Journal of Contaminant Hydrology 156 (2014) 38–51

Contents lists available at ScienceDirect

Journal of Contaminant Hydrology

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

A three-dimensional numerical model for linkingcommunity-wide vapour risks

Nizar Mustafa a, Kevin G. Mumford b,⁎, Jason I. Gerhard a, Denis M. O'Carroll a

a Department of Civil and Environmental Engineering, University of Western Ontario, London, Ontario N6A 5B9, Canadab Department of Civil Engineering, Queen's University, Kingston, Ontario K7L 3N6, Canada

a r t i c l e i n f o

⁎ Corresponding author. Tel.: +1 613 533 6325; faxE-mail address: kevin.mumford@civil.queensu.ca (

0169-7722/$ – see front matter © 2013 Elsevier B.V. Ahttp://dx.doi.org/10.1016/j.jconhyd.2013.10.003

a b s t r a c t

Article history:Received 21 December 2012Received in revised form 11 October 2013Accepted 16 October 2013Available online 30 October 2013

A three-dimensional (3D) numerical model that couples contaminant transport in the saturatedzone to vapour transport in the vadose zone and vapour intrusion into buildings was developed.Coupling these processes allows the simulation of vapour intrusion, arising from volatilization atthe water table, associated with temporally and spatially variable groundwater plumes. Inparticular, the model was designed to permit, for the first time, 3D simulations of risk at receptorslocated in the wider community (i.e., kilometre scale) surrounding a contaminated site. Themodelcan account for heterogeneous distributions of permeability, fraction organic carbon, sorption andbiodegradation in the vadose and saturated zones. The model formulation, based upon integrationof a number of widely accepted models, is presented along with verification and benchmarkingtests. In addition, a number of exploratory simulations of benzene and naphthalene transport in a1000 m long domain (aquifer cross-section: 500 m × 14 m) are presented, which employedconservative assumptions consistent with the development of regulatory guidance. Under theseconservative conditions, these simulations demonstrated, for example, that whether houses in thecommunity were predicted to be impacted by groundwater and indoor air concentrationsexceeding regulatory standards strongly dependedon their distance downgradient from the sourceand lateral distance from the plume centreline. In addition, this study reveals that the degreeof reduction in source concentration (i.e., remediation) required to achieve compliance withstandards is less if the risk receptor is in the wider community than at the site boundary. How-ever, these example scenarios suggest that, even considering community receptors, sources withinitially high concentrations still required substantial remediation (i.e., N99% reductions in sourceconcentration). Overall, thiswork provides insights and a new tool for considering the relationshipsbetween contaminated site source zones and community-wide risk assessment that allows fordevelopment of policies and technical approaches for contaminated site management. It isanticipated that this coupled model not only will allow significant convenience, for examplein running suites of Monte Carlo simulations for complex scenarios, but will also allow theinvestigation of vapour intrusion under conditions where soil gas concentrations may change overthe same timescale as an evolving plume.

© 2013 Elsevier B.V. All rights reserved.

Keywords:Reactive transportVapour intrusionRisk assessmentCommunity scale

1. Introduction

Contaminated groundwater associated with improperhandling or disposal of hazardous industrial liquids such as

: +1 613 533 2128.K.G. Mumford).

ll rights reserved.

nonaqueous phase liquids (NAPLs) is common, with adverseeffects on the environment and human health. Risk assess-ment is a critical tool for quantifying the potential impactsand has become widely used to support regulatory frame-works and site-specific decisions (e.g., Ma, 2002; Maxwelland Kastenberg, 1999). For sites contaminated with volatileorganic compounds (VOCs), vapour intrusion into buildingshas emerged as a critical component of risk assessments, and

39N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

thereby a controlling factor in determining mitigation andremediation activities (Abreu and Johnson, 2005, 2006;Bozkurt et al., 2009; DeVaull et al., 2002; Fischer et al.,1996; Hers et al., 2001; Johnson and Ettinger, 1991; Lahvis etal., 1999; Ostendorf and Kampbell, 1991; Roggemans et al.,2002; Wang et al., 2012; Yao et al., 2013; Yu et al., 2009).

In many jurisdictions soil and groundwater standards forVOCs have been established using risk-based criteria applied atthe boundaries of the contaminated site; for example, inOntario, Canada this distance is assumed to be 43 m. However,often the critical contaminant pathways and risk receptors arein the adjacent community at distances greater than theseassumed site boundary distances. Confining risk assessmentsto the site boundaries may be overly conservative in somecases, requiring remediation to a degree that is technicallyimpractical, and thereby impeding brownfield redevelopment.Risk-based approaches evaluated at the community levelmay reveal significantly different compliance criteria than thestandard approach at the property level, supporting achievablesite closure while still remaining protective of human health(e.g., Malina et al., 2006;Weiss et al., 2006;Wycisk et al., 2003).Conducting community-based (or ‘area-wide’) risk assessmentstudies, therefore, requires the ability to predict a number ofcontaminant pathways at the community scale, including thoserelated to groundwater and vapour intrusion.

Models to simulate vapour intrusion have been developedand widely used (e.g., Abreu and Johnson, 2005; Bozkurt et al.,2009; DeVaull et al., 2002; Johnson and Ettinger, 1991). Some ofthese models are based on analytical solutions and necessarilyemploy simplified assumptions about the scenario (e.g., homo-geneous subsurface, one-dimensional geometry), which canlimit their applicability to more complex site conditions.Numerical models (Abreu and Johnson, 2005, 2006; Abreu etal., 2009; Bozkurt et al., 2009) are able to overcome some ofthese restrictions, employing three-dimensional domains toconsider the vapour pathway from the water table to thebottom of the basement slab and then to indoor air. Thesemodels are capable of simulating a wide range of conditions(USEPA, 2012); however, studies that have used these modelshave typically focused on smaller domains (e.g., up to100 m × 100 m)with uniform, constant-concentration sourceslocated either directly below or laterally offset from the buildingfoundation. These conditions may not be applicable tocommunity-scale scenarios where groundwater plumes extendover larger distances and concentrations will be temporallyvariable near each house. For example, the time required for soilgas concentration profiles to reach steady-state may be on thescale of changes in groundwater concentrations for plumes notat steady state, particularly for fine-grained soils with highsorption (USEPA, 2012) or where biodegradation in the vadosezone is significant. Under these conditions, coupling evolvinggroundwater concentrations to soil gas concentrations andvapour intrusion would be advantageous.

Yu et al. (2009) employed a multiphase, multi-componentmodel to examine the fate of trichloroethylene (TCE) vapoursfrom a NAPL source zone in a heterogeneous aquifer. Theirsensitivity studies explored the influence of capillary fringethickness, slab fracture aperture, infiltration rate, and indoorair pressure drop within the house. The numerical modelincluded significant complexity in processes (e.g., NAPL migra-tion, heterogeneity in the subsurface, contaminant transport

across a capillary fringe, and pressure fluctuations in the indoorair). As a result, their study employed some simplifications inorder to achieve manageable simulation times, such as (i) atwo-dimensional cross-section domain, (ii) equilibrium masstransfer between phases, (iii) no sorption in the vapour phase,and (iv) no degradation in the water and the vapour phases. Inaddition, the domain size in that study was limited to theimmediate vicinity of the source zone such that the receptorwas 50 m downgradient and the aquifer was only 8 m deep.Wang et al. (2012) performed additional simulations using themodel of Yu et al. (2009), which were extended to threedimensions and investigated multiple houses in domains160 m long × 50 m wide × up to 15 m deep.

The goal of this study was to investigate the prediction ofrisk from VOC-contaminated sites on groundwater and vapourreceptors in the surrounding community and to demonstrate acoupled modelling approach to asses these receptors withinthe same simulation. The specific objectives of the study wereto (i) consider the difference in the degree of source zoneremediation required when risk assessment was conductedat the site boundary versus at receptors in the community,and (ii) evaluate the sensitivity of this difference to twofactors: the physicochemical properties of the contaminant(comparing benzene and naphthalene), and whether ground-water or indoor air was considered the risk driver. In this study,risk was assessed by comparing groundwater and indoor airconcentrations against a regulatory standard. In order topursue these objectives, a new numerical model, GW-VAP3D,was developed. This model couples three existing models tosimulate three-dimensional contaminant transport in bothgroundwater and vapour phases, including (i) sorption andbiodegradation in both phases, (ii) vapour intrusion intobuildings, and (iii) spatial heterogeneity of subsurface param-eters (e.g., hydraulic conductivity, fraction organic carbon).One novel feature of this model is that it is designed to becomputationally inexpensive even for community-scale do-mains (i.e., on the scale of 100's to 1000's of metres). Sevenexample simulations were conducted to demonstrate thisnew model using conditions similar to those used to developregulatory standards in Ontario, Canada which are designed tobe protective of human health and the environment (MOE,2011). As such, the simulation conditions used here are nec-essarily very conservative and are not intended to be repre-sentative of typical contaminated sites. However, it is expectedthat the approach will be of interest to other jurisdictionsinterested in risk-based approaches applied at a communitylevel (e.g., Malina et al., 2006; Weiss et al., 2006; Wycisk et al.,2003).

