Release rates of chlorinated hydrocarbons from contaminated sediments

transcript

Netherlands Journal of Sea Research 29 (4): 297-310 (1992)

R E L E A S E RATES OF C H L O R I N A T E D H Y D R O C A R B O N S F R O M C O N T A M I N A T E D S E D I M E N T S

KEES BOOIJ 1, ERIC P. ACHTERBERG 2 and BJORN SUNDBY 3

1Netherlands Instffute for Sea Research, RO. Box 59, 1790 AB Den Burg, Texel, The Netherlands 2University of Liverpool, Oceanography Laboratories, P.O. Box 147, Liverpool L69 3BX, UK

31NRS-Oceanologie, Centre Oceanographique de Rimouski, 310 allde des Ursufines, Rimouski (Qu6bec), Canada G5L 3A1

ABSTRACT

The release rates of chlorinated hydrocarbons from contaminated intertidal sediments were measured with an apparatus that consists of a microcosm with calibrated flow regime, a continuous counterflow liquid-liquid extractor, and an aerator column. Sediment-water partition coefficients were also determined. A set of design criteria was developed and used to asses the validity of the flux measurements. The release rates of halogenated hydrocarbons (octanol-water partition coefficient 10 4 to 10 7) conform to a diffusion model with an effective diffusion coefficient of the order of 1 to 4.10 -8 m2.s -1, 40 times higher than the estimated molecular diffusion coefficient. It is difficult to understand why the effective diffusion coefficient is so high. Using a numerical transport model that assesses the effect of pore water flushing, we have ruled out the possibility that the result is due to pore water irrigation by the secondary flow that exists in the microcosm, but more work will be needed to understand the transport mechanism. The magni- tude of the fluxes suggests that contaminated sediments, such as those used in this experiments, may act as weak but persistent sources of chlorinated hydrocarbons to the water column.

1. INTRODUCTION

Deposits of contaminated sediments, which can now be found in many of the world's coastal regions, have become the subject of environmental concern because the contaminants they contain may be released to the water column and enter the tissue of living organisms. Decisions to remove such deposits by dredging, or to cap them with a cover of uncon- taminated sediment, or simply to leave them un- touched, can best be made if one understands the properties of the contaminants that are adsorbed on

the sediments. In particular, one needs to understand whether or not the contaminants are released from such deposits and, if they are, what the release rates are and what controls them.

It is now well established that sediments can be a source of hydrophobic chlorinated hydrocarbons (HALTER & JOHNSON, 1977; ELDER et al., 1979; LARS- SON 1983, 1984, 1985; SODERGREN & LARSSON, 1982; FISHER et al., 1983; KARICKHOFF & MORRIS, 1985; WOOD et al., 1987). The mechanisms involved are less well established. For example, both FISHER et al. (1983) and KARIOKHOFF & MORRIS (1985) found that the release of chlorinated hydrocarbons from sediments could be described by a diffusion model. KARICKHOFF & MORRIS (1985) estimated the diffusion coefficient for penta- and hexa-chlorobenzene to be slightly higher than their molecular diffusion coefficients whereas FISHER et al. (1983) calculated values for the diffusion coefficient that were orders of magni- tude lower than the molecular diffusion coefficients. The latter authors also found that the dependence of the flux on the hydrophobicity of the compounds was the opposite of the model predictions.

In this paper we present the results of a laboratory study to measure the release rates of chlorinated hydrocarbons from coastal marine sediments. The sediments were obtained from the Dutch Wadden Sea and were used in their original, already contaminated form. We thus avoided the potential problems that may arise when contaminants are added to the sediment in the laboratory.

Acknowledgements.--We thank M.Th.J. Hillebrand for valuable assistance and C.J.M. Wolf and Dr. A.P. Woodbridge for their continuing interest and support of the project. This research was supported by a grant from Shell Internationale Research Maatschap- pij. The experiments used a patented microcosm whose inventor is Dr. G. Gust (University of South Florida). His permission for its use in this project is acknowledged.

298 K. BOOIJ, E.P. ACHTERBERG & B. SUNDBY

2. EXPERIMENTAL DESIGN CRITERIA

A laboratory set-up to measure fluxes across the sediment-water interface consists of a tank containing sediment and overlying water, a stirrer, and a device to extract the contaminants continuously from the water. The extractor may be an adsorbent, a gas stripping column, or a liquid-liquid extractor. The flux is sampled by extracting the overlying water continuously during a certain time interval.

In the following discussion on how to assess the validity of flux measurements, we will make use of the mass transfer coefficient (k), which links the flux (j) to the difference between the concentrations in the pore water at infinite depth (CP) and the water column (C w) (BIRD et al., 1960; LEVENSPIEL, 1962; BOUDREAU & GUlNASSO, 1982; VAN RAAPHORST et al., 1988):

j=k (CP-C w) (1)

The mass transfer coefficient, which has the dimen sion of velocity, is also known as the piston velocity because the transfer process can be treated as an apparent fluid flow (BOUDREAU & GUlNASSO, 1982; MACKAY & PATERSON, 1982). Per unit area and unit time, this apparent flow pumps an amount k-C w into the sediment and an amount k.CP out of the sediment. The product of the mass transfer coefficient and the surface area of the sediment (k-A) is mathe- matically equivalent to a volume rate of flow.

