist task. DGXI therefore engaged a team of consultants, led by AEA Technology, to help the Working Group with this part of their assignment. They were asked: to assess likely concentrations of benzene across the Union in the year 2010, taking into account the effects of existing and planned legislation; to determine whether further action would be needed to reach targets of 10 µg/m3, 5 µg/m3 and 2 µg/m3 by that date; and if further action would be needed, to identify the least cost means; to assess the benefits of meeting these targets.

4.2 Results of economic analysis

125

A full report of the study is available in a separate document2. A number of difficulties were encountered during the course of the work, which the consultants considered were priorities for further data collection and research activities. Firstly, they found that, as found also in Chapter 1 of this document, emission inventories between different countries are inconsistent. Secondly, the cost-benefit assessment was made extremely difficult by the lack of good exposure-response data for benzene. A third difficulty concerned the overall framework for the analysis. The consultants suggest that it would be preferable in future analyses to take a broader approach, considering a larger number of pollutants (e.g. SO2, NO2, benzene, and ozone) simultaneously. This would allow better integration of the secondary effects of abatement measures for individual pollutants, for example through reductions in emissions of other pollutants under consideration. Overall this should lead to a more optimal selection of abatement technologies. Analysis concentrated on three were extrapolated to the rest of the European Union. The Steering Group for this work requested that the following values be investigated as possible limits in this study: • 2, 5 and 10 ug/m3 as annual average to be attained by 2010.

These limits were to be investigated as they applied in both urban background and hot-spot locations. The sole benefit considered is reduction in the risk of contracting leukaemia. There are suggestions that benzene could have other effects, such as lung cancer. However, as noted by the Ad Hoc Group of Experts on the risks of benzene there are only isolated studies showing such effects and there is no agreement amongst experts as to potential mechanisms nor potential unit risks. Benzene has no recorded direct effects on the environment. It is an ozone precursor, though its ozone creation potential is relatively low. The methodology for this study follows to a large extent extrapolation of the results of the Auto-Oil programme. Auto-Oil provided detailed modelled assessments of urban background air quality across 7 cities, these cities being broadly representative with respect to air quality of all cities in the European Union. Auto-Oil also provides a set of data and

2

AEA Technology: Economic Evaluation of Air Quality Targets for CO and Benzene – available from DGXI

126

assumptions that have been widely reviewed, discussed and agreed by European decision makers and other interested parties already. The baseline scenario used here incorporated the Auto-Oil Directives on fuel quality (including the reduction in maximum benzene content of petrol to 1%) and vehicle emissions and the first daughter Directive on SO2, PM, NO2 and lead, using results from an earlier study conducted for the Commission. There is emerging evidence that the measures considered in that earlier study would not be sufficient to meet the limits contained in the first daughter Directive. Given the commonality of sources for the pollutants considered, and the particular effect of transport in areas where exceedences are most likely it is possible that this will mean that the present analysis overestimates future benzene levels and therefore costs. Emissions of benzene are expected to fall greatly by 2010 as a result of the legislation included in the baseline scenario. The study predicts that background concentrations will fall to some 3.2 µg/m3 even in Athens, the most polluted of the cities studied in Auto-Oil. Benzene in the future will therefore be a problem only in hotspots – primarily heavily trafficked urban streets. Aggregated results for urban background and hot-spot locations are shown in Table 4.1. A number of uncertainties have been explored in these tables: • Emission scenarios. Three scenarios were examined, requiring high (H), medium (M) or low (L) further reductions to meet targets. • Relationship between urban background and hot-spot concentrations • The risk factor determining the number of cancers likely to be linked to a given level of benzene • Whether measures are strongly targeted (optimised abatement case - opt) or weakly targeted (generalised abatement case - gen) at the locations where exceedences occur.

