Influence of environmental parameters on the accuracy of nitrogen dioxide passive diffusion tubes for ambient measurement

Abstract

Two studies at three sites in the UK provided confirmation that systematic positive bias in NO₂ diffusion tube measurement occurred because of changes to “within-tube” chemistry, rather than eddy diffusion at the mouth of the tube. In the first study in Cambridge, UK, sampler overestimation for 1 and 2 week exposures was compared to corresponding time-averaged monitor measurements (NO–NO₂–NO_x, O₃) and weather variables. Nonlinearity between sampler and monitor NO₂ measurements was interpreted in terms of spatial and temporal variations in relative and absolute availability of NO, NO₂ and O₃ at the site. A maximum overestimation occurred for an exposure mean NO₂/NO_x ∼ 0.5. The separate contributions of reduced NO₂ photolysis and eddy diffusion were compared in Study II using samplers of two materials, acrylic and quartz, and of different lengths (40, 55, 71 and 120 mm) at three sites: Norwich background, Cambridge intermediate, London kerbside. For compared sites, NO₂ measured by acrylic samplers was significantly higher than for equivalent quartz samplers. For quartz samplers [NO₂]_mean was only just above the monitor at Norwich and London; sampler/monitor NO₂ = 1.04 (P = 0.59) and 1.01(P = 0.76), respectively. For acrylic samplers the order of [NO₂]_mean was 40 mm > 120 mm > 71 mm ≥ 55 mm. Excepting 40 mm samplers, this accords with a chemical bias where co-diffusing NO and O₃ molecules in longer tubes have more time to react to form excess NO₂. Bias in 40 mm samplers is discussed. Eddy diffusion is negligible for standard samplers because [NO₂]_mean was equivalent for 55 mm and 71 mm acrylic samplers and close to monitor NO₂ for 71 mm quartz tubes. Both studies showed that sampler accuracy was dependent on location. Significantly, overestimation was greatest (∼3–4 ppb) where the NO₂ annual mean was ∼20 ppb, close to the UK and EU air quality standard of 21 ppb.

---

Introduction

Nitrogen dioxide (NO₂) is a key urban pollutant for which routine review and assessment is required by local authorities in the UK under the terms of the Environment Act 1995. Passive diffusion tubes are widely used for ambient NO₂ measurement to determine the extent of polluted zones^1,2 and for comparison with national and international air quality standards.^3,4 Unfortunately, validation of the technique for outdoor use is incomplete and uncertainties remain regarding its accuracy and reliability for air quality assessment.^5–8

In a recent paper,⁹ we have identified a source of inaccuracy arising from differences in sampler preparation methods and advised the adoption of a universal standard method to eliminate this type of error. However, removal of systematic biases, as reported in previous outdoor studies,^5–8 is more problematical. These sources of bias need to be clarified and their influence on sampler accuracy assessed for different sampling conditions. For instance, trials in rural areas have shown good agreement with the reference chemiluminescence technique^10,11 but urban studies have found significant overestimation of NO₂ concentrations.^5,7,8,12 Two studies have also found an aging effect whereby NO₂ uptake declined as sampler exposure time increased.^7,8 The reason for this loss is not known but may be due to limitations in adsorbent capacity or photolytic degradation of bound nitrite. Incomplete conversion of NO₂ gas to nitrite at the triethanolamine surface may also contribute to the aging effect.

One approach to “correcting” sampler NO₂ uptake for these conflicting biases might be to use an effective sampling rate, based on laboratory tests with standard NO₂ atmospheres. This would avoid errors incurred by using the theoretical equation based on Fick's diffusion law. However, it will be shown that positive bias in outdoor sampler measurement is a consequence of constantly changing environmental variables. Apart from the difficulties of reproducing these conditions in laboratory experiments, this implies that the sampling rate would vary according to atmospheric conditions.

Two potential sources of environmental interference have been proposed to explain positive bias in NO₂ measurement. Firstly, eddy diffusion caused by air turbulence at the entrance to the sampling tube could lead to a shortening of the diffusion path.^5,6 Secondly, attenuation of UV radiation through the acrylic walls could reduce daytime photolysis of NO₂ molecules diffusing along the sampler.¹³ Reaction between co-diffusing NO and O₃ molecules continues, causing excess NO₂ formation at the absorbent. A model of sampler diffusion, incorporating changes to “within-tube” chemistry, has been developed by Heal and Cape.¹³ This has been applied to monitoring data from two independent studies in Edinburgh⁷ and Cambridge.⁸ In both studies, modelled NO₂ and sampler NO₂ measurements were in good agreement, indicating that reduced NO₂ photolysis accounted for most of the excess NO₂ recorded by the acrylic samplers. However, wind-turbulence could not be ruled out as an additional source.

The influence of environmental sampling conditions on the accuracy of NO₂ diffusion tubes is explored in this paper using data from two separate studies: the Cambridge study⁸ mentioned above and a three-centre study in London, Cambridge and Norwich. Both investigate the respective contributions of eddy diffusion and “within-tube” chemistry to excess NO₂ formation in acrylic samplers. Results provide strong support for “within-tube” chemistry, showing how excess NO₂ formation in acrylic samplers is modified by spatial and temporal changes in sampling conditions. Relevant environmental measurements taken over the study periods include weather parameters and pollutants (NO, NO₂, NO_x and O₃), obtained as continuous hourly mean measurements and averaged over corresponding sampler exposure times.

Experimental

Sampler preparation, deployment and analysis have been reported previously,⁹ together with the monitoring sites relevant to this paper. Sampling protocols for the two studies are described below; neither has been previously reported.

Study I. Cambridge

At this urban intermediate site, situated 20 m from a main road, NO₂ measurements by standard length (71 mm) acrylic diffusion tubes were compared to those from chemiluminescence over 22 months. A concurrent time–exposure comparison of 1, 2 and 4 week acrylic samplers took place over 11 months of this long-term study. The Cambridge monitoring site was equipped with gas analysers for NO–NO₂–NO_x and O₃, located at ground level, and a roof-top weather station (height = 10 m). Pollutant and weather parameters were collected and stored as continuous hourly mean measurements. Statistical analyses of data sets were performed using MINITAB version 10.5 for Windows.

In the long-term study consecutive 2 week exposures of two replicate samplers, placed on the inlet arm of the monitoring station, were obtained over 22 months commencing March 1995. Two additional samplers were placed within the rain hood protecting the inlet, concurrent with these exposures, for a period of 17 months. The open ends of the tubes were level with the bottom edge of the hood to allow free movement of air whilst sheltering the samplers from wind turbulence. Data from the 4 sampler positions was without loss over these periods but the chemiluminescence data was only available for approximately 17 months out of the 22 month survey because of discontinuities in operation.

