Modeling of gas generation from the river adjacent to the manufactured gas plant

Abstract

Ebullition of gas bubbles through sediment can enhance the migration of gases through the subsurface and potentially affect the emission of important greenhouse gases to the atmosphere. To better understand the parameters controlling ebullition, investigations of gas ebullition in the Grand Calumet river (Indiana, USA) were conducted. We found that gas ebullition might shift and change with different vertical hydraulic gradients and temperatures. CO₂ and CH₄ flux for each site increased with an increase in temperature. A comparatively simple linear relationship existed between the gas flux and the measured parameters (G_F = 0.316T + 300.66i, R² = 0.82). The gas flux in the sand cap was more variable than that in sediment. Moreover, the total field gas fluxes varied from 10 to 180 mmol m⁻² d⁻¹ for sediment and from 5 to 35 mmol m⁻² d⁻¹ for the sand cap, which proved an in situ sand cap could provide effective remediation. The analysis presented here has shown that gas fluxes and reactive transport modeling can provide effective means of investigating ebullition and quantifying gas transport.

---

1. Introduction

Pollution of freshwater resources by various organic and inorganic contaminants such as polycyclic aromatic hydrocarbons (PAHs) and other chemicals has attracted considerable attention from researchers. Studies were conducted to investigate gas ebullition and contaminants such as PAHs and to better characterize the existing hydrologic conditions around and within a river adjacent to a former manufactured gas plant (MGP) site at the Grand Calumet river, Indiana, USA.^1,2 Surface water of the Grand Calumet river indicated that tar and oil droplets migration from sediment was occurring near the former gas station. These indications were problematic, because tar was a dense, non-aqueous-phase liquid (DNAPL); thus, once it was deposited in the riverbed sediment, one would not typically expect it to float up from the riverbed to the water surface.

Different aspects of facilitated migration of contaminants from sediments have been investigated by many researchers, in particular gas generation from sediment. Gas migration from sediment was found to be a function of changes in air pressure.³ Changes in hydrostatic pressure due to changing elevation were also found at several field sites to influence rates of gas migration.⁴ Sediment temperature was found to influence gas migration from sediment in Lake Suwa, Japan, on a seasonal basis.⁵ Long-term trends in methane mass in water were evaluated at Onondaga Lake in New York, USA, and it was found that methane increased through the spring and summer, peaked in early fall, and rapidly decreased in late fall to winter.^6,7

Although ebullition was accepted as a potentially important mechanism for the fate of contaminants, no comprehensive studies have been reported in the literature related to the effects of temperature and elevation on ebullition. Palermo et al.⁸ found that gas ebullition can have a significant effect on sediment stability. Ebullition was the result of a series of processes in which excess gases were generated by microorganisms from organic matter. The gases released from contaminated sediments were generally methane (46–95%), nitrogen (3–50%), and trace amounts of hydrogen, carbon dioxide, ammonia, and hydrogen sulfide.⁹ Most of the gas bubbles originated from the upper 10–20 cm of the sediment column.¹⁰ Martens and Klump¹¹ reported a range of bubble sizes between 0.062 cm and 0.37 cm with a mean volume of 0.104 ml at a water depth of 7.5 m. Bubbles grew until a pressure threshold was reached as they had to build up a certain amount of buoyancy to overcome the cohesive strength of the sediment and migrate upward. Gas ebullition generally occurred episodically due to changes in pressure and water level, which influenced the sediment matrix and thus affected gas bubble release.¹⁰ Increased hydraulic gradients and atmospheric pressure changes led to a sudden release of gas bubbles, which ceased after the excess pressure was relieved. Data revealed that the size of the gas bubbles depended on the amount of gas in the bubbles, temperature and pressure; moreover, temperature strongly affected microbial activity as well as saturation of the gas.⁸ Therefore, the purpose of this study was to investigate the rate of gas ebullition and to build a gas ebullition model using multivariate regression analysis. In addition, understanding the facilitated migration process of the NAPL is necessary to formulate remedies to reduce the risk from tarry sediment.

