Analysis of the spatial characteristics and driving forces of underground consumer service space in Chinese megacities based on multi-source data

Full text

Analysis of the spatial characteristics and driving forces of underground
consumer service space in Chinese megacities based on multi-source data
Yuxiao Tang
a
, Yudi Tang
b,*
a
School of Architecture and Urban Planning, Suzhou University of Science and Technology, Suzhou 215009, PR China
b
School of Minerals and Energy Resources Engineering, The University of New South Wales, Sydney, NSW 2052, Australia
ARTICLE INFO
Keywords:
Underground consumer service space
Spatial characteristic
Driving force
Geographical detector
Chinese megacities
ABSTRACT
Underground consumer service spaces (UCSS) offer new solutions for urban residents’ daily needs, but existing
studies on their distribution and driving forces are often fragmented and overshadowed by research on other
underground spaces, lacking targeted analysis. This study examines UCSS in the central urban areas of seven
representative Chinese megacities. Using spatial analysis methods like kernel density estimation, multi-distance
spatial clustering, and geographical detectors, the spatial characteristics and driving forces of UCSS are analyzed
alongside aboveground consumer service spaces (ACSS). Results show that both ACSS and UCSS exhibit multi-
centered, concentric spatial patterns, though UCSS demonstrates higher spatial aggregation. Unlike other un-
derground public spaces (UPS), UCSS relies more on service industry agglomeration and market factors, while
other UPS are more inuenced by surrounding development intensity. UCSS follows the core principles of central
place theory but deviates from the market-driven patterns typical of ACSS. Socioeconomic conditions and
transportation infrastructure form the foundational basis for UCSS distribution, while service industry agglom-
eration, market dependence, and land development intensity exert more direct inuence. The commercial at-
mosphere and existing underground space development play critical roles in UCSS distribution. Two key spatial
scales for understanding UCSS distribution are the strong inuence zones of shopping malls and metro stations,
and high-density urban areas.
1. Introduction
The development and utilization of urban underground spaces
globally have generated numerous positive external effects (Dong et al.,
2021a; Ma & Peng, 2021; Zargarian et al., 2018), signicantly contrib-
uting to the sustainable development of cities (Dong et al., 2021b; Peng
et al., 2021; Qiao et al., 2019). Initially focused on disaster prevention
and municipal utilities (Chen & Liu, 2011; Makana et al., 2016), un-
derground spaces have gradually evolved to serve public needs,
expanding their functions to include transportation and commerce
(Broere, 2016; Liu et al., 2024). As part of this transformation, Under-
ground Consumer Service Spaces (UCSS) have emerged as a crucial
component in the comprehensive use of underground spaces (Bobylev,
2016a; Chen et al., 2018; Peng & Peng, 2018), offering services like
retail, dining, and entertainment (Besner, 2017; Peng et al., 2020),
thereby providing new options for residents’ daily needs. By integrating
these services into accessible underground locations, UCSS reduces
travel time and enhances convenience for urban residents (Debrock
et al., 2023; Sun & Leng, 2021; Xu & Chen, 2021a). Furthermore, these
spaces alleviate surface-level congestion, improving urban space utili-
zation efciency (Bobylev, 2016b; Fan & Cui, 2023; Li et al., 2024).
However, the layout of UCSS often follows the design and spatial ar-
rangements of other types of underground spaces, lacking more targeted
theoretical guidance, which limits its potential to optimize urban func-
tionality and improve residents’ quality of life.
Research on the spatial characteristics and inuencing factors of
underground spaces at the urban scale has primarily focused on un-
derground public spaces (UPS) (Dong, et al., 2021b; Peng et al., 2019; Xu
& Chen, 2021b), which excludes independently placed functional spaces
such as disaster prevention or municipal utilities (Jia et al., 2024b). UPS
emphasizes accessibility and support for social activities among urban
residents (Mehta, 2014; Schmidt et al., 2011) and includes not only
UCSS but also underground transportation and pedestrian spaces (Cui,
2021). However, UCSS, as a prot-oriented space driven by the private
sector (Ma et al., 2022), has development dynamics that differ signi-
cantly from transportation and pedestrian spaces, which are led by the
* Corresponding author.
E-mail addresses: yuxiaotangup@outlook.com (Y. Tang), yudi.tang@unsw.edu.au (Y. Tang).
Contents lists available at ScienceDirect
Sustainable Cities and Society
journal homepage: www.elsevier.com/locate/scs
https://doi.org/10.1016/j.scs.2024.105924
Received 12 June 2024; Received in revised form 11 October 2024; Accepted 17 October 2024
Sustainable Cities and Society 116 (2024) 105924
Available online 18 October 2024
2210-6707/© 2024 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC license (
http://creativecommons.org/licenses/by-
nc/4.0/ ).

public sector (Dong, et al., 2023b). The existing studies often group
UCSS with other types of UPS, overlooking its unique characteristics as a
consumer space. Traditional consumer space layout theories, such as
central place theory (Christaller, 1966), have been widely applied to
above-ground environments (Jia et al., 2024a). By classifying consumer
spaces into different levels (Ren et al., 2024; Yi et al., 2024), these
theories have optimized the spatial layout of urban consumer services,
becoming an important basis for planning consumer spaces (Lamb,
1985; Sparks et al., 2020; Yan, 2017). However, whether these theories
are applicable to UCSS remains uncertain, as the physical conditions and
development costs of the underground environment limit the exibility
of UCSS distribution. Therefore, independently analyzing the spatial
distribution and inuencing factors of UCSS is crucial, as it not only
expands existing consumer space layout theories but also provides
practical guidance for the appropriate siting and development of UCSS.
To bridge the aforementioned research gap, this study set out to
answer three key questions: (1) Does UCSS have a unique spatial se-
lection logic compared to other types of UPS? (2) What are the spatial
distribution and agglomeration patterns of UCSS, and how do its spatial
distribution and patterns differ from aboveground consumer service
spaces (ACSS)? Do they align with traditional consumer space layout
theories? (3) How do economic, social, and urban functional factors
inuence the spatial agglomeration of UCSS? To explore the growing
agglomeration of UCSS in cities and provide further research support for
their development, the central urban areas of seven representative
megacities in China were selected. Methods such as kernel density
estimation, multi-distance spatial clustering, and geographic detectors
were employed to conduct an in-depth analysis of the spatial charac-
teristics and inuencing factors of UCSS at the urban scale, from a
comparative perspective of both ACSS and UCSS.
2. Literature review
2.1. Central place theory and urban consumer spaces
The distribution patterns and inuencing factors of traditional con-
sumer spaces are core issues in commercial geography. Early studies
primarily used qualitative methods to describe the characteristics of
consumer spaces, leading to the formation of consumer space layout
theories centered around central place theory (Lamb, 1985). This theory
was initially used to explain the spatial distribution of cities, towns, and
villages of different sizes, as well as how they provided goods and ser-
vices to surrounding areas (Giuliano & Small, 1991; Wyckoff, 1989). As
research progressed, scholars found that the fundamental principles of
central place theory were equally applicable to explaining the spatial
structure within cities (Davies, 1959; McMillen, 2001; Potter, 1976). By
emphasizing consumers’ sensitivity to spatial distance across different
levels of commercial services (Christaller, 1966; Malczewski, 2009),
central place theory explains the formation of hierarchical consumer
spaces and commercial districts (Tammiksaar et al., 2018). High-level
central places, such as large shopping malls, provide high-end services
because consumers are willing to travel longer distances for them, while
lower-level central places, such as community shopping centers, meet
residents’ daily needs, with consumers preferring to shop at closer lo-
cations. With advances in technology and changes in business models,
research on consumer space distribution has become increasingly
diverse. Scholars have analyzed urban commercial center systems
(Boratinskii & Osaragi, 2024; Hou & Chen, 2021; Ren et al., 2024) and
consumer space structures (Liu et al., 2015a; Zhong et al., 2014) through
different types of businesses, such as retail (Carter & Potter, 1983; Fang
et al., 2021) and food service (Qin et al., 2014; Tian et al., 2023), further
validating the core principles of central place theory. Moreover, key
factors inuencing the distribution of urban consumer spaces, such as
population density (Applebaum, 1966; Merino & Ramirez-Nafarrate,
2016), economic development (Meijers et al., 2017; Watkins, 2014),
business composition (Paul & Rosenbaum, 2020; Teller & Reutterer,
2008), market demand (Drennan et al., 2011; Pennerstorfer & Penner-
storfer, 2019) and transportation conditions (Chen & Wang, 2022; Melo
& Graham, 2018), have also been conrmed in these studies. In practice,
urban planners over time have used these principles to rationally plan
commercial districts of different levels, avoiding the over-concentration
of resources (Yang et al., 2019; Zhou et al., 2023b), ensuring the
reasonable distribution of consumer services within the city, and
signicantly improving land-use efciency (Jia et al., 2024a; Qiang
et al., 2024).
2.2. The spatial distribution of UCSS
For a long time, research on UCSS has typically been integrated into
the UPS framework, mainly focusing on the physical spaces formed by
the coupling of transportation, pedestrian networks, and various ser-
vices such as dining, retail, and entertainment (Ma et al., 2022; Peng
et al., 2020). Most studies concentrate on the spatial environment, such
as the physical environment (Darbani et al., 2020; Dong et al., 2021;
Qiao et al., 2023) and walkability (Cui et al., 2015; Durmisevic & Sar-
iyildiz, 2001; Gu et al., 2019). Although some studies have expanded to
larger-scale analyses to assess the spatial performance (Jia et al., 2024b;
Ma & Peng, 2023a) and vitality (Li et al., 2024; Xu & Chen, 2023) of
UPS, the focus remains on improving specic micro-level environments
rather than systematically analyzing their siting logic. In recent years, as
research on UPS has deepened, some studies have begun to address the
spatial characteristics and inuencing factors of UPS. For example, Peng
et al. analyzed the distribution of UPS around subway stations in Osaka,
Japan, and found that passenger ow and land prices were the core
inuencing factors (Peng et al., 2019). Dong et al. (2023b) through
studies on metro-led UPS in China, conrmed the impact of economic,
environmental, and development scale factors on UPS distribution. Cui
et al. (2013) demonstrated the impact of economic development, sub-
way systems, and city size on underground pedestrian systems in 19
cities worldwide. Additionally, some research has focused on under-
ground parking spaces, discussing how urban functions and land
development intensity inuence parking space distribution (Dong et al.,
2021b; 2023a). Most existing studies focus on UPS near subway stations,
viewing these spaces as auxiliary facilities of transportation nodes, thus
overlooking UCSS in non-subway station areas. This results in incom-
plete research objects that fail to account for the spatial selection logic of
market-driven UCSS. Furthermore, the public sector’s dominance in
subway stations and pedestrian pathways further weakens UCSS’s
uniqueness in combined studies, reducing the explanatory power of
research ndings on UCSS spatial distribution. Although underground
parking spaces exhibit some self-organizing characteristics, their auxil-
iary function means that their spatial selection logic differs signicantly
from that of UCSS.
UCSS possesses the dual attributes of both underground space and
consumer space. Existing studies on its spatial distribution have largely
focused on the perspective of underground space, often overlooking its
spatial selection logic as a consumer space. This has led to an incomplete
understanding of the spatial agglomeration patterns of UCSS. Although
numerous studies have shown that ACSS in above-ground environments
generally align with traditional consumer space layout theories,
applying these theories to UCSS still requires further exploration. To
ensure the integrity of UCSS as an independent research subject, this
study separates it from the broader UPS framework and systematically
compares the spatial distribution of UCSS and ACSS, revealing the key
differences between them. The results will not only provide new per-
spectives for the existing theoretical framework but also, in the context
of increasing urban space constraints, help enhance the utilization ef-
ciency of UCSS through scientic planning, promote the integration of
above-ground and underground consumer spaces, and contribute to the
construction of a multi-dimensional consumer service network in future
cities.
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
2

