Oxygenated gasoline release in the unsaturated zone — Part 1: Source zone behavior

Journal of Contaminant Hydrology 126 (2011) 153–166

Contents lists available at SciVerse ScienceDirect

Journal of Contaminant Hydrology journal homepage: www.elsevier.com/locate/jconhyd

Oxygenated gasoline release in the unsaturated zone — Part 1: Source zone behavior Juliana G. Freitas ⁎, James F. Barker Department of Earth and Environmental Sciences, University of Waterloo, 200 University Ave. W., Waterloo, Ontario, Canada N2L 3G1

a r t i c l e

i n f o

Article history: Received 17 October 2010 Received in revised form 4 May 2011 Accepted 6 July 2011 Available online 2 August 2011 Keywords: Ethanol Gasoline Unsaturated zone Capillary fringe

a b s t r a c t Oxygenates present in gasoline, such as ethanol and MTBE, are a concern in subsurface contamination related to accidental spills. While gasoline hydrocarbon compounds have low solubility, MTBE and ethanol are more soluble, ethanol being completely miscible with water. Consequently, their fate in the subsurface is likely to differ from that of gasoline. To evaluate the fate of gasoline containing oxygenates following a release in the unsaturated zone shielded from rainfall/recharge, a controlled field test was performed at Canadian Forces Base Borden, in Ontario. 200 L of a mixture composed of gasoline with 10% ethanol and 4.5% MTBE was released in the unsaturated zone, into a trench 20 cm deep, about 32 cm above the water table. Based on soil cores, most of the ethanol was retained in the source, above the capillary fringe, and remained there for more than 100 days. Ethanol partitioned from the gasoline to the unsaturated pore-water and was retained, despite the thin unsaturated zone at the site (~ 35 cm from the top of the capillary fringe to ground surface). Due to its lower solubility, most of the MTBE remained within the NAPL as it infiltrated deeper into the unsaturated zone and accumulated with the gasoline on top of the depressed capillary fringe. Only minor changes in the distribution of ethanol were noted following oscillations in the water table. Two methods to estimate the capacity of the unsaturated zone to retain ethanol are explored. It is clear that conceptual models for sites impacted by ethanol-fuels must consider the unsaturated zone. © 2011 Elsevier B.V. All rights reserved.

1. Introduction Oxygenates are typically added to gasoline to improve urban air quality or decrease greenhouse gas emissions. Methyl tert-butyl ether (MTBE) was the most common oxygenate in North America, but it is increasingly being replaced by ethanol, usually added to gasoline in fractions of around 10% (E10, often termed gasohol). Brazil has been using 20 to 25% ethanol mixed with gasoline for 30 years. Since fuels are a common source of groundwater contamination, the behavior of ethanol in the subsurface and the potential impacts

⁎ Corresponding author. Tel.: + 1 519 888 4567x35372; fax: + 1 519 746 7484. E-mail addresses: [email protected] (J.G. Freitas), [email protected] (J.F. Barker). 0169-7722/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.jconhyd.2011.07.003

it might have on the fate of hydrocarbon contaminants from gasoline are of concern. Ethanol is completely miscible with water, unlike toxic aromatic hydrocarbons in gasoline, which have low aqueous solubility (Table 1). Due to ethanol's hydrophilic nature, when an E10 mixture contacts groundwater, ethanol will partition preferentially into the aqueous phase (Oliveira, 1997; Powers et al., 2001). Once dissolved, ethanol travels at the groundwater velocity, without being retarded by sorption (Zhang et al., 2006). Interestingly, despite ethanol's high solubility and mobility, it is rarely found in high concentrations at sites contaminated with gasohol (Freitas et al., 2010a; McDowell et al., 2003). It appears that the unsaturated zone and the capillary fringe play an important role in the fate of ethanol in the subsurface, helping explain why spilled ethanol does not appear extensively in groundwater (Dakhel et al., 2003; Freitas et al., 2010a; McDowell et al., 2003).

154

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

Table 1 Physical–chemical properties of the compounds investigated.

Ethanol MTBE Benzene Toluene Ethylbenzene p,m-Xylene o-Xylene 1,3,5-TMB 1,2,4-TMB 1,2,3-TMB Naphthalene

Molecular weighta

Densitya (g/L)

Solubilitya (mg/L)

Henry constanta (atm L/mol)

Vapor pressure (atm)

Cosolvency powerb

46.068 88.148 78.112 92.139 106.165 106.165 106.165 120.191 120.191 120.191 128.171

789.3 735.3 876.5 866.8 862.6 858.1 880.2 861.5 875.8 894.4 1025.3

Infinite 44,000 1780 531 161 171 171 50 57 70 86

0.00507 0.70 5.57 6.60 8.43 7.1 5.51 7.81 5.69 3.43 0.43

0.08d 0.33d 0.1259e 0.04e 0.0126e 0.011e 0.009e 0.0033e 0.0027e 0.0022e 0.000112d

0.3 1.5 1.5 2 2 2 2 2 2 2

Sorption distribution coefficient (Kd)c (L/kg)

0.089 0.088 0.14 0.187 0.128

Sources: a Lide (2008). b Estimated from Corseuil et al. (2004), Freitas and Barker (2008) and Morris et al. (1988). MTBE cosolvency power was determined from laboratory measurements (see Section 3). c Hubbard (1992). Naphthalene solubility obtained from Lide (2008) was corrected from the solid to the subcooled liquid (Schwarzenbach et al., 2003). Average values were adopted for p,m-xylene. d SRC (2010). e Schwarzenbach et al. (2003).

In general, the capillary fringe is defined as the region above the water table where the pores are virtually saturated, but with negative gauge pressure (Berkowitz et al., 2004). The lower limit of the capillary fringe is consistently defined in the literature as the water table; however, different definitions of the upper limit of the capillary fringe have been proposed (see Berkowitz et al., 2004). In this study, the upper boundary of the capillary fringe is defined where water saturation reaches around 95%. Visualization laboratory tests indicated that ethanol tends to accumulate in the upper part of the capillary fringe due to its low density (Capiro et al., 2007; McDowell and Powers, 2003; Stafford et al., 2009). Under these conditions, advective horizontal transport in the capillary fringe is likely (Abit et al., 2008; Berg and Gillham, 2010; Berkowitz et al., 2004; Henry and Smith, 2003; Ronen et al., 1997). Since ethanol acts as a cosolvent, increasing hydrocarbon solubility, it likely would carry high concentrations of hydrocarbons with it in the capillary fringe. Cosolvency and transport in the capillary fringe are addressed in Part 2 (Freitas et al., this issue) and Freitas and Barker (2011). As the concentration of ethanol in the aqueous phase increases, both the surface tension between air and water and the interfacial tension between the aqueous phase and the NAPL (non-aqueous phase) decrease (McDowell and Powers, 2003; Oliveira, 1997). Ethanol surface tension (around 22 mN/m) is approximately three times less than that of water (around 72 mN/m) (Lide, 2008). Some of the consequences of this decrease in surface tension include depression of the capillary fringe and induced flow by capillary pressure gradients (Henry and Smith, 2003; Jawitz et al., 1998; Smith and Gillham, 1999). The decrease in interfacial tension can decrease NAPL entrapment, increase NAPL saturation and make NAPL more mobile (McDowell and Powers, 2003; Oliveira, 1997). Besides the issues of transport in the capillary fringe and change in NAPL distribution, McDowell and Powers (2003)

