{"paper_id":"2ef7d29e-54ff-4319-a219-60c0ea29074b","body_text":"Ultraviolet filters (UVFs)\nare widely used in numerous personal\ncare and hygiene products. 1  Benzophenones,\nsuch as benzophenone-3 (BP-3), benzophenone-4 (BP-4), and avobenzone\n(AVO) are among the most used organic UVFs. These compounds, as other\nUVFs, are characterized as photostable and lipophilic, 2  thus bioaccumulating and biomagnifying along the trophic\nweb. 1 , 3  Benzophenones and their derivatives have\nbeen reported as endocrine disruptors. 4  They cause adverse effects in fish and rodents’ fecundity, 5  neurotoxicity, cytotoxicity, and Hirschsprung\ndisease. 6  In humans, they have been associated\nwith estrogen-dependent diseases such as endometriosis. 7  Despite their lipophilic nature, they have been\ndetected in all types of water: surface, seawater, 8  wastewater, 9  tap water, 10  and groundwater. 11  UVFs have also been found in solid environmental matrices such sewage\nsludge, 12  sediments, 13  and biota (fish, marine mammals, birds, and invertebrates). 14 − 16\nManaged aquifer recharge (MAR) is considered a potential source\nof UVFs into groundwater bodies. 17  This\nis especially true when recharged water is the treated effluent of\na wastewater treatment plant (WWTP). While MAR contributes to augmenting\nwater resources, regulatory concerns have arisen regarding the possibility\nof aquifer contamination. 18  Nonetheless,\nMAR can be integrated as a tertiary treatment process in conjunction\nwith a WWTP, termed Soil Aquifer Treatment (SAT), since infiltration\nthrough the porous media promotes the natural attenuation of many\nrecalcitrant compounds. Valhondo et al. proposed the installation\nof reactive barriers at the bottom of infiltration ponds to further\nenhance degradation and sorption processes during SAT. 19  This reactive barrier consists of a mixture\nof natural materials providing a range of sorption sites (neutral,\ncationic, and anionic) and enhancing biodegradation by the release\nof labile organic carbon, which in turn promotes a broad spectrum\nof reduction–oxidation (redox) states, thereby expanding the\npathways for degradation and the removal of contaminants of emerging\nconcern (CECs). 20\nSorption of UVFs,\nas other CECs, is governed by their affinity\nto the immobile organic phases (as characterized by octanol–water\npartition constant,  K ow , or organic carbon–water\npartition constant,  K oc ) and, when ionized,\nby interactions with charged solid surfaces. 21 − 23  In turn, these\nsorption mechanisms are affected by numerous chemical parameters of\nthe solution (ionic strength, pH, concentrations of competing ions)\nand the solid surfaces (their composition, surface charge, and the\norganic matter age). 24  Ionizable organic\nmolecules, with p K a  within the experimental\npH range, may change their ionic state and therefore their sorption\nmechanism. In groundwater, the pH usually ranges between 6 and 8.\nThis is relevant for some UVFs, such as benzophenones, whose p K a s are mostly within this range ( Table  1 ), causing both neutral and\nionic forms to coexist.\nUVFs biodegrade in groundwater under aerobic and anaerobic\nconditions,\nbut mainly by cometabolism. 25 − 27  Therefore, the presence of labile\norganic carbon enhances their biodegradation. The biodegradation pathway,\ndetermined by the redox state and the electron acceptor availability,\nis a key parameter controlling the type of transformation product\n(TP). 25  Biodegradation occurs mainly inside\nthe biofilms, which are an assemblage of microorganisms comprising\nmicrobial species attached to a surface and encased in a self-synthesized\nmatrix with water and extracellular polymeric substances (EPS). They\ncan act as active sorbents to organic compounds in porous media. 28 , 29  EPS is both positively and negatively charged, favoring the retention\nof anionic and cationic species. There are also lipids that promote\nthe retention of lipophilic compounds. Although there is some evidence\nof organic micropollutants retention in biofilm in WWTPs, 22  there is no experimental evidence about their\nretention in porous media, and neither regarding biotransformation\nand retention mechanisms with respect to changes in redox conditions\npromoted by the reactive barriers.