A Heritage-Oriented Evaluation Framework for Differentiated Conservation of Mountainous Traditional Villages: A Case Study of Three Villages in the Southern Taihang Mountains, China

Abstract

Mountainous traditional villages may possess valuable inherited landscapes and cultural heritage while differing markedly in their capacity to maintain, interpret, and appropriately use these resources. This mismatch means that a lower overall evaluation score does not necessarily indicate weaker conservation value, and that villages within the same region may require fundamentally different conservation priorities. To explore this issue, this study developed a heritage-oriented resource–service diagnostic framework and applied it to Xiaodianhe, Liyu, and Xuebaizhuang villages in the Southern Taihang Mountains, China. A three-criterion, 15-indicator system combined expert-derived AHP weights with five-grade assessments from 180 respondents, supported by field investigation, UAV imagery, and resource inventories. Traditional heritage resources received the highest criterion weight (0.5490), while the overall scores of Xiaodianhe, Liyu, and Xuebaizhuang were 4.163, 3.751, and 3.168, respectively. More importantly, these scores concealed three distinct conservation situations: Xiaodianhe showed relative resource–service coordination; Liyu retained strong inherited resources but was constrained by targeted deficiencies in interpretation, internal routes, and stay-support conditions; and Xuebaizhuang retained resource potential despite substantial gaps in accessibility, management, and community participation. The comparison demonstrates that differentiated conservation should respond to the internal relationship between inherited resources and service-support conditions rather than to composite rankings alone. Accordingly, conservation actions should be sequenced according to whether a village primarily requires heritage-quality maintenance, targeted service improvement, or basic support-capacity enhancement, while safeguarding authenticity, integrity, landscape continuity, and community continuity.

1. Introduction

Traditional villages are living cultural landscapes shaped by long-term interactions among topography, hydrology, settlement form, vernacular construction, agricultural production, local knowledge, and everyday community life [1]. Their conservation is therefore important not only for retaining historic buildings and spatial patterns, but also for sustaining regional identity, cultural continuity, and locally adapted human–environment relationships. Mountainous traditional villages are particularly vulnerable because terrain constraints, population decline, infrastructure modernization, tourism development, and changing livelihoods can simultaneously affect heritage integrity and the conditions required for continued use.

Contemporary heritage conservation involves several related debates. One concerns the balance between material authenticity and living continuity. Strict fabric-centred protection may restrict necessary adaptation, whereas excessive functional transformation may weaken authenticity, integrity, and regional character [2,3]. A second debate concerns conservation and development: improved access, interpretation, maintenance, and local economic opportunities may support continued heritage use, but standardized construction and tourism-oriented commercialization can produce spatial homogenization. A third concerns decision-making authority. Experts, residents, visitors, and professional-background respondents (P) may evaluate the same village differently because they emphasize heritage value, everyday use, cultural attachment, or service experience to different degrees [4,5,6]. Heritage-oriented evaluation must therefore distinguish inherited resources, public perceptions, service-support needs, and planning decisions rather than treating them as interchangeable.

Existing traditional-village studies have evaluated heritage value, tourism facilities, cultural inheritance, socio-ecological patterns, spatial accessibility, public-space quality, and residents’ everyday landscapes [7,8]. Fang and Li classified villages according to cultural-ecological heritage values [9], while Wang et al. examined how tourism facilities affect living protection [10]. Tang et al. evaluated cultural inheritance under tourism development [11], Jiang et al. developed socio-ecological protection types [12], Zhang et al. analysed regional distribution and accessibility [13], and Li et al. interpreted heritage value through residents’ daily lives and the Historic Urban Landscape approach [14]. These studies provide important conceptual and methodological foundations, but their principal outputs are generally comprehensive scores, rankings, spatial types, or dimension-specific recommendations. They rarely retain inherited resources and service-support conditions as separate decision dimensions after aggregation or use their mismatch to determine the sequence of conservation actions. A detailed comparison is provided in Supplementary Table S1.

Alternative analytical methods also serve different purposes. Entropy weighting derives weights from data variation, but statistical dispersion does not necessarily represent conservation importance, particularly when only three villages are compared. TOPSIS is effective for ranking alternatives by distance from an ideal solution, yet compensatory aggregation may allow strengths in one dimension to offset serious deficiencies in another. Principal-component and factor analyses are better suited to identifying latent structures in large datasets, while cluster analysis requires substantially more cases for statistically meaningful classification. In contrast, AHP was selected in this study to derive expert-based weights within the predefined hierarchical indicator system, while fuzzy comprehensive evaluation was used to aggregate the five-grade questionnaire distributions. The methodological rationale, procedural steps, and supporting literature for AHP are detailed in Section 2.4.1. Their combination is appropriate for a predefined hierarchical system, a small comparative sample, and five-grade public assessments because it keeps expert-based weighting analytically distinct from questionnaire-based condition assessment while retaining criterion-level information for subsequent resource–service diagnosis. The methodological contribution of this study does not lie in combining these established methods, but in retaining criterion-level information for subsequent resource–service diagnosis.

Against this background, this study develops a heritage-oriented resource–service diagnostic framework for Xiaodianhe, Liyu, and Xuebaizhuang villages in the Southern Taihang Mountains. Natural and landscape resources and traditional heritage resources are treated as the inherited resource base, whereas accessibility, internal routes, accommodation, management, catering, and community participation are treated as service-support conditions. The study addresses three questions: RQ1, what relative importance is assigned to the three criterion layers and their indicators? RQ2, how do the three villages differ in overall scores, criterion structures, resource–service relationships, and high-weight strengths and deficiencies? RQ3, how can these structural profiles inform sequenced and village-specific conservation priorities?

The study makes three contributions. First, it shifts evaluation from composite ranking toward structural diagnosis by retaining resource and service dimensions after aggregation. Second, it combines expert-derived weights with five-grade public assessments while preserving criterion- and indicator-level information. Third, it translates resource–service mismatches into phased, low-impact conservation priorities constrained by authenticity, integrity, landscape continuity, and community continuity. The results identify Xiaodianhe as relatively resource–service coordinated, Liyu as resource-advantaged but affected by targeted service deficiencies, and Xuebaizhuang as possessing resource potential constrained by substantial service gaps. These findings indicate that differentiated conservation should respond to internal resource–service structures rather than composite scores alone.

The remainder of this paper is organized as follows. Section 2 introduces the study area and case-selection rationale and describes the evidence sources, indicator system, AHP-weighted fuzzy comprehensive evaluation, resource–service structural diagnosis, and robustness procedures. Section 3 presents the evaluation results, including the weight structure, criterion- and target-layer evaluations, robustness analyses, cross-village resource–service differences, diagnostic profiles, and the resulting conservation priorities. Section 4 discusses the findings in relation to the heritage-oriented decision logic and previous studies, examines their implications for differentiated conservation, and addresses the limitations, transferability, and directions for future research. Finally, Section 5 summarizes the principal findings, methodological contribution, and conservation implications.

2. Study Area and Methods

This section describes the study area, case-selection rationale, evidence sources, participant recruitment, indicator system, AHP-weighted fuzzy comprehensive evaluation, resource–service diagnostic measures, and robustness procedures. Field investigation, UAV imagery acquired using a DJI Mini 3 (SZ DJI Technology, Inc., Shenzhen, China), and resource inventories were used to operationalize and verify the assessment content, whereas expert pairwise comparisons and questionnaire ratings provided the direct numerical inputs. The UAV imagery was used primarily as visual evidence to support field verification and resource identification, and no specialized photogrammetric or three-dimensional modelling software was used for image processing. Questionnaire data organization, AHP matrix calculations, fuzzy comprehensive evaluation, resource–service diagnostic calculations, and robustness analyses were performed using WPS Office 2025 (Beijing Kingsoft Office Software, Inc., Beijing, China). Figure 1 summarizes the analytical workflow, and the following subsections provide the information required to reproduce the case comparison and calculations.

2.1. Study Area and Case Selection

The study adopted purposive case selection rather than statistical sampling. Xiaodianhe, Liyu, and Xuebaizhuang were selected according to four criteria: (1) location within the same township and Southern Taihang mountain system, thereby limiting broad differences in administrative, climatic, and regional cultural contexts; (2) retention of identifiable mountainous settlement forms, vernacular architecture, agricultural landscapes, and local cultural practices; (3) observable variation in heritage preservation, accessibility, service provision, public-space improvement, and community operation; and (4) availability of comparable field observations, UAV imagery, resource inventories, expert judgments, and questionnaire data. The cases were not selected to predetermine the final diagnostic profiles, which were derived from the subsequent criterion- and indicator-level results.

The three villages are located in Shibaotou Township, Weihui City, Henan Province, on the eastern foothills of the Southern Taihang Mountains (Figure 1). Their proximity provides a broadly comparable geographical, administrative, and regional-policy context, while their differences in inherited resources and service-support conditions permit controlled comparison. The cases are analytically representative of mountainous villages characterized by mountain–valley terrain, vernacular settlement structures, agricultural cultural landscapes, and uneven service provision, but they are not statistically representative of all traditional villages in China. Application to other regions would require recalibration of the indicators, weights, and diagnostic thresholds.

Liyu covers approximately 520 ha and has an average elevation of about 505 m. Its cultivated land, woodland, water systems, traditional courtyards, stone-paved routes, and the Huanggu’an site form an integrated mountain–water–farmland–settlement landscape. Xiaodianhe is located approximately 26 km from central Weihui and retains a Qing-dynasty residential complex of about 50,000 m², comprising 10 courtyards, 23 quadrangle compounds, and 86 individual buildings. It is designated as a Chinese Traditional Village and was included in the second batch of National Key Rural Tourism Villages in 2020. Xuebaizhuang, situated deeper in the mountains, has a settlement history of more than 660 years and retains a Ming-dynasty stone gate, the Xiulou structure, stone dwellings, and an agricultural landscape historically associated with apple cultivation.

The three cases combine contextual similarity with differences relevant to the proposed framework. Xiaodianhe presents comparatively strong heritage preservation and accessibility; Liyu is characterized by an integrated valley, settlement, architectural, and agricultural landscape; and Xuebaizhuang retains distinctive mountain and stone-built heritage but has comparatively limited accessibility, routine maintenance, and service support. Figure 2 provides representative UAV and field-survey evidence of these conditions. These preliminary descriptions establish the comparative context and should not be interpreted as final evaluation results. The Chinese characters visible on the signage indicate the local village name and public information at the survey site.

2.2. Evidence Sources, Participants, and Quality Control

Candidate indicators were initially identified through a literature review of Web of Science, China National Knowledge Infrastructure, and related academic and policy sources. Searches used terms including “traditional villages”, “village landscapes”, and “heritage conservation”, with priority given to empirical studies, policy documents, research reports, and representative conservation cases published during the preceding decade. The preliminary indicator pool covered natural and landscape resources, traditional heritage resources, accessibility, facilities, management, and community participation. Field investigation, UAV imagery, and resource inventories were subsequently used to verify whether the candidate indicators were observable and applicable to all three villages, while the expert panel reviewed their conservation relevance and practical applicability. In the final screening, three considerations were used jointly: explicit support in the reviewed literature, observability and applicability across all three cases based on field, UAV, and inventory evidence, and expert-assessed conservation relevance and practical applicability. The final indicator system and operational definitions are presented in Section 2.3.

Five complementary evidence streams were used: field investigation, UAV imagery, resource inventories, expert pairwise comparisons, and questionnaire ratings. Field investigation included georeferenced observation, photographic documentation, route inspection, facility inspection, and problem recording. UAV imagery was used to verify settlement morphology, land-use relationships, water and valley systems, vegetation patterns, road connections, building distribution, and spatial continuity. Resource inventories recorded the presence, condition, distribution, and functional status of natural, heritage, and service-support elements. These three evidence streams supported indicator operationalization, preparation of standardized assessment materials, condition verification, and interpretation of the results, but they were not assigned independent coefficients or entered directly into the numerical model. The direct numerical inputs were the expert pairwise comparisons used to determine AHP weights and the questionnaire ratings used to construct fuzzy membership matrices.

The expert panel comprised nine specialists with 15–25 years of relevant research or professional experience (mean: 19.7 years): four professors, two associate professors, two researchers, and one rural-construction management practitioner. Their disciplinary backgrounds included landscape architecture (two experts), architecture (two), urban and rural planning (two), and rural tourism (three), and all nine were familiar with both traditional villages and the study area. The panel was purposively selected to provide complementary expertise across the three evaluation domains: landscape and environmental assessment for B1, vernacular architecture and settlement conservation for B2, and spatial planning, tourism services, management, and community participation for B3. Professional experience, disciplinary relevance, and familiarity with the study context were used as the principal selection criteria. Each expert independently completed four reciprocal pairwise-comparison matrices, and the corresponding judgments were aggregated using the geometric mean after individual consistency testing. Detailed anonymized expert profiles are provided in Supplementary Table S6.

