SUPPORTING INFORMATION

Transcription

1 SUPPORTING INFORMATION Elevated mobility of persistent organic pollutants in the soil of a tropical rainforest Qian Zheng, a,b Luca Nizzetto, *, c,d Xiang Liu, *,a Katrine Borgå, c,e Jostein Starrfelt, c,e Jun Li, a Yishan Jiang, a Xin Liu, a Kevin C. Jones, f and Gan Zhang a a State Key Laboratory of Organic Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, Guangzhou , China. b Graduate University of the Chinese Academy of Sciences, Beijing , China. c Norwegian Institute for Water Research, Oslo, 0349, Norway. d Research Centre for Toxic Compounds in the Environment, Brno, 62500, Czech Republic. e Department of Biosciences, University of Oslo, 0316 Oslo, Norway. f Lancaster Environment Centre, Lancaster University, Lancaster, LA1 4YQ, UK. 1

2 Contents Text S1. Air sampling system and approach....3 Text S1. Chemical analysis and QA/QC....3 S1.1 Details on methods... 3 S1.2 Quality Assurance/Quality Control (QA/QC)... 4 S1.2.1 Method detection limit definition... 4 S Procedural laboratory blanks... 5 S1.2.3 QA/QC standards and recovery results... 5 S1.2.4 Consistency... 6 Text S2. Performance of spiking....6 Text S3.Mass balance model for process rate determination....6 Text S4. Statistical methods....8 Figure S1. Scheme of an individual lysimeter and sampling system....9 Figure S2. Evolution over time of the vertical profile of labelled PCB Figure S3. Comparison of OC dependence of soil concentration Figure S4. Correlation between scaled leaching fluxes of labelled and native PCB congeners Figure S4b. Phase distribution of labelled PCB in the leachate Figure S5. Leaching masses of labelled compounds and their time trend Figure S6. Occurrence of 13 C-labelled PCBs in the passive mini PUF plugs during the three deployment periods Figure S7. Results of litter spiking Figure S8. Calibrated model fitting and errors Figure S9. Distribution of 13 C-labelled PCBs in the 3 active mini PUF plugs Table S1a. Summary of results (data reported in pg per compartment) Table S1b. Distribution of labelled PCB in the leachate samples Table S2. Estimated process rate and half-life values Table S3. Masses of C12 native congeners in litter soil and leachate (pg) Table S4. Soil layer dry masses and Total Organic Carbon (TOC) content Table S5. Details of leachate analysis results (pg) References

3 Text S1. Air sampling system and approach. The sampling system encompassed a 10 mm i.d. PTFE tube connected to a 2.5 cm i.d. glass tube hosting three 7 cm long PUF sorbents, deployed in series and connected to a peristaltic pump continuously operating at a flow rate of 6 L min -1. The inlet of the PTFE tube was placed inside the lysimeter opening at a distance of 5 cm from the spiked litter surface (Figure S1). Based on the lysimeter head space geometry, the active suction generated a residence time of the air immediately above the soil core (e.g. the air in the volume (7.3 L) of the portion of the external cylinder exceeding the ground) approximately of 75s. Since the experiment was conducted under a dense canopy (e.g. LAI > 5) and the spiked litter was protected by the walls of two concentric cylinders of respective heights 15 and 30 cm, the soil core surface never experienced exposure to direct wind flow. Text S1. Chemical analysis and QA/QC. S1.1 Details on methods A standard solution containing a range of 13 C PCBs was selected as RCs (EC-4058, Cambridge Isotope Laboratories, USA). The solution encompassed the following 13 C labelled congeners: PCB 28, PCB 52, PCB 101, PCB 138, PCB 153, PCB 180, PCB 209. The samples (litter and soil) from the lysimeters and the initial litter aliquotes (analyzed to define values used in the mass balance calculation) were freeze dried and then Soxhlet extracted with 300 ml mixture of DCM and acetone (v:v=5:1) for 36 h. Before extraction, a mixture containing TCmX, PCB30, PCB198, PCB209 was added as surrogate recovery standards. Activated copper granules were added to the collection flask to remove elemental sulphur. The extract was concentrated to 1 ml using a rotary evaporator and solvent-exchanged into hexane. Then the extract subsequently loaded onto a multiple layer clean-up column (8mm i. d.) packed with aluminum oxide (1cm, 3% deactivated), neutral silica gel (1cm, 3% deactivated), 50% sulfuric acid silica (3cm) and anhydrous sodium sulphate (1cm) and eluted with 20 ml hexane and DCM (v:v=1:1). The PUF disks and mini PUF plugs were Soxhlet extracted for 36 hours with DCM and acetone (v:v=5:1). The extract was concentrated to 1 ml using a rotary evaporator and solvent-exchanged into hexane. Then the extract subsequently cleaned on a column (8mm i. d.) packed with aluminum oxide (1cm, 3% deactivated), neutral silica gel (1cm, 3% deactivated), 50% sulfuric acid silica (3cm) and anhydrous sodium 3