The time–exposure study consisted of sampler exposures over consecutive 4 week periods between February and November 1996. For this study the samplers were located on holders spaced at 100 mm intervals along a horizontal baton, fixed behind the hood of the monitoring inlet. The baton was parallel to the road so that exposed samplers were equidistant from the traffic emission source. For each 4 week period, 14 samplers were deployed as follows: 2 samplers exposed for 4 weeks (weeks 1–4); 4 samplers exposed for 2 week periods (2 each for weeks 1–2, 3–4); 8 samplers exposed for 1 week periods (2 each for weeks 1, 2, 3 and 4).

For the duration of this study the 2 week samplers also contributed to the long-term dataset, replacing the samplers attached directly to the monitoring inlet arm. Hence, 6 samplers were located beside the monitor inlet at any one time and a 96% return was achieved on all duplicate samplers. Reported data were the means of duplicate measurements. The 6 sampler positions were changed each week using a randomised layout to avoid any bias due to position. Sampler exchanges took place at the same time each week so that exposure times were exact multiples of 168 h (1 week). Both sites were well protected from vandalism and losses were therefore minimal. Collected samplers were sealed and stored at low temperature in polythene bags and analysed at the end of each 4 week period. Storage time before and after exposure was minimised to avoid contamination and for consistency of procedures. For each 4 week set, 2 unexposed samplers were retained as blanks for laboratory analysis.

Study II. London, Cambridge and Norwich

Experimental design. In this study, NO₂ uptake was compared for samplers of different length (to assess turbulence) and of two materials, acrylic and quartz (to assess chemical overestimation). Three different locations were used to capture the variation in relative and absolute NO, NO₂ and O₃ concentrations found within urban areas, to determine whether location affected sampler accuracy.

Wind-driven turbulence as a cause of NO₂ overestimation was examined, as it has been elsewhere,^5,6 by varying the length of the diffusion tubes [40 mm, 55 mm, 71 mm (standard length) and 120 mm]. If eddy diffusion contributed to overestimation then the reduction in the diffusion path length would be of the same magnitude for all samplers, irrespective of their overall length. This would have the greatest proportional effect on enhanced NO₂ uptake for the shortest tubes, reducing as the tube length increased (i.e. the shortest tubes would record the highest NO₂ concentrations). There would also be no difference in NO₂ measured by acrylic and quartz tubes of the same length.

To determine whether reduced NO₂ photolysis contributed to sampler overestimation, two sets of samplers were deployed composed of two different materials: acrylic (UV-blocking) and quartz (UV-transmitting). Heal and Cape¹³ have estimated that quartz tubing transmits ∼90% of light of wavelengths appropriate to NO₂ photolysis, compared with only ∼20% light transmission for acrylic tubing. A chemical overestimation would therefore be indicated by higher NO₂ concentrations for acrylic than for corresponding quartz samplers. In addition, greater overestimation by the longest acrylic sampler, 120 mm, would occur compared to shorter lengths because of the extra time available for co-diffusing NO and O₃ molecules to react to form excess NO₂. For quartz samplers this would cause little or no difference in accumulated NO₂ for different tube lengths because NO₂ photolysis was taking place. However, some excess NO₂ might form as UV transmission through quartz would be <100% and this might show a length dependence. This experimental design should identify and distinguish between the two potential causes of bias and also determine whether location is a key factor.

Sampler deployment. The three sites were chosen to fit each of the categories defined in the UK NO₂ diffusion tube survey³ by distance from the roadside: urban background (Norwich), intermediate (Cambridge) and kerbside (London). This enabled a wider range of NO₂ measurements to be collected than from one site alone. It was not possible to hold the trials concurrently, with trials commencing at Cambridge before those at Norwich and London. Exposure sets were deployed as detailed below.

Cambridge: Winter, 8 sets, December 1997–March 1998; Summer, 6 sets, March–August 1998.

Norwich: Summer, 7 sets, March–June 1998 terminated because of access changes.

London: Summer, 9 sets, March–August 1998.

In Study II gas analyser measurements were only available at the Norwich and Marylebone Road sites.

As the study was concerned with very small differences between groups of samplers, measurements for each type and length were obtained in triplicate. Calculation of the NO₂ concentrations from extracted nitrite masses¹⁰ took account of changing tube dimensions (length and cross-sectional area) in accordance with the following expression derived from Fick's diffusion law:

where Q = extracted nitrite mass, D_NO2 = diffusion coefficient for NO₂ in air, z = length of diffusion tube, A = internal cross-sectional area of tube and t = collection time.

Tube dimensions were determined as mean measurements obtained for each group used in the study, by material and length.

Blank absorbances. An important criterion was that all the acrylic and quartz tubes fitted securely into the same standard end caps. Leakage around the end caps was tested by exposures of sealed tubes in the laboratory, as well as a set of field blanks exposed at the polluted London site. In both cases, very low mean NO₂ measurements of <1 ppb showed that leakage of air into the samplers via the end caps was not a significant source of error. During the study, blank absorbances were obtained from unexposed samplers, reserved from each prepared batch for each tube type and length, to correct the absorbances of exposed samplers. The blank absorbances obtained from routine exposures gave an equivalent nitrite mass ranging from 0.003–0.025 µg, which were acceptable levels for 2 week exposures at the lowest recorded mean [NO₂] of 9 ppb.

Exposure length. Two week exposures were chosen to accommodate the range of NO₂ mass uptake by the diffusion tubes of 0.25 µg (120 mm tubes at Norwich) to 5.11 µg (40 mm tubes at Marylebone Road). The total nitrite mass accumulating on the TEA absorbent had to be sufficient for analytical determination and on this basis it was not practicable to use a 1 week exposure. The lowest nitrite accumulation at the Norwich site of 0.25 µg was only a factor of 10 above the highest blank value for unexposed samplers.

The use of replicate tube arrays⁹ has been validated by the authors to show that samplers record equivalent NO₂ measurements. As in the earlier time–exposure study (Study I), parallel batons were positioned close to monitor inlets with samplers spaced at 100 mm intervals. Each exposure set comprised 18 samplers: 3 each of 40 mm, 71 mm and 120 mm acrylic and 40 mm, 71 mm and 120 mm quartz samplers. Sets at the Norwich and Marylebone Road sites also included 3 each of 55 mm acrylic samplers.

