FLOW BEHAVIOR WITHIN FRACTURED

MEDIA

K. KOSTARELOS*, J. WADE**, J. TAIWO**, S. YOON**, D. DERMATAS***

*
Civil & Envrionmental Engineering, 75 Kallipoleos Street, University of Cyprus,
Nicosia, CY-2003 * corresponding author, dino@ucy.ac.cy +357 2289 2214
**
Civil Engineering, Polytechnic University, 6 MetroTech Center, Brooklyn, NY,
USA 11201
***
Waste Management Authority, .... .., N. Plastira 6, Komotini,
69100, Greece.

SUMMARY: A transparent physical model with fractures of 3 aperture sizes (0.5, 0.2, and 0.1
mm) was used to demonstrate the flow behavior in fractured media. The fractures were initially
filled with a NAPL (octanol) and two fluids were used to displace the NAPL. The experiments
show that a viscosifer that increased the endpoint mobility ratio M resulted in a significant
improvement of sweep efficiency and a resulting increase in NAPL displacement. These results
were expected based on prior work in the petroleum industry for improved oil recovery, but to
date have not been implemented for environmental remediation purposes. Computer simulations
(UTCHEM) explored the sensitivity of flow within fractured media to mobility ratio and to
gravity number, concluding that the mobility ratio is a key consideration in both improving the
delivery of remedial fluids in a fractured rock environment and when attempting to displace
NAPLs from fractured rock. Density (i.e., buoyancy) effects are much less important in fracture
flow when delivering remedial fluids.

1. INTRODUCTION

Recovery of nonaqueous phase liquids (NAPLs) from the subsurface is problematic, and a
serious concern for the environmental community. Many treatment approaches are focused on
insitu approaches, either insitu treatment or recovery of NAPLs, that rely on injecting remedial
fluids into the subsurface. To date, most research has focused on recovery from porous media
primarily sands and sandy soilsand little focus on fractured media. Many practitioners
perceive fractured rock as a more difficult subsurface environment compared to sandy aquifers
and are reluctant to apply new, innovative technologies. Consequently, an understanding of
flow behavior within fractured media, and the ability to numerically model flow in fractured
rock, is of special concern for environmental practitioners and the focus of this paper.
Many NAPL sites are within bedrock or are underlain by bedrock. Of the DNAPL sites being
managed by the US Department of Energy (DOE), about 35% have been documented to have
contamination within fractured bedrock (DOE, 1998). The US EPAs Technology Innovation
Program (TIP, formerly TIO) developed a database containing profiles of fractured rock sites
(USEPA, 2008). Currently, 126 site profiles from 24 states and two foreign countries are
included in the database and they represent the single best source of information about
contamination in bedrock.NAPL movement (mobilization and migration) within fractured,
porous media
When discussing the movement of contaminants classed as NAPLs, it is useful to distinguish
between the mobilization of the NAPL itself, which typically occurs due to a gravity gradient,
and the migration of the dissolved NAPL components within the aqueous phase, which can
occur due to a variety of gradients (e.g., concentration, advective, density, etc.) (Kostarelos,
1998). When a DNAPL moves through fractured porous media, it will enter the fracture first
because it offers the least resistance to flow due to a lower entry pressure as compared to the
surrounding porous media (Esposito and Thomson, 1999; Charbeneau, 2000; Reynolds and
Kueper, 2001; Reynolds and Kueper, 2004). In the case when fractures are of a different size,
the DNAPL movement will prefer fractures with larger apertures than smaller ones due to the
lower entry pressure of the larger aperture fracture. The mobilization of the DNAPL will depend
on the permeability of the medium, the physical and chemical properties of the DNAPL and the
porous medium, and the size of the release (Reynolds and Kueper, 2004).
In many field situations there is a pathway through a lowpermeability matrix, such as a
fractures, root holes, dessication cracks, or unsealed boreholes. Their existence is only realized
when dissolved contaminants are detected in confined aquifers. The contamination arises from a
DNAPL source within the subsurface that should be removed in order to eliminate the
contaminant source. Thus, DNAPL movement within fractured media has been the subject of
study in recent years. Many researchers have found that DNAPLs will find fractures that
allow for mobilization through the otherwise low permeability deposits (Esposito and Thomson,
1999; Parker et al., 2004; Reynolds and Kueper, 2004). Vertical fractures in a low permeability
matrix increases both the DNAPL mobilization and migration of dissolved phases, resulting in
the contamination spreading into confined aquifers within a much shorter time period than
expected (Reynolds and Kueper, 2004). This understanding leads one to reasonably conclude
that removal of the DNAPL effectively eliminates the contamination source and thereby
eliminates groundwater cleanup times.

