(This is a sample cover image for this issue. The actual cover is not yet available at this time.)

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Assessing performance and closure for soil vapor extraction: Integratingvapor discharge and impact to groundwater quality

Kenneth C. Carroll a,⁎, Mart Oostrom a, Michael J. Truex a, Virginia J. Rohay b, Mark L. Brusseau c,d

a Pacific Northwest National Laboratory, Richland, WA, USAb CH2M Hill Plateau Remediation Company, Richland, WA, USAc Department of Soil, Water, and Environmental Science, The University of Arizona, Tucson, AZ, USAd Department of Hydrology and Water Resources, The University of Arizona, Tucson, AZ, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 13 May 2011Received in revised form 22 September 2011Accepted 4 October 2011Available online 20 October 2011

Soil vapor extraction (SVE) is typically effective for removal of volatile contaminants fromhigher-permeability portions of the vadose zone. However, contamination in lower-permeability zonescan persist due to mass transfer processes that limit the removal effectiveness. After SVE hasbeen operated for a period of time and the remaining contamination is primarily located inlower-permeability zones, the remedy performance needs to be evaluated to determine whetherthe SVE system should be optimized, terminated, or transitioned to another technology to replaceor augment SVE. Numerical modeling of vapor-phase contaminant transport was used to investi-gate the correlation between measured vapor-phase mass discharge, MFr, from a persistent,vadose-zone contaminant source and the resulting groundwater contaminant concentrations.This relationship was shown to be linear, and was used to directly assess SVE remediation pro-gress over time and to determine the level of remediation in the vadose zone necessary to protectgroundwater. Although site properties and source characteristics must be specified to establish aunique relation between MFr and the groundwater contaminant concentration, this correlationprovides insight into SVE performance and support for decisions to optimize or terminate theSVE operation or to transition to another type of treatment.

© 2011 Elsevier B.V. All rights reserved.

Keywords:Soil vapor extractionSVEConcentration reboundMass fluxRemediationNAPLVOCVadose zone

1. Introduction

Soil vapor extraction (SVE) has been the presumptiveremedy for volatile organic compounds (VOCs) in the vadosezone for approximately 15 years (U.S. EPA, 1996a). While ini-tially SVE tends to be a highly effective method, it is recog-nized that SVE operational efficiency typically becomeslimited over time primarily due to mass-transfer constraintsassociated with contaminant mass residing within lower-permeability portions of the vadose zone (Brusseau et al.,2010; Carroll et al., 2009; DiGiulio et al., 1998; Hoier et al.,2009; Oostrom et al., 2010; Switzer et al., 2004; Truex et al.,2009; U.S. EPA, 1996b; Yoon et al., 2009). For most SVE

systems, a decision point eventually develops regardingwhether to continue under the reduced-efficiency conditions,to adjust the extraction protocol, or to cease operations andpotentially switch to other remediation methods. The U.S.Army Corps of Engineers (US Army Corps of Engineers,2002) and the U.S. Environmental Protection Agency (EPA)(U.S. EPA, 2001) provide guidance for assessing transitionand closure of SVE systems. A key analysis in this process isthe determination of contaminant mass flux, or mass dis-charge, to groundwater and the resultant groundwater con-centration at monitoring locations of interest. SVE closure/transition decisions related to meeting groundwater goalsmust consider the impact of persistent vadose-zone contam-inant sources remaining after SVE termination.

The contaminant mass flux/discharge, also referred to assource strength, for a source zone is now recognized forgroundwater as a primary metric for assessing risk and

Journal of Contaminant Hydrology 128 (2012) 71–82

⁎ Corresponding author. Tel.: +1 509 371 7222.E-mail address: Kenneth.Carroll@pnnl.gov (K.C. Carroll).

0169-7722/$ – see front matter © 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.jconhyd.2011.10.003

Contents lists available at SciVerse ScienceDirect

Journal of Contaminant Hydrology

j ourna l homepage: www.e lsev ie r .com/ locate / jconhyd

Author's personal copy

remediation performance, because it relates source-zone andplume dynamics (Brusseau et al., 2007, 2008, 2011; Carrolland Brusseau, 2009; DiFilippo and Brusseau, 2008; DiFilippoet al., 2010; DiGiulio et al., 1999; Einarson and Mackay, 2001;Falta, 2008; Falta et al., 2005a, 2005b; Freeze and McWhorter,1997; ITRC, 2002; Schwarz et al., 1998; Suthersan et al., 2010;U.S. EPA, 2003). Contaminant mass discharge, rather than con-taminant mass or concentration, is most directly related to theimpact of vadose-zone contaminant sources on groundwatercontaminant concentrations. However, to date, only a fewstudies have evaluated mass-discharge behavior for persistentsources in the vadose zone (Brusseau et al., 2010; DiGiulioand Varadhan, 2001; DiGiulio et al., 1998, 1999; Truex et al.,2009).

Recently, a vadose-zone characterization method was de-veloped to quantify the overall vapor-phase contaminantmass-discharge rate emanating from persistent (e.g., diffusioncontrolled) VOC sources within the SVE treatment volumeusing data collected from cyclic SVE operations (Brusseauet al., 2010). Vapor-phase concentrations tend to increase dur-ing the no-flow period, or rebound time, due to diffusive masstransfer from persistent sources. The mass-transfer-limiteddischarge during rebound is analogous to conditions thatwould persist if the SVE system were to remain shut off (i.e.after SVE closure). Thus, it is of interest to predict the ground-water contaminant concentration that would result over timefrom this type of persistent, vadose-zone contaminant massdischarge. With this type of prediction, a relatively short-termmeasurement of vadose-zone source mass discharge could beused to evaluate the impact of the source on the groundwatercontaminant concentration, for instance, at a down-gradientcompliance well, as part of remedy decisions for the SVEsystem.

A numerical model was used to investigate the correlationbetween vapor-phase mass discharge from a persistent,vadose-zone contaminant source and the resulting ground-water contaminant concentrations. The study also evaluatedhow uncertainties in the vadose-zone source characteristics(e.g., size, location, and concentration) and different valuesfor groundwater flow rate, sorption characteristics, and re-charge rate impact this correlation. A waste site contaminatedwith carbon tetrachloride (CT) at the Department of EnergyHanford Site was used to demonstrate the methodology.

2. Methodology and case study

2.1. Case study site

At the Hanford Site in Washington State, dense nonaqu-eous phase liquids (DNAPL) consisting of carbon tetrachlo-ride (CT) mixed with lard oil, tributyl phosphate, anddibutyl butyl phosphonate were disposed at the 216-Z-9Trench (Rohay, 2007; Rohay and McMahon, 1996). A recentconceptual model developed based on multifluid flow simu-lations (Oostrom et al., 2007a) showed that CT in theDNAPL migrated primarily in a vertical direction below thedisposal site and that some CT DNAPL likely migrated intothe regional aquifer. Over time, the CT contamination withinthe more permeable sediments has been removed due toactive SVE remediation. Initial vapor concentrations of CT(1993)measured from the SVE systemdischarge had an annual

average of 15,939 ppmv (35,520–168 ppmv), which decreasedto an average of 38 ppmv (66–11ppmv) in 1996 (1 mg/L=159ppmv CT at 25 °C). Starting in 1996, SVE was operated in re-peating, annual-operation cycles of about 6 months of SVE ex-traction followed by 6 months of no extraction. Reboundconcentrations of CT in the SVE extraction system have de-creased over time, and ranged from approximately 3 to 11ppmv in 2011. This rebound behavior of CT concentrations fol-lowing the periods of no flow, and its persistence despite activeremediation, suggests that sources of CT mass remain in amass-transfer-limited region of the vadose zone (Brusseau etal., 2010). The persistent source of this CT contamination hasbeen interpreted to be located within the Cold Creek Unit(hereafter termed CCU), which is an approximately 5 m thicklower-permeability silt layer located mid-depth in the 70 mthick vadose zone. Groundwater was also contaminated withCT from the Z-9 Trench Site and other disposal areas, andhas been undergoing active remediation via extraction andtreatment at the land surface. Beneath the Z-9 Trench Site thegroundwater CT concentrations in 2011 varied between 2 and0.43 mg/L with expectations that concentrations will remainhigh for decades until the groundwater is treated. Thus, thegroundwater contaminant concentration is not a direct indica-tor for the impact of vadose-zone contaminant conditions atthe Z-9 Trench Site on the groundwater.

