AI Meets the Eye of the Storm - Machine Learning-Driven Insights for Hurricane Damage Risk Assessment in Florida

Aug 5, 2026

Upload by Aurelia

Florida’s exposure to hurricanes from both the Atlantic Ocean and the Gulf of Mexico makes hurricane damage risk assessment essential for reducing potential impacts. The study presents a machine-learning risk assessment focused on hurricane-induced flood damage, using granular building-level insurance data from 1985–2024 combined with remote-sensing derived variables. A stacked ensemble model predicts damage with MAE 11.3% at the ZIP Code Tabulation Area (ZCTA) level, and explainability highlights key drivers such as building property value, construction year, occupancy type, and flood zone designation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages21

Reading Time53 min

AccessFree to download

Abstract

Study goal and data

Model performance and explainability

Key predictors and spatial patterns

Future research directions

Study limitations

Introduction

Hurricane hazards and scales

Storm surge and damage mechanisms

Importance of risk assessments and existing methods

Role of machine learning

Earth Systems and Environment [https://doi.org/10.1007/s41748-025-00571-9](https://doi.org/10.1007/s41748-025-00571-9) ORIGINAL ARTICLE AI Meets the Eye of the Storm: Machine Learning-Driven Insights for Hurricane Damage Risk Assessment in Florida Sameera Maha Arachchige1 · Biswajeet Pradhan1 Received: 13 August 2024 / Revised: 18 November 2024 / Accepted: 6 January 2025 © The Author(s) 2025 Abstract Due to Florida’s exposure to hurricanes originating from both the Atlantic Ocean and the Gulf of Mexico, hurricane risk assessments serve as a critical tool for mitigating potential impacts. This is the first novel study to develop a machine learning based risk assessment for hurricane induced flood damage, which demonstrates the potential of granular building level insurance data from 1985 to 2024, enriched with remote sensing derived variables. The stacked ensemble machine learning model predicted hurricane flood damage with an MAE of 11.3% at a granular ZIP Code Tabulation Area level (ZCTA) . The model’s explainability tools determined that building property value was a significant predictor of hurricane damage, as it correlated with property size, complex architectural design, and proximity to waterfront locations, all of which affect potential repair costs. Other predictive factors include construction year, occupancy type, and flood zone designation. Partial dependency plots (PDPs) identified that northwest Florida is particularly susceptible to hurricane damage, attributed to the Gulf of Mexico’s warm and shallow waters than eastern Florida’s cooler Atlantic conditions and steep ocean floor. Miami’s significant coastal urbanisation, rendered it a hotspot despite southeast Florida’s overall low hurricane risk. Similarly Jacksonville in north-eastern Florida was identified as a hotspot due to compounded flooding from storm surge and nearby river systems. Partial dependency plots also quantified the significant positive impact of 1970s building code regulation. Future studies should examine coastal morphology, landfall angle, and proximity to barrier islands. A study limitation is that insurance data may be an imperfect representation of Florida, due to underinsurance and inability to afford insurance. Keywords Damage Model · Hurricane Vulnerability · Insured Losses · Machine Learning · Risk Assessment · Storm Surge 1 Introduction Hurricanes are one of the most destructive natural catastrophes. On average 80 hurricanes take place around the world yearly (Sobel et al. 2021) . In the Atlantic and the Eastern Pacific Ocean, they are termed hurricanes, while in the West Pacific Ocean, they are known as typhoons and are known as cyclones in regions such as Australia. Due to factors such as heavy rainfall and storm surges, hurricanes 􀀍 Biswajeet Pradhan [Biswajeet.Pradhan@uts.edu.au](Biswajeet.Pradhan@uts.edu.au) 1 Centre for Advanced Modelling and Geospatial Information Systems (CAMGIS), School of Civil and Environmental Engineering, Faculty of Engineering and IT, University of Technology Sydney, Sydney 2007, Australia are destructive by nature. Generally, populations in coastal areas are the worst affected by hurricanes, and as it moves inland the intensity of the hurricane decreases (Saxena et al. 2013) . Hurricanes have caused severe damage to infrastructure, environment and communities in coastal areas (Knutson et al. 2010) . The Saffir-Simpson hurricane scale is widely used to categorise potential hurricane damage. This scale is based only on sustained wind speed; however, there are several other factors that impact damage. For example, in 2005 Hurricane Katrina generated catastrophic damage with over $100 billion in damages and 1833 fatalities but was only a category 3 hurricane (Byrant and Akbar 2016) . Hence, it is evident that wind speed is not able to solely predict hurricane damage. Hurricane damage arises from various hazards including turbulent winds, changes in wind direction, wind uplift p

FAQs

What does the study focus on in Florida hurricane risk assessment?

It develops a machine-learning based assessment for hurricane-induced flood damage risk in Florida using insurance data enhanced by remote-sensing derived variables.

How accurate is the stacked ensemble machine learning model?

The model predicts hurricane flood damage with a mean absolute error (MAE) of 11.3% at the ZIP Code Tabulation Area (ZCTA) level.

Which factors were found to be most predictive of hurricane damage?

Explainability indicates building property value is a significant predictor, along with construction year, occupancy type, and flood zone designation; factors related to waterfront proximity and property characteristics also play roles.

Aug 5, 2026

Upload by Aurelia