Vipin Kumar’s Knowledge-Guided Machine Learning Framework Improves Operational Flood Forecasting

Department of Computer Science & Engineering Regents Professor Vipin Kumar co-led two new studies that examine how machine learning can improve the prediction of floods. The interdisciplinary team includes researchers from the University of Minnesota’s College of Science and Engineering and the College of Food, Agricultural, and Natural Resource Sciences, as well as Pennsylvania State University. The studies, published in Water Resources Research and the Proceedings of the IEEE International Conference on Data Mining, demonstrate how "knowledge-guided" artificial intelligence can assist forecasters in saving lives and protecting infrastructure as the frequency of extreme weather increases. 

Currently, flood forecasters at the National Weather Service use physics-based models that require manual, real-time adjustments based on field observations. This can be an extremely labor-intensive process and is difficult to scale during widespread flood emergencies. The research team developed a new model that combines traditional scientific models with new machine learning techniques to automatically analyze the state of a river’s watershed from the observed data. This hybrid approach can predict streamflow and flood levels with greater accuracy than current methods used across the United States. Additionally, this closely integrated research in hydrology and AI has resulted in innovations in both fields. 

"The knowledge-guided approach allows the model to learn from real-world data while still respecting the fundamental laws of hydrology,” said Kumar, a senior author on the papers. “This is not just about improving statistical accuracy. It is about providing reliable, actionable forecasts that emergency managers and forecasters can trust when making high-stakes decisions.” 

Ongoing research is focused on improving the method and applying the model in practice to support conservation and emergency management in Minnesota. This work is being carried out in a collaboration that includes Kumar’s lab in the College of Science and Engineering and the University of Minnesota Climate Adaptation Partnership. 

Kumar pioneered the integration of machine learning and data science into climate and Earth system science, helping establish data-driven approaches as a central pillar of modern environmental research. To overcome the limitations of conventional machine learning in complex environmental systems, he developed Knowledge-Guided Machine Learning, a framework that embeds scientific principles into modern AI to advance scientific discovery. This flood forecasting project showcases how knowledge-guided artificial intelligence can assist forecasters in saving lives and protecting infrastructure as the frequency of extreme weather increases. 

This research was supported by the National Science Foundation, State of Minnesota Weather Ready Extension, and Minnesota Pollution Control Agency; and done in collaboration with the University of Minnesota Data Science Initiative and AI-LEAF (National AI Research Institute for Land, Economy, Agriculture & Forestry). 

Learn more about the project in the article published by the College of Science and Engineering. Read the entire papers on the Institute of Electrical and Electronics Engineers website, with the hydrologic companion piece at Water Resources Research and the corresponding story on the College of Food, Agricultural, and Natural Resource Sciences website.