An Investigation into the Depth
and Rate of Weathering on Gold Tailings Dam Surfaces as Key Information
for Long Term AMD Risk Assessment
Report No. 1347/1/05
November 2005
EXECUTIVE SUMMARY
BACKGROUND
AND MOTIVATION
The oxidation of sulphide minerals in gold Fine Residue
Deposits (FRDs) of the Witwatersrand and its potential effect on water
resources is arguably one of the major regional strategic environmental
issues for the gold mining industry and the South African government.
Seepage from these FRDs tends to be alkaline to neutral during the
operational phase, but can become acidic in the post operational phase
as Acid Mine Drainage (AMD) processes become significant when the
piezometric surface subsides.
International research and field observations have shown that an
oxidised outer cap forms on FRDs, below which limited oxidation takes
place. Enough acidity may be produced in this outer cap to consume all
available neutralisation potential within an FRD body, however, as
oxidation slows, the risk associated with AMD decreases.
The key question is thus how the oxidised zone will develop over time
and how much acidity will be generated relative to available
neutralisation potential in the operational phase of an FRD. This will
determine the likelihood that a particular FRD will start generating
acidic seepage.
OBJECTIVES
A top down approach was used to evaluate practical methods
for evaluating the acidification risk associated with Witwatersrand
FRD’s. The study was guided by three study objectives,
including:
- To understand the relationship between the depth and rate
of weathering and the potential mass of acidity at gold FRDs through
empirical analysis.
- Develop procedures to conduct rapid accurate AMD risk
assessments of gold FRDs.
- Evaluate mitigation or risk reduction measures by making
use of the AMD risk assessment procedure and provide industry
guidelines.
SUMMARY OF FINDINGS
Characteristics
of FRD’s
Acid Mine Drainage (AMD) originates due to the combined
effects of flow or transport processes, and geochemical processes.
Transport processes include transport of liquids and gases through the
mineral matrix of a tailings deposit and geochemical processes include
all mineral reactions and interactions between gases, liquids and
solids within a deposit.
The parameters relating to each of these processes are determined by
the physico-chemical nature of the tailings material in question. It
was the purpose of this study to discuss the transport and geochemical
processes as they pertain specifically to gold FRDs of the
Witwatersrand, as these deposits are relatively well constrained in
character and composition.
The nature of FRD varies throughout its lifecycle and depends amongst
other factors on the method by which it was constructed, its
mineralogy, climate, particle size distributions local topography and
geology, age, dimensions and geochemical character. Available data
indicate that the tailings materials of the Witwatersrand do not vary
greatly in terms of their physico-chemical characteristics. Only two
depostitional methods were recorded in the literature, i.e. paddock
systems and cycloned tailings deposits.
Witwatersrand FRDs are typically very large deposits. They are
typically greater than 1km in length and breadth and up to 60m high.
Mineralogy is dominated by quartz (70% and 90%), phylosilicates (mainly
sericite) (10% to 30%) and minor amounts of primary minerals (1% to
2%), including pyrite and calcite, which are key minerals in
controlling AMD.
Some particle size fractionation may occur by hydraulic separation,
thus resulting in differential permeabilities within these deposits.
This effect is enhanced in cycloned deposits, where the wall sections
are constructed with thickener underflow. The effect is expected to be
relatively limited in Witwatersrand tailings, as the particle size
distribution tends to fall within a relatively narrow range, when
compared to tailings from some other mineral deposits.
The moisture content within an FRD or between different FRDs may be
highly variable. Factors determining moisture content include climatic
conditions and physical properties of the FRD on a macro and micro
scale. Fine-grained layers, which may form capillary breaks are
important in FRDs, as they cause local variation in moisture contents
and can significantly affect the flow of gases. Existing measurements
on the study site indicated that moisture tends to vary between 20 and
40% by volume in the material. Along with porosity, the moisture
content plays a major role in determining the diffusion coefficient
within a waste body, as it has been determined that the diffusion
coefficient through waste is proportional to the air filled porosity.
This is largely a result of the fact that the diffusion coefficient of
oxygen in water is four orders of magnitude slower than the diffusion
coefficient of oxygen through air.
TRANSPORT PROCESSES
Water and gas transport are the main processes affecting AMD and
pollution mobilisation at gold FRDs. These processes are expected to be
rate limiting in that flow processes determine the rate at which
contaminants are mobilised from the tailings material into the
receiving environment, whereas gas transport processes limit the rate
at which AMD processes take place within the tailings deposit.