Table 4.1 . Aggregated results of the analysis for benzene, estimating costs and benefits throughout the EU. Cells with ‘0’ entered represent cases where no exceedence is predicted.
Limit µg/m3 2 Impacts (cases) Benefits (kEuro) Costs (kEuro) Location Emissions Low High Low High urban back L - opt 0.0003 0.034 0.67 119 890 M - opt 0.0059 0.59 12 2,107 71,800 H - opt 0.010 1.0 20 3,583 207,850 hot spot L - gen 0.081 8.1 162 14,344 9,774,150 M - gen 0.13 13 265 23,151 23,059,710 H - gen 0.16 16 330 28,691 34,491,730

2

127

2

5

5

10

10

L - opt M - opt H - opt urban back L - opt M - opt H - opt hot spot L - opt M - opt H - opt urban back L - opt M - opt H - opt hot spot L - opt M - opt H - opt

hot spot

0.037 0.077 0.13 0 0 0 0.0030 0.010 0.019 0 0 0 0 0 0.0040

3.7 7.7 13 0 0 0 0.30 1.0 1.9 0 0 0 0 0 0.40

74 153 263 0 0 0 5.9 20 37 0 0 0 0 0 7.9

6,879 12,463 20,713 0 0 0 1,058 3,511 5,619 0 0 0 0 0 1,404

976,220 3,539,690 8,389,050 0 0 0 24,330 226,360 465,280 0 0 0 0 0 45,590

There is a tendency for costs to significantly exceed benefits where exceedences are likely to occur, except in the case of a target of 10 µg/m3 with abatement closely targeted on areas of exceedance. There are several points that should be noted here however: 1. In many cases it is likely that concentrations estimated here are overestimated. Further abatement to that considered here in the baseline scenario may well follow without further action to control benzene. This may arise through measures needed to control other air pollutants, congestion, and emissions of greenhouse gases following the agreement of the Kyoto protocol and the need to reduce other impacts of transport. A further factor is the potential introduction of new technologies, which could make an impact on benzene levels by 2010. 2. The consultants did not consider direct secondary effects of abating benzene, for example on ozone levels, or indirect secondary effects via reductions in concentrations of other VOCs, NO2 and particulate matter. They considered the logic for including such effects (see table) within the context of the Framework Directive unclear. In this case it is likely that the so-called secondary effects would be much larger than the primary effects. If this were so it would probably be more efficient to develop abatement strategies primarily to address these larger benefits and to ensure that the potential for reducing benzene concentrations was considered simultaneously.
Abatement measure Traffic calming, public transport subsidies, etc. Burden affected Emission of all transport pollutants (SO2, NOx, PM10, VOCs, CO2 etc.. generation of related secondary pollutants such as ozone) Risk of accidents Congestion Impacts affected Effects on health, materials, ecology Death and injury, material damage Travel time

128

Emission constraints

Noise Emission of all transport pollutants (SO2, NOx, PM10, VOCs, CO2 etc.)

Amenity Effects on health, materials, ecology

3. In many cases the extent of exceedence was tightly constrained, geographically and with respect to the amount of exceedence. This makes it more likely that the measures taken for reasons raised at [1] could lead to certain limits being met without controls being directed specifically at benzene. The clear exception to this was the limit of 2 µg/m3 applied in hot-spots.
4. There are significant problems in the quantification of benzene risks. The unit risks

found in the scientific literature vary a great deal. Following the advice of the Ad Hoc Group of Experts on health risks and the recommendation of the Working Group a range or unit risks differing by a factor of 100 was explored in this study. Further research on the mechanism of benzene effects may reduce this uncertainty in the near future. As already explained above there are also at present few data on personal exposure to benzene. This too seems likely to improve in future years.

4.3 Discussion
Cost benefit analysis is a relatively new tool in analysis of carcinogenic compounds. There are few published studies in the literature. Genotoxic carcinogens are also a very sensitive political and public issue. It is important therefore to be clear about the role of such analysis in helping legislators to arrive at decisions. Firstly, it should be understood that the above results are not an estimate of the costs and benefits of reducing benzene concentrations from today’s levels to the target concentrations included in the analysis. Concentrations will fall by 2010 as a result of legislative decisions already taken – most notably the Auto-Oil Directives, the IPPC Directive, the proposed Directive setting limit values for sulphur dioxide, nitrogen dioxide, particulate matter and lead and the proposed Solvents Directive. The figures above include neither the costs nor the benefits relating to these results. They are the estimated costs and benefits for further action beyond these existing plans. Secondly it is extremely important also to understand the uncertainties identified by the consultants and the extent to which they were able to deal with these in their analysis. The most important of these uncertainties were the quantification of both current and future emissions, model error, the relationship between background concentrations (for which