2. Model development

2.1. Conceptual framework

An example contaminated site is shown in Fig. 1, where aNAPL below the water table presents a long-term contami-nation source. Dissolution of the NAPL generates an aqueousplume of VOCs subject to volatilization to the soil gas in thevadose zone, and both the aqueous and gaseous VOC plumesare subject to advection, dispersion, sorption and biodegra-dation. In the vadose zone, pressure differences betweenthe interior of buildings and the subsurface induce flow of

Fig. 1. Conceptual model depicting the fate and transport of volatile organic compounds from a source zone located below the water table.

40 N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

soil gas into buildings through cracks in their foundations(Loureiro et al., 1990). Both elevated aqueous VOC concen-trations in the groundwater and elevated gaseous VOCconcentrations in buildings present a risk to humans: viaingestion and other exposure (e.g., bath/shower) for ground-water, and via inhalation for indoor air. Because groundwaterplumes can be extensive (Freeze and Cherry, 1979), there isthe potential for groundwater and vapour impacts at sig-nificant distances from the site where the source resides.Therefore, groundwater and vapour intrusion can result inhigher-than-acceptable risk over these same distances. Thedeveloped model, GW-VAP3D, is designed to simulate thekey processes controlling the concentration of contaminantcompounds observed at the receptor (via groundwater andindoor air) in Fig. 1 at the community scale. It is noted thatNAPL flow and dissolution, as well as the receptors (e.g.,humans), were not modelled explicitly. Instead, the simula-tions in this study focused on the reactive transport betweenthe source and receptors.

GW-VAP3D is comprised of a coupling of several publishedand well-accepted numerical models, namely MODFLOW-2005(Version 1.8) for groundwater flow (Harbaugh et al., 2005),MT3DMS (Version 5.3) (Zheng et al., 2010) for aqueous con-taminant transport, and Abreu and Johnson (2005) for vapourintrusion. MODFLOW/MT3DMS solves the three-dimensionalgroundwater flow and aqueous contaminant transport equa-tions with a finite-difference scheme including advection,anisotropic dispersion, diffusion, biodegradation, and sorption.It was modified in this work to incorporate heterogeneity ofpermeability and fraction organic carbon (foc) at the scale of thedomain's discretization, allowing input from a spatially corre-lated random field generator.

The Abreu and Johnson (2005) model couples a continuityequation governing the vapour-phase pressure distributionand the resulting soil gas velocity field,with a gaseous chemicaltransport equation that accounts for diffusion, advection, andbiodegradation. It assumes that the vapour receptors aresingle-family dwellings that exhibit a crack in the foundation.Advective soil gas flow is induced by a reduced pressure inthe building relative to the atmosphere due to temperature

differences, wind conditions, or venting from the building(USEPA, 2012). The solution of these equations generatesthree-dimensional pressure, soil vapour velocity, and chemicalconcentration fields that are then processed to predict thesoil vapour flow rate into the building, the chemical transferrate into the building, and thus the resulting indoor air con-centration. In GW-VAP3D, due to the linkage with MODFLOW/MT3DMS, the soil vapour concentrations along the water tableare dictated locally by the adjacent groundwater VOC concen-trations. Thus, simulations conducted using GW-VAP3D differfrom the study by Abreu and Johnson (2005) in that the vapoursource boundary conditions are spatially and temporallyvariable. It further differs in that the vapour concentrationsare directly linked to the source zone via the evolvinggroundwater plume. Furthermore, GW-VAP3D permits het-erogeneity of soil properties both above and below the watertable. Compared to the simulations of Yu et al. (2009) andWang et al. (2012) the presented model is less complex insome respects (e.g., no NAPL migration, no capillary fringe,constant indoor air pressure) andmore complex in others (e.g.,sorption and biodegradation considered in both vapour andwater phases). The guiding principle in this model's designwas to include the majority of critical processes for simulatingconditions where a source zone is depleting over time, gen-erating evolving groundwater and vapour plumes (i.e., con-centrations increasing, then decreasing) over hundreds ofmeters in a realistic subsurface environment.

2.2. Governing equations

2.2.1. Groundwater flow and aqueous contaminant transportThe three-dimensional flow of groundwater of constant

density through an unconfined aquifer is described by (Rushtonand Redshaw, 1979):

Ss∂h∂t ¼ ∇ � K∇hð Þ þW ð1Þ

where K is the hydraulic conductivity tensor (LT−1), h isthe hydraulic head (L), W is a source/sink term (T−1), Ss is

41N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

the specific storage (L−1), and t is time (T). The generalthree-dimensional, contaminant transport equation for a tran-sient groundwater flow system solved by MT3DMS, assumingequilibrium linear sorption to organic carbon in aquifer solids, is(Zheng et al., 2010):

1þ ρbKoc f ocθw

� � ∂Cw

∂t ¼ −∇ � qwθw

Cw

� �þ∇ � D∇Cwð Þ

þ qssCss−Rs ð2Þ

where Cw is the concentration of chemical in the aqueousphase (ML−3), ρb is the bulk density of the soil matrix(ML−3), Koc is the sorption coefficient of chemical to organiccarbon (MMoc

−1·L3M−1), foc is the mass fraction of organiccarbon in the soil (MocMsoil

−1), qw is the groundwater specificdischarge vector (LT−1), θw is thewater-filled porosity (L3L−3),D is the dispersion tensor (L2T−1), qss is the volumetric flowrate per unit volume of aquifer representing fluid sources/sinks(T−1), Css is the concentration of the source/sink (ML−3);and Rs is the chemical reaction term for the saturated zone(ML−3T−1). The specific discharge is given by Darcy's law(qw = −K∇h) based on the solution to Eq. (1) and thedispersion tensor is calculated in MT3DMS considering molec-ular diffusivity and dispersivity, including cross-terms (Bear,1979). This formulation assumes that changes in the concen-tration field will not affect the flow field significantly, which isreasonable for the scenarios under consideration. The reactionterm, Rs, is typically employed to simulate biodegradation inthe saturated zone and can be implemented with a range ofexpressions including first-order or Monod kinetics.

2.2.2. Vapour flow and gaseous contaminant transportThe governing equations for flow of soil gas leading to

vapour intrusion have been presented in previous studies(e.g., Abreu and Johnson, 2005; Yao et al., 2013; Yu et al.,2009). GW-VAP3D uses those presented by Abreu andJohnson (2005), which are summarized here. Derived fromthe soil gas continuity equation, the three-dimensionaldisturbance pressure field equation assuming heterogeneoussoil properties is (Massmann et al., 1991):

∂P∂t ¼ Patm

θgμg∇ � kg∇P

� �ð3Þ

where P is the absolute pressure (ML−1T−2), Patm (ML−1T−2)is the mean pressure, approximated here by the atmosphericpressure, kg is the soil gas permeability tensor (L2), θg is thegas-filled porosity (L3L−3), and μg is the soil gas dynamicviscosity (ML−1T−1). Note that this equation neglectsdensity-driven flow effects, which are not expected to besignificant in these scenarios (Abreu and Johnson, 2005). Thesoil gas specific discharge field can be obtained from thesolution to Eq. (3) and Darcy's law applied to the soil gas:

qg ¼ 1μg

∇ � kgP� �

ð4Þ

where qg is the soil gas specific discharge vector (LT−1). Thevapour concentration is given by the chemical transport

equation (Bear, 1979), assuming water velocity in the vadosezone is negligible:

