AIR SPARGING REMEDIATION: A STUDY ON HETEROGENEITY AND AIR MOBILITY REDUCTION 1S.S. Di Julio and 2A.S. Drucker 1California State University, Northridge, Northridge, CA 91330; 1Phone: (805) 667-2496, 1Fax: (805) 667-7062, 1E-mail: [email protected]; 2NFESC, US Navy; 2Phone: (805) 982-4847, 2E-mail: [email protected]. ABSTRACT Contaminated groundwater is a widespread problem often requiring innovative technology to remediate. The purpose of this paper is to present the laboratory results of air sparging models. Initial tests used very fine porous media (glass beads-packed column) to represent a relatively homogeneous soil samples. Subsequent testing employed budded core samples taken from a site of interest to represent more realistic, heterogeneous samples. 1,1,1 Trichloroethane (TCA) was used as the dissolved contaminant to represent BTEX/ gasoline contamination; however, results obtained here can be applied to any NAPL-dissolved phase. A technique based on foam injection is proposed and is demonstrated to reduce air mobility. This reduction in air mobility has potential to improve contaminant removal. Laboratory results are compared with predictions of a numerical model, which is an advection-diffusion air sparge simulation model. Sensitivity analysis of the numerical model provides the range of some key parameters used to screen/evaluate air sparging as the remediation method for a given contaminated site of interest. Eventual scaleup of the model to an actual site application can be justified by the favorable results presented in this paper. Key words: remediation, air sparging, NAPL, foam

BACKGROUND scopic capillary air channeling is estimated by The efficiency of air sparging as a ground- Clayton (1998) to occur at air-entry pressure of water remediation process depends to a large about 15 to 20 cm of water. Since this is a low extent on the contact time and contact area of air-entry pressure, it is very likely that air air with contaminated water. A number of channeling occurs, as was seen in all 18 labora- investigators have conducted laboratory or field tory experiments carried out by Clayton (1998). studies, or numerical simulation, in order to It is important to realize that the mecha- better understand air distribution and a few have nism of contaminant removal, in such by-passed conducted studies to measure the removal rate regions both in homogeneous and heteroge- of contaminants from groundwater. A majority neous porous media, is severely diffusion limited of these studies chose mostly homogeneous (Clayton, 1998; Ji and Ahfled, 1993; Clayton porous media, either 2D/3D glass-beads or and Nelson, 1995; Clark, 1996; Choa, 1998; sand packs, or core samples for laboratory Brusseau, 1991). Plummer et al. (1997) studies, and/or homogeneous strata for field observed channeling in their 2-D homogeneous, studies. Ji et al. (1993), Ahfeld et al. (1994), medium-grained glass beads model and the and Clayton (1998) have demonstrated the homogeneous sand pack of comparable perme- tendency of air channels developing in response ability, density, and porosity representing both to heterogeneity at both pore and larger scale in horizontal and vertical well configurations. The coarse to fine homogeneous sand. The transi- air distribution was more uniform for the hori- tion from pore-scale viscous fingering to macro- zontal well, suggesting that more of the porous

256 Proceedings of the 2000 Conference on Hazardous Waste Research media is impacted by air flow. We have also 1998), as well as us; our sensitivity runs, dis- observed (discussed in Results Section) that in cussed later in this paper, indicate that contami- our core studies the contaminant recovery is nant recovery near the interface increases as more efficient in the horizontally cut cores than the air channel density and VOC diffusivity the vertically cut, indicating the adverse effect of increase. However, as our results indicate, the soil stratification on contaminant recovery rate. overall contaminant recovery efficiency McKay and Acomb (1996) and Schima et decreases when air channels or bypassing al. (1996) used neutron moisture probe and occurs. This is mainly due to overall decrease cross-bore hole resistivity, respectively, to in air saturation, and as seen in our laboratory measure percentage of fluid displaced and air results, the contaminant recovery time will distribution during air sparging at two wells in a increase drastically. Hence it is best to reduce homogeneous formation consisting of uniform or eliminate channeling as proposed by foam sands. They observed an initial rapid lateral injection, discussed later in this paper. We also expansion followed by consolidation of the share the observation made by Chao et al. region. They also observed inconsistent read- (1998) that there seem to be an optimum mass ings in less permeable, heterogeneous forma- transfer flow rate. tions, indicating the inconsistent behavior of air Ahlfed et al. (1994) have conceptually flow in such formations. described the air sparging process and they note Chao et al. (1998) have developed water- that in heterogeneous, stratified formations in to-air mass transfer for a number of VOCs which sparging is often applied, the pattern of during air sparging in soil columns packed with air movement through the subsurface is com- coarse, medium, or fine sand or glass bead. plex. This complexity is largely driven by They used a reaction numerical model and variation in grain size, capillary resistance, and assumed concentration in the bulk phase re- intrinsic permeability of the porous media. In mains constant due to slow diffusion of VOC in addition, operating parameters such as airflow the aqueous phase to the air-water interface as rate, injection pressure, and depth and cross- compared to rapid volatilization of VOCs at the sectional area of injection will also affect the air-water interface. Therefore, they have contaminant recovery process. As far as the modeled the interface mass transfer alone. authors know, no laboratory study has specifi- Their results indicated that, depending on the cally involved heterogeneous core studies. VOC sparged, the estimated fraction of total Hence we hope that our laboratory results on volume affected by air sparging varied from 5 to comparison between contaminant recoveries in 20% for fine sand, but may be as high as 50% relatively homogeneous and heterogeneous for coarse sand, where more channels are porous media, and the proposed foam injection, expected to form. This observation has been instead of air injection, will aid in improving the made by others (Ji et al. 1993, and Clayton contaminant recovery for an air sparging pro-