129

most data are available and which can be modelled most accurately) and hotspot concentrations, the size of the risk associated with benzene, the extent to which the population is actually exposed to different concentrations in different types of location, the valuation of benefits, the integration of secondary costs and benefits of abatement options, and the extent to which results for the three cities studied can be extrapolated to the EU as a whole. As far as possible, the consultants integrated these factors into their study during the main part of their work or during sensitivity analysis. This is reflected in the ranges of their estimates. The largest uncertainty factor incorporated into the analysis is the risk associated with benzene. With a factor of 100 between the highest and lowest unit risk estimates this dominates the benefits results. Some uncertainties could not be incorporated. Among these the two most important are is the extent of secondary benefits due to the implementation of measures that would reduce benzene concentrations and the extent to which measures which would reduce concentrations of benzene will be implemented anyway for other reasons. The consultants estimated that the potential reduction in the number of cancers per year (beyond the reduction due to existing plans) would be relatively low for all limit values tested (see table 4.1 above). Estimated costs exceed estimated monetary benefits for both the 2 and 5 µg/m3 targets at the kerbside. For the 10 µg/m3 target, benefits were found to be larger than costs, but only when abatement measures are closely targeted to the areas where exceedance occurs. This picture would be likely to change if either secondary benefits or measures to meet other targets were included. Potential secondary benefits include abatement of other pollutants and in the case of non-technical measures, alleviation of congestion and noise. These benefits could be substantial. However, other important pollutants are being dealt with by separate legislation. For some, including particulate matter, Council has just adopted a Directive setting limit values. As far as possible the effects of this are already built into this analysis. The remaining secondary benefits would be those of going further than the Common Position. There are not sufficient data available to take into account reducing congestion or noise. In any case, an economic analysis of reducing benzene concentrations which showed that the great majority of the benefits would be obtained for example, from reductions in other pollutants which are regulated separately, would seem to be of limited use in analysing what action should be taken on benzene itself.

130

On the other hand, it can be argued that some of the measures which would reduce concentrations of benzene beyond current plans will in fact be taken in coming years for other reasons. If so, the costs should not be assigned wholly to benzene. The proposals recently adopted by the Commission on ambient air quality standards for ozone and national emission ceilings will, for example, require reductions in emissions of VOCs generally. Measures taken to reduce traffic congestion would also reduce benzene concentrations. Implementation of the Kyoto protocol will affect global energy consumption and hence further decrease benzene emissions. It is not possible to make any precise estimate of the extent to which such actions would reduce the costs of achieving benzene targets on the basis of presently available data. But in the view of the Working Group it is likely that estimated costs would begin to exceed estimated benefits at a concentration below 10 µg/m3.

131

Chapter 5

Benzene: Discussion and final recommendations

132

5.1 Summary of Chapters 1-4
5.1.1 Chapter 1
Chapter 1 of this document discusses available data on present day benzene emissions and concentrations in the European Union and expected trends in coming years. The major source of benzene across the EU is combustion, of which traffic is the largest component. Emissions are expected to fall substantially in the period to 2010 compared with 1990 levels. There are no systematic data on present day concentrations, but cities studied in the Auto-Oil programme showed concentrations in the urban background of 6 – 12 µg/m3 for 1990. Measurements show that concentrations have already decreased in some cities. It is expected that these concentrations will be reduced to below 4 µg/m3 in all cities by 2010, primarily as a result of new emission limit values for vehicles and a reduction in the maximum benzene content of petrol to 1% from 2000. Other data shown in Chapter 1 show that concentrations can vary a great deal across the Union, particularly when hotspot situations are taken into account. There are a number of potential reasons for this. Some, such as the age of industrial plant or of the vehicle fleet, can in principle be managed. Some, such as climate or local topography cannot.