Precision of replicate measurement

In both studies precision, determined as the relative standard deviation (RSD) of replicate measurements, compared favourably with reported precision from previous studies in Cambridge⁹ and elsewhere.^5,10 For Study I, the long-term mean precision was 3.8% (n = 41) for 2 week exposures outside the monitoring inlet (DT_OUT) and 3.9% (n = 33) for samplers inside the inlet hood (DT_IN). Reproducible NO₂ measurements were also indicated by high correlations between replicates at each position: r = 0.89 for both DT_OUT and DT_IN. The RSD of the difference in replicate measurements was 1.6 ppb and 1.4 ppb for DT_OUT and DT_IN, respectively. Paired Student t-tests indicated there was no difference in NO₂ measured by replicates at the same position so a mean reading was used for subsequent analyses. For Study II mean precision is reported in Table 2 (see below), by type and length of sampler. For both acrylic and quartz samplers, at all lengths, the mean precision was acceptably low at <10% (RSD).

Results and discussion

Accuracy of NO₂ diffusion tube measurement was determined by comparing sampler and equivalent time-averaged chemiluminescence monitor measurements. Sampler/monitor NO₂ ratios (S/M_NO2) provided a measure of the difference between the two techniques and the degree of over- or underestimation by the diffusion tubes. The effects of other variables, both pollutant and weather, was determined by comparing concurrent time-averaged measurements to the corresponding S/M_NO2 values. This approach has assumed that the reference technique was itself free from error and this may not have been valid for all conditions. However, sources of bias discussed in Heal et al.⁸ are thought to be small for UK conditions. In any case the most significant source, over-reading due to interference from peroxyacetylnitrate and nitrous acid, would affect both sampling techniques in the same way.

Study I. Comparison of exposed and sheltered 2 week samplers

Time series for samplers and monitor NO₂. Fig. 1 depicts the time series for NO₂ measured by diffusion tubes DT_IN and DT_OUT, and monitor. The rise and fall in the NO₂ measured by the monitor is mirrored by that of the diffusion tubes but DT_OUT measurements are higher for almost all exposures (32 out of 34 exposures). Diffusion tube measurements were well correlated with those from the monitor (r = 0.74 and 0.70 for DT_IN and DT_OUT, respectively). Mean S/M_NO2, 95% confidence intervals and number of observations were as follows: DT_OUT/monitor = 1.16 ± 0.05 (n = 34) and DT_IN/monitor = 1.03 ± 0.06 (n = 23). Hence exposed samplers were overestimating NO₂ by 16% on average compared to only 3% for sheltered samplers.

Fig. 1 Time series of 2 week NO₂ exposures measured by chemiluminescence monitor and diffusion tubes in exposed and sheltered positions. Diffusion tube data are the means of duplicate measurements.

Paired t-tests of the difference of means found a very highly significant difference for exposed samplers and monitor (P = 0.0005) but no significant difference for sheltered samplers and monitor (P = 0.614). In addition, the two sets of sampler data were very well correlated with each other (r = 0.84, n = 33), but paired t-tests showed a very highly significant difference of means (P = 0.0005).

Differences between winter and summer NO₂ measurements. There was a marked decline in mean winter (October–March) S/M_NO2 compared to summer (April–September): winter mean = 0.98 (n = 11) and 1.13 (n = 17) and summer mean = 1.08 (n = 12) and 1.18 (n = 18) for DT_IN and DT_OUT, respectively. This winter/summer difference in excess NO₂ formation is discussed further in the section Seasonal differences in NO₂ uptake by samplers (below).

Relationship between sampler/monitor NO₂ ratio and weather variables. The differences in NO₂ measurements shown by the exposed and sheltered samplers suggested that wind-driven turbulence was contributing to overestimation. However, no significant correlations between wind speed and S/M_NO2 were found. Wind speed was assessed by using time-averaged exposure means and by categories (% of exposure time above fixed wind speeds of 3, 4 and 5 m s^-1). Using log-transformed data, regression coefficients for S/M_NO2versus mean windspeed were: DT_OUT, R² = 0.28 (n = 24) and DT_IN, R² = 0.24 (n = 23), indicating that windspeed accounted for less than 30% of the variation in S/M_NO2. More importantly, there was no difference in R² for exposed and sheltered samplers. This implied that enhanced NO₂ uptake in exposed tubes was not due to wind turbulence as this would be indicated by a higher correlation with wind speed for exposed than for protected samplers.

Other weather parameters, mean temperature and solar radiation, were also compared to S/M_NO2, but no significant correlations were found. However, lack of a relationship may reflect inappropriate weather measurements, as time-averaged parameters provided only a gross approximation to actual conditions. Air flows at ground-level would be reduced by frictional drag and surface roughness features, such as vegetation and buildings,¹⁴ and might not be adequately represented by wind flows measured at a height. Therefore further investigation was needed of the wind turbulence contribution to enhanced NO₂ uptake by samplers.

Study I. Influence of other pollutant measures on sampler NO₂ measurement

Relationship between sampler and monitor NO₂ measurements. Scatterplots of long-term 2 week data, DT_OUT and DT_IN, against equivalent monitor measurements did not show a 1∶1 correspondence but correlations, r = 0.70 and 0.74 respectively, showed that they were varying in the same way. This indicated that diffusion tubes provided a good measure of relative NO₂ concentrations but not of absolute concentrations. At the lower end of the NO₂ data range (15 ppb to ∼23 ppb) the sampler NO₂ generally exceeded monitor NO₂. However, as NO₂ increased towards a maximum of 29 ppb, there was a tendency for samplers to underestimate monitor NO₂.

For the 1 week data, less affected by the second bias of time-dependent loss of NO₂,⁸ nearly all data points lay above the line of 1∶1 correspondence. However, there was considerable variation in the extent of overestimation by the samplers (mean = 1.24; range = 0.95–1.73). As with the 2 week data, there was a tendency for sampler data at the upper end of the NO₂ range (>23 ppb) to be closer to the monitor NO₂.

It was found that linear regression applied to the 1 and 2 week datasets gave large positive intercepts on the sampler NO₂ axis when monitor NO₂ = 0 ppb. A quadratic fit, indicating a non-linear sampler–monitor relationship, might have been more appropriate but it was concluded that regression analysis was inappropriate for the limited data range.

