Slope Stability 2011: International Symposium on Rock Slope Stability in Open Pit Mining and Civil Engineering,

Vancouver, Canada (September 18-21, 2011)

Keynote Paper:

Investigating the Effects of Mining-Induced Strain in Open Pit Slopes

N.D. Rose Piteau Associates Engineering, Ltd., North Vancouver, BC, Canada

Abstract
Modern open pit slope design is increasingly requiring meaningful forecasts of slope deformation behaviour
during mining, in addition to traditional design approaches used to satisfy slope stability. As open pit mines
become deeper and stability mechanisms become more complex, optimizing ore extraction while maintaining
stability may require detailed design investigations to maximize slope angle while maintaining acceptable levels
of risk. In situations where ultimate pit mining encroaches on surface or underground infrastructure, design
acceptability criteria require incorporation of deformation tolerances in addition to traditional factor of safety
or probability of failure criteria. This requires numerical modeling techniques, such as distinct element
modeling, that incorporate a detailed representation of the engineering geology and groundwater conditions and
the mining excavation sequence. Validation of model input parameters is established through careful calibration
of model displacements with actual slope monitoring data during back analysis. Slope deformation forecasts can
then be used to define acceptable movement magnitude, rate and strain threshold criteria.
This paper describes the use of detailed distinct element modeling to investigate the influence of strain-induced
degradation (disturbance) of the rock mass that occurs in slopes as a result of stress relief and relaxation during
mining. Recognizing the potential adverse effects of mining-induced strain is important in defining appropriate
design strength parameters, and hence, successful geotechnical slope designs. As current analytical methods
may be limited in addressing the time-dependent strength behaviour of geological materials, deformation
forecasts must be continually evaluated with slope displacement monitoring to confirm the validity of analytical
forecasts. To safeguard against destabilizing slope conditions, if they were to occur, empirical failure time
prediction and runout assessment methods can be used to define slope monitoring thresholds and hazard zones.

Introduction

In jointed rock masses, structural geology often plays an important, if not controlling, role in slope stability.
Representing the influence of structural discontinuities in the design of stable bench, interramp and overall slope
geometries can be critical in defining and optimizing slope angles. Where geological structure is continuous
relative to the scale of an excavation, simple planar, wedge and toppling failures can be assessed using basic
limit equilibrium methods (Hoek and Bray, 1981). However, where geologic structure is discontinuous and
involves complex interaction with the rock mass, more complex failure modes may occur, such as step path or
complex wedge failures that combine shear or tensile failure through intact rock or rock bridges along the failure
path. Limit equilibrium analysis methods can be used to evaluate step path failures or instability that involves
rock bridging, but these methods do not provide an evaluation of deformation and strain related changes in shear
strength and rock mass conditions that precede progressive failure development.
This paper provides an overview of the capabilities of distinct element modeling in providing retroactive
assessment and predictive forecasts of slope deformation and stability mechanisms in open pit mine slopes.
General observations from modeling can provide valuable insight into the influences of strain-induced yielding
that can occur as a result of mining. Depending on the orientation and continuity of geological structure, mininginduced strain or creep displacements may lead to destabilizing conditions that result in failure over time, unless
they are recognized in advance and mitigated. Empirical methods provide a useful means of estimating the
timing and potential impacts of failure, and improve safety. Examples of these conditions are presented.

Distinct element modeling of complex slope stability and deformation

2.1 Background
Distinct element numerical modeling codes such as UDEC (Itasca 2011) and its three-dimensional equivalent
3DEC (Itasca 2003), provide the ability to model complex geological structure, rock mass and groundwater
conditions, the interaction of mining activity and related displacement behaviour, and the potential for
progressive failure development. This capability is advantageous for the investigation of slope movements that
develop prior to, and coincident with, a failure condition, or in environments where design tolerances to slope
deformations are low, such as in areas of permanent mine facilities, critical haul roads or crown pillars. Due to
the number of input parameters required in distinct element models, a high level of input data quality is required.
A more detailed description of the UDEC numerical modelling methodology summarized in this paper is
presented in Rose and Scholz (2009).

2.2 Assessment of slope conditions to determine numerical modeling requirements

In consideration of whether detailed numerical modeling may be appropriate for a given slope, an evaluation
should first be carried out to determine whether stability conditions can be adequately explained using simpler,
more traditional stability analysis methods. Factors that require consideration include the stability mechanism,
structural complexity and whether adequate information exists to calibrate a detailed numerical model and
provide meaningful modeling results. As shown in Figure 1, a series of checks can be applied to determine
whether detailed numerical modeling may be required:

Stabilitymechanism/
Slopeconditions

Simple

Complex

(e.g.,planarfailure,
brittlerock,no
infrastructure)

(e.g.,flexuraltoppling,large
moduluscontrasts,
infrastructure)

Standardkinematicsand
limitequilibriumanalysis
(e.g.,smallopenpitsor
strongrockmasses)

Detailednumerical
modelingnotrequired

Figure 1.

