A Screening Approach to Simulation of Aquifer Contamination by
Fuel Hydrocarbons (BTEX and MTBE)
James W. Weaver
Ecosystems Research Division
National Exposure Research Laboratory
United States Environmental Protection Agency
Athens, Georgia
Randall J. Charbeneau
Center for Research in Water Resources
The University of Texas at Austin
Austin, Texas
Abstract
Subsurface contamination by light nonaqueous phase liquids (LNAPLs) is a
common occurrence as evidenced by more than 397,000 confirmed releases from
underground storage tanks across the United States (USEPA, 2000). Because of
generally limited resources, common biodegradation of contaminants, and programmatic
policies, there is an emphasis on risk-based corrective action for these releases. This
approach implies a predictive modeling capability. This chapter describes data from a set
of LNAPL cases studies, drawn from underground storage tank program files from state
environmental agencies and the U.S. Department of Defense. These illustrate data
availability under realistic conditions. Against this background, a simplified model for
exposure assessment is described. This model is called the Hydrocarbon Spill Screening
Model (HSSM). The mathematical basis of the model is given, and the underlying
assumptions are discussed. Application of the model to a field site is described. This
case has extensive data set that was analyzed to generate input parameter values for the
model. The approach included an estimate of mass of contaminants, the location of center
of mass, and the gasoline volume. By treating the model inputs as fitting parameters,
order-of-magnitude matches to these data sets were achieved. The model provides a
means of completing the conceptualization of each site by providing a plausible source
and transport scenario, which may not be directly observed from site data.
Introduction
Fuels cause contamination in the subsurface by their presence as a separate
phase and through contamination of soil, subsurface air and water. Aquifer contamination
reflects dissolution of contaminants (benzene, toluene, ethylbenzene and xylenes-BTEX,
-i-
-------
or methyl tert-butyl ether-MTBE) from the fuel, transport in the aquifer and loss
mechanisms. Each of these represents a component of mass balance for the
contaminants. Mass entering a contaminant plume originates in the fuel (LNAPL) phase,
which may be mobile depending on the nature of the release. Along with the hydrologic
processes, this phase controls the rate of release of mass to the aquifer. Once in the
aquifer, the contaminants are transported by flowing ground water and are subject to
sorption. The apparent dilution observed in monitor wells is often characterized by the
aquifer dispersivities. Reduction in concentration can also be due to biodegradation
which has been established to occur very commonly for BTEX. Thus observed
contaminant plumes reflect each of these three generalized processes: dissolution from
the LNAPL source, transport in the aquifer and degradation or other loss mechanisms.
The model described below provides a means for estimating concentrations of fuel
hydrocarbon constituents at downgradient receptors in aquifers based on this scenario.
The Hydrocarbon Spill Screening Model (HSSM)
The Hydrocarbon Spill Screening Model (HSSM) was intended as a simplified
model for estimating the impacts of petroleum hydrocarbons on subsurface water
resources (Weaver et al., 1994b; Charbeneau et al., 1995). The model includes the major
elements described above: presence and motion of an LNAPL, dissolution of
contaminants from LNAPL, transport in the aquifer and degradation. Figure 1 shows the
release of LNAPL from near the ground surface through a mildly heterogeneous vadose
zone. The path followed by the LNAPL is shown to be determined by the distribution of
heterogeneities. Figure 2 shows, in contrast, the scenario used in HSSM. The primary
differences between the two figures are that the vadose zone and aquifer are assumed to
be uniform, and preferential spreading of the LNAPL in the direction of the water table
gradient is ignored in Figure 2 and HSSM. The model focuses on downgradient receptor
concentrations, rather than the details of the LNAPL distribution in the source. The
simplifications were made based upon data availability and an interest on downgradient
receptors. In many cases, aquifer contamination occurs after the LNAPL has reached the
water table, and there are little or no data on the distribution history of LNAPL in the
vadose zone. Vadose zone transport was included in HSSM, however, to assure
completeness, to allow assessment of arrival times at the water table, and to include the
effect of transient LNAPL flux to the water table. As implied by Figure 2, HSSM uses
assumptions of
homogeneous subsurface properties
one dimensional flow in the vadose zone
radial spreading of the LNAPL in the capillary fringe.
Transport in the aquifer, as noted below is assumed to be two-dimensional in the plane,
-2-
-------
but only a fraction of the aquifer thickness is contaminated.
In the HSSM scenario, the LNAPL flows downward through the vadose zone and
forms a lens in the capillary fringe. The LNAPL is assumed to be composed of two
components; one is the contaminant of interest, and the other is a slightly soluble oil
(LNAPL). The properties of the LNAPL phase (density, viscosity, LNAPL/water partition
coefficient) remain constant throughout the simulation. The contaminant (usually a BTEX
compound or MTBE) can dissolve from the LNAPL into the flowing ground water and
diffuse aquifer recharge that is assumed to flow through the lens. At receptor locations in
the aquifer, contaminant concentrations follow breakthrough curves as illustrated in Figure
3. The shape of these curves is determined by advective-dispersive transport in the
aquifer and by the history of mass released to the aquifer. The asymmetry evident in
Figure 3 is caused by the mass flux to the aquifer increasing with lens radius (see
equations 10 and 11 below) as the lens is formed. Later, a slow decline in mass flux
occurs as the constituent is gradually leached from the lens. This results in the tailing
shown in Figure 3. A symmetric input would produce a symmetric breakthrough curve at a
receptor under the assumptions of linear equilibrium partitioning as are used in the HSSM
aquifer module.
The HSSM model consists of three modules that treat transport in the vadose zone,
formation and decay of an oil lens in the capillary fringe, and transport of soluble
constituents of the LNAPL in the aquifer to receptor locations. The modules are the
Kinematic Oily Pollutant Transport Module for the vadose zone; the OILENS module for
LNAPL lens motion and dissolution of constituents, and the Transient Source Gaussian
Plume Module for aquifer transport (Figure 2). The model uses semi-analytical solutions
of the transport equations so much of the otherwise required numerical evaluation is
avoided. This Section contains a review of the theoretical background of HSSM and is
based upon the material presented by Weaver et al. (1994a) and Charbeneau et al.
(1995). The background documentation, along with the model and example data files, can
be downloaded from http://www.epa.gov/athens/hssm1.htm.
The Kinematic Oily Pollutant Transport Module
The Kinematic Oily Pollutant Transport (KOPT) module was derived from the phase
conservation equation for an LNAPL in the presence of a fixed amount of water and air in
the pore space. The amount of water is determined from the diffuse recharge rate and the
amount of air is estimated from an observation that the water phase conductivity is only
about 50% of its maximum value during infiltration (Bouwer, 1966). Depending on the
boundary condition, LNAPL flow can be driven by gravity and pressure during the release.
After the end of the release, flow is assumed to be driven by gravity only. The resulting
conservation equation for the LNAPL phase is
-3-
-------
as
dS0 di
where n is the porosity, Ke0(S0, SW(avg)) is the effective conductivity to the LNAPL, which is
a function of S0, the LNAPL saturation, and SW(avg), the recharge-determined water
saturation, z is the depth below the surface, and t is the time. Saturation is defined as the
fraction of the pore space occupied by a fluid. Equation 1 is a first order hyperbolic
equation that has the method-of-characteristics solution
Equation 2 is called the classical method-of-characteristics solution of equation 1.
Because the effective conductivity function is nonlinear, it must be supplemented by a
generalized, or shock, solution which is given by
where qi and q2 are the LNAPL fluxes on either side of the leading edge of the invading
LNAPL (see Figure 4, left), and S0i and S02 are the corresponding LNAPL saturations.
Equations 2 and 3 are implemented in the KOPT module. During a release under ponded
conditions, the fluxes in equation 3 are determined by the Green-Ampt Model (Green and
Ampt, 1911) to include gravity and pressure effects. For releases that occur at rates
below the effective conductivity of the soil, and for times after the end of the release, the
fluxes are equal to the effective conductivity (Weaver et al., 1994a).
