Deliverable D4.1 – Performance and assessment of drinking water treatment trains for removal of PFAS and industrial chemicals

Project ID N°: 101036449 Call: H2020-LC-GD-2020-3 Topic: LC-GD-8-1-2020 - Innovative, systemic zero-pollution solutions to protect health, environment, and natural resources from persistent and mobile chemicals Preventing Recalcitrant Organic Mobile Industrial chemicalS for Circular Economy in the soil-sediment-water System Start date of the project: 1st November 2021 Duration: 42 months Main author: Fiona Rückbeil (BWB), Alexander Sperlich (BWB), Christian Dietrich (BAFG), Alicia Hartmann (BAFG), Jochen Kuckelkorn (UBA), Anna von Wichert (UBA), Peter Behnisch (BDS), Harrie Besselink (BDS) Lead Beneficiary: BWB Type of delivery: R Dissemination Level: Public Filename and version: PROMISCES_D4.1_Drinking-water-treatment-trains (version 1) Website: https://promisces.eu/ Due date: 28 February 2025 (M40) D4.1 – Performance and assessment of drinking water treatment trains for removal of PFAS and industrial chemicals D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 2 © European Union, 2025 No third-party textual or artistic material included on the publication without the copyright holder’s prior consent to further dissemination by other third parties. Reproduction is authorized provided the source is acknowledged Disclaimer The information and views set out in this report are those of the author(s) and do not necessarily reflect the official opinion of the European Union. Neither the European Union institutions and bodies nor any person acting on their behalf may be held responsible for the use which may be made of the information contained therein. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 3 Document History This document has been through the following revisions: Authorisation Distribution This document has been distributed to: Version date Author/Reviewer Description 0.01 13/12/2024 F. Rückbeil & A. Sperlich (BWB) Initial version 0.02 04/02/2025 J. Kuckelkorn & A. von Wichert (UBA), C. Dietrich & A. Hartmann (BAFG) Supplemented to initial version 0.1 04/02/2025 F. Rückbeil & A. Sperlich (BWB) First draft for review 0.2 12/02/2025 P. Schumann (KWB) First draft review back to authors 0.3 25/02/2025 F. Rückbeil & A. Sperlich (BWB), J. Kuckelkorn & A. von Wichert (UBA), C. Dietrich & A. Hartmann (BAFG) Corrections Final version of the deliverable 0.4 27/02/2025 F. Sermondadaz (IEIC) Quality control 1.0 27/02/2025 J. Lions (BRGM) Final Version for distribution Authorisation Name Status Date Review P. Schumann Reviewer 12/02/2025 Validation U. Miehe/ V. Zhiteneva WP 4 leader 25/02/2025 Quality Control F. Sermondadaz Administrative and financial manager 27/02/2025 Approval J. Lions Project Coordinator 27/02/2025 Name Title Versio n issued Date of issue BWB, UBA, BAFG, BDS Task 4.1 partners Version 1 03/03/2025 D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 4 Executive Summary The aim of this deliverable is to enhance the understanding of competitive adsorption and ion exchange of perand polyfluoroalkyl substances (PFAS) and other industrial persistent, mobile and toxic substances (iPMT) in the presence of dissolved organic matter (DOM) and to optimize their removal from groundand drinking water in fixed-bed filters. To achieve this, laboratory and pilotscale experiments were combined with monitoring of a large-scale groundwater treatment plant with activated carbon at a legacy contaminated site. In order to identify an optimal treatment train for PFAS removal, different high performance and tailored adsorbents were tested at laboratory scale and benchmarked against state-of-the-art granular activated carbon (GAC). Initially, a comprehensive adsorbent screening was conducted that involved a series of equilibrium jar tests using 18 different adsorbents and spiked drinking water (including 13 PFAS and 7 iPMT). These adsorbents were selected based on a literature research for high performance adsorbents for PFAS and iPMT removal and included several activated carbons, strong basic anion exchange resins as well as novel adsorbents such as surface modified clays, a cyclodextrin polymer, and a zeolite. The results of the jar tests revealed that strong-base anion exchange resins (IX) were the most effective at removing PFAS. However, these IX were particularly vulnerable to competition from nontargeted DOM components. Activated carbons were found to be the most effective adsorbents in removing a wide range of iPMT, but they were ineffective at removing short-chain PFAS (which are by definition of the OECD all PFAS with less than 6 perfluorinated carbon atoms (ECD, 2021)). For the alternative adsorbents tested, there were significant variations in their ability to remove PFAS and iPMT, with the bentonite-based surface modified clays (SMC) showing considerable potential due to their high PFAS selectivity. In line with current literature, the results showed that PFAS removal increased with the length of the perfluorinated carbon chain. For compounds of the same chain length, those with a sulfonate group were removed more effectively than those with a carboxylate group. Based on the results from the jar tests, three GAC products, two SMC, and two IX were selected for testing in laboratory rapid small-scale column tests (RSSCT) with spiked drinking water, including 15 PFAS and 8 iPMT. The RSSCT dataset was used to determine the breakthrough curves of all investigated compounds and to shortlist the adsorbents for subsequent piloting. In addition, the data was analyzed to identify possible synergies between adsorbents to identify the most promising treatment train. Usage rates were quantified for all adsorbents based on the RSSCT data. In the subsequent piloting, an IX and a SMC, which had excelled in the laboratory tests, were piloted alongside a large-scale activated carbon plant. In addition, the most promising treatment train was tested, consisting of a combination of activated carbon and ion exchange. All process variants were tested at the same site for their ability to remove PFAS from legacy contaminated groundwater. The target analysis in the pilot tests was supplemented by additional sampling campaigns for bioassay testing using TTR-TRβ CALUX (PFAS CALUX) tests and for suspect screening. These results allow a more holistic evaluation of the adsorptive processes. All adsorbents tested were able to reduce the PFOA-equivalents in the PFAS CALUX tests. In the suspect screening, a lower number of substances was found in the GAC effluent than in the effluents of the SMC adsorber and the IX. This may indicate less selective removal by the activated carbon filter or leaching from the other adsorbents, but further studies may be of interest. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 5 In the subsequent cost analysis, it became clear that the choice of a suitable adsorptive process is highly site-dependent in regards to the composition of the PFAS and iPMT, DOC concentration, and targeted effluent values. UV254 was found to be a potential surrogate parameter for real-time monitoring of PFAS removal by AC and IX. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 6 Table of contents 1 Introduction ................................................................................................................................. 10 2 Methodology ................................................................................................................................ 11 2.1 Investigated PFAS, iPMT and other OMP .............................................................................. 11 2.2 Adsorbent selection .............................................................................................................. 12 2.3 Laboratory scale adsorbent testing....................................................................................... 12 2.3.1 Jar tests .......................................................................................................................... 12 2.3.2 RSSCT .............................................................................................................................. 13 2.3.3 RSSCT breakthrough curve fitting and calculation of usage rates ................................. 15 2.4 Largeand pilot-scale groundwater treatment..................................................................... 