Comparing PFAS removal across multiple groundwaters for eight GACs and alternative adsorbent

Eight granular activated carbons (GACs) and one alternative adsorbent (AA) were evaluated using rapid small‐scale column tests (RSSCTs) to remove low level per‐ and polyfluoroalkyl substances (PFAS) from several groundwaters. Results suggested variability among waters for adsorbents to reach breakthrough. Time to reach breakthrough appeared to be inversely proportional to the background dissolved organic carbon (DOC). Bituminous GACs (particularly F400 and UC1240LD) were more effective than non‐bituminous. The elution order for PFAS was PFHxA (C6) > PFBS (C4) > PFOA (C8) > PFHxS (C6) > PFOS (C8). Multivariate regression predicted bed volumes at which F400 reached significant exhaustion (defined here as 60%) for PFOA using only two parameters (humic acid, DOC). This merits further study as these parameters could potentially be incorporated into models for predicting PFAS breakthrough. VOCs presence negatively impacted PFAS adsorption on GAC. Relative to GACs, the AA was not nearly as impacted by DOC and showed superior performance.


| INTRODUCTION
Per-and polyfluoroalkyl substances (PFAS) are a class of aliphatic organofluoride compounds in which hydrogen is partially or fully substituted by fluorine. PFAS have been used in a variety of stain and water repellant applications, non-stick cookware, food-contact papers, and firefighting foams for >60 years (Post et al., 2013;Zareitalabad et al., 2013). Due to their widespread use, low volatility, high water solubility, and extreme resistance to degradation, they are widely detected in the environment (Post et al., 2012) including surface and groundwater (Hoffman et al., 2011;Hu et al., 2016;Post et al., 2013). While PFAS exposure to humans can occur in various ways, ingestion of contaminated drinking water may be an important pathway (Hoffman et al., 2011). PFAS exposure has been linked to potential cancer, liver/kidney damage and developmental effects in mammals (Flynn et al., 2019;Jian et al., 2018). In the last decade, multiple PFAS have been detected in the drinking water supplies of several municipalities globally. In North Carolina, Gen-X was detected at a concentration of $4500 ng/L (Hopkins et al., 2018). Gen-X was also found in Rhine -Meuse delta (Netherlands), The Cape Fear River (United States), and Xiaoqing River (China) (Bao et al., 2018;Brandsma et al., 2019;Strynar et al., 2015). Legacy PFAS compounds like PFOS and PFOA have been detected in drinking water supplies of numerous municipalities of Australia, Europe and India (Coggan et al., 2019;Phong Vo et al., 2020;Semer ad et al., 2020;Sharma et al., 2016;Yong et al., 2021).
Drinking water regulations and advisories for PFAS have been developed in numerous U.S. states (ITRC, 2022;USEPA, 2022) to recognize potential negative health effects. A sub-group of PFAS, perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid (PFOS) have received the greatest attention with respect to regulatory activity due to their widespread occurrence. The United States Environmental Protection Agency (USEPA) issued a proposed Maximum Contaminant Level (MCL) at 4 ng/L for both PFOA (perfluorooctanoic acid) and PFOS (perfluorooctane sulfonic acid). This concentration matches EPA's Unregulated Contaminants Monitoring Rule (UCMR5) in which EPA determined that the lowest concentration that can be reliably measured is 4 ng/L. The EPA also proposed to regulate four additional chemicals: PFNA (perfluorononanoic acid), PFHxS (perfluorohexane sulfonate), PFBS (perfluorobutane sulfonic acid) and GenX (HFDO-DA, hexafluoropropylene oxide-dimer acid) using a hazard index calculation which will be utilized to calculate combined potential risk of these four PFAS (USEPA, 2023).
California's Division of Drinking Water (DDW) established advisory Notification Levels (NLs) for PFOA, PFOS, PFHxS and PFBS at 5.1, 6.5, 3, and 500 ng/L, as well as corresponding Response Levels (RLs) at 10, 40, 20, and 5000 ng/L, respectively (DDW, 2022). NLs and RLs are health-based advisory levels for drinking water established by DDW for chemicals that lack enforceable Maximum Contaminant Levels (MCLs). If PFAS concentrations exceed the RL, the state recommends that the water source should not be used until appropriate treatment is implemented to reduce concentrations.
The Orange County Water District (OCWD) manages the groundwater basin of north and central Orange County, California, which serves as the primary drinking water supply for 19 major public water systems. Groundwater wells in 11 of these systems were impacted by the occurrence of PFOA above or near the RL, and thus were elected to remove more than 50 water supply wells from the service. The impacted aquifer system is within the Orange County Groundwater Basin (basin), which has been replenished for decades using surface water from the Santa Ana River (SAR) through managed aquifer recharge. PFAS have been detected in the river and basin. Other potential sources are currently being investigated. Despite having played no role in releasing PFAS into the environment, cities and water agencies in Orange County must find ways to remove them from groundwater to restore the local drinking water supply. Groundwater from the basin provides on average approximately 77% of local demand in the OCWD service area .
Treatment by separation technologies, namely adsorption on granular activated carbon (GAC) or anion ion exchange (IX) resins, is currently the most practical and cost-effective way to remove PFAS from water; these technologies have matured for other target organic contaminants and have been implemented at full-scale including for PFAS (Bertanza et al., 2020;Woodard et al., 2017). Studies have shown that PFAS adsorption to GAC is dependent on chain length and functional group and short-chain PFAS generally breakthrough earlier than long-chain PFAS (Belkouteb et al., 2020;McCleaf et al., 2017). Belkouteb et al. (2020) explored the PFAS removal by GAC at full-scale and observed that GAC removal effeiciency was affected by GAC type and surface loading rates. McCleaf et al. (2017) investigated the removal of 14 different PFAS on one GAC and one IX in terms of effects of chain length, functional group as well as isomer structure (branched or linear). Alternative adsorbents (AA) like β-cyclyodextrin polymers and clay-based materials have recently exhibited potential for PFAS removal from groundwater (Alsbaiee et al., 2016;Klemes et al., 2019;Medina et al., 2022;Xiao et al., 2017;Yan et al., 2020).
Many previous studies (Belkouteb et al., 2020;Kempisty et al., 2022;Liu et al., 2019;Najm et al., 2021;Park et al., 2020;Wu et al., 2020;Yan et al., 2020;Yuan et al., 2022) have investigated PFAS removal by GAC, and less frequently by AA (Najm et al., 2021), either at bench-or pilotscale or full-scale but have used either a single water source and/or studied only a few adsorbents. Different GAC products can vary widely in their adsorption capacity, and their performance for the particular target organic contaminant depends on empty bed contact time (EBCT), hydraulic loading rate, influent quality as well as flow rate of water flowing through the bed (Rahman et al., 2014). Yuan et al. (2022) suggested that EBCT was an important factor affecting GAC efficiency. McCleaf et al. (2017) suggested that linear versus branched PFAS can be adsorbed Article Impact Statement Time to reach breakthrough on GAC varied by PFAS type and was inversely proportional to the background DOC and humic acid content of water. AA was superior to GAC for PFAS adsorption.
differently. Further, adsorption on GAC is highly influenced by the content of total and dissolved organic carbon (DOC) in the water as high organic content can compete with PFAS for adsorption sites and reduce GAC efficiency (Gonzalez et al., 2021;Yu et al., 2009).
When designing full-scale GAC treatment for PFAS, it is advantageous to conduct a laboratory (or pilot) assessment of different GAC products available in the market; if treatment is required for multiple waters (i.e., different regional drinking water production wells, even if in the same aquifer, as in the present study), it is also advantageous to further include all or a subset of representative waters with different DOC concentrations and characteristics.
The objective of our study was to determine which GAC and AA products among several candidates are most efficient at removing PFAS for Orange County water agencies for a range of regional groundwaters considering the local groundwater geochemistry. Treatment of PFAS at drinking water wells throughout north and central Orange County will allow the water agencies to utilize local groundwater instead of purchasing more costly treated imported surface water. Rapid Small Scale Column Testing (RSSCT) results are herein reported for eight different GACs and one AA from testing for all or a subset of 10 production well waters and an irrigation well (total of 11 groundwaters).
The relative benchmarking of adsorbents against one another at laboratory scale (i.e., comparing time to significant PFAS breakthrough within days for RSSCT) allowed for screening for the most efficient media (i.e., longest lasting media) for each well site; further, RSSCT results were modeled to project the estimated media life at full-scale (i.e., months). Medina et al. (2022) reports the complementary pilot-scale study completed in parallel with the present RSSCT study. The pilot was operated using groundwater from the above irrigation well based on its water quality, which is considered representative of the range of the production well water quality across the basin. The pilot-scale study evaluated the same GAC products and further reports comparative performance of IX resins; IX resins were not included in the RSSCT in the present laboratory-scale study.

