 Home
 Assessments
 Bioregional Assessment Program
 Hunter subregion
 2.6.2 Groundwater numerical modelling for the Hunter subregion
 2.6.2.8 Uncertainty analysis
 2.6.2.8.1 Factors included in formal uncertainty analysis
Section 2.6.2.7 described the available observations, predictions required of the model and sampling of parameter space as design of experiment to train emulators, as well as the sensitivity analysis based on these model runs. The same parameters evaluated in the sensitivity analysis are considered in the formal uncertainty analysis, although the sensitivity analysis indicated that a limited number of these parameters influence the predictions.
As described in companion submethodology M09 (as listed in Table 1) for propagating uncertainty through models (Peeters et al., 2016) and in Section 2.6.2.1, the parameter space is constrained by the observations relevant to the predictions through a Markov chain Monte Carlo sampling using the Approximate Bayesian Computation (ABC) methodology (Beaumont et al., 2002; Vrugt and Sadegh, 2013). As in any Bayesian methodology, a set of prior parameter distributions needs to be defined that encapsulates the current information and knowledge, including correlation or covariance of parameters. This is described in Section 2.6.2.8.1.1.
The prior parameter distributions are constrained by observations with the ABC methodology to generate posterior parameter distributions. The posterior parameter distributions are then used to generate the final set of predictions from which the uncertainty of drawdown due to additional coal resource development (i.e. maximum difference in drawdown between the CRDP and baseline) (dmax) and year of maximum change (tmax) predictions can be characterised. The process of constraining the prior parameter distributions by observations is described and discussed in Section 2.6.2.8.1.2, while the resulting posterior predictive distributions are detailed in Section 2.6.2.8.1.3.
The uncertainty analysis is focused on predictions of dmax and tmax due to the additional coal resource development. The surface water – groundwater fluxes generated along the river network within the groundwater model domain are inputs to the Hunter river model (see companion product 2.6.1 for the Hunter subregion (Zhang et al., 2018)) and included in the characterisation of uncertainty of streamflow predictions in that product. Results from the groundwater modelling are summarised as a water balance across the entire modelling domain in companion product 2.5 for the Hunter subregion (Herron et al., 2018a).
2.6.2.8.1.1 Prior parameter distributions
The model parameters in the sensitivity and uncertainty analysis are not varied directly, but rather through a set of offsets, multipliers or coefficients of depthdependent relationships. The initial values of the model parameters and the ranges of the offsets, multipliers and coefficients listed in Table 8 are based on information available about the groundwater systems, summarised in companion product 2.12.2 (Herron et al., 2018b) and companion product 2.3 (Dawes et al., 2018) for the Hunter subregion. The ranges of parameter values reflect the uncertainty in characterising the system due to spatial variability and incomplete knowledge of the system.
The information available was not deemed sufficient to justify any other distribution than a uniform distribution. The prior parameter distributions of the parameters listed in Table 8 are all uniform distributions with ranges corresponding to the minimum and maximum values given in the table. No covariance between parameters is specified as insufficient information is available to specify such covariance. Assuming no correlation between parameters is a conservative assumption as it will result in wider predictive intervals.
2.6.2.8.1.2 Posterior parameter distributions
The posterior parameter distributions are obtained through a Markov chain Monte Carlo sampling of the prior parameter distributions with a rejection sampler based on the ABC methodology. The rejection sampler only accepts a parameter combination, randomly drawn from the prior parameter distribution, if a model run with these parameters satisfies a predefined objective function threshold. The objective functions for the groundwater model summarise its performance in predicting groundwater levels and baseflow rates against observations of groundwater levels and baseflows estimated from streamflow observations. An example of an objective function is the mean absolute difference between observed and simulated groundwater levels. The ABC methodology requires not only definition of an objective function, but also the threshold value above which the parameter set is deemed to be acceptable. Ideally, this threshold is based on an independent estimate of the observation error.
For the Hunter subregion two types of such observations are available to constrain the parameters for the predictions (see Section 2.6.2.7.1):
 groundwater level observations
 observations of total streamflow.
The sensitivity analysis (Section 2.6.2.7.3) showed that both groundwater level predictions and surface water – groundwater flux are dominated by the drainage level assigned to the river model nodes (d_riv), while predictions of dmax due to the additional coal resource development are mostly sensitive to mine pumping rates and the hydraulic properties, and not to d_riv. The sensitivity analysis also showed considerable variation in the sensitivity indices, reflecting spatial variability in both the observations and model predictions.
