Download as pdf or txt
Download as pdf or txt
You are on page 1of 55

Page 1 of 55

For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Canadian Geotechnical Journal (Author Accepted Manuscript)

Development and outcomes of a tailings slope stability comparative

design exercise

David Reid, PhD (Corresponding author)

ORCID: 0000-0002-1867-1676

The University of Western Australia, Perth, Australia


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

david.reid@uwa.edu.au

Andy Fourie, PhD

ORCID: 0000-0001-7250-2104

The University of Western Australia, Perth, Australia

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 2 of 55

ABSTRACT
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

Tailings storage facilities (TSFs) have undergone an unacceptably high failure rate for

decades, leading to an increased focus on improving the safety of these structures. One

aspect that has not received sufficient attention is examining how reliably and consistently

engineers analyse the stability of TSFs. An understanding of the current state of practice is

needed as there are a range of analytical methods available to engineers, while the high
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

failure rate of these structures strongly points to some deficiencies in practice. To examine

some of these issues, a tailings-focussed slope stability comparative design exercise was

organised to compare the methods and results used by tailings engineers to analyse the

same slope within which the phreatic surface was to rise under a specified pattern. Twenty-

eight practitioners participated in the exercise to predict at what phreatic surface level the

slope would fail. A wide range of predictions were made, ranging from assuming the slope

would liquefy and fail instantly (i.e. before any rise in phreatic surface), to a range of

techniques based on (i) yield strength ratios, (ii) stress path methods, and (iii) numerical

analyses, each predicting various failure levels, and finally some predictors who applied

drained strengths, resulting in predictions that the slope would not fail.

Keywords: slope stability, tailings, liquefaction

© The Author(s) or their Institution(s)


Page 3 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

INTRODUCTION
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

A number of significant tailings storage facility (TSF) failures - with subsequent investigations

of the causes - have occurred over the past decade, including Fundão (Morgenstern et al.

2016), Cadia (Jefferies et al. 2019) and Feijão (Robertson et al. 2019, Arroyo and Gens

2021). In each of these failures, a static rather than a seismic trigger was identified as the

primary cause. The specific causes of static triggering were identified by subsequent
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

investigations - being the stress path induced in contractive tailings from the deformation of

underlying materials (Fundão and Cadia), and for Feijão, one investigation suggesting loss

of near-surface suction and drained creep (Robertson et al. 2019), with the second indicating

drilling activities (Arroyo and Gens 2021). Importantly, for all the TSF failures listed, pre-

failure stability analyses had been carried out. While not all these pre-failure analyses are

publicly available, it seems reasonable to infer from the activities underway at each site that

imminent failure and subsequent loss of containment was not expected by site personnel, or

the engineers involved. This raises important questions about the current state of practice of

TSF stability analyses, a point highlighted by Jefferies (2021).

In order to assess the current state of practice of a particular sub-discipline of geotechnical

engineering, including how practitioners approach the inherent uncertainties required in

many different forms of analysis, comparative design exercises have proven useful

(Schweiger 1998, 2006, Orr et al. 2011, Bolzon et al. 2021). Given the current state of

practice for the analysis of TSF slope stability evidenced by the high failure rate, and the

ongoing focus on the improvement of TSF safety (ICMM 2020, Morrison 2022), it seems

clear that further investigation of the state of practice and comparison of techniques adopted

by engineers would be useful. To this end, the current paper presents the development,

organisation and results of a TSF slope stability comparative design exercise that was

recently carried out to assess some aspects of the current state of practice to infer the in situ

state of TSFs and analyse their stability.

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 4 of 55

PROBLEM DEVELOPMENT
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

General approach and trigger mechanism

Common features across many recent TSF failures include: (i) loose/contractive, saturated

tailings, (ii) static triggering of contractive undrained shearing (i.e. static liquefaction) leading

to flow. Static triggering herein refers to events preceding the flow event (often referred to
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

as the “trigger mechanism”) wherein the in situ stress ratio 𝜂=q/p' increased to an instability

value 𝜂IL for at least some portions of the in situ loose saturated tailings, leading to the

initiation of instability and resulting transition from drained to undrained conditions, brittle

strain softening and subsequent flow (Jefferies and Been 2015, Been 2016, Jefferies 2021).

Owing to the prevalence of these general features in recent failures, the comparative design

exercise was developed to allow a focus on design and analysis tools relevant to predicting

the initiation of instability and slope stability of loose contractive tailings. While there has

been much study of the potential for pre-failure instability of loose soils, including static

liquefaction, over the preceding decades (Sladen et al. 1985, Lade 1992, Sasitharan et al.

1993, Yamamuro and Lade 1997, Andrade 2009, Monkul et al. 2011, Jefferies and Been

2015) the authors’ hypothesis in commencing this exercise was that some aspects of

instability – particularly of sandy tailings – are insufficiently appreciated in tailings

engineering practice.

In terms of static triggers leading to the initiation of instability and transition from drained to

undrained conditions, three commonly discussed mechanisms include (i) placement of

additional fill, often at a high rate of placement such that increments of undrained loading

occur in underlying saturated materials, (ii) a rising phreatic surface, i.e. a constant shear

drained (CSD) stress path (Brand 1985, Anderson and Riemer 1995, di Prisco and

Imposimato 1997, Chu et al. 2012), and (iii) underlying deformations (Casagrande 1965,

Smith 1969, Hicks and Boughrarou 1998), often referred to as basal or lateral extrusion

(Jefferies and Been 2015, Morgenstern et al. 2016). Of course, while it is useful to

© The Author(s) or their Institution(s)


Page 5 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

conceptualise these mechanisms in isolation, each may be contributing to the stress path
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

followed by in situ elements of contractive and saturated tailings within a TSF over their

typically decades-long development.

In this exercise, the CSD stress path resulting from a rising phreatic surface was selected as

the trigger mechanism because of the ease with which the changing phreatic surface could

be defined unambiguously for participants and used as a discrete value for the point at which
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

failure was predicted by participants. This selection was further informed by the authors’

experience as third party reviewers of TSF stability analysis works, where we have seen

significant uncertainty and doubts raised regarding the plausibility of static liquefaction of

sandy soils, particularly on a CSD stress path - an issue which appears to be common (e.g.

Jefferies 2021). The scepticism remains despite decades of evidence, including element

tests (Sasitharan et al. 1993, Skopek et al. 1994, Wanatowski et al. 2010, Chu et al. 2012,

Reid and Fourie 2019, Fanni et al. 2022), documented case histories (Olson et al. 2000,

France et al. 2021), physical models (Eckersley 1990) and their interrogation through

numerical analysis (e.g. Cuomo et al. 2019) – all of which unambiguously point to the ability

of loose saturated soils to undergo pre-failure instability on a CSD stress path at values of η

lower than the critical state stress ratio M.

Scenario development

Various options as to the form of the exercise were available, including back-analysis of a

documented failure, use of existing 1g or centrifuge model failures (Take et al. 2004, Zhang

and Askarinejad 2020), or development of a hypothetical scenario. Direct analysis of a

documented failure case history was avoided, as the available data and conclusions of

published investigations would invariably influence the outcome. While 1g or centrifuge

models brought to failure are appealing given their directly observed behaviour, it was

desired that the participants would be required to infer in situ state and strengths based on

cone penetration test (CPT) results – by far the most common in situ test currently used in

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 6 of 55

the investigation of TSFs. This further ruled out the use of available centrifuge or model
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

studies of the CSD stress path, which generally do not include CPT data of the failed slopes.

Therefore, development of a hypothetical TSF slope that incorporated many salient features

of real TSFs was selected as the basis of the exercise.

The geometry of the idealised TSF used in this exercise is presented in Figure 1. Although

this is an idealised geometry rather than a real slope, the general features of the slope are
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

typical of many TSFs. The slope was 50 m high with 1V:3H side slopes. A 10m wide bench

was included at a height of 25m, making the overall average slope 1V:3.2H. An initial

phreatic surface slope of 1V:9.2H was present in the “initial conditions” of the slope,

commencing from 30 m upstream of the toe. This is a common situation for fallow TSFs

after the end of deposition. However, while a slow decrease in phreatic surface following

closure is common to TSFs in arid or semi-arid climates, several conceivable subsequent

events such as recommissioning of the facility or a degradation of surface water

management structures leading to an increased presence of surface water could result in a

subsequent rise in the phreatic surface. Indeed, the Simmergo TSF (Blight 2010) was an

inactive facility that, following recommissioning, underwent a rapid rise in phreatic surface

and subsequent failure, plausibly involving a CSD stress path.

Included in Figure 1b is the final phreatic surface profile of relevance to the exercise. The

outline of the program provided to participants indicated that the phreatic surface would rise

steadily (without quantifying the actual time over which this would occur) from the initial to

the final profiles shown in Figure 1b. To enable a simple definition of the phreatic surface

conditions and level, it was to be assumed that the phreatic surface at X=230m would not

change, this point would thus act as a “pivot”. For example, the initial conditions being a

phreatic surface level of Y=25m, and the final conditions Y=77m. It is further noted, as

highlighted in Figure 1b, that while the phreatic surface slope would be extrapolated up to

values of Y=77m, the pore pressures would be capped as 0 kPa at the surface of the TSF,

© The Author(s) or their Institution(s)


Page 7 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

i.e. a phreatic surface at ground level on the TSF beach represents the maximum condition,
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

and a depth of standing water was not to be considered.