\nThis study aims to assess,\nthrough both experimental methods and\na numerical model, the role of biofilm in the sorption and degradation\nprocesses of specific benzophenone-type UV filters found in a real\nWWTP effluent. Two soil aquifer treatment systems were utilized for\nthis investigation, with one of them incorporating a reactive barrier\nat the infiltration pond. The selected UVFs were BP-3 and its main\nTPs, benzophenone-1 (BP-1), 4,4′-dihydroxybenzophenone (4DHB),\n4-hydroxybenzophenone (4HB), and 2,2′-dihydroxy-4-methoxybenzophenone\n(DHMB), BP-4, and AVO. Analyzed UVFs samples comprised all of the\nphases present in the SAT system: water, aquifer sediments, and biofilm.\n\nTwo pilot SAT systems were tested at the Palamós\nWWTP (Girona, Spain). Each consists of a constructed aquifer coupled\nto an infiltration basin fed with a WWTP secondary effluent ( Figure  1 ). One of the two\nsystems, referred to as “CT” (compost-treatment), contained\na 1 m thick reactive barrier that had been installed 2 years before\nthe UVF experiment. The barrier was made up of sand (0.15 to 0.4 mm\nparticle sand to provide structure, 49% in volume), vegetal compost\n(from tree pruning to provide sorption sites for lipophilic compounds\nand release DOC, 49%), and clay (providing cation sorption sites 2%).\nThe other system is denoted “ST” (sand-treatment) as\nit reflects the conventional SAT systems, consisting solely of sand.\nThe resulting overall porosities were 0.38 for ST and 0.48 for CT.\nSoil aquifer\ntreatment (SAT) pilot system at Palamós (Girona,\nSpain). (a) Location and picture of the facility with the six systems.\n(b) Scheme description of each system. Black dots represent the sand\nsampling points, whereas the red ones represent the biofilm and water\nsampling points.\nBoth systems were equipped with PVC piezometers\nfor monitoring\nalong the flow path ( Figure  1 ): 9 in the aquifer (sections A-C in  Figure  1 ), and a fully screened inclined piezometer\nat the base of the barrier (oblique piezometer, referred to as “O”\nhereafter) to collect water exiting the barrier. The outflow was integrated\nand collected by a discharge pipe installed at the base of the aquifer.\nPiezometers were sampled with submersible centrifugal pumps (nontoxic\nABS plastics, Stainless-steel impeller, and silicone delivery pipe).\nGroundwater level, electrical conductivity (EC), and temperature were\nregularly monitored. A detailed description of the pilot systems can\nbe found in Valhondo et al., 2020. 30\nOur experiment consisted of the continuous monitoring of UVFs and\nTPs ( Table  1 ) along\nthe SAT systems (inflow, section A, section B, section C, and outflow\n( Figure  1 )), within\nthe three environmental compartments: water, aquifer sediments, and\nbiofilm. The monitored UVFs were those naturally present in the recharged\nwater, i.e., the secondary treatment effluent (STE) from the Palamós\nWWTP. UVFs monitoring started after the injection of lithium acetate\n(LiAc), acting as a tracer of the experiment. The monitoring extended\nover 86 days, during which UVFs, LiAc, and hydrochemical parameters\nwere meticulously characterized.\nPrior to the experiment, the\nSAT systems were operated for one\nmonth, with a flow rate of 1 m 3 /d. The purpose of this\npre-experimental phase was to establish a steady state flow and to\ncharacterize conservative transport and arrival times at monitoring\npoints (details in  S2  in the SI). The same\nflow conditions were kept during the whole experiment (86 days). The\nwater level at the outlet was set at 135 cm for both systems, resulting\nin an approximate total residence time in the tanks of around 12 days.\nTracer injection (day 0) consisted of 57 L of water from the STE\nspiked with LiAc at 17.8 mg/L in the infiltration basins of the ST\nand CT systems. LiAc was selected because Li +  acted as\na sorbing tracer and Ac –  as a source of easily degradable\norganic carbon, favoring redox processes, as well as UVF degradation.\nLi +  was an adequate tracer because: (1) it did not interact\nwith UVFs determinations as colorimetric tracers or bromide; 31  (2) it was absent in the STE, and, (3) it was\nnondegradable. Water, sediments, and biofilm samples were collected\nat scheduled times during the experiment ( S2  in the SI).