Google’s AI detects methane from space
Google and NASA have introduced `MAPL-EMIT`, an AI model that detects methane emissions from space. Trained on 3.6 million simulations, it identifies 50% more plumes than human experts and locates more than 23,000 additional sources.
Google and NASA’s Jet Propulsion Laboratory (JPL) have introduced MAPL-EMIT, an artificial intelligence model that detects methane emissions worldwide using images captured from space.
The study, published in PNAS, tackles a difficult problem: distinguishing methane plumes from complex terrain, noise, and other signals in the images. To train the model, the researchers generated 3.6 million physics-based simulated methane plumes.
The result is a tool that detects 50% more emissions than human experts and identifies more than 23,000 additional sources worldwide. These include 24 of the 25 highest-emitting landfills in the world.
Why methane matters
Methane is an especially potent greenhouse gas. Over a 100-year period, its warming potential is 30 times greater than that of carbon dioxide.
The problem is not only how much methane is emitted, but how difficult it can be to find the exact source. A leak at a gas facility, mine, or landfill can go undetected without frequent observations and tools capable of analyzing them at scale.
MAPL-EMIT helps turn those images into a map of specific sources. That makes it possible to prioritize inspections and act sooner at the sites causing the most pollution, instead of searching blindly.
Open data for research and action
Google published the global methane plume dataset in Earth Engine, its platform for analyzing large volumes of geospatial data. It also launched an application for viewing the results.
The open-source models are also available on Kaggle, and the tools for running inferences have been published on GitHub. The goal is to make them available to:
- Researchers studying emissions.
- Public agencies designing climate policies.
- Companies and operators that need to locate leaks.
The tool does not eliminate emissions on its own. Its role is to make the first part of the process faster and more precise: finding out where emissions are and which sources require urgent attention.
The next question is how often this map can be updated and how its detections translate into real repairs. Detecting thousands of sources from space is a practical step forward, but its climate impact will depend on governments and operators acting on that information.