HYDROCARBON SPILL EXPOSURE
ASSESSMENT MODELING
Ground Water Management
Book 4 of the Series
Proceedings of the
1990
Petroleum Hydrocarbons and Organic Chemicals
in Ground Water. Prevention, Detection, and
Restoration
Abstract
This bound volume contains papers presented at the 1990 Petroleum Hydrocarbons Conference. The meeting
was held October 31-November 2, 1990, in Houston, Texas.
'The conference is annually co-sponsored by the Association of Ground Water Scientists and Engineers and
the American Petroleum Institute. A combined total of 52 oral and poster presentations was presented in 1990
focusing on prevention of contamination, detection, and monitoring of hydrocarbons in soils and ground
water, transport and fate, free-phase hydrocarbon recovery, defining remediation levels, remediation of residual
phase hydrocarbons (by vapor extraction and bioventing),' and biodegradation.
Invited presenters included Jay Lehr, National Water Well Association; Bob Hockman, Amoco Corp.; Rob
Booth, Environment Canada; and Iris Goodman, U.S. EPA. Attendees represented a broad spectrum of interests
and specializations. f
The co-sponsors hope environmental scientists find the papers in this publication to be of great value.
Conference on Petroleum Hydrocarbons and Organic Chemicals in Ground Water-Prevention,
Detection, and Restoration (1990 ; Houston, Tex.)
Proceedings of the 1990 Conference on Petroleum Hydrocarbons and Organic Chemicals in
Ground Water-Prevention, Detection, and Restoration : October 31-November 2, 1990, the
Westin Galleria, Houston, Texas / sponsored by the American Petroleum Institute and the
Association of Ground Water Scientists and Engineers ...
p. cm. - (Ground water management, ISSN 1047-9023 ; bk. 4)
"Produced by Water Well Journal Publishing Co."
1. Water, Underground-Pollution-Congresses. 2. Petroleum chemicals-Environmental
aspects-Congresses. 3. Hydrocarbons-Environmental aspects-Congresses. 4. Organic
compounds-Environmental aspects-Congresses. \. American Petroleum
Institute. II. Association of Ground Water Scientists and Engineers (U.S.)
m. Title. IV. Series.
TD426.C64 1990
628.1 ' 6833-dc20
-------
Ground Water Management
A journal for rapid dissemination of ground water research.
To minimize delay of new findings and insights, Ground Water Management publishes papers
exactly as submitted without technical and grammatical editing or peer review. It is the belief of the
editor that these papers have technical merit. Complete accuracy or technical viability cannot, however,
be assured. It is believed, nevertheless, that early publication and rapid dissemination outweigh any
possible reduction in quality that may be encountered.
Papers published by Ground Water Management were presented at conferences sponsored by the
Association of Ground Water Scientists and Engineers, a division of the National Water Well Association.
Staff
Jay H. Lehr, Ph.D., Editor
Anita Stanley, Managing Editor
Chris Miller, Manuscript Coordinator
Pat Alcom, Business Manager
Robert Ison, Circulation
Suzy Colley, Graphics
Subscription Rates
Subscriptions to Ground Water Management ?re for any six issues. Special coupon books are issued to
each subscriber to redeem for issue topics of interest. Each subscriber will receive notification of the
printing of each issue of Ground Water Management.
A subscription expires when all six coupons have been redeemed. A six-issue subscription is $140
($192.50 to non-members), a savings of $70 when compared to the single issue price. Subscribers who
attend NWWA-sponsored conferences have the option of using a subscription coupon for a second
copy of the conference's proceedings, or saving it for a future issue.
Responsibility
The Association of Ground Water Scientists and Engineers, a division of NWWA, is not responsible for
statements and opinions advanced in its meetings or printed in publications.
Copyright
©1990 Water Well Journal Publishing Co.
All rights reserved. ISSN: 1047-9023
Claims
Claims for missing issues must be made in writing within three months of publication and will be
subject to availability of back issues.