To accommodate this, the model parameters for the predictions of dmax and tmax at each model node are constrained individually using an objective function (OF) that consists of two components: (i) a spatially weighted sum of the residuals between observed and simulated groundwater levels, and (ii) a baseflow constraint based on historical streamflow data, defined as:

(14) 
In this equation, is the average historical simulated surface water – groundwater flux at gauge , is the number of gauges with observations of total streamflow, and and are the 20th and 70th percentile of observed streamflow over the historical record at gauge respectively. Any parameter combination that results in an value of less than 0.9 is deemed unrealistic and excluded from the posterior parameter distribution. In other words, only parameter sets which simulate an average historical surface water – groundwater flux between the negative of the 20th percentile of observed streamflow and the 70th percentile of observed streamflow at 90% of the gauges are retained.
River systems in the Hunter subregion are predominantly gaining (see companion product 2.12.2 (Herron et al., 2018b) and companion product 2.3 (Dawes et al., 2018) for the Hunter subregion). However, estimates of the groundwater contribution to total streamflow vary widely and are influenced by the estimation method (e.g. digital baseflow separation, salt balance, environmental tracers). Unless the uncertainty in the estimated surface water – groundwater flux can be characterised robustly and independently, the adoption of a very precise objective function based on the mismatch between the observed and simulated surface water – groundwater fluxes is not justified (Kavetski et al., 2006). Here, a less stringent objective function has been defined to eliminate unrealistic parameter combinations. Based on estimates of baseflow for Hunter subregion streams (see companion product 1.1 (McVicar et al., 2015) and companion product 2.3 (Dawes et al., 2018) for the Hunter subregion), it is considered unlikely that the longterm average groundwater contribution to total streamflow is more than the 70th percentile of the streamflow hydrograph. High percentiles generally correspond to high flows dominated by direct runoff from rainfall. While the system is described as predominantly gaining, some reaches are ephemeral and will at times be losing streamflow. It is unlikely that the average loss from surface water to groundwater averaged over the historical period (1983 to 2012) would exceed the 20th percentile of total streamflow for that period. Streamflow losses greater than this are inconsistent with the magnitude of river losses from the regulated river system.
The first component of the objective function consists of a distanceweighted sum of the residuals between observed and simulated groundwater levels. Thus, an observation close to a prediction location has greater potential to constrain that prediction than an observation further away. The objective function takes this into account by weighting the difference between observed and simulated values of groundwater level based on the distance between the prediction location and the observation and the distance of the observation point to the nearest blue line network – that is, the mapped major river network. The additional criterion for predictions of change in groundwater level in layer 1 is therefore defined as:

(15) 
where is the criterion for prediction j, is the ith observed groundwater level and is the simulated equivalent to this observation. is the distance (in km) of observation to the nearest blue line network while is the distance (in km) between observation and prediction . Coefficient controls the distance at which the weight of observation drops to zero. To account for transient observations, the weights are divided by , the number of observations at the observation location. The tanh function allows the weight of an observation to decrease almost linearly with distance and to gradually become zero at a distance of approximately . This is illustrated in Figure 34 for different values of .
d = distance between observation and prediction (km); w = shortest distance between observation location and blue line network (km)
Data: Bioregional Assessment Programme (Dataset 1)
Ideally, the distance weighting function is chosen based on a spatial analysis of the available groundwater level observations. The sparsity of observations, especially transient observations, did not allow such an analysis. In the uncertainty analysis of the groundwater model for the Hunter subregion the value of is set equal to 5 km, which means that an observation has a weight of zero if the distance between the prediction and the observation is more than 15 km (a pragmatic choice by the project team).
The threshold for each prediction is defined as:

(16) 
This means any parameter combination that results in an average difference between observed and simulated groundwater level equal to or less than 10 m is deemed acceptable. In a traditional calibration this corresponds to a normalised root mean squared error of 2% for groundwater level observations that span a range of about 500 m (Bioregional Assessment Programme, Dataset 1). In other words, each parameter combination that is accepted in the Monte Carlo analysis, will lead to a groundwater model with a normalised root mean squared error of no more than 2%.
Figure 35 Groundwater level objective function threshold
Data: Bioregional Assessment Programme (Dataset 1)
Figure 35 shows the acceptance threshold for the groundwater level objective function for each model node. The value of the threshold is not that important: high values indicate a prediction is within the zone of influence of one or more observations, and this means the parameter distribution used for that prediction can potentially be constrained by groundwater level observations; low values indicate predictions far removed from any observation, which therefore have a low potential to be constrained.
