Constructing an Evaluation Framework and Developing Optimization Strategies for Public Cultural Service Spaces in Metro Station Areas: A Scene Theory

Abstract

Metro station areas are transitioning from single-purpose transportation transfer nodes into multifunctional public spaces that integrate public services, cultural displays, and place-based experiences. However, there remains a lack of systematic evaluation tools for assessing public cultural service quality in metro station areas that integrate cultural meanings, public identity, and experiential expressions. An Analytic Network Process (ANP) network structure was established based on expert judgments to determine the weights, and Technigue for Order Preference by Similarity to Ideal Solution (TOPSIS) was applied for comprehensive evaluation and ranking using the scores provided by professional evaluators for 13 sampled stations in Foshan, Xi’an, and Shenzhen. Analytic Hierarchy Process (AHP)-TOPSIS was used for robustness testing, and dimensional weight perturbation analysis was conducted to assess model stability. Six types of Point of interests (POIs) were further incorporated to characterize the external functional contexts of the sampled stations. The results showed that the weights of Authenticity, Legitimacy, and Theatricality were 0.498, 0.296, and 0.206, respectively. Traditional, Self-expressive, Charismatic, Exhibitionistic, and Utilitarian were identified as the key indicators with relatively high weights. The 13 sampled stations showed significant differences in overall performance. Higher-performing cases demonstrated more balanced performance across the three scene dimensions of Authenticity, Legitimacy, and Theatricality. In contrast, lower-performing cases showed relatively weaker performance on some key indicators. The robustness test based on AHP-substituted weights and Spearman correlation analysis verified the stability of the evaluation framework, while the sensitivity analysis showed that the evaluation results remained relatively stable under weight perturbations. This study extends the application of scene theory in metro station area research and provides an evaluation framework and practical reference for optimizing public cultural service spaces in transit-oriented urban environments.

1. Introduction

Against the backdrop of continuous urban population growth and the rapid expansion of metro networks, metro systems have become essential infrastructure for alleviating urban transportation pressure, supporting daily commuting, and guiding urban spatial restructuring [1]. As metro development shifts from large-scale expansion toward quality-oriented improvement, metro station areas are gradually evolving from single-purpose transit interchange nodes into multifunctional public spaces integrating public services, cultural displays, and social interactions. Elements such as cultural walls, public art, Exhibitionistic facilities, and activity spaces have further enabled these areas to serve as important carriers for perceiving urban culture, identifying local characteristics, and shaping place experiences [2]. Therefore, systematically identifying the key factors influencing the quality of public cultural service spaces in metro station areas and developing an operational evaluation framework are of practical necessity.

Existing studies on metro station areas have primarily focused on passenger experience, spatial environments, travel behaviors, and cultural expression. Jasińska and Kłosek-Kozłowska [3] examined the effects of spatial characteristics of underground metro stations on passenger comfort and perceived safety. Yang et al. [4] and Buchunde et al. [5] respectively examined the relationships between spatial environments and travel behaviors from the perspectives of platform waiting area design, station design, and passenger flow organization. Meanwhile, the cultural attributes of metro spaces have gradually attracted increasing attention. Wei et al. [2] used multi-source data to analyze the development characteristics of cultural spaces in the Wuhan metro system and public perceptions of these spaces. Zhang et al. [6] investigated the relationships among cultural landscapes, environmental design, and place perception in the Xi’an metro system. González Martínez [7] examined how metro spaces contribute to urban cultural heritage narratives from the perspective of heritage customization. Studies on urban space evaluation have further demonstrated that factors such as accessibility, land use, environmental experience, and social activities are closely associated with public perceptions and urban vitality [8,9,10,11]. These studies provide an important foundation for understanding the spatial quality and cultural value of metro station areas.

However, existing studies still have three major limitations: (1) Existing studies have mainly focused on spatial design, passenger experience, cultural landscapes, or heritage narratives, while systematic evaluations of public cultural service spaces in metro station areas, which integrate cultural expression, public services, and everyday use, remain limited. (2) Cultural meanings, public identity, and experiential presentation have often been discussed separately, while a theoretical framework that integrates these three aspects into a unified structure remains lacking. (3) Existing evaluation studies have paid relatively limited attention to the potential interdependencies and feedback relationships among different evaluation factors, making it difficult to fully capture the connections among the multidimensional attributes of public cultural service spaces.

Based on these limitations, this study addresses two research questions: First, how can Scene Theory be operationalized into an evaluation framework applicable to public cultural service spaces in metro station areas? Second, what is the relative importance of the three dimensions—Authenticity, Legitimacy, and Theatricality—and their evaluation indicators? How do the overall performances of different sampled stations vary, and what differences exist in their surrounding urban functional environments?

This study aims to develop an evaluation framework for public cultural service spaces in metro station areas and identify the relative importance of different scene dimensions and evaluation indicators. It further compares the overall performance of sampled stations and differences in the performance of individual evaluation indicators, and proposes targeted spatial optimization strategies accordingly. Scene Theory explains cultural meanings, public identity, and experiential expression in urban living environments through three dimensions: Authenticity, Legitimacy, and Theatricality [12]. Based on Scene Theory and related empirical studies, this study developed 15 evaluation indicators through the aggregation of bipolar attributes and operationalized them within the spatial contexts of metro station areas. Building on this framework, an ANP network structure was developed based on the theoretical relationships, and the comprehensive weights of the dimensions and evaluation indicators were calculated through expert judgments. Subsequently, the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) was applied for comprehensive evaluation and ranking based on the scores assigned by professional evaluators to 13 sampled stations in Foshan, Xi’an, and Shenzhen. To examine the stability of the evaluation results, this study conducted robustness and sensitivity analyses. A descriptive spatial comparison of the urban functional environments surrounding the sampled stations was also performed based on six POI categories: road and transportation facilities, industrial facilities, public management and public services, residential areas, green spaces and squares, and commercial services.

The contributions of this study are reflected in three aspects: (1) This study applies Scene Theory to public cultural service spaces in metro station areas characterized by high mobility, short stays, and strong management, thereby expanding its application context in micro-scale transportation public spaces. (2) This study operationalizes Authenticity, Legitimacy, and Theatricality into 15 evaluation indicators and applies ANP to address the interdependencies and feedback relationships among evaluation factors. (3) This study integrates ANP and TOPSIS to establish a continuous evaluation pathway from indicator construction and weight calculation to station comparison and difference analysis, providing a methodological basis for comparative evaluation and differentiated optimization of public cultural service spaces in metro station areas.

The remainder of this paper is organized as follows. Section 2 reviews related studies and develops an evaluation framework based on Scene Theory. Section 3 describes the case selection, the design of the two-round questionnaires, and the ANP, TOPSIS, and related validation methods. Section 4 presents the evaluation weights, robustness and sensitivity analyses, comprehensive evaluation of sampled stations, and descriptive comparisons of the built environments. Section 5 discusses the interpretation and extension of Scene Theory in public cultural service spaces in metro station areas and proposes optimization strategies. Section 6 summarizes the main research findings, theoretical and practical implications, cross-context applicability, as well as research limitations and future directions.

2. Literature Review

2.1. Scene Attributes of Public Cultural Service Spaces in Metro Station Areas

Public cultural service spaces in metro station areas are centered on rail transit stations, integrating station interiors with surrounding public spaces to form urban spatial carriers that accommodate everyday public activities. In Transit-Oriented Development (TOD)-related studies, station areas are regarded as fundamental spatial units where transport accessibility, land-use structures, and pedestrian environments are mutually coupled [13]. The Node-Place-Design framework broadens the evaluation perspective by extending the comprehensive assessment of stations to three dimensions: network functions, place values, and spatial design quality [14]. Together, these two analytical perspectives demonstrate that the public value of metro station areas is not solely dependent on passenger flow organization; spatial use patterns and environmental quality also provide essential foundations for such value. From the perspective of public cultural service spaces, the historical information, regional characteristics, and spatial narrative cues conveyed by cultural landscapes can shape place meanings [15]. Meanwhile, users’ perceptions, emotions, and behavioral practices also influence place identity and spatial attachment [16]. Based on the above discussion, this study considers the scene attributes of public cultural service spaces in metro station areas to be primarily reflected in three aspects: operational order, local identification, and public experience. Specifically, operational order concerns circulation, transfers, and spatial orientation. Local identification is established through cultural symbols, public art, and historical memories to form regional characteristics. Public experience focuses on the spatial atmosphere generated by people’s staying, social interactions, and cultural participation.