Figures 4 and 5 illustrate schematically the solution obtained from the KOPT model. While
the LNAPL is infiltrating, the leading edge of the LNAPL is represented as a sharp front
(Figure 4, left). Spreading associated with capillary gradients would tend to produce a
smooth front is neglected in equation 1. At each time, the position of the sharp front
(Figure 5) is given by the solution of equation 3. After the end of the release, the
2
—- = 0 along
dt
dz _ 1
dt ~ n f*sD
rfz _ 1 " 9a
dt ' n s% - s,3
-4-
-------
redistribution of the LNAPL is governed by gravity (equations 2 and 3). The resulting
distribution of LNAPL behind the front is smooth (Figure 4, right) and there is a gradual
decrease in saturation from the front to the ground surface. Because the LNAPL
saturation is reduced over time at the front, the speed given by equation 3 is also
gradually reduced. Figure 5 shows the gradual slowing of the front as time goes on.
The dissolved constituent of the LNAPL (i.e., constituent benzene of LNAPL
gasoline) is simulated by the solution of a mass conservation equation. Here dispersion is
neglected so the equation becomes a first order hyperbolic equation which is also solved
by a method-of-characteristics approach. The conservation equation is
^kff\ dcw
rjir T ft-*4
where ko is the equilibrium linear partition coefficient between the water and LNAPL
phases (k0 = c0 / Cw), Pb is the bulk density, Cs is the soil phase concentration, kd is the
equilibrium linear partition coefficient between the soil and water phases (kd = Cs / Cw), and
q0 and qw are the LNAPL and water fluxes, respectively. The method-of-characteristics
solution is
dz
ig — = -
dt 5
which is implemented in KOPT. Since the conservation equation 4 is linear, no shock
solution analogous to equation 3 is needed.
The OILENS Module
The OILENS model simulates the flow of LNAPL and its constituent in a lens at the
water table. The distributions of water, LNAPL and air are idealized in accordance with
the theory described in Weaver et al. (1994b), which gives an equivalent uniform LNAPL
saturation in the lens (S0(max))- By following this procedure the LNAPL saturation is
-5-
-------
averaged over the capillary fringe and the averaged saturation is used in the model as an
idealized constant LNAPL saturation applied over an equivalent thickness of the lens. This
usage eliminates vertical variation in saturation in the lens from the simulation. Building
from this assumption, a mass conservation equation can be written for the LNAPL lens.
Two conservation equations are used: one for a cylinder that is located directly below the
LNAPL source, and another for the entire lens (Figure 6). The equation for the cylinder
gives
dh„
—77- = Qitdpt *
at
- a
where Rs is the radius of the LNAPL source, S0(max) is the LNAPL saturation in the lens, (3,
the buoyance factor, is defined as (3 = pw/(pw - Po), hos is the LNAPL head at the source,
Qkopt is the volumetric inflow to the lens, Qradjai is the volumetric outflow from the central
cylinder, and Q|0SS is the sum of the volumetric losses due to dissolution and LNAPL phase
trapping in the saturated and vadose zones. The lens height at any radius is determined
by the Dupuit assumption where the head is constant along vertical sections (Bear, 1972).
The continuity equation for the lens volume, VL, is
dVL
" = afC
where Qout is the loss of LNAPL from dissolution and trapping at residual saturation, and Rt
is the lens radius. Trapping of LNAPL occurs as the lens collapses after the influx to the
lens stops (illustrated by the cross-hatched area on Figure 7). The lens volume is
determined from
2
n p
TT^
4
-6-
-------
The basic KOPT and OILENS equations (2, 3, 5, 6, and 7) form a system of coupled
ordinary differential equations. These are solved numerically by a Runge-Kutta technique
with automatic time-stepping control (Felhberg, 1969). The key to the efficient solution of
the lens equations (6 and 7) is the analytical expression for the lens volume, VL, given in
equation 8.
The KOPT model generates both the LNAPL flux to the lens and the constituent
concentration in the LNAPL as functions of time. Dissolution of the constituent into the
aquifer is assumed to be caused by contamination of recharge water moving through the
lens, and by contact with ground water flowing beneath the lens. The mass flux from
recharge, minfii is estimated as
where qwi is the recharge rate, and Cwo is the equilibrium water phase concentration of the
constituent. Cwo is calculated by assuming that Raoult's law applies to the partitioning of
hydrocarbons from gasoline. Cline et al, 1991, show laboratory data on partitioning from
31 gasoline samples that indicate the Raoult's law assumptions holds. Flow in the aquifer
contributes to mass flux, because of the vertical dispersion of contaminants from the
LNAPL lens to the flowing ground water. By solving the an equation of time-dependent
vertical dispersion as flow goes under the lens, and integrating over the area of the lens,
the following expression for mass flux, mdiSs, was developed (Charbeneau et al, 1995; for
similar approaches, Chrysikopoulos, 1995 and Hunt, 1988)
where v is the seepage velocity, and av is the vertical dispersivity of the aquifer. The
integral in equation 10 is approximately equal to 0.87402.
The mass flux to the aquifer is given by the sum of equations 9 and 10. This
quantity varies with time because it depends upon the radius of the LNAPL lens and upon
the amount of the dissolved constituent in the LNAPL. Mass flux increases with radius
(RT) as the lens expands and decreases with declining constituent concentration (c^) as
the mass is depleted from the lens.
In OILENS, the LNAPL lens is assumed to be circular with no elongation in the
direction of ground water flow. Generally, downgradient migration of the fuel is limited by
rtc '
-i-
-------
1) entrapment of residual LNAPL and 2) reduction in effective conductivity to the NAPL
because of the relatively low LNAPL saturations achieved in lenses. Typical average
LNAPL saturations in lenses are on the order of 0.2 to 0.4, which correspond to relatively
low fluxes. For the Hagerman Ave example given below, the flow rate of the LNAPL is 41
times lower than that of the water given the hydraulic conductivity, gradient, and the
LNAPL's average saturation, density and viscosity. Likewise the gradient from radial flow
in the lens is four times greater than the water table gradient. This shows that the flow is
driven radially with a smaller component in the direction of the water table gradient.
Exhaustion of the free LNAPL by trapping limits the possible downgradient motion of the
lens. As shown in the example, the model applies to cases without significant
downgradient spreading of the LNAPL lens. This condition has been found to be the case
for many releases.
The Transient Source Gaussian Plume (TSGPLUME) Module
With the LNAPL lens located in the capillary fringe, the source of contamination
remains near the top of the aquifer (Figure 2). The TSGPLUME model reflects this
behavior by assuming that the contaminants only are present over a certain thickness of
the aquifer, called the penetration thickness. That thickness is determined from the size of
the lens, the recharge rate, the ground water velocity, and the vertical dispersivity
(Charbeneau et al., 1995). In HSSM, the contaminant in the aquifer is averaged over the
penetration thickness and concentrations vary in two dimensions-longitudinally and
transversely in the horizontal plane.
Two-dimensional solute transport with first order decay obeys
where Rd is the retardation coefficient, c is the concentration, t is time, DL and DT are the
longitudinal and transverse dispersion coefficients, respectively, x is longitudinal distance,
X| is the coordinate of the downgradient edge of the LNAPL lens, y is the distance
transverse to the plume centerline in the horizontal plane, v is the seepage velocity, and K
is the first order decay constant. The boundary conditions applied define the gaussian
source
-8-
-------
;(x-x/y,0) = 0
c[x,yj) = cDr
x-x„-»t) = c[:
where o is the standard deviation of the contaminant distribution transverse to the plume,
and c0 is the peak concentration. The gaussian boundary condition is used to distribute
the mass flux across the width of the LNAPL lens, reflecting variability in the width and
strength of the source. The width, w, of the LNAPL lens is incorporated into the boundary
condition (equation 12) by assuming that w is equal to four times the standard deviation of
the gaussian boundary condition.
