Chemosphere 72 (2008) 1455–1461

Contents lists available at ScienceDirect

Chemosphere

journal homepage: www.elsevier .com/locate /chemosphere

Black carbon-dominated PCDD/Fs sorption to soils at a formerwood impregnation site

Gerard Cornelissen a,b,*, Ian T. Cousins a, Karin Wiberg c, Mats Tysklind c, Henning Holmström d,Dag Broman a

a Department of Applied Environmental Sciences (ITM), Stockholm University, 10691 Stockholm, Swedenb Norwegian Geotechnical Institute (NGI), P.O. Box 3930 Ullevaal, N-0806 Oslo, Norwayc Department of Chemistry, Umeå University, SE-90187 Umeå, Swedend Geological Survey of Sweden (SGU), P.O. Box 16247, SE-10324 Stockholm, Sweden

a r t i c l e i n f o

Article history:Received 25 January 2008Received in revised form 30 April 2008Accepted 6 May 2008Available online 17 June 2008

Keywords:DioxinsFuransSorptionSoilBlack carbon

0045-6535/$ - see front matter � 2008 Elsevier Ltd. Adoi:10.1016/j.chemosphere.2008.05.010

* Corresponding author. Address: Norwegian GeotBox 3930 Ullevaal, N-0806 Oslo, Norway. Tel.: +47 22

E-mail address: gerard.cornelissen@itm.su.se (G. C

a b s t r a c t

The influence of black carbon (BC) on the sorption of 17 native polychlorinated-p-dibenzodioxins(PCDDs) and polychlorinated dibenzofurans (PCDFs) was studied in five soil samples from a sawmill sitewhere wood used to be impregnated with chlorophenol preservatives. The presence of BC caused mea-sured total organic carbon (TOC)–water distribution ratios (KTOC) to be a median factor of 51 (interquar-tile range 18–68, n = 85) higher than modeled amorphous organic carbon (AOC)–water distribution ratios(KAOC). KTOC was a factor of 73 ± 27 above KAOC for PCDFs (n = 10) and a factor of 20 ± 13 (n = 7) for PCDDs.The reason for this difference is probably that attaining a planar configuration after sorption to BC is lessthermodynamically favorable for PCDDs than for PCDFs. BC–water distribution ratios were calculatedfrom KTOC, KAOC and BC contents, and ranged from 109.9 (2,3,7,8-Tetra-CDD) to 1011.5 l kg�1 (Octa-CDF).More than 90% of the PCDD/Fs in the soil was calculated to be BC-sorbed. Dissolved organic carbon(DOC)–water distribution ratios were measured to be in the same order of magnitude as KAOC. This studyshows that strong sorption to BC should be included when assessing ecotoxicological risk or modelingtransport to groundwater of PCDD/Fs in soil.

� 2008 Elsevier Ltd. All rights reserved.

1. Introduction

Soil pollution as a result of wood impregnation with tetra- andpenta-chlorophenol preservatives occurs at hundreds of locationsin various countries such as Sweden, Finland and Canada (Jensen,1996). The contamination usually consists of chlorophenols, poly-chlorinated dibenzofurans (PCDFs), polychlorinated-p-dibenzodi-oxins (PCDDs), chlorinated phenoxyphenols and chlorinateddiphenylethers (Persson et al., 2007). Total PCDD/Fs contents insuch soils can amount up to three mg World Health OrganizationToxic Equivalents (WHO-TEQ) per kg dry weight (Persson et al.,2007). As a result, PCDD/Fs can leach to groundwater (Frankkiet al., 2006) and accumulate in organisms. Leaching (Huang et al.,2003; Cornelissen et al., 2005; Frankki et al., 2006), bioaccumula-tion (Moermond et al., 2005) and toxicity (Knauer et al., 2007)can be reduced by the strong binding of PCDD/Fs to carbonaceousgeosorbents (CGs) including black carbon (BC), unburned coal andkerogen. Since no coal mining has existed in the area, BC is ex-pected to dominate possible strong PCDD/F sorption to CG.

ll rights reserved.

echnical Institute (NGI), P.O.023159; fax: +47 22230448.ornelissen).

For PAHs and PCBs in sediments, BC and other carbonaceousgeosorbents have been shown to enhance sorption by a factor ofup to 100 (Jonker and Koelmans, 2001; Bärring et al., 2002; Huanget al., 2003; Cornelissen et al., 2005; Lohmann et al., 2005). ForPCDD/Fs in soils, the observations are scarce. Three studies indicatethat BC could enhance sorption of these compounds as well. First,the BC–water distribution ratios (KBC) of spiked mono-, di-, tri- andtetra-substituted PCDD/Fs in diesel soot were 20–300 times higherthan amorphous organic carbon (AOC)–water distribution ratios(KAOC) (Bärring et al., 2002). Second, the KBCs of native di-CDDand tri-CDD in New York Harbor sediment were two orders ofmagnitude above KAOC (Lohmann et al., 2005). Third, theTOC–water distribution ratios (KTOC) of native tetra-, penta-,hexa-, hepta- and octa-chlorinated CDD/Fs in Norwegian Mg-smel-ter contaminated sediments were a factor of 40–3330 above KAOC

(Persson et al., 2002). In the third study freely dissolved aque-ous concentrations (CW,free) were not measured directly butdetermined by correcting total aqueous concentrations (CW,total)for dissolved organic carbon (DOC) sorption, using non site-specificliterature KDOC-values.

In contrast, no effect of BC on PCDD/F sorption was observedfor native PCDD/Fs in a soil sample from a wood impregnationsite in Northern Sweden (Sikeå) (Frankki et al., 2006). Possible

1456 G. Cornelissen et al. / Chemosphere 72 (2008) 1455–1461

explanations for the apparent lack of BC-sorption observed in thisstudy are discussed later in the manuscript.

