A Comparative Study of the Performance of Passive Samplers

ABSTRACT

Atmospheric concentrations of benzene, toluene, ethylbenzene, and xylenes (o-xylene and m,p-xylene) were assessed in the Tricity area (Gdańsk-Sopot-Gdynia, Poland) with the use of two types of passive samplers: permeation (homemade passive samplers) and diffusive (Radiello and Orsa 5). Samples were collected during 2008 at selected sites in the Tricity area at monitoring stations belonging to the agency of Regional Air Quality Monitoring Foundation. The field study was conducted to compare the performance of these two different types of passive samplers. A statistical approach was formulated, and the experimental data were evaluated using the paired t test, Wilcoxon signed rank-sum test, and Friedman analysis of variance. All the statistical results confirm the hypothesis that the differences between the performances of the three sampling devices are highly significant. Despite the fact that data obtained with the homemade passive sampler indicated that the results were higher compared with those for the Radiello and Orsa 5 diffusive samplers, the authors note that all differences between the homemade permeation sampler and the Radiello and Orsa 5 diffusive samplers are positive.

IMPLICATION

This study examined the performance of two types of passive samplers (permeation and diffusive type) for the long-term characterization of time-weighted mean concentration of analytes in a given environmental compartment. The results of this evaluation demonstrated that passive sampling was found to be suitable for consideration as part of an emerging strategy for monitoring a wide spectrum of priority pollutants in ambient air. The samples of analytes obtained can serve as a source of information in reference to creating the precise distribution of pollutant concentrations in both time and space and analyzing trends in pollutant concentrations in outdoor air. The data obtained may aid in the identification of analyte sources and modeling air pollution transport.

INTRODUCTION

The elevated levels of volatile organic compounds (VOCs) in the air over many conurbations have spawned numerous attempts on various scales to clean up the air. Air quality management systems have been set up to supply information on concentration levels of selected atmospheric pollutants. Levels of analytes detected in samples collected at particular places and times are increasingly being used in models for predicting the spread of pollutants and in compiling distribution maps of pollutant levels. For the persons responsible for environment quality control, the biggest problem is choosing the proper field sampling technique, which should have the following features: detection of an appropriate concentration range of target analytes, known accuracy and precision of measurements, clearly defined influence of interferents on measurement results, simplicity and rapidity of measurement, and low running costs.

Passive sampling is currently one of main areas of development in analysis, particularly in the monitoring of environmental pollutants. The main advantage of passive sample collection/enrichment technology is that in situ sample collection is very simple, which is particularly important in the case of long-term measurements and is also the reason that passive sampling appears to be an interesting alternative to the routinely used dynamic methods. Passive air sampling can provide atmospheric concentrations of pollutants over a period of months and simultaneously at a number of sites at low cost. With this method, pollutant concentrations are integrated over the whole exposure time, making them immune to accidental, extreme variations. Passive sampling can be used to investigate the atmospheric concentration of organic pollutants not only at the local scale1,2 but also at the continental and global scales.3 Information obtained in this way is suitable for producing a long-term overview of pollutant concentration levels.4,5 The time-weighted average concentrations of analytes obtained by passive sampling give a better picture than short-term concentrations of the effect of pollutants on human health, because they reflect the long-term influence of pollutants on human health.

The principle of operation of passive samplers is based on mass transport, described by Fick's first law of diffusion. Passive samplers fall into one of two categories, depending on the type of diffusion barrier used6,7: passive diffusion samplers in which the transport of analytes takes place by way of free diffusion of the analyte through a stagnant gas layer; and passive permeation samplers, in which the transport of analytes takes place by way of permeation of the analyte through a semipermeable membrane.

To a large extent, the type of analytical information obtained using passive sampling technology depends on the accumulation regimen in which passive samplers operate during field exposure. In general, a passive sampling device is designed to operate in two different accumulation regimens:

•	In the kinetic and time-integrative uptake phase (rate of mass transfer to the receiving phase is linearly proportional to the difference in chemical potential of the contaminant in the receiving phase and sample). These kinds of passive samplers are called linear uptake passive samplers.
•	In the at-equilibrium regimen, described by the partition coefficient between the receiving phase and the sample matrix. These kinds of passive samplers are called equilibrium passive samplers.

