CDA 2019 Annual Conference

Congrès annuel 2019 de l’ACB

CANADIAN DAM ASSOCIATION Calgary, AB, Canada
ASSOCIATION CANADIENNE DES BARRAGES October 6-10, 2019

A CASE STUDY: THE IMPORTANCE OF UNDERSTANDING SPATIAL VARIATION

OF GEOLOGICAL PROFILES IN OIL SANDS TAILINGS DAM DESIGN

Tiequn Feng, P.Eng., Imperial Oil Resources Limited, Calgary, AB, Canada
Paul Cavanagh, P.Eng., Imperial Oil Resources Limited, Calgary, AB, Canada
Jason Polak, P.Eng., Imperial Oil Resources Limited, Calgary, AB, Canada

ABSTRACT

Design of the External Tailings Area (ETA) for Imperial’s Kearl Mine was based, in part, on seepage analyses that
used generalized geologic information combined with discrete boreholes, soil samples and piezometric information.
The scale and locations of the available information could not practically describe or account for the geologic
variability at the site, which led to discrepancies between the measured seepage rates and the rates predicted in design.
This resulted in the installation of “fit for purpose” seepage control measures at locations along the toe of the dam
soon after the start of reservoir filling. A series of field investigations and seepage back-analyses were undertaken to
better understand the seepage mechanism. The effort focused mainly on how to match the measured and predicted
seepage rates and incorporation of important features not initially taken into account.

This case study presents a practical approach to evaluate the seepage mechanism, which includes a series of transient
seepage back-analyses, establishing relationships for piezometric responses in time and space, and a study on
geological borehole locations. The proposed approach substantiated through this case study can be applied to other
similar projects.

RÉSUMÉ

La conception du bassin de rétention de résidus miniers pour la mine Kearl de l’Impériale s’est basée, en partie, sur
des analyses d’infiltration des eaux utilisant des données géologiques généralisées en combinaison avec l’information
provenant de trous de forage, des échantillons de sol et des données piézométriques. L’échelle et la position des
données disponibles ne permettaient pas de représenter adéquatement la variabilité géologique présente sur le site,
résultant ainsi en une grande variation entre les taux d’infiltration prédits lors de la conception et ceux mesurés sur le
terrain. Les incertitudes observées menèrent à l’instauration de mesures spécifiques aux besoins de la mine Kearl
permettant de mesurer le débit d’infiltration des eaux tout au long du pied du barrage, immédiatement après la mise
en service du bassin de rétention. Par la suite, une série d’investigations effectuées sur le terrain et d’analyses
d’infiltration rétrospectives ont été entreprises afin d’approfondir la compréhension des mécanismes d’infiltration.
Les efforts ont principalement visé à obtenir une meilleure corrélation entre les taux d’infiltration mesurés et prédits
ainsi que l’incorporation de paramètres initialement omis.

Cette étude présente une approche pratique servant à évaluer le mécanisme d’infiltration des eaux incluant une série
d’analyses rétrospectives de l’infiltration dynamique, l’établissement de corrélations temporelles et spatiales à des
réponses piézométriques et une étude sur l’emplacement de trous de forage géologique. L’approche développée,
décrite dans cette étude de cas, peut être appliquée à d’autres projets similaires.

CDA 2019 Annual Conference, Calgary, AB, Canada

1 INTRODUCTION

Imperial Oil Resources Limited has developed an oil sands mine at its Kearl lease, located approximately
70 km north of Fort McMurray, Alberta (Figure 1). The mine requires an External Tailings Area (ETA) to
accommodate storage of the tailings generated by the mine. The ETA is divided into West ETA (WETA)
and East ETA (EETA).

McClelland Lake
Kearl Oil 
Sands Project 

Kearl Site 

Fort McMurray
SCALE = 1:500,000 

Province of Alberta 

Figure 1: Kearl site location plan

Design of the ETA was based, in part, on seepage analyses that used generalized geologic information
combined with discrete boreholes, soil samples and piezometric information. The design sections in 2D
seepage analyses are commonly selected from the same sections as the slope stability analyses, which
typically focus on weak units with relatively low permeability. However, seepage issues usually occur in
soil units with higher permeability.