TAILINGS MATERIAL PROPERTIES

General

The use of data from recent failures was selected for the exercise, as opposed to more
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

general tailings data from the geotechnical literature, as the failure case histories are

particularly relevant to the investigation of tailings stability analyses. Further, their use may

serve to avoid subsequent debates as to whether the tailings material properties provided

are consistent with those likely to, when saturated and loose, undergo pre-failure instability,

static liquefaction, and flow. The publicly available data from the Fundão and Cadia TSF

failure investigations were therefore utilised for this exercise. These data were relevant as

(i) they represent a reasonable spectrum of tailings gradations, and (ii) the Fundão CPT

probing conditions were drained, whereas probing at Cadia was undrained, enabling both

types of behaviour to be examined in the study while avoiding partially drained CPT data

which otherwise introduces additional uncertainties in interpretation (e.g. Reid and Smith

2021).

Element test data

To avoid complications related to interpretation of laboratory data and thus retain a greater

focus on slope stability analysis methods, laboratory data was provided as critical state

parameters and other inputs to the NorSand constitutive model (Jefferies 1993) similar to

those developed by the Fundão and Cadia Panels as outlined in Table 1. These inputs were

slightly modified from those developed by the failure panels to obscure their origins. The

NorSand constitutive model was selected as the “vehicle” to provide the laboratory data as

there are publicly-available NorSand tools to examine predicted element behaviour (the

“VBA spreadsheet”) and to develop material-specific relationships between CPTu results

and the state parameter - for example, see Jefferies and Been (2015) and Shuttle and

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 8 of 55

Jefferies (2016). It is emphasised that this form of data provision was not to specifically
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

promote the use of NorSand in the exercise, but rather was a consequence of NorSand’s

common usage in TSF engineering practice and availability of a simple tool to extract

various element test data from provided numerical model fits. As will be seen, the use of

NorSand in this context did not appear to lead to those who elected to use advanced

numerical models favouring NorSand over other alternatives.


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

The element behaviour predicted by NorSand for the parameters listed in Table 2 is

presented in Figure 2 for isotropically consolidated undrained (CIU) triaxial compression

element simulations with p’c= 100 kPa and initial state parameter Ψ0 = +0.05. The greater

initial stiffness for Scenario A and more brittle behaviour (primarily from a less compressible

CSL slope λe) are clear. This contrast in stress-strain behaviour, particularly post-peak

brittleness, was explicitly sought for the purposes of this study to examine how different

participants would approach the two scenarios.

By carrying out a number of element simulations as per Figure 2, and adopting the initial

peak as yield strength su(yield) and either the quasi steady-state QSS (where relevant), or

final critical state as liquefied shear strength su(LIQ), trends obtained for both scenarios are

presented as a function of Ψ0 in Figure 3. For specimens exhibiting QSS, the strength at

this point has been taken as representing su(LIQ) owing to frequent arguments as to the

likely inability for in situ soils to undergo post-QSS dilation, and thus the relevance of the

QSS measured in element tests to field-scale behaviour (Meneses-Loja Jorge et al. 2000,

Reid and Fourie 2014, Sadrekarimi 2014, Jefferies and Been 2015, Reid et al. 2022) – also

consistent with observations of flow liquefaction in deposits such as Nerlerk and Fundão with

significant proportions of in situ Ψ < 0 . The expected tendency of increasing strength and

reduced brittleness with decreasing Ψ0 are clear from the results. Also evident is the lower

magnitude of brittleness produced by Scenario B, and the greater propensity for QSS

behaviour to extend to lower (i.e. slightly negative, in this case) values of Ψ0 for Scenario A.

© The Author(s) or their Institution(s)


Page 9 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

In situ penetration test data - development


For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

The CPT data from both case histories were processed for the exercise through (i) use of

stress-normalised tip resistance Q, Friction ratio Fr, and normalised excess pore pressure Bq

from the original datasets to allow calculation of entire CPT soundings across the range of

stresses relevant to each hypothetical CPT in the exercise, (ii) application of a small random-

number scaling to slightly modify the values (while maintaining the same approximate
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

characteristic values as the source data), and (iii) increase of the tip resistance values above

the phreatic surface owing to the likely effects of suction/partial saturation (e.g. Russell et al.

2022) – with the source CPTs having been almost entirely saturated, therefore requiring this

modification given the specified initial phreatic surface for this exercise. The produced

results for CPTu 1 are presented in Figure 4 for both scenarios. Further details on the data

processing of the CPT data, along with the electronic data provided to participants, is

included in a project-specific data repository as indicated in the paper’s data availability

statement.

The data developed for both scenarios is compared in terms of state parameter Ψ (Been and

Jefferies 1985) in Figure 5 for all three CPTu probes (see Figure 1 for probe

locations). State parameter is estimated using the approach outlined by Plewes et al. (1992)

(i.e. the “Plewes Method”) for the purpose of this comparison, with a site-specific value of Mtc

of 1.45 and an assumed K0 of 0.7. Results from above the phreatic surface are presented in

Figure 5 also using the Plewes method, simply to show what range of values would be

obtained if one applied this method to unsaturated conditions without further consideration of

the method’s inapplicability under such conditions. The data for both scenarios are also

presented in a Q(1-Bq)+1 vs Fr plot in Figure 6 as average values of the saturated and

unsaturated portions of each sounding for comparison purposes.

SCREENING-LEVEL SLOPE STABILITY ANALYSES

To assess the characteristics of the slope for initial benchmarking and illustration, a series of

limit equilibrium (LE) stability analyses were carried out using the Rocscience Slide2 code,

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 10 of 55

with optimised failure surface search routines using the Morgenstern-Price method of slices.
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

Figure 7a shows the slope with a yield strength ratio (su(yield)/σ’vc) required to produce a

FoS of unity for the slope, indicating a value of 0.16 is required – this being below that

inferred from back analysis of any static slope failures of which the authors are aware (e.g.

Olson and Stark 2003). This suggests that stability analyses using su(yield)/σ’vc -based

approaches would be unlikely to suggest failure until at least some phreatic surface increase
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

had occurred. Figure 7b presents the maximum height of the phreatic surface with the

corresponding su(yield)/σ’vc that will produce a FoS of unity. Failure of the slope - using this

screening-level approach - is therefore suggested for any su(yield)/σ’vc below 0.32. This

su(yield)/σ’vc is higher than essentially all the back analysed failures of which the authors are

aware. Therefore, participants who recognised the contractive nature of the tailings and

carried out some form of stability analysis using su(yield)/σ’vc – based approaches were likely

to conclude that the slope will fail prior to the final phreatic surface being reached.

An additional analysis carried out with the final phreatic surface (77m at X=0) using the

critical state friction angle of 36 degrees (i.e. a drained analysis) produced a FoS of 1.9 as

shown in Figure 7c, thus suggesting that participants who selected such a strength input for

the tailings would conclude that the slope would not fail. In other words, if the slope were

analysed solely in terms of drained strength parameters, failure was unlikely to be predicted,

even for the maximum height of phreatic surface.

PARTICIPANT SUBMISSIONS

A total of 28 participants made submissions for the round robin: 23 for Scenario A, and 25

for Scenario B. Participant information, along with the nominated failure values submitted by

each of the participants are listed in Table 3 and summarised in histogram format in Figure

8. A wide range of results for both scenarios is apparent, with some participants suggesting

the slope would fail under initial conditions (Y = 25m), while others concluded the slope

would not fail even upon reaching the highest level relevant to the exercise

© The Author(s) or their Institution(s)


Page 11 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

(Y=77m). Clearly, this represents a significant variation in results, suggesting that different
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

methods and assumptions were applied by participants.

To organise the results in a rational framework, the entries are classified into the following

categories based on the analysis method and/or strength input values assigned in a

particular method, as also indicated in Table 3.


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

- LE Analyses:

o Drained strengths

o Yield strength ratios, su(yield)/σ’vc

o Liquefied strength ratios, su(LIQ)/σ’vc

- “Stress path” analyses, where in situ stresses were estimated either using

spreadsheet-based methods or carrying out a relatively simple elastic analysis of the

slope geometry to allow interrogation of in situ η values and their proximity to

potential triggering ηIL values.

- Finite element / finite different (FE/FD) analyses, where participants attempted to

model the entire process of phreatic surface rise, potential initiation of instability, and

post-peak shearing response in a single FE/FD analysis.

However, prior to interrogation of the different methods adopted, the technique used by

participants to assess Ψ are discussed, given the prevalence of such approaches by

participants and to frame the discussion in a similar order to the analysis steps taken by

many.

State parameter estimates

State parameter was analysed by 21 of the participants, either for direct use in numerical

models, stress path calculations, to provide an input to Ψ vs strength profiles, or simply to

give a screening-level understanding of the in situ state of the tailings. The state parameter

methods applied are outlined as follows:

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 12 of 55

 Been et al. 1987 method, based on comparison of CPTu calibration chamber results
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

to laboratory derived CSLs for the same sands. In this approach, a fit between mean

stress normalised tip resistance Qp and Ψ is developed using the CSL slope for the

material under consideration, drawing on the calibration chamber data for other

sands. In this method, the CSL slope relevant to the CPTu data being examined is

required.
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

 The Plewes et al. (1992) screening method (often referred to as the “Plewes

Method”), which extends the previous work of Been et al. (1987) to include both

partially drained and undrained penetration. In this method, the required input of

CSL slope λ10 can be taken either from available laboratory testing or is inferred

through empirical correlations to Fr. A value of Mtc is also required, which while

initially taken as 1.2 in the development of this method is more commonly now taken

as a site-specific value or estimate, given the generally higher values of Mtc relevant

to tailings (e.g. Reid 2014).

 The Robertson (2010a) screening method, where Ψ is calculated based on a

relationship to clean sand corrected dimensionless penetration resistance Qtn,cs, the

derivation of which is described by Robertson (2009). It is noted that this method

was developed with a focus on sandy soils where CPTu probing will be drained, and

that in analyses of TSFs that include both drained and undrained probing conditions

it is common to omit use of this method for undrained soundings (e.g. Robertson et

al. 2019).