2.2. Applicability of Scene Theory for Evaluating Cultural Spaces in Metro Station Areas

Transportation space evaluation has long relied on indicator systems to describe the overall performance of transportation systems and station areas. Yang et al. [17] developed a multi-level indicator system for the integrated design evaluation of transportation infrastructure and urban spaces, which supports the comparison of different planning and design alternatives. Sdoukopoulos et al. [18] systematically analyzed sustainable transport indicator systems and demonstrated that mobility, safety, and accessibility represent key evaluation themes within these systems. At the station-area scale, indicators such as density, land-use mix, street design, and accessibility are also commonly used to explain the relationship between the built environment and travel behaviors [1]. The above indicator systems primarily focus on transportation operations, accessibility, the built environment, and sustainability performance. These approaches provide an important foundation for the quantitative evaluation of station areas. However, public cultural service spaces in metro station areas also involve local cultural meanings, public identity, and experiential expression. These aspects cannot be fully captured by conventional transportation performance and built environment indicators. To address this limitation, this study introduces Scene Theory to develop an evaluation indicator system oriented toward the quality of public cultural service spaces.

Beyond being perceived spaces, public cultural service spaces in metro station areas also encompass public service provision, cultural symbol expression, and the organization of everyday activities. The evaluation of such spaces involves greater complexity and requires a more operational framework. Scene Theory conceptualizes places as multidimensional configurations of meanings embedded in material practices, emphasizing that facilities, activities, symbols, and values jointly shape place characteristics [19]. The advantage of this theory lies in its ability to interpret the material conditions of space together with cultural expression, public norms, and experiential creation within a unified framework.

2.3. Development of Evaluation Dimensions Based on Scene Theory

Based on Scene Theory [12], this study categorizes the scene characteristics of public cultural service spaces in metro station areas into three dimensions: Authenticity, Legitimacy, and Theatricality. Among them, Authenticity emphasizes the role of local culture, historical memories, and identity characteristics in supporting spatial meanings. Legitimacy emphasizes the value foundation of spatial practices being recognized, accepted, and continuously used by the public. Theatricality emphasizes the visibility, interactivity, and experiential presentation of scenes.

Authenticity, Legitimacy, and Theatricality jointly constitute the core dimensions for evaluating public cultural service spaces in metro station areas and form interrelated relationships (Figure 1). The relationship between Authenticity and Legitimacy is reflected in the connection between construction and regulation. Local identity and community participation can enhance the legitimacy and inclusiveness of spatial expressions. This suggests that cultural expressions need to respond to public participation and social acceptance [20]. The relationship between Authenticity and Theatricality is reflected in the interaction between construction and representation. Authenticity originates not only from historical materials and local symbols but is also continuously generated through specific experiential processes [21]. Spatial authenticity can be transformed into experiential value perceived by the public through place atmosphere and everyday practices [22].

The interaction between Legitimacy and Theatricality is reflected in the interplay between regulation and representation. Material environments, performative practices, and audience participation jointly contribute to the presentation of spatial value [23]. Legitimacy primarily concerns whether cultural expressions, public services, and spatial practices can gain recognition and acceptance within specific social norms and value systems. It can be understood as the overall judgment of social actors regarding whether certain actions or arrangements are appropriate and worthy of recognition [24]. Theatricality emphasizes how cultural meanings are perceived and experienced by the public through Exhibitionism, interaction, and spatial representation. This distinction is also consistent with the operationalization of related attributes in metro cultural scenes [25].

The same spatial element in a metro setting may simultaneously possess two types of attributes. For example, a community art Exhibition may reflect Legitimacy through providing opportunities for public Self-expression, while also embodying Theatricality through its publicly visible Exhibitionistic format. Such overlaps reflect the multiple scene attributes of the same spatial practice rather than a conceptual equivalence between the two dimensions. Based on this understanding, this study maintains the primary theoretical attribution of each indicator while allowing for inter-dimensional dependencies and feedback relationships. These relationships are subsequently addressed in the ANP network structure.

2.4. Determination and Operational Definitions of Evaluation Indicators

After determining Authenticity, Legitimacy, and Theatricality as the three evaluation dimensions, this study uses the original three-dimensional structure of Scene Theory and related empirical studies as the theoretical basis [19,25,26]. These dimensions are adopted as the primary categories for indicator extraction. Silver and Clark [19] established 15 pairs of bipolar scene attributes across the three dimensions, with each pair consisting of two opposing value endpoints, resulting in a total of 30 classic value endpoints. Klekotko [27] systematically organized this structure from the perspective of urban scene measurement. Navarro et al. [28] applied the scene framework in an empirical study of Spanish cities. Wang and Zhou [25] applied Authenticity, Legitimacy, and Theatricality and their specific attributes to metro cultural scene research. The above studies provide a coherent theoretical and empirical basis for assigning the 15 indicators to their respective dimensions in this study.

Based on this foundation, this study follows the bipolar dimensional logic of Scene Theory [19]. Two opposing value endpoints within the same continuous dimension are paired and merged while maintaining their original theoretical attribution. For example, locality–globality, Charismatic–ordinary, and traditional–novelty each represent two opposing endpoints of the same scene value continuum. Accordingly, they are not treated as separate evaluation indicators. Through bipolar aggregation, 15 evaluation indicators were ultimately derived. Their operational definitions were developed by integrating the original theoretical definitions, relevant empirical studies, and the spatial characteristics of metro station areas, as shown in

Table 1. Construction and Operationalization of the Evaluation Indicator System.

Dimension	Selection Basis	Original Bipolar Attribute	Selection Basis	Indicator Layer	Definition	Definition Basis
Authenticity (A)	[19,25,26]	Local	[19,25,27,28]	Local (A1)	Place-based identity, local authenticity, local memory, and culturally recognizable characteristics of a place.	[29,30]
		Global
		Ethnic		Ethnic (A2)	Collective cultural identity, ethnic or group-related symbols, shared memory, and culturally distinctive meanings.	[31]
		Non-ethnic
		State		State (A3)	Public authority, institutional order, civic regulation, official identity, and state-related symbolic representation.	[27,32]
		Anti-state
		Corporate		Corporate (A4)	Commercial organization, brand presence, market participation, and corporate involvement in spatial production and service provision.	[33]
		Independent
		Rational		Rational (A5)	Logical order, functional clarity, efficiency, and rule-based organization in spatial or social arrangements.	[25,32]
		Non-rational
Theatricality (B)		Exhibitionistic		Exhibitionistic (B1)	Public display, visibility, performance, self-presentation, and the outward presentation of individual or collective identity.	[28]
		Reserved
		Glamorous		Glamorous (B2)	Aesthetic appeal, visual attraction, stylishness, distinction, and desirable symbolic atmosphere.	[19,34]
		Ordinary
		Neighbourly		Neighbourly (B3)	Social familiarity, friendliness, mutual support, everyday interaction, and community-oriented relations.	[28]
		Distant
		Transgressive		Transgressive (B4)	Unconventionality, experimentation, boundary-crossing, and resistance to ordinary or established norms.	[35]
		Conformist
		Formal		Formal (B5)	Standardization, rule-governed behavior, institutional regulation, orderliness, and formal modes of cultural expression.	[27,35]
		Informal
Legitimacy (C)		Traditional		Traditional (C1)	Historical continuity, inherited meanings, cultural heritage, customs, and established local values.	[30]
		Novel
		Utilitarian		Utilitarian (C2)	Practical usefulness, functional value, service efficiency, convenience, and orientation toward everyday needs.	[27]
		Unproductive
		Egalitarian		Egalitarian (C3)	Equality, equal access, non-hierarchical relations, inclusive participation, and shared opportunities.	[27]
		Particular
		Self-expressive		Self-expressive (C4)	Personal identity, creativity, individuality, cultural attitudes, and opportunities for expressive participation.	[33,36]
		Scripted
		Charismatic		Charismatic (C5)	Symbolic attraction, iconicity, emotional appeal, cultural aura, and the capacity to generate recognition or attachment.	[36]
		Routine

Source: Statistics from the author based on the results.

