Environ. Sci. Technol. 2003, 37, 1803-1810
Mechanisms Affecting the Infiltration and Distribution of Ethanol-Blended Gasoline in the Vadose Zone CORY J. MCDOWELL AND SUSAN E. POWERS* Center for the Environment, Clarkson University, Potsdam, New York 13699-5710
One- and two-dimensional experiments were conducted to examine differences in the behavior of gasoline and gasohol (10% ethanol by volume) as they infiltrate through the unsaturated zone and spread at the capillary fringe. Ethanol in the spilled gasohol quickly partitions into the residual water in the vadose zone and is retained there as the gasoline continues to infiltrate. Under the conditions tested, over 99% of the ethanol was initially retained in the vadose zone. Depending on the volume of gasoline spilled and the depth to the water table, this causes an increase in the aqueous-phase saturation and relative permeability, thus allowing the ethanol-laden water to drain into the gasoline pool. Under the conditions tested, the presence of ethanol does not have a significant impact on the overall size or shape of the resulting gasoline pool at the capillary fringe. Residual gasoline saturations in the vadose zone were significantly reduced however because of reduced surface and interfacial tensions associated with high ethanol concentrations. The flux of ethanol in the effluent of the column ranged from 1.4 × 10-4 to 4.5 × 10-7 g/(cm2 min) with the LNAPL and from 6 × 10-3 to 3.0 × 10-4 g/(cm2 min) after water was introduced to simulate rain infiltration. The experimental results presented here illustrate that the dynamic effects of ethanol partitioning into the aqueous phase in the vadose zone create an initial condition that is significantly different than previously understood.
Introduction Current environmental legislation in the United States requires the addition of oxygenates to gasoline in some parts of the country in an attempt to reduce emissions of carbon dioxide and other pollutants that result from the combustion of gasoline in automobiles (1). Recent problems associated with groundwater contamination by methyl tert-butyl ether (MTBE), the most commonly used oxygenate, has led to the introduction of legislation in the United States Congress to phase out the use of MTBE as an oxygenate in gasoline (2, 3). As a replacement and to increase our Nation’s use of biofuels, legislation currently before the Senate would increase current ethanol use from 1.5 billion to over 5 billion gal over the next 5 years (4). The addition of ethanol to gasoline affects two primary properties governing the infiltration, redistribution, and * Corresponding author e-mail:
[email protected]; phone: (315)268-6542; fax: (315)268-7985. 10.1021/es025976l CCC: $25.00 Published on Web 04/04/2003
2003 American Chemical Society
dissolution of the gasoline that is spilled to the subsurface. The hydrophilic nature of ethanol causes it to almost completely partition into the aqueous phase (5). As ethanol concentrations in the aqueous phase increase, the solubility of BTEX (benzene, toluene, ethylbenzene, and xylenes) and other gasoline components also increase; a phenomenon termed cosolvency. Significant reductions in the surface and interfacial tensions also occur as ethanol concentrations in the aqueous phase increase (6, 7). Although the effects of ethanol on the equilibrium concentrations and surface tension have been well-documented, little is known about the dynamic aspects of the partitioning process as gasoline infiltrates through the unsaturated zone. Current mathematical models describing the fate and transport of gasohol spills in the subsurface have assumed that the composition of the gasoline present at the capillary fringe is similar in ethanol content to that of what was spilled (8-11), although no field evidence is available to support this assumption. These modeling efforts have found that the addition of ethanol at 10% can increase dissolved contaminant plumes by 25-150% depending on existing conditions at the site and model assumptions. The most significant cause of the plume extension is the reduction in the degradation rates of BTEX because of the consumption of electron acceptors and nutrients as the ethanol is degraded, thereby limiting their availability for BTEX degradation (8). A significant level of uncertainty is inherent in these modeling efforts with respect to the fate of ethanol itself as well as its impact on the migration of gasoline as a separate phase and the dissolution and transport of BTEX components in the groundwater (12). The net impact of ethanol may vary depending on phenomena occurring in a series of steps, beginning with the infiltration of the gasoline through the unsaturated zone, followed by spreading of the gasoline at the capillary fringe, dissolution of constituents into water in both saturated and unsaturated zones, and transport as dissolved constituents in groundwater. Uncertainties in each of these steps must be addressed to fully develop an accurate model for the fate and transport of ethanol and BTEX in groundwater systems. At this point, there has not been any work done to understand the beginning steps of this processs infiltration and spreading in the unsaturated zone. The objective of this research was to determine the significance of adding 10% ethanol by volume to gasoline on the infiltration phenomena and the resulting distribution of gasoline in the vadose zone and at the capillary fringe in the subsurface. Gasoline and gasohol were spilled into a onedimensional (1-D) column and two-dimensional (2-D) sand tank to quantify and observe their migration and distribution through the vadose zone. To enhance the visibility of the spills, dyes were added to the ethanol and gasoline. Images of the spills from the 2-D tank were then compared to determine any differences in the behavior of the two spills to improve the understanding of the mechanisms that affect the migration of gasoline containing ethanol in the subsurface. The column experiments allowed the concentration and flux of ethanol to be quantified following drainage through unsaturated media.