3. Study area and data
3.1. Study area
Based on the comprehensive consideration of factors such as the
socio-economic conditions of the cities, the maturity of underground
space development and regional differences, we selects seven repre-
sentative mega-cities in China for case studies: Beijing, Chengdu,
Guangzhou, Shanghai, Shenyang, Wuhan, and Xi’an. The Hu Huanyong
Line will remain a demographic, economic, and environmental dividing
line in China for the coming decades (Zhang et al., 2020). As shown in
Fig. 1a, the selected cities are all located east of the Hu Huanyong Line,
where urbanization levels are higher (Chen et al., 2019), and the char-
acteristics of a high-density built environment are prominent. Table 1
provides a detailed overview of the selected case cities in terms of
socio-economic development and the maturity of underground space
development. Data on population density, per capita GDP, urbanization
rate, and the proportion of the tertiary industry were sourced from the
respective 2023 statistical yearbooks of seven major Chinese cities.
Metro operating line length data and data on underground space
development were obtained from authoritative reports. The case cities
exhibit high levels of economic development, which grants them the
economic strength to lead in the utilization of underground space.
Moreover, the tertiary industry in these cities accounts for more than 58
% of their economies, indicating a continued demand for underground
space in the service sector. In terms of population density, the case cities
have population densities far exceeding the national average of 147
people per square kilometer, with urbanization rates reaching around 80
%, highlighting the severe land-use pressure and making underground
space an essential means of expanding urban development. Regarding
underground space development, these cities are leading in both the
intensity of underground space development and the length of opera-
tional metro lines. Finally, the case cities have each implemented a se-
ries of regulations and policies to guide the development and utilization
of underground space, demonstrating their consistent efforts in pro-
moting underground space development. Table 2 presents recent pol-
icies and plans related to underground space development in the case
cities, highlighting a strong similarity in principles, objectives, and
measures across these cities.
Additionally, all the case cities are located in different geographical
regions, representing the urban development of their respective areas,
and they exhibit certain differences in terms of economy, culture,
climate, and terrain. These differences also lead to certain variations in
their approaches to underground space development. According to
recent editions of the China Underground Space Development Blue Book
(China Academy of Engineering Strategic Consulting Center, 2024), the
selected case cities consistently rank at the top in terms of comprehen-
sive underground space development levels, thereby signicantly
reecting the development level of underground spaces in China.
To achieve similar spatial scales in each city, the central urban areas
enclosed by the city’s ring roads were selected as the specic study
areas. As shown in Fig. 1, the central urban areas are the most densely
populated regions in these case cities, with the most developed con-
sumer service functions and the highest concentration of underground
space development. Furthermore, the main parts of each city’s subway
network are within the central urban areas, making these regions ideal
for obtaining clearer and more accurate research conclusions. Table 3
provides specic details on the location and scale of the selected case
cities. In summary, diversity, variability, and comparability were fully
considered in the selection of study areas in this paper, allowing for a
more comprehensive analysis of the spatial characteristics and driving
forces of UCSS.
3.2. Data collection
Compared to traditional data, big data is characterized by high
precision, strong real-time capability, large volume, and extensive
coverage, enabling more rened urban studies (Liu et al., 2015b). We
primarily collect three types of urban data, which are used to measure
the development of UCSS, conduct driving force analysis, and create
results maps. The specic details are shown in Table 4. These data were
mostly collected in September 2023, thus accurately reecting the real
situation at a specic point in time. All data underwent preprocessing
steps such as data cleaning, data validation, standardization, and spatial
projection coordinate calibration. All spatial data were projected to the
corresponding UTM zone in ArcGIS.
3.2.1. Consumer service space data
Dianping, established in 2003, is China’s leading local lifestyle in-
formation and transaction platform and the country’s earliest third-
party consumer review website. User reviews of shops, processed by
the website’s system, are available for reference by all potential con-
sumers. Additionally, the website arranges for dedicated personnel to
verify various types of information, such as shop locations. Due to its
short information update cycle, it can more comprehensively reect the
status of urban consumer service facilities. Existing research has
conrmed the applicability of Point of Interest (POI) data in urban
functional zoning (Hu & Han, 2019; Yang et al., 2019) and spatial
pattern analysis (Jiao et al., 2024; Wan et al., 2024). A considerable
number of studies on underground spaces have already used POI data as
a primary data source (Dong et al., 2023; 2023b; Ma & Peng, 2023a).
Compared to POI data from other online maps like Amap, Dianping’s
POI data is more specialized in local consumer service facilities and has
been widely used in the study of urban consumer service spaces (Qin
et al., 2014; Yan, 2017; Zhao et al., 2024).
Dianping POI data was used to characterize the spatial distribution of
Fig. 1. Location of study area.
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
3

urban consumer service spaces in this paper. The data was collected
from Dianping as of September 2023, yielding approximately 3.32
million records for the selected case cities. Each Dianping POI data entry
includes information such as ID, name, category, number of reviews,
rating, address, and latitude and longitude. As shown in Fig. 2, the raw
data is divided into 16 categories, with the vast majority belonging to
the commercial services industry, while the remaining part belongs to
the public services sector but also has strong consumer attributes. The
Dianping POI data preprocessing procedure is as follows: rst, the raw
data were deduplicated using the pandas library in Python based on a
unique ID. Then, spatial datasets were generated in ArcGIS by adding XY
data according to the latitude and longitude elds, and Dianping POI
data located within the study area were extracted. Finally, we removed
POIs that do not belong to the consumer service industry, primarily a
small number of transportation facility. After preprocessing the raw
data, wildcard searches were used in ArcGIS to lter out facilities
located underground from all life service facilities. The ltering method
is as follows: consumer service facilities with address elds containing
terms such as ’Negative ’, ’Underground’, ’LG’, ’lg’, ’B1
′
, ’B2
′
, ’B3
′
, ’b1
′
,
’b2
′
, ’b3, etc., were selected. Invalid elds, such as cases where ’B’ or ’b’
does not indicate a oor level, were removed. Finally, the results were
checked and conrmed. After obtaining the underground consumer
service facilities dataset, these entries were removed from the original
dataset to create the aboveground consumer service facilities dataset.
The number of aboveground and underground consumer service facil-
ities obtained in the study areas of all case cities is shown in Table 5,
along with the calculated underground facility density (the number of
Table 1
Description of selected megacities (Data Source: 2023 Statistical Yearbooks of Each City, 2022 Urban Rail Transit Annual Statistics and Analysis Report, and 2023
China Urban Underground Space Development Blue Book).
Indicator Beijing Chengdu Guangzhou Shanghai Shenyang Wuhan Xi’an
Geographical Region North Southwest South East Northeast Central Northwest
City Status Municipality Provincial
Capital
Provincial
Capital
Municipality Provincial
Capital
Provincial
Capital
Provincial
Capital
Population Density (people/km
2
) 1331 1484 2588 3905 711 1603 1287
Per Capita GDP (10,000 RMB) 19.03 9.81 15.36 17.99 8.43 13.78 8.88
Urbanization Rate ( %) 87.60 % 79.90 % 86.50 % 89.30 % 85.00 % 84.70 % 79.60 %
Proportion of Tertiary Industry ( %) 83.90 % 66.40 % 71.40 % 74.10 % 58.10 % 61.90 % 61.70 %
Development Intensity of Underground Space in
Built-up Areas (10,000 m
2
/km
2
)
7.48 5.3 7.06 10.86 4.42 7.93 3.86
Metro Operating Line Length (km) 722.08 518.54 519.08 795.37 114.07 460.84 272.12
Number of Regulations and Policies in Underground
Space Development
17 12 8 24 8 10 9
Table 2
Key recent policies and plans governing underground space development in
selected Chinese cities (Source: Policies and plans sourced from respective
municipal government publications).
City Key Policies and Plans Year Development principles
Beijing Beijing Negative List for
Underground Space Use
(2022)
2023 Improve urban functions,
regulate development content
Underground Space
Planning and Design
Guidelines
2020 Promote integrated use,
ensure safety and protection
Guiding Opinions on
Utilizing Underground
Space
2018 Encourage commercial
services, enhance
convenience
Chengdu Chengdu Underground
Space Management
Measures
2021 Regulate development,
prioritize public functions
Opinions on Encouraging
Underground Space Use
(Trial)
2020 Simplify processes, prioritize
public functions, offer
nancial support
Chengdu Central Urban
Area Underground Space
Plan
2015 Ensure public facility
construction, promote
integration
Guangzhou Guangzhou Underground
Space Plan
2022 Ensure safety, promote
integration
Guangzhou Underground
Space Development
Measures
2019 Promote integrated use,
enhance efciency
Shanghai Shanghai Urban Renewal
Regulations
2021 Promote underground-
overground coordination,
protect historical areas
Shanghai Underground
Space Development
Regulations
2020 Promote integrated use,
ensure public facility
prioritization
Shenyang Shenyang Underground
Space Development
Management Law
2022 Regulate development,
promote integrated use
Wuhan Wuhan Underground
Space Management
Provisions
2022 Regulate development,
promote integrated use,
prioritize public services
Xi’an Xi’an Underground Space
Management Measures
2018 Promote development,
enhance public services
Xi’an Three-Year Action
Plan for Underground
Space
2018 Optimize public welfare,
accelerate parking lot
construction
Table 3
Description of selected study area.
City Description of study area Scale
(km
2
)
Beijing Area enclosed by fth ring expressway 667.77
Chengdu Area enclosed by fourth ring expressway 541.26
Guangzhou Area enclosed by the administrative boundary and ring
expressway
716.57
Shanghai Area enclosed by outer ring road 663.31
Shenyang Area enclosed by ring expressway 455.12
Wuhan Area enclosed by third ring expressway 525.07
Xi’an Area enclosed by ring expressway 458.50
Table 4
Basic information of research data.
Classication Data Data Source Collection
Time
Consumer Service
Industry Data
POI Data Dianping Sep. 2023
Driving Force Data Housing Price Data Lianjia.com Sep. 2023
LandScan
Population Data
Oak Ridge National
Laboratory (ORNL)
Jul. 2023
POI Data Amap Sep. 2023
Road Network
Data
Open Street Map Sep. 2023
Building Outline
Data
Amap Sep. 2023
Basic Geography
Data
Administrative
Data
Baidu Map Sep. 2023
Green Space Data Baidu Map Sep. 2023
Water System Data Baidu Map Sep. 2023
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
4

underground facilities divided by the study area) and the underground
facility ratio (the number of underground facilities divided by the total
number of facilities). Their spatial distribution is depicted in Fig. 3.
As shown in Fig. 4, the recorded timestamps of POIs in UCSS illus-
trate the development trajectory of UCSS in the case cities. Since the
1980s, China’s underground space has entered an expansion stage (Chen
et al., 2018), with UCSS emerging alongside the rapid development of
metro-led underground transportation spaces. The coverage period of
the dataset reects the expansion of UCSS, indicating that its scale has
greatly surpassed the initial stage.
3.2.2. Driving force analysis data
The data used to characterize the driving forces mainly fall into three
categories: socioeconomic data, urban function data, and urban
morphology data. These data include both raster and vector formats.
Socioeconomic data includes housing price and population. Housing
prices are often used as an indicator of a region’s economic conditions in
studies on consumer space and UPS distribution (Jia et al., 2024b; Peng
et al., 2019), as they closely correlate with economic prosperity and
residents’ income levels. Higher housing prices generally reect stron-
ger purchasing power, which in turn inuences the location selection of
consumer services. The housing price data is sourced from Lianjia.com,
Fig. 2. Dianping POI categories.
Table 5
Number and distribution of facilities in consumer services across the central urban areas of case cities.
City Beijing Chengdu Guangzhou Shanghai Shenyang Wuhan Xi’an
Aboveground facilities count 221,275 278,714 309,957 284,155 157,200 220,394 231,472
Underground facilities count 13,515 6124 8052 14,585 2276 5090 5350
Underground facility density (units/km
2
) 667.77 541.26 716.57 663.31 455.12 525.07 458.50
Underground facility ratio ( %) 5.76 % 2.15 % 2.53 % 4.88 % 1.43 % 2.26 % 2.26 %
Fig. 3. Spatial distribution of consumer services POIs in the case city.
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
5

one of China’s leading real estate service platforms. Lianjia.com’s
housing price data is characterized by its wide coverage, fast updates,
and high transparency, and it has been widely used to reect urban
housing price levels (Li et al., 2019; 2021a). Over 110,000 residential
community records for the case cities were obtained through the Lianjia.
com data interface, with key elds including name, address, average
price, and number of properties for sale, among others. LandScan pop-
ulation data is used to represent the population distribution of the case
city. The data is sourced from ORNL and is considered the most accurate
and reliable global population dynamics analysis database worldwide. It
is widely used in various urban studies related to population distribution
(He et al., 2024; Kyaw et al., 2023). Considering that the overall eco-
nomic and social development of the selected case city is relatively
stable, the latest LandScan population data released in July 2023, which
reects the population distribution as of 2022, is still timely enough to
meet the research needs.
POI data, representing various urban functions, were obtained from
the Amap platform, totaling 6.7 million records for the case cities. The
data were thoroughly veried, and no further deduplication was
necessary. Amap’s POI data are divided into 23 major categories, 15 of
which are highly relevant to urban functions. These include shopping
services, consumer services, dining services, companies and enterprises,
transportation facilities services, government agencies and social orga-
nizations, accommodation services, educational and cultural services,
healthcare services, sports and recreational services, business and resi-
dential services, automotive services, public facilities, nancial and in-
surance services, and scenic spots. Each of these categories also includes
subcategories and sub-subcategories, with POI data located within the
study area extracted, and the proportions of each part shown in Table 6.
The table reveals signicant differences in the types and distribution of
POIs across cities. For instance, Guangzhou has the highest proportion of
companies and enterprises, while Beijing shows higher percentages of
transportation and government services. Wuhan and Xi’an have rela-
tively larger shares of educational and cultural services, reecting the
functional differences among these cities. Using wildcard searches, we
extracted the relevant POI data based on the driving force factors
selected for the study, with the corresponding relationships shown in
Table 7.
Urban morphology data mainly includes road networks and building
footprints. Using the application interface provided by Amap, the
complete data was obtained. The total length of road networks in the
selected case cities is 47,312 km, and the total building footprint area is
731 km
2
.
3.2.3. Basic geographic data
Additionally, three types of basic geographic data are involved in this
paper: administrative boundary data, urban water system data, and
urban green space data. These data were all obtained from the
Fig. 4. The recording time of underground consumer service POIs.
Table 6
Proportional distribution of POIs by category in the central urban areas of case cities.
Category Beijing Chengdu Guangzhou Shanghai Shenyang Wuhan Xi’an
Shopping Services 19.46 % 24.74 % 28.34 % 21.22 % 29.85 % 28.21 % 26.93 %
Consumer Services 16.09 % 18.65 % 16.28 % 16.93 % 16.56 % 17.12 % 17.90 %
Dining Services 13.22 % 16.69 % 14.69 % 14.78 % 16.04 % 15.79 % 16.31 %
Companies and Enterprises 10.00 % 9.13 % 12.54 % 10.67 % 6.09 % 7.21 % 7.32 %
Transportation Facilities Services 9.81 % 5.65 % 5.02 % 7.74 % 5.84 % 5.45 % 5.49 %
Government Agencies and Social Organizations 6.86 % 2.60 % 3.56 % 5.45 % 3.35 % 3.93 % 2.70 %
Automotive Services 2.40 % 2.98 % 2.56 % 2.16 % 4.11 % 3.25 % 3.46 %
Accommodation Services 2.42 % 5.02 % 3.53 % 2.81 % 2.78 % 3.45 % 5.68 %
Educational and Cultural Services 5.00 % 2.98 % 3.21 % 3.59 % 3.87 % 3.63 % 3.31 %
Healthcare Services 2.96 % 3.94 % 2.75 % 2.45 % 5.15 % 3.70 % 3.79 %
Sports and Recreational Services 3.11 % 2.60 % 2.30 % 3.17 % 2.60 % 3.20 % 2.39 %
Business and Residential Services 3.60 % 2.95 % 2.80 % 4.47 % 1.81 % 2.65 % 2.53 %
Public Facilities 2.23 % 0.74 % 0.72 % 2.16 % 0.53 % 0.61 % 0.64 %
Financial and Insurance Services 2.02 % 1.05 % 1.08 % 1.85 % 1.15 % 1.22 % 1.15 %
Scenic Spots 0.83 % 0.28 % 0.63 % 0.54 % 0.28 % 0.58 % 0.40 %
Total count 373,712 391,891 520,478 428,935 249,292 311,487 320,832
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
6

