India Is Using AI and Satellites to Map Urban Heat Vulnerability Down to the Building Level

WIRED

6 hours ago

Jun 23, 2025 5:00 AM Remote-sensing data and artificial intelligence are mapping the most heat-vulnerable buildings in cities like Delhi, in an effort to target relief from extreme temperatures at a granular level. Photograph:Zubaida starts her day at eight in the morning, sorting discarded plastics, glass, and chemicals with her bare hands, to collect items she can sell. With waste-segregation centers in this part of East Delhi currently shut down, she and other waste-pickers from the Seemapuri slum work outside by a dusty road through the hottest hours of the day, under the blazing sun. There is no fan or shade. With Delhi's heat wave season here, they are constantly exposed to intense high temperatures.
On June 11, the India Meteorological Department (IMD) issued a red alert for Delhi, warning of a high risk of heat illness and heat stroke. Temperatures that week pushed beyond 45 degrees Celsius (113 degrees Fahrenheit) across northern India, with the real-feel temperature, which factors in how humidity augments the effects of heat, spiking at an alarming 54 degrees Celsius. Prolonged exposure to such high heat can raise a person's body temperature to dangerous levels. It's potentially fatal.
In the face of this inferno, the IMD advised people to avoid going outdoors during the daytime, and markets and streets were deserted during the day. But even when extreme heat strikes, Delhi's large, informal workforce continues to work. Street vendors, rickshaw pullers, and waste pickers have reported dizziness, infections, breathing difficulties, and persistent fatigue during recent heat waves. 'We cannot stop,' says Zubaida. 'We earn each day to eat each day. If we miss a day, we miss a meal.'
At present, heat action plans (HAP) are India's primary approach for managing heat waves and keeping essential services running. Developed annually by state, district, and city governments—the Delhi government released its citywide HAP for 2025 in April—these plans are designed to help cities prepare for, respond to, and recover from extreme heat. And they're not working.
This has driven nonprofits and research organizations to show how HAPs can be improved—with one key idea being to use geographic information systems (GIS) that combine satellite imagery with local data to provide cities with a granular, building-by-building views of their heat. Action plans, often plagued by generalization, can then be tailored to better protect those at the highest risk.
Places like Ahmedabad and Tamil Nadu have seen falls in heat-related illnesses and deaths since introducing their plans (Ahmedabad's was the first to be implemented in India, back in 2013). But with most HAPs there are still major gaps when it comes to their implementation, coordination, and funding. A 2023 study found that 95 percent of HAPs do not have detailed mapping of heat-prone areas or vulnerable populations. Without this kind of granular data, it becomes difficult to allocate resources where they are needed most.
What is still missing is a hyperlocal targeted response to risk, says Radhika Khosla, associate professor at the Smith School of Enterprise and the Environment at the University of Oxford. She explains that heat risk is really the intersection of three things: the hazard—where temperature and humidity are highest within different city regions; exposure—which communities are most exposed to heat; and their vulnerability—factors like socioeconomic status, health conditions, levels of outdoor work, access to electricity, and the conditions of the built environment. Right now, most plans aren't collecting and analyzing this sort of data, Khosla says. 'Often where the actions are taken may not be where the most vulnerable are.'
Heat doesn't just differ between cities; it varies significantly within them. The urban heat island effect—the phenomenon of cities being hotter than surrounding rural areas—is often discussed at the city scale, but what isn't understood well enough is how differently it is felt at ground level. A slum settlement with a tin roof can be about 6 degrees Celsius hotter than a nearby independent house with a garden, even though they are both on the same street. The way heat is absorbed, trapped, and felt can shift block by block, depending on land use, construction materials, density, and the presence or absence of green cover, says architect and climate researcher Ujjvala Krishna. Because heat is granular, city-level HAPs are too broad to provide solutions that are specific to the most vulnerable.
To understand just how local heat stress can be, Krishna and her team at the nonprofit Ashoka Trust for Research in Ecology and the Environment (ATREE) conducted a study in a ward in north Bengaluru of just two square kilometers. That ward has a mix of land use and building types—high-rise apartments, low-income informal settlements, large independent houses, old industrial zones, and commercial areas. 'Because the built environment and green cover were so heterogeneous, we saw land-surface temperature differences of up to 9 degrees Celsius within the same ward,' Krishna says.
Krishna and her team at ATREE first plotted the different land uses within the Marappanapalya ward in north Bengaluru. Image: Ujjvala Krishna/ATREE
They then used satellite imagery to map surface heat across the ward, and investigated how this correlated with land use. Image: Ujjvala Krishna/ATREE
'And that is exactly how most cities are built,' Krishna says. Urban land use is shifting rapidly, and large plots are being converted to high-rise towers, which increases density, reduces airflow, and traps heat for more people in a smaller area.
Delhi's 2025 heat action plan made some progress towards addressing the data gap. It included high-resolution, ward-level land surface temperature maps that identified the city's most heat-vulnerable areas—neighborhoods that are also lacking basic infrastructure to cope with extreme temperatures. However, the quality and application of that data remain limited. 'It's a step forward, but GIS and remote-sensing analysis still need to be more advanced to influence decision-making,' says Raj Bhagat Palanichamy, senior program manager of geoanalytics at WRI India, an independent research organization that works with India's local and national governments.
