Oral C2 AI, Automation and Agent Platform

2:30 PM - 2:45 PM JST(5:30 AM - 5:45 AM UTC)

[GdsyDt6e-03] ROMSA: An Agentic AI Framework for Regional Ocean Model Setup

*Xing Zhou (1. Fujian Ocean Innovation Center (China))

Mechanistic coastal and regional ocean models remain essential for understanding and predicting ocean processes. However, setting up such models is still a highly specialized task that requires substantial technical expertise and practical experience. Recent advances in AI agents provide an opportunity to encode accumulated model setup knowledge and experience into agent skills, enabling a pathway toward automation of model configuration and setup.

Here we introduce ROMSA (Regional Ocean Modeling Setup Agent), an agentic framework designed to translate scientists' intent in natural language into a structured regional ocean model setup. ROMSA aims to lower the technical barrier to regional ocean modeling while retaining scientific oversight throughout the workflow. ROMSA adopts a multi-agent architecture. A questionnaire agent first interacts with the user to identify the scientific objectives and required model configuration, and then multiple model-building agents use the structured questionnaire output together with curated modeling knowledge, preprocessing tools, and established procedures to construct the numerical experiment.