London, November 26 (IANS): A team of researchers has developed a tool that uses machine learning with hyperspectral data to automatically detect methane plumes on Earth from orbit.
This could help identify excess methane “superemitters” and enable more effective measures to reduce greenhouse gas emissions.
While net-zero targets focus on carbon dioxide (CO2) emissions, combating methane emissions is 80 times more effective at trapping heat than CO2. Therefore, acting quickly to reduce methane emissions from anthropogenic sources will have an immediate impact on slowing global warming and improving air quality.
However, until now, methods for easily mapping methane plumes from aerial imagery have been very limited and the processing steps are extremely time-consuming.
This is because methane gas is transparent to both the human eye and the spectral range used by most satellite sensors.
Even when satellite sensors are operating in the correct spectral range to detect methane, the data is often obscured by noise and requires painstaking manual effort to effectively identify plumes.
A new machine learning tool developed by researchers at the University of Oxford in partnership with Trillium Technologies’ NIO.space solves these problems by detecting methane plumes in data from hyperspectral satellites.
These detect a narrower band than typical multispectral satellites, making it easier to tune into specific features of methane and filter out noise.
However, the amount of data generated is much larger and difficult to process without artificial intelligence (AI).
The researchers trained their model using 167,825 hyperspectral tiles (each representing an area of 1.64 km2) captured by NASA’s airborne sensor AVIRIS over the Four Corners region of the United States. The algorithm then uses other hyperspectral sensors in orbit, such as data collected from NASA’s new hyperspectral sensor EMIT (Earth Surface Mineral Dust Source Investigation Mission), which is mounted on the International Space Station and covers nearly the entire Earth’s globe. Applied to data from spectral sensors. .
Overall, the model was more than 81% accurate in detecting large methane plumes and 21.5% more accurate than the previous most accurate approach. The method, published in the journal Scientific Reports, also significantly improved the false positive rate for tile classification, reducing it by about 41.83 percent compared to the previous most accurate approach.
The researchers are now considering whether the model could run directly on the satellite itself, allowing other satellites to make follow-up observations as part of the NIO.space initiative.
“Such on-board processing could mean that initially only priority alerts need to be sent back to Earth, for example if an identified methane such as a text alert signal containing the coordinates of the source.”
“Furthermore, this allows a swarm of satellites to work together autonomously. The initial weak detection acts as a tip-off signal to other satellites in the constellation, directing the imager’s focus to the area of interest. You can adapt it to a location,” Ruzicka added.