Field Evaluation of the Sentinel™ Integrative Passive Sampler for the Measurement of Perfluoroalkyl and Polyfluoroalkyl Substances in Water Using a Modified Organosilica Adsorbent

A passive sampler specifically designed to measure perfluoroalkyl and polyfluoroalkyl substances (PFAS) in water was tested in four study areas (Ellsworth and Peterson Air Force bases, CO and SD; the Ohio River, OH, WV, KY, IN; and the Santa Ana River, CA). Locations included both groundwater and surface water locations. Over the 2‐year study, 96 passive samplers were deployed at 33 sample locations and were compared with co‐collected grab samples, all of which were measured for 19 PFAS analytes by HPLC–MS/MS. Correlations were observed (typically within 2× difference) between aqueous PFAS concentrations measured by passive versus discrete grab samples across over 5 orders of magnitude in concentration (0.5 to 150,000 ng/L). Overall relative percent difference between grab and passive results displayed a median of 18% (interquartile range of −19 to 73%). Detection limits were around 1 ng/L for a 2‐week sampling time with sampling rates ranging from 12 to 70 mL/day in flowing systems. Duplicate samplers were deployed in all study areas which indicated a 14 to 42% (median 24%) relative standard deviation in the precision of passive sampling. Larger variances were seen with sites with higher and potentially more variable water flows. A sub‐set of duplicate samplers were measured by a commercial laboratory which returned equivalent data to research laboratory measurements (43 [±26 SD]% relative percent difference). Standardized protocols and calculation methods were developed to facilitate expanded testing and future broader use of passive sampling for PFAS by site investigators.


Introduction
Perfluoroalkyl and polyfluoroalkyl substances (PFAS) are a large class of persistent contaminants routinely monitored in the hydrosphere for both scientific (Barzen-Hanson et al. 2017;Vo et al. 2020) and regulatory (Whitaker et al. 2021) purposes.There are numerous sources of PFAS to the environment (Buck et al. 2011), including landfills, biosolids, air emissions from industrial combustion processes, manufacturing and chemical production facilities, and Class B aqueous film-forming foams (AFFFs) used to extinguish flammable liquid-based fires (Glüge et al. 2020;Saawarn et al. 2022).In particular, at many military installations and other industrial facilities, the historical use of AFFF for firefighter training and emergency response has resulted in multiple PFAS species including PFOS, PFOA, PFHxA, PFBA, PFOSA, and other PFAAs including precursors compounds that are susceptible to biological and abiotic transformation impacting groundwater, surface water, and stormwater (Moody and Field 2000;Schultz et al. 2006; N. Wang et al. 2009;Plumlee et al. 2009;Place and Field 2012;Houtz et al. 2013;D'Agostino and Mabury 2014;Anderson et al. 2016;Hu et al. 2016).Analytical methods have been developed or are in progress to measure PFAS analytes in collected samples of drinking water including US EPA Method 537.1 (US EPA 2020) and environmental waters including US EPA Method 8327 (US EPA 2022).The concentrations measured using a discrete (i.e., grab, spot, snapshot) sample reflect a momentary condition which may differ from a representative concentration needed to understand average contaminant mass discharge/load, long-term remedy performance, support risk assessment and exposure modeling, or for regulatory decision-making.As an alternative or complement to the use of discrete or composite samples, passive sampling is emerging as an environmental monitoring tool due to advantages of reduced cost and ability to provide time averaged data (Alvarez et al. 2004;Stuer-Lauridsen 2005;Booij et al. 2007;Taylor et al. 2021;Edmiston et al. 2023).
Passive samplers are devices that typically contain a receiving phase, such as a solid adsorbent, liquid, or gas phase, which collects one or more analytes from the environment during deployment.Adsorbed analytes are desorbed after return to the lab to quantify the concentration in the sampling matrix.Sampling is typically performed in either an equilibrium mode where the amount of analyte in the adsorbent comes to equilibrium during the deployment time period or in an integrative mode where continuous non-equilibrium adsorption provides a time-weighted average of the analyte in the matrix (i.e., water) based on the known analyte uptake rate and deployment time.Integrative response produces a time-weighted average concentration and this information may be preferred or complementary to data from discrete-time sampling.In some cases, cumulative adsorption in an integrative passive sampler can allow lower effective limits of detection compared with equilibriumtype passive samplers.In addition to these potential technical advantages, passive sampling can also offer cost savings compared with conventional grab sampling methods due to reduced field labor time, lower equipment and shipping costs, and minimal waste generation.
Various passive sampling designs have been utilized for groundwater sampling for a range of contaminants for decades (Divine et al. 2005;IRTC 2005;Godlewska et al. 2021).Recently, both equilibrium and integrative passive sampler designs have been evaluated and optimized for PFAS measurement (Cao et al. 2018;Dixon-Anderson and Lohmann 2018;Gobelius et al. 2019;Kaserzon et al. 2019;Fuchte et al. 2020;Fang et al. 2021;Hartmann et al. 2021;Gardiner et al. 2022;Kaltenberg et al. 2023; for a review see Lai et al. 2019).One equilibrium design used a water-filled polypropylene vial sealed with a PFAS permeable membrane to sample groundwater via diffusive exchange (McDermett et al. 2022), which is similar to acquiring a grab sample of over of the equilibration time and lacks an adsorbent to concentrate analytes.An alternative equilibrium design employs a graphene-based hydrogel monolithic adsorbent which is placed in water (Becanova et al. 2021).The graphene concentrates PFAS with measured equilibrium partition coefficients >100.One of the potential limitations of equilibrium devices is the inability to determine a time-weighted average concentration for the entire deployment period because equilibrium samplers are disproportionally representative of the most recent concentrations during the deployment period (e.g., Divine and McCray, 2004) present models and experimental data that highlight this behavior).Several integrative passive samplers for PFAS have recently been reported, including those based on the diffusive gradient in thin film (DGT) approach (Fang et al. 2021;Wang et al. 2021) or designs using a polar organic chemical integrative sampler (POCIS; Kaserzon et al. 2012;Alvarez 2013).DGT-based samplers disperse an adsorbent such as OASIS-WAX in a film of agarose to create a diffusive barrier to adsorption.Diffusion leads to a characteristic rate of uptake and an integrative response (Fang et al. 2021); however, the development of biofilms have been reported by some researchers (Wang et al. 2021), which can alter sampling rates (R s ) and lead to errors when calculating the PFAS aqueous phase concentration (C w ).POCIS samplers (Alvarez 2013) place an adsorbent behind a diffusion-limiting (sub-micron pore size) membrane which is contact with the water.Designs using POCIS have been successful in measuring PFAS in aqueous samples, generally yielding accurate results; however, uptake rates can vary depending on flow rate and pH (Kaserzon et al. 2013).Novel PFAS adsorbents have been developed for use with POCIS ionic liquids (Wang et al. 2017) and hydrogels (Urík and Vrana 2019) providing improvements in PFAS binding.POCIS has practical disadvantages; for instance, standard sampling systems are relatively large making them harder to deploy in narrow well casings and the apparatus may be visible to members of the public when deployed in surface water.Also, the relatively slow rates of analyte diffusion through the membrane may preclude their use in short-duration sampling events, especially when concentrations are low.
While commercially available passive samplers are widely used by the practitioner community for sampling a wider range of volatile organic compounds (VOCs) and other constituents in groundwater (Taylor et al. 2021), various design complexities, implementation challenges, and performance uncertainties have limited the development and widespread adoption of recent passive sampler designs for PFAS by the practitioner community.Thus, a focus on making a simple robust device that is economical with reliable performance was central to the development of a PFAS-specific passive sampler evaluated in this effort.First, the sampler was designed to exhibit integrative response with a relatively fast and predictable adsorption rate to shorten deployment times and enhance sensitivity through analyte enrichment.Second, sampler analysis was intended to be convenient so third-party laboratories, including commercial laboratories, would be able to conduct analysis using standard labware and with work times less than or equal to comparable grab sample processing.Third, analysis would involve the use of isotopically labeled surrogates for isotopic dilution and QA/ QC protocols analogous to methods developed for analysis of water samples.Fourth, the samplers needed to be small and lightweight to allow versatile deployment in multiple environmental settings and configurations and to minimize shipping costs.Moreover, the samplers had to be deployable without the requirement of special tools, refrigeration, or solvents in the field.Finally, the sampler needed to be amenable to inexpensive mass production from readily available materials, permitting rapid commercial availability of the final sampler design.
The sampler design is summarized in the next section (Methods).The study described herein focused on field testing the passive sampler in both groundwater monitoring wells (Peterson Air Force Base [AFB], Colorado) and at multiple sampling locations in several different flowing surface water systems (specifically, Ellsworth AFB in South Dakota, the Santa Ana River and tributaries in southern California, and the Ohio River with stations along the borders of Pennsylvania, West Virginia, Ohio, Kentucky, and Indiana).The aims of the study were to evaluate the performance and applicability of the sampler by assessing the precision/ accuracy of passive samples compared with grab samples, determining the ruggedness of the passive sampler design in field conditions, and validating an integrative response.In a separate work, we conducted a parallel laboratory study that demonstrated the passive sampler has an integrative response to most analytes for >90 days (Edmiston et al. 2023) The study herein compared the results of passive sampler data compared with matched grab samples.Third party analysis of splits and duplicates of both water samples and passive samplers were also performed for cross-laboratory comparisons and assessing the ease of use and analysis.

