Variable PFAS removal by adsorbent media with sufficient prediction of breakthrough despite reduced contact time at pilot scale

One alternative adsorbent (AA) and five ion exchange (IX) resins were tested for the removal of per‐ and polyfluoroalkyl substances (PFAS) from groundwater in pilot‐scale columns for up to 19 months using empty bed contact times (EBCTs) representative of full‐scale treatment. For the six detected PFAS in the pilot feed water, the long‐chain PFAS (perfluorooctanoic acid [PFOA], perfluorooctanesulfonic acid [PFOS], and perfluorohexanesulfonic acid [PFHxS]) were well removed with only PFOA, which is a perfluoroalkyl carboxylic acid (PFCA) eventually breaking through as the media became exhausted. Perfluorobutanesulfonic acid (PFBS), a short‐chain perfluorosulfonic acid (PFSA), was also well removed, whereas short‐chain PFCAs (perfluoropentanoic acid [PFPeA] and perfluorobutanoic acid [PFBA]) were not removed (i.e., immediate breakthrough). Overall, IX and AA demonstrated superior removal of PFSAs compared to PFCAs (i.e., later breakthrough of PFSAs translating to longer media life). Media life varied, ranging from 6 to 15 months before adsorbents reached a significant PFOA breakthrough. The performance of the two adsorbents piloted at shorter EBCT reasonably predicted the longer (representative) pilot EBCT results (within ±20–30%) for the same adsorbents following data scaling. This suggests that pilot‐scale testing may be conducted at a faster pace and therefore more economically.


INTRODUCTION
Per-and polyfluoroalkyl substances (PFAS) are a class of anthropogenic contaminants that have been detected in drinking water at trace levels (Crone et al., 2019;Kurwadkar et al., 2022).PFAS feature a hydrophobic peror polyfluorinated carbon chain and a hydrophilic functional group which gives them surfactant properties for industrial (Kucharzyk et al., 2017) as well as consumer uses (Post et al., 2013;Zareitalabad et al., 2013).PFAS have widespread occurrence and long half-life in the environment, and several PFAS are suspected to have harmful health effects (Gaballah et al., 2020;Pelch et al., 2019;Tucker et al., 2015).Removal of PFAS is a challenge as conventional water treatment does not eliminate PFAS (Houtz et al., 2016;Pan et al., 2016;Plumlee et al., 2008;Rostvall et al., 2018).
Several state regulatory agencies have proposed either Maximum Contaminant Levels (MCLs) or advisory levels for drinking water for several PFAS.The United States Environmental Protection Agency (USEPA) issued a proposed (draft) MCL at 4 ng/L for both PFOA (perfluorooctanoic acid) and PFOS (perfluorooctanesulfonic acid) (USEPA, 2023).This concentration aligns with Environmental Protection Agency (EPA)'s Unregulated Contaminants Monitoring Rule 5 (UCMR5) (USEPA, 2022) determination of the lowest concentration that can be reliably and consistently measured analytically.The EPA also proposed to regulate four additional chemicals: PFNA (perfluorononanoic acid), PFHxS (perfluorohexanesulfonic acid), PFBS (perfluorobutanesulfonic acid), and GenX (HFDO-DA, hexafluoropropylene oxide-dimer acid) using a Hazard Index calculation to calculate a combined potential risk (USEPA, 2023).USEPA's final PFAS regulation for drinking water is expected to be promulgated in 2024.
The water agency-led, pilot-scale treatment study that is the subject of this article took place in California.California's Division of Drinking Water (DDW) has established advisory Notification Levels (NLs) for PFOA, PFOS, PFHxS and PFBS at 5.1, 6.5, 3, and 500 ng/L, respectively; and corresponding Response Levels (RLs) at 10, 40, 20, and 5000 ng/L, respectively (DDW, 2022).NLs and RLs were established by DDW as interim drinking water advisories for chemicals prior to the establishment of enforceable MCLs.If PFAS concentrations exceed the RLs, the state recommends that the water source not be used until appropriate measures are implemented to reduce concentrations.Once the federal USEPA MCLs are finalized, states must either adhere to the federal regulations or make their state regulations more stringent.
PFAS are amenable to several conventional ex situ adsorption treatments including granular activated carbon (GAC), ion (anion) exchange (IX), and some newly developed alternative adsorbents (AAs).AAs are neither GAC nor IX and may be either earth materials (modified) or synthesized materials such as cyclodextrin polymers (Wu et al., 2020).Several recent publications have demonstrated successful removal of PFAS from water using the above-mentioned technologies in pilot or laboratory settings (Ellis et al., 2022;Grieco et al., 2021;Medina et al., 2022;Najm et al., 2021;Pannu et al., 2023).
