Content uploaded by Shanqi Zhang
Author content
All content in this area was uploaded by Shanqi Zhang on Sep 06, 2022
Content may be subject to copyright.
Manuscript
EPB: Urban Analytics and City Science
2022, Vol. 0(0) 1–15
© The Author(s) 2022
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/23998083221118570
journals.sagepub.com/home/epb
Towards a 15-minute city: A
network-based evaluation
framework
Shanqi Zhang, Feng Zhen and Yu Kong
School of Architecture and Urban Planning, Nanjing University, China; Key Laboratory of Monitoring, Evaluation and Early
Warning of Territorial Spatial Planning Implementation, Ministry of Natural Resources, China; Provincial Engineering
Laboratory of Smart City Design Simulation & Visualization, China
Tashi Lobsang
School of Architecture and Urban Planning, Yunnan University, China; Provincial Engineering Laboratory of Smart City
Design Simulation & Visualization, China
Sicong Zou
School of Architecture and Urban Planning, Nanjing University, China; Provincial Engineering Laboratory of Smart City
Design Simulation & Visualization, China
Abstract
Developing 15-minute cities, where people can access to living essentials within a 15-minute trip,
has become a global effort. In addition to practical exercise, researchers have paid attention to the
evaluation of 15-minute cities using home-based accessibility approaches. However, existing ap-
proaches do not account for human mobility, an important indicator of how people access and
interact with urban amenities. In this study, we propose a novel network-based framework that
assesses a 15-minute city considering human mobility patterns. We assume that there exists an
optimal mobility network, which would maximize human mobility under the constraints of the
current distribution of amenities. Locations where the provision of urban amenities does not match
local needs are first identified based on the comparison between optimal mobility patterns and their
actual counterparts. Built environment, demographic, and network structure factors that con-
tribute to identified mismatch issues are then examined. The empirical study of Nanjing, China,
suggests that the proposed framework could enable a dynamic evaluation of 15-minute cities and
could provide important insights on policies and intervention strategies of planning and developing
15-minute cities.
Keywords
15-min city, human mobility, network science, accessibility
Corresponding author:
Feng Zhen, School of Architecture and Urban Planning, Nanjing University, 22 Hankou Road, Nanjing 210093, China.
Email: zhenfeng@nju.edu.cn
Introduction
The concept of “N-minute city”(most used “20-minute city”or “15-minute city”)first came out in
Portland, Oregon, in early 2010s (Mackness and White, 2021). It suggests a novel planning scheme
that people can access to living essentials within a 15- or 20-minute trip, ideally by walk (Moreno
et al., 2021). The concept adds to the rhetoric of building sustainable cities, as proximity to essential
services will reduce people’s dependence on vehicles and therefore decrease carbon emissions (Da
Silva et al., 2020). More recently, the emergence of COVID-19 global pandemic has put the
spotlight on how urban planning might have generated inequalities, given that movement re-
strictions may result in varying, often unequitable, accessibility to local service and amenities (Liu
et al., 2021b). The 15-minute city approach has gained salience as it would improve health and
wellbeing through the design of 15-minute walkable neighborhoods and will contribute to an
increased equality by providing ubiquitous access to service and facilities that satisfy people’s basic
living needs (Graells-Garrido et al., 2021;Weng et al., 2019). A range of global cities, including
Melbourne, Ottawa, Paris, and Tempe in Arizona, has adopted the approach and made their own
twists for implementation.
In the context of China, different terminologies such as “15-minute living circle,”“15-minute life
unit”or “15-minute neighborhood”are used to describe the efforts of building a 15-minute city.
“Building a 15-minute living circle”has become a national effort since the State Council puts
forward the requirement in 2016. The concept of “15-minute living circle residential area”was
incorporated into the new edition of “National Standard of Planning and Designing Urban Res-
idential Neighborhoods (GB50180-2018)”. In 2021, Ministry of Natural Resources released
“Spatial Planning Guidance: Community Life Unit”that further clarified the specifications of
planning and building 15-minute living circles. Multiple cities, including Beijing, Shanghai,
Guangzhou, Nanjing, Changsha, Hangzhou, and Wuhan, have carried out local planning policies
for allocating services and facilities within a 15-minute walking zone of local inhabitants.
The increasing acceptance of the 15-minute city concept and extensive global exercise lead to
new research questions with regard to the extent to which essential services are accessible within a
15-minute travel zone. Measuring accessibility of urban amenities and opportunities is not new in
the studies of social and transport-related equity issues (Kwan, 1999;Xing et al., 2020;Yigitcanlar
et al., 2010). Researchers applied home-based accessibility measurements to assess the accessibility
of essential services in various urban contexts against the background of 15-minute city (Da Silva
et al., 2020;Shen et al., 2020). Attention has been paid to developing methods of quantifying living
convenience based on the spatial distribution of facilities (Zhong et al., 2020) and disparities in
accessibility (Shen et al., 2020). While accessibility measurements provide useful insights on the
provision of essential services, they do not account for how facilities are actually used. In reality,
there might exist a divergence between how services can potentially be used based on their
provision and how services are actually used (Frank et al., 2019).