Relationship between sampler/monitor NO₂ and monitor NO₂. To examine how the relationship between sampler and monitor measurements changed with ambient NO₂ concentrations the excess NO₂ formation, expressed by S/M_NO2, was compared with monitor NO₂, Fig. 2. An inverse linear dependence on monitor NO₂ concentration was found, regression coefficients indicating that a significant proportion of the variation in sampler measurement was explained by the change in monitor measurement: 62% for 1 week exposures; 51% and 54% for 2 week exposures DT_OUT and DT_IN, respectively. The form of the relationship, linear or otherwise, was a matter of debate but the overall trend was clear, S/M_NO2 was highest at low [NO₂] concentrations and lowest at high [NO₂] concentrations (S/M_NO2 ≥ 1 for [NO₂] = 15–20 ppb and S/M_NO2 ≤ 1 for [NO₂] > 25 ppb).

Fig. 2 Correlations between sampler/monitor NO₂ ratios and monitor NO₂ for sampler exposures: (a) 1 week; (b) 2 week; (c) 2 week, summer; and (d) 2 week, winter. The line of equal sampler/monitor correspondence is included.

These findings gave support to “within-tube” excess NO₂ formation, a source which was sensitive to relative and absolute ambient concentrations of the reactants, NO and O₃. In Fig. 2, exposure mean NO₂ concentrations in the range 15–20 ppb were indicative of a daytime atmosphere where both NO and O₃ were available to react to form excess NO₂ within the diffusion tube. (This was supported by statistical modelling¹⁵ of the automated hourly NO₂, NO and O₃ datasets.) Conversely, raised NO₂ concentrations at this site were associated with increased levels of NO_x (NO + NO₂) from nearby traffic emissions. This most frequently occurred in winter when dispersion was inhibited by greater atmospheric stability, and NO₂/NO_x ratios tended towards values associated with tailpipe emissions¹⁶ (∼0.2). A consequence of increased NO_x concentrations was that available O₃ in the roadside atmosphere was rapidly depleted through reaction with NO, inhibiting excess NO₂ formation within samplers. This effect was shown by S/M_NO2 approaching unity at higher [NO₂]. Lack of O₃ availability in winter was also a consequence of reduced photochemical activity, with lower mean [O₃] than in summer.¹⁷

At lower concentrations, [NO₂] < 5 ppb, typifying rural conditions remote from NO_x emission sources, lack of NO in the atmosphere could also inhibit the NO + O₃ reaction in the sampler.⁸ Hence, excess NO₂ formation within exposed samplers will not take place at either low or high [NO₂] when the atmosphere becomes NO-limited or O₃-limited, respectively. This implies that, whilst linear regression lines are depicted in Fig. 2 to show trends, a nonlinear relationship is probably more valid. Indeed, 2 week ratios in Fig. 2 are actually below unity at higher NO₂ concentrations. This may well reflect loss of absorbed NO₂, the opposing systematic bias to that previously described and largely absent in 1 week exposure data, Fig. 2(a). It appears that samplers can both over- and underestimate ambient NO₂ concentrations, depending on length of exposure and proximity to NO_x emission sources. This implies that accuracy is both highly variable and dependent on monitoring location.

Seasonal differences in NO₂ uptake by samplers. Overestimation by samplers appears to be lessened in winter compared to summer [see Differences between winter and summer NO₂ measurements (above)], probably due to seasonal changes in the relative abundance of NO, NO₂ and O₃. This sensitivity may also explain the trends in Figs. 2(c)–(d), where S/M_NO2 has a weaker correlation with monitor NO₂ in summer (r = 0.59) than in winter (r = 0.82). The explanation for this difference is that in summer increased convective turbulence within the low-level atmosphere promotes more rapid dilution of NO emissions away from the roadside.¹⁶ In addition daytime [O₃] is higher than in winter whilst [NO] is lower. Sufficient NO and O₃ is available at this intermediate site over a wider range of [NO₂], hence S/M_NO2 has a weaker association with ambient [NO₂], Fig. 2(c). We have described in the previous section how, in winter, higher [NO₂] near the roadside is accompanied by high [NO] and low [O₃]. But even in winter, as [NO₂] declines improving O₃ availability enhances excess NO₂ formation in samplers. Correlation between S/M_NO2 and ambient NO₂ is therefore sharper in winter and less pronounced in summer.

These seasonal differences in correlations also appear for sheltered samplers, Figs. 2(c)–(d), indicating that samplers have responded in a similar manner to unprotected samplers. This accords with a reduced NO₂ photolysis occurring in all acrylic samplers used in these trials, irrespective of whether sheltered or not.

Heal et al.⁸ have shown that NO, NO₂ and O₃ interactions at this site provide for maximum “modelled” excess NO₂ formation within samplers. The variation in relative and absolute concentrations of NO, NO₂ and O₃ over short timescales is maximised at this location, as highlighted by a principal components model of the NO₂–NO_x–O₃ system derived from the hourly monitoring data.¹⁵ It is expected that a seasonal influence on sampler accuracy would be less obvious for sampler data from background or kerbside sites which were further from, or closer to, NO_x emission sources.

Relationship between sampler/monitor NO₂ and monitor NO/NO₂ and O₃/NO₂ ratios. The following exposure mean relationships were found to apply for nearly every sampling period at the Cambridge site: in summer [NO]_av < [O₃]_av and in winter [O₃]_av < [NO]_av. This shows that atmospheric conditions were NO-limiting in summer and O₃-limiting in winter. The effect of this on the extra NO₂ generated in the tube is examined by comparing S/M_NO2 to the NO/NO₂ ratio for summer data, Fig. 3(a) and (c), and against the O₃/NO₂ ratio for winter data, Fig. 3(b) and (d). Correlation coefficients for 1 week exposures are as follows: r = 0.63 in summer (n = 15); r = 0.74 in winter (n = 12). For 2 week data almost identical correlations occur: r = 0.64 in summer (n = 17); r = 0.70 in winter (n = 17). Although time-averaged data provided only a crude estimate of the real-time fluctuations in NO₂, NO and O₃ in ambient air, the two ratios still accounted for between 40 and 55% of the variability in the sampler overestimation. The strength of the relationships serve to illustrate the underlying nature of the overestimation (because of the relative proportions of NO₂, NO and O₃ found at this location). Even for 2 week exposures the propensity for a chemical overestimation is demonstrated, despite being tempered by the opposing bias of reduction in accumulated NO₂ on the absorbent.

Fig. 3 Correspondence of the sampler/monitor NO₂ ratio to the NO/NO₂ ratio in summer, (a) and (c), and to the O₃/NO₂ ratio in winter, (b) and (d), for 1 and 2 week exposures, respectively.