When nondimensionalized, equations 11 and 12 become
d2C r\d2C
r + D^-=- +i3
ax2 a y2
and
,Y,T) = exp(- 14
with the nondimensional variables defined by
-9-
-------
X =
Y =
7 =
A =
D =
C =
V(X-X/)
Dl
V
a
v*t
RDl
RKDl
DJh-
oV
c_
15
Application of Fourier and LaPlace transform techniques gives the solution for the time-
invariant boundary condition as (Smith and Charbeneau, 1990)
. At 2 + 4ti6
l/4nf3(1
As noted above for OILENS, the mass flux to the aquifer is allowed to be time dependent,
so that the boundary condition is incorporated in TSGPLUME by using Duhamel's
Principle (Carslaw and Jaeger, 1959):
tTB(T- 14)^,7
J D
where B(T) is the time-dependent mass flux from equations 9 and 10.
Although equation 17 accounts for time dependency in mass flux, the size of the
source also varies with time as the LNAPL lens expands. In some situations, namely a
relatively high viscosity LNAPL or high hydraulic conductivity formation, the flow of the
LNAPL occurs mostly during the initial part of the event. The LNAPL effectively reaches its
maximal extent fairly early in the event. Under this scenario, the maximum LNAPL lens
-10-
-------
size might be picked for the size of the aquifer source. In many cases, however, the flow of
the LNAPL does not necessarily cease quickly and the LNAPL lens may continue to
expand over a long period of time. Further, flow of the LNAPL clearly continues for a long
time when the leak is assumed to occur over a long time period. To estimate the peak
concentration, a rule used in HSSM is that the lens size picked for the TSGPLUME
boundary condition is the lens size that occurs when the mass flux to the aquifer is also a
maximum. This ensures that the peak source flux enters the aquifer through an
appropriately sized boundary condition.
Data Sets
Twenty four case studies of petroleum hydrocarbon contamination were collected
from various state underground storage tank programs, private industry and the U.S.
Department of Defense. The releases occurred in 12 States and the District of Columbia.
The data sets represent a range of site and release conditions. The purpose of the review
was to determine what data are collected at typical sites and what model input parameters
must be estimated. Table 1 a lists the reasons given for the site investigations. Most were
associated with tank removals and only four were investigated because of ground water
contamination or vapor accumulation. Reflecting diversity in responsible party finances,
state program requirements, and the expertise of the investigators, each data set was
unique and the amount of information varied significantly. These cases were, however,
taken from relatively small scale releases and the data reflect this. Large scale releases of
petroleum hydrocarbons as might occur at refineries, fuel depots, truck terminals, etc., are
not represented in these cases. Thus the data are biased toward the small sites, which
have relatively modest investments in data collection. Tables 1b and 2 list general
categories of input (Table 1 b) and output (Table 2) parameters for the HSSM, along with
the number of sites with at least a single measured value. Although the parameters
represent specific input required for HSSM, they also represent the parameters needed for
other multiphase, multicomponent models. The tables show clearly that all necessary
parameters were not measured at these sites. Although not indicated explicitly in the
tables, varying numbers of measurements were made for each parameter, so the degree
of spatial and temporal variability characterized at these sites also varied significantly.
The mass of petroleum product released and duration of the release are important
input parameters because they define the boundary condition for the model. Major
uncertainties in site evaluation are introduced by not having this fundamental information.
In the two cases where there was a single known release, one was a catastrophic tank
failure and the other was a plane crash. Even in these, the volumes are not precisely
known. For the plane crash, some of the fuel burned so the amount released can only be
bounded by an upper limit. For the tank failure, only an estimated fuel volume was
available. In five other cases there were known releases, but other uncharacterized
releases also occurred, so the total volume of the release and its timings could not be
determined. In contrast to the few cases of known release, the more typical situation is
-ii-
-------
where undetected releases occurred that were discovered during some latter event, such
as a tank upgrade, property transfer or contamination of a well (Table 1 a). Because this is
the nature of most releases, firm bounding of the release date(s) may only be possible by
using the beginning and ending dates of active life of the facility. These were reported for
nine and eleven of the sites, respectively. As discussed below for the Hagerman Avenue
site, MTBE contamination may be used roughly to date a release. MTBE usage began in
1979 (USEPA, 1998), so MTBE plumes must originate after that date. BTEX
contamination at these sites, however, could originate from earlier releases.
Another component of the boundary condition is the area over which the release
occurred. For tanks, it may be reasonable to approximate the area by the backfilled area
of the tank pit. Releases from leaking tanks or piping systems may distribute through this
region before entering the vadose zone. Leaks originating in piping or overfills outside the
tank pit may be assumed to occur over a smaller area.
In addition to the fuel volume, the mass of each chemical released to the subsurface
depends on the composition of the fuel. Fuel composition varies with the product, crude oil
source, refiner, season of the year, geography and by regulatory requirement (e.g., Alberta
Research Council, 1994, Gustafson et al., 1997, and Neff etal., 1994). When ground
water contamination is detected many years after the release, it is obviously not possible
to determine the composition of the original fuel. Hydraulic properties of the fuel—density,
viscosity, and surface tension also depend on its type or composition. For most fuels,
however, there are typical and literature values available for these parameters (Gustafson
et al. 1997 and Neff et al., 1994), but the lack of site specific measurements introduces a
moderate degree of uncertainty in any simulation results. In one of the cases in Table 1 a,
general classes of compounds (i.e., paraffins, iso-paraffins, naphthenes, aromatics, and
olefins) were identified for fingerprinting the fuel source. In most other cases the fuel type
was taken as the product stored in the tanks. Diesel and gasoline each may have been
released in several cases, without being differentiated by the investigators.
In none of the cases were data collected on the vadose zone conductivity or the
moisture retention (capillary pressure) properties. This is presumably reflective of the
emphasis on ground water contamination occurring after the fuel has flowed through the
vadose zone, lack of application of vadose zone models, and lack of apparent need for
inclusion of vadose zone processes in site evaluation. In contrast, the aquifers were more
highly evaluated. In each case, information on the geologic structure was provided, either
through studies of regional geology, boring logs or both. In each case the depth to water
was determined in the course of site evaluation. The porosity and aquifer dispersivities
were not measured at any site. Hydraulic conductivity was measured at 16 sites and
estimated from the literature at five more. In some cases conductivities were measured at
multiple locations in the aquifer to show spatial variability. The organic carbon content was
determined in five cases, sometimes at different locations.
The hydrology of the sites is represented in three parameters of HSSM. First is the
-12-
-------
recharge rate, which was not measured at any site. Annual precipitation was reported for
four cases and used to assume the amount of recharge. Aquifer thicknesses were
determined from boring logs in two cases and estimated from geologic literature in six
others. In all cases the ground water gradient was determined in conjunction with
measured water levels in observation wells.
LNAPL parameters-the LNAPL/water partition coefficient and the LNAPL hydraulic
properties-were not measured at the sites. Neither were the contaminant partition
coefficients. In one case a half life was estimated for the a dissolved contaminant and in
five cases the distribution of electron acceptors and metabolic byproducts were
determined. The latter provide evidence for biodegradation of the contaminants, although
they do not give a half life for use in equation 11.
