UNDERSTANDING THE HYDROGEOLOGY OF BAUXITE RESIDUE DISPOSAL AREAS AT THE WORSLEY ALUMINA REFINERY, COLLIE, WESTERN AUSTRALIA
R. F. Wright1, S. C. Davidge2, A. M. Speechly1, P. B. Johnston1
1. Worsley Alumina Pty Ltd, Collie, Western Australia, 6225
2. Golder Associates 441 Vincent St West, Leederville WA 6007
Detailed hydrogeological characterisation of a bauxite residue disposal area (BRDA) at Worsley Aluminas Collie Refinery has been used to support closure design investigations. The investigations have included the drilling and installation of piezometers, water balance calculations, field and laboratory estimation of hydraulic parameters, and groundwater modelling.
Hydraulic parameter estimations are most important for the reasonable prediction of future underdrainage flow rates from the BRDA and the success of re-vegetation. These data have been estimated from a variety of field and laboratory methods and confirmed by water balance calculations and groundwater model calibration.
The water balance work indicated that the capillary fringe does not follow the BRDA phreatic surface fluctuations. The fringe stretches between winter (when it is close to the phreatic surface) and summer (when it is at its most distant).
The development of a satisfactory conceptual groundwater model was critical to the successful development of a predictive (mathematical) groundwater model.
Predictive groundwater modelling has been performed to estimate the likely future underdrainage flow from the BRDA. These predictions were necessary for planning of post-closure liquor collection and treatment (or other disposal method).
hydrogeology, bauxite residue, residue storage, modelling, decommissioning
UNDERSTANDING THE HYDROGEOLOGY OF BAUXITE RESIDUE DISPOSAL AREAS AT THE WORSLEY ALUMINA REFINERY, COLLIE, WESTERN AUSTRALIA
R. F. Wright, S. C. Davidge, A. M. Speechly, P. B. Johnston
Worsley Alumina Pty Ltd (WAPL) has bauxite mining and refining operations in the Darling Range, near Collie, Western Australia. Bauxite ore refining residue (bauxite residue) is a waste product from the refining of bauxite ore to alumina.
Environmental legislation demands safe containment of the bauxite residue, which in this case is achieved by deposition into large dams, known as bauxite residue disposal areas (BRDAs), situated within the refinery lease. Worsley Alumina is legally bound to develop a decommissioning plan for their BRDAs. One BRDA in particular is approaching decommissioning stage. A sound knowledge of the hydrogeology of this BRDA is essential for the development of a satisfactory decommissioning plan.
Research into the hydrogeological system that occurs within and near to BRDAs at the Collie Refinery has been carried out over the last years.
Specifically, a detailed study of BRDA1 has been used:
- as a guide to understanding the hydrogeology at all the BRDAs.
- for closure planning of BRDA1, which will be the first BRDA to be closed.
When BRDA1 is closed and rehabilitated, it will be capped with between 300 to 500 mm of soil and re-vegetated. An interception drain and capillary break will underlie the soil cap, so that it will be non-infiltrating (rather than "impermeable").
The objectives of the hydrogeological study were:
- to develop a clear understanding of the hydrogeological processes within a BRDA that are occurring now and will occur after closure.
- to develop preliminary estimates of the water balance that will occur in a closed BRDA and how it can be managed to meet the timetable for eventual refinery closure.
- to support selection of a conceptual design for the closure of BRDA1.
Bauxite residue is a by-product of the Bayer process in which bauxite ore is physically and chemically transformed to extract alumina, leaving a viscous slurry by-product. The bauxite residue is dewatered to give a slurry of approximately 62-64% solids. The solids consist of combined sand and fine silt to clay sized fractions, of which iron (oxy)hydroxides are the dominant mineral. Before and after drainage, the residue remains alkaline, sodic and saline. Consistent with residue chemistry, the drainage liquor is alkaline, saline and contains humic substances.
The alkalinity, salinity and sodicity of residue and drainage liquor is due to digestion of bauxite ore with caustic soda in the Bayer process.
Residue slurry is contained in large dams constructed within naturally occurring valleys of the Augustus River Catchment. BRDA1 deposition commenced in 1984.
Residue was placed within a natural valley, across which an embankment was constructed to contain the residue. The embankment was constructed in six stages, being the initial embankment and five raises. Each raise was constructed with the upstream construction method using locally sourced materials. As the raises progressed, it became necessary to construct upstream embankments to encircle the stored residue. Deposition ceased behind the latest raise in 1995. Figure 1 shows a cross section of BRDA 1 and the drainage system.
Cross Section through BRDA1
The main features of the BRDA design are summarised below.
Bauxite residue is transported as a slurry to the BRDA and deposited from a single spigot that is moved regularly. Residue dries and consolidates in BRDAs via solar evaporation and drainage.
The refinery site comprises a gently undulating lateritic granite upland. In places the granite bedrock is close to the surface and thin gravelly soils occur over it. The most common soil profile is that of thin soil over a saprolite sequence which overlies the massive granite basement.
