Advances in natural-language processing and large language models (LLMs), such as those utilized by ChatGPT, are transforming how geoscientists interact with complex datasets, enabling efficient and intuitive scientific analyses. As part of the Tracking and Communicating on Sea Level Pacific RISA project, PI Widlansky and the University of Hawaiʻi Sea Level Center (UHSLC) are developing tools that use LLMs to allow researchers to ask questions in everyday language and receive clear explanations and data analyses in response, minimizing the need for time-consuming tech support for project managers. One such tool, called the Station Explorer Assistant (SEA), draws on the UHSLC’s extensive databases and uses AI to analyze sea level data, compare water levels to normal conditions, and predict potential flooding. It even writes and runs its own analysis software, which it shows the user to check that its results are accurate. By making sea level science easier to understand and access, SEA can support communities adapting to rising seas and other coastal challenges. 

SEA technology is also generalizable across geoscience domains, through a framework called an Intelligent Data Exploring Assistant (IDEA), which can be demonstrated by asking it to analyze atmospheric observations from Mars collected by NASA’s InSight Mission (Try it!). By combining LLM capabilities with robust domain-specific customizations, SEA and the IDEA example generate accurate analyses, visualizations, and insights through natural-language prompts. This study highlights the potential of IDEA frameworks to lower technical barriers, enhance educational opportunities, and transform geoscientific workflows while addressing the limitations and uncertainties of current LLM technology. PI Widlansky’s work also highlights how AI can enhance scientific research and communication, and helps us to envision how the creation of similar tools can support scientists in many fields.

SEAinfo page, with several YouTube video demonstrations and presentations by PI Widlansky
https://uhslc.soest.hawaii.edu/research/SEAinfo/
GitHub IDEA page
https://github.com/uhsealevelcenter/IDEA
IDEA manuscript with a plain language summary and abstract
https://uhslc.soest.hawaii.edu/research/SEAinfo/IDEA_manuscript_latest.pdf