*>EPA
/
EPA 600/S-21/061 April 2021
Research BRIEF
\
www.epa.gov/research
| INNOVATIVE RESEARCH FOR A SUSTAINABLE FUTURE
The Formation, Transport, and
Breakup of Submerged Oil-Particle
Aggregates in Great Lakes Riverine
Environments
(alphabetical) John Berens1 Michel C. Boufadel2, Faith A. Fitzpatrick3, Marcelo H.
Garcia1, Jacob Hassan4, Earl Hayter5, Lori Jones6, Susan Mravik7, David M. Waterman8
Background
The formation, transport, and resuspension of oil-particle aggregates (OPA) in freshwater environments are of much
interest to oil spill responders and scientists, especially as transportation of light and heavy crude oils has
substantially increased across river corridors and coasts in the Great Lakes Basin (Great Lakes Commission, 2015).
The persistent sheening from accumulated OPA along 60 km of the Kalamazoo River in Michigan's lower peninsula
resulted in a lengthy and expensive cleanup following the 2010 Enbridge Line 6B pipeline rupture (USEPA, 2016). The
interaction of oil with river sediment and organic matter and its long-term fate depend on the physical properties of
the oil and particles as well as the environmental setting of river, its climate, morphology, currents and mixing
opportunities (Lee et al., 2001; 2002; Fitzpatrick et al., 2015a).
Laboratory experiments and model simulations conducted during the Kalamazoo River response helped to describe
how accumulated OPA river bottom sediment was resuspended and transported (Fitzpatrick, et al., 2015a, b). This
information was used in containment and recovery plans as well as endpoint determinations. In 2016-18 laboratory
experiments and model applications were continued to more broadly describe OPA fate and transport over a larger
range of river and sediment conditions.
This research brief presents results from the OPA laboratory experiments and fate and transport modeling conducted
under a Regional Applied Research Effort. This research involved the U.S. Environmental Protection Agency (USEPA),
the U.S. Geological Survey (USGS), U.S. Army Corp of Engineers (USACE), the University of Illinois Ven Te Chow
Hydrosystems Laboratory (VTCHL), and the New Jersey Institute of Technology (NJIT). Research objectives included
expanded laboratory experiments of aggregate characteristics with Cold Lake Blend diluted bitumen (hereafter
referred to as Cold Lake Blend) and a range of sediment particle sizes, adding an OPA formation algorithm to an
existing sediment contaminant transport model, and development of a simplified, particle-tracking based rapid
response model of OPA formation, transport, and deposition. A description of formulas developed for determining
mixing energy in rivers in terms of river properties is also included.
¦ University of Illinois, Champaign-Urbana, IL, 2UNew Jersey Institute of Technology, Newark, NJ,
3U.S. Geological Survey, Madison, Wl, 4U.S. EPA, Chicago, IL, 5U.S. Army Corp of Engineers, Engineer Research and
Development Center, Vicksburg, MS, formerly with the University of Illinois, Champaign-Urbana, IL,
7U.S. EPA Office of Research and Development, Center for Environ. Solutions & Emergency Response, Ada, OK,
8Civil and Environmental Engineering, South Dakota School of Mines and Technology, Rapid City, SD
-------
Introduction
The presence of oil-particle aggregates (OPA) from
the mixing of oil and sediment in marine coastal
environments has been recognized since the 1990s
(Lee et al,, 2001; Owens et al,, 1994; Bragg and
Owens, 1995; Lee et al., 1997a,b,c; Owens and Lee,
2003; and Gustitus and Clement, 2017). Bench-
scale tests and wave tanks, as well as numerical
modeling for describing OPA formation along marine
shorelines, has been important in the spill response
community for containment and recovery. Clay
particles have been added to spilled oil at sea for
increased physical dispersion and for increasing
potential for biodegradation (Lee et al., 1997a, Lee
et al., 2003). However, less is known about OPA
formation in river spills (Lee et al., 2015; Fitzpatrick
et al., 2015a) and sinking agents are not permitted
for use in the U.S. Oiled sediment was found along
60 km of the Kalamazoo River downstream of the
2010 Line 6B release of diluted bitumen into the
Kalamazoo River. Oiled sediment was especially
problematic in impounded sections and along
riverbanks and side channels (USEPA, 2016). In 2012,
science advisors recognized the release of oil from
oil-laden river sediment containing OPA as the cause
for continued oil sheening in slow moving areas (Lee
et al., 2012) and by 2014 dredging was employed to
remove the remaining large deposits (Dollhopf et
al., 2014; USEPA, 2016). Ultraviolet-epifluorescence
microscopy was employed for visual determinations
of oil droplet and sediment mixtures (Figure 1).
2
^ #
• t"
50 Mm
Figure 1. UV epifluorescence photo from Lee et al. (2012) of Kalamazoo River bottom
sediment and Cold Lake Blend.
-------
The formation of OPA in rivers is co-dependent on
characteristics of the oil as well as the environmental
setting that provides the particles and the mixing
energy (Fitzpatrick et al., 2015b). Oil droplets
form from the slick as a result of imbalance between
the destructive forces of turbulence in water in
comparison to the resisting forces from interfacial
tension and oil viscosity (Zhao et al., 2014a). In
rivers, the formation of oil droplets is expected to
result from one of two mechanisms: 1) entrainment
of oil blobs from the slick due to (turbulent)
horizontal shear from sudden changes in the water
surface and 2) advection of oil towards the
streambed following riffles or sudden water drop.
While the buoyancy of the oil droplets causes them
to rise to the surface (Zhao et al., 2015), rivers with
sufficient turbulent energy could keep oil droplets
suspended in the water column, including in the
vicinity of the streambed, where the energy
dissipation rate is highest.
Oil droplets attach to particles in the water column
through Pickering emulsions (when there is more oil
than sediment) or nesting (when there is more
sediment than oil). The traditional approach of OPA
formation assumes that the particles attach to the oil
droplet until complete coverage occurs, after which
no attachment takes place (Hill et al. 2002, Khelifa et
al. 2002, Khelifa et al. 2004, Ajijolaiya et al. 2006).
The term used here to describe the interaction of oil
with particles to form oil particle aggregates (OPA) in
a river transcend the nomenclature proposed
by Gustitus and Clement (2017) to include both
microscopic and macroscopic sizes in various shapes
and configurations (Fitzpatrick et al., 2015a). In
situations where the particles were platy (e.g., clay
particles), it was assumed that the particles deposit
flat on the droplet (Khelifa et al., 2005). However,
recently Zhao et al. (2017) found through confocal
microscopy that platy particles of size 10 microns
penetrate deep (almost the whole length) into
the oil droplets (Figure 2), and cause the OPAs to
disintegrate after a few hours of mixing.
Additionally, the presence of particles and
aggregation cause the oil droplets to reduce in size
as shown for three examples after 3 hours (Figure
3a-c) and one example after 24 hours(Figure 3d)
(Zhao et al., 2017).
PROJECTILE MECHANISM
Oil - Green Particle - Red
Figure 2. Particles (length 10 microns) penetrate the oil droplet (size around 50 microns) from Zhao et a! (2017).
3
-------
(b) ^
jflir
The time required for complete coverage of oil
droplets by particles (tc) is related to oil droplet
diameter (DJ, sediment diameter (D ), oil droplet
concentration (N:), sediment concentration (N.)
sediment concentration at t=0 (N), oil-sediment
coalescence efficiency (aos), turbulent-kinetic energy
dissipation rate (€), and Kinematic viscosity of water
(v) (Hill et al., 2002). Zhao et al. (2016) developed the
conceptual-numerical model A-DROPto help predict
the amount of oil that might be trapped in the oil-
particle interaction process. The A-DROP model was
built off the framework of VDROP, which describes oil
droplet formation, the process needed before oil-
particle interactions can happen (Zhao et al., 2014a,
2014b, 2015).
The shape and structure of OPA are affected by the
mixing energy and both oil and particle properties
(Boufadel et al., 2019). An increase in mixing energy
results in smaller but more oil-particle interactions
(Zhang et al., 2010). Small particles, generally silt
and clay, play the most important role in coating
oil droplets, because, per unit mass of sediment,
they have a larger number and a larger surface
area (Ajijolaiya et al., 2006). As oil concentration
increases, the concentration of sediment also needs
to increase, and thus it is common to represent the
number of OPAs as function of the ratio of particle to
oil concentration. Usually, the optimum efficiency for
oil-particle interaction occurs when the ratio of
concentrations is close to 1.0. Bacteria and other
biogenic-based particles may become part of the
particulates that form an aggregate (Bragg and
Owens, 1995).
Some of the simplified assumptions of OPA used in
modeling are oil droplets and particles are spherical,
particles cover an oil droplet as a monolayer, and
that the oil droplet size distribution (DSD) has
reached equilibrium prior to interaction with the
particles. This was motivated by offshore oil spills
where the sediment content offshore (at the water
surface) is small and increases rapidly at the
shoreline. The contact between droplets and
particles depends on their diameters and mixing
energy, whereas the coalescence (i.e., to form an
OPA) depends also on oil viscosity, hydrophobicity of
the particle, and the interfacial tension or interaction
between oil and particles.