| Adsorbent media selection and rapid small-scale column testing (RSSCT)
A survey of commercially available adsorbent products was performed to identify candidates for RSSCT. The media selection was based on effective PFAS adsorption, consistent performance, commercial availability, and maturity of the products. The GAC and AA medias reported herein were from vendors: Calgon, Cabot, Evoqua and Jacobi (for GAC), and CETCO (for AA). A complete list is provided in Table 1.
The RSSCT is a flow-through test configuration that uses a small quantity of adsorptive media packed into small columns. The key to an RSSCT is that the adsorptive media size is reduced prior to column testing (via crushing), which lessens the time needed to complete the test relative to the timescales of pilot-or full-scale testing (Crittenden et al., 1986). The ratio between the sizereduced media and full-scale media was fixed, and a direct scale-up factor was established. The RSSCT can simulate months to years of full-scale operations in a relatively short period of time (several days or weeks). This reduces the time for testing, the amount of water required, and the waste produced.
RSSCTs were designed based on the constant diffusivity (CD) model. The CD model was chosen over proportional diffusivity (PD) because CD is a more conservative approach for predicting the shape of PFAS breakthrough curves (Park et al., 2020;Zeng et al., 2020).
RSSCTs were completed at Battelle (Columbus, Ohio) and configured in up-flow mode under a modified ASTM D6586-03 method, in which materials that may adsorb or leach PFAS (i.e., Teflon or PTFE) were eliminated to the extent practical and were replaced with other materials (i.e. polypropylene, PVC, and stainless steel) (ASTM, 2014). Per recommendation from USEPA ICR Manual for Bench-and Pilot-Scale Treatment Studies (USEPA, 1996), pre-filtration was accomplished utilizing a 0.20 micrometer (μm) filter. To avoid false positives from the use of Teflon, pre-filtration was accomplished utilizing an organicscertified, 10-inch, double open-ended, polypropylene, 0.20-μm cartridge filter. All test columns were constructed with equal length columns, approximately equal number of borosilicate glass beads for columns with identical column depths, and equal masses of glass wool to equalize PFAS losses between the GAC columns due to glass sorption. Configuration of each RSSCT column is described in the Table S1.
The GAC RSSCT columns were designed to simulate a full-scale EBCT of 10 min, typical for PFAS treatment. The EBCT for the AA, FLUORO-SORB ® 200 (herein abbreviated as FS200), was designed to simulate 2 min, which was selected in collaboration with the manufacturer. By running different GACs plus AA in parallel using one or two groundwaters at a time, a total of twelve sets of RSSCTs were conducted on 10 local groundwaters collected from different drinking water production wells and one irrigation well. The irrigation well was selected as a regionally representative, higher-DOC groundwater to serve as the source well for the separately reported pilot-scale study of the same GACs, AA, and other media (Medina et al., 2022). The various groundwaters are herein named Well 1 through 10, all of which are production wells except for the pilot well, Well 4. This manuscript focuses on the RSSCT results of four of these wells (Wells 1, 2, 3 and 4), while the results for the remaining wells (Wells 3a and 5 through 9) are provided in Figures S1 and S2.
For the first four groundwaters tested (Well 1, 2, 3 and 4), the RSSCTs were operated for 15 days and evaluated eight media types simultaneously for one groundwater by using eight parallel columns (one column for each media). The eight media (columns) were seven GAC products (see first seven GACs listed in Table 1) and the one AA, FS200. Two of the seven GACs were the same (Calgon FILTRASORB 400 [F400]) but one a reactivated form of F400 to evaluate any impact on performance with the less costly reactivated product.
Upon examination of results from the first four waters, the results were very similar across the groundwaters in terms of relative performance of each media. Thus, given the significant expense and effort associated with RSSCTs for PFAS, the remaining five waters only tested the two best performing GAC media (Calgon F400 and Evoqua UltraCarb 1240LD) as well as FS200. In addition, the RSSCT duration was extended to 24 days for GAC and 45 days for FS200 to provide additional resolution in the results given the late breakthrough of PFAS observed in testing of first four waters. Further, for the remaining groundwaters, given the preliminary data that emerged from the pilot operations being completed in parallel, an additional GAC (Jacobi) was added to the list of RSSCT products.