IX offers a cost-effective drinking water treatment for certain regulated PFAS as discussed in our previous studies (Medina et al., 2022;Pannu et al., 2023;Plumlee et al., 2022).IX synthetic resins are usually made of polymer plastic beads activated with chemical solutions (e.g., trimethylamine) to create negatively charged surfaces with ions (e.g., Cl-) combined with a positively charged functional group on the resin surface.As the negatively charged PFAS contacts the resin, the negatively charged ion is exchanged for PFAS, leaving PFAS attached to the surface.Several studies have established the effectiveness of IX in removal of PFAS at laboratory and pilot scales (Franke et al., 2019;Gao et al., 2017;Woodard et al., 2017), including our prior phase one pilot study (Medina et al., 2022).Our study presented herein was designed with the primary objective of comparing five new IX products as candidate resins for local fullscale PFAS treatment systems.These products have not been reported on pilot scale previously and are from several different manufacturers.These products vary in type of resin and PFOA adsorption capacities.Our performance driver was PFOA given its relatively elevated levels in local drinking water supplies (groundwater) compared to California PFOA RL and tendency for earlier breakthrough of this PFAS as compared to the other long-chain PFAS.Resin details are provided in Table 1.Out of the six PFAS evaluated in this pilot study, three were long-chain (PFOA, PFOS, and PFHxS), and three were short-chain (PFBS, perfluoropentanoic acid [PFPeA], and perfluorobutanoic acid [PFBA]) PFAS, as measurable by EPA Method 533.
Pilot studies for treatment media generally require several months or over a year to complete to reach the target effluent (treated water) concentration (ideally reaching at least 50-60% exhaustion of media to establish a meaningful breakthrough curve, i.e., effluent concentration reaching that percent of the influent PFAS concentration as indicated by C/C 0 ).The extended duration of pilots is due to the key design parameter of Empty Bed Contact Time (EBCT) being selected to match (or be close to) the anticipated EBCT of the full-scale design.The EBCT, or time required for the water to pass through the (theoretical) empty media column or vessel, is calculated by dividing the volume of the empty bed by the flowrate.The contact time between the water and the adsorbent influences amount of PFAS adsorbed; therefore, it is important that this contact time be long enough to adsorb PFAS and capture the mass transfer zone (MTZ, the region of the adsorbent bed that is still adsorbing PFAS).Full-scale EBCT for drinking water PFAS treatment systems is typically near 2-3 min for IX and near 10 min for GAC adsorbents.
In real-world applications, using full-scale EBCT during pilot testing of adsorbent media can lead to relatively long test durations (e.g., greater than a year to determine because of large PFAS capacity of adsorbents or low influent concentrations prolonging time to breakthrough).Hence, facilities may elect not to complete site-specific field testing of adsorbents prior to full-scale investment, instead making media selections based on performance claims, cost, and/or experiences at other sites; or facilities may pursue more rapid bench-scale testing to derive a performance estimate (Grieco et al., 2021;Najm et al., 2021;Pannu et al., 2023).Some pilot studies have utilized shorter EBCTs to expedite testing based on the assumption that the results can be scaled up; only one study has reported the PFAS IX adsorption at different EBCTs to determine the MTZ for various PFAS (Murray et al., 2021).Consequently, a secondary objective of the present study was to test the hypothesis that pilot-scale media columns operated at half the full-scale EBCT (leading to half the normal pilot runtime) can accurately predict the kinetics of the same pilot run or full-scale operations at full-scale EBCT via a simple linear extrapolation of the 50% EBCT results to 100% EBCT, for both IX and AA.

Column design
The pilot system consisted of a pre-filtration skid and a test media skid provided by Evoqua Water Technologies (Figure 1).The media skid held a total of six primary adsorbent columns as well as two secondary columns that tested a shorter EBCT for one of the IX and the AA.The pre-filter skid comprised two 5-μm cartridge filters connected in parallel upstream of the adsorbent media skids to capture suspended particles from the groundwater serving as influent to the pilot to prevent potential adsorbent media column plugging.Secondary cartridge filters were installed ahead of the media skid to provide redundancy.Influent sampling ports were located before and after the pre-filtration skid.A previous pilot performed using the same skid setup and prefiltration system suggested that the filters only captured suspended solids and do not affect the PFAS concentrations reaching the columns (Medina et al., 2022).
