Date of Award

12-2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Civil Engineering

Committee Chair/Advisor

Pamela Murray-Tuite

Committee Member

Thomas Sharkey

Committee Member

Seth Guikema

Committee Member

Mashrur Chowdhury

Abstract

This dissertation advances a household-centered understanding of post-disaster recovery by linking evacuation behavior, restoration preferences, and optimization-based recovery planning. First, it extends the empirical evacuation literature by modeling both pre- and post-impact evacuation decisions using survey data collected after Hurricane Ida (2021). A sequential logit model shows that official evacuation warnings influence both stages of evacuation, but the determinants diverge thereafter: socio-economic and housing characteristics shape pre-impact evacuation, whereas post-impact evacuation is driven primarily by situational conditions—especially residential damage and prolonged utility outages. Durations of power and water outages exhibit additive effects, significantly increasing the likelihood of post-impact departure. These findings challenge the implicit planning assumption that residents who shelter in place will remain and suggest that prolonged utility loss can induce a second wave of evacuation not typically captured in conventional evacuation or recovery plans. The second study examines short- to mid-term restoration preferences through a stated-preference survey across four large U.S. metropolitan areas (Houston, New York City, Washington, D.C., and Miami). Using rank-ordered logit models, including a latent class specification, the analysis reveals meaningful heterogeneity in recovery priorities. Three preference classes emerge: younger, flexible workers emphasize employment and financial services; older working-age households without children prioritize health and food; and a more generalized group demonstrates diffuse preferences across services. These results illustrate that restoration priorities vary systematically by demographic, socio-economic, and work-related conditions, reinforcing that “one-size-fits-all” recovery strategies may fail to meet resident-defined needs. The final study integrates these insights within an interdependency-aware mixed-integer linear programming model embedded in the CLARC virtual community testbed. The model formalizes household satisfaction as a function of restored utilities and services weighted by resident preferences, and compares three restoration strategies: a traditional population-based priority (PBP), a household-centric priority (HCP), and a phased restoration strategy with household departure (PRHD). Under a Category 3 hurricane, all strategies converge to near-complete satisfaction. Under a more severe Category 5 event, however, the household-driven strategies outperform PBP (end-of-horizon satisfaction: HCP ≈ 76.3%, PRHD ≈ 72.8%, vs. PBP ≈ 71.5%), particularly for recovery of food, employment, education, home repair, and banking services. PRHD additionally reduces post-impact evacuation by 27–37% by prioritizing restoration for households who remain in place. Taken together, these three studies demonstrate that post-impact evacuation is a behavioral adaptation linked to infrastructure failure; that residents’ recovery priorities are heterogeneous; and that embedding these behavioral and preference insights into restoration models can meaningfully improve recovery satisfactions under severe disruption. This dissertation advances a resident-centered framework for disaster recovery that integrates household behavior, diverse recovery preferences, and infrastructure interdependencies, offering a more considerate and operationally relevant alternative to traditional population-based restoration planning.

Author ORCID Identifier

https://orcid.org/0000-0003-4062-2130

Download

Available for download on Thursday, December 31, 2026

Share

COinS