Biodegradable dissolved organic carbon profiling reveals capacity of carbon‐based potable reuse treatment over a range of operating conditions

Biological treatment is gaining ground as a means to enhance removal of total organic carbon (TOC) as part of a multi‐barrier treatment train for water reuse. Here we applied biodegradable dissolved organic carbon (BDOC) analysis to evaluate the extent of removal of various TOC fractions through a pilot‐scale water reuse train employing flocculation/sedimentation, ozone, biologically active carbon (BAC), and granular activated carbon (GAC). BDOC analysis highlighted GAC and ozone treatments as critical to non‐biodegradable dissolved organic carbon removal and the need to optimize BAC performance to maximize GAC adsorption capacity. BDOC analysis was further applied to benchmark process performance to operational conditions, such as empty bed contact time (EBCT), occurrence of nitrification, and operational upsets. The lower EBCT proved to be less resilient to nonideal conditions. Overall, BDOC analysis proved an asset for understanding and improving operation of ozone/BAC/GAC treatments for water reuse.


| INTRODUCTION
Advanced water treatment (AWT) for the purpose of producing high quality water for potable reuse can take many forms, typically employing multiple physical and chemical treatment barriers.Reverse osmosis is often thought of as an essential treatment barrier for potable reuse (Bernados, 2018;Leverenz et al., 2011), given its ability to remove a wide range of contaminants, including salinity.However, reverse osmosis presents several drawbacks, including high energy demands, loss of efficiency due to fouling, high implementation costs, and need to treat and dispose of the brine waste stream (Schimmoller & Kealy, 2014;Subramani & Jacangelo, 2014;Wenten & Khoiruddin., 2016).Further, the quality of water may not be compatible with the aquifer geochemistry for recharge.This has brought about interest in membrane-free, AWT trains as an attractive alternative.In particular, carbon-based AWT trains employ ozone and biologically active carbon (BAC) coupled with granular activated carbon (GAC).This alternative harnesses the natural biodegradative capacity of microorganisms inhabiting the BAC to provide biological degradation of trace organic contaminants, such as pharmaceuticals and personal care products, which are inefficiently removed in conventional drinking water and wastewater treatment trains (S.D. Kim et al., 2007;Ternes et al., 2002;Westerhoff et al., 2005).To expand the application of ozone-BAC-GAC, it is necessary to gain a better understanding of its capacity to remove various forms of organic carbon including total organic carbon (TOC), dissolved organic carbon (DOC), biodegradable dissolved organic carbon (BDOC), nonbiodegradable dissolved organic carbon (NBDOC), and particulate organic carbon (POC), in the face of various long-term operational changes and upsets.
BAC filtration has become widely applied in the context of conventional drinking water treatment for removing organic carbon (Carlson & Amy, 1998;Hozalski et al., 1995;W. H. Kim et al., 1997;Terry & Summers, 2018;Yapsakli & Çeçen, 2010), including precursors to disinfection byproduct formation (Vatankhah et al., 2019).The capacity of BAC filtration can be further enhanced by employing an upstream advanced oxidation process, such as ozone, which acts to increase the bioavailability of complex organic compounds ahead of the biofilter (Nishijima & Speitel, 2004).However, while the knowledge base of BAC filtration for conventional drinking water treatment translates to other contexts to some degree, there is a need to assess the boundaries for its application for potable reuse.For example, the organic matter makeup in a wastewater's tertiary effluent consists largely of soluble microbial products and is distinct from the natural organic matter encountered in more conventional drinking water sources (Garner et al., 2016;Rittmann & Snoeyink, 1984).Further, a key consideration for the ozone-BAC-GAC configuration is to maximize biodegradation of DOC in the BAC to the extent possible to maximize the lifespan of the downstream GAC, where BDOC and (much more difficult to remove) NBDOC may compete for GAC adsorption sites, leading to the costly need to regenerate GAC.The effectiveness of biofiltration has been previously linked to changes in empty bed contact time (EBCT), ozone dose, temperature (and related kinetics), nutrient and substrate matrixes, and lesser understood microbial interactions (Basu et al., 2016;Carlson & Amy, 1998;Gerrity et al., 2018;Hallé et al., 2015;Peterson & Summers, 2021).Treatment train performance can be further improved through the implementation of upstream, synergistic processes such as coagulation-flocculation-sedimentation (Floc-Sed), which can decrease contaminant loads to subsequent ozonation and biofiltration by removing BDOC, NBDOC, and POC (Üstün et al., 2011;C. Volk et al., 2000;C. J. Volk & Lechevallier, 2002).
TOC is often monitored as a gross indicator of water quality through a treatment train, as it is typically cost prohibitive to monitor each individual trace organic contaminant of concern (Vaidya et al., 2020).However, to optimize biological treatment, it is essential to be able to assess the biodegradable TOC fraction.For example, approximately 80% of conventional TOC encountered in surface water-sourced drinking water (Terry & Summers, 2018) and secondary wastewater effluent (Khan et al., 1998) is nonbiodegradable.Various means of assessing BDOC have been applied to characterizing potable water (Escobar & Randall, 2001;Servais et al., 1989;van der Kooij, 1992;van der Kooij et al., 1982).The assimilable organic carbon test measures cellular growth of two model strains of bacteria that commonly occur in drinking water, relating it to measured loss of TOC and yielding units of mg acetate equivalents/L (LeChevallier et al., 1993;C. J. Volk & LeChevallier, 2000).However, a drawback of this approach is that it is unlikely that the two strains employed will be capable of degrading the full spectrum of biodegradable constituents represented by a given TOC measurement.The BDOC test circumvents this problem by employing an inoculum of microorganisms that is native to the system being tested and thus is more likely to yield a representative assessment of the true proportion of TOC that is biodegradable.Conventionally, the BDOC test is applied in an analogous fashion as the biochemical oxygen demand test, where TOC (rather than dissolved oxygen) is monitored in the bulk water with time.However, a solid media-based approach was developed that is more relevant to biofilm-based treatments, such as BAC filtration (Page & Dillon, 2007;Park et al., 2004;Trulleyov a & Rulík, 2004).When applied across an AWT train, the BDOC test could ideally serve to inform with respect to the percent biodegradable carbon removed by each treatment stage, providing a direct measurement of the theoretical maximum possible DOC removal via biodegradation, the nonbiodegradable fraction, and overall capacity of biological treatment.
Here, BDOC analysis was used to assess the capacity for removal of TOC, DOC, BDOC, NBDOC, and POC through

