Available online at www.sciencedirect.

com

Journal of Hazardous Materials 154 (2008) 79–87

Floods from tailings dam failures

M. Rico a,∗ , G. Benito b , A. Dı́ez-Herrero c
a
CSIC - Instituto Pirenaico de Ecologı́a, Zaragoza, Spain
bCSIC - Centro de Ciencias Medioambientales, Madrid, Spain
c Research Unit on Geological Hazard and Risks, Instituto Geológico y Minero de España (IGME), Madrid, Spain

Received 6 September 2006; received in revised form 26 September 2007; accepted 27 September 2007
Available online 2 October 2007

Abstract
This paper compiles the available information on historic tailings dam failures with the purpose to establish simple correlations between tailings
ponds geometric parameters (e.g., dam height, tailings volume) and the hydraulic characteristics of floods resulting from released tailings. Following
the collapse of a mining waste dam, only a part of tailings and polluted water stored at the dam is released, and this outflow volume is difficult
to estimate prior the incident. In this study, tailings’ volume stored at the time of failure was shown to have a good correlation (r2 = 0.86) with
the tailings outflow volume, and the volume of spilled tailings was correlated with its run-out distance (r2 = 0.57). An envelope curve was drawn
encompassing the majority of data points indicating the potential maximum downstream distance affected by a tailings’ spill. The application of the
described regression equations for prediction purposes needs to be treated with caution and with support of on-site measurement and observations.
However, they may provide a universal baseline approximation on tailing outflow characteristics (even if detailed dam information is unavailable),
which is of a great importance for risk analysis purposes.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Tailings pond; Mine waste; Dam failure; Dam break; Risk analysis

1. Introduction during emplacement, construction and operation of the dam, and

(5) the high cost of remediation works following the closure of
Dams are structural barriers built mainly for water man- mining activities.
agement (for example, irrigation, hydroelectric power and/or Ever since the earliest dams were built, there have been
flood control) or the storage of industrial and mineral process- dam failures. However, most studies of dam-break floods have
ing waste. Tailings dams are a particular type of dam built focused on water-storage dams, with only a few exceptions
to store mill and waste tailings from mining activities. Cur- [1–4]. Tailings dam failures result from a variety of causal
rently, thousands of tailings dams worldwide contain billions mechanisms (e.g., flooding, piping, overtopping, liquefaction,
of tonnes of waste material from mineral processing activity or a combination of several) spilling out polluted water and
at mine sites. A number of particular characteristics make tail- tailings with a variety of textural and physical–chemical prop-
ings dams more vulnerable to failure than water storage dams, erties, which may impact over the downstream socio-economic
namely: (1) embankments formed by locally derived fills (soil, activities and ecological systems. A good example of the high
coarse waste, overburden from mining operations and tailings); socio-economic cost associated with tailings dam disasters is
(2) multi-stage raising of the dam to cope with the increase in the Los Frailes (Aznalcollar) accident, a large scale sulphide
solid material stored and effluent (plus runoff from precipitation) tailings pond spill that occurred in April 1998, with ca. D 152
released; (3) the lack of regulations on specific design criteria; million in socio-economic losses [5–7]. About D 147 million
(4) dam stability requiring a continuous monitoring and control was spent to correct the negative environmental and agricultural
impacts, including restoration of the area’s natural resources
(average impact of 5.7 × 106 D /km2 ) and ca. D 5 million was
∗ Corresponding author. Tel.: +34 976 716118; fax: +34 976 716019. dedicated to mitigate socio-economic and socio-labor impacts
E-mail address: mayterico@ipe.csic.es (M. Rico). in the affected municipalities. In addition, uncountable impacts

0304-3894/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhazmat.2007.09.110
80 M. Rico et al. / Journal of Hazardous Materials 154 (2008) 79–87

