Environ Sci Pollut Res

DOI 10.1007/s11356-016-7931-3

RESEARCH ARTICLE

Two-dimensional numerical modelling of sediment and chemical

constituent transport within the lower reaches of the Athabasca
River
Shalini Kashyap 1 & Yonas Dibike 1 & Ahmad Shakibaeinia 1 & Terry Prowse 1 &
Ian Droppo 2

Received: 8 May 2016 / Accepted: 17 October 2016

# Springer-Verlag Berlin Heidelberg 2016

Abstract Flows and transport of sediment and associated river bed concentration levels, resulting in higher outflows with
chemical constituents within the lower reaches of the increases in their concentration in the bed sediment.
Athabasca River between Fort McMurray and Embarrass
Airport are investigated using a two-dimensional (2D) numeri- Keywords Athabasca River . Water quality . Sediment
cal model called Environmental Fluid Dynamics Code (EFDC). transport . Numerical modelling . River engineering . EFDC
The river reach is characterized by complex geometry, including
vegetated islands, alternating sand bars and an unpredictable
thalweg. The models were setup and validated using available The Athabasca River is located in Alberta, Canada and originates
observed data in the region before using them to estimate the at the base of the Columbia Icefield in Jasper National Park. It has
levels of cohesive sediment and a select set of chemical constit- alengthofmorethan1500 km and drains~156,000km2 areafrom
uents, consisting of polycyclic aromatic hydrocarbons (PAHs) its headwaters, through to the Peace–Athabasca Delta. The lower
and metals, within the river system. Different flow scenarios reaches of the river are situated between Fort McMurray and the
were considered, and the results show that a large proportion delta in an area that has seen significant changes in land use due to
of the cohesive sediment that gets deposited within the study increasing industrial activities that have been taking place since
domain originates from the main stem upstream inflow bound- 1967 (Dillon et al. 2011). These reaches pass through natural oil
ary, although Ells River may also contribute substantially during sands formations where bitumen is found very close to the earth
peak flow events. The floodplain, back channels and islands in surface and is likely to be affected by natural erosional and depo-
the river system are found to be the major areas of concern for sitional processes. The river geometry is complex, as it contains
deposition of sediment and associated chemical constituents. vegetated islands, alternating sand bars, and has an unpredictable
Adsorbed chemical constituents also tend to be greater in the thalweg.Between FortMcMurray and Old Fort,the river has been
main channel water column, which has higher levels of total characterized as being somewhere between a meandering and a
suspended sediments, compared to in the flood plain. braided river (Conly et al. 2002).
Moreover, the levels of chemical constituents leaving the river Sediment transport within the lower Athabasca River
system are found to depend very much on the corresponding (LAR) is of particular interest due to its ability to transport
chemical constituents such as polycyclic aromatic hydrocar-
Responsible editor: Marcus Schulz bons (PAHs) and metals. The river bed within this reach is
composed of a mixture of gravel, sand and cohesive sediment
* Shalini Kashyap (Doyle 1977; Shaw and Kellerhals 1982), and it is the cohe-
shalini.kashyap@gmail.com sive sediment that favours the adsorption of chemical constit-
uents (Garcia-Aragon et al. 2011). To date, there have been
1
Environment and Climate Change Canada, Water & Climate Impact
few studies that have estimated sediment budgets within the
Research Centre, 3800 Finnerty Road, Victoria, BC V8P 5C2, LAR (Carson 1990; Conly et al. 2002), and those studies have
Canada been based on data collected from limited field sampling sta-
2
Environment and Climate Change Canada, Canada Centre for Inland tions. The sediment budget carried out by Carson (1990) for
Waters, 867 Lakeshore Rd, Burlington, ON L7S 1A1, Canada the period between 1976 and 1984 was based on suspended
 Environ Sci Pollut Res

sediment data collected at sampling stations within the main River between Windfall to Lake Athabasca, again using only rect-
stem and tributaries and accounted for sediment loadings from angular cross sections to represent channel geometry. Andrishak
the upstream sources and erosion from the main stem bed and et al. (2008) not only used River1D to model both hydraulics and
banks. The sediment budget of Conly et al. (2002) also ice for a ~80 km reach between Fort McMurray and Bitumount but
accounted for sediment originating from erosion of the main also relied on limited cross-sectional data. More recently,
stem bed and banks. The sediment budget in the present study Shakibaeinia et al. (2014, 2016) completed a 1D modelling of the
uses numerical modelling to estimate sediment entering and Athabasca River between Fort McMurray and Old Fort to predict
leaving the river reach based on continuous simulations. It trends and levels in dissolved oxygen by incorporating additional
also identifies depositional areas and quantities of cohesive detailed bathymetry from Geoswath and other data sources.
sediment originating directly from the main stem upstream Detailed 2D numerical models such as River2D (Van der Vinne
boundary near Fort McMurray, as well as the tributary sources 2003) and CRISSP2D (Wojtowicz 2010) have also been used to
during a peak flow event. model hydraulics and ice, although these studies have focussed on
Selected chemical constituents consisting of two PAHs and shorter reaches (<10 km in length) within the LAR. To date, there
two metals are considered for the transport modelling runs in have only been two attempts to model chemical constituent trans-
this study. The first PAH is the parent PAH, phenanthrene, and port along the Athabasca River. Both of these attempts not only
the second is the alkylated PAH c1-benzo[a]anthracene/chrys- involved using the 1D Water Quality Analysis Simulation
ene. Both of these PAHs are recommended for monitoring by Program (WASP) to model naphthenic acids and other priority
the Joint Canada Alberta Implementation Plan for Oil Sands PAHs (Kannel and Gan 2013; McCauley 1997) but also relied on
Monitoring (Environment Canada 2011). Phenanthrene is a limited cross-sectional bathymetry data.
three-ring PAH (molecular mass = 178 g/mol) and has been Given the present and projected changes in land use in the
classified by the Environmental Protection Agency as a vicinity of the LAR, there is a need for increased understanding
Priority Pollutant (Irwin 1997). Bed sediment concentration of the spatial variability and trends in both sediment and chem-
levels of phenanthrene measured by Regional Aquatics ical constituent distributions in the river system. Such knowl-
Monitoring Program (RAMP) (2013) between October 1997 edge is difficult to obtain from field measurements alone due to
and September 2012 have sometimes been greater than the their low spatial and temporal resolution and the high costs of
maximum level recommended by Canadian Council of obtaining such measurements. This paper demonstrates the use
Resource and Environment Ministers (CCREM) (1999) for of high-resolution 2D numerical modelling as a tool to help
the protection of aquatic life (0.0419 mg/kg). The concentrations predict sediment and chemical constituent distributions and pat-
have also been shown to increase as one moves downstream terns along the LAR. The 2D numerical model developed in this
through the natural bitumen deposit within tributaries (Yergeau study incorporates both high-resolution Geoswath bathymetry
et al. 2013). C1-Benzo[a]anthracene/chrysene is an alkylated and LIDAR topography data of the LAR, which greatly im-
PAH (molecular mass = 242 g/mol), and due to its relatively proves the reliability of the results. The specific objectives of
high molecular weight, it is expected to be persistent in the river. this modelling study are as follows:
Kirk and Manzano (2013) have also identified that in 2012,
levels of c1-benzo[a]anthracene/chrysene were found to be high 1. To calculate sediment budgets for peak flow events and
in snow samples close to oil sands developments. The two identify the corresponding potential locations for high levels
metals considered in the transport modelling study are lead of sediment deposition within the main stem of the LAR
and vanadium. While their concentration levels in the bed 2. To assess (based on numerical results) the percentage con-
sediment and water column measured between October 1997 tribution of each tributary of the LAR towards deposition-
and September 2012 by RAMP (2013) did not exceed CCREM al loads during a peak flow event
(1999) guidelines, they are still considered important com- 3. To assess sediment and associated chemical constituents
pounds for an environmental monitoring program and are with- budgets within the LAR and their sensitivity to sediment
in the list of parameters to be monitored by the Joint Oil Sands bed concentrations and sediment parameters for various
Monitoring Program (JOSMP) (Environment Canada 2011). flow scenarios
Onlyalimitednumberofstudieshavebeencompletedinvolving
numerical modelling of river flow and transport processes within
the LAR and the majority of these studies have involved the use of
1D models to simulate the flow and transport processes. Khanna Methodology
(2003) made an early attempt to model flow between Fort
McMurray and Embarrass Airport using a 1D model called The study reach and bathymetry
CDG1-D which incorporated rectangular sections to approximate
river cross sections. Pietroniro et al. (2011) also used a 1D model The lower Athabasca River reach considered in this study
called River1D to model flow along ~1000 km of the Athabasca extends for about 160 km from just downstream of Fort
Environ Sci Pollut Res