Fig. 1 illustrates the conceptual model for the case study sitesubsurface, representing the major sedimentary units, the gen-eral source location, and pertinentmass-transfer processes. Theapproximately 70-m-thick vadose zone consists of the perme-able Hanford Formation at the top and the permeable RingoldFormation at the bottom, which are separated by the lower-permeability CCU. The configuration shown in Fig. 1 may begenerally applicable to other sites with persistent contaminantslocated within lower-permeability zones, surrounded by clean-er, high-permeability sediments. Contaminants within thelower-permeability zones are likely to be persistent even withcontinued SVE operations, and represent a long-term sourceto groundwater (or vapor intrusion) that needs to be consid-ered for closure or transition decisions.

2.2. Measuring mass discharge

An analysis of contaminant mass discharge at the Z-9Trench Site was presented by Brusseau et al. (2010) for the cy-clic operations of the SVE system, where periods of no flow(typically 0.5 to 1 year “rebound” time) were imposed toallow diffusive mass flux to occur between periods of activeSVE operation. An example of typical CT concentration reboundduring flow interruption and decreases during vapor extractionat the Z-9 Trench Site is shown in Fig. 2. Using data such asthose presented in this figure, Brusseau et al. (2010) derivedseveral source zone mass-discharge values (M/T), includingthe average rebound mass discharge (MFr), calculated as thetotal mass of contaminant (MPV) released from the sourcezone during rebound divided by the rebound time.

The MFr term is determined based upon the assumptionthat the mass transferred from the lower permeability sourcezone into the higher permeability vadose zone during a re-bound period is the same mass that is collected from the va-dose zone pore space when the SVE system is restarted. So,the mass discharge during the rebound period can be

72 K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82

Author's personal copy

evaluated by tabulating the mass removed during extractionof the first pore volume of the vadose zone soil gas collectedby the SVE system just after it is turned back on after a re-bound period. The amount of mass (MPV) discharged duringthe no flow time is determined as the mass recovered bythe SVE system within the 1st pore volume after SVE systemrestart as (Brusseau et al., 2010):

MPV ¼Xn

1

CQTs ð1Þ

where n is the sample where one pore volume of gas has beenextracted, C is the contaminant concentration in the extractedsoil gas (M/L3), Q is the extraction flow rate (L3/T), and Ts isthe interval between sample times (T). The pore volume ofthe vadose zone accessed by the Z-9 Trench Site SVE systemwas estimated by Rohay and McMahon (1996) to be approx-imately 600,000 m3. Discussion of uncertainties associatedwith these mass-discharge terms, and of the underlying com-ponents, is presented in Brusseau et al. (2010).

Of the mass-discharge values introduced by Brusseauet al. (2010), the MFr is the most attractive for use in a

remediation quantification and endpoint analysis because(1) it is an estimate of mass discharge associated with diffu-sive mass transfer of contaminant from the lower-permeability source zone to the advective domain duringthe rebound period between each operation cycle, represent-ing the mass-discharge behavior under natural-gradient (i.e.closure) conditions, (2) an equivalent quantity can be rou-tinely obtained from numerical-model simulations of re-bound by integrating transient source-zone boundary fluxesand then dividing by the rebound time, and (3) a related av-erage rebound concentration (Cr) can be calculated by multi-plying MFr by the rebound time and then dividing by theestimated SVE pore volume. An overview of measured MFrand related Cr values for the twelve SVE cycles from 1996 to2010 at the Z-9 Trench Site is listed in Table 1. The datashow that the field measurements of source-zone mass dis-charge were collected on a relatively continuous basis usingan approximately half-year rebound time, although a few ofthe cycle rebound times were approximately one year or lon-ger (Table 1). These data provide an exceptional long-termset of field-scale source mass-discharge measurements thatfacilitate quantification of remediation progress.

Fig. 1. Conceptual model of the Hanford Site Z-9 Trench case study with a persistent CT source zone in the low-permeability CCU layer.

0

20

40

60

80

100

120

140

Time

Car

bo

n T

etra

cho

rid

e V

apo

r

July 1998 through September 1998

March 1999 through June 1999Asymptotic

Maximum Rebound

180 day

Co

nce

ntr

atio

n (

pp

mv)

Fig. 2. Example of SVE operations concentration rebound observed at theHanford 200-West Z-9 Trench (after Brusseau et al., 2010). Note 1 mg/L=159ppmv for CT at 25oC.

Table 1Measured CT rebound mass flux values (MFr) and concentration (Cr) for the12 SVE cycles at the Hanford Site Z-9 Trench Site.

SVEcycle

Time period(date in years)

Rebound time(months)

MFr(g/day)

Cr(mg/L)

1 96–97 8.2 1008.0 0.4192 97–98 9.1 1003.6 0.4633 98–99 5.9 949.4 0.2844 99–01 24.3 – –

5 01–02 5.9 1561.1 0.4676 02–04 26.5 – –

7 04–05 4.9 1366.7 0.3398 05–06 12.0 249.3 0.1529 06–07 5.9 561.1 0.16810 07–08 10.6 204.3 0.11011 08–09 5.9 183.3 0.05512 09–10 5.9 150.0 0.045

73K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82

Author's personal copy

2.3. Numerical model configuration

Numerical simulations of a variety of scenarios were con-ducted using the three-dimensional (3D) water–oil–air oper-ational mode of the STOMP simulator (White and Oostrom,2006). This fully implicit, integrated finite difference numericalmodel has been used to simulate various multiphase systems(Oostrom et al., 2005, 2007b, 2010; White et al., 2004). Forthe Hanford Site case study described in this paper, a multi-phase numerical analysis is necessary, because vapor transportcontributes significantly to contaminantmovement. For sites atwhich aqueous transport dominates (e.g., high recharge), ananalytical approach to the predictive analysis may be used(DiGiulio et al., 1999; Truex et al., 2009).

A schematic of the model domain, which is a representa-tion of the conceptual model of the Hanford Site Z-9 Trench(Fig. 1), is shown in Fig. 3. The domain has a length of ap-proximately 2000 m, a width of 1200 m and a height of86 m, and was discretized into 64×56×66 grid cells. The nu-merical model configuration shown in Fig. 3 consisted of a5 m thick lower-permeability layer, representing the CCU,surrounded by high-permeability sediments representingthe Hanford Formation and the Ringold Formation aboveand below the CCU, respectively. The water table was locatedat 20 m above the model base. Hydraulic properties of thetwo porous media used in the simulations and other modelparameters are listed in Table 2. The property values wereprimarily obtained from Hanford Site data bases, as reportedby Oostrom et al. (2010) and Truex et al. (2009). The sourcezone was represented by a rectangular block centrally locat-ed in the CCU lower-permeability layer. Taking advantage ofa plane of symmetry through the middle of the source zoneparallel to the groundwater flow direction, only half of thefull domain was included in the simulations. Considerable re-finement was imposed near the water table and around thesource zone. The final discretization was obtained after astandard procedure where several simulations with in-creased refinements were conducted until no changes wereobserved in the source mass flux and concentration mixingbelow the water table.