The scale and locations of the available information from drilling programs could not practically describe
or account for the geologic variability at the site. Seepage flows are sometimes impacted by variability
perpendicular to the section, which introduces flow externally into the 2D seepage system. Without
accounting for this external flow in a 2D seepage analysis, the modelled permeability of the dyke core had
to be adjusted to an unreasonably high value in order to match the observed flow rates. Based on these
model results, “fit for purpose” seepage control measures at specific locations along the toe of the dam were
installed soon after the start of reservoir filling to mitigate the perceived risks of a compromised dyke core.

This case study presents a practical approach to evaluate the seepage mechanism, which includes a series
of transient seepage back-analyses, establishing relationships for piezometric responses in time and space,
and a study on geological borehole locations. The proposed approach substantiated through this case study
can be applied to other similar projects.

CDA 2019 Annual Conference, Calgary, AB, Canada 2

2 WEST ETA DESIGN OVERVIEW

The West ETA comprises an overburden Starter Dyke to El. 358 m and is raised via conventional upstream
construction to ultimate design El. 420 m. The typical Starter Dyke and Ultimate Dyke design sections are
shown in Figure 2.

Pressure Relief System

Figure 2: Typical starter dyke and ultimate dyke design sections

2.1 Site Geology

Geomorphology of the Kearl lease is a result of a complex glacial history consisting of multiple glacial ice
advances and retreats, glacial thrusting and the associated sedimentation, melt-water processes, and
remobilization of sediments. Quaternary deposits unconformably overlie Cretaceous deposits of Middle
and Lower Members of the McMurray Formation. Upper McMurray and the Clearwater Formation have
been eroded and are absent from the ETA footprint. The Quaternary geology within the Kearl lease is a
complex distribution of glacial deposits including outwash sands and gravels, lacustrine clays and silts, and
sandy and clayey diamictons, commonly referred to as tills.

CDA 2019 Annual Conference, Calgary, AB, Canada 3

2.2 Starter Dyke

2.2.1 Foundation Conditions

Surficial geology of the area comprises patchy Holocene peat and a mix of Pleistocene moraine and glacio-
fluvial deposits. The West ETA is divided into sectors with representative sections (e.g., J-M), differentiated
on the basis of the foundation conditions, as shown in Figure 3. The red rectangular area is the seepage
area to be discussed in the following sections.

Figure 3: Plan view of design sections

2.2.2 Seepage Controls

The main considerations for External Tailings Area (ETA) seepage include:
 Internal drainage and piping control for the overburden Starter Dyke, which was required to
impound up to 20 m of water during start-up conditions
 Internal drainage control for the ETA during operations as the dyke is raised using hydraulic cell
and beaching constructions methods
 Interception of foundation seepage to mitigate off-lease migration of contaminants
 Uplift pressure mitigated by a pressure relief system

The internal drainage and piping control system for the overburden starter dyke includes a chimney drain,
drainage blanket, passive pressure relief system, low-permeability core and horizontal downstream barrier
as shown in Figure 2.

CDA 2019 Annual Conference, Calgary, AB, Canada 4

3 SEEPAGE BACKGROUND

3.1 Seepage Flow Rates

The majority of the drain-outtake pipes within the seepage area (as shown in Figure 3) began flowing upon
commencement of reservoir filling, beginning in July 2011, and the observed flow volumes were higher
than predicted in the design. A comprehensive investigation program was conducted shortly thereafter,
which included water chemistry testing, 2-D seepage analyses, and a CPT investigation.

3.2 Water Chemistry

Water samples were collected in August 2012 from the ETA pond and foundation drainage pipes along
with the available groundwater data. Based on the major ion patterns (shown in Figure 4) the water collected
from ditches and pipes compared well with the shallow groundwater values.

EXPLANATION
ETA/Athabasca river
Ditch and Pipe
Groundwater obs well

Figure 4: Piper graphs for the water chemistry

3.3 Seepage Analysis

In order to assess the higher than anticipated flow rates observed in the area as shown in Figure 3, a
geotechnical assessment including interpretation of possible seepage sources and the seepage back-analysis
was conducted.

Seepage analysis models were developed from available piezometric information and as-built records of
the starter dyke. The foundation conditions in Section L, as shown in Figure 3 and Figure 5, comprised
thick Pl clays, which were assessed for geotechnical stability. Section L was also selected as a typical
section for seepage back-analysis. The modelled hydraulic conductivities of the foundation and the dyke
were presented in Table 1.

