Next Article in Journal
Exploring the Joint Influence of Built Environment Factors on Urban Rail Transit Peak-Hour Ridership Using DeepSeek
Previous Article in Journal
Synergistic Effects of Alkali Activator Dosage on Carbonation Resistance and Microstructural Evolution of Recycled Concrete: Insights from Fractal Analysis and Optimal Threshold Identification
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Investigation into the Distribution Features and Determinants of Underground Commercial Spaces in Qingdao City

College of Civil Engineering and Architecture, Shandong University of Science and Technology, Qingdao 266590, China
*
Author to whom correspondence should be addressed.
Buildings 2025, 15(10), 1743; https://doi.org/10.3390/buildings15101743
Submission received: 11 April 2025 / Revised: 10 May 2025 / Accepted: 19 May 2025 / Published: 21 May 2025
(This article belongs to the Section Architectural Design, Urban Science, and Real Estate)

Abstract

With the gradual increase in the total volume of underground commerce in cities, underground commercial spaces are increasingly becoming a key carrier for breaking the constraints of land resources and reconfiguring the relationship between people and land. This paper quantifies and visualizes the layout and scale of underground commercial spaces in the central urban area of Qingdao by using kernel density, multi-distance spatial clustering, and spatial autocorrelation analysis and analyzes the influencing factors by using the geographical detector and MGWR model. The research results show that the underground commercial spaces in the central urban area present a “multi-core–multi-level” layout pattern, and high-density areas are more likely to cluster, with the most significant clustering scale being 3.39 km. Commercial supporting facilities, development of underground space, and population heat value are the core driving factors. The impact of rail transit, centrality, commercial supporting facilities, and development of underground space on the east coast urban area is much greater than that on the west and north urban areas. Finally, corresponding strategies are proposed from the perspectives of business districts, station areas, supply and demand, and planning and management to optimize the development and layout of underground commercial spaces, so as to promote the organic integration of underground commercial spaces and urban spaces.

1. Introduction

With the continuous advancement of urbanization, issues such as traffic congestion, environmental pollution, and uncontrolled population growth have rapidly intensified, leading to increasingly scarce land resources and heightened pressure on urban ground space [1,2]. Actively developing and utilizing urban underground space can not only alleviate the land use pressure caused by urbanization but also significantly enhance urban resilience and improve the quality of life for residents [3]. As a critical component of urban underground infrastructure, underground commercial spaces provide essential public service areas for residents and cities, extend the spatial dimensions of ground-level commercial activities, and lay the foundation for deeper underground development.
The study of urban commercial space patterns originated from the exploration of the distribution of urban functional spaces. For instance, Alfred Marshall identified three key factors contributing to commercial agglomeration within his theory of external economies [4]. Alfred Weber elucidated the spatial logic of industrial agglomeration through the concept of “Isodapane” [5], while Walter Christaller introduced the central place theory, which delineates the centrality, range, and hierarchy of commercial services [6]. These foundational theories provided a theoretical framework for subsequent research on the spatial distribution of urban commercial spaces. Building on this foundation, scholars have employed various analytical methods, including kernel density analysis [7,8], hot spot analysis [8], space syntax [9,10], and cluster analysis [11], to empirically investigate the layout patterns and agglomeration factors of urban commercial spaces. As urban underground spaces have expanded in volume, underground commercial spaces have emerged as significant carriers of public activities beneath the city surface [12], offering a wide array of services such as shopping, dining, and leisure [13], and many scholars have conducted in-depth analyses and discussions on underground commercial spaces from both quantitative and qualitative perspectives. For example, regarding development and architectural design, Hu Bin et al. proposed strategies to enhance the consumption vitality of underground commercial spaces by focusing on indoor pedestrian streets, anchor stores, atrium settings, and interface designs [14]; Li Wei et al. explored the significance of developing and utilizing underground commercial streets in old urban areas for promoting three-dimensional development and expanding commercial space, while also proposing improvement strategies from the aspects of civil air defense design and commercial planning [15]; Guo Baili systematically categorized four forms of connection between underground commercial streets and external spaces in China and established design standards for the number of floors, floor height, and public pedestrian passages [16]. In terms of spatial vitality, Ma Guimin et al. measured the vitality of underground commercial spaces near metro stations and identified traffic organization and business type arrangement as critical factors for enhancing their vitality [17]; Shang Qian et al. defined passenger flow and public node usage as indicators of economic and social vitality, respectively, and analyzed the influencing factors and mechanisms of underground commercial space vitality in Beijing from the perspectives of location and connectivity [18]. Regarding business types, Yuan Hong et al. employed literature collection and social survey methods to clarify the business formats of underground streets in various commercial districts in Chongqing and conducted a comparative analysis with underground commercial spaces in Japan [19]. In terms of comfort, Li Fan et al. utilized statistical correlation analysis, regression modeling, and questionnaire surveys to demonstrate that pedestrian comfort is significantly influenced by the detour coefficient of underground commercial spaces, the width of linear spaces, and the density of signage [20].
At present, rail transit exerts a significant driving effect on the agglomeration of commercial spaces [17]. With the rapid development of rail transit and the increasingly sophisticated planning of urban underground spaces, many cities have seen the formation of substantial underground commercial areas. The commercial nature of ground-level business areas serves as the foundation for determining the development drivers, demand intensity, and appropriate functional types of underground commercial spaces [21]. Underground commercial spaces heavily rely on the commercial flow generated by ground-level business centers [22]. Ground-level commercial areas can effectively channel people flow into underground commercial spaces via internal and external vertical transportation systems, thereby enhancing their vitality. Moreover, from a three-dimensional commercial perspective, ground-level commercial areas also play a crucial role in promoting the development and construction of underground commercial spaces [23]. These aspects collectively demonstrate the radiation effect of ground-level commercial areas on underground commercial spaces. As an extension of ground-level commercial spaces, the overall spatial distribution of underground commercial spaces generally aligns with that of their above-ground counterparts. However, these underground spaces are not independent business forms and still rely on existing ground-level commercial districts [22]. Their spatial agglomeration is influenced by multiple factors, including rail transit networks, metro station density, ground-level commercial density, and the intensity of underground development [24,25,26]; in addition, the existence of issues such as unclear three-dimensional land registration and non-standardized usage rights registration in underground spaces [27] may lead developers to avoid areas with complex three-dimensional property rights. This, in turn, restricts the development and interconnectivity of underground space projects [13], thereby influencing the spatial location selection of underground commercial areas to some extent. The asynchronous development of different urban areas results in spatial heterogeneity in the distribution of underground commercial spaces. This means that the impact of various factors on the spatial distribution of underground commercial spaces varies across different regions within a city. Using OLS regression from a global perspective can lead to local interpretation failure [28], necessitating a more localized approach. Therefore, the geographically weighted regression (GWR) model and the multi-scale geographically weighted regression (MGWR) model are employed to address spatial non-stationarity. Compared to the GWR model, the MGWR model allows each explanatory variable to independently optimize its bandwidth, thereby capturing scale heterogeneity [29] and more accurately reflecting the spatial scale of each variable’s influence [30]. However, the application of MGWR models in studies of underground commercial space distribution remains limited.
Previous studies have examined various aspects of underground commercial space design, vitality, and development drivers. However, when analyzing spatial distribution patterns, these studies predominantly focus on continuous geographical regions or entire cities as research objects, often overlooking the multi-factorial spatial heterogeneity arising from specific geographical characteristics in localized areas. The central urban area of Qingdao encompasses both the historic old city and the primary metropolitan core. The three urban districts—east coast, west coast, and north coast—are arranged in a ring around Jiaozhou Bay, with relatively weak terrestrial connectivity. Regarding maritime connections, key infrastructures such as the Jiaozhou Bay Tunnel, Jiaozhou Bay Bridge, and the southern section of Metro Line 1 were all inaugurated after 2011. While these developments have addressed some gaps in maritime connectivity among the three districts, the long-standing geographical separation has resulted in uneven distributions of multiple elements within the central urban area, thereby inducing significant spatial heterogeneity in the distribution of underground commercial spaces. This unique geographical context provides an ideal backdrop for studying the spatial distribution of underground commercial spaces. Based on this, this study uses the central urban area of Qingdao as a case study and employs spatial analysis methods such as kernel density estimation and spatial autocorrelation to identify the agglomeration areas and scales of underground commercial spaces. It incorporates influencing factors as analytical variables to investigate the driving mechanisms of various influencing factors and utilizes the MGWR model to further uncover the spatial heterogeneity in the distribution of underground commercial spaces and influencing factors across the east, west, and north coast urban areas. Additionally, this study proposes optimization strategies for underground commercial spaces from three perspectives: business district agglomeration-driven strategy, station-area synergistic development, supply-and-demand coupling measurement, and 3D cadastre proactivity, aiming to provide a scientific basis for decision-making in optimizing urban underground commercial layouts.

2. Research Area and Data Sources

2.1. Research Area

Qingdao, a prefecture-level city under the jurisdiction of Shandong Province, holds sub-provincial status and is designated as a city with independent planning authority. In September 2023, the State Council approved the “Shandong Province Spatial Planning (2021–2035)”, emphasizing the need to leverage the leading role of the Shandong Peninsula Urban Agglomeration and enhance the core functions of both the Jinan Metropolitan Area and the Qingdao Metropolitan Area [31]. Subsequently, in November 2024, the State Council endorsed the “Qingdao City Spatial Planning (2021–2035)” [32], officially designating Qingdao as an important coastal central city. Qingdao lies along the coastal underground space development axis and is situated within the concentrated development zone for underground space in the Shandong Peninsula Urban Agglomeration. By the end of 2022, Qingdao ranked among the top 10 cities nationally in terms of urban underground space development potential and social dominance of underground space utilization [33]. Since 2014, Qingdao has issued several key documents, including the “Comprehensive Utilization Planning for Underground Space Resources of Qingdao City” and the “Regulations on the Development and Utilization of Underground Space in Qingdao City”, establishing a comprehensive underground space planning system. Notably, early developments in underground commercial spaces can be found in business districts such as SiFang, TaiDong, LiCun, Zhongshan Road, and Railway Station, showcasing a rich variety and strong representativeness.
Considering the distribution of underground commercial spaces and the current development status of each district, this study focuses on the central urban area of Qingdao (Figure 1). According to the “Qingdao City Spatial Planning (2021–2035)”, the central urban area comprises the east coast urban area, the west coast urban area, and the north coast urban area. This region connects the city’s north–south coastal development axis and concentrates the urban functions of the Bay Area Metropolitan Area. It is the most densely populated area in Qingdao and hosts 88% of the city’s underground commercial facilities, serving 47% of its permanent residents.

