A LLM-Based Platform for Flood Risk Education and Weather Alerts in SIDS

Small Island Developing States (SIDS) are particularly vulnerable to the impacts of climate change, including hurricanes, flooding, and other extreme weather events. Effective flood risk education and near real-time weather alerts are critical for these regions. This paper presents a Proof of Concept (PoC) platform that leverages a Large Language Model (LLM) to process and deliver both static and real-time, country-specific information aimed at educating the public and generating targeted weather alerts. The LLM is augmented with climate, flooding, and weather data from various organizations and disaster management agencies, enabling it to provide context-specific, localized in-formation and educational resources. Unlike traditional methods, our approach integrates Retrieval-Augmented Generation (RAG) to enhance the accuracy and relevance of the generated outputs by referencing a fact-checked knowledge base before responding. Preliminary results demonstrate that the LLM, even at a smaller scale, delivers more accurate and context-specific information compared to standard models, underscoring the potential of LLM technology in addressing the unique challenges faced by SIDS. Future work will focus on scaling the model to improve performance and expanding its application to broader disaster management scenarios.