2. Materials and methods

2.1 Gas collection and analysis

The volume of gas released from the sediment was measured using “gas tents” at five locations within Reach 6, six locations within Reach 7, and one location within Reach 5. In addition, gas was collected from four sand cap test cells with seepage meter domes in Reach 6. The test cells were installed by Purdue in 2008, and the two sand-only caps and two sand-organoclay caps were monitored. The sediment near each of these test cells was sampled at the same time as each of the sand caps for comparison. Fig. 1 shows the general location of the gas sampling activities.

Fig. 1 Locations of gas sampling activities. RC6 (top) is near the Hohman Avenue Bridge and the railroad bridge is evident from the change in surface topography over the river.

Each gas “tent” (Fig. 2) consisted of a frame made from PVC pipe (schedule 40, 7.6 cm ID) and PVC film (0.2 mm thickness). The area of each gas tent was 6.5 m² (3.05 m by 2.13 m). Each tent was held in place by 4 PVC pipes pushed into the sediment. In this way, each gas tent could move vertically as the river elevation changed by floating on the surface of the river. A closable sampling port (vent) was installed through the film near one of the corners. Upon sampling, the input port of an electric pump was attached to the port of the tent, and the output end of the pump to a 0.5, 8.1, or 20.3 L Tedlar gas sampling bag. Bags in series were filled completely by squeegeeing the gas (under the PVC film) to the corner where the gas at the vent was actively pumped. To sample the gas released from the sand cap test cells and the adjacent sediment in Reach 6, seepage meter domes were placed in the sand and adjacent sediment, respectively. The area covered by a seepage meter dome was 0.3 m², and the port that normally would be connected to a seepage meter flow tube was capped such that all gas emitted from the sediment was collected in a gas sampling bag attached to the central pipe through a connector valve and flexible tubing (Fig. 2). For each gas sampling event in both the gas tent experiments and the sand cap experiments, gas was collected for at least 7 days, removing (collecting) the gas every few days into the Tedlar bags to measure the volume. Gas ebullition was evident, as streams of bubbles were often observed emanating from the river after a sufficient quiescent period. Moreover, during the summer, gas release could be induced simply by disturbing the water or sediment surface with a boat oar.

Fig. 2 Schematic of sand cap test cell and photograph of gas collection tent at RC3, with gas evident under the PVC film.

Gases were collected into gas bags and then 1 ml gas was subsampled into evacuated 12 ml Labco exetainers (Labco Limited, UK). The analysis was carried out using a PDZ Europa trace gas analyzer (TGA) interfaced to a PDZ Europa 20/20 isotope ratio mass spectrometer (IRMS) (Sercon, Crewe, UK).

2.2 Seepage meter

The interfacial flow measuring system consisted of a dome with a flow tube and vent, a circuit board, and a computer (Fig. 3). The dome, made from stainless steel, had an OD of 61 cm, a height of 19.1 cm, and a volume capacity of 28.4 L. The gas vent was a 1.27 cm diameter PVC pipe attached to the top of the dome with a bulkhead flange of sufficient length to extend above the water surface, allowing gas to escape and water to rise within the pipe to the river's water table level. Closed-cell polyurethane foam attached to the rim of the dome ensured a watertight seal. The flux meter and the dome were connected directly using a flow tube. This allowed water to flow between the river and the flux meter at a volumetric rate equal to the rate across the sediment–water interface. As water flowed through the tube, four thermocouples positioned within the tube at different positions sensed the temperature change as a function of time. The volumetric flow rate in the tube was calculated from the temperature–time profile measured by the two thermocouples downstream from the heater, as described in our previous study.¹

Fig. 3 Schematic of field implementation of the seepage meter system.

Sediment at the experimental site on the Grand Calumet river was generally composed of silt-sized particles with high organic matter, consisting of both natural organic matter and coal tar, as reported in our previous study.¹ A fine- to medium-grained sand layer occurred below this organic-rich top layer. Interfacial flow (Darcy flux) was measured at 7 locations (3 within Reach 6 and 4 within Reach 7) (Table S1†). Measurements were made at each location between 4 and 8 times over a 15 month period. The first measurement was made near RC3 on March 28, 2011, and the last measurement was made near RC12 on May 25, 2012. On each day at each location, generally between 2 and 6 measurements were made, each requiring approximately 30–40 minutes with the reported Darcy velocity being the average of all values recorded.