Article Impact Statement
Biodegradable dissolved organic carbon analysis can help to optimize performance of carbonbased potable reuse trains.a pilot-scale carbon-based AWT potable reuse treatment train employing coagulation/flocculation/sedimentation (Floc-Sed), ozone, BAC, GAC, and UV disinfection.Specifically, the objectives of this research were to (1) evaluate how each treatment stage influences overall TOC removal and the proportion of POC, BDOC, and NBDOC removed and (2) characterize how changes in operating conditions (e.g., ozone dose, monochloramine addition, nutrient addition, and temperature) and upset events (i.e., individual treatment processes operating outside of intended performance range or uncharacteristic changes in water quality parameters) are reflected in BDOC measurements and overall performance over long-term operation.Of particular interest to this study was the effect of EBCT on organic carbon profiles following BAC filtration and GAC contacting, especially when the GAC is operated long-term without regeneration.This pilot-scale study was conducted as part of the ongoing Sustainable Water Infrastructure for Tomorrow (SWIFT) initiative, in which the goal is to achieve water of a quality suitable for indirect potable reuse via aquifer recharge and permit limit of 4 mg/L TOC.

| Site description, sample collection, and preservation
Samples were collected from a 4.3 gpm AWT pilot plant operated at the Hampton Roads Sanitation District's York River Wastewater Treatment Plant (Figure 1) over an $18-month period.The pilot-scale plant uses York River's tertiary effluent (secondary effluent treated via a methanol supplemented dentification filter -TETRA™ DeepBed™ filters (De Nora, Milan, Italy)) and employs coagulationflocculation-sedimentation (Floc-Sed), monochloramine supplemented ozonation for bromate suppression (Wert et al., 2007), BAC filtration, GAC contacting, and ultraviolet disinfection.To assess the impact of EBCT, the pilot flow was split between two different BAC-GAC trains: one with 5 min BAC EBCT (BAC5) followed by 10 min GAC EBCT (GAC10) and the other with 10 min BAC EBCT (BAC10) followed by 20 min GAC EBCT (GAC20).The BAC effluent was not combined; the 10 min GAC EBCT unit received flow from only the 5 min BAC EBCT unit.In addition to the treatment processes themselves, various chemicals and operational conditions were adjusted periodically.Major changes and upset conditions are outlined in Table 1 with additional design parameters provided in Tables S1-S5 and Figures S1 and S2.Prior to this period of testing, the pilot was operated under relatively stable conditions from June 2016 until beginning of the addition of preformed monochloramine in September 2016, substantially before the initial sampling in January of 2017.Further, BAC filters were exhausted prior to the pilot commissioning through continuous feeding of secondary effluent until organics removal plateaued (determined via bulk TOC and DOC measurements).Sampling events were also conducted at intervals ≥3 weeks following process modification, with most occurring $7 weeks.Operating conditions were stable according to standard water quality parameter monitoring, unless otherwise noted.Additional details on pilot operation during this study period are available in prior publications (Pruden et al., 2020;Sun et al., 2019;Vaidya et al., 2019Vaidya et al., , 2020Vaidya et al., , 2021) ) with summarized water quality data for the pilot provided in Tables S1-S4.Bulk water samples for BDOC testing were collected at the effluent of each treatment, aligning hydraulic residence time to the extent possible to ideally capture the same plug of water as it traveled through the pilot.Sampling events and dates are reported in Table 1.For each sample, two 1-L samples were collected in acid-washed, baked glass bottles with triplicate 25-mL samples collected in acid-washed, baked 40-mL glass vials for both TOC and DOC analysis.DOC samples were syringefiltered using a sterile PVDF 0.45-μm syringe filter (MilliporeSigma™ Millex™, Darmstadt, Germany), which was rinsed immediately before collection with sample water to prevent carbon leaching from the filter membrane itself.Both TOC and DOC samples were acid preserved (pH <2) using 85% phosphoric acid before they, and the nonacid preserved 2-liter samples, were stored on ice for transit.

| Operational conditions and organic carbon profiling through the carbon-based AWT train
Table 1 provides a timeline of the 10 sampling dates examined in this study as they relate to operational conditions and upset events.Each sampling event is named sequentially (S01-S10) and further differentiated according to whether the event represented stable (e.g., SS01) or upset (e.g., SU02) conditions or any other relevant defining information (e.g., SUMeth02, SUTurbid09).Monochloramine addition (3 mg/L Cl 2 ) and ozone ratios were sequentially adjusted, to suppress formation of bromate, but required dechlorinating with sodium bisulfite or calcium thiosulfate to prevent disinfectant from entering microbially active BAC filters.Following months of stable operation targeting 1:1 ozone:TOC dose, the ozone ratio was decreased to assess its impact on BDOC generation.Sampling events following SS06 sequentially targeted ozone:TOC ratios of 0.5, 0.25, 1, and 0.8.Outside of the intentional operational variation, performance was also likely affected by upsets of increased DOC loading during SUMeth02 and SUMeth04, due to carry-over of methanol from the upstream denitrification unit, and during SUTurbid09, due to a Floc-Sed failure and elevated turbidity.Also, ozone dose variability (i.e., standard deviation) decreased from SS05 onwards, when more precise control measures were implemented (Spreadsheet in Supporting Information).GAC performance indicated dynamic shifts in carbon removal with time, consistent with adsorption sites being exhausted and shift to biological treatment as a function of bed volumes passed through the filter (Appendix A in Supporting Information).Almost all samples were collected after 20,000 bed volumes, the point at which GAC would be expected to be largely exhausted and begin transitioning to a steadystate BAC filter (Peterson & Summers, 2021).This was supported in our own assessment of DOC removal, Appendix A in Supporting Information.Table S6 summarizes the observed organic carbon profiles for each sampling date, including TOC, DOC, BDOC, NBDOC, and POC.