affected the region’s production structure which produced a drop and the United Nations Environmental Programme [15] (last
in sales from milk producers, farmers and fish industries. updated 4 March 2006) in which a selection of 83 major tailings
To date, most numerical models for dam-break analysis have dam failures was compiled. The analysis of tailings dam per-
been developed for water-storage dams. The purpose of these formance provides important knowledge on key design factors
models has been to predict the flood characteristics (flood hydro- on dam stability [12], including on-site characteristics (geol-
graph, peak discharge, flood wave propagation time, etc.), this ogy, seismicity, climate, upstream catchment area), selection of
depending upon dam type (for example, earth, rockfill, concrete embankment and construction sequence type, as well as hazard
gravity, concrete arch, etc.), break mechanisms and breach size. assessment (heavy rain, flooding, earthquake). Empirical models
A great effort is still needed, however, to establish a reliable of dam failure impact and tailings outflow movement are more
general methodology for coping with hazard prediction from difficult to establish in a universal way due to the severe lack
tailings dam failures, which may serve to classify tailings ponds of reliable data from previous incidents and high levels of com-
according to their potential downstream damages. In tailings plexity on dam-break mechanisms, breach size, water-tailings
dam accidents, flow numerical models need to account for high concentration and spilled volume.
sediment concentration floods. Jeyapalan et al. [2] applied a The e-EcoRisk database (e-EcoRisk—A Regional Enterprise
Bingham plastic model (TFLOW computer program) both for Network Decision-Support System for Environmental Risk and
the Aberfan case and the Gypsum Tailings Dam incident (case Disaster Management of Large-Scale Industrial Spills) was fed
12 in Table 1) showing laminar flow behavior of the tailings with the information mentioned above, after a process of revi-
flood. sion, cross-checking and updating of the data using existing
Apart from complex hydraulic calculations applied to specific bibliographic sources. During this process, a detailed literature
case studies, more simple estimations can be performed based on review was conducted to gather as much information on tailings
generic empirical relationships. In these equations, key hydro- dam failures as possible. New data were added and informa-
logical parameters associated with dam failures (e.g., outflow tion gaps were completed using different data sources such as
volume, peak discharge, mine waste run-out distance) can be journals, conference proceedings, reports, published and unpub-
estimated from pre-failure physical characteristics of the dam lished dissertations and web pages [2–4,16–28, among others].
(dam height, reservoir volume, etc.), based on reported historic Data from one of the European dams were obtained by interview
dam failures. This approach has been successfully applied to with one dam failure witness and mine worker. The compilation
estimate peak discharge and flood volume resulting from water- of data from different European countries was achieved through
retention dam failures [8–10]. With tailings dams, empirical the collaboration of the e-EcoRisk partners.
relationships are very limited probably due to the scarcity of The reports on tailings dam failures are incomplete and heav-
reliable historical data. Lucia et al. [11] proposed a method to ily biased. There is no (complete) worldwide database of all
estimate the potential mine waste run-out distance (at slopes historical failures. This database is, therefore, a subset of the
<4◦ ), based on historical tailings dam failures and using the value actual number of tailings dam incidents in the world. The major-
of the residual strength of liquefied tailings. Although these ity of tailings dam incidents remain unreported, especially in
are based on simple empirical equations, the model requires developing countries. In cases where a known accident did occur
some detailed geotechnical data of the material contained in the it is often difficult to access basic information regarding the tail-
tailings ponds as well as on the geometry of the downstream val- ings dam and its condition prior to the incident (e.g., dam height,
ley, which are not always available or provided by the mining tailings storage volume, tailings thickness, water content, etc.).
companies. To date, 250 cases of tailings dam failures in the world have been
The aim of this paper is to propose a set of simple empirical compiled. For each case as much information as possible was
equations, to describe tailings pond characteristics and outflow sought and documented. In the majority of cases, the information
hydraulic behavior, based on data from historic dam failures. obtained is scarce; therefore, only 28 accidents (Table 1) with
Although accuracy of these estimations should be approached complete information on tailings outflow volume and flood run-
with great caution, the equations provide a first and universal out distance have been used in the correlation analysis presented
way for simple estimations on potential risk and impact related in this study.
to tailings dam breaks, using basic physical dam information.
3. Results
2. Methodology
In water-storage dams, sensitivity studies have indicated that
Several investigations have attempted to summarise the reservoir volume as well as dam height are critical factors in the
causes of major tailings dam failures throughout the world. The magnitude of dam failure hydrographs [9,29]. The simplest pro-
most recent and main synthesis was by ICOLD [12] (221 tailings posed relations involve those two parameters, considering that
dam incidents), based on the previous database by USCOLD most of the water volume stored at the reservoir was released.
[13], which collected a large amount of information on inci- In the case of tailings ponds, the tailings outflow volume from
dents in the USA (185 tailings dam incidents) that occurred dam incidents depends on the liquefaction process, break time,
during the period 1917–1989. This database has been supple- breach size and the amount of water content in the pond at the
mented by the U.S. Environmental Protection Agency [14] with failure time. In most failure cases, tailings ponds are never emp-
recent mining and mineral processing damage cases in the USA tied and, indeed, only a limited portion of the mine waste is
 M. Rico et al. / Journal of Hazardous Materials 154 (2008) 79–87 81

Table 1
Historical tailings dam failures used in the correlation analysis
Ref. no. Name Date of Type Dam Impoundment Run-out Dam factor Released
of the failure of dam height volume distance (H × VF ) volume
dam (year) (m) (×106 m3 ) (km) (×106 m3 )

1 Arcturus (Zimbawe) 1978 RING 25 1.7–2.0 Mt 0.3 0.5 0.0211

2 Bafokeng (South Africa) 1974 RING 20 13 45 60 3
3 Baia Mare (Romania) 2000 UPS 7 0.8 0.18 0.7 0.1
4 Bellavista (Chile) 1965 RING 20 0.45 0.8 1.4 0.07
5 Buffalo Creek (USA) 1972 UPS 14–18 0.5 64.4 7–9 0.5
6 Cerro Negro No.3 (Chile) 1965 UPS 20 0.5 5 1.7 0.085
7 Cerro Negro No.4 (Chile) 1985 MXSQ 40 2 8 20 0.5
8 Churchrock (USA) 1979 WR 11 0.37 96.5–112.6 4.07 0.37
9 Cities Service (USA) 1971 WR 15 12.34 120 135 9
10 El Cobre Old Dam (Chile) 1965 UPS 35 4.25 12 66.5 1.9
11 Galena Mine (USA) 1974 UPS 9 0.61 0.034 0.0038
12 Gypsum Tailings Dam (USA) 1966 UPS 11 7 Mt 0.3 0.88–1.43 2 × 105 t
13 Hokkaido (Japan) 1968 UPS 12 0.3 0.15 1.08 0.09
14 Itabirito (Brazil) 1986 Gravity 30 12 3 0.1
15 La Patagua New Dam (Chile) 1965 RING 15 5 0.525 0.035
16 Los Frailes (Spain) 1998 RING 27 15–20 41 53.51 4.6
17 Los Maquis (Chile) 1965 UPS 15 0.043 5 0.315 0.021
18 Merriespruit (South Africa) 1994 RING 31 7.04 2 18.6 2.5 Mt
19 Mochikoshi No.1 (Japan) 1978 UPS 28 0.48 8 2.24 0.08
20 Mochikoshi No.2 (Japan) 1978 UPS 19 0.15 0.057 0.003
21 Ollinghouse (USA) 1985 WR 5 0.12 1.5 0.125 0.025
22 Omai (Guyana) 1995 WR 44 5.25 80 184.8 4.2
23 Phelps-Dodge (USA) 1980 UPS 66 2.5 8 132 2
24 Sgurigrad (Bulgaria) 1966 UPS 45 1.52 6 9.9 0.22
25 Stancil (USA) 1989 UPS 9 0.074 0.1 0.342 0.038
26 Stava (Italy) 1985 RING 29.5 0.3 4.2 5.605 0.19
27 Tapo Canyon (USA) 1994 UPS 24 0.18
28 Unidentified (USA) 1973 UPS 43 0.5 25 7.31 0.17
29 Veta del Agua N◦ 1 (Chile) 1985 MXSQ 24 0.7 5 6.72 0.28