Proceedings of the 2000 Conference on Hazardous Waste Research 257 cess. The proposed process may also decrease pressure and pressure drop across the column. the remediation or cleanup time, which is of The water vapor in the air was removed prior to important concern during field operation. injection and its flow rate was controlled and measured accurately by a digital flow system. EXPERIMENTAL AND NUMERICAL MODELS The effluent air was sampled by a gas chro- matograph. At the end of each run, it was a. Experimental Setup and Procedure extremely important to remove residual con- To study contaminant recovery in a rela- taminant. This was done by flushing the tively homogeneous porous media, a glass-bead column with warm water and hot air for a packed column, constructed of clear acrylic number of days. pipe, packed with beads of an average grain To study contaminant recovery in heteroge- diameter of 67 microns, was used. Both ends neous porous media, a composite core model of the column were sealed with recessed end was used. The core samples taken from the site caps around the injection and production ports of interest were composed of silty sand, sandy to prevent leakage. To allow a dispersed flow siltstone, silty sandstone, and fine to coarse- of air through the column, a 40-micron sintered- grained sandstone. The composite core assem- bronze filter, 0.32 cm in diameter and length, bly consisted of three core samples, cut either was positioned on the centerline at the bottom vertically or horizontally. A sleeve of heat- of the glass bead-packed column. This filter shrinkable Teflon™ tubing was slid over the represented the slotted portion of a sparge well, composite cores with screens and end plates at distributing air along its length. Positioned at the the two ends to hold the sand grains in place; the top of the column was a 11.4 cm long acrylic composite cores were then housed in a Hassler pipe, which was sealed with another end cap. pressure cell. Mineral oil was used for applica- This void space was created to allow the water tion of overburden pressure. Properties of all to swell (rise due to displaced volume by air). three laboratory models are given in Table 1. A positive pressure transducer and a differential pressure transducer were used to measure inlet Table 1. Porous Media Properties Porous Media Glass Bead Vertical Horizontal Properties Pack Composite Core Composite Core

L1ength (cm) 64164.1 15.0

D2iameter (cm) 54.7 24.5 2.5

P9ermeability, k (md) 02.3 115. 52.

Porosity, φ (3%) 346. 2535.

P8ore volume, ml 414 256. 23.

258 Proceedings of the 2000 Conference on Hazardous Waste Research Benzene, toluene, ethyl benzene, and and the experimental results can be applied to xylene are all constituents of BTEX, the largest both LNAPL and DNAPL remediation. regulated component of gasoline contamination. Both the glass-bead pack and the com- In an effort to reduce the number of variables in posite cores were saturated with a solution of the laboratory study, focus was primarily given water and TCA at three different concentrations to the remediation of benzene, the most strin- of 10, 25, and 50 ppm. Low-pressure air of gently regulated contaminant. However, due to 1.41 - 1.68 atm (6-10 psig) was injected the health hazard associated with exposure to through the bottom of the saturated glass bead- benzene, 1,1,1 trichloroethane (TCA) was packed column or the composite-core assem- selected as a suitable alternative to benzene, bly. Contaminant removal rate at three different having similar solubility, vapor pressure, and flow rates of 15, 20, and 30 ml/min were Henry’s constant (ratio of vapor pressure to studied. Air then flowed to the top where it was solubility). Other properties such as molecular sampled intermittently, at intervals of 3 - 14 weight, density, boiling point, melting point, and minutes, by a gas chromatograph to measure the specific heat were also considered for compari- concentration of TCA in the effluent, as shown son. Table 2 gives a comparison of the chemi- in Figure 1. cal properties of benzene with cyclohexane, b. Numerical Model Description toluene, and 1,1,1 trichloroethane; chemicals Analytical and numerical models provide used as suitable alternatives. Because the insight into the air sparging mass transfer pro- contaminant is in dissolved phase only, its cess. Rabideau and Blayden (1998) have specific gravity does not play an important role

