Richard P. Hamilton and Mathew R. Heal*
School of Chemistry, University of Edinburgh, West Mains Road, Edinburgh, UK EH9 3JJ. E-mail: m.heal@ed.ac.uk; Fax: +44 (0)131 6504743; Tel: +44 (0)131 6504764
First published on 17th November 2003
This study was carried out in response to suggestions that the measurement of NO2 by Palmes-type passive diffusion tubes (PDT) is affected by the method of preparation of the triethanolamine (TEA) absorbent coating on the grids. The following combinations of factors were investigated: TEA solvent (acetone or water), volume composition of TEA in solvent (50% or 20%), and grid coating method (dipping in solution prior to assembly or pipetting solution on after assembly). Duplicate PDTs prepared by each of the 8 methods were exposed in parallel, in urban air, for a total of 80 separate 1 week exposures. NO2 concentrations derived from PDTs prepared by pipetting methods were significantly less precise than concentrations from dipping methods, with mean RSDs for duplicate measurements of 13.8% and 8.5%, respectively (n = 316 each category). Pipetting methods using solutions of 50% TEA composition were particularly imprecise (mean RSD 17.2%). Data from PDTs prepared by pipetting methods were systematically more poorly correlated with each other and with data from co-located chemiluminescence analysers, than corresponding data from PDTs prepared by dipping methods, indicating that more consistent accuracy was also obtained by the latter PDTs. The statistical evidence suggested that PDTs prepared by pipetting 50% TEA in water generally gave lower NO2 concentrations. Although this is in agreement with a previous study, it is also possible that such an observation here may be a statistical artefact given the demonstrably poorer precision of this method. The general tendency of PDTs to show positive bias in NO2 measurement in urban air in 1 week exposures was again evident in this study (mean biases at roadside and urban centre locations of +35% (n = 475) and +18% (n = 112), respectively) consistent with augmentation of within-tube NO2 flux by chemical reaction between co-diffusing NO and O3. Overall, it is recommended that the pipetting method of PDT grid preparation is avoided, or at least investigated further, because of the apparent degradation in precision and accuracy of NO2 measurement. Potential reasons for the effect are discussed.
Continuous analysers, such as the chemiluminescence analyser, are usually specified as the reference method for measurement of NO2, but passive diffusion samplers (particularly the Palmes-type passive diffusion tube,1 PDT) remain widely used for indicative assessment of spatial variations of longer-term average NO2 concentrations because of their ease of use and low cost. For example, an NO2 concentration map for the entire UK is interpolated from a national network of PDTs.2
The fundamental principle of PDT measurement is that NO2 molecules are captured by triethanolamine (TEA) absorbent, N(CH2CH-2OH)3, coating the grids inside the closed end of the tube, with the rate of capture of NO2 determined by the diffusive flux along the tube. Since TEA is extremely viscous at room temperature, it is dissolved in a solvent for application to the PDT grids. After exposure, trapped NO2 is extracted into aqueous solution as nitrite ions, NO2−, and quantified.
