Published online 17 May 2007
Published in Vadose Zone J 6:397-405 (2007)
DOI: 10.2136/vzj2005.0131
© 2007 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
SPECIAL SECTION: SAVANNAH RIVER SITE
Soil Vapor Extraction Performance in Layered Vadose Zone Materials
Christine Switzerb and
David S. Kossona,*
a Dep. of Civil and Environmental Engineering, Vanderbilt Univ., Box 1831, Station B, Nashville, TN 37235
b current address: Inst. for Infrastructure and Environment, School of Engineering and Electronics, Univ. of Edinburgh, Edinburgh, UK
* Corresponding author (david.kosson{at}vanderbilt.edu).
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.
1 The soil vapor extraction and air sparging system was designed and installed by the contractor, Westinghouse Savannah River Company. 
Received 14 November 2005.
 |
ABSTRACT
|
|---|
A pilot soil vapor extraction (SVE) system was installed at a small landfill within the Savannah River Site to address trichloroethylene (TCE) contamination present in the vadose zone. The SVE system has been operating since September 1999 and numerous tests have been performed on the system. A model was developed to simulate SVE at this site, incorporating the effects of contaminant behavior in a layered subsurface as well as the effects of contaminant diffusion into and out of soil aggregates. The objectives of this study were to: (i) compare the field data from the site with predictions from this mechanistic model; and (ii) establish the case for closure based on field observations and model predictions. A dense non-aqueous-phase liquid TCE source was discovered at the site during the course of operation. Location of this source compares well with the predicted residual source from the application of the diffusion components of the SVE model to soil gas TCE concentration rebound observations. Collectively, the field observations and the model predictions strongly support the observations that a substantial portion of the source contamination at the site has been removed by the SVE system and that the criteria for site closure have been met.
Abbreviations: AS, air sparging • CBRP, C-Area Burning Rubble Pit • DNAPL, dense non-aqueous-phase liquid • NAPL, non-aqueous-phase liquid • SVE, soil vapor extraction • TCE, trichloroethylene.
|
INTRODUCTION
|
|---|
Soil vapor extraction is a common in situ method for the removal of volatile organic compounds from the vadose zone. In each SVE installation, a series of wells is screened in the subsurface according to the information that is known about site stratigraphy and spill distribution. Vacuum is applied to the configuration of wells, drawing the contaminant vapors through the subsurface to the ground surface where they are either treated or released, depending on contaminant mass removal rate. The problem of poorly defined contamination sources exists at many sites with causes such as undetected spills or insufficiently documented chemical waste disposal. Source amount, distribution, and time spent in contact with the subsurface are very important parameters in the design and operation of SVE systems, as well as in the determination of the required cleanup time (Massmann et al., 2000). Equally important is subsurface heterogeneity, as the presence of different soil types can impact SVE effectiveness and total remediation time (Mackay and Cherry, 1989; Benson et al., 1993; Rodriguez-Maroto et al., 1994; Brusseau et al., 2000).
A number of models have been proposed to describe SVE remediation (USEPA, 1995). Typically, SVE models fit into two groups: local equilibrium and nonequilibrium. Nonequilibrium models assume that mass transfer between phases is rate controlled, whereas local-equilibrium models assume that the phase transfer is fast or instantaneous. Previous work suggests that the local-equilibrium assumption is valid for non-aqueous-phase liquid (NAPL) evaporation, but probably not valid for other processes such as diffusion through an aqueous phase or sorption (Wilson et al., 1994; Ng and Mei, 1996).
In terms of sorption processes, linear partitioning is a common assumption (Ng and Mei, 1996; Kaleris and Croise, 1997). Brusseau (1995) found that the assumption of linear partitioning provides an insufficient representation of contaminant behavior when the Freundlich exponent is <0.9. Nonlinear partitioning has been investigated by the inclusion of Freundlich, Brunauer–Emmett–Teller (BET) or Langmuir nonlinear sorption models into the larger scale SVE model (Kaleris and Croise, 1997; Fischer et al., 1998). Nonlinear partitioning in itself has not captured adequately the observed effects of "tailing" during desorption and "rebound" of soil gas contaminant concentrations after termination of advective mass removal (Massry, 1997). Depending on the soil type(s) present, other processes may contribute more to these observed phenomena in the field.
Subsurface heterogeneities or apparent heterogeneities such as SVE-induced preferential pathways, soil aggregates, cracks and fractures, and nonuniform moisture distribution have been represented in subsurface transport and SVE models. Typically, these models are dual domain in nature (Gerke and van Genuchten, 1993; Ng and Mei, 1996; Feehley et al., 2000). Advection, diffusion, and dispersion are the primary transport mechanisms in the mobile regime, while diffusion alone is the transport mechanism in the aggregate or immobile regime. These regimes can be interfaced by a mass balance.
The inclusion of NAPL in SVE models varies according to factors such as the age of the spill and the information known about it. In models of fresh spills, NAPL would be included, as the contaminant did not have enough time to distribute through the subsurface. In the case of older or less defined spills, a NAPL component might be excluded from the modeling approach (Ng and Mei, 1996, Kaleris and Croise, 1999, Ng, 1999). DiGiulio et al. (1999) include a NAPL distributed through the subsurface. This distribution was represented as partitioning between the NAPL and air phases. The NAPL and water interactions are neglected. Laboratory studies by Schaefer et al. (1998) suggested that pore filling by a NAPL had a similar impact on diffusional limitation as that of pore filling by water.