4 sulphate (1cm), eluted with 20 ml mixture hexane and DCM (v:v=1:1). Water samples were first passed through baked (450 C for 4 h) glass fiber filters (GF/F, 14.2 cm diameter). The water phase was spiked with surrogate standards and liquid liquid extracted 5 times using a separator funnel and 50 ml DCM for each extraction. The extracts were combined and then dried with baked anhydrous sodium sulfate. After that, each sample extract was concentrated to 1mL by a rotary evaporator and solvent-exchanged into hexane. Then the elute was cleaned up through an 8 mm i.d. alumina/silica column packed with aluminum oxide (1cm, 3% deactivated), neutral silica gel (1cm, 3% deactivated), 50% sulfuric acid silica (3cm) and anhydrous sodium sulphate (1cm), eluted with 20 ml mixture hexane and DCM (v:v=1:1). The glass fiber filters were Soxhlet extracted for 36 hours with DCM and acetone (v:v=5:1). The extract was concentrated to 1 ml using a rotary evaporator and solvent-exchanged into hexane. Then the extract subsequently cleaned on a column (8mm i. d.) packed with aluminum oxide (1cm, 3% deactivated), neutral silica gel (1cm, 3% deactivated), 50% sulfuric acid silica (3cm) and anhydrous sodium sulphate (1cm), eluted with 20 ml mixture hexane and DCM (v:v=1:1). and concentrated to 0.5 ml under a gentle stream of nitrogen. A GPC column was used for the final clean-up step. 6 g of Bio-Beads S-X3 were used in a 15 cm glass column (2 cm i. d.). The concentrated samples were loaded and eluted with 55 ml a of a mixture hexane and dichloromethane (v:v=1:1). The first 15 ml were discarded and next 40 ml containing the PCBs was collected and concentrated under a gentle N 2 flow. Prior to GC-MS-MD analysis, 2 ng of 13 C labelled PCB128 were added as internal standards for spiked reference PCBs and native PCBs qualification. Briefly, an Agilent-5975N GC-MSD system equipped with a 50 m 0.25 mm i. d. CP-Sil 8CB capillary column (0.25 µm film thickness) was used and operated under selected ion monitoring (SIM). Filtered helium was used as the carrier gas at 1.2 ml min -1. The injector temperature was set at 250 C and the oven temperature increased from 60 C to 290 C at a rate of 4 C min -1. Data were acquired and processed on HP Chemstation software and the 8 13 C labelled PCBs and the following 27 native PCBs were quantified: PCB28, PCB37, PCB44, PCB52, PCB60, PCB66, PCB70, PCB74, PCB77, PCB82, PCB87, PCB99, PCB101, PCB105, PCB114, PCB118, PCB126, PCB128, PCB138, PCB153, PCB166, PCB169, PCB179, PCB180, PCB183, PCB187, PCB189. S1.2 Quality Assurance/Quality Control (QA/QC) S1.2.1 Method detection limit definition PUF field blanks were routinely included (one every 6 samples) in the analysis to look for occurrence of traces of 13 C labelled PCBs as result of cross contamination during sampling handling and analysis. No detectable levels of the RCs were found. 4