15 2.4.1 Pilot site description ...................................................................................................... 15 2.4.2 Sampling campaign for PFAS CALUX and suspect screening ......................................... 16 2.5 Analytical methods ................................................................................................................ 17 2.5.1 Target analyses of PFAS, iPMT and other OMP ............................................................. 17 2.5.2 Quantification and characterization of DOM ................................................................ 17 2.5.3 PFAS CALUX bioassay ..................................................................................................... 18 2.5.4 Suspect screening .......................................................................................................... 18 3 Results and discussion ................................................................................................................. 20 3.1 Laboratory-scale adsorbent testing ...................................................................................... 20 3.1.1 Competitive adsorption of DOM ................................................................................... 20 3.1.2 Removal of PFAS, iPMT and other OMP ........................................................................ 22 3.2 Large and pilot-scale groundwater treatment ...................................................................... 30 3.2.1 PFAS removal, cost analysis and UV254 correlation ....................................................... 30 3.2.2 PFAS CALUX bioassay results ......................................................................................... 33 3.2.3 Suspect screening .......................................................................................................... 37 4 Conclusion .................................................................................................................................... 40 References ........................................................................................................................................... 41 Annex ................................................................................................................................................... 49 D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 7 List of Tables Table 1 Design parameters of the CD-RSSCT. ...................................................................................... 14 Table 2 Overview of sampling IDs, times and points for suspect screening and PFAS CALUX. ........... 17 Table 3 Chromatographic gradient used for suspect screening of PFAS and iPMTs in the BAFG suspect screening method. ............................................................................................................................... 19 Table 4 Specific adsorbent media costs. .............................................................................................. 33 Table 5 PFAS CALUX results of GW sample extracts. ........................................................................... 35 List of Figures Figure 1 RSSCT flow diagram with six columns in parallel. .................................................................. 14 Figure 2 Flow diagram of the large-scale GAC plant and pilot-scale SMC and IX plant for groundwater treatment. ............................................................................................................................................ 16 Figure 3 Pictures of the large-scale GAC plant during construction works in 2021 (left) and installation of the IXand SMC-pilot (right) in 2024. .............................................................................................. 16 Figure 4 Removals of DOC and abatements of UV254 in jar tests with spiked DW after 48 h equilibration time at 20 and 50 mg/L adsorbent dosage (dry weight). .................................................................... 20 Figure 5 LC-OCD-UVD chromatograms from jar tests for both adsorbent doses (20 and 50 mg/L) of selected adsorbents in direct comparison with the reference sample c0. .......................................... 21 Figure 6 Removals of OMP in test water from jar tests at adsorbent dosages of 20 and 50 mg/L (dry weight) after 48 h equilibration time. ................................................................................................. 24 Figure 7 OMP breakthroughs during RSSCT experiments with test water quantified by BV25, BV50 and BV95. ..................................................................................................................................................... 26 Figure 8 Normalized effluent concentration of Triflouromethanesulfonic acid (TFMSA). .................. 27 Figure 9 Normalized effluent concentration of 2,4-dimethylbenzene-1-sulfonic acid (DMBSA)........ 28 Figure 10 Usage rates based on number of bed volumes until 50 % breakthrough (UR50) derived with RSSCT data. .......................................................................................................................................... 29 Figure 11 PFAS influent and effluent data of the IX and SMC pilot and first GAC adsorber (B1) of largescale GW treatment plant. ................................................................................................................... 31 Figure 12 Comparison of normalized effluent UV254 absorbance and normalized effluent concentration of PFBA, PFOS and PFAS-20. ......................................................................................... 32 Figure 13 Typical PFAS CALUX concentration-response curves of selected water-extracts analysed by BDS. ..................................................................................................................................................... 34 Figure 14 Comparison of PFAS target analysis (original sample (‘target’) and rediluted extracts (‘target_eluate’)) for selected pilot plant and large-scale GW treatment plant samples ................... 36 Figure 15 PFOA-equivalents in the extracts for both sampling campaigns (1 and 2) at the large-scale GW treatment plant and the pilot plant.. ............................................................................................ 37 Figure 16 Cumulated intensity of substances detected in samples from the pilotand large-scale GW treatment plant analysed via database-assisted suspect screening and number of detected substances. ........................................................................................................................................... 38 Figure 17 PFAS compounds detected in samples from the pilotand large-scale GW treatment plant analysed via database-assisted suspect screening. ............................................................................. 39 D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 8 List of Abbreviations AA Alternative adsorbent AAMPS Sodium acryloyldimethyltaurate AC Activated carbon ACE Acesulfame ADONA Amino 2,2,3-trifluoro-3-[1,1,2,2,3,3-hexafluoro-3- (trifluoromethoxy)propoxy]propionate ATA Amantadine BAFG Bundesanstalt für Gewässerkunde BB Building blocks BDMA Benzyldimethylamine BDS BioDetection Systems BEA Beta zeolite BETMAC Benzyltrimethylazanium BEQ Biological PFOA equivalents BP Biopolymer BTA Benzotriazole BWB Berliner Wasserbetriebe CBM Carbamazepine CDP Cyclodextrin polymer CEQ Chemical PFOA equivalents CG Cyanoguanidine DCF Diclofenac DIOTOG Ditolylguanidine DMBSA 2,4-dimethylbenzene-1-sulfonic acid DOC Dissolved organic carbon DOM Dissolved organic matter DTA Diatrizoate DW Drinking water EBCT Empty bed contact time GAC Granular activated carbon GW Groundwater H4-PFOS 3,3,4,4,5,5,6,6,7,7,8,8,8-tridecafluorooctane-1-sulfonic acid HA Humic acids HFPO-DA Perfluoro-2-propoxypropanoic acid HPLC-MS/MS High-performance liquid chromatography with mass spectrometry HRMS High-resolution mass spectrometry IX Ion exchange resin LC-OCD Liquid size-exclusion chromatography with UV254 and online organic carbon detection LMWA Low molecular weight acids LMWN Low molecular weight neutrals LOQ Level of quantification MEL Melamine MPSA Sodium 2-methylprop-2-ene-1-sulfonate OMP Organic micropollutant D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 9 PFBA Perfluorobutyric acid PFBS Perfluorobutanesulfonic acid PFDA Perfluorodecanoic acid PFHpA Perfluoroheptanoic acid PFHxA Perfluorohexanoic acid PFHxS Perfluorohexanesulfonic acid PFHpS Perfluoroheptanesulfonic acid PFNA Perfluorononanoic acid PFOA Perfluorooctanoic acid PFOS Perfluorooctanesulfonic acid PFPeA Perfluoropentanoic acid PRI Primidone RSSCT Rapid small-scale column tests SAC Saccharine SF Sand filter SF Scaling factor SMC Surface modified clay SMX Sulfamethoxazole SOG Activated carbon surface oxygen groups TEQ Trigger PFOA equivalents TFMSA Triflouromethanesulfonic acid TR Thyroid hormone receptor TTR Transthyretin; transport protein for thyroxine (T4) TTR-TR CALUX Bioassay (PFAS CALUX) UBA Umweltbundesamt (German Environment Agency) VSA Valsartan acid D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 16 Figure 2 Flow diagram of the large-scale GAC plant and pilot-scale SMC and IX plant for groundwater treatment. Figure 3 Pictures of the large-scale GAC plant during construction in 2021 (left) and installation of the IXand SMC-pilot (right) in 2024. 