Table 1 a Reasons Given for Detection of Contamination or
Site Investigation
Reason
Number of
Cases
Tank Removal
9
Observed Surface Releases
3
Property Transfer
3
Subsurface Fuel Vapors
2
Contaminated Ground Water
2
Investigation of Nearby Release
1
Inventory Reconciliation
1
Excavation for Construction
1
Unstated
2
-13-
-------
Table 1 b General categories of HSSM input parameters
Item
Parameterization
Number of
Cases with
Site-specific
Measurement
Contaminant Source
Source Mass (Duration and Volume)
2
Free Product Recovery Volumes
9
Tank Installation/Beginning of Operations
9
Tank Removal/Ending of Operations
11
Fuel Composition
0
Fuel Properties (density, viscosity, surface
tension)
0
Vadose Zone
Hydraulic Conductivity
0
Moisture Retention Curve Parameters
0
Aquifer
Geologic Cross Section or Description
24
Depth to Ground Water
24
Porosity
0
Hydraulic Conductivity
16
Dispersivities
0
Orqanic Carbon Content
6
Hydrologic
Recharqe Rate
0
Aquifer Thickness
2
Ground Water Gradient
24
Contaminant
NAPL/water Partition Coefficient
0
Hydraulic Properties (relative permeability and
capillary pressure curves)
0
Soil/water Distribution Coefficient
0
Half Life
0
Electron Acceptor/Metabolic Byproduct
Concentrations
5
-14-
-------
Table 2 General categories of HSSM output
Item
Parameterization
Number of Cases
with Site-Specific
Measurement
Vadose Zone
Time Dependent LNAPL Saturation
0
NAPL Saturation: Soil Cores
22
Free Product Levels in Wells
14
Time-Dependent Lens Radius
0
Time-Dependent Lens Thickness
0
Saturated Zone
Mass Flux to Aquifer
0
Receptor Concentrations
24
Not unexpectedly, an emphasis in data collection was on water and soil samples
(Table 2). These are the standard analyses that are widely used for characterizing
subsurface contamination. Soil core data were collected from 22 sites. Characterization
of contamination from soil samples could under certain circumstances be used to
characterize the LNAPL phase. Other information that would characterize the time-
dependent distribution of the LNAPL phase were lacking (lens radius and thickness). Free
product was observed at 14 sites. These data could be used to establish roughly the foot
print of the LNAPL contaminated zone, but are not likely to provide detailed information
needed for evaluating the model predicted LNAPL distribution in very many cases. All sites
had water sample data. As for the soil core data, the spatial and temporal density of the
samples varied greatly. At one site, one monitoring well was installed in the first phase of
investigation. When free product was detected in that well, eight others were installed.
For a few sites, multiple sample rounds generated a portion of the breakthrough curves at
the receptor wells. For the most part, these curves were not complete because the
releases generally occurred years before any site investigation was undertaken.
Although the data are limited with regard to simulation models as described above,
these same data were judged acceptable for the purposes of the respective State or
Federal programs. These purposes include listing of a site, assessing contaminant
impacts and developing corrective action plans. These data sets, however, do not contain
all of the inputs required for even a simple solute transport model (note the lack of
measured dispersivities, porosities and degradation rates). More parameter values are
required for inclusion of the LNAPL in the model to obtain realistic treatment of the source.
By including the contaminant source, the model of a site can include the entire mass of
contaminants present in the subsurface. From this, an appropriate release rate to the
-15-
-------
aquifer, the expected duration of contamination and LNAPL imposed limitations to
remediation can be simulated. As illustrated in the following case study, uncertainties in
input data and field observations of contaminant distributions will, however, limit the
application of models to these sites.
Application of HSSM to a Field Site
The HSSM model was applied to a leaking underground storage tank site, where
data were drawn from the State Agency case file (Sosik, 1996). Modeling of the spill was
not considered an essential or integral assessment activity and modeling was not a part of
the site assessment. Table 3 lists general features of the release and investigation. The
objectives of the model application were to determine whether HSSM could reproduce the
observed contaminant distributions and to demonstrate the effect of data gaps on model
results.
Table 3 General features of the spill site used to
demonstrate HSSM
Item
Hagerman Ave
Release Date
Unknown
Release Volume
Unknown
Gasoline Composition
Unknown
Mass in Ground Water
Estimate
Cores Analyzed
30
Monitor Wells
48(a)
Sample Rounds
3
Data Points Per Sample
Round
210
Slug Tests
13
Pump Tests
1
(a) 26 multilevel samplers and 22 screened wells (b) 6 samples only from MW-2
-16-
-------
Nature of the Hagerman Avenue Plumes
Subsurface contamination was detected at E. Patchogue, New York, when water from a
residential well on Hagerman Avenue became undrinkable. The site investigation began at
the well and expanded through the drilling of monitoring wells in the upgradient and
downgradient directions (Figure 8). The purpose of the drilling was to delineate the extent
of contamination and locate the suspected source. Ultimately, the source was traced back
to an abandoned service station approximately 1200 m (4000 ft) upgradient from the
Hagerman Avenue residence. Soil borings in the area of the service station confirmed the
presence of hydrocarbon contamination. The service station's tanks were removed in
1988. In 1994 and 1995, the contaminant plume was mapped from samples taken from 26
multilevel samplers and 22 monitoring wells. Water samples from three sample rounds
were analyzed for BTEX and MTBE. Total organic carbon contents were determined on 11
clean core samples.
The Hagerman Avenue site is unusual in the sense that the benzene plume is long
compared to average values cited from plume studies. Such studies have summarized
plume lengths that were derived either through the application of solute transport models or
data evaluation techniques applied to data sets from leaking underground storage tank
sites. Generally the results showed that the average length of benzene or total BTEX
plumes from these data are on the order of 200 ft long (Newell and Connor, 1997, Rice et
al., 1995, Mace et al., 1997 and Groundwater Services Inc., 1997). These studies do not
preclude longer plumes; particularly as noted by Newell and Connor (1997) where the
longest BTEX plume was greater 3000 ft long. With the possible exception of Newell and
Connor (1997) which was based on a nationwide survey, these studies were undertaken in
geologic environments unlike the coarse sand and gravel aquifers of Long Island, and thus
do not necessarily represent contaminant behavior on Long Island.
One important key to understanding the length of the Hagerman Avenue benzene plume
is the vertical characterization that was undertaken. The Hagerman Avenue plume dives
into the aquifer as it is transported away from the gasoline source. If the plume was only
characterized by sampling the top ten feet of the aquifer, then the plume would falsely be
assumed shorter than it actually was (Weaver et al., 1999). This observation would be
made because the diving plume would drop below the bottom of the sampling network. If
sampled in this fashion the benzene plume would have been thought to be about one fifth its
actual length.
A study of California MTBE plumes, showed that they ranged from 0.18 to 3.4 times the
length of benzene plumes (Happel et al., 1998). MTBE plumes were included in the study
only if they were adequately delineated by a monitoring network designed for benzene.
One possible reason for the variability of the results is that gasoline composition is variable.
Thus MTBE may or may not be present in fuel that was released at a specific time.
Releases at these sites could consist of varying patterns of gasoline with and without
-17-
-------
MTBE. Sites with MTBE plumes shorter than benzene plumes could be the result of a
continuing series of releases that only contained MTBE in later years. The MTBE plume
may be shorter because it was released later and had not yet had time to extended further
from its source.
The Hagerman Ave MTBE plume is apparently detached from the gasoline source
(Figure 9), whereas the benzene plume is not (Figure 10). Rather than approach this from
the viewpoint that there is a statistical relationship between the lengths of the plumes as in a
plume study, applying a model like HSSM uses the viewpoint that the release scenario,
biodegradation rate, ground water velocity, chemical and other parameters determines the
relationship between the plumes at various times throughout the simulation.
Analysis of Data from Hagerman Avenue, East Patchogue, New York
Published studies of groundwater flow on Long Island indicate that a regional ground
water divide lies along the length of the island and to the north of the geographic centerline
(Eckhardt and Stackelberg, 1995). South of the divide, flow is generally toward the Atlantic
Ocean. Buxton and Modica (1993) estimate that the hydraulic conductivity of the upper
glacial aquifer is on the order of 8.1x10"2 cm/sec (230 ft/day) in the outwash section near
the southern shore, with estimated ground water velocities of 3.5x10"4 cm/sec (1 ft/day) or
greater. Based on a regional water balance (Franke and McClymonds, 1972) estimated
the average recharge rate to be 17 cm/year (22 inches/year).
The use of methyl tert-butyl ether, MTBE, began on Long Island in the late 1970s, after
EPA approved its usage as an octane enhancer. Initial usage of MTBE on Long Island was
likely in the range of 5% by volume. Oxygenated additives were mandated to reduce
carbon monoxide emissions during the winter months in various locations, including New
York City and Long Island communities, by the 1990 amendments to the Clean Air Act as
implemented in the Oxygenated Fuel (Oxyfuel) Program (USEPA, 1998). State of New
York regulations have required use of fuel with oxygen content between 2.7% and 2.9% in
the winter months since 1992 (State of New York, 1995). The most commonly used
oxygenated additive is MTBE, which provides the required oxygen content at about 15%
MTBE by volume. In 1995, the U.S. EPA initiated the Reformulated Gasoline Program
(RFG) with requires the year round addition of 2% oxygen by weight to reduce ozone and
smog. The New York area currently participates in this program (USEPA, 1998).