RING: ring dyke; WR: water retention; UPS: dams subsequently raised upstream; MXSQ: dam comprising different raising typology (upstream, centreline and
downstream); H: dam height; VF : volume of tailings released.

released (see Table 1). Tailings outflow volume seems a more

appropriate means to establish correlations with flood related
parameters instead of reservoir volume. In tailings ponds, the
height difference between the crest of the dam and the decant
surface, known as freeboard, is rather small, and essentially the
dam height equals the thickness of the tailings deposit. Accord-
ingly, dam height provides a good approximation of the tailings
thickness and its potential energy during the dam failure.
In Fig. 1, a plot of dam height (H) of the known historical
tailings dam failures versus the outflow run-out distance (Dmax )
is shown. The graph shows a great dispersion of data and a poor
relationship between these variables, described by the following
regression equation:

Dmax = 0.05 × H 1.41 , r2 = 0.16 (1)

The envelope curve for run-out distance from all the tailings Fig. 1. Graph showing the dam height (H in meters) at time of failure, vs. the
dams has the equation: run-out distance of tailings following dam breach (Dmax in km) for the historical
failure cases, with numbers keyed to Table 1.
Dmax = 0.01 × H 3.23 (2)

This poor correlation shows that run-out distance depends A better relationship was obtained using mine waste outflow
on other factors not considered in the equation, such as outflow volume (VF ), as the independent variable (note that VF includes
mine waste volume, gradient and topography to which tailings tailings and water); although great data dispersion is still present
flood debouches. (Fig. 2). A plot of these variables encompassing all the data
82 M. Rico et al. / Journal of Hazardous Materials 154 (2008) 79–87

run-out distance, and (2) above the line comprise those cases
with a longer downstream dispersion of tailings outflow. The
first group includes incidents with a high viscosity of the spilled
mine waste (cases 3, 12, 13); a limitation on the tailings extent
due to obstacles, topographic restrictions and barriers (cases 1,
23, 25); and those with very low slope gradient downstream of
the pond. This is the case of Gypsum Tailings Dam (case 12)
emplaced over a flat surface, containing non-plastic silt tailings
with an average field water of about 30% [2]. In the case of
Merriespruit dam, South Africa (case 18), a tailings spill of wet
materials traveled 2 km before being halted and contained by
an ornamental lake [30]. Other examples of a tailings release
with short run-out distance is the Stancil dam (USA, case 25)
where the tailings flow slide was blocked at a creek near the
embankment toe [12].
The second cluster of data points (above regression line in
Fig. 2. Graph showing the tailings outflow volume due to tailings dam incidents Fig. 2) comprises those waste ponds which stored a larger vol-
(VF , in millions of m3 ) vs. the run-out distance of tailings from historical failure
ume of water with the tailings as well as those failures related to
cases (Dmax in km), with numbers referred in Table 1.
heavy rains and dam overtopping (cases 5, 14, 15, 17 in Table 1).
Some examples include La Patagua New Dam (case 15) and Los
points from Table 1 except case 8 (Churchrock, USA) provided
Maquis (case 17), both in Chile, where water contents of the tail-
the following regression equation:
ings reached 70 and 85%, respectively [19]. A different case is
Dmax = 14.45 × VF0.76 , r2 = 0.56 (3) incident 28 (Unidentified, USA; Table 1) in which an anomalous
tailings outflow volume (about one-third of the impoundment
with a standard error (S.E.) of 0.62%. The envelope curve volume [12]) was reported, which apparently was not related to
obtained for this data set follows the equation: the large water volume stored behind the dam.
Dmax = 112.61 × VF0.81 (4) Hagen [9] and the Committee on the Safety of Existing Dams
[31] developed a criterion for water-storage dams to estimate
Data dispersion may be explained by the water content at the peak discharge based upon the product of the dam height (H)
decant pond at the time of failure, the water/tailings ratio at the and reservoir volume (V), also known as dam factor (H × V).
run-out flood, as well as the drainage slope. This factor is a crude index of the energy expenditure at the dam
The Churckrock dam (case 8) was an earthfill dam type when it fails [10]. In most of the tailings dam incidents only a part
located at a uranium mine. A major difference with other tail- of the pond volume was released, which is a major difference
ings dams is that the Churckrock dam retained only clear water in behavior with water-storage dam accidents. Therefore, for
and the outflow peak following the failure moved downstream tailings dams a better relationship was found when dam factor
a larger distance in relation to the dam height and outflow vol- was based upon the tailings spilled volume (VF ). In Fig. 3, a plot
ume than conventional tailings dam failures. This indicates that of a “tailings dam” factor (H × VF ) versus the outflow run-out
this spill behaved in the manner of a water-storage dam failure
and, accordingly, it was neither used in the regression equations
(Eqs. (3) and (5)) nor on the envelope curves (Eqs. (4) and (6)).
Other historic mine waste dam failures referred to in biblio-
graphic sources as water-storage dams (WR in Table 1) at mine
complexes have been marked on the different plots. These cases
include dams from Cities Services (USA), Ollinghouse (USA)
and Omai (Guyana) (cases 9, 21 and 22 in Table 1). Contrary to
the Churchrock dam, these examples did contain tailings and so
these case studies were included in Fig. 2.
In Fig. 1 (outflow run-out distance versus dam height) tailings
ponds of water retention type (incident cases 8, 9, 21, 22 in
Table 1) or containing a large volume of water at the failure time
(incident cases 2, 5) are plotted close to the envelope curve,
whereas those tailings ponds with lower water content plot in
the lower part of the graph (incident cases 1, 13, 20, 25, 27 in
Table 1). In Fig. 2 (outflow run-out distance versus mine waste
outflow volume) data points are represented around a regression Fig. 3. Graph showing “tailings dam” factor (H × VF ; H in meters and VF in
line which separate two data groups: (1) below the regression line 106 × m3 ) vs. run-out distance of tailings (Dmax in km) for the historical failure
data points correspond to incidents with a shorter tailings outflow cases, with numbers keyed to Table 1.
 M. Rico et al. / Journal of Hazardous Materials 154 (2008) 79–87 83