.

3. Research Design and Methods

3.1. Research Process

The methodological design of this study consists of two components: determining indicator weights and measuring station performance, as shown in Figure 2. Based on Scene Theory, this study established an evaluation framework consisting of Authenticity, Legitimacy, and Theatricality and developed fifteen evaluation indicators. On this basis, this study first obtained the relative importance of each evaluation indicator through expert judgments. The study then obtained the actual performance of the sampled stations across each indicator through station evaluations. Finally, the two types of data were integrated to calculate the comprehensive evaluation results of different public cultural service spaces in metro station areas. Data collection was conducted in two rounds. The first round involved the ANP expert judgment questionnaire, and the second round involved the questionnaire for evaluating typical metro station sites. During the data processing stage, ANP was applied to calculate the comprehensive weights of the evaluation indicators, which were then used as the weight vector in the TOPSIS model. The station evaluation data were then normalized in the TOPSIS procedure, and the distances between each station and the positive and negative ideal solutions were calculated to obtain the comprehensive ranking of the sampled stations. All data processing, weight calculations, TOPSIS analysis, and statistical computations were conducted using Microsoft Excel 2021.

3.2. Case Study Sites and Data Sources

China’s urban rail transit systems have entered a stage of large-scale networked operation, with the number of stations and spatial typologies continuing to expand. As of 31 December 2025, a total of 54 cities in mainland China had operational urban rail transit systems, with 343 operating lines, a total operating mileage of 11,710.3 km, and 6680 stations [37]. During the transition from the Pearl River Delta (PRD) to the Guangdong–Hong Kong–Macao Greater Bay Area (GBA), the region has exhibited distinct patterns of multi-scale governance and differentiated urban development pathways [38]. Studies on the formation of the Pearl River Delta urban region have also indicated that functional divisions and development disparities exist among different cities within the region [39]. Urban renewal and spatial development in Xi’an have long been influenced by the preservation of historical and cultural heritage and the reproduction of cultural values [40]. This study selected 13 metro stations from the urban rail transit systems of Shenzhen, Foshan, and Xi’an as empirical samples (Figure 3) and adopted a multi-case comparative approach for case organization. Sample selection primarily considered urban contexts, station types, data availability, and cross-case comparability. Shenzhen, Foshan, and Xi’an exhibit substantial differences in urban development pathways, cultural backgrounds, and spatial environments. These differences provide diverse contexts for the cross-city comparison of the evaluation framework.

After selecting the three cities, this study selected stations that exhibited relatively distinct differences in cultural expression, public service functions, spatial organization, and scene presentation. This ensured that the cases included stations with prominent cultural identification and Exhibitionistic characteristics. It also included stations primarily serving daily transportation and basic public services, where cultural expression was relatively limited. Multiple stations were selected in each city to avoid representing an entire city with a single station and to enable comparisons of case differences both across cities and within the same city [41,42]. Ultimately, 13 sampled stations were selected, including five in Xi’an, four in Shenzhen, and four in Foshan, as shown in Figure 4. The selected cases were designed to form an exploratory case group with diverse characteristics by incorporating heterogeneity in cultural expression, spatial organization, and public service features. This provides empirical cases for comparing different types of public cultural service spaces in metro station areas and testing the applicability of the evaluation framework.

3.3. Two-Round Questionnaire Design and Data Collection

This study used two rounds of questionnaires to obtain indicator weights and station performance data, respectively. The first-round questionnaire was distributed to experts with more than 10 years of relevant research or practical experience. A total of 12 questionnaires were distributed, and 10 valid responses were obtained. The first-round questionnaire required experts to systematically assess the relative importance and interrelationships among the three dimensions of Authenticity, Legitimacy, and Theatricality and the fifteen evaluation indicators. Since the results were directly used for calculating the ANP network weights, participants were required to have substantial professional experience. Expert identification should generally be defined based on domain knowledge, professional experience, and the specific judgment tasks involved [43]. In studies applying fuzzy AHP to transportation and infrastructure evaluation, some researchers have also used at least 10 years of professional work experience as a criterion for selecting senior experts [44]. Accordingly, this study set the required research or practical experience of the first-round experts at more than 10 years (

Table 2. Expert profiles for the ANP panel.

Number	Professional Title	Work Years	Expertise Area
Expert 1	Senior Engineer	≥10 years	Urban Planning/Urban Design
Expert 2	Professor-level Senior Engineer	≥10 years	Human Geography
Expert 3	Engineer	≥10 years	Urban Planning/Urban Design
Expert 4	Senior Engineer	≥10 years	Architecture/Environmental Design
Expert 5	Senior Engineer	≥10 years	Transportation Planning/Transportation Engineering
Expert 6	Senior Engineer	≥10 years	Transportation Planning/Transportation Engineering
Expert 7	Senior Engineer	≥10 years	Transportation Planning/Transportation Engineering
Expert 8	Professor-level Senior Engineer	≥10 years	Transportation Planning/Transportation Engineering
Expert 9	Senior Engineer	≥10 years	Architecture/Environmental Design
Expert 10	Researcher	≥10 years	Transportation Planning/Transportation Engineering

Source: Statistics from the author based on the results.

). The complete first-round ANP expert questionnaire is provided in Appendix B.

The second round involved a station evaluation survey, with 17 professional evaluators who had more than 5 years of relevant research or practical experience participating in the assessment. The complete second-round station evaluation questionnaire is provided in Appendix C. This round required evaluators to assess the actual performance of familiar stations across the fifteen evaluation indicators based on on-site visits, image materials, design documents, or their overall understanding of the stations. The assessment focused on specific case sites. Unlike the first round, which required judgments on the relative importance of indicators and network dependency relationships, the second-round evaluation required less extensive experience in systematic decision-making. Considering the professional experience requirements reported in relevant expert selection studies [45], this study adopted at least 5 years of relevant research or practical experience as the basic selection criterion for second-round evaluators. The professional backgrounds of the evaluators were mainly concentrated in spatial design-related fields, including environmental design, architectural design, urban planning, and public art. Among them, evaluators with an environmental design background accounted for the largest proportion (

Table 3. Evaluator details (station assessment group).

Number	Professional Background	Work Years	Area of Expertise
Expert 1	Planning and Design Professional	5–10 years	Environmental Design
Expert 2	Planning and Design Professional	5–10 years	Environmental Design
Expert 3	Public Art and Cultural Planning Professional	5–10 years	Environmental Design
Expert 4	Public Art and Cultural Planning Professional	5–10 years	Public Art
Expert 5	Planning and Design Professional	5–10 years	Environmental Design
Expert 6	Public Art and Cultural Planning Professional	5–10 years	Architectural Design
Expert 7	Planning and Design Professional	5–10 years	Architectural Design
Expert 8	Public Art and Cultural Planning Professional	≥10 years	Environmental Design
Expert 9	Planning and Design Professional	≥10 years	Urban Planning
Expert 10	University Faculty/Researcher	≥10 years	Architectural Design
Expert 11	Public Art and Cultural Planning Professional	5–10 years	Environmental Design
Expert 12	Planning and Design Professional	5–10 years	Environmental Design
Expert 13	Rail Transit Professional	≥10 years	Urban Planning
Expert 14	Planning and Design Professional	≥10 years	Urban Planning
Expert 15	University Faculty/Researcher	≥10 years	Architectural Design
Expert 16	Public Art and Cultural Planning Professional	5–10 years	Environmental Design
Expert 17	Public Art and Cultural Planning Professional	≥10 years	Environmental Design

Source: Statistics from the author based on the results.