CDA 2019 Annual Conference, Calgary, AB, Canada 5

 Hl (Wick Drains Installed) 

Filter Drain  Collection Ditch 
Dyke Shell 
Dyke Core  Blanket Drain  Hl  
Pl (Wick Drains Installed)  Pl (Wick Drains Installed)  Pl  
Relief Trench Well  Pfs 

Figure 5: 2D Steady-state Seepage model of section L

Table 1: Hydraulic conductivities
Material kh (m/s) kh/kv
Hl Clay 1×10-8 5
Hl Clay (Wick Drains Installed) 1×10 -4 1*
Pl Clay 1×10-8 5
Pl Clay (Wick Drains Installed) 1×10 -4 1
Pfs Sand 1×10 -5 1
PGL Till 1×10-7 1
Till Core (Zone 1) 1×10-8 1
Dyke Shell (Zone 2) 5×10 -6 1
Chimney Filter (Zone 3) 1×10-4 1
Relief Trench/Well System 1×10-3 1
Upstream Blanket Drain for Wick Drain 1×10 -4 1
*for Simplicity, the wick drains are assumed to be isotropic for this study and
the anisotropic behaviour of the conductivity will be investigated in the future

Analyses to evaluate flow and piezometric levels under steady-state conditions and to reproduce observed
conditions were undertaken. The results generally matched the measured flows and piezometric response
below and immediately behind the core when using a modeled hydraulic conductivity in the core of
1×10-4 m/s (compared to the design value of 1×10-8 m/s).

3.4 CPT Investigation

Based on the seepage analysis as described in section 4.3, a Cone Penetration Test (CPT) program was
completed in late 2012 to assess the geotechnical condition of the core, which included:
 Continuous soil type profiles with inferred geotechnical conditions and properties such as
consistency, hydraulic conductivity (permeability) and location of phreatic surface;
 Conditions within the core, specifically, hydraulic conductivity;
 Confirmation of the core base hydraulic conductivity; and
 Confirmation of the foundation soils hydraulic conductivity.

The key findings from the CPT program were summarized in the following:
 Based on CPT pore pressure dissipation, estimated core conductivity ranges from 1×10-4 m/s to
1×10-9 m/s, the operational overall core conductivity is estimated from 1×10-6 m/s to 1×10-7 m/s
(compared to 1×10-8 m/s in the initial design);
 Sandy zones exist above the core base but they are likely isolated sandy lenses within the core;
 The CPTs did not measure unexpectedly high pore-water pressures close to the base of the core;

CDA 2019 Annual Conference, Calgary, AB, Canada 6

  The CPT testing did not confirm the presence of high conductivity materials in the order of
1×10-4 m/s close to the core base that would be required in order to convey flows similar to those
observed in the drainage discharge pipes.

3.5 Remediation Works

The CPT results concluded that seepage through the core and at the base of the core is not considered to be
a major contributor to the observed seepage and core remediation work was not required. However, a
remedial gravel toe blanket was still recommended as a precautionary measure to mitigate uncertainty
related to the source and the work was completed in 2013. A typical remediation section is shown in
Figure 6.

ZONE 3 MATERIAL: FILTER DRAIN SAND (LESS THAN 5% PASSING 75 MICRON) 

Figure 6: Typical section of installed drainage blanket (ZONE 3)

4 PROPOSED APPROACH

The process described in Section 3 generally followed the traditional observational method (Peck 1969)
and the remediation works were implemented to increase the factor of safety with regards to an internal
erosion failure mode. However, the following considerations were not included in the analysis:
 Transient (vs steady state) seepage back-analysis focusing on piezometer responses (vs flow rates);
 Spatial (i.e., 3D) relationship of the piezometer responses;
 An in-depth historical review of geological information.

4.1 Transient Seepage Back-Analyses Based on Piezometer Response

The seepage analysis discussed in Section 3.3 focused on the flow rates under steady state conditions. In
order to further investigate the seepage mechanism, a series of 2D transient seepage sensitivity analyses on
the same model at Section L were performed. Three scenarios were examined according to the piezometric
responses in specific locations, varying hydraulic conductivity to explore:
 Scenario 1: sensitivity of the core;
 Scenario 2: sensitivity of the passive pressure relief system and downstream wick drains;
 Scenario 3: sensitivity of upstream of wick drains.

CDA 2019 Annual Conference, Calgary, AB, Canada 7

Scenario 1: sensitivity of the core

The steady state seepage analysis results as discussed in section 3.3 indicated that the conductivity of the
core had to increase to 1×10-4 m/s from 1×10-8 m/s in order to match calculated flow rates with the observed
ones. To further investigate the time-dependent piezometric head at selected Tips LP23 and LP10 (Figure
7), transient seepage analyses for the two cases were examined.

As shown in Figure 8, the calculated piezometric heads at LP23 with core k=1×10-8 m/s were insensitive to
pond fill while those with core k=1×10-4 m/s increased with pond fill. On the other hand, the observed
piezometric heads decreased with time which was due to dissipation of the excess pore pressures, which
can be considered to be insensitive to the pond fill as well.

Hl (Wick Drains Installed) 

Filter Drain  Collection Ditch 
Dyke Shell 
Dyke Core  Blanket Drain  Hl  
LP23 Pl (Wick Drains Installed) 
Pl (Wick Drains Installed)  Pl  
Relief Trench Well  LP10 Pfs 

Figure 7: 2D Transient Seepage model of section L

Figure 8: Comparison of calculated and observed piezometric elevation at LP23

CDA 2019 Annual Conference, Calgary, AB, Canada 8

 Figure 9: Comparison of calculated and observed piezometric elevation at LP10

The other calculated piezometric heads at LP10, as shown in Figure 9, were much lower than observed in
both cases and they were insensitive to the core conductivity. Although the seepage volumes were matched
between the calculated and observed flow rates by increasing the core conductivity, the unanticipated results
of piezometric responses suggested further analysis of the seepage field, which is discussed in
Scenario 2 and 3.

Scenario 2: sensitivity of the passive pressure relief system and downstream wick drain

One method to increase piezometric heads at LP10 to match the observed values is to lower the hydraulic
conductivities of the downstream soil units, especially those with high conductivities. The ranges of the
hydraulic conductivities used for the relief system and downstream wick drain (DSWD) were between
1×10-4 m/s and 1×10-6 m/s, and cases with different combinations were examined. It should be noted that
the conductivity of the core is set to 1×10-8 m/s based on the Scenario 1.

Figure 10 presents the results of the cases with the DSWD k=1×10-5 m/s and the case with relief system
k=2.5×10-5 m/s showed the best fit to the observed data. The similar trends can be seen in the cases with
DSWD k=1×10-4 m/s and k=1×10-6 m/s as well (Figures 11-12).

CDA 2019 Annual Conference, Calgary, AB, Canada 9

 Figure 10: Sensitivity analyses of relief system and DSWD (DSWD k=1×10-5 m/s)

Figure 11: Sensitivity analyses of relief system and DSWD (DSWD k=1×10-4 m/s)

CDA 2019 Annual Conference, Calgary, AB, Canada 10

 Figure 12: Sensitivity analyses of relief system and DSWD (DSWD k=1×10-6 m/s)

Scenario 3: sensitivity of the upstream wick drain

The best fit case in Scenario 2 was used to investigate the sensitivity of upstream wick drains (USWD), with the
results plotted in Figure 13. The parameters in the best fit case of Scenario 1 were used as well (i.e., USWD k=1×10-
4
).

Figure 13: Sensitivity analyses of USWD

Based on the above transient seepage analyses, the parameters of the core, relief system, the wick drains
are summarized in Table 2. It should be noted that this solution is non-unique and other combinations for
parameters could also match the data. As it can be seen, the conductivities of core, DSWD and relief system
are lower than those used in steady state seepage analysis in Section 4.3, which means the seepage volumes
from the transient seepage analysis cannot match the observed flow rates.

CDA 2019 Annual Conference, Calgary, AB, Canada 11

 Table 2: Comparison of Hydraulic conductivities between Steady-state and Transient Analysis
Material Steady state kh Transient kh
(m/s) (m/s)
USWD (Hl clay) 1×10-4 1×10-4
USWD (Pl clay) 1×10-4 1×10-4
DSWD (Hl clay) 1×10-4 1×10-5
DSWD (Pl clay) 1×10 -4
1×10-5
Till Core (Zone 1) 1×10 -4 1x10-8
Relief System 1×10-3 2.5x10-5

4.2 Spatial Relationship of the Piezometer Responses

Although the above 2D transient seepage analyses could explain the piezometric response at some locations,
the anticipated high seepage volume remained unresolved. For this reason, further instrumentation data
review was performed at Sections J-M (Figure 3). Two representative vibrating wire piezometer (VWP)
tips at each section were selected for investigation:
 Tip 1: located within the sand blanket near the toe of dyke core (Tip IDs: JP13, KP21, LP24 and
MP24);
 Tip 2: located within the sand blanket and approximately 70 m downstream of Tip 1 (Tip IDs: JP14,
KP22, LP25, MP25)

