Uncertainties of geochemical codes and thermodynamic databases for predicting the impact of carbon dioxide on geologic formations

Numerical codes are applied to calculate chemical reactions following geologic carbon sequestration in deep formations and CO2 leakage in shallow formations. However, using different thermodynamic databases generates variations in the simulation results, which are referred to as the model uncertainty. The PHREEQC and The Geochemist's Workbench codes were used to simulate anorthite dissolution for storage, retention, transfer, and near-surface formation waters in the respective geological units. For each of the formation waters, a simple one-dimensional scenario was simulated using eight different thermodynamic databases. Groundwaters in shallow aquifers commonly exhibit low ionic strengths (<0.5 mol/kgw) and low temperatures, whereas storage formation waters are characterized by high ionic strength (>1.0 mol/kgw) and high temperatures. In storage formations, mineral trapping is the most efficient process for long-term CO2 storage. However, with respect to the geological formations and the time needed for anorthite dissolution, the model uncertainties associated with using different combinations of numerical codes and thermodynamic databases were largest (similar to 90%) for the storage formation waters at 58 degrees C and I = 6.5 mol/l. Conversely, in near-surface formation waters, the model uncertainty was less than 1%. Due to CO2 dissolution, the calculated pH of the formation waters decreased to a range between pH 4.0 and 5.5. In this pH range, the dissolution mechanism of anorthite switches from the slow neutral mechanism to the faster acid mechanism, causing dissolution time length variations. The calculated pH variation further increased with rising ionic strength. A detailed examination of the reasons revealed the activity coefficient calculation method of the main aquatic species to have the largest impact on the simulated model results. The calculation method of the CO2 activity coefficient had the second largest impact. Via calibration with the experimental data, a specific thermodynamic database can be chosen to represent these experimental results. However, the calibration of thermodynamic databases is not possible for all potential reactions in more complex geological systems at large ranges of temperature, ionic strength and pressure conditions. The uncertainties associated with using thermodynamic databases quantified in this study for CO2 storage systems will therefore persist independently from previously conducted calibrations of thermodynamic databases with experimental or field data. In view of these model uncertainties, the modeller is encouraged to include a routine in the simulations for quantification of the model uncertainty depending on the specific scenario or to assess the simulation results as a range of values that represent a soft outcome. (C) 2016 Elsevier Ltd. All rights reserved.