application interface provided by Baidu Maps. The administrative
boundary data serve as the source of spatial analysis units, with a total of
696 subdistricts in the case cities. The urban water system and green
space data are primarily used for creating result maps.
4. Methodology
Multi-source data and various spatial analysis methods are utilized to
examine the spatial distribution characteristics of UCSS in Chinese
megacities, and the driving forces behind its development are further
analyzed. As shown in Fig. 5, the research framework is divided into four
parts: data preparation, data processing, spatial characteristics analysis,
and driving force analysis. First, indicators of ACSS and UCSS and
related driving force analysis data are collected. These data mainly
include POI data, basic geographic data, socioeconomic data, and urban
morphology data. Second, the attributes of various geographic data are
spatially analyzed to derive the indicators for measuring consumer
service spaces and the driving force factor indicators. Then, multiple
methods are employed to analyze the spatial characteristics of ACSS and
UCSS. Finally, a comparative analysis method is used to conduct the
driving force analysis for a total of three dependent variable indicators
for both ACSS and UCSS.
4.1. Selection of indicators for UCSS
The development level of UCSS is measured using two indicators:
underground consumer service density (UCSD) to represent the absolute
development level of UCSS, and underground consumer service ratio
(UCSR) to represent the relative development level of UCSS. UCSD
directly measures the scale of UCSS; however, this scale does not fully
reect the degree of underground penetration of the service industry in
the area. UCSR, on the other hand, indicates the extent to which con-
sumer service spaces penetrate underground within a specic spatial
unit. Therefore, using both indicators together provides a more
comprehensive reection of the development status of UCSS.
When analyzing the driving forces of spatial elements at the urban
scale, subdistrict units and grid units are two commonly used basic
analysis units. As observed in Fig. 3, compared to ACSS, UCSS exhibits
signicant spatial gaps across all case cities, indicating a strong spatial
discontinuity. If grid units are used for analysis, there could be situations
in urban core areas where adjacent grids have similar levels of various
driving factors but vastly different levels of UCSS development. This
would greatly reduce the explanatory power of the analysis model.
Subdistrict-level units typically have a larger spatial scale and greater
Table 7
Driving force dataset and corresponding POI categories.
Driving Force Dataset POI Data Category
Metro Stations Subway stations in the Transportation Facilities category
Bus Stops Bus stops in the Transportation Facilities category
Public Service
Facilities
Educational and Cultural Services, Healthcare Services,
Government Agencies and Social Organizations, and
Sports and Recreational Services (sports venues)
Commercial Service
Facilities
Dining Services, Shopping Services, Financial and
Insurance Services, Automotive Services, Consumer
Services, Accommodation Services, Sports and
Recreational Services (excluding sports venues)
Residential Areas Residential areas in the Business and Residential Services
category
Business Ofce
Facilities
Business facilities in the Business and Residential Services
category
Underground Facilities All facilities located underground, retrieved using
wildcard searches
Fig. 5. Workow diagram.
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
7

self-contained socioeconomic organization, which allows them to more
effectively capture distribution characteristics within concentrated
areas. Furthermore, more focus is placed on the comparison between
UCSS and ACSS rather than on comparisons across different analytical
units when analyzing the driving forces of UCSS. Therefore, the sub-
districts were selected as the basic analysis units in this study. The
calculation formulas for the two types of UCSS development indicators
for each subdistrict unit are as follows:
ACSD =Numa/A (1)
UCSD =Numb/A (2)
UCSR =Numb/(Numa+Numb)(3)
where Num
a
and Num
b
represent the number of aboveground and un-
derground consumer service facilities, respectively, and A is the area of
the subdistrict unit.
4.2. Selection of driving force indicators
Underground consumer service is a type of urban consumer services
industry located in underground spaces, characterized by two basic at-
tributes: underground space and consumer economy. Therefore, when
selecting the driving forces of UCSS, it is necessary to consider both the
factors inuencing the layout of underground spaces and the factors
inuencing the layout of traditional commercial spaces.
Currently, research on the driving forces of UCSS is relatively scarce,
and a comprehensive system of driving force studies has yet to be
established. Therefore, the selection of driving force indicators for UCSS
mainly references existing studies that focus on the entire underground
spaces or other categories of underground spaces. Given the very high
construction costs of underground spaces, economic conditions are a
crucial factor in the development and utilization of underground spaces
(Chen et al., 2018). Related studies also indicate a positive correlation
between population distribution and the scale of underground spaces
(Chen et al., 2022; Zhao et al., 2016), making socioeconomic indicators
very important. The consumer services industry caters to the daily lives
of urban residents, requiring highly convenient transportation. Thus,
public transportation systems dominated by subways and buses are
equally critical (Dong et al., 2023b). Existing research shows that un-
derground spaces tend to be located in high-density built environments
(Bobylev, 2016b), making land development intensity, road network
density, and similar factors important indicators as well (Peng et al.,
2019). As a subcategory of underground spaces, the development in-
tensity of other underground spaces will obviously affect the scale of
UCSS.
The driving force indicators for traditional comsumer spaces overlap
with those for underground spaces in aspects such as socioeconomic
conditions and transportation (Fang et al., 2021; Wang et al., 2018), so
these indicators will not be reiterated here. Aggregation is one of the
important principles of urban economic development (Feng et al.,
2023), so public service spaces and commercial service spaces related to
consumer services have a signicant impact on UCSS. Additionally,
because consumer service spaces are highly dependent on the consumer
market (Daniel & Hernandez, 2024), and residential and ofce areas
(Zhou et al., 2023a) can largely reect the spatial distribution of the
consumer population, these factors also need to be considered.
In summary, it is posited that the driving force indicators should
include ve components: socioeconomic conditions, transportation
status, service industry aggregation, market dependency, and land
development intensity. A total of 11 representative factors were
selected, as shown in Table 8. Socioeconomic conditions are represented
by housing price and population density, reecting the overall devel-
opment level of the subdistricts. In terms of transportation conditions,
the density of metro stations, bus stops, and road networks were selected
to represent different types of transportation conditions. The density of
public service facilities and commercial service facilities were used to
reect the aggregation of the commercial atmosphere. Market de-
pendency is represented by the density of ofce facilities and residential
areas. Finally, the oor area ratio and the density of existing under-
ground facilities were used to measure the development intensity of the
subdistricts.
Additionally, due to the low spatial density of certain data, such as
metro stations located near the boundaries of two subdistricts, which
signicantly affect both subdistricts but are counted only once in sta-
tistics, kernel density estimation is used for population and metro station
indicators, following existing research (Li et al., 2021b). The kernel
density values were calculated using a 1500 m search radius and sum-
med using a 200m*200 m grid, replacing the original data for the
calculation of driving forces.
4.3. Spatial analysis
4.3.1. Kernel density estimation
The kernel density estimation method can intuitively show the
clustering or dispersion distribution characteristics of spatial point
datasets, effectively reecting the distance decay effect of geographical
elements’ spatial distribution (Okabe et al., 2009). It has been widely
used in the study of the spatial distribution of geographical elements
(Qin et al., 2014; Yu et al., 2015). In this paper, kernel density analysis is
conducted on the distribution of ACSS and UCSS in the central urban
areas of the seven case cities. The natural breaks classication method is
Table 8
Driving factors of UCSS.
Indicator
Dimension
Indicator Code Description
Socioeconomic
Conditions
Housing Price X1 Ratio of the total unit price of
houses for sale within the
subdistrict to the total number
of houses for sale
Population Density X2 Ratio of total population
kernel density value within
the subdistrict to the
subdistrict area
Transportation
Conditions
Metro Station
Kernel Density
X3 Ratio of total metro station
kernel density value within
the subdistrict to the
subdistrict area
Bus Stop Density X4 Ratio of the number of bus
stops within the subdistrict to
the subdistrict area
Road Network
Density
X5 Ratio of total road network
length within the subdistrict
to the subdistrict area
Service Industry
Aggregation
Public Service
Facility Density
X6 Ratio of the number of public
service facilities within the
subdistrict to the subdistrict
area
Commercial
Service Facility
Density
X7 Ratio of the number of
commercial service facilities
within the subdistrict to the
subdistrict area
Market
Dependency
Residential Area
Density
X8 Ratio of the number of
residential areas within the
subdistrict to the subdistrict
area
Ofce Facility
Density
X9 Ratio of the number of ofce
facilities within the subdistrict
to the subdistrict area
Land
Development
Intensity
Floor Area Ratio X10 Ratio of total building area
within the subdistrict to the
subdistrict area
Underground
Facility Density
X11 Ratio of the number of
underground facilities within
the subdistrict to the
subdistrict area
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
8

used to layer and color the analysis results, and the visualized results are
presented in a three-dimensional manner.
4.3.2. Nearest neighbor analysis
Nearest neighbor analysis determines the spatial clustering degree of
point features by analyzing the ratio of the observed nearest neighbor
distance to the expected nearest neighbor distance, known as the
Nearest Neighbor Index (NNI) (Wu et al., 2016). It has been widely used
in the study of the spatial clustering degree of geographical elements
(Shi et al., 2019; Xu et al., 2024). When NNI is less than 1, the point
dataset exhibits a clustered distribution, and the smaller the NNI, the
higher the spatial clustering degree. When NNI equals 1, the point
dataset shows a random distribution. When NNI is greater than 1, the
point dataset shows a dispersed distribution. The calculation formula is:
NNI =d0/de(4)
do=1
N
N
i=1
di(5)
de=1
2
N
A
(6)
Where: d
i
is the distance from a specic point i to its nearest neigh-
boring point, d
o
is the average of the nearest neighbor distances d
i
for all points in the dataset, also known as the observed mean nearest
neighbor distance, d
e
is the expected nearest neighbour distance, the
theoretical distance assuming points are randomly distributed, N is the
number of points in the dataset, and A is the area of the study region.
4.3.3. Multi-distance spatial clustering
Nearest neighbor analysis can determine the overall spatial clus-
tering characteristics of point datasets but cannot assess the clustering
characteristics at different spatial scales (Hao et al., 2018). The distri-
bution state of point datasets at a global scale does not necessarily match
their distribution at a local scale. Multi-distance spatial clustering, also
known as Ripley’s K function, is an analytical method for point data
patterns that assesses the degree of clustering of spatial point datasets at
various distances. This analysis ultimately reveals the clustering or
dispersion characteristics of point datasets at different observational
scales (Ge et al., 2021; Self et al., 2023). The calculation formulas are
(Ripley, 1977):
K(d) = A
n
i=1
n
i=1
i∕=j
δdij
n∗n(7)
L(d) = 
K(d)
π
−d(8)
where A is the area of the study region, d
ij
is the distance between spatial
objects i and j within the study region, d is the spatial scale, δ(d
ij
) is the
indicator function, and n is the number of spatial objects within the
study region. L(d) is a linear transformation of the square root of K(d).
When L(d) equals 0, it indicates a random distribution of point features;
when L(d) is greater than 0, it indicates a clustered distribution of point
features; when L(d) is less than 0, it indicates a dispersed distribution of
point features.
4.3.4. Spatial autocorrelation
The aforementioned spatial statistical methods are directly applied
to point datasets. Due to the self-organizing nature of urban societies,
spatial distribution of geographical elements is often analyzed using
subdistricts as spatial units. Therefore, spatial autocorrelation is intro-
duced to further analyze the clustering characteristics of UCSS. Spatial
autocorrelation analysis includes global spatial autocorrelation and
local spatial autocorrelation, using Moran’s I and Local Moran’s I sta-
tistics respectively (Anselin, 2010; Wrigley, 1982). The calculation for-
mulas are as follows:
I=nn
i=1n
j=1wij(xi−x)xj−x
n
i=1n
j=1wijn
i=1(xi−x)2(9)
Ii=xi−x
σ
2
n
j=1
wijxj−x2(10)
where I represents the global Moran’s I statistic, I
i
represents the local
Moran’s I statistic, n is the number of observations, w
ij
is the spatial
weight between elements i and j, and x
i
is the observed value of the i-th
element. The spatial autocorrelation tools in ArcGIS are used to calculate
Moran’s I and to further identify local autocorrelation patterns,
including high-high, low-low, low-high, and high-low patterns.
4.3.5. Geographical detector
The geographical detector is a set of statistical methods used to
detect spatial differentiation and reveal its driving forces (Wang et al.,
2016). It has been successfully applied in various elds, including urban
studies (Wang et al., 2010; 2021; Yang et al., 2024). One of its unique
advantages is its ability to detect and identify the interaction effects of
two factors on the dependent variable. The geographical detector is used
to analyze the driving forces of UCSSs. The calculation formula is as
follows:
qx=1−1
N
σ
2
L
h=1
Nh
σ
2
h(11)
where q
X
is the inuence of factor X on the density of commercial fa-
cilities, with a value range of [0,1]. The larger the q value, the greater
the inuence of factor X on the density of consumer service spaces. L is
the stratication of the two types of consumer service space density Y or
inuencing factor X; Nh and N are the number of units in layer h and the
entire area, respectively;
σ
2
h
and
σ
2
are the variances of Y values in layer
h and the entire area, respectively.
Subdistrict units are used to conduct detection analysis on the three
dependent variables: ACSD, UCSD, and UCSR. The stratication of fac-
tors is mainly determined by the parameter optimization method (Wang
et al., 2016). Various classication methods such as equal intervals and
natural breaks were tested, and the natural breaks method was ulti-
mately adopted. All inuencing factors were stratied into seven layers.
5. Results
5.1. Spatial characteristics of UCSS
5.1.1. Spatial distribution patterns
Using kernel density estimation to analyze the spatial distribution
patterns of ACSS and UCSS, the search radius was set to 800 m to ac-
count for both overall and micro scales. The results are displayed in
three dimensions in Figs. 6 and 7. The heights of the aboveground and
underground kernel densities cannot be directly compared but represent
the relative heights of the kernel density values within each respective
category. Taking the central urban area of Shanghai as an example, it
can be seen that the UCSS forms a multi-centered distribution pattern.
High-density centers are formed in areas such as East Nanjing Road,
Zhongshan Park, Maotai Road, and Shangcheng Road. Medium-density
centers are formed in areas like Huamu Road, Jinke Road, and Wuzhong
Road. Low-density centers are formed in areas like Wanda Plaza. Similar
multi-centered spatial distribution characteristics are observed in the
central urban areas of other six cities.
Comparing the spatial distribution patterns of UCSS and ACSS, it can
be observed that the overall distribution patterns of the two types of
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
9