Table 2. Chemical Properties of Various Contaminants Cehemical Properties Beenzen Ceyclohexan Teoluen Trichloroethan S7olubility (g/l) 18.7 5304.5 4. V7apor pressure (kPa) 112. 183. 35. 16. M1olecular weight (g/mol) 768.1 844.1 942.1 133. M5elting point (°C) 55. 65. -09 -3 B1oiling point (°C) 810. 81141 7 D5ensity (g/ml) 05.876 09.778 03.866 1.330 S4pecific heat (J/g-K @ 25 °C) 14.7 11.8 18.7 1.0 Henry's Constant 5.6x10-3 1.9x10-1 6.6x10-3 4.3x10-3 (Bar-m/mol @ 25 °C) Data from CRC Handbook of Chemistry and Physics, CRC Press, 73 Edition 1992-1993

Proceedings of the 2000 Conference on Hazardous Waste Research 259 A reactor model based upon past visual studies, which simulated flow dynamics of air sparging (Ji et al., 1993) and Wilson’s (1992) proposed general n-compartment model, was developed. Between the advective air channel regions of the soil column, VOC liquid phase transport, as simulated by the model, was assumed to be diffusion limited. To facilitate data analysis, the model was written using Visual Basic Application (VBA). The numerical model had the capability to simulate the cycling on and Figure 1. Schematic of Experimental Setup off of the sparge system, a practice some agree reviewed the literature on modeling work done has the potential to reduce costs and increase on air sparging and have categorized the nu- remediation efficiency (Hinchee, 1994; Acomb merical models into two types: and McKay, 1996). In addition, the model can 1. Mechanistic models, or multi-dimen- be used to help predict the extent of post- sional, PDE, known also as compositional remedial contaminant rebound by simulating multidimensional models for fluid flow and groundwater contaminants as they diffuse contaminant removal studies. These numerical towards equilibrium concentrations. About models are commonly used in the oil industry 60% of the 32 case studies evaluated showed due to availability of extensive site characteriza- poor performance (not sufficient for site closure) tion. Such data are usually not cost-effective for due to substantial rebound following an initial remediation sites where operation costs are contaminant concentration reduction (Bass, minimized and comprehensive site characteriza- 1996). Generally a period of 6-12 months is tions are not available. required for rebound to fully develop. In some 2. Reactor models, on the other hand, are cases this rebound may be related to rise in simpler and can handle mass removal, volume of water table, and hence desorption of contami- fluid circulating through the source zone, and air nant. Our results showed that the slow diffusion channel development. of contaminant towards air channels may also A number of investigators have used be responsible for rebound of dissolved con- multiphase flow simulation to model air sparging. taminant concentration. But as Clayton (1998) points out, multiphase Because the water velocity is negligible flow simulations are unable to handle air chan- (McCray and Falta, 1977), the aqueous-phase neling without special considerations to repre- dispersion is neglected. The contaminant vapor- sent flow in individual stream tubes, which are izes at the interface, and its distribution within not interconnected. the air channel is assumed to be instantaneous

260 Proceedings of the 2000 Conference on Hazardous Waste Research Figure 2. Annular Aqueous and Air Channel Figure 3. Annular Shells Within a Single Aque- Regions of Model ous Region and thus is considered an equilibrium process, length (∆z) sections as shown in Figure 2. described by Henry’s Law (Equation 1), which Figure 3 illustrates how each cylindrical aqueous is the best-case assumption. Our measurements region in each soil column element is divided also indicated that equilibrium is not reached in into equal-thickness (∆r) annular shells. Addi- the early part of contaminant recovery curves, tionally, a tortuosity parameter (Lpath or path when advection forces are predominant. length coefficient) was used to make the parallel air channels resemble the tree-like air channels gl= CKCh (1) observed by investigators. where: Combining Fick’s first law relationship Cg = Molar concentrations of compund in gas of binary diffusion with Henry’s Law, along with phase - (g moles/m3) equations to describe the geometry, results in Cl = Molar concentration of compound in liquid the following equation, which represents the phase - (g moles/m3) diffusion transport of the contaminant through