2.2. Data Sources

The commercial facilities data for Qingdao City were obtained from Dianping (https://www.dianping.com) (accessed on 20 September 2024). Each store’s data include name, address, star rating, number of reviews, geographic coordinates (longitude and latitude), and the business district it belongs to. After cleaning, deduplication, and classification, the commercial facilities in Qingdao City were categorized into seven types: catering services, shopping services, science and education cultural services, accommodation services, business support, living support, and leisure and entertainment. By filtering the “name” and “address” fields with conditions such as “underground”, “negative”, “B1 floor”, “B2 floor”, “B3 floor”, and “B4 floor”, this research identified underground commercial facilities, thereby distinguishing between above-ground and underground commercial facilities. The geographic coordinates of the commercial facilities were initially in the Baidu09 coordinate system and were subsequently converted to the WGS1984 coordinate system to obtain the accurate Points of Interest (POIs) and their distributions.
The administrative region vector maps were obtained from the National Basic Geographic Information Center (https://cloudcenter.tianditu.gov.cn/) (accessed on 20 December 2024). Basic geographic data, including road networks, green spaces, and river systems, were sourced from Open Street Map (www.openstreetmap.org) (accessed on 22 December 2024). Commercial property rental data for Qingdao and housing price data were collected via web crawlers from Anjuke (https://qd.anjuke.com) (rental data accessed on 22 December 2024; housing price data accessed on 30 December 2024). Population grid data were retrieved from WorldPop (www.worldpop.org) (accessed on 22 December 2024), while population heat map data were obtained from Baidu Heat Map (https://lbsyun.baidu.com/) (accessed between 2 and 8 September 2024). Subway station data were sourced from Amap (www.amap.com) (accessed on 23 December 2024).
The software information is presented below: ArcGIS Pro (version 2.8, Esri, Redlands, CA, USA), SPSS (version 27.0.1, IBM, Armonk, NY, USA), ArcGIS (version 10.2, Esri, Redlands, CA, USA), MGWR (version 2.2.1, Arizona State University, Tempe, AZ, USA), RStudio (version 2024.12.0, Posit, Boston, MA, USA), and R (version 4.4.2, The R Foundation for Statistical Computing, Vienna, Austria).

3. Methodology

This study applies kernel density estimation to identify the core areas of underground commercial zones, with grids summarizing kernel density values serving as geographical analysis units. Based on this, spatial autocorrelation analysis is conducted to assess the global and local agglomeration characteristics of underground commercial activities. The scale of agglomeration for underground commercial activities is determined using the multi-distance spatial clustering function. Subsequently, influencing factors are selected, and the geographical detector method is employed to identify the dominant driving factors. Finally, the MGWR model is utilized to calculate the spatial regression coefficients of the influencing factors, thereby analyzing the spatial heterogeneity of underground commercial activities and their relationship with each influencing factor (Figure 2).

3.1. Kernel Density Estimation (KDE)

Kernel density estimation (KDE) is a non-parametric statistical method widely applied in spatial analysis [34]. It enables the visualization of the spatial distribution density of data points and reveals their underlying spatial patterns. In geography, KDE is frequently employed to investigate the spatial arrangement of various phenomena. The continuous density surface generated through KDE can approximate the continuous nature of density distributions [35], taking into account the influence of each point on its surrounding space. This approach facilitates a more intuitive representation of the spatial agglomeration characteristics of urban elements represented by POIs. In this study, the kernel density analysis of POI data was performed using ArcGIS Pro 2.8, with a search radius set at 400 m. To ensure accuracy in reflecting the true spatial distribution of kernel density, the geodesic method was adopted for measurement [36]. The formula for kernel density estimation is as follows:
f x = i = 1 n k ( x - c i ) h / h 2
In the formula, f x denotes the observed value at point x in space (expressed as a kernel density), h represents the bandwidth, n is the number of POI points contributing to the calculation of the observed value at x , k is the kernel function, and c i indicates the position of the i -th POI point.

3.2. Data Gridding

Data gridding is a process in spatial data processing that transforms irregularly distributed spatial data into regularly arranged grid cells [37]. In this study, a 300 m × 200 m grid cell scale was applied to construct a fishnet for the central urban area. By employing techniques such as replication, translation, merging, and feature-to-point conversion, the center points of hexagonal grids (POIs) were derived. Using these center point POIs as input features, Thiessen polygon grids were generated. Subsequently, the grid data were integrated with kernel density estimates, POI distributions, and other relevant datasets for value assignment. Finally, the results were visualized, and an analysis of influencing factors was conducted.

3.3. Spatial Autocorrelation Analysis

Spatial autocorrelation encompasses global spatial autocorrelation and local spatial autocorrelation [38]. Global spatial autocorrelation can reflect the spatial distribution characteristics of a certain attribute value throughout the entire region [39,40], that is, systematic spatial association, while local spatial autocorrelation can reveal the heterogeneity of a certain attribute value among different spatial units [41]. In this paper, global Moran’s I is used to calculate the differentiation characteristics of underground commercial space in the central urban area to determine whether the underground commercial space has a spatial agglomeration attribute as a whole. Then, Local Moran’s I is used to calculate the local characteristics of the agglomeration degree of underground commercial space, revealing the spatial heterogeneity and differences in local areas. The calculation formulas for the two types of spatial autocorrelation are as follows:
Global   Moran s   I = i = 1 m j = 1 m x i x ¯ x j x ¯ W i j / S 2 i = 1 m j = 1 m W i j
Local   Moran s   I = x i x ¯ j = 1 m W i j x j x ¯ / S 2
In the formula, m denotes the number of grids involved in the calculation, S 2 represents the variance of the observed values, x ¯ indicates the mean of the observed values, W i j signifies the spatial weight between the i -th and j -th elements, and x i and x j correspond to the observed values of grids i and j , respectively.

3.4. Multi-Distance Spatial Clustering (Ripley’s K) Analysis

Ripley’s K function is typically used in cluster analysis. This method determines the distribution patterns of point features in space at different scales by comparing the observed values and expected values at specific distances [42,43,44]. During the calculation of Ripley’s K, multiple Monte Carlo simulations [45] are employed to generate a confidence interval with a confidence level of 99%, quantifying the critical scale at which point features are aggregated or dispersed. The calculation formula is as follows:
L d = A i = 1 n j = 1 , j i n k i , j π n n 1
In the formula, d denotes the distance, n represents the total number of POIs involved in the calculation, A indicates the area of the region where POIs are calculated, and k i , j signifies the weight value.

3.5. Geographical Detector

The geographical detector is a statistical method designed to analyze the spatial heterogeneity of geographical phenomena and their driving factors. Factor detection in the geographical detector quantifies the consistency between the spatial distribution of factors and the dependent variable, thereby assessing the explanatory power of these factors for the dependent variable. This process can be used to screen key driving factors [46]. Interaction detection, on the other hand, quantitatively characterizes the relationship between paired factors and the dependent variable, determining whether interactions exist between factors, as well as the strength and form of such interactions [47]. Compared with traditional statistical methods, the geographical detector imposes fewer assumptions and constraints during its application [48], making it widely applicable in urban geography research areas such as urban ecology [49], regional patterns [50], and urbanization [51]. The calculation formula is as follows:
P D , G = 1 1 n σ G 2 i = 1 m n D , i σ D , i 2
In this formula, P D , G represents the explanatory power of factor D on the spatial differentiation of dependent variable G , with values ranging from 0 to 1. Here, n denotes the total sample size within the study area; σ G 2 represents the global variance of dependent variable G ; m indicates the number of classifications of factor D ; n D , i refers to the sample size in the i -th category of factor D ; and σ D , i 2 signifies the variance of dependent variable G within the i -th category of factor D .

3.6. Multi-Scale Geographically Weighted Regression (MGWR)

The multi-scale geographically weighted regression (MGWR) model is an extension of the global regression model (ordinary least squares, OLS) and the geographically weighted regression model (GWR). It enables the quantitative analysis of geographical variation in the relationship between explanatory variables and response variables [52]. As a local regression model, MGWR is specifically designed to capture spatial heterogeneity in data. In contrast, the traditional OLS model assumes that relationships between variables are constant across all spatial locations, thereby ignoring the inherent spatial heterogeneity of geographical data. This limitation necessitates the introduction of GWR to address spatial non-stationarity issues [53]. Although GWR allows regression coefficients to vary with geographical location, it has the drawback of assigning the same optimal bandwidth to all explanatory variables [54]. The MGWR model improves upon GWR by calculating each regression coefficient through localized regressions [55], incorporating geographical location information into regression parameters for estimating local variations [56]. Furthermore, MGWR distinguishes the scales of influence for different explanatory variables and permits each variable to have its own independent optimal bandwidth [29], thus enhancing the interpretability and flexibility of the model. The calculation formula is as follows:
Y i = j = 1 k β b w j u i , v i x i j + β 0 u i , v i + ε i
In the formula, Y i represents the observed value at grid i , x i j is the value of the j -th independent variable at grid i , u i , v i are the spatial coordinates of the center point of grid i , β 0 u i , v i is the intercept term for grid i , β b w j u i , v i is the local regression coefficient of the j -th independent variable at grid i , and ε i is the random error.