The initial and boundary conditions were designed to repre-sent the conceptual model including the dominant processes(Fig. 1). A constant atmospheric gas pressure was imposed atthe top of the domain without the consideration of baromet-ric pressure fluctuations. CT mass was allowed to be trans-ported across the lateral and upper boundaries in the gasphase above the water table and across lateral boundariesin the aqueous phase below the water table. Groundwaterflow in the saturated zone was imposed with a constantflux boundary condition over the lower 20 m at the upgradi-ent side of the domain and a specified pressure boundarycondition at the downgradient side. To establish an essentiallyhorizontal water table at 20 m from themodel base and similarwater desaturation behavior as a function of elevation through-out the domain, the hydraulic conductivity of the saturatedzone was assumed to be four orders of magnitude largerthan the high-permeability sediments. This assumption allowsfor a straightforward calculation of mass fluxes into thegroundwater and consistent comparisons between the consid-ered scenarios. A uniform and constant recharge rate was im-posed at the top of the domain with a constant-flux boundarycondition.

The model simulations consisted of two phases. In thefirst phase, the initial conditions for CT vapor transportwere computed by allowing saturated and unsaturated flowto reach steady-state conditions over a period of10,000 years. After computing these initial steady-state flowconditions, the second phase of the simulations was con-ducted by restarting the model with a constant CT vapor con-centration in the source area to simulate transient vapormass discharge from the source zone and transport in the va-dose zone and groundwater. The source vapor concentrationwas maintained at a constant value throughout the sourcezone during the simulation through equilibrium interphasepartitioning from an immobile organic liquid phase emplacedin the source zone. The presence of organic liquid within the

Fig. 3. Numerical model configuration of the Z-9 Trench Site illustrating thelocation and distribution of the persistent CT source zone in the low-permeability CCU layer.

Table 2Material properties and other parameter values used for the Hanford Site Z-9Trench Site simulations.

Parameter Value

CT diffusion coefficient in water (m2/day)a 8.25×10−5

CT diffusion coefficient in air (m2/day)a 0.715Henry's law coefficient for CT (−) 0.813CT aqueous solubility (mg/L) 800

High permeability sedimentHydraulic conductivity (cm/s)b 5.73×10−3

Van Genuchten αa (1/cm) 1Van Genuchten na (−) 2.5Irreducible water saturationa (−) 0.0583Porositya (−) 0.3

Low permeability sedimentHydraulic conductivity (cm/s)b 1.38×10−4

Van Genuchten αa (1/cm) 0.1Van Genuchten na (−) 2.5Irreducible water saturationa (−) 0.0583Porositya (−) 0.3

a Truex et al. (2009); Oostrom et al. (2010).b Khaleel et al. (2001).

74 K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82

Author's personal copy

source zone was maintained through slow injection duringthe simulation. To ensure that diffusion and tortuosity werenot affected by the organic liquid saturation in the sourcezone, the dependency of these parameters on organic liquidsaturation was ignored. These imposed initial conditions areconsistent with a system where the gas phase contaminationhas been removed by SVE from the entire vadose zone, ex-cept for within the source zone. The CT transport simulationswere conducted for a time period of 1000 years, althoughsteady-state conditions were often reached within10–50 years. The SVE process itself was not simulated.

Model outputs included CT gas and aqueous concentra-tions, as well as vapor mass discharge from the source zoneas a function of time. Groundwater impacts were assessedas the groundwater concentrations mixed over a 10 m aqui-fer thickness, representing concentrations averaged over thewell-screen interval of a monitoring compliance well. Theconcentration in this well was computed as the average con-centration over the grid cells in the upper 10 m of the aquiferat the location where this value was at a maximum. At a re-mediation site, the location of compliance wells may be se-lected based on other factors, and the numerical analysismay be adjusted accordingly.

2.4. Analysis approach

A numerical simulation matrix was developed to repre-sent different conceptual models for the vadose-zone sourceconfiguration and properties (e.g., source size, location, andconcentration), contaminant properties, and other hydrogeo-logic conditions (e.g., groundwater flow rate). The simulationresults were used for evaluating the impact of uncertainty inthese parameters on the resultant predicted long-termgroundwater concentration at the selected compliance well.

For this demonstration of the analysis approach, the se-lected simulated scenarios included a Base Case and eight ad-ditional simulations investigating the effects of sourcedimension, sorption coefficient, recharge, and groundwaterflow Darcy velocity (Table 3). Other potentially uncertainprocesses or parameter variability not evaluated here includebarometric pressure fluctuations, hydraulic properties, dis-persivity values, and different distributions of hydraulic con-ductivity. In this approach, the assumptionwas made that the

subsurface of the Hanford Site Z-9 Trench is sufficiently char-acterized such that a conceptual model as depicted in Fig. 1could be derived with a reasonable level of confidence, in-cluding the hydraulic properties of the major porous media.In this context, it is recognized that the capillary fringe con-figuration may have a considerable effect on mass flux intothe water table (Klenk and Grathwohl, 2002; Mccarthy andJohnson, 1993), including proper discretization of this zonein numerical models (Maier et al., 2008a, 2008b; Oostromet al., 2010). However, given that capillary fringe effects oncontaminant transport are not yet fully understood, varia-tions in capillary fringe configurations were not included inthe current analysis approach.

The Base Case configuration used a source size equivalentto the footprint of the Z-9 Trench (20×10×5 m), as proposedby Oostrom et al. (2007a) who, using numerical simulations,observed minimal lateral movement of the DNAPL betweenthe disposal trench and the CCU. The Base Case also used asorption coefficient (Kd) of 0.0 mL/g, consistent with the find-ings of Wellman et al. (Wellman et al., 2007) for sedimentscollected near the Site, a recharge rate of 4 mm/year, and agroundwater Darcy velocity of 5.45×10−3 m/day as listedin the Z-9 Trench Site feasibility study (DOE/RL-2007-27,2011), which is believed to be representative for this casestudy despite being relatively low, in general.

The investigated range in source size included a sourcesize one order of magnitude larger than for the Base Case(Medium Source scenario: 50×40×5 m) and a large sourcezone approximately equal in area to the footprint of the SVEwell system (Large Source scenario: 150×150×5 m). Con-sideration of increasing source zone size was based on thepossibility of lateral spreading above and into the CCU duringwaste infiltration. Uncertainty in source size and distributionis generally a significant limitation at most sites, and the eval-uation of multiple source sizes herein was used to considerthe impact of this uncertainty on the results. The range inKd values included the Large Kd scenario using a reasonablevalue for several types of sediments based on a literature re-view (Truex et al., 2001) that was applied in previous Z-9Trench Site modeling studies (e.g., Oostrom et al., 2007a),and a value that is one order of magnitude smaller (SmallKd scenario), which was intended to represent semi-arid/arid sites with low organic matter. The imposed rechargerange was also selected from the Z-9 Trench Site feasibilitystudy and includes the Small Recharge scenario, representingthe presence of a surface barrier, and the Large Rechargescenario, simulating a condition where the surface is deliber-ately kept vegetation free. The two alternative Darcy ground-water velocities listed in Table 3, although higher than theBase Case, reflect a typical range of values relevant to severalother sites (DOE/RL-2007-27, 2011; Oostrom et al., 2010;Truex et al., 2009).