While it is accepted that NO2 PDTs are not as precise as continuous analysers, considerable debate remains regarding the accuracy of NO2 PDTs in the field. This debate surrounds the significance or not of specific factors giving rise to specific biases in deriving NO2 concentrations. Some investigators have concluded that NO2 PDTs are accurate within acceptable tolerances of precision,3 whilst others have argued that specific phenomena such as wind-induced turbulence at the entrance to the tube,4 or chemical production of additional NO2 by co-diffusing NO and O3 within the tube5,6 contribute to an effective over-measurement of NO2 by PDTs, or that an exposure-duration related loss of trapped NO2 contributes to an effective under-measurement.7
More recently, two studies8,9 have suggested that the accuracy of NO2 PDT measurement may also be affected by the way in which the TEA absorbent at the end of the tube is prepared. The observation has potentially major consequences for comparing NO2 PDT data since there is no generally-recognised standard method of tube preparation. The variables in preparation are that the TEA absorbent may be dissolved in acetone or deionised water, in volume ratios ranging from 20%–50%, and the solution applied to the grids either by dipping the grids into the solution prior to PDT assembly, or by pipetting a known volume of solution onto the grids after assembly. The report9 produced by NETCEN, the UK body currently responsible for collating and validating data from the UK national network of NO2 PDTs, recommended that the preferred preparation method was to pipette an unspecified volume of a 20% TEA:water solution onto grids already assembled within the PDT cap. The recommendation was based on the observation that this preparation method yielded NO2 measurements closest in value to those of a chemiluminescence analyser, whilst PDTs prepared by dipping grids into a 50% TEA:acetone solution (probably the most widespread method), or by pipetting a 50% TEA:water solution onto assembled grids, were reported to yield NO2 measurements consistently higher, or lower, respectively, than the analyser. The study was conducted in a laboratory chamber. The field investigation by Kirby et al.8 also reported that tubes prepared by pipetting 50% TEA:water solution onto assembled grids yielded lower NO2 measurement but that there was no difference in NO2 measurement from tubes prepared using smaller volume ratios of TEA in water or using grids dipped in 50% TEA in acetone. Conversely, a recent review of routine PDT data from UK local authority networks reported no discernible effect of tube preparation method on PDT measurements.10
One difficulty in trying to rationalise observations concerning the effect, or not, of absorbent preparation method is uncertainty regarding the actual mechanism of NO2 complexation by TEA. A mechanism proposed by Glasius et al.,11 that yields a 1∶1 ratio between NO2 in air and NO2− in solution, is shown in eqn. (1):
2NO2 + N(CH2CH2OH)3 + 2OH− → 2NO2− + −O–+N(CH2CH2OH)3 + H2O | (1) |
The required hydroxyl ions are postulated to arise from the dissociation of TEA in water molecules present in the air, so the reaction will not take place in completely dry air, in accord with observations that TEA hydration is important for quantitative NO2 sampling.12 The OH− ions may also derive from water molecules in the absorbent solution itself (if present), and it has been suggested that the extent of this process may account for the apparent lower trapping efficiency of different molar ratio TEA:water absorbent solutions.8
It is not clear from the studies cited above whether the physical process itself of coating grids (i.e. dipping or pipetting) contributes to differences in NO2 measurement. It is also difficult to rationalise how different preparation methodologies may influence measurement. Given the ambiguity of the above studies and the continued importance of PDTs for ambient NO2 measurement, the impetus for the current work was a detailed investigation of whether, and how, absorbent preparation method affects NO2 concentrations derived from PDTs. Eight different grid preparation methods were compared in simultaneous exposures in urban air. Uniquely, this study also compared a sub-set of the PDT NO2 measurements with the NO2 concentration predicted to be measured by a PDT when within-tube production of NO2 along the diffusion path of the tube is also taken into account.5 To limit the impact on data interpretation of the further confounding issue of the loss of nitrite evident during longer exposure periods,7 all exposures were of 1 week duration only.
Eight different methods of applying TEA to PDT grids were compared, comprising all combinations of two choices of three two-level factors: TEA solvent (acetone or deionised water), TEA concentration (50% or 20% by volume in the solvent), and application method of solution to grid (dipping or pipetting). The preparation methods, and their labelling nomenclature, are summarised in Table 1. For dipped method preparations, grids were immersed in the appropriate solution for approximately 5 min, laid on tissue paper, and patted gently with another piece of tissue paper to remove excess solution. Tubes were then assembled and capped. For pipetted methods, a volume of 50 µl of solution was pipetted directly onto the two grids within the capped end of a tube already assembled. Tubes were left to stand for approximately 3 min before the remaining open end was capped.
Absorbent solution composition | Grid preparation method | |
---|---|---|
Dipped | Pipetted | |
50% TEA, 50% acetone | A | E |
20% TEA, 80% acetone | B | F |
50% TEA, 50% deionised water | C | G |
20% TEA, 80% deionised water | D | H |
Tubes were exposed for 1 week at three sites in central Edinburgh: Princes Street gardens (PSt), Castle Street (CSt), and Haymarket (HMt). The PSt site is classified as an urban centre (city centre but not roadside), while CSt and Hmt are roadside sites (1–5 m from a busy road).
PDTs were always deployed in duplicate. The eight pairs of tubes (plus a field blank of method A) were arranged in random order at each site, adjacent to the inlet of a NOx chemiluminescence analyser. The PSt site also had an O3 analyser. Hourly averaged NO, NO2 and O3 data at this site were used as input to a numerical model5,13 to calculate the amount of additional NO2 created and trapped within the PDTs, during each exposure, by chemical reaction between NO and O3 also diffusing inside the tube.