This study involved the application of a previously developed SVE model (Switzer, 2004) to a field scenario where the initial conditions are poorly defined. The SVE model couples the advective removal of TCE to a multipore regime mass transfer model, allowing for the inclusion of the important aggregate-scale effects in the bulk model. This model captures the observed slow desorption and rebound phenomena as well as the effect of aging in slowing the SVE remediation process.
The objectives of this study were to apply the SVE model to evaluate the field data collected from the site, to estimate the extent of contamination, and to support the decision to end active SVE at the site.
|
Materials and Methods
|
|---|
Site Background and Installation History
A pilot scale SVE and air sparging project was performed at the C-Area Burning Rubble Pit (CBRP) within the U.S. Department of Energy's Savannah River Site to examine the effectiveness of SVE in layered soils. The CBRP was used for the disposal of debris and solvents (primarily TCE) from the 1950s to the 1980s, when it was filled with natural materials and closed. Records of its use were not kept, so the distribution and amount of the TCE contamination present were not known.
The subsurface stratigraphy at the CBRP consists of interbedded layers of sands and clays. A thick, low-permeability clay layer is present at a depth of 9 m (30 ft). The shallow water table aquifer begins at a depth of approximately 21 m (70 ft). At a depth of 27 m (88 ft), a Tan Clay confining unit separates this aquifer from the deeper aquifer (Westinghouse Savannah River Company, 1997).
The plume from the CBRP extends for nearly 1.6 km (1 mi) to the surface water discharge point, where TCE concentrations exceed the maximum contaminant level specified by the USEPA. Soil vapor extraction was selected to remediate the source TCE contamination in the vadose zone at CBRP and air sparging (AS) was selected to remediate the shallow groundwater near the original pit.
In spring 1999, 39 SVE wells and 17 AS wells were installed at the CBRP. Over the original pit, SVE wells were installed in a clustered configuration of three wells per location (Fig. 1). The deepest wells, designated as A screens, were intended to operate in conjunction with the AS system. The B screens were placed in the deeper vadose zone, below the stiff clay formation. The C screens were placed in the shallow vadose zone, at or above the stiff clay formation. Well locations and depths were selected according to geology and contaminant distribution.1
View larger version (46K):
[in this window]
[in a new window]
|
FIG. 1. (a) Subsurface stratigraphy at Soil Vapor Extraction (SVE) Site 19C translated to (b) the bulk SVE model and (c) the intraaggregate regime.
|
|
Locations SVE 18, 19, and 22 were selected for extensive characterization because these wells were believed to be located in the region with the highest levels of contamination (Fig. 2). In May 1999, soil cores were taken from six locations between these three SVE locations. In June 1999, a team from the USEPA, Region 4, installed 26 soil vapor monitoring probes at various locations and depths between these three SVE locations. The SVE wells 18, 19, and 22 are shown with respect to the subsurface stratigraphy in Fig. 3. Soil core and monitoring probe locations are indicated. In April 2001, that team returned to install an additional set of 34 soil vapor monitoring probes. These probes were installed across a broader area in the original pit, as well as downgradient from it.
View larger version (25K):
[in this window]
[in a new window]
|
FIG. 2. Plan view of the installed soil vapor extraction (SVE) system at the C-Area burning rubble pit (CBRP), showing the location of the SVE 18–19–22 study area and the trichloroethylene non-aqueous-phase liquid (NAPL) that was discovered by soil coring.
|
|
View larger version (74K):
[in this window]
[in a new window]
|
FIG. 3. Subsurface stratigraphy at soil vapor extraction (SVE) sites 18, 19, and 22 as interpreted from cone penetrometer measurements made at the C-Area burning rubble pit. The location where the trichloroethylene non-aqueous-phase liquid (NAPL) was discovered is indicated.
|
|
After preliminary testing, the SVE system began operating in October 1999 at SVE 18B, 18C, 19B, 19C, 22B, and 22C only. The A screens were flooded, so those wells were inoperable. In December 1999, operation was expanded to the full well field. The SVE system has been operated on a near-continuous schedule since then. Periodically, the SVE system was shut down to conduct soil gas TCE rebound tests and perform routine maintenance on the system. The soil gas TCE concentration rebound studies involved the purposeful shutdown of the SVE system to observe the diffusion-induced rise of TCE concentrations in the absence of SVE-induced advection. In cases where maintenance might interfere with the test results, maintenance began after the rebound test was conducted. An operational history for the CBRP can be found in Table 1. Because the AS system was not used extensively during this study, the data related to its use are not presented here. Its operational schedule is included in Table 1, however, because of its potential effects on the data collected for the SVE system
Soil Gas Trichloroehylene Concentration Rebound Testing
The rebound phenomenon has been observed widely in the laboratory and in the field (DiGiulio, 1992; Armstrong et al., 1994; Ng et al., 1999; Bass et al., 2000; Mackay et al., 2000). The SVE operation flushes accessible contaminant vapors from the vadose zone. After SVE operation has been terminated, subsurface soil gas contaminant concentrations increase as contaminant diffuses from less accessible pore space into the SVE-accessible pore space. Subsurface soil gas TCE concentration rebound testing was the primary performance assessment tool for the CBRP since SVE operation began. For each rebound test, the SVE system was shut down and the subsurface was allowed several weeks to reequilibrate. Soil gas TCE concentrations were recorded at the SVE wells and monitoring probes several times during each test. The first test lasted 2 wk, the second through fifth tests lasted approximately 1 mo each, and the sixth rebound test lasted 6 mo. A more detailed explanation of the test methodology as well as interpretation for the first three rebound tests conducted at the CBRP are available elsewhere (Switzer et al., 2004).
