🔬 physics

Bidirectional yet asymmetric causality between urban systems and traffic dynamics in 30 cities worldwide

This study introduces a novel spatio-temporal causality framework to analyze 30 global cities, revealing that while urban systems and traffic dynamics exhibit bidirectional but asymmetric relationships where urban form and function primarily drive mobility, distinct causal archetypes emerge that can guide context-sensitive sustainable planning.

Original authors: Yatao Zhang, Ye Hong, Song Gao, Martin Raubal

Published 2026-07-08

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yatao Zhang, Ye Hong, Song Gao, Martin Raubal

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a city as a giant, living organism. It has a skeleton (the roads), a shape (how buildings and parks are arranged), and a personality (what people do there, like working, shopping, or playing). Now, imagine the traffic as the blood flowing through this organism.

For a long time, scientists have tried to figure out the relationship between the city's body and its blood flow. Do the roads dictate where the traffic goes? Or does the traffic, over time, force the city to change its shape?

This paper, titled "Bidirectional yet asymmetric causality between urban systems and traffic dynamics in 30 cities worldwide," acts like a high-tech detective story. The researchers looked at 30 major cities across six continents (from Singapore to New York, from London to Rio) to solve this mystery.

Here is the simple breakdown of what they found:

1. The Two-Way Street (But One Way is Wider)

The study confirms that the relationship is bidirectional, meaning it's a two-way street.

City $\rightarrow$ Traffic: The layout of the city definitely changes how traffic moves.
Traffic $\rightarrow$ City: Over a long time, the way people move can actually force the city to change its layout (like building a new subway line because roads are too clogged).

However, the key discovery is that this street is asymmetric.
Think of it like a heavy truck and a bicycle. The truck (the city's structure) can easily knock the bicycle (traffic) off course. But the bicycle has a much harder time knocking the truck over.

The Finding: In most cities, the city's design (roads, building shapes, and land use) has a much stronger influence on traffic than traffic has on the city. The city "pushes" the traffic around far more than the traffic "pushes" the city back.

2. The "Triple Helix" of the City

The researchers didn't just look at "roads." They broke the city down into three distinct parts, like a three-stranded DNA helix:

Structure: The skeleton (the road network).
Form: The shape (how dense the buildings are, how the city looks from above).
Function: The activity (where the shops, offices, and homes are).

The Surprise:
When they just looked at correlation (how closely things move together), the Roads (Structure) seemed to be the most important. It makes sense: cars need roads to drive on.

But when they looked at causation (what actually causes the other to happen), the story changed.
Form and Function (where people live, work, and play, and how the city is shaped) were actually the stronger drivers of traffic congestion.
Analogy: Imagine a party. The roads are the hallway (Structure). The party guests (Form/Function) are the people. You can have a huge hallway, but if you put 500 people in a tiny room at the end of it, the hallway gets jammed. The people (Form/Function) are the real cause of the jam, not just the hallway itself.

3. The "Rest Day" vs. "Work Day" Difference

The researchers found that the city behaves differently on weekends compared to weekdays.

Rest Days (Weekends): The city is more like a flexible dance floor. People move around freely, and the city's shape and function strongly dictate where they go. The "push" from the city to the traffic is very strong here.
Work Days (Weekdays): The city becomes a rigid factory assembly line. Everyone is rushing to the same places at the same time (commuting). The traffic gets so clogged and predictable that the city's shape matters a little less because the "rush" overrides the design. In some cities, the traffic even starts to "push back" harder against the city design during these chaotic hours.

4. Three Types of Cities

Based on how these cities interact, the researchers grouped them into three "personality types":

The Tightly Coupled (e.g., Singapore, New Delhi): The city and traffic are best friends. They move in perfect sync. Changing the city layout here would immediately fix traffic.
The Pattern-Heterogeneous (e.g., London, Paris): The relationship is complex and varies by neighborhood. Some parts of the city are tightly linked, others are not. You can't use a "one-size-fits-all" solution here; you need a custom plan for each district.
The Workday-Attenuated (e.g., Zurich, Sydney): On weekends, the city controls the traffic well. But on workdays, the connection weakens. The traffic becomes so chaotic that the city's design struggles to influence it. To fix traffic here, you need to manage the people's schedules (demand), not just the roads.

The Big Takeaway

The paper concludes that we can't just look at traffic and say, "We need more roads."

Old Thinking: "The roads are too narrow, let's build wider ones."
New Insight: "The city's shape and where we put our offices and homes are actually the main reasons for the traffic. If we want to fix the flow, we need to redesign the city's shape and function, not just its skeleton."

In short: The city designs the traffic, but the traffic doesn't design the city very easily. To make cities smoother, we need to look at the "personality" of the city (where people live and work) rather than just the "skeleton" (the roads).

Technical Summary: Bidirectional Yet Asymmetric Causality Between Urban Systems and Traffic Dynamics

Problem Statement
Understanding the co-evolution of urban systems and traffic dynamics is essential for sustainable and resilient city planning. While existing literature has extensively mapped statistical associations between built-environment factors (such as road networks, land use, and morphology) and traffic states, these studies often fail to distinguish correlation from causation. A critical gap remains in comprehensively characterizing the bidirectional feedback mechanisms and their potential asymmetries across diverse global contexts. Furthermore, existing causal inference methods often struggle to simultaneously address spatial heterogeneity, temporal variation, and the complex interplay between urban structure, form, and function. Current approaches either focus on temporal causality in time-series data or spatial causality in static cross-sections, lacking an integrated framework capable of capturing the dynamic, spatio-temporal nature of urban mobility.