 Cavity expansion techniques, where NorSand is used to simulate cavity expansion to

obtain the limiting cavity pressure under either drained (Shuttle and Jefferies 1998,

Ghafghazi and Shuttle 2008) or undrained (Shuttle and Cunning 2007, Shuttle and

Jefferies 2016) conditions, and the results are scaled to those of the CPT using the

Cq scaling factor as outlined by Shuttle and Jefferies (2016).

© The Author(s) or their Institution(s)


Page 13 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

A summary of the methods applied by participants in this study is outlined in Table 4. The
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

Robertson (2010a) method was adopted by eight participants, 11 adopted the Plewes

method (Plewes et al. 1992), one used the method proposed by Been et al. (1987), and

three carried out cavity expansion simulations and development of material-specific

calibration as per the methods of Shuttle and Jefferies (2016). It is noteworthy that six

participants applied the Robertson (2010a) method to Scenario B, despite this scenario
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

featuring undrained penetration conditions likely outside the range of applicability of the

method. Another interesting observation is that all the participants who used the Plewes

method adopted a site-specific value of Mtc sourced from the available laboratory data, which

appears likely to be common industry practice and should produce more accurate results.

Another aspect of inferring Ψ that warrants discussion is the assumed value of K0, as this is

required for the use of Been et al. (1987), the Plewes method, or cavity expansion

approaches, as all are formulated in terms of Qp. There are persuasive arguments for taking

a value of 0.7 as a first assumption (Jefferies and Been 2015), based on geological

processes that occur in situ that are not reproduced in typical normally consolidated

laboratory element tests. As indicated by the results in Table 4, there does appear to be a

tendency for participants to adopt values higher than the ~0.5 that a reconstituted K0 test or

the Jaky equation would suggest.

Limit equilibrium methods


Drained strengths

All participants made use of drained strengths in their analyses in some form - most

commonly derived on the basis of the laboratory-measured value of Mtc, and then assigned

to material above the phreatic surface. However, of note are five participants who adopted

drained strengths for their analyses of Scenario A - including for saturated contractive

material below the phreatic surface (refer Table 5). This outcome with respect to Scenario A

is arguably one of the most important insights from this exercise. It is yet again emphasised

that in the program description it was made clear that the term “drained”, where used, was

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 14 of 55

only relevant to the CPTu probing conditions. This important outcome was expected by the
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

authors and seems consistent with other commentaries on the state of geotechnical practice

related to tailings (Davies et al. 2002, Jefferies 2021).

Yield strength ratios

Analysis of slopes comprising saturated contractive materials using a su(yield)/σ’vc approach

has a long history. For example, clayey soils have been characterised using undrained
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

strength ratios for decades (e.g. Ladd and Foott 1974), while a method for carrying out

triggering assessments for static triggers (such as the CSD stress path) was formalised by

Olson and Stark (2003) with various refinements incorporating loading direction (Sadrekarimi

2014) and more complex methods to incorporate brittle strength redistribution later proposed

by Sadrekarimi (2016). The use of an su(yield)/σ’vc approach was adopted by 11

participants, with their specific methods outlined in Table 5.

The primary input required for use of this approach is a su(yield)/σ’vc value. Three

techniques were adopted by participants using a su(yield)/σ’vc approach:

 Use of empirical CPTu techniques based either on Olson and Stark (2003) or

Sadrekarimi (2014). It is noted that the case histories used to develop these

correlations are dominated by drained CPTu soundings, and thus are likely to be

more directly applicable to Scenario A.

 Use of the “Nkt” method to estimate su(yield)/σ’vc directly from the CPTu. It is

emphasised that the Nkt method is only relevant (at least in its common form) to

undrained CPTu penetration.

 Estimating Ψ in situ from the CPTu results and developing a Ψ vs su(yield)/σ’vc

relationship by means of NorSand element simulations.

As indicated in Table 5 the application of these methods was broadly similar across those

participants adopting a su(yield)/σ’vc approach. There is a logical tendency by the

participants to adopt the Olson and Stark (2003) and Sadrekarimi (2014) approaches more

© The Author(s) or their Institution(s)


Page 15 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

commonly when analysing Scenario A and the Nkt for Scenario B, consistent with the CPTu
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

drainage conditions and applicability of the two techniques to each scenario. Where the Nkt

method was adopted for Scenario B, Nkt values ranging from 14 to 16 were adopted,

consistent with typical recommendations when other data to calibrate this value is not

available (e.g. Robertson 2009). Finally, as the Ψ -based methods are conceptually

independent of CPTu drainage conditions, they find application in both scenarios.


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

The entries adopting LE techniques with su(yield)/σ’vc strengths are summarised in Figure 9

for both scenarios. Where participants adopted a range of strengths, either to accommodate

shearing mode or variation in in situ conditions, a range and average value are indicated.

Participants who only applied a particular strength to small zones within the saturated

tailings or did not select strengths on the basis of stress-normalised format, are excluded for

clarity. Included in Figure 9 are benchmarking analyses carried out by the authors wherein

strengths were iteratively adjusted for all tailings below the phreatic surface to obtain FoS=1

(using both circular and composite/optimised failure surfaces) for a range of phreatic

surfaces. While the results of the entries included in Figure 9 share the same trend,

generally, to the benchmarking curves, there is appreciable scatter. Examination of the

entries indicates that this variation is largely a result of participants either adopting a range of

su(yield)/σ’vc values (either from CPT analysis, loading mode, or both) or through differing

assumptions regarding the exact extent of where contractive material was located (e.g.

including zones above the phreatic surface, or extending the contractive zone closer to the

toe). Additional information on the geometric strength assumptions of participants, where

available, is included in the paper’s data repository.

It is finally noted that for certain cases with a low phreatic surface and low strength the

provided geometry was of insufficient width to avoid constraining circular failure surfaces.

Both the constrained (using provided geometry) and an extended geometry (to avoid

constraints) benchmarking curves are provided in Figure 9. Fortuitously, this geometric

constraint did not affect the outcome of any of entries received.

© The Author(s) or their Institution(s)


For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Canadian Geotechnical Journal (Author Accepted Manuscript) Page 16 of 55

Liquefied strength ratios

It is increasingly advocated that, given the clear risk posed by failure of brittle, contractive

materials such as tailings, ensuring stability of the structure even if liquefied strengths are

mobilised is essential. For example, Robertson (2010b) makes this case, with reference to

previous arguments put forward by Silvis and de Groot (1995) regarding loose sand
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

deposits. Morgenstern (2018) argues that for the purpose of preliminary design, liquefaction

(and thus mobilisation of a su(LIQ)/σ’vc) should be assumed. Similarly, Been (2016) and

Jefferies (2018) make similar points on the basis of the difficulty in accounting for post-peak

brittle stress redistribution in engineering practice. Recently published international

guidelines on tailings standards can also be interpreted to generally advocate this position

(ICMM 2020).

In the current study, three participants indicated failure under the initial phreatic surface

(Y=25m) because the slope, in this initial condition, had a FoS less than 1.0 when analysed

with their estimates of su(LIQ)/σ’vc applied to tailings below the phreatic surface. These

entries are distinct from those who analysed the slope adopting liquefied strengths as an

outcome of a stress path analysis (discussed subsequently). While, as noted previously,

adopting such an assumption in the design of high-risk structures containing loose, brittle

materials may be prudent, a distinction must be drawn with the explicit stated purpose of this

exercise - which was to identify when participants thought the slope would actually fail. It

could therefore be concluded that the increasing advocacy of design methods based on the

assumption of liquefied strength mobilisation has led participants to assume that liquefaction

will occur always, regardless of whether in situ stress conditions have reached a state where

triggering is likely to occur or not.

Table 5 includes the techniques adopted by participants who used su(LIQ)/σ’vc in LE analysis

solely on the basis of the contractive nature of the tailings highlighted. Liquefied strengths

© The Author(s) or their Institution(s)


Page 17 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

were inferred either using empirical relationships to CPTu data (Olson and Stark 2002,
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

Robertson 2010b, Sadrekarimi 2014) or adopting friction sleeve measurements as

remoulded strength for fine-grained soils (e.g. Robertson 2010, 2021). It is noted that

essentially any assumed value of su(LIQ)/σ’vc would produce a FoS < 1.0 at Y=25m if applied

to all tailings below the phreatic surface, which accounts for this answer provided by three of

the participants (17, 18, and 20) adopting su(LIQ)/σ’vc. Participant 25 assumed liquefaction
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

on the basis of the contractive state of the tailings but predicted a much higher phreatic

surface as being required for failure owing to applying liquefied strengths to limited, discrete

portions of the tailings.

Stress path methods

Stress path methods (e.g. Yamamuro and Lade 1997) were adopted by six participants,

wherein using either spreadsheet-based analytical calculations or simple elastic models of

the TSF, the in situ stress state was estimated and its proximity to ηIL was estimated It is

important to first note that there are a number of important assumptions required when

carrying out this type of analysis, many of which can have a significant effect on the outcome

(e.g. Reid et al. 2021):

 What in situ stress conditions are assumed and/or produced (for example by use of a

particular constitutive model inputs), in terms of geostatic stress ratio K0 (at a

minimum) or, more broadly, intermediate principal stress ratio b and principal stress

angle 𝛼 below the slope in areas where an assumption of axisymmetric conditions

with vertical principal stress angle is no longer reasonable.

 Will liquefied strengths be assumed in all contractive zones once one element has

reached the selected value of 𝜂IL, some proportion/quantity of elements, or direct

modelling of brittle stress redistribution as different elements are triggered and shed

their shear stress to adjacent elements, potentially generating a progressive failure.

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 18 of 55

A summary of the assumptions and methods made by each of the five participants who
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

adopted stress path approaches are outlined in Table 6, along with the method for liquefied

strength estimation in triggered zones for subsequent LE stability analyses, where relevant.

Axisymmetric conditions are used for the presentation in Table 6 as most participants made

such a simplifying assumption in spreadsheet-based analyses they carried out.

