Subscriber access provided by CORNELL UNIVERSITY LIBRARY
Article
Passive air samplers as a tool for assessing long-term trends in atmospheric concentrations of semivolatile organic compounds Ji#í Kalina, Martin Scheringer, Jana Boruvkova, Petr Kukucka, Petra P#ibylová, Pernilla Bohlin-Nizzetto, and Jana Klánová Environ. Sci. Technol., Just Accepted Manuscript • Publication Date (Web): 23 May 2017 Downloaded from http://pubs.acs.org on May 25, 2017
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 31
Environmental Science & Technology
1
Passive air samplers as a tool for assessing long-
2
term trends in atmospheric concentrations of
3
semivolatile organic compounds
4
Jiří Kalina1, Martin Scheringer*1, 2, Jana Borůvková1, Petr Kukučka1, Petra Přibylová1, Pernilla
5
Bohlin-Nizzetto3, Jana Klánová1
6
1 Research Centre for Toxic Compounds in the Environment RECETOX, Kamenice 5, 625 00
7
Brno, Czech Republic
8
2 Institute for Chemical and Bioengineering, ETH Zürich, 8093 Zürich, Switzerland
9
[email protected] 10
3 Norwegian Institute for Air Research NILU, PO box 100, 2027 Kjeller, Norway
11
KEYWORDS: Persistent organic pollutant, POP, passive sampling, sampling rate, active
12
sampling, time series, temporal trend.
13
ACS Paragon Plus Environment
1
Environmental Science & Technology
Page 2 of 31
14
ABSTRACT: Many attempts have been made to quantify the relationship between the amount of
15
persistent organic pollutants sequestered by passive air sampling devices and their actual
16
concentrations in ambient air. However, this information may not be necessary for some
17
applications. In this study, two sets of 30 ten-year-long time series of simultaneous passive and
18
high-volume active air sampling carried out at the Košetice observatory in the Czech Republic
19
were used for a comparison of temporal trends. Fifteen polyaromatic hydrocarbons, seven
20
polychlorinated biphenyls and eight organochlorine pesticides were investigated. In most cases, a
21
good agreement was observed between the trends derived from passive and active monitoring
22
with the exception of several compounds obviously affected by sampling artifacts. Two sampling
23
artifacts were observed: breakthrough of high-volume sampler filters for penta- and
24
hexachlorobenzene and semi-quantitative values for PAHs with a high molecular weight. It has
25
been suggested before that annually aggregated results of passive air monitoring may be used
26
directly for the assessment of the long-term behavior of these compounds. The extensive set of
27
long-term data used in this study allowed us to confirm this finding and to demonstrate that it is
28
also possible to derive temporal trends and the compounds’ half-lives in air from the passive-
29
sampling time series.
ACS Paragon Plus Environment
2
Page 3 of 31
30
Environmental Science & Technology
TOC Art
31
32
33
ACS Paragon Plus Environment
3
Environmental Science & Technology
34
Page 4 of 31
Introduction
35
Passive air sampling is an increasingly common alternative to the conventional active air
36
sampling of semivolatile organic compounds (SVOCs), mainly persistent organic pollutants
37
(POPs) in ambient air. While active sampling relates sequestered amounts of analytes to the
38
measured volume of air in order to derive chemical concentrations in air, this volume is uncertain
39
for passive sampling. Therefore it is necessary to determine a sampling rate (i.e. a characteristic
40
volume of air that is stripped by the passive sampling medium per unit of time). The calculation
41
or estimation of compound-specific or generic sampling rates and determination of the most
42
important parameters affecting these rates have been a subject of numerous studies1–5. Such
43
estimations were based on calibration studies comparing the amount of analyte sequestered by a
44
passive air sampler (PAS) to its air concentration derived from an active air sampler (AAS)
45
employed in parallel6 or application of so called "performance reference" or "depuration"
46
compounds7. The PAS geometry, the type of sampling media, and the meteorological conditions
47
(temperature and wind speed) are factors affecting significantly the particle-gas and PAS-gas
48
partitioning of compounds of interest, the particle sampling efficiency of PAS, and thus the
49
sampling rates. Therefore reliable measurements of physicochemical properties of the analytes
50
(partitioning) or simultaneous measurements of meteorological conditions are required3,8,12.
51
However, even the best methods currently used to estimate sampling rates are associated with
52
substantial uncertainties (published coefficients of variation (standard deviation divided by
53
mean) reach values of up to 2)3,12–16. Several studies published in recent years have shown that
54
primary results of passive sampling may be used to identify time trends, which would help avoid
55
the calculation of the sampling rate17,18. However, these studies covered relatively few years of
56
sampling and did not quantify the time trends in terms of half-lives, nor did they confirm the
ACS Paragon Plus Environment
4
Page 5 of 31
Environmental Science & Technology
57
trends found in PAS-derived data by comparison with an independent sampling method such as
58
active air sampling.
59
In line with the studies by Shunthirasingham et al.17 and Gawor et al.18, we investigate whether
60
it is possible to skip the conversion of passive sampling results to air concentrations and to
61
compute reliable trend statistics from the primary data of the passive monitoring. Our aim is to
62
show that a sufficiently long time series of passive monitoring data makes it possible to reliably
63
quantify time trends of SVOCs because using annual average values eliminates the effect of
64
seasonal weather conditions on the partitioning of chemicals between air and passive sampling
65
media. We use time series of passive sampling data to estimate half-lives of the compounds in air
66
and evaluate them by comparison with half-lives estimated from active sampling results. Finally,
67
we discuss limitations of this approach by identifying chemicals for which sampling artifacts
68
occur.
69 70
Materials and methods
71
Sampling and Analysis. In order to compare passive and active monitoring trends, we used data
72
from a program that is part of the monitoring networks MONET (Monitoring Network of POPs
73
contamination in ambient air in Europe) for the passive sampling and of EMEP (European
74
Monitoring and Evaluation Programme) for the active sampling. Passive air sampling was
75
conducted from October 1, 2003, to December 25, 2013, which is more than 10 years of
76
continuous monitoring. The frequency of active sampling was every 7 days for 24 hours, which
77
corresponds to a total number of 534 samples, each of about 600 m3 of sampled air. The
78
frequency of passive sampling was every 28 days for the whole period of 28 days, which is
ACS Paragon Plus Environment
5
Environmental Science & Technology
Page 6 of 31
79
equivalent to 132 samples. Active sampling data represent a concentration in units of ng/m3 of
80
air, passive sampling data represent a mass of the analyte in units of ng/PUF/28 days.