θg þθwH

þ ρbKoc f ocH

� �∂Cg

∂t ¼ ∇ � Dgw∇Cg

� �−∇ � qgCg

� �−Ru

ð5Þ

where H is the Henry's law constant (ML−3·M−1L3), Cg isthe concentration of chemical in the gas phase (ML−3) assumedto be in equilibrium with the aqueous concentration accordingto Henry's law (Cg = HCw) (Abreu and Johnson, 2005; Bedientet al., 1994; Bozkurt et al., 2009; Davis et al., 2009), Dgw is theoverall diffusion coefficient (L2T−1), and Ru is the chemicalreaction term for the vadose zone (ML−3T−1). The overalldiffusion coefficient, based on effective porous mediadiffusion (Millington and Quirk, 1961), is given by:

Dgw ¼ Dg θg10=3

θt2 þ Dw

Hθw

10=3

θt2 ð6Þ

where Dg is the molecular diffusion coefficient of the chemicalin air (L2T−1), Dw is the molecular diffusion coefficient of thechemical in water (L2T−1), and θt is the total porosity (L3L−3).Biodegradation is assumed to occur in the soil moisture and, asin the saturated zone, the reaction term can be of several forms.In the example simulations presented here, it is described by afirst-order rate expression:

Ru ¼ θwλuCw ð7Þ

where λu is the first-order reaction rate constant (T−1). It iscommonly accepted that a first-order biodegradation reaction isappropriate for petroleum hydrocarbons as long as the oxygenconcentration in the vadose zone exceeds aminimum threshold(e.g., DeVaull, 2007). Implementing this in the model requiresthe simulation of oxygen as a separate species (e.g., Abreuand Johnson, 2006). While the model supports this option,implementing it has significant computational cost implicationsfor community-wide simulations. Since the oxygen thresholdhad a minor effect on indoor vapour concentrations in a ma-jority of simulated cases (Abreu and Johnson, 2006) it was notincluded in the example simulations presented here.

2.2.3. Indoor air concentrationThe indoor air concentration is determined from a steady-

state mass balance on the enclosed space assuming rapidmixing of the indoor air, no indoor emission sources, and noindoor reactions (Abreu and Johnson, 2005), and assumingthe chemical enters the building only from the soil gas andnot from the air outside the building, given by:

Cia ¼1

VbAex þ Qs∫Lck

Qck

expQck

WckDckdck

� �Cg

ck−Cia

expQck

WckDckdck

� �−1

dLck ð8Þ

where Cia is the indoor air concentration (ML−3), Vb is thevolume of the enclosed space within the building (L−3), Aex isthe air exchange rate of the enclosed space (T−1), Qck is thesoil gas flow rate per unit length of crack (L3T−1L−1), Lck is

42 N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

the total crack length (L), Qs is the soil gas flow to theenclosed space based on advection and diffusion throughcracks in the foundation calculated by integrating Qck over Lck(Abreu and Johnson, 2005) (L3T−1), Wck is the crack width(L), dck is the foundation thickness (L), Dck is the effectivediffusion coefficient for transport through the crack, which isassumed to be equal to Dgw (L2T−1), and Cg

ck is the concen-tration of chemical in the gas phase at the soil-foundationcrack interface (ML−3). Integral calculations were based onthe trapezoid rule for numerical integration, and the indoorair concentrations from the previous time step were used tosolve the implicit definition given by Eq. (8).

2.3. Equations solution and model coupling

MODFLOW uses a finite difference formulation and amixed Eulerian–Lagrangian approach to solve for groundwa-ter flow (Eq. (1)), and MT3DMS, coupled via a split-operatorapproach (Barry et al., 2000), solves for contaminant transportin the saturated zone (Eq. (2)) (Zheng et al., 2010). In thiswork, the vapour intrusion equations (Eqs. (3)–(8)) werecoded in FORTRAN,whichwere also coupled toMODFLOWandMT3DMS using a split-operator approach. As a result, thegroundwater transport step calculates the heterogeneousdistribution of aqueous contaminant concentrations for thenodes corresponding to the water table, each of which thenserves as a vapour transport boundary condition. Groundwaterconcentrations are updated in these nodes using computedmass loss to the vapour phase prior to the next time step. Thiswork neglects the presence of a capillary fringe, assuminga sharp interface between the water table and a constantmoisture content in the vadose zone. This was to reducecomputational expense and is considered conservative in thatvapour concentrations, the dominant risk driver inmany cases,are expected to be reduced with a significant capillary fringe(Yu et al., 2009) or with high-moisture content layers betweenthe vapour source and the building foundations (USEPA, 2012).

MODFLOW normally permits heterogeneity to be definedonly as stratigraphic layers. For this work, the code wasmodified to allow heterogeneity discretized at the nodal scale.This permitted input of a spatially correlated hydraulic con-ductivity field generated using the random field generator FGEN(Robin et al., 1993). FGEN generates autocorrelated and cross-correlated three-dimensional random fields using the directpower spectral estimation method (i.e., by applying an inverseFourier transform to randomized spectral fields obtained fromthe three-dimensional power spectral density functions and thecross-spectral density functions of the field parameters) (Robinet al., 1993). An exponential autocorrelation function was usedin this work. In practice, this is an accepted method of creatingnumerical domains that replicate the natural heterogeneity at avariety of scales associated with unconsolidated, near surfaceaquifers (Cvetkovic and Shapiro, 1990; Cvetkovic et al., 1998;Dagan, 1994; Gelhar and Axness, 1983; Hess et al., 1992;LeBlanc et al., 1991; Neuman and Tartakovsky, 2009; Sudicky,1986; Woodbury and Sudicky, 1991).

2.4. Model verification

MODFLOW/MT3DMS has been extensively tested and iswidely accepted (e.g., Barry et al., 2002; Prommer et al.,

2003). To verify modifications made as part of GW-VAP3D(Eqs. (3)–(8)), preliminary simulations were comparedagainst solutions using the model of Abreu and Johnson(2005). These test cases assumed a constant contaminantconcentration at the water table, with and without first-orderbiodegradation in the vadose zone water-filled porosity,using input parameters from Abreu and Johnson (2005).Fig. 2 presents the match of the two solutions. In addition,verification was performed for the model coupling via massbalance calculations on all phases. For example, Scenario A(discussed further in Section 3) over a 20-year simulationperiod showed an error of 0.83% between mass entering (viagroundwater) and exiting (via soil gas) a control volumedefined as a 1 m × 20 m × 20 m region immediately abovethe water table under a centrally located house.

3. Model simulations

3.1. Simulation conditions

The community scenario employed considered multiplehouses located within 1000 m downgradient and within250 m laterally of the centreline of a site contaminated with aNAPL source. Many of the simulation conditions, including thedimensions for the contaminated site and source area, werechosen to be consistent with those used in the screeningmodelemployed by the Ontario Ministry of the Environment forestablishing their cleanup criteria (MOE, 2011) (Table 1). Themodel domain (Fig. 3) is 250 m wide × 1000 m long in planview (see shaded area); by employing a no-flow boundary anda ‘half domain’ that cuts longitudinally through the source zoneand parallel to the dominant flow direction it is possible to takeadvantage of symmetry along the central vertical transect ofthe plume and thereby represent a scenario 500 m wide whilelimiting computational expense. Thus, throughout this work,the term “centreline” refers to the line of symmetry that bisectsthe source zone and groundwater plume (i.e., no-flow domainboundary on left side when looking downgradient). Thetotal depth is 14 m, comprising a 4 m deep vadose zoneunderlain by a 10m thick aquifer and an aquitard upper surfacecoincident with the bottom boundary. The surface of thedomain is parallel to the water table. A mean hydraulicgradient of 0.003 was applied from left to right via constanthead boundary conditions. No flow boundary conditions wereassigned at all other boundaries except the downgradientboundary that allowed the free exit of both groundwater andvapour. The model discretization was Δx = 1 m, Δy = 1 m,and Δz = 0.5 m, which corresponds to a total of 7.0 × 106

total nodes; this discretization was confirmed to provideexcellent mass balance and minimal numerical dispersionvia comparison to analytical solutions (Aziz et al., 2000) ofthree-dimensional advective–dispersive groundwater flowand transport.

The source zone was a 13 m × 13 m × 2 m deep regionpositioned adjacent to the left-hand boundary, the top ofwhich was coincident with the water table. In this sourcearea a constant concentration of dissolved contaminant wasspecified for a sufficient period (the ‘active source’ period) toensure that steady-state concentration of the target VOC wasobserved at the downgradient boundary. An ‘active source’period of 200 years was used for a benzene source, and an

Fig. 2. Simulation results of soil vapour intrusion with and without biodegradation in the unsaturated zone. Source concentration versus indoor air concentrationis plotted using both the Abreu and Johnson (2005) model and the GW-VAP3D model developed here.