Dense Non-Aqueous-Phase Liquid Soil Coring
In September 2001, a soil coring expedition was performed at the CBRP in an effort to locate a suspected dense non-aqueous-phase liquid (DNAPL) source at the site (Rossabi, personal communication, 2003). Soil cores were collected according to a direct push method involving the cone penetrometer test wireline system that allowed continuous vertical sampling at each sampled location without repeated removal and reinsertion of the coring rods into the subsurface (Farrington et al., 1999). This method uses push rods equipped with removable core barrels so that when the rod is pushed to the desired depth, the tip can be unlocked and removed through the rod with a wire. The soil core barrel is dropped into place through the rod. The rod is advanced to collect the soil core and then removed quickly to the surface by its wire tether. Another core barrel can be inserted into the rod to collect a soil core at the next depth interval, or a blank or piezocone tip can be inserted to the end of the rod to push the rod to the next depth interval. At the CBRP, soil cores were collected at 0.3-m (1-ft) intervals to depths of 18 m (60 ft) at 16 locations (Rossabi, personal communication, 2003). The advantages to using this direct-push approach include an increased number of samples per day compared with traditional removal and reinsertion methods, as well as decreased cost and decreased waste associated with each soil coring expedition.
|
Numerical Modeling
|
|---|
Model Development
A single soil vapor extraction well in layered strata is modeled in two dimensions at the center of a cylinder of height b and radius RO (Fig. 1). Airflow converges to the SVE well at the center. No mass flux is assumed at the upper axial boundary. The TCE source is represented as a continuous DNAPL pool located at the lower axial boundary. The DNAPL evaporation is assumed instantaneous at the lower boundary (Wilson et al., 1994; Ng and Mei, 1996). The source is represented as a constant vapor concentration at the lower boundary and is placed at a distance RI from the SVE well, where RI varies from 0 to 7.6 m (0 to 25 ft). The outer radial boundary is set at a distance RO from the SVE well. Typically, this distance is set to 9 m (30 ft). Diffusion into and out of soil aggregates is also included, since this mechanism is believed to be an important source of lingering contamination observed in the field (Grathwohl and Reinhard, 1993; Ng and Mei, 1996; Arands et al., 1997; Massry, 1997; Ng et al., 1999; Werth and Hansen, 2002).
This scenario is represented by the advective–dispersive equation:
 | [1] |
where Rv,INTER is the interaggregate retardation coefficient, CvI is the vapor-phase TCE concentration in the high-permeability compartment, t is time, Deff is the effective vapor-phase diffusivity of TCE in the high-permeability compartment, vr is the radial velocity, r is the radial distance coordinate, vz is the axial velocity, z is the axial distance coordinate, k is the aggregate regime (1, 2, ..., m – 1, m), 
is the porosity of the high-permeability soil, 
k is the readius of aggregate k, DINTRA,k is the diffusivity of TCE within aggregate regime k, 
L,kII is the aqueous-phase concentration in aggregate regime k, and 
is the radial coordinate in aggregate regime k. Equation [1] is subject to the initial and boundary conditions
 | [2] |
 | [3] |
 | [4] |
 | [5] |
 | [6a] |
 | [6b] |
as developed in Switzer (2004), where C0 is the initial concentration of TCE in the high-permeability compartment, rw is the SVE well radius, and Cvsat is the saturated vapor concentration.
The SVE flow path is considered the interaggregate regime and the pore space where advective flow does not occur is considered the intraaggregate regime. Retardation in the interaggregate regime is defined as
 | [7] |
where 
is moisture content, H is the Henry's Law coefficient, Fk is the fraction of mass in aggregate regime k, 
p is the particle density, and 
CS/CV is the solid/vapor partition coefficient.The effective diffusivity used in this study corrected the molecular diffusivity of TCE in air, D0, by the porosity and moisture content in the Millington–Quirk equation:
 | [8] |
where the exponents m and n are 10/3 and 4/3, respectively, because the volume of gas-filled pores divided by the volume of the soil bed is >0.2 in this case (Schaefer et al., 1998).
For the rebound intervals, the velocity components vr and vz are set to 0 to represent soil gas contaminant concentration rebound; for an operating SVE well, an existing airflow model can be used to determine the velocity components. The final term in Eq. [1] represents the mass contribution that results from contaminant diffusion into or out of the k soil aggregate regimes.
The boundary conditions in Eq. [3] through [6] invoke symmetry at the well, no contaminant in the far field, no flux across the upper boundary, and the NAPL source emplaced at the lower boundary. The boundary condition of Eq. [6b] establishes the NAPL as a ring source, which is relevant to field conditions when a portion of the ring is considered.
