Operational machine learning for remote spectroscopic detection of CH4 point sources

More details will be added.

Please refer to the paper preprint "Operational machine learning for remote spectroscopic detection of CH4 point sources", arXiv preprint arXiv:2511.07719 (2025).

Slides from AGU2025 • Quick inference demo with EMIT • Datasets • Trained models

---

Operational machine learning for remote spectroscopic detection of CH4 point sources

Abstract

Mitigating anthropogenic methane sources is one of the most cost-effective levers to slow down global warming. While satellite-based imaging spectrometers, such as EMIT, PRISMA, and EnMAP, can detect these point sources, current methane retrieval methods based on matched filters produce a high number of false detections requiring laborious manual verification. To address this challenge, we deployed a machine learning system for detecting methane emissions within the Methane Alert and Response System (MARS) of the United Nations Environment Programme's International Methane Emissions Observatory. This represents the first operational deployment of automated methane point-source detection using spaceborne imaging spectrometers, providing regular global coverage and scalability to future constellations with even higher data volumes.

Adapting plume detection for operational deployment required several technical advances. First, we created one of the largest and most diverse and global machine learning ready datasets to date of annotated methane plumes from three imaging spectrometer missions, and quantitatively compared different deep learning model configurations. Second, we extended prior evaluation methodologies from small, tiled datasets to full granules that are more representative of operational use. This revealed that deep learning models still produce a large number of false detections, a problem we addressed with model ensembling, which reduced false detections by over 74%. Deployed in the MARS pipeline, our system processes scenes and proposes plumes to analysts, accelerating the detection and analysis process. During eleven months of operational deployment, it processed more than 25,000 hyperspectral products faciliting the verification of 2,851 distinct methane leaks, which resulted in 834 stakeholder notifications. We further demonstrate the model's utility in verifying mitigation success through case studies in Libya, Argentina, Oman, and Azerbaijan. Our work represents a critical step towards a global AI-assisted methane leak detection system, which is required to process the dramatically higher data volumes expected from current and future imaging spectrometers.

Dataset

Datasets presented in this paper are uploaded to huggingface inside this collection: UNEP-IMEO/mars-hyperspectral

Code examples

Install

# Option 1:
conda env create -f environment.yml
conda activate hyper_mars

# Option 2:
conda create -c conda-forge -n hyper_mars python=3.11
conda activate hyper_mars
pip install -r requirements.txt

# Option 3: run our examples on Google Colab - they include minimal library dependencies setup

Inference

To start using our model for inference, you can check the demo with AVIRIS data in Colab Inference on EMIT scenes .

Our trained models are stored in Hugging Face🤗 at UNEP-IMEO/MARS-Hyperspectral-models.

Training

To reproduce the same training process as reported in the paper, you will need to download the full MARS-Hyperspectral EMIT dataset from Hugging Face, and prepare the coding environment.

# Check possible parameters with:
!python3 -m scripts.run_train --help

# And run the prepared training script used for the paper models (remember to download and adjust the paths to the training datasets)
./bash/train_UNET_RGB_WMF.sh

Citation

If you find our models and datasets useful in your research, please consider citing our work.

@article{ruzicka2025operational,
  title={Operational machine learning for remote spectroscopic detection of CH $ \_ $\{$4$\}$ $ point sources},
  author={Růžička, Vít and Mateo-Garcia, Gonzalo and Irakulis-Loitxate, Itziar and Johnson, Juan Emmanuel and Martin, Manuel Montesino San and Allen, Anna and Guanter, Luis and Thompson, David R},
  journal={arXiv preprint arXiv:2511.07719},
  year={2025}
}

Acknowledgments

We would like to thank the analysts of MARS UNEP IMEO for their valuable contributions in identifying methane plumes. In addition to the authors of this paper, we gratefully acknowledge the efforts of Marc Watine and Adriana Valverde. We acknowledge the rest of the MARS team Meghan Demeter, Tharwat Mokalled, Giulia Bonazzi, Florencia Carreras, Konstantin Kosumov and Queen Safari, for their work in country engagement and in the notification process. A portion of this work was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration (80NM0018D0004). Part of Vít Růžička's research was supported by an appointment to the NASA Postdoctoral Program at the Jet Propulsion Laboratory, administered by Oak Ridge Associated Universities under contract with NASA.