Researchers from the University of Oxford partnered with NIO.space of Trillium Technologies to develop a tool that automatically picks up Earth's methane plumes from orbit. The device mainly utilized machine learning alongside hyperspectral data.

greenhouse gas
(Photo: Pixabay / marcinjozwiak)

Novel AI Tool For Detecting Earth Methane Plumes From Space

The novel tool can help with the identification of grave methane emitters. It may also offer actions to reduce emissions of greenhouse gases. The innovation was documented in the "Semantic Segmentation of Methane Plumes with Hyperspectral Machine Learning Models" study.

The AI tool specifically utilizes the data taken from hyperspectral satellites to pick up methane plumes. These are capable of picking up narrower bands compared to multispectral satellites. Because of this, it is easier to filter noise and tune to the specific methane signature. However, the data produced is massively larger. This, in turn, makes it hard to further process without the assistance of artificial intelligence (AI).

The model was trained with 167,825 hyperspectral tiles taken by NASA's aerial AVIRIS sensor over the US' Four Corners. The algorithm was then used for other hyperspectral sensor data, such as the ones taken from NASA's EMIT hyperspectral sensor, which is linked to the International Space Station. The EMIT sensor, which stands for Earth Surface Mineral Dust Source Investigation mission, offers nearly complete Earth coverage.

The model's accuracy reaches over 81% when picking up large methane plumes. It was also found to be 21.5% more accurate than the previous approach, which was considered the most accurate.

The innovation also significantly helped with detecting false positives in tile classification. This went down by 41.83% compared to the other most accurate method.

For further methane detection research, the code and annotated dataset have been open-sourced by researchers on a GitHub project page. The researchers are now considering whether the model can work aboard the satellite and allow others to perform further observations.

ALSO READ: Human-Induced Methane Emissions Contribute to Greenhouse Gases; Agriculture, Livestock, Gas, and Coal Are Primary Culprits


Methane Emissions in Global Warming

Last 2021, methane was responsible for 21% of greenhouse gas emissions across the US that are attributed to human activity. These methane-releasing human acitvities include livestock raising and natural gas system leaks. Natural sources, including termintes, also release methane, while other natural processes in soil and atmospheric chemical reactions also lead to the gaseous release. 

Compared to carbon dioxide, methane gas has 80 times higher effectiveness when trapping heat. However, its atmospheric lifetime of seven to 12 years is remarkably short of that of CO2.

Because of this, proactively reducing methane emissions from sources of anthropogenic roots could immediately affect the slowing down of global warming and boost air quality. Cutting emissions of methane is the fastest way to do so. Hence, this opportunity is something that humanity cannot afford to miss out on. 

Estimates have shown that achievable methane emission reductions could offer almost 0.3 degrees Celsius of avoided global warming in the next twenty years.

Check out more news and information on Environment & Climate in Science Times.