ArticlePDF Available

Some considerations in the stability analysis of upstream tailings dams


Abstract and Figures

Upstream constructed tailings dams represent a significant challenge to the geotechnical engineer in terms of analysis of their stability, in large part because the shear strength of the loose sands and fine grained or "slimes" components of such structures is open to considerable uncertainty. In particular, it is critical that the behavior of the tailings under shear (contractant or dilatant, failure under drained or undrained conditions) be understood, and that this understanding be incorporated into stability analyses. The continued development of critical state soil mechanics has served to focus attention on this issue, but this focus has generally been directed towards problems involving seismic liquefaction. However, contractant versus dilatant behavior in shear is equally important in terms of static stability, particularly for upstream tailings dams which so often are constructed of primarily contractant, potentially liquefiable materials. This paper presents a review of drained versus undrained methods of static stability analysis of upstream tailings dams, and how these relate to dilatant versus contractant behavior. The authors support the view that for upstream dams constructed of contractant, potentially liquefiable tailings, both undrained strength analysis and steady state strength analysis should be considered, and that effective stress analysis for such structures can be fundamentally incorrect and unsafe.
Content may be subject to copyright.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 1
Some considerations in the stability analysis of upstream tailings dams
T.E. Martin
AGRA Earth & Environmental Limited, Burnaby, B.C. Canada
E.C. McRoberts
AGRA Earth & Environmental Limited, Edmonton, Alberta, Canada
ABSTRACT: Upstream constructed tailings dams represent a significant challenge to the
geotechnical engineer in terms of analysis of their stability, in large part because the shear strength
of the loose sands and fine grained or “slimes” components of such structures is open to
considerable uncertainty. In particular, it is critical that the behavior of the tailings under shear
(contractant or dilatant, failure under drained or undrained conditions) be understood, and that this
understanding be incorporated into stability analyses. The continued development of critical state
soil mechanics has served to focus attention on this issue, but this focus has generally been
directed towards problems involving seismic liquefaction. However, contractant versus dilatant
behavior in shear is equally important in terms of static stability, particularly for upstream tailings
dams which so often are constructed of primarily contractant, potentially liquefiable materials.
This paper presents a review of drained versus undrained methods of static stability analysis of
upstream tailings dams, and how these relate to dilatant versus contractant behavior. The authors
support the view that for upstream dams constructed of contractant, potentially liquefiable tailings,
both undrained strength analysis and steady state strength analysis should be considered, and that
effective stress analysis for such structures can be fundamentally incorrect and unsafe.
The upstream method has been employed for many tailings dams. Two idealized typical sections
of upstream tailings dams are shown in Figure 1. The past and continued attraction of the method
is obvious, as it represents the most economical of all methods of tailings dam construction. Mill
tailings are hydraulically separated - either by cycloning or spigotting on beaches or in cells - with
the coarser sand sized fraction used to build a retention shell or dam and the fines collecting in a
pond. In some cases when beaches are short, sands are placed subaqueously and as a result are
very loose. As the dam raises the more stable subaerial beaches [which are often strengthened
by desiccation] step out over loose subaqueous sands or fines (slimes.) As dams are usually
raised against some sort of regional slope, or in a valley, water pools against the starter dam. This
combination of circumstances usually occurs where the dam is highest. Thus as the dam is raised
loose sands and weak slimes in a wide range of relative proportion and degree of intermixing are
trapped in the downstream section, often with disastrous results.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 2
Figure 1. Idealized sections of upstream tailings dams.
Upstream tailings dams are unforgiving structures, and any one or combinations of improper
design, construction and operation have resulted in a number of well-known, catastrophic failures,
that have in some instances caused loss of life, such as the Stava failure in Italy (Berti et al., 1988,
Chandler and Tosatti, 1995). The United States Committee on Large Dams (USCOLD, 1994)
published a review of tailings dam failure records available to them in 1994. This review found
that upstream-constructed tailings dams have recorded the largest share of documented failures.
The authors believe that many of these failures are due to the entrapment of loose sands and/or
weak slimes in the downstream section. In many, and including recent designs reviewed by the
authors, this factor is overlooked. This, as shall be described, is due to fundamental errors in the
understanding of the operative strength of loose sands and weak slimes.
The susceptibility of upstream tailings dams to liquefaction and flow failures under seismic
loading conditions is well known, and a number of case histories of such failures from the 1965
Chilean earthquake (Dobry and Alvarez, 1967), the 1978 Isu-Ohshima earthquake (Marcuson et
al., 1979), and the 1985 Chilean earthquake (Castro and Troncoso, 1988) are well-documented. It
is well understood that loose sands or weak slimes are contractant when sheared, and are
therefore susceptible to this seismic triggering. What often appears to be not so well understood,
and is the focus of this paper, is that undrained shear under static loading conditions can have the
same consequences. As discussed in detail by McRoberts and Sladen (1992) there is little
practical difference in the magnitude of the shear strength induced by seismic or undrained static
The stability of tailings impoundments increases considerably with time once their operational
life ends. This is particularly true for upstream tailings dams, and is due to the following factors:
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 3
1. Surface water is usually absent, particularly for impoundments regraded to shed runoff,
allowing levels of saturation within the outer shell (and possibly the slimes) to gradually reduce.
2. Excess pore pressures induced by the raising of the impoundment will gradually dissipate,
resulting in an increase in strength in the tailings slimes.
3. Capping of impoundments for closure reduces infiltration and allows for further reduction in
saturation levels, particularly in dry climates.
4. Aging effects related to cementation and oxidation processes of tailings in the unsaturated
zone, which may increase the liquefaction resistance of tailings by as much as 250% over 30
years (Troncoso, 1988, 1990).
5. Aging effects related to particle rearrangement resulting in macro-interlocking of particles
and micro-interlocking of surface roughness (Joshi et al., 1995), a particularly significant
mechanism given the angularity of tailings particles.
There is nothing fundamentally wrong with upstream tailings dams provided that key principles are
adhered to in the design, construction, and operation of such dams. Upstream dams were
originally designed in an empirical manner by mine operators, without the insights derived from
geotechnical principles (Vick, 1992, Casagrande and MacIver, 1970). The vast majority of
upstream tailings dams have performed satisfactorily. Based on experience, both successful and
unsuccessful, this empirical design approach identified the importance of several key fundamentals
for upstream tailings dams (Lenhart, 1950, Vick, 1992):
1. spigotting of a wide, sand (drained) tailings beach from the embankment crest;
2. avoiding situations whereby the dam slope is underlain by fine tailings (slimes) deposited
within the water pond;
3. prevention of seepage emerging on the dam face; and
4. having a well-drained foundation.
The dam section shown in Figure 1(a) satisfies the criteria above, while the section shown in
Figure 1(b) clearly does not.
These experienced-based principles remain valid, but do not directly address the criticality of
operating and monitoring practices in maintenance of the safety of upstream dams. Nor do they
directly address the issue of characterization of shear strengths in terms of drained or undrained
behavior under shear. The authors therefore expanded the checklist above to the following eight
fundamental rules for design, construction and operation of upstream tailings dams:
1. A sufficiently wide beach, relative to the ultimate height of the dam, must be maintained at all
times, to achieve segregation of the coarser tailings sizes and to form a relatively strong, wide,
drained (unsaturated), and/or dilatant (non-contractant during shear) outer shell. The dam slope
must not be underlain by tailings slimes, unless the designer has satisfied Rule 4 below. The shell
must be of sufficient width to retain the “bursting pressures” [see Casagrande and MacIvor] of
the upstream contractant beach sands or slimes if they liquefy.
2. The rate of raising of the dam must be sufficiently slow such that there is a sufficient degree
of dissipation of excess pore pressures in the outer shell and in the slimes, and such that excess
pore pressure buildup does not occur in foundation materials.
3. There must be sufficient underdrainage (drainage blanket, finger drains) and/or a pervious
foundation to maintain the sand shell in a relatively drained condition, and to prevent seepage from
issuing from the face of the tailings dam.