). This may be attributed to the fact that environmental design emphasizes spatial form, visual expression, and place experience, which are highly relevant to the evaluation of metro scenes in this study. Most evaluators also had practical experience related to metro stations, rail transit public spaces, or station-area environmental design, providing relevant expertise for conducting station evaluations. Because each evaluator was allowed to assess one to three sampled stations, the number of station evaluation records was higher than the number of participants. To avoid speculative judgments on unfamiliar cases, the study required evaluators to rate only one to three sampled stations that they had visited or were sufficiently familiar with.

3.4. Weight Calculation and Comprehensive Evaluation Methods

This study employed a combined ANP and TOPSIS approach for comprehensive evaluation. The combination of multi-criteria weighting methods and TOPSIS has been applied in urban comprehensive evaluation studies [46] In this framework, ANP was used to address the dependency and feedback relationships among evaluation indicators and determine their comprehensive weights, whereas TOPSIS was applied to calculate the comprehensive evaluation values and ranking results of the sampled stations.

3.4.1. ANP-Based Weight Calculation

ANP is suitable for addressing multi-criteria decision-making problems involving interdependencies and feedback relationships among evaluation elements [47]. Based on the established three dimensions, 15 evaluation indicators, and network relationships, this study obtained judgments on the relative importance and interrelationships among the indicators through the first-round expert questionnaire.

• Step 1: Construction of Pairwise Comparison Matrices

The questionnaire used the Saaty 1–9 scale for pairwise comparisons to construct the comparison matrices:

$$\mathbf{A}=(a_{ij}{)}_{n\times n}=\left[\begin{array}{cccc}1& a_{12}& \cdots & a_{1n}\\ a_{21}& 1& \cdots & a_{2n}\\ ⋮& ⋮& \ddots & ⋮\\ a_{n1}& a_{n2}& \cdots & 1\end{array}\right]$$

(1)

where $a_{ij}$ represents the judgment value of evaluation element $j$ relative to evaluation element $j$, and satisfies:

$$a_{ii}=1,\hspace{1em}{a}_{ji}={\displaystyle \frac{1}{a_{ij}}}$$

(2)

• Step 2: Aggregation of Group Judgment Matrices

The geometric mean method was used to aggregate the judgments from multiple experts:

$$\overline{a_{ij}}=({\displaystyle \prod _{k=1}^K}{a}_{ij}^{(k)}{)}^{1/K}$$

(3)

where ${\overline{a}}_{ij}$ represents the judgment value provided by the $k$th expert regarding the relative importance or influence between evaluation elements $i$ and $j$; $k$ represents the number of valid experts; and $\overline{a_{ij}}$ represents the aggregated judgment value after synthesizing expert judgments.

The group pairwise comparison matrix is obtained as follows:

$$\overline{\mathbf{A}}=(\overline{a_{ij}}{)}_{n\times n}$$

(4)

• Step 3: Derivation of Local Priority Weights

The eigenvector method was used to calculate the local priority weights of the aggregated judgment matrices:

$$\overline{\mathbf{A}}\tilde{\mathbf{w}}={\lambda }_{\max } \tilde{\mathbf{w}}$$

(5)

After normalization, the eigenvector was converted into the local priority weights:

$$w_i={\displaystyle \frac{\widetilde{w_i}}{{\displaystyle {\sum }_{j=1}^n}\widetilde{w_j}}}$$

(6)

The following conditions must be satisfied:

$${\displaystyle \sum _{i=1}^n}{w}_i=1$$

(7)

where $w_i$ represents the local priority weight of evaluation indicator $i$ under the corresponding comparison relationship. The local weight vectors obtained from the pairwise comparison matrices collectively form the corresponding submatrices of the ANP unweighted supermatrix.

• Step 4: Construction of the Unweighted Supermatrix

Considering the interdependencies and feedback relationships among evaluation indicators, the local priority weights obtained from the judgment matrices were arranged to form the unweighted supermatrix:

$${\mathbf{W}}_u=\left[\begin{array}{cccc}{\mathit{W}}_{11}& {\mathit{W}}_{12}& \cdots & {\mathit{W}}_{1q}\\ {\mathit{W}}_{21}& {\mathit{W}}_{22}& \cdots & {\mathit{W}}_{2q}\\ ⋮& ⋮& \ddots & ⋮\\ {\mathit{W}}_{q1}& {\mathit{W}}_{q2}& \cdots & {\mathit{W}}_{qq}\end{array}\right]$$

(8)

where ${\mathbf{W}}_{pq}$ represents the local priority weight matrix formed by the evaluation elements in cluster $q$ under the control of cluster $p$. If there is no direct influence relationship between two clusters, the corresponding submatrix is a zero matrix. The complete unweighted supermatrix is presented in

Table A1. Unweighted supermatrix of the ANP model.

ANP	A1	A2	A3	A4	A5	B1	B2	B3	B4	B5	C1	C2	C3	C4	C5
A1	1.000	0.000	0.000	0.000	0.000	0.063	0.075	0.030	0.024	0.046	0.035	0.037	0.053	0.059	0.064
A2	0.000	1.000	0.000	0.000	0.000	0.079	0.094	0.038	0.030	0.057	0.036	0.038	0.055	0.061	0.066
A3	0.000	0.000	1.000	0.000	0.000	0.053	0.063	0.025	0.020	0.038	0.031	0.034	0.048	0.054	0.059
A4	0.000	0.000	0.000	1.000	0.000	0.027	0.033	0.013	0.011	0.020	0.011	0.011	0.016	0.018	0.020
A5	0.000	0.000	0.000	0.000	1.000	0.042	0.050	0.020	0.016	0.030	0.027	0.029	0.042	0.046	0.050
B1	0.098	0.096	0.075	0.034	0.035	1.000	0.000	0.000	0.000	0.000	0.048	0.052	0.074	0.082	0.090
B2	0.081	0.080	0.062	0.029	0.029	0.000	1.000	0.000	0.000	0.000	0.039	0.041	0.059	0.066	0.072
B3	0.037	0.036	0.028	0.013	0.013	0.000	0.000	1.000	0.000	0.000	0.018	0.019	0.027	0.030	0.033
B4	0.032	0.032	0.025	0.011	0.011	0.000	0.000	0.000	1.000	0.000	0.017	0.019	0.027	0.030	0.032
B5	0.041	0.040	0.031	0.014	0.015	0.000	0.000	0.000	0.000	1.000	0.018	0.019	0.027	0.030	0.033
C1	0.068	0.067	0.052	0.024	0.024	0.045	0.054	0.022	0.017	0.033	1.000	0.000	0.000	0.000	0.000
C2	0.027	0.027	0.021	0.010	0.010	0.040	0.047	0.019	0.015	0.029	0.000	1.000	0.000	0.000	0.000
C3	0.045	0.044	0.035	0.016	0.016	0.044	0.052	0.021	0.017	0.032	0.000	0.000	1.000	0.000	0.000
C4	0.091	0.089	0.069	0.032	0.032	0.071	0.085	0.034	0.027	0.051	0.000	0.000	0.000	1.000	0.000
C5	0.058	0.057	0.045	0.021	0.021	0.065	0.077	0.031	0.025	0.047	0.000	0.000	0.000	0.000	1.000

Note: A1–A5 represent the Authenticity criteria: A1 = Local, A2 = Ethnic, A3 = State, A4 = Corporate, and A5 = Rational. B1–B5 represent the Theatricality criteria: B1 = Exhibitionistic, B2 = Glamorous, B3 = Neighbourly, B4 = Transgressive, and B5 = Formal. C1–C5 represent the Legitimacy criteria: C1 = Traditional, C2 = Utilitarian, C3 = Egalitarian, C4 = Self-expressive, and C5 = Charismatic. Source: Compiled by the authors based on the study results.

in Appendix A.