identified that the unsaturated zone above the capillary fringe will also impact the fate of ethanol. In one and twodimensional experiments simulating E10 spills, they noticed that some of the ethanol partitioned out of the NAPL and was retained in the unsaturated zone. They described that following 400 and 600 mL E10 releases into a 44 cm long column with water at residual saturation, at least 99% of the ethanol was retained in the unsaturated zone. It was hypothesized that ethanol retention in the unsaturated zone with slow release to the saturated zone due to the decrease in surface tension could decrease the impact of ethanol on hydrocarbon biodegradation, as it results in lower ethanol concentrations reaching the saturated zone (McDowell and Powers, 2003). Volatilization and biodegradation can also impact the fate of aromatic hydrocarbons, ethanol and MTBE in the unsaturated zone (Dakhel et al., 2003; Rivett et al., 2011). Volatilization is likely to be less significant for ethanol than for some aromatic hydrocarbons and MTBE, which have higher vapor pressures. Ethanol biodegradation in the unsaturated zone was reported by Dakhel et al. (2003) as a significant process for attenuation of ethanol vapor from an E5 source. Most of the field studies to date (Freitas et al., 2010b; Mackay et al., 2006), have focused on the saturated zone and on the impacts ethanol might have on hydrocarbon biodegradation. To evaluate the fate of ethanol and gasoline compounds after an E10 spill, including both the vadose and saturated zone, a controlled release of E10 was performed at field scale. This is the first controlled field study to evaluate the fate of ethanol in the vadose zone following a gasohol spill. The results of the study are presented in two parts. This paper is focused on the processes in the source zone that control ethanol, MTBE and gasoline hydrocarbon distribution. Simplified methods to assess the potential for ethanol retention in the unsaturated zone are also developed. A companion paper discusses the transport and biodegradation of ethanol and hydrocarbons in the capillary fringe and groundwater.

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

2. Field test 2.1. Methods The field test was conducted in a test cell at Canadian Forces Base Borden, located 80 km northwest of Toronto. The Borden aquifer is an unconfined sand aquifer of glaciofluvial origin, composed of well sorted, fine to medium grained sand (Mackay et al., 1986). Distinct bedding features were identified by Mackay et al. (1986), primarily horizontal and parallel, with some cross bedding and convoluted bedding. The geometric average hydraulic conductivity is 7.2 × 10 − 5 m/s (10 °C), but it can vary spatially by one order of magnitude (Sudicky, 1986). The water saturation profile for Borden sand is shown in Fig. 1. The test cell, which was also used in other studies (Freitas et al., 2010b; Molson et al., 2008), is 7.35 m wide (perpendicular to the direction of groundwater flow) and is isolated on the sides by 7 m deep sheet pilling, but allows natural flow along the axis of the cell and natural variation in the depth to the water table (Fig. 2). The site was covered with a roof to avoid direct recharge. The NAPL mixture released comprised 171 L of API 91–01 gasoline (Prince et al., 2007), 20 L of ethanol (99.9% purity, from Commercial Alcohols Inc.) and 9 L of MTBE (99.8% purity, from Sigma Aldrich). This resulted in an E10 mixture, with 10% ethanol and 4.5% MTBE (volume basis). To aid in the identification of NAPL in soil samples, the E10 mixture was dyed with 0.1 g/L of Oil-Red-O. Some concern has been raised about possible changes of NAPL properties due to the inclusion of dye (Tuck et al., 2003). The surface tension of the gasoline mixture with and without dye was measured by the pendant drop method (VCA 2500, AST Products, Massachusetts) at the Porous Media Laboratory, in the Department of Chemical Engineering, at the University of Waterloo.

Fig. 1. Pressure head–saturation profiles for clean (Mickle, 2005) and gasoline contaminated Borden sand (estimated based on interfacial tension reduction).

155

The surface tension with and without the dye was not significantly different (21.5 ± 0.1 dyn/cm and 21.2 ± 0.2 dyn/cm respectively). The release was made into a trench, 1.5 m wide × 0.8 m in the flow direction and 0.2 m deep. The sides of the trench were supported by steel sheets and the trench sides and bottom were lined with a plastic sheet. On August 21, 2008 (Day 0), the ethanol and the premixed gasoline containing the MTBE were poured into the trench, mixed, the plastic layer removed, and the E10 allowed to infiltrate into the unsaturated zone through the bottom of the trench. The water table was 52 cm below ground surface (bgs) and 32 cm below the bottom of the trench (which was 0.2 m deep). To minimize losses by volatilization, the trench was covered immediately with plywood. To verify the gasoline composition during the period of infiltration, 8 NAPL samples were collected from the trench, one from the initial mixture and the others at intervals of 20 to 50 min. No difference in composition was observed between samples. After 5 h, when all the gasoline mixture infiltrated, the steel sides were removed and the trench was refilled with Borden sand. To evaluate the contaminant distribution in the sand, soil cores were collected close to the source zone 14, 47 and 77 days after the release (cores 1 to 3 in Fig. 2). A 2.8 cm ID aluminum tube was hammered to the desired depth; the tube with the core inside was retrieved and the holes were backfilled with sand or bentonite. The tube wall was drilled every 3 cm and two samples of around 2 mL of soil were collected at each position using a 5 mL syringe with the tip removed (Hewitt, 1996; Schumacher and Minnich, 2000). One sample was used for hydrocarbon analysis and the other was analyzed for oxygenates. The soil samples were inserted into 20 mL pre-weighted VOA vials with 5 mL of solvent. For hydrocarbon analysis the solvent was methylene chloride, and for ethanol, MTBE and TBA (tert-butyl alcohol) analysis the solvent was water with 0.05 mL of a 10% solution of sodium azide. Core 4 (see Fig. 2), 1.5 m long, was collected in a 5.08 cm (2″) aluminum pipe on day 341 for saturated hydraulic conductivity (Ks) measurements. The tube was cut in 10 cm segments and the soil collected, dried and homogenized before triplicate falling head permeameter tests (Reynolds, 2008). Soil gas sampling probes were installed around the source zone, 1 to 2 m apart (Fig. 2), with vertical sampling ports spaced at 20 cm, from 0.2 to 0.8 m bgs. The sampling probes were constructed using 3 mm ID steel tubing with screen positioned at the bottom. Similar designs were applied by Conant et al. (1996) and Thomson and Flynn (2000). Samples were collected using a gastight syringe fitted with Hamilton® miniature inert plug valves, transferred underwater to a 22 mL headspace crimp top vial initially water-filled. The water was displaced by the gas and, still underwater, the vial was capped using a gray butyl, Teflon-faced stopper and then sealed with an aluminum crimp. Samples were stored upside down with the cap and neck under water until analyzed. Preliminary lab tests indicated minor loss (from 14 to 17%) of benzene, toluene, ethylbenzene and p,m-xylenes (BTEX) due to the underwater sample transfer. Soil gas samples were collected 12, 14, 16 and 21 days after the release. The shallow water table (0.2 to 0.7 m bgs) during the test limited the depth from which soil gas samples were retrieved.

156

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

Fig. 2. Field site plan view. Spill trench (orange rectangle), coring locations, vapor sampling probes (gray squares) and multilevel wells for groundwater sampling (blue circles) are indicated. Inferred contamination lateral extent is also presented based on dye ground stain, core results, and contamination in the downgradient groundwater.