Kh = Henry’s law constant - dimensionless the aqueous region: dCw M2π∆ zD  Based on work performed by Wilson et. == −w + +()ww − φ∆ ∆ rCjj+1( j 1) rCCjj−1 al. (1992) and Roberts et. al. (1993), the soil dt Vol r Vj  where: column was modeled as a composite of evenly spaced cylindrical air channels with a surround- φ= Soil porosity - (L3/ L3) ing nonadvective aqueous region (a radius of “b”) illustrated in Figure 2. All air channel and Vol = Volume - (L3) cylindrical aqueous regions are to be equal in Cw = Concentration of the contaminant length and circumference. The soil column of in the solvent (water) - (m/ L3) height, “h”, is to be divided into equal vertical-

Proceedings of the 2000 Conference on Hazardous Waste Research 261 The following equation represents the which can result in improved contaminant contaminant transport once it has entered the air recovery. Foams are dispersion of gas bubbles channel: in liquids. Such dispersions are normally quite unstable, unless surfactant is added to the liquid, −−gggw + − w 4/DC()ih K C n V c() Cii1 C dC =+ which greatly improves the stability. Previous φ∆ dt a r φπa2 ∆ z() Channel # foam studies in reservoir engineering have For details of the numerical model, please demonstrated the tendency of foams to prefer- refer to Drucker (1995,1996). entially plugging channels or higher permeable c. Foam Injection to Reduce Air Mobility regions in porous media. Additionally air The capacitance or dead-end pore model mobility is reduced via an increase in air viscos- was originally proposed to explain the concen- ity and decrease in air-relative permeability (a tration “tail” observed in breakthrough curves of reduction as high as 200-600 fold), while gas displacements. This tail is more pronounced in saturation remains unchanged (Khan, 1965; carbonate than in sand stones because the pore Bernard and Holm, 1964). This is attributed to structure of a typical carbonate is more hetero- blocking of pore throats due to gas films. In a geneous (Raimondi and Torcaso, 1964; parallel core flood study, Di Julio (1989)

Stalkup, 1970; Shelton and Schneider, 1975; demonstrated the ability of foam to reduce CO2 Spence and Watkins, 1980). Similarly we have gas mobility by plugging air channels in higher

observed the more pronounced tailing phenom- permeable core and to diverting CO2 to lower enon here when the results of contaminant permeability core. The foam injection resulted recovery for glass beads are compared with in an incremental oil recovery of 33.6%. It is that of the core, even though the contaminant expected that foam injection in the air sparge exists as a dissolved phase only. This leads one process reduces air mobility and hence provides a to believe that in our study, air channeling or by- significant reduction in contaminant recovery time. passing in heterogeneous cores resulted in the RESULTS inefficiency in contaminant recovery, leading into longer recovery times than the relatively homo- a. Model Sensitivity Analysis A sensitivity analysis was performed by geneous glass bead packs. This is also sup- varying several input parameters and observing ported by the results of sensitivity analysis their impact on normalized residual mass of on “a” (air channel radius) as seen in the contaminant. The input variables used to Results section. conduct the analysis were “a” (air channel This problem may be remedied by using radius), “D” (diffusivity constant), “Kh” foaming surfactants, which tend to encapsulate (Henry’s constant), and “Vc” (air injection flow air and form foam. The reduction of air mobility rate). Results of this analysis are presented in increases air residence time and the contact Figures 4 - 7. area between air and the contaminated water,

262 Proceedings of the 2000 Conference on Hazardous Waste Research Figure 4. Normalized Residual Mass vs. Figure 5. Normalized Residual Mass vs. Remediation Time (“a” is varied) Remediation Time (“D” is varied)