The adsorbent media skid consisted of a steel frame that housed the columns, sampling ports, flowmeters, piping, and secondary 5-μm filter.All columns were made of schedule 40 transparent polyvinyl chloride (PVC) plastic; fittings and tubes were made from stainless steel or PVC plastic.The skid included columns with a nominal diameter of 5.08 cm (2 in) and nominal length of 91.4 cm (36 in).Each column was equipped with one variable area flow meter with a range of 5.68-56.8L per hour (0.025-0.25 gpm) and an effluent sampling port.The media skid was equipped with an influent pressure regulator and pressure gauge for the common influent as well as a pressure gauge for the common effluent.It also included one totalizer for the whole skid and a 5-μm filter on the outlet side to prevent any accidental release of PFAS laden material through the outlet water.
Primary columns were installed at a media bed depth of 79 to 91 cm and flow rate of 0.20 to 0.23 gpm resulting in an EBCT of 2 min based on vendor recommendations for each media; for those IX media, performance was only evaluated at a 2-min EBCT.For two other media, IX-4 resin) and AA, an in-series two-column arrangement (primary and secondary), was installed at a bed depth of 40 to 41 cm (average depth = 40.5 cm) in each column and a flow rate of 0.2 gpm resulting in EBCT of 1 min for each in-series column.Effluent from the 1-min EBCT column (primary) was the influent to the secondary column to achieve a total of 2-min EBCT for the overall (primary plus secondary) column.This arrangement allowed direct comparison of the 50% and 100% EBCT performance for these media, where media life results from the 50% EBCT column may be scaled to predict 100% EBCT by doubling the observed time to breakthrough.For IX-4, the 2-min EBCT column (the second column in the two-column in-series set) failed prematurely because of packing and flow issues and did not reach significant PFOA breakthrough.

Source water
The source water for the pilot study came from a non-potable irrigation groundwater well in Anaheim, California.The well has a maximum depth of 92 m below ground surface (bgs) and a screened interval from 52.4 to 57.6 m bgs.Two submersible 3.55 cm (1.4 in.) pumps (Grundfos 5 SQ05-140, Brookshire, TX) were placed inside the well casing at a depth of 55 m bgs.One pump provided sufficient flow and pressure of non-chlorinated groundwater to the pilot system, whereas the other pump provided redundant service.The pilot site and well water were selected as sufficiently representative of the north/ central Orange County aquifer general water quality and PFAS concentrations, recognizing some variability in local aquifer water quality may influence media performance at groundwater treatment sites elsewhere in the region overlying this aquifer.More details about Orange County Water District (OCWD) and how OCWD manages the groundwater in its service area are provided in Plumlee et al. (2022), including impacts from PFAS and the agency's response.

Resin selection and analytical method
Six commercially available, single use (i.e., non-regenerable) adsorbents were included in the study in which five were IX (LanXESS Lewatit ® TP108DW, Purolite D9279, Evoqua APR-2, Evoqua PSR2+, and ResinTech SIR-110-HP) and one AA (FLUORO-SORB ® FLEX 200, Table 1).Table 1 describes the different media and their properties.The adsorbents were selected based on an early 2021 survey of promising products marketed for removal of PFAS; FLEX 200 has since been discontinued.The adsorbents previously tested in our Phase 1 work (Medina et al., 2022) from a prior market survey were not repeated in this study.Details on hydraulic loading rates, flow rates, and EBCT are provided as Table S1.
Influent and effluent samples were collected monthly from the pilot using recommended field sampling protocols from California State Water Resources Control Board (SWRCB) DDW and Division of Water Quality (DWQ) for PFAS.Samples were analyzed either using EPA Method 537.1 (18 analytes) or EPA Method 533 (25 analytes) at the OCWD Philip L. Anthony Water Quality Laboratory in Fountain Valley, California under certification from the state's Environmental Laboratory Accreditation Program (ELAP).The reporting limit for all PFAS was 2 ng/L.Quality assurance/quality control (QA/QC) samples included one sample duplicate, one sample spike, and one spike duplicate, during every sampling event.Field reagent blanks (FRBs) were also collected as F I G U R E 1 Schematic of the pilot system showing skid with six columns.Column four (IX-4) and column six (AA) have both primary and secondary columns to measure effluent for 1-min EBCT and 2-min EBCT, respectively.EBCT, empty bed contact times.required for EPA Method 533.Quarterly, additional water samples were collected to test for general waterquality parameters.

