In hydrologic environments, physical and geochemical variables such as TSS, sediment grain size, sediment %moisture and sediment total organic carbon (%TOC) are all important parameters to consider when evaluating the presence, dispersal and distribution of micropollutants. Sediment %moisture varied from 30% (site 6) to 67% (site 4), with an overall average of 52% (Fig. S4). Sediment %moisture showed a weak inverse correlation (Spearman’s Rho 0.44, p<0.05) with sediment grain size. TSS was measured at all sites where surface water was collected (Fig. S5). The detection limit for TSS was determined to be 0.1 mg/L. TSS values ranged from 0.11–261 mg/L with an overall median average of 26.3 mg/L. There was no significant difference in TSS among all sampling sites (Kruskal-Wallis, p>0.05). Sediment texture varied both spatially and temporally in the TFPR (Fig S6). The sediments were predominantly classified as sandy-silt for Hunting Creek (sites 6-10) and as silty for Four Mile Run (sites 3 and 4) and Gunston Cove locations (sites 12 and 13). However, there was a significant difference in grain size among the sediments across all sites (Kruskal-Wallis, p<0.05); for upstream Hunting Creek (site 5), a greater percentage of sand composition relative to the other downstream sites was observed. %TOC in sediment (Fig. S7) varied minimally both spatially and temporally in the TFPR, ranging from 1.2 - 2.0% (wt/wt) with a mean value of 1.6% across all the sediment sites. There was no statistical difference in %TOC among the sites (Kruskal-Wallis, p>0.05).
PPCP Quantitation Frequency
The quantitation frequencies observed for the PPCPs in both surface water and sediments are provided in Table S5. Overall, 33 of 91 total PPCPs were quantified in water (S33-wPPCP) and 39 of 91 were quantified in sediments (S39-sPPCP) collected in the TFPR. Altogether, 52 of 91 PPCPs were quantified in either water or sediments. The remaining 39 PPCPs were not quantified in either matrix. The PPCPs were grouped by quantitation frequency (QF) into high (³60%), moderate (³40%), and low (£25%) categories in any single matrix to characterize abundance (Table S5). Three PPCPs were quantified at high frequency in both matrices, including fexofenadine, metoprolol, and tramadol. Those PPCPs detected primarily in water (³60%) included sulfamethoxazole, caffeine, nicotine, and carbamazepine. Conversely, the PPCPs found extensively in sediments (³60%) included diphenhydramine, escitalopram, methadone, sertraline, and fluoxetine. High QF PPCPs were those commonly quantified in both matrices, moderate frequency PPCPs were those quantified primarily in a single matrix, but not both, matrices, and low frequency PPCPs were rarely quantified £25% in either matrix.
Spatial Distribution of SPPCPs
The S33-wPPCP concentrations (ng/L) showed significant differences in surface water spatially along the TFPR (Kruskal-Wallis, p<0.05), with the greatest median concentrations observed near the WTP outfalls at sites 3 and 6 (Fig. 1a). The lowest S33-wPPCP median concentrations occurred in the mainstem Potomac River or at fluvial sites upstream from the WTP outfalls. The five mainstem TFPR sites (1,9,10, 13 and 14) showed significant differences in surface water concentrations (Kruskal-Wallis, p>0.05), with site 14 being an outlier. However, when site 14 was removed the four remaining mainstem locations were not significantly different in median S33-wPPCP concentrations (Kruskal-Wallis, p<0.05). Site 14 was clearly influenced by the presence of a WTP located on the Quantico Marine Base, which discharge into Quantico Creek upstream of the sampling point. The mainstem TFPR showed the smallest median S33-wPPCP concentrations (sites 1, 9, 10 and 13), ranging from 75-113 ng/L, with the fluvial entry point to the TFPR at Chain Bridge showing the greatest upper-end-member concentration among the mainstem locations. It appeared the median concentrations decreased slightly in the downstream direction in the mainstem TFPR, although the differences were not statistically significant. This observation was likely because of dilution arising from additional surface water contributed by nine tributaries flowing into the mainstem TFPR, adding ~4% of additional flow volume below Chain Bridge. Although the high-capacity WTPs
increased surface water concentrations of PPCPs in the immediate vicinity of the outfall zones, with a downstream influence of ~1 km radius, the WTPs did not appear to alter the concentrations of S33-wPPCP in the mainstem Potomac River along the entire length of the TFPR from Chain Bridge to Quantico. The concentrations of S33-wPPCP in the mainstem of the TRPR were consistent along its longitudinal axis, and any net increase or decrease was undetectable.
