Google and NASA Release MAPL-EMIT, an AI Model for Tracking Methane Emissions from Satellite Data

Google Research and NASA's Jet Propulsion Laboratory (JPL) announced the development of MAPL-EMIT, an AI model that tracks global methane emission sources using data from the NASA EMIT satellite. The study, published in PNAS, focused on improving the accuracy of detecting large-scale emission sources, considering that methane's global warming potential is 30 times higher than that of carbon dioxide over a 100-year period.

MAPL-EMIT was trained on 3.6 million physically simulated methane plume datasets. In environments with complex terrain and noise, it detects 50% more plumes than human experts and has identified over 23,000 additional emission sources globally. Notably, it accurately located 24 of the 25 largest landfills, which are the world's biggest emission sources.

Google has made this data available on Earth Engine along with a visualization app. The open-source model is distributed on Kaggle, and inference tools are available on GitHub, supporting researchers, policymakers, and operators in utilizing these resources for climate mitigation efforts.