2.4.2 Sampling campaign for PFAS CALUX and suspect screening Samples for suspect screening and PFAS CALUX were taken in February and March 2024 (sampling campaign 1 and 2) at the pilot and large-scale GW treatment plant using 1 L glas bottles. Samples of every influent and effluent of the pilot plant as well as the last GAC filter of the B line (GAC B4, Figure 2) were taken. An overview of all samples taken is given in Table 2. The samples were transported refrigerated to UBA on the same day and further processed there as described in section 2.5.3. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 17 The extracts obtained were analyzed using PFAS CALUX, suspect screening, and PFAS target analysis (‘target_eluate’). In addition, the original water samples (without additional enrichment) from all sampling points except B4 were also analyzed using PFAS target analysis (‘target’). Table 2 Overview of sampling IDs, times and points for suspect screening and PFAS CALUX. Sample ID Sampling date Description SFB_1 19/02/24 Effluent of rapid sand filters in line B; influent of IX1B and SMC1 SFB_2 12/03/24 IX1A_1 19/02/24 Effluent of IX column IX1A IX1A_2 12/03/24 IX1B_1 19/02/24 Effluent of IX column IX1B IX1B_2 12/03/24 SMC1_1 19/02/24 Effluent of SMC1 column SMC1_2 12/03/24 B2_1 19/02/24 Effluent of second GAC filter in line B (B2); influent of IX1A B2_2 12/03/24 B4_1 19/02/24 Effluent of last GAC filter in line B (B4) B4_2 12/03/24 The biological PFOA-equivalents derived using the PFAS CALUX (BEQ) were compared with PFOAequivalents derived by converting the PFAS target analysis results with relative potency factors based on PFAS CALUX (CEQ), as described by Behnisch et al. (2021), as well as relative potency factors based on animal data and in-silico modelling (TEQ), as proposed in the amendment of the EU Water Framework Directive COM EU 440 (2022). 2.5 Analytical methods 2.5.1 Target analyses of PFAS, iPMT, and other OMP All PFAS analyses during the jar test experiment (except for TFMSA analysis) were conducted using a high-performance liquid chromatography with mass spectrometry (HPLC-MS/MS) method described by Ingold et al. (2023). All PFAS samples obtained from RSSCT, pilotand large-scale adsorbent tests were analyzed by a HPLC-MS/MS method described by Zietzschmann et al. (2023). The other OMP (including TFMSA) were analyzed by a HPLC MS/MS method described by Zeeshan et al. (2023). 2.5.2 Quantification and characterization of DOM DOC was quantified with a TOC analyzer that uses a catalytic oxidation method (TOC-L, Shimadzu, Kyoto, Japan) in accordance with a standard method (DIN EN 1484). Ultraviolet (UV) absorption at 254 nm (UV254) was quantified with a Lambda 12 dual beam spectral photometer (Perkin-Elmer, USA) with 10 mm quartz cuvettes (Suprasil, Hellma, Germany). The DOM in all samples was characterized using size exclusion chromatography with UV254 and online organic carbon detections (LC-OCD-UVD model 8, DOC-Labor Huber, Germany). The method uses advanced oxidation/mineralization. A detailed description is given by Huber et al. (2011). The LC-OCD peaks were allocated as described by Haberkamp et al. (2007) and assigned to different DOM fractions: biopolymers (BP), humic acids (HA), building blocks (BB), low molecular weight acids (LMWA), and neutrals (LMWN). D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 18 2.5.3 PFAS CALUX bioassay The water samples were prepared by UBA with weak anion exchange solid phase extraction in accordance with the corresponding standard operating procedure (enrichment factor 10,000) developed by BDS for further testing by UBA and BDS (bioassays), BWB (target analysis), and BAFG (suspect screening). The extracts were diluted in DMSO for the bioassays and in methanol for the chemical analysis. The test strategy for toxicity testing of potentially PFAS-contaminated water samples was adapted from Deliverable 1.5 (Behnisch et al. 2021). The total PFAS content of the extracted water samples was determined using the TTR-TRβ CALUX by BDS and UBA. The CALUX bioassays are based on genetically modified cell lines, each of which has a receptor-specific response element and a dependent reporter gene for luciferase. When a substance binds to the receptor, luciferase is ultimately formed. Through addition of luciferin, the appropriate substrate for luciferase, light is emitted. The amount of light produced is proportional to the amount of ligand-specific receptor binding, which is benchmarked against the specific reference compounds. The results are expressed as bioanalytical equivalents (BEQs), i.e. the substances in the samples have an effect like the corresponding equivalent concentration of PFOA (as the reference substance in the TTR-TRβ CALUX). The TTR-TRβ CALUX bioassay is based on competition between PFAS and the thyroid hormone T4 for binding sites on transthyretin (TTR) transport proteins. In the TTR-binding assay, binding competition between a fixed concentration of T4 and a dilution series of sample extracts is studied. Increasing concentrations of relevant substances capable of competing with T4 for TTR-binding sites will result in a decreased amount of TTR-bound T4. Following separation of TTR-bound and unbound compounds (T4 and relevant substances), the amount of TTR-bound T4 is determined using the TRβ CALUX bioassay. An inhibition of binding of T4 to the TTR greater than 80% is considered an effect. On each 96 well plate, complete calibration curves for each respective bioassay were also analysed using the relevant reference compounds. Ultrapure water was used as negative control. 2.5.4 Suspect screening Water samples prepared by UBA (cf. compare section 2.5.3) were analysed via liquid chromatography coupled to mass spectrometry (LC-MS) in a database-assisted suspect-screening approach. A detailed description of the method is provided in Deliverable D1.4 (Dietrich et al. 2024). Samples were sent to BAFG by project partners and were stored in the refrigerator at 5°C until analysis. The methanolic extracts (enrichment factor 10,000, cf. section 2.5.3) were diluted 1:2000 in ultrapure water and spiked with a mix of three surrogate standards used for instrumental quality control at concentration levels of 1000 ng/L for bezafibrate-d4, 500 ng/L for olmesartan-d6, and 2000 ng/L for iopromide-d3. Aqueous samples from the pilot plant were filtered through 0.45 µm celluloseacetate-filters prior to shipping. At BAFG facilities, samples were spiked with surrogate standards, analogously to SPE extracts, without further dilution. For chromatographic separation, LC-MS grade acentonitrile was obtained from Merck (Darmstadt, Germany). Formic acid (LC-MS grade) was received from Sigma-Aldrich (Seelze, Germany). Ultrapure water was generated with a Milli-Q water purification system (Merck Millipore, Merck, Darmstadt, Germany). Surrogate standards bezafibrate-d4, olmesartan-d6, and iopromide-d3 were purchased from TRC (North York, Canada). Chromatographic separation was achieved using an Agilent 1260 infinity (Agilent Technologies, Waldbronn, Germany) system equipped with a Zorbax Eclipse Plus C18 column (2.1 x 150 mm, 3.5 µm, 95 Å; Agilent Technologies, Waldbronn, Germany) coupled to a Phenomenex Security GuardTM AQ C18 column (3 x 4 