2.3 Installation of piezometers

Fifty piezometers, gas collector sheets and eight stream gauges were installed to monitor the local hydrology and gas ebullition along a 2.5 km stretch of the Grand Calumet river (Fig. 1). Piezometers were constructed from 4.45 cm OD polyethylene pipe with 20 holes (0.95 cm diameter) drilled into the pipe within 15.2 cm of the capped end. These holes were wrapped in a porous geotextile and aluminum wire mesh to avoid sediment inflow and clogging. The stream gauges (i.e., piezometers at depth 0) were constructed in the same manner with the holes and screen located within a 60 cm segment at a sufficient distance from the capped end to ensure they would be located above the sediment–water interface after installation. The piezometers and stream gauges were installed manually by pushing to the target depth. In the river, six piezometer clusters were installed, with each cluster consisting of two piezometers pushed to depths of 1.2 and 2.4 m below the sediment–water interface and one stream gauge, each located approximately 15 cm apart. The location of each cluster is shown on Fig. 1 and 3.

3. Results and discussions

3.1 Effect of gas ebullition

As shown in Fig. 4, below the water surface and 2.0–3.3 m organic-rich sediment layer was 0.3–0.6 m of a fine- to medium-grained sand layer over a continuous, less permeable clay layer. The sand layer was extensive enough to be connected, but not evenly distributed over the site. Fig. 4 presents a schematic of the basic hydro-biogeochemical processes occurring within the sediment as it currently exists (eqn (1) and (2)).

Fig. 4 Possible pathway for sediment contaminant transport by gas bubble ebullition. (a) Gas collection by PVC tent; (b) gas collection by dome.

Active aerobic condition:

Active anaerobic condition:

The measured total field gas fluxes varied from 10 to 180 mmol m⁻² d⁻¹ for sediment and from 5 to 35 mmol m⁻² d⁻¹ for the sand cap (Fig. 5). Overall, CH₄, N₂, and CO₂ comprised 54.44 ± 8.22%, 39.72 ± 8.92%, and 5.83 ± 1.21% (values are means ± SD) of the gas by volume, respectively. Gas ebullition was higher for Reach 6, Reach 7 and sediment, whereas release of gas for Reach 5 and the sand cap was comparatively lower. McLinn and Stolzenburg⁵ reported that the low activity and presence of PAHs in the sand cap layer on contaminated sediment are the key factors for higher gas ebullition. It is clear from Fig. 5b that the gas flux value of the sediment dome was lower than that in the sediment PVC device. The possible reason was that, in comparison to the sediment–water interface, there was a gas flux at the water–air interface due to the decomposition by microbes in the river water. In addition, N₂ has a high percentage content in air (78.12%) and its solubility is 1/50 by volume. N₂ dissolved in the water could be collected with the PVC device using flowing water. Analysis of simulated CH₄ transport provides a measure of the significance of ebullition as a transport mechanism. At RC7b, the total gas produced was 128 mmol m² d⁻¹. These large episodic releases indicated that they do not commonly coincide with short-term changes in water table elevation. Therefore, ebullition fluxes could be largely indeterminable by chamber and tower based measurements and by methods that rely on changing water table elevations to estimate methane fluxes.^12,13

Fig. 5 Quantity of gas released (a) at different locations, CH₄, N₂, and CO₂ comprised 54.44 ± 8.22%, 39.72 ± 8.92%, and 5.83 ± 1.21% (values are means ± SD) of the gas by volume, respectively; and (b) gas collection at different devices e.g. sediment dome and PVC tents.

As shown in Fig. 6, the CO₂ and CH₄ flux for each site increased with an increase in temperature. Higher temperatures in spring and summer led to a higher G_F compared to the fall. The relationship between the CH₄ flux of sediment and the water temperature was consistent with the results of DelSontro et al.,¹² and it was clear that the CO₂ and CH₄ flux values of sediment were higher than those in the sand cap. Comparing Fig. 6c with Fig. 6f, gas flux in the sand cap was more variable than that in sediment. The reason may be the complex system in the sand cap. No bioturbation exists at the cap-sediment interface and chemical migration processes are also much slower. Therefore, the upward migration of contaminants goes through the sand cap in sediment. The cap materials prevent pollutants from entering into the water by adsorption, entrapment, bonding and degradation.^14–16 This function is very similar to the active cap layer or active permeable wall in the treatment of groundwater contamination.¹⁴ Compared to sediment, the permeability of the sand cap is higher but the volume of the gas flux is lower, which could be due to the aerobic microbial activity of the sand cap being higher; however, anaerobic conditions exist in sediment, therefore the volume of the gas flux in sediment was significantly higher than that in the sand cap (Fig. 5).

