Multidimensional Spatial–Cultural Clustering of Traditional Villages in Northwestern Yunnan Based on a Four-Dimensional Analytical Framework for Sustainable Conservation

Abstract

Traditional villages in ecologically fragile and multi-ethnic frontier regions are increasingly threatened by rapid urbanization and socio-economic transformation. Northwestern Yunnan, located in the longitudinal valleys of the Hengduan Mountains, represents a key cultural landscape of plateau agropastoral civilization and ethnic interaction, yet its spatial organization and clustering mechanisms remain insufficiently understood. This study develops a four-dimensional analytical framework integrating four dimensions—spatial morphology (village distribution patterns and density), geomorphological conditions (elevation, slope, and terrain features), cultural attributes (ethnic composition and historical-cultural corridors), and architectural typologies (dominant residential structure types) to examine 246 officially recognized traditional villages. Using GIS-based spatial statistics, kernel density estimation (KDE), spatial autocorrelation, and a hierarchical overlay model, the study identifies the spatial structure (distribution patterns and density gradients), environmental adaptability (relationships with elevation, slope, and hydrological conditions), and multidimensional clustering characteristics (integrated clustering intensity across four analytical dimensions) of settlements. The results reveal a highly uneven and a statistically significant clustered spatial pattern ($R$ = 0.606, Moran’s $I$ = 0.251, p < 0.05) characterized by a “two corridors–six clusters–multiple nodes” structure. Settlement distribution demonstrates strong coupling with mid-elevation plateau basins, river valley systems, and trade-cultural corridors shaped by the Ancient Tea Horse Road. Multidimensional integration further classifies villages into three typologies—comprehensive, specialized, and general clusters—reflecting different levels of coordination among spatial, environmental, cultural, and architectural dimensions. These findings reveal the spatial regularities and multidimensional clustering characteristics of officially recognized traditional villages in Northwestern Yunnan, and suggest that environmental setting, historical corridors, and cultural-architectural features jointly shape the current recognized heritage landscape. The proposed framework provides a context-sensitive basis for differentiated heritage conservation and rural management in mountainous multi-ethnic regions.

1. Introduction

1.1. Background

Under the backdrop of globalization and rapid urbanization, traditional villages are facing multiple impacts and challenges, particularly in ecologically sensitive and multi-ethnic regions [1]. As significant carriers of human civilization, traditional villages hold rich historical and cultural values as well as unique regional wisdom [2]. Their spatial distribution patterns are shaped through complex interactions with the natural environment, socio-economic conditions, and cultural heritage [3,4,5].

Northwestern Yunnan is one of China’s core regions for traditional villages. Located in the longitudinal valleys of the Hengduan Mountains, the area features large elevation differences and a dense river network. Villages are often distributed along rivers and mountains, forming an important ecological barrier and a crossroads of diverse ethnic cultures in southwestern China [6]. Historically, this region was also a vital commercial area along the Ancient Tea Horse Road [7]. Such unique geographical and cultural contexts have given rise to distinctive architectural forms like “Three Houses and One Wall” and “Four Courtyards and Five Wells” [8,9], making Northwestern Yunnan a living heritage of plateau agropastoral civilization and Sino-Tibetan cultural exchange [10]. The spatial distribution characteristics of its traditional villages effectively reflect the long-term human-environment adaptation; therefore, its distinctive geographical and cultural features require deeper exploration.

Despite its importance, systematic research on traditional villages in Northwestern Yunnan remains insufficient. Existing studies are mostly concentrated on eastern, southern coastal, and central regions, often at the provincial or meso-regional scale [11,12]. Less attention has been paid to border regions with complex geographical environments and diverse ethnic cultures. In particular, research on multi-dimensional clustering and conservation strategies at smaller spatial scales is lacking. Most studies focus on individual village cases, such as Shuhe in Lijiang or Nuodeng in Dali, while regional-scale clustered studies remain limited [13]. Scholars such as Zheng et al. [14] have emphasized the need to analyze traditional village distribution in Northwestern Yunnan with respect to environmental factors like high rainfall and low transportation accessibility. Meanwhile, Yuan et al. [15] identified Northwestern Yunnan as a high-density traditional village zone but did not explore the influence of multi-ethnic interactions on shaping the spatial “genes” of villages.

To address these issues, this study proposes a four-dimensional analytical framework—including spatial morphology, geomorphological features, cultural attributes, and architectural typology—and applies a hierarchical overlay model to systematically investigate the spatial distribution and clustering types of traditional villages in Northwestern Yunnan [16].