Limitequilibriumanalysis
ofrockbridgingorstep
pathmechanismsbased
onbackanalysis
DesignAcceptability
Criteriaaremet.
Deformationlevels
tolerable.

Deformationtolerancelow
ormechanismnotexplained
withstandardanalysis
approaches

Detailednumerical
modelingrequired

Chart illustrating general process for determining whether detailed numerical modeling is warranted.

Information that is required to carry out detailed UDEC modeling is summarized as follows:

a detailed geological model defining the distribution of major lithologies, faults and/or structural zones;
structural discontinuity fabric defining the orientations of predominant structural sets, persistence and
spacing, and peak/residual shear strength characteristics;
geotechnical properties defining the rock mass shear strength and deformation properties of the main
geotechnical units, as derived from laboratory testing and rock mass classification;
a comprehensive database of displacement monitoring data;
detailed historical and forecasted mining sequences;
piezometric monitoring and groundwater modeling results defining the distribution of pore pressures during
various stages of mine development and closure; and
information defining the insitu stress regime, if available.

2.3 Model representation of physical slope conditions

2.3.1 Structural discontinuity fabric and shear strength
Development of structural fabric in distinct element models requires scaled representation of semi-continuous to
continuous structural features that are important to kinematically possible failure modes or may influence slope
deformation. Non-important structural features such as discontinuous joints are generally accounted for in the
characterization of rock mass strength, as represented by finite difference zones (Section 2.3.2). Peak and
residual strength conditions should be incorporated to account for the effects of shear strain on discontinuity
shear strength. Figure 2 illustrates the modeling of discontinuities that exhibit peak-residual behaviour such as
joints, as compared to major fault zones that may only exhibit residual shear strength behaviour.
peak

peak
residual
residual

'
joint

residual
displacement
peak
residual

c'

fault
Model Structure

residual
displacement

Figure 2.

Illustration of peak and residual Mohr-Coulomb strength relationships for major and minor
discontinuities. Solid lines represent faults and dashed lines represent joints.

Model discontinuity (joint) properties that are subject to peak and residual slip criterion include:

joint friction angle (');

joint cohesion (c');
joint tensile strength, which is typically zero;
joint normal stiffness (jkn); and
joint shear stiffness (jks).

2.3.2 Rock mass strength conditions and yield softening criteria

As illustrated in Figure 3, rock mass conditions are represented in model finite difference zones using a MohrCoulomb elastic-plastic constitutive model. Mohr-Coulomb strengths are derived from the 2002 Hoek-Brown
criterion at maximum confining stresses ('3max), or alternatively normal stresses ('n), that are representative of
interramp to overall slope conditions. Strain softening criteria are simulated with FISH routines to update the
constitutive parameters of yielded finite difference zones after each excavation from undisturbed to disturbed
rock mass strength conditions defined by Hoek-Brown Disturbance (D) factors of 0 and 1, respectively.
Softening of rock mass modulus (Em) to disturbed values is carried out following each excavation step for each
finite difference zone that is indicated to be in a current state of yield or has yielded in the past. Rock mass
dilation angle is generally approximated as one-quarter to one-eighth of the rock mass friction angle, except for
fault zones or low quality rock masses where dilation angle is zero. Poissons ratio is estimated using laboratory
testing results or empirical estimates.
'1
Hoek-Brown (2002)
Mohr-Coulomb

t
rock
mass

'3

D=0

D=1
Undisturbed (D=0)

'1 or

*
Finite Difference
Zones

Disturbance
transition

Yield

'1

representative
'3max or 'n

t
'3

'3 or'n

Disturbed (D=1)

Figure 3.

Illustration of rock mass strength conditions represented in finite difference zones with MohrCoulomb yield criterion (strength envelopes represented by dashed lines). The Hoek-Brown (2002)
strength criterion is used to define undisturbed and disturbed conditions (solid lines).

As illustrated on the stress plot in Figure 3, the transition between disturbed and undisturbed conditions (thick
grey dashed line) in an excavated slope occurs as a function of the tensile or shear damage that develops from
stress relief and slope relaxation during mining. This transition can occur over significant depths and is sensitive
to the structural, rock mass and groundwater conditions in the slope. The general shape of the disturbance
transition curve may exhibit sub-linearity relative to the disturbed and undisturbed non-linear rock mass strength
envelopes. This sub-linearity may justify the use of Mohr Coulomb or bilinear strength criterion, or alternatively,
s-shaped strength criterion or rock damage transition behaviour described by others (e.g., Kaiser and Kim 2008,
Diederichs et al. 2007, and Carter et al. 2008). As shown on Figure 4, the range of stress or alternatively depth of
the disturbance transition zone can be estimated from the percentage of yielded elements and contacts in distinct
element models. This information can be used to estimate appropriate ranges of Hoek-Brown D factor for limit
equilibrium modeling. Selection of appropriate depths of disturbance requires consideration of whether slope
stability conditions are dominated by geologic structure, rock mass conditions, or a combination of both.