-------
In river channels, the average mixing energy (i.e.,
the average energy dissipation rate) is related to
river properties of channel shape and size (hydraulic
radius), slope, and roughness coefficient (Boufadel
et al., 2019). Similarly, the average energy dissipation
rate is related to the bottom shear stress (typically
used in river hydraulic and sediment transport
models) by knowing the water density, total water
depth, and roughness height of the streambed,
which is typically related to bed material particle
size (Boufadel et al., 2019; Garcia, 2008). These
characteristics provide the hydraulic building blocks
for understanding the fate and transport of OPA in
rivers (Jones and Garcia, 2018). The water depth in
rivers can change quickly in response to precipitation-
induced runoff events.
Modeling the fate and transport of OPA over different
flow scenarios was one of six techniques that also
included geomorphic mapping, field assessments
of submerged oil, oil sheen tracking, forensic oil
chemistry and net environmental benefit analysis in a
multiple-lines-of-evidence approach used in the
cleanup of submerged oil in the Kalamazoo River
(Dollhopf et al., 2014; USEPA, 2016). Activities during
the Kalamazoo River response also included
collection of new field and laboratory data to support
the OPA models. These data describe OPA formed
from the combination of weathered Cold Lake Blend
and fine-grained Kalamazoo River sediment
regarding oil content, size, density, settling velocity,
and erodibility (Perkey et al., 2014; Waterman and
Garcia, 2015; Waterman et al., 2015).
As part of the Kalamazoo River response
several models were developed to simulate the
hydrodynamic conditions needed for resuspension
and settling of submerged oil, OPA and oil-
contaminated sediment (Fitzpatrick et al., 2015b).
The main assumption for these models is that OPA
was previously formed and existed in the riverbed at
concentrations pre-determined based on submerged
oil assessments. A two dimensional (2D)
hydrodynamic sediment transport model
(LimnoTech, 2015; Hayter et al., 2014; Jones and Lick,
2001) was used to simulate resuspension and settling
of existing OPA in the Kalamazoo River during a high
baseflow scenario. A three-dimensional (3D)
hydrodynamic model was coupled with a 3D
Lagrangian particle tracking model to simulate the
movement of OPA in an impounded section of the
Kalamazoo River under the same
flow conditions (Zhu and Garcia, 2015; Zhu et al.,
2018). The 3D model included the flow releases
associated with a hydro-power dam along with wind
effects to assess the transport and fate of OPA in the
impoundment formed by the dam, known as Morrow
Lake.
In order to represent river sediment in the models
appropriately, Perkey et al. (2014) conducted
cohesive sediment erosion testing by use of a
field-deployable sedflume on nine cores collected
along the 60 km reach of the Kalamazoo River with
submerged OPA. A sedflume consists of a rectangular
flume with the capability of applying bed stresses in
the range of 0.1 to 12 Pa on a riverbed core as it is
raised up through an opening in the bottom of the
flume into the recirculating water currents. Sensors
in the flume continuously record erosion rates as the
sediment density, texture, and cohesiveness change
with depth in the core. From these data
representative equations for erosion rates relative to
shear stress applied to the river bed were developed.
SEDZLJ, which is the mixed sediment bed model in
the 2D hydrodynamic sediment transport model
mentioned above, was modified to incorporate the
physical characteristics of OPAs (Hayter et al., 2014;
LimnoTech, 2015). Bulk density and particle size
distribution were analyzed.
Results included in this research brief encompass
new developments in research concerning the
formation, transport, and fate of OPA in riverine
environments. These results build off the laboratory
experiments and fate and transport modeling
completed for the 2010 Line 6B pipeline release into
the Kalamazoo River. This brief includes the results
from expanded laboratory experiments of aggregate
characteristics with Cold Lake Blend and a range of
sediment sizes and additional modeling of OPA
formation, transport, and deposition.
-------
Laboratory Experiments
Laboratory experiments conducted by the VTCHL at
the University of Illinois in Urbana, built off those
conducted for the 2010 Kalamazoo River spill of
diluted bitumen, following the same approach used
by Waterman and Garcia (2015). This approach
concentrated on the product of OPA formation to
allow parameterization of numerical models
representative of the remaining OPA deposits in the
Kalamazoo River from about 2012-14. Evaluation of
the mixing energy was intended to aid in the
development and validation of process-based
numerical models of OPA formation.
Experiment Methods
The experiments included orbital shaker/baffled
flask tests with different mixing speeds and oil
weathering, tensiometer, and epifiuorescence
analyses. Orbital shaker baffled flask tests were
performed with Cold Lake Blend and kaolinite clay
to explore the effects of mixing speed and mixing
time on OPA formation. The ceramic grade EPK
Kaolin used in the experiments was 97% kaolinite
with a 325 mesh size (particle size less than 44
micron). For all tests, 100 |iL of oil and 50 mg of
kaolinite were mixed in 120 mL of tap water. The
specific conductance of water was 635 |iS/cm2 and
typical of representative conditions found in natural
rivers, including the Kalamazoo River. After mixing,
oil deposits were removed and analyzed under
ultraviolet epifiuorescence microscopy. The
population of OPA was characterized in terms of
their size distribution.
The mixing speed experiments used speeds of 160,
180, and 200 RPM. For the same experiments, the
weathered state was also varied with oils of 0, 10,
and 17.5% mass loss for a total of 9 tests. Each test
was mixed for 1 hour and with air, water, and oil
temperatures between 21°C and 22°C. The time-
dependent experiments had mixing times of 4, 8, 16,
32, and 60 minutes. The weathered state for these
tests was 17.5% mass loss, mixing speed of 200 RPM,
and temperatures of 22.7°C. Each mixing time
condition was conducted three times for a total of 15
test results.
Orbital shaker control tests were also conducted with
the EPK kaolinite clay without the presence of Cold
Lake Blend. For these initial tests, clay alone was
mixed to see how the particle size distribution of clay
aggregates was affected by varying mixing speed,
sediment concentration, and water density. Mixing
speeds tested with clay alone were 160, 180, 200
RPM, and concentrations of clay were 5, 20, and 80
mg/L. Saltwater and two types of freshwater (distilled
and tap water) were used as the medium for mixing.
Each combination of the three conditions was tested
resulting in 27 size distributions of kaolinite clay.
Tensiometer tests (Figure 4) were conducted to
examine the surface tension of Cold Lake Blend
diluted bitumen at both the oil-water and oil-air
interfaces. A DuNouy Tensiometer was used with
beakers and 50 mL of the ASTM Standard D971-12
(ASTM, 2013), The tensiometer was calibrated to the
ASTM Standard and the surface tension of distilled
water was measured. Surface tension was measured
by lowering the beaker away from the tensiometer
ring until it separated from the fluid. The same
procedure was repeated with a 10 mm thick layer of
oil on the surface. Three measurements were taken
with three different samples of oil for this test.
Figure 4. Ring tensiometer post-measurement
of oil-air surface tension test.
6
-------
An opposite configuration was used in the
measurement of the oil-water interface where the
same thickness of oil was placed at the bottom of
the beaker and water was gently placed above it
(Figure 5). The ring of the tensiometer was uniformly
submerged in the oil and then the beaker lowered
until the ring broke away from the oil and was only
submerged in distilled water. Three measurements
were taken for each sample, but due to the time-
consuming nature of the tests, only two samples of
oil were used.
Epifluorescence techniques followed those of Lee
et al. (2012) and Waterman and Garcia (2015).
Samples of OPAs were extracted from shaker
tests and photographed using epifluorescence.
With each experimental trial, only a subsample
of the deposit was prepared for epifluorescence
photomicrography. Subsamples were taken due to
the overly large number of images that would be
necessary to evaluate the entire deposit, although
enough subsamples were obtained to ensure that the
subsamples were representative.
Extensive efforts were undertaken to develop new
methods to obtain samples in such a manner that
photomicrographs could be readily analyzed using
an automated algorithm. The key confounding
factor in the earlier image analysis of Waterman
and Garcia (2015) was the presence of sheen from
the supernatant water overlying the deposit, not
associated with OPA, which was very challenging to
distinguish from actual oil aggregated to particles
in the sediment deposit using automated image
processing procedures. In an image captured using
a single fluorescent wavelength, all fluorescent
material appeared similar, with subtle differences
and personal judgment necessary to distinguish
sheen attached to the Fluorodish glass from OPA.
Eliminating or minimizing the presence of sheen in
samples was thus a key objective of developing a
modified and more objective method of preparing
samples.
A variety of adsorbent materials were experimented
with to eliminate sheen from the supernatant
water to prepare "clean" OPA deposits for image
analysis. Such adsorbents included diatomaceous
earth, polypropylene plastic sorbents, ground corn
cob sorbents, and paper shop towels. None of the
adsorbent materials was adequate after a single
surface application to remove enough sheen from
the samples to greatly improve the image quality.
The most important factor was found to be repeat
Figure 5. a) Ring parallel to the water/oil surface before the test and b) the oil about to
break away from the tensiometer ring near the surface.
