groundwater assessment - terramin australia limited · groundwater discharge mechanism (i.e., via...
TRANSCRIPT
Groundwater Assessment
Appendix F – Numerical Model Development
Numerical Model Development
F1
Appendix F – Numerical Model Development
1. Modelling overview and objectives The potential impacts of mining on groundwater (largely caused by mine inflows) include effects upon
existing users of groundwater, as well as the interaction between surface water and groundwater
resources such as losses in baseflow to watercourses. The possible magnitude and extent of impacts
related to the Bird-in-Hand Gold Project (Project) was assessed through the development of a three-
dimensional numerical groundwater flow (and solute transport) model.
The objectives of the groundwater model were to:
• Simulate pre-mining groundwater elevations and seasonal fluctuations due to recharge and
groundwater abstraction, to ensure the model can replicate the current groundwater flow regime.
• Determine the rate of groundwater inflows into the proposed mine and inform mine water
management design.
• Evaluate the groundwater related impacts (groundwater levels and salinity) of mining on private
wells (identified during the well census) and the baseflow to the Inverbrackie Creek.
• Evaluate the effectiveness of water management options, such as grouting of the underground
mine to minimise groundwater inflows and MAR of mine inflows.
• Assess the project against the water allocation criteria of the Western Mount Lofty Ranges
(WMLR) Water Allocation Plan (WAP), including estimation of groundwater abstraction from the
WMLR and Eastern Mount Lofty Ranges (EMLR).
The numerical model is a tool to identify risks, such as inflows (and the associated drawdown) caused by
unplanned interception of the hanging wall fracture. An unexpected inrush of groundwater onto the mine
which is higher than anticipated could threaten the safety of mine workers. Likewise, a mine inflow rate
substantially higher than predicted may indicate a greater water level or quality impact to the surrounding
groundwater users. Identification of the nature of such risks through the numerical modelling process
allows for implementation of conservative mitigation and contingency measures.
Evaluation of the potential mine inflows and groundwater drawdown which can emanate some distance
from the proposed mine required the development of a model with a high degree of confidence,
particularly as the proposed mine is situated within a complex FRA system. The confidence level of any
numerical groundwater model is based on the amount of available data.
A three-dimensional finite-difference numerical groundwater model of the study area was developed
using MODFLOW (Harbaugh, 2005), which is widely regarded as the industry standard software package
for numerical groundwater modelling. The numerical model construction was based on the conceptual
model of the study area, this being a simplified representation of the real system that includes the most
important geological units and hydrogeological processes. This is summarized in the following section.
Numerical Model Development
F2
2. Conceptual model
This section provides a summary of the conceptual model of the study area hydrogeology and proposed
mine.
The conceptual hydrogeological model was based on data gathered through exploration and investigation
drilling, geophysical logging, analysis of baseline hydrological data, hydraulic testing and regional
groundwater monitoring. Full details of the development of the various aspects of the conceptualisation of
the system are provided in the main body of the report (Section 4).
Figures F1 depicts the model domain extent in the context of regional surface geology and a schematic
geological cross section.
Numerical Model Development
F3
Figure F1. Model domain extent in the context of regional surface geology and schematic geological cross section.
Numerical Model Development
F4
The primary aquifer in the area is a Fractured Rock Aquifer, comprised of different geological
formations (Figure F1). Almost half of the sub-catchment (eastern part) is underlain by the
Kanmantoo Group which has been identified as a very poor aquifer and is not developed for
irrigation. Younger Adelaidean geological sequences such as the Tapley Hill Formation,
consisting of blue-grey laminated siltstone and slate, and the Tarcowie Siltstone are most
prominent in the rest of the catchment. Groundwater drawn from these formations is generally
< 1,500 mg/L with yields in the order of 5-10 L/s across the sub-catchment. Occasionally higher
yields are obtained.
The general regional groundwater flow is from the northeast to the southwest through the
Inverbrackie Creek sub-catchment. The main groundwater receptors include several operational
wells and the Inverbrackie Creek (and its tributaries), which receives baseflow in its lower
reaches, mostly during winter when the groundwater levels are higher.
Figure F2a presents the conceptual hydrostratigraphic sequence of the Bird-in-Hand Project site
in more detail, followed by the proposed mine plan with respect to the key hydrostratigraphic
components (Figure 2b).
Numerical Model Development
F5
Figure F2a. Conceptual hydrogeological cross-section (west to east).
Figure F2b. Cross-section (west to east) showing underground mine plan with respect to key
hydrostratigraphic components.
Hanging wall fracture
Marble
Decline (within Tapley Hill Fm)
Historic mine workings
West East
West East
Numerical Model Development
F6
The proposed underground mine would access the deep extension of the Marble which contains the gold
deposit that was first mined in the 1880’s. Underlying the Marble and outcropping to the north and west of
the site is the Tapley Hill Formation, which supports the majority of groundwater users. Overlying the
Marble are the Tarcowie Siltstone – which contains a significant water-bearing fracture, refered to as the
hanging wall fracture – and Cox Sandstone units. The hanging wall fracture, when modelled using drill
core data, is interpreted to extend towards the old Bird-In-Hand mine workings, roughly coinciding with
the depth (~125 m) where high water inflows were encountered.
The proposed underground mine has a life of 5.5 years and comprises of a decline within the Tapley Hill
Formation and a series of horizontal drives to access the gold within the Marble (Figure 2b). The layout of
the mine decline and drives has been configured to minimize the rate of groundwater inflows by avoiding
the major water-bearing structures that have been identified from the resource and test well drilling.
The numerical model developed for this project represents is a simplified representation of the real
system taking into account the main hydrogeological processes and features. Figure F3 provides
additional key conceptual model details. This figure was constructed via overlay on a cross-section from
the numerical model and therefore provides a visual overview of the numerical implementation of the
conceptual model components discussed above (and in the main body of the report).
The key elements of the conceptual model that were incorporated into numerical model are reiterated as
follows:
• Steeply west-east dipping stratigraphy;
• Water-bearing fracture zone in the hanging wall (HW);
• Groundwater divide east of the Project site separating the Inverbrackie Creek and Dawesley
Creek sub-catchments;
• Existence of higher groundwater salinity to the east of groundwater divide, within the less
productive Kanmantoo Group;
• Inverbrackie Creek, which interacts with the regional groundwater system, primarily acting as a
groundwater discharge mechanism (i.e., via baseflow);
• Groundwater flow barriers including fault structures and zones of weathering; in particular
significant barriers to horizontal groundwater flow to the north and south of the Project area.
Full details of the numerical model development are provided in the following sections.
Numerical Model Development
Figure F3. Conceptual hydrogeological cross section (west to east).
Numerical Model Development
3. Model domain and layer structure
The model domain covers an area of approximately 26.5 km2. The model domain covers the
Inverbrackie Creek sub-catchment area in the WMLR, and part of the Dawesley Creek sub-
catchment in the EMLR. The model comprises seven layers and approximately 900,000 grid
cells. Cell dimensions range from 50 m x 50 m to 5 m x 5 m within progressive refinement in the
vicinity of the mine area. Figure F4 displays the model domain in plan view, including the BiH
mine area (yellow box), as well as two representative cross-sections (indicated by the lines A-A’
and B-B’).
The upper boundary of the model domain is defined by the regional topography based on
available digital elevation model (DEM) data (2 m resolution).
Model layer geometry was guided by available information pertaining to the site geological
features. The mine-site geometry and thickness is based on the geological and structural model
developed by Terramin. The geological model was based on drill core information obtained
from >50 exploration holes and improved with information obtained from the hydrogeological
investigation bores (IB). Rock quality designation (RQD) information together with acoustic
downhole geophysical logs of the investigation bores (IB) was used to construct a 3D model of
the main fracture (water bearing) zones. The regional layer geometry, including the dip of the
hydrostratigraphic units and outcrop of the Brighton Limestone (Marble) and Tapley Hill
Formation, was extrapolated based on the geological records obtained from other historic
mines including ‘Nest Egg’, ‘The Ridge’ mine and ‘Two in The Bush’.
The Tapley Hill Formation was divided across model layers 6 and 7. Throughout the main
irrigation area layer 6 represents the productive zone of the Tapley Hill Formation, which is
targeted by private wells. The deeper and thicker portion of the Tapley Hill Formation is
represented as layer 7. Layer 6 was assigned aquifer properties in accordance with the regional
steady-state and transient calibration processes (described under Section 5), supported by
some pumping tests of private wells. Layer 7 was assigned K values of one order of magnitude
lower than Layer 6 to represent the reduction of K with depth (corresponding to fracture
extinction).
At the Project site layer 6 incorporates the footwall, which contains the mine decline, and was
assigned aquifer properties based on the pumping test of IB5. The marble, which hosts the gold
deposit (and mine drives), was assigned to layer 5. The hydraulic properties of this layer varied
with depth, in accordance with the presence of caving (i.e., higher K), competent geology
(representing the broader-scale K of the Brighton Limestone / Marble formation) and alteration
to clay (lower K). Note that this zonal reduction in K with depth presents as a lateral variation
within Layer 5 (see Figures F7 and F8) due to the steeply dipping nature of the stratigraphy in
this area (see Figures F2, F3 and F4).
Layers 2 to 4 represent the Tarcowie Siltstone, which also includes the Cox Sandstone units.
The hanging wall fault in the Tarcowie Siltstone (shown in Figure 2) was represented as Layer
3. A zone of less fractured rock (between the marble and the hanging wall fault) was assigned
to Layer 4 (in accordance with the geological model; see Figures F2a and F2b).
To the east of the site Layer 1 represents the Kanmantoo Formation, where it is interpreted to
exist based on geological maps (e.g. Figure F1). This formation has been identified as a very
poor aquifer and is not developed for irrigation, and is thus represented as an area of low
permeability. This formation has previously been characterized by other studies to have very
low bulk permeability (e.g., Coffey, 2007).
Numerical Model Development
The lateral boundaries of the domain generally follow the boundaries of the Inverbrackie Creek
sub-catchment. However, it is necessary to ensure that areas of key simulated stresses (e.g.,
mine area depressurization) are located far enough from model boundaries such that they are
not controlled by them. For this reason it was necessary to extend the model domain east of
BiH and beyond the south-eastern boundary of the Inverbrackie Creek sub-catchment and into
the neighbouring Dawesley Creek sub-catchment.
Numerical Model Development
Figure F4. Model domain, discretization and layer geometry, including stratigraphic unit annotation (Bird-in-Hand Gold Project area indicated by the yellow box; Inverbrackie Creek sub-catchment boundary represented by pink line).
Numerical Model Development
4. Model inputs
4.1. Boundary conditions and recharge
Figure F5 provides a graphical overview of the boundary conditions and recharge distribution
employed in the model. The lateral boundaries of the model domain are assigned either as
head-controlled boundaries employing the MODFLOW general head boundary (GHB) package,
or as no-flow boundaries.
Figure F5. Model boundary conditions and recharge including zones indicating model
layer to which recharge is applied (pink line represents Inverbrackie Creek sub-
catchment boundary).
The assignment of GHB and no-flow boundaries was guided by the characteristics of the
regional groundwater flow field, which was inferred through interpolation (kriging) of available
groundwater level information. Figure F6a depicts the interpolated regional flow field that was
used for this interpretation process, with Figure F6b showing the resultant values assigned to
GHB cells.