Fig. 5. Graph the dam height (H in meters) at time of failure, vs. peak discharge
Fig. 4. Graph showing the tailings outflow volume from the tailings dam (VF in
(m3 s−1 ) for constructed and landslide dams (dates from Costa [10]) and tailings
106 × m3 ) vs. the volume of tailings stored at the dam (VT in 106 × m3 ) at the
dams (case 5: Buffalo Creek, USA; case 16: Los Frailes, Aznalcollar, Spain).
time of the incident.

distance (Dmax ) is shown. The regression line including all the relationships obtained using dam height (H), water volume con-
data except case 8 (Fig. 3) is the following: tained at the reservoir (V) and the resulting flood peak discharge
Dmax = 1.61 × (HVF )0.66 , r2 = 0.57 (5) (Q) when it failed (see Figs. 5 and 6). A major contribution
of the Costa [10] paper is the analysis of floods from natural
where H is dam height in meters and VF is waste outflow volume dams including ice dams (water impounded within or behind
in 106 × m3 . glacial ice), moraine dams, volcanic flow dams and landslide
An envelope curve encompassing all the historical tailings dams. Landslide dams typically involve large volumes of sed-
dam failures except case 8 has the equation: iments producing wide abutments, which are partly eroded by
water during dam overtopping and breaching [34]. According
Dmax = 12.46 × (HVF )0.79 (6)
to Costa [10] this explains that flood peaks from failed land-
A major limitation to the application of this equation to risk slide dams appear to be smaller than constructed dam failures
analysis of standing tailings dams consists of the uncertainty with the same dam height and reservoir volume (Figs. 5 and 6).
of potential tailings outflow volume (VF ) in the case of failure.
Another empirical relationship (Fig. 4) was found between the
tailings storage volume and the tailings released at the incidents
(using incident cases of Table 1), which has the equation:
VF = 0.354 × VT1.01 , r2 = 0.86 (7)
The above equation shows, that in average, one-third of the
tailings and water at the decant pond is released during dam
failures. The envelope curve represents the maximum tailings
volume that can be released in the most extreme situation in
which pond volume was emptied following the dam break, as
is the case of water-storage dam accidents or those of industrial
(diluted) waste ponds.
Available historic data on peak discharges from failures of
tailings dams are scarce. In this study, only three cases have been
found: (1) from Buffalo Creek (USA; case 5), with an estimated
discharge of 1415 m3 s−1 using the slope-area method [10]; (2)
from Los Frailes (Spain; case 16), with a recorded discharge Fig. 6. Graph the dam factor (H × V) vs. peak discharge (m3 s−1 ) for constructed
of 811 m3 s−1 , using a rating curve obtained for clear water at and landslide dams (data from Costa [10]) and tailings dams (Case 5: Buf-
a gauge station located 11 km downstream of the tailings dam falo Creek, USA; Case 16: Los Frailes, Aznalcollar, Spain (numbers keyed to
[32], and (3) from McLauren gold Mine (USA) [33], with an Table 1). Dam factor (H × V) for Los Frailes was plotted using both the total
associated discharge of ca. 200 m3 s−1 , estimated using the dam tailings volume at decant pond (case 16 point 2) and the tailings outflow volume
(case 16 point 1). In the Los Frailes incident, a discharge-dam factor relation-
factor/discharge equation proposed by Costa [10]. ship similar to landslide dams is obtained from a dam factor calculated from the
On the basis of historical water-storage (constructed and spilled tailings volume instead of the total dam tailings volume which is used
natural) dam breaks, Costa [10] proposed several empirical on water-retention dam failures relationships.
84 M. Rico et al. / Journal of Hazardous Materials 154 (2008) 79–87