-------
application of a cleaning and rinsing procedure, as
opposed to the specific type of adsorbent. The paper
shop towels were found to perform as well as the
other materials when the repeat procedure was
used. The repeat procedure involves the following:
absorbing as much sheen as possible from the surface
of the beaker using the adsorbent; removing most
of the supernatant water using a pipette; gently
rinsing both sides and bottom of the beaker with a
laboratory wash bottle containing tap water until the
water level in the beaker returned to its original level.
Rinsing the bottom of the beaker was done with care
by tipping the beaker and allowing the sediment to
settle to one side to ensure the jet of wash water
does not disrupt the structure of the sediments
in the deposit. The procedure was found to yield
photomicrographs containing oil globules in OPA, but
without sheen.
Each of these parameters warranted experimentation
over a range of values to establish the relative effects
of each parameter, however, each incremental
expansion of the parameter space exponentially
increased the number of images that needed to be
processed. This description is intended to highlight
the issue that to begin building a data set not limited
to a very narrow parameter range, a robust image-
processing algorithm is required, in addition to a
robust method to obtain samples in a manner that
the images can be readily analyzed. The method to
process images used by Waterman and Garcia (2015)
is extremely labor intensive; such methods inherently
limit the scope of experimental trials to a limited
parameter space, while the need clearly exists for
experimental data analyzed in such a way to help
parameterize numerical models over a wide range of
potential prototype conditions.
While the method for extracting a size distribution
of OPAs in Waterman and Garcia (2015) from
epifluorescence microscopy was accurate, it was
time consuming and limited how many tests could
be completed. The development of an algorithm for
the post-processing of images taken with the same
epifluorescence methodology to extract the size
distribution of OPA deposits in a timelier fashion was
initiated but not completed. The algorithm used free
ZEN lite software from Zeiss and a combination of
wavelength bands to brighten oil content in images
(Carl Zeiss Microscopy GmbH, 2014). Brightness
thresholds were then used to identify oil droplets
against the background noise and the size distribution
of the droplets was produced by assuming spherical
particles. Using the resolution of each image, a size
distribution was constructed from a large series of
images.
Experiment Results and Discussion
Effects of Mixing Speed on OPA Formation
Results from the orbital shaker baffled flask tests
showed that increasing mixing energy decreased the
median (D50) aggregate size of OPAs for weathered
diluted bitumen (Table 1). This test also varied the
weathered state of the oil, but the tests did not have
a noticeable difference between 10 and 17.5 percent
mass loss. These results imply that higher mixing
energy conditions (higher river velocities) provide an
environment for smaller OPA to form.
Table 1. Laboratory results for the effects of mixing
speed and weathered state on the median (Dso)
aggregate size of OPAs formed during orbital flask
shaker tests
Median (Dso) aggregate size
Mixing Speed
Weathered State (% Mass Loss)
(RPM)
0
10
17.5
Mean
160
28.2
40.5
39
35.9
180
31.8
27.7
30.7
30.1
200
30.2
22.3
16.3
22.9
Effects of Mixing Time on OPA Formation
Samples from the mixing time experiments of Cold
Lake Blend and kaolinite clay have yet to be analyzed
and are awaiting validation of the OPA identifying
algorithm. Analysis the epifluorescence images of
these tests will provide great insight to the
development of OPA through time. The mixing times
that were tested were 4, 8, 16, 32, and 60 minutes.
Kaolinite Clay Control Aggregation
Control tests were conducted to see how water
salinity, kaolinite clay concentration, and mixing
speed affected the resulting particle aggregate size
distribution of kaolinite clay without oil (Figure 6a-c).
Salt water tended to generate size distributions with
larger particle aggregates than those in the tap or
distilled water (Figure 6a).
8
-------
Aggregate size distributions tended to vary less for
lower concentrations of kaolinite clay (Figure 6b).
Generally, mixing speed had the least effect on
aggregate size (Figure 6c). These test results provide
information on how aggregate size differs when
formed in the presence of oil compared to just clay
alone- under the same laboratory conditions.
10" 10'
size (/im)
Figure 6. Size distributions of kaolinite clay samples for varying a) water density, b) kaolinite
clay concentration, and c) mixing speed.
Surface Tension
Using weathered Cold Lake Blend and tap water with
specific conductance of 640 (iS/cm, the surface
tension at the oil-air and oil-water interfaces were
measured (Tables 2 and 3). The mean surface tension
was 38.6 mN/m at the oil-air interface. The tension
at the oil-water surface was 21.3 mN/m, or about
half that of the oil-air interface. These values could
be used to quantify oil droplet breakup and
understand the process by which the breakup of an
oil slick occurs at a river surface and below or at the
bottom. The density of the weathered Cold Lake
Blend in the laboratory was less than but close to the
density of water, indicating that some of the oil might
be in the water column or submerged.
Table 2, Oil-air surface tension
measurements (mN/m)
Measurements (mN/m)
Sample
1
2
3
1
40.5
38.0
38.5
2
39.0
39.0
38.5
3
38.3
38.0
38.0
Table 3. Oil-water surface tension
measurements (mN/m)
Measurements (mN/m)
Sample
1
2
3
1
24.0
23.0
20.0
2
23.0
20.0
18.0
-------
Automation of Epifluorescence-Based
Aggregate Size Determinations
During the laboratory tests conducted by VTCHL
during the Kalamazoo River spill response (Waterman
and Garcia, 2015) the OPA size determinations were
done manually and were overly time consuming.
Because of the additions of multiple tests for this
study, development of an automated procedure was
highly desirable but fuil development of an algorithm
for a range of mixing speed experiments was outside
the scope of this study. Once analyzed, the
experimental data set will be unique in that it
focuses on time-dependent development of OPA.
More importantly, the techniques and algorithms
developed will allow the processing of much larger
data sets to begin to generalize experimental results
instead of being limited to specific conditions
due to limitations in the ability to analyze large
experimental data sets. Before the time-dependent
shaker test results can be analyzed, the algorithm
will be validated for the mixing speed experiments
that were analyzed manually. An example of results
from the initial testing of the algorithm are shown
in Figure 7. Like other photogrammetry techniques, a
shortcoming of this method is that it assumes
spherical particles, meaning that the size of non-
spherical particles can be over or under estimated
depending on the thresholds chosen. Another
shortcoming is that the algorithm works only for
particles in a 2D plane. In some images, particles can
be stacked on top of each other which the algorithm
cannot identify. However, this algorithm has
provided some promising results in samples with
bright, spherical OPA deposits. Figure 8 demonstrates
how the algorithm performs under nearly ideal
circumstances for a test from the mixing speed
experiment.
•v #">•
Figure 7. Example of OPAs identified in an image taken with UV epifluorescence.
The red circle is the diameter taken for a size distribution.
Manual Sampling
Image Processing Algorithm
160 RPM /17.5% Mass Loss
Particle Diameter (/jm)
Figure 8. Size distribution of
OPAs found in a deposit mixed
at 160 RPM of Cold Lake Blend
weathered to 17.5 % mass loss
and kaolinite clay extracted both
manually and with the image
processing algorithm.
-------
OPA Fate and Transport Models
Two approaches were used to further develop
modeling applications for OPA fate and transport
in rivers. These included adding an OPA formation
algorithm to a 2D sediment transport model, and the
development of a ID particle-tracking based rapid
response model of OPA formation, transport, and
deposition. The goal of both models was to
incorporate the oil droplet formation numerical
model VDROP (Zhao et al., 2014b) and oil particle
formation model A-DROP (Zhao et al., 2016).
Simulation of Oil Droplet and OPA
Formation in a 2D Sediment Transport
Model
One of the original research objectives of this
Regional Applied Research Effort (RARE) study was to
combine the Environmental Fluid Dynamics Code
(EFDC) and its related 2D sediment transport model
SEDZU (Hamrick, 1992; LimnoTech, 2015; Hayteret
al. 2014) with the oil droplet formation VDROP (Zhao
et al., 2014b) and ultimately the oil particle
formation model A-DROP (Zhao et al., 2016). The
SEDZU model included the river channel and river
channel plus floodplain version. Model inputs
included bathymetric data, tributary flows,
suspended sediment rating curves, dam rating
curves, and bed sediment characteristics along the 38
miles of spill effected reach between Marshall and
Kalamazoo, Ml (LimnoTech, 2015).
Hayter et al. (2014) describes the modifications made
to LimnoTech's 2D EFDC model (LimnoTech, 2015)
that includes the SEDZLJ (Jones and Lick, 2001)
sediment bed model updated for simulated OPA
transport during the Kalamazoo River spill response.
The oiled sediment, in the form of oil particle
aggregates (OPA), was simulated as distinct particles
from the mixed sediment particles represented in
SEDZLJ. The changes made to EFDC and SEDZLJ to be
able to represent the transport of both sediment and
OPA were to add a separate transport module for the
OPA to EFDC, 2) incorporate the percentages of OPA
types present in the sediment bed along the modeled
reach in the SEDLZJ layered bed model, and 3) modify
the mass balance routines in EFDC to calculate time-
and space-averaged mass balances of the simulated
classes of OPA. The modifications added the ability to
represent a specified number of sediment size classes
as well as a specified number of OPA classes or types.