| BDOC analysis preparation and setup
BDOC analysis was conducted in accordance with previously published methods utilizing attached growth microbial communities (Escobar & Randall, 2001;Joret et al., 1991;Joret & Levi, 1986) with minor alterations to sample volumes and inoculum preparation.These methods were adapted from conventional suspended growth BDOC assays (Servais et al., 1987(Servais et al., , 1989) ) to better account for BDOC available to complex microbial communities present in biofilms (Page & Dillon, 2007;Park et al., 2004;Trulleyov a & Rulík, 2004) and were selected due to the nature of BAC filtration.At least 1 week prior to sample collection, a biologically active sand column (BAS) was set up and seeded with conditioning water, a one-time seed slurry of HACH polyseed and nutrients, and then continuously fed pilot influent water to develop an attached microbial community for inoculation into the BDOC testing bottles.The initial sand (effective size ranging from 0.45 to 0.55 mm with a uniformity coefficient of 1.5) was rinsed with DI water and baked at 550 C for 5.5 hours to burn off residual carbon.After collection, water samples and the BAS reactor were transported to the laboratory where the samples and BAS were combined.Initially, the BAS was gently rinsed three times with deionized water to remove residual organic carbon from the feeding phase and prevent contamination of the BDOC samples, especially those with lower DOC concentrations.After the first wash, $50 g BAS (wet weight) was inoculated into acid-washed and baked BDOC bottles (triplicate 250 mL-bottles for each sample location).Following inoculation, each triplicate set of BDOC bottles and inoculated sand was rinsed three times using the first liter of sample water, to further prevent contamination from the inoculated BAS which was originally conditioned with pilot feed water.After the second washing, each bottle was filled with approximately 200 mL of its corresponding sample, gently shaken, and sampled for a reactor DOC measurement.All sample bottles were placed on a shaker table set at 100 rpm for approximately 30 days at 20 ± 2 C and sampled for DOC measurements every $7 days.

| Laboratory TOC and DOC measurements
All TOC/DOC acid preserved samples were analyzed using a Sievers 5310 C Laboratory TOC Analyzer (Suez, Boulder, CO), which was purged prior to operation to prevent run to run contamination.Prior to analysis, each sample was sparged with nitrogen gas for 3 min at 10 PSI to remove inorganic carbon.Standards and nanopure water blanks were run at the beginning, end, and every 10-12 samples.Each sample was measured three times, with the first measurement being considered a "rinse" sample and the remaining two TOC and DOC measurements averaged to represent the measurement for each biological replicate.Biological replicates were averaged for subsequent analysis.

| Water quality and operational conditions data collection
BDOC measurements were compared with operational and water quality data.This comparison leverages periodic water quality monitoring occurring at the SWIFT facility, utilizing a variety of standard methods and online operational data collected continuously during treatment operation with inline sensors.
Because water quality data were collected periodically, sampling dates were used to sort and identify data collected around each sampling event, specifically the day of and the 3 days preceding and following each sampling date.Because all parameters were not measured every day of operation, selection criteria were developed that prioritized: (1) data collected on the day of sampling, (2) average of the day before and after sampling, (3) the only value in the data range if only one existed, or (4) a blank value.After selection, the data were used to benchmark BDOC measurements to water quality characteristics.
General operational data and parameters were collected by online analyzers at approximately 5-min intervals.Daily summary statistics, including averages, standard deviations, minima, and maxima were calculated and used to benchmark treatment operational data to subsequent BDOC measurements.

| Data analysis and statistics
DOC measurements over the $30-day BDOC test were plotted to produce carbon degradation curves (e.g., Figure S3).The initial and final DOC measurements were subtracted to calculate the amount of organic carbon remaining after treatment that was biodegradable (BDOC, mg/L).The ratio of the BDOC to the initial DOC measurement was calculated as the biodegradable fraction, with the final DOC measurement indicating the amount of NBDOC.The change in initial DOC measurements (i.e., effluent DOC from each treatment stage) between treatment processes was used to calculate the bulk DOC removal achieved in the pilot (mg/L) and the total treatment efficiency (% removal) by taking the ratio of DOC removed to initial DOC.Further, initial TOC and DOC measurements were used to calculate the amount and fraction of carbon that was found in the dissolved and particulate phases.Combinations of the above were utilized to fractionalize the carbon profile through the AWT train and assess carbon removal performance.
Nonparametric Wilcoxon signed-rank tests were conducted in R version 4.0.2(RStudio Team, 2020) to compare changes in organic carbon components between sampling location.Following the application of Wilcoxon signed rank tests on all samples, outliers were removed when justified by cross-referencing to sampling notes and the identification of documented upset conditions, prior to conducting exploratory correlation analysis.The distribution of each variable was tested for normality utilizing the Shapiro-Wilk test in RStudio.Normally distributed variables were correlated to water quality and operational data using the Pearson linear correlations.To correct for false discoveries associated with multiple comparisons, the Benjamini-Hochberg (BH) procedure was applied to each set of correlation tests at a liberal FDR set at 0.20, per exploratory recommendations (Benjamini & Hochberg, 1995).Significance for all statistical tests was set at p < .05.