The site has a Mediterranean climate with cool wet winters and hot dry summers. Approximately 80% of annual rainfall is received in the six month period May-October. Average annual rainfall for the refinery is about 1200 mm. Evaporation exceeds rainfall in the months between October and April and on an annual basis.
The groundwater quality beneath the site is variable but consistently fresh to brackish.
The study comprised a combination of field, laboratory and desktop activities. In particular they included:
- piezometer installation and monitoring.
- hydraulic conductivity measurements.
- estimating storage characteristics.
- computer-based groundwater modelling.
A total of 29 piezometers and twelve vertical permeability test holes were installed to depths ranging from 0.39 m to 18.08 m at various locations across BRDA1.
Several of the piezometers were installed in clusters of two or three, to enable vertical groundwater gradients to be measured at these locations. The piezometers were installed according to a design (25 mm casing with slots and gravel pack) that enables estimation of horizontal hydraulic conductivity by falling or rising head tests. The vertical permeability test holes comprised a design (100 mm casing driven into the saturated zone) that enabled estimation of vertical hydraulic conductivity by the same tests.
All the test work was concentrated in the northern part of BRDA1 so that greater detail could be obtained for the sections and maps. It was assumed for this purpose that the hydrogeological conditions were symmetric about an axis along the length of the storage.
Figure 2 shows a plan view of BRDA1 and the location of installed piezometers.
BRDA 1 Location Plan
The finite element groundwater model, SEEP/W, was selected to predict long-term baseflows and recession times of the liquor contained in BRDA1. The model included representation of the residue, the embankments (including the central embankment), the sand drains, the clay liner and drainage layers, the clay foundation underlying the BRDA, and the weathered basement aquifer.
The model was calibrated to measurements of water levels in the BRDA1 piezometers and the measurement of flow at the Pipehead Dam. The measured distributions of hydraulic head constrained the relative distribution of hydraulic conductivity (i.e. areas where it is low and areas where it is high), and the magnitude of the baseflow constrained the absolute values of hydraulic conductivity.
Results and discussion
Groundwater occurs within the BRDA as pore-water in the deposited and consolidated residue. Based on assessment of the piezometer data, it appears that the saturated residue is in continuous hydraulic connection with the underlying "native" groundwater system. Therefore, the ground immediately underlying the BRDA is saturated and the BRDA does not appear to comprise a "perched" system of its own.
The results of field and laboratory measurements indicate that hydraulic conductivity decreases with depth in the eastern and central part of the BRDA, and that hydraulic conductivity is relatively constant with depth in the western part of the BRDA (i.e. close to the main embankment).
Estimates of horizontal hydraulic conductivity were within the range of 3x10-6 m/s to 8x10-10 m/s in all parts of the BRDA tested. The results also indicated that vertical hydraulic conductivity is approximately an order of magnitude less than horizontal hydraulic conductivity in the eastern and central parts of the BRDA, and that they are approximately equal in the western part of the BRDA.
The measured relationship between pressure and volumetric water content indicated a specific yield in the order of 0.2, and an air entry pressure between 0 and 5 kPa suction.
An air entry value between 0 and 5 kPa suction was interpreted to be unrealistically low for the fine-grained residue, and inconsistent with observations of the hydro-dynamics of the BRDA. Accordingly, for the modelling work, storage characteristic curves for the residue were developed from experience-based assessment of the likely values.
When shallow water pressures are at their peak in late July, the water levels in the upper most piezometers are close to or above the surface of the BRDA, indicating a groundwater mound in this area, with water flowing laterally away from this mound. At this time, water is ponded on the surface around the central decant towers. In early April, water pressures in the western portion of the BRDA are also greater than in the east, indicating mounding in the west and flow towards the east.
Downward vertical groundwater gradients exist at all times over the entire length of the BRDA (indicating downward vertical movement of groundwater). Vertical gradients are not uniform across the BRDA, reflecting the difference in depositional history to the east and west of the central embankment, and the differences in hydraulic conductivity.
Baseflow from BRDA1 Drainage System
Flow measurements of the BRDA underdrainage are made at the Pipehead Dam. However, the volume of water that is collected by the drainage system at the base of BRDA1 cannot be measured accurately. This is because the drainage system is connected in series with a portion of the drainage systems from two other BRDAs, prior to the point at which the flow measurement is made.
Based on the available information, it was estimated that the flow from the base of BRDA1 is between 50 and 80 kL/day, and that it remains relatively constant throughout the year. Since the outflow from the BRDA should currently be near to steady state in balance with the infiltration at the surface, this implies that the average infiltration at the surface is also currently between 50 and 80 kL/day. This represents approximately 4 to 8% of incident rainfall.
With knowledge of the groundwater flow rates and hydraulic parameters it was possible to develop a conceptual model of the hydrogeological system within BRDA 1.