43N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

‘active source’ period of 400 years was used for a naphtha-lene source. Following this ‘active source’ period, the sourceconcentration was reduced instantly to zero, and a period ofequal length was simulated in order to consider the influenceof source removal by remediation or natural attenuation.Run times for simulations using the 1000 m × 250 m × 14 mdomain discretized to 7 × 106 nodes over a period of 400 yearswere approximately 3 days using the SHARCNET (SharedHierarchical Academic Research Computing Network) groupof high-performance computing clusters.

The aquifer properties employed in this work are presentedin Table 1. The random, spatially correlated hydraulic conduc-tivity field (Robin et al., 1993) employed aquifer correlationlengths of 17.2 m × 7.4 m in plan view and 1.0 m in height(Sudicky et al., 2010). The hydraulic conductivity propertieswere based on the North Bay aquifer (Sudicky et al.,2010), with a geometric mean hydraulic conductivity of3.5 × 10−3 cm/s and variance of ln(K) = 1.79. This representsa relatively heterogeneous aquifer with soil ranging fromgravel to silt and a mean grain size of medium sand. A singlepermeability field was generated and applied over the entiredomain, including the saturated zone and the vadose zone. Thesingle realization of this hydraulic conductivity field employedfor all simulations in this work is presented in Fig. 4. For solutetransport, local dispersivity values of 10 m, 1 m, and 0.1 m inthe longitudinal, transverse horizontal, and transverse verticaldirections, respectively, were used (Table 1). These values arelarger than those typically applied in heterogeneous simula-tions, but were chosen here as part of the conservativeapproach adopted by theMOE (2011) tomaximize the numberof houses in the domain that might be impacted by the plume.Use of these dispersivity values resulted in a plume that spreadapproximately 30 m further from the centreline compared tosimulations that used local dispersivity values equal to zerounder the conditions of Scenario A (described below). MonteCarlo simulations that explore ensemble results for a single setof aquifer statistics, aswell as sensitivity of model results to the

mean and variance of the field, will be presented in a follow-upwork.

The receptors of interest are two lines of six houses each:one on the centreline and the other offset laterally by 125 m(Fig. 3). Six of the houses, which are discussed further inSection 4, have been labelled H1, H4, H6, H7, H10, and H12.H1 is located 43 m from the upgradient boundary (i.e., 30 mfrom the downgradient edge of the source), coincident withthe contaminated site boundary. H4 and H6 are located at500 m and 1000 m into the community along the centrelinewhile H7, H10, and H12 are located at these same threedistances from the source but laterally offset 125 m from thecentreline. Each house is 10 m × 10 m in plan view and ischaracterized by a crack at the junction between the floorand the walls along the perimeter of the concrete basement(Abreu and Johnson, 2005). Each house is assigned thephysical characteristics presented in Table 1, which alsodetails the vapour transport parameters employed in allsimulations considered in this study. A zero-flow boundarycondition for the air phase was imposed at all externaldomain boundaries and at the envelope of the buildingfoundation except at the cracks. The top of the domain (soilsurface) was assigned atmospheric pressure and zero con-taminant concentration, while a slightly negative pressure(5 Pa below atmospheric) was assigned inside each house(Table 1). Results confirmed that the domain was chosen tobe large enough so that there were negligible soil gasvelocities and no VOC-contaminated vapour in the vicinityof the side boundaries.

Benzene and naphthalene were considered as representa-tive compounds in this study to investigate a range ofpartitioning behaviour characterized by different values ofaqueous solubility, Henry's coefficient, and organic carbonsorption coefficient (Table 1). Benzene has been used previ-ously, in part, because its physicochemical properties aresimilar to many compounds of interest for vapour intrusion(USEPA, 2012). Naphthalene was selected because it is

Table 1Model parameters used in the numerical simulations.

Parameter Value

Flow and transport parametersSaturated zone thickness 10 mVadose zone thickness 4 mPorosity1 0.3Dry bulk density1 1.81 g/cm3

Fraction of organic carbon1 0.0003Mean horizontal hydraulicgradient1

0.003

Recharge rate 0 mm/yWater-filled porosity in vadosezone2

0.054

Longitudinal dispersivity 10 mTransverse horizontal dispersivity 1 mTransverse vertical dispersivity 0.1 mSource zone thickness1 2 mDynamic viscosity of air2 0.0648 kg/(m·h)

Basement parameters2

Basement depth 2 mFoundation thickness 0.15 mEnclosed space volume 244 m3

Indoor air mixing height 2.44 mAir exchange rate 0.5 h−1

Perimeter crack width 0.001 mTotal crack length 39 mAtmospheric concentration 0 μg/m3

Building pressure(disturbance pressure)

5 Pa below atmospheric

Chemical parameters Benzene Naphthalene

Aqueous solubility at 15 °C1 1800 mg/L 32 mg/LDiffusivity in air3 8.8 × 10−2 cm2/s 5.9 × 10−2 cm2/sDiffusivity in water3 9.8 × 10−6 cm2/s 7.5 × 10−6 cm2/sHenry's Coefficient at 15 °C4 0.145 0.01Log Koc

5 1.56 2.97Groundwater standard1 5 μg/L 11 μg/LIndoor air standard6 3.1 μg/m3 3 μg/m3

1 MOE (2011).2 Abreu and Johnson (2005).3 USEPA (2003).4 USEPA (2001).5 USEPA (1996).6 USEPA (1998).

44 N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

significantly less volatile and more highly attenuated bynatural organic matter than benzene (Table 1). All housesshown in Fig. 3 were considered as receptors of vapourintrusion and groundwater. Groundwater concentrations rele-vant to each house were computed as the maximumgroundwater concentration observed as a function of timeover a 10 m wide × 10 m deep control plane orientedperpendicular to the groundwater flow direction and locateddirectly under the centre of each house. The simulatedconcentrations were compared to the Ontario MOE standardfor potable groundwater and the United Sates EnvironmentalProtection Agency (USEPA) standard for indoor air (Table 1).

By choosing simulation conditions that were consistentwith those employed for the development of regulatorystandards, the results of this study are expected to be con-servative (i.e., lead to groundwater and indoor air concentra-tions that are much higher than those at typical contaminatedsites). Simulation conditions expected to lead to conservativeresults include a shallow water table, no recharge, no bio-degradation in the saturated zone, and high aqueous

concentrations in the source zone. The effect of water tabledepth (USEPA, 2012) and recharge (Barber et al., 1990;Nazaroff et al., 1985; Wang et al., 2012; Yu et al., 2009) onindoor air concentrations have been presented elsewhere.While these simulations are expected to be relatively conser-vative, some attenuation mechanisms (sorption, biodegrada-tion) were included to increase the relevance to typical fieldscenarios. First-order biological degradation was assumed tooccur in themoisture retained in the vadose zone. Although themodel is capable of accounting for heterogeneous distributionsof organic carbon and first-order decay coefficients, both wereassumed to be spatially invariant in these simulations; futurework will examine the influence of heterogeneity of each ofthese on vapour intrusion at the community scale.

3.2. Scenarios

A total of seven simulation scenarios (A–G) were con-sidered to demonstrate the model and investigate risks togroundwater and indoor air for different source types andconcentrations. In Scenario A the VOC of concern wasbenzene and its concentration in groundwater at the sourcewas specified to be 900 mg/L. This concentration representshalf the aqueous solubility of benzene, which is the maxi-mum aqueous concentration that cannot be exceeded inOntario even if site-specific information is used in risk-basedcalculations (MOE, 2011), and is representative of a NAPLsource. Similarly high concentrations have been used previ-ously to investigate vapour intrusion (Abreu and Johnson,2006). The first-order biodegradation rate in the vadose zonewas specified to be 0.83 d1. This value is on the order of thelowest value reported by DeVaull (2007), and is in the rangeof degradation rate constants investigated by others studyingvapour intrusion (Abreu and Johnson, 2006; Abreu et al.,2009). An investigation of different vadose zone biodegrada-tion rates (results not shown) confirmed the findings of theseprevious studies, and showed that indoor air concentrationswere sensitive to changes in degradation rates and decreasedwith increasing degradation rates. This value of degradationrate constant is considered a conservative value given thatpetroleum hydrocarbon degrading aerobic organisms arerelatively ubiquitous in soils, and because significant atten-uation of hydrocarbon vapours via aerobic biodegradationhas been observed in natural settings away from buildings(Laubacher et al., 1997; Ostendorf and Kampbell, 1991;Roggemans et al., 2002) and in experimental soil microcosmssettings (Baker et al., 1997; DeVaull et al., 2002; Hohener etal., 2003; Ostendorf et al., 2000). The ‘active source’ periodwas 200 years, to ensure steady-state conditions at thedowngradient boundary (i.e. 1000 m downgradient), follow-ed by a second 200 year period for which the source wasremoved. While these times may appear long, they arerelevant to the natural, rate-limited dissolution of NAPLsource zones and groundwater plume transport times at thecommunity scale; moreover, it is noted that MOE risk assess-ment guidelines require considering 300 years for calculatingsource remediation based on off-site risk (MOE, 2011).