Source water characterization
Source water (groundwater influent to the pilot) analyses consistently detected the presence of six of the 25 PFAS reported with EPA Method 533 at levels above the reporting limit of 2 ng/L.The average pilot influent concentrations for these PFAS are presented in Table 2.The long-chain PFAS detected were PFOA (19 ng/L), PFOS (24 ng/L), and PFHxS (10 ng/L).The short-chain PFAS were PFBS (14 ng/L), PFPeA (7 ng/L), and PFBA (29 ng/L).Table 2 also includes water-quality parameters and concentrations of relevant groundwater coconstituents detected in the source water including dissolved organic carbon (DOC) at 1.2 mg/L.The concentrations of inorganic constituents (Cl À , HCO 3 À , Mn 2+ , Mg 2+ , NO 3 À , PO 4 3À , SO 4 À ) did not change between influent and effluents (see Table S4), indicating no/minimal removal by the adsorbents tested in the current study as might be expected for IX media.This suggests PFAS specificity for these IX and AA adsorbents and underscores the contribution of hydrophobic interaction.DOC removal did occur initially for the IX medias (none for AA), but there was minimal DOC removal after 6 months of pilot runtime.

Breakthrough of long-chain PFAS
Figure 2 shows breakthrough curves for three long-chain PFAS (PFOA, PFOS, and PFHxS) for the five IX and one AA tested at an EBCT of 2 min.PFOA eventually broke through all the adsorbent columns reaching between 70% and 90% exhaustion by the end of the 19-month study period.For this discussion, "significant" breakthrough is herein defined as approximately 50% breakthrough (T 50% ) (i.e., time required for effluent PFAS concentration to reach 50% of influent PFAS concentration).The 50% breakthrough chosen in this study is a conservative milestone because based on information obtained from several full-scale wellhead treatment systems recently commissioned and operating in OCWD's service area, the water retailers typically replace the spent IX media with fresh media when the lead vessel effluent to influent PFOA concentration ratio (C/C 0 ) is in the range of 68-84%.The dual vessel configuration (lead-lag) was selected to achieve overall water-quality goals and meet state regulatory requirements.Any residual PFOA detected between the lead and lag vessels is adsorbed by the lag vessel, that is, non-detect PFOA in lag (system) effluent.The operating protocol with respect to timing of media change outs was developed based on breakthrough results from pilot-scale media testing (OCWD, 2021) imputed to lead-lag modeling.Additional discussion of the full-scale treatment facilities operation including media change out is provided in Section S1.Other pilotscale evaluations for PFAS treatment media may only be operated long enough to reach initial breakthrough of the target PFAS to identify longest-lasting media.However, this may not be representative if the full-scale treatment system media will be operated to a greater degree of exhaustion (to significantly reduce costs associated with more frequent media replacement) such as in the lead vessels for lead-lag systems in Orange County, and since the media exhibiting the longest time (performance life) to 50% exhaustion may not necessarily be the same as exhibiting the longest time to initial breakthrough.
Removal efficiency for PFOA varied across adsorbents, but they all achieved significant breakthrough (T 50% ) between 5 and 13 months which corresponded to bed volumes (BV) treated of 110,000-200,000.The earliest significant breakthrough (T 50% ) for the tested water source occurred in IX-5, and the longest lasting adsorbents (latest breakthrough) were AA and IX-4 (previous study, Medina et al., 2022).We note that relative performance of adsorbents can vary by site/water source, that is, the best-or worst-performing media in our case may not be the same for another study.Our results were similar to our previously published work using pilot-scale columns at the same site (same water source) that demonstrated an AA of a similar composition (also from CETCO FLUORO-SORB 200) was superior for PFOA removal as compared to other adsorbents tested during that study (Medina et al., 2022).The earlier breakthrough trend of PFOA as compared to other long-chain PFAS is also consistent with other previous studies (Franke et al., 2021;Pannu et al., 2023;Zaggia et al., 2016).
The other two long-chain PFAS (PFOS and PFHxS) did not experience breakthrough during the duration of the study period, that is, non-detect effluent concentrations after 19 months in pilot operation.This was expected as previous studies have shown PFSAs have a strong affinity to adsorb on IX resins as compared to PFCAs (like PFOA) (Medina et al., 2022;Woodard et al., 2017).In our previous pilot study (Medina et al., 2022), we observed that both IX and AA had more than 3 times higher adsorptive capacity of PFOS than PFOA.Furthermore, PFOS has been shown to competitively displace the lower affinity PFAS for adsorption on IX resins (Gao et al., 2017;McCleaf et al., 2017).Thus, the results observed in the present study are consistent with previous studies that these adsorbents are better at removing per/polyfluorinated sulfonic acids as compared to carboxylic acids (Franke et al., 2021;Pannu et al., 2023;Woodard et al., 2017).

