Published Research

This study developed an artificial intelligence system that uses high-resolution satellite images to automatically map oil and gas well pads and storage tanks, which are important sources of methane emissions. When tested in two major U.S. oil-producing regions, the system accurately identified most known well sites and discovered more than 70,000 previously unlisted well pads and over 169,000 storage tanks, showing that satellite-based machine learning tools can help build more complete and transparent databases to better track and reduce methane pollution.

Authors independently evaluated five major aircraft-based methane sensing platforms, including MethaneAIR, through over 700 single-blind controlled methane releases ranging from 1 to 1,500 kg CH₄/h. Most platforms reliably detected and quantified emissions above 10 kg/h, indicating strong agreement with metered release rates. The findings demonstrate that aircraft-based methane monitoring has substantially matured and is well positioned to support industrial methane management and climate policy.

Authors describe details of the processor used in MethaneAIR, which converts raw (Level 0) sensor data into calibrated, geolocated, and scientifically usable radiance data (Level 1B) through dark current correction, noise estimate, stray-light removal, spectral calibration, radiometric correction, and orthorectification. It sets the foundation for operational MethaneSAT Level 0 to Level 1B processor.

Presents a probabilistic method for detecting and removing shadows in hyperspectral imagery from MethaneAIR. The approach estimates shadow fraction at each pixel and corrects spectra while preserving key methane-sensitive features. By reducing shadow-related artifacts, particularly in critical CO₂ and CH₄ absorption bands, the method expands usable area for methane detection and strengthens accuracy of emissions monitoring from airborne and satellite platforms.

Describes two independent quantification methods for methane plumes detected by MethaneAIR and demonstrates their accuracy using controlled release experiments. Authors quantified methane emissions using a modified integrated mass enhancement method and a divergence integral method. Comparison with controlled release experiments in 2021 and 2022 show that the accuracy of the sensor and algorithms is better than 25% for sources emitting 200 kg h−1 or more.

Authors created the first large, openly accessible global database of oil and gas infrastructure locations to help track methane emissions, a major driver of climate change. The database includes about 6 million records—such as wells, pipelines, refineries, and compressor stations, along with details about their operations and ownership. By linking this information with satellite and aircraft measurements, the database helps identify specific methane sources and supports better monitoring and targeted emission reductions worldwide.

Reviews how current and upcoming satellites can measure methane emissions, from global scale down to individual facilities. Some satellites, such as GOSAT, TROPOMI, and MethaneSAT track emissions over larger regions and areas, while others like GHGSat can spot large leaks from specific sites. Together, these systems are becoming powerful tools for monitoring methane emissions and supporting climate agreements, with new satellites expected to improve coverage and accuracy even further.

MethaneAIR, the airborne precursor to the MethaneSAT satellite, is designed to detect and quantify methane emissions using two high-resolution imaging spectrometers. This study details the laboratory-based spectral calibration process, including stray light correction, wavelength registration, and the construction of instrument spectral response functions. The resulting calibration was validated using flight data over Colorado, establishing a foundation for MethaneSAT’s future calibration and performance.

Accurate global measurements of CO₂ and methane from space require simultaneous oxygen observations to determine the path length of backscattered sunlight. Although the oxygen singlet Delta (a1Δg) band at 1.27 μm has been considered unusable due to strong atmospheric airglow, this study shows that the airglow signal can be effectively separated from reflected sunlight using limb satellite observations. The findings suggest that using this band could simplify instrument design, lower costs, and improve the accuracy of future missions.