mm; Phenomenex, Aschaffenburg, Germany). D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 19 100 µL of sample were injected into the system. Ultrapure water with 0.1% formic acid (eluent A) and acetonitrile with 0.1% formic acid (eluent B) were used as mobile phases at a flow rate of 0.3 mL/min and a column temperature of 40°C, with the gradient shown in Table 3 . Table 3 Chromatographic gradient used for suspect screening of PFAS and iPMTs in the BAFG suspect screening method. Time Eluent A Eluent B 0 min 98% 2% 1 min 98% 2% 2 min 80% 20% 16.5 min 0% 100% 22 min 0% 100% 22.1 min 98% 2% 25 min 98% 2% Mass spectrometric analysis was performed with a TripleToF 6600 hybrid quadrupole time-of-flight spectrometer (Q-ToF-MS/MS; Sciex, Darmstadt, Germany) equipped with an electron spray ionization source, and operated in positive and negative ionization mode in two separate runs. Data acquisition was done by means of full scan experiments ranging from 100 to 1200 Da. Subsequently, MS2 spectra of the eight most intense peaks were recorded via information dependent acquisition (IDA) with a collision energy of 40 V and collision energy spread of 15 V. Acquisition time of full scan and the IDA experiments were 150 ms and 30 ms. For data processing, data received from the measurements were first transferred to the open data format mzXML using Proteo Wizard 3.0 (proteowizard.sourceforge.io). Peak picking was performed by an algorithm in R as described in Dietrich et al. (2022). Following peak picking, features detected in different samples were aligned based on their similarity in retention time and mass-to-charge ratio (m/z) (m/z-tolerance 20 ppm, retention time tolerance 20 s). Identification of suspects was achieved by matching high resolution mass, MS2 spectra, and retention time of aligned features with a collective spectral library (CSL). The CSL was fed with the respective information from analyses of authentic reference standards, mostly on the same instruments or by cooperating laboratories, and currently consists of about 1,400 substances (January 2025). D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 20 3 Results and discussion 3.1 Laboratory-scale adsorbent testing 3.1.1 Competitive adsorption of DOM The DOC and UV254 data from the jar test in Figure 4 show that the AC and, in particular, the IX, remove large proportions of the DOM, while the AA show substantially lower removals or even DOC leaching. Among the AC, highest DOC removals and UV254 abatements were seen with charcoal-based AC5, followed by lignite AC6, and very similar results were obtained for the different bituminous AC (AC14). The IX from different manufacturers showed only slight differences in the removals of DOC and abatements of UV254 with very low standard deviations (e.g. 2 % for DOC and 3 % for UV254 at 50 mg/L). This indicates very similar affinities towards DOM which could be attributed to similar product characteristics. PFAS-specific IX are less affected by DOM competition than conventional IX, removing up to 4-5 times less DOC, depending on IX dose and DOM composition (Dixit et al. 2021a). SMC4 showed an inhomogeneous pattern: while the highest DOC removal of all AA was found at an adsorbent dose of 50 mg/L (28 %), a two-digit negative removal was quantified at the lower adsorbent dose of 20 mg/L (-34 %). CDP also showed very negative removals of DOC and UV254 at both doses. Previous laboratory scale studies with CDP did not investigate the decline of DOC or UV254 (Wang et al. 2020; Ching et al. 2020), whereas a pilot-scale column study with CDP reported TOC removals below 10 % (Hwang et al. 2021). Low DOC and UV254 removals by BEA could be attributed to size exclusion of most DOM fractions, as suggested by Fu et al. (2022). Figure 4 Removals of DOC and abatements of UV254 in jar tests with spiked DW after 48 h equilibration time at 20 and 50 mg/L adsorbent dosage (dry weight). Reference sample c0: 4.5 mg/L DOC, 9.4 m-1 UV254 (Rückbeil et al. 2025, submitted). D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 21 Integration of peak areas within selected detection time ranges of the LC-OCD-UVD chromatograms (Figure 5) showed that the DOM in the test water consists of approximately 45 % HA, 29 % BB, 13 % LMWN, 13 % LMWA, and no BP. Removal by AC was highest for LMWA (on average 65 % at 50 mg/L), followed by the BB (on average 50 % at 50 mg/L). Previous studies revealed that adsorption competition between OMP and DOM on AC is dominated by LMWA and LMWN, especially the UV254 active part of LMWA and LMWN, which lower adsorption capacities compared to water with lower DOM content (Zietzschmann et al. 2016; Schumann et al. 2023). For IX, highest removals were observed for HA (on average 82 % at 50 mg/L) and BB (on average 51 % at 50 mg/L), therefore a high proportion of HA and BB appears to be particularly detrimental to the performance of the IX. These findings are also supported by Hu et al. (2014). Removals of HA, BB, and LMWA by IX is dominated by electrostatic interactions between positively charged quaternary ammonia groups of the IX and deprotonated acid groups of the DOM fractions, whereas the removal of LMWN can be attributed to physical adsorption processes (Cornelissen et al. 2008). It can be concluded that although the IX achieved the highest capacities for many PFAS among the adsorbents tested, especially short-chain compounds, it must be assumed that there is considerable competition from components of the DOM, and that these reduce the capacity of the IX, as discussed further in section 2.3.1. Figure 5 LC-OCD-UVD chromatograms from jar tests for both adsorbent doses (20 and 50 mg/L) of selected adsorbents in direct comparison with the reference sample c0. With biopolymers (BP), humic acids (HA), building blocks (BB), low molecular weight acids (LMWA), and low molecular weight neutrals (LMWN), UV: ultraviolet absorbance at 254 nm, OC: organic carbon (Rückbeil et al. 2025, submitted). D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 22 The LC-OCD-UVD chromatograms of samples after treatment with BEA, SMC1, SMC3, and SMC5 were so similar to the reference sample that no quantitative calculation of the removal of the individual organic carbon (OC) fractions was performed. The LC-OCD results were hereby consistent with the very low DOC and UV254 removals shown in Figure 4. From this it can be concluded that SMC removes PFAS particularly selectively and that competition for adsorption sites by DOM is negligible. Nevertheless, it is still possible that high DOM contents slow down the adsorption kinetics and reduce the capacity via pore blocking and fouling when operating a fixed bed filter. The LC-OCD-UVD chromatograms of samples after treatment with CDP showed elevated peaks, which overlap with the detection times of HA, BB, LMWA, and LMWN in the OC data series and with the detection times of HA, BB and LMWA in the UV254 data series. This is consistent with the high negative DOC removal and UV254 abatement of CDP shown in Figure 4. Elevated DOC concentrations after adsorbent dosage might be related to adsorbent leaching or impurities introduced during transport or storage. Further studies are recommended to investigate the increase of DOC and UV254 after addition of SMC2, SMC4, and CDP. 3.1.2 Removal of PFAS, iPMT and other OMP Jar test results The removal of all OMP included in the jar test experiment is shown in Figure 6. A clear correlation between perfluorinated carbon chain length and PFAS removal was observed across all adsorbents. Bituminous AC (AC1-AC4) outperformed charcoal-based AC (AC5) and lignite-based AC (AC6), especially for short-chain PFCA and HFPO-DA. Differences among bituminous AC were minimal, suggesting similar AC properties. The only macroporous IX (IX1) was the most effective at removing short-chain PFCA. All IX removed PFCA with more than six perfluorinated carbon atoms completely at both doses, making it difficult to assess performance differences. IX were generally more efficient for long-chain PFCA than the tested AC and AA adsorbents. IX also excelled at removing HFPO-DA, with only slight differences between products (5% standard deviation at 50 mg/L). The two bentonite-based products (SMC1 and SMC2) showed similar or better PFAS removal than AC. In contrast, the three palygorskite-based SMC products (SMC3, SMC4, and SMC5) were