PDT exposures were carried out between November 2001 and March 2002 and again between November 2002 and March 2003, yielding data for a total of 30 1 week exposures for each preparation method at each site, except for the PSt site at which the continuous analyser ceased operation in December 2002.
After exposure, trapped nitrite in the tubes was extracted into deionised water and quantified by the standard sulfanilamide/NEDA colorimetric method (absorbance measurement at 540 nm). Independent duplicate sets of nitrite calibration standards were prepared each week and the average ambient [NO2] during the exposure calculated from the NO2− calibration curves using 0.154 cm2 s−1 as the diffusion coefficient of NO2 in air.
![]() | ||
Fig. 1 Scatter plot of duplicate determinations of NO2 concentration for PDTs prepared by (a) all dipping methods, (b) all pipetting methods. Tubes in each replicate pair are randomly assigned the label 1 or 2. |
Fig.1 clearly shows that measurement precision from tubes prepared by dipping methods A–D is considerably better than precision from tubes prepared by pipetting methods E–H. Precision is particularly poor for methods E and G (Fig. 1b) in which an absorbent solution of 50% TEA in acetone or water, respectively, is pipetted directly onto the grids after PDT assembly. The two other pipetting preparation methods F and H show better precision although instances of poor precision remain.
Table 2 summarises the precision RSD data by preparation method. Method E has the worst precision, with mean RSD of 20.9%. The mean RSD of method G is also high (13.5%), although a median RSD of 8.2% indicates that the mean is adversely affected by some instances of very poor precision (visible in Fig. 1b). The three most precise methods are dipping methods. The mean and median RSD values for all dipping methods (8.5% and 6.7%, respectively) are significantly better than the corresponding values (13.8% and 9.0%) for all pipetting methods. With the exception of methods E and G, mean RSD values of 7.7–10.6% for each of the other 6 methods compare well with previously quoted mean RSD values of 8%,14 ∼10%15 and <4%6 for PDT precision.
Preparation method | Mean % RSD | Median % RSD | n |
---|---|---|---|
A (dip, 50%, acetone) | 10.4 | 7.7 | 80 |
B (dip, 20%, acetone) | 7.7 | 6.5 | 80 |
C (dip, 50%, water) | 8.2 | 7.4 | 77 |
D (dip, 20%, water) | 7.7 | 5.3 | 79 |
E (pip, 50%, acetone) | 20.9 | 14.1 | 79 |
F (pip, 20%, acetone) | 10.4 | 7.4 | 80 |
G (pip, 50%, water) | 13.5 | 8.2 | 77 |
H (pip, 20%, water) | 10.6 | 7.5 | 80 |
All dipped | 8.5 | 6.7 | 316 |
All pipetted | 13.8 | 9.0 | 316 |
All | 11.2 | 7.8 | 632 |
Time-series comparisons between NO2 concentrations derived from each PDT preparation method and the co-located chemiluminescence analyser for all exposure periods are shown in Fig. 2. The entire dataset comprises up to 80 exposures for each preparation method.
![]() | ||
Fig. 2 Time-series of the mean (of duplicate) NO2 concentration determined from PDTs prepared by 8 different methods and exposed at 3 different urban locations in Edinburgh: (a) Princes St, (b) Castle St, (c) Haymarket. Also shown is the exposure-average NO2 concentration measured by co-located chemiluminescence analysers and, for the Princes St site, the NO2 concentration determined by a computer model of PDT sampling that also incorporates reaction within the tube between NO and O3. Identifiers 1–15 and 18–32 correspond to 1 week exposure periods between Nov 01–Mar 02, and Nov 02–Mar 03, respectively. |
The first observation from Fig. 2 is the highly significant trend for PDT NO2 values to exceed the exposure-average chemiluminescence NO2 concentration (positive bias), regardless of the PDT preparation method (P < 0.001, paired t-tests). The mean overestimation of analyser NO2 by PDT across all data (n = 587) is 32%. The mean overestimations by PDT at each site are 18% for PSt (n = 112), 41% for CSt (n = 238) and 29% for HMt (n = 237).