| Fate of DOC in the carbon-based AWT train
Performance of the carbon-based AWT train was evaluated across all 10 sampling dates according to the conventional measure of DOC removal (Figure 2) (Table S6).Floc-Sed significantly reduced DOC, with a median reduction of 1.65 mg/L (24% of the influent carbon).Ozonation did not significantly reduce DOC, with a median change of À0.02 mg/L (Table S7) but did result in an average UVT increase of 8.80% ± 2.55 as a result of breakdown of organic compounds.Both BAC5 and BAC10 provided significant reductions of DOC, with median removals of 0.65 mg/L (11% treatment efficiency) and 1.19 mg/L (19% treatment efficiency), respectively.The observed improvement in DOC removal with elevated EBCT was significant.Similarly to the BAC filters, both GAC10 and GAC20 provided significant reductions in DOC, with median reductions of 0.93 mg/L (19% treatment efficiency) and 1.04 mg/L (27% treatment efficiency), respectively.In contrast to BAC filtration, increased EBCT resulted in only marginal additional reduction of DOC concentrations by GAC treatment.This lack of improved performance within GAC's at higher EBCT was somewhat surprising, given that with extended operation GAC begins to convert to BAC, which demonstrably improved in performance with extended EBCT.The authors postulate that additional limiting factor(s) (e.g., nutrients) could have precluded further biological removal at elevated EBCTs.Alternatively, the interplay between the amount of easily biodegradable carbon in the GAC influent and its EBCT could result in similar net removals for the EBCTs tested.In other words, it is possible that GAC20 received less readily bioavailable carbon (because of BAC10s improved performance), but facilitated greater removals of less bioavailable carbon at its higher EBCT while GAC10 primarily degraded the more readily bioavailable carbon left over from BAC5 during its shorter EBCT, resulting in similar total DOC removals.Consistent with this, the BAC10/ GAC20 combination did provide significantly improved DOC removal relative to the BAC5/GAC10 combination.
Under the conditions of this study, increased EBCT resulted in increased DOC removal during BAC filtration.Increased contact time likely increased interactions between microorganisms and BDOC and facilitated biological degradation of organics with slower kinetics (Wu & Xie, 2005;Zhiteneva et al., 2020).Additionally, increased EBCT may enhance breakdown of microbial inhibitors, such as ozone and associated by-products, effectively reducing their impact on the microbes.Notably, DOC removal did not markedly increase with increased EBCT in the GAC contactors, as both GAC10 and GAC20 yielded similar net reductions in DOC concentrations.However, since upstream BAC performance was improved by EBCT, this resulted in a greater percent removal of the remaining DOC during GAC contacting.This highlights the importance of optimizing upstream BAC filtration to maximize GAC contacting (Summers et al., 2020) and hypothetically prolong its operational lifespan.Still, the remaining organic carbon in the GAC10 and GAC20 effluents were 29% and 22% BDOC, respectively, indicating that there is still potential for biofouling and microbial regrowth in the final water (de Vera & Wert, 2019;Escobar & Randall, 2001;Joret et al., 1991;Thayanukul et al., 2013;van der Kooij, 1992).

| Fate of POC in the carbon-based AWT train
It was found that POC concentrations in the feed were very low (at most 10% of the TOC loading and typically below 5%) and did not significantly change during treatment, even after biological treatment where sloughed biomass is of concern.This was likely the result of upstream, tertiary denitrifying sand filtration prior to pilot operation.Thus, POC results are not discussed further, but are reported in full in the SI (Appendix B in Supporting Information).

| Fate of BDOC in the carbon-based AWT train
BDOC analysis provided deeper insight into the composition of the TOC through the treatment train, the role of each treatment stage, and performance with time.Figure 3 provides an example (Sampling Event SUMeth02 and SS-05) of the relative proportions of DOC that were BDOC and NBDOC through the treatment train and how much of each was removed by each stage.Table S8 summarizes statistical comparisons.Figure 4 provides a more detailed accounting of BDOC concentrations and their dynamics throughout the AWT train across the 10 sampling dates.
Aside from two of the upset sample events (SUMeth02, SUMeth04), pilot influent BDOC concentrations were relatively stable, ranging from 1.75 to 3 mg/L, representing 25%-40% of the total DOC concentration (Figure 4e).Floc-Sed significantly decreased BDOC, with an average removal of 1.78 mg/L (Table S7) and average percent removal of BDOC greater than 60%.Interestingly, the BDOC fraction was much more effectively reduced by Floc-Sed than the NBDOC fraction.This suggests that the particulates removed by Floc-Sed, such as micro-flocs lingering after upstream biological treatment, were largely biodegradable in nature.
Consistent with expectation, ozonation was the only the process that resulted in a significant increase in BDOC concentrations and percent BDOC relative to total DOC.This indicates that ozonation successfully served one of its purposes by increasing the biodegradability of incoming organic matter, a result that could not be discerned from DOC analysis alone.Accordingly, both BAC5 and BAC10 achieved substantial reduction in BDOC with BAC10, on average, yielding more extensive DOC removal (Table S7).The GAC treatment stage provided further significant reduction in BDOC, with GAC10 averaging 0.86 mg/L of BDOC removed and GAC20 averaging 0.80 mg/L of BDOC removed.

