Article pubs.acs.org/est
Modeling Water Column Partitioning of Polychlorinated Biphenyls to Natural Organic Matter and Black Carbon Richard W. Greene,† Dominic M. Di Toro,*,‡ Kevin J. Farley,§ Kathy L. Phillips,∥ and Cynthia Tomey⊥ †
Watershed Assessment Branch, Delaware Department of Natural Resources and Environmental Control, 820 Silver Lake Boulevard, Suite 220, Dover, Delaware 19904, United States ‡ Department of Civil and Environmental Engineering, University of Delaware, DuPont Hall, Newark, Delaware 19716, United States § Department of Civil and Environmental Engineering, Manhattan College Riverdale, New York, New York 10471, United States ∥ Geosyntec Consultants, 289 Great Road, Suite 105, Acton, Massachusetts 01720, United States ⊥ Axys Analytical, 2045 Mills Road, Sidney, British Columbia V8L 3 S8, Canada S Supporting Information *
ABSTRACT: High volume in situ surface water samples were collected from a tidal tributary of the Delaware Estuary using an Infiltrex sampling system equipped with a 1 μm particle filter and a XAD-2 resin column. Particulate and dissolved phase polychlorinated biphenyl (PCB) congeners were analyzed using high resolution gas chromatography/high resolution mass spectrometry to obtain detection levels in the femtograms per liter range. The data were fit to a four-phase equilibrium partitioning model including freely dissolved PCB, PCB bound to particulate organic carbon (POC), PCB bound to dissolved organic carbon (DOC), and PCB bound to black carbon (BC). Isotherms were assumed to be linear for POC and DOC and nonlinear for BC. The partition coefficient between BC and dissolved PCB was assumed to depend on the dihedral angle between the phenyl rings. Following parameter optimization, the correlation coefficient between the log of the modeled and measured apparent distribution coefficient Kp,app was 0.94, and the RMSE was 0.189 log units. Including BC in the model reduces the dissolved PCB phase concentration in the water column for all congeners, especially for the non-ortho and mono-ortho substituted congeners.
■
standard filters. Another development in PCB modeling that occurred in the early 1990s involved expressing PCBs as homologue groups and then using the results to drive a detailed food chain bioaccumulation model.8 By the mid to late 1990s, the three-phase partitioning approach had been incorporated into teaching texts9 and had been applied to significant modeling efforts on the Hudson River.10,11 During the decade beginning in 2000, Schwarzenbach et al.12 had fully integrated the three-phase partitioning approach into their text. Finally, the three-phase partitioning approach was recently used to describe partitioning of PCBs in the Delaware Estuary.13,14 While these efforts reflected what was known at the time, there is a growing awareness that certain PCBs, particularly those lacking chlorines in the four inner (ortho) positions (Figure S1, Supporting Information) or those with only a single chlorine in an ortho position, are subject to enhanced sorption to black carbon (BC).15−20 These observations have been largely restricted to bedded sediments rather than surface
INTRODUCTION Understanding the fate and effects of PCBs requires an understanding of partitioning behavior in the water column. The level of detail used to describe and model partitioning of PCBs and similar nonionic apolar to weakly polar organic chemicals in the water column has advanced over the years. Early modeling efforts in the late 1970s and early 1980s employed two-phase partitioning between dissolved and particulate forms,1−4 where the dissolved form was defined operationally to include all PCB passing through a filter of a given size (e.g., 0.45 or 1 μm) and the particulate form included PCB sorbed onto suspended matter. At the time, PCBs were expressed as total PCB, typically in terms of Aroclors or other commercial mixtures. In the late 1980s and early 1990s, research began to emerge on the three-phase partitioning approach.5−7 In this approach, partitioning is assumed to occur between the dissolved phase, chemical sorbed into POC, and chemical associated with DOC. In this framework, POC and DOC are used to represent natural organic matter. The inclusion of DOC was intended to parse the apparent dissolved phase into two subfractions: the truly (or freely) dissolved phase versus chemical that is sorbed to microparticles (colloids) and organic macromolecules (DOC) that readily pass through © 2013 American Chemical Society
Received: Revised: Accepted: Published: 6408
February April 15, April 17, April 17,
21, 2013 2013 2013 2013
dx.doi.org/10.1021/es400817c | Environ. Sci. Technol. 2013, 47, 6408−6414
Environmental Science & Technology
Article
