Empirical Hurricane Vulnerability and Fragility Curves

This project develops a reproducible and transparent workflow for analyzing the vulnerability and fragility of buildings subjected to hurricane winds. Using a suite of interconnected Jupyter notebooks, the workflow transforms raw hazard and exposure datasets into processed information, computes building-level or group-level damage ratios, and generates empirical vulnerability matrices. These matrices are then fitted into continuous vulnerability and fragility curves, which quantify how expected damage ratios and probabilities of exceeding damage thresholds evolve with wind speed. The notebooks also provide tools for comparing curves across different building classes, construction types,etc, enabling both validation and deeper insights into resilience.

The data produced in this project can be reused in multiple ways. The outputs (CSV tables and PNG figures) are readily applicable to catastrophe modeling for insurance and reinsurance, where vulnerability and fragility curves form a critical input for risk assessment. They can also support academic research by serving as benchmarks for methodological comparisons, or by being integrated into multi-hazard frameworks. In addition, policy makers and code developers can leverage these datasets to evaluate how construction practices influence structural performance, guiding the design of more resilient standards. Because the entire pipeline is coded in Python and documented step by step, users can easily adapt the notebooks to new datasets, hazard events, or structural inventories, ensuring long-term reusability.

The uniqueness of this project lies in its transparency, flexibility, and modularity. Traditional catastrophe models often operate as black boxes, limiting validation and adaptability. In contrast, this framework provides open-source, step-by-step tools that combine empirical data processing, statistical rigor (outlier detection, logistic regression, performance metrics), and visualization in a single reproducible environment. Users may execute the full sequence of notebooks for a complete study, or selectively run modules to focus on specific tasks such as curve fitting or comparative analysis. This adaptability makes the workflow valuable for both detailed case studies and broad comparative research.

The primary audience for this project includes researchers in wind engineering, structural safety, and risk analysis who require transparent methods for vulnerability modeling. It is also intended for insurance and reinsurance professionals seeking empirically derived curves for underwriting, pricing, and capital allocation. Policy makers and engineers can use the results to support evidence-based decision making and improvements in building codes. Finally, educators and students can adopt the notebooks as teaching material for courses on catastrophe modeling, vulnerability analysis, and disaster risk reduction.