7/8/2026 Lois Yoksoulian
A new study led by Civil and Environmental Engineering Professor Lei Zhao develops a new, AI-based framework that produces detailed maps of air temperature across cities. In contrast with the satellite-based data typically used to estimate heat in urban areas, Zhao's model uses physical data and data-driven methods to show the heat that resident's actually experience, even capturing how the temperature can vary block by block.
Written by Lois Yoksoulian
CHAMPAIGN, Ill. — Cities are often described as “heat islands,” with media reports warning that some neighborhoods can be 20°F hotter than others. But those temperatures are often based on satellite data rather than the conditions people actually experience, due to the dearth of near-surface urban observations. This data gap hinders understanding public health risks during heat waves, planning for energy demand, infrastructure resilience and climate adaptation.
A new study aims to change that. The research, led by University of Illinois Urbana-Champaign civil and environmental engineering professor Lei Zhao, Illinois graduate student Yiwen Zhang and Pierre Gentine at Columbia University, develops the first high-resolution urban air temperature dataset that shows what the heat in cities really feels like to people and infrastructure — not just what satellites see from space.
The study is published in the journal Nature Communications.
It’s easy to assume that cities would be rich sources of environmental data, but in reality, the opposite is true, the study reports. International guidelines from the World Meteorological Organization require that standard weather stations be in open, unobstructed areas, far from buildings and other structures that could affect measurements. As a result, many urban weather stations are located at airports or nearby rural land, which do not represent conditions in dense neighborhoods where most people live and work.
“Yes, cities are hot, and some neighborhoods are hotter than others,” Zhao said. “But not always as extreme as some surface-temperature-based maps suggest.”
To address this, the research team developed a physics-informed transfer learning model — an artificial intelligence framework that blends physical understanding of the atmosphere with data-driven methods. This model can estimate near-surface air temperature in cities at very high spatial resolution, across more than 380 cities in the contiguous United States.
The Urban High-Resolution Air Temperature dataset, or U-HAT, provides detailed maps of air temperature across cities, capturing how heat varies block by block. It can be used for public health studies, urban climate research, energy planning and machine-learning applications.
“This data allows for pixel-by-pixel comparison between satellite land surface temperature and true urban air temperature,” Zhao said. “And it shows that satellite-based data often overestimates heat stress and exaggerates disparities between neighborhoods, helping explain why some past maps and media stories may have unintentionally misled the public about how extreme urban heat differences really are.”
The new framework can be applied beyond cities, in regions around the world where weather and climate observations are limited. The researchers emphasize that the method is not only about improving science in wealthy, data-rich countries. It also has major implications for global equity.
“Many regions, particularly in the Global South, are highly vulnerable to climate change and extreme heat and lack dense networks of weather stations,” Zhao said. “These are often the very places that most need reliable climate and weather information for planning, health and development decisions.”
By using physics-informed AI to “fill in” missing data, this framework can help create more accurate, decision-relevant information for communities that currently have few or no local measurements without requiring costly, large-scale deployments of physical sensors.
Zhao also is affiliated with the Institute for Sustainability, Energy, and Environment, the National Center for Supercomputing Applications, the Department of Climate, Meteorology and Atmospheric Sciences and Gies College of Business at Illinois. The National Science Foundation, NASA and the Department of Energy support this research.
Editor’s note:
To reach Lei Zhao, email leizhao@illinois.edu. The paper “Transfer learning reveals large discrepancies between air and land surface temperatures in cities“ is available online. DOI: 10.1038/s41467-026-73716-7. Civil and environmental engineering is part of The Grainger College of Engineering at Illinois.