Article pubs.acs.org/est
Effect of Dissolved Oxygen Manipulation on Diffusive Emissions from NAPL-Impacted Low Permeability Soil Layers Lisa M. Clifton, Paul R. Dahlen, and Paul C. Johnson* School of Sustainable Engineering and the Built Environment, Ira A. Fulton Schools of Engineering, Arizona State University, Tempe, Arizona 85287, United States ABSTRACT: Aquifer physical model experiments were performed to investigate if diffusive emissions from nonaqueous phase liquid (NAPL)-impacted low-permeability layers into groundwater moving through adjacent NAPL-free high-permeability layers can be reduced by creating an aerobic biotreatment zone at the interface between the two, and if over time that leads to reduced emissions after treatment ceases. Experiments were performed in two 1.2-m long × 1.2-m high × 5.4 cm wide stainless steel tanks; each with a highpermeability sand layer overlying a low-permeability crushed granite layer containing a NAPL mixture of indane and benzene. Each tank was water-saturated with horizontal flow primarily through the sand layer. The influent water was initially deoxygenated and the emissions and concentration distributions were allowed to reach near-steady conditions. The influent dissolved oxygen (DO) level was increased stepwise to 6.5−8.5 mg/L and 17−20 mg/L, and then decreased back to deoxygenated conditions. Each condition was maintained for at least 45 days. Relative to the near-steady benzene emission at the initial deoxygenated condition, the emission was reduced by about 70% when the DO was 6.5−8.5 mg/L, 90% when the DO was 17−20 mg/L, and ultimately 60% when returning to low DO conditions. While the reductions were substantial during treatment, longer-term reductions after 120 d of elevated DO treatment, relative to an untreated condition predicted by theory, were low: 29% and 6% in Tank 1 and Tank 2, respectively. Results show a 1−2 month lag between the end of DO delivery and rebound to the final near-steady emissions level. This observation has implications for post-treatment performance monitoring sampling at field sites.
■
INTRODUCTION Permeability contrasts are present in most subsurface systems, whether they are associated with layers of different unconsolidated materials or fractures in rock matrices. This poses challenges for soil and groundwater remediation as many technologies rely on fluid flow for contaminant extraction or reactant delivery (e.g., soil vapor extraction, pump and treat, in situ chemical oxidation, air sparging, enhanced bioremediation). The outcome is often preferential treatment of the higher permeability zones, and ineffective remediation of the lower permeability regions.1 Source zone natural depletion beneath the water table will also over time result in preferential contaminant mass loss from higher permeability zones and persistent residuals in lower permeability zones because the dominant depletion processes (dissolution and biodegradation) are dependent on fluid flow for removal of dissolved compounds and delivery of electron acceptors and donors.2 When low-permeability zones or otherwise hydraulically isolated regions are not remediated or naturally depleted, they can continue to serve as decades-long sources of groundwater contamination because chemicals slowly diffuse out of those zones and into groundwater moving through the adjacent higher permeability zones. The issue of diffusive emissions from lower permeability layers in unconsolidated settings and from © 2014 American Chemical Society
secondary porosity in fractured rock has received attention in recent years, particularly in relation to dissolved chlorinated solvents in a range of geologic settings.3−7 Remedial strategies for sites with persistent emissions from lower permeability zones often focus on groundwater impacted by the emissions because of the difficulties in treating the sources; those strategies may make use of combinations of hydraulic containment, reactive flow-through barriers, and monitored natural attenuation of the dissolved contaminant plume. Because the source is not being remediated, those strategies must be maintained long-term until emissions from the low permeability zones decline naturally to acceptable levels. Remediation of the low permeability zones may be attempted, and options for that include excavation and in situ thermal heating to drive contaminants out of the lower permeability zones. The strategy examined in this work lies between the two extremes discussed above: reduction of the diffusive emissions to an acceptable level of groundwater impact via partial reactive Received: Revised: Accepted: Published: 5127
December 27, 2013 April 6, 2014 April 9, 2014 April 9, 2014 dx.doi.org/10.1021/es405775q | Environ. Sci. Technol. 2014, 48, 5127−5135
Environmental Science & Technology
Article
Figure 1. Conceptualization of interface treatment and its effects on emissions from low permeability zones.