The questionnaire survey was conducted from April to June 2026. Each village contributed 60 valid responses, yielding 180 responses in total. For sampling and subsequent respondent-group analyses, respondents were assigned to three mutually exclusive recruitment categories: (1) residents or respondents with long-term familiarity with local conditions (R/F; n = 30 per village), (2) on-site visitors (V; n = 15 per village), and (3) respondents recruited on the basis of relevant professional or disciplinary backgrounds (P; n = 15 per village). Residents or respondents with long-term familiarity were approached during household and public-space surveys with assistance from local contacts and community organizations. Visitors were recruited at village entrances, historic streets, and principal residential or heritage areas after completing a substantive visit. Professional-background respondents (P) were invited through Henan Institute of Science and Technology and had training or experience in landscape architecture, architecture, rural tourism, urban and rural planning, or related fields. Participation was voluntary and anonymous, and the study purpose and assessment procedures were explained before participation. Ethical review and approval were waived because the study involved only non-interventional, anonymous questionnaire and expert assessments and posed no more than minimal risk to participants.

The professional-background respondents (P) completed the same public-rating questionnaire as the R/F and V groups and did not participate in the AHP weighting process; they were therefore analytically distinct from the nine experts who provided the pairwise comparisons used to derive the AHP weights. All respondents evaluated the current condition of the same 15 indicators using the same five ordered grades: excellent, good, moderate, poor, and very poor. For each village, the frequency distribution of the five grades was converted into an indicator-level fuzzy membership vector. The resulting membership matrices were subsequently combined with the AHP weight vectors, as described in Section 2.4.

Of the 180 valid responses, 159 were based on direct on-site observation and 21 were completed through image-assisted assessment. The image-assisted responses were provided by professional-background respondents (P) who could not complete an on-site assessment, including six for Xiaodianhe, six for Liyu, and nine for Xuebaizhuang. For each village, respondents received a standardized package containing UAV images of overall settlement morphology, ground-level photographs of the principal natural, heritage, and service-support conditions, and a concise factual village profile. The same indicator definitions, image materials, presentation sequence, written instructions, and five-grade scale were used for respondents assessing the same village. Researchers clarified only the meanings of indicators or rating categories when necessary and did not suggest particular scores or comparative judgments.

Respondent-group membership, reported professional background, and assessment mode were treated as separate variables. For all respondent-group analyses, group membership was defined by the three mutually exclusive recruitment categories described above (R/F, V, and P), rather than simply by whether a respondent reported relevant professional training. The sample comprised 81 male and 99 female respondents. In total, 49 respondents reported relevant professional training or experience; this number exceeded the 45 respondents assigned to the P group because four respondents recruited into the R/F or V groups also reported relevant professional experience. Full distributions by village, age, sex, occupation, respondent group, professional background, visit experience, and assessment mode are provided in Supplementary Table S2. The variable definitions and coding used in the respondent-group and assessment-mode robustness analyses are provided in Supplementary Table S7.

The questionnaire was not designed as a psychometric scale measuring a single latent construct. Instead, the 15 indicators were treated as formative and independently interpretable criteria defined by observable and non-interchangeable components. Cronbach’s alpha was therefore not used as the primary measure of validity. Content validity was supported through literature-based indicator development, field investigation, resource inventories, and expert review of indicator relevance and applicability. Accordingly, the indicator, criterion, and target-layer results should be interpreted as questionnaire-derived, evidence-informed comparative perceptions of observable village conditions rather than direct technical audits of building condition, heritage authenticity, ecological performance, tourism-management effectiveness, or governance capacity.

2.3. Heritage-Oriented Resource–Service Framework and Indicator System

The framework evaluates the current resource–service structure of each village rather than heritage value, tourism performance, or conservation effectiveness alone. Natural and landscape resources (B1) and traditional heritage resources (B2) constitute the inherited environmental and cultural resource base. Service-support conditions (B3), including accessibility, internal routes, accommodation, management, catering, and community participation, represent the enabling conditions for maintenance, interpretation, day-to-day access, and appropriate use. The three dimensions are analytically related but not interchangeable: strong service provision cannot compensate for the loss of authenticity, integrity, or landscape continuity, while a low service-support score does not necessarily indicate low heritage value.

Differentiated conservation was guided by four principles: heritage-value priority, landscape continuity, resource–service compatibility, and gradual low-impact intervention. Conservation priorities were therefore derived by first identifying the strength and observable condition of inherited resources and then examining whether service-support conditions were broadly commensurate with that resource base. The framework does not assume that additional facilities or higher composite scores necessarily produce better conservation outcomes.

Figure 3 summarizes the analytical sequence from evidence acquisition and indicator construction to AHP weighting, fuzzy comprehensive evaluation, structural diagnosis, and sequenced conservation priorities. AHP was used to determine the relative importance of the criterion and indicator layers, while fuzzy comprehensive evaluation aggregated the five-grade questionnaire distributions. Criterion-level information was retained after aggregation to identify resource–service gaps, structural imbalance, and high-weight deficiencies.

The framework comprises one target layer, three criterion layers, and 15 indicators (

Table 1. Indicator system, diagnostic roles, and literature basis of the heritage-oriented resource–service framework.

Criteria Layer	Indicator Layer	Indicator Description	Diagnostic Role	Literature Basis
Natural and Landscape Resources (B1)	Mountain Landforms and Visual Landscape (C1)	Mountain landforms, terrain characteristics, and visual landscape quality	Resource base	[15,16]
	Water Bodies and Gully Landscapes (C2)	Streams, ponds, seasonal water features, gullies, and associated landscape conditions	Resource base	[17]
	Vegetation and Ecological Landscape (C3)	Vegetation cover, structural and visual diversity, and landscape continuity	Resource base	[18,19]
Traditional Heritage Resources (B2)	Traditional Settlement Environment (C4)	Traditional settlement patterns, street texture, spatial continuity, and overall village environment	Resource base	[20,21]
	Traditional Mountain Vernacular Architecture (C5)	Traditional dwellings, regional architectural forms, materials, and historic building remains	Resource base	[22,23]
	Mountain Agricultural Landscape and Production Culture (C6)	Terraced fields, cultivated landscapes, farming traditions, and production practices	Resource base	[24,25]
	Folk Activities and Local Cultural Memory (C7)	Local customs, festivals, rituals, collective activities, and cultural memory	Resource base	[26,27]
	Material Cultural Remains (C8)	Historic structures, inscriptions, stone components, archaeological remains, and related material heritage	Resource base	[21,28]
	Local Products and Culinary Resources (C9)	Local agricultural products, food traditions, culinary knowledge, and production techniques	Resource base	[29,30]
Service-Support Conditions (B3)	External Transport Accessibility (C10)	Regional accessibility, travel convenience, and connections to external transport routes	Enabling and support capacity	[31,32]
	Accessibility of Internal Roads and Tourist Routes (C11)	Internal roads, walking routes, route continuity, and local mobility	Enabling and support capacity	[33,34,35]
	Accommodation and Stay Conditions (C12)	Availability, hygiene, comfort, operational condition, and compatibility of accommodation	Enabling and support capacity	[36]
	Tourism Service and Management Capacity (C13)	Reception, interpretation, environmental maintenance, visitor management, and coordination	Enabling and support capacity	[37]
	Catering Provision and Service Conditions (C14)	Catering availability, hygiene, service quality, operational stability, and environmental compatibility	Enabling and support capacity	[38,39]
	Cultural Tourism Operations and Community Participation (C15)	Local operation, resident participation, governance, employment, coordination, and benefit sharing	Enabling and support capacity	[5,6]

). The indicator system was derived from literature-based screening and subsequently refined through field investigation, resource inventories, UAV-supported condition verification, and expert review. To make the academic basis of the framework explicit, Table 1 identifies the principal literature supporting each indicator. B1 and B2 describe the inherited resource base, whereas B3 describes enabling and support capacity. Field investigation, UAV imagery, and resource inventories were used to verify the observable content and applicability of the indicators rather than to assign independent numerical coefficients. Detailed operational definitions, evidence sources, grading guidance, and model-entry procedures are provided in Supplementary Table S3.

RQ1 was addressed through criterion- and indicator-level weights, RQ2 through the criterion-, target-, and structural-diagnostic results, and RQ3 through the translation of these results into village-specific conservation priorities.

2.3.1. Natural and Landscape Resources

Natural and landscape resources describe the inherited environmental setting and its observable landscape condition rather than serving as direct measurements of biodiversity or ecosystem functioning. The mountain–valley terrain, water and gully systems, woodland, cultivated land, and settlement relationships of the three villages provided the empirical basis for evaluating C1–C3.

• Mountain landforms and visual landscape (C1)assess the recognizability of ridges, slopes, valleys, and terrain structure, together with visual openness, spatial depth, landscape layering, and access to viewing points. Higher ratings indicate a clearly legible and visually coherent mountainous setting, while lower ratings indicate weak landform recognition, obstructed views, or intrusive spatial disturbance [15,16].
• Water bodies and gully landscapes (C2)include permanent and seasonal water features, gullies, dry channels, drainage systems, and their associated banks and slopes. The assessment considered continuity, visibility, naturalness, accessibility, visual coherence, and observable disturbances such as sedimentation, litter, excessive hard revetment, or intrusive construction [17].
• Vegetation and ecological landscape (C3)evaluate vegetation coverage, structural and visual diversity, continuity, condition, and coordination with the settlement and surrounding farmland. The resulting score represents respondents’ comparative perception of vegetation and ecological landscape quality rather than a technical ecological or biodiversity assessment [18,19].

2.3.2. Traditional Heritage Resources

Traditional heritage resources comprise the tangible and intangible cultural elements inherited through the long-term development of the villages. These include settlement patterns, vernacular buildings, agricultural landscapes, local customs, material remains, products, and culinary knowledge. The six indicators assess their observable condition, continuity, legibility, and local distinctiveness rather than determining statutory heritage significance.

• Traditional settlement environment (C4)evaluates settlement morphology, street texture, spatial continuity, environmental coordination, and the extent of visually intrusive construction. Higher ratings indicate that the historic settlement structure remains legible and coherent within its landscape setting [20,21].
• Traditional mountain vernacular architecture (C5)assesses the retention, condition, integrity, and recognizability of traditional dwellings, regional construction forms, materials, and historic architectural fabric. The rating reflects observable architectural condition and character rather than a complete professional assessment of heritage significance [22,23].
• Mountain agricultural landscape and production culture (C6)cover terraced fields, orchards, cultivated land, production spaces, and associated agricultural practices. The assessment considered landscape continuity, visibility, continued use, and the recognizability of traditional production activities [24,25].
• Folk activities and local cultural memory (C7)include customs, festivals, rituals, oral traditions, collective activities, and other locally transmitted practices. Ratings focused on visibility, participation, community recognition, and perceived continuity rather than measuring cultural identity as a single latent construct [26,27].
• Material cultural remains (C8)include historic structures, bridges, wells, inscriptions, stone components, archaeological remains, and production-related features. The assessment considered diversity, physical condition, legibility, maintenance, spatial distribution, and relationships with the wider settlement environment [28].
• Local products and culinary resources (C9)include distinctive ingredients, agricultural products, processed goods, traditional dishes, preparation knowledge, and associated local practices. Ratings considered diversity, local distinctiveness, availability, continuity, authenticity, and presentation [29,30].

2.3.3. Service-Support Conditions

Service-support conditions describe the enabling capacity for routine maintenance, interpretation, access, visitor use, and community involvement. They do not represent heritage value and cannot compensate for the loss of inherited resources. The six indicators cover external access, internal movement, accommodation, management, catering, and community participation. Figure 4 provides illustrative field evidence of the observable conditions assessed under C10–C15; the photographs were not entered into the model as independent numerical variables.

External transport accessibility (C10) evaluates travel convenience, travel time, road connection and condition, reliability, and observable routine safety when reaching the village from external transport routes. It reflects practical access conditions rather than regional transport development as a whole [31,32].

Accessibility of internal roads and tourist routes (C11) assesses route continuity, walkability, legibility, signage, normal-use safety, and connections among residential areas, public spaces, and principal heritage or landscape nodes [33,34,35].

Accommodation and stay conditions (C12) evaluate the availability, capacity, hygiene, comfort, operational condition, and compatibility of accommodation with village character. The score reflects perceived support for overnight or extended visits rather than the commercial performance of accommodation businesses [36].

Tourism service and management capacity (C13) covers reception, information provision, interpretation, environmental maintenance, routine visitor management, and operational coordination. Questionnaire ratings represent observable service experiences rather than a comprehensive institutional audit of local governance [37].

