The fate and persistence of polychlorinated biphenyls in soil Stephen Ayris and Stuart Harrad* Environmental Health, Institute of Public & Environmental Health, University of Birmingham, Birmingham, UK B15 2TT . Tel:+44 (0) 121 414 7298; Fax:+44 (0) 121 414 3078; E-mail: S.J.Harrad@bham.ac.uk Received 15th April 1999, Accepted 25th May 1999 The fate and persistence of PCBs 28, 52, 101, 138, and 180 artificially introduced into three soils was studied under a variety of field conditions for up to 415 d following initial contamination.A relationship was detected between ln Koa (octanol/air partition coeYcient) and the experimentally observed first-order loss rate constant that was statistically significant at at least the 90% level in all but one instance. In nearly all experiments, PCB persistence was greater in soils of higher organic carbon content.Soil temperature and moisture content were also indicated as important influences on persistence. Significantly longer half-lives were observed in a soil in which initial PCB contamination had occurred ca. 1 year previously. A mass balance showed the most likely mechanism of loss to be volatilisation. Losses attributable to aerobic biodegradation could not be ruled out, but those due to leaching, uptake by biota, and soil erosion were demonstrated to be negligible.First-order rate constants (kv) were determined for volatilisation of the same congeners from soil under a variety of controlled laboratory scenarios. Multiple linear regression analysis (MLRA) showed the most important influences on kv to be ln Koa (adjusted for soil temperature) and soil organic carbon content.Limited evidence was observed for a relationship between kv and soil moisture content, but not water flux. When tested against field measurements, the MLRA-derived relationship between kv and independent variables predicted to within a factor of 2.5, the persistence of PCBs 28, 52, and 101.However, it did not account for the influence of the age of contaminant association with the soil, soil moisture content or water flux, and failed to function for soils of high organic content, or where ln Koa exceeds ca. 23. The laboratory-based experiments directly measured volatil- Introduction isation of PCBs from soil. Their principal aim was to evaluate In recognition of the continued potential health threat posed the relationship between the first-order volatilisation rate by polychlorinated biphenyls (PCBs),1 the UK government constant (kv) and the octanol air partition coeYcient (Koa) recently announced an action plan to phase out and destroy adjusted for soil temperature, soil organic carbon content remaining stocks.2 While this is to be welcomed, there remains ( foc), soil moisture content (MC), and the water evaporation a considerable burden of PCBs associated with topsoil,3 which flux (WF).These data were subjected to multiple linear represents an important reservoir of material that is essentially regression analysis (MLRA), to provide an equation to predict beyond human control. One route via which PCBs enter soils the persistence of PCBs in soil.The accuracy of this equation is by accidental spills of PCB-contaminated fluids associated was tested against our field measurements. with transformers and large capacitors. If the risk arising from such incidents is to be understood and eVectively managed, Experimental methods the rate of decline in soil contamination must be accurately predicted, along with the relative significance of the diVerent Field experiments fate mechanisms involved, i.e., volatilisation, biodegradation, General.For each experiment, 30 cm diameter and 12 cm leaching, uptake by biota, and physical mass transport. depth containers were filled to 5 cm depth (between ca. 1.25 Furthermore, volatilisation from soil has been identified as the and 2.5 kg depending on soil density) with soil. This layer was most important single source of PCBs to the UK atmosphere.3 covered with a further 5 cm of the same soil artificially enriched It is therefore important that our understanding of the kinetics with a known quantity of PCBs 28, 52, 101, 138, and 180.of soil air exchange of PCBs is improved, if the contribution This was achieved by applying a mixture of these congeners of such volatilisation to the contemporary emission inventory (dissolved in 50 ml acetone) to the soil using a spray gun is to be accurately determined, and the likely influence of pesticide applicator with thorough mixing.To permit volatilis- current eVorts to reduce environmental contamination by ation of acetone and more intimate PCB binding to soil, 72 h reducing primary emissions is to be assessed.elapsed before the first soil samples were taken and the This paper reports the results of experiments monitoring experiments begun. One hundred 1 mm diameter holes