Per-and poly-fluoroalkyl substances ( PFAS ) in river discharge : modeling 1 loads upstream and downstream of a PFAS manufacturing plant in the Cape 2 Fear watershed , North Carolina

Authors: Pétré M-A, Salk KR, Stapleton HM, Ferguson PL, Tait G, Obenour DR, Knappe DRU, 4 Genereux DP 5 6 Department of Marine, Earth, and Atmospheric Sciences, North Carolina State University, Raleigh, North 7 Carolina, United States 8 Now at Geological Survey of Finland, Espoo, Finland 9 Tetra Tech Center for Ecological Sciences, Research Triangle Park, North Carolina, United States 10 Nicholas School of the Environment, Duke University, Durham, North Carolina, United States 11 Department of Civil and Environmental Engineering, Duke University, Durham, North Carolina, United 12 States 13 Department of Civil, Construction, and Environmental Engineering, North Carolina State University, 14 Raleigh, North Carolina, United States 15 16 *Corresponding author: Marie-Amélie Pétré, marie-amelie.petre@gtk.fi 17

Environmental studies monitoring PFAS contamination in urban watersheds typically report PFAS 63 concentrations in surface waters, but PFAS load (riverine mass flux), the product of concentration and 64 river discharge, may be better suited to assessing and managing PFAS sources. PFAS load can be used to 65 quantify the mass of chemical passing monitoring stations and entering downstream waterbodies, such as 66 reservoirs and estuaries. Accurately estimating loads is challenging as it requires continuous monitoring of 67 river discharge and frequent co-located sampling of river water to capture the temporal variability in 68 PFAS concentration (Lee et al. 2019). 69 Previous studies have highlighted the limitations of estimating PFAS loads from rivers. In some previous 70 studies, PFAS load was based on the product of measured river PFAS concentration and long-term mean 71 river discharge for the month of PFAS sampling, rather than measured discharge at the time and place of 72 PFAS sampling (Ahrens et al. 2009, Pistocchi andLoos 2009, McLachlan et al. 2007). This could give 73 rise to error from temporal differences in river discharge, and from the locations of PFAS sampling 74 differing from the locations for the long-term average discharge values. Some recent studies have utilized 75 "snapshot" or seasonal sampling campaigns rather than frequent and long-term monitoring of PFAS, e.g., 76 Munoz To address the methodological challenges of characterizing the temporal variability in PFAS loads and 80 river discharge, we used a sampling scheme with relatively frequent (daily to weekly) PFAS sampling 81 conducted over a relatively long monitoring period (13 months at one station, 28 at another) for a 82 significant list of PFAS analytes (13 at one station, 43-57 at another), including perfluoroalkylsulfonic 83 acids (PFSA), perfluoroalkylcarboxylic acids (PFCA), per-and polyfluoroalkyl ether acids (PFEA), 84 fluorotelomer sulfonates (FTS) and sulfonamides. Critical to load estimation, PFAS sampling was 85 conducted in close proximity to continuous discharge gaging stations operated by the US Geological 86 Survey (USGS 2021). In addition, PFAS data were collected at or very near drinking water intakes in the 87 study area, providing relevance to PFAS exposure in the affected communities. 88 The study was undertaken in the Cape Fear River  to July-August 2020 and at the Kings Bluff raw water intake in the Cape Fear River between 12 116 September 2018 and 1 February 2021 (28 months) (Figures 1, S1). 117 In the Haw River watershed, 28-42 water samples were collected at each station and 13 PFAS were 118 targeted (Table 1). The sampling interval typically ranged from 6-8 days. Due to COVID-19, sample 119 collection was reduced to three stations between April 14 and June 22, 2020: Bynum, Burlington 120 Upstream, and Burlington Downstream ( Figure S1). The latter two stations are located directly upstream 121 and downstream of the Burlington wastewater treatment plant. The Bynum sampling station is adjacent to 122 the water intake for the city of Pittsboro, NC, and about 40 km downstream of Burlington Downstream. 123 Water samples were collected in 1-liter pre-cleaned polyethylene bottles, either by wading into the middle 124 of the channel to fill the bottle, or lowering a bucket from a bridge and then filling the bottle from the 125 bucket (at the Cane Creek sites only, Figure S1). Details on PFAS analyses and quality assurance/quality 126 control (QA/QC) protocols are provided in Texts S1 and S2. 127 The Kings Bluff sampling station is at the CFPUA water intake. It is located 88 km downriver of the 128 Fayetteville Works and delivers water to the CFPUA's Sweeney Water Treatment Plant in Wilmington. At 129 Kings Bluff, a total of 120 river water samples were collected by the utility and analyzed by a commercial 130 lab (Text S1 targeted (as part of the 43 PFAS targeted in total throughout the study), and the remaining 10 were 138 subsequently added to the list of analytes in September 2020 . The contribution of the PFAS associated  139  with the plant to the total quantified PFAS at Kings Bluff was calculated by dividing the sum of the 10  140  site-related PFAS concentrations by the total quantified PFAS concentration summed over the 43 PFAS  141  targeted during the entire study period at Kings Bluff, Σ43PFAS (for consistency over the 28-month  142  monitoring period, Σ43PFAS was used for this calculation even during late 2020 when 57 PFAS were  143 targeted). However, the calculation of the additional contribution of the 10 remaining compounds from 144 September 2020 is presented in section 3.2.2. 145 2.2 River discharge 146 Daily river discharge data were obtained from three USGS gaging stations (Figure 1, Figure S1) located:  Spearman's correlation coefficients ranging from 0.02 to 0.96 (Table S4). This suggests that PFCA and 213 PFSA in the Haw River originate from common (or similar) loading sources. 214 The highest Σ13PFAS measured in the Haw River basin (1,197 ng/L, Table S1) was found at Burlington 215 Downstream ( Figure S1) in September 2019. The lowest Σ13PFAS was found in samples collected in 216 Jordan Lake and at station CC1, the most upstream station on Cane Creek ( Figure S1). PFOA+PFOS at 217 CC2, CC3 and CC4 was higher than the USEPA Health Advisory Level ( respectively. FTS and sulfonamides were not detected, except on one sampling date each (October and  242 November 2020, respectively). Σ43PFAS ranged from 40 to 377 ng/L, with an average of 143 ng/L. Total 243 concentration of targeted PFEA ranged from 12 to 274 ng/L. GenX was detected in all samples with 244 concentrations from 3 to 76 ng/L (mean 14.8 ng/L), below the NC Health Goal of 140 ng/L (Table S3) River at Kings Bluff (Table S5)  conducted by Chemours identified a total of 257 unknown PFAS in their process wastewater samples and 288 discharge samples from locations "that may reach the Cape Fear River" (The Chemours Company, 2020). 