81
In total 30 substances were analyzed, divided in three groups: organochlorine pesticides
82
(OCPs), polychlorinated biphenyls (PCBs), and polycyclic aromatic hydrocarbons (PAHs). A list
83
of the analytes is given in the SI. The number of concentration values is in total 16,020 for active
84
sampling and 3,960 for passive sampling. Several parts of this extensive dataset have been
85
published in recent works (all analytes 1996–2005 active)9, (all analytes 1996–2005 active)10,
86
(PAHs 1996–2011 active)11, (OCPs 2003–2012, PAHs 2006–2012 both active and passive)12,
87
(PAHs 2011–2014 active)19.
88
For high-volume active sampling, a device PS-1 Tisch Environmental TE1000 (air flow: 15
89
m3h-1) with a quartz fiber filter (QFF) (Whatmann, fraction > 2 µm) and polyurethane foam
90
(PUF) plug (6 cm diameter, 8 cm thickness) of type N 3038 (Gumotex, Břeclav) was used up to
91
31 December 2010, and from 1 January 2011 a device Digitel DPM10/30/00 Environ-sense was
92
used (air flow: 30 m3h-1) with a similar quartz fiber filter (QFF) (Whatmann, fraction > 2 µm)
93
and PUF plug (11 cm diameter, 10 cm thickness) of type N 3038 (Gumotex, Břeclav). For
94
passive sampling, PUF disks of type T 3037 (Molitan, Břeclav) were used (15 cm diameter,
95
1.5 cm thickness). The density of all PUFs was 0.0303 g·cm-3.
96
Both the PUF disks from passive sampling and PUF plugs and QFF filters from active
97
sampling were extracted with dichloromethane (Büchi System B-811 automatic extractor).
98
Surrogate recovery standards were spiked to filters before extraction, namely d8-naphthalene,
99
d10-phenanthrene, and d12-perylene for PAHs and PCBs 30 and 185 for PCBs/OCPs. Sample
100
extracts were concentrated under nitrogen and then one half was used for PAHs analysis and the
101
other half for PCBs/OCPs analysis.
ACS Paragon Plus Environment
6
Page 7 of 31
Environmental Science & Technology
102
A clean-up was conducted in the next step, using a sulfuric acid/silica gel column eluted with
103
30 mL of 1:1 hexane/dichloromethane prior to PCB/OCP analysis and a silica gel column (30 cm
104
length; 1 cm i.d.; 5 g silica) eluted with 10 mL of hexane (discarded), followed by 20 mL of
105
dichloromethane for PAH analysis.
106
Terphenyl was used as an internal standard for PAHs and PCB-121 as an internal standard for
107
PCBs/OCPs. Both passive and active samples were analyzed for PCBs and OCPs using a gas
108
chromatograph coupled with a mass spectrometer (GC-MS) (HP5975) with J&W Scientific
109
fused silica column (DB-5MS). Conversely, PAH were determined in all samples using a GC-
110
MS (HP 6890, HP 5972 and 5973) with J&W Scientific fused silica column (DB-5MS).
111 112
Data treatment. A part of the data is left-censored, i.e., includes values below the limit of
113
quantification, LoQ (semiquantitative values). The LoQ is calculated as 3.3 times the limit of
114
detection, LoD. The LoD, in turn, is determined for each compound as the concentration at the
115
lowest point on its calibration curve where the compound is still reliably detected20. The LoQ
116
was determined as 0.0025 ng/m3 for active and 0.5 ng/PUF/28 days for passive samples of PAHs
117
and 0.0005 ng/m3 for active and 0.1 ng/PUF/28 days for passive samples of OCPs and PCBs. On
118
average, for passive sampling 21% of the values were below the LoQ, and this proportion varies
119
from 0% for some PAHs with lower molecular weight up to 65% for indeno(1,2,3-c,d)pyrene.
120
There are also 9% of values below the LoQ in the active sampling data, ranging from 0% for
121
PAHs with lower molecular weight up to 48% for PCB-118. For the analysis the values below
122
the LoQ were replaced by half of the LoQ (compound specific).
123
Because the number of semiquantitative values is relatively low for most of the compounds, all
124
30 analytes were included in the statistical analysis. However, in the cases of indeno(1,2,3-
ACS Paragon Plus Environment
7
Environmental Science & Technology
Page 8 of 31
125
c,d)pyrene, benzo(g,h,i)perylene, benzo(a)pyrene, PCB-118, and β-HCH the portion of values
126
below LoQ exceeded 50%, which could affect the concentration averages and the resulting time
127
trends. In addition, there are several effects (sampling artifacts) that influence specifically the
128
passive or active sampling only, restricting their comparability. These limitations and their
129
effects on the results are discussed in detail in the Results and discussion section together with
130
other sampling artifacts.
131 132
Distributions of all primary passive and active data are presented as both time plots (Figure S1 and Figure S2) and box-and-whisker plots (Figure S3 and Figure S4) in the SI.
133 134
Annual aggregation. Given the strong seasonal changes of the concentration-time series due to
135
weather conditions and variable intensity of primary emission sources (higher concentrations of
136
PAHs in winter due to combustion) and revolatilization from secondary sources21–23 (higher
137
concentrations of OCPs in summer due to revolatilization and agricultural activity), statistical
138
evaluations that would be sensitive to seasonal fluctuations cannot be performed on the primary
139
data. Therefore we averaged the values over time within each year to achieve the same
140
granularity of both groups of data series with values not influenced by within-year fluctuations.
141
A standard method of a unification of monitoring time series with different granularity is
142
annual aggregation, usually by an arithmetic mean as suggested by UNEP24, which leads to the
143
same number of values in all time series and no seasonal fluctuations in individual years. The
144
annual aggregation makes it possible to avoid incomparable data caused by different rates of
145
particle infiltration of the active and passive samplers. Particle phase sampling efficiency is
146
about 100% for active sampling, but could be significantly lower for different types of passive
147
samplers25. Because the gas-particle partitioning of SVOCs is sensitive to temperature changes,
ACS Paragon Plus Environment
8
Page 9 of 31
Environmental Science & Technology
148
this leads to a variable dependence of the sampling rate on the temperature and wind velocity.