The assumptions made by participants adopting the stress path approach clearly vary
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

widely. For example, the K0 values, either explicitly assumed, or resulting from an input of

Poisson’s ratio to an elastic model, varied from 0.41 to 0.80. This assumption itself

drastically affects how much phreatic surface rise is required for the predicted stress path to

reach ηIL. To highlight this distinction, the initial stress state and selected value of 𝜂IL for four

participants using stress path approaches is outlined in Figure 10 (Participant 24 is excluded

as they adopted a mobilised su(yield)/σ’vc trigger criterion).

Also contrasted in Figure 10 are the stress paths the participants assumed – with most

adopting a constant deviator stress (CSD stress path), whereas one assumed a stress path

“perpendicular” to the critical state friction line (Mtc = 1.45) - i.e., assuming that the rising

phreatic surface led to some additional evolution of stresses beyond simply a reduction in

mean effective stress. In addition to the initial stress state and stress path assumptions, the

values of 𝜂IL adopted vary significantly - for example, for Scenario B from 0.82 to 1.45 (i.e.

Mtc), with this variation appearing to result not from different inferred Ψ values but owing to

different fundamental views by these participants as to whether instability would occur at η

lower than critical on a CSD path. Owing to the variation seen in each of these forms of

inputs, a meaningful trend of predicted failure phreatic surface and any single input is not

discernible in the stress-path method results.

It is interesting that the assumption by some participants of Mtc as the stress ratio wherein

instability would initiate, was despite the saturated material within both scenarios clearly

being contractive, and therefore having the potential to undergo instability at lower values of

© The Author(s) or their Institution(s)


Page 19 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

η. Further, adoption of Mtc ignores the likely effects of b on critical friction ratio M in below-
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

slope plane strain conditions (Wanatowski et al. 2010, Jefferies and Shuttle 2011, Reid

2020, Shuttle et al. 2021). Clearly, current engineering practice has not developed a

consensus as to the application of the stress path approach for triggering analyses.

FE/FD analyses – Models other than NorSand

Four participants made use of constitutive models other than NorSand, namely the Plastic
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Hardening and Strain softening models included with FLAC and the Sanisand model

(Dafalias and Manzari 2004, Taiebat and Dafalias 2007) implemented as a user defined

model in Plaxis. A summary of the methods used by these four participants are outlined in

Table 7. One participant (Entry 26) carried out undrained perturbations of the model by

increasing gravity after each phreatic surface increase to check model equilibrium, while the

remainder increased the phreatic surface while continuously monitoring model equilibrium

without applying an undrained perturbation. The stress-strain behaviour of the calibrated

models used by these participants for Scenario B are presented in Figure 11 for an

undrained simple shear element test from an isotropic consolidation effective stress of 100

kPa.

The element behaviour shown in Figure 11 indicates significant differences in the calibrated

model behaviours, both with respect to yield/liquefied strengths and the rate of post-peak

strength loss. Indeed, the Plastic Hardening model – at least as calibrated by the participant

adopting this model - exhibited perfectly plastic behaviour without strain softening. Similar

contrast is also seen in the K0 values that would develop in the model using the selected

inputs, ranging from 0.25 to 0.60, the importance of which in the context of such analyses

was highlighted previously. The variation in methods adopted were manifested in the results

from those adopting these techniques, with failure indicated anywhere from Y=25m to

Y=70m for Scenario B, for example.

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 20 of 55

The intricacies of FE/FD modelling and the merits of different available constitutive models
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

are beyond the primary scope of this paper. However, the comparison provided here

indicates that significant variation can occur with different approaches that were all based on

the same source data and problem statement, and points to the critical need for the

functionality and calibration of advanced constitutive models to be well understood and

carried out when applied in engineering practice (Brinkgreve and Engin 2013, Lees 2016).
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

FE/FD analyses – NorSand constitutive model

Two participants (5 and 28) made use of the NorSand constitutive model, with details

provided in Table 8. An important first point regarding these two entries is that in neither

case did the participant identify “failure” in the sense of significant deformations such that the

grid geometry of their model became unusable, a criterion that is often used to designate

onset of failure for such a model. Rather, the failures identified by the participants were

selected as the point when the observed displacement rate at the crest increased

significantly. However, in both cases, the maximum crest displacement reached less than

0.5 m, with the model still achieving equilibrium. This is inconsistent with a flow event and

warrants further investigation both with respect to these two entries, and to provide more

context on element behaviour on a CSD stress path.

Both NorSand users did so with the commercially available software FLAC 2D 8.1 (Itasca

2019a) or FLAC3D v7.0 (Itasca 2019b) - the 3D version being used with a single element

width model to simulate plane strain conditions. Although NorSand has shown promising

potential to simulate soil response on a CSD path (Jefferies and Been 2015), several

important considerations are relevant to its use to identify failure of a slope. First, and

perhaps most importantly, NorSand will only provide a reasonable simulation of the

behaviour of loose soil on a CSD stress path if the implementation used includes yielding on

the inner cap (Jefferies 1997) – i.e. where plastic yielding can occur during mean effective

stress reduction. It is therefore interesting to note that inclusion of the inner cap in the

FLAC/FLAC3D implementation of NorSand only occurred in May 2020 as an “experimental”

© The Author(s) or their Institution(s)


Page 21 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

feature, and the first of the two NorSand entries - being completed prior to this date –
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

therefore did not include the inner cap. It is also noteworthy that it appears neither

participant adopting NorSand carried out, or at least indicated, element simulations to

examine the behaviour of the model on a CSD stress path.

To provide further background on the use of NorSand in the context of the CSD stress path,

Figure 12 presents two single element triaxial simulations carried out using FLAC v8.1,
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

where the inner cap is activated in only one of the simulations. The Scenario B parameters

are adopted, with the CSD stress path commencing from initial conditions of p’ = 160 kPa

and q = 60 kPa. As can be seen in Figure 12, the model without the inner cap only

undergoes elastic swelling during unloading, with the stress ratio achieved far exceeding Mtc

- indeed, 𝝈’2 and 𝝈’3 cross into tension - before numerical instability occurs. This element

behaviour, without an inner cap, is therefore consistent with the prediction of no failure by

Entry 5 for Scenario B.

Turning to element behaviour with the inner cap, contraction is seen to commence when a

stress ratio of 0.97 is achieved. Eventually, at a stress ratio of 1.31, the stress path can no

longer be maintained in the simulation and numerical instability occurs. This more closely

aligns to experimental observations of the CSD path. However, it still suggests a relatively

high value of 𝜂 can be sustained in a saturated element above the CSL while the specimen

has begun contracting towards the CSL - something that seems inconsistent with available

experimental evidence for loose soils (Sasitharan et al. 1993, Reid and Fourie 2019, Fanni

et al. 2022).

Experimental data is useful to provide context around element behaviour up to and at

instability. The test programs carried out by Sasitharan et al. (1993) and Skopek et al.

(1994) on Ottawa sand both involved triaxial CSD tests, where deviator stress was applied

using dead weights. This is an important distinction, as the deviator stress applied with

weights is “relentless” - it is not dependent on the speed at which an actuator can apply load

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 22 of 55

as, for example, the specimen’s deformation rate begins to increase. If the specimen cannot
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

bear the deviator stress being applied, rapid failure will occur. The work of Sasitharan et al.

involved specimens saturated using standard triaxial methods, whereas that of Skopek et al.

used dry sand. The response of one test from each of these programs are contrasted in

Figure 13, as both a Cambridge plot and state diagram in the same form as Figure

12. Although no two tests from these programs had identical density and stress conditions,
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

the two presented in Figure 13 are sufficiently close that comparisons can be made.

In the test by Sasitharan et al. (1993), once a particular stress ratio is reached, essentially

instant collapse of the specimen occurs - the authors of that work describe the dramatic

effect of the falling weights at failure. Prior to this occurrence, only slight volumetric

expansion is seen, consistent with what might be expected from elastic

unloading. Alternatively, for the Skopek et al. (1994) test, at a similar stress ratio to where

the Sasitharan et al. test failed, the specimen begins to contract rapidly, with the authors

using the term “structural collapse”. The range of ηIL seen in these tests where either

collapse of the sample (falling of weights) or rapid contraction occurs are consistent with

those seen in other element studies of the CSD stress path, and which importantly are

significantly lower than the Mtc of Ottawa sand. For the purposes of the current work, and

field-scale CSD behaviour more generally, the critical aspect of these works as summarised

in Figures 12 and 13 is that it is quite unclear as to whether - in a field scale, stress-

controlled situation - in situ elements of soil would be able to sustain the CSD stress path

beyond the point at which contraction has commenced. Physical model evidence suggests

that the rate of contraction is such that drainage cannot occur fast enough, even in coarse

grained soils (Eckersley 1990). This is consistent with field-scale behaviour of the CSD

stress path on similarly sandy materials (Olson et al. 2000, France et al. 2021).

The various element and model data outlined therefore raise important questions as to

whether the behaviour of the NorSand model with inner cap - where the CSD stress path

can be successfully followed up to an η of 1.31 while contraction is occurring - is a realistic

© The Author(s) or their Institution(s)


Page 23 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

representation of what will occur for a loose, saturated soil in situ. The importance of the
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

verification of numerical models, as noted previously, is again highlighted. Clearly, further

1g or centrifuge model investigation of the CSD stress path would be useful (e.g. Ng et al.

2023). It would also likely be of value for additional benchmarking exercises of this type to

be carried out in the future, perhaps under the auspices of relevant tailings- and dam-safety

international committees.
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

CONCLUSIONS

A comparative design exercise was organised to assess the different methods that would be

applied, and results obtained, to the analysis of an idealised TSF slope within which the

phreatic surface was rising. Two scenarios were developed to span much of the range of

gradation and CPTu drainage conditions relevant to recent prominent TSF failures. The

CPTu data and engineering behaviour of the two tailings were derived from those of the

recent Feijão and Cadia TSF failure investigations.