In Scenario B the VOC was changed to naphthalene at asource concentration of 16 mg/L (half its aqueous solubility),representative of a heavy hydrocarbon or coal tar contaminat-ed site. This simulation required an ‘active source period’ of

500

m

1000 m

Source zone = (13 m x 13 m x 2 m)

Community Boundary

House 10 m x 10 m

43 m

125

mNo flow boundary

No flow boundary

Hea

d =

13

m

C=Co=900 mg/L

Hea

d =

10

m

Groundwater Flow

125

m

125 m

500 m 250 m 250 m

125

m

250 m

H1

H10H7 H12

H4 H6

Fig. 3. Community scenario of 1000 m length and 500 m width. Using a symmetry, no-flow boundary along the centreline, the grey-shaded 250 m × 1000 mmodel domain is employed. The source zone is represented as a red square, characterized by a constant concentration period (i.e., the ‘active source’ period)followed by a zero concentration period; all other boundary conditions are marked. Two lines of six houses each are included in the model domain; one on theplume centreline and the other 125 m off the centreline. Each house is assumed to have a basement with slightly negative pressure and is a receptor for bothgroundwater and indoor air.

45N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

400 years to reach steady-state andwas followed by a 400 yearperiod for which the source was removed. All other conditionswere the same as used in Scenario A. In Scenarios C–G thesource concentration of benzenewas changed to 0.009, 0.0158,0.63, 0.95, or 100 mg/L, respectively. These can be consideredthe result from differing spill conditions, source zone architec-tures, ageing characteristics, or the effects of differing types ordegrees of partial mass removal by engineered or intrinsicremediation. For example, higher source concentrations arerepresentative of pure benzene spills with little mass trans-fer limitation during source dissolution, whereas lower

ln(K) (K in m/s)

Fig. 4. The random, spatially correlated, heterogeneous permeability field realizationallow the internal structure to be viewed. Statistical aquifer properties are provided

concentrations are representative of sorbed benzene or ben-zene in NAPL mixtures (e.g., petroleum fuels) with greatermass transfer limitation during source dissolution. For thegroundwater flow conditions used here, the highest sourceconcentration considered (Scenario A – source concentration of900 mg/L for 200 years) requires a release of approximately 85drums (55 gal each) of pure benzene, which would distributewithin the 13 m × 13 m × 2m source at an average saturationof 17%. The lowest source concentration considered (Scenario C— 0.009 mg/L for 200 years) requires 177 mL of benzene. Allother conditions were the same as used in Scenario A.

that was employed for all simulations. The front top quadrant is removed toin the text.

46 N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

4. Results and discussion

4.1. Scenario A: high concentration benzene source

The groundwater plume for Scenario A at steady-state isshown in plan view at the water table (Fig. 5), and incross-section along (i) the plume centreline (Fig. 6a) and(ii) the transect 125 m from the centreline (Fig. 6b). Fig. 5reveals a lateral plume spread of approximately 150 m at thedowngradient boundary, using the groundwater standard of0.005 mg/L. Fig. 6 reveals, as expected, that the magnitudeand trend in concentration with distance downgradient isa function of lateral distance from the source. For example,on the centreline groundwater concentrations decreaseddowngradient from the source, while at 125 m off-centrethe groundwater concentrations increased up to the 1000 mextent of the domain, and would decrease at greater dis-tances if the domain were larger. Vapour concentrations areobserved to reflect the upwards flow pathways associatedwith vertical pressure gradients leading to indoor air intru-sion into the basements. They also mirror the groundwaterconcentrations, as expected; for example, Fig. 6a reveals thatboth groundwater and soil vapour concentrations decreaseby approximately 5 orders of magnitude with distance alongthe centreline from the source zone and Fig. 6b revealsthat both increase 3.5 orders of magnitude along the off-settransect. The vapour concentrations at greater distances down-gradient of the source are due to the transport of benzene in thegroundwater (i.e., primarily horizontal migration) followed bypartitioning and transport, primarily vertical, in the soil gas.Vapour concentrations are not due to transport through the soilgas alone between the source and the houses (i.e., horizontalgas phase migration is limited).

Groundwater and indoor air breakthrough curves forbenzene at houses H1 and H7 (43 m), H4 and H10 (500 m)and H6 and H12 (1000 m) are plotted in Fig. 7. Results indicatethat the first breakthrough of benzene along the centreline (i.e.,exceeding 0.005 mg/L), occurred after 3 years, 32 years and64 years at 43 m, 500 m and 1000 m, respectively. Steady-stateconcentrations were achieved after 13 years, 70 years and140 years at the three distances, respectively (Fig. 7a and b). A

300 100 50

Fig. 5. Plan view of the distribution of benzene groundwater concentrations at the wforms a line of symmetry through the source zone and plume centreline. The dottedH12. The outside edge of the final concentration contour represents the drinking w

similar trend was observed for the indoor air concentrations.Steady state groundwater and indoor air concentrationsexceeded their respective standards at all house locationsexcept H7, located 43 m downgradient of the source and125 m off of the plume centreline. In agreement with Fig. 6,these data confirm thatwhile the steady-state groundwater andindoor air concentrations decreased between 500 m and1000 m from the source along the centreline, they increasedalong the 125 m off-centre transect due to the increasing effectof transverse dispersion with distance downgradient. Thispattern is observed in the results, to varying degrees, for all ofthe benzene simulations conducted in this work. While this isnot an unexpected result, it emphasizes the need to identifyreceptor locations relative to groundwater flow directions,particularly at the community scale. Thus, houses in a com-munity that are located along a groundwater flow path down-gradient from a contaminated site must be consideredimportant candidates for exceeding risk thresholds but thelikelihood of exceedance decreases with distance from thesource. However, receptors that are laterally offset from thisadvective groundwater pathmust also be considered candidatesfor exceeding risk thresholds if they are in a key range ofdistances downgradient: not so close that plume dispersion isnot significant and not so far that dilution/sorption/degradationhave depleted groundwater concentrations. For example, Wanget al. (2012) identified that regulatory standards for indoor airmay not be exceeded in houses offset from the plume centrelineby only 10 m, but their study focused on houses within 100 mdowngradient of the source. The key range of distances iden-tified in this study, for example between 500 and 1000 mdowngradient at a lateral distance of 125 m in Scenario A, arespecific to these simulation conditions.

4.2. Scenario B: high concentration naphthalene source

Naphthalene groundwater and indoor air breakthroughcurves at the houses located 43 m, 500 m and 1000 m fromthe source (Fig. 8) show longer arrival times than observedfor benzene, consistent with naphthalene's higher Koc value(Table 1). Both groundwater and indoor air show a decreasein the steady-state concentration with increased distance

10 5 1 0.1 0.01 0.005 0

ater table for Scenario A. The top of the figure is the no-flow boundary whichline is at 125 m off the centreline and corresponds with houses H7, H10 andater standard of 0.005 mg/L.

GroundwaterConcentration (mg/L)

VapourConcentration (µg/m3)

500

105

250

104

100

103

10

102

1

101 100

0.1 0.001

10-3

A

B

0

0

Fig. 6. Cross-section, longitudinal transects showing benzene concentration in the groundwater (blue) and vapour (red) phases along (a) the plume centrelineand (b) 125 m laterally off-set from the centreline. Vapour concentrations downgradient of the source are dominated by contaminant transport in groundwaterfollowed by the partitioning of benzene between the groundwater and soil gas in the vicinity of the houses (i.e., horizontal gas phase migration is limited).