For 300 days, the depth to the water table was monitored at 6 hour intervals using pressure transducers installed in four piezometers (see Fig. 2). Surface ground penetrating radar (GPR) was also used to assess the LNAPL infiltration and redistribution (McNaughton, 2011). Soil samples were analyzed for oxygenates and hydrocarbons at the Organic Chemistry Laboratory, Department of Earth and Environmental Sciences, University of Waterloo. For ethanol, MTBE and TBA analyses the vials with soil and water were shaken by hand for 20 min and allowed to settle for 30 min. A 2 mL aliquot of the aqueous solution was transferred to an auto-sampler vial and placed on a 7673A HP Autosampler Gas Chromatograph (GC). The GC was equipped with a flame ionization detector and a 3 m long by 0.318 cm inner diameter column packed with 3% SPI500 on Carbopack B (80/100 mesh). The detection limits for the three compounds were below 0.5 mg/kg. Hydrocarbon (BTEX, trimethylbenzene isomers (TMBs) and naphthalene) analyses were performed by solvent extraction with methylene chloride followed by gas chromatography. The vials with the soil samples and methylene chloride were shaken vigorously (350 rpm) for 18 h and then settled for around 3 weeks. The samples were reweighted to ensure there was no solvent loss during this period. A 0.7 mL aliquot of extraction solvent was placed in a Teflon-sealed autosampler vial and injected into a HP 5890 capillary GC equipped with a 0.25 mm× 30 m long DB5 capillary column with a stationary phase film thickness of 0.25 μm, a HP7673A autosampler, and a flame ionization detector. The method detection limits for the aromatics tested were below 0.05 mg/kg for soil samples. The relative standard deviation of replicated standards was below 10% for all chemicals in the range of concentrations tested. Soil gas samples were analyzed for pentane, hexane, ethanol, BTEX and TPH (total petroleum hydrocarbons) on a Hewlett Packard 5890 gas chromatograph equipped with a split injection port, a HP-624 capillary column, 30 m

long × 0.32 mm I.D. and 1.8 μm film thickness, a PID and a Varian Genesis headspace autosampler. Peak areas were measured by a HP 3392A integrator and an external standard method of calibration was used. TPH, Fraction 1 (C5-C10) was determined by summing all the peak areas from retention time 1.9 to 15.05 min and dividing by the average response factor of hexane and toluene. Calibration standards were prepared by spiking the same 22 mL vials with methanolic stock solutions and sealed. Detection limits were found to range from 2 to 5 μg/L. 2.2. Results and discussion The resulting distribution of residual LNAPL and oxygenates in the field test was assessed by direct methods, such as coring, and also by the groundwater concentrations measured downgradient (Freitas et al., this issue). Soil cores 1–3 were collected in the source zone for chemical analysis (Fig. 2). The number of cores and position was selected to minimize disturbance in the source. The first core was obtained 14 days after the release, while the water table was at 67 cm bgs. During the period between the release and the first coring the water table had generally moved downwards (Fig. 3). Between the first and the second coring events (day 14 to day 47), the water table oscillated between 70 and 20 cm bgs. The water table was at 53 cm bgs when core 2 was collected. From the time of the second to the last coring (day 117) the water table continued oscillating, and was higher (36 cm bgs) by the time core 3 was collected. The results from the surface GPR surveys were consistent with the LNAPL distribution derived from these core results (McNaughton, 2011). 2.2.1. NAPL distribution Twenty-four days after the E10 release there was a major precipitation event and the water table rose to around 20 cm bgs (Fig. 3). In the presence of gasoline, the capillary fringe is

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

157

Fig. 3. Water table position and coring dates (core 4 was collected on day 305, not shown).

expected to be around 15 cm thick, with the transition zone extending another 20 cm (Fig. 1). Therefore, with the water table rise to 20 cm bgs, the mobile gasoline moved upwards close to the ground surface and a stain from the gasoline red dye was left on the ground. From one corner of the trench a finger of stain extended more than 0.5 m upgradient (Fig. 2). This clearly shows that the contaminant distribution was not symmetrical. Subsequent excavation found a tree root at about 35 cm depth extending the length of the gasoline stain. After that first event when the water table rose to very shallow depths, the water table went back down to 60 cm bgs, and then in December, 2008 (after about day 85) it moved up and stayed around 45 cm bgs. A darker stain was noticed on the ground surface around the edge of the trench on December 03, 2008 possibly caused by differences in packing when the trench was filled. In January, 2009, the water table rose slightly, reaching about 30 cm bgs. The dye stain on the ground became darker, following the water table oscillations. The evolving stain pattern in the ground is shown in Fig. 2. A good correlation between the lateral extent of gasoline occurrence inferred from core data, surface stain extent transverse to groundwater flow and the lateral extent of the MTBE plume observed at Row A (Freitas et al., this issue) indicates that the stain left on the ground represents the lateral source distribution (Fig. 2). By inference, the longitudinal distribution of the source is taken to be represented by the distribution of the surface stain. Based on the surface stain the source area was estimated to be 4.2 m 2. The GPR surveys indicate similar results, and also identified NAPL upgradient of where the stain was noticed (McNaughton, 2011). The results for ethanol, MTBE and total hydrocarbons in the sand, based on the soil cores, are presented in Fig. 4. In core 1, the highest hydrocarbon concentrations were found around 32 to 41 cm bgs, about 26 cm above the water table, consistent with the expectation of accumulation on top of a

depressed capillary fringe (see Fig. 1). In core 2, hydrocarbons were smeared from 12 to 30 cm bgs, a consequence of the water table oscillations before the coring. In the third core higher concentrations were measured at much shallower depths, from 12 to 2 cm bgs, indicating that the water table rise had mobilized the gasoline upwards. In all soil cores, the highest hydrocarbon concentrations were always located at around 24 cm above the water table at the time of coring, consistent with the estimated top of the capillary fringe. The presence of NAPL in the soil samples was assessed by two methods. The first method, from Feenstra et al. (1991), calculates the maximum concentration that could be present in the dissolved and sorbed phases for each sample for toluene, ethylbenzene and o-xylene. If the measured concentration was higher than the calculated value, then NAPL was likely present in the sample. In the second method, the ratios of selected hydrocarbons in soil samples to the ratios in the released gasoline were calculated. The concentrations of organic compounds in the gasoline released were determined previously (Prince et al., 2007), and so the initial ratios of each compound to the total BTEX + TMBs + Naphthalene were calculated (termed expected ratio). Then, the same ratios were calculated for each soil sample (termed measured ratio). The plots of the measured ratios over the expected ratios are presented in Fig. 5 for core 1. If NAPL is present in the soil sample that phase dominates the total mass of each hydrocarbon found in the soil and so the measured ratio tends to be close to the ratio in the initial gasoline, i.e., the ratio values tend to one. Where NAPL is not present, the values deviate significantly from one, with the ratios for the compounds with higher effective solubility (like benzene and toluene) being higher than one, as they are present in higher concentrations in the dissolved phase. The results are consistent with the results obtained from the method of Feenstra et al. (1991). The results also show an enrichment in

158

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

Fig. 4. Concentration in soil cores (mg/g wet soil). Depth is on the y-axis (cm bgs). Hydrocarbon is the sum of BTEX, TMBs and naphthalene. Vertical arrows indicate the range of water table oscillation from the release to the day of the core.

naphthalene and loss of benzene close to the ground surface, which indicates losses of the more volatile compounds by volatilization.