The unverified results from the simulation of OPA
transport in the modified 2D SEDZLJ model during
a 13-day period in October-November 2011 for the
Kalamazoo River are shown in Figure 9 (Hayter et al.,
2014). Five size classes of sediment and three OPA
classes were included. The results show the change
in the mass of the simulated OPA classes in the
surface bed layer over the 13-day period. Figure 9
shows time series over the last 12 days of the
simulation of the predicted discharge and total OPA
concentrations at the following locations: a) Mile
Post (MP) 9.25; b) MP 14.1; c) MP 18.1; and d) the
35th Street bridge. The simulated OPA concentration
at the most downstream location (35th Street bridge)
is the lowest of the modeled river reach. The
numerous spikes seen in the OPA concentrations at
all four locations over the last 12 days of the model
run are caused by erosion of surficial sediment bed
and entrainment of embedded OPAs into the water
column. These results as well as results at other
locations along the modeled reach of the river
confirmed that OPA was entrained at most non-
impounded locations several times over the 13-day
run. The simplified OPA transport module developed
during the cleanup was viewed as a first-generation
model, upon which more advanced modeling
algorithms of OPA formation, transport, deposition,
resuspension and potentially breakup could be
added to during later research opportunities.
For the RARE study, an attempt was made to add the
oil particle formation model VDROP (Zhao et al.
2014a,b) to the Kalamazoo River 2D OPA modified
SEDZLJ (Hayter et al. 2014) by treating the VDROP
numerical model as a source/sink term in the SEDZLJ
equations.
-------
1.0E+02
1.0E+01
i l.OE+OO
g 1.0E-01
u
c
o
u
1.0E-02
1.0E-03
Time (days)
MP 9.25 MP 14.1 MP 18.1 35th St Bridge Q
Figure 9. Total OPA
Concentration Time
Series at the Specified
Four Locations. Day 665
corresponds to Oct 29.
Within each grid cell and within each time step
EFDC would call VDROP; however, this work did
not conclude with a working model. We were only
provided the executable of VDROP, and not the
code, and despite many efforts to link it to EFDC,
we were not successful.
Simplified ID Modeling of OPA
Formation, Transport, and Fate
A simplified ID model was developed, called
IDHydroOPA, to assist in rapid response in the event
of a riverine oil spill (Jones, 2018; Jones and Garcia,
2018). The goal of the development of this model
was to be able to quickly simulate the formation,
transport, and fate of oil as it mixes with sediment
and forms OPA in a river. The model was applied to
the spill reach of the Kalamazoo River. Two types
of simulations were done - one with a simplified
rectangular channel and another with channel cross
sections from a USGS flood inundation model (Hoard
etal., 2010).
Methods
The formation and transport of OPAs were modeled
in IDHydroOPA by applying a random walk particle
tracking algorithm (Jones and Garcia, 2018) to the
previously published numerical models for oil droplet
formation VDROP (Zhao et al,, 2014b) and OPA
formation A-DROP (Zhao et al., 2016). First, the model
was tested in a straight, rectangular channel before
being tested in a HEC-RAS model of the Kalamazoo
River (Hoard et al., 2010). The hydraulics of the
Kalamazoo River model was developed using a steady
state HEC-RAS model (Brunner, 2016) from previously
developed HEC-RAS models of the river completed
during the spill response (Hoard et al., 2010).
The A-DROP model (Zhao et al., 2016) provided the
size distribution of OPAs based on the oil droplet size
distribution and one characteristic size of particles.
Additionally, the A-DROP model did not account
for the penetration of particles in the oil droplets
nor did it account for the fragmentation of OPAs,
both discovered afterwards (Zhao et al., 2017). The
IDHydroOPA model assumed a steady state Rouse-
Vanoni suspended sediment concentration profile
(Garcia, 2008) at each iocation along the reach. Other
assumptions included:
1. The oil droplets themselves do not breakup or
adhere to each other. The model accepts an
input of the distribution of the oil droplet
diameters, and this distribution is treated as
the final size distribution of the oil droplets after
the initial droplet formation occurred following a
spill event (Zhao et al., 2016).
-------
2. When the particles attach to the oil droplet, they
cover the oil droplet uniformly and, in a
monolayer (Zhao et al., 2016).
3. Once particles attach to an oil droplet, they do
not detach.
4. The only type of collisions that result in
coagulation are collisions between oil droplets
and sediment particles, and OPA particles and
sediment particles. Thus, a collision of an oil
droplet and an OPA particle will not result in
coagulation.
5. The suspended sediment concentration in the
river is assumed to be at steady state; thus,
sediment particles are not tracked in the model,
but rather the suspended sediment
concentration is described by a Rouse-Vanoni
profile throughout the river (Garcia, 2008).
6. Once an OPA particle settles onto the river
bed, it is no longer tracked in the model. Thus,
resuspension of OPAs is not included in this
version of the model.
7. The model assumes that the river is a
rectangular channel, and the flow is steady and
uniform.
The steady state distribution of oil droplets used in
the lDHydroOPA model were assumed to be the
steady-state distribution of the oil droplets after an
initial spill of oil has occurred. The distribution of oil
droplet sizes used in this study is shown in Figure 10,
and in each simulation 1,000 oil particles were used.
0.35
0.3
0.25
c 0.2
o
o
CO
U- 0.15
0.1
0.05
0
0 20 40 60 80 100 120 140 160 180
Oil Droplet Diameter (^m)
Figure 10. Initial oil droplet distribution used as
input to the lDHydroOPA model with a simplified
rectangular channel for the Kalamazoo River spill.
To simulate the movement of oil particles and OPAs
through the domain, the random walk particle
tracking method was used to move each particle
through (x, y, z) space (Zhu, 2015; Garcia et al., 2013;
Visser, 1997). The random walk scheme simulates the
movement of oil particles and OPAs due to advection
and turbulent diffusion. The settling velocity of oil
particles determined by Zhao et al. (2016) was used
and recomputed as sediment particles attached.
Suspended sediment concentration in the vertical
was described by a Rouse-Vanoni profile (Rouse,
1938) and was assumed to be at steady state.
With this profile, the volumetric concentration of
suspended sediment can be found at any vertical
location in the water column. With this highly
simplified approach, sediment particles did not need
to be individually tracked. The three types of oil
with a range of densities tested by Zhao et al. (2016)
were included in the rapid response model - South
Louisiana crude oil, Silicon oil, and benzene-carbon
tetrachloride (Table 4). These oil types have densities
that span the range of unweathered (932 kg/m3) and
weathered Cold Lake Blend (993 kg/m3) similar to
what was spilled in the 2010 pipeline release into the
Kalamazoo River (Waterman and Garcia, 2015).
Table 4. Types of oil tested and their density
(Jones and Garcia, 2018)
Type of Oil
Density
(kg/m3)
South Louisiana crude oil
(Zhao et al., 2016)
820
Silicon oil
(Zhao et al., 2014b)
968
Benzene-carbon tetrachloride
(Zhao et al., 2014b)
1000
The coagulation algorithm between oil droplets
and sediment was taken from the Zhao et al.
(2016) A-DROP model and coded for inclusion in
lDHydroOPA. Simulations were run for 5 hours while
varying the oil density, flow velocity, and sediment
size. Table 4 shows the three oil types that were
simulated. The physical characteristics of the channel
that were tested are shown in Table 5 as the velocity
was increased a Courant condition was employed to
set the time step used for the simulation, where the
Courant condition is set to unity and given by,
uAt
-------
To ensure that oil droplets are not transported
outside of the domain of the river, the model
established boundary conditions at the water
surface, river bed, and the sides of the river. For all
these boundaries, if over-passed, the particle was
placed back at the boundary. However, if the particle
passed the bed or the side walls, it was considered
settled and no longer tracked. Table 5 shows the
flow conditions of the experiments as well as the
geometry that creates the domain of the river.
Table 5. Flow characteristics used as input to the
model (Jones and Garcia, 2018)
u (m/s)
v (m/s)
w (m/s)
0 (m3/s)
H( m)
S
0.2
0
0
60
3
0.001
0.3
0
0
90
3
0.001
0.4
0
0
120
3
0.001
0.5
0
0
150
3
0.001
0.6
0
0
180
3
0.001
0.7
0
0
210
3
0.001
to air, a process called weathering occurred, where
some of the hydrocarbon groups volatilized, resulting
in the diluted bitumen becoming denser and more
viscous (Waterman and Garcia, 2015). Thus, the
density of the oil, and the distribution of oil droplets
that the oil will result in when it breaks up, depends
on how much the diluted bitumen has weathered.
