Alberta Wells Dataset: Pinpointing Oil and Gas Wells from Satellite Imagery

Pratinav Seth, Michelle Lin, BREFO DWAMENA YAW, Jade Boutot, Mary Kang, David Rolnick

TL;DR: We introduce a large benchmark dataset for pinpointing abandoned oil and gas wells, a significant contributor to climate change and pollution.

Abstract: Millions of abandoned oil and gas wells are scattered across the world, leaching methane into the atmosphere and toxic compounds into the groundwater. Many of these locations are unknown, preventing the wells from being plugged and their polluting effects averted. Remote sensing is a relatively unexplored tool for pinpointing abandoned wells at scale. We introduce the first large-scale Benchmark dataset for this problem, leveraging high-resolution multi-spectral satellite imagery from Planet Labs. Our curated Dataset comprises over 213,000 wells (abandoned, suspended, and active) from Alberta, a region with especially high well density, sourced from the Alberta Energy Regulator and verified by domain experts. We evaluate baseline algorithms for well detection and segmentation, showing the promise of computer vision approaches but also significant room for improvement.

Lay Summary: Millions of abandoned oil and gas wells worldwide leak potent methane and other toxins, yet many are unmapped, hampering cleanup. To help solve this, we present the Alberta Wells Dataset, the first large-scale machine learning benchmark specifically designed for pinpointing these onshore wells – whether abandoned, suspended, or active. Our dataset offers over 200,000 well instances from Alberta, Canada, paired with high-resolution satellite imagery, aiming to drive breakthroughs in identifying these sites. We've framed the challenge for AI as detecting well locations and outlining their exact shapes, and our initial tests show promise but also the need for more advanced methods. By providing this benchmark, we enable the research community to develop more scalable and accurate AI techniques. The ultimate impact is to significantly reduce environmental harm: improved AI, built using this dataset, can help locate even unrecorded wells and identify high-emission sites that urgently need remediation. This directly contributes to mitigating climate change by helping stop potent methane leaks from these often hidden sources, paving the way for cleaner air and protected groundwater.

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Link To Code: https://github.com/RolnickLab/Alberta_Wells_Dataset

Primary Area: Applications->Everything Else

Keywords: Remote Sensing, Satellite Imagery, Climate Change, AI for Good

Submission Number: 6985