Subsequent to its introduction MTBE contamination has been found in ground and surface
waters (Squillace et al., 1995).
The subsurface behavior of MTBE is notable for two reasons. First, MTBE is highly
water soluble. As a measure of the solubility, the fuel/water partition coefficient for MTBE is
about 23 times lower than that for benzene and 280 times lower than those for o- or p-
xylene. The release of MTBE from gasoline, therefore, is expected to be more rapid than
-18-
-------
the release of BTEX. Secondly, MTBE is recalcitrant to biodegradation. Microcosm
studies conducted with three soils showed no degradation of MTBE over a 250 day study
period under anaerobic conditions (Yeh and Novak, 1994). Degradation was induced
under anaerobic conditions with the addition of nutrients, a hydrogen source and molybdate
in an organic-poor soil. In organic rich soils degradation of MTBE could not be induced.
Horan and Brown (1995) concluded MTBE degradation might occur at a very low rate,
however, under aerobic conditions. In a controlled field study, gasoline with 10% MTBE,
and an 85% methanol/15% gasoline blend were released in the same aquifer (Hubbard et
al., 1994). MTBE was found to be recalcitrant to degradation, while methanol and BTEX
were degraded. Further, the MTBE had no measurable effect on the degradation of the
other compounds.
Moments Analysis
The relatively large number of monitoring wells and multilevel samplers at Hagerman Ave
generated a three-dimensional data set, which was analyzed by calculating the moments of
each concentration distribution. The moments, Mijk, are defined by
fff is
where x, y, and z are the moment arms, n is the porosity, and C(x,y,z) is the concentration.
These moments can be used to estimate the mass of the contaminant distribution, given by
the zeroth moment, M000. Likewise the first moments can be used to determine the center
of mass of the distribution:
X -
M000
Mmd
Ve 19
z =
c • iDDD
M'
where Xc, yc, and Zc are the x, y, and z coordinates of the center of mass of the distribution.
-19-
-------
The challenge in applying equation 19 to field data lies in evaluating the integrals. The
moment estimates were developed by dividing the contaminant plume into a set of nearest-
neighbor polygons. The polygons represent zones of influence of each well. In essence,
the polygons replace the explicit interpolation schemes between sampling locations that
have been used in other analyses (Freyberg, 1986, among others). For most of the plume,
the wells cross the entire width of the plume. In some upgradient locations, however,
monitoring wells with relatively high contaminant concentrations are located on the edge of
the sampling network (MW-12, MW-30, MW-38, MW-39), which causes some of the BTEX
to be omitted from the following estimates. Because the MTBE is located entirely
downgradient of MW-30, MW-38, and MW-39, its mass estimates were not greatly
impacted by this problem.
Table 4 shows the mass estimates and the distance of the center of mass of the
contaminant distribution from the contaminant source, dCOm- Since the samples in round
one were taken over a long time period, contaminants sampled upgradient may have been
transported to downgradient receptor wells before they were sampled. The order of
sampling, however, proceeded upgradient from the discovery point (MW-1) to the
suspected source, followed by the wells downgradient from MW-1, mitigating the problem
somewhat.
Table 4 Mass and location of center of mass for the Hagerman Avenue data
Chemica
I
Sample Round One
July, 1994 to March 1995
(average date: Dec 16, 1994)
Sample Round Two
April 11 - April 20, 1995
Sample Round Three
Oct 10-Oct 24, 1995
M™
M»«
dmm
M™
M»«
dmm
M«
WL
dmm
kq
kq
m
kq
kq
m
kq
kq
m
MTBE
268
24
1387
386
34
1557
229
20
158
3
B
241
156 (122)
991
117
76(59)
1004
58
38 (29)
106
1
T
108
253 (217)
230
65
152 (130)
298
60
141
(120)
306
E
29
249 (153)
347
24
206 (127)
347
21
180
(111)
326
X
149
1041 (804)
222
95
663 (513)
277
92
643
(497)
272
Table Notes
M^is the mass of chemical x dissolved in ground water.
is the estimated mass of chemical x sorbed to aquifer solids.
The first value was estimated by using the value of Koc reported by Mercer and Cohen (1990); the value given
in parenthesis used the estimate from US EPA (1990).
d^ is the distance from the suspect source to the center of mass of the contaminant distribution.
Each of the chemicals listed in Table 4 has some tendency for sorption, which must
-20-
-------
be included in estimates of the total mass. Chemicals sorb in proportion to the fraction of
organic carbon in the aquifer material, foc, and the chemical's organic carbon partition
coefficient, Koc. Sorption was assumed to follow the linear equilibrium isotherm as given by
where Cxs is the sorbed concentration of contaminant x expressed per unit mass of aquifer
solids, and Cxw is dissolved concentration of chemical x. The sorbed mass
of contaminants was estimated from
= ~ ^ae ?oc My1
n
where Mxs and Mxw are the respective sorbed and dissolved masses of chemical x, and pb
is the bulk density. Organic carbon contents were determined for 11 uncontaminated
samples taken from 4.88 m to 8.23 m (16 ft to 27 ft) below the ground surface near the
source. The arithmetic average of foc was 0.126%, with range of 0.009% to 0.627% and
standard deviation of 0.190%. The porosity and solids density were assumed to equal 0.30
and 2.65 g/cm3, respectively, giving a bulk density of 1.86 g/cm3. The Koc values were
taken from Table 5 which lists the density, p, solubility, S, organic carbon partition
coefficient, Koc, fuel/water partition coefficient, K0, and the mass fraction in gasoline, x, of
MTBE and the BTEX compounds. Koc values were taken from Mercer and Cohen, (1990)
and US EPA (1990). The fuel/water partition coefficient and mass fraction data were
measured by Cline et al. (1991) on 31 samples of gasoline from Florida. The range
reported covers the variation in measured mass fractions in samples from other parts of
the continent and from lists of typical gasoline compositions (see e.g., Cline et al., 1991,
Corapcioglu and Baehr, 1987).
-21-
-------
Table 5 Chemical parameter values
Chemical
Density
Solubility
Koc (a)
K0(b)
Gasoline mass
fraction (c)
Federal MCL
q/ml
mq/L
L/kq
%
Uq/L
MTBE
0.74
48000
11.2
15.5
—
none
benzene
0.88
1750
83 (65)
350
1.73(0.7-3.8)
5
toluene
0.87
535
300 (257)
1250
9.51 (4.5-21.0)
1000
ethylbenzene
0.87
152
1100 (676)
4500
1.61 (0.7-2.8)
700
m-xylene
0.86
130
982 (691)
4350
(d)
10000
o-xylene
0.86
196
879 (691)
4350
(d)
10000
p-xylene
0.86
175
830 (691)
3630
2.33(1.1-3.7)
10000
Table notes:
(a) Organic carbon partition coefficient reported by Mercer and Cohen (1990), the second value (in
parenthesis) from US EPA (1990).
(b) Average fuel/water partition coefficient reported by Cline et al. (1991).
(c) Mass fraction of chemical in gasoline reported by Cline et al. (1991).
(d) o- and p-xylene were not differentiated, they composed 5.95% with range of 3.7% to 14.5%.
The estimated mass of benzene, toluene, ethyl-benzene and the xylenes decreased
between each sample round (Table 4). Each of these compounds is expected to undergo
biodegradation in the aquifer, but each continued to dissolve into the aquifer through at
least sample round three collected in October 1995. The latter is established by the
persistence of BTEX concentrations near the source. The mass of MTBE, however,
appeared to increase between the first two sample rounds, then decreased between the
second and third sample rounds. The distribution of MTBE was such that in all sample
rounds, no MTBE was found between the source and a point approximately 600 m (2000
ft) downgradient (Figure 9). Thus, it appears that the MTBE was almost entirely leached
from the gasoline in the source.
