Portuguese two-lane highways: modelling crash frequencies for different temporal and spatial aggregation of crash data

da Costa, Jocilene Otilia; Maria, Alice Prudêncio Jacques; Pereira, Paulo António Alves; Freitas, Elisabete Fraga; Soares, Francisco Emanuel Cunha · 2015 · OpenAlex-citations

DOI: 10.3846/16484142.2015.1073619

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the lack of recent Crash Prediction Models (CPMs) for Portuguese two-lane rural highways, aiming to identify factors contributing to fatal and injury crash frequencies. The research focuses on national roads NR-14, NR-101, and NR-206 in Northern Portugal. The primary objectives were to determine significant crash contributory factors and to analyze how different temporal and spatial aggregations of crash data impact model performance and factor identification. The methodology utilized data from 88 two-lane road segments, each initially defined as 200 meters long, covering the period from 1999 to 2010. The dataset included Average Annual Daily Traffic (AADT) and geometric characteristics such as lane width, shoulder width, lateral offset, horizontal and vertical sinuosity, and density of access points. Due to a high prevalence of zero-crash records (70% in the disaggregated data), the authors explored various spatial aggregations (grouping segments into 400-meter units) and temporal aggregations (1 to 12 years) to reduce zero-inflation. The study employed Generalized Estimating Equations (GEE) with a negative binomial link function to model the longitudinal data, testing independent, exchangeable, and autoregressive correlation structures. The results indicated that observations within each road segment followed an exchangeable correlation structure. The analysis revealed that aggregating data spatially and temporally significantly reduced the proportion of zero-crash records, thereby improving model stability. The major contributing factors identified for crash frequencies were traffic volume (AADT), lane width, vertical sinuosity, and the density of access points. Other geometric variables, such as shoulder width and horizontal sinuosity, were not found to be statistically significant in the final models. The study determined that an acceptable CPM could be developed for 400-meter-long segments over a cumulative two-year period, which provided a robust balance between data availability and model accuracy. The significance of this work lies in providing the first recent CPMs for Portuguese two-lane rural highways, offering transportation agencies a tool to estimate expected crash frequencies and prioritize safety improvements. By demonstrating the impact of data aggregation on model capability, the study highlights the importance of selecting appropriate spatial and temporal units when dealing with sparse crash data. The findings suggest that traffic volume and specific geometric features like lane width and access density are critical targets for safety interventions on these roadways.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• incidence prevalence

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: crash risk outcomes
• Methodological Resource: dataset resource