In situ calibration of a tube passive sampler in wastewater effluent with adjustable volumetric flow for assessment of micro-pollutants with fluctuating concentrations

In situ calibration of a tube passive sampler in wastewater 1 effluent with adjustable volumetric flow for assessment of 2 micro-pollutants with fluctuating concentrations 3 Tobias Hensela, Jörg-Helge Heinb, Thorsten Reemtsmac, Alexander Sperlicha, 4 Regina Gnirßa, Frederik Zietzschmanna* 5 aBerliner Wasserbetriebe, Neue Jüdenstraße 1, 10179 Berlin, Germany 6 bGCI GmbH, Bahnhofstraße 19, 15711 Königs Wusterhausen, Germany 7 cHelmholtz-Centre for Environmental Research (UFZ), Department of Analytical 8 Chemistry, Permoserstrasse 15, 04318 Leipzig, Germany 9 *[email protected] 10 This document is the unedited Author’s version of a Submitted Work that was 11 subsequently accepted for publication in ES&T Water, copyright © 2024 The Authors. 12 Published by American Chemical Society after peer review. To access the final edited 13 and published work see https://doi.org/10.1021/acsestwater.4c00348. 14 15 Abstract 16 We present a versatile flow-through tube passive sampling device (TPS), with 17 controllable feed water volumetric flow, that can be calibrated in situ against the feed 18 water load of organic micro-pollutants (OMPs). This semi-passive approach has the 19 advantage of a determinable water load feeding the sampling device. The design of 20 the TPS allows for new sampling scenarios in closed piping while providing stable and 21 controlled sampling conditions. The calibration referencing an OMP’s feed water load 22 can describe the uptake behavior from wastewater treatment plant effluent with 23 potentially highly fluctuating OMP concentrations. The TPS and its load-dependent 24 calibration under realistic environmental conditions proves possible for a variety of 25 organic trace substances in a challenging matrix. Nine of the 20 monitored 26 representative OMPs could be calibrated load-dependently, leading to a good 27 agreement between the calculated concentration from the TPS and the average 28 concentration of corresponding direct measurements. Due to the simple measuring 29 principle and the membrane-less discs, many influencing factors such as diffusion, 30 turbulence and lag time phenomena can be neglected. The TPS could support the 31 existing online measurement analytics in a (process-) water treatment plant by 32 delivering integrated water concentrations for discharge monitoring. 33 Keywords 34 passive sampling, load dependent calibration, wastewater, organic contaminants, 35 fluctuating concentrations, organic micro-pollutants, flow-through. 36 Synopsis 37 The study introduces a versatile tube passive sampling device that allows controlled 38 flow and in situ calibration for monitoring organic micro-pollutants in wastewater, 39 demonstrating effective load-dependent calibration under fluctuating conditions. 40 Contribution 41 T. Hensel: methodology, investigation, software and writing original draft, J.-H. Hein: 42 methodology, investigation and software, T. Reemtsma: writing - review, A. Sperlich: 43 project administration, R. Gnirß: funding acquisition, F. Zietzschmann: methodology, 44 supervision, review & editing. 45 1 Introduction 46 Organic micro-pollutants (OMP) play a central role in the quality assessment of various 47 compartments of the water cycle; wastewater in particular accounts for a high 48 proportion of the introduction of OMPs into water bodies1. Particularly in highly 49 urbanised areas, it is necessary to consider not only the concentration of these 50 substances but also the associated load which the aquatic environment is confronted 51 with2–5. 52 Several options for wastewater treatment plant (WWTP) effluent monitoring exist: 53 • Grab samples which provide accurate snapshots for monitoring but lack any 54 information outside the individual sampled point in time6, 55 • composite samples which provide average data over the chosen time period but 56 require costly and laborious continuous maintenance and surveillance, 57 • passive samplers (PS) which provide cheap and readily available qualitative time58 integrative data over the application period but tend to loose accuracy if OMP 59 concentrations vary substantially7,8. 