Materials Porous Media. Two different mixtures of medium to coarse quartz sand from U.S. Silica (Ottawa, IL) comprised the porous media for these experiments (Table 1). The two sands had similar median sizes, although the higher coefficient of uniformity for sand 1 indicates that it had a much wider range of grain sizes. The second sand mixture had a narrower VOL. 37, NO. 9, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1803
TABLE 1. Porous Media Properties ((95% Confidence Interval) property
mixture 1
mixture 2
mean diameter (d50 (mm)) coeff of uniformity (d60/d10) hydraulic conductivity (cm/s) bubbling pressure (cm of water) residual saturation of water porosity
0.035 2.466 0.037 ( 0.013 20.0 0.07 0.376 ( 0.006
0.033 1.333 0.079 ( 0.003 23.3 0.04 0.391 ( 0.012
grain size distribution, with a majority of the fine diameter grains eliminated. This resulted in a much more uniform sand packing with sand 2 than sand 1. Other characterizations of these media are included in the Supporting Information. To characterize the sand mixtures, the hydraulic conductivity, bubbling pressure, grain densities, and grain size distribution were measured (Table 1). The porosity was evaluated in a glass column (32 cm long × 2.44 cm diameter) that was used for measuring the hydraulic conductivity of the media (ASTM D2434-68). Capillary pressure curves for each sand were generated using a standard tempe cell (Soil Moisture, Santa Barbara, CA) with a 0.5-bar high-flow ceramic plate to determine bubbling pressure and residual saturation. The sands had similar bubbling pressures although sand 1 had a higher residual saturation, as was expected for a wellgraded sand. Fluids. Standard 87 grade gasoline purchased from a gasoline station in Potsdam, NY, was used for the 2-D tank experiments. Ethanol (Pharmco Products, Brookfield, CT) was added to this gasoline to generate gasohol (10% ethanol by volume). The gasoline was dyed with a hydrophobic dye (0.1 mg/L Oil red-O; Aldrich Chemical Co.) prior to being injected into the system. For the gasohol mixture, the ethanol was dyed with a hydrophilic dye (0.05 mg/L fluorescein; Aldrich Chemical Co.) prior to mixing it into the dyed gasoline. Tracer experiments proved that neither the fluorescein nor the ethanol was retarded by the sand (13). A mixture of 10% ethanol and isooctane was used for the column experiments to simplify chemical analysis of ethanol in the effluent. Fluid properties are presented in Table 2.