Since the mass-transfer coefficient may be time dependent, it is best determined by keeping the con centration in the water column constant, and prefera- bly zero in order to maintain a maximum driving force for the flux. This condition can be evaluated from the mass balance for the water column (see Fig. 1). The mass balance equation for the water column is:

dC w_ k A (CP-CW)-Qw ~ C w (2) dt V V

where A is the surface area of the sediment, V is the volume of the overlying water, Qw is the volume rate of flow of the water, and ~ is the extraction efficiency for single through flow. Assuming that during one flux measurement k is approximately constant over the sampling interval, the steady state solution of equation (2)is

C w k A

Cp - C w Qw ~ (3)

A practical criterion for C TM being close to zero is that Cw,~cP. This leads to the first design criterion:

kA - - < 1 ( 4 )

A 100o/0 extraction efficiency (~=1) is therefore neither necessary nor sufficient. Instead, the rate of removal (Qw E) by the extracting device needs to be much larger than the rate of transport from the sediment to the water column (k A). As a check on the validity of the assumption of steady-state, the characteristic time scales for the attainment of steady-state were determined by numerical integration of equation (2). These time scales are smaller than the flushing time of one volume of overlying water. During the experiments this flushing time was 20 min, which is small compared to even the smallest sampling interval (2.4 hours). Thus, the steady state assumption is reasonable.

A second criterion that should be satisfied con- cerns the acceptability of the first flux measurement after starting the experiment. The amount desorbed from the sediment during the first time interval should be much larger than the amount originally present in the water column at t=0. Assuming that the contaminant dissolved in the water column initial- ly is in equilibrium with the pore water (C w =CP) this condition can be expressed by the equation:

j A At = kCPA At ~> V C w = VCP (5)

or, after rearranging,

V ~1 (6)

k A n t

These criteria will be used to accept or reject the mass transfer coefficients calculated from the flux measurements.

--~x-t ract-oT--i

Qw CW I_ _ I Qw

) seaiment

, , - , ,

C~(l-~)

Fig 1. Contributions to the mass balance of the water column, k.A is the apparent volume rate of flow out of the

sediment, Qw is water flow rate through the extractor.

RELEASE RATES OF CHLORINATED HYDROCARBONS 299

TABLE 1 Characteristics of the sediments used in the release experiments. PCBs are coded by their IUPAC numbers.

Sediment 1 2a 2b 3 4 type sand sand sand mud sandy mud

fom 0.016 0.010 0.009 0.113 0.039 ~,-HCH (pg g- l ) 4 6 7 13 59 hexachlorobenzene (pg g - 1) 33 12 17 184 NA CB 28 (pg g- l ) 94 68 81 430 NA CB 52 (pg g- l ) 98 57 53 422 760 CB 101 (pg g- l ) 134 92 78 557 1125 CB 138 (pg g- l ) 182 135 119 716 1329 CB 153 (pg g- 1) 244 185 165 984 2594 CB 180 (pg g- l ) 136 83 71 583 943 K d experiment no no no yes yes flux experiment yes yes yes yes no

3. EXPERIMENTAL PROCEDURES

3.1. SAMPLING AND ANALYSIS

Sediment samples were collected between January and June 1987 from tidal flats in the western Wadden Sea (The Netherlands). The top 3 cm of the sediment was collected and mixed after removal of large shells.

The organic matter content of the sediment was determined as ash-free dry weight, by drying at 105°C to constant weight, followed by combustion at 600°C for 3 hours. This method is not without problems (see for instance the review by DANKERS & LAANE, 1983). The errors involved include the loss of water adsorbed to clay minerals, the loss of crystal water from clays, and the partial dissociation of car- bonates. The larger errors seem to be related to high clay contents, low initial drying temperatures, and high ignition temperatures. Despite these uncertain- ties, many of the authors cited in the above review paper have attained ratios of ash-free dry weight to organic carbon that are close to 2.5. We feel therefore that the errors in our organic matter determination are minimal because we used a relatively high drying temperature (105°C) and a moderate ignition temper ature (600°C).

The characteristics of the sediments used in the experiments on partition coefficients and in the flux measurements are summarized in Table 1.

A schematic presentation of the experimental set- up is shown in Fig. 2. A stainless steel microcosm of 30 cm diameter and 35 cm height was filled with sediment to a height of 25 cm. The sediment was allowed to settle for at least four hours. The water column was stirred by a 20 cm diameter rotating disk with a 4 cm high skirt. Water was removed from the centre of the disk, fed into a liquid-liquid extractor and returned on top of the disk. The combined action of disk revolu- tion and water flow produces a homogeneous distri- bution of the skin friction (bottom shear stress minus

NJo2--P---

" - ¢ " 1 = ,

• i • 1

• t i • °1 3 O

• O O t ,

• t e l O o O i e l

o = = 1 o

, •= i " O O

or•' ' 0 0 0

• el 0

• , 0 ~ 0

i;:i. ° 0 o

oqbo E

0 0 °

0 0 o 0

t t I o • o-•1 0

: ' :1 o o o • • , O O

"'-I oOO • 0

t i e o • l 0 0 0

t t , O O

= e l 0 0 ~ 1 . . . . . . . .

Fig. 2. Schematic representation of the experimental set-up used for the flux measurements. Water is removed from the centre of the microcosm (M) and pumped by a gear pump (P1) via a heat exchanger (H) to the extractor (E). In the extractor, organic contaminants are removed from the water phase with hexane. Water is returned to the microcosm via the stripping column (S), where hexane is removed. Hexane is partly recycled via gear pump (P2) and partly distilled in the distillation set-up (D). The water column in the microcosm is stirred by the rotating disk (R) and the circulating

water.

300 K. BOOIJ, E.R A C H T E R B E R G & B. S U N D B Y

i,, t li

i ' %, .ji",~,,,,~ j!,d+ i ' i,~. P.\ '!,,. .......