4. Results

4.1. Spatial Layout Features of Underground Commercial Spaces in the Central Urban Area

4.1.1. The Underground Commercial Space Exhibits a Layout Pattern Characterized by “Multiple Cores–Multiple Levels”

By applying the natural break classification method to grade the kernel density values of underground commercial spaces in the central urban area, it is evident that the underground commercial facilities exhibit a “multi-cores, multi-levels” distribution pattern. The seven core areas with higher distribution densities are LiCun, XinDuXin, JinJiaLing, TaiDong, Hong Kong Middle Road, RongChuangMao, and Zhengyang Road (Figure 3). In terms of agglomeration intensity, the underground commercial facilities in the LiCun business district have the highest distribution density, significantly surpassing other core areas. Its spatial structure is not singularly centralized but forms a dual-density center around the LiCun subway station, radiating outward. The XinDuXin, JinJiaLing, TaiDong, Hong Kong Middle Road, RongChuangMao, and Zhengyang Road business districts also exhibit relatively high distribution densities. Among these, the underground commercial facilities in the core areas of the eastern shore urban area radiate outward, forming medium-to-high-density zones in surrounding regions. In contrast, areas such as Lanxiu City, Railway Station, JiMiYa, TangDao Bay, and XiangJiang Road have smaller-scale underground commercial facilities with medium-to-low distribution densities. Along certain subway lines, small-scale micro-clusters with low density and scale have emerged, indicating that the development of underground commercial facilities in these areas is still in its infancy.
In terms of agglomeration form, the central urban area has developed a north–south-oriented development belt. Among these, the east coast urban area, being the older part of Qingdao, exhibits a relatively high level of development. The underground commercial facilities here exhibit patch-distributed clusters, with high-density cores interconnected by medium-to-high-density corridors extending in both the north and south directions. This configuration results in strong spatial continuity within the development belt centered on the eastern shore urban area. However, spatial connectivity between the east and west coast urban areas is limited due to their sole connection via the Jiaozhou Bay Tunnel. Additionally, the lack of large-scale business districts between the east coast and north coast urban areas weakens the spatial continuity of the transition zone. Consequently, only small-scale medium-to-high-density agglomeration zones have emerged at the end of the north–south development belt.

4.1.2. The Agglomeration Features of Underground Commercial Spaces Are Highly Pronounced

In the results of global Moran’s I, the range of the Moran’s I is [−1, 1]. The closer its absolute value is to 1, the stronger the spatial correlation becomes. Positive and negative values, respectively, indicate positive and negative correlations. The Z-value represents the multiple of the standard deviation; when the Z-value exceeds 2.58, it signifies that the confidence level of element aggregation reaches 99%. According to the global Moran’s I (Table 1), the seven types of underground commercial spaces in the central urban area exhibit varying degrees of agglomeration distribution characteristics. All types of underground commercial spaces have a global Moran’s Index greater than 0, a Z-value greater than 2.58, and a p-value less than 0.05, indicating significant spatial agglomeration features across all categories. In terms of spatial agglomeration intensity, the seven types of underground commercial spaces rank as follows: living support (0.550), business support (0.511), catering services (0.501), science and education cultural services (0.479), shopping services (0.470), leisure and entertainment (0.105), and accommodation services (0.088). The global Moran’s I for underground commercial spaces overall is 0.819, which approaches 1, suggesting a strong positive spatial correlation in the distribution of underground commercial spaces in the central urban area.
Based on the global spatial autocorrelation analysis, the Local Moran’s I was calculated to conduct a local spatial autocorrelation analysis. Cluster analysis reveals that the predominant clustering pattern is high–high clustering, with low–high clustering being the primary outlier type. No high–low or low–low clustering areas were identified. Additionally, the scatter plot of the Local Moran’s I exhibits a positive slope, indicating a positive spatial correlation in the density distribution of underground commercial spaces (Figure 4). Based on these findings, the following conclusions can be drawn:
(1)
In terms of quantity, high–high clustering areas constitute a substantial proportion. Areas with higher underground commercial densities are more likely to cluster and exhibit continuous spatial distributions.
(2)
Spatially, underground commercial spaces form point-like clusters in multiple locations. Despite considerable spatial distances between clusters, a limited number of low–high clusters exist, suggesting that high-density areas exert weak radiation effects on surrounding regions, which diminish rapidly with increasing distance.
(3)
From a zoning perspective, the east coast urban area features larger high–high clustering areas, where the distribution of underground commercial spaces aligns closely with surface commercial layouts and metro line directions. In the west coast urban area, underground commercial spaces are primarily concentrated along Metro Line 1 and in the JiMiYa business district. In the north coast urban area, only the Zhengyang Road business district contains a relatively concentrated underground commercial space.

4.1.3. The Underground Commercial Space Exhibits a Core Agglomeration Scale of 3.39 km

Although spatial autocorrelation analysis has identified the clustering patterns of underground commercial spaces, it lacks the ability to quantify the specific scale of clustering and its changing trends over different distances. Consequently, multi-distance spatial clustering analysis is employed to examine the underground commercial spaces in the central urban area across various scales. In the results of this analysis, L(d) denotes the observed value, ExpK represents the expected value (i.e., the theoretical distance), and the value of DiffK indicates the difference between the observed value and the expected value. The clustering state of the elements is most pronounced when DiffK reaches its maximum value. Specifically, a significant clustering state occurs when L(d) > ExpK and the high confidence value exceeds the low confidence value. Conversely, a significant dispersed state is observed when L(d) < ExpK and the high confidence value falls below the low confidence value. Based on the observation results (Figure 5), it is evident that the high confidence value consistently surpasses the low confidence value, indicating that spatial clustering at all distances is statistically significant. At 3.39 km, DiffK attains its peak value, signifying that the clustering of underground commercial spaces is most significant at this scale, with a clustering intensity of 2.69. Beyond 3.39 km, as the observation distance increases, the growth rate of clustering intensity gradually diminishes. At 12.39 km, L(d) equals ExpK, marking the point where underground commercial spaces achieve their maximum clustering intensity of 12.39. Beyond this distance, the spaces no longer exhibit a clustering pattern.

4.2. Analysis of Influencing Factors Shaping the Layout of Underground Commercial Spaces

4.2.1. Selection of Influencing Factors

The layout and agglomeration of underground commercial spaces result from the combined effects of multiple factors, including land cost [22], regional planning, rail transit [22,57], and population density [22]. Drawing on prior research and considering the availability and quantifiability of data, this study identifies and selects 10 influencing factors affecting the layout of underground commercial spaces across four dimensions: location and transportation, functional space, development status, and population factors. Specifically, they are as follows: (1) location and transportation includes three influencing factors—rail transit, road network density, and centrality—which reflect the radiation effect of rail transit and large commercial districts as well as the support capacity of road infrastructure; (2) functional space includes two influencing factors—commercial supporting facilities and development of underground space—which indicate the driving force of ground-level commercial activities and the development level of underground space; (3) development status includes three influencing factors—ground development intensity, shop rent, and land cost—which represent the degree of ground-level development, the cost of space usage, and the cost of space development; (4) population factors include two influencing factors—permanent resident population density and population heat value—which capture static consumption demand and dynamic spatial demand (Table 2). Given the challenges in obtaining data for land cost and underground shop rent, this study uses new house prices and ground-level shop rent as proxies, respectively.
The Pearson correlation coefficient method was employed to examine the correlation between the underground commercial space agglomeration value (Y) and the influencing factors (×1–×10). The results (Figure 6) indicate that the underground commercial space agglomeration value is significantly correlated with all ten influencing factors. Specifically, it exhibits significant positive correlations with road network density, commercial supporting facilities, development of underground space, ground development intensity, shop rent, land cost, permanent resident population density, and population heat value, with correlation coefficients of 0.045, 0.497, 0.565, 0.193, 0.074, 0.094, 0.254, and 0.452, respectively. Notably, the highest correlation coefficient (0.565) is observed for development of underground space. In contrast, the agglomeration value of underground commercial space exhibits a significantly negative correlation with proximity to the nearest subway station and central business districts, with correlation coefficients of −0.112 and −0.165, respectively. The negative correlation suggests that as the distance from the nearest subway station and major commercial areas increases, the distribution of underground commercial space becomes more dispersed. This indirectly verifies the critical role of subway accessibility and central location in promoting the agglomeration of underground commercial space.

4.2.2. The Detection Results of the Influencing Factors

The selected influencing factors from the preceding text were used as detection factors, with grids serving as the sample units. Five adaptive classification methods—equal interval, natural breaks, quantile, geometric interval, and standard deviation—were employed to construct an optimal-parameter geographical detector in RStudio 4.4.2 for conducting factor detection and interaction detection on the factors and dependent variable.
In factor detection, the q-value represents the spatial explanatory power of a factor on the dependent variable, while the p-value reflects the confidence level of the factor’s spatial explanatory power. The results of the factor detection (Table 3) indicate that the p-values of all 10 factors are less than 0.001, suggesting that their explanatory power for the spatial differentiation of underground commercial areas lies within the 99.9% confidence interval. Considering the q-values, the q-values of commercial facilities (×4), development of underground space (×5), and population heat value (×10) are 0.2653, 0.3890, and 0.2801, respectively, explaining 26.53%, 38.90%, and 28.01% of the spatial differentiation of underground commercial areas. Their explanatory power is significantly higher than that of other factors, qualifying them as core driving factors.
Interaction detection identifies interactions between different factors and reveals the form and intensity of these interactions. The interaction detection results show three types of interactions: double-factor enhancement (EB), nonlinear enhancement (EN), and single-factor nonlinear weakening (WU), with no independent interaction factors identified (Table 4). Except for the interaction between shop rent and population heat value, which slightly reduces explanatory power, the explanatory power of any two factors after interaction detection exceeds their original individual explanatory power. This indicates that paired factors significantly enhance the explanatory power for the spatial layout of underground commercial areas. The explanatory power range of the interaction between development of underground space and the other nine factors is (0.3986, 0.5286), surpassing the explanatory power of other two-factor interactions, demonstrating that the combined explanatory power of underground space development with other factors is more pronounced. Notably, the interaction detection between underground space development and population heat value achieves the highest explanatory power at 0.5286, jointly explaining 52.86% of the spatial layout of underground commercial areas from the perspectives of underground facility development and consumer demand. Additionally, the explanatory power of other factors significantly increases after interacting with rail transit and central location, with average increases of 37.35% and 44.2%, respectively. This highlights that rail transit and the radiation capacity of ground commercial areas can effectively couple with other factors and jointly influence the spatial layout of underground commercial areas.

