Legacy halogenated organic contaminants in urban-influenced Legacy halogenated organic contaminants in urban-influenced waters using passive polyethylene samplers: Emerging evidence waters using passive polyethylene samplers: Emerging evidence of anthropogenic land-use-based sources and ecological risks of anthropogenic land-use-based sources and ecological risks

Abstract


Introduction
Organochlorine pesticides (OCPs) and polychlorinated biphenyls (PCBs) are two classes of halogenated persistent organic pollutants (POPs) characterized by high lipophilicity, bioaccumulation, toxicity, and resistance to environmental degradation processes (Stringer & Johnston, 2001;Venier et al., 2016).They have been widely used in industrial, agricultural, commercial products from 1920s-1970s in the U.S. and been phased out under the Stockholm Convention (Choi et al., 2010).Despite the imposed ban, OCPs and PCBs are still residual and ubiquitous in the environment because of the ongoing land-based sources.PCBs might be released from improper disposal and leaks from hazardous waste (e.g.old PCB containing equipment, landfills) sites, as byproducts formed during waste incineration, in wastewater and from pigment manufacture in urban areas (Rodenburg et al., 2010).Since 2000s, non-Aroclor congeners, such as PCB-11 and PCB-209 were detected in a variety of environmental compartments, linked to azo pigments, (Hu et al., 2008;Basu et al., 2009;Anezaki et al., 2015;Jahnke & Hornbuckle, 2019).Sharing similar timeline as PCBs, OCPs have been widely used in agriculture worldwide for several decades (Barbiero et al., 2018).
Their secondary sources resulting from large burdens stored in contaminated soils and lateral remobilization associated with non-point sources across the built urban environment have become increasingly influential.Therefore, PCBs and OCPs are continually posing potential (eco-)toxicological threats and drawing extensive attention from scientists and the public as legacy POPs (Holma-Suutari et al., 2016;Hu et al., 2020).
Narragansett Bay (NB) is located on the Atlantic coast of Rhode Island, USA, which is an example of urbanized estuary area.PCBs and OCPs represent legacy contamination in NB, with PCBs remaining an important contaminant class in the aquatic environment (Häder et al., 2020).Previous studies showed that a history of industrial activities, such as textiles, local jewelry and metal working in the urban upper reaches of the bay, had contributed large quantities of PCBs to the bay's sediments (Desbonnet & Lee, 1991;Latimer & Quinn, 1996).PCBs in migratory fishes from NB had been found to consistently exceed the safe-eating guidelines by the U.S. EPA in this century, which indicated the elevated local inputs of PCBs (Morgan & Lohmann, 2010).DDT and Chlordane were likewise extensively used in homes and agriculture, possibly reaching the bay through run-off (Hartmann et al., 2005).Furthermore, anthropogenic activities with land use change were able to affect the river input, wastewater discharge, and airwater exchange of contaminants, creating a dynamic situation in which changes the concentrations in the watershed-estuary-bay area could alter the direction of the resulting flux (Zhao et al., 2018).Populations moving toward the more south rural areas along the shoreline of the bay had potential negative impacts on the mid and lower bay areas (Hartmann et al., 2005).Thus, understanding how freely dissolved PCBs and OCPs are distributed and affected by population pressures and land use changes spatially are important for assessing the risk to organisms and human that depend on Narragansett Bay.
The data discussed here were acquired by low density polyethylene sheets (LDPE), which had been successfully used as passive samplers and considered as an effective, simplest, and cheapest tool for monitoring gaseous and truly dissolved POPs.LDPEs were deployed in the surface water during June-July in 2014 to (i) investigate the spatial trends of the freely dissolved PCBs and OCPs in Narragansett watershed-coast area, (ii) identify their possible sources, (iii) examine the influence of the population and land use patterns on their spatial variations, and (iv) lastly deriving and assessing the PCB and OCP toxicity.

Materials and Methods
Detailed materials and methods pertaining to the passive sampling and analytical methods employed in this study were described in Zhao et al. (2018).Information specific to the PCB and OCP analytes are specified below and in the Supporting Information (SI).