Passive Sampler Design
The specific design of the Sentinel™ PFAS passive sampler (patents pending; Figure 1) evaluated here has been described previously (Hartmann et al. 2021) and is briefly summarized.The device is constructed of polyethylene (2.5 cm wide by 4.5 cm long by 0.2 mm thick) containing 100 mg swellable organically modified silica (commercially available as Osorb®, 180 to 250 μm particle size; Burkett et al. 2008) and is infused with crosslinked polyethylenimine and Cu(II) to enhance PFAS adsorption and function as a biocide to limit potential biofilm formation.The adsorbent Cu(II)-PEI-SOMS was synthesized using weak anion exchange amine groups and divalent copper ions to promote the adsorption of anionic PFAS.The adsorbent is housed in direct contact with the water inside a 1 cm through-hole held in place by polyethylene mesh screens on opposite sides.The Cu(II)-PEI-SOMS is pre-wetted during sampler construction allowing direct deployment in the field without preconditioning.Two ¼″ threaded mounting points allow for multiple modes of attachment and the device are shipped in standard 50 mL centrifuge tubes at 4 °C.Further details of the design of the sampler and laboratory testing results are presented in Hartmann et al. (2021), Edmiston et al. (2023), and SERDP project report for ER20-1127.
Sampler deployment in groundwater consists of removing the passive sampler from the centrifuge tube (shipping container), attaching deployment line to one of the threaded holes in the passive sampler housing, attaching a weight to the second threaded hole in the passive sampler housing, and securing to the well vault.Although there are many potential applications for using the passive sampler in surface water, deployment in surface water in general consists of removing the passive sampler from the centrifuge tube, attaching a deployment line to one of the threaded holes in the passive sampler housing, and attaching the line to a weight, stake, or other attachment point.After a period of deployment (days to weeks) in groundwater or surface water, the sampler is retrieved, placed into the centrifuge tube, and shipped to the laboratory on ice.A complete description of deployment, retrieval, and extraction procedures for the Sentinel™ passive PFAS sampler can be found in Edmiston (2015).

Study Area Description and Passive Sampler Deployment
The passive samplers were deployed at four study areas (Figure 2, Table 1) with widely varying field conditions.

Peterson AFB (Denver, Colorado)
The study area is located near an active fire training area where groundwater impacted with micrograms per liter-(μg/L-)-levels of PFAS occurs in alluvial paleochannel deposits.Treatment via a Horizontal Reactive Media Treatment Well (HRX Well®; e.g., Divine et al. 2018) demonstration project is underway.During two events in summer (July) and fall (October) 2021, passive samplers were deployed for 14 days and paired low-flow grab samples were collected from seven monitoring wells in the HRX Well demonstration network.The sample locations included background plume groundwater as well as groundwater in the treatment zone.
Ellsworth AFB (Rapid City, South Dakota) Remedial investigation at Ellsworth AFB identified PFAS in soil, groundwater, and surface water related to fire training activities.A permeable adsorptive barrier (PAB) is being demonstrated to treat groundwater discharging to surface water at ppb levels of PFOS/PFOA.Passive samplers were deployed in two events.The first event comprised the pilot deployment of the passive samplers in late spring (June) 2021.Passive samplers were deployed in a small stream in the PAB project area at five locations for 7 and 14 days, and paired grab samples were collected at sampler retrieval.During the second event in early spring (April) 2022 passive samplers were deployed at five stream locations for 4 days, and paired grab samples were collected at the beginning and end of deployment.In addition, to gauge variability over the deployment period, twice daily grab samples were collected at two of the five locations.Continuous temperature and stream gauge logging was also performed in a flume installed for the PAB project over the duration of the deployment.To further evaluate variability in passive sampler results, replicate passive samplers were deployed at each location.One passive sampler was deployed near the water surface (attached to a fishing float) and one passive sampler was deployed beneath the water surface.A duplicate passive sampler was deployed below the water surface, and a third passive sampler below the water surface was sent to a commercial laboratory for analysis.

