Hourly air quality archive and live dashboard for the Kashmir Valley — 28 towns from Uri to Pahalgam, Gurez to Banihal.
Live Dashboard: syedhamidali.github.io/aqikashmir — MapLibre map of current AQI per station, click a point for readings + history. A toggleable shaded AQI surface interpolates between stations and draws filled contour bands across the valley. On open, the dashboard fetches live readings directly from the Open-Meteo API (badge shows "Live"); if that's unreachable it falls back to the last hourly-logged snapshot (badge shows "Offline"). Each station panel has a "Download CSV" button for that station's time series.
Where the numbers come from. There is no public network of live ground sensors covering Kashmir, so this project does not scrape physical monitors. Readings are sampled from the ECMWF CAMS global atmospheric composition model via the Open-Meteo Air Quality API at each town's coordinates. That is a ~40 km modelled estimate, not a calibrated instrument reading — good for regional trends and day-to-day comparison, not a substitute for a CPCB reference station. See Why there are no ground sensors here.
Data
- Latest snapshot (one point per station):
output/geojson/aqi-latest.geojson - Full time series, one CSV per day:
output/csv/ - Station/file index:
output/manifest.json
A GitHub Actions workflow runs every hour and:
- Fetches hourly air quality + weather for all 28 stations from Open-Meteo (two batched requests, no API key)
- Writes
output/geojson/aqi-latest.geojson— a snapshot with the latest reading per station - Appends only genuinely new readings to a daily CSV under
output/csv/, deduplicated against the per-station high-water mark inoutput/state/last-timestamps.json - Updates
output/manifest.json, which the dashboard uses to know which stations and CSV files exist
Because each run requests a full past day and only appends what is new, a missed run is not lossy.
npm install # no external dependencies today, but keeps this future-proof
npm test # checks the API is reachable and returning values for Kashmir
npm start # fetch latest data + append to today's CSV logTo seed history on a fresh checkout (the API serves up to 92 past days):
rm -f output/csv/*.csv output/state/last-timestamps.json # clear the dedup high-water marks first
AQI_PAST_DAYS=92 npm startThe same backfill is available from the Actions tab — run the Refresh AQI Data workflow manually and set past_days.
python3 -m http.server 8080
# open http://localhost:8080/index.htmlEdit src/stations.js — the dashboard reads its station list from the generated snapshot, so no frontend change is needed. Keep station_id stable once it has been published, or that station's archived history stops joining up. Points closer together than about a model grid cell (~40 km) will report near-identical values.
This was investigated with a live WAQI API token on 23 August 2026, and the answer is that there is currently no usable ground-monitoring data for the Kashmir Valley:
- A bounding-box query across all of Jammu & Kashmir and Ladakh (32–36°N, 73–80°E) returns zero stations.
- The one CPCB reference monitor in the valley — Rajbagh, Srinagar (WAQI uid 13694) — last reported on 23 June 2026 (its final reading was AQI 52). It is not currently publishing.
- The nearest station reporting live data is 608 km away, in Kashgar, China.
Separately, WAQI's terms of service state that their data "can not be redistributed as cached or archived data" — which is incompatible with this project, whose entire purpose is a public CSV archive. WAQI is therefore only usable for live on-page display, not for the archive.
OpenAQ remains the one route that would permit archiving (its data is CC BY 4.0), so if the Srinagar station comes back online, that is where to integrate it. The hook would be in src/fetch-aqi.js: fetch measured stations alongside the model, tag each feature measured vs modelled, and style the two differently on the map.
Until then, the CAMS model is genuinely the best available data for the region — which is the reason this project exists.
output/geojson/aqi-latest.geojson — one Point feature per station. Properties include every sensor field (aqi, pm25, pm10, co, co2, no2, o3, so2, dust, uv, temp, hum, plus <field>_unit), station_id, name, district, online, and time (timestamp of the latest reading).
output/csv/aqi-YYYYMMDD.csv (bucketed by IST calendar day) — columns: time, station_id, name, latitude, longitude, aqi, pm25, pm10, co, co2, no2, o3, so2, dust, uv, temp, hum. One row per station per hour.
aqi is the US EPA AQI, matching the colour breakpoints in the dashboard legend (0/51/101/151/201/301).
The optional shaded layer inverse-distance-weights (power 2) the 28 station values onto a raster and colours it with the same EPA breakpoints as the legend, with isolines at an interval chosen to suit the current spread. It is an interpolation, not a measurement: between stations it is a smooth guess that ignores terrain, and Kashmir's air quality is strongly shaped by terrain. It deliberately fades out more than ~70 km from any station rather than filling its bounding box, so it never implies coverage it doesn't have.
Structure, dashboard and data-logging approach adapted from publicmap/goa-aqi-logger. Weather and air quality data by Open-Meteo (CC BY 4.0), derived from ECMWF CAMS. Basemap by OpenFreeMap / OpenStreetMap contributors.