4. Design analyses must include both undrained strength analysis (USA) and effective stress
analysis (ESA), with design controlled by the analysis type giving the lowest factor of safety. A
wide range of factors including material type, degree of consolidation and stress path must be
assessed in assigning the appropriate USA.
5. A high degree of regular performance monitoring, reviews, and ongoing involvement by the
designer is essential to check that design intent is being satisfied, to confirm design assumptions,
and to identify any design changes that may be required.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 4
6. Conventional upstream dams cannot be considered for areas of moderate to high seismicity.
Improved upstream construction, involving a combination of compaction of the outer shell and
good internal drainage, can be used in such areas.
7. The design must be consistent in terms of design requirements (e.g. minimum beach width)
versus operational requirements (e.g. pond size required for clarification, storm storage and
freeboard). The geotechnical design of upstream tailings dams cannot be carried out in ignorance
of operating constraints.
8. Seepage conditions within the dam must be well-defined, requiring a good understanding of
pore pressure profiles and hydraulic gradients. The distinction between pore pressure measured
at a given point, and saturation level, must be well understood and correctly applied in stability
analyses, especially in instances where there is strong downward drainage.
The following sections of this paper are focused primarily on illustration and support of Rule 4
above. Rule 4 merits particular attention for the following reasons:
1. It becomes critical when Rule No. 1 is violated, as is frequently the case for upstream
tailings dams;
2. A great many upstream tailings have been, and continue to be, designed and/or evaluated
based on limit equilibrium slope stability analyses assuming only ESA parameters;
3. ESA analysis for upstream dams in many instances is based on fundamentally incorrect
assumptions regarding pore pressures prior to and during shear failure, and can result in large
overestimates of the factor of safety; and
4. USA analysis should be applied for staged construction, and upstream tailings represent a
classic case of staged-construction.
Tailings deposited in upstream impoundments, and particularly tailings slimes, are generally
loose to very loose, contractant during shear, and strain-softening (brittle). Exceptions occur
where the outer shell of an upstream dam is compacted (e.g. Martin and Tissington, 1996), or
where desiccation in arid climates leads to overconsolidated and unsaturated conditions within the
tailings beach (Blight, 1988). Both of these exceptions are cases where the tailings are dilatant
during shear (and/or unsaturated), and therefore the undrained strength is higher than the drained
strength (negative pore pressures generated during undrained shear). For contractant tailings, the
opposite is true, and the question then becomes whether shear will occur under drained (no shear-
induced increase in pore pressure) or undrained (shearing induces increased pore pressure)
3.1 Definitions of strength characterization
ESA, following Ladd (1991) and Carrier (1991) is defined as limit equilibrium analysis that
assumes effective stresses during shear are unchanged from those that existed immediately prior
to the onset of shear. That is to say, measured insitu pore pressures describe the conditions at
failure. In other words, ESA explicitly assumes that shear occurs slowly enough, and/or the
material being sheared is sufficiently free-draining, that there are no shear-induced pore pressures.
An ESA method of analysis is either correct if one is sure that no positive pore pressures are
generated on shearing, or is conservative if it is known that shearing is dilatant. On the other hand
if one did an ESA analysis using a method to predict the pore pressure response during shearing
[i.e., at the moment of failure] then a correct answer would be obtained. What can be
fundamentally wrong about ESA analysis is using existing as measured pore pressures in a dam to
represent the conditions at failure. This simple fact is unfortunately not understood by many
geotechnical engineers.
USA is an analytically economical way of accounting for the pore pressures generated by
undrained shearing. USA is defined in this paper as limit equilibrium analysis that assumes shear
occurs under undrained conditions. This type of analysis therefore accounts for positive shear-
induced pore pressures during shear of contractant materials, and negative pore pressures in the
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 5
case of dilatant materials. A convenient way to describe the USA strength is the undrained
strength ratio (cu/p’). The effective stress p’ denotes the operative effective stress [or
consolidation stress] that exists in a soil element at the moment that failure begins.
3.2 Behavior of non-dilatant or contractant material during shear
Traditional soil mechanics has and continues to develop strength models for two basic classes of
materials, clays and sands. Many typical mill tailings streams produce relatively unique artificial
materials that tend to be bounded by these more traditional models. However as the guidance
offered by this rich literature database often seems to be ignored by tailings dam designers it is
useful to briefly review some fundamental aspects of it.
The term USA-D is introduced, (D designates ductile) to denote in a simple way the response
of normally consolidated clays during shearing. During shear excess pore pressures are generated
but with straining the available strength does not fall off or reduce. Key references are for
example the work of Ladd (1991) and Wroth (1984). This literature makes several key points:
1. The USA-D strength of normally consolidated clays as a first estimate is in the range of
cu/p’ = 0.20 to 0.25.
2. The USA-D strength is dependent on stress path. For example see Mesri (1989) who
indicates that for triaxial extension (TE) the c
u/p’ can be from 50% to 70% of the triaxial
compression (TC) strength and the strength in direct simple shear (DSS) is intermediate between
these limits.
3. The USA-D is dependent on the method used to measure undrained strength. Different
results are obtained by different methods (Wroth, 1984).
These issues are well known in geotechnical practice, the lesson being that the determination of
USA-D strength is not straightforward. Specification of pore pressure response at failure in an
ESA type approach is even less straightforward, as demonstrated by Carrier (1991).
The term USA-SS is used here to designate the SS or steady state of loose sands [also called
the residual strength.] Following Casagrande (1975), Poulos (1988) and as discussed more
recently by McRoberts and Sladen (1992) loose sand can exhibit an undrained strength response
at high strain, and a response that is highly brittle. The strength of clean loose sands at high strain
or at steady state can also be expressed in a normalized manner, see for example Wride et. al.
(1998). These authors follow others in normalizing the back-calculated strength of several case
records in which the residual or steady state strength can be obtained. The normalized strength
from these case records is as low as 0.01 and ranges up to about 0.20.
A recent series of USA-SS laboratory test on loose clean sands has been reported by Yoshimine (1998) for TC, TE, and DSS modes. This work following on earlier studies by others
indicates that at the relatively low strains measurable in the laboratory that - and entirely similar to
USA-D behavior for clays - the mode of shear has a strong influence on the magnitude of the
normalized cu/p’.
3.3 Collapse Surface Approach
The collapse surface approach presented by Sladen et al. (1985), as an extension of critical state
theory, provides a useful framework to illustrate the authors’ contention that USA analysis is
mandatory for upstream tailings dams where contractant materials are involved. The concept is
particularly useful in this application because it ties together undrained behavior in shear, the need
for triggering of undrained behavior, and the brittle, flowsliding nature so often associated with
tailings dam failures.
The concept of a collapse surface is illustrated in Figure 2a, which shows that for a given void
ratio (density), there is a unique condition of stress state at which collapse of the sand structure
takes place and undrained failure (liquefaction) is initiated. This unique state of stress is termed
the collapse surface, and it exists below the drained failure envelope for contractant tailings.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 6
Figure 2a illustrates that stress states lying on or above the collapse surface are highly unstable, as
liquefaction can be triggered by even minor disturbance. Stress states below the collapse surface
line are stable.
Figure 2. Collapse surface framework for comparison of USA and ESA approaches.
The collapse surface occurs in three-dimensional void ratio-shear stress-mean normal stress
space. Figure 2b shows the effect of increasing density (decreasing void ratio) on the collapse
surface, as projected onto two-dimensional shear stress - mean normal stress space. Also shown
is a drained-loading stress path. Although point 2 lies above the collapse surface that corresponds
to point 1, the fact that loading has taken placed under drained conditions, in concert with
consolidation to a lower void ratio, has resulted in a shift upwards in the collapse surface line.
Therefore, as long as undrained behavior is not triggered during the loading, a stable stress state
can exist above the original collapse surface line, but only because the position of the collapse
surface line has been shifted upwards during slow, drained loading.