The information used to construct the interpolated regional flow field was based on 66 wells,
including 39 wells that are regularly monitored by Terramin (displayed in Figure F6, and
detailed in Table 3 and Figure 7 of the main body of the report), along with 27 additional wells
for which historical data is available through the Department of Environment, Water and Natural
Resources (DEWNR) online groundwater database ‘WaterConnect’ (DEWNR, 2017). The latter
(represented as grey points in Figure F5a) are less reliable data, as most comprise a single
historical groundwater level measurement, for which both seasonal fluctuation and potential
long-term changes are a source of uncertainty. Nonetheless, this data is considered to
Numerical Model Development
constitute a reasonable approximation of the broader-scale native regional groundwater flow
pattern, which is useful for informing model boundary conditions.
As can be seen through inspection of Figure F6, GHBs were assigned to model domain
boundaries in areas where the regional flow field enters or exits the model domain area. Along
the southern and eastern model domain boundaries where flow enters and exits the model
domain, a component of flow is also parallel to the boundary, with the interpolated contours in
Figure 6a indicating a near-linear hydraulic head gradient at and outside these boundaries.
Linearly gradated GHBs were assigned along these boundaries to represent this regional flow
behaviour (see Figure 6b).
GHB conductance values were inferred through the steady-state calibration process and range
from 0.2 m2/d to 5 m2/day. The same GHBs were assigned to all model layers. GHB
conductance C (given by C = KhA/dl, where A is cross-sectional surface area and dl is distance
to the specified GHB head) is typically parameter associated with high uncertainty (e.g.,
Knowling et al., 2015). Based on estimated Kh ranges and the variability in cross-sectional
surface area of model boundary cells along which GHBs are assigned (from 25 m2 to 3e4 m2),
KhA ranges between approximately 2.5e-4 and 1.1e4. Specified GHB head values were
assigned based on kriging of measured hydraulic heads outside the model domain area (see
Figure F6), the locations of which range from immediately adjacent to the model boundary, up
to 2 km away. Considering a representative distance dl of 1 km gives a range for KhA/dl of 2.5e-
7 – 11. The range of C values obtained through calibration is therefore considered reasonable.
No-flow boundaries were assigned where flow is largely parallel to model boundaries. These
boundaries are generally aligned with native groundwater divides, and which correspond to
boundaries of the Inverbrackie sub-catchment itself. Due to the proximity of the proposed mine
location to the sub-catchment boundary as discussed above, the model domain extends
beyond the sub-catchment boundary in the southeast and therefore beyond the native
groundwater divide (indicated in Figure F6a by the divergence of the interpolated flow field in
this area), which therefore requires explicit representation within the modelled regional flow
distribution. The existence of this groundwater flow divide is central to the current
understanding of the regional hydrology. Inverbrackie Creek and Dawesley Creek serve as
groundwater sinks in the adjacent Inverbrackie Creek and Dawesley Creek sub-catchments,
respectively. A flow divergence in order for groundwater to flow towards each of these
effectively parallel gaining streams is therefore a natural consequence.
As demonstrated in Figure F6a, Inverbrackie Creek (and its tributaries) are represented
explicitly within the model (discussed in greater detail in various sections below). Dawesley
Creek (and its tributaries) fall outside of the eastern boundary of the model domain (see
Figure F6a). However, its existence is implicitly represented by the assigned GHB values, as
the low end of the linear GHB gradation in the south-eastern corner of the domain is defined by
groundwater levels measured in close proximity to Dawesley Creek (these wells have unit
numbers 6628-14069 and 6628-21783). In the absence of measured stream stage data for
Dawesley Creek, the shallow groundwater levels (SWLs of 1.5-4 m) are assumed to be
approximately representative of the stream stage in this area.
Representation of the groundwater flow divide within the model is therefore achieved primarily
as a natural consequence of the assigned GHB distribution represented in Figure 6b, which is
based on the interpolated regional flow field of Figure F6a. Simulated recharge influences the
groundwater flow divide, but is not necessary for its occurrence in the model. Simulated
recharge, its influence upon the regional flow field and the sensitivity of model predictions to
recharge is explored in detail in section 9.2 below.
Numerical Model Development
The abovementioned groundwater discharge to Inverbrackie Creek (baseflow) is one of the
primary natural groundwater discharge processes in the study region. When the groundwater
level within the adjacent aquifer is higher than the stream stage elevations, the stream gains
water (at a volumetric flow rate as dictated by the streambed hydraulic conductivity).
Comparison of simulated baseflow with available information including a previously published
baseflow estimate and post-summer spring observations was undertaken as part of the model
validation process (see Section 6 below).
Inverbrackie Creek and its tributaries within the sub-catchment are simulated via the
MODFLOW river package to represent stream-aquifer interaction. Streambed elevations and
stream stage heights are estimated based on topography. The effect of streambed sediments
upon the hydraulic interaction between the creek the aquifers is controlled by the specified
streambed hydraulic conductivity and thickness parameters. Due to a lack of field-based data
pertaining to streambed hydraulic conductivity and thickness, these parameters were subject to
adjustment during the steady-state calibration process (detailed below), within bounds
considered to be realistic.
A summary of the model river package parameters is as follows:
• River stage varies from 443.2 m at modelled river origin in the northeast (460.1 m in the
tributary at the easternmost domain extent, to 342.7 m at southwest extent;
• Average riverbed hydraulic conductivity 0.5 m/day, varying between approximately
0 m/d (i.e., 2e-19) and 1.5 m/d (no field data is available to support these values,
however the range is considered reasonable and simulated total baseflow is
comparable to a previously published estimate – see section 6 below);
• Specified stream width of 2.5 m;
• Specified streambed thickness of 0.5 m.
Spatially distributed recharge is represented in the model via piecewise constant zones, as
shown in Figure F4. Zone boundaries were guided by the chloride mass balance (CMB)-based
estimates (refer to Figure 42 and section 4.6.1 in the main body of the report), as well as study
area soft knowledge such as the observed ponding of surface water within the topographic
depression immediately east of the mine site, likely contributing to elevated recharge to the
shallow/perched system in this area.
Modelled recharge rates vary spatially between 0.5% and 11% of rainfall (i.e., between 4 mm/y
and 80 mm/y), with spatially averaged annual recharge representing 2.5% of average annual
rainfall (equivalent to 18 mm/yr).
Throughout the majority of the model domain recharge is applied directly to the Tapley Hill
formation (model layer 6), as this is the uppermost saturated hydrogeological unit. In the
southeastern region of the model domain where the regional Tapley Hill aquifer dips steeply
(see Figures F2 and F3), recharge is applied to model layer 1. The zones pertaining to differing
recharge application layer are also indicated on Figure F5.
Numerical Model Development
Figure F6. a) regional groundwater flow pattern as inferred (through kriging) from wells monitored by Terramin (high confidence) and additional WaterConnect database wells (lower confidence); and b) model GHB (mAHD) and no-flow boundaries as inferred from a).
a) b)
Numerical Model Development
4.2. Hydraulic properties
Hydraulic conductivity and specific storage of the model layers have been assigned through a
combination of field-based hydraulic testing (described in section 2.4/Table 5 of the main body of the
report) and the calibration processes (detailed below). A summary of the model parameters assigned
to each layer is provided in Table F1. Hydraulic conductivity values (both horizontal and vertical) are
spatially distributed and are therefore displayed in Figure F7 and Figure F8, respectively. (As per
common practice, the logarithms of hydraulic conductivity values are displayed to delineate important
relative differences between values at the low end of the scale as well as the high end.)
Table F2 and Table F3 provide hydraulic property summaries in terms of key hydrostratigraphic units,
and comparison between model-based and field-test values, respectively.
Table F1. Summary of model parameters.
Model
layer Hydrostrat. Unit
Approx. layer
thickness (m)
*
Kh Kv Ss
(1/m) Sy ηe
1 Tarcowie Siltstone /
Kanmantoo Formation
15-350 Fig. F6 Fig. F7 1e-4 0.01 0.2
2 Cox Sandstone /
Tarcowie Siltstone 0-50 Fig. F6 Fig. F7 8e-7 0.03 0.05
3 Tarcowie Siltstone
(highly fractured) 0-25 Fig. F6 Fig. F7 3e-4 0.1 0.2
4 Tarcowie Siltstone 0-25 Fig. F6 Fig. F7 1e-6 0.03 0.05
5 Brighton Limestone
(Marble) 50 Fig. F6 Fig. F7 1e-7 0.1 0.1
6 Tapley Hill Fm. (footwall and
productive zone) 40 Fig. F6 Fig. F7 2.5e-5 0.05 0.05
7
Tapley Hill Fm. (below fracture
extinction depth) 125-500
Fig. F6 (equivalent zones
to Layer 6, one order of
magnitude lower)
Fig. F7 2.5e-6 0.01 0.01
*Zero thickness is represented by a very small value in the model, as MODFLOW requires spatially continuous layers.
Numerical Model Development
Table F2. Model hydraulic property summary by key hydrostratigraphic component.
Description Layer Layer
thickness (m) Kh (m/d) Kv (m/d) Sy (-)
Ss
(1/m)
Overburden – Tarcowie Siltstone
at Project site and Kanmantoo
Formation to the east
Layer 1 15 – 400 7e-5 1e-4 0.01 1e-4
Tarcowie Siltstone )including Cox
Sandstone) Layer 2 5 – 50 0.1 0.025 0.03 8e-7
Tarcowie Siltstone Hanging Wall
– Highly Fractured Layer 3 5 – 25 3 2.5 0.1 3e-4
Tarcowie Siltstone Hanging Wall
– Less Fractured Layer 4 5 – 25 0.25 1 0.03 1e-6
Marble – Upper Fracture Zone Layer 5 50 1 1 0.1 1e-7
Marble – Competent Layer 5 50 0.05 0.1 0.1 1e-7
Marble – Alteration to clay Layer 5 50 1e-4 1e-4 0.1 1e-7
Tapley Hill – Footwall Layer 6 40 0.03 –
0.05 5e-4 0.05 2.5e-5
Tapley Hill – Productive zone
(regional irrigation area) Layer 6 40
1.5e-3 –
3.7 5e-4 0.05 2.5e-5
Tapley Hill – Fracture extinction
zone Layer 7 125 - 600
1.5e-4 –
0.37 1e-6 0.05 2.5e-6
4.3. Comparison of model hydraulic properties and field estimates
Table F3 presents a comparison between field hydraulic property estimates and hydraulic property
parameters employed in the model in the vicinity of several key wells for which field tests were
performed. The field hydraulic property estimates are sourced from Table 5 in the main body of the
report.