1.2. Literature Review

Traditional villages refer to rural settlements that have preserved their historical appearance, characteristic traditional architecture, and unique cultural heritage [17]. According to the Evaluation and Identification System of Traditional Villages jointly released in 2012 by the Ministry of Housing and Urban-Rural Development and three other ministries, traditional villages must meet three basic criteria: integrity of traditional architectural style, preservation of traditional site selection and spatial layout, and the living inheritance of intangible cultural heritage [18].

At present, the research on traditional village conservation in China has generally experienced three developmental stages: the initial phase (2000–2010), which focused on individual building protection; the development phase (2010–2018), emphasizing regional linkage protection; and the current phase (2018–present), which advocates for holistic and clustered conservation [19]. Recently, academic perspectives have shifted from the protection of material space to the integrated preservation of cultural ecosystems [20,21].

In the field of traditional village cluster research, several scholars have proposed valuable theoretical approaches. Wang et al. [22] developed the “Three-Level Jump” theory, which has been empirically supported using spatial clustering methods. Bi et al. [23] introduced the “Human Geography Zoning Method”, which has shown notable value in Northwestern Yunnan. Yang et al. [24] proposed the “Cultural Landscape Atlas” paradigm, which deepens the understanding of village cluster culture from three innovative dimensions: morphological characteristics, cultural genes, and spatial narratives.

In recent years, a growing number of studies have focused on the spatial distribution characteristics of traditional villages at different regional scales. At the national level, traditional villages in China exhibit significant spatial heterogeneity and clustering patterns, mainly distributed in southeastern regions and forming several high-density agglomeration zones [25]. At the regional scale, studies in Guangxi, Hunan, and the Tibetan Plateau reveal that traditional villages are typically concentrated in mountainous and hilly areas, with strong dependence on river systems and relatively low accessibility to transportation networks [26,27,28].

Moreover, comparative studies across provinces indicate clear regional differences in distribution balance and concentration. For example, Anhui Province shows the highest concentration and uneven distribution, whereas Yunnan Province presents a relatively lower concentration but larger total number of traditional villages [29]. At the macro scale, the distribution of traditional villages is influenced by a combination of natural factors (e.g., topography, river density, and climate) and socioeconomic factors (e.g., population density and transportation accessibility) [30,31].

Based on the current research, the spatial distribution of traditional villages in Northwest Yunnan presents unique patterns shaped by both environmental and cultural factors. Compared with other regions in China, the spatial distribution of traditional villages in Northwest Yunnan shows both similarities and distinct regional characteristics. Similar to Guangxi and Hunan, traditional villages in Northwest Yunnan are predominantly distributed in mountainous and river-adjacent areas, reflecting a strong dependence on natural environmental conditions such as topography and water systems [26,27]. However, unlike eastern regions such as Zhejiang and Anhui, where traditional villages are more concentrated and exhibit higher accessibility, Northwest Yunnan demonstrates a relatively dispersed yet clustered distribution pattern within localized high-density zones. This pattern is more comparable to plateau regions such as the Tibetan Plateau, where traditional villages exhibit a “large dispersion–small aggregation” pattern due to environmental constraints and ethnic settlement structures [28]. Furthermore, Northwest Yunnan is characterized by multi-ethnic settlements and complex terrain, which leads to stronger spatial heterogeneity compared with lowland or economically developed regions. This indicates that natural constraints and cultural diversity jointly shape the distribution of traditional villages in this region.

From a quantitative perspective, significant regional differences can also be observed. For instance, Yunnan Province contains a large number of traditional villages (over 700), but its spatial concentration is lower compared to provinces such as Anhui, where traditional villages are more unevenly distributed but highly concentrated [32]. At the national scale, traditional villages exhibit strong spatial imbalance and clustering tendencies, mainly distributed in southeastern China, with multiple high-density clusters [33]. In contrast, Northwest Yunnan presents a relatively fragmented distribution with localized clustering, reflecting the influence of mountainous terrain and ethnic settlement patterns. Additionally, previous studies have shown that traditional villages are typically located in areas with elevations below 1000 m and gentle slopes, while also being closely associated with river systems [34]. However, in Northwest Yunnan, the elevation and terrain conditions are more extreme, suggesting that traditional villages in this region exhibit stronger adaptability to harsh natural environments.