100
ACCUMULATED ELEMENT YIELD
ACCUMULATED CONTACT YIELD

90

ACCUMULATED ELEMENT AND

CONTACT YIELD

80

YIELD (%)

70
60
50
40
30
20
10
0
0

50

100

150

200

250

300

350

400

DEPTH (m)

Figure 4.

Example of cumulated percentage of element and contact yield versus depth in the 450m high pit
slope shown on Figure 7. Percentage yield can be used to estimate zonation of Hoek-Brown D
factor.

2.4 Calibration of existing slope conditions based on back analysis

2.4.1 Simulation of the actual stress-strain path of a slope with bench-by-bench mining
Due to the amount of time required to develop detailed distinct element models, a common approach is to model
slope conditions with a single excavation stage under elastic conditions, followed by analysis of elastic-plastic
conditions. The main disadvantage in simplifying the mining excavation sequence is that slope deformation and
stability is assessed relative to displacement endpoints rather than the actual stress-strain path that a slope
undergoes during mining. This is illustrated in the following example:
Bench excavation sequence

Displacement

Initial bench excavations

exhibiting regressive
displacement

10

11

12

13

Final bench excavations

exhibiting progressive
displacement

14

Actual
Modeled multi-bench
Modeled single stage

Time

Figure 5.

Theoretical displacement-time graph for multi-benched slope with acceleration indicated during
mining of the final benches.

Consideration of a single staged excavation may not identify the acceleration trend in slope displacements during
mining of the final benches. Acceleration trends in displacement rates can be important in identifying potential
for strain-induced degradation of structural and rock mass strength conditions, or progressive failure
development. This, in turn, could be important in the consideration of long-term slope stability conditions based
on the observation of increasing strain rates and magnitudes, even if the final slope configuration is determined
to be stable.
2.4.2 Representing the excavation sequence
Simulation of multi-bench mining conditions should include the staged representation of at least one prior
mining phase. This is important to represent the actual mass removed during mining and to limit inertial shock
effects in the model that may adversely (unrealistically) affect the modeling results. As shown on Figures 6 and
7, calibration is established through back analysis of slope displacements on the existing mining phase by first
removing the previous slope on a bench-by-bench basis. Following each excavation, time stepping is carried out
to equilibrium to verify that conditions are stable before proceeding to the next excavation stage.

Bench 1
2
3
4
5
6
7
8
8
10
11
12
13
14
15
16

Figure 6.

Calibration
Monitoring
Point

Forecast
Monitoring
Point

Previous Mining Phase

Current Mining Phase
Forecast Mining Phase

rock
mass
joint
fault

Illustration of benched mining sequence including representation of the previous and current mining.
Once model calibration is achieved, model forecasts are carried out for subsequent mining.

One advantage of simulating bench-by-bench mining is the ability to forecast incremental bench velocities as a
function of incremental model displacements relative to the bench mining schedule, as illustrated in Figure 7.

2.5 Forecasts of slope deformation and stability

Once modeled slope conditions are back analyzed and calibrated based on slope monitoring data (grey shaded
areas on inset graphs in Fig. 7), displacement and velocity forecasts can be developed for comparison with actual
monitoring installations during mining. These can be developed in the form of surface monitoring points for
comparison with prism monitoring data or as subsurface displacements that can be compared with inclinometer
monitoring data. Provided that the modeling results meet acceptability criteria, this provides a useful means of
defining threshold movement criteria that can be evaluated on an ongoing basis. Trigger action response
procedures (TARPs) can be developed to include re-evaluation of slope conditions, if actual displacements or
velocities exceed model forecasts.

Figure 7.

Calibrated UDEC model from south wall of Barrick Goldstrike Mine, Nevada, USA after Rose and
Scholz (2009), showing modeled (red) versus actual (grey) prism horizontal and vertical
displacements and slope distance velocities. Forecasted versus actual data (non-shaded areas on inset
graphs) show close agreement.

2.6 Identification of potential instability

Identification of potential indicators of instability from modeling results can be important with respect to both
short-term and long-term stability of a slope. A number of stability indicators can be used to identify the
potential for instability, including:

model plasticity and discontinuity slip defining patterns that are consistent with potential mechanisms of
instability;
strength reduction factor (SRF) indicating a low factor of safety;
elevated levels or contrasts in shear strain and/or a high percentage of yielding in rock bridges; and
elevated model velocity, indicating potential for continued creep-displacements or strain.