One of the key model features, which is important to predictions of potential mine inflows, is the HW
fracture zone (see Figure F2b-d). This is targeted by investigation well IB-4, and hydraulic properties
were evaluated through a 5-day CRDT at a volumetric extraction rate of 17 L/s (see section 3.2.4 of
the main body of the report for full details), as well as several short-duration (400 min) airlift tests
conducted at higher rates up to 38 L/s (see section 3.1.3 of the main body of the report). Field
estimates of T for the HW fracture zone (IB-4) were made using multiple analysis methods and range
from 65-126 m2/d. Late-time drawdown data indicated lower T values on the order of 45-60 m2/d (see
Figure 19 in the main body of the report). The value of T for the HW fracture zone employed in the
model, following the CRDT transient calibration process (see below) is 75 m2/d, which is consistent
with the field test-based range. The sensitivity analysis process included a doubling of the HW
fracture zone T value to 150 m2/d, this being a conservative worst-case scenario value with respect to
the overall field test-based range.
Inspection of Table F3 indicates some minor discrepancies between field test-based and model
hydraulic properties. This includes higher model T values for wells IB-2 and IB-5 compared with field
test-based values. These wells target the Marble (containing mine drives) and Tapley Hill (containing
the mine decline) units, respectively, which contributes to the conservative nature of simulated rates
of mine water inflow. The model value for T at the location of IB-2 (Cox Sandstone) is also
conservatively high.
Numerical Model Development
The high field test transmissivity value obtained for well 6628-20475 is attributable to the underground
network of historical mine workings in its vicinity. This level of detail (especially at this distance from
the mine location) is not included in the model, and therefore the model T value of 5 m2/d represents
the broader-scale value for the Tarcowie Siltstone employed in this region (see Figure F7).
Finally, as discussed in section (5.3) below, spatially variable storage parameters were not introduced
into the model. Where significant discrepancies between field test-based and model values for
storativity exist, they involve a smaller model-based value such that model predictions of the
magnitudes of potential drawdown caused by mining are conservative.
Numerical Model Development
Table F3. Comparison of model hydraulic parameters with field test results (the latter are sourced from Table 5 in the main body of the report)*.
Well Parame
ter Field-test
values Model values
Sensitivity test values/range
Model Layer
Formation(s) Comments
IB3
T 0.1 5 10
Layer 2
Tarcowie Siltstone and
Cox Sandstone units
IB3 was screened across the lower Cox Sandstone band as this band contained groundwater. However Layer 2 also incorporates the TS above and below the CS resulting in higher T being calculated for this model layer (due to greater thickness).
K 0.003 0.1 0.2
b 28 50 -
IB1, IB2, IB4, IB5
T 17 - 125 (67) ** 75 150
Layer 3
Tarcowie Siltstone
(including hanging wall fracture - L3)
Aquifer properties were sourced from Table 5 in the main body of the report. For all tested wells, T ranged from 17 to 125 with a geometric mean of 67 m2/d. The range of T for IB4 alone was 49 to 125 m2/d with a geometric mean of 90 m2/d. In the model, the TS was split into two layers (Layer 3 and 4) with a combined thickness of 50 m. Layer 3 is the HW fracture zone (25 m thick) and layer 4 is a zone of less fractured rock (25 m thick) between the HWF and the marble. The airlift and pumping test was done over both of these zones (layers)
K 0.3 - 2.5 (1.3) 3 6
S 0.0004 - 0.005
(0.002) 0.0075 7.5e-4 - 0.075
b 50 25 -
T - 6.25 12.25
Layer 4
K - 0.25 0.5
S - 2.50E-05 2.5e-5 - 2.5e-3
b - 25 -
IB1
T 7 - 9 2.5
(fracture zone 50) 5 - 10
(fracture zone 100)
Layer 5
Marble
T of competent marble ranged from 0.05 to 9 m2/day. A zone of fracturing in the upper section of the marble was assigned a T of 50 m2/d (K = 1 m/day). This zone was not tested in the field.
The geometric mean of the field test-based values of T across both IB1 and IB2 is 1.6 m2/d, i.e., giving an overall range and mean for the Marble of 0.05 - 9 (1.6).
K 0.11 - 0.15 0.05
(fracture zone = 1) 0.1 - 0.2
(fracture zone = 2)
S 0.0001 0.0001 0.00001 - 0.001
b 63 50 -
IB2
T 0.05 - 0.64 2.5 10
Layer 5 K 0.001 - 0.016 0.05 0.2
S - 0.0001 0.00001 - 0.001
b 40 50 -
IB5
T 0.54 - 1.5 1 0.5 - 2.4
Layer 6 Tapley Hill -
Footwall
The TH Fm was split across Layer 6 and 7 to separate the footwall for the mine decline (which was assigned to Layer 6). The deeper portion of the TH is represented by Layer 7 and is several hundred metres deep. T and K for Layer 6 was based on the water bearing section of the footwall intercepted by IB 5 (open hole from 270-294 m).
K 0.02 - 0.06 0.025 0.0125 - 0.06
b 24 40 -
Private (100.58 m)
T 11 1.7 0.64 - 16.75
layer 1 Kanmantoo
This bore targets the upper most water bearing section of the Kanmantoo, which deepens to several hundred metres. The field-test T is reflective of the upper water bearing section, whilst the model T is lower as the entire thickness of the formation is represented as a single model layer.
Coffey (2007) indicate that the rock mass hydraulic conductivity is in the order of 0.01 m/day but could be as low as 0.001 m/day. The pumping test conducted on bore 6628-8301 indicated localised areas of higher permeability are possible in areas of faulting.
6628-8301 K 0.001 - 0.01
(Coffey, 2007) 0.005 0.0019 - 0.05
b - 335 -
Numerical Model Development
Private (50 m) T 5 5 -
Layer 2 Tarcowie Siltstone (shallow)
In the absence of bore casing details, b of 6628-8945 was based on SWL - Bore depth (7.89m -50m).
6628-8945 K 0.09 0.1 -
b 54 50 -
‘Nest Egg’ 6628-20475
(90m)
T 175 75 -
Layer 3 Tarcowie
Siltstone (old mine workings)
Private well intercepts old mine workings, likely resulting in elevated K values compared with broader-scale uniform value employed in model (local-scale heterogeneity not represented due to distance from Project site).
K - 3 -
b No bore info 25 -
Private T 6 5 -
Layer 6 Tapley Hill Production
Interval
The TH Fm was separated into two layers: Layer 6 (productive zone) and thicker Layer 7 (fracture extinction zone), which was assigned K values of one order of magnitude lower than the values of Layer 6. The T values of private well represent the upper water bearing production zone of the TH (Layer 6) and not conditions several hundred meters deep (Layer 7). Across the main irrigation area, the TH Fm is targeted by private pumping wells at an average depth of 85 m, which were assigned to Layer 6.
6628-8939 (80 m)
K 0.09 0.12 -
b 68.22 40 -
* Refer to Table 5 in the main body of the report for comprehensive documentation of field test-based hydraulic parameters.
** () = geometric mean.
Numerical Model Development
Figure F7. Log10 of horizontal hydraulic conductivity (Kh) for all model layers.
Layer 2
Log10(Kh)
Layer 1
Layer 3 Layer 4
Layer 5 Layer 6/7
Numerical Model Development
Figure F8. Log10 of vertical hydraulic conductivity (Kv) for all model layers.
Log10(Kv)
Layer 1 Layer 2
Layer 3 Layer 4
Layer 5 Layer 6/7
Numerical Model Development
5. Model calibration
Calibration is the process by which model parameters and boundary conditions are
adjusted, within realistic limits, to attain an acceptable match between simulated and
observed/measured field data. The realistic limits on parameter values and boundary
conditions are constrained by the ranges of measured values from pumping tests and
other hydrogeological investigations, as well as more general hydrogeological
knowledge.
Model calibration performance is assessed via several metrics, both quantitative (e.g.,
explicit field measurements) and qualitative (e.g., pattern-matching). A summary of
the calibration processes undertaken in the present work is as follows, with complete
details of each process provided in the subsequent sections:
• Steady-state calibration: comparison of modelled and observed hydraulic
heads for pre-mining regional groundwater conditions. This process primarily
informed hydraulic conductivity, recharge, GHB package and river package
parameters.
• Local-scale transient calibration: comparison of simulated and observed
hydrographs for 17-day constant rate drawdown test (CRDT) at BiH (i.e., five
days of pumping followed by 15 days of recovery). This process primarily
informed local-scale (in close proximity to the mine site) hydraulic conductivity
and specific storage parameters.
• Regional scale transient calibration: comparison of simulated and observed
hydrographs for a two-year span including seasonal pumping-induced
seasonal groundwater level fluctuations. This process primarily informed
regional-scale specific storage.
Figure F9 depicts the locations of calibration targets employed for each of the above
calibration processes.
Whilst each of the above calibration processes largely targeted different groups of
model parameters/inputs, it was a necessarily iterative process. For example, model
parameter/input adjustments made during the regional transient calibration process
were fed back into the earlier calibration processes to ensure consistency.
It is common practice to withhold part of a transient groundwater level dataset from a
transient calibration process and subsequently use the withheld data to validate a
model’s predictive capabilities. However, the record-length of the available regional
transient groundwater level dataset was considered to be too short (spanning two
years) to justify this practice. The entire dataset was considered to be more usefully
dedicated to the regional transient calibration process.
Given that the record length of available regional data has substantially increased
since the time of the transient regional calibration process presented herein, it is
recommended that the transient regional calibration process is repeated in future. This
may also include introduction of seasonally variable model inputs such as recharge
(based on rainfall records) and pumping (if such estimates are possible) as a means
of attempting to reproduce temporal fluctuations with greater accuracy.
Numerical Model Development
In lieu of a transient model validation process by means of withholding a portion of the
dataset, additional available information provided an opportunity for model validation.
This included:
• Estimated baseflow to Inverbrackie Creek obtained through catchment
modelling undertaken by the Department of Water, Land and Biodiversity
Conservation (DWLBC) (Zulfic et al., 2002);
• Mapped locations of springs observed during a ground survey in April 2015
(details presented in Section 4.3 and Figure 34 of the main body of the
report);
• Anecdotal information pertaining to historical mining operations in two nearby
mines (i.e., ‘The Ridge’ and ‘Two in the Bush’).
5.1. Steady-state calibration
Where possible, it is common modelling practice to establish steady-state
groundwater conditions through simulation of an aquifer system in its “pre-
development” state (i.e., in the absence of anthropogenic stresses). A steady-state
simulation provides a basis for establishing model boundary conditions and estimating
parameters such as hydraulic conductivity through calibration. It also provides initial
conditions for subsequent transient modelling (e.g., transient calibration and predictive
mining simulations).
The system under study has been exposed to historical stresses including past mining
activities and ongoing private groundwater pumping. However, data suggests that
there exist no relic impacts from previous mining activities. Furthermore, temporal
groundwater levels indicate that the groundwater system effectively makes a full
annual recovery from private pumping stresses, which occur during the summer
months (see Figure 26 of the main body of the report). For these reasons, the
groundwater conditions in spring (i.e., when the system has recovered from the
previous summer’s pumping and prior to the commencement of the next summer’s
pumping) is considered to be an effective representation of the pre-development
steady state of the system. The observed groundwater conditions in late August 2014
are adopted for this purpose.
Steady-state calibration was carried out using data from 35 bores within the model
domain with available records of potentiometric head measured in late August 2014. A
steady-state approximation is considered reasonable as long-term regional
groundwater level trends are generally stable (see Figure 25 and Figure 26 in the
main body of the report). September groundwater level measurements were used to
represent pre-mining steady-state conditions in order to minimize the influence on the
water table caused by regional groundwater pumping from private users (which
occurs predominantly during summer months, with the system effectively recovering
completely over the winter period). That is, pumping for irrigation causes seasonal
fluctuations, but no long-term decline.