Figure 2. Schematic of physical model tank experiment. Internal tank thickness is about 5.4 cm.
achieve >50% fractional reduction in mass flux required from 30−90% fractional mass removal, depending on the experimental geometry and treatment method. Our work focused on investigating if dissolved hydrocarbon emissions from petroleum nonaqueous phase liquid (NAPL)-
treatment of the low permeability matrix. Marble et al.8 explored one version of this concept in small-scale laboratory tests involving TCE and in situ chemical oxidation using permanganate. They reported empirically derived relationships between flux reduction and fractional mass reduction; to 5128
dx.doi.org/10.1021/es405775q | Environ. Sci. Technol. 2014, 48, 5127−5135
Environmental Science & Technology
Article
impacted low-permeability regions can be reduced by the twostage process illustrated conceptually in Figure 1: (a) in the near-term, by the creation of a reaction zone at the high-low permeability zone interface, and (b) in the long term, by creating a NAPL-depletion zone in the interface region of sufficient thickness that the concentration gradient, emission rate, and groundwater impact are reduced after active treatment ceases. This was investigated through the use of laboratory-scale physical models for the case of benzene (C6H6) emissions from NAPL-impacted soils and an aerobic interface reaction zone. It was hypothesized that the reaction zone shown conceptually in Figure 1 could be created by delivering water containing dissolved oxygen (DO) along the high/low permeability zone interface because benzene is aerobically biodegradable and benzene-degrading organisms are relatively ubiquitous in soils.9
for the upper support brace. DO sampling ports were installed in the tank inlet and outlet, and water sparging chamber. Vertical manifolds at the tank inlet and outlet sides spanned only the high-permeability layer and were constructed of 53 cm long × 0.64 cm diameter stainless steel mesh Geoprobe soil gas implants clamped to 0.64 cm stainless steel support tubing. This, in combination with the 1000× permeability contrast, ensured that flow was primarily through the high permeability layer and insignificant in the low permeability layer as confirmed by dye and bromide tracer tests. Petroleum Hydrocarbon Mixture. Benzene was chosen because it is a groundwater contaminant of regulatory concern at petroleum spill sites and, like many dissolved petroleum components, biodegrades under aerobic conditions.9 To simulate benzene release from a gasoline NAPL mixture into groundwater, benzene was mixed with indane (C9H10), a polycyclic aromatic lower-solubility hydrocarbon (S = 109 mg/ L-H2O vs 1780 mg/L-H2O for benzene). The relative proportions of the two compounds were selected so that the dissolved benzene concentration in water in equilibrium with the NAPL would be in the range 10−100 mg/L-H2O, which is comparable to total dissolved petroleum hydrocarbon concentrations in NAPL source zones at gasoline spill sites. Aquifer Materials. To create a two-layer contrastingpermeability system with native benzene-degrading microorganisms, two materials were selected: 20−40 mesh Colorado quartz sand and quarter-minus decomposed Yavapai granite sifted through a No. 16 sieve (1.2 mm), with hydraulic conductivities of 1.9 × 10−1 cm/s and 3.5 × 10−4 cm/s, respectively. Hydraulic conductivities were determined by constant-head permeameter tests.10 The fraction of organic carbon was similarly low for both materials, ranging from about 0.002−0.004 g-organic carbon/g-soil. Aerobic benzene degrader presence and activity was verified in both materials by soil-slurry microcosm tests using clean soil and about 1 mg/LH2O benzene solutions.11 Post-test benzene degradation activity was also verified in the sand using samples collected from within the tanks at shutdown. Inhibition of degrading activity was noted at higher dissolved concentrations (approximately 150 