Methods 2-D Experimental Spills. A 2-D stainless steel tank (1.12m × 0.68m × 0.03 m) with a glass front was used to simulate gasoline and gasohol spills in the unsaturated zone (Figure 1). The glass front allowed for photographic documentation and image analysis. The center portion of the tank was packed with quartz sand. Wells, which were connected to a constant head tank, were located on each end of the tank to maintain the water level in the system. The tank was packed with quartz sand under water in approximately 3-cm lifts using a funnel and long hose. Once the packing was complete, the water level was dropped in 5-cm increments and allowed to equilibrate. This process was continued until the capillary fringe reached the desired
FIGURE 1. Schematic of 3-cm-wide stainless steel tank used for experiments. Dark shaded area represents the wells on either side of the area packed with sand (light shading). Constant head tanks are located on each end of the tank. level. The gasoline was injected into the top of the tank at an approximate flow rate of 35 mL/min. In comparison, Rice et al. (12) report that underground gasoline tanks can leak at a rate of 3 gal/day (8 mL/min) without detection by standard methods. Thus, the rate we used is representative of a more catastrophic spill. Digital images were taken throughout the duration of the spill to document the migration and behavior of the gasoline. Following equilibration (approximately 24 h), the tank was cleaned using pure ethanol. The tank was then continuously flushed with water (8-10 pore volumes) to remove any residual ethanol that remained in the tank. The spill process was then repeated for a gasohol mixture. Pictures taken throughout each spill were compared to determine any differences in the behavior of the spills. Table 3 shows the experimental conditions for each set of spills. The tank porosity was estimated by dividing the tank volume by the sand added to the tank. The capillary fringe was determined visually as the average height prior to any gasoline entering the system. These heights were within a few centimeters of the bubbling pressure determined via standard tempe cell. Residual saturations measurements were made by collecting sand samples from the unsaturated zone of the tank and drying them to determine the moisture content. 1-D Column Experiments. 1-D column experiments were conducted to quantify the retention of ethanol in the unsaturated zone. Isooctane was substituted for gasoline in these experiments to minimize interference of the more soluble gasoline components with the UV absorbance technique used to quantify ethanol concentrations. The partitioning properties of ethanol in isooctane and gasoline are very similar (5). Fluorescein was also found to interfere with the analytical methods. One experiment was conducted with fluorescein added to the ethanol to verify the general trends in the partitioning process. Observations of this spill
TABLE 2. Fluid Properties ((95% Confidence Interval Where Available) measured property (22 °C) fluid
density (g/cm3)
viscosity (cP)
surface tension (dyn/cm)
water water with 10% ethanol water with 50% ethanol gasoline (87 grade) gasohol (87 grade + 10% ethanol) isooctane (10% ethanol)
1.0064 0.9936 0.9374 0.743 ( 0.001 0.756 ( 0.001 0.699 ( 0.0001
1.01 1.42 2.40 3.14 ( 0.09 3.27 ( 0.05 0.60 ( 0.05
71.8 46.0 28.7 22.2 ( 0.2 22.0 ( 0.1 19.2 ( 0.1
standard method
ASTM D 1217-93
ASTM D 445-97
ASTM D 971
1804
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 9, 2003
FIGURE 2. Photograph (a) and corresponding image (b) of a gasohol pool.
TABLE 3. Matrix of Experimental Conditions properties
spill 1
spill 2
spill 3
spill 4
media used flow rate (mL/min) spill volume (mL) depth to capillary fringe (cm) tank porosity capillary fringe height (cm) residual saturation via tank samples (water)
sand 1 38.0 400 30.6 0.373 19.5 nma
sand 2 36.0 400 32.5 0.386 20.3 0.056
sand 2 31.0 400 31.0 0.386 20.5 nm
sand 2 37.5 600 30.7 0.386 20.8 0.054
a
nm, not measured.
were identical to those seen in the tank experiments. Each spill thereafter was conducted without fluorescein to enable analytic quantification of ethanol concentrations A glass column (10.5 cm diameter × 44.5 cm long, tapering to 5 cm diameter over bottom 4.5 cm; Ace Glass, Vineland, NJ) was packed with quartz sand containing an aqueous residual saturation (5%). The porosity achieved in the column was slightly higher (0.42) than that of the tank. The tapered end of the column contained a screw-on Teflon cap fitted with an opening so that the effluent from the column could be collected. The volume of this column (2750 cm3) was roughly equivalent to the volume of the unsaturated zone in the 2-D tank that was contaminated by gasohol. The length of the column was equivalent to the depth of the unsaturated zone in the tank. This allowed the column experiments to be conducted under almost identical conditions as the tank experiments. The volume injected into the column was equal to that of the tank spills (400 and 600 mL) and contained the same ethanol fraction. The LNAPL that migrated downward through the column to the effluent tube was collected in 40-mL sealed vials to reduce vaporization. The ethanol in 10
mL of the collected LNAPL was extracted into 5 mL of water. The aqueous phase was then separated, and the ethanol concentration was analyzed using UV spectrophotometry (Shimadzu Scientific Instruments, UV-2401 (PC), Columbia, MD) at 192 nm. Standards were also prepared as an organic phase and extracted with water. Ethanol retained within the unsaturated zone was mobilized with water added to the column to simulate rainfall (approximately 250 mL over a duration of 15 min). The water was allowed to infiltrate through the column and then was collected in 15-mL samples and analyzed. Image Analysis. Pictures taken during the spill experiments were analyzed using image analysis software (Scion Image, Scion Corp., Frederick, MD) to quantify various characteristics. A threshold of the image was created to isolate the gasoline pool from the background (Figure 2). This differentiated the pool and the background according to userdefined pixel values. Once the pool was highlighted, a pixel count was done to determine the area, length, and height of the pool in pixels. The saturation of the pool was calculated by further differentiating between the areas of high and low saturation. The lower saturated areas were removed by increasing the sensitivity of the analysis. This allowed the average pool saturation to be determined independent of isolated NAPL blobs smeared at the periphery of the pool. Image analysis results for the saturation were verified by comparison with a mass balance on gasoline in the system. A sensitivity analysis indicated that the quantitative assessment of the NAPL saturation was not very sensitive to the user-defined cutoff point. It was much more sensitive to the defined edge of the LNAPL pool, which generally had a low saturation. To overcome this uncertainty, the outer boundary of the pool was defined by image analysis at least three separate times. The reported saturations are the average of these triplicate analyses. VOL. 37, NO. 9, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1805
FIGURE 3. Time-series depiction of a gasoline spill. The blue line represents the water table, and the black line represents initial capillary height in all photos. Upper left: 5 min after injection initiated; lower left, 15 min after injection initiated; upper right, 60 min after injection initiated; lower right, 21 h after injection initiated. Photographs include the entire visible height of the tank (57 cm) by 75 cm of the width (72% of the visible area).
