SIMULATION OF DNAPL DISTRIBUTION RESULTING FROM
MULTIPLE SOURCES
Michael Fishman (Dynamac Corp., Ada, OK); Joseph Guarnaccia (Ciba
Specialty Chemicals Corp., Toms River, NJ); Lynn Wood (US EPA/NRMRL,
Ada, OK); Carl Enfield (US EPA/NRMRL, Cincinnati, OH)
ABSTRACT: A three-dimensional and three-phase (water, NAPL and gas)
numerical simulator, called NAPL, was employed to study the interaction between
DNAPL (PCE) plumes in a variably saturated porous media. Several model
verification tests have been performed, including a series of 2-D laboratory
experiments involving the migration of PCE through a variably saturated,
homogeneous sand. A comparison of the experimental data to the model results
illustrates the effect and importance of fluid entrapment and saturation hysteresis.
The NAPL model was used to simulate a 3-D multi point PCE source
release within a contained test cell at the Groundwater Remediation Field
Laboratory (GRFL) in Dover, Delaware. In this experiment, the migration of PCE
in the unsaturated and saturated zones, under various infiltration scenarios, was
simulated. The modeling of multiple injection points in a homogeneous aquifer
shows that the ultimate distribution of PCE depends on the injection point
locations and the time-varying release rates, and the depth to the water table. In
general, an intermittent, slow, injection rate caused narrow, deeply penetrating
DNAPL plumes. On the other hand, higher injection rates resulted in a wider
horizontal distribution and more interaction between neighboring plumes, thus
creating non-symmetric distributions and an increase in the flow rate and depth of
penetration.
INTRODUCTION
A 3-D NAPL model was developed to investigate the movement of
organic compounds in both homogeneous and heterogeneous porous media.
Particular attention was paid to the development of a sub-model that describes
three-phase hystereric permeability-saturation-pressure relationships, and the
potential entrapment of fluids when they are displaced. Several laboratory
experimental data sets have been compared to simulator predictions, including a
series of 2-D laboratory experiments involving PCE release conducted by the
authors at the R. S. Kerr Laboratory, Ada OK, and described in Fishman et al.
(1998) and Guarnaccia et al. (1997),
The results of modeling runs were used to develop a DNAPL release
strategy for remediation technology demonstrations at the Dover National Test
Site (DNTS), Dover, Delaver, and to predict the DNAPL movement from
injection points under a range of hydrodynamic release conditions. Because the
hydrodynamic properties and spatial variability of these properties strongly
influences the behavior of DNAPL in the subsurface, simulations were done for a
range of hydrodynamic conditions in both homogeneous and heterogeneous
porous media.
-------
The attributes (parameters, processes) that control DNAPL distribution in
subsurface granular porous media include: fluid properties (density, viscosity,
interfacial tension, wettability), soil properties (hydraulic conductivity,
heterogeneity), source conditions (release rate and proximity of multiple sources).
The focus of this paper is on the issue of source conditions', how release rates and
source proximity affect DNAPL distribution in the subsurface.
NUMERICAL MODEL
A numerical model, called NAPL (Guarnaceia et al., 1997), was used to
simulate the experiment described above. The model has the following conceptual
and computational attributes, which are assumed to be relevant to the physical
experiment to be modeled:
1. simultaneous flow of water, NAPL, and gas;
2. the three-phase relative permeability-saturation-capHlary pressure (k-S-P)
relationships are based on fluid phase wettablility considerations and two-phase
data; wettability follows, from most to least, water-NAPL-gas;
3. the three-phase k-S-P model reduces to the appropriate two-phase model when
appropriate;
4. the k-S-P model includes flow-path-history-dependent ftinetionals (hysteresis);
5. the k-S-P model includes a mechanism for fluid entrapment during drainage
(residual emplacement), where the amount entrapped is a function of the
maximum imbibed saturation;
6. the S-P model employs capillary pressure scaling to account for variable porous
medium and fluid properties;
1. at each boundary node, one can specify either a no flow condition (can also be
coupled with a point source or sink 'well' rate for one or more phases), or any one
of the three phase pressures known, or all the primary variables known (i.e., one
pressure and two saturations);
8. a numerical 'peclet criterion* can be employed to ensure that the scale of the
saturation-capillary pressure functional is compatible to the scale of the grid.
The model employs the collocation finite element method to approximate
the system of governing equations spatially, and an implicit finite difference
approximation in time. The non-linear system of governing equations is solved
using a sequential solution algorithm. Details of the numerical methods can be
found in Guamaccia et al. (1997).