149
Partitioning between the gas and particle phases is governed by the octanol-air partition
150
coefficient, KOA12. If the partitioning changes because of changing environmental conditions,
151
active and passive samplers provide different results during one year. This plays a role especially
152
for compounds with log KOA between 8.8 and 11.3. Nevertheless, under the assumption of long-
153
term stable conditions and thus similar seasonal partitioning changes, the annually aggregated
154
values are similarly affected every year.
155
To obtain one representative aggregated value for each year and compound, it was necessary to
156
use data covering the whole year. Therefore, we excluded the data from 2003 from the analysis
157
since in this year the data are available from October-December only. The data of the remaining
158
ten years were aggregated to obtain 2 × 30 time series of length of 10 values, which were used
159
for the subsequent statistical processing.
160 161
Statistical tests applied. The robust Mann-Kendall test was used to check for presence of a
162
trend. This test indicates the statistical significance of a concentration decrease or increase
163
without determining the magnitude of the change. The test is designed specifically for time-
164
series analysis and does not presume any specific demands on the data. In addition, we also
165
tested the hypothesis of a normal distribution of the concentration values for all compounds by
166
the Kolmogorov-Smirnov test of normality; we obtained a negative results (p > 0.05), which
167
means that normality can neither be ruled out nor expected. Thus, for our 10-year long series, the
168
assumptions of the parametric Pearson test of trend are not violated. We therefore used also the
169
Pearson test and the Spearman test of trend and compared their results with the Mann-Kendall
170
test. Descriptions of the Pearson and Spearman tests and their results are provided in the
ACS Paragon Plus Environment
9
Environmental Science & Technology
Page 10 of 31
171
Supporting Information in Table S3. Box plots illustrating the distribution of the data points in
172
all time series are shown in Figures S3 and S4 in the Supporting Information.
173
The Mann-Kendall test26 makes it possible to identify the temporal trend as a correlation
174
(measured by the Kendall correlation coefficient, τ27) between time and the quantity of the
175
analyte (concentration or mass per PUF plug).
176
! = 2∙(!!!− !!!)/!∙(!−1)
177
where nCP is a number of concordant pairs of time and concentration values (for earlier time
178
points the concentration is lower than for later time points = increasing concentration) and nDP is
179
a number of discordant pairs (for earlier time point the concentration is higher than for later one
180
= decreasing concentration). The Kendall τ follows a distribution tabulated for small n and
181
approximated by a normal distribution for n > 30. Thus confidence intervals and p-values can be
182
computed.
183 184
Results and discussion
185
Trend identification
186
Correlation coefficients τ and the resulting p-values of the Mann-Kendall test, specifying their
187
statistical significance, are listed in Table 1.
188 189 190
Table 1: Trend identification by Mann-Kendall test for time series derived from passive and active sampling at Košetice. Results that are significant at a 95% confidence level are highlighted in green. Results from Mann-Kendall trend test chemical
τ, active
p, active p
τ, passive
p, passive
naphthalene
0.422
0.107
0.200
0.474
acenaphthylene
0.289
0.283
-0.022
1.000
ACS Paragon Plus Environment
10
Page 11 of 31
Environmental Science & Technology
acenaphthene
0.200
0.474
-0.111
0.721
fluorene
-0.111
0.721
-0.156
0.592
phenanthrene
-0.156
0.592
-0.244
0.371
anthracene
-0.333
0.210
-0.200
0.474
fluoranthene
0.022
1.000
0.067
0.858
pyrene
0.156
0.592
-0.022
1.000
benzo(a)anthracene
0.422
0.107
0.022
1.000
chrysene
0.156
0.592
0.244
0.371
benzo(b)fluoranthene
0.333
0.210
0.200
0.474
benzo(k)fluoranthene
0.156
0.592
-0.067
0.858
benzo(a)pyrene
0.378
0.152
0.022
1.000
indeno(123cd)pyrene
0.156
0.592
0.045
0.928
benzo(ghi)perylene
0.244
0.371
0.067
0.858
PCB 28
-0.556
0.032
-0.556
0.032
PCB 52
-0.600
0.020
-0.689
0.007
PCB 101
-0.422
0.107
-0.422
0.107
PCB 118
-0.689
0.007
-0.644
0.012
PCB 138
-0.556
0.032
-0.689
0.007
PCB 153
-0.600
0.020
-0.689
0.007
PCB 180
-0.556
0.032
-0.556
0.032
α-HCH
-0.422
0.107
-0.600
0.020
β-HCH
-0.289
0.283
-0.333
0.210
γ-HCH
-0.511
0.049
-0.778
0.002
p,p‘-DDE
-0.111
0.721
-0.244
0.371
p,p‘-DDD
-0.600
0.020
-0.600
0.02
p,p‘-DDT
-0.333
0.210
-0.289
0.283
PeCB
-0.200
0.474
-0.556
0.032
HCB
-0.067
0.858
-0.289
0.283
191
ACS Paragon Plus Environment
11
Environmental Science & Technology
Page 12 of 31
192
There is generally high agreement between active and passive sampling results: in all cases
193
where a significant trend was found for at least one variant of passive or active sampling, the
194
orientation of the trend is the same in both active and passive sampling (only decreasing trends
195
were found to be significant).
196
For 27% of the chemicals, the trends from both active and passive sampling are significant,
197
and in 67% of the cases both trends are insignificant. There are no cases where a trend identified
198
by active sampling is not recognized by the passive sampling and only two cases where a
199
significant trend was found in the passive-sampling data when it is not significant in the active-
200
sampling data. These cases are PeCB and α-HCH (and for HCB in the case of the Pearson test).
201
These discrepancies are most likely caused by sampling artifacts, as discussed below. The
202
results of the Pearson and Spearman tests listed in the Supporting Information are almost
203
identical with those from the Mann-Kendall test.
204 205
Linear trends, exponential trends and half-lives estimation
206
The simplest method of a trend quantification is a linear fit that specifies a constant slope of the
207
trend. Slopes were estimated separately for the time series of active and passive sampling by
208
means of least squares optimization. As in the trend identification above (Table 1), there were no
209
significant trends for any of the PAHs. For all other chemicals (except for α-HCH and HCB),
210
active and passive sampling gave the same results for significant or insignificant trends. α-HCH
211
and HCB showed significant trends in the passive sampling, but not in the active sampling data.
212
Detailed results including plots are provided in the Supporting Information as Table S5 and
213
Figure S6.
ACS Paragon Plus Environment
12
Page 13 of 31
Environmental Science & Technology
214
Next, we performed an exponential fit, which makes it possible to check for a first-order
215
chemical kinetics governing the concentrations and to estimate the half-lives of the compounds
216
in ambient air. The underlying differential equation describing the concentration-time trend is !" = −! ∙ ! !"