Catering provision and service conditions (C14) assess facility availability, food variety, local character, hygiene, service efficiency, operational stability, and compatibility of the dining environment with the village setting [38,39].

Cultural tourism operations and community participation (C15) evaluate organized tourism operation, resident participation, local governance, employment and entrepreneurship opportunities, collaboration, coordination, and perceived benefit-sharing. Higher scores indicate more visible and organized local involvement rather than proof of equitable long-term outcomes [5,6].

For C10–C13, references to safety concern only observable and perceived day-to-day conditions of road travel, pedestrian movement, accommodation, and visitor management under normal operating conditions. The framework did not assess hazard probability, exposure, physical vulnerability, emergency response, evacuation, or post-disaster recovery. These indicators should therefore not be interpreted as measures of disaster safety or village resilience.

2.3.4. Indicator Operationalization, Scoring Rules, and Model Entry

Each indicator was operationalized through observable components identified from the literature and verified through field investigation, UAV imagery, and resource inventories. These evidence sources defined what respondents evaluated, confirmed the applicability of the indicators to the three villages, and supported the preparation of standardized assessment materials. They were not assigned independent numerical coefficients and did not directly determine village scores.

All respondents evaluated the same 15 indicators using the ordered evaluation set $\mathrm{V} = \{\mathrm{excellent}, \mathrm{good}, \mathrm{moderate}, \mathrm{poor}, \mathrm{very} \mathrm{poor}\}$, corresponding to the score vector $\mathrm{H} = (5, 4, 3, 2, 1)$. For each village and indicator, the numbers of responses in the five categories were divided by the 60 valid village-specific responses to form a five-element membership vector, whose elements summed to 1. Higher grades consistently represented stronger condition, continuity, accessibility, adequacy, or participation, depending on the indicator, whereas lower grades represented deterioration, fragmentation, insufficiency, weak accessibility, or limited organization. The indicator-specific observable components and grade directions are detailed in Supplementary Table S3.

Expert judgments and questionnaire ratings performed different numerical functions. The aggregated expert pairwise comparisons determined the criterion- and indicator-level AHP weights, whereas the questionnaire-derived membership vectors represented the assessed condition of each village. The local indicator weights were combined with the corresponding membership matrices to obtain the B1, B2, and B3 criterion-layer results, after which the criterion-layer weights were used to derive the target-layer evaluation. The detailed equations and matrix operations are presented in Section 2.4.

This procedure maintained a clear distinction between contextual evidence and numerical model inputs. Field investigation, UAV imagery, and resource inventories supported indicator definition and result interpretation, while only expert-derived weights and questionnaire-derived membership vectors entered the AHP-weighted fuzzy comprehensive evaluation directly. The AHP supporting materials, including the aggregated pairwise-comparison matrices, anonymized individual expert judgments, consistency results, and full-precision weights, are provided in Supplementary Table S4, whereas the complete questionnaire-derived frequencies and membership values for all 15 indicators across the three villages are provided in Supplementary Table S5.

2.4. Resource Evaluation Model for the Three Traditional Villages

Throughout this study, the analytic hierarchy process (AHP) refers exclusively to the derivation of criterion- and indicator-level weights from reciprocal pairwise-comparison matrices, whereas fuzzy comprehensive evaluation refers to the aggregation of questionnaire-derived five-grade membership vectors. The term fuzzy AHP (FAHP) is not used because the expert judgments were expressed using Saaty’s conventional 1–9 reciprocal scale rather than fuzzy-number comparisons. The combined procedure is therefore termed AHP-weighted fuzzy comprehensive evaluation. The analytical model comprised four stages: AHP weight determination, fuzzy membership construction and score aggregation, resource–service structural diagnosis, and robustness analysis.

The methodological design was selected to match three characteristics of the research problem. First, the predefined heritage-oriented hierarchy required a transparent procedure for deriving normative relative priorities among criteria and indicators, for which AHP provides pairwise comparison and consistency testing. Second, the village conditions were assessed through five ordered questionnaire grades rather than precise physical measurements; fuzzy comprehensive evaluation therefore preserves the distribution of judgments across the five grades instead of reducing them immediately to a single value. Third, the research objective was structural diagnosis rather than ranking alone, requiring the B1, B2, and B3 results to remain separately interpretable after aggregation. AHP weighting, fuzzy comprehensive evaluation, and the subsequent resource–service diagnosis therefore perform complementary rather than interchangeable analytical functions.

2.4.1. AHP Weight Determination and Consistency Testing

The analytic hierarchy process (AHP) was used to derive the relative weights of the three criterion layers and the 15 indicators within the predefined hierarchical structure [40]. Each of the nine experts independently completed four reciprocal pairwise-comparison matrices: matrix $\mathrm{A}$, comparing natural and landscape resources (B1), traditional heritage resources (B2), and service-support conditions (B3); matrix ${\mathrm{B}}_1$, comparing C1–C3; matrix ${\mathrm{B}}_2$, comparing C4–C9; and matrix ${\mathrm{B}}_3$, comparing C10–C15. This procedure produced 36 individual judgment matrices.

AHP was selected for the weighting stage because it provides a structured means of decomposing a multidimensional decision problem into a hierarchy, eliciting relative priorities through pairwise comparisons, and examining the internal consistency of expert judgments [40]. These characteristics are particularly relevant to the present study, in which the relative conservation importance of three criterion layers and 15 indicators had to be determined within a predefined hierarchical framework.

Step 1. Construction of Pairwise-Comparison Matrices

For expert $\mathrm{e}$, the reciprocal judgment matrix was expressed as:

$${\mathbf{A}}^{\left(\mathrm{e}\right)}={\left[a_{ij}^{\left(e\right)}\right]}_{n\times n}, {\mathrm{a}}_{ii}^{\left(e\right)}=1, {\mathrm{a}}_{ji}^{\left(e\right)}={\displaystyle \frac{1}{{\mathrm{a}}_{ij}^{\left(e\right)}}}$$

(1)

where ${\mathrm{a}}_{\mathrm{i}\mathrm{j}}^{\left(\mathrm{e}\right)}$ denotes expert $\mathrm{e}\mathrm{’}\mathrm{s}$ judgment of the importance of element $\mathrm{i}$ relative to element $\mathrm{j}$ within the same hierarchical level. Pairwise importance was assessed using Saaty’s 1–9 scale and its reciprocals, as shown in

Table 2. Saaty’s 1–9 pairwise-comparison scale and reciprocal judgments.

Importance Scale	Meaning
1	Elements $\left(i\right)$ and $\left(j\right)$ are equally important.
3	Element $\left(i\right)$ is moderately more important than element $\left(j\right).$
5	Element $\left(i\right)$ is strongly more important than element $\left(j\right).$
7	Element $\left(i\right)$ is very strongly more important than element $\left(j\right).$
9	Element $\left(i\right)$ is extremely more important than element $\left(j\right).$
2, 4, 6, 8	Intermediate values between two adjacent judgments.
Reciprocal	If $a_{ij}=x,$ then $a_{ji}=1/x.$

[40].

Step 2. Aggregation of Expert Judgments

The elicitation consisted of one independent round rather than a Delphi process. All experts were assigned equal importance. The consistency of each individual matrix was examined before aggregation. Corresponding pairwise judgments were then aggregated entry by entry using the geometric mean:

$${\overline{a}}_{ij}={\left({\displaystyle \prod _{\mathrm{e}=1}^9}{a}_{ij}^{\left(e\right)}\right)}^{1/9}$$

(2)

where ${\mathrm{a}}_{\mathrm{i}\mathrm{j}}^{\left(\mathrm{e}\right)}$ is the judgment provided by expert $\mathrm{e}$, and ${\overline{\mathrm{a}}}_{\mathrm{i}\mathrm{j}}$ is the corresponding group judgment. The geometric-mean values formed the aggregated group matrix:

$$\overline{\mathit{A}}={\left[{\overline{a}}_{ij}\right]}_{n\times n}$$

(3)

The same procedure was applied separately to matrices A, B1, B2, and B3. The complete aggregated matrices, individual- and group-level consistency results, and full-precision weights are provided in Supplementary Table S4. Supplementary Table S4h–k additionally report all independent pairwise-comparison judgments provided by the nine anonymized experts. Because the diagonal elements of each reciprocal judgment matrix equal 1 and the remaining reciprocal entries satisfy $a_{ji}=1/a_{ij}$, these values permit reconstruction of all 36 individual judgment matrices and independent reproduction of the geometric-mean aggregation.

Step 3. Derivation of Local and Global Weights

The priority vector of each aggregated judgment matrix was derived using the principal eigenvector method:

$$\overline{\mathbf{A}}\mathbf{v}={\mathrm{\lambda }}_{\mathrm{m}\mathrm{a}\mathrm{x}}\mathbf{v}$$

(4)

where ${\mathrm{\lambda }}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ is the maximum eigenvalue of the aggregated matrix and $\mathrm{v}$ is its corresponding principal right eigenvector. The eigenvector was normalized to obtain the weight vector:

$${\omega }_i={\displaystyle \frac{v_i}{{\displaystyle {\sum }_{j=1}^n}{v}_j}}, {\displaystyle \sum _{i=1}^n}{\omega }_i=1$$

(5)

Matrix $\mathrm{A}$ generated the criterion-layer weight vector ${\mathbf{W}}_{\mathrm{A}}={\left({\mathrm{\omega }}_{\mathrm{B}1},{\mathrm{\omega }}_{\mathrm{B}2},{\mathrm{\omega }}_{\mathrm{B}3}\right)}^{\mathrm{T}}$, whereas matrices ${\mathrm{B}}_1$, ${\mathrm{B}}_2$, and ${\mathrm{B}}_3$ generated the corresponding local indicator-weight vectors. The global weight of indicator ${\mathrm{C}}_{\mathrm{i}}$ was calculated as:

$${\omega }_{C_i}^G={\omega }_{B_k}\hspace{0.17em}{\omega }_{C_i\mid B_k}$$

(6)

where ${\mathrm{\omega }}_{{\mathrm{C}}_{\mathrm{i}}\mid {\mathrm{B}}_{\mathrm{k}}}$ is the local weight of indicator ${\mathrm{C}}_{\mathrm{i}}$ under its parent criterion ${\mathrm{B}}_{\mathrm{k}}$, and ${\mathrm{\omega }}_{{\mathrm{C}}_{\mathrm{i}}}^{\mathrm{G}}$ is its global weight in the target-layer evaluation. Full-precision weights were used in all calculations, while rounded values were used only for presentation.

Step 4. Consistency Testing

The logical consistency of each individual and aggregated judgment matrix was evaluated using the consistency index $\mathrm{C}\mathrm{I}$ and consistency ratio $\mathrm{C}\mathrm{R}$:

$$\mathrm{C}\mathrm{I}={\displaystyle \frac{{\lambda }_{\mathrm{m}\mathrm{a}\mathrm{x}}-n}{n-1}}$$

(7)

$$\mathrm{C}\mathrm{R}={\displaystyle \frac{\mathrm{C}\mathrm{I}}{\mathrm{R}\mathrm{I}}}$$

(8)

where $n$ is the matrix order and $\mathrm{R}\mathrm{I}$ is the average random consistency index corresponding to that order. The $\mathrm{R}\mathrm{I}$ values used in the consistency test are provided in

Table 3. Random consistency index (RI) values by matrix order.

n	1	2	3	4	5	6	7	8	9
RI	0.00	0.00	0.58	0.90	1.12	1.24	1.32	1.41	1.45

. A judgment matrix was regarded as acceptably consistent when $\mathrm{C}\mathrm{R}<0.10$; otherwise, the corresponding pairwise judgments required reassessment.

In this study, matrices $\mathrm{A}$ and ${\mathrm{B}}_1$ were of order 3 and therefore used $\mathrm{R}\mathrm{I}=0.58$, whereas matrices ${\mathrm{B}}_2$ and ${\mathrm{B}}_3$ were of order 6 and used $\mathrm{R}\mathrm{I}=1.24$. The same CR < 0.10 criterion was applied to all 36 individual matrices and the four aggregated group matrices. Individual and group-level consistency results are reported in Supplementary Table S4e,f, while the anonymized pairwise judgments underlying the 36 individual matrices are provided in Supplementary Table S4h–k.

2.4.2. Fuzzy Comprehensive Evaluation and Score Derivation

Fuzzy comprehensive evaluation combined the AHP-derived weights with the five-grade questionnaire distributions. For each village, only questionnaire ratings entered the membership and score calculations; field investigation, UAV imagery, and resource inventories were used to define and verify the observable indicator content and to support result interpretation rather than as additional numerical variables.