\nWater samples were collected from inflow\n(STE), piezometers (O, A2, B2, and C2), and outflow of the two SAT\nsystems, after purging the piezometers using drive pumps. Physicochemical\nparameters (EC, pH, redox potential ( Eh ), and temperature)\nwere measured  in situ  with a multiparameter probe\n(YSI, Inc. Yellow Springs, OH). Alkalinity was measured also  in situ  with a test kit (Merck Millipore, Darmstadt, Germany).\nWater hydrochemistry characterization consisted in the analysis of\ndissolved organic carbon (DOC), NH 4 + , major\ncations, and anions (sampling and analytical details in  S3 ). Samples for UVFs analysis were collected\nin amber glass bottles of 150 mL, immediately frozen, and kept in\nthe dark to prevent photo- and biodegradation. The analyzed UVFs are\nlisted in  Table  1 ,\nand analytical determination is described in  Section  2.3 .\nNine sediment samples were collected from each SAT system at different\nlocations (see  Figure  1 ) and at different experimental times (see  S2  in the SI). Samples were taken at 55 cm depth, which was the same\ndepth as the screened interval of the piezometer. Sampling was done\nusing a 110 cm long and 1 cm wide drill. The sediment samples were\ncollected to determine the UVFs concentrations ( Section  2.3 ) as well as the fraction\nof sedimentary organic carbon ( f oc ) ( S6  of SI).\nBiofilm in the SAT systems was\ncharacterized using biotraps installed in the piezometers (O, A2,\nand B2) at 55 cm depth of each SAT system ( Figure  1 ). Biotraps consisted of sandbags packed\ninto a protective plastic mesh. The siliceous sand was the same as\nthe aquifer and was previously muffled for 5 h at 600 °C to remove\nall sedimentary organic matter (SOM) and to ensure that any observed\nretention of UVFs in the biotraps was solely attributed to the biofilm,\nas little interaction with the silica sand was anticipated. 28  Biotraps were installed one month before the\ntest to promote bacterial growth and biofilm formation. UVFs concentration,\nbacterial density, and EPS (as an amount of biofilm measurement) were\ndetermined in the biotrap samples (see  Section  2.3  and  S6  in the\nSI).\nUVFs determination in the sediments and water matrices followed\npreviously developed methods. 33 , 34  However, a new methodology\nwas developed and validated for biofilm analysis. The method consisted\nof extraction and purification using QuEChERS and further analysis\nby liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS).\nSeparation and quantification of the target analytes were performed\nby high-performance liquid chromatography in a Hibar Purosher STAR\nHR R-18 (50 mm × 2.0 mm, 5 μm) column using a Symbiosis\nPico instrument from Spark Holland (Emmen, The Netherlands) attached\nto a 4000 QTRAP mass spectrometer from Applied Biosystems-Sciex (Foster\nCity). The achieved limits of detection (LODs) ranged between 0.18\nand 0.87 ng/g of dw, and the limits of quantification (LOQs) ranged\nbetween 0.60 and 2.89 ng/g of dw, showing the method’s good\nsensitivity. The relative standard deviation (RSD) values were below\nor equal to 20%, indicating good precision. Accuracy was evaluated\nby the recovery rates of each standard spiked in a representative\nmixture of biofilms at two concentration levels. Recovery rates at\n5 ng/g dw spike level were between 72.48 and 131.2%, and at 100 ng/g\ndw, recovery rates were between 86.4 and 122.8%. Further details about\nthese analytical methods are provided in the SI ( S6.4 section ), including the instrumental analysis, method\nvalidation, and QA/QC for the biofilm analysis.\nThe fate of UVFs and most pollutants\nin porous media is mostly controlled by advection/mixing-dispersion\nand by sorption and degradation processes. Inside the biofilm, the\nmost important processes are the different retention mechanisms, diffusion,\nas well as degradation 35  ( Figure  2 a).\n(a) Conceptual representation\nof UVFs partitioning processes, where\n(1) refers to the ionization, exchange between neutral (n) and ionic\n(i) forms, controlled by pH and p K a ; (2)\nexchange between mobile (m) and immobile waters (im), controlled by\nmolecular diffusion; and (3) retention into the sediment characterized\nby ionic interactions or by affinity to organic matter. (b) Two-compartment\nmodel adopted for equilibrium and mass balance calculations and (c)\nbiodegradation pathways of BP-3 and transformation products. * Compounds\nalso formed in anoxic conditions.