Incorporating Process-Based Models into NeuralHydrology
Day 3 Session 3 (10:30AM)
Presenters:
Daniel McKenzie (Colorado School of Mines)
Ziyu Li (Colorado School of Mines)
Andy Wood from Colorado School of Mines
This workshop introduces differentiable modeling approaches that integrate process-based hydrologic models within the neuralhydrology deep-learning framework. Participants will explore implementation strategies using the CFE conceptual model, examine key integration challenges, and learn best practices for designing hybrid modeling workflows.
The session highlights how combining physically based hydrologic models with machine learning can improve model robustness, interpretability, and predictive performance. Emphasis is placed on practical considerations for deploying hybrid models in operational forecasting environments.
Learning Outcomes:
- Understand core concepts behind differentiable hybrid hydrologic modeling
- Implement process-based components within neuralhydrology workflows
- Evaluate the performance of hybrid process-informed ML models
- Apply best practices for deploying hybrid models in operational settings