were the fate of PCB congeners 28, 52, 101, 138 and 180 in soils drilled into the base of each pot to facilitate leachate collection under field and controlled laboratory conditions.The aims of these experiments were: (i) to assess the relative significance via a plastic funnel draining directly into an amber glass bottle. of potential mechanisms influencing PCB fate in soil, viz, Leachate samples were collected on an approximately volatilisation, leaching, soil erosion, uptake by biota, and monthly basis.degradation; (ii) to evaluate the influence of edaphic and Experimental conditions. These are summarised in Table 1. pollutant properties on PCB persistence in soil; and (iii) to Soil A was a sandy loam, soil B a clayey loam, while soil C determine how closely PCB loss from soil approximates to a first-order process over the time scales studied. was a commercially available peat-based compost.J. Environ. Monit., 1999, 1, 395–401 395Table 1 Conditions for field experiments Experiment/ Mean Mean moisture Duration (no. of soila temperature/°C foc/g g-1 content/g g-1 measurements) 1/A 10 0.0113 0.363 13/5/96–5/7/97 (24) 1/B 10 0.0213 0.353 13/5/96–5/7/97 (24) 1/C 10 0.207 0.736 13/5/96–5/7/97 (24) 2/A 3 0.0113 0.425 11/11/96–5/2/97 (12) 2/B 3 0.0213 0.403 11/11/96–5/2/97 (12) 2/C 3 0.207 0.87 11/11/96–5/2/97 (12) 3/A 14 0.0113 0.19 5/5/97–27/7/97 (9) 3/WA 14 0.0113 0.271 5/5/97–27/7/97 (9) aWA-‘weathered’ soil A.Sampling procedures Topsoil monitoring. Soil sampling was initially conducted with bidaily frequency but as the experiment proceeded this frequency was reduced. To ensure only the contaminated layer was sampled, the total depth to the base of each pot was measured and samples taken to this depth less 5 cm.Each sampling event comprised five cores taken over the soil surface using a steel tube of 1 cm diameter. These sub-samples were Fig. 1 Schematic diagram of the volatilisation chamber. homogenised and a 1.5 g sample taken for analysis, with unused sample being returned to the pot. using a Hi-Vol sampler, with volatilised PCBs retained on a Mass balance measurements.To provide an additional single polyurethane foam (PUF) plug. The sampling flow rate measurement of PCB persistence, a mass balance was conduc- and duration were identical for all samples (0.75 m3 min-1 ted for each soil in experiment 1. In summary, the mass and 24 h, respectively). balance for each pot is described by eqn.(1): To ensure that PCBs detected in the sampling PUF were due to volatilisation from the soil sample alone, the air entry Mu=M0+MD-ML-MS-Mt=415 (1) for the chamber was fitted with two PUF plugs to remove in which Mu is the ‘unaccounted for’ congener mass loss; M0 PCBs from incoming air. The PCB removal eYciency, is the congener soil burden at time zero (the start of the expressed as the percentage of incoming PCB removed by the experiment); MD is the increase in congener mass due to bulk ‘pre-chamber’ cleaning PUFs, was measured by conducting a atmospheric deposition; ML is the congener mass loss due to ‘chamber blank’, i.e., conducting an experiment but in the leaching; MS is the congener mass loss from sampling; and absence of soil, after each five experiments.Mean ‘chamber Mt=415 is the congener soil burden at the end of the experi- blank’ values for each congener were subtracted from all ment (415 d). experimental data. To facilitate mass balance calculation, at the end of experi- Prior to use, the chamber was thoroughly cleaned and dried. ment 1,Mt=415 was calculated by determining the PCB content The soil heater was installed on the base of the soil tray and of a single 5 cm diameter core to the base of the pot, and the experimental soil loaded into the soil tray to a depth of multiplying this by the total soil mass.M0 was determined by 7 cm. The platinum film detectors and the thermocouple probe measuring the soil bulk density, volume, and the PCB concen- were laid on the surface of this soil layer before being evenly tration in both the contaminated and non-contaminated soil covered to a depth of 1 cm with soil artificially enriched with layers, ML was estimated by summing the leached masses a known quantity of PCB congeners 28, 52, 101, 138, and 180.recorded in each sample over the duration of the experiment, MS was taken to be the total mass of PCB removed over all Measurement of edaphic properties the soil sampling events, and MD was calculated as the product Soil temperature was measured at a 1 cm depth via four evenly of the exposed soil surface area and the bulk atmospheric spaced sheathed platinum film detectors connected to a cali- deposition fluxes measured at the sampling site during the brated digital voltmeter.The temperature over the duration experiment.Our atmospheric deposition sampling method is of each experiment was maintained within ±2 °C. described