Each of the scenarios listed in Table 3 was simulated forthree different CT source vapor concentrations (0.1, 1, and10 mg/L), to evaluate source concentration impacts onvapor mass discharge. Changes in vapor pressure of theemplaced organic liquid were used to generate the differ-ences in source vapor concentration. The consideration ofthese relatively low concentrations, compared to the vaporpressure of neat CT (~650 mg/L), is consistent with the mix-ture nature of the Hanford Site organic liquid and the fact

Table 3Summary of simulation scenarios. For each scenario, simulations were con-ducted with 0.1, 1.0, and 10 mg/L CT source concentrations.

Scenarioname

Sourcedimensionsa

(m)

Darcyvelocity(m/day)

Kd

(mL/g)Recharge(cm/year)

Base case 20×10 5.45×10−3 0.0 0.4Medium source 50×40 5.45×10−3 0.0 0.4Large source 150×150 5.45×10−3 0.0 0.4Small Kd 20×10 5.45×10−3 0.02 0.4Large Kd 20×10 5.45×10−3 0.2 0.4Small recharge 20×10 5.45×10−3 0.0 0.05Large recharge 20×10 5.45×10−3 0.0 6.3Small Darcy velocity 20×10 3.00×10−2 0.0 0.4Large Darcy velocity 20×10 3.00×10−1 0.0 0.4

a Height of source zone is 5 m for all simulations.

75K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82

Author's personal copy

that the liquid has likely undergone considerable weatheringsince its disposal ~40–50 years ago. The source-zone aqueousconcentrations were maintained at interphase partitioningequilibrium with the source vapor concentration throughHenry's Law.

It should be noted that for consistency in the analysis, thesource size and concentration were constant during the tran-sient simulations, which may be considered to be unrealisticfor remediation operations. The constant-source assumptionwas imposed, because it leads to conservative results, as de-creases in source area and concentration would also reducethe vapor discharge and long-term groundwater impacts.Natural attenuation of the source may also occur, but thetime scale over which it will occur is likely to be relativelylong with respect to typical remediation planning periodsfor sites with persistent sources (e.g., Devlin et al., 2002). Itshould also be kept in mind that the focus in the analysis isto determine closure criteria at sites that would likely havepersistent sources even after closure occurs.

The analysis of the numerical results has two major com-ponents. First, short-term average rebound mass-dischargevalues were computed to allow for a direct comparisonwith the field-derived mass discharges and concentrations.These short-term vapor mass discharge predictions, concep-tually equivalent to the field-measured MFr values, werecomputed by integrating the mass released from the sourcezone over the specific rebound period and then dividingthat number by the duration of the rebound period. Consis-tent with SVE rebound periods typically used at the HanfordSite, numerical ½-year and 1-year MFr values were comput-ed. In direct association with these short-term mass-discharge values, average rebound concentration predictions,equivalent to the field Cr measurements, were obtained bydividing the integrated mass released from the source zoneduring the rebound period by the estimated pore volume of

600,000 m3. Field-measured rebound and model-simulatedMFr and Cr relations can then be compared with the intentto show the relative position of the various scenarios with re-spect to the field observations.

In the second part of the data analysis, long-term predic-tions of the CT distributions in the groundwater allow forthe development of correlations between short-term vapor-phase concentration rebound mass discharge and futuregroundwater concentrations at designated groundwatercompliance wells. This endpoint approach thus quantifiesthe functional relationship between the short-term (mea-sureable) vapor mass-discharge rate in the vadose zone andthe long-term (predicted) maximum groundwater concen-trations, taking into account processes such as mass transferand attenuation in both the saturated and unsaturated portionsof the subsurface. This relationship can be used to directly as-sess SVE remediation progress over time and to determinethe level of remediation in the vadose zonenecessary to protectgroundwater. In addition, regression analysis may be used todetermine the short-term measureable MFr required to obtainlong-termgroundwater concentrations below regulatory limitsrepresenting the SVE closure criteria.

3. Results and discussion

3.1. Simulated CT behavior in the gas phase and groundwater

The general behavior of the simulated CT plumes in boththe vadose zone and groundwater is similar for all casesthat were considered. For example, Figs. 4 and 5 illustrateCT transport from the CCU source zone for the 1 mg/L sourceconcentration (Base Case) simulation under conditionswhere starting concentrations in the higher-permeabilityHanford and Ringold portions of the vadose zone were zero.These initial conditions for the higher-permeability zones ap-

-35

-30

-25

-20

-15

-10

-5

0

5

10

15

20

25

30

35

0.00 0.20 0.40 0.60 0.80 1.00 1.20

Vapor VOC Concentration (mg/L)

Dis

tan

ce F

rom

So

urc

e Z

on

e (m

) A

lon

g

Ver

tica

l Tra

nse

ct in

So

urc

e A

rea

17.5

22.5

27.5

32.5

37.5

42.5

47.5

52.5

57.5

62.5

67.5

72.5

77.5

82.5

87.5

Ele

vati

on

Ab

ove

Bas

e o

f M

od

el (

m)

0.0

0.5

1.0

3.0

5.0

10.0

50.0

500.0

1000.0

Time Years

Fig. 4. Base Case scenario (1 mg/L source) simulation results of rebound or post-SVE transient vapor concentration vertical profiles for a transect through the centerof the source zone.

76 K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82

Author's personal copy

proximate the removal of contaminant from these zonesprior to either a rebound period or prior to shut down ofthe SVE system.

In Fig. 4, simulated vadose-zone vapor concentrations forseveral simulation times are presented as a function of timealong a vertical profile through the center of the source area.Initially, the concentration gradient is highest between theconstant concentration source zone in the lower-permeabilitylayer and the CT-free, higher-permeability sediments aboveand below the source zone. The vapor concentrations abovethe lower-permeability source zone increase rapidly over thefirst few years due to the large concentration gradient betweenthe constant source and the overlying, initially clean high-permeability zone. The upward diffusive mass transfer out ofthe source zone approaches a steady-state concentration distri-bution after approximately five years with a concentration gra-dient between the source zone and the assumed concentrationof zero at the ground surface. Although the concentration gra-dient between the source zone and the surface is constant,the steady-state concentration distribution is not linear be-cause of the 3D nature of the domain configuration.

Fig. 4 also shows that the development of pseudo steady-state conditions is considerably slower between the sourcezone and the water table and that the concentration distribu-tions are different. Below the source zone, the diffusion processis initially governed by concentration gradients similar to thoseabove the source zone. However, once vapors approach thewater table, limited mass transfer from the vadose zone tothe groundwater controls the development of the concentra-tion profile (Truex et al., 2009), and a non-zero vapor phaseconcentration at the water table persists. Though not directlyshown, the aqueous concentration of CT in the vadose zonepore water is in equilibrium with the vapor phase concentra-tion (i.e., as calculated by Henry's Law).

The simulated steady-state groundwater plume, aqueousCT concentrations at the water table, and aqueous CT concen-tration mixed over the upper 10 m of the aquifer in the verti-cal x–z plane bisecting the source zone are shown for the1 mg/L source concentration Base Case in Fig. 5. The cross-sectional concentration plot indicates that a plume developsin the direction of the groundwater flow and that spreadingoccurs due to hydrodynamic dispersion. The contour plotalso indicates that CT concentrations at the water table be-come lower than the depth-integrated concentrations withdowngradient distance from the source zone, primarily as aresult of re-volatilization from the groundwater into the va-dose zone. This observation is supported by the line plotsshowing a rapid decrease in water table concentration at ei-ther end of the source and a more gradual decrease in themixed downstream concentrations. The general behaviordepicted in Figs. 4 and 5 is relevant to the scenarios investi-gated in this study, but may be somewhat different forother sites depending on the vadose zone and source condi-tions and the groundwater Darcy velocity.