Santa Ana River (SAR) and Tributaries (Orange County, California)
The SAR basin is one of Southern California's largest watershed drainage areas south of the Sierra Nevada mountains and is located in a highly urbanized, highly regulated setting.At about 100 miles long and with more than 50 tributaries, the SAR spans parts of San Bernardino, Riverside and Orange counties as it drains 2840 mile 2 of land.From May through October, much of the baseflow is typically tertiary-treated wastewater from treatment plants in San Bernardino and Riverside Counties.Previously measured PFAS concentrations in the SAR are in the nanograms per liter (ng/L) range, compared with the μg/L levels reported at the two AFB sites.Passive samplers were deployed by Orange County Water District (OCWD) personnel during routine sampling activities at six locations in July through October 2021.Two locations were sampled twice monthly (for a total of six events) and four locations were sampled once.For the two locations sampled multiple times, a new passive sampler was deployed at the time of retrieval of the previous passive sampler.Grab samples were collected at the time of passive sampler retrieval.
Ohio River (Pennsylvania, Ohio, West Virginia, Kentucky, Indiana) Passive samplers were deployed by US EPA personnel conducting a study of ambient PFAS levels in the Ohio River commissioned by Ohio River Valley Water Sanitation Commission (ORSANCO).Passive samplers were deployed for 28 to 50 days at 15 sampling sites points along the length of the Ohio River (nearly 1000 miles; see Figure S1 for the sample locations).In conjunction with passive sampling, composite samples were collected by US EPA using the equal discharge increment sampling method (Wilde 2008) and for those samples PFAS concentrations were analyzed in US EPA laboratories (ORSANCO 2022).PFAS concentrations for 28 analytes were measured in the water samples by US EPA including all 19 of the analytes measured in passive samplers described herein.Passive samplers were analyzed by the laboratory at The College of Wooster.The results were blinded and compared when the grab sampling results were released to the public in July 2022.

Laboratory Handling of Grab Samples
Grab samples (250 mL) were obtained in tandem with passive sampling and were analyzed using EPA 537.1 (see below) yielding a concentration factor of 250×.HDPE bottles were cleaned twice with methanol, rinsed with Type 1 DI water, and dried prior to sampling.Field blanks and laboratory blanks were prepared using Type 1 DI water.

Passive Sampler Analysis
Upon retrieval and return to the laboratory samplers were rinsed with PFAS-free DI water to remove debris.(Note: The adsorbent was not removed from the sampler body during analysis and all steps are done at 25 °C.)Each sampler was placed in a 50 mL conical centrifuge tube along with 15 mL of DI water containing isotopically labeled surrogates (50 ng each, Table S2) and mixed for 7 to 8 h on a platform shaker table.Note: 7 to 8 h incubation times were optimized to quantitively bind surrogates without desorption of analytes (Edmiston et al. 2023).Samplers were removed from surrogate solution and residual liquid was spun off by centrifugation for 20 s at 1500 rpm.PFAS was extracted by placing each sampler in a 50 mL centrifuge tube containing 20 mL of methanol + 2% NH 4 OH and shaking it 4 h.Samplers were next removed from the methanol solution and placed in a new 50 mL centrifuge tube and centrifuged for 20 s at 1500 rpm to spin down residual solution.The methanol recovered by centrifugation was combined with the extraction solution and evaporated to dryness under N 2 .Residue was reconstituted in 1.00 mL of 96% methanol:4% water (v/v) containing internal standards (see Table S2) and centrifuged at 14,000 rpm prior to analysis by HPLC-MS/MS.
The concentrations of PFAS in water samples were analyzed using solid phase extraction (SPE) with isotopic dilution.Samples (250 mL) were spiked with 50 ng isotopically labeled surrogate (Table S2).Samples were drawn through the SPE cartridges using an automated SPE at 2 mL/min.After the sample was added to the cartridges, the cartridges were then rinsed with 5 mL of DI water.The elution was 4 mL of 0.1% v/v ammonium in methanol.Eluents were then evaporated by nitrogen gas and reconstituted with 1.0 mL methanol with internal standards (50 ng/mL).PFAS analyte standards were obtained from Wellington Laboratories.Solvents (see Table S1) were purchased from Pharmco/Aaper and used as received.The complete list of PFAS analytes, isotopically labeled surrogates, and isotopically labeled internal standards is listed in Table S2.Ammonium acetate (LC-MS grade) was obtained from Sigma-Aldrich.All other reagents and supplies were obtained from Fisher Scientific and used as received.
PFAS concentrations were analyzed by HPLC-MS/MS using an Agilent 1200/6410 HPLC-MS/MS (QqQ) with using Infinity Lab C18 Poroshell 120 21 × 100 mm column, particle size 2.7 μm with a Restek PFAS delay column.Mobile phases were A: 5 mM ammonium acetate in water B: 95% methanol + 5 mM ammonium acetate with a flow rate of 0.250 mL/min at temperature of 45 °C (Table S1).A calibration curve was measured with each batch of samples.Method blanks, instrument blanks, and 10 ng/mL laboratory control standards were run in accordance to DoD QSM Table B-15.Limit of detection (LODs) and method detection limits (MRLs) for each analyte are reported in Table S4.A subset of passive sampler duplicates was split between our lab and a commercial laboratory (Eurofins Scientific) and analyzed by the same method described above.Calculation of PFAS concentrations in aqueous phase was determined from the mass of each analyte adsorbed during deployment.The calculation method and standard operating procedures have been reported elsewhere (Hartmann et al. 2021) and are included in Data S1.