Understanding human mobility patterns has become a central focus in urban planning as it is
informative for developing human-centered policies and essential for making cities more sus-
tainable (Graells-Garrido et al., 2021). In the context of 15-minute city, it is suggested that the
allocation of facilities should account for how people access and use local service and amenities
rather than merely considering population size (Chai et al., 2021). Specifically, scholars have
recommended taking into account people’s diversified needs and behavioral preferences through the
analysis of people’s spatiotemporal behavior (Chai et al., 2021;Mackness and White, 2021). Human
activity at neighborhood level, together with its contributing built environment and socioeconomic
factors, has long been an important research topic in urban studies. Abundant work has focused on
measuring neighborhood vibrancy, based on various indicators including the density of residential
and employment population, mixture of land use, the intensity of human activities, and de facto
2EPB: Urban Analytics and City Science 0(0)
population estimation (Kang, 2020;Xia et al., 2020). Dynamics in population densities or human
activities are considered as being able to reveal the demand for urban services such as health,
transportation, and infrastructure (Ratti et al., 2006). More recently, the availability of large human
mobility datasets makes it possible to monitor how people move within the city and interact with
urban amenities. Using mobile phone signaling data, Graells-Garrido et al. (2021) examined the
difference between human mobility and accessibility of urban amenities and suggested that suf-
ficient education and business amenities are conducive to increased human mobility at the
neighborhood level. Despite the efforts, studies regarding how the planning of a 15-minute city
aligns with human mobility within the city are still limited.
To bridge this research gap, this study proposes a methodological framework for evaluating 15-
minute city based on network science approaches. Network science approaches have been widely
adopted for studying human mobility patterns (Batty, 2012). Particularly, studies have approved the
efficiency of network science methods for identifying urban structures and spatial interaction
patterns (Liu et al., 2015;Lobsang et al., 2021). Some recent work focuses on identifying anomalous
spatial interaction patterns (e.g., excess commuting trips) by comparing optimal and actual travel
patterns (Liu et al., 2021a;Zhou and Murphy, 2019). In this study, a network-based approach is
proposed based on the assumption that there exists an optimal mobility network, which would
maximize human mobility under the constraints of the current distribution of amenities (Liu et al.,
2021a). Two research questions are addressed: (1) how the current provision of urban amenities
aligns with human mobility patterns? (2) What built environment and demographic factors con-
tribute to mismatch between service provision and human mobility? The first research question is
examined based on the comparison between actual mobility patterns, derived from mobile phone
signaling data, and optimal ones, which are modeled based on current amenity and population
distribution. Supply-demand mismatch issues, identified by the differences in actual and optimal
mobility network, are examined. Further, the second question investigates factors associating with
identified issues using Exponential-family Random Graph Models (ERGMs).
Collectively, this study makes two major contributions to current studies and practices of 15-
minute city. From a methodological perspective, this study proposes a network-based approach to
evaluating 15-minute city. This approach differs from mostly used accessibility measurements by
accounting for human mobility patterns, which can indicate how people interact with urban
amenities. As such, the proposed approach could provide a more dynamic perspective to the
understanding and assessment of 15-minute cities. From an empirical perspective, the case study
carried out in Nanjing, China, has implications on policies and intervention strategies for improving
the supply of urban amenities so that they can better meet the needs of local inhabitants. Other cities
could learn from the empirical study for better planning and development of 15-minute cities.
The remainder of this paper is organized as follows. Next section introduces study area and data,
followed by a detailed description of the proposed approach and results of the case study. The paper
concludes with a summary of major contributions, key findings, limitations, and directions for
future work.
Study area and data
Study area
The case study was carried out in the central city of Nanjing, the capital city of Jiangsu province,
China. According to the seventh national population census of 2020, the city has an administrative
area of 6600 km
2
and a total population of 9.3 million. The master plan of Nanjing (2018–2035)
defines a central city of Nanjing where the built up area is most concentrated and population is
densely distributed. The central city consists of four parts: the main city, Xianlin sub-city, Dongshan
Zhang et al. 3
sub-city, and Jiangbei new central city (Figure 1(a)). The study area is discretized into a series of
500-meter grid cells for further analysis.
Data
Mobile phone signaling data. Mobile phone signaling data have been widely used to estimate
population distribution and investigate human mobility patterns (Deville et al., 2014;Gonz´
alez
et al., 2008). In this study, we use mobile phone signal data during June 1
st
to June 30
th
2019 to
estimate working and employment population as well as obtaining actual mobility network (i.e., the
origin-destination (OD) network between home and non-work activity locations). The data were
provided by China Unicom, one of the largest telecommunication operators in China. Due to the
needs of protecting data privacy, the data can only be accessed through a DaaS platform developed
by China Unicom.
The anonymous dataset records the locations of mobile phone users using the position of base
stations. In urban areas, the distance between any two mobile phone base stations is usually within
the range of 50–200 meters (Dong et al., 2015). If an individual stays at one location for more than
30 minutes, the location is considered as an activity point. For each individual, his/her residential
and work locations are recognized according to the following rules: the residential location is the
activity point an individual visits the most during the time period of 9pm to 8am in one consecutive
month; the work location is the activity point an individual visits the most during the time period of
9am to 5pm in one consecutive month. The residential population and employment population of
each grid cell are then calculated based on identified residential and working locations (Figure 1(c)
and (d)).
