methane-plume-detector

Methane Plume Detector

Description

This repository contains a computer vision model to detect methane plumes in Sentinel-2 satellite imagery. The model is UNet-64 based and was trained on a dataset of synthetic plumes embedded in Sentinel-2 observations of the Permian Basin (TX, USA). It is hosted on HuggingFace for easy access and use by the community.

Model Details

More information on how the model was trained can be found in this Master Thesis report.

The model's architecture can be found in ch4_unet.py

Usage

This model takes in input 512x512x11 Sentinel-2 images. The 11 bands must be the following (in this order)

• Band 2 (Blue)
• Band 3 (Green)
• Band 4 (Red)
• Band 8 (NIR)
• Band 10 (SWIR - Cirrus)
• Band 11 (SWIR1)
• Band 12 (SWIR2)
• Normalized Difference Methane Index (NDMI)
• Normalized Difference Building Index (NDBI)
• Normalized Difference Vegetation Index (NDVI)
• Bare Soil Index (BSI)

For more information of preprocessing steps such as resolution upsampling, normalized difference indeces computation or band standardization, please refer to section 3 in the above linked report.

License

This model is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. For more details, see the LICENSE file in this repository or visit the Creative Commons website.

Citation

If you use this model in your research, please cite it as follows:

@misc{Methane Plume Detector,
  author = {Francois Martin-Monier},
  title = {Methane Plume Detector},
  year = {2024},
  howpublished = {\url{https://huggingface.co/fmartinmonier/methane-plume-detector}}
}

Contact

For any questions or feedback, please open an issue in this repository.

Acknowledgements

This model was developped during a Master Thesis at Harvard University's Atmospheric Chemistry Modeling Group. I'd like to thank Prof. Daniel Jacob and Daniel Varon for their guidance, support and expertise.