The flow-weighted average (FWA) concentrations of the PPCPs showed good agreement with the time-weighted median (TWM) concentrations as shown in Fig. 1a. That FWA » TWM implies a flat C-Q relationship between solute concentration and stream flow (Chorover et al. 2017; Moatar et al. 2017), which is not unexpected since all sampling occurred during predominantly baseflow conditions from May through September 2018.
The S39-sPPCP concentrations in sediment also varied spatially, with the greatest concentrations observed within a 1 km radius of WTP discharge (Fig 1b). However, the distinct feature of sediment was that concentrations maximized in the embayments and not directly adjacent to the discharge zone as shown in Fig. 1b. The sediment spatial presence is best exemplified by Hunting Creek (sites 6-8), whereby the maximum concentrations were found in the deposition zone of the Lower Hunting Creek (sites 7-8). The deposition zone occurs where Lower Hunting Creek empties into its shoal and forms a bayhead delta. This sediment deposition zone clearly traps PPCPs emerging from WTP discharge undergoing downstream transport. Such a sedimentary process creates greater ecotoxicological risk in the benthic shoal community in the TFPR from PPCPs entering through the tributaries. However, the presence of relatively large S39-sPPCP concentrations in the Lower Hunting Creek shoal was a localized phenomenon because the mainstem TFPR sediments were much lower in concentration relative to the embayments (Fig. 1b).
Individual PPCPs in WTP Effluent and Nearby Discharge Zones
The PPCPs were evaluated individually by box plots, and in this case only those with quantitation frequencies >50% were considered in statistical analysis. Fexofenadine, an antihistamine, and caffeine had the greatest concentrations in the surface water samples followed by desvenlafaxine (antidepressant), nicotine (stimulant), MDA (illicit), hydrochlorothiazide (diuretic), metoprolol (b-blocker) and sulfamethoxazole (antibiotic), all present with median concentrations greater than 20 ng/L (Figs. 2 and 3). Carbamazepine (anticonvulsant) and bupropion (antidepressant) were observed with median concentrations greater than 10 ng/L. The remainder of the individual PPCPs in surface water were found at concentrations less than 10 ng/L.
The PPCPs measured in the effluents of WTP 1 and WTP 2 are illustrated in Figs. S8 and S9. The individual PPCP concentrations (log transformed) were well correlated between the two effluents (Spearman's Rho = 0.861, p<<0.05), indicating the PPCP constituent compositions among the two waste treatment discharges were similar, which was expected given the similar sociodemographics of the two sewer-sheds in Northern Virginia. But, the total PPCP concentrations were generally greater in WTP 2 effluent. The important implication of this observation is that the PPCP chemical-specific inputs into the TFPR were seemingly consistent among the major WTPs. However, the composition of PPCPs in WTP effluents differed significantly from the PPCP compositions observed in the corresponding TFPR discharge zones (Figs. 2 and 3), which was true for WTP 1 and site 3 (Spearman's Rho = 0.15, p>0.05) and WTP 2 and site 6 (Spearman's Rho = 0.27, p>0.05). There were 43 PPCPs quantified in the WTP effluents and 33 PPCPs correspondingly quantified in the WTP discharge zones. PPCPs were detected at concentrations up 14,000 ng/L in WTP effluent, while averaging up to 400 ng/L in discharge waters, yielding a ~40-fold or greater dilution upon entering the TFPR.
The substantial change in chemical composition of the PPCPs between effluent and nearby receiving waters shows that dispersal forces and reactivity act on PPCPs rapidly following discharge. The most likely physical processes that act on PPCPs immediately are dilution and sorption in geosolids, with air-water partitioning likely occurring to a minor extent with respect to the total PPCP composition. PPCPs for the most part have negligible Henry’s law constants. However, some PPCPs are highly particle-reactive, and it has been shown previously that substantial PPCP fluxes occur into bed sediments within the discharge zone of Hunting Creek (site 6) (Foster and Leahigh 2021). Degradative pathways such as biotransformation, hydrolysis and photolysis are also likely to alter the PPCP compositions in surface waters. For example, photolysis of hydrochlorothiazide, which was a predominant PPCP in our study, occurs rapidly in water with a half-life of 0.43 hr (Baena-Nogueras et al. 2017). Diphenhydramine, fexofenadine, sertraline, and escitalopram have previously been shown to be highly sorptive in sediments as reported by other studies (Stein et al. 2008; Li et al. 2011; Yu et al. 2013; Xu et al. 2021). Atenolol, metoprolol, caffeine and carbamazepine can be rapidly degraded by residual chlorine alone or in combination with UV-light (Cheng et al. 2019). Effluent has very moderate residual chlorine concentrations, which is used as a disinfectant in tertiary treatment at WTPs. Ranitidine is rapidly transformed into a nitrosamine by-product in the presence of chlorine and UV-light (Seid et al. 2021). Bupropion undergoes rapid hydrolytic degradation in aqueous solution at pH >5 to its most prominent degradation pathway that involves a hydroxide-catalyzed catalysis of the neutral base form (O’Byrne et al. 2010). The pH of receiving waters reported for the TFPR estuary have ranged from 6.8 to 7.8 depending on the season (Jones et al. 2014) promoting hydrolysis. All these examples show how geochemical distribution and partitioning along with degradative forces act on PPCPs discharged into surface waters, contributing in many cases to rapid and extensive alterations in chemical compositions.