5 PUF disk field blanks were also used as surrogate field blanks for all the set of samples (soil, litter, and PUF used to trap leachate). PUF disk field blanks were transported in the field, exposed to air for short time (e.g. 30 minutes) and carried back to the laboratory. Method detection limits for PUFs (both for C 13 and C 12 PCBs) were defined as a) the average of the field blanks + 3 standard deviations, or b) the lower point of the calibration curve of the RC compounds, in case no traces of the target analytes were detected in the field blanks. Expectedly our field blanks did not show detectable levels of the labelled compounds. The method detection limits for this set of analytes were therefore obtained from definition b and was 1 pg g -1 for soil, litter and PUF samples and 0.5 pg L -1 for leachate samples. No detectable levels of the labelled PCBs were found too in the procedural and field blanks, therefore the method detection limit for these congeners in the litter and soil samples was also obtained from definition b. S Procedural laboratory blanks Procedural laboratory blanks were obtained by running the analysis without the addition of any sample in the soxhlet extractor (only the recovery and the internal standards were added). They were included routinely (every 5 real samples) during the analyses to exclude cross contamination during sampling handling and analysis. The procedural laboratory blanks had no detectable levels of the 13 C PCBs, for native PCBs, low levels (typically less than 5% of the levels in the real samples) of PCB 18 and trace levels of PCB28 and PCB 52 (see table M1). S1.2.3 QA/QC standards and recovery results In order to validate the quantitative method, the mixture of labelled compounds used as field spikes was routinely analyzed. Results were in agreement with the nominal concentration within a factor of 0.83 ± 0.15 and recovered level did no show any relationship with time (e.g. run or sample number) (p<0.05). Recovery values from the procedural laboratory blanks were also used to assess method performance in absence of a sample matrix. We compared these results with the recovery from individual real samples. Laboratory procedural blanks consistently showed recoveries 5 to 7 % higher than those of the real samples (depending on congeners) as a possible result of matrix effect on contaminant extraction. Recovery for individual samples were therefore corrected by these factors to achieve final recovery values as follows: 52.6±19.2%, 74.2±12.1%, 91.7±7.8%, 94.9±8.2% for TCmX, PCB30, PCB198, PCB209 recovery standards, respectively. Since the lighter PCBs had lower recovery and a relatively higher variance we decided to correct individual samples for recoveries. 5

6 S1.2.4 Consistency A subset of 3 duplicated analysis of the same litter sample showed agreement within 30% for all chemicals. Text S2. Performance of spiking. At the beginning of the experiment (Day 0) the mass of labelled PCBs measured in the litter accounted for 57-91% (depending on congeners) of their nominal spiked amount, owing to the rapid volatilization during the 24h pre-experiment phase (Figure S4). This behavior was the same observed in previous studies 1, 2. These pilot studies demonstrated consistency between mobility of artificially added labelled POPs and airborne native POPs, at the litter-soil interface. Text S3.Mass balance model for process rate determination. Assuming atmospheric depositions of labelled compounds to be negligible, the change of the masses (ng) and (ng) of a given labelled reference compound present in the litter and the soil core respectively, can be mathematically described as: = = (1) = = 1 + (2) where,, and (ng m -2 d -1 ) are the volatilization, litter-soil transfer, degradation and leaching fluxes, respectively,,, and (d -1 ) are rates of volatilization, litter-soil transfer and degradation, respectively. In order to obtain estimates of process rates we forced the model to find solutions of k x that better match the following constraints: ( ) = ( ) = ( ) = ( ) = (3) where ( ) are the experimentally measured masses of the labelled PCBs in the different compartments at time y={120, 270, 360}. To take into account the uncertainty surrounding the efficiency of the AAS to trap volatilization of 13 C PCBs a correction factor (q) was estimated simultaneously to the 6

7 process rate. The model prediction of 13 C PCB masses in the AASs ( ) is, in fact: =, (4) The uncertainty factor q was randomly extracted from uniform distribution in the range Such a range is extremely conservative since a value of q = 0.01 is equivalent to the assumption that the AAS only captured 1/100 of the real volatilization flux, a situation infringing mass balance closure by an off-set of 2 orders of magnitude higher than the mass of the added labelled congeners. The cumulative amount of labelled compounds undergoing leaching from the bottom of the soil core is experimentally given by: = + 5) where is the cumulative mass of labelled PCBs measured in PUF A, PUF B, PUF C and that measured in the leachate at a given time point. Finally, the mass of labelled POPs that underwent downward litter-to-soil transport during time ( ) is given by: = + 6) where is the mass of labelled congeners found in the soil core at a given time point. The Bayesian parameter estimation searches for the sets of values of, and q that minimize the sum of squared deviation between observed values at the 3 sampling time points (day 120, 270 and 360) and those predicted by integrating the model at the respective times (assuming the k value to be constant over time). All together the mathematics presented above describes an over-determined system with 5 unknown parameters (namely:,,, and q). In addition the model estimates an error variance, capturing the assumed normal deviation (on a log scale) of the observed versus the modelled values. The degradation half-life (t½, h) and its confidence boundaries were traced from the distribution of k degr assuming first order decay, as follows: t½=24 (10) 7