The majority of the 35 steady-state calibration bores are within the Tapley Hill
formation (model layer 6), with a smaller number in shallower layers including the
Brighton Limestone (Marble; model layer 5), Tarcowie Siltsone (model layers 3 and 4),
Cox Sandstone (model layer 2) and Kanmantoo Formation (model layer 1).
Numerical Model Development
Figure F9. Locations of calibration targets.
Numerical Model Development
Calibration was sought through adjustment of hydraulic conductivity, recharge and
boundary conditions (including general head boundaries and riverbed conductance).
This was effected largely through manual parameter adjustment. Part of the
calibration process was expedited through use of state-of-the-art automated
calibration software PEST (Doherty, 2016), with appropriate user-intervention to
ensure PEST-adjusted variables remained realistic. Furthermore, particular care was
taken to ensure the simulated regional flow pattern reflects the current understanding
of key conceptual components of the hydrogeological environment, including key
recharge/discharge zones and general regional flow patterns.
A scatterplot of modelled versus measured pre-mining (post-Winter) groundwater
elevations (at the locations denoted in Figure F9) is presented as Figure F10. A
commonly employed statistic for assessing the goodness-of-fit resulting from a
calibration process is SRMS (scaled root mean squared error). The SRMS for the
steady-state calibration scatterplot in Figure F10 is 6.0%. The Australian groundwater
modelling guidelines (Barnett et al., 2012) highlight that goodness-of-fit is subjective.
However, an SRMS of <10% is referred to as an example target within these
guidelines, which is satisfied by the current steady-state calibration process.
Figure F11 displays the post-calibration simulated steady-state hydraulic head
distributions in model layers 1-6 (Figure F11b represents layers 2-5 as these head
distributions are equivalent), and Table F4 provides the steady-state water balance.
Figure F12a displays an overlay of the simulated pre-mining steady-state hydraulic
head distribution upon the interpolated measured hydraulic head distribution displayed
in Figure 6a. This provides a means of spatial comparison between simulated and
measured hydraulic heads. Good agreement is demonstrated across the majority of
the study area (noting that some aspects of the interpolated measured head
distribution, particularly areas of high curvature, may be an artefact of the kriging
method and not reflective of the actual head distribution).
A notable discrepancy in Figure F12a between the observed and simulated head
distribution occurs in the form of the interpolated ridge of groundwater elevations
greater than 420 mAHD (peaking at over 430 mAHD) in the central-southeast, which
is not observed in the groundwater elevations within the Tapley Hill formation. The
groundwater level observations pertaining to this ridge are from bores completed in
the Kanmantoo Formation, below which the Tapley Hill dips to the east of the mine
site (see Figures F1-F3). These formations (the stratigraphic structure and
characteristics of which are not currently well known) are currently represented
collectively by model layer 1. The vertical connectivity between these units and the
underlying units is believed to be relatively low, as reflected by the Kv value for this
layer (see Figure F8). Recharge in this area is indicated by CMB estimates to be
relatively high in this area, believed to be caused by localised surface water ponding,
which results in a local groundwater level high in this area within the upper layer.
Figure 12b displays the model layer 1 groundwater level contours overlayed on the
interpolated regional elevations, showing relatively good agreement in this region.
Numerical Model Development
Figure F10. Scatterplot of simulated versus observed hydraulic head for pre-mining steady-state calibration.
Table F4. Water balance for pre-mining steady-state calibration model.
Flow term In (ML/yr) Out (ML/yr) In-Out (ML/yr)
Recharge 475 0 475
River leakage 16 732 -716
General head boundaries 937 686 251
Sum 1428 1418 10
Discrepancy (%) 0.72
Numerical Model Development
Figure F11. Simulated post-calibration pre-mining steady-state (August/September) hydraulic head distribution (mAHD) in a) the Kanmantoo Formation (Layer 1), b) the Cox Sandstone, Tarcowie Siltstone and Brighton Limestone formations, and c) the Tapley Hill Formation.
Numerical Model Development
Figure F12a. Overlay of interpolated (kriged) observed regional groundwater flow pattern and simulated pre-mining steady-state (Spring, 2014) flow pattern in the Tapley Hill. (Figure F6 displays locations of wells used for observed groundwater level kriging and Figure F9 displays locations of wells used as steady-state calibration targets.)
Numerical Model Development
Figure F12b. Overlay of interpolated (kriged) observed regional groundwater flow pattern and simulated pre-mining steady-state (Spring, 2014) flow pattern in the Kanmantoo Formation. (Figure F6 displays locations of wells used for observed groundwater level kriging and Figure F9 displays locations of wells used as steady-state calibration targets.)
Numerical Model Development
5.2. Transient calibration – CRDT
One of the two transient calibration processes utilized groundwater level data
obtained during a 5-day CRDT (and subsequent 15 days of recovery) at the Project
site. The groundwater level change in 36 wells, including 21 private wells (see Figure
6 in the main body of the report), was recorded as water was pumped from the HW
fracture zone (via well IB-4) at a rate of 17 L/s (1.5 ML/day). These bores target a
range of different hydrostratigraphic units, with a range of depths spanning
approximately 30-470m bgl (specific details are provided in Table 3 of the main body
of the report). This data was used for comparison with simulated groundwater
response for calibration purposes. The CRDT performed on IB-4 was simulated in the
model via the well package in MODFLOW.
Figure F13 presents the comparison between simulated and observed drawdowns in
response to pumping from IB-4. 25 wells exhibited no groundwater level response to
the pumping of IB-4, of which only the nearest 5 to the Project site are included in
Figure F13 for the sake of clarity (4 of which are private wells).
Figure F13 shows that the rates and magnitudes of observed drawdowns in most
wells are reasonably well represented by the model, with the exception of three wells,
namely BH35, BH36 and BH42, for which the fit is comparatively poor (3-4 m). BH35
and BH36 are both located approximately 100 m north-west of the Project site, in the
vicinity of the old underground mine workings. This may be causing a localised
influence resulting in the drawdown error in these two wells. Moreover, model
hydraulic conductivities are presently represented as isotropic. The underestimation of
drawdown at these two wells in addition to BH42 may indicate local anisotropy in
hydraulic conductivity with a northwest-southeast primary axis. This is common in
fracture zones due to local-scale directional fracture connectivity. (An equivalent effect
may be caused by a greater degree of aquifer compartmentalization caused by flow
barriers.)
Figure F13 demonstrates that groundwater level recovery is well-represented in most
wells. Some small discrepancies exist in the form of model-based underestimation of
the recovery rate, indicating that the model is conservative in this manner.
Numerical Model Development
Figure F13. Simulated versus observed drawdowns for transient CRDT calibration.
Figure F14 provides a conceptual representation of the local drawdown in the HW
fracture zone, overlayed upon the currently simulated drawdown and including the
spatial locations of BH35, BH36 and BH42. It is clear that this would cause a relative
increase in drawdown in all three of these wells (and thus improved calibration).
Whether or not this local-scale anisotropy is represented in the model is not expected
to significantly influence the simulation of key mining impacts such as mine inflows
and drawdown in the Tapley Hill formation. Nonetheless, incorporation of this
anisotropy should be a consideration for future model updates.
Numerical Model Development
Figure F14. Conceptual representation of locally anisotropic drawdown in HW fracture zone.
Figure F15 isolates modelled versus observed drawdowns for the transient CRDT
calibration for the 5 nested investigation (IB-) bores as well as BH41, each of which
target a different depth (and formation). This highlights the quality of calibration
performance in terms of capturing interconnectivity between hydrostratigraphic units
at the Project site, which is critical to predicting the contribution of mine inflows from
other units that are not directly depressurized by mine decline and drive construction
(i.e., Tapley Hill and Marble). Thus the CRDT behaviour of the wells isolated in
Figure F15 in particular was key in inferring values for Kv (see Figure F8).
The SRMS for the CRDT transient calibration across all wells is 12.6%. SRMS
considering only these 6 nested wells is 6.7%.
Numerical Model Development
Figure F15. Simulated versus observed drawdowns for transient CRDT calibration, isolating 6 nested wells nearest to (and including) IB-4.
5.3. Transient calibration – regional abstraction
Regional transient calibration was undertaken utilizing irrigation pumping data
(estimates of which were obtained during the groundwater census – see Figure 33 of
the main body of the report) and routinely recorded transient groundwater level data
spanning a two-year period (from late-August 2014 to September 2016) in 28
monitoring bores throughout the Inverbrackie Creek sub-catchment (locations
displayed in Figure F9).
The simulated steady-state groundwater conditions (calibrated to late-August 2014
groundwater level data as described above) were used as initial conditions for the
transient model. The equivalent two-year period was simulated with seasonal
irrigation pumping data supplied as a model input (via MODFLOW’s well package).
Model-versus-measurement comparison for the regional transient calibration
concerned deviation of groundwater level relative to observed/simulated late-August
2014 groundwater level (i.e., the time adopted to represent pre-development steady-
state conditions as discussed above). Any consistent offsets in transient trends are a
function initial RSWL discrepancies. Figure F16 provides an example of this offset in
two wells.
Numerical Model Development
Figure F16. Example of initial model-versus-measurement RSWL offset in two
wells.
The minimization of these initial offsets was achieved to the greatest extent possible
during the steady-state calibration process above. Such offsets are therefore not
relevant to the transient calibration. Thus, deviations in groundwater level are
presented rather than absolute RSWL (this is equivalent to the comparison of
modelled and measured drawdown for the CRDT transient calibration above).
Figure 17 displays the model-measurement comparisons for all 28 monitoring bores.
The hydrographs presented in Figure F17 are presented in order of increasing
distance from the Project site (nearest bores presented first). Inspection of Figure F17
demonstrates that, despite a few exceptions, the model generally reproduces
seasonal pumping-induced fluctuations. The model-versus-measured fits that appear
to be (qualitatively) weakest correspond generally with bores that are located farthest
from the Project site (highlighted in Figure F18), and thus they are less relevant to
mining impact predictions.
6628-23611 IB-3
Numerical Model Development
Figure F17. Simulated and observed hydrographs (mAHD) for regional pumping transient
calibration. Stratigraphic unit targeted by each bore is indicated (CS – Cox Sandstone; TS –
Tarcowie Siltstone; BL – Brighton Limestone; TH – Tapley Hill; MQ – Mitcham Quartzite).
Numerical Model Development
Figure F17 (cont.). Simulated and observed hydrographs (mAHD) for regional pumping
transient calibration. Stratigraphic unit targeted by each bore is indicated (CS – Cox
Sandstone; TS – Tarcowie Siltstone; BL – Brighton Limestone; TH – Tapley Hill; MQ – Mitcham
Quartzite).
Numerical Model Development
Figure F17 (cont.). Simulated and observed hydrographs (mAHD) for regional pumping
transient calibration. Stratigraphic unit targeted by each bore is indicated (CS – Cox
Sandstone; TS – Tarcowie Siltstone; BL – Brighton Limestone; TH – Tapley Hill; MQ – Mitcham
Quartzite).
