Quantitative Estimation and Analysis of Spatiotemporal Delay Effects in Expressway Traffic Accidents

Yunfei Zhang, Zhengrui Pan, Fangqi Zhu, Chaoyang Shi, Xue YangISPRS Int. J. Geo Inf 2024 · DOI

TL;DR

This paper introduces a novel method to quantify and analyze the spatiotemporal delay effects of expressway traffic accidents by integrating multi-source geographic data and a Traffic Performance Index (TPI). The approach effectively identifies delay hotspots and reveals significant correlations between accident characteristics and the severity and spread of traffic disruptions, offering valuable insights for traffic management.

What Problem Does This Paper Address?

Traffic accidents on expressways are a major cause of unpredictable, non-recurrent congestion, leading to significant delays, economic losses, and disruptions to transportation systems. Accurately estimating the spatial extent and temporal duration of these delays, and understanding the factors that influence them, is crucial for effective traffic management and emergency response. Previous methods often relied on idealized assumptions or struggled with integrating diverse data sources. This paper aims to bridge this gap by providing a more comprehensive and data-driven approach to analyze these spatiotemporal delay effects.

Methods

The study employs a multi-step methodology:

  • Data Preprocessing and Fusion: Standardizing coordinate systems and refining accident locations by matching trajectory data with a road network map using a shortest-path-based algorithm.
  • Traffic Performance Index (TPI): A novel metric is introduced to quantify congestion severity. The TPI is calculated as the ratio of actual travel time to free-flow travel time, accounting for varying road conditions.
  • Spatiotemporal Delay Impact Assessment:
    • Identifying impactful accidents based on speed reductions within a defined time window.
  • Quantifying key impact metrics: Accident Impact Area (Sa), Accident Impact Duration (Ta), and Accident Impact Degree (Da).
  • Spatial Analysis:
    • Fuzzy C-Means (FCM) Clustering: To group accidents based on their spatiotemporal delay patterns (duration, extent, severity), identifying distinct accident categories.
  • Kernel Density Estimation (KDE): To visualize and identify spatial hotspots of impact duration and length.
  • Regression Analysis:
    • Spatial Error Model (SEM): Selected as the optimal model (based on LM tests) to analyze the relationship between various accident characteristics and spatiotemporal delay impacts (duration and length).

Datasets

  • Expressway Accident Dataset: Containing details on accident locations, mileage, time of occurrence, types, casualties, and weather conditions.
  • Vehicle Trajectory Data: Including license plates, timestamps, coordinates, and speeds for over 534,000 vehicles.
  • Traffic Road Network Map: Containing segment identifiers, geometry, lanes, speed limits, and road types.

The study utilized data from expressways in Hunan Province, China, for October 2018, with 3474 traffic accidents recorded.

Key Insights and Ideas

  • TPI as a Robust Congestion Measure: The TPI offers a more nuanced assessment of congestion than simple speed thresholds, as it accounts for the inherent differences in road characteristics.
  • Spatiotemporal Patterns are Key: Accidents don't just cause localized delays; they create ripple effects that spread both spatially and temporally, influencing traffic far beyond the immediate accident site.
  • Accident Characteristics Matter Significantly: The type of collision, presence of casualties, road type (bridges/tunnels vs. regular roads), and occurrence of secondary accidents are strong predictors of delay severity and duration.
  • Hotspots Reveal Vulnerable Corridors: Identifying specific expressway segments (e.g., Changsha-Zhangjiajie, Beijing-Hong Kong-Macao) as persistent delay hotspots highlights areas requiring targeted infrastructure improvements and traffic management strategies.
  • Pre-accident Conditions Influence Impact: Higher speeds before an accident might paradoxically lead to shorter impact durations, possibly due to lower traffic volumes and quicker clearance.

Key Results

  • The study identified 3474 traffic accidents in Hunan Province in October 2018.
  • Fuzzy C-Means clustering categorized 2179 accidents into eight distinct groups based on delay impact, with clusters 1, 3, 4, and 6 representing the highest severity levels (level 4).
  • Kernel density analysis revealed significant delay hotspots concentrated around Changsha and along major expressways like the Beijing-Hong Kong-Macao Expressway.
  • The SEM model showed that collisions with stationary vehicles and fixed objects lead to significantly longer delay durations and impact lengths compared to collisions with moving vehicles.
  • Accidents on bridges and tunnels exhibited shorter impact lengths (44.8% and 58.2% less, respectively) compared to regular roads, likely due to faster response times.

Limitations and Future Work

  • Data Availability: The study notes that data are unavailable due to privacy or ethical restrictions, which might limit the broader applicability or replication of the dataset.
  • Urban Environments: The current methodology is focused on expressways. Future research should extend this to urban traffic environments, which present more complex dynamics.
  • Enhanced Estimation: Integrating video monitoring and other supplementary data sources could further enhance delay estimation accuracy.
  • Road Network Resilience: Comprehensive evaluations of road network resilience using detailed models of spatiotemporal delay effects from various disruptions are suggested.
  • Real-time Strategies: The ultimate goal is to support the development of robust, real-time traffic management strategies.

Summary updated

BibTeX
@article{zhang2024quantitative,
  author = {Yunfei Zhang and Zhengrui Pan and Fangqi Zhu and Chaoyang Shi and Xue Yang},
  title = {{Quantitative Estimation and Analysis of Spatiotemporal Delay Effects in Expressway Traffic Accidents}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2024},
  doi = {10.3390/ijgi13110407},
}
Figure