elsewhere,4,5 but in summary, bulk (i.e., wet and PCB concentrations in the top 1 cm soil layer were deter- dry, vapour and particulate) deposition was collected on a mined at the beginning and end of each experiment, by monthly basis using an Al foil-lined inverted frisbee connected analysing a representative sample of approximately 1 g.This to an amber glass vessel. was obtained by taking ten sub-samples of 1 g each in an X pattern across the soil surface. After homogenization and Measurement of edaphic properties removal of the sample for analysis, the remaining soil was Mean air temperatures were taken from a weather station spread evenly over the surface layer.located near the experimental site. The soil foc was measured The soil foc was measured by elemental analysis, while the by elemental analysis, and the soil MC, determined by drying soil MC was taken as the mean of that determined for to constant mass, taken as the mean of that determined for representative soil samples taken before and after each experieach sampling event.ment. The water flux (WF) was calculated using eqns. (2) and (3). Laboratory experiments WF=ydkw (2) Volatilisation chamber. PCB volatilisation fluxes were studied kw=( ln W0-ln Wt)/t (3) using a custom-built chamber, illustrated in Fig. 1. Essentially, air was pulled via suction over the surface of a soil sample, where yd is the vertical diVusion distance assumed to be 396 J.Environ. Monit., 1999, 1, 395–4010.005 m (half the distance of the enriched soil layer); kw is the Results and discussion water flux rate constant in d-1; W0 and Wt are the masses of Field experiments water present (kg) in the soil at the beginning and end of each experiment, respectively, calculated as the product of MC and Applicability of first-order kinetics.PCB loss from a number the total mass of the spiked soil layer; and t is the duration of of Arctic soils was reported to follow first-order kinetics for the experiment (d). at least the first 5 years after initial contamination.6 However, Twenty-two laboratory experiments were conducted. The it has been suggested that first-order kinetics may not relevant conditions of each are detailed in Table 2.After each adequately describe long term PCB loss.7 To assess whether experiment, the PCB content of the sampling PUF was deter- first-order kinetics were appropriate to describe PCB loss from mined and converted to kv values using eqns. (4) and (5): soil over the time scales studied in this project, we plotted the concentration of each congener n in soil at time=t ( ln Ctn) kv=( ln M0-ln Mt)/t (4) against time (t) for each experiment.Table 3 summarises the kinetic information obtained from such plots for each congener Mt=M0-Mv (5) and experiment. Out of 40 such plots, linear correlation at the 99.9% confidence level was observed in 31 cases, and in only where M0 and Mt are the congener masses (mg) present in the one instance was the confidence level below 90%.These data soil at the beginning and end of each experiment, respectively; clearly show that PCB loss from soil over the relatively short and Mv is the congener mass (mg) present on the sampling duration of these experiments is a first-order process, with PUF (corrected for chamber blank). apparent deviations most likely due to a combination of soil inhomogeneity, particularly likely for soil C, and an inadequately low ratio of experiment duration to the rate of PCB Sample purification and analysis.PCB analyses were conduc- loss from a given soil. ted using well-validated GC-MS-based procedures.4,5 Soil samples were homogenised with Na2SO4 and Soxhlet extracted Comparison of half-life measurement methods.Table 4 sumwith hexane5acetone (40560 v/v), bulk deposition samples marises the mass balance data obtained. Half-lives obtained were filtered through a glass fibre filter (GFF) and PUF in by this method are consistently longer than those obtained via series before soxhlet extraction with dichloromethane, leachate the ‘topsoil monitoring’ approach (Table 3). We believe the samples were passed through a PUF prior to Soxhlet extraction discrepancies are predominantly due to the influence of soil with dichloromethane, and PUF samples of volatilised PCB compaction during experiment 1, which when combined with were Soxhlet extracted with dichloromethane.Following con- the sampling method adopted for the ‘topsoil monitoring’ centration, extracts were subjected to acid washing, Florisil approach means that as the experiment progressed, PCB chromatography, lipid removal via solvent exchange between concentrations were measured in an increasingly narrower DMSO and hexane, and concentration prior to GC-MS analy- depth of soil from which PCB loss via volatilisation would be sis on a Fisons’ MD-800 instrument, fitted with an HP-5 Trace more facile.It is therefore considered that the half-lives Analysis column (60 m×0.25 mm i.d.