Machine learning models based on the U-Net architecture are proposed for the detection of methane leaks from anthropogenic sources.
NASA's EMIT instrument has identified over 750 emission sources of greenhouse gases, particularly methane, since August 2022.
The data collected by EMIT could aid in achieving the global methane-reduction pledge, which aims to reduce emissions by at least 30% of 2020's levels by 2030.
The largest total emissions were observed for Turkmenistan.
NASA's Earth Surface Mineral Dust Source Investigation (EMIT) instrument, launched in July 2022 and installed on the International Space Station (ISS), has been successful in identifying point-source emissions of greenhouse gases, particularly methane, from its position in space. The instrument has identified over 750 emission sources since August 2022, exceeding the expectations of its designers. The data collected by EMIT is helping to improve understanding of how dust that gets lofted into the atmosphere affects climate. The data collected by EMIT is available for use by the public, scientists, and organizations.
The EMIT instrument was originally designed to map 10 crucial minerals in Earth's dry regions, but it has also proven effective in detecting plumes of greenhouse gases, methane and carbon dioxide. The sensor combines the precision of airplane-mounted instruments with the wide coverage of satellites, enabling it to identify and quantify fine-scale methane and carbon dioxide emissions to specific sources. The study focuses on emissions from the oil and gas, waste, and energy sectors in countries with significant production and limited reporting. The largest total emissions were observed for Turkmenistan.
The data collected by EMIT could aid in achieving the global methane-reduction pledge, which aims to reduce emissions by at least 30% of 2020's levels by 2030. The use of machine learning models for the detection of methane leaks from anthropogenic sources is also discussed. The authors propose small and efficient machine learning models based on the U-Net architecture that use established representations for both hyperspectral and multispectral data. The proposed system could ease the work of experts in the field by sifting through vast amounts of data and proposing locations of interest for manual confirmation and release through relevant agencies.
The data collected by EMIT could aid in achieving the global methane-reduction pledge, which aims to reduce emissions by at least 30% of 2020's levels by 2030.
The article discusses the use of machine learning models for the detection of methane leaks from anthropogenic sources.
The authors propose small and efficient machine learning models based on the U-Net architecture that use established representations for both hyperspectral and multispectral data.
The proposed system could ease the work of experts in the field by sifting through vast amounts of data and proposing locations of interest for manual confirmation and release through relevant agencies.