Scientists at the Berkeley National Laboratory in California have developed a tool that uses artificial intelligence (AI) to locate oil and gas wells that were never officially recorded.
The U.S. is estimated to have between 300,000 and 800,000 undocumented wells. Many of these wells, drilled before regulations existed, remain unplugged, posing risks of oil and methane leaks into the environment.
The study, published in Environmental Science & Technology, combines AI with historical maps from as early as 1884. These maps often show old oil and gas well locations. Fabio Ciulla, the study’s lead author, described the project as a “beautiful mix of the old and the new.”
Since 2011, the U.S. Geological Survey has scanned over 190,000 maps created between 1884 and 2006. While locating a single well on a map is easy due to clear symbols, identifying undocumented wells scattered among documented ones is like “finding a needle in a haystack,” according to co-author Charuleka Varadharajan.
The Berkeley team trained an AI program to recognize well symbols on any map. Once identified, the program provided the coordinates, which were confirmed using satellite images.
The tool was tested in four counties known for early oil production: Los Angeles and Kern in California, and Osage and Oklahoma in Oklahoma. Over 1,000 undocumented wells were discovered, with 29 verified manually so far. The researchers plan to use drones in future efforts to speed up the verification process.
Source: thechemicalengineer