Figure 2c presents stress paths for contractant and dilatant materials inherent to both USA and
ESA limit equilibrium analysis. For a contractant material, USA assumes failure at point 2, with
the mean normal effective stress at failure reduced from that at point 1 because of positive pore
pressures generated during shear. USA also indicates a post-peak reduction in shear strength to
point 3, which corresponds to the residual, or steady state strength. Stress path 1-4 represents
that which corresponds to a conventional ESA, which by definition assumes slow shearing with
complete dissipation of any shear-induced pore pressures during failure. Therefore, ESA assumes
the mean normal effective stress at failure (point 4) to be the same as that at the initiation of
shearing (point 1). It also assumes that failure occurs slowly. Figure 4c shows the shear stress at
failure (and therefore the factor of safety) for ESA to be about twice that for USA, a typical
result (Ladd, 1991) when comparing the two methods of analysis.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 7
Assuming the material to be dilatant, then for undrained shear (assumed by USA), negative
pore pressures developed by shearing would result in mean normal effective stress at failure (point
5) higher than that at the initiation of shear (point 1). Under fully drained shear assumed by ESA,
failure would occur at point 4. Therefore, USA predicts a higher shear stress at failure (and
therefore a higher factor of safety) than ESA for dilatant materials. In this case, ESA represents
the more critical (and proper) method of analysis.
The differences between the stress paths assumed in the two types of analyses for contractant
versus dilatant materials represent an important point. Advocates of ESA for upstream tailings
dams, irrespective of whether saturated zones are contractant or dilatant, ignore completely the
physical behavior of the materials under shear, surely one of the most important principles in soil
mechanics. Furthermore, there is an inherent contradiction in accounting for contractant response
when analyzing the seismic safety of upstream tailings dams, but failing to do so for static loading
conditions. Finally, since ESA is the type of analysis applied for compacted embankments, it
seems counterintuitive to blindly apply this type of analysis to uncompacted, loose tailings.
Figure 2d shows a case whereby loading from a point below (point 6) to above (point 7) the
collapse surface occurs without failure, because undrained shear was not triggered during the
loading. A stress state can therefore exist in the unstable zone, which in this case would be better
referred to as the metastable zone, where spontaneous collapse (static liquefaction) can occur
with even slight disturbance. USA for this circumstance would yield a factor of safety of less
than 1 (failure at a shear stress corresponding to point 7A), and interpreted literally would suggest
that such a stress state could not exist. Therefore, USA carried out in isolation as a standard limit
equilibrium analysis ignores the need for some disturbance to trigger undrained behavior. Note
again that ESA for such a metastable state ignores completely the potential for collapse of the soil
structure, and gives a completely misleading impression as to the safety of the dam.
3.4 Evidence Of USA Response In Upstream Tailings Deposits
In a previous section we have reviewed the soil mechanics background and shown that for the
typical range of soft normally consolidated clay soils to loose sands that a USA type strength
response is often encountered. What then is the evidence from tailings deposits?
Probably the first reference on the normalized strength of copper tailings slimes is from Castro
and Troncoso (1988) as summarized in Table 1.
Table 1. Data from Castro and Troncoso (1988)
Dam Undrained Peak
Strength Ratio
Undrained Steady
State Ratio
Cerro Negro CN4 0.27 0.07 Slimes, slightly plastic clayey
silt PI of 5-20%
Veta deAgua VA1 0.21 0.11 Slimes, clayey silt.
El Cobre EC4 0.29 0.08 Slimes, no details.
Note: Tests at steady state done with rapid undrained vane tests.
This testing indicates both the USA-D type response as well as the substantially lower steady
state or USA-SS mode.
A summary of other case records is given in Table 2.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 8
Table 2. Summary of USA Response in Tailings
Case Ratio cu/p’ Reference
Aluminum red mud (residual or SS) 0.025 Poulos et. al. (1985)
Lead – Zinc non-plastic slimes 0.2 to 0.22 Vick (1990)
Copper tailings slimes , PI = 6% 0.26 to 0.33 Bromwell (1984)
Copper tailings slimes, PI = 10+/-3% 0.275 Ladd (1991)
Copper tailings slimes N1,60 of from 1 to 5 >0.20 Vidic et. al. (1995)
South African slimes 0.25 CPT
0.17 to 0.4 Vane Blight (1997)
Writing on possible mechanisms for the runout of tailings derived mudflows, Blight (1997)
presents conflicting information on tailings strength from Merriespruit which clearly demonstrates
the USA vs ESA issue. The author states that:
“Tailings consist of sand and silt sized particles of milled rock. The tailings referred to in this
paper contain hardly any clay-size particles and, when sheared, behave as frictional,
cohesionless materials with angles of shearing resistance in the range of 29-35°.”
Data from laboratory undrained dynamic shear tests report dilatant behavior, and Blight (1997)
concludes that the low shear strengths that occurred in the field cannot be explained by postulating
a form of undrained shear mobility of the tailings. However the author also presents the results of
in situ vane and CPT tests (see Table 3 of Blight, 1997). The cone tip resistances are interpreted
by Blight (1997) to give a shear strength gradient of about 2 kPa/m in the interior of the
impoundment i.e., slimes or fine tailings. For an effective vertical stress gradient of about 8.0
kPa/m the USA-D ratio is therefore about 0.25. It is reasonable to think that these CPT probes
are essentially undrained. The author also presents the results of vane testing. The vane peak
tests, which are likely drained give a high strength/vertical effective stress ratio of 0.75 or more
indicating dilatant response for drained shear. Remoulded vanes probably undrained give cu/p’ of
0.17 in the upper 15 m and about 0.40 in the lower 15-30 m of the deposit. It is not known if the
laboratory tests were on undisturbed or remoulded samples. Experience indicates that obtaining
either undisturbed samples of loose tailings or re-creating the in situ fabric by laboratory
techniques is difficult. The field evidence in the form of vane and CPT probes clearly indicates
undrained response, as does the flowslide nature of the failure itself. More exotic explanations of
the failure hardly seem necessary.
The authors had occasion to investigate the recent failure of a tailings dam in South America
designed in the early 1990’s by an internationally known company experienced in tailings dam
design. This dam was designed solely with an ESA framework. After the failure which was due
to undrained loading and resulting shear failure a major site investigation was including SPT and
CPT testing was undertaken by the designer. During this investigation slimes were encountered
and described by the designer as follows:
“Fine tailings (slimes) that induce porewater pressure response during cone penetration are
interpreted by this classification as clayey silts to clay. This interpretation may be correct in
interpreting predominant particle sizes, but suggests a degree of plasticity which is not present
in non-plastic rock flour”
The designers went on to analyze the structure using an ESA approach and determined that it
was safe, notwithstanding the recent failure, to raise the structure as the investigation had
determined that the design parameters were consistent with the original design. Analysis of the
CPT data using the same procedures as Vidic et. al. (1995) by the authors indicated in situ cu/p’
values of about 0.20. The use of this magnitude of strength readily explained the failure which
occurred during a construction lift of the dam. In the authors’ opinion this case record offers a
classic example of the ESA versus USA issue and the absolute fallacy of assuming that “non
plastic rock flour” can only have a so-called drained strength. In this particular case the designers
recognized that the cone testing in slimes induced a pore pressure response but chose to
rationalize the fact away by the “rock flour” model. The dam was condemned.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 9
3.5 Discussion
Ladd (1991) discusses the differences between ESA and USA for staged construction, and
argues convincingly that USA is the correct approach, because when failure of an upstream
tailings dam does occur, it will be under rapid, undrained conditions if the tailings are contractant
under shear, as is so often the case. Carrier (1991), in a paper that should be required reading for
any engineer involved with upstream tailings dams, extends Ladd’s argument specifically to the
case of upstream tailings dams, for which he too convincingly advocates use of USA analysis.
Carrier (1991) also pointed out the fallacy of justifying the assumption of drained shear in tailings
on a high cv value (i.e. assuming a relatively free-draining material would necessarily undergo
shear under fully drained conditions). Comparing an upstream tailings dam to a stress-controlled
triaxial test, Carrier pointed out that, once the peak deviator stress is reached, the strain rate to the
steady state (residual) strength for an undrained, contractant sand can be about 170% per second
(Been and Jefferies, 1985). At such a strain rate, even a highly free-draining material would
undergo shear under undrained conditions.
