Environ. Sci. Technol. 2005, 39, 549-556
Field Observation and Modeling of Dissolved Fraction Sediment-Water Exchange Coefficients for PCBs in the Hudson River MICHAEL J. ERICKSON Blasland, Bouck and Lee, Inc., 455 E. Eisenhower Parkway, Suite 260, Ann Arbor, Michigan 48108-3324 CARRIE L. TURNER Limno Tech, Inc., 501 Avis Drive, Ann Arbor, Michigan 48109 LOUIS J. THIBODEAUX* Gordon A. and Mary Cain Department of Chemical Engineering, Louisiana State University, South Stadium Drive, Jesse Coates Hall, Baton Rouge, Louisiana 70803
Chemical fate and transport models that simulate sediment-water exchange of contaminants typically employ empirically determined sediment-water exchange coefficients for the dissolved fraction to describe the net effect of poorly understood mechanisms. This paper presents field-derived observations of the coefficient for 12 PCB congeners and two PCB mixtures in the Thompson Island Pool, Hudson River, and also presents an evaluation of a theoretical sediment-water exchange model. An extensive PCB data set was used to compute apparent coefficients for PCBs in the pool. Average exchange coefficients for the 12 congeners ranged from 2.6 to 18.8 cm/ day, and results showed a strong seasonal dependence. Peak coefficient values occurred in mid-May to early July, preceding peak water temperatures by 1 month and lagging the spring high-flow period. The coefficients increase with increasing partition coefficients, suggesting a dependence on congener properties. The large magnitude of the coefficients and the variation among the congeners is inconsistent with the pore-water moleculardiffusion transport process. A theory-based, mechanistic twolayer model reproduces the nonlinear relationship between the sediment-water exchange coefficients and partition coefficients. This model includes transfer through the mixed sediment layer by bioturbation and diffusion transfer through a water-side boundary layer governed by flow velocity. Results suggest that this algorithm can provide increased accuracy to future system-level fate and transport models for hydrophobic chemicals. The seasonal variation in the transfer coefficient appears to be a poorly understood interaction of physical and biological processes and merits further study.
Introduction Quantifying the release of accumulated contaminants from sediment is important in evaluating contaminated sediment * Corresponding author phone: 225 578 3055; fax: 225 578 1476; e-mail:
[email protected]. 10.1021/es034520g CCC: $30.25 Published on Web 12/03/2004
2005 American Chemical Society
remediation alternatives and water-quality management alternatives. This includes the total maximum daily load (TMDL) development for water bodies impaired by historical sediment contamination. These activities often rely on the use of contaminant fate and transport models to predict impacts of contaminated sediments on water quality and long-term trends in sediment contamination levels for risk assessment purposes. Commonly used models typically lump the effects of various sediment-release processes by means of an effective exchange coefficient applied to the contaminant dissolved fraction gradient between the pore water and water column. It is then imputed and used to calculate releases to the water column over time. Review of sedimentwater exchange formulations in a number of these models reveals a common limitation: a lack of any means to estimate exchange coefficients from system variables, requiring reliance on calibration or use of default values (1). Development of accurate mechanistic formulations of sediment-water exchanges would improve model acceptance and reliability in decision making. The potential importance of resuspension to contaminant mobility in sediments has long been recognized; however, other sediment-water transfer processes can be equally important. These include molecular diffusion, biodiffusion/ bioirrigation, desorption, bioturbation, pore-water advection due to groundwater pressure or pressure fluctuations across bottom-form structures, ebullition of diagenetic gases, bioresuspension, emergence or uprooting and decomposition of aquatic plants, and various physical disturbances. Magnitudes of several of these processes have been estimated for the Fox River, Wisconsin (2). Collectively, these processes can cause significant sediment-water transfer of contaminants. According to a mass balance analysis using field data for the period 1993-1997 (3), approximately 65% of the observed sediment-water PCB flux in the Thompson Island Pool (TIP) of the Upper Hudson River, New York, occurs during nonresuspending flows. Achieving a quantitative, mechanistic understanding of these processes is important in predicting the fate and transport of sediment contaminants. This paper presents a methodology and results for quantification, from field data, of the magnitude and annual variation of effective sediment-water exchange coefficients simultaneously for 12 PCB congeners or coeluting groups and two PCB mixtures for the Thompson Island Pool (TIP), the reach with the highest PCB sediment concentrations in the Upper Hudson River. In addition, a mechanistic model of sediment-water exchange proposed by Thibodeaux et al. (4) is evaluated using these “field-observed” exchange coefficients. Development of a PCB model for the Hudson River provided the initial basis for this work (3). An objective in development of the Hudson River model was simultaneous calibration to total PCB (∑PCB), a PCB mixture, and five PCB congeners, spanning a wide range of PCB physical-chemical properties such as molecular weight, sediment-water partition coefficients, and air-water exchange coefficients. This effort found that the sedimentwater exchange coefficient must be adjusted or determined from field data for each PCB form to match water-column data (3). Also, the sediment-water exchange coefficients determined through calibration for the five congeners were nonlinear with respect to their sediment-water partition coefficients, suggesting that processes in addition to dissolved-phase diffusion were mediating transfer. The recently published model for the effective sedimentwater exchange coefficient (Kf ) by Thibodeaux et al. (4) VOL. 39, NO. 2, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
549
FIGURE 1. Site map of the Thompson Island Pool, Upper Hudson River, New York. River flow is to the south. suggests that the heavier, more strongly partitioning congeners are moved more rapidly by particle-based processes (e.g., particle mixing by bioturbation) because they exhibit higher adsorption coefficients, whereas lighter congeners diffuse faster through pore water and the benthic boundary layer. This gives rise to differences in how particle-based processes and dissolved-phase processes affect sedimentwater exchanges of different congeners.
