Analysis of Association Between Demographic, Socioeconomic, and Built Environment Factors and Pedestrian Safety Using Traditional and AI Approaches

Pedestrian safety is a critical concern, particularly in urbanized areas where increasing population densities and heavy reliance on motorized transportation elevate risks for vulnerable road users who travel on foot. Creating safe, walkable environments is not only a public health priority but also vital for sustainable urban development. To better understand pedestrian crash risks, this dissertation explores the relationship between built-environment and socio-economic factors and their influence on crash risks at the Census Block Groups (CBGs) level in Austin, San Antonio, and Dallas. Given the spatial nature of CBGs, the spatial distribution of pedestrian crash risks within urban areas, such as Austin, is also examined to understand how built-environment and socio-economic factors contribute to this variability. Additionally, the study identifies distinct individual-scale pedestrian crash clustering patterns by severity using data from California. The dissertation is organized around three major studies, each addressing a specific research question: What demographic, socioeconomic, and built environment factors are associated with pedestrian safety? What are the spatial variations in pedestrian crash risks at the census block group level in urban areas? Are individual-level pedestrian crashes spatially clustered? Specifically, the first study investigates the impact of socio-economic, built-environment, transit, and trip characteristics on pedestrian crashes in the Texas cities of Austin, San Antonio, and Dallas. It seeks to identify key factors influencing pedestrian crash risks across CBGs and evaluates the effectiveness of machine learning models, specifically SHapley Additive exPlanations (SHAP), in explaining how these factors affect Equivalent Property Damage Only (EPDO) rates. The findings reveal that auto-oriented network density is consistently associated with higher pedestrian crash risks, while pedestrian-oriented network density and sidewalk coverage generally have negative associations. Transit frequency and socio-economic factors, such as the percentage of zero-car households, also significantly impact pedestrian safety. The study underscores the need for targeted interventions in disadvantaged CBGs with higher levels of zero-car households, more mixed land use, and denser transit networks, but lower percentages of high-wage workers and certain community services. The second study examines the spatial variability of pedestrian crash risks within urban areas, focusing on Austin using Multiscale Geographically Weighted Regression (MGWR). The analysis reveals that higher percentages of two-plus-car households are associated with lower crash risks, possibly due to reduced pedestrian exposure. Conversely, areas with higher employment rate and household entropy, auto network density, and higher transit frequency exhibit greater crash risks, likely driven by increased interactions between pedestrians and vehicles in more densely populated areas, transit-oriented environments. The third study identifies patterns in pedestrian crash severity using a clustering framework enhanced by explainable artificial intelligence (AI) techniques. This analysis, based on data from California, uncovers distinct patterns in crash severity by examining factors associated with fatal, injury, and non-injury crashes. It also explores how societal and demographic factors differ in their association with varying levels of crash severity, highlighting the disparities between underserved and more resilient communities. The most impactful factors in fatal crashes include pedestrian sobriety impairment, lighting conditions, and macro-scale traffic fatalities. In less severe crashes, broader societal and demographic influences are more prominent. The dissertation may provide insight for policymakers seeking to improve pedestrian traffic safety.

Name thesis:degree_name

Doctor of Philosophy

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Geographic Information Science

Grantor

Texas State University

Year dc:date.issued

2024

Author dc:creator

• Liu, Jinli

Advisors dc:contributor.advisor

• Zhan, F. Benjamin
• Das, Subasish

• pedestrian safety
• machine learning
• SHAP

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/10877/19982

OAI identifier oai:identifier

oai:digital.library.txst.edu:10877/19982