For upstream dams composed of loose, contractant tailings, failure will therefore typically occur
very rapidly and under undrained conditions, and this is borne out by the fact that upstream tailings
dam failures so often take the form of massive flowslides. Unfortunately, such field behavior is
often not duplicated in laboratory testing, where, out of testing convenience and equipment
limitations, testing is carried out under strain-controlled rather than stress-controlled conditions.
Casagrande (1975) pointed out that rapid (i.e. undrained) collapse deformation is an important
behavior related to flow liquefaction in the field that cannot be observed in strain-controlled tests in
the laboratory. Zhang and Garga (1997), describing the results of triaxial tests on loose sands
carried out under stress-controlled conditions with a rapid data collecting system, suggested that
laboratory equipment limitations and procedures may inhibit collapse deformation behavior,
supporting Casagrande’s view. This work suggests some caution in the adoption of USA-SS
strength characterization from laboratory testing, and especially of the strain-controlled variety.
The advocacy of USA for upstream tailings dams presented in this paper is therefore a well
known approach, best supported by Carrier’s (1991) back analysis of the well known Tyrone
failure case history. The Tyrone failure provided a classic example of the misconceptions as to
tailings strength. This seems to have its origin in the old soil mechanics precepts that sands were
frictional and clays cohesive. The misconception is that if one has non-plastic rock flour such
material is obviously frictional and an ESA analysis applies. The flawed corollary is that if slimes
are not cohesive, then a USA is not required - irrespective of whether or not the slimes are
contractant in shear, and potentially liquefiable. Given these misconceptions, it is not surprising
that tailings dam failures continue to occur. These lessons, shorn of theoretical aspects, were well
understood many years ago by the likes of Casagrande and MacIver (1970), Smith (1972), and
Lenhart (1950), and yet still have not entirely permeated the practice of tailings dam design and
ESA analysis typically overestimates the factor of safety by a factor of two relative to USA
analysis. This in turn would suggest that a great many upstream tailings dams have USA factors
of safety of less than one (since they are rarely designed using ESA to a factor of safety of 2 or
more), and should have failed (based on limit equilibrium analysis), but have not. This in turn
suggests that in many cases an undrained trigger has been absent.
The discussion above for Figure 2d addresses this apparent contradiction. Vick (1992), in a
discussion of Ladd’s Terzaghi lecture (Ladd, 1991) also addressed it, pointing out that for
upstream dams constructed as shown on Figure 1a, which satisfy Rule No. 1 above, ESA had
served well. Vick (1992) also discussed how the breakout of seepage on the dam slope could lead
to rapid, undrained progressive failure even of relatively coarse (free-draining) but loose
(contractant) sands. In this case, the seepage breakout triggered undrained shearing in a
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 10
contractant material, even though drained conditions existed immediately prior to failure and the
material was free-draining, a result consistent with the studies by Eckersley (1990). This
reinforces the notion that some trigger is required to initiate undrained shear failure in upstream
tailings dams.
For upstream tailings dams, unfortunately, potential triggers of undrained failure are numerous,
and include those listed in Table 3.
Table 3. Triggering mechanisms for undrained failures of upstream tailings dams.
Mechanism Trigger
Oversteepening at toe due to: Erosion (intense storm runoff, pipeline break causing washout)
localized, initially drained sloughing
construction activities (excavation)
Overloading due to: rapid rate of impoundment raising
steepening at crest
construction activities at crest
Changes in pore pressures due to: seepage breakout on face of dam
deterioration in performance of underdrainage measures
inhibited volumetric creep
concentrated tailings discharge from one location for extended
leakage/rupture of low level outlet
accelerated rate of construction
foundation and/or embankment movement
intense rainstorms
increased pond levels
Triggering collapse surface by
reduction in mean effective stress
(see Figure 3)
Consider an element of soil below the collapse surface with a low
shear stress and high mean effective stress due to low or absent
phreatic surface. Saturating the slope reduces mean effective
stress, but leaves shear stress constant. The reducing mean
stress results in contact with the collapse surface and
liquefaction is triggered.
Overtopping due to: severe storm runoff
failure of diversion dams/ditches
blockage and failure of spillways/decants
seismic deformation and loss of freeboard
Acceleration/vibrations due to: Earthquakes
construction traffic
It is noteworthy that earthquakes represent perhaps the only of the above triggers for which
analytical methods account for, and are indeed based on, the distinction between contractant
versus dilatant material in shear.
Ladd (1991) observed that failure for contractant materials will occur under undrained
conditions even if drained conditions prevailed immediately prior to the failure, an assertion
supported by work by Eckersley (1990), in his modelling of flowslides. The most elegant
demonstration of this, and the role of Sladen’s collapse surface and drained triggering mechanisms
can be found in a series of tests reported by Sasitharan et. al. (1993). Figure 3 reports on one of
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 11
these tests. In this Figure the state boundary or collapse surface for a clean Ottawa sand is given
at a void ratio of 0.809. In the lab a new sample of sand at a void ratio of 0.804 and at a stress
state represented by a deviator stress of 100 kPa and mean consolidation stress of 300 kPa was
prepared. This sample exists well below the state boundary surface and any minor excursions of
shear stress or reduced or increased mean effective stress would result in a stable response to
loading. However by appropriate manipulation of the loading conditions and back pressures, the
mean effective stress on the sample was reduced but the deviator stress was kept constant. As
the mean stress was reduced in a series of drained increments, the collapse surface was reached.
At this time the void ratio had increased to 0.809, the collapse surface was reached and an
undrained collapse of the sample was triggered. As shown on Figure 3 the deviator stress that
could be supported by the sample reduced to about 20 kPa. The USA-SS strength for this sample
is therefore about 20/300 or 0.067.
Figure 3. State boundary defined by void ratio 0/809 and constant deviator (q=100kPa) drained stress path.
After Sasitharan et al (1993).
Consider then an element of soil in the relatively loose shell of a tailings dam. Assume a low
phreatic surface and a resulting stress state below the collapse surface. This stress state has
reasonable high shear stress and high mean effective stress due to low or absent phreatic surface.
Assume further that the slope saturates due to heavy rainfall, lateral migration of the phreatic
surface due to dam raising or some combination of these or other events. Saturating the slope
reduces mean effective stress, but leaves shear stress constant. If the reducing mean stress
results in contact with the collapse surface liquefaction is triggered. Right up to the initiation of
collapse the soil elements have been drained. This is an example of drained loading triggering an
undrained collapse.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 12
The authors therefore strongly concur with Carrier’s (1991) assertion that upstream dams should
be assessed based on both USA and ESA analysis. However, in recognition of the metastable
nature of upstream tailings dams, the USA analysis results must be interpreted in the context of
the potential triggers of undrained shear listed in Table 1, rather than a limit equilibrium factor of
safety in isolation. Unless considerable effort is expended in the design process and due
consideration is given to all failure modes it is considered appropriate to assume that if a soil can
liquefy it will.
Given that many of these triggers represent operational rather than geotechnical factors, it
follows that it is essential that operational aspects (e.g. rate of raising) be integrated into any
stability assessment and design of an upstream tailings dam. Furthermore, the margin of stability
against shear failure of upstream tailings dams is better expressed jointly in terms of both the
probability of undrained behavior being triggered, and a limit-equilibrium factor of safety.
The authors consider that, for a great many upstream tailings dams, especially those where the
dam slope is underlain by slimes (i.e. violates Rule No. 1), ESA analysis has given the right
answer (i.e. a stable dam), but for the wrong reasons. That is an undrained mechanism never
existed. The fact that such dams have remained stable cannot be considered an endorsement of
universal application of ESA analysis. Fortunate happenstance is no substitute for good design
and analytical practice. Only where Rule No. 1 is satisfied, and the dam configuration is as shown
on Figure 1a, can reliance solely on ESA be justified. Continued reliance on such an approach for
upstream dams that violate Rule No. 1 ignores the fundamentals underlying triggering and pore
pressure response during undrained shear in contractant materials, and is likely to lead to future
failures of upstream tailings dams.