Fig. 6 (a, b, d and e) CO₂ and CH₄ flux in sediment and sand cap with respect to temperature. (c and f) Relationship between the CH₄ and CO₂ flux in sediment and sand cap.

3.2 Effect of hydraulic head and Darcy flux

The piezometric head h is a measurement of the hydraulic head at the point of measurement referenced to some standard elevation. From h values measured at the river stage (h₀), 1.2 m deep (h_1.2), and 2.4 m deep (h_2.4) piezometers in each piezometer cluster, vertical hydraulic head gradients i (m m⁻¹), were calculated by dividing the hydraulic head differences between the piezometers by the depth difference dz,where the subscripts a and b refer to the position within the sediment where the head was measured relative to the sediment–water interface (h_a is always river level; h_b is the head in one of the piezometers; dz = 1.2 m). As shown in Fig. 7, the water levels within the 1.2 m piezometers were generally higher than the river elevation over the 4 months of continuous measurement from early May to early September, 2012. Temporal changes in the gradients were minimal, except after high-rainfall events when the changes in elevation of the river (h₀) were more significant than the changes in water levels within the 1.2 m piezometers (p > 0.05) (e.g., May 8, June 17 and August 26).

Fig. 7 Comparative graph of the percentage of time under discharge conditions and under recharge conditions for the seven river clusters with data loggers at different hydraulic heads, (a) at 1.2–2.4 m; and (b) at 0–1.2 m.

The vertical Darcy velocities or specific discharges, q, measured at each river cluster position are reported in Table S1.† The vertical hydraulic head gradients (i) at each respective river cluster were measured manually on the same day when q was measured. Over all measurements, the range for the specific discharge (q) was 0.24–2.53 cm d⁻¹.

Note that the reported values of the vertical change in hydraulic head (i.e. magnitude of positive or negative head) were reflected in the corresponding seepage rates that were measured according to Darcy's Law. The vertical hydraulic conductivity, K_v (cm d⁻¹), within the top 1.2 m sediment layer is calculated by dividing q by the corresponding hydraulic head gradient, i_0–1.2,

where dh is the piezometric head difference between the stream gauge and the 1.2 m piezometer and dz is the elevation difference between the sediment–water interface and the 1.2 m piezometer screen (i.e., 1.2 m). The calculated values of K_v are reported in Table S1† with values ranging from 1.26 × 10⁻⁵ to 2.93 × 10⁻³ cm s⁻¹. The major advantage of the interfacial flow meter system described in this study is the ease with which it can be deployed to measure relatively low flow rates across the sediment–water boundary. Under the lowest flux conditions (e.g., q = 0.29 cm d⁻¹ for RC11 on 5/22/2012; Table S1†) discharge across the sediment–water interface through the area circumscribed by the collar was 0.49 ml min⁻¹. With an accurate addition of 3.00 ml min⁻¹ using the pump, the fraction of flow due to groundwater discharge was nearly 30%, providing an accurate measurement even at this low flow rate. In some cases, measurements were made with the flow of water into the dome in alternate directions at the same arbitrary flow rate such that the measured rate of discharge through the tube was either (a) actual flow + pump flow or (b) actual flow − pump flow. In the latter case, the net flow direction was into the dome, requiring T_m (the maximum temperature occurring at each thermocouple) to be measured at the thermocouple on the other side of the heater, nearest to the heater. In this case, the net groundwater flow rate is simply the sum of the two measured flow rates divided by 2, avoiding the need to accurately determine the pump flow rate.

