- Home
- Assessments
- Bioregional Assessment Program
- Hunter subregion
- 2.6.1 Surface water numerical modelling for the Hunter subregion
- 2.6.1.5 Uncertainty
- 2.6.1.5.1 Quantitative uncertainty analysis
The aim of the quantitative uncertainty analysis is to provide probabilistic estimates of the changes in the hydrological response variables due to coal resource development. A large number of parameter combinations are evaluated and, in line with the Approximate Bayesian Computation outlined in companion submethodology M09 (as listed in Table 1) for propagating uncertainty through models (Peeters et al., 2016), only those parameter combinations that result in acceptable model behaviour are accepted in the parameter ensemble used to make predictions.
Acceptable model behaviour is defined for each hydrological response variable based on the capability of the model to reproduce historical, observed time series of the hydrological response variable. For each hydrological response variable, a goodness of fit between model simulated and observed annual hydrological response variable is defined and an acceptance threshold defined.
The ensemble of predictions are the changes in hydrological response variable simulated with the parameter combinations for which the goodness of fit exceeds the acceptance threshold. The resulting ensembles are presented and discussed in Section 2.6.1.6.
Parameter sampling
For AWRA-L, the parameters varied in the uncertainty analysis are those used in the calibration, with the addition of the parameter for scaling effective porosity, ne_scale. For AWRA-R, two of eight calibrated parameters, K_rout and Lag_rout, are included in the uncertainty analysis. The remaining AWRA-R parameters are set to their values from the low-streamflow calibration and not varied in the uncertainty analysis: variability in surface water – groundwater interactions (two parameters) and catchment runoff (one parameter) is already captured in inputs from the groundwater model and AWRA-L simulations; the three irrigation parameters are fixed as changes in irrigation management are not in the scope of this study.
Table 9 lists the parameters used in the uncertainty analysis and the range uniformly sampled in the design of experiment. The Australian Water Resources Assessment (AWRA) landscape model (AWRA-L) and AWRA river model (AWRA-R) parameters in Table 9 are explained in the AWRA-L v4.5 documentation (Viney et al., 2015) and AWRA-R v5.0 documentation (Dutta et al., 2015), respectively. Parameters with a large order of magnitude range in parameter bounds or which are thought to be particularly sensitive to low parameter values are transformed logarithmically to ensure that values near the minimum of the range are adequately sampled.
Three thousand parameter combinations are generated from the AWRA-L and AWRA-R model parameters according to the ranges and transformations shown in Table 9 using a space-filling Latin Hypercube sampling algorithm (Santner et al., 2003) that efficiently covers the sample space. These ranges and transformations are chosen by the modelling team based on previous experience in regional and continental calibration of AWRA-L (Vaze et al., 2013) and AWRA-R (Dutta et al., 2015). These mostly correspond to the upper and lower limits of each parameter that are applied during calibration.
The parameter combinations generated include all the parameter combinations for the groundwater model (see companion product 2.6.2 for the Hunter subregion (Herron et al., 2018)). This linking of parameter combinations allows the results to consistently propagate from one model to another, as outlined in the model sequence section (Section 2.6.1.1).
Each of the 3000 parameter sets is used to drive AWRA-L to generate a runoff time series at each 0.05 x 0.05 degree (~5 x 5 km) grid cell. The resulting runoff is accumulated to the scale of the AWRA-R subcatchments and is used – in conjunction with the sampled AWRA-R parameters – to drive AWRA-R.
Table 9 Summary of AWRA-L and AWRA-R parameters for uncertainty analysis
AWRA-L = Australian Water Resources Assessment landscape model; AWRA-R = Australian Water Resources Assessment river model; na = not applicable (dimensionless)
Observations
Predictions and observations from 22 streamflow gauges whose catchments contribute flow to the surface water modelling domain in the Hunter subregion are used for uncertainty analysis. Selection of the 22 catchments is based on three criteria: (i) data length more than 10 years; (ii) not subject to major open-cut and underground mine impacts; and (iii) not subject to major dam control. For these catchments, historical observations of streamflow are summarised into the nine hydrological response variables for all years. The equivalent historical simulated hydrological response variable values are computed from the 3000 design of experiment runs. The goodness of fit between these observed and simulated historical hydrological response variable values is used to constrain the 3000 parameter combinations and select the best 10% of replicates (i.e. 300 replicates) for each hydrological response variable. These 300 replicates are used for predictions in Section 2.6.1.6.
Predictions
For each of the 65 model nodes the post-processing of design of experiment results in 3000 time series with a length of 90 years (2013–2102) of hydrological response variable values for baseline, , and coal resource development pathway (CRDP) conditions, .
These two time series are summarised through the maximum raw change (amax), the maximum percent change (pmax) and the year of maximum change (tmax). The percent change is defined as:
|
(1) |
As the predictions include the effect of surface water – groundwater interaction through coupling with the groundwater models, groundwater parameters will also affect the surface water predictions.
Selection of parameter combinations
The acceptance threshold for each hydrological response variable is set to the 90th percentile of the average goodness of fit between observed and simulated historical hydrological response variable values obtained from nodes at 22 streamflow gauging sites. This means that out of the 3000 model replicates, the 300 best (or 10% best) are selected for each hydrological response variable.
The selection of the 10% threshold is based on two considerations: (i) guaranteeing enough prediction samples to ensure numerical robustness; and (ii) their performance approaching that obtained from the high- and low-streamflow model calibrations. Furthermore, it is expected that the full 3000 replicates contain many with infeasible parameter combinations (caused, for example, by parameter correlations that are not considered in the independent random sampling) and that these are likely to be filtered out by sampling only the best 10% of replicates. Nevertheless, selecting the best 10% of replicates is determined arbitrarily, and the strength and weakness of this decision are further discussed in Section 2.6.1.5.2.
Product Finalisation date
- 2.6.1.1 Methods
- 2.6.1.2 Review of existing models
- 2.6.1.3 Model development
- 2.6.1.3.1 Spatial and temporal dimensions
- 2.6.1.3.2 Location of model nodes
- 2.6.1.3.3 Choice of seasonal scaling factors for climate trend
- 2.6.1.3.4 Representing the hydrological changes from mining
- 2.6.1.3.5 Modelling river management
- 2.6.1.3.6 Rules to simulate industry water discharge
- References
- Datasets
- 2.6.1.4 Calibration
- 2.6.1.5 Uncertainty
- 2.6.1.6 Prediction
- Citation
- Acknowledgements
- Currency of scientific results
- Contributors to the Technical Programme
- About this technical product