The weathering of the diluted bitumen is considered
in the analysis of the Kalamazoo River by considering
two cases: no weathering (0% mass loss) and the
maximum weathering which resulted in 17.4% mass
loss (Waterman and Garcia, 2015). In the 0% mass
loss case, the density of the oil is 932.0 kg/m3 and
the distribution of oil particles is shown in Figure
11 (Waterman and Garcia 2015). In the 17.4% mass
loss case, the density of the oil is 992.6 kg/m3 and
the distribution of oil particles is shown in Figure
12 (Waterman and Garcia 2015). The distribution
of oil particles was determined via orbital shaker
experiments at 180 RPM, which was found to be
representative of the mixing energy of open channel
flows (Waterman and Garcia, 2015).
The model was validated by comparing results to
the same set of experimental results from Sun et
al. (2010) that the modeling results of the A-DROP
model, developed by Zhao et al. (2016) was
compared to. A detailed look at validation can be
found in Jones and Garcia (2018).
lDHydroOPA Simulation with Simple Rectangular
Channel
The motivation behind the creation of the rapid
response lDHydroOPA model described in this study
was the 2010 oil spill in the Kalamazoo River (Jones
and Garcia, 2018). Thus, this model was applied
to the Kalamazoo River to develop an estimate of
how much oil would settle and where the oil would
settle given the hydraulic conditions of the river
at the time of the spill. The rapid response model
for the Kalamazoo River still employed a simplified
rectangular channel and was run under steady,
uniform flow, thus the results obtained from this
application were simply a quickly generated first
estimate example of what could be expected in such
a situation during initial phases of the response.
As a reminder, the type of oil that entered the
Kalamazoo River in 2010 was the Cold Lake Blend
bitumen diluted with a benzene rich natural gas
condensate (Dollhopf et al., 2014). When the diluted
bitumen was exposed
0.18
0.16
0.14
0.12
| °-1
i? 0.08
0.06
0.04
0.02
0
0 50 100 150 200 250
Oil Diameter (/xm)
Figure 11. The steady state oil droplet size distribution for the
diluted bitumen with 0% mass loss used in the lDHydroOPA
model (from Waterman and Garcia, 2015; Jones, 2018).
0.25
0.2
0.15
c
o
E
Li.
0.1
0.05
0
0 50 100 150 200 250 300 350
Oil Diameter (/im)
j^Hnn. fl n. n . n ^
Figure 12. The steady state oil droplet size distribution for the
diluted bitumen with 17.4% mass loss used in the lDHydroOPA
model (Waterman and Garcia, 2015; Jones, 2018).
-------
The velocity of the Kalamazoo River is variable, with
flows ranging from less than 0.1 m/s to greater than
2.0 m/s. The median diameter of the suspended
sediment in the river is approximately 30 |im (Reneau
et a!., 2015). The channel is wide and shallow (a
width of 50 m and a depth of 1 m) with a very mild
longitudinal slope (0.06%). Keeping the width, depth,
slope, and sediment size constant, the velocity was
varied from 0.15 m/s to 2.0 m/s to capture all the
expected velocity conditions along the Kalamazoo
River. The varying velocity scenarios were conducted
with both the 0% mass loss and 17.4% mass loss
diluted bitumen for a total elapsed time of 5 hours.
Linking Kalamazoo River IDHydroOPA and
HEC-RAS Models
The IDHydroOPA model with the simplified
rectangular channel and steady, uniform flow meant
that the model was not applicable to rivers with many
tributary inflows and irregular channel geometries.
To improve on these simplifications, the model
was altered to accept channel geometry and flows
from a steady flow Hydrologic Engineering Center's
River Analysis System (HEC-RAS) simulation of the
Kalamazoo River (Hoard et al., 2010; Jones and
Garcia, 2018). The following changes to the previous
model were made:
1. River geometry was extracted from a HEC-RAS
model, and no longer assumed to be rectangular.
2. Flow rates, average velocities, bed shear stress,
and water surface elevations along the river
were extracted from the HEC-RAS model.
3. The law of the wall was used to calculate the
longitudinal velocity with respect to the position
within the water column to account for the
effect of a rough bottom.
4. To account for resuspension of the OPAs from
the bed, the Shields number to induce significant
suspension of the OPAs from the bed served as
an estimate for OPA resuspension.
The structures included in the Kalamazoo River
HEC-RAS model are 16 bridges, two dams, and one
culvert (Hoard et al., 2010). The 2010 Enbridge spill
happened in Talmadge Creek (a Kalamazoo tributary),
but in this model, it was assumed to occur at the
confluence of the creek and the main river. Figure 13
shows a map of the modeled reach. Figure 14 shows
the bed elevation profile.
Figure 13. The location
of the Kalamazoo River
in Michigan, and the
Kalamazoo River shown
with Battle Creek and
Talmadge Creek. The
IJSGS gages used as the
upstream and downstream
boundary conditions in
the HEC-RAS model are
marked. The spill is
modeled at the Talmadge
Creek and Kalamazoo
River confluence.
-------
Distance Downstream (km)
Figure 14. The bed elevation profile of the
Kalamazoo River from Marshall, Ml to Battle Creek,
Ml. The two steep drops in elevation are from dams
(from Hoard et al., 2010).
The peak flow conditions of the flood event that
occurred when the Enbridge pipeline ruptured
were used for flow inputs in the lDHydroOPA
simulation. The flood was approximated by the USGS
to be a 25-year event with the discharge increasing
from approximately 50 m3/s at Marshall, Ml to
approximately 86 m3/s at Battle Creek, Ml due to
tributary inflows (Hoard et al., 2010). The weathered
Cold Lake Blend was used and assumed
to have the 17.4% mass loss (a density of 992.6 kg/
m3 and the same droplet size distributions as that
shown in Figure 12) for simplicity. The model included
introduction of a pulse of 4,000 oil droplets at the
time of the spill and the river suspended sediment
size of 30 pirn used was the same as the previous
application.
lDHydroOPA Simulation Results
Results from Jones (2018) and Jones and Garcia
(2018) for the lDHydroOPA model simulation for
three crude oils with a range of densities (Table 4,
Zhao et al., 2016) are summarized in this section.
The three oil types used by Zhao et al. (2016) were
modeled to explore how oil density affected OPA
formation, transport, and settling characteristics over
a range of typical river velocities and time periods.
lDHydroOPA Simulation with Simple Rectangular
Channel
For the simplistic 6-km long straight rectangular
channel, the percent of settled OPA increased with
increasing velocities, similarly across all three oil
types, with just under 70% of the OPA deposited
in the 5-hr. experimental run time (Table 6). The
centroid of the extent of settled OPAS were also
similar for the three oil types, with the OPAs
transported and settled out further downstream for
the small velocities (over 5,000 m) than the faster
velocities (just under 600 m). Figure 15 shows the
oil droplets, OPA in suspension, and settled OPA for
multiple time steps for the lightest crude oil (South
Louisiana) at a velocity of 0.5 m/s.
Table 6. Percent of oil settled and centroid of settled OPA longitudinally along a hypothetical
straight rectangular channel for six velocity simulations after a time period of 5 hours in the
lDHydroOPA model (Jones 2018)
Oil
Density
(kg/m3)
Parameter
0.2 m/s
0.3 m/s
0.4 m/s
0.5 m/s
0.6 m/s
0.7 m/s
890
% of Settled Oil
0
2.0%
24.7%
47.0%
68.0%
68.1%
Centroid of
Settled OPAs (m)
N/A
5467.0
2684.1
1249.6
1150.4
592.4
968
% of Settled Oil
0
0.5%
35.3%
51.3%
68.0%
68.1%
Centroid of
Settled OPAs (m)
N/A
5080.0
2975.1
1589.0
1145.0
593.1
1000
% of Settled Oil
0
0.06%
21.3%
47.1%
68.0%
68.1%
Centroid of
Settled OPAs (m)
N/A
5387.0
2800.6
1172.9
1171.6
582.7
16
-------
1
0.5 -
0.5 -
1
Time = 0 hours
+ Suspended Oil Droplet
« Suspended OPA
d Settled OPA
100
50
1000 2000
3000 4000 5000
Time = 1 hours
6000
1000 2000
3000 4000 5000
Time = 2 hours
6000
100
2000 3000 4000 5000 6000
Time = 3 hours
100
1000 2000 3000 4000 5000 6000
Time = 4 hours
Lateral Extent of Flume (m)
100
2000 3000 4000 5000 6000
Time = 5 hours
0 . 5 -
0
0 1000
2000 3000 4000 5000 6000
Distance Downstream (m)
100
Figure 15. Snapshots in time of the oil droplets and OPAs in the river domain using South Louisiana crude oil
and a flow velocity of 0.5 m/s in the lDHydroOPA model with a simplified rectangular channel (from Jones
and Garcia, 2018).
17
-------
The distribution of suspended oil droplets, suspended
OPA, and settled OPA in four of the six different
velocity cases after 5 hours are shown for the case
of the low-density oil (South Louisiana Crude Oil) in
Figure 16. As velocity increased, the amount of oil
droplets that mixed with sediment and formed OPA
increased, but with higher velocities the OPA settled
out of the water column faster. Figure 17 shows
that increasing the velocity increases the assumed
suspended sediment concentration, which causes
more OPAs to form.
The results from the simple straight rectangular
channel runs above indicates that as the velocity
increases, the particles settle out of the water
column faster, resulting in more particles settled out
further upstream of the leading edge of the plume.
