RathFelder 1991

Journal of Contaminant Hydrology, 8 (1991) 263-297 263
Elsevier Science Publishers B.V., Amsterdam
Mathematical simulation of soil vapor extraction

systems: model development and numerical examples
Klaus Rathfelder", William W-G. Yeh a and Douglas Mackay b

~Department of Civil Engineering, UCLA, Los Angeles, CA 90024, USA
bSchool of Public Health, UCLA, Los Angeles, CA 90024, USA
Received September 13, 1989; revised version accepted March 27, 1991
ABSTRACT
Rathfelder, K., Yeh, W.W-G. and Mackay, D., 1991. Mathematical simulation of soil vapor
extraction systems: model development and numerical examples. J. Contam. Hydrol., 8:
263-297.
This paper describes the development of a numerical model for prediction of soil vapor
extraction processes. The major emphasis is placed on field-scale predictions with the objective
to advance development of planning tools for design and operation of venting systems. The
numerical model solves two-dimensional flow and transport equations for general n-component
contaminant mixtures. Flow is limited to the gas phase and local equilibrium partitioning is
assumed in tracking contaminants in the immiscible fluid, water, gas, and solid phase. Model
predictions compared favorably with analytical solutions and multicomponent column venting
~xperiments. Sensitivity analysis indicates equilibrium phase partitioning is a good assumption
in modeling organic liquid volatilization occurring in field venting operations. Mass transfer
rates in volatilization from the water phase and contaminant desorption are potentially rate
limiting. Simulations of hypothetical field-scale problems show efficiency of venting operations
is most sensitive to vapor pressure and the magnitude and distribution of soil permeability.
INTRODUCTION
Nonaqueous Phase Liquids (NAPL'S)immobilized as residual saturation in

the unsaturated zone are a potential source for groundwater pollution due to
leaching and transport from percolating recharge. Residual saturation can
vary from 3-51 m -3 for highly permeable soil up to 30-501 m - 3 for soils with
low permeability (Schwille, 1984). A traditional remediation practice for
contaminated soils is excavation and subsequent treatment or disposal;
however, economic and physical constraints often restrict applicability of
excavation operations. Soil vapor extraction (also known as soil venting) is an
alternative remediation strategy targeting the removal of volatile contami-
nants from the unsaturated zone. This method has the advantage that it is
implemented in place, eliminating excavation requirements. Soil vapor
0169-7722/91/$03.50 © 1991 Elsevier Science Publishers B.V. All rights reserved.

264 K. RATHFELDER ET AL
vent / vapor
injection
(optional) isepal
Typical Desian Criteria

well material: 2-4 in. PVC
screen interval: 5-20 ft.
flowrate: 50-2000 cfm
well spacing: 10-50 ft.
Fig. I. Typical soil vapor extraction system.
extraction, under suitable circumstances, is a superior remediation method

applicable at large or deep contamination sites or sites hindered by physical
obstacles.
The soil vapor extraction methodology generates advective vapor fluxes
through the contaminated soil using vadose zone wells or horizontal piping
together with air blowers or vacuum pumps. Vapor fluxes induce volatil-
ization of the contaminants which are then removed from the soil via the
extracted gas phase. Above ground the gases are treated or discharged directly
to the atmosphere, depending on local ordinances. A schematic of a typical
soil vapor extraction system is shown in Figure 1.
The soil vapor extraction technology was developed in the late 1970's and
early 80's and has evolved quickly to widespread field application. Initial
feasibility studies were performed by the Texas Research Institute (1980) and
laboratory investigations are reported by Thornton and Wootan (1982),
Texas Research Institute (1984), Marley and Hoag (1984) and Bloes et al.
(1991). Crow et al. (1985, 1987) describe a highly monitored field-scale
experiment. Each of these studies provides evidence that soil vapor extraction
methods are effective for control of subsurface vapors and removal of con-
taminant residual saturation. Successful application to field-scale contami-
nation sites is reported by Agrelot et al. (1985), Camp, Dresser and McKee,
Inc (1987), Connor (1988), Regalbuto et al. (1988), Hutzler et al. (1988), and
Downey and Elliott (1990).
A large number of physical and system variables impact the applicability,
design, operation, and in general, the efficiency and success of soil vapor
extraction systems. We classify these variables as chemical, field, and system
variables. Chemical parameters govern the contaminants potential to
MATHEMATICAL SIMULATION OF SOIL VAPOR EXTRACTION SYSTEMS 265
partition into the moble gas phase. These include the contaminant vapor
pressure, Henry's Law constant, and sorption affinity. Gas flow capacities
through the contaminated soil are controlled by field conditions, including the
soil type and distribution, moisture distribution, and NAPE distribution.
System parameters effect the efficiency of venting operations. Important
design and management parameters are the positioning of the vadose zone
wells, pumping rates and pumping locations.
The development of site-specific design and operation criteria is hampered
by the variability of field parameters and complexity of interacting transport
phenomena. Frequently, the design and implementation of soil vapor
extraction systems is accomplished by empirical methods based upon engi-
neering judgement and experience. Development of design guidelines have
been aided by the application of numerical models to investigate various
parameter sensitivities. Numerical models also provide the means to investi-
gate the physical and system parameters controlling site-specific soil vapor
extraction operations, to the extent that data requirements can be fulfilled.
A variety of numerical models have been developed for prediction of
induced advective vapor flow and contaminant transport in porous media.
Several researchers have used analytical solutions to develop general response
expressions. Massmann (1989) developed an expression for vapor flux that is
similar in form to the transient groundwater flow equation. The equation was
solved analytically and results were compared with field data from a gas
extraction test. Johnson et al. (1990) developed analytical expressions for a
number of system responses: time required for steady-state vapor flow, height
of water table upwelling, temperature effects on vapor flow, venting-induced
vapor and liquid compositional changes, and venting rates in heterogeneous
soils with free product boundaries.
Other researchers have applied numerical simulation methods to study
advective transport of contaminant vapor~ "Iere the focus was primarily on
the control of subsurface contaminant vapors using soil venting techniques;
the role of entrapped residual NAPE'S was not considered in these studies.
Wilson et al. (1987) developed a two-dimensional model tracking a two-
component vapor phase; the source of the contaminant vapors was a floating
product pool. Predictions compared favorably with experimental results. A
major conclusion was vapor extraction processes should be modeled as a
multicomponent mixture for both vapor and liquid phases. Krishnayya et al.
(1988) developed a two-dimensional finite-element model for predictions of
flow and pressure fields resulting from soil vapor extraction operations.
Predicted pressure fields were compared against laboratory measurements
showing close agreement.
Several researchers have developed predictive models linking vapor and
liquid contaminant transport processes. Stephanatos (1988) developed a
266 K. RATHFELDER ET AL.
two-dimensional finite-difference model for transport of a single volatile

species in a two-phase air-water system. The model was applied to a field
contamination site. Independent predictions of venting removal rates for
toluene and xylene show generally good agreement with measured rates,
although tailing effects were not captured. Also, as pointed out by Wilson
et al. (1987), solution of a single transport equation does not adequately
represent compositional changes in the liquid phases. Baehr et al. (1989)
developed and tested a one-dimensional multicomponent compositional
transport model. This model has the capability to track multicomponent
vapor flux in a three-phase air-water-NAPL system. The model was tested with
comparisons against gasoline venting experiments in laboratory soil columns.
Tracking five representative gasoline constituents, predictions of the total
hydrocarbon effluent showed good agreement with measured values.
Rathfelder (1989) developed and tested a two-dimensional numerical multi-
component compositional transport model for simulating soil vapor
extraction systems.
The objective of this study was the development of a numerical model for
investigation of field-scale vapor extraction operations. The model described
here solves two-dimensional flow and transport equations for general n-
component contaminant mixtures. The model can be used to study problems
with heterogeneous soil and fluid distributions; conditions typifying field
settings. Two-dimensional solutions also provide the capability to study
design and management parameters. A theoretical analysis describing the
mathematical formulation, assumptions, and solution scheme is presented.
The model is then applied to one- and two-dimensional hypothetical test
problems. Illustrative examples demonstrate the influence of chemical, field,
and system parameters in vapor extraction operations.
MATHEMATICAL DEVELOPMENT
Organic contaminants released to the subsurface can potentially partition

in a number of phases in the unsaturated zone: volatilization to the gas phase,
dissolution into soil water, adsorption onto the solid particles and capillary
retention of nonaqueous phase liquids (NAPL). Depending on chemical
characteristics, contaminant partitioning in the water and solid phases may
slow contaminant removal by soil vapor extraction methods. Accordingly,
the mathematical formulation accounts for mass storage in the water, solid,
NAPL and vapor phases. For convenience of notation, single or multi-
component NAPL'S are referred to as "oil" in this paper.
The porous medium is conceptualized by four overlapping continua
denoted by a: air (a), water (w), oil (o) and solid (s). Each phase can be
comprised by a number of components denoted by 7- A general macroscopic
mass balance equation is expressed for a representative y-component in a

a-fluid phase:
m
~t (nS~p~.) = -V'nS~(p~;.Vm~ + J~ + J';.) + nS~p~F~;. - I~'~ (1)
where t is time [T]; n is the dimensionless soil porosity; S~ is the dimensionless