Normalized values of residual mass, M/ Henry’s constant, Kh, has a pronounced Mo, for a given range of “a”, are shown in but limited effect on air sparging remediation Figure 4. As the value of the air channel radius, efficiency, as shown in Figure 6. In this particu- a, is changed, it proportionally affects the value lar case, Henry’s constant has an upper limit of of “b,” the radius of the aqueous region sur- approximately 0.35 (i.e., Kh values larger than rounding the air channel. For instance, if the 0.35 will not improve upon the remediation value “a” is decreased, the value of “b” is also rate). The lower boundary of Kh values, such as decreased (due to an increase in air channel Kh = 0.01, illustrates how a relatively low number density), causing the remediation rate to volatilization rate limits the contaminant removal improve. Therefore, to improve contaminant rate. Therefore, an analysis such as this can be removal efficiency, it is desirable to promote a useful in determining the effectiveness of air relatively large number of air channels, having small sparging on the removal rate of contaminants radii, within a given volume of saturated soil. In the with lower values of Kh, such as semi-volatile limit, this may be represented by evenly distributed organic solvents. air saturation within a given zone. The volumetric flow rate of injected air has The diffusivity constant, D, has no direct a direct bearing on the air saturation within bearing on the channel geometry but has a saturated soil. It is expected that an increase of significant effect on the contaminant removal airflow rate will increase the air saturation and rate. Most VOC diffusion coefficients will fall in hence increase the number and diameter of the the range between 2.54E-7 to 3.0E-6 cm2/s. air channels. However, to investigate only the The contaminant removal rates in the form of M/ effect of air volumetric flow rate on rate of Mo are illustrated in Figure 5. From the figure it contaminant removal, air saturation is held is evident that the increase in value of diffusivity constant, while the flow rate is raised or low- improves the contaminant removal rate. This ered, as shown in Figure 7. The lower limit of result is expected due to the model being air-flow, Vc = 0.016 cm3/s, in Figure 7, is an diffusion limited. example of flow rate becoming too small to

Proceedings of the 2000 Conference on Hazardous Waste Research 263 Figure 6. Normalized Residual Mass vs. Figure 7. Normalized Residual Mass vs.

Remediation Time (“Kh” is varied) Remediation Time (“Vc” is varied)

efficiently remove the contaminant volatilized in the aqueous phase, at which time the diffusion the air channel. Lower limits of airflow such as process will begin to dominate the removal rate. this are encountered at the perimeter of a sparge These two regions of flow regimes: advection- well’s region of influence. The upper limit of controlled and diffusion-controlled, are seen as effectiveness of raising the airflow rate is real- the initial peaks followed by an asymptotic ized at or slightly less than Vc = 0.295 cm3/s. behavior in all of our contaminant removal Therefore, air injection greater than 0.295 cm3/s curves, respectively. will only serve to increase the cleanup costs Contaminant recovery in glass bead packs associated with sparging the contaminated were measured at three different air-injection groundwater. This is an important observation, flow rates (10, 20, and 30 ml/min) for three since the air sparging process is a diffusion- different initial concentrations of (10, 25, and 50 limited process. An increase in air-injection ppm TCA in water). Figure 8 shows the flow rate will not have a significant effect on the contaminant recoveries at three different flow rate of contaminant diffusion and removal, but rates for initial concentration of 25 ppm. Figure may only improve the vaporization (according to 9 shows a similar curve for 50 ppm initial Henry’s Law) and hence its subsequent advec- concentration. The increase in air-injection flow tion by air. rate increases the contaminant recovery slightly. b. Laboratory Results Comparing the contaminant recovery for 50 The removal rate is believed to be con- ppm and 25 ppm at air-flow rate of 20 ml/min, trolled by two distinct processes of advection the higher peak for 50 ppm, indicates higher and diffusion. Initially the rate of contaminant convective recovery very early on. However, removal is controlled by advection. Contami- this higher recovery is not sustained during the nant is removed by relatively quick vaporization second portion of the curve, which is diffusion- from the air channel wall and subsequent limited, shown as the asymptotic recovery, advection by air, until the contaminant concen- leading to lower overall percentage recovery. tration in the air channel reduces below that in

264 Proceedings of the 2000 Conference on Hazardous Waste Research Figure 8. Cont. Recovery for Initial TCA Figure 9. Cont. Recovery for Initial TCA Concent. of 25 ppm—Glass Bead Pack Concent. of 50 ppm—Glass Bead Pack