Let $\mathrm{i}=1,2,3$ denote Xiaodianhe, Liyu, and Xuebaizhuang, respectively; $\mathrm{k}=1,2,3$ denote the criterion layers; $\mathrm{j}=1,\dots ,{\mathrm{p}}_{\mathrm{k}}$ denote the indicators within criterion layer $\mathrm{k}$; and $\mathrm{g}=1,\dots ,5$ denote the evaluation grades. The factor sets were defined as:

$${\mathrm{B}}_1=\left\{{\mathrm{C}}_1,{\mathrm{C}}_2,{\mathrm{C}}_3\right\},{\mathrm{B}}_2=\left\{{\mathrm{C}}_4,{\mathrm{C}}_5,{\mathrm{C}}_6,{\mathrm{C}}_7,{\mathrm{C}}_8,{\mathrm{C}}_9\right\},{\mathrm{B}}_3=\left\{{\mathrm{C}}_{10},{\mathrm{C}}_{11}, {\mathrm{C}}_{12}, {\mathrm{C}}_{13}, {\mathrm{C}}_{14}, {\mathrm{C}}_{15}\right\}$$

The evaluation set was $\mathrm{V}=\left\{{\mathrm{V}}_1,{\mathrm{V}}_2,{\mathrm{V}}_3,{\mathrm{V}}_4,{\mathrm{V}}_5\right\}$, corresponding to excellent, good, moderate, poor, and very poor, with the score vector:

$$\mathbf{q}={\left(5,4,3,2,1\right)}^{\mathrm{T}}.$$

(9)

For village $\mathrm{i}$, the membership degree of indicator $\mathrm{j}$ in criterion layer $\mathrm{k}$ to evaluation grade $\mathrm{g}$ was calculated from the questionnaire frequency as:

$${\mathbf{r}}_{\mathrm{i}\mathrm{k}\mathrm{j}\mathrm{g}}={\displaystyle \frac{{\mathrm{m}}_{\mathrm{i}\mathrm{k}\mathrm{j}\mathrm{g}}}{n_{\mathrm{i}\mathrm{k}\mathrm{j}}}},n_{\mathrm{i}\mathrm{k}\mathrm{j}}=60.$$

(10)

where ${\mathrm{m}}_{\mathrm{i}\mathrm{k}\mathrm{j}\mathrm{g}}$ is the number of respondents selecting grade $\mathrm{g}$, and $n_{\mathrm{i}\mathrm{k}\mathrm{j}}$ is the number of valid ratings for the corresponding village and indicator. The indicator-level membership vector was expressed as:

$${\mathbf{r}}_{\mathrm{i}\mathrm{k}\mathrm{j}}=\left({\mathrm{r}}_{\mathrm{i}\mathrm{k}\mathrm{j}1},{\mathrm{r}}_{\mathrm{i}\mathrm{k}\mathrm{j}2},{\mathrm{r}}_{\mathrm{i}\mathrm{k}\mathrm{j}3},{\mathrm{r}}_{\mathrm{i}\mathrm{k}\mathrm{j}4},{\mathrm{r}}_{\mathrm{i}\mathrm{k}\mathrm{j}5}\right).$$

The membership vectors of all indicators within criterion layer $\mathrm{k}$ were arranged by row to form the village-specific membership matrix.

$${\mathbf{R}}_{\mathrm{i}\mathrm{k}}={\left[\begin{array}{c}\begin{array}{c}\begin{array}{c}{\mathbf{r}}_{\mathrm{i}\mathrm{k}1}\\ {\mathbf{r}}_{\mathrm{i}\mathrm{k}2}\end{array}\\ ⋮\end{array}\\ {\mathbf{r}}_{\mathrm{i}\mathrm{k}\mathrm{j}\mathrm{g}}\end{array}\right]}_{\mathrm{p}\mathrm{k}\times 5,}$$

(11)

where ${\mathrm{p}}_1=3$ and ${\mathrm{p}}_2={\mathrm{p}}_3=6$.

Let

$${\mathbf{w}}_{\mathrm{k}}={\left({\mathrm{w}}_{\mathrm{k}1},{\mathrm{w}}_{\mathrm{k}2},\dots ,{\mathrm{w}}_{\mathrm{k}{\mathrm{p}}_{\mathrm{k}}}\right)}^{\mathrm{T}}$$

denote the local AHP weight vector of the indicators within criterion layer $\mathrm{k}$. The corresponding criterion-layer fuzzy evaluation vector was calculated as:

$${\mathbf{b}}_{\mathrm{i}\mathrm{k}}={\mathbf{w}}_{\mathrm{k}}^{\mathrm{T}}\times {\mathbf{R}}_{\mathrm{i}\mathrm{k}}$$

(12)

The largest component of ${\mathrm{b}}_{\mathrm{i}\mathrm{k}}$ identifies the dominant evaluation grade of village $\mathrm{i}$ at criterion layer $\mathrm{k}$.

Let

$$\mathbf{W}={\left({\mathrm{W}}_1,{\mathrm{W}}_2,{\mathrm{W}}_3\right)}^{\mathrm{T}}$$

denote the criterion-layer AHP weight vector. The three criterion-layer evaluation vectors were arranged as:

$${\mathbf{C}}_{\mathrm{i}}=\left[\begin{array}{c}\begin{array}{c}{\mathbf{b}}_{\mathrm{i}1}\\ {\mathbf{b}}_{\mathrm{i}2}\end{array}\\ {\mathbf{b}}_{\mathrm{i}3}\end{array}\right].$$

The target-layer fuzzy evaluation vector was then calculated as:

$${\mathbf{a}}_{\mathrm{i}}={\mathbf{W}}^{\mathrm{T}}\times {\mathbf{C}}_{\mathrm{i}}$$

(13)

The largest component of ${\mathrm{a}}_{\mathrm{i}}$ identifies the dominant target-layer evaluation grade.

Indicator, criterion, and target-layer scores were obtained by defuzzifying the corresponding five-grade membership vectors:

$$s_{ikj}={\mathit{r}}_{ikj}\mathit{q}, B_{ik}={\mathit{b}}_{ik}\mathit{q}, T_i={\mathit{a}}_i\mathit{q}$$

(14)

where ${\mathrm{s}}_{\mathrm{i}\mathrm{k}\mathrm{j}}$ is the score of indicator $\mathrm{j}$, ${\mathrm{B}}_{\mathrm{i}\mathrm{k}}$ is the score of criterion layer $k$, and ${\mathrm{T}}_{\mathrm{i}}$ is the target-layer score for village $i$. The maximum-membership grade and the defuzzified score were interpreted jointly: the former identifies the dominant grade, whereas the latter summarizes the complete five-grade distribution.

The complete village-specific membership matrices are provided in Supplementary Table S5, and the complete numerical calculation for the Xiaodianhe ${\mathrm{B}}_1$ criterion is retained in Section 3.2.1. All calculations used full-precision numerical values. Principal criterion- and target-layer scores and AHP weights are reported to four decimal places, whereas robustness-scenario summary scores are reported to three decimal places for compact cross-scenario comparison. Inferential test statistics are reported as specified in the corresponding tables.

2.4.3. Resource–Service Structural Diagnosis

The structural diagnosis used the defuzzified criterion-layer scores $B_{i1}$, $B_{i2}$, and $B_{i3}$ obtained from Equation (14), where $i$ denotes the village. No additional weights, questionnaire data, or membership matrices were introduced at this stage.

The inherited resource base was represented by the arithmetic mean of the natural and landscape resource score and the traditional heritage resource score:

$${RB}_i={\displaystyle \frac{B_{i1}+B_{i2}}{2}}$$

(15)

The equal mean treats ${\mathrm{B}}_1$ and ${\mathrm{B}}_2$ as two co-constitutive dimensions of the inherited resource base. It serves only as a transparent diagnostic reference and does not replace the AHP-weighted target-layer score. Using the original AHP weights at this stage would allow the higher weight of traditional heritage resources to dominate the resource-base reference, whereas Equation (15) addresses the separate question of whether service support is broadly commensurate with both inherited resource dimensions.

The service-support score and relative service gap were defined as:

$$S_i=B_{i3}$$

(16)

$${\Delta }_i=S_i-{RB}_i$$

(17)

A negative ${\Delta }_i$ indicates that service-support conditions lag behind the resource base, whereas a positive value indicates that the service-support score exceeds the resource-base reference. This difference is a descriptive diagnostic measure rather than a statistical test.

To test whether the diagnosis depended on the equal-mean specification, two alternative resource-base calculations were used. First, the AHP weights of ${\mathrm{B}}_1$ and ${\mathrm{B}}_2$ were renormalized within the resource domain:

$$R{B}_i^{\left(w\right)}={\displaystyle \frac{w_1{B}_{i1}+w_2{B}_{i2}}{w_1+w_2}}$$

(18)

$${\Delta }_i^{\left(w\right)}=B_{i3}-R{B}_i^{\left(w\right)}$$

(19)

Second, the better-performing resource dimension was used as a conservative reference:

$${\Delta }_i^{\left(\mathrm{m}\mathrm{a}\mathrm{x}\right)}={\mathrm{B}}_{i3}-\mathrm{m}\mathrm{a}\mathrm{x}\left(B_{i1},B_{i2}\right)$$

(20)

These alternative calculations were used only for sensitivity analysis. They did not replace the principal equal-mean gap, the criterion-layer scores, or the target-layer evaluation.

Criterion-level dispersion was calculated as

$${\mathrm{D}}_i=\mathrm{m}\mathrm{a}\mathrm{x}\left(B_{i1},B_{i2},B_{i3}\right)-\mathrm{m}\mathrm{i}\mathrm{n}\left(B_{i1},B_{i2},B_{i3}\right)$$

(21)

A smaller ${\mathrm{D}}_i$ indicates a relatively balanced criterion structure, whereas a larger value indicates a more pronounced internal imbalance.

Diagnostic profiles were identified through the joint interpretation of ${\Delta }_i$, ${\mathrm{D}}_i$, the fuzzy evaluation grade of service-support conditions, and indicator-level deficiencies, particularly deficiencies involving relatively important indicators. On the five-grade scale, 0.10 points correspond to one tenth of the interval between adjacent grades and were used to distinguish near parity from a practically meaningful difference; 0.50 points correspond to half of one grade interval and were used to identify a substantial gap. These values are transparent operational rules for comparing the three cases rather than statistically validated or universally applicable cut-off values (

Table 4. Operational thresholds and interpretation rules for resource–service diagnostic profiles.

Diagnostic Profile	Relative Service Gap ${\Delta }_{\mathit{i}}$	Criterion-Level Dispersion ${\mathbf{D}}_{\mathit{i}}$	Supporting Interpretation
Relative resource–service coordination	$(-0.10<{\Delta }_{\mathit{i}}<+0.10)$	$({\mathrm{D}}_i<0.10)$	Resource and service scores are approximately balanced, without concentrated service-support deficiencies.
Resource advantage with targeted service deficiencies	$(-0.50<{\Delta }_{\mathit{i}}\le -0.10)$	($0.10\le {\mathrm{D}}_i<0.50$)	Service support falls moderately below the resource base, with deficiencies concentrated in identifiable service indicators.
Resource potential constrained by substantial service gaps	$({\Delta }_{\mathit{i}}\le -0.50)$	$({\mathrm{D}}_i\ge 0.50)$	Service support falls substantially below the resource base, with multiple service deficiencies and a clearly imbalanced criterion structure.

).

The thresholds were not applied mechanically. A diagnostic profile was retained only when the direction of the gap remained consistent under the equal-mean, AHP-weighted, and strongest-resource specifications and when it was supported by the service-support grade and indicator-level evidence. No case in this study exhibited ${\Delta }_{\mathrm{i}}\ge 0.10$; such a result would not automatically indicate an optimal condition, because excessive or incompatible service development could still threaten heritage authenticity and landscape continuity. The resulting profiles are descriptive and do not demonstrate that the proposed interventions will necessarily improve conservation outcomes.

3. Results and Analysis

The results are presented in the order of the three research questions. Section 3.1 reports the criterion- and indicator-level weight structure for RQ1. Section 3.2 and Section 3.3 compare the overall scores, fuzzy-grade distributions, criterion structures, resource–service relationships, and indicator-level strengths and deficiencies for RQ2. Section 3.4 translates the resulting diagnostic profiles into sequenced village-specific conservation priorities for RQ3.

3.1. Weight Structure and Relative Importance of Evaluation Dimensions

All 36 individual judgment matrices completed by the nine experts satisfied the consistency requirement of $\mathrm{C}\mathrm{R}<0.10$ and were therefore retained (

Table 5. Individual and aggregated AHP consistency results for the four judgment matrices.