The components of the water balance for the BRDA are illustrated in Figure 3. Rainfall that falls on the surface of the BRDA will infiltrate if the pores of the residue are not saturated at the surface. Once the pores become saturated, surface ponding and runoff commence. Water that infiltrates contributes to changes in storage in the vadose zone, and potentially beyond the vadose zone to the phreatic surface (recharge).
Conceptual Model of BRDA1 Groundwater Flow
Water that is stored in the vadose zone is available for transpiration by plant roots, and at the surface itself is available for direct evaporation to the atmosphere. Partitioning of water to the vapour phase and subsequent vapour transport to the surface is also possible. The water in the vadose zone and the saturated zone can be regarded as a continuum, and transfer of water within and between the two zones occurs in response to hydraulic gradients. The relief of pressure at the base of the BRDA by the provision of a drainage system open to the atmosphere induces downward gradients that cause downward flow within the saturated zone.
Information on the outflow from the base of the BRDA (discussed above) enabled an assessment of whether recession in piezometer water levels over the dry season reflects a drop in the capillary fringe, or whether it reflects a change in water pressure beneath an extending capillary fringe.
An approximate analysis of the baseflow data suggested that the air entry value is significantly less than zero, and that during the summer, the water table recedes, the top of the capillary fringe recedes much less, and the capillary fringe extends. The summer recession characteristics would therefore depend on the value of air entry pressure and on the specific storage. The longer-term recession under reduced infiltration conditions will require drainage of pores, and therefore will depend on the value of specific yield.
Predictive modelling was carried out to estimate the long-term recession of the saturated thickness of residue and the flow from the basal drainage system. For the predictions of long-term recession, the flow rate at the upper boundary of the model was specified as various percentages of the estimated current recharge rate.
The long-term recession behaviour in BRDA1 will depend on values of hydraulic conductivity, and the volumes of water which are released from storage for given reductions in pressure. Release of water from storage as a result of pressure reduction will depend mainly on air entry pressure and specific yield. Recession in the phreatic surface would be most rapid for large values of hydraulic conductivity and low storage, whereas recession would be slowest for small values of hydraulic conductivity and large storage.
Based on the range of results from field and laboratory measurements, and on results of model calibration, three sets of parameters were chosen for the predictive modelling:
- a combination of the "most likely" hydraulic conductivity values and "most likely" storage characteristics
- a combination of the "upper bound" values of hydraulic conductivity and the "lower bound" storage characteristics (corresponding to rapid recession rate)
- a combination of the "lower bound" values of hydraulic conductivity and the "upper bound" storage characteristics (corresponding to slow recession rate).
Only the recession predictions from the "most likely" parameter set are presented in this paper.
The adopted most likely values of specific yield, air entry pressure and specific storage are provided in Table 1. Horizontal hydraulic conductivity values between 1x10-6 m/s and 1x10-9 m/s were used in the modelling to most reasonably represent the measured variation. Vertical hydraulic conductivities were set at values between 1 and 1/10th of horizontal.
Storage Parameters for Predictive Modelling
Specific Yield (-)
Air Entry Pressure (Kpa)
Specific Storage (m-1)
"Most Likely" Storage
When recharge to the phreatic surface is reduced below its current average value by the proposed residue capping, water levels in the residue pile will drop to new levels which are in equilibrium with the altered recharge rate, and the baseflow will similarly decrease to equal the altered recharge. Recession of the top of the capillary fringe provides a better measure of the progress of recession than does recession of the phreatic surface, since it indicates the depth of residue that remains saturated.
Since baseflow will approach a new steady state value asymptotically, an operational definition was adopted for "recession time," as "the time required for baseflow to drop by 90% of the expected total change". The predicted recession time using the "most likely" set of hydraulic parameters and 0% infiltration is approximately 75 years, which corresponds to a baseflow of approximately 8 kL/day (Figure 4). At this time the top of capillary fringe is predicted to have fallen by 13 m in the western section of BRDA1 and 7m in the eastern section.
Predictions of Recession of Baseflow as a Function of Time
The hydrogeological study of BRDA1 has yielded valuable insight into the short and long-term behaviour of groundwater flow in the BRDAs at the refinery. The most important conclusions that were drawn from this study are:
- only a small proportion of incident rainfall on the surface of BRDA1 infiltrates and eventually leaves the BRDA by basal drainage
- the capillary fringe stretches between winter and summer, resulting in generally low discharge of baseflow
- after capping of the BRDA and re-vegetation, near zero infiltration is expected to be achieved. This will result in a reduction of baseflow to about 10 % of currently estimated values within about 75 years
- recession of the capillary fringe is expected to take up to 100 years before most of the recession has occurred
- the development of the conceptual hydrogeological model was important, because it enabled the development of a mathematical model which incorporated best estimates of storage and hydraulic conductivity
- the development of a mathematical model of the system has been useful in quantifying some predictions that relate to on-going requirement for post closure liquor collection and treatment.