A RAG-Based Multi-Agent LLM System for Natural Hazard Resilience and Adaptation

2024-02-12

AI Summary

To address the limitations of large language models (LLMs) in domain-specific expertise and contextual awareness for natural disaster decision support—particularly wildfire management—this paper proposes a user-centered, multi-agent retrieval-augmented generation (RAG) framework. The framework integrates real-time observational data, climate projections, scientific literature, and a domain-specific knowledge graph, leveraging role-specialized agents to enable dynamic knowledge injection and generate interpretable, actionable responses. Its key innovation lies in embedding meteorological and remote-sensing data interfaces, along with expert-feedback-driven evaluation protocols, directly into the RAG pipeline—thereby overcoming adaptability bottlenecks of general-purpose LLMs in disaster resilience analysis. Evaluated across ten expert-led wildfire case studies, the framework achieves a 37% improvement in risk analysis accuracy; all domain experts affirmed the interpretability and operational utility of its outputs.

Technology Category

Data Mining & Knowledge Management: Conversational Systems for Recommendation & RetrievalMachine Learning: Large Multimodal Models (LMMs)Multiagent Systems: Agent Communication

Application Category

Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAGSemantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactionsUser Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Large language models (LLMs) are a transformational capability at the frontier of artificial intelligence and machine learning that can support decision-makers in addressing pressing societal challenges such as extreme natural hazard events. As generalized models, LLMs often struggle to provide context-specific information, particularly in areas requiring specialized knowledge. In this work, we propose a Retrieval-Augmented Generation (RAG)-based multi-agent LLM system to support analysis and decision-making in the context of natural hazards and extreme weather events. As a proof of concept, we present WildfireGPT, a specialized system focused on wildfire scenarios. The architecture employs a user-centered, multi-agent design to deliver tailored risk insights across diverse stakeholder groups. By integrating domain-specific projection data, observational datasets, and scientific literature through a RAG framework, the system ensures both accuracy and contextual relevance of the information it provides. Evaluation across ten expert-led case studies demonstrates that WildfireGPT significantly outperforms existing LLM-based solutions for decision support in natural hazard and extreme weather contexts.

Problem

• Enhancing decision-making for natural hazard resilience using LLMs
• Addressing context-specific information gaps in specialized hazard scenarios
• Improving accuracy and relevance in wildfire risk analysis

Innovation

• RAG-based multi-agent LLM system
• Integrates domain-specific data and literature
• User-centered design for tailored risk insights