\nFor a semiquantitative interpretation of observations\nand to evaluate\nthe role of the processes involved in the biofilm retention, we built\na simple (in that processes are represented by simplified equations)\nmodel yet complex (in that numerous variables are involved). The model\nsimplifies the SAT system by representing it as a single cell (or\nbox) with two compartments: the mobile one and the immobile ( Figure  2 b), in a similar\nand simplified way as multi-rate mass transfer model. 36  The immobile compartment represents water in biofilms,\nmicroorganisms, and extracellular biological material, sedimentary\norganic matter, and isolated pores. Therefore, sorption (both absorption\ninto organic and biological solids and adsorption onto mineral surfaces\nand charged organic matter) occurs primarily in the immobile compartment.\nSimilar to previous works, 37  it is also\nassumed that microbial communities responsible for degradation reactions\nlive in biofilms and they are mature so that degradation is limited\nby the concentration of the compound and can be taken as first order.\nThis assumption was reasonable in our system since the SAT systems\nwere run for more than 2 years. Instead, the mobile compartment represents\nfree-flowing water so that sorption and degradation are neglected.\nThese compartments are characterized by (1) their exchange rate,\nα, inverse of the mean residence time in the immobile water\ncompartment (promoted by diffusion through the immobile compartment),\n(2) the mobile and immobile porosities, ϕ m  and ϕ im , respectively, both referred to the total volume of the\nmedium, so that the total porosity (assumed to be 0.25) is ϕ\n= ϕ m  + ϕ im , (3) sorption properties\n( K ow  for absorption into lipophilic substances\npresent in biofilms,  K oc  for absorption\ninto aquifer organic matter, and  K d  as\na lumped parameter for adsorption onto ionic surfaces), blended into\nthe retardation factor,  R im, j  (ratio of total, sorbed plus dissolved, to dissolved mass\nin the immobile compartment, computed from partition parameters as\ndiscussed in the  SI ), and (4) degradation\nrates λ p , j  [T –1 ] (of parents  p  to daughters  j ). The latter describe the degradation pathways, which\nare complex and dependent on redox conditions. The model cannot reproduce\nredox conditions, which would require multiple immobile zones (the\nmost reducing conditions are reached in the least accessible portions\nof the medium). Therefore, a simplified degradation network, using\nonly the analyzed species and neglecting redox state, has been adopted 25 , 38 , 39  ( Figure  2 c). This network is further discussed in  Section  3.2 .\nThe\nresulting model, equilibrium, and mass balance calculations\nas well as all used parameters are described in  S8  in the SI. Sorbed concentrations ( S j im ) were derived from the observed dissolved concentrations as given\nby 1a 1b where  c j  and  c j im  (M V –1 ) are\nthe concentrations of the  j -th solute in the mobile\nand immobile compartments, respectively,  S j im  (M V –1 ) is the mass retained in solid phases per\nunit volume of sediments, and λ p , j R  (T –1 ) is the effective degradation rate\nof each parent defined as 2 which depends on λ j , T R  = ∑ d =1 N dj λ j , d R , the\neffective total rate (sum of all degradation paths). These calculations\nwere performed using the mean observed water concentration at piezometers\nin section B (representative of mean conditions) for  c j  in  eq  1a  and then to obtain  S j im . The modeled results were compared to the measured retained mass\nin the sediment samples since they contained both biofilm and sedimentary\norganic matter. They were calculated by averaging spatial concentrations\nsince results were comparable before and after the acetate injection,\nbesides this, they were considered more representative of the overall\naquifer system than those of biotraps ( Table S10 ). The mass balance of BP-3 and TPs ( Figure  2 ) is given by (see the  SI  for details) 3 where θ =  Q /( Q  +  V αϕ im ) relates\nthe advective flux of solutes ( Q  is the mean flow\nrate) and the total flux, advective plus mobile immobile ( V  is the volume of sediments), and  c j Inp  is the input concentration (average of the 15 inflow samples, using\nhalf of the detection limit for samples below). This equation was\napplied sequentially, starting with BP-3, using  eq  3  to compute the outflow and (1a) to compute\nthe immobile water concentrations for the computation of daughter\ncompounds in the chain of  Figure  2 . These outflow concentrations were compared to the\naverage of the 6 outflow samples (see  Tables S5–S7  in the SI for the detailed inflow and outflow UVFs concentrations).