For the samples where there was indication of NAPL presence, its saturation (SNAPL) was calculated assuming equilibrium between phases: sol

SNAPL ¼

si · ρb −Ci;dis · xi ðn · Sw þ Kdi · ρb Þ vi;NAPL · n · ρi

ð1Þ

where si is the dry soil concentration of compound i; ρb is the soil bulk density (1.81 g/cm3; Mackay et al., 1986); C i,soldis is the solubility of compound i in the aqueous phase; xi is molar fraction of compound i in the NAPL, assumed to be the same as in the original NAPL; n is the porosity (0.33; Mackay et al., 1986); Sw is water saturation; Kdi is the soil–water partitioning coefficient for compound i; vi, NAPL is the volume fraction of compound i in the NAPL and ρi is the density of compound i. The relevant properties of the compounds are presented in Table 1. In general, the values are within the range expected for residual saturation in the unsaturated zone (3% to 7%; Parker et al., 1995) and saturated zone (12%; Oliveira, 1997). It appears that the NAPL became smeared over time, likely due to the oscillations in the water table (Fig. 6). In core 3 NAPL was also present below the water table at the time of coring, reflecting the trapping of gasoline residuals by capillary forces below a rising water table (Illangasekare et al., 2005).

Fig. 5. Measured ratio of Ci/Ct over the expected ratio of Ci/Ct. Deviation from 1 indicates the absence of NAPL (data from soil core 1, obtained on 04-Sept-08).

2.2.2. MTBE and ethanol distribution The vertical distribution of MTBE in the soil cores was similar to that of the hydrocarbons, while ethanol appeared at higher concentrations where hydrocarbons were at lower concentration. This difference is likely a consequence of the much higher affinity of ethanol for soil water. As the E10 migrated in the unsaturated zone, oxygenates partitioned to the

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

159

Fig. 7. Ethanol pore-water concentration (mass fraction). See text for details. Fig. 6. Calculated NAPL saturation profile (average of values determined based on toluene, ethylbenzene and o-xylene) (Eq. (1)). Error bars represent 2 standard deviations. Water table position at each core date is color coded on the right side.

pore-water, but most of the MTBE mass was still retained within the NAPL reaching the capillary fringe, while the ethanol was lost to the soil water. To confirm this hypothesis, the distribution of MTBE between the NAPL and the aqueous phase was estimated using the soil cores results. The maximum MTBE concentration in the aqueous phase is the equilibrium concentration when the NAPL contains 4.5% MTBE. This concentration was determined in laboratory as approximately 1800 mg/L, which is consistent with values estimated based on Raoult's Law. Using this concentration and the water saturation profile estimated for each core, the maximum mass of MTBE that could be present in the aqueous phase was obtained. In core 1, at least 76% of the MTBE was present in the NAPL, and in soil cores 2 and 3 a minimum of 60% of the MTBE mass was still within the NAPL. Further evidence that confirms the hypothesis that MTBE remained in the NAPL while ethanol partitioned to the aqueous phase, is the result of one NAPL sample recovered from the site 21 days after the release, at 40 cm bgs, through a steel tube previously installed for vapor monitoring (Fig. 2). The recovered sample was almost completely depleted of ethanol (only 0.02% compared to 10% in the initial E10), while it still contained 3% MTBE (4.4% in the initial mixture). Ethanol pore-water concentrations (ce, q) were calculated assuming equilibrium partitioning between the NAPL and the aqueous phase (Appendix A). In all cores most of the ethanol (around 85%) was present in the aqueous phase, contrary to what was estimated for MTBE. The aqueous ethanol concentration profiles in soil cores 1 and 2 were similar (Fig. 7), with peak concentrations of around 80 g/L (8%) occurring around 15 to 25 cm deep. The profile in the third soil core is similar to the others from 10 cm bgs and deeper. From 10 cm bgs to the ground surface, the profile is significantly different, with much higher concentrations,

reaching values around 14%. While the change in position of the maximum values can be attributed to the shallower water table when the core was collected, a two times increase in pore-water concentration seems unlikely. A rise of the water table height could cause an increase in ethanol concentration in the capillary fringe as the ethanol present in the unsaturated zone would dissolve into the capillary fringe and remain on it as the water table moves up. However, in this case the total ethanol mass would have to remain the same. However, the overall increase in ethanol concentration in core 3 indicates a significant increase in the ethanol mass over the whole depth (Fig. 4). The observed discrepancy could be a consequence of a heterogeneous source zone, as discussed previously. The ethanol concentrations are lower than expected based on the preferential partitioning of ethanol to the aqueous phase (see Section 3). Several reasons for the relatively low ethanol concentrations measured in the soil can be considered. First, all the cores were collected outside the trench area (Fig. 2) and lateral spreading of ethanol due to diffusion and surface tension gradients could cause a decrease in ethanol concentrations (Henry and Smith, 2003). Ethanol biodegradation and volatilization could also result in lower concentrations in the field (Dakhel et al., 2003). Microcosm tests using Borden aquifer material and groundwater, with excess oxygen and without nutrient addition, indicated no ethanol loss after 64 days when ethanol concentrations were 1.5% (Araujo, 2000), suggesting that biodegradation would not have a great impact on ethanol concentrations in the source zone at early times. Volatilization is also not expected to be a significant process for ethanol mass loss due to ethanol's relatively low Henry constant (Table 1). This was evident in the soil gas, where samples commonly had high MTBE concentrations (up to 14,000 μg/L at 20 cm bgs, 65 cm away from the release trench) while no ethanol was detected (b5 μg/L) in any sample. Ethanol concentrations higher than 10%, as were estimated, are expected to increase the solubility of the gasoline hydrocarbons due to cosolvency effects (Corseuil et al., 2004).

160

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

Hydrocarbon solubilities considering ethanol cosolvent effects (Cim) were calculated using the log-linear model (Eq. (2); Heermann and Powers, 1998; Morris et al., 1988), where Ci,soldis is the hydrocarbon solubility in pure water, f c is the volume fraction of the cosolvent (ethanol) and σ is the cosolvency power, for which values were obtained from previous studies (Table 1). m

sol

log Ci ¼ log Ci;dis þ σ · f

c

ð2Þ

Based on Eq. (2), we can estimate the increase in hydrocarbon solubilities at the highest ethanol concentration (14%). For example, benzene solubility would increase from 1780 mg/L to 2890 mg/L (effective solubility going from 26.7 to 43.4 mg/L). More hydrophobic compounds have even greater relative increases; for example, o-xylene solubility would increase from 205 to 391 mg/L (effective solubility going from 6.7 to 13 mg/L). The NAPL saturation was reassessed considering the new solubility values, but the changes were always less than 0.05%. The distribution of phases in the ternary diagram inferred from soil samples also reveals some information about the processes that took place in the unsaturated zone (Fig. 8). The presence of some samples in regions with the total ethanol fraction higher than 10% cannot be explained by simple dilution with water. If an E10 gasoline mixture was being mixed with water, the system composition would follow the dashed line shown in Fig. 8. The presence of samples above this line indicates that, after the gasoline mixture contacted water and two phases were formed (aqueous and

oleic), the two phases were transported differently in the subsurface, resulting in system compositions that are different than what would be expected based solely on mixing with water. This is consistent with our conceptual model that the aqueous phase with high ethanol was retained in the unsaturated zone while the gasoline-MTBE phase continued traveling downwards. 2.2.3. Potential for ethanol transport and redistribution Ethanol was found at higher concentration in the soil at about the same depths in all soil cores (around 15 to 25 cm bgs in cores 1 and 2 and from 5 to 25 cm bgs in core 3). On the first two core dates, this range is clearly in the unsaturated zone, above the top of the capillary fringe. The depth where ethanol was found in higher concentration corresponds to the depth of the trench (20 cm), indicating that most of the ethanol was retained at a depth close to where it was released. Based on the groundwater data (Freitas et al., this issue) the lateral distribution of ethanol was similar to that of MTBE. Ethanol presence at high concentrations at such high elevations outside the trench area suggests that ethanol (but not gasoline NAPL) moved laterally in the unsaturated zone. This behavior has been observed in 2D lab tests (Henry and Smith, 2003; Yu et al., 2009) and could have resulted from ethanol diffusion in the aqueous phase or pressure gradients created by changes in surface tension (Henry and Smith, 2003). The retention of ethanol at the same depth indicates that oscillations in the water table had minor impact on the ethanol distribution. Ethanol mobilization during water table oscillation could be minimized by ethanol's lower surface

Fig. 8. Gasoline–ethanol–water ternary diagram (adapted from Oliveira, 1997 as reproduced by Powers et al., 2001) with the phase compositions inferred from soil cores. Also shown are the calculated composition trend for E10 gasoline being diluted by mixing with water (dashed line) and model results (solid curve) at 300 min (see Section 3). Arrows indicate increasing depth.