• Step 5: Construction of the Weighted Supermatrix

Since the sub’matrices in the unweighted supermatrix are derived from different local comparison relationships, the sum of each column in the entire matrix does not necessarily equal 1. Therefore, the unweighted supermatrix needs to be weighted according to the relative influence weights among clusters to satisfy the column stochasticity condition.

Let $c_{pq}$ denote the normalized influence weight of cluster $q$ under the control of cluster $p$. The weighted supermatrix can then be expressed as:

$${\mathbf{W}}_w=\left[c_{pq}{\mathbf{W}}_{pq}\right]$$

(9)

where $({\mathbf{W}}_w{)}_{ij}$ represents the element in the $i$th row and $j$th column of the weighted supermatrix.

• Step 6: Derivation of the Limit Supermatrix and Final Weights

Subsequently, continuous power iterations were performed, and the limit supermatrix was obtained when the matrix converged:

$${\mathbf{W}}_{\infty }=\underset{t\to \infty }{\lim }({\mathbf{W}}_w{)}^t$$

(10)

This study extracted and normalized the limit priorities of the nodes corresponding to the three dimensions and 15 evaluation indicators. The final comprehensive weight of the $j$th evaluation indicator can be expressed as:

$$w_j={\displaystyle \frac{{\omega }_j}{{\displaystyle {\sum }_{h=1}^{15}}{\omega }_h}}$$

(11)

The following conditions are satisfied:

$${\displaystyle \sum _{j=1}^{15}}{w}_j=1$$

(12)

where ${\omega }_j$ represents the stable priority value corresponding to the $j$th evaluation indicator in the limit supermatrix, and $w_j$ represents the normalized final comprehensive weight.

Since ANP incorporates the interdependencies and feedback relationships among dimensions and evaluation indicators into the network calculation, the final weights of dimension nodes and indicator nodes are jointly influenced by the entire network structure. This distinguishes ANP from the AHP weight calculation approach based on the assumption of strict hierarchical independence.

3.4.2. TOPSIS-Based Comprehensive Evaluation

TOPSIS determines the relative performance of alternatives by comparing their distances from the positive and negative ideal solutions [48]. All 15 indicators in this study are benefit-type indicators, where higher scores indicate better performance of the corresponding attributes.

• Step 1: Construction of the Decision Matrix

For stations evaluated by multiple evaluators, the average score of the same indicator was used as the original evaluation value. Assume that there are $m$ sampled stations and $n$ evaluation indicators. In this study, $m=13$ and $n=15$, and the original evaluation matrix is expressed as:

$$\mathbf{X}=(x_{ij}{)}_{m\times n}=\left[\begin{array}{cccc}{x}_{11}& x_{12}& \cdots & x_{1n}\\ x_{21}& x_{22}& \cdots & x_{2n}\\ ⋮& ⋮& \ddots & ⋮\\ x_{m1}& x_{m2}& \cdots & x_{mn}\end{array}\right]$$

(13)

where $x_{ij}$ represents the average score of the $i$th sampled station on the $j$th evaluation indicator.

• Step 2. Normalization of the Original Evaluation Matrix

To eliminate the influence of different measurement scales among evaluation indicators on the comprehensive evaluation, the vector normalization method was adopted to standardize the original evaluation matrix. The normalized values were calculated as follows:

$$r_{ij}={\displaystyle \frac{x_{ij}}{\sqrt{{\displaystyle {\sum }_{k=1}^m}{x}_{kj}^2}}}$$

(14)

The normalized evaluation matrix is obtained as follows:

$$\mathbf{R}=(r_{ij}{)}_{m\times n}$$

(15)

where $r_{ij}$ represents the normalized evaluation value of the ith sampled station on the jth evaluation indicator.

• Step 3. Construction of the Weighted Normalized Matrix

The comprehensive weights $w_j$ of the 15 evaluation indicators obtained from ANP were incorporated into the normalized evaluation matrix to calculate the weighted normalized values.

$$v_{ij}=w_j{r}_{ij}$$

(16)

The weighted normalized evaluation matrix is obtained as follows:

$$\mathbf{V}=(v_{ij}{)}_{m\times n}$$

(17)

where $w_j$ represents the ANP comprehensive weight of the $j$th evaluation indicator, and satisfies:

$${\displaystyle \sum _{j=1}^n}{w}_j=1$$

(18)

• Step 4. Determination of the Positive and Negative Ideal Solutions

Since all indicators are benefit-type indicators, the positive and negative ideal solutions are defined as the maximum and minimum values of each indicator, respectively:

$$A^{+}=\{v_1^{+},v_2^{+},\dots ,v_n^{+}\},\hspace{1em}{v}_j^{+}=\underset{1\le i\le m}{\max }{v}_{ij}$$

(19)

$$A^{-}=\{v_1^{-},v_2^{-},\dots ,v_n^{-}\},\hspace{1em}{v}_j^{-}=\underset{1\le i\le m}{\min }{v}_{ij}$$

(20)

where $v_j^{+}$ represents the positive ideal value of the $j$th evaluation indicator, and $v_j^{-}$ represents the negative ideal value of the $j$th evaluation indicator.

• Step 5. Calculation of the Distances from the Positive and Negative Ideal Solutions

The Euclidean distance was used to calculate the distances between the $i$th sampled station and the positive and negative ideal solutions as follows:

$$D_i^{+}=\sqrt{{\displaystyle \sum _{j=1}^n}(v_{ij}-v_j^{+}{)}^2}$$

(21)

$$D_i^{-}=\sqrt{{\displaystyle \sum _{j=1}^n}(v_{ij}-v_j^{-}{)}^2}$$

(22)

A smaller value of $D_i^{+}$ indicates that the station is closer to the positive ideal solution; a larger value of $D_i^{-}$ indicates that the station is farther from the negative ideal solution.

• Step 6. Calculation of the Relative Closeness Coefficient

The relative closeness is calculated as follows:

$$C_i={\displaystyle \frac{D_i^{-}}{D_i^{+}+D_i^{-}}}$$

(23)

where $C_i$ represents the relative closeness of the $i$th sampled station. A value of $C_i$ closer to 1 indicates that the station is closer to the positive ideal solution and has better overall performance in the public cultural service space of the metro station area. Conversely, a value of $C_i$ closer to 0 indicates that its overall performance is closer to the negative ideal solution. Finally, the 13 sampled stations were ranked in descending order according to $C_i$, resulting in the ANP-TOPSIS comprehensive evaluation results.

3.4.3. Robustness Test

To examine the impact of the ANP network relationship settings on the evaluation results, this study constructed an AHP comparison model. The AHP model retained the hierarchical structure of “overall goal–three dimensions–15 evaluation indicators” and did not incorporate the interdependencies and feedback relationships among dimensions and cross-dimensional indicators [47]. Subsequently, the same expert judgment data used in ANP were aggregated using the geometric mean method to calculate the AHP weights, which were then used to replace the ANP weights for the TOPSIS evaluation. The Spearman rank correlation coefficient was used to compare the station rankings obtained from ANP-TOPSIS and AHP-TOPSIS to examine the robustness of the evaluation results.

In AHP, the global weight of the $j$th evaluation indicator is:

$$w_j^{AHP}=w_g^D\times w_{j(g)}^L$$

(24)

where $w_g^D$ represents the weight of the $g$th dimension to which the indicator belongs; $w_{j|g}^L$ represents the local weight of the $j$th indicator within its corresponding dimension $g$.

The AHP weights were used to replace the ANP weights, while the original evaluation matrix, normalization method, and TOPSIS calculation procedure were kept exactly the same to recalculate the comprehensive evaluation results of the 13 sampled stations.

