Detecting anomalous spatial interaction patterns by maximizing urban population carrying capacity

https://doi.org/10.1016/j.compenvurbsys.2021.101616Get rights and content

Highlights

  • A weighted origin-destination bipartite network is constructed to structure spatial interactions between urban space areas.

  • The optimal interactions between urban space areas which can maximize the urban population carrying abilities are obtained.

  • Anomalous interaction patterns causing both overload and underload are detected and analyzed.

Abstract

Rapid urbanization in China has prompted plenty of urban facilities to be constructed with the expectation of harmonizing with the rapid growth of urban population. However, regarding the spatial interactions produced by cross-area human mobility, the diversity and variability of residents' trip requirements inevitably cause the deviations of the real interaction patterns from the optimal status determined by the current allocation of urban facilities. To maximize the utility of urban facility allocation, we designed a bipartite network-based approach to explore anomalous spatial interaction patterns within cities. First, considering the potential area attractiveness, a weighted origin-destination bipartite network was constructed to structure the spatial interactions between traffic analysis zones. Then, a branch and bound (BnB) based augmenting path algorithm was proposed to optimize the distribution of spatial interactions, which can maximize the urban population carrying capabilities. Finally, anomalous interaction patterns causing both overload and underload were detected through comparisons between the actual and optimal spatial interaction distribution. The experimental results show that the two types of anomalous interaction patterns have significantly different spatial distribution characteristics. Through further analyzing the relationships between the two types of anomalous interaction patterns and urban evolution process, this study can also provide targeted decision supports for the accommodating of urban facility allocations to the distributions of resident trips in space.

Introduction

The basic manifestation of urbanization progress is the continuous growth of urban populations. According to the United Nations World Urbanization Prospects (United Nations, 2019), 68% of the world population is projected to be residing in urban areas by 2050. The rate of urbanization is even more pronounced in China. Over the past 40 years, China has experienced the largest and fastest urbanization in the world. In 2019, the urbanization rate in China reached 60.6% with a permanent urban population of nearly 850 million (National Bureau of Statistics of China, 2020). Urban agglomeration resulted in the gradual emergence of a series of social problems and rapid expansion of urban spaces. Therefore, improving urban population carrying capacity has attracted widespread attention from researchers and government officials (Shi, Shi, & Wang, 2019).

Existing researches have confirmed that the aggregated distribution of urban facilities has an agglomeration economics effect, which can significantly increase production efficiency and create more economic benefits. The influence of those areas with complete facilities can be radiated to surrounding areas and even the entire city. The powerful attractions of these areas usually aggregate crowd flow that exceed their population carrying capacities and result in overload-related anomalous interaction patterns. At the same time, in order to alleviate urban problems caused by sharp population growth, massive infrastructure projects have been developed in some suburb areas of a city. However, these newly-built areas may become “ghost cities” if there is not enough population to utilize the relatively complete facilities (Williams, Xu, Tan, Foster, & Chen, 2019). In this case, the lack of expected popularity crowd flow will produce underload-related anomalous interaction patterns in these areas.

The above-mentioned phenomena can be summarized as the spatial mismatch problem, meaning that the unreasonable distribution of urban infrastructures leads to that residents have to bear higher transportation costs to satisfy their multiple demands with the utilization of public facilities (Andersson, Haltiwanger, Kutzbach, Pollakowski, & Weinberg, 2014; Perle, Bauder, & Beckett, 2010). Nowadays, the spatial mismatch problem caused by the re-layout of urban spatial structure has gradually become the main factor of restricting the resident employment, traffic operation efficiency, economic development and so on (Sun & Lv, 2020; Wang et al., 2020). With the help of multi-source trip trajectory big data, the discovery of spatial interaction patterns within cities can provide latent knowledge for the spatial layout optimization of urban facilities (Wu et al., 2019); thus, it has been widely employed in the fields of complex networks, intelligent transportation, and urban geography (Hu, Wang, Wu, & Stanley, 2020; Zhang, Chong, Li, & Nie, 2020).