fluid saturation; p~r is the mass concentration of the ,/-component in the a-fluid
phase [M/L3], such that:
~~ p ~ , = p~,
7,=1
where p~ is the mass density of the a-fluid [M/L 3] and N, is the number of
representative components comprising the a-phase; p~. Vm,, J~., JS are the
7-component mass flux in the a-phase by advection, kinematic dispersion and
molecular diffusion, respectively [M/LZT]; F,~, is the rate of internal 7-
component mass production per unit mass of a-fluid [l/T], and I~ represents
interphase mass transfer between the a-phase and all other adjacent phases,
/~[M/L 3T]. A similar equation is expressed for the solid phase with volumetric
concentration expressed as PsPs~., where Ps is the bulk soil density and p~..is the
solid phase mass concentration in weight per weight.
Advective fluxes
Soil vapor extraction is assumed to be implemented after discovery

and elimination of the contaminant source. New contaminants will not be
introduced and contaminants are assumed to have thoroughly drained by
gravity through the unsaturated zone with residual quantities retained and
immobilized by capillary forces. These oil-air capillary forces are assumed
sufficient to overcome momentum transfer from the mobile gas phase. Defor-
mation of the soil matrix is also considered negligible.
Soil venting operations extract water and air when conducted in regions of
large soil moisture content, such as near the water table or following drainage
episodes of water recharge. Generally, water extraction is undesirable because
water recovered from the contamination region will likely require treatment.
In practice, water extraction can be minimized by use of an impermeable
ground cover, large-diameter extraction wells, and secondary water wells to
lower the water table (Hutzler et al. 1988). In this paper, we assume field
conditions and design features are such that soil moisture remainsat or below
residual saturation (field capacity) and is unaffected by the mobile air phase.
Thus, the water phase is immobile, and model applicability is consequently
restricted to scenarios with low moisture content.
Under the assumptions above, the gas phase is the sole mobile phase. Gas
phase pressures are assumed to be predominantly controlled by applied

stresses at the extraction wells and independent of liquid-phase pressures.
Furthermore, oil saturation below the residual level cannot be related to the
capillary pressure. As a result, pressures of the immobile liquid phases are not
required in the mathematical development.
The vapor-phase specific discharge is evaluated with Darcy's Law,
assuming the gravity potential is negligible:
nSa Vm~ -- kkra V P (2)

#a
where k is the intrinsic permeability [L2]; kr, is the dimensionless gas phase
relative permeability;/t a is the gas phase dynamic viscosity [M/LT]; and P is
the gas phase pressure [M/LT2]. In alluvial soils with good permeability and
a total liquid saturation below field capacity, the impedance to gas flow from
the liquid phases will be small. However, for illustrative purposes the gas
relative permeability was evaluated by:
kra ~-- Sal/2[1 - (1 - 8 a ) 1 - 4 3 ] 1'4 (3)
At low pressures, gas flux through soil may violate Darcy's Law due to
slip-flow conditions (Klinkenberg effect), that effectively increase the air-
phase permeability (Bear, 1972). Corey (1986) states errors contributed by the
Klinkenberg effect, are "usually small in the case of coarse sands, but huge in
the case of silts or clay." In this study the Klinkenberg effect is ignored by
restricting analysis to soils with relatively high conductivity; however, the
simulation code can easily accommodate permeability correction factors to
account for the Klinkenberg effect.
The air-phase mixture viscosity was estimated with a semi-empirical
formula given by Bird et al. (1960):
P~ = ~ u (4)
j=l
in which
where ~ and Xj are air-phase mole fractions; /~, and pj are component
viscosities at the system temperature and pressure; and M~, and Mj are the
component molecular weights.
Dispersive fluxes
Dispersive fluxes are assumed significant only in the gas phase; liquid and
surface diffusion are neglected. A traditional groundwater modelling
approach is used in applying Ficks Law to evaluate combined processes of
kinematic dispersion and molecular diffusion:
,j(d _[_ "/a~ = --Dah~Vpa7 (6)
where Dh~ [L2/T] is the gas phase dispersion tensor. For an isotropic medium
the dispersion tensor is evaluated by (Bear, 1972):
h v, vj ,
(O~7)i j = avlv]~ij + (a L -- a T ) - ~ + D.:, (7)
where aL and aT are longitudinal and transverse aquifer dispersivities [L],

respectively; Viiand Vj are the components of the absolute value of the Darcy
velocity IV]; 5ij is the Kronecker delta; and D.*, is the effective gas phase soil
molecular diffusion coefficient [L2/T]. The effective soil molecular diffusion
coefficient is commonly related to measured diffusion coefficients for unob-
structed fluids. For an isotropic medium, Millington (1959) relates the area
available for flow to the pore volume by the expression:
D.*. _ ,0/3 [nSa],O/3
_ n. _ (8)
Da~, n2 n2
where Da~ is the unobstructed vapor-phase molecular diffusion coefficient and

na = nSa) is the air-filled porosity. Published values of binary air-contami-
nant diffusion coefficients are used as estimates for the mixture diffusion
coefficient.
Internal sources and sinks
Organic contaminants can undergo chemical or biological transformations

in the subsurface (e.g. Vogel et al., 1987). Such reactions could theoretically
reduce the modeled contaminant mass, or if the transformation products were
themselves contaminants being modeled, increase the mass. These reactions
were neglected for the following reasons.
The rates of chemical transformations of many of the contaminants
amenable to soil venting are slow relative to the duration of the venting
process. For the aromatic compounds typically of regulatory concern in fuel
spills, chemical transformations are considered negligible (Harris, 1982). For
chlorinated solvents, the reported half-lives for chemical transformations
range from several months to many years (Harris, 1982; Vogel et al., 1987),
and in some cases the transformation products may themselves be toxic
volatile compounds (Mackay and Vogel, 1985). The conservative assumption

in both cases is likely to be that no reaction occurs (i.e. that reactions occur
very slowly or create volatile contaminants that themselves must be removed).
Similar issues and uncertainties plague the prediction of biotransformation
of contaminants during soil venting. For chlorinated solvents known to be
resistant to aerobic degradation and sometimes susceptible to transformation
in the absence of oxygen, the venting process would possibly slow or stop the
biotransformations. For aromatic contaminants known to be readily
degradable aerobically, the venting process can accelerate biotransformation
in the subsurface, which is generally thought to be limited by oxygen avail-
ability, unless the presence of the NAPL inhibits microbial activity. Given the
conceptual difficulties in predicting biotransformation rates we choose to
make the conservative assumption in neglecting these reactions.
Interphase mass transfer
In summary, the macroscopic mass transport equations are expressed as:

Vapor phase
0 h
kkr. VP + nS~D~./Vp,,] + [~7
oZ (n& Pa,,) = V. p~, (9)
Pa
Liquid phases
0
aZ (nS, ) = I2~ ~ = w, o (10)
Solid phase
0t (PsPsT) = If~ (ll)
where I~ represents mass transport of the V-component across interfaces

separating the a-phase from contiguous phases, ft. Interphase mass transport
processes potentially have significant influence on the rate and distribution of
contaminant removal. The importance of interphase mass transfer is
examined with separate models using equilibrium and kinetic phase partition-
ing. In both models we assume the soil remains completely water wet and the
oil phase is partitioned as films between the water and air phases. Such
configurations have been observed with styrene polymer blob casts studied by
Wilson et al. (1988).
Equilibrium model
In the absence of rate data, equilibrium phase partitioning is commonly
assumed (e.g., Baehr et al., 1989). The 7-component transport equations are
summed over the four phases, eliminating all interphase terms and resulting
in a single 7-component transport equation:
-~[nSap~, + nSoPo~ + nSwPw~ + PsPsr] = V" P~r VP + nSaDha~,VP.r

~a
(12)
Supplementary equilibrium partitioning relationships are needed to relate
the phase concentrations. These equilibrium coefficients are evaluated by
expressions described in Corapcioglu and Baehr (1987), and are developed
under the assumption of isothermal conditions and ideal fluid behavior.
The air-oil equilibrium relationships are evaluated from Raoult's Law
together with the Ideal Gas Law:
~o - p~ - P~ (13)
Po.~ R T ~ Pot
7=IM~/~
where P~ is the ~-component vapor pressure at the system temperature, T; R
is the universal gas constant and M r is the ~,-component molecular weight.
Capillarity influence on vapor pressure is ignored.
For common organic compounds with low water solubilities, Henry's Law
and the Ideal Gas Law are combined to evaluate the air-water partition
coefficients:
/Qaw -- Pa), __ M/K~h

Pw~ RT (14)
where/~ is the 7-component Henry's Law constant at the system temperature.