Communications and Business Correspondence
Ground Water Management, 6375 Riverside Dr., Dublin, OH 43017, (614) 761-3222.
v-. „.: .• •• -
Materials appearing in thi'J publication are indexed to Ground Water
• PnrLin.ij?, jhe. data base of the*National Ground Water Information
Center'at(614)76Y-l7l'l. ' •" •-.-•.••'
-------
Hydrocarbon Spill Exposure Assessment Modeling
James W. Weaver
Robert S. Ken Environmental Research Laboratory
United States Environmental Protection Agency
Ada, Oklahoma
Randall J. Charbeneau
Center for Research in Water Resources
The University of Texas at Austin
Austin, Texas
Abstract
Hydrocarbon spills impact drinking water supplies at down gradient locations.
Conventional finite difference and finite element models of multiphase, multicornponent flow have
extreme requirements for both computer time and she data. Site data and the intent of the
modeling often do not warrant the application of such models. An alternative approach is
proposed which is based on semi-analytic models for vertical product Infiltration, radial spreading
on the water table, and transport of aqueous phase contaminants in the aquifer. Three Individual
models for these processes are Bnked to estimate exposure at a down gradient well. The
Infiltration model is based on kinematic wave theory supplemented by approximate dynamic
relationships. In the vidntty of the water table the product moves laterally, and is approximated
by considering the effects of buoyancy and assuming that the shape of the oil lens is given by the
Qhyben-Herzberg-DupuK assumptions. The flux of dissolved chemicals to the aquifer is related
to the recharge rate, the size of the lens, the groundwater seepage velocity, and the oil-water
partition coefficient of the dissolved constituents. The boundary condition for the aquifer model
must be able to handle transient mass flux. This is accomplished by integrating the analytic
solution of the advection-dispersion equation with the source as a time-variable boundary
condition. A time record of concentration can be determined for any desired location In the
aquifer. The presentation outlines the methodology and examines parameter sensitivity within
the combined vadose zone, oil lens, and aquifer transport models.
Introduction
When fluids that are immiscible with water (the so-called non aqueous phase liquids or
NAPLs) are released in the subsurface, they remain distinct fluids, flowing separately from the
233
-------
water phase. Fluids less dense than water (L-NAPLs) migrate through the unsaturated zone, but
when reaching the water table, tend to form lenses on top of the aquifer. Generally the fluids are
composed of complex mixtures, so that aquifer contamination results from the dissolution of
various components of the NAPL This paper outlines the methodology for a screening model for
releases or spills of L-NAPLs.
Components of the Model
The model proposed for hydrocarbon exposure assessment modeling is composed of
three parts. The first two parts are transport of the hydrocarbon product downward through the
unsaturated zone and subsequent radial transport along the water table. These two are
simulated by the KOPT/OILENS model, where the acronym KOPT stands for Kinematic Oily
Pollutant Transport. The third part of the model is transport through the aquifer of dissolved
contaminants, which is simulated by TSGPLUME (Transient Source Gaussian Plume). All of the
models are in the form of semi-analytical and analytical solutions of the governing equations. The
conceptual basis of the models are discussed in the following paragraphs. The mathematical
details of KOPT/OILENS may be found in Weaver and Charbeneau (1989) and Charbeneau et al.
(1990), while those of TSGPLUME can be found in Charbeneau and Johnson (1990).
The KOPT/OILENS model is intended to address the specific problem of L-NAPL
transport from the ground surface to a water table aquifer. Since ultimate Interest lies with water
quality, an emphasis of the model is the determination of the oil lens size and the mass flux of
contaminants into the aquifer. These quantities are not predictable without considering three-
phase flow in the unsaturated zone. Thus the model is intended for use In characterizing the
magnitude of the source of aquifer contamination. In modeling terms, KOPT/OILENS provides a
time-variable boundary condition for an aquifer model.
A chemical constituent dissolved in both the NAPL and water phases is also tracked by
KOPT/OILENS. Once that chemical reaches the water table it contaminates the aquifer by
contaminating the recharge water and by dissolution from the L-NAPL lens. The TSGPLUME
model takes the dissolution mass flux from the KOPT/OILENS model and calculates the expected
concentrations at a number of down gradient receptor wells. Notably the mass flux from
KOPT/OILENS is time varying, so that the aquifer model must be capable of simulating a time
varying source condition.
Usage of the Approximate Model
The purpose of the methodology described herein is use as a screening tool. For
example, the model can be used to estimate the effects of L-NAPL loadings, partition coefficients,
groundwater flow velocities, etc., on pollutant transport. Since approximations are used for
developing the model, the model results must be viewed as approximations. If simulation of
complex heterogeneous sites is needed, or other approximations made in the models are
unacceptable, then a more inclusive model, such as the MOFAT code developed at VPI
(Kuppusamy et al., 1987), should be used instead of, or in addition to, KOPT/OILENS and
TSGPLUME.