McMurray to approximately 40 km downstream of the 6. Digital Elevation Model (DEM) data for topography of
Firebag River confluence (see Fig. 1). The reach has complex the flood plain and islands not covered by LIDAR data
bathymetry and an unpredictable thalweg and receives lateral (Geobase 2012)
inflows from up to six major tributaries. The tributaries con-
sidered for model development are the Ells, Firebag, MacKay, Post-processing techniques were used to transform the 127
Muskeg, Steepbank and Tar rivers. The river bathymetry and rectangular cross sections into parabolic-type sections in order
floodplain topography used for setting up the models were to prevent sudden changes in water surface elevations (WSEs)
compiled from the following six data sources (in order of due to contraction and expansion of the flow. Interpolated
priority in case of overlap): sections with spacings of 100 to 200 m were also prepared
using the HEC-RAS software. The bathymetry data from all
1. Approximately 40 km of high-resolution Geoswath bathym- these sources were combined to construct a continuous ba-
etry (5 m spacing) collected and prepared for this study by thymetry with a resolution ranging from 10 to 25 m, which
Environment Canada (http://jointoilsandsmonitoring.ca) was then interpolated onto the model computational grids.
2. Fifty-four detailed surveyed bathymetry cross sec-
tions between Crooked Rapids and Steepbank
River (Hicks 2011) The numerical model
3. Six high-resolution (variable spacing) surveyed bathyme-
try reaches (collected with a Raytheon Fathometer echo The 2D numerical modelling system used for this study
sounder) from CEMA (2012) is the Environmental Fluid Dynamics Code (EFDC)
4. One hundred twenty-seven rectangular bathymetry sec- (Hamrick 1992), with the EFDC Explorer 7.1 pre-/
tions (1 km intervals) located between Steepbank River post-processor developed by DSI Consulting Group
and Embarrass Airport from the Mackenzie River Basin (Craig 2013). The EFDC model has been extensively
Hydraulic Model (Pietroniro et al. 2011) tested and used in over 100 modelling studies (Ji
5. High-resolution LIDAR data (5 m spacing) for the flood- 2008). Some river applications using EFDC include
plain and islands obtained from Alberta Environment and the following: the San Joaquin River in California (for
Sustainable Resources Development (AESRD) (2014) sediment transport and water quality), the Columbia

Fig. 1 Location of the lower

Athabasca River reach including
the inflow/outflow boundaries
and the main tributary locations
considered in this study
 Environ Sci Pollut Res

River in Portland Oregon (for toxin transport), Third– where ws is the settling velocity (m/s) which may be set
Fourth Creeks in Tennessee, the Yangtze River in China to a constant value or may be determined based on a
(for toxin transport), the Red River in Vietnam (for known method such as Huang and Metha etc. (Hamrick
hydrodynamics), the Ohio River in Ohio and the 2007b). Pk is a probability of deposition function which
Belleville Locks Dam River in Ohio (for hydrodynam- is described in Hamrick (2007b). Sd is the near bed
ics) (Dynamic Solutions International LLC 2016). A re- solid concentration (mg/L); τb is the bed shear stress
cent application of EFDC was also completed for the (N/m 2 ); and τ cd is the critical bed shear stress for
Northern Saskatchewan River which investigated water deposition.
quality (McDonald and Muricken 2009). EFDC solves The resuspension flux (J0r, g/m2s) of cohesive sediment is
the depth averaged (hydrostatic), free-surface, Reynolds- given by:
averaged Navier–Stokes equations (RANS) using a
second-order accurate finite difference scheme on a
a
dme τ b −τ ce
staggered grid. The time integration employs a three- J r0 ¼ wr S r ¼ : τ b ≥ τ ce ð3Þ
time level finite differencing scheme with internal- dt τ ce
external mode splitting. The external mode solution is
semi-implicit. The internal momentum equation is im- where wr is the rate of resuspension (m/s); Sr is the dry density
plicit with respect to vertical diffusion. Time splitting of the sediment (g/m3); dme/dt is the base erosion rate (g/m2s);
in the three-time level scheme is controlled by insertion τce is the critical shear stress for surface erosion; and a is the
of a second-order accurate two time-level trapezoidal erosion exponent.
step (Ji 2008). The main EFDC modules used for this
study included hydrodynamics, sediments and chemical
constituent transport. For brevity, the hydrodynamic EFDC chemical constituent model
equations in the EFDC model are not presented in this
paper but can be found in Hamrick (2007a). Chemical constituents both in the sediment bed and in
the water column were considered. The model is capa-
EFDC sediment transport model ble of simulating the transport of chemical constituents
adsorbed to suspended sediment, those freely dissolved
EFDC is able to simulate cohesive and noncohesive suspended in the water column and dissolved chemical constituents
sediment transport of multiple size classes. The general 2D which are complexed with dissolved organic carbon
transport equation for suspended sediment is (Hamrick 2007b): (DOC) (Hamrick 2007b). The fractions of chemical con-


 stituents in dissolved and adsorbed phase are determined
δHC i δHuC i δHvC i δ δC i δ δC i
þ þ ¼ HK x þ HK y þ QC by equilibrium partitioning coefficients. For the portion
δt δx δy δx δx δy δy
of the chemical constituents freely dissolved in the wa-
ð1Þ ter phase, the model accounts for advection, vertical
diffusion, desorption from the sediment and other dis-
where H is the water depth; u and v are the depth-averaged velocity
solved phases (DOC) and adsorption to the sediment
components in the x and y directions respectively; Ci is the depth-
and dissolved phases (DOC). For the portion of the
averaged concentration of component i; Kx and Ky are the diffusion
dissolved chemical constituents adsorbed to dissolved
coefficients in the x and y directions respectively; and QC repre-
constituents in the water phase (DOC), the model ac-
sents the external source (erosion) and sink (deposition) terms.
counts for advection, vertical diffusion, sorption to the
The near bed equilibrium of noncohesive suspended sedi-
dissolved phase and desorption from the dissolved
ment can be determined by a number of methods including the
phase. For the portion of the chemical constituents
equations of García and Parker (1991), Van Rijn (1984) and
adsorbed to suspended sediment in the water column,
Smith and McLean (1977). These formulas all require that the
the model accounts for advection, vertical diffusion, ad-
critical Shields shear stress be specified for each sediment size
sorption to the sediment phase and desorption from the
class. The Van Rijn (1984) equation was chosen for all simu-
sediment phase.
lations in this study.
The interaction of chemical constituents between the
For cohesive sediment, the deposition flux (Jd, g/m2s) at
water column and the bed plays a critical role in eval-
the bed is given by:
uating the net fluxes of chemical constituents leaving
 the river system. In EFDC, the boundary condition at
−ws Pk S d : τ b ≤ τ cd the water column sediment bed interface is given by
Jd ¼ ð2Þ
0 : τ b ≥ τ cd (Hamrick 2007b):
Environ Sci Pollut Res

0 0 X j 11
Ab X X
 

i
J
 C þ
w C
j d AA
− δz C − w i i
f C ¼ @ max J i X i ; 0 þ εmax SBS ; 0 @
H i s s i SBS s
ρs
i φ
SB
0 0 X j 11 0 0 X j 11
X
 i 

i
J
 C þ
w C d X

i
J
 C þ
w C
j d AA
þ @ max J X ; 0 þ εdep min
i SBS
;0 @
j AA þ @ εmax SBB
;0 @
i SBS s
ρs
i φ i ρs
i φ
WC SB
0 0 X j 11 0 0 X 11 0 0 X 11 ð4Þ
X