PPCP Composition in Surface Water and Sediment
There were 18 individual PPCPs detected at concentrations above the QL in >50% of the samples, which included nicotine, caffeine, triamterene, metoprolol, tramadol, desvenlafaxine, bupropion, sulfamethoxazole, dextromethorphan, venlafaxine, diphenhydramine, carbamazepine epoxide, fexofenadine, carbamazepine, methadone, and celecoxib. In addition, there were several PPCPs detected in Hunting Creek (sites 5-8) and Four Mile Run (sites 2-4) that were not present at Gunston Cove (sites 11-13). These PPCPs included cotinine, atenolol, propranolol, diltiazem, hydrochlorothiazide, and furosemide. The PPCP MDA was found exclusively in surface water samples from the Four Mile Run area. There were no individual PPCPs that were unique to only the Gunston Cove area (sites 11-13).
The identity of PPCPs found at sites remote from the WTP discharge zones in the TFPR were identical to the individual PPCPs shown in Figs. 2 and 3, but with fewer of them and at lower concentrations. Along with a reduction in the concentrations of PPCPs remote from the WTP discharge zone, it was observed the compositional profile of PPCPs also changed markedly. PPCP concentrations in water were converted to mole fraction-PPCP concentrations and analyzed by principal component analysis (PCA) to visualize chemical composition trends. The compositional profiles of the PPCPs varied substantially throughout the TFPR, which is demonstrated by the PCA plot (Fig. 4) highlighting a spread of eigenvalues across a compositional arc beginning with the WTP effluents and culminating in the mainstem TFPR and upland locations (lower right quadrant of PCA). The PCA can be divided into 4 compositional segments (Fig. 4), including (i) the WTP effluent from WTPs 1 and 2 (upper left quadrant in Fig. 4), (ii) WTP discharge zone, (iii) transition mixture where the embayments flow into the mainstem TFPR and (iv) the mainstem TFPR itself in combination with the upland watersheds (lower right quadrant). The first two PCAs accounted for 87% of the compositional variability. The sites that cluster in the PCA zones are also shown in Fig. 4. The WTP effluent composition clusters in the upper left quadrant of the PCA plot, and as indicated above the two effluents were well correlated in PPCP composition. The TFPR sites nearest the WTP outfalls showed a composition on the PCA in closest proximity to the effluents (e.g., sites 3, 4, 6-9, 12 and 14), but trending down and to the right. The opposite end of the PCA in the lower right quadrant included the sites directly upstream (above the head of tide) from the WTP outfalls (e.g., sites 2, 5, and 11) and some of the mainstem TFPR sites (e.g., sites 1, 10 and 13). The mixed zone included several sites that were primarily mainstem sites (e.g., sites 9, 10, and 13) that trended upward toward the discharge zone in the PCA. The PCA was very useful in proving that site 14 was impacted by nearby WTP effluent at Quantico (site 14).
The composition of the PPCPs in surface water was distinctly different between the TFPR mainstem and the tributary discharge zones as described above. Further insight into the factors controlling the change in PPCP composition during downstream transport is further highlighted in Fig. 5. Desvenlafaxine, for example, was a major PPCP constituent in WTP effluent and was also prominent in the WTP discharge zone of Hunting Creek, but it was not detected in filtered surface water at sites 9, 10 or 13 in the mainstem TFPR. In addition, fexofenadine was replaced by caffeine as the most prominent PPCP in the mainstem TFPR. The likely explanation for this alteration in composition in the Hunting Creek region is sorption to geosolids followed by particle deposition within the Hunting Creek shoal. The sediment concentrations of PPCPs are shown in Fig. 6 with site 8, the site with the greatest PPCP observed concentrations in TFPR sediments, showing the dominant presence of desvenlafaxine and fexofenadine. The log Kow values of these PPCPs are listed in Table 1, where it is clear from inspection of the pKa values (also in Table 1) that a significant fraction of these particular PPCPs is expected to be positively charged at ambient pH based on the observed log Kow > log Dow in combination with at least one pKa > 7.6 in each case (7.6 was the average observed pH in surface water at this location). It is well known that a significant portion of PPCPs sorption occurs via mineral complexation of positively charged, protonated nitrogen species (Loeffler et al. 2005; Kiecak et al. 2019).