\n\nGroundwater flow and conservative transport in the\ntwo SAT systems were comparable, showing similar arrival times of\nboth EC and lithium at the different sampling points ( Figure S3a,b and c,d  respectively). In the two\nsystems, preferential flow dominated transport through the recharge\nzone, with a fast-early arrival and a broad range of residence times\nat the first observation point of the two systems (O-piezometer),\nsimilarly to Valhondo et al. 40  On the other\nhand, transport in the aquifer was what might be expected from a relatively\nhomogeneous aquifer with moderate dispersion for short transport length\n(up to A2), but more significant dispersion up to B2 ( Figure S3 ).\nRedox zonation ( pe  in  Figure S3 ) was largely controlled\nby the oxidation of DOC and ammonium in the recharged water. The breakthrough\ncurves (BTCs) of DOC ( Figures  3  and  S3e,f ) were similar to those\nof Li +  ( Figure S3c,d ) in that\nthey displayed a fast peak and a much lower concentration for the\nCT than for the ST. The main difference between them is that the tails\nof the DOC BTCs dropped much faster than the tails of Li +  BTCs, which we attribute to strong biogeochemical activity in the\nslow flow and immobile portions, especially in the CT. In fact, DOC\nwas mostly degraded by the time the plume reached the B2 point (fully\ndegraded in the CT). That is, the presence of the reactive barrier\nenhanced the biological activity and the organic carbon oxidation,\nwhich is consistent with observations of Valhondo et al. 30 , 41 , 42  Recall that the reactive barrier\nhad been installed 2 years before the slug injection, and one of its\ngoals was to release DOC to favor reducing conditions. By the end\nof this 2-year period, it was expected that this release had been\nlargely depleted. 43 , 44  Therefore, we attribute the higher\ndegradation in the CT system to a richer bacterial activity as was\nestablished in Hellman et al. 45  The greater\nreactivity in CT is also supported by the Li +  breakthrough\ncurves ( Figure S2 ). The tail of Li +  lasts longer than that of CE, and more so for the CT, while\nthe peak arrivals are similar, which reflects both that Li +  is adsorbed and that adsorption is noninstantaneous (otherwise the\nLi +  peak arrival would have been retarded).\n(a) Evolution of the\nconcentration of DOC, BP-3, and TPs at the\ndifferent sampling points vs the cumulative injected water volume\nalong the two SAT pilot systems. The corresponding volumes for the\nsampling points are approximately 1(O), 3(A2), and 9(B2) m 3 . The shadow rectangle represents the range of influent concentrations.\n(b) Mass balance of BP-3 and TPs in the two SAT systems. The “inflow”\nrepresents the cumulative mass input during the initial 12 days of\nthe experiment. The “outflow” represents the mass present\nin the effluent after undergoing the residence time within the SAT\nsystem for the same duration.\nThe redox potential ( Figure S3g,h ) was\nquantified as  pe  (computed as  pe  =  EhF /2.3 RT , where  Eh  is the measured redox potential,  F  is the Faraday\nconstant,  R  is the gas constant, and  T  is the temperature). Redox potential was measured during sampling,\nthus representing the mixture of captured waters, which explain the\nobserved fluctuations. The injection of easily degradable DOC led\nto reduced conditions (i.e., a fast drop in  pe ).\nA lower  pe  was reached in the ST, where  pe  values decreased from 8 to 3–4 (associated to Mn reducing\nconditions), whereas in the CT,  pe  dropped to the\nrange of 4–7. Regarding pH, while the pH of the inflow was\naround 7.5, it was kept between 6.5 and 7 in the whole system. A detailed\ndiscussion of the evolution of the different terminal electron acceptors\nis provided in  Section S10  in the SI ( Figures S4 and S5 ). Note that although the acetate\ninjection was equal in the two systems, the response was significantly\ndifferent. The CT system displayed a larger capacity for DOC and ammonium\noxidation as observed in previous works, 30 , 42  promoting higher diversity of the redox conditions and, therefore,\nincreased density and diversity of bacterial populations being, all\nof these, attributed to the presence of the compost reactive barrier\n( Table S10 ).