consumer service spaces tend to be consistent, with both types exhibit-
ing a concentric decline from their centers. However, there are certain
differences between the two. To enable a clear and accurate comparison
of the spatial distribution patterns between UCSS and ACSS, the Min-
Max normalization method was applied to the kernel densities of both
UCSS and ACSS, using the central urban area of Shanghai as a case study.
The same prole line was employed for the analysis, with the line
carefully positioned to pass through the peak density points of both
UCSS and ACSS as closely as possible. The results are presented in Fig. 8.
Firstly, it is easily noticeable that the distribution area of UCSSs is
signicantly smaller than that of ACSSs in all case cities, indicating that
UCSS have a weaker ability to spread to surrounding areas. Secondly,
there is a displacement of the two types of consumer service spaces at the
micro scale. East Nanjing Road is the main center for ACSSs, with its
kernel density value being much higher than other areas. However, in
the underground part, East Nanjing Road loses this advantage and be-
comes a regular center. Similar cases include areas like Xujiahui and
Shangcheng Road, among others. In contrast, the importance of UCSSs
on Wujiaochang, Maotai Road, Zhongshan Park, and Jinke Road is
enhanced compared to their aboveground counterparts. Similar char-
acteristics can be observed in the central urban areas of other six case
cities.
5.1.2. Spatial clustering characteristics
The nearest neighbor index was used to quantitatively analyze the
spatial clustering characteristics of ACSS and UCSS. To ensure consis-
tency in the spatial scale of the analysis for both types of data, the default
analysis area for ACSS in the case cities was used as the common study
area for both ACSS and UCSS. The analysis results for all seven cities
rejected the hypothesis of random distribution with very low z-values
and small p-values. The nearest neighbor index (NNI) calculation results
for the point datasets are shown in Table 9. The nearest neighbor index
for consumer service spaces in all cities is less than 1, indicating a sig-
nicant clustering state for both types of consumer service spaces in the
central urban areas of all case cities. Furthermore, we can observe that
the nearest neighbor index for UCSS is lower than that for ACSSs in the
central urban areas of all case cities. Since the further the nearest
neighbor index deviates from 1 (in the negative direction), the stronger
the clustering characteristic of the dataset. Therefore, the overall spatial
clustering of UCSSs is considerably higher compared to ACSS.
Using the multi-distance spatial clustering analysis method, we
observed the spatial clustering characteristics of the two types of con-
sumer service spaces at different scales. The analysis results for the
aboveground and underground parts are shown in Fig. 9 and Fig. 10,
respectively. In the gures, the horizontal axis represents the observa-
tion scale, and the vertical axis represents the L(d) value, which in-
dicates the degree of clustering. The L(d) values for both types of
consumer service spaces in the central urban areas of all case cities are
consistently greater than 0, indicating that both aboveground and un-
derground consumer service spaces exhibit clustered distribution pat-
terns across all observation scales. For ACSSs, the maximum clustering
occurs within the 6–8 km range. For UCSS, the maximum clustering
occurs within the 2–8 km range. Except for Guangzhou, the UCSS in the
central urban areas of other case cities reach their maximum clustering
at smaller observation distances. In terms of clustering intensity, the
maximum clustering of UCSS is higher than that of ACSSs in the central
Fig. 6. Kernel density of consumer service spaces in the central urban area of Shanghai.
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
10

urban areas of Beijing, Guangzhou, Shanghai, and Xi’an. In the central
urban area of other three cities, the maximum clustering of UCSSs is
slightly lower than that of ACSS. Overall, UCSSs achieve higher
clustering at smaller observation scales compared to ACSS.
To further explore the differences in clustering between the two
types of consumer service spaces in the central urban areas of case cities,
Fig. 7. Kernel density of consumer service spaces in the central urban areas of six other cities.
Fig. 8. ACSS & UCSS kernel density prole in Shanghai.
Table 9
NNI of urban consumer service space.
Beijing Chengdu Guangzhou Shanghai Shenyang Wuhan Xi’an
ACSS 0.39 0.35 0.34 0.38 0.35 0.32 0.35
UCSS 0.24 0.21 0.23 0.20 0.19 0.19 0.22
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
11

we used the simple curve operation tool in Origin software to calculate
the difference between the L(d) value curves of the two types of con-
sumer service spaces (underground L(d) value minus aboveground L(d)
value). The resulting new curve is shown in Fig. 11, where the horizontal
axis represents the observation scale, and the vertical axis represents the
difference in clustering degree (L(d)) between the two types of consumer
service spaces. Due to the higher density of underground space layout
around metro stations, each central urban area of case city reaches the
rst extreme point within the 270 m–1000 m observation range. Sub-
sequently, the clustering degree difference decreases until it reaches
zero and eventually becomes negative. Here, it can be observed that the
central urban areas of Guangzhou and Shanghai do not follow this
pattern; their UCSS maintains a higher clustering degree than ACSS
across the entire observation scale.
5.2. Spatial characteristics of UCSR
5.2.1. Spatial distribution patterns
UCSR was calculated for each subdistrict unit and classied using the
natural breaks method, as shown in Fig. 12. It can be observed that areas
with relatively high UCSR fall into two categories: those located in
central regions and those in peripheral regions. In the central urban area
of Beijing, both central and northern regions have areas with high UCSR.
In the central urban area of Chengdu, high UCSR areas are found in both
Fig. 9. Multi-distance spatial clustering analysis results of ACSS.
Fig. 10. Multi-distance spatial clustering analysis results of UCSS.
Fig. 11. Difference of concentration degree between UCSS and ACSS.
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
12

the central and southern regions. The central urban areas of Shanghai,
Shenyang, Wuhan, and Xi’an all exhibit characteristics of having high
points in both central and peripheral areas. Only the central urban area
of Guangzhou does not have particularly high UCSR areas in its pe-
ripheral regions. There is a noticeable variation in the maximum UCSR
values across the subdistrict units of the seven case cities. Both the
central urban areas of Beijing and Shanghai have maximum UCSR values
exceeding 17 %, while the central urban areas of Chengdu and
Guangzhou reach 13.1 % and 11 % respectively. The maximum UCSR
values in the the central urban areas of other three cities are relatively
lower, all below 10 %.
5.2.2. Spatial correlation
First, a global spatial autocorrelation analysis was conducted on the
UCSR of the central urban areas in seven case cities. The results are
shown in Table 10. Chengdu, Shenyang, Wuhan, and Xi’an did not
exhibit signicant spatial correlation, with Shenyang and Wuhan having
negative Moran’s I values, indicating that the UCSR in these cities is
randomly distributed spatially. Beijing, Guangzhou, and Shanghai
showed signicant spatial correlation, with Moran’s I values around
0.15, indicating that the UCSR in these three cities is spatially clustered.
Further analysis of the local spatial autocorrelation results, as shown
in Fig. 13, reveals the following patterns. High-high clusters are mostly
located in the central regions, while low-low clusters are primarily found
in the peripheral regions. In Xi’an, a signicant portion of high-high
clusters is located in the southern peripheral areas. Each case city has
low-high clusters, mainly appearing around high-high clusters and at the
edges of the central urban areas. High-low clusters are distributed more
randomly. In the central urban areas of Beijing, Wuhan, and Shanghai,
they are located near the central regions, while in Chengdu and Xi’an,
they appear in the peripheral regions. Guangzhou and Shenyang do not
have high-low clusters.
5.3. Driving force analysis
5.3.1. Factor detection
Fig. 14 shows the factor detection results for the three dependent
variables, ACSD, UCSD, and UCSR, including p-values and q-statistics,
which represent signicance and explanatory power, respectively. For
the analysis of ACSD, except for housing prices (X1) in the central urban
areas of ve cities and bus stop density (X4) in the central urban area of
Xi’an, which have excessively high p-values, the detection results for all
other factors in the case cities are statistically signicant (p < 0.05).
Regarding q-statistics, the explanatory power of the latter six factors is
generally higher than that of the rst ve factors, with commercial
service facility density (X7) being the strongest explanatory factor in the
central urban areas of all case cities. There are some differences in the
explanatory power of each indicator between the case cities, but the
overall trend is similar. This indicates that the selected factors can
adequately explain the distribution characteristics of ACSS.
For the analysis of UCSD, the majority of factors are statistically
signicant. However, there are relatively more non-signicant in-
dicators in the central urban areas of Wuhan, Shenyang, and Xi’an,
mainly concentrated on housing prices (X1), bus stop density (X4), and
residential area density (X8). Regarding q-statistics, the explanatory
power of almost all inuencing factors is lower for UCSD than for ACSD,
except for the density of underground facilities (X11). Commercial ser-
vice facility density (X7) and underground facility density (X11) are the
two factors with the highest explanatory power, corroborating that
UCSSs are indeed most inuenced by existing underground space con-
struction and traditional commercial spaces. There is considerable
variation in the explanatory power of different factors across the central
urban areas of case cities. For instance, the q-statistic for commercial
service facility density (X7) is 0.814 in Chengdu but only 0.379 in
Wuhan. Overall, the UCSD factors exhibit characteristics similar to those
of ACSD, with the latter six indicators related to service industry ag-
gregation, market dependency, and land development intensity showing
Fig. 12. Spatial distribution pattern of UCSR.
Table 10
Spatial autocorrelation analysis of UCSR.
Beijing Chengdu Guangzhou Shanghai Shenyang Wuhan Xi’an
Moran’s I 0.154 0.004 0.124 0.162 -0.029 -0.013 0.070
p-value 0.000 0.668 0.000 0.000 0.685 0.951 0.174
z-score 4.339 0.429 3.983 4.054 -0.405 -0.062 1.360
status Clustered Random Clustered Clustered Random Random Random
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
13

