MAPL-EMIT

MAPL-EMIT is an AI model that automatically finds invisible methane gas leaks in satellite images. It was built by Google Research together with NASA's Jet Propulsion Laboratory, using data captured by an instrument called EMIT mounted on the International Space Station.

Think of it like scanning a wide field with a thermal camera to spot where smoke is leaking out — work that used to require a person to scan every image by eye, now done automatically by a computer. Methane is invisible to the naked eye, but it leaves a distinctive signature by absorbing specific wavelengths of light. In the past, this signature was often confused with ground surface signals that happened to look similar, leading to frequent false positives. MAPL-EMIT was trained to look beyond a single point and examine the surrounding area as a whole, learning to recognize the true shape of a gas plume as it disperses in the wind.

As a result, it detects more leak sources than previous methods, while also calculating how much is being emitted and exactly where the leak is located. The instrument itself was originally launched to map minerals in arid regions like deserts, but it turned out to be useful for methane detection as well, expanding its intended use.

News articles describe it along the lines of: "Google Research, together with NASA's Jet Propulsion Laboratory, unveiled MAPL-EMIT, a deep learning model that automatically identifies methane leak sources from satellite hyperspectral imagery." It's important not to confuse MAPL-EMIT with hardware like a satellite or camera — it is a software model that analyzes data captured by satellites.