Sampler Robustness
In total, 96 passive samplers were field deployed during the course of the study.All samplers were physically unharmed during deployment except for two devices.One sampler deployed in the Santa Ana River was damaged when an unknown member of the public apparently attempted to tear it from the mounting point.One sampler deployed in the stream at Ellsworth AFB developed a small tear in the mesh screen holding the adsorbent which led to loss of resin during return shipment.Examination indicated that the screen was damaged during construction or had a defect.Biofilm formation was not apparent on resins, although biofilms or similar films were occasionally observed on the plastic parts of the devices and could be removed using a PFAS-free lab wipe.Copper ions in the adsorbent resin likely inhibited biofilm growth.Sediment particles and other debris were sometimes found in the adsorbent housing; these were rinsed off under a stream of DI water prior to analysis.All samplers remained in good condition during analysis which included centrifugation and agitation in methanol solvent.Samplers are buoyant due to the polyethylene construction and need to be sufficiently weighted during deployment.Unexpected very low measured concentrations and the results from isotopic dilution for two samplers indicate these may have failed to be fully submerged in the groundwater monitoring well.

Sampling Rates
Knowledge of R s is necessary to accurately calculate C w from measured accumulated masses of PFAS in an integrative passive sampler.Rate of uptake is controlled by Fick's Law of diffusion and is linear until approximately half the capacity of the receiving medium is achieved (Davison et al. 2012;Caban et al. 2021).Maximum sampling times are therefore defined by the capacity.Here, a maximum sampling time is >3 months due to a high capacity PFAS-specific adsorbent that uses a swellable resin to create additional pore volume (Edmiston and Underwood 2009).Uptake rate is proportional to the aqueous phase concentration which defines the difference in chemical potential that drives the accumulation.However, integrative response is a kinetic measurement; thus, several factors can potentially affect the sampling rate (a rate constant) including flow velocity, temperature, and water chemistry.
Laboratory scale measurements were performed under controlled conditions as a first step first used to determine R s values (Hartmann et al. 2021) across environmentally relevant conditions (Figure 3).Analytes included 19 PFAS that were a mix of short chain and long PFAS encompassing most of the priority EPA 537.1 list.Recently, we have measured R s values for additional PFAS on the EPA 1633 list (Edmiston et al. 2023).Field data comparing passive samplers to paired discrete samples were used to confirm the accuracy of the laboratory derived R s values and make refinements.Across the field sites there was variability in flow rates and water chemistry (Table 1) that assisted in the validation of R s constants.It was found that the two predominant factors that affected the sampling rate were the water flow rate and temperature (Figure 3).A dependence on temperature is expected since uptake is due to diffusion.The increase of R s with increasing flow rate is likely due to turbulence decreasing the thickness of the boundary layer on the microscale (Raupach et al. 1991) or increased fluid mixing into the bed of adsorbent on the macroscale.Variables which exhibited a lesser effect on R s values included total dissolved solids (TDS), total organic carbon (TOC), pH, sulfate, salinity, or added ascorbic acid, the latter of which led to a reducing chemical potential (for complete list of R s values under different environmental conditions see Hartmann et al. 2021).The relative insensitivity of sampling rate to water chemistry is helpful in that correction factors are generally not required even when measuring sites with diverse water chemistry.Interestingly, the relatively minor changes in R s due to water chemistry may even cancel each other out in some circumstances since there is not a unidirectional concomitant increase of R s with an increasing amount of solute.Inaccuracies in the applied R s values will translate to bias in the calculated concentrations; however, these inaccuracies will not affect precision (i.e., repeatability).If higher certainty of the calculated PFAS concentrations is needed for particular sites, it is possible to apply corrections to R s via calibration with split grab samples.
Water flow rate can vary more than six orders of magnitude between the sampling contexts of quiescent groundwater versus rapidly flowing surface water streams (e.g., flow velocities in some groundwater systems are <10 ft/yr while velocities in medium-gradient streams commonly exceed 1 ft/s).Temperature can also vary significantly according to geography, season, and even time of day.Towards the development of simple tools for analysis, a concise table of field-refined R s values was determined that can be used for standard situations encountered in environmental monitoring (Table 2).Laboratory measurements showed that for flow rates above 2 cm/min the R s values did not change appreciably, presumably due to the receiving phases being in direct contact with the water.Therefore, sampling rate values were divided into two calibration ranges: (1) low flow rates for groundwater monitoring and (2) high flow rates for surface water stream monitoring.Both lab and field data demonstrated that there was a nonlinear dependence on temperature which enables temperature effects to be grouped in two conditions for typical environmental ranges: above and below 10 °C.The cut-off of 10 °C was selected based on the complete set of field and laboratorybased measurements.If water at a site averages <10 °C a PFAS-specific correction factor is multiplied to reduce the sampling rate to account for the slower kinetics at the lower temperature regime.Simplification reduces the accuracy of the calculated results; however, in practice we have found the R s values listed yield concentrations that are generally acceptably accurate for site monitoring requirements (±15 to 40%; see further discussion below).If necessary, a user can extrapolate between the values in Table 2 or conduct experiments to calibrate at conditions relevant to the specific site(s) under investigation.

