Towards Operational Automated Greenhouse Gas Plume Detection and Delineation

B.D. Bue, J.H. Lee, A.K. Thorpe, P.G. Brodrick, D. Cusworth, A. Ayasse, V. Mancoridis, A. Satish, S. Xiong, R. Duren
arXiv preprint https://arxiv.org/abs/2505.21806
Code repository: https://github.com/JPLMLIA/operational-ghg

Summary

This repository contains the ML-ready datasets used to train methane plume detection and delineation models from the AVIRIS-NG and EMIT missions. These are 256x256 pixel tiles sampled from the columnwise matched filter methane retrieval flightlines and scenes. Please refer to the above manuscript for complete details.

• COVID.tar
  
  • COVID_train.csv, COVID_test.csv - tiles and labels for the training and test sets, respectively.
  • COVID/ang*.tif - GeoTIFF tiles of methane retrieval concentrations of plumes and background
  • COVID/ang*_label.tif - PNG corresponding binary masks produced as described in the manuscript
• CACH4.tar
  • Same as above with CACH4 prefixes
• Permian.tar
  • Same as above with Permian prefixes
• multicampaign.tar
  • This dataset contains only the combined train/test dataset definitions that include all three above datasets. 
• EMIT.tar
  • Same as the AVIRIS-NG dataset with EMIT prefixes

Acknowledgements

The Earth surface Mineral dust source InvesTigation (EMIT) plume delineations were performed using a tool built with the Jet Propulsion Laboratory Multi-Mission Geographical Information System. Carbon Mapper acknowledges the generous support of its philanthropic donors. The High Performance Computing resources used in this investigation were provided by funding from the JPL Information and Technology Solutions Directorate. EMIT is supported by the National Aeronautics and Space Administration Earth Venture Instrument program, under the Earth Science Division of the Science Mission Directorate. The research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration (80NM0018D0004). © 2025. California Institute of Technology. Government sponsorship acknowledged.