1.1. Improving Sweep Efficiency within Fractured Rock

The use of viscosifiers during surfactant flooding improves mobility control and the overall
sweep efficiency (Manning et al., 1983; Lake, 1989). For example, xanthan gum serves as a
viscosifier and can improve the sweep efficiency of injected remediation fluids in fractured
bedrock. That is, the injected fluids will be less affected by heterogeneities and be distributed
more evenly within the fracture network. Viscosifiers can also be used to create a more
favorable mobility ratio between a displacing fluid and a resident fluid that will result in
improved sweep efficiency (Lake, 1989; Pitts et al., 1993). A favorable mobility ratio, which is
a ratio less than one, can eliminate viscous fingering.
The term fingering is used to describe the bypassing of a resident fluid by a displacing
fluid in a homogeneous medium. This encompasses instabilities caused by both viscous forces
(viscous fingers) and gravity forces (gravity fingers) but does not include bypassing due to
permeability heterogeneities. Viscosifiers have a favorable effect on the front stability between
solutions that are injected (Martel et al., 2004) and resident fluids.
The volumetric sweep efficiency is an important factor that describes the flow of DNAPL
when a visosifier is injected into porous media. Volumetric sweep efficiency is defined as the
ratio of the volume of resident fluid contacted by the displacing fluid to the volume of resident
fluid originally in place (Lake, 1989). The volumetric sweep efficiency can be considered as the
product of vertical sweep efficiency and areal sweep efficiency.
 The vertical sweep efficiency (EI) is the ratio of the cross-sectional area contacted by the
displacing agent and the total crosssectional area. The areal sweep efficiency (EA) is the ratio
of area contacted by the displacing agent and the total area. The areal sweep efficiency (EA) is a
function of the injectionpumping pattern and the injection pumping rate. EA and EI depend on
fluid viscosity and on porous medium heterogeneity. EI is also influenced by fluid density. A
decrease in mobility ratio will improves DNAPL remediation effort by increasing vertical sweep
efficiency and areal sweep efficiency, i.e., the amount of DNAPL contacted by remediation
fluid. Both the vertical (EI) and areal (EA) sweep efficiencies are controlled by the endpoint
mobility ratio (M) between the displacing fluid ( 1 ) and the displaced fluid ( 2 ) (Lake, 1989).
r1