 C þ C Cw þ Cj Cw þ Cj
@ εdep min J iSBB w j d AA j d j d
þ ;0 @ þ @maxðqw ; 0Þ@ AA þ@minðqw ; 0Þ@ AA
i ρis φdep φ φdep
WC SB WC
00 X 1 0 X 1 1
Cw þ Cj
j d
Cw þ Cj
j d
− qdiff @@ A −@ A A
φdep φ
WC SB

where Ab is the vertical eddy diffusivity (m2/s), which is cal- the bathymetric and topographic data by a plain representing
culated based on the Mellor and Yamada (1982) turbulence the high flow water surface plus 1 m of freeboard. The first
closure scheme, modified by Galperin et al. (1988) and mesh was a fine Cartesian grid with 46,203 elements (element
Blumberg et al. (1992). H is the total water column depth; C size of 75 × 75 m). This grid was used for modelling the peak
is the total concentration of chemical constituent (mg/L), and flow scenario sediment transport runs, which were used to
the subscripts w and d refer to the freely dissolved and the determine depositional patterns and sediment budgets. The
dissolved organic carbon bound concentrations respectively; second mesh was a curvilinear grid with 10,985 elements
ws is the settling velocity associated with the particulates (element size of ~60 × 300 m). This relatively coarser grid
(m/s); fs is the sediment bound fraction of chemical constitu- was used for all of the main chemical constituent runs due to
ent; JSBS and JSBB are fluxes of suspended load and bedload their higher computational requirements. In summary, valida-
respectively, defined as positive from the bed (kg/(m2s)); Xs is tion runs for hydraulics and transport of sediment were com-
the mass of chemical constituent sorbed to sediment class i per pleted on both grids, while the scenarios for sediment trans-
mass of sediment; ε is the bed sediment void ratio; ρs is the port (including hydrodynamics) were run on the fine mesh and
sediment density (kg/m3); φ is the bed sediment porosity; qw the validation and scenarios for chemical constituents
is the water specific discharge due to bed consolidation and
groundwater interaction, defined as positive from the bed
(m/s); and qdiff is the diffusion velocity, which incorporates
molecular diffusion, hydrodynamic dispersion and biological
induced mixing (m/s). The subscripts WC and SB refer to
conditions in the water column immediately above the bed
and conditions in the top layer of the sediment bed respective-
ly. The subscript dep refers to newly deposited sediment.
As can be seen through Eq. 4, the rate at which chemical
constituents are transferred between the bed sediment and the
water column depends on the suspended load and bedload
fluxes JSBS and JSBB, the water specific discharge qw and the
diffusion velocity qdiff.

Numerical setup

Model grids

Two sets of grids were used for the numerical simulations

(Fig. 2). The meshes for both types of grids were created from Fig. 2 Plan view of the modelled reach showing a the lateral extent as
a shape file containing both the lower Athabasca River main well as close-ups of the computational grid resolutions for the b fine and c
channel and flood plain. The shape file was created by cutting coarse grids
 Environ Sci Pollut Res

(including hydrodynamics and sediments) were run on the all cases. The time steps for the various runs varied from 0.25
coarse grid. It should be noted that trials were initially com- to 4 s for the coarse grid and from 0.5 to 4 s for the fine grid.
pleted to verify that the size of the grids were adequate, and the
results presented for the second mesh (curvilinear mesh) com- Hydraulic and sediment transport simulations
pared well to a local model completed on a 25 × 25 m element
size mesh. The constant flow hydraulic validation runs corresponded to con-
ditions on 15 September 2004, a date with simultaneous field
Simulation runs and basic setup measurements of WSE and depth averaged velocity data obtained
from CEMA (2012). Boundary conditions for the validation runs
Model simulations consisting of steady state validation runs are given in Table 1. The constant flow sediment transport vali-
and unsteady scenario runs were conducted for hydrodynam- dation runs with an erodible bed (where bedload was present)
ics, sediment transport and chemical constituent transport in corresponded to conditions on 18 September 2004. This date
the LAR. The following three cases were defined for the sce- was chosen for availability of field data for both boundary
nario runs: case I (historical peak flow), case II (extreme peak conditions and validation close to this date. The distribution of
flow), and case III (historical constant low flow). Each of these bed sediment along the modelled reach was determined from
cases was run for a period of 23 days in order to understand cores obtained from the Water Survey of Canada (2013) and
the effects of high loadings of sediment and/or chemical con- Shaw and Kellerhals (1982). In upstream of Shott Island (0–
stituents under high and low flow conditions. 124 km downstream of Fort McMurray), the bed sediment was
For all simulations, the main stem upstream boundary for considered to be a trimodal mixture of gravel, sand and cohesive
flow and transport of sediment and chemical constituents was sediment, while in downstream of Shott Island, it was considered
based on measurements (either direct or determined from rat- a bimodal sand and cohesive sediment mixture. For these 2D
ing curves) at the Water Survey of Canada Station 07DA001 models, the sand was considered uniform with a D50 of
located slightly downstream of Fort McMurray (Water Survey 0.16 mm. The gravel was also considered uniform with a D50 of
of Canada 2013; RAMP 2013). The downstream water level 1.5 cm. The properties (settling velocity (ws) [0.0001 m/s] and
boundary, however, was taken from a well validated 1D hy- critical shear stresses for erosion (τcε) [0.40 N/m2] and deposition
drodynamic model of the lower Athabasca River setup using (τcd) [0.35 N/m2]) of the cohesive sediment were estimated from a
the Mike-11 software (Shakibaeinia et al. 2014). Lateral in- combination of measurements in laboratory experiments by
flows were also considered from six tributaries (Ells, Firebag, Droppo et al. (2015, 2016) and Garcia-Aragon et al. (2011) and
MacKay, Muskeg, Steepbank and Tar Rivers), and their data calibration runs. The τcε and τcd adopted for this model setup were
were obtained from a combination of RAMP ( 2013) and greater than those found through the experiments of Droppo et al.
Water Survey of Canada (2013) hydrometric stations close (2015, 2016), and Garcia-Aragon et al. (2011) as the EFDC de-
to the tributary confluences to the lower Athabasca River main fines the τcd to be the upper limit, above which no deposition
stem. Wetting and drying were set on for all simulations, and occurs, and τcε to be when substantial erosion occurs (Hamrick
the effective bed roughness, ks value of 0.5 mm, was used in 2007b). The cohesive reference surface erosion rate (Wrsp) (0.5 g/

Table 1 Boundary conditions for the different constant flow validation cases

Upstream Inflow Ells River Firebag River MacKay River Muskeg River Steepbank River

Hydraulics validation (15 September 2004), WSE at outlet = 218.00 m

Flow (m3/s) 1110.00 2.17 10.60 1.49 0.78 1.87
Sediment transport validation (18 September 2004), WSE at outlet = 218.05 m
Flow (m3/s) 1090.00 2.35 21.80 1.95 1.14 2.05
Cohesive suspended sediment (mg/L) 142.55 5.39 7.13 3.25 4.41 7.29
Noncohesive suspended sediment (mg/L) 15.84 0.60 0.79 0.36 0.49 0.81
Chemical constituent validation (7 September 2011), WSE at outlet = 216.94 m
Flow (m3/s) 588.00 4.42 15.10 7.67 0.48 1.18
Cohesive suspended sediment (mg/L) 5.08 10.48 3.84 13.45 1.09 3.71
Noncohesive suspended sediment (mg/L) 45.71 1.16 0.43 1.49 0.12 0.41
C1-Benzo[a]anthracene/chrysene (μg/L) 0.002535 0.007320 0.001690 0.01110 0.000203 0.003790
Phenanthrene (μg/L) 0.001150 0.002090 0.000920 0.001780 0.001290 0.002220
Lead (μg/L) 0.728300 0.300000 0.136000 0.200000 0.100000 0.129500
Vanadium (μg/L) 2.846700 1.350000 0.663000 0.959500 0.387000 3.712500
Environ Sci Pollut Res

