Climate Change, Hurricanes, and Property Loss: A Machine Learning Approach to Studying Infrastructure Sustainability

Hurricanes have intensified and become more persistent under a changing climate, increasing the risk of infrastructure damage and property loss in coastal regions, threatening their sustainability. This study examines how hurricane intensity and persistence influence infrastructure loss, contributing to a more comprehensive understanding of climate-related risks. Using data from the National Oceanic and Atmospheric Administration (NOAA) Storm Events Database from 1996 to 2024, we develop a series of machine learning models to predict property losses based on storm characteristics and contextual vulnerability factors. Narrative-based text analysis and time-series feature engineering were applied to extract meteorological and temporal attributes, while regression and ensemble models were used for predictive evaluation. Results show that storm intensity alone explains only a small portion of loss variance, with persistence influencing damage primarily through rainfall and hydrological effects. The findings highlight that vulnerability, exposure, and cumulative risk dynamics are essential for accurate long-term prediction and for assessing infrastructure sustainability. Overall, the study demonstrates that combining machine learning techniques with climate and vulnerability data can inform future research on infrastructure sustainability. The quantified vulnerability-versus-intensity breakdown presented here can support post-disaster resource allocation, insurance risk modeling, and the prioritization of infrastructure maintenance in hurricane-prone regions.