The flow of water determines how much of the contaminant load is
mobilised and also determines how fast and in which direction it will
move. Additionally, it also determines the moisture content within the
tailings body at any one position in time and space and thus influences
the oxidation rate by determining the fraction of air-filled pore space
available for air to move through.
In fine tailings material, where oxygen diffusion is the major gas
transport mechanism related to pyrite oxidation, gas transport rates
are essential in determining the rate at which sulphides can be
oxidised. This amounts to a direct measurement of the rate at which
contaminants (acidity and toxic elements) are generated.
The rate of contaminant generation, along with the rate of contaminant
transport will determine the concentration and volume of seepage from
the FRD’s and thus the environmental impact.
Water flow processes
The hydraulic conductivity in Witwatersrand gold FRDs has
been found to be highly anisotropic (the ratio of horizontal
permeability vs. vertical permeability is significantly different),
largely due to compaction of the sediment column in a water-saturated
environment. Hydraulic conductivity in the horizontal plane is
considerably higher than in the vertical plane. The anisotropy is
caused mostly by the stratification within the sediment and shrinkage
cracks. Anisotropy coefficients of between 5 and 10 are considered
normal, but values up to 200 have been reported (Rösner et al.
(2001).
During the operational phase, seepage from an FRD in the vertical
direction is largely dependant on the permeability of the material at
the base of the FRD and the presence of subsurface drains. Where the
base material is less permeable in the vertical direction than the
tailings material, seepage will move horizontally towards the walls of
the FRD. The results of this movement are observed in toe seepage. When
tailings material is less permeable than the underlying material,
seepage will tend to move downward, through the foundation into the
local groundwater system.
Both saturated and unsaturated flow processes occur within an FRD,
especially during the operational phase. In the post closure phase,
unsaturated flow processes become progressively more dominant as the
piezometric surface subsides. Future recharge then occurs under
unsaturated conditions. Previous modelling studies found that recharge
would occur at a rate of between 1 and 6% of MAP after decommissioning.
Flow under saturated conditions can be represented by a simple form of
Darcy’s equation and is applicable below the piezometric
surface within an FRD. This zone extends from the edges of the pool
along the hydraulic gradient towards the edges of the FRD. If the
piezometric surface intersects the wall of the FRD, toe seepage occurs.
Morin and Hutt (1997) indicated that flow in the walls of an FRD is
likely to be more rapid than near the pool. This is because of the
coarser grain sizes and the shorter flow path length. Anisotropy can
also be expected to play a part. In the post closure phase, it has been
shown that the piezometric surface subsides at approximately 0.5m per
year, which is an indication of the saturated hydraulic conductivity.
The saturated hydraulic conductivities of Witwatersrand tailings have
been determined to be between 10-6 and 10-8
ms-1.
Unsaturated flow also obeys Darcy’s law, however, it is more
complex to calculate, because it is dependent on the degree of moisture
saturation which typically varies over time. This is because large
pores drain more quickly than small pores and water filled pores
conduct water more easily than air filled pores. Empirically,
unsaturated hydraulic conductivity is a non-linear function of
volumetric water content and is also dependent on hydraulic head. It is
commonly expressed as a function of soil suction (K(θ), pore
water
pressure (K(Uw) and/or effective pore radius (K(R). The effect of
hystersis has resulted in the preferential use of the moisture
retention (soil suction) curve, as the differences between wetting and
drying curves are less than for pore water pressure. A large number of
empirical models have been developed to make use of these parameters.
Gas transport processes
Gas transport is generally described as occurring either
by advective
or diffusive flow. In tailings materials it has been shown that
advective flow, including barometric pumping (movement of air due to
changes in atmospheric pressure), does not contribute significantly to
oxygen transport, due to the fine-grained and hence relatively
impermeable nature of tailings material. Where preferential pathways
exist in the near surface environment, a limited amount of advective
flow may occur, however, this is not expected to significantly affect
the overall oxygen transport rate within an FRD. Throughout the
literature surveyed, diffusion was recognised as the major mechanism
for oxygen transport through an FRD, and also serves as the basis upon
which oxidation models for fine tailings are based.
Diffusion in tailings is driven by the concentration gradient between
the tailings surface and the matrix, caused by weathering reactions
that consume oxygen, e.g. the oxidation of pyrite. Oxidation is usually
confined to a few metres below the surface as a result of oxygen
transport limitations due to diffusion, with oxygen concentrations
becoming progressively lower with depth.