Methodology for Empirical Rate Determination The Hudson River TIP is an impounded 6-mile reach bounded upstream by Fort Edward and downstream by Thompson Island (TI) Dam (Figure 1). Historically, the TIP received PCB loads from upstream manufacturing facilities; these are no longer significant. TIP currently contains high PCB concen550
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 2, 2005
trations, and its sediment bed is the focus of a remediation plan (5). The chemical, environmental and physical conditions have been extensively studied and documented since the late 1970s. A number of factors make TIP an ideal field site for observing sediment-water exchange coefficients: (i) extensive, high-quality data sets exist (6, 7); (ii) upstream sources are negligible (3, 7); (iii) tributary influences are small (3); and (iv) the PCB signature from the sediment source is strong (3, 7). The field-observed PCB sediment-water exchange coefficients were derived from measured increases in the load carried by the river between the upstream and downstream ends of the TIP. A three-mixed-tanks-in-series hydraulic model provides a realistic dispersion coefficient (26 m2/s average) for the pool. The physical and geometric charac-
The term σn has the form
TABLE 1. Segment Geometry and Sediment Properties for Mass Balance Model Segments
area (ha) 1 94.9 2 52.2 3 35.6
σn )
sediment propertiesa
segment geometry
dry bulk water hydraulic depth radius volume sediment density (g/cm3) porosity (m) (m) (ha‚m) foc 2.03 2.76 2.60
1.90 2.48 2.41
192 144 92.5
0.0125 0.0149 0.0124
1.29 1.18 1.25
0.512 0.553 0.527
*All samples were classified as fine or coarse sediment. Mean concentrations of each class were assigned to the fine and coarse sediment areas mapped by side scan sonar surveys (10) to compute the area weighted average surface-sediment PCB concentration for each of the three segments of TIP used in this analysis. Segment-averaged values for surface-sediment organic carbon (11), porosity and dry bulk density (13) were also computed using an area weighted average of the site-specific fine and coarse sediment properties.
teristics of the three segments are reported in Table 1. Three steady-state water-column mass balance equations were solved to quantify the effective, pool-wide coefficients for selected data pairs during nonresuspending flows. The three segments represent variations in depth, sediment type, and sediment PCB concentration along the length of the pool. Model Equations. A Lavoisier species mass balance describing PCB movement rates in the water column in each volume segment is
dCT Q ) (CT′ - CT) + Sf + Sr - Sd - Sv dt V
(1)
where CT is the TIP exiting and CT′ is the TIP incoming total chemical concentration (M/L3); Q is the flow (L3/t); V is the segment volume (L3); t is the time (t); and S denotes the source/sink terms for dissolved chemical flux from the bed (Sf), resuspension of chemical sorbed to particles (Sr), deposition of chemical sorbed to particles (Sd), and volatilization of freely dissolved chemical (Sv) in (M/L3‚t). Only data collected during nonresuspending flows were used, so Sr ) 0. Assuming that the dissolved fraction rate constant from the bed to the water is spatially constant and applying eq 1 to each segment results in a system of three equations for the water PCB concentrations in the TIP segments that have the form
dCTn Q Kfo kv s C - C f - C f ) (CTn-1 - CTn) + dt Vn hn PWn hn Tn dn hn Tn pn (2) where the subscript n denotes the number; Kfo (L/t) is the field-observed sediment-water exchange coefficient; h is the water depth (L); CPW is the pore-water concentration; kv is the air-water transfer rate (L/t); fd and fp are the dissolved and particulate fraction (computed from the two-phase equilibrium partitioning equations), respectively, of PCB in the water column; and s is the settling rate(L/t). In segment 1, CTn-1 ) CT′, the entering upstream PCB concentration at Fort Edward, and CT3 is the water leaving TIP at the dam. The three unknown values are Kfo, CT1, and CT2. Assuming steadystate conditions and combining all three equations to eliminate CT1 and CT2 yields the following expression for Kfo
(
)
Q3 V1V2V3 Kfo ) 2 C C CPW Q PW Q PW + σ1 + σσ V2V3 h 1 V3 h 2 h 3 1 2 CT3σ1σ2σ3 - CT′
( )
( )
( )
(3)
where the subscripts 1-3 denote the segment number.