Analysis of the extracts was performed on a high-resolution mass spectrometer (HRMS) coupled to a high-resolution gas chromatograph (HRGC) equipped with a SPB-octyl chromatography column (30 m, 0.25 mm i.d., 0.25 μm film thickness). The analytical method was carried out in accordance with EPA Method 1668A with changes and corrections through August 20, 200325 and additional modifications developed by AXYS.26 Results for the filters and the resins were reported as picograms per liter for all 209 PCB congener,; 50 of which were coelutions of two or more congeners. TSS was analyzed using EPA Method 160.2; POC was analyzed using EPA Method 440.0; and DOC was analyzed by EPA Method 415.1. Bottom sediments were analyzed for grain size using the hydrometer method of Bouyoucos.27 Black carbon analyses were based on the Lloyd Kahn TOC method with modifications for black carbon as suggested by Gustafsson et al.28 and Accardi-Dey and Gschwend.29 Quality Assurance/Quality Control. Quality control samples for the PCB analyses included separate filter and XAD resin procedural (laboratory) blanks, ongoing precision and recovery (OPR) samples, and OPR duplicates. Linearity, calibration verification, OPR, OPR duplicate, and labeled compound recovery specifications were met with few exceptions (Table S3, Supporting Information). Detections of individual congeners in the laboratory blanks (filter and resin) were all less than 0.0001 ng/L, with total PCB concentrations of 0.00061 ng/L and 0.00046 ng/L in the filter blank and resin blank, respectively. Detections in the trip blanks were similar and well correlated to the laboratory blanks. All individual PCB congeners detected in the field samples were present at concentrations at least 10 times higher than the corresponding concentrations in the laboratory blanks. Field duplicate and secondary resin columns were not part of the QA/QC samples analyzed in the study due to budget constraints. Total PCB results from the Infiltrex samples were within the observed range for the Delaware Estuary and so provide some support that the Infiltrex results are valid. Apparent Distribution Coefficient. Preliminary analysis of the PCB data was performed using apparent distribution coefficients. The apparent distribution coefficient Kp,app is the ratio of the solid phase PCB concentration to the apparent dissolved concentration.12
waters, most likely due to the lack of high quality data sets for the more dilute concentrations of both sorbate and sorbent concentrations in the water column compared to bottom sediments. Recently, Howell et al.21 used a four-phase partitioning model to predict PCB distribution in the water column of the Houston Ship Channel. For that application, partition coefficients were assigned based upon published linear free energy relationships (LFERs), a default value was used for the Freundlich exponent, and BC concentrations were likewise assigned. The current work advances the four-phase partitioning approach for the water column by determining partition coefficients for POC, DOC, BC, and the BC Freundlich exponent for a high quality data set. A new model for BC partitioning is introduced that uses the dihedral angle between the phenyl rings to parametrize structural features that influence the strength of black carbon sorption.
■
METHODS Study Area. The Christina Basin is a 1466 km2 tributary to the Delaware Estuary in the Mid-Atlantic region of the eastern United States. Land use and land cover in the Christina Basin is mixed with the lower tidal reaches consisting largely of urban use associated with the city of Wilmington. Historically, there has been significant industrial activity along the tidal waterfront of the lower Christina Basin.22 Activities that likely used or encountered PCBs include shipbuilding, railcar maintenance, pigment manufacturing, rubber casting, and metal salvage. Federal and state waste cleanup programs have remediated many sites known to be contaminated with PCBs along the tidal waterfront, although many PCB-impacted sites remain.23 Surface water, sediment, fish, and other aquatic resources in the tidal Christina Basin remain contaminated with PCBs.24 Field Sampling. In October of 2007, in situ high volume water samples (400 L) were collected from three contrasting locations in the tidal Christina Basin using an Infiltrex 300 trace organic sampling system (AXYS Environmental Systems Ltd.). A map (Figure S2) and general attributes of the three sampling locations (Table S1) are provided in the Supporting Information. Infiltrex samples were collected from the middle of the water column at a nominal pumping rate of 1.25 L/min. Water drawn into the equipment was first passed through a glass fiber filter with a nominal pore size of 1 μm, and then the filtrate was passed through an Amberlite XAD-2 polymeric adsorbent column. Grab samples of the water column were also collected at the beginning, middle, and end of the Infiltrex sampling using a Beta bottle lowered to mid-depth. The individual samples were analyzed for nonfilterable residue (i.e., total suspended solids (TSS)), POC, and DOC. Composite samples of surficial bottom sediments were collected at each site at the same time as the Infiltrex samples using a stainless steel Ted Young modified Van Veen grab sampler. The sediment samples were analyzed for grain size, total organic carbon (TOC), and BC. A calibrated hand-held multimeter was used to measure water temperature and salinity. Analytical Procedures. Infiltrex filters and resin columns were analyzed by AXYS Analytical Ltd. in Sidney, British Columbia, Canada. Samples were Dean−Stark Soxhlet extracted. The extracts were spiked with 13C-labeled PCB surrogates and 13C-labeled PCB cleanup standards and then cleaned up on a series of chromatographic columns. The extracts were reduced in volume and spiked with isotopically labeled recovery (internal) standards prior to instrumental analysis. Final extract volume was 20 μL; 1 μL was injected.