mg/L-H2O dissolved benzene and 100 mg/L-H2O dissolved indane) representative of the interior of the NAPL-impacted low permeability layer. This is consistent with the inhibition at comparable concentrations previously reported by Alvarez et al.9 NAPL-Impacted Low Permeability Media Preparation. A 99 L (3.5 ft3) capacity cement mixer was used to mix approximately 200 g NAPL (20 g benzene, 180 g indane) with 93 kg of low-permeability granite. These amounts were selected based on partitioning model projections to yield a final total NAPL concentration of about 1000−2000 mg/kg-soil and dissolved benzene concentrations of 10−100 mg/L-H2O after considering volatilization losses anticipated during tank preparation. The granite and NAPL were mixed for 20 min, with the mixer opening sealed with two layers of heavy-duty aluminum foil held in place with cloth and wire. After mixing, an aluminum scoop was used to transfer the NAPL-impacted granite into six 13 L (3.5 gal) steel buckets with clamp-on lids. Tank Packing Procedure. The tank interior was purged with nitrogen (N2) prior to packing; the O2 level was measured at 3.3% v/v using a YSI 550A DO meter. Tank packing procedures were performed in quick succession to minimize hydrocarbon volatilization losses. After emptying each bucket into the tank, the granite was leveled and tightly packed by
■
MATERIALS AND METHODS Conceptual Experimental Design. Proof-of-concept experiments were performed using two two-dimensional aquifer physical model tanks shown schematically in Figure 2. Each tank was prepared similarly with two water-saturated soil layers of equal thickness: an upper high-permeability layer and a lower low-permeability layer. The low permeability layer was mixed with a two-component NAPL prior to packing in the tank. The high-low permeability interface ran horizontally across the middle of the tank. The aquifer materials were selected to achieve a 1000× permeability contrast, so that lateral flow would occur primarily through the high permeability layer with insignificant flow through the low permeability layer. Influent, effluent, and in-tank dissolved concentrations of hydrocarbons, oxygen, as well as flow rate and pressure differential were monitored with time as the inlet DO concentration was manipulated through the sequence: deoxygenated water (15 cm) below the high-low permeability interface near sampling port rows 17 and 19. The dissolved hydrocarbon concentrations in this area were relatively constant during the experiments and there was no evidence of DO transport to these depths. Postexperiment, each tank was rotated to a horizontal position, the viewing window was removed, and soil samples were collected. Approximately 20 g of soil were placed into a 40 mL VOA vial with 20 mL of methanol. After shaking and settling, a 1 μL methanol sample was injected into a gas chromatograph equipped with a flame ionization detector (GC-FID) and calibrated for benzene and indane using 1, 10, 100, 1000, and 2500 mg/L-methanol standards, with analytical results reported as mg-hydrocarbons/kg-dry soil. Tank 1 benzene and indane soil concentrations averaged 19 ± 10 and 1600 ± 710 mg/kgsoil, respectively. Tank 2 benzene and indane soil concentrations averaged 56 ± 26 and 1500 ± 720 mg/kg-soil, respectively. Variability in benzene concentrations is to be expected given the soil handling procedures. The approximately 3× lower initial benzene soil concentrations in Tank 1 vs Tank 2 were consistent with the differences in average dissolved benzene concentrations measured at the lower rows of ports in the two tanks. Consistency in the soil concentration data was also checked by 5130
dx.doi.org/10.1021/es405775q | Environ. Sci. Technol. 2014, 48, 5127−5135
Environmental Science & Technology
Article
Figure 3. Tank 1 data, including dissolved oxygen level changes and effect on emission flux.
Figure 4. Tank 2 data, including dissolved oxygen level changes and effect on emission flux.