×0.25 mm film thick- reported in Table 3 are underestimates where soil compaction ness). One ml of sample extract was injected in the splitless occurs. mode (injector and transfer line temperatures both 300 °C), and the oven program was 140 °C for 2 min; 5 °Cmin-1 to Comparison with previously reported half-lives. The variation 200 °C; and 2 °Cmin-1 to 280 °C.Ten ions were monitored in in half-lives that we observed for diVerent congeners, soils, EI selected ion monitoring (SIM) mode (ionisation voltage= and climatic conditions, is consistent with the range reported 70 eV) for the analysis of the trichlorinated biphenyls through by other studies. Individual PCBs in artificially contaminated the heptachlorinated biphenyls, two ions for each homolog Arctic soils displayed half-lives of 0.7–2.3 years,6 half-lives in group.These ions were 255.95, 257.95, 289.95, 291.95, 325.90, sewage sludge-amended UK agricultural soils ranged from <1 327.90, 359.90, 361.90, 393.85, and 395.85. to 8.5 years for PCBs 18, 28 and SPCB,7 those for di-, tri-, and tetrachlorobiphenyls were 0.5, 0.75, and 3.2 years, respectively, in field studies of PCB loss from sewage sludge-amended Quality control and quality assurance.Sampling eYciency soil,8 while data extrapolated from temporal trends in PCB standards (SESs, PCB congeners 19 and 147) were added to levels in UK soils indicated a mean half-life for PCBs in 5 the bulk deposition sampler, the leachate capture vessel, and diVerent soils of ca. 5 years.9 It is therefore interesting that the sampling PUF used in the laboratory experiments before the average half-life for congeners 28, 52, 101, 138, and 180 each measurement. SES recoveries (which reflect analyte losses in ‘weathered’ soil A in experiment 3, which we consider the due to both sampling and analysis) ranged between 45 and most applicable to ‘typical’ UK conditions, is ca. 5.5 years. It 75% for the field experiments and between 55 and 86% for the is important to note, however, that our experiments measured laboratory experiments. Sample data were not corrected for PCB loss in the absence of vegetative cover, which appears to SES recoveries. Recoveries of internal standards added to retard PCB loss from soil,6 and thus are likely to underestimate check analyte losses during sample analysis alone (PCB con- PCB persistence in many environmental scenarios.geners 34, 62, 119, 131, and 173) ranged between 47 and 90% for all samples. To measure the combined uncertainty due to Congener-specific diVerences in persistence. Table 3 reveals a marked increase in persistence with increasing chlorine soil sampling and analysis combined, five replicate soil samples were analysed for each experiment.The coeYcients of variation number. The octanol air partition coeYcient (Koa), which is negatively correlated with increasing chlorination within a for individual congeners ranged between 4 and 16% (arithmetic mean=6.8%). compound class, has been suggested as the most suitable physico-chemical parameter for predicting semivolatile organic compound (SOC) partitioning between air and organic environmental media, such as soil.10 To evaluate the relation- Multiple linear regression analysis (MLRA).MLRA was conducted using SPSS for Windows version 7 in a ‘retro- ship between Koa and PCB persistence in our experiments, we derived Koa values for each congener at the mean temperature analytical’ manner.Inclusion of independent variables in the final regression equation depended on significant T not of each experiment using appropriate algorithms,11,12 and plotted ln Koa against k. The relationship between ln Koa and exceeding 0.1. J. Environ. Monit., 1999, 1, 395–401 397Table 2 Conditions and volatilization rate constants (kv d-1) for laboratory experiments Mean soil concentration of congener/mg kg-1 kv for congenerd Experiment foc a MCb WFc T /K 28 52 101 138 180 28 52 101 138 180 1 0.011 0.40 0.24 303 46.8 51.9 37.3 42.1 48.5 0.0094 0.0073 0.0029 0.0031 0.0003 2 0.011 0.40 0.26 303 49.1 47.3 36.1 41.5 55.0 0.0113 0.0085 0.0037 0.0012 0.0003 3 0.207 0.63 0.02 303 116.5 134.5 126.5 113.7 131.8 0.0037 0.0021 0.0008 0.0002 0.0001 4 0.021 0.26 0.07 303 30.0 30.9 34.0 31.7 30.9 0.0145 0.0111 0.0046 0.0023 0.0005 5 0.011 0.33 0.16 285 44.6 48.8 41.0 44.8 48.7 0.0058 0.0043 0.0018 0.0003 0.0001 6 0.011 0.40 0.06 286 30.7 32.9 31.5 29.2 31.9 0.0082 0.0061 0.0025 0.0004 0.0002 7 0.011 0.25 0.29 278 28.2 25.0 23.1 27.0 23.6 0.0034 0.0009 0.0013 0.0004 0.0004 8 0.011 0.18 0.07 293 32.8 31.3 27.5 32.1 29.7 0.0035 0.0054 0.0021 0.0006 0.0003 9 0.011 0.36 0.41 286 20.0 24.1 22.3 22.7 23.3 0.0098 0.0076 0.0030 0.0004 0.0001 10 0.011 0.19 0.16 286 16.5 21.1 19.6 21.7 20.3 0.0033 0.0029 0.0010 nc nc 11 0.011 0.20 0.25 286 16.9 19.8 18.3 18.0 