To study the effect of heterogeneity on contaminant is increased, the optimum flow rate, contaminant recovery, measurements on glass for which contaminant recovery is maximum, bead pack may be compared with those from also increases. For example, we observed that the cores and also measurements on horizontally the optimum flow rate for 10 ppm was 10 ml/ cut cores may be compared with those for the min, while that of 25 ppm was about 20 ml/min. vertically cut cores, which are more heteroge- We obtained reasonably good reproducibility of neous and stratified. the recovery profile and the percent of total Figures 10 and 11 show results of three air contaminant removed. But we had difficulty sparging runs for the horizontal and vertical with cleaning the cores after each run. composite cores, respectively. Concentrations Our experimental results on glass bead of the contaminant, TCA, in effluent air, as a packs show less channeling than our core function of time for air injection rates of 15, 20, studies, as seen in the shape of contaminant and 30 ml/min are compared. Note that these recovery profile and the shorter recovery time results are for initial concentration of 25 ppm (about 4 hr recovery time for the glass bead TCA in water, used to saturate the core prior to pack versus 24 hr recovery time for the com- air injection. Results for the horizontal compos- posite core, estimated based on cumulative ite core look more like the glass bead pack run; recoveries), considering that the glass bead the contaminant removal curves are flatter at pack has a pore volume which is about 22.4 generally higher concentrations of contaminant times that of the composite cores. This can also and with less tailing effect, as compared with be observed when experimental results are those for the vertical composite core. This compared with the numerical model predictions, behavior could be attributed to the higher which assume the contaminant recovery takes heterogeneity of the vertical composite core, place via air channels; the predicted profile where more channeling/bypassing takes place matches the core studies better due to its and hence the removal efficiency is inconsistent inherent heterogeneity and hence predominant and lower. As the initial concentration of recovery via air channels.

Proceedings of the 2000 Conference on Hazardous Waste Research 265 Figure 10. Cont. Recovery for Initial TCA Figure 11. Cont. Recovery for Initial TCA Concent. of 25 ppm—Horizontal Composite Concent. of 25 ppm—Vertical Composite Core Core

Results of two experiments on glass bead uted to dispersion and/or adsorption processes and composite cores are compared with the which were not included in our numerical model. numerical model prediction as shown in Figures Even with this discrepancy, one may still acquire 12 and 13. Figure 12 for the glass bead pack is a conservative prediction of the air sparge at 10 ppm and 15 ml/min, while Figure 13 is for remediation rate, if one were to scale up the vertical composite core at 50 ppm and 15 ml/ numerical model to simulate air sparging at the min initial TCA concentration and air-injection field scale. The existing model has considerable flow rate, respectively. The predicted concen- potential for evolving into an accurate field-scale trations are matched against the experimental model. Once developed, the field-scale numeri- values by varying the parameter a, air channel cal model could be used in the design of a full- radius, which in effect also changes the density size air sparge system by determining air sparge of air channels. Table 3 shows the input and well spacing and placement, air injection rate, calculated variables used for simulation. The injection pressure, and rate of remediation. peak values of contaminant concentration are Since the numerical model allows for the approximately 1/4 to 1/10 of the estimated contaminant removal through air channels only, equilibrium values, based on Henry’s Law, the better match between measurements in the which assumes instantaneous distribution of core samples and predictions indicates the contaminant within the air channels. As ex- existence of channels in heterogeneous samples, pected, Henry’s law provides the upper limit of and hence the need for reduction of air channel- contaminant concentration in the effluent air. ing. One such remedy may be foam injection. The laboratory measurement results in Figure 13 c. Result of Foam injection show a leveling off at 3.5 ppm, after 3 1/3 In the case considered here, the average hours, while the model prediction tends to bubble size is larger than the pore diameter and asymptotically approach zero-contaminant thus foam flows as a progression of films that concentration. This discrepancy may be attrib-

266 Proceedings of the 2000 Conference on Hazardous Waste Research Figure 12. Comparison of Laboratory and Figure 13. Comparison of Laboratory and Numeric Model Results—Glass Bead Pack Numeric Model Results—Composite Core separate individual gas bubbles. Some of the column where surfactant is present. We ob- undesirable features of surfactant, such as sensi- served the formation of foam within the column tivity to highly saline brine, temperature, contami- and the increase in pressure drop by a factor of nant type, and retention, need to be considered 20 to 40 on the average, at flow rates of 53 and when surfactants are selected for a given site. 20 ml/min, respectively. This drastic increase in We used the surfactant Steol CA-460, pressure drop is attributed to decrease in air manufactured by Stepan Company in relative permeability, which results from reduc- Northfield, Ill. Steol CA-460 is an alcohol- tion of air channels. As seen in previous studies ethoxy sulfate consisting of 15% denatured ethyl ,the reduction of gas mobility can improve its alcohol and 2% ammonium sulfate. To study saturation distribution and hence increase both the reduction of air mobility in the air sparging the contact area and contact time between the process, the glass bead-pack column was air and the contaminant, resulting in an improved saturated with water and the pressure drop recovery of the contaminant from porous media. across the column was measured as air was CONCLUSION injected into the column at a given flow rate. Laboratory results from glass bead packed This provided the baseline pressure-drop column and composite cores have demonstrated profile. The column was then saturated with a the effectiveness of air sparging as a remediation solution of 1%-by-weight surfactant in water process. Two distinct regimes of advective- and again the pressure drop across the column controlled and diffusion-controlled flows are was measured as air was injected. Note that observed. Measurements on composite cores we expected some air channeling or bypassing showed more channeling than the relatively in the glass bead pack, which would be less homogeneous glass bead packs, and hence than that in the composite cores. Figure 14 required a much longer time for air injection and shows an example of the measured increase in subsequent contaminant removal. pressure drop when air is injected into the

