Date of Award
8-2026
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Civil Engineering
Committee Chair/Advisor
Dr. Weichiang Pang
Committee Member
Dr. Yongjia Song
Committee Member
Dr. Chao Fan
Abstract
This thesis develops an integrated tract-level framework for estimating post-hurricane displacement, short-term shelter demand, long-term shelter demand, and shelter allocation under hurricane wind hazards. The framework combines parametric wind-field modeling based on HURDAT2 and EBTRK hurricane data with HAZUS-based fragility and loss functions, uninhabitability relationships, demographic factors, and an insurance-sensitive recovery model. Tract-level insurance adequacy is estimated using American Community Survey data, National Association of Insurance Commissioners premium data, and the National Risk Index, using five scenario-based models; the combined home-value and hazard-risk model is adopted for final analyses. To evaluate post-disaster housing capacity, the study estimates the number of habitable vacant hotel rooms and then applies both a greedy heuristic and a Mixed-Integer Linear Programming (MILP) model for shelter allocation.
The framework is validated using Hurricanes Laura (2020) and Ida (2021) in Louisiana. Results show strong agreement with HAZUS for peak gust wind speed, displaced households, and short-term shelter needs, and reasonable agreement with FEMA Individual Assistance–based indicators for long-term shelter need. The model achieved wind-field correlations of 0.971 for Laura and 0.939 for Ida, while tract-level correlations for displaced households and short-term shelter need ranged from 0.89 to 0.91. Long-term shelter need showed correlations of 0.70 with FEMA-assisted housing outcomes for Laura and 0.80 for Ida. The framework was then applied to a deterministic landfall-shifting analysis of Hurricane Hugo across 100 candidate landfall locations along the South Carolina coast. The worst-case scenario was identified near Edisto Beach, producing an estimated 42,103 displaced households, 13,458 short-term shelter-need households, 20,290 long-term shelter-need households, and $38.26 billion in residential losses. Under a 25-mile relocation constraint, the MILP allocation model achieved 88.53% shelter coverage, leaving 2,328 households with unmet needs compared to 2,413 in the greedy approach. Most significantly, the MILP model reduced total travel distance by over 60%, totaling 56,205 household-miles (average 6.05 miles) compared to 141,773.80 miles (average 15.44 miles) for the greedy algorithm. These results demonstrate the value of integrating hazard, insurance, shelter-demand, and allocation modeling to support risk-informed hurricane housing recovery planning.
Recommended Citation
Adhikari, Nirmal, "An Integrated Model for Post-Hurricane Displacement, Shelter Demand and Shelter Allocation" (2026). All Theses. 4801.
https://open.clemson.edu/all_theses/4801