Sorption by geologic media can significantly impact contaminant transport
and fate in the subsurface (Mackay, 1990). Sorption is known to vary for a
given contaminant from site to site (Mackay, 1990) and spatially within a site
(Mackay et al., 1986). In addition, sorption varies among contaminants for a
given sample of geologic medium. Although measurements of the sorption
properties are clearly preferable for input to venting simulations, they are
rarely available in practice.
For illustrative purposes in this development, we use the approach
commonly applied for estimating the equilibrium sorption partition coef-
ficient of soils and sediments:
K~s'w = Kocfoc - Ps~ (15)

P~r
where foe is the organic carbon content of the soil and Koc is the organic-
carbon-normalized partition coefficient. Numerous methods have been
272 K. R A T H F E L D E R ET AL.
proposed to estimate Koc (Lyman, 1982). Here we use the equation reported
by Karickhoff et al. (1979) to correlate Koc with the contaminant water
solubility for ten organic compounds, mostly aromatic:
Log Koc = - 0 . 5 4 Log S + 0.44 (16)
where the ~,-component water solubility S is expressed in mole fraction. The
approach outlined is strictly useful only if foe is well above 0.1% (Karickhoff,
1984), and sorption can be assumed to be at equilibrium. Based on the few
available data, it appears the organic carbon content of sandy aquifers is
typically well below 0.1% (Mackay, 1990). We assume the carbon contents are
sufficiently high to warrant use of equations (15) and (16) but caution that this
assumption must be verified. Other correlations can be easily incorporated in
the simulation code.
Substituting the equilibrium partitioning expressions into (12), the 7-
component equilibrium transport equation is given as:
Ot [nSaPa~,+ nSoI~oPa7+ nSwK~wPa.~+

__ Y Y 7
ps/C.Jw/C~wPaT]
= V. [ Pay kkr"vP
#a + nSaD~,Vpa,] (17)
Nonequilibrium model
When interphase mass partitioning rates deviate significantly from the
equilibrium assumption, then interphase terms in (9)-(11) must be evaluated
and the complete set of coupled transport equations must be solved. A model
of this type was developed to assertain the potential of rate-limiting effects
during venting operations. This model focuses on processes of contaminant
volatilization from the liquid phases and contaminant desorption. Oil/water
interactions resulting after start up of venting operations are neglected.
The interphase mass transfer terms are evaluated by the dual resistance
model. This model assumes: (1) are mass transfer is controlled by the rate of
diffusion on each side of the interface, and (2) no resistance to transfer is
encountered at the interface (Welty et al., 1984). Frequently, the diffusional
resistance on one side of the interface is considered dominant so that mass
transfer rates can be described with a single overall mass transfer coefficient.
In the dual resistance theory, the overall mass transfer coefficients encompass
resistance from diffusion through static fluids only, but in practice are
commonly measured in mass transfer systems where hydrodynamic conditions
affect on regulate diffusional impedence. The overall mass transfer coefficients
are strictly applicable for the measured system only and should be extended
to other systems with caution.
For oil volatilization gas phase control is assumed. The air/oil interphase
term can be expressed as:
2ao[Pa~ -- K~aoPoe] (18)

where
Kga (19)
is the air-oil transfer factor [l/T], Kg is the overall mass transfer coefficient
[L/T] based upon gas phase control [mass transferred/(time) (interfacial area)
(APa~,)], and a = A/Vt [I/L] is the specific surface area [air-oil interfacial
contact area per total volume of soil]. The product K~aoPo~is the air-phase
equilibrium concentration with the oil phase.
For volatile organic compounds, liquid-phase resistance is commonly
assumed to control mass exchange across the air/water interface. Munz and
Roberts (1984) and Roberts et al. (1985), however, demonstrate that gas phase
resistance becomes significant for moderately volatile compounds. For
the volatile contaminants in this study we assume liquid-phase control in
expressing the air/water transfer term as:
law = nSa2aw[KawPar - Pw~,] (20)
Similarly, we assume that water-phase resistance controls the rate of contami-
nant desorption:
Isw = nSw2ws[Pw - Ps~/K~w] (21)
Flow equation
A flow equation describing the gas phase pressure field is formulated from
the general gas phase mass balance expression:
a--t [nSaPa] = -V~nSaflaVma + IaB+ nSaFa (22)
where Iaa is the total interphase mass transfer to the gas phase and F a are the
gas phase source/sinks. We neglect the interphase term, assuming that mass
transfer to the gas phase has negligible effect on the pressure field over the
course of a single time step. Employing Darcy's Law and equating air density
with the Ideal Gas Law, the gas phase flow equation is expressed as:
~--IC[nSa
3 ----~jMaPl= -V " [MaP
L RT kkra#aVP] + nSal-"a (23)
where M a is the gas phase molecular weight.

274 K. RATHFELDER ET A L
Totality conditions
Totality conditions provide additional expressions necessary to close the

system of equations. By definition fluid phase saturations sum to unity:
S~ + So + Sw = 1 (24)
Similarly, for the incompressible liquid phase at constant temperature, the
component volumetric fractions sum to unity, for oil:
~ Po~ _ 1 (25)
7= I P7
and for water:

Pw___~w+ ~ Pw~ _ 1 (26)
Pw ~.- i P~
where N is the number of components comprising the oil phase mixture, P7 is
the V-component density as a pure compound, and Pww is the clean (pure)
water component.
The gas components sum to the gas phase density which is evaluated by the
Ideal Gas Law:
N paMa
Pa. + Paw + Z P.;, = (27)
.1=1 RT
where Paa and Paw are the dry air and water vapor components, respectively.
NUMERICAL SOLUTION SCHEME
The coupled flow and transport equations are solved with fully implicit
finite-difference approximations together with Newton-Raphson iteration.
These methods have been found to provide stable solutions with time-step
restrictions limited only by the initial guess for the iteration procedure. The
flow and transport equations were solved sequentially on different size
meshes. The gas phase pressure distribution was calculated first from the flow
equation (23), assuming that temporal changes in saturation and gas phase
composition have negligible effect on the flow field over the time step. These
solutions are generated on a large grid, so that boundary conditions for gas
pressure are sufficiently far from the contaminant region to minimize their
influence in the solution of the transport equations. The updated air pressure
distribution is then used to solve the transport equations on a smaller grid
encompassing only the region where contaminant transport processes are
significant.
A point distributed finite-difference scheme with variable grid spacing is
used. The finite-difference analogs for the flow equation (23) and the equi-
librium transport equation (17) in one space dimension are respectively:
n, Sa, Ma,, 1 1
• " A,-Pi,+I
RT 2 R T mxi
X [(ma, a)i+il2,,A+P, ,,,+l
2 -- (maJ, a)i_it2,,A?P ,,,+,]
2 + n,&<,r.:,.,+, (28)
n i
IliA t (SaPaT)i,t +, d- --~ [(sog~aoPa?)i,t "-}-(SoPo?)i,t + (SwI~aawPaT)i,t +I -- (SwPwr)i,t]
PSi
q-- -~ [(~aaw-K':wsPa$ )i,t + l -- Psi,,,]
1
- Ax~ [(P"~a)~+ U2.t+,G+ P~.,+, -- (P.<ta)~-,I2.,+,A;P~,+,.
A + h
+ (nSaA~rxx)i+ 1/2,t+ I x Par,,,+, -- (nSaDaTxx)i-,/2,,+ IAxPar/,,+,]
(29)
where A + and A- are the forward and backward difference operators respect-
ively, i and t are indices of the space and time coordinates, and:
kkra
L = (30)
is the gas phase mobility. Gas phase concentration in the transport equation
at the interblock nodal locations are assigned upwind values. Remaining
space-dependent variables are evaluated as arithmetic means.
Newton-Raphson iteration procedures are used to solve the nonlinear
discretized flow and transport equations, employing the Harwell Sparce
Matrix Package (Duff, 1979). Primary dependent variables are gas pressure in
the flow equation, and fluid saturations and phase concentrations in the
transport equations. Secondary dependent variables are lagged; the partition
coefficients are lagged by one iteration, while small variability in the molecular
weight, mobility, and dispersion coefficients justifies delaying their updates
until the end of each time step. Convergence is established with a relative
maximum change criteria:
max IAxil
< e (31)
max Ixel
w h e r e ~ w a s 1 x 10 -4 .
Material balance checks were calculated as a measure of the simulator's
ability to correctly solve the flow and transport equations. A global mass
balance error is calculated for each component using:
error/ ,00f, i=|