\nFigure  3 a displays\nthe evolution in space (three observation points) and time (cumulative\ninjected water volume) of concentrations of BP-3 and its TPs in water\nsamples. Sampling was scheduled according to the travel times determined\nin a previous tracer test (S2) so that the same water mass was sampled\nas it passed through the different sampling points. Similarly,  Figure  3 b shows the overall\nmass balance of BP-3 and its TPs of the same sampled water in the\ninflow and outflow of the two SAT systems.\nOverall, UVF concentrations\nin the sampling points ( Figure  3 a) did not exhibit a clear trend, which reflects: (1) the\ninherent variability of UVF concentration in the STE ( Table S5 ), especially at the O-piezometer (this\nnoise is dampened at the downstream observation points due to sorption\nand dispersion); (2) the complexity of the BP-3 transformation pathway,\ncompromising back and forward reactions conditioned to the redox potential\n( Figure S2 ); and (3) the low concentrations,\nclose to the method limit of quantification, especially for TPs. Despite\nthese difficulties, two important issues emerge in  Figure  3 .\nFirst, CT displayed\noverall lower concentrations than ST for BP-3\nand its TPs in water ( Figure  3 b). We attribute this to a higher retention and thus degradation\ncapacity of this system induced by the reactive barrier in a similar\nway as reported for DOC degradation (see  Section  3.1 ). The higher degradation capacity of the\nCT system can be associated with (1) the geochemical conditions promoted\nby the reactive barrier facilitated the degradation of UVFs; (2) the\nCT induced a different microbial population capable of degrading these\ncompounds; and (3) a higher sorption capacity of the system induced\nby the reactive barrier.\nSecond, BP-3 and most TPs concentrations\ndropped in both systems\njust after the acetate injection, concurring with the tracer peaks,\nbut then rebounded back to levels comparable to the inflow concentration\n( Figure  3 a). This would\nmean that the presence of acetate enhanced its biodegradation and\nenlarged the TPs’ production, as reported by Liu et al. 25  The rebound is explained because: (1) after\nthe acetate peak, the system tended to the prior conditions, especially\nin ST, and (2) a potential release of BP-3 and TPs from the biofilm\nphase to the water, promoted by decay of microbial communities and\ntheir detachment from the solid surface after the acetate peak, which\nwould favor the dragging of EPS and all sorbed substances.\nRegarding\nTPs formation processes, BP-3 was first biotransformed\ninto BP-1 via demethylation of the methoxy substituent 39  ( Figure  2 c). This transformation pathway can occur both in oxic and\nanoxic conditions. 25  Note that BP-1 was\nalso present in the influent water ( Figure  3 , blue shadow zone), meaning it was already\npresent in the inlet water of the WWTP or it was formed there, as\nobserved in Mao et al. 39  After the injection\nof LiAc, BP-1 dropped in the ST and remained below the method limit\nof detection at the CT.\nBP-1 biodegradation can lead to the\nformation of 4DHB. 25  Similarly and under\noxic conditions, BP-3 can\nform DHMB 46  ( Figure  2 c). Both were identified in pore waters and\nfollowed a behavior similar to that of BP-3 and BP-1. Prior to the\ninjection, their concentrations were comparable to those in the inflow\nalong the ST and were below the limit of detection in CT. Also, the\nconcentrations of both dropped as the DOC peak reached every observation\npoint and rebounded back to initial operation concentrations after\nthe peak in the ST, but remained comparable to inflow concentrations\nin A2 and B2 points of the CT.\nFinally, 4HB can be formed from\n4DHB or, directly, from BP-1 by\na hydroxyl group loss 39  ( Figure  2 c). Interestingly, this compound\nwas not present in the inflow water. Thus, it was formed in the SAT\nsystems, demonstrating that our systems enhanced the biodegradation\nof BP-3 regardless of the reactive barrier. Similarly, its concentration\nwas lower in CT, which was completely removed since it was not detected\nin the outflow ( Figure  3 b).