The analysis of spatial distribution patterns has become central to traditional village studies. Methodologically, scholars widely adopt kernel density analysis [35,36,37], classification-based coloring [38], or combine kernel density with spatial autocorrelation, buffer zone analysis, spatial Gini coefficient, and imbalance index, to illustrate distribution patterns or evolution of traditional villages [39]. Grid-based classification has also been used to extract spatial characteristics [40]. Other methods such as GIS, remote sensing visualization, fractal theory, space syntax, and network analysis are also employed to explore influencing factors and mechanisms of village distribution [41,42,43,44]. Meanwhile, unmanned aerial vehicle (UAV)-based sensing and intelligent control technologies have also become important tools for spatial data acquisition, environmental perception, and fine-scale analysis in complex geographic environments [45,46,47].

Kernel density estimation (KDE) has been widely applied to identify hotspot clusters of traditional villages. For example, Liu et al. [48] employed ArcGIS 10’s Kernel Density tool to classify the spatial density of 1561 traditional villages into four levels, identifying the areas within the highest density tier (35.128–75.613 units/10,000 km²) as the four major cluster areas across China: Northwestern Yunnan, southeastern Guizhou, the Central Plains, and the Anhui-Zhejiang border. Results showed that Northwestern Yunnan and southeastern Guizhou have significantly higher kernel density values than other regions, confirming the general existence of a “core-periphery” distribution structure. Further spatial autocorrelation analysis reveals deeper spatial interrelations among village distributions. Methodologically, KDE has been extensively adopted in broader geostatistical modeling to estimate spatial intensity surfaces and to analyze uncertainty in irregularly distributed phenomena [49]. These studies highlight the robustness of kernel-based spatial estimation techniques in capturing clustered spatial structures across heterogeneous environments, thereby reinforcing their applicability to traditional settlement analysis.

Multidimensional clustering methods have also been introduced to evaluate the distributional balance of villages. These approaches classify similar village units into the same group based on multiple indicators, maintaining high homogeneity within groups and clear heterogeneity between groups. This enables comprehensive classification and intuitive identification of sample features [50]. For instance, Xue et al. [51] used nearest neighbor index to identify an agglomerated distribution pattern in the Yellow River Basin. Li et al. [27] combined imbalance index and Lorenz curves to confirm the concentration of traditional villages in western and southern Hunan, which negatively correlated with road density and GDP levels, indicating a “sheltering effect” of socio-economic underdevelopment on spatial preservation.

These studies show that the spatial morphology of traditional villages results from both internal spatial construction principles and external factors such as environmental conditions and local residents’ agency [52]. Northwestern Yunnan’s border location and ecological sensitivity further highlight its uniqueness [53]. As key nodes of national cultural security and ecological protection, traditional villages must seek a balance between heritage conservation, tourism development, and community sustainability [54]. Future research should integrate methods such as geographical detectors and social network analysis to further examine the interactions among environmental constraints, cultural diffusion, and policy intervention. This may provide a more robust basis for regionally differentiated conservation and management strategies in multi-ethnic frontier areas [55].

This study provides new insights into the spatial distribution of traditional villages in Northwest Yunnan. It shows that the distribution is mainly influenced by the region’s complex topography and multi-ethnic settlements, with natural and cultural factors playing a stronger role than in more economically developed areas. The study also highlights the more dispersed distribution of villages in Northwest Yunnan, with localized high-density zones, reflecting the region’s unique characteristics. These findings improve our understanding of how natural and cultural factors shape village distribution and support region-specific conservation strategies.

2. Materials and Methods

2.1. Study Area

The northwestern region of Yunnan Province is located in China’s southwestern frontier and includes four administrative areas: Lijiang City, Dali Bai Autonomous Prefecture, Nujiang Lisu Autonomous Prefecture, and Diqing Tibetan Autonomous Prefecture. Geographically, the region begins in the southern section of the Hengduan Mountains, extends east to the Yalong River Basin, west to the eastern slopes of the Gaoligong Mountains on the China–Myanmar border, north to the southeastern edge of the Qinghai–Tibet Plateau, and south to the northern foothills of the Ailao and Wuliang Mountains. The area features a diverse range of landforms and significant elevation differences, with terrain generally higher in the north and lower in the south. As of the end of 2020, the total area reached 87,193 square kilometers, accounting for approximately 22% of Yunnan Province (Figure 1).