Numerical Model Development
Figure F17 (cont.). Simulated and observed hydrographs (mAHD) for regional pumping
transient calibration. Stratigraphic unit targeted by each bore is indicated (CS – Cox
Sandstone; TS – Tarcowie Siltstone; BL – Brighton Limestone; TH – Tapley Hill; MQ – Mitcham
Quartzite).
Numerical Model Development
Figure F17 (cont.). Simulated and observed hydrographs (mAHD) for regional pumping
transient calibration. Stratigraphic unit targeted by each bore is indicated (CS – Cox
Sandstone; TS – Tarcowie Siltstone; BL – Brighton Limestone; TH – Tapley Hill; MQ – Mitcham
Quartzite).
Numerical Model Development
Figure F18. Regional transient calibration targets (equivalent to those displayed in Figure F8), indicating locations of poorest fits (green
arrows).
Numerical Model Development
The goal of transient calibration is to improve model-based reproduction of temporal head variation,
which is important as aquifer dynamics are critical to model forecasting ability (e.g., Peeters, 2011).
Figure F17 demonstrates that the responsiveness to regional abstraction and the magnitudes of
fluctuation are reasonably well represented.
As well as aquifer dynamics, spatial gradients (defined by absolute head values) are also of high
importance to model predictive ability. Characterization of these spatial gradients is the role of the
steady-state calibration process (presented above in Section 5.1). Figure F19 presents a comparison
of the simulated head distribution in the Tapley Hill formation between Spring steady state conditions
and post-summer (maximum regional abstraction-induced drawdown) conditions. This qualitatively
demonstrates that spatial gradients are not significantly altered by regional abstraction. Furthermore,
Figure F20 presents the modelled-versus-observed head scatterplot for post-summer 2015, overlayed
on the Spring 2014 scatterplot (equivalent to Figure F8) for comparison. This demonstrates that the
regional comparison between simulated and observed heads, and thus regional spatial gradients,
remains adequate. (Figure F20 is associated with a slight increase in SRMS from 6.0% to 7.8%
between the Spring and post-summer GWLs.)
The chief intention of the regional transient calibration process was to ensure that there exist no major
systematic errors in simulated groundwater level fluctuations. That is, bores in which larger
fluctuations are observed generally exhibit larger simulated fluctuations, and the muted seasonal
responses in a number of wells (e.g., as observed in the site investigation bores) is also reflected
within the model.
Uncertainty in regional groundwater abstraction data precluded a rigorous transient regional
calibration process. It was not possible to wholly quantify all groundwater use within the sub-
catchment, for example private abstraction for domestic use is not accounted for (this comprises a
minor proportion of overall groundwater use). The observed groundwater level behaviour (compared
with simulated behaviour) within some observation bores suggests an influence from pumping that
was not captured during census. These comparisons may be used for future revisions of regional
abstraction estimates. Furthermore, details of pumping frequency and intensity are not known. It is
presently assumed that pumping is distributed uniformly across 6 months of each year.
It should be noted that the model does not exhibit errors in the form of both overestimation and
underestimation of observed data. Almost every case of model-to-measurement disagreement occurs
in the form of model underestimation of observed groundwater level fluctuation. The tendency for
underestimation is expected as it is known that groundwater use within the sub-catchment is not
wholly accounted for as discussed above, and thus total abstraction is known to be underestimated
within the model to some degree.
Finally, a slight overall downward trend in observed groundwater levels exacerbates misfit in a number
of cases. This is likely due to climatic influences (e.g., inter-annual recharge variation) that are not
presently included in the regional transient calibration simulation.
Numerical Model Development
Figure F19. Simulated Spring and Summer groundwater elevations (mAHD).
Numerical Model Development
Figure F20. Comparison of model-versus-observed head scatterplot between Spring and post-summer.
During the regional transient calibration process, specific storage (Ss) was adjusted,
but unlike some other model parameters which are spatially distributed (e.g., hydraulic
conductivity, recharge), all storage parameters (which control the magnitude of
groundwater level fluctuations) presently remain spatially uniform within each model
layer. Heterogeneous storage parameters could be introduced in order to improve the
match between simulated and observed seasonal fluctuations in individual wells.
However, due to the uncertainty in seasonal pumping model inputs, spatially variable
storage parameters may therefore play surrogate roles and simply “mask” model
imperfections arising from other sources. This has the potential to degrade model
predictive ability. Additionally, as discussed above, the length of available hydraulic
head timeseries data does not support a highly robust transient calibration process.
Figure F21 displays the change in SRMS with order-of-magnitude variation in Ss of
the Tapley Hill Formation. The change in SRMS is relatively small, nonetheless the
“final” model (labelled as “base case” in Figure F21) was selected as the local
minimum in the SRMS curve. Predictive mining impact simulations for both a one
order-of-magnitude increase and decrease relative to the base case were conducted
as part of the sensitivity analysis (presented in Section 9.5 below).
Numerical Model Development
Figure F21. SRMS variation with change in Tapley Hill Formation specific
storage Ss (+/- represents order-of-magnitude increase/decrease relative to base
case (see Table F1)).
6. Model validation
As discussed above, additional available information pertaining to aspects of the
hydrogeological behaviour of the Inverbrackie Creek sub-catchment provided some
opportunities for model validation. This information included a previously published
2002 estimate baseflow to Inverbrackie Creek. Augmenting the validation process
regarding baseflow is the mapped locations of springs based on an April 2015 ground
survey. Additionally, anecdotal information is available pertaining to historical mining
operations in two nearby mines. The details and results of the model validation phase
are presented in this section.
6.1. Baseflow
The simulated baseflow to Inverbrackie Creek (and its tributaries) in pre-mining
steady-state conditions is approximately 732 ML/y. A long-term average baseflow
estimate of 874 ML/y for the Inverbrackie Creek sub-catchment is published by
DWLBC (Zulfic et al., 2002). Baseflow simulated by the current model is adequately
comparable with the published DWLBC value, particularly considering the latter is
itself uncertain as it was estimated based on catchment modelling rather than through
streamflow data (due to Inverbrackie Creek being ungauged).
Figure F22 displays simulated baseflow to Inverbrackie Creek for the duration of the
two-year regional transient calibration period, which is shown to fluctuate seasonally
between approximately 2 ML/d and approximately 1.7 ML/d due to seasonal
groundwater abstraction.
Numerical Model Development
Figure F22. Modelled baseflow (ML/d) to Inverbrackie Creek for transient regional calibration simulation period.
6.2. Springs
As discussed in Section 4.3 of the main body of the report, a ground survey
conducted in April 2015 identified a number of pools along the Inverbrackie Creek
(photographs are presented in Appendix D2).
Comparison of the locations of simulated baseflow at the end of Summer 2015 (based
on the transient regional calibration simulation – see section 5.3 above) with the
locations of observed springs adds a qualitative spatial component to the above
quantitative model validation regarding baseflow. This comparison is presented in
Figure F23. Higher negative river leakage values (i.e., baseflow) within river package
model cells occur in similar locations to observed springs. Some simulated summer
baseflow occurs in locations where springs were not mapped in April 2015. Further
field investigations are recommended to fine-tune this aspect of the model during
future updates.
Figure F23. Comparison of a) spring locations mapped during an April 2015
ground survey (reproduced from Figure 34 of the main body of the report) and
b) simulated post-Summer 2015 river leakage.
Numerical Model Development
6.3. Historical anecdotal observations
Anecdotal information pertaining to historical mining operations provided an
opportunity for model validation. Diary entries were made by the mine manager during
the period spanning the 1880’s to 1930’s regarding mine inflows and the groundwater
declines observed in two nearby mines, namely ‘The Ridge’ and ‘Two in the Bush’.
The locations of these historic mines is indicated in Figure 1 of the main body of the
report (The Ridge mine is labelled as ‘Ridge Gold’ in this figure).
Anecdotal records revealed that active operations commenced on the 2nd of January
1891 with pumping commencing on 17th January. The mine manager noted that in
March deepening of the Victoria Shaft below 5 level commenced. On the 23rd May the
groundwater level was noted at 18 feet below 5 level (equating to ~113 m). The mine
manager noted on that day the water levels since pumping began had lowered 40 feet
(approx. 12 metres) at Two in the Bush (located 614 m to the north) and 15 feet
(approx. 4.5 metres) at The Ridge (located 363 m to the south).
The equivalent pumping scenario was simulated within the model (see Figure F24)
and the groundwater level drawdowns at the locations of these two mines compared
with the anecdotal evidence. The predicted drawdowns are considered to be in
adequate agreement with the reported anecdotal observations.
Figure F24. Simulated drawdown at ‘Two in the Bush’ and ‘The Ridge’ mines for historical validation.
Additionally, the hanging wall fracture targeted by IB-4 is interpreted to intercept the
old mine workings. It is thought that this water-bearing structure was the potential
cause of high inflows which ultimately led to the suspension of mining at the historical
Bird-in-Hand mine in the 1930’s. This was also emulated within the calibrated model
via placement of a 10 m section of mine drive (mine drive simulation details are
described in the next section) within the model layer representing the hanging wall
Numerical Model Development
fracture (i.e., modal layer 3). The resultant simulated inflows are presented in
Figure F25. Initial inflows upon hanging wall fracture interception are slightly greater
than 60 L/s, gradually decreasing to slightly less than 40 L/s over a 125-day
simulation period. (Note that the overall time axis in Figure F25 is not relevant in this
case, as it relates so the base simulation from which the validation simulation was
developed.)
Figure F25. Simulated drawdown at ‘Two in the Bush’ and ‘The Ridge’ mines for historical validation.
7. Predictive modelling For all predictive scenarios, the simulated steady-state groundwater flow conditions
were employed as initial conditions. The annual regional extraction cycle was not
included in predictive simulations in order to isolate the simulated impacts caused by
underground mining and depressurization activity. All predictive mining impact
simulations include 5 years 3 months of mining, followed by a subsequent 5 years of
post-mining recovery.
7.1. Underground mine representation
Underground mining and depressurization activity is represented in the model using
the drain package in MODFLOW, this being a common approach (Zaidel et al., 2010).
Drain cells within the Tapley Hill formation and Brighton Limestone (Marble) are used
to simulate the decline and mine drives, respectively. (This refers to the base case –
mine drive segments within the HW fracture zone are also simulated as part of the
sensitivity analysis.)
Drain cell geometries emulate mine workings plans and are temporally variable,
progressing in accordance with the mine plan as detailed by Figure 46 in the main
body of the report. Figure F26 provides an example of the drain cell-based
representation of the underground mine progression in the model. Table F5 details the
model stress period structure for predictive mining impact simulations.
As discussed in the main body of the report, one of the proposed water management
options for minimizing groundwater inflows into the underground mine includes
grouting ahead of mine development (see Section 5.2.1 of the main body of the
Numerical Model Development
report). This mitigation measure was evaluated via the numerical modelling, the
numerical implementation of which is explained in the next subsection.
Table F5. Predictive mining simulation stress periods and time stepping.
