EVALUATION OF A NOVEL PASSIVE SAMPLER FOR POLY- AND PERFLUOROALKYL SUBSTANCES IN AQUATIC ENVIRONMENTS

Polyand perfluoroalkyl substances (PFASs) are of growing concern worldwide, due to their ubiquitous presence and adverse health effects in humans and the environment. Surface waters in the northeastern United States in particular have displayed elevated concentrations of PFASs. Passive sampling devices are excellent monitoring tools, that accumulate contaminant loadings through passive diffusion and adsorption to the sampler, and provide a long-term, time-weighted average of the contaminant over large temporal and spatial scales. Here we utilize a novel integrative passive sampler—a microporous polyethylene (PE) tube filled with Hydrophilic-Lipophilic-Balanced sorbent—to gain a better understanding of its function, utility, and uptake rates in field environments. Three sampling campaigns were conducted in the fall of 2017 and summer 2018, deploying a total of seventy-two PE tube passive samplers across nine sites in a well-mixed estuary and in two wastewater treatment plant effluents for one month’s duration. Twenty-four PFASs (including carboxylic acids, sulfonates, and precursors) were measured across all sites in the passive samplers, as well as complementary water samples, using Ultra Performance Liquid Chromatography/Mass Spectrometry. In the estuary, the PE tube samplers accumulated a sum PFASs of 2 to 15 ng sampler, and in the waste water treatment plant effluent 60 to 70 ng sampler. In situ sampling rates, which are essential when needed to calculate the contaminant concentrations in water, were characterized using a first order kinetic model, yielding sampling rates of 10-50 mL day. Results from this study imply that these passive samplers can be successfully used to determine dissolved concentrations of PFASs in surface waters, though the sampling rate seems to vary with external water flow velocity.