Similarly, when a tailings dam is overtopped by runoff excess the total tailings volume contained at the reservoir when dam
from the drainage basin, there is commonly large water content failed.
for tailings sediment before a full breach is developed which
would be added to the tailings outflow volume. A tailings flood 4. Discussion
is commonly composed of water with high sediment concentra-
tion which provides a wide range of fluid behavior from debris In this paper, empirical relationships of mine decant pond
flow to muddy floodwater. This wide range of flow behavior is geometric characteristics with hydrologic parameters of floods
reflected in the representation of the few available historical data from mine waste spills from historic dam failures were obtained.
points (Figs. 5 and 6), which corresponds to Buffalo Creek (case These relationships provide a first approximation to estimate
5) and Los Frailes (case 16). On the one hand, Buffalo Creek the tailings outflow volume and the mine waste run-out dis-
dam incident plots on the regression line corresponding to water- tance from a dam failure, although these assessments may
storage dams due to a dam failure caused by flow overtopping in contain large uncertainties considering the high standard errors
relation to extreme rainfall. The run-out flood contained a mix- of the regression equations. These errors result from a large
ture of water and carbon residua with a turbulent flow behavior variety of parameters affecting the mine waste flow, including
[2], which indeed plots on the regression line of incidents, related sediment load, fluid behavior (Newtonian or Bingham-plastic)
to constructed earthfill dams for water-storage. Regarding Los which depends on the type of failure (e.g., seismic action,
Frailes tailings dam incident, it was released a high viscous flow static liquefaction, slide, etc.), particle-dependent rheology of
containing a large proportion of tailings. The mine waste flood the suspension, topography and valley gradient and presence of
discharge was recorded at a gauge station located 11 km down- obstacles impeding the slurry to flow among others. Other source
stream of the tailings dam. The hydrograph shape comprised of uncertainty is related to the lack of data related with the water
two peaks, with the first one reaching 811 m3 s−1 and a sec- volume existing at the time of failure either stored at the decant
ond one of 294 m3 s−1 , according to a rating curve calibrated pond or linked to the meteorological causes triggering the dam
for water flow, which probably overestimated the mine waste failure (intense rainfall, hurricanes, rapid snowmelt, ice accumu-
discharge due to its higher viscosity [32]. The first hydrograph lation in the tailings dam, etc.), which may change indeed the
peak was composed of a water-dominant flow, mainly from the hydrologic conditions (peak discharge, tailings outflow volume)
upper water-laid layer at the reservoir, whereas the second con- and the run-out distance of the tailings. Those conditions are site
tained a high tailings load in a high viscosity flow (Fig. 7). As and event specific, and therefore, difficult to be considered by
shown in Fig. 5, Los Frailes tailings dam incident (case 16) plots empirical relationships.
on the regression line obtained from historical failures of land- These uncertainties may be reduced considering the most
slide dams after Costa [10]. The discharge estimation from the similar case in the historical database. This analysis can also
landslide dam failure equation is 1189 m3 s−1 , which is a simi- be used as a deterministic tool to provide information on the
lar order of magnitude to the 811 m3 s−1 recorded at the gauge largest peak discharge and/or potential distance of the tailings
station for the first hydrograph peak [32]. flood by using the proposed envelope curves. Note that even
Fig. 6 shows the empirical relationship between dam fac- in this case, a greater distance may be reached if the released
tor and peak discharge of historical dam failures of constructed tailings is mixed and diluted within a river water flow, as hap-
and natural dams by Costa [10], in which dam factor (H × V) pened in various historical tailings dam incidents. In spite of the
for Los Frailes tailings dam failure was plotted using both the described limitations, the method provides preliminary data on
total tailings volume stored at the decant pond (case 16 point the tailings outflow hydraulic characteristics using basic avail-
2) and the outflow volume (case 16 point 1). It is shown that able tailings dam and embankment parameters, which may be
for tailings dam failures more robust results are obtained con- used, for instance, to classify the failure risk associated to a large
sidering the spilled tailings volume than the ones applying population of tailings dams.
These equations were applied on two case studies of the
e-Ecorisk project (Table 2): (1) Los Frailes dam (Aznalcol-
lar, Spain, case 16, Table 1) which failed in April, 1998, and
(2) the inactive Ashes Dam at the Almagrera mine (Spain), an
example of an un-breached tailings dam. On 25 April 1998,
the rupture of the dam containing the tailings reservoir of the
pyrite mine of Aznalcollar caused a spill of 4.5 × 106 m3 of
acid water and pyrite ore. The resulting mine waste, contain-
ing high concentrations of heavy metals (mean particle size
12 ␮m), was of a solid content of some 0.88 × 106 m3 and
2.97 × 103 Mg mass [6]. The accident provoked the inundation
of the floodplains of the rivers Agrio and Guadiamar (aver-
age stream gradient of 0.00146) and was close to affecting
Fig. 7. Slurry flood hydrograph of the Los Frailes spill in the El Guijo gauge
station, 11 km downstream of the Los Frailes tailings dam, in the Guadiamar the Guadalquivir marshlands posing a substantial threat to the
River. The hydrograph comprises two peaks related to a first water pulse and a ecosystems of the Doñana Natural and National Parks. The
second tailings flow (after Ayala-Carcedo [32]). tailings flood led to the sedimentation of pyrite contaminated
 M. Rico et al. / Journal of Hazardous Materials 154 (2008) 79–87 85

Table 2
Tailings dam characteristics of the selected examples, and channel slope of the major river connected to the tailings pond

Dam name Dam Height (m) Tailings volume at decant pond (×106 m3 ) Average slope downstream (m m−1 )

Los Frailes Dam (Spain) 27 15–20 0.00146

Almagrera Ashes Dam (Spain) 34.5 2.8 0.00154

Table 3
Tailings volume for different dams, and observed spilled volume and distance run-out of tailings in relation to the Los Frailes (Aznalcollar) dam incident

Dam name VT (×106 m3 ) VF observed VF from Eq. D observed (km) D’ (km) estimated from Eq. (5) using:
(×106 m3 ) (7) (×106 m3 )
VT VF

Los Frailes Dam (case no. 16 in Table 1) 15–20 4.6* 5.4–7.3 41 82–99 42–51
Almagrera Ashes Dam 2.8 1 32 16