19.3 0.0054 0.0049 0.0016 0.0002 nc 12 0.011 0.18 0.39 286 16.7 18.1 17.1 18.2 18.5 0.0027 0.0027 0.0008 nc nc 13 0.011 0.12 0.28 286 14.6 15.9 14.1 9.2 14.9 0.0009 0.0010 0.0008 0.0001 nc 14 0.011 0.34 0.35 286 61.7 63.3 59.9 58.8 58.4 0.0051 0.0042 0.0023 0.0005 nc 15 0.011 0.29 0.20 287 73.7 74.7 70.3 69.5 69.8 0.0044 0.0039 0.0024 0.0006 nc 16 0.011 0.27 0.14 286 79.4 79.5 78.5 77.1 77.8 0.0042 0.0039 0.0023 0.0006 nc 17 0.044 0.33 0.43 303 9.6 7.7 10.0 11.3 14.0 0.0146 0.0109 0.0044 0.0007 0.0003 18 0.017 0.08 0.26 303 8.0 7.6 7.5 9.1 7.7 0.0144 0.0101 0.0036 0.0014 0.0012 19 0.021 0.11 0.29 303 9.3 8.9 9.1 11.9 8.8 0.0271 0.0195 0.0102 0.0016 0.0013 20 0.207 0.87 0.03 303 44.0 43.2 43.5 37.7 39.0 0.0016 0.0014 0.0007 0.0002 0.0001 21 0.011 0.31 0.17 303 18.3 21.5 20.3 17.4 12.4 0.0178 0.0128 0.0051 0.0029 0.0015 22 0.011 0.27 0.18 303 13.2 14.6 14.4 12.6 12.7 0.0217 0.0162 0.0061 0.0035 0.0014 afoc is organic carbon content of soil (g g-1).bMC is moisture content of soil (g g-1). cWF is water evaporation flux (cm d-1). dnc is not calculable. 398 J. Environ. Monit., 1999, 1, 395–401Table 3 Summary of kinetic data for field experiments Half-life (rangea)/d Experiment soilb Congener 28 Congener 52 Congener 101 Congener 138 Congener 180c 1/A 94 (90–98) 100 (95–107) 126 (120–133) 173 (160–188) 217 (204–230) 1/B 91 (88–95) 105 (99–112) 139 (133–144) 198 (178–223) 210 (198–224) 1/C 178 (157–204) 187 (169–211) 257 (210–329) 239 (198–303) 433 (347–578) 2/A 148 (122–187) 193 (131–363) 231 (165–385) 239 (173–386) 224 (161–367) 2/B 161 (147–179) 217 (187–258) 301 (247–386) 347 (257–533) 85 (301–535) 2/C 267 (217–346) 239 (187–332) 495 (315–1151) 630 (435–1146) 866 (385–nc) 3/A 122 (117–127) 154 (144–166) 217 (192–249) 315 (289–346) 408 (364–463) 3/WA 193 (178–209) 267 (247–290) 462 (432–497) 495 (434–576) 8664 (5776–17239) aRange calculated using percent error in k.bWA=‘weathered’ soil A. cnc is not calculable. Table 4 Mass balance data for field experiment 1 Soil Congener M0/mg MD/mg ML/mg MS/mg Mt=415/mg MU/mg Half-life (rangea)/d A 28 65.1 0.29 0.13 0.50 6.1 58.7 121 (112–133) A 52 70.6 0.37 0.10 0.61 7.8 62.5 131 (120–144) A 101 62.5 0.24 0.09 0.51 15.0 47.1 202 (175–239) A 138 63.7 0.17 0.07 0.57 29.1 34.1 367 (298–477) A 180 91.4 0.07 0.05 0.72 51.4 39.2 500 (402–661) B 28 74.2 0.29 0.02 0.75 5.9 67.8 114 (106–122) B 52 77.4 0.37 0.02 0.87 10.2 66.7 142 (134–151) B 101 79.5 0.24 0.03 1.00 20.7 58.1 214 (199–231) B 138 67.2 0.17 0.01 1.03 31.0 35.3 372 (308–469) B 180 99.8 0.07 0.01 1.33 53.9 44.7 467 (402–558) C 28 41.2 0.29 0.02 0.89 15.0 25.5 285 (223–395) C 52 50.3 0.37 0.02 1.12 24.2 25.3 393 (302–563) C 101 41.1 0.24 0.01 0.97 28.9 11.5 817 (469–3172) C 138 45.2 0.17 0.01 1.06 34.9 9.5 1112 (495–nc)b C 180 54.3 0.07 0.01 1.34 49.3 3.2 3325 (781–nc)b aRange calculated from measured combined sampling and analytical coeYcient of variation for each congener and each soil.bnc=not calculable. k for soil A over the entire duration of experiment 1 is significant at the 99.9% confidence level. Similar plots for other soils and experiments reveal linear relationships between ln Koa and k that, for all but one experiment, are significant at at least the 90% confidence level.Clearly, our field experiments indicate PCB persistence to be positively linearly correlated with ln Koa. Influence of soil organic carbon content. With very few exceptions, PCB persistence in the three soils studied in experiments 1 and 2 followed the order soil C&soil B>soil A. While there are insuYcient data to permit firm conclusions to be drawn, this observation is consistent with the hypothesis that SOC persistence in soils is positively correlated with foc.Fig. 2 PCB half-lives in soil A over the first 86 d of experiments 1 and 3. Influence of temperature and soil moisture content. Table 3 demonstrates that over the full durations of experiments 1 and 2 the half-lives of all congeners in all soils were significantly longer in experiment 2.As discussed above, there appears to used for the 2 experiments were not absolutely identical with be a positive correlation between PCB persistence and Koa. respect to their PCB sorption potential, the only other detected The latter is negatively correlated with temperature, and it is diVerence between the two experiments was that the mean soil thus unsurprising that PCB persistence is greater in experimoisture contents were 0.363 and 0.190 g g-1, respectively, ment 2 (mean temperature=3 °C) than experiment 1 (mean over the first 86 days of experiments 1 and 3.temperature=10 °C). However, there is evidence that a Furthermore, half-lives in soil A were comparable in experi- straightforward temperature-related seasonal eVect on PCB ments 2 and 3, despite the higher mean temperature observed persistence does not occur in all instances, and that edaphic