| Water quality and PFAS analysis
Bulk water (in multiple 55-gal polyethylene closed-top drums) was collected from each groundwater well. The water was first homogenized, and an aliquot was collected from the homogenized bulk water sample and analyzed for PFAS (EPA Method 537.1), TOC (SW 9060A), alkalinity (EPA Method 2320B), pH (pH meter), iron, calcium, magnesium, and manganese (SW 6020A). The raw water sample was then filtered using a 0.2-micron polypropylene cartridge filter prior to performing the RSSCT. A post-filtration sample was collected and analyzed for PFAS (Method 537.1), TOC, DOC (SM 5310B), ultraviolet (UV) absorbance (UVA or UV-254 absorbance) (SM 5910B), pH using standard methods and Excitation Emission Matrix spectroscopy (EEMs) (Horiba Aqualog instrument) to assess water quality of the RSSCT influent. Samples for EEMs were analyzed using a Horiba Aqualog spectrophotometer, which can simultaneously measure absorbance and fluorescence in multiple wavelength matrices. Samples were analyzed using a 4 second integration time from 200 to 800 nm excitation wavelength in 5 nm increments and 83-631 emission wavelength in 3.55 nm increments with a CCD gain of medium. Fluorescence intensities were blank subtracted and corrected for Rayleigh masking and inner filter effects. EEM fraction values were computed for each sample using parallel factor (PARAFAC) analysis. PARAFAC analysis is used to separate multi-dimensional data into distinct components which in this case correspond to dissolved organic matter (DOM) fractions of humic-like substances, fulvic-like substances, tryptophan, and tyrosine. PFAS in RSSCT effluent (Method 537.1) was sampled at regular intervals to monitor breakthrough from each column; the laboratory provided data below the 2 ng/L reporting limit hence these estimated points are included in the results to visualize breakthrough (e.g., Figure 1). Historical water quality data (2010-2020) for the production wells is provided in the Table S2 and Table S3 presents water quality data for homogenized raw and post filtered water. The PFOA, PFOS and PFBS ranges measured in the bulk samples from various RSSCTs were 7.6-66 ng/L, 12-51 ng/L and 3.0-14 ng/L, respectively (Table S3). The TOC levels in the bulk samples ranged from 0.34 to 2.3 mg/L. The nature of the total and dissolved organics varied between wells. Per Table S3, the TOC was equivalent to DOC for 9 of the 12 waters, indicating that all the organics are dissolved for those wells; for the other wells with higher TOC than DOC, this indicates that a portion of the organics is particulate (undissolved). Overall, the DOC at the irrigation well (Well 4) (1.6 mg/L) was greater than the other wells tested (0.21-0.92 mg/L) by 75% or more, except for Well 3 which had a post-filter DOC of 1.4 mg/L. However, this was higher than the bulk raw water TOC (0.6 mg/L) and the historical TOC values for this well (0.22-0.56 mg/L from 2007 to 2019) and thus considered a suspect result. The UVA value was greater for the Well 4 and lesser UVA at the production wells. The greater UVA is indicative of the presence of larger and more absorbable organics (fulvics and humics), while the lesser UVA indicates the presence of smaller hydrophilic organics that are less adsorbable.
A loss of PFOA (5%-67%) and PFOS (10%-78%) was observed during the pre-filtration step before eight of the 12 RSSCTs. No loss of PFOA or PFOS was observed during the pre-filtration step before the other four RSSCTs. Loss greater than 20% would be beyond the range of expected analytical variability and represents measurable loss across the pre-filter. A loss of PFOS is often observed during 0.2 μm pre-filtration before RSSCTs and is typically greater than any observed loss of PFOA. These losses have been attributed to adsorption of PFAS to suspended solids and/or particulate organics removed by the 0.2 μm prefilter. The observed losses do not affect the RSSCT data analysis because the post-filter PFAS that serves as the influent to the RSSCTs is still within the same order of magnitude as the raw water concentration and is adequate to assess the efficacy of the GAC and AA in adsorbing and removing PFAS.

| Selective RSSCT and VOCs spiking RSSCT on best GAC
One of the water production wells (Well 3a) was impacted by both PFAS and volatile organic compounds (VOCs). As such, one RSSCT set for this well water included VOC spiking prior to RSSCT to evaluate impact of VOCs on PFAS treatment. Two GAC media were evaluated using RSSCT with and without VOC spiking. VOC spiking was conducted using a method similar to that provided by Kempisty (Kempisty et al., 2019) and described in more detail elsewhere (OCWD, 2021). VOCs were measured using EPA 524.2.