Data: Bioregional Assessment Programme (Dataset 1)
Figure 36 shows the fraction of evaluated parameter combinations of the design of experiment that meets the groundwater level acceptance criterion. Values in excess of 0.9 indicate that most parameter combinations evaluated in the design of experiment produce groundwater level predictions that agree with the observations in that region (within the specified acceptable range). In other words, the simulated groundwater levels in these regions are not very sensitive to the parameter values. In the groundwater model for the Hunter subregion, the high acceptance rates along the Hunter River alluvium are probably caused by proximity to surface water boundaries. Simulated groundwater levels close to the model domain’s river network are affected by the river boundary conditions, rather than the parameter values.
Very low acceptance rates indicate that only a limited number of parameter combinations produce simulated groundwater levels that correspond to observed values at the observation locations. As parameter bounds in the design of experiment are deliberately chosen to be wide, low acceptance values indicate localised shortcomings of the conceptual model – that is, either it is not possible to get an acceptable correspondence between observed and simulated values, or some of the observations are not representative of the regional groundwater flow system. In some extreme cases, the acceptance rate is equal to zero, which occurs in the north of the subregion and east of Mudgee. The latter are predictions within overlapping zones of influence of two observation locations for which the simulated values cannot simultaneously agree with both observed groundwater levels. Although this can relate to shortcomings of the spatially uniform parameter multipliers within a layer, it is also likely due to the boundary conditions, the local conceptualisation, or even artefacts or errors in the observation database.
Figure 36 can be used as a proxy for the conceptual model uncertainty of the groundwater model. Very high acceptance rates indicate insensitivity to parameter values, although the groundwater level predictions broadly agree with observations. Low values, on the other hand, indicate regions where the groundwater model is unlikely to make acceptable groundwater level predictions and the conceptualisation is locally inadequate. Although the extent and shape of these regions is determined by the weighting function, the presence of these zones indicates that for some sets of observations the model is not able to simultaneously match the observations within the prescribed error threshold. A detailed forensic examination of both the observation database and the groundwater model to attribute the mismatch to either observation uncertainty or conceptual model issues has not been carried out.
In the Markov chain Monte Carlo sampling, random parameter combinations are selected from the uniform prior parameter combinations. These parameter combinations are evaluated with the emulator for the objective function for groundwater levels for that prediction location and with the emulator for the surface water – groundwater flux objective function. Only if a parameter combination yields objective function values that meet both thresholds is the parameter combination accepted in the posterior parameter combination for this prediction of dmax and tmax. For predictions where the acceptance rate of the groundwater level observation objective function of the design of experiment parameter values is less than 0.1, the objective function for groundwater levels is not used. As explained earlier, low acceptance rates can be an indication of conceptual model issues or observation error. If only a small fraction of parameter combinations leads to acceptable groundwater levels, it is likely that the observation is overfitted and the uncertainty under estimated. By using only the baseflow objective function to constrain parameter sets for predictions with a groundwater level acceptance rate of less than 0.1, overfitting and underestimating uncertainty are avoided.
The Markov chain Monte Carlo sampling is repeated until 10,000 samples are accepted in the posterior parameter combinations. This number was chosen to be as large as was computationally possible within the time frame of the project. Verification that 10,000 samples is sufficient to obtain robust estimates of the 5th, 50th and 95th percentiles of dmax and tmax is provided in the next section.
The histograms in Figure 37 show the 5th, 50th and 95th percentiles of the 1566 posterior parameter combinations (one for each prediction). Parameters are uniformly sampled so those not constrained by the observations will have the 5th and 95th percentiles close to the minimum and maximum of the prior distribution, while the 50th percentile will be close to the centre of the distribution. The posterior distributions for parameters Q_mine, C_riv, ne_lambda, ne and K_ramp are very similar to their prior distributions and are therefore not constrained by the observations. The posterior distributions of d_riv, RCH, K_lambda, KvKh and Kh differ from their priors. The general trends are for the depth of riverbed below topography to be deeper (d_riv more positive), recharge to be higher and the hydraulic conductivity to be lower (higher K_lambda and lower Kh) and more isotropic (higher KvKh).
Figure 38 shows the spatial distribution of the median of the posterior parameter distributions from which a number of spatial patterns emerge:
 the southeast region, close to Newcastle, trends to higher recharge and lower Kh, indicating groundwater levels are under estimated
 the southwest region indicates lower recharge and higher Kh, indicating groundwater levels are over estimated
 clusters of higher or lower d_riv indicate localised underestimation or overestimation of groundwater levels.