5.1 Estimation of Drained Shear Strength Parameters for Analysis
For assessments of the stability of a number of upstream tailings dams at base metal mines in
which the authors were recently involved, little or no data was available with which to estimate the
drained and undrained strength parameters for the tailings dams assessed. There is abundant case
history experience to draw upon in terms of drained strength parameters. Effective friction angle
(φ’) values typically range between 25° and 35° for base metal tailings, with the lower portion of
this range applying for silt/clay tailings slimes, and the upper portion applying for coarser tailings
deposited on beaches. Table 2.8 of Vick (1990) gives a summary of typical values of φ’ for
various types of tailings.
5.2 Estimates of Undrained Shear Strength Parameters
The brief review presented earlier for normally consolidated clays and loose sands indicate two
forms of USA behavior: ductile and brittle or steady state / residual modes. We have also
reviewed a series of references where many examples of undrained strength ratios (cu/p’)
reported in the literature, where a range of 0.2 to 0.3 appears to cover most of the cases for
tailings at least for USA-D type response. This suggests that many tailings have an USA
response very similar to normally consolidated clays.
However the test data presented by Castro and Troncoso (1988) introduces the considerable
caution that a USA-SS response may also be present. Much of the data available and discussed
above is based on insitu interpretations of CPT response.
What we have called USA-SS mode or strength response has been discussed in detail by
Poulos (1988) who recommends an approach more conservative that the USA approach (which is
based on peak undrained shear strength). For the USA-SS mode design analyses are carried out
based on the steady state, or residual, undrained strength. This approach assumes that the worst
case scenario (straining of the contractant tailings sufficient to cause undrained collapse of the soil
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 13
structure) governs design, reducing shear strength from peak (USA analysis, point 2 on Figure 2c)
to residual (Poulos method, point 3 on Figure 2c) levels. Carrier (1991) supports consideration of
this approach. He emphasizes that it is necessary to consider the stress-strain characteristics of
the tailings, with brittle, cohesionless tailings (strain to peak about 1% typically) more relevant and
critical for a Poulos type analysis that clayey, ductile, cohesive slimes (strain to peak much larger
than 1%).
The authors advocate that, for upstream dams, the Poulos approach should also be considered
for both loose sands and well as slimes. This forces the stress-strain characteristics of the
materials forming the dam to be considered in the analysis. The steady state approach is widely
applied in analyzing the seismic stability of upstream tailings dams. Given that seismic loading is
but one of many potential mechanisms of undrained loading (McRoberts and Sladen, 1992), it is
inconsistent that the steady state approach not be at least considered under static loading
conditions. Application of the method is not without its problems, particularly with respect to
determination of the appropriate steady state strength, which has been the subject of much
research and debate within the geotechnical profession for many years. Considering the
difficulties of predicting the appropriate USA strength is however a step in the right direction;
assuming the designer has at least abandoned a sole reliance on the ESA mode.
5.3 Pore pressure conditions for stability analyses
Use of both ESA and USA requires an estimation of the effective consolidation stresses at any
given time, which in turn require a good understanding of the pore pressure conditions (hydrostatic
versus downward drainage, normally consolidated versus excess pore pressures) within the dam.
Pore pressure conditions within upstream tailings dams are very often complex, misunderstood,
and improperly incorporated into stability analysis. Misinterpretation of piezometer data can easily
occur if adequate piezometer coverage does not exist. Vick (1990) suggests that, for rates of
impoundment rise of between 15 and 30 ft/year, excess pore pressures are usually assumed to
dissipate as rapidly as the load is applied, and therefore a normally consolidated state (i.e. zero
excess pore pressure) can be assumed. Mittal and Morgenstern (1976) also suggested this range
as being sufficient to generate excess pore pressures in slimes.
The authors caution that these experience-based criteria on rate of rise can be safely applied
only in cases with good underdrainage (permeable foundation relative to the tailings slimes),
relatively smaller embankments (35 m in height or less) with relatively short drainage paths, and
slimes free of significant clay content and plasticity. For example, the authors are aware of one
large upstream dam in which very high excess pore pressures exist in the clayey slimes despite a
rate of rise of only about 7 ft/year. Another example that emphasizes the need for caution is the
Tyrone tailings dam, which failed under undrained conditions at a construction rate of 12 to 15
5.4 Recommended approach
To summarize, the authors provide the following general recommendations for any static stability
analysis of an existing upstream tailings dam:
1. Determine whether or not the dam slope is comprised of materials that are contractant or
dilatant under shear. Characterize pore pressure conditions within the dam, to properly determine
effective stresses. If the materials are dilatant and/or fully drained (unsaturated), then only a ESA
is required.
2. If the dam slope is fully or partially composed of contractant materials, then both ESA and
USA should be carried out. The factor of safety from the USA will better represent the margin
of safety of the dam. Considerable care must be exercised in the selection of the appropriate
USA strength mode. Careful attention must be given to whether or not a USA-D or USA-SS
mode might be triggered.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 14
3. Review the operational and design factors necessary to assess the probability of undrained
shear being triggered by the various mechanisms listed in Table 1. The probability of these
mechanisms should be considered jointly with the USA factor of safety in evaluating the safety of
the dam. Unless one is very sure that undrained triggers absolutely do not exist, it is considered
prudent to assume the worst.
4. Evaluate the stress-strain behavior of the tailings comprising the dam. If the tailings are both
contractant and brittle (low strain to peak strength, and significant post-peak reduction in strength),
then a Poulos (steady state strength or USA-SS mode) analysis should be carried out.
The results of this analysis should be considered jointly with the assessed probabilities of
undrained shear triggering mechanisms.
A dam configuration typical of several of those recently reviewed by the authors is shown on
Figure 4. This dam was designed in the early 1990’s and includes a compacted starter dam with
drainage and filter zones as shown. The dam is being raised about 25 m above the starter dam
crest, at a slope of 3H:1V, using spigotted tailings produced from a flotation circuit (base metals
mine). The design specified that a minimum 30 m wide beach be maintained at all times.
Fortunately, the operators were able to maintain somewhat wider beaches than this in all but one
of the cases.
Figure 4. Typical design section of upstream dams reviewed.
The designer’s stability analyses were of the ESA type, and assigned both the sand shell and
the slimes an effective friction angle of 30°. Therefore, by assuming the same shear strength for
both zones, the only effect of the beach width in terms of the design analyses was on the assumed
location of the phreatic surface. The designer’s analysis also assumed hydrostatic pore pressure
The stability analysis geometry and results obtained by the authors are illustrated graphically on
Figures 5 and 6, for the design minimum beach width of 30 m. They clearly show that ESA
provides an acceptable factor of safety to dam heights of up to 30 m (above the starter dam
crest), while USA-D analysis indicates inadequate factors of safety. The USA analysis results
shown on Figure 6 may well be slightly unconservative because they assume ESA can be applied
for the outer sand shell. However, this material may also be contractant in shear, so undrained
strengths could be more appropriate in the saturated portions of this zone. The analyses also do
not account for the stress path dependence of USA-D strength discussed in Section 3.2.
Moreover, depending on the ability of the tailings to absorb a degree of straining a lower USA-SS
strength may be appropriate
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 15
Figure 5. Stability analysis section.
Figure 6. Results of preliminary stability analyses.
Upstream tailings dams are complex structures for which shear strength and pore pressure
conditions are difficult to predict in advance. It is essential that stability analyses of these
structures be predicated on a thorough understanding of the behavior of the tailings under shear,
and an understanding of the likelihood of potential triggering mechanisms of undrained failure.
Stability analysis of upstream tailings dams should be carried out using both drained strength
analysis and undrained strength analysis. In some cases it may also be necessary to consider
steady state strength analysis. This approach forces the designer to come to grips with the issue
of contractant versus dilatant behavior in shear, and with the stress-strain characteristics of the
material, two of the most fundamental precepts of soil mechanics that are ignored at the
geotechnical engineer’s peril.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 16
Been, K., & M.G. Jefferies 1985. A state parameter for sands. Geotechnique, Vol. 35, June,
Berti, G., Villa, F., Dovera, D., Genevois, R., and J. Brauns 1988. The disaster of Stava/northern
Italy. Hydraulic Fill Structures, ASCE Spec. Public. No. 21, pp. 492-510.
Blight, G.E. 1997 Destructive mudflows as a consequence of tailings dyke failures. Pro. Inst.