in experiment 3 (14 °C compared with 3 °C).Again, mean soil properties other than temperature may play an important ro� le. moisture content was higher in experiment 2 (0.425 g g-1) To illustrate, Fig. 2 shows the half-lives of PCBs measured in than in experiment 3 (0.190 g g-1).Such eVects on persistence soil A over the first 86 days of both experiment 1 and 3. It is associated with increased soil moisture content, are consistent apparent that despite the mean temperatures over both periods with the concept that such increases enhance competition being similar (13 °C and 14 °C for experiments 1 and 3, between water and PCB molecules for sorptive sites in the respectively) PCBs were clearly more persistent in experiment 3.Aside of the possibility that the sub-samples of soil A soil matrix. J. Environ. Monit., 1999, 1, 395–401 399Influence of soil ‘weathering’. While the majority of our Extrapolation to field conditions. To test the field validity of eqn. (6), we used it to predict half-lives for individual con- experiments studied PCB fate in ‘freshly-enriched’ soils, experiment 3 compared PCB loss from two portions of soil A (one geners under the conditions prevailing in the field experiments.Table 5 compares these predicted values with those observed ‘freshly-enriched’, the other subjected to 1 year of ‘weathering’, i.e., exposure to environmental conditions following artificial in field experiments 1, 2, and 3.Clearly, eqn. (6) is only applicable under specific conditions; it functions poorly, if at enrichment in the same fashion as the other soils). Half-lives in the ‘weathered’ soil are clearly longer for all congeners all, for soils with a high foc, is inapplicable to both PCBs 138 and 180 under the conditions of our field experiments, and despite the fact that its mean moisture content was higher than that of the ‘freshly-contaminated’ soil.This is consistent cannot distinguish between PCB persistence in a freshlycontaminated soil and an identical ‘weathered’ soil. Despite with the suggestion that SOC persistence in soils is strongly dependent on the length of the SOC–soil interaction, i.e., the such limitations, eqn.(6) predicts the persistence of congeners 28, 52, and 101 to a reasonable degree of accuracy (mean extent of ‘weathering’, and that laboratory studies where PCBs are added as spikes underestimate PCB persistence.7 observed5predicted ratio=1.04; s=0.55; range=0.10–2.32). Evidently, although eqn. (6) may only be used under certain conditions, it is capable of predicting to within at least a factor Relative significance of potential loss mechanisms.Five possof 2.5, the persistence of PCBs 28, 52, and 101 in soils of ible mechanisms may account for PCB loss during our experitypical foc values at the annually averaged temperatures prevail- ments: leaching, soil erosion, uptake by biota, degradation, ing in the UK. There are several likely factors why this and volatilisation.Table 4 shows that PCB losses due to equation is not more accurate, the majority associated with leaching (ML) were minimal compared with the ‘unaccounted the influence of excluded variables. These are: MC, WS, and for’ mass loss (MU). A quantitative expression of this obserconsideration of the length of PCB association with the soil vation was obtained by calculating ML/MU for each congener matrix (‘weathering’).Each of these is likely to exert an and soil. Values ranged from 0.0002 to 0.0031, with an influence on PCB persistence under field conditions, and their arithmetic mean and s of 0.0013 and 0.00086, respectively. omission from eqn. (6) inevitably impairs its eYcacy. It should Similarly, we consider losses due to soil erosion to be minimal also be noted that predicted half-lives consider loss due to for two reasons: first, the magnitude of MU could only be volatilisation only.The correspondence between predicted and assigned to soil erosion if a significant proportion of the total observed half-lives therefore provide further indirect evidence soil mass was lost during the experiments, such loss was not that volatilisation was the major loss mechanism operating observed; secondly, soil erosion would aVect all congeners during the field experiments. equally, and thus significant losses via this mechanism are Given the mechanistic simplicity and potential wider generic inconsistent with the observed congener-specific losses.Uptake applicability of eqn. (6), it is instructive to compare its ability by biota is not considered significant in our experiments as to forecast PCB fate in soil with that of the more mechan- weeds were removed regularly and PCB translocation from istically detailed mathematical modelling approach developed soil to foliage is considered minimal, and also because fauna by Jury and coworkers to predict volatilisation fluxes of such as earthworms