A summary of the main simulation results for all nine sce-narios is presented in Table 4, including the vapor mass-discharge rate (MFr) from all surfaces of the source zone inte-grated for both half-year and one-year rebound periods, andthe associated CT vapor concentration (Cr). Table 4 also con-tains steady-state values for vapor mass discharge and themaximum long-term CT concentrations mixed over 10 mbelow thewater table. Comparison of the numerically obtainedMFr and Cr values (Table 4) to fieldmeasurements (Table 1) forsimilar rebound times and simulated rebound times is pre-sented in the next section. The maximum mixed groundwaterconcentrations obtained at the end of the 1000 year simulationperiod were subsequently used to assess the impact of vadose-zone source configurations on groundwater quality.

Fig. 5. Base Case scenario simulation (1 mg/L source) results of the aqueous concentration contours and values along a lateral transect in the groundwater flowdirection with a comparison of concentrations at the water table to those mixed through a vertical depth average over 10 m (e.g., compliance well).

77K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82

Author's personal copy

The results in Table 4 show that for each scenario, the re-lations between source concentrations and the various mass-discharge values and between source concentrations andmaximum mixed concentrations are linear. This linearity isexpected as the range in simulated source concentrationswould not yield considerable nonlinearities due to densitydriven vapor advection, which should be considered forhigher concentrations (Oostrom et al., 2010). Although thesource sizes for the Medium Source and Large Source scenar-ios are about, respectively, 10 and 100 times larger than forthe Base Case, the resulting mass discharges and maximummixed aqueous concentrations do not show the same ratios.The nonlinear increase with source size is directly related tothe 3D nature of the CT gas plumes, the size of the vadosezone, and proximity of the ground surface. As the sourcesize increases, the ratio of the CT gas plume volume and thesource volume decreases and will approach a constant valuewhen the source size becomes relatively large. The plumesize is not allowed to increase directly with the source size,because it is limited by the thickness of the vadose zoneand the proximity of the ground surface (zero concentrationboundary condition). For the range in source sizes investigat-ed in this analysis, the volume ratios are such that linearity inmass discharge has not been achieved.

The increased Kd values in the Small and Large Kd scenariosraise short-term mass-discharge rates, because concentrationgradients remain larger for a longer period of time comparedto the Base Case as the discharged CT is both transported inthe gas phase, and it is also sorbed to the solid phase. Overtime, the gradients diminish and steady-state mass-dischargevalues become similar to those for the Base Case.

The alternative recharge scenarios yield increased mass-discharge rates for the Small Recharge scenario and de-creased rates for the Large Recharge scenario compared tothe Base Cases. In general, water saturations in the lower-permeability source zone increase with increasing rechargerates while, consequently, gas saturations and effective diffu-sion in the gas phase decrease, leading to reductions in mass-discharge rates. Conversely, when recharge rates increase,more dissolved CT is being advectively transported into thegroundwater, resulting in larger mixed aqueous concentra-tions below the water table.

Finally, increases in the Darcy velocity have a limited im-pact on the mass-discharge rate, because the fraction trans-ported across the water table is relatively small for thisparticular vadose-zone configuration. However, maximummixed groundwater concentrations are reduced when theDarcy velocity increases due to increased dilution.

The temporal behavior of vapor mass-discharge rates as afunction of time for the Base Case and all other 1 mg/L sourceconcentration scenarios is presented in Fig. 6. As was shownin Table 4, the mass discharge relations for the Base Case sim-ulation are proportional to the imposed source concentra-tion. The Large Source and Medium Source scenariosproduce larger vapor mass discharge as a result of the in-creased source volumes. Also consistent with the data inTable 4 are the close proximity of the results for the 1 mg/Lsource concentration Base Case and the alternative Kd and re-charge scenarios. Comparing the initial and long-term mass-discharge rates, all of the scenarios showed a decrease of be-tween 1 and 2 orders of magnitude. Over time with no addi-tional SVE (e.g., closure), the vapor concentration increases in

Table 4Summary of simulation results for each of the scenarios.

Scenario ID Source vaporconcentration

Cr 1/2 year(mg/L)

Cr 1 year(mg/L)

MFr 1/2 year(g/day)

MFr 1 year(g/day)

Steady state MF(g/day)

Steady state maxgroundwater concentrationmixed over 10 m (μg/L)

Base case 0.1 mg/L 0.0004 0.0007 1.43 1.18 0.63 1.7Base case 1 mg/L 0.0043 0.0072 14.31 11.79 6.05 16.1Base case 10 mg/L 0.0435 0.0719 143.3 118.1 61.2 170.0Medium source 0.1 mg/L 0.0022 0.0034 7.37 5.65 2.08 3.9Medium source 1 mg/L 0.0224 0.0344 73.72 56.48 20.85 38.9Medium source 10 mg/L 0.2239 0.3443 738.17 565.63 209.15 402.7Large source 0.1 mg/L 0.0186 0.0274 61.48 44.99 11.61 7.9Large source 1 mg/L 0.1865 0.2739 614.86 450.00 116.17 79.1Large source 10 mg/L 1.8671 2.7429 6155.3 4505.9 1162.4 805.2Small Kd 0.1 mg/L 0.0004 0.0007 1.48 1.22 0.63 1.7Small Kd 1 mg/L 0.0045 0.0074 14.83 12.17 6.05 16.2Small Kd 10 mg/L 0.0450 0.0742 148.5 121.9 59.4 170.0Large Kd 0.1 mg/L 0.0006 0.0009 1.85 1.49 0.63 1.7Large Kd 1 mg/L 0.0056 0.0090 18.50 14.86 6.04 16.2Large Kd 10 mg/L 0.0562 0.0906 185.2 148.8 59.4 177.2Small recharge 0.1 mg/L 0.0005 0.0008 1.65 1.36 0.76 1.6Small recharge 1 mg/L 0.0050 0.0083 16.49 13.63 6.97 14.7Small recharge 10 mg/L 0.0501 0.0831 165.1 136.5 61.8 136.9Large recharge 0.1 mg/L 0.0003 0.0006 1.14 0.94 0.50 3.7Large recharge 1 mg/L 0.0035 0.0057 11.43 9.43 4.75 34.9Large recharge 10 mg/L 0.0347 0.0575 114.4 94.4 43.8 336.2Small Darcy velocity 0.1 mg/L 0.0004 0.0007 1.43 1.18 0.64 0.9Small Darcy velocity 1 mg/L 0.0043 0.0072 14.31 11.79 5.96 8.5Small Darcy velocity 10 mg/L 0.0435 0.0719 143.3 118.1 59.5 89.4Large Darcy velocity 0.1 mg/L 0.0004 0.0007 1.43 1.18 0.64 0.3Large Darcy velocity 1 mg/L 0.0043 0.0072 14.31 11.79 6.02 2.7Large Darcy velocity 10 mg/L 0.0435 0.0719 143.3 118.1 60.1 28.5

78 K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82

Author's personal copy

the high-permeability zone due to the diffusive flux out ofthe source zone, resulting in a decreased concentration gradi-ent out of the source zone and associated corresponding de-crease in the diffusive vapor-discharge rate. Fig. 6 illustratesthat the initial decrease occurs primarily in the first10 years, after which the mass-discharge rates are approxi-mately constant. This behavior, visible for all scenarios, sug-gests that long-term groundwater impacts are likely to becontrolled by the asymptotic discharge rates. However, typicalfield measurements, such as the ones listed in Table 1 for theZ-9 Site, will only feasibly capture short-term (e.g., ½–1 year)vapor-discharge behavior from a rebound test. Thus, a predic-tive analysis is needed to estimate long-termbehavior and com-pute groundwater concentrations at compliance locations.