20 30 4O 50 t ( ra in )

Fig. 3. Examples of chromatograms obtained in the experiments, a. pentane extract of 75 g of sediment 1, b. blank for microcosm and extractor without sediment (10 h), c. continuous water extract from a flux experiment (sediment 1, 27 h).

form drag) at the sediment water interface (GUST, 1988). A detailed description of the microcosm and the hydrodynamical conditions prevailing in the water column can be found in Boou et al. (1991). The experiments were carried out at a friction velocity of 0.75 cm.s -1. In the natural environment, the friction velocity ranges from several mm.s-1 in the deep sea (WIMBUSH & MUNK, 1970) tO several cm.s -1 in estu- aries (HEATHERSHAW, 1979; ANWAR, 1983; SCHAUER, 1987). The temperature of the water was maintained at 15+2°C.

Organic contaminants were continuously extracted from the water column with hexane, using a counterflow liquid-liquid extractor. Dissolved hexane was removed from the effluent in a gas stripping column placed in line with the extractor. A mixture of oxygen and purified nitrogen was used as the stripping gas. This procedure, which also served to maintain the water column in the microcosm aerated, kept the concentration of hexane in the return flow less than 0.6 g.m -3, as measured by GC/MS. The concentra tion of chlorinated hydrocarbons in the sediments was determined by drying the sediment at 60°C to constant weight (< 36 h) and extracting with pentane in a Soxhlet extractor overnight (40 to 200 g dry weight). The extraction efficiency was checked by extracting the sediments a second time with dich- Ioromethane. The amounts of polychlorinated biphenyls (PCBs) present in the dichloromethane extract amounted to less than 5% of that in the pentane extract. However, recent experiments have shown that extractions with acetonelhexane (30170) general- ly result in 30% higher yields.

Using the procedure of DUINKER & HILLEBRAND (1983), the extracts were cleaned up, separated into one fraction containing PCBs, p,p'-DDE and chlorinated benzenes and another containing pesticides, and analysed with capillary gas chromatography with electron capture detection. Typical chromatograms of microcosm blanks, sediment extracts and continuous water extracts are shown in Fig. 3. The pesticide fraction could not be analysed because of interfering peaks, except for c~- and .y-hexachlorocyclohexane. Fluxes of pentachlorobenzene were not considered because of high blanks, originating from the stripping gas.

3.2. DETERMINATION OF SEDIMENT-WATER PARTITION COEFFICIENTS

Sediment-water partition coefficients were determined on two sediment samples from the study area (organic matter fraction: fore=0.039 and 0.113). A spike solution, prepared in hexane, contained 26 g.m-3 Clophen A30, 20 g.m-3 Clophen A60 and 0.3 g.m -3 of decachlorobiphenyl, pentachlorobenzene, hexachlorobenzene and ~- and .y-hexachlorocyclo-

hexane. Partition coefficients were determined at two contaminant Ioadings. Volumes of 0.4 and 1.0 cm 3 of the spike solution were plated onto the walls of 100 cm 3 glass bottles, after which 4 g of dried sediment and 70 cm 3 of seawater were added. After an incu- bation period of 7 days at 14°C, the bottles were cen- trifuged for 20 rain at 3000 rpm. The water phase was siphoned off to a level of 5 mm above the surface of the solids and extracted by shaking with 20 cm 3 hexane in a separatory funnel followed by a 24-h phase separation. This extraction was repeated twice, and the subsequent extracts were combined. The sediment phase was extracted as described above. Root mean square deviations for K d values obtained at high and low contaminant Ioadings were 0.07 log units (= 17%).

3.3. REJECTION PROCEDURE FOR FLUX MEASUREMENTS

A database of the chromatographically resolved peaks and peaks of co-eluting PCBs with equal numbers of chlorine atoms was prepared. Components with retention time deviations of more than 0.03 min from the external standard were deleted. We estimate the lower limit for reliable quantification to be 0.4 pg injected. All amounts less than this value were also deleted from the database. As a second step, the apparent mass transfer coefficients (kapp) were calculated from the amounts (n) in the water extracts and the concentrations C s in the sediment phase, As- suming Cw,~cP and CP=CS/Kd, it follows from equation (1):

_Jaoo Kd_ n K d kapp- C s A ,~t C s (7)

All apparent mass transfer coefficients that did not conform to the acceptance criteria (equations (4) and (6)) were deleted. The flux measurement was accept- ed when the left-hand sides of these equations were smaller than 0.3. No fluxes of PCBs with more than 7 chlorine atoms passed the rejection procedure.

Octanol-water partition coefficients were obtained from SHIU & MACKAY (1986). For PCB congeners not listed, the logarithmic average of listed congeners with the same number of chlorine atoms was adopted. For co-eluting PCBs with the same number of chlorine atoms, the average log (Kow) for the individual congeners was used.

4. RESULTS AND DISCUSSION

4.1. PARTITION COEFFICIENTS

In order to check if the octanol-water partition coefficient (Kow) and the organic matter fraction are valid

I0 S -

(cm 3 g-l)

10 4 --

++ + +o ++

~lIH~. I l l=nm I I I,nm I I I~lH. I ,lIra, I l lIra11 10 3 10 4 l0 5 10 6 i0 7 i0 8 l0 g

Fig. 4. Plot of the sediment-water partition coefficient normalized to the organic matter fraction versus the octanol- water partition coefficient, for two sediments with

fom =0.039 (circles) and 0.113 (crosses).

descriptors of equilibrium partitioning between sediment and water, we determined sediment water partition coefficients (Kd) on two of our sediments. K d values are frequently normalized to the organic matter fraction (fom) or the organic carbon content (foc) of the sediment (KARICKHOFF et al., 1979; SCHWARZEN- BACH & WESTALL, 1981; MEANS et al., 1980; CHBOU et al., 1983):

