Trajectory-Based Surrogate Safety Assessment for Sustainable Freeway Management: Traffic-Condition Associations and Segment Heterogeneity Using a Bayesian Hierarchical Negative-Binomial Model

Sustainable freeway management requires safety monitoring that protects road users and supports reliable mobility. High-resolution vehicle trajectories support the screening of potentially hazardous interactions and loss-of-control conditions when independently validated crash labels are unavailable. This study develops an event-specific framework using 5623 trajectories from an approximately 700 m freeway corridor observed during a 60 min dry-weather daylight period. Seven segments and twelve five-minute intervals define a nominal grid of 84 segment–time cells. Potential loss-of-control events (S) are screened using SVCI > 1, and potential multi-vehicle conflicts (M) using positive TTC below 3 s. Qualifying frames are consolidated into distinct episodes before aggregation. The specified analysis uses separate Bayesian hierarchical negative-binomial models with fixed slopes, one segment random intercept per outcome, and positive dispersion. The framework distinguishes surrogate screening from confirmed crashes and contemporaneous association from causation. Complete posterior summaries, diagnostics, and comparisons on a common retained dataset are required before empirical conclusions about coefficient uncertainty, nonlinear turning points, or predictive improvement can be supported. By distinguishing event types and segment heterogeneity, the framework offers a methodological basis for prioritizing safety inspections and monitoring resources, linking surrogate safety assessment to sustainable transportation. Crash, delay, emission, and cost reductions are not evaluated.