Fig. 2a also shows the model-derived NO2 concentration expected for a PDT at the PSt site when additional within-tube chemical production of NO2 (from reaction between the known ambient concentrations of NO and O3 during each exposure period) is also included. The average ratio between the modelled PDT NO2 concentration and analyser NO2 concentration for this sub-set of exposures is 1.26 (n = 14, range 1.13–1.40), confirming again the intrinsic capacity for PDTs to significantly overestimate NO2 because of within-tube chemistry at locations near fluctuating strong sources of NO (e.g. near roads).5–7 The average PDT:analyser NO2 ratio, regardless of PDT preparation method, for exposures for which a modelled value is available is 1.18 (n = 112), which compares well with the mean ratio of 1.26 predicted by the chemical overestimation model. (Note that the observed ratio is likely to be slightly lower, on average, than the model-predicted ratio because of some exposure-duration decline in PDT efficacy even over the 1 week exposures used in this study7). In fact, for this sub-set of PDT data which can be compared with model-predicted values, NO2 concentrations derived from PDT methods A–F do not differ significantly from model concentrations, whilst concentrations from methods G and H are significantly lower (paired t-test).
The smaller positive bias, on average, of PDTs at the PSt site, compared with the CSt and HMt sites is again expected, since the former site is situated further from the roadside source of NO allowing more time for oxidation of NO to NO2 in the air mass before it enters the tube.
The Spearman rank correlation coefficients between NO2 concentrations derived from each PDT preparation method and the continuous analyser are shown in Table 3. The correlations within dipped PDT data (all r > 0.73) and between dipped PDT data and analyser data (r values 0.58–0.75) are consistently greater than the correlations within pipetted PDT data (all r < 0.67) or between pipetted PDT data and analyser data (r values 0.55–0.68). The poorer correlations associated with data from pipetting methods are probably partly a consequence of the lower precision associated with these data (Section 3.1), but must also reflect a greater inaccuracy of pipetting-method data to track the variation of NO2 concentration with exposure/location.
This latter observation is clearly evident in Fig. 2 which shows that, although PDTs prepared by all 8 methods show positive bias in NO2 concentration, there are systematic variations in the accuracy, i.e. the extent to which PDT measurements from different preparation methods are clustered together along a common trend. In general, NO2 data derived from PDTs prepared by dipping methods A–D (solid symbols) are both more closely clustered to each other, and follow a more tightly-constrained trend of values with respect to the analyser values, than the NO2 data from PDTs prepared by pipetting methods E–H (open symbols). These latter data are considerably more erratic in accuracy. Fig. 2 shows that data from method E (but also methods F and G) are particularly erratic. However, there needs to be a lot of caution in making statements concerning the accuracy of method E in particular, since the very poor precision associated with this method (Fig. 1b) implies very low confidence in the absolute values.
The existence of differences in NO2 concentration with tube preparation method is confirmed statistically (P < 0.001) by the non-parametric Friedman test, in which the NO2 values from the 8 preparation methods from each exposure are ranked from 1–8 and the sum of the ranks assigned to each method over the 76 complete sets compared. The preparation method(s) that differ significantly are determined by comparing the differences between the rank sums for the methods with an appropriate critical value (Table 4). The table shows that NO2 concentrations from preparation method G are consistently lower than for other methods and significantly lower than for methods F and B.
The comparison of PDT absorbent preparation methods shows a very clear tendency for data from PDTs prepared by dipping the grids in TEA-solvent prior to tube assembly to be more precise than data from PDTs prepared by pipetting the TEA-solvent on to the grids after tube assembly (Fig. 1). There is also a persistent tendency for data from PDTs prepared by pipetting to show greater fluctuation in accuracy relative to each other and to continuous analyser measurements (Fig. 2 and Table 3). In so far as it is possible to make any statement regarding systematic trend in relative accuracy of preparation method (because of the demonstrable variation in precision with preparation method), it appears that method G (pipetting 50% TEA in water) generally yields the lowest concentrations, and method F (pipetting 20% TEA in acetone) the highest (Table 4). Although the observation of a trend for lower NO2 concentrations derived from tubes prepared by pipetting 50% TEA in water on grids is consistent with the two previous investigations,8,9 it may also here be simply a statistical artefact arising from poorer precision.