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

tension, lower density and higher viscosity in comparison to pure water. The lower surface tension can cause a lower aqueous saturation in the ethanol-rich zones in comparison to the adjacent zones. This can result in a lower hydraulic conductivity, which would direct groundwater flow around the ethanol-rich zone. Also, solutions with high ethanol concentration would have lower density and higher viscosity. Although viscosity and density values for pure ethanol deviate only around 20% from pure water, solutions with intermediate fractions of ethanol have more than 2 times higher viscosity than water (Ageno and Frontali, 1967; Belda et al., 2004). For ethanol fractions ranging from 30% to 60%, the changes in density and viscosity cause a decrease in hydraulic conductivity by around 65%. Besides the changes in hydraulic conductivity, displacement of ethanol by water is subject to fingering due to the unequal viscosities and densities, which will cause a decrease in the mass of ethanol that is carried by the water (Dullien, 1979). Although it was not possible to discern the occurrence of fingering in the field, it likely happens since, based on the viscosity and density differences between pure water and ethanol-water mixtures, downward displacement of ethanol by water will always be unstable while upward displacement might be stable, depending on the velocity of the displacement and on the ethanol fraction in the ethanol-water phase (Kueper and Frind, 1988). Ethanol could also be mobilized by recharge. Since the research site was under a roof, the effect of recharge on the source zone was not evaluated. This is not unrealistic as fuel stations and distribution terminals have recharge minimized by the presence of roofs and pavement. Also, McDowell and Powers (2003) indicated that ethanol is not readily flushed by recharge; in a column test, several recharge events were necessary to mobilize ethanol retained in the unsaturated zone. The retention of ethanol even in the presence of recharge is likely a consequence of instabilities arising from the density difference between ethanol and water (Kueper and Frind, 1988) and differences in surface tension. The potential for ethanol being transported laterally downgradient by groundwater flow can be accessed by comparing the concentration profile with the effective hydraulic conductivity for the aqueous phase (Kq) as shown in Fig. 9. To estimate the relative permeability (kr) based on the water saturation, the combined van Genuchten–Mualem model (van Genuchten, 1980) was applied utilizing parameters from Mickle (2005). To incorporate the effect of ethanol on the hydraulic conductivity, experimental data of density and viscosity as a function of ethanol concentration from Belda et al. (2004) was used to obtain a polynomial equation of the relative permeability (kr) as a function of ethanol mass fraction ( fe) (Eq. (3)). The relative permeability function reaches its minimum of 0.3 at around 40% ethanol. −8

4

kr ¼ 8 · 10 · fe −2 · 10 þ 0:9846

−5

3

2

· fe þ 0:0015 · fe −0:0538 · fe

ð3Þ

It can be seen from Figs. 7 and 9 that the high ethanol concentrations were observed in regions where the effective hydraulic conductivity was zero, caused mainly by the low aqueous saturation. Since the high ethanol concentrations were always located in regions with extremely reduced

161

Fig. 9. Effective hydraulic conductivity for the aqueous phase (m/s). Dashed line indicates Kq which considers effects of solution density and viscosity.

hydraulic conductivity, the additional reduction caused by the ethanol changes in viscosity and density was not significant. The presence of ethanol in regions where the hydraulic conductivity is essentially nil supports the inference that most of the ethanol is not being transported downgradient, and it is being retained in the unsaturated zone in the release region. Soil core 3, which was collected when the water table was at a higher elevation, showed a high percentage of ethanol, up to 14%, more than 20 cm bgs (Fig. 7), where the hydraulic conductivity starts to increase. It is possible for ethanol to be transported laterally downgradient in the capillary fringe under these conditions. However, the low density and high viscosity due to ethanol's presence in the aqueous phase could still delay its downgradient transport. Also, if ethanol is in regions of lower hydraulic conductivity as indicated in 2-D laboratory tests (Freitas et al., this issue), a further delay in ethanol transport is possible. The effect of water table oscillation to the downgradient transport of ethanol is assessed in Part 2 (Freitas et al., this issue). The retention of ethanol in the unsaturated zone could also be facilitated by heterogeneities in the porous medium. Although fairly homogeneous, the Borden aquifer has horizontal bedding features, with the hydraulic conductivity varying by one order of magnitude (Mackay et al., 1986). In the unsaturated zone, a layer with smaller pore size, and consequent higher capillary pressure, would have higher water saturation, and therefore would be more likely to retain a higher fraction of the ethanol, as observed in laboratory tests (Freitas et al., this issue). The hydraulic conductivity measurements from the E10 spill site (Freitas and Barker, 2011) vary from 4 × 10 − 5 to 2 × 10 − 4 m/s. However there is no clear evidence of a significant contrast in hydraulic conductivity at the shallower depths, although thin or discontinuous lenses could have been missed.

162

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

3. Potential for ethanol extraction from NAPL into the pore-water As gasoline with ethanol infiltrates in the unsaturated zone, above the capillary fringe, it will contact the porewater retained in the soil by capillarity. Ethanol will tend to partition out of the gasoline into the water, in a liquid–liquid extraction process. When ethanol partitions from the gasoline into the groundwater, water drainage occurs, due to two factors. First, the volume of the aqueous phase, now composed of water and ethanol, will increase. Since the volume of an aqueous phase that can be retained in the pore space is limited, the increase in volume alone results in drainage. Second, the increase in ethanol concentration in the aqueous phase results in a decrease of surface tension (McDowell et al., 2003), and consequent decrease in capillary pressure and water saturation (Kueper and Frind, 1991; Leverett, 1941; Smith and Gillham, 1999). Again, drainage results. The volume of ethanol retained in the unsaturated zone above the capillary fringe is not easily established, and it will depend on several factors. Two methods to estimate ethanol retention were developed and tested: a one-dimensional (1-D) mathematical model and a simplified equation. 3.1. Method 1: 1-D mathematical model 3.1.1. Model description First, a simplified mathematical model of the system was defined (Fig. 10). It is 1-D, with E10 gasoline flowing downward. The soil type was defined as Borden sand, which has been well characterized (Mackay et al., 1986; Sudicky, 1986). The water saturation profile was estimated assuming that the

water saturation is a function of the capillary pressure between gasoline and water (Lenhard and Parker, 1990), scaling the Borden sand capillary pressure–saturation curve considering a decrease in surface tension from 72 dyn/cm (clean water surface tension) to 23 dyn/cm (water–gasoline interfacial tension), as presented in Fig. 1. This profile was assumed to be constant in time, with the water table at the bottom of the column. This simplification does not consider a transient water saturation profile, as would be expected based on the dependence of the surface tension with ethanol concentration. It was assumed that equilibrium between the NAPL and the aqueous phase is reached within each layer of height dz. Based on the pseudo-ternary diagram presented in Oliveira (1997), Eq. (4) was developed to estimate the ethanol content in the aqueous phase (ce,q), defined as the ratio between the ethanol mass in the aqueous phase and the total aqueous phase mass, as a function of the fraction of gasoline ( fg) and the fraction of ethanol ( fe) in the system (mass base). To derive the equation, the coordinates of each point in the ternary diagram were transformed into a Cartesian coordinates system (Freitas, 2009). The applicability of Eq. (4) was verified by comparing eight experimental values from Oliveira (1997) to values calculated using Eq. (4). The relative error ranged from 8.9% to − 18%, with a mean of 6.7%. ∘

1−fg − fe cos 60 1 ∘ ¼ þ cos 60 ce;q fe

ð4Þ

MTBE was also included in the model, assuming that its partitioning to the aqueous phase follows Raoult's law. Based on the ternary diagram for water, ethanol and MTBE by Ashour (2005), it was concluded that MTBE has a greater

Fig. 10. Conceptual model to estimate ethanol and MTBE retention in the unsaturated zone. Subscripts N and q refer to NAPL and aqueous phase respectively; subscripts e and M refer to ethanol and MTBE components.