3.2. Comparison of simulated results to field vapor-dischargemeasurements

Fig. 7 combines the ½-year field measured rebound andmodel-simulated rebound vapor mass-discharge rates (MFr)and CT concentrations (Cr). A positive correlation is evident forthe ½-year field rebound data, indicating that the field concen-trations are proportional to vapor-discharge rates. Although,by definition, the simulated MFr and Cr are also linearly corre-lated (Cr=MFr×182.5 rebound days/600,000 m3 pore vol-ume), plots like Fig. 7 have value because they allow a directcomparison of numerical and field data and visually show therelative position of the various simulation scenarios (e.g., com-bination of source concentration, size, recharge, Kd, etc.) withrespect to the field observations.

Most recent field data and simulations have ½-year MFrvalues below 200 g/d, and Fig. 7b focuses on this lower rangeto examine trends around thesemost recent field observations.The figure shows that simulations for all of the 10 mg/L smallsource zone scenarios produce mass-discharge rates relativelyclose to the recent field observations, indicating the limited ef-fects of recharge, sorption, and groundwater velocity for the

ranges investigated in this analysis. However, aMediumSourcescenariowith a source concentration between 1 and 10 mg/L ora Large Source scenario with a source concentration between0.1 and 1 mg/L also produce simulated results close to thefield observations. This type of comparison of simulated andfield results can be used to identify conceptual model alterna-tives (simulation scenarios) that most closely represent thefield measurements of mass discharge and concentrationdata. The best-fit scenarios can then be used to evaluate the im-pact to groundwater, as described below. If the range ofgroundwater impacts associated with these scenarios (or dueto conceptual model uncertainty) is too large to support a deci-sion, then additional characterization to narrow the range ofconceptual model possibilities would be warranted.

For contaminant sources in lower-permeability regions, thedesired impact of SVE remediation is to propagate from a start-ing point with an initially higher MFr and Cr along this correla-tion line in the direction toward the origin (lowerMFr and Cr) ifremediation is decreasing the source size or concentration. Thepertinent remediation closure question is: How low does MFrhave to be to protect groundwater? Note that the trend of theperiodic rebound test MFr results for the case study site is to-ward the origin (with the exception of the 1998–1999 yearmeasurement), suggesting that the source size or concentra-tion has been decreasing (between 2000 and 2010) due tothe SVE remediation. Using this type of plot, periodic reboundtest data can provide an indication of remediation system effi-ciency. For instance, if the trend towards the origin with SVEoperation stagnates at levels that do not meet the remediationgoal, sites should consider a transition to alternative remedia-tion methods or a modification of current SVE operations.

3.3. Performance and closure assessment

Long-term predictive simulations can be used to quantifythe groundwater impacts associated with persistent sourcesin the vadose zone using a correlation of the relationship

0.1

1.0

10.0

100.0

1,000.0

10,000.0

0.001 0.010 0.100 1.000 10.000 100.000 1000.000Rebound or Closure Elapsed Time (years)

Vap

or

Mas

s D

isch

arg

e R

ate

(g/d

ay)

Base Case (0.1 mg/L) Base Case (1 mg/L)

Base Case (10 mg/L) Medium Source (1 mg/L)

Large Source (1 mg/L) Small Kd (1 mg/L)

Large Kd (1 mg/L) Small Recharge (1 mg/L)

Large Recharge (1 mg/L)

Steady-State Vapor Discharge

Integrated Area for Nominal1/2 year Vapor Discharge

Fig. 6. Simulated transient rebound or post-SVE transient vapor discharge rates for selected scenarios.

79K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82

Author's personal copy

between a short-termMFr and a long-termmaximum ground-water concentration. The correlation of short-term (e.g.,½ year) MFr with long-term maximum mixed groundwaterconcentration for each of the simulated scenarios is presentedin Fig. 8. The latest (09–10) measured MFr for the case studysite (~150 g/d) is represented by the dashed vertical line, andthe hypothetical remediation goal of 5 μg/L for groundwater isrepresented by the horizontal line. The results presented inFig. 8 show that the correlation between short-term vapor-discharge rate and long-term groundwater impact for a specificsimulation scenario (e.g., conceptual model alternative) islinear. For a given site, simulations can be conducted for plausi-ble conceptual model scenarios for a relevant range of MFrvalues. Because of the linearity in the correlation, a regressionanalysis can be conducted using the simulation results andused to compute the MFr value that is predicted to meet thegroundwater remediation goal (e.g., MCL concentration). The re-gression analysis enables calculation of the conditions that arecompliant without the need to conduct trial and error simula-tions to find theMFr that meets the goal.

An example regression analysis and resulting correlationequation is shown for the Base Case scenario in Fig. 8. This typeof regression analysis can then be used to determine that asource MFr of approximately 4 g/d, measured for a ½-year re-bound period, will result in a maximum groundwater contami-nant concentration that would meet a remediation goal of5 μg/L over a mixing zone depth interval of 10 m. For the pur-pose of this study, the groundwater remediation goal of 5 μg/Lwas arbitrarily imposed at the location where the simulationpredicted the highest groundwater concentration. Note that dif-ferent simulations scenarios can result in similar MFr and Crvalues as for the Base Case scenario. Thus, it is important tohave sufficient characterization data to identify the plausibleconceptualmodel for the site. For instance, sourceMFrof approx-imately 10 and 40 g/d, measured for a ½-year rebound period,would also meet the remediation goal for the medium andlarge source sizes.

Based on the historical performance of the SVE system, thetime required for continued SVE operation to reduce MFrfrom the current value of 150 g/d to a target value that meetsthe groundwater remediation goal can be computed using a re-gression analysis as presented by Brusseau et al. (2010). In thisanalysis, the regression equation for MFr (g/d) field data fromTable 1 with a ½-year rebound time is:

MFr ¼ 1899e−0:189 tð Þ ð2Þ

with a correlation coefficient (R2) of 0.65, where t is totaloperations time in years. As an example of this process, reduc-ing the current 150 g/d MFr to 4 g/d, the MFr that meets theremediation goal for the Base Case scenario, would require ap-proximately 20 additional years based on continuation of thecurrent SVE operation procedure.

3.4. Uncertainty considerations

The predictive modeling analysis described above pro-vides quantitative input to selection of SVE remediation clo-sure criteria based on the correlation of a measured MFr inthe vadose zone to a long-term groundwater impact. Predictivemodeling results, as depicted in Fig. 8, are also useful to evalu-ate how uncertainties, associated with the site conceptualmodel as described by the simulation scenario variations, im-pact this correlation.

These results illustrate that source strength, orMFr, is highlydependent on both the source distribution or size and thesource concentration. The source concentration has the mostsignificant impact on results because it is important with re-spect to the magnitude of both ½-year MFr and long-termgroundwater concentration. Increasing source size for a fixedsource concentration created considerable MFr increases withonly limited groundwater concentration increases. Increasesin groundwater velocity caused direct decreases in groundwa-ter concentrations through dilution, and increased rechargecaused small decreases in MFr with associated increases ingroundwater concentration. Alternatively, sorption created aminor increase in the MFr with negligible long-term change ingroundwater concentration.

These trends illustrate the relative impact of either uncer-tainty or variability in site specific conditions, and are important

Fig. 7. Correlations of simulated ½-year MFr versus Cr with a comparison tofield measurements (red square symbols) from the Hanford Site Z-9 TrenchSite ½ year rebound data for (a) MFr values less than 2000 g/d, and (b) MFrvalues less than 200 g/day. The field measurement date (years) of the re-bound cycles (Table 1) are presented in parenthesis for the Z-9 Trench Sitedata, and simulation source concentrations are in brackets. (For interpretationof the references to color in this figure legend, the reader is referred to the webversion of this article.)