| Thomas model projection
There are various empirical models developed to fit breakthrough data generated from column operations. One particular model that is widely used, known as the Thomas model, was fit to the data to generate a PFAS breakthrough curve (Xu et al., 2013) (Thomas, 1944). This model was also used to extrapolate data to project performance beyond the measured data. More details about the Thomas model and the equations are described in our complementary study (Grieco et al., 2021). For our study, use of the Thomas Model provided a reasonable fit to develop breakthrough curves for single vessel adsorption based on the RSSCT data.

| RESULTS AND DISCUSSION
Breakthrough data are plotted as PFAS concentrations (ng/L) versus number of RSSCT bed volumes or time (Figures 1 and 2). The number of bed volumes (BV) represents the amount of water passing through the columns normalized by the volume of the empty column which is occupied by the media. Initial breakthrough is defined here as first detection of PFAS above the reporting limit (which was 2 ng/L for all detected PFAS). Like our previously published and related pilot-scale work (Medina et al., 2022), significant breakthrough is herein defined as approximately 60% breakthrough (i.e., time or BV required for effluent PFAS concentration to reach 60% of the influent PFAS concentration for a particular PFAS) and is denoted as BV 60% . This was considered a key metric for comparing media performance, rather than initial breakthrough, because minimizing the time to significant breakthrough in the range of approximately 50%-80% (herein using 60%) is anticipated to be more important than time to initial breakthrough for media selection for a full-scale lead-lag system (OCWD, 2021).

| Breakthrough of long-chain PFAS (PFOA, PFOS, PFHxS) for GAC
The relevant long-chain PFAS in this study were PFOA, PFOS and PFHxS due to the remaining detectable PFAS having quite low concentrations (and intermittent non-detect levels) in the local well waters tested. Figures 1 F I G U R E 2 Breakthrough of 2A) PFHxS (left panel) and 2B) PFBS (right panel) in four groundwaters (Well 1, 2, 3 and 4) treated with seven or eight different GAC products plotted versus RSSCT bed volumes. The black dashed line indicates significant breakthrough here defined as the time or bed volumes when effluent PFAS concentration reaches 60% of the influent PFAS concentration (BV 60% ). The red dashed line indicates the PFAS detection limit at 2 ng/L corresponding to initial breakthrough. For Well 1, PFHxS breakthrough at 20,000 bed volumes for F400 was unexpected and likely an outlier. and 2 present the RSSCT breakthrough curves of PFOA, PFOS and PFHxS in four well waters showing effluent concentrations from seven GAC products plotted versus the number of RSSCT bed volumes (or eight GAC products tested in the case of Well 4).
Although there was variability among the four waters in terms of PFOA breakthrough time, the two best performing GACs for all four waters were F400 and UC1240LD. Breakthrough times for PFOA reported in other studies are more similar to our breakthrough results for Well 4 (earlier breakthrough) versus our results from the other three groundwaters showed breakthrough at a much larger bed volume. These studies reported that most GACs reached 60% PFOA exhaustion between BV of 5000-45,000 with a mean of 25,000 BV (Belkouteb et al., 2020;Franke et al., 2021;Liu et al., 2019;Rodowa et al., 2020). Our complementary pilot study, (Medina et al., 2022) on Well 4 also suggest a significant breakthrough (BV 60% ) between 35,000-55,000 BV which falls in the range of what we found in the present laboratory-scale study for Well 4 (20,000-65,000).

| Breakthrough of PFOS
PFOS breakthrough in all four groundwaters was later than PFOA. For Wells 1-3, groundwaters (Figure 1b), two products (G400, F600) showed an initial breakthrough of PFOS at BV of approximately 30,000-40,000 whereas the remaining five GAC showed initial breakthrough between 50,000 to >100,000 BV, except that for Well 1, F400 and F400R exhibited a temporary, very early breakthrough assumed to be an outlier. Overall, the performance of the seven GAC products was roughly similar in Wells 1, 2 and 3, with no significant breakthrough (effluent concentration >60% of influent or BV 60% ) observed for PFOS for the duration of the study up to approximately 100,000 BV (Figure 1b).
The difference in media performance among the different wells for PFOA and PFOS was likely due to the DOC content. The DOC content in the various waters is shown in Table S3. Well 4 had the highest DOC (1.6 mg/L), followed by Well 2 (0.91 mg/L), Well 1 (0.85 mg/L), and Well 3 (0.51 mg/L). The DOC content appeared to be inversely proportional with the media performance with higher DOC corresponding with lower media performance (i.e., earlier PFAS breakthrough in Well 4). Our results agree with previously published literature suggesting DOC competition with PFAS for adsorption on GAC (Grieco et al., 2021;Son et al., 2020). Elevated iron and manganese (Table S3) are not directly relevant to media performance in an RSSCT test due to the prefiltration performed to prevent column clogging and because iron/manganese fouling would not be anticipated to impact PFAS adsorption in short term operations.
In general, for the four groundwaters, PFHxS breakthrough was faster than PFOS and slower than PFOA.

| Breakthrough of short-chain PFAS (PFBS, PFHxA)
The short-chain PFAS that were detectable in our test waters were PFBS and PFHxA. Figure 2b plots the PFBS breakthrough curves for all GAC products in four waters. PFHxA data is included in Data S1.

