A new deep learning model maps global methane emissions from space (blog.google)

A groundbreaking deep learning model, MAPL-EMIT, has been developed to map global methane emissions from space, offering a significant advancement in climate monitoring. Methane, a greenhouse gas that warms the planet 30 times more effectively than carbon dioxide over a century, makes detecting its emissions critical. By leveraging 3.6 million physics-simulated methane plumes, MAPL-EMIT surpasses human capabilities, identifying 50% more plumes and uncovering over 23,000 additional emissions sources worldwide, including nearly all of the largest landfills. This enhanced detection ability paves the way for faster and more effective climate mitigation strategies. Google has made this innovative model even more accessible by publishing a global plume database on Earth Engine and creating an app to visualize the data. Open-source models are available on Kaggle, with inference tools provided on GitHub, enabling researchers, policymakers, and environmental operators to utilize this technology in their efforts to combat climate change. The release of MAPL-EMIT represents a significant step forward for the AI and machine learning community, showcasing how advanced analytics can be harnessed to address critical environmental challenges.