The tip locations within the sand blanket at Section L are shown in Figure 14. Section L is selected here for
the purpose of comparison with those discussed previously, while the other sections behaved similarly. The
following features can be interpreted and are shown in Figures 15 and 16:
 Piezometric elevations at both LP24 and LP25 increased with pond filling, but the most important
feature was that the piezometric elevations of LP25 were higher than that of LP24 after pond filling,
which indicates that water flowed from the downstream side towards the dyke core, and as a result,
the hypotheses concerning a leaky dyke core was not supported. See Figure 15.
 This circumstance was not observed at Section J, where the seepage gradient demonstrated flow
from the dyke core towards downstream side (i.e., from JP13 to JP14). See Figure 16 and the station
number in shown in Figure 19.
 Also of note in Figure 16 is that KP22 displayed the highest piezometric elevation among all the
tips, which suggests that there might be a discharge point in that area.

Hl
Pl
Pga, PGL1/2
Pfs, Pfsa
LP24 Pfg
LP25 Tip location

Figure 14: Illustrative tip locations within the sand blanket at Section L

CDA 2019 Annual Conference, Calgary, AB, Canada 12

 Figure 15: Piezometer response to pond infill in Section L

Figure 16: Piezometric elevations within sand blanket measured on Oct 6, 2012

4.3 Geological data review

As mentioned in Section 4.2, evidence suggested a discharge area near KP22, which is located within the
sand blanket approximately 70 m away from dyke core toe. The area near KP22 was considered to be a
discharge area because the hydraulic heads of KP22 are higher than those of other tips in the area.
However, none of the geological sections J-M in the vicinity supported this hypothesis. As shown in
Figure 14 for Section L and Figure 17 for Section K, continuous Pl/Hl layers existed beneath the dyke.

After interviewing personnel who previously worked on the ETA foundation preparation, the authors
became aware of a pre-existing waterbody, known as Hitchcock Lake in the seepage area as shown in
Figure 18, though the borehole logs from the area did not suggest any high hydraulic conductivity units.
When the boreholes were plotted on a plan of the Hitchcock Lake area (Figure 19) it was observed that
none of the boreholes had been drilled in close proximity to KP22, where the hydraulic heads were the
highest (illustrated as a dashed red circle in Figure 19). It was confirmed through an instrumentation
review that this area was located in the center of the former lake and was inaccessible for a drill rig at that

CDA 2019 Annual Conference, Calgary, AB, Canada 13

time. Despite the lack of geologic information at this precise location, it is conceivable that this was a
historical discharge area prior to dyke construction activities.

Hl
Pl
Pga, PGL1/2
Pfs, Pfsa/b
Pfsg
KP21
Tip location
KP22

Figure 17: Illustrative geological Section K

Figure 18: A photo of Hitchcock Lake prior to dyke construction

CDA 2019 Annual Conference, Calgary, AB, Canada 14

 . Borehole location
Discharging point (KP22)
Hitchcock Lake boundary
SCALE 
1:5000 

Figure 19: Plan view of former Hitchcock Lake with borehole and discharging locations

5 CONCLUSIONS

For the seepage issue discussed in this case study, a 2D steady state seepage analysis focusing on the flow
rates could not correctly map actual piezometric conditions when the observed seepage volume was
introduced to the 2D system. A transient seepage analysis focusing on the seepage field (e.g., piezometric
heads) was applied to seepage volume calibration modelling to better understand the seepage mechanism.
In addition, the spatial relationship of the piezometric heads provided evidence of a discharge area, in the
center of former Hitchcock Lake. Lastly, the sections selected for slope stability did not represent conditions
for seepage evaluation. The inclusion of small but important geologic features was vital to understanding
the behaviours observed. The proposed approach can be generally applied to similar projects.

6 ACKNOWLEDGEMENTS

The authors gratefully acknowledge the permission of Imperial Oil Resources to publish the findings of this
case study and the previous design engineers from Wood (formerly AMEC) and Golder Associates for their
comprehensive engineering services.

7 REFERENCES

Peck R., B. 1969. “Advantage and limitations of the observational method in applied soil mechanics” Géotechnique,
Vol.19 No.2: pp 171-187.

CDA 2019 Annual Conference, Calgary, AB, Canada 15