4.2.3. Model Development and Comparative Analysis of Results

The selected influencing factors were used as explanatory variables, and OLS, GWR, and MGWR models were respectively constructed and analyzed using SPSS 27.0.1, ArcGIS 10.2, and MGWR 2.2.1 software for regression analysis. The model fitting information and MGWR model statistics were subsequently obtained.
The results (Table 5) indicate that the MGWR model exhibits superior performance, with an R2 value of 0.859 and an adjusted R2 value of 0.834, both surpassing those of the OLS and GWR models. These values suggest that the MGWR model can explain 94.5% of the variation in underground commercial agglomeration, demonstrating a higher degree of model fit. Furthermore, after performing multi-scale geographically weighted regression, the residual sum of squares (RSS) and corrected Akaike information criterion (AICc) values were found to be 1016 and 8872, respectively, which are significantly lower than those of the OLS and GWR models. This indicates that the MGWR model achieves a better balance between model fit and complexity, resulting in predicted values that more closely approximate the true values.
A multicollinearity test was performed on the explanatory variables, and the results showed that the VIF values for all explanatory variables were below 5. This confirms the absence of multicollinearity among the data. The GWR model results indicate that all explanatory variables share a uniform bandwidth of 722, implying that regression analysis in this model is conducted using a fixed bandwidth. In contrast, the MGWR model reveals varying bandwidths for different explanatory variables. Specifically, centrality, commercial supporting facilities, development of underground space, shop rent, land cost, and permanent resident population density exhibit relatively small bandwidths, with respective values of 46, 47, 48, 50, 145, and 48. These smaller bandwidths suggest that their influence on the agglomeration of underground commercial space is predominantly localized; rail transit and population heat value demonstrate larger bandwidths of 1155 and 406, respectively, indicating a medium-scale influence on underground commercial space agglomeration; road network density and ground development intensity possess the largest bandwidth at 7210, reflecting a global influence on the agglomeration of underground commercial space (Table 6).

4.2.4. Regression Results of the MGWR Model

The regression coefficients of each explanatory variable are spatially visualized, and the values of underground commercial space agglomeration are mapped one-to-one to the values of each explanatory variable in the form of bivariate color zoning maps.
The influence coefficient of rail transit is −1.3246 to 0.2165. The regression coefficients of the underground commercial agglomeration areas in the commercial districts such as LiCun, TaiDong, and Zhongshan Road in the eastern urban area are all negative values, and the influence coefficient reaches the lowest value at the double-density center of the LiCun commercial district (Figure 7). The underground commercial space agglomeration value and the distance to the subway station mostly show a “high-low” coupling relationship. From the eastern urban area to the western and northern urban areas, the regression coefficients gradually increase, and the maximum value appears in the northern urban area. The underground commercial space agglomeration value and the distance to the subway station mostly show a “high-medium” and “high-high” coupling relationship. The meaning of this explanatory variable is the distance from the underground commercial space to the nearest subway station. A negative regression coefficient indicates that the greater the distance, the lower the degree of underground commercial agglomeration, that is, the closer the distance to the subway station, the higher the degree of underground commercial agglomeration. The eastern urban area has a larger number of rail transit lines and stations, and the construction time of rail transit is much earlier than that of the western and northern urban areas. The underground commercial space is strongly influenced by the radiation effect of rail transit during the formation and development process. Moreover, the absolute values of the regression coefficients in the eastern urban area are generally higher than those in the western and northern urban areas, indicating that the impact of rail transit on the underground commercial agglomeration in this area is much greater than in other areas. Generally, the “high-low” coupling areas are concentrated in small clusters within a 500 m to 600 m radius around subway stations. This suggests that subway stations significantly enhance the agglomeration of underground commercial activities within this range; however, their influence diminishes sharply as distance increases beyond this threshold.
The influence coefficient of the centrality ranges from −1.8535 to 5.5280. In the LiCun commercial district, all regression coefficients are positive and reach their maximum value, indicating a “high-low” coupling relationship between the underground commercial space agglomeration value and the distance to the nearest ground-level business district. By contrast, in other commercial districts such as XinDuXin and Zhongshan Road, the regression coefficients are negative, reflecting a promoting effect of the ground-level commercial district on underground commercial agglomeration that diminishes with increasing distance (Figure 8). This explanatory variable represents the distance from underground commercial spaces to the nearest large ground-level commercial district. A positive regression coefficient signifies an inhibitory effect, while a negative coefficient indicates a promoting effect. The unique case of the LiCun commercial district arises due to its early unified planning of underground commercial spaces. Large-scale projects such as Zhongfang Commercial Street and Weike Star City were developed around the LiCun subway station and gradually expanded outward. As a result, the development of underground commercial spaces in LiCun is not constrained by the influence of the ground-level commercial district, leading to the formation of a large-scale double-density underground commercial center. From the perspective of urban planning, forward-looking underground space planning can not only enable underground commercial development to transcend the traditional “ground-to-underground radiation” pattern but also strengthen its capacity for independent growth.
The influence coefficient of commercial supporting facilities ranges from −0.8758 to 5.2304. The regression coefficient reaches its peak in the LiCun business district of the eastern urban area. In this district, the underground commercial space agglomeration value and commercial supporting facilities exhibit a “high-high” coupling relationship. The spatial distribution of underground and above-ground commercial spaces demonstrates consistency, with commercial supporting facilities significantly promoting the agglomeration of underground commercial spaces. Moving southward or northward from the LiCun business district, the regression coefficient gradually decreases, indicating a weakening driving effect of commercial supporting facilities. Nevertheless, the coupling relationship between the underground commercial space agglomeration value and commercial supporting facilities remains predominantly “high-high” (Figure 9). Overall, there is a positive correlation between commercial facilities and underground commercial agglomeration. Regions with high ground-level commercial density tend to develop underground commercial agglomeration zones more readily. Nevertheless, in most areas, underground commercial activities remain non-independent business forms, relying on the development of above-ground commercial activities. As above-ground commercial activities exhibit a circular distribution pattern, underground commercial activities similarly demonstrate this characteristic but within a smaller spatial range.
The eastern urban area is the earliest developed region in Qingdao. Shinan District and Shibei District, located at the southern end of this area, represent the prototype of Qingdao City. Early development and construction established the commercial pattern here, resulting in a significantly higher ground commercial density compared to other regions. However, as urban construction preceded underground space planning, the influence of commercial supporting facilities on underground commercial spaces exhibits spatial instability and unevenness. In contrast, the development of Licang District at the northern end of the eastern urban area occurred later than that of Shinan and Shibei Districts. This allowed for reasonable underground space planning and coordinated development of above-ground and underground spaces, enabling the simultaneous development of both above-ground commercial patterns and underground commercial spaces. Consequently, the influence of commercial supporting facilities on underground commercial spaces in Licang District is notably stronger than in other areas. The West Coast New Area in the western urban area and Chengyang District in the northern urban area were established in 2014 and 1994, respectively. Due to their relatively late development, the explanatory power of commercial supporting facilities for underground commercial agglomeration is weaker in these regions. The commercial land use of the ground area provides underground commerce with essential infrastructure support, including rail transit, public transportation, drainage systems, and power supply. Ground-level commerce channels pedestrian traffic into underground spaces via internal and external vertical transportation systems, facilitating the transfer of consumer demand. Moreover, the synergistic development of above-ground and underground commerce drives the advancement of three-dimensional commercial networks. Overall, commercial supporting facilities form the foundation for the growth of underground commerce.
The influence coefficient of development of underground space ranges from −4.4908 to 5.2502. The relatively small bandwidth value (48) of this explanatory variable in the MGWR model indicates that its impact on underground commercial activities is predominantly localized. Consequently, the influence of development of underground space on the agglomeration of underground commercial spaces exhibits significant regional disparities (Figure 10). In the “high-high” coupling areas between the agglomeration value of underground commercial spaces and development of underground space, the central area of the east coast urban area shows a relatively low absolute regression coefficient with fewer positive values, suggesting a weaker promotion effect of development of underground space on underground commercial activities. By contrast, the peripheral areas of the east coast urban area, as well as the west coast and north coast urban areas, exhibit predominantly positive regression coefficients in “high-high” coupling areas, indicating a markedly stronger promotion effect of development of underground space on underground commercial activities. This phenomenon can be attributed to the asynchronous development patterns across different urban areas. The central area of the east coast urban area is densely equipped with various underground facilities, such as subway stations, indoor underground facilities, and public underground infrastructure. However, as the old urban core of Qingdao, this area faces greater challenges in constructing new underground facilities due to constraints imposed by the existing built environment. As a result, the construction of related underground facilities has only a limited promoting effect on the development of underground commercial activities. In contrast, the peripheral areas of the east coast urban area, along with the west coast and north coast urban areas, experienced later development and construction of underground facilities, benefiting from less restrictive built environments. This allowed the construction of various underground facilities to effectively promote the clustered development of underground commercial spaces, thereby enhancing the overall promoting effect of underground space development on the agglomeration of underground commercial activities in these regions. It is evident that the promoting and supporting role of underground facilities in fostering underground commercial activities must be grounded in the multi-dimensional synergy of factors, including the alignment of temporal sequences and policy coupling between the two.
The influence coefficients of shop rent and land cost range from −1.6053 to 3.9096 and from −4.7234 to 4.9364, respectively. From the coupling relationship between the explanatory variables and the agglomeration value of underground commercial space, the central area of the eastern urban zone is predominantly characterized by “medium-high” coupling, while a “high-high” coupling belt has formed in the peripheral areas of the eastern urban zone. In contrast, the western and northern urban zones exhibit mostly “high-medium” and “high-low” coupling relationships (Figure 11 and Figure 12).
According to the interpretation of the indices, shop rent and land cost represent the usage cost and development cost of underground commercial space, respectively. Generally, rising costs tend to impede development activities. Based on the cost transmission mechanism and yield curve theory, an increase in land cost directly drives up rental prices. The high-rent screening effect tends to exclude low-margin and high-premium business formats, thereby making it challenging for underground businesses to achieve complementary clustering with above-ground counterparts. Moreover, high rents can lead to lower returns on investment for underground commercial ventures. For example, in low-cost areas such as the Zhengyang Road commercial district, where the coupling relationship is characterized as “high-low” and exhibits a negative regression coefficient, increased costs hinder the clustering of underground commercial activities. However, costs are influenced by multiple factors. Higher operational and land costs often correspond to superior facilities, prime locations, and greater consumption potential, which can enhance the attractiveness of commercial clustering. For instance, in high-cost areas like the Olympic Sailing Center and Haixin Square in Shinan District, where the coupling relationships are categorized as “high-high” or “high-medium” with positive regression coefficients, higher costs actually promote the clustering of underground commercial activities. Combining the results of interaction detection, the explanatory power of the two variables—shop rent and land cost—after interaction analysis is 0.0222. From a purely cost-based perspective, these factors alone struggle to provide a robust explanation for the clustering mechanisms of underground commercial spaces. Future research could explore the coupling relationships between costs and other factors, such as business formats, facility quality, and consumption potential, to better elucidate the underlying mechanisms.
The influence coefficient of permanent resident population density ranges from −2.8828 to 2.1630. When considering the bandwidth (48) of this explanatory variable, it is evident that the regression coefficient of permanent resident population density exhibits strong regional characteristics. Specifically, both high-value and low-value areas of the regression coefficient are distributed in a point-like pattern, whereas low-value areas predominantly display a surface-like distribution. The highest regression coefficient occurs in the central area of the east coast urban area, where underground commercial space agglomeration and the explanatory variable demonstrate a “high-high” coupling relationship. This suggests that an increase in permanent resident population density promotes the agglomeration of underground commercial space. However, the regression coefficient decreases progressively from the central area of the east coast urban area toward the southern and northern directions, indicating a gradual reduction in the explanatory power of permanent resident population density regarding the agglomeration of underground commercial space (Figure 13 and Figure 14). From the perspective of the Pearson correlation coefficient, the correlation between permanent resident population density and commercial supporting facilities is 0.672, reflecting a positive relationship. By analyzing the distribution of regression coefficients for both explanatory variables, it becomes apparent that these coefficients generally exhibit opposite trends. This implies that an increase in permanent resident population density primarily enhances the development of above-ground commercial space. Given that underground commercial space represents a non-independent business form dependent on above-ground commercial development, its growth is constrained by the driving force of above-ground commercial activities. Consequently, the consumption potential reflected by population primarily influences the development and agglomeration of above-ground commercial space, which has become the dominant form of commercial supply within the region. Furthermore, the response mechanism of underground commercial space to population factors demonstrates notable limitations. In the interaction analysis, the explanatory power of permanent resident population density is significantly enhanced only when it interacts with commercial facilities and underground space development. This suggests that population factors must be integrated with the current regional development context to achieve multi-dimensional promotion of underground commercial space development. Overall, the population factor must be integrated with regional commercial stock and current supply capacity through spatial planning optimization and dynamic supply–demand regulation mechanisms. This enables precise alignment of underground commercial supply and demand, thereby facilitating multi-dimensional development of underground commercial space.
The absolute values of the regression coefficients for road network density and ground development intensity remain consistently near 0.01, with spatially homogeneous and smooth transitions in the regression coefficients (Figure 15 and Figure 16). Additionally, the bandwidth value of 7210 represents a global bandwidth, and the standard deviation of the regression coefficients is less than 0.002. These findings suggest that the influence of these two factors on the agglomeration of underground commercial space is relatively minor, with their effects being more globally distributed and lacking significant spatial heterogeneity. Consequently, the impacts of these factors are better suited for exploration within macro-oriented models and studies.