( ) ( )
kvfd,w Q + Vn h
+
n
sfp,w h
(4)
n
where the subscript n is the segment number. With this final result the field-observed Kfo values are obtained primarily from measured flows and concentrations, with slight corrections for volatilization and particle settling. Model Inputs. An extensive data set is available for the TIP as a result of numerous studies performed by federal and state agencies and General Electric (GE) (6, 7). Congenerlevel PCB quantitation is available; only data from September 24, 1996, to December 22, 1999, were used to eliminate two potential sources of uncertainty related to an unrepresentative sampling station at TI Dam (8) and a pulse of PCB released from the Hudson Falls plant site in 1991 (8). Twelve individual or coeluting congeners plus the mixtures ∑PCB and PCB3+ (the sum of tri- through decahomologues ) were used in this study. The latter two mixtures were created from the congeners (see Table 2) to approximate historic measurements of PCB concentrations. The calculation of Kfo (eq 3) assumes no sediment resuspension, so the PCB data used were limited to both low flow and low total suspended solids (TSS) concentrations. Only same-day pairs of upstream-downstream samples collected below the flow of 283 m3/s (10 000 cfs) and TSS of 10 mg/L were used. Up to 51 data pairs were identified to meet these criteria. The hydraulic residence times in TIP were short (4-14 h), which supports the assumption that same-day (8-10 h) samples approximate chemical steadystate conditions. Typically, for steady-state conditions the downstream-measured concentration was larger than the upstream-measured concentration because of the steady input from the bed. A small number of data pairs where the downstream concentration was either below the detection limit or less than the upstream value yielded negative coefficients that were excluded from the analysis. The exclusion causes an unknown but small positive bias in the coefficients. Other data needed for eq 3 appear in Table 1. Average surface-sediment ∑PCB concentrations were estimated from 1998 GE measurements (9); they were grouped by sediment type and weighted on the basis of fine/course distributions determined from side scan sonar survey data (10). The average concentrations in segments1-3 were 11.5, 20.6, and 11.3 mg/kg, respectively. The 1998 values were corrected yearly using the approximate long-term rate of decline observed for PCBs in TIP surface sediments of 0.07 year-1 (3). This assumption is consistent with evidence that dechlorination is essentially complete (7). To compute the dissolved and particulate fractions in the water column, a previously developed empirical correlation between foc and flow was used (3). The daily average discharges recorded at Fort Edward by the U.S. Geological Survey were utilized; tributary inflows were neglected, as they were typically less than 5% of the total (3). Water and air temperatures were available from monitoring studies and wind-speed was obtained from the Glens Falls, NY, airport (3). The mass losses from TIP, quantified by the kv and s parameters in eq 4, indicate that volatilization and deposition account for 1030% of the PCBs released from within TIP, and therefore, Kfo is not significantly affected by these processes. The small volatilization fraction is rather uniform, reflective of the fairly small range of H for these congeners (see Table 2). The deposition fraction varied considerably among congeners, reflective of the wide range of Koc values. On average, deposition had the largest effect for BZ#101/90, 17% of the total sink term, and the lowest effect for BZ#4/10, 1%. Site-specific suspended sediment-water partition coefficients, KDw, were obtained from the U.S. Environmental VOL. 39, NO. 2, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
551
TABLE 2. PCB Chemical Properties, Sediment-Water Partition Coefficients, and Percentages in TIP Sediment Bed partition coefficients, log Koc (L/kg of organic carbon) PCB ID
no. of Cla
mol wt (mol)
Henry’s constant (atm‚m3/mol)
measured Kocsb
measured Kocw
theoretical KDETc
∑PCB (%)
PCB3+ (%)
BZ#1d BZ#4/10d BZ#5/8 BZ#15/18d BZ#28/50d BZ#31 BZ#44/104 BZ#52/73d BZ#56/60 BZ#66/93/95 BZ#70/76/61 BZ#101/90d total PCB (∑PCB) PCB3+
1 2 2 2, 3 3, 4 3 4, 5 4 4 4, 5, 5 4 5 NA NA
187 223 223 249 258 258 292 292 292 294 292 326 269 286