8 Text S4. Statistical methods. Normality test was conducted using the Lilliefors test. Correlation analysis for non-normally distributed dataset was conducted using the Spearman rank correlation analysis. P-values were calculated using exact permutation distribution. The Mann Whitney U test was used to compare mode and distributions of different process rates. Slopes of regression curves for the relationship between OC and concentration of labelled and native PCBs was conducted using ANOVA after checking for normality. Parameter optimization for the model parameterization was carried out using Markov Chain Monte Carlo techniques in a Bayesian framework. 3 Priors for all rates were assumed to be log-uniform. 4 In addition to the process rates, a standard deviation of the error model (i.e. the equation linking the simulated outcomes of the model with the observations), and the likelihood was defined as normal using log transformed values. Essentially such a procedure arrives at a chain of parameter values which make up the posterior distribution. The posterior is a combination of the prior and the likelihood based of the observations given the parameters. Obtained parameter estimates therefore fully take into account the variance of observation deriving from all sources of model conceptual, operational and analytical errors. The model predictions were checked by using median estimates of process rates to calculate a Spearman rank correlation coefficient between estimated and observed masses in the different compartments (Figure S8) 8

9 Figure S1. Scheme of an individual lysimeter and sampling system. The study encompassed 3 lysimeters sampled every fourth month. 9

10 Figure S2. Evolution over time of the vertical profile of labelled PCB. Labelled PCB concentrations (black points) and OC content in individual soil layers (grey squares) for two congeners at the opposite ends of the physical-chemical property range of PCBs (taken as examples). Vertical distribution of all labelled PCB tended to increase its dependence on OC content over time. This is shown by the P value observed at day

11 Figure S3. Comparison of OC dependence of soil concentration. Comparison of OC dependence of soil concentration between native 12 C-PCB (all soil cores) and labelled PCB (at t360 soil core C3) Scaling was performed by dividing each group of data for individual congeners by its maximum value in the group. 11

12 Figure S4. Correlation between scaled leaching fluxes of labelled and native PCB congeners. Note: the analysis included all pairs of selected congeners. Scaling was performed by dividing each dataset by the respective average leaching value. R (Spearman)= P< % prediction band 95% confidence 12

13 Figure S4b. Phase distribution of labelled PCB in the leachate. Data presented as percentage of the total cumulative leached mass 13

14 Figure S5. Leaching masses of labelled compounds and their time trend. Maximum losses were observed at day 270 coinciding with the wet season. Unexpectedly the more hydrophobic congeners on the right side of the chart experienced the highest leaching fluxes. 14

15 Figure S6. Occurrence of 13 C-labelled PCBs in the passive mini PUF plugs during the three deployment periods Amount (pg)g (B) R1-PAS minipuf R2-PAS minipuf R3-PAS minipuf 0 Compounds 15

16 ) Figure S7. Results of litter spiking pg nominal spike 13C-PCB28 13C-PCB52 13C-PCB101 13C-PCB138 13C-PCB153 13C-PCB180 13C-PCB209 Compounds 16

17 Figure S8. Calibrated model fitting and errors Note: the analysis was performed considering simultaneously predictions for all labelled congeners pools in litter, soil and leachate. Regression parameters refer to the log-log plot. 17

18 Figure S9. Distribution of 13 C-labelled PCBs in the 3 active mini PUF plugs. Amount (pg) R1-C1-PUFa R1-C1-PUFb R1-C1-PUFc 0 Compounds 18

19 Table S1a. Summary of results (data reported in pg per compartment) Lysimeter1 (t 0 -t 120 ) 13 C-PCB28 PCB52 PCB101 PCB138 PCB153 PCB180 PCB209 Added amounts (pg) Mass in AAS (cumulative) Mass in litter Mass in soil layer A Mass in soil layer B Mass in soil layer C Mass in soil layer D Mass in soil layer E Mass in lower PUF disk (cumulative) ( ) Mass in leachate (cumulative) ( ) Added amount t0 ( ) (pg) Mass in AAS (cumulative) Lysimeter2 (t 0 -t 270 ) 13 C-PCB28 PCB52 PCB101 PCB138 PCB153 PCB180 PCB Mass in litter Mass in soil layer A Mass in soil layer B Mass in soil layer C Mass in soil layer D Mass in soil layer E Mass in lower PUF disk (cumulative) ( ) Mass in leachate (cumulative) ( )