(F~.,- Fout,)At,
t
where M{ and M ° is the 7-component mass storage at time t and time zero,
respectively; (Fin, - Fout,)Ati is the net mass entering the domain during time
step Ati; and NI is the number of time steps from zero to t.
VERIFICATION
The numerical simulation model was verified by comparisons with

analytical expressions and against laboratory results from one-dimensional
column experiments. The comparisons with analytical expressions demon-
strate validity of solutions from the transport equations in the absence of
contaminant partitioning. Comparisons against experimental data demon-
strate the model's ability to predict compositional transport between a
residual organic and mobile gas phase in a laboratory soil column.
The one-dimensional advection-dispersion equation for isothermal
transport in a saturated uniform soil column of length L, with constant Darcy
velocity q, can be expressed as:
0p,~. _ q ~Pa~. h 32P~' (33)
~t n ~X + Da:, 3x z
subject to:
p,:. = 0 0 <~ x <<, L, t = O
p,:, = p*, x = 0, t > 0 (34)
p,~. = 0 x = L,t > 0
For these conditions the solution is (Bear, 1972):
p,,. l [ e r f c ( x - qt/n~ (qx/n) (x_ + qt/n~]
p*, - 2 \ - ~ / + exPkDh ~)erfck~/J (35)
The numerical model was adapted to fit this problem by defining a uni-
form pressure gradient and solving only the transport equation (the flow)
equation was not solved). Oil and water saturations were set to zero, and
consequently computed partition coefficients were also zero. Remaining
parameter values are shown in Table 1. Results are plotted in Figure 2 for a
pure diffusion and an advective-dispersion problem. In both cases, close
agreement with the analytical solution was obtained indicating the numerical
model is correctly solving the transport equation given the pressure field.
TABLE 1
Parameters used in comparisons with analytical expressions
k = 10 d a r c y #a = 0 . 0 1 7 8 c p L = 50cm
n = 40% pa* = 198mg/l Ax = 0.5cm
kra = 1 Da~, = 0 . 0 0 8 8 c m 2 / s (free air) t = 0.5 min
~L = lcm
Pure-diffusion case
~P/Ox = 0 q = 0 D *a ; = 1.55cm2/min
Advective-dispersion case
~P/Ox = 1 x 1 0 - S a t m / c m Q/A = 0 . 3 3 7 c m / m i n D a'/
h = 1.89cm2/min
Mackay et al. (1990) conducted laboratory experiments to investigate

individual component behavior during venting of multicomponent organic
liquids from soil columns. In these experiments, mixtures of benzene, toluene
and TCE liquids with known mass, were emplaced at residual levels in a dry
porous medium of either glass beads or Borden sand. The organics were
vented from the column with nitrogen gas under positive gage pressure.
Effluent vapor concentrations of the individual constituents were measured
throughout the venting experiments.
The conditions of the laboratory experiments were simulated with the
one-dimensional equilibrium model. In these simulations, measured values of
permeability, porosity, and initial fluid saturations were assumed uniform
along the column. The gas pressure at the inlet boundary was evaluated with
the known nitrogen flowrate and atmospheric pressure was used at the outlet
1
"0 -]t'~
-'d~~ & AD - simulation (t=20 min)
] ~ ~ ~ AD- analytical(t=20 min)
0 8 -I ¢'~ ~ ---- PD - simulation (t-120 min)
0.4
0.2
0.0 -
0 10 20 30 40 50
position (cm)
Fig. 2. Comparison of numerical and analytical solutions for advective-dispersion (AD) and pure diffusion
(PD) problems.
278 K. RATHFELDER ET A L
TABLE 2
Parameters used in simulations of laboratory column experiments
Glass Sand
Physical
k (darcy) 50.0 50.0
n (%) 37.3 39.0
~tL (cm) 0.1 0.1
initial So 5 % 8.51 7.97
Vapor pressure (atm)

benzene 0.118 0.118
TCE 0.092 0.092
toluene 0.029 0.029
Initial mole fraction

benzene 0.467 0.432
TCE 0.277 0.255
toluene 0.257 0.313
relative permeability (kra) -~ 1

c o l u m n length (L) = 3 0 c m
c o l u m n d i a m e t e r -- 2 . 5 c m
flow rate (Q) = 30 m l / m i n
temperature ( T ) = 23°C
Model Ax = 1.0cm
At = 0 . 0 5 h
boundary:
OP -Q #a
- t>O,x = 0
~x A kk~a
(36)
P = latin t > O,x = L
where Q is the measured gas flow rate and A is the column cross-sectional
area. The gas composition at the inlet boundary was pure nitrogen, and
contaminant transport across the outlet boundary was assumed to occur b y
advection only:
Pay = 0 t > 0, X = 0,~ = 1, N
(37)
Opal,
t~X -- 0 t > O,X = L, 7 = I,N
Remaining parameters used in these simulations are given in Table 2.

Figures 3 and 4 compare experimental and simulated effluent vapor
250-
[] Benzene
~200 1 • TCE
o Toluene
8 iso,i Predi~ed
~ 100"
3
50:
• , . , • ,
2 4 6 8 10 12 14
Time (hrs)
Fig. 3. Experimental and predicted effluent vapor concentrations during venting from glass beads.
concentrations during venting from glass beads and Borden sands, respectively.
Mass balance results are presented in Table 3. Both experimental and
predicted results show that the more volatile constituents are removed ahead
of successively less volatile components. The experimental results also found
evidence of contaminant adsorption from the gas phase, manifested by tailing
of effluent concentrations in the soil column. Mechanisms of solid-vapor
interaction, however, were not a part of the numerical code and consequently
were not captured in the simulation results.
Overall, comparisons appear to be reasonably good. This suggests that in
the laboratory experiments, equilibrium vapor concentrations were estab-
lished prior to exiting the column, and the equilibrium model captured the
essential organic phase compositional changes necessary to predict effluent
vapor concentrations. The experimental results, however, do not provide
~" 200
a Benzene
~150 • TCE
o Toluene
i ~ Predicited
E 100
g
o
~ 5o
Q.
• , . , . , . L , ~ . , .~, ,
2 4 8 10 12 14 16 18
Time (hrs)
Fig. 4. Experimental and predicted effluent vapor concentrations during venting from Borden sand.
280 K. RATHFELDERET AL.
TABLE 3
Cumulative mass balance errors in the simulations of laboratory experiments

Problem Component % Error
L a b o r a t o r y c o l u m n experiment - Benzene 2.1

Glass Bends Toluene 1.9
Time = !3 h o u r s Xylene 1.I
L a b o r a t o r y c o l u m n experiment - Benzene 2.2

B o r d o n Sand Toluene 2.0
Time = 17 h o u r s Xylene 1.1
information about the rate at which equilibrium concentrations were esta-

blished, and accordingly, predictions of effluent vapor concentration do not
provide model validation for spatial and compositional changes in the organic
phase.
The equilibrium model was next applied to hypothetical field-scale
problems in order to demonstrate the behavior of vapor extraction processes
in representative field settings, and to investigate sensitivities to various design
and field parameters. These simulations, unlike the laboratory problems, also
tested the model under conditions with highly nonuniform flow fields and
contaminant partitioning in water and soil phases. Contaminants selected for
these simulations were benzene, toluene and xylene, representing commonly
encountered organic pollutants with a moderate range in vapor pressure.
Table 4 lists chemical data used in these simulations.
The equations were solved first in one-dimensional radial coordinates. Here
solutions may represent axial symmetric horizontal flow to a single extraction
well positioned at the center of cylindrical zone of contaminated soil. The flow
equation was solved in the domain from the well screen to the radius of
influence; rwdI to r~. Atmospheric pressure was assumed at r~, while at the
well screen, gas pressure was evaluated from the specified vapor extraction
rate, Q:
Q _ Prwe,,kkra OP[ (38)

A Patm #a Or I rwe,l
The transport equations were solved in the domain from the well screen to a
radius, re, where transport processes were considered insignificant. At this
radius, contaminant vapor concentrations were taken to be zero. Contaminant
transport at the well screen boundary was assumed to occur by advection only:
Op,~0r r . e , , = 0 (39)
TABLE 4
Chemical properties of test contaminants
Benzene Toluene o-Xylene Water Dry air
Molecular weight (g/mole) 78.1 92.1 106.2 18.0 28.98