6.92 × 10-4 2.30 × 10-4 2.30 × 10-4 2.30 × 10-4 2.00 × 10-4 1.90 × 10-4 1.40 × 10-4 2.00 × 10-4 1.52 × 10-4 1.20 × 10-4 1.00 × 10-4 9.00 × 10-5 1.85 × 10-4 1.69 × 10-4
4.40 ( 0.52 4.92 ( 0.70 5.68 ( 0.47 5.75 ( 0.45 6.17 ( 0.54 5.92 ( 0.57 5.71 ( 0.54 5.72 ( 0.61 5.55 ( 0.50 5.72 ( 0.55 5.78 ( 0.56 5.54 ( 0.52 5.42 5.55
5.21 5.03 5.50 6.12 5.82 5.77 5.78 5.81 5.96 5.92 6.08 6.15 5.84 5.81
4.35 4.76 4.83 4.91 5.31 5.31 5.64 5.91 5.67 5.74 5.73 6.14 -
10.3 26.6 7.62 5.95 3.37 3.23 00.99 2.43 0.72 1.96 0.76 0.44 -
12.6 7.13 6.83 2.1 5.15 1.52 4.15 1.6 0.94 -
a Number of chlorine atoms. For coeluting group having congeners with different chlorine contents, the number of chlorines is given for each in the order of their occurrence in the PCB ID. b Mean values ( one standard deviation. c Values for each congener were obtained from Brunner et al. (24), and values for each congener mixture were computed on the basis of congener mass fractions determined from site data. These constants were temperature-corrected using the relationship developed for PCBs in Green Bay (25). d Denotes representative congeners.
Protection Agency (EPA) water-column data set collected in 1993 (7), and bed sediment-water partition coefficients, KDs, were obtained from the 1991 GE data (11). The GE samples were composited and frozen prior to analysis, which might have affected the accuracy of the KDs values. In both cases, the two-phase partition coefficients were determined by regression of paired particulate and filtrate PCB measurements collected from multiple stations along the river. The variability of the measured partition coefficients was reduced significantly when placed on an organic-carbon basis, yielding organic-carbon-to-water partition coefficients for the sediment, Kocs, and water column, Kocw. Obtaining reliable estimates of three-phase partition coefficients, including dissolved organic carbon as a third phase, was pursued but proved difficult (13). Theoretical organic carbon-based, twophase partition coefficients, KDET, were compiled for individual congeners from Burkhard, as cited by Mackay (14), for comparison to the site-specific KDs values. The 14 PCB forms and their relevant properties are listed in Table 2.
TABLE 3. Statistical Summary of Observed Kf Values (cm/day)
Results and Discussion
cm/day and a Kocs value of log 5.75 L/kg. The largest mean Kfo, 18.79 cm/day, belongs to BZ#28/50, which also has the largest Kocs value of log 6.17 L/kg. The low average Kfo values for BZ#5/8 and BZ#15/18 deviate from the trend of increasing Kfo with increasing Kocs that is otherwise nearly monotonic. Kfo results for PCB3+ range from 1.04 to 64.58 cm/day, with a mean of 12.79 cm/day, which agrees well with PCB3+ Kfo values determined from 1993-1996 data for the EPA model, which ranged from 1.96 to 44.69 cm/day, with a mean value of 12.15 cm/day (7). These results are also similar to other noncongener data reported by Connolly et al. (13, 15), who used 1998 PCB3+ data to develop Kfo values, with a range from 3 cm/day in the winter to a peak in the late spring to early summer of 10-14 cm/day. Our Kfo values for ∑PCB range from 0.76 to 12.09 cm/day, with an average of 5.36 cm/day. The average observed Kfo values are more than 2 orders of magnitude larger than the effective transfer coefficient due to molecular-diffusion coefficients for hydrophobic organic chemicals in porous sediment (16, 17). This indicates that molecular diffusion is not a significant mechanism of PCB release from TIP sediments and that other processes dominate. A gross time-of-year variation of Kfo for ∑PCBs and PCB3+ but not individual congeners has been reported previously
Using the above mass balance (eqs 1-4), field-observed sediment-to-water exchange coefficients, denoted kfo (cm/ day), were obtained. This section contains a statistical summary and discussion of observed coefficient patterns. Following this, calculations using the two-layer, bed-side and water-side, mass-transfer theory (eq 5) are compared to the field-observed values. Using eq 3, a total of 512 observations of Kfo were obtained for the approximately 4-year study period. Because the values for H and Kocs are less variable for individual congeners than for PCB mixtures, they were used to explore Kfo dependence on chemical properties. Six congeners, representative of the Koc range of all 12 congeners, were selected for detailed discussion of results: BZ#1, BZ#4/10, BZ#15/18, BZ#52/73, BZ#101/90, and BZ#28/50, which had the highest average Kocs. Table 3 presents a statistical summary of the Kfo values; the six representative congeners are denoted with an asterisk. The minimum Kfo value was 0.15 cm/day for BZ#1, and the maximum was 79.61 cm/day for BZ#44/104. Of the mean Kfo values for each congener, BZ#1, 2.62 cm/day, and BZ#4/10, 4.13 cm/day, are the lowest. These two congeners also have the lowest Kocs values, log 4.40 and log 4.92 L/kg, respectively. In the middle range is BZ#15/18, with a mean Kfo of 7.98 552