Results Gasoline and Gasohol Spills. The gasoline spills behaved as expected under the conditions tested. When the gasoline entered the sand, it traveled downward until it reached the capillary fringe where it began to pool and spread outward (Figure 3). This pool contained a much higher saturation than seen anywhere else in the spill area. Low saturations were visible in the vadose zone in the vicinity of the spill, as was evident by the very light color of the dye. As time increased, capillary suction caused the unsaturated zone above the periphery of the spill to become more contaminated. This phenomenon is caused by the intermediate wetting characteristics of gasolinesbetween air and water and the positive spreading coefficient of the gasoline in this system (14). This combination of characteristics creates a capillary fringe of gasoline in the same manner that water does. A slight depression (0.5-1 cm) of the capillary fringe was observed in the area of the spill because of the pressure head imparted by the gasoline and the lowered surface tension of water contaminated by the gasoline (see Figure S-1 in Supporting Information). A time-series of photographs for a gasohol spill are presented in Figure 4. On the basis of similarity in the partitioning characteristics between the dyes and the contaminant that they represent, it is assumed that the red represents the area contaminated by gasoline and that the yellow represents the area that is affected by the ethanol at 1806
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 9, 2003
the given point in time. These pictures illustrate that ethanol as an additive in gasoline at 10% infiltrated through the vadose zone in a manner very different than the bulk gasoline. As the gasohol entered the unsaturated zone, two separate observations were made that indicated that the ethanol quickly partitioned to the residual water saturation in the vadose zone, creating regions of high ethanol concentrations. Within the first 5 min of the spill, the fluorescein was seen to accumulate in the unsaturated zone, with partitioning characteristics close to that of ethanol, this suggests similar behavior for the ethanol as well. In addition, the gasoline saturationssas defined by the intensity of the red colors near the periphery of the yellow regions were reduced near the gasoline injection site, suggesting that the high concentrations of ethanol were sufficient to reduce the gasoline interfacial tension, thereby promoting drainage and redistribution of the gasoline. The gasoline continued to travel through the sand until it reached the capillary fringe, where it began to pool and spread laterally through the tank. The ethanol, however, remained in the unsaturated zone as a component within the residual water. For the conditions tested here, it appeared that a high fraction of the gasoline that reached the capillary fringe was depleted of its ethanol content. The size and shape of the gasoline pool itself remain essentially unaffected by the addition of ethanol to the gasoline. The apparent saturation of the gasoline pool at the
FIGURE 4. Distribution of gasoline (red) and ethanol (yellow) following a 400-mL spill of gasohol (spill 3). The black line represents initial capillary height in all photos. The water table is approximately at the bottom of each photograph. Photographs include 51 cm of the height (90% of the visible area) by 75 cm of the width (72%). capillary fringe did not change noticeably with the addition of ethanol. After 2 h, the clean sand halo around the visible yellow region suggests that the apparent area of the unsaturated zone that was affected by the ethanol was larger than visible by the fluorescein dye. This is an indication that the ethanol is actually traveling farther than the dye/visualization technique suggests. This is most likely the result of low ethanol and fluorescein concentrations that were not visible in the system. The clean area surrounding the lightly shaded area can be seen, but the ethanol-impacted region still did not reach the capillary fringe. The major impact of the ethanol illustrated in these experiments was an increase in ethanol concentrations and a significant reduction in the residual saturation of gasoline retained in the unsaturated zone. Thus,