234
-------
Complex models, however, may not always be the most desirable tool for a given
problem. Such models require large amounts of computer time and available memory. Further,
there may be a significant investment In training users to set up the model and run tt properly.
Additionally, a large amount of field data is required to run such a model because the expense of
running the model is not warranted if adequate site data are not available. In addition to the
parameters for aqueous phase solute transport (such as hydraulic conductivity, dispersivity,
sorption parameters), multiphase transport parameters are needed (interphase partition
coefficients, capillary pressures and relative permeabilities) for each different zone or material
present in the field. The latter properties are not well understood and are difficult to obtain for
field problems. Site data is usually incomplete because of monetary, safety and regulatory
Imitations. Historical records of pollutant releases are often nonexistent, although such
knowledge should be precisely defined in a model. Sampling limitations often result in situations
where the total mass of contaminants cannot be defined. These limitations are likely to require
approximations to be made even when running a complex model. Certain problems may warrant
the use of alternative simplified models.
KOPT/OILENS and TSGPLUME have an advantage in some of these areas; the models
execute rapidly on small computers, require little memory and are designed to be run easily. The
advantage is that they are based on semi-analytic approaches, which do not require discretization
of the domain nor iterative solution of the non-linear governing equations. These advantages are
achieved at the cost of flexibility in accommodating heterogeneities and other phenomena Clear
recognition should be made that for the sake of efficiency and robustness, accuracy and/or the
ability to simulate various situations Is being given up. At some point there is a limit to the
phenomena that can be treated In a simplified context; beyond that limit, the complex models
must be used.
Assumptions of the Model
The following paragraphs describe the conceptual model upon which KOPT/OILENS and
TSGPLUME are based. This discussion is intended to give a dear understanding of the
assumptions and limitations of the models.
The flow system is idealized as consisting of a circular source region overlying an aquifer
at specified depth. Although flow in the unsaturated zone is three-dimensional, the KOPT model
treats transport through the unsaturated zone as one-dimensional. Lateral spreading of
contaminants by capillary forces is neglected, as is spreading due to heterogeneity, since the soil
Is assumed to be of uniform composition. For situations where the contaminants are applied over
relatively large areas, the flow becomes nearly one-dimensional in the center. For contaminant
sources that are of small area! extent the lateral transport of contaminants may be significant
and the assumption of one-dimensional flow Is less applicable. In some sense, however, one-
dimensional modeling leads to a conservative model as all of the pollutant is assumed to move
downward. In reality, some may be left behind due to entrapment by layering or lateral
spreading.
The spill or release of the L-NAPL phase may be simulated in three ways. First is a
release of a known L-NAPL flux for a specified duration. The release occurs at the ground
surface. Based on an approximate capillary suction relationship, some of the L-NAPL may run
off at the surface if the flux exceeds the maximum effective L-NAPL conductivity. Second, a
235
-------
constant depth of ponded L-NAPL, for a known duration, may also be specified. This case
represents a slowly leaking tank, or a leaking tank within an embankment Lastly, a known
volume of L-NAPL may be placed over a specified depth of the soil. This last scenario represents
either a land treatment operation or a landfill containing a known amount of contaminants at the
beginning of the simulation.
L-NAPL phase transport occurs within the soil which contains a constant water saturation
(saturation is defined as the per cent of the pore space filled by a given fluid), corresponding to
the average annual recharge rate, or some other water saturation. The temporal effects of
cBmate: temperature, rainfall, relative humidity, and wind speed are neglected, as average values
of these parameters are used. The average (annual) recharge rate is used to represent the
actual time dependent recharge. Justification of this approach comes from the fact that the soil
moisture profile remains relatively uniform except near the surface. Much data is required to
simulate the time record of rainfall events to develop the non-uniform soD moisture profile. The
effort Involved is not warranted in light of the assumptions used in developing the models.