Breakthrough of short-chain PFAS
The short-chain PFAS detected in the source water were PFBS, PFBA, and PFPeA.Although PFBS breakthrough has been reported previously in the literature, PFBA and PFPeA are less studied as these PFAS are the newly added PFAS in the EPA Method 533 list of 25 PFAS analytes.Although we do not have data for the first month, appears as though all three PFAS compounds were initially removed by all adsorbents (Figure 3), PFBA and PFPeA broke through quickly at just 1 month and <25,000 bed volumes.PFPeA had a bit more sustained removal by IX-4 for up to 3 months and also F I G U R E 2 Long-chain PFAS (PFOA, PFOS, PFHxS) breakthrough on six adsorbents tested at 2-min EBCT.The y-axis shows C /C 0 (effluent concentration divided by instantaneous influent concentration).The lower x-axis shows time in months elapsed, and the upper x-axis shows cumulative bed volumes treated during the study period.There was no breakthrough of PFOS or PFHxS over the 19-month duration of the pilot.EBCT, empty bed contact times; PFAS, per-and polyfluoroalkyl substances; PFHxS, perfluorohexane sulfonate; PFOA, perfluorooctanoic acid; PFOS, perfluorooctane sulfonic acid.reached 100% exhaustion by month six.These data are similar to a previously reported study where both PFBA and PFPeA broke through within 50,000 BVs for the various adsorbents tested (Ellis et al., 2022).
In contrast, PFBS was completely removed by all adsorbents for approximately 5 months (non-detect PFBS in all six effluents) and was well removed overall because PFBS did not reach significant breakthrough (T 50% ) for any of the five IX resins tested but did reach T 50% at $10 months ($220,000 BV) for the AA (CETCO FLUORO-SORB FLEX 200).
Our data suggest that single use IX resins are more efficient at removing PFSAs (PFOS, PFHxS, and PFBS) as compared to PFCAs (PFOA, PFBA, and PFPeA), which is consistent with literature reported before (Ellis et al., 2022;Medina et al., 2022).Murray et al. (2021) suggested that PFAS removal is chain length and functional group dependent.Ellis et al. (2022) observed that regenerable resins fared equally or were superior in removing short-chain PFCAs like PFBA as compared to single use resins.Earlier breakthrough of short-chain PFAS could be because of lack of hydrophobic and van der Waals forces that attract the hydrophobic tail (nonpolar) of PFAS to the hydrophobic portions of the IX resins.Shorter tail length (shorter chained) PFAS results in lower hydrophobicity leading to earlier breakthrough (Boyer et al., 2021;Wang et al., 2019) as compared to long-chain PFAS.Ellis et al. (2022) suggest that facilities concerned with long-chain PFAS may opt for single use IX resins (similar to the ones tested in this study) but those concerned with short-chain PFAS may opt for testing regenerable resins at bench or pilot scale.

Months to significant PFOA breakthrough
To summarize the key observations from the breakthrough curves presented in Figures 2 and 3, Figure 4 presents a bar chart of time (in months) to significant PFOA breakthrough on various adsorbents tested at a 2-min EBCT representative of full scale.In addition to Figure 4 presenting the time to significant breakthrough, time (months) and bed volumes to initial and 25% breakthrough are presented as Table S2.However, the time to more significant breakthrough is a more meaningful result for identifying the longest-lasting media for fullscale application because allowance of a fair degree of exhaustion (80-90%) for the lead bed is expected in a lead-lag configuration before media replacement.Choice of the limit of lead bed exhaustion percentage will drive the media change-out frequency and therefore O&M for PFAS treatment costs.Per Figure 4, four of the six adsorbents reached significant breakthrough (T 50% ) between 6 and 8 months of pilot runtime, whereas the two superior adsorbents reached significant breakthrough at 17 months for IX-4 and 13 months for AA-indicating the benefit of pilot testing to identify superior media.For IX-4, the 2-min EBCT column (the second column in the two-column in-series set) failed prematurely because of packing and flow issues and did not reach significant PFOA breakthrough.Hence, an estimate of T 50% for IX-4 F I G U R E 3 Short-chain PFAS (PFBS, PFPeA, PFBA) breakthrough on six adsorbents tested at 2-min EBCT.The y-axis shows C/C 0 (effluent concentration divided by instantaneous influent concentration).The lower x-axis shows time in months elapsed and the upper x-axis shows cumulative bed volumes treated during the study period.EBCT, empty bed contact times; PFAS, per-and polyfluoroalkyl substances; PFPeA, perfluoropentanoic acid; PFBS, perfluorobutanesulfonic acid; PFBA, perfluorobutanoic acid.
was projected using available 1-min EBCT data by multiplying time elapsed by two.