The objective of this study was to obtain information on the level of atmospheric air pollution due to BTEX (benzene, toluene, ethylbenzene, and xylenes) in the Tricity area (Gdańsk-Sopot-Gdynia, Poland) by means of passive sampling (dosimetry). Because of the diversity of passive samplers available on the market, an attempt has been made to assess the applicability of these samplers at different sampling rates of analytes and on the basis of different phenomena (diffusion and permeation) of mass transport for continuous urban air quality control. To accomplish this goal, a statistical approach was formulated, and the experimental data were compared using the paired t test, Wilcoxon signed rank-sum test, and Friedman analysis of variance (ANOVA).

MATERIALS AND METHODS

Study Design

In the assessment of atmospheric air quality, the choice of sampling points determines whether the sample will be representative and whether it will reflect the real state of the monitored air. Therefore, to obtain reliable estimates of background BTEX pollution of the atmospheric air in and around the Tricity conurbation, sampling points that would reflect the actual state of the environment had to be chosen. The locations of the sampling sites were chosen so that no point or linear sources of emission were in the direct vicinity of the sites and surrounding areas. Thus, 10 monitoring stations designated by the Regional Air Quality Monitoring Foundation were used. On the basis of results of long-term studies, the localization of these stations was determined by factors such as weather conditions, human population density, and existing databases on the emissions of pollutants from point and surface sources.

To evaluate the atmospheric average urban background BTEX concentration in the Tricity area, 12 1-month passive sampling campaigns were conducted (for the Orsa 5 passive sampler and homemade permeative passive sampler). To avoid saturation of the sorbent, the exposure time of the Radiello passive sampler was shortened to 2 weeks and, consequently, 24 exposures were conducted.

Passive Sampler

For isolation and enrichment of the analytes from atmospheric air, three sampling devices were used:

•	A homemade, badge-type permeation passive sampler equipped with a 50-μm-thick semipermeable membrane of polydimethylsiloxane (SSPM100, Speciality Silicone Products Inc.). The sorbent was activated charcoal (40–60 mesh; Gryfskand, Gryfino, Poland), which has a large specific surface area (>1000 m2/g) and hence a large sorption capacity and is not particularly sensitive to variations in temperature or atmospheric pressure. These properties of activated charcoal make it ideally suited for long-term sample collection; analyte enrichment, a significant parameter where passive sampling is concerned, takes place concurrently.8,9
•	An Orsa 5 (Dräger, Lübeck, Germany), a tube-type diffusive passive sampler with a short diffusion path length (L). The sorbent is activated charcoal (made from coconut husks; granulation 0.4–0.8 mm), bounded on either side by a cellulose acetate diffusion barrier, which defines the diffusion length within the sampler.10
•	A Radiello radial diffusive passive sampler, designed and developed by the Fondazione Salvatore Maugieri (Padua, Italy). This sampling system is made up of a cylindrical adsorbing cartridge housed coaxially inside a cylindrical diffusive body of polycarbonate and microporous polyethylene. It combines different characteristics common to both types: the diffusive surface is the cylinder itself, the sampling area is large, and the diffusion length is short. Diffusion is radial, and the analyte passes across a microporous cylinder before reaching an inner stainless steel net cylinder containing 300 mg of Carbo-graph 4 (a thermally desorbable adsorbent) as an adsorbent. Because of the high sampling rate, care must be taken not to saturate the sorbent with the component of interest.11,12 Radiello diffusive samplers filled with a thermally desorbable adsorbent have been evaluated for the monitoring of BTEX according to the European Standard EN 13528-2 as a reference passive sampler.13

The use of passive samplers under real conditions for monitoring and assessing atmospheric air quality requires a knowledge of the parameters responsible for the rate of analyte sampling by a particular type of sampler. Then it becomes possible to determine the time-weighted average concentrations of analytes in the air from a knowledge of the sampler exposure time and the mass of compounds retained by the sampler sorbent during exposure. For the commercially available passive diffusion samplers such as the Radiello and Orsa 5, routinely used for monitoring atmospheric air quality, the parameters responsible for the analyte sampling rate are supplied by their manufacturers. The results obtained using the Radiello and Orsa 5 samplers were corrected for temperature to give the numerical value of the rate constant of sampling analytes from atmospheric air.14,15

Passive permeation samplers fitted with a polydimethylsiloxane semipermeable membrane have hitherto been used only to assess the quality of indoor air. For these samplers, the parameter responsible for the rate of analyte sampling from the air, the numerical value of the calibration constants (k) for each compound determined, was measured experimentally on the basis of tests in exposure chambers in an atmosphere of standard gas mixtures exclusively at a temperature of 23 °C.16,17 Although the dosimeter calibration constants for each compound were not corrected for temperature, the possibility of extending the range of applicability of these samplers was examined.