In the present study, sorption of PCDD/Fs to TOC, BC and DOCwas studied in five soils affected by chlorophenol wood impregna-tion. All 2,3,7,8-chlorinated PCDD/F congeners were included in thestudy. These congeners show extremely high toxicity through aspecific mode of toxic action through binding to the aryl hydrocar-bon receptor. Novel elements were added to the knowledge pre-sented in the abovementioned studies by (i) providing KBC-valuesfor 2,3,7,8-substituted PCDD/F congeners in soils, (ii) measuringfreely dissolved aqueous concentrations with a passive samplermethod, avoiding effects of incomplete phase separation, and (iii)measuring both CW,free, CW,total and DOC concentrations, providingreliable data on KDOC for these compounds. Sorption to environ-mental BC has not been studied before for 2,3,7,8-substituted tet-ra- to octa-CDD/Fs, and PCDD/F sorption to BC has not beenstudied before for soil samples.

2. Materials and methods

2.1. Materials

All solvents used were of glass-distilled purity (Burdick and Jack-son, Muskegon, MI, USA). All 13C-labelled and 12C-PCDD/Fs were ob-tained from Wellington Laboratories, Ontario, Canada.Polyurethane foam (PUF) adsorbents were obtained from Ekornes,Norway, and precleaned by twice 24-h extraction in subsequentlytoluene and acetone. POM (polyoxymethylene) of 55-lm thickness(POM-55, obtained in �1-kg cylinder-shaped blocks and sliced on alathe equipped with a high-precision razor blade) from Astrup AS(Norway) was pre-cleaned by washing in hexane and methanol fol-lowed by drying overnight at 60 �C. Seventeen 13C-labelled PCDD/Fs(all 2,3,7,8-substituted) were used as internal standards. The quan-tification was made by the isotope dilution method and by using a12C-mixture of 17 PCDD/Fs (all 2,3,7,8-substituted). Quantificationmethods from Danielsson et al. (2005) were exactly followed.

2.2. Soil samples

Samples were collected at a former sawmill site at Marieberg,Sweden (62�590800, E 17�4705100), where around 260 kg WHO-TEQPCDD/F are present. Samples I and II consisted of soil from the in-ner sawmill area (<50 m from the sawmill), sample III of sawdustfrom the inner sawmill area, and samples IV and V of soil fromthe outer sawmill area (350 and 450 m from the sawmill, respec-tively). Samples (0–50 cm depth; >5 kg) were collected in summer2006 with a spade, sieved over a 5-mm steel sieve, homogenizedusing a blender, and analyzed fresh, without prior drying.

2.3. BC and TOC contents

Total organic carbon (TOC) contents were determined with cat-alytic combustion elemental analysis at 1030 �C after micro-acidi-fication (1 M HCl) to remove inorganic carbonates. Black carbon(BC) contents were determined using the same method on smallsamples (10-mg) that were ball-ground to optimize oxygen access.Samples were combusted at 375 �C for 18 h under abundant oxy-gen access, following exactly the procedures outlined by Gustafs-son et al. (1997). The temperature was controlled independentlyand shown to be 375 ± 1 �C.

2.4. POM calibration for PCDD/Fs

All POM calibration studies were carried out in triplicate, in1000-ml all-glass flasks, in the presence of 1 g NaN3 (biocide alsoproviding a constant ionic strength). The principle of the POM

equilibrium passive sampling method is that one can deduce freelydissolved aqueous concentrations (CW) from the PCDD/F contentsin POM with concentration-independent POM–water distributionratios (KPOM) (Jonker and Koelmans, 2001).

In the present study, KPOM was measured in POM–water sys-tems with spiked PCDD/Fs, without soil. KPOM was measured foreight non-2,3,7,8-substituted congeners of variable degree of chlo-rination (2,8-DiCDF, 1,6-DiCDD, 2,4,8-TriCDF, 1,3,6,8-TetraCDF,1,3,6,8-TetraCDD, 1,2,3,8,9-PentaCDF, 1,2,3,4,6,9-HexaCDD and1,2,3,4,6,7,9-HeptaCDD), and deduced KPOM for the toxic, 2,3,7,8-substituted congeners from the KPOM–KOW linear free-energy rela-tionship (LFER) obtained by linear regression of measured KPOM vs.KOW for the non-2,3,7,8-substituted congeners.

KPOM was measured at a range of methanol–water ratios (0–50%methanol) and derived KPOM for pure water by extrapolating to 0%methanol. Measurement in methanol–water cosolvent systems,similarly to Smedes (2007), was chosen because of the difficultyin measuring aqueous PCDD/F concentrations at pg to ng per lranges in 1 l pure water.

PCDD/F stock solution (prepared by dissolving the compoundsin acetone and spiking <500 ll acetone per l water) and POM(25–250 mg) were shaken horizontally (180 rpm; 20 ± 1 �C;100 d). This 100-d equilibration time is longer than equilibrationtimes of around 30 d used for similar experiments with nine timesthicker POM (500 lm) in earlier studies (Jonker and Koelmans,2001; Cornelissen and Gustafsson, 2004). After equilibration, indi-vidual PCDD/F concentrations were 0.1–20 ng l�1, depending oncongener and methanol content. The POM-strips were taken outof the water-methanol solutions using tweezers and treated ex-actly as described in Cornelissen and Gustafsson (2004). Briefly,POM strips were dried with a tissue and extracted by horizontalshaking (180 rpm; 20 ± 1 �C; 48 h) with 20 ml hexane in the pres-ence of internal standard (see above). The aqueous phase was ex-tracted with 2 � 25 ml hexane. Hexane extracts were cleanedand analyzed according to the procedures described below.

2.5. Freely dissolved concentrations and KTOC

All sorption studies were carried out in triplicate, at 20 ± 1 �C, in250-ml all-glass flasks, with NaN3 (1 g l�1). Soil (non-dried, 2 g dryweight) was shaken horizontally (180 rpm, 14 d) with distilled water(250 ml). Subsequently POM (2 g) was added and the systems wereshaken for another 30 d. No PCDD/Fs were added as native in situcompounds were studied. It was assumed that 30 d was long enoughfor equilibration of the systems where soil and water were pre-equil-ibrated. The equilibration time of 14 + 30 d was longer than the 10 dearlier shown to be sufficient for equilibration of native six-ring PAHsin a sediment–water–POM-55 system (Cornelissen et al., 2008). Four-teen days sufficed for equilibration of PAHs, PCBs and trichloro-PCDD/Fs in 51-lm thick LDPE (Lohmann et al., 2005). After equilibration, thePOM strips were extracted with hexane as described above. Cleanupand analysis of the extracts was as described below.