NUMERICAL EXPERIMENTS
Model Testing. A two-dimensional artificial aquifer experiment was conducted
to study how DNAPL (PCE) migrates through a variably saturated homogeneous,
isotropic, porous medium. A video image of the experiment was analyzed to
define the DNAPL saturation at the pixel-scale as a function of time. Once
constructed, the image was compared to the solution of the numerical model
designed to simulate the same experiment. When the image was averaged to the
same spatial scale as the model grid, a qualitatively similar solution was observed.
-------
Figure 1 compares the image
of experimental distribution of PCE
and the numerical results at T =
6Qmin (as the PCE is migrating
through the saturated zone). Note
that the solutions are
qualitatively similar, indicating that
the numerical model effectively
captures the physics of the problem
as defined at the scale of a grid cell.
This information can be used in
practice to define appropriate
dissolution and vaporization mass
transfer rates at the grid scale.
Details of the experiment and the
video imaging analysis can be
found in Rshman et al. (1998).
SIMULATION
limns
FIGURE I. Experiment and model simulation.
Effects of Heterogeneity on DNAPL Distribution. It is well known that
DNAPL distribution in the subsurface is very sensitive to heterogeneity in
hydraulic conductivity (see for example Poulsen and Kueper. 1992, and Kueper et
al., 1993). The effect of soil
heterogeneity on PCE
distribution is illustrated in
Figure!. The same PCE
flood experiment was run in
a variably saturated domain
with three hydraulic
conductivity (K)
realizations: a homogeneous
domain (part a), a horizontal
low K lens located below
the water table (part b), a
vertical high K lens located
below the water table (part
c). The low K horizontal
lens exerts a very strong
influence on the PCE
migration pattern. The
inability to penetrate this lens is due to a combination of preferential flow through
high K sand and the fact that the capillary pressure above the lens is less than the
entry pressure for the lens (Figure 2 b). Conversely, the high K vertical lens
enhances vertical PCE migration via the combination of preferential flow and the
low entry pressure associated with the lens (Figure 2 c).
The simulations show that heterogeneity causes the shape and internal
FIGURE 2.PCE saturation distribution.
a-homogcneous media, b-single horizontal
lense. c- single vertical lense.
-------
structure of the DNAPL plume to differ significantly from that predicted using the
homogeneous realization. At the field scale, because of uncertainties in defining
soil properties, and computational issues regarding discretization, DNAPL
modeling becomes one of a sequential screening exercise. Specifically, initial
modeling with a mean homogeneous K-fkld is combined with data generated
from a groundwater quality profile located down-gradient of the source (see
Pardieck and Guarnaccia, 1999). An appropriate 'modeling K-field' is defined
using the groundwater quality profile signature.
Effects of Source Rate, and Multiple Release Locations. In this section we
present results of several 3-D PCE flood numerical experiments aimed at
providing insight into the importance of source area characterization in assessing
DNAPL distribution in the subsurface. Specifically, we consider the variables of
source release rate and the number and proximity of sources. The 3-D domain is
homogeneous, while the source distribution is heterogeneous.
Soil-and Fluid-Phase Parameters, The model parameters, which are presented
in Table 1, were determined in the laboratory or obtained from literature.
TABLE 1. Parameters used in the DNAPL spill simulations.
Soil Properties
porosity - 0.37; hydraulic conductivity - 10 m/day; volume density -1.7 g/cm3.
van Genuchten k-S~P parameters; a,j - 4.00/m; % - 6.00/m; n-6.35;
DNAPL residual saturation as a nonwetting phase - 0.14.
DNAPL properties
viscosity - 0.90 Cp; density -1.626 g/ml; interfacial tension, PCE -water- 47.5
dyne/cm; interfacial tension, PCE- air - 31.74 dyne/cm
Water properties
viscosity - 0-99 Cp; density - 0.99 g/ml; interfacial tension, air-water - 72.75
dyne/cm.
Air properties
viscosity - 0.02 Cp; density -
0.00129 g/ml.
u»
ii
The Flow Domain. The
dimensions of the flow domain
selected for 3-D multi-well
simulations are 200 cm high, 460
cm long, and 300 cm wide. The
domain is discretized into an
irregular grid of 10 by 26 by 24
elements with spacing that varies
from 5.0 by 5.0 by 5.0 cm in
proximity to the potential source
1.1
1.1
W 1
Ml
w*
Wl
NL
Wjj,
WL i
Hi!
usn,
_»U
WL
Ml
C"
l¥ J
f.M.
a
"tj
J U
8.5 f.» U
HGURE3.