5. Discussion and Strategies

5.1. Discussion

This study determines the layout and agglomeration scale of underground commercial spaces in the central urban area of Qingdao using spatial analysis methods such as kernel density estimation and multi-distance spatial clustering. Additionally, it quantifies and visualizes the influencing factors of underground commercial space agglomeration through geographical detector and multi-scale geographically weighted regression (MGWR), analyzing their scope and scale of influence. Compared with prior studies that examined the vitality [18] and business types [19] of underground commercial areas from a micro perspective, this paper investigates the influencing factors of underground commercial areas from a macro-regional perspective, using spatial heterogeneity as the entry point. The findings indicate that in the layout of underground commercial areas, central location, commercial facilities, underground space development, shop rent, land cost, and resident population density exert localized effects, whereas rail transit and population heat values have medium-scale impacts. Additionally, this paper employs the MGWR model to visualize the degree of influence of these factors through regression coefficients and conducts coupling analysis to determine their specific ranges of influence. For example, unlike previous studies that used subway station kernel density as a factor for geographical detection [22], this study incorporates the more granular factor of “distance to the nearest subway station” and extends the analysis by integrating MGWR modeling. This approach clarifies that the influence of subway stations on the agglomeration of underground commercial areas is confined to a 500–600 m station area radius.
From the perspective of its distribution characteristics, underground commercial spaces exhibit a “multi-core, multi-level” layout pattern. Since underground commercial spaces are not independent business forms but are influenced by the radiation effects of above-ground commercial spaces, multiple high-density cores have formed in large commercial districts. Between these cores, medium-density belts of underground commercial spaces have gradually developed. The spatial distribution of underground commercial spaces demonstrates significant positive spatial autocorrelation, forming a “high-high” clustering pattern driven by factors such as rail transit and ground commercial districts. Among the seven types of underground commercial spaces, living support, business support, and catering services are used more frequently in residents’ daily lives and are essential for meeting basic living needs. These three types exhibit higher global Moran’s Index values (0.550, 0.511, 0.501), indicating significantly stronger spatial agglomeration compared to other types of underground commercial spaces. Multi-distance spatial clustering divides the scale of underground commercial space agglomeration into two ranges: 0–3.39 km represents the core scale of underground commercial layout, where accessibility is high, foot traffic is concentrated, and the linkage between above-ground and underground spaces is strong. Therefore, this range requires enhanced construction and optimization of business configuration and spatial connectivity. In the 3.39–12.39 km range, the agglomeration effect weakens with increasing distance, representing the spillover range of the core scale agglomeration effect. Within this range, many medium-density belts of underground commercial spaces have formed, which rely heavily on the radiation and driving effects of rail transit stations and ground commercial districts. However, walkability and attractiveness decrease significantly in this range. During development and construction, it is necessary to anchor on rail transit stations and ground commercial districts to achieve coordinated and three-dimensional development.
The layout of underground commercial spaces results from the synergy of multiple factors acting at different scales. Spatial differences in the distribution of various factors across urban areas fundamentally explain the spatial heterogeneity of the layout and regression coefficients of underground commercial spaces. In the eastern shore urban area, a mature urbanized region, underground facilities, above-ground commercial facilities, and rail transit infrastructure are relatively complete. This area exhibits higher commercial district coverage, population density, and consumption potential compared to the western and northern shore urban areas. The spatial distribution of various factors has transitioned from homogeneous expansion to structural differentiation, leading to a highly significant discrete and regional distribution of regression coefficients for factors such as rail transit, central location, and commercial support in the eastern shore urban area. Interaction detection reveals that 53% of cases belong to the nonlinear enhancement type (EN), indicating that well-developed rail transit and high-coverage ground commercial districts not only form a fully connected network but also interact with other factors. The interaction among factors shifts from linear superposition to nonlinear coupling, resulting in the formation of multiple high-density cores of underground commercial spaces in the eastern shore urban area. In contrast, the western and northern shore urban areas have yet to establish a fully connected network. The absolute values of the regression coefficients of influencing factors in these areas are relatively small, with no obvious spatial differentiation. Rail transit and ground commercial districts play dominant roles, replacing the weights of other influencing factors. Underground commercial spaces in these areas exhibit single-line and isolated layouts around rail transit lines and ground commercial districts. Overall, the agglomeration value of underground commercial spaces directly reflects the density of main influencing factors. Areas with diverse and dense influencing factors exhibit higher degrees of underground commercial space agglomeration, while areas with fewer influencing factors show lower degrees of agglomeration.
Unlike ground-level commercial activities, underground commercial operations are constrained by spatial limitations and rely on the infrastructure provided by ground-level commercial facilities, consumer traffic, and the rapid dispersal capabilities of rail transit. The complementary agglomeration of underground and ground-level business forms suggests that the supply of underground commercial spaces must better align with population demands. Therefore, appropriate commercial support, a well-developed rail transit system, and precise alignment of underground commercial demands are key drivers for the successful commercialization of underground commercial spaces. Simultaneously, planning guidance and regional economic levels play crucial roles in shaping macro policies and determining available economic investments. However, these factors can act as double-edged swords, as the development and construction of underground spaces involve higher costs and relatively lower rental returns. Additionally, future considerations for underground interconnectivity complicate the process further. Given that the development of underground spaces is highly reliant on policy support, delayed underground space planning and underdeveloped regional economies may result in fragmented underground commercial layouts and significantly hinder development efficiency.
As an integral part of underground public spaces, underground commercial spaces present significant challenges to urban planning, decision-making, and governance. This paper proposes that governments and planning departments should take the lead in developing specialized plans for underground commercial spaces. During the planning process, democratic decision-making mechanisms and public participation procedures should be incorporated, starting from the actual needs of the population regarding underground commercial activities and business forms. Based on key factors such as subway stations, commercial supporting facilities, and large commercial districts that significantly drive underground commercial activities, a comprehensive evaluation system should be established. This system should include multiple indicators, such as location selection, functional setting, capacity allocation, and connectivity methods for underground commercial spaces. It should serve as a tool for comprehensively assessing the feasibility of underground commercial space development and promoting its reasonable advancement. Although multi-source data were utilized in this study, many cities lack rail transit coverage; additionally, economic factors, which are among the macro-level influencing factors, were not adequately considered. This presents significant challenges for studying underground commercial activities. Therefore, in cities without rail transit, the combination of “distance to the nearest bus stop” and “taxi GPS trajectory density” can substitute for the “distance to the nearest subway station” indicator to represent accessibility to underground commercial spaces. Simultaneously, the development of underground spaces incurs higher costs compared to above-ground spaces, and the progress of underground commercial activities is greatly influenced by economic conditions. Future research could adopt street- or county-level regions containing economic data as geographical units and use the investment in underground commercial space development (or GDP per unit area) of each unit as an economic indicator to explore the impact of economic investment capacity on the layout and location selection of underground commercial spaces.