Preparation and deployment of LDPE passive samplers
The 50-μm thick LDPE samplers were cut and cleaned for yielding passive samplers.Brominated biphenyls (2,2,2',5,2,2',4,5',and octachloronaphthalene) were used as performance reference compounds (PRCs) to infer the equilibrium concentration of compounds in the passive samplers as in previous work (Booij et al., 2002(Booij et al., , 2010;;Mcdonough et al., 2014).Both passive samplers and field blanks were picked up by trained volunteers and deployed in surface waters throughout the Narragansett Bay watershed in June-July 2014, as a part of the University of Rhode Island Watershed Watch program (http://www.uri.edu/ce/wq/ww/index.htm).The detail methodology, map of monitoring sites, and other related monitoring summary were provided in the supporting information (Figure S1 & Table S1).
PCBs and OCPs were analyzed on a Waters Quattro micro GS Micromass MS-MS and quantified using Waters QuanLynx V4.1 software as detailed elsewhere (Sacks and Lohmann, 2012).Further details are given in the Supporting Information.Samples were analyzed for 29 PCBs and 22 OCPs (Table S2).

Quality assurance/Quality control
Method blanks were prepared with each batch of samples to monitor for laboratory contamination.Samples were blank-corrected by subtracting the average of the method and field blank concentrations.Target analytes were quantified by internal calibration relative to surrogates.Average surrogate recoveries for PCBs and OCPs ranged from 57%-107% and 57%-104%, respectively (Tables S3).Method detection limits (MDLs) were calculated as 3 times the standard deviation of the average field blank concentration.For compounds that measured above the MDL in ≥ 80% of samples, concentrations < MDL were reported as half of the MDL as recommended by Antweiler & Taylor (2008), to minimize bias in the statistical analysis.Compounds that were measured above the MDL in < 20% of samples were omitted from the discussion.

Calculations/Data analysis
• Determination of sampling rate and ambient concentration The fraction of equilibrium (f) achieved for each compound was determined by fitting the equilibrium of the PRCs and their temperature-corrected log KPE values to a model curve derived as, Where Rs (L/day) is the sampling rate defined as the amount of water that encounters with the sampler per day; t is deployment time (days); MPE is the PE weight (kg); and KPEw is the LDPE-water partitioning coefficient (L/kg).The values of Rs can be estimated using nonlinear least squares methods, by considering f as a continuous function of KPEw, with Rs as an adjustable parameter using Excel Solver to obtain the best fit (Booij and Smedes, 2010).The average aqueous sampling rate of PCBs and OCPs was 16 ± 7 L/day.
Freely dissolved concentrations, CW (ng/L) of compounds were calculated from the equation, Where CPE is the PE-normalized concentrations (ng/L).For more details, see the Supporting Information and Tables S1.

• Source identification
The diagnostic ratios of TC/CC, TN/TC, DDT/(DDD + DDE), DDD/DDE, and Principle components analysis (PCA) were performed using the dissolved OCP concentrations in the surface water.PCA computation was done using the software SPSS 22.0.As a receptor model, positive matrix factorization (PMF) was able to weight each data point individually and provide better quantification of impact of its improved resolved sources (Paatero and Tapper, 1994;Du et al., 2008;Rodenburg et al., 2020).
Combing with PCA, PMF model was applied to estimate composition of possible sources of PCBs and the United States Environmental Protection Agency (USEPA) PMF 2.0 software was used.For more details, see the Supporting Information.
• Land-use regression The Narragansett watershed area in Rhode Island was 2077.6 km 2 including the estuarine waters (43.6% total watershed).The land coverage pattern around the bay was dominated by forest (58%), urban or built-up land (28%), agriculture (5.4%), water (4%), wetland, barren land and rangeland (each < 2%, see Figure S1).Land-use regressions (LUR) models are GIS-based spatial models that relate locations specific data on pollutant concentrations to location-specific source and environment data using regression (Briggs et al., 1997).
The following land-use classes were considered: residential areas, commercial and industrial areas, facilities and institutes, and recreational areas occupied by human activities (Figure S2).Other variables included population density (Pop/km 2 ), buildings, roads, sewer areas, and impervious surface areas.All geospatial data were retrieved from the RIGIS web site (http://www.edc.uri.edu/rigis).The analyses were conducted using ArcGIS version 10.2 software using coverage and grid data formats whenever possible.Statistical analyses were performed using IBM SPSS Statistics 22. Linear regression was chosen as an appropriate model based on the correlation between dissolved concentrations and major descriptor variables.