To ensure that the samples of atmospheric air were representative and homogeneous, two exposed samplers of each kind were placed 20 cm from one another at the measuring stations during the monitoring campaignn. They were mounted at a height of ∼3 m above the ground in specially designed acid-resistant steel screens to protect them from the elements (precipitation and strong winds). In addition, in accordance with Good Laboratory Practice, one unexposed (i.e., isolated from the external environment) passive sampler of each type was installed at each station to obtain field blanks. After exposure, the passive samplers were sealed and isolated from the outside air, in accordance with the procedures relevant to each type. Each Orsa 5 passive sampler was placed in a separate glass vessel, which was then sealed off with a polyethylene screw top. The sorption medium in the Radiello samplers was removed from the cylindrical diffusion casing and placed in a glass vessel, which was then tightly stoppered with a polyethylene bung. The permeation samplers were closed with a polyethylene lid. All of the samplers were taken to the laboratory where they were stored at 5 °C until analysis. The analytes were released from the sorbent, and the final determinations were done no later than 1 week after the conclusion of the exposure period.

Sample Analysis

Liberation of Analytes from Sorption Bed and Chromatographic Conditions

For the Orsa 5 passive samplers and the homemade permeation passive samplers, all experiments were performed using an HP 5890 gas chromatograph (Hewlett-Packard, Palo Alto, CA) equipped with a flame ionization detector (280 °C) and a splitless injector (inlet temperature 200 °C; injection volume 2 μL). The analytes were liberated from the sorbent bed by extraction with 1 mL of CS₂. To confirm the identity of BTEX, randomly selected CS₂ extracts were analyzed by means of a gas chromatograph-mass selective detector.

The analytes in the Radiello diffusive samplers were liberated from the sorbent bed by two-step thermal desorption. All experiments were performed using an Agilent Technologies 6890 gas chromatograph coupled with a 5873 network inert mass spectrometer (mass spectrometry) and a thermal desorber (limited version 2.00, Unity Markes International, South Queensferry, U.K.).

Further details of the measurement methods were described in previous publications.18,19

Chemical Standards

External calibration was performed with a volatile calibration mix (VCM) containing 13 VOCs (including BTEX) at 2000 μg/mL (Supelco). Six standard solutions were prepared by diluting the VCM in methanol for gas chromatography (Merck). For solvent extraction, the concentration of standard solutions ranged between 1 and 25 μg/mL and for thermal desorption ranged between 50 and 2000 μg/mL. Calibration solutions were freshly prepared just before calibration.

Quality Control/Quality Assurance

The quality assurance and quality control procedures included analysis of field blanks and parallel samples (for each type of passive samplers) and triplicate measurements of CS₂ extracts (for the homemade and Orsa 5 samplers). The precision estimated from replicate analyses of the standards and samples was within ±10%. For field blanks, unexposed passive samplers were analyzed for BTEX under conditions identical to those used in the analysis of samplers exposed to atmospheric air. The method limit of detection (MDL) was defined using the equations to calculate BTEX concentrations in atmospheric air for each analyte and each type of passive sampler, where the mass of analyte collected on the sorbent bed is the average mass obtained from the unexposed samplers (blanks) being used.18,19 The MDL for benzene for the Orsa 5 and passive permeation samplers for a 1-month exposure was 0.6 μg/m³; for the other compounds monitored it was 0.3 μg/m³. Consequently, the method quantification limit for benzene for the Orsa 5 and the homemade passive permeation sampler was 1.7 μg/m³; for the other compounds monitored it was 1 μg/m³. In general, BTEX concentrations measured using the Orsa 5 and the homemade passive permeation sampler were substantially higher in the samples than in the field blanks.