| Fate of NBDOC in the carbon-based AWT train
As summarized in Figures 3 and 4, the recalcitrance of NBDOC through the carbon-based AWT train was apparent.Only ozonation, GAC10, and GAC20 yielded any notable reductions in NBDOC with average reductions of 1.07, 0.40, and 0.65 mg/L, respectively (Table S7).Notably, ozonation did not provide measurable reduction in DOC, but the BDOC test revealed its ability to transform NBDOC into BDOC.Interestingly, although Floc-Sed substantially reduced BDOC, it failed to significantly change the NBDOC concentration.This suggests that NBDOC entering the pilot was not amenable to charge neutralization and/or sorption to the sedimented flocs that formed.
In order to be able to discern removal mechanisms of biodegradation versus sorption through the treatment train, the GAC implemented in the BAC filters was intentionally exhausted prior to commencing this study, resulting in theoretically negligible sorption-based removal.Correspondingly, BAC5 and BAC10 were found to achieve minimal reductions of NBDOC, averaging 0.05 mg/L and 0.11 mg/L, respectively, while ozonation accounted for an average reduction of 1.07 mg/L.Further, BAC5 effluent NBDOC concentrations were not significantly different than the influent ozonated water, while BAC10 achieved slight, but statistically significant reduction.
Outside of significant changes to EBCT, further reductions of NBDOC were found to rely on optimizing the ozonation process (adjustment of ozone dosing) and subsequent BAC filtration, resulting in decreased BDOC loading onto the GAC contactors and maximizing availability of adsorption sites for NBDOC.This important interplay between BAC and GAC performance, especially as it relates to NBDOC removal, was highlighted by the removal dynamics observed throughout the pilot treatment system.Of the total carbon removed during GAC treatment, 68.3% was biodegradable at the lower EBCT (GAC-10) while that percentage was decreased to 55.1% at the higher EBCT (GAC-20) with improved BDOC removal in the preceding BAC-10 filter.Considering that total DOC removal throughout both GAC contactors was fairly consistent, the increased reduction of BDOC during BAC filtration allowed GAC-20 to more effectively remove NBDOC.Theoretically, the complete removal of BDOC prior to GAC, that is, through optimized BAC filtration, should result in the lowest effluent DOC concentration and greatest reductions in NBDOC.This should extend the lifespan of the GAC, which, as time goes on, will lose adsorptive capacity and eventually functionally transition into BAC filters and need to be regenerated or replaced (Servais et al., 1994).In the present study, the GAC units were operating primarily beyond 20,000 bed volumes, at which point preliminary testing (Appendix A in Supporting Information) and outside review (Peterson & Summers, 2021) indicated that adsorptive capacity is largely spent as GACs transition to BACs.Yet, the continued NBDOC removal achieved by GAC contacting revealed through the BDOC analysis in this study suggests that substantial adsorption was still occurring well beyond 20,000 BV and notably more than either BAC filter.Therefore, notable performance differences exist between BAC filters operated in series with GAC contactors subjected to extended operation.The authors hypothesize that the operational differences between both treatment processes select for distinct microbial communities and functional gene profiles that subsequently result in distinct treatment performance.For example, it is possible that GAC-adapted microbes degrade sorbed BDOC, effectively regenerating sorption sites.Determining the mechanistic nature of longer term NBDOC removal within GAC contactors would be of interest for future research.

| Correspondence of organic carbon profiling to water quality data and ability to assess operational changes and diagnose upset events
The above trends and analyses were integrated across nearly 18 months of operation, including various operational changes and upsets (Table 1), indicating that overall organic carbon removal performance was robust in the face of prolonged challenge testing.However, it was of interest to determine if the organic carbon profiling approach applied here could effectively discern operational changes and diagnose upsets.To examine whether this was the case, correlations were assessed between TOC, DOC, BDOC, NBDOC, and POC measurements and available water quality and operational data.The full results of this analysis are available in the supplemental spreadsheet, with a summary of key findings with respect to each stage of treatment summarized below.

| Pilot influent
Among non-upset sampling events, the typical percent DOC in the influent represented by BDOC ranged from 25.7% to 41.5% (Table S6).However, during the methanol leaching events associated with SUMeth02 and SUMeth04, the pilot influent's DOC fraction was 55.9% and 56.2% BDOC, respectively.The conclusion that the percent influent BDOC was elevated by methanol was further supported by the fact that the NBDOC displayed minimal variability during these events and throughout operation (average NBDOC = 4.62 mg/L, SD = 0.459 mg/L, n = 8).Overall, the BDOC analysis provided insights into upstream plant performance and the nature of upsets that more conventional TOC or DOC measurements would not be able to capture.Apart from the two sampling events influenced by methanol carry-over, influent concentrations of TOC, DOC, BDOC, and BDOC:NBDOC ratio were relatively uniform and did not correlate with any water quality parameters.

| Floc-Sed
Floc-Sed was operated with effluent turbidity limits set to 0.5 NTU with online measurements taken every 5 min via integrated sensors.Coagulation was facilitated by dosing aluminum chlorohydrate (ACH) and a cationic polymer at 35 and 0.75 mg/L, respectively.Floc-Sed operation was consistent throughout the testing period, except for one major upset event, SUTurbid09, where the aluminum chlorohydrate stock ran out and temporary adjustments were made to the coagulant dosing to compensate.Even with consistent operation, performance was noticeably variable during initial SUmeOH periods, especially with regards to BDOC removal, underscoring the importance of influent characteristics.The doses applied were found to be positively correlated with the change in POC concentration (Pearson: r = .80),with doses >32-35 mg/L resulting in increased POC in the filter effluent.Further, the change in the percent BDOC through Floc-Sed was negatively correlated with influent turbidity (Pearson: r = À.87) and turbidity removal achieved through Floc-Sed (Pearson: r = À.92).Correspondingly, the change in the NBDOC fraction was inversely correlated.Overall, the results demonstrate that elevated turbidity was effectively managed by the Floc-Sed stage, while BDOC fractions were disproportionately decreased when turbidity was elevated, relative to the NBDOC fraction.The authors hypothesize that improved BDOC removal, relative to NBDOC removal, was related to the composition of natural organic matter (SUVA values less than 2 L/mg-m) and potential contributions to BDOC from upstream denitrification (e.g., low levels of remaining methanol and cellular debris from the transition away from anaerobic denitrification).More specifically, the composition of natural organic matter lacking hydrophobic, refractory humic substances may allow for more efficient removal of BDOC though additional research would be of interest.