60 The principle of PS is the enrichment of analytes from a donor phase (typically liquids 61 like aqueous phases; or gases9) onto a sorbent and subsequent analysis of the 62 sorbent for sorbed analytes after a certain period. PS-enriched OMP for subsequent 63 analysis should be neither degradable nor volatile. There is a variety of passive 64 sampler types for aqueous use, specializing in the type of contaminants they are 65 targeting, like the well-described membrane-encapsulated sampler “polar organic 66 chemical integrative sampler” (POCIS)10 or the organic-diffusive gradients in thin-films 67 (o-DGT)11 that are targeting polar organic compounds. There are also approaches that 68 assess viruses12, use different materials like silicone rubber13, apply tube shaped 69 devices14, or try more complex devices with a forced flow through15,16. To date, the 70 passive samplers described in the literature are limited with respect to possible points 71 of uses. For example, passive sampler-based monitoring of wastewater streams 72 requires non-pressurized, free-flowing, accessible sampling locations17,18 like ponds, 73 basins, streams etc. In addition, many approaches use laboratory-based calibration, 74 thus potentially missing important environmental factors19. In drinking and ground 75 water (with relatively stable OMP concentrations), the concentration statements of per76 and polyflouroalkyl substances (PFAS) from POCIS with laboratory flow-through 77 calibration led to a good agreement to the accompanied composite samples20. In 78 surface water, quantitative applications of laboratory-calibrated PS appear 79 controversial, leading to only qualitative statements21. Applications in wastewater 80 introduce an extremely inhomogeneous and challenging matrix22 compared to 81 drinking, ground, or surface water. Laboratory PS calibration under controlled 82 conditions cannot provide sufficient reliability for such highly complex water matrices 83 of strong compositional fluctuations. Known impacts on the OMP-uptake like salinity, 84 temperature23, pH value, and the concentration of dissolved organic matter (DOC)24, 85 all of which can change rapidly, are mostly neglected. For these reasons, in situ 86 calibration is much more accurate25 and leads to good quantitative statements in 87 wastewater with passive sampler applications (for OMPs with non-fluctuating 88 concentrations)8. However most passive sampling devices are calibrated in vitro in 89 their kinetic uptake phase26. 90 To compensate for changing peripheral parameters (temperature, turbulence, etc.), 91 isotope-labelled performance reference compounds (PRC) can be used27. The use of 92 PRCs though has its limitations (e.g. different release behaviour of PRC as compared 93 to uptake behaviour of target OMP)28,29. 94 Many authors commonly avoid the use of PRCs and rely on in situ calibration30–32, 95 obviating the need for PRCs, whereas all influences on the uptake are present.33,34 96 The present study compares an in situ time-dependent calibration with a load97 dependent approach which relates the uptake of analytes into the sorbent to the feed 98 water load of a substance to the PS. With this new calibration approach the goals of 99 this study are: 100 • Introduce a controlled flow-through sampling device (tube passive sampler, 101 TPS) that has an adjustable feed water flow, 102 • present a quantitative load-dependent calibration approach for 20 organic micro 103 pollutants (OMPs) from WWTP effluent, 104 • compare the concentration results of 11 OMPs derived by the presented TPS 105 to OMP concentrations from conventional composite sampling in a WWTP 106 effluent. 107 2 Materials and Methods 108 2.1 Design of the Tube Passive Sampler 109 The design of the TPS (cf. Fig. 1) aims at highly controllable and reproducible sorption 110 of OMPs onto PS discs, supressing influences like turbulence27, flow velocities35, 111 additional membranes10, and avoiding PRCs36. The TPS can access open water 112 surfaces by use of a gear pump. The design also allows to conveniently sample in 113 pressurised lines by enclosing the sorption discs in a special designed pipe (Fig. 1C), 114 thus in any monitoring setup that uses slight over pressure as found in treatment 115 plants, process tubing, or pumping stations; in such scenarios, an active pump is not 116 necessary. 117 118 Fig. 1: A: Schematic cross-section and zig-zag-flow diagram of the miniaturised TPS. Circles with dots facing 119 towards, circles with crosses facing away. B: Photograph of the TPS. C: Integration into a process water system. 