In accordance with common soil science practice, the effect of the air flow on the L-NAPL
phase transport is neglected. The effect of the presence of the water and air phases on the L-
NAPL phase transport is included by the usage of a non-hysteretic three-phase relative
permeability model. This model is a reasonable approximation of the pore-scale phenomena
occurring in three-phase flow, but the actual nature of these relationships is a major cause of
uncertainty in this and most other multiphase flow models. The model uses measured properties
of the soil (capillary pressure curve parameters) to approximate the relative permeability. The
model does not include transport in fractures or macropores.
Efficiency is achieved in running the model primarily because the gradients of the
capillary pressure are neglected. This causes the governing equations to become hyperbolic
equations, which can be solved by the generalized method of characteristics (Smoller, 1983).
One major effect of this assumption on the simulation results is that the leading edge of the L-
NAPL moving into the son is idealized as a sharp front (Figure 1). Some experimental results
show soil moisture profiles which have nearly sharp fronts. Reible and Illangasekare (1989) show
Infiltration results for an L-NAPL phase where the fronts are nearly sharp. For cases where the
front truly spreads, Charbeneau (1984). for example, presents a theoretical proof that the mean
displacement speed of the sharp and true fronts is the same. Smith (1983) presents a numerical
result for water flow showing that a numerical solution of Richards equation was tracked by a
sharp front solution.
Since the capillary gradient has a dramatic impact on the infiltration capacity of the soil,
an approximate model (the Green-Ampt model) is used to estimate the infiltration capacity during
the application of the L-NAPL phase. This gives the proper flux in the soil, given a flux or
constant head ponding condition at the surface.
H a large enough volume is supplied, the L-NAPL reaches the water table. Typically this
occurs in a relatively short time for L-NAPLs with high mobilities, like gasoline. If sufficient head
is available, the water table is displaced downward, lateral spreading begins, and the OILENS
portion of the model is triggered. OILENS is based on three major approximations. First the L-
NAPL spreading is purely radial, which Implies that the slope of the regional groundwater table is
small enough to be unimportant for the lens motion. The second major assumption is that the
thickness is determined by buoyancy only (Ghyben-Herzberg relations). Thirdly, the shape of the
236
-------
Depth
Figure 1 Idealzed sharp front profile
Groundwater
Flow
Figure 2 Idealized lens profile during
hydrocarbon tens growth
Hydrocarbon Trapped
ta the UncaturaUd
Zone
ten* Daring
Lena Attor Source
Hydrocarbon Trapped Below
ttM Water Table
Groundwatar
Flow
Figure 3 Idealized lens profile during
hydrocarbon lens decay
237
-------
lens is given by the Duputt assumptions, where flow is assumed horizontal and the gradient is
approximated by the change in head over a horizontal distance. The three major assumptions
lead to a very efficient formulation of the model, which is reflected in its low computational
requirements.
Typical results from OILENS include the following features. The lens thickness increases
(Figure 2) during the initial phase of spreading. Initially, the L-NAPL enters the lens faster than Its
volume is spread radially, hence the increase in lens height Later, the lens thins while continuing
to spread laterally. Residual hydrocarbon is left both above and below the actively spreading lens
during this period (indicated in Figure 3 as the lens after source control).
In KOPT/OILENS the L-NAPL is treated as a two-component mixture. The L-NAPL itself
is assumed to be soluble in water and sorbing. Due to the effects of the recharge water and
contact with the ground water, the L-NAPL may be dissolved. This may be significant for highly
soluble L-NAPL phases. The L-NAPL's transport properties (density, viscosity, capillary
pressure, relative permeability), however, are assumed to be unchanging. The second
component is a chemical constituent which can partition between the L-NAPL phase, water phase
and the soil, and can volatilize. This constituent of the L-NAPL is considered the primary
contaminant of interest.
A kinematic approach is used within KOPT/OILENS for the dissolved chemical transport,
which results in a model that neglects dispersion. The chemical motion is assumed to be due to
the advection of water and L-NAPL only. The chemical, which is the second component of the L-
NAPL phase discussed above, is assumed to partition between the NAPL, water and soil
according to equilibrium, linear partitioning relationships. Non-equilibrium and diffusion-limited
partitioning are neglected. Chemical mass flux into the aquifer comes from recharge water being
contaminated by contact with the lens and from dissolution occurring as groundwater flows under
the lens. The concentration of the chemical in the aquifer is limited by its water solubility.
Volatilization is treated as diffusion from the top most part of the contaminated zone. The only
vapor phase flux that is considered is that due to Fickian diffusion. The volatilization of the
chemical from the lens is neglected.