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 2, 2005
PCB ID BZ#1a BZ#4/10a BZ#101/90a BZ#56/60 BZ#5/8 BZ#44/104 BZ#66/93/95 BZ#52/73a BZ#15/18a BZ#70/76/61 BZ#31 BZ#28/50a ∑PCB PCB3+ a
log Kocs (L/kg of mean median min organic C) Kf Kf Kf 4.40 4.92 5.54 5.55 5.68 5.71 5.72 5.72 5.75 5.78 5.92 6.17 5.42 5.55
2.62 4.13 13.67 8.71 5.48 16.96 12.63 16.11 7.98 18.24 15.53 18.79 5.36 14.52
2.19 3.67 11.74 5.99 4.33 14.29 11.43 15.05 7.19 16.39 12.07 16.28 5.14 12.15
0.15 0.65 2.32 2.20 0.98 2.62 3.20 2.30 1.54 2.81 1.21 2.17 0.76 1.96
max standard count Kf deviation no. 13.02 9.73 37.69 45.84 29.75 79.61 31.73 46.31 27.14 59.33 57.97 56.4 12.09 44.69
2.09 1.91 8.41 8.92 4.47 12.94 6.70 9.40 5.17 11.65 10.51 11.81 2.59 8.37
48 49 35 34 47 39 36 43 46 36 45 42 49 45
Denotes representative congeners.
FIGURE 2. Time-of-year variation in observed Kfo for six representative congeners. The dark dots are the measured Kf values; see the scale to the left. Water temperature (T), noted by the thin solid line, and monthly average “base” flow (Q), noted by the thick solid line, are superimposed on the y axis; see the scales to the right. The vertical dashed lines demark seasonal intervals. for the Hudson River (3, 13, 15), the Fox River (2), and other streams containing PCBs (18). Seasonal variation in Kfo is shown in Figure 2 for only 6 of the 12 congeners studied. The coefficient annual cycle along with monthly average base flow (computed by excluding days with flows larger than 283 m3/s) and water temperature are used as an aid to evaluating the Kfo seasonality. The vertical lines mark the approximate “seasonal” changes in Kfo observations. A broad, distinct peak in Kfo occurs in the mid-May to early July period for all congeners, about 1 month after the peak flow period and 1 month prior to peak water temperature. The peak is followed by a decrease in Kfo during the low-flow, warm summer months. This is followed by an increasing trend in late September into October, especially for BZ#1 and BZ#4, which appears as a second peak, although it is generally lower than the springtime one. Whereas the spring peak behavior is clearly apparent for all congeners, the increase in the fall values is not, particularly for BZ#15/18 and BZ#28. The lowest Kfo values occur in the winter in late November and December with the exception of BZ#1 and BZ#101/90. Because of its low mass fraction in ∑PCB, BZ#101/90 Kfo values might be most subject to analytical uncertainty. Therefore, it is not entirely unexpected that these results show a higher variability and a noisy seasonal pattern compared to the other congeners. Although peak Kfo values are not fully explained by flow variations, as can be seen in Figure 2, base flows in the early spring and summer period tend to be higher compared to those in the rest of the year. Elevated base-flow velocities, although insufficient to cause appreciable resuspension, might enhance transfer of dissolved PCB across the water-side boundary layer due to high turbulence. It appears that physical-chemical processes governed directly by water temperature and base-flow variations do not explain the observed seasonal variation in Kfo or the timing of the peak values. Evaluated together, it appears likely that additional factors play a significant role. We hypothesize that the seasonal trend in Kfo is governed by a combination of physical and biological factors that vary seasonally as a function of flow velocities, temperature, bioturbation, fish activities and types, etc. Potential cause-and-effect mechanisms of some of these factors on the apparent Kfo values are discussed below. The occurrence of peak Kfo values in mid-May to early July coincides with rapid water warming, as seen in Figure 2. This warming period is expected to cause increased benthic organism activity, enhancing bioturbation; diagensis; and resulting gas ebullition (19) and, consequently, PCB release.