217
where c denotes the concentration, t time and k a rate constant, indicating a fraction of
218
pollutant loss per unit of time. This corresponds to a situation of negligible emissions together
219
with ongoing loss processes of the pollutants such as degradation, deposition and transport9,28–31.
220
The exponential trend fit was performed by least-squares optimization, providing results of
221
relative decrease, expressed as a half-life, t1/2, and its significance. The half-life is expressed as
222
follows: !!/! =
223
ln 2 !
Characteristics of the exponential trend models are listed in Table 2 and the results of the
224
exponential fits are plotted in Figure S6 in the Supporting Information.
225
Table 2: Results of exponential trend fitting. Green-colored cells indicate statistically significant
226
results (p < 0.05). For fits that are not significant, the confidence intervals include positive values
227
(range above upper bound) and negative values (range below lower bound).
pollutant
active
passive
half-life [years]
half-life [years]
(95% confidence interval)
p
(95% confidence interval)
p
naphthalene
-11.5 (87.8; -5.4)
0.076
-32.81 (12.7; -7.2)
0.539
acenaphthylene
-12.6 (16.6; -4.6)
0.226
64.56 (4.9; -5.8)
0.855
acenaphthene
-19.8 (32.3; -7.6)
0.191 -177.63 (16.2; -13.7)
0.852
ACS Paragon Plus Environment
13
Environmental Science & Technology
fluorene
Page 14 of 31
391.0 (12.2; -13.1)
0.942
70.94 (11.6; -17.3)
0.663
68.3 (9.9; -14.0)
0.706
65.29 (11.6; -17.9)
0.633
anthracene
28.36 (7.1; -14.3)
0.462
12.51 (4.6; -17.1)
0.219
fluoranthene
-25.7 (19.1; -7.7)
0.354
-33.22 (18.2; -8.7)
0.439
pyrene
-37.0 (13.4; -7.8)
0.558
-141.87 (10.0; -8.8)
0.883
benzo(a)anthracene
-10.5 (67.1; -4.9)
0.081
-23.75 (6.7; -4.3)
0.624
chrysene
-19.7 (16.9; -6.2)
0.319
-14.44 (22.5; -5.5)
0.198
benzo(b)fluoranthene
-12.6 (46.2; -5.6)
0.108
-18.91 (7.0; -4.0)
0.550
benzo(k)fluoranthene
-36.8 (13.6; -7.8)
0.551
140.07 (5.0; -5.4)
0.934
benzo(a)pyrene
-12.8 (46.0; -5.6)
0.109
67.78 (3.5; -3.9)
0.903
indeno(123cd)pyrene
-11.8 (28.6; -4.9)
0.141
-32.21 (2.5; -2.2)
0.871
benzo(ghi)perylene
-11.7 (29.3; -4.9)
0.138
31.27 (2.2; -2.6)
0.865
PCB 28
6.7 (3.8; 29.7)
0.018
8.0 (4.4; 42.8)
0.022
PCB 52
3.1 (2.1; 6.3)
0.002
3.1 (2.2; 5.1)
0.000
PCB 101
4.3 (2.2; 119.2)
0.044
3.8 (2.1; 21.8)
0.023
PCB 118
3.4 (2.4; 6.1)
0.001
3.9 (2.3; 13.3)
0.011
PCB 138
3.1 (1.9; 10.5)
0.011
2.6 (1.7; 5.7)
0.003
PCB 153
4.4 (2.5; 17.0)
0.014
3.6 (2.0; 14.1)
0.015
PCB 180
3.6 (2.1; 14.9)
0.016
3.4 (2.0; 9.8)
0.008
α-HCH
9.3 (4.4; -80.1)
0.073
5.7 (3.5; 16.9)
0.008
β-HCH
2.2 (1.1; 32.0)
0.038
5.2 (1.9; -6.9)
0.224
γ-HCH
3.0 (1.5; -139.8)
0.054
4.6 (3.3; 7.6)
0.001
p,p‘-DDE
67.5 (10.1; -14.3)
0.697
19.6 (6.2; -16.7)
0.319
p,p‘-DDD
3.9 (2.3; 12.3)
0.010
4.38 (2.9; 8.5)
0.001
p,p‘-DDT
5.8 (2.4; -14.9)
0.136
6.62 (2.5; -10.3)
0.198
22.4 (7.6; -23.9)
0.269
9.92 (5.17; 123.0)
0.036
phenanthrene
PeCB
ACS Paragon Plus Environment
14
Page 15 of 31
Environmental Science & Technology
HCB
12.8 (3.6; -8.9)
0.392
6.3 (2.9; -38.2)
0.083
228 229
As Table 2 shows, there is a similar pattern of correspondence of statistically significant trends
230
from active and passive sampling as in the trend identification (Table 1). There is also an
231
accordance of the estimated half-life values between the active and the passive sampling. Except
232
for γ-HCH, in all cases the passive-sampling half-lives are located in the 95% confidence
233
interval of the active-sampling estimates and vice versa. For 30% of the chemicals, both trends
234
are significant, for 60% both trends are insignificant, there is one case (3%; β-HCH) with a
235
significant trend from active sampling only, and there are two cases (7%; α-HCH, PeCB) with
236
significant trends from passive sampling only. However, even in these cases, the estimated half-
237
lives fall in the 95% confidence interval of their counterparts. Negative values of the half-lives
238
represent increasing trends, but these are in all cases statistically insignificant.
239
The concentrations of PCBs in ambient air approximately follow a first-order kinetics; this has
240
been observed at different sites in earlier work28–30 and is also visible from our results. Moreover,
241
for all PCBs assessed, the half-lives of PCBs in ambient air are very similar for the active and
242
passive trends, see Figure 1. Our estimated half-lives vary from 3.1 years for active sampling and
243
2.6 years for passive sampling (PCB 138) to 6.7 years for active sampling and 8.0 years for
244
passive sampling (PCB 28), which is in good agreement with similar results estimated at sites in
245
Germany28, UK29 and US30, with median half-lives (based on values from 10 sites) from 5.2 to
246
8.3 years (the shortes half-life was 2.3 years, the longest 330 years) for the same set of PCBs.