Diffusion in Soil Aggregates
Contaminant diffusion into and out of soil aggregates is represented as one-dimensional spherical diffusion (Grathwohl and Reinhard, 1993; Ng and Mei, 1996; Arands et al., 1997; Ng et al., 1999; Werth and Hansen, 2002). Contaminant molecules will diffuse into the soil aggregates when the concentration outside the soil aggregates exceeds the concentration inside the aggregates. When the concentration inside the aggregates exceeds the concentration outside the aggregates, contaminant molecules will diffuse out of the soil aggregates. The soil aggregates are assumed to be water saturated, so contaminant diffusion within the soil aggregates is modeled as aqueous diffusion (Massry, 1997). At the interface between the vapor and aqueous phases, Henry's Law behavior is assumed.
Previous aggregate models that have been used in conjunction with SVE have assumed a single soil aggregate regime (Grathwohl and Reinhard, 1993; Ng and Mei, 1996; Arands et al., 1997). The current approach assumes multiple soil aggregate regimes distinguished by pore size. From Massry (1997), the equation for each aggregate regime k is
 | [9] |
subject to the initial and boundary conditions
 | [10] |
 | [11] |
 | [12] |
Equations [1] through [6] were solved by explicit finite difference with distance steps in both r and z of 7.62 cm (0.25 ft) and a time step of 8 s. Equations [9] through [12] were solved implicitly at each step of the explicit advective–dispersive solution, with concentration evaluated at five distance steps within the aggregate. This system simplifies to five equations and five unknowns, forming a five by five matrix that was solved by a Gaussian elimination procedure (Carnahan et al., 1969; Constantinides, 1987). Within each 7.62- by 7.62-cm block, all soil aggregates were assumed to contribute equally to the advective–dispersive equation.
Model Solution
The finite-difference SVE model was formulated and solved in Matlab, Release 13 (Mathworks, Natick, MA). Translation of the CBRP stratigraphy to model compartment can be found in Fig. 1. Rebound data were simulated individually for SVE 18C, 19C, and 22C, as well as the monitoring probes, assuming the same stratigraphic layering for all locations. During rebound, the concentration at each SVE well was determined by averaging the concentrations of all cell blocks at the SVE well interface. The concentration at each monitoring probe was represented by a single cell block at the respective depth for each probe.
The aggregate diffusion parameters were determined by model simulation of column desorption experiments with selected soils from the CBRP (Berler, 2001). Desorption profiles from contaminated soils from the CBRP were calibrated to a finite-difference advective–dispersive model similar to Eq. [1] that was solved for column geometry. Pore size distribution determined from Hg porosimetry was used to establish the pore regimes. A boundary condition similar to Eq. [10] formed the convergence criteria for each time step of the model, which was solved for the aggregate diffusivities DINTRA,k. The parameters determined from the column experiments that are relevant to the SVE model are listed in Table 2.
Results from rebound-only simulations of the SVE model were applied to the study at the CBRP. The advection terms were neglected, leaving just diffusion as the mechanism for mass transfer. Simulations were performed with and without the aggregate diffusion component. A DNAPL source was placed at the lower boundary and its distance from the monitoring point (SVE well or monitoring probe) was varied from 0 to 9 m (0–30 ft). This scenario was applied to the data collected during several of the rebound tests conducted at the CBRP.
|
Results and Discussion
|
|---|
Location of the DNAPL Source
Approximately 700 soil cores were collected from 16 locations at the CBRP in September 2001 (Rossabi, personal communication, 2002). Non-aqueous-phase liquid was discovered at one location in two depth intervals (Fig. 2 and 3). These data, coupled with the absence of a confirmed NAPL source at other core locations and depth intervals, suggest that the highest concentrations reside in the immediate area of the stiff clay formation that divides the vadose zone, and much less residual contamination remains below this clay layer.
Mass Removed
A TCE DNAPL source was detected in September 2001 when soil cores were extracted at the CBRP. This single core had soil concentration measurements indicative of TCE DNAPL presence (>2000 mg TCE/kg soil) in two 0.3-m (1-ft) sections. These two sections of core were located at and immediately above the stiff clay formation present at a depth of 9 m (30 ft) below the surface.
The capacity within the SVE 18–19–22 triangle, assuming it is an equilateral triangle filled with clay of porosity 0.466, is approximately 15 000 kg TCE if the DNAPL pool extends 0.6 m downward and approximately 7600 kg TCE if the pool extends 0.3 m downward. This calculation assumes that the pore space is completely saturated with TCE liquid. Based on moisture content measurements from previously extracted soil cores, the relative saturation of the clay layer is approaches 1, so more likely the TCE DNAPL source exists in discrete, disperse locations within the 18–19–22 triangle. Therefore, this capacity calculation is expected to overestimate substantially the source present within the triangle.
In 3 yr of operation, the SVE system at the CBRP has removed approximately 800 kg of TCE. Comparing this removal amount to the capacity of the SVE 18–19–22 triangle, the SVE system has removed 5 to 11% of the TCE capacity within the triangle. Considering that the capacity calculation overestimates the potential mass present near these wells, the mass removed by the SVE system at the CBRP may represent a more substantial percentage of the actual TCE mass that was present.
