Subsurface fluid injection associated with hydraulic fracturing, or 'fracking' -the high-pressure injection of fluids into deep rock formations to release oil and gas-can trigger manmade or 'induced' earthquakes. These events cause injury and damage, and pose risks to critical infrastructure such as pipelines and oil refineries. Induced earthquakes are becoming increasingly prevalent in the UK, and a data-driven understanding is urgently needed to inform efficient regulation and decision-making. However, due to the inherent complexity of the processes and time lags involved, it has not yet been possible to resolve how different operational and geological factors combine to induce earthquakes.
We will apply Bayesian data mining to analyse the complex, time-dependent interplay between fluid injection, geological factors affecting susceptibility, and seismicity, in major fracking operations. The project capitalises on innovative new software, recent mathematical advances, and high resolution geologic, seismic and injection data.
This project will apply machine learning to a problem of increasing national and global importance. It complements a number of the Turing lnstitute's key research areas, including data-driven modelling of complex systems (as applied to policy and decision making), understanding complex structures in spatial and temporal data, and automating forecasting processes using innovative Al techniques.
Principal Investigator: Dr Thomas Gernon (Southampton)
Co Investigator: Dr Theo Hincks