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

affinity for ethanol than for the water. The mole fraction of MTBE in the water phase increased from 0.78% to 3.94% while the ethanol fraction in the water went from 0 to 13%. A mole fraction of ethanol N0.2 resulted in a single phase. The cosolvent effect of ethanol on MTBE solubility was included using a log-linear model (Eq. (2)). The MTBE aqueous solubility (66 g/L) and cosolvency power (0.3) were estimated based on laboratory tests (data not shown) where the concentrations of MTBE and ethanol in water equilibrated with mixtures of API-91-01 gasoline, ethanol and MTBE were determined. Ethanol transport in the aqueous phase due to diffusion and the surface tension gradient was not considered. Also, it was assumed that ethanol and water volumes are cumulative, while Oliveira (1997) showed some volume reduction (a maximum of 3.5%, when the ethanol mass fraction was around 23%) when ethanol was added to water. The solution scheme is presented in Appendix B, and the calculations were performed in a spreadsheet. The geometry of the modeled system mimicked the field test described in Section 2. In the vertical direction, twenty layers of dz = 1.8 cm were used, totaling 36 cm, corresponding to the distance between the bottom of the release trench and the water table. We recognize that if ethanol reaches the capillary fringe it would accumulate due to density effects and would be transported by advection (Freitas and Barker, 2011; Stafford et al., 2009). However, these processes were not included in this simplified model, which is, therefore, more adequate to simulate the processes above the top of the capillary fringe. The source area was defined as 1.7 m 2, corresponding to the trench area plus 10 cm in each horizontal direction. The NAPL flux was set as 23.5 L m − 2 h − 1 for 300 min and then decreased to zero. The flux was calculated from the time required in the field

163

to infiltrate the 200 L of E10. Porosity was defined as 0.33, representative of Borden sand (Mackay et al., 1986). The calculations were done for a total time of 1500 min, to ensure steady state was reached. 3.1.2. Results from method 1 Most of the ethanol (maximum ethanol concentration, 67%) remained in the top layers above the capillary fringe (Fig. 11a) over the 1500 min. Below 45 cm bgs ethanol concentrations were close to zero. Where ethanol concentrations were higher, MTBE was also found at higher concentrations, due to cosolvency (Fig. 11b). Unlike ethanol however, MTBE was present throughout the profile with concentrations equal to and often greater than its effective solubility determined in the absence of ethanol. The volume of ethanol retained per layer at the steady-state condition is shown in Fig. 11c. All the ethanol mass is in the aqueous phase, and only 15% of the ethanol mass is below 40 cm bgs, where the water saturation is higher and some horizontal transport would be anticipated (Fig. 11a). Only 7% of the MTBE is in the aqueous phase (Fig. 11c), the remaining 93% left the model domain (reached the depth of 55 cm) still in the NAPL. This result is consistent with the ethanol and MTBE distribution in soil cores described in Subsection 2.2.3, where most of the ethanol was retained above the capillary fringe while most of the MTBE was within the NAPL. Therefore, although several simplifying assumptions were made, the model showed that the unsaturated zone above the capillary fringe has the potential to retain most of the ethanol mass released in this experiment in the absence of recharge. The results from the model at 300 min, corresponding to the end of injection, were plotted on the ternary diagram (Fig. 8). The model results did not match well with the field

Fig. 11. Simulation results. (a) Ethanol volume fraction in the aqueous phase and the calculated water saturation profile (dashed line). (b) MTBE concentration (mg/L) in the aqueous phase. The dotted vertical line represents the maximum MTBE concentration expected in the absence of ethanol. (c) Profile of ethanol and MTBE retained in the aqueous phase after 1500 min, as percentage of the total released. The depth of 20 cm bgs corresponds to the bottom of the trench and the water table is at 55 cm bgs. See Section 3 for details.

164

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

data, likely because the composition was modeled after 300 min, just when the gasoline infiltration stopped, while the cores were taken at later times, after the gasoline had drained farther and after the influence of water table oscillations. This explains why the gasoline fraction is higher in the model results than what was found in the later cores. Also, the model does not consider gasoline retention in the unsaturated zone, which may contribute to the poor match. Even though the model results do not match the field data, the model supports the interpretation that overall ethanol fractions in water higher than 10% are produced when ethanol is retained in the unsaturated zone while the oleic phase migrates downward. Even though the magnitude of the concentrations measured in the field do not match the simple model values, the pattern of ethanol concentration with depth seems to agree with the model expectations. 3.2. Method 2: Simplified equation A second simplified method can be used to estimate the volume of ethanol that can be retained in the unsaturated zone. Based on the phase behavior as defined by the ternary diagram (Fig. 8), when the mixture spilled has an ethanol mass fraction less than 70%, the maximum ethanol content in the aqueous phase (ce, q) will always be around 70% (defined by the plait point on the ternary diagram). For spills where the ethanol fraction in the fuel is higher than 70%, the system is composed of a single phase and aqueous concentrations as high as the concentration injected can be anticipated. Therefore, the maximum volume of ethanol that can be retained in the unsaturated zone (Vemax) can be approximated by Eq. (5), where A is the spill area, h is the distance between the spill and the top of the capillary fringe, n is porosity and S¯w is the average water saturation considering a decrease in surface tension due to the presence of ethanol. Ve max ¼ A · h · n · S¯w · Ce;q

ð5Þ

Application of Eq. (5) results in a prediction of the maximum volume of ethanol that could be retained in the unsaturated zone, assuming that equilibrium between phases is reached during gasoline infiltration through the unsaturated zone. However, in some conditions this maximum capacity might not be attained. For example, heterogeneities in the unsaturated zone (with preferential flow through fractures, macropores or high permeability zones) could affect the volume of porous media that is contacted by the infiltrating NAPL, limiting the applicability of this estimate. Therefore, Eq. (5) provides an estimate of the maximum volume of ethanol that can be retained. To illustrate how this estimate can be used, Eq. (5) was applied to the controlled release of E10 at the Borden aquifer described in Section 2. The values adopted were 1.7 m 2 for the spill area, 20 cm as the distance to the top of capillary fringe, porosity of 33% and average water saturation of 30%. The maximum ethanol volume that could be retained in the unsaturated zone based on Eq. (5) is 25 L. The volume of ethanol released was 20 L. Therefore the unsaturated zone had the potential to retain all the ethanol released. This is in good agreement with the field findings and also with the previous model results. This also indicates