Point of interest data. Point of interest (POI) data are used to measure the availability of local urban
amenities (Graells-Garrido et al., 2021). The data are collected using the API of Gaode Maps
(https://www.amap.com/), one of the largest Chinese web map service providers. POI is classified
based on the “National Standard of Planning and Designing Urban Residential Neighborhoods
(GB50180-2018),”which suggests that business service, recreational, sports, culture, education,
and healthcare amenities are related to people’s daily activities. The recent published industry
standard “Spatial Planning Guidance: Community Life Unit”further suggests that the planning of
community circle should differentiate service amenities that satisfy people’s basic needs such as
grocery shopping, banking, and mailing, and those may improve the of local people’s sense of
happiness, including shopping center, caf´
e, and others. As such, we further classify business service
POI into two sub-categories of living and business. In addition, transport-related amenities, such as
bus station and subway stations, are closely related to people’s travel. Traveling to these stations
Figure 1. Study area.
4EPB: Urban Analytics and City Science 0(0)
would generate trips around neighborhoods. To make actual mobility and optimal mobility
comparable, transport POI is also included in the study. A summarization of the definition and
counts of POI in each category is included in Table S1 in the Supplementary Material.
Methods
We propose a network-based evaluation framework that consists of three parts (Figure 2). First,
optimal mobility network is estimated based on the spatial distribution of urban amenities and
population using a maximum flow algorithm following Liu et al. (2021a). Second, the actual OD
network is obtained using mobile phone signaling data. Third, the differences between actual OD
network and the optimal one are measured to provide insights on the extent to which human
mobility patterns, as a reflection on the usage patterns of urban amenities, match or do not match the
schemes of urban planning and construction. Built environment, demographic, and network
structure factors that contribute to supply-demand imbalance are identified using ERGM. Supply-
demand imbalance issues are further identified to inform planning and intervention strategies.
Estimating optimal mobility network
The optimal mobility network can be modeled as a multiple-source multiple-sink maximum flow
problem (Liu et al., 2021a). The standard single-source single-sink maximum flow problem aims to
find a solution that enables maximum flow be sent from a given source to a given sink through a
network. In an urban system, there are multiple sources and sinks, indicating where people move in
and out (Liu et al., 2012). The typical solution to multiple-source multiple-sink maximum flow
problem is reducing it to a single problem by adding a virtual super source node Sand a virtual super
sink node T(Ford and Fulkerson, 1962). Virtual node Sconnects to all source nodes uwith edges of
infinite capacity, and virtual node Tconnects to all sink nodes vin the same way.
The problem is formulated by defining a graph G¼ðV,EÞwith Vdenoting all nodes in the graph
and Eall edges, a set sof nodes designated as sources, a set tof nodes designated as sinks, and a
Figure 2. Proposed network-based framework for evaluating 15-minute city.
Zhang et al. 5
capacity function that assigns a finite capacity cðeÞ≥0 to every edge e2E(Nussbaum, 2010). In our
study, cðeÞis defined using
(distij >1:5km →ceij¼0 (1)
distij ≤1:5km →ceij¼Tij (2)
A key to planning 15-minute city is defining the accessible area within the 15 minute time threshold.
In this study, we consider walking as the main travel mode. Assuming an average walking speed of
6 km per hour (Zou et al., 2021), we use 1.5 km as the threshold for the facility to deliver service.
The Euclidean distance is used here rather than road network data considering the computational
complexity. The calculation of the capacities of edges is conditioned by the distance between two
grids an edge connects. If the distance is further than 1.5 km, the capacity equals to zero, as no flows
would be generated (equation (1)).
For two grids that are less than 1.5 km away, the capacity of edges connecting the grids are
calculatedbased on a gravity model. Gravity model has been widelyused to model mobility flows and
is advantageous for its effectiveness (Liu et al., 2014;Schl¨
apfer et al., 2021). A general form of gravity
model is Tij ¼αθiθjfðrijÞ,whereTij denotes the flow between location iand location j;θi,θjare
factors relate to the number of trips starting from ior arriving at j;fðrijÞis the cost effect of traveling
from ito j, usually represented as a function of distance and αis a constant (Barbosa et al., 2018).
Gravity model often assumes that movement between an origin and a destination is proportional
to the population of the two locations (Simini et al., 2012). Other studies suggest that the op-
portunities provided by a destination influence how many people move to it (Drezner, 2019;Liu
et al., 2021a). In this study, we consider the number of trip from a source node ito sink node jis
proportional to how attractive the destination is to the origin and the population at origin and
destination. The attractiveness of node jto node iis quantified by
ωij ¼X
M
m¼1X
N
n¼1
λmnpoimpopn(3)
where λmn is the weight of the mth type of POI preferred by nth type of population groups. Previous
studies have suggested that spatiotemporal mobility patterns vary vastly among different population
groups and that different population groups have diversified needs regarding urban amenities
(Zhang et al., 2021;Zou et al., 2021). Therefore, we consider four population segments, namely,
youth (age<18), young adults (age 19–39), middle-aged adults (age 40–59), and seniors (age >60),
and calculate their preferences of each type of urban amenities based on survey data (see details in
supplementary material S2.1). Accordingly, the capacity of edges are calculated using the following
equations.