Examination of these ratios gives insight into the nature of NO₂ overestimation but their utility as correction factors is limited, necessitating the use of gas monitors to obtain accurate NO₂, NO and O₃ concentrations and thus making diffusion tubes redundant.

Study I. Effect of different protective screens on NO₂ uptake by samplers

In Fig. 2(b)–(d) there is an almost constant difference in NO₂ measured by exposed and sheltered samplers, which is independent of NO₂ concentration and must therefore have a different origin. The mean absolute difference in NO₂ concentration is 1.94 ppb or 8.7% of the mean of DT_IN and DT_OUT data sets. Eddy diffusion occurring at the mouth of exposed samplers could be indicated by this difference. Alternatively, the difference could be explained by a stagnation of the air under the inlet hood, reducing NO₂ uptake by protected samplers. Both of these explanations would be independent of ambient NO₂ concentration. However, given the open aspect of the monitoring site, air stagnation is not seen as a likely cause. Differences of UV transmittance for samplers in the two positions is also unlikely, as the Edinburgh study⁷ found no significant difference in NO₂ measured by standard and foil-wrapped samplers: both were UV-blocking.

A third explanation is proposed for the difference between NO₂ uptake by exposed and sheltered samplers, based on a short study using different types of protective screen. Emissions from traffic are subject to both dilution (through transport and mixing) and chemical reaction, taking place through time and space as air moves away from the roadside. Daytime concentrations of NO, NO₂ and O₃ are controlled predominantly by the reactions of the photostationary state equilibrium.¹⁷ Close to the roadside this equilibrium is continually perturbed by fresh NO_x emissions. However, as air moves away from the roadside, both dilution and chemical reaction cause changes in pollutant concentrations. Eventually, at a sufficient distance from emission sources, background concentrations of NO, NO₂ and O₃ are attained. In fact, the results from a spatial NO₂ study in Cambridge¹ and from other studies¹⁶ show that background NO₂ concentrations are often achieved at a short distance from the roadside, ∼20 m.

The placing of a barrier near the mouth of the sampler will impede the natural air flow, creating extra turbulence and delaying entry into the tube. This additional mixing and time delay allows further dilution and chemical reaction of the polluted air before reaching the sampler. On average then, the protected sampler receives air which is closer to a background state of equilibration than for an identical unprotected sampler. If the sampler has maximum protection, while still being in contact with the free atmosphere, the NO₂ recorded may approach that of local background. If the degree of screening is less, sampler NO₂ will lie between that of an equivalent unprotected sampler and the background concentration.

This explanation is supported by the results of a short-term trial reported in Table 1. For equivalent exposures well-protected samplers, placed within a miniature Stevenson screen next to the inlet at the Cambridge site, recorded the same [NO₂] as samplers at an urban background site. At the same time [NO₂] recorded by samplers on the monitor inlet, protected by simple plastic mesh surrounds, was slightly lower than for unprotected samplers. For samplers under the monitor hood, [NO₂] was even lower. The order of measured [NO₂] can be seen as reflecting the level of protection of the samplers. However, for samplers exposed at the urban background site, no difference was found in [NO₂] recorded by unprotected samplers and those protected by the plastic mesh surrounds (Table 1). In the well-equilibrated atmosphere at the background site, well away from direct sources of NO_x emissions, no further reduction in [NO₂] is expected. Despite the brevity of the survey (four consecutive 2 week exposures), these results give support to a process of dynamic change taking place at near-road locations, which leads to lower NO₂ levels for sheltered diffusion tubes.

Table 1 Comparison of NO₂ diffusion tube measurements for wind-protected and exposed sites at intermediate and urban background sites in Cambridge (4 consecutive exposures)^a

Location	Exposure dates
	20/11/95–4/12/95	4/12/95–18/12/95	18/12/95–8/01/96	8/1/96–22/1/96
a All positions, except roadside, are the means of duplicate tubes; NO2 concentrations in ppb.
East Road monitor—
Outside hood	28.0	21.9	21.9	27.1
Inside hood	26.6	20.8	21.7	23.1
Outside, mesh-screened	27.1	21.2	20.9	25.3
Inside Stevenson screen	21.0	17.3	16.8	20.2
East Road kerbside	38.8	31.2	29.4	32.6
Urban background—
Unsheltered	23.3	17.5	16.0	23.1
Mesh-screened	23.2	17.8	16.6	24.4

---

In the long-term survey, NO₂ concentrations from the samplers under the monitor hood appeared to be closer to the corresponding monitor NO₂ values. If the explanation presented above is valid, sheltered samplers will not provide greater accuracy of measurement. In fact, external protection may not be advisable, as it may alter the composition of the air being sampled. The extent of change will depend on the type of screen used and the proximity to direct sources of NO_x emissions. Other studies in the UK¹² and elsewhere² provide evidence of this variability in wind-protected samplers, underestimating by ∼10% on average compared to monitor measurements.

One implication of a “within-tube” chemical bias in sampler NO₂ measurement is that overestimation only takes place during daylight hours. After sunset, NO₂ photolysis ceases and the chemistry of NO and O₃ within the diffusion tube is equivalent to that taking place in the ambient atmosphere. This suggests that diffusion tubes act as efficient monitors during the night hours and has important implications for indoor monitoring, where NO₂ photolysis is also absent. Diffusion tube measurements indoors should be more comparable to NO₂ concentrations by chemiluminescent monitors.

Study II. Comparison of eddy diffusion and “within-tube” chemistry as sources of excess NO₂ in diffusion tube measurement

For the sites in the 3-centre study there were no losses of samplers, although a small number (<3%) were found to be defective during laboratory analyses. Reported data were the means of triplicate measurements except where only duplicate sampler measurements were available.

Evaluation of sources of overestimation. For each site, measurement differences between samplers in individual exposure sets were compared firstly by tube length (120 mm, 71 mm, 55 mm, 40 mm) and then by tube material (acrylic or quartz). At all three sites, [NO₂] measurements within individual exposure sets were always in the order 40 mm ≥ 120 mm ≥ 71 mm ≥ 55 mm for both acrylic and quartz samplers. This order is also reflected in sampler mean [NO₂] measurements in Table 2 for each location. At the Norwich and London sites, all sampler NO₂ measurements were referenced to concurrent time-averaged chemiluminescence measurements (Table 2). This was not possible at the Cambridge site where measurement differences between different types of samplers were expressed by corresponding mean NO₂ ratios (Tables 3 and 4). At all sites the significance of measurement differences was determined using Student's paired t-tests (2-tailed) on log-transformed data, with P-values indicating the degree of confidence to be placed in the results.