120 Within the TPS, a constant sample water flow passes laterally along the passive 121 sampling discs (Fig. 1A and Fig. S6A, including TPS & housing dimensions). A mass 122 flow controller (MFC) sets the volume flow for each experiment. The device consists 123 of stainless steel and was designed and constructed by GCI GmbH Königs 124 Wusterhausen (GER), patent DE 10 2016 003 8430, 2017. Four different commercially 125 available and unmodified 47 mm-discs were used for evaluation (Atlantic® HLB 126 (Biotage), Atlantic® DVB (Biotage), Atlantic® C18 (Biotage) and Resprep-C18 127 (Restek)). All discs came with glass fibre layers applied by the manufacturers. The 128 stainless steel disc retainer fits the diameter of the discs as well as the thickness 129 (5 mm) of the discs exactly and holds the discs with a stainless steel mesh-plate (mesh 130 diameter ca. 7 mm) on both exposed sides (for more details see Fig. S6B), leaving 131 1735 mm2 disc area per side exposed to the sample water. 132 2.2 Calculation of the load-dependent calibration of the tube passive sampler 133 Our TPS calibration bases on the classical time-dependent calibration approach 134 discussed in detail by Booij et al.26, cf. equation (1) below, and applied in a variety of 135 studies8,16, with constant concentrations. This approach proved only sparsely 136 applicable on OMPs with strongly fluctuating concentrations8,26. It uses a time137 weighted average concentration (cw,twa) and follows equation (1). 138 𝑀𝐷,𝑡𝑖𝑚𝑒(𝑡)=𝑅𝑆∗∫𝑐𝑊𝑑𝑡 (1) MD,time (g) is the amount of analyte on the disc after the exposure time t (d), 𝑅𝑆 139 represents the sampling rate in the dimension of (𝐿∗𝑑−1), ∫ cw dt = t * cw,TWA. 140 To differentiate between the time-dependent and the load-dependent calibration we 141 introduce the collection ratio (𝑅𝐶, dimensionless), related to the sampling rate through 142 equation (2) 143 𝑅𝑆=𝑅𝐶∗𝑄 (2) with 𝑄 being the (adjustable) fixed water flow through the sampler. 144 This difference allows the calculation of the TPS’s feed water load 𝑀𝑊,𝑐𝑢𝑚 via equation 145 (3). 146 𝑀𝑊,𝑐𝑢𝑚(𝑡)=𝑄∗∫𝑐𝑊 𝑑𝑡 (3) Using equations (2) and (3) in equation (1) leads to equation (4) to calculate the 147 compound-specific collection ratio 𝑅𝐶. 148 𝑅𝐶=𝑀𝐷(𝑡) 𝑀𝑊,𝑐𝑢𝑚(𝑡) (4) To calculate the time weighted average concentration 𝑐𝑊,𝑇𝑊𝐴 from the sampling ratio 149 𝑅𝐶, equations (1) and (2) can be combined to equation (5). 150 𝑐𝑊,𝑇𝑊𝐴 =𝑀𝐷(𝑡) 𝑅𝐶∗𝑡∗𝑄 (5) 2.3 Experimental design 151 Evaluation and in situ calibration of the TPS took place in a WWTP treating domestic 152 wastewater with some industrial wastewater shares in Berlin, Germany. The WWTP 153 has a capacity of 42 500 m³ per day which is equivalent to a population equivalent of 154 ~350 000. The TPS was placed inside the process monitoring station at the WWTP 155 outlet channel. Branching from a pressurized effluent pipe with continuous flow inside 156 the station, two overflow containers were fed with moderate overpressure (~1.2 bar), 157 for the setup see Fig. S2. For the calibration experiments, an automatic water sampler 158 WS Porti 24T (WaterSam GmbH, Balingen, Germany) was placed just outside the 159 monitoring station. The sampler rinsed its hose automatically before it took 160 consecutive 8 hour composite samples (20 mL/6 min) of the effluent throughout each 161 complete experiment. The samples were cooled at 4 °C until transfer to the laboratory. 162 All passive sampling experiments took place consecutively one after another, all 163 accompanied with their individual composite samples. For the order and duration of 164 the TPS experiments, see Fig. S7. 165 2.4 Preparation of the sorption discs 166 Disc conditioning involved careful placement of the DVB and HLB sorption discs in a 167 solid phase extraction (SPE) disc manifold, followed by rinsing with LC-MS-grade 168 solutions of 5 mL methanol with 1 % (v/v) formic acid, 5 mL methanol, 5 mL of methanol 169 with 1 % (v/v) ammonia and 20 mL ultra-pure water before storage in ultra-pure water 170 until usage. Mounting of the discs within the TPS comprised careful fixation into the 171 disc retainer, insertion of the disc retainer into the TPS, and closing of the lids. The 172 mass flow controller of the electronic control unit yielded a steady volumetric water 173 flow through the TPS. After the experiment, short rinsing with ultra-pure water removed 174 deposits and particles on the discs (while inside the retainer), followed by 175 disassemblage of the retainers and transfer