Historically, this region served as a key corridor of the Ancient Tea Horse Road, a traditional trade network linking southwestern China with Tibet and surrounding regions, facilitating the exchange of tea, horses, and other goods. In addition to its economic function, it also acted as an important channel for cultural interaction among multiple ethnic groups and is therefore considered an important historical factor in understanding the spatial distribution and cultural landscape of traditional villages in Northwestern Yunnan. More than ten ethnic groups have historically inhabited this area, including the Han, Bai, Yi, Lisu, Naxi, Tibetan, Hui, Pumi, Nu, and Drung peoples. Due to deep historical accumulation and geographical isolation, the region has developed a uniquely diverse cultural landscape, exhibiting rich and varied social and cultural forms [56].

2.2. Data Sources

This study references the official lists of traditional villages, historical and cultural villages, and distinctive ethnic minority settlements released by the Ministry of Housing and Urban-Rural Development of China and the Department of Housing and Urban-Rural Development of Yunnan Province. Based on these sources, a systematic collection of officially recognized traditional villages in Northwestern Yunnan was conducted. As of 2025, a total of 246 traditional villages have been confirmed in this region, forming the core dataset for this study. Each village is associated with both spatial data (e.g., geographic location, elevation, slope, aspect, and distance to rivers) and attribute data (e.g., ethnic composition, architectural type, and cultural characteristics).

Multi-level administrative boundary data for China and Northwestern Yunnan were obtained from the National Geographic Information Resource Directory Service System (National Geomatics Center of China. National Geographic Information Resource Directory Service System. Available online: https://www.webmap.cn/main.do?method=index (accessed on 16 July 2025)). The 12.5 m digital elevation model (DEM), slope data, aspect data, and hydrological data were sourced from the Geospatial Data Cloud (Computer Network Information Center, Chinese Academy of Sciences. Geospatial Data Cloud. Available online: https://www.gscloud.cn/search (accessed on 21 July 2025)).

In terms of attribute data, information such as ethnic composition, types of traditional architecture, and cultural customs was extracted from multiple sources, including 12 government bulletins, 18 news reports, 9 local gazettes, and 21 academic publications. These data were systematically organized, classified, and cross-checked to ensure consistency and reliability. A classification system was then established to standardize these attributes. All attribute data were matched and linked with the spatial data of each village using Excel.

The final village-level attribute matrix contained one record for each of the 246 villages, including administrative location, geographic coordinates, elevation, slope, aspect, distance to major water bodies, ethnic affiliation, architectural type, and cultural-corridor relation. This unified dataset was used as the common analytical basis for the statistical and spatial analyses.

It should be noted that the sample in this study consists of officially recognized traditional villages rather than the full set of rural settlements in Northwestern Yunnan. Accordingly, the spatial patterns identified here describe the clustering characteristics of the recognized heritage sample under the current designation system. They should not be interpreted as a complete representation of the regional settlement system, nor as evidence that such attributes are produced by designation itself.

2.3. Research Methods

2.3.1. Research Framework

This study adopts a multidimensional analytical approach to systematically examine traditional villages in Northwestern Yunnan. First, spatial analysis is conducted using the ArcGIS Pro 3.4 platform (Esri. ArcGIS Pro, Version 3.4), focusing on three main aspects: (1) the spatial distribution equity of traditional villages; (2) characteristics of spatial clustering; and (3) spatial correlations with natural geographical factors such as elevation, slope, aspect, and hydrology. Second, a four-dimensional analytical framework is established, consisting of four dimensions: (1) spatial morphology, referring to the spatial distribution patterns and density characteristics of villages; (2) geomorphological features, including elevation, slope, and terrain conditions derived from DEM data; (3) cultural attributes, represented by ethnic composition and proximity to the Ancient Tea Horse Road as a cultural corridor; and (4) architectural forms, defined by dominant residential structure systems (log cabin, qionglong, and timber frame). Each dimension is analyzed in detail. Finally, a hierarchical overlay model is applied to integrate these four dimensions, constructing a typological classification system for traditional villages with distinctive regional characteristics.

In this study, a “cluster” is defined as a statistically significant spatial aggregation of traditional villages identified through quantitative spatial analysis methods. Specifically:

• A clustered pattern is identified when the Average Nearest Neighbor Index ($R$) is less than 1 and statistically significant (p < 0.05);
• A positive and statistically significant Global Moran’s $I$ (p < 0.05) indicates spatial clustering;
• In kernel density analysis, areas within the top 50% of density values are defined as high-density cluster zones.

Therefore, “cluster” in this study refers to statistically validated spatial aggregation rather than a descriptive expression.