Spatial interaction is a general term for the movement of people, goods, or information over space (Fotheringham, 2001). In this study, spatial interaction is specifically defined as the human mobility across distinct urban sub-areas. As an important external manifestation of urban spatial mismatch, the process of spatial interaction can also drive the change of spatial distribution patterns of urban facilities (Timmermans, 2001). Consequently, spatial interaction related researches have received unprecedented attention in the era of big data (Shu et al., 2020; Song et al., 2019).Various studies have been conducted to theoretically analyze the spatial interaction patterns by considering dynamic evolution, spatiotemporal resolution, and mobility uncertainties (Espín Noboa, Lemmerich, Singer, & Strohmaier, 2016; Yin, Gao, Du, & Wang, 2016). For example, taxi service ranges were analyzed to guide the optimization of ride-sharing schemes (Zhang et al., 2017; Dong, Wang, Li, & Zhang, 2018). Saberi, Mahmassani, Brockmann, and Hosseini (2017) analyzed travel demand based on multi-layer complex network for detecting characteristic origin-destination areas of urban spatial interactions. Moreover, several measurement methods have been proposed to predict spatial interaction fluxes from the perspective of travel cost and urban structures (Sarkar et al., 2017; Shi, Zhou, Wang, & Wang, 2018). Although the spatial layout of urban functional areas is generally inconsistent with the governments' expectations (Liu, Wang, Xiao, & Gao, 2012; Zhong, Arisona, Huang, Batty, & Schmitt, 2014; Long & Thill, 2015), the characteristics of urban block functions and urban structure can be accurately deducted from spatial interaction patterns (Kong, Xia, Ma, Li, & Yang, 2019), to improve the efficiency of city operations (Kang & Qin, 2016).

Regarding various spatial interactions patterns, the detection of those anomalous ones is conducive to revealing the sources of spatial mismatches and further guide the adjustments of unreasonable urban spatial layouts (Liu et al., 2012). Excess commuting is a representative form of anomalous interaction patterns in cities, which is defined as the differences of the actual and the theoretical minimum commuting distances between workplaces and residential areas (Ma & Banister, 2006). Amounts of studies have been carried out to investigate the formation mechanism of excess commuting by taking multiple factors into consideration (Kanaroglou et al., 2015; Leith et al., 2016; Nick & Karolien, 2020). Additionally, some other studies aimed at evaluating the extent of actual commuting excess relative to the optimal status (Zhou & Murphy, 2019). In addition to excess commuting, some studies have been conducted to detect anomalous interaction patterns using spatial statistical approaches. Han, Liu, and Omiecinski (2013) detected anomalous trajectories of mobile objects on the road network by evaluating the direction and distance similarities with others. From the dynamic microscopic perspective, some researchers modeled the moving behavior on single or multiple objects mathematically to recognize anomalous inter-area interactions (Huang, 2015; Yuan, Liu, & Wei, 2017). Shi et al. (2018) designed an approach to identify anomalous patterns from traffic flow data by constructing dynamic neighborhoods (Shi, Deng, et al., 2018). Liu, Wu, et al. (2020) developed a network-constrained clustering method to statistically detect significant source or sink areas. Cai et al. (2020) mined the anomalous regions according to the spatial co-distribution pattern (Cai, Deng, Guo, Xie, & Shekhar, 2020; Cai, Xie, et al., 2020). And others discovered the convergence and divergence patterns of traffic flow over time (Fang, Yang, Xu, Shaw, & Yin, 2017; Huang, Wang, Zhang, Gao, & Schich, 2018). More recently, several novel methods for detecting arbitrarily shaped clusters of OD flows with different scales in flow space are proposed and have excellent capabilities in aggregation flow detection (Shu et al., 2020; Song et al., 2019).

In summary, the objectives of minimum travel time or distances in the classical excess commuting studies may produce inconsistent optimization results to the actual service ranges of urban facilities, especially in large cities. Additionally, other existing studies of anomalous interaction pattern mining mostly aim to detect the significant aggregation and sparsity of crowds or origin-destination flows by comparing with that in spatial adjacent areas (Tao & Thill, 2018; Jeong, Yin, & Wang, 2018). In other words, they are not suitable to detect those anomalous interaction patterns with significant deviations from the optimal status constrained by current urban facility allocations. This type of anomalous interaction pattern is a critical indicator for the spatial mismatch problem and can provide feasible guides for optimizing both the distribution of resident trips in space and the spatial layout of urban facilities. To overcome these problems, this study measures the influences of urban spatial layout on resident trips by quantifying the attraction of sub-areas, and aims to detect anomalous interaction patterns by successively answering the following questions:

  • (1)

    What is the optimal interaction flux between any two urban blocks?