Civil Engrs. Geotech. Engng. 125, Jan., 9-18.
Blight, G.E. 1988. Some less familiar aspects of hydraulic fill structures. Hydraulic Fill
Structures, ASCE Spec. Public. No. 21, pp. 1000-1027.
Bromwell, L.G. 1984. Stability investigation of Tailings Dam 1, Tyrone Mine, Tyrone, New
Mexico. Report to Phelps Dodge Corporation, Tyrone Branch, Tyrone, New Mexico, report
prepared by Bromwell and Carrier, Inc.
Carrier, W.D. 1991. Stability of tailings dams. XV Ciclo di Conferenze di Geotecnica di
Torino, Italy, November.
Casagrande, A. 1975. Liquefaction and cyclic deformation of sands: a critical review. Harvard
Soil Mech. Series No. 88.
Casagrande, L., and MacIver, B.N. (1970). Design and construction of tailings dams.
Proceedings, Symposium on Stability in Open Pit Mining.
Castro, G. and Troncoso, J., 1988 Effects of 1985 Chilean earthquake on three tailings dams.
%th Chilean Conference on Seismology and Earthquake Engineering.
Chandler, R.J., and G. Tosatti, 1995. The Stava tailings dam failure, Italy, July 1985. Proc. Inst.
Of Civil Eng., Vol. 118, April, pp. 67-79.
Dobry, R., and L. Alvarez, L. 1967. Seismic failures of Chilean tailings dams. Journ. Soil Mech.
& Found. Eng. Div., ASCE, Vol. 93, No. SM6.
Eckersley, D. 1990. Instrumented laboratory flowslides. Geotechnique, 40(3), pp. 489-502.
Joshi, R.C., Achari, G., Kaniraj, S.R., and Wijeweera, H. (1995). Effect of aging on the
penetration resistance of sands. Canadian Geotech. Journ., Vol. 32(5), pp. 767-782.
Ladd, C.C. 1991. Stability evaluation during staged construction. The 22nd Karl Terzaghi Lecture,
Boston, 1986, Journ. of Geotech. Eng., ASCE, Vol. 117, April, pp. 537-615.
Lenhard, W. 1950. Control of tailings from washing plants. Rock Products (July), pp. 72-80.
Martin, T.E., and I. Tissington 1996. Design evolution of tailings dams at Inco Sudbury.
Proceedings, Tailings & Mine Waste ’96, January, pp. 49-57.
Poulos, S.J. (1988). Strength for static and dynamic stability analysis. Hydraulic Fill Structures,
ASCE Spec. Public. No. 21, pp. 452-474.
Poulos, S., Robinsky, E., and Keller, T 1985 Liquefaction resistance of thickened tailings. J.G.E.,
Vol. 111, No12 1380-1393.
McRoberts, E.C, and Sladen, J., 1992 Observations on static and cyclic sand liquefaction
methodologies. Canadian Geotech. Journ., 29, 605-665.
Mesri, G. 1989 a revaluation of su(mob) = 0.22σp using laboratory shear tests. Canadian Geotech.
Journ., Vol. 26(1), 162-164.
Mittal, H.K. and N.R. Morgenstern 1976. Seepage control in tailings dams. Canadian Geotech.
Journ., Vol. 13(3), August, pp. 277-293.
Sasitharan , S, Robertson, P.K., Sego, D., and Morgenstern, N.R. 1993. Collapse behavior of
sand. Canadian Geotech Journ., Vol. 30, pp. 569-577.
Sladen, J.A., D’Hollander, R.D., and J. Krahn 1985. The liquefaction of sands, a collapse surface
approach. Canadian Geotech. Journ., Vol. 22, pp. 564-578.
Smith, E.S. 1972. Tailings disposal - failures and lessons. Proceedings, 1st International Tailing
Symposium, Tucson, Arizona, pp. 356-376.
Troconso, J. 1988. Evaluation of seismic behaviors of hydraulic fill structures. Hydraulic Fill
Structures, ASCE Spec. Public. No. 21, pp. 475-491.
Troncoso, J. 1990. Failure risks of abandoned tailings dams. Proc. Int. Symposium on Safety and
Rehabilitation of Tailings Dams, ICOLD.
USCOLD 1994. Tailings dam incidents. U.S. Committee on Large Dams, 82 p.
T.E. Martin, E.C. McRoberts: Some considerations in the stability analysis 17
Vick, S. 1990. Planning, design, and analysis of tailings dams. Bitech Publishers, ISBN 0-
Vick, S. 1992. Stability evaluation during staged construction - discussion. ASCE, Journ. of
Geotech,. Eng., Vol. No. 118, No. 8, August, pp. 1282-1289.
Vidic S, Mayne P., and Beckwith Geo., 1995 Profiling mine tailings with CPT Proc. Int. Symp.
On CPT, CPT’95, vol2 607-612.
Wride, C.E., McRoberts, E.C., and Roberston, P.K. Reconsideration of Case Histories for
Estimating Undrained Shear Strength in Sandy Soils. Manuscript submitted to the Canadian
Geotech. Journ., April 1998.
Wroth, C.P., 1984. The interpretation of insitu soil tests. Geotechnique 34, No. 4, 449-489.
Yoshimine, M., Robertson, P.K. and Wride, C.E. 1998 Undrained shear strength of clean sands.
Manuscript submitted to the Canadian Geotech. Journ. April 1998.
Zhang, H. and V.K. Garga 1997. Quasi-steady state: a real behavior? Canadian Geotech.
Journ., Vol. 34(5), October, pp. 749-761.
... The practice of hydraulically disposing tailings into a large wet impoundment, without allowing for the necessary time for settlement and consolidation, may promote a loose, saturated tailings structure that is susceptible to brittle stress-strain behaviour and liquefaction following a triggering mechanism (Martin and McRoberts, 1999;Blight, 2010;Jefferies and Been, 2016). Numerous case studies (e.g. ...
... The state parameter (ψ) is the difference between the existing void ratio and the critical void ratio (i.e. the void ratio at which no volume change occurs during shear as determined from reconstituted samples) of the tailings at the same mean effective stress (Jefferies and Been, 2016). In saturated conditions, loose, contractive materials (+ψ) are potentially liquefiable upon undrained shearing whereas dilative materials (− ψ) are more compact and demonstrate greater resistance to liquefaction (Vick, 1983;Martin and McRoberts, 1999;Blight, 2010;Morgenstern et al., 2016). During flow liquefaction, impounded tailings undergo rapid strain-softening (i. ...
... Failures of some upstream-raised dams have been attributed to partial founding on soft, compressible, poorly consolidated slimes that consequently underwent lateral extrusion, undrained creep and development of excess pore water pressures to the point of failure (Dobry and Alvarez, 1967;Troncoso et al., 1993;AECOM, 2009;Morgenstern et al., 2016). However, as counter-argued by Vick (1992), Martin and McRoberts (1999) and Morgenstern (2018), the prominent causes of these instabilities may lie in improper design, construction and operation of the facilityincluding, but not limited to, the overabundance of contractive materials, insufficient drainage and seepage control and a loose embankment corerather than the general upstream method itself. Fig. 4 presents a temporal bar chart distribution of major causative variables for 46 tailings impoundment failures that produced mass flows. ...
Full-text available
Tailings dam failures have received significant attention in recent years due to the catastrophic downstream consequences, as evidenced by the 2019 Feijão disaster in Brazil and numerous precedents. This paper presents a timely review of tailings flows with the support of a comprehensive global database of 63 cases that have been remotely analyzed through a compilation of satellite imagery, digital elevation models and literature. The synthesis provides insight into the influence of impoundment conditions, preconditioning and trigger variables, failure mechanisms and the downstream environment on the field behaviour of tailings flows. The database also sheds light on the limitations of data quality and availability in the public domain. Magnitude-frequency statistics indicate that tailings dam breaches that have produced catastrophic mass flows with total outflow volumes of ≥1 M m³ have occurred at a mean recurrence interval of 2–3 years over the period 1965–2020. Weather hazards and impoundment drainage issues are identified as major causative variables. The occurrence of liquefaction and/or the incorporation of free water are sufficient conditions to trigger extremely rapid, highly mobile behaviour. Travel path confinement and steeper bed slopes enhance flow velocities (peak of 25–30 m/s) and kinetic energy, whereas flow mobility appears to be exacerbated along major rivers. Although general trends may be observed in empirical observations, such efforts are prone to substantial uncertainty due to the complexity and variability of site conditions (that are typically unaccounted for in broad statistical approaches) as well as poor data availability and/or quality for many of the selected cases. This highlights the importance of performing site-specific investigations through numerical models, laboratory tests and field observations to better predict post-breach behaviour (ideally within a probabilistic framework) when undertaking site assessments.