Oligochaetes, in particular, and other macroscopic fauna increase in numbers and activity level in early spring. Growth in benthic populations and organism activity can also be stimulated by replenishment of surface sediments with fresh organic and nutrient material deposited during runoff periods in the spring freshet and fall rainy season. In early summer, sediment-water exchange rates decline, possibly because of declining flows and the appearance of submerged aquatic vegetation (SAV), all of which reduce boundary-layer velocities and allow increased sedimentation. Additionally, macrophytes are reported to absorb PCBs (20), which would contribute to lower apparent Kfo values in the summer months. The lower values could also result from diminished bioturbation resulting from declining labile organic carbon pools in the surficial sediments. Declining redox conditions as water-column temperature and dissolved oxygen levels change might also be a contributing factor. The fall peak in the apparent Kfo values could be related to uprooting, decomposition, and disappearance of SAV and algal beds as temperature and light levels drop. SAV uprooting can destabilize surface sediments and contribute to sediment release. SAV disappearance reexposes sediments to higher shear forces, reducing boundary layer resistance and possibly introducing resuspension even within the low flow range used in this investigation. The decline in Kfo values during cold winter months is consistent with reduced biological activity at near-freezing water temperatures. Once ice cover is established, increased flow resistance and turbulence induced from the ice might increase near-bottom velocities and enhance PCB transfer. The above paragraph contains much speculation about processes that highlights the lack of understanding on how they interact to yield the observed seasonal variation in the sediment-water chemodynamics of contaminants and the need for further research. Understanding the mechanisms will afford greater confidence in using models to predict the long-term effects of these mechanisms and increased confidence in extrapolating information from extensively studied sites to other sites. The annual variation in the Kfo vs Koc relationship was assessed by evaluating its behavior within the five “seasonal” intervals in Figure 2: “early spring” (JD 105-130), “spring” (JD 131-190), “summer-fall” (JD 191-265), “fall” (JD 266320), and “winter” (JD 321-365). The Koc numerical values are not evenly distributed over the range 104.52-106.29 L/kg, and any regression analysis performed will be unduly biased by 8 of the 12 congeners having Koc values between 105.6 and VOL. 39, NO. 2, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
553
FIGURE 3. Model results versus average observed Kfo for five congener groups for the five seasonal intervals. The solid line shows the two-layer model result. Error bars on the average observed Kfo for each group are (1 standard deviation.
TABLE 4. Transport Parameters: Two-Layer Theoretical Model Regression Data season (Julian days)
season description
105-130 131-190 191-265 266-320 321-365
early spring spring summer fall winter
a
β D ba two-layer linear (cm/day) (cm2/day) model R2 model R2 18.7 32.4 51.8 10.5 35.5
0.03020 0.01910 0.00956 0.00336 0.00898
0.77 0.96 0.99 0.74 0.78
0.54 0.82 0.96 0.27 0.79
The bioturbated depth, z ) 10 cm, was used.
106.1 L/kg. Therefore, the data set was reduced to five evenly spaced groups defined by ranges of Koc values so that correlations could be developed to better reflect the Kf relationship over the full range of Koc values. Group 1 includes BZ#1 and BZ#4/10; group 2 includes BZ#101/90 and BZ#56/ 60; and group 3 includes five congeners: BZ#5/8, BZ#44/ 104, BZ#52/73, BZ#66/93/95, and BZ#15/18. Group 4 includes BZ#70/76/61 and BZ#31. Congener BZ#28/50 is the only one in group 5. Average and standard deviations of the congener Kfo values in each group are plotted in Figure 3 versus the average Kocs value for each of the five seasonal periods described previously. Graphically, the results indicate that the Kfo vs Koc relationship is nonlinear and the slope of the relationship might be seasonally variable. Kfo generally increases with Koc until leveling off at high Koc values. The apparent nonlinear function is illustrated by contrasting a linear regression of the Kfo vs Koc data; the dashed lines appear in Figure 3, and the R2 regression statistic is reported in Table 4. Generally, the theoretical curve gave a better fit to the data than the linear one. The average Kfo values for congeners 554
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 2, 2005
with middle range Koc values, groups 2-4, tend to lie above the regression line, whereas the values for groups 1 and 5 tend to lie below the line. This behavior is unexpected if Kfo is assumed to be proportional to the congener molecular diffusivity in pore water; it should decrease slightly with increasing Koc and molecular size. A similar dependence of Kfo on Kd, the sediment-water partition coefficient, for pesticides and polynuclear aromatic hydrocarbons in laboratory studies has been reported (4).