of the discs into a polypropylene box lined 176 with a sheet of aluminium foil for transportation. In the laboratory, the discs dried under 177 a gentle airflow in a fume hood for 2 days. The discs were then transferred in an SPE 178 disc manifold and were rinsed consecutively with 5 mL of methanol with 1 % (v/v) 179 formic acid, 5 mL methanol, 5 mL of methanol with 1 % (v/v) ammonia collecting all 180 three filtrates of one disc in one vial. The combined filtrates were concentrated to 0.5 181 mL in a water bath under a constant nitrogen flow, transferred in a 1.5 mL-vial and 182 reconstituted with methanol to 1 mL. The vials were stored at -25 °C. Each disc 183 represented an individual sample. 184 2.5 Analysis 185 Eluates were measured in aqueous dilutions of 1:500 and 1:50.000, water samples in 186 aqueous dilutions of 1:5 and 1:100, respectively. The compounds were analysed using 187 an Exactive+ Orbitrap system with Dionex eluent pumps from Thermo Fisher Scientific 188 (Waltham, USA). The injection volume onto the online SPE column (Thermo Fisher 189 Scientific Hypersil GOLD aQ 2.1x20 mm) was 1 mL. The analytical column was an 190 Acquity UPLC HSS T3 (2.1x50 mm, 1.8 µm, Waters Corp., Milford USA). Compound 191 separation resulted from a constant flow of 0.6 mL per minute, using a binary eluent 192 gradient of ultra-pure water (1 % (v/v) methanol, 0.1 % (v/v) formic acid) (A) and 193 methanol (0.1 % (v/v) formic acid) (B) with a gradient elution from 1 % B ramping 194 linearly to 95 % B at 7 minutes. Column rinsing uses 99 % B for 2.5 minutes, followed 195 by a reconstitution step on starting conditions at 1 % B for 3 minutes. Ionisation was 196 achieved using an electron spray ionisation (ESI) in positive and negative mode. The 197 constant concentration over time like metoprolol, the load values show a similar x-axis 328 distribution compared to the time-dependent graph, despite different axis units (arrows 329 from A to C). Nonetheless, even small differences in concentration among the 330 consecutively conducted experiments of the same exposure time lead to different 331 loads, thus small shifts on the x-axis (Fig. 4C). When considering a compound with 332 highly variable concentrations, like tolyltriazole, the load-dependent calibration (Fig. 333 4D) demonstrates pronounced differences as compared to the time-dependent 334 calibration (Fig. 4B): The 4 arrows to the left of the graph (from Fig. 4 B to D) depict 335 the splitting of each of the two consecutive experiments with 2 and 7 days exposure. 336 The concentration variations of tolyltriazole result in different feed water loads in 337 experiments that had the same duration (2 days or 7 days) but were conducted at 338 different times (i.e. at different feed water concentrations). This observation also 339 occurs for the other observed OMPs’ calibration curves, see Fig. S8. 340 Within Fig. 4, panels B & D, the two rightmost arrows indicate a lower tolyltriazole disc 341 uptake during the 21 days experiment compared to the 14 days experiment. While the 342 time-dependent calibration leads to the conclusion that the 14 day experiment marks 343 the maximum allowable exposure time (corresponding to the highest OMP disc 344 uptake), the feed water load-dependent calibration demonstrates that longer exposure 345 times are possible – with almost linear increases of the disc uptake up until the 21st 346 day (despite not integrated in the calibration due to our conservative 2/3 threshold 347 approach). This behavior applies to other substances whose concentrations fluctuate, 348 like industrial chemical benzotriazole (data not shown). Compounds that have a mostly 349 constant concentration over time (e.g. metoprolol) show no comparable change in data 350 point sequence (on the x-axis) when plotted over the load, although a spreading of 351 x-axis values occurs in the load-dependent plot (cf. other calibration plots, e.g. AMPH 352 and DEET in Fig. S8). The plotting of disc-load/cW,TWA8,26 vs. time, rearranges the data 353 points on the y-axis depending on their specific cW,TWA values, which influences the 354 linear regression, cf. Fig S10. The load-dependent representation provides a 355 resolution of each individual experiment according to its specific measured load. 356 The feed water load-dependent calibration is powerful in scenarios of largely varying 357 OMP concentrations – such as WWTP effluents with hourly, daily, weekly, monthly etc. 358 patterns of industrial/household chemicals or pharmaceutical agents (Fig. 2; and 359 weekly variation of x-ray contrast media38,39). In cases of substantially fluctuating OMP 360 concentrations, the disc uptake might occur rapidly (high OMP concentrations, cf. 361 tolyltriazole 14 d experiment in Fig. 4) or slowly (low OMP concentrations, cf. 362 tolyltriazole 21 d experiment). Note that the 14 d experiment and the 21 d experiment 363 were not executed in parallel but consecutively. Hence, not the exposure time is 364 generally the limiting variable, but the maximum disc uptake is the limiting variable. 