2.3.2. Methods for Analyzing Spatial Distribution Characteristics

(1)

Spatial Distribution Equity Analysis

The spatial distribution equity is evaluated using the geographic concentration index and the imbalance index based on the Lorenz curve. These indices help assess the overall fairness of village distribution and reveal disparities in resource allocation across administrative units.

$$G=100\sqrt{\sum _{i=1}^n{\left({\displaystyle \frac{X_i}{T}}\right)}^2}$$

where $G$ is the geographic concentration index, $X_i$ is the number of traditional villages in city $i$, $T$ is the total number of villages in the study area, and $n$ is the number of administrative units.

$$S={\displaystyle \frac{100{\sum }_{i=1}^n{Y}_i-50(n+1)}{100n-50(n+1)}}$$

where $S$ is the imbalance index, and $Y_i$ is the cumulative percentage of the object in the $i$th administrative unit (sorted in descending order), with $n$ representing the total number of units.

(2)

Spatial Clustering Analysis

The spatial clustering pattern is analyzed using the Average Nearest Neighbor Index, which helps to determine whether the distribution exhibits a clustered pattern ($R$ < 1), a random pattern ($R$ ≈ 1), or a dispersed pattern ($R$ > 1). A Z-score test is applied to evaluate statistical significance.

$$R={\displaystyle \frac{r_1}{r_E}}=2\sqrt{r_{1}D}$$

where $R$ is the nearest neighbor ratio, $r_1$ is the observed mean distance between neighboring villages, $r_E$ is the expected distance, and $D$ is the point density.

The Global Moran’s $I$ is used to measure spatial autocorrelation among villages:

$$I={\displaystyle \frac{n{\sum }_{i=1}^n{\sum }_{j\ne i}^n{w}_{ij}\left(X_i-\overline{X}\right)\left(X_j-\overline{X}\right)}{{\sum }_{i=1}^n{\sum }_{j\ne i}^n{w}_{ij}\sum _i^n{\left(X_i-\overline{X}\right)}^2}}$$

where $I$ is the Global Moran’s $I$, $n$ is the number of units, $X_i$ and $X_j$ are the number of villages in units $i$ and $j$, $\overline{X}$ is the mean value, and $w_{ij}$ is the spatial weight matrix.

(3)

Spatial Distribution and Natural Environmental Influences Analysis

To analyze spatial patterns and environmental influences, the Extract Values to Points tool in ArcGIS Pro 3.4 is used to extract elevation, slope, and aspect values at village locations. Elevation is obtained from DEM pixel values, while slope is calculated from elevation changes along the x and y axes:

$$S=\arctan \left(\sqrt{{\left({\displaystyle \frac{{\partial }_z}{{\partial }_x}}\right)}^2+{\left({\displaystyle \frac{{\partial }_z}{{\partial }_y}}\right)}^2}\right)$$

where $z$ is elevation, and ${\displaystyle \frac{{\partial }_z}{{\partial }_x}}$, ${\displaystyle \frac{{\partial }_z}{{\partial }_y}}$ represent elevation gradients in east–west and north–south directions, respectively.

Aspect is derived from the direction of the slope gradient:

$$\theta =\arctan 2\left({\displaystyle \frac{{\partial }_z}{{\partial }_y}},-{\displaystyle \frac{{\partial }_z}{{\partial }_x}}\right)$$

where $\theta$ is the aspect angle (0–360°); in ArcGIS Pro, 0° represents due north and increases clockwise.

The Near tool in ArcGIS Pro is used to calculate the Euclidean distance from each village to the nearest river:

$$D=\underset{j}{\min }\sqrt{{\left(x_i-x_j\right)}^2+{\left(y_i-y_j\right)}^2}$$

where $D$ is the shortest distance from point $i$ to the nearest river point $j$, and $(x_i,y_i)$, $(x_j,y_j)$ are the coordinates of the village and river points.

Although the study area features significant vertical relief, 2D planar distance remains a consistent and reliable proxy for identifying spatial association patterns and heritage clusters at a regional scale, as 3D physical distances do not significantly alter the relative spatial proximity of settlements to water systems in macro-scale analysis.

2.3.3. Methods for Analyzing Clustering and Differentiation Characteristics

(1)

Spatial Proximity Clustering Analysis Method

Kernel Density Estimation is applied to analyze spatial proximity. Given the region’s complex terrain, low population density, and scattered settlements, along with the obstructive effects of natural features like ridgelines and rivers, the bandwidth is set to 20 km. The results are divided into 10 equal intervals, and the top 5 density levels are used to identify cluster zones of traditional villages.

$$f_n\left(x\right)={\displaystyle \frac{1}{nh}}\sum _{i=1}^{n}k\left({\displaystyle \frac{X-X_i}{h}}\right)$$

where $f_n\left(x\right)$ is the estimated density at location $x$, $X_i$ is the coordinate of observation $i$, $h$ is the bandwidth, and $k$ is the kernel function.