5.1.2 Chapter 2
Chapter 2 deals with the risks associated with benzene. Most authorities agree that benzene should be classified as a known human genotoxic carcinogen. There are open questions about its mechanism of action, particularly at low doses, but at present no threshold for effects has been identified. It is impossible to estimate precisely the risks associated with exposure to environmental benzene. The largest body of evidence is concerned with exposure of industrial workers to relatively high concentrations for relatively short periods of time. There are many uncertainties when extrapolating from these data to exposure of the whole population to lower concentrations over a lifetime. WHO in developing its 1996 guidelines used a linear model to extrapolate. They noted some uncertainties but did not feel it possible to recommend any other way of modelling the data. An ad hoc group of experts which reviewed evidence published since WHO carried out their work was of the view that this evidence does not allow the uncertainties identified by WHO to be removed. That is, it is still not possible to say what is the best model for extrapolating from worker exposure to environmental exposure of the whole population.

133

Nor, despite evidence that benzene is an unusual carcinogen, do animal and other data allow a threshold to be identified below which effects are not expected. The ad hoc group concluded that though it was not possible on present evidence to give a precise estimate of the risk associated with benzene it was possible to define a range within which that risk was likely to lie. The procedure followed by the WHO working group was considered to result in the highest plausible estimate of risk – an excess lifetime risk of leukaemia at an air concentration of 1 µg/m3 of 6 x 10-6. Unit risks in the order of 10-10 and 10-11 derived by some authors, were felt to be too far outside the general range of estimates to be relied upon with good supporting evidence for the models used. The lowest unit risk which the group felt was likely to be plausible was in the order of 5 x 10-8. They did not feel that present evidence supported the setting of a short-term limit value. Consideration could be given to this in future of more evidence accumulated. In the meantime however, reducing annual average concentrations should also reduce shorter-term peak concentrations. Working Groups on individual pollutants are asked by the Commission to base their initial recommendations for goals for air quality on their assessment of risk. This is particularly difficult in the case of carcinogens. Converting unit risks to goals means deciding on a level of risk which will be tolerated. The European Parliament and the European Council, has agreed a proposal for a Directive on Drinking Water, in which an excess lifetime risk of 1 in a million is taken as the starting point for developing limit values. Taking this as a precedent, the range of unit risks given above (6 x 10-6 to 5 x 10-8) has been converted into annual average concentrations which are estimated, over a lifetime, to equate to an excess risk of contracting leukaemia of 1 in a million. The resultant range of concentrations is 0.2 to 20 µg/m3. The Working Group notes that there are epidemiological studies suggesting risks of other effects of benzene, from both long-term and short-term exposure. The number of such studies dealing with any particular effect is small and they are not at present supported by work suggesting plausible mechanisms. There is no consensus on the extent of any such risks. There is much research now underway on the mechanism by which benzene causes leukaemia and on whether benzene has other effects on human health. The Working Group recommends that the accumulating scientific evidence on benzene should be reviewed again in approximately 5 years’ time.

134

5.1.3 Chapter 3
Chapter 3 deals with assessment of benzene concentrations and recommends methods of measuring it. It is important to be aware when setting standards of the accuracy with which concentrations can be measured or otherwise assessed. As Chapter 3 shows, there are uncertainties associated with all methods of assessment, including measurement, and these tend to increase as the concentrations to be assessed decrease.

5.1.4 Chapter 4
Chapter 4 summarises the results of the consultancy study on the economic evaluation of air quality limit values for benzene. The consultants point out a number of important uncertainties in the study. Benefits estimates are highly uncertain owing to lack of agreement on risks and lack of data on human exposure and costs are probably overestimated. Caution is therefore needed in interpreting the results. The consultants found that the estimated potential reduction in the number of cancers per year beyond the substantial reduction which will occur due to existing legislation would be relatively low for all limit values tested. This is the case even if the most conservative attitude is adopted to risk and exposure. Estimated costs exceed estimated monetary benefits for both the 2 and 5 µg/m3 targets at the kerbside. For the 10 µg/m3 target benefits were found to be larger that costs, but only when the most conservative attitude is taken to risks and when abatement measures are closely targeted to the areas where exceedance occurs. It should be noted that costs do not fall evenly across Member States. The highest costs would be borne by Mediterranean cities such as Athens. Secondary benefits of measures to reduce concentrations of benzene, such as reductions in concentrations of other pollutants, or in emissions of greenhouse gases were not included. They are potentially much larger than the benefits associated with further reductions in benzene concentrations, but tend to be the subject of separate legislative action.. On the other hand if further action were taken with one or more of these benefits as a primary goal, then this would reduce the cost of achieving limit values for benzene. Such further action will be needed in order for example to implement the Kyoto protocol, and to reduce acid rain and ozone formation. Meeting limit values should therefore be easier to attain than the results suggest for all Member States. 5.2 Considerations to be taken into account in setting limit values for benzene