Kom- Kd (8) fore

The results of the measurements (Fig. 4) show that Iog(Korn) increases linearly with Iog(Kow ) for Iog(Kow)<6.1. The slope of the linear portion of the graph is not significantly different from 1 (p=0.05). The difference between the Iog(Kom ) values for the two sediments (with fom =0.113 and fom =0.039) was less than 0.08 log units (= 20%). The line drawn in Fig. 4 corresponds to the equation

log I~ -~ t =log (Kow) (9)

a=0.062 cm3.g -1, N=41, ~=0.17

When a ratio of organic matter to organic carbon of 2.5 is assumed, partition coefficients calculated from eq. (9) are 0.6 log units lower than the values reported by KARICKHOFF et al. (1979) and by MEANS et al. (1980) and are approximately equal to the values

reported by SCHWARZENBACH & WESTALL (1981) and by CHtOU et al. (1983).

For Kow values greater than 106, Kom rises tess steeply and levels off at about 2.105 cm.g -1. This could be due to the presence of colloidal particles (GSCHWEND & WU, 1985; BROWNAWELL & EARRING- TON, 1986), slow attainment of sorption equilibrium for the more hydrophobic compounds (KARICKHOFF, 1984), or sorption saturation. Considering the good correspondence between the K d values obtained at the high and the low contaminant loading, we do not think that the reason is sorption saturation. If the leveling-off were due to the presence of colloidal particles in the water phase, the maximum value for Kom would be close to the reciprocal dissolved organic matter concentration (BROWNAWELL & FARRINGTON, 1986; BROWNAWELL, 1986). The Kom maximum of 2.105 cm-g -1 could then be explained by a dissolved organic matter concentration of 5 g.m -3, which is a reasonable value. In the further analysis of the data we assume eq. (9) to be valid for all compounds identified in the flux experiments (104< Kow< 107).

4.2. FLUX MODEL

FISHER et al. (1983) and KARICKHOFF & MORRIS (1985) have modelled contaminant fluxes from sediments using a diffusion model according to which the flux of a hydrophobic compound is inversely related to time and hydrophobicity. The assumptions under- lying this model are instantaneous sorption equilibrium between the pore water and the sediment particles, homogeneity of the diffusion coefficient (D), the porosity (e) and the sediment water partition coefficient (Kd), and no advection. In addition it is assumed that at t ~<0 the concentration of contaminants in the sediment is homogeneous and that at t=0 the concentration at the sediment water interface changes instantaneously to a constant value. With these assumptions an analytical solution for the flux exists (BERNER, 1980; FISHER et al., 1983; KARICK- HOFF & MORRIS, 1985):

j= e~D(1 +K) (CP - C w) (10)

K= (1- ~)0 Kd (11)

Assuming Cw~C p, and K~>I, and substituting CS=KdCP in equation (10), gives:

j _e l /D ( I + K ) ~ v ' ~ (1-~)~o

C s ~, ~/-~t K d l~ ~/-~t ~ d (12)

i0 -i _

(g ~2 S-l)

10 -2 -

10 .3 _

sediment #2b

+ o -~ +

+ :~ Z %

# -~ + _--

sediment #3

' ' ' ' " ' 1 I ' ' ' ' ' " 1 ' ' ' ' ' " 1 ' ' ' ' ' ' " 1 $ l0 s 10 6 i0 s 10 6

t (s) t (s)

Fig. 5. Examples of normalized fluxes of hexachlorocyclohexanes (circles) and heptachlorobiphenyls (crosses) versus time for two sediments with forn--0"0085 (left graph) and 0.113 (right graph)

Fig. 5, which shows the fluxes of hexachlorocyclohexanes and heptachlorobiphenyls, normalized to their concentrations in the sediment, suggests that our results agree qualitatively with this model: the fluxes decrease with time, and higher normalized fluxes correspond to compounds of lower hydrophobicity and to lower organic matter content in the sediment. (The degree of scatter for compounds of similar hydrophobicity is about the same for all the flux measurements, viz. 0.1 to 0.5 log units.) Transforming equation (12) to logarithms gives

,oo C l-o., ,oo IDo( - o,o.I _o., ,oo(,,-O.,,oO(,d, (13)

Thus the diffusion model requires that Iog(j/C s) should be proportional to -0.5 log(t) and to -0.5 Iog(Kd). Furthermore, since K d is proportional to both Kow and fore, I°g(Kow) and Iog(fom ) have theoretical regression coefficients of -0.5.

4.3. APPLICATION OF THE MODEL

In order to compare the results with the model we carried out a multiple regression analysis of the normalized fluxes, using time, the octanol-water partition coefficient and the organic matter fraction as independent variables. The results, including the 95% confidence intervals of the regression coefficients, can be summarized by

log I-~---s~ o~-(0.44_+0.05)1og (Kow)-(0.41+0.09) log(t)

-(0.78+0.11)1og (forn) (14)

R=0.88, N=219, (~N_4=0.31

The regression coefficients are close to -0.5, which is the theoretical value required by the diffusion model.

We also carried out a regression analysis of the data for the individual sediments, using time and the octanol-water partition coefficient as independent variables. The results of this analysis (Table 2) show

TABLE 2 Regression coefficients (:t:95% confidence intervals) for the multiple regression analysis of Iog0/C s) on log(t) and Iog(Kow ).