Empirical assessment of strain-induced instability

The ongoing effects of strain from mining-induced ground relaxation and stress relief, as well as seasonal
variations in pore pressure, can lead to time-dependent deformation and degradation of slope conditions leading
to progressive slope failure, as described by Martin (1993) and Zavodni (2000). To accommodate uncertainties
in time-dependent behaviour of slopes that may not be effectively accounted for in analytical models, empirical
methods can be used to safeguard against progressive slope failure by predicting the time and associated runout
distances of a potential slope instability.

3.1 Forecasting potential slope failure time using linear estimates of inverse-velocity trends

INVERSE VELOCITY (1/V)

The inverse-velocity method, developed by Fukuzono (1985), provides a useful tool for interpreting slope
monitoring data, with the objective of anticipating or predicting slope failure. The concept of inverse velocity for
predicting the time of slope failure was developed based on previous Japanese work and on large-scale wellinstrumented laboratory tests simulating rain-induced landslides in soil. The conditions simulated in the
laboratory were considered to be characteristic of accelerating creep (i.e., slow continuous deformation) under
gravity loading. When the inverse of the observed rate of displacement (inverse velocity) was plotted against
time, its values approached zero as velocity increased asymptotically towards failure. A trend-line through
values of inverse-velocity versus time could be projected to the zero value on the abscissa (x-axis), predicting the
approximate time of failure, as shown on Figure 8.

Failure

t0

Figure 8.

TIME

tf

Inverse velocity versus time relationships preceding slope failure after Fukuzono (1985).

Fukuzono presented three types of plots fitted to the laboratory data (i.e., concave, convex or linear), defined by
the following equation:
1

1
[ A( 1)] 1 (t f t ) 1
V

[1]

where t is time, A and are constants and tf is the time of failure. In the laboratory measurements preceding
failure, was found to range between 1.5 and 2.2. As shown on Fig. 8, the curve of inverse velocity is linear
when =2, concave when <2 and convex when >2. Based on the results of laboratory testing, Fukuzono
concluded that a linear trend fit through inverse-velocity data usually provided a reasonable estimate of the time
of failure, shortly before failure occurred.
As documented by successful predictions of slope failures in Rose and Hungr (2007), the assumption of linear
trend fits of inverse-velocity data can be applied to estimate the time of slope failure, provided that forecasts are
adjusted to accommodate potential trend changes. Figure 9a is a plot of inverse velocity versus time showing the
trends of nine survey prisms located at various elevations on the southeast wall of the Barrick Goldstrike BetzePost open pit, Nevada, USA, over the last six weeks preceding the S-01-A slope failure in August 2001. Filtering
(data smoothing) of two-hour robotic total station prism monitoring measurement data was achieved by
calculating six-day average (incremental) slope distance velocities to reduce the effects of instrument error in
low-level velocity values. As displacement rates increased, a clear inverse-velocity trend developed and began to
converge on a failure time of August 29, 2001. Linear regression was applied to the inverse velocity values,
which defined a coefficient of determination (R2) of 99% for all nine prisms. The failure occurred on the
predicted date encompassing an overall slope height of 550m and an estimated 47 million tonnes. The instability
occurred over several hours as a series of nested wedge failures and rock avalanches.

1.0

35

Predicted velocity curves (based

on inverse-velocity fits) compared
with actual velocity data

30

0.8

INVERSE VELOCITY (days/cm)

VELOCITY (cm/day)

0.9

0.7

25
20

15
10
5

0.6

45

0.5

40

35 30 25 20 15 10
TIME BEFORE FAILURE (DAYS)

0.4
S-189
S-190
S-205
S-219
S-220
S-221
S-249
S-262
S-265

0.3

0.2

0.1

Regression coefficient (R2) = 99%

for all inverse-velocity fits

0.0
45

40

35

30

25

20

15

10

TIME BEFORE FAILURE (DAYS)

18-Jul-01

Figure 9.

25-Jul-01

1-Aug-01

8-Aug-01

15-Aug-01

22-Aug-01

29-Aug-01

Plot of 6-day average slope distance inverse velocity (a) and velocity (b) versus time for nine prisms
from Zone S-01-A (time 0 was the observed time of failure). The right figure (c) shows the prism
locations, mean azimuth and plunge of displacement vectors (orange) and plane intersections of the
interpreted multi-planar complex wedge (dashed lines) after Rose and Hungr (2007).