135

Benzene is the first carcinogen to be considered under the new framework for ambient air quality limit values set up by the Directive on Ambient Air Quality Assessment and Management. There are no existing ambient air quality limit values for carcinogens, although such limit values have been developed for drinking water. There are a number of factors to be taken into account in framing legislation on benzene. These include: • • • • • • the risks to human health; the obligation under the Treaty to implement the precautionary principle; the extent to which risk is acceptable; the measures that can be taken to reduce benzene concentrations; the speed with which they can be implemented across the whole of the EU; the extent to which society is prepared to bear the costs.

The primary consideration is the risk to human health. The Working Group has not found it possible to estimate this risk precisely and has given a range within which the unit risk of contracting leukaemia from exposure to benzene (the additional risk of contracting leukaemia if exposed to 1 µg/m3 for a lifetime) is most likely to lie. This range is 6 x 10-6 to 5 x 10-8. Even if the risk could be quantified an extra step is however required in order to determine the basis for a limit value. Since on current evidence no threshold can be identified setting a target means deciding what level of risk is tolerable. This is not a decision that can be taken on technical grounds alone. The Working Group notes the decision by the European Council and the European Parliament in the context of drinking water that an additional lifetime risk of 1 in a million should be taken as the starting point for determining limit values. Translating the range given above into an annual average concentration which equates to an additional lifetime risk of contracting leukaemia of 1 in a million gives a range of concentrations of 0.2 to 20 µg/m3. This wide range cannot be narrowed on scientific grounds without further evidence about the mechanism by which benzene causes leukaemia. The Working Group considers it reasonable to expect that further evidence will become available over the next five years or so which will throw more light on this question. It notes that further research is also underway on other possible effects of benzene. It considers that the results should be reviewed in approximately 5 years time. It the meantime it recommends that exposure to benzene in ambient air should be as low as reasonably achievable.

136

This leads inevitably to the question of what is reasonably achievable. The analysis of the costs and benefits of meeting limit values for benzene described in Chapter 4 is a useful input, but the answer depends also on the attitude adopted to risk in the light of uncertainty, and on willingness to undertake further actions to reduce risk given that uncertainty. Clearly the decisions to be made are socially and politically important and the Working Group considers them beyond its remit. However, in order further to assist both the Commission in developing its proposal for a Directive and those involved in the adoption of legislation the Working Group has decided to put forward three illustrative options, developed on the basis of its technical analysis.

5.3 Options for consideration
The Working Group has adopted the following principles in developing the options below: • concentrations of benzene in the air shall not increase; • exposure to benzene in the ambient air should be as low as possible; an annual average limit value should be set; the limit value should apply everywhere including hot spots; • there is not sufficient evidence for setting a separate short-term limit value. An annual average limit value applying in hotspots will however reduce peak exposure.

• •

5.3.1 Option 1
Decision makers may consider that a conservative attitude should be adopted towards the risk associated with benzene and the precautionary principle implemented to the maximum possible extent. A concentration of 0.2 µg/m3, which equated to a additional risk of 1 in a million on the most conservative view of risk (see 5.1.2 above) is close to that found in remote areas such as Antarctica and is not achievable in areas of economic activity. Given however that people do not spend entire lifetimes in areas of maximum benzene concentration the Working Group considers that a limit value in the lowest end of the range, i.e. 2 - 5 µg/m3, to apply in hotspots as well as the urban background, would provide a very high level of protection to the population.