| Breakthrough of PFBS
For Wells 1 and 2 groundwater, four products (H4000, G400, AC1230CX, F400R) showed an initial breakthrough of PFBS at 15,000-30,000 BV whereas the remaining three products had initial breakthrough between 40,000-50,000 BV. Well 3 groundwater, having the lowest DOC, exhibited later breakthrough. The best performing product (F400) identified for long-chain PFAS was also best performing for PFBS with significant breakthrough (BV 60% ) at $75,000 to >90,000 BV for the three groundwaters (Well 1, 2 and 3) (Figure 2b).

| Breakthrough of PFHxA
For PFHxA ( Figure S3), the breakthrough patterns were similar to PFBS. Both PFHxA and PFBS are considered short-chain PFAS and showed earlier breakthrough in Well 4 water versus similar breakthrough times in the other three wells.
When comparing the breakthrough curves for different PFAS on the same chart ( Figure S4), the PFAS with shorter carbon chains (PFBS) and carboxylic functional groups (PFHxA) generally elute faster than PFAS with longer carbon chain (PFOA) and a sulfonic functional group (PFOS). This is because PFAS with shorter chain and a carboxylic functional group are more hydrophilic and are therefore less adsorbable to GAC. It is also important to note that a carboxylic compound with a longer chain can elute faster than a sulfonic compound with a shorter chain. The elution order for the PFAS was PFHxA (C6) > PFBS (C4) > PFOA (C8) > PFHxS (C6) > PFOS (C8). The results from our study are similar to previous studies that observed more efficient removal of longer chain PFAS than shorter chain which they suggested was promoted by hydrophobic effects of longer chain PFAS with GAC (Du et al., 2014;Murray et al., 2021;Park et al., 2020). Gagliano et al. (2020) suggested that adsorption capacity of short-chain is lower than that of long-chain PFAS.
As noted previously, the number of GAC products in RSSCTs for Wells 5 through 9 was reduced as these tests focused on only the best performing GACs (F400 and UC1240LD) determined from testing of the first four wells (Figures 1 and 2). It was observed that F400 was still the longest lasting media across all groundwaters tested, except for Well 2 (Figures S1, S2).

| Comparison among GAC adsorbents
The time to reach significant breakthrough of PFAS as defined herein as BV 60% is valuable to estimate media changeout frequency and therefore O&M treatment cost. In this study, PFOA was the PFAS of interest to comparatively evaluate the performance of GAC products, as with the complementary pilot study (Medina et al., 2022). This is because PFOA levels in the impacted sites were often greater than California's Response Level of 10 ng/L combined with the earlier breakthrough of PFOA relative to other regulated PFAS; thus, PFOA is the key PFAS driving frequency of media changeout (and therefore treatment cost) for the drinking water treatment program in Orange County.
To summarize the key observations from the previously presented breakthrough curves (Figures 1 and 2), Figure 3 presents a histogram showing predicted time (in months) to initial and significant breakthrough (BV 60% ) of PFOA for various GAC products in four of the tested well waters assuming a 10 min full-scale EBCT. The bed volumes (days) from Figures 1 and 2 were converted to full-scale time (months) using the measured breakthrough curves directly or by utilizing the extrapolated Thomas model data fit. Thomas model extrapolation was used for those products that did not reach 60% exhaustion within the duration of the RSSCT. The time (days) was calculated using the conversion below: Although the performance of GAC products varied among the different groundwaters, the longest lasting GACs were F400, UC1240LD and AC1230CX. These GAC products behaved similarly in the three lower-DOC waters (Wells 1, 2 and 3) obtained from drinking water production wells, requiring 23-38 months for significant exhaustion (breakthrough at BV 60% ); whereas for Well 4 (non-potable, higher-DOC water), these products reached significant exhaustion in 14-16 months. The difference in months to significant breakthrough among the different wells for PFOA was likely due to the DOC content (Table S2), which was inversely proportional with media performance. Accordingly, F400 lasted 30-32 months for Wells 1 and 2 (similar DOC of 0.85-0.91 mg/L), 38 months for Well 3 (lowest DOC of 0.51 mg/L), and 14 months for Well 4 (highest DOC of 1.6 mg/L).
In addition to differences in performance among the GAC products, it was observed that months to initial and significant breakthrough (BV 60% ) of new F400 compared to reactivated F400 (F400R) were very similar (e.g., 26-36 months to BV 60% for the three lower-DOC waters), suggesting that the reactivation process likely produces a GAC which is as effective as a new GAC for removing PFAS from groundwater sources. Medina et al. (2022) suggested that using reactivated GAC may result in substantial cost savings as the reactivated product may cost significantly less than new GAC, though drinking water regulatory barriers exist to such media reuse in some states.

| Total PFAS breakthrough
The overall performance of different GACs in different groundwaters is compared by plotting total measured F I G U R E 3 Months to initial and significant (BV 60% ) breakthrough of PFOA for various GAC adsorbents and AA in four groundwaters (three for AA) scaled from RSSCT. Initial breakthrough was defined as the time when effluent PFOA concentration was detected above the 2 ng/L detection limit. Significant breakthrough was defined as the time when effluent PFOA concentration reached 60% of influent concentration, here plotted inclusive of initial breakthrough.
PFAS concentration (sum of five PFAS measured via EPA Method 537.1) in RSSCT effluent versus time ( Figure 4). The influent total PFAS levels in the tested waters were 94 ng/L in Well 1, 97 ng/L in Well 2, 46 ng/L in Well 3 and 65 ng/L in Well 4 water. Wells 1 and 2 had similar PFAS loads and similar DOC levels (0.8-0.9 mg/L) and thus the breakthrough pattern of different PFAS were similar. DOC levels in Well 3 were almost half of the DOC of Wells 1 and 2 and thus the breakthrough of total PFAS was later, whereas Well 4 groundwater had the highest DOC and thus the earliest breakthrough (Figure 6). Shorter chain PFAS (PFBS, pink bar), and PFAS with carboxylic group (PFHxA, light green bar) generally eluted faster than longer chain PFAS (PFOA, orange bar) and sulfonic PFAS (PFOS, dark green bar). Thus, the adsorption capacity of GAC depends on the individual PFAS compound (e.g., PFOA, PFOS, PFBS, etc.) and the quality of the water being treated.