INTRODUCTION
This thesis is presented in manuscript format, and is currently in preparation for submission to the journal Environmental Science and Technology.
PFASs enter the environment primarily through waste water treatment plant (WWTP) effluent, septic system leaking and discharge, and aqueous film forming foams (AFFF) used by airports, military bases and fire training areas (Möller et al. 2010, Moody & Field 2000, Schaider et al. 2016. One of the biggest environmental sources of these compounds are through manufacturing discharge. An estimated 2,610-21,400 tons of perfluoroalkyl carboxylic acids were produced globally from 1951-2015, and 65%-98% of this was discharged into surface waters (Wang et al. 2014). These compounds then leach into the ground water, aquifers and enter drinking water supplies. The northeastern United States, in particular, has shown to have elevated levels of PFAS in its surface and groundwaters (Zhang et al. 2016;Weber et al. 2017).
Passive sampling tools have been successfully used for decades to easily and time-effectively monitor different groups of contaminants in the environment. These sampling tools work through the passive diffusion of analyte molecules from the sampled medium to a collecting medium due to a difference in chemical potentials (Górecki & Namieśnik 2002). Passive samplers are inexpensive, small, easy to use, and can be deployed over large spatial and temporal scales, without the need of daily environmental sampling. Passive samplers accumulate compounds over the deployment period and generate a reliable time weighted average of those compounds in water/air (Vrana et al. 2005), which is more representative of the environmental variation of contaminant concentration over time, as compared to an active grab sample which only represents a single moment in time. Due to the accumulation and concentration of the compounds onto the sampler overtime, contaminant concentrations can often be detected in the sampler that would otherwise be undetectable in a grab sample at lower concentrations.
Another benefit of passive samplers is that they can mimic the portion of contaminants bioavailable to organisms and help predict biomonitoring results (Vrana et al. 2005).
These sampling devices are especially useful when studying remote areas, large spatial scales or long-term monitoring projects, because they vastly decrease the amount of field and lab work needed to determine environmental contaminant concentrations.
Traditional passive samplers, such as polyethylene sheets or polyurethane foam, are excellent monitoring tools for nonpolar contaminants, such as dioxins, polychlorinated biphenyls (PCBs), and polycyclic aromatic hydrocarbons (PAHs) (Lohmann et al. 2012).
For these traditional samplers, Performance Reference Compounds (PRCs) are added prior to sampler deployment, to determine uptake rates (the rate at which the target analytes are taken up by the sampler), which are greatly dependent on environmental conditions, such as flow rate, temperature and biofouling. PRCs are non-interfering, mass-labeled compounds which slowly diffuse out of the sampler, while the compounds of interest sorb to the sampler. The known PRC loss is related to the uptake of the target compound, and thus an uptake rate can be derived (Huckins et al. 2002).
Due to their polar nature, and high affinity for water, PFASs do not sorb strongly to traditional passive samplers, which typically target non-polar analytes. A Polar Organic Chemical Integrative Sampler (POCIS) has been developed for polar compounds, including PFASs. This device incorporates a metal ring sandwiching a charged powered adsorbent which binds the polar compound, between a thin polyethersulfone membrane (Alvarez et al. 2004). The main drawback of the POCIS sampler is that the thin membrane is very permeable to water, thus causing the sampling uptake rate to be strongly dependent on the flow rate of the medium the sampler is deployed in (Gobelius et al. 2019;Kaserzon et al. 2012;Kaserson et al. 2013).
Additionally, PRCs cannot be used for these polar samplers, because the PRC affinity to the adsorbent powder is much too great, and the compounds would not diffuse out of the sampler.
Another, much newer and less widely used passive sampler has been developed for polar compounds (see Figure 1). This sampler consists of a hollow microporous polyethylene (PE) tube filled with charged powdered adsorbent and sealed at both ends (Kaserzon et al. 2019). The PE tubes have thick porous walls (2 mm) to reduce the effects of the flow rate of the medium. Instead, uptake is presumably limited by the passive diffusion of the compounds (i.e. their chemical properties) through the polyethene walls.
These passive samplers have been previously used for the detection of polar herbicides (Fauvelle et al. 2017a) and PFASs in ground water (Kaserzon et al. 2019), but more work still needs to be performed to better characterize these samplers.
The research performed for this MS thesis consists of initial field studies to determine if these samplers are a promising tool for detection of dissolved PFASs in surface water. If these tube samplers prove useful, further studies should be conducted assessing these specific variables to better understand the applicability of these samplers for PFASs uptake. For example, to be able to predict the uptake rate of individual PFASs, we would need to understand the permeability of the tube and mass transfer of the compounds by studying the porosity of the tube, the particle size of the analyte, the membrane resistance and therefore resistance to flow rate of the medium as discussed by Fauvelle et al. (2017b).
This study will evaluate and field validate the PE tube passive sampler as a sampling tool for PFASs in aquatic environments. While drinking water is a main human exposure source, studying PFASs concentrations in surface waters is critical. Surface water can often be predictive of contaminant levels in ground water (Heberer et al. 1998;Thurman et al. 1991) as the aquifers recharge with human use and rainfall. Additionally, marine and fresh water organisms bioaccumulate PFASs from the environment, which humans are later exposed to through consumption of seafood and shellfish (Berger et al. 2009 Table A and B in the Appendix for further standard information.

Waste water treatment plant deployment
Two waste water treatment plants (Field's Point and Bucklin Point, see Figure 3) servicing Providence, RI, were used as the study site for an in situ calibration deployment for these novel passive samplers. A pilot study was conducted for a one-month duration in the fall of 2017, Field's Point and Bucklin Point, Providence, RI. . At each WWTP, 3 samplers were deployed sequentially for consecutive 10-day periods (for a total of 6 passive samplers between the two sites). Water samples were also collected in precleaned polyethylene bottles, during deployment and recovery of the passive samplers (for a total of 8 water samples between the two sites). Field blanks were collected for quality assurance-optima-grade water was brought along and transferred at the site of water collection into another pre-cleaned polyethylene bottle, to mimic the water collection process and account for any potential environmental contamination.
In the late spring of 2018, a second deployment was conducted in the final WWTP effluents. At each WWTP, passive samplers were deployed in triplicate for 2, 4, 8, 16, and 29 day periods, to gain an understanding of the kinetic PFASs uptake by the passive samplers (5 time periods x 3 samplers = 15 samplers per site) daily composite water sample (sub samples collected every hour for the twenty-four-hour period and combined) was collected every day. Field blanks were also collected for quality assurance as described above.