Sediment N regression coefficient for Iog(Kow) regression coefficientfor log(t) R (TN-3

1 44 -- 0.71 + 0.15 -- 0.43 + 0.15 0.89 0.36 2a 90 - 0.38:1:0.07 - 0.40 + 0.14 0.80 0.34 2b 24 -0.25:1:0.12 -0.54+0.12 0.82 0.24 3 61 - 0.53 + 0.09 - 0.22 :!: 0.20 0.86 0.32

that the fluxes from the individual sediments are less well described by the diffusion model, The regression coefficients range from -0.25 to -0.71 for Iog(Kow ), and from -0.22 to -0.54 for log(t). Four out of eight regression coefficients differ significantly from -0.5 (p=0.05).

Several phenomena that may affect the fluxes are not included in the diffusion model. These are slow desorption kinetics, boundary layer resistance, transport of contaminants associated with colloidal material, and reworking of sediment by organisms. We will now discuss the possible importance of each of these phenomena in relation to the diffusion model.

If slow desorption kinetics were important, the pore water would be depleted more rapidly than the diffusion model predicts because the sediment phase cannot deliver contaminants as fast to the pore water as they are lost by diffusion. Thus, a more negative value of the regression coefficient for log(t) would be expected. The coefficient for Iog(Kow ) would also be more negative than -0.5 because larger molecules are released more slowly than smaller ones (Wu & GSCHWEND, 1986). Inspection of Table 2 shows that such is not the case.

If the benthic boundary layer completely controlled the flux, the release rate would not be limited by depletion of the pore water, but by the rate with which the contaminants were removed from the sediment- water interface. The flux would then be independent of time and linearly proportional to the concentration in the pore water, hence inversely proportional to K d. This would result in regression coefficients of 0 and -1 for log(t) and Iog(Kow ), respectively. Other than possibly for sediment 3 (regression coefficient for log(t)= -0.22+0.20) it does not appear that the boundary layer resistance was important.

Because organic colloids show a high affinity for hydrophobic contaminants (BOEHM & QUlNN, 1973; WHITEHOUSE, 1985; BROWNAWELL & FARRINGTON, 1986; WIJAYARATNE & MEANS, 1984), association with colloids might increase the transport of contaminants. This could be important with dispersive processes other than molecular diffusion, such as mixing of pore water by currents and biota. If such particle associated transport of contaminants were the only transport mechanism, the normalized flux (j/C s) would be independent of the partition coefficient since colloids and sediment show similar ad- sorption characteristics. However, the diffusion of the colloidal particles themselves might be described by the diffusion model. Co-diffusion can be identified by regression coefficients for Iog(Kow ) that are less negative than -0.5 and by regression coefficients for log(t) that are equal to -0.5. Inspection of Table 2 shows that this is the case for sediments 2a and 2b.

The same reasoning can be applied to the contribution of sediment reworking, if the mixing can be

described by a particle mixing coefficient. Particle mixing and the diffusion of colloidal particles in the pore water are basically the same phenomenon, but for different particle sizes. Hence, the deviations from the model by sediments 2a and 2b could also be due to particle mixing.

4.4. ESTIMATES OF DIFFUSION COEFFICIENTS

Since the diffusion model represents the data reasonably well, we can use the model to estimate diffusion coefficients. By rearranging equation (12) and substituting the correlation between K d, fom and Kow (equation 9), we get

D=Tr t f°m a K°w,! C s ' 2 (15) e (1- e) e

With porosities of 0.5 for sediments 1, 2a, 2b, and 0.8 for sediment 3, and a sediment grain density of 2.6 g.cm-3 we obtain diffusion coefficients in the range of 1 to 4.10 -8 m2.s -1 (Table 3). These estimates are 20 to 80 times larger than molecular diffusion coefficients for PCBs, which are approximately 4 to 6.10 -1° m2.s -1, m 2 calculated from the Wilke-Chang equation (DANCKWERTS, 1970). KARICKHOFF & MOR- RiS (1985) obtained estimates of the diffusion coefficient in the pore water in the range of 1 to 3-10-9m2.s -1 for pentachlorobenzene and hexach- Iorobenzene, also higher than the molecular diffusion coefficient. Estimates by FISHER et al. (1983) were lower, and in the range 10 -11 to 10 -14 m2-s -1 for tri-, tetra- and pentachlorobiphenyls. Although FISHER et al. (1983) did find that the flux was inversely proportional to the square root of time, as predicted, the dependence on hydrophobicity was the opposite of the prediction, i.e. the more hydrophobic contaminants were more mobile.

TABLE 3 Logarithmic averages of the diffusion coefficient in the pore water obtained from eq. (15). The listed ranges correspond to the standard deviation of the log-averaged estimates.

Sediment effective diffusion coefficient (10-8.m2s- 7)

log averaged range

1 1.2 0.2- 7 2a 4.0 0.8 - 21 2b 1 1 0.2 - 5 3 1.2 0.3- 6 total 2.0 0.4 - 11

v v i v

Fig. 6. Relation between the experimental set-up and the linear column used for calculating the effect of pore water flushing on the flux. Top: experimental set-up with the original streamlines of the secondary flow. Middle: intermediate transformation. Bottom: sediment column with linear streamlines, used in the numerical model. The length L of the linear column is taken as the distance between the centres of the regions of inflow and outflow, which is ap-

proximately half the microcosm radius.

4.5. EXPLANATION OF THE HIGH DIFFUSION COEFFICIENT

The diffusion coefficients obtained with the data of the present study are surprisingly high: 20 to 80 times higher than estimated molecular diffusion coefficients. One reason might be that the chlorinated hydrocarbon concentration in the sediment may have been underestimated by 30% (see experimental methods section), which would cause an overesti- mation of the diffusion coefficient by a factor of 1.7. Increased mixing of the pore water in the upper mil-

limetres of the sediments by the flow in the water column could also contribute to the high diffusion coefficient. For example, estimates of the effective diffusion coefficient of silica in the pore water of surface sediments in the southern North Sea by VAN- DERBORGHT et al. (1977) were of the order 10 -8 m2-s -1. This high value was attributed to random mixing induced by the flow in the overlying water.