5.2. Strategies for Optimizing the Layout of Underground Commercial Spaces

In the central urban area of Qingdao, various factors influencing the layout of underground commercial spaces exhibit strong spatial heterogeneity across the urban space, leading to significant regional differentiation in the development of underground commercial spaces. Therefore, achieving coordinated optimization of different elements in different regions is crucial for transforming underground commercial spaces from “passive dependence” to “active empowerment”.

5.2.1. Business District Agglomeration-Driven Strategy: Activating Underground Commercial Clusters via Large-Scale Business Districts as Spatial Catalysts

Large business districts are characterized by high densities of ground-level commercial facilities, high ground-level development intensity, and relatively high densities of underground facilities. A well-developed built environment can promote the development of underground commercial spaces and enhance the agglomeration of underground commercial facilities. The closer the spatial and temporal proximity to large business districts, the stronger the radiation and driving effects of ground-level commercial activities. Thus, in future underground commercial space development, areas containing large business districts should be prioritized as key development zones. Regional underground commercial space development plans should be formulated to guide their spatial forms, strengthen interactions with ground-level commercial activities, and form underground commercial clusters.

5.2.2. Station–Area Synergistic Development: Building a Passenger–Flow–Business–Space Symbiotic Interface Through Three-Dimensional Transit-Oriented Development

Rail transit exerts a strong guiding effect on the agglomeration of underground commercial spaces but has a limited influence range and requires high coordination between above-ground and underground development. Therefore, it is necessary to further assess the development capacity of underground facilities within station areas, clarify the development conditions of underground commercial facilities, and leverage the efficient passenger flow distribution of rail transit to provide stable customer sources and commercial vitality for underground commercial spaces. This will enhance the accessibility of underground commercial spaces. Moreover, underground commercial spaces based on rail transit can improve public transportation efficiency and achieve coordinated development between rail transit and underground commercial spaces.

5.2.3. Supply-and-Demand Coupling Measurement: Guiding the Flexible Layout of Underground Commercial Spaces via the “Population-Commercial Supply” Assessment Mechanism

Population density reflects commercial demand, while both above-ground and underground commercial facilities contribute to commercial service supply. The balance between supply and demand significantly influences the distribution of underground commercial facilities. Therefore, it is essential to consider the actual commercial activities of urban residents as demand capacity and the total development volume of commercial spaces as supply capacity. Detailed statistics and analysis should be conducted on the resident population and potential consumer groups in relevant areas. From the perspective of balancing supply and demand, the comprehensive accessibility of underground commercial facilities should be determined to provide critical guidance for the spatial location and scale design of underground commercial facilities.

5.2.4. 3D Cadastre Proactivity: Enhance the Top-Level Design by Integrating Forward-Looking Planning with the Development of a Three-Dimensional Cadastre System

Develop a specialized plan for underground commercial space, leveraging the binding force of statutory planning to ensure connectivity and interoperability of underground spaces. This approach provides macro-level flexibility for future underground commercial development. Simultaneously, it is essential to establish a three-dimensional cadastral management system to clearly define the legal rights, boundaries, and spatial extents of underground spaces, which will address the challenges posed by ambiguities in underground space usage rights or property rights that may impede the formulation and implementation of plans.

6. Conclusions

Based on the POI data of underground commercial spaces and other multi-source datasets, this study takes the central urban area of Qingdao as a case example. By employing methods such as kernel density estimation and spatial autocorrelation analysis, it investigates the spatial layout of underground commercial facilities. Furthermore, the geographical detector and MGWR models are utilized to quantify and visualize the influencing factors. The following conclusions are drawn:
(1)
The underground commercial spaces in the central urban area exhibit a “multi-core–multi-level” distribution pattern, with seven core areas and a hierarchical division into “high-density core–medium-density belt–low-density points–nascent areas”. Areas with higher underground commercial densities are more likely to agglomerate, and overall, they display a “high-high” clustering pattern, with the most significant agglomeration occurring at a scale of 3.39 km.
(2)
Commercial supporting facilities, development of underground space, and population heat value serve as the core driving factors for the clustering of underground commercial facilities. The interaction between development of underground space and population heat value exhibits the highest explanatory power, jointly accounting for 52.86% of the spatial layout of underground commercial facilities from the perspectives of underground facility development levels and consumer demand. The explanatory power of other factors significantly increases after interacting with rail transit and central location, indicating that these two factors exhibit the strongest coupling effects with other influencing factors.
(3)
The influences of centrality, commercial supporting facilities, development of underground space, shop rent, land cost, and permanent resident population density on the agglomeration of underground commercial space is predominantly localized; the influences of rail transit and population heat value exert a medium-scale influence on underground commercial space agglomeration. The spatial differentiation of various influencing factors is the fundamental cause of the spatial heterogeneity of underground commercial spaces. Therefore, the development of underground commercial spaces requires forward-looking planning to address these challenges. By establishing and improving a three-dimensional land registration management system, obstacles caused by unclear underground space usage rights can be effectively mitigated. Furthermore, specialized development can be achieved by leveraging the radiation and driving effects of regional advantages such as ground-level commercial areas and rail transit systems.
This study uses subway station POI data to calculate the distance between underground commercial spaces and subway stations. Future research could employ subway station card swipe data or passenger flow data to deepen the analysis from a distance-based perspective to an evaluation of passenger flow capacity at subway stations. Additionally, for cities without rail transit coverage, the combination of “distance to the nearest bus stop” and “taxi GPS trajectory density” can serve as an alternative measure of accessibility for underground commercial spaces. When expanding the research scope to provincial, metropolitan, or national levels, administrative units such as streets, districts, counties, or provinces can be used as geographical units, with economic data associated with these units incorporated to conduct economic analyses.

Author Contributions

Conceptualization, J.Z. and Y.S.; methodology, J.Z., Y.S. and H.W.; software, Y.S. and H.W.; validation, J.Z., Y.S. and Y.Z.; formal analysis, Y.S. and H.W.; investigation, Y.S., H.W. and H.L.; resources, Y.S., H.W. and H.L.; data curation, Y.S.; writing—original draft preparation, J.Z., Y.S. and H.W.; writing—review and editing, J.Z. and Y.S.; visualization, Y.S. and H.W.; supervision, J.Z. and Y.Z.; project administration, J.Z.; funding acquisition, J.Z. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the China Center for Urban and Small-Town Reform and Development Project (Grant No: ZXKJ20190096), the Qingdao Philosophy and Social Sciences Planning Project (QDSKL2401105), and the Qingdao Double Hundred Research Project (2024-B-136).

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors on request.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
GWRGeographically weighted regression
MGWRMulti-scale geographically weighted regression
KDEKernel density estimation
POIPoint of Interest
Ripley’s KMulti-distance spatial clustering
OLSOrdinary least squares
L(d)The observed value
ExpKThe expected value
DiffKThe difference between the observed value and the expected value
EBDouble-factor enhancement
ENNonlinear enhancement
WUSingle-factor nonlinear weakening
R2The goodness of fit
Adjusted R2The adjusted goodness of fit
RSSResidual sum of squares
AICcAkaike information criterion with correction for small sample sizes
VIFVariance Inflation Factor

