H53K-01
Model-based analysis of push-pull experiments in deep aquifers to predict large-scale impacts of CSG product water reinjection

Friday, 18 December 2015: 13:40
3016 (Moscone West)
Henning Prommer, University of Western Australia, Crawley, Australia; CSIRO Land and Water, Floreat, Australia
Abstract:
Over the next two decades coal seam gas production in Australia will require the management of large quantities of production water. For some sites the most viable option is to treat the water to a high standard via reverse osmosis (RO) and to inject it into deep aquifers. The design and implementation of these field-scale injection schemes requires a thorough understanding of the anticipated water quality changes within the target aquifers. In this study we use reactive transport modeling to integrate the results of a multi-scale hydrogeological and geochemical characterization, and to analyze a series of short-term push-pull experiments with the aim to better understand and reliably accurately predict long-term water quality evolution and the risks for mobilizing geogenic arsenic.

Sequential push-pull tests with varying injectant compositions were undertaken, with concentrations recorded during the recovery phase reaching levels of up to 180 ppb above the ambient concentrations observed prior to the push-pull experiments. The highest As concentrations were observed in conjunction with the injection of aerobic water, while de-oxygenation of the injectant lowered As concentrations significantly. The lowest As concentrations were observed when the injectant was de-oxygenated and acid-amended. The latter was underpinned by complementary laboratory As sorption experiments using sediments from the target aquifer at various pHs, which, consistent with literature, show a decrease in As sorption affinity under alkaline conditions.

In the model-based analysis of the experimental data, model parameters for each conceptual model variant were estimated through an automatic calibration procedure using Particle Swarm Optimization (PSO) whereby bromide and temperature data were used to constrain flow, solute and heat transport parameters. A series of predictive model scenarios were performed to determine whether advanced manipulation of the injectant composition is required.