Meet IDEA: An AI assistant to help geoscientists explore Earth and beyond

A new artificial intelligence tool developed by researchers at the University of Hawai‘i (UH) at Mānoa is making it easier for scientists to explore complex geoscience data—from tracking sea levels on Earth to analyzing atmospheric conditions on Mars. Called the Intelligent Data Exploring Assistant (IDEA), the software framework combines the power of large language models, like those used in ChatGPT, with scientific data, tailored instructions, and computing resources. By simply providing questions in everyday language, researchers can ask IDEA to retrieve data, run analyses, generate plots, and even review its own results—opening up new possibilities for research, education, and scientific discovery. Their work was published recently in the Journal of Geophysical Research: Machine Learning and Computation.

“We built a prototype assistant that lets scientists ask plain-language questions and get back working code, clear explanations, and even publication-ready figures—in minutes,” said Matthew Widlansky, lead author of the study and associate director of the UH Sea Level Center, which is part of the Cooperative Institute for Marine and Atmospheric Research in the School of Ocean and Earth Science and Technology. “Our goal was to lower the barrier between geophysical data and the people trying to understand it."

Widlansky and Nemanja Komar, co-author on the study and the software engineer behind the project, designed the Station Explorer Assistant—or SEA, as it’s called at the UH Sea Level Center—as a prototype built on the broader IDEA framework. SEA demonstrates how the framework can be applied to global sea level observations, helping researchers and students explore coastal change through natural language interactions.

“With the Station Explorer Assistant, users don’t need to write a single line of code to analyze tide gauge data, track sea level rise, or assess flooding occurrence,” said Widlansky.

“An exciting part of this work is how easily the IDEA framework can be adapted to explore new datasets,” added Komar. “We even shifted from sea level records to dust storms on Mars—just by changing the instructions and data source.”

Still, the researchers caution that AI-generated analyses aren’t foolproof. “SEA and other IDEA-based applications can make mistakes, like miscalculating a trend,” Widlansky noted. “That’s why human oversight remains essential—we’re building tools to assist scientists, not replace them.”

Creating a tide gauge data assistant

To build the SEA tool, Widlansky and Komar connected a large language model service from OpenAI, similar to what powers ChatGPT, with access to read from the UH Sea Level Center’s data archive. They also provided the model with domain-specific instructions: essentially a virtual user manual for analyzing tide gauge data. A secure computing environment at UH then runs any code the model generates.

This setup allows the assistant to analyze coastal water level observations, assess sea level trends, and even describe results—without the user writing computer code.

“By incorporating tide gauge measurements with an interactive, expert AI assistant, we give scientists and students a new way to explore how rising seas and high‑tide flooding affect the world’s coastlines—no specialized software or coding ability required,” continued Widlansky. 

The work illustrates UH’s role in translating advanced research into practical tools for island resilience and STEM training in Hawai‘i.

Expanding applications for IDEA

While SEA focuses on sea level data, the underlying IDEA framework is designed to work across a wide range of geoscience domains. In one example from the study, the researchers applied IDEA to atmospheric data from Mars—an area they had never worked with before—and were surprised by how easily the assistant adapted to the new dataset with just a change in instructions and data source.

This flexibility is central to IDEA’s design. As an open-source, general-purpose framework, it can be customized for different research problems, from ocean forecasting to land use change, or even planetary science.

Although still a prototype, SEA is available online for scientists or university students to try out and test. Developers are encouraged to explore the IDEA framework on GitHub and experiment with adapting it to their own data or using it with other large language model services. The team welcomes feedback and collaboration to help improve IDEA and expand its scientific applications. Users of SEA and IDEA can provide feedback by emailing idea-dev-grp@hawaii.edu.

Looking ahead, the researchers plan to expand IDEA’s capabilities and user base. Future improvements include automated checks to reduce plotting errors, support for additional data sources, and a new feature that will help users build their own assistants for other geoscience challenges. As AI tools like SEA and IDEA continue to evolve, Widlansky and Komar hope they will serve as accelerators of discovery and as gateways toward making scientific exploration more accessible to students, educators, and researchers in Hawai‘i and beyond.