Tij ¼8
>
>
>
<
>
>
>
:
poij>0→
ωijpj
dist2
ij
(4)
poij¼0→
pipj
dist2
ij
(5)
where piand pjis the population of node iand node j, respectively; distij refers to the distance
between node iand node j. When node jhas POI, the capacity of the edge connecting node iand
node jis calculated using the modified gravity model that account for the attractiveness of node jto
node i(equation (3)). Otherwise, a conventional gravity model is used (equation (3)). We use 2 as
the distance decay parameter according to previous studies (Drezner, 2019).
6EPB: Urban Analytics and City Science 0(0)
Maximum flow is calculated using a flow function fðeÞthat satisfies the following constraints
8
<
:
0≤fðeÞ≤cðeÞ"e2E(6)
X
e2inðvÞ
fðeÞ¼0"e2V∖S[T(7)
For any given edge, the flow is less than the capacity of the edge (equation (4)). In addition, we
consider all flows in the network follow the path of S→u→v→T, where S,Tare virtual source
and sink nodes and u,vare actual source and sink nodes in the network. Equation (4) ensures that the
number of incoming flow matches that of outgoing flows so that no residual flows left in the network
(Liu et al., 2021a). Among various algorithms for solving maximum flow problem, we adopt Dinic’s
algorithm considering its computational efficiency (Derigs and Meier, 1989). A further introduction
of the pseudo code of the algorithm can be referred to Supplementary Material (S2.2).
Obtaining actual mobility network
As introduced in Data, residential and employment locations of each individual are recognized
using mobile phone signaling data. Commute trips, referring to trips that are either from home to
work or from work to home, are excluded here as employment opportunities are spread out spatially
and can hardly be allocated within a 15-minute radius (Da Silva et al., 2020). For each individual,
non-working activity points within the 1.5 km radius are connected with his/her residential location.
The trips are then aggregated at the grid level. For each pair of grid ða,bÞ, if there exists people
living in grid aand visiting grid b, then a connection is formed between aand b. A directed network
Ga¼ðV,E,WÞis constructed, where Vare grids that are lived or visited by mobile phone users, E
is the edge that connects where people live and visit, and Wis the weight of the edge which is
calculated by the number of times a grid bis visited by people living in grid ain 1 month.
Comparing the two networks
To make actual and optimal networks comparable, the number of flows between any two grids are
normalized using fn¼fofmin
fmaxfmin , where fois the flow volume between any two grids, fmax ,fmin are the
maximum and minimum values of actual or optimal networks. Two networks are compared by
calculating the discrepancy between the two networks using Diff ij ¼Actualij Optimalij, where
Actualij and Optimalij denote the actual flow and optimal flow from grid ito grid j, respectively (Liu
et al., 2021a). Following the approach presented by Liu et al. (2021a), we consider interactions with
difference values below or above two standard deviations to be anomalous (i.e., actual mobility
volume largely deviates from what is expected) as the distribution of the difference values between
actual and optimal mobility networks is close to a normal distribution (Figure S1). We consider
these anomalous interactions as an imbalanced network, which indicates a possible mismatch
between the supply of urban amenities and the demand of local residents. As a comparison,
connections where the volume of actual OD network is close to that of the optimal network are
considered to constitute a balanced network.
Identifying factors associated with supply-demand mismatch. We then identify factors that are asso-
ciated with the formation of the imbalanced network using ERGM (a detailed introduction of
ERGM methods is included in the Supplementary Material S2.3). We focus on demographic and
built environment variables as they are considered as critical determinants of human mobility
(Cervero and Kockelman, 1997). Population density and percentage of senior population are used to
present local demographic characteristics. For built environment variables, we use the wide adopted
Zhang et al. 7
conceptualization of 3Ds (i.e., density, diversity, and design). POI density, the entropy of POI and
the design of road and sidewalk networks are used as key explanatory variables. The definition and
descriptive statistics of selected variables are shown in Supplementary Material S2.4. In addition to
demographic and built environment variables, two endogenous variables of the network, namely,
edges and mutual, are considered in the model. Among them, edges variable is a network statistics
that represents the number of edges in the network (Morris et al., 2008). It is similar to the intercept
of linear regression model. The term mutual reveals the possibility of a pair of nodes having a mutual
relationship. To understand if and how these explanatory variables differently impact the formation
of anomalous interactions versus their counterparts that are close to expected, two models were
developed for imbalanced network and balanced network, respectively.
Results
The characteristics of actual and optimal mobility networks
Actual and optimal mobility networks are obtained using methods described in Methods. For both
networks, higher volumes of flow are concentrated at areas with high population densities (Figure
3). The descriptive statistics of the two networks suggest differences between the networks (Table
1). The number of nodes represents the number of grid cells that have people move in or out. In this
study, this number is influenced by where population and public amenities locate. Actual OD
network has a slightly lower number of nodes, indicating that some grid cells supposed to have flows
of people are missing. The number of edges represents the number of connections existing between
different grids. Optimal network has a smaller number of edges compared to its actual counterpart,
suggesting that the interactions between certain pairs of nodes are absent (Liu et al., 2014). The
density of a network measures the density of intra-node interaction in a network (Lazega et al.,
1995). Influenced by the lower numbers of edges, the optimal network also has a lower network
density.
