Selected Talks
Understanding Parameter Uncertainty in CLM5 with Machine Learning for Enhanced Hydrological Predictions
Event: The Scientific comPUting and Data Science (SPUDS) Seminar
Location: Boise, ID, USA
Date: February 18, 2025
- Evaluated soil moisture variability across CONUS using EOF analysis
- Clustered moisture patterns with EOF–SOM to refine parameter calibration
- Quantified epistemic and aleatoric uncertainties in LAI via EDNNs
- Extended EDNN framework to predict hydrological variables in ungauged basins
Emulating and Evaluating Uncertainty in CLM5 Simulations Using Evidential Deep Neural Networks
Event: American Geophysical Union 2024 Annual Meeting
Location: Washington, D.C., USA
Date: December 2024
- Presented a novel approach utilizing EDNN to deliver enhanced uncertainty estimates in CLM5 simulations and reduce computational costs.
Assessing Soil Hydraulic Parameter Impacts and Performing Uncertainty Quantification in Climate Model Simulations Using Advanced Analytical Techniques
Event: The Scientific comPUting and Data Science (SPUDS) Seminar
Location: Boise, ID, USA
Date: October 22, 2024
- Identified dominant soil moisture variability modes with EOF analysis
- Clustered EOF modes by soil texture and climate via SOM
- Captured aleatoric & epistemic uncertainties in perturbed ensembles using EDNNs
- Tested transferability of methods in CAMELS basins
History Matching with Evidential Deep Neural Networks to Improve Terrestrial Water Storage Simulations in the Community Land Model
Event: HydroML 2024 Symposium
Location: Richland, WA, USA
Date: May 2024
- Introduced an EDNN-based history matching method that provides richer uncertainty estimates in terrestrial water storage simulations, improving confidence in model outputs.
Emulating and Evaluating Uncertainty Estimates in CLM5 Terrestrial Water Storage Simulations Using Evidential Deep Neural Networks
Event: 29th Annual CESM Workshop
Location: Boulder, CO, USA
Date: June 2024
- Highlighted how pinpointing high-uncertainty regions guides better decision-making and model refinements in climate simulations.
Combining Empirical Orthogonal Functions and Self-Organizing Maps to Understand Soil Parameter Influence in CLM5
Event: College of Engineering Fall Design Showcase
Location: Boise, ID, USA
Date: November 2023
- Applied EOF and SOM analysis to explore relationships between soil properties and moisture dynamics in CLM5.
Assessing the Effect of Soil Parameter Setups on Soil Moisture in CLM5 over the Continental U.S. Using EOF Analysis
Event: Boise State University Graduate Showcase
Location: Boise, ID, USA
Date: April 2023
- Demonstrated how different soil parameter configurations alter moisture patterns, informing improvements to model calibration and prediction accuracy.
Integrating Empirical Orthogonal Functions into Machine Learning Workflows
Event: Research Computing Days
Location: Boise, ID, USA
Date: April 2023
- Showcased EOF integration for enhanced climate-data feature extraction and interpretability.
- Organized a coding competition to foster collaboration and innovation among participants.