Despite all participants having identical information, a wide range of results were obtained -

from predictions that the slope would fail immediately, without any rise in phreatic surface, to

those indicating that the slope would remain stable under the maximum phreatic surface

prediction relevant to the exercise. The submissions to the exercise and general outcomes

are summarised as follows:

 For Scenario A, in particular, widely divergent views were seen:

o Five participants used drained strengths and concluded that the slope would

not fail up to the final phreatic surface profile of Y=77m. The implicit (and in

this case incorrect) assumption was that if the CPT sounding was drained,

only drained stability analyses were required.

o Alternatively, two participants assumed mobilisation of liquefied strengths for

Scenario A under initial phreatic conditions (Y=25m) for all saturated tailings.

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 24 of 55

 Five participants adopted stress path approaches with initial stress states based on
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

K0 ranging from 0.41 to 0.80 and 𝜂IL inferred to be anywhere from 0.82 to 1.45,

resulting in a wide range of outcomes with this approach. This demonstrates both

the significant uncertainty in the estimation of in situ stresses, and the different views

on selecting representative values of 𝜂IL from an identical dataset.

 A wide range of constitutive models were adopted by different participants who


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

analysed the exercise using FE/FD techniques (strain softening, Plastic Hardening,

SaniSand and NorSand). A wide range of outcomes were seen with the different

approaches, further pointing to the importance of validation of geotechnical numerical

models, with a particular need for such work focussed on the CSD stress path.

The results of the exercise would suggest that further study of the CSD stress path by

means of 1g and centrifuge experiments would be of value. Similarly, additional

comparative design exercises of this kind, perhaps held at regular intervals under the

auspices of an appropriate international organisation, could be beneficial.

ACKNOWLEDGEMENTS

The authors express their significant appreciation to all the participants for their contributions

to this work, which would not have been possible without such efforts. This work forms part

of TAILLIQ (Tailings Liquefaction), which is an Australian Research Council (ARC) Linkage

Project supported by financial and in-kind contributions from Anglo American, BHP,

Freeport-McMoRan, Newmont, Rio Tinto, and Teck. The TAILLIQ project is being carried

out at The University of New South Wales (UNSW), The University of South Australia, The

University of Western Australia (lead university), and The University of Wollongong.

DATA AVAILABILITY STATEMENT

Data provided to participants of the program at the initiation of this study are available in the

OSF repository at https://osf.io/p4v37/?view_only=ec3aa0115e8b4feba0e4de6ca5739a3c.

© The Author(s) or their Institution(s)


Page 25 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

Additional data generated or analysis during the study after receiving participant entries are
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

available from the corresponding author upon reasonable request.

COMPETING INTERESTS STATEMENT

Competing interests: The authors declare there are no competing interests.

REFERENCES
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Anderson SA, Riemer MF. 1995. Collapse of Saturated Soil Due to Reduction in
Confinement. Journal of geotechnical engineering 121: 216–220.
https://doi.org/10.1061/(ASCE)0733-9410(1995)121:2(216)

Andrade JE. 2009. A predictive framework for liquefaction instability. Géotechnique 59: 673–
682. http://dx.doi.org/10.1680/geot.7.00087
Arroyo M, Gens A. 2021. Computational analyses of Dam I at the Corrego de Feijao mine in
Brumadinho - Final Report.
Been K. 2016. Characterizing mine tailings for geotechnical design. Australian
Geomechanics Journal 51(4).
Been K, Jefferies MG. 1985. A state parameter for sands. Géotechnique 35: 99–112.
https://doi.org/10.1680/geot.1985.35.2.99
Been K, Jefferies MG, Crooks JHA, Rothenburg L. 1987. The cone penetration test in sands:
part II, general inference of state. Géotechnique 37: 285–299.
https://doi.org/10.1680/geot.1987.37.3.285
Blight GE. 2010. Geotechnical Engineering for Mine Waste Storage Facilities. CRC Press.
Bolzon, G., Sterpi, D., Mazza, G, Frigerio, A. 2021. Numerical analysis of dams –
Proceedings of the 15th ICOLD International Benchmark Worksop. Springer.
Brand EW. 1985. Some thoughts on rainfall induced slope failures. Proceedings 10th
International Conference on Soil Mechanics and Foundation Engineering. p. 373–376.
Brinkgreve RBJ, Engin E. 2013. Validation of geotechnical finite element analysis.
Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical
Engineering. p. 677–682.
Casagrande A. 1965. Roll of the “Calculated Risk” in Earthwork and Foundation
Engineering. Proceedings, American Society of Civil Engineers 91: 1–40.
Chu J, Leong WK, Loke WL, Wanatowski D. 2012. Instability of Loose Sand under Drained
Conditions. Journal of Geotechnical and Geoenvironmental Engineering 138: 207–216.
https://doi.org/10.1061/(ASCE)GT.1943-5606.0000574
Cuomo S, Ghasemi P, Martinelli M, Calvello M. 2019. Simulation of Liquefaction and
Retrogressive Slope Failure in Loose Coarse-Grained Material. International Journal of
Geomechanics 19: 04019116. https://doi.org/10.1061/(ASCE)GM.1943-5622.0001500

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 26 of 55

Davies MP, McRoberts EC, Martin TE. 2002. Static liquefaction of tailings - fundamentals
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

and case histories. Tailings Dam 2002: Proceedings of the Joint ASDSO/USSD Specialty
Conference. p. 233–255.
Eckersley D. 1990. Instrumented laboratory flowslides. Géotechnique 40: 489–502.
https://doi.org/10.1680/geot.1990.40.3.489
Fanni R, Reid D, Fourie AB. 2022. Effect of principal stress direction on the instability of
sand under the constant shear drained stress path. Geotechnique Ahead of print.
https://doi.org/10.1680/jgeot.22.00062
France JW, Alvia I, Williams JL, Miller A, Higinbotham S. 2021. Investigation of failures of
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Edenville and Sandford Dams - Interim Report.


Ghafghazi M, Shuttle D. 2008. Interpretation of sand state from cone penetration resistance.
Géotechnique 58: 623–634. https://doi.org/10.1680/geot.2008.58.8.623
Hicks MA, Boughrarou R. 1998. Finite element analysis of the Nerlerk underwater berm
failures. Géotechnique 48: 169–185. https://doi.org/10.1680/geot.1998.48.2.169
ICMM. 2020. Global Industry Standard on Tailings Management.
Idriss IM, Boulanger RW. 2008. Soil Liquefaction during Earthquakes. EERI MNO-12
Itasca. 2019a. Fast Lagrangian Analysis of Continua (FLAC), v8.1 User’s Guide.
Itasca. 2019b. Fast Lagrangian Analysis of Continua (FLAC) 3D v7.0, User’s Guide.
Jefferies M. 1997. Plastic work and isotropic softening in unloading. Géotechnique 47: 1037–
1042. https://doi.org/10.1680/geot.1997.47.5.1037
Jefferies M. 2018. Critical state soil mechanics: 125 years of history to current use. Jenning
Lecture 2018.
Jefferies M. 2022. Improving governance will not be sufficient to avoid dam failures.
Proceedings of the Institution of Civil Engineers - Geotechnical Engineering 1–46.
https://doi.org/10.1680/jgeen.21.00105
Jefferies M, Been K. 2015. Soil Liquefaction: A Critical State Approach, Second Edition.
CRC Press.
Jefferies M, Morgenstern NR, Van Zyl DV, Wates J. 2019. Report on NTSF Embankment
Failure, Cadia Valley Operations, for Ashurst Australia.
Jefferies MG. 1993. Nor-Sand: a simple critical state model for sand. Géotechnique 43: 91–
103. https://doi.org/10.1680/geot.1993.43.1.91
Kulhawy FH, Mayne PW. 1990. Manual on estimating soil properties for foundation design.
EPRI EL-6800
Ladd CC, Foott R. 1974. New Design Procedure for Stability of Soft Clays. Journal of the
Geotechnical Engineering Division 100: 763–786. https://doi.org/10.1061/AJGEB6.0000066
Lade PV. 1992. Static Instability and Liquefaction of Loose Fine Sandy Slopes. Journal of
Geotechnical Engineering 118: 51–71. https://doi.org/10.1061/(ASCE)0733-
9410(1992)118:1(51)
Lees A. 2016. Geotechnical Finite Element Analysis. ICE.