In principle this result is counterintuitive. However,
the particles settling faster with higher flows is likely
due to greater shear velocities (or bed shear stresses)
that result from higher flows, resulting in more
entrainment of sediment from the bed, which in turn
results in faster formation of the negatively buoyant
OPAs. The change in the suspended sediment
concentration profile with increasing flow velocity
is shown in Figure 17. As can be seen, greater flow
velocities result in much more sediment entrained
from the bed, which in turn increases collision and
coagulation rates.
u = 0.3 m/s
ISuspended Oil Particles
HSuspenced OPA Particles
ISetSed Particles
1000 2000 3000 4000 5000
u = 0.5 rn/s
_ 25
¦5 15-
6000
|Suspended OH Particles
^Suspended OPA Particles
|Settled Particles
_ 50
£
u = 0.4 m/s
¦Suspended Oil Particles
3 Suspended OPA Particles
|Settled Particles
0 1000 2000 3000 4000 5000 6000 7000 8000
u = 0.6 m/s
¦Suspended Oil Particles
3 Suspended OPA Particles
|Settled Particles
1
2000 4000 6000 8000
Distance Downstream (m)
2000 4000 6000 8000 10000 12000
Distance Downstream (m)
Sediment Volume Concentration
Figure 16. The distribution
of suspended oil droplets,
suspended OPA, and settled
OPA after 5 hours for four
different longitudinal
velocities (u) using the
South Louisiana crude
oil in the IDHydroOPA
model with a simplified
rectangular channel (Jones
2018).
Figure 17. The volumetric
suspended sediment concentration
profiles for the five different
velocities used in IDHydroOPA
simulation of a simplified
rectangular channel, using a
sediment grain size of 50 (im where
z/H is the non-dimensional height
from the bed (Jones 2018).
-------
Just like the observation previously stated, the
distribution of settled and suspended OPA and oil
droplets in the heavy density oil case and medium
density oil cases are practically identical (see Jones
and Garcia, 2018). Additionally, the mass centroid of
the settled particles for the heavy density oil case is
slightly more upstream in the river than the medium
and light oil density cases, as expected. These
observations indicate that the amount of OPA that
settles out of the water column is most dependent
on suspended sediment concentration, mean flow
velocity, the river hydraulics, controlled by the flow
depth, bottom slope and channel roughness as well
as the distribution of oil droplet sizes, and not the
density of the oil itself.
In addition to suspended sediment concentration,
sediment particle size is also a parameter that has a
dramatic impact on OPA settling (Jones and Garcia,
2018). For small grain sizes, all the oil droplets will
coagulate and settle, but if the grain size is too large,
none will settle. This model only considers type 1
OPAs where the oil particles are covered by smaller
grain sizes, so OPAs cannot form when sediment size
approaches the size of oil particles. Figure 18 shows
how the fraction of oil that settles decreases with
increased grain size.
s
0.8
CO
c.
CO
0.7
-C
w
0.6
0)
o
XL
0.5
ro
D.
—
0.4
O
2
0.3
O
o
0.2
c
o
o
0.1
rc
ul
0
» • » •
10 20 30 40 50 60
Sediment Grain Size (pm)
70
80
The density of the oil was shown to not meaningfully
affect the rate of oil settling out of the water column.
Figure 19 shows how different density oil particles
settle with increased velocity in a 5-hour long
simulation.
0.8 ¦
0.7 ¦
0.6 -
0.5 -
0.4 •
0.3 ¦
0.2 -
0.1 ¦
« Density=820 g/L
Q Density=968 g/L
a Density= 1,000 g/L
O
B
(0
hm
LL
0.1 0.2 0.3 0.4 0.5 0.6
Velocity (m/s)
0.7
0.8
Figure 19. Fraction of oil settled with increased
velocity for three oil densities in a 5-hour
simulation of the lDHydroOPA model with a
simplified rectangular channel (from Jones and
Garcia, 2018).
Kalamazoo River lDHydroOPA Simulation with
HEC-RAS Channel Geometry
The lDHydroOPA model was linked with the more
complicated channel cross sections from a HEC-RAS
model of the Kalamazoo River. Using the flood flows
at the time of the 2010 Enbridge Line 6B pipeline
rupture, it was found that the leading edge of the oil
droplet and OPA plume reached Battle Creek, Ml
after approximately 2.6 hours (Jones and Garcia,
2018). An assumption of the model is that the
assumed weathered Cold Lake Blend entered the
Kalamazoo River atTalmadge Creek.
Just as observed in hypothetical straight rectangular
channel, with an increased velocity in the simple
representation of the Kalamazoo River, more OPAs
settled out of the water column for both the 0% and
17.4% mass loss cases (Figure 20). In agreement
with observations made previously for a straight
rectangular channel, increasing the velocity from
0.15 to 0.5 m/s results in a steep increase in
settled particles, but increasing the velocity past
0.5 m/s does not significantly increase the number
of settled particles (although a slight increase in
Figure 18. Fraction of total oil settled as OPA
after a 5-hour simulation for a range of sediment
sizes for the lDHydroOPA model with a simplified
rectangular channel (from Jones and Garcia, 2018).
-------
settled particles is noticed when the flow velocity is
increased from 1.0 m/s to 2.0 m/s). Thus, as expected
from the previous results, the Kalamazoo River
reaches an approximate asymptote on the number of
particles that will settle at a flow velocity of 0.5 m/s.
The change in the centroid of the settled particles
as a function of the velocity is shown in Figure 20
for the 0% and 17.4% mass loss cases. As observed
previously, as the velocity increases from 0.15 to 1.0
m/s, the location of the centroid of settling location
decreases. However, when the velocity increases to
2.0 m/s, the iocation of the centroid dramatically
increases. The increase in the location of the centroid
of the settled particles downstream and the slight
increase in number of settled particles observed
when the flow velocity is 2.0 m/s for both mass loss
cases is since at 2.0 m/s the suspended sediment
concentration is dramatically increased. Thus, at
2.0 m/s, the increase in the suspended sediment
concentration allows more of the oil particles to form
OPAs and settle out of the water column (Figure 20).
This increase in suspended sediment also allows
the OPAs to keep forming further downstream
(and thus settling out of the water column further
downstream), increasing the location of the centroid
of the settled particles (Figure 21). Additionally, there
are not significant differences in the results of the 0%
and 17.4% mass loss cases, which was expected due
to the previous findings that density did not play a
factor in how the oil particles settled out of the water
column. The slight differences in the results from the
two mass loss cases is most likely due to the slightly
difference oil droplet size.
0.8
03
0.7
.C
"D
o
.*05
r
<0
0>
Q_ CO 0.4
h-
O
raO-3
o
-C
o
0.2
(0
LL
0.1
0
O
~
8
o
~
8
0
Oo% Mass Loss
~ 17.4% Mass Loss
0.5
1 1.5
Velocity (m/s)
3000
2500
*£ 2000
¦a
O 1500
c
-------
modeled is large (near the 25-year event), the shear
stresses along the river bed were large enough
to continuously exceed the Shields number for
significant resuspension. However, this process
is not well understood, and the actual amount of
resuspension would rely on a statistical probability
of the oil coagulating with sediment in the river bed,
and the energy requirements to break the bonds
between oil and sediment in newly formed OPA.
0.25
0 1 2 3 4 5 6 7
Distribution of Particles (%)
20 30 40 50 60
Distance Downstream of Spill Site (m)
Figure 23. The longitudinal distribution of settled
OPA in the Kalamazoo River simulation
(Jones 2018).
Figure 24. The vertical distribution of suspended oil
droplets and suspended OPAs in the dimensionless
water column at the end of the Kalamazoo River
simulation (Jones 2018).
In the absence of the knowledge of the energy
requirement for breaking oil and sediment
connections in OPA, a better estimate of the possible
amount of oil that would be expected to be in the
river bed by the time the oil plume reaches Battle
Creek is to examine the vertical distribution of the
oil in the water column. Figure 24 shows the vertical
distribution of all suspended oil droplets and OPAs at
the end of the 2.6-hour simulation. The percentage
of all the oil introduced into the river that is in the
lower quarter of the water column at the end of the
simulation is 41.95% (23.20% oil droplets and 18.75%
are OPA). Thus, with the 0.4% that settled out at the
beginning of the simulation, a total of 42.35% of the
oil is close to settling on the river bed.
21
-------
Advances in Understanding, Lessons Learned
and Future Research Needs
The research conducted for this study and other related
research by the study team from 2016-18 reflects the
beneficial expansions of coordinated applied research
in OPA formation, transport, and fate following oil spills
in freshwater riverine environments. The following
advances were made for understanding OPA fate and
transport.
1) Mixing energy can be estimated for rivers based on
the Manning's Equation (Boufadel et al. 2019). The
energy-dissipation rate, which is used in modeling
oil droplet formation, can be calculated from river
characteristics of water depth, slope, and streambed
roughness. In rivers it can be typically higher than
what has been observed in the open sea with
regular waves.
