MozFest House Amsterdam

MethaneMapper: An AI transformer model to track methane emissions from aerial imagery
2024-06-12 , Room D - Water Studio

Methane emissions pose a significant, yet often underestimated, threat to our environment. This potent greenhouse gas, while less prevalent than carbon dioxide, is 85 times more effective at trapping heat in the atmosphere, making it a critical target in the fight against global warming. The dangers of methane are exacerbated by its invisibility and the difficulty in detecting and quantifying its sources effectively.

The urgency to address methane emissions cannot be overstated. Due to methane’s high global warming potential, immediate action is imperative to prevent its accumulation in the atmosphere. The current climate crisis demands technologies that can swiftly and accurately identify and quantify methane emissions to mitigate their impact effectively.

Equally important is the democratization of this information. Everyone, irrespective of their location or background, deserves access to data about environmental pollutants like methane. Such transparency empowers communities, particularly those in vulnerable regions, to advocate for their health and environmental safety. It also fosters a broader understanding and engagement with environmental issues, paving the way for global cooperation in sustainability efforts.

In response to these challenges, our team at EyeClimate Inc. developed "MethaneMapper," an innovative tool designed to revolutionize how we detect and manage methane emissions. MethaneMapper is an open-source, AI-powered hyperspectral imaging tool that provides accurate, wide-area detection of methane emissions by analyzing overhead, airborne hyperspectral images. Unlike traditional methods that rely on ground or localized measurements prone to manipulation, MethaneMapper uses advanced AI algorithms, including a hyperspectral transformer network and contrastive learning, to trace emissions back to their sources while minimizing false positives. This tool was recognized for its groundbreaking contributions at the prestigious CVPR2023 conference.

The implications of deploying MethaneMapper are profound. By making invisible methane emissions visible, we can hold responsible parties accountable and drive the adoption of cleaner practices. This technology arms frontline communities, environmental activists, and policymakers with the data necessary to push for stricter regulations and reparations. Ultimately, MethaneMapper’s long-term impact lies in its potential to transform pollution monitoring globally, fostering sustainable management practices and informed policymaking that protect human health and the environment.

EyeClimate CEO and Schmidt Science Fellow: Satish Kumar is a seasoned expert with 12 years of research experience, 7 of which lie at the intersection of climate and AI. About to graduate with a PhD in computer vision, he has conducted cutting-edge research, having taken on projects alongside NASA-JPL, the Smithsonian Institute, The Aerospace Corporation, Harvard University and more. He has successfully completed projects in GHG emissions, wildlife tracking in the Maasai Mara and Serengeti. He has published papers in premier conferences such as CVPR, NeurIPS, BMVC, WACV and delivered talks on AI for Climate Change and Wildlife at CVPR, NeurIPS, Mara Elephant Project in Kenya. His GHG detection work was given a special highlight at CVPR. Satish has five years of hands-on experience in OS development at Samsung R&D, both in South Korea and India.