Mapping global methane emissions from space with deep learning

What is it about?

A study using deep learning to analyze data from the Earth Surface Mineral Dust Source Investigation (EMIT) instrument on the International Space Station finds that combining spectral data with spatial context allows for mapping point sources of methane with sensitive detection limits and the ability to delineate overlapping plumes; the authors trained the deep learning model on synthetic data representing 3.6 million plumes and applied the resulting model to all data ever produced by EMIT automatically identifying thousands of methane sources that were previously undetected.

Featured Image

Photo by Kevin Stadnyk on Unsplash

Photo by Kevin Stadnyk on Unsplash

Why is it important?

Methane point-sources drive near-term climate forcing, yet tracking them globally remains challenging. We present a deep-learning framework that advances emission monitoring by directly analyzing satellite-based spectroscopic data and demonstrate its application to data from the Earth Surface Mineral Dust Source Investigation (EMIT) instrument. This approach provides the spatial context required to map gas concentrations across landscapes, significantly lowering detection limits and automating plume delineation and source localization, even for overlapping methane plumes from multiple sources. We validate the model against synthetic and real-world benchmarks, and demonstrate that it captures known emitters while identifying numerous previously undetected plumes. The automated end-to-end modeling pipeline enables application to the full EMIT archive, providing a compelling foundation for climate mitigation and facility-level accountability.

Perspectives