Berkeley AI uncovers forgotten oil and gas wells in historical maps

The discovery and management of hundreds of thousands of undocumented oil and gas wells across the United States presents significant environmental and safety challenges, with researchers now leveraging artificial intelligence to locate these potentially hazardous sites.

The critical challenge: Undocumented orphaned wells (UOWs) pose significant environmental risks through potential chemical leaks into water sources and methane emissions into the atmosphere.

• Experts estimate between 310,000 and 800,000 undocumented wells exist across the United States
• These wells can leak toxic substances like benzene and hydrogen sulfide
• Methane emissions from these wells are particularly concerning, as methane is 28 times more potent than carbon dioxide at trapping heat in the atmosphere

Innovative detection approach: Berkeley Lab researchers have developed an AI system that analyzes historical USGS topographic maps to identify previously unknown well locations.

• The AI system examines maps from 1947-1992 that used consistent symbols for oil and gas wells
• Researchers trained the algorithm using nearly 100 manually marked maps from California
• The system successfully identified 1,301 potential undocumented wells across four counties in California and Oklahoma

Verification methods: Multiple technologies are being deployed to confirm well locations and assess their environmental impact.

• Ground-based verification uses magnetometers to detect buried metal well casings
• Drone-mounted sensors equipped with various detection technologies survey larger areas
• Low-cost sensor combinations measure methane emissions to prioritize well remediation efforts

Collaborative efforts: The Consortium Advancing Technology for Assessment of Lost Oil & Gas Wells (CATALOG) coordinates research across multiple national laboratories and stakeholders.

• The program includes teams from five national laboratories
• Partnerships with Native American tribes, including the Osage Nation, provide essential feedback
• The initiative aims to develop cost-effective tools that can be widely adopted across the United States

Looking ahead: The combination of historical data analysis and modern technology represents a promising approach to addressing this environmental challenge.

• The current findings likely underestimate the total number of undocumented wells
• Ongoing development of new detection methods continues to improve identification capabilities
• The initiative demonstrates how integrating past and present technologies can help address legacy environmental issues

The success of this multi-layered approach to well detection suggests that similar strategies could be applied to other environmental challenges where historical data might help inform modern remediation efforts.