m2s) was also first estimated from measured data, and the final (ws) = 0.0001 m/s, τcε = 0.40 N/m 2, τ cd = 0.35 N/m 2,
value is determined through calibration. Total suspended sedi- Wrsp = 0.5 g/m2s and D50 (sand) = 0.16 mm. The LSP set
ment (TSS) measurements were obtained from RAMP (2013), used the following values: ws = 0.001 m/s, τcε = 0.40 N/m2,
although the percentage of cohesive sediment in the TSS was not τcd = 0.35 N/m2, Wrsp = 0.5 g/m2s and D50(sand) = 0.75 mm.
available. Consequently, an initial TSS consisting of 90 % co- For all scenarios, the incoming suspended sediment was con-
hesive sediment and 10 % noncohesive sediment was consid- sidered to be 90 % cohesive sediment and 10 % noncohesive
ered based on literature (Droppo et al. 2016). On the other sand as it has been explained earlier.
hand, the initial bed material was considered to be 90 %
noncohesive and 10 % cohesive on the basis of available sed- Chemical constituent transport simulations
iment bed core sample data. Sediment loads at the upstream
boundary and tributary inflows were obtained from sediment Two PAHs (phenanthrene and c1-benzo[a]anthracene/chrys-
versus discharge rating curves. TSS measurements in the ene) and two metals (vanadium and lead) were considered for
lower Athabasca River collected by RAMP (2013) were used constituent chemical transport modelling in the lower
for model validation where simulated TSS was compared to Athabasca River. Constant flow validation runs (including
measured data for the period between 16 and 20 September suspended and bedload sediment transport) for each of these
2004 when data was readily available. chemical constituents were completed for the conditions on 7
Two peak flow scenario simulations were conducted for September 2011 obtained from RAMP (2013) and Water
analysis of sediment deposition patterns. Here, the river bed Survey of Canada (2013). Boundary conditions for these val-
was considered to be nonerodible (rigid); bedload was not pres- idation runs are given in Table 1. Additional validation runs
ent; and suspended sediment transport was simulated so that for the metals were also completed for 5 June 2012. Hydraulic
depositional patterns from the incoming sediments could be and sediment properties (e.g. ks, D50, ws, τcε and τcd) and
easily identified. Case I (historical peak flow) was based on sediment distribution parameters were the same as those used
measured historical flow data for both the upstream inflow for the validation of hydraulics and sediment transport pre-
and tributaries during the period between 9 and 31 July 2011. sented in the BHydraulic and sediment transport simulations^
Case II (extreme peak flow) was also based on the same histor- section. TSS concentrations at the upstream and lateral inflow
ical flow data for the inflow (9–31 July 2011); however, each boundaries were determined from sediment versus discharge
tributary inflow had been assigned a constant value equal to the rating curves, once again with 90 % considered to be cohesive
maximum peak flow observed in the stream during the period and 10 % considered to be noncohesive. Concentrations of
between October 1957 and December 2011. The upstream in- chemical constituents at the inflow boundaries were kept
flow hydrograph dates of 9–31 July 2011 were also chosen as constant and the values were derived from RAMP (2013)
this was the period of peak hydrograph during the historical measured data for the 7 September 2011 runs and from
observation period between October 1957 and December 2011. discharge rating curves for the 5 June 2011 runs. Validation
Sediment loads at the upstream main stem and each tribu- data for chemical constituents were obtained from RAMP
tary boundary were calculated using sediment discharge rating (2013) for the 7 September 2011 runs and from the Joint Oil
curves. The maximum flows and sediment loads for each case Sands Monitoring Program (Government of Canada and
are presented in Table 2. In addition, two sets of sediment Alberta Government 2015) for the 5 June 2012 runs.
parameters were considered; namely ‘high suspension proba- There were several loss terms and diffusion coefficients that
bility (HSP)’ and ‘low suspension probability (LSP)’ param- needed to be specified in the EFDC chemical constituent model.
eters. The HSP set used the following: settling velocity Exchange of chemical constituents between the bed sediment

Table 2 Model inflow boundaries and the corresponding concentrations of cohesive suspended sediment for cases I and II (based on sediment-
discharge rating curves)

Source Max flow from Max. cohesive suspended Peak flow Cohesive suspended Constant Cohesive suspended
hydrograph (m3/s) sediment released (mg/L) (m3/s) sediment released (mg/L) flow (m3/s) sediment released (mg/L)

Case I Case II Case III

Upstream Inflow 4410.00 2081.59 4410.00 2081.59 323.00 15.15
Steepbank River 3.62 14.60 80.00 835.99 1.46 4.82
Ells River 30.04 232.24 237.00 6533.00 9.50 36.14
Firebag River 24.50 8.68 238.00 399.53 14.9 3.76
Muskeg River 1.63 4.95 66.10 65.65 1.58 4.90
MacKay River 33.60 62.25 – – 5.8 10.06
 Environ Sci Pollut Res

and the water column is dependent on the diffusion coefficient at moving, which can lead to further transport of chemical constit-
the sediment/water interface. It can be estimated approximately uents. In addition, macro- and microorganisms living in the top
the same as or less than the diffusion coefficient for the given layers of the sediment may lead to further mixing of the sedi-
chemical in water. For phenanthrene, which is a parent PAH, ment. The processes of burrowing, ingestion, tube building,
this diffusion coefficient in water is reported by GSI defecation, and biodeposition by these organisms are referred
Environmental Inc. (2014) to be 7.47 × 10−10 m2/s. While, to as bioturbation, and will lead to additional transport of chem-
values of the diffusion coefficients for alkylated PAHs are not ical constituents (Ji et al. 2002). In EFDC, the combined trans-
as well known, they should decrease with increasing molecular port of chemical constituents at the sediment bed due to biotur-
weight. Therefore, the diffusion coefficient for c1- bation and sediment mixing processes is accounted for by using
benzo[a]anthracene/chrysene should be less than that for a particle mixing diffusion coefficient. Typical values for the
benzo[a]anthracene and chrysene which are 3.80 × 10−10 and particle mixing diffusion coefficient are small, and the model
3.77 × 10−10 m2/s respectively (Environment Agency 2008). results are not highly sensitive to this coefficient in comparison
The diffusion coefficients at the sediment/water interface for to the other diffusion coefficients. Some reported values for this
both metals were considered to be zero based on diffusion co- parameter have been 1 × 10−9 m2/s (Ji et al. 2002) and
efficients in water reported by GSI Environmental Inc. (2014). 6.94 × 10−11 m2/s (Dynamic Solutions LLC 2013).
Moreover, the chemical constituent water decay rate, sediment Selected model input parameters for the chemical con-
bed decay rate, photolysis rate and volatilization rates for all stituent runs are given in Tables 3 and 4. These are the
PAHs and metals were all assumed to be zero. These assump- initial parameter values that were used both for the vali-
tions are made as the simulation time in this experiment was dation and scenario runs. It should be noted that sensitivity
short enough that no substantial loss is expected because of analysis was completed to determine how sensitive the
decay. The photolysis and volatilization rates were also set to results were to the diffusion coefficient by varying it from
zero, as these processes were not considered in this study. the value used in this paper to 0 m2/s. The difference in the
Transport of chemical constituents within the sediment bed is total concentration within the water column was less than
due to many mechanisms including transfer between the sedi- 2 %. Sensitivity analysis was also conducted for the
ment and the pore water, sediment mixing due to physical pro- validation and the scenario runs to determine how the
cesses and bioturbation (García 2008). The transfer of chemical results would change with variations in the initial bed
constituents between the sediment and pore water depends on concentrations. The fractions of organic carbon in the
the diffusion coefficient in the sediment bed pore water. bed sediment and water column were 0.01 and 0.10
Generally, this coefficient is less than that for water due to the respectively, and were determined from an average of
porosity and tortuosity of the sediment (García 2008). RAMP (2013) measurements in the main stem.
Therefore, for phenanthrene and c1-benzo[a]anthracene/chrys- The equilibrium soil organic carbon water partitioning coef-
ene, these values should be less than or equal to 7.47 × 10−10 and ficient (Koc) is defined by the following equation (Lick 2009):
3.8 × 10−10 m2/s respectively, and they should be zero for the
metals. If there is sediment transport (bedload, resuspension and Kp
K oc ¼ ð5Þ
deposition) at the bed, then the sediment or pore water may be f oc

Table 3 Selected input

parameters for the PAHs Parameter Recommended value
considered in this study
Phenanthrene C1-Benzo[a]anthracene/
chrysene

Partitioning coefficient Koc (water column) (L/mg) 0.2691 1.741

Partitioning coefficient Koc (bed) (L/mg) 0.2691 1.741
Volatilization molecular weight (g/mol) 178 242
Diffusion coefficient for sediment bed pore water (m2/s) 7.47 × 10−10 3.80 × 10−10
Diffusion coefficient at sediment/water interface (m2/s) 7.47 × 10−10 3.80 × 10−10
Particle mixing diffusion coefficient (m2/s) 1 × 10−9 m2/s 1 × 10−9 m2/s
Particle mixing active depth in bed (m) 0.10 0.10
Bed initial concentration (mg/kg) validation run 0.0122 0.4028
Water column initial concentration (μg/L) validation run 0.00115 0.00254
Bed initial concentration (mg/kg) main run 0.0122 0.4028
Water column initial concentration (μg/L) main run 0.003980 0.012860
Environ Sci Pollut Res

Table 4 Selected input

parameters for the Metals Parameter Recommended Value
considered in this study
Lead Vanadium