Figure 37 Histograms of the 5th, 50th and 95th percentile of posterior parameter combinations
The prior distributions for each parameter are uniform. A posterior parameter distribution that is not constrained by the available observations (such as C_riv) will have the 5th percentile close to the minimum of the parameter range, the 50^{th} percentile close to the middle and the 95th percentile close to the maximum. Posterior parameter distributions of parameters that are constrained by the available observations will deviate from that, as for example shown for Kh.
Data: Bioregional Assessment Programme (Dataset 1)
Figure 38 Spatial distribution of the median of the posterior parameter distributions
Data: Bioregional Assessment Programme (Dataset 1)
2.6.2.8.1.3 Predictions
At each of the 1566 model nodes, an emulator is trained to predict dmax and tmax using the 1500 evaluated parameter combinations during the design of experiment. Each of the 10,000 parameter combinations in the posterior parameter distribution for each prediction is evaluated with the corresponding emulator to produce 10,000 predictions of dmax and tmax. Results are summarised using the 5th, 50th and 95th percentiles. Figure 39 shows how the 95th percentile of dmax evolves with increasing sample size for two prediction locations, probe66 and GW044912. At both locations the 95th percentile of dmax stabilises after about 6,000 samples.
dmax= maximum difference in drawdown between the coal resource development pathway (CRDP) and baseline, due to additional coal resource development
Data: Bioregional Assessment Programme (Dataset 1)
Figure 40a and Figure 40b show the histograms of the 5th, 50th and 95th percentile of dmax and tmax for the 1566 predictions. A log_{10} scale is used on the yaxis as the hydrological change is very skewed. More than threequarters of the output locations show dmax less than 2 m; more than twothirds of output locations have dmax less than 0.2 m. Figure 40c provides increased resolution of the number of predictions with dmax between 0 m and 2 m. At about 20 output locations, the median of drawdown is larger than 10 m, with no median drawdowns in excess of 40 m. The 95th percentile of dmax is larger than 40 m for about ten output locations with the maximum 95th percentile equal to 75 m.
The histograms of tmax (Figure 40b) show that the median tmax is evenly distributed throughout the simulation period. The 5th percentiles of tmax are slightly skewed to earlier in the simulation period, whereas the 95th percentiles are skewed to later in the simulation period. A very large proportion of model nodes have a tmax equal to 2102. This is the value assigned to all output locations for which dmax is equal to 0 m or dmax is not realised within the simulation period.
dmax = maximum difference in drawdown between the coal resource development pathway (CRDP) and baseline, due to additional coal resource development
tmax = year of maximum change
The 5th percentile of dmax is not entirely visible as it is covered by the line for the 50th percentile. A tmax value of 2102 is assigned to all predictions in which the dmax is not realised during the simulation period or the dmax is equal to 0 m.
Data: Bioregional Assessment Programme (Dataset 1)
Results from the groundwater model nodes were spatially interpolated to obtain valid posterior distributions at other locations across the modelling domain. A Delaunay Triangulation (DT) was generated in the R package ‘tripack’ (Renka et al., 2015). For any new location within the DT, the quantiles of maximum drawdown at the nodes of the triangle enclosing the new location are linearly interpolated to the new location. A forwardbackward cubedroot transform is applied during the interpolation process to improve performance over potentially nonlinear surfaces. The maps in Figure 41 and Figure 42 show the probability of dmax exceeding 0.2 m and 2 m respectively in the Hunter subregion. In each figure, the two smaller maps show the probability of drawdown under baseline and CRDP, while the larger map shows the probability of the drawdown due to the additional coal resource development. It can be seen that the hydrological changes are localised around the mine footprints and that the probability of exceeding both thresholds is higher under the CRDP.
For the purposes of defining the groundwater zone of potential hydrological change, that is the area in which the magnitude of predicted drawdowns from the additional coal resource development is deemed significant in terms of potentially impacting groundwaterdependent landscape classes and assets, a greater than 5% probability of dmax exceeding 0.2 m has been adopted. In Figure 41, this is shown by the change from purple to white. At its maximum, this zone does not extend more than about 20 km from the mining developments.
dmax = maximum difference in drawdown for one realisation within an ensemble of groundwater modelling runs, obtained by choosing the maximum of the time series of differences between two futures.
The difference in drawdown between CRDP and baseline is due to the additional coal resource development (ACRD). Drawdown under the baseline is relative to drawdown with no coal resource development; likewise, drawdown under the CRDP is relative to drawdown with no coal resource development.
