Identifying High-Risk Roadways for Infrastructure Investment Using Naturalistic Driving Data

Wolshon, Brian; Parr, Scott A.; Mousavi, Seyedeh Maryam · 2015 · ROSA P / Maritime Transportation Research and Education Center (MarTREC)

archive: archived pipeline: cataloged verified

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

Summary

This research addresses the limitations of traditional roadway safety assessment, which relies on retroactive analysis of historic crash data. Because crashes are rare events, this approach often identifies high-risk locations only after injuries or fatalities have occurred. The study investigates whether "surrogate safety measures," specifically clusters of high-magnitude negative jerk events (the rate of change of acceleration), can proactively identify unsafe roadway segments before crashes happen. The authors hypothesize that concentrations of abrupt braking maneuvers, indicative of near-crashes or evasive actions, correlate with long-term crash rates and can serve as a predictive indicator for infrastructure investment. The methodology utilized naturalistic driving data collected via GPS sensors from 31 participants in Baton Rouge, Louisiana, over a six-month period. The data was processed to calculate acceleration and jerk values, which were then geo-located to two specific roadways: LA 42 and LA 1248. Using Geographic Information Systems (GIS) linear referencing, the researchers matched GPS data points to the road network and compared them against five years of historic crash data (2009–2013). A sensitivity analysis determined the optimal parameters for the study, identifying a jerk threshold of -3 ft/s³ and a quarter-mile segment length as yielding the highest correlation between jerk events and crash rates. Negative binomial models were then employed to estimate crash frequency, incorporating variables such as average daily traffic (ADT), road curvature, and jerk-rate. The results demonstrated a significant statistical relationship between jerk-clusters and historic crash frequencies. In the negative binomial models for both roadways, jerk-rate was the only variable significantly related to long-term crash frequency. While ADT also showed a correlation for LA 1248, it was not significant for LA 42, and road curvature was insignificant in both models. The sensitivity analysis confirmed that quarter-mile segments provided the most accurate correlation, whereas shorter or longer segments reduced the statistical strength of the relationship. Heat maps visualizing jerk-rates and crash rates showed strong spatial alignment, indicating that locations with high concentrations of negative jerks corresponded to areas with higher historic crash rates. The study concludes that jerk-clusters are a stronger indicator of roadway safety than traditional metrics like ADT or geometric features. This finding suggests that naturalistic driving data can be used to predict safety problems proactively, allowing transportation agencies to allocate infrastructure investments to prevent crashes before they occur. The authors note that future advancements in GPS-enabled devices, such as smartphones, could enable crowd-sourced data collection, further enhancing the ability to identify high-risk roadways in real-time. This approach offers a potential shift from reactive to proactive highway safety management, potentially reducing damage, injury, and loss of life.

Key finding

Clusters of high magnitude negative jerk events were significantly correlated with long-term crash rates, providing a stronger indicator of safety than traditional variables such as average daily traffic and road curvature.

Methodology

naturalistic

Sample size: 31

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. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

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

• naturalistic crash near crash
• telematics crash prediction
• incidence prevalence
• induced exposure
• exposure measurement
• causation analyses

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, observational prevalence
• Methodological Resource: dataset resource