Table 2 Summary of NO₂ measurements by diffusion tubes (by type and length) and chemiluminescent monitor^a

Site	Monitor	Acrylic tubes				Quartz tubes
		120 mm	71 mm	40 mm	55 mm	120 mm	71 mm	40 mm
a NO2 concentrations are quoted in ppb.
Cambridge—
Winter mean (8 sets)		24.4	23.2	30.1		25.3	24.7	32.2
Summer mean (6 sets)		21.3	18.4			17.9	17.5
RSD (%) (all sets)		2.6	5.3	5.9	6.2	6.4	6.9	8.1
Marylebone Road—
Mean of 8 sets	44.7	51.2	46.7		47.6	46.3	45.2
Mean of 5 sets	44.4	49.5	45.4	53.4	46.2	44.0	42.2	52.3
RSD (%) (all sets)		2.1	4.1	3.4	6.9	5.6	3.9	5.7
Norwich—
Mean of 6 sets	10.4	12.6	11.6	13.0	11.7	10.8	10.6	12.1
RSD (%)		3.0	4.4	5.9	4.2	6.0	6.3	5.7

---

Table 3 Cambridge: comparison of NO₂ measurements by acrylic and quartz diffusion tubes for each length

Tube length	A/Q
	Winter	Summer
120 mm—
No. of observations	8	6
Mean A/Q ratio and 95% confidence interval	0.97 ± 0.14	1.21 ± 0.14
Significant paired t-test difference: P-value	0.239	0.009
Correlation (r)	0.95	0.87
71 mm—
No. of observations	8	6
Mean A/Q ratio and 95% confidence interval	0.94 ± 0.06	1.07 ± 0.12
Significant paired t-test difference: P-value	0.041	0.203
Correlation (r)	0.92	0.77
40 mm—
No. of observations	8
Mean A/Q ratio and 95% confidence interval	0.94 ± 0.006
Significant paired t-test difference: P-value	0.136
Correlation (r)	0.66

---

Table 4 Cambridge NO₂ measurements by diffusion tubes, lengths compared for each tube type

Ratio of diffusion	Acrylic/acrylic		Quartz/quartz
	Winter	Summer	Winter	Summer
120 mm/71 mm—
No. of observations	8	6	8	6
Mean ratio and 95% confidence interval	1.05 ± 0.03	1.15 ± 0.08	1.02 ± 0.04	1.02 ± 0.07
Significant paired t-test difference: P-value	0.007	0.004	0.351	0.448
Correlation (r)	0.98	0.96	0.97	0.98
120 mm/40 mm—
No. of observations	8		8
Mean ratio and 95% confidence interval	0.81 ± 0.07		0.79 ± 0.10
Significant paired t-test difference: P-value	0.0005		0.002
Correlation (r)	0.77		0.68
71 mm/40 mm—
No. of observations	8		8
Mean ratio and 95% confidence interval	0.77 ± 0.06		0.77 ± 0.08
Significant paired t-test difference: P-value	0.0005		0.0005
Correlation (r)	0.83		0.75

---

Cambridge datasets

At the Cambridge site there was a marked change in the behaviour of the samplers between winter and summer. In winter the mean [NO₂] for quartz (Q) samplers was slightly higher than for acrylic (A) samplers of equivalent length. This order was reversed in summer with mean [NO₂]_A > mean [NO₂]_Q. In Table 3, comparative data for acrylic and quartz samplers of the same length showed that the mean A/Q NO₂ ratio was below unity in winter and above unity from April onwards. This abrupt change in the ratio coincided with a sudden rise in mean daily temperatures and marked the onset of summer.

Comparison of samplers by type of material. The differences in winter/summer sampling were investigated using paired t-tests of means and 95% confidence intervals of mean A/Q ratios. The results, Table 3, showed that in winter acrylic and quartz samplers provided equivalent measurements for all sampler lengths: [NO₂]_A ≈ [NO₂]_Q for 120 mm, 71 mm and 40 mm tubes. By contrast in summer the mean A/Q ratios indicated that acrylic samplers were recording much higher [NO₂] than quartz samplers of the same length. The difference was greatest for the longest tube length (21% for 120 mm; 7% for 71 mm). Paired t-tests confirmed a highly significant difference between the acrylic and quartz 120 mm tubes (P = 0.009) but not for 71 mm samplers where the mean difference was quite small.

Comparison of samplers by tube length. Results for length comparisons for each material, separated into winter and summer data, are reported in Table 4. For NO₂ measurements by acrylic samplers, in both seasons there were highly significant differences between different tube lengths (P ≤ 0.01). Further, the A_120 mm/A_71 mm [NO₂] ratio increased markedly from winter (1.05) to summer (1.15). For quartz samplers there was no significant difference in NO₂ for 120 mm and 71 mm lengths, either in summer or winter (1.02 in both seasons). Quartz samplers, for these lengths, were recording equivalent NO₂ measurements. However, there was an enhanced uptake of NO₂ for 40 mm compared to 120 mm and 71 mm samplers for both acrylic and quartz samplers.

Norwich and Marylebone Road datasets

The range of NO₂ reflected the site location: 9 to 19 ppb at the urban background site at Norwich and 38 to 61 ppb at Marylebone Road, the London kerbside site. The results of these sites were consistent with the summer data from Cambridge; the order of [NO₂] concentrations for individual exposures was predominantly 40  > 120  > 71  ≥ 55 mm for both acrylic and quartz samplers. Mean exposures (Table 2) and sampler/monitor NO₂ ratios (Table 5) also followed this order.