The Spearman rank correlation coefficient was used to compare the two sets of station rankings obtained from ANP-TOPSIS and AHP-TOPSIS. Since the 13 sampled stations in this study had no tied ranks, the Spearman rank correlation coefficient was calculated as follows:

$$\rho =1-{\displaystyle \frac{6{\displaystyle {\sum }_{i=1}^m}{d}_i^2}{m(m^2-1)}}$$

(25)

where $\rho$ represents the Spearman rank correlation coefficient; $m$ represents the number of sampled stations, where $m=13$ in this study; $d_i$ represents the rank difference in the $i$th sampled station between the ANP-TOPSIS and AHP-TOPSIS rankings. The value of $\rho$ ranges from −1 to 1. A larger absolute value indicates a higher degree of consistency between the two rankings.

3.4.4. Sensitivity Analysis

To examine the sensitivity of ANP weight results to variations in scene dimension weights, this study conducted a sensitivity analysis using the one-at-a-time (OAT) perturbation method. This method adjusts the weight of one dimension while maintaining the weight proportions of the other dimensions unchanged, and observes changes in the global weights of evaluation indicators and station rankings.

Let $W_i$ denote the original weight of the kth scene dimension and ${W_i}^{\prime }$ denote its perturbed weight. Here, ${W_j}^{\prime }$ represents the perturbed weight of the $j$th dimension, and $W_j$ represents the original weight of the $j$th dimension. The adjusted weights of the other dimensions are calculated as follows:

$${W_j}^{\prime }=W_j\times {\displaystyle \frac{1-{W_i}^{\prime }}{1-W_i}}$$

(26)

Based on this, the weights of Authenticity, Legitimacy, and Theatricality were adjusted separately, and the global weights of the 15 evaluation indicators were recalculated to compare changes in indicator rankings under different perturbation scenarios. When the weight of an indicator exceeds that of the current highest-weighted indicator, the corresponding weight value is defined as the critical point of rank reversal. This critical point is used to assess the sensitivity of the evaluation framework to changes in dimension weights.

4. Results

4.1. ANP Weight Results

After completing the aggregation of group judgment matrices and the limit calculation of the weighted supermatrix, the comprehensive weights of the three scene dimensions—Authenticity, Legitimacy, and Theatricality—and the 15 evaluation indicators were obtained. The results indicate that the relative importance of different dimensions and indicators varies significantly in the evaluation of public cultural service spaces in metro station areas, as shown in

Table 4. Weights of evaluation indicators for public cultural service spaces in metro station areas.

Dimension	Dimension Weight	Criterion	Global Weight
Authenticity (A)	0.498	Local (A1)	0.027
		Ethnic (A2)	0.032
		State (A3)	0.028
		Corporate (A4)	0.021
		Rational (A5)	0.041
Theatricality (B)	0.206	Exhibitionistic (B1)	0.091
		Glamorous (B2)	0.064
		Neighbourly (B3)	0.063
		Transgressive (B4)	0.070
		Formal (B5)	0.046
Legitimacy (C)	0.296	Traditional (C1)	0.138
		Utilitarian (C2)	0.082
		Egalitarian (C3)	0.078
		Self-expressive (C4)	0.127
		Charismatic (C5)	0.092

Source: Statistics from the author based on the results.

.

At the dimension level, Authenticity received the highest weight (0.498), followed by Legitimacy (0.296), while Theatricality had a relatively lower weight (0.206), indicating that the three scene dimensions differ in their relative priorities within the overall evaluation. It is worth noting that although Authenticity received the highest weight at the dimension level, the top five indicators in terms of global weight were mainly distributed across the Legitimacy and Theatricality dimensions. This indicates that the weight distribution at the dimension level does not fully correspond to the ranking of individual indicators at the indicator level.

4.2. ANP–AHP Robustness and Sensitivity Tests

This study recalculated the evaluation weights using AHP to examine whether the ANP network relationship settings substantially affected the main evaluation results. The results showed that the dimension weights of Authenticity, Legitimacy, and Theatricality under AHP were 0.552, 0.281, and 0.166, respectively. Their relative importance followed the same order as the ANP results, with Authenticity ranking first, followed by Legitimacy and Theatricality. At the indicator level, the two methods showed some differences in weight allocation, indicating that the interdependencies and feedback relationships incorporated in ANP influenced the relative importance of some indicators. Using the ANP-derived global weights, the TOPSIS model was further applied to evaluate the comprehensive performance of the 13 sampled stations. The resulting relative closeness values are presented in

Table 5. TOPSIS results of public cultural service spaces at the sampled metro stations.

City	Metro Line	Sample Station	Distance from the Positive Ideal Solution $({\mathbf{D}}_{\mathbf{i}}^{+})$	Distance from the Negative Ideal Solution $({\mathbf{D}}_{\mathbf{i}}^{-})$	Relative Closeness to the Ideal Solution $({\mathbf{C}}_{\mathbf{i}})$
Xi’an	Line 8	Qingshaonianzhongxin Station	0.018	0.064	0.782
Shenzhen	Line 16	Shuanglong Station	0.020	0.063	0.755
Shenzhen	Line 16	Pingshanwei Station	0.021	0.062	0.750
Shenzhen	Line 16	Tianxin Station	0.026	0.062	0.704
Xi’an	Line 8	Hanyao Station	0.029	0.058	0.669
Xi’an	Line 8	Jingshangcun Station	0.029	0.050	0.633
Shenzhen	Line 16	Universiade Center Station	0.032	0.055	0.629
Foshan	Line 3	Daliang Zhonglou Station	0.033	0.048	0.593
Xi’an	Line 8	Zhiwuyuan Station	0.035	0.049	0.585
Xi’an	Line 8	Mutasixi Station	0.039	0.039	0.500
Foshan	Line 3	Dongping Station	0.041	0.038	0.483
Foshan	Line 3	Shunde OcT Harbour PLUs Station	0.061	0.020	0.242
Foshan	Line 3	Lunjiao station	0.0670	0.016	0.183

Source: Statistics from the author based on the results.

.

Under the assumption of hierarchical independence, AHP determines the global weights of indicators based on the dimension weights and the local weights within each dimension. Therefore, the relatively high weight of the Authenticity dimension was further transmitted to its internal indicators, with Local, Ethnic, and State ranking among the top three indicators in terms of global weight. In contrast, indicators such as Traditional, Self-expressive, Charismatic, and Exhibitionistic received relatively lower global weights under AHP than under ANP (

Table 6. Comparison of indicator weights and rankings between AHP and ANP.

Criterion	ANP Global Weight	Rank	AHP Global Weight	Rank	Ranking Change
Local (A1)	0.027	14	0.163	1	13
Ethnic (A2)	0.032	12	0.154	2	10
State (A3)	0.028	13	0.126	3	10
Corporate (A4)	0.021	15	0.052	9	6
Rational (A5)	0.041	11	0.057	7	4
Exhibitionistic (B1)	0.091	4	0.043	10	−6
Glamorous (B2)	0.064	8	0.056	8	0
Neighbourly (B3)	0.063	9	0.019	14	−5
Transgressive (B4)	0.070	7	0.018	15	−8
Formal (B5)	0.046	10	0.029	13	−3
Traditional (C1)	0.138	1	0.034	12	−11
Utilitarian (C2)	0.082	5	0.037	11	−6
Egalitarian (C3)	0.078	6	0.066	6	0
Self-expressive (C4)	0.127	2	0.071	5	−3
Charismatic (C5)	0.092	3	0.072	4	−1

Source: Statistics from the author based on the results.

). Such differences in indicator-level weights and rankings are methodologically justified by the distinct weighting mechanisms of the two methods. Despite these differences, the final rankings of the 13 sampled stations showed a high degree of consistency when the two sets of weights were separately applied to the same TOPSIS evaluation matrix (Spearman’s ρ = 0.874). This indicates that although the indicator weight structure changed, the overall relative performance of the sampled stations remained relatively stable, as shown in

Table 7. Robustness comparison of station comprehensive evaluation results between ANP-TOPSIS and AHP-TOPSIS.

