Methane is a focus for reducing greenhouse gas emissions due to its high global-warming potential. It exceeds carbon dioxide by over 80-fold in the short term. However, tracking methane emissions and measuring their amounts has been difficult because of limitations with current detection methods. Scientists have created a way to automatically detect methane emissions on a global level, which could potentially provide frequent and detailed methane detection from specific sources.
Global warming potential in the short term exceeds carbon dioxide by over 80-fold. Monitoring methane emissions and compiling their quantities have been challenging due to limiting trade-offs with existing detection methods. Researchers have created a method to automatically detect methane emissions at a global scale, potentially providing methane detection at high frequency and resolution from point sources. As global temperatures rise to record highs, the pressure to curb greenhouse gas emissions has intensified. Methane is particularly targeted because of its significant global-warming potential in the short term.The amount of methane emitted in the short term is more than 80 times greater than carbon dioxide. However, it has been difficult to monitor and quantify methane emissions due to limitations with current detection methods. A research team from Kyoto University and Geolabe, USA has developed a new method to automatically detect methane emissions on a global scale. Lead author Bertrand Rouet-Leduc of KyotoU’s Disaster Prevention Research Institute and Geolabe says that their approach has the potential to detect methane at a high frequency and resolution from point sources, which could lead to a systematic quantification method.Rouet-Leduc’s method could potentially be used to prioritize and validate efforts to reduce methane emissions in the atmosphere, which currently contribute to about one third of global warming.
In recent years, multispectral satellite data has become an effective tool for detecting methane, allowing for regular monitoring of methane emissions on a global scale every few days. However, the data is often affected by significant noise, and previous detections have been limited to large emissions and required human verification.
On the other hand, the team has trained an AI to automatically detect methane detections without human intervention.methane leaks over 200kg/h, accounting for more than 85% of the methane emissions in well-studied, large oil and gas basins.
“When using satellite measurements, there are trade-offs to consider in terms of spatial coverage, spatial and temporal resolution, and spectral resolution, as well as associated detection accuracy. AI helps to partially balance out these trade-offs,” explains co-author Claudia Hulbert, also of Geolabe.
Methane plumes are invisible and odorless, so they are typically detected with specialized equipment such as infrared cameras. It is extremely challenging to find these leaks from space, similar to finding a needle in a haystack. Leaks are difficult to discern without the right tools. rnrnThe group’s collaborative work is a significant advancement in the precise and systematic monitoring of methane emissions globally, with most methane plumes being small and easily overlooked in satellite data. “Automation is crucial for analyzing large areas. We were amazed to find that AI can automate the process and surpass the human eye in detecting small methane plumes,” Rouet-Leduc said. In the next phase, the group plans to integrate additional satellites in a global study of methane emissions. The Journal Reference is not available.
<p>The following article was recently published in <strong>Nature Communications</strong>:</p>
<ol class="journal">
<li>Bertrand Rouet-Leduc, Claudia Hulbert. <strong>Automatic detection of methane emissions in multispectral satellite imagery using a vision transformer</strong>. <em>Nature Communications</em>, 2024; 15 (1) DOI: <a href="http://dx.doi.org/10.1038/s41467-024-47754-y" style="text-decoration: underline" rel="noopener noreferrer" target="_blank">10.1038/s41467-024-47754-y</a> </li>
</ol>
<hr />