efforts for unsaturated zone remediation, such as soil vapor extraction, would have a different efficiency following a spill of gasohol than expected based on experience with gasoline spills. The experimental conditions illustrated in Figures 3 and 4 would most likely represent a small volume of gasoline spilled to the subsurface. One set of spills (no. 4) (Figure 5) was also conducted using a larger volume of gasoline to determine the extent that ethanol is retained in the unsaturated zone under these conditions. The ethanol behaved very similarly and quickly partitioned into the residual water in the vadose zone. Adding ∼60 mL of ethanol to the residual saturation of water in the region near the injection point resulted in a significant increase in the aqueous-phase saturation. The increased saturation and related decrease in VOL. 37, NO. 9, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1807
FIGURE 5. Distribution of gasoline (red) and ethanol (yellow) of a 600-mL spill of gasohol (spill 4). The black line represents initial capillary height in all photos. The water table is approximately at the bottom of each photograph. Photographs include 51 cm of the height (90% of the visible area) by 75 cm of the width (72%). the aqueous-phase surface tension reduced the capacity for capillary forces to retain this phase in the unsaturated zone. Thus, after approximately 1 h, gravity forces caused the ethanol-ladened water to drain into the existing gasoline pool. As this water drained into the existing gasoline pool, concentrations of ethanol were likely elevated because of the large volume of ethanol relative to the volume of pore water in the contaminated region of the unsaturated zone. This caused localized areas within the pool to have a reduced interfacial tension and, therefore, higher gasoline saturations. The darker area shown in the closeup photograph (Figure 6) represents the increased gasoline saturation of the larger volume spill. This phenomenon was also observed in early experiments conducted with different and less effective dyes (13). Despite the localized increases in the LNAPL saturation 1808
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 9, 2003
FIGURE 6. Closeup of gasohol picture at 52 h (spill 4). The photograph includes an area 46 cm wide by 20 cm high. in this region of the pool, the overall size and shape of the pool did not appear affected by the presence of the ethanol.
TABLE 4. Range of Ethanol Concentrations (% vol) in Effluent time
spill 1 (400 mL)
spill 2 (400 mL)
spill 3 (600 mL)
spill 4 (600 mL)
before rainfall (mostly LNAPL) after rainfall (mostly in aqueous phase)
0.004-0.022 7.1-7.7
0.004-0.052 6.2-7.1
0.022-0.22 7.6-9.1
0.006-0.16 6.0-7.9
However, the area of reduced residual contamination in the unsaturated zone was larger in comparison with the smaller volume spills. Ethanol concentrations in the unsaturated zone were estimated assuming that all ethanol partitions into the residual water in the gasoline-contaminated region. A residual water saturation of 5% was assumed (Table 3). The smaller volume spills would have had associated ethanol concentrations on the order of 40 vol %, with approximately 55% in the higher volume spills. Significant decreases in the interfacial tension (see Figure S-1 in Supporting Information) and increased solubility of the gasoline at these high ethanol concentrations (5) could both have contributed to the drainage of the ethanol-laden water to the capillary fringe and subsequent increase in LNAPL saturation. The rate at which this aqueous phase drains will vary due to the dependence of the solution density and viscosity on the ethanol content (Table 2). The increased viscosity and reduced density would result in a saturated hydraulic conductivity for a 50% ethanol solution that is approximately 40% of clean water. Predicting this rate of infiltration would require that the compositional effects of the ethanol be incorporated into the capillary, viscous and buoyancy forces acting on the fluid. An important variable that may affect the nature of gasohol infiltration is the depth to the capillary fringe. In the experiments conducted for this research, the depth was kept relatively constant. If this depth were greater, it would be expected that more ethanol would be retained within the unsaturated zone because more residual water would be available for the ethanol to partition into. This could result in less ethanol reaching the gasoline pool at the capillary fringe. Quantitative Image Analysis. The final photographs of the gasoline and gasohol spills were analyzed to quantify differences in the geometry of the pool and NAPL saturation using image analysis software (Figure 7). The graph represents the values of each gasohol spill normalized to the values for each corresponding gasoline spill. The area and length of the gasohol spills were slightly less (normalized values