U 13 2.5 If 1* 13 JJ
-------
areas to 30.0 by 30.0 by 30.0 cm at the perimeter of the domain, A plan view of
the domain is shown in Figure 3 and a cross section is shown in Figure 4,
Initial, Point Source, and Boundary Conditions. The domain is initially free of
PCE. PCE flood simulations were performed in which PCE was allowed to
infiltrate from as many as twelve point sources located below ground surface, but
above the water table, at a rate between 0.25 L/rnin and 0.9 L/min (total volume
injected at any one well ranges from 5,0 to 9.0 L). The locations of the point
sources are shown in Figure 3, The boundary conditions are no-flow along all
boundaries except the top, where a constant atmospheric pressure is prescribed,
The full experiment was modeled as a sequential series of three sub-
models. Each sub-model solves a part of the overall flow problem:
1. The initial conditions for sub-model 1 were full water saturation. At time>0,
the water table was lowered between 20 and 40 cm and the two-phase system
(water and gas) was allowed to approach steady-state conditions. The final
distribution of water saturation was adopted as the initial conditions for second
sub-model.
2. Using the initial conditions (distribution of water saturation) determined from
sub-model 1, DNAPL was released at a predefined rate from the source(s) until a
predetermined total volume was released.
3. After the DNAPL was released to the formation the spill was allowed to
redistribute for the duration of the five day simulation period.
For multiple simulations involving changing parameters (e.g., source
conditions), this structure allows for restart at known intermediate flow
conditions, and thus, it is a time saving measure.
RESULTS AND DISCUSSION
Effect of Injection Rate on DNAPL Distribution. Figure 4 (cross-section)
shows the solution after a ten minute release of PCE and five days of
redistribution from two
point sources: at the
source on the left, 5.0L
of PCE were injected,
while at the source on
the right 9.0 L were
injected. The results
show that as the release
rate increases the
DNAPL plume tends
to spread more
laterally. The higher
volume injected on the
right and resulted in
deeper penetration into the saturated zone. For the case of equivalent volume
4.6m
FIGURE
-------
injected at different rates (not shown), the slower rate resulted in a plume which
had a smaller effective radius and greater penetration. This result is consistent
with that described in Pouisen and Kueper (1992).
The Multi Wells Simulations. A 3-D simulation was performed in which PCE
was allowed to infiltrate from twelve point sources located in the vadose zone (see
Figure 3, plan view). Each
well injected between 5.0 and
9.0 L of PCE in
approximately 10 minutes (the L
volume for each well is I"
shown in figure 3). After
injection the PCE was
allowed to redistribute for five
days simulation time.
Below the water tabie
the extent of the PCE plume
reached the bottom and in
three areas of PCE release
(W_2;W_liandW_l2)it
accumulated as a 'pool' of FIGURE I i&wi«uoii 1)
(Figures 5 and 6). Pools were created due to the proximity of the wells (20cm) to
the impervious walls. In the rest of the area, with a release amount of 9.0 L, the
W 4
W 6
W 9
W 10
W 12
FIGURE 6- Distribution ofPCE saturation at crosi section W_4 - WL12
plume may reach the bottom but no pool will be created.
Figure 7 (a) presents the simulation when distance between wells is 1-Om
and less than spreading area (1.4m) shown in part b of Figure 7. In this case we
-------
have more interaction between neighboring plumes, thus creating non-symmetric
distributions and an increase in the flow rate and depth of penetration. The free
phase of DNAPL will accumulate at the bottom of the cell.
W_12Q = 9.0L
FIGURE 7. PCE saturation distribution.
a - distance between wells is 1.0m. b » distance between wells b 1.4m,
CONCLUSIONS
Three important conclusions have been drawn from the modeling experiments:
a) The depth and the width of spreading PCE are influenced by the release
rate and the amount of release- A smaller rate of PCE release gives deeper
penetration and a smaller radius of influence than from an instantaneous spill.
Our laboratory and modeling experiments confirmed this conclusion;
b) The proximity of multiple sources on DNAPL distribution has an
important effect on overall DNAPL distribution, in that coalescence of individual
plumes will result in deeper penetration into the saturated zone;
c) Comparison to both laboratory and field DNAPL flood and
redistribution experiments indicates that this and other NAPL models are capable
of capturing the physics of the problem. However, at the field scale, uncertainty in
the most important 'driving' parameters, hydraulic conductivity and source
conditions (location and release rate history), as well as, computational issues
(discretization limitations), requires that DNAPL modeling be combined with
groundwater quality profiling data. The profiling data provides an indication of
DNAPL location, and the model is used to obtain a meaningful realization using
'effective" parameters for screening and remediation purposes.
REFERENCES
Broholm K., S.Feenstra, J. A. Cherry. 1999. "Solvent Release into a Sandy
-------
Aquifer. I. Overview and Dissolution Behavior." Environ. Set. Technol.