Density (g/l) (20°C) 879.0 867.0 980 998.0 1.21
Vapor pressure (atm)
20°C 0.102 0.0294 0.0092 0.00975
10°C 0.061 0.0171 0.0051 0.0091
Water solubility (g/l)
20°C 1.78 0.52 0.175
10°C 1.5 0.37 0.15
Henry's Constant (atml/g)
20°C 0.0573 0.0565 0.0524
I 0°C 0.0407 0.0422 0.04
Vapor viscosity (cpoise)
20°C 0.0075 0.0069 0.007 0.0097 0.0181
10°C 0.007 0.0065 0.0065 0.009 ! 0.0178
Binary diffusion coeff. (cm2/s)
20°C 0.088 0.077 0.062 0.245
The soil properties and all initial contaminant distributions were assumed to
be uniform. The initial oil phase composition was specified with mole
fractions for each representative contaminant. Based upon this composition,
equilibrium partitioning relationships were used to evaluate the initial con-
taminant gas, water, and solid phase concentrations.
The flow and transport domains were discretized with an irregular mesh.
Near the extraction well a highly refined grid is required to capture the steep
pressure gradients and large radial velocities in the gas phase. The pressure
gradients, however, diminish rapidly from the well where larger spacings were
used to reduce computational effort. The grid points were determined with a
geometric progression such that:
Ari+ j ,,, tart1 '/~N~- ')
Ari - LAriJ (40)
where Aq and Arr are the initial and final block sizes and Ng is the number of
grid points (Aziz and Settari, 1979). Grid sensitivities were performed to
insure solutions were sufficiently independent of mesh spacing and the radius
of influence.
In the first test problem, a benzene, toluene, and xylene (BTX) mixture was
vented from a sandy soil with a constant extraction rate of 100's cfm with
respect to standard atmospheric conditions. Table 5 lists parameter values used
in this simulation. To observe effects from solid and water-phase partitioning,
282 r. RATHFELDERETAL
TABLE 5
Parameters used in one-dimensional radial and two-dimensional radial vertical simulations
1D Radial 2D Radial-Vertical
Configuration
well radius (rwell) 50era
radius of soil contamination (re) 10m
radius of influence (r~) 300 m
number of mesh points 75
Arj 2.5 cm
Arc 29 m
Field parameters
k (darcy) 10 Zone 1-3 10,
Zone 4 0.01
n (%) 40 Zone 1-3 41,
Zone 4 42
p~ (g/cm 3) 1.75 1.75
at (cm) 10 10
At (%) 0.05 0.05
Q (scfm) 100 100
T (°C) 20 20
Initial saturations
Sw (%) 15 15
So (%) l 1
Initial oil phase mole fractions (%)
benzene 50% 50%
toluene 25% 25%
xylene 25% 25%
the predicted contaminant mass removed from each phase during venting is
plotted against time in Figure 5. Predictions show that 27 days of continuous
venting are required to remove all contaminants from the 20-m diameter
contamination zone, and consistent with results observed in column venting
experiments, early operations produce the greatest total contaminant mass
removal rates, diminishing as the oil phase becomes enriched with the less
volatile constituents. Contaminant removal from the water and solid phases
also aligns with this sequential pattern under the equilibrium partitioning
conditions. Benzene comprises the greatest initial contaminant mass fraction in
the water phase and being the most volatile constituent is removed the fastest.
Consequently, removal of the benzene constituent early in the extraction
operations is accompanied by removal of the largest proportion of the water
contaminant mass. Conversely, the more hydrophobic components, toluene
and xylene, comprise the greatest proportion of absorbed contaminant mass
and due to their lower volatility are removed in the later stages of the venting
100
~ 8O
g
80
r
4o
total
O~
0 5 10 15 2O 25
time (days)
Fig. 5. Contaminant mass removal rates from one-dimensional radial simulations.
operations. Greatest contaminant desorption rates are therefore predicted

during tail end operations. The removal of the oil phase parallels closely with
the total mass removal rate, because contaminant mass in the water and solid
phases comprises a small fraction of the total initial mass, and is removed
quickly under equilibrium conditions.
Design and operation of soil vapor extraction systems is aided by
knowledge of the contaminant removal process. Plotting the predicted con-
taminant distribution at varying time intervals, Figure 6 shows the pro-
gression of contaminant stripping. For the one-dimensional configuration, gas
flow is in the horizontal plane, radially inward. Air absent of contaminant
mass (clean air) contacts the contaminated soil only at the outside radius, and
because clean air has the greatest contaminant assimilation capacity, the bulk
of contaminant volatilization occurs along this outside fringe. Consequently,
6000-
u
5ooo-
. ~ 0.0 days
2.7 days
4000" 4.7 days
N 6.7 days
• 10.7 days
3000" 15.7 days
& 20.7 days
! 24.7 days
2000"
o
1000"
0
0 2 4 6 8 10 12
radius (m)
Fig. 6. Contaminant soil distribution during one-dimensional radial simulations.

• rcap = 10 m
3m
t zone 1
3.25.~
6.25 m
rwell = 50 cm zone 4 1,75 m

rc= lOm
t-
r~=40 m
Fig. 7. Physical domain for two-dimensional radial-vertical problems.
in Figure 6 the contaminants are removed from the outside inward toward the
extraction well. This radially inward progression, however, appears as a
disperse front in the multicomponent case due to unequal component volatil-
ization rates. Close to the extraction well, the increasing gas flux together
with the stipulated equilibrium condition cause additional contaminant
volatilization.
The plots in Figure 6 illustrate flow of clean air through the contaminated
soil is a principal factor controlling the rate of contaminant removal. In field
settings, gas flow to the extraction well is three-dimensional, determined by
field and system variables. To investigate the effects of vertical gas flow, the
equilibrium model was extended to two dimensions solving the equations in
a radial-vertical coordinate system. The physical domain, shown in Figure 7,
describes a three-dimensional system with flow to a single extraction well,
assuming axial symmetry. The problem scenario is an idealized representation
of an underground tank leak resulting in a uniform zone of contaminated soil
spread over a clayey stratum. To remediate the site an extraction well is
positioned in the center of the cylindrical contamination zone and screened
over the thickness of the contaminated soil. An impermeable cap is placed on
the ground surface.
The boundary conditions in the radial direction were identical to the
one-dimensional problem. Atmospheric conditions were specified at ro~, and
second type conditions were specified at rwe~; within the screened well casing
pressure is evaluated from the known flux using equation (38) and along the
blank well casing and below the well, no flow conditions were used. In the
vertical direction no flow conditions were specified along the bottom
boundary and along the impermeable cap at the ground surface. Atmospheric
conditions were used beyond the radius of the impermeable cap. The initial
soil and chemical properties were uniform within respective zones and
identical to those used in the one-dimensional radial problem. Parameter
values are listed in Table 5.
Figure 8 shows the predicted steady-state gas velocity field through the
M A T H E M A T I C A L S I M U L A T I O N O F SOIL V A P O R E X T R A C T I O N SYSTEMS 285
0.0
1.0
w . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.0
mu,,, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iiiiiiiiiiiiiiiiiiiii i i i i i i 3.0
T"
4.0 b--
O-
klJ
C-1
5.0
I
iiiiiiiiiiiiii i
........ J ........ I I 1 " I t
. .
I
. .
I
. .
"
.
I " I I •
6.0
7.0
0.0 2.0 4.0 6.0 B.0 10.0 12.0
ao6 M/DRY RROIUS (Ml

)
MRXIMUM VECTOR
Fig. 8. Predicted steady-state gas velocity field.
contaminated soil zone. The gas velocities in this sandy soil system approach
quasi steady-state quickly, in less than a few hours. The gas velocity increases
toward the extraction well with clean air entering the contamination zone
along the lateral and horizontal boundaries. The progression in contaminant
stripping is shown in soil concentration contours at 2.25, 12.25 and 22.25 days
in Figure 9. Clean air entering the contaminated soil near the well leaves
behind a wedge-shaped contamination zone, while at the radial boundary,
horizontal gas flow strips the contaminants in a fashion similar to the one-
dimensional radial case. Figure 10 compares the contaminant mass removal
rates for the one- and two-dimensional cases. The influx of clean air along the
horizontal boundaries hastens the removal of contaminants in the early stages
of venting operations. During the later stages, vertical influx close to the well
reduces the flow through the remainder of the soil contamination zone. In
comparison with the one-dimensional case a tailing effect is observed. The
time for complete removal is comparable for the one- and two-dimensional
cases.
Table 6 lists results of material balance calculations for simulations of the
one- and two-dimensional hypothetical field problems. In all cases,
cumulative mass balance errors were less than 5%. Grid sensitivity runs also
showed no appreciable effect on the final solutions and mass balance results.
The material balance checks and the reasonably good predictions of the
laboratory column experiments provide confidence that the numerical model
is correctly solving the coupled flow and transport equations.
2,5 ] I I f I f I l ] I
3.5
4.5
5.5
6.5 i I i I ~ r t I I
.0 2,0 4.0 6.0 8.0 10.0 12.0
2.5 ] I ] I t I i I I
3.5
-T-
4.5
-I-
F-
13_ 5.5
LLJ
C3 • -- 100.~
6.5 I t I I I l
0,0 2,0 4.0 6.0 8.0 10.0 12.0
2.5 I I I I f I
3.5
4.5
5.5
6.5 I t I I I I
0.0 2.0 4.0 6.0 0.0 10.0 12.0
RFIDIUS lM}
Fig. 9. Predicted soil concentration distribution at 2.25, 12,25 and 22.25 days• Contour
interval = 1000ppm.
100
80'
o
60
E 4O
= ~]~P ~ water - 2D case
~ ll~/~ ~ solid- 2D case
20 .r~ ~ total - 213 case
o. ~1~ a . total -
total,- 1 D c_se
1D case
0 P
0 " 1'0 20 3'0
time (days)
Fig. 10. C o m p a r i s o n o r c o n t a m i n a n t mass r e m o v a l rates f r o m one- and t w o - d i m e n s i o n a l simulations.