Mobility ratio M = (where 1 denotes displacing fluid) (1)
r2

k ri
mobility of fluid i ri = (where k r denotes endpoint relative permeability (2)
i
Dyes (1954) illustrated areal sweep efficiency with the use of three charts from a scaled physical
model of oil displacements to show the effect of mobility ratio and time on sweep efficiency.
The illustrated data are for three well patterns: (1) a confined five-spot pattern (2) a confined
direct line drive pattern and (3) a staggered line drive pattern. For each well pattern, the
percentage of area swept was plotted against pore volume for varying reciprocal mobility ratio.
Lake (1989) discussed the effect of mobility ratio on the cumulative displaceable pore volume of
the displacing agent required to achieve similar swept areas. At a fixed mobility ratio, EA is
equal to the displaceable pore volumes injected until breakthrough. It was noted that a decrease
in EA with increasing M resulted for similar injected PV and viceversa.
Lake (1989) explains how a displacing agent will behave if it is injected into a two-layer
reservoir that has two different permeabilities. He explained that, if a displacing agent is injected
into the layers of the reservoir at the same time, the overall sweep efficiency of the area will be
poor and the displacing agent will reach the output at two different times due to the permeability
contrast. The fast layer that has M < 1 will be filled up with the injected fluid faster than the
slow layer. However, if the mobility ratio of the injected fluid is decreased by using a
viscosifier, the vertical sweep efficiency EI of the reservoir will improve and the injected fluid
will reach the output with less difference in their travel times compared to the case of a
displacing agent that has the same viscosity as the resident fluid.
Displacement efficiency (ED) is another important parameter that contributes to the recovery
of oil/NAPL in fractured bedrock. It is defined as the ratio of the amount of DNAPL displaced
and the amount of DNAPL contacted by a displacing agent. ED is bounded between 0 and 1. The
rate at which it approaches 1 is strongly affected by the initial conditions, the displacing agent,
and the amount of displacing agent. Other factors that influence ED are the fluid, rock, and fluid-
rock properties. If the displacement is such that the displacing agent will contact all the DNAPL
initially present in the medium, the volumetric sweep efficiency will be unity, and ED will
become the Recovery Efficiency (ER).
In theory, the ultimate displacement efficiency using water or viscosified water is usually
between 50% and 80% of the contacted oil/NAPL due to the presence of residual oil/NAPL and
connate water saturation. Thus, while improved sweep efficiency can be achieved when
attempting to contact NAPL with remedial fluids, methods that intend to recover residual
oil/NAPL must rely on something other than the mobility ratio concept, such as displacing
oil/NAPL with miscible agents or lowering the wateroil interfacial tension.
The justification for the use of a viscosifier to improve sweep efficiency is to reduce the
endpoint mobility ratio and also to enhance front stability between the displacing agent and the
resident fluid in fractured bedrock. Using a polymer solution as a viscosifer, mobility control of
displaced NAPL is improved and volumetric sweep of injected remediation fluids is increased.
When polymer is added to fluids injected into the fracture medium, the amount of displacing
agent that is needed is minimized and the overall time of the remediation will be reduced,
thereby reducing overall project costs.

2. METHODOLOGY AND RESULTS

2.1. Physical model

A physical fracture model was constructed using two Plexiglaspanels that measured 73.0 cm, by
42.9 cm, and 2.54 cm thick. Plexiglas was used for the model because non-aggressive, non-
toxic, immiscible liquids were planned for use in the flow experiments to mimic NAPL/water
behavior. A groove was cut on the inside face of one Plexiglas panel that resulted in a fracture
when the two panels were placed together. The fracture sizes 0.5 cm, 0.2, and 0.1 were cut on
one face of a single Plexiglas panel using a glass cutter. Sand paper was used to smoothen the
fractures walls before a thin coating of vacuum grease was applied to the second Plexiglas
panel surface. When assembled, the grease creates an airtight seal around the fractures. Any
excess grease was removed with ethanol and a cloth. Figure 1 shows the fracture pattern created
on the Plexiglas panel.
The two Plexiglas panels were then screwed together, forming a network of fractures. To
further seal the model, silicon sealant was applied to the four edges and allowed to dry for at
least twenty-four hours. In order to inject and withdraw fluids into the fracture model, four
Swagelok fittings were installed through one of the tank panels. Two fittings provided access
to the top of the fracture network, and two fittings accessed the bottom. Each fitting was
connected to Peforax tubing that could then be connected to either a pump or collection reservoir
as illustrated in Figure 1. The dualpiston pump was, inturn, connected to either of two
reservoirs; one reservoir contained degassed, deionized water while the other contained either
dyed, degassed water or polymer solution, depending on the experiment.
Prior to introducing any water into the fracture model, carbon dioxide gas was used to
displace any air, preventing the formation of trapped air pockets. Then, water was allowed to
flow slowly into the model using a low hydraulic gradient (gravity) that was raised until all the
fractures were saturated. The pump was used to flow water through the tank and check for leaks.