© The Author(s) or their Institution(s)


Page 27 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

Meneses-Loja Jorge, Ishihara Kenji, Towhata Ikuo. 2000. Flow Failure of Saturated Sand
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

under Simultaneous Monotonic and Cyclic Stresses. Journal of Geotechnical and


Geoenvironmental Engineering 126: 131–138. https://doi.org/10.1061/(ASCE)1090-
0241(2000)126:2(131)
Monkul MM, Yamamuro JA, Lade PV. 2011. Failure, instability, and the second work
increment in loose silty sand. Canadian Geotechnical Journal 48: 943–955.
https://doi.org/10.1139/t11-013
Morgenstern N. 2018. Geotechnical Risk, Regulation, and Public Policy (Victor de Mello
Lecture). Soils and Rocks 41: 107–129.
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Morgenstern NR, Vick SG, Viotti CB, Watts BD. 2016. Fundão Tailings Dam Review Panel:
Report on the immediate causes of the failure of the Fundão Dam.
Morrison KF. 2022. Tailings Management Handbook: A LifeCycle Approach. Society for
Mining, Metallurgy & Exploration. SME.
Ng CWW, Crous PA, Jacobsz SW. 2023. Centrifuge and Numerical Modeling of Liquefied
Flow and Nonliquefied Slide Failures of Tailings Dams. Journal of Geotechnical and
Geoenvironmental Engineering 149: 04023075. https://doi.org/10.1061/JGGEFK.GTENG-
10800
Olson SM, Stark TD. 2002. Liquefied strength ratio from liquefaction flow failure case
histories. Canadian Geotechnical Journal 39: 629–647. https://doi.org/10.1139/t02-001
Olson SM, Stark TD. 2003. Yield Strength Ratio and Liquefaction Analysis of Slopes and
Embankments. Journal of Geotechnical and Geoenvironmental Engineering 129: 727–737.
https://doi.org/10.1061/(ASCE)1090-0241(2003)129:8(727)
Olson SM, Stark TD, H. WW, Castro G. 2000. 1907 Static Liquefaction Flow Failure of the
North Dike of Wachusett Dam. Journal of Geotechnical and Geoenvironmental Engineering
126: 1184–1193. https://doi.org/10.1061/(ASCE)1090-0241(2000)126:12(1184)
Orr TLL, Bond AJ, Scarpelli G. 2011. Findings from the 2nd Set of Eurocode 7 Design
Examples. p. 537–547. ISGSR, Proceedings of the 3rd International Symposium on
Geotechnical Safety and Risk. Munich, Germany.
Plewes HD, Davies MP, Jefferies MG. 1992. CPT based screening procedure for evaluation
liquefaction susceptibility. 45th Canadian Geotechnical Conference. p. 41–49.
di Prisco C, Imposimato S. 1997. Experimental analysis and theoretical interpretation of
triaxial load controlled loose sand specimen collapses. Mechanics of Cohesive‐frictional
Materials: An International Journal on Experiments, Modelling and Computation of Materials
and Structures 2: 93–120. https://doi.org/10.1002/(SICI)1099-1484(199704)2:2%3C93::AID-
CFM22%3E3.0.CO;2-Z
Reid D, Chiaro G, Fanni R, Fourie AB. 2022. Arguments for increasing the use of hollow
cylinder testing in the characterization of tailings. Tailings and Mine Waste 2022. Denver,
Colorado.
Reid D, Dickinson S, Mital U, Fanni R, Fourie A. 2021. On some uncertainties related to
static liquefaction triggering assessments. Proceedings of the Institution of Civil Engineers -
Geotechnical Engineering 175(2): 181-199. https://doi.org/10.1680/jgeen.21.00054

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 28 of 55

Reid D, Fourie A. 2019. A direct simple shear device for static liquefaction triggering under
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

constant shear drained loading. Géotechnique Letters 9: 142–146.


https://doi.org/10.1680/jgele.19.00011
Reid D, Fourie AB. 2014. Assessing the post-liquefaction shear strength of thickened tailings
in the design stage - a review and update. 17th International Seminar on Paste and
Thickened Tailings. p. 429–444. Vancouver, B.C.
Reid D, Smith K. 2021. Interpretation of state parameter in partially drained tailings – a case
history examination. Géotechnique Letters 11: 1–20. https://doi.org/10.1680/jgele.21.00066
Robertson PK. 2009. Interpretation of cone penetration tests — a unified approach.
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Canadian Geotechnical Journal 46: 1337–1355. https://doi.org/10.1139/T09-065


Robertson PK. 2010a. Estimating in-situ state parameter and friction angle in sandy soils
from the CPT. 2nd International Symposium on Cone Penetration Testing.
Robertson PK. 2010b. Evaluation of Flow Liquefaction and Liquefied Strength Using the
Cone Penetration Test. Journal of Geotechnical and Geoenvironmental Engineering 136:
842–853. https://doi.org/10.1061/(ASCE)GT.1943-5606.0000286
Robertson PK, Campanella RG. 1983. Interpretation of cone penetration tests. Part I: Sand.
Canadian Geotechnical Journal 20: 718–733. https://doi.org/10.1139/t83-078
Robertson PK, de Melo L, Williams DJ, Wilson GW. 2019. Report of the Expert Panel on the
technical causes of the Failure of Feijão Dam I.
Russell AR, Vo T, Ayala J, Wang Y, Reid D, Fourie AB. 2022. Cone penetration tests in
saturated and unsaturated silty tailings. Géotechnique 1–15.
https://doi.org/10.1680/jgeot.21.00261
Sadrekarimi A. 2014. Effect of the Mode of Shear on Static Liquefaction Analysis. Journal of
Geotechnical and Geoenvironmental Engineering 140(12).
https://doi.org/10.1061/(ASCE)GT.1943-5606.0001182
Sadrekarimi A. 2016. Static Liquefaction Analysis Considering Principal Stress Directions
and Anisotropy. Geotechnical and Geological Engineering 34: 1135–1154.
https://doi.org/10.1007/s10706-016-0033-7
Sasitharan S, Robertson PK, Sego DC, Morgenstern NR. 1993. Collapse behavior of sand.
Canadian Geotechnical Journal 30: 569–577. https://doi.org/10.1139/t93-049
Schweiger HF. 1998. Results from Two Geotechnical Benchmark Problems. A p. 645–654.
Application of Numerical Methods to Geotechnical Problems. Proceedings of the Fourth
European Conference on Numerical Methods in Geotechnical Engineering Numge98. Udine,
Italy October 14–16, 1998
Schweiger HF. 2006. Results from the ERTC7 benchmark exercise. 7th European
Conference on Numerical Methods in Geotechnical Engineering - Trondheim, Norway. p. 3–
8.
Shuttle D, Jefferies M. 1998. Dimensionless and unbiased CPT interpretation in sand.
International Journal for Numerical and Analytical Methods in Geomechanics 22: 351–391.
https://doi.org/10.1002/(SICI)1096-9853(199805)22:5%3C351::AID-NAG921%3E3.0.CO;2-8
Shuttle D, Jefferies M. 2016. Determining silt state from CPTu. Geotechnical Research 3:
90–118. https://doi.org/10.1680/jgere.16.00008

© The Author(s) or their Institution(s)


Page 29 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

Shuttle DA, Cunning J. 2007. Liquefaction potential of silts from CPTu. Canadian
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

Geotechnical Journal 44: 1–19. https://doi.org/10.1139/t06-086


Silvis F, de Groot MB. 1995. Flow slides in the Netherlands: experience and engineering
practice. Canadian Geotechnical Journal 32: 1086–1092. https://doi.org/10.1139/t95-107
Skopek P, Morgenstern NR, Robertson PK, Sego DC. 1994. Collapse of dry sand. Canadian
Geotechnical Journal 31: 1008–1014. https://doi.org/10.1139/t94-115
Sladen JA, D’Hollander RD, Krahn J. 1985. The liquefaction of sands, a collapse surface
approach. Canadian Geotechnical Journal 22: 564–578. https://doi.org/10.1139/t85-076
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Smith ES. 1969. Tailings disposal and liquefaction. Transactions, Society of Mining
Engineers, AIME 244: 179–187.
Take WA, Bolton MD, Wong PCP, Yeung FJ. 2004. Evaluation of landslide triggering
mechanisms in model fill slopes. Landslides 1: 173–184. https://doi.org/10.1007/s10346-
004-0025-1
Wanatowski D, Chu J, Loke WL. 2010. Drained instability of sand in plane strain. Canadian
Geotechnical Journal 47: 400–412. https://doi.org/10.1139/T09-111
Yamamuro JA, Lade PV. 1997. Static liquefaction of very loose sands. Canadian
Geotechnical Journal 34: 905–917. https://doi.org/10.1139/t97-057
Zhang W, Askarinejad A. 2020. Centrifuge modelling of static liquefaction in submarine
slopes: scaling law dilemma. Canadian Geotechnical Journal 58: 200–209.
https://doi.org/10.1139/cgj-2019-0417

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 30 of 55

FIGURE CAPTIONS
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

Figure 1: Idealised TSF geometry, (a) CPTu locations and initial phreatic
surface, (b) final phreatic surface level for consideration by participants
Figure 2: Triaxial compression element test data produced using NorSand VBA
for Scenario A and B, CIU p’c= 100 kPa, Ψ0 = +0.05: (a) deviator stress vs. axial
strain (with inset) and (b) deviator stress vs mean effective stress
Figure 3: Peak and minimum post-peak strengths for both scenarios as a
function of Ψ0 as predicted through NorSand triaxial compression element
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

simulations outlined in Figure 2


Figure 4: Normalised CPTu data for CPTu1: (a) Scenario A, (b) Scenario B
Figure 5: State parameters based on Plewes Method with material-specific Mtc
of 1.45: (a) CPTu1, (b) CPTu2, (c) CPTu 3. Both scenarios are shown for each
CPTu. Phreatic surface shown is that relevant for initial conditions (Y=25m)
relevant to the fictional CPTu probing.
Figure 6: Summary of provided CPT data for the exercise in Q(1-Bq)+1 vs Fr
format after Shuttle and Jefferies (2016, average values for saturated and
unsaturated portions of the CPT soundings show for clarity
Figure 7: LE stability analyses of the idealised slope, all using optimised
failure surface geometry, Morgenstern-Price method: (a) Back-analysed
su(yield)/σ’vc required to produce a FoS of unity for initial phreatic surface
conditions, (b) Back-analysed su(yield)/σ’vc required to produce a FoS for unity
for final phreatic surface conditions, and (c) Stability analysis assuming
effective friction strengths for the saturated tailings under final phreatic
surface conditions.
Figure 8: Predicted failure phreatic surface heights histogram: (a) Scenario A,
(b) Scenario B
Figure 9: Comparison of adopted strength ratios (and ranges, where relevant)
with predicted failure phreatic surface elevation (at X=0m), including reference
values produced by authors adopting circular and optimised failure surfaces
where strength ratio is applied to all tailings below phreatic surface and
friction angle of 36 degrees above phreatic surface
Figure 10: Schematic stress path representation of the implications of different
K0 and ηIL assumptions made by participants. All idealised stress paths
assume initial vertical effective stress of 200 kPa and axisymmetric conditions,
for illustration purposes.
Figure 11: Element simulation stress-strain behaviour of three participants
adopting constitutive models other than NorSand for FE/FD analysis
Figure 12: NorSand element simulation of axisymmetric CSD test, adopting
Scenario B parameters: (a) Cambridge space, (b) state diagram.