Table 5 Comparison of NO₂ measured by diffusion tubes with monitor NO₂ at Norwich and Marylebone Road

	Acrylic tube/monitor				Quartz tube/monitor
	120 mm	71 mm	40 mm	55 mm	120 mm	71 mm	40 mm
Marylebone Road—
No. of sets	9	9	7	8	8	8	5
Mean ratio and 95% confidence interval	1.15±0.07	1.05±0.09	1.20±0.16	1.04±0.07	1.04±0.09	1.01±0.09	1.18±0.16
Signif. paired t-test difference: P-value	0.003	0.35	0.002	0.30	0.36	0.763	0.02
Correlation (r)	0.38	0.37	0.62	0.32	0.65	0.36	-0.05
Norwich—
No. of sets	6	6	6	6	6	6	6
Mean ratio and 95% confidence interval	1.24±0.14	1.14±0.15	1.26±0.15	1.12±0.14	1.06±0.13	1.04±0.15	1.19±0.11
Signif. paired t-test difference: P-value	0.005	0.09	0.053	0.098	0.35	0.59	0.006
Correlation (r)	0.93	0.88	0.88	0.89	0.88	0.41	0.95

---

Comparison of acrylic and quartz samplers with monitor NO₂ measurements. The mean sampler/monitor NO₂ ratios in Table 5 showed that while there was some degree of overestimation in both types of sampler, irrespective of tube length, this was greatest for acrylic samplers. For the standard acrylic length, 71 mm, a 14% average overestimation at Norwich was close to the 16% average for Cambridge from Study I. By contrast, 71 mm quartz samplers recorded only a 4% overestimation at this site. At Marylebone Road overestimation was much less than at Norwich for all sampler types: 5% and 1% for 71 mm acrylic and quartz samplers, respectively.

Significant paired t-tests, Table 5, showed whether the differences between sampler and monitor NO₂ measurements were significant. At both sites, [NO₂]_A was significantly different from [NO₂]_M for 120 mm samplers. However, excess NO₂ formation in 71 mm acrylic samplers was not sufficient to be confirmed by t-tests. For 120 mm and 71 mm quartz samplers at both sites, t-tests and sampler/monitor ratios indicated that [NO₂]_Q was not different from [NO₂]_M. Results also showed that for both acrylic and quartz 40 mm samplers, sampler and monitor [NO₂] were different.

Comparison with Cambridge data. So that results could be compared to Cambridge data, sampler NO₂ measurements were compared with each other, firstly by type of material and then by tube lengths.
Acrylic/quartz sampler differences. The difference between acrylic and quartz NO₂ measurements was greater for 120 mm than for 71 mm samplers and more marked at the Norwich site than at Marylebone Road. Paired t-tests confirmed these differences for 120 mm samplers (P ≤ 0.01) at both sites and for 71 mm samplers (P = 0.005) at Norwich only.
Comparison of sampler tube lengths. For acrylic samplers there were significant differences between 120 mm, 71 mm and 40 mm lengths: P ≤ 0.02. However for quartz samplers, at both sites, no significant difference was found between 120 mm and 71 mm tubes. Differences between 40 mm and both 120 mm and 71 mm quartz samplers were significant (P ≤ 0.01). All results were in agreement with the Cambridge summer data set.

Summary of evidence. Results from Study II give strong support to reduced NO₂ photolysis as the main source of measurement bias in acrylic samplers. The bias is substantially removed in the UV-transparent quartz samplers. The evidence is summarised as follows.
Acrylic–quartz comparison. At all three sites the mean NO₂ measurements by acrylic samplers were significantly higher than quartz samplers of equivalent length.

The difference between acrylic and quartz showed a length dependence, i.e., it was greatest for 120 mm lengths, slightly less for 71 mm and small but apparent for 40 mm lengths.

The mean NO₂ from quartz samplers was only just above the monitor NO₂ at Norwich and Marylebone Road; the sampler/monitor NO₂ ratio = 1.04 (P = 0.59) and 1.01 (P = 0.76), respectively. The slight positive bias remaining, which was greater in 120 mm tubes, was explained by a small loss of UV radiation through quartz tube walls (see Experimental design).

Acrylic length comparison. The mean NO₂ measured by acrylic samplers of different lengths was in the order 40 mm ≥ 120 mm ≥ 71 mm ≥ 55 mm, which, excepting 40 mm, corresponded to a chemical bias (where co-diffusing NO and O₃ molecules in the longer tubes had more time to react to form excess NO₂).

Results for 40 mm samplers suggested enhanced NO₂ uptake caused by eddy diffusion at the mouth of the tube. However, this could also have been due to (i) increased atmospheric turbulence for very short samplers caused by the proximity of the sampler holders or (ii) a breakdown of the diffusion process for this length.

More importantly, the results showed that eddy diffusion had a negligible effect for all other tube lengths because mean NO₂ was equivalent for 55 mm and 71 mm acrylic tubes (P = 0.766 and 0.335 for Norwich and Marylebone Road, respectively), and mean NO₂ for 71 mm quartz samplers was close to monitor NO₂.

Explaining differences in sampler accuracy

Effects of season and location. Winter/summer differences in the behaviour of acrylic samplers have been found in both studies. In Study II the seasonal differences in Cambridge data can be explained, as in Study I, by enhanced NO and O₃ availability in summer, which leads to greater overestimation by acrylic samplers. In winter, the similarity between acrylic and quartz sampler NO₂ measurements can be attributed to an O₃-limited atmosphere in which little or no excess NO₂ formation occurred.

Evidence from both studies indicates that, for outdoor NO₂ measurement, sampler accuracy varies both within and between different locations. This is because accuracy is dependent on constantly changing NO, NO₂ and O₃ concentrations of the monitored atmosphere. Factors that influence the level of accuracy of exposed samplers are the proximity to NO_x emission sources, atmospheric chemistry and prevailing meteorology as these determine relative and absolute NO and O₃ concentrations.¹⁶ This also explains differences in sampler accuracy found in earlier studies between rural^10,11 and urban^5,7,8,12 sites. The complexity of interactive processes means that it is extremely difficult to apply a simple correction factor to sampler NO₂ measurements.

Estimating accuracy from the sampler/monitor NO₂ relationship. The studies presented here have shown that the sampler/monitor NO₂ relationship is not a reliable method of estimating excess NO₂ formation as opposing systematic biases can lead to both over and underestimation of NO₂ concentrations. This is demonstrated by the nonlinearity of the sampler/monitor NO₂ relationship, shown in Fig. 4 using data from all three sites. It is clear from Fig. 4(a) that linear data fits for individual sites are inappropriate and forcing regression lines through the origin¹² is probably unjustified. Data variability may be more appropriately described, Fig. 4(b), using an outer-limit envelope to enclose data points. This is supported by comparative sampler and monitor data compiled by Campbell et al.,⁵ the scatterplot having a remarkably similar pattern.

Fig. 4 Sampler/monitor NO₂ ratio versus monitor NO₂/NO_x for (a) 1 week and (b) 2 week sampler exposures. The line of equal sampler/monitor correspondence is included.