experiments were
Figure 1: Photograph of physical model (left) and dimensions of the physical model and the model
used in simulations. Fractures at the top and bottom are 5 mm in size.
At this time, the pressure and flow rate was measured and recorded. Two flow experiments were
performed: the first used water to displace 1-octanol from the fracture model, while the second
employed a xanthan gum polymer solution to displace the octanol. The dynamic viscosity at
laboratory room temperature (20C) of the resident fluid 1-octanol is about 8.5 cP (Peters, 1992),
whereas the displacing fluid water is about 1 cP and and the viscosified water about 10 cP. After
making the reasonable assumption that the endpoint permeability for the fracture network is
nearly equal for each phase, the resulting endpoint mobility ratio is about 8.5 when the
displacing fluid is water and about 0.9 when it is viscosified water. The experiments were
photographed at small time intervals in order to compare the performance of each fluid (water or
polymer solution) in displacing an immiscible fluid. Pressure and flow measurements were
made throughout each experiment.

2.1.1. General Observations

To illustrate the effect of lowering the endpoint mobility number, a comparison is made between
the photographs of the physical model in Figure 2 and Figure 3. The two cases show similar
degree of NAPL saturation, specifically within the finest fracture to the lower right of the model,
and within the larger fractures at the top. However, the time required to achieve these
comparable states of NAPL displacement are about 60 minutes using water and only 14 minutes
using the viscosified water. The use of a viscosifier improved the sweep efficiency when
displacing 1-octanol within the fractured bedrock model; the octanol was recovered from the
fracture model using less displacing fluid. The time (and displacing fluid volume) required to
displace octanol using a viscosifier is considerably less when compared to water. The
significance for field recovery of NAPLs from fractured rock is that a considerable savings could
be realized when a viscosifier is used rather than plain water to achieve a favourable (less than
unity) endpoint mobility ratio, M.
The efficiency improvement using a viscosifier is twofold. The viscosifier will reduce the
amount of the displacing fluid to displace NAPL in the fractured bedrock. Second, the
viscosifier will reduce the overall time for the NAPL recovery process. This will lead to a
savings in total cost for the NAPL recovery within the bedrock.
Moreover, these results imply that the use of viscosifier along with other remediation
technologies will improve the delivery of remediation fluids within the fractured bedrock. The
viscosifier improves the sweep efficiency of the fracture network, thereby delivering remediation
fluids with less wasted fluid and in shorter time. Thus, even if an environmental practitioners
goal is NOT displacement of the resident NAPL, but rather delivery of a remedial fluid for the
purpose of degrading the NAPL insitu, employing the mobility ratio concept will result in better
contact between the two fluids and as such a shorter remediation time.

Figure 2: Comparison of physical model to simulated model at t=14 minutes. Resident fluid (1-
octanol) in red, displacing fluid (10 cP viscosified water) in green. (Note: 0.7 minutes
of dead time before polymer solution injection into the physical model)
Figure 3: Comparison of physical model to simulated model at t=60 minutes. Resident fluid (1-
octanol) in red, displacing fluid (1 cP water) in green. (Note: 1.3 minutes of dead time
before polymer solution injection into the physical model)