Freely dissolved soil porewater concentrations (CW,free) were de-duced from the PCDD/F contents in the POM samplers with theconcentration-independent KPOM values determined in the POMcalibration experiments. The passive samplers turned out to de-plete 0.1–5% (average < 1%) of the PCDD/Fs in the soil phase. Thereported aqueous concentrations were corrected for this depletion.No correction of KTOC for the ‘‘salt effect” caused by the biocide con-centration (0.015 M), as the correction was <0.005 log-unit andthus within the experimental error.

2.6. Total dissolved concentrations and KDOC

Soil (1000 g) was slowly percolated with 0.01 M CaCl2 in water(50 ml/h) for 14 days in a 10-cm glass column until a L/S (liquid/

Table 1LogKOW and logKPOM for non-2,3,7,8-substituted (measured) and 2,3,7,8-substituted(derived from logKPOM–log KOW relationship) PCDD/F congeners

5.2

5.4

5.6

5.8

6

6.2

6.4

5 5.5 6 6.5 7.5 8 8.5log K OW

log

KP

OM (

L/kg

)

2

3

4

5

6

7

0 10 20 30 40 50f methanol

log

KP

OM (L

/kg

)

26-DiCDD 248-TriCDF123469-HxCDD 1234679-HpCDD

7

a

b

Fig. 1. (a) LogKPOM vs. fmethanol, the fraction of methanol in water (%). Example forfour of the seven measured non-2,3,7,8-substituted congeners, (b) logKPOM–logKOW

relationship for the seven measured non-2,3,7,8-substituted PCDD/F congeners.

G. Cornelissen et al. / Chemosphere 72 (2008) 1455–1461 1457

solid) ratio of 15–50 was reached. The eluate was filtered for par-ticles by a 1-lm glass fiber filter (GFF, Whatman, Germany) andsubsequently led over a PUF column absorbing all freely dissolvedand DOC-sorbed PCDD/Fs. Thus, DOC was operationally defined asthe TOC in water passing a 1-lm filter. DOC concentrations in theeluate were measured after 1, 5, 10 and 14 days. This eluate(4 � 15 ml) was sampled directly after passage of the 1-lm GFF fil-ter and before passage of the PUF absorbent. Eluate samples wereacidified with 0.3 mmol HCl and analyzed for DOC contents in aShimadzu 5000 TOC analyzer. The total amount of water neededfor these analyses (60 ml) was negligible to the total amount ofwater that passed the PUF absorbents (16 l) and thus did not affectthe PCDD/F quantifications in the PUF. A GFF-PUF setup has previ-ously been tested and verified for PCDD/F analysis (Persson et al.,2002). Amounts passing the PUF filters were checked for sampleI and were <10% of the amounts retained on the filters.

2.7. Extraction, cleanup and PCDD/F analysis

The method followed the procedure in Danielsson et al. (2005).Non-dried soil mixed with anhydrous Na2SO4 (2–5 g dry weight)and PUF absorbents were extracted with toluene (400 ml) andinternal standard in a Soxhlet-Dean–Stark for 17 h. Two open li-quid chromatographic columns were used for the clean-up of theextracts. First, a multilayer column (Silica 60, Merck, Darmstadt,Germany) with KOH–silica, neutral silica, H2SO4–silica and Na2SO4

was used. The second column contained a activated carbon(AX21):Celite mix (carbon from Anderson Development Co., MI,USA and Celite 545� from Fluka, Switzerland), where the PCDD/Fs were collected in the second fraction by reversed elution withtoluene. Analyses were performed using an HRGC/HRMS system(an Agilent 6890N gas chromatograph coupled to a Waters Auto-spec Ultima mass spectrometer (Milford, MA, USA) equipped witha 60 m � 0.32 mm DB-5MS column with a 0.25 lm thick film, J&WScientific, CA, USA). The limit of detection (LOD) was calculated asthree times the average area of noise in GC–MS chromatogramsfrom > five samples. The LOD varied from 0.3 to 3 pg/sample forthe various congeners.

Non-2,3,7,8-substituted congeners LogKOWa LogKPOM extrapolatedb

2,8-DiCDF 5.34 5.35 ± 0.071,6-DiCDD 5.4 5.21 ± 0.052,4,8-TrCDF 5.9 5.79 ± 0.061,3,6,8-TCDF 6.35 5.82 ± 0.141,3,6,8-TCDD 7 5.89 ± 0.141,2,3,8,9-PeCDF 6.76 5.8 ± 0.31,2,3,4,6,9-HxCDD 7.92 6.1 ± 0.21,2,3,4,6,7,9-HpCDD 8.42 6.21 ± 0.13

2,3,7,8-Substituted congeners LogKOWa LogKPOM derivedc

2,3,7,8-TCDF 6.41 5.71,2,3,7,8-PeCDF 6.74 5.82,3,4,7,8-PeCDF 6.8 5.81,2,3,4,7,8-HxCDF 7.46 6.01,2,3,6,7,8-HxCDF 7.56 6.01,2,3,7,8,9-HxCDF 7.44 6.02,3,4,6,7,8-HxCDF 7.43 6.01,2,3,4,6,7,8-HpCDF 7.81 6.11,2,3,4,7,8,9-HpCDF 7.92 6.1OCDF 8.43 6.32,3,7,8-TCDD 7.05 5.91,2,3,7,8-PeCDD 7.06 5.91,2,3,4,7,8-HxCDD 7.93 6.11,2,3,6,7,8-HxCDD 7.93 6.11,2,3,7,8,9-HxCDD 7.91 6.11,2,3,4,6,7,8-HpCDD 8.42 6.3OCDD 8.85 6.4

a From (Sacan et al., 2005).b From five measurements in the range 10–50% methanol, extrapolated to 0%

MeOH.c From logKPOM–logKOW relationship for non-2,3,7,8-substituted congeners.

3. Results and discussion

3.1. POM–water distribution ratios

KPOM values for PCDD/Fs have not been reported before andwere thus established in the present study. Relationships betweenKPOM and methanol content were obtained in the range 10–50%methanol, and assumed to be linear (Fig. 1a). The curves leveledoff at 0% methanol (pure water). This underestimation is probablydue to erroneously high measured PCDD/F contents in the metha-nol-free aqueous phase. Small amounts of sorbing solids in theaqueous phase can lead to overestimations of CW of extremelyhydrophobic compounds such as PCDD/Fs in pure water in labora-tory systems.