All PDTs were subject to identical environmental conditions during each exposure, so factors such as chemical augmentation, shortening of diffusion path length, and exposure-dependent degradation cannot explain differences in precision and accuracy associated with grid preparation method. Factors that can be postulated to contribute are the chemical conversion of NO2 at the absorbent, or the surface area over which NO2 absorption is assumed to occur.
The molar ratio of TEA added to the grids in any of the methods studied should be well in excess of the amount of NO2 collected during an exposure period. This was shown in this study by weighing the grids pre- and post-addition of the TEA solution, and is in agreement with similar calculations by Kirby et al.8 Therefore, there should be sufficient TEA present for 100% conversion of NO2 whatever the method of preparation.
The mechanism of complexation in eqn. (1) was proposed by Glasius et al.11 to accord with the observation (using FAB-MS) that TEA N-oxide was the only TEA-derived product of the reaction. The mechanism can also be written stoichiometrically equivalently as:
2NO2 + N(CH2CH2OH)3 + H2O → 2NO2− + −O–+N(CH2CH2OH)3 + 2H+ | (2) |
The second factor to consider is surface area of absorption. When calculating the ambient NO2 concentration from the total NO2− captured by a PDT it is assumed that NO2 is complexed by TEA absorbent across a surface area equal to the internal cross-section of the tube. It seems reasonable to assume that surface tension effects will ensure that a grid submerged in TEA-solvent solution will be consistently and evenly coated with solution on each occasion. In contrast, it seems less clear that using a pipette to introduce a small volume of solution onto grids within a cap (whether the tube has already been inserted into the cap or not) will always result in an absorbent surface area exactly equal to the value assumed in the calculation of NO2 concentration. Thus, failure to coat the entire grid surface with TEA will reduce the flux of captured NO2, leading to a negative bias in derived NO2 concentration, while any “creep” of solution up the inside wall of an assembled tube will increase the absorbent surface area and lead to a positive bias in derived NO2 concentration. The latter may occur in the pipetting methodology if tubes are tilted or inverted too quickly after pipetting and solution runs down the inside walls of the tube. (It is assumed that the physical act of patting dry dipped grids prior to assembly means that there is no possibility of the above issues for dipped preparation methods). Thus it is proposed that variation from one PDT to the next in the practical action of assembling tubes via a pipetting methodology may explain the greater imprecision of data derived from these preparation methods. The two pipetting methods that yield least precision (methods E and G) both use 50% TEA solutions. Pure TEA is extremely viscous at room temperature, which is the reason it has to be dissolved in a solvent in the first place. A 50% TEA solution is still fairly viscous, so the greater difficulty in reproducibly dispensing such a viscous solution from the pipette may also contribute to the greater imprecision for these specific preparation methods. In the work presented here, tubes were prepared by two different analysts, so whatever the specific cause(s) of the greater imprecision in pipetting methods, it was not analyst specific.
Finally, it is important to emphasise that any effect of preparation method on PDT precision and accuracy will be independent of the effect of any other operational factor(s) that may influence PDT accuracy (for example, a decline in measured NO2 from longer exposure, and/or an enhancement of measured NO2 from within-tube chemistry, or wind-induced turbulence). The NO2 measurement derived from a PDT exposure will be the composite of all relevant influences. The identified influences of these factors do not negate the continued use of PDTs for indicative NO2 measurement, but it is important that PDT data are always interpreted with due consideration of all parameters associated with their exposures.
NO2 concentrations derived from PDTs prepared by pipetting TEA-solvent solutions onto the grids after assembly are also, in general, more variable in accuracy than concentrations from PDTs prepared by dipped methods. There is only weak evidence that one or more method gives systematically different NO2 across all exposures.
For 1 week exposures at these roadside and urban centre locations, NO2 concentrations from PDTs prepared by all methods are greater than chemiluminescence analyser NO2 concentrations, with greater discrepancy at the roadside locations.
Overall, it is concluded from this study that PDT performance is influenced by the physical method of grid preparation, whereby pipetting absorbent solution onto grids after tube assembly (particularly pipetting solutions of high (50%) TEA composition) leads to an apparent degradation in precision and accuracy of NO2 measurement. It is recommended that these preparation methods be avoided, or at least investigated further.
This journal is © The Royal Society of Chemistry 2004 |