Partitioning coefficient Koc (water column) (L/mg) 2.0013 2.0297

Partitioning coefficient Koc (bed) (L/mg) 2.0013 2.0297
Volatilization molecular weight (g/mol) 50.94 207.20
Diffusion coefficient for sediment bed pore water (m2/s) 0 0
Diffusion Coefficient For Sediment Bed Pore Water (m2/s) 0 0
Particle mixing diffusion coefficient (m2/s) 1 × 10−9 m2/s 1 × 10−9 m2/s
Particle mixing active depth in bed (m) 0.10 0.10
Bed initial concentration (mg/kg) validation run 5.9786 16.61430
Water column initial concentration (μg/L) validation run 0.4563 2.5025
Bed initial concentration (mg/kg) main run 5.9786 16.61430
Water column initial concentration (μg/L) main run 7.5000 3.2000

where foc is the fraction of organic carbon in the sediment, and average of the values obtained from RAMP ( 2013) be-
Kp is the soil–water partition coefficient for inorganic constit- tween 6 October 1997 and September 2012. The initial
uents, which is the ratio of chemical constituents adsorbed to water column concentrations were set to the first values
sediment to freely dissolved chemical constituents. Kp is de- for the inflow concentration time series data.
fined by the following equation (Ji 2008): For case III (historical low flow), the flow represented a
constant low flow period during open water conditions
ðmass of adsorbed contaminantsÞ=ðmass of sedimentsÞ (i.e. without ice). This case was chosen in order to under-
Kp ¼ ð6Þ
ðmass of dissolved contamnantsÞ=ðvolume of waterÞ stand the effects of high chemical constituent loadings
under a low flow condition. Discharges were set to flows
In this study, Koc values were estimated from a combination of obtained on 20 September 2001 from Water Survey of
RAMP (2013) measured data and literature values (Hawthorne Canada (2013) and RAMP (2013), which was 323 m3/s
et al. 2006). for the main stem upstream inflow boundary. The corre-
The main chemical constituent runs were completed for sponding inflow boundaries for the Ells, Firebag, Mackay,
case II, HSP and LSP and for case III (historical low flow, Muskeg and Steepbank Rivers were 9.5, 14.9, 5.8, 1.58
see description below), HSP. It should be noted that, for and 1.46 m3/s respectively. Sediment loads (TSS) were
these chemical constituent runs, the sediment bed was set also determined based on discharge rating curves, with
to be erodible rather than rigid (and bedload was present). 10 % being noncohesive and 90 % being cohesive.
This was required to get adequate transfer of chemical Chemical constituent concentrations for the tributaries
constituents between the bed sediment and the water are shown in Table 5. For the main stem inflow boundary,
column, which is critical when considering mass peak chemical constituent concentration levels were used,
balance, as the bed sediment is usually the dominant which for phenanthrene, c1-benzo[a]anthracene/chrysene,
source of chemical constituents. The values for chemical vanadium and lead were 0.0142, 0.0405, 25.56 and
constituent concentration levels at the tributaries were 12.23 μg/L respectively. The corresponding constant
kept constant at the maximum observed levels between values for chemical constituent concentration levels at
February 1997 and September 2012 obtained from the tributaries were taken as the maximum observed levels
RAMP (2013) (Table 5). The initial values for concentra- between February 1997 and September 2012 obtained
tions of chemical constituents in the sediment bed were an from RAMP (2013) (see Table 5).

Table 5 Levels of chemical

constituents at the tributary inflow Tributary Phenanthrene (μg/L) C1-Benzo[a]anthracene/ Vanadium (μg/L) Lead (μg/L)
boundaries corresponding to the chrysene (μg/L)
case II and case III runs
Ells 0.00171 0.00597 13.10000 2.70000
Firebag 0.00130 0.00518 1.50000 0.90000
Muskeg 0.00129 0.00142 11.60000 1.20000
Steepbank 0.13000 0.08510 6.80000 10.30000
 Environ Sci Pollut Res

Results and discussions was conducted with inflow boundary conditions correspond-
ing to constant average flows of 15 September 2004. Figure 3
Validation results shows comparison of simulated WSEs and depth averaged
velocities with the observed data collected by CEMA (2012)
The EFDC hydrodynamic model validation, which was run within a reach near Northlands from 9 to 16 September 2004.
over the full lower Athabasca River model domain (~160 km), There are fairly good agreements between simulated and mea-

Fig. 3 a Plan view showing ADCP data points and 2D cross-section sections XS 01 and XS 05 respectively. These results correspond to the
locations for CEMA (2012) reach near Northlands. b, d and c, e hydrodynamic model validation for flow conditions on 15 September
Measured and simulated velocity magnitudes (Vel) and WSEs for 2004 on a fine grid for b, c and a course grid for d, e
Environ Sci Pollut Res

sured WSEs with differences less than 0.5 m for the fine grid,
while it goes up to 2 m for the coarse grid. These differences
are due in part to error associated with interpolating very high-
resolution bathymetry data onto a relatively coarse resolution
grid. Fairly good agreements and similar distribution trends
are also found between simulated and measured depth aver-
aged velocities, especially for the finer grid, with measured
velocities for the courser grid being higher than simulated
once. These differences seem to be due to the differences in
WSEs and possible errors in bed elevations, particularly in
areas where there is no detailed bathymetry data.
Figure 4 shows simulated total and noncohesive suspended
sediment along the channel thalweg for the sediment transport
Fig. 4 Longitudinal variations of simulated TSS and noncohesive validation run. Measured TSSs at specific locations (i.e. left
suspended sediments along the channel thalweg. A comparison bank, center and/or right bank) along the river reach (RAMP
between measured and simulated TSS is shown at specific locations at
a number of cross sections. These results correspond to the sediment
2013) are also plotted on Fig. 4 with the corresponding values
transport validation for conditions on 18 September 2004. This model of simulated TSSs extracted at the same locations. The results
considers both bedload and suspended load transport indicate that high levels of TSS are predicted close to the
inflow boundary, which are likely due to the change in bed

Fig. 5 Longitudinal variations in total, adsorbed, and dissolved left bank and right bank) at the specified distance along the river thalweg.
concentrations for a phenanthrene, b c1-benzo[a]anthracene/chrysene, c Simulated TSSs are also shown. These results correspond to the chemical
vanadium and d lead. Total local measured/simulated chemical constituent transport model validation for conditions on 7 September
constituent values are shown for data taken along the cross sections (i.e. 2011. These models consider both suspended load and bedload transport
 Environ Sci Pollut Res

slope present close to Fort McMurray. In downstream of to the fact that the initial bed sediment concentrations of chem-
Steepbank River, levels of TSS do not show either an increas- ical constituents are considered to be constant throughout the
ing or a decreasing trend. Generally, most of the simulated reach, while in reality, we should expect some variation in
results are within the same order of magnitude as the measured these values along the river. The simulation results also reveal
data and some of the discrepancies can be attributed to local that the magnitudes of chemical constituent concentration in
variations in the actual bed material. the water column are very much dependent on the concentra-
The results for the chemical constituent validation runs are tion of chemical constituents in the sediment bed. Simulation
shown in Figs. 5 and 6. Simulated trends of total, adsorbed results across two sections on Fig. 7 also show adsorbed
and dissolved chemical constituents, as well as TSS are shown chemical constituents to be higher in the main channel, and
on these graphs. For all runs, there is a direct relationship lower in the flood plain. This can be explained by the fact that
between total chemical constituents in the water column and TSS concentrations are higher in the main channel than in the
the corresponding TSS. The sum of adsorbed and dissolved flood plain. The situation for dissolved chemical constituents
chemical constituents is equal to the total chemical constitu- is quite different in that they vary little between the main
ents and the proportion of each of these is determined by the channel and the floodplain (see Fig. 7b, c). This seems to be
partitioning coefficient (Koc). RAMP (2013) and JOSMP because the concentration of dissolved chemicals does not
(2015) measurements of total chemical constituents are also depend on the amount of suspended solids in the water
plotted on the graphs. For the most part, locally measured column.
values are within the same order of magnitude as the locally
simulated values (Figs. 5 and 6). Discrepancies are due mainly Sediment budget and depositional patterns