City	Metro Line	Sample Station	ANP-TOPSISRelative Closeness (Ci)	Rank	AHP-TOPSIS Relative Closeness (Ci)	Rank	Ranking Change
Xi’an	Line 8	Qingshaonianzhongxin Station	0.782	1	0.915	1	0
Shenzhen	Line 16	Shuanglong Station	0.755	2	0.637	7	5
Shenzhen	Line 16	Pingshanwei Station	0.750	3	0.721	5	2
Shenzhen	Line 16	Tianxin Station	0.704	4	0.768	2	−2
Xi’an	Line 8	Hanyao Station	0.669	5	0.743	3	−2
Xi’an	Line 8	Jingshangcun Station	0.633	6	0.730	4	−2
Shenzhen	Line 16	Universiade Center Station	0.629	7	0.621	8	1
Foshan	Line 3	Daliang Zhonglou Station	0.593	8	0.688	6	−2
Xi’an	Line 8	Zhiwuyuan Station	0.585	9	0.592	9	0
Xi’an	Line 8	Mutasixi Station	0.500	10	0.540	10	0
Foshan	Line 3	Dongping Station	0.483	11	0.496	11	0
Foshan	Line 3	Shunde OcT Harbour PLUs Station	0.242	12	0.168	12	0
Foshan	Line 3	Lunjiao station	0.183	13	0.160	13	0

Note: Spearman’s ρ = 0.874. Source: Statistics from the author based on the results.

.

To examine the sensitivity of the evaluation results to changes in dimension weights, this study employed the one-at-a-time (OAT) perturbation method to adjust the weights of the three scene dimensions, namely Authenticity, Legitimacy, and Theatricality, and observed changes in the ranking of the global weights of the 15 evaluation indicators.

The perturbation results for the Authenticity dimension showed that the ranking of the core indicators remained stable even when the weight of Authenticity was continuously increased from its initial value of 0.498. When the weight of Authenticity increased to 0.770, Rational, an internal indicator of the Authenticity dimension, surpassed Traditional and became the highest-weight indicator. Within a relatively wide range of changes in the Authenticity weight, the major indicator rankings remained stable. This indicates that the model exhibited high robustness to perturbations in the Authenticity dimension weight.

The perturbation results for the weight relationship between the Legitimacy and Theatricality dimensions showed a rank reversal when the weight of Legitimacy decreased from 0.296 to 0.217. The global weight of Exhibitionistic exceeded that of Traditional for the first time under this condition. This indicates that the relative weight changes between the Legitimacy and Theatricality dimensions constitute the main sensitivity area affecting the ranking of key indicators.

Overall, the current ANP weight structure remained relatively stable under a wide range of weight perturbations. Core indicators such as Traditional, Self-expressive, and Exhibitionistic were not significantly affected by ordinary levels of weight changes. The evaluation framework exhibits good structural robustness under weight perturbations.

4.3. Comprehensive Evaluation Differences Among Sampled Stations and Descriptive Comparison of Built Environments

To examine potential subjective differences in the scores provided by professional evaluators, this study conducted a quadratic weighted Cohen’s Kappa test on the four sampled stations with repeated evaluations. The results showed that the mean weighted Kappa values for Shuanglong Station, Universiade Center Station, Hanyao Station, and Dongping Station were 0.060, 0.015, 0.106, and 0.031, respectively (

Table 8. Mean quadratic weighted Kappa values.

City	Metro Line	Sample Station	Number of Evaluators	Weighted Kappa
Shenzhen	Line 16	Shuanglong Station	2	0.060
Shenzhen	Line 16	Universiade Center Station	2	0.015
Xi’an	Line 8	Hanyao Station	3	0.106
Foshan	Line 3	Dongping Station	5	0.031

Source: Statistics from the author based on the results.

). The complete unweighted supermatrix is presented in Table A1 in Appendix A. The overall consistency level was relatively low, and noticeable differences existed among some evaluators’ scores. This result suggests that judgments of the same station’s scene attributes by professional evaluators with different backgrounds may be influenced by their individual experiences, professional expertise, and the way they apply scoring scales. The limited number of evaluators may further increase the uncertainty of station evaluations. For stations with repeated evaluations, the mean scores from all evaluators were used to construct the original evaluation matrix for subsequent TOPSIS analysis, thereby reducing the direct influence of individual judgments on the comprehensive evaluation results.

Based on the original evaluation matrix, the ANP-derived global weights were incorporated into the TOPSIS model. A comprehensive comparison of the data in Table 6 shows that highly rated stations were characterized by relatively balanced performance across multiple scene attributes. The advantages of Qingshaonianzhongxin Station mainly stem from the coordinated performance of cultural recognition, public attributes, and spatial order. Shuanglong Station, Pingshanwei Station, and Tianxin Station achieved their comprehensive advantages through stronger visual presentation, experiential attraction, and opportunities for public expression. In contrast, lower-ranked stations exhibited more concentrated weaknesses in local cultural recognition, neighbourhood interaction, Self-expressive, and emotional appeal. Improved transportation functions or the incorporation of isolated cultural decorations alone are insufficient to support high-quality public cultural service spaces. Stations need to establish relatively coordinated scene combinations among cultural content, public services, and experiential expression.

In addition, this study conducted a city-scale spatial overlay analysis between the sampled stations and six POI categories, including Road & Transport, Industrial, Public Services, Residential, Green Space & Squares, and Commercial Services, to describe the urban functional environments surrounding the sampled stations. The spatial overlay results are presented in Figure A1, Figure A2 and Figure A3 in Appendix A.

The POI spatial comparison showed that the sampled stations in Xi’an were surrounded by relatively concentrated urban functions. The Shenzhen samples exhibited an interwoven pattern of transportation, industrial, residential, and public service functions, while the functional environments of the Foshan samples varied considerably. Overall, stations with different evaluation levels were distributed across diverse urban functional environments. No stable and consistent spatial correspondence was observed between any single POI category and the comprehensive evaluation results. Therefore, this analysis was only used to describe the external functional contexts of the sampled stations and was not intended to infer statistical associations or causal relationships.

5. Discussion

5.1. Interpretation and Extension of Scene Theory in Metro Station Public Cultural Service Spaces

TOD primarily focuses on the integration of public transport and land use, development density, functional mix, and pedestrian accessibility [49]. The Node–Place–Design framework further evaluates station-area development conditions from the perspectives of node functions, place values, and design qualities [49,50]. Place attachment focuses on emotional bonds, identity recognition, and environmental perception between individuals and places [51]. The above theories explain different aspects of station areas, including transport performance, spatial coordination, and psychological connections. Despite these contributions, they provide relatively limited explanations of how cultural content acquires local meanings, gains public recognition, and is transformed into spatial experiences. Scene Theory conceptualizes Authenticity, Legitimacy, and Theatricality as corresponding to cultural meaning foundations, public norm recognition, and experiential expression, respectively. It integrates cultural expression, public services, and public experiences into a unified analytical framework. In this context, Scene Theory provides a more suitable framework for interpreting the metro station public cultural service spaces examined in this study.

The weighting results reveal the high importance of clear and recognizable cultural meaning sources in metro environments characterized by high mobility, short stays, and transport-oriented functions. Indicators such as Traditional, Self-expressive, Charismatic, and Exhibitionistic received relatively high global weights. The results further demonstrate that the quality of public cultural service spaces depends not only on local cultural foundations but also on cultural continuity, public participation, emotional recognition, and information presentation.

The extension of Scene Theory in this study is mainly reflected in the expansion of its application context and analytical scale. The application of Scene Theory is extended from urban neighbourhoods, cultural consumption facilities, and community scenes to metro station public cultural service spaces. The context is further expanded to commuting environments characterized by high mobility, short stays, and strong management. The analytical scale is further refined to micro-scale transit spaces, including concourses, platforms, transfer areas, and entrances and exits. In this study, Scene Theory is extended from cultural scene identification and description to indicator construction, weight calculation, station comparison, and design optimization, providing a new practical pathway for its application in micro-scale transit public spaces.