Matrix	Order ($\mathit{n}$)	Individual $\mathbf{C}\mathbf{R}$ Range	(${\mathit{\lambda }}_{\mathbf{m}\mathbf{a}\mathbf{x}}$)	$\mathbf{C}\mathbf{I}$	$\mathbf{R}\mathbf{I}$	Group $\mathbf{C}\mathbf{R}$	Result
Target layer ($\mathrm{A}$)	3	0.0079–0.0462	3.0012	0.0006	0.58	0.0011	Pass
($B_1$) Natural and landscape resources	3	0.0079–0.0462	3.0066	0.0033	0.58	0.0056	Pass
($B_2$) Traditional heritage resources	6	0.0385–0.0855	6.0321	0.0064	1.24	0.0052	Pass
($B_3$) Service-support conditions	6	0.0391–0.0795	6.0259	0.0052	1.24	0.0042	Pass

Note: Individual CR ranges show the minimum and maximum consistency ratios among the nine experts. Group matrices were obtained through geometric-mean aggregation of corresponding pairwise judgments.

). Corresponding pairwise judgments were aggregated entry by entry using the geometric mean. For example, the aggregated comparison of traditional heritage resources (${\mathrm{B}}_2$) with natural and landscape resources (${\mathrm{B}}_1$) was 1.8761. This value represents the geometric mean of the nine experts’ corresponding Saaty-scale judgments rather than a score assigned by a single expert. The complete aggregated matrices, individual- and group-level consistency results, and full-precision weights are provided in Supplementary Table S4a–g, while Supplementary Table S4h–k report the anonymized expert-level pairwise judgments required to reconstruct all 36 individual matrices.

The maximum individual consistency ratio was 0.0855 for the ${\mathrm{B}}_2$ matrix, and the four group consistency ratios ranged from 0.0011 to 0.0056. Thus, no individual matrix required exclusion or reassessment, and all four aggregated matrices were accepted for weight derivation.

The resulting criterion-layer weights were 0.3030 for natural and landscape resources (${\mathrm{B}}_1$), 0.5490 for traditional heritage resources (${\mathrm{B}}_2$), and 0.1480 for service-support conditions (${\mathrm{B}}_3$). The local indicator weights were multiplied by their corresponding criterion-layer weights to obtain the global weights shown in

Table 6. Local and global AHP weights and rankings of the 15 indicators.

Criterion and Criterion Weight	Indicator	Local Weight	Global Weight	Rank
($B_1$) (0.3030)	C1	0.3689	0.1118	5
	C2	0.4727	0.1432	2
	C3	0.1584	0.0480	8
($B_2$) (0.5490)	C4	0.2432	0.1335	3
	C5	0.2807	0.1541	1
	C6	0.0699	0.0383	10
	C7	0.1107	0.0608	6
	C8	0.2242	0.1230	4
	C9	0.0713	0.0392	9
($B_3$) (0.1480)	C10	0.3900	0.0577	7
	C11	0.2257	0.0334	11
	C12	0.1075	0.0159	12
	C13	0.1053	0.0156	13
	C14	0.0874	0.0129	14
	C15	0.0840	0.0124	15

; the local and full-precision weights are reported in Supplementary Table S4.

Within ${\mathrm{B}}_1$, water bodies and gully landscapes (C2, 0.4727) received the highest local weight, followed by mountain landforms and visual landscape (C1, 0.3689). Within ${\mathrm{B}}_2$, traditional mountain vernacular architecture (C5, 0.2807), the traditional settlement environment (C4, 0.2432), and material cultural remains (C8, 0.2242) were the three most influential indicators. Within ${\mathrm{B}}_3$, external transport accessibility (C10, 0.3900) and internal-road and visitor-route accessibility (C11, 0.2257) together accounted for 0.6157 of the criterion-layer weight, indicating the relative importance assigned to basic access and route continuity.

After multiplication by the corresponding criterion-layer weights, the five highest global weights were assigned to traditional mountain vernacular architecture (C5, 0.1541), water bodies and gully landscapes (C2, 0.1432), the traditional settlement environment (C4, 0.1335), material cultural remains (C8, 0.1230), and mountain landforms and visual landscape (C1, 0.1118). Together, these five indicators accounted for 0.6656 of the total weight. These results answer RQ1 by showing that the expert panel prioritized the physical, spatial, and environmental foundations of mountainous heritage conservation.

The lower weight assigned to ${\mathrm{B}}_3$ should not be interpreted as indicating that service deficiencies are unimportant. Rather, it reflects the conceptual distinction between inherited resources as the basis of heritage value and service-support conditions as enabling capacity. The adequacy of service provision relative to the inherited resource base is therefore examined separately through the resource–service structural diagnosis.

3.2. Overall and Criterion-Level Evaluation Results

This section reports the fuzzy comprehensive evaluation results at the criterion and target layers. Section 3.2.1 presents the criterion-layer evaluation vectors and a complete calculation example for the Xiaodianhe ${\mathrm{B}}_1$ criterion, while Section 3.2.2 compares the target-layer membership distributions and composite scores of the three villages. These results are interpreted as questionnaire-derived, evidence-informed comparisons of observable village conditions rather than direct technical audits of heritage authenticity, ecological performance, management effectiveness, or conservation outcomes.

3.2.1. Criterion-Layer Evaluation Results

The local AHP weights reported in Table 6 were combined with the village-specific membership matrices using Equation (12). The complete questionnaire frequencies and membership values for all 15 indicators in the three villages are provided in Supplementary Table S5. To demonstrate the calculation procedure without reproducing all nine criterion-layer matrices in the main text, the natural and landscape resource criterion (${\mathrm{B}}_1$) of Xiaodianhe is retained as a complete numerical example.

The local weight vector for C1–C3 was

$${\mathbf{W}}_{B_1}^T=(0.3689, 0.4727, 0.1584)$$

Based on the questionnaire-derived membership values, the corresponding matrix was

$${\mathbf{R}}_{\mathrm{B}1}^{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}=\left[\begin{array}{ccccc}0.4167& 0.3500& 0.1667& 0.0500& 0.0167\\ 0.3667& 0.4667& 0.1500& 0.0167& 0.0000\\ 0.2667& 0.4833& 0.1833& 0.0667& 0.0000\end{array}\right].$$

The rows correspond to C1–C3, and the columns correspond to excellent, good, moderate, poor, and very poor. The displayed membership values are rounded to four decimal places; the calculations used the exact questionnaire frequencies divided by the 60 valid Xiaodianhe ratings and the full-precision AHP weights.

The criterion-layer fuzzy evaluation vector was calculated as

$${\mathbf{b}}_{{\mathrm{B}}_1}^{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}={\mathbf{w}}_{{\mathrm{B}}_1}^{\mathrm{T}}{\mathbf{R}}_{{\mathrm{B}}_1}^{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}=\left(0.3693,0.4263, 0.1614, 0.0369, 0.0061\right).$$

For example, the membership degree associated with the good grade (${\mathrm{V}}_2$) was

$${\mathbf{b}}_{{\mathrm{B}}_1,{\mathrm{V}}_2}^{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}=0.3689\times {\displaystyle \frac{21}{60}}+0.4727\times {\displaystyle \frac{28}{60}}+0.1584\times {\displaystyle \frac{29}{60}}=0.4263.$$

Because 0.4263 was the largest component of the evaluation vector, the dominant grade for Xiaodianhe ${\mathrm{B}}_1$ was good. Using the score vector $\mathrm{q}={(5,4,3,2,1)}^{\mathrm{T}}$, its defuzzified score was:

$${\mathrm{B}}_1^{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}={\mathbf{b}}_{{\mathrm{B}}_1}^{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}\mathrm{q}\approx 4.1156$$

The maximum-membership grade and the defuzzified score provide complementary information. The former identifies the dominant evaluation category, whereas the latter incorporates the complete five-grade membership distribution. The same procedure was applied to the remaining criterion layers and villages, producing the results in

Table 7. Criterion-layer fuzzy evaluation vectors, dominant grades, and scores for the three villages.

Village	Criterion Layer	Evaluation Vector (Excellent, Good, Moderate, Poor, Very Poor)	Maximum-Membership Grade	Score
Xiaodianhe	(${\mathrm{B}}_1$) Natural and landscape resources	(0.3693, 0.4263, 0.1614, 0.0369, 0.0061)	Good	4.1156
	(${\mathrm{B}}_2$) Traditional heritage resources	(0.3751, 0.4518, 0.1531, 0.0141, 0.0059)	Good	4.1761
	(${\mathrm{B}}_3$) Service-support conditions	(0.4106, 0.4168, 0.1500, 0.0151, 0.0075)	Good	4.2080
Liyu	(${\mathrm{B}}_1$) Natural and landscape resources	(0.1334, 0.4991, 0.3262, 0.0412, 0.0000)	Good	3.7248
	(${\mathrm{B}}_2$) Traditional heritage resources	(0.1887, 0.4851, 0.2776, 0.0463, 0.0023)	Good	3.8116
	(${\mathrm{B}}_3$) Service-support conditions	(0.1152, 0.4631, 0.3282, 0.0728, 0.0207)	Good	3.5793
Xuebaizhuang	(${\mathrm{B}}_1$) Natural and landscape resources	(0.1325, 0.4106, 0.2351, 0.1569, 0.0649)	Good	3.3889
	(${\mathrm{B}}_2$) Traditional heritage resources	(0.1082, 0.3073, 0.4045, 0.1111, 0.0689)	Moderate	3.2746
	(${\mathrm{B}}_3$) Service-support conditions	(0.0038, 0.1406, 0.2537, 0.3778, 0.2241)	Poor	2.3222

Note: Membership vectors and criterion scores are reported to four decimal places. All calculations used full-precision weights and questionnaire-derived membership values. Complete indicator-level frequencies and membership values are provided in Supplementary Table S5.

.

Xiaodianhe was classified as good across all three criteria, with criterion scores ranging from 4.1156 to 4.2080. Liyu was also classified as good across the three criteria, although its service-support score was lower than its two inherited-resource scores. Xuebaizhuang showed a more uneven structure: natural and landscape resources were classified as good, traditional heritage resources as moderate, and service-support conditions as poor. For Xuebaizhuang ${\mathrm{B}}_3$, the combined membership degree of the poor and very poor grades reached 0.6019. These results identify the principal criterion-level differences relevant to RQ2; their resource–service relationships and indicator-level causes are examined in Section 3.3. Figure 5 presents the indicator-level membership distributions underlying these criterion-layer results.

3.2.2. Target-Layer Evaluation Results

The three criterion-layer evaluation vectors were aggregated using the criterion-layer AHP weights according to Equation (13). Continuing the Xiaodianhe example, the criterion-layer evaluation matrix was

$${\mathrm{C}}^{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}=\left[\begin{array}{ccccc}0.3693& 0.4263& 0.1614& 0.0369& 0.0061\\ 0.3751& 0.4518& 0.1531& 0.0141& 0.0059\\ 0.4106& 0.4168& 0.1500& 0.0151& 0.0075\end{array}\right]$$

where the rows correspond to ${\mathrm{B}}_1$, ${\mathrm{B}}_2$, and ${\mathrm{B}}_3$, and the columns correspond to excellent, good, moderate, poor, and very poor. The criterion-layer weight vector was

$$\mathbf{W}={(0.3030,0.5490,0.1480)}^{\mathrm{T}}$$

The target-layer evaluation vector was therefore

$${\mathbf{a}}^{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}={\mathbf{W}}^{\mathrm{T}}{\mathbf{C}}^{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}=(0.3786,0.4389,0.1551,0.0212,0.0062)$$

Using the score vector $\mathrm{q}={(5,4,3,2,1)}^{\mathrm{T}}$, the target-layer score was

$${\mathrm{T}}_{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}={\mathbf{a}}^{\mathrm{X}\mathrm{i}\mathrm{a}\mathrm{o}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\mathrm{h}\mathrm{e}}\mathrm{q}\approx 4.1625$$

The largest membership degree was 0.4389 for the good grade; Xiaodianhe was therefore classified as good at the target layer. The same procedure was applied to Liyu and Xuebaizhuang, producing the results in

Table 8. Target-layer fuzzy evaluation vectors, dominant grades, scores, and rankings for the three villages.

Village	Target-Layer Evaluation Vector (Excellent, Good, Moderate, Poor, Very Poor)	Maximum-Membership Grade	Score	Rank
Xiaodianhe	(0.3786, 0.4389, 0.1551, 0.0212, 0.0062)	Good	4.1625	1
Liyu	(0.1611, 0.4861, 0.2998, 0.0487, 0.0043)	Good	3.7509	2
Xuebaizhuang	(0.1001, 0.3139, 0.3308, 0.1645, 0.0907)	Moderate	3.1683	3

Note: The target-layer evaluation vectors and scores were calculated using full-precision criterion-layer weights and evaluation vectors. Values are reported to four decimal places.

. Displayed vectors are rounded to four decimal places, whereas all calculations used full-precision criterion-layer weights and membership values.