The first set of sediment transport simulations was run by

assuming a rigid bed river channel and using conservative
sediment parameters (LSP) to gain a better understanding of
cases with high sediment deposition. The spatial distribution
of deposited cohesive sediment from all sources (upstream
boundary and tributary inflows) at the end of the simulation
period is shown in Fig. 8a for case II (LSP) that corresponds to
extreme peak–flow scenario. The major areas of cohesive sed-
iment deposition are circled in red, indicating that most of the
deposition occurs within the floodplains and islands (Fig. 8b,
d). The general tendency for sediments to deposit in river
floodplains, where water velocities are approaching zero, is
well known and has been noted by Lair et al. (2009). As
shown in Fig. 8c, e, total cohesive suspended sediment is also
predicted to be higher in the main channel and lower in the
surrounding floodplains and islands. Such results are to be
expected as flow velocities in the main channel are mostly
higher than in the floodplains, resulting in a greater tendency
for cohesive sediment to remain in suspension. Some analyses
were also completed for cases I and II for flood inundation at
the major tributaries, but no discernible differences could be
seen in water depths during peak flow between the two cases.
According to the mass balance calculations for the conser-
vative sediment runs (sases I (LSP) and II (LSP)), a large
percentage (>50 %) of sediment released from each individual
source (tributaries and main stem) is deposited within the river
Fig. 6 Longitudinal variations in total, adsorbed and dissolved
model domain (see Fig. 9a, b). The cohesive sediment prop-
concentrations for a vanadium and b lead. Total local measured/ erties used in these two LSP scenarios lead to a greater ten-
simulated chemical constituent values are shown for data taken along a dency for sediment with a settling velocity of 1 mm/s to settle.
cross section at the specified distance along the river thalweg. Simulated This finding indicates that in some instances, there is a poten-
TSSs are also shown. Measured data are obtained from the JOSMP
(2015). These results correspond to chemical constituent transport model
tial for a large proportion of bed sediment and associated
validation for conditions on 5 June 2012. These models consider chemical constituents originating from the different sources
both suspended load and bedload transport to be deposited within the main stem of the LAR. Moreover,
Environ Sci Pollut Res

Fig. 7 a Plan view showing

locations of cross sections S1 and
S2 (flood plain shown in red,
while main channel is shown in
black). b, c The lateral variations
in total, dissolved and adsorbed
c1-benzo[a]anthracene/chrysene
for cross sections S2 and S1
respectively. These results
correspond to chemical
constituent transport model
validation for conditions on 7
September 2011. This validation
model considers both bedload and
suspended load transport

looking at the contribution of each source towards the total compared to the tributaries. For case II (LSP) that corresponds
deposited cohesive sediment, the majority of total deposited to extreme peak–flow scenario, however, the tributaries (par-
cohesive sediment originates from the main stem upstream ticularly Ells River) also contribute substantially towards de-
boundary (99.9 % for case I (LSP), and 61.4 % for case II posited sediment (up to 34.7 %; see Fig. 10b). While the
(LSP)). This is not surprising as much more sediment enters above results are consistent with the modelling assumptions,
the model domain from the main stem upstream boundary it may also be possible that more sediment could pass through

Fig. 8 a Plan view of model domain showing areas A1 and A2 and bed mass in A1 and e cohesive suspended sediment levels in A1. Results
simulated deposited cohesive bed mass. Major areas of deposition are shown correspond to the end of the high flow simulation run (31
circled in red. These areas include b deposited cohesive bed mass in July 2011) for (case II (LSP)). These models do not consider bedload
A2, c cohesive suspended sediment levels in A2, d deposited cohesive transport but do consider suspended load transport
 Environ Sci Pollut Res

Fig. 10 Percentage of cohesive sediment contribution from each source

relative to the total amount of sediment deposited from all sources for a
case I (LSP), b case II (LSP) and c case II (HSP). These results correspond
to model simulations that do not consider bedload transport but do con-
sider suspended load transport, sediment deposition and resuspension

Canada (2013) measurements. Nevertheless, these results and

comparisons are important as they show a range of possible
results based on a probable range of sediment parameters
(LSP and HSP).

Distributions of chemical constituents and mass balance

The model validation results in the previous section have

shown the very high levels of adsorbed chemical constituents
in the main channel. Also, the levels of total chemical constit-
uents within the water column tended to be higher in the main
channel than in the floodplain. Similar results are shown in
Fig. 11 for the distributions of total and dissolved c1-
benzo[a]anthracene/chrysene in the water column correspond-
ing to case II (HSP), which is similar to the distributions found
for other chemical constituents. Consistent with the validation
results, the total values for the concentration of this PAH
Fig. 9 Percentage (and amounts) of deposited sediment relative to the
amount released from each source for a case I (LSP), b case II (LSP) and c (Fig. 11a, b) are substantially greater in the main channel
case II (HSP). These results correspond to models that do not consider compared to those in the floodplain, while there is little vari-
bedload transport but do consider suspended load transport, sediment ation in the dissolved values between the main channel and the
deposition and resuspension
floodplain (Fig. 11c). This confirms the fact that the concen-
trations of chemical constituent within a water column depend
the system without substantial deposition within the model on the amount of TSS.
domain. This has been illustrated with case II (HSP), where Figure 12 presents the results of an analysis on the sensi-
approximately 75 % of total cohesive sediment entering the tivity of chemical constituent mass balance (inflow/outflow)
system through the upstream boundary and tributaries leaves within the model reach to the corresponding chemical concen-
the system at the outflow boundary. The amount of sediment tration in the bed sediment and the sediment erosional param-
deposited from each individual source ranges from ~14 to eters. The results indicate that the mass of each chemical con-
31 % (Fig. 9c) with the sediment inflow at the main stem stituent leaving the system is highly dependent on its concen-
upstream boundary still constituting the largest proportion to- tration within the river bed. The maximum bed concentration
wards total cohesive sediment deposited (~70 %) in the sys- for each chemical constituent shown in Fig. 12 is the average
tem, although the Ells River also contributes substantially of measured bed concentrations within the model reach. As
(27 %) (see Fig. 10c). Care must be taken in the interpretation the bed concentrations decrease from these measured values,
of these results, however, as the input sediment loads from the mass outflow to mass inflow ratios also decrease linearly
each source are based on sediment discharge rating curves, until the mass outflow approaches zero when the bed concen-
which are created from RAMP ( 2013) and Water Survey of tration is set to zero. These results indicate that the bed
Environ Sci Pollut Res

Fig. 11 Plan views of c1-

benzo[a]anthracene/chrysene in
the water column at the end of run
case II (HSP). These models
consider both suspended and
bedload transport. a Total values
throughout the model domain
(scale x/y = 3.5:1). b Total values
within the red box indicated in a.
c Dissolved values within the red
box indicated in a. The scale in b
and c is x/y = 1:1

sediment is the main source of chemical constituents being the water column is mainly due to the processes of diffusion,
transferred to the water column, and a large portion of what sediment suspension and water-specific discharge (see Eq. 4).
is leaving the model reach originates from the river bed. Such The results in Fig. 12a, b for phenanthrene and c1-
a transfer of chemical constituents from the bed sediment to benzo[a]anthracene/chrysene also show that the ratio of

Fig. 12 Sensitivity analysis for

suspended sediment mass
balances at the main stem inflows
and outflows for a total
phenanthrene and b total c1-
benzo[a]anthracenes/chrysenes.
All trials are run for case II (HSP),
case II (LSP) and case III (HSP).
Resuspension occurs for all of
these models, but bedload
transport is considered for cases I
and II but not case III. The results
show chemical mass balance
integrated over a 23-day
simulation time period
 Environ Sci Pollut Res

chemical constituent mass outflow to inflow is lower for case benzo[a]anthracene/chrysene, which has a bed concentration of
II (LSP) as compared to case II (HSP). This result is consistent 0.40280 mg/kg (33 times that of phenanthrene), the mass outflow
with the expected low levels of suspended sediment concen- is 517,855.14 g (27 times that of phenanthrene). The same is true
tration in the case of LSP compared to that of HSP. Moreover, for metals if we see in Table 6 that for lead, which has a bed
at the same bed sediment and upstream and tributary inflow concentration of 5.97 mg/kg, the mass outflow is 9,133,334 g,
concentrations, the ratio of mass outflow to mass inflow for while for vanadium, which has a bed concentration of 16.6 mg/
case III (HSP) that corresponds to low flow condition, is less kg (2.8 times that of lead), the mass outflow is 23,186,306 g (2.5
than that of case II (HSP) that corresponds to maximum flow times that of lead). It is also important to note that the mass outflow
condition. This is because the amount of cohesive sediment to inflow ratio may also depend on the magnitude of the inflow
leaving the system for case III (HSP) is less than 1 % of that concentration. For example the ratios of mass outflow to inflow for
for the case II (HSP), implying that less adsorbed and dis- metals not only increase with increasing bed concentration but also
solved chemicals are leaving the system during low flow con- depend on the magnitude of the inflow concentration. That is the
ditions. These findings are important as they demonstrate that reasonwhytheoutflowtoinflowratiosforthePAHsandthemetals
peak flow rates are critical in transporting chemical constitu- inTable6arequitedifferentasthetotalmassinflowismuchgreater
ents from the river system down towards the delta and Lake for the metals compared to the PAHs.
Athabasca.
Another important finding here is that the magnitude of chem-
ical constituent mass outflow is highly dependent on the concen- Summary and conclusions
tration of chemical constituents in the bed. Examining case II
(HSP)resultsinTable6,forphenanthrenewithabedconcentration Two-dimensional hydrodynamic, sediment and chemical
of 0.01220 mg/kg, the mass outflow is 18,989.47 g. For c1- constituent transport models for the LAR have been

