Teaching AI the Physics of Soil

For the past two decades, we have treated artificial intelligence like a child left alone to learn everything from scratch through pure observation (i.e., data). This only works if the child has millions of experiences to learn from, but soil science is different. Taking soil samples is expensive and time-consuming; we simply do not have “millions” of observations, and this will likely remain the case for the foreseeable future.

In his PhD research, Sarem Norouzi tried to introduce a “parent” to this learning process: soil physics. In the same way parents teach a child different rules, he “teaches” AI the fundamental laws of soil physics, such as how water is held and how it moves in soil as a result of various interacting factors, such as retention curves and particle size distribution. These physics-based “parents” allow AI to learn much faster and with far less data. Most importantly, they prevent AI from making “childish” mistakes, such as predicting outcomes that are physically impossible in the real world. The result is a new way to understand and map soil that is both reliable and cost-effective. These tools can support better management of agriculture and help protect the environment in a future where soil data remain scarce.

The PhD study was completed at Soil Physics and Hydropedology Section, Department of Agroecology, Faculty of Technical Sciences, Aarhus University.

This summary was prepared by the PhD student.

Time: Monday, 10 August 2026 at 13:00
Place: Auditorium, AU Viborg, Aarhus University, Blichers Alle 20, 8830, Tjele
Title of PhD thesis: Scientific Machine Learning for Soil Hydrology: From Enhanced Prediction to Process Discovery
Contact information: Sarem Norouzi, e-mail: sarem.nrz@agro.au.dk, tel.: +45 50 17 61 95
Members of the assessment committee:
Distinguished Professor Ty P.A. Ferré, Department of Hydrology and Atmospheric Sciences, The University of Arizona, USA
Professor Karsten Høgh Jensen, Department of Geosciences and Natural Resource Management, University of Copenhagen, Denmark
Professor Diego Abalos (chair), Department of Agroecology, Aarhus University, Denmark
Main supervisor: Professor Lis Wollesen de Jonge, Department of Agroecology, Aarhus University, Denmark
Co-supervisor: Professor Mogens Humlekrog Greve, Department of Agroecology, Aarhus University, 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.