Another, potentially more serious, possibility is that the diffusion coefficient is artificially high because the radial pressure gradient that exists in devices with a rotating water column causes flushing of the pore water. Because of the pressure gradient, water may enter the sediment near the wall of the microcosm and leave at the centre. For sandy sediments the interstitial flow velocity of this irrigation flow is be tween 0.4 and 10 #m.s-1 at friction velocities of 0.37 and 0.95 cm.s -1, respectively (BOOu et al., 1991). These authors showed that the high estimate of the diffusion coefficient of oxygen in a sandy sediment obtained with an identical microcosm was artificially high because of this phenomenon.

We applied a numerical diffusion-advection model to assess the importance of pore water flushing on the flux of the chlorinated hydrocarbons in our exper iment. To reduce the complexity of this model we have simplified the geometry of the physical system to a linear sediment column of length L and a uniform flow velocity u of the pore water, as shown in Fig. 6. The differential equation to be solved is (cf. BERNER, 1980):

, c = _ ± [u,C]+ o r, cT_ ,10, #t I+K I+K Lo~x 2/

for the boundary conditions

C(x,t)=0 x=0 C(x,t)=0 x=L

and the initial condition

(17a) (17b)

C(x,0) = CP t = 0 , 0 < x < L (17c)

Eq. (16) was rewritten in non-dimensional form by making the substitutions

D t T= (18a)

L2(1 + K)

x = x (18b) L

C c = - - (18c)

Substituting the transformations (18a, b, c) in eq. (16) gives

clC=_pe i o~c ] [ °~20 ] a,- L~-J + (19) L ax 2 J

where Pe is the column Peclet number

pe=U__L D (20)

Eq. (19) was integrated, using the Crank-Nicholson algorithm (CRANK, 1957). For each integration the mass balance was evaluated by comparing the change in the inventory with the integrated flux for each time interval r - r + & . Time- and length steps were decreased until the mass balance was satisfied within 5%. The numerical solution for Pe=0 was checked against the analytical solution given by CRANK (1957).

The average flux from the sediment is obtained by averaging the fluxes at x=0 and at x=L:

aC j= -~ D /

~xxd x=0, or x=L (21)

Substituting transformations (18a, b, c) in eq. (21) gives the dimensionless flux (J):

j = j L dcj D CP d X X=0, or X= l (22)

The effect of pore water flushing can now be evaluated by comparing the dimensionless flux at various Peclet numbers. The results of the integration at Pe=15 and Pe=50 are shown in Fig. 7. For comparison, the results of the vertical diffusion model (eq. 10) are also shown. This model can be written in dimensionless form by substituting eqs (18a, b, c) in eq. (10):

1 J = - - (23)

As can be seen in Fig. 7, the flux at Pe>0 approaches the flux predicted by the vertical diffusion model as r approaches 0. The reason for this is that at r=0 the concentration gradient at the sediment- water interface is infinite. Therefore, any contribution from advection is overruled by diffusion. For longer time periods the flux is larger than predicted by the vertical diffusion model. The slope of log(J) versus log(r) is then larger than -0.5. A sharp decrease of the flux is expected when the sediment has been flushed with (I+K) pore volumes, i.e. at time

t= L (1+ K___~) (24) U

1o:, -

i0;" i j L

~ D C P

i0-2-~ .... ---- .......... T .... ~---

10 -6 10 "-~

, I i i

1 0 -2 1 0 o

L2(I+K)

Fig. 7. Dimensionless flux as a function of dimensionless time at two values of the P~clet number (Pe=15 and Pe=50). The relation obtained from the vertical diffusion model (Pe=0) is shown as the straight line. The filled circles indicate the time at which the sediment is flushed with I+K

pore volumes (see text).

or at dimensionless time

D 1 r = - (25)

u L Pe

These values for r are marked by closed circles on the graphs for Pe=15 and Pe=50.

In order to see if pore water flushing has affected our measurements seriously, we calculated the dimensionless experimental variables J and r from the data. To do so, we adopted the diffusion coefficients listed in Table 3, and a column length of 7.5 cm, which corresponds to half the microcosm radius. We expect the flow velocity in the pore water to be com- parable to the estimates given in BOOIJ et al. (1991), viz. 8 #m.s-1, except for sediment 3 which is a mud with a lower permeability to water flow. Relevant P~clet numbers are in the range 15 to 60. The results are plotted in Fig. 8. It can be concluded from this figure that, except for sediment 2a, all measurements are in the range where the vertical diffusion model still applies: Hence, the diffusion coefficient for sediment 2a (4-10 -8 m2-s -1) is overestimated. For all other sediments advection can be neglected.

Another question related to the effect of pore water flushing is whether the effective diffusion coefficient increased because of increased dispersion. An estimate of the effect of water flow on the diffusion coefficient can be obtained from the results of KOCH & BRADY (1985). Assuming a median grain size of 200

#DC p 101

i0 °_

i0 2_ jL

i0 °-

~ o O~o g D o

0 0 0 0

sediment #i

I . 0 I I I I

0 0 0 0

0 0 0 0 0 0

° ° \

sediment #2b

o o o ~ 0 O 0 0 0

o Q " < o \ o

sediment #2a

I I I I I

0 0 O0

~ - ~ o o o o .

sediment #3

I I I I I I I I I I 3.0 -6 10 -4 10 -2 10 -6 10 -4 10-2

D t D t

L2(I+K) L2(I+K)

~ % o ol

~o ° o

Fig. 8. Experimental values of the dimensionless flux versus dimensionless time. The relations for the diffusion-advection model (Pe=15 and Pe=50) are shown as drawn lines for reference.