Model for Soluble Fraction Release A theoretical model based on the well-known resistancein-series concepts for chemical transport between phases is proposed (4, 16). It employs transport coefficients for which independent measurements are available (1, 21, 22). The in-bed process on the sediment side is predominantly particle biodiffusion, whereas that on the water side is based on the well-known benthic boundary layer mass-transport coefficient concept. Individually, these processes have a sound theoretical basis, with numerous observational data sets both in laboratory studies and in the field. When combined in the resistance-in-series fashion, the so-called “two-layer model” takes the form of the following algorithm for Kf vs Koc
Kf )
1 z 1 + β DbKocfocFb
(5)
Development and derivation of this equation appears elsewhere in the literature (16). This model includes a porewater diffusion term, although a good approximation is obtained by omitting it, as in eq 5, because transfer of PCBs by particle mixing is much faster. The benthic boundarylayer coefficient, β in cm/day, is controlled primarily by the
water-to-bed friction velocity, v (cm/day), and the chemical Schmidt number, Sc (dimensionless). One proposed (21) formulation of this relationship is
β)
0.114υ Sc2/3
(6)
The Db (cm2/day) and z (cm) parameters in eq 5 are the particle biodiffusion coefficient and the bioturbation depth, respectively; Fb is the bulk dry density of the bed (kg/m3). A recent literature review summarizes the reported data on these bioturbation parameters for both freshwater and marine sites (1). As shown in the following section, the observed Kfo values can be fitted to eq 5 using Koc as the independent variable to yield field-observed Db and β values for the PCBs. The five seasonal periods shown in Figure 2, each containing five sets of Kfo and Koc averages, were fitted to eq 5 to evavuate the proposed model. It can be linearized algebraically and placed in the form y ) Sx + I with slope S and intercept I. It becomes
1 1 )S +I Kfo Kocs
(7)
with the slope S ) z/(DbfocFb) and the intercept I ) 1/β. Linear regression analysis was used to extract both S and I from the linearized equation, from which the transport coefficients Db and β were subsequently determined (Table 4). The Kf vs Koc line traces for the theoretical two-layer model for each seasonal period appear in Figure 3. The model’s algebraic form reproduces the nonlinear, field-observed Kf vs Koc relationship quite well, achieving statistical R2 values ranging between 74% and 99% for the five periods. The values of the particle biodiffusion coefficient, Db, and the benthic boundary layer coefficient, β, are within the range of values reported elsewhere for these basic transport parameters (1, 22). Db ranged from 0.003 to 0.03 cm2/day, and β ranged from 10.5 to 51.9 cm/day for the five seasonal periods. The Db values in Table 4 are the first reported biodiffusion coefficients based on water-column chemistry data rather than bed-solids particle-tracer data such as radionuclides. The seasonal dependence of β and Db is considered below. Theoretically, β should vary linearly with water velocity, as shown in eq 6, and should be positively correlated to water temperature through the Sc dependence. Both the flow and temperature functions inherent in eq 6 are displayed in Figure 2; however, neither of these trends is suggested in the β values of Table 4. The season β values range from 10 to 50 cm/day, and seasonal variation appears erratic, although this might just be a result of “noise” in the analysis. Percentwise, the seasonal variations in Db are much larger than those of β. Db should have a strong benthic-biologydriven behavior, and it is known to be a strong positive function of the number of benthic organisms and of temperature (23). As the seasons change from spring to fall, Db decreases by approximately a factor of 10 with a slight rebound occurring in early winter. Increased water turbulence due to under-ice flows can explain the increases in both β and Db in the winter compared to the fall. As to the numerical magnitudes in the Hudson River, a Db value of 0.0027 cm2/ day has been reported for Foundary Cove and Lents Cove on the Lower Hudson River (25). Although based on particletracer studies, that result is in good agreement with the values in Table 4. As demonstrated here, the theoretical two-layer model provides a chemodynamic description of the PCB sedimentwater release process that is consistent with known individual mechanisms. It reproduces the variation in observed Kfo values among PCB congeners and provides the basis to study
the seasonality in the basic transport coefficients that underpin the exchange coefficient. This study provides the theoretical basis for the empirically determined sedimentwater exchange coefficients commonly used for the dissolved fraction released from the bed. It also highlights a need for further research to fully understand the other physical, chemical, and biological factors that control the magnitude and seasonal behavior of these site-specific chemical transport parameters.