that if a spill in this setting contained more than 25 L of ethanol a portion of the ethanol would reach the top of the capillary fringe. For example, in a 300 L E10 spill, at least 5 L of the ethanol released would reach the capillary fringe, with a maximum of 25 L retained in the unsaturated zone. The same rationale can also be applied to other compounds, replacing ce, q by the effective solubility of that compound. The limiting factor for the retention in the unsaturated zone will be the solubility of the compound. For example, MTBE is also a very soluble compound. However, the MTBE concentration that could be retained in the unsaturated zone pore-water for the E10 released at Borden, which contained 4.5% MTBE, is around 3 g/L, or 0.3%. This results in a maximum of 0.1 L that could be retained in the unsaturated zone, of a total of 9 L. The lower retention of MTBE in the unsaturated zone was seen in the controlled E10 release, where most of the MTBE reached the capillary fringe with the NAPL. Therefore, in order to have a significant retention of MTBE in the unsaturated zone, the unsaturated zone must have a very large retention capacity, likely only in sites with a deep water table. 4. Conclusions and implications After an E10 release most of the ethanol was retained in the unsaturated zone, above the capillary fringe. This retention was caused by ethanol partitioning out of the NAPL to the pore-water, in regions where transport is limited due to the low water saturation and so low hydraulic conductivity. Ethanol retention was significant even with a thin (around 35 cm) unsaturated zone. In contrast, most of the MTBE and hydrocarbons were transported downwards within the fuel NAPL phase. Based on soil core results, oscillations in the water table did not have a major effect on the distribution of ethanol that was retained above the capillary fringe. One possible reason for that is a decrease in the effective hydraulic conductivity of zones rich in ethanol due to decreased surface tension and therefore decreased aqueous saturation. Also, flow instabilities due to differences in density and viscosity might have interfered. Being in zones of low effective hydraulic conductivity, ethanol remained in the unsaturated zone above the capillary fringe for more than 100 days. The effect of direct recharge on the source was not evaluated. We anticipate that ethanol flushing by recharge will also be affected by decreased hydraulic conductivity and flow instabilities. Research aiming to quantify the extent of ethanol flushing from the unsaturated zone following recharge would be important for a comprehensive understanding of ethanol behavior in the unsaturated zone. The extent of ethanol retention in the unsaturated source zone is dependent on the amount of water available between the release and the capillary fringe. Two methods to estimate the retention of ethanol in the unsaturated zone were developed based on equilibrium partitioning between the NAPL and the aqueous phase. One is a 1-D mathematical model and the other is a simplified equation. The results from these estimates were consistent with the field results. Even in settings where the unsaturated zone does not have capacity to retain most of the ethanol present in the fuel, the retention of ethanol in the unsaturated zone is likely to

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

cause the composition of the fuel reaching the capillary fringe to vary in time. Initially, the mixture will be more depleted in ethanol due to ethanol retention. As the ethanol retention in the unsaturated zone approaches its maximum capacity, the concentration of ethanol in the fuel-NAPL increases. The variability in the ethanol content of the fuel-NAPL that reaches the saturated zone will result in a heterogeneous distribution of ethanol in the source. The retention of ethanol in the unsaturated zone evident in this controlled field release shows that the unsaturated zone must be considered in monitoring strategies and in conceptual and numerical models for ethanol-fuel impacted sites. The consequences for transport, fate and remediation are being addressed in Part 2 (Freitas et al., this issue) and in ongoing studies. Supplementary materials related to this article can be found online at doi:10.1016/j.jconhyd.2011.07.003. Acknowledgments The authors would like to acknowledge the financial support of the American Petroleum Institute, the NSERC CRD program, the Canadian Petroleum Products Institute, Water and Earth Science Associates Ltd. (WESA), Conestoga Rovers and Associates (CRA) and the Ontario Ministry of Environment. Juliana Freitas was supported by a scholarship from the Brazilian Government (CAPES, Brazil). Analyses were performed by Marianne Vandergriendt and Shirley Chatten at UW. References Abit, S.M., Amoozegar, A., Vepraskas, M.J., Niewoehner, C.P., 2008. Solute transport in the capillary fringe and shallow groundwater: field evaluation. Vadose Zone J. 7, 890–898. Ageno, M., Frontali, C., 1967. Viscosity measurements of alcohol–water mixtures and the structure of water. Proc. Natl. Acad. Sci. U.S.A. 57, 856–860. Araujo, D.B., 2000. Effect of fuel ethanol on subsurface microorganisms and its influence on biodegradation of BTEX compounds. MSc thesis, Department of Biology, University of Waterloo. Ashour, I., 2005. Liquid–liquid equilibrium of MTBE plus ethanol plus water and MTBE plus 1-hexanol plus water over the temperature range of 288.15 to 308.15 K. J. Chem. Eng. Data 50, 113–118. Belda, R., Herraez, J.V., Diez, O., 2004. Rheological study and thermodynamic analysis of the binary system (water/ethanol): influence of concentration. Phys. Chem. Liq. 42, 467–479. Berg, S., Gillham, R.W., 2010. Studies of water velocity in the capillary fringe: the point velocity probe. Ground Water 48, 59–67. Berkowitz, B., Silliman, S.E., Dunn, A.M., 2004. Impact of capillary fringe on local flow, chemical migration, and microbiology. Vadose Zone J. 3, 354–548. Capiro, N.L., Stafford, B.P., Rixey, W.G., Bedient, P.B., Alvarez, P.J.J., 2007. Fuelgrade ethanol transport and impacts to groundwater in a pilot-scale aquifer tank. Water Res. 41, 656–664. Conant, B.H., Gillham, R.W., Mendoza, C.A., 1996. Vapor transport of trichloroethylene in the unsaturated zone: field and numerical modeling investigations. Water Resour. Res. 32, 9–22. Corseuil, H.X., Kaipper, B.I.A., Fernandes, M., 2004. Cosolvency effect in subsurface systems contaminated with petroleum hydrocarbons and ethanol. Water Res. 38, 1449–1456. Dakhel, N., Pasteris, G., Werner, D., Hohener, P., 2003. Small-volume releases of gasoline in the vadose zone: impact of the additives MTBE and ethanol on groundwater quality. Environ. Sci. Technol. 37, 2127–2133. Dullien, F.A.L., 1979. Porous Media–Fluid Transport and Pore Structure, second ed. Academic Press, San Diego. Feenstra, S., Mackay, D.M., Cherry, J.A., 1991. A method for assessing residual NAPL based on organic chemical concentrations in soil samples. Ground Water Monit. Rem. 11, 128–136.