  • (2)

    What are the characteristics of anomalous interaction patterns between the blocks?

  • (3)

    What are the differences between overload- and underload-related anomalous interaction patterns at different scales?

This study has the following three contributions:

  • The proposed anomalous interaction pattern detection approach based on bipartite networks and taxi data provides a new perspective for exploring the possibility of optimizing urban population carrying capacities.

  • A branch and bound (abbreviated as BnB) based augmenting path algorithm weighted by the attractiveness degree is proposed to obtain the optimal interaction fluxes between blocks.

  • The interaction flux adjustment plan is designed by comparing and analyzing two types of anomalous interaction patterns, namely overload and underload interactions, from multiple perspectives, which can provide feasible suggestions for the spatial layout of urban facilities.

The remainder of this paper is structured as follows. Section 2 describes the research framework and the proposed method for anomalous interaction pattern detection based on bipartite networks. The study area and the data used in this research are elaborated in Section 3. Section 4 reports the results of the experiments conducted to analyze the relationship between two types of anomalous interaction patterns. Section 5 presents the conclusions and future research plans.

Section snippets

Overview of the research framework

By coupling the geospatial network with mathematical optimization, a bipartite network-based framework was developed to detect the anomalous interactions between urban subareas through global maximization of urban population carrying capacities. The proposed framework is mainly composed of four steps (Fig. 1). First, a fully connected bipartite network is constructed using traffic analysis zones (TAZs) as the network nodes, so as to quantitatively represent the spatial interactions between

Study area and data description

Wuhan City, located in Central China, is a typical city with a polycentric spatial structure. In this study, we selected the core built-up area of Wuhan as the study area in the experiments (Fig. 3). Moreover, in order to involve the fine-grained information of resident trips, we selected TAZs obtained the division of the study are by road network as the basic spatial unit.

Static data with fundamental geographic information, e.g., the road network structure, points of interest (POI)

Optimal interaction fluxes between TAZs

To obtain the optimal fluxes between TAZs, attractiveness intensities were quantified using the gravity-based attraction model defined in Section 2. Considering the different scales among population densities, POI densities, and road network distances involved in the model, the Z-score value was used to normalize the data used in this study.Z=XX¯σwhere X¯ and σ denote the mean and standard deviation of the target dataset. Moreover, to facilitate the subsequent analysis, the original Z values

Conclusions

Intra-urban anomalous interaction patterns that result in the overload or underload of the areal population play a key role in crowd diversion and urban facility reallocation. To accurately detect anomalous interaction patterns, this study first constructed a bipartite network of TAZs. This network was weighted by the attractiveness between TAZ pairs measured by using a gravity-based attraction model. The optimal interactions between TAZs were obtained by solving the bipartite network maximum

Declaration of Competing Interest

The authors declare that this research was conducted in the absence of any relationships that could be construed as a potential conflict of interest.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (No. 41730105, No. 42071452), the Natural Science Foundation of Hunan Province (No. 2020JJ4696), the Open Fund of Key Laboratory of Urban Land Resources Monitoring and Simulation, Ministry of Natural Resources (No. KF-2019-04-032), the Hunan key research and development project (No. 2018SK2052), the Fundamental Research Funds for the Central Universities of Central South University (No. 2018zzts198), and the Key

References (61)

  • Y. Long et al.

    Combining smart card data and household travel survey to analyze jobs–housing relationships in Beijing

    Computers, Environment and Urban Systems

    (2015)
  • L.J.C. Ma

    Economic reforms, urban spatial restructuring, and planning in China

    Progress in Planning

    (2004)
  • Y. Shi et al.

    Detecting anomalies in spatio-temporal flow data by constructing dynamic neighbourhoods

    Computers, Environment and Urban Systems

    (2018)
  • Y. Shi et al.

    Reconsideration of the methodology for estimation of land population carrying capacity in Shanghai metropolis

    Science of the Total Environment

    (2019)
  • T. Sun et al.

    Employment centers and polycentric spatial development in Chinese cities: A multi-scale analysis

    Cities

    (2020)
  • H. Timmermans

    Spatial choice models

  • Y. Wang et al.

    Analysing the spatial configuration of urban bus networks based on the geospatial network analysis method

    Cities

    (2020)
  • Y.H.D. Wei

    Restructuring for growth in urban China: Transitional institutions, urban development, and spatial transformation

    Habitat International

    (2012)
  • S. Williams et al.

