The University of Hawaiʻi at Mānoa will lead a three-year, nearly $1 million project using artificial intelligence to predict how rising seas and stronger storms threaten coastal freshwater supplies and ecosystems. The National Science Foundation awarded a $500,000 grant under its Collaborations in Artificial Intelligence and Geosciences program, with UH Mānoa partnering with the University of Texas at Austin.
Associate Professor Jonghyun "Harry" Lee of the UH Mānoa Water Resources Research Center and Department of Civil, Environmental and Construction Engineering serves as principal investigator. The project addresses a growing problem: saltwater contamination of drinking water aquifers and degradation of coastal ecosystems affect millions of people, along with agriculture, infrastructure, and local economies.
Faster models for groundwater-ocean interactions
The research team will build AI models that simulate how water moves between coastal aquifers and the ocean. By combining recent advances in AI with traditional environmental modeling, they aim to predict seawater intrusion into freshwater aquifers and freshwater discharge from land into the ocean with greater speed and accuracy than current methods allow.
"This project will improve our understanding and prediction of how groundwater and the ocean interact in Hawaiʻi's coastal aquifers, helping communities better protect freshwater resources and coastal ecosystems from challenges, such as seawater intrusion and coastal inundation," Lee said.
The approach creates high-speed "surrogate models" that run complex simulations at a fraction of the cost while preserving essential physics laws at the land-sea boundary. This matters for researchers working in AI for Science & Research, where computational efficiency often determines whether a model can be deployed for real-world decisions.
Project scope and deliverables
The three-year effort includes several defined goals:
- Flexible AI architecture: Combining different AI tools to analyze long-term ocean patterns and short-term events such as storm surges.
- Real-world testing: Validating models on benchmark applications in Hawaiʻi and Texas coastal systems.
- Near-real-time forecasts: Enabling rapid assessments of seawater intrusion and coastal groundwater discharge to support digital twin models of complex ecosystems.
- Practical decision tools: Providing actionable insights for water management, infrastructure planning, and long-term coastal resilience.
- Open science and education: Developing open-source software, interactive visualization tools, and interdisciplinary training for students and early-career researchers.
Lee said students and postdocs on the project will receive hands-on training at the intersection of groundwater hydrology, coastal science, computational modeling, and artificial intelligence. The open-source component means researchers outside the two partner institutions can access and build on the tools. Those looking to develop relevant skills can explore AI Research Courses that cover modeling and analysis techniques used in scientific contexts.
Why this matters for science and research professionals
The project signals a shift in how geoscience research gets funded and conducted. NSF's Collaborations in Artificial Intelligence and Geosciences program is explicitly designed to pair domain scientists with AI specialists, and the grant structure - $500,000 to the lead institution within a larger collaborative award - reflects the expectation that no single lab holds all the necessary expertise. For researchers in adjacent fields, the surrogate modeling approach offers a template: physics-informed AI models that preserve accuracy while cutting computational costs, tested against real coastal systems rather than synthetic benchmarks.
Your membership also unlocks:
