ABSTRACT: Catastrophic flow failures occur in mine tailings dams and dumps of discards and other mine waste with alarming
frequency. In recent years catastrophic flow failures have also occurred in dumps of municipal refuse and even in what were
considered to be carefully controlled and well engineered landfills. Apart from the environmental devastation caused by these flows,
they are also dangerous to human life and society. For examples the Buffalo Creek disaster in the USA in 1972 killed 118 people,
made 4000 homeless and destroyed 50 million US dollars worth of property and facilities, the flow slide that occurred in the
Umraniye-Hekimbasi refuse dump in Turkey in 1993, killed 39 people, destroying their homes in the process. The paper will briefly
review some of the more typical flow slides in waste materials, analysing the mechanics of failure and pointing to ways of preventing
this type of failure by a combination of sound design and operating procedures. In the case of existing deposits modified operating
procedures can be adopted, reducing the probability of failure as well as constructing deflecting structures to protect communities and
facilities from the consequences of failure.
surface, along the valley. The sewage that poured from the sewer
pipe was dammed by the slide debris and formed a lake of
sewage on the upstream side of the obstruction. Since this
occurrence, two and possibly three more flow failures of
municipal solid waste deposits have been reported (Hendron et al
1999 (17, Table 1), Brink et al 1999 (18, Table 1)).
Deviator
Stress
(kPa)
Deviator
Stress
(kPa)
Pore
Pressure
u
(kPa)
Pore
Pressure
u
(kPa)
Figure 10: The effect of initial relative density (Dr) on the shape
of the stress-strain curves of consolidated undrained tests on
saturated sand. (Castro, 1969)
increasing strain. The strength of the intermediate specimen
remained more or less constant.
.3 In the loose specimen, the pore pressure rapidly rose with
increasing strain, to reach a constant maximum. In the dense
specimen, after rising to a peak, the pore pressure reduced
continuously and the shear strength increased.
Figure 11 :
a) Stress-strain curves for stress-controlled
anisotropically consolidated undrained tests on saturated loose
sand ( s 31c = 400kPa)
b) Corresponding stress paths (Castro, 1969).
Figure 12: Stress path for dynamic shear test on loose saturated
natural sandy silt.
Some gold mines in South Africa sluice their coarse wastes with
waste mine water as a means of disposing of waste water. The
sluicing causes the rock to compact, reducing its tendency to
contract, but may unfortunately increase acid seepage from the
base of the dumps, leading to undesirable surface and ground
water pollution.
During the quake a cloud of dust arose from the dried surface
of the only periodically used impoundment. The flow failure
continued for 20 minutes after the quake had ended, as 1.9 x
106m3 of a total storage of 4.25 x 106m3 of tailings flowed down
a dry valley for a distance of 12km. A town in the path of the
flow was annihilated with 300 deaths occurring.
As shown by Figure 19, the dam was constructed on sloping
ground with a slope angle of 3 and the average slope of the postfailure profile through the breach was only 3.5. The flow was
reported to have covered its 12km course in a few minutes. This
is too imprecise to allow the speed of the flow to be estimated,
but it must have been about 20kmh-1 (see Section 4).
Figure 20: Sections through failure at Merriespruit showing postfailure equilibrium surface.
causing the noise and being followed by a wave of mud (see
Figure 5).
Figure 20 shows sections of the post-failure equilibrium
surface for the failure basin of the Merriespruit tailings
impoundment. Section E'E' is the pre-failure section normal to
the wall and EE is a section through the breach. Section FF
runsat right angles to EE, and GG runs at 45 to EE. The
intersections of FF and GG with EE are marked in Figure 20. The
slope of the tongue of escaped tailings was 2, which is very
similar to the slope of large portions of FF and GG. In other
words, the post-failure surface had flattened to a general slope of
2-3, with some portions around the perimeter of the failure scar
being as steep as 10-20. Presumably, these areas had formed
late in the failure process, had been subjected to lesser shear
strains because they were not so high, and were therefore stable
at steeper surface slopes.
Because of the disturbance caused by the failure, it is very
difficult to know from what depth in the impoundment the
material that composes the post-failure surface originated. The
surface is also too soft to be accessible after a failure until a
drying crust has formed. Hence it is not possible to sample a
post-failure surface straight after the failure to help identify its
depth of origin. It seems likely, however, that the tailings that
move out of the breach will consist of the upper, more recently
deposited layers, and that the post-failure surface will consist of
deeper layers exposed as the slope of the failure basin is flattened
by the outward flow of the tailings.