The hydrocarbon and chemical contaminant can biodegrade within the lens, based on
oxygen supply from the ground water. Both can degrade, based on their stoichiometric oxygen
consumption. Degradation is neglected in KOPT, i.e., unsaturated zone transport, and reaeration
of the groundwater from the unsaturated zone is neglected. Degradation in the aqueous phase is
modeled in TSGPLUME using first-order decay.
The aquifer transport of the dissolved contaminant is simulated by using a two-
dimensional vertically averaged analytic solution of the advection-dispersion equation. The
boundary conditions are placed at the down gradient edge of the lens and take the form of a
Gaussian distribution with the peak directly down gradient of the center of the lens. The peak
concentration of the Gaussian distribution adjusts through time so that the simulated mass flux
from the lens equals that into the aquifer. The width of the Gaussian distribution remains
constant and is taken so that four standard deviations are equal to the ultimate diameter of the
lens. Although the size of the lens varies with time, a constant representative lens size is used
for the aquifer source condition. Note that in many cases the lens reaches its maximum size
rather rapidly compared with the transport in the aquifer, so that the use of the maximum lens
size will not introduce large errors.
238
-------
The depth of penetration of the plume is calculated from vertical dispersion beneath the
lens plus the advection flow due to infiltration through the lens, following the approach outlined in
the background documents for the EPACML model (USEPA, 1988). If the calculated penetration
depth exceeds the aquifer thickness, then the plume fully penetrates the aquifer and the model
allows for dilution of the plume by diffuse recharge. If the penetration depth is less than the
aquifer thickness, then the plume thickness is taken as the penetration depth. The thickness
remains constant and the plume is not diluted by diffuse recharge, representing the case where
recharge simply pushes the plume deeper into the aquifer. The analytical solution for the plume
concentration at any time and location is found through use of Romberg integration.
Example Simulation
Table 1 contains the parameters for an example which will be discussed subsequently.
Since there are large uncertainties associated with parameter values for subsurface models, the
approach will be to Investigate the sensitivity of the simulated results to parameter variations.
First, a fairly detailed set of results from the models will be presented for a baseline case. Due to
space limitations, parameter uncertainty will be illustrated with TSGPLUME results only.
The base case consists of a gasoline release from an above ground storage tank which
fails, but the product is retained within a surrounding dike. The fate of xylene, which is assumed
to compose approximately 1% of the mass, is of interest In the base case, the radius of the
source region is 4.0 meters. The gasoline is assumed ponded at an average depth of 1 cm for a
period of 3 days. The water table is 10 meters below the land surface. The unsaturated zone
and aquifer are composed of a dean sand.
KOPT/OILENS predicts that the 3-day duration of the release results in a total of 69.9
cubic meters of gasoline entering the subsurface. The KOPT/OILENS simulation of 1000 days,
with a 5-day maximum time step, required approximately 1 minute and 49 seconds of time to run
on an AST Premium 386/16 PC with an Intel math co-processor and disk caching. The
TSGPLUME simulation with 4 receptor well locations and 120 output times required
approximately 3 minutes and 17 seconds.
Figure 4 shows total liquid profiles for several times prior to the gasoline front reaching
the water table. These profiles illustrate the percent of the pore space filled by the liquids
(saturation) as a function of depth for given times. The water flux of 0.001 m/d, representing the
average annual infiltration rale, results in 32.6% of the pore space being filled by water. The
remaining portion of the pore space is available for the gasoline. At 3 days, the end of the
release, Figure 4 shows that the gasoline occupies 55.7% of the pore space from the surface to a
depth of 6.18 meters. After the end of the release, the saturations decrease with time throughout
the profile, as indicated for days 4 and 5. During the entire simulation, the gasoline flux
decreases with time. The cumulative gasoline flux into the lens (Figure 7) shows slow drainage
of the gasoline into the lens over the simulation time.