\nRegarding the other monitored UVFs, we did not observe\na clear\ncorrelation with the acetate injection ( S10  in SI). BP-4 was detected in water samples at concentrations between\n1 and 2 orders of magnitude higher than the other UVFs, which is consistent\nwith its low degradability both in the WWTP (inflow concentrations\nwere high) and in the SAT systems. AVO, on the other hand, was neither\npresent in the influent water nor in the outflow, but it was detected\nin water samples from the two SAT systems. As far as we know, there\nis no literature about the potential formation of AVO as a derivative\nproduct of other UVFs. Therefore, we conjecture (1) that AVO reverts\nback from a conjugate (as a glucuronide conjugate typical from Phase\nII metabolism) or (2) a potential desorption from the biofilm and\nsediments, resulting from the dragging of EPS.\nAmong the target UVFs, only 4HB, BP-4, and\nAVO were detected in sediments and biofilm samples. As we did not\nobserve any space correlation for UVFs retained in the solid phases\nor for retention parameters (bacterial density, EPS,  f oc , or  f ow ,  Table S10  in the SI), we presented the data in average form\nbefore and after the acetate injection ( Table  2 ). Note that the UVFs detected in both biofilm\nand sediment samples are associated with their accumulation/retention\nin the immobile phase, that is, interaction with SOM and/or biofilm.\nAs biotraps were previously muffled, the detected UVFs in those samples\nwere only associated with the retention promoted by biofilm ( Table  2 ). Therefore, our\nresults experimentally confirm that biofilm in the porous media is\ncapable of retaining certain UVFs.\nConcentrations are referred to grams\nof sampled sediments, where ND means nondetected.\nA further step is to compare the relative importance\nof mobile\nand immobile compartments in the fate of UVFs. To do this, we have\nmodeled the concentrations of UVFs in the two compartments using the\ndual-domain model and compared it with the experimental information\n( Figure  4 ), using the\nbiomass concentrations in  Table S10 . In\nthis analysis, we included all of the UVFs, although we only detected\n4HB, BP-4, and AVO. The comparison with the nondetected compounds\nwas done using the detection limit (depicted as white dots in  Figure  4 ). Modeled results\nfor undetected compounds were lower (or very close) to the detection\nlimit (as seen by the white points in  Figure  4 ). This reflects that concentrations are\nvery low and suggests that the current detection limit is still too\nhigh for detecting sorbed concentrations in biofilm/sediments.\nConcentrations\nin mobile and immobile compartments (expressed in\nng per m 3  of aquifer). Points represent measured concentrations\n(EXP), whereas bars represent model computations (MOD) in the two\ncompartments. White points refer to undetected compounds (ND) plotted\nat the detection limit. The salmon represents the BP-3 and TPs compounds.\nTwo features deserve attention from  Figure  4 . First, observed (or modeled\nwhen not detected)\nretained UVFs mass in immobile phase is between 1 and 4 orders of\nmagnitude larger than dissolved, with variations depending on the\ncompounds sorption parameters and degradation rates ( Figure  4 ). This accumulation in aquifer\nsediments was in line with the partition coefficients ( K ow ,  K oc , and  K d ) of these compounds ( Table  1 ). That is biofilm and sedimentary organic\nmatter present in sediment samples act as real sinks for dissolved\nUVFs. Second, beyond the affinity of the compounds to the solid phases\n(values of  K ow ,  K oc , and p K a ), the accumulation\nof TPs also results from the imbalance between its production and\ndegradation. For example, 4HB, which had not been detected in the\ninflow and had similar solid affinity as BP-3 and the other TPs (see  Table  1 ), displays a high\naccumulation (its distribution coefficient had to be increased to\nmatch observations) and is the last derivative produced along the\ntransformation processes of the BP-3. This shows that the imbalance\nbetween its degradation, production, and exchange with the mobile\nphase is responsible for its retention in biofilm, which is consistent\nwith the findings of Wang et al. 35  Interestingly,\n4HB was not initially present in the biofilm samples and at very low\nconcentration in the sediments ones.