Nature blind test: satellites reliably spot methane leaks

Methane is the second most important greenhouse gas released by humans after CO2, and much of it escapes from oil and gas facilities. Whether satellites can reliably find such leaks has long been disputed. A team led by PhD student Frances Reuland and Professor Adam Brandt of Stanford University tested this in a year-long blind test, published in "Nature" on 14 October. At a test site in Casa Grande, Arizona, methane was deliberately released 481 times in 2025, at rates between 9.5 and around 1,184 kilograms per hour, timed to overpasses by twelve satellites. In addition, there were 265 overpasses without releases as controls; 15 analysis teams from nine countries did not know when gas was flowing.

Instruments purpose-built for methane detected 92 to 100 per cent of releases under clear skies, with 0 to 11 per cent false alarms. Repurposed instruments such as the German satellite EnMAP achieved 62 to 88 per cent, general land-imaging satellites only 29 to 37 per cent. Most quantity estimates were off by up to 50 per cent, with a mean error of around 44 to 46 per cent, but over- and underestimates largely balanced out on average. Under ideal conditions, the detection limit is about 50 to 200 kilograms per hour depending on the system, close to the 100-kilogram threshold above which facilities are reviewed as "super-emitters".

The study also shows limits: with more than 20 per cent cloud cover, the performance of all systems dropped, and 79 per cent of false alarms came from AI-based analyses by commercial teams. According to Stanford, more research is needed for non-desert terrain, varied weather and open water. "There has been significant skepticism over the years about whether methane tracking really works," Brandt said. Even if the amount is only roughly known, you still know a leak needs fixing, the researcher said. In total, less than 50 tonnes of methane were released during the test.