AirAtlas — India Air Quality + Weather Dashboard
Live site → sarvesh-tewari.github.io/AirAtlas
A free, open-source, public dashboard for Indian cities showing air quality (AQI)
and weather, with multi-year seasonal trends, severity at a glance,
pollution-vs-weather correlation, and city-to-city comparison.
The default AQI standard is India's NAQI, with a user toggle to US EPA AQI
and EU EAQI — all computed from raw pollutant concentrations, so identical air
can be compared across standards.
Why this exists
AQI is a formula, not a measurement. The same air reads "Moderate" under India's
NAQI but "Unhealthy" under the US EPA standard, because the breakpoint tables differ.
AirAtlas stores raw pollutant concentrations and computes any standard on demand,
making those differences transparent for a policy and public audience.
What you can do
- See the current AQI for 285+ Indian cities (geolocates to your nearest covered city).
- Toggle between NAQI, US EPA, and EU EAQI and watch the same air re-classify.
- Explore multi-year daily history — seasonal trends, monthly heatmaps, year-by-year summaries.
- Break the index down into its individual pollutants (PM2.5, PM10, NO2, SO2, CO, O3, NH3).
- See how many days each year breached each category, and how weather tracks pollution.
- Compare cities side by side.
Methodology (summary)
- Air quality comes from CPCB monitoring stations (Central Pollution Control Board),
sourced via OpenAQ, which keeps multi-year history of the same stations. The headline
shows the latest available reading and refreshes daily, so it can lag by about a day.
- Weather (current + historical) comes from Open-Meteo.
- AQI for all three standards is computed in
pipeline/aqi/ from raw concentrations,
respecting each standard's averaging windows and breakpoint tables.
- A city's value is the mean pollutant concentration across its stations, computed
identically for every standard so the comparison is apples-to-apples.
Full methodology, breakpoint versions, and the source reconciliation rule live on the
About page in the app and in SOURCES.md.
Architecture
Static site + scheduled data pipeline — no always-on server.
- Pipeline (Python): fetch → compute AQI → aggregate → reconcile → write Parquet/JSON.
Runs on GitHub Actions cron (hourly live + daily history).
- Data: two-tier Parquet (multi-year daily + last ~90 days hourly) plus a small live
JSON snapshot. The published Parquet files are the open dataset.
- Frontend: React + Vite + TypeScript, querying the Parquet in-browser via DuckDB-WASM.
Charts with Apache ECharts; map with Leaflet.
- Hosting: GitHub Pages.
Repository layout
pipeline/ Python data pipeline: AQI engine, ingestion, transform, orchestration
data/ Published data — live/ (JSON), history/ + recent/ (Parquet), meta/
web/ React + Vite + TypeScript frontend
.github/ Scheduled refresh + deploy workflows
Local development
Prerequisites: Python 3.11+, Node.js 22+, Git.
# Pipeline
cd pipeline
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
pytest # run the pipeline + AQI engine tests
# Frontend
cd ../web
npm install
npm run dev
Copy .env.example to .env and add your API keys for local pipeline runs.
Never commit real keys.
Contributing
Contributions, bug reports, and data-accuracy feedback are welcome, all through GitHub.
See CONTRIBUTING.md for setup, the checks to run before a PR, and conventions.
Licensing
- Code: MIT
- Data: CC BY 4.0 — attribute CPCB, OpenAQ, and Open-Meteo.
Acknowledgements
Air-quality data © CPCB (Central Pollution Control Board), distributed via OpenAQ
(CC BY); weather data © Open-Meteo (CC BY 4.0).
