Estimate traffic cyclist crashes using Poisson-Tweedie models.

Area of Science:

• Urban planning and transportation safety.
• Statistical modeling for crash analysis.
• Geospatial analysis of traffic incidents.

Background:

• Cyclist safety is a major urban transportation concern, influenced by infrastructure and spatial factors.
• Understanding cyclist crash patterns is crucial for effective safety interventions.
• Previous studies highlight the need for detailed analysis of contributing factors.

Purpose of the Study:

• To estimate cyclist crash frequencies in Lisbon using advanced statistical models.
• To identify key infrastructure and spatial variables associated with cyclist crashes.
• To evaluate the predictive performance of different modeling approaches for cyclist safety.

Main Methods:

• Utilized Poisson-Tweedie models for overdispersed count data of 541 cyclist crashes (2015-2019).
• Spatially structured crash data into 250x250 meter grid cells, incorporating covariates like road length and intersection types.
• Developed and compared base (aggregated) and disaggregated models, including spatial autocorrelation.

Main Results:

• Intersection density and road length showed strong associations with cyclist crash frequency.
• Cycleway length had a significant but more modest effect on crash rates.
• The disaggregated model provided better interpretability but not superior predictive accuracy compared to the base model.

Conclusions:

• Improving intersection design offers greater safety benefits than solely increasing cycling infrastructure length.
• Predictive modeling can identify high-risk zones for proactive cyclist safety planning.
• Results offer actionable insights for data-driven urban mobility and cyclist safety management.