Literature Cited (1) Thoms, S. R.; Matisoff, G.; McCall, P. L.; Wang, X. Models for Alteration of Sediment by Benthic Organisms; Project 92-NPS2-Report; Water Environmental Research Foundation: Alexandria, VA, 1995. (2) Thibodeaux, L. J.; Reible, D. D.; Valsaraj, K. T. Non-particle resuspension chemical transport from stream beds. In Chemicals in the Environment: Fate, Impacts, and Remediation; ACS Symposium Series 802; Lipnick, R. L., Mason, R. P., Phillips, M. L., Pittman, C. U., Jr., Eds.; American Chemical Society: Washington, DC, 2002; Chapter 7. (3) Phase 2 ReportsReview Copy: Revised Baseline Modeling Report; Hudson River PCBs Reassessment RI/FS; U.S. Environmental Protection Agency, U.S. Government Printing Office: Washington, DC, 1999. (4) Thibodeaux, L. J.; Valsaraj, K. T.; Rieble, D. D. Bioturbationdriven transport of hydrophobic organic contaminants from sediment. Environ. Eng. Sci. 2001, 18 (4), 215-223. (5) Hudson River PCBs Site, New York/Record of Decision; U.S. Environmental Protection Agency, U.S. Government Printing Office: Washington, DC, 2001. (6) Phase 2 ReportsReview Copy: Further Site Characterization and Analysis Database Report; Hudson River PCBs Reassessment RI/ FS; U.S. Environmental Protection Agency, U.S. Government Printing Office: Washington, DC, 1995. (7) Phase 2 ReportsReview Copy: Data Evaluation and Interpretation Report; Hudson River PCBs Reassessment RI/FS; U.S. Environmental Protection Agency, U.S. Government Printing Office: Washington, DC, 1997. (8) Thompson Island Pool Sediment PCB Sources; Report Prepared for the General Electric Company; Quantitative Environmental Analysis: Albany, NY, 1998. (9) 1998 Upper Hudson River Sediment Coring Program; Prepared for the General Electric Company; O’Brien and Gere Engineers, Inc.: Albany, NY, 1999. (10) Flood, R. D. Analysis of Side-Scan Sonar, Bathymetric, Subbottom, and Sediment Data from the Upper Hudson River between Bakers Falls and Lock 5; Report to TAMS Consultants, Inc. for the Hudson River PCB Reassessment RI/FS; State University of New York at Stony Brook, Marine Science Research Center: Stony Brook, NY, 1993. (11) 1991 Sediment Sampling and Analysis Program; Prepared for the General Electric Company; O’Brien and Gere Engineers, Inc.: Albany, NY, 1993. (12) Brown, M. P.; Werner, M. B.; Carusone, C. R.; Klein, M. Distribution of PCBs in the Thompson Island Pool of the Hudson River: Final Report of the Hudson River PCB Reclamation Demonstration Project Sediment Survey; NYSDEC: Albany, NY, 1988. (13) Connolly, J. P.; Zahakos, H. A.; Benaman, J.; Ziegler, C. K.; Rhea, J. R.; Russell, K. A model of PCB fate in the upper Hudson River. Environ. Sci. Technol. 2000, 34, 4076-4087. (14) Mackay, D.; Shiu, W. Y.; Ma, K. C. Illustrated Handbook of Physical-Chemical Properties and Environmental Fate of Organic Chemicals. Monoaromatic Hydrocarbons, Chlorobenzenes, and PCBs. Lewis Publishers: Ann Arbor, MI, 1992; Vol. 1. (15) PCBs in the Upper Hudson River; Report Prepared for the General Electric Company; Quantitative Environmental Analysis: Albany, NY, 1999. (16) Thibodeaux, L. J. Environmental Chemodynamics; John Wiley & Sons: New York, 1996. (17) Reible, D. D.; Valsaraj, K. T.; Thibodeaux, L. J. Chemodynamic models for transport of contaminants from sediment beds. In The Handbook of Environmental Chemistry, Hutzinger, O., Ed.; Springer-Verlag: Berlin, 1991; Vol. 3, Part F. (18) A Comprehensive Characterization of the Lower Grasse River; Alcoa Corporation, Pittsburgh, PA, 1999; pp 10-11. VOL. 39, NO. 2, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
555
(19) DiToro, D. Sediment Flux Modeling; John Wiley: New York, 2002. (20) Vanier, C.; Planas, D.; Sylvestre, M. Equilibrium partitioning theory applied to PCBs in macrophytes. Environ. Sci. Technol. 2001, 34, 4830-4833. (21) Christy, P. S.; Thibodeaux, L. J. Spill of soluble high-density immiscible chemicals on water. Environ. Prog. 1982, 2, 126129. (22) Boudreau, B. P. Diagenetic Models and Their Interpretation; Springer-Verlag: New York, 1986; pp 178-189. (23) Matisoff, G. Animal Sediment Relations. In Mathematical Models in Biochemistry; McCall, R. L., Tevasz, M. I. S., Eds.; Plenum Press: New York, 1982; Chapter 8.
556
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 2, 2005
(24) Brunner, S.; Hornung, E.; Santi, H.; Wolff, E.; Piringer, O. G. Henry’s Law Constants for PCBs: Experimental Determination and Structure Property relationships. Environ. Sci. Technol. 1990, 24 (11), 1751-1754. (25) Achman, D. R.; Brownawell, B. J.; Zhang, L. Exchange of polychlorinated bi-phenyls between sediment and water in the Hudson River Estuary. Estuaries 1996, 19(4), pp 950-965.
Received for review May 23, 2003. Revised manuscript received October 1, 2004. Accepted October 14, 2004. ES034520G