    Ghost cities of China: Identifying urban vacancy through social media data

    Cities

    (2019)
  • L. Wu et al.

    Inferring demographics from human trajectories and geographical context

    Computers, Environment and Urban Systems

    (2019)
  • Y. Yuan et al.

    Exploring inter-country connection in mass media: A case study of China

    Computers, Environment and Urban Systems

    (2017)
  • Y. Yue et al.

    Exploratory calibration of a spatial interaction model using taxi gps trajectories

    Computers, Environment and Urban Systems

    (2012)
  • W. Zhang et al.

    Spatial patterns and determinant factors of population flow networks in China: Analysis on Tencent location big data

    Cities

    (2020)
  • J. Zhou et al.

    Day-to-day variation in excess commuting: An exploratory study of Brisbane, Australia

    Journal of Transport Geography

    (2019)
  • F. Andersson et al.

    Job displacement and the duration of joblessness: The role of spatial mismatch

    (2014)
  • C. Bongiorno et al.

    Core of communities in bipartite networks

    Physical Review E

    (2017)
  • G. Borradaile et al.

    Multiple-source multiple-sink maximum flow in directed planar graphs in near-linear time

    SIAM Journal on Computing

    (2017)
  • J. Cai et al.

    Discovering regions of anomalous spatial co-locations

    International Journal of Geographical Information Science

    (2020)
  • T. Champion

    Urbanization, suburbanization, counterurbanization and reurbanization

  • National Bureau of Statistics of China

    Statistical communique of the People's Republic of China on the 2019 national economic and social development

    People's Daily

    (2020)
  • Cited by (10)

    • Uncovering the spatiotemporal motif patterns in urban mobility networks by non-negative tensor decomposition

      2022, Physica A: Statistical Mechanics and its Applications
      Citation Excerpt :

      Studies in this regard have made commendable contributions in modeling human mobility [4–7], describing urban development [8–10], easing traffic congestion [11,12], and linking mobility characteristics with social economy [13,14]. Characterizing cities as complex systems using complex network tools has been a major research field [8,15–24]. However, existing studies on urban characteristics based on large-scale passenger flow data mostly take first-order single trips as the fundamental units for constructing human mobility networks, or only focus on the mobility of individuals or collectives themselves but neglect their interactions with urban spaces.

    • Detecting spatial flow outliers in the presence of spatial autocorrelation

      2022, Computers, Environment and Urban Systems
      Citation Excerpt :

      Different from the flow clusters, which indicate strong correlations in spatial flow data, SFOs refer to unusual local anomalies that break the spatial stationarity in flow neighborhoods. Discovering SFOs is of substantial importance in revealing anomalous spatial interactions and locating the unusual changes underlying dynamic spatial processes in many real-world applications, such as detecting emerging spatial transmission of infectious disease in spatial epidemiology and identifying unusual human mobility across urban sub-areas in transportation (Liu et al., 2021a). To avoid the methodological issue caused by inappropriate distribution assumptions, we propose a novel statistical interpretation of spatial outliers by taking spatial autocorrelation into account in the null hypothesis, which has not been considered by the current research on spatial outlier detection.

    • Revealing spatiotemporal correlation of urban roads via traffic perturbation simulation

      2022, Sustainable Cities and Society
      Citation Excerpt :

      Urban road networks facilitate the transport of individuals and materials from one place to another, which defines the operational efficiency of an urban transportation system (Duan et al., 2013; Mattinzioli et al., 2020). Moreover, it is critical to identify the correlation of urban road networks by quantifying the correlation and closeness between road segments, as it can promote the development of intelligent transportation systems, including traffic flow prediction (Liu & Wu et al., 2021; Lu et al., 2015; Zeroual et al., 2019), traffic demand estimation (Liu et al., 2016; Liu & Deng et al., 2021), path recommendation (Chen et al., 2012), road network generalization (Yu et al., 2020), and missing data interpolation (Ran et al., 2016; Zou et al., 2012). Traffic flow is an extremely complex dynamic process due to non-linear interactions between several traffic components such as road segments, requiring large-scale information and calculations to understand these interrelationships in detail (Ermagun et al., 2017; Zhang et al., 2020; Li, Batty & Goodchild, 2020).

    View all citing articles on Scopus
    View full text