365 Resulting from defined time values in time-dependent calibration approaches, the 366 corresponding models include fewer x-axis values than the feed water load-dependent 367 approach (cf. differentiation of data points when changing from top graphs to bottom 368 graphs in Fig. 4). Passive samplers of “classical” designs – thus lacking the highly 369 controlled characteristics of the here-presented TPS – additionally suffer from 370 uncertainties induced by either unknown or fluctuating volumetric flows. Such cases 371 occur when deploying passive samplers freely into water streams in ponds/basins etc. 372 For some OMPs (acesulfame, propyphenazone, 4-dimethylaminoantipyrine, iomeprol, 373 tributylphosphate, tris-(2-chloroisopropyl)-phosphate, triphenylphosphate), an 374 evaluation was not possible due to a lack of data from the water samples. For valsartan 375 acid however, a special behavior occurs: This OMP occurs continuously in all aqueous 376 samples and also on all PS discs but the application of both the time-dependent and 377 the load-dependent calibration fails, see Fig. S9. This behavior could result from 378 (microbial) transformation of valsartan and other sartans into valsartan acid, 379 accumulating not only from the feed water but also being formed directly in the passive 380 sampler from precursors40,41. Unstable and/or potentially degradable substances are 381 therefore not suitable for assessments via passive sampling. For 382 2-methylthiobenzothiazole (MTBT), no calibration was possible, potentially due to the 383 volatility of this thioether and long dwell time on the sampling disc. 384 The classical sampling rate (RS in L/d) and here-introduced collection ratio (RC with no 385 dimension) for all evaluable OMPs are summarized in Tab. 1 under “large prototype”. 386 The sampling rates (RS) and collection ratios (RC) result from the slope of the 387 respective linear regression and for RS the mean concentration over all experiments 388 included in the linear regression corresponding to Fig. 4. The units of RC allow for 389 direct inference of the amount of an OMP collected from the feed water flow to the 390 TPS. Since the coefficient of determination will not give an adequate interpretation on 391 the error of the used no-intercept regressions used to determine RS and RC42, we use 392 the absolute residual error (ARE, in mg) of the disc uptake (y-values) to interpret the 393 quality of fit, see also Section S3. For the large prototype, 7 of the 11 OMPs ARE are 394 comparable if not better for the load-dependent calibration than for time-dependent 395 one (highlighted in green). Note that both the time-dependent and the load-dependent 396 calibration use the benefits of the TPS (flow-through device, steady flow conditions, 397 constant temperature). Furthermore, we used consecutive experiments, which differs 398 from other studies that start different experiments at the same time and only vary the 399 exposure length8. Future studies should assess the TPS’ benefits (load-dependent 400 calibration, time-dependent calibration) against classical passive sampling without 401 volumetric flow control (e.g., POCIS, o-DGT) with corresponding time-dependent 402 calibration. 403 Tab. 1: Performance parameters (sampling rate 𝑅𝑆 and collection ratio 𝑅𝐶) of the in situ calibration of the two TPS, 404 better performing calibration according to absolute residual error highlighted in green; ARE: Absolute residual error 405 (ARE compares modelled and measured disc uptake data points, see section S3). 