289 Concentrations of the main PFEA found at Kings Bluff (GenX, PFMOAA and PFO2HxA) generally 290 followed the same temporal variations until mid-September 2020, but PFMOAA concentrations increased 291 noticeably after that ( Figure S4, CFPUA 2021). The causes of this increase are unclear and might be due 292 to a process at or near the Fayetteville Works, the mobilization of PFMOAA from groundwater, or a 293 combination of these and other factors. 294 It is possible that some PFAS reaching the river may become associated with river sediments and this may 295 affect the PFAS concentrations in river water (Harfmann et al. 2021). In addition, semi-labile PFAS such 296 as FTS and sulfonamides are precursor compounds and can transform during their transport in the river, 297 forming PFCA and PFSA as terminal products (Liu and Mejia Avendaño, 2013). These processes merit 298 further study in general; the extent of their influence on PFAS in the Cape Fear River is not fully known. 299 3.2.3 PFAS concentration relationships with river discharge 301 At both Bynum and Kings Bluff, total quantified PFAS concentration was negatively correlated with river 302 discharge in each study year. Discharge and PFAS concentration were negatively correlated across years 303 and sampling sites, indicating a diluting relationship (Figure 4). At Bynum, the concentration-discharge 304 relationship was not significantly different among years. Discharge and year explained more than half of 305 the variability in PFAS concentration at Bynum (Figure 4a; ANCOVA, F(5, 164) = 41.74, R 2 = 0.55). At 306 King's Bluff, the slope of the concentration-discharge relationship was not significantly different among 307 years, but the intercepts among years showed a decreasing trend over time, indicating that at a given 308 discharge, PFAS concentrations were expected to be higher in 2013 and 2018 than in 2019 and 2020. 309 Discharge and year explained 2/3 of the variability in PFAS concentration at King's Bluff (Figure 4b; 310 ANCOVA, F(7,141) = 43.77, R 2 = 0.67). 311 Thus, the overall PFAS concentration differed among years, but the impact of discharge on PFAS 312 concentration was remarkably similar across years. Also, for the mean discharge at Kings Bluff during the 313 study period (409 m 3 /s), the PFAS concentration given by each successive best-fit line is lower over time 314 (Figure 4b). This decreasing trend is consistent with the flow-weighted mean concentrations calculated at 315 Kings Bluff (Section 4.3). 316 317 3.3 Mass fluxes 318 At Kings Bluff, Σ43PFAS load (i.e., the cumulative river export of 43 PFAS from the watershed) 319 determined on the sampling dates ranged from 459 g/day to 17,300 g/day (mean 3,440 g/day). At Bynum, 320 measured Σ13PFAS load ranged from 28 to 949 g/day (mean 256 g/day). PFAS load generally increased 321 with increasing river discharge ( Figure S5). Despite the typically lower concentration during high flow, 322 the highest PFAS mass transport occurred at high discharge due to the higher volume of water moving 323 through the system. In particular, the Σ43PFAS load at Kings Bluff was highest (6,500-17,300 g/day) 324 during and their loading at Bynum was not sensitive to discharge at the Burlington wastewater treatment plant. 373 We estimated the PFAS load to the Haw river from the Burlington wastewater treatment plant by 374 subtracting the PFAS load at the "Burlington Upstream" station from that at the "Burlington Downstream" 375 station for the same 42 sampling days from 10 June 2019 to 20 July 2020. PFAS input to the Haw River 376 from the wastewater treatment plant was highly variable during this time, from 9 to 444 g/day (mean value 377 of 122 g/day). This variability in treatment plant effluent complicates the use of load estimation programs 378 such as LOADEST, especially for 10 of the 13 PFAS targeted in this study whose loads in the Haw River 379 are controlled partly by the wastewater treatment plant effluent.  (Table S7). 387 PFAS loads between Bynum and Kings Bluff were compared using daily load estimates from LOADEST 388 during the common monitoring period of the two stations (10 June 2019-20 July 2020) and including only 389 the 13 PFAS targeted at both Bynum and Kings Bluff (Table 1). Σ13PFAS load at Kings Bluff was 1,024 390 g/day on average (Figure 7), 3.6 times higher than in Bynum (285 g/day). The mean river discharge at 391 Kings Bluff was about four times higher than at Bynum. Thus, PFAS input to the Cape Fear River 392 between Bynum and Kings Bluff was estimated to be 739 g/day, including a substantial input of "legacy" 393 PFAS (558 g/day of PFCA+PFSA) and the PFEA input from the Fayetteville Works (181 g/day of GenX  downward trend is encouraging, the rate is slow. Both PFEA and legacy PFAS continue to reach the river 461 in significant quantities, and that seems likely to continue for years. 462 Persistent high PFEA at Kings Bluff, up to 3 years after the termination of process wastewater discharge 463 to the river at the Fayetteville Works, likely reflects the importance of discharge of contaminated 464 groundwater to the river and its tributaries (baseflow contribution). The occurrence and distribution of 465 legacy PFAS indicate continuing inputs to the river system despite the phase out of PFOS and PFOA 466 production over a decade ago in North America. The load estimation program LOADEST was a useful 467 tool in quantifying individual and total quantified PFAS loads at Kings Bluff, however, its use was limited 468 at the upstream Bynum station where PFAS levels in the river were affected by variable inputs from a 469 wastewater treatment plant. On average, 3.4 kg/day of total quantified PFAS (1,256 kg/year) passed the 470 Kings Bluff station on the Cape Fear River to enter coastal marine waters during the study period. 471 Continued long-term monitoring of PFAS concentration is recommended. Persistence of PFAS in surface 472 water and drinking water supply suggests that up to 1.5 million people in NC might be exposed, and raises 473 technical and financial challenges for drinking water utilities that are faced with costly treatment upgrades. 474

