Date of Award

8-2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Civil Engineering

Committee Chair/Advisor

Dr. Chao Fan and Dr. Pamela Murray-Tuite

Committee Member

Dr. Thomas Sharkey

Committee Member

Dr. Abdul Khan

Abstract

The core of reframing and operationalizing disaster resilience with a human-centered lens is to incorporate concepts from socio-ecological resilience into engineering resilience to better understand the humans’ capability for disaster adaptation. Existing studies have drawn practical implications by identifying actionable thresholds for infrastructure systems under disasters, which can be easily applied by policymakers, emergency managers and municipal agencies. However, how individuals interact with, respond to, or adapt under these infrastructure thresholds remain understudied. This hinders the operationalization of disaster resilience at the human scale.

First, I examined exposure by analyzing how configuration and distribution of urban infrastructure systems, such as tree cover, roads and buildings, contribute to the heat exposure. I developed a physic-informed deep learning model to quantify spillover effects within urban systems, which allows planners and policymakers to identify the most effective heat mitigation strategies across five U.S. cities.

Second, I assessed economic vulnerability of households under consecutive heavy rainfall in the hot-humid southeastern U.S., with a focus on understanding how economic hardships accumulate as the frequency of heavy-to-extreme rainfall events increases. By constructing a precipitation-hardship curve, it allows us to predict future population vulnerabilities over region. The findings of this study can inform disaster managers in developing proactive strategies, such as targeted financial support, before households reach critical tipping points and become trapped in prolonged economic hardships.

The third study investigates how household adapt under two consecutive tropical storms. We developed a novel hybrid framework that integrates random forest-derived insights into binary logit models to uncover how infrastructure disruptions interact with socioeconomic conditions of households to shape their adaptation decisions. This method is robust to “small-n-large-p” settings (such as insufficient survey responses) and can handle multicollinearity. This makes it suitable for behavioral studies using survey data with limited observations and highly correlated variables. It offers a robust and replicable methodology for future studies that aim to improve the understanding of human interactions with infrastructure systems.

Author ORCID Identifier

0009-0002-1036-0168

Download

Available for download on Monday, August 31, 2026

Share

COinS