comparing Raoult’s Law predicted dissolved benzene concentrations (using the average soil concentrations above) to the averages of dissolved concentrations in the lower ports. The Tank 1 measured and predicted concentrations were 26 ± 10
and 30 mg-benzene/L-H2O, respectively. The Tank 2 measured and predicted concentrations were 87 ± 23 and 98 mgbenzene/L-H2O, respectively. Each chemical was distributed between NAPL, sorbed, and dissolved phases. The measured 5131
dx.doi.org/10.1021/es405775q | Environ. Sci. Technol. 2014, 48, 5127−5135
Environmental Science & Technology
Article
Table 1. Benzene and Indane Mass Flux Results Summary Tank 1 14-day average flux at end of period [% reduction from end of initial low DO period: % reduction from extrapolation of untreated condition] DO condition (mg/ LH20)
days
low DO < 1 med. DO 6.5−8.5 high DO 17−20 low DO 0.9−2.5
0−118 118−196 196−255 255−311
benzene (mg/cm2-d) × 103 3.5 0.74 0.21 1.3
± ± ± ±
0.6 [NA: NA] 0.06 [79%: 69%] 0.08 [94%: 90%] 0.2 [63%: 29%]
indane (mg/cm2-d) × 102 1.5 0.63 0.19 0.80
± ± ± ±
0.2 [NA: NA] 0.09 [59%: 24%] 0.05 [87%: 73%] 0.10 [47%: 0%]
Tank 2 14-day average flux at end of period [% reduction from end of initial low DO period: % reduction from extrapolation of untreated condition] days 0−64 64−119 119−184 184−228
benzene (mg/cm2-d) × 103 13 4.7 1.3 6.1
± ± ± ±
2 [NA: NA] 0.5 [63%: 48%] 0.2 [90%: 81%] 0.7 [52%: 6%]
indane (mg/cm2-d) × 102 2.2 1.6 0.58 0.99
± ± ± ±
0.4 [NA: NA] 0.2 [30%: 9%] 0.06 [74%: 58%] 0.1 [56%: 22%]
Flux from Low-Permeability Layer. As discussed above, emission fluxes Fi [mg-i/cm2-d] from the low- to highpermeability layer were determined by mass balance knowing the effluent flow rate and effluent dissolved hydrocarbon concentrations (Fi = Q [L/d] × effluent dissolved concentration [mg/L]/interface area [cm2]). The results are presented in Figures 3 and 4 for Tank 1 and Tank 2, respectively, and key features are summarized in Table 1. As mentioned previously, this Fi calculation yields an attenuated emission rate under elevated DO conditions, reflecting combined emission and any subsequent reaction between the source and effluent monitoring point. Each experiment was operated similarly in that the emission rate was monitored under deoxygenated conditions until a near-steady “baseline” emission flux was observed. The baseline benzene emission rate was about 4× greater in Tank 2 than Tank 1, but is explained by the differences in averaged initial dissolved concentrations in the low permeability zone discussed above. The ratio of the initial Tank 2/Tank 1 emission rates (3.7) is very similar to the ratio of averaged initial dissolved benzene concentrations (3.3). Increasing the influent DO level to 6.5−8.5 mg/L resulted in about 79% (Tank 1) and 63% (Tank 2) reductions in benzene emission, with the reduction being evident within about a week of DO transport across each tank. The emission reduction in Tank 1 was initially about an order of magnitude, but effluent concentrations rebounded after about 30 days, and then declined again to a level intermediate between the two conditions. Similar behavior was not observed in Tank 2. The subsequent increase in the influent DO level to 17−20 mg/L resulted in about 94% (Tank 1) and 90% (Tank 2) reductions in benzene emission rate relative to the baseline emission, with the increased reduction being evident sooner after the DO influent change than was observed with the increase from deoxygenated conditions to 6.5−8.5 mg/L. Upon return to low DO conditions after 137 days (Tank 1) and 120 days (Tank 2) of oxygenation, the emissions in both tanks rebounded and stabilized similarly to about 63% (Tank 1) and 52% (Tank 2) of the baseline benzene emission. Results also show one- to two-month lags between the end of oxygenation and rebound to the final near-steady emission rates. This has implications for post-treatment performance monitoring at field sites. Theory−projected Emission Flux Reduction without Treatment. As nontreated control experiments were not conducted, theory was used to extrapolate the early time pretreatment data to longer-term emission rates. Mathematical solutions for problems involving dissolved concentrations that are proportional to their soil concentration at any point (as is the case for a relatively soluble NAPL component i partitioning from a relatively insoluble NAPL base and Raoult’s Law partitioning behavior), one-dimensional transport in a semi-
values above with an estimated sorption coefficient was used to calculate the initial percentage mass distribution of indane between NAPL, sorbed, and dissolved phases. The majority of mass (about 70%) was in the NAPL phase, followed by 30% sorbed mass, and