Soil Gas Trichloroethylene Concentration Rebound Tests
Six soil gas TCE concentration rebound tests were performed at the CBRP. For each test, the SVE system was shut down for a fixed amount of time (Test 1: 2 wk; Tests 2–5: 4 wk; Test 6: 6 mo). Subsurface soil gas TCE concentrations at the SVE wells and monitoring probes were recorded several times during each test. The first three tests are discussed in detail elsewhere (Switzer et al., 2004). In the first test, subsurface soil gas TCE concentrations rebounded to levels at or near the initial concentrations at many of the sampling locations. This test formed the baseline for comparison of subsequent rebound tests. During the 3 yr of operation, the progression of rebound tests showed a decline in observed rebound, suggesting progress in removing the TCE source from the vadose zone (Fig. 4). The small to negligible rebound observed at most locations during the sixth rebound test suggests that, while some source remains in the subsurface, a diminishing return has been reached in terms of mass removal by the SVE system.
View larger version (35K):
[in this window]
[in a new window]
|
FIG. 4. Soil gas trichloroethylene (TCE) concentration rebound tests conducted at soil vapor extraction (SVE) wells (a) 18C, (b) 19C, and (c) 22C above the stiff clay and (d) 18B, (e) 19B, and (f) 22C below this layer.
|
|
Diffusion Modeling
The extent of contamination was estimated by diffusion modeling. Two approaches were investigated. The first approach examined two-dimensional transient Fickian-type diffusion (Eq. [1–6]), less the advection and aggregate terms). The second approach included the effects of contaminant diffusion into and out of soil aggregates (Eq. [9–12]), less the advection terms). Rebound test data from the SVE wells and the soil gas monitoring probes were compared with predicted rebound profiles using each of these models. The soil gas TCE concentration achieved in approximately 14 d of the actual rebound test was converted to a distance based on each model. Many of the rebound tests in this study had very few data points and, in some cases, 14 d was the earliest reliable data point. Had more data been collected, regressions would have been possible. The SVE wells and monitoring probes above the stiff clay formation were the only locations involved in this study.
Because a conservative estimate of diffusivity was used, including corrections only for tortuosity and moisture content to account for slower diffusion of TCE in the subsurface, the resulting prediction from the two-dimensional Fickian-type case was expected to be too generous in terms of a distance estimate to the source. The inclusion of the aggregate diffusion term in the second case was expected to produce a less generous estimate of distance to the contaminant source, but some effect from the conservative diffusivity estimate was expected to remain.
The two-dimensional transient Fickian-type diffusion model applied to the first three rebound tests shows a depleting source with time. The initial condition was assumed to be a DNAPL pool completely surrounding all wells and monitoring probes in the SVE 18–19–22 triangle. The diffusion model applied to the rebound test converted the 14-d concentration at each point to a distance to source, and this distance was represented visually by a "clean" circle with this distance as its radius (Fig. 5). Comparison of the predicted remaining source (area not deemed clean by the diffusion model) to the actual location of the confirmed DNAPL suggests that the model provides a reasonable prediction of source location, with effective diffusivity offering a too-generous estimate of actual diffusion behavior. In the absence of more information for a diffusivity estimate, the basic Fickian-type diffusion model would be a useful tool for planning future coring efforts or designing a SVE system, if rebound tests were conducted at the beginning of the site characterization process.
View larger version (62K):
[in this window]
[in a new window]
|
FIG. 5. Soil gas trichloroethylened 14-d rebound concentrations translated to distance-to-source estimates. The subsurface was assumed contaminated unless diffusion modeling showed otherwise. The actual location of detected dense non-aqueous-phase liquid (DNAPL) is shown for comparison.
|
|
The Fickian-type diffusion model coupled with the aggregate diffusion model predicts a slower cleanup than the Fickian-type-only model (Fig. 5). Distance estimates in the first three tests show significant differences in the predicted "clean" areas by the two models; however, by the February 2002 test (data not shown), the predictions are indistinguishable. The advantages of using either model occur in the early stages of the SVE cleanup. The advantage of the coupled model is that it offers a more mechanistic interpretation of subsurface behavior. The trade-off is that more information is required to apply this model. Column experiments and separate model simulations of these experiments are required to determine the aggregate diffusion parameters, and the number of aggregate regimes, so more extensive characterization of the soil is required.
In the model formulation, a diminished soil gas TCE concentration rebound is attributed to a more distant DNAPL source in the subsurface. Other factors may contribute to the diminished observed rebound in the subsurface, including slow mass transfer from the residual source to the advective pathways near the observation point, sequestration of the residual source in lower permeability materials, and reduced source contamination in the subsurface. The impact of these effects may be significant and may contribute to the overall error in the model predictions, especially in the later rebound evaluations when a substantial portion of the available source in the subsurface has been removed.
Extent of Contamination Based on Mass Removal Data
The mass removal rates measured at the CBRP were correlated with distance-to-source estimates from the rebound profiles to estimate an extent of contamination. This estimate, compared with the extent of contamination proposed by the diffusion modeled rebound tests, suggests an extent of dispersion of the DNAPL source within the SVE 18–19–22 triangle.