Lessons learned/research needs: The turbulence in
most rivers is enough to breakup oil slicks into oil
droplets and is enhanced by natural variability in
rivers along the banks and bed from wood and rocks
as well as artificial structures such as dams. Simple
river characteristics reflecting energy dissipation
rates may be applied to nationally available stream
networks and applied in rapid time of travel models
such as ICWater.
2) Platy OPAs that form with clays may break apart
faster than expected with spherical shape
assumptions (Zhao et al., 2017). With increasing
mixing energy, the size of oil droplets and OPAs may
decrease (Zhao et al., 2017).
Lessons learned/research needs: Simple hydraulic
equations can be used as a rapid estimation tool for
OPA formation in rivers. Assumptions of OPA size
remaining stable during transport help to simplify
transport calculation but likely size distributions and
longevity of OPAs in natural channels are more
complicated.
3) Laboratory experiments with Cold Lake Blend
further the understanding of OPA formation.
Higher mixing energies yielded smaller oil droplets
and smaller OPA. Initial tests on kaolinite clay-only
aggregation and varying mixing speeds showed that
saline water generated larger aggregates than
freshwater, aggregate size distributions varied less
for lower concentrations of kaolinite clay, and
mixing speed had the least effect on aggregate size.
OPA aggregation tests with varying mixing speeds
are awaiting validation of an automated OPA
identification algorithm. Mean values of Cold Lake
Blend surface tension were 38.6 mN/m at the oil-air
interface and 21.3 mN/m at the oil-water surface.
A new algorithm applied by VTCHL to experimental
results shows potential to drastically increase the
variable space that can be explored but the method
needs more validation.
Lessons learned/research needs: The laboratory
results offer further confirmation that OPA conform
quickly after a spill depending on initial conditions.
A repository is needed for photomicrographs and a
method for automating the photogrammetry to
quantify the results of each time step for a range of
other oil and sediment types.
4) Linking existing OPA formation models with existing
2D sediment transport models can be difficult.
Lessons learned/research needs: During this study
proprietary model codes were encountered and
sharing code couldn't be overcome among
universities and federal agencies. Nonproprietary
codes should be developed to prevent this problem
from occurring again.
5) Rapid models such as the lDHydroOPA model offer
possibility for quickly estimating the potential for
OPA formation, transport, fate and resuspension
based on simplified channel hydraulics
(Jones and Garcia, 2018).
Lessons learned/research needs: Flow velocity is a
major parameter in determining how fast the oil
droplets mix with particles and form OPAs and
settle out of the water column. With an increased
flow velocity, the particles will tend to settle out
of the water column faster, due to the increase in the
suspended sediment concentration which increases
the rate of coagulation and its efficiency. The density
of the oil as a function of weathering was shown to
not meaningfully affect the number of particles that
settle out of the water column. The application of rapid
model for the 2010 Kalamazoo River spill resulted in
estimations of the amount of settled oil, and the
location of the centroid of the settled oil particles, that
would be expected for different flow velocities.
Currently, the developed model does not include
effects of oil breakup or re-entrainment of settled OPA
particles from the bottom of the river (Jones and Garcia,
2018). This model has the potential for these effects
to be included in future work, which would result in a
relaxation of the current assumptions made. However,
given a known type of oil, sediment in the river, and
flow conditions during a given oil spill event, this model
has the potential to aid in oil cleanup efforts, acting as
a simple screening model of where most of the settled
oil is located downstream of the spill location.
22
-------
Acknowledgments
We would like to acknowledge Civil and Environmental Engineering students (UIUC) past and present; USACE
programmer contributions, and NJIT students. Funding for this project was provided by the U.S. Environmental
Protection Agency (USEPA), Office of Research and Development (ORD), Regional Applied Research Effort Program,
USGS Interagency Agreement DW-014-92452901-0. Project management was provided by the Center for Environmental
Solutions & Emergency Response (CESER), Groundwater Remediation and Characterization Division (GCRD), Ada,
Oklahoma.
Notice
The U. S. Environmental Protection Agency through its Office of Research and Development funded the research
described here. This research brief has been subjected to the Agency's peer and administrative review and approved for
publication as an EPA document. Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.
23
-------
References Cited
Ajijolaiya, L. 0., P. S. Hill, A. Khelifa, R. M. Islam and K. Lee
(2006). "Laboratory investigation of the effects of mineral
size and concentration on the formation of oil-mineral
aggregates." Mar Pollut Bull 52(8): 920-927.
ASTM, 2013, Standard Test Method for Interfacial Tension
of Oil against Water by the Ring Method, D971-12, ASTM
International, PA, 3 p.
Boufadel, M. C., F. Fitzpatrick, F. Cui and K. Lee (2019).
"Computation of the Mixing Energy in Rivers for Oil
Dispersion." Journal of Environmental Engineering
145(10): 06019005.
Bragg, J. R. and E. H. Owens (1995). Shoreline cleansing
by interactions between oil and fine mineral particles.
Proc. Int. Oil Spill Conference.
Brunner, G.W., 2016, HEC-RAS River Analysis System
user's manual, version 5.0: U.S. Army Corps of Engineers,
Institute for Water Resources, Hydrologic Engineering
Center (HEC), Davis, CA. https://www.hec.usace.army.mil/
software/hec-ras/documentation/HEC-RAS%205.0%20
Users%20Manual.pdf
Carl Zeiss Microscopy GmbH, 2014, Software Guide,
ZEN 2 (blue edition): Gottingen, Germany, Carl-Zeiss
Microscopy GmbH, 309 p. https://www.unige.ch/
medecine/bioimaging/files/3914/3765/2064/ZEN_2_
blue_edition_-_Software_Guide.pdf
Dollhopf, R.H., Fitzpatrick, F.A., Kimble, J.W., Capone,
D.M., Graan, T.P., Zelt, R.B., Johnson, R., 2014, Response
to heavy, non-floating oil spilled in a Great Lakes river
environment: a multiple-lines-of-evidence approach for
submerged oil assessment and recovery: Proceedings,
2014 International Oil Spill Conference, Savannah, GA,
May 7-9, 2014, p. 434-448. http://ioscproceedings.org/
doi/pdf/10.7901/2169-3358-2014.1.434
Fitzpatrick, F.A., Boufadel, M.C., Johnson, R., Lee, K.,
Graan, T.P., Bejarano, A.C., Zhu, Z., Waterman, D., Capone,
D.M., Hayter, E., Hamilton, S.K., Dekker, T., Garcia,
M.H., and Hassan, J.S., 2015a, Oil-particle interactions
and submergence from crude oil spills in marine and
freshwater environments—Review of the science and
future science needs: USGS Open-File Report 2015-1076,
33 p.
Fitzpatrick, F.A., Johnson, R., Zhu, Z., Waterman, D.,
McCulloch, R., Hayter, E., Garcia, M.H., Boufadel, M.C.,
Dekker, T., Hassan, J.S., Soong, D.T., Hoard, C.J., and Lee,
K., 2015b, Integrated modeling approach for fate and
transport of submerged oil and oil-particle aggregates
in a freshwater riverine environment: Proceedings,
SEDHYD2015 Conference, Reno, NV.
Garcia, M.H. (Editor), 2008, Manual of Engineering
Practice 110: Sedimentation Engineering: Processes,
Measurements, Modeling and Practice, 1150, ASCE,
Reston, VA.
Garcia, T., Jackson, P.R., Murphy, E.A., Valocchi, A.J., and
Garcia, M.H., 2013, Development of a fluvial egg drift
simulator to evaluate the transport and dispersion of
Asian carp eggs in rivers: Ecol. Modell. 263(l):211-222.
Great Lakes Commission, 2015, Issues and trends
surrounding the movement of crude oil in the Great
Lakes-St. Lawrence River Region: Ann Arbor Ml, February
2015, 47 p.
Gustitus, S. A. and T. P. Clement (2017). "Formation,
Fate, and Impacts of Microscopic and Macroscopic Oil-
Sediment Residues in Nearshore Marine Environments: A
Critical Review." Reviews of Geophysics 55(4): 1130-1157.
Hayter, E.J., R. McCulloch, T. Redder, and M. Boufadel.
2014. "Modeling the Transport of Oil Particle Aggregates
and Mixed Sediment in Surface Waters," Letter Report.
Environmental Laboratory, U.S. Army Engineer Research
and Development Center, Vicksburg, MS.
Hamrick, J. M. (1992). "A Three-Dimensional
Environmental Fluid Dynamics Computer Code:
Theoretical and Computational Aspects", Special
Report No. 317 in Applied Marine Science and Ocean
Engineering, Virginia institute of Marine Science,
Gloucester Point, VA., pp. 1-64.
Hill, P., A. Khelifa and K. Lee (2002). "Time scale for oil
droplet stabilization by mineral particles in turbulent
suspensions." Spill Science & Technology Bulletin 8(1):
73-81.
Hoard, C.J., Fowler, K.K., Kim, M.H., Menke, C.D., Morlock,
S.E., Peppier, M.C., Rachol, C.M., and Whitehead, M.T.,
2010, Flood-inundation maps for a 15-mile reach of the
Kalamazoo River from Marshall to Battle Creek, Michigan,
2010: U.S. Geological Survey Scientific Investigations Map
3135, 6 p. pamphlet, 6 sheets, scale 1:100,000.