TABLE 6
Cumulative mass balance errors in one- and two-dimensionalsimulations

Problem Component % Error
I-D radial Benzene 4.3
time = 26.0 days Toluene 4.3
Xylene 3.5
Water 4.1
2-D radial Benzene 0.8
time = 25.0 days Toluene 2.3
Xylene 2.5
Water 2.4
SENSITIVITY STUDIES
Equilibrium phase partitioning

Substantiation of the local equilibrium assumption with measured rate
data in systems representative of field-scale venting operations has not been
accomplished. In general, there is a scarcity of rate data in soil systems,
especially with regard to volatilization processes. Consequently, a nonequi-
librium model was used to examine sensitivity of the local equilibrium
assumption in vapor extraction systems.
We examine first the mass transfer rates associated with volatilization of the
immiscible liquid. Johnson et al. (1990) examined the validity of the local
equilibrium assumption with analytical solutions to a one-dimensional
diffusion equation. Assuming that diffusional resistance is the sole mechanism
of impedance, they show that for air flowing through contaminated soil in a
typical venting system, saturated gas concentrations result over a travel
distance of less than 1 cm. This assertion is supported by successful applica-
tion of the local equilibrium assumption in simulating column venting
experiments (e.g., Baehr et al., 1989; Bloes et al., 1991). In convective mass
transfer systems, however, mass transfer rates depend not only on the rate of
diffusion but also on the fluid dynamics of the system. Petrovic and Thodos
(1968) measured mass transfer coefficients for vaporization of water and
heavy hydrocarbons in packed beds of porous spheres (0.0721-0.37 inches in
diameter) with flows of low Rynolds number (specific discharge 0.25-5.2 cm
min-~). Using these data, mass transfer factors were calculated and are
plotted in Figure 11 as a function of specific discharge. Power functions fit to
these results roughly comply with the qualitative rule-of-thumb of Edwards
(1984) which states that 2,o is proportional to the 0.7 power of gas flux.
The one-dimensional radial model was modified to accommodate nonequi-
librium partitioning. In this model the coupled transport equations (9)-(11)
~, = 1365 qO.76 r2 = 0.97
k = 477 qO.83
r2 = 0.80
4" k = 773 qO.SS

r2 = 0.90
qO.59 r 2 = 0 . 9 9
_00 3 ¸
[] d e c a n e (0.0721 inch)
• d e c a n e ( 0 . 1 2 1 5 inch)
z~ o c t a n e ( 0 . 0 8 5 9 inch)
• w a t e r ( 0 . 0 8 5 9 inch)
log q ( m / d a y )
Fig. 11. Calculated mass transfer factors for pure liquid volatilization from experimental data of Petrovic
and Thodos (1968).
were solved using (18) to evaluate interphase mass transfer. The system
configuration and field parameters were identical to the one-dimensional
radial case listed in Table 5.
Effects from nonequilibrium effect in organic liquid volatilization were
examined by setting Sw to zero and ignoring solid-vapor adsorption. The
transfer factor was calculated with the relation, 2ao = Cq °7 suggested by
Edwards (1984) and supported with data from Petrovic and Thodos (1968).
This relation does not directly account for varying interfacial contact area
with time, but presumably this area diminishes slowly as the oil remains
spread as wedge-shaped films between the water phase and the nonwetting air
phase. The constant C was varied by three orders of magnitude from 500 to
0.5, and corresponding transfer factors ranged from 19,300 to 2.5 day -~.
Figure 12 shows the influence of reduced mass transfer rates on the con-
taminant distribution after 50 days of venting pure xylene with an initial
saturation of 1%. As the mass transfer rate is reduced, the gas phase travel
distance over which saturated vapor concentrations are established (i.e. the
region of contaminant stripping) increases. Consequently, the evacuation
front becomes disperse. For C = 500, roughly representative of the corre-
lations developed from the data of Petrovic and Thodos (1968), the contami-
nant distribution compares favorably with the equilibrium results, suggesting
in this case equilibrium concentrations are established relatively quickly.
For C = 0.5 the mass transfer coefficients are reduced to the extreme that
contaminant stripping occurs over the entire path through contaminated soil.
Figure 13 compares the mass removal rates from the pure xylene venting
problem, as well as from the venting of pure benzene, also with an initial
saturation of 1%. The comparisons show diminished mass transfer rates
have little effect on the time for complete contaminant removal so long as
6000]
5000
g 40001
i 3oooJ rium
~ 2000
1000
0.0 2.0 4.0 6.0 8.0 10.0

radius (m)
Fig. 12. Xylene soil concentration for varying air--oil transfer factors after 50 days of continuous venting:
(PI) C = 500; (P2) C = 50; (P3) C = 5; and (P4) C = 0.5.
equilibrium gas concentrations are established prior to reaching the extraction

well. Under the conditions of the one-dimensional radial simulations, this
occurs with mass transfer factors two orders of magnitude smaller than values
obtained from Petrovic and Thodos (1968). These results, collectively with the
observation of equilibrium vapor concentrations in effluent from column
venting experiments, imply that local equilibrium with regard to organic
volatilization is a reasonable assumption in the modeling of field-scale venting
operations.
Field vapor extraction operations ordinarily exhibit tailing in contaminant
effluent vapor concentrations. Tailing effects have been explained, in part, by
reduced mass transfer rates in solid phase desorption and in volatilization
•/benzene i xylene
40 ]lj~' t . ~ ~-- P1 um
~ =/" =,~ • P2
irll, i . . ~
0~ . . . . . . . ,t . . P4. ,
0 40 80 120
time (days)
Fig. 13. Benzene and xylene mass removal rates with varying air-oil transfer factors: (PI) C = 500; (P2)
C = 50; (P3) C = 5; and (P4) C = 0.5.
loo] ~ . ,........
~ . . . . . . . . ,_.,.
801 /7
~ 60
~ 4o
-~ N f " total- P1
~ d/ a~ --"&"- total- P2
20 -~ ~1 L~- ~ total- P3
1.1/ aE" -----o---- wa!er - P3
o g . J ' , , • ,
0 10 20 30 40
time (days)
Fig. 14. Benzene mass removal rates with varying air-water and water-solid transfer factors:
PI = (2,w = 200 day -I , 2ws = 1 0 d a y l ) , p 2 = (2a~ = 20 day ], 2~s = 1 . 0 d a y - I ) , P3 = (2aw = 2.0
day 1,2w~ = 0.1 d a y - I ) .
from the water phase (Hutzler et al., 1988). The nonequilibrium model was
used to examine the rate of contaminant mass evacuation under conditions of
reduced water-phase interphase transfer rates. In this case nonequilibrium
partitioning from the water phase was evaluated with the relations (20)-(21),
and organic phase volatilization was evaluated with equilibrium expressions.
The selected contaminant was pure benzene with an initial saturation of 1%
and remaining parameters are listed in Table 5. Figure 14 compares benzene
mass removal rates for varying mass transfer factors. The plots show for
decreasing air/water and water/solid transfer factors a greater effort is
required to remove contaminants from the water and solid phases. With
2aw = 200 day-1 and 2ws = 10 day-i mass removal rates compared favorably
with the equilibrium model. An order of magnitude reduction in transfer
factors (2,w = 20 day -], 2ws = 1.0 day -~) roughly doubles the time required
for complete removal, and further reduction (2aw = 2.0 day -~, 2ws = 0. I
day -j ) increases evacuation time to well beyond 40 days. These results
illustrate mass transfer from the water and solid phases are potentially rate
limiting processes in venting operations.
Permeability and flow field

Field and laboratory evidence indicates the success of vapor extraction
operations depends principally on the ability to generate and control gas flux
through the contaminated soil. The two-dimensional equilibrium model was
used to examine how the flow field and clean up rate are affected by per-
meability, temperature, and some design alternatives. In these comparisons,
the two-dimensional simulation results described in the section "'Verification"
are considered the base case scenario. The system configuration remains as
MATHEMATICALSIMULATIONOFSOILVAPOREXTRACTIONSYSTEMS 291
TABLE 7
Base case variations in two-dimensional simulations
Problem number Permeability Changes from base case
Zone Value (darcy)
Base case 1-3 l0

4 0.01
2 1-3 l0
4 0.01
3 1 l0
2 l
3 5
4 0.01
4 1-2 100
3 10
4 0.01
5 1-2 100 well screen shortened
3 10 to 1.25 m at depth 5-6.25 m
4 0.01
6 1-3 10 temperature reduced
4 0.01 from 20°C to I O°C
s h o w n in F i g u r e 7 a n d all p a r a m e t e r values are listed in T a b l e 5, except for

