Lab
DeployedMethane Minder
A live full-stack methane plume dashboard built on Carbon Mapper's published data.
- Capabilities
- Data visualisation · Full-stack engineering · Cloud deployment
- Evidence
- public
- Disclosure
- general
The problem
Methane plume detections are published, but they live in scientist-facing files. A policymaker or operator who wants to ask “which regions look concerning, and what is the trend?” has to open the raw data themselves.
Methane Minder answers a narrower question well: show the plumes on a map, and for a country, compare a statistical forecast of methane emissions against its historical baseline.
Intended user
Someone who wants a fast, honest picture of methane activity: where plumes were detected, how concentrated they are, and whether a country’s emissions trajectory is trending above what its own history would suggest. The tool informs a look, it does not replace analysis.
Architecture and interface
The build is a Vue 3 + TypeScript frontend over a FastAPI backend. The map renders plume detections with Leaflet; the risk panel uses Holt-Winters exponential smoothing from Statsmodels to forecast next-year emissions for a selected country and labels the result against its historical baseline. The historical chart is built from Our World in Data’s methane emissions series.
The deployed architecture is deliberately split: the API runs on Google Cloud Run in London, and the frontend is a static build served by Cloudflare Pages that calls the API directly with a CORS whitelist. The serving image excludes the geospatial pipeline dependencies (GeoPandas, Shapely, PyArrow live behind an optional extra), which keeps cold starts fast and the attack surface small. GitHub Actions lint, build and smoke-test both halves on every push.
Data provenance and attribution
The plume dataset is a snapshot derived from Carbon Mapper, used under its educational and research terms with visible attribution in the interface, alongside Our World in Data and OpenStreetMap credits. The data dictionary shipped with the repository documents every field. Datasets in src/data are regenerable through the documented pipeline script.
Limitations
This is a snapshot service, not a live feed: the plume set reflects the published data at pipeline run time, and forecasts are statistical extrapolations from historical emissions, not physical models. The risk label (low, medium, high) is a transparent threshold rule against the baseline, not a sophisticated risk engine — and the interface does not pretend otherwise. Non-commercial use only, per the source data licence.
Source availability
The source is Not public. The live deployment is openly visitable at methane-minder.pages.dev, the API responds at methane-minder-backend-880712225874.europe-west2.run.app with OpenAPI docs at /docs, and the README documents the deployment commands and data provenance. The repository itself stays private.