Oxygen is the most significant gas in determining reactions in tailings
material. The ability to determine the oxygen diffusion rate through an
FRD over time will thus enable long-term prediction of AMD potential.
Along with knowledge of water flow, calculation and/or modelling of
pyrite oxidation rates provides the tools for determining pollutant
concentrations within and around a tailings deposit over time.
Gas diffusion is mathematically described by Fick’s law.
Previous studies have showed that the limiting parameters controlling
pyrite oxidation through diffusion are the diffusion coefficient and
the kinetic rate constant. As the diffusion depth increases, oxygen
diffusion becomes a more dominant parameter than the inherent pyrite
reaction rate in determining oxygen flux.
Investigations into diffusion controlled pyrite oxidation in FRDs
internationally has led to the conclusion that determinations that are
made on bulk tailings tend to overestimate reaction rates. This is
caused by local variation in moisture saturation levels within the
tailings profile, due to the presence of more saturated fine-grained
layers and less saturated coarser grained layers.
Geochemical processes
Mineral
reactions
Witwatersrand gold tailings are constrained within a
relatively narrow
compositional range. Minerals that contribute to the Acid Potential
(AP) are present in relatively low concentrations, however, AP is
generally higher than the Neutralisation Potential (NP).
Acid Potential is predominantly ascribed to pyrite, as it is the most
dominant sulphide mineral, with other sulphides occurring less
frequently. Pyrite may react along different evolutionary paths,
depending on the pH and redox conditions in which the reaction takes
place.
Neutralisation Potential is ascribed primarily to carbonates of Ca and
Mg. Silicates may also contribute significantly to the NP in a tailings
deposit, depending on their inherent neutralisation potential and
reactivity. Some silicate minerals are capable of neutralising acidity,
especially where reaction rates are slow.
The formation of secondary minerals was also shown to be of major
importance as the reaction stoichiometries of neutralisation reactions
are critically dependent on mineral reactions that occur. Formation of
ferrihydrite after pyrite oxidation, for example, results in double the
requirement in neutralisation potential to prevent the reaction
solution from acidifying.
Mineral reaction kinetics
Mineral reaction kinetics focus mostly on the oxidation of
pyrite. The
shrinking core model has commonly been used to calculate oxidation
rates of pyrite grains over time. An oxidised layer, which forms
through reaction of pyrite has been measured in the laboratory and
calibrated
empirically to have a secondary diffusion coefficient in the region of
approximately 7.8x10-12 m2s-1
Bacteria such as Thiobacillus
ferrooxidans play an important part in
determining oxidation rates of pyrite, however, in a diffusion
controlled environment this has a more limited effect. It has been
determined that under conditions of less than approximately 8% oxygen
these bacteria do not contribute significantly to pyrite oxidation
reactions.
FIELDWORK
Fieldwork for this research concentrated on supplementing
field data
that were collected for Anglogold Limited by Pulles, Howard and de
Lange in 2002. The approach that was followed, was to generate
additional data and to discover relevant field related variables that
would be of interest in constructing an oxygen diffusion based model
for assessing acidification risk. Parameters that were determined in
the field and laboratory included sample descriptions, particle sizes,
X-Ray diffraction mineralogical analyses (XRD) and shake flask tests.
Additional parameters, such as ABA analyses, moisture contents,
particle size distributions, porosities and sulphur concentrations were
obtained from the literature.
Observations and data.
Surface features were found to be abundant. Shallow mud cracks, the
growth of plants and changes in particle size distributions were
observed across the surface of the deposit. The presence of mud cracks
and plants is expected to result in some variation in permeability
across the surface of the FRD through macropore spaces. In the case of
particle size distributions, variation is expected between
conductivities in the wall and beach sections of the deposit.
Subsurface features included the presence of a leached zone directly
below the surface, a hardpan horizon below this and an oxidation front,
where iron stained sediments contacted grey, apparently unoxidised
tailings.
The leached horizon is expected to be the result of rainfall leaching
of oxidised products. The hardpan horizon is a result of mobilised iron
from pyrite oxidation forming secondary minerals. Evidence of variable
oxygen concentrations and preferential flow paths was noted by the
presence of mottling and accentuated staining respectively.
The phenomenon of the oxidation front was researched and it was found
that this transition is related to the presence of oxidised Fe and that
it does not necessarily define the point of zero oxygen concentration.