the p a r a m e t e r v a r i a t i o n s listed in T a b l e 7. T h e predicted c o n t a m i n a n t mass
e v a c u a t i o n rates for each p r o b l e m v a r i a t i o n are c o m p a r e d in F i g u r e 15.
T h e i m p a c t o f s h o r t circuiting at the g r o u n d surface was studied in p r o b l e m
60 so]u basecase
~]Lt# " prob!em 2
E 404~ • problem 3
•~-
c ~j~r- ----o-- problem 4
~j~ • problem 5
o
0 10 20 30 40 50 60 70 80 90
time (days)
Fig. 15. Total contaminant mass removal rates from two-dimensional simulations.
292 K. R A T H F E L D E R E T AL.
2 (parameter variatons are listed in Table 7). When the ground cover is
removed from the base case setting, pressure gradients between the well head
and the well screen increase, augmenting the flow along this path. Conse-
quently gas flow through the contaminated soil is diminished and a greater
total venting effort is needed to comply with clean up standards. Figure 15
shows that an additional 8 days of continuous venting were needed under the
conditions of this simulation. In general, short circuiting effects at the ground
surface are less important with deeper contamination zones, but can be highly
significant with shallower contamination zones. Therefore, the ground cover
is regarded as a primary design feature that guards against contaminant
leaching, and additionally can improve efficiency of venting operations.
Soil permeability controls the pressure drawndown in the extraction well
and ultimately limits the gas flow through the soil. Plotting pressure
drawndown at a single extraction well as a function of extraction rate and
varying permeability, Baehr et al. (1990) show limiting extraction rates vary
substantially with permeability. Problem 3 examines the influence of soil
meability on pressure drawndown and clean up time. Soil permeability in
the contamination zone was reduced by one-half from the base case and a
thin stratum of less permeable sand bounding the top of the contaminated
soil was added. The resulting maximum pressure drawdown was 0.65atm
compared with 0.8atm in the base case at the same extraction rate. The
steeper pressure gradients and confining effect of the bounding soil stratum
enhanced air flow through the contaminated soil, reducing the total venting
time by 5 days as shown in Figure 15. The improved venting efficiency was
obtained at the expense of increasing the vacuum capacity needed to establish
equal extraction rates. Greater permeability reductions will eventually restrict
extraction rates below 100 cfm, decreasing efficiency, and alternatively for the
base case setting larger system components could be applied to increase
extraction rates above I00 cfm. The numerical model can potentially serve in
economic analyses to determine trade-offs between system sizing and operation
costs.
Heterogenities in soil permeability potentially result in preferential gas flow
paths that by-pass the target contamination zone. In problem 4 the per-
meability above the contaminated soil was reduced by an order of magnitude,
augmenting the flow path around the contamination zone. Figure 15 shows
short circuiting more than doubles the total venting duration to 52 days, as
compared with 22 days for the base case. Short circuiting is evident in total
soil concentration contours plotted in Figure 16. In an effort to diminish the
impacts from short circuiting the well screen in problem 5 was shortened by
approximately one-half. The shorter well screen increased pressure drawdown
in the well, but was not restrictive for this scenario. In distancing the well
screen from the reduced permeability zone, flow through the contamination
2.5 i I i [ I I [ I i
3.5
4.5
5.5
6.5 I i I i I i I i
0.0 2.0 4.0 6.0 B.O 10.0 12.0
2.5 I i I I [ I I i I
3.5
4.5
q-
I.--
O_ 5.5
I..l.J
r"l
6.5 I I I i [ I I I I
0.0 2.0 4.0 6.0 8.0 10.0 12.0
2.5 I 1 I i [ i I i I
3.5
4.5
~@@ .@
5.5
6.5 I I I i I i I I I
0.0 2.0 4.0 6.0 B.O 10.0 12.0
RRDIUS (M}
F i g . 16. P r e d i c t e d soil c o n c e n t r a t i o n distributions in p r o b l e m 5 at 6.75, 2 1 . 2 5 a n d 4 1 . 2 5 d a y s . C o n t o u r

interval = 1000ppm.
zone was increased and consequently the total venting duration was shortened
by 8 days as shown in Figure 15. Collectively these problems demonstrated
that efficiency of soil vapor extraction operations is highly sensitive to the
magnitude and distribution of soil permeability.
Temperature
The contaminant vapor pressure bounds the rate of mass partitioning to the
gas phase and consequently is a limiting factor in venting efficiency. For
typical organic contaminants at common environmental conditions, contami-
nant vapor pressure is a strong function of temperature. Temperature sensi-
tivity was studied in problem 6, reducing the base case temperature condition
was reduced from 20 to 10°C. The corresponding reduction in vapor pressure
substantially lengthened the total venting duration by almost a
factor of four, from 22 days to 86 days. Venting efficiency is clearly sensitive
to system temperature and consequently proper specification is essential

in modeling efforts. The results also suggest that benefits can be derived
from field procedures that can economically elevate contaminant vapor
pressure.
SUMMARY AND CONCLUSIONS
Soil vapor extraction techniques are a popular and potentially efficient

remediation alternative for removal of volatile contaminants from the unsatu-
rated zone. Specific system effectiveness depends upon field, design, and
management variables. Design and operation of soil vapor extraction systems
is often based upon engineering judgement and experience. A numerical
model was developed to aid in investigation of field-scale venting projects. The
model was verified by mass balance checks and comparisons with analytical
solutions and laboratory column experiments. Verification with field venting
data was not undertaken.
Sensitivity studies and laboratory evidence suggests that at the field-scale,
equilibrium vapor concentrations from organic liquid volatilization are estab-
lished in relatively short travel distances. Consequently, the local equilibrium
assumption provides an adequate description of organic liquid volatilization
in modeling field-scale venting operations. Sensitivity studies also show non-
equilibrium desorption and contaminant volatilization from the water phase
can potentially reduce venting efficiency.
Simulations of hypothetical field-scale problems demonstrate efficiency of
venting operations is highly sensitive to the magnitude and distribution of soil
permeability. Soil heterogeneities are characteristic of field settings and,
consequently, design and operation of venting systems may be significant
factors determining clean-up costs. Numerical models can provide the
necessary predictive capabilities in developing planning tools for improved
design and operation of venting systems.
ACKNOWLEDGEMENTS
The research was supported by grant CDR-8-22184 through the Engineer-

ing Research Center for Hazardous Substance Control at the University of
California, Los Angeles. The authors would like to thank the two anonymous
reviewers for their critical but constructive comments and suggestions.
REFERENCES
Agrelot, J.C., Malot, J.J. and Visser, M.J., 1985. Vacuum: Defense system for ground water
VOC contamination. In: Proceedings, Fifth National Symposium and Exposition on
Aquifer Restoration and Ground Water Monitoring. National Water Well Association,
Columbus, OH, pp. 484-494.
Aziz, K. and Settari, A., 1979. Petroleum Reservoir Simulation. Applied Science Publishers,
London, 476 pp.
Baehr, A.L., Hoag, G.E. and Marley, M.C., 1989. Removing volatile contaminants from the
unsaturated zone by inducing advective air-phase transport. J. Contam. Hydrol., 4:
1-26.
Bear, J., 1972. Dynamics of Fluids in Porous Media. Elsevier, Amasterdam-New York, 764 pp.
Bird, R.B., Steward, W.E. and Lightfoot, E.N., 1960. Transport Phenomena. John Wiley and
Sons, New York, NY, 780 pp.
Bloes, M.B., Rathfelder, K.M. and Mackay, D.M., 1991. Laboratory Studies of Vapor
Extraction for Remediation of Contaminated Soil: In: Proceedings of the International
Association of Hydrologists Conference on Surface Contamination by Immiscible Fluids,
Calgary, Alta., April 17-20, 1989 (in press).
Camp, Dresser and McKee, Inc., 1987. Interim report, Assessment of Vacuum Extraction
Technology Application: Bellview, Florida LUST Site. Prepared for: U.S. Environmental
Protection Agency, Edison, N J, Contract No. 68-03-3409, Camp, Dresser, and McKee, Inc.,
Boston, MA, 114 pp.
Conner, J.R., 1988. Case study of soil venting. Pollution Eng., 20(7): 74-78.
Corapcioglu, M.Y. and Baehr, A.L., 1987. A compositional multiphase model for groundwater
contamination by petroleum products 1. Theoretical conditions. Water Resour. Res., 23(1):
191-200.
Corey, A.T., 1986. Mechanics of Immiscible Fluids in Porous Media. Water Resources Publi-
cations, Fort Collins, CO, 255 pp.
Crow, W.L., Anderson, E.R. and Minugh, E., 1985. Subsurface Venting of Hydrocarbon
Vapors from an Underground Aquifer. American Petroleum Institute, Washington, D.C.
Crow, W.L., Anderson, E.R. and Minugh, E., 1987. Subsurface venting of hydrocarbon
emanating from hydrocarbon product on groundwater. Ground Water Monitor. Rev.,
Winter, pp. 51-57.
Downey, D.C. and Elliott, M.G., 1990. Performance of selected in situ soil decontamination
technologies: An Air Force perspective. Environm. Progress, 9(3): 169-173.
Duff, I.S., 1979. MA28 - A Set of Fortran Subroutines for Sparse Unsymmetric Linear
Equations. Report No. AERE-R.8730, AERE, Harwell, 88 pp.
Edwards, W.M., 1984. Mass transfer and gas adsorption. In: R.H. Perry, D.W. Green and J.O.
Maloney (Editors), Perry's Chemical Engineers' Handbook. McGraw-Hill, New York, NY,
40 pp.
Harris, J.C., 1982. Rate of hydrolysis. In: W.J. Lyman, W.F. Rheel and D.H. Rosenblatt
(Editors), Handbook of Chemical Property Estimation Methods: Environmental Behavior
of Organic Compounds. McGraw-Hill, New York, NY, 48 pp.
Hutzler, N.F., Murphy, B.E. and Gireke, J.S., 1988. State of Technology Review Soil Vapor
Extraction Systems. Cooperative Agreement CR-814319-01-1, Hazardous Waste Engineer-
ing Research Laboratory, Environmental Protection Agency, Cincinnati, OH, 36 pp.
Johnson, P.C., Kemblowski, M.W. and Colthart, J.D., 1990. Quantitative analysis for the
cleanup ~f hydrocarbon-contaminated soils by in-situ soil venting. Ground Water, 28(3):
403-412.
Karickhoff, S.W., 1984. Organic pollutant sorption in aquatic systems. J. Hydraul. Eng.,
110(6): 707-735.
Karickhoff, S.W., Brown, D.S. and Scott, T.S., 1979. Sorption of hydrophobic pollutants on
natural sediments. Water Res., 13: 241-248.
Krishnayya, A.V., O'Connor, M.J., Agar, J.G. and King, R.D., 1988. Vapor extraction systems
factors affecting their design and performance. In: Proceedings, Petroleum Hydrocarbons
and Organic Chemicals in Groundwater Prevention, Detection and Restoration. National