In evaluation of the inverse-velocity trend lines in Figure 9a, it is noteworthy that the shallower sloping trend
lines, defining the minimum values of A, corresponded to prisms located higher on the slope where
displacement rates were greater and tension cracks were first observed. The steeper trend lines, representing
larger values of A, corresponded to prisms located lower on the slope that exhibited lower displacement rates,
but higher rates of acceleration up to the time of failure. These observations have proven useful in the
identification of conditions that could potentially lead to progressive slope failure in other slope instabilities, as
discussed in Section 3.3.

3.2 Assessment of debris runout potential and hazard zoning

Evaluation of the potential impacts of slope failure requires consideration of the potential failure mechanism, the
type and nature of the materials involved (e.g., rock or soil, peak or residual strength), the degree of saturation,
and the surface roughness or effective friction angle of the runout surface. Analytical codes, such as the
Dynamic Analysis code, DAN-W, developed by Hungr (1995), are available to evaluate the potential impacts of
debris runout from slope failure or rock avalanches. However, as is the case for other analytical methods, these
models require proper calibration, which may not be possible if deteriorating slope conditions are observed and
available time for modeling is limited. As a starting point, empirical methods can be used to estimate potential
failure runout distance based on empirical databases of historical failures or landslides.
One empirical method was developed by Tianchi (1983), who carried out statistical analysis of data from 76
major rock avalanches that occurred in the European Alps between 1901 and 1974. He found that a strong
correlation exists between avalanche volume (V) and the maximum vertical drop (H) to the maximum horizontal
distance travelled (L) by the avalanche, a relationship previously considered by Heim (1932) and Scheidegger
(1973). The tangent of the H/L ratio is referred to as the Fahrbeschung (), overall travel angle or runout angle,
and is defined by the angle measured from the top of the failure scarp to the limit of the runout debris. Figure
10a shows the graphical relationship defined by Tianchi (1983), with data points for four failures added from the
south wall of the Barrick Goldstrike Betze-Post open pit. Figure 10b shows the same relationship, but in terms of
the runout angle on the y-axis.

50
1

10
9
5

0.6

14

21
19 26

0.5

13

11

0.4

Ratio (H/L)

S-01-A
S-05-B
S-07-B
S-09-B

12
8

0.3

16

0.2

2
1
9

39

47

41
17 22 27 31
43
24

S-01-A
S-05-B
S-07-B
S-09-B
10

23
29
28
32
18
20 25

15

40

35

Runout Angle ()

0.9
0.8
0.7

50
55

48
44

61
66

52
34 40 46 51 57 58 63
33
38 45 49
60 62 64 70
37
56
67
36
54
69
53
42
65
71
68
30
59

75

74

14

30

21
3

19 26
7

11

20

12
8

76

10

73

13

23
28

29

18 20

25

32

39
47

41 43 50 55 61
17 22 27 31
48
15 24 35
66
44
52
16
34 40 46 51 57 58 62 63
33 38 45 49
60
64 70
56
37
67
54
36
69
53
42
65
71
68
30
59

75

74
76

0.1

73

0.09
0.08
0.07

72

72

0
0.1

10
100
Volume (106 m3)

1000

10000

0.1

10
100
Volume (106 m3)

1000

10000

Figure 10. Correlation between the H/L ratio and rock avalanche volume (a) compiled from a database of 76
European rock avalanches by Tianchi (1983). Data points for four failures are added from the south
wall of the Betze-Post open pit, Nevada, USA (inset legend). The right graph (b) shows runout angle
derived from the tangent of (H/L) on the y-axis.
Equation [2] was derived from the best fit least squares regression line on Figure 10a, using a logarithmic
relationship (with base 10) between the H/L ratio and volume in cubic metres. A standard deviation of 0.118
delineates the lower bound and upper bound dashed lines relative to the solid mean fit line.
log(H/L) = 0.664 0.1529 logV

[2]

An important observation from Figure 10b is that small failure volumes of about 100,000 m3 have a mean runout
angle of 38, which is a typical angle of repose for fragmented rock. By rewriting Eqn. [2] in terms of L, mean
horizontal runout distance in metres can be calculated as a function of slope height from Eqn. [3]:

L mean

H
10

0.6640.1529logV

[3]

An upper bound runout distance can be estimated from the minus one standard deviation H/L ratio from Eqn [4]:

L upperbound

H
10

0.5460.1529logV

[4]

As seen on Figure 10, the majority of the historical landslide data plot above the minus one standard deviation
line for volumes less than about 25 million cubic metres. This indicates that provided that failure conditions do
not define fluid runout potential, this relationship may provide an upper bound estimate of runout distance from
the known location of tension cracks on an unstable slope.