0.E+00

1.E+05

2.E+05

3.E+05

4.E+05

0

100

200

300

400

500

0

A

Ind

oo

r ai

r co

nce

ntr

atio

n (

µg

/m3 )

Gro

un

dw

ater

co

nce

ntr

atio

n (

mg

/L)

Time (year)

0

2

4

6

8

0

0.002

0.004

0.006

0.008

0.01

0

B

Ind

oo

r ai

r co

nce

ntr

atio

n (

µg

/m3 )

Gro

un

dw

ater

co

nce

ntr

atio

n (

mg

/L)

Time (year)

Groundwater (43 m) Groundwater (500 m) Groundwater (1000 m)Indoor air (43 m) Indoor air (500 m) Indoor air (1000 m)Groundwater standard Indoor air standard

100 200 300 400 100 200 300 400

Fig. 7. Benzene (Scenario A) groundwater and indoor air breakthrough curves for the houses identified in Fig. 3 along (a) the centreline (H1, H4, and H6) and(b) 125 m off-set from the plume centreline (H10 and H12). Results are not shown for H7, located 125 m off-set from the plume centreline and 43 mdowngradient, as groundwater and indoor air concentrations were negligible at that location.

47N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

0

0.01

0.02

0.03

0.04

0.E+00

5.E-05

1.E-04

2.E-04

2.E-04

3.E-04

Ind

oo

r ai

r co

nce

ntr

atio

n (

µg

/m3 )

Gro

un

dw

ater

co

nce

ntr

atio

n (

mg

/L)

Time (year)

B

0

200

400

600

800

1000

1200

0

1

2

3

4

5

6

0

A

Ind

oo

r ai

r co

nce

ntr

atio

n (

µg

/m3 )

Gro

un

dw

ater

co

nce

ntr

atio

n (

mg

/L)

Time (year)

Groundwater (43 m) Groundwater (500 m) Groundwater (1000 m)

Indoor air (43 m) Indoor air (500 m) Indoor air (1000 m)

800600400200 0 800600400200

Fig. 8. Naphthalene (Scenario B) groundwater and indoor air breakthrough curves for the houses identified in Fig. 3 along (a) the centreline (H1, H4, and H6) and(b) 125 m off-set from the plume centreline (H7, H10 and H12).

48 N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

downgradient of the source along the centreline (Fig. 8aand b). At 125 m off-set from the centreline (Fig. 8c and d),both groundwater and indoor air concentrations increasedthen decreased with increasing distance downgradient ofthe source. The maximum groundwater concentration washigher than the naphthalene drinking water standard of0.011 mg/L (MOE, 2011) at all houses along the centreline,while it was lower than the standard at all houses 125 m offthe centreline. The naphthalene plume at steady-state didnot impact as much of the aquifer laterally relative to thebenzene plume due to the lower source zone concentration(i.e., lower solubility) and the higher threshold groundwaterconcentration (i.e., higher drinking water standard) used todefine the extent of the plume. This example illustrates thatcompounds characterized by a lower ratio of their aqueoussolubility to their regulatory standard (e.g., 2.8 × 103 fornaphthalene in this study) are less likely to pose a risk toreceptors that are laterally offset from the plume centrelinethan those compounds with higher ratios of aqueous solu-bility to regulatory standard (e.g., 3.6 × 105 for benzene inthis study).

A close inspection of the data in Figs. 7 and 8 shows thatfollowing complete source removal (i.e., after the simulatedsource concentrations were set to zero at the end of the ‘activesource’ period) the order in which compliance with thegroundwater and indoor air standards were achieved differedbetween the two representative compounds. For example, at adistance of 500 m downgradient of the source along the plumecentreline (i.e., at house H4), the naphthalene groundwaterconcentration decreased below the standard 173 years aftersource clean up, and the indoor air concentration decreasedbelow the standard 4 years later. For benzene, the groundwa-ter concentration decreased below the standard 25 years aftersource clean up, and the indoor air concentration decreasedbelow the standard 2 years earlier. Thus, for naphthalene,groundwater concentrationswere in compliancewith drinkingwater standards before indoor air concentrations were incompliance with indoor air standards and the opposite trendwas observed for benzene. This is due to the volatility andsorption properties of both contaminants: benzene is highlyvolatilewith low sorption,while naphthalene has low volatility

with high sorption. In the cases shown, the differences in thetimes required to reach compliance in groundwater and indoorair are quite similar (i.e., within 4 years for naphthalene andwithin 2 years for benzene). However, if a different standardwere used (e.g., by using indoor air standards that were lowerthan those in Table 1 by one order of magnitude) the dif-ferences in the times required to reach compliance in ground-water and indoor air are more substantial (27 years and39 years for benzene and naphthalene, respectively). Underconditions where more substantial discrepancies exist, it isimportant to consider both the groundwater and indoorair concentrations to determine when compliance could beachieved after source mass reduction.

4.3. Scenarios C–G: lower concentration benzene source

Fig. 9 presents the steady-statemaximum benzene ground-water and indoor air concentrations for the houses at 43 m and1000 m from the source for a range of source strengthsfrom 0.009 mg/L to 100 mg/L (Scenario C–G) and 900 mg/L(Scenario A). These results show that to achieve a groundwaterconcentration less than the standard of 0.005 mg/L at 43 m(site boundary) a source zone concentration of less than0.0158 mg/L (i.e., 0.0018% of the aqueous solubility) wasrequired. However, to achieve an indoor air concentrationless than the standard of 3.1 μg/m3 in a house at 43 m requiresa further reduction to 0.0090 mg/L (i.e., 0.001% of the aqueoussolubility) (Fig. 9a). For high-concentrations sources, such asthose created by the dissolution of single-component NAPLsource zones, these results represent the extent of remediationrequired to meet MOE generic standards at the propertyboundary. As expected, less reduction in source zone concen-trations are required to meet standards at receptors locatedfurther into the surrounding community. Simulation resultsshow that to achieve a groundwater concentration less thanthe standard at the house located 1000 mdowngradient on thecentreline a source zone concentration of less than 0.95 mg/L(i.e., 0.053% of the aqueous solubility) was required, while toachieve an indoor air concentration less than the standardrequired a further reduction to 0.63 mg/L (i.e., 0.035% of theaqueous solubility) (Fig. 9b). These values depend strongly on

Fig. 9. Steady-state groundwater and indoor air concentration (log scale) for different source strengths of benzene (Scenarios A and C–G) at (a) house H1 (43 mdistance, on the centreline), and (b) house H6 (1000 m distance, on the centreline).

49N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

the choice of parameter values that control the lateralspreading of the plume (i.e., dispersivity and distribution ofhydraulic conductivity), with less spreading resulting insmaller differences between source removal goals to protectreceptors at the property boundary compared to those furtherinto the community. These values also depend on the con-servative assumptions adopted in the regulatory context ofthis work, with a shallow water table, no recharge, and nobiodegradation in the saturated zone leading to higher vapourconcentrations and, therefore, greater removal required.

The need for more substantial remediation to satisfyrequirements for the protection of indoor air receptors ishighly dependent on the specific regulatory standards beingapplied. However, the results shown in Fig. 9 highlight twoimportant considerations. First, remediation requirementsmaydiffer depending on the pathway/receptor being considered;i.e., cleaning up to groundwater standards will not necessarilysatisfy indoor air standards. Thismay be particularly true whenremediation endpoints are based on risk to receptors locatedin a surrounding community. The difference identified inthis example may not appear to be significant, (i.e., 99.89%reduction to meet groundwater standards versus 99.93%reduction to meet indoor air standards). However, it is likelythat other compounds (e.g., less soluble, more volatile) ordifferent source/site characteristics (e.g., more heterogeneous)would result in a difference in remediation goals that isgreater. Moreover, at a site where natural attenuation is beingemployed as the remediation strategy and where sourcedissolution is controlled by mass transfer limitations – a situa-tion common to many NAPL sites – an increase in remediationgoal of a fraction of a percent in groundwater concentrationmay still represent a substantial period of time. Second,this example indicates that the remediation of higher-concentration sources to satisfy regulatory limits for ground-water or indoor air concentrations likely requires substantialreductions in source concentration (e.g., greater than 90% oreven 99%) even when receptors in the wider community are

considered. For lower-concentration sources, such as thoseproduced by mass-transfer limited source zones, sorbed mass,or multi-component NAPLmixtures, as extensive removal maynot be necessary.