After 5.65 days, the lens begins forming at the water table. Figures 5 and 6 illustrate two
configurations of the lens, corresponding to the schematic representations in Figures 2 and 3. At
7.5 days (Figure 5) the lens height is still increasing. By 50 days (Figure 6) the lens is thinning
while continuing to spread radially. Figure 6 shows the extent of the gasoline that is trapped at
residual saturation, after the lens has thinned. Figure 7 shows the gasoline inflow and radius as a
239
-------
-10
0 0
0.5
Total Liquid Saturation
(% Pore Space)
1 0
-8
-9--
-10--
-11--
-12-
0
Variably Saturated with Gasoline
Actively Spreading Gasoline
10 15
Radius (meters)
20
Figure 4 Total liquid saturation profiles
in the unsaturated zone
Rgure 5 Gasoline lens profile at 7.5 days
-8
-9--
I
-11-J-
-12-
Variably Saturated with Gasoline
/*'
Resfdualfy Saturated with -
Gasoline @ 5%
Actively Spreading
Reskfualfy Saturated Gasoline
with Gasoline @ 15%
10 15
Radius (meters)
Figure 6 Gasoline lens profile at 50 days
20
Gasoline Radius
Effective Xylene Radius 6Q|.
Inflow
500
Time (days)
•100
1000
Rgure 7 Gasoline radius and volume inflow
with time
240
-------
function of time. Note that the maximum extent of the lens is largely attained within the first 200
days of the simulation.
Table 1 Parameters for Example Simulation
SoU and Aquifer Characteristics
Saturated horizontal hydraulic* conductivity (water)
Saturated vertical hydraulic conductivity (water)
Pore size distribution (Brooks and Corey)
Residual water saturation
Thickness of capillary fringe
Porosity
Bulk density
Longitudinal dispersrvfty
Transverse dispersMty
Vertical dispersMty
Qroundwater seepage velocity
Average water infiltration rate
Aquifer thickness
Penetration Depth
Hydrocarbon characteristics
Density
Dynamic viscosity
Residual oil saturation above tons
Residual oil saturation below tons
Hydrocarbon solubility
Surface tension
Contaminant characteristics
Initial concentration in gasoline
Hydrocarbon-water partition coefficient
Soil-water partition coefficient
Contaminant solubility in water
Contaminant theoretical oxygen demand
5.0 m/d
1.0 m/d
1.0
0.10
0.10m
0.40
1.65g/cc
10m
2m
0.10m
0.25 m/d
0.001 m/d
5m
2.8m
0.70 g/cc
0.30 cp
0.05
0.15
200mg/l
30 dyne/cm
7000 mg/l
100
2.0 I/kg
175mg/1
0.032
The effective tons radius for the contaminant (xytone) is not necessarily the same as the
tons radius. This Is due to the partitioning of the contaminant between the L-NAPL, water and
sofl. In the tons, motion is assumed to be due to the L-NAPL only, so that there Is an effective
retardation of the contaminant caused by Its partitioning to the water and soil. The amount of
retardation depends on the L-NAPL/water and water/sod partition coefficients. Figure 7 shows
the effective contaminant radius for the baseline case.
241
-------
The dissolution rate of xylene into the aquifer, caused both by contact with the flowing
groundwater and infiltrating water is shown in Figure 8. Initially, the rate of dissolution is low, due
largely to the small radius of contact between the lens and the aquifer. As the size of the lens
increases so does the dissolution rate. Later, as the water leaches the xylene out of the lens, the
rate drops. Figure 9 shows the mass of xylene in the lens as a function of time. The mass of
xylene contained within the lens increases for approximately the first 100 days, since the mass
flux out of the lens is smaller that the xylene loading. As the lens increases in size, the mass in
the lens drops due to leaching. Since mass enters the lens throughout the 1000 days depicted,
the sum of the two curves in Figure 9 is not a constant
Xylene concentration histories from TSGPLUME at receptor wells located 50,100, 150,
and 200 meters down gradient along the centeriine of the plume are shown in Figure 10. As
expected, dispersion decreases the peak concentrations as the distance from the source
increases. At 50 meters the concentration history shows a rapid rise in concentration, followed by
a long tailing at tow concentration. This effect is due to the tailing in the mass flux input to the
aquifer (Figure 8), which occurs even though equilibrium linear partitioning is assumed between
the phases. In this example, no degradation or chemical attenuation of the xylene is assumed to
occur, resulting in the long tailing of the contaminant at each location. The observed tailing is not
due to non-equilibrium in partitioning, rather it is caused by the distribution of mass flux to the
aquifer. Histories at the down gradient distances also show tailing, but dispersion reduces the
sharpness of the rising concentration.