Estuary field trial deployment
Narragansett Bay (NB) is a well-mixed, tidally influenced estuary in Rhode Island, USA (Pilson 1985

Extraction of water grab samples
Samples were stored in 1 L pre-rinsed polyethylene bottles at −20°C and thawed to room temperature for extraction. 300 to 500 mL of the sample were spiked with 10 ng (100 uL at 0.05 ng mL -1 ) 25 mix mass-labeled PFAS mixture, as surrogate standards for quantification. The water samples were then filtered using glass fiber filters, and extracted using Oasis Weak Anion EXchange (WAX) solid phase extraction cartridges (6 mL, 150 mg of sorbent), collecting the final methanol elution as the sample extract. The water extraction procedure follows EPA method LOP-AED/PEB/DK17-01-00.

Extraction of polyethylene tube samplers
Passive samplers were cleaned with deionized water to remove algal growth. Whole passive samplers were centrifuged three times for three minutes at 4,000 rpms, to remove remaining water trapped inside the tube. Passive samplers were then transferred to precleaned 15 mL Falcon centrifuge tubes, filled with 6mL methanol, spiked with 10 ng internal PFAS standard and let sit for 24 hours. The methanol was then transferred to another precleaned Falcon tube, and the PE Tube were fill with another 6 mL methanol and the process repeated three more times, for a total of four extraction steps per passive sampler. All individual sampler extracts were combined as the sample extract. to handle/store the samples, only high-density polyethylene and polypropylene were used (as they have much less reactive surfaces to prevent compound sorption). Mass-labeled internal standard were always added to the sample prior to any manipulation, to account for any losses in the method.

Instrument analysis and QA/QC
Field blanks, process blanks and instrument blanks were performed at every step of collection, extraction and analysis to account for any outside commination or sample loss.
Matrix samples were also performed to determine how the water matrix effects the PFASs behavior and recovery.

Data analysis
A five-point calibration curve was made and analyzed on the instrument to determine linearity of the MS detector and derive sample concentrations. All integrations and chromatogram analysis were done using Mass/Target Lynx software. Sample concentrations were then volume/mass, blank and recovery corrected.
Histograms of water and passive sampler concentrations were computed, and a heat map of Narragansett Bay PFASs concentrations was created by extrapolating the known concentrations from the 9 sampling sites on Ocean Data View computing software. Sampler concentrations of PFASs were compared to water parameters (temperature, salinity, dissolved oxygen, etc.) through a generalized linear model on R to determine if environmental conditions effect uptake rate of the samplers.
The sampler concentration of the WWTP time series experiment were plotted against the samplers' deployment time, to observe the PFASs uptake kinetics, and to determine whether the uptake remains linear. Regressions were ran to calculate the linearity of the uptake rate, and determine an optimal deployment time, by calculating the time needed to reach equilibrium The optimal deployment time was based on how long the uptake is linear (before it reaches equilibrium with the environment), and if the samplers have accumulated a high enough concentration of PFASs to pass the detection limits of the instrument.
Compound specific uptake rates were analyzed to discern if there is differential uptake between chain lengths and functional groups (carboxylic acids vs sulfonates).
Shorter chain compounds are expected to have higher uptake rates, due to their higher diffusivity when compared to long chain compounds.

Waste water treatment plant
The initial pilot study demonstrated that the passive samplers, the extraction and analytical method could be successfully used for the detection for 24 target analytes in both WWTPs (with the exception of HFPODA-Gen-due to instrument interference). The detection of these compounds in both the effluent ( Figure 4) and in the passive samplers with in the WWTP-and these will be discussed in greater depth later.