Statistical Analysis

Statistical analysis was performed using Statistica 7 for Windows (Stat-Soft, Inc.). Pearson and Spearman rank correlation tests were used to examine the strength of associations among the BTEX concentrations. The differences among the samplers were compared with the paired t test, Wilcoxon signed rank-sum test, and Friedman ANOVA nonparametric tests.20–22

RESULTS AND DISCUSSION

The main goal of the research was the field comparison of passive samplers at different sampling rates of analytes (Table 1) based on different phenomena (diffusion and permeation) to verify the possibility of extending the range of applicability of a homemade permeation passive sampler. Summary statistics (arithmetic mean and median, minimum, maximum, SD, and also skewness and kurtosis) for the final results (12 months, 10 monitoring stations) are shown in Table 2 and plotted in Figure 1. The patterns in Figure 1 are in good agreement with the results in Table 2 and suggest that there is considerable similarity between the Radiello and Orsa 5 diffusive samplers but a large difference between these two and the permeation homemade sampler with respect to the central tendency (mean and median) and the spread of results (SD range, including skewness and kurtosis). Higher concentrations of all compounds were observed for the permeation homemade sampler. In addition, the correlations between the BTEX concentrations and meteorological parameters presented in Table 3 illustrate a low influence of atmospheric conditions on the results generated.

Table 1. Sampling rate of passive samplers for a given analyte

Type of Passive Devices	Q (ml/min)
	B	T	EB	oX	m,pX
R	26.8	30.0	24.6	26.6	24.6
O	6.3	5.6	5.1	5.3	5.3
Hm	6.0	6.9	8.5	10.5	8.8
Notes: Q = sampling rate; B = benzene; T = toluene; EB = ethylbenzene; oX = o-xylene; m,pX = m,p-xylene; R = Radiello; O = Orsa 5; Hm = homemade.

Table 2. Statistical results

Compound	N	Mean (μg/m^3)	Median (μg/m^3)	Minimum (μg/m^3)	Maximum (μg/m^3)	SD (μg/m^3)	Skewness	Kurtosis
B^Hm	120	8.581	8.440	1.366	19.054	4.379	0.374	−0.637
B^R	120	1.430	1.107	0.163	4.418	1.125	0.900	−0.239
B^0	120	2.605	2.152	0.122	11.216	1.855	1.509	3.583
T^Hm	120	12.092	11.224	1.742	37.824	5.400	1.254	3.450
T^R	120	2.172	1.840	0.225	5.819	1.234	0.788	−0.071
T^0	120	2.670	2.193	0.877	9.772	1.486	1.819	4.706
EB^Hm	120	4.451	3.730	0.667	14.521	2.696	1.498	2.167
EB^R	120	0.725	0.502	0.084	3.998	0.700	2.069	5.271
EB^0	120	1.032	0.840	0.271	3.608	0.657	1.447	2.023
oX^Hm	120	3.763	3.127	0.580	13.922	2.473	1.754	3.134
oX^R	120	0.647	0.443	0.063	4.140	0.653	2.555	9.009
oX^0	120	0.881	0.736	0.228	3.596	0.546	2.175	6.825
m,pX^Hm	120	15.588	12.167	3.784	67.182	10.955	2.146	5.374
m,pX^R	120	2.511	1.957	0.134	10.807	2.085	1.766	3.847
m,pX^0	120	2.683	1.971	0.572	14.901	2.393	2.514	7.856
S^Hm	120	19.352	15.197	4.364	81.104	13.303	2.048	4.849
S^R	120	3.141	2.535	0.134	14.455	2.695	1.924	4.773
S^0	120	3.564	2.708	0.834	18.497	2.921	2.462	7.774
Notes: N = number of samples; B = benzene; T = toluene; EB = ethylbenzene; oX = o-xylene; m,pX = m,p-xylene; Hm = homemade; R = Radiello; 0 = Orsa 5; S = sum of xylenes.