Therefore KPOM for pure water (0% methanol) was derived forthe non-2,3,7,8-substituted congeners through extrapolation ofthe 10–50% methanol range (Fig. 1a, Table 1). All measured KPOM

values at the various MeOH fractions including 0% MeOH are pre-sented in Table S1. Extrapolated KPOM values as used in the remain-der of the paper were approximately a factor of three higher thanvalues measured directly in pure water. Using the latter KPOM val-ues in the calculations would result in KTOC and KBC values a factorof three (0.5 log-unit) lower than reported.

Subsequently KPOM for the 2,3,7,8-substituted PCDD/Fs was de-rived from the logKPOM–logKOW relationship derived for the non-2,3,7,8-substituted congeners (Fig. 1b, Table 1). The logKPOM–logKOW

1458 G. Cornelissen et al. / Chemosphere 72 (2008) 1455–1461

relationship for non-2,3,7,8-substituted PCDD/Fs (Fig. 1b, r2 = 0.84)was used to derive KPOM for the 2,3,7,8-substituted congeners.Since sorption to POM is a partitioning process rather than adsorp-tion, values for non-2,3,7,8-substituted PCDD/Fs were expected tobe usable to estimate KPOM for the 2,3,7,8-substituted ones. The to-tal error in the derived KPOM values was estimated to be in the or-der of 0.4 log-units, a summation of the errors in (i) the regressionsof logKPOM vs. fmethanol (Fig. 1a), where the median value of thestandard deviation in KPOM (the intercept at 0% methanol) was0.12 log-units for the seven measured congeners, (ii) the logK-POM–logKOW regression (0.1 log-unit), and (iii) the literature log-KOW values from Sacan et al. (2005) (estimated to be in the orderof 0.2 log-units).

3.2. TOC, BC and PCDD/F contents

TOC contents showed a large variation for the five soils(0.31–44%; Table 2), with the highest value for sawdust sampleIII. BC was present in all samples (0.04–0.4%; Table 2). BC was0.6–13% of TOC (Table 2); these BC/TOC ratios were in accor-dance with literature values (median literature value for 90 soils

Table 2TOC, BC, PCDD/F, PAH and oil contents of the soils (triplicate measurements with standar

I II I

Soil sawmill area Soil sawmill area S

TOC (%) 7.7 ± 0.5 1.8 ± 0.4 4BC (%) 0.096 ± 0.005 0.035 ± 0.004 0BC/TOC (%) 1.24 ± 0.15 1.9 ± 0.7 0PCDD/Fa (lg WHO-TEQ kg�1) 30.4 ± 1.5 1.34 ± 0.15 0PAHb (mg kg�1 16-EPA) 0.44 0.56 0Oil (mg kg�1)

C10–C12 <10 <10 <C12–C16 <10 <10 <C16–C35 <10 15 2

a Contents of individual compounds in Table S1.

Table 3LogKOW, logKAOC, logKTOC, logKBC at in situ PCDD/F concentrations, and logKF,BC derived witand KBC values are median values for five soil samples (range for the five soils in brackets

LogKOWa LogKAOC

b LogKTOCc LogKBC at in situ

CW,free

Ln

LW Loct�1 LkgAOC�1 LkgTOC�1 LkgBC�1 ðp

2,3,7,8-TCDF 6.41 5.93 7.9 (7.6–8.7) 10.0 (9.6-10.5)2,3,7,8-TCDD 6.74 6.26 8.0 (7.7–8.6) 9.9 (9.7–10.3)1,2,3,7,8-PeCDF 6.80 6.32 8.1 (7.9–9.4) 10.1 (10.0–10.8)2,3,4,7,8-PeCDF 7.46 6.98 8.5 (8.0–9.3) 10.2 (10.1–11.0)1,2,3,7,8-PeCDD 7.56 7.08 8.2 (7.7–9.0) 10.2 (9.9–10.7)1,2,3,4,7,8-HxCDF 7.44 6.96 8.8 (8.0–9.5) 10.4 (10.2–11.2) 11,2,3,6,7,8-HxCDF 7.43 6.95 8.8 (8.0–9.5) 10.4 (10.2–11.2)1,2,3,7,8,9-HxCDF 7.81 7.33 8.8 (8.2–9.7) 10.6 (10.3–11.2)2,3,4,6,7,8-HxCDF 7.92 7.44 8.8 (8.1–9.5) 10.4 (10.2–11.2) 11,2,3,4,7,8-HxCDD 8.43 7.95 8.5 (8.1–9.4) 10.5 (10.1–11.1)1,2,3,6,7,8-HxCDD 7.05 6.57 8.7 (8.1–9.5) 10.4 (10.2–11.3) 11,2,3,7,8,9-HxCDD 7.06 6.58 8.6 (8.1–9.7) 10.5 (10.2–11.4)1,2,3,4,6,7,8-

HpCDF7.93 7.45 9.1 (8.3–9.9) 10.8 (10.4–11.5) 1

1,2,3,4,7,8,9-HpCDF

7.93 7.45 9.2 (8.5–10.1) 11.0 (10.5–11.6) 1

1,2,3,4,6,7,8-HpCDD

7.91 7.43 9.0 (8.4–10.0) 10.9 (10.4–11.6) 1

OCDF 8.42 7.94 9.7 (8.8–10.6) 11.5 (10.8–11.9) 1OCDD 8.85 8.37 9.4 (8.8–10.3) 11.2 (10.8–12.0) 1

a From Sacan et al. (2005).b Derived with logKAOC = logKOW � (0.48 ± 0.42) (Seth et al, 1999).c Measured values.d Units ((pg kg�1

BC )/(pg l�1)n) represent KBC at 1 pg l�1.e Units ((lg kg�1

BC )/(lg l�1)n) represent KBC at 1 lg l�1.

4% with interquartile range, IQR, 2–13% (Cornelissen et al.,2005)). This BC probably consists of mainly soot, since charsdo not survive the CTO-375 BC isolation method (Elmquistet al., 2006).