Discharge and recharge of water from sediment had a large influence on the release of gas from tar-contaminated riverbeds. Fig. 7 reveals the recharge and discharge trends for different study sites at different depths of river sediment. As shown in Fig. 7a, vertical gradient data for 1.2–2.4 m at RC6, RC5, RC7, RC8 and RC11 show a higher vertical discharge, whereas recharge gradient values were found to be very low at these sites. The gas ebullition rate was also very high at sites where the vertical gradient discharge was high. This trend indicated that release of gas also depended on the vertical gradient discharge. A possible reason for this trend might be that pressure release at the site resulted in a higher release of gas. The vertical gradient discharge and recharge rates at 0–1.2 m were not similar to those at 1.2–2.4 m. The deep sediments had high pressure that was responsible for the release of gas. Increased hydraulic pressure and vertical gradient or hydrostatic pressure led to a higher release of gas bubbles, which ceased after the excess pressure was relieved.

In the case of high pressure and discharge rates in sediments, the sediment layer could force newly generated gas bubbles to migrate through the available pores, resulting in breaking up larger bubbles into smaller ones. These bubbles then broke out of the sediment layer and flew upward according to the vertical gradient discharge trend and pressure in the sediments. The size of the gas bubbles and their release rate depended on the amount of gas present in the sediments and the ambient temperature and pressure.

3.3 Effect of gas flux, elevation, and tar migration

Water elevation played a very important role in gas ebullition. It generated pressure in sediment that was responsible for forcing gas bubbles to move upwards. Fig. 8 shows that gas flux fluctuated according to changes in river elevation. It reveals that gas flux decreased as the river level increased. The value of gas flux measured by the plastic film was highest for sediment, and the value from the test cell for sediment was higher than for the sand cap at the same time. High river elevation generated pressure on the sediment layer or sand cap, which forced gas bubbles to move upwards. It could also be concluded that the sand cap was still very effective in reducing upward flow of contaminants. The impact of gas flux and seepage depended on the rate of the fluxes and therefore in situ measurements of the fluxes were required.

Fig. 8 Fluctuation of gas flux, (a) in sediment and sand cap; and (b and c) fluctuation in river elevation and gradient.

Pore water flow through the sediments was presumably driven by piezometric head gradients that varied over time due to hydrologic processes. In estuaries, the effects may exhibit shorter time responses due to tidal fluctuations which can create short-term variations in head differences. The highest groundwater discharge corresponds to periods of low water level and it could potentially even reverse direction. In the Grand Calumet river, reported measurements of seepage rates ranged from 0.24 to 2.53 cm d⁻¹. The groundwater seepage phenomenon could indirectly affect the stability of sediments by altering the consolidation rates in the sediment and changing their bulk density and thus their erosion resistance. Simon and Collison²³ stated that, in addition to advective-flow-induced shear stresses on cohesive stream beds, another mechanism contributing to the detachment of cohesive aggregates is upward-directed seepage forces. The range of groundwater fluxes reported in the literature varies significantly, depending on the sediment type of the bed and other characteristics of the site. Spatial and seasonal variations at the sites where seepage measurements were collected also affected the ranges. It should be noted that, since an upward vertical gradient of greater than 1 implies rapid movement, the bed can not be stable due to the seepage effects. One possible explanation is that non-Darcy flow through channels may be responsible for the primary transport of pore water through the sediments. This preferred flow would change many aspects of sediment resuspension and mass transfer of contaminants, which should be carefully considered for its relevance at a particular site. Experimental investigations studying seepage effects should be performed considering the possibility of channel formation.

3.4 Linear regression analysis describes gas ebullition

A comparatively simple linear relationship existed between gas flux and the measured parameters, as shown by multivariate regression analysis:

G_F = 0.316T + 300.66i, R² = 0.82 (5)

where G_F is the molar gas flux (mmol m⁻² d⁻¹), T is the pore water temperature (°C) and i represents the vertical hydraulic gradient (m m⁻¹).