Table 3. Pearson and Spearman rank correlation coefficients between the concentrations of BTEX compounds and meteorological parameters

	Pearson Correlation Coefficient				Spearman Rank Correlation Coefficient
	T (°C)	H (%)	WS (m sec^−1)	P (hPa)	T (°C)	H (%)	WS (m sec^−1)	P (hPa)
B^R	−0.537	−0.040	−0.047	0.443	−0.525	0.078	0.103	0.444
T^R	0.321	−0.446	−0.268	0.540	0.322	−0.466	−0.513	0.616
EB^R	0.214	−0.229	−0.136	0.408	0.239	−0.347	−0.302	0.548
oX^R	0.240	−0.219	−0.168	0.465	0.269	−0.317	−0.349	0.534
m,pX^R	0.314	−0.359	−0.260	0.412	0.332	−0.449	−0.432	0.560
S^R	0.303	−0.331	−0.243	0.436	0.318	−0.421	−0.414	0.560
B^Hm	−0.197	0.034	0.046	0.330	−0.209	0.123	−0.092	0.311
T^Hm	−0.592	−0.076	0.115	−0.122	−0.606	−0.030	0.409	−0.069
EB^Hm	−0.010	−0.578	−0.124	0.222	−0.042	−0.260	0.163	0.042
oX^Hm	0.091	−0.322	−0.051	0.135	0.128	−0.321	−0.024	0.140
m,pX^Hm	−0.042	−0.021	−0.043	−0.120	−0.156	0.025	0.237	−0.121
S^Hm	−0.023	−0.069	−0.046	−0.084	−0.127	−0.018	0.224	−0.098
B^0	−0.382	−0.073	−0.000	−0.135	−0.443	−0.026	0.314	−0.053
T^0	−0.497	−0.275	−0.203	0.129	−0.494	−0.218	0.183	0.171
EB^0	−0.380	−0.375	−0.170	0.251	−0.408	−0.237	0.087	0.308
oX^0	−0.411	−0.341	−0.227	0.043	−0.393	−0.386	0.180	0.112
m,pX^0	−0.427	−0.295	−0.130	0.118	−0.392	−0.254	0.199	0.246
S^0	−0.438	−0.320	−0.164	0.100	−0.440	−0.253	0.236	0.214
T	1.000	−0.391	−0.398	0.069	1.000	−0.470	−0.641	0.040
H	−0.391	1.000	0.460	−0.383	−0.470	1.000	0.347	−0.391
WS	−0.398	0.460	1.000	−0.316	−0.641	0.347	1.000	−0.494
P	0.069	−0.383	−0.316	1.000	0.040	−0.391	−0.494	1.000
Notes: T = temperature; H = relative humidity; WS = wind speed; P = atmospheric pressure; B = benzene; T = toluene; EB = ethylbenzene; oX = o-xylene; m,pX =m,p-xylene; R = Radiello; Hm = homemade; 0 = Orsa 5; S = sum of xylenes.

Figure 1. Box and whisker plot of the results obtained with the three samplers.

A low influence of atmospheric conditions on the data also results from the data showing fluctuations in monthly average BTEX concentrations. Figure 2 shows monthly average concentrations of analytes at station 10, where the BTEX are measured by BTEX monitor (Chrompack CP 7001).23 It is interesting to note that despite the differences in the BTEX concentration values obtained with different types of passive samplers, the concentration profiles of the compounds monitored for particular stations are similar.

Figure 2. Monthly fluctuations in average concentrations of BTEX in atmospheric air at station number 10 in 2008.

The comparison of monthly BTEX concentrations determined using active sampling and each type of passive samplers was examined by means of the linear regression method.24,25 The BTEX concentrations determined with the Chrompack CP 7001 monitor were plotted versus those determined with each type of passive sampler. The results are presented in Figures 3 and 4. In addition, the confidence band of the monthly BTEX concentrations determined using active sampling is plotted in each case to help visualize the estimated range of values.26

Figure 3. Comparison of the time-weighted monthly average benzene concentrations in the atmospheric air obtained with passive samplers and the Chrompack CP 7001 monitor at station number 10 (– – –, confidence bands; ——, regression curve).

Figure 4. Comparison of the time-weighted monthly average toluene concentrations in the atmospheric air obtained with passive samplers and the Chrompack CP 7001 monitor at station number 10 (– – –, confidence bands; ——, regression curve).

For the difference between the two values to be insignificant, the dependence should be linear (y =bx +a), the line should pass through the origin of the coordinate system, and the slope should be close to unity. In other words, the parameters used for the validation of the proposed approach to compare monthly concentrations are the slope b, the intercept a, and the regression coefficient r. The previous conditions with respect to the slope b (t_b calc [(1 − b)/sec _b ] ≤ t _cr =2.014) were not fulfilled for all passive samplers (Table 4) with the exception of the Orsa 5 sampler for ethylbenzene. Thus, the slope b values were significantly different from the expected values of β₀ =1, which means that the differences between the BTEX concentrations determined using passive samplers were statistically significant at the probability level p = 95% and for f = n – 2 = 46 df.