Surrogate Recovery
Quality assurance methodology can be accomplished with passive sampling (Caban et al. 2021).It has been reported that the accuracy of POCIS-type integrative samplers mainly depends on the precision of the R s value (Poulier et al. 2014).However, such precision assumes complete recovery of adsorbed analytes.Near-quantitative desorption of PFAS from adsorbents has been demonstrated (Fang et al. 2021); however, recoveries are often highly dependent on solvent composition (Dixit et al. 2021).Most PFAS methods thus involve the use of surrogates and/or isotopic dilution when conducting solid-phase extraction; thus, we included a similar isotopic dilution methodology in our analysis protocol.Isotopically labeled surrogates were quantitatively adsorbed in the laboratory post-field deployment but before desorption.After desorption the amount of surrogate recovered was determined using a calibration curve.A percent recovery can be calculated to be reported for QA/QC purposes similar to values used in solid-phase extraction.The amount of a PFAS analyte was then determined by isotopic dilution.
Recoveries during the field deployment tests ranged from 60 to 110% but are typically <100% meaning the strong affinity of the resin led to incomplete recovery of PFAS (Table S5).In particular, the recovery of neutral species PFOSA was typically <50% which indicates a particularly strong adsorption affinity.The neutrality of PFOSA may decrease the effectiveness of methanol-ammonia as a desorption solvent from the weak anion exchange sites of the Cu-PEI-SOMS resin.Longer chain compounds including perfluorododecanoic acid (PFDoA) and perfluorotetradecanoic acid (PFTeDA) also showed reduced recovery, presumably due to the increased affinity of the fluoroalkyl groups to the resin.These data highlight the usefulness of surrogates and isotopic dilution to improve the accuracy of measurements of PFAS amounts accumulated in the samplers.
Another quality assurance method sometimes used for passive sampling is the application of performance reference compounds (PRCs): non-analyte standards added to the receiving phase which become depleted via diffusion into the sampling matrix during deployment.PRCs are useful in establishing the volume of water sampled.PRCs have been evaluated for the sampler and it was found that desorption rates for a very slow a long-chain PFAS analogue and are therefore considered unnecessary; however, they may be useful for short-chain analogues (work in progress).The primary value of the PRC would be to validate whether contact with water was occurring during deployment through an expected drop in PRC amount post-recovery.For instance, PRC depletion would be a helpful assurance method in wells where the submersion of a sampler cannot be directly observed.Limit-of-detection for a flowing stream, 10-day sampling time, and analysis using LC-MS instrumentation and methods as described herein.
The passive sampler is integrative (response assessed previously, see Edmiston et al. 2023), representing a timeweighted average concentration, and may therefore yield different but not inaccurate results when compared with grab samples.Passive methods can yield different results than grab or purge-and-bail methods based on lithology, in well-mixing and other processes (Britt 2005;Divine et al. 2005).Nevertheless, because for most groundwater systems, temporal concentration changes on the scale of the deployment times are usually small we believe a comparison between the passive sampler and grab results is useful, although we do not consider the grab result the "gold standard."The passive and grab groundwater samples collected in July and October 2021 showed overall an excellent 1:1 correspondence, with 56% of 164 paired passive and grab sample detections within 1.5× of each other, 71% within 2× and 97% within 5×.(A ratio of 1.5× is approximately equivalent to a 40% relative percent difference [RPD]; a ratio of 2× if ~67% RPD, and a ratio of 5× is approximately a 133% RPD.) Several potential outlier results contained higher PFAS concentrations in the grab than passive samples (Figure 4).One well location appeared to be associated with multiple outlier results in both sampling events but specific site-specific differences between this location and other sample locations (e.g., geochemical, saturated screen length in well) were not identified.The passive sampler was found to have similar accuracy for both short-and longchain PFAS (Figure 4).
Statistical evaluation of the Peterson AFB groundwater data was performed using the non-parametric Passing-Bablok regression (Figure S3).For each sample location and date, individual PFAS analyte results for passive and grab samples were added together into Σ 19 PFAS, which was subjected to regression analysis.Although the statistical evaluation is somewhat limited by the relatively small data set, a strong correspondence was found between the passive and grab results (Pearson's r of 0.79), and the slope of the correspondence was not statistically different from 1 (0.968 with 95% confidence interval around slope of 0.612 to 1.71).

Surface Water-Ellsworth AFB Site, South Dakota
The PFAS concentrations in Ellsworth AFB surface water samples differed somewhat between the 2021 pilot deployment prior to PAB reactive mat installation and the 2022 deployment completed after PAB installation (Figure S2).In  3) are within the size of the symbol.The 1.5:1 and 5:1 ratio lines for passive to grab and grab to passive ratios correspond approximately to RPD of 40% and 133%, respectively.
The first sampling event in surface water at Ellsworth AFB was the pilot field deployment of the passive sampler.Samplers were deployed at each of five locations for ~1 or 2 weeks, and paired grab samples were collected at the time of passive sampler retrieval.The paired data (Figure 5a) indicate overall a 1:1 correspondence but with more scatter than in the Peterson AFB groundwater data.No difference in performance was observed between the samplers deployed for 1 versus 2 weeks.
The second sampling event at Ellsworth AFB sought to account for variability in surface water concentrations over the deployment period as well as provide a measure of precision in the passive sampler results.During this sampling event, grab samples were collected at the beginning and end of the deployment at five sampling locations.In addition, twice daily grab samples were collected from two of the five sampling locations over the course of the passive sampler deployment period to estimate a time-weighted average concentration for comparison to passive sampler results.Stream flow and temperature were also monitored during the event using a transducer located within a flume for measure flow rate.Data from a pressure transducer demonstrated diurnal fluctuations in both temperature and flow rate, with a gradual decrease in average flow rate and increase in water temperature over the week of deployment.No significant rainfall events occurred during the deployment period.Given the relatively low temperature (average ~11.0 °C ± 4.8 °C), a low-temperature correction factor was applied to the R s values used for concentration calculations.Overall, the Ellsworth AFB surface water study area is considered to be an end member case in terms of temperature fluctuations over a sampling duration (Table 2).Most sampling applications would experience smaller temperature variations and accordingly less potential variation in R s during a sampling event.
A total of four passive samplers were deployed per location.One was deployed near the water surface to capture potential stratification effects.Three were deployed connected together, mid-depth, approximately several inches to 1 ft below the surface.Two of the mid-depth passive samplers were analyzed by the College of Wooster, and one was analyzed by a commercial laboratory (discussed below).Data from the three passive samplers analyzed by the College of Wooster were compared with the grab sample results (Figure 5b).Because the passive sampler operates in an integrative manner representing a time-weighted average concentration over the deployment period, it is preferable to compare the passive sampler results to averaged grab results.This approach is particularly appropriate for surface waters like this one, where large concentration fluctuations are possible over the passive sampler deployment period.
Figure 5b shows the correspondence between the passive and grab sample results collected during the deployment period.For each sample location, the averages of the passive samples and grab samples are plotted (filled circle symbol), and error bars represent the overall range (i.e., minimum and maximum) of the respective passive/ grab results for that location during the deployment period.Overall, there is strong 1:1 correspondence between grab and passive sample results, with slightly greater variability in the ranges of the passive sample results compared with the range of the grab results over the deployment period.Most analytes were measured at concentrations >10 ng/L and passive and grab samples matched in terms of detection/non-detection.For five analytes measured at C w < 10 ng/L (long chain perfluorocaboxylates PFUdA to PFTeDA, and HFPO-DA), the rate of detect/non-detect agreement was highly variable (14 to 100%), potentially reflecting analytical challenges associated with extracting some of the longer chain compounds especially given the at the low concentrations (<5 ng/L) observed at the sites.In the spring 2022 Ellsworth AFB data set, 58% of the averaged passive sampler results and 47% of individual passive sampler results fell within 1.5× of the averaged grab sample results.Approximately 77% of the averaged passive sampler results and 70% of the individual passive sampler results fell within 2× of the averaged grab results, demonstrating good correspondence; and 99% of the results fell within 5× as depicted on the graph (Figure 5b).