Parameter Sensitivity
Invariably, input parameters for subsurface models are uncertain. In this section, the
effects of selected input parameter variation on exposures are assessed through the model
results.
The lens source width is a parameter for TSGPLUME which must be taken as constant,
despite the fact that the width varies as the lens enlarges. Figure 11 shows the effect on the
estimated concentrations of using the same mass flux but with half the ultimate lens size. There
is a slight increase in peak concentrations due to the narrower plume for the baseline case (see
Figure 10). At 50 meters the peak concentration increases from 12.0 to 14.0 mg/l. Further down
gradient, however, the effect of source size becomes negligible, as at 200 m the peak only
increases from 2.97 to 3.07 mg/l.
Often the loading pattern of contaminants is a major source of uncertainty. By varying
the duration of the release so that the gasoline volume varies from 49.5 to 90.0 cubic meters, the
peak concentrations at 50 meters increase as illustrated in Figure 12. Increasing the volume of
gasoline introduced into the subsurface increases the peak concentrations. The arrival times
are unaffected in this example, because the gasoline is applied in the same way In all three
cases. Since the gasoline volume increases the aquifer source radius, there may be some
tendency for concentration to be reduced near the source. In this example, however, the latter
effect is unimportant relative to the increased volume.
The L-NAPL/water partition coefficient, K , plays a major roie in determining the mass flux
of contaminant into the aquifer (Figure 13). °As the coefficient increases, the amount of
contaminant entering the water phase in a given time decreases, since the contaminant tends to
242
-------
0.4
250
0.0
2000
Time (days)
4000
Xytene Mass Dissolved..
5001000
Time (days)
Rgure 8 Xytene mass flux into aquifer
Rgure 9 Xytene mass In lens with time
15--
15--
0
5000 10000
Time (days)
5000 10000
Time (days)
Rgure 10 Estimated concentration histories
for receptor wefts at 50.100,
150, and 200 meters down gradient
Rgure 11 Source size impact on estimated
concentration history at 50
meters
243
-------
IS-
8
15--
5000 10000
Time (days)
Rgure 12 Gasoline loading impact on
estimated concentration
history at 50 meters
5000
Time (days)
Rgure 13 Gasoline/water partition coefficient
impact on estimated concentration
history at 50 meters
15--
SO'OO 10600 15000 20000 25600
Time (days)
Rgure 14 Hydraulic conductivity impact on estimated
concentration history at 50 meters
244
-------
reside within the gasoline. With enough time, however, the same mass of contaminant will leach
into the groundwater. Thus the duration of the mass influx and the time to peak increase as the
contaminant becomes more hydrophobic (Table 2). The effective size of the gasoline lens also
increases with K , as more of the dissolved contaminant flows radially in the gasoline phase.
Peak concentrations decrease, with increasing K , since the mass enters the aquifer over a
longer period of time at lower mass flux. As indicated by Figures 10 and 11, the effect of
increased radius also contributes to the lowering of the peak concentration with increasing K .
Table 2 L-NAPUWater Partition Coefficient Effects
L-NAPL7 Effective Influx Peak Concentration
Water Xytene Duration at 50 meters
Partition Lens Radius
Coefficient (meters) (days) (mg/l)
50 14.2 1910 12.7
100 16.1 3060 12.0
200 16.9 5015 9.1
The effect of changing the aquifer hydraulic conductivity by orders of magnitude is
illustrated in Figure 14. For comparison purposes, the hydraulic conductivity of the unsaturated
zone is unchanged, so that the gasoline arrives at the water table at the same time in all cases.
The rate at which the gasoline spreads, the mass flux of xylene to the aquifer, the peak
concentration, and the advective travel time in the aquifer all strongly depend on the aquifer's
hydraulic conductivity (table 3). Figure 14 shows that for the high hydraulic conductivity case (K -
50.0 m/d), the peak arrives fastest due to the highest seepage velocity.
The height and shape of the concentration histories depend on the distribution of xylene
mass flux to the aquifer and the water flux under the lens. The xylene entering the aquifer mixes
with the groundwater in a mixing zone to determine the aquifer concentration under the lens. The
water volume flux through the mixing zone depends on its size, which is determined by the
effective radius of the source and the penetration depth, and the seepage velocity. The
combination of mass flux from the lens, seepage velocity, and mixing zone size for the high
hydraulic conductivity case results in relatively low initial concentrations, compared with the other
cases. At the 50 meter receptor well location, the peak concentration is lower than that for the
base case (K » 5 m/d). The concentration history in the high hydraulic conductivity case is
narrow, since the xylene enters the aquifer over a short time period. The duration is related to the
seepage velocity.
