Climate modelling with Generative AI

From low-resolution global models to high-resolution local impacts: Global climate models describe large-scale atmospheric patterns and processes, but many climate impacts depend on local conditions. This project investigates how AI-based climate downscaling can help bridge that gap.

26 June 2026 by Kristoffer Kristensen

During her PhD studies, Thea Quistgaard researched the use of generative artificial intelligence (AI) in climate modelling. Modern climate models often operate at spatial scales too coarse to capture local rainfall patterns and extreme weather events in detail. Generative AI has recently attracted attention for its ability to create realistic, high-resolution climate data while also representing uncertainty and variability.

In her research, Thea Quistgaard developed and tested AI-based models to generate detailed local climate information over Denmark from coarse global climate data. The work examined how physical information influences model behaviour, how such systems should be evaluated, and under which conditions AI-generated climate data can be considered scientifically trustworthy.

The research highlights both the potential and the limitations of generative AI in climate science and argues that good statistical performance alone is not enough for responsible scientific use.

The PhD study was completed at Department of Environmental Science, Faculty of Technical Sciences, Aarhus University.

This summary was prepared by the PhD student.

Time: Friday, 26 June 2026 at 13:00
Place: Building 112, H. H. Koch Auditorium, Department of Environmental Science, Aarhus University, Frederiksborgvej 399, 4000 Roskilde
Title of PhD thesis: Generative Diffusion Models for Climate Downscaling: Building scientifically valid tools
Contact information: Thea Quistgaard, e-mail: tquistgaard@envs.au.dk, tel.: +45 22631634
Members of the assessment committee:
Professor Martin Drews, DTU MANAGEMENT, Technical University of Denmark (DTU), Denmark
Senior Researcher Ramón Fuentes Franco, Rossby Centre, Swedish Meteorological and Hydrological Institute (SMHI), Sweden
Senior Researcher Niels Bohse Hendriksen (chair), Department of Environmental Science, Aarhus University, Denmark
Main supervisor: Professor Peter L. Langen, Department of Environmental Science, Aarhus University, Denmark
Co-supervisor: Professor Simon Stisen, Department of Hydrology, Geological Survey of Denmark and Greenland, Denmark
Language: The PhD dissertation will be defended in English

The defence is public.
The PhD thesis is available for reading at the Graduate School of Technical Sciences/GSTS, Ny Munkegade 120, building 1521, 8000 Aarhus C.

Climate modelling with Generative AI - promises and pitfalls