Water Well Association, Houston, TX, pp. 547-567.
Lyman, W.J., 1982. Adsorption coefficients for soils and sediments. In: W.J. Lyman, W.F.
Rheel and D.H. Rosenblatt (Editors), Handbook of Chemical Property Estimation
Methods: Environmental Behavior of Organic Compounds. McGraw-Hill, New York, NY,
33 pp.
Mackay, D.M., 1990. Characterization of the distribution of behavior of contaminants in the
subsurface. In: Remediating Groundwater and Soil Contamination Remediation. Report on
a Colloquium sponsored by the Water Science and Technology Board, National Academy
Press, Washington, D.C., pp. 70-90.
Mackay, D.M. and Vogel, T.M., 1985. Groundwater contamination by organic chemicals:
uncertainties in assessing impact. In: Proceeding of the Second Canadian/American
Conference on Hydrology. National Water Well Association, Worthington, OH, pp. 50-
59.
Mackay, D.M., Ball, W.P. and Durant, M.G., 1986. Variability of aquifer sorption properties
in a field experiment on groundwater transport of organic solutes: methods and preliminary
results. J. Contam. Hydrol., l(l): 119-132.
Marley, M.C. and Hoag, G.E., 1984. Induced soil venting for recovery/restoration of gasoline
hydrocarbons in the vadose zone. In: Proceedings, Petroleum Hydrocarbons and Organic
Chemicals in Ground Water- Prevention, Detection and Restoration. National Water Well
Association, Houston, TX, pp. 473-503.
Massmann, J.W., 1989. Applying groundwater flow models in vapor extraction system design.
J. Environ. Eng., 115(1): 129-149.
Millington, R.J., 1959. Gas diffusion in porous media. Science, 130: 100-103.
Munz, C. and Roberts, P.V., 1984. The ratio of gas-phase to liquid-phase mass transfer
coefficients in gas-liquid contacting processes. In: W. Brutsaert and G.H. Kirka (Editors),
Gas Transfer at Water Surfaces. D. Reidel, Dordrecht, pp. 35-45.
Petrovic, L.J. and Thodos, G., 1968. Mass transfer in the flow of gasses through packed beds.
I and EC Fundamentals, 7(2): 274-280.
Rathfelder, K., 1989. Numerical simulation of soil vapor extraction systems. Ph.D. Disser-
tation, Department of Civil Engineering, University of California, Los Angeles, CA.
Regalbuto, D.P., Barrera, J.A. and Lisiecki, J.B., 1988. In-situ removal of VOC's by means of
enhanced volatilization. In: Proceedings, Petroleum Hydrocarbons and Organic Chemicals
in Groundwater Prevention, Detection and Restoration. National Water Well Association,
Houston, TX, pp. 571-590.
Roberts, P.V., Hopkins, G.D., Munz, C. and Riojas, A.H., 1985. Evaluating two-resistance
models for air stripping of volatile organic contaminants in a countercurrent packed
column. Environ. Sci. Technol., 19(2): 164-173.
Schwille, F., 1984. Migration of organic fluids immiscible with water in the unsaturated zone.
In: B. Yaron, G. Dagan and J. Goldsmid (Editors), Pollutants in Porous Media. Springer-
Verlag, Berlin, pp. 27-48.
Stephanotos, B.N., 1988. Modeling the transport of gasoline vapors by an advective-diffusion
unsaturated zone model. In: Proceedings, Petroleum Hydrocarbons and Organic Chemicals
in Ground Water: Prevention, Detection and Restoration. National Water Well Associ-
ation, Houston, TX, pp. 591-611.
Texas Research Institute, 1980. Examination of Venting for Removal of Gasoline Vapors from
Contaminated Soil. American Petroleum Institute, Washington, D.C., 25 pp.
Texas Research Institute, 1984. Forced Venting to Remove Gasoline Vapors from a Large-
Scale Model Aquifer. American Petroleum Institute, Washington, D.C., 60 pp.
M A T H E M A T I C A L S I M U L A T I O N O F SOIL V A P O R E X T R A C T I O N S Y S T E M S 297
Thornton, J.S. and Wootan, W.L., 1982. Venting for the removal of hydrocarbon vapors from
gasoline contaminated soil. J. Environ. Sci. Health, A17(l): 31-44.
Vogel, T.M., Criddle, C.S. and McCarty, P.L., 1987. Transformations of halogenated aliphatic
compounds. Environ. Sci. Technol., 12(8): 722-736.
Welty, J.R., Wicks, C.E. and Wilson, R.W., 1984. Fundamentals of~Momentum, Heat and
Mass Transfer. John Wiley and Sons, New York, NY, 803 pp.
Wilson, D.E., Montgomery, R.E. and Sheller, M.R., 1987. A mathematical model for removing
volatile subsurface hydrocarbons by miscible displacement. Water, Air, Soil Pollution,
33(3-4): 231-255.
Wilson, J.L., Conrad, S.H., Hagan, E., Mason, W.R. and Peplinski, W., 1988. The pore level
spatial distribution and saturation of organic liquids in porous media. In: Proceedings,
Organic Chemicals and Petroleum Hydrocarbons in Groundwater, National Water Well
Association, Houston, TX, pp. 107-132.

RathFelder 1991

Uploaded by

Copyright:

Available Formats

RathFelder 1991

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

RathFelder 1991

Uploaded by

Copyright:

Available Formats

Journal of Contaminant Hydrology, 8 (1991) 263-297 263

Elsevier Science Publishers B.V., Amsterdam

Mathematical simulation of soil vapor extraction

Klaus Rathfelder", William W-G. Yeh a and Douglas Mackay b

Nonaqueous Phase Liquids (NAPL'S)immobilized as residual saturation in

0169-7722/91/$03.50 © 1991 Elsevier Science Publishers B.V. All rights reserved.

Typical Desian Criteria

Fig. I. Typical soil vapor extraction system.

extraction, under suitable circumstances, is a superior remediation method

two-dimensional finite-difference model for transport of a single volatile

Organic contaminants released to the subsurface can potentially partition

mass balance equation is expressed for a representative y-component in a

where t is time [T]; n is the dimensionless soil porosity; S~ is the dimensionless

Soil vapor extraction is assumed to be implemented after discovery

phase pressures are assumed to be predominantly controlled by applied

nSa Vm~ -- kkra V P (2)

where aL and aT are longitudinal and transverse aquifer dispersivities [L],

where Da~ is the unobstructed vapor-phase molecular diffusion coefficient and

Internal sources and sinks

Organic contaminants can undergo chemical or biological transformations

volatile compounds (Mackay and Vogel, 1985). The conservative assumption

Interphase mass transfer

In summary, the macroscopic mass transport equations are expressed as:

0t (PsPsT) = If~ (ll)

where I~ represents mass transport of the V-component across interfaces

-~[nSap~, + nSoPo~ + nSwPw~ + PsPsr] = V" P~r VP + nSaDha~,VP.r

/Qaw -- Pa), __ M/K~h

where/~ is the 7-component Henry's Law constant at the system temperature.

K~s'w = Kocfoc - Ps~ (15)

Ot [nSaPa~,+ nSoI~oPa7+ nSwK~wPa.~+

2ao[Pa~ -- K~aoPoe] (18)

Isw = nSw2ws[Pw - Ps~/K~w] (21)

a--t [nSaPa] = -V~nSaflaVma + IaB+ nSaFa (22)

where M a is the gas phase molecular weight.

Totality conditions provide additional expressions necessary to close the

and for water:

NUMERICAL SOLUTION SCHEME

balance error is calculated for each component using:

error/ ,00f, i=|

The numerical simulation model was verified by comparisons with

Parameters used in comparisons with analytical expressions

Mackay et al. (1990) conducted laboratory experiments to investigate

Parameters used in simulations of laboratory column experiments

Vapor pressure (atm)

Initial mole fraction

relative permeability (kra) -~ 1

Remaining parameters used in these simulations are given in Table 2.

Cumulative mass balance errors in the simulations of laboratory experiments

L a b o r a t o r y c o l u m n experiment - Benzene 2.1

L a b o r a t o r y c o l u m n experiment - Benzene 2.2

information about the rate at which equilibrium concentrations were esta-

Q _ Prwe,,kkra OP[ (38)

Chemical properties of test contaminants

Benzene Toluene o-Xylene Water Dry air

Molecular weight (g/mole) 78.1 92.1 106.2 18.0 28.98

Parameters used in one-dimensional radial and two-dimensional radial vertical simulations

Fig. 5. Contaminant mass removal rates from one-dimensional radial simulations.

operations. Greatest contaminant desorption rates are therefore predicted

Fig. 6. Contaminant soil distribution during one-dimensional radial simulations.

rwell = 50 cm zone 4 1,75 m