VT : total tailings volume at dam; VF : tailings outflow volume observed; D: run-out distance of tailings following dam breach. VF ’ and D’ Tailings outflow volume
(VF ’) and pollutant run-out distance (D’) estimated using Eqs. (7) and (5).
* Note that tailings spill volume includes acid water and pyrite ore.

waste [35,36] along the floodplain of the Agrio-Guadiamar to anticipate the volume of water that may be involved or the
system, covering 2616 ha from the failed dam to the start thickness of the tailings contained at the decant pond at time of
of the Guadalquivir marshlands. Heavy metal pollution (zinc, failure. In those cases, basic dam information (dam height and
lead, and cadmium among others; [37]) mostly affected the volume), which can be obtained from national dam databases
superficial soil layers (0–20 cm), although in some coarser grain- or by remote sensing analysis, may provide preliminary data to
sized soils pollution may penetrate down to at least 50–80 cm assess potential risk to downstream socio-economic activities
[38]. and environmental impacts.
The empirical regression equations proposed in this study
provide an estimation of mine waste outflow volume rang- 5. Conclusions
ing between 5.4 and 7.3 × 106 m3 (from Eq. (7); Table 3),
whereas the actual Aznalcollar incident released 4.6 × 106 m3 Dam failure impact depends on the tailings outflow travel dis-
which includes acid water and pyrite ore. The mine waste run- tance and path, and on the exposure and vulnerability of goods,
out distance estimated from Eq. (5) ranges between 42 and population, land use, water use and environmental values of the
51 km (Table 3). The actual run-out distance downstream of natural areas located downstream. The diversity of the tailings
the Guadiamar River was between 40 and 45 km (according to dam characteristics (dam type, dam situation, type of sequen-
Antón-Pacheco et al. [39]) which is similar to the empirically tially raised tailing dam, dam foundation, dam fill material, state
estimated tailings outflow travel distance. of activity, storage volume, tailings dam height, tailings’ den-
The Almagrera mine, the second test case, is within the same sity and pond water volume, among others), make any universal
metalogenetic province and mining context of the Aznalcollar prediction assessing dam failure impacts very speculative. In
mines. At the Almagrera mine, the ashes dam is 34.5 m in height addition, detailed risk assessments involve timely and costly
with an actual tailings volume of 2.8 × 106 m3 , although it was geotechnical, hydrological and hydraulic studies which can only
designed to store about 3.2 × 106 m3 of milled waste [40]. In the be completed with either or both the complicity of mining com-
case of dam failure, the proposed empirical regression lines show panies and political authorities.
an estimated run-out distance between 16 and 32 km (Table 3) for In this study, a worldwide database of historic failures was
present conditions (volume stored of 2.8 × 106 m3 ). The Alma- collected from which a regression analysis was carried out in
grera dam is connected with a tributary stream which joins the order to establish relationships and trends. The resulting empiri-
Odiel River (stream gradient 0.00154) ca. 3.5 km downstream. cal correlations among physical parameters (namely dam height
According to this estimated run-out distance, the tailings waste and pond volume) of the tailings reservoir and tailings flood
may reach the Odiel River and pollutants can be dispersed by the characteristics can be applied to provide a first estimate on the
river’s stream flow, in a similar way to the Aznacollar incident. volume of tailings spills, tailings run-out distance and, with the
The proposed equations should be considered as a first appropriate measurements and observations, even to assess the
approach in tailings dam risk studies, but more precise determin- outflow peak discharge. In these relationships, two possible solu-
istic models are required to provide precise site-by-site tailings tions have been provided: (1) a conservative estimation based
flood impacts. Errors may be large on extreme cases either for on the regression equations of the historic tailings dam failures,
tailings dam containing high water volume, or for hypercon- and (2) the worst case scenario estimation based on the envelope
centrated flows. Obstacles or barriers along the slurry pathway curves developed from historic dam failures from tailings dams
should also be considered. On the positive side, there is a lack and water retention dams.
of detailed data on most of the existing tailings dams of the Tailings outflow volume from a dam failure can be reason-
world. In addition, for operative tailings dams is not possible ably estimated if total tailings pond volume is known. Then,
86 M. Rico et al. / Journal of Hazardous Materials 154 (2008) 79–87