| Ozone
Ozone dosing was controlled via integrated sensors with targeted O3:TOC ratios identified in Table 1.Ozone dosing, off-gas concentrations, transferred dose, and ozone residual were measured every 5 min via integrated sensors.Change in the percent BDOC following ozone treatment was positively correlated with three operational indicators: ozone:TOC (Pearson: r = .78),ozone:TOC with nitrite adjustment (Pearson: r = .81),and applied ozone dose (Pearson: r = .80).These strong correlations are consistent with the understanding that elevated ozone increases conversion of organic matter to a biodegradable form (Hammes et al., 2006;Loganathan et al., 2022;Reungoat et al., 2012).Nitrite is a known ozone sink (1 mg/L of nitrite consumes 3.43 mg/L of ozone) (Lee et al., 2013;Li et al., 2015;Stapf et al., 2016), and factoring in the corresponding reduction in overall oxidative capacity resulted in the strongest correlation.
Change in NBDOC concentration was found to be positively correlated with influent turbidity (Pearson: r = .83),where transformation of NBDOC was less effective at higher influent turbidity.This suggests either that the influent POC is not very biodegradable or that POC imparted an ozone demand and diminished efficacy against NBDOC.The effluent NBDOC was also negatively correlated with the ozone:TOC dose (Pearson: r = À.72), consistent with less effective conversion to BDOC at lower ozone dose.

| BAC filtration
BAC filtration relies on microbial treatment and thus was expected to be the most sensitive to changing conditions in the system.The BAC filtration units therefore were tested under a range of operating conditions (Table 1).At this point in the treatment train, flow was split between the BAC5 and BAC10 treatment units, enabling assessment of effects of EBCT.BAC filters had effluent turbidity and headloss limits set at 0.3 NTU and 10 ft, respectively.Both were monitored via integrated sensors at 5 min intervals along with filter runtime, filter loading, and filter levels.Automated back washing was conducted if either effluent turbidity or maximum headloss were exceeded.Both BAC units showed similar trends over time and indicated consistent nitrification, that is, ammonia being removed and nitrite/nitrate being produced.In terms of organic carbon removal, BAC10 performed consistently better than BAC5.Also, correlation analysis of the BAC10 filter revealed a wider range of relationships between organic carbon profiles and water quality/ operational data than the BAC5 filter.This suggests that the lower EBCT was not as stable, making it more vulnerable to high loading rates and shifting influent conditions, resulting in greater variability in the data.
Correlation analysis of the BAC10 unit revealed a strong and unexpected relationship with nitrogen species.Specifically, changes in BDOC concentrations were negatively correlated with changes in NH 4 + concentration (Pearson: r = À.94) and total inorganic nitrogen concentration (Pearson: r = À.84) and positively correlated with effluent NO 3 À concentration (Pearson: r = .77)and its change during treatment (Pearson: r = .8436).In other words, BDOC removal was enhanced when there was less removal of NH 4 + and total inorganic nitrogen.The fact that the correlation was also strong with NO 3 À suggests that the driver of this relationship is not merely physical removal of NH 4 + and total inorganic nitrogen, but biological nitrification.While numerous studies have investigated ammonia removal during BAC filtration, most have focused on drinking water applications (Basu et al., 2016) with fewer indexed to reuse contexts (dos Santos & Daniel, 2020).Studies examining enhanced organics removal through nutrient supplementation have indicated mixed results (Hamidi et al., 2020), with studies reporting both positive (Dhawan et al., 2017;Granger et al., 2014) and negligible impacts (Azzeh et al., 2015;Fu et al., 2017;Nemani et al., 2018) on performance.In the present study, the authors hypothesize that the development of an efficient nitrifying community may negatively impact the metabolic activity of heterotrophic bacteria primarily responsible for organics biodegradation through the creation of a nitrogen limited environment.Percent removal of DOC was also correlated with the change (removal) in NH 4 + concentration during BAC10 treatment (Pearson: r = .77),temperature (Pearson: r = .79),and monochloramine concentrations (Pearson: r = .89).The latter relationship suggests that monochloramine, when coupled with dechlorination prior to BAC filtration, may have been more of an indicator of residual ammonia reducing competition for nitrogen between heterotrophs and nitrifiers, than acting as a disinfectant.However, when monochloramine was unquenched, BDOC removal was negatively impacted by the loading of monochloramine directly onto the filter (SSNoQuench-10).
In the BAC5 filter, the effluent BDOC concentration was also positively correlated with both monochloramine (Pearson: r = .87)and ammonia concentrations (Pearson: r = .88),while the ammonia concentration was further positively correlated with effluent TOC (Pearson: r = .85).No operational parameter or conditions correlated strongly with treatment efficiency or bulk reductions of organic carbon.
Upset events provided the opportunity to assess the resilience of BAC filtration to changing operating conditions and to gain deeper insight into functionality of the units.During the higher BDOC loadings associated with the methanol carry-over event (SUMeth02 and SUMeth04), effluent TOC and DOC were elevated.However, net DOC removal was also elevated, indicating that the treatment train had the capacity to absorb the shock of the elevated BDOC load, with comparable BDOC across sampling events from the effluent of the BAC filters onward (Figure S4).Effluent TOC and DOC were also elevated following Floc-Sed upset, which subjected the filters to increased influent turbidity (SUTurbid09).In this case, DOC removal was negatively impacted, but BAC10 was successfully able to continue treatment at a reduced rate, highlighting the robustness of BAC filters operated at longer EBCT.BAC5 was less successful at continued treatment, highlighting the susceptibility of shorter EBCT to operational variability and the need for more effective controls.There were also two instances of unintentional media loss from BAC5, following errors in backwashing.In both instances, media were replaced and/or the influent flow rate was adjusted to normalize EBCT.After the first instance of media loss (April 01, 2017), media from BAC10 were transferred into BAC5 prior to adjusting both filters flow rate to keep EBCT consistent.Following the transfer of media, SS-05 through SS-08 experienced increased removal of DOC and treatment efficiencies within BAC5 (Figure 2) by approximately 0.24 mg/L and 12%, respectively.BAC5's improved performance decreased following the upset of the Floc-Sed process (SUPolymerTurbid-09) and loading of monochloramine without quenching onto the filters (SSNoQunech-10).Conversely, BAC10's performance remained relatively unchanged.
Outside of process upsets, performance may also be linked to changes in operational parameters, as each sampling event with measurable changes in removal of the various organic carbon fractions was associated with changes in treatment, such as chlorine loading onto the biofilters (SUMeth02 and SS10), high turbidity (SUTurbid09), and filter polymer addition (SUTurbid09 and SSPolymer10).Notably, filter polymer addition was associated with elevated percent POC in the BAC10 (variability in data likely masked this relationship for BAC5), potentially the result of increased biofilm development on polymer scaffolding.BDOC removal was most efficient during the stable operation period associated with SS05 through SS08, resulting in the lowest effluent TOC and DOC concentrations.Peak performance occurred during SS05 and SS06, when the temperature also peaked (>26 C) (all other sampling events were <18 F), which likely relates to elevated microbial metabolism with elevated temperature.Just prior to SS07, temperature control was implemented at 15 C to eliminate this variable and create a "worst" case for process operation (especially biodegradation kinetics), but BDOC removal remained high.Nutrients were also supplemented at this point, to ensure that the BAC filters were not nutrient-limited, but it is not possible to evaluate the extent that each variable affected BDOC removal because they were changed at the same time.
Benchmarking of BDOC results to operational conditions and water quality data proved useful in explaining BAC performance variability.Other studies examining the fate of individual trace organic contaminants via ozonation and biofiltration have similarly found that removal depends on several operational and environmental conditions.These include the following: • Upstream ozone dose where increased dosing resulted in greater breakdown of organic contaminants (Arnold et al., 2018;Gifford et al., 2018;Huber et al., 2003;Lee et al., 2013;Reungoat et al., 2012;Snyder et al., 2006;Sundaram & Pagilla, 2020;Ternes et al., 2002).• Upstream chemical additions such as prechlorination which has been identified to suppress biological degradation (Liu et al., 2001;Urfer et al., 1997) and peroxide addition which increased oxidative action of microorganisms (Lauderdale et al., 2012).• Decreased temperatures which stifle microbial activity and decrease biological degradation (Basu et al., 2016).• Media type where GAC has been identified as providing a more robust alternative to sand or anthracite media (Arnold et al., 2018;Basu et al., 2016;LeChevallier et al., 1992).• Filter age where GAC contacting is notably more effective prior to exhaustion, but depending on water quality goals can still provide effective organics removal over prolonged operation (Sun et al., 2018).• Nutrient concentrations which have been both improved biofilter performance in some applications (Dhawan et al., 2017;Granger et al., 2014;Lauderdale et al., 2012) and resulted in negligible improvement in others (Azzeh et al., 2015;Fu et al., 2017;Nemani et al., 2018).• Longer EBCTs which typically improve removals with diminishing returns (Arnold et al., 2018;Müller et al., 2017;Sundaram & Pagilla, 2020;Zearley & Summers, 2012).
Here it was found that BAC performance was strongly tied to EBCT and nitrification as well as the avoidance of operational upsets and limiting of microbial toxic compounds.Specifically, carbon removal during BAC5 was decreased during the Floc-Sed process upset (SUPolymerTurbid-09) and chlorine loading without quenching (SSNoQuench-10).The lower EBCT proved to be less resilient to nonideal conditions.This was further supported by the result that organic carbon removal, the primary focus of BAC filtrations, was negatively correlated with ammonia oxidation and ultimately nitrification, potentially because of competition for ammonia.This highlights the sensitivity of biological treatment systems to factors such as temperature and nitrogen.In particular, NH 4 + availability appeared to be key, while monochloramine acted both as an inhibitory compound, but also as an indicator of the availability of NH 4 +.Interestingly, BAC5's performance was improved and sustained following the transfer of media from BAC10 after media loss occurred prior to SS-05.This observation warrants additional research, as it is plausible that the transferred media seeded BAC5 with a more established, efficient microbial community resulting in improved DOC removal.