Table 6 Magnitudes of chemical

constituent mass inflows and Case Bed concentration Diffusion coefficient at Mass Mass
outflows for the different (mg/kg) sediment/water inflow (g) outflow (g)
simulation runs interface (m2/s)

Phenanthrene
Case II (LSP) 0.01220 7.47 × 10−10 43,571.49 6006.09
0.00488 7.47 × 10−10 43,571.49 2831.64
0 7.47 × 10−10 43,571.49 1515.69
Case II (HSP) 0.01220 7.47 × 10−10 43,571.49 18,989.47
0.00488 7.47 × 10−10 43,571.49 8632.99
0 7.47 × 10−10 43,571.49 1718.25
Case III (HSP) 0.01220 7.47 × 10−10 12,899.34 1518.23
0.00488 7.47 × 10−10 12,899.34 866.81
0 7.47 × 10−10 12,899.34 431.85
C1-Benzo[a]anthracene/chrysene
Case II (LSP) 0.40280 3.80 × 10−10 141,914.71 88,690.65
0.16112 3.80 × 10−10 141,914.71 30,615.76
0 3.80 × 10−10 141,914.71 2701.19
Case II (HSP) 0.40280 3.80 × 10−10 141,914.71 517,855.14
0.16112 3.80 × 10−10 141,914.71 211,930.71
0 3.80 × 10−10 141,914.71 3884.13
Case III (HSP) 0.40280 3.80 × 10−10 30,185.31 27,866.99
0.16112 3.80 × 10−10 30,185.31 11,272.18
−10
0 3.80 × 10 30,185.31 238.76
Vanadium
Case II (HSP) 16.6143 0 80,285,046.28 23,186,306.36
Case III (HSP) 16.6143 0 16,757,628.33 766,750.99
Lead
Case II (HSP) 5.9786 0 36,378,430.79 9,133,334.65
Case III (HSP) 5.9786 0 7,964,959.88 328,390.11
Environ Sci Pollut Res

setup using bathymetric information and other observed coverage of the high-resolution Geoswath bathymetry in the
data from various sources. The models were successful- models will help improve simulation results. As the study has
ly validated with observed hydrodynamic, sediment and shown that concentrations of constituent chemicals in the bed
chemical constituent data. A number of model simula- sediment have a major impact on the corresponding concentra-
tion scenarios have been designed and implemented to tions in the water column, more spatially resolved field sam-
investigate hydrodynamics, sediment and chemical con- pling of chemical constituent concentrations in bed sediments
stituent transport characteristics, inflow/outflow mass would greatly improve the accuracy of model predictions. The
balances and depositional patterns within the LAR. model may also be extended to include additional chemical
The following are some of the major conclusions that constituents to get a more complete picture of the transport
are drawn from the results of this study: phenomenon in the lower Athabasca River system.

1. For all three high flow sediment transport scenarios (case Acknowledgments The financial support for this study was provided
I (LSP) and case II (LSP and HSP)), a large proportion of by the Government of Alberta and Environment Canada Joint Oil-Sands
Monitoring Program (JOSMP). The authors would like to thank Drs. Fred
the cohesive sediment deposited within the LAR origi-
Wrona and Anil Gupta of Alberta Environmental Monitoring and Science
nates from the main stem upstream inflow boundary that Division as well as Drs. Patricia Chambers and Malcolm Conly of
is close to Fort McMurray. In another words, tributary Environment and Climate Change Canada for their contributions and
contributions towards cohesive sediment deposition in constructive suggestions at the various stages of this project. The authors
acknowledge the help of Ali Oveisy in developing the model grids and
the LAR are relatively small except for the case of Ells
Paul Craig from Dynamic Solutions International for his support of the
River (cases II (LSP and HSP)), which may contribute EFDC software. The authors would like to thank Dr. Fay Hicks for pro-
substantially (up to 27–35 %) to the sediment deposition viding some of the lower Athabasca River cross-section data and Dr.
depending on specific flow conditions and the sediment Roderick Hazewinkel from Alberta Environment and Parks for facilitat-
ing access to the LIDAR data used for preparing the bathymetry for the
parameters used.
Athabasca River and its flood plains.
2. For all sediment simulation scenarios, the patterns of co-
hesive sediment deposition were similar in that they were
mostly deposited within the flood plains and around
References
islands where flow velocities were very low. The results
also indicated that the major depositional areas within the
Alberta Environment and Sustainable Resource Development (AESRD)
LAR were upstream and downstream of Steepbank River, (2014) Personal communication
close to Firebag River, and downstream of Firebag River Andrishak R, Abarca JN, Wojtowicz A and Hicks F (2008) Freeze-up
near the outflow boundary. study on the lower Athabasca River (Alberta, Canada), Proc. 19th
3. The chemical constituent transport models were able to IAHR International Symposium on Ice, Vancouver, BC, v.1., 77–88
Blumberg AF, Galperin B, O’Connor DJ (1992) Modeling vertical struc-
predict concentrations of water column chemical constit-
ture of open-channe flow. J Hyd Engr 118:1119–1134
uent values within the same order of magnitude compared Canada-Alberta JOINT Oil Sands Environmental Monitoring
to the measured data by using average observed bed sed- I n f o r m a t i o n P o r t a l ( J O S M P ) ( 2 0 1 5 ) h t t p : / / w w w.
iment concentration values. Sensitivity analysis has also jointoilsandsmonitoring.ca. Accessed 10 Oct 2015
revealed that the levels of simulated water column chem- Canadian Council of Ministers of the Environment (CEMA) (2012) Lower
Athabasca data. "http://ftp.cciw.ca/incoming/LowerAthabasca_data_
ical constituents in the river system depend very much on
March05.12/, Accessed on Nov. 10, 2012
the corresponding bed sediment concentrations. Canadian Council of Resource and Environment Ministers (CCREM)
Moreover, adsorbed chemical constituent levels tended (1999) Canadian water quality guidelines for the protection of aquatic
to be higher in the main channel compared to that in the life, polycyclic aromatic hydrocarbons (PAHs), "http://www.google.
floodplain. ca/url?sa=t&rct=j&q=&esrc=s&source=web&cd=5&cad=rja&uact=8
&ved=0CCsQFjAE&url=http%3A%2F%2Fceqg-%2Fdownload%2
4. The results of the various simulation scenarios also indi- Fen%2F201&ei=sUiTVIWKEZTloAShwYKgCw&usg=
cated that high river flow conditions correspond to higher AFQjCNFDsHWOah3TsUxPSqZ_ZnR8v6-x8w&sig2=y3IV2
sediment loadings and thus higher concentration of usMLa_3O1fGeTkySg&bvm=bv.82001339, d.cGU, Accessed on
chemicals constituents as compared to low flows. Dec. 18, 2014
Carson, M.A. (1990) Sediment budget for the Athabasca River Basin,
Consequently, high flows appear to be the main drivers
Alberta, between Fort McMurray and Embarras, 1876–1986. Report
for transport of chemical constituents to the downstream for the Inland Waters Directorate, Environment Canada, Calgary,
delta and lake. Alberta
Conly FM, Crosley RW, Headley JV (2002) Characterizing sediment
In general, the two-dimensional numerical model developed sources and natural hydrocarbon inputs in the lower Athabasca
River, Canada. J Environ Eng Sci 1(3):187–199
in this study was successfully applied to perform quantitative Craig PM (2013) User’s manual for EFDC_Explorer7.1: a pre/post pro-
analysis of sediment and constituent chemical transport within cessor for the Environmental Fluid Dynamics Code (Rev 00),
the lower Athabasca River. Further improvement and increased Dynamic Solutions International, Edmonds, Washington
 Environ Sci Pollut Res