Stress period Stress period length Time (days) Time steps Mine stage
1 90 90 10 Decline
2 120 210 10
Decline + 3 Drives 3 120 330 10
4 125 455 10
5 120 575 10
Decline + 6 Drives 6 120 695 10
7 125 820 10
8 120 940 10
Decline + 9 Drives 9 120 1060 10
10 125 1185 10
11 120 1305 10
Decline + 12 Drives 12 120 1425 10
13 125 1550 10
14 120 1670 10
Decline + 15 Drives 15 120 1790 10
16 125 1915 10
17 1826 3741 50 Recovery
Numerical Model Development
Whilst transient evolution of the mine was simulated at an overall rate in accordance
with the current mine plan (as detailed in Table F5 and exemplified by Figure F26), its
progression is unavoidably stepwise in nature (without employing an exorbitantly large
number of stress periods). For example, the beginning of a given stress period
instantaneously gives rise to a string of drain cells representing a new mine decline
segment and accompanying entire length of drive. That is, for each period detailed in
Table F5, a total of up to several hundred metres of underground mine length
effectively appears “overnight”. This is an unavoidable consequence of constructing a
model that is numerically tractable with practically reasonable run times.
The use of arbitrarily large drain conductance values results in temporary extreme
“spikes” in simulated inflows during the first timestep of each stress period (with a
duration in the order of one day of simulation time) due to the instantaneous
availability of an unrealistically large surface area through which groundwater inflow
may occur (rather than the more gradual mine progression that will occur in reality).
This is associated with spurious model behaviour including numerical oscillations (i.e.,
unrealistic model-generated fluctuations in inflow rate that do not have a physical
basis). Further compounding this unrealistically large underground mine surface area
are model layer thicknesses (25 m for the Marble and 40 m for the Tapley Hill) which
are substantially greater than the expected height of mine workings (approximately
5 m).
In order to manage the spurious early-time behaviour of simulated inflows caused by
instantaneous stepwise increases in underground mine surface area, reduced drain
conductance values were employed. Multiple simulations were undertaken in a trial-
and-error process, starting with an arbitrarily large value for drain conductance and
progressively reducing its value until simulated inflows stabilized and exhibited typical
recession curve characteristics.
Figure F27a provides a graphical example of the trial-and-error reduction of drain
conductance undertaken in order to attain a stable simulation of mine inflows. (A
snapshot of approximately one year of inflows on a truncated scale is displayed for
the sake of visual clarity.) Following this analysis, a drain cell conductance value of
1 m2/d was employed for simulation of unmitigated mine inflows.
Figure F27b displays the relationship (for the unmitigated case) between drain cell
conductance and total mine inflows for the duration of the simulated mine life (5
years). This demonstrates that larger initial mine inflow “spikes” caused by larger
conductance values (and which, as discussed above, are deemed to be an unrealistic
artefact of the simplified stepwise mine progression employed in the model) do not
result in greater total simulated inflows. In fact, total mine-life inflows for C = 100 m2/d
are slightly lower than inflows for the C = 1 m2/d value adopted for the base case
simulation. It follows that simulated post-mining drawdown impacts would not be be
increased through adoption of the larger drain cell conductance values displayed in
Figure 27b.
Numerical Model Development
Figure F26. Time-variant drain (yellow) cell input example snapshots to simulate underground mine progression. Drain cell structure evolves at four-monthly increments as per the mine plan.
Tapley Hill (layer 6) ~ 1 yr Marble (layer 5) ~ 1 yr
Tapley Hill (layer 6) ~ 5 yrs Marble (layer 5) ~ 5 yrs
Marble (layer 5) ~ 2 yrs Tapley Hill (layer 6) ~ 2 yrs
Numerical Model Development
Figure F27a. Mine drain cell conductance C (m2/d) determination example.
Figure F27b. Total inflows (GL) for the simulated mine life (5 years) versus mine drain cell
conductance C (m2/d).
Numerical Model Development
7.2. Grouting scenarios
For representation of the underground mine grouting scenarios, conductance values were reduced
through trial-and-error until the desired percentage reduction in total mine inflows was achieved (i.e., as
dictated by the grouting scenario – i.e.,). Drain cell conductance values representing the 70% and 90%
grouting effectiveness are 0.018 m2/d and 0.0035 m2/d, respectively.
Post-mining recovery is simulated through deactivation of all drain cells. Table details the various
simulated mine grouting scenarios.
Table F6. Mining and grouting scenarios.
Mining Scenario
Decline –
grouting
effectiveness (%)
Active drives –
grouting
effectiveness (%)
Backfilled drives –
grouting effectiveness
(%)
No Mitigation (base case) 0 0 0
No Mitigation (base case) 0 0 100 (post-mining)
Grouting of decline and drives 70 70 70
Grouting of decline and drives 90 90 90
7.3. Managed aquifer recharge (MAR)
In addition to grouting, managed aquifer recharge (MAR) is an additional strategy under consideration for
managing any groundwater that seeps into the mine and offsetting groundwater impacts from mine
inflows. MAR is simulated via the well package in MODFLOW. Several predictive scenarios involving
MAR were simulated, as detailed by Table F7.
Simulated MAR well locations and injection rates are presently conceptual only. Figure F28 displays the
presently adopted locations of MAR wells (8 in total; 3 within the fractured region of the Tarcowie
Siltstone and 5 in the Tapley Hill formation). The simulated combined injection rate of MAR wells during
each stress period is equal to the average total mine inflows for the same stress period. Volumetric
injection rates are distributed evenly amongst all MAR wells, with the exception of ‘MAR2’ in the Tarcowie
Siltsone, which injects 30% of the inflows (all other MAR wells inject 10% each).
The current MAR configuration does not induce artesian conditions, with the exception of up to 4 m within
the cell containing “MAR3_TS” (with 1 m artesian head extending a maximum distance of approximately
100 m to the west of the injection well).
Numerical Model Development
Figure F28. Time-variant drain (yellow) cell input example snapshots
Table F7. Grouting and MAR scenarios.
Mining Scenario
Decline -
grouting
effectiveness (%)
Active drives -
grouting
effectiveness (%)
Backfilled drives-
grouting
effectiveness (%)
Grouting of decline and drives + MAR 70 70 70
Grouting of decline and drives + MAR 90 90 90
Depressurisation of Hanging wall fault and MAR 70 70 70
7.4. Solute transport modelling
The risk of saline groundwater intrusion from the higher salinity FRA of the Eastern Mount Lofty Ranges
(EMLR) (i.e., the adjacent Dawesley Creek sub-catchment) was assessed by solute transport modelling.
Solute transport modelling was undertaken using MT3DMS (Zheng and Wang, 1998). Non-reactive
transport was simulated using the finite difference solution method, employing upstream weighting.
Longitudinal dispersivity αL was set to 100 m and transverse dispersivity αT was set to 10 m (effective
porosity ηe values are displayed in Table F1).
Groundwater salinities were broadly assigned in accordance with salinities obtained from baseline well
sampling. Representative initial salinity zones of 2,400 mg/L and 1,000 mg/L were applied to the EMLR
Layer 3
(Tarcowie)
Layer 6
(Tapley Hill)
Log10(Kh)
Numerical Model Development
and WMLR areas, respectively. The boundary between these two initial salinity zones coincides with the
Backstairs Passage Formation surface geology boundary, as displayed on Figure F37 in which salinity
migration results are shown (the Backstairs Passage Formation is a member of the Kanmantoo
Group/Formation (Geoscience Australia, 2017)).
8. Results This section of the report presents the model predictions associated with simulated mining activity (and
subsequent post-mining recovery). These predictions include mine inflows, drawdown, baseflow to
Inverbrackie Creek, migration of more saline water in the adjacent Dawesley Creek sub-catchment, and
abstraction form the WMLR and EMLR. The predictions are presented in the context of the suite of
different mitigation scenarios (including grouting and MAR) detailed in Section 7 above (and further
discussed in the main body of the report).
8.1. Predicted mine inflows
Simulated mine inflows for the key predictive scenarios summarized in Table F6 and Table F7 are
displayed in Figure F29. The “no mitigation” base case is included in each figure for reference. The late-
time tapering of inflows is a result of mine drives encountering the low-K clay zone within the Marble (see
Figure F7).
As discussed in the previous section, the metric selected to define grouting effectiveness was percentage
of total inflows. That is, 90% effective grout is represented by a MODFLOW drain cell conductance that
results in 10% of the total mine inflow volume simulated during the 5 year mine life. As a result, simulated
mine inflows at any given point in time does not necessarily reflect the specified effectiveness with
respect to percentage. For example, the respective mine inflow rates at the end of each simulation period
are observably greater than 70% and 90% of the unmitigated case. Equivalently, early-time predicted
mine inflows are less than 70% and 90% of the unmitigated case inflow rates.
As is evident from Figure F29, the simulated MAR results in only a very minor increase in predicted mine
inflows (negligible for the 90% effective grouting scenario).
Numerical Model Development
Figure F29. Predicted mine inflows for grouting and MAR mitigation scenarios relative to
unmitigated mine inflows.
8.2. Predicted drawdowns
Figures F30a and F30b display plan-view predicted drawdowns in the Tapley Hill and Tarcowie Siltstone,
respectively, for all simulated mining scenarios. Figure F30c provides an example of mine
depressurization-induced drawdown from a cross-sectional perspective (the unmitigated case is shown
for 1 year, 2 years and 5 years into the mine life).
From Figure F30a and Figure F30b the influence of the various mitigation scenarios in reducing the
predicted drawdown impacts is clear. In particular, the reinjection of mine water through MAR is predicted
to markedly reduce the magnitude and extent of drawdown.
Figures F30 through F33 provide a means of direct comparison between mitigation scenarios. They
display overlays of the 1 m, 5 m and 10 m drawdown contours, respectively, for all mining impact
simulations.
Notably, the predicted drawdown extent is somewhat similar for both the 90% and 70% effective grout
cases when MAR is included. This is attributable to the fact that the volume of water reinjected through
MAR scheme for each stress period is calculated as the volume of mine inflows simulated for the same
stress period (i.e., a greater MAR volume accompanies the 70% effective grouting case relative to the
90% effective grouting case as described in the previous section).
Numerical Model Development
Figure F30a. Predicted drawdowns in the Tapley Hill after 5 years of mining, for all simulated mining scenarios (ML displayed in orange).
Numerical Model Development
Figure F30b. Predicted drawdowns in the Tarcowie Siltstone after 5 years of mining, for all simulated mining scenarios.
Numerical Model Development
Figure F30c. Example of drawdown cross-section (unmitigated scenario) after a) 1 year, b) 2 years and c) 5 years of mining.
a)
b)
c)
Numerical Model Development
Figure F31. Mining impact 1 m drawdown contour range across mining scenarios, Tapley Hill.
Numerical Model Development
Figure F32. Mining impact 5 m drawdown contour range across mining scenarios, Tapley Hill.
Numerical Model Development
Figure F33. Mining impact 10 m drawdown contour range across mining scenarios, Tapley Hill.