Estuary field deployment
The Narragansett Bay PFASs surface water concentrations (mean of the two grab samples collected on deployment and recovery of the passive samplers) were mapped and extrapolated to cover the Bay as a whole ( Figure 10). The PFASs distribution is in line with what was predicted, with the highest concentrations being north near a large human population and industry, and lower concentrations towards the mouth of the Bay.
Throughout the Bay, total PFASs were also dominated by carboxylic acids, with sulfonates comprising the next biggest fraction ( Figure 11, Table 4). Some of the spatial distribution can be explained: Site 1 (Phillipsdale Landing) is located just south of the Bucklin Point WWTP outfall, and while it is lower than the WWTP effluent itself (42 ng L -1 compared to 130 ng L -1 ) it is the second highest concentration we observed in the Bay. Site 2 (Field's Point Bay) is located near the Fields Point WWTP outfall, and also displayed elevated PFASs levels of 22 ng L -1 . Site 3 (Pawtuxet River) is located at the mouth of a tributary meeting Narragansett Bay, and displays the highest PFASs levels (62 ng L -1 ) we observed in the Bay. There is a big industry presence up stream of the Pawtuxet River, most notably electrical and metal plating, which use some PFASs containing surfactants in their production, and it is likely that industry discharge is leading to these elevated concentrations (Clara et al. (2008); Lin et al. (2009) respectively, compared to the average of the other sites 1.3% and 2.6% respectively).
These two precursors are known additives to aqueous film forming foams (AFFF), which are pervasively used across the military for fire training, and a recent fire training activity could lead to this unique PFAS signature observed at site 7 (Barzen-Hanson et al. (2017); Houtz et al. (2013). The remaining five sites are located in the wider part of the Bay with more mixing with the Atlantic Ocean water and lower human population density leading to lower PFASs levels than observed at the previous four sites. This broad range of PFASs concentrations and field conditions creates a good opportunity to test how well samplers work across a range of environmental settings.
In the Narragansett Bay field deployments, the passive samplers accumulated a sum 24 PFASs of 2.5 to 15 ng sampler -1 ( Figure 12, Table 5). In general, the trend is that higher PFASs water concentration lead to a higher passive sampler concentration. The accumulation of PFASs by the passive samplers seemed to conserve their general contribution in the water, with carboxylic acids being the dominant fraction. A notable exception, is that the passive samplers at site 7 were not dominated by 6:2-FTS and EtFOSAA like the water grab, which will be further explored in a later section.

Reproducibility and sampler type
The twelve passive samplers deployed to test the reproducibility of the PE tubes, accumulated an average of 7.1 ± 0.3 ng sampler -1 , and had consistent concentrations across all replicates ( Figure 13, Table 6). For fourteen out of the twenty-four compounds observed, the relative percent difference between replicates was between 20% and 30%, and as low as 5% for PFDS and 4:2-FTS, indicating very good reproducibility for the more abundant compounds. For the longer chain compounds (chain length greater than ten) and precursors, greater variability is observed. This variability is likely due to the compounds being present at such low concentrations (two or three orders of magnitude lower than the dominate compounds), that any fluctuation has a much greater relative impact. This variability is likely exacerbated by instrumental detection limits and poor optimization for these larger compounds with much longer retention times. Generally, the PE tube samplers appear to be reproducible, with reproducibility being mainly limited by instrumental detection limits and acquisition.
The uptake of the two different sampler types was also compared. For the Bay study, 'caged' samplers (PE tubes wrapped in polyethylene mesh) were deployed throughout the Bay, in hopes of preventing biofouling. While 'naked' samplers (bare PE tubes) were used in both WWTPs, in anticipation of less biofouling due to the sterilization of the treatment process. For all twenty-four compounds, there was found to be no significant difference (p > 0.05) in mean uptake between the two sampler types (Table 6). By visual observation, the cage did not reduce biofouling, so the simpler naked design should be used in future studies for ease of use. Additionally, in most cases the reproducibility of the naked samplers was better (lower standard error) than of the caged passive samplers (Table 6). Lastly, this lack of difference between the two samplers, enables the comparison of the Bay and WWTP sampler uptake without needing to correct for the different sampler design.