3.3 Documented instability on the South Wall of the Barrick Goldstrike Betze-Post Pit
Over the last ten years, four large slope instabilities have occurred on the south wall of the Barrick Goldstrike
Betze-Post open pit. The timing of each of these instabilities was predicted using the inverse-velocity method
described in Rose and Hungr (2007). The runout limits of the two most recent failures, S-07-B and S-09-B, were
predicted using the mean Fahrbeschung relationships defined by Tianchi (1983) shown on Figure 10 and in
Eqns. [2] and [3]. Two of the instabilities, S-01-A and S-09-B, occurred on ultimate pit walls following the
completion of mining. The S-05-B and S-07-B failures occurred during mining, but were forecasted in advance
and had no adverse impacts on the mining operation.
Table 1 summarizes the volume, mass and plan source area of the instabilities. Also included are the minimum
values of -A from the inverse-velocity trend lines that were used to forecast the failures. These trends were
derived from prism slope distance inverse velocity with data averaging increments ranging from two to six days.
Table 1.

Summary of Slope Instability Details on the Betze-Post South Wall between 2001 and 2009.
Failure

Volume
(Mm3)

Mass
Area
(Mtonnes) (103m2)

-A(min)
(cm-1)

Runout
Angle ()

Rock type

S-01-A

19

47

367

0.010

22.5

Intrusives

S-05-B

51

0.042

27

Metasediments

S-07-B

0.2

0.5

29

0.165

38

Metasediments

S-09-B

1.1

2.7

42

0.079

29

Metasediments

S-05-B

S-09-B

S-08-B

Intrusives

S-07-B

Metasediments
S-01-A

Figure 11. Plan map showing limits of Zones S-01-A, S-07-B and S-09-B. Zone S-05-B was mined out.

3.3.1 Relationship between failure size and inverse-velocity trends

Figure 12a shows the correlation between the minimum value of A and failure volume for the four historical
failures on the south wall (Table 1). The minimum values of A correspond to prisms in the active zones of the
instabilities, in the area of tension cracks in the upper portions of the failures. The relationship between A and
failure volume is defined by Eqn. [5], with an R2 of 98%.
A cm1 384V 0.628

[5]

Figure 12b is a plot of the minimum value of A versus the plan area of three instabilities in metasedimentary
rocks on the south wall, two of which are seen on Figure 13. As all of these failures involved a component of
shearing along bedding with consistent dips of about ten degrees towards the pit, the assumption was made that
failure size could be related to failure area based on an assumed relationship between depth and areal extent. The
relationship between A and plan area (square metres) of the failures is defined by Eqn [6], with an R2 of 99%.
A cm1 1.01e

6.18E05 PlanArea

[6]

S-07-B
S-07-B

S-09-B

0.1

S-05-B

S-09-B

0.1

-Amin (cm-1)

-Amin (cm-1)

S-05-B

0.01

S-01-A

0.01

Intact Ground
Failed Debris

0.001

0.001

0.0001
0.1

1
Volume (106 m3)

10

10

20

30
40
50
Plan Area (103 m2)

60

70

80

Figure 12. Correlation between the slope of linear inverse-velocity trend lines (A) and failure volume (a) for
four historical instabilities on the south wall of the Betze-Post pit. The right graph (b) shows A
versus the plan area of three failures in metasediments.
The data from the failures supports an inverse relationship between failure size and rate of acceleration, as
indicated by values of A. Larger failures tend to move at lower rates of acceleration than smaller failures, and
are therefore easier to detect. Similarly, creep displacements in previously failed (residual) materials commonly
move at low acceleration rates. This is illustrated by the blue symbol on Figure 12b that represents residual
materials within Zone S-09-B (Fig. 13). This suggests that inverse-velocity trends, once detected, can be
compared to potential failure size to determine the nature of the slope movement materials and associated
hazards. Intact materials may define brittle behaviour and therefore require lower monitoring thresholds to
accommodate higher acceleration rates and greater runout potential. Conversely, different monitoring thresholds
may be justified for residual materials on the basis of creep behaviour and low rates of acceleration.

S-07-B

IN-PIT WASTE
BACKFILL
S-08-B

S-09-B

WASTE ROCK
BUTTRESS/
ACCESS

Figure 13. View of the south wall of the Betze-Post pit showing S-07-B, S-08-B and S-09-B instabilities looking east.
3.3.2 Assessment of slope monitoring thresholds for intact ground
To determine whether established monitoring thresholds will provide adequate warning time of impending slope
failure, monitoring thresholds can be calculated from Eqn. [7], according to required warning time (T) prior to
failure.
MonitoringThreshold cm/day

1
T days

1
A cm1

[7]