... In 2019, a massive catastrophic flow failure of an iron ore tailings dam (Feijao I) took place in Brazil causing the pollution of the Paraopeba river and many fatalities. This tailings dam was constructed using the upstream method which builds accumulating vertically the tailings through successive disposals over soft tailings (Kossoff et al., 2014).Upstream tailings dams are considered at high risk because the stability of the tailings facility diminishes substantially the operations cease (Martin & McRoberts, 1999). Static liquefaction of iron tailings was the underlying collapse mechanism of the Feijao I dam which was triggered by loss of suction and creep (P. ...
Full-text available
Large-scale mining operations generate vast quantities of tailings that are placed in hydraulic-fill tailing dams. The stability of these tailings dams against flow failure triggered by liquefaction is of paramount importance to protect lives, infrastructure, and the surrounding environment. In 2019, a massive catastrophic flow failure of an iron ore tailings dam took place in Brazil raising concerns on the safety of this type of facilities around the world. The phenomena of flow failure triggered by either static or dynamic liquefaction on Bauxite and Iron ore tailings dams have not been studied extensively. Unlike sandy tailings, the clayey mineralogy of Bauxite tailings due to its residual formation pose a challenge on geotechnical characterization. Furthermore, effects of ageing and cementation are not clearly understood. This paper aims to study the flow liquefaction susceptibility of Bauxite tailings of a case study located in a country with low seismic hazard and therefore where static liquefactions is prevalent. The framework of critical state soil mechanics was applied using data from a comprehensive site investigation including several Cone Penetration Testing (CPTu). The purpose is to enhance the understanding of the behavior of Bauxite tailings at large deformations and to verify the applicability of well-known correlations. Results show that current screening methods (Plewes and Robertson) for the defining whether a bauxite layer is either contractive or dilative are mostly not in agreement. Hence, a word of caution is recommended when applying this methods to Bauxite tailings being necessary to conduct further analysis.
... Fine tailing storage is a difficult problem that must be managed in tailing treatment technology, and this exits in the fields of gold, copper, leadzinc, and other materials. Martin and McRoberts [1] pointed out that the stability of tailing dams constructed using the upstream method is a challenging problem due to the great uncertainty of the shear strength of fine tailings or cement. Fine tailings have the following engineering characteristics: high moisture content, high compressibility, difficult deposition, and difficult consolidation. ...
Full-text available
In light of the engineering characteristics of fine-grained tailings, such as their high moisture content, high compressibility, difficult deposition, and difficult consolidation, and based on the summary and analysis of the experience of tailing dam construction in China and abroad, a new comprehensive treatment technology suitable for dam construction on a weak beach, which is composed of strengthening foundation with the soilbag, subdam stabilization, and collaborative drainage, is proposed. Taking a dam construction project on the weak beach of a tailing pond in China as the research object and through the study of the reinforcement mechanism of the soft beach treated using the soilbag method, it can be seen that the soilbag body has the mechanism of strength growth by sand fixation and water permeability and constraint formation and can achieve an overall stable effect of extrusion and silt replacement, stacking and surcharge, and flexible embedding in a weak beach. Using this method, combined with relevant monitoring technology, a good demonstration of dam construction speed of up to 30 m/a and a safe dam construction height of up to 60 m on a weak beach surface is achieved in a tailing dam construction project, which provides a good reference for the design and construction of fine tailing dam engineering and the cooperative disposal of adverse geological conditions.
... In tailings dams, there are several plausible triggers (e.g. Martin & McRoberts 1999). Where the consequence of failure is high, it is often considered prudent to assume that strength loss will be triggered (e.g. ...
Conference Paper
Full-text available
Soil liquefaction is a major concern for structures constructed with or on saturated, contractive sandy soils as it can result in sudden loss of strength and subsequent flow failure. The first step in designing against liquefaction is to assess the liquefaction potential of the soil. There are several methods available, however, many of these methods were developed for specific soil types (e.g. sands or silts). In this study, four such methods were assessed with a specific focus on sites where mixed soils are present. In particular CPTu probes through gold tailings and the Salzburg Basin material (natural sandy alluvium underlain by clay) were assessed. Using the CPTu as a screening tool, the state parameter with depth was estimated and zones of material susceptible to liquefaction were identified for each material. Good correlation in terms of identifying material zones susceptible to liquefaction was found between the method based on drained probing (Been et al. 1987) and the method based on undrained probing (Shuttle & Cunning 2007). It appears there is little difference in the estimated state parameter whether soil type is considered or not. This suggests that for screening purposes, there may be no need to consider soil type when conducting a liquefaction potential assessment.
... Vick [3] addressed the need to manage the pond filling rates of an upstream tailings dam to prevent the build-up of excess pore water pressure that can reduce the shear strength of the fill material. Excessive rates of the pond filling would cause a trigger for static liquefaction that has been the underlying cause for many upstream tailings impoundment failures [19]. ...
Full-text available
One of the challenges in upstream tailings dam projects is to ensure the allowable rate of deposition of tailings in the pond (i.e., pond filling rate) while maintaining the stability of the dam. This is due to the fact that an upstream tailings dam is constructed by placing dikes on top of previously deposited soft tailings, which could lead to a decrease in dam stability because of the build-up of excess pore water pressure. The main purpose of this work is to investigate the effects of pond filling rates on excess pore water pressure and the stability of an upstream tailings dam by a numerical study. A finite element software was used to simulate the time-dependent pond filling process and staged dam construction under various pond filling rates. As a result, excess pore water pressure increased in each raising phase and decreased in the subsequent consolidation phase. However, some of the excess pore water pressure remained after every consolidation phase (i.e., the build-up of excess pore water pressure), which could lead to a potentially critical situation in the stability of the dam. In addition, the remaining excess pore water pressure varied depending on the pond filling rates, being larger for high filling rates and smaller for low filling rates. It is believed that the approach used in this study could be a guide for dam owners to keep a sufficiently high pond filling rate but still ensure the desirable stability of an upstream tailings dam.
... Apart from the increase in the height of the structure itself, it is necessary to take into account the non-stationary state of filtration, in particular, pore pressure distribution within the sediments themselves, as well as the dam body and the subsoil (Jiang & Tang, 2015;Li et al., 2009;Sitharam & Hegde, 2017). Stability is affected by the pore pressure (Martin & McRoberts, 1999), in particular, the water table localization (Dong et al., 2017;Tang et al., 2016;Yin et al., 2011). It is also very important to correctly determine the parameters of waste and substrate, which constitute input data for the stability verification calculations. ...
Full-text available
In this paper, a qualitative safety analysis of the Żelazny Most tailings pond is addressed. This object is one of the largest facilities of this type in the world being a crucial element in the technological line of copper production in KGHM Polska Miedź S.A. The assessment of the effectiveness of two types of preventive measures, i.e., relief wells and loading berms, is investigated based on displacement and stability analysis of two 2D cross-sections in a technical section of the dam. The study shows that the considered preventive measures generally have a positive impact on increasing the safety level of the structure during its further raise. In particular, their effectiveness is most evident when they are applied simultaneously. It is eventually suggested that the selection of final solutions to be applied on the facility should be based on the quantitative 3D analysis.
... In upstream dams, the additional layers of structure are placed on top of tailings depositions, therefore the construction's safety depends on the integrity of tailings for stability (Davies and Martin, 2002). This type of dam requires greater ongoing scrutiny (Martin and McRoberts, 1999). Figure 1: Tailings dam construction method: conceptual illustration. ...