At the other extreme, the low conductivity case (K - 0.5 m/d) gives a low mass flux from
the gasoline lens over a long time period, resulting in a wide xylene concentration history. The
peak concentration is relatively high, since the mass flux from the lens enters slowly moving
groundwater.
245
-------
Table 3 Hydraulic Conductivity Effects
Horizontal Effective Water Maximum Peak Approx
Hydraulic Xylerw Flux Mass Cone. Time of
Conductivity Radius Under Flux at 50m Peak
Legs
(m/d) (meters) (m /d) (kg/d) (mg/l) (days)
0.5 16.6 1.66 0.071 8.90 13700
5.0 16.1 9.02 0.306 12.0 1700
50.0 14.2 68.4 1.980 10.8 200
Conclusions
A methodology has been presented for estimating contaminant exposures resulting from
spills of hydrocarbons (L-NAPLs). The KOPT/OILENS model approximates the transport of
hydrocarbon and a dissolved groundwater pollutant of interest by kinematic approaches, coupled
with approximate dynamic techniques when necessary. Transport along the water table Is
simulated by the Qhyben-Herzberg-Dupuit assumptions. Transport within the aquifer is
simulated by an analytic approach for two-dimensional vertically averaged solute transport,
TSGPLUME. The simulation results show tailing of concentrations in the aquifer, because of the
distribution of mass flux to the aquifer. The tailing occurs even though equilibrium partitioning is
assumed between all phases.
Sensitivity of estimated exposure concentrations was investigated using the models
described. The greatest effect was found to be caused by varying the seepage velocity (hydraulic
conductivity) of the aquifer. Peak concentration, arrival time and width of the concentration
distribution are all dramatically impacted. Spill volume primarily affects the peak concentrations,
while the L-NAPL/water partition coefficient affects both peak concentration and peak arrival
times. Several of the these results involve the effective contaminant radius, which can affect the
peak concentration. As the distance to the receptor well increases, this effect becomes
insignificant. Thus the assumption of constant size of the lens is relatively unimportant for
TSGPLUME simulations. The results of these limited comparisons suggest that the hydraulic
conductivity dominates the model results.
Acknowledgement
Although this work was partially funded by the United States Environmental Protection
Agency through the participation of the first author, It has not been subjected to the agency's peer
review process, therefore no official endorsement should be inferred. The mention of trade
names does not imply official endorsement.
Bibliography
Charbeneau, RJ. 1984. Kinematic Models for Soil Moisture and Solute Transport. Water
Resour. Res., 20:6, pp. 699-706.
246
-------
Charbeneau, R.J. and W.M. Johnson. 1990. A Transient Source Gaussian Plume Groundwater
Assessment Model. Submitted for publication in ASCEJ. Hydr. Engrg.
Charbeneau, R.J., S.R. Shultz and J.W. Weaver. 1990. Hydrocarbon Source Term
Characterization for Groundwater Models. Submitted for publication in ASCE J. Hydr.
Engrg.
Kuppusamy, T., J. Scheng, J.C. Parker and RJ. Lenhard. 1987. Finite-Element Analysis of
Multiphase Immiscible Row Through Soils. Water Resour. Res., 23:4, pp. 625-631.
Reibte, D.D. and T.H. Illangasekare. 1989. Subsurface Processes of Nonaqueous Phase
Contaminants. In Intermedia Pollutant Transport Plenum Publishing Corp. pp. 237-254.
Smith, R.E. 1983. Approximate Soil Water Movement by Kinematic Characteristics. So/7 Sci.
See. Am. J., 47, pp. 3-8.
Smoller, J. 1983. Shock Waves and Reaction - Diffusion Equations. Springer-Vertag. New York,
pp. 265-305.
U.S. Environmental Protection Agency. 1988. Background Document on Subsurface Fate and
Transport Model. Office of Solid Waste.
Weaver, J.W. and R.J. Charbeneau. 1989. A Kinematic Model of Oil Drainage in Soils.
Submitted for publication in Water Resour. Res.
247
------- |