Total PCDD/F soil contents of samples I, II and III (30, 1.3 and0.8 lg WHO-TEQ kg�1 dw, respectively) exceeded the Swedishgeneric guideline value for less sensitive land use (0.25 lgWHO-TEQ kg�1 (Swedish EPA, 1998)). The dominant congenerswith regard to mass were 1,2,3,4,6,7,8-HpCDF and OCDF (account-ing for 71–95% of total mass and for 35–69% of total TCDD-equiv-alent toxicity; Table S2).

3.3. TOC–water distribution ratios

TOC–water distribution ratios for all 17 PCDD/Fs in all five soilsamples, KTOC, were higher than values obtained through thewidely used LFER logKAOC = logKOW–(0.48 ± 0.42) (Seth et al.,1999) (Table 3; Fig. 2a). The difference is a median factor of 51(IQR 18–68, n = 85). Note that an additional uncertainty exists inthis factor 51; this is the factor 2.5 bandwidth in LFER of Sethet al. (1999).

d deviations, single measurements for PAH and oil), as well as BC/TOC ratios

II IV V

awdust sawmill area Soil outer sawmill area Soil outer sawmill area

3.7 ± 0.3 0.31 ± 0.01 7.4 ± 0.4.40 ± 0.01 0.039 ± 0.01 0.045 ± 0.002.92 ± 0.04 13 ± 3 0.61 ± 0.06.83 ± 0.08 0.14 ± 0.02 0.14 ± 0.03.41 <0.3 0.19

10 <10 <1010 <10 <101 20 14

h various nF,BC values (0.5, 0.61 and 0.8, respectively) and in different units. KTOC, KF,BC

)

ogKF,BCd

F,BC = 0.61LogKF,BC

d

nF,BC = 0.5LogKF,BC

d

nF,BC = 0.8LogKF,BC

e

nF,BC = 0.61

gkgBC�1 Þ=ðpgL�1Þn ðpgkgBC�1 Þ=ðpgL�1Þn ðpgkgBC�1 Þ=ðpgL�1Þn ðlgkgBC�1 Þ=ðlL�1Þn

9.1 (8.8–9.5) 8.8 9.5 6.88.9 (8.6–9.2) 8.6 9.4 6.69.1 (9.0–9.6) 8.7 9.6 6.89.6 (9.2–10.1) 9.3 10.0 7.39.3 (8.7–9.9) 9.1 9.7 7.10.1 (9.4–10.4) 9.8 10.3 7.89.9 (9.4–10.3) 9.6 10.2 7.69.9 (9.4–10.3) 9.6 10.3 7.60.0 (9.5–10.5) 9.8 10.3 7.89.4 (8.9–10.1) 9.1 9.9 7.20.0 (9.4–10.8) 9.8 10.2 7.79.8 (9.2–10.6) 9.5 10.1 7.51.0 (10.3–11.3) 11.0 10.7 8.7

0.2 (9.6–10.6) 9.9 10.5 8.0

0.4 (9.7-11.3) 10.2 10.5 8.1

1.2 (10.5–11.6) 11.4 11.3 9.00.7 (10.0–11.4) 10.5 10.9 8.4

6

7

8

9

10

11

6 6.5 7 7.5 8 8.5 9log KOW

6 6.5 7 7.5 8 8.5 9log KOW

log

KT

OC (

L/k

g)

LFER-KAOC TOC-I TOC-IITOC-III TOC-IV TOC-VTOC average LFER limit

6

7

8

9

10

log

KD

OC (

L/k

g)

LFER-KAOC DOC-I DOC-IIDOC-III DOC-IV DOC-VDOC average LFER limit

a

b

Fig. 2. (a) LogKTOC vs. logKOW for 17 PCDD/Fs in five soils (n = 85). The solid line isfor LFER-derived logKAOC (see text). The dotted line shows the variation in KAOC

(Seth et al., 1999). The dashed line is the average logKTOC value for each PCDD/F inthe five soils (line thus based on 17 data points), (b) LogKDOC vs. logKOW for 17PCDD/Fs in five soils. The solid line is for LFER-derived logKAOC (Seth et al., 1999).The dotted line shows the variation in KAOC (Seth et al., 1999). The dashed line is theaverage logKDOC value for the five soils.

G. Cornelissen et al. / Chemosphere 72 (2008) 1455–1461 1459

The observed high logKAOC values are in accordance with earlierstudies on freely dissolved PAH, PCB and PCDD/F concentrations incontaminated harbor sediments (Jonker and Koelmans, 2001; Cor-nelissen et al., 2005; Lohmann et al., 2005). BC probably causedthis strong sorption (Jonker and Koelmans, 2001; Cornelissenet al., 2005; Lohmann et al., 2005). The uncertainty of approxi-mately 0.4 log-unit in KPOM will directly work through in the KTOC

estimates. Therefore standard errors in the values in Table 3 are>0.4 log-units. The soil with the highest BC/TOC ratio (soil IV)showed highest KTOC values. However, soil II also showed high KTOC

values with BC/TOC ratios in the same order of magnitude as theother three samples.

KTOC values of PCDD/Fs in a similar aged wood impregnation soilfrom Northern Sweden (Sikeå) were 1–1.5 log-units below the cur-rently observed values (Frankki et al., 2006), albeit for only one soilsample. Two factors might explain this difference. Most impor-tantly, CW was determined in solutions containing DOC contentsof around 10 mg l�1 after flocculation. As the authors them-selves indicated, this DOC can substantially reduce apparent KTOC

values for extremely hydrophobic compounds such as tetra- toocta-CDD/Fs. In the present passive sampling approach, such DOCeffects are avoided. In addition, the BC content in the soil (0.6%of TOC (Frankki et al., 2006)) was lower than that in four of the fivepresently studied samples.

KTOC-values high above KAOC (factor of 40–3330) were observedfor in situ sorption of many native tetra- to octa-CDD/Fs in a

Norwegian Mg-smelter contaminated sediment (Persson et al.,2002), with BC/TOC ratios similar to the currently studied soil sam-ples (2–15%).