475
Funding for the Haw River sampling and analyses were supported by a Duke University Collaboratory 476 funding opportunity. We thank Sharon Zhang, Duncan Hay, Nick Herkert, Jake Greif, and Analise 477 Lindborg for their assistance in the field, laboratory, and data analysis portions of this work. The work      (Sun et al. 2016), and 2018-2020 (this study).

Figure 5 a) River discharge (m 3 /s) in the Cape Fear River at Kings Bluff, b) Σ43PFAS concentration (ng/L) and Σ57PFAS concentration (ng/L) and c) Observed and estimated Σ43PFAS load (kg/d) in the Cape Fear
River at Kings Bluff.   Table 1 for the list of PFAS.

Text S1 Chemicals and standards/ extraction and analysis
Water samples collected in the Haw River watershed were processed as in Herkert et al. (2020). Samples were stored in a 4°C refrigerator until analysis, and were filtered under vacuum using a glass fiber filter . Laboratory blanks (800 mL of LC-MS water) were processed in each batch of water samples. All samples were spiked with an isotopically labelled GenX [2,3,3,3-Tetrafluoro-2- (1,1,2

Text S2 Quality Assurance and Quality control
Laboratory processing blanks were included in every batch of samples. Method detection limits (MDL) were determined for each batch of samples and were calculated using three times the standard deviation of laboratory processing blanks. MDLs ranged from 0.02 for GenX to 1.1 ng/L for PFOS among the batches. Average recoveries for labelled PFAAs were 74%.