Estimation of the Mass of Gasoline Released
The mass of contaminants in the aquifer can be used to place bounds on the volume
of gasoline released. The total (aqueous + sorbed) estimated mass of MTBE in the
aquifer is 292 kg for sample round one and 420 kg for sample round two. Since the tanks
were pulled in 1988, the gasoline was released before the Clean Air Act mandates, so
MTBE was assumed to comprise 5% by volume of the gasoline. The corresponding
volume of gasoline for these estimated masses would be 7.89 m3 (2080 gallons) and 11.35
m3 (2999 gallons). Because of the apparent complete leaching of MTBE from the gasoline,
this estimate would represent the entire volume of MTBE-enhanced gasoline released to
the aquifer.
-22-
-------
The BTEX data suggest the volumes of gasoline listed in Table 6, assuming that the
density of the gasoline was 0.72 g/cm3. In the absence of specific knowledge concerning
the composition of the released gasoline, the estimates developed by Cline et al. (1991)
(see Table 5) were used in estimating the gasoline volumes in Table 6. Unlike MTBE,
each of the BTEX chemicals persists in gasoline at the source (Figure 10). More of the
benzene originally contained in the gasoline, however, would be in the aquifer than any of
the other BTEX compounds because of benzene's lower fuel/water partition coefficient. A
greater fraction of each of T, E and X remain in the gasoline because of their higher
affinities for the gasoline phase (expressed in Table 5 by their lower water solubilities and
higher fuel/water partition coefficients). Because of this and the biodegradation of bezene
in the aquifer, the benzene-based gasoline volume estimate is a minimum. To contrast with
the Cline et al. (1991) average benzene mass fraction of 1.7%, the often-used estimate of
1 % by mass gives an estimated gasoline volume of 55.1 m3 (14600 gallons) for a benzene
mass of 397 kg and 50.4 m3 (13300 gallons) for a benzene mass of 363 kg.
Table 6 Constituent mass estimates and gasoline volume estimates
from sample round one
Chemical
Mass estimate
mass
Gasoline volume
estimate
Mass fraction from Cline
et al. (1991)
kg
low
middle
high
benzene
(a)
397
20807
8567
3832
(b)
363
19025
7834
3505
toluene
(a)
361
2943
1393
631
(b)
325
2650
1255
568
ethylbenzene
(a)
278
14624
6335
3643
(b)
182
9539
5273
2399
xylenes
(a)
1190
9096
5273
2399
(b)
953
7284
4223
1921
(a) Sorbed mass estimate using Koc of Mercer and Cohen (1990)
(b) Sorbed mass estimate using Koc of USEPA (1990)
Simulation of Hagerman Avenue
The data set from Hagerman Avenue was used to estimate the ground water flow
-23-
-------
velocity, the volume of gasoline released and the mass of BTEX and MTBE released to the
aquifer. The MTBE in the aquifer traveled from the source zone to its center of mass in
1994 and 1995 in 16 years or less. Using the centers-of-mass from Table 4, the MTBE
plume advanced at the rates listed in Table 7. These rates suggest that the average
transport rate is nearly constant for distances between 1387 m and 1583 m from source,
which were the locations of the MTBE center of mass. The rate would have been 0.65 m/d
if the entire release occurred on December 31, 1988, and the 0.25 m/d if the release
began on January 1, 1979. The data also suggest a release volume of at least 13,200
gallons, which contains 397 kg of benzene. The volume estimate from the MTBE data
(2000 to 3000 gallons) may be significantly lower because of intermittent MTBE use prior to
1992, and seasonal use thereafter. The release date, or dates if from a series of releases,
is unknown. Because of the MTBE, some gasoline must have been released after 1979.
The tanks were pulled in 1988, thus providing a definitive ending date.
Table 7 Average velocities from MTBE concentration data.
Sample
Date
Distance
Days Since
Velocity
Days Since
Velocity
m
Dec 31,1988
m/d
Jan 1,1979
m/d
1
Dec 16,1994
1387
2176.25
0.64
5828.75
0.24
2
ApriM 5,1995
1557
2297.5
0.68
5950.0
0.26
3
Oct 17,1995
1583
2481.5
0.64
6134.0
0.26
With the estimated release parameters, the approach taken to simulating the site
was to attempt to match sample round two (April 15, 1995) data for total xylenes, benzene
and MTBE. The strategy was as follows. The values of the aquifer hydraulic parameters
and release dates were set by matching the model to the MTBE data. The previously
described release volume and MTBE mass were to be used without adjustment and MTBE
was assumed not to biodegrade. Once the MTBE date were adequately matched,
benzene and total xylenes were simulated. The goal was that only chemical parameters,
the mass fraction in the fuel and biodegradation rates would be adjusted. After calibration
to one set of data, ideally the hydraulic parameters should not be adjusted if only the
chemical has changed. (A potential exception are the dispersivities, which may depend on
the flow length and thus sorption.) A further assumption was that a single release scenario
would be adequate for all chemicals, i.e., there were no prior releases of non-MTBE
gasoline which created the benzene and total xylenes plumes. The three chemicals were
picked for simulation as they represent a range of effective solubilities and each plume had
a different qualitative appeareance.
MTBE Simulation
-24-
-------
The centerline MTBE concentrations were simulated as shown in Figures 11 and 12.
The HSSM result is plotted against the average MTBE concentration from wells on the
approximate centerline of the plume. The objectives of the simulation were to match the
qualitative distribution of MTBE-detached from the source (Figure 9) and to come close to
the observed concentrations. The release dates were chosen to be January 1, 1984 to
September 15, 1988, which gave a reasonable match to the leading and trailing edges of
the MTBE plume. The 13200 gallons of gasoline were released during this time period. To
obtain the estimated 420 kg of MTBE in the aquifer, its mass fraction in gasoline was set to
1.2%. The hydraulic conductivity of the aquifer was set to 118 m/d with a gradient of
0.0013 and porosity of 0.30 (Figure 11), giving a seepage velocity of 0.51 m/d. Parameter
values for the simulation are listed in Tables 8a and 8b along with their source. The
parameters listed with "calibration" as the source were determined from a number of
simulations that resulted in order of magnitude matches to the observed data. Chemical
properties of xylene, benzene and MTBE were taken from Table 5 and were not adjusted in
attempting to match the field data. The critical parameters for simulating the site were
found to the duration of the release, ground water flow velocity, aquifer dispersivity,
degradation rate constants, and water table fluctuation. These parameters are important to
the simulation results because they determine the mass of contaminants released to the
aquifer, the advective transport rate, the spreading of the contaminant plume, the rate of
loss of mass, and size of the aquifer boundary condition, respectively.
In this simulation the MTBE distribution appears to be off-set from the field data by
about 1500 feet, although the concentrations were in the correct range. By increasing the
seepage velocity to 0.63 m/d, the better temporal match of Figure 12 was obtained. The
reduced simulated concentrations of Figure 12 result from increased dispersion at higher
velocity.
-25-
-------
Table 8a Parameters for the Hagerman Avenue simulations
Item
Value
Data Source
Hydraulic Conductivity
118, 122 m/d
Calibration
Hydraulic Gradient
0.0013
Field Data
Depth to Water
6.1 m
Field Data
Aquifer Thickness
30 m
Field Data
Porosity
0.30
Estimate
Fraction Organic Carbon
0.126 %
Laboratory Measurement
Water Table Fluctuation
0.3 m
Calibration
Recharge Rate
20 in/yr
Literature
Capillary
Pressure
Curve
parameters
Brooks and Corey lambda
1.69
Laboratory Measurement
Air Entry Head
10 cm
Laboratory Measurement
Residual Water Saturation
0.01
Laboratory Measurement
Dispersivity
Longitudinal
10 m
Calibration
Transverse
0.23 m
Calibration
Vertical
0.1 m
Calibration
Mass
Fraction in
Gasoline
MTBE
1.2%
Estimate
benzene
1.14%
Calibration
xylenes
4%
Calibration
Half Life
MTBE
Infinite
Estimate
benzene
912 days
Calibration
xylene
365 days
Calibration
-26-
-------
Table 8b Parameters forthe Hagerman Ave Simulations (continued)
Item
Value
Source
Gasoline
Density
0.72 g/ml
Literature
Viscosity
0.45 cp
Literature
Surface Tension
35 dyne/cm
Literature
Aquifer Residual Saturation
0.15
Estimate
Vadose Zone Residual
Saturation
0.05
Estimate
Source
Area
15 m2
Estimate
Volume
15 m3 (13200
gallons)
Estimate
Beginning Date
Jan 1,1984
Calibration
Ending Date
Sep 15, 1988
Estimate
Benzene Simulation
Benzene was simulated using the seepage velocity of 0.51 m/d (Figure 13). The
benzene mass fraction in gasoline was 1.14% and its half life was 912 days. The latter
value was arrived at by matching observed concentrations. The benzene distribution shows
that the concentrations are low for roughly 3000 feet downgradient from the source (Figure
10). Higher concentrations are then encountered, which reflect the early release of mass
into the aquifer. Benzene released later has the low concentrations observed near the
source because the source becomes depleted. The model reproduced this semi-detached
plume behavior and produced concentrations close to those observed. The modeled
plume, however, extends further downgradient than do the data. The downgradient edge of
the plume reflects the beginning time of the release. This time was not increased for these
simulations, because the MTBE gasoline released in January 1984 would certainly contain
benzene. Assuming a later beginning time for the benzene simulation would use the implicit
assumption that MTBE was released without benzene.