3.4. BC-inclusive sorption

KTOC was reinterpreted with a BC-inclusive Freundlich sorptionmodel, analogous to the modeling in Cornelissen and Gustafsson(2004) and Lohmann et al. (2005)

CS ¼ fAOCKAOCCW;free þ fBCKF;BCCnF;BCW;free ð1Þ

where CS is the concentration in the soil (lg kg�1 dw), fAOC and fBC

are the soil mass fractions of AOC and BC, respectively, KAOC is theAOC–water distribution ratio (l kg�1), CW,free is the freely dissolvedaqueous concentration (lg l�1), KF,BC is the concentration-indepen-dent Freundlich BC–water distribution ratio ((pg kg�1

BC )/(pg l�1)n)and nF,BC is the Freundlich exponent of BC sorption.

All additional sorption over AOC sorption is hereby assumed tobe caused by BC. Wood chips and oil are other materials that couldincrease sorption strength. However, wood chips were not expectedto contribute significantly to total sorption since wood shows muchweaker sorption than BC and even AOC. Measured K-values forwood lignin were 1.0–1.5 log-unit below AOC for PCDD/Fs; logKlig-

nin = 0.74logKOW � 0.04 (MacKay and Gschwend, 2000). Weatheredoil residues may show equally strong sorption as BC (Jonker et al.,2003). However, oil contents in the soils, <0.01–0.021 g kg�1, werelower than BC contents, 0.3–4 g kg�1 (Table 2).

A recent elaborate literature study (Cornelissen et al., 2005)showed that the median nF,BC of 21 measurements on several typesof BC (lacustrine sediment, soil, marine sediment, pure char, iso-lated char, kerogen char, diesel soot, oil soot (Cornelissen et al.,2005)) was 0.61 for phenanthrene (IQR 0.58–0.64). Since no litera-ture values on nF,BC of PCDD/F are available, this value was used forthe calculation of KF,BC from the sorption data (Table 3), under theassumption that nF,BC for various PCDD/Fs in the same BC is similarto the phenanthrene value. This way, median logKF,BC values forthe five soils in units of ((pg kg�1

BC )/(pg l�1)n) were derived to bein the order of 8.6–9.2 for the least hydrophobic congener(2,3,7,8-TCDD) to 10.5–11.6 for the most hydrophobic one (OCDF;Table 3).

To investigate the validity of the assumption that nF,BC forPCDD/Fs equals nF,BC for phenanthrene, the sensitivity of the de-rived KF,BC-values to variation in nF,BC was analyzed. Therefore KF,BC

was additionally calculated on the basis of nF,BC-values of 0.5 and0.8 (Table 3). The value of KF,BC obtained with nF,BC of 0.5 and 0.8,respectively, deviated maximally 0.2 log-unit from the value calcu-lated with nF,BC of 0.61 (Table 3). Additional uncertainty in KF,BC-values (0.42 log-unit or a factor of 2.5) is also caused by the bandwidth of Seth’s KOW–KAOC LFER (Seth et al., 1999).

The values of KF,BC were rather constant (within a factor of 5–10for each individual PCDD/F in five soil samples) despite the strongvariation in BC contents in the five soils (BC 0.04–0.4%. This showsthat the BC in the five soils was probably similar. However, thisconclusion should be interpreted with caution since similar sam-ples were studied. BC sorption strength to different types of BChas been shown to vary widely, by several orders of magnitude(Jonker and Koelmans, 2001; Cornelissen et al., 2005; Hawthorneet al., in press).

LogKBC at the in situ PCDD/F concentrations were calculatedwith

CS ¼ fAOCKAOCCW þ fBCKBCCW ð2Þ

logKBC ranged from 9.6 to 12.0 for the various PCDD/Fs and wasapproximately 2 orders of magnitude above the observed logKTOC

values.

1460 G. Cornelissen et al. / Chemosphere 72 (2008) 1455–1461

BC dominated total sorption, with 60–99% of PCDD/Fs sorbed toBC (Table S5). The lowest percentages were observed for soil V(median for 17 PCDD/Fs 90% vs. 94–99.7% for the other four soils;Table S5) containing the lowest BC/TOC ratio (BC 0.6% of TOC, Table2). The dominance of BC to total sorption is in agreement withobservations that the spatial distribution of PCDD/Fs in suspendedparticles showed stronger correlations with BC than with bulk TOCcontents (Persson et al., 2002).

However, the KBC data can be compared to those in two previ-ous studies, one on native non-2,3,7,8-substituted congeners andone on pure diesel soot. One native di-CDD (2,7-DCDD) and onetri-CDD (2,3,7-TCDD) were studied (Lohmann et al., 2005), presentin New York, NY, Harbor sediment at concentrations similar to thepresently studied soil I (10 pg l�1 for 2,7-DCDD and 0.5 pg l�1 for237-TCDD, respectively). logKBC-values were 7.4 and 8.3 for 2,7-DCDD and 2,3,7-TCDD (Lohmann et al., 2005), respectively, aroundtwo orders of magnitude above their respective logKOW values of5.7 and 6.3 (KOW from Sacan et al. (2005)). The observed valuesfor KBC in soil I were 2–3 orders above logKOW (Table 3 and TableS3) and slightly higher than Lohmann’s values (2005). The pres-ence of a cocktail of other organic compounds in Lohmann’s NYsediment (mainly PAHs from diffuse sources, at around100 mg kg�1 dw (Lohmann et al., 2005)) can probably explain thisdifference between soil I and NY sediment. In contrast, PCDD/Fsfrom one point source were the main contaminants in the soilsand PAH contents (<1 mg kg�1; Table 2) were a factor of 100 lowert‘han in NY sediment. Competition between PAHs and PCDD/Fs forBC sorption sites could explain that the KBC values for NY sedimentwere relatively lower than the currently observed values.

The present KF,BC values were on average 106.8 and 106.6

(lg kg�1BC )/(lg l�1)n for 2,3,7,8-TCDF for 2,3,7,8-TCDD, respectively

(Table 3). Similar logKF,BC values of 107.2 for 1368-TCDF and 107.1

for 1368-TCDD (0.4–0.5 log-unit) were reported earlier (Bärringet al., 2002). The measurements in Bärring et al. (2002) were doneat CW,free = 0.03 and 0.11 lg l�1, respectively, and recalculated to1 lg l�1, assuming nF,BC = 0.61. This subtle difference in KF,BC be-tween the present study and Bärring et al.’s one (2002) can proba-bly be explained by competition effects as well, since the currentlymeasured values apply to BC in the presence of a cocktail of PCDD/Fs and biomolecules, whereas competition for BC sorption sites isabsent in Bärring’s laboratory systems containing one compoundsorbed to pure diesel soot (Bärring et al., 2002).