Numerical Model Development
8.3. Drawdowns at private well locations
Figure F34 displays, for several mitigation scenarios, timeseries’ of simulated
drawdown at the locations of operational private wells for the duration of the mine life
(and recovery). The influence of simulated MAR is clear. Some wells exhibit a 5-20 m
drawdown for the 70% effective grout case. This is reduced to a maximum of
approximately 9 m with the inclusion of the simulated MAR scheme (detailed in
section 7.3 above). However, it should be noted that the abrupt drawdown in early
2024 is a result of the simulated cessation of MAR coinciding precisely with the
simulated end of mine life. The small time lag associated with the propagation of
drawdown causes an increase in drawdown over a subsequent period of
approximately 50 days. It would therefore be recommended that MAR is continued for
1-2 months following the cessation of mining to negate the potential impacts of
drawdown time lag. Assuming the brief post-mining increase in drawdown is avoidable
in this manner in practice, the inclusion of MAR reduces maximum drawdown to no
more than 2 m for all wells except for 6628-23182, which reaches approximately 4 m.
Moreover, simulated groundwater mounding due to MAR does not exceed 1-2 m at
any private well locations with the exception of 6628-8936, which is in close proximity
to a MAR well in the current hypothetical configuration (this would be avoided in
practice).
Numerical Model Development
Figure F34. Simulated drawdowns at private well locations for 70% effective
grout case, 90% effective grout case, and 70% effective grout case including
MAR.
0
5
10
15
20
25
30
35
40
2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
Dra
wd
ow
n (m
)
Year
6628-18637 6628-22028 6628-20632 6628-15320 6628-9153 6628-10699
GTZ 6628-10245 6628-8940 6628-8950 6628-8936 6628-10944
6628-9154 6628-23182 6628-9152 6628-10249
0
5
10
15
20
25
30
35
40
2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
Dra
wd
ow
n (m
)
Year
6628-18637 6628-22028 6628-20632 6628-15320 6628-9153 6628-10699
GTZ 6628-10245 6628-8940 6628-8950 6628-8936 6628-10944
6628-9154 6628-23182 6628-9152 6628-10249
-10
-5
0
5
10
15
20
25
30
2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
Dra
wd
ow
n (m
)
Year
6628-18637 6628-22028 6628-20632 6628-15320 6628-9153 6628-10699
GTZ 6628-10245 6628-8940 6628-8950 6628-8936 6628-10944
6628-9154 6628-23182 6628-9152 6628-10249
70% effective grout
90% effective grout
70% effective grout + MAR
6628-8308
6628-8308
6628-8308
Numerical Model Development
8.4. Baseflow impacts
Figure F35 presents simulated transient baseflow to Inverbrackie Creek for the
duration of the mine life. Note that (as detailed above) seasonal abstraction is not
included in the mining impact simulations, therefore the plots in Figure F35 represent
changes relative to post-Winter baseflow.
By the end of the mine life, simulated baseflow is reduced by approximately 20%.
Following 5 years’ recovery post-mining, simulated baseflow has partially recovered
but remains 10% lower than its pre-mining level.
Maximum baseflow reduction is reduced to approximately 10% and 5% for the 70%
and 90% effective grout mitigation scenarios, respectively. For both cases a near-full
recovery occurs 5 years post-mining (a 5% reduction remains for the 70% case).
The inclusion of simulated MAR serves to effectively negate any reduction in baseflow
casued by mining, irrespective of whether grouting is 70% or 90% effective.
(Regarding the small abrupt reduction in baseflow immediately following the end of
mining at approximately 5.5 years, the equivalent discussion of MAR cessation
applies to that presented in section 8.3 above.)
Figure F35. Simulated timeseries of baseflow to Inverbrackie Creek for a range
of mining scenarios.
8.5. Recovery
Figure F36 displays, for a range of scenarios, residual drawdowns following simulated
mine closure. The scenario involving 70% effective grout and MAR results in almost
no observable impact 5-years post-mining.
Numerical Model Development
Figure F36. Residual drawdown 5-years post-mining (ML displayed in orange)..
Numerical Model Development
8.6. Solute transport
The simulated post-mining salinity distribution in the Tapley Hill for the unmitigated scenario (i.e., no
grouting or MAR is displayed in Figure F37. The 1,100 mg/L contour line is chosen as a representative
boundary as it defines a small predicted increase in groundwater salinity (i.e., 100 mg/L) relative to the
background salinity. It is clear from Figure F37 that the predicted movement of the saline interface is very
minor (the initial position of which was defined by the western border of the displayed Backstairs Passage
Formation surface geology as described above).
Figure F37. Simulated post-mining salinity distribution.
Numerical Model Development
8.7. Estimated groundwater abstraction from WMLR and EMLR
Section 8 of the main body of the report addresses the proximity of the Project site to the boundary
between the WMLR and EMLR Prescribed Water Resources Areas (PWRA). Assessment the Bird-in-
Hand Gold Project against the WMLR Water Allocation Plan (WAP) requires prediction of groundwater
abstraction volumes from the WMLR PWRA, including effective abstraction from the adjacent EMLR
PWRA in the form of groundwater transfer between the two PWRAs.
Predicted groundwater abstraction values are calculated as averages for the simulated 5.5 year mine life.
Mining-induced transfer of groundwater from the EMLR to the WMLR (the representative zones defined in
the model are displayed in Figure F38) represents abstraction from the EMLR. Mining-induced transfer is
calculated as the change in simulated groundwater flux relative to the simulated pre-development steady-
state net flux of 25 ML/y from the EMLR to the WMLR. Abstraction from the WMLR is calculated as the
average mine inflow rate, minus the calculated ELMR abstraction rate. Table F8 provides the resultant
calculated WMLR and EMLR abstraction values for the four key management scenarios considered in the
present study.
Figure F38. WMLR and EMLR zones as defined in the model
EMLR
PWRA
WMLR
PWRA
Numerical Model Development
Table F8. Model-predicted groundwater abstraction from WMLR and EMLR.
Scenario Groundwater abstraction from
WMLR (ML/y) Groundwater injection
into WMLR (ML/y) Groundwater abstraction from
EMLR (ML/y)
70% grouting -316 - -64
90% grouting -110 - -20
70% grouting + MAR -430 +430 +10 (receiving)
90% grouting + MAR -123 +123 +7 (receiving)
9. Sensitivity/uncertainty analysis
9.1. Summary
Predictive modelling in the present study considers a broad scope of scenarios, ranging from completely
unmitigated mining impacts, to grouting of varying effectiveness coupled with MAR (refer to Table 11 and
Table 12 in the main body of the report). This range of scenarios provides a full spectrum of hypothetical
impacts. Additional sensitivity analysis undertaken to assess the robustness of the model predictions,
providing insight into the potential degree of uncertainty in key model predictions owing to conceptual
model assumptions and potential hydraulic property inaccuracies. This analysis involved assessment of:
• The sensitivity of predictive mining impact simulations to the area of interpreted elevated
recharge immediately to the east of the mine area.
• The significance of the hanging wall fault, which will produce large amounts of groundwater if
exposed. To assess this a 10 m mine-drive segment (i.e., two drain cells) was simulated within
the fracture zone (layer 3) at two representative depths, these being 130 m bgl and 300 m bgl,
and the mine water inflows over a 125-day stress period were quantified.
• The influence of the two main interpreted groundwater flow barriers (represented as low-K strips
north and south of the Project site – see Figure F7 and Figure F8) upon inflows and drawdowns.
This was assessed by removing these barriers from the model (by increasing their K values to
match adjacent K values) and repeating key mining impact simulations (i.e., 70% grouting
effectiveness case with and without MAR) and quantifying mine inflows and drawdowns.
• The effect on mine inflows caused by the presence of the upper weathered zone above the
Tapley Hill Formation, surrounding the decline. Its potential influence was assessed by reducing
Kh in the model to an arbitrarily low value of 1e-4 m/d (resembling the alteration to clay as
simulated at depth within in the Marble – see Table F2).
• Sensitivity to aquifer properties (tested by perturbing Kh, Kv and Ss). Of particular interest is the
influence of the permeability of fracture zones upon mine inflows. This was analysed by
increasing the K of the upper fracture zone (in the Marble) and the hanging wall fracture (in the
Tarcowie Siltstone).
Numerical Model Development
9.2. Recharge
The existence of an area of relatively high recharge immediately east of the mine area was inferred
through CMB analysis and soft-knowledge of localized surface water ponding in this area. Given the
proximity of this elevated recharge feature to the Project site, it was considered pertinent to examine its
potential influence upon key mining impact predictions. To this end, simulated recharge within this zone
was reduced to a value equal to the adjacent zone within which the mine area itself is located (i.e., a
reduction from 80 mm/yr to 10 mm/yr). This degrades steady-state calibration performance, to the extent
that a rigorous statistical uncertainty analysis may discount this simulation due to the low likelihood
associated with the increase in model-to-measurement misfit. (Figure F39 shows the steady-state
calibration scatterplots for both the base case and reduced recharge case; the corresponding increase in
SRMS is from 6.0% (base case) to 8.8% for the reduced recharge case, equating to a 47% increase in
SRMS.) The effect upon this recharge reduction upon predicted mining impacts was nonetheless tested
by re-running the base case (no mitigation) scenario. This required alteration to the pre-mining steady-
state simulation in order to generate the corresponding initial head distribution for the transient mining
impact simulation.
Table F9 provides the water balance for the steady-state calibration simulation with reduced recharge
applied east of the mine site. Notably, simulated baseflow is reduced to 704 M/y (from 732 ML/y for the
base case – see Table F4) and therefore compares less favourably with the DWBLC catchment modelling
estimate of 874 ML/y (Zulfic et al., 2002).
Figure F39. Steady-state calibration scatterplot comparison for the base case (80 mm/yr recharge
applied east of the mine location) and following reduction to 10 mm/yr recharge applied east of
the mine location.
Numerical Model Development
Figure F40 provides a comparison of the simulated pre-development steady-state hydraulic head
distribution for the reduced recharge case with the base case. This demonstrates that the general pattern
of simulated groundwater flow (in particular the representation of the groundwater flow divide between the
WMLR and EMLR) is not heavily controlled by the interpreted zone of elevated recharge east of the
Project site. Absolute heads, however, are somewhat altered by the change in simulated recharge. The
unmitigated mining scenario simulation was repeated for the reduced recharge case to test the sensitivity
of impact predictions to the recharge reduction.
Figure F41 provides a comparison of predicted drawdowns (after 5 years of mining) with and without the
area of elevated recharge immediately east of the mine location. Figures F42a and F42b display plots of
simulated transient mine inflows and baseflows, respectively, for the same two simulations (the latter also
including the simulated 5-year post-mining recovery period). It is clear that this recharge zone is
insignificant in the context of model-based predictions of mining impacts.
Figure F40. Simulated pre-development steady-state head distribution for the reduced recharge
case (a) as compared with the base case (b).
Numerical Model Development
Table F9. Water balance for steady-state calibration model with reduced recharge east of mine.
Flow term In (ML/y) Out (ML/y) In-Out (ML/y)
Recharge 320 0 320
River leakage 18 704 -686
General head boundaries 973 611 362
Sum 1310 1315 -4
Discrepancy (%) -0.31
Figure F41. Simulated 5-year mining impacts in the Tapley Hill (no mitigation scenario) for a)
80 mm/yr recharge applied east of the mine location (i.e., base case), and b) 10 mm/yr recharge
applied east of the mine location.
a) b)
Numerical Model Development
Figure F42a. Simulated mine inflows (no mitigation scenario) for the base case (i.e., 80 mm/yr
recharge applied east of the mine location), and the reduced recharge scenario (i.e., 10 mm/yr
recharge applied east of the mine location).