Palanichamy points out that many land-surface temperature maps currently used are based on satellite datasets originally designed for agricultural monitoring, which might not be calibrated for complex multi-use urban landscapes. They may be incapable of the specificity needed to support targeted interventions in a city like Delhi.
To try to better capture that variability, SEEDS, a Delhi-based disaster-preparedness nonprofit, along with Chintan, a nonprofit working on waste management and livelihoods, have used GIS mapping of indoor heat risks across homes in low-income settlements in Delhi, located near landfills and industrial zones, and inhabited by waste pickers.
A Delhi heat map created by SEEDS and Chintan using maximum land surface temperatures for the month of May, 2013–2021. The team then used this map to identify high heat density areas to target with hyperlocal interventions. Illustration: SEEDS India and Landsat 8
They used an artificial intelligence model called Sunny Lives—built by SEEDS and Microsoft—to assess indoor heat risk across different types of buildings. 'Our model looks at how heat is experienced inside individual homes,' says Anshu Sharma, cofounder of SEEDS. 'If the outside temperature is 40 degrees Celsius, a house with a tin roof can reach 45 degrees Celsius indoors. And that's where the most vulnerable people—infants, elderly, and the unwell—are stuck during the hottest times of the day.'
SEEDS and Chintan also calculate the wet-bulb temperature, a measure that combines heat and humidity to indicate the effectiveness of evaporative cooling—or in human terms, sweating. The higher the humidity, the higher the wet-bulb temperature, and the less effective sweating is at keeping someone cool—meaning they are more likely to experience heat stress.
The team tracked conditions in homes topped with tin roofs, tiles, plastic sheeting, and concrete to figure out which are at the highest risk of extreme indoor heat. 'That trained the model to recognize patterns, which we then combined with satellite imagery to identify roof types and building materials at scale. We can now assess indoor heat exposure for individual buildings even without installing sensors for similar geographic locations.'
A Sunny Lives assessment of a dense Delhi neighborhood. In this example, the AI has highlighted all houses in the neighborhood at low-risk from heat. Photograph: Anshu Sharma/SEEDS
Sharma believes this kind of micro-level resolution should guide the next generation of heat action plans in India. 'We now have the tech to zoom in down to individual buildings and clusters, and it's available and affordable. The same temperature doesn't impact everyone the same way, and our planning has to reflect that,' he says.
'What we need are ward-level plans, because that's the smallest administrative unit where the nuances of heat can actually be addressed,' Krishna says. At that level, resources can be better allocated, responsibilities assigned to local bodies, and interventions coordinated with residents' welfare associations, slum group associations, hospitals, and schools. She also suggests building a heat vulnerability index to take stock of everyone in the area based on age, gender, and socioeconomic status and identify the most vulnerable groups.
'The more granular your data, the more targeted your response can be,' Krishna adds. Instead of generic advisories like 'drink more water,' localized data could drive specific actions: change market or factory hours, deploy cost-effective heat shelters, or set up oral rehydration solution stations in high-footfall areas. 'If you know who's most vulnerable, you can act faster,' she adds.
But as Delhi goes through another searing summer, against a backdrop of rising heat-related deaths and increasing climate stress, many of the relief measures already promised by the government—such as placing 3,000 water coolers in public places, altering timings of construction work, and building daytime shelters for outdoor workers and homeless people—are yet to be carried out.
This year, the heat feels even more unbearable for Zubaida. 'My blood pressure drops, I get splitting headaches,' she says. Frequent and prolonged power cuts in her neighborhood also mean there's little respite at home. 'We need proper shelter and some shade to work under.'
Part of the problem is that heat action plans aren't legally binding, says Tamanna Dalal, a senior research associate at the New Delhi-based Sustainable Futures Collaborative, an environmental research organization. 'Heat waves aren't recognized as state-specific disasters in most parts of India,' she explains. 'Right now, only about eight states have formally declared heat waves as disasters. This means that local authorities aren't obligated to prioritize heat action unless they get direct rapid response guidelines during the heat wave from higher levels.'
The national government also doesn't recognize heat waves as 'notified' disasters, meaning they can't trigger financial assistance under the country's disaster-management legislation.
As a result, whatever measures are taken tend to be short term and reactive. Temporary measures like school closures ordered by the education department or oral rehydration solution stockpiling orders by health departments are being repeated each year. But these measures don't do anything to build structural resilience for cities to adapt to worsening heat conditions.
It is ultimately about building more heat-resilient cities that can adapt and mitigate simultaneously, while integrating heat with other policy goals such. as energy, water, job creation, and air quality, explains Khosla.
Some of the funding pathways already exist. 'We found that 18 centrally sponsored schemes have direct linkages to long-term solutions listed in HAPs, like piped water supply, rooftop solar, et cetera,' says Dalal. But many local officials don't know these resources are available. A recent amendment to India's national disaster mitigation fund now helps states finance heat-related actions but awareness remains very low. India is still in the early stages of heat planning in the country, and large-scale capacity building is urgently needed to help implementers see heat as a chronic threat in a warming world.
'We are soon reaching the threshold of 1.5 degrees Celsius, which is going to be irreversible,' Dalal says, referring to the targeted limit for global warming set out in the Paris Agreement, which the world is almost certain to break. 'This will impact every aspect of our daily lives. The next few years are critical in implementing some of the long-term solutions, because they take years to implement and even longer to have an impact.'