Redox conditions were found to be a function of the Fe2+/Fe3+
concentration. The rapid transition, colour transition that is normally
observed, along with occasional evidence of pyrite formation in cracks
below the transition zone, indicates that the colour transition is
likely to represent the position of the transition from oxic to anoxic
conditions.
The presence of moist, fine-grained layers was observed. These layers
are known to retard oxygen diffusion and were found to occur at
different levels within an oxidation profile. Few moist layers were
found in the top two meters of the tailings profiles investigated,
probably as a result of desiccation during the long preceding dry
season.
A number of yellow stained strata were observed in tailings auger
cuttings that were collected. These were ascribed to syn-depositional
oxidation. Due to rapid oxidation of near surface pyrite grains, such
layers form rapidly during any drying cycle experienced across an FRD.
A significant finding that was observed from hydrometer particle size
distributions, was that a significant proportion of analyses indicated
that vertical variability in particle sizes is at least as significant
as horizontal variability. This is important with respect to hydraulic
conductivity and moisture retention, two key parameters in modelling
impacts from tailings dams.
MODELLING
A conceptual model was developed to represent the
generation of acidity
from pyrite oxidisation within a typical Witwatersrand FRD. According
to the model, wall sections of a tailings dam will undergo rapid
acidification, as a result of horizontal influx of oxygen across the
profile. The bulk of an FRD will, however, undergo oxidation through
vertical ingress of oxygen by diffusion through the tailings profile. A
large store of unreacted material, with associated NP occurs below the
oxidation zone and potentially provides a means to delay the onset of
acidic seepage from the base of the FRD for an extended period.
Use was made of the dual diffusion PYROX model to develop envelopes for
oxidation rates of typical Witwatersrand FRD’s. Making use of
the PYROX modelling results, a spreadsheet engineering type model was
developed to model the likelihood of acidification from an FRD due to
AMD process. The purpose of the probabilistic spreadsheet model
(Tailings Acidification Prediction or TAP model) that was constructed
as part of this project, is to make use of this concept to predict the
likely rate of acidification of seepage at the base of an FRD. The
model consists of two components, i.e. an oxygen diffusion component to
predict trends in pyrite oxidation rates and a neutralisation component
to predict the rate of consumption of NP in the FRD.
Acidification of seepage is considered to be a major impact, relative
to neutral drainage, as a result primarily of the potential for metal
leaching to the receiving environment, but also as a result of acidity
being harmful in itself.
Actual impacts will be a function of flow processes, coupled with the
rate of contaminant generation within an FRD and the sensitivity of the
receiving environment. Receptor impacts, however, will have to be
evaluated on a case-by-case basis.
PYROX modelling
The PYROX model developed at the University of Waterloo is
based on a
double diffusion model developed by Davis and Ritchy (1986) and has
been used extensively in international tailings investigations. The
model was used to perform sensitivity analyses on relevant variables on
the control of pyrite oxidation rates. Parameters were tested within
the known compositional range of Witwatersrand tailings, in order to
assist in developing a simplified empirical model, which could be used
for assessing the likely time span for generation of acidic seepage
from the base of a Witwatersrand gold FRD.
Typical to conservative values of sulphur content, particle size,
moisture content, porosity, temperature, secondary diffusion
coefficient and the thickness of the tailings profile were used as a
baseline for determining sensitivities of other parameters.
It was found that within the ranges of these parameters within a
Witwatersrand FRD, the absolute values of pyrite oxidation rates vary
as expected. It was also, however, found that the rate of change of the
oxidation rate over time fell within a relatively narrow range of
fitted functions, especially when viewed over the long term. These
findings were used to construct the TAP model as a function of static
diffusion rates at initial compositions.
The TAP model was constructed from the point of view that the impact
due acidification of seepage from the base of an FRD is dependent on
the time span over which it is expected to occur. Consideration of the
likely lifespan of suggested mitigation measures is a major
consideration in this respect, as well as the likelihood of major
impacts occurring after an extended period of weathering, when
contaminant generation rates are likely to be much lower.
The model makes use of humidity cell secondary oxidation rates,
porosity, the degree of moisture saturation, pyrite concentrations, the
depth of a tailings profile and the NP:AP ratio of the tailings to
predict a time to acidification of seepage.
Sensitivity analyses that were run for the model, indicated that most
FRDs are likely to take more than 1000 years before acidic seepage will
be generated from their bases. Exceptions were found to occur for FRDs
with low sulphide concentrations and for facilities of low height.