(2)

Geomorphological Similarity Clustering Analysis Method

Geomorphological similarity is identified by combining DEM data with clustering and outlier detection. Elevation values for each village are input into Cluster and Outlier Analysis (Anselin Local Moran’s $I$). Villages classified as high-high (HH) or low-low (LL) elevation clusters are assigned a value of 1; those marked as high-low (HL), low-high (LH), or without significant terrain association are assigned 0.

$$I_i=Z_i\sum _{i=1}^N{W}_i{Z}_j$$

where $I_i$ is the spatial autocorrelation index, $Z_i$, $Z_j$ are the standardized values for observations $i$ and $j$, and $W_{ij}$ is the spatial weight.

(3)

Cultural Integration Clustering Analysis Method

In the cultural integration analysis, due to the predominance of ethnic minority settlements and agrarian culture across the study area, which display high spatial homogeneity, these are not used as separate indicators. Instead, the Ancient Tea Horse Road is designated as the core cultural corridor. A 5 km buffer zone is constructed along this route, and villages within this buffer are defined as part of the cultural influence zone. Cultural clustering is determined by whether villages within the buffer exhibit a continuous, linear distribution along the route.

(4)

Architectural Form Clustering Analysis Method

Based on the classification of 246 traditional villages, three primary residential types are identified: log cabin system, qionglong system, and timber frame system. Each type is encoded numerically (1–3). Kernel density analysis is applied to each type separately, using a 20 km bandwidth, to identify architectural clustering patterns for each residential form.

(5)

Integrated Multifactor Clustering Analysis Method

To comprehensively identify the multidimensional clustering characteristics of traditional villages in Northwestern Yunnan, this study employs a hierarchical overlay model to integrate four analytical dimensions: spatial proximity, geomorphological similarity, cultural integration, and architectural form. At the unit clustering stage, each village is assigned a binary code for each dimension. A value of 1 is given if a village falls within a high-density spatial cluster, belongs to a high-high or low-low geomorphological cluster, is located within the cultural buffer zone along the Ancient Tea Horse Road and displays linear clustering, or is part of an architectural type-based density cluster. A value of 0 is assigned otherwise. In the integrated clustering stage, following the methodology proposed by Mukhametzyanov [57], all dimensions are assigned equal weights, with each factor given a weight of 1. The Intersect tool in ArcGIS is used to overlay the binary clustering results from the four dimensions, producing a composite score matrix ranging from 0 to 4.

Based on the composite scores, traditional village clusters in Northwestern Yunnan are categorized into three types. Villages with a score of 4 are identified as comprehensive clusters, characterized by strong synergistic features across all four dimensions, reflecting a highly integrated and multifaceted settlement pattern. Following this, villages with scores of 3 or 2 are defined as characteristic clusters, shaped by the presence of three or two dominant features. Finally, villages with scores of 1 or 0 are classified as general clusters, exhibiting only limited or singular clustering features. This typology serves as a scientific foundation for formulating differentiated conservation and development strategies according to the specific characteristics of each village type.

2.3.4. Definition of “Cluster”

In this study, the term “cluster” is used in a strictly operational and quantitative sense, rather than as a descriptive or visual interpretation. To avoid ambiguity, clustering is defined differently across analytical dimensions based on specific statistical or spatial criteria.

(1)

Spatial proximity clustering

A village is identified as belonging to a spatial cluster when it is located within the top 50% kernel density values derived from Kernel Density Estimation (KDE) with a bandwidth of 20 km. These high-density zones represent statistically significant concentration areas of settlements.

(2)

Spatial statistical clustering

At the global level, clustering is evaluated using the Average Nearest Neighbor Index ($R$ < 1) and Global Moran’s $I$ ($I$ > 0, p < 0.05), indicating a clustered spatial pattern. At the local level, clustering is identified using Anselin Local Moran’s $I$, where High–High (HH) and Low–Low (LL) types are defined as clustered units.

(3)

Geomorphological clustering

Villages are classified as geomorphological clusters if they belong to statistically significant HH or LL categories in elevation-based Local Moran’s $I$ analysis.