higher explanatory power.
For the analysis of UCSR, the majority of factors are statistically
signicant. However, compared to UCSD, the signicance of the factors
is noticeably reduced, and the q-statistics for UCSR are also signicantly
lower than those for UCSD. Housing prices (X1) and population density
(X2) are not signicant in around half of the central urban areas in case
cities, indicating that socioeconomic conditions do not adequately
explain the distribution characteristics of UCSR. The density of under-
ground facilities (X11) has the highest explanatory power, indicating
that there is still a certain correlation between the relative level and the
absolute scale of UCSS. Other factors have lower explanatory power
compared to underground facility density (X11) and vary between cities.
For instance, the central urban areas of Guangzhou and Xi’an are
signicantly more inuenced by metro station kernel density (X3)
compared to other case cities, while Wuhan is signicantly less inu-
enced by road network density (X5) than other cities.
Overall, service industry aggregation, market dependency, and land
development intensity are common driving forces for both ACSD and
UCSD. This suggests that UCSS and ACSS share similar spatial selection
logic to some extent. However, UCSR does not synchronize with the
other two dependent variable indicators in terms of service industry
aggregation, market dependency, and land development intensity. This
indicates that UCSR has a relatively unique spatial selection logic to
some extent.
5.3.2. Factor interaction detection
An interaction detector was used to further analyze the combined
effect of any two inuencing factors on the spatial distribution patterns
of UCSS. As shown in Figs. 15 and 16, the values on the diagonal
represent the original q-statistics of single factors, while the values in the
lower half of the diagonal represent the q-statistics of pairwise in-
teractions. The upper half of the diagonal indicates the type of interac-
tion between two factors. Both UCSD and UCSR exhibit two types of
interactions: enhanced bivariate (EB) and nonlinear enhancement (EN).
UCSD shows more EB interactions, while UCSR shows more EN in-
teractions. It can be observed that the explanatory power of factor in-
teractions for both UCSD and UCSR signicantly increased in the central
urban areas of all case cities, indicating that the combined explanatory
power of two factors is markedly enhanced, and there are no mutually
independent factors.
In terms of UCSD, the interaction between socioeconomic factors
such as housing prices (X1) and population density (X2) with multiple
factors has signicantly improved explanatory power in the central
urban areas of several case cities. Notably, there was a substantial in-
crease in explanatory power for interactions such as Beijing’s housing
prices (X1) with bus stop density (X4) and public service facility density
(X6), as well as housing prices (X1) with commercial service facility
density (X7) in Shenyang, Wuhan, and Xi’an. Regarding other in-
dicators, in the central urban area of Guangzhou, the q-statistic for the
Fig. 13. Local spatial autocorrelation analysis of UCSR.
Fig. 14. Factor detection of ACSD, UCSD and UCSR.
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
14

interaction between bus stop density (X4) and commercial service fa-
cility density (X7) reaches 0.89. In the central urban area of Shenyang,
while the q-statistic for bus stop density (X4) is only 0.27 and for ofce
facility density (X9) only 0.47, their combined q-statistic rises to 0.95. In
the central urban area of Wuhan, the interaction among bus stop density
(X4), road network density (X5), and commercial service facility density
(X7) demonstrates an explanatory power exceeding 0.95. Similarly, in
Wuhan, residential area density (X8) and oor area ratio (X10) also
exhibit signicant interaction effects.
The improvement in the explanatory power of the q-statistics for the
interaction of UCSR inuencing factors is generally lower than that of
UCSD. The enhancement in explanatory power for socioeconomic fac-
tors after interaction is quite signicant, especially in interaction with
transportation conditions and service industry aggregation, which is
similar to UCSD. In the central urban areas of Beijing, Shanghai, and
Xi’an, the interaction between residential area density (X8) and ofce
facility density (X9) shows a noticeable increase, indicating a high
correlation between UCSR and market dependency. In the central urban
area of Xi’an, the interaction between metro station density (X3) and
commercial service facility density (X7) results in a q-statistic of 0.78.
Additionally, the interaction between road network density (X5) and
both commercial service facility density (X7) and residential area den-
sity (X8) results in q-statistics exceeding 0.74. In the central urban area
of Shenyang, the interaction between bus stop density (X4) and ofce
facility density (X9) results in a q-statistic of 0.8.
Combining the results from both factor detection and interaction
detection, the factors inuencing UCSS development can be divided into
two categories: foundational factors and dominant factors. Foundational
factors show relatively low explanatory power when analyzed inde-
pendently, but their explanatory power increases signicantly when
Fig. 15. Interaction detection of UCSD.
Fig. 16. Interaction detection of UCSR.
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
15

interacting with other factors. Dominant factors, on the other hand,
exhibit relatively high explanatory power even in individual detection,
and their explanatory strength further improves with interaction. For
example, in the UCSD analysis results, the factors in the socioeconomic
conditions and transportation conditions dimensions generally demon-
strate lower explanatory power in independent analysis but see a sig-
nicant increase in explanatory strength (more than 0.2) when
interacting with other factors. These factors are classied as founda-
tional factors. Most of the factors within the service industry agglom-
eration, market dependency, and land development intensity
dimensions show relatively higher explanatory power in individual
detection and continue to perform well in interactions. However, resi-
dential area density (X8), which is part of the market dependency
dimension, behaves differently, exhibiting characteristics more aligned
with foundational factors due to its lower explanatory power in the in-
dependent analysis. In interaction detection, its explanatory power
signicantly improves in interaction detection. As a result, residential
area density (X8) is treated as a foundational factor in this context, while
the remaining factors in the service industry agglomeration, market
dependency, and land development intensity dimensions are classied
as dominant factors.
6. Discussion
6.1. Understanding the spatial characteristics of UCSS development
6.1.1. Key scales of UCSS distribution
A comparative analysis of the spatial characteristics of ACSS and
UCSS reveals that both types of consumer service spaces exhibit multi-
centered, concentric spatial distribution patterns, with clear hierarchi-
cal and agglomeration characteristics in the central urban areas of case
cities, consistent with existing research on the distribution of consumer
service spaces (Tian et al., 2023; Yan, 2017). However, there are also
certain differences in the spatial distribution of ACSS and UCSS. UCSS
generally has a higher degree of spatial agglomeration, and at the micro
level, there is a certain degree of spatial misalignment between the two.
We suggest that two key spatial scales can explain these differences: the
strong radiation range of shopping malls and subway stations, and the
scope of high-density built-up urban areas.
Figs. 6 and 7 illustrate that the underground oors of shopping malls
and subway stations are the two primary carriers of UCSS, and they
largely overlap. To achieve higher efciency, large underground com-
plexes are typically constructed at the intersections of commercial dis-
tricts and major subway lines (Peng et al., 2019). Due to the high cost of
underground space development, the extent to which UCSS can expand
outward, whether within shopping malls or subway stations, is signi-
cantly limited. So, what is the extent of this constrained spatial scale? As
shown in Fig. 11, the difference in the degree of agglomeration between
UCSS and ACSS initially increases and then decreases, with the peak
value occurring at distances of 270 m–1000 m in the central urban areas
of case cities. This range can be considered the strong radiation zone of
shopping malls and subway stations. In southern Chinese cities such as
Chengdu and Guangzhou, this scale exceeds 800 m, while in other cities,
it is generally less than 450 m, which is consistent with the scale (400
m–800 m radius from subway stations) identied in previous studies on
the surrounding areas of underground spaces in cities like Shanghai
(Dong et al., 2023; Ma et al., 2022).
The density of shopping malls and subway stations is higher in city
centers, which gives UCSS a stronger preference for central distribution.
Thus, the size of high-density built-up areas is another important scale
for understanding UCSS. Outside of high-density agglomeration zones,
the distribution density of ACSS decreases, but its relative scale remains
substantial. In contrast, as the number of shopping malls and subway
stations in the urban periphery of the central urban areas sharply de-
creases, the distribution density of UCSS also drops dramatically. As
shown in Fig.11, as the observation scale continues to expand, the gap in
agglomeration between UCSS and ACSS becomes smaller, and eventu-
ally, the agglomeration of ACSS surpasses that of UCSS. This pattern has
been observed in the central urban areas of all cities except Guangzhou
and Shanghai. It is expected that if the observation distance continues to
increase, similar patterns will emerge in these two cities as well.
6.1.2. Spatial mismatch between UCSD and UCSR
Fig. 17 shows the subdistrict unit statistics for UCSD, revealing that
high-value units for UCSD are fewer and more concentrated in the
central regions compared to UCSR. Table 11 presents the bivariate
Moran’s I results for UCSD and UCSR, indicating that the spatial cor-
relation between UCSD and UCSR across the central urban areas of even
case cities has a Moran’s I value of less than 0.25, suggesting overall low
correlation. The central urban areas of Beijing and Shanghai have the
highest correlations, with values of 0.244 and 0.202 respectively,
showing the smallest spatial differences among the seven cities. The
central urban areas of Guangzhou and Xi’an follow, while Chengdu,
Shenyang, and Wuhan have Moran’s I values below 0.1, indicating the
largest spatial differences between UCSD and UCSR in these cities. The
correlation between the density and ratio of underground parking spaces
generally exceeds 0.4 (Dong et al., 2021b; 2023a), indicating that the
spatial difference between UCSD and UCSR is greater. Unlike UCSS, the
main carriers of underground parking spaces include not only com-
mercial areas but also a signicant number of residential areas, which
may explain the higher correlation between parking density and parking
ratio.
Although both ACSS and UCSS tend to be distributed in central re-
gions, UCSS still lags far behind ACSS in terms of scale. This means that
while UCSS may be large in the central regions, it remains relatively
small compared to the overall size of ACSS. Conversely, in peripheral
regions, although UCSS is smaller in scale, the base of ACSS is even
smaller, leading to a relatively high degree of undergroundization of
consumer services in some peripheral regions. This may be due to the
challenges posed by the existing built environment (Deng et al., 2024;
Lin et al., 2021), making it more difcult to expand underground spaces
in central regions, whereas in peripheral regions, it is easier to construct
new or expand existing underground spaces for consumer services.
6.1.3. Factors inuencing UCSS distribution
A comparative analysis of the driving forces of ACSS and UCSS across
ve dimensions is conducted in this study. The driving forces of ACSS
align closely with previous studies on the consumer service industry (Jia
et al., 2024a; Tian et al., 2023; Wang et al., 2021), while UCSS exhibits
similar development patterns but with certain differences. Both foun-
dational and dominant factors demonstrate higher explanatory power
for ACSS distribution compared to UCSS. Consistently, dominant factors
exhibit greater independent explanatory power than foundational fac-
tors for both ACSS and UCSS. However, for ACSS, residential area
density (X8) emerges as a dominant factor. This suggests that the cor-
relation between UCSS and residential areas is signicantly weaker than
that of ACSS, indicating that UCSS is rarely designed to serve residential
communities. Socioeconomic factors reect the economic strength,
consumption capacity, and demand scale of a region, making them
necessary conditions for promoting UCSS development, consistent with
existing research on UPS (Chen et al., 2022; Dong et al., 2023b; Peng
et al., 2019). Although subway stations are one of the major carriers of
UCSS, transportation conditions, including subway station density, do
not have high explanatory power for UCSS on their own. However, the
interaction between transportation factors and service industry
agglomeration has a high explanatory power in the central urban areas
of all case cities, indicating that UCSS tends to concentrate around
subway stations within commercial areas, suggesting that UCSS should
be planned in subway station zones with signicant commercial land use
(Ma et al., 2022; Peng et al., 2019). Unlike underground parking spaces
(Dong et al., 2021b), the distribution of business ofce facilities has a
much greater impact on UCSS than residential areas, as underground
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
16

parking is common in residential areas, whereas UCSS is rarely found in
communities. In the central urban areas of all case cities, the density of
underground facilities emerges as the most important inuencing factor,
with the intensity of other types of UPS signicantly impacting the
spatial distribution of UCSS (Dong et al., 2023a; Peng et al., 2020; Xu &
Chen, 2021b). Compared to existing studies that integrate UCSS within
UPS inuencing factors (Dong et al., 2023b; Peng et al., 2019), foun-
dational factors show lower explanatory power for UCSS. However,
except for oor area ratio (X10), the dominant factors exhibit higher
explanatory power. This suggests that the distribution of UCSS is more
reliant on service industry aggregation and market factors, while other
types of UPS are more signicantly affected by development intensity of
the surrounding environment. In summary, there are two types of for-
mation patterns for UCSS: (1) In urban central regions, UCSS forms
around subway stations near commercial districts, serving as a numer-
ical supplement to ACSS. (2) In urban peripheral regions, UCSS relies on
subway stations to form smaller-scale clusters, complementing the
functions of consumer services in suburban areas.
Previous studies have identied the main driving forces behind the
development of UPS, including housing prices, population, and land
development intensity (Chen et al., 2022; Cui & Nelson, 2019; Xu &
Chen, 2021b). However, there has been a lack of horizontal comparison
across different cities. Table 5, which examines UCSD and UCSR, reveals
differences in the overall development levels of UCSS among various
case cities. The distribution patterns of UCSS are inuenced by a com-
bination of factors. The interaction detection in Fig. 15 indicates that the
central urban area of Shanghai exhibits the most balanced performance
in terms of inuencing factors; apart from the density of underground
facilities (X11), no pair of factors has an interaction explanatory power
exceeding 0.75. This may be related to the relatively high maturity of
UCSS development in Shanghai. The development of UCSS in the central
urban areas of Beijing, Chengdu, and Xi’an is more reliant on their
commercial environments, with smaller UCSS scales around subway
stations located in non-commercial areas. In the central urban areas of
Guangzhou and Wuhan, UCSS is more commonly used for residential
entertainment and leisure rather than business ofces, which may be
related to the composition of UCSS business types in these cities. The
development of UCSS in the central urban area of Shenyang is still in a
relatively early stage, with most UCSS located in areas with good
foundational conditions and close proximity to the city’s CBD.
6.1.4. UCSS spatial distribution in central place theory
According to central place theory, the spatial layout of consumer
spaces should exhibit two key characteristics (Malczewski, 2009; Yan,
2017): hierarchy and agglomeration. Clearly, UCSS, like ACSS, dem-
onstrates both of these features across the central urban areas of all case
cities. Taking the central urban area of Shanghai as an example, the
standardized kernel density values of clustering centers in Fig. 8 are
divided into four levels: 0–0.3, 0.3–0.6, 0.6–0.8, and 0.8–1. ACSS forms
a hierarchical sequence of "2, 4, 13, 40," while UCSS forms a sequence of
"1, 7, 25, 52." Central Place Theory suggests that, under market princi-
ples, centers should follow a k =3 sequence (1, 2, 6, 18), and ACSS
largely conforms to this pattern, which is consistent with previous
studies on above-ground environments (Qin et al., 2014; Tammiksaar
et al., 2018; Tian et al., 2023). However, UCSS tends to follow a k =4
(transportation-based) and k =7 (administrative-based) sequence,
indicating that UCSS more closely aligns with the layout patterns gov-
erned by transportation and administrative principles. Additionally, the
location of the centers has shifted. For example, in central Shanghai, the
primary centers for ACSS are located around East Nanjing Road and
Shangcheng Road, while UCSS centers have shifted northward to
Wujiaochang, deviating from the original theoretical predictions.
Furthermore, UCSS’s spatial distribution surpasses the triangular diffu-
sion pattern predicted by central place theory (Christaller, 1966;
Fig. 17. Spatial distribution pattern of UCSD.
Table 11
Spatial autocorrelation analysis of UCSD & UCSR.
Beijing Chengdu Guangzhou Shanghai Shenyang Wuhan Xi’an
UCSD 0.430 0.321 0.237 0.505 0.037 0.193 0.390
UCSR 0.154 0.004 0.124 0.162 -0.029 -0.013 0.070
UCSD &UCSR 0.244 0.085 0.163 0.202 0.017 0.024 0.184
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
17

