Improving Inspection Efficiency and Compliance Through Advanced Technology
The meteoric rise of shale oil and gas drilling in the United States poses significant challenges for reducing greenhouse gas emissions. An estimated 2 percent (13 billion kilograms) of natural gas production is leaked from U.S. facilities each year, contributing aggressively to climate change. Yet regulators and many firms are unable to comprehensively monitor emissions – relying on infrequent on-site facility inspections to detect leaks. Without reliable estimates of methane emissions, regulators cannot efficiently target leaks and incentivize their prevention. Now, advances in remote sensing technology offer a low-cost alternative, enabling precise measurement of methane emissions at scale.
The epicenter for work on this challenge is largely the state of Colorado. Home to a sizeable and growing shale oil & gas industry, the state was also the first to pass regulations aimed at reducing methane emissions from oil and gas operations and remains on the forefront of methane regulation. Realizing Colorado’s unique position, the E&E Lab partnered with the Colorado Department of Public Health and Environment to identify tools that can successfully address the gap in methane emissions monitoring and enforcement.
This groundbreaking project leverages predictive modelling, remote sensing technology, and field evaluations to improve inspection targeting, increase operator compliance, and, ultimately, to reduce methane emissions from oil and gas facilities. The E&E Lab has developed a machine learning model that leverages historical state inspection data to predict the location of methane leaks. The state began incorporating the model’s predictions as part of its inspection targeting strategy in 2022. The E&E Lab has worked with the state to design and implement three randomized control trials that vary the information provided to firms about 1) leak risk, 2) detected leaks, and 3) increased regulatory scrutiny. The research measured changes in methane emissions from oil and gas facilities by applying a novel monitoring technology.
The researchers found that in 2021, 5.6% of state oil and gas inspections identified violations. Once Colorado adopted the machine learning model to target inspections, the hit rate tripled – increasing to 17%. The researchers also found that informing operators which facilities are at higher risk of leaks and increased regulatory scrutiny decreases fugitive emissions by 50%. Colorado is continuing to use the machine learning model to target its oil and gas facility inspections, and the U.S. EPA has adopted a similar machine learning model to target hazardous waste inspections. At EPA’s request, the research team is now adapting this machine learning approach to Clean Air and Clean Water Act inspections in the U.S. The machine learning model framework is ready to be adapted and scaled in new states and countries beyond the US where enforcement resources are limited.