AI-Enhanced Flood Storytelling for Mobilizing Action and Building Resilience in Underserved Communities

Objective:

This research addresses the challenge of fragmented flood-risk information by centralizing data from government agencies, weather systems, historical records, and social media to improve communication through compelling, actionable stories. The intended product is a web-based platform powered by AI that organizes semantic and quantitative data into a knowledge graph, enabling automated generation of contextualized infographics, map visualizations, and communication toolkits tailored for community leaders and socioeconomically vulnerable populations.

Approach:

To efficiently centralize fragmented flood information, data from sources such as NOAA, FEMA, USGS, ArcGIS, social media, and web-scraped content—including communication guides—is identified, collected, processed, and structured into an ontology, or "blueprint," of core concepts. This ontology creates a shared, scalable vocabulary across diverse data types. A knowledge graph is then constructed by populating this schema with real-world data. Modern artificial intelligence techniques, including retrieval-augmented generation supported by vector embeddings, large language models, and agentic systems are used to ground the automatic creation of "stories," such as contextualized infographics and map-based narratives, directly from the knowledge graph. A web interface, built on a modern technology stack, enables users to interact with these outputs and explore the underlying data and reasoning. Validation is conducted through expert interviews and iterative user feedback, ensuring that the platform's visual and textual content is not only accurate, but also locally relevant and usable for community leaders and emergency forecasters.

Impact:

Abstract:

Flooding disproportionately impacts vulnerable communities, burdening local leaders who must navigate fragmented and overwhelming flood-risk information from multiple disconnected sources and communication channels. Our research tackles this challenge by developing an innovative AI-powered web platform that centralizes and clearly communicates complex flood-risk data. The platform integrates information from government agencies such as NOAA, FEMA, and USGS, as well as historical records and social-media data. It then applies advanced artificial intelligence—including large language models, knowledge graphs, and retrieval-augmented generation—to automatically synthesize and visualize actionable insights. This work advances disaster communication by (1) creating a standardized semantic vocabulary (ontology) that links different types of flood information; (2) developing a dynamic knowledge graph that provides real-time contextual insights; (3) generating clear, story-oriented infographics and interactive maps that distill complex flood-risk data into engaging content; and (4) deploying an accessible web interface designed for community leaders and socioeconomically vulnerable populations, thereby enhancing decision-making and reducing information overload. By producing contextualized flood-risk narratives and streamlined information workflows, the platform will improve emergency preparedness, crisis response, and long-term recovery. Ultimately, the project establishes a scalable, standardized, and commercially viable foundation for disaster-response applications, strengthening community engagement and resilience while saving lives through clear, timely, and actionable communication.