The target-layer score order was Xiaodianhe, Liyu, and Xuebaizhuang. Xiaodianhe and Liyu were both classified as good, although their membership structures differed: the difference between the highest and second-highest memberships was 0.0603 for Xiaodianhe and 0.1862 for Liyu. For Xuebaizhuang, the moderate and good memberships differed by only 0.0169, indicating that its dominant target-layer grade was less distinct than those of the other two villages.

These results provide the overall comparative baseline for RQ2 but do not independently determine the conservation profiles. Target-layer aggregation can conceal whether a lower score originates from inherited-resource conditions or service-support deficiencies and can allow stronger dimensions to compensate for weaker ones. The criterion-level resource–service gaps, dispersion, and indicator-level deficiencies are therefore examined separately in Section 3.3. The stability of the village ranking across respondent groups and assessment conditions is evaluated in the following robustness analyses.

3.2.3. Respondent-Group and Assessment-Mode Robustness

The respondent-group analyses followed the three mutually exclusive recruitment categories defined in Section 2.2: residents or respondents with long-term familiarity with local conditions (R/F), on-site visitors (V), and professional-background respondents (P) (P). To test whether the cross-village ranking depended on the original quota composition (R/F: 50%; V: 25%; P: 25% within each village), target-layer scores were recalculated using the original quota, equal group weights, and each respondent group separately (

Table 9. Target-layer scores and village rankings under alternative respondent-composition scenarios.

Respondent-Group Composition	Xiaodianhe Score (Rank)	Liyu Score (Rank)	Xuebaizhuang Score (Rank)	Village Order
Original quota: 0.50/0.25/0.25	4.163 (1)	3.751 (2)	3.168 (3)	Xiaodianhe $>$ Liyu $>$ Xuebaizhuang
Equal group weights: 1/3 each	4.169 (1)	3.751 (2)	3.170 (3)	Xiaodianhe $>$ Liyu $>$ Xuebaizhuang
Residents/familiar respondents only	4.143 (1)	3.750 (2)	3.163 (3)	Xiaodianhe $>$ Liyu $>$ Xuebaizhuang
On-site visitors only	4.139 (1)	3.770 (2)	3.229 (3)	Xiaodianhe $>$ Liyu $>$ Xuebaizhuang
professional-background respondents (P) only	4.224 (1)	3.734 (2)	3.118 (3)	Xiaodianhe $>$ Liyu $>$ Xuebaizhuang

Note: The original quota comprised 30 residents or respondents familiar with local conditions, 15 on-site visitors, and 15 professional-background respondents (P) per village. Scenario scores are reported to three decimal places because they are used for robustness comparison rather than the principal evaluation.

). The same full-precision AHP weights were used in all scenarios.

The scores varied moderately across respondent compositions, but the order Xiaodianhe $>$ Liyu $>$ Xuebaizhuang remained unchanged in all five scenarios. This indicates that the principal ranking was not produced by the original 0.50/0.25/0.25 quota alone.

Respondent-level criterion- and target-layer scores were then compared within each village across the same three groups (R/F, V, and P) using Kruskal–Wallis tests.

Table 10. Kruskal–Wallis tests and Dunn–Holm post hoc comparisons of respondent-group differences in criterion- and target-layer scores.

Village	Layer	R/F, Median (IQR)	V, Median (IQR)	P, Median (IQR)	(H)	(p)	Dunn–Holm Result
Xiaodianhe	(${\mathrm{B}}_1$)	4.236 (3.803–4.683)	4.000 (3.921–4.369)	4.000 (3.764–4.264)	2.691	0.260
	(${\mathrm{B}}_2$)	4.029 (3.862–4.336)	4.315 (4.064–4.422)	4.393 (4.317–4.463)	10.725	0.005	P > R/F
	(${\mathrm{B}}_3$)	4.225 (3.952–4.589)	4.087 (3.911–4.482)	4.184 (4.018–4.466)	0.565	0.754
	(A)	4.142 (3.986–4.257)	4.106 (4.030–4.344)	4.228 (4.136–4.322)	2.231	0.328
Liyu	(${\mathrm{B}}_1$)	3.842 (3.486–4.000)	3.842 (3.658–4.078)	3.738 (3.158–4.027)	1.092	0.579
	(${\mathrm{B}}_2$)	3.829 (3.571–4.125)	3.678 (3.538–4.082)	3.798 (3.553–4.047)	0.583	0.747
	(${\mathrm{B}}_3$)	3.500 (3.358–3.638)	3.700 (3.471–3.831)	3.916 (3.453–4.042)	9.177	0.010	P > R/F
	(A)	3.761 (3.575–3.856)	3.692 (3.587–3.862)	3.706 (3.496–3.944)	0.071	0.965
Xuebaizhuang	(${\mathrm{B}}_1$)	3.475 (3.055–3.683)	3.473 (3.055–4.131)	3.211 (2.765–3.657)	1.313	0.519
	(${\mathrm{B}}_2$)	3.232 (2.933–3.602)	3.245 (3.121–3.444)	3.398 (3.128–3.550)	0.922	0.631
	(${\mathrm{B}}_3$)	2.121 (1.806–2.718)	2.631 (2.240–3.137)	2.168 (1.880–2.359)	6.669	0.036	No pairwise difference remained significant
	(A)	3.121 (3.014–3.339)	3.296 (3.085–3.358)	3.177 (3.048–3.263)	1.069	0.586

Note: R/F = residents or respondents with long-term familiarity with local conditions; V = on-site visitors; P = professional-background respondents (P); A = target layer; df = 2 for all Kruskal–Wallis tests. Respondent-group membership follows the mutually exclusive recruitment categories defined in Section 2.2. Respondent-level scores were calculated using the same full-precision AHP weights as the main evaluation.

reports the group medians and interquartile ranges. Significant omnibus tests were followed by Dunn pairwise comparisons with Holm adjustment. Table 10 reports the group medians and interquartile ranges. Significant omnibus tests were followed by Dunn pairwise comparisons with Holm adjustment.

No significant respondent-group difference was found for the target-layer score of Xiaodianhe $(H=2.231,p=0.328)$, $\mathrm{L}\mathrm{i}\mathrm{y}\mathrm{u}(H=0.071,p=0.965)$, or Xuebaizhuang $(H=1.069, p=0.586)$. At the criterion level, professional-background respondents (P) assigned higher traditional heritage-resource scores to Xiaodianhe than the resident/familiar group and higher service-support scores to Liyu than the resident/familiar group. Xuebaizhuang also showed an omnibus difference in ${\mathrm{B}}_3$, but no pairwise comparison remained significant after Holm adjustment. Thus, respondent identity affected several dimension-specific judgments without changing the overall village ranking. These results directly address the reviewer’s concern that residents, visitors, and professional-background respondents (P) may evaluate heritage and service conditions differently.

Assessment-mode sensitivity was examined within the professional-background subgroup because all 21 image-assisted questionnaires were completed by respondents in this category. The image-assisted and on-site subsamples comprised 6 and 9 respondents for Xiaodianhe, 6 and 9 for Liyu, and 9 and 6 for Xuebaizhuang, respectively. None of the C10–C15 indicator-level differences remained significant after Holm adjustment; the smallest adjusted p-value was 0.090 for Liyu C10.

At the criterion level, no significant assessment-mode difference in B3 was found for Xiaodianhe $(U=34.0,p=0.456)$ or Xuebaizhuang $(U=39.0,p=0.181)$. For Liyu, image-assisted respondents reported a higher median ${\mathrm{B}}_3$ score than on-site professional-background respondents (P)—4.042 versus 3.572 $(U=8.0,p=0.026)$. No assessment-mode difference was found in the target-layer score of Xiaodianhe $(U=27.0,p=1.000)$, Liyu $(U=20.0,p=0.456)$, or Xuebaizhuang $(U=35.0,p=0.388)$.

When C10–C15 were recalculated using only on-site ratings, the ${\mathrm{B}}_3$ scores were 4.226 for Xiaodianhe, 3.522 for Liyu, and 2.370 for Xuebaizhuang; the corresponding target-layer scores were 4.165, 3.742, and 3.175. The village order remained unchanged, and exclusion of the image-assisted service ratings did not reverse the broad resource–service interpretation of the three cases. Assessment mode therefore influenced a limited service-support judgment in Liyu but did not determine the overall ranking or the principal cross-village differentiation.

3.2.4. Integrated Robustness Analysis

The robustness analyses examined four potential sources of result dependence: indicator-weight specification, the influence of individual experts, respondent composition and assessment mode, and the specification of the resource-base score.

Table 11. Robustness of village rankings and resource–service diagnoses under alternative analytical scenarios.

Robustness Scenario	Xiaodianhe	Liyu	Xuebaizhuang	Stability Result
Baseline target-layer score	4.163	3.751	3.168	Xiaodianhe $>$ Liyu $>$ Xuebaizhuang
Indicator-weight perturbation: 60 scenarios	4.153–4.172	3.743–3.759	3.156–3.181	Ranking unchanged
Leave-one-expert-out: 9 scenarios	4.160–4.164	3.747–3.755	3.160–3.188	Ranking unchanged
Respondent-composition scenarios	4.139–4.224	3.734–3.770	3.118–3.229	Ranking unchanged; two boundary classifications were sensitive
On-site-only ratings for C10–C15	4.165	3.742	3.175	Ranking and broad diagnostic profiles unchanged
Alternative resource-base specifications	${\Delta }_i=+0.032$ to $+0.062$	${\Delta }_i=-0.232$ to $-0.189$	${\Delta }_i=-1.067$ to $-0.993$	Direction of the resource–service gap unchanged

Note: Baseline and robustness scores are reported to three decimal places for cross-scenario comparison; the principal evaluation results in Table 8 are reported to four decimal places. The indicator-weight perturbation analysis separately changed each of the 15 global indicator weights by −20%, −10%, +10%, and +20%, with the remaining weights proportionally renormalized, yielding 60 scenarios.

summarizes the results. Detailed respondent-group and assessment-mode comparisons are reported in Section 3.2.3.

The village order remained unchanged in all 60 indicator-weight perturbation scenarios. Xiaodianhe remained first, Liyu second, and Xuebaizhuang third. Within the matched perturbation scenarios, the minimum score differences were 0.402 between Xiaodianhe and Liyu and 0.572 between Liyu and Xuebaizhuang, indicating that the ranking was not determined by a near tie.

The leave-one-expert-out analysis also retained the same ordering in all nine scenarios. After each expert was removed and the group matrices were reconstructed, the criterion-layer weights ranged from 0.275 to 0.323 for natural and landscape resources, from 0.527 to 0.576 for traditional heritage resources, and from 0.134 to 0.161 for service-support conditions. Thus, neither the principal ranking nor the priority assigned to traditional heritage resources depended on a single expert. Because the underlying expert-level judgments are provided in Supplementary Table S4h–k, each of the nine leave-one-expert-out aggregation scenarios can be independently reconstructed from the reported numerical inputs.

Respondent composition and assessment mode affected several criterion-specific judgments, as reported in Section 3.2.3, but did not reverse the overall village ranking. Excluding image-assisted ratings for C10–C15 also retained the broad cross-village differentiation. Similarly, the equal-mean, AHP-weighted, and strongest-resource specifications produced the same gap direction: Xiaodianhe remained close to resource–service parity, Liyu retained a limited service-support lag, and Xuebaizhuang retained a substantial service-support gap.

Two boundary sensitivities should nevertheless be acknowledged. Under the resident/familiar-respondent-only scenario, Xiaodianhe’s maximum-membership grade shifted from good to excellent. Under the professional-background-only scenario, Liyu’s resource–service relationship shifted from a limited service-support lag to near parity. These changes did not alter the village ranking or the broad contrast among the three cases, but they indicate that maximum-membership grades and profile assignments close to the operational thresholds should not be treated as completely invariant.

Overall, the comparative ordering and the principal cross-village resource–service differentiation were robust to the tested analytical choices. This robustness concerns the internal stability of the three-case comparison; it does not establish external validity for other regions or demonstrate that the proposed conservation interventions will necessarily produce improved outcomes.

3.3. Cross-Village Resource–Service Differences and Diagnostic Profiles

This section further addresses RQ2 by examining the internal resource–service structure of each village rather than relying on the target-layer ranking alone. Section 3.3.1 compares the inherited resource base, service-support conditions, relative service gap, and criterion-level dispersion. Section 3.3.2 and Section 3.3.3 then identify the high-weight indicators and village-specific strengths and deficiencies underlying these structural differences. The resulting profiles represent descriptive diagnoses of the three cases rather than causal assessments of conservation effectiveness.

3.3.1. Resource–Service Gap and Criterion-Level Balance

The diagnostic profiles were identified using the operational rules defined in Section 2.4.3. The relative service gap ${\Delta }_i$ provided the primary numerical distinction, while criterion-level dispersion ${\mathrm{D}}_i$, the fuzzy grade of service-support conditions, alternative resource-base specifications, and indicator-level evidence were used as supporting information.

