Real-time flood forecasting plays a crucial role in enabling timely and effective emergency responses. However, a significant challenge lies in bridging the gap between complex numerical flood models and practical decision-making. Decision-makers often rely on experts to interpret these models for optimizing flood mitigation strategies. And the public requires complex techniques to inquiry and understand socio-cultural and institutional factors, often hinders the public's understanding of flood risks. To overcome these challenges, our study introduces an innovative solution: a customized AI Assistant powered by the GPT-4 Large Language Model. This AI Assistant is designed to facilitate effective communication between decision-makers, the general public, and flood forecasters, without the requirement of specialized knowledge. The new framework utilizes GPT-4's advanced natural language understanding and function calling capabilities to provide immediate flood alerts and respond to various flood-related inquiries. Our developed prototype integrates real-time flood warnings with flood maps and social vulnerability data. It also effectively translates complex flood zone information into actionable risk management advice. To assess its performance, we evaluated the prototype using six criteria within three main categories: relevance, error resilience, and understanding of context. Our research marks a significant step towards a more accessible and user-friendly approach in flood risk management. This study highlights the potential of advanced AI tools like GPT-4 in democratizing information and enhancing public engagement in critical social and environmental issues.

翻译：实时洪水预报在实现及时有效的应急响应中发挥着关键作用。然而，复杂数值洪水模型与实际决策之间的鸿沟仍是一大挑战。决策者常需依赖专家解读这些模型以优化防洪策略，而公众则需要复杂技术手段来探究和理解社会文化及制度因素，这往往阻碍了公众对洪水风险的认知。为克服这些难题，本研究提出了一项创新解决方案：基于GPT-4大语言模型定制开发的AI助手。该AI助手旨在无需专业知识的前提下，促进决策者、公众与洪水预报人员之间的有效沟通。新框架利用GPT-4先进的自然语言理解与函数调用能力，提供即时洪水预警并回应各类洪水相关问询。我们开发的原型系统整合了实时洪水预警、洪水地图与社会脆弱性数据，并能将复杂的洪泛区信息有效转化为可操作的风险管理建议。为评估其性能，我们从相关性、错误恢复能力和上下文理解三大类共六项指标对原型进行了测试。本研究标志着洪水风险管理向更易获取、更用户友好的方向迈出了重要一步，同时揭示了GPT-4等先进AI工具在促进信息普及和提升公众对关键社会与环境问题参与度方面的巨大潜力。