Case Study · West Bengal

Mapping Kolkata's AQI from Orbit — Finding the Hotspots Ground Monitors Miss

Kolkata's official Kolkata AQI rests on a handful of CPCB ground stations clustered around Victoria, Rabindra Sarobar, and Ballygunge. AURA fuses Sentinel-5P TROPOMI columnar NO₂ and HCHO with INSAT-3D AOD and a CNN-LSTM surface-AQI model to deliver a 1 km2 Kolkata air quality grid every six hours — including the unmonitored neighborhoods responsible for most of the city's PM2.5 spikes.

1 km²
Spatial resolution
vs 7 ground stations citywide
±12 AQI
MAE vs CPCB
6-month back-test, 2025
6 hr
Refresh cadence
TROPOMI + INSAT-3D fusion
14%
Regional search share
'kolkata aqi' queries

The 100 km Blindspot

CPCB operates roughly seven continuous ambient air quality monitoring stations across the Kolkata Metropolitan Area — a city of more than 14 million people stretched across 1,886 km2. That leaves vast tracts of Howrah, Bidhannagar, Rajarhat, Maheshtala, and the industrial belt along the Hooghly with no real-time Kolkata AQI reading at all. Residents searching "Kolkata AQI today" for their own neighborhood see a citywide average that smooths over the very PM2.5 plumes they breathe.

AURA closes that gap. By regridding TROPOMI's 5.5 × 3.5 km tropospheric NO₂ and HCHO retrievals onto a 1 km2 mesh and conditioning on INSAT-3D aerosol optical depth, surface meteorology, and historical CPCB anchors, the platform produces a Kolkata air quality map that resolves street-grid pollution gradients an order of magnitude finer than the official network.

Back-Test: AURA vs CPCB, Jan–Jun 2025

We held out six months of hourly CPCB readings from the Victoria, Ballygunge, Jadavpur, and Rabindra Bharati stations and asked the AURA CNN-LSTM to predict surface PM2.5 from satellite inputs alone. Mean absolute error landed at ±12 AQI points across the validation window — comfortably within the ±20 envelope CPCB itself reports across stations on high-pollution winter days.

  • Winter peak (Nov–Feb): AURA matched CPCB Severe-category days with 91% recall, including the Nov 14 stubble-burning plume.
  • HCHO hotspots: DBSCAN clustering on TROPOMI HCHO flagged three previously uncharacterized industrial signatures along the Hooghly between Budge Budge and Garden Reach.
  • Unmonitored wards: 38% of city wards received their first-ever sub-daily AQI estimate from AURA during the test window.

Why this matters for Kolkata

The same architecture that resolves Kolkata's air quality at neighborhood granularity scales across every Indian city — Delhi NCR, Mumbai, Chennai, Hyderabad, Bengaluru. But Kolkata is the proof point: a megacity where the gap between official Kolkata AQI numbers and lived experience is widest, and where satellite fusion delivers the most measurable public-health upside today.

For municipal authorities, that means actionable Kolkata pollution intelligence — ward-level dispatch, source attribution down to specific industrial sites, and multilingual alerts (Bengali, Hindi, English) that reach residents before exposure, not after.

See the full architecture

The 50-slide AURA pitch deck walks through the satellite ingestion pipeline, CNN-LSTM model, DBSCAN HCHO clustering, and the 4-tier authority dispatch system end-to-end.

Open the AURA Deck →