Sampling rate
The sampling rate is needed to back-calculate the water concentration of the environment the passive sampler was deployed in. The sampling rates (Rs) for individual compounds were calculated using a first-order kinetic model: Across all three study locations, sampling rates displayed similar values, ranging from 10 to 50 mL day -1 (Figure 15) for the nine compounds examined (PFBS, PFHxS, PFOS, PFPeA, PFHxA, PFHpA, PFOA, PFNA, and PFDA). However, Bucklin Point displays significantly greater sampling rates (40 ± 5.1 mL day -1 ) than both Fields Point (29 ± 2.9 mL day -1 ) and Narragansett Bay (23 ± 4.2 mL day -1 ), and Fields point displayed significantly greater Rs than the Bay as well (Table 7). Sampling rates at all study locations also seem to be linked to the chain length of the compound (Figure 16), where compounds with longer chain lengths, exhibit higher sampling rates. Similar results were also noted by Kaserzon et al. (2019). An increase of Rs with increasing chain length is contrary to expectations, as molecular diffusivity decreases with increasing chain length.
It might indicate the importance of PFASs adsorbing to the sampler surface, as partitioning constants will increase with increasing chain length (Urik and Vrana (2019).
The sampling rate was further explored using a more simplified linear uptake model: The sampling rates calculated from both equations were compared ( Figure 17, Table 8), and at all three study sites, were not found to be significantly different (p-value > 0.32), with Fields Pt WWTP linear calculation being 29 ± 2.2 mL day -1 and 29 ± 2.9 mL day -1 for the kinetic model, 40 ± 4.1 mL day -1 for Bucklin's linear and 40 ± 5.1 mL day -1 for the kinetic, and 22 ± 3.0 mL day -1 for the Bay's linear and 23 ± 4.2 mL day -1 for the kinetic. This is useful when wanting to calculate the sampling rate for compounds whose partitioning constants (Ksw) are not known, we can estimate the sampling rates by using this more simplified linear calculation.

Linear uptake
The duration of linear uptake is an important consideration for integrative passive samplers, such as this PE tube, and helps determine the optimal deployment period. The  Table 9). Calculating the time the sampler takes to reach equilibrium (teq) is a useful metric to determine the maximum deployment period: Using the mean Ksw, 201368 L kg -1 , of the nine compounds analyzed, sorbent mass (ms) of 6x10 -4 kg, and mean Rs of 0.02921 L day -1 , we get a teq value of 4166 days.
Indicating that our thirty-day deployment was well within the linear uptake phase. The optimal deployment period is typically constrained by temporal resolution, over-coming instrumental detection limits, and biofouling, so finding a compromise between these factors (likely one to several months) would be a useful deployment period for the future.

Environmental factors
Five of the Narragansett Bay fields deployment sites were located next to long term monitoring buoys that record temperature, salinity, density, dissolved oxygen, pH and chlorophyll a. When these environmental factors were plotted against the mean sampling rates at each site ( Figure 19), regressions suggest there are probable links between sampling rate and environmental factors, most notable pH (R 2 =0.62) and density (R 2 =0.52). To explore these relationships further, a generalized linear model and Akaike Information Criterion (AIC) likelihood of fit indices were analyzed (Table D).
Chlorophyll (which can be used as a proxy for organic carbon) and pH generated the strongest relationship with the sampling rate. However, this data interpretation is likely skewed by one low outlier, so this sample set is too small to derive useful conclusions on these variables. sampling rate for ground water (3.2 mL day -1 ) was an order of magnitude lower than the sampling rates for the Bay (23 mL day -1 ) and WWTP (37 mL day -1 ). These results indicate that environmental flow rate is likely more impactful on the sampling rate than initially expected. Where higher flow environments facilitate a higher sampling rate, by reducing the water boundary layer between the sampler and medium (Fauvelle et al. 2017b).

Back-calculating surface water concentrations using the sampling rate
Knowing that the one month's deployment is well within the linear uptake phase, and that the sampler design (caged vs naked) does not affect the uptake, we can use the  (Table 10).
Water grab concentrations of PFASs at deployment and recovery of the passive sampler in the Bay were plotted independently, along with the calculated surface water values ( Figure 21). Quite a bit of variability is observed between the two water grabs, particularly for site 7 (Quonsett Point), which displayed extremely high levels of 6:2-FTS and EtFOSAA only during the first grab sampling event. This was possibly due to release from AFFF application event, due to the proximity of the air force base, and that these concentrations dissipated quickly, thus not being represented in the passive samplers or in second grab sample. Sites 1 and 3 also display some variability in the two different water grab concentrations, while sites 6, 8, and 9 have more consistent PFASs levels, perhaps due to being located further from direct point sources and therefore are only exposed to consistent background levels. These observations further the notion that estuaries fluctuate quite a bit, and that a long-term passive sampler could be incredibly beneficial for more representative data.
The calculated surface water concentrations appear to underestimate the observed PFASs concentrations, particularly for the sites in the north of the Bay (1, 2, and 3) ( Figure 21 and 22). Suspecting that the flow rate of the system does indeed factor in to the sampling rate of the PE tube, the Rs values from Field's Point are likely not representative of what is observed in the specific nine sites throughout Narragansett Bay, especially with tidal and freshwater influence. Moving forward, measuring the flow rate of a deployment site is recommended in order to derive a flow rate specific sampling rate and derive more accurate PFAS concentrations.