K p,app, =
C Tp/m TSS C Td
(1)
For the field data, CTp is the particulate phase PCB concentration reported for the Infiltrex filter. This concentration is assumed to include both PCB sorbed into POC and PCB sorbed to BC. mTSS is the average total suspended solids concentration for the site calculated from the three individual grab samples collected at the beginning, middle, and end of the Infiltrex sampling. Finally, CTd is the apparent dissolved PCB concentration reported for the Infiltrex resin, which is assumed to include freely dissolved PCB and PCB associated with DOC. Measured apparent distribution coefficients were determined on a PCB congener-specific basis in situations where both apparent dissolved and particulate phase concentrations were reported above the detection limit. Apparent distribution coefficients were compared to octanol−water partition coefficients (log KOW) on a congener-specific basis. Log KOW values were taken from Hawker 6409
dx.doi.org/10.1021/es400817c | Environ. Sci. Technol. 2013, 47, 6408−6414
Environmental Science & Technology
Article
flat (a.k.a. planar), these particular congeners appear to bond more strongly with sorbents that have planar moieties such as black carbon through π−π bonding above and below the phenyl rings.32 This increased bonding strength is reflected through the second term in eq 5, which causes KBC to increase as the dihedral angle decreases. Ultimately, the motivation for our KBC model was to capture the behavior of the observed field data. Method of Determining Model Parameters. The parameters α, β, a0, a1, and n were estimated by minimizing the sum of squares of the differences between the measured and modeled (eqs 2d, 3−5) apparent distribution coefficient for all of the congeners. This was done using Microsoft Excel’s SOLVER with settings GRG (generalized reduced gradient) algorithm, a maximum of 200 iterations, a precision of 1 × 10−7, automatic scaling, and central differencing of numerical derivatives. Further, α was constrained to values between 0 and 0.1; n was constrained to values between 0 and 1; and β, a0, and a1 were set as positive. An initial guess for α was 0.06 based on the value given by Burkhard specifically for PCBs.33 An initial guess for β was 0.35 based on Seth et al.,34 and an initial guess for n was 0.7 based on Schwarzenbach et al.13 Finally, initial estimates for a0 and a1 were made by assuming an order of magnitude value for log KBC of 5.5 based on Bucheli and Gustafsson16 and Koelmans et al.,20 a dihedral angle of 56.6° (mean for mono-orthos), and that the two terms on the righthand side of eq 5 are equal. This results in an initial estimate of 5.2 for log a0 and an initial estimate of 3.7 for log a1. Once a local optimum solution was identified using SOLVER, the individual parameter values were systematically perturbed to verify that the local optimum solution was in fact the global optimum solution.
and Connell30 with no adjustment for temperature or salinity. Averages were used for coeluters. Four-Phase Equilibrium Partitioning Model. Final analysis of the PCB data was performed considering fourphase equilibrium partitioning. Notation and associated definitions used in the partitioning model appear at the end of this paper. Substituting the individual concentrations that make up the particulate and dissolved concentrations into eq 1 yields K p,app =
1 C POC + C BC m TSS Cfd + C DOC
(2a)
Substituting the partition coefficient models for POC, DOC, and BC yields =
n 1 mPOCKPOCCfd + mBCKBCCfd m TSS Cfd + mDOCKDOCCfd
(2b)
and in terms of the dry weight fractions of POC (f POC) and BC (f BC) yields =
fOC KPOCCfd + fBC KBCCfdn Cfd + mDOCKDOCCfd
(2c)
or =
fOC KPOC + fBC KBCCfdn − 1 1 + mDOCKDOC
(2d)
The partition coefficients in eq 2d were assumed to be of the form: KPOC = βK OW
(3)
KDOC = αK OW
■
(4) o
KBC = a0 + a1(90 − dihedral angle)
RESULTS Total PCB Concentrations. Table S2(Supporting Information) presents a summary of the total PCB, total apparent dissolved PCB, and total particulate PCB concentrations at the three sampling locations. Table S3 (Supporting Information) presents the individual PCB congener results for the Infiltrex filters and XAD-2 resins together with associated QA/QC results. For purposes of calculating totals, PCB congeners reported as not detected or congeners that had a peak but that did not meet quantification criteria were assigned a value of zero. Total PCB concentrations (1.67−5.38 ng/L) were within the range reported for the tidal Delaware River (1.2−6.5 ng/ L)35 and the Houston Ship Channel (0.49−12.49 ng/L),21 lower than the range reported for New York Harbor (6.7−9.4 ng/L),12 and higher than the range reported for Lake Michigan (0.34−1.74 ng/L)36 and San Diego Bay (