In the recharge direction, there would be no relationship between gas flux and the vertical hydraulic gradient. As shown in Fig. 7 and 8, all vertical gradient data exhibited higher vertical discharge and the recharge gradient values were very low in our studies. This indicated that the release of gas also depended on the vertical gradient discharge. As shown in Fig. 9b, methane flux increased as the vertical hydraulic gradient (discharge) rose. These results were consistent with the previous trend described by Huls and Costello.¹⁷ In previous studies, if the discharge rate was high, gas ebullition was also high, which could affect erosion rates.¹⁸

Fig. 9 Comparison of measured versus model-predicted gas flux: (a) compares measured gas flux values to regression-predicted gas flux (the dashed line represents a 1 : 1 slope); (b) compares methane flux values versus vertical hydraulic gradient (discharge) values.

Exploratory factor analysis was performed to assess the robustness of this regression. There are few studies reported in the literature related to temperature and gas flux. However, cases are rare in the literature about the parameters G_F, sediment temperature, and vertical hydraulic gradient. Moreover, no gas ebullition model was reported regarding the relationship between the vertical hydraulic gradient and gas flux in the field. There are two models available in the literature corrected for gas fluxes.¹⁹ The model results were converted from a volumetric basis, assuming that the top 1 m of sediment at the field sites is ebullition-active, to provide a consistent comparison with the measured G_F values and the regression. Both literature models predict substantially higher G_F values than those observed in the field. As shown in Fig. 9a, the calculated gas flux matched the measured values better (R² = 0.82). Compared with previous research, these two parameters were able to predict G_F better than all the other measured parameters, explaining 82% of the variation (Fig. 9a). The choice of model for estimating the volume of gas within the soil at a bog site depends on the elasticity of the semipermeable, semi-confining layers that allow an interval of overpressured sediment to exist at depth. Rosenberry et al.¹³ provided a gas ebullition model based on barometric efficiency; this model is not sensitive to the difference between the porosity and volumetric moisture content of the sediment. Since the Grand Calumet river has a substantially overpressured interval at depth, it may be reasonable to prefer the overpressure (or hydraulic head) model over the barometric efficiency model.¹³ Beckwith and Baird²⁰ carried out a laboratory column study using time-domain reflectometry probes that indicated a gas content between 5% and 10% by volume. These methods do not provide a direct, in situ measurement of gas volume. The methods described above are based on a destructive/invasive sampling methodology. Sediment samples were collected from peatlands, some of which had been dewatered by natural or man-made causes.¹³ They were also instantaneous measurements. The methods used in this paper are non-invasive and non-destructive. In addition to providing an in situ measure of gas volume, the methods presented here provide continuous data, allowing the response of gas volumes at different sites to be related to climatic drivers. There is no paper reporting on the difference between air–water interface and sediment–water interface gas collection.

Gas bubbles from the sediment with the tar deposit were generated by anaerobic degradation of organic matter, consisting of organic material in the riverbed (sawdust and other detritus), as well as low-molecular-weight (LMW) PAHs in the tar, as discussed by Godsy et al.²¹ Compounds which may be lost from the sediment due to gas ebullition could include mono-, di- and trichlorobiphenyl congeners, toxaphene and other semi-volatile environmentally persistent organic compounds. Tar migration to surface water was mostly observed in those areas where both ebullition and tarry sediment were observed. However, ebullition occurred only in a portion of the tar deposit in which the water was relatively shallow and there was sufficient organic matter in the sediment. It is well known that ebullition is a dynamic equilibrium between degradation of organic carbon, water depth, and sediment strength, such that no one parameter will control gas bubble generation.¹³ It could be concluded from the results that gas ebullition increased with Darcy flux and temperature. A possible reason could be that microbial growth increased with increasing temperature.¹⁰ Microbial activity also increased the amount of gas, leading to the formation of new gas in deep sediments.¹⁰ Thus, gas production was higher in the current study due to the increase in Darcy flux, temperature and microbial growth in contaminated river sediments, which is inconsistent with previous results.^12,22 The final phase of bubble or contaminant transport from sediments was bubble ejection from the sediments, movement through the water column, and release to the atmosphere. Based on the studies, it was apparent that the transport of sediment-associated organic compounds by way of sediment bubbles may be an important pathway. This should be considered in toxic-chemical management plans and models for the Great Lakes Basin. It is a mechanism by which pollutants in place in sediment may be recycled within the basin and could also represent a pathway whereby contaminants could be transported outside the basin.