165

Freitas, J.G., 2009. Impacts of Ethanol in Gasoline on Subsurface Contamination. PhD thesis, Department of Earth Sciences, University of Waterloo. Freitas, J.G., Barker, J.F., 2008. Sampling VOCs with porous suction samplers in the presence of ethanol: how much are we losing? Ground Water Monit. Rem. 28, 83–92. Freitas, J.G., Barker, J.F., 2011. Monitoring lateral transport of ethanol and dissolved gasoline compounds in the capillary fringe. Ground Water Monit. Rem. 31, 95–102. Freitas, J.G., Fletcher, B., Aravena, R., Barker, J.F., 2010a. Methane production and isotopic fingerprinting in ethanol fuel contaminated sites. Ground Water. doi:10.1111/j.1745-6584.2009.00665.x. Freitas, J.G., Mocanu, M.T., Zoby, J.L.G., Molson, J.W., Barker, J.F., 2010b. Migration and fate of ethanol-enhanced gasoline in groundwater: a modelling analysis of a field experiment. J. Contam. Hydrol. doi:10.1016/j. jconhyd.2010.08.007 Freitas, J.G., Doulatyari, B., Molson, J.W., Barker, J.F., this issue. Oxygenated gasoline release in the unsaturated zone — Part 2: Downgradient transport of ethanol and hydrocarbons. Heermann, S.E., Powers, S.E., 1998. Modeling the partitioning of BTEX in water-reformulated gasoline systems containing ethanol. J. Contam. Hydrol. 34, 315–341. Henry, E.J., Smith, J.E., 2003. Surfactant-induced flow phenomena in the vadose zone: a review of data and numerical modeling. Vadose Zone J. 2, 154–167. Hewitt, A.D., 1996. Effect of collection and handling practices on concentrations of volatile organic compounds detected in soil subsamples. In: Wang, W., Schnoor, J.L., Doi, J. (Eds.), Volatile Organic Compounds in the Environment. : ASTM STP 04-012610-16. American Society for Testing and Materials, West Conshohocken, pp. 170–180. Hubbard, C.E., 1992. Transport and fate of dissolved methanol, methyltertiary-butyl-ether, and monoaromatic hydrocarbons in a shallow sand aquifer. MSc thesis, Department of Earth Sciences, University of Waterloo. Illangasekare, T.H., Jensen, K.H., Javandel, I., Mayer, A.S., 2005. Migration and distribution. In: Mayer, A.S., Hassanizadeh, S.M. (Eds.), Soil and Groundwater Contamination: Nonaqueous Phase Liquids. American Geophysical Union, Washington, pp. 47–95. Jawitz, J.W., Annable, M.D., Rao, P.S.C., 1998. Miscible fluid displacement stability in unconfined porous media: two-dimensional flow experiments and simulations. J. Contam. Hydrol. 31, 211–230. Kueper, B.H., Frind, E.O., 1988. An overview of immiscible fingering in porous media. J. Contam. Hydrol. 2, 95–110. Kueper, B.H., Frind, E.O., 1991. Two-phase flow in heterogeneous porous media. 2. Model Application. Water Resour. Res. 27, 1059–1107. Lenhard, R.J., Parker, J.C., 1990. Estimation of free hydrocarbon volume from fluid levels in monitoring wells. Ground Water 28, 57–67. Leverett, M.C., 1941. Capillary behavior in porous solids. Trans. Am. Inst. Min. Metall. Pet. Eng. 142, 152–169. Lide, D.R. (Ed.), 2008. CRC Handbook of Chemistry and Physics, 88th ed. CRC Press/Taylor and Francis, Boca Raton. Mackay, D.M., Freyberg, D.L., Roberts, P.V., 1986. A natural gradient experiment on solute transport in a sand aquifer: 1. Approach and overview of plume movement. Water Resour. Res. 22, 2017–2029. Mackay, D.M., De Sieyes, N.R., Einarson, M.D., Feris, K.P., Pappas, A.A., Wood, I.A., Jacobson, L., Justice, L.G., Noske, M.N., Scow, K.M., Wilson, J.T., 2006. Impact of ethanol on the natural attenuation of benzene, toluene and o-xylene in a normally sulfate-reducing aquifer. Environ. Sci. Technol. 40, 6123–6130. McDowell, C.J., Powers, S.E., 2003. Mechanisms affecting the infiltration and distribution of ethanol-blended gasoline in the vadose zone. Environ. Sci. Technol. 37, 1803–1810. McDowell, C.J., Buschek, T., Powers, S.E., 2003. Behaviour of gasoline pools following a denatured ethanol spill. Ground Water 41, 746–757. McNaughton, C.H., 2011. Monitoring a Shallow Gasoline Release Using GPR at CFB. MSc thesis, Department of Earth and Environmental Sciences, University of Waterloo. Mickle, R.J., 2005. A Coupled Hydrogeological-Petrophysical Analysis of Geophysical Variation in the Vadose Zone. MSc thesis, Department of Earth Sciences, University of Waterloo. Molson, J.W., Mocanu, M.T., Barker, J.F., 2008. Numerical analysis of buoyancy effects during the dissolution and transport of oxygenated gasoline in groundwater. Water Resour. Res. 44, W07418. Morris, K.R., Abramowitz, R., Pinal, R., Davis, P., Yalkowsky, S.H., 1988. Solubility of aromatic pollutants in mixed solvents. Chemosphere 17, 285–298. Oliveira, E., 1997. Ethanol flushing of Gasoline Residuals: Microscale and Field Experiments. Ph.D dissertation, Department of Earth Sciences, University of Waterloo. Parker, J.C., Waddill, D.W., Johnson, J.A., 1995. UST corrective actions technologies: engineering design of free product recovery systems. EPA/ 600/SR-96/031. U.S. Environmental Protection Agency, Washington.

166

J.G. Freitas, J.F. Barker / Journal of Contaminant Hydrology 126 (2011) 153–166

Powers, S.E., Hunt, C.S., Heermann, S.E., Corseuil, H.X., Rice, D., Alvarez, P.J.J., 2001. The transport and fate of ethanol and BTEX in groundwater contaminated by gasohol. Crit. Rev. Env. Sci. Technol. 31, 79–123. Prince, R.C., Parkerton, T.F., Lee, C., 2007. The primary aerobic biodegradation of gasoline hydrocarbons. Environ. Sci. Technol. 41, 3316–3321. Reynolds, W.D., 2008. Chapter 75 — saturated hydraulic properties: laboratory methods. In: Carter, M.R., Gregorich, E.G. (Eds.), Soil Sampling and Methods of Analysis. CRC Press, Boca Raton, pp. 1013–1024. Rivett, M.O., Wealthall, G.P., Dearden, R.A., McAlary, T.A., 2011. Review of unsaturated-zone transport and attenuation of volatile organic compound (VOC) plumes leached from shallow source zones. J. Contam. Hydrol. 123, 130–156. Ronen, D., Scher, H., Blunt, M., 1997. On the structure and flow processes in the capillary fringe on phreatic aquifers. Transp. Porous Media 28, 159–180. Schumacher, B.A., Minnich, M.M., 2000. Extreme short-range variability in VOC contaminated soils. Environ. Sci. Technol. 34, 3611–3616. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic Chemistry, 2nd ed. John Wiley & Sons, Hoboken. Smith, J.E., Gillham, R.W., 1999. Effects of solute concentration-dependent surface tension on unsaturated flow: laboratory sand column experiments. Water Resour. Res. 35, 973–982.

SRC (Syracuse Research Corporation), 2010. Interactive PhysProp Database. http://www.syrres.com. Stafford, B.P., Cápiro, N.L., Alvarez, P.J.J., Rixey, W.G., 2009. Pore water characteristics following a release of neat ethanol onto pre-existing NAPL. Ground Water Monit. Rem. 29, 93–104. Sudicky, E.A., 1986. A natural gradient experiment on solute transport in a sand aquifer: spatial variability of hydraulic conductivity and its role in the dispersion process. Water Resour. Res. 22, 2069–2082. Thomson, N.R., Flynn, D.J., 2000. Soil vacuum extraction of perchloroethylene from the Borden aquifer. Ground Water 38, 673–688. Tuck, D.M., Iversen, G.M., Pirkle, W.A., 2003. Organic dye effects on dense nonaqueous phase liquids (DNAPL) entry pressure in water saturated porous media. Water Resour. Res. 39, 1207. van Genuchten, M.T., 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44, 892–898. Yu, S., Freitas, J.G., Unger, A.J.A., Barker, J.F., Chatzis, J., 2009. Simulating the evolution of an ethanol and gasoline source zone within the capillary fringe. J. Contam. Hydrol. 105, 1–17. Zhang, Yi, Khan, I.A., Chen, X.H., Spalding, R.F., 2006. Transport and degradation of ethanol in groundwater. J. Contam. Hydrol. 82, 183–194.