Uncertainty analysis: influence of hydraulic fracturing on overlying aquifers in the presence of leaky abandoned wells

Abstract

Model uncertainty analysis can quantify uncertainty both prior to calibration and postcalibration if the calibration dataset appropriately informs parameter estimates and model predictions. In certain cases calibration data (observations or measurements) may not be immediately apparent, but calibration datasets can be developed from related data for model interrogation and quantification and minimization of uncertainty. This study applies a series of techniques to investigate uncertainty in a simple numerical model of upward flow (leakage) through an abandoned oil and gas well converted into a water well in hydraulically fractured shale. Model calibration was achieved by developing a limited calibration dataset from well-specific measurements at a horizontal well in the Eagle Ford Shale play. Uncertainty in the calibrated model was interrogated using sensitivity, linear, and nonlinear analyses available in the PEST suite. Sensitivity analysis suggests that flowback after hydraulic fracturing could be crucial in reducing leakage. Linear analyses indicate horizontal-well production rates and long-term reservoir pressures are valuable measurements to collect when evaluating potential leakage. Nonlinear analyses identify the range in predictive uncertainty of potential leakage. The results underscore the need to evaluate and include additional types of well data in public records, such as flowback and produced water volumes. Overall, the results of this study illustrate the utility of uncertainty analyses with a limited calibration dataset applied to a simple model.