Estimating accuracy using the location mean NO₂/NO_x ratio. It may be possible to use the exposure mean NO₂/NO_x ratio, which is a measure of the relative availability of NO and O₃ in the atmosphere, as a marker for overestimation by samplers. This is explored in Fig. 5 by plotting S/M_NO2 against monitor NO₂/NO_x using the Cambridge data from Study I together with Norwich and Marylebone Road datasets from Study II. The three sites capture the full range of NO₂/NO_x ratios: Cambridge mean = 0.54 (range 0.74–0.39); Norwich mean = 0.81 (range 0.88–0.66); Marylebone Road mean = 0.26 (range 0.34–0.22). Overestimation changes with location with a maximum potential occurring for mean NO₂/NO_x∼0.5, confirming the same relationship modelled by Heal et al.⁸ Availability of both NO and O₃ is greatest around this mid-point value and is further evidence of a chemical source of bias in sampler NO₂ measurement.

Fig. 5 Sampler NO₂ against monitor NO₂ for 3 locations: (a) comparing correlations and regression lines by individual sites and (b) replacing regression lines with an outer-limit data envelope over all sites.

The magnitude of the NO₂/NO_x ratio reflects location type in terms of proximity to NO_x emission sources and dispersion characteristics of the locality.^15,16 If this ratio is known for a particular site, the extent of overestimation by diffusion tube measurement may be estimated.

Use of NO₂ diffusion tubes for assessment against air quality standards

Diffusion tube measurements are now directly comparable to the new UK and EU annual mean standards, both set at 21 ppb. Mean absolute sampler overestimations at the three study sites are as follows: Norwich ∼1.4 ppb, Cambridge ∼3.3 ppb, Marylebone Road ∼2.3 ppb. This indicates a maximum overestimation of 3–4 ppb occurring when [NO₂]_annual is close to the limit level; illustrated by Cambridge 1996 annual means of 20.9 ppb and 23.5 ppb, monitor and sampler NO₂, respectively. This indicates that for a sampler [NO₂]_annual = 25 ppb, air quality may actually be in compliance with EU and UK standards.

Trends in NO₂ may be difficult to identify in long-term surveys³ because overestimation increases as O₃ availability improves: a direct consequence of declining NO concentrations. Real reductions in NO₂ of a few ppb may be masked by greater overestimation by samplers.

It is important to note that, even for worst case conditions, the magnitude of the systematic biases is still within accuracy limits recommended by the EU Daughter Directive for indicative monitoring of ±25%. A maximum bias for urban intermediate and background sites of ∼20% (based on 2 week exposures from the Cambridge site) would therefore be acceptable. However, for samplers exposed for longer periods (4–5 weeks) at heavily trafficked roadside sites, accuracy may fall below acceptable limits.

Optimising sampler accuracy. Sampler accuracy could be improved either by the use of quartz tubing or an alternative material which does not absorb UV within the critical wavelength range for NO₂ photolysis (280–430 nm). This needs further investigation. Other improvements to design which warrant further investigation are a shortening of tube length (to reduce NO₂ gain) combined with use of a narrower tube (to preserve Fick's law for shorter samplers). Optimum exposure times have been shown to be 1 or 2 weeks⁸ as longer exposures are more susceptible to a time-dependent loss of accumulated NO₂. Underestimation by samplers is a possibility at heavily trafficked roadside sites, particularly in winter when [O₃] is frequently depleted.

Positioning samplers within shelters is not recommended as this may interfere with the composition of the monitored atmosphere. In the studies presented here sites were reasonably sheltered, however, and this may have helped to eliminate eddy diffusion as a significant source of excess NO₂. By optimising sampler design as described, an accuracy for NO₂ samplers of ±10% compared to chemiluminescence measurements should be achievable.

Acknowledgements

The provision of a bursary to C. Kirby by Anglia Polytechnic University is gratefully acknowledged. Laboratory facilities and materials were made available by the Department of Geography, Cambridge University. Thanks are also due to staff from the Environmental Health Departments of Norwich City Council and Westminster City Council, London, for site access. Guidance on gas monitor operation was provided by the South East Institute of Public Health, King's College, London, as well as access to the Marylebone Road site.

References

1. C. Kirby, A. Greig and T. Drye, Environ. Monit. Assess., 1998, 52, 65.
2. M. Glasius, M. F. Carlsen, T. S. Hansen and C. Lohse, Atmos. Environ., 1999, 33, 1177.
3. T. Bush, D. Mooney and K. Stevenson, UK Nitrogen Dioxide Survey 1998. Report AEAT 4565, AEA Technology, National Environmental Technology Centre, Culham, UK, 2000..
4. H. Van Reeuwijk, P. H. Fischer, H. Harssema, D. J. Briggs, K. Smallbone and E. Lebret, Environ. Monit. Assess., 1998, 50, 37.
5. G. W. Campbell, J. R. Stedman and K. J. Stevenson, Atmos. Environ., 1994, 28, 477.
6. A. J. Gair and S. A. Penkett, Atmos. Environ., 1995, 29, 2529.
7. M. R. Heal, M. A. O'Donoghue and J. N. Cape, Atmos. Environ., 1999, 33, 513.
8. M. R. Heal, C. Kirby and J. N. Cape, Environ. Monit. Assess., 2000, 62, 39.
9. C. Kirby, M. Fox and J. Waterhouse, J. Environ. Monit., 2000, 2, 307.
10. D. H. Atkins, J. Sandalls, D. V. Law, A. M. Hough and K. J. Stevenson, Harwell Laboratory Report AERE-R12133, National Environment Technology Centre, Culham, UK, 1986..
11. K. J. Hargreaves, PhD Thesis, University of Nottingham, UK, 1989..
12. T. Bush, K. J. Stevenson, S. Moorcroft and S. Smith, Validation of nitrogen dioxide diffusion tube methodology, Stanger Science and Environment/National Environmental Technology Centre, Culham, UK, 1998. Available on DETR web page http://www.environment.detr.gov.uk/airq/aqinfo.htm..
13. M. R. Heal and J. N. Cape, Atmos. Environ., 1997, 31, 1911.
14. T. R. Oke, Boundary Layer Climates, Cambridge University Press, Cambridge, UK, 1993..
15. C. Kirby, PhD Thesis, Anglia Polytechnic University, Cambridge, UK, 1999..
16. Quality of Urban Air Review Group, Urban Air Quality in the United Kingdom, First Report of QUARG, Department of the Environment, London, UK, 1993..
17. UK Photochemical Oxidants Review Group, Ozone in the United Kingdom, Department of the Environment, London, UK, 1987..

---

Footnote

† Address for correspondence: 11 Henry Morris Road, Impington, Cambridge, UK CB4 9YG. E-mail: kirby_carolyn@hotmail.com

---