3.5. PCDD vs. PCDF sorption

Higher KBC values and higher KTOC values for PCDFs than forPCDDs of similar KOW have been reported (Bärring et al., 2002;Persson et al., 2002). Also in the present study PCDFs exhibitedhigher KTOC compared to KAOC than PCDDs: KTOC was a factor of73 ± 27 above KAOC for PCDFs (n = 10) and a factor of 20 ± 13(n = 7) for PCDDs (significant difference at a 95% confidence level;two-sided t-test). Also KF,BC was a factor of 513 ± 434 above KOW forPCDFs and a factor of 94 ± 53 for PCDDs (no significant differenceat the 95% confidence level). 98.5 ± 0.4% of PCDFs was sorbed toBC vs. 93.5 ± 2.7% for PCDDs (Table S5; difference significant at95% confidence level).

Differences in total soil contents between PCDFs and PCDDscannot be the cause of this phenomenon. Since sorption to BC isstrongly nonlinear, the importance of BC decreases with increasingPCDD/F concentration. However, the total soil contents of the tenmeasured PCDFs were 3–77 times higher than those of the sevenmeasured PCDDs (median factor 25, calculated from Table S2).

Thus, the present results are in accordance with the earlierobservation that PCDFs, as compared to PCDDs with similar KOW,exhibited higher laboratory-derived KBC values for spiked PCDD/Fs (Bärring et al., 2002) as well as higher KTOC values for native,

in situ PCDD/Fs (Persson et al., 2002). Ab initio molecular modelingof the energy-minimized torsional angle suggests that this angle issignificantly less than 180 �C for PCDDs, but not for PCDFs (Bärringet al., 2002). Thus attaining a planar configuration when sorbing toBC is probably less thermodynamically favorable for PCDDs thanfor PCDFs.

3.6. DOC–water distribution

DOC contents (operationally defined as passing a 1-lm filter) ofthe soil eluates were 7.8–18.5 mg l�1 (Table S9). Other methods forOC–water separation (e.g. flocculation, centrifugation) might haveresulted in (slightly) different values for the total aqueous concen-tration CW,total and DOC–water distribution ratios KDOC. From thedifferences in CW,total (freely dissolved plus DOC-bound) and CW,free

(only freely dissolved), KDOC was determined with

KDOC ¼CDOC

CW;free¼ 1

fDOC

CW;total

CW;free� 1

� �ð3Þ

where fDOC is the DOC fraction in water (kg DOC l�1 water) and CDOC

is the PCDD/F concentration in the DOC (ng PCDD/F per kg DOC).KDOC (Fig. 2b; Table S10) was clearly lower than KTOC (Fig. 2a),

and in the same order of magnitude as KAOC. These values are lar-gely in agreement with a literature survey (Krop et al., 2001) of 900KDOC values for many different types of DOC, where KDOC for PCDD/Fs with logKOW > 5 was found to be 0.5 to 2 orders of magnitudebelow KOW (and thus 0–1.5 orders of magnitude below KAOC). Re-cently observed KDOC values for PCDD/Fs in soils were as much as1–2.5 log-units below KOW (Frankki et al., 2006). Even thoughDOC contents were modest (7.8–18.5 mg l�1), almost all PCDD/Fin the liquid phase was bound to the DOC, especially for soil IV(95–99%; Table S9). Another observation from Fig. 2b is that KDOC

increased only slightly with increasing KOW (slope 0.27 ± 0.06). Pre-vious KDOC–KOW relationships have, without exception, shownslopes below one (Krop et al., 2001; Frankki et al., 2006). For exam-ple, a slope around 0.3 was recently observed for 17 PCDD/Fs in soil(calculated from Frankki et al., 2006).

3.7. Ecotoxicological and groundwater transport risks

On the basis of total soil contents, three of the five soils showpotential risks since CS is over Swedish generic guideline valuefor less sensitive land use of 0.25 lg WHO-TEQ kg�1 (Table 2,where soils I, II and III exceed the criterion by factors of 121, 5and 3, respectively). However, total soil contents do not take intoaccount limited bioavailability due to strong sorption to BC (Corne-lissen et al., 2005; Ehlers and Luthy, 2003). The presence of BC hasbeen shown to reduce bioaccumulation (Moermond et al., 2005)and toxicity (Knauer et al., 2007). Therefore actual ecotoxicologicalrisk assessment is better based on CW,free (Ehlers and Luthy, 2003).

No Swedish risk limit for serious groundwater contaminationexists. The Dutch value is 1 pg l�1 WHO-TEQ. CW,free in the pore-water (Table S3) exceeded this risk limit in soil I (by a factor 6)but not in soils II and III (factors 11 and 16 below risk limit,respectively).

Transport via groundwater occurs via both the freely dissolvedand the DOC-bound phase. The total dissolved PCDD/F concentra-tions in the leachates from all five soils exceeded the criterion forserious groundwater contamination (1 pg l�1 WHO-TEQ total dis-solved in water), by factors of 1.5–216 (Table S8). One should keepin mind that laboratory-measured leaching is here compared to agroundwater field criterion. The tentative conclusion can be drawnthat serious risk for spreading of the contamination via the ground-water probably exists at all five sampling sites, due to DOC-facili-tated transport.

G. Cornelissen et al. / Chemosphere 72 (2008) 1455–1461 1461

Acknowledgements

This study was financed by the County administration of Väst-ernorrland. The authors would like to thank Kerstin Grunder,Yngve Zebuhr, Anna Kitti and Maria Hjelt for help with the exper-imental work. We also would like to thank Viktor Lundgren at theCounty administration of Västernorland and the Municipality ofKramfors for providing us with samples from Marieberg.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.chemosphere.2008.05.010.

References

Bärring, H., Bucheli, T.D., Broman, D., Gustasson, Ö., 2002. Soot-water distributioncoefficients for polychlorinated dibenzo-p-dioxins, polychlorinateddibenzofurans and polybrominated diphenylethers determined with the sootcosolvency-column method. Chemosphere 49, 515–523.