ineffective, with SMC3 and SMC5 showing no removal of short-chain PFAS, though some long-chain PFAS removal was observed. SMC4 achieved 20% removal of short-chain PFAS, such as PFPeA, at 50 mg/L. BEA and CDP showed poor PFBA removal, and only low removals for short-chain PFCA like PFPeA and PFHxA (<8% at 50 mg/L for BEA). CDP performed similarly to AC for these substances, but neither adsorbent removed HFPO-DA as effectively as AC or IX. Previous studies have shown that bituminous AC can achieve higher PFAS loadings than other AC types, likely due to its favorable pore distribution for PFAS molecular sizes (Cantoni et al. 2021). Hydrophobic interactions and electrostatic forces are the primary mechanisms for PFAS removal by AC (Grieco et al. 2021; Gagliano et al. 2020). Unlike AC and most AA, all IX were effective at removing TFMSA, with macroporous IX1 showing particular efficiency. Other PFSA were 100% removed by all IX at 20 mg/L. IX1 differs from other IX primarily in its macroporous structure, which may provide easier access for PFAS to active sites D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 23 compared to the gelular IX. Other studies support this influence of IX structure (Boyer et al. 2021; Dixit et al. 2021b). Gellular IX products from different manufacturers showed similar performance for all OMP, suggesting that manufacturer choice is less important. All PFAS-specific IX adsorbents rely on electrostatic interactions between PFAS and quaternary ammonia groups, as well as hydrophobic interactions with the PFAS alkyl chain and, for long-chain PFAS, the IX polystyrene backbone. These mechanisms explain the better retention of long-chain PFAS and the difficulty of regenerating PFASspecific resins via conventional brine regeneration (Dixit et al., 2019). Surface modification with cationic surfactants, like aliphatic amines, enhances the hydrophobicity and anion exchange capacity of SMC (Jiang et al. 2023; Mukhopadhyay et al. 2021). The removal of anionic PFAS is then primarily driven by hydrophobic interactions, though electrostatic interactions and ligand exchange can also play a role (Jiang et al. 2023; Mukhopadhyay et al. 2021; Liu et al. 2024). Despite being grouped as SMC, SMC1, SMC2, SMC3, and SMC5 show significant differences in PFAS removal effectiveness. These variations might stem from differences in the clay mineral properties (e.g., fibrous vs. layered) or the surface functionalization. Manufacturers provided no details on the functionalization process. Literature suggests that PFSA are more easily removed than PFCA with similar perfluorinated C-chain lengths (Plumlee et al. 2022; Cantoni et al. 2021; Dixit et al. 2019; Jiang et al. 2023; Ching et al. 2020). This may be due to the stronger acidity of sulfonic acids, which enhances interactions with amines. Additionally, the presence of an ether group (as in HFPO-DA) reduces removability compared to PFOA, likely because it interferes with hydrophobic interactions. This study confirms several findings from a large US pilot column study on PFAS removal from GW with varying DOC levels (0.3 to 2.3 mg/L, pH 7.3 – 8.3): 1) bituminous AC outperforms AC from other raw materials; 2) IX and SMC1 achieve higher PFAS loadings than AC; 3) gellular IX from different manufacturers perform similarly; and 4) CDP is less effective at removing PFAS? than AC, IX, or SMC (Plumlee et al. 2022; Pannu et al. 2023; Hwang et al. 2021). Hwang et al. (2021) found a negative correlation between days until PFAS breakthrough and DOC concentration for AC, while SMC1 and CDP performed similarly across DOC levels. The DOC impact on AC is significant, as shown by isotherm studies in DOC-free water (Olhansky et al. 2022). Olshansky et al. (2022) conducted isotherm studies with AC2 in deionized water, finding Kd values up to 2000 times higher than in this study. Most other OMP were primarily removed by AC, especially those with aromatic rings such as BDMA, BETMAC, BTA, and CBM. Polar compounds like CG and MPSA were poorly removed by AC. IX removed BTA, but less effectively than AC. BDMA and BETMAC were completely removed by BEA and SMC4 (100% at 50 mg/L), while the other adsorbents had no effect. SMC4 was particularly effective for BTA removal (100% at 20 mg/L). CDP was the only adsorbent that removed CG (71% at 50 mg/L). DIOTOG was removed by SMC4 and BEA (100%), while MPSA was not removed by any adsorbent. IX effectively removed anionic pharmaceuticals and sweeteners (DCF, SMX, VSA, ACE, SAC), but not neutral or positively charged compounds. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 24 Figure 6 Removals of OMP in test water from jar tests at adsorbent dosages of 20 and 50 mg/L (dry weight) after 48 h equilibration time. Reference concentrations are listed in Table SI 2 (Rückbeil et al. 2025, submitted). RSSCT results An overview of all BV25, BV50, and BV95 values derived during the RSSCT experiment is given in Figure 7. The breakthrough curves with fitted model curves for six adsorbents for one PFAS (TFMSA) and one iPMT (DMBSA) are also shown in Figure 8 and Figure 9 as examples for data evaluation. All data sets in which a breakthrough could be quantified (normalized effluent concentration > 0.05) were accurately described by the applied S-curve model (coefficient of determination R2 always > 0.59, mean R2 = 0.91). The results of the RSSCT confirm important findings of the jar tests: short-chain PFAS are generally more difficult to remove than long-chain PFAS, and sulfonic acids are better removed than carboxylic acids. AC cannot be used to achieve economic runtimes for short-chain PFAS removal such as PFBA and TFMSA, as these substances achieved 95 % breakthrough with all three AC after just a few thousand BV. The highest throughputs until breakthrough of PFBA and TFMSA were achieved with IX1, but other IX or SMC may also be viable alternatives to AC to remove these short-chain compounds. While for all PFSA with at least two perfluorinated C atoms no breakthrough within the first 100,000 BV for all IX and SMC tested was seen, SMC1 and SMC2 proved to be the best adsorbents for removing long-chain PFCA. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 25 Emerging PFAS, such as the tested PFECA, HFPO-DA, and ADONA, and the polyfluorinated substance H4-PFOS, are more difficult to remove than the now banned substances PFOA and PFOS they substitute. This illustrates that a ban on a single perfluorinated substance is insufficient as long as it is simply replaced by a very similar but not banned substance. On the contrary, the substitutes bring additional uncertainties, as no information is available on their removal, and analytical methods must first be developed to be able to quantify them in environmental samples (Evich et al. 2022; Hale et al. 2020a, 2020b). The UR50 derived with the RSSCT are visualized in Figure 10, where a dashed horizontal line indicates a UR50 of 25 mg/L to identify strongly adsorbing compounds, as proposed by Kempisty et al. (2022). This limit for a feasible UR50 should not be taken as an absolute value to define feasabilty of GAC treatment, but as a practice-oriented size range derived from discussions among water professionals to avoid excessive GAC change out during filter operation (Kempisty et al., 2020). Based on the respective bed densities of the RSSCT, a UR of 25 mg/L refers to 9,000-11,000 bed volumes of IX treatment, 13,000-15,000 BV of GAC treatment, and 13,000-20,000 BV of SMC treatment. It can be seen that feasible UR50 can be achieved for PFSA with 5 or more pefluorinated C-atoms as well as PFCA with 6 or more perfluorinated C-atoms. The IX and SMC that were tested were able to achieve significantly lower UR for all PFAS. For the iPMT and other OMP, a large part of the UR50 could only be determined for the GAC, as the removals by IX and SMC were too small, which once again emphasizes the broader removal effect of the GAC. ATA, AAMPS, and DTA are examples of OMP for which no feasible UR could be achieved with any adsorbent under the experimental conditions described. These substances should be classified as particularly critical for adsorptive water treatment. In Figure 10, the UR50 classification for GAC has been used for all adsorbents, which enables direct comparability with reference to the same adsorbent mass. However, it could just as well make