20 Lysimeter3 (t 0 -t 360 ) 13 C-PCB28 PCB52 PCB101 PCB138 PCB153 PCB180 PCB209 Added amount t0 ( ) Mass in AAS (cumulative) Mass in litter -* Mass in soil layer A Mass in soil layer B Mass in soil layer C Mass in soil layer D Mass in soil layer E Mass in lower PUF disk (cumulative) ( ) Mass in leachate (cumulative) ( ) *Due to a laboratory accident the extract of the litter samples for the t370 went lost. Table S1b. Distribution of labelled PCB in the leachate samples. Data are presented as cumulative pools (pg) in the three sampling stages. 13 C-PCB28 PCB52 PCB101 PCB138 PCB153 PCB180 PCB209 t120 t270 t360 dissolved particulate dissolved particulate dissolved particulate

21 Table S2. Estimated process rate and half-life values. Bold numbers in the rate columns are medians from the posterior distributions; numbers in italics give the mode (the most probable value in the posterior distributions). Ranges in parentheses are 95% credibility intervals (formally the 95% highest posterior density interval). Half-life values for degradation rates were calculated using the 85 th percentile. which is approximately where the density reaches the plateau (see Figure 3). 13 C-PCB28 Volatilization (d -1 ) t½ (d) 188 Export from litter to Leaching Degradation upper soil t½ t½ t½ ( ) ( ) ( ) ( ) C-PCB ( ) ( ) ( ) ( ) C-PCB ( ) ( ) ( ) ( ) C-PCB ( ) ( ) ( ) ( ) C-PCB ( ) ( ) ( ) ( ) C-PCB ( ) ( ) ( ) ( ) C-PCB ( ) ( ) ( ) ( ) 21

22 Table S3. Masses of C12 native congeners in litter soil and leachate (pg) PCB28 PCB52 PCB101 PCB138 PCB153 PCB180 t120 Mass in soil layer A Mass in soil layer B Mass in soil layer C Mass in soil layer D Mass in soil layer E Cumulative mass in leachate Mass in litter t270 Mass in soil layer A Mass in soil layer B Mass in soil layer C Mass in soil layer D Mass in soil layer E Cumulative mass in leachate Mass in litter t360 Mass in soil layer A Mass in soil layer B Mass in soil layer C Mass in soil layer D Mass in soil layer E Cumulative mass in leachate

23 Table S4. Soil layer dry masses and Total Organic Carbon (TOC) content Weight (g) Total OC (%) layers 120 d 270 d 360 d 120 d 270 d 360 d soil layer A soil layer B soil layer C soil layer D soil layer E litter n.a. n.a. n.a. Table S5. Details of leachate analysis results (pg) t120 t270 t360 Mar Apr-Jun Jul Aug Sep Nov 13C-PCB C-PCB C-PCB C-PCB C-PCB C-PCB C-PCB Volume of sample

24 References 1. Liu, X.; Ming, L.-L.; Nizzetto, L.; Borgå, K.; Larssen, T.; Zheng, Q.; Li, J.; Zhang, G., Critical evaluation of a new passive exchange-meter for assessing multimedia fate of persistent organic pollutants at the air-soil interface. Environmental Pollution 2013, 181, (0), Nizzetto, L.; Liu, X.; Zhang, G.; Komprdova, K.; Komprda, J., Accumulation Kinetics and Equilibrium Partitioning Coefficients for Semivolatile Organic Pollutants in Forest Litter. Environmental Science & Technology 2014, 48, (1), Gelman, A.; Carlin, J. B.; Stern, H. S.; Rubin, D. B., Bayesian data analysis. Chapman & Hall/CRC: Boca Raton, FLA, 2003; p MacLeod, M.; Fraser, A. J.; Mackay, D., Evaluating and expressing the propagation of uncertainty in chemical fate and bioaccumulation models. Environmental Toxicology and Chemistry 2002, 21, (4),