• Toxicity assessment
For PCBs, the toxicity equivalent (TEQ), established by the World Health Organization (WHO 2005), was calculated by summing the multiplication of dioxinlike PCB congener concentrations with the corresponding Toxic Equivalency Factors (TEFs) (Montuori et al., 2020).For OCPs, an ecological risk assessment for water organisms was conducted using the risk quotient (RQ) model.RQ is established based on the equation, Where, MEC is the measured environmental concentration and PNEC is the predicted no effect concentration.PNEC is derived from the lowest toxicity value (i.e., no-observed effect concentration (NOEC) value) observed for the most sensitive species.When NOEC values were not available, we used the median lethal concentration (LC50) or the median effective concentration (EC50) after correction by an assessment factor intended to extrapolate from acute to chronic toxicity and for removing the uncertainty arising from the extrapolation from intra-and inter-species variability in sensitivity (Table S13).The levels of risk were divided into four categories according to the value of RQ: insignificant risk (RQ < 0.01), low risk (0.01 < RQ < 0.1), medium risk (0.1 < RQ < 1), and high risk (RQ > 1) (Sah et al., 2020).For more details, see the Supporting Information.

Results and Discussion
Freely-dissolved PCBs in surface water
PCB concentrations derived here were more than 100-1000 times higher than those obtained by passive sampling in the Great Lakes (USA/Canada) (Liu et al., 2015;Ruge et al., 2018).The concentrations in surface water in the northern basin of this study area were about 10 times higher than those reported in the seawater of Narragansett Bay (Morgan, 2008), indicating that runoff acted as sources of PCBs from watershed to the Narragansett Bay (Table S7).However, our freely dissolved PCB values were much lower than those found in River Ravi, Pakistan (Baqar et al., 2017), and nine rivers in South China (Yang et al., 2015), though different sampling approaches were used.(Table S14).The cosine theta similarity metric (cos θ) was used for the congener pattern matching.Factor 1 explained 44% of the total PCBs and showed high loadings on PCB 11, 44, 153, and 138.PCB 11 was known to be a non-Aroclor congener and was the dominant congener in this factor, constituting 23% of ΣPCBs.It was produced as a byproduct from the manufacture of the yellow pigment widely used in ink, paint, and textile printing (Rodenburg et al. 2010;Khairy et al. 2015), and was also used as a tracer to identify factors associated with stormwater and wastewater (Litten et al., 2002;Du et al., 2008).Moreover, this factor contains the highest proportion of PCB 44 at 15 % of ΣPCBs.PCB 44 has been noted as an intermediate product of PCB dechlorination occurring in sewers (Magar et al., 2005;Rodenburg et al., 2010).Accordingly, we speculated this factor might be associated with wastewater/stormwater inputs.
Factor 2 explained 32% of the total variability in the data and was heavily loaded   S6).Dissolved OCP concentrations were dominated by DDTs (average 230 pg L −1 ), followed by chlordanes (average 230 pg L −1 ), and HCB (average 22 pg L −1 ) (Figure S5), which were detected at all sites and nearly 2-10 times higher than those obtained by passive sampling in the Great Lakes (Khairy et al., 2014;Ruge et al., 2018) (Table S8).OCP concentrations in north urban area were significantly greater than it in most south rural area.Elevated OCP concentrations were observed along the northern coast of Narragansett Bay, such as sites GB7 (chlordane, heptachlor epoxide, α-endosulfane, oxychlordane), Buckeye Brook near Knowles Brook (BB_KB) (DDTs), and Woonasquatucket River near Donigian Park (WR_DP) (HCB).Heptachlor epoxide and α-endosulfane were mainly found in the upper bay, like GB4 and GB7, where heptachlor epoxide was the main contaminant.α-Endosulfan showed an inconsistent trend and was concentrated in northern NB watershed, averaging 38.9 pg L -1 , which was comparable with the concentration reported by Venier et al. (2014).

• Potential sources of dissolved OCPs
Technical chlordane (sum of trans-chlordane (TC), cis-chlordane (CC), and transnonachlor (TN)) was used as an insecticide, herbicide, and termiticide until the early 1980s, when most of these uses were restricted in the United States and Canada in response to environmental and safety concerns.In our study, the ratios of TC/CC averaged 0.90 suggested that detected chlordane was historical and weathered residues from commercial sources except at sites GB13, S Brook, BH8, S Lake, NR01, GR3, FR_A (Figure S6) (Bidleman et al., 2002;Yu et al., 2014;Sah et al., 2020).These results were similar with the ratios in rural Great Lakes suggesting tributaries and sediments as likely sources of aged chlordane to the water column (Gouin et al., 2007;Ruge et al., 2018).TN/TC ratio ranged from 0.23 to 2.11, aquatic enrichment of TN in the sites near forest and agricultural land confirming the aged chlordane in the study area, which indicated the volatilization sources from soil containing weathered chlordane residues and deposition into the surface water (Khairy et al., 2014).
In addition to the former application to as an antifungal agent in agriculture, HCB is also an unintentionally byproduct of chlorinated compounds from industry.The average concentration of HCB (30 pg L -1 ) in north urban area was double than it in most south rural area (13 pg L -1 , see Figure 2), indicating that anthropogenic inputs other than atmospheric deposition were the dominant sources.α-Endosulfan was used on fruits, vegetables, cotton, tobacco, and trees as insecticide and banned in 2010 in the U.S. α-Endosulfan was the only analyte indicating fresh inputs by runoff and atmospheric deposition of technical-grade endosulfan.
Principal component analysis (PCA) of dissolved OCPs distinguished two groups accounting for 84% of total variance, while PCA of sampling sites also extracted two components that captured 81% of total variance (Figure 4).The sampling sites involved in PC-1 were in south rural areas (Q river, S lake, C brook, FR_A), the transition area

