In a groundbreaking effort to tackle Delhi's persistent air quality crisis, researchers from the Indian Institute of Technology (IIT) Kanpur have deployed a state-of-the-art mobile laboratory equipped with artificial intelligence. The initiative, led by a dedicated team of over 20 scientists, aims to move beyond generic pollution readings and pinpoint exactly where and when toxic emissions originate in different parts of the capital.
The Mobile Lab: A Laboratory on Wheels
At the heart of this project is a custom-built, heavy-duty van that functions as a fully equipped atmospheric research facility. Professor Sachchida Nand Tripathi, Dean of the Kotak School of Sustainability and Project Director of the AI Centre of Excellence for Sustainable Cities at IIT-Kanpur, describes it as a "lab on wheels." This mobile unit, valued at over Rs 22 crore, houses an array of sophisticated instruments rarely found outside advanced labs.
The van's arsenal includes a high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS) for near-molecular analysis of airborne particles, a real-time metal monitor that acts as a chemical fingerprint to distinguish between sources like road dust and vehicle exhaust, and regulatory-grade gas analysers. It also carries aethalometers for black carbon, particle-size analysers, and high-precision meteorological sensors.
Revealing Delhi's Divergent Pollution Stories
Between May and June 2025, the mobile lab was stationed at two contrasting locations: the notoriously polluted Anand Vihar junction and the calmer suburban area of Dwarka, roughly 30 km away. The data revealed starkly different pollution profiles, challenging the notion of a single, city-wide problem.
At Anand Vihar, the lab recorded an average PM 2.5 concentration of about 63 μg/m³, with more than half being organic matter. The chemical breakdown showed a complex mix: 34% from road dust, 26.9% from sulphur-rich particles, and 16.7% from chlorine-rich emissions. This indicated a combination of local traffic and dust with regionally transported pollution.
In Dwarka, the average PM 2.5 was lower at 38 μg/m³, but over 80% of it consisted of secondary organic aerosols—particles formed in the air from gaseous precursors. Black carbon levels were consistently low, painting a picture of a background environment rather than an intense local hotspot.
How AI Bridges the Data Gap
The true innovation lies in combining this detailed chemical data with AI. For 10-12 days at each site, the advanced mobile lab operates alongside networks of low-cost sensors. Machine learning models are trained to establish a relationship between the simple sensor readings and the complex molecular data from the lab.
"Once trained, the model needs only sensor data to identify sources," explained Professor Tripathi. This approach, already tested with over 90% accuracy in Lucknow, can then be scaled. Instead of installing crore-rupee labs everywhere, trained low-cost sensors can provide hyper-local, real-time source apportionment across neighbourhoods—a capability current regulatory stations lack.
Tripathi emphasizes that Delhi's challenge is not a lack of solutions, but a lack of precision. The goal is to enable authorities to know exactly where to act, when to act, and how narrowly action can be focused. This project builds upon earlier work like the Real-Time Delhi Air Quality Experiment (2018-2022) but marks a significant leap in scale, speed, and the application of artificial intelligence for environmental management.