Fishman M., J. Guarnaccia, C. Enfieid, L- Wood. 1998."DNAPL infiltration and
Distribution in Subsurface; 2D Experiment and Modeling Approach."
Nonaqueous Phase Liquids. Remediation of Chlorinated and Recalcitrant
Compounds. C 1-2:37-42.
Guamaccia, J. F., G. F. Finder, and M. Fishman. 1997, "NAPL: Simulator
Documentation" EPA 600 R-97/102, http://www.epa.gov/ada/napl_sim.html.
Kueper, B H,, D. Redman, R. C, Starr, S. Reitsma, and M. Man 1993. "A Field
Experiment to Study the Behavior of Tetraehloroethylene Below the Water Table;
Spatial Distribution of Residual and Pooled DNAPL." Ground Water.31(5):756-
766.
Pardieck, D. and J. Guarnaccia. 1999. "Natural Attenuation of Groundwater
Plume Source Zones: A Definition." Journal of Soil Contamination, 8(l):9-15>
Poulsen, M. M. and B. H. Kueper. 1992 "A Field Experiment to Study the
Behavior of Tetraehloroethylene in Umaturated Porous Media." Environ. Set.
rec/m0U<5(5);889-895.
ACKNOWLEDGMENTS
The work upon which this paper is based was supported by US EPA through its
Office of Research and Development and based on funding from SERDP CU 368.
It has not been subjected to Agency review and therefore does not necessary
reflect the view$ of the Agency and no official endorsement should be inferred.
-------
NKM1L-ADA-00115
1. REPORT NO.
.PA/600/A-01/1Q3
TECHNICAL REPORT DATA
2.
4. TITLE AMD SUBTITLE
Simulation of DNAPL Distribution Resulting From Multiple Sources
7. AUTHOR (S)
*• Michael Fishman, 2' Joseph Guarnaccia, '• Lynn Wood, *• Carl Enfield
9. PERFORMING ORGANIZATION NAME AND ADDRESS
1 Dynamac Corporation, Ada, OK
2' Ciba Specialty Chemicals Corporation, Toms River, HJ
'• NRMRL, ORD, US EPA, Ada, OK
4- NRMRL, ORD, US EPA, Cincinnati, OH
12. SPONSORING AGENCY NAME AND ADDRESS
US Environmental Protection Agency
Office of Research and Development
National Risk Management Research Laboratory
Subsurface Protection and Remediation Division
Ada, OK 74820
3.
5. REPORT DATE
6. PERFORMING ORGANIZATION CODE A
8. PERFORMING ORGANIZATION REPORT NO.
NRMRL-Ada-001 1 5
10. PROGRAM ELEMENT NO.
1538
11. CONTRACT/GRANT NO. 68-C4-0031
13. TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
Project Officer: Lynn Wood, 580/436-8552
16. ABSTRACT
A three-dimensional and three-phase (water, NAPL and gas) numerical simulator, called NAPL, was employed
to study the interaction between DNAPL (PCE) plumes in a variably saturated porous media. Several model
verification tests have been performed, including a series of 2-D laboratory experiments involving the
migration of PCE through a variably saturated, homogeneous sand. A comparison of the experimental data to the
model results illustrates the effect and importance of fluid entrapment and saturation hysteresis.
The NAPL model was used to simulate a 3-D ntulti point PCE source release within a contained test cell at
the Qroundwater Remediation Field Laboratory (GRFL) in Dover, Delaware. In this experiment, the migration of
PCE in the unsaturated and saturated zones, under various infiltration scenarios, was simulated. The modeling
of multiple injection points in a homogeneous aquifer shows that the ultimate distribution of PCE depends on
the injection point locations and the time-varying release rates, and the depth to the water table. In
general, an intermittent, slow, injection rate caused narrow, deeply penetrating DNAPL plumes. On the other
hand, higher injection rates resulted in a wider horizontal distribution and more interaction between
neighboring plumes, thus creating non-symmetric distributions and an increase in the flow rate and depth of
penetration.
17.
A. DESCRIPTORS
18. DISTRIBUTION STATEMENT
Release to the Public
KEY WORDS AND DOCUMENT ANALXSIS
B. IDENTIFIERS/OPEN ENDED TERMS
19. SECURITY CLASS (THIS REPORT)
Unclassified
20. SECURITY CLASS {THIS PAGE)
Unclassified
C. COSATI FIELD, GROUP
21 . NO . OF PAGES
8
22. PRICE
EPA FORM 2220-1 (REV.4-77)
PREVIOUS EDITION IS OBSOLETE
------- |