Table 12. Criterion-layer scores, resource–service gaps, dispersion, and diagnostic profiles for the three villages.

Village	${\mathbf{B}}_1$	${\mathbf{B}}_2$	$\mathbf{R}{\mathbf{B}}_{\mathbf{i}}$	${\mathbf{S}}_{\mathit{i}}\left({\mathbf{B}}_3\right)$	${\Delta }_{\mathit{i}}$	${\mathbf{D}}_{\mathit{i}}$	Diagnostic Profile
Xiaodianhe	4.1156	4.1761	4.1459	4.2080	+0.0621	0.0923	Relative resource–service coordination
Liyu	3.7248	3.8116	3.7682	3.5793	−0.1889	0.2323	Resource advantage with targeted service deficiencies
Xuebaizhuang	3.3889	3.2746	3.3318	2.3222	−1.0096	1.0667	Resource potential constrained by substantial service gaps

Note: ${\mathrm{B}}_1$ denotes natural and landscape resources; ${\mathrm{B}}_2$ denotes traditional heritage resources; $R{B}_i=\left(B_{i1}+B_{i2}\right)/2$; $S_i=B_{i3}$; ${\Delta }_i=S_i-R{B}_i$; and ${\mathrm{D}}_{\mathrm{i}}=\mathrm{m}\mathrm{a}\mathrm{x}\left({\mathrm{B}}_{i1},{\mathrm{B}}_{i2},{\mathrm{B}}_{i3}\right)-\mathrm{m}\mathrm{i}\mathrm{n}\left({\mathrm{B}}_{i1},{\mathrm{B}}_{i2},{\mathrm{B}}_{i3}\right)$. All calculations used full-precision criterion-layer results, whereas displayed values are reported to four decimal places.

and Figure 6 summarize the resulting resource–service structures.

Xiaodianhe had a resource-base score of 4.1459 and a service-support score of 4.2080, producing a relative service gap of $+$0.0621 and criterion-level dispersion of 0.0923. Both values were within the operational 0.10-point range for near parity. Xiaodianhe was therefore interpreted as relatively resource–service coordinated rather than as having a meaningful service surplus. The positive gap should not be understood as evidence that additional facilities are necessarily required or desirable, because further construction could still affect heritage authenticity, settlement character, and landscape continuity.

Liyu recorded a resource-base score of 3.7682 and a service-support score of 3.5793. Its relative service gap of $-$0.1889 and criterion-level dispersion of 0.2323 fell within the operational ranges for a limited service-support lag. Because its service-support criterion remained within the good grade, this result indicates targeted deficiencies rather than a general absence of service capacity. Liyu was therefore diagnosed as having a resource advantage accompanied by specific service-support deficiencies.

Xuebaizhuang showed the most pronounced internal imbalance. Its resource-base score was 3.3318, whereas its service-support score was 2.3222, producing a relative service gap of $-$1.0096 and criterion-level dispersion of 1.0667. Both values exceeded the operational 0.50-point threshold for a substantial imbalance. In addition, its service-support criterion was classified as poor, and the combined membership of the poor and very poor grades reached 0.6019. Xuebaizhuang was therefore diagnosed as having resource potential constrained by substantial service gaps.

The direction of the three resource–service relationships remained unchanged under the equal-mean, AHP-weighted, and strongest-resource specifications examined in Section 3.2.4. Nevertheless, the three profiles should not be interpreted as universally applicable categories. They are case-specific, descriptive diagnoses based on operational thresholds and do not demonstrate that the subsequent conservation recommendations will necessarily produce improved outcomes. The indicator-level evidence supporting the profiles is examined in Section 3.3.2 and Section 3.3.3.

3.3.2. Inter-Village Differences in High-Weight Indicators

To identify the indicator-level differences with the greatest potential influence on the weighted evaluation, the five indicators with the highest global weights were selected for cross-village comparison: traditional mountain vernacular architecture (C5), water bodies and gully landscapes (C2), the traditional settlement environment (C4), material cultural remains (C8), and mountain landforms and visual landscape (C1). Their cumulative global weight was 0.6656. Figure 7 and

Table 13. Questionnaire-derived scores and inter-village differences for the five highest-weight indicators.

Indicator	Global Weight	Xiaodianhe	Liyu	Xuebaizhuang	Maximum Difference
C5 Traditional mountain vernacular architecture	0.1541	4.4333	3.9000	3.4500	0.9833
C2 Water bodies and gully landscapes	0.1432	4.1833	3.6167	3.2000	0.9833
C4 Traditional settlement environment	0.1335	4.1167	4.0167	3.2000	0.9167
C8 Material cultural remains	0.1230	4.0833	3.7667	3.0167	1.0667
C1 Mountain landforms and visual landscape	0.1118	4.1000	3.8167	3.5833	0.5167

compare the questionnaire-derived scores of these indicators across the three villages.

Xiaodianhe received the highest questionnaire-derived score for each of the five indicators. Its strongest result was traditional mountain vernacular architecture (C5, 4.4333), followed by water bodies and gully landscapes (C2, 4.1833), the traditional settlement environment (C4, 4.1167), mountain landforms and visual landscape (C1, 4.1000), and material cultural remains (C8, 4.0833). These results indicate that respondents evaluated its architectural, settlement, and mountain–water conditions comparatively favorably, but they do not by themselves demonstrate a causal synergy among these dimensions.

Liyu was closest to Xiaodianhe in the traditional settlement environment: the difference in C4 was only 0.1000. Its score for traditional mountain vernacular architecture was also comparatively high at 3.9000. Larger differences occurred in water bodies and gully landscapes and material cultural remains, suggesting that respondents perceived its settlement and architectural character more favorably than the visibility, condition, maintenance, or presentation of some landscape and material-heritage elements.

Xuebaizhuang’s highest score among the five indicators was mountain landforms and visual landscape (C1, 3.5833), whereas material cultural remains received the lowest score (C8, 3.0167). This pattern indicates that its mountainous setting was evaluated more favorably than the observable condition, legibility, maintenance, and presentation of its material remains. It should not be interpreted as evidence that its statutory heritage value is low.

Across the five indicators, material cultural remains showed the largest inter-village difference at 1.0667, while mountain landforms and visual landscape showed the smallest difference at 0.5167. The common mountainous setting therefore produced comparatively limited variation in C1, whereas more pronounced differences occurred in the observable retention and presentation of buildings, settlement environments, water–gully landscapes, and material remains.

These high-weight comparisons clarify the indicator-level basis of the overall cross-village differences, but they do not constitute a complete diagnosis. Indicators with lower global weights may still represent urgent local deficiencies, particularly when they concern access, routine maintenance, management, or community participation. Section 3.3.3 therefore considers all 15 indicators when identifying village-specific strengths and priority deficiencies.

3.3.3. Relative Strengths and Priority Deficiencies

Figure 8 identifies the three highest- and three lowest-scoring indicators within each village based on questionnaire-derived mean scores. These within-village rankings were used to distinguish relative strengths from lower-scoring fields and to explain the indicator-level basis of the diagnostic profiles. They did not replace the AHP weights or independently determine conservation priorities. A lower position within one village should therefore not automatically be interpreted as an absolute deficiency, particularly when the corresponding score remained comparatively high.

Xiaodianhe’s three highest-scoring indicators were traditional mountain vernacular architecture (C5, 4.4333), external transport accessibility (C10, 4.4167), and water bodies and gully landscapes (C2, 4.1833). Its three lowest scores occurred for local products and culinary resources (C9, 3.7333), vegetation and ecological landscape (C3, 3.9500), and accommodation and stay conditions (C12, 3.9667). Because these lower-scoring indicators remained close to or above the good-grade range and the village’s service-support criterion was also classified as good, they represent relative fields for improvement rather than substantial deficiencies. This pattern is consistent with the diagnosis of relative resource–service coordination.

Liyu’s relative strengths were concentrated in the traditional settlement environment (C4, 4.0167), traditional mountain vernacular architecture (C5, 3.9000), and vegetation and ecological landscape (C3, 3.8333). Its lowest-scoring indicators were folk activities and local cultural memory (C7, 3.3500), accessibility of internal roads and tourist routes (C11, 3.4000), and accommodation and stay conditions (C12, 3.5833). These results indicate that its comparatively coherent physical resource base is accompanied by more specific weaknesses in cultural continuity, route organization, and support for extended stays, providing indicator-level evidence for the profile of resource advantage with targeted service deficiencies.

Xuebaizhuang’s three highest-scoring indicators were mountain agricultural landscape and production culture (C6, 4.0167), local products and culinary resources (C9, 3.8667), and mountain landforms and visual landscape (C1, 3.5833). By contrast, external transport accessibility (C10, 2.1500), tourism service and management capacity (C13, 2.2000), and cultural tourism operations and community participation (C15, 2.2667) received the lowest scores. The concentration of low scores in accessibility, management, and community-based operation shows that the village’s overall limitation does not result from an absence of identifiable resources, but from inadequate service support for their maintenance, interpretation, and appropriate use.

Together, these indicator-level patterns further answer RQ2 by showing that the same target-layer ranking conceals different structural conditions. Xiaodianhe primarily contains relative quality-improvement fields, Liyu contains targeted cultural and service-support weaknesses, and Xuebaizhuang combines identifiable resource strengths with multiple substantial service deficiencies. These findings are descriptive and do not demonstrate that any particular intervention will necessarily improve conservation outcomes. Section 3.4 translates them into provisional and sequenced conservation priorities.

3.4. Sequenced Conservation Priorities Derived from the Diagnostic Profiles

This section addresses RQ3 by translating the three diagnostic profiles into sequenced conservation priorities (

Table 14. Evaluation evidence, sequenced conservation priorities, and planning propositions for the three villages.

Village and Diagnostic Profile	Measured Basis	Sequenced Evidence-Based Priorities	Planning Propositions Requiring Further Assessment
Xiaodianhe: relative resource–service coordination	$({\Delta }_{\mathit{i}}=+0.0621)$; (${\mathrm{D}}_i$ = 0.0923); all three criteria were classified as good. C5, C10, and C2 were the three highest-scoring indicators, whereas C9, C3, and C12 were relatively lower but remained close to or above the good-grade range.	(1) Maintain traditional mountain vernacular architecture, water–gully landscapes, and settlement–landscape continuity; (2) prioritize the quality and heritage compatibility of existing service support rather than general facility expansion; and (3) improve local-product provision, vegetation management, and accommodation conditions only where compatible with heritage conservation.	The scale, form, materials, and location of new or upgraded facilities require site-specific design studies and heritage-impact assessment. The present evaluation did not test particular architectural-control standards, accommodation models, or facility locations.
Liyu: resource advantage with targeted service deficiencies	$({\Delta }_{\mathit{i}}=-0.1889)$; (${\mathrm{D}}_i$ = 0.2323). C4, C5, and C3 were relative strengths, whereas C7, C11, and C12 received the three lowest scores.	(1) Retain the traditional settlement, vernacular architecture, and vegetation landscape; (2) strengthen cultural interpretation and the continuity and legibility of internal routes; and (3) improve visitor-stay conditions without weakening the existing settlement character.	The alignment of routes linking valleys, streets, courtyards, and agricultural spaces, together with the form and content of interpretation and accommodation facilities, requires detailed spatial analysis, visitor-behavior observation, feasibility assessment, and conservation planning.
Xuebaizhuang: resource potential constrained by substantial service gaps	$({\Delta }_{\mathit{i}}=-1.0096)$; (${\mathrm{D}}_i$ = 1.0667); B3 was classified as poor, with a combined poor and very poor membership of 0.6019. C6, C9, and C1 were relative strengths, whereas C10, C13, and C15 were the principal deficiencies.	(1) Prioritize routine accessibility, maintenance and service management, and organized community participation before broader resource-use initiatives; (2) retain mountain landscapes, agricultural production culture, and local products as the principal resource fields; and (3) consider only gradual, low-impact use after basic conservation-support capacity has improved.	Specific wayfinding systems, tourism products, operating models, development intensity, and benefit-sharing arrangements require feasibility studies and community consultation. Because hazard probability, evacuation, emergency response, and disaster vulnerability were not evaluated, access interventions also require a separate risk assessment.

Note: The measured basis combines criterion-layer results, resource–service diagnostic measures, questionnaire-derived indicator scores, and field-verified observable conditions. The recommendations identify priority domains rather than validated projects. The present analysis does not demonstrate that the proposed measures will necessarily improve conservation outcomes.

). The sequence was determined through the joint interpretation of the resource–service gap, criterion-level dispersion, service-support grade, and indicator-level strengths and deficiencies. Evidence-based priorities refer to conservation or service domains directly supported by the measured results, whereas general planning propositions refer to possible spatial, design, or operational responses that were not independently tested. The latter require additional physical investigation, heritage-impact assessment, feasibility analysis, community consultation, and post-implementation monitoring. Service-support improvement is treated as an enabling condition and not as a substitute for heritage authenticity, integrity, landscape continuity, or community continuity.

