How AI could help scientists spot ultra-emission methane plumes fasterfrom space

Date:
Mon, 27 Nov 2023 20:00:00 +0000

Description:
Reducing leaks of the potent greenhouse gas could alleviate global warming by as much as 0.3 degrees Celsius over the next two decades. The post How AI could help scientists spot ultra-emission methane plumes fasterfrom space appeared first on Popular Science .

FULL STORY ======================================================================


Reducing damaging ultra-emission methane leaks could soon become much easierthanks to a new, open-source tool that combines machine learning and orbital data from multiple satellites, including one attached to the International Space Station.

Methane emissions originate anywhere food and plant matter decompose without oxygen, such as marshes, landfills, fossil fuel plantsand yes, cow farms . They are also infamous for their dramatic effect on air quality. Although capable of lingering in the atmosphere for just 7 to 12 years compared to
CO2s centuries-long lifespan , the gas is still an estimated 80 times more effective at retaining heat. Immediately reducing its production is integral to stave off climate collapses most dire short-term consequencescutting emissions by 45 percent by 2030, for example, could shave off around 0.3 degrees Celsius from the planets rising temperature average over the next twenty years.

[Related: Turkmenistans gas fields emit loads of methane .]

Unfortunately, its often difficult for aerial imaging to precisely map real time concentrations of methane emissions . For one thing, plumes from so-called ultra-emission events like oil rig and natural gas pipeline malfunctions (see: Turkmenistan ) are invisible to human eyes, as well as
most satellites multispectral near-infrared wavelength sensors. And what aerial data is collected is often thrown off by spectral noise, requiring manual parsing to accurately locate the methane leaks.

A University of Oxford team working alongside Trillium Technologies NIO.space has developed a new, open-source tool powered by machine learning that can identify methane clouds using much narrower hyperspectral bands of satellite imaging data. These bands, while more specific, produce much more vast quantities of datawhich is where artificial intelligence training comes in handy.

The project is detailed in new research published in Nature Scientific
Reports by a team at the University of Oxford, alongside a recent university profile . To train their model, engineers fed it a total of 167,825 hyperspectral image tileseach roughly 0.66 square milesgenerated by NASAs Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) satellite while orbiting the Four Corners region of the US. The model was subsequently
trained using additional orbital monitors, including NASAs hyperspectral EMIT sensor currently aboard the International Space Station.

The teams current model is roughly 21.5 percent more accurate at identifying methane plumes than the existing top tool, while simultaneously providing nearly 42 percent fewer false detection errors compared to the same industry standard. According to researchers, theres no reason to believe those numbers wont improve over time.

[Related: New satellites can pinpoint methane leaks to help us beat climate change .]

What makes this research particularly exciting and relevant is the fact that many more hyperspectral satellites are due to be deployed in the coming
years, including from ESA, NASA, and the private sector, Vt Rika, lead researcher and a University of Oxford doctoral candidate in the department of computer science, said during a recent university profile . As this satellite network expands, Rika believes researchers and environmental watchdogs will soon gain an ability to automatically, accurately detect methane plume events anywhere in the world.

These new techniques could soon enable independent, globally-collaborated identification of greenhouse gas production and leakage issuesnot just for methane, but many other major pollutants. The tool currently utilizes already collected geospatial data, and is not able to currently provide real-time analysis using orbital satellite sensors. In the University of Oxfords recent announcement, however, research project supervisor Andrew Markham adds that the teams long-term goal is to run their programs through satellites onboard computers, thus making instant detection a reality.

The post How AI could help scientists spot ultra-emission methane plumes fasterfrom space appeared first on Popular Science . Articles may contain affiliate links which enable us to share in the revenue of any purchases made.



======================================================================
Link to news story: https://www.popsci.com/environment/methane-plume-ai-detection/


--- Mystic BBS v1.12 A47 (Linux/64)
* Origin: tqwNet Science News (1337:1/100)