24
-------
Jones, L., 2018, Development of a rapid response riverine
oil-particle-aggregate (OPA) formation, transport, and
fate model: University of Illinois at Urbana-Champaign
Master's Thesis.
Jones, L., and Garcia, 2018, Development of a rapid
response riverine oil-particle-aggregate (OPA) formation,
transport, and fate model: J. Environ. Eng. 144(12):
04018125-1 to -16.
Jones, C., and Lick, W., 2001, SEDZLJ, A sediment
transport model: Final Report, Santa Barbara, CA:
Department of Mechanical and Environmental
Engineering, University of California. May 29, 2001.
Khelifa, A., P. S. Hill and K. Lee (2005). A comprehensive
numerical approach to predict oil-mineral aggregate
(OMA) formation following oil spills in aquatic
environments. International Oil Spill Conference,
American Petroleum Institute.
Khelifa, A., K. Lee, P. Hill and L. Ajijolaiya (2004).
Modelling the effect of sediment size on OMA formation.
27th Proceedings of Arctic and Marine Oil Spill Program
Technical Seminar, Edmonton, Alberta, Canada,
Environmental Canada.
Khelifa, A., P. Stoffyn-Egli, P. S. Hill and K. Lee (2002).
"Characteristics of oil droplets stabilized by mineral
particles: effects of oil type and temperature." Spill
Science & Technology Bulletin 8(1): 19-30.
Lee, K., Bugden, J., Cobanli, S., King, T., Mclntire, C.,
Robinson, B., Ryan, S., Wohlgeschaffen, G., 2012, UV-
Epifluorescence Microscopy Analysis of Sediments
Recovered from the Kalamazoo River: Center for Offshore
Oil, Gas and Energy Research (COOGER) Technical Report
for the Enbridge Line 6B Spill, October 24, 2012.
Lee, K., M. Boufadel, B. Chen, J. Foght, P. Hodson, S.
Swanson and D. Venosa Albert (2015). The Behaviour
and Environmental Impacts of Crude Oil Released into
Aqueous Environments. The Royal Society of Canada,
Ottawa.
Lee, K., P. Stoffyn-Egli, G. H. Tremblay, E. H. Owens, G.
A. Sergy, C. C. Guenette and R. C. Prince (2003). "Oil-
Mineral Aggregate Formation on Oiled Beaches: Natural
Attenuation and Sediment Relocation." Spill Science &
Technology Bulletin 8(3): 285-296.
Lee, K., Lunel, T., P. Wood, R. Swannell and P. Stoffyn-
Egli (1997a) Shoreline clean up by acceleration of clay-
oil flocculation processes. In: Proceedings of the 1997
International Oil Spill Conference (Prevention, Behavior,
Control and Cleanup. Fort Lauderdale, Florida, April 7-10,
1997. pp. 235-240.
Lee, K., Lunel, T., Wood, P., Swannell, R., Stoffyn-Egli, P.,
1997b. Clay-oil flocculation during surf washing at the
Sea Empress incident. In: Proceedings of the 20th Arctic
and Marine Oil Spill Program Technical, Seminar, pp.
1085-1103.
Lee, K., P. Stoffyn-Egli and E. H. Owens. 2002. The OSSA
II pipeline oil spill: Natural mitigation of a riverine oil
spill by oil-mineral aggregate formation. Spill Science &
Technology Bulletin 7:149-154.
Lee, K. and P. Stoffyn-Egli (2001) Characterization of oil-
mineral aggregates. Proceedings of the 2001 International
Oil Spill Conference, Tampa Florida, USA, March 26-29,
2001. pp. 991-996.
Lee, K., A.M. Weise and T. Lunel (1997c) Marine oil spills:
Enhanced biodegradation with mineral fine interaction.
In Situ and On-Site Bioremediation 4: 365-370.
LimnoTech, 2015, Kalamazoo River hydrodynamic and
sediment transport model documentation: LimnoTech
Technical Report, January 26, 2015.
Owens, E. H., B. Humphrey and G. A. Sergy (1994).
"Natural cleaning of oiled coarse sediment shorelines in
Arctic and Atlantic Canada." Spill Science & Technology
Bulletin 1(1): 37-52.
Owens, E. H. and K. Lee (2003). "Interaction of oil and
mineral fines on shorelines: review and assessment."
Marine Pollution Bulletin 47(9): 397-405.
Perkey, D.W., Smith, S.J., and Kirklin, T., 2014, Cohesive
sediment erosion field study: Kalamazoo River,
Kalamazoo, Michigan: USACE ERDC Letter Report, April
24, 2014.
Reneau, P.C., Soong, D.T., Hoard, C.J., and Fitzpatrick, F.A.,
2015, Hydrodynamic-assessment data associated with the
July 2010 Line 6B spill into the Kalamazoo River, Michigan,
2012-14: U.S. Geological Survey Open-File Report 2015-
1205, 26 p., http://dx.doi.org/10.3133/ofr20151205
Rouse, H., 1938, Experiments on the mechanics of
sediment suspension: in , Proc. 5th International Congress
on Applied Mechanics, Hartog & Peters, Editors, p. 550-
554, published by John Wiley & Sons, Inc.
Sun, J., Khelifa, A., Zheng, X., Wang, Z., So, L.L, Wong, S.,
Yang, C., and Fieldhouse, B., 2010, A laboratory study
on the kinetic of formation of oil-suspended particulate
matter aggregates using the NIST-1941b sediment: Mar.
Pollut. Bull. 60(10):1701-1707.
U.S. Environmental Protection Agency, 2016, FOSC
Desk report for the Enbridge Line 6b oil spill, Marshall,
25
-------
Michigan, U.S. Environmental Protection Agency, April
2016, 241 p.
Visser, A.W., 197, Using random walk models to simulate
the vertical distribution of particles in a turbulent water
column: Mar. Ecol. Prog. Ser. 158:275-281.
Waterman, D.M., Fytanidis, D., and Garcia, M.H., 2015,
Kalamazoo River in situ flume bed erosion study: Ven Te
Chow Hydrosystems Laboratory, Civil Engineering Studies,
Hydraulic Engineering Series No. 105.
Waterman, D.M., and Garcia, M.H., 2015, Laboratory
tests of oil-particle interactions in a freshwater riverine
environment with Cold Lake Blend Weathered Bitumen,
2015, Ven Te Chow Hydrosystems Laboratory, Civil
Engineering Studies, Hydraulic Engineering Series No.
106.
Zhang, H.P., Khatibi, M., Zheng, Y., Lee, K., Li, Z.K., and
Mullin, J.V., 2010, Investigation of OMA formation and the
effect of minerals: Mar. Pollut. Bull. 60(9):1433-1441.
Zhao, L., M. C. Boufadel, E.E. Adams, S. A. Socolofsky,
T. King, K. Lee, and T. Nedwed, 2015, Simulation of
scenarios of oil droplet formation from the Deepwater
Horizon blowout: Marine Pollution Bulletin. 101(1):304-
319.
Zhao, L., Boufadel, M.C., Geng, X., Lee, K., King, T.,
Robinson, B., and Fitzpatrick, F., 2016, A-DROP: A
predictive model for the formation of oil particle
aggregates (OPA): Marine Pollution Bulletin 106(1-2):245-
59.
Zhao, L., J. Torlapati, T. King, B. Davidson, M. Boufadel
and K. Lee (2014a) A numerical model to simulate the
droplet formation process resulting from the release of
diluted bitumen products in marine environment. 2014
International Oil Spill Conference Proceedings, Savannah,
Georgia, USA. May 5-8, 2014. pp. 449-462.
Zhao, L., M. C. Boufadel, J. Katz, G. Haspel, K. Lee, T. King
and B. Robinson (2017). "A new mechanism of sediment
attachment to oil in turbulent flows: Projectile particles."
Environmental Science & Technology 51(19): 11020-
11028.
Zhao, L., J., Torlapati, M. C., Boufadel, T. King, B. Robinson
and K. Lee (2014b). "VDROP: A comprehensive model
for droplet formation of oils and gases in liquids-
Incorporation of the interfacial tension and droplet
viscosity." Chem. Eng. J. 253: 93-106.
Zhu, Z., 2015, Modeling the transport and fate of oil-
particle aggregates after an oil spill in inland waterways:
Ph.D. Dissertation, Dept. of Civil and Environmental
Engineering, Univ of Illinois at Urbana-Champaign.
Zhu, Z., and Garcia, M.H., 2015, Three-dimensional
numerical modeling of Morrow Lake in Kalamazoo River,
Michigan: Ven Te Chow Hydrosystems Laboratory Report.
Zhu, Z., Waterman, D., and Garcia, M.H., 2018, Modeling
the transport and fate of oil-particle aggregates
resulting from an oil spill in a freshwater environment:
Environmental Fluid Mechanics https://doi.org/10.1007/
si 0652-018-9581-0.
26
------- |