137

The Group notes that achieving such a limit value in the relatively short timescale of 2010 could entail considerable costs in addition to those which will be incurred because of recently agreed legislation on benzene, such as new fuel quality standards. If decision makers consider that a limit value should be set for 2010 they may consider that a limit value of 5 µg/m3 would implement the precautionary approach whilst making allowance for the practicalities of the short timescale. The Working Group notes that costs are likely to be lower than those estimated in the study described in Chapter 4 because of action which will be taken to reduce concentrations of other pollutants such as ozone. It is not however possible to make an estimate of the extent to which costs might be lower.

5.3.2 Option 2
Decision makers may feel that in the light of the uncertainties associated with the risks of benzene and the expectation that further evidence on its mechanism of action will become available, and given the information also on costs, a more measured approach should be taken. Concentrations of benzene will decline from present day levels as a result of legislation already agreed. They will not however decline at the same rate or to the same extent across the European Union. The conclusion may therefore be that an initial limit value should be set to ensure good progress everywhere whilst making explicit provision for a re-examination of the evidence when more data are available. In that case a two stage process could be set up, analogous to that agreed for particulate matter. A first stage limit value would be set to be met as soon as possible – for example 10 µg/m3 to be met in 2007. A second stage limit value would be set for 2015, at the lower end of the range examined (2 – 5 µg/m3) but this would be reviewed at some suitable point. A review should take place when more scientific data are available on which to judge the risks associated with benzene, and when more data are available from Member States on concentrations and practicalities as a result of implementation of the legislation. The Working Group notes that a possible disadvantage of a two-stage process is that action plans for benzene would for the first few years relate only to the Stage 1 limit value. Ideally information should be gathered during Stage 1 on measures which might be taken to reach the Stage 2 limit value. The Working Group recommends that if a two-stage approach is chosen, decision- makers should consider whether Stage 1 should include provision for gathering such information.

138

The Working Group recommends also that the Commission and legislators should consider integrating a future review of benzene with work on other pollutants. This would avoid some of the difficulties in economic analysis of having to assign costs of measures that may be taken on other grounds wholly to benzene.

5.3.3 Option 3
Option 3 is similar to Option 2 in giving greater weight than Option 1 to the uncertainties associated with benzene and making explicit provision for review. Option 2 gives a relatively short time horizon to planners and less certainty over the longer term. It also means, as noted in 5.3.2, that information needed for Stage 2 may not be collected in time for a review. Decision-makers may feel that it would be better to determine a single relatively ambitious limit value, in the range 2 – 5 µg/m3, to be met over a longer timescale, for example 2015. Provision should however be made to review this limit value in ample time before it is met. In that case action plans would relate to this limit value and would provide information on potential abatement options which is not presently available. A potential disadvantage of this option is that actions that could be taken relatively quickly may be deferred, in the knowledge that a review will be carried out well before the limit value must be met. Although the Air Quality Framework Directive includes provision for setting a margin of tolerance, which then declines over time to meet the limit value, it does not require that concentrations should be reduced immediately to below the margin of tolerance. Member States’ obligation is to meet the limit value by the target date. However, the Commission will publish each year a list of areas that are above the margin of tolerance. These areas must also prepare within two years an action plan that shows how they will meet the limit value by the target date and this action plan will be made public. It may be considered that these transparency provisions will be sufficient to ensure early action.

5.4 Reviewing limit values
The difficulties in estimating the costs of action to reduce concentrations of benzene arise primarily from considering benzene in isolation. As this paper makes clear, actions, which would reduce concentrations of benzene, would have substantial secondary benefits. This can be looked at from another point of view. If such actions will be required anyway in order to meet other targets such as reductions in greenhouse gas emissions and

139

emissions of other air pollutants it does not seem sensible to assign all their costs to benzene abatement. Problems of this sort are to some extent inevitable - it is impossible to undertake an integrated analysis of all potential targets and measures to meet them. Nevertheless, it is recommended that any review of limit values for benzene should be carried out as part of an integrated review, considering other important pollutants simultaneously.

140