Figure F42b. Simulated baseflow to Inverbrackie Creek (no mitigation scenario) for the base case
(i.e., 80 mm/yr recharge applied east of the mine location), and the reduced recharge scenario (i.e.,
10 mm/yr recharge applied east of the mine location).
As a further examination of the potential influence of the interpreted zone of elevated recharge upon
model predictions, an additional sensitivity test was undertaken. This involved a reduction in recharge
rate of this zone, accompanied by a proportional reduction in hydraulic conductivity in order to maintain a
constant ratio between recharge and K (this being a widely recognized source of non-uniqueness in
groundwater models). Maintaining this ratio effectively preserves hydraulic heads and therefore does not
affect calibration performance. Therefore, the resultant model predictions are theoretically of equal
likelihood (in the absence of additional data to constrain either recharge or K).
Numerical Model Development
The recharge zone with a base case value of 80 mm/yr was reduced to 30 mm/yr (the latter aligning with
the low-end of the CMB recharge estimates for this zone – see Figure 43 of the main body of the report).
This 62.5% reduction in recharge lead to an increase in SRMS for the steady state calibration from 6.0%
to 7.1%. Hydraulic conductivity in model layer 1 was reduced by a factor of 62.5% (i.e., Kh was reduced
from 0.005 m/day to 0.0019 m/day and Kv was reduced from 0.001 m/day to 3.75e-4 m/day). As per
theoretical expectation, this conservation of the ratio between recharge and K (approximately) preserved
original steady state hydraulic heads hydraulic heads and therefore model-to-measurement misfit
(achieving an SRMS of 6.1% as compared with the original 6.0% for the base case). Figure F43 displays
a comparison of the steady-state calibration scatterplots for each case.
Figure F43. Steady-state calibration scatterplot comparison for the base case (80 mm/yr recharge
applied east of the Project site), case involving reduction to 30 mm/r recharge, and case involving
reduction of both recharge (to 30 mm/yr) and hydraulic conductivity to preserve recharge/K ratio.
Figure F44 displays a comparison of simulated unmitigated mine inflows between the base case and the
case involving reduction of both recharge (to 30 mm/yr in the zone east of the Project site) and hydraulic
conductivity in order to preserve the recharge/K ratio. Figure F44 demonstrates that this has a negligible
impact upon predicted mine inflows. Comparison of simulated maximum (post-mining) drawdowns also
demonstrates negligible differences as compared with the base case, with the exception of model layer 1.
The reduction in Kh and Kv in this layer causes a reduction in simulated drawdown. At the location of the
sole operational well within this layer (i.e., ‘6628-8308’ in Figure F34), maximum drawdown caused by
simulated unmitigated mining is reduced from 9.3 m to 4.6 m. Thus, in this respect, the adopted base
case model is a conservative predictor of mine impacts.
Numerical Model Development
A negligible difference in predicted baseflow for the two cases compared in Figure F44 was observed.
Additionally, the transient CRDT calibration results were compared for these cases, also yielding a
negligible difference. These results are not displayed for the sake of brevity.
Figure F44. Simulated mine inflows (no mitigation scenario) for the base case (i.e., 80 mm/yr
recharge applied east of the Project location), and the scenario involving reduction of both
recharge (to 30 mm/yr) and hydraulic conductivity to preserve recharge/K ratio.
9.3. Hanging wall fracture interception
Figure F45 displays the results of the component of sensitivity analysis targeting the significance of the
hanging wall fracture, showing the simulated inflows from a 10 m local (non-grouted) area within the
hanging wall fracture zone at two representative mining depths: 130 m and 300 m bgl. The simulated
inflows indicate that the hanging wall fracture may produce large amounts of water on the order of 50 L/s
to 150 L/s, if exposed.
The predicted inflows at 130 m bgl are consistent with the high inflows reported during dewatering of the
historic Bird-in-Hand mine (40 to 60 L/s) when the hanging wall fracture was intercepted at roughly the
same depth of 130 m bgl. (Note that the 130 m bgl case also provided an additional means of model
validation against anecdotal evidence, as discussed above.)
Figure F46 displays the simulated additional drawdown in the Tapley Hill caused by each of the HW
fracture interception scenarios, if the inflows (presented in Figure F45) are allowed to persist for a period
of 2 weeks. Drawdown difference is presented as an alternative to comparison of absolute drawdowns, as
absolute drawdowns with and without the simulated fracture interception are essentially indistinguishable
from one another (absolute drawdowns at various simulation times are displayed in Appendix G). From
Figure F46 it can be seen that simulated additional drawdown at 2 weeks following HW fracture
interception (for the 70% effective grout including MAR mitigation case) does not exceed 1 m at the
location of any currently operational wells.
Numerical Model Development
Figure F45. Simulated inflows for hanging wall fracture interception at two representative depths.
Numerical Model Development
Figure F46. Drawdown difference (m) at 2 weeks following simulated interception of hanging wall fracture at 130 m (blue) and 300 m
(orange). (Tapley Hill formation; 70% effective grout incl. MAR scenario.)
Numerical Model Development
9.4. Significance of flow barriers
The potential uncertainty in model predictions associated with the main interpreted horizontal
groundwater flow barriers located to the north and south of the proposed mine could be assessed through
multiple simulations employing various configurations of, and hydraulic properties of, the flow barriers. In
lieu this, predictive impact simulations were repeated with the barriers completely removed in order to
obtain a conservative end-member for the predictive uncertainty range caused by this aspect of the
conceptual model.
Figure F47 displays a comparison of the predicted drawdown distribution in the Tapley Hill with and
without the presence of the key interpreted flow barriers discussed above and in the main body of the
report. Figure F48 presents the equivalent results for the Tarcowie Siltstone. Figure F49 provides an
indication of the influence of the flow barriers upon predicted mine inflows and predicted changes in
baseflow to Inverbrackie Creek.
Whilst the sensitivity of the model predictions to the presence of the flow barriers is observable for the
unmitigated mining scenario, the discrepancies in predictions appear to reduce significantly with the
involvement of mitigation strategies (grouting and MAR).
Numerical Model Development
Figure F47. Comparison of predicted 5-year drawdowns in the Tapley Hill with (top) and without (bottom) flow barriers present.
Numerical Model Development
Figure F48. Comparison of predicted 5-year drawdowns in the Tarcowie Siltstone with (top) and without (bottom) flow barriers present.
Numerical Model Development
Figure F49. Comparison of simulated mine inflows (top) and baseflows to Inverbrackie Creek
(bottom) with and without interpreted flow barriers present.
0
10
20
30
40
50
60
70
80
90
0 1 2 3 4 5 6
Sim
ula
ted
min
e in
flo
w (
L/s)
Time (y)
No mitigation No mitigation, no barriers
70% effective grout 70% effective grout, no barriers
70% effective grout + MAR 70% effective grout + MAR, no barriers
1
1.2
1.4
1.6
1.8
2
2.2
0 2 4 6 8 10 12
Sim
ula
ted
bas
eflo
w (
ML/
d)
Time (y)
No mitigation No mitigation, no barriers
70% effective grout 70% effective grout, no barriers
70% effective grout + MAR 70% effective grout + MAR, no barriers
Numerical Model Development
9.5. Sensitivity to hydraulic parameters
Figure F50 displays the envelope of temporal mine inflows obtained through the sensitivity
analysis with respect to key model hydraulic parameters. The inclusion of the upper weathered
zone in the Tapley hill results in a large early-time difference in inflows. The difference in
maximum inflows across the range of perturbed hydraulic parameters is approximately 10-
15 L/s (i.e., approximately 25% of maximum inflows). It should be emphasized that the
sensitivity analysis was performed on the unmitigated mining simulation in order to best
highlight model sensitivities for the purpose of acquiring understanding. As previously
discussed, mitigation strategies (grouting and MAR) will be employed in practice, which are
shown to substantially reduce mine inflows and drawdowns, and therefore the sensitivity of
simulated inflows and drawdowns caused by hydraulic parameter changes will be
proportionately smaller relative to the differences for the unmitigated scenario presented herein.
In order to reduce clutter, Figure F50 does not display the results of all sensitivity analysis
simulations undertaken. A select group of key simulations has been chosen, which highlight key
sensitivities, including (most importantly) those which define the upper and lower bounds of the
uncertainty envelope.
Figure F50. Envelope of mine inflows based on hydraulic parameter sensitivity analysis.
Figures F51, F52 and F53 display the 1 m, 5 m and 10 m drawdown contours, respectively, for
selected key sensitivity analysis simulations which encompass the best-case and worst-case
scenarios (i.e., for non-mitigated mining).
Numerical Model Development
Figure F51. Mining impact 1 m drawdown contour range based on parameter sensitivity analysis, Tapley Hill (‘OM’ = order of magnitude).
Numerical Model Development
Figure F52. Mining impact 5 m drawdown contour range based on parameter sensitivity analysis, Tapley Hill (‘OM’ = order of magnitude).
Numerical Model Development
Figure F53. Mining impact 5 m drawdown contour range based on parameter sensitivity analysis, Tapley Hill (‘OM’ = order of magnitude).
Numerical Model Development
The results of the sensitivity/uncertainty analysis, in particular comparison of Figures F51-F53 with
Figures F31-F33, suggests that the mining impact uncertainty owing to uncertainties in key model
hydraulic parameters and conceptual uncertainties is relatively insignificant as compared with the
differences in mining impacts that may be achieved through mitigation strategies including varying
degrees of grouting and/or commissioning of reinjection of mine water via MAR.
10. Modelling limitations • As discussed above, the transient CRDT calibration results display a relatively poor model-to-
measurement fit for some wells. As also discussed, this is likely attributable to local-scale
anisotropy. This may be introduced into the model during future upgrades to improve local-scale
drawdown representation, but is not expected to significantly influence larger-scale mining impact
simulation results.
• Data availability to support a transient regional calibration process is relatively limited. That is, the
available two-year timeseries is relatively short (however this is now approaching three years and
therefore transient recalibration using the expanded dataset is a consideration for future model
upgrades), and historical abstraction information is limited due to lack of metering.
• Some model parameters are presently based on a relatively coarse parameterization scheme, for
example:
o Storage parameters within each layer are homogeneous. Introduction of spatially
distributed specific storage would likely enable improved transient calibration
performance (particularly regional variation in seasonal fluctuation magnitude). However,
the merit of doing so is restricted by the currently limited transient data availability
discussed above.
o The current mode of spatial variability for parameters that are spatially distributed (e.g.,
hydraulic conductivity) is piecewise constant zones. This effectively limits the degree of fit
between observed and simulated hydraulic heads that is attainable through calibration
(as compared with a continuous parameterization mechanism such pilot points, for
example). However, as discussed above, the introduction of spatially variable storage
parameters would only be useful if accompanied by reliable, holistic groundwater
extraction data.
• Whilst key model parameters were targeted and highly conservative simulations were
undertaken, the sensitivity/uncertainty analysis could be extended further, incorporating a greater
range of tested parameters, in order to more accurately constrain model prediction likelihoods.