| Bituminous and non-bituminous GAC products for PFOA removal performance
Out of the seven GAC products tested (eight including the F400R), four were bituminous coal-based products and three were non-bituminous products. Figures 1 and 2 (breakthrough curves) show that bituminous GACs may generally be more effective at removing PFAS, in that they had later initial breakthrough and lower exhaustion over longer times as compared to the non-bituminous GACs. For PFOA, this is illustrated in Figure 3 where the bituminous GACs are sorted first. For this study, only PFOA was used for a detailed comparison between bituminous and non-bituminous products. Figure 5 compares the performance of the bituminous and non-bituminous products based on volume of water treated (bed volumes filtered to 60% exhaustion). The comparison illustrates that the bituminous products slightly outperformed the non-bituminous products, but the performance difference was not significantly different (large error bars, ANOVA, p = .65). The non-bituminous products tested in our study were either coconut based, lignite based, or enhanced blended GAC. Bituminous GACs may be expected to have superior performance than nonbituminous GACs (like coconut based AC1230CX) because non-bituminous GACs are inherently microporous, which is not ideal for sorption of larger contaminants such as PFAS (Liu et al., 2019).
These results were similar to previously reported isotherm results (Qiu et al., 2007;Rahman et al., 2014) as well as RSSCT breakthrough results (McNamara et al., 2018). Pilot results (Medina et al., 2022) from our parallel study using the same GACs for Well 4 water (pilot location) were consistent with our RSSCT results, suggesting that as compared to non-bituminous GAC, bituminous GAC were more effective in PFAS removal. McNamara et al. (2018) suggested that although the bituminous GACs contain fewer high-energy pores relative to coconut-based carbons, the larger transport pores allow access to adsorption sites effective at adsorbing PFAS compounds, whereas non-bituminous carbons have relatively narrow pore structure which restricts access to high energy pores for abundant adsorption.
F I G U R E 5 Media life comparison between bituminous (n = 4) and sub-bituminous/ non-bituminous (n = 3) GAC products indicating total water volume treated (as number of bed volumes) before significant (BV 60% ) breakthrough of PFOA was reached (i.e., when effluent PFOA concentration reaches 60% of the influent concentration), hence taller bars indicate longer lasting products. The bars represent the average media life, and the error bars indicate the longest and shortest media life of the GACs tested.
F I G U R E 6 Impact of VOCs on PFOA adsorption for groundwater (Well 3a) spiked with known concentrations of VOCs (dimmed lines) compared to non-spiked (dark lines) as determined from RSSCT for two GACs. The primary y-axis (left axis) shows normalized PFOA concentration (i.e., RSSCT effluent PFOA concentration divided by influent PFOA concentration) and the secondary y-axis (right axis) shows the PFOA effluent concentration (ng/L).
3.6 | Impact of VOCs on PFOA removal performance in GAC For this study, only PFOA was used for a detailed comparison of impact of VOCs. Impact of VOCs on PFOA adsorption was evaluated using RSSCT for one of the well waters (Well 3a) for two GACs with and without VOC spiking to the groundwater influent. VOC spiking to ug/L levels of VOCs was performed with 1,1-dichloroethene (1,1-DCE), tetrachloroethene (PCE), and trichloroethene (TCE). Influent VOC concentrations (in ng/L range) are presented in Table S4. Figure 6 presents the Well 3a RSSCT data and Thomas model projection curves for F400 and AC1230CX with and without VOC spiking.
The results showed that the presence of VOCs significantly shortened the time to breakthrough for PFOA (dimmed color lines), compared to the non-VOCs spiked RSSCT (solid color lines). For the VOC-spiked RSSCT columns, PFOA reached significant breakthrough (BV 60% ) at approximately 50,000 BV (corresponding to approximately 12 months of full-scale equivalent operation) as compared to 165,000 and 275,000 BV for UC1240LD and F400 (projected), respectively (corresponding to approximately 37-62 months, respectively, of full-scale operation) ( Figure 6). With respect to VOCs removal, TCE and PCE were removed by GAC and only reached 10% or less of GAC adsorptive capacity (i.e., 10% exhaustion) at 40,000-70,000 BV. Thus, VOCs will negatively impact PFOA adsorption and corresponding time to GAC changeout at sites with VOCs co-contamination with PFAS. To our knowledge, this is the first study to report the competitive effects of VOCs and PFAS for adsorption sites on GAC media.