Based on the results from the soil coring study, where DNAPL was found in a 0.6-m (2-ft) region immediately above the stiff clay formation, the mass removed by the SVE system was assumed to occupy a 0.6-m- (2-ft-) tall cylinder. The cylinder was assumed to be filled completely with porous sand, and the pore space of the sand was assumed to be filled completely with liquid-phase TCE. For comparison, a 0.3-m- (1-ft-) tall cylinder was included in the study also. The completely filled 0.6-m-tall cylinder represented the worst case. The 0.3-m cylinder represented an improved case. Because the recovered DNAPL had varying concentration in each 0.3-m interval, the third case involved assumed that the higher of these concentrations was continuous.
All removed TCE was assumed to have existed as a continuous DNAPL source before removal by SVE. The first and second cases are shown in Fig. 6. The known location of the DNAPL source is indicated and the points established as DNAPL-free by soil coring are shown for comparison. The third case (not shown) translates to a cylinder approximately 18 m (60 ft) in diameter, which has a footprint more than seven times the area of the SVE 18–19–22 triangle. The translation of the TCE mass removal from the site into an extent of contamination supports strongly the field observation that a substantial portion of the source contamination at the CBRP has been removed by the SVE system.
View larger version (50K):
[in this window]
[in a new window]
|
FIG. 6. Approximate mass removed from the C-Area burning rubble pit in 3 yr of operation expressed as (a) 0.3-m- and (b) 0.6-m-tall cylinders with pore space completely filled with trichloroethylene dense non-aqueous-phase liquid (DNAPL). For comparison purposes, the two cylinders are shown with the location and source remaining after the April 2001 rebound test as estimated by the soil vapor extraction (SVE) model's diffusion component.
|
|
Site Closure
Previous efforts at defining site closure criteria have centered on cost–benefit analysis (Hirata et al., 1998), removed fraction of total initial mass (Armstrong et al., 1994; Kaleris and Croise, 1997), or the asymptotic limit of SVE during operation (Mohr and Merz, 1995). None of these approaches are entirely appropriate for application at the CBRP because the initial mass at the CBRP is not known and the asymptotic limit achieved during operation may reflect dilution rather than source depletion.
The closure criteria for the CBRP were established around a set of soil gas TCE concentration rebound tests. The rebound test was the primary performance assessment method for the SVE system at the CBRP, with the first test forming the baseline rebound for comparison with future test results. The original study (Switzer et al., 2004) reported the results of three rebound tests. Since then, new monitoring probes were installed, a DNAPL source location was identified by soil coring, four new SVE wells were installed near the lingering hotspots, and three additional rebound tests were conducted, including data at the new locations.
The original closure strategy was developed around the rebound tests, tracking rebounds until two successive tests showed little or no rebound in the subsurface. Soil coring would be used to verify the observations of the rebound tests. Mass transfer modeling would be used to support both sets of field observations, helping to fill in the gaps of these point assessment techniques.
According to the soil gas TCE rebound tests, subsurface TCE concentrations have decreased substantially since operation began. In the first test, soil gas TCE concentrations at many of the monitoring probes and SVE wells rebounded to levels observed before operation began. Subsequent tests showed a decline in observed rebound. The first round of soil cores was collected in September 2001, several months after the third rebound test. One of these core locations confirmed a DNAPL source at the CBRP, so an additional SVE well was installed specifically at this location. Subsequent soil gas TCE concentrations and rebound tests showed a continued decline in observed concentrations, suggesting that the DNAPL source that was discovered at the CBRP was not very large. Mass transfer modeling supported these observations. The sixth rebound test showed slight rebound at a few locations, but little rebound at most locations (Fig. 4). Mass removal rates also continued to decline. Although some TCE source does remain at the CBRP, possibly still as a DNAPL, continued SVE operation was deemed economically impractical. The SVE system was scheduled for transition to passive operation in calendar year 2004, after a second round of soil cores confirmed the absence of a substantial residual source at the CBRP. In 2004, the contractor ceased active SVE operation. Instead of transitioning to passive vapor extraction, the contractor installed solar-powered blowers to continue TCE remediation at the site.
|
Conclusions
|
|---|
Field observations and model predictions suggest that a substantial portion of the TCE contamination at the CBRP has been removed by the SVE system. The predictions of a TCE diffusion model that includes the effects of slow diffusion into and out of soil aggregates correlate well with the location of a DNAPL source that was identified during the course of operation. A diminishing return of mass removal was demonstrated in the field by the low mass removal rate and absence of observed rebound during the 6-mo rebound test. Model tests support the field observations, suggesting that a substantial portion of the original contamination has been removed by SVE operation. The remaining TCE contamination is trapped in the low-permeability and SVE-inaccessible regions of the subsurface, where it will diffuse slowly to the more accessible regions of the subsurface.