Table 4. Statistical parameters of the correlation analysis of results (the concentration levels of BTEX) obtained using a Chrompack CP 7001 monitor and passive dosimeters on the stage of the sampling of analytes from ambient air

Analyte	b	sb	tb calc	a	sa	ta calc	r
B^R	0.66	0.06	6.12	0.42	0.11	3.81	0.97
T^R	0.69	0.11	2.89	0.50	0.27	1.85	0.90
EB^R	0.57	0.10	4.38	0.52	0.12	4.43	0.88
S^R	0.44	0.16	3.43	2.86	0.70	4.11	0.65
B^0	−0.06	0.11	9.27	1.53	0.44	3.45	−0.14
T^0	0.58	0.21	2.02	0.59	0.56	1.04	0.66
EB^0	0.67	0.26	1.24	0.27	0.33	0.83	0.63
S^0	0.62	0.15	2.43	2.10	0.66	3.20	0.79
B^Hm	0.17	0.04	20.34	−0.16	0.40	0.41	0.80
T^Hm	0.13	0.03	25.24	0.37	0.46	0.80	0.78
EB^Hm	0.05	0.08	12.50	0.77	0.45	1.71	0.21
S^Hm	−0.03	0.04	23.59	5.25	1.10	4.80	−0.20
Notes: b = slope; sb = SD of the slope; a = intercept; sa = SD of the intercept; r = regression coefficient; B = benzene; T = toluene; EB = ethylbenzene; oX = o-xylene; m,pX = m,p-xylene; R = Radiello; 0 = Orsa 5; Hm = homemade; S = sum of xylenes.

The previous conditions with respect to the intercept a (t_a calc [(a − 0)/sec _a ] ≤ t _cr =2.014) were not fulfilled in most cases with the Radiello and Orsa 5 samplers (Table 4) with the exception of toluene (Radiello) and toluene with ethylbenzene (Orsa 5). Thus, the intercept a values were significantly different from the expected values of β₀ =0, which means that the differences between the BTEX concentrations determined using Radiello and Orsa 5 passive samplers were statistically significant. For the homemade passive sampler, it was found that the intercept a was not significantly different from the expected values for benzene, toluene, and ethylbenzene.

The regression coefficients r for the Radiello and homemade passive samplers were higher than r _cr at the probability level p = 95% and for f = n – 2 =46 df. For the Orsa 5 sampler, it was found that the values of coefficient r were higher than r _cr for toluene, ethylbenzene, and xylenes.

On the basis of the linear regression method if can be concluded that the differences between the concentrations determined using passive samplers and the active sampling technique are statistically significant in most cases at the probability level p = 95% and for f = n – 2 =46df, but the data are statistically correlated.

With regard to the experimental design and data structure, the paired t test and nonparametric tests evidently seem to be the best suited to demonstrate similarities or differences between the results obtained with the three samplers. The statistical results for the paired t test (Table 5) indicate that the individual concentrations are quite different; that is, the differences between the results from each sampler are highly significant (p < 0.00001), except for the difference between the concentrations of m,p-xylene collected with the Radiello and Orsa 5 diffusive samplers, which is not significant (p > 0.12056). In addition, all differences between the homemade permeation sampler and the Radiello and Orsa 5 diffusive samplers are positive. The Wilcoxon signed rank-sum test was also applied to the differences between measurements. Here, the differences are first ranked without regard for sign. The smallest difference is given rank 1, and ties are assigned average ranks. The sign of the difference is assigned to each rank. The positive and negative ranks aresummed and the absolute value of the smaller rank (T) is then compared with the standard critical values for testing the hypothesis. Because in this case the number of measurements was very large, the statistic Z, which is supposed to follow a standard normal distribution, was used. The results (p < 0.00001, except for the difference between the concentrations of m,p-xylene collected with the Radiello and Orsa 5 diffusive samplers, which is not significant [p > 0.117635]) are very similar to the results obtained with the paired t test and strongly endorse the same conclusion that the difference between the permeation and diffusive samplers is highly significant. Finally, the Friedman ANOVA, a test based on a nonparametric multiple comparison, was applied to compare the differences among the samplers. The results confirm the hypothesis that the differences between the results of the three sampling techniques are highly significant (p <0.00001).