In a directed graph, node degree is the number of links adjacent to a node. Optimal network has a
slightly higher value of average node degree but a lower network density. A further examination on
Figure 3. Spatial distributions of actual and optimal mobility networks.
8EPB: Urban Analytics and City Science 0(0)
the distribution of node degree shows that actual mobility network has a skewed degree distribution,
while the degree distribution of the optimal network is close to a normal distribution (Figure S2).
This result confirms the assumption that ideally, the planning of a 15-minute city aims to achieve a
more even distribution of urban amenities. People are encouraged to use nearby facilities and as a
result, a more evenly distributed mobility pattern would be realized. The actual mobility network
shows that a considerable amounts of nodes possessing a large number of links in the network (i.e.,
nearly 15% of the nodes have a node degree of 54). Urban places often have different levels of
attraction, and fewer urban places have higher levels of attraction and accordingly attracts more
people (Schl¨
apfer et al., 2021). The identified skewed distribution suggests that grids with high node
degrees have stronger attractiveness compared with others and can potentially be overloaded with
excessive flows of people.
The comparison of two mobility networks
Anomalous interactions are further identified based on the comparison between actual and optimal
mobility networks. Figure 4 shows the spatial distribution and descriptive statistics of the im-
balanced network. Overall, the imbalanced network is mostly distributed at areas with high
population density. In the network, positive and negative connections represent that the volume of
actual mobility is larger than and smaller than that of optimal mobility, respectively. Only 65 nodes
and 56 edges are associated with negative connections. Spatially, these connections are concentrated
in the central business district in the main city. Positive connections relate to 964 nodes and 2531
edges. They are mainly distributed at the center of three sub-cities and the outskirts of the main city.
Table 1. Descriptive statistics of actual and optimal mobility networks.
# of nodes # of edges Density Avg. degree of nodes
Actual mobility network 3364 56,380 0.00498 16.76
Optimal mobility network 3508 30,500 0.00139 17.39
Figure 4. (a) Spatial distributions and descriptive statistics of imbalanced network and (b) street images of
typical places with supply-demand mismatch.
Zhang et al. 9
Both negative and positive networks exist in the central part of the main city but with different
directions, indicating a mismatch between supply and demand. In three sub-cities, positive networks
are identified, suggesting that actual mobility is greater than expected. ERGM was further applied to
identify key factors associated with the mismatch issue. Considering the co-existence of positive
and negative networks (i.e. there are common factors collectively contributing to the emergence of
imbalanced networks), two models were developed for imbalanced and balanced network, re-
spectively (Table 2). Here, the balanced network refers to connections where the volume of actual
mobility network is close to that of optimal mobility network.
For endogenous variables, the variable mutual is significantly positive in both networks. That is,
in these two networks, if there is a connection from node ito node j, it is highly probable that there
exists a connection from node jto node i. For exogenous variables, entropy of POI and road density
are significantly associated with the formation of balanced network. Both variables have negative
effects, suggesting that balanced network is less likely to exist in areas where the diversity of POI
and road density are low. Population, POI, and road densities as well as the diversity of POI are
positively correlated with the formation of imbalanced network. These results suggest that the
mismatch between supply and demand is more likely to happen in population-dense areas, as these
areas are mostly associated with high density and diversity of built environment factors. This is
consistent with Figure 4, which shows that the imbalanced networks are mostly distributed at
residents’concentrated areas (i.e., the center of main cities and new cities). Taking Xinjiekou CBD
district as an example (example 1 in Figure 4), it is the busiest business district in Nanjing and
attracts people from all around Nanjing and even nearby cities. Despite the abundant urban
amenities in the district, some of the amenities, such as business for luxury shopping or high-end
catering, may not accommodate local residents’daily needs. Residents may instead go to amenities
that are not nearby for proper service. As a result, a mismatch between supply and demand is
observed.
Additionally, the proportion of senior population is negatively associated with the formation of
balanced network. The more the senior population, the less likely supply-demand mismatch is
observed. This explains the imbalanced networks identified in the new cities. In the city of Nanjing,
specialized zones, such as university towns, industrial parks, and new city centers, have been
developed in the three sub-cities through mega project development. These mega projects have
greatly boosted local economy and restructured the city spatially and socioeconomically (Qian,
2011). The sub-city attracted significant numbers of young workers and local demographic structure
has been greatly changed. Although a significant number of urban amenities have been arranged,
Table 2. Results for ERGM.
Variable Balanced network Imbalanced network
Endogenous variables Edges 6.83*** 10.57***
Mutual 8.50*** 6.32***
Demographic structure Population density 0.19 1.99***
Percentage of senior population (60+) 0.05 2.06***
Density Density of POI 0.19 2.23***
Diversity Entropy of POI 0.13** 2.05***
Design Road density 0.39*** 0.95***
Sidewalk density 0.15 0.04
AIC 646,446 35,712
BIC 646,561 35,827
Significant level: ***p< 0.001; **p< 0.01.
10 EPB: Urban Analytics and City Science 0(0)
they may not fully accommodate the needs of the increasing population. Additionally, the mismatch
issue is identified near industrial parks, where young workers are concentrated (examples 3, 4 in
Figure 4). These relatively newly built parks are mostly developed for high-tech companies and are
equipped with urban services for workers. However, there might be differences between the needs of
workers and those of local residents. Taking Suning industrial park as an example, it is located on the
border of old central city and the new sub-city. Around the park, there are mixes of new and high-
end commodity residential area primarily for upper-level managers, workers with relatively high
income levels and affordable housing for low-income people. The large discrepancy between
population groups may result in differences in the choice of urban services.