These results answer RQ3 by showing that differentiated conservation concerns the sequence and role of intervention rather than the allocation of different projects according to composite rank alone. Xiaodianhe requires quality control and compatibility management within a relatively coordinated structure; Liyu requires targeted completion of cultural interpretation, internal routes, and stay-support conditions; and Xuebaizhuang requires the prior establishment of basic conservation-support capacity before broader activation. These priorities remain provisional and should be refined through site-specific investigation, participatory decision-making, and subsequent monitoring.

4. Discussion

This section interprets the empirical findings in relation to the three research questions and situates them within the broader context of heritage-oriented conservation. It first discusses the relative importance of the evaluation dimensions and the corresponding heritage-oriented decision logic, then examines cross-village resource–service differences and the implications of profile-based sequencing for differentiated conservation. Finally, the limitations and transferability of the proposed framework are considered.

4.1. Relative Importance and the Heritage-Oriented Decision Logic

RQ1 concerned the relative importance and decision roles of the three evaluation dimensions. Traditional heritage resources (B2) received the highest criterion-layer weight (0.5490), followed by natural and landscape resources (B1, 0.3030) and service-support conditions (B3, 0.1480). Within the three criteria, water bodies and gully landscapes (C2) received the highest local weight under B1, traditional mountain vernacular architecture (C5) under B2, and external transport accessibility (C10) under B3. C10 and internal-road and visitor-route accessibility (C11) together accounted for 0.6157 of the B3 weight. At the global level, C5, C2, the traditional settlement environment (C4), material cultural remains (C8), and mountain landforms and visual landscape (C1) were the five highest-weight indicators, jointly accounting for 0.6656 of the total weight. This weight structure places inherited architectural, settlement, and mountain–water characteristics at the centre of conservation decisions.

This result is consistent with studies emphasizing locality preservation and the relationships among landscape, daily life, settlement form, material fabric, authenticity, and integrity [21,22,23]. Research on cultural-tourism spatial quality and public-space experience has also shown that accessibility and environmental conditions affect how village heritage is perceived and used [7,8]. The present framework connects these perspectives by assigning them different decision roles: inherited resources constitute the primary conservation basis, whereas service-support conditions enable maintenance, interpretation, ordinary access, and appropriate use. The lower ${\mathrm{B}}_3$ weight therefore does not imply that service deficiencies are unimportant, but it prevents stronger service performance from compensating for the deterioration or loss of core traditional heritage resources.

The analytical method reflects this heritage-oriented logic. Entropy weighting would primarily respond to statistical variation and might assign greater importance to indicators that differ most among the three cases, even when such variation does not represent conservation significance. TOPSIS could provide an efficient overall ranking, but its closeness coefficient would not readily distinguish whether a lower position resulted from weak inherited resources or inadequate service support. AHP was therefore appropriate for deriving the expert-based priority structure because it uses hierarchical decomposition, pairwise comparisons, and consistency testing to establish relative weights [40]. Fuzzy comprehensive evaluation retained the five-grade questionnaire distributions, while preserving the B1, B2, and B3 criterion-level results for the subsequent resource–service diagnosis.

The framework nevertheless remains dependent on the predefined hierarchy, expert judgments, questionnaire design, and five-grade scale. It should therefore be interpreted as a transparent comparative decision-support approach for the three cases rather than as an objective or universally optimal measurement of traditional-village heritage value.

4.2. Cross-Village Resource–Service Differences

RQ2 examined whether the target-layer ranking adequately represented the differences among the three villages. The structural results show that it did not. Xiaodianhe was close to resource–service parity, with a relative service gap of $+$0.0621; Liyu retained a moderate service-support lag of $-$0.1889; and Xuebaizhuang showed a substantial gap of $-$1.0096. These differences indicate three distinct conditions: relative resource–service coordination, resource advantage with targeted service deficiencies, and resource potential constrained by substantial service gaps. In particular, Xuebaizhuang’s lower overall score should not be interpreted as evidence of uniformly low heritage value, because its principal limitations were concentrated in accessibility, management, and community-based operation rather than in all inherited-resource dimensions.

This differentiation is consistent with studies showing that traditional villages within similar regional settings may nevertheless require different conservation responses. Jiang et al. [12] derived classified protection strategies from socio-ecological spatial differences, while Zhang et al. [13] demonstrated the influence of accessibility and spatial distribution on village differentiation. Research on cultural-tourism spatial quality and public spaces has likewise shown that heritage perception depends on spatial organization and environmental experience as well as resource presence [7,8]. The present findings provide a more specific village-level explanation: Liyu’s lower internal-route, cultural-memory, and accommodation scores correspond to studies emphasizing mobility, spatial legibility, connectivity, and stay quality [32,33,34,35], whereas Xuebaizhuang’s service gap reinforces the importance of external access [31], routine management [10], and organized resident participation [5,6]. Unlike approaches based mainly on total scores or regional spatial types, the resource–service diagnosis distinguishes whether an unfavorable result arises from inherited-resource conditions, service-support limitations, or an imbalance between them.

Stakeholder comparisons further qualify these findings. Residents or respondents familiar with local conditions, visitors, and professional-background respondents differed in several criterion-layer assessments. However, no significant respondent-group difference occurred in the target-layer score of any village, and the order Xiaodianhe $>$ Liyu $>$ Xuebaizhuang remained unchanged across the five respondent-composition scenarios. This partly supports Shi et al. [4], who identified differences between expert and resident perspectives in living-heritage assessment. In the present study, respondent identity influenced the emphasis placed on particular heritage or service dimensions but did not change the overall comparative conclusion. Assessment mode also produced a limited difference in Liyu’s service-support criterion, yet excluding image-assisted service ratings did not reverse the target-layer ranking or broad cross-village differentiation.

The broad structural contrast was also retained under the equal-mean, AHP-weighted, and strongest-resource specifications, as well as under indicator-weight perturbation, leave-one-expert-out analysis, and on-site-only service ratings. However, exact grade or profile assignments close to the operational thresholds were not completely invariant: Xiaodianhe’s dominant grade and Liyu’s near-parity interpretation changed under individual respondent-group scenarios. The robustness results therefore support the stability of the overall ranking and the principal resource–service contrast, rather than proving that every boundary classification is fixed. Moreover, the gaps and thresholds remain descriptive, case-specific diagnostic measures and should not be interpreted as universally validated classifications or causal evidence of conservation outcomes.

4.3. Profile-Based Sequencing in Conservation Decision-Making

RQ3 concerned how the structural diagnoses could inform differentiated conservation decisions. The results indicate that intervention should be sequenced according to the source and severity of the resource–service imbalance rather than applied through a uniform improvement model. For Xiaodianhe, relative resource–service coordination supports an emphasis on heritage-compatible quality control and the avoidance of unnecessary facility expansion. For Liyu, the limited service lag supports targeted improvement in cultural interpretation, internal-route continuity, and stay-support conditions while retaining its settlement and landscape character. For Xuebaizhuang, the substantial service gap indicates that basic accessibility, routine maintenance and management, and organized community participation should precede broader resource activation.

This sequence is consistent with the broader shift from object-centred preservation toward integrated and locality-sensitive conservation. The Historic Urban Landscape approach and locality-preservation studies emphasize the maintenance of inherited spatial relationships, everyday functions, local identity, and landscape continuity during adaptation. The priorities identified for Liyu are also consistent with research on intravillage mobility and accommodation quality [36], while those for Xuebaizhuang correspond to studies emphasizing external accessibility [31], resident participation, and institutional coordination [5,6]. The present study extends these findings by placing them within a profile-dependent sequence: quality control follows relative coordination, targeted service completion responds to a limited service lag, and basic conservation-support capacity precedes resource activation where the service deficit is substantial.

The framework therefore identifies priority intervention domains and their relative order, but it does not determine specific engineering measures, route alignments, facility locations, architectural forms, investment scales, operating models, or development intensity. Such decisions require additional physical surveys, heritage-impact assessment, feasibility and cost analysis, consultation with communities and management authorities, and post-implementation monitoring. The proposed sequence should consequently be interpreted as a decision-support result rather than evidence that any particular intervention will necessarily improve conservation outcomes.

4.4. Limitations, Transferability, and Future Research

Several limitations should be acknowledged. First, the comparative design included only three traditional villages within the same township. This strengthened contextual comparability but limited external generalization. The analysis was also based on cross-sectional judgments from nine experts and 180 quota-sampled respondents, with only 15 visitors and 15 professional-background respondents (P) per village. The subgroup analyses should therefore be interpreted as exploratory, and the three profiles represent analytically transferable diagnostic patterns rather than statistically validated or universally applicable village types.

Second, 21 professional-background respondents (P) completed image-assisted assessments using standardized UAV images, ground-level photographs, and village profiles. These materials could not fully reproduce direct experience of route continuity, accommodation, catering, environmental maintenance, management interaction, or community participation. Excluding the image-assisted ratings did not change the village ranking or the broad cross-village differentiation, but an assessment-mode difference was identified for Liyu’s service-support criterion. Future applications should therefore prioritize on-site evaluation for experience-dependent indicators or supplement image materials with virtual-route records, service-use videos, interviews, and structured operational data.

Third, the resource-base score, relative service gap, criterion-level dispersion, and diagnostic thresholds are descriptive and case-specific. They identify structural differences but do not establish causal relationships among inherited resources, service support, and conservation outcomes. The framework also does not directly assess detailed building condition, authenticity, integrity, demographic continuity, governance capacity, or disaster vulnerability.

No systematic data were collected on multi-hazard exposure, evacuation, emergency access, or community response capacity. The results therefore apply only to observable resource conditions and routine service support under ordinary operating conditions and should not be interpreted as evidence of village safety or resilience. Moreover, the proposed priorities were not implemented or longitudinally evaluated; their effects on heritage condition, landscape continuity, community well-being, visitor experience, and management performance remain unverified.

Future research should test the framework across larger and more diverse samples using multi-period field surveys, building-condition assessments, resident interviews, visitor-behavior observations, demographic and governance records, and village-operation data. In the present study, field investigation, UAV imagery, and resource inventories supported indicator operationalization and result verification but did not enter the model as independent numerical variables. Subsequent studies could integrate directly measured indicators, including GIS-based accessibility, vegetation coverage, building-condition ratios, facility density, population continuity, and a separate multi-hazard risk module. Longitudinal post-implementation monitoring is also required to determine whether the diagnostic rules remain stable and whether profile-based priorities produce measurable conservation benefits.

5. Conclusions

The central finding of this study is that the three villages should not be understood simply as higher- or lower-scoring examples of the same conservation condition. Instead, the statistical analysis revealed three distinct resource–service structures: Xiaodianhe combines a strong inherited resource base with broadly commensurate service support; Liyu retains substantial natural and heritage resources but is constrained by targeted deficiencies in interpretation, internal-route continuity, and stay-support conditions; and Xuebaizhuang retains resource potential while facing a pronounced gap in accessibility, management, and community participation. This distinction is important because the overall scores of 4.1625, 3.7508, and 3.1683 do not constitute a simple ranking of heritage value. Traditional heritage resources received the highest criterion-layer weight (0.5490), but the criterion-level results show that conservation priorities depend on how inherited resources relate to the conditions required for their maintenance, interpretation, accessibility, and appropriate use. The broad differentiation among the three villages remained stable across the tested weight, expert, respondent-composition, assessment-mode, and alternative resource-base scenarios, supporting the robustness of this structural interpretation.

These structural differences imply different sequences of conservation action rather than a uniform development model. Xiaodianhe should emphasize the conservation of traditional architecture, settlement character, and mountain–water landscapes while improving the quality and heritage compatibility of existing services rather than pursuing general facility expansion. Liyu should retain its settlement, architectural, agricultural, and landscape resources while giving earlier attention to cultural interpretation, internal-route continuity, and stay-support conditions. Xuebaizhuang should prioritize basic accessibility, routine environmental maintenance and management, wayfinding, and organized community participation before broader resource-use or tourism-development initiatives. These priorities represent evidence-informed planning directions rather than validated conservation outcomes.

The principal contribution of this study lies not in combining AHP and fuzzy comprehensive evaluation as established methods, but in retaining the distinction between inherited resource conditions and service-support capacity after aggregation. This shifts traditional-village evaluation from composite ranking toward structural diagnosis and sequenced conservation decision-making, while avoiding the assumption that stronger service provision can compensate for losses in authenticity, integrity, or landscape continuity. Taken together, the findings indicate that differentiated conservation of mountainous traditional villages should be guided by the internal relationship between inherited resources and service-support conditions rather than by composite scores alone.