Full-text available
The use of Internet of Things (IoT) technologies is becoming a preferred solution for the assessment of tailings dams' safety. Real-time sensor monitoring proves to be a key tool for reducing the risk related to these ever-evolving earth-fill structures, that exhibit a high rate of sudden and hazardous failures. In order to optimally exploit real-time embankment monitoring, one major hindrance has to be overcome: the creation of a supporting numerical model for stability analysis, with rapid-enough response to perform data assimilation in real time. A model should be built, such that its response can be obtained faster than the physical evolution of the analyzed phenomenon. In this work, Reduced Order Modelling (ROM) is used to boost computational efficiency in solving the coupled hydro-mechanical system of equations governing the problem. The Reduced Basis method is applied to the coupled hydro-mechanical equations that govern the groundwater flow, that are made non-linear as a result of considering an unsaturated soil. The resulting model's performance is assessed by solving a 2D and a 3D problem relevant to tailings dams' safety. The ROM technique achieves a speedup of 3 to 15 times with respect to the full-order model (FOM) while maintaining high levels of accuracy.
... Martin et al. (2002) provide 10 basic rules of good practice for the design, construction, and management of upstream TDs that should also be followed for centerline and downstream types. These 10 rules are by no means new, and are restatements of principles outlined by previous authors (Casagrande 1950(Casagrande , 1975Casagrande and MacIver 1970;Ladd 1986;Lenhart 1950;Martin and McRoberts 1999;Martin et al. 2002;Robertson 2010;Robertson et al 2019;Rodríguez 2018;Smith 1969;Vick 1992). In all TD failures, more than three of these 10 basic rules have been broken Oldecop and Rodríguez 2006;Robertson 2010). ...
Full-text available
Based on research carried out at 67 tailings dams in Spain: (1) tailings dams contain alternating sedimentary layers with contractive and dilative geomechanical behaviours; (2) tailings saturate quickly but drain more than 10 times slower due to the high-suction capacity of the porous sediments (2–300 MPa); and (3) over the long-term, a stationary flow regime is attained within a tailings basin. Four temporal and spatial conditions must all be present for a tailing dams flow failure to occur: (1) the tailings must experience contractive behaviour; (2) the tailings must be fully saturated; (3) the effective stress due to static or dynamic load must approach zero; and (4) the shear stress must exceed the tailings residual shear stress. Our results also indicate that the degree of saturation (Sr) is the most influential factor controlling dam stability. The pore-pressure coefficient controls geotechnical stability: when it exceeds 0.5 (Sr = 0.7), the safety factor decreases dramatically. Therefore, controlling the degree of tailings saturation is instrumental to preventing dam failures, and can be achieved using a double drainage system, one for the unconsolidated foundation materials and another for the overlying tailings.
Failure of soil slopes is often associated with instability of soils. Instability refers to a behavior in which large plastic strains are generated rapidly when a soil element sustains a given load or stress. Currently, the research related to instability of soils is primarily conducted at saturated conditions through undrained triaxial tests on loose saturated soils (e.g., Lade, J Geotech Eng 118:51–72, 1992; Leong et al., Geotech Test J 23:178–192, 2000; Yang, Geotechnique 52:757–760, 2002) and drained constant shear tests on saturated medium and dense sands (Chu et al., Can Geotech J 40:873–885, 2003). However, many natural soil deposits encountered in engineering practice are often unsaturated. During rainfall infiltration, a reduction in soil suction causes a decrease in the shear strength, which leads to the development of plastic strains and ultimately to the instability of the soil. This process can be idealized as a wetting path along which the shear stress and net mean stress keep constant, but the suction decreases over time. However, the instability behavior of unsaturated granular soils along the wetting path has seldom been investigated.
The use of Internet of Things (IoT) technologies is becoming a preferred solution for the assessment of tailings dams safety. Real-time sensor monitoring proves to be a key tool for reducing the risk related to these ever-evolving earth-fill structures, that exhibit a high rate of sudden and hazardous failures. In order to optimally exploit real-time embankment monitoring, one major hindrance has to be overcome: the creation of a supporting numerical model for stability analysis, with rapid-enough response to perform data assimilation in real time. A model should be built, such that its response can be obtained faster than the physical evolution of the analyzed phenomenon. In this work, Reduced Order Modeling (ROM) is used to boost computational efficiency in solving the coupled hydro-mechanical system of equations governing the problem. The Reduced Basis method is applied to the coupled hydro-mechanical equations that govern the groundwater flow, that are made non-linear as a result of considering an unsaturated soil. The resulting model’s performance is assessed by solving a 2D and a 3D problem relevant to tailings dams safety. The ROM technique achieves a speedup of 3 to 15 times with respect to the full-order model (FOM) while maintaining high levels of accuracy.
This book describes the processes by which tailings are produced, and why the solids must be separated from them. The usual method is by allowing the tailings slurries to settle in ponds, retained by temporary dams. These dams are described in all their aspects, both the theoretical behaviour of fluids in the ponds, and engineering aspects of their placing, design and construction, and the behaviour and prevention of seepage. (U.K.)
The expression Su(mob) = 0.22σ′p for the average undrained shear strength mobilized on a slip surface in the field resulted from in situ vane Su/σ′vo and odeometer σ′p/σ′vo data, combined with a correction factor μ obtained from the computed factor of safety of unstable embankments, footings, and excavations. It is shown here that the same expression for mobilized undrained strength of soft clays is also obtained from laboratory shear tests by taking into account anisotropy and time effects. This result is highly significant, since the laboratory undrained shear strength data, as well as the correction factor for the time effect, are completely unrelated to the in situ data that previously resulted in the expression for field mobilized undrained shear strength.
Staged construction uses controlled rates of load application to increase the foundation stability of structures founded on soft cohesive soils and to improve the slope stability of tailings dams. Because construction causes positive excess pore pressures and because actual failures usually occur without significant drainage, stability analyses should compute the factor of safety against an undrained failure as the most critical and realistic condition. This requires an undrained strength analysis (USA) that treats predicted or measured in situ effective stresses as equal to consolidation stresses in order to calculate variations in undrained shear strength during construction. The recommended USA methodology requires a detailed evaluation of changes in vertical stress history profiles, uses undrained strength ratios obtained from CK0U tests to account for anisotropy and progressive failure, and is more rational than stability evaluations based on UU and CIU triaxial compression testing. Conventional effective stress analyses should not be used for staged construction because the computed factor of safety inherently assumes a drained failure that can give highly misleading and unsafe estimates of potential instability.
A new method is presented for analysis of the potential for triggering liquefaction, i.e., a flow slide, in liquefiable soil masses. The method is based on the principle of steady state deformation. In conventional mine tailings disposal operations, the tailings are pumped as a water suspension into extensive flat containment ponds. These ponds generally are formed by construction of large tailings dams. However, if the tailings are thickened substantially before discharge, the tailing slurry will form a sloping deposit. Maximum slope angles of 3.5° are normally recommended. Use of thickened tailings is less costly and has much less environmental impact than conventional tailings because tailings dams and their associated slime ponds are eliminated. In this paper, the resistance to liquefaction due to earthquakes of a proposed bauxite tailings deposit placed at a 2.9° slope is analyzed. The water content of the thickened tailings is high enough to make them susceptible to liquefaction even when placed at such gentle slopes. However, due to the clay content and thixotropic nature of these tailings, earthquakes that induce 0.1 g peak ground acceleration do not cause enough strain to trigger liquefaction.
Disastrous mudflows, involving the movement of large volumes of semi-liquid tailings, have followed the failure of many tailings containments. For instance, the mudflow that followed the failure of a tailings dam complex at Stava, Italy, killed nearly 300 people. However, in other cases tailings dams have failed with very little post-failure movement or escape of tailings. This paper examines the behaviour and consequences of failures in five tailings ringdykes in southern Africa and attempts to define the circumstances under which a mudflow will occur. It is shown that the occurrence of a mudflow is closely associated with the condition (dry or wet) of the ground surface on to which the escaping tailings move. If the ground surface is dry, it is likely that the tailings will not move far whereas if it is wet, a mudflow is much more likely to ensue.