Although our prior pilot study at the same water source (Medina et al., 2022) already evaluated the resin IX-4, we included it again in the present study's pilot evaluation that was operated 25 months later alongside other medias to verify its superior performance given its broad implementation for treatment facilities in Orange County (Plumlee et al., 2022).This is a novel aspect of the present study because repeating a pilot run for the same resin is not generally pursued to our knowledge (given the time and expense).It is generally unknown how reproducible any pilot observation of media life may be; for instance, if piloting predicts a media to have longer life than another, would a duplicate test (separate batch and/or later time) reveal that the difference is in the range of uncertainty?Per Figure 4, we observed that the repeat evaluation of IX-4 resulted in a similar breakthrough curve performance as the prior test, albeit earlier initial breakthrough of PFOA was observed in the present study.However, the key observation of time to significant breakthrough (T 50% ) was within 20% of the prior pilot study (Medina et al., 2022) at 18 months in the prior study and 16 months in the present study.Changes in water quality over time may contribute to the slight performance difference; however, general water-quality and PFAS concentrations did not change significantly between the two phases of the study (PFAS data are presented in Table S3), and PFAS breakthrough was normalized as C/C 0 .Ideally, more pilot-scale media evaluations will include duplicate columns and even repeat testing in a subsequent trial to shed light on reproducibility.

Impact of EBCT
PFAS selective adsorbents (such as tested herein) evaluated to obtain breakthrough data are very efficient in removing PFAS, and thus, pilot tests take a substantial time (several months) to complete when design criteria include EBCTs representative of (matching) EBCT used at full-scale, for example, 2 min for IX.As a result, decision-making for PFAS treatment may be less timely and more costly.Reducing the EBCT (e.g., 25%, 50%, 75% of the design EBCT) carries the risk that the shorter EBCT and shorter bed depth of the pilot column may not contain the MTZ for target PFAS and in particular the shorter chain and less adsorptive PFAS (Murray et al., 2021).Further, a smaller EBCT may cause a greater potential for preferential flow path development and premature breakthroughs.To our knowledge, there is no data or guidance in the literature indicating whether it is appropriate to pilot using shortened EBCTs and then scale up the findings to the design EBCTs.
F I G U R E 4 Months (Mo) and bed volumes (BV) to reach 50% (C/C 0 ) breakthrough of PFOA for six adsorbents piloted at 2-min EBCT.Significant breakthrough was defined for this study as when effluent PFOA concentration reached 50% of influent concentration because this approximately corresponds to when the lead bed media would be replaced in a full-scale lead-lag treatment system.For IX-4, because the 2-min column failed prematurely, the bar represents 1-min EBCT data projected to 2-min by simply multiplying time (months elapsed) by two.EBCT, empty bed contact times; PFOA, perfluorooctanoic acid.
Hence, one of the novel concepts conducted as a part of this study was testing the hypothesis that pilot-scale columns run at half the design EBCT (requiring half the runtime) can (by a simple scaling extrapolation) accurately predict the kinetics of a pilot column run at the design EBCT, and therefore by extension be useful for predicting performance of a full-scale system that uses the design EBCT.We tested this hypothesis on AA and IX-4 for the breakthrough of one long-chain PFAS (PFOA) and one short-chain PFAS (PFBS).Shorter EBCTs require a shorter bed depth and result in faster PFAS breakthrough.Because half (1 min) of the design EBCT (2 min) was utilized, it was expected that if 1-min EBCT sufficiently contains the MTZ, the breakthrough time observed for a given C/C 0 at 1-min EBCT will be half of the breakthrough time at 2-min EBCT.Breakthrough time results of 1-min EBCT columns were therefore projected to 2-min EBCT by multiplying the elapsed time by two.The data for PFOA and PFBS are presented in Figures 5 and 6, respectively.A scatterplot smoothing method known as local regression (LOESS) was used to aid visualization of individual column breakthrough.LOESS is a non-parametric method which fits multiple linear least squares regressions to localized subsets of the data serving as a moving average that reduces noise and variability.