sense to adjust the limit for a feasible UR50 depending on the adsorbent price. This means that if, for example, the price of an IX is 5 times higher, then a feasible UR50 for the IX would have to be 5 times lower than that of the GAC in order to be competitive. However, as the prices for adsorbents fluctuate and are subject to uncertainty, this evaluation has been omitted here. Specific adsorbent costs are discussed in more detail in section 3.2.1. In summary, IX and SMC in particular remove PFAS much more selectively than AC. The macroporous IX1 proved to be the best adsorbent for short-chain PFAS, while SMC1 was the best adsorbent for long-chain PFAS. However, as already indicated in the jar test, AC was the only adsorbent group investigated that also achieved a broader removal of other OMP, such as MEL or DPG. The choice of the best adsorptive treatment option is therefore strongly dependent on the composition of the OMP and the treatment objective, and should be decided individually according to site specific conditions. While AC and in particular IX also showed high removals of DOM, no competition for adsorption sites by DOM could be determined for the SMC. Based on the laboratory results presented, IX1 and SMC1 were selected for subsequent piloting, with IX1 showing highest removal for short-chain PFAS and SMC being especially PFAS-selective. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 32 The normalized effluent UV254 absorbance is compared with the normalized effluent concentration of PFAS-20, PFBA, and PFOS in Figure 12. Previous studies have already shown that for iPMT and other OMP, correlations with the removal of UV254 can be observed during removal by AC (Schumann et al. 2023; Jekel and Zietzschmann 2018; Altmann et al. 2016). The data from the pilotand largescale GW treatment plant were used to make similar estimates for the removal of PFAS by GAC, IX, and SMC. With the exception of two outliers, a good correlation was observed with GAC between normalized UV254 absorbance, PFOS concentration, and PFAS-20, but not for the poorly removable PFBA compound. With the IX, a correlation between normalized PFBA concentration and normalized UV254 absorbance was also evident. With the SMC, no correlation could be observed due to the very low DOM removal. In practical application, the siteand water quality-specific correlation between target PFAS and UV254 removal needs to be established first. Afterwards, UV254 real time monitoring can provide a fast and cost-effective monitoring method to estimate PFAS breakthrough in GAC and IX treatment. However, it cannot be recommended as an option for monitoring PFAS breakthrough in SMC adsorbers. Figure 12 Comparison of normalized effluent UV254 absorbance and normalized effluent concentration of PFBA, PFOS, and PFAS-20. Very different data on the adsorbent costs of GAC, IX, and SMC were found in literature (Riegel et al. 2023; Barr Engineering Co. and Hazen and Sawyer 2023). The numbers given in Table 4 therefore only represent an initial range and should only be used with a safety margin in further estimates. Additionally, it should be noted that other operation and maintenance costs like reactivation (GAC) or disposal or incineration (IX, SMC), as well as investment costs, are not considered. Assuming IX is 2-6 times more expensive per m3 than GAC, it would need to treat a correspondingly higher number of BV in order to be more economical than GAC. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 33 This would be the case for PFAS-4 and PFAS-20 at the investigated pilotand large-scale GW treatment plant. But SMC may be more cost effective for use on site. However, the high selectivity of the SMC also means that other iPMT or pharmaceuticals are hardly or not at all removed, as our laboratory tests have shown. The use of SMC is therefore limited to waters in which PFAS are the primary compound to be removed. Table 4 Specific adsorbent media costs. Adsorbent Specific adsorbent media costs (€/m3) References GAC 2,000-4,300 Riegel et al. (2023); Barr Engineering Co. and Hazen and Sawyer (2023) IX 8,000-13,500 Riegel et al. (2023); Barr Engineering Co. and Hazen and Sawyer (2023) SMC 5,500-7,600 Barr Engineering Co. and Hazen and Sawyer (2023) For a broader removal of iPMT and other OMP from DOC-rich source water, processes with GAC or a combination of GAC and IX are particularly powerful, as our data show. However, due to the high specific adsorbent costs of IX, a precise cost analysis needs be carried out based on site-specific conditions (like treatment target, DOM composition) and the costs and benefits compared. No general recommendation can be made for this treatment train. A potentially more cost-effective combination of GAC and SMC could also be interesting for subsequent studies. However, when applied in the DW sector, national regulations must also be considered in regards to which adsorbents are actually permitted. In Germany, for example, this would currently only be GAC, as defined by the annex of the German drinking water ordinance (TrinkwV 2023). Modeling tools such as the Planning & design tool for drinking water treatment for PFAS & industrial chemicals developed in PROMISCES can be a cost-effective and fast way to avoid more expensive and lengthy piloting of additional treatment trains (Rückbeil et al., 2024). The disadvantage of modeling solutions is always the high level of uncertainty and the lack of validation with real data. For this reason, it makes sense to combine piloting with modelling as recommended by Burkhart et al. (2022). The model can be calibrated and validated with existing data and then used to predict new scenarios. In this way, both methods can complement each other and form a flexible and validated decisionmaking tool (Burkhardt et al. 2022). 3.2.2 PFAS CALUX bioassay results In Table 5, the TTR-TRβ CALUX (i.e. PFAS CALUX) results for the received groundwater sample extracts are summarised. For all samples, the PFAS CALUX results are expressed as the concentration of reference compound equivalents (µg PFOA-equivalents) per liter of water samples. Typical graphical representation of concentration-response curves in the TTR-TRβ CALUX bioassay from GW samples (analysed by BDS) are shown in Figure 13. In general, the results in Table 5 show good agreement between the independent testing done at UBA and BDS. The majority of the samples showed no effects above the LOQ. There are deviations in four samples, with samples B4_1 and SMC1_1 being only marginally above the LOQ at BDS. SFB_1 (2.5 µg PFOA eq/L) and SFB_2 (2.7 µg PFOA eq/L) show increased activity in the PFAS CALUX in BDS. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 34 In the SFB_2 sample, UBA found no effects, while for the SFB_1 sample, a concentration dependent effect was seen. This effect did not fall below the 80 % threshold of inhibition, so no activity could be calculated. However, the use of higher concentrations was not possible at UBA because of the cytotoxic effects of DMSO. The results in the bioassays match the processing stages well. Effects after the first treatment step of sand filtration were to be expected. The following additional treatment steps reduce the effects, i.e. the substances contained, below the LOQ. In principle, the results also fit well with the results of chemical analysis (see Figure 15). The slight deviations in results obtained by UBA and BDS (Table 5) affect the following water samples: BDS recorded four samples above the LOQ (0.99). B4_1 and SMC_1 indicated activities of 1.0 µg PFOA-eq/L, which is marginally above the LOQ. The two SFB samples were slightly higher at 2.5 and 2.7. In the SFB_2 sample, UBA found no effects, while for the SFB_1 sample, a concentrationdependent effect was seen. This effect did not fall below the 80 % threshold of inhibition, so no activity could be calculated. However, the use of higher concentrations was not possible at UBA because of the cytotoxic effects of DMSO. PFAS CALUX (BDS lab-results) 10 - 1 0 10 - 8 10 - 6 10 - 4 0 20 40 60 80 100 120 10 0 10 1 10 2 10 3 PFOA Ref SFB_1 (BDS) Reinstw asser 1 (BDS) Reinstw asser 2 (BDS) Reinstw asser 3 (BDS) SFB_2 (BDS) Reinstw asser 4 (BDS) Reinstw asser 5 (BDS) Reinstw asser 6 (BDS) mol PFOA/l incubate l sample/l incubate Relative induction (% of max) Figure 13 Typical PFAS CALUX