AI model enables pinpoint visualisation of sites of future heat stress in cities
Freiburg, 20/08/2025
A new AI model from the University of Freiburg and the Karlsruhe Institute of Technology (KIT) is the first to calculate how heat stress will develop metre-by-metre in a city. The model can help cities adapt to climate change and provides a stimulus for climate-conscious urban planning.
An interdisciplinary research team from the University of Freiburg and the Karlsruhe Institute of Technology has developed an AI model that can calculate how heat stress will develop in a city. This is the first model not only to offer high resolution with metre-by-metre analysis, but also to cover long periods of time. The researchers took Freiburg as the pilot study for their model. The AI system combines geodata such as building heights and vegetation structures with weather forecasting or climate projection data such as air temperature and radiation. The model can be used to predict a variety of climate scenarios, ranging from a climate that only heats up slightly thanks to significant climate protection activities, through to a far hotter climate as a result of very high greenhouse gas emissions.
“Our AI model allows us to analyse heat development literally everywhere in Freiburg.”
Prof. Dr. Andreas Christen
Environmental Meteorologist, University of Freiburg
“Our AI model allows us to analyse heat development literally everywhere in Freiburg,” says Professor Dr. Andreas Christen, environmental meteorologist at the University of Freiburg. “Since every city has a structure of its own, with its building density, greenery and location, it’s critical to calculate heat stress as precisely as possible – we can only develop appropriate measures that protect people from extreme heat if we do so.” Following validation and adaptation to specific urban characteristics the model can be tailored to any other city and applied there.
Heat stress will increase in Freiburg
Using the AI model, the future Freiburg city climate for the period 2070 to 2099 was simulated according to three scenarios. The most pessimistic scenario involved up to 307 hours of severe heat stress with a perceived daytime temperature of over 32 °C per year. In the reference period from 1990 to 2019 there were 135 such hours a year. The number of hours of very severe heat stress of over 38 °C perceived temperature could even increase ten-fold: to 71 hours a year in the period 2070 to 2099, as against seven hours annually in the reference period.
Up to 307 hours
possible with extreme heat exposure
By comparison, in the scenario with the lowest heat development, the hours of severe heat stress rise to just 149 per year. The number of hours of very severe heat stress were limited in this scenario to twelve.
The effect of heat stress varies within a city. “Factors such as building density, vegetation and air circulation determine whether an area remains comparatively cool or extreme heat builds up,” explains Dr. Ferdinand Briegel, lead author of the study and research associate in the Regional Climate and Weather Hazards work group at KIT’s Institute for Meteorology and Climate Research.
“Factors such as building density, vegetation and air circulation determine whether an area remains comparatively cool or extreme heat builds up.”
Dr. Ferdinand Briegel
Research Associate in the Regional Climate and Weather Hazards work group at KIT’s Institute for Meteorology and Climate Research
The study measured the heat stress using representative districts of Freiburg in the form of an industrial area, a residential area with mature trees, and the historical inner city with medium-height buildings and less vegetation. The results show that the hotter hours of the day rise rapidly in industrial areas especially, as they are mostly made up of hard surfaces with little shade. “Densely developed areas with mature trees can ensure a smaller rise in hotter hours during the day, as they offer shade. At night, however, this building and tree structure slows cooling and retains the heat for longer,” says Briegel.
Further Information
- Original Publication: Briegel, F., Schrodi, S., Sulzer, M., Brox, T., Pinto, J.G. & Christen, A. (2025). Deep learning enables city-wide climate projections of street-level heat stress. Urban Climate. DOI: https://doi.org/10.1016/j.uclim.2025.102564
- The researchers’ interdisciplinary work involved the Chair of Environmental Meteorology and the Department of Computer Science at the University of Freiburg and KIT’s Institute of Meteorology and Climate Research (IMKTRO) with its Regional Climate and Weather Hazards work group.
- The Helmholtz Association, the largest scientific organisation with a focus on the natural sciences, technology and biology/medicine, organises its research through programme-oriented funding. In preparation for the next programme-oriented funding period, PoF V, the Helmholtz Association has introduced the central topic of urban research. Shared approaches like this between the University of Freiburg and KIT show how important networked research is for major cutting-edge issues.