\nBP-4 was found more frequently\nand homogeneously in aquifer sediments\nthan in biofilm ( Table  2 ). The presence of BP-4 in water has increased recently because it\nis substituting BP-3 in some personal care products in order to increase\nwater solubility. 44  This substitution was\nmotivated by the low affinity of BP-4 for the organic phase (log  K ow  = 0.37) being mainly present in its ionic\nform (p K a  = −2.42). Therefore,\nits retention in sediments is explained by ionic interactions with\npositive surfaces such as Fe and Mn oxides, which are present in our\nSAT systems (see  S9 ). Chang et al. 100  described an interaction between Mn oxides\nand BP-4. Furthermore, positive surfaces have been described inside\nbiofilms. 35\nAVO was detected in most\nof the biofilm samples from both systems\nand at similar concentrations ( Table  2 ). AVO is not a transformation product of BP-3; thus,\nits accumulation in biofilm is explained by its high affinity to the\norganic phase in the biofilm (log  K ow  = 4.51) and slow degradability. Note that AVO did not show temporal\nor spatial variability, showing that it reached an equilibrium concentration\nthat was not dependent on redox conditions or the presence of reactive\nbarrier.\nIn this work, we have detected\nand quantified, for the first time in porous media, the amount of\nUVF retained in the nonmobile phases. The developed model reproduced\nqualitatively the retained mass in the nonmobile phase ( Figure  4 ), in the sense that it reproduces\nnot only the broad range of observed concentrations and yields low\nconcentrations for undetected compounds but also because it simulates\nthe anticipated and observed localization of reactions (4HB being\npresent in the immobile compartment despite its absence in the inflow\nand outflow). Besides this, it also shows that the compost SAT system\npresents lower masses of UVFs, despite the fact that its biofilm is\nmore active and SOM content larger. This implies a higher removal\ncapacity. As shown in  eqs  1a , b  and  2 , the degradation\nrate is effectively multiplied by the retardation coefficient, because\nso is the residence time. As a result, even compounds that are highly\nrecalcitrant to conventional wastewater treatments may be extensively\ndegraded in biofilms. In short, the suite of observations and model\ncalculations implies that hydrophobic compounds will tend to accumulate\nin biofilms, which promotes biodegradation even if the degradation\nrate is very slow (their effective residence time is orders of magnitude\nlarger than that of water). These conclusions are illustrated in  Figure  5 , which summarizes\nthe mass balance of the various processes controlling the UVF fate\nof BP-3 and its TPs in the CT (a similar figure for the ST is shown\nin  Figure S5 ), which underscores the central\nrole of biofilm in the fate of UVFs. This aligns perfectly with the\nfindings in the recent study by Markale et al., where they observed\na significant impact of biofilm permeability heterogeneity on biologically\ndriven reactions. 47\nSummary of retention\nand degradation processes computed for BP-3\nand its TPs in the CT. (a) Mass balance terms (ng/m 3 /day)\nfor CT, with cold colors for inputs (inflow and production from parent\ncompounds) and cold colors for outputs (outflow and degradation) and\n(b) mass retained in the aqueous phase and solid phase (biofilm and\naquifer sediments).\nOverall, this study demonstrates that implementing\na reactive barrier\nin an infiltration basin improves the degradation of benzophenone-type\nUVFs and, especially, of BP-3 and its TPs, and that biofilm acts as\nan additional environmental compartment favoring the retention and\ndegradation of UVFs in porous media. It does not sway us that these\nprocesses can be assumed for other hydrophobic compounds. With this,\nwe want to emphasize that the current understanding of the organic\ncompounds’ fate should incorporate biofilm as a pool capable\nof bioaccumulating these compounds. Beyond aquifers, the role of biofilms\nas an additional environmental compartment would imply that aquatic\necosystems would be exposed to a higher dose of UVFs and likely to\nmany other organic pollutants in comparison to those calculated through\nthe measured concentrations in water, suspended particulate matter,\nsediments, and biota.","source_license":"CC-BY-4.0","license_restricted":false}