Mapping global methane emissions from space with deep learning

A deep-learning model called MAPL-EMIT can automatically detect and locate methane emissions from satellite imagery by analyzing the chemical fingerprints of methane gas in hyperspectral data. Methane is a potent greenhouse gas responsible for approximately 25% of human-induced warming since the industrial era, and because it has a short atmospheric lifespan, reducing emissions quickly offers a critical pathway to slowing global temperature rise. Over 125 countries have committed to reducing methane emissions by 30% by 2030, creating urgent need for tools that can track emissions from oil and gas infrastructure, agricultural facilities, and landfills at a global scale. The model was trained on 3.6 million synthetic methane plumes injected into real satellite scenes and can simultaneously measure methane amounts, delineate plume boundaries, and pinpoint emission sources even when multiple plumes overlap in dense industrial regions.