| GAC contacting
Organic carbon profiling corroborated the understanding that the GAC contactors are strongly influenced by performance of the BAC filters.GAC contactors were subjected to similar operational controls and measurements as the BAC filters with additional manual monitoring of effluent TOC (4 mg/L threshold) and other contaminants of interest.Both GAC contactors exhibited elevated effluent TOC and DOC concentration during SUMeth02, SUMeth04, SUTurbid09, and SSPolymer10, highlighting the importance of optimizing upstream treatment for GAC contactors to perform optimally.Otherwise, both GAC contactors provided consistent organic carbon removal throughout the testing period.
BDOC analysis allowed estimation of NBDOC removal, that is, the intended purpose of the GAC units.When combining the data from both GAC units, NBDOC removal negatively correlated with average GAC filter runtime (Pearson: r = À.66) and temperature (Pearson: r = À.71).GAC20's effluent NBDOC also negatively correlated with average filter runtime (Pearson: r = À.88) and temperature (Pearson: r = À.87) and additionally was positively correlated with NO 2 À concentration (Pearson: r = .88).GAC contactor bed volumes were also negatively correlated with DOC removal (Pearson: r = À.62) and treatment efficiency (Pearson: r = À.75).These results are consistent with well-established principles of GAC exhaustion being driven by the total loading of organic carbon in the influent with time.
Similarly to BAC filters, BDOC removal by the GAC contactors was correlated with the fate of nitrogen species.Interestingly, change in percent NBDOC positively correlated with the change in NO 3 À and total inorganic nitrogen concentrations for both GAC units, when assessed independently and jointly (Pearson: r = À.89 and .81,for NO 3 À and total inorganic nitrogen, respectively; p and r values presented for the joint analysis, all values reported in SI materials).Additionally, change in BDOC concentrations similarly correlated with changes in NO 3 À (Pearson: r = .92)and total inorganic nitrogen concentration (Pearson: r = .87).Treatment efficiency (percent DOC removal relative to influent DOC) negatively correlated with changes in NO 3 À (Pearson: r = À.93) and total inorganic nitrogen (Pearson: r = À.79) concentration.Relationships between BDOC and nitrogen removal suggest that although the primary purpose of the GAC contactors is to remove NBDOC via sorption, there was some degree of biological removal occurring as well.