4. Conclusion

There is no paper reporting on the difference between air–water interface and sediment–water interface gas collection. Hydraulic head gradients and temperature data can be used to estimate the volumes of gas bubbles in sediment. Results showed that a comparatively simple linear relationship existed between gas flux and measured parameters (G_F = 0.316T + 300.66i, R² = 0.82). The methods used in this paper are non-invasive and non-destructive. In addition to providing an in situ measure of gas volume, the methods presented here provide continuous data, allowing the response of gas volumes at different sites to be related to climatic drivers. In addition, these results have implications for capping designs in ebullition-active sediment sites, and proved an in situ sand cap could provide effective remediation for tar contamination. The analysis presented here has shown that gas fluxes and reactive transport modeling can provide effective means of investigating ebullition and quantifying gas transport. Further work on this modeling using different sites and gas production rates will lead to a better understanding of controlling toxic chemicals by way of sediment bubbles.

Acknowledgements

Funding for this study was supported by Weston Inc., and Purdue University, the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD), and the Natural Science Foundation of Jiangsu Province, P.R. China (BK20131287).

References

1. S. Hyun, C. T. Jafvert, B. Jenkinson, C. Enfield and B. Johnson, Chemosphere, 2007, 68, 1020–1029.
2. C. T. Jafvert, D. Lane, L. S. Lee and P. S. C. Rao, Chemosphere, 2006, 62, 315–321.
3. M. D. Mattson and G. E. Likens, Nature, 1990, 347, 718–719.
4. J. P. Chanton, C. S. Martens and C. A. Kelley, Limnol. Oceanogr., 1989, 34, 807–819.
5. E. L. McLinn and T. R. Stolzenburg, Environ. Toxicol. Chem., 2009, 28, 2298–2306.
6. D. A. Matthews, S. W. Effler and C. M. Matthews, Archiv fur Hydrobiologie, 2005, 163, 435–462.
7. J. Zeikus and M. Winfrey, Appl. Environ. Microbiol., 1976, 31, 99–107.
8. M. Palermo, T. Thompson and F. Swed, Response to a document by the Johnson Company: Ecosystem-based rehabilitation plan–An integrated plan for habitat enhancement and expedited exposure reduction in the lower Fox River and Green Bay, 2002.
9. N. J. Fendinger, D. D. Adams and D. E. Glotfelty, Sci. Total Environ., 1992, 112, 189–201.
10. J. Joyce and P. W. Jewell, Environ. Eng. Geosci., 2003, 9, 167–178.
11. C. S. Martens and J. Val Klump, Geochim. Cosmochim. Acta, 1980, 44, 471–490.
12. T. DelSontro, D. F. McGinnis, S. Sobek, I. Ostrovsky and B. Wehrli, Environmental Science & Technology, 2010, 44, 2419–2425.
13. D. O. Rosenberry, P. H. Glaser, D. I. Siegel and E. P. Weeks, Water Resour. Res., 2003, 39, 1066.
14. D. Reible, D. Lampert, D. Constant, R. D. Mutch Jr and Y. Zhu, Biorem. J., 2006, 17, 39–53.
15. P. H. Jacobs, Water Res., 2002, 36, 3121–3129.
16. K. Yin, P. Viana, X. Zhao and K. Rockne, Sci. Total Environ., 2010, 408, 3454–3463.
17. H. H. Huls and M. Costello, Third International Conference on Remediation of Contaminated Sediments, New Orleans, LA, 2005.
18. R. Jepsen, J. McNeil and W. Lick, J. Great Lakes Res., 2000, 26, 209–219.
19. P. Z. Viana, K. Yin and K. J. Rockne, Environ. Sci. Technol., 2012, 46, 12046–12054.
20. C. W. Beckwith and A. J. Baird, Water Resour. Res., 2001, 37, 551–558.
21. E. M. Godsy, D. F. Goerlitz and D. Grbic-Galic, Ground Water, 1992, 30, 232–242.
22. R. T. Amos and K. U. Mayer, Environ. Sci. Technol., 2006, 40, 5361–5367.
23. A. Simon and A. J. Collison, Earth Surf. Processes Landforms, 2001, 26, 1421–1442.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4ra06627h

---