5. Conclusions

Groundwater contamination can pose a risk to communitiesthat surround contaminated sites. Assessing this risk, andplanning remedial action, requires tools that link hydrogeologic,physical, and chemical processes to contaminant transport todrinking water and indoor air. The principal contribution of thisstudy is the development of themodel GW-VAP3D, based on thecoupling of three existing models, which allows the simulationof contaminant transport in the saturated zone and vapourtransport in the vadose zone followed by vapour intrusion intobuildings. This model includes an array of key processes and itscomputational efficiency permits rapid (and, thus, multiple)simulations at the community scale. It is anticipated that thismodel, and the coupled approach used in this study, could beused to examine the risk associated with multiple buildingslocated within a large-scale (kilometres) community, such asthose considered in area-wide risk assessments. It will beparticularly useful for the investigation of heterogeneity andparameter uncertainty using multiple realizations (i.e., MonteCarlo) and under conditions where groundwater concentra-tions can evolve over timescales similar to those required forthe development of soil gas profiles above the plume.

To demonstrate thismodel and the coupled approach, a suiteof demonstration simulations was presented. These simulationsused conservative assumptions consistent with those used todevelop regulatory guidance (MOE, 2011) and a range of sourceconcentrations. Demonstration simulations using a high con-centration source zone (i.e., 50% of the aqueous solubility)showed that regulatory standards may be exceeded in ground-water and indoor air at receptors located at substantial lateraldistances from the plume centreline at key intermediate

50 N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

distances where plume spread is significant. For the conditionsinvestigated here these key distances were between 500 m and1000 m downgradient at a lateral distance of 125 m from theplume centreline. It is expected that the area at risk will dependon site-specific hydrogeological, physical, and chemical condi-tions, which is the subject of future work. The coupled modelcan help quantify how the extent of source remediationrequired to meet acceptable levels of risk depends on wherethat risk is assessed. These simulations suggest that, while theamount of necessary source remediation is reduced whencommunity receptors were considered instead of at the siteboundary, nevertheless large (i.e., N99.9%) reductions in sourceconcentrations may be necessary when high-concentration(e.g., those resulting from NAPL dissolution) source zones areconsidered. Finally, the demonstration simulations showedthat different contaminant physicochemical properties, shownhere for benzene and naphthalene, can control whethercontaminant transport to groundwater or indoor air controlsthe overall risk at a community receptor, and which standardmay bemet first following remedial action taken at the source.

It is acknowledged that the development of this coupledmodel, and the presentation of these demonstration simula-tions, are only a further step on the path to more fullyconsidering community-wide risk assessment in the contextof site remediation. Additional simulations using conditionsthat are less conservative than the very conservative ones(i.e., shallow water table, no biodegradation in the saturatedzone, and no recharge) adopted here within a regulatorycontext are required to capture behaviour at typical contam-inated sites. Model validation, specifically the comparison ofpredicted results to actual concentrations in a relevant fieldscenario, would provide confidence that all of the keyprocesses are included. It is hoped that this work will helplead to such datasets being published. Furthermore, it isacknowledged that additional laboratory and field studieswould be beneficial for better statistical characterization ofthe parameter values that need to be populated in the model.Finally, it is clear that suites of Monte Carlo simulations arerequired to correlate aquifer statistical properties (e.g., meanand variance of permeability, retardation, biodegradationrates) to the probability of exceeding risk at communityreceptors; this will be explored in subsequent publications.

Acknowledgements

Funding for this research was provided by the OntarioMinistry of the Environment, Best in Science-Research GrantProject 89011. The authors would like to thank the projectsteering committee. We acknowledge the contributions fromSHARCNET, a consortium of Canadian academic institutionswho share a network of high performance computers. Wealso thank Tarek Rashwan for his assistance. Finally, theauthors thank Dr. Greg Davis and two anonymous reviewersfor their helpful and insightful comments.

References

Abreu, L.D.V., Johnson, P.C., 2005. Effect of vapor source - building separationand building construction on soil vapor intrusion as studied with athree-dimensional numerical model. Environ. Sci. Technol. 39 (12),4550–4561.

Abreu, L.D.V., Johnson, P.C., 2006. Simulating the effect of aerobic biodegra-dation on soil vapor intrusion into buildings: influence of degradationrate, source concentration, and depth. Environ. Sci. Technol. 40 (7),2304–2315.

Abreu, L.D.V., Ettinger, R., McAlary, T., 2009. Simulated soil vapor intrusionattenuation factors including biodegradation for petroleum hydrocar-bons. Ground Water Monit. Rem. 29 (1), 105–117.

Aziz, C.E., Newell, C.J., Gonzales, J.R., Haas, P., Clement, T.P., Sun, Y., 2000.BIOCHLOR—Natural attenuation decision support system v1.0, user'smanual. U.S. Environmental Protection Agency Report EPA/600/R-00/008.EPA Center for Subsurface Modeling Support (CSMOS), Ada, Oklahoma.

Baker, J.R., Mihelcic, J.R., Luehrs, D.C., Hickey, J.P., 1997. Evaluation ofestimation methods for organic carbon normalized sorption coefficients.Water Environ. Res. 136–145.

Barber, C., Davis, G.B., Briegel, D., Ward, J.K., 1990. Factors controlling theconcentration of methane and other volatiles in groundwater and soil-gas around a waste site. J. Contam. Hydrol. 5 (2), 155–169.

Barry, D.A., Bajracharya, K., Crapper, M., Prommer, H., Cunningham, C.J.,2000. Comparison of split-operator methods for solving coupledchemical non-equilibrium reaction/groundwater transport models.Math. Comput. Simul. 53 (1–2), 113–127.

Barry, D.A., Prommer, H., Miller, C.T., Engesgaard, P., Brun, A., Zheng, C., 2002.Modelling the fate of oxidisable organic contaminants in groundwater.Adv. Water Resour. 25 (8–12), 945–983.

Bear, J., 1979. Hydraulics of Ground Water. Bismarck, N.D.(v, 84 p. pp.).Bedient, P.B., Rifai, H.S., Newell, C.J., 1994. Ground water contamination:

transport and remediation. Prentice-Hall International, Inc.Bozkurt, O., Pennell, K.G., Suuberg, E.M., 2009. Simulation of the vapor

intrusion process for nonhomogeneous soils using a three-dimensionalnumerical model. Ground Water Monit. Rem. 29 (1), 92–104.

Cvetkovic, V.D., Shapiro, A.M., 1990. Mass arrival of sorptive solute inheterogeneous porousmedia. Water Resour. Res. 26 (9), 2057–2067.

Cvetkovic, V., Dagan, G., Cheng, H., 1998. Contaminant transport in aquiferswith spatially variable hydraulic and sorption properties. Proc. R. Soc.Lond. A 454, 2173–2207.

Dagan, G., 1994. The significance of heterogeneity of evolving scales totransport in porous formations. Water Resour. Res. 30, 3327–3336.

Davis, G.B., Patterson, B.M., Trefry, M.G., 2009. Evidence for instantaneousoxygen-limited biodegradation of petroleum hydrocarbon vapors in thesubsurface. Ground Water Monit. Rem. 29 (1), 126–137.

DeVaull, G.E., 2007. Indoor vapor intrusionwith oxygen-limitedbiodegradationfor subsurface gasoline source. Environ. Sci. Technol. 41, 3241–3248.

DeVaull, G., Ettinger, R., Gustafson, J., 2002. Chemical vapor intrusion fromsoil or groundwater to indoor air: significance of unsaturated zonebiodegradation of aromatic hydrocarbons. Soil Sediment Contam. 11 (4),625–641.

Fischer, M.L., Abra, J.B., Kristie, A.D., Alfred, T.H., William, W.N., Richard, G.S.,Joan, M.D., 1996. Factors affecting indoor air concentrations of volatileorganic compounds at a site of subsurface gasoline contamination.Environ. Sci. Technol. 30 (10), 2948–2957.

Freeze, R.A., Cherry, J.A., 1979. Groundwater. Prentice-Hall.Gelhar, L.W., Axness, C.L., 1983. Three-dimensional stochastic analysis of

macrodispersion in aquifers. Water Resour. Res. 19 (1), 161–180.Harbaugh, A.W., Banta, E.R., Hill, M.C., McDonald, M.G., 2005. MODFLOW-

2000, the U.S. Geological Survey Modular Ground-Water Model—UserGuide to Modularization Concepts and the Ground-Water Flow Process.U.S. Geological Survey Open File Report 00-92.