| Excitation emission matrix spectroscopy (EEMs)
Excitation Emission Matrix (EEM) fluorescence spectroscopy has been widely used to characterize dissolved organic matter in water (Chen et al., 2003). EEM spectroscopy results are delineated into five excitation-emission regions: aromatic proteins (Regions I and II), fulvic acids (Region III), hydrophilic acids (Region IV), and humic acids (Region V). The EEMs output is provided in arbitrary fluorescence units (AFU) in the Table S5.
Excitation-emission regions (x-and y-axis) are coupled with relative intensity (color spectrum) to provide integrated sample results on a single threedimensional plot as shown in Figure 7 which indicates the intensity of different fluorescing DOC regions in the groundwaters of Wells 1 through 4. The area distribution of fulvic-type substances (Region III) is greater in Well 4 water supporting the contribution of Region III to the overall DOC. Conversely, waters from Wells 1, 2 and 3 have greater presence of hydrophilic acids or soluble microbial products (Region IV) than Well 4 water. Hydrophilic organics such as present in Region IV generally do not adsorb to GAC (Engel & Chefetz, 2015;Wei et al., 2008) and thus are not expected to negatively influence the adsorption of PFAS to the same degree as other types of DOC. Our RSSCT results support this expectation because GAC products lasted longer in Wells 1, 2 and 3, which featured lower DOC but also less-sorbing DOC, compared to shorter GAC life for Well 4 water having greater DOC that was more sorbing (e.g., humic and fulvic acids).
To further evaluate this relationship between EEMs character and GAC life, a multivariate linear regression was conducted using number of bed volumes treated to reach 60% exhaustion of PFOA as the dependent variable, and EEMs output (presented as relative fulvic acid, humic acid, and protein contents) (AFU units) and dissolved organic carbon (DOC as mg/L) in groundwater as independent variables, for F400 GAC, presented in Figure 8 (R 2 = 0.955). More detail regarding how the AFUs were utilized to predict the linear relationship is provided in Tables S6 and S7. The statistical analysis suggests that humic acid and DOC are the relevant statistically significant variables (p < .05) to adequately describe the bed volumes to reach 60% exhaustion across the data set, which closely matches the regression analysis using all four variables. Figure 8 also shows the predicted bed volumes using humic acid and DOC only (R 2 = 0.948). The significant relationship seen in this dataset (measured by EEMs output as relative humic acid content) is notable because it suggests that in lieu of costly and time-consuming RSSCT or pilot-scale evaluation, knowledge of groundwater quality alone (DOC and humic acid) may be able to accurately predict the bed volumes (or full-scale months) to reach GAC exhaustion.

| Breakthrough of PFOA on AA
While most of the medias tested in this study were GAC, findings are also reported from contemporaneous benchscale RSSCT of the AA, FS200. The RSSCT was designed to simulate 2-min EBCT to correspond with anticipated full-scale design for FS200. A subset of this FS200 data was previously presented elsewhere (Grieco et al., 2021). For this study, only PFOA was used for a detailed comparison of breakthrough on AA.
FS200 was tested on eight of the 10 well waters, and Figure 9 presents the RSSCT breakthrough curves and Thomas model prediction curves for PFOA for these eight waters. Each RSSCT reached a C/C 0 of approximately 20%-40% within the duration of testing except for Well 5 which reached a C/C 0 of over 80%. Even though the selected testing duration did not achieve greater breakthrough and media exhaustion (due to the long media life of the product) for seven of the eight tests, RSSCT experimental results (Figure 9) generally provided a good fit to the Thomas model. Additionally, except for Well 5, the RSSCTs performed on different well waters yielded reasonably similar results (+/À 50%) including the higher-DOC case (Well 4), indicating that FS200 performance for PFOA is not affected by DOC competition with PFAS to nearly the same extent as GAC. Well 5 had similar DOC levels (0.49 mg/L) as most of the other well waters tested, however, EEMs characterization (Table S5) revealed it has a different character (greater humic-like fraction) which may have led to the poor performance (shorter media life).
A previous study showed FS200 to be highly effective in removing diverse types of PFAS from AFFF-impacted groundwater (Yan et al., 2020). Yan et al. (2020) also found that co-contaminants such as diesel and 1,4-dioxane did not affect the sorption capacity, which is in line with the present study results that DOC did not appear to significantly influence media life for FS200 (i.e., did not appear to compete with PFAS for adsorption sites on this AA). Our results were also in agreement with our related studies that tested FS200 at laboratory and pilot scale for the Well 4 groundwater (Grieco et al., 2021;Medina et al., 2022). At pilot-scale, Medina et al. (2022) found superior performance of FS200 over GAC and certain IX resins. Additionally, Medina et al. (2022) observed that while GAC tended to remove DOC from influent groundwater (along with PFAS) until the DOC removal capability was exhausted, FS200 did not remove DOC, which is F I G U R E 7 Excitation emission matrix spectroscopy (EEMS) data for four drinking water production well waters evaluated in the RSSCT study. The last panel is a legend showing the different excitation and emission regions (Chen et al., 2003). consistent with lack of significant competitive adsorption. Thus, FS200 may provide superior performance for PFAS removal, though it may not be as suitable for utilities seeking co-contaminant removal. Figure 10 compares the PFOA RSSCT results for FS200 from Well 4 and Well 2 to the best performing GAC products (F400, UC1240LD) for the same water. The Well 4 was selected to represent an elevated-DOC background water quality and Well 2 represents a low-DOC water. The data show that for Well 4, FS200 significantly outperformed the GAC products, with its Thomas model curve reaching a C/C 0 of 60% at more than 400 L of water treated. Whereas, for Well 2, due to low DOC, the performance difference between FS200 and the GAC products is less pronounced. This is because GAC performance is highly dependent on background DOC because of competition, while FS200 is less influenced by background DOC (Grieco et al., 2021). Thus, we conclude that competitive adsorption influence from DOC is much less for FS200 than GAC.
F I G U R E 8 Comparison of actual (blue squares) versus predicted (dashed lines) treatment capacity for PFOA on GAC F400 as a function of independent variables, where prediction is based on multivariate regression indicating that humic acid and DOC are the significant variables related to bed exhaustion. Two data points for Well 4 represent data from two rounds of RSSCT performed on this water. The wells are ordered from lowest to greatest according to measured BV (60%).
F I G U R E 9 RSSCT breakthrough of PFOA for multiple groundwaters treated with FS200 (alternative adsorbant) plotted as effluent PFOA concentration normalized by influent PFOA concentration (C/C 0 ) versus bed volumes or time (days). The dashed lines represent the Thomas Model curve fit.
3.9 | Comparing RSSCT to pilot results Medina et al. (2022) reports the complementary pilotscale study completed in parallel with the present RSSCT study for the same GACs and AA (FS200). In general, the scale-up of the RSSCT (laboratory-scale data) for Well 4 (same groundwater as pilot) for GACs compared fairly well to the pilot-scale performance. This is an important result since it suggests that RSSCT is a suitable method for predicting full-scale performance of GACs for PFAS. Figure 11 presents a comparison of the initial and subsequent PFOA breakthroughs. The RSSCT-predicted time to initial breakthrough did not match the actual performance (where pilot results are assumed to predict fullscale performance) for GACs or AA, where the RSSCT predicted faster breakthrough (which is conservative). However, at later breakthroughs (at or above BV 60% ), the GACs RSSCT accurately predicted the pilot result, suggesting that RSSCT is predictive of full-scale performance closer to the point of typical changeout exhaustion. Time (media life) to later breakthrough such as BV 60% is arguably more relevant than time to initial breakthrough for media selection for treatment since allowance of a higher degree of exhaustion for the lead bed is expected in full-scale operation in a lead-lag configuration before media replacement.
However, in contrast to the GACs, the FS200 (AA) RSSCT did not predict the pilot performance accurately at significant (nor initial) breakthrough ( Figure 11) in our study. The RSSCT predicted a much larger media life (not conservative) compared to what was seen in the pilot. Further work is needed to evaluate whether our findings hold true for other water qualities to verify whether RSSCT is appropriate for this adsorbent (e.g., assessing constant diffusivity assumptions) and to improve RSSCT methods for FS200.
F I G U R E 1 0 Comparison of the best performing GACs and an alternative adsorbent (FS200) for treatment of two groundwaters featuring lower versus higher DOC concentrations, plotted as RSSCT effluent PFOA concentration normalized by influent PFOA concentration (C/C 0 ) versus volume of water treated over time. The dashed lines represent the Thomas Model curve fit. The darker colored lines are for Well 4 with DOC of 1.6 mg/L and the lighter colored lines are for Well 2 with a DOC of 0.91 mg/L. F I G U R E 1 1 Comparison of PFOA breakthrough in RSSCT (this study, blue bars) and pilot scale study (complementary study, green bars) for seven GAC and one AA on Well 4 at initial breakthrough (left panel) at 21,600 BV for GACs and 200,000 for AA and at significant PFOA breakthrough (right panel) at 56,000 BV for GAC and 450,000 for AA.