2.2. Numerical model (UTCHEM)

UTCHEM was developed at the University of Texas at Austin and is a 3-D muticomponent,
multiphase, compositional model of chemical flooding processes. It accounts for complex
phase behavior, chemical and physical transformations, and heterogenous porous media
properties, and uses advanced concepts in highorder numerical accuracy and dispersion control
and vector and parallel processing (Delshad et al., 1996; RERPC, 2000).
The UTCHEM simulations were designed by having areas of low and high permeability
matching the physical model in Figure 1. In order to make the model, variable gridblock sizes
were used, and the model was divided into 70 gridblocks wide by 60 gridblocks high. The
gridblcoks ranged in width of 1 mm to 3 mm, depending on the location. Generally, for the
gridblocks between fractures, a 1.5 mm gridblock size was used. In certain horizontal grooves,
two gridblocks were used, a smaller and a larger part. This was done to accommodate areas
where a 1 or 2 mm fracture was at the same horizontal depth as the 5 mm fracture. The resulting
gridblock composition at these depths would be the use of a 1 and 4 mm, or a 2 and 3 mm
gridblock to form the fracture.
In order to simulate the physical experiments, 3 components were injected: water, oil, and
viscosified water. Two extraction wells were placed on the two upper corners, while an
injection well was placed on the lower left corner to model the injection point of water, oil, or
viscosified water. The resulting pressure gradient provided an upward gradient. The well
properties were identical to the physical experiment.
The UTCHEM simulations were setup (not calibrated) and run exactly as the physical
experiments: they consisted of injecting the NAPL into the initially watersaturated fractures and
then running the two trials to displace the NAPL, both water and viscosified water. Output from
UTCHEM was viewed by using MVIEW software. A comparison was made with the results
from the physical model. After the initial runs verified reasonable match with the physical
model, the viscosity was modified for several more runs, and the density of the polymer solution
was changed to run a trial with density less than and greater than water to examine parameter
sensitivity. These results are not discussed here, but will be presented in a future publication.

2.2.1. General Observations

One of the most important things with computer simulations is its ability to accurately depict
actual experiments or field conditions. Usually, this is more difficult under field conditions due
to the higher degree of uncertainty in the field, but the laboratory is easier to simulate since a
physical model can be constructed with little uncertainty. The result is a perfect opportunity to
compare laboratory experiments with computer simulations.
Figure 4: Comparison of physical model to simulated model at time = 2 minutes. Resident fluid
(1-octanol) in red, displacing fluid (10 cP viscosified water) in green.(Note: 0.7 minutes
of dead time before polymer solution injection into the physical model)

First, to compare flow behavior models is visually compared at various times. Both models
shown in Figure 4, which depict the injection of polymer solution into the NAPL saturated
fracture network, the progression of the polymer solution is the same 2 minutes after injection
commenced. A comparison between models presented in the previous section 2.1.1 (Figure 2)
made after 14 minutes of injecting polymer solution, the simulation confirms the difficulty in
displacing the NAPL in the lower right corner of the actual model. Another location where
NAPL displacement is difficult is near the top center where a small amount of NAPL remains.
The displacement of NAPL from this area was simulated accurately, where a low NAPL
saturation remained for some time using both models.
Figure 3 (also presented in Sec 2.1.1) compares the results of the physical and simulated
models using water. It is observed in both models that some pockets of NAPL remained in the
uppermost horizontal fractures when using water to displace. Furthermore, no NAPL was
displaced from the lower right corner of both the physical and simulated model. Aside from
small differences in amount of NAPL displaced, the two models compared very well.

3. SUMMARY & CONCLUSIONS

Using the simulator to study multiphase flow behavior within fractured media can be an
important tool. The simulations reduce the number of actual experimental runs that need to be
performed and provide a means of studying the effects of parameters such as viscosity and
density. For example, the hour of laboratory experiments were done within 15 minutes using the
computer simulation. Time was saved because once the initial simulation input was designed,
changing a viscosity is simply a matter of changing a value in the input file. In the lab, the
physical model must prepared for each run and then an hour of observations with recording of
data, all taking the experimentalist several days. In the physical experiments, photographs every
10 or 20 seconds for an hour were necessary. Once the simulation has started, the user can walk
away and come back later to view the output at any point in time.
Confirming the adequacy of the simulator by comparing with laboratory experiments instills
confidence in the effectiveness of the simulator for field simulations. Although, uncertainties in
the field may hinder the potential to perfectly model field situations, it still can be useful in
predicting the outcome in the field.
REFERENCES

