methaneset-s2-finetune/README.md

MethaneSET-S2 Finetune: Verified Methane Plume Events from Sentinel-2 for Supervised Learning

methaneset-s2-finetune is the verified plume subset of MethaneSET-S2, designed for supervised fine-tuning of methane detection and segmentation models. This subset contains Sentinel-2 imagery with manually verified methane plumes, binary segmentation masks, and methane enhancement maps (ΔXCH₄ in ppb). Unlike MARS-S2L which provides only six common bands, MethaneSET retrieves all 13 Sentinel-2 L1C bands at 10m GSD (200x200 pixel chips), enabling research with coastal aerosol, water vapour, cirrus, and red edge channels. Each sample includes target and reference image pairs, plume segmentation masks, CH4 enhancement images, Cloud Score+ masks, wind vectors (ERA5-Land onshore, GEOS-FP offshore), solar/viewing geometry, emission rates with uncertainties, elevation (Copernicus DEM GLO-30), and 64-dim AlphaEarth Foundation embeddings.

Dataset Information

Version: 1.0.0

License: CC-BY-4.0

Keywords: methane, finetune, supervised, segmentation, plume-detection, remote-sensing, Sentinel-2, MSI, earth-observation, deep-learning

Tasks: segmentation, classification, detection

Dataset Overview

Partitions: 21 files Spatial coverage: [-103.98, -36.37, 116.52, 49.97] (WGS84) Temporal coverage: 2018-01-05 to 2024-12-30

Dataset Structure (Root-Sibling Uniform Tree)

Root: FOLDER (3,612 samples)

Hierarchy:

• Level 1: FILE → FILE → FILE → FILE → FILE (18,060 samples)

Metadata Fields

LEVEL0

Field	Type	Description
id	string	Unique sample identifier within parent scope. Must be unique among siblings.
type	string	Sample type discriminator (FILE or FOLDER).
stac:crs	string	Coordinate reference system (WKT2, EPSG, or PROJ)
stac:tensor_shape	list<item: int64>	Raster dimensions [bands, height, width]
stac:geotransform	list<item: double>	GDAL affine transform
stac:time_start	timestamp[us]	Start timestamp (μs since Unix epoch, UTC)
stac:centroid	binary	Center point in EPSG:4326 (WKB)
stac:time_end	timestamp[us]	End timestamp (μs since Unix epoch, UTC)
stac:time_middle	timestamp[us]	Middle timestamp (μs since Unix epoch, UTC)
detection:isplume	bool	Whether a methane plume is present
detection:ch4_fluxrate	float	Methane flux rate (kg/h)
detection:ch4_fluxrate_std	float	Standard deviation of flux rate
detection:sector	string	Emission sector (Oil and Gas, Coal, Waste, etc.)
detection:offshore	bool	Whether location is offshore
detection:wind_source	string	Wind data source (e.g. ERA5-Land, GEOS-FP)
detection:case_study	string	Case study area name (e.g. Permian Basin)
satellite:platform	string	Satellite platform (S2A, S2B, LC08, LC09)
satellite:tile	string	Product identifier
satellite:vza	float	Viewing zenith angle (degrees)
satellite:sza	float	Solar zenith angle (degrees)
satellite:background_tile	string	Reference image product identifier
quality:percentage_clear	float	Percentage of clear pixels (0-100)
quality:observability	string	Image quality classification
quality:notified	bool	Whether observation has been notified
quality:last_update	string	Last registry modification timestamp (ISO format)
plume:geometry	binary	Plume extent as WKB geometry
site:country	string	Country of the emission source
site:location_name	string	Site location identifier
meteo:wind_u	float	U-component of wind at 10m (m/s)
meteo:wind_v	float	V-component of wind at 10m (m/s)
split	string	Dataset partition identifier (train, test, or validation)
majortom:code	string	MajorTOM spherical grid cell identifier (e.g., 0100km_0003U_0005R) with ~dist_km spacing
geoenrich:elevation	float	Mean elevation in meters (GLO-30 DEM)
geoenrich:temperature	float	Mean annual temperature in °C estimated from MODIS LST data
geoenrich:population	float	Population density from HRSL. Facebook High Resolution Settlement Layer
geoenrich:admin_countries	string	Country name at centroid location
geoenrich:admin_states	string	State/province name at centroid location
geoenrich:admin_districts	string	District/county name at centroid location
internal:current_id	int64	Current sample position at this level (0-indexed). Enables O(1) random access and relational JOINs (ZIP, FOLDER, TACOCAT).
internal:parent_id	int64	Foreign key referencing parent sample position in previous level (ZIP, FOLDER, TACOCAT).