Planning implications
The spatial distribution of imbalanced network suggests that the main city and three sub-cities face
different challenges of achieving a 15-minute city. In the main city, public amenities serve not only
residents nearby but also residents from the entire city. Despite abundant high-level facilities, local
residents or workers’daily needs may not be properly accommodated. The old city is also ex-
periencing transformation resulted from the new urbanization. Dynamic spatial and socioeconomic
restructuring driven by initiatives such as urban revitalization may result in a mismatch between the
arrangement of service facilities and people’s needs. As such, effective and dynamic assessments of
how people interact with urban amenities are necessary to shed light on people’s needs and to guide
the arrangement of public services.
Sub-cities are going through rapid developments and are facing challenges of accommodating
the needs of a changing population. Dongshan sub-city has the longest history of development and
largest population among three sub-cities. The abundance in urban amenities and closeness to the
main city have attracted more people moving to the sub-city. This continuing growth of population
generates more demands and further results in insufficient supplies of amenities. Xianlin sub-city
was first developed as university towns and later attracted more residents in recent years. During the
transformation, amenities used to accommodate needs of university students need to be rearranged.
In the Jiangbei sub-city, a large number of commodity housing, the price of which is relatively cheap
compared to the main city, was developed at the identified area (example 2 in Figure 4) and made the
area a new center of the sub-city. New workers or residents are attracted to the new zone, yet their
size and socioeconomic structure could be different from the master plan. In addition, some people
move from old central city to sub-city for cheaper housing, and they may still work in the old central
city and is inhabited to use urban services in the old city. It is worthwhile to further investigate local
people’s preferences to better arrange service amenities.
Discussion and conclusion
This paper proposes a network-based approach to evaluating 15-minute city from a human mobility
perspective. While traditional accessibility measurements are effective for investigating the equity
and equality of people’s potential access to opportunities including urban amenities, they do not
account for people’s spatiotemporal behavior. The proposed approach builds upon the assumption
that there exists an optimal flow pattern within the city where human mobility could match the
capacity of urban amenities. By measuring the difference between actual and optimal mobility
patterns, useful insights can be obtained regarding whether human mobility patterns within the city
match the ideology of developing a 15-minute city or not. Issues of supply-demand mismatch and
their ascribe factors can also be identified for further improvement.
The proposed approach provides a new perspective to examining the 15-minute city practice.
Currently, the planning of a 15-minute city primarily builds upon a static view of allocating urban
Zhang et al. 11
amenities based on population size (Da Silva et al., 2020). The case study, however, suggests that
this static view could be problematic as the actual situation could be considerably different from the
planning scheme. Identifying spatial mismatch issues that hinder the achievement of 15-minute city
and their ascribe factors will be critical for developing proper intervention strategies. The proposed
approach is also conducive to dynamic optimization of urban amenities and services. In the context
of China, the undergoing rapid urbanization process has spurred meta-development projects, in-
cluding the development of new sub-cities and metropolitan regions. How cities can better ac-
commodate the needs of a changing population has become an important yet challenging topic for
urban researchers and planners. This study presents an attempt that enable a timely evaluation of the
extent to which public amenities meet inhabitants’needs using human mobility datasets. The
evaluation can be carried out regularly (e.g., by year or by season) to monitor the dynamics between
the availability of public amenities and public needs, so that further guidance on optimizing urban
services could be provided.
This study has several limitations that should be addressed in future studies. First, this study uses
1.5 km as a proxy for distance can be reached in 15 minutes by walk. This absolute threshold can be
adjusted by the walkability of the local environment, as walkability can affect the willingness of
people to walk around. Multi-modal transportation, such as bike and public transit, can also be
included in future studies, depending on local planning context. Second, while we removed
commute trips and filtered discretionary trips by distance using mobile phone signaling data, we do
not know accurately the purpose of those trips. There could be noises in the data as people may work
at a location within 15-minute walk of their home location or travel more than 15-minute for
discretionary purposes. A more detailed GPS tracking data or travel-diary data could supplement the
presented study by identifying what kinds of urban amenities people use within the 15-minute
walking zone.
There are a few research directions future work can expand upon. First, results suggest that node
degrees of the optimal mobility network follow a normal distribution pattern. This raises an in-
teresting question regarding what would be the ultimate goal we would like to achieve with the
planning and development of 15-minute cities. For instance, would a 15-minute city mean a more
equitable distribution of urban amenities at the neighborhood level, and accordingly more even
distributions of human mobility patterns? Second, the implementation of a 15-minute may in fact
depends on local contexts. However, this study makes one step further towards the quantitative
evaluation of 15-minute cities. Future studies may account for various policy contexts and more
complicated human activity patterns when estimating the optimal mobility pattern. Third, the
increasing popularity of delivery services has changed how people interact with public amenities.