80 K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82

Author's personal copy

to consider for site characterization and remediation operationdecisions. As an example, reducing recharge at a site (e.g., sur-face cover installation) with significant recharge could poten-tially shift the site-specific relationship to a decreased level ofgroundwater impact such that the MFr required for closurecould be reached faster or with a more limited effort. However,the results underscore the importance of site characterizationinformation. For example, with characterization data that re-duces uncertainty in the source size, calculation of the sourceconcentration from site MFr measurements will be improvedwith a corresponding improvement in the estimated ground-water contaminant concentrations and prediction of the MFrthat will meet the target groundwater contaminant concentra-tion. Because of this strong correlation of vadose-zone sourcecharacteristics and resultant groundwater contaminant con-centrations, data are needed to describe the distribution of con-taminantswithin the vadose zone at the time of the decision forSVE closure/transition as recognized by the USACE (US ArmyCorps of Engineers, 2002) and EPA (U.S. EPA, 2001).

There are inherent uncertainties in applying a predictive ap-proach to evaluating fate and transport of contaminants in thesubsurface.Whenever possible, thedirect evaluation of ground-water impacts associated with persistent sources in the vadosezone should be evaluated. However, use of groundwater datamaynot be viable depending on the conditions. Due to the tran-sient nature of transport processes in both the vadose zone andgroundwater system, a lag time exists between changes insourceMFr in the vadose zone and the associated change in im-pact to groundwater. Additionally, many sites will have sourceswithin the groundwater system, and thus groundwater concen-trations will be influenced by mass flux from both vadose-zoneand groundwater sources. In fact, SVE systems may remediatethe vadose zone prior to the remediation of the groundwatersystem. These situations make direct evaluation of groundwa-ter impact associated with vadose zone sources problematic.

4. Summary and implications

The predictive analysis approach used for this study dem-onstrates a correlation between vadose-zone volatile contami-nant sources and groundwater contaminant concentrations. Inparticular, the study examined vadose-zone contaminantsources that are recalcitrant to SVE treatment. While thesesources may persist, the need for their treatment depends inpart onwhether or not groundwater will be negatively impact-ed in the absence of treatment. For a specific site scenario (con-ceptual site model) and the transport processes considered inour analysis, there is a linear function that describes thegroundwater contaminant concentration that results from avadose-zone volatile contaminant source as quantified by thevapor-phase mass discharge from the source, MFr. The MFrcan be measured from analysis of data from cyclic operationof the SVE system. For a unique relation between MFr and thegroundwater contaminant concentration, site properties andsource characteristics must be specified. However, once thecorrelation is established either for a specific site conditions,or for a range of conditions such as in a sensitivity analysis,the correlation provides insight into SVE performance, andsupport for decisions to optimize or terminate the SVE opera-tion or to transition to another type of treatment.

Acknowledgments

The Pacific Northwest National Laboratory is operated byBattelle Memorial Institute for the Department of Energy(DOE) under Contract DE-AC05-76RL01830. The authorswould like to thank the U.S. Department of Energy — Office ofEnvironmental Management, EM-32 Office of Groundwaterand Soil Remediation, and the U.S. Department of DefenseEnvironmental Security Technology Certification Program(ER-201125), for their support of this research.

y = 1.189x0.996

R2 = 1.000

0.1

1.0

10.0

100.0

1,000.0

1 10 100 1,000 10,000

MFr Over 1/2 Year Rebound Time (g/d)

Max

imu

m 1

0 m

Mix

ed G

rou

nd

wat

er V

OC

C

on

cen

trat

ion

(µg

/L)

Base Case

Medium Source

Large Source

Small Kd

Large Kd

Small Recharge

Large Recharge

Small Darcy Velocity

Large Darcy Velocity

MCL = 5µg/L

Darcy Flux

Kd

Recharge

Source Size

(09-10 MFr)

Fig. 8. Correlations of simulated ½ year MFr versus the predicted long-term maximum groundwater VOC concentrations mixed through a vertical depth averageover 10 m, which can be used operationally to determine SVE progress and predict continued time until closure calculated using regression equations with fieldmeasurements of ½ year MFr. The vertical dashed line represents the most recently measured (~150 g/d), and the angled, solid, black line and equation presentregression results for the Base Case.

81K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82

Author's personal copy

References

Brusseau, M.L., Nelson, N.T., Zhang, Z., Blue, J.E., Rohrer, J., Allen, T., 2007.Source-zone characterization of a chlorinated-solvent contaminatedSuperfund site in Tucson, AZ. Journal of Contaminant Hydrology 90(1–2), 21–40.

Brusseau, M.L., DiFilippo, E.L., Marble, J.C., Oostrom, M., 2008. Mass-removaland mass-flux-reduction behavior for idealized source zones with hy-draulically poorly-accessible immiscible liquid. Chemosphere 71 (8),1511–1521.

Brusseau, M.L., Rohay, V., Truex, M.J., 2010. Analysis of soil vapor extractiondata to evaluate mass-transfer constraints and estimate source-zonemass flux. Ground Water Monitoring and Remediation 30 (3), 57–64.

Brusseau,M.L., Carroll, K.C., Allen, T., Baker, J., DiGuiseppi,W., Hatton, J.,Morrison,C., Russo, A., Berkompas, J., 2011. Impact of in situ chemical oxidationon con-taminant mass discharge: linking source-zone and plume-scale character-izations of remediation performance. Environmental Science & Technology45 (12), 5352–5358.

Carroll, K.C., Brusseau, M.L., 2009. Dissolution, cyclodextrin-enhanced solu-bilization, and mass removal of an ideal multicomponent organic liquid.Journal of Contaminant Hydrology 106 (1–2), 62–72.

Carroll, K.C., Taylor, R., Gray, E., Brusseau, M.L., 2009. The impact of composi-tion on the physical properties and evaporative mass transfer of a PCE-diesel immiscible liquid. Journal of Hazardous Materials 164 (2–3),1074–1081.

Devlin, J.F., McMaster, M., Barker, J.F., 2002. Hydrogeologic assessment of insitu natural attenuation in a controlled field experiment. Water Re-sources Research 1002 (1), 1002.

DiFilippo, E.L., Brusseau, M.L., 2008. Relationship between mass-flux reduc-tion and source-zone mass removal: analysis of field data. Journal ofContaminant Hydrology 98 (1–2), 22–35.

DiFilippo, E.L., Carroll, K.C., Brusseau, M.L., 2010. Impact of organic-liquiddistribution and flow-field heterogeneity on reductions in mass flux.Journal of Contaminant Hydrology 115 (1–4), 14–25.

DiGiulio, D.C., Varadhan, R., 2001. Analysis of water saturation, NAPL con-tent, degradation half-life, and lower boundary conditions on VOC trans-port modeling: implications for closure of soil venting systems. GroundWater Monitoring and Remediation 21 (4), 83–95.

DiGiulio, D.C., Brusseau, M.L., Ravi, V., 1998. Use of diffusion modeling to aidassessment of rate-limited vapor transport for SVE closure. Published In:Wickramanayake, G.B., Hinchee, R.E. (Eds.), Physical, Chemical, andThermal Technologies: Remediation of Chlorinated and RecalcitrantCompounds (C1-5). Battelle, Columbus, OH.

DiGiulio, D.C., Ravi, V., Brusseau, M.L., 1999. Evaluation of mass flux to andfrom ground water using a vertical flux model (VFLUX): application tothe soil vacuum extraction closure problem. Ground Water Monitoringand Remediation 19 (2), 96–104.

