Characterizing oil and gas wells with fugitive gas migration through Bayesian multilevel logistic regression

Abstract

Oil and gas wells are engineered with barriers to prevent fluid movement along the wellbore. If the integrity of one or more of these barriers fails, it may result in subsurface leakage of natural gas outside the well casing, a process termed fugitive gas migration (GM). Knowledge of the occurrence and causes of GM is essential for effective management of associated potential risks. In the province of British Columbia, Canada (BC), oil and gas producers are required to report well drilling, completion, production, and abandonment records for all oil and gas wells to the provincial regulator. This well data provides a unique opportunity to identify well characteristics with higher likelihoods for GM to develop. Here we employ Bayesian multilevel logistic regression to understand the associations between various well attributes and reported occurrences of GM in 0.6% of the 25,000 oil and gas wells in BC. Our results indicate that there is no association between the occurrence of GM and hydraulic fracturing. Overall, there appears to be no well construction or operation attribute in the study database that is conclusively associated with GM. Wells with GM more frequently exhibit indicators of well integrity loss (i.e., surface casing vent flow, remedial treatments, and blowouts) and geographic location appears to be important. We ascribe the spatial clustering of GM cases to the local geologic environment, and we speculate that there are links between particular intermediate gas-bearing formations and GM occurrence in the Fort Nelson Plains Area. The results of this study suggest that oil and gas wells in high GM occurrence areas and those showing any attribute associated with integrity failure (e.g., surface casing vent flow) should be prioritized for monitoring to improve the detection of GM.