People can now do grocery shopping online and get grocery delivered to their home or to a nearby
pick-up location. The virtual services may partly substitute traditional on-site services. It is
worthwhile to further investigate the intertwinement of virtual and on-site services and understand
its implications for developing 15-minute cities.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or
publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or pub-
lication of this article: This study was supported by National Nature Science Foundation of China (Grant
No.52008201), Key Laboratory of Monitoring, Evaluation and Early Warning of Territorial Spatial Planning
12 EPB: Urban Analytics and City Science 0(0)
Implementation, Ministry of Natural Resources (Grant No. LMEE-KF2021007) and the Jiangsu Shuangchuang
(Mass Innovation and Entrepreneurship) Talent Program (Grant No. JSSCBS20210046).
ORCID iD
Shanqi Zhang https://orcid.org/0000-0002-0745-879X
Supplemental Material
Supplemental material for this article is available online.
References
Barbosa H, Barthelemy M, Ghoshal G, et al. (2018) Human mobility: models and applications. Physics Reports
734: 1–74. DOI: 10.1016/j.physrep.2018.01.001.
Batty M (2012) Building a science of cities. Cities. DOI: 10.1016/j.cities.2011.11.008.
Cervero R and Kockelman K (1997) Travel demand and the 3Ds: density, diversity, and design. Transportation
Research Part D: Transport and Environment 2(3): 199–219. DOI: 10.1016/S1361-9209(97)00009-6.
Chai Y, Li C and Zhang Y (2021) A new time-geography research framework of community life circle.
Progress in Geography 39(12): 1961–1971.
Da Silva DC, King DA and Lemar S (2020) Accessibility in practice: 20-minute city as a sustainability planning
goal. Sustainability (Switzerland) 12(1): 1–20. DOI: 10.3390/SU12010129.
Derigs BU and Meier W (1989). Implementing Goldberg’s max-flow-algorithm - A computational inves-
tigation. ZOR Methods and Models of Operations Research 33: 383–403.
Deville P, Linard C, Martin S, et al. (2014) Dynamic population mapping using mobile phone data. Proceedings
of the National Academy of Sciences 111(45): 15888–15893. DOI: 10.1073/pnas.1408439111.
Dong H, Wu M, Ding X, et al. (2015) Traffic zone division based on big data from mobile phone base stations.
Transportation Research Part C: Emerging Technologies 58: 278–291. DOI: 10.1016/j.trc.2015.06.007.
Drezner T (2019) Gravity models in competitive facility location. International Series in Operations Research
and Management Science 281: 253–275. DOI: 10.1007/978-3-030-19111-5_11.
Ford LR and Fulkerson DR (1962) Flows in Networks. Princeton University Press.
Frank LD, Mayaud J, Hong A, et al. (2019) Unmet demand for walkable transit-oriented neighborhoods in a
midsized Canadian community: market and planning implications. Journal of Planning Education and
Research. DOI: 10.1177/0739456X19831064.
Gonz´
alez MC, Hidalgo CA and Barab´
asi AL (2008) Understanding individual human mobility patterns. Nature
453(7196): 779–782. DOI: 10.1038/nature06958.
Graells-Garrido E, Serra-Burriel F, Rowe F, et al. (2021) A City of Cities: Measuring How 15-minutes Urban
Accessibility Shapes Human Mobility in Barcelona.1–32 http://arxiv.org/abs/2103.15638.
Kang CD (2020) Effects of the human and built environment on neighborhood vitality: evidence from Seoul,
Korea, using mobile phone data. Journal of Urban Planning and Development 146(4): 05020024. DOI:
10.1061/(asce)up.1943-5444.0000620.
Kwan MP (1999) Gender and individual access to urban opportunities: A study using space–time measures.
Professional Geographer. DOI: 10.1111/0033-0124.00158.
Lazega E, Wasserman S and Faust K (1995) Social network analysis: methods and applications. Revue
Française de Sociologie. DOI: 10.2307/3322457.
Liu B, Deng M, Yang J, et al. (2021a) Detecting anomalous spatial interaction patterns by maximizing urban
population carrying capacity. Computers, Environment and Urban. Systems 87: 101616. doi: 10.1016/j.
compenvurbsys.2021.101616.
Liu J, Gross J and Ha J (2021b) Is travel behaviour an equity issue? Using GPS location data to assess the
effects of income and supermarket availability on travel reduction during the COVID-19 pandemic.
International Journal of Urban Sciences 25(3): 1–20. DOI: 10.1080/12265934.2021.1952890.
Zhang et al. 13
Liu X, Gong L, Gong Y, et al. (2015) Revealing travel patterns and city structure with taxi trip data. Journal of
Transport Geography 43: 78–90. DOI: 10.1016/j.jtrangeo.2015.01.016.
Liu Y, Sui Z, Kang C, et al. (2014) Uncovering patterns of inter-urban trip and spatial interaction from social
media check-in data. PLoS One 9(1): e86026. DOI: 10.1371/journal.pone.0086026.
Liu Y, Wang F, Xiao Y, et al. (2012) Urban land uses and traffic“source-sink areas”: evidence from GPS-
enabled taxi data in Shanghai. Landscape and Urban Planning 106(1): 73–87. DOI: 10.1016/j.
landurbplan.2012.02.012.
Lobsang T, Zhen F, Zhang S, et al. (2021) Methodological framework for understanding urban people flow
from a complex network perspective. Journal of Urban Planning and Development 147(3): 1–13. DOI:
10.1061/(asce)up.1943-5444.0000689.