Wyckoff, 1989), exhibiting more diverse micro-level spatial distribution
patterns. This phenomenon has also been observed in recent studies of
ACSS (Jia et al., 2024a; Liu et al., 2015a; Zhong et al., 2014), reecting
the growing complexity of factors (Chen & Wang, 2022; Meijers et al.,
2017; Pennerstorfer & Pennerstorfer, 2019) inuencing consumer space
distribution as cities evolve.
Although the spatial distribution of UCSS is more inuenced by
factors related to service industry agglomeration and market depen-
dence according to the driving force analysis, the impact of these two
dimensions is signicantly weaker compared to ACSS. As a result, the
relative importance of transportation and administrative factors be-
comes more prominent in UCSS distribution. However, socioeconomic
and transportation factors do not exhibit as direct an effect, which
conrms that UCSS spatial distribution reects a complex interplay of
multidimensional inuences. Overall, the ndings on UCSS support the
core principles of central place theory, maintaining both hierarchical
and agglomeration characteristics. However, the spatial distribution of
UCSS challenges the assumption that ACSS typically adheres to the k =3
market principle, revealing a stronger inuence of transportation and
administrative factors, which also leads to a shift in the hierarchical
status of agglomeration centers. Moreover, the diversity in UCSS’s
micro-level distribution raises new questions about the applicability of
central place theory at the micro- level.
6.2. Planning and policy implications
As China’s urbanization stabilizes, the conict between population
growth and limited land resources becomes increasingly evident, posi-
tioning UCSS as a critical component for the future development of
consumer services. This study provides valuable insights into the spatial
distribution, inuencing factors, and theoretical underpinnings of UCSS,
offering practical guidance for urban planning and policymaking.
In China, the spatial layout of UPS is heavily inuenced by Transit-
Oriented Development (TOD) principles (Calthorpe, 1993; Cervero
et al., 2009), with compact, mixed-use developments centered around
metro stations (Dong et al., 2022). However, as seen in the central urban
areas of multiple case cities, the impact of development intensity on
UCSS is diminishing, while the inuence of urban functionality is
becoming more prominent. Therefore, when planning UCSS, it is
essential to consider not only the development intensity of the sur-
rounding area but also the fundamental service-oriented nature of UCSS.
Given the agglomerative nature of consumer services, effectively inte-
grating UCSS with existing ACSS is crucial for success (Admiraal &
Cornaro, 2016; Peng & Peng, 2018). UCSS and ACSS can complement
each other in terms of service types and target demographics, with ACSS
typically providing higher-end services while UCSS offers more acces-
sible, everyday options. The future should involve implementing more
strategies to promote a healthy complementary relationship between
UCSS and ACSS.
Among the factors inuencing UCSS, service industry agglomeration,
market dependence, and land development intensity emerge as domi-
nant drivers, playing a pivotal role in shaping UCSS distribution.
Different from other types of UPS (Dong et al., 2023; Jia et al., 2024b;
Ma et al., 2023b), urban planning and policy should prioritize these
dominant factors, while still acknowledging the inuence of founda-
tional factors. Although UCSS has already demonstrated its utility in
urban centers like CBDs (Peng et al., 2020), its potential at the com-
munity level remains underutilized. While dispersed UCSS development
may involve higher costs and face challenges in attracting consumer
trafc, concentrating UCSS in community commercial hubs could still
enhance service diversity and convenience for residents.
The development patterns of UPS (Cui et al., 2013; Dong et al.,
2023a), including UCSS, should respect the unique characteristics of
each case city, and planning strategies must be tailored accordingly. For
example, in cities like Beijing and Chengdu, where the commercial
environment is well-established, leveraging the existing commercial
vitality can help create unique business types and functions that enrich
UCSS. This could involve connecting UCSS with ACSS to scale up and
align with local consumer demand. Conversely, in cities like Guangzhou
and Wuhan, where the service demand is more specialized, a targeted
development strategy focused on entertainment and leisure-oriented
UCSS can create a complementary relationship with ACSS. Compre-
hensive cities should exibly combine both development models,
adjusting layouts to meet the specic needs of different urban areas.
6.3. Contributions and limitations
The contributions of this study are as follows: (1) This study applies
central place theory to UCSS, revealing the enhanced applicability of
transportation and administrative principles, in addition to the tradi-
tional market principle. This not only deepens the understanding of
spatial layout patterns in different types of consumer spaces but also
demonstrates that the spatial logic of consumer spaces can signicantly
change when constrained by physical limitations, thus expanding the
application of central place theory. (2) The study identies signicant
differences in the spatial distribution of UCSS compared to other UPS,
emphasizing its unique characteristics as a consumer space and moving
beyond a development model solely based on development intensity.
This provides new guidance for urban planning, helping planners better
balance consumer demand with urban functional layouts. (3) By incor-
porating UCSS into the theoretical framework of consumer space layout,
this study reveals key differences between UCSS and ACSS, deepening
the understanding of UCSS’s spatial siting logic. This not only provides a
systematic theoretical basis for UCSS’s spatial layout but also offers
important insights for the integration of UCSS and ACSS in the future,
contributing to the development of a three-dimensional urban consumer
service network. (4) The study introduces a systematic analytical
framework that categorizes the factors inuencing UCSS layout into
foundational and dominant factors, offering a new perspective for
policy-making. This framework emphasizes the interaction between
different factors, helping planners to more precisely optimize UCSS
layouts, improve the efciency of underground space utilization, and
promote coordinated development with above-ground consumer spaces.
However, Some limitations are acknowledged in this study. First, in
the selection of driving force factors, due to the lack of research spe-
cically analyzing the driving forces of UCSSs, the spatial distribution
drivers of underground spaces and commercial spaces were primarily
referenced in this study, with factors selected based on experience,
which introduces certain limitations. Additionally, a dynamic research
perspective is lacking in this study. As consumer services increasingly
penetrate underground spaces, the distribution patterns and inuencing
factors of underground consumer service spaces are still evolving.
Future research could incorporate multiple temporal cross-sections to
better capture the dynamic evolution characteristics of UCSS.
7. Conclusions
The central urban areas of seven representative megacities in China
are used as examples to compare the different distribution and clustering
characteristics of ACSS and UCSS, and to explore the factors inuencing
the distribution patterns of UCSS. The specic conclusions are as
follows:
(1) Both ACSS and UCSS exhibit hierarchical and agglomerative
characteristics, forming multi-centered, concentric spatial distribu-
tion patterns. However, the distribution range of UCSS is much
smaller than that of ACSS, and UCSS shows a higher overall degree of
agglomeration compared to ACSS.
(2) The spatial selection logic of UCSS differs from that of other types
of UPS. While UPS layouts are largely inuenced by TOD theory and
typically organized around public transportation hubs based on
development intensity, UCSS, due to its unique consumer space
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
18