(4)

Cultural clustering

Villages located within a 5 km buffer zone along the Ancient Tea Horse Road and forming a continuous linear distribution are considered part of a cultural cluster.

(5)

Architectural clustering

Villages are identified as clustered if they fall within high-density zones (top 50%) of kernel density surfaces generated for each architectural typology.

(6)

Integrated clustering

A composite clustering score (0–4) is calculated by overlaying the four dimensions. Villages with higher scores indicate stronger multidimensional clustering.

3. Results

3.1. Spatial Distribution Analysis

3.1.1. Spatial Distribution Equity

The analysis reveals that the Gini Index (G) for the spatial distribution of traditional villages in Northwestern Yunnan is 26.4803, significantly higher than the theoretical equilibrium threshold (G₀ = 20.41), indicating a higher concentration spatial inequality. To further illustrate this pattern, the Lorenz curve (Figure 2) is employed to characterize the cumulative distribution of villages across counties. As shown, the curve deviates markedly from the line of equality, forming a typical concave shape, which suggests that traditional villages are unevenly distributed and exhibit a clear concentration effect.

Specifically, Jianchuan County, Yulong County, the Old Town District of Dali City, and Yunlong County exhibit higher concentrations of traditional villages, as indicated by their cumulative proportions in the Lorenz curve, which deviate more significantly from the line of equality compared with other counties.

3.1.2. Spatial Clustering Analysis

To examine the spatial clustering of traditional villages in Northwestern Yunnan, both nearest neighbor analysis and spatial autocorrelation analysis were applied. The results of the nearest neighbor analysis indicate that the observed average distance between villages is 7.905 km, which is significantly shorter than the theoretical expected distance of 13.040 km. This yields a nearest neighbor index R = 0.606 (where R < 1 indicates clustering). The Z-score of −11.825 and a significance level of p < 0.01 confirm the presence of a statistically significant clustered distribution pattern, as indicated by R = 0.606 (<1), Z = −11.825, and p < 0.01, meaning that the observed inter-village distances are significantly shorter than those expected under a random distribution (Figure 3a).

The results of the spatial autocorrelation analysis further support these findings. The calculated Global Moran’s $I$ index is 0.251, much higher than the expected value of −0.043, indicating a moderate to strong positive spatial autocorrelation within the range of −1 to 1. The test statistics show a variance of 0.017, a Z-score of 2.229, and a p-value of 0.026, all of which pass the significance threshold, thereby providing further evidence of spatial clustering (Figure 3b).

The two spatial analysis methods yield highly consistent results, both confirming the clustered nature of traditional village distribution, as evidenced by (1) significantly reduced nearest neighbor distances ($R$ < 1), and (2) positive spatial autocorrelation (Moran’s $I$ = 0.251, p < 0.05), indicating that villages with high densities tend to be located near each other rather than being randomly distributed in Northwestern Yunnan. These findings provide a robust theoretical foundation for subsequent typological classification and conservation planning.

3.1.3. Spatial Distribution and Environmental Influencing Factors Analysis

(1)

Relationship between Traditional Villages and Elevation

Based on spatial visualization of the digital elevation model (DEM) in ArcGIS Pro, traditional villages in Northwestern Yunnan exhibit significant variation in elevation distribution. The results show that 86.18% of traditional villages are concentrated in the mid-elevation zone (1500–2500 m), which mainly comprises small plateau basins and river valley landforms. This elevation band is characterized by a temperate monsoon climate with favorable hydrothermal conditions, well-developed soils suitable for agriculture, relatively gentle terrain with high accessibility, and strong ecological carrying capacity. These advantageous natural conditions collectively supported the formation and development of traditional settlements.

In contrast, only 8.94% of villages are located in high-altitude areas (above 2500 m), where settlement development is constrained by steep terrain, limited accessibility, and cold climatic conditions. The proportion of villages in low-altitude zones (below 1500 m) is the smallest, accounting for only 4.88%, and these are mainly found in specific geomorphic units such as hot-dry river valleys and humid thermal basins (Figure 4,

Table 1. Elevation Distribution of Traditional Villages in Northwestern Yunnan.

Altitude h [m]	Number	Percentage
1000 < h ≤ 1500	12	4.88%
1500 < h ≤ 2000	95	38.62%
2000 < h ≤ 2500	117	47.56%
2500 < h ≤ 3000	17	6.91%
3000 < h ≤ 3500	5	2.03%

).