Mackness BYK and White I (2021) Towards the 20-minute city. Build 183: 71–72.
Moreno C, Allam Z, Chabaud D, et al. (2021) Introducing the “15-minute city”: sustainability, resilience and
place identity in future post-pandemic cities. Smart Cities 4(1): 93–111. DOI: 10.3390/
smartcities4010006.
Morris M, Handcock MS and Hunter DR (2008) Specification of exponential-family random graph models.
Journal of Statistical Software 24(4): 1548–7660. http://statnetproject.org/
Nussbaum Y (2010) Multiple-source multiple-sink maximum flow in planar graphs. Computing Research
Repository.1–10. arxiv:1012.4767 [cs.DM].
Qian Z (2011) Building Hangzhou’s new city center: mega project development and entrepreneurial urban
governance in China. Asian Geographer 28(1): 3–19. DOI: 10.1080/10225706.2011.577977.
Ratti C, Frenchman D, Pulselli RM, et al. (2006) Mobile landscapes: using location data from cell phones for
urban analysis. Environment and Planning B: Planning and Design 33(5): 727–748. DOI: 10.1068/b32047.
Schl¨
apfer M, Dong L, O’Keeffe K, et al. (2021) The universal visitation law of human mobility. Nature
593(7860): 522–527. DOI: 10.1038/s41586-021-03480-9.
Shen G, Zhou L, Wang Z, et al. (2020) Home-based locational accessibility to essential urban services: the case
of wake county, North Carolina, USA. Sustainability (Switzerland) 12(21): 1–22. DOI: 10.3390/
su12219142.
Simini F, Gonz´
alez MC, Maritan A, et al. (2012) A universal model for mobility and migration patterns. Nature
484(7392): 96–100. DOI: 10.1038/nature10856.
Weng M, Ding N, Li J, et al. (2019) The 15-minute walkable neighborhoods: measurement, social inequalities
and implications for building healthy communities in urban China. Journal of Transport and Health
13(129): 259–273. DOI: 10.1016/j.jth.2019.05.005.
Xia C, Yeh AGO and Zhang A (2020) Analyzing spatial relationships between urban land use intensity and
urban vitality at street block level: a case study of five Chinese megacities. Landscape and Urban
Planning 193: 103669. DOI: 10.1016/j.landurbplan.2019.103669.
Xing L, Liu Y, Wang B, et al. (2020) An environmental justice study on spatial access to parks for youth by
using an improved 2SFCA method in Wuhan, China. Cities 96: 102405. DOI: 10.1016/j.cities.2019.
102405.
Yigitcanlar T, Rashid K and Dur F (2010) Sustainable urban and transport development for transportation
disadvantaged: a review. The Open Transportation Journal 4(1): 1–8.
Zhang S, Yang Y, Zhen F, et al. (2021) Understanding the travel behaviors and activity patterns of the
vulnerable population using smart card data: an activity space-based approach. Journal of Transport
Geography 90: 102938. DOI: 10.1016/j.jtrangeo.2020.102938.
Zhong T, Lü G, Zhong X, et al. (2020) Measuring human-scale living convenience through multi-sourced urban
data and a geodesign approach: buildings as analytical units. Sustainability (Switzerland) 12(11): 1–19.
DOI: 10.3390/su12114712.
14 EPB: Urban Analytics and City Science 0(0)
Zhou J and Murphy E (2019) Day-to-day variation in excess commuting: an exploratory study of Brisbane,
Australia. Journal of Transport Geography 74: 223–232. DOI: 10.1016/j.jtrangeo.2018.11.014.
Zou S, Zhang S and Zhen F (2021) Measurement of community daily activity space and influencing factors of
vitality based on residents’spatiotemporal behavior: taking Shazhou and Nanyuan streets in Nanjing as
examples. Progress in Geography 40(4): 580–596.
Author Biographies
Shanqi Zhang is a research associate professor in the School of Architecture and Urban Planning at
Nanjing University and the technical director of Provincial Engineering Laboratory of Smart City
Design Simulation & Visualization, Jiangsu, China. Her main research interests include spatio-
temporal modeling of human mobility, and big data-supported smart planning.
Feng Zhen is a professor in the School of Architecture and Urban Planning at Nanjing University
and the director of Provincial Engineering Laboratory of Smart City Design Simulation & Vi-
sualization, Jiangsu, China. His research expertise include urban geography, urban, urban and
regional planning, and smart cities.
Yu Kong is a PhD student in the School of Architecture and Urban Planning at Nanjing University
and Provincial Engineering Laboratory of Smart City Design Simulation & Visualization, Jiangsu,
China. His main research interests include ICT and mobility, and urban planning.
Lobsang Tashi is a lecturer in the School of Architecture and Urban Planning at Yunnan University
and the adjunct senior researcher of Provincial Engineering Laboratory of Smart City Design
Simulation & Visualization, Jiangsu, China. His main research interests include network science,
human mobility modeling and urban planning.
Sicong Zou is master student in the School of Architecture and Urban Planning at Nanjing
University and Provincial Engineering Laboratory of Smart City Design Simulation & Visuali-
zation, Jiangsu, China. Her main research interests include big data analysis and urban planning.
Zhang et al. 15