© The Author(s) or their Institution(s)


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 31 of 55

© The Author(s) or their Institution(s)


Skopek et al. (1994): (a) Cambridge plot, (b) State diagram
Canadian Geotechnical Journal (Author Accepted Manuscript)

Figure 13: CSD element tests carried out by Sasitharan et al. (1993) and
Canadian Geotechnical Journal (Author Accepted Manuscript) Page 32 of 55

LIST OF NOTATIONS
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

𝜎2 ― 𝜎3
b intermediate principal stress ratio ( = 𝜎1 ― 𝜎3)

Bq dimensionless excess pore pressure ( = (𝑢2 ― 𝑢0)/(𝑞𝑡 ― 𝜎𝑣)

ecs Void ratio on critical state

Fr stress-normalised CPT friction ratio


𝜎′ℎ
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

K0 geostatic stress ratio ( = 𝜎′𝑣)

G elastic shear modulus

H0 Reference plastic hardening modulus at Ψ=0

Hy Change in plastic hardening modulus with Ψ

Mtc critical state friction ratio, triaxial compression conditions

N volumetric coupling parameter

p’ mean effective stress (= (𝜎′1 + 𝜎′2 + 𝜎′3)/3)

q deviator stress (= 𝜎1 ― 𝜎3)

Q normalised dimensionless tip resistance normalised by vertical stresses

Qp dimensionless tip resistance normalised by mean stresses

qt corrected cone tip resistance

u0 in situ equilibrium pore pressure

u2 CPT shoulder pore pressure

α principal stress angle


𝑞
η stress ratio ( = 𝑝′)

𝑞
ηIL instability stress ratio ( = 𝑝′ 𝑎𝑡 𝑖𝑛𝑠𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦)

v Poisson’s ratio

σh total horizontal stress

σ’h effective horizontal stress

σv total vertical stress

σ’v effective vertical stress

© The Author(s) or their Institution(s)


Page 33 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

σ’vo effective vertical overburden stress


For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

φ'cs critical state friction angle

Γ CSL intercept at p’ = 1 kPa

λe slope of partial-logarithmic CSL, base e

Ψ state parameter

χ State dilatancy constant


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

© The Author(s) or their Institution(s)


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript)

Figure 1: Idealised TSF geometry, (a) CPTu locations and initial phreatic surface, (b) final phreatic surface level for consideration by participants
Page 34 of 55
Page 35 of 55
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Canadian Geotechnical Journal (Author Accepted Manuscript)
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Figure 2: Triaxial compression element test data produced using NorSand VBA for Scenario A and B, CIU p’c= 100 kPa, Ψ0 = +0.05: (a) deviator stress vs. axial strain (with inset) and (b)
deviator stress vs mean effective stress

© The Author(s) or their Institution(s)


For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Canadian Geotechnical Journal (Author Accepted Manuscript) Page 36 of 55
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Figure 3: Peak and minimum post-peak strengths for both scenarios as a function of Ψ0 as predicted through NorSand triaxial compression element simulations outlined in Figure 2

© The Author(s) or their Institution(s)


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 37 of 55

Figure 4: Normalised CPTu data for CPTu1: (a) Scenario A, (b) Scenario B

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript)
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Canadian Geotechnical Journal (Author Accepted Manuscript) Page 38 of 55
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Figure 5: State parameters based on Plewes Method with material-specific Mtc of 1.45: (a) CPTu1, (b) CPTu2, (c) CPTu 3. Both scenarios are shown for each CPTu. Phreatic surface
shown is that relevant for initial conditions (Y=25m) relevant to the fictional CPTu probing.

© The Author(s) or their Institution(s)


Page 39 of 55
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Canadian Geotechnical Journal (Author Accepted Manuscript)
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Figure 6: Summary of provided CPT data for the exercise in Q(1-Bq)+1 vs Fr format after Shuttle and Jefferies (2016, average values for saturated and unsaturated portions of the CPT
soundings show for clarity

© The Author(s) or their Institution(s)


For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Canadian Geotechnical Journal (Author Accepted Manuscript) Page 40 of 55
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Figure 7: LE stability analyses of the idealised slope, all using optimised failure surface geometry, Morgenstern-Price method: (a) Back-analysed su(yield)/σ’vc required to produce a
FoS of unity for initial phreatic surface conditions, (b) Back-analysed su(yield)/σ’vc required to produce a FoS for unity for final phreatic surface conditions, and (c) Stability analysis
assuming effective friction strengths for the saturated tailings under final phreatic surface conditions.

© The Author(s) or their Institution(s)


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 41 of 55

Figure 8: Predicted failure phreatic surface heights histogram: (a) Scenario A, (b) Scenario B

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript)
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Canadian Geotechnical Journal (Author Accepted Manuscript) Page 42 of 55
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Figure 9: Comparison of adopted strength ratios (and ranges, where relevant) with predicted failure phreatic surface elevation (at X=0m), including reference values produced by
authors adopting circular and optimised failure surfaces where strength ratio is applied to all tailings below phreatic surface and friction angle of 36 degrees above phreatic surface

© The Author(s) or their Institution(s)


Page 43 of 55
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Canadian Geotechnical Journal (Author Accepted Manuscript)
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Figure 10: Schematic stress path representation of the implications of different K0 and ηIL assumptions made by participants. All idealised stress paths assume initial vertical effective
stress of 200 kPa and axisymmetric conditions, for illustration purposes.

© The Author(s) or their Institution(s)


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript)

Figure 11: Element simulation stress-strain behaviour of three participants adopting constitutive models other than NorSand for FE/FD analysis
Page 44 of 55
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 45 of 55

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript)

Figure 12: NorSand element simulation of axisymmetric CSD test, adopting Scenario B parameters: (a) Cambridge space, (b) state diagram.
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

diagram

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript)

Figure 13: CSD element tests carried out by Sasitharan et al. (1993) and Skopek et al. (1994): (a) Cambridge plot, (b) State
Page 46 of 55
Page 47 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

Table 1: Critical state and NorSand parameters provided to participants

Value
Parameter
Scenario A Scenario B
Γ 0.920 0.740
λe 0.021 0.050
Mtc 1.45 1.45
φ'cs 36 36
N 0.30 0.25
χ 6.3 7.2
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

H 125 - 690 ψ 45 - 350 ψ


G* 50 * (p'/100)0.42 15*(p'/100)0.72
*p' input in kPa, G result in MPa
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 48 of 55

Table 2: 20th, 50th, and 80th percentile Ψ values, Plewes Method with Mtc = 1.45, λ10 based on each Scenario’s specific CSL
slope, and K0 = 0.7

CPT 1 CPT 2 CPT 3


Scenario A B A B A B
20th -0.016 +0.085 -0.008 +0.064 -0.036 +0.064
Percentile 50th +0.014 +0.121 +0.004 +0.121 +0.000 +0.098
80th +0.050 +0.153 +0.032 +0.153 +0.029 +0.147
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

© The Author(s) or their Institution(s)


Page 49 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

Table 3: Program and participant summary


Team description Scenario A Scenario B
Failure Failure
No. of Academics / phreatic phreatic
Entry Analysis
team consultants / surface surface Analysis type
type
members operators identified (at identified (at
X=0) X=0)
1 2 Consultants No entry 46 Stress path
2 1 Consultants 41 LE 75 LE
3 2 Consultants No entry 43 LE
4 2 Consultants 67 LE 39 LE
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

5 1 Consultants 69 FE/FD Does not fail FE/FD


6 1 Consultants Does not fail LE 47 Stress path
7 1 Consultants Does not fail LE No entry
8 1 Consultants No entry 47 LE
9 3 Consultants 60 FE/FD 70 FE/FD
10 2 Consultants Does not fail LE No entry
11 1 Consultants 60 Stress path 40 LE
12 3 Consultants 48 LE 70 LE
13 2 Consultants 55 LE Does not fail LE
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

14 2 Consultants 50 Stress path 40 Stress path


15 1 Consultants 63 LE* 60 LE
16** 1 Academic Synthesis of Scenario 1 and 2, Y=55m, LE
17 1 Consultants 25 LE 39 LE
18 1 Consultants No entry 25 LE
19 2 Academic Does not fail LE* 50 FE/FD
20 1 Consultants 25 LE 25 LE
21** 4 Consultants Various analyses under different conditions without single entry
22 2 Consultants 36 LE No entry
23 1 Consultants Does not fail LE 50 LE
24 3 Operators No entry 25 Stress path
25 3 Consultants 75 LE 26 LE
26 1 Consultants 25 FE/FD 33 FE/FD
27 1 Consultants 31 FE/FD 25 FE/FD
28 1 Consultants Does not fail FE/FD 65 FE/FD
*Used SSR rather than LE, but groups with LE entries for simplicity
**Not included in subsequent analyses owing to varied techniques and/or synthesis of the two scenarios

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 50 of 55

Table 4: Comparison of methods used by participants to estimate state parameter

Methods Assumed Ψ Ψ value or range


Entry K0 (when λ source Mtc source selection
Scenario A Scenario B required) process A B

Direct
1 - Robertson (2010a) n/a n/a n/a from CPT - Varied with depth
data
50 - 80th
2 Cavity expansion widget (Shuttle and Jefferies 2016) 0.70 CSL provided Mtc provided percentile Unknown Unknown
("by eye")
80th
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