Impact on long-chain PFAS breakthrough (PFOA)
Figure 5a shows the actual and predicted PFOA breakthrough data for IX-4.Unfortunately, as previously described, the 2-min EBCT column (pink line) failed at  (Medina et al., 2022) is compared to this study (pink) at same 2-min EBCT.The 2-min EBCT breakthrough curve (purple) is projected from the 1-min EBCT observations (data not shown) by multiplying time (months elapsed) by two.For AA, the observed 2-min EBCT breakthrough curve (green) is compared to the 2-min EBCT breakthrough curve (orange) projected from the 1-min EBCT data.EBCT, empty bed contact times; PFAS, per-and polyfluoroalkyl substances; PFOA, perfluorooctanoic acid.9 months because of packing and flow issues in the columns and only reached a PFOA breakthrough of approximately 30%.Thus, for comparison, we utilized the data obtained at 2-min EBCT from our previous publication (black "Phase 1" line) (Medina et al., 2022) from a pilot conducted at the same site using same source water and 2-min EBCT.Separately, as noted previously, the 1-min breakthrough elapsed time was multiplied by two to project it to a 2-min EBCT prediction (purple line), for example, for measurements of 1-min EBCT at 0, 1, 2, and 3 months, the projected breakthrough time for 2-min EBCT column would be at 0, 2, 4, and 6 months.
The data on IX-4 suggest that the column operated with 1-min EBCT projected to 2-min EBCT (purple line) underpredicted the performance of the resin with an earlier initial breakthrough than was observed with the Phase 1 two-minute EBCT data (black line) as well as the first 6 months of operation of the 2-min EBCT column (pink line).The projected T 50% for IX-4 to a 2-min EBCT from the 1-min EBCT pilot was reached at 16.5 months for this test, which differs less significantly (9% lower) from the T 50% reached at $18 months for Phase 1 data operated at 2-min EBCT.With respect to the initial breakthrough, the 1-min EBCT showed initial breakthrough at $2 months (which projects to $4 months) compared to the (Phase 1) 2-min EBCT column initial breakthrough of 6 months; these values are here rounded to one significant figure because sampling occurred monthly and thus the actual observed breakthrough could theoretically be as much as $3-4 weeks prior.Although the initial breakthrough time was therefore underpredicted by $30%, the T 50% prediction using 1-min EBCT was within 20% of the actual 2-min EBCT observation from the prior test and is more relevant than F I G U R E 6 Pilot breakthrough of short-chain PFAS (PFBS) for (a) IX-4 and (b) AA.Points are direct observations, and lines represent data smoothing using local regression.For IX-4, Phase 1 breakthrough curve (black) previously observed for same pilot site (Medina et al., 2022) is compared to this study (pink) at same 2-min EBCT.The 2-min EBCT breakthrough curve (purple) is projected from the 1-min EBCT observations (data not shown) by multiplying time (months elapsed) by 2. For AA, the observed 2-min EBCT breakthrough curve (green) is compared to the 2-min EBCT breakthrough curve (orange) projected from the 1-min EBCT data.EBCT, empty bed contact times; PFAS, per-and polyfluoroalkyl substances; PFBS, perfluorobutanesulfonic acid.time to initial breakthrough for projecting full-scale treatment costs (media life) and comparing candidate resins.These findings for 1-min EBCT may be considered acceptable uncertainty to justify a faster, less costly pilot test.
For AA, Figure 5b presents the projected breakthrough data for 1-min EBCT (orange line) in comparison to 2-min EBCT (green line).It matched fairly well up to approximately T 25% breakthrough after which the projection somewhat underpredicts the time to significant breakthrough at T 50% , that is, the actual time to T 50% was 12.5 months compared to 10 months (within 20%) in the projection.After >50% breakthrough, the actual breakthrough curve demonstrated considerable variation, but overall, the projection continued to underpredict breakthrough time.
Collectively, the fair degree of similarity between the projected and actual breakthrough times for both IX and AA products indicates that the MTZ for PFOA may be sufficiently captured in the shorter 40-cm bed depth associated with 1-min EBCT, and that 1-min EBCT columns reasonably projected the PFOA breakthrough time at least up to T 50% of 2-min EBCT columns.Thus, a shorter pilot EBCT can be used to predict PFOA breakthrough, though the prediction may only be within $20-30% of actual performance, which may be acceptable.The finding also suggests that vessel breakthrough projected from 50% sampling ports in full-scale systems may be sufficiently predictive.Consistent with the present study, a previous study by Ellis et al. (2022) compared 2-and 3-min EBCT for regenerable resin at pilot-scale and concluded that breakthrough data obtained at the shorter EBCT (2-min) accurately predicted the breakthrough for the 3-min EBCT, which was the recommended EBCT for this resin.