DOE/RL-2007-27, 2011. Feasibility Study for the Plutonium/Organic-RichProcess Condensate/Process Waste Group Operable Unit: Includes the200-PW-1, 200-PW-3, and 200-PW-6 Operable Units. U.S. Departmentof Energy, Richland Operations Office, Richland, Washington.

Einarson, M.D., Mackay, D.M., 2001. Predicting impacts of groundwater con-tamination. Environmental Science & Technology 35, 67A–73A.

Falta, R.W., 2008. Methodology for comparing source and plume remedia-tion alternatives. Ground Water 46 (2), 272–285.

Falta, R.W., Basu, N., Rao, P.S., 2005a. Assessing impacts of partial mass de-pletion in DNAPL source zones: II. Coupling source strength functionsto plume evolution. Journal of Contaminant Hydrology 79 (1–2), 45–66.

Falta, R.W., Rao, P.S., Basu, N., 2005b. Assessing the impacts of partial massdepletion in DNAPL source zones — I. Analytical modeling of sourcestrength functions and plume response. Journal of Contaminant Hydrol-ogy 78 (4), 259–280.

Freeze, R.A., McWhorter, D.B., 1997. A framework for assessing risk reduc-tion due to DNAPL mass removal from low-permeability soils. GroundWater 35 (1), 111–123.

Hoier, C.K., Sonnenborg, T.O., Jensen, K.H., Gudbjerg, J., 2009. Model analysisof mechanisms controlling pneumatic soil vapor extraction. Journal ofContaminant Hydrology 103 (3–4), 82–98.

ITRC, 2002. DNAPL Source Reduction: Facing the Challenge. Interstate Tech-nology and Regulatory Council Dense Nonaqueous Phase Liquids Team.

Khaleel, R., Jones, T.E., Knepp, A.J., Mann, F.M., Myers, D.A., Rogers, P.M.,Serne, R.J., Wood, M.I., 2001. Modeling data package for S-SX fieldinvestigation report. RPP-8296, Rev. 0, CH2MHill. Richland, Washington:Hanford Group Inc.

Klenk, I.D., Grathwohl, P., 2002. Transverse vertical dispersion in groundwa-ter and the capillary fringe. Journal of Contaminant Hydrology 58 (1–2),111–128.

Maier, U., Beyer, C., Susset, B., Grathwohl, P., 2008a. Modelling the Dilution ofSolutes due to Mass Transfer Across the Capillary Fringe, 324. IAHS Publ.,pp. 86–93.

Maier, U., Beyer, C., Susset, B., Grathwohl, P. (Eds.), 2008b. Modelling of Sol-ute Mass Transfer Across the Capillary Fringe. : Advances in SubsurfacePollution of Porous Media: Indicators, Processes and Modelling, 14. Balk-ema, Boca Raton, FL, pp. 15–30.

Mccarthy, K.A., Johnson, R.L., 1993. Transport of volatile organic-compoundsacross the capillary-fringe. Water Resources Research 29 (6), 1675–1683.

Oostrom, M., Dane, J.H., Wietsma, T.W., 2005. Removal of carbon tetrachlo-ride from a layered porous medium by means of soil vapor extractionenhanced by desiccation and water table reduction. Vadose Zone Journal4 (4), 1170–1182.

Oostrom, M., et al., 2007a. Carbon tetrachloride flow and transport in thesubsurface of the 216-Z-9 trench at the Hanford Site. Vadose Zone Jour-nal 6 (4), 971–984.

Oostrom, M., Truex, M.J., Thorne, P.D., Wietsma, T.W., 2007b. Three-dimen-sional multifluid flow and transport at the Brooklawn site near BatonRouge, LA: a case study. Soil & Sediment Contamination 16 (2), 109–141.

Oostrom, M., Truex, M.J., Tartakovsky, G.D., Wietsma, T.W., 2010. Three-di-mensional simulation of volatile organic compoundmass flux from the va-dose zone to groundwater. GroundWater Monitoring and Remediation 30(3), 45–56.

Rohay, V.J., 2007. Performance Evaluation Report for Soil Vapor ExtractionOperations at the 200-PW-1 Operable Unit Carbon Tetrachloride Site,Fiscal Year 2006, SGW-33746. Fluor Hanford, Inc., Richland, WA.

Rohay, V.J., McMahon, W.J., 1996. Airflow Modeling Report for Vapor Extrac-tion Operations at the 200-ZP-2 Operable Unit (Carbon TetrachlorideExpedited Response Action), BHI-00882, Rev. 0. Bechtel Hanford Inc.,Richland, WA.

Schwarz, R., Ptak, T., Holder, T., Teutsch, G., 1998. Groundwater risk assess-ment at contaminated sites: a new investigation approach. In: Herbert,M., Kovar, K. (Eds.), GQ'98 Groundwater Quality: Remediation and Pro-tection. IAHS Publication, pp. 68–71.

Suthersan, S., Divine, C., Quinnan, J., Nichols, E., 2010. Flux-informed remedi-ation decision making. Ground Water Monitoring and Remediation 30(1), 36–45.

Switzer, C., Slagle, T., Hunter, D., Kosson, D.S., 2004. Use of rebound testingfor evaluation of soil vapor extraction performance at the SavannahRiver Site. Ground Water Monitoring and Remediation 24 (4), 106–117.

Truex, M.J., et al., 2001. Assessment of Carbon Tetrachloride GroundwaterTransport in Support of the Hanford Carbon Tetrachloride InnovativeTechnology Demonstration Program. PNNL-13560. Pacific NorthwestNational Laboratory, Richland, WA.

Truex, M.J., Oostrom, M., Brusseau, M.L., 2009. Estimating persistent massflux of volatile contaminants from the vadose zone to ground water.Ground Water Monitoring and Remediation 29 (2), 63–72.

U.S. EPA, 1996a. User's Guide to the VOC's in Soils Presumptive Remedy.EPA/540/F-96/008.

U.S. EPA, 1996b. Engineering Forum Issue Paper: Soil Vapor Extraction Im-plementation Experiences. EPA/540/F-95/030.

U.S. EPA, 2001. Development of Recommendations andMethods to Support As-sessment of Soil Venting Performance and Closure. EPA/600/R-01/070.

U.S. EPA, 2003. The DNAPL Remediation Challenge: Is There a Case for SourceDepletion?.

US Army Corps of Engineers, 2002. Engineering and Design: Soil Vapor Ex-traction and Bioventing. EM 1110-1-4001, Appendix B.

Wellman, D.M., Riley, R.G., Rodriguez, E.A., Mitroshkov, A.V., Parker, K.E.,2007. Carbon Tetrachloride Partition Coefficients Measured by AqueousSorption to Hanford Sediments from Operable Units 200-UP-1 and 200-ZP-1. PNNL-16100, Rev. 1, Pacific Northwest National Laboratory, Rich-land, WA.

White, M.D., Oostrom, M., 2006. STOMP, Subsurface Transport Over MultiplePhases. Version 4.0. User's Guide, PNNL-15782, Pacific Northwest Na-tional Laboratory, Richland, WA.

White, M.D., Oostrom, M., Lenhard, R.J., 2004. A practical model for mobile,residual, and entrapped NAPL in water-wet porous media. GroundWater 42 (5), 734–746.

Yoon, H., Oostrom, M., Wietsma, T.W., Werth, C.J., Valocchi, A.J., 2009. Nu-merical and experimental investigation of DNAPL removal mechanismsin a layered porous medium by means of soil vapor extraction. Journalof Contaminant Hydrology 109 (1–4), 1–13.

82 K.C. Carroll et al. / Journal of Contaminant Hydrology 128 (2012) 71–82