247
ACS Paragon Plus Environment
15
Environmental Science & Technology
Page 16 of 31
248 249
Figure 1: Ranges of half-lives of PCBs in ambient air determined in Košetice and at other
250
sites28,29. Crosshairs: individual half-lives from the literature; red squares: medians over the
251
literature-derived half-lives; dots: half-lives estimated in Košetice from passive (orange) and
252
active (green) sampling.
253 254
For the PAHs, the exponential fit yields virtually straight lines (Figure S6), which suggests that
255
the PAH concentrations do not follow first-order kinetics. PAH concentrations are, in addition to
256
loss processes, controlled also by ongoing emissions, as confirmed by other research works11,32.
257
For all assessed PAHs the trends were not significant (oscillating) and for half of them even
258
increasing, which is in accordance with the latest results from Central Europe34.
259
Half-life estimates for OCPs are shown in Figure 2. For substances where both trends are
260
significant (γ-HCH and DDD, see Table 2), the estimates from passive and active sampling are
261
close (orange and green dots). For substances with greater differences between the two estimates,
262
no (p,p’-DDE, HCB) or only one (β-HCH, PeCB) significant trend was observed. Finally, DDT
263
has no significant trend, but very similar half-life estimates, and α-HCH has one significant trend
264
(passive sampling) and the half-life estimates are relatively close. Half-life data from the
ACS Paragon Plus Environment
16
Page 17 of 31
Environmental Science & Technology
265
literature are mainly available for α- and γ-HCH35–38; for both HCHs, our estimates are within the
266
range spanned by the earlier data. Moreover, the difference between the passive and active
267
sampling estimates is less than the range spanned by all data.
268
269 270
Figure 2: Ranges of half-lives of OCP air concentrations in Košetice and other sites.35–38
271
Individual half-lives are depicted as crosshairs, medians over all values are depicted as red
272
squares and values estimated in Košetice are depicted as green dots for active and orange dots for
273
passive sampling.
274 275
Sampling artifacts
276
There are two groups of compounds that show systematic differences between the trends derived
277
from the active sampling and passive sampling time series. The first group includes the most
278
volatile compounds, which are mostly in the gas-phase due to their low octanol-air partition
279
coefficient (log KOA at 25 °C about 7 and lower): naphthalene, acenaphthylene, acenaphthene,
280
and to some extent possibly also pentachlorobenzene and hexachlorobenzene. For the trends of
ACS Paragon Plus Environment
17
Environmental Science & Technology
Page 18 of 31
281
the three light PAHs, there is typically a slower concentration decrease (or even an increase) in
282
the trends from the active sampling than in the trends from the passive sampling.
283
The second group includes three higher-molecular-weight PAHs with high PUF-air partition
284
coefficients, which are mostly bound to particles (log KOA at 25 °C about 9 and higher), namely
285
indeno(1,2,3-c,d)pyrene, benzo(a)pyrene and benzo(g,h,i)perylene. Due to the relatively low
286
particle infiltration efficiency of the deployed passive samplers25, the amount of these
287
compounds sequestered by the samplers during a 28-day period is relatively low, near to the
288
quantification limit (0.5 ng/PUF/28 days).
289
These two groups of compounds are related to two different sampling artifacts. The artifacts
290
affect either the active or the passive sampling time series and, thereby, make the time series
291
incomparable.
292
1. Semiquantitative values: a significant number of semiquantitative values (left-
293
censored values lower than the limit of quantification of the analytical method) and
294
their substitution by ½ of the limit can influence the annually aggregated values and the
295
resulting trend39. This was the case for passive sampling of indeno(1,2,3-c,d)pyrene,
296
benzo(a)pyrene and benzo(g,h,i)perylene, which are present almost entirely in the
297
particle phase and are, therefore, only partially sequestered by the passive sampler.
298
Average concentrations of these compounds were close to the quantification limit and
299
the number of semiquantitative results was too high to provide meaningful results (see
300
Table S7 in the SI). Regardless of the censoring, trends of these compounds were
301
insignificant for both the active and the passive sampling.
302
ACS Paragon Plus Environment
18
Page 19 of 31
Environmental Science & Technology
303
2. Breakthrough. This occurs for compounds for which the sorption capacity of the PUF
304
plug in the high volume active air sampler is not sufficient with respect to the total
305
volume of air passing through the sampler. Latest research40 shows that PAHs with low
306
molecular weight (2 and 3 rings), PeCB and HCB tend to show breakthrough even for
307
ca. 700 m3 of sampled air with a PUF plug of ca. 1,000 cm3. This is more probable
308
when higher concentrations are present (typically in summer for OCPs because of
309
higher temperatures and in winter for PAHs because of higher emissions).
310
The effect of the breakthrough is loss of material from the sample and, therefore, a
311
reduction of both the concentration raw data and the annually aggregated concentration
312
values.
313
Regarding breakthrough effects in our active-sampling data, the most important
314
incident was the change from a PS-1 sampler, deployed from 2004 to 2010, to a Digitel
315
sampler deployed from 2011 to 2013. The Digitel sampler deploys a PUF plug with a
316
volume of 950 cm3 (11 cm diameter, 10 cm thickness) compared to 226 cm3 (6 cm
317
diameter, 8 cm thickness) of the older PS-1 PUF plug made of the same material. The
318
airflow per unit area of the Digitel sampler is 1.7 times lower than for the PS-1 (see the
319
calculation in the SI). Together with the greater thickness of the Digitel plug (i.e.
320
higher number of theoretical chromatography plates), this results in reduced risk of
321
breakthrough40 for the Digitel and thus considerably fewer breakthrough events
322
between 2011 and 2013. Accordingly, the trend derived from active sampling is
323
affected by unrealistically low results caused by breakthrough in the initial phase
324
(2004–2010). This is visible for NAP, ACY and ACE in Figure 3; there is an apparent
325
effect of low sorption capacity in the initial period (2004–2010). Whereas the low
ACS Paragon Plus Environment
19
Environmental Science & Technology
Page 20 of 31
326
values remain similar from 2004 to 2013, the highest peaks are clearly lower in the
327
earlier period (2004–2010), see also Figures S1 and S2 in the SI.
328
For PeCB and HCB, the highest concentrations in the first three years are probably
329
affected by breakthrough; afterwards (2007–2013) the concentrations are generally
330
lower (as it is confirmed by the passive sampling results) and breakthrough is no longer
331
an issue.
332
Figure 3: Primary active sampling data for compounds tending to show breakthrough.
333
ACS Paragon Plus Environment
20
Page 21 of 31
Environmental Science & Technology
334
Breakthrough can be avoided by shorter periods of active air sampling, by using more efficient
335
sorbents (e.g. XAD) or by using PUF plugs of higher sorption capacity, i.e. plugs with larger
336
surface (as from 2011).
337 338
Environmental Significance
339
The effectiveness evaluation of the Stockholm Convention on POPs based on Article 16 has been
340
the main driving force for rapid development of passive air sampling techniques in the last
341
decade. In order to obtain reliable temporal trend data, the Global Monitoring Plan (GMP) was
342
established under the SC. While active air samplers could only provide accurate time-resolved
343
data from very few monitoring sites across the globe due to the high costs and logistic
344
constraints, passive samplers offer a simple and cheap alternative allowing for establishment of
345
sustainable air monitoring programs. Several global and regional large-scale efforts (GAPS,
346
MONET, LAPAN) stem from this advancing technology and generate data as required under the
347
GMP. However, it has been found that interpretation of the resulting data is not straightforward
348
as there are several ways how to translate amounts of chemicals sequestered by the passive
349
sampling devices into relevant air concentrations. Numerous studies introduced these calibration
350
methods revealing at the same time their uncertainties. In this paper, we demonstrate that for an
351
assessment of the long-term trends this step can be avoided and long-term trends can be derived
352
directly from concentrations of target chemicals in sampling media. This approach further
353
reduces the costs of the long-term programs by avoiding using labeled performance reference
354
compounds or parallel active samplers and reduces uncertainties of the resulting temporal trends.
355
The approach demonstrated in this study is applicable in all cases where the actual air
356
concentration of a pollutant is not an ultimate goal of the analysis, i.e. where the goal is
ACS Paragon Plus Environment
21
Environmental Science & Technology
Page 22 of 31
357
identification of long-term trends and possibly half-lives. Whereas high-volume active air
358
samplers may be prone to breakthrough during a 24-hour deployment, the linear uptake phase of
359
the PUF plugs in the Digitel samplers is not exceeded when the samplers are deployed for 28
360
days. This renders passive samplers a suitable tool for long-term assessment of lighter (gas-
361
phase) semivolatile organic compounds.
362 363 364
ASSOCIATED CONTENT The Supporting Information is available free of charge on the ACS Publications website at
365
DOI: XXX
366
Time plots of primary data from active (Figure S1) and passive (Figure S2) sampling in
367
Košetice, box-and-whisker plots of primary data from active (Figure S3) and passive (Figure S4)
368
sampling in Košetice, tables of annually aggregated values of ambient air concentrations of
369
target compounds derived from active (Table S1) and passive (Table S2) monitoring in Košetice,
370
description of Pearson and Spearman tests of trend and their results on annually aggregated data
371
(Table S3), description and numeric results of linear trends (Table S4 and Figure S5), similar
372
plots for exponential trend (Figure S6) and comparison of half-lives of the target compounds
373
between Košetice and several sites from other studies.
374
AUTHOR INFORMATION
375
Corresponding Author
376
*Email:
[email protected] Phone: +420 549496698
377
Notes
378
The authors declare no competing financial interest.
ACS Paragon Plus Environment
22
Page 23 of 31
Environmental Science & Technology
379
ACKNOWLEDGMENT
380
This work was supported by the RECETOX and ACTRIS research infrastructures (Ministry of
381
Education, Youth and Sports of the Czech Republic, LM2015051, LM2015037 and
382
CZ.02.1.01/0.0/0.0/16_013/0001761),
383
(CZ.02.1.01/0.0/0.0/15_003/0000469), and by a grant from Iceland, Liechtenstein and Norway
384
(EHP-CZ02-OV-1-029-2015).
385
References
the
CETOCOEN
PLUS
project
386
(1) Harner, T.; Bartkow, M.; Holoubek, O.; Klánová, J.; Wania, F.; Gioia, R.; Moeckel, C.;
387
Sweetman, A. J.; Jones, K. C.; Passive air sampling for persistent organic pollutants:
388
Introductory remarks to the special issue. Environmental Pollution. 2006, 144 (2), 361–364; DOI
389
10.1016/j.envpol.2005.12.044.
390
(2) Melymuk, L.; Bohlin, P.; Sáňka, O.; Pozo, K.; Klánová, J. Current Challenges in Air
391
Sampling of Semivolatile Organic Contaminants: Sampling Artifacts and Their Influence on
392
Data Comparability. Environmental Science & Technology. 2014, 48 (24), 14077–14091; DOI
393
10.1021/es502164r.
394
(3) Tuduri, L.; Harner, T.; Hung, H. Polyurethane foam (PUF) disks passive air samplers:
395
Wind effect on sampling rates. Environmental Pollution. 2006, 144 (2), 377–383; DOI
396
10.1016/j.envpol.2005.12.047.
397
(4) Herkert, J. N.; Martinez, A., Hornbuckle, K. C. A Model Using Local Weather Data to
398
Determine the Effective Sampling Volume for PCB Congeners Collected on Passive Air
399
Samplers.
400
10.1021/acs.est.6b00319.
Environmental
Science
&
Technology.
2016,
ACS Paragon Plus Environment
50,
6690–6697;
DOI
23
Environmental Science & Technology
Page 24 of 31
401
(5) Petrich, N. T.; Spak, S. N.; Carmichael, G. R.; Hu, D.; Martinez, A.; Hornbuckle, K. C.
402
Simulating and explaining passive air sampling rates for semi-volatile compounds on
403
polyurethane foam passive samplers. Environmental Science & Technology. 2013, 47 (15),
404
8591–8598; DOI 10.1021/es401532q.
405
(6) Shoeib, M.; Harner, T. Characterization and Comparison of Three Passive Air Samplers for
406
Persistent Organic Pollutants. Environmental Science & Technology. 2002. 36(19), 4142–4151;
407
DOI 10.1021/es020635t.
408
(7) Moeckel, C.; Harner, T.; Nizzetto, L.; Strandberg, B.; Lindroth, A.; Jones, K. C. Use of
409
Depuration Compounds in Passive Air Samplers: Results from Active Sampling-Supported Field
410
Deployment, Potential Uses, and Recommendations. Environmental Science & Technology.
411
2009. 43(9), 3227–3232; DOI 10.1021/es802897x.
412
(8) Melymuk, L.; Robson, M.; Helm, P. A.; Diamond, M. L. Evaluation of passive air sampler
413
calibrations: Selection of sampling rates and implications for the measurement of persistent
414
organic pollutants in air. Atmospheric Environment. 2011. 45 (10), 1867–1875; DOI
415
10.1016/j.atmosenv.2011.01.011.