| CONCLUSIONS
Here the authors refined and tested a BDOC protocol adapted for water reuse treatment trains employing biological removal via media with attached growth.Overall, this study demonstrated that BDOC testing can provide critical insight into drivers of organic carbon removal via carbon-based AWT, which can further inform efforts to improve reliability and performance for producing highquality water for potable reuse.Unlike traditional TOC and DOC analysis, BDOC analysis proved to be capable of yielding key insights into the fate of biodegradable and nonbiodegradable fractions of dissolved organic carbon during various stages of AWT.BDOC analysis validated some fundamental expectations regarding carbon removal by various processes but also generated new insights and hypotheses.For example: • Modest BDOC removal was achieved during Floc-Sed, an effect previously reported for drinking water treatment (C.Volk et al., 2000), but BDOC was disproportionately reduced relative to NBDOC.• Ozonation did not reduce DOC, but could reduce NBDOC (average reduction of 1.07 mg/L) by proportionally increasing BDOC, as expected, by 0.36-1.81mg/L.• BAC filtration effectively removed BDOC (average reduction of 0.82 mg/L), but did not significantly remove NBDOC (average reduction of 0.07 mg/L), as expected.
• GAC contacting was unexpectedly able to achieve 0.53 mg/L of average reduction in NBDOC, even with >20,000 treated bed volumes.However, inefficient BDOC removal during BAC treatment will detract from the ability of GAC to remove NBDOC.• Improved performance was achieved at a longer EBCT.
The longer EBCT treatment train (BAC10/GAC20) outperformed the shorter EBCT treatment train in terms of overall effluent BDOC concentrations (0.76 mg/L vs. 1.21 mg/L in the effluent, respectively).• Benchmarking of BDOC measurements to operational conditions proved useful in identifying process upsets and explaining BAC performance variability.For example, BDOC analysis was effective at identifying elevated BDOC entering the pilot after methanol leaching upstream.Further, it provided insights into BAC filtrations resiliency to upsets by tracking the fate of these elevated concentrations and subsequent elevated removal rates during filtration.BAC5's performance was also improved following the inoculation of media from BAC10.Generally, BAC performance and BDOC removals were strongly tied to EBCT, the avoidance of elevated turbidity in Floc-Sed effluent, and limiting microbial toxic compounds (e.g., chlorine concentrations).• Correlation of water quality data to BDOC measurements were also found to provide valuable insights into process performance.For example, significant relationships were found between BDOC removal and nitrification/monochloramine concentrations in both BAC5 and BAC10.During ozonation, change in BDOC percentage was found to be positively correlated with multiple indicators for ozone dosing and ratios of ozone to TOC.With respect to NBDOC removal, GAC contactor performance was found to be negatively correlated with markers for contactor runtime while BDOC removal was correlated to nitrogen concentrations (similarly to the BAC filters), bringing to light unexpected biological removal of organic carbon occurring during GAC treatment.
Ideally, a carbon-based AWT train should produce finished water that is as biologically stable as possible (Ren & Chen, 2021;Sambo et al., 2020), which can best be assessed by BDOC, rather than TOC or DOC.Also, BDOC analysis can be applied to ensure that upstream BDOC removal is maximized prior to GAC treatment, thus optimizing lifespan of the GAC and minimizing the cost needed for regeneration or replacement.It should be noted, however, that BDOC analysis requires substantially more time (15-30 days vs. <1 h) and effort (<1.5 h vs. <0.5 h) than conventional TOC analysis.Therefore, its application should ideally be evaluated and completed in conjunction with conventional methods.

F
I G U R E 1 Hampton Roads Sanitation District SWIFT Pilot Facility's process schematic.Sampling locations indicated with red arrows.The pilot facility was fed tertiary treated denitrified effluent from the York River wastewater treatment plant at a flowrate of 4.3 gallons per minute.

F
I G U R E 2 (a) Total organic carbon (TOC) and (b) dissolved organic carbon (DOC) concentrations through the pilot over the six sampling points.DOC removal of each treatment stage relative to the prior treatment stage for all 10 sampling events: (c) mass change of DOC (mg/L) and (d) percent DOC removal relative to DOC measured in prior treatment stage effluent.

F
I G U R E 3 Dissolved organic carbon (DOC) measured in the effluent of each indicated treatment stage along the treatment train that is nonbiodegradable or biodegradable.The gray shaded area, "Treatment Removal," indicates the portion of DOC removed relative to the prior treatment stage.The line plot corresponds to the secondary y-axis and indicates the Percent Biodegradable DOC in the effluent of the indicated treatment stage.Exemplar sampling events include (a) SUMeOH-02, representative an upset condition impacted by upstream methanol leaching and (b) SS-05, representative of stable operation.F I G U R E 4 Biodegradable organic carbon (BDOC) and non-biodegradable organic carbon (NBDOC) measurements at each treatment stage: (a) BDOC concentration (mg/L), (b) change in BDOC concentration (mg/L) relative to prior stage of treatment, (c) NBDOC concentration (mg/L), (d) change in NBDOC concentration (mg/L) relative to the prior stage of treatment, (e) percentage (%) of BDOC, relative to DOC, remaining after treatment, and (f) change in the percentage (%) of BDOC achieved by each stage of treatment.
Overview of sampling events, operational conditions, and upset events.
a Not included in analysis.