Using temporal detrending to observe the spatial correlation of traffic

Area of Science:

• Transportation Science
• Traffic Engineering
• Network Analysis

Background:

• Understanding spatial correlations between traffic links is crucial for accurate traffic flow prediction.
• Existing short-term traffic forecasting models often overlook the nuanced spatial relationships under varying traffic conditions.
• Freeway systems with grid-like topologies offer unique opportunities to study these correlations.

Purpose of the Study:

• To investigate the spatial correlation of freeway traffic links across different traffic regimes.
• To develop a robust algorithm for measuring traffic link correlations by removing temporal trends.
• To identify how network topology and traffic conditions influence spatial correlation patterns.

Main Methods:

• Empirical analysis of 140 freeway traffic links in the Minneapolis-St. Paul freeway system.
• Development of a novel algorithm to eliminate temporal trends in hourly, weekly, and system dimensions.
• Juxtaposition of positive and negative spatial correlations within a grid-like network topology.

Main Results:

• A stronger negative spatial correlation between traffic links is observed during rush hours, particularly in parallel links and upstream segments, due to congestion influencing route choices.
• A consistent strong positive correlation exists between upstream and downstream links, irrespective of time and day.
• This positive correlation is more pronounced in uncongested traffic regimes where flow is less impeded.

Conclusions:

• The spatial correlation structure of traffic links varies significantly with traffic regimes and network topology.
• The findings provide a deeper understanding of traffic dynamics, especially the interplay between congestion and route choice.
• Incorporating the extracted spatial correlation patterns can enhance the accuracy and reliability of short-term traffic forecasting models.