Inside Delhi’s urban heat island: How AI and community action are combating the crisis

In Delhi’s Vivekananda Camp, a summer afternoon isn’t just hot—it’s oven-level scorching. The tin roofs glare, the walls radiate heat, and even shade feels warm, so the  children play in narrow alleys under makeshift shade. Homes in such settlements are often built with whatever materials are available: tin sheets, asbestos, plastic, or unplastered brick. They absorb heat quickly and release it slowly, turning living spaces into ovens. But in 2024, something unexpected happened. A tech-powered, community-led experiment turned this bustling basti into a lab for climate resilience. And no, it didn’t involve air conditioners or skyscraper gardens. Instead, it used AI, old saris, jute, and a lot of local hustle. With the help of AI and grassroots collaboration, Vivekananda Camp became the testing ground for a pilot to combat extreme urban heat.

A growing crisis written in Celsius

In a paper titled “Building Heat-Resilient Communities: A Collaborative Approach to Beat the Heat,” global heat extremes are becoming the new normal. Authors Haripriya Kesavan, Samhita Rb, and Anshu Sharma present a first-of-its-kind framework that blends cutting-edge technology with community participation to address urban heat risks. “The consequences of rising urban heat extend beyond discomfort as it adversely impacts health and psychosocial wellbeing,” they write. “Not all urban residents experience these impacts equally.”

In 2023 alone, 78% of the world’s population endured at least a month of extreme heat. Once-a-decade heatwaves now occur eight times per century, and with just 2°C of global warming, they could become 5.6 times more frequent. And in India, urban heat islands (UHI)—areas where concrete, traffic, and zero ventilation stew up the heat—are making cities like Delhi especially vulnerable.

When AI meets jugaad

To combat these localised risks, Resilience AI, a Gurgaon-based Tech4Impact start-up co-founded by the authors, developed Resilience360™, an AI-powered risk management platform. Within it is ResSolv™, a proprietary tool that uses satellite imagery, geoclimatic data, and machine learning to map heat risk down to the building level.

“Our approach blends the technical precision of large language models (LLMs) with ground-truthed built-environment databases,” the paper explains, highlighting the use of advanced AI models like UNet and Inception v3 to detect risk patterns. “Machine learning processes over 14 geo-climatic, geo-spatial and built-form parameters… to pinpoint heat ‘hotspots.’”

Cool roofs, hot data

Armed with super-specific, rooftop-level heat data, the team got to work—not with fancy gadgets, but with humble materials like jute, cardboard, insulation sheets, and recycled tin. They rolled out three different “cool roof” models, each one built with local, low-cost resources and tested against a regular, heat-trapping tin-roofed control building.

The results? Nothing short of sizzling success. Model 2 emerged as the coolest kid on the block, clocking an indoor temperature of just 33.1°C, while Model 1 wasn’t far behind at 33.8°C. Meanwhile, the control building baked at a sweltering 45.8°C. Even Model 3—the budget-friendly version—cut peak temperatures by a solid 19%.

Shade, Saris, and Sensors

Beyond structural changes, the authors stress the importance of behavioural and social resilience. The project supported “community-driven street shading initiatives,” where residents used recycled materials—scarves, saris, and plastic bottles—to create shaded zones. Drinking water stations, long defunct, were revitalised through “collaborative and community-led efforts” to ensure access to safe hydration during peak heat hours.

But perhaps the most transformative component was the development of a community-based early warning system (EWS). “Using Automatic Weather Stations (AWS) for metrics like temperature and humidity, a community-centred warning service was managed by local volunteers,” the authors note. Residents were trained to operate data loggers, record heat stress data, and issue tailored advisories based on a heat stress chart placed visibly across the neighbourhood.

From Heat Hacks to Heat Policy

So what’s next? The authors make a compelling case for scaling this model nationally. “Reimagining heat action demands a comprehensive approach that resonates with the realities on the ground,” they argue. They call for policy shifts that include:

• Short-term measures, like releasing risk-based heatwave advisories and deploying cooling kits;

• Medium-term interventions, such as adopting natural cooling through urban greenery, following the “3-30-300” rule (3 trees per person, 30% green cover, a park within 300 meters);

• And long-term strategies, including statutory mandates for climate vulnerability assessments and the development of an indoor heat index.

They also underscore how ResSolv™ accelerates planning by offering “hyperlocal risk insights (100m² zones in <72 hours)”—a feature that can help authorities prioritise high-risk areas and cut recovery costs dramatically.

Heat resilience is climate justice

At the heart of this initiative is a powerful idea: that climate resilience must be people-centred. “People experience heat in different ways,” the authors remind us. A child in a tin-roof home, a street vendor working through a heatwave, or a menopausal woman without access to cooling—each person lives a different reality. Effective solutions, they argue, must reflect this diversity.

In the crowded lanes of Vivekananda Camp, where heatwaves once spelt helplessness, now there are shaded streets, restored water taps, and rooftop innovations—all thanks to a community that learnt to fight fire with data, dignity, and determination.