the estimated tailings spill can be used in combination with [9] V.K. Hagen, Re-evaluation of design floods and dam safety, in: Proceedings
the dam factor (height and volume) and the pond’s tailings of the 14th International Congress on Large Dams, vol. 1, Commission on
volume for prediction purposes on the estimation of run-out Large Dams Congress, Rio de Janeiro, 1982, pp. 475–491.
[10] J.E. Costa, Floods from dam failures, in: V.R. Baker, et al. (Eds.), Flood
distance of standing dams. Two case studies included in the Geomorphology, Wiley, New York, 1988, pp. 439–463.
European Commission e-Ecorisk project, namely the Los Frailes [11] P.C. Lucia, J.M. Duncan, H.B. Seed, Summary of research on case histories
(Aznalcollar) historic dam failure and the standing tailings dam of flow failures of mine tailings impoundments, in: Technology Transfer
at Almagrera (Ashes dam) complex have been discussed in Workshop on Mine Waste Disposal Techniques, Information Circular 8857,
view of the results provided by the proposed regression equa- U.S. Bureau of Mines, Denver, CO, 1981, pp. 46–53.
[12] ICOLD. Tailings Dams—Risk of dangerous occurrences, lessons learnt
tions. It is evident that the results need to be treated with from practical experiences, Bulletin 121, United Nations Environmental
caution, due to the uncertainty present in documentary evidence Programme (UNEP), Division of Technology, Industry and Economics
and the diversity of the tailings dams. However, the estimates (DTIE) and International Commission on Large Dams (ICOLD), Paris,
obtained for the Aznacollar dam failure provided similar values 2001.
in terms of tailings outflow volume and run-out distance to the [13] USCOLD, Tailings Dam Incidents, U.S. Committee on Large Dams
(USCOLD), Denver, CO, 1994.
field observations. In summary, the proposed regression equa- [14] EPA, Damage Cases and Environmental Releases from Mines and Mineral
tions provide a first approach to assess some risk parameters Processing Sites, U.S. Environmental Protection Agency, Office of Solid
at dams with a lack of information or prior to further detailed Waste, Washington, DC, 1997.
studies. [15] UNEP. Environmental and safety incidents concerning tailings dams at
mines: results of a survey for the years 1980–1996 by the Mining Jour-
nal Research Services, Report prepared for United Nations Environment
Acknowledgments Programme, Industry and Environment, Paris, 1996.
[16] S.C. Bourcy, R.E. Weeks, Stream morphology and habitat restoration of
Pinto Creek, Gila County, Arizona, in: Tailings and Mine Waste 2000,
The authors are very grateful to Varyl Thorndycraft (Royal
Proceedings of the Seventh International Conference on Tailings and Mine
Holloway, University of London) for the critical review of the Waste 2000, Fort Collins, Colorado, USA, 23–26 January 2000, Balkema,
original manuscript, and for his very useful comments and Rotterdam, 2000, pp. 467–475.
suggestions. This research has been funded by the European [17] D. Brink, The long-term repair of the Merriespruit tailings dam, in: Tailings
Commission through the project “A regional enterprise network and Mine Waste 1998, Proceedings of the Fifth International Conference
on Tailings and Mine Waste 1998, Fort Collins, Colorado, USA, 26–28
decision-support system for environmental risk and disaster
January 1998, Balkema, Rotterdam, 1998, pp. 953–957.
management of large-scale industrial spills”, e-Ecorisk Project [18] R.J. Chandler, G. Tosatti, The Stava tailings dams failure, Italy, July 1985,
(contract no. EVG1-2002-0068) and by the Spanish Ministry of Proc. Inst. Civ. Eng. 113 (2) (1995) 67–79.
Science and Education (HP2006-0072). [19] R. Dobry, L. Alvarez, Seismic failures of Chilean tailings dams, J. Soil
Mech. Found. Div. 93 (6) (1967) 237–260.
[20] A.B. Fourie, G. Papageorfiou, G.E. Blight, Static liquefaction as an
References explanation for two catastrophic tailings dam failures in South Africa,
in: Tailings and Mine Waste 2000, Proceedings of the Seventh Inter-
[1] J.K. Jeyapalan, J.M. Duncan, H.B. Seed, Summary of research on analyses national Conference on Tailings and Mine Waste 2000, Fort Collins,
of flow failures of mine tailings impoundments, in: Information Circular Colorado, USA, 23–26 January 2000, Balkema, Rotterdam, 2000,
8857, Technology Transfer Workshop on Mine Waste Disposal Techniques, pp. 149–158.
U.S. Bureau of Mines, Denver, CO, 1981, pp. 54–61. [21] L.A. Hansen, N.J. LaFronz, M.B. Yasin, Stabilization of the Pinto Valley
[2] J.K. Jeyapalan, J.M. Duncan, H.B. Seed, Investigation of flow failures of tailings impoundment slide, in: Tailings and Mine Waste 2000, Proceedings
tailings dams, J. Geotech. Eng. Div. 109 (2) (1983) 172–189. of the Seventh International Conference on Tailings and Mine Waste 2000,
[3] G.E. Blight, Destructive mudflows as a consequence of tailings dyke fail- Fort Collins, Colorado, USA, 23–26 January 2000, Balkema, Rotterdam,
ures, Proc. Inst. Civ. Eng. 125 (1997) 9–18. 2000, pp. 477–487.
[4] F.V.M. Wagener, H.J. Craig, G. Blight, G. McPhail, A.A.B. Williams, J.H. [22] R.N. Kostaschuk, J.M.T. Wilkinson, Predicting the deformation of tail-
Strydom, The Merriespruit tailings dam failure, in: Tailings and Mine Waste ings dams resulting from earthquake liquefaction, in: Tailings and Mine
1998, Proceedings of the Fifth International Conference on Tailings and Waste 1999, Proceedings of the Sixth International Conference on Tailings
Mine Waste 1998, Fort Collins, CO, USA, 26–28 January 1998, Balkema, and Mine Waste 1999, Fort Collins, Colorado, USA, 24–27 January 1999,
Rotterdam, 1998, pp. 925–952. Balkema, Rotterdam, 1999, pp. 237–248.
[5] CMA, El accidente minero de Aznalcóllar, Medio ambiente en Andalucı́a, [23] L.F. Harder, J.P. Stewart, Failure of Tapo Canyon tailings dam, J. Perform.
Informe 1998, Consejerı́a Medio Ambiente (Junta Andalucı́a), Sevilla, Const. Fac. 10 (3) (1996) 109–114.
1998. [24] G. McPhail, R.J. Stuart, D. Venter, The disaster of Merriespruit and its
[6] F. Gallart, G. Benito, J.P. Martı́n Vide, A. Benito, J.M. Prió, D. Regüés, consequences: remediation-making safe, in: Tailings and Mine Waste 1998:
Fluvial geomorphology and hydrology in the dispersal and fate of pyrite proceedings of the Fifth International Conference on Tailings and Mine
mud particles released by the Aznalcóllar mine tailings spill, Sci. Total Waste 1998, Fort Collins, Colorado, USA, 26–28 January 1998, Balkema,
Environ. 242 (1–3) (1999) 13–26. Rotterdam, 1998, pp. 959–965.
[7] G. Benito, A. Benito-Calvo, F. Gallart, J.P. Martı́n-Vide, D. Regüés, E. [25] D.J. Miller, Failure modes and effects analyses of potential ground move-
Bladé, Hydrological and geomorphological criteria to evaluate the disper- ments at Golden Sunlight Mine, in: Tailings and Mine Waste 1999,
sion risk of waste slurry generated by the Aznalcollar mine spill (SW Spain), Proceedings of the Sixth International Conference on Tailings and Mine
Environ. Geol. 40 (4–5) (2001) 417–428. Waste 1999, Fort Collins, Colorado, USA, 24–27 January 1999, Balkema,
[8] G.W. Kirkpatrick, Evaluation guidelines for spillway adequacy, in: The Rotterdam, 1999, pp. 335–344.
Evaluation of Dam Safety, Proceedings of the Engineering Foundation [26] A.D. Penman, Risk analyses of tailings dam construction, in: Conference
Conference, Asilomar Conference Grounds, Pacific Grove, California, Proceedings: Safety of Mining Dams Seminar, Gällivare, 20–21 Septem-
November 28-December 3, 1976, Am. Soc. Civ. Eng., New York, 1977, ber 2001, Swedish Mining Association, Natur Vards Verket, European
395-414. Commission, Technical Papers, 2001, pp. 37–53.
 M. Rico et al. / Journal of Hazardous Materials 154 (2008) 79–87 87