406 Large prototype Miniaturised prototype Time-Dependent LoadDependent Time-Dependent LoadDependent Compound RS (mL/d) ARE (mg) Rc (ng/µg) ARE (mg) RS (mL/d) ARE (mg) RC (ng/µg) ARE (mg) Phenazone 29 0.017 0.035 0.01 38 0.058 0.094 0.053 AAA 13 0.21 0.014 0.18 37 0.33 0.075 0.25 AMPH 19 0.028 0.016 0.031 30 0.069 0.077 0.069 FAA 27 0.28 0.026 0.38 35 1.10 0.075 0.99 Diclofenac 40 0.045 0.039 0.054 46 0.36 0.097 0.44 Metoprolol 43 0.063 0.043 0.044 33 0.11 0.071 0.15 Iomeprol 21 0.33 0.014 0.24 Valsartan 37 1.1 0.023 0.39 DEET 39 0.008 0.047 0.006 32 0.017 0.082 0.011 Benzotriazole 13 1.70 0.023 0.92 15 0.58 0.037 0.48 Tolyltriazole 34 0.18 0.037 0.23 44 1.0 0.104 0.88 407 3.3 Miniaturisation and integration in online measurement 408 One of the goals of this study was to miniaturize the TPS in a way that it could be used 409 permanently side-by-side existing monitoring equipment in WWTPs, whilst reducing 410 the number of deployable passive sampler discs to two. We conducted consecutive 411 experiments with the miniaturized TPS prototype at exposure times of 3, 4, 5, 7, 8 and 412 14 days, now receiving WWTP effluent as feed water directly from a pressurized 413 bypass of the online measurement devices in the monitoring station of the examined 414 WWTP. The experiments lasted ≤14 d due to the insight of the previous experiments 415 that 2/3 of the maximum disc uptake generally occurred during that time. Again, the 416 subsequent calibration calculations only used data points below the 2/3 threshold of 417 the observed maximum disc uptake. 418 The in situ time-dependent (top) and load-dependent (bottom) calibration plots of the 419 miniaturized TPS (mini TPS) are shown in Fig. 5. The comparison of the time420 dependent and the load-dependent calibrations for metoprolol (Fig. 5 A and C) shows 421 no particular change in order on the x-axis, as indicated by the arrows. In contrast, the 422 tolyltriazole data (Fig. 5 B and D) rearrange on the x-axis upon plotting load423 dependently (Fig. 5 B) opposed to time-dependent (Fig. 5 B). This behavior is similar 424 to the plots of tolyltriazole in the large TPS (Fig. 4 B and D). Such behavior also applies 425 to the calibration plots of diclofenac, benzotriazole, and AMPH, see Fig. S5, which also 426 contains the data for the other tested OMPs. When plotting disc-load/cW,TWA vs. time, 427 a data point shift is observed on the y-axis, see Fig. S11. The re-arrangement of disc428 uptake data points demonstrates the importance of considering potentially variable 429 OMP concentrations. Even if a given OMP typically exhibits relatively constant 430 concentrations, potential variations may still occur, necessitating in situ calibration with 431 continuous monitoring of the OMP concentration. The 𝑅𝐶 as well as the 𝑅𝑆 results of 432 the corresponding calibrations are summarized in Tab. 1 under “miniaturized 433 prototype”. 434 435 Fig. 5: Top: Metoprolol (A) and tolyltriazole (B) amounts on disc of the miniaturized TPS over exposure time in 436 experiments of variable durations (3, 4, 5, 7, 8, 14 days) and bottom: Same amounts on disc over metoprolol (C) 437 and tolyltriazole (D) feed water loads (as obtained from continuous 24 h composite samples) through the mini-TPS 438 during the corresponding experimental run times; regressions only including values below threshold; red arrows 439 linking same values in the different graphs. Data points result from means of duplicates. 440 3.4 Validation of the tube passive sampler 441 In order to test the load-dependent calibration (Fig. 5 C and D), we carried out an 442 additional 7-day test (we chose this experimental run time so that the disc uptake 443 would likely not exceed the threshold and thus lie within the calibration range). We 444 used the measured disc uptake from this experiment to calculate the feed water load 445 resulting from the linear regression equation in Fig. 5. Accompanying continuous 446 composite samples provided the necessary information on the actual concentrations 447 in the feed water during the experiment as a reference. The calculation results using 448 𝑅𝐶 along with equation (5), the results using 𝑅𝑆 and the results from the composite 449 samples are compared in Tab. 2. For MTBT no concentration resulted from the TPS. 450 Tab. 2: OMP average concentrations (cW) obtained for a 7 day experiment of the mini-TPS in WWTP effluent 451 calculated from time-dependent and load-dependent TPS calibration compared to mean concentrations from 7 day 452 composite water samples. 