References

  1. Broere, W. Urban Underground Space: Solving the Problems of Today’s Cities. Tunn. Undergr. Space Technol. 2016, 55, 245–248. [Google Scholar] [CrossRef]
  2. Yi, R.; Yan, H.; Qi, M.; Zhang, Z.; Dong, Z.; Wang, Y.; Jiang, Y.; Zeng, J.; Jia, K. Suitability Evaluation of Urban Underground Space Development Based on Urban Planning. Geol. Explor. 2024, 60, 339–347. [Google Scholar]
  3. Qiao, Y.; Peng, F.; Sabri, S.; Rajabifard, A. Socio-Environmental Costs of Underground Space Use for Urban Sustainability. Sustain. Cities Soc. 2019, 51, 101757. [Google Scholar] [CrossRef]
  4. Alfred, M. Principles of Economics; The Macmillan Company: London, UK, 1938. [Google Scholar]
  5. Alfred, W. Theory of the Location of Industries; J.C.B. Mohr (Paul Siebeck): Tübingen, Germany, 1909. [Google Scholar]
  6. Walter, C. Central Places in Southern Germany; Gustav Fischer Verlag: Jena, Germany, 1933. [Google Scholar]
  7. Wei, Z.; Su, H.; Huang, R. A POI Data-Based Analysis of Commercial Agglomeration Characteristics of Xi’an. J. Southwest Univ. Sci. Ed. 2020, 42, 97–104. [Google Scholar] [CrossRef]
  8. Chen, H.; Yang, D.; Li, J.; Wu, R.; Huo, J. Distribution Characteristics and Influencing Factors of Commercial Center and Hotspots Based on Big Data: A Case of the Main Urban Area of Urumqi City. Prog. Geogr. 2020, 39, 738–750. [Google Scholar] [CrossRef]
  9. Zhou, Q.; Zheng, Y. Research on the Spatial Layout Optimization Strategy of Huaihe Road Commercial Block in Hefei City Based on Space Syntax Theory. Front. Comput. Neurosci. 2023, 16, 1084279. [Google Scholar] [CrossRef]
  10. Zhang, J.; Song, J.; Fan, Z. The Study of Historical Progression in the Distribution of Urban Commercial Space Locations—Example of Paris. Sustainability 2023, 15, 14499. [Google Scholar] [CrossRef]
  11. Li, W.; Zhang, M.; Duan, J. The Research of Nanjing Urban Spatial Pattern Based on POI Data. World Reg. Stud. 2020, 29, 317–326. [Google Scholar]
  12. Tang, Y.; Tang, Y. Analysis of the Spatial Characteristics and Driving Forces of Underground Consumer Service Space in Chinese Megacities Based on Multi-Source Data. Sustain. Cities Soc. 2024, 116, 105924. [Google Scholar] [CrossRef]
  13. Peng, F.; Qiao, Y.; Zhao, J.; Liu, K.; Li, J. Planning and Implementation of Underground Space in Chinese Central Business District (CBD): A Case of Shanghai Hongqiao CBD. Tunn. Undergr. Space Technol. 2020, 95, 103176. [Google Scholar] [CrossRef]
  14. Hu, B.; Li, H. Design of Improving the Business Vigor of Underground Commercial Space. Chin. J. Undergr. Space Eng. 2014, 10, 1575–1579. [Google Scholar]
  15. Li, W.; Yin, F.; Zhang, M. Study on Development and Design of Underground Commercial Street in Old City—Take Jianjun Road Underground Street in Yan City as Example. Chin. J. Undergr. Space Eng. 2015, 11, 541–546. [Google Scholar]
  16. Guo, B. Discussion and Analysis on Designing of Underground Commercial Street. Chin. J. Undergr. Space Eng. 2014, 10, 1571–1574. [Google Scholar]
  17. Ma, G.; Yan, J.; Yang, X. The Research on ActivitIies of Underground Commercial Space of Rail Transit Station Based on Coordinated Development: Analysis of Underground Commercial Space of Jinhui Squara in Tianjin. Mod. Urban Res. 2016, 31, 100–105. [Google Scholar]
  18. Shang, Q.; Meng, F. Factors Affecting the Vitality of Underground Commercial Space. Chin. J. Undergr. Space Eng. 2024, 20, 1073–1085. [Google Scholar] [CrossRef]
  19. Yuan, H.; Zuo, F.; Zhang, L. Commercial Conditions and Realignment Measures for CBD’s Underground Space in Chongqing. Chin. J. Undergr. Space Eng. 2017, 13, 1157–1164+1172. [Google Scholar]
  20. Fan, L.; Cui, X. Analysis of the Factors Influencing the Comfort of Different Behaviors in Underground Commercial Space: A Case Study of Track Connection to Underground Space. Iran. J. Sci. Technol. Trans. Civ. Eng. 2023, 47, 2485–2495. [Google Scholar] [CrossRef]
  21. Shao, J.; Hu, Z. Study on Multi-coupling between Urban Underground and Aboveground. Chin. J. Undergr. Space Eng. 2017, 13, 1431–1443. [Google Scholar]
  22. Tang, Y.; He, L.; Wang, R.; Tang, F.; Zhang, Z. Spatial Distribution Patternand Influence Factors of Underground Commerce in Nanjing. Chin. J. Undergr. Space Eng. 2023, 19, 368–379. [Google Scholar]
  23. Dong, L.; Wei, C. Research on Stereo Design of Traditional Business District in City: Taking the Commercial District Around Kunshan Commercial Building as an Example. Zhuangshi 2022, 139–141. [Google Scholar] [CrossRef]
  24. Fang, H.; Shen, Z.; Yu, B.; Li, Y.; Luo, K. Spatial and Temporal Evolution of Underground Commercial Space in Chengdu Based on POI Data: A Case Study Based on North Railway Station Area, Chunxi Road Area and Global Center Area. South Archit. 2022, 85–93. [Google Scholar] [CrossRef]
  25. Zhang, D.; Xia, H.; Liu, Y. The Application Research and Assessment Method of Underground Commercial Space of Rail Transit. Huazhong Archit. 2019, 37, 59–62. [Google Scholar] [CrossRef]
  26. Yang, Q.; He, D.; Gao, P. Spatial Pattern and Influencing Factor Analysis of Experience Business in Shanghai. Urban Probl. 2018, 34–41. [Google Scholar] [CrossRef]
  27. Qiao, Y.; Peng, F. Advances and Development Thoughts on Three-Dimensional Urban Underground Cadastre. Chin. J. Undergr. Space Eng. 2023, 19, 359–367, 409. [Google Scholar]
  28. Yan, J.; Yin, C.; An, Z.; Zhang, S.; Wen, Q.; Chen, W. Integrated Impacts of Urban Spatial Form on Thermal Environment and Zonal Regulation under the Perspective of Spatial Heterogeneity. Trop. Geogr. 2025, 45, 143–154. [Google Scholar] [CrossRef]
  29. Song, Y.; Gao, M.; Wang, J.; Xu, Z. Spatial Prediction and Influencing Factors Analysis of Soil Salinization in Coastal Area Based on MGWR. Environ. Sci. 2024, 45, 4293–4301. [Google Scholar] [CrossRef]
  30. Hu, Y.; An, R.; Yang, J.; Liu, Y. Measurement and Driving Mechanism of Land Use Conflict Based on Landscape Pattern—A Case Study of Wuhan Metropolitan Area. Res. WaterConservation 2024, 31, 354–364. [Google Scholar] [CrossRef]
  31. The State Council’s Approval of the “Shandong Province Territorial Spatial Planning (2021–2035)” (Guo Han [2023] No. 102). Available online: https://www.gov.cn/zhengce/zhengceku/202309/content_6906387.htm (accessed on 10 March 2025).
  32. The State Council’s Approval of the “Qingdao City Territorial Space Planning (2021–2035)” (Guo Han [2024] No. 169). Available online: https://www.gov.cn/zhengce/zhengceku/202411/content_6986714.htm (accessed on 10 March 2025).
  33. The 2023 China Blue Book on the Development of Urban Underground Space. Available online: http://www.csrme.com/content/article/show/id/4617.do (accessed on 15 March 2025).
  34. Yang, Y.; Li, Q. Distribution Pattern and Lts Formation Mechanism of Public Recreational Space Based on POl Data: A Case Study of the Main Urban Area of Changsha City. Mod. Urban Res. 2021, 91–97. [Google Scholar] [CrossRef]
  35. Yang, Z.; He, X.; Sui, X.; Zhang, J. Analysis of the Evolution of Urban Center Space Based on POI:A Case Study of Main Area in Kunming. Urban Dev. Stud. 2019, 26, 31–35. [Google Scholar]
  36. Liu, W.; Lv, S.; Liang, F. Kernel Density Discriminant Method Based on Geodesic Distance. J. Fuzhou Univ. Nat. Sci. Ed. 2011, 39, 807–810. [Google Scholar]
  37. Chen, P.; Yuan, L.; Sun, Y. Research on the Spatial and Temporal Evolution of Urban Structure Based on POI Data and Nighttime Light Data: A Case Study of Fuzhou City. Mod. Urban Res. 2024, 39, 49–54. [Google Scholar]
  38. Meng, B.; Wang, J.; Zhang, W.; Liu, X. Evaluation of Regional Disparity in China Based on Spatial Analysis. Sci. Geogr. Sin. 2005, 25, 393–400. [Google Scholar]
  39. Zhang, S.; Zhang, K. Comparison between General Moran’ s Index and Getis-Ord General G of Spatial Autocorrelation. ACTA Sci. Nat. Univ. Sunyatseni 2007, 46, 93–97. [Google Scholar]
  40. Jin, C.; Lu, Y. Evolvement of Spatial Pattern of Economy in Jiangsu Province at County Level. ACTA Geogr. Sin. 2009, 64, 713–724. [Google Scholar]
  41. Liu, D.; Hu, J.; Chen, J.; Xu, X. The Study of Spatial Distribution Pattern of Traditional Villages in China. China Popul. Resour. Environ. 2014, 24, 157–162. [Google Scholar]
  42. Li, X.; Li, H.; Man, W.; Mao, D.; Wang, Z. Process and Driving Factors of Urban Land Expansion in Harbin-Changchun City Cluster. Sci. Geogr. Sin. 2018, 38, 1273–1282. [Google Scholar] [CrossRef]
  43. Wang, Z.; Xu, H.; Xu, W.; Lin, J.; Liu, X.; Li, Z.; Zhu, J. Distribution Dynamics of Red-Crowned Crane Population in Zhalong Wetland by the Point Pattern Analysis. Sci. Silvae Sin. 2017, 53, 168–174. [Google Scholar]
  44. Tang, Y.; Wu, Z.; Chen, H. Spatial Characteristics and Influencing Factors of Suzhou’s Catering Industry in the Internet Era. Trop. Geogr. 2022, 42, 1904–1917. [Google Scholar] [CrossRef]
  45. Wang, H.; Wu, J.; Gao, Y.; Liu, L.; Yang, W.; Peng, Z.; Guan, Q. Spatial Fitness of Urban Public Resources and Population Distribution: Taking Shenzhen as an Example. Acta Sci. Nat. Univ. Pekin. 2021, 57, 1143–1152. [Google Scholar] [CrossRef]
  46. Wang, J.; Xu, C. Geodetector: Principle and Prospective. ACTA Geogr. Sin. 2017, 72, 116–134. [Google Scholar]
  47. Lv, C.; Lan, X.; Sun, W. A Study on the Relationship between Natural Factors and Population Distribution in Beijing Using Geographical Detector. J. Nat. Resour. 2017, 32, 1385–1397. [Google Scholar] [CrossRef]
  48. Ding, Y.; Cai, J.; Ren, Z.; Yang, Z. Spatial Disparities of Economic Growth Rate of China’s National-Level ETDZs and Their Determinants Based on Geographical Detector Analysis. Prog. Geogr. 2014, 33, 657–666. [Google Scholar]
  49. Guo, F.; Tong, L.; Qiu, F.; Li, Y. Spatio-Temporal Differentiation Characteristics and Influencing Factors of Green Development in the Eco-Economic Corridor of the Yellow River Basin. ACTA Geogr. Sin. 2021, 76, 726–739. [Google Scholar]
  50. Yang, R.; Luo, X.; Chen, Y. Spatial Pattern and Influencing Factors of Rural Multifunctionality at County Level in China. Prog. Geogr. 2019, 38, 1316–1328. [Google Scholar] [CrossRef]
  51. Liu, Y.; Yang, R. The Spatial Characteristics and Formation Mechanism of the County Urbanization in China. ACTA Geogr. Sin. 2012, 67, 1011–1020. [Google Scholar]