For example, Figure 21 shows profiles of vane shear strength
measured in an operating gold tailings impoundment. In the
event of the outer dyke being breached, it is obvious from their
relatively low strength that the top 10m of tailings would tend to
flow off more readily than the deeper layers. Figure 21 also
demonstrates the loss of in situ strength of the tailings when
disturbed, with a sensitivity ratio or strength reduction factor
(undisturbed/remoulded strength) of about 2.7.
10
waste was 9m. The speed of the flow was estimated to have been
15 to 30kmh-1.
3.3 Flow failures in municipal solid waste
The flow failure at Istanbul (11, Table 1) will be taken as the
archetypical example of this type of flow failure (Kocasoy and
Curi 1995). It is remarkable not only for the destruction it
wrought, but also for the lack of common sense of the authorities
that established and operated the landfill. Figure 8 shows that the
landfill must have been sited where it was, purely for reasons of
expediency. Given some flexibility in siting, no engineer in his
right mind would have sited a waste deposit on a 27 slope. The
waste was dumped near the edge of the slope, sorted through by
informal reclaimers (i.e. scavengers), and then pushed over the
edge by dozer where it came to rest at an angle of repose of 45.
There was no attempt to compact the waste and no attempt to
cover it either. As a result, the waste absorbed all the rain that fell
on it, as well as the runoff from the dumping platform. The waste
Figure 23: Progression of failure of Dona Juana landfill (Hendron, etal, 1999).
11
The zone of the landfill that failed (see Figure 23) was lined with
a 1mm PVC geomembrane resting on either compacted clay or in
situ soils. A sand drainage layer and a protective soil layer were
above the liner. A horizontal soil cover layer was provided on top
of each 2.5m lift of compacted waste, while the lifts of waste
between cover layers were interconnected with rock-filled drains
to allow leachate to percolate downwards to the drain above the
liner. There was also a passive gas venting system consisting of
vertical perforated pipes on a 50m grid. A leachate recirculation
system was installed consisting of horizontal perforated pipes
placed on top of each waste lift before placing the cover layer.
The object of this piping was to inject leachate, collected from
the base of the landfill, back into the waste, so as to operate the
landfill as a biological waste reactor, thus purifying the leachate
before releasing it into the nearby river.
The investigation of the failure concluded that it had been
triggered by high liquid pore pressures caused by the re-injection
of leachate. The zone that failed was the only zone where
leachate recirculation had been applied. The design stability
analysis had assumed that no pore pressures would occur in the
waste. The inset on Figure 23 shows how the calculated factor of
safety for the failed section must have declined as the waste
thickness increased during the initial 22 months prior to the start
of leachate injection (Caicedo et al 2002). The additional pore
pressures caused by re-injection caused the already low factor of
safety to fall to 1.0 and the failure followed.
The failure investigation reached the obvious conclusion that
when designing a landfill where leachate is to be re-circulated,
pore pressures must be properly evaluated and their effect must
be considered in the stability analysis
(1)
(1a)
(1b)
(2)
In equation (1)
W = [2h - L(tan - tani)? L/2
and
(3)
(5)
12
Table 1: 22 flow failures of mine waste tips (or dumps), tailings dams and municipal solid waste landfills that have resulted in deaths,
major environmental damage, or major damage to structures and infrastructure
(Note: Entries have been selected, list is not comprehensive)
Year &
Number
Volume of
Flow
Location
Waste
Cause of Failure
1928
(1)
Barahona, Chile
copper tailings
8.2 Richter
earthquake
environmental
devastation
1961
(2)
Jupille,
Belgium
fly ash
removal of toe
support of dump
11 deaths, houses
destroyed
1965
(3)
El Cobre (2
impoundments)
copper tailings
7.5 Richter
earthquake
1) 1.9 x 106m3
2) 0.5 x 106m3
fine tailings
1966
(4)
Aberfan, UK
coal waste
dumping of waste
over spring
1970
Mufulira
copper tailings
(5)
Zambia
collapse of tailings
dam into workings
1972
(6)
Buffalo Creek,
USA
coal waste
overtopping of waste
impoundment
1974
(7)
Bafokeng,
South Africa
platinum tailings
overtopping of
tailings dam
13 deaths, extensive
