Cities started setting record-high temperatures in May, and July 22, 2024, was the hottest day ever in the history of the planet, breaking the latest record set the day before.

“As temperatures keep increasing, it’s critical we alert those most vulnerable to extreme heat waves,” says Etienne Vos, a research manager at IBM. Vos is part of a consortium developing an early warning system for extreme heat events, one of many groups across the world creating new AI-powered tools to battle extreme heat.

Heat islands and early health alert systems

Urban planners have been aware of heat islands—urban areas that experience higher temperatures than outlying areas—for decades. However, characterizing and forecasting heat islands accurately, at scale, and at high resolution remains an unsolved problem, or a very expensive one at best. Some Earth-orbiting satellites can measure thermal radiation, which serve as ground-truth observations from which hot- and cool-spots within a city can be identified, but due to long time-gaps between satellite overpasses, these observations are quite sparse.

Now researchers like Vos, part of the Heat and Health African Transdisciplinary Center (HE²AT Center), are leveraging IBM’s Earth Observation foundation models to develop a system that can characterize surface urban heat islands and issue alerts about when and where extreme heat hotspots appear within a city. The academic and industry partners from the HE²AT Center are combining this information with health outcome data, socio-economic vulnerability information, geolocations, and other factors that contribute to heat exposure risk. Using an app, people will receive information and alerts regarding heat exposure based on their individualized risk profile.

Cool roofs

For over a decade, researchers including a group of engineers at Princeton University have been investigating “cool roofs,” lighter-colored roofs that stay cooler than darker roofs. For example, the Princeton researchers measured the temperature of a black roof when the air temperature was in the 90s. The black roof reached a scorching 130 degrees Fahrenheit (warm enough to fry an egg). A white roof in the same air temperature, however, ranged between 90-100 degrees Fahrenheit.

Using aerial imagery with AI algorithms, Google is generating high-resolution solar reflectivity measurements which in turn can help urban developers determine which areas would benefit most from cool roofs.

Cool roofs “reflect sunlight and absorb less heat. That leads to reduced indoor temperatures and improved health outcomes,” said Mansi Kansal, a senior product manager at Google. “Cool roofs have enormous potential in low-income communities that don’t have access to air conditioning,” said Kansal.

Tree power

The city of Los Angeles has made big strides in reducing the risk of extreme heat using another tool that is powered by AI: the Tree Canopy Lab. Using Google aerial imagery collected from planes as well as satellite imagery of Earth, the algorithm can pinpoint all the trees in a city and measure their density.

A specialized AI system then automatically scan images, detects the presence of trees and produces a map that shows the density of tree cover or the “tree canopy.”

Typically, getting information on tree coverage is expensive and time-intensive as people manually survey cities block-by-block and often can only count trees in public spaces.

The AI-powered tree canopy tool can help city planners identify which residential blocks would benefit most from planting trees and locate sidewalks that are vulnerable to higher temperatures given low canopy coverage.

But how do you decide whether a cool roof or more trees is the best solution for a given high heat zone? Experts say there are multiple factors to consider, such as recent research that suggests that tree canopies, for instance, have the greatest positive impact in arid climates rather than humid ones.

Digital twins

Digital twins—full digital replicas of cities and locations powered by AI—are another new frontier in tools to battle extreme heat. Dr. Iphigenia Keramitsoglou, a research director at the National Observatory in Athens, is developing a digital twin of Athens that uses machine learning to integrate diverse urban data and predict heat-related mortality. This allows for real-time and long-term heatwave management.

So, for instance, emergency responders could allocate ambulances to the areas at greatest risk so they could have a better chance of saving the lives of people most likely to suffer in extreme heat. After Athens, Keramitsoglou and her team, who have created a host of tools to fight heat called Extrema Global, plan to create digital twins of Prague and Budapest.

Beyond the technical solutions, learning how to avoid extreme heat can be approachable—even engaging. Another new tool the Extrema Global team is developing is a digital game for kids to learn about heat safety. “Based on the conditions of the moment, a child can decide which is the right clothing to wear so they can stay safe. There’s no reason staying cool can’t also be fun,” says Keramitsoglou.