164 - Towards Democratized Flood Risk Management: An Advanced AI Assistant Enabled by GPT-4 for Enhanced Interpretability and Public Engagement

Real-time flood forecasting stands as a pivotal component in facilitating timely and efficacious emergency interventions. Nevertheless, the esoteric nature of specialized numerical models that produce these forecasts can obstruct immediate and well-informed decision-making and affect public situational awareness. This challenge is further complicated by the necessity for decision-makers to consult with domain experts in forecasting, in order to optimize an array of flood mitigation strategies. Intriguingly, the effectiveness of such systems transcends mere technical prowess, as the public's perception of risk is also modulated by an array of sociocultural and institutional variables. To address these multifaceted challenges, our study leverages the computational strengths of the GPT-4 model to develop a tailored AI Assistant focused on flood risk evaluation and preventive action. This Assistant is integrated with extant Geospatial tools, notably QGIS, resulting in a prototype that utilizes a flood-oriented large language model. This innovative system is capable of translating colloquial queries from decision-makers and the lay public into the specialized framework of numerical modeling and GIS. In doing so, it bridges the often-challenging divide between intricate technical forecasts and actionable, comprehensible information. Subsequently, germane data gleaned from the modeling outputs are transmuted back into layman’s terms, thereby fostering clarity and accessibility for non-specialist audiences. Further augmenting its capabilities, our AI system synchronizes vulnerability index mapping with pertinent flood data, thereby enabling nuanced discussions around resilience and social justice. As a future avenue, we plan to explore the system's capacity to initiate additional numerical simulations based on user queries that cannot be addressed by pre-existing/standard simulation outcomes. Our pioneering research lays the groundwork for a holistic, user-centric paradigm in flood risk management. We will dissemine our findings during the forthcoming conference.