concentration-response curves of selected water-extracts analysed by BDS. Inhibition of the binding of T4 to the TTR protein is expressed as induction relative to the maximum induction of the PFOA reference series (relative induction RI%). “Reinstwasser”: ultrapure water A comparison of target PFOA equivalents based on different conversion methods is shown in Figure 14. It can be observed that the BEQ values obtained by BDS (based on PFAS CALUX) are above the CEQ and TEQ (calculated by BWB) for samples SFB_1, SMC1_1 and B4_1 (sampling campaign 1) as well as SFB_2 (sampling campaign 2). This could indicate the presence of unknown PFAS compounds in the inflow of the pilotand large-scale plant or measurement uncertainties. The BEQ values of UBA align with all CEQ and TEQ values, indicating PFOA equivalents below 0.99 µg/L. As expected, the PFOA equivalents are reduced by adsorptive treatment. Both SFB samples (which represent the first effluent after the sand filters) show the highest PFOA equivalents. The samples SMC1_1 and B4_1 showed elevated PFOA equivalents in the PFAS CALUX test method. A comparison with the target analysis results of these samples (original sample and extracts in Figure 14) shows that there is an increased PFPeA and PFHxA concentration in the extracts which was not found in the original sample. This could indicate impurities during concentration or sample storage. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 35 Table 5 PFAS CALUX results of GW sample extracts. * UBA tested PFOA and ultrapure water only in one concentration as an internal control, therefore no activity could be calculated. Sample TTR - TRβ CALUX (UBA) TTR - TRβ CALUX (BDS) Activity LOQ Activity LOQ (µg PFOA-equivalents/L sample) (µg PFOA-equivalents/L sample) B4_1 <LOQ 0.99 1.0 0.99 SFB_1 <LOQ 0.99 2.5 0.99 IX1_1 <LOQ 0.99 <LOQ 0.99 IX2_1 <LOQ 0.99 <LOQ 0.99 SMC_1 <LOQ 0.99 1.0 0.99 B2_1 <LOQ 0.99 < LOQ 0.99 B4_2 <LOQ 0.99 <LOQ 0.99 SFB_2 <LOQ 0.99 2.7 0.99 IX1_2 <LOQ 0.99 <LOQ 0.99 IX2_2 <LOQ 0.99 <LOQ 0.99 SMC_2 <LOQ 0.99 <LOQ 0.99 B2_2 <LOQ 0.99 <LOQ 0.99 PFOA 1 - * - 1.1 0.99 PFOA 2 - * - 1.9 0.99 PFOA 3 - * - 1.3 0.99 PFOA 4 - * - <LOQ 0.99 PFOA 5 - * - 1.8 0.99 PFOA 6 - * - <LOQ 0.99 Ultrapure water 1 -* - <LOQ 0.99 Ultrapure water 2 -* - <LOQ 0.99 Ultrapure water 3 -* - <LOQ 0.99 Ultrapure water 4 -* - <LOQ 0.99 Ultrapure water 5 -* - <LOQ 0.99 Ultrapure water 6 -* - <LOQ 0.99 D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 36 Figure 14 Comparison of PFAS target analysis (original sample (‘target’) and rediluted extracts (‘target_eluate’)) for selected pilot plant and large-scale GW treatment plant samples from the sampling campaign for PFAS CALUX and suspect screening. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 37 Figure 15 PFOA-equivalents in the extracts for both sampling campaigns (1 and 2) at the large-scale GW treatment plant and the pilot plant. PFOA-equivalent calculation based on PFAS CALUX (BEQ, LOQ = 0.99 µg/L) or conversion of PFAS target analysis using relative potency factors based on PFAS CALUX (CEQ, LOQ = 0.003 µg/L) or animal data (TEQ, LOQ = 0.003 µg/L). Values below LOQ are displayed as 0 PFOA equivalents. 3.2.3 Suspect screening Database assisted suspect screening of samples from the pilotand large-scale GW treatment plant resulted in the identification of four perfluorinated compounds, four industrial chemicals, ten pharmaceuticals and metabolites thereof, and six substances mainly used in personal care products and/or as household chemicals. The latter case is not easily distinguishable from the group of industrial chemicals, as the same products might as well be used in industrial applications, such as laundries. An overview of the detected substances is provided in Figure 16. It should be noted that the applied screening technique does not provide absolute concentrations. Therefore, cumulative intensities provided in Figure 16 must not be interpreted as a sum of concentrations. Different compounds of the same concentrations might appear at different intensities. However, intensity trends within a single compound may provide insight into its actual concentration trend. During GW treatment plant operation, intensities of three of the four PFAS detected in the effluent of the rapid sand filters (SFB_1 and SFB_2) and effluent of the second GAC filter (B2_1 and B2_2) (i.e. PFPrS, PFHxS, and PFHpS) are reduced and could not be detected after SMC and IX treatment (pilot plant) (Figure 17). Surprisingly, the intensity of the PFHxA was found to be significantly increased in one sample after SMC treatment (SMC1_1). This finding was supported by target analysis showing elevated PFHxA concentrations in the same extract sample. However, since target analysis did not detect PFHxA above 1 ng/L in the original sample, it may be an impurity introduced during sample storage or preparation (compare PFAS target analysis results for extract and original sample in Figure 14). The lowest number of detected substances was found after treatment with four GAC filters in series in the large-scale GW treatment plant (sample B4_1 and B4_2), or after a combination of GAC and IX (sample IX1A_1 and IX1A_2) in the pilot setup. This is supported by findings from the laboratory scale adsorbent screening (compare section 3.1.2), revealing that broad removal of iPMT D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 38 is best achieved with AC (or treatment trains combined with AC). The SMC samples showed intensities of dimethyl-ditetradecylammonium that did not appear in other samples. Figure 16 Cumulated intensity of substances (stacked columns, left y-axis) detected in samples from the pilotand large-scale GW treatment plant analysed via database-assisted suspect screening and number of detected substances (black points, right y-axis). Samples were extracted with methanol (1:10,000) and rediluted in ultrapure water (1:2000). Intensities must not be interpreted as concentrations: different compounds of the same concentration might appear at different intensities. Color code: yellow: industrial chemicals, red: PFAS compounds, pink: pharmaceuticals, blue: household and personal care. Grey arrows: flow path of the largescale GAC plant and pilot-scale SMC and IX plant. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 39 Figure 17 PFAS compounds detected in samples from the pilotand large-scale GW treatment plant analysed via database-assisted suspect screening. Samples were extracted with methanol (1:10,000) and rediluted in ultrapure water (1:2000). Intensities must not be interpreted as concentrations; different compounds of the same concentration might appear at different intensities. Grey arrows: flow path of the large-scale GAC plant and pilot-scale SMC and IX plant. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 40 4 Conclusion To improve knowledge on the removal of PFAS, iPMT, and other OMP from waters with high DOM concentrations (4-5 mg/L DOC), the performance of IX and high-performance adsorbent media specifically tailored to remove PFAS was tested and compared to AC. A broad screening of available tailored adsorbent products included six AC, five SMC, one CDP, and one BEA zeolite. Laboratoryscale equilibrium jar tests showed that PFAS-specific IX and SMC can be relevant alternatives to GAC for PFAS removal. Dynamic breakthrough behavior of PFAS was studied in RSSCTs, which confirmed jar test results. It was shown that GAC remains the benchmark adsorbent for removing a wide range of iPMT, although it exhibits limited performance in the removal of short-chain PFAS. Strong-base IX were identified as highly effective for PFAS removal, yet their efficiency was significantly compromised by the presence of competing DOM. Bentonite based SMC demonstrated promising potential for PFAS removal, particularly through enhanced selectivity for longer-chain PFAS and compounds with sulfonate groups, and offers a viable option for targeting specific PFAS compounds in complex water matrices. Pilot-scale column testing of IX and SMC confirmed the laboratory findings and enabled direct comparison with GAC performance in a large-scale plant operated in parallel. These results indicate that treatment trains combining GAC with IX or SMC may offer a more efficient solution for PFAS and iPMT removal in the presence of DOM. However, the choice of the most efficient treatment train is highly dependent on the composition of the PFAS and iPMT, the water matrix, the treatment objective, and economic considerations (e.g. adsorbent media costs). Bioassays with the PFAS CALUX test (TTR-TRβ CALUX) showed no effects above the LOQ in the majority of samples, particularly after adsorption treatment. Database-assisted suspect screening of samples from the pilot-scale IX/SMC and large-scale GAC columns confirmed PFAS removal performance determined via target analysis. The integration of PFAS CALUX tests and suspect target analysis allowed for a more comprehensive assessment of the adsorptive processes, highlighting the importance of detecting unforeseen contaminants and evaluating the effectiveness of adsorbents beyond targeted PFAS. Overall, this study advances the understanding of competitive adsorption and ion exchange of PFAS and iPMT, and provides valuable insights into the potential of alternative absorbents. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 41 References Altmann, Johannes; Massa, Lukas; Sperlich, Alexander; Gnirss, Regina; Jekel, Martin (2016): UV254 absorbance as real-time monitoring and control parameter for micropollutant removal in advanced wastewater treatment with powdered activated carbon. In: Water research 94, S. 240–245. DOI: 10.1016/j.watres.2016.03.001. Arp, Hans Peter H.; Hale, Sarah E. (2022): Assessing the Persistence and Mobility of Organic Substances to Protect Freshwater Resources. In: ACS environmental Au 2 (6), S. 482–509. DOI: 10.1021/acsenvironau.2c00024. Barr Engineering Co.; Hazen; Sawyer (2023): Evaluation of Current Alternatives and Estimated Cost Curves for PFAS Removal and Destruction from Municipal Wastewater, Biosolids, Landfill Leachate, and Compost Contact Water. Hg. v. Minnesota Pollution Control Agency. Available at: https://www.apwa.org/wp-content/uploads/Evaluation-of-Current-Alternatives-andEstimated-Cost-Curves-for-PFAS-Removal-and-Destruction-from-Municipal-WastewaterBiosolids-Landfill-Leachate-and-Compost-Contact-Water.pdf, last access 31.01.2025. Behnisch, Peter A.; Besselink, Harrie; Weber, Roland; Willand, Wolfram; Huang, Jun; Brouwer, Abraham (2021): Developing potency factors for thyroid hormone disruption by PFASs using TTR-TRβ CALUX® bioassay and assessment of PFASs mixtures in technical products. In: Environment international 157, S. 106791. DOI: 10.1016/j.envint.2021.106791. Behnisch, Peter; Sosnowska, Anita; Mombelli, Enrico; Kuckelkorn, Jochen (2021): D1.5 – Set of novel QSAR models/grouping/read-across and in vitro bioassay approaches predicting relevant toxicological endpoints for PFAS/iPM(T) chemicals. Online verfügbar unter https://ec.europa.eu/research/participants/documents/downloadPublic?documentIds=080 166e511e8a9eb&appId=PPGMS. Boyer, Treavor H.; Fang, Yida; Ellis, Anderson; Dietz, Rebecca; Choi, Youn Jeong; Schaefer, Charles E. et al. (2021): Anion exchange resin removal of perand polyfluoroalkyl substances (PFAS) from impacted water. A critical review. In: Water research 200, S. 117244. DOI: 10.1016/j.watres.2021.117244. Brunn, Hubertus; Arnold, Gottfried; Körner, Wolfgang; Rippen, Gerd; Steinhäuser, Klaus Günter; Valentin, Ingo (2023): PFAS. Forever chemicals—persistent, bioaccumulative and mobile. Reviewing the status and the need for their phase out and remediation of contaminated sites. In: Environ Sci Eur 35 (1), S. 242. DOI: 10.1186/s12302-023-00721-8. Burkhardt, Jonathan B.; Burns, Nick; Mobley, Dustin; Pressman, Jonathan G.; Magnuson, Matthew L.; Speth, Thomas F. (2022): Modeling PFAS Removal Using Granular Activated Carbon for FullScale System Design. In: J. Environ. Eng. 148 (3), S. 1–11. DOI: 10.1061/(asce)ee.19437870.0001964. Cantoni, Beatrice; Turolla, Andrea; Wellmitz, Jörg; Ruhl, Aki S.; Antonelli, Manuela (2021): Perfluoroalkyl substances (PFAS) adsorption in drinking water by granular activated carbon. Influence of activated carbon and PFAS characteristics. In: The Science of the total environment 795, S. 148821. DOI: 10.1016/j.scitotenv.2021.148821. Ching, Casey; Klemes, Max J.; Trang, Brittany; Dichtel, William R.; Helbling, Damian E. (2020): βCyclodextrin Polymers with Different Cross-Linkers and Ion-Exchange Resins Exhibit Variable D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 48 Zeng, Chao; Atkinson, Ariel; Sharma, Naushita; Ashani, Harsh; Hjelmstad, Annika; Venkatesh, Krishishvar; Westerhoff, Paul (2020): Removing perand polyfluoroalkyl substances from groundwaters using activated carbon and ion exchange resin packed columns. In: AWWA Water Science 2 (1). DOI: 10.1002/aws2.1172. Zhang, Dongqing; He, Qiaochong; Wang, Mo; Zhang, Weilan; Liang, Yanna (2021): Sorption of perfluoroalkylated substances (PFASs) onto granular activated carbon and biochar. In: Environmental technology 42 (12), S. 1798–1809. DOI: 10.1080/09593330.2019.1680744. Zietzschmann, Frederik; Hensel, Tobias; Anne, Togola; Sebastien, Bristeau; Alda, Miren Lopez de; Llorca, Marta et al. (2023): Deliverable D1.1 - Methods for PFAS in waters and complex matrices. Online verfügbar unter https://doi.org/10.5281/zenodo.14332693. Zietzschmann, Frederik; Stützer, Christian; Jekel, Martin (2016): Granular activated carbon adsorption of organic micro-pollutants in drinking water and treated wastewater--Aligning breakthrough curves and capacities. In: Water research 92, S. 180–187. DOI: 10.1016/j.watres.2016.01.056. D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 49 Annex Table SI 1 Information on jar test water (DOM, anions, hardness). pH UV254 DOC SO42ClNO3o-PO43FlKS4.3 KB8.2 hardne ss (CO32-) hardness (Ca2++ Mg2+) 1/m in mg/ L in mg/L in mg/ L in mg/L in µg/L in mg/ L in mmol/L in mmol/L in °dH in °dH 7.5 9.8 4.4 89.0 56. 0 2.7 18.0 0.2 4.02 0.25 11.1 16.4 Table SI 2 OMP reference concentrations in the test water of the jar test experiment. Parameter c0 unit UV254 9.4 1/m DOC 4.5 mg/L PFBA 9870.0 ng/L PFPeA 9599.2 ng/L PFBS 873.3 ng/L PFHxA 7022.5 ng/L GenX 7768.3 ng/L PFHpA 686.7 ng/L PFHxS 1062.5 ng/L PFOA 891.7 ng/L PFNA 936.7 ng/L PFOS 528.3 ng/L PFDA 285.8 ng/L CG 883.1 ng/L BTA 968.5 ng/L Parameter c0 unit BDMA 1535.8 ng/L BETMAC 1779.2 ng/L ATA 1350.8 ng/L PRI 1440.0 ng/L DIOTOG 1233.3 ng/L DTA 1073.4 ng/L VSA 1337.5 ng/L CBM 1386.7 ng/L DCF 1377.5 ng/L SMX 1142.9 ng/L MPSA 1008.3 ng/L TFMSA 748.7 ng/L ACE 1577.5 ng/L SAC 1035.3 ng/L AAMPS 1031.8 ng/L Table SI 3 Information on RSSCT test water (DOM, anions, hardness). pH UV254 DOC SO42ClNO3o-PO43FlKS4.3 KB8.2 hardne ss (CO32-) hardness (Ca2++ Mg2+) 1/m in mg/ L in mg/L in mg/ L in mg/L in µg/L in mg/ L in mmol/L in mmol/L in °dH in °dH 7.5 10.0 4.5 99.0 54. 0 2.6 17.0 0.2 4.3 0.22 11.0 16.3 D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 50 Table SI 4 OMP influent concentrations (c0) of the test water during RSSCT experiments in ng/L. AC IX SMC AAMPS 1098 1050 1063 ACE 1562 1480 1457 ADONA 1308 1312 1427 ATA 1280 1355 1307 BDMA 2517 2660 4457 BETMAC 5098 3455 3695 CBM 1305 1318 1393 DCF 1407 1293 1447 DIOTOG 1267 1400 1350 DMBSA 2258 2405 2722 DPG 717 945 1132 DTA 854 960 873 H4-PFOS 663 720 444 HFPO-DA 982 989 840 MEL 930 895 1130 MPSA 998 929 975 PFBA 1781 2000 1523 PFBS 1252 1258 1113 PFDA 405 587 934 PFHpA 1203 1160 975 PFHxA 1297 1200 1028 PFHxS 1587 1141 1222 PFNA 882 734 1120 PFOA 1288 917 1309 PFOS 626 606 1046 PFPeA 1275 1320 1091 PFPeS 1303 1059 1159 PRI 1250 1352 1323 SAC 1016 899 1011 SMX 1305 1283 1307 TFMSA 780 947 772 VSA 1265 1215 1282 D4.1 –Performance of drinking water treatment trains for removal of PFAS and industrial chemicals 51 Table SI 5 Mean influent concentrations (PFAS, inorganic anions, dissolved organic carbon) during full-scale GAC operation (prior to first GAC exchange) and SMC and IX pilot operation. Only PFAS above LOQ shown. *included in sum parameter PFAS-20, °included in sum parameter PFAS-4. parameter GAC7 (B1) IX1A, SMC1 IX1B unit PFAS-20 582.1 556.5 136.1 ng/L PFAS-4 503.9 468.3 94.8 ng/L PFBA 4.8* 6.4* 6.5* ng/L PFBS 12.8* 19.0* 9.0* ng/L PFHpA 4.1* 4.2* 2.2* ng/L PFHpS 12.4* 10.5* 1.2* ng/L PFHxA 24.6*° 28.4*° 20.2*° ng/L PFHxS 202.4*° 207.0*° 45.0*° ng/L PFOA 19.9* 19.7* 6.7* ng/L PFOS 280.8*° 232.2*° 29.0*° ng/L PFPeA 6.8* 7.9* 7.2* ng/L PFPeS 15.2* 17.6* 5.1* ng/L Cl46.0 46.0 mg/L Fl0.2 0.2 mg/L HCO3 - 251.3 251.1 mg/L NO3 - 1.5 1.5 mg/L SO4 2 - 114.4 115.5 mg/L PO4 3 - 23.0 22.6 µg/L DOC 4.3 4.3 3.7 mg/L UV254 8.2 9.1 7.0 1/m