DOE, 1998. Data collected from DOE sites and summarized by T.O. Early, Subsurface
Contaminants Focus Area. As cited in Strategies for Monitoring the Performance of DNAPL
Source Zone Remedies, Technical/Regulatory Guidelines, Prepared by the Interstate
Technology & Regulatory Council, DNAPL Team, 2005.
Dyes A.B, B.H. Caudle, et al. 1954. Oil Production After BreakthroughAs Influenced by
Mobility Ratio, Transactions of the American Institute of Mining and Metalurgical Engineers,
201:81-86. As cited in Lake, Larry W., 1989. Enhanced Oil Recovery. Englewood Cliffs:
Prentice Hall, pp.191-193.
Jackson, 2001. DNAPL Remediation Which New Paradigm Will Prevail?, GWMR, 54-57.
Kostarelos, 1998. PhD Thesis, The University of Texas at Austin, 2-8.
Charbeneau R.J., 2000. Groundwater Hydraulics and Pollutant Transport. Upper Saddle River:
Prentice Hall.
Delshad M.., G.A. Pope, et al. 1996. A Compositional Simulator for Modeling Surfactant
Enhanced Aquifer Remediation, 1 Formulation. Journal of Contaminant Hydrology. 23, 303-
327.
Esposito S.J. and N.R. Thomson, 1999. Two-Phase Flow and Transport in a Single Fracture-
Porous Medium System. Journal of Contaminant Hydrology. 37, 319-341.
Lake L.W., 1989. Enhanced Oil Recovery. Englewood Cliffs: Prentice Hall.
Manning R.K., G.A. Pope et al. 1983. A technical survey of polymer flooding projects, US
Department of Energy DOE/BETC/10327-19, Bartlesville, Oklahoma. p. 344.
Martel R., A. Hebert, et al. 2004. Displacement and sweep efficiencies in a DNAPL recovery
test using micellar and polymer solutions injected in a five-spot pattern Journal of
Contaminant Hydrology 75:1 29.
Parker, Beth L., J.A. Cherry, et al. 2004. Field Study of TCE Diffusion Profiles Below DNAPL
to Assess Aquitard Integrity. Journal of Contaminant Hydrology. 74, 197-230.
Peters, R.A.. Alcohols, higher aliphatic. In: Kirk-Othmer encyclopedia of chemical technology.
4th ed. Vol. 1. John Wiley and Sons, 1992. p. 865- 893
Pitts M.J., K. Wyatt, et al. 1993. Utilization of chemical-enhanced oil recovery technology to
remove hazardous oily waste from alluvium. SPE Int. Symp. Oilfield Chem., pp. 3344.
(SPE/DOE paper no. 25153).
Reservoir Engineering Research Program Center (RERPC) for Petroleum and Geosystems
Engineering for The University of Texas, 2000. Volume II: Technical Documentation for
UTCHEM-9.0 A Three Dimensional Chemical Flood Simulator. Texas.
Reservoir Engineering Research Program Center (RERPC) for Petroleum and Geosystems
Engineering The University of Texas, 2000. Volume I: Users Guide for UTCHEM-9.0 A
Three Dimensional Chemical Flood Simulator. Texas.
Reynolds D.A. and B.H. Kueper, 2000. Multiphase Flow and Transport in Fractured Clay/ Sand
Sequences. Journal of Contaminant Hydrology. 51, 41-62.
Reynolds D.A. and B.H. Kueper. 2004. Multiphase Flow and Transport Through Fractured
Heterogeneous Porous Media. Journal of Contaminant Hydrology. 71, 89-110.
Taiwo J. (2005). Masters Thesis. Polytechnic University, Brooklyn, NY.
US EPA, 2008. Remediation databases, http://www.epa.gov/tio/databases/
Wade J. (2005). Masters Thesis. Polytechnic University, Brooklyn, NY.