A new deep learning model maps global methane emissions from space.
Google and NASA JPL have released a deep learning model designed to quantify global methane emissions using data from the Earth Surface Mineral Dust Source Investigation (EMIT) instrument. The model automates the identification of methane point sources by processing hyperspectral imagery captured from the International Space Station.
Verified State Diff
Impact & Verification Analysis
Climate scientists, environmental policy makers, and global sustainability organizations.
This provides a scalable, verifiable method for identifying high-emission methane leaks, enabling targeted mitigation efforts and providing transparent data for global climate reporting.
Full Fact Overview
The collaboration leverages the EMIT instrument, which captures hyperspectral data across the visible and short-wave infrared spectrum. By applying a deep learning architecture to this high-dimensional data, the model can isolate the specific spectral signature of methane against complex surface backgrounds. This represents a shift from manual or less precise atmospheric monitoring to an automated, AI-driven pipeline capable of mapping emissions at a global scale with high spatial resolution.