Table 5. Results obtained from applying the paired t test

Pairs	D	s d	t	df	p
B^R-B^Hm	−7.151	3.723	−20.774	116	0.000000
B^R-B^0	−1.194	2.268	−5.671	115	0.000000
B^Hm-B^0	5.960	4.156	15.443	115	0.000000
T^R-T^Hm	−9.921	4.890	−21.942	116	0.000000
T^R-T^0	−0.515	1.332	−4.169	115	0.000059
T^Hm-T^0	9.310	4.390	22.838	115	0.000000
EB^R-EB^Hm	−3.726	2.333	−17.269	116	0.000000
EB^R-EB^0	−0.322	0.621	−5.584	115	0.000000
EB^Hm-EB^0	3.342	2.229	16.150	115	0.000000
oX^R-oX^Hm	−3.146	2.146	−15.652	113	0.000000
oX^R-oX^0	−0.260	0.443	−6.237	112	0.000000
oX^Hm-oX^0	2.831	2.137	14.265	115	0.000000
m,pX^R-m,pX^Hm	−13.076	9.638	−14.676	116	0.000000
m,pX^R-m,pX^0	−0.202	1.391	−1.564	115	0.120561
m,pX^Hm-m,pX^0	12.662	8.980	15.187	115	0.000000
S^R-S^Hm	−16.210	11.604	−15.109	116	0.000000
S^R-S^0	−0.4663	1.718	−2.923	115	0.004167
S^Hm-S^0	15.494	10.980	15.198	115	0.000000
Notes: sb = SD of the slope; B = benzene; T = toluene; EB = ethylbenzene; oX = o-xylene; m,pX = m,p-xylene; R = Radiello; Hm = homemade; 0 = Orsa 5; S = sum of xylenes.

CONCLUSION

The concentrations of VOC analytes in the atmosphere can vary widely in time as well as spatially. Therefore, proper equipment and analytical methodology for measuring pollution levels have to be designed to obtain reliable and valid data for the given environmental compartments. In recent years, passive sampling has gained broad application as a tool for air quality assessment in urbanized areas. This is a relatively simple and inexpensive technique for the sampling and enrichment of analytes collected from outdoor air. The samples of analytes obtained can serve as a source of information in reference to creating the precise distribution of pollutant concentrations in both time and space and determining the time-weighted average concentration of an analyte. It was confirmed that the passive technique can be used to isolate and enrich analytes to get long-term information on the concentration of an individual BTEX in atmospheric air in the Tricity agglomeration.

The results of investigations of passive sampler performance with regard to BTEX are reported. The analytical data obtained in northern Poland during 2008 (12 months and 10 sampling stations) were evaluated by the paired t test, Wilcoxon signed rank-sum test and Friedman ANOVA. All the statistical results confirm the hypothesis that the differences between the performances of the sampling procedures are highly significant. The homemade permeation sampler seems to be the most efficient concerning the amount retained, whereas the Radiello and Orsa 5 diffusive samplers seem to be similar.

The greater differences observed with the permeation sampler compared with the Orsa 5 and Radiello samplers may result from the fact that in the former the influence of temperature on calibration constant (k) and, hence, on sampling rate was not taken into account. The parameter responsible for the sampling rate in the gaseous phase was determined at a constant temperature (23 °C), and the influence of temperature on the k constant has been not examined in a very wide range of temperatures. For this reason, to extend the range of applicability of these samplers, new dosimeter calibration constants must be assessed. In addition, the greater variation between the passive samplers could be explained by differences in the sampling rates.

On the basis of the linear regression method, it can be concluded that the differences between the concentrations determined using passive samplers and the active sampling technique are statistically significant, in most cases at the probability level p = 95% and for f = n – 2 =46 df, but the data are statistically correlated. This research on the quality of the atmospheric air in the metropolitan Tricity area confirmed the good capability of passive samplers to show fluctuations in BTEX concentrations in atmospheric air, which makes them applicable for air monitoring on the local scale.

ACKNOWLEDGMENTS

The authors are grateful to the agency of Regional Air Quality Monitoring Foundation for access to the monitoring stations.