Proceedings of the 2000 Conference on Hazardous Waste Research 267 A numerical reaction model was used to contaminant removal agreed fairly well with the conduct a sensitivity analysis by varying input laboratory measurement results and indicated parameters such as air channel radius (a), the prominent existence of channels in heteroge- diffusion constant (D), Henry’s Law constant neous samples, and hence the need for reduc- (Kh), and injection flow rate (Vc). Results of tion of air channeling. Use of foaming surfac- the analysis illustrate the significance of each tants is suggested as a method to reduce air parameter with respect to air sparging feasibility mobility in channels. This can lead into reduc- and remediation rate. Model predictions of tion of air channeling and may improve contami- Table 3. Input and Calculated variables for Simulation Glass Bead Input Variables

veariable vsalu unnit descriptio

r0ad 1..00 ien radius of influenc

h02.4.000 ihn sparge pt. - water table dept

V0c 0s.0152 ceuin/ control panel volumetric flow rat

K3h 0a.71 nt/ Henry's constan

C0jo 5b0.0 pnp initial contaminant concentratio

v60n.3 cyuin/cui total soil porosit

w30n.3 cyuin/cui water-filled soil porosit

D71s.50E-0 stqin./ diffusion constan

C0sat 1b,100,000.0 pnp contaminant saturation concentratio

a27..00E-0 isn air channel radiu

L4path 1n. i1n/i path-length coefficient ( ≤ Lpath ≤ 2)

d0t 4sl.2 time interva

Glass Bead Calculated Variables

vnariable cealculatio vsalu unnit descriptio

d#z h0/cell row 2n.40 itcell heigh

b5r4ad/(Chanl#Lpath)^. 0n.40 isnonadvective channel radiu

d9r (7b-a)/ 0n.03 isnonadvective shell thicknes

arfactor a2/d 1n.88 isn/i air channel radius/shell thicknes

chhannel# (3v-w) x (rad/a) ^2/Lpat 4a37 nt/ airchannel number per elemen

pspbfactor .1001 if using ppb' 0a.00 nr/ ppb calculation facto

zjdaz x nn/ itcell heigh

268 Proceedings of the 2000 Conference on Hazardous Waste Research Bass, D.H., and R. A. Brown, 1996. Air sparging case study database update. Proceedings of the First International Symposium on In Situ Air Sparging for Site Remediation. October 24 and 25, Las Vegas, Nevada. Brusseau, M.L., 1991. Transport of organic chemicals by gas advection in structured Figure 14. Foam-Induced Nobility Control or heterogeneous porous media: devel- Indicated by Increase in Pressure Drop opment of a model and application to column experiments. Water Resour. nant recovery by diverting foam to less perme- Res., Dec., 27, 12, 3189-3199. able zones. Bernard, G.G., and L.W. Holm, 1964. Effect of foam on permeability of porous ACKNOWLEDGMENTS media to gas. SPE 986, presented at We wish to thank Southern California Gas the Fall technical Conference and Co. for providing project funding, the California Exhibition of the Society of petroleum Engineers, Houston, Texas. State University-Northridge engineering stu- dents who helped construct the laboratory Chao, K.P., S.K. Ong, and A. Protopapas, 1998. Water-to-air mass transfer of apparatus and conduct the applicable tests, and VOCs: Laboratory-scale air sparging Bill Shallenberger for guidance and review of system. J. of Environmental Engi- the analysis and laboratory results. neering, 124, no. 11, pp1054-1060. Andrew S. Drucker Clayton, W.S., 1998. A field and laboratory ([email protected]) is currently an investigation of air fingering during air engineer at the Naval Facilities Engineering sparging. Ground Water Monitoring and Remediation, 18, no. 3,134-145. Service Center, Point Huneme, Calif. He completed the numerical analysis presented here Di Julio, S. S., and A.S. Emanuel, 1989. Laboratory study of foaming surfactant as his master’s thesis at the aforementioned for CO2 mobility control. 1989. SPE university. Shoeleh Di Julio (the correspondent Reservoir Engineering, May, 136-142 author, [email protected]) is a professor at Drucker, A.S., 1995. Aliso Canyon site in situ California State University-Northridge, Me- remediation. M.S. Thesis. California chanical Engineering Department, Northridge, State University, Northridge. CA 91330-8348. Drucker, A.S., and S.S. Di Julio, 1996. Groundwater cleanup, in situ air REFERENCES sparging: development of a model Ahlfed, D.P., A. Dahmani, and W. Ji, 1994. A and application to saturated soil conceptual model of field behavior of air column experiments. Water Envi- sparging and its implication for applica- ronment Federation, tion. Ground Water Monitoring and WEFTECH’96, 69th Annual Con- Remediation, 14, no. 4: 132-139.

