Research
Research in the Abeshu Hydrosystems Intelligence Lab centers on connected directions that use physics-informed AI, multi-sensor Earth observation, and decision analytics to improve water resilience from local infrastructure to global Earth-system models.
Research Pillars
Three pillars, built for interpretation, transfer, and use
The lab develops learning-based methods that respect physical structure, combine multiple sources of evidence, and produce decision-relevant outputs. The work is organized around three pillars, from understanding how water moves to supporting the people and systems that manage it.
Pillar 1
Process-Aware Hydrologic Prediction
Predicting how water moves and is stored across watersheds, aquifers, rivers, and lakes, with learning-based methods that respect physical structure.
Physics-Informed AI for Connected Water Extremes
Modeling how droughts, floods, heat, and water-storage anomalies emerge, interact, and propagate across watersheds, aquifers, reservoirs, and communities.
- Graph learning for cascading hydroclimate extremes across connected water systems.
- Multi-sensor fusion using satellite, in-situ, hydrometric, and reanalysis data.
- Risk mapping that links hazards with infrastructure exposure and community vulnerability.
Applications: Probabilistic forecasts and hotspot analyses to support preparedness in systems such as the Rio Grande, Elephant Butte, Caballo, and irrigated landscapes.
Learning-Enhanced Earth System and Hydrologic Modeling
Building physics-guided machine learning components that improve regional and global hydrologic models under changing climate, land, and infrastructure conditions.
- Differentiable soil moisture, groundwater, surface water, and storage representations.
- Model-data fusion with GRACE, SMAP, SWOT, streamflow records, and global networks.
- Integration with modeling systems such as E3SM, MOSART, and Xanthos.
Applications: Faster, more reliable model components that preserve physical constraints while improving prediction under non-stationary conditions.
Lake Dynamics, Remote Sensing, and Earth-System Integration
Using multi-sensor observations and physics-guided learning to monitor and predict lake storage, extent, temperature, ice, and mixing.
- Optical, radar, altimetry, and thermal sensing for global and regional lake dynamics.
- Data assimilation methods that honor heat, water, and mass-balance constraints.
- Improved lake boundary conditions and parameterizations for hydrologic and Earth-system models.
Applications: Scalable datasets and forecasts for lake-dependent water availability, arid-land planning, and climate feedback assessment.
Pillar 2
Water Infrastructure and Operational Intelligence
Turning prediction into operating decisions for the reservoirs, dams, and supply systems that communities depend on.
Adaptive Operations for Resilient Water Infrastructure
Developing AI-enabled decision tools for reservoirs, hydropower, flood control, water supply, and environmental flows under deep uncertainty.
- Hybrid forecasting that blends hydraulic routing, hydrologic prediction, and regime-shift detection.
- Multi-objective optimization for flood protection, supply reliability, equity, and ecosystem needs.
- Operator-centered decision support for drought-to-flood transitions and other rapid changes.
Applications: Open-source tools, scenarios, and training resources that support adaptive risk management with agencies, utilities, and stakeholder groups.
Pillar 3
Human–Water Resilience and Adaptation
Studying how people and water systems shape each other, and what makes adaptation effective and equitable.
Equitable Water Security and Sociohydrologic Intelligence
Combining AI, Earth observation, hydrologic modeling, and socioeconomic data to identify water access gaps and emerging risks.
- Fine-scale mapping of water insecurity, quality stressors, and service gaps.
- Scenario modeling for climate, development, governance, and public health outcomes.
- Transparent equity analytics for rural, underserved, and colonia communities.
Applications: Decision-ready benchmarks that help agencies and partners prioritize water infrastructure, policy intervention, and community resilience investments.
Cross-Cutting Methods
Methods used across all three pillars
The same methodological core supports every pillar, which is what lets results transfer between them.
Illustrations on this page are AI-generated diagrams used to introduce each pillar. They are illustrative only and do not present research results or data.