Hers, I., Zapf-Gilje, R., Li, L., Atwater, J., 2001. The use of indoor airmeasurements to evaluate intrusion of subsurface VOC vapors intobuildings. J. Air Waste Manag. Assoc. 51 (9), 1318–1331.

Hess, K.M., Wolf, S.H., Celia, M.A., 1992. Large-scale natural gradient tracertest in sand and gravel, Cape-Cod, Massachusetts.3. Hydraulic conduc-tivity variability and calculated macrodispersivities. Water Resour. Res.28 (8), 2011–2027.

Höhener, P., Duwig, C., Pasteris, G., Kaufmann, K., Dakhel, N., Harms, H., 2003.Biodegradation of petroleum hydrocarbon vapors: laboratory studies onrates and kinetics in unsaturated alluvial sand. J. Contam. Hydrol. 66 (1),93–115.

Johnson, P.C., Ettinger, R.A., 1991. Heuristic model for predicting the intrusionrate of contaminant vapors into buildings. Environ. Sci. Technol. 25 (8),1445–1452.

Lahvis, M.A., Baehr, A.L., Baker, R.J., 1999. Quantification of aerobicbiodegradation and volatilization rates of gasoline hydrocarbons nearthe water table under natural attenuation conditions. Water Resour. Res.35 (3), 753–765.

Laubacher, R.C., Bartholomae, P., Velasco, P., Reisinger, H.J., 1997. AnEvaluation of the Vapor Profile in the Vadose Zone above a GasolinePlume. The Petroleum Hydrocarbons and Organic Chemicals in GroundWater: Prevention, Detection and Remediation Conference, Houston, TX,pp. 396–409.

51N. Mustafa et al. / Journal of Contaminant Hydrology 156 (2014) 38–51

LeBlanc, D.R., Garabedian, S.P., Hess, K.M., Gelhar, L.W., Quadri, R.D.,Stollenwerk, K.G., Wood, W.W., 1991. Large-scale natural gradient tracertest in sand and gravel, Cape Cod, Massachusetts, 1. Experimental designand observed tracer movement. Water Resour. Res. 27, 895–910 (no. 3).

Loureiro, C.O., Abriola, L.M., Martin, J.E., Sextro, R.G., 1990. Three-dimensionalsimulation of radon transport into houses with basements under constantnegative pressure. Environ. Sci. Technol. 24 (9), 1338–1348.

Ma, H.W., 2002. Stochastic multimedia risk assessment for a site withcontaminated groundwater. Stoch. Env. Res. Risk A. 16 (6), 464–478.

Malina, G., Krupanek, J., Sievers, J., Grossmann, J., ter Meer, J., Rijnaarts, H.,2006. Integrated management strategy for complex groundwatercontamination at a megasite scale. Viable Methods of Soil and WaterPollution Monitoring, Protection and Remediation, 69, pp. 567–577.

Massmann, J., Freeze, R.A., Smith, L., Sperling, T., James, B., 1991.Hydrogeological decision-analysis.2. Applications to groundwater con-tamination. Ground Water 29 (4), 536–548.

Maxwell, R.M., Kastenberg, W.E., 1999. Stochastic environmental riskanalysis: an integrated methodology for predicting cancer risk fromcontaminated groundwater. Stoch. Env. Res. Risk A. 13 (1–2), 27–47.

Millington, R., Quirk, J., 1961. Permeability of porous solids. Trans. FaradaySoc. 57, 1200–1207.

MOE (Ministry of the Environment), 2011. Rationale for the development ofsoil and ground water standards for use at contaminated sites in Ontario.Prepared by: Standards Development Branch, Toronto. April 15, 2011,PIBS 7386e01.

Nazaroff, W., Feustel, H., Nero, A., Revzan, K., Grimsrud, D., 1985. Radontransport into a detached one-story house with a basement. Atmos.Environ. 19 (1), 31–46.

Neuman, S.P., Tartakovsky, D.M., 2009. Perspective on theories of anomaloustransport in heterogeneous media. Adv. Water Resour. 32, 670–680.

Ostendorf, D.W., Kampbell, D.H., 1991. Biodegradation of hydrocarbonvapors in the unsaturated zone. Water Resour. Res. 27 (4), 453–462.

Ostendorf, D.W., Hinlein, E.S., Lutenegger, A.J., Kelley, S.P., 2000. Soil gastransport above a jet fuel/solvent spill at Plattsburgh Air Force Base.Water Resour. Res. 36 (9), 2531–2547.

Prommer, H., Barry, D.A., Zheng, C., 2003. MODFLOW/MT3DMS-based reactivemulticomponent transport modeling. Ground Water 41 (2), 247–257.

Robin, M.J.L., Gutjahr, A.L., Sudicky, E.A., Wilson, J.L., 1993. Cross-correlatedrandom field generation with the direct Fourier Transform Method.Water Resour. Res. 29 (7), 2385–2397.

Roggemans, S., Bruce, C.L., Johnson, P.C., 2002. Vadose zone natural attenuationof hydrocarbon vapors: an empirical assessment of soil gas vertical profiledata. American Petroleum Institute Technical Report. Washington, DC.

Rushton, K.R., Redshaw, S.C., 1979. Seepage and Groundwater Flow.Numerical Analysis by Analog and Digital Analysis. John Wiley &Sons0471997544 (327 pp.).

Sudicky, E.A., 1986. Natural gradient experiment on solute transport in asand aquifer: spatial variability of hydraulic conductivity and its role inthe dispersion process. Water Resour. Res. 22, 2069–2082.

Sudicky, E.A., Illman, W.A., Goltz, I.K., Adams, J.J., McLaren, R.G., 2010.Heterogeneity in hydraulic conductivity and its role on the macroscaletransport of a solute plume: from measurements to a practicalapplication of stochastic flow and transport theory. Water Resour. Res.46, W01508. http://dx.doi.org/10.1029/2008WR007558.

USEPA, 1996. Soil Screening Guidance: technical background document. EPA540-R-95-128 (July 1996).

USEPA, 1998. OSWER draft guidance for evaluating the vapor intrusion toindoor air pathway from groundwater and soils (subsurface vaporintrusion guidance) tables. EPA 530-D-02-004 (November 2002).

USEPA, 2001. Fact Sheet: Correcting Henry's Law Constant for SoilTemperature (June 2001).

USEPA, 2003. User's Guide for Evaluating Subsurface Vapor Instrusion intoBuildings. 68-W-01-058 (June 2003).

USEPA, 2012. Conceptual Model Scenarios for the Vapor Intrusion Pathway.EPA 530-R-10-003 (February 2012).

Wang, X., Unger, A.J., Parker, B.L., 2012. Simulating an exclusion zone forvapour intrusion of TCE from groundwater into indoor air. J. Contam.Hydrol. 140, 124–138.

Weiss, H., Daus, B., Heidrich, S., Kaschl, A., Schirmer, M., Wycisk, P., 2006.Management options for regionally contaminated aquifers: a case studyat Bitterfeld, Germany. Viable Methods of Soil and Water PollutionMonitoring, Protection and Remediation, 69, pp. 579–589.

Woodbury, A.D., Sudicky, E.A., 1991. The geostatistical characteristics of theBorden aquifer. Water Resour. Res. 27, 533–546.

Wycisk, P., Weiss, H., Kaschl, A., Heidrich, S., Sommerwerk, K., 2003.Groundwater pollution and remediation options for multi-sourcecontaminated aquifers (Bitterfeld/Wolfen, Germany). Toxicol. Lett. 140,343–351.

Yao, Y., Shen, R., Pennell, K.G., Suuberg, E.M., 2013. A review of vaporintrusion models. Environ. Sci. Technol. 47 (6), 2457–2470.

Yu, S., Unger, A.J.A., Parker, B., 2009. Simulating the fate and transport of TCEfrom groundwater to indoor air. J. Contam. Hydrol. 107 (3–4), 140–161.

Zheng, C., Weaver, J., Tonkin, M., 2010. MT3DMS: a modular three-dimensional multispecies transport model: user guide to the hydrocar-bon spill source (HSS) package. Athens, Georgia: Prepared undercontract to the U.S. Environmental Protection Agency.