  52. Zhou, L.; Ou, G.; Wang, J.; Xu, H. Light Saturation Point Determination and Biomass Remote Sensing Estimation of Pinus Kesiya Var. Langbianensis Forest Based on Spatial Regression Models. Sci. Silvae Sin. 2020, 56, 38–47. [Google Scholar]
  53. Wang, C.; Luo, J.; Wang, Z.; Zhang, Q.; Xie, L. Spatial Pattern Evaluation and Spatial Optimization of Natural Protected Areas in Taihang Mountains Based on the MGWR Model. Chin. J. Ecol. 2025, 44, 325–336. [Google Scholar] [CrossRef]
  54. Zhu, C.; Zheng, L.; Zhang, Y. Spatial Heterogeneity Analysis of Bikesharing Based on MGWR. Logist. Sci-Tech 2024, 47, 72–77. [Google Scholar] [CrossRef]
  55. Liu, Y.; Yang, Z.; Xu, G.; Liu, B.; Zhang, P.; Chi, J. Impacts of Urbanization on Habitat Quality Using MGWR Models in Wanjiang City Belt. Sci. Geogr. Sin. 2023, 43, 280–290. [Google Scholar] [CrossRef]
  56. Wang, E.; Zhou, J.; Yang, J.; Wang, Y.; Yang, P.; Wang, X. Impact of Built Environment on Spatial Differentiation of Urban Vitality at the Subdistrict Level Based on MGWR: A Case Study of of Shenyang Central Urban Area. Sci. Geogr. Sin. 2024, 44, 1322–1331. [Google Scholar] [CrossRef]
  57. Wang, Y.; Ye, S.; Jin, Y.; Shi, J. A Study on the Underground Commercial Layout of Hangzhou Rail Transit Stations. Railw. Transp. Econ. 2018, 40, 95–100. [Google Scholar] [CrossRef]
Figure 1. Location map of the research area (map review approval number: GS(2024)0650).
Figure 1. Location map of the research area (map review approval number: GS(2024)0650).
Buildings 15 01743 g001
Figure 2. Methodology framework.
Figure 2. Methodology framework.
Buildings 15 01743 g002
Figure 3. Kernel density analysis of underground commercial space.
Figure 3. Kernel density analysis of underground commercial space.
Buildings 15 01743 g003
Figure 4. Analysis of the Local Moran’s I.
Figure 4. Analysis of the Local Moran’s I.
Buildings 15 01743 g004
Figure 5. The multi-distance spatial clustering results of underground commercial spaces.
Figure 5. The multi-distance spatial clustering results of underground commercial spaces.
Buildings 15 01743 g005
Figure 6. The correlation between the agglomeration value of underground commercial spaces and the respective influencing factors.
Figure 6. The correlation between the agglomeration value of underground commercial spaces and the respective influencing factors.
Buildings 15 01743 g006
Figure 7. The spatial distribution of the regression coefficient and the bivariate color plot for rail transit.
Figure 7. The spatial distribution of the regression coefficient and the bivariate color plot for rail transit.
Buildings 15 01743 g007
Figure 8. The spatial distribution of the regression coefficient and the bivariate color plot for centrality.
Figure 8. The spatial distribution of the regression coefficient and the bivariate color plot for centrality.
Buildings 15 01743 g008
Figure 9. The spatial distribution of the regression coefficient and the bivariate color plot for commercial supporting facilities.
Figure 9. The spatial distribution of the regression coefficient and the bivariate color plot for commercial supporting facilities.
Buildings 15 01743 g009
Figure 10. The spatial distribution of the regression coefficient and the bivariate color plot for development of underground space.
Figure 10. The spatial distribution of the regression coefficient and the bivariate color plot for development of underground space.
Buildings 15 01743 g010
Figure 11. The spatial distribution of the regression coefficient and the bivariate color plot for shop rent.
Figure 11. The spatial distribution of the regression coefficient and the bivariate color plot for shop rent.
Buildings 15 01743 g011
Figure 12. The spatial distribution of the regression coefficient and the bivariate color plot for land cost.
Figure 12. The spatial distribution of the regression coefficient and the bivariate color plot for land cost.
Buildings 15 01743 g012
Figure 13. The spatial distribution of the regression coefficient and the bivariate color plot for permanent resident population density.
Figure 13. The spatial distribution of the regression coefficient and the bivariate color plot for permanent resident population density.
Buildings 15 01743 g013
Figure 14. The spatial distribution of the regression coefficient and the bivariate color plot for population heat value.
Figure 14. The spatial distribution of the regression coefficient and the bivariate color plot for population heat value.
Buildings 15 01743 g014
Figure 15. The spatial distribution of the regression coefficient and the bivariate color plot for road network density.
Figure 15. The spatial distribution of the regression coefficient and the bivariate color plot for road network density.
Buildings 15 01743 g015
Figure 16. The spatial distribution of the regression coefficient and the bivariate color plot for ground development intensity.
Figure 16. The spatial distribution of the regression coefficient and the bivariate color plot for ground development intensity.
Buildings 15 01743 g016
Table 1. The results of the global Moran’s I for underground commercial spaces.
Table 1. The results of the global Moran’s I for underground commercial spaces.
TypeUnderground Commercial SpacesCatering ServicesShopping ServicesScience,
Education, and Cultural Services
Accommodation
Services
Business Support ServicesLife
Support Services
Leisure and Entertainment
Services
Global Moran’s I0.8190.5010.4700.4790.0880.5110.5500.105
p-Value0.000.000.000.000.000.000.000.00
Z-Score203.158.0755.3948.057.0648.7465.997.959
Table 2. Influencing factors and corresponding descriptions for the layout of underground commercial spaces.
Table 2. Influencing factors and corresponding descriptions for the layout of underground commercial spaces.
DimensionsInfluencing FactorDescription
Location
and
Transportation
Rail transit (×1)Distance from the grid centroid to the nearest subway station
Road network density (×2)The ratio of road network length to the buffer zone area within a
500 m walking buffer
Centrality (×3)Shortest distance from the grid centroid to a
major commercial district 1
Functional SpaceCommercial supporting
facilities (×4)
The ratio of road network length to the area of the
500 m walking buffer
Development of
underground space (×5)
The ratio of the number of underground POIs to the area of the
500 m walking buffer
Development StatusGround development
Intensity (×6)
The ratio of built-up areas to the area of the
500 m walking buffer
Shop rent (×7)Average shop rent within the grid
Land cost (×8)Average transaction price per unit area for new housing
within the grid
Population FactorsPermanent resident
population density (×9)
The ratio of permanent resident population to the area of the
500 m walking buffer
Population heat value (×10)The average population heat value within the
500 m walking buffer
1 The eight major commercial areas in Qingdao comprise JinJiaLing Commercial District, Zhengyang Road Commercial District, Hong Kong Central Road Commercial District, Zhongshan Road Commercial District, LiCun Commercial District, TaiDong Commercial District, JiMiYa Commercial District, and XinDuXin Commercial District.
Table 3. Factor detection results of the geographical detector.
Table 3. Factor detection results of the geographical detector.
Factor×1×2×3×4×5×6×7×8×9×10
q-value0.00510.01040.07010.26530.38900.06320.00870.01760.11170.2801
p-value<0.001<0.001<0.001<0.001<0.001<0.001<0.001<0.001<0.001<0.001
Table 4. The interaction detection results of the geographical detector.
Table 4. The interaction detection results of the geographical detector.
Factor×1×2×3×4×5×6×7×8×9×10
×10.0551EBENENEBENENENENEB
×20.06380.0104ENENEBENENEBENEB
×30.13370.09260.0701EBEBENENENENEB
×40.32060.37070.32100.2653EBEBENENEBEB
×50.43590.39860.42860.51900.3890EBENENEBEB
×60.12330.08230.13630.32740.44530.0632ENENEBEB
×70.0730.01970.09270.30000.40080.07990.0087EBENWE
×80.07510.02550.09840.28990.42990.09030.02220.0176ENEN
×90.17460.13940.23400.33640.45640.15980.12780.14400.1117EB
×100.3190.28860.31250.35690.52860.30500.26110.30130.31530.2801
Table 5. Comparison of model results: OLS, GWR, and MGWR.
Table 5. Comparison of model results: OLS, GWR, and MGWR.
ModelOLSGWRMGWR
R20.4180.6770.859
Adjusted R20.4170.6760.834
RSS419923321016
AICc16,58812,6598872
Table 6. Standardized regression coefficients and bandwidths for explanatory variables in MGWR.
Table 6. Standardized regression coefficients and bandwidths for explanatory variables in MGWR.
Explanatory VariablesMeanMinimumMaximumStandard
Deviation
Bandwidth
(GWR)
Bandwidth
(MGWR)
VIF
Rail transit−0.615−1.3250.2160.46072211151.564
Road network density0.0020.0000.0050.00172272101.091
Centrality0.421−1.8535.5281.317722462.050
Commercial supporting facilities0.538−0.8765.2300.934722472.654
Development of underground space0.798−4.4915.2501.097722481.332
Ground development intensity−0.017−0.019−0.0130.00272272101.730
Shop rent0.028−1.6053.9100.492722501.596
Land cost0.087−4.7234.9361.5577221452.149
Permanent resident population density−0.144−2.8832.1630.608722483.085
Population heat value−0.109−1.0430.0990.2397224062.393
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhao, J.; Wang, H.; Sun, Y.; Li, H.; Zhu, Y. Investigation into the Distribution Features and Determinants of Underground Commercial Spaces in Qingdao City. Buildings 2025, 15, 1743. https://doi.org/10.3390/buildings15101743

AMA Style

Zhao J, Wang H, Sun Y, Li H, Zhu Y. Investigation into the Distribution Features and Determinants of Underground Commercial Spaces in Qingdao City. Buildings. 2025; 15(10):1743. https://doi.org/10.3390/buildings15101743

Chicago/Turabian Style

Zhao, Jingwei, Heqing Wang, Yu Sun, Haoqi Li, and Yinge Zhu. 2025. "Investigation into the Distribution Features and Determinants of Underground Commercial Spaces in Qingdao City" Buildings 15, no. 10: 1743. https://doi.org/10.3390/buildings15101743

APA Style

Zhao, J., Wang, H., Sun, Y., Li, H., & Zhu, Y. (2025). Investigation into the Distribution Features and Determinants of Underground Commercial Spaces in Qingdao City. Buildings, 15(10), 1743. https://doi.org/10.3390/buildings15101743

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop