Environ. Sci. Technol. 2009, 43, 5383–5390
Field Performance of Seven Passive Sampling Devices for Monitoring of Hydrophobic Substances
require further work. Finally, the usefulness of passive samplergenerated contaminant concentrations is demonstrated through the comparison with institutional monitoring and with European Water Framework Directive Environmental Quality Standards (EQS).
I A N J . A L L A N , * ,†,‡ K E E S B O O I J , § ALBRECHT PASCHKE,| BRANISLAV VRANA,⊥ GRAHAM A. MILLS,# AND RICHARD GREENWOOD‡ Norwegian Institute for Water Research, Gaustadalle´en 21, NO-0349 Oslo, Norway, School of Biological Sciences, University of Portsmouth, King Henry I Street, Portsmouth, PO1 2DY, U.K., Royal Netherlands Institute for Sea Research, PO Box 59, 1790 AB Texel, The Netherlands, Department of Ecological Chemistry, UFZ Helmholtz Centre for Environmental Research, Permoserstrasse 15, 04318 Leipzig, Germany, National Water Reference Laboratory for Slovakia, Water Research Institute, Nabr. arm. gen. L. Svobodu 5, 81249 Bratislava, Slovakia, and School of Pharmacy and Biomedical Sciences, University of Portsmouth, White Swan Road, Portsmouth, PO1 2DT, U.K.
Introduction
Received February 27, 2009. Revised manuscript received May 22, 2009. Accepted May 26, 2009.
The performance of seven passive sampling devices for the monitoring of dissolved concentrations of polycyclic aromatic hydrocarbons (PAHs), polychlorinated biphenyls (PCBs), hexachlorobenzene, and p,p′-DDE was evaluated through simultaneous field exposures of 7-28 days in the River Meuse (The Netherlands). Data from the Chemcatcher, low density polyethylene membranes, two versions of the membrane-enclosed sorptive coating (MESCO) sampler, silicone rods, silicone strips and semipermeable membrane devices (SPMD) was assessed through rate of dissipation of performance reference compounds (PRCs), mass of analyte absorbed by the samplers and time-weighted average concentration (CTWA) data. Consistent PRC data throughout the range of samplers tested here confirmed the transition from membrane- to boundary layer-controlled exchange at log KOW 4.5-5.0. The comparison of sampler surface area-normalized masses absorbed for analytes under boundary layer-control showed some variability between samplers that can be attributed to the conformation and deployment of the various samplers and to the uncertainty associated with the analysis conducted in different laboratories. Despite different modes of calculation, relatively consistent CTWA were obtained for the different samplers. The observed variability is likely to be due to the uncertainty of sampler-water partition coefficients and the extrapolation of analyte uptake rates at the high log KOW range (under boundary layer-controlled exchange) from a narrow PRC data range, and these issues * Corresponding author phone: +47 22 18 5100; fax: +47 22 18 52 00; e-mail:
[email protected]. † Norwegian Institute for Water Research. ‡ School of Biological Sciences, University of Portsmouth. § Royal Netherlands Institute for Sea Research. | UFZ Helmholtz Centre for Environmental Research. ⊥ Water Research Institute. # School of Pharmacy and Biomedical Sciences, University of Portsmouth. 10.1021/es900608w CCC: $40.75
Published on Web 06/16/2009
2009 American Chemical Society
Many nonpolar organic substances such as polycyclic aromatic hydrocarbons (PAHs) and polychlorinated biphenyls (PCBs) may cause adverse effects in aquatic environments (1). Since these hydrophobic contaminants readily sorb to bottom sediments, concentrations in surface waters are generally in the low ng L-1 to pg L-1 range. Consequently, the regulatory monitoring and risk assessment of hydrophobic contaminants in surface waters is generally hampered by the inability to measure reliably these low (and sometimes fluctuating) concentrations (2). Under the Water Framework Directive (WFD) currently in force across the European Union, environmental quality standards (EQS) are defined for a set of hydrophobic priority substances that include, for example, PAHs, organochlorine pesticides and brominated flame retardants (2). Bottle sampling for the measurement of contaminant concentrations in water can become particularly challenging depending on sample pretreatment such as filtration or storage, the extraction technique used and levels of suspended solids or dissolved organic matter present in the water. The reporting of values below poor limits of detection (LOD) is unlikely to support current legislation. Since the introduction of passive sampling two decades ago, the focus has increasingly been on the determination of time-weighted average concentrations (CTWA) of hydrophobic contaminants dissolved in water (3-5). Contaminant accumulation into passive sampling devices is a diffusive process resulting from the difference in chemical activity of the contaminant dissolved in water and that in the sampler. These integrative samplers are generally composed of a receiving phase for contaminant accumulation and a membrane to limit mass transfer. Mass transfer itself depends on the characteristics of the contaminant of interest such as the size of the molecule, its affinity for the membrane/receiving phase material, and transport across phases, namely the membrane layer, the diffusive boundary layer and any biofilm layer developing at the surface of the sampler during extended exposures. Calibration experiments are generally conducted in the laboratory to determine contaminant uptake rates (RS) by exposing samplers under constant conditions of contaminant concentration, water temperature and turbulences at the surface of the sampler. Since the application of laboratory-determined RS to field situations is unreliable, the dissipation of performance reference compounds (PRCs), non-naturally occurring chemicals spiked into the sampler prior to deployment, allows RS calibration in situ (6, 7). This is only possible when there is an isotropic exchange of chemicals between the sampler and water. The intercomparison of sampling procedures in the wider context of quality control schemes is often overlooked. Nonetheless, such intercomparisons when applied to passive sampling can address the reproducibility of sampler preparation, extraction and analysis, field deployment procedures by different teams and the accuracy and precision of CTWA. While a range of passive samplers at various stages of their development are available and have been the subject of much testing separately, they have seldom been evaluated alongside each other. Here, the simultaneous deployment of seven passive sampling devices was undertaken in the River Meuse VOL. 43, NO. 14, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5383
5384
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 14, 2009
321 8.02 ACE-d10 PHE-d10 FLUO-d10 CHRY-d12 CB004 CB155 CB204 324 1.77 ACE-d10 PHE-d10 FLUO-d10 CHRY-d12 CB004 CB155 CB204 Triolein 460 4.95 ACE-d10 FLUE-d10 PHE-d10 CHRY-d12g B[a]A- d12 receiving phase material surface area (cm2) sampler volume, VS (cm3) performance reference compoundse,f
a See the Materials and Methods section for full names of samplers. b GC-MS at the University of Portsmouth. c GC-MS for PAHs and GC-ECD for PCBs at NIOZ. d Thermal desorption-GC-MS at UFZ, Leipzig. e PRC for which elimination rates could be used for further data interpretation are in italic. f BIP: biphenyl; ACE: acenaphthene; FLUE: fluorene; PHE: phenanthrene; ANT: anthracene; PYR: pyrene; FLUO: fluoranthene; CHRY: chrysene; B[a]A: benzo[a]anthracene. g PRC elimination rate significant only for the 28 day exposure.
0.64 0.031 PHE-d10 ANT-d10 FLUO-d10 PYR-d10 B[a]A-d12 silicone 1 0.047 PHE-d10 ANT-d10 FLUO-d10 PYR-d10 B[a]A-d12
silicone LDPE (100 µm)
cellulose acetate silicone 0.33 0.016 LDPE (40 µm) LDPE (70 µm) membrane material
Chemcatcher, SPMD
C18 Empore disk/octanol 17 0.6 BIP-d10 ACE-d10 FLUE-d10 PHE-d10 PYR-d10 B[a]A-d12
LDPE
silicone (500 µm)
MESCO I (m)d silicone stripc
passive sampling devicesa
c b b
TABLE 1. Characteristics of the Passive Sampling Devices Tested
Field Site. Passive samplers were deployed between 12th April and 10th May 2005 at a monitoring station on the River Meuse (The Netherlands) situated downstream of the border with Belgium (50°46′46.1 “N; 5°41′58.9”E). The Meuse river water is classified as hard (CaCO3 ∼ 250 mg L-1) and during this field study, values of temperature, pH, and total/dissolved organic carbon (TOC/DOC) were in the range 14-17 °C, 7.7-7.9, 4-7 mg L-1, and 2-4 mg L-1, respectively (more detailed data is provided in the Supporting Information (SI)). Passive Sampling Devices. The seven passive samplers tested included semipermeable membrane devices (SPMDs), a version of the Chemcatcher designed for sampling hydrophobic compounds, low-density polyethylene membranes (LDPEs), silicone strips, two types of membraneenclosed sorptive coating samplers (MESCOs) and silicone rods (7-11). While both SPMDs and the Chemcatcher use an LDPE membrane, their receiving phases are a thin film of triolein and a C18 Empore disk loaded with 1-octanol, respectively. LDPE membrane, silicone strip, and silicone rod samplers are single-phase samplers. The original MESCO (referred to as MESCO I (m) with m for modified) uses a cellulose acetate dialysis membrane filled with water, however, it differs from the original design as a result of the replacement of the Gerstel Twister by a silicone rod (10). MESCO II is based on an LDPE envelope around a silicone rod with an additional air layer separating the two phases (11). Characteristics of the samplers are given in Table 1. Chemcatcher, SPMD, LDPE membrane, silicone strip and rod, and MESCO II devices were all spiked with PRCs with log KOW values in the range 3.9-7.3 to allow the estimation of contaminant exchange kinetics between water and the sampler (see Table 1 for PRCs used for each sampler). Sampler Preparation, Processing and Analysis for PAHs and PCBs. Chemcatcher devices with a Teflon sampler body (University of Portsmouth, UK) were prepared, extracted and analyzed for PAHs by gas chromatography-mass spectrometry (GC-MS) as described previously (9). Standard size SPMDs (92 cm long; 2.5 cm wide) were purchased from Exposmeter AB (Tavelsjo, Sweden) and extracted following published procedures (12). Briefly, SPMDs were dialyzed (2 × 24 h in n-hexane) and the triolein removed from the extract through a size-exclusion chromatographic column with dichloromethane as mobile phase. Finally, the solvent was exchanged to n-hexane and extracts reduced and analyzed by GC-MS for PAHs and PCBs (8). LDPE membranes (64.4 cm long; 2.5 cm wide) were prepared from lay-flat tubing purchased from Brentwood Plastics Inc. (St Louis, MO), preextracted with n-pentane overnight before spiking with a series of PRCs by incubating them in a PRC methanol-water solution (80/20 v/v) (8). Silicone strips were made from 0.5 mm thick sheets (Rubber BV, Hilversum, The Netherlands) and were of a similar dimension to LDPE membranes. These were precleaned by Soxhlet extraction with ethyl acetate (16 h) and methanol (2 h), and a similar procedure to that used for LDPE membranes was employed to spike PRCs into silicone strips. Following exposure both types of sampler were wiped with a damp paper tissue to remove biofilms and then extracted in 100 mL n-pentane (once for LDPE membranes and twice for silicone strips). Extracts were reduced, cleaned-up with silica (2 g, deactivated with 6%
MESCO IId
Materials and Methods
LDPE (112 µm)
silicone rodd
(The Netherlands) for exposures of 7-28 days to evaluate (i) the legitimacy of PRC-based in situ RS calibration across the range of samplers, (ii) the effects of factors such as the method used to calculate concentrations, and (iii) the influence of the exposure time on CTWA generated by the various samplers. The extraction and analysis of passive samplers for PAHs and PCBs in three different laboratories provides an additional dimension to this study.
water; elution with n-pentane) and analyzed by GC-MS for PAHs. An electron capture detector was used for the detection and quantification of PCBs, hexachlorobenzene and p,p′DDE. MESCO I (m) was prepared by inserting a precleaned silicone rod (1 cm long; 2 mm diameter, Goodfellow Ltd., UK) into a dialysis membrane bag (18 mm flat width and 30 mm long) made from regenerated cellulose (Spectra/Por 6, molecular weight cutoff 600 Da) filled with Milli-Q water (10). Diffusion-limiting envelopes of MESCO II were composed of air-filled nonporous LDPE membrane (purchased from Polymer-Synthesewerk, Rheinberg, Germany) containing a 1.5 cm long silicone rod (from Goodfellow GmbH, Bad Nauheim, Germany) of 2 mm diameter as receiving phase spiked with PRCs (11). The bare silicone rods used were 8 cm long and 2 mm diameter. Following sampler retrieval, silicone rods were removed from the MESCO membranes and stored in glass vials at -20 °C until analysis. Bare silicone rods were quickly washed under tap water and dried with tissue paper before storage at -20 °C. The combined processing and analysis of silicone rods (1.5 cm long from the MESCOs and 1 cm pieces cut from silicone rods) consisted of a thermal desorption step followed by GC-MS analysis. A thermo-desorption unit (TDU) from Gerstel (Mu ¨ lheim a.R., Germany) was placed on top of an Agilent 6890 GC (Agilent Technologies, Palo Alto, CA) equipped with a cold injection system CIS-4 (Gerstel) and a mass spectrometric detector (MSD) 5973N (Agilent). Full details of the analysis can be found elsewhere (11). In all cases, quality assurance procedures such as the use of internal standards for the extraction and analytical steps and the assessment of analyte recoveries were conducted. Sampler Deployment and Retrieval. All prepared samplers were stored at -20 °C and the temperature maintained below 0-4 °C during transport to and from the field site. Preparation and trip control samplers were prepared and transported in a similar way to exposed samplers and opened to the air during deployment and retrieval procedures. During deployment, controls were stored in closed containers at -20 °C. Samplers were mounted onto stainless steel cages, and moorings kept them 1 m below the surface of the water. In most cases, triplicate passive sampling devices of each type were exposed for a period of 7 days, two consecutive 14 day periods (14 days (1) and 14 days (2), respectively), and an overlapping 28 day exposure (further details on replication in SI). In addition, silicone strips, were deployed for four consecutive 7 day exposures. The 7 day sampling period for Chemcatcher, SPMDs and LDPE membranes was not undertaken in cages and samplers were therefore exposed to higher water turbulences.
Results and Discussion Adequacy of the Performance Reference Compound Approach. The measurement of PRC dissipation provides information on contaminant exchange kinetics between water and the sampler and allows the estimation of RS values in situ (6). Analytes for which the concentration in the sampler approaches equilibrium with the concentration in the water are characterized by significant or even complete elimination of PRC with similar log KOW. However, negligible or little PRC dissipation is indicative of rates in the linear phase of uptake. The threshold between these two regimes is generally found for PRCs with log KOW of 4.5-5 for exposure periods of several weeks (13, 14). In addition, using multiple PRCs with a range of log KOW makes it possible to establish when kinetics of uptake into the sampler are membrane- or boundary layercontrolled. The overall resistance to mass transfer (1/kO) into the samplers can be expressed as the sum of the water (δW/ DW) and membrane-side (δM/KMWDM) resistances:
δW δM 1 ) + kO DW KMWDM
(1)
with KMW the membrane-water partition coefficient, δW and δM the boundary and membrane layer thicknesses (m), and DW and DM (m2 s-1) analyte diffusion coefficients in water and the membrane, respectively. Amounts of analytes absorbed by the samplers follow a first-order approach to equilibrium: N ) KSWVCTWA[1 - exp(-ket)]
(2)
where N is the amount of analyte absorbed (ng), KSW the sampler-water partition coefficient (L L-1), V the volume of the sampler (L), ke the exchange rate constant (h-1), t the exposure time (h), and CTWA is in ng L-1. PRC dissipation also follows first-order kinetics: NPRC ) N0,PRC exp(-ket)
(3)
where NO,PRC and NPRC are PRC masses in the samplers prior to and following exposure, respectively and where ke is given by ke )
kOA RS ) KSWV KSWV
(4)
where kO is the overall mass transfer coefficient (see eq 1), A the surface area of the sampler (m2), V the volume of the sampler (L) and RS the analyte uptake rates (L d-1). PRC elimination rates, ke, were calculated for the various exposures and samplers and their statistical significance tested using a procedure described previously (15). Overall, it was possible to use most PRC data; however, data were not used when release was either close to 100% or insignificant, or when amounts remaining in trip controls were significantly lower than in fabrication controls (see SI for further details). Since configurations of the devices differ widely (Table 1) and ke is proportional to A/V (eq 4), elimination rates were normalized to this ratio. The relationship between keV/A values for 14 and 28 day exposures and log KOW is presented in Figure 1A. The spread of the data across the range of samplers is less than one log unit and the apparent plateau for PRCs with log KOW < 5 is indicative of membranecontrolled mass transfer (13). The overlap of Chemcatcher and SPMD (both using LDPE membrane material) data and generally higher keV/A values for silicone strips and MESCO II for PRCs with log KOW < 5 reflects higher diffusion coefficients in the silicone material compared with LDPE (16). Overall mass transfer coefficients (kO) determined as the product of keV/A and KSW (eq 4), were plotted as a function of log KOW (Figure 1B). KSW for nondeuterated PRC analogues were used (see following section for a detailed list of references). The transition between membrane-controlled mass transfer, where kO increases with increasing PRC hydrophobicity, to boundary layer-control becomes more apparent with the bell-shaped relationship between log kO and log KOW (Figure 1B). Under boundary layer-controlled mass transfer, RS is expected to decrease with increasing hydrophobicity. Here, a decrease can be observed for silicone strips (phenanthrene-d10 and fluoranthene-d10) and LDPE membranes (fluoranthene-d10 and chrysene-d12). The transition between membrane- and water-sidecontrol of mass transfer appears to occur for compounds with log KOW between 4.5 and 5.0 (Figure 1B) and confirms previously observed cutoff points (13, 14). One would expect similar kO values for fluoranthene-d10 under boundary layercontrolled exchange for LDPE membranes and silicone strips since both types of samplers have a similar configuration VOL. 43, NO. 14, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5385
if the reduction in mass transfer coefficients was solely the result of a decrease in analyte diffusion coefficients in water with increasing molecular weight (14). These slopes are, however, similar to those from log kO-log MW regressions obtained in sediment slurries (17). This sharp decrease in kO values has been observed previously during sampler calibration experiments (13, 15) and attributed to (i) the transfer in the membrane of the Chemcatcher or SPMD becoming ratelimiting again owing to an increasing difficulty for larger molecules to diffuse in the LDPE, or (ii) contaminant sorption to DOC that reduces the fraction available to the samplers and results in the underestimation of RS values of large molecular weight PAHs for example. Since the present data is based on PRC elimination rather than analyte uptake, log kO-log MW relationships suggest that a significant reduction in analyte diffusion coefficients in the membrane materials tested here is possible and contributes the strong decrease in uptake rates for compounds with log KOW > 5. Calculation of TWA Concentrations. Concentrations of dissolved contaminants in the Meuse river water were calculated using the following equation (combination of equations 2 and 4): CTWA )
[
N
(
KSWV 1 - exp -
FIGURE 1. (A) First-order performance reference compound elimination rates, ke, normalized to the sampler surface to volume ratio (A/V ) for five different passive samplers. (B) Mass transfer coefficients, kO, as the product of keV/A and sampler-water partition coefficients, KSW. Data are for 14 (1st and 2nd successive exposures) and 28 day exposures. Lines are intended as a guide to the eye only. and were disposed randomly in the same cages during exposure. Accounting for an uncertainty of log KSW of around 0.3 log units, differences observed here are not likely to be significant (16). PRC mass transfer coefficients were regressed (using Minitab version 14) against log KOW for those under membrane layer-controlled kinetics and molecular weight (MW) for those under boundary layer-limited exchange (Table 2). The observation of similar slopes for log kO versus log KOW regressions for Chemcatcher and SPMDs is not unexpected since both samplers use an LDPE membrane. The steeper slopes observed for both samplers for the 7 day exposure under higher water turbulences indicate that resistance to mass transfer in the boundary layer is of a similar order of magnitude to that in the membrane for analytes with log KOW near 4.5. According to eq 1, kO is influenced by both KSW and the analyte diffusion coefficient DM for the membrane material when mass transfer is membrane-controlled. With slopes of log KSW-log KOW relationships close to unity, observed log kO-log KOW slopes of 0.7-0.95 as shown in Table 2 are plausible. Slopes for silicone strips, however, are significantly lower. It is likely that resistance to mass transfer in the boundary layer is not negligible and contributes to the overall resistance to mass transfer of PRCs with higher log KOW used in these regressions. This is important since with an accurate knowledge of KSW and DM values, estimates of kW (DW/δW) may be obtained from PRCs under “membrane-controlled uptake”. PRC-based information on boundary layer-controlled uptake is available only for LDPE membranes and silicone strips (Table 2). A linear regression of log kO on log MW gave values in the range -3 to -8.9 which are over an order of magnitude higher than the slope of -0.35 predicted 5386
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 14, 2009
RSt KSWV
)]
(5)
Further details of the calculation of CTWA are available in the SI. Literature values for KSW for each sampler are needed and exposure-specific RS have to be determined. KSW values for the Chemcatcher, LDPE membranes, MESCO I (m), MESCO II, silicone strips, silicone rods and SPMDs were obtained from refs 15, 14, 18-21 (using the experimental design as described in ref 22), and 13, respectively. RS for the PRCs were calculated from RS ) ke,PRC KSW V. Since PRC-based RS are for a limited log KOW range, models relating RS to analyte properties were used to estimate RS for compounds outside the PRC range. A full description of the calculation of RS values is provided in the SI. Briefly, sampling rates of the PRCs were fitted to the empirical log RS-log KOW relationships reported for the Chemcatcher (15) and SPMDs (13). For all other samplers, these relationships are not available and sampler-specific methods were used. For silicone strips, the PRC-based linear relationship between ke,PRC and KSW-1 was used to extrapolate the RS value for analytes with log KOW < 4.6. For those above this threshold, boundary layer-controlled uptake was assumed and RS-PRC for fluoranthene-d10 was used to extrapolate uptake rates for the remaining compounds according to RS ∼ (Vm)-0.39 where Vm is the analyte molar volume at boiling point (13). For LDPE membranes, the empirical KOW-RS model developed by Booij and co-workers (14) based on SPMD/LDPE membrane experimental calibration data was used to estimate RS for all analytes (see SI). Offloading of fluoranthene-d10 and chrysene-d12 and literature data were used to estimate two empirical parameters BW and Bm representative of mass transfer in the boundary and membrane layers, respectively. The product of the mass transfer coefficient obtained and the surface area of the sampler is RS. For MESCO I (m), no PRC data was available. Instead mean values of laboratorybased RS were used. These were corrected according to eq 4 to account for the use of a different receiving phase (with different V and KSW) (10, 23). For MESCO II, the overall mass transfer coefficients were calculated from the sum of theoretical mass transfer coefficients for the various layers of the sampler as previously undertaken (19). Waterside mass transfer was adjusted using available PRC data. Finally, analyte RS for silicone rods were also estimated from semiempirical mass transfer coefficients calculated
TABLE 2. Slopes of Linear Regressions of Log kO on Log KOW and Log kO on Log MW for Each of the Samplers and Exposure Period of 7, 14, and 28 Days membrane-controlled uptake ((∆log kO)/(∆log KOW)), (SE)a boundary layer-controlled uptake ((∆log kO)/(∆log MW)), (SE)a exposure (days) Chemcatcher SPMD LDPE silicone strip
7 0.95b (0.07) 0.93b (0.29)
14 (1) 0.91 (0.08) 0.71 (0.17)
14 (2) 0.79 (0.06) 0.78 (0.05)
28 0.85 (0.10) 0.78 (0.17)
-c
0.18 (0.04)
0.39 (0.05)
0.47 (0.03)
7
14 (1)
14 (2)
28
-8.5 (2.6) -c
-5.6 (2.0) -8.8 (1.6)
-8.9 (1.8) -8.5 (2.4)
-4.8 (0.5) -3.1 (1.2)
a SE ) standard error of the slope. b Deployment outside the cage resulting in higher mass transfer for PRC with log KOW ∼ 5. c Insufficient replication available.
for the membrane and boundary layer according to eq 1. A 10 µm boundary layer thickness based on ke for fluoranthene-d10 was adopted. Interestingly, this value is similar to that obtained for MESCO II. In order to evaluate the performance of the various samplers, we compared (i) masses of analytes absorbed (normalized to the respective sampler surface areas) for all analytes that were in the linear phase of uptake, (ii) calculated CTWA, and (iii) the precision of these CTWA estimates. To compare surface area-normalized amounts of analytes, we first calculated the average amounts for each analyte and each sampler for the 7 day exposure. This was repeated for the 14 and 28 day exposures. These values were then divided by the corresponding values obtained for the LDPE membrane samplers. LDPE membrane samplers were selected based on the fact that the largest number of analytes was detected with this sampler. The size of data sets used to create the box-plots (Figure 2A) is indicative both of the number of analytes in the linear phase of uptake for the various samplers and of the relative method quantification limits (MQLs) of the various methods. These show that MQLs generally increase in the order LDPE membrane ∼ silicone strip ∼ SPMD < silicone rod ∼ MESCO II < MESCO I (m) ∼ Chemcatcher. Generally samplers with large surface areas such as LDPE membranes, silicone strips and SPMDs enabled the quantification of all target compounds. The very similar mean analyte masses accumulated in silicone strips and in LDPE membranes result from the analysis being conducted in the same laboratory and the samplers having almost identical sizes and similar mounting in deployment cages. The uncertainty in the normalized mean ratio for SPMDs combines that associated with the analysis being conducted in a different laboratory with those due to differences in turbulences around the samplers resulting from their larger dimensions. Similar factors influence the data obtained for the other samplers. Some variability can be observed for these samplers though the significantly smaller size of data sets is likely to affect these results. The particularly small data set for MESCO I (m) is the result of membrane rupture in exposures of over 14 days. To compare CTWA values, we first calculated the geometric mean of CTWA for each compound and each exposure taken over all seven samplers. Ratios of individual CTWA estimates over the geometric mean were then calculated. Dissolved contaminant concentrations varied over 3 orders of magnitude with low molecular weight PAHs at the ng L-1 level down to PCBs found at concentrations of tens of pg L-1. Marked differences in CTWA generated by the various samplers can be observed in Figure 2B. CTWA estimated by LDPE membranes, MESCO II and SPMDs are closest to respective mean concentrations. Concentrations measured by the Chemcatcher appear generally higher than mean concentrations. This could be explained by a reduction in uptake rates (as shown by PRC elimination rates) with increasing exposure time. Data obtained with MESCO I (m) and the silicone rods consistently under predict mean concentrations and
appear much lower than those generated by the Chemcatcher, LDPE membrane or silicone strips. This could be the result of possible bias induced by the method used to calculate TWA concentrations from analyte masses accumulated or uncertainty in the PRC data for the silicone rods. Since the uptake of many of the analytes detected and quantified by these two samplers had reached a significant degree of equilibrium, most of the variability in CTWA may be linked to the variability of KSW values (16). An uncertainty (or bias) of 0.3 log units is not impossible and would result in error equivalent to a factor of 2 when calculating CTWA for analytes close to equilibrium. It should be noted here that Figure 2B reflects the variability among samplers and among laboratories. Finally, to compare the precision of CTWA values, CTWA for each analyte and each sampler were log-transformed before calculating standard deviations. The antilog of these standard deviations can be interpreted as an uncertainty factor and provides a comparison of the overall precision of the different passive sampling methods used here (Figure 2C). The observed variability for all samplers was in the range 1.2-1.5. The smallest variability is generally exhibited by the Chemcatcher and LDPE membranes. The precision of analytical measurements decreases with decreasing analyte concentration. Concentrations of these analytes in passive sampler extracts are closest to analytical LODs where analytical precision is worst. In contrast with the less hydrophobic PAHs (close to equilibrium), the calculation of CTWA for analytes in the linear phase of uptake relies significantly more on PRC elimination rates. Therefore, the precision of CTWA for these compounds cumulates errors from more sources since it includes differences in the physical preparation of the samplers, in masses accumulated by the samplers, in the PRC elimination rates and finally in the extraction and analytical measurements (generally close to analytical LODs) conducted in two different laboratories. Interestingly, here the spread of the LDPE membrane data is much lower than for silicone strips and SPMDs. Effect of Sampler Exposure Time. Sampler exposure time has an impact not only on PRC dissipation but also on masses of analyte accumulated. Longer deployments generally result in the accumulation of higher masses of contaminants that are in the linear phase of uptake (i.e., far from equilibrium) facilitating their analytical measurement while bringing sampler concentrations of less hydrophobic ones closer to equilibrium with the water phase. However, membrane fouling by biofilm-forming microorganisms or accumulation of suspended matter on the sampler surface may affect the exchange of analytes and PRCs between water and passive sampler when samplers are exposed long enough for these phenomena to occur. Exposures of 14 and 28 days resulted in significant biofouling comprising a large proportion of sediment particles. This may be due to a combination of relatively small openings on the cages used for deployment and the “zigzag” mounting of samplers within the cages that facilitates sediment particles settlement inside the cage. VOL. 43, NO. 14, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5387
FIGURE 3. Ratio of analyte masses accumulated over 28 days to the sum of masses accumulated during the two successive 14 day exposures. Reference lines at y ) 1 and 0.5 indicate ideally compounds for which uptake is linear over the 28 days and those that have reached equilibrium, respectively.
FIGURE 2. (A) Box-plots of sampler surface area-normalized amounts absorbed for analytes under boundary layer controlled uptake (N/A). These were normalized with respect to those for LDPE membrane samplers ((N/A)LDPE membrane) and calculated for each analyte and each exposure. (B) Ratios of time-weighted average concentrations (CTWA) measured by the different samplers to the geometric mean concentration (from all sampler replicates) for each analyte and exposure time. (C) Box-plot of standard deviations of log-transformed samplerspecific CTWA calculated for each analyte and exposure. Values on the box-plots represent the sample size on which the box-plot is based. Dots are 5/95 percentiles. Significantly less fouling of the samplers was observed for the 7 day exposure outside the cages. Our approach here was to compare masses of analytes accumulated and this was possible for the 28 day or consecutive 14 day deployments since sampler-specific exposure conditions were identical. Figure 4 shows the ratio of the mass of contaminant accumulated over 28 days to the sum of masses accumulated over the two successive 14 day exposures. In the case of compounds in the linear phase of uptake (generally with log KOW > 5) during these 28 days, a 5388
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 14, 2009
ratio of one would be expected. However, for those close to reaching equilibrium, a ratio of 0.5 should be obtained if the dissolved analyte concentration in the water phase did not change noticeably during the field test. For analytes with log KOW > 5, ratios are in the range 0.5-1.0 with most values in the range 0.6-0.95 (Figure 3). Since PRC data demonstrated that most of these compounds were in the linear phase of uptake during the 28 days, ratios of one should be observed. The lower values seen here may result from increasing fouling of the samplers over time during exposure. When considering analytes that have neared equilibrium, most ratios are well below 0.5. While this could be explained by radical changes in dissolved analyte concentrations during exposures, contaminant masses accumulated during four successive 7-day exposures of silicone strips did not demonstrate such changes in concentration (data not shown). Masses of contaminants with log KOW < 5.2 accumulated in all samplers appear to decrease with increasing exposure time and increasing membrane fouling, possibly as a result of degradation of the less hydrophobic PAHs. However, (photo-) degradation of analytes sorbed onto the receiving phase of the samplers is unlikely. Concerns can be raised when estimating CTWA of the more mobile and degradable compounds when heavy membrane fouling is observed during long passive sampler deployment. Additional work is required to understand such a process and to estimate its overall importance in the determination of CTWA. For the more hydrophobic contaminants, generally linear uptake was observed and exposure time/heavy fouling induced only minor changes in estimates of CTWA. While minimal effects of biofouling have previously been observed (24), changes in uptake rates during exposure can be compensated since biofouling is expected to affect PRC release in a similar manner to analyte uptake (13). Here, only the PRC elimination data for the Chemcatcher showed a reduction in uptake rates when increasing sampler exposure from 14 to 28 days. Regulatory Use of Passive Sampling Data. Clear objectives and readily available methods with adequate limits of detection, precision and accuracy are required for regulatory monitoring. Water quality monitoring of hydrophobic organic contaminants as defined in the European WFD is based on the comparison of samples with “whole water” EQS. Since passive sampling measures the truly dissolved fraction of contaminants in water data generated by this method cannot be compared directly
with currently set WFD EQS, even though the fraction sampled is more toxicologically relevant. Nevertheless, comparisons with “whole water” EQS values are possible after a further data manipulation to account for sorption to DOC and suspended particulate matter data (see SI). DOC-water (KDOC) and OC-water (KOC), partition coefficients (25, 26) may be used to calculate “whole water” concentrations from passive sampler-based CTWA. Despite the high uncertainty of KOC and KDOC, the use of conservative values will result in an overestimation of “whole water” concentrations. If these are still well below EQS, compliance may be demonstrated. Passive sampling-based whole water concentrations were compared with those obtained using bottle sampling collected during the field trial and with monthly institutional monitoring data for the period 2002-2005 (Table S5). Additionally, “whole water” concentrations were estimated from data obtained from monitoring of suspended particulate matter and of the fraction of organic carbon for the same 2002-2005 period. Bottle sampling was characterized by many measurements below limits of detection (LODs) that varied by a factor of 2-7. When comparing concentrations measured by bottle sampling with EQS values (Table S5), it is important to take account of limits of quantification, particularly for larger molecular weight PAHs (e.g., for benzo[ghi]perylene) since for a method to be considered fit-for-purpose these values should not exceed one-third of the EQS. Mean whole water concentrations of benzo[ghi]perylene and indeno[1,2,3-cd]pyrene estimated from passive sampling are very close to proposed WFD annual average EQS. Most mean concentrations estimated from suspended particulate matter monitoring for 2002-2005 were variable and close to or above EQS (2). Passive samplers generally provide data that is less variable than that from “whole water” sampling since the latter may be strongly influenced by levels of suspended particulate matter. This lower variability is an attractive characteristic in the monitoring of water quality and the detection of temporal trends in concentrations. The present study showed that the CTWA estimated by the different samplers varied by a factor of 2 on average while short-term within-sampler variability was a factor of 1.3. Efforts should focus on quantifying the long-term within-sampler variability and understanding and reducing the variability between different types of samplers. LODs of passive samplers with large surface area are likely to be well below typical concentrations encountered across Europe for analytes with log KOW < 7.5, and this enables their use for monitoring tasks such as comparison with EQS or the monitoring of trends (4, 27). For other samplers such as Chemcatcher and MESCO, screening for larger molecular weights PAHs can be undertaken with “field” LODs in a similar range to EQS levels. Investigative monitoring tasks or monitoring at sensitive sites or where elevated concentrations are expected (e.g., sewage/stormwater effluents) are therefore most appropriate applications for these devices.
Acknowledgments We thank Nel Frijns and the RIZA monitoring team at Eijsden (The Netherlands), Uwe Schro¨ter in Leipzig for the analysis of MESCO/silicone rod samplers and guidance with the sizeexclusion chromatography, and Ronald van Bommel for the extraction and analysis of silicone strips and LDPE membranes at NIOZ. We acknowledge financial support from the European Union’s Sixth Framework Programme (Contract SSPI-CT-2003-502492; http://www.swift-wfd.com). Views presented here are those of the authors alone.
Supporting Information Available Additional details on water quality, sampler replication data, lists of chemicals analyzed and detected by the various
samplers, and details of the calculation of time-weighted average concentrations.This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) Warren, N.; Allan, I. J.; Carter, J. E.; House, W. A.; Parker, A. Pesticides and other micro-organic contaminants in freshwater sedimentary environmentssA review. Appl. Geochem. 2003, 18 (2), 159–194. (2) Lepom, P.; Brown, B.; Hanke, G.; Loos, R.; Quevauviller, P.; Wollgast, J. Needs for reliable analytical methods for monitoring chemical pollutants in surface water under the European Water Framework Directive. J. Chromatogr., A 2009, 1216, 302–315. (3) Vrana, B.; Mills, G. A.; Allan, I. J.; Dominiak, E.; Svensson, K.; Knutsson, J.; Morrison, G.; Greenwood, R. Passive sampling techniques for monitoring pollutants in water. TrAC, Trends Anal. Chem. 2005, 24 (10), 845–868. (4) Allan, I. J.; Vrana, B.; Greenwood, R.; Mills, G. A.; Roig, B.; Gonzalez, C. A “toolbox” for biological and chemical monitoring requirements for the European Union’s Water Framework Directive. Talanta 2006, 69 (2), 302–322. (5) Kot-Wasik, A.; Zabiegala, B.; Urbanowicz, M.; Dominiak, E.; Wasik, A.; Namiesnik, J. Advances in passive sampling in environmental studies. Anal. Chim. Acta 2007, 602 (2), 141– 163. (6) Booij, K.; Sleiderink, H. M.; Smedes, F. Calibrating the uptake kinetics of semipermeable membrane devices using exposure standards. Environ. Toxicol. Chem. 1998, 17 (7), 1236–1245. (7) Huckins, J. N.; Petty, J. D.; Lebo, J. A.; Almeida, F. V.; Booij, K.; Alvarez, D. A.; Clark, R. C.; Mogensen, B. B. Development of the permeability/performance reference compound approach for in situ calibration of semipermeable membrane devices. Environ. Sci. Technol. 2002, 36 (1), 85–91. (8) Booij, K.; Smedes, F.; van Weerlee, E. M. Spiking of performance reference compounds in low density polyethylene and silicone passive water samplers. Chemosphere 2002, 46 (8), 1157–1161. (9) Vrana, B.; Mills, G. A.; Dominiak, E.; Greenwood, R. Calibration of the Chemcatcher passive sampler for the monitoring of priority organic pollutants in water. Environ. Pollut. 2006, 142 (2), 333–343. (10) Vrana, B.; Popp, P.; Paschke, A.; Schuurmann, G. Membraneenclosed sorptive coating. An integrative passive sampler for monitoring organic contaminants in water. Anal. Chem. 2001, 73 (21), 5191–5200. (11) Paschke, A.; Schwab, K.; Brummer, J.; Schuurmann, G.; Paschke, H.; Popp, P. Rapid semi-continuous calibration and field test of membrane-enclosed silicone collector as passive water sampler. J. Chromatogr., A 2006, 1124 (1-2), 187–195. (12) Vrana, B.; Schuurmann, G. Calibrating the uptake kinetics of semipermeable membrane devices in water: Impact of hydrodynamics. Environ. Sci. Technol. 2002, 36 (2), 290–296. (13) Huckins, J. N.; Petty, J. D.; Booij, K., Monitors of organic chemicals in the environment: Semipermeable membrane devices. Springer: New York, 2006. (14) Booij, K.; Hofmans, H. E.; Fischer, C. V.; Van Weerlee, E. M. Temperature-dependent uptake rates of nonpolar organic compounds by semipermeable membrane devices and lowdensity polyethylene membranes. Environ. Sci. Technol. 2003, 37 (2), 361–366. (15) Vrana, B.; Mills, G. A.; Kotterman, M.; Leonards, P.; Booij, K.; Greenwood, R. Modelling and field application of the Chemcatcher passive sampler calibration data for the monitoring of hydrophobic organic pollutants in water. Environ. Pollut. 2007, 145 (3), 895–904. (16) Rusina, T. P.; Smedes, F.; Klanova, J.; Booij, K.; Holoubek, I. Polymer selection for passive sampling: A comparison of critical properties. Chemosphere 2007, 68 (7), 1344–1351. (17) Booij, K.; Hoedemaker, J. R.; Bakker, J. F. Dissolved PCBs, PAHs, and HCB in pore waters and overlying waters of contaminated harbor sediments. Environ. Sci. Technol. 2003, 37 (18), 4213– 4220. (18) Paschke, A.; Popp, R. Solid-phase microextraction fibre-water distribution constants of more hydrophobic organic compounds and their correlations with octanol-water partition coefficients. J. Chromatogr., A 2003, 999 (1-2), 35–42. (19) Wennrich, L.; Vrana, B.; Popp, P.; Lorenz, W. Development of an integrative passive sampler for the monitoring of organic water pollutants. J. Environ. Monit. 2003, 5 (5), 813–822. VOL. 43, NO. 14, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5389
(20) Yates, K.; Davies, I.; Webster, L.; Pollard, P.; Lawton, L.; Moffat, C. Passive sampling: partition coefficients for a silicone rubber reference phase. J. Environ. Monit. 2007, 9 (10), 11161121. (21) Paschke, A.; Hanke, K.; Schu ¨u ¨ rmann, G. in preparation. (22) Paschke, A.; Brummer, J.; Schuurmann, G. Silicone rod extraction of pharmaceuticals from water. Anal. Bioanal. Chem. 2007, 387 (4), 1417–1421. (23) Vrana, B.; Paschke, A.; Popp, P. Calibration and field performance of membrane-enclosed sorptive coating for integrative passive sampling of persistent organic pollutants in water. Environ. Pollut. 2006, 144 (1), 296–307. (24) Booij, K.; van Bommel, R.; Mets, A.; Dekker, R. Little effect of excessive biofouling on the uptake of organic contaminants by
5390
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 14, 2009
semipermeable membrane devices. Chemosphere 2006, 65 (11), 2485–2492. (25) Karickhoff, S. W. Semiempirical estimation of sorption of hydrophobic pollutants on natural sediments and soils. Chemosphere 1981, 10 (8), 833–846. (26) Burkhard, L. P. Estimating dissolved organic carbon partition coefficients for nonionic organic chemicals. Environ. Sci. Technol. 2000, 34 (22), 4663–4668. (27) Allan, I. J.; Mills, G. A.; Vrana, B.; Knutsson, J.; Holmberg, A.; Guigues, N.; Laschi, S.; Fouillac, A. M.; Greenwood, R. Strategic monitoring for the European Water Framework Directive. TrAC, Trends Anal. Chem. 2006, 25 (7), 704–715.
ES900608W