were either absent or included in the final surface applied chemicals from soils.14 As we had not measured ‘mass balance’ measurement.Although we did not measure all of the input parameters required to evaluate the Jury any PCB degradation products, aerobic degradation has been approach, direct evaluation of its predictive performance with shown to be significant under favorable conditions for lower respect to our data was not possible.Instead, we referred to chlorinated PCBs,7 with one study reporting that although no the work of Cousins et al.15 who assessed the ability of the degradation of Aroclor 1260 was observed, 4–7% of the PCB Jury approach to predict PCB volatilisation fluxes measured associated with a soil contaminated with 100 mg kg-1 of under controlled laboratory conditions for a number of sewage Aroclor 1242 was aerobically degraded by indigenous microsludge- amended soils.For surface layer sludge applications, flora over a 60 d period.13 Over the first 63 d of experiment 1, they measured volatilisation fluxes of PCBs 18, 33, 47, 52, 87, k for the loss of congener 28 from soil A was 0.0143.This and 101 over three consecutive time periods. Using the Jury equates to a loss of ca. 60% of the initial loading of PCB 28 in 63 d, and thus the estimated maximum contribution of aerobic degradation to the overall loss of congener 28 is 12%. Table 5 Observed field and predicted half-lives using equation 6a In addition to biodegradation, there may also have been losses Observed field due to photolysis, but we have been unable to measure the half-life Predicted significance of this pathway.By implication therefore, we Experiment Soila Congener (rangeb)/d half-life/d conclude that the principal PCB loss mechanism in our field experiments is volatilisation; a hypothesis supported by our 1 A 28 121 (112–133) 115 observed correlation between ln Koa and k. 1 A 52 131 (120–144) 132 1 A 101 202 (175–239) 467 1 B 28 114 (106–122) 115 Laboratory experiments 1 B 52 142 (134–151) 132 1 B 101 214 (199–231) 467 In seven experiments the volatilised mass of PCB 180 was 1 C 28 285 (223–395) 2963 below the average ‘chamber blank’ value, while in the remain- 2 A 28 148 (122–187) 188 der, volatilised mass was frequently close to the blank level. 2 A 52 193 (131–363) 210 2 B 28 161 (147–179) 204 We therefore eliminated data for this congener before con- 2 B 52 217 (187–258) 231 ducting a multiple linear regression analysis on this reduced 3 A 28 122 (117–127) 99 data set (Table 2). This analysis yielded eqn. (6) relating kv, 3 A 52 154 (144–166) 115 to foc and temperature-dependent ln Koa. 3 A 101 217 (192–249) 295 3 WA 28 193 (178–209) 99 kv=5.17×10-2-2.21×10-3 ln Koa-2.97×10-2 foc (6) 3 WA 52 267 (247–290) 115 3 WA 101 462 (432–497) 295 n.(6) was significant at the 99.99% confidence level. b aWA denotes ‘weathered’ soil A. bOnly those experiments for which values were derived for both independent variables, and predictions were calculable are shown. For all other experiments, revealed ln Koa (b=-0.79) to exert amore significant influence predicted half-lives were not calculable.on k, than foc (b=-0.31). 400 J. Environ. Monit., 1999, 1, 395–401model provided predicted fluxes that consistently underesti- ditions. Multiple linear regression analysis of these data revealed the most important independent variables to be ln mated measured values by a mean factor of 12.4 (s=9.99; range=3.4–40.7).Clearly, although we were unable to conduct Koa (adjusted for soil temperature), and to a lesser extent, foc. The multiple linear regression equation derived was tested direct comparison of the relative eYcacy of our empirically derived equations with that of the Jury model, when used against our field measurements of PCB persistence in soils, and found to perform well for PCBs 28, 52, and 101 under appropriately eqn.(6) appears to be a useful ‘screeninglevel’ tool. conditions consistent with those typically prevailing in the UK. However, caution is needed if the findings of this work Observations on the influence of specific parameters on kv are to be extrapolated to other situations. Specifically, the regression equation does not account for the influence of the Soil organic carbon content.The foc was negatively correlated age of contaminant association with the soil (‘weathering’), with kv, a finding consistent with previous theoretical treatthe eVect of varying soil moisture content, and fails to function ment.16 Further support for this relationship is supplied by for soils of high organic content, or in situations where ln Koa other experimental studies, each demonstrating a negative exceeds ca. 23, e.g., for congeners 138 and 180 at soil tempera- relationship between SOC evaporative loss from soil and tures approaching 0 °C.foc.17–19 Soil moisture content and water evaporation