Since 2010 the Office of Groundwater Impact Assessment (OGIA) has been responsible for assessing the impact of coal seam gas (CSG) extraction activities on groundwater resources in the Surat Cumulative Management Area (CMA) and preparing an Underground Water Impact Report (UWIR) every three years.
As part of the 2019 UWIR, a regional groundwater flow model has been redeveloped to assess aquifer impacts from current and future CSG development. The groundwater flow model is constructed using a modified version of MODFLOW-USG and simulates 34 layers of geological strata over a 450 km x 650 km area. It is a large, complex model, containing over 1.3 million active cells and accounting for dual phase flow, faults, aquifer reinjection and the partial completion of CSG wells into non-CSG reservoirs. The model was calibrated using a highly-parameterized, regularized inversion approach implemented using PEST and requiring adjustment of more than 18,000 parameters on the basis of nearly 65,000 observations. This dataset of observations included transient observations of head and head changes at 480 monitoring locations throughout the Surat CMA. Model calibration was followed by the generation of 450 alternative calibration-constrained parameter fields using the PEST-supported null space Monte Carlo method in order to explore post-calibration predictive uncertainty.
Predictive uncertainty results included ranges of predicted short and long term CSG impacts on springs, water supply bores and overall aquifer water budgets. These results are used to support make good arrangements for water supply bores, inform spring impact management strategies and estimate aquifer recovery times.
This research demonstrates how the results of a predictive uncertainty analysis for a complex regional groundwater flow model can be used to support assessment of cumulative CSG groundwater impacts for a key groundwater system in Australia. Although important predictions are obtained to support CSG impact assessment, it is becoming increasingly apparent that regional modelling should be supplemented by predictive uncertainty modelling at smaller scales to better represent local aquifer interconnectivity features and capture local groundwater system dynamics.