453 CW in µg/L Compound time-dependenta load-dependenta autosamplerb Phenazone 0.23 ± 0.01 0.21 ± 0.01 0.33 ± 0.01 AAA 2.8 ± 0.2 2.8 ± 0.2 3.0 ± 0.2 AMPH 0.19 ± 0.01 0.19 ± 0.01 0.35 ± 0.01 FAA 9.5 ± 0.7 9.4 ± 0.7 11.6 ± 0.9 Diclofenac 3.8 ± 0.3 3.8 ± 0.3 4.0 ± 0.3 Metoprolol 2.0 ± 0.1 1.9 ± 0.1 1.7 ± 0.1 DEET 0.12 ± 0.01 0.11 ± 0.01 0.13 ± 0.01 Benzotriazole 42.8 ± 1 35.5 ± 0.9 16.6 ± 0.4 Tolyltriazole 11.5 ± 0.1 10.1 ± 0.1 7.9 ± 0.1 a Calculated value ± absolute error, based on the individual error, 454 b Mean value ± absolute error, based on the individual error. 455 The OMP concentrations calculated via the load-dependent and time-dependent TPS 456 calibration are ranging in two orders of magnitude from 0.11 µg/L, 0.12 µg/L, 457 respectively for DEET and 35.5 µg/L, 42.8 µg/L, respectively for benzotriazole and are 458 in similar ranges as concentrations obtained from auto-sampling. The values 459 generated from the PS are mostly in good agreement with those from the autosampler, 460 e.g. 3.8 µg/L (load-dependent TPS) and 4.0 µg/L (autosampler) for diclofenac. 461 Deviations >30% occur for AMPH and benzotriazole, yet the TPS still allows for rough 462 quantitative assessments for such OMPs. We further assume that more calibration 463 data points and more routine in deployment of the mini-TPS will allow for increased 464 accuracy. The concentrations derived from the time-dependent and the 465 load-dependent calibrations are comparable, while both approaches likely profit from 466 the overall highly constant conditions facilitated by the TPS. The TPS is a valuable 467 asset for long-term monitoring of WWTP effluents where autosampling is often very 468 timeand resource-consuming. Another advantage of the TPS compared to 469 autosamplers is its small size, allowing for deployment in monitoring booths or running 470 process lines. This feature also means that temperatures below freezing impact the 471 TPS much less than autosamplers which malfunction easily due to freezing water in 472 intermittently ponding supply tubes; the constantly running water flow in the TPS 473 avoids malfunction by freezing. Due to its versatility of using any commercial 47 mm 474 SPE disc material, the TPS opens up many possibilities for monitoring chemically 475 variable target compounds while being integrated into the existing analytical 476 equipment. 477 4 Conclusion 478 This study tested a semi-passive sampler concept with precisely adjustable volumetric 479 flow, containing standard 47 mm SPE discs, to monitor organic micro-pollutants in 480 WWTP effluent. A miniaturized design with two passive sampler discs was 481 implemented in pressurized lines of a WWTP monitoring station. After a load482 dependent calibration of the system, an additional test run showed applicable and 483 determined average WWTP effluent concentrations for a variety of OMPs, which were 484 highly similar to concentrations obtained from measured composite samples. Given 485 the straightforward handling once installed, and – more importantly – the 486 good/accurate quantitative results, this semi-passive system is, to our knowledge, the 487 first to calibrate passive sampler discs with exact feed water loads of potentially 488 strongly fluctuating concentrations, combined with the possibility to operate in 489 pressurized lines. The load-dependent calibration and the time-dependent calibration 490 both agreed well with corresponding composite samples. Further experiments should 491 compare the load-dependent and time-dependent calibrations within the TPS with 492 “classical” passive sampling without volumetric flow control. Many use-cases appear 493 feasible for the TPS, like quantitative continuous monitoring which spares – once the 494 TPS is calibrated – the need for extensive and laborious auto-sampling. Due to the 495 versatility of the TPS, further application could result from continuous monitoring of 496 industrial discharges by either the industry themselves or authorities43,44. 497 Supporting Information 498 Additional experimental details, materials, and methods, including photographs of 499 experimental setup 500 5 Acknowledgements 501 This study was partially founded by the European Union’s Project Horizon 2020 502 research and innovation programme “PROMISCES” under the grand agreement No. 503 101036449. Additional in-house funding from Berliner Wasserbetriebe is also 504 gratefully acknowledged. The design and construction of the TPS is property of GCI 505 GmbH (patent No. DE 10 2016 003 8430, 2017) and was kindly made available for the 506 experiments. We thank Mr. Robert Bloch, Mr. Paul Simmons and the team of Berliner 507 Wasserbetriebe’s WWPTs laboratories for valuable discussions and analytical 508 assistance. We also like to thank the reviewers for their revealing insights. 