Impact on short-chain PFAS breakthrough (PFBS)
Like the prediction for long-chain PFAS (PFOA), shortchain PFAS (PFBS) breakthrough was evaluated at a reduced EBCT.Figure 6 presents the breakthrough of PFBS for both AA and IX-4.For IX-4, purple line is the 1-min EBCT projected to 2-min EBCT, the actual 2-min EBCT breakthrough is the pink line, and the black line is the Phase 1 line.For AA, green line is the actual data and orange line is 1-min data projected to 2-min data.
For IX-4, the PFBS did not have significant breakthrough during the study period for either the 1-min or 2-min EBCT columns (Figure 6a); thus, without significant breakthrough, no observations can be made about how well the overall breakthrough curve shape of the projection would match actual.On the other hand, the lack of breakthrough (non-detect PFBS) for the first $22 months is in perfect agreement for the projection to actual, suggesting that the faster, 1-min EBCT pilot was sufficient to represent 2-min EBCT performance.The projection suggests initial breakthrough of PFBS after $22 months (purple line), which is consistent with the actual observations (albeit variable) from the Phase 1 two-minute EBCT pilot (black line).Hence, overall, the local regression of the 2-min projection aligns quite closely with the local regression of the Phase 1 2-minute data.
Figure 6b shows that the projected breakthrough on AA using the 1-min EBCT (orange line) was earlier than the breakthrough in actual 2-min EBCT (green line).
Comparing the time to a significant breakthrough (T 50% ), the projected breakthrough was at $7.5 months, and the actual breakthrough was at $10 months (i.e., approximately 25% difference).The breakthrough trend after the C/C 0 of 50% continued to be underpredicted (earlier breakthrough) by the 1-min EBCT while the overall profile was similar.
Predicting short-chain PFAS breakthrough within 20-30% may be acceptable for utility planning purposes if using shorter EBCT pilot columns.Relatively poor prediction of short-chain breakthrough when using shorter EBCT pilot columns relative to long-chain PFAS may be expected because hydrophilic short-chain PFAS move faster through the beds to reach unsaturated layers (in terms of PFAS adsorption) of the bed creating a longer mass transfer zone.Short-chain PFAS can also be displaced by stronger adsorbing PFAS reaching the same layers creating overlapping mass transfer zones and making it difficult to determine adsorption capacities for short chains (Li et al., 2020).Reducing EBCT has been previously shown to negatively impact removal of micropollutants especially those with slow adsorption kinetics (Kearns et al., 2021;Park et al., 2020).Short chains are typically not the treatment target for utilities because of lesser bioaccumulation potential and associated human health risk (Chi Thanh, 2022), though PFAS regulations may continue to evolve.

CONCLUSIONS
Six adsorbents evaluated at pilot scale in this study were more efficient at removing long-chain PFAS (later breakthrough) as compared to short-chain PFAS.Additionally, PFSAs showed superior removal (i.e., media were longer lasting) compared to PFCAs.Significantly, a shorter EBCT and a shorter bed depth in pilot columns reasonably predicted the PFOA and PFBS breakthrough of the same media using full-scale design EBCT (longer EBCT), suggesting that the MTZ was contained in the shorter bed depth for both long-and short-chain PFAS.
Acknowledging that breakthrough at shorter EBCT may not be predictive of full-scale contact time for all longand short-chain PFAS, the two PFAS that experienced sufficient breakthrough in this study (PFOA and PFBS) generally showed agreement between results for shorter versus full-scale EBCTs.This observation suggests that practitioners have the flexibility to conduct pilot-scale testing at a faster pace and more economically by designing pilots at shorter EBCTs.
Poor removal performance by adsorbents for some short-chain PFCAs (i.e., PFBA and PFPeA) may mean that full-scale treatment units that have short-chain removal treatment goals (such as based on regulatory drivers) must be configured to achieve longer contact times or operated with more frequent media change outs.

F
I G U R E 5 Pilot breakthrough of long-chain PFAS (PFOA) for (a) IX-4 and (b) AA.Points are direct observations, and lines represent data smoothing using local regression.For IX-4, Phase 1 breakthrough curve (black) previously observed for same pilot site Properties of IX and alternative adsorbent media selected for the pilot.
Water quality of the pilot source water.
T A B L E 2