Surface Water-Santa Ana River and Tributaries, California
The concentrations of PFAS measured in the Santa Ana River basin samples were much lower than the AFB sites (on average <70 ng/L).Passive samplers were deployed in diverse field environments, for example, concrete-lined channels and stream beds consisting of sand or large cobbles and concrete riprap.Deployments and retrieval were successfully performed by third party sampling personnel from OCWD.One passive sampler was damaged apparently due to disturbance by the public.Variability in passive sampler data compared with grab sample data in samples from this field site may be related to heavy suspended sediment and organic matter load in some locations.For analytes measured at >10 ng/L (PFBA, PFPeA, PFHxA, PFHpA, PFOA, PFBS, PFHxS, PFOS), there was close to 100% agreement of passive and grab samples in terms of detection versus nondetection.Of 190 paired detections 35% of passive and grab results matched within 1.5×, 54% matched within 2×, and 82% matched within 5×.For analytes measured at concentrations <10 ng/L, the rate of detect/non-detect agreement was >70% for 7 of 10 analytes; overall at low concentrations there were more likely to be detections in grab than passive samples.It is noted again that grab samples are discrete samples while passive samples represent a time-averaged concentration.

Surface Water-Ohio River Study
The Ohio River study was a useful evaluation of the samplers since deployments were made by a third party at 15 locations (including two locations with duplicate samplers).Co-collected composites samples were run by a separate lab and blinded until the analysis was complete.Overall, there was good agreement between the composite grab samples and the passive sampling results (Table 3, Table S6).Comparison of PFOA and PFOS concentrations showed matching values for 20 of 26 data points with the passive sampling points that, although were not statistical matches, they yielded values that would have minimal impact on decisionmaking; for example PFOA at Ohio River mile 109.6 was determined to be 2 ng/L in passive sampling compared with <1.01 ng/L in the corresponding grab sample (Table 3).Across all 19 measured PFAS analytes in passive samplers 269 of 285 data points matched within ±2× of the US EPA reported value, 13/269 were 2× to 4× different, and 3 of 269 data points had a >4× difference.However, it should be noted that many of the matching data represent situations where an analyte was non-detect in both samples.In addition to being a method to compare discrete versus passive sampling these data organized by ORSANCO are the first comprehensive study of PFAS concentrations across the Ohio River main stem.Additional reports on the Ohio River study are expected from US EPA in the future.

Precision of Passive Samplers
Selecting monitoring methodology and interpreting data requires knowledge of the precision (i.e., repeatability) of a measurement technique.Replicate grab samples and passive samplers were used to determine the analytical relative standard deviation (RSD) of the measurement (Table 4).Replicate injections of a 10 ng/mL standard to measure instrument precision showed 3.4% RSD.Replicate sampling and analysis of duplicate grab samples (>70 ng/L) had an average RSD 5.1 to 7.1% establishing a benchmark for the analysis of discrete samples.Replicate injections of a passive sampler extract obtained from EAFB site (to measure instrument performance) yielded RSD of 1.1% for analytes with C w 's determined to be >70 ng/L.The lower RSD for replicate injections of the passive sampler extract relative to replicate injection of standards is likely due to the higher concentration of PFAS in the final extract compared with 10 ng/mL standard.Duplicate passive samplers deployed twice at Peterson AFB in 2021 yielded an average RSD of 14% for compounds >70 ng/L; representing 2× higher variance.Replicates collected in surface water samples yielded an average RSD of 30% for compounds measured at concentrations >70 ng/L (Ellsworth AFB) and 30 to 42% for compounds <70 ng/L (Ellsworth AFB, Ohio River, and Santa Ana River basin).Higher RSDs measured in surface water samples likely reflect higher variability in water flow rate and temperature, especially at Ellsworth where the water stream was in a 10 to 20 cm ditch-like structure (fast flow, low volume).Overall, the passive sampler measurements of replicate deployments in the same location were found to be less precise than that of grab samples, especially in fast moving water systems.Variations in the mass of sorbent and the orientation of the sampler to the flow direction may be responsible for higher variability which was observed in lab testing to be 20% RSD (Hartmann et al. 2021).However, the precision and accuracy of passive samplers (i.e., RSDs ≤30 to 45% in flowing systems at low PFAS concentration) should be sufficient for many project data quality objectives, similar to many other passive samplers, in cases where decision-making does not involve concentrations near a critical regulatory threshold.

Third-Party Analysis of Passive Samplers
Duplicate passive samplers were deployed at five sampling locations at EAFB in April 2022 and submitted to an external laboratory (Eurofins Lancaster Laboratory Environment) for analysis using standard operating procedures Note: Passive sampler deployed by EPA and measured at The College of Wooster.EPA composite sample collected and analyzed by US EPA (ORSANCO 2022).J: The reported result is an estimate.The value is less than the minimum calibration level but greater than the estimated detection limit (EDL).U: The analyte was not detected in the sample at the estimated detection limit (EDL).
1 Average of triplicate replicate injections.
2 Average of duplicate samples.
described above.The same R s values were used to calculate the aqueous phase concentration from both laboratories.There was a strong correlation between the results obtained from our internal research laboratory compared with the external commercial laboratory (Figure 6, Table S7).The average difference in PFOA concentrations measured between the two laboratories was 29% (range 12 to 43%) in the range where the C w values ranged from 150 to 500 ng/L.The average difference in PFOS concentrations measured between the two laboratories was 32% (range 13 to 55%) in the range where C w 's was 400 to 1500 ng/L.There were 20 data points that matched as non-detect by both laboratories (plotted as the point 1,1 in Figure 6).Analytical sensitivity between labs differed leading to vertically stacked series of points at low concentrations depicted at 1 ng/L and led to the inability to perform correlations at C w < 10 ng/L.The limits of detection for the internal lab were lower than the limit of quantitation reported by commercial lab.Thus, there were 23 data points that were qualified as estimated ("J-flagged") or non-detect in the commercial lab yet were detectable in the research lab with a median C w of 6 ng/L.J-flagged results correlated well with passive sampler values with all values falling in a range of 1 to 5 ng/L.There were no pairs of data that were non-detect in the research lab with a detected value in the commercial lab, again owing to the slightly higher sensitivity by research lab.The correlation plot (Figure 6) shows a linear dependence with a degree of bias to higher values measured by the external laboratory.
Overall the relative percent difference (RPD) of paired detected results in the external and internal laboratory was 43 (±26 SD)%.Results are satisfactory for an initial test evaluation; however, further refinements to the protocols and reporting methods may be necessary to ensure accuracy across multiple laboratories.