Correlations of human activities and freely dissolved concentrations
The effect of the land use patterns on the spatial distribution of dissolved PCBs and OCPs was investigated using a Land-use regression (LUR) (Table S10).Population density, roads, buildings, and residential areas within 2 km radius explained 36-73% of PCB variability (Table 1).Decreasing residential coverage and increasing population density might lead to greater amounts of PCBs' entering the environment, especially for penta-, hexa-, and hepta-CBs.Furthermore, there was a positive correlation existing between buildings and tetra-CBs.Hence, we can infer that legacy PCB levels in the surface water might be associated with PCBs used in building and equipment materials like sealants, paints, adhesives, etc. (Kohler et al., 2005;Liu et al., 2016;Marek et al., 2017), which could still release with building aging (Demirtepe et al., 2019).
Correlations between PCB congeners varied between north urban area and south rural area (Figure 5).In the north urban area, correlation between di-PCB (CB-11) and deca-CB(CB-209) was significant and strong (p < 0.01), indicating the potential sources of unintentionally produced PCBs coming from inadvertent production during paint, colorant, and pigment manufacturing or other industrial activities (Jahnke & Hornbuckle, 2019;Mao et al., 2021).While correlations among di-CB and penta-PCBs Urban areas within 2 km radius were major predictors of increased concentrations of chlordanes, DDTs, and heptachlor epoxide in the Narragansett watershed-coast area (Table 1 & S9).The apparent urban signal further indicated volatilization of past OCPs use from urban soil could be emitted to the Narragansett watershed (Sun et al., 2006;Venier & Hites, 2009).Agricultural areas within 5 km radius accounting for ~30-35% of the total variability in the dissolved concentrations of HCB and α-Endosulfane.The correlation coefficients were negative which suggested the application of endosulfan and HCB might decrease, and the changes of agricultural area surrounding each site could affect its spatial distribution pattern in a certain extent (Khairy et al., 2014).
No correlation was found among OCPs and PCBs, implying the declined impact of past indoor uses of OCPs for indoor insect control in north urban area.However, we found significant correlations among HCB, heptachlor epoxide and PCBs (tetra-, penta-, hexa-, and hepta-CBs) in south rural area, possibly indicating the historical residues of byproducts during the synthesis of chlorinated pesticides, including HCB and PCB mixtures (Figure 5 & Table S11-12).compared with the south rural area (Figure S7).Among all pesticides, DDE showed a RQ > 0.01 at all monitored sites and had medium to high ecological risks near Pawtuxet River (Pawtuxet R), Buckeye Brook (BB_MC, BB_KB), Greenwich Bay (GB7, GB13), and Sand hill Brook (S Brook).Furthermore, DDD, DDT and α-Endosulfane also showed medium to low risks at the majority sites in north area, while chlordane, HCBz, oxychlordane, and heptachlor epoxide showed low to negligible risks in all zones.

Conclusions and Implications
This were the major reasons for this distribution pattern.Land-use regression models could also be used to better understand sources, distributions of pollutants with the aim of reducing emissions, and guide sampling site selection (Melymuk et al., 2013).It is important to pay a continued attention to the potential emissions of legacy contaminants and their relationship with changes in land use and human activities by applying this method in future work.

