Explainable AI for Noah-MP Land Surface Modeling
A major research project funded by NSF NCAR, focusing on implementing explainable AI/ML to improve physics-based Noah-MP land surface modeling of plant-rock-water interactions.
Project Overview
This project represents a significant advancement in geoscience modeling, combining traditional physics-based approaches with cutting-edge AI techniques. As Project Admin for NCAR/CISL Allocation Project UTAA0012, I coordinate technical leadership and implementation of innovative workflows that enhance our understanding of land surface processes.
Project Details
- Funding: NSF NCAR/CISL Allocation Project UTAA0012
- Title: “Explainable AI for Improving Physics-Based Noah-MP Land Surface Modeling of Plant–Rock–Water Interactions”
- Role: Project Admin and Research Assistant
- PI: Prof. Zong-Liang Yang (UT Austin)
- Collaborators: Dr. Daniella Rampe, Dr. Ashley Matheny
Computational Resources
The project has been awarded substantial computational resources:
- GPU Computing: 1,000 GPU hours on NSF NCAR Derecho-GPU
- CPU Computing: 20,000 CPU core-hours on NSF NCAR Derecho
- Additional Computing: 2,000 CPU core-hours on Casper
- Storage: 2 TB campaign storage for data management
Technical Approach
Physics-Based Modeling
- Noah-MP Framework: Advanced land surface model implementation
- Plant Hydraulics: Integration of plant hydraulic stress mechanisms
- Rock-Water Interactions: Modeling of subsurface hydrological processes
- Multi-scale Processes: From leaf-level to ecosystem-scale interactions
AI Integration
- Explainable AI: Implementing interpretable machine learning techniques
- Model Calibration: ML-assisted parameter optimization
- Pattern Recognition: AI-driven identification of complex process interactions
- Uncertainty Quantification: AI-enhanced error estimation and propagation
Research Innovations
Framework Development
- Architecting frameworks for precise, ready-to-use datasets
- Integration of traditional physics-based models with AI capabilities
- Development of hybrid modeling approaches for enhanced accuracy
Scientific Impact
- Advancing understanding of plant-rock-water interactions
- Improving predictions of land surface processes
- Contributing to climate model enhancement
- Supporting sustainable water resource management
Technical Skills Demonstrated
- High Performance Computing: Utilizing large-scale computational resources
- Land Surface Modeling: Advanced understanding of Noah-MP, CTSM, HRLDAS
- Machine Learning: Application of AI techniques to geoscience problems
- Data Management: Handling large-scale environmental datasets
- Project Management: Coordinating multi-institutional research efforts
Expected Outcomes
- Enhanced Noah-MP model performance through AI integration
- Improved understanding of plant hydraulic stress in land surface systems
- Development of transferable AI techniques for Earth system modeling
- Publications in high-impact geoscience and AI journals
Broader Impact
This project contributes to:
- Climate Science: Better representation of land-atmosphere interactions
- Water Resources: Improved hydrological predictions
- Agriculture: Enhanced understanding of plant-water stress
- AI in Science: Demonstrating explainable AI applications in geoscience
Current Status
Actively implementing rock and wood moisture components into the Noah-MP model while developing AI-enhanced calibration techniques. The project represents a cutting-edge intersection of artificial intelligence and Earth system science.