[27] H.J. Van Niekerk, M.J. Viljoen, Causes and consequences of the Mer- [34] K.L. Lee, J.M. Duncan, Landslide of April 25, 1974 on the Mantaro River,
riespruit and other tailings-dam failures, Land Deg. Dev. 16 (2) (2005) Peru, National Academy of Sciences, Washington, DC, 1975.
201–212. [35] E. López-Pamo, D. Barettino, C. Antón-Pacheco, G. Ortiz, J.C. Arránz, J.C.
[28] M. Willow, C. tenBraak, Survey of three hard-rock acid drainage treatment Gumiel, B. Martı́nez-Pledel, M. Aparicio, O. Montouto, The extent of the
facilities in Colorado, in: Tailings and Mine Waste 1999, Proceedings of Aznalcóllar pyritic slurry spill and its effects on soils, Sci. Total Environ.
the Sixth International Conference on Tailings and Mine Waste 1999, Fort 242 (1–3) (1999) 57–88.
Collins, Colorado, USA, 24–27 January 1999, Balkema, Rotterdam, 1999, [36] ITGE, Cartografı́a y cubicación de lodos mineros en la cuenca del rı́o
pp. 759–767. Guadiamar, Aznalcollar-Entremuros. Internal report, RM-001-98, Instituto
[29] A.W. Petrascheck, P.A. Sydler, Routing of dam break floods, Int. Water Tecnológico Geominero de España (ITGE), Madrid, 1998.
Power Dam Const. 36 (1984) 29–32. [37] J.O. Grimalt, M. Ferrer, E. Macpherson, The mine tailing accident in
[30] G.E. Blight, A.B. Fourie, A review of catastrophic flow failures of Aznalcollar, Sci. Total Environ. 242 (1–3) (1999) 3–11.
deposits of mine waste and municipal refuse, Introductory report [38] F. Cabrera, L. Clemente, E. Dı́az Barrientos, R. López, J.M. Murillo, Heavy
of the International Workshop on Occurrence and mechanisms of metal pollution of soils affected by the Guadiamar toxic flood, Sci. Total
flows in natural slopes and earthfills, IW-FLOWS2003, Sorrento, Italy, Environ. 242 (1–3) (1999) 117–129.
14–16 May 2003, Associazione Geotecnica Italiana (AGI), 2003, [39] C. Antón-Pacheco, J.C.Arranz, D. Barettino, G. Carrero, M. Giménez, J.A.
(http://www.unina2.it/flows2003/flows2003/articoli/articoli.htm). Gómez, J.C. Gumiel, E. López-Pamo, J.A. Martı́n Rubı́, B. Martı́nez Pledel,
[31] Committee on the Safety of Existing Dams, Safety of existing dams: evalu- E. De Miguel, J. Moreno, G. Ortiz, J.G. Rejas, A. Silgado, E. M. Vázquez,
ation and improvement, National Academy Press, Washington, DC, 1983. Actuaciones para el reconocimiento y retirada de los lodos depositados
[32] F. Ayala-Carcedo, La rotura de la balsa de residuos mineros de Aznalcóllar sobre el terreno, y su restauración edáfica y morfológica, Mapping and
(España) de 1998 y el desastre ecológico consecuente del rı́o Guadiamar: removal of the sludge deposited on land, Soil remediation and morpho-
causas, efectos y lecciones, Boletı́n Geológico y Minero 115 (4) (2004) logical restoration of the fluvial system, Boletı́n Geológico y Minero 112
711–738. (Special issue) (2001) 93–122.
[33] W.A. Marcus, G.A. Meyer, D.R. Nimmo, Geomorphic control of persis- [40] Knight Piésold Consultiong, Almagrera S.A. Normas de Seguridad Estruc-
tent mine impacts in a Yellowstone Park stream and implications for the tural y Plan de Emergencia, ITC 08.02.01 de 22 de Abril de 2000, Depósito
recovery of fluvial systems, Geology 29 (4) (2001) 355–358. de estériles, Depósitos de cenizas (Internal report), Huelva, 2001.