There were 16 PPCPs detected at concentrations above the QL in sediments at ³50% frequency. These included all the PPCPs shown in Fig. 6 for site 8. All sites showed PPCPs that were composed of subsets of these 16 constituents, except for Pohick Bay where triclocarban was quantified at <50% frequency in sediments. PPCP chemical composition in sediments diverged from that found in surface water. Further, the PPCPs detected in sediments showed no significant correlation (Spearman’s Rho = 0.15, p>0.05) between log Dow (n-octanol/water distribution constant at pH 7.4) and the measured conditional distribution constant, Kd-cond (Kd-cond = Cs/Cw for PPCP concentrations quantified in sediments (Cs) and surface water (Cw)). The dynamic interaction of PPCPs with sediments is only partially explained by log Dow because sorption to sediment through electrostatic complexation mechanisms occurs in addition to organic matter partitioning (Khetan and Collins 2007; Pan et al. 2009; Martínez-Hernández et al. 2014; Yamamoto et al. 2018). The Kd-cond estimates were often much larger than expected based upon log Dow, especially for PPCPs predicted to be positively charged at ambient pH. Another possible reason for lack of correlation with log Dow is because of rapid transformation that may be taking place in the environment (yielding low water concentrations). Furthermore, the sediment concentrations were not normalized to organic carbon levels because there was no observed correlation between Kd-cond and %TOC (Spearman’s Rho 0.10, p>0.05). It is generally assumed that organic micropollutants partition primarily into natural organic matter based on polarity and the (increasing) magnitude of Dow. Interactions of PPCPs between water and geosolids is a mixed complexation process, and the role the organic carbon plays in geochemical fate is not dominant.
Seasonal (i.e., May through September) differences in PPCP concentrations were exemplified in the TFPR for caffeine and fexofenadine (Kruskal-Wallis, p<0.05). Two distinct seasonal profiles were apparent in our study. The examples were derived from site 6 near WTP 2 effluent because this is where the greatest concentrations of PPCPs were observed in surface waters (Fig. 7). Mole fraction-PPCP concentrations are expressed in Fig. 7 to represent compositional changes in the PPCP mixture. Caffeine displayed reduced concentrations in June and July relative to May and September (Fig. 7). The caffeine profile was negatively correlated (Spearman’s rho = -0.68, p<0.05) with temperature, which was the only significantly correlated ancillary parameter identified. Other PPCPs that mirrored the caffeine seasonal profile included nicotine, metoprolol, atenolol, propranolol, celecoxib, desvenlafaxine, venlafaxine, triamterene and tramadol. Previous reports have documented the seasonal or annual variability in PPCP concentrations in surface waters (Sui et al. 2011; Sun et al. 2014; de Jesus Gaffney et al. 2017; Khasawneh and Palaniandy 2021; Singh and Suthar 2021), and suggest variables such as precipitation and stream flow, water temperature, WTP tertiary treatment technology, and dosage patterns influence seasonal variability. Another observed seasonal profile of concentration differences in surface water (Kruskal-Wallis, p<0.05) was demonstrated by fexofenadine (Fig. 7). Fexofenadine showed larger concentrations in May, June and July relative to September, which was a rather unique seasonal trend among this assemblage of PPCPs. Cotinine was the only other PPCP to show a similar seasonal trend to fexofenadine. The remaining PPCPs showed no significant seasonal trends. Sun et al. (2014) showed that caffeine in WTP receiving waters in China was present at the greatest concentrations in February, which was correlated with greater PPCP removal efficiencies from sewage treatment in the warmer months. Caffeine has shown a high removal efficiency (>90%) during sewage treatment indicating it is fairly labile (Kosma et al. 2014). However, even PPCPs previously reported to be recalcitrant to degradation by sewage treatment, such as metoprolol (de Jesus Gaffney et al. 2017; Khasawneh and Palaniandy 2021), showed a significantly lower composition in the summer months implying greater degradation in the WTPs during warmer temperatures.