Data: Bioregional Assessment Programme (Dataset 1)
dmax = maximum difference in drawdown for one realisation within an ensemble of groundwater modelling runs, obtained by choosing the maximum of the time series of differences between two futures.
The difference in drawdown between CRDP and baseline is due to the additional coal resource development (ACRD). Drawdown under the baseline is relative to drawdown with no coal resource development; likewise, drawdown under the CRDP is relative to drawdown with no coal resource development.
Data: Bioregional Assessment Programme (Dataset 1)
2.6.2.8.1.4 Comparison with results from other models
Section 2.6.2.2 provides a list of groundwater models that have been developed on behalf of various mining companies in the Hunter subregion for the purposes of estimating mine water make and modelling the impacts of pumping on local groundwater levels. These groundwater models are typically site scale, have a fine grid resolution and are underpinned by sitespecific knowledge of the lithology, geophysics and hydrogeology of the model domain that is held by the mine. They are also deterministic models, which means they provide a single estimate of hydrological change based on a single parameter combination that is considered optimal, whereas the Hunter subregion groundwater modelling package is designed to provide probabilistic ensembles of predictions, based on a range of likely parameter combinations. The primary predictions from the Hunter subregion groundwater model are dmax (i.e. the maximum difference in drawdown between baseline and CRDP) and tmax, whereas the minescale models are designed to provide changes in groundwater levels and fluxes at selected times in the future. Further complicating direct comparisons between model outputs are the differences in conceptualisation, boundary conditions and, critically, the implementation of coal resource development.
Given these points of difference, it is not warranted to make direct comparisons between results from these localscale models and from the groundwater model developed for the Hunter subregion.
However, Figure 43, Figure 44 and Figure 45 show the 5th, 50th and 95th percentiles respectively of dmax at the regional watertable under baseline and CRDP (relative to no coal resource development), and dmax at the regional watertable due to additional coal resource development (the difference in results between CRDP and baseline). These maps provide the range of drawdown predictions from the numerical modelling and can be used as reference to compare individual localscale model results.
dmax = maximum drawdown, the maximum difference in drawdown between the coal resource development pathway (CRDP) and baseline due to additional coal resource development.
Data: Bioregional Assessment Programme (Dataset 1)
dmax = maximum difference in drawdown due to the additional coal resource development (ACRD), obtained from the time series of differences between CRDP and baseline. Drawdowns under the baseline and the CRDP are relative to drawdown with no coal resource development.
Data: Bioregional Assessment Programme (Dataset 1)
dmax = maximum drawdown, the maximum difference in drawdown between the coal resource development pathway (CRDP) and baseline due to additional coal resource development.
Data: Bioregional Assessment Programme (Dataset 1)
While a direct comparison with results of localscale models is not warranted, the database of model runs created during the design of experiment permits a rapid assessment of the effect of local geological and hydrogeological information on the predictions for an area. This is illustrated here for the proposed Wallarah 2 mine.
The Wallarah 2 development is chosen because results from the surface water and groundwater modelling, based on the regional parameterisation, indicate potentially large hydrological changes in the Wyong River catchment, which are attributable predominantly to this development. However, local geological and hydrogeological information indicate that for some parameters, the range represented in the regional modelling is too wide and a locally more accurate prediction can be obtained through constraining the results from the regional model.
The Wallarah 2 proposal is a greenfield underground coal mine west of Wyong. Groundwater modelling undertaken for the Environmental Assessment assumes that the horizontal and vertical hydraulic conductivities of the overburden layers are sufficiently low that groundwater depressurisation due to mining would not propagate through to the watertable (Mackie Environmental Research, 2013). The Wallarah 2 groundwater modelling also assumes that hydraulic enhancement above the longwall panels will not extend into the top 120 m of overburden, thus nearsurface hydraulic conductivities will be unaffected by mining. The NSW Planning Assessment Commission critically examined the available data and modelling and concluded that the available data and modelling are consistent with a lowpermeable interburden and that propagation of depressurisation to the watertable aquifer is likely to be limited (Shepherd et al., 2014).
The prior parameter distributions specified in Section 2.6.2.8.1.1 represent a conservative range of parameters, designed to account for hydrogeological conditions at the regional scale varying from a tight, impermeable overburden to a more permeable interburden. The hydraulic conductivity data from the proposed Wallarah 2 mining area shows that hydraulic conductivity decreases more rapidly with depth than is the case in the regional dataset (Figure 46; Bioregional Assessment Programme, Dataset 2). The range of the parameters controlling the horizontal hydraulic conductivity, Kh and K_lambda is adjusted such that the upper limit of horizontal K values reflects the locally measured values.