Discussion
Quantitative analytical performance is of primary importance when considering the use of passive sampling over discrete water sampling.First, passive sampling should provide equivalent data to grab sampling in terms of the frequency of positive and negative detections, and especially the latter (i.e., minimal "false negatives").Second, the precision (reproducibility) and sensitivity (limit-of detection) of passive sampling should be similar to grab sampling or at least sufficient to meet data quality objectives.Third, the passive sampling results need to be unbiased, meaning limited systematic error due to site characteristics or PFAS structure.Relatedly, the results need to be sufficiently accurate (i.e., comparable to grab samples or acceptable alternate sampling method) to allow appropriate decision-making based on the purpose of the monitoring program.The characterization of sufficient accuracy and precision depends on a project's data qualitative objectives.The accuracy and precision of the passive sampler tested here was considered with respect to ability to provide data to guide long-term trend analysis and support site characterization.Methodology in this study focused heavily on field trials with extensive num-bers of paired passive sampler versus grab sample data was used to evaluate passive sampler performance with respect to these criteria.
The agreement between frequency of detections and non-detections was evaluated across the four study areas.At C w 's five times above the analytical quantitation limit (i.e., >10 ng/L), there was near 100% agreement of detections in both passive and grab samples (606 of 610 cases).At lower concentrations (<10 ng/L), there were more frequent cases where detects and non-detects in paired grab and passive samples disagreed, and passive samples were approximately twice as likely to result in a non-detect value compared with a grab sample in the Peterson AFB, Ellsworth AFB, and Santa Ana River basin locations.In the Ohio River study 269 of 285 grab vs. passive sampler data points matched within ±2×, many of which were matching non-detects.The frequency of disagreement between detect and non-detect results corresponded more with site or event than with particular analytes; it is noted that the quantitation limit for the passive samples is tied to the mass accumulated on the sampler, rather than directly to C w , and therefore is more variable than the aqueous quantitation limit.Grab samples are more likely to contain entrained particulate matter that may contain adsorbed PFAS and contribute to additional variability and bias, especially at low concentrations.Furthermore, grab samples, even when averaged together, represent discrete point-in-time measurements, while the passive sampler is constantly exposed to water and represents the average concentration during deployment.
Precision of the passive sampler was evaluated in field duplicates of passive and grab samples collected at all sites and passive samplers collected in triplicate in the 2022 Ellsworth AFB sampling event.The RSD of measurements doubled from 5.1 to 7.1% in grab samples to 14% in passive samples of groundwater for compounds with C w > 70 ng/L, and approximately doubled again in passive samples of surface water at C w both >70 ng/L (30% RSD) and <70 ng/L (30 to 42% RSD).The dynamic nature of surface water systems may help explain the lower precision of passive sampler measurements in this environment.Higher RSDs measured in surface water samples may reflect higher variability in specific sorbent-water interactions in potentially turbulent surface water versus quiescent groundwater environment.Stratification of PFAS within surface water systems may also be a factor affecting surface water variability (Field et al. 2022).Although passive samplers deployed near the water surface at Ellsworth AFB did not yield statistically different concentrations from the passive samples collected below the surface, many compound concentrations were higher in the shallowest passive samples.Overall, the range of compound concentrations measured in replicates of Ellsworth AFB passive samples was similar to the range of concentrations measured in multiple grab samples from the same locations over the period of deployment.
A summary of overall passive sampler accuracy (presented as agreement between passive and grab results) is provided in Table 5. RPDs between passive and grab results were calculated for three study areas (Peterson AFB, Ellsworth AFB, and Santa Ana River), and for a combined measure.The Ohio River results were not included in the calculation, which considered paired detections only, due to the low frequency of detections.The RPDs were calculated as positive or negative, versus the typical presentation as absolute value, to allow for description of systematic bias that is, one method providing a consistently higher or lower concentration.Overall, the calculated RPD of paired passive and grab detections had a median value of 18% [interquartile range of −19 to 73%].The positive value of the median represents a small potential bias, with passive sampler results being on average slightly lower than grab results.Several potential sources of bias include the differences due to the integrative passive sampler versus averaged time-discrete grab samples, potential presence of suspended solids in the grab samples, or inherent differences in the sampling methods.The RPD for the Peterson AFB groundwater samples (median 3.7%) represented minimal bias and the narrowest interquartile range, suggesting that the relatively consistent conditions over the sampler deployment period may contribute to greater accuracy.
Paired passive and grab results were further compared at different concentration ranges to evaluate for bias.At C w > 70 ng/L, evidence of bias was not observed in the perfluorocaboxylate compounds, but perfluorosulfonate compounds tended to have higher concentrations in grab samples.In the 10 to 70 ng/L range, concentrations in grab samples were more frequently higher than in passive samples.At C w < 10 ng/L, mixed detections and non-detections were more common than at higher concentrations.Among paired detections at C w < 10 ng/L, evidence of bias was inconsistent.For example, concentrations of fluorotelomer sulfonates in grab samples were more frequently greater than the concentrations in passive samples, while concentrations of PFDoA in passive samples tended to be higher than in grab samples.Future field applications should continue to evaluate evidence for systematic bias in relation to site flow and geochemical conditions.Site-specific adjustment of R s values could be performed to reduce bias, should it be observed in site data.
The results from passive and grab samples generally showed a 1:1 correspondence over 5 orders of magnitude in C w , and a statistically strong correlation in Σ 19 PFAS measured in passive and grab samples was determined for the Peterson AFB groundwater results.Across the four study areas, approximately two thirds of paired passive and grab sample detections matched within 2× and nearly all paired detect results matched within 5×.Furthermore, strong agreement was observed between a set of passive samples analyzed at the research laboratory and splits analyzed at an external commercial laboratory 43 [±26 SD]% RPD.
Sampling and analysis can be complicated by the presence of suspended solids.Compared with an aqueous grab sample, the passive sampler was found to entrain less suspended sediment or PFAS associated organic matter.Though samplers deployed in some stream samples were covered with silt and/or organic matter upon retrieval, the material was removed with a wipe and water rinse, and   surrogates bound satisfactorily.Removal of sediment particles was facilitated by the open mesh design.Minimal biofilm formation was observed due to open mesh, the free movement of adsorbent particles in flowing systems, and likely the use of copper ions in the adsorbent.Grab samples were not subjected to filtration or centrifugation in this study.Future field evaluations may compare passive results to both filtered and unfiltered grab samples to better understand this component of agreement between passive and grab results, and the degree that suspended sediment may induce variability and positive bias in grab samples compared with passive samplers.