Total Xylenes Simulation
-27-
-------
Total xylenes were simulated by assuming the same seepage velocity as benzene
(0.51 m/d), total xylene mass fraction of 4% and half life of 365 days. The total xylene
distribution remains closest to the source as it has the highest fuel/water partition
coefficient, highest tendency for sorption and shortest assumed half life. Both the model
and field data show similar contaminant distributions.
Conclusions
Reports from leaking underground storage tank sites reveal that site information is
usually limited for model application. Many of the transport parameters needed for
simulating the releases are not available, even those needed for simplified models like
HSSM. Release dates, fuel compositions and volumes are rarely known. These form,
however, critical boundary conditions for model application, which then must become
subject to calibration or estimation. Since at most sites contamination is detected years
after releases occur, much of the data corresponding to model results (e.g., aqueous
contaminant concentrations) is of limited duration relative to the lifetime of the
contamination event.
Following the typical pattern the release that occurred at Hagerman Avenue
occurred at unknown times and intervals, and much about the contamination at the site
remains unknown. The extensive monitoring network at this site allowed estimation of the
mass and moments of the contaminant distributions. These were used to estimate certain
input parameter values for the model (gasoline volume and contaminant mass). Their
accuracy, however, depends upon the sampling network, the duration of sample events,
and the accuracy of the procedure used for forming the estimates. During calibration of the
model these were not varied, because an adequate fit to the data was obtained. Release
timing is also required for specifying the boundary condition. In some cases regulatory
actions taken at sites can provide bounding dates. For Hagerman Ave the most
reasonable starting assumption was that the gasoline was released over some portion of
the active life of the service station. Tank removal dates and historical usage of MTBE
provided a rough means of bounding the release. Further refinement of the release dates
was made in running the simulations. In the case of detached plumes, the location of the
center of mass, leading edge and trailing edge allow estimation of the transport time. As
for the release volume, release dates could be selected to model the behavior of the
plumes. The simulations showed that HSSM had the capability of simulating the detached
MTBE plume, the semi-detached benzene plume and the attached total xylenes plume.
These three types of plumes occurred at the same site and were simulated with roughly the
same hydraulic inputs. Using a different seepage velocity for MTBE might be justified by
noting that the MTBE plume is in a different part of the aquifer than the benzene plume.
With this exception the model results showed that the differences between the plumes were
caused by differing chemical properties, biodegradation rates and mass fractions in
-28-
-------
gasoline. Although parameter sets were found for simulating each of the plumes, alternate
values of various parameters can be found that give equivalent matches to the field data.
This variability is cause by limited field data, limited accuracy of site characterization
approaches, model assumptions, and treatment of input data as fitting parameters. Thus,
the input parameter sets and the approximations in the model are not completely
representative of contaminant transport at the site. One clear example is found in the
vertical profiles (Figures 9 and 10). Here the plume is shown to vary across the vertical
while HSSM uses vertically averaged concentrations (Equation 11). Since the components
of HSSM do not include heterogeneity, a use for the model in some situations is to
generate a mass input function for the aquifer, but then to simulate aquifer contamination
with a numerical solute transport model, such as MT3D (Zheng and Bennett, 1995). Using
a numerical model would provide the flexibility to simulate the effects of pumping wells,
multidimensional ground water flow and realistic aquifer boundary conditions.
The simulations provide insights into contaminant behavior at the site that are not
obtained from field data alone. Application of the model to the sites is valuable for showing
rough agreement between the data and model results and developing plausible release
scenarios that essentially extrapolate information from the observed contaminant
distribution. At most sites, inferences concerning contaminant behavior must often be
drawn from case histories with little detailed information on the source of contaminants and
limited information on the distribution of contaminants in the aquifers. Application of
models to the sites do not eliminate these limitations, but provide insight into contaminant
behavior based on the observed data. The model constructs an entire contaminant
release, flow, and transport scenario that when matched to field observations provides a
conceptualization of the entire contamination event.
Acknowledgment
The information in this document has been funded wholly or in part by the United
States Environmental Protection Agency. It has been subjected to Agency review and
approved for publication.
The authors thank Joseph Haas, New York State Department of Environmental
Conservation, and Charles Sosik of Environmental Assessment and Remediation of
Patchogue, New York for providing the Hagerman Ave data set.
-29-
-------
References
Alberta Research Council, 1994, Composition of Canadian Summer and Winter Gasolines
1993, Canadian Petroleum Products Institute and Environment Canada, Report No. 94-5.
Bear, J., 1972, Dynamics of Fluids in Porous Media, Dover, New York, 764 pp.
Bouwer, H., 1966, Rapid field measurements of the air entry value and hydraulic
conductivity of soil as significant parameter in flow system analysis, Water Resources
Research, 2, 729-738.
Buxton, H.T., and E.Modica, 1993, Patterns and rates of ground-water flow on Long Island,
New York, Ground Water, 30(6), 857-866.
Carsaw, H.S. and J.C. Jaeger, 1959, Conduction of Heat in Solids, Oxford University
Press, Oxford, England, 2nd ed., 510 pp.
Charbeneau, R.J., J.W.Weaver, and B.K.Lien, 1995, The Hydrocarbon Spill Screening
Model (HSSM) Volume 2: Theoretical Background and Source Codes, US Environmental
Protection Agency, Ada, OK, EPA/600/R-94/039b.
Chrysikopoulos, C.V., 1995, Three-dimensional analytical models of contaminant transport
from nonaqueous phase liquid pool dissolution in saturated subsurface systems, Water
Resources Research, 31(4), 1137-1145.
Cline, P.V., J.J.Delfino, and P.S.C.Rao, 1991, Partitioning of aromatic constituents into
water from gasoline and other complex solvent mixtures, Environmental Science and
Technology, 23, 914-920.
Corapcioglu, M.Y., and A. Baehr, 1987, A compositional multiphase model for groundwater
contamination by petroleum products, Water Resources Research 23, 201-243.
Eckhardt, D.A.V. and P.E.Stackelberg, 1995, Relation of ground-water quality to land use
on Long Island, New York, Ground Water, 33(6), 1019-1033.
Felhberg, E., 1969, Low-Order Classical Runge-Kutta Formulas Wth Stepsize Control and
Their Application to Some Heat Transfer Problems, National Aeronautics and Space
Administration, Washington, D.C., TR-315.
-30-
-------
Franke, O.L. and N.E. McClymonds, 1972, Summary of the Hydrologic Situation on Long
Island, New York, as a Guide to Water-Management Alternatives, United States Geological
Survey, Professional Paper 627-F.
Freyberg, D.L., 1986, A natural gradient experiment on solute transport in a sand aquifer:
2. Spatial moments and the advection and dispersion of nonreactive tracers, Water
Resources Research, 22(13), 2031-2046.
Green W. H. and G. A. Ampt, 1911, Studies on soil physics, J. Agric. Sci., 4, 1-24.