Dillon P, Dixon G, Driscoll C, Giesy J, Hurlbert S, Nriagu J (2011) Irwin R (1997) Environmental contaminants encyclopedia, Phenanthrene
Evaluation of four reports on contamination of the Athabasca river entry, "http://www.nature.nps.gov/hazardssafety/toxic/list.cfm,
system by oil sands operations, Water Monitoring Data Review Accessed on May 8, 2015
Committee, Alberta, 46p Ji Z-G (2008) Hydrodynamics and water quality: modeling rivers, lakes,
Doyle PF (1977) Hydrologic and hydraulic characteristics of the and estuaries. Wiley, New Jersey, USA, pp 704
Athabasca River from Fort McMurray to Embarrass. Open File Ji Z-G, Hamrick JH, Pagenkopf J (2002) Sediment and metals modeling
Report 1977-13, Alberta Research Council, Edmonton, Alberta in Shallow River. J Environ Eng 128(2):105–119
Droppo IG, D’Andrea L, Krishnappan BG, Jaskot C, Trapp B, Basuvaraj Kannel PR, Gan TY (2013) Application of WASP for modelling and
M, Liss SN (2015) Fine-sediment dynamics: towards an improved management of naphthenic acids along Athabasca River, Alberta,
understanding of sediment erosion and transport, J. Soils and Canada. Water Air Soil Pollution 224:1764 16p
Sediments 15(2):467–479 Khanna VK (2003) Application of the edg1-d model in the lower
Droppo IG, Krishnappan BG, Lawrence JR (2016) Microbial interactions Athabasca river basin to estimate high flows during open-water
with naturally occurring hydrophobic sediments: influence on sedi- season, contract# 2002–0028. Cumulative Environmental
ment and associated contaminant mobility. Water Res 92:121–130 Management Association, Fort McMurray, Alberta
Dynamic Solutions International LLC (2016) EFDC explorer, River Kirk, J., and Manzano, C. (2013). Personal communication.
N e t w o r k s ( 1 D & 2 D ) , " h t t p : / / w w w. e f d c - e x p l o r e r. Lair GJ, Zehetner F, Fiebig M, Gerzabek MH, van Gestel CAM, Hein T,
com/applications/river-networks-(1d-and-2d).html, Accessed on Hohensinner S, Hsu P, Jones KC, Jordan G, Koelmans AA, Poot A,
July 3, 2016. Slijkerman DME, Totshe KU, Bondar-Kunze E, Barth JAC (2009)
Dynamic Solutions LLC (2013) Draft: lake thunderbird report for nutri- How do long-term development and periodical changes of river-
ent, turbidity, and dissolved oxygen TMDLs, Prepared for floodplain systems affect the fate of contaminants? Results from
Oklahoma Department of Environmental Quality Water Quality European rivers. Environmental Pollution 157:3336–3346
Division, "http://www.deq.state.ok.us/wqdnew/tmdl/Lake%20 Lick W (2009) Sediment and contaminant transport in surface waters.
Thunderbird%20TMDL%20ReportJune6Draft.pdf, Accessed on CRC Press, Taylor & Francis Group, Boca Raton, FL
Dec. 18, 2014 McCauley E (1997) A review and evaluation of water quality and quan-
Environment Agency (2008) Using science to create a better place, com- tity models used by the Northern River Basins Study, Northern
pilation of data for priority organic pollutants for the derivation of River Basins Study project report no. 82., Northern River Basins
soil guideline values, science report: SC050021/SR7. Almondsbury, Study, Edmonton, Alberta
United Kingdom McDonald D, and Muricken D (2009) North Saskatchewan River water
Environment Canada (2011) Lower Athabasca water quality monitoring quality model: Alberta environment technical report, Version 1.1,
program, phase 1, Athabasca River mainstem and major tributaries, Alberta Environment
Minister of the Environment, Gatineau, QC Mellor GL, Yamada T (1982) Development of turbulence closure model for
Galperin B, Kantha LH, Hassid S, Rosati A (1988) A quasi-equilibrium geophysical fluid problems. Rev Geophys Space Phys 20:851–875
Pietroniro A, Hicks F, Andrishak R, Watson D, Boudreau P and Kouwen N
turbulent energy model for geophysical flows. J Atmos Sci 45:55–62
(2011) Hydraulic routing of flows for the Mackenzie River, Environment
García, M. H. (Ed.) (2008) Sedimentation engineering: processes, mea-
Canada, University of Alberta and National Research Council, 78p
surements, modeling and practice. ASCE manuals and reports on
Regional Aquatics Monitoring Program (RAMP) (2013) Monitoring data-
engineering practice, 110, ASCE: Reston. ISBN 978-0-7844-0814-
b a s e : s e d i m e n t q u a l i t y " h t t p : / / w w w. r a m p - a l b e r t a .
8. xx, 1132 pp.
org/data/Sediment/sediment.aspx, Accessed on Nov. 2013
García M, Parker G (1991) Entrainment of bed sediment into suspension.
Shakibaeinia A, Dibike YB, Prowse TD (2014) Numerical modelling of
J. Hyd. Engrg. 117:414–435
dissolved-oxygen in a cold-region river, Proceedings of the 7th Intl.
Garcia-Aragon J, Droppo IG, Krishnappan B, Trapp B, Jaskot C (2011) Congress on Env. Modelling and Software, ed. Ames, D.P., Quinn,
Experimental assessment of Athabasca river cohesive sediment de-
N.W.T., and Rizzoli, A.E., San Diego, CA
position dynamics. Water Qual Res J Can 46(1):87–96 Shakibaeinia A, Kashyap S, Dibike Y, Prowse T (2016) An integrated
Geobase (2012) Canadian Digital Elevation Data, "http://www.geobase. numerical framework for water quality modelling in cold-region
ca/geobase/en/index.html, Accessed on Oct. 10, 2012 rivers: a case of the lower Athabasca River. Sci Total Environ
Government of Canada and Alberta Government (2015) Canada-Alberta 569–570:634–646. doi:"10.1016/j.scitotenv.2016.06.151
oil sands environmental monitoring information portal, Shaw J and Kellerhals R (1982) The composition of recent alluvial
"http://www.jointoilsandsmonitoring.ca/default.asp?lang=En&n=5 gravels in Alberta river beds, Bulletin 41, Alberta Research Council
B6AC9E5-1, Accessed on May 1, 2014 Smith JD, McLean SR (1977) Spatially averaged flow over a wavy bed. J
GSI Envrionmental Inc (2014) Chemical database <"http://www.gsi-net. Geophys Res 82:1735–1746
com/en/publications/gsi-chemical-database.html> (Dec. 16, 2014) Van der Vinne G (2003) Flow simulations and fish habitat evaluation for
Hamrick JM (1992) A three-dimensional environmental fluid dynamics the Athabasca River at Bitumount (reach #4), Cumulative
computer code: theoretical and computational aspects. In: Special Environmental Management Association (CEMA). Fort
report 317, The College of William and Mary. Virginia Institute of McMurray, Alberta 41p
Marine Science, Gloucester Point, VA Van Rijn LC (1984) Sediment transport, part II: suspended load transport.
Hamrick, J. M. (2007a) The environmental fluid dynamics code theory J Hyd Engrg 110:1613–1641
and computation. Volume 1: hydrodynamics and mass transport, Water Survey of Canada (2013) "http://www.ec.gc.ca/rhc-wsc/ .
Tetra Tech, Inc., Fairfax, VA Accessed 10 Jul 2013
Hamrick JM (2007b) The environmental fluid dynamics code theory and Wojtowicz AM (2010) 2-D modeling of freeze-up processes on the
computation. Volume 2: sediment and contaminant transport and Athabasca River downstream of Fort McMurray, Alberta, M.Sc.
fate. Tetra Tech, Inc., Fairfax, VA Thesis, University of Alberta, 157pp
Hawthorne SB, Grabanski CB, Miller DJ (2006) Measured partitioning Yergeau E, Lawrence JR, Sanschagrin S, Roy JL, Swerhone
coefficients for parent and alkyl polycyclic aromatic hydrocarbons GDW, Korber DR, Greer CW (2013) Aerobic biofilms grown
in 114 historically contaminated sediments, part 1 KOC values. from Athabasca watershed sediments are inhibited by increas-
Environ Toxicol Chem 25(11):2901–2911 ing bituminous compounds concentrations. Appl Environ
Hicks F (2011) Personal communication Microb 79(23):7398e7412