5.2. Optimization Strategies Based on Evaluation Results

Based on the ANP weighting results and TOPSIS evaluation results, the optimization of public cultural service spaces at metro stations should not adopt a unified design template. Intervention priorities should instead be determined according to dimension weights, indicator weaknesses, and spatial and functional characteristics. For highly rated stations, multidimensional qualities should be maintained in a balanced manner, while scene vitality should be sustained through content updates and operational activities. Medium-rated stations should implement targeted improvements for low-scoring indicators. Low-rated stations should prioritize strengthening cultural recognition and basic public services, rather than merely adding visual decorations before establishing fundamental qualities.

Authenticity-oriented optimization should be based on locality and cultural continuity. Before design implementation, the origins of station names, surrounding histories, collective memories, and representative cultural resources should be identified and analysed. These elements should then be translated into recognizable spatial narratives rather than directly reproducing traditional patterns. Specifically, continuous cultural cues can be created across entrances and exits, concourses, transfer passages, and external public spaces, allowing materials, colours, signage information, public art, and Exhibition content to collectively express the identity of the station’s locality. Public art can serve as a medium for placemaking, strengthening spatial characteristics and community identity, but it cannot replace the improvement of overall spatial quality [52].

Legitimacy-oriented optimization should focus on addressing high-weight indicators, including Traditional, Self-expressive, Charismatic, Utilitarian, and Egalitarian. Without interfering with pedestrian flow, safety, and operational management, cultural displays can be integrated with signage, waiting areas, rest spaces, public services, and accessible facilities. Such integration can enhance the everyday usability of cultural services. Flexible Exhibitions, community artwork displays, and small-scale interactive nodes can also be introduced to provide opportunities for participation and expression among different groups.

Theatricality-oriented optimization should emphasize the perceptual interface role of Exhibitions. Given the short stay duration and high information density of metro environments, clear visual hierarchies and narrative sequences should be established. Distinctive cultural information should be placed at key turning points, transfer nodes, and waiting areas. Experiential qualities can be enhanced through lighting, digital media, and interactive installations. The amount of information should be controlled to avoid competing with the wayfinding system. Overall, optimization strategies should follow the principle that Authenticity establishes cultural foundations, Legitimacy coordinates public use, and Theatricality enhances perceptual communication. Regular evaluations should be conducted to continuously adjust design contents and operational approaches.

6. Conclusions

6.1. Main Findings

This study developed an evaluation framework for public cultural service spaces at metro stations based on Scene Theory, consisting of three dimensions—Authenticity, Legitimacy, and Theatricality—and 15 evaluation indicators. The framework was empirically applied to 13 sampled stations in Foshan, Xi’an, and Shenzhen using ANP and TOPSIS.

The results showed that Authenticity had the highest weight at the dimension level, followed by Legitimacy, while Theatricality received a relatively lower weight. At the indicator level, Traditional, Self-expressive, Charismatic, Exhibitionistic, and Utilitarian constituted the key indicators with relatively high global weights. ANP-TOPSIS and AHP-TOPSIS produced highly consistent station rankings. The evaluation results remained robust to changes in the weighting method. Weight sensitivity analysis revealed that key indicator rankings remained generally stable under a wide range of dimension weight perturbations. The evaluation results were not significantly affected by changes in a single dimension weight.

Station comparisons revealed significant differences in comprehensive performance among the sampled stations. Highly rated stations generally exhibited coordinated combinations of cultural recognition, public services, and experiential expression, whereas lower-rated stations showed concentrated weaknesses in specific indicators. Within this analytical scope, POI analysis was used only to describe the external functional contexts of the sampled stations and did not infer statistical associations or causal relationships.

6.2. Theoretical, Methodological, and Practical Contributions

This study applies Scene Theory to metro station public spaces characterized by high mobility, short stays, and strong management. Authenticity, Legitimacy, and Theatricality are used to examine local cultural foundations, public recognition, and spatial experiences, respectively. The application of Scene Theory is extended from identifying and describing urban cultural characteristics to an evaluation framework for public cultural service spaces at metro stations, demonstrating the contextual expansion of its applicability.

Methodologically, this study operationalized Scene Theory indicators and integrated ANP and TOPSIS to develop an evaluation approach linking indicator construction, network relationship modelling, weight calculation, station comparison, and comparative analysis. Bipolar attributes were consolidated into 15 evaluation indicators. ANP was used to capture the interdependencies and feedback relationships among indicators, and TOPSIS was applied to compare the comprehensive performance of different stations. Such an approach provides methodological support for identifying differences in station performance and determining optimization priorities.

From a cross-city and cross-country application perspective, the proposed evaluation framework provides a structured benchmarking basis for other metro station public cultural service spaces. For cross-context applications, several benchmarking questions can be considered: whether station cultural expressions reflect local histories, collective memories, and cultural identities; whether public cultural services consider everyday use, equal participation, and public expression; whether cultural content can be clearly identified and effectively perceived in metro environments characterized by high mobility and short stays; and which scene dimensions and evaluation indicators contain the main advantages and weaknesses compared with similar stations. The framework can provide references for evaluating public cultural service spaces at metro stations in different cities and countries and can be adaptively adjusted according to specific contexts.

In addition, the evaluation framework can provide decision-making references for different stakeholders. The specific beneficiaries and their approaches are presented in

Table 9. Main Stakeholders and Their Potential Benefits in This Study.

Stakeholder	Applicable Research Outcome	Specific Beneficiary Approach
Urban rail transit operators	15 evaluation indicators, weighting results, and station evaluation results	Identifying weaknesses in cultural displays, public services, and spatial experiences to provide a basis for station maintenance, content updates, and operational optimization
Urban planning, transport, and cultural management departments	Comprehensive evaluation framework, station rankings, and comparative analysis of differences	Assisting in identifying the construction and operational needs of different stations, and providing references for public cultural service resource allocation and related policy formulation
Planning, architecture, and environmental design institutions	Three-dimensional evaluation structure, key indicator weights, and station evaluation results	Determining priorities for cultural recognition, public services, public participation, and experiential expression in design and renovation projects, supporting differentiated optimization
Urban public cultural institutions and community organizations	Key indicators including Traditional, Self-expressive, Charismatic, and Exhibitionistic	Clarifying intervention directions for local cultural displays, community participation, and public expression, and providing references for Exhibitionistic, public art, and cultural activity organization
Metro passengers and other space users	Public cultural service spaces after evaluation and optimization	Potentially gaining clearer local cultural recognition, improved public services, more inclusive participation opportunities, and better spatial experiences

Source: Compiled by the authors based on the study results.

.

6.3. Limitations and Future Research

Although this study provides meaningful insights, several limitations remain. The empirical analysis selected 13 sampled stations in Foshan, Xi’an, and Shenzhen, covering different urban contexts and station types, while the sample size remains limited. Differences existed among evaluators’ scores for some stations. Individual professional experience, scoring scale interpretation, and the limited number of evaluators may increase the uncertainty of station evaluation results. Future research can expand the sample of professional evaluators and improve the balance of evaluator numbers across different stations to enhance the stability of evaluation results.

This study constructed the evaluation indicator system based on Scene Theory and related studies and used ANP to address the interdependencies and feedback relationships among indicators. Although AHP was employed as a comparative model to examine the stability of evaluation results under different weighting structures, it mainly verifies the consistency of comprehensive rankings and cannot fully reveal the specific mechanisms of interactions among indicators. Future research can incorporate larger samples of experts and public participants to further examine differences in perceptions of indicator relationships among different evaluation groups and explore the dynamic changes in interactions among indicators.

POI analysis was only used to describe the external functional contexts of the sampled stations. Quantitative indicators such as POI density, land-use mix, green space ratio, and public service accessibility were not incorporated, and their statistical relationships with station comprehensive evaluations were not examined. Future research can incorporate buffer-based built environment indicators and apply correlation, regression, or spatial statistical methods to further verify the relationship between the two.