Methodology
To address these challenges, the authors propose a spatio-temporal causality framework that integrates Spatio-Temporal Weighted Regression (STWR) with a novel Spatio-Temporal Convergent Cross-Mapping (STCCM) method.

Data and Spatial Units: The study analyzes 30 cities across six continents, selected based on data availability, congestion levels, and global distribution. Traffic dynamics are measured using a congestion indicator derived from HERE Technologies' jam factors. Urban systems are characterized through a "triple helix" of components:
- Structure: Road network topologies (e.g., node density, centrality).
- Form: Landscape-level morphological metrics (e.g., patch density, edge density).
- Function: Land-use attributes (e.g., commercial, residential, transport areas).
  Spatial units are defined as Buffered Voronoi Cells (BVCells), constructed by clustering road intersection nodes and applying a 500m buffer to capture localized traffic dynamics and surrounding environments.
Spatio-Temporal Weighted Regression (STWR): This step quantifies the spatially heterogeneous and temporally varying relationships between urban system features and traffic dynamics. STWR generates local coefficients for each feature, which are then smoothed and aggregated to create composite indicators for urban structure, form, and function. This allows the model to capture how the strength of association varies across space and time (e.g., between workdays and rest days).
Spatio-Temporal Convergent Cross-Mapping (STCCM): Building on the composite indicators, STCCM is employed to infer bidirectional causal relationships. Unlike standard Convergent Cross-Mapping (CCM) or Geographical CCM (GCCM), STCCM jointly embeds spatial dependency and temporal variation into state-space reconstruction.
- Mechanism: The method reconstructs shadow manifolds for urban systems ( $X$ ) and traffic dynamics ( $Y$ ) using spatial lags and temporal snapshots, specifically focusing on congested periods.
- Inference: Causality is determined by the convergence of cross-mapping skill ( $\rho$ ) as the library size ( $L$ ) increases. If the predictability of $Y$ from $X$ ( $\rho_{X \to Y}$ ) converges and exceeds the reverse ( $\rho_{Y \to X}$ ), a dominant causal direction is inferred.
- Asymmetry: The magnitude of the difference ( $\Delta\rho$ ) between the two directions quantifies the asymmetry of the causal relationship.
Clustering: A causality-driven clustering approach uses Dynamic Time Warping (DTW) for curve shape similarity and Jaccard distance for magnitude overlap to group cities into distinct causal archetypes.

Key Contributions

Methodological Innovation: The development of STCCM, which extends causal inference to the spatio-temporal domain by integrating spatial dependencies (via spatial lags in state-space reconstruction) with temporal dynamics, overcoming limitations of previous GCCM and CCM models.
Conceptual Framework: The application of the "triple helix" (structure, form, function) to systematically disentangle the distinct mechanisms through which urban systems influence traffic.
Global Empirical Analysis: A comprehensive analysis of bidirectional causality across 30 diverse cities, moving beyond single-city case studies to identify global patterns and archetypes.

Results

Bidirectional but Asymmetric Causality: The study confirms that causality flows in both directions between urban systems and traffic dynamics. However, a clear asymmetry exists: urban systems generally exert a stronger causal influence on traffic dynamics than the reverse. This is statistically significant across the majority of cities (26–29 out of 30, depending on the component).
Dominance of Form and Function: While urban structure (road networks) shows the highest statistical correlation ( $R^2$ ) with traffic dynamics, urban form and function often demonstrate stronger causal influence ( $\rho$ ). For instance, in Singapore, structure had the highest $R^2$ (0.95), but form showed the highest causal strength ( $\rho = 0.65$ ). This suggests that while road networks constrain movement, the spatial layout and land-use distribution (demand-side factors) are the primary drivers of congestion patterns.
Temporal Variation: Causal patterns differ between rest days and workdays. Rest days show stronger and more distinct causal signals, likely due to self-organized travel behavior. Workdays exhibit "attenuated" causality in some cities, where fixed commuting schedules lead to saturation, reducing the distinctiveness of causal pathways.
Three Causal Archetypes: The 30 cities were clustered into three distinct groups:
1. Tightly Coupled: Cities (e.g., New Delhi, Singapore) with strong bidirectional causality, where urban variables explain significant traffic variance.
2. Pattern-Heterogeneous: Cities (e.g., Helsinki, Vienna) with moderate causality but distinct internal variations, requiring diagnostic, context-specific planning.
3. Workday-Attenuated: Cities (e.g., Zurich, Sydney) where causal links weaken significantly on workdays, suggesting static urban levers are insufficient without operational demand management.

Significance and Claims
The paper claims that moving from statistical association to causal inference is critical for effective urban and transport planning. The findings challenge the reliance on supply-side solutions (e.g., expanding road networks) by demonstrating that demand-side factors (urban form and function) often drive traffic dynamics more profoundly. The identified causal archetypes provide a basis for city-to-city learning, allowing planners to tailor interventions based on whether a city is tightly coupled, heterogeneous, or attenuated. The authors emphasize that their framework offers a scalable blueprint for translating causal diagnostics into targeted, context-sensitive strategies for congestion mitigation and sustainable development, while acknowledging limitations regarding the reliance on vehicular data and the need for future work on multimodal integration and sub-regional heterogeneity.