Examining macro-level traffic crashes considering nonlinear and spatiotemporal spillover effects

Area of Science:

• Transportation Science
• Urban Planning
• Geospatial Analysis

Background:

• Traditional traffic crash studies often assume linearity and spatial dependence, potentially underestimating risks.
• Effective traffic safety management requires understanding complex, non-linear, and spatiotemporal crash influences.

Purpose of the Study:

• To investigate the nonlinear and spatiotemporal spillover effects on traffic crashes.
• To develop an advanced model for macro-level crash analysis incorporating these effects.
• To identify key influencing factors and their impact on vehicular crash occurrences.

Main Methods:

• Utilized a geographically and temporally weighted method to capture spatiotemporal spillover effects.
• Integrated internal and spillover factors as independent variables.
• Employed gradient boosting decision trees to model nonlinear relationships and accumulated local effect plots for interpretation.
• Conducted a case study in New York City (2016-2019) using diverse data sources.

Main Results:

• Model performance significantly improved by incorporating nonlinear and spatiotemporal spillover effects.
• Identified significant nonlinear effects for factors like mixed land uses, sidewalks, and junction density.
• Highlighted spatiotemporal spillover effects from factors such as building density, bike parking density, and education attainment.

Conclusions:

• Nonlinear and spatiotemporal spillover effects are crucial for accurate traffic crash analysis.
• Findings provide actionable insights for developing targeted safety countermeasures and policies.
• Emphasizes the need for inter-regional collaboration in urban safety planning.