attributes, is more strongly affected by service industry agglomera-
tion and market factors.
(3) The spatial layout of UCSS supports the core principles of central
place theory but also introduces some deviations. While ACSS largely
aligns with the market principles of central place theory, UCSS fol-
lows a pattern more closely associated with transportation and
administrative principles. In many locations, the hierarchical struc-
ture of UCSS centers does not correspond with that of ACSS. More-
over, UCSS exhibits greater diversity in micro-level spatial layout
patterns.
(4) The spatial distribution of UCSS is inuenced by two categories of
factors: foundational and dominant. Socioeconomic and trans-
portation factors provide the basic conditions, while service industry
agglomeration, market dependence, and land development intensity
have a more direct impact on UCSS distribution. The commercial
atmosphere and existing underground space development are the
two most important factors in UCSS layout, with UCSS often being
developed in conjunction with other types of underground spaces.
(5) The strong radiation range of shopping malls and subway sta-
tions, as well as the scope of high-density built-up areas, are crucial
scales for understanding the distribution of UCSS. Due to cost con-
straints, the expansion of UCSS is limited, with the strong radiation
range of shopping malls and subway stations typically falling be-
tween 400 and 800 m. The distribution density of UCSS decreases
signicantly faster than ACSS outside of high-density built-up areas.
CRediT authorship contribution statement
Yuxiao Tang: Writing – original draft, Visualization, Validation,
Resources, Methodology, Formal analysis, Conceptualization. Yudi
Tang: Writing – review & editing, Resources, Methodology, Formal
analysis.
Declaration of competing interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Acknowledgments
This work was nancially supported by the China Scholarship
Council (202208320010).
Data availability
Data will be made available on request.
References
Admiraal, H., & Cornaro, A. (2016). Why underground space should be included in urban
planning policy – And how this will enhance an urban underground future.
Tunnelling and Underground Space Technology, 55, 214–220. https://doi.org/
10.1016/j.tust.2015.11.013
Anselin, L. (2010). Local indicators of spatial association—LISA. Geographical Analysis,
27, 93–115.
Applebaum, W. (1966). Guidelines for a store-location strategy study. Journal of
Marketing, 30(4), 42–45. https://doi.org/10.1177/002224296603000409
Besner, J. (2017). Cities think underground – Underground space (also) for people.
Procedia Engineering, 209, 49–55. https://doi.org/10.1016/j.proeng.2017.11.129
Bobylev, N. (2016a). Transitions to a high density urban underground space. Procedia
Engineering, 165, 184–192. https://doi.org/10.1016/j.proeng.2016.11.750
Bobylev, N. (2016b). Underground space as an urban indicator: Measuring use of
subsurface. Tunnelling and Underground Space Technology, 55, 40–51. https://doi.org/
10.1016/j.tust.2015.10.024
Boratinskii, V., & Osaragi, T. (2024). Modelling of urban activity centers in Tokyo from
alternative indicators of urban activity. Cities, 151, Article 105108. https://doi.org/
10.1016/j.cities.2024.105108
Broere, W. (2016). Urban underground space: Solving the problems of today’s cities.
Tunnelling and Underground Space Technology, 55, 245–248. https://doi.org/
10.1016/j.tust.2015.11.012
Calthorpe, P. (1993). The next American metropolis: Ecology, community, and the American
dream. New York: Princeton Architectural Press.
Carter, H., & Potter, R. (1983). The urban retailing system: Location, cognition and
behaviour. The Geographical Journal, 149, 371.
Cervero, R., Sarmiento, O. L., Jacoby, E., Gomez, L. F., & Neiman, A. (2009). Inuences
of built environments on walking and cycling: Lessons from Bogot´
a. International
Journal of Sustainable Transportation, 3(4), 203–226. https://doi.org/10.1080/
15568310802178314
Chen, D., Zhang, Y., Yao, Y., Hong, Y., Guan, Q., & Tu, W. (2019). Exploring the spatial
differentiation of urbanization on two sides of the Hu Huanyong Line—Based on
nighttime light data and cellular automata. Applied Geography, 112, Article 102081.
https://doi.org/10.1016/j.apgeog.2019.102081
Chen, Q., & Wang, M. (2022). Opening of high-speed rail and the consumer service
industry: Evidence from China. Economic Analysis and Policy, 76, 31–45. https://doi.
org/10.1016/j.eap.2022.07.010
Chen, Y., Chen, Z., Guo, D., Zhao, Z., Lin, T., & Zhang, C. (2022). Underground space use
of urban built-up areas in the central city of Nanjing: Insight based on a dynamic
population distribution. Underground Space, 7(5), 748–766. https://doi.org/
10.1016/j.undsp.2021.12.006
Chen, Z. L., Chen, J. Y., Liu, H., & Zhang, Z. F. (2018). Present status and development
trends of underground space in Chinese cities: Evaluation and analysis. Tunnelling
and Underground Space Technology, 71, 253–270. https://doi.org/10.1016/j.
tust.2017.08.027
Chen, Z., & Liu, H. (2011). Master planning of urban underground space. Master Planning
of Urban Underground Space (in Chinese).
China Academy of Engineering Strategic Consulting Center. (2024). China Society for
Rock Mechanics and Engineering Underground Space Committee, & Urban Planning
Society of China. 2023 blue book on urban underground space development in China.
Beijing, China: China Science Publishing & Media Ltd. (in Chinese).
Christaller, W. (1966). Central places in southern Germany. Prentice-Hall, Inc. (C. W.
Baskin, Trans.).
Cui, J. (2021). Building three-dimensional pedestrian networks in cities. Underground
Space, 6(2), 217–224. https://doi.org/10.1016/j.undsp.2020.02.008
Cui, J., Allan, A., Taylor, M. A. P., & Lin, D. (2013). Underground pedestrian systems
development in cities: Inuencing factors and implications. Tunnelling and
Underground Space Technology, 35, 152–160. https://doi.org/10.1016/j.
tust.2012.12.009
Cui, J., Allan, A., Taylor, M. A. P., & Lin, D. (2015). An examination of Pedestrian trip
behaviour in underground Pedestrian systems. International Planning Studies, 20(3),
209–227. https://doi.org/10.1080/13563475.2014.984662
Cui, J., & Nelson, J. D. (2019). Underground transport: An overview. Tunnelling and
Underground Space Technology, 87, 122–126. https://doi.org/10.1016/j.
tust.2019.01.003
Daniel, C., & Hernandez, T. (2024). What retail apocalypse? A Delphi forecast of
commercial space demand in the Toronto region. Journal of Retailing and Consumer
Services, 77, Article 103670. https://doi.org/10.1016/j.jretconser.2023.103670
Darbani, E. S., Parapari, D. M., Mirheydar, S., & Mastiani, S. (2020). Climatic design
principles of underground spaces using CFD simulation. Civil and Environmental
Engineering, 16(2), 340–350. https://doi.org/10.2478/cee-2020-0034
Davies, D. H. (1959). Boundary study as a tool in CBD analysis: An interpretation of
certain aspects of the boundary of Cape Town’s central business district. Economic
Geography, 35(4), 322–345.
Debrock, S., Van Acker, M., & Admiraal, H. (2023). Design recommendations for
sustainable urban underground spaces. Tunnelling and Underground Space Technology,
140, Article 105332. https://doi.org/10.1016/j.tust.2023.105332
Deng, F., Cheng, T., Huang, Y., Chen, Z., & Han, Q. (2024). Evaluation of urban
underground space via automated constraint identication and hybrid analysis.
Tunnelling and Underground Space Technology, 153, Article 106005. https://doi.org/
10.1016/j.tust.2024.106005
Dong, X., Wu, Y., Chen, X., Li, H., Cao, B., Zhang, X., Yan, X., Li, Z., Long, Y., & Li, X.
(2021). Effect of thermal, acoustic, and lighting environment in underground space
on human comfort and work efciency: A review. Science of the Total Environment,
786, Article 147537. https://doi.org/10.1016/j.scitotenv.2021.147537
Dong, Y. H., Peng, F. L., Bao, Z. H., & Qiao, Y. K. (2021b). Identication of the spatial
distribution pattern and driving forces of underground parking space based on multi-
source data: A case study of Fuzhou City in China. Sustainable Cities and Society, 72,
Article 103084. https://doi.org/10.1016/j.scs.2021.103084
Dong, Y. H., Peng, F. L., Du, Y., & Men, Y. Q. (2023). Automatic identication and feature
recognition of the metro-led underground space in China based on point of interest
data. Underground Space, 9, 186–199. https://doi.org/10.1016/j.undsp.2022.07.008
Dong, Y. H., Peng, F. L., Li, H., & Men, Y. Q. (2023a). Spatial autocorrelation and spatial
heterogeneity of underground parking space development in Chinese megacities
based on multisource open data. Applied Geography, 153, Article 102897. https://doi.
org/10.1016/j.apgeog.2023.102897
Dong, Y. H., Peng, F. L., Li, H., & Men, Y. Q. (2023b). Spatiotemporal characteristics of
Chinese metro-led underground space development: A multiscale analysis driven by
big data. Tunnelling and Underground Space Technology, 139, Article 105209. https://
doi.org/10.1016/j.tust.2023.105209
Dong, Y. H., Peng, F. L., Qiao, Y. K., Zhang, J. B., & Wu, X. L. (2021a). Measuring the
monetary value of environmental externalities derived from urban underground
facilities: Towards a better understanding of sustainable underground spaces. Energy
and Buildings, 250, Article 111313. https://doi.org/10.1016/j.enbuild.2021.111313
Dong, Y. H., Peng, F. L., Zha, B. H., Qiao, Y. K., & Li, H. (2022). An intelligent layout
planning model for underground space surrounding metro stations based on NSGA-
II. Tunnelling and Underground Space Technology, 128, Article 104648. https://doi.
org/10.1016/j.tust.2022.104648
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
19

Drennan, M. P., & Kelly, H. F. (2011). Measuring urban agglomeration economies with
ofce rents. Journal of Economic Geography, 11(3), 481–507.
Durmisevic, S., & Sariyildiz, S. (2001). A systematic quality assessment of underground
spaces – public transport stations. Cities, 18(1), 13–23. https://doi.org/10.1016/
S0264-2751(00)00050-0
Fan, L., & Cui, X. (2023). Analysis of the factors inuencing the comfort of different
behaviors in underground commercial space: A case study of track connection to
underground space. Iranian Journal of Science and Technology, Transactions of Civil
Engineering, 47(4), 2485–2495. https://doi.org/10.1007/s40996-023-01045-6
Fang, Y., Mao, J., Liu, Q., & Huang, J. (2021). Exploratory space data analysis of spatial
patterns of large-scale retail commercial facilities: The case of Gulou District,
Nanjing, China. Frontiers of Architectural Research, 10(1), 17–32. https://doi.org/
10.1016/j.foar.2020.02.002
Feng, Y., Lee, C. C., & Peng, D. (2023). Does regional integration improve economic
resilience? Evidence from urban agglomerations in China. Sustainable Cities and
Society, 88, Article 104273. https://doi.org/10.1016/j.scs.2022.104273
Ge, Y., Pu, Y., & Sun, M. (2021). Alternative measure of border effects across regions:
Ripley’s K-function method. Papers in Regional Science, 100(1), 287–303. https://doi.
org/10.1111/pirs.12565
Giuliano, G., & Small, K. A. (1991). Subcenters in the Los Angeles region. Regional Science
and Urban Economics, 21(2), 163–182. https://doi.org/10.1016/0166-0462(91)
90032-I
Gu, Z., Osaragi, T., & Lu, W. (2019). Simulating pedestrians’ spatio-temporal distribution
in underground spaces. Sustainable Cities and Society, 48, Article 101552. https://doi.
org/10.1016/j.scs.2019.101552
Hao, F., Wang, S., Feng, Z., Yu, T., & Ma, L. (2018). Spatial pattern and industry
distribution of commercial spaces in Changchun based on POI data. Geographical
Research, 37(2), 366–378 (in Chinese).
He, X., Zhou, Y., Yuan, X., & Zhu, M. (2024). The coordination relationship between
urban development and urban life satisfaction in Chinese cities—An empirical
analysis based on multi-source data. Cities, 150, Article 105016. https://doi.org/
10.1016/j.cities.2024.105016
Hou, G., & Chen, L. (2021). Regional commercial center identication based on POI big
data in China. Arabian Journal of Geosciences, 14(14), 1360. https://doi.org/
10.1007/s12517-021-07597-z
Hu, Y., & Han, Y. (2019). Identication of urban functional areas based on POI data: A
case study of the Guangzhou economic and technological development zone.
Sustainability, 11(5), 1385. https://doi.org/10.3390/su11051385
Jia, L., He, X., & Sun, S. (2024a). Exploring factors inuencing the agglomeration of
implicit consumption space: A comparison with explicit consumption space. Applied
Geography, 169, Article 103340. https://doi.org/10.1016/j.apgeog.2024.103340
Jia, X., Wang, J., & Yan, B. (2024b). Evaluation of existing underground public space
supply based on the node-place-experience model. Tunnelling and Underground Space
Technology, 144, Article 105517. https://doi.org/10.1016/j.tust.2023.105517
Jiao, J., Jin, Y., & Yang, R. (2024). An approach to exploring the spatial distribution and
inuencing factors of urban problems based on Land use types. Sustainable Cities and
Society, 104, Article 105321. https://doi.org/10.1016/j.scs.2024.105321
Kyaw, A. K., Hamed, M. M., Kamruzzaman, M., & Shahid, S. (2023). Spatiotemporal
changes in population exposure to heat stress in South Asia. Sustainable Cities and
Society, 93, Article 104544. https://doi.org/10.1016/j.scs.2023.104544
Lamb, R. F. (1985). The morphology and vitality of business districts in upstate New York
villages 1. The Professional Geographer, 37(2), 162–172. https://doi.org/10.1111/
j.0033-0124.1985.00162.x
Li, H., Chen, P., & Grant, R. (2021a). Built environment, special economic zone, and
housing prices in Shenzhen, China. Applied Geography, 129, Article 102429. https://
doi.org/10.1016/j.apgeog.2021.102429
Li, H., Wei, Y. D., Wu, Y., & Tian, G. (2019). Analyzing housing prices in Shanghai with
open data: Amenity, accessibility and urban structure. Cities, 91, 165–179. https://
doi.org/10.1016/j.cities.2018.11.016
Li, X., Zhou, L., Jia, T., Liu, F., & Zou, Y. (2021b). Analysis of urban spatial risk factors
during the "COVID-19" pandemic: A case study of Wuhan. Urban Planning, 45(8),
78–86 (in Chinese).
Li, Z., Lu, Y., Zhuang, Y., & Yang, L. (2024). Inuencing factors of spatial vitality in
underground space around railway stations: A case study in Shanghai. Tunnelling and
Underground Space Technology, 147, Article 105730. https://doi.org/10.1016/j.
tust.2024.105730
Lin, S. S., Shen, S. L., Zhou, A., & Xu, Y. S. (2021). Risk assessment and management of
excavation system based on fuzzy set theory and machine learning methods.
Automation in Construction, 122, Article 103490. https://doi.org/10.1016/j.
autcon.2020.103490
Liu, S. C., Peng, F. L., Qiao, Y. K., & Dong, Y. H. (2024). Quantitative evaluation of the
contribution of underground space to urban resilience: A case study in China.
Underground Space, 17, 1–24. https://doi.org/10.1016/j.undsp.2023.11.007
Liu, X., Gong, L., Gong, Y., & Liu, Y. (2015a). Revealing travel patterns and city structure
with taxi trip data. Journal of Transport Geography, 43, 78–90. https://doi.org/
10.1016/j.jtrangeo.2015.01.016
Liu, Y., Liu, X., Gao, S., Gong, L., Kang, C., Zhi, Y., Chi, G., & Shi, L. (2015b). Social
sensing: A new approach to understanding our socioeconomic environments. Annals
of the Association of American Geographers, 105(3), 512–530. https://doi.org/
10.1080/00045608.2015.1018773
Ma, C. X., & Peng, F. L. (2021). Monetary evaluation method of comprehensive benets
of complex underground roads for motor vehicles orienting urban sustainable
development. Sustainable Cities and Society, 65, Article 102569. https://doi.org/
10.1016/j.scs.2020.102569
Ma, C. X., & Peng, F. L. (2023a). Evaluation of spatial performance and supply-demand
ratios of urban underground space based on POI data: A case study of Shanghai.
Tunnelling and Underground Space Technology, 131, Article 104775. https://doi.org/
10.1016/j.tust.2022.104775
Ma, C. X., Peng, F. L., Qiao, Y. K., & Li, H. (2022). Evaluation of spatial performance of
metro-led urban underground public space: A case study in Shanghai. Tunnelling and
Underground Space Technology, 124, Article 104484. https://doi.org/10.1016/j.
tust.2022.104484
Ma, C. X., Peng, F. L., Qiao, Y. K., & Li, H. (2023b). Inuential factors of spatial
performance in metro-led urban underground public space: A case study in
Shanghai. Underground Space, 8, 229–251. https://doi.org/10.1016/j.
undsp.2022.03.001
Makana, L. O., Jefferson, I., Hunt, D. V. L., & Rogers, C. D. F. (2016). Assessment of the
future resilience of sustainable urban sub-surface environments. Tunnelling and
Underground Space Technology, 55, 21–31. https://doi.org/10.1016/j.
tust.2015.11.016
Malczewski, J. (2009). Central place theory. International encyclopedia of human
geography (pp. 127–131). Elsevier. https://doi.org/10.1016/B978-0-08-102295-
5.10332-4
McMillen, D. P. (2001). Nonparametric employment subcenter identication. Journal of
Urban Economics, 50(3), 448–473. https://doi.org/10.1006/juec.2001.2228
Mehta, V. (2014). Evaluating public space. Journal of Urban Design, 19(1), 53–88.
https://doi.org/10.1080/13574809.2013.854698
Meijers, E., Burger, M., Camagni, R., Capello, R., & Caragliu, A. (2017). Static vs.
Dynamic agglomeration economies. Spatial context and structural evolution behind
urban growth. Papers in Regional Science, 95(1), 133–158.
Melo, P. C., & Graham, D. J. (2018). Transport-induced agglomeration effects: Evidence
for US metropolitan areas. Regional Science Policy & Practice, 10.
Merino, M., & Ramirez-Nafarrate, A. (2016). Estimation of retail sales under competitive
location in Mexico. Journal of Business Research, 69(2), 445–451. https://doi.org/
10.1016/j.jbusres.2015.06.050
Okabe, A., Satoh, T., & Sugihara, K. (2009). A kernel density estimation method for
networks, its computational method and a GIS-based tool. International Journal of
Geographical Information Science, 23(1), 7–32. https://doi.org/10.1080/
13658810802475491
Paul, J., & Rosenbaum, M. (2020). Retailing and consumer services at a tipping point:
New conceptual frameworks and theoretical models. Journal of Retailing and
Consumer Services, 54, Article 101977. https://doi.org/10.1016/j.
jretconser.2019.101977
Peng, F. L., Qiao, Y. K., Sabri, S., Atazadeh, B., & Rajabifard, A. (2021). A collaborative
approach for urban underground space development toward sustainable
development goals: Critical dimensions and future directions. Frontiers of Structural
and Civil Engineering, 15(1), 20–45. https://doi.org/10.1007/s11709-021-0716-x
Peng, F. L., Qiao, Y. K., Zhao, J. W., Liu, K., & Li, J. C. (2020). Planning and
implementation of underground space in Chinese central business district (CBD): A
case of Shanghai Hongqiao CBD. Tunnelling and Underground Space Technology, 95,
Article 103176. https://doi.org/10.1016/j.tust.2019.103176
Peng, J., & Peng, F. L. (2018). Layout patterns of underground space comprehensive
utilization around metro station for urban commercial area. In Proceedings of the
GeoShanghai international conference.
Peng, J., Peng, F. L., Yabuki, N., & Fukuda, T. (2019). Factors in the development of
urban underground space surrounding metro stations: A case study of Osaka, Japan.
Tunnelling and Underground Space Technology, 91, Article 103009. https://doi.org/
10.1016/j.tust.2019.103009
Pennerstorfer, A., & Pennerstorfer, D. (2019). How small are small markets? Local
market size for child care services. Regional Science and Urban Economics, 77,
340–355. https://doi.org/10.1016/j.regsciurbeco.2019.06.006
Potter, R. B. (1976). The structural characteristics of the urban retailing system and the nature
of consumer behaviour and perception : A case study based on Stockport. Doctoral
dissertation. Royal Holloway, University of London.
Qiang, W. W., Wen, T., Luo, H., Huang, B., & Lee, H. F. (2024). Does a more compact
urban center layout matter in reducing household carbon emissions? Evidence from
Chinese cities. Land Use Policy, 146, Article 107320. https://doi.org/10.1016/j.
landusepol.2024.107320
Qiao, R., Wu, Z., Gao, S., Jiang, Q., Liu, X., Huang, C., Xia, L., & Chen, M. (2023).
Towards inclusive underground public transportation: Gender differences on
thermal comfort. Building and Environment, 242, Article 110569. https://doi.org/
10.1016/j.buildenv.2023.110569
Qiao, Y. K., Peng, F. L., Sabri, S., & Rajabifard, A. (2019). Socio-environmental costs of
underground space use for urban sustainability. Sustainable Cities and Society, 51,
Article 101757. https://doi.org/10.1016/j.scs.2019.101757
Qin, X., Zhen, F., Zhu, S., & Xi, G. (2014). Study on the spatial distribution pattern of
catering industry in Nanjing urban area based on online reputation — Taking
Dianping.com as an Example. Geographical Sciences, 34, 810–817. https://doi.org/
10.13249/j.cnki.sgs.2014.07.011 (in Chinese).
Ren, Z., Seipel, S., & Jiang, B. (2024). A topology-based approach to identifying urban
centers in America using multi-source geospatial big data. Computers, Environment
and Urban Systems, 107, Article 102045. https://doi.org/10.1016/j.
compenvurbsys.2023.102045
Ripley, B. D. (1977). Modelling spatial patterns. Journal of the Royal Statistical Society, 39,
172–212. Ripley, B. D. 1977. Modelling spatial patterns. J.r.Stat.Soc, 39(2), 172–192.
Schmidt, S., Nemeth, J., & Botsford, E. (2011). The evolution of privately owned public
spaces in New York City. URBAN DESIGN International, 16(4), 270–284. https://doi.
org/10.1057/udi.2011.12
Self, S., Overby, A., Zgodic, A., White, D., McLain, A., & Dyckman, C. (2023).
A hypothesis test for detecting distance-specic clustering and dispersion in areal
data. Spatial Statistics, 55, Article 100757. https://doi.org/10.1016/j.
spasta.2023.100757
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
20