By substituting Eqn [5] for A in Eqn [7], slope monitoring velocity thresholds for intact ground on the south
wall are plotted in Figure 14a according to different advanced warning times and potential failure volumes.
Figure 14b shows calculated monitoring velocity thresholds according to plan surface area from Eqn. [6]. These
plots indicate that lower thresholds are required to provide adequate warning time for smaller failures.
As an example of slope monitoring thresholds, Warning and Critical movement rates of 38 and 76 mm/day
(1.5 and 3 inch/day) were defined for the south wall, represented by the orange and red dashed lines on Figure
14, respectively. For a potential failure volume of about 100,000 m3, these thresholds correspond to calculated
advanced warning times of one day and 12 hours, respectively. As discussed in Section 3.2 and shown on Figure
10b, this volume defines the approximate limit where materials run out at an angle of repose of 38. Previous
instabilities of this size, or smaller, were contained by available catchment on the slope, but still require
consideration of rockfall potential. Zone S-08-B, seen on Figures 11 and 13, provides an example of these
conditions.

15

10

Critical
5

Warning

Warning Time
12 Hour
1 Day
2 Day
3 Day

15

7
6
5
4

10

Critical
5

3
2

Warning

Monitoring Threshold (in/day)

20

Warning Time
12 Hour
1 Day
2 Day
3 Day

Monitoring Threshold (cm/day)

Monitoring Threshold (cm/day)

20

0
0

0.2

0.4
0.6
Volume (106 m3)

0.8

0
0

10
20
30
Plan Area (103 m2) - Bedding Instability

40

Figure 14. Plot of monitoring threshold (intact ground) versus volume (a) and plan surface area (b) of slope
movement on the south wall for warning times of 12 hours to three days. A minimum of one to two
days advanced warning is considered necessary to provide adequate safety.

3.3.3 Hazard zoning and protection measures

Hazard zones are defined across the south wall, below areas of observed tension cracks, bounded by major
faults. Details of hazard zoning assessments carried out for mining on the 11th West Layback are discussed in
Armstrong and Rose (2009). Since mining was completed on the 11th West Layback in late 2009, a 70 to 100m
wide by 125m high waste rock buttress has been constructed to stabilize the lower south wall and to provide
access to the in-pit backfill dumps on the east side of the pit (Fig. 13). An approximately 8m deep catchment
ditch and rockfall impact berm is maintained on the inside of the buttress against the toe of the south wall.
Rockfall and debris runout modeling is carried out to evaluate potential hazards and design mitigation measures.
The south wall is continually monitored with a Reutech MSR 300 slope monitoring radar and a robotic total
station prism monitoring system. Both systems send warning messages to Mine Dispatch and geotechnical and
mine engineering senior personnel in the event of a threshold level being exceeded. Standard operating
procedures and TARPs are clearly defined to evacuate and barricade access to the slope within a distance of
180m of the highwall in the event of a Critical threshold being confirmed. The haul road to the in-pit backfill
dumps is periodically closed during elevated slope movement periods. Continual evaluation of slope conditions
is carried out to assess potential rockfall and runout hazards, and mine operating procedures are adjusted, as
required.

Conclusions

Rock slope stability mechanisms in large open pit slopes involve the complex interaction of geologic structure,
rock mass, groundwater and in-situ and induced stress conditions. Stress relief and slope relaxation related to
excavation induces strain that can result in dynamic changes in strength conditions over the life of a mined slope.
Even following mining, seasonal changes in pore pressures can result in strain-induced degradation of slope
conditions. To accommodate uncertainties in time-dependent behaviour of geologic materials that may not be
effectively accounted for in analytical models, empirical methods can be used to safeguard against progressive
slope failure, if it were to occur. The inverse-velocity relationship by Fukuzono (1985) and Fahrbeschung

relationships developed by Tianchi (1983) help improve safety by allowing prediction of the timing of slope
failure and associated runout distances. However, in the application of these methods, it must be recognized that
these tools can aid in the engineering judgement process, but should not be applied on the basis of misperceived
accuracy. This is particularly important if fluid runout potential may be present. The empirical relationships
developed by Tianchi (1983), and shown on Figure 10, are based on a database of fragmental rock avalanches
and should not be applied to saturated weak rock or soil-like materials.
When dealing with the potential for slope failure, a strict philosophy should be maintained where mining is not
allowed to continue on or below a slope that it is expected to fail. Adequate time (e.g., days) should be provided
for evacuation of established hazard areas, thereby safeguarding against the possibility that empirical or
analytical predictions are wrong. Not all acceleration trends lead to slope failure (e.g., regressive displacement,
strain hardening), and as such, false alarms should be anticipated as part of the ground control management
process. In that regard, the occurrence or recurrence of false alarms in slope failure time predictions should not
lead to complacency with respect to mine operating procedures or the arbitrary adjustment of threshold criteria in
unstable slope areas.
Data from four significant instabilities on the south wall of the Barrick Betze-Post open pit indicate an inverse
relationship between the minimum negative slope of the inverse-velocity trend line, an indicator of acceleration
rate, and failure size. This relationship can be used to define appropriate threshold monitoring levels to provide
adequate warning time of impending slope failure. Although experience with this approach is currently specific
to the south wall of the Barrick Goldstrike mine, it is presented with the anticipation that the general approach
may be successfully applied in other mining operations that have documented monitoring data from prior slope
instability. Further development of this method will hopefully expand on this experience.