416
(9) Holoubek, I.; Klánová, J.; Jarkovský, J.; Kohoutek, J. Trends in background levels of
417
persistent organic pollutants at Košetice observatory, Czech Republic. Part I. Ambient air and
418
wet deposition. Journal of Environmental Monitoring. 2007. 9(6), 557–563; DOI
419
10.1039/b700750g.
420
(10) Dvorská, A.; Lammel, G.; Klánová, J.; Holoubek, I. Košetice, Czech Republic – ten years
421
of air pollution monitoring and four years of evaluating the origin of persistent organic
422
pollutants. Environmental Pollution. 2008. 156(2), 403–408; DOI 10.1016/j.envpol.2008.01.034.
ACS Paragon Plus Environment
24
Page 25 of 31
Environmental Science & Technology
423
(11) Liu, L.-Y.; Kukučka, P.; Venier, M.; Salamova, A.; Klánová, J.; Hites, R. A.; Differences
424
in spatiotemporal variations of atmospheric PAH levels between North America and Europe:
425
Data from two air monitoring projects. Environment International. 2014. 64, 48–55; DOI
426
10.1016/j.envint.2013.11.008.
427
(12) Holt, E.; Borůvková, J.; Kalina, J.; Melymuk, L.: Bohlin, P.; Klánová, J. Using long-term
428
air monitoring of semi-volatile organic compounds to evaluate the uncertainty in polyurethane-
429
disk passive sampler-derived air concentrations. Environmental Pollution. 2017. 220 part B,
430
1100–1111; DOI 10.1016/j.envpol.2016.11.030.
431
(13) Wania, F.; Shen, L.; Ying D. L.; Teixeira, C.; Muir, D. C. G. Development and
432
Calibration of a Resin-Based Passive Sampling System for Monitoring Persistent Organic
433
Pollutants in the Atmosphere. Environmental Science & Technology. 2003, 37 (7), 1352–1359;
434
DOI 10.1021/es026166c.
435
(14) Pozo, K.; Harner, T.; Wania, F.; Muir, D. C. G.; Jones, K. C.; Barrie, L. A. Toward a
436
Global Network for Persistent Organic Pollutants in Air:
Results from the GAPS Study.
437
Environmental Science & Technology. 2006, 40 (16), 4867–4873; DOI 10.1021/es060447t.
438
(15) Gouin, T.; Wania, F.; Ruepert, C.; Castillo, L. E. Field Testing Passive Air Samplers for
439
Current Use Pesticides in a Tropical Environment. Environmental Science & Technology. 2008,
440
42 (17), 6625–6630; DOI 10.1021/es8008425.
441
(16) Bohlin, P.; Audy, O.; Škrdlíková, L.; Kukučka, P.; Přibylová, P.; Prokeš, R.; Vojta, Š.;
442
Klánová, J. Outdoor passive air monitoring of semi volatile organic compounds (SVOCs): a
443
critical evaluation of performance and limitations of polyurethane foam (PUF) disks.
444
Environmental Science: Processes & Impacts. 2014, 16, 433–444; DOI 10.1039/C3EM00644A.
ACS Paragon Plus Environment
25
Environmental Science & Technology
Page 26 of 31
445
(17) Gawor, A.; Shunthirasingham, C.; Hayward, S. J.; Lei, Y. D.; Gouin, T.; Mmereki, B. T.;
446
Masamba, W.; Ruepert, C.; Castillo, L. E.; Shoeib, M.; Lee, S. C.; Harner, T.; Wania, F. Neutral
447
polyfluoroalkyl substances in the global Atmosphere. Environmental Sciences: Processes and
448
Impacts. 2014, 16 (3), 404–413; DOI 10.1039/C3EM00499F.
449
(18) Shunthirasingham, C.; Oyiliagu, C. E.; Cao, X.; Gouin, T.; Wania, F., Lee, S.-C.; Pozo,
450
K.; Harner, T., Muir, D. C. G. Spatial and temporal pattern of pesticides in the global
451
atmosphere. Journal of Environmental Monitoring. 2010, 12 (9), 1650–1657; DOI
452
10.1039/C0EM00134A.
453
(19) Shapoury, P.; Lammel, G.; Šmejkalová, A. H.; Klánová, J.; Přibylová, P.; Váňa, M.
454
Polycyclic aromatic hydrocarbons, polychlorinated biphenyls, and polychlorinated pesticides in
455
background air in central Europe – investigating parameters affecting wet scavenging of
456
polycyclic aromatic hydrocarbons. Atmospheric Chemistry and Physics. 2015, 15, 1795–1805;
457
DOI 10.5194/acp-15-1795-2015.
458
(20) Wenzl, T.; Haedrich, J.; Schaechtele, A.; Robouch, P.; Stroka, J. Guidance Document on
459
the Estimation of LOD and LOQ for Measurements in the Field of Contaminants in Feed and
460
Food. European Union Reference Laboratories for Polycyclic Aromatic Hydrocarbons, Dioxins
461
and PCBs in Feed and Food, Heavy Metals in Feed and Food and Mycotoxins, 2016.
462
(21) Sweetman, A. J.; Valle, M. D.; Prevedouros, K.; Jones, K. C. The role of soil organic
463
carbon in the global cycling of persistent organic pollutants (POPs): interpreting and modelling
464
field data. Chemosphere. 2004, 60 (7), 959–972; DOI 10.1016/j.chemosphere.2004.12.074.
465
(22) Komprda, J.; Komprdová, K.; Sáňka, M.; Možný, M.; Nizzetto, L. Influence of Climate
466
and Land Use Change on Spatially Resolved Volatilization of Persistent Organic Pollutants
ACS Paragon Plus Environment
26
Page 27 of 31
Environmental Science & Technology
467
(POPs) from Background Soils. Environmental Science & Technology. 2013, 47 (13), 7052–
468
7059; DOI 10.1021/es3048784.
469
(23) Harner, T.; Bidleman, T. F.; Jantunen, L. M. M.; Mackay, D. Soil—air exchange model of
470
persistent pesticides in the United States cotton belt. Environmental Toxicology and Chemistry.
471
2001, 20 (7), 1612–1621; DOI 10.1002/etc.5620200728.