Shi, X., Wang, S., Hao, F., & Zhou, S. (2019). Spatial characteristics and causes of O
2
O
catering industry in Changchun based on comparison of physical stores. Human
Geography, 2, 10 (in Chinese).
Sparks, K., Thakur, G., Pasarkar, A., & Urban, M. (2020). A global analysis of cities’
geosocial temporal signatures for points of interest hours of operation. International
Journal of Geographical Information Science, 34(4), 759–776.
Sun, L., & Leng, J. (2021). Research on inuencing factors of travel in underground space
based on multi-source data: Spatial optimization design for low-carbon travel. Energy
and Buildings, 253, Article 111524. https://doi.org/10.1016/j.enbuild.2021.111524
Tammiksaar, E., Jauhiainen, J. S., Pae, T., & Ahas, R. (2018). Edgar Kant, Estonian
geography and the reception of Walter Christaller’s central place theory, 1933–1960.
Journal of Historical Geography, 60, 77–88. https://doi.org/10.1016/j.
jhg.2018.02.001
Teller, C., & Reutterer, T. (2008). The evolving concept of retail attractiveness: What
makes retail agglomerations attractive when customers shop at them? Journal of
Retailing and Consumer Services, 15(3), 127–143. https://doi.org/10.1016/j.
jretconser.2007.03.003
Tian, C., Luan, W., Wang, H., Yang, D., Su, M., Zhang, Z., & Li, Z. (2023). Spatio-temporal
agglomeration and morphological causes of Shanghai catering clusters. Applied
Geography, 152, Article 102898. https://doi.org/10.1016/j.apgeog.2023.102898
Wan, B., Wan, D., & Sheng, Q. (2024). Evaluating changes in business distribution within
urban rail transit hubs in Beijing via Point of Interest (POI) data analysis
(2008–2020). Frontiers of Architectural Research, 13(2), 305–318. https://doi.org/
10.1016/j.foar.2023.12.002
Wang, J., Li, X., Christakos, G., Liao, Y., Zhang, T., Gu, X., & Zheng, X. (2010).
Geographical detectors-based health risk assessment and its application in the neural
tube defects study of the Heshun Region, China. International Journal of Geographical
Information Science, 24(1), 107–127. https://doi.org/10.1080/13658810802443457
Wang, J. F., Zhang, T. L., & Fu, B. J. (2016). A measure of spatial stratied heterogeneity.
Ecological Indicators, 67, 250–256. https://doi.org/10.1016/j.ecolind.2016.02.052
Wang, N., Wu, J., & Peng, Z. (2021). Spatial pattern and inuencing factors of residential
service industries in Shenzhen based on multi-source big data. Tropical Geography, 41
(5), 956–967. https://doi.org/10.13284/j.cnki.rddl.003381 (in Chinese).
Wang, T., Wang, Y., Zhao, X., & Fu, X. (2018). Spatial distribution pattern of the
customer count and satisfaction of commercial facilities based on social network
review data in Beijing, China. Computers, Environment and Urban Systems, 71, 88–97.
https://doi.org/10.1016/j.compenvurbsys.2018.04.005
Watkins, A. R. (2014). The spatial distribution of economic activity in Melbourne,
1971–2006. Urban Geography, 35(7), 1041–1065. https://doi.org/10.1080/
02723638.2014.930574
Wrigley, N. (1982). Review of Spatial processes: Models and applications. Ord.
Geographical Journal, 148(3), 383–385.
Wu, K., Zhang, H., Wang, Y., Wu, Q., Ye, Y., Kangmin, W., Hongou, Z., Yang, W.,
Qitao, W., & Yuyao, Y. (2016). Identication and spatial pattern of multiple types of
commercial centers in Guangzhou. Progress in Geography, 35(8), 963–974. https://
doi.org/10.18306/dlkxjz.2016.08.005 (in Chinese).
Wyckoff, W. (1989). Central place theory and the location of services in Colorado in
1899. The Social Science Journal, 26(4), 383–398. https://doi.org/10.1016/0362-
3319(89)90003-7
Xu, H., Jiang, Y., Fang, Y., & Huang, X. (2024). Spatial pattern characteristics and
accessibility of public sports facilities in Beijing. Journal of Arid Land Resources and
Environment, 38(5), 190–198. https://doi.org/10.13448/j.cnki.jalre.2024.111 (in
Chinese).
Xu, Y., & Chen, X. (2021a). Factors inuencing the vitality of underground commercial
spaces in metro station areas and strategies to enhance vitality. Modern Urban
Research, 12, 70–76 (in Chinese).
Xu, Y., & Chen, X. (2021b). Quantitative analysis of spatial vitality and spatial
characteristics of urban underground space (UUS) in metro area. Tunnelling and
Underground Space Technology, 111, Article 103875. https://doi.org/10.1016/j.
tust.2021.103875
Xu, Y., & Chen, X. (2023). Uncovering the relationship among spatial vitality, perception,
and environment of urban underground space in the metro zone. Underground Space,
12, 167–182. https://doi.org/10.1016/j.undsp.2023.02.010
Yan, L. (2017). The impact of online take-out service on restaurant location
characteristics—A study based on open data in Shanghai inner city. Urban Planning
Forum, 04, 113–119. https://doi.org/10.16361/j.upf.201704013 (in Chinese).
Yang, J., Zhu, J., Sun, Y., & Zhao, J. (2019). Delimitating urban commercial central
districts by combining kernel density estimation and road intersections: A case study
in Nanjing City, China. ISPRS International Journal of Geo-Information, 8(2), 93.
https://doi.org/10.3390/ijgi8020093
Yang, T., Pan, H., Hewings, G., & Jin, Y. (2019). Understanding urban sub-centers with
heterogeneity in agglomeration economies—Where do emerging commercial
establishments locate? Cities, 86, 25–36. https://doi.org/10.1016/j.
cities.2018.12.015
Yang, Z., Wang, H., & Chen, B. (2024). Assessment of urban waterlogging-induced road
trafc safety risk and identication of its driving factors: A case study of Beijing.
Transportation Research Part A: Policy and Practice, 183, Article 104080. https://doi.
org/10.1016/j.tra.2024.104080
Yi, C., Nam, J., Kim, J., & Lee, J. S. (2024). Comparison of the distributions of centrality
indices: Using spatial big data to understand urban spatial structure. Cities, 150,
Article 105046. https://doi.org/10.1016/j.cities.2024.105046
Yu, W., Ai, T., & Shao, S. (2015). The analysis and delimitation of Central Business
District using network kernel density estimation. Journal of Transport Geography, 45,
32–47. https://doi.org/10.1016/j.jtrangeo.2015.04.008
Zargarian, R., Hunt, D. V. L., Braithwaite, P., Bobylev, N., & Rogers, C. D. F. (2018).
A new sustainability framework for urban underground space. Proceedings of the
Institution of Civil Engineers - Engineering Sustainability, 171(5), 238–253. https://doi.
org/10.1680/jensu.15.00013
Zhang, W., Chong, Z., Li, X., & Nie, G. (2020). Spatial patterns and determinant factors of
population ow networks in China: Analysis on Tencent Location Big Data. Cities, 99,
Article 102640. https://doi.org/10.1016/j.cities.2020.102640
Zhao, J. W., Peng, F. L., Wang, T. Q., Zhang, X. Y., & Jiang, B. N. (2016). Advances in
master planning of urban underground space (UUS) in China. Tunnelling and
Underground Space Technology, 55, 290–307. https://doi.org/10.1016/j.
tust.2015.11.011
Zhao, N., Shi, C., & Wang, C. (2024). Effects of the “double reduction” policy on the
commercial tutoring sector in China. International Journal of Educational Development,
105, Article 102989. https://doi.org/10.1016/j.ijedudev.2024.102989
Zhong, C., Arisona, S. M., Huang, X., Batty, M., & Schmitt, G. (2014). Detecting the
dynamics of urban structure through spatial network analysis. International Journal
of Geographical Information Science, 28(11), 2178–2199. https://doi.org/10.1080/
13658816.2014.914521
Zhou, L., Wang, C., & Zhen, F. (2023a). Exploring and evaluating the spatial association
between commercial and residential spaces using Baidu trajectory data. Cities, 141,
Article 104514. https://doi.org/10.1016/j.cities.2023.104514
Zhou, L., Xiao, W., Zheng, Z., & Zhang, H. (2023b). Commercial dynamics in urban China
during the COVID-19 recession: Vulnerability and short-term adaptation of
commercial centers in Shanghai. Applied Geography, 152, Article 102889. https://
doi.org/10.1016/j.apgeog.2023.102889
Y. Tang and Y. Tang
Sustainable Cities and Society 116 (2024) 105924
21