509 6 References 510 (1) Muschket, M.; Zahn, D.; Neuwald, I.; Knepper, T.; Schumann, P.; Rabe, L.; Ruhl, 511 A. S.; Jekel, M.; Kuckelkorn, J.; Schnitzer, G.; Schulze, H.; Dölchow, U.; Fink, A.; 512 Reemtsma, T. Persistente mobile organische Chemikalien in der aquatischen Umwelt: 513 Quellen, Vorkommen und technische Möglichkeiten zu ihrer Entfernung in der 514 Trinkwasseraufbereitung (PROTECT). DOI: 10.5281/ZENODO.7683983. 515 (2) Jelic, A.; Gros, M.; Ginebreda, A.; Cespedes-Sánchez, R.; Ventura, F.; Petrovic, 516 M.; Barcelo, D. Occurrence, partition and removal of pharmaceuticals in sewage water 517 and sludge during wastewater treatment. Water Research 2011, 45 (3), 1165–1176. 518 DOI: 10.1016/j.watres.2010.11.010. Published Online: Nov. 13, 2010. 519 (3) Loos, R.; Carvalho, R.; António, D. C.; Comero, S.; Locoro, G.; Tavazzi, S.; 520 Paracchini, B.; Ghiani, M.; Lettieri, T.; Blaha, L.; Jarosova, B.; Voorspoels, S.; Servaes, 521 K.; Haglund, P.; Fick, J.; Lindberg, R. H.; Schwesig, D.; Gawlik, B. M. EU-wide 522 monitoring survey on emerging polar organic contaminants in wastewater treatment 523 plant effluents. Water Research 2013, 47 (17), 6475–6487. DOI: 524 10.1016/j.watres.2013.08.024. Published Online: Sep. 15, 2013. 525 (4) Neuwald, I. J.; Muschket, M.; Seelig, A. H.; Sauter, D.; Gnirss, R.; Knepper, T. P.; 526 Reemtsma, T.; Zahn, D. Efficacy of activated carbon filtration and ozonation to remove 527 persistent and mobile substances - A case study in two wastewater treatment plants. 528 The Science of the total environment 2023, 886, 163921. DOI: 529 10.1016/j.scitotenv.2023.163921. Published Online: May. 8, 2023. 530 (5) Seelig, A. H.; Zahn, D.; Reemtsma, T. Sources of persistent and mobile chemicals 531 in municipal wastewater: a sewer perspective in Leipzig, Germany. Environ Sci Pollut 532 Res 2024. DOI: 10.1007/s11356-024-33259-0. Published Online: Apr. 18, 2024. 533 (6) Bartelt-Hunt, S. L.; Snow, D. D.; Damon-Powell, T.; Brown, D. L.; Prasai, G.; 534 Schwarz, M.; Kolok, A. S. Quantitative evaluation of laboratory uptake rates for 535 pesticides, pharmaceuticals, and steroid hormones using POCIS. Environmental 536 toxicology and chemistry 2011, 30 (6), 1412–1420. DOI: 10.1002/etc.514. Published 537 Online: Mar. 25, 2011. 538 (7) Huckins, J. N.; Petty, J. D.; Booij, K. Monitors of Organic Chemicals in the 539 Environment: Semipermeable Membrane Devices; Springer Science+Business 540 Media, 2006. DOI: 10.1007/0-387-35414-X. 541 (8) Mackie, R.; Hawker, D. W.; Ghorbani Gorji, S.; Booij, K.; Qu, X.; Bowles, K.; Shea, 542 S.; Higgins, C. P.; Kaserzon, S. Application of a Microporous Polyethylene Tube 543 Passive Sampler for Monitoring Perand Polyfluoroalkyl Substances in Wastewater 544 Influent and Effluent. ACS EST Water 2024, 4 (5), 2281–2291. DOI: 545 10.1021/acsestwater.4c00125. 546 (9) Dong, H.; Wang, X.; Xu, L.; Ding, J.; Wania, F. A Flow-through Passive Sampler 547 for Microplastics in Air. Environmental Science & Technology 2023, 57 (6), 2362–2370. 548 DOI: 10.1021/acs.est.2c07016. Published Online: Jan. 16, 2023. 549 (10) Alvarez, D. A.; Petty, J. D.; Huckins, J. N.; Jones-Lepp, T. L.; Getting, D. T.; 550 Goddard, J. P.; Manahan, S. E. Development of a passive, in situ, integrative sampler 551 for hydrophilic organic contaminants in aquatic environments. Environmental 552 toxicology and chemistry 2004, 23 (7), 1640–1648. DOI: 10.1897/03-603. 553 (11) Challis, J. K.; Almirall, X. O.; Helm, P. A.; Wong, C. S. Performance of the organic554 diffusive gradients in thin-films passive sampler for measurement of target and suspect 555 wastewater contaminants. Environmental Pollution 2020, 261, 114092. DOI: 556 10.1016/j.envpol.2020.114092. Published Online: Jan. 29, 2020. 557 (12) Li, J.; Verhagen, R.; Ahmed, W.; Metcalfe, S.; Thai, P. K.; Kaserzon, S. L.; Smith, 558 W. J. M.; Schang, C.; Simpson, S. L.; Thomas, K. V.; Mueller, J. F.; Mccarthy, D. In 559 Situ Calibration of Passive Samplers for Viruses in Wastewater. ACS EST Water 2022, 560 2 (11), 1881–1890. DOI: 10.1021/acsestwater.1c00406. 561 (13) Martin, A.; Margoum, C.; Randon, J.; Coquery, M. Silicone rubber selection for 562 passive sampling of pesticides in water. Talanta 2016, 160, 306–313. DOI: 563 10.1016/j.talanta.2016.07.019. Published Online: Jul. 8, 2016. 564