POCIS versus Polyethylene Tube Sampler
Two previous studies have been conducted using POCIS to monitor PFASs in aquatic environments. A Swedish study used a traditional POCIS in a drinking water treatment plant, and found an average sampling rate of 45 mL day -1 (Gobelius et al. 2019), and correcting for the surface area (46 cm 2 ), its uptake rate is 0.98 mLday -1 cm -2 .
Another Australian study used a modified POCIS (smaller surface area and greater sorbent amount than a traditional POCIS) and found an uptake rate of 16.8 mLday -1 cm -2 (270 mLday -1 over 16 cm 2 ) (Kaserzon et al. 2012). Comparatively, the uptake rate produced by the PE tube sampler in Narragansett Bay was 1.28 mLday -1 cm -2 (23 mLday -1 over 19 cm 2 ). While each study has its own limitations for comparison-the Gobelius study taking place in a drinking water treatment plant, thus without the environmental and tidal fluctuations of an estuary, and the Kaserzon study, while in an estuary, however did use a modified POCIS-the PE tube sampler seems comparable to the POCIS, and has a much lower sampling rate than the modified POCIS, stipulating that the PE tubes have a reduced effect of flow rate in comparison. These PE tube samplers with a lower Rs and reduced flow rate dependence are more desirable in field deployments in order to control for environmental impacts as much as possible. However, to gain a better understanding of how these two passive samplers compare, a side by side deployment would be helpful to determine how the sampling rates correlate.

Limitations
An uncertainty of this study is the uptake pattern of the Bucklin Point time series, the effluent-and these differences in treatment could lead to differences in particle size and effluent composition which could lead to differences in the sampler uptake. Visually, the effluent flow at Bucklin Point appeared to be faster than at Fields Point, which could explain why Bucklin exhibited higher sampling rates than Fields, but without measuring the effluent with a flow meter, no definite conclusions can be drawn. Within the effluent, there are likely hundreds of other compounds, beyond PFASs, that could interfere with the sampler uptake. In order to rule out the potential effect of interference, whole effluent samples should be analyzed and screened for other compounds, in particularly pharmaceuticals, which are polar and would accumulate in the PE tubes samplers as well.
Another limitation of this study is the lack of flow rate measurements, especially in light of the data presented here that the flow of the water is a driver of the sampling rate.
The WWTP flow is likely fairly stable, but the Bay is tidally influenced, and knowing the flow rate at the specific sites would be helpful to calibrate the sampling rate.
Additionally, PFASs concentration is inherently correlated with other environmental factors in the Bay (salinity, temperature, etc.) due to the physical set up of the estuary.
The generalized linear model accounted for some of the variability of the PFAS water concentration, but a study with the same PFASs concentration and a wider range of environmental factors would be better to discern specific environmental effects.

Conclusions
Twenty-four PFASs were detected in situ in the polyethylene tube sampler, across both waste water treatment plants and Bay deployments, with the samplers being optimized for carboxylic acid and sulfonate groups. Sampling rates vary from 10 to 50 mL day -1 across all sites, with mean rates of 40 ± 5.1 mL day -1 for Bucklin Point WWTP, 29 ± 2.9 mL day -1 for Fields Point WWTP, and 23 ± 4.2 mL day -1 for Narragansett Bay.
Unexpectedly, sampling rates tended to increase with increasing chain length of the compounds, possibly due to adsorption of the longer-chain compounds. The samplers show good agreement with the PFASs water concentrations, with the most abundant compounds (PFBA, PFPeA, PFHxA, PFHpA, and PFOA) in the active grab samples, also being the most abundant in the passive samplers. The one-month deployment was well within the linear uptake phase (which could last for years). The optimal deployment will hence be a compromise between the desired temporal resolution, limitations from biofouling and amassing sufficient PFAS to overcome detection limits easily. The samplers appear to be reproducible across all sites, with PFASs uptake being 2 to 15 ng per sampler in the Bay and 60 to 70 ng per sampler in the waste water treatment plants.