In addition, the ratio of vertical to horizontal hydraulic conductivity is limited to 0.5. This means that the vertical hydraulic conductivity is at most half of the horizontal hydraulic conductivity. A final local adjustment is the hydraulic enhancement after longwall mine collapse. In the regional parameterisation the hydraulic enhancement extent varies from 100 m to 500 m above a mine. Based on the geomechanical analysis in Mackie Environmental Research (2013), the upper limit of this range is limited to 250 m.
Figure 47 shows the effect on the contours of dmax of this local parameterisation for the Wyong area. There is a considerable reduction of the area enclosed by the 0.2 m contour for the 50th percentile map where maximum dmax values do not exceed 2 m. At the 5th percentile no dmax values in excess of 0.2 m are present. At the 95th percentile with local parameterisation there is still drawdown simulated, but the magnitude and extent is less than what is simulated as median value in the regional parameterisation.
Boxplots show the range of Kh values from the regional dataset (Bioregional Assessment Programme, Dataset 2) and the Wallarah 2 groundwater assessment (Mackie Environmental Research, 2013) grouped by 25 m depth intervals. The area of lighter shading represents the range of Kh values at this location in the BA groundwater model for the regional parameterisation. The dark shading indicates the range of Kh values updated with local information, where the lower limit is left unchanged, but the upper limit is more constrained.
The drawdown area has been clipped to the Wyong River catchment boundary. dmax = maximum drawdown, the maximum difference in drawdown between the coal resource development pathway (CRDP) and baseline due to additional coal resource development
Data: Bioregional Assessment Programme (Dataset 1)
The Wallarah 2 example illustrates how the stochastic results of a conservative parameterisation can be updated in a straightforward way based on local information.
Cautious comparisons can also be made between the Hunter subregion BA drawdown results and the potential impact zone defined in the Mid Hunter Groundwater study (EMM, 2015). Key differences between this study and the Hunter subregion BA modelling are discussed in Section 2.6.2.2. In the EMM (2015) study, analysis of modelled drawdowns from sitescale groundwater modelling at a number of opencut mining sites in the Hunter Coalfield led to the adoption of a generalised 4 km buffer around mining areas to define the potential surface water impact zone (drawdowns >1 m). In Figure 48, a 4 km buffer has been generated around each of the baseline, CRDP and additional coal resource developments to delineate the zone within which drawdowns are assumed to be >2 m using the EMM (2015) approach. It can be seen that a generalised 4 km buffer provides a useful first approximation of the >2 m drawdown zone. When compared to the >2 m drawdown area from the BA model results, 4 km underpredicts the extent of this area in some areas and overpredicts it in others. In particular, the drawdowns from underground mines in the Newcastle Coalfield are underpredicted, although when local hydrogeological and geological data are used to constrain regional model results around Wallarah 2 (as shown in Figure 47), a 4 km radius overpredicts the potential impact zone. The difference between the Newcastle Coalfield and the two other coalfields, may be because the buffer was defined based on the results from modelling opencut mines only and/or because physical differences between the Newcastle and Hunter coalfields mean the 4 km buffer is not appropriate in areas beyond where it was derived. The BA groundwater model also results in asymmetric drawdown zones around mines, which reflects the spatial variability across the model domain (e.g. from topography; the alluvium and stream network; thickness of overburden). Because a onesizefitsall buffer approach does not take account of cumulative impacts, the buffer approach tends to underpredict the >2 m drawdown area under the CRDP in the Hunter Coalfield, but overpredict this area for just the additional coal resource development. In other words, a generalised buffer is not sensitive to the intensity of coal resource development.
dmax = maximum drawdown, the maximum difference in drawdown between the coal resource development pathway (CRDP) and baseline due to additional coal resource development
Data: Bioregional Assessment Programme (Dataset 1)
Product Finalisation date
 2.6.2.1 Methods
 2.6.2.2 Review of existing models
 2.6.2.3 Model development
 2.6.2.4 Boundary and initial conditions
 2.6.2.5 Implementation of the coal resource development pathway
 2.6.2.6 Parameterisation
 2.6.2.7 Observations and predictions
 2.6.2.8 Uncertainty analysis
 2.6.2.9 Limitations and conclusions
 Citation
 Acknowledgements
 Currency of scientific results
 Contributors to the Technical Programme
 About this technical product