Conor Murphy

Research

Assessment of hazard and risk due to induced seismicity for underground CO2 storage and oil and gas production assets

Overview

Production of oil and gas can in some circumstances cause seismic activity. Problematic levels of seismic activity are rare and occur only in a small minority of oil and gas fields. However, the potential impact of high levels of seismic activity justifies careful monitoring and modelling of hazards and risks. Such induced earthquakes also arise from a new strategy used to combat climate change, specifically where CO­2 is captured and stored in underground reservoirs. Accurate forecasting of hazards under scenarios for future extraction/injection is vital in ensuring the process is operated safely. Statistical methodology plays an important role in the design of monitoring strategies and in the assessment of these hazard and risks. Our work relies on a framework in Statistics known as Extreme Value Theory. In this area, we aim to model and forecast the most extreme events of a process, such as heavy rainfall, extreme waves, or in this case, induced earthquakes, with the goal of protecting against such events.

This project aims to improve upon current statistical models of marked point pattern data to inform the assessment of seismic hazards. A difficulty arises in this setting due to the development of the sensor network over time. Improvements to the geophone network at the site over time have allowed smaller magnitude earthquakes to be detected. This leads to the problem of missing data from the years where the geophone network was not dense enough to detect small magnitude seismic events. Previous work has explored temporal features of the process and a time-varying threshold selection method has been developed for an extreme value analysis of the induced earthquake magnitudes. This project will build upon this work by exploring spatio-temporal properties.

The project will initially be focused on data from the Groningen gas field in the Netherlands but the inclusion of seismic data from an underground CO2 storage site will also be explored. The Groningen gas field is important because extraction over the past few years has been radically reduced as seismic activity posed too great a threat. Despite this, there are still substantial hazards and risks related to seismicity in the area, requiring potentially large investment into mitigating measures such as house strengthening. This natural resource plays an important role as a back-up for gas production in the event of a cold Winter.  It is important to see if improvements to statistical models and seismicity forecasting can inform future extraction strategies so that the occurrence of larger earthquakes is reduced.

Current Research

Motivated by the complex setting of induced earthquakes, we’re currently focussed on a novel methodology for automated threshold selection for univariate extremes. We have explored the use of this method for constant threshold selection for iid, continuous data and found that it outperforms existing approaches. We are also investigating extensions of this methodology to more difficult settings. In any setting, the choice of threshold is the most fundamental aspect of a threshold-based extreme value analysis. Too low a threshold can violate the asymptotic basis of the extreme value model, inducing bias into the model fit, while too high a threshold leads to increased uncertainty in parameter estimation. Current methods acknowledge this fundamental bias-variance trade-off without directly accounting for it. Our method can be seen to directly tackle this challenge. We have applied our method to simulated datasets and compared the performance against existing methods. The R functions used to implement the method are available here and the preprint is available here.

Furthermore, we are currently focussed on extending the time-varying threshold used for the Groningen earthquake catalogue to also include spatial variation. We are exploring the inclusion of spatio-temporal covariates in our modelling procedure in order to make use of more of the available data and improve hazard forecasts.

Conference and workshop contributions

Conor Murphy
STOR-i PhD Student