4 Plewes et al. (1992) - 0.70 λ10 = Fr /10 Mtc provided +0.04 Did not use Ψ
percentile
50th Various ranges applied in sensitivity
5 Robertson (2010a) n/a λ10 = Fr /10 n/a
percentile analyses
50th Examined only to screen for
6 Robertson (2010a) n/a n/a n/a
percentile contractive behaviour
Plewes et al. (1992) and 50th
8 - 0.80 λ10 = Fr /10 Mtc provided - +0.04
Robertson (2010a) percentile
Scenario
A - 0.48 80th
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

9 Plewes et al. (1992) λ10 = Fr /10 Mtc provided Unknown Unknown


Scenario percentile
B - 0.54
Examined only
Plewes et al. (1992)
50th to screen for
11 and Robertson - 0.70 λ10 = Fr /10 Mtc provided -
percentile contractive
(2010a)
behaviour
+0.06 (0-25m Dilative (0-25m
85th depth) depth)
12 Cavity expansion widget (Shuttle and Jefferies 2016) 0.70 CSL provided Mtc provided
percentile +0.11 (>25m +0.05 (>25m
depth) depth)
50th
13 Robertson (2010a) n/a n/a n/a Unknown Unknown
percentile
Direct
14 Robertson (2010a) n/a n/a n/a from CPT Varied with depth
data
Examined only to
Cavity expansion widget 80th screen for
18 - 0.45 CSL provided Mtc provided -
(Shuttle and Jefferies 2016) percentile contractive
behaviour
50th
19 Plewes et al. (1992) - 0.50 λ10 = Fr /10 Mtc provided Unknown -
percentile
Examined only
50th
to screen for
20 Plewes et al. (1992) - 0.60 λ10 = Fr /10 Mtc provided percetile -
contractive
("by eye")
behaviour
50th
22 Plewes et al. (1992) - 0.80 λ10 = Fr /10 Mtc provided Unknown -
percentile
80th Examined only to screen for
23 - Plewes et al. (1992) 0.70 λ10 = Fr /10 Mtc provided
percentile contractive behaviour
Examined only to
50th
screen for
24 Plewes et al. (1992) 0.67 CSL provided Mtc provided percetile -
contractive
("by eye")
- behaviour
Examined only
to screen for
25 Robertson (2010a) - n/a n/a n/a n/a -
contractive
behaviour
λ10 = Fr /10, 50th
26 Plewes et al. (1992) 0.6 - 0.7 and CSL Mtc provided percetile Unknown Unknown
provided ("by eye")
λ10 = Fr /10, 50th
27 Plewes et al. (1992) 0.6 - 0.7 and CSL Mtc provided percetile +0.07 +0.13
provided ("by eye")
Direct input of 50th percentile
50th
28 Been et al. (1987) 0.50 CSL provided Mtc provided value in incremental depth "bins"
percentile
into FLAC

© The Author(s) or their Institution(s)


Page 51 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

Table 5: Summary of LE methods adopted by participants


Scenario A Scenario B
Saturated tailings strength input and Source for
Failure Saturated tailings strength input and methods Failure drained
methods
Entr phreatic phreatic strength,
y surface surface above
Method to su (yield) / σ'vc or Method to su (yield) / σ'vc or
identified Input type identified Input type phreatic
estimate su (LIQ) / σ'vc estimate su (LIQ) / σ'vc
(at X=0) (at X=0) surface

NorSand
NorSand
su (yield) / element
2 41 su (yield) / σ'vc element 0.17 - 0.25 75 0.24 Mtc provided
σ'vc simulation
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

simulations
s
su (yield) /
3 No entry 43 Nkt - 14 0.21 Mtc provided
σ'vc
Ψ-
YSR/USR 0.35 (TC) 0.25 (TC)
su (yield) /
4 67 su (yield) / σ'vc trends 0.26 (DSS) 39 Nkt - 15 0.19 (DSS) Mtc provided
σ'vc
from 0.18 (TE) 0.13 (TE)
literature
Does not Drained Mtc
6 n/a Not LE technique Mtc provided
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

fail strength provided


Robertson
Does not Drained
7 and Cabal n/a No entry Mtc provided
fail strength
(2015)
16 degrees
adopted
NorSand
based on
su (yield) / element
8 No entry 47 0.27 strain
σ'vc simulation
compatibility
s
consideration
s
Inferred
from CPTu
results, Inferred from
synthesis CPTu results,
of: synthesis of:
Robertson Robertson &
& Campanella
Does not Drained
10 Campanell n/a No entry (1983),
fail strength
a (1983), Kulhawy and
Kulhawy Mayne (1990)
and Mayne and Jefferies
(1990) and and Been
Jefferies (2015)
and Been
(2015)
0.22 at depth,
then linearly
su (yield) / transition to
11 Not LE technique 40 Nkt - 15 Mtc provided
σ'vc drained strength
at shallower
depths
NorSand
su (yield) /
12 48 su (yield) / σ'vc element 0.19 70 Nkt - 16 0.20 Mtc provided
σ'vc
simulations
NorSand
NorSand
Does not su (yield) / element
13 55 su (yield) / σ'vc element 0.27 0.48 Mtc provided
fail σ'vc simulation
simulations
s
NorSand
NorSand
Function of su (yield) / element Function of
15* 63 su (yield) / σ'vc element 60 Mtc provided
depth σ'vc simulation depth
simulations
s
Synthesis Estimated
of baesd on
33 degrees,
Robertson su (yield) / Qtn of 3-4
17 25 su (LIQ) / σ'vc 0.07 39 0.22 based on
(2010) and σ'vc (Robertson
Mayne (2006)
Sadrekarim and Cabal
i (2014) 2015)
Friction
18 No entry 25 su (LIQ) / σ'vc sleeve 0.12 Mtc provided
resistance

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 52 of 55

Does not Drained Mtc


19* n/a Not LE technique Mtc provided
fail strength provided
Friction Friction
20 25 su (LIQ) / σ'vc sleeve 0.15 25 su (LIQ) / σ'vc sleeve 0.15 Mtc provided
resistance resistance
Olson and
Stark
(2003) and
22 36 su (yield) / σ'vc 0.13** No entry Mtc provided
NorSand
element
simulations
Does not Drained Mtc su (yield) /
23 n/a 50 Nkt - 14 0.25 Mtc provided
fail strength provided σ'vc
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

Olson and
su (LIQ) / σ'vc su (yield) /
25 75 Stark 0.10** 26 Nkt - 14 0.15 - 0.26 Mtc provided
*** σ'vc
(2002)
*Used SSR rather than LE

**Reduced owing to inferred brittle nature of tailings

***Only applied to discrete zones of saturated tailings


Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

© The Author(s) or their Institution(s)


Page 53 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

Table 6: Summary of stress path methods adopted by participants


For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

Method to estimate su (LIQ) / σ'vc


Predicted failure Initial K0 value Initial K0
Entry Method Trigger point Triggering criteria Stress path assumed in zones where triggering
adopted rationale
identified
Scenario A Scenario B
Once one zone had
triggered, switch all
1 n/a 46 Spreadsheet 0.80 Assumed At CSL, ηIL = 1.45 CSD Robertson (2010b)
contractive zones to
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

liquefied strength

Once one zone had


Synthesis of Olson and Stark
triggered, switch all
6 n/a 47 Spreadsheet 0.41 Jaky Equation ηIL = 1.2 CSD (2002), Idriss and Boulanger
contractive zones to
(2008), Robertson (2010b)
liquefied strength

Assumed, Once one zone had


Jefferies and triggered, switch that
11 60 n/a Elastic FE/FD model 0.70 ηIL = 0.82 CSD Robertson (2010b)
Been (2015) zone to liquefied
recommendations strength

Move to LE stability
analyses with liquefied
Llano-Serna et al.
14 50 40 Spreadsheet 0.56 At CSL, ηIL = 1.45 strengths once 50% of Perpendicular to Mtc slope Robertson (2010b)
(2017)
elements along CPT
reach triggering criterion

Move to LE stability
Not applicable, as
Output of analyses with liquefied
elements already at
assumed su / σ'vc = 0.28 to strengths based on
24 n/a 25 Elastic FE/FD model 0.50 selected trigger under Robertson (2010b)
Poisson's ratio of 0.30 qualitative assessment of
initial phreatic surface
0.33 below-phreatic surface
condition
element stress conditions

© The Author(s) or their Institution(s)


Canadian Geotechnical Journal (Author Accepted Manuscript) Page 54 of 55

Table 7: Details regarding other constitutive models applied by participants other than NorSand
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

Predicted failure Initial K0 value Initial K0


Entry Constitutive model adopted Failure criteria
Scenario A Scenario B adopted rationale
Output of
9 60 70 Plastic Hardening Model 0.48 - 0.54
model
Assumed
19 n/a 50 Scenario 2 - Strain softening 0.43 Poisson's ratio
of 0.3 Lack of
convergence
26 25 33 SaniSand 0.60 Assumed
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

Assumed
27 31 25 Strain Softening 0.25 Poisson's ratio
of 0.2

© The Author(s) or their Institution(s)


Page 55 of 55 Canadian Geotechnical Journal (Author Accepted Manuscript)

Table 8: Details of methods used by participants that adopted NorSand constitutive model
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.

Predicted failure Software code K0 selection


Inner cap yielding in model
Entry used, and date Rationale for Failure criteria
Scenario A Scenario B implementation used? Initial value
released selection
FLAC 8.10.486, Lack of convergence
5 69 Does not fail No Unknown
13 February 2020 (where relevant)
Increased displacement
rate at slope crest.
FLAC3D 7.00.119,
28 Does not fail 65 Yes 0.50 Assumed However, maximum crest
13 May 2020
Can. Geotech. J. Downloaded from cdnsciencepub.com by University of Western Australia on 11/21/23

displacement of 0.3m at
"failure" (Scenario B)

© The Author(s) or their Institution(s)

You might also like