damage to mine
installation and
environment
1978
(8)
Mochikoshi,
Japan
gold tailings
7.0 Mercalli
earthquake
80 x 103m3 fine
tailings
environmental
devastation
1985
Stava, Italy
fluorite tailings
shear failure of
retaining dyke
1985
(10)
Quintette
Mamot, BC,
Canada
coal waste
pore pressure
resulting from
collapse settlement
2.5 x 106m3
environmental
damage - river valley
filled with waste for
2.5km
1993
(11)
Istanbul,
Turkey
(UmraniyeHekimbasi)
municipal solid
waste
shear instability of
uncompacted waste
1.2 x 106m3
39 deaths, 11 houses
destroyed, main
sewer fractured,
sewer flow dammed
by slide debris
1993
(12)
Saaiplaas,
South Africa (3
failures in 3
days)
gold tailings
minimal
environmental
damage. Not
reported by news
media
1994
(13)
Merriespruit,
South Africa
gold tailings
overtopping of
tailings dam
17 deaths, extensive
damage to housing
and environment
1995
(14)
Omai, Guyana
gold tailings
piping erosion of
retaining dyke
80km of river
devastated
1995
(15)
Surigao del
Norte,
Philippines
gold
dyke failure
50 x 103m3
12 deaths, coastal
pollution
1996
(16)
Sgurigrad,
Bulgaria
overtopping of
retaining dyke
220 x103m3
107 deaths,
environmental
devastation
(9)
13
Consequences
89 miners killed
underground
Year &
Number
Location
Waste
Cause of Failure
Volume of
Flow
1997
(17)
Bogota,
Colombia
municipal solid
waste
800 x103m3
river dammed by
debris
1997
(18)
Durban, South
Africa
municipal solid
waste
160 x 103m3
slide contained
within boundary of
site
1998
(19)
Los Frailes,
Spain
foundation failure of
tailings dam
4 x 106m3 slurry
environmental
devastation
1999
gold
Tailings slurry
escaping from burst
pipe
700 x 103m3
(20)
Surigao del
Norte,
Phillippines
17 houses destroyed,
agricultural land
devastated
2000
(21)
Inez, Kentucky,
USA
coal wastes
950 x103m3
120km of rivers
devastated by slurry
2000
(22)
Manila,
Philippines
municipal solid
waste
Consequences
shear failure
not known
minimum of 218
following heavy
deaths
typhoon rains
At least 1 400 deaths in 72 years (a maximum of perhaps 20 per year) compared with thousands of millions killed by war, disease,
famine, traffic accidents, etc. in the same period.
Table 1 was drawn from a number of sources, most of which appear in the reference list. For post 1991 failures, the list given by
Fahey et al (2002) has been useful.
Table 2: Summary of observed post-failure surface slopes and
corresponding ground/tailings interfacial shear strengths for
flow failures in tailings impoundments
Tailings dam
Post failure
surface
slope
Ground
slope
i
At rest
interfacial
shear
strength, t
(kPa)
Bafokeng (Figure 2)
Bafokeng (Figure 3)
Arcturus
Saaiplaas (Figure 9)
(After rain)
(No rain)
(No rain)
Merriespruit (Figure 4)
(Flow slide)
(Failure basin)
4
2
3
1.5
1.3
1.5
5.2
1.6
2.6
3
2.3
3
1
-0.5
-0.5
2.3
3.4
3.6
2
2
1.5
0
1.0
1.8
(6)
the failure that occurred after rain was 65% of that corresponding
to flows over a dry ground surface. At Merriespruit, the fugitive
tailings flow over wet ground had an interfacial shear strength of
55% of that of the final surface of the failure basin. Figures 24b
and c show some data on the shear strength required for stability
(zero acceleration) on various ground slopes (b) and also the
acceleration that will occur if these shear strengths are not met
(c). The data correspond to a simple case in which the surface of
the flowing waste is parallel to the ground surface, but via
equation (2) give some idea of the speed with which a flow slide
can move. For example, if the acceleration from rest is only
0.1ms -2 and this is maintained for 1 minute, the flow will
accelerate to 6ms -1 or 20kmh-1 in this period. The consequences
of higher rates of acceleration are frightening. In the flow failure
at Bafokeng, the flow velocity a short distance after leaving the
breach in the impoundment was estimated from stagnation flow
heights on damaged buildings (by equating the potential energy
14
?
?
?
?
Side-hill dumps are often opted for because the top of a ridge
may be easily accessible, and dumping can proceed by building
out a horizontal platform using edge-tipping with gravity to
transport the waste down the hill, over the "wasteland". This was
the reason for the choice of the Istanbul site and several others
like it, as well as the Quintette Marmot site.
The Durban Bulbul landfill (18, Table 1) was sited in a steepsided valley. This caused seepage from the hillside to be directed
towards the waste body in addition to providing a steep base for
the landfill to rest on.
