Google Research's latest work applies deep learning to satellite imagery to map methane emissions anywhere on Earth. Methane is a short-lived but potent greenhouse gas, and identifying individual sources from space has historically required time-consuming manual review of images. The new model automates plume detection, making it possible to scan broad regions consistently.
The source describes the model as a step toward turning raw satellite observations into emission maps that can inform climate action. Because the pipeline is automated, it can be repeated as new images become available, offering a more up-to-date view of methane hotspots than field surveys alone. With a single source, there are no differing claims to weigh.