Groundwater Services Inc., 1997, Florida RBCA Planning Study: Impact of RBCA Policy
Options on LUST Site Remediation Costs, Florida Partners in RBCA Implementation
(PIRI).
Gustafson J.B., J.G. Tell, and D Orem, 1997, Selection of Representative TPH Fractions
Based on Fate and Transport Considerations, Total Petroleum Hydrocarbon Criteria
Working Group Series, Volume 3, Amherst Scientific Publishers, Amherst, Massachusetts.
Horan, C.M., and E.J. Brown, 1995, Biodegradation and inhibitory effects of Methyl-
Tertiary-Butyl Ether (MTBE) added to microbial consortia, Proceedings of the 10th Annual
Conference on Hazardous Waste Research, May 23-24, Kansas State University,
Manhattan, Kansas, 11-19.
Happel, A.M., E.H. Beckenbach, and R.U. Halden, 1998, An Evaluation of MTBE Impacts to
California Groundwater Resources, Lawrence Livermore National Laboratory, University of
California, UCRL-AR-130897.
Hubbard, C.E., J.F. Barker, S.F. O'Hannesin, M. Vandegriendt, and R.W. Gillham, 1994,
Transport and Fate of Dissolved Methanol, Methyl-Tertiary-Butyl-Ether, and Monoaromatic
Hydrocarbons in a Shallow Sand Aquifer, American Petroleum Institute, Washington D.C.,
Health and Environmental Sciences Publication 4601.
Hunt, J.R., N. Sitar, K.S. Udel, 1988, Nonaqueous phase liquid transprot and cleanup: 1.
Analysis of mechanisms, Water Resources Research, 24(8), 588-598.
Mace, R.E., R.S. Fisher, D.M. Welch and S. P. Parra, 1997, Extent, Mass and Duration of
Hydrocarbon Plumes from Leaking Petroleum Storage Tank Sites in Texas, The University
of Texas at Austin, Bureau of Economic Geology, Circular 97-1.
-31-
-------
Mercer, J.W., and R.M. Cohen, 1990, A review of immiscible fluids in the subsurface:
Properties, models, characterization and remediation Journal of Contaminant Hydrology, 6,
107-163.
Neff J. M., D. E. Langseth, E. M. Graham, T. C. Sauerand S. C. Gnewuch, 1994, Transport
and Fate of Non-BTEX Petroleum Chemicals in Soils and Groundwater, American
Petroleum Institute, No. 4593, Washington, D.C.
Newell C.J. and J.A. Connor, 1997, Characteristics of Dissolved Petroleum Hydrocarbon
Plumes: Results from Four Studies, American Petroleum Institute, Tech Transfer Bulletin
Rice, D.W., R.D. Grose, J.C. Michaelsen, B.P. Dooher, D. H. MacQueen, S.J. Culen, W. E.
Kastenberg, L.G. Everett, and M.A. Marino, 1995, California Leaking Underground Fuel
Tank (LUFT) Historical Case Analyses, Lawrence Livermore National Laboratory,
University of California, UCRL-AR-1222207.
Smith V.J. and R. J. Charbeneau, 1990, Probabilistic soil contamination exposure
assessment procedures, American Society of Civil Engineers, Journal of Environmental
Engineering 116(6), 1143-1163.
Sosik, C., 1996, Subsurface Investigation Report, East Patchogue, NY, NYSDEC Spill #94-
04094, Environmental Assessment and Remediation, Patchogue, New York.
State of New York, 1995, Official Compliation of Codes, Rules and Regulations of the State
of New York, Title 6 Environmental Conservation Subpart 255-3, Albany, New York.
Squillace, P.J., J.S.Zogorski, W.G.Wilber, and C.V.Price, 1995, A Preliminary Assessment
of the Occurrence and Possible Sources of MTBE in Ground Water of the United States,
1993-94 U.S. Geological Survey Open File Report 95-456.
United States Environmental Protection Agency, 1998, MTBE Fact Sheet #1, Office of
Underground Storage Tanks, Washington, D.C., EPA/510/F-98/001.
United States Environmental Protection Agency, 2000, UST Corrective Action Measures
for Second Half FY 99, Office of Underground Storage Tanks, Washington, D.C.,
http://www.epa.gov/swerust1/catmarchv.htm.
United States Environmental Protection Agency, 1990, Subsurface Remediation Guidance
Table 3, U. S. Environmental Protection Agency, Washington, D.C., EPA/540/2-90/011b.
-32-
-------
Weaver, J.W., R. J. Charbeneau, and B. K. Lien, 1994a, A screening model for
nonauqeous phase liquid transport in the vadose zone using Green-Ampt and kinematic
wave theory, Water Resources Research, 30(1), 93-105.
Weaver, J.W., R.J.Charbeneau, J.D.Tauxe, B.K.Lien and J.B.Provost, 1994b, The
Hydrocarbon Spill Screening Model (HSSM) Volume 1: User's Guide, US Environmental
Protection Agency, Ada, OK, EPA/600/R-94/039a.
Weaver, J.W., J.E. Haas, C.B. Sosik, 1999, Characteristics of Gasoline Releases in the
Water Table Aquifer of Long Island, National Ground Water Association/American
Petroleum Institute, Proceedings of the Petroleum Hydrocarbons and Organic Chemicals in
Ground Water, Houston, Texas.
Yeh, C.K. and J. T. Novak, 1994, Anaerobic biodegradation of gasoline oxygenates in
soils, Water Environment Research, 66(5), 744-752.
Zheng, C and G.D. Bennett, 1995, Applied Contaminant Transport Modeling, Van Nostrand
Reinhold, New York, 440pp.
-33-
-------
Land Surface
Aquifer
Figure 1 LNAPL release into a mildly heterogeneous vadose zone, illustrating
irregular downward migration of the LNAPL and pooling on the water table.
Downgradient migration of the LNAPL may occur in response to the water table
gradient. Water table fluctuation may create a smear zone.
-------
Land Surface
Figure 2 Idealized LNAPL release scenario used in HSSM. Assumptions include one-
dimensional flow in the vadose zone, approximate treatment of the smear zone, and
formation of a radially symmetric lens.
-------
Time
Figure 3 Typical breakthrough curve (concentration history) at a HSSM receptor point.
-------
Saturation
Saturation
Figure 4 Idealized LNAPL profiles during infiltration (left) and redistribution (right).
-------
napl Release
Pressure and
Gravity
Gravity
Figure 5 Schematic illustration of KOPT model solution.
-------
'KOPT
radial
'KOPT
Hi
^ Qout
Figure 6 Lens and central cylinders used in formulating the 01 LENS module.
-------
Figure 7 LNAPL lens during lens decay. LNAPL is trapped above and below the lens at
the vadose zone and aquifer residual saturations, respectively.
-------
Figure 8 Hagerman Avenue site plan.
-------
O O O O th Q
o o o o a o
o a o a o
^ a 3 a s
y uonavara
o
£
T3
c
3
o
!-h
(L)
S3
C/3
c
_o
"3
'C
w
PQ
H
c\
a
¦-
S
otj
-------
fien-ZRne ('ooh*!
1000 150D 2000
3000 3500 4D00
Di&lance (ft)
4 50D 5000
6000 6500
Figure 10 Benzene distribution in sample round two.
-------
4
H3SM Result
~—~ Field Data
0
0
2000
4000
6000
Distance from Source (ft)
Figure 11 HSSM simulation of centerline MTBE concentration using seepage velocity of 0.51 m/d.
Field data and model results both show detachment of the MTBE plume from the source at x = 0.
-------
4
H3SM Result
~—~Field Data
0
Distance from Source (ft)
Figure 12 HSSM simulation of centerline MTBE concentration with seepage velocity of 0.63 m/d.
Lowered concentration is a result of increased dispersion with greater velocity.
-------
1
H3SM Result
~—~Field Data
Distance from Source (ft)
Figure 13 HSSM simulation of centerline benzene concentration with seepage velocity of 0.51 m/d.
The semi-detached nature of the benzene plume is shown by the low concentrations near the source at
distances less than 3000 feet.
-------
o
d
------- |