Siting of waste deposits in swampy areas has been the root
cause of many failures (e.g. Blight 1997). In 1970 a tailings dam
collapsed into underground workings in Zambia, trapping and
killing 89 miners in the workings, and this was also the cause of
the failure at Inez, Kentucky (21, Table 1) in 2000. The
Bafokeng tailings dam was sited with one of its outer dykes on
the bank of a dry valley, and it was the presence of this valley,
carrying water after rain, that caused the 42km long flow of the
escaped tailings.
5.2 Design
Because of the long-term nature of waste deposition operations,
and because the characteristics of the waste will inevitably
change during the deposit's operating life, pre-construction
designs are really site preparation designs, based on available
knowledge of the waste characteristics. Design for stability must
be reviewed and, if necessary, revised once the installation is
operating, waste has been deposited and its in situ properties
have been measured. It is also quite likely that the envisaged
method of deposition will prove unsuccessful or unsuitable and
will have to be changed. For example spigot deposition of
unthickened tailings from a ring delivery main may be replaced
by paddock deposition or thickened tailings, or placing of dry
waste by mechanical stacker may be replaced by spreading from
bottom-dump trucks.
However, to avoid failure of a (suitably sited) waste deposit,
and in particular, failure resulting in a destructive flow, the
design should provide for:
5.1 Siting
Many waste deposits whether of hydraulic fill tailings, "dry"
mine waste or municipal solid waste are sited in positions that
invite the occurrence of disasters. Examples are the Jupille,
Aberfan and Quintette Marmot waste dumps (2, 4 and 10, Table
1), the El Cobre, Mochikoshi, Stava and Merriespruit tailings
impoundments (3, 8, 9 and 13, Table 1), all of which were sited
on hillsides or hill crests above villages, the Bafokeng (7, Table
1) tailings impoundment, sited 200m from an unprotected mine
shaft and the Istanbul MSW dump (11, Table 1) sited on the crest
of a very steep slope. These are obviously unacceptable sites for
structures of this type. In all likelihood, most of these sites were
chosen for reasons of cost saving, or to use land that was
regarded as waste land, unsuitable for any other use.
15
5.3 Operation
5.4 Review
Reviewing and measurements of at least the following should be
made six-monthly:
? Properties of wastes disposed (grading, shear strength,
consolidation parameters of the waste for mine wastes and
composition for municipal solid waste).
? Properties of wastes as placed (slurry density, beach slopes and
gradings down the beach, in situ shear strength and dry
density for tailings, in situ densities and water contents for
dry mine wastes, in situ densities for MSW).
? Dimensions of deposit (slope angles, heights and rates of rise).
? Effluents from deposits (quantities and rates of flow for return
water from tailings dams, rates of flow of leachate from
landfills, seepage from all waste deposits, erosion from
slopes).
? Weekly maximum pool levels and minimum freeboards.
? Weekly return water reservoir or leachate pond levels.
? Measurements from instruments (pore pressure, settlement,
movement of slopes).
? Meteorological data, rainfall, evaporation, wind speed and
direction.
? Seismic data (whether natural or seismically induced.)
? A detailed site inspection by an independent engineer or panel
of engineers.
The design should then be reviewed by the engineer or
engineering panel in the light of the current design for the waste
deposit, including reviews of:
? the water balance for the deposit;
? the stability of the slopes in terms of geometry, height, rate of
rise, in situ shear strength and results of instrument
measurements;
? minimum free boards and maximum return water reservoir
levels.
Any deficiencies in the performance of the deposit or its
operation must then be corrected immediately, and the
corrections reported at the next review. If and where necessary,
amendments must be made to the design and to the operating
manual, for immediate implementation.
REFERENCES
Anonymous 1967. Report of the tribunal appointed to enquire
into the disaster at Aberfan. HMSO, London, UK.
Bishop, A.W. 1973. The stability of tips and spoil heaps.
Quarterly Journal of Engineering Geology, 6, 3 & 4: 335-376.
Blight, G.E. 1979. Editorial: The disposal of mining and
industrial waste. The Civil Engineer in South Africa, June: 133.
Blight, G.E. 1990. The effect of dynamic loading on underground
fill in South African gold mines. De Mello Tribute Volume,
Editoria E. Blucher, Sao Paulo, Brazil: 37-44.
6. CONCLUDING REMARKS
Tailings impoundments, dry mine waste dumps and landfills are
different from natural slopes in that they all are, or should be
16
17