|
Appendix
|
|---|
| b |
thickness (depth) of the high-permeability compartment (cm) |
| C0 |
initial concentration of TCE in the high-permeability compartment (g/cm3) |
| L,kII |
aqueous-phase concentration in aggregate regime k (g/cm3) |
| CvI |
vapor-phase TCE concentration in the high-permeability compartment (g/cm3) |
| Cvsat |
saturated vapor concentration (g/cm3) |
| D0 |
free diffusivity of TCE in air (cm2/s) |
| Deff |
effective vapor-phase diffusivity of TCE in the high-permeability compartment (cm2/s) |
| DINTRA,k |
diffusivity of TCE within aggregate regime k (cm2/s) |
| Fk |
fraction of mass in aggregate regime k |
| H |
Henry's Law coefficient [(mg/L vapor)/(mg/L liquid)] |
| k |
aggregate regime |
| r |
radial distance coordinate (cm) |
| k |
radial coordinate for each intraaggregate regime k (cm) |
| rw |
SVE well radius (cm) |
| RI |
radial distance to the inner boundary of the TCE DNAPL source (cm) |
| Rk |
aggregate radius of regime k (cm) |
| RO |
radial distance to the outer boundary of the TCE DNAPL source (cm) |
| Rv,INTER |
interaggregate retardation coefficient |
| Rv,INTRA |
intraaggregate retardation coefficient |
| t |
time (s) |
| vr |
radial velocity (cm/s) |
| vz |
axial velocity (cm/s) |
| z |
axial distance coordinate (cm) |
|
porosity of the high-permeability soil |
| p |
particle density (g/cm3) |
| CS/CV |
solid/vapor partition coefficient ([mg/kg soil]/[mg/L vapor]) |
|
|
ACKNOWLEDGMENTS
|
|---|
We gratefully acknowledge the financial support of the Consortium for Risk Evaluation with Stakeholder Participation (a cooperative agreement with the Department of Energy [DOE]) Grant DE-FG01-03EW15336 and the USEPA for the donation of the monitoring probes, as well as the assistance of colleagues Jerry Nelsen, Mike Morgenstern, Ihab Massry, Dan Berler, Mary Harris, Greg Flach, Greg Rucker, Ron Falise, John Bradley, Keith Hyde, and Johnny Simmons. Special thanks to Greg Flach and Mary Harris for the interpretations of the cone penetrometer data that led to Fig. 3 and also to Joseph Rossabi for the data related to the DNAPL coring. This paper reflects the opinions of the authors and does not reflect the policies or practices of the DOE or USEPA. The use of trade names is for identification purposes only and does not imply endorsement by either agency.
|
REFERENCES
|
|---|
- Arands, R., T. Lam, I. Massry, D.H. Berler, F.J. Muzzio, and D.S. Kosson. 1997. Modeling and experimental validation of volatile organic contaminant diffusion through an unsaturated soil. Water Resour. Res. 33:599–609.[CrossRef]
- Armstrong, J.E., E.O. Frind, and R.D. McClellan. 1994. Nonequilibrium mass transfer between the vapor, aqueous, and solid phases in unsaturated soils during vapor extraction. Water Resour. Res. 30:355–368.[CrossRef]
- Bass, D.H., N.A. Hastings, and R.A. Brown. 2000. Performance of air sparging systems: A review of case studies. J. Hazard. Mater. 72:101–119.[CrossRef][Web of Science][Medline]
- Benson, D.A., D. Huntley, and P.C. Johnson. 1993. Modeling vapor extraction and general transport in the presence of NAPL mixtures and nonideal conditions. Ground Water 31:437–445.[CrossRef][Web of Science]
- Berler, D.H. 2001. The prediction of observed diffusivity coefficients in natural organic and aquifer materials. Ph.D. diss. Rutgers Univ., New Brunswick, NJ.
- Brusseau, M.L. 1995. The effect of nonlinear sorption on transformation of contaminants during transport in porous media. J. Contam. Hydrol. 17:277–291.[CrossRef][Web of Science]
- Brusseau, M.L., N.T. Nelson, M. Oostrom, Z. Zhang, G.R. Johnson, and T.W. Wietsma. 2000. Influence of heterogeneity and sampling method on aqueous concentrations associated with NAPL dissolution. Environ. Sci. Technol. 34:3657–3664.
- Carnahan, B., H.A. Luther, and J.O. Wilkes. 1969. Applied numerical methods. John Wiley & Sons, New York.
- Constantinides, A. 1987. Applied numerical methods with personal computers. McGraw-Hill, New York.
- DiGiulio, D.C. 1992. Evaluation of soil venting application. J. Hazard. Mater. 32:279–291.[CrossRef][Web of Science]
- DiGiulio, D.C., V. Ravi, and M.L. Brusseau. 1999. Evaluation of mass flux to and from ground water using a vertical flux model (VFLUX): Application to the soil vacuum extraction closure problem. Ground Water Monit. Rem. 19:96–104.[CrossRef]
- Farrington, S.P., M.L. Gildea, and J.C. Bianchi. 1999. Development of a Wireline CPT System for Multiple Tool Usage. P2.11. In Proc. Ind. Partnerships to Deploy Environ. Technol. Conf., Springfield, VA. Natl. Tech. Inf. Serv., Morgantown, WV.
