Innovative risk assessment approach for storm surge losses using adaptive models and machine learning algorithms

The increasing frequency and intensity of natural hazards have caused weatherrelated losses in the United States of more than $2.7 trillion from 1980 to 2024, with coastal communities being especially vulnerable.Beyond economics is the social and longer-term economic disruption resulting in a lack of resilience for many of these communities.This study addresses the limitations of current loss functions used in storm surge assessments, which often fail to propagate uncertainties or account for diverse building types.By using dynamically adaptive models to simulate the properties of contents, non-structural, and structural elements of buildings, a robust element-wise approach is achieved that considers both the uncertainty and variation in building properties.Monte Carlo simulations are used to derive loss functions for individual buildings, along with machine learning algorithms that predict building losses more efficiently by capturing the relationships between relevant variables, resulting in a more cost-effective model.This integration of AI-driven optimization and predictive modeling leads to the development of risk maps based on the existing hazard maps and building inventory data available for coastal communities.The results help improve decision-making for stakeholders and policymakers, leading to more accurate risk maps and better-informed mitigation strategies for coastal communities.