flux. In contrast Acknowledgements to MLRA of the whole data set, analysis of laboratory We gratefully acknowledge the Royal Society and the West experiments 9–14 inclusive, suggests that kv may be positively Midlands Regional OYce of the National Health Service linearly related to MC.In these experiments, MC was varied, Executive for supporting our work on PCBs, and Gian Marco while other independent variables were essentially held con- Currado and Lee Hoon Lim for assistance with sampling stant.Plots of kv against MC for PCBs 28, 52, and 101, atmospheric deposition. respectively, in these experiments (there were insuYcient data to permit similar plots for congeners 138 and 180) revealed good Pearson correlation coeYcients (0.947>R>0.831, i.e., References significant at the 95–99% confidence level ) for these congeners. These findings suggest that MC does exert an influence on kv, 1 F.X. R. van Leeuwen and M. Younes, Organohalogen Compd., 1998, 38, 295. but that the eVect detected in our laboratory experiments is 2 Department of the Environment, United Kingdom action plan for very small compared to that of ln Koa and foc. the phasing out and destruction of polychlorinated biphenyls (PCBs) It has also been suggested that PCB volatilisation fluxes are and dangerous PCB substitutes, Department of the Environment, positively correlated with water flux.16 However, neither the London, 1997.MLRA conducted on the full data set, nor regression of WF 3 S. J. Harrad, A. Sewart, R. Alcock, R. Boumphrey, V. Burnett, versus kv for experiments 9–14, revealed any significant R. Duarte-Davidson, C. Halsall, G.Sanders, K.Waterhouse, S. R. Wild and K. C. Jones, Environ. Pollut., 1994, 85, 131. relationship between these two variables. 4 S. Ayris, G. M. Currado, D. Smith and S. Harrad, Chemosphere, 1997, 35, 905. Octanol air partition coeYcient. In addition to our field 5 S. Harrad and D. Smith, Sci. Total Environ, 1998, 212, 137. experiments, other recent experimental studies support the 6 S.L. Grundy, D. A. Bright, W. T. Dushenko and K. J. Reimer, broad validity of eqn. (6). Chiarenzelli et al.20 observed that Environ. Sci. Technol., 1996, 30, 2661. approximately 60%, 55%, 35% and 20% of the PCBs contained 7 R. E. Alcock, J. Bacon, R. D. Bardget, A. J. Beck, P. M. in Aroclors 1242, 1248, 1254, and 1260, respectively, volatilised Haygarth, R. G. M. Lee, C. A.Parker and K. C. Jones, Environ. Pollut., 1996, 93, 83. over one day from subaqueous sand. Cousins et al.15 measured 8 R. Gan and P. M. Berthoux, Water Environ. Res., 1993, 66, 54. volatilisation fluxes of individual tri-through pentachloro PCBs 9 R. E. Alcock, A. E. Johnston, S. P. McGrath, M. L. Berrow and from sewage sludge ameliorated soils. MLRA of this dataset K. C. Jones, Environ. Sci. Technol., 1993, 27, 1918. showed that fluxes were positively correlated with the sludge 10 T. Harner and D. Mackay, Environ. Sci. Technol., 1995, 29, 1599. concentration and negatively correlated with log Koa. 11 T. Harner and T. F. Bidleman, J. Chem. Eng. Data, 1996, 41, 895. This study shows that over periods of up to 415 d, PCB loss 12 R. L. Falconer and T. F. Bidleman, Atmos. Environ., 1994, 28, 547. 13 B. Guilbeault, M. Sondossi, D. Ahmad and M. Sylvestre, M. Int. from soil under field conditions typically prevailing in the UK Biodeterior. Biodegrad., 1994, 33, 73. follows first-order kinetics. It reports first-order rate constants 14 W. A. Jury, W. F. Spencer and W. J. Farmer, J. Environ. Qual., for the loss of five PCB congeners from three soils, and 1983, 12, 558. provides evidence that they are correlated with ln Koa and 15 I. T. Cousins, N. Hartlieb, C. Teichmann and K. C. Jones, influenced by temperature as well as soil moisture and organic Environ. Pollut., 1997, 97, 229. carbon content. The age of PCB association with one soil 16 G. H. Eduljee, Chemosphere, 1987, 16, 907. 17 B. Lindhardt and T. H. Christensen,Water, Air, Soil Pollut., 1996, (‘weathering’) was shown to influence persistence, with signifi- 92, 375. cantly longer PCB half-lives observed in soil subjected to ca. 18 R. Haque and D. W. Schmedding, J. Environ. Sci. Health, Part B, 1 year’s prior ‘weathering’. Indirect evidence suggests that the 1976, 11, 129. primary loss mechanism is volatilisation, and although the 19 Y. Iwata, W. E. Westlake and F. A. Gunther, Bull. Environ. significance of aerobic biodegradation could not be dismissed, Contam. Toxicol., 1973, 9, 204. losses due to leaching, soil erosion, and uptake by biota were 20 J. R. Chiarenzelli, R. J. Scrudato and M. L. Wunderlich, Environ. Sci. Technol., 1997, 31, 597. shown to make negligible contributions to the overall loss. First-order rate constants for the volatilisation of PCBs from soil were determined under controlled laboratory con- Paper 9/03017D J. Environ. Monit., 1999, 1, 395–401 401