- Feehley, C.E., C. Zheng, and F.J. Molz. 2000. A dual-domain mass transfer approach for modeling solute transport in heterogeneous aquifers: Application to the Macrodispersion Experiment (MADE) site. Water Resour. Res. 36:2501–2515.[CrossRef]
- Fischer, U., C. Hinz, R. Schulin, and F. Stauffer. 1998. Assessment of nonequilibrium in gas–water mass transfer during advective gas-phase transport in soils. J. Contam. Hydrol. 33:133–148.[CrossRef][Web of Science]
- Gerke, H.H., and M.Th. van Genuchten. 1993. A dual-porosity model for simulating the preferential movement of water and solutes in structured porous media. Water Resour. Res. 29:305–319.[CrossRef]
- Grathwohl, P., and M. Reinhard. 1993. Desorption of trichloroethylene in aquifer material: Rate limitation at the grain scale. Environ. Sci. Technol. 27:2360–2366.
- Hirata, T., N. Egusa, O. Nakasugi, S. Ishizaka, and M. Murakami. 1998. Cost efficiency of subsurface remediation using soil vapor extraction and groundwater extraction. Water Sci. Technol. 37:161–168.[Web of Science]
- Kaleris, V., and J. Croise. 1997. Estimation of cleanup time for continuous and pulsed soil vapor extraction. J. Hydrol. 194:330–356.[CrossRef]
- Kaleris, V., and J. Croise. 1999. Estimation of cleanup time in layered soils by vapor extraction. J. Contam. Hydrol. 36:105–129.[CrossRef][Web of Science]
- Mackay, D.M., and J.A. Cherry. 1989. Groundwater contamination: Pump-and-treat remediation. Environ. Sci. Technol. 23:630–636.
- Mackay, D.M., R.D. Wilson, M.J. Brown, W.P. Ball, G. Xia, and D.P. Durfee. 2000. A controlled field evaluation of continuous vs. pulsed pump-and-treat remediation of a VOC-contaminated aquifer: Site characterization, experimental setup, and overview of results. J. Contam. Hydrol. 41:81–131.[CrossRef][Web of Science]
- Massmann, J.W., S. Shock, and L. Johannesen. 2000. Uncertainties in cleanup time for soil vapor extraction. Water Resour. Res. 36:679–692.[CrossRef]
- Massry, I.W. 1997. The impact of micropore diffusion on contaminant transport and biodegradation rates in soils and aquifer materials. Ph.D. diss. (Diss. Abstr. 9800280) Rutgers Univ., New Brunswick, NJ.
- Mohr, D.H., and P.H. Merz. 1995. Application of a 2D air flow model to soil vapor extraction and bioventing case studies. Ground Water 33:433–444.[CrossRef][Web of Science]
- Ng, C.-O. 1999. Macroscopic equations for vapor transport in a multi-layered unsaturated zone. Adv. Water Resour. 22:611–622.[CrossRef]
- Ng, C.-O., and C.C. Mei. 1996. Aggregate diffusion model applied to soil vapor extraction in unidirectional and radial flows. Water Resour. Res. 32:1289–1297.[CrossRef]
- Ng, C.-O., C.C. Mei, and D.W. Ostendorf. 1999. A model for stripping multicomponent vapor from unsaturated soil with free and trapped residual nonaqueous phase liquid. Water Resour. Res. 35:385–406.[CrossRef]
- Rodriguez-Maroto, J.M., C. Gomez-Lahoz, and D.J. Wilson. 1994. Soil cleanup by in situ aeration: XVIII. Field scale models with diffusion from clay structures. Sep. Sci. Technol. 29:1367–1399.
- Schaefer, C.E., R.R. Arands, and D.S. Kosson. 1998. Modeling of the gaseous diffusion coefficient in the presence of NAPL. J. Contam. Hydrol. 33:431–437.[CrossRef][Web of Science]
- Switzer, C. 2004. Evaluation of soil vapor extraction performance for trichloroethylene (TCE) remediation in layered heterogeneous soils. Ph. D. diss. (Diss. Abstr. 3117638) Rutgers Univ., New Brunswick, NJ.
- Switzer, C., T. Slagle, D. Hunter, and D.S. Kosson. 2004. Use of rebound testing for evaluation of soil vapor extraction performance at the Savannah River Site. Ground Water Monit. Rem. 24:106–118.
- USEPA. 1995. Review of mathematical modeling for evaluating soil vapor extraction Systems. EPA/540/R-95/513. Natl. Ctr. for Environ. Publ. and Inf., Cincinnati, OH.
- Werth, C.J., and K.M. Hansen. 2002. Modeling the effects of concentration history on the slow desorption of trichloroethylene from a soil at 100% relative humidity. J. Contam. Hydrol. 54:307–327.[CrossRef][Web of Science][Medline]
- Westinghouse Savannah River Company. 1997. RCRA facility investigation/remedial investigation report with baseline risk assessment for the C-Area burning/rubble pit. WSRC-RP-96-170. Westinghouse Savannah River Co., Aiken, SC.
- Wilson, D.J., J.M. Rodriguez-Maroto, and C. Gomez-Lahoz. 1994. Soil cleanup by in-situ aeration. XIX. Effects of spill age on soil vapor extraction remediation Rates. Sep. Sci. Technol. 29:1645–1671.
This article has been cited by other articles:
|
|
|
|
 
J. C. Seaman, B. B. Looney, and M. K. Harris
Research in Support of Remediation Activities at the Savannah River Site
Vadose Zone J.,
May 17, 2007;
6(2):
316 - 326.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
