Luca Brugaletta
Add HF metadata and clean links
55755eb
---
tags:
- methane
- sentinel-2
- image-segmentation
license: cc-by-nc-4.0
---
# MethaneS2CM – Plume Masks Only (Derived Dataset)
This dataset is a **plume-mask-only** derivative of
[H1deaki/S2CM_RAW](https://huggingface.co/datasets/H1deaki/S2CM_RAW).
It contains the methane plume segmentation masks (`plume.tif`) without redistributing the Sentinel-2 imagery (`s2.tif`, `s2_pre.tif`, `s2_pre_pre.tif`).
The goal is to provide a compact (~2 GB) resource focused on plume labels for users who already have access to Sentinel-2 data or who want to combine these masks with their own imagery.
---
## 1. Origin and Relationship to MethaneS2CM
The original **MethaneS2CM** dataset is a large-scale benchmark for methane emission detection from Sentinel-2 multispectral imagery, built from real plume annotations from CarbonMapper and co-located Sentinel-2 observations.
Each sample in MethaneS2CM includes:
- Sentinel-2 imagery at three time points: leak day (T), T-90, and T-365, and
- a corresponding plume mask and plume label.
This derived dataset:
- Keeps only the **plume masks** (`plume.tif`),
- Preserves the original indexing / IDs where applicable (so that users can cross-reference with the full MethaneS2CM dataset),
- Does **not** include any Sentinel-2 pixel data.
For full details on the construction and intended use of the original dataset, please refer to:
- Original dataset: [H1deaki/S2CM_RAW](https://huggingface.co/datasets/H1deaki/S2CM_RAW)
- Paper: “MethaneS2CM: A Dataset for Multispectral Deep Methane Emission Detection” (KDD 2025).
---
## 2. Data Contents
### 2.1 File structure
The repository contains only plume masks, stored as GeoTIFFs (`plume.tif`).
A typical directory layout is:
```text
.
├── README.md
└── plumes/
├── ang20160915t180022-B_0/plume.tif
├── GAO20210731t175256p0000-B_1/plume.tif
├── emi20220827t060753p04013-C_0/plume.tif
└── ...
```
Each `plume.tif` is a raster mask with:
- Single channel (binary):
- `1` (or 255) = plume pixels
- `0` = non-plume pixels
- Spatial reference / georeferencing inherited from the original Sentinel-2 product used in MethaneS2CM (e.g. UTM/WGS84).
> Note: The exact folder structure and naming convention may differ; check this README and the repository tree for the final, authoritative layout.
### 2.2 Coordinate system and resolution
Plume masks are aligned with the original MethaneS2CM Sentinel-2 patches:
- Pixel size: 20 m (Sentinel-2 SWIR resolution),
- Projection: the same as in MethaneS2CM (UTM zone corresponding to the plume location, WGS84 ellipsoid).
This ensures that users can:
- Directly overlay these plume masks on Sentinel-2 imagery downloaded from Copernicus / Google Earth Engine, or
- Combine them with the original MethaneS2CM dataset.
---
## 3. Intended Use
This plume-only dataset is suitable for:
- Training and evaluating **segmentation** models for methane plume detection when you supply your own input imagery,
- Using the masks as labels while:
- Redownloading Sentinel-2 data from the official sources, or
- Using Sentinel-2 data from the original MethaneS2CM dataset (if you also have a local copy),
- Performing analyses that focus on:
- Plume shape and morphology,
- Spatial statistics,
- Label quality and uncertainty, independently of a specific input feature set.
If you require:
- Full Sentinel-2 inputs (T, T-90, T-365) and
- The predefined dataset splits and Parquet metadata,
please use the original dataset: [H1deaki/S2CM_RAW](https://huggingface.co/datasets/H1deaki/S2CM_RAW).
---
## 4. Citation
If you use this dataset in your work, **please cite the original MethaneS2CM paper**, as this dataset is entirely derived from it:
```bibtex
@inproceedings{liu2025methanes2cm,
title = {MethaneS2CM: A Dataset for Multispectral Deep Methane Emission Detection},
author = {Liu, Hongxuan and Leung, Juliana Y and Niu, Di},
booktitle = {Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2},
pages = {5640--5651},
year = {2025}
}
```
Additionally, you may optionally reference this derived dataset as:
```bibtex
@misc{methanes2cm_plume_only,
title = {MethaneS2CM – Plume Masks Only (Derived Dataset)},
author = {<Your Name>},
howpublished = {\url{https://huggingface.co/datasets/Lucab95/MethaneS2CM_plume_only}},
year = {2025},
note = {Derived from H1deaki/S2CM_RAW under CC-BY-NC 4.0}
}
```
Please replace `<Your Name>` and `<yourname>` with your actual information.
---
## 5. License
The original **MethaneS2CM** dataset is released under the **CC-BY-NC 4.0** license.
This derived plume-only dataset is therefore also distributed under:
> **License:** CC-BY-NC 4.0 (Attribution–NonCommercial)
Key points (informal, not legal advice):
- **Attribution (BY):**
You must credit:
- the MethaneS2CM authors (Liu, Leung, Niu) and
- this derived dataset (if used explicitly),
and retain references to the original dataset and this repository where appropriate.
- **NonCommercial (NC):**
You may not use this dataset or derivatives of it for **commercial purposes** without obtaining separate permission from the rights holders.
For the full legal text, see the [CC-BY-NC 4.0 license](https://creativecommons.org/licenses/by-nc/4.0/).
---
## 6. Acknowledgments
This dataset is a derivative work of:
- MethaneS2CM: [H1deaki/S2CM_RAW](https://huggingface.co/datasets/H1deaki/S2CM_RAW).
We acknowledge:
- The **European Space Agency (ESA)** for providing Sentinel-2 data, and
- The **CarbonMapper** initiative for plume annotations and their open data efforts,
following the acknowledgments of the original MethaneS2CM dataset.