Google and NASA model spots methane leaks from space
Using data from a sensor on the space station, the MAPL-EMIT model found 50% more methane leaks than human experts.
The model
Google and NASA's Jet Propulsion Laboratory introduced MAPL-EMIT, an AI model that tracks methane emissions from space. It uses data from EMIT, an imaging spectrometer mounted outside the International Space Station. The research appeared in PNAS earlier in September.
What it found
The model detected 50% more leaks than human experts and found more than 23,000 previously hidden methane plumes worldwide, including 24 of the 25 largest-emitting landfills. It was trained on 3.6 million physics-simulated plumes, which helps it pick out gas over cluttered landscapes.
Why it matters
Over 100 years, methane's warming potential is about 30 times that of carbon dioxide. Finding the source of a leak has been the main bottleneck. Automating detection and location makes faster, more targeted fixes possible.