#m, and a molecular diffusion coefficient of 5.10 -1° m2.s -1, the effective diffusion coefficient would be 1.3 and 3 times larger than the molecular diffusion coefficient with transverse and longitudinal diffusion, respectively. In view of the estimates from the flux experiments this contribution would be insignificant. We therefore have no reason to believe that pore water flushing induced by the radial pressure gradient has affected our measurements.

The final question regarding the high diffusion coefficients is the depth range to which the estimates apply. The depth (d) to which the concentration gra-

dients extend into the sediment can be estimated from penetration theory:

¢ ° d = ~r - - t (26) I +K

Using a time period of 300 hours and the values of the diffusion coefficients listed in Table 3, the depth scales are in the order of 5 to 16 mm for trichloro biphenyls and 1 to 2 mm for heptachlorobiphenyls. Consequently, the high diffusion coefficients apply to only the first few millimetres of the sediment.

5. REFERENCES

ANWAR, H.O., 1983. Turbulence measurements in stratified and well-mixed estuarine flows.--Estuar, coast. Shelf Sci. 17: 243-260.

BERNER, R.A., 1980. Early diagenesis. A theoretical approach. Princeton University Press, Princeton (N J): 9-89.

BIRD, R.B, W.E. STEWART & E.N. LIGHTFOOT, 1960. Trans- port phenomena. John Wiley & Sons, New York: 636-684.

BOEHM, P.D. & J.G. QUINN, 1973. Solubilization of hydrocarbons by the dissolved organic matter in sea water.-- Geochim, Cosmochim. Acta 37: 2459-2477.

BOOlJ, K., W. HELDER & B. SUNDBY, 1991. Rapid redistribu- tion of oxygen in a sandy sediment induced by changes in the flow velocity of the overlying water.-- Neth. J. Sea Res. 28: 149-165.

BOUDREAU, B.P. & N.L. GUINASSO, 1982. The influence of a diffusive sublayer on accretion, dissolution an diagenesis at the sea floor. In: K.A. FANNING & FT. MANHEIM. The dynamic environment of the ocean floor. Lexing- ton Books, Toronto: 115-145.

BROWNAWELL, B.J., 1986. The role of colloidal organic matter in the marine geochemistry of PCBs. Ph.D. thesis. MIT/WHOI, WHOI-86-19: 70-122.

BROWNAWELL, B.J. & J.W. FARRINGTON, 1986. Bio- geochemistry of PCBs in interstitial waters of a coastal marine sediment.--Geochim. Cosmochim. Acta 50: 157-169.

CHIOU, c.m., RE. PORTER & D.W. SCHMEDDING, 1983. Parti- tion equilibria of nonionic compounds between soil organic matter and water.--Environ. Sci. Technol. 17: 227-231.

CRANK, J., 1957. The mathematics of diffusion. University Press, London: 186-218.

DANCKWERTS, P.V., 1970. Gas-liquid reactions. McGraw-Hill Book Company, London: 6-21.

BANKERS, N. & R.W.P.M. LAANE, 1983. A comparison of wet oxidation and loss on ignition of organic material in suspended matter.--Environ. Technol. Letters 4: 283-290.

DUINKER, J.C. & M.T.J. HILLEBRAND, 1983. Determination of selected organochlorines in seawater. In: K. GRASHOFF, M. EHRHARDT & K. KREMLING. Methods of seawater analysis. Verlag Chemie, Weinheim: 290-309.

ELDER, D.L, S.W. FOWLER & G.G. POLIKARPOV, 1979. Remobilization of sediment-associated PCBs by the worm Nereis diversicolor.--Bull, environ. Contam. Toxicol. 21: 448-452.

FISHER, J.B., R.L. PETTY & W. LICK, 1983. Release of polychlorinated biphenyts from contaminated lake sediments: flux and apparent diffusivities of four individual PCBs.--Environ. Pollut. Ser. B 5: 121-132.

GSCHWEND, P.M. & S.C. WU, 1985. On the constancy of sediment-water partition coefficients of hydrophobic organic pollutants.--Environ. Sci. Technol. 19: 90-96.

GUST, G., 1988. Verfahren und Vorrichtung zum Erzeugen von definierten Bodenschubspannungen. Bundes- republik Deutschland, Patentschrift DE 3717969 Cl. Bundesdruckerei 05.88 808 128/463.

HALTER, MT. & H.E. JOHNSON, 1977. A model system to study the desorption and biological availability of PCB in hydrosoils. In: EL. MAYER & J.L. RAMELINK. Aquatic toxicology and hazard evaluation. ASTM STP 634. Am. Soc. Test. Mat.: 178-195.

HEATHERSHAW, A.D., 1979. The turbulent structure of the bottom boundary layer in a tidal current.--Geophys. J. R. astr. Soc. 58: 359-430.

KARICKHOFF, S.W., 1984. Organic pollutant sorption in aquatic systems.--J, hydraulic Eng. 110: 707-735.

KARICKHOFF, S.W. & K.R MORRIS; 1985. Impact of tubificid oligochaetes on pollutant transport in bottom sediments.--Environ. Sci. Technol. 19: 51-56.

KARICKHOFF, S.W., D.S BROWN & T.A. SCOTT, 1979. Sorp- tion of hydrophobic pollutants on natural sediments.--Water Res. 13: 241-248.

KOCH, DL. & J.E BRADY, 1985. Dispersion in fixed beds.--& Fluid Mech. 154: 399-427.

