Machine learning-based emission rate estimates of global  methane super-emissions

Roberts, Clayton; Maasakkers, Joannes D.; de Jong, Tobias A.; Schuit, Berend J.; Sharma, Shubham; Huegens, Theo; van den Berg, Anne-Wil; Houweling, Sander; Aben, Ilse

doi:10.5194/amt-19-5659-2026

Articles | Volume 19, issue 17

https://doi.org/10.5194/amt-19-5659-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/amt-19-5659-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 19, issue 17

Research article

|

08 Sep 2026

Research article |

| 08 Sep 2026

Machine learning-based emission rate estimates of global methane super-emissions

Clayton Roberts, Joannes D. Maasakkers, Tobias A. de Jong, Berend J. Schuit, Shubham Sharma, Theo Huegens, Anne-Wil van den Berg, Sander Houweling, and Ilse Aben

Data sets

Simulated WRF-Chem Plume Output Resampled to TROPOMI Pixel Footprints Clayton Roberts et al. https://doi.org/10.5281/zenodo.21808941

Simulated HYSPLIT Plume Output Resampled to TROPOMI Pixel Footprints Clayton Roberts et al. https://doi.org/10.5281/zenodo.21792154

Dataset: all TROPOMI detected plumes for 2021. [Schuit et al. 2023: Automated detection and monitoring of methane super-emitters using satellite data] Berend J. Schuit et al. https://doi.org/10.5281/zenodo.8087133

ERA5 hourly data on pressure levels from 1940 to present Hans Hersbach et al. https://doi.org/10.24381/cds.bd0915c6

ERA5 hourly data on single levels from 1940 to present Hans Hersbach et al. https://doi.org/10.24381/cds.adbb2d47

Model code and software

ML-SPERE Trained CNN Model File for TROPOMI Methane Super-Emitter Emission Rate Estimation Clayton Roberts et al. https://doi.org/10.5281/zenodo.21786440

ML-SPERE Codebase for TROPOMI Methane Plume Emission Rate Estimation (v1.0.4, Publication Release, for use within SRON) C. Roberts https://doi.org/10.5281/zenodo.21934190

Short summary

Methane is a powerful greenhouse gas, and cutting its emissions can slow climate change in the near term. We have created ML-SPERE, a new machine-learning based method that uses TROPOMI observations and weather data to better estimate emission rates for large methane plumes. It has proved more accurate than the Integrated Mass Enhancement (IME) method when tested on simulated satellite observations. When applied to real TROPOMI plumes, ML-SPERE produces estimates that agree with other methods.