Cornelissen, G., Gustafsson, Ö., 2004. Sorption of phenanthrene to environmentalblack carbon in sediment with and without organic matter and native sorbates.Environ. Sci. Technol. 37, 148–155.

Cornelissen, G., Gustafsson, Ö., Bucheli, T.D., Jonker, M.T.O., Koelmans, A.A., VanNoort, P.C.M., 2005. Extensive sorption of organic compounds to black carbon,coal, and kerogen in sediments and soils: mechanisms and consequences fordistribution, bioaccumulation, and biodegradation. Environ. Sci. Technol. 39,6881–6895.

Cornelissen, G., Pettersen, A., Broman, D., Mayer, P., Breedveld, G.D., 2008. Fieldtesting of equilibrium passive samplers to determine freely dissolved nativePAH concentrations. Environ. Toxicol. Chem. 27, 499–508.

Danielsson, C., Wiberg, K., Korytar, P., Bergek, S., Brinkman, U.A.T., Haglund, P., 2005.Trace analysis of polychlorinated dibenzo-p-dioxins, dibenzofurans and WHOpolychlorinated biphenyls in food using comprehensive two-dimensional gaschromatography with electron-capture detection. J. Chromatogr. A 1086, 61–70.

Ehlers, L.J., Luthy, R.G., 2003. Contaminant bioavailability in soil and sediment.Environ. Sci. Technol., 295A.

Elmquist, M., Cornelissen, G., Kukulska, Z., Gustafsson, Ö., 2006. Distinct oxidativestabilities of char versus soot black carbon: implications for quantification andenvironmental recalcitrance. Global Biogeochem. Cycles 20, 1–11.

Frankki, S., Persson, Y., Tysklind, M., Skyllberg, U., 2006. Partitioning of CPs, PCDEs,and PCDD/Fs between particulate and experimentally enhanced dissolvednatural organic matter in a contaminated soil. Environ. Sci. Technol. 40, 6668–6673.

Gustafsson, Ö., Haghseta, K., Chan, F., McFarlane, A., Gschwend, P.M., 1997.Quantification of the dilute sedimentary soot phase: implications for PAHspeciation and bioavailability. Environ. Sci. Technol. 31, 203–210.

Hawthorne, S.B., Grabanski, C.B., Miller, D.J., in press. Measured partitioningcoefficients for parent and alkyl PAHs in 114 historically contaminatedsediments: part 2. Testing the KocKbc two-carbon type model. Environ.Toxicol. Chem., doi: 10.1897/07-087.

Huang, W., Peng, P., Yu, Z., Fu, J., 2003. Effects of organic matter heterogeneity onsorption and desorption of organic contaminants by soils and sediments. Appl.Geochem. 18, 955.

Jensen, J., 1996. Chlorophenols in the terrestrial environment. Rev. Environ. Contam.Toxicol. 146, 25–51.

Jonker, M.T.O., Koelmans, A.A., 2001. Polyoxymethylene solid phase extraction as apartitioning method for hydrophobic organic chemicals in sediment and soot.Environ. Sci. Technol. 35, 3742–3749.

Jonker, M.T.O., Sinke, A.J.C., Brils, J.M., Koelmand, A.A., 2003. Sorption of polycyclicaromatic hydrocarbons to oil contaminated sediment: unresolved complex?Environ. Sci. Technol. 37, 5197.

Knauer, K., Sobek, A., Bucheli, T.D., 2007. Reduced toxicity of diuron to thefreshwater green alga Pseudokirchneriella subcapitata in the presence of blackcarbon. Aquat. Toxicol. 83, 143–148.

Krop, H.B., Van Noort, P.C.M., Govers, H.A.J., 2001. Determination and theoreticalaspects of the equilibrium between dissolved organic matter and hydrophobicorganic micropollutants in water (KDOC). Rev. Environ. Contam. Toxicol. 169, 1–122.

Lohmann, R., MacFarlane, J.K., Gschwend, P.M., 2005. Importance of Black Carbon tosorption of native PAHs, PCBs, and PCDDs in Boston and New York Harborsediments. Environ. Sci. Technol. 39, 141–148.

MacKay, A.A., Gschwend, P.M., 2000. Sorption of monoaromatic hydrocarbons towood. Environ. Sci. Technol. 34, 839–845.

Moermond, C.T.A., Zwolsman, J.J.G., Koelmans, A.A., 2005. Black carbon andecological factors affect in situ biota to sediment accumulation factors forhydrophobic organic compounds in flood plain lakes. Environ. Sci. Technol. 39,3101.

Persson, N.J., Gustafsson, Ö., Bucheli, T.D., Ishaq, R., Naes, K., Broman, D., 2002. Soot-carbon influenced distribution of PCDD/Fs in the marine environment of theGrenlandsfjords, Norway. Environ. Sci. Technol. 36, 4968–4974.

Persson, Y., Lundstedt, S., Öberg, L., Tysklind, M., 2007. Levels of chlorinatedcompounds (CPs, PCPPs, PCDEs, PCDFs, PCDDs) in soils at contaminated sawmillsites in Sweden. Chemosphere 66, 234–242.

Sacan, M.T., Özkul, M., Erdem, S.S., 2005. Physico-chemical properties of PCDD/Fsand phthalate esters. SAR QSAR Environ. Res. 16, 443–459.

Seth, R., Mackay, D., Muncke, J., 1999. Estimating the organic carbon partitioncoefficient and its variability for hydrophobic chemicals. Environ. Sci. Technol.33, 2390–2394.

Smedes, F., 2007. Monitoring of chlorinated biphenyls and polycyclic aromatichydrocarbons by passive sampling in concert with deployed mussels. In:Greenwood, R., Mills, G.A., Vrana, B. (Eds.), Passive Sampling Techniques inEnvironmental Monitoring, Comprehensive Analytical Chemistry, Vol. 48.Elsevier Science & Technology Books, Oxford, UK (Chapter 19).

Swedish Environmental Protection Agency, 1997. Development of generic guidelinevalues – models and data used for the development of generic guideline valuesfor contaminated soils in Sweden. Report 4639.