LARSSON, P, 1983. Transport of 14-C labelled PCB compounds from sediment to water and from water to air in laboratory model systems.--Water Res. 17: 1317-1326.

- - - - , 1984. Transport of PCBs from aquatic to terrestrial en- vironments by emerging chironomids.--Environ. Pol- lut. Ser. A. 34" 283-289.

- - - - , 1985. Contaminated sediments of lakes and oceans act as sources of chlorinated hydrocarbons for release to water and atmosphere.--Nature 317: 347-349.

LEVENSPIEL, O., 1962. Chemical reaction engineering. John Wiley & Sons, New York: 332-489.

MACKAY, D. & S. PATERSON, 1982. Fugacity revisited, the fugacity approach to environmental transport.--Environ. Sci. Technol. 16: 654A-660A

MEANS, J.C., S.G. WOOD, J.J. HASSETT & W.L. BANWART, 1980. Sorption of polynuclear aromatic hydrocarbons by sediments and soils.--Environ. Sci. Technol. 14: 1524-1528.

RAAPHORST, W. VAN, P RUARDtJ & A.G. BRINKMAN, 1988. The assessment of benthic phosphorus regeneration in an estuarine ecosystem model.--Neth. J. Sea Res. 22: 23-36.

SCHAUER, U., 1987. Determination of bottom boundary layer parameters at two shallow sea sites using the profile method.--Cont. Shelf Res. 7: 1211-1230.

SCHWARZENBACH, a.P & J. WESTALL, 1981. Transport of nonpolar organic compounds from surface water to groundwater. Laboratory sorption studies.--Environ Sci. Technol. 15: t360-1367.

SHIU, W.Y. & D. MACKAY, 1986. A critical review of aqueous solubilities, vapor pressures, Henry's law constants, and octanol-water partition coefficients of the polychlorinated biphenyl&--J. Phys. Chem. Ref. Data 15: 911-929.

SODERGREN, A. & P. LARSSON, 1982. Transport of PCBs in aquatic laboratory model ecosystems from sediment to the atmosphere via the surface microlayer.--Ambio 11: 41-45.

VANDERBORGHT, J.P., R. WOLLAST & G. BILLEN, 1977. Kinet- ic models of diagenesis in disturbed sediments. Part 1. Mass transfer properties and silica diagenesis.-- Limnol. Oceanogr. 22: 787-793.

WHITEHOUSE, B., 1985. The effects of dissolved organic matter on the aqueous partitioning of polynuclear aromatic hydrocarbons.--Estuar, coast. Shelf Sci. 20: 393-402.

WIJAYARATNE, R.D. & J.C. MEANS, 1984. Sorption of polycy- clic aromatic hydrocarbons by natural estuarine col- Ioids.--Mar. environ. Res. 11" 77-89.

WIMBUSH, M. & W. MUNK, 1970. The benthic boundary layer. In: A.E. MAXWELL. The sea. Part IV. Interscience, New York: 731-758.

WOOD, L.W., G.Y. RHEE, B. BUSH & E. BARNARD, 1987. Sedi- ment desorption of PCB congeners and their bio- uptake by dipteran larvae.--Water Res. 21: 875-884.

Wu, S.C. & P.M. GSCHWEND, 1986. Sorption kinetics of organic compounds to natural sediments and soils.- Environ. Sci. Technol. 20" 717-725.

(received 31 July 1991; revised 20 May 1992)

s y m b o l

SYMBOLS

q u a n t i t y u n i t

A CP C s C w

foc fom J J k kapp K K d

Kom Kow L n N Pe

Qw R t U X X

propor t iona l i ty constant between Kow and Kom . . . . . . . . . . . . . . . . . . . . . . . . . m 3.kg-1 area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m 2 concentra t ion in pore water at t = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . mo l .m -3 concentra t ion in sed iment phase at t = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . mol .kg -1 concentrat ion in water co lumn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . mol .m -3 d imens ion less concentrat ion C / CP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-] length scale of removal f rom the sed imen t . . . . . . . . . . . . . . . . . . . . . . . . . m effect ive di f fusion coeff ic ient in the pore water . . . . . . . . . . . . . . . . . . . . . . . . . m2.s -1 mass fract ion organ ic carbon in sed iment phase . . . . . . . . . . . . . . . . . . . . . . [-] mass fract ion organ ic matter in sed iment phase . . . . . . . . . . . . . . . . . . [-] f lux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m o l . m - 2 . s -1 d imens ion less f lux j L/(e D CP) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [.] mass transfer coeff ic ient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m.s-1 apparent mass transfer coeff ic ient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m . s - 1 (1-q~) e Kd /~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-] sed iment water part i t ion coeff ic ient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m3.kg -1 sed iment water part i t ion coeff ic ient normal ized on fract ion organ ic matter , m3.kg -1 octanol water part i t ion coeff ic ient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-] co lumn length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m

amount of con taminant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . mot number of observat ions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-] co lumn P6cl~t number u L / D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-] water vo lume rate of f low . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m3.s -1 correlat ion coeff ic ient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-] t ime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . s

f low veloci ty in the pore water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m.s -1 hor izontal coord inate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m d imens ion less hor izontal coord inate x/L . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-]

GREEK SYMBOLS

extract ion ef f ic iency for s ingle f low through . . . . . . . . . . . . . . . . . . . . . . . . . [-] 3,14159 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-] s tandard deviat ion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-] porosi ty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-] dens i ty of the sed iment par t ic les . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . kg.m -3 d imens ion less t ime t D / [L 2 ( I+K) ] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [-]

Release rates of chlorinated hydrocarbons from contaminated sediments

Documents