These findings are ascribed to the deeper penetration of oxygen in a
low sulphide scenario, where oxygen consumption occurs more slowly and
the relatively large proportion of oxidised tailings relative to
unoxidised tailings in a shallow profile.
DISCUSSION OF RESULTS AND
IMPLICATIONS FOR THE MINING INDUSTRY
Modelling and field observations indicate that weathering
in the top of
the tailings profile in a typical Witwatersrand FRD is rapid for
approximately 10 years after decommissioning and that an oxidised
horizon of up to 5m deep will form during this time. As discussed
above, variations in physico-chemical parameters can modify these
figures to form shallower or deeper profiles. Modelling results show
that pyrite oxidation is only likely to consume available NP after
approximately 1000 years in tailings of typical composition.
Considering the particle size distributions of tailings, it is likely
that engineered surface covers will not significantly reduce oxidation
rates on an FRD, except in the short period after decommissioning. The
value of a cover may be realised in maintaining higher moisture levels
and in reducing recharge through an FRD, however.
PROPOSED ASSESSMENT AND
DESIGN CRITERIA FOR MITIGATION OF ACIDIFICATION
IN WITWATERSRAND FINE RESIDUE DEPOSITS.
Field measurements
Field measurements need to concentrate on providing
calibration data
for the TAP model. Relevant measurements, which will assist in this
regard include:
- Directly determined oxygen
flux measurements, which can be checked against modelled values. These
measurements should be made regularly to determine temporal variations.
- Moisture contents, which are
important for determining diffusion coefficients and can be determined
with geophysical methodologies. Measurements should be ongoing to
determine temporal variations.
- Porosity.
- Pyrite concentrations in
representative samples below the oxidised zone.
- NP measurements from samples
above the piezometric surface, but below the zone where acid reactions
have already occurred.
Sampling should be conducted such that oxidation reactions are
prevented, through exclusion of air and if possible the inclusion of
inert gases in samples.
Potential mitigation
options
A number of potentially promising mitigation options were
identified on
the basis of the conceptual model. The various options need to be
subjected to technical and financial feasibility assessments before
they can be considered as guidelines, however. A list of identified
measures is included below:
- Alkaline addition
– Addition of limestone, especially in the wall section, will
decrease the AP:NP ratio and will reduce the likelihood of
acidification of seepage due to the acid-base balance.
- Depyritisation –
This technique has been documented in the literature, especially with
regard to cycloned tailings. Pyrite can be separated and included in a
lower risk environment elsewhere or can be reprocessed for a value
added product.
- Engineered covers
– These are commonly used in the industry and are designed to
reduce oxygen ingress through maintaining high moisture levels and to
reduce recharge and consequent seepage through the store and release
principle.
- Fly ash – Fly ash
is alkaline and fine grained, as well as having pozzolanic properties
in many cases. Mitigation with this material may therefore be possible,
however, caution needs to be exercised due to the commonly high
leachable metal concentrations in fly ashes.
- Bottom liner –
Where seepage is generated, a bottom liner can be used to collect
affected seepage for treatment. This will be especially of relevance
where acidic seepage will be generated at an early stage in the wall
section of an FRD.
The suggested measures need to be considered in the context of
available opportunities throughout the lifecycle of an FRD, for
instance, a liner must be installed prior to construction for maximum
efficiency, whereas a cover can be constructed only after
decommissioning of the facility.
CONCLUSIONS AND
RECOMMENDATIONS
The research has shown that the depth and rate of
weathering in
Witwatersrand FRDs is highly dependent on physico-chemical properties,
however, it was shown that facilities of typical composition and
morphology should take more than 1000 years, before acidic seepage is
released to the receiving environment below the base of the facility.
The objectives for the study were partially achieved, in so far as it
was found that accurate modelling of oxidation rates is relatively
complex and data intensive. Providing a general model is therefore
difficult due to the various site-specific variations in parameters
affecting AP and NP realization on an FRD. The recommended mitigation
measures need to be subjected to technical and financial feasibility
assessments before they can be accepted as best practice measures.
It is thus recommended that further work should be conducted in
evaluating these mitigation measures. Other recommendations include:
- Verification of findings by
application of other models.
- Collection of field data to
develop a database against which calibration can take place and which
can serve as a record. Parameters to be determined include:
- Pyrite concentrations
- Intrinsic pyrite oxidation rates
(humidity cells)
- Empirical determination of diffusion
coefficients
- Moisture profile determinations
- Depth dependent oxygen concentration
profile determinations for time line calibration
- Oxygen flux determinations for
oxidation rate verification.