Proceedings of the 2000 Conference on Hazardous Waste Research 269 ference & Exposition, Dallas, Texas, Water Monitoring and Remediation, October 5-9, 1996. Fall 1998, pp120-130. Hinchee, R.E., 1994. Air sparging for site Raimondi, P., and M. A. Torcaso, 1964. Distri- remediation. CRC Press Inc., bution of oil phase obtained upon Boca Raton. imbibition of water. Transaction of the American institute of Mining and Ji, W., A. Dahmani, D.P. Ahlfeld, J.D. Lin, and Mmetallurgical Engineering, 231: 49- E. Hill,1993. Laboratory study of air 55. sparging: air flow visualization. Ground Water Monitoring & Remediation 13, Roberts L.A., and D.J. Wilson, 1993. no.4: 115-126. Groundwater cleanup by in situ sparging I. modeling of dense, non- Johnson, R.L., P.C. Johnson, D.B. aqueous-phase liquid droplet removal. McWhorter, R.E. Hinchee, and I. Separation Science And Technology, Goodman, 1993. An overview of in 28, 5, 1127-1143. situ air sparging. Ground Water Monit. Rev., Fall, 127-135. Schima, S., D.J. Labrecque, and P.D. Lundegrad, 1996. Monitoring air Kearl, P.M., and N.E. Korte, 1991. Vapor sparging using resistivity tomography. extraction experiments with laboratory Ground Water Monitoring and soil columns—implications for field Remediation 16, no.2: 131-138. programs. Waste Management, 121, 231-239. Stalkup, Fred I., 1970. Displacement of oil by solvent at high water saturation. Soci- Khan, S. A., 1965. The flow of foam ety of Petroleum Engineers Journal, through porous media. M.S. thesis, 10:337-348. Stanford University. Shelton, J.L., and F.N. Schneider, 1975. The McCray, J.E., and R.W. Falta, 1997. Numeri- effect of water injection on miscible cal simulation of air sparging for flooding methods using hydrocarbons remediation of NAPL contamination. and carbon dioxide. Society of Petro- Ground Water 35, no. 1: 99-110. leum Engineers Journal, 15: 217-226. McKay, D. J., and L. J. Acomb, 1996. Neu- Spence, A.P. Jr., and R.W. Watkins, 1980. tron moisture probe measurements of The effect of microscopic core fluid displacement during in situ air hetrogeneity on miscible flood residual sparging. Ground Water Monitoring oil saturation. SPE 9229, presented at and Remediation 16, no.4: 86-94. the 55th Annual Fall Technical Confer- Plummer, C. R., J. D. Nelson, and G. S. ence and Exhibition of the Society of Zumwalt. 1997. Horizontal and vertical Petroleum Engineers, Dallas, Texas. well comparison for in situ air sparging. Wilson, D.J., S. Kayano, R.D. Mutch, and Ground Water Monitoring and A.N. Clarke, 1992. Groundwater Remediation 17, no. 1: 91-96. cleanup by in situ sparging, I, math- Rabideau, A.J., and J.M. Blayden, 1998. ematical modeling. Separation Sci- Analytical model for contaminant mass ence and Technology, 27 no. 8 & 9: removal by air sparging. Ground 1023-1041.

270 Proceedings of the 2000 Conference on Hazardous Waste Research