Model sensitivity to different combinations of parameters was assessed during calibration of the regional steady-state model and transient sub-model using the parameter estimation (PEST) suite of software in a two-stage process as described by GHD (GHD, 2012, pp. 72–73):
- Calculation of the sensitivities of each of the parameters to be estimated with respect to all observation targets used in the calibration process. This task involved the computation of a Jacobian Matrix holding information on the sensitivity of each observation to each parameter. This first calibration step required the model to be run as many times as the number of base parameters (i.e. 3,366 times for the regional model and 895 times for the transient sub-model);
- Use of the Singular Value Decomposition Assist (SVDA) methodology and Tikhonov Regularisation to optimise the model parameters. A key part of this process was the production of a user defined number of ‘super-parameters’ i.e. linear combinations of base parameters. For the final regional model calibration 430 such ‘super-parameters’ were adopted based on the maximum number recommended by the PEST SUPCALC utility. Each iteration of the model optimisation process involves at least as many runs as ‘super parameters’ (i.e. 430 in the case of the final regional model calibration) and multiple iterations are typically required. A detailed explanation of the SVDA methodology is outside the scope of this Report, thus the interested reader is directed to the PEST manual (Doherty, 2010) for more information.
Calibration of the OGIA model also considered model sensitivity to how the Walloon Coal Measures were represented, recharge estimates and model robustness. ‘The potential impact of different layering systems for representation of the Walloon Coal Measures has not been assessed at this stage. However, the sensitivity of model predictions to different parameterisation of the three Walloon Coal Measures layers has been assessed as part of the model uncertainty analysis’ (GHD, 2012, p. 106). ‘It is recognised that revisions to recharge quantities may not significantly affect predictions of CSG impacts, at least in the short to medium-term. The sensitivity of model predictions to different recharge estimates has been considered further as part of the model uncertainty analysis’ (GHD, 2012, pp. 106–107). Model robustness is evidenced by the use of 200 calibration-constrained parameter sets that all achieve a reasonable match to the adopted calibration data.
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 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
- Glossary
- Citation
- Acknowledgements
- Contributors to the Technical Programme
- About this technical product