Inferring urban polycentricity from the variability in human mobility patterns

Carmen Cabrera-Arnau, Chen Zhong, Michael Batty, Ricardo Silva, Soong Moon KangScientific 2023 · DOI

TL;DR

This paper introduces a novel probabilistic framework to quantify urban polycentricity—the degree to which a city has multiple functional centers—by analyzing smart travel card data. By applying Poisson mixture models to journey lengths in London and Seoul, the researchers demonstrate that while both cities show complex structures, London remains significantly more monocentric than the highly polycentric Seoul.

What Problem Does This Paper Address?

Urban planners have long championed the "polycentric city" as a solution to the "monocentric" woes of congestion, high rents, and poor accessibility. However, "polycentricity" is a notoriously "fuzzy" concept. Traditional methods for measuring it often rely on static data, like population density or employment locations from censuses, which are updated infrequently and fail to capture the dynamic reality of how people actually use urban space.

The authors argue that urban structure is best revealed through human mobility patterns. The challenge lies in moving beyond simple averages to account for the inherent variability in travel. If a city were perfectly monocentric, almost all trips would originate from a single central "nucleus." In reality, trips are a messy blend of local errands and city-wide commutes. This paper seeks to turn that "messiness" into a formal metric for urban structure.

Methods

The researchers treat the public transport system as a complex network, where stations are nodes and transit lines are edges. Distance is measured not in kilometers, but in "network distance" (the number of stops between stations).

The core innovation is the use of Poisson mixture models (PMM). Instead of assuming a single average trip length for each station, the PMM assumes that the journeys terminating at a station come from different statistical "sub-populations":

  1. Proximal Component: Short, local journeys (suggesting the presence of a nearby sub-center).
  2. Distal Component: Long, city-wide journeys (suggesting trips from the main nucleus or far-flung areas).
  3. Medial Component: Intermediate journeys (used in their 3-component model).

By comparing actual travel data against a "monocentric hypothesis"—a null model where the expected journey length to any station equals its distance from the city center—the authors can quantify how much a city "departs" from a single-center model.

Datasets

The study leverages high-resolution "smart card" data from two major metropolises:

  • London (UK): Oyster card data from five full weekdays in January 2014, encompassing ~3.22 million journeys across 382 stations.
  • Seoul (South Korea): T-money card data from four full weekdays in December 2012, encompassing ~5.96 million journeys across 512 stations.
  • Basemaps: Geographic data for both cities was sourced from OpenStreetMap.

Key Insights and Ideas

  • Variability is the Signal: The authors found that a single-component model (a simple average) fails to capture urban reality, especially in Seoul. The variability in trip lengths is what actually reveals the hidden polycentric structure.
  • The "Proximal" Constant: A striking finding was that the "proximal" (local) trip length remains remarkably consistent—around 5 stops—regardless of how far a station is from the city center. This suggests that even in peripheral areas, people rely on local hubs for daily socioeconomic activities.
  • Spectrum, Not Binary: Rather than labeling a city as "monocentric" or "polycentric," the method treats urban structure as a spectrum. This allows for a more nuanced comparison of how different cities evolved.
  • Historical Path Dependency: The results hint at the impact of history. London’s transport network began in the 19th century (favoring a monocentric hub-and-spoke model), while Seoul’s began in the 1970s, a period when urban planning had already begun shifting toward decentralization.

Key Results

  • London is more Monocentric: London showed a much stronger correlation ($R = 0.758$) with the monocentric hypothesis. Its travel patterns still heavily revolve around a central nucleus (defined in the study as Piccadilly Circus).
  • Seoul is more Polycentric: Seoul showed a significantly lower correlation ($R = 0.585$) and a much flatter regression slope for journey lengths. This indicates that Seoul has successfully developed a more distributed network of centers (with City Hall as the primary nucleus).
  • Model Fit: The 2- and 3-component Poisson mixture models provided a much better fit for the data than a single Poisson distribution, successfully "unpicking" the different types of urban travel.

Limitations and Future Work

The authors acknowledge several constraints:

  • Unknown Origins: The specific dataset used only provided the number of stops traveled, not the specific origin station. This prevented the use of Euclidean distance or travel time.
  • Modal Blindness: The study focuses exclusively on public transport. It does not account for private car travel, walking, or cycling, which are significant components of urban mobility.
  • Temporal Resolution: The analysis looks at a "typical weekday" but doesn't yet explore how these patterns shift between morning peaks, evening peaks, or over several decades of urban growth.
  • Future Directions: The researchers suggest applying this framework to historical data to see how cities "transition" from monocentric to polycentric over time, which could provide invaluable insights for future urban policy.

Summary updated

Abstract

The polycentric city model has gained popularity in spatial planning policy, since it is believed to overcome some of the problems often present in monocentric metropolises, ranging from congestion to difficult accessibility to jobs and services. However, the concept 'polycentric city' has a fuzzy definition and as a result, the extent to which a city is polycentric cannot be easily determined. Here, we leverage the fine spatio-temporal resolution of smart travel card data to infer urban polycentricity by examining how a city departs from a well-defined monocentric model. In particular, we analyse the human movements that arise as a result of sophisticated forms of urban structure by introducing a novel probabilistic approach which captures the complexity of these human movements. We focus on London (UK) and Seoul (South Korea) as our two case studies, and we specifically find evidence that London displays a higher degree of monocentricity than Seoul, suggesting that Seoul is likely to be more polycentric than London.
BibTeX
@inproceedings{cabreraarnau2023inferring,
  author = {Carmen Cabrera-Arnau and Chen Zhong and Michael Batty and Ricardo Silva and Soong Moon Kang},
  title = {{Inferring urban polycentricity from the variability in human mobility patterns}},
  booktitle = {Scientific},
  year = {2023},
  doi = {10.1038/s41598-023-33003-7},
}
Figure