Deep learning models have shown promise in environmental and Earth system modeling applications, with many examples in weather forecasting and flood prediction that sometimes outperform traditional physics-based models across several dimensions. However, they remain black boxes, with parameters and processing that are not easily interpretable. Key challenges—such as their ability to extrapolate to data not seen during training, transfer to new locations, demonstrate robustness, and characterize predictive uncertainty—remain active areas of research.
In this presentation, I will focus on efforts to tackle these challenges by aligning deep learning models with physical principles for river flow and flood prediction, aiming to develop more robust and transferable models.
Bio
Caleb Buahin conducts research on advancing integrated
hydrologic and hydrodynamic modeling frameworks for
water systems ranging from large river reservoir/lake
systems to urban collection systems. He also focuses on
the intersection between these models and emerging
artificial intelligence/machine learning (AI/ML) to
improve the degree of fidelity and accuracy of processes
represented and for optimizing the management of
water systems.
Caleb was Lead Hydroinformatics Engineer for Xylem
Inc.’s Wastewater Network Optimization team, where he
employed hybrid traditional physics/process-based
models and artificial intelligence/machine to support
Long Term Control Plan (LTCP) development for various
municipalities and for real time control and decision
support systems for collection systems. Before Xylem
Inc., he worked for Environmental Resources
Management Inc. as a Project Engineer, where he developed and applied hydrologic models and various
1D, 2D, and 3D hydrodynamic models for environmental impact assessment studies for sectors including
the oil and gas, power, and mining industries looking at the transport and fate of various pollutants (e.g.,
nutrients, oil spills, produced water, sediment, thermal discharges, etc.) in water bodies.
Caleb obtained his BS and MS degrees in Civil Engineering from Brigham Young University, Provo Utah in
2010 and his Ph.D. in Civil Engineering Utah State University, Logan Utah in 2017.