472
(24) Draft revised guidance on the global monitoring plan for persistent organic pollutant;
473
Conference of the Parties to the Stockholm Convention on Persistent Organic Pollutants. United
474
Nations
475
meetingdocs/inf14/GMP%20Guidance%20CD/Guidance.pdf.
Environment
Program,
2011;
http://www.pops.int/documents/meetings/cop_3/-
476
(25) Markovic, M. Z.; Prokop, S.; Staebler, R., M.; Liggio, J.; Harner, T. Evaluation of the
477
particle infiltration efficiency of three passive samplers and the PS-1 active air sampler.
478
Atmospheric Environment. 2015, 112, 289–293; DOI 10.1016/j.atmosenv.2015.04.051.
479
(26) Cortes, D. R., Hites, R. A. Detection of Statistically Significant Trends in Atmospheric
480
Concentrations of Semivolatile Compounds. Environmental Science & Technology. 2000. 34,
481
2826–2829; DOI 10.1021/es990466l.
482 483
(27) Kendall, M. A New Measure of Rank Correlation. Biometrika. 1938. 30 (1–2), 81–89; DOI 10.1093/biomet/30.1-2.81.
484
(28) Bruckmann, P.; Hiester, E.; Klees, M.; Zetzsch, C. Trends of PCDD/F and PCB
485
concentrations and depositions in ambient air in Northwestern Germany. Chemosphere. 2013,
486
93 (8), 1471–1478; DOI 10.1016/j.chemosphere.2013.07.029.
ACS Paragon Plus Environment
27
Environmental Science & Technology
Page 28 of 31
487
(29) Schuster, J. K.; Gioia, R.; Sweetman, A. J.; Jones K. C. Temporal Trends and Controlling
488
Factors for Polychlorinated Biphenyls in the UK Atmosphere (1991-2008). Environmental
489
Science & Technology. 2010, 44 (21), 8068–8074; DOI 10.1021/es102134d.
490
(30) Ping, S.; Basu, I.; Hites, R. A. Temporal Trends of Polychlorinated Biphenyls in
491
Precipitation and Air at Chicago. Environmental Science & Technology. 2006, 40 (4), 1178–
492
1183. DOI: 10.1021/es051725b.
493
(31) Holt, E.; Kočan, A.; Klánová, J.; Assefa A.; Wiberg, K. Spatiotemporal patterns and
494
potential sources of polychlorinated biphenyl (PCB) contamination in Scots pine (Pinus
495
sylvestris) needles from Europe. Environmental Science and Pollution Research. 2016, 23 (19)
496
19602–19612; DOI 10.1007/s11356-016-7171-6.
497
(32) Tobiszewski, M.; Namieśnik, J. PAH diagnostic ratios for the identification of pollution
498
emission
sources.
Environmental
499
10.1016/j.envpol.2011.10.025.
Pollution.
2012,
162,
110–119;
DOI
500
(33) Kamens, R. M.; Guo, Z.; Fulcher, J. N.; Bell, D. A. The influence of humidity, sunlight,
501
and temperature on the daytime decay of polyaromatic hydrocarbons on atmospheric soot
502
particles.
503
10.1021/es00166a012.
Environmental
Science
&
Technology.
1988,
22
(1),
103–108;
DOI
504
(34) Garrido, A.; Jiménez-Guerrero, P.; Ratola, N. Levels, trends and health concerns of
505
atmospheric PAHs in Europe. Atmospheric Environment. 2014, 99, 474–484; DOI
506
10.1016/j.atmosenv.2014.10.011.
ACS Paragon Plus Environment
28
Page 29 of 31
Environmental Science & Technology
507
(35) Liu, L.-Y.; Salamova, A.; Hites R. A. Interstudy and Intrastudy Temporal Trends of
508
Polychlorinated Biphenyl, Pesticide, and Polycyclic Aromatic Hydrocarbon Concentrations in
509
Air and Precipitation at a Rural Site in Ontario. Environmental Science & Technology. 2014,
510
1 (4), 226–230; DOI 10.1021/ez5000572.
511 512
(36) Choi M.-K.; Chun M.-Y. Half lives of Gaseous Organochlorine Pesticides in Atmosphere. Journal of Environmental Toxicology. 2007, 22 (2), 177–184.
513
(37) Hung, H.; Halsall, C. J.; Blanchard, P.; Li, H. H.; Fellin, P.; Stern, G.; Rosenberg, B.
514
Temporal Trends of Organochlorine Pesticides in the Canadian Arctic Atmosphere.
515
Environmental Science & Technology. 2002, 36 (5), 862–868; DOI 10.1021/es011204y.
516
(38) Buehler, S. S.; Basu, I.; Hites, R. A. Causes of Variability in Pesticide and PCB
517
Concentrations in Air near the Great Lakes. Environmental Science & Technology. 2004, 38 (2),
518
414–422; DOI 10.1021/es034699v.
519
(39) Antweiler, R. C. Evaluation of Statistical Treatments of Left-Censored Environmental
520
Data Using Coincident Uncensored Data Sets. II. Group Comparisons. Environmental Science &
521
Technology. 2015, 49, 3439−13446; DOI 10.1021/acs.est.5b02385.
522
(40) Melymuk, L.; Bohlin-Nizzetto, P.; Prokeš, R.; Kukučka, P.; Klánová, J. Sampling artifacts
523
in active air sampling of semivolatile organic contaminants: Comparing theoretical and measured
524
artifacts and evaluating implications for monitoring networks. Environmental Pollution. 2015;
525
DOI 10.1016/j.envpol.2015.12.015.
526 527
(41) Pearson, K. Notes on regression and inheritance in the case of two parents. Proceedings of the Royal Society of London. 1895, 58, 240–242.
ACS Paragon Plus Environment
29
Environmental Science & Technology
Page 30 of 31
528
(42) Pankow, J. F. Overview of the gas phase retention volume behavior of organic compounds
529
on polyurethane foam. Atmospheric Environment. 1989, 23 (5), 1107–1111. DOI 10.1016/0004-
530
6981(89)90311-9.
531
(43) Rahman, N. A. A Course in Theoretical Statistics. 1968. ISBN 978-0852640685.
532
(44) Choi, S. C. Tests of Equality of Dependent Correlation Coefficients. Biometrika. 1977,
533
64(3), 645–647. DOI: 10.1093/biomet/64.3.645.
ACS Paragon Plus Environment
30
Page 31 of 31
Environmental Science & Technology
78x44mm (300 x 300 DPI)
ACS Paragon Plus Environment