Acknowledgements

The author would like to thank Barrick Goldstrike Mines Inc. and Barrick Gold Corp. for their permission to
present the case examples presented in this paper. Acknowledgement is also given with thanks to J. Armstrong
and R. Sharon of Barrick and A. Stewart and M. Scholz of Piteau Associates for their review of this manuscript
and their involvement and support with the material presented in this paper.
Figures 8 and 9 are reprinted from the International Journal of Rock Mechanics and Mining Sciences, Rose,
N.D. and Hungr, O., Forecasting potential slope failure in open pit mines using the inverse-velocity method, pp.
308-320, Copyright (2007), with permission from Elsevier.

References

Armstrong, J., Rose, N.D. (2009). Mine operation and management of progressive slope deformation on the south wall of
the Barrick Goldstrike Betze-Post open pit. Proceedings of Slope Stability 2009 Conference, Santiago, 9 p.
Carter, T.G., Diederichs, M.S., Carvalho, J.L. (2008). Modified Hoek-Brown transition relationships for assessing strength
and post yield behaviour at both ends of the rock competence scale. In Stacey et al. (eds.), Proc. of 6th Int. Symp. on
Ground Support in Mining and Civil Engineering Construction, SAIMM, Cape Town. pp. 37-60.
Diederichs, M.S., Carvalho, J.L., Carter, T.G. (2007). A modified approach for prediction of strength and post-yield
behaviour for high GSI rock masses in strong, brittle rock. In Eberhardt et al. (eds.), Rock Mechanics: Meeting
Societys Challenges and Demands, Proceedings 1st Canada U.S. Rock Mechanics Symposium, Taylor and
Francis, Leiden, 1, pp. 277285.
Fukuzono, T (1985). A new method for predicting the failure time of a slope. In Proceedings of the Fourth International
Conference and Field Workshop on Landslides. Japan Landslide Society, Tokyo, pp. 145-150.
Hoek, E., Bray, J.W. (1981). Rock Slope Engineering, 3rd Edition. Institution of Mining and Metallurgy, London.
Hoek, E., Carranza-Torres, C.T., Corkum, B. (2002). Hoek-Brown failure criterion - 2002 edition. In Hammah et al. (eds.),
Proceedings of the Fifth North American Rock Mechanics Symposium, Toronto. University of Toronto Press,
Toronto, pp. 267273.
Heim, A. (1932). Bergsturz und Menschenleben. Fretz and Wasmuth, Zrich, 218p.

Hungr, O. (1995). A model for the runout analysis of rapid flow slides, debris flows, and avalanches. Canadian Geotech
Journal 32(4): 610-623.
Itasca (2011). Universal Distinct Element Code (UDEC) Users Guide, Version 5.0. Itasca Consulting Group, Inc.,
Minneapolis, Minnesota.
Itasca (2003). 3D Distinct Element Code (3DEC) Users Guide, Version 3.0. Itasca Consulting Group, Inc., Minneapolis,
Minnesota.
Kaiser, P. K., Kim, B. H. (2008). Rock mechanics challenges in underground construction and mining. In Potvin et al.
(eds.), 1st Southern Hemisphere International Rock Mechanics Symposium, ACG, Perth, pp. 23-38.
Martin, D.C. (1993). Time dependent deformation of rock slopes. Ph. D. Thesis, University of London, August.
Rose, N.D., Hungr, O. (2007). Forecasting potential slope failure in open pit mines using the inverse-velocity method. Int.
Journal of Rock Mechanics and Mining Sciences 44: 308-320.
Rose, N.D., Scholz, M.F. (2009). Analysis of complex deformation behaviour in large open pit mine slopes using the
Universal Distinct Element Code (UDEC). In Proceedings of Slope Stability 2009 Conference, Santiago, 11 p.
Scheidegger, A.E. (1973). On the prediction of the reach and velocity of catastrophic landslides. Rock Mechanics 5: 231236.
Tianchi, L. (1983). A mathematical model for predicting the extent of a major rockfall. Zeitschrift fr Geomorphologie
27(4): 47382.
Zavodni, Z.M. (2000). Time-dependent movements of open-pit slopes. In Hustrulid et al. (eds.), Slope Stability in Surface
Mining. Society for Mining, Metallurgy, and Exploration, Inc., Littleton, Colorado, pp. 81-88.