Good to get this sorted out before it becomes a hobby horse of denial for interested parties.
"We generated a data set that comprised 746 release or non-release events at a fixed location near Casa Grande, Arizona, in 2025. This included measurements of 481 controlled releases with rates ranging from 10 to 1,200 kilograms of CH4 per hour, which is typical of real-world point sources such as landfills and oil and gas facilities, and 265 blank control events, in which no methane was released when the satellite passed over the site. The procedures for gas release and data reporting followed established protocols3.
Our team provided gas flow rates and a suite of meteorological measurements in a publicly accessible data set4. The satellite data were analysed independently by 15 teams from a broad global consortium of academic, government, non-profit and commercial organizations.
Satellites that were purpose-built for methane detection, such as Tanager-1 and GHGSat, performed best, with high true positive rates, largely unbiased quantification and few to no false positives for releases ranging from less than 100 to more than 1,000 kilograms of CH4 per hour (Fig. 1). Several teams that analysed multipurpose hyperspectral instruments, such as the German EnMAP and the Italian PRISMA satellites, also performed well.
Methane detection outcomes across satellites and analysis teams reveal that dedicated methane-monitoring instruments had the highest detection rates and that performance varied for multipurpose and global-scanning satellites.
Teams using AI-based methods reached detection accuracies of 0.62–0.77 (well above chance). However, most of the false positives came from commercial teams that applied AI methods to data from the European satellite Sentinel-2 and Landsat (a satellite run by the US Geological Survey and NASA), underscoring the need for continued development to improve reliability."
#ClimateChange
#methane
#RemoteSensing
#ghg
https://www.nature.com/articles/d41586-026-03075-2