| Application of data
RSSCT results for GAC and the AA, as well as the pilotscale findings reported by Medina et al. (2022) for the same adsorbents plus IX resins, were utilized by water agencies in north and central Orange County, CA to select media type (i.e., GAC or IX) and particular product (i.e., media manufacturer) for PFAS-impacted production well sites requiring treatment. As noted previously, the overall treatment objective was to address PFAS occurrence above California regulatory guideline values, driven by PFOA occurrence. As described by Plumlee et al. (2022), the majority of impacted water retailers in this region selected IX over GAC based on a combination of cost and (smaller) footprint.
An advantage of the installed vessel systems selected by OCWD and the impacted retailers is their compatibility with both GAC, IX, and potentially similar AAs, allowing for future possibility of switching to a different media product based on cost, superior expected performance, or other factors. As a new product to the drinking water industry, potential use of the AA is under review in consultation with California regulators and is being considered by other drinking water utilities for PFAS in other states (personal communication, 2022).
Under full-scale system operations for GAC, once the treatment media is spent and needs to be replaced, there are two options: (1) dispose of the spent GAC and replace with new GAC product, or (2) send the spent GAC for custom batch reactivation and return the reactivated product back into service. Given the lower cost of the reactivated GAC product (compared to new), use of reactivated GAC could result in a significantly lower lifecycle cost if it is able to provide comparable adsorption to the new product. To demonstrate this, reactivated F400 was tested in this study and compared to the new F400 in RSSCT. Results suggested that new F400 only slightly outperformed the reactivated product, though sometimes (other waters) more significantly. However, small differences in projected life between the two products may be within the sensitivity of the modeling projections and the RSSCT scale-up precision, suggesting an overall nominal difference between performance of new and reactivated products.

| CONCLUSIONS
The objective of our study was to determine which media (eight GAC, one AA) are most efficient at removing PFAS for Orange County water agencies considering the range of local groundwaters requiring treatment. Consistent with other studies, we observed that the adsorption capacity of GAC depends on the individual PFAS compound (e.g., PFOA, PFOS, PFBS, etc.) and the quality of the water being treated. The two best performing GACs were F400 and UC1240LD. The elution order for the PFAS was PFHxA (C6) > PFBS (C4) > PFOA (C8) > PFHxS (C6) > PFOS (C8). GAC media life was negatively impacted by the presence of background DOC in the water which competes for adsorption sites, i.e., competitive adsorption. Bituminous GACs may have been slightly more effective than non-bituminous with respect to media life for removing PFAS but the difference between the two GAC types was not significant in all well waters due to variability. Evaluation of the impact of VOCs showed that PFOA adsorption was negatively impacted by VOCs presence as VOCs significantly shortened the time to PFOA breakthrough. To our knowledge, this is the first study to report the competitive effects of VOCs and PFAS for adsorption sites on GAC media.
The AA, FS200, tested in this study was not impacted by the presence of background DOC and showed superior performance as compared to the GACs with respect to media life based on review of RSSCT findings (present study) together with pilot-scale observations (Medina et al., 2022). Hence this media merits further consideration for full-scale installation and testing. More work is needed to evaluate and improve RSSCT methods for FS200 given that, in contrast to the GACs, the FS200 RSSCT did not predict the pilot performance accurately at significant breakthrough (here defined as BV 60% ) nor initial breakthrough.
A multivariate regression analysis suggested that characterization of DOC and humic acid fraction in the PFAS-impacted groundwater can be used to predict the bed volumes (or full-scale months) at which GAC reaches significant breakthrough (BV 60% , which may correspond to approximate media change-out time in lead-lag configuration). This observation merits further study as these parameters could be potentially incorporated into predictive models for PFAS breakthrough.