Conclusions
The field study data demonstrate that the Sentinel passive sampler can provide reliable results for a wide range of PFAS compounds, with sufficient accuracy and precision for many applications.Field deployments highlighted several practical advantages of passive samplers in general, and the Sentinel PFAS passive sampler, specifically.Field teams found the passive sampler simple to use and readily adaptable for deployment in various environmental sites.The small size of the Sentinel passive sampler facilitates its field adaptability, and, importantly, results in substantial reductions in shipping costs compared with aqueous samples.Correlations between co-collected grab samples and passive samplers indicated that analytical results are generally equivalent between the two sampling approaches with overall RPD between grab and passive results of median 18% [interquartile range of −19 to 73%].Grab versus passive sampler correlations were good across a diverse range of sample sites with PFAS concentrations ranging over 5 orders of magnitude (0.5 to 150,000 ng/L) and PFAS chain lengths from C4 to C14.Reproducibility of passive samplers was found to 14 to 42% depending on sampling sites which should provide adequate data quality for many site investigations.
Additional future work is needed to further assess the Sentinel passive sampler.Interpretations of passive and grab sample agreement in the current study were limited quantitatively by the relatively small range in individual analyte concentrations observed at a given site (typically 1 to 2 orders of magnitude), while the overall range in PFAS compound concentrations was larger.Future field demonstrations will seek to compare the passive and grab sample concentrations of individual compounds over several orders of magnitude to provide a better measure of agreement.Future field work should also further investigate potential geochemical and physical (flow, temperature) effects on R s and the potential for improved precision and accuracy with site-specific calibration.The generic R s values developed herein were designed for ease of use by practitioners, but in some cases tighter precision or accuracy may be provided by site-specific values.
The integrative nature of the passive sampler is advantageous for long-term monitoring, especially when characterization of average concentrations over 3 to 60 day timeframes (Edmiston et al. 2023)

Figure 1 .
Figure 1.Schematic (left) and photograph (right) of the passive sampler constructed of HDPE.

Figure 2 .
Figure 2. Locations of the field testing.

Figure 3 .
Figure 3. Tornado plot indicating the range and sensitivity of average sampling rate constant across C4-C9 perfluorocaboxylates and C4-C8 perfluorosulfonates, to variables.The low, nominal, and high conditions are listed.

Figure 4 .
Figure 4. Correspondence between grab and passive sample PFAS analyte results (ng/L) in Peterson AFB groundwater samples collected in July and October 2021.X and Y error bars depicting the RSD of replicate passive and grab samples (see Table3) are within the size of the symbol.The 1.5:1 and 5:1 ratio lines for passive to grab and grab to passive ratios correspond approximately to RPD of 40% and 133%, respectively.

Figure 5 .
Figure 5. Correspondence between grab and passive sample PFAS analyte results (ng/L) in Ellsworth AFB surface water.(a) 2021 pilot deployment for 7 or 17 days indicated no difference due to deployment length.(b) 2022 deployment, plotting averaged passive and grab sample results from each of five sample locations.Error bars depict the absolute range of individual passive and grab sample results collected during the 4-day deployment period.The 1.5:1 and 5:1 ratio lines for passive to grab and grab to passive ratios correspond approximately to RPD of 40% and 133%, respectively.
Abbreviations: AFB, Air Force Basel; ng/L, nanograms per liter; SPE, solid phase extraction. 1 Average RSD for PFAS analytes measured. 2Peak area variance, replicate injections (n = 10) of a 10 ng/L sample. 3PFAS concentration variance, replicate grab samples (n = 27) collected at Ellsworth AFB surface water study area for analytes with concentration >70 ng/L. 4PFAS concentration variance, duplicate grab samples (n = 2) collected at Peterson AFB groundwater study area for analytes with concentration >70 ng/L. 5PFAS concentration variance, replicate injections (n = 3) of two passive sampler extracts <70 ng/L. 6PFAS concentration variance, passive sampler duplicates (n = 2) at two locations measured at Ohio River and replicates (triplicate samples, n = 4) at four locations at Ellsworth AFB for individual PFAS with concentrations <70 ng/L. 7PFAS concentration variance, passive sampler duplicates (n = 10) for 10 samples from six locations at Santa Ana River, CA for individual PFAS with concentrations <70 ng/L. 8PFAS concentration variance, passive sampler duplicates (n = 2) collected at Peterson AFB groundwater study area for individual PFAS with concentrations >70 ng/L. 9PFAS concentration variance, passive sampler replicates (triplicates, n = 4) collected at four locations at Ellsworth AFB surface water study area for individual PFAS with concentrations >70 ng/L.

Figure 6 .
Figure 6.Comparison of passive sampler PFAS concentrations as measured by the internal research laboratory compared with an external third-party laboratory (Eurofins Lancaster Laboratory Environment).Data is from five duplicate passive samplers deployed at EAFB in April 2022.Limit-of-detection for internal lab (LOD) and limit-of-quantitation for external lab are indicated.The dashed line represents ±30% difference.

Table 4 Relative Error of Grab and Passive Sampling Analysis Measure of Error Relative Standard Deviation (%) 1
Abbreviations: AFB, Air Force Basel; ng/L, nanograms per liter; SPE, solid phase extraction. 1 Average RSD for PFAS analytes measured. 2Peak area variance, replicate injections (n = 10) of a 10 ng/L sample. 3PFAS concentration variance, replicate grab samples (n = 27) collected at Ellsworth AFB surface water study area for analytes with concentration >70 ng/L.

Table 5 Summary of Agreement Between Passive and Grab Results (Paired Detections) Expressed as Relative Percent Difference
Relative percent difference (RPD) calculated as: (grab minus passive)/(average of passive and grab).RPD of 0 indicates equivalent result with no bias.Positive RPD >0 indicates grab result greater than passive result.Negative RPD indicates passive result greater than grab result.
1 Ohio River results not included in the tabulation due to low quantity of paired detections at very low values.2Sample size (n) denotes number of detect passive-grab pairs.3 is useful.Long-term integrative verification studies are in progress and will be reported in future work.Long deployments will evaluate if reproducible detections of low-concentration (low ng/L) compounds are achievable.The Sentinel passive sampler may also assist in sampling of storm water systems with sporadic flows, and research into its applicability in sediment pore water is in progress.Finally, protocols should be adapted to comply with evolving analytical methods including draft EPA Method 1633.Overall, the results reported here represent a step forward in PFAS measurement by passive sampling.