Mapping global methane emissions from space with deep learning
| Source: Google Research Blog
Tags: MAPL-EMIT, Google Research, methane detection, hyperspectral imaging, climate AI, satellite data, NASA EMIT
Google Research's MAPL-EMIT uses deep learning on NASA's EMIT hyperspectral satellite to automatically detect and quantify methane plumes globally, achieving 84% recall on expert-annotated data — turning satellite imagery into actionable emissions monitoring for the Global Methane Pledge's 30% cut target.
Details
Methane carries 30x the warming potential of CO2 over a 100-year period and has driven ~25% of human-induced warming since industrialization. Its short atmospheric lifespan makes rapid emission cuts a fast-action climate lever — the basis for the Global Methane Pledge, where 125+ countries committed to a 30% reduction by 2030. Google Research built MAPL-EMIT (Methane Analysis and Plume Localization with EMIT), a deep-learning framework atop NASA's EMIT hyperspectral sensor aboard the ISS. EMIT records hundreds of spectral bands per pixel, enabling detection of the chemical fingerprints of otherwise invisible methane plumes. MAPL-EMIT automates the full pipeline: detection, enhancement quantification, and source attribution. Published in PNAS, the model achieves 84% recall on expert-annotated plumes and outperforms existing matched-filter methods on signal-to-noise ratio. Google is releasing all artifacts open-source: a global plume database on Earth Engine, enhancement maps, a visualization app, trained model weights, a synthetic plumes dataset, and an inference library. The practical upshot: regulators, operators, and NGOs can now monitor methane leaks across oil/gas infrastructure, farms, and landfills at global scale — without manual expert review of raw satellite imagery.