Publication Date
12-2023
Date of Final Oral Examination (Defense)
October 2023
Type of Culminating Activity
Thesis
Degree Title
Master of Engineering in Civil Engineering
Department Filter
Civil Engineering
Department
Civil Engineering
Supervisory Committee Chair
Yang Lu, Ph.D.
Supervisory Committee Member
Bhaskar Chittoori, Ph.D.
Supervisory Committee Member
Ron Fisher, Ph.D.
Abstract
Natural hazards pose significant threats to communities, impacting livelihoods and property. Socially vulnerable communities are disproportionately affected by these hazards, highlighting the importance of considering social vulnerability in hazard risk assessment. Research on multi-hazards risk assessment, and social vulnerability can provide a comprehensive understanding of hazards and local vulnerability to support the reduction of disaster suffering overall. This thesis presents a comprehensive study conducted in Idaho, US, to examine the regional multi-hazards risk and assess the progression of social vulnerability metrics in wildfire-affected areas in a case study.
The first part of the research proposes a multi-hazards risk assessment methodology that integrates social vulnerability analysis and machine learning models. Five machine learning methods, including Naïve Bayes, K-Nearest Neighbors, Logistic Regression, Random Forest, and K-Means, are utilized for hazard mapping and characterization of multi-hazards (Flooding, Wildfires, and Seismic). Additionally, a composite index tailored to the case-study area is employed to evaluate social vulnerability.
The second part of the research focuses on the analysis of social vulnerability trends over a 10-year period in wildfire-affected areas as a case study. By considering short-term and long-term trends, indicators of social vulnerability are examined across different levels of wildfires risk. The results indicate a correlation between high wildfires risk levels and elevated social vulnerability, highlighting the importance of understanding the interplay between hazards and vulnerability.
The results indicate that RF model performs best in both hazard-related and social vulnerability datasets. The most cities at multi-hazards risk account for 34.12% of total studied cities (covering 20.80% land). Additionally, high multi-hazards level and socially vulnerable cities account for 15.88% (covering 4.92% land). This study generates a multi-hazards risk map which show a wide variety of spatial patterns and a corresponding understanding of where regional high hazards potential and vulnerable areas are. Additionally, the case study reveals that high wildfires risk levels generally follow the highest levels of social vulnerability. There are cities where high or low social vulnerability remains persistent over time, even in the absence of wildfire effects. Comparing regions impacted by wildfires to similar non-wildfire affected areas, the case study identifies a 2.74% rise in civilian unemployment rates and a 0.77% decline in households earning over $200,000 annually in the wildfire-affected regions.
The thesis emphasizes the urgent need for information-based prioritization and effective policy measures to reduce natural hazards risks. The research contributes to a better understanding of the distribution of multi-hazards risk and social vulnerability, which can aid in enhancing natural hazards risk perception, preparedness, and resilience in Idaho and similar regions. The findings can inform decision-making processes, support targeted interventions, and promote proactive measures to mitigate the impacts of multi-hazards and enhance community resilience.
DOI
https://doi.org/10.18122/td.2204.boisestate
Recommended Citation
Wang, Donglei, "Understanding Social Vulnerability to Natural Hazards Through Machine Learning and Dynamic Spatiotemporal Analysis" (2023). Boise State University Theses and Dissertations. 2204.
https://doi.org/10.18122/td.2204.boisestate