Figure 1
Figure 1 Dissolved ∑29PCB and individual PCB profiles in the Narragansett onPCB 101, 11, 52, 138, and 187.This pattern was close to the profile of PCBs in volatilized Aroclor 1260 (Cos θ = 0.90) and other emission sources from sealant (Cos θ = 0.79), municipal solid waste incineration facility (Cos θ = 0.76), and cement plant (Cos θ = 0.72) with exclusion of PCB11.This factor was most prevalent near Bristol Harbor (BH1-81% and BH12-99%) and GB13 (83%), and dominant at sites of ST#3, Pawtuxet R, GB13, S Lake, and NR07, where it constitutes 23-41% of ΣPCBs.This likely represents sources of volatilized PCBs from the Harbor area and deposited throughout the watershed-estuary-bay area.Factor 3 resembled Aroclor 1260 (cos θ = 0.86).It was present in the urban areas including sites of S Brook (79% of ΣPCBs), WR_DP (58%), Pawtuxet R (35%), and BB_MC (45%), which might represent a few localized contaminated sites where Aroclor 1260 was used.Aroclor 1260 represented about 11% of US production of PCBs and was distinct from the other Aroclors.Its use was limited to transformers, hydraulic fluids, as a plasticizer in synthetic resins, and dedusting agents (Rodenburg et al. 2010).Consequently, this factor was helpful to pinpoint the locations of sites with Aroclor 1260 contamination.All three factors contained dechlorination (PCB 44) and wastewater input (PCB 11) signals.Thus, there are major inputs of urban stormwater or treated wastewater that might carry pigment and volatilization-associated PCBs from Narragansett watershed to the bay.Additionally, it seems to be shifting from the legacy Aroclor-like signature to the current use (non-Aroclor) sources.There might also be some uncertainties associated with source attribution being affected by physio-chemical processes such as partitioning or degradation.The high similarity of factor 2 and 3 with Aroclor 1260 could possibly indicate the lack of weathering.However, the lack of gaseous and particle-bound PCB datasets and measuring a limited number of PCB congeners might overestimate contributions of the wastewater/stormwater source and the non-Aroclor congeners and underestimate contributions from Aroclor congeners.

Figure 2
Figure 2 Three resolved factors by positive matrix factorization (PMF) model.
from north to south (GB13, S Brook), and Bristol Harbor, suggesting these areas had similar sources of legacy OCPs.The continued emissions of racemic chlordane from the transition area (Cluster 1) and aged chlordane from the rural areas (Cluster 2), might be primarily responsible for the current levels of chlordanes in the surface water in the Narragansett watershed-estuary-bay area.The metabolites of heptachlor, endosulfan, HCB, and chlordane presented a potential tendency to degrade in the surface water in Bristol Harbor and south rural area (Cluster 2).The north urban areas represented in PC-2 overlapped with DDTs (Cluster 3), indicating that the secondary sources of past usage and historic contamination were expected to re-enter the surface water via runoff, atmospheric transport, and deposition.These results were consistent with the above ratio diagnostics results.

Figure 3
Figure 3 Dissolved ∑22OCP and individual OCP in Narragansett watershed-

(
, hexa-CBs, nona-CB (CB-206) were also significant but relatively weaker (p < 0.05), suggesting similar ways of these contaminants entering the surface water, such as stormwater discharging and volatilization, and further supporting the results of PMF model.On the other hand, dissolved concentrations of di-CB and tetra-CBs (CB-52, -44, -66), penta-CBs, and hexa-CBs correlated significantly (p < 0.01) in south rural area, possibly related to their vapor pressor and presented similar environmental transport from north to south area.

Figure 5
Figure 5 Spearman correlation results between OCPs and PCBs.Solid lines work represented the extended research of dissolved PCBs and OCPs using passive samplers around the Narragansett Bay watershed, deployed by a long-running volunteer network.It further confirmed that LDPE passive sampling could easily and accurately detect the active portion of trace contaminants (e.g., freely dissolved PCBs and OCPs) to help determine time-integrated exposure in aquatic environments.We tracked the potential stormwater/wastewater inputs of PCBs to the surface water combing the sampling technology and PMF model.There were still several sites contaminated with Aroclor 1260 and ongoing chlordane.Surprisingly, several sites in north area still displayed medium to high toxic effects based on the presence of DDE, DDD, DDT and α-Endosulfane long after their use stopped.The sampling approach should be applied to more worldwide areas and environment reservoirs to build routine monitoring programs.Combining passive sampling with land-use regressions in Narragansett watershedbay area enabled the identification the emission sources of legacy PCBs and OCPs from various human activities.Results presented here suggested that the spatial variability of PCBs and OCPs concentrations and potential land-based sources were related to the land use pattern and population distribution in the urbanized estuary's watershed.The contaminant contents in the urban and built-up regions were higher than those from the other areas, indicating that influences of anthropogenic activities with land use change