LEVEL1

Field	Type	Description
id	string	Unique sample identifier within parent scope. Must be unique among siblings.
type	string	Sample type discriminator (FILE or FOLDER).
geotiff:stats	list<item: list<item: float>>	Per-band statistics (List[List[Float32]]): categorical mode returns class probabilities, continuous mode returns [min, max, mean, std, valid%, p25, p50, p75, p95]
taco:header	binary	Binary TACOTIFF header (35 bytes + tile counts) for fast reading without IFD parsing
internal:current_id	int64	Current sample position at this level (0-indexed). Enables O(1) random access and relational JOINs (ZIP, FOLDER, TACOCAT).
internal:parent_id	int64	Foreign key referencing parent sample position in previous level (ZIP, FOLDER, TACOCAT).
internal:relative_path	string	Relative path from DATA/ directory. Format: {parent_path}/{id} or {id} for level0 (ZIP, FOLDER, TACOCAT).

Usage

Python

# pip install tacoreader
import tacoreader

ds = tacoreader.load("methaneset-s2-finetune.tacozip")
print(f"ID: {ds.id}")
print(f"Version: {ds.version}")
print(f"Samples: {len(ds.data)}")

R

# Coming soon: R support is planned but not yet available
# install.packages("tacoreader")
library(tacoreader)

ds <- load_taco("methaneset-s2-finetune.tacozip")
cat(sprintf("ID: %s\n", ds$id))
cat(sprintf("Version: %s\n", ds$version))
cat(sprintf("Samples: %d\n", nrow(ds$data)))

Julia

# Coming soon: Julia support is planned but not yet available
# using Pkg; Pkg.add("TacoReader")
using TacoReader

ds = load_taco("methaneset-s2-finetune.tacozip")
println("ID: ", ds.id)
println("Version: ", ds.version)
println("Samples: ", size(ds.data, 1))

Data Providers

UNEP IMEO — producer

Source Cooperative — host

Dataset Curators

Name	Organization	Email
Cesar Aybar	Universitat de València, Image and Signal Processing (ISP) Group	cesar.aybar@uv.es

Publications & Citations

If you use this dataset in your research, please cite:

DOI: 10.48550/arXiv.2411.15452

Vaughan, A., Mateo-Garcia, G., Irakulis-Loitxate, I., Watine, M., Fernandez-Poblaciones, P., Turner, R. E., Requeima, J., Gorroño, J., Randles, C., Caltagirone, M., & Cifarelli, C.* (2024). AI for operational methane emitter monitoring from space. arXiv preprint arXiv:2411.15452.

Operational MARS-S2L system for global methane monitoring from Sentinel-2 and Landsat 8/9.

---

DOI: 10.48550/arXiv.2511.21777

Vaughan, A., Mateo-Garcia, G., Irakulis-Loitxate, I., Watine, M., Fernandez-Poblaciones, P., Turner, R. E., Requeima, J., Gorroño, J., Randles, C., Caltagirone, M., & Cifarelli, C.* (2024). Artificial intelligence for methane detection: from continuous monitoring to verified mitigation. arXiv preprint arXiv:2511.21777.

Extended operational deployment demonstrating 1,015 stakeholder notifications across 20 countries and verified permanent mitigation of six persistent emitters.

---

DOI: 10.5194/essd-13-4349-2021

Muñoz-Sabater, J., et al. (2021). ERA5-Land: a state-of-the-art global reanalysis. Earth System Science Data, 13, 4349-4383.

---

DOI: 10.1029/2014JD022685

Lucchesi, R. (2013). GEOS-5 FP (Forward Processing) File Specification. NASA GMAO Technical Report.

---

BibTeX

@dataset{methaneset-s2-finetune1,
  title = {MethaneSET-S2 Finetune: Verified Methane Plume Events from Sentinel-2 for Supervised Learning},
  author = {Cesar Aybar},
  year = {2018},
  version = {1.0.0},
  publisher = {Universitat de València, Image and Signal Processing (ISP) Group}
}

---

Generated with ❤️ using TacoToolbox v0.26.9