(2)

Relationship between Traditional Villages and Slope

Northwestern Yunnan is well-known for its Three Rivers Parallel Region, characterized by sharp vertical terrain differences and complex surface morphology. Slope analysis conducted in ArcGIS Pro reveals that topographic slope is a significant factor affecting village distribution patterns. The findings indicate that 54.47% of villages are located in relatively flat areas with slopes of 0–5°, mainly concentrated in plateau basins such as the Lijiang and Heqing basins, as well as lakeside areas around Erhai and Jianhu. These areas provide flat terrain and favorable agricultural conditions, making them core zones for village aggregation.

Another 29.67% of villages are found on gently sloped land (5–15°), often situated on river terraces and terraced fields, maintaining a moderate buffer from major water systems. This distribution pattern balances agricultural needs while minimizing flood risks. Only 15.86% of villages are located in steep-slope areas above 15°, primarily on mountain slopes within the Three Rivers Parallel Region. These villages tend to be smaller in scale and more sparsely distributed due to terrain limitations. The results demonstrate a clear gradient response between slope and village distribution, offering valuable insights into the human–environment relationship in the formation of traditional settlements (Figure 5,

Table 2. Slope Distribution of Traditional Villages in Northwestern Yunnan.

Slope S [°]	Number	Percentage
0 ≤ S ≤ 2	68	27.64%
2 < S ≤ 5	66	26.83%
5 < S ≤ 15	73	29.67%
15 < S ≤ 25	30	12.20%
25 < S ≤ 35	9	3.66%

).

(3)

Relationship between Traditional Villages and Aspect

The aspect analysis in ArcGIS Pro shows that traditional villages in Northwestern Yunnan exhibit a marked orientation preference. Influenced by the region’s north–south mountain orientation and the limited proportion of north-facing slopes, combined with the impact of reduced solar radiation during winter and traditional Chinese geomantic principles—specifically the concept of “embracing the sun and avoiding the shadow”—villages are significantly less likely to be located on north-facing slopes (p < 0.05). In contrast, the distribution of villages across other slope aspects appears relatively balanced, with no statistically significant differences (p > 0.05).

This analysis indicates that although aspect exerts a certain influence on village site selection under complex topographic conditions, its impact is relatively limited compared to more dominant factors such as elevation and slope. This finding provides new empirical evidence for understanding how traditional settlements adapt to their surrounding terrain (Figure 6,

Table 3. Slope Direction Distribution of Traditional Villages in Northwestern Yunnan.

Slope Aspect θ [°]	Number	Percentage
North (0 ≤ θ ≤ 22.5/337.5 < θ ≤ 360)	7	2.85%
Northeast (22.5 < θ ≤ 67.5)	30	12.20%
East (67.5 < θ ≤ 112.5)	49	19.92%
Southeast (112.5 < θ ≤ 157.5)	38	15.45%
South (157.5 < θ ≤ 202.5)	44	17.89%
Southwest (202.5 < θ ≤ 247.5)	41	16.67%
West (247.5 < θ ≤ 292.5)	28	11.38%
Northwest (292.5 < θ ≤ 337.5)	9	3.66%

).

(4)

Relationship between Traditional Villages and Major Water Systems

Using the Near analysis tool in ArcGIS Pro, it was found that approximately 47% of traditional villages are located within 5 km of major water systems. Specifically, 21.95% are within 0–2 km, and about 25.2% are within the 2–5 km range. Interestingly, the proportion of villages declines initially with increasing distance but then rises again, with 19.11% situated between 5 and 10 km and a notable 30% located within the 10–20 km range. Beyond 20 km, only about 4% of villages are present.

This non-linear distance decay pattern reflects a nuanced dependency on water resources. It also illustrates traditional wisdom in settlement site selection that seeks to balance proximity to water with avoidance of flood risks—a concept often described as “close to water but not too close.” This spatial pattern provides further insights into traditional environmental adaptation strategies (Figure 7,

Table 4. Distance Distribution of Traditional Villages from Major Water Bodies.

Distance to River x [km]	Number	Percentage
0 ≤ x ≤ 2	54	21.95%
2 < x ≤ 5	62	25.20%
5 < x ≤ 10	47	19.11%
10 < x ≤ 20	73	29.67%
20 < x ≤ 35	10	4.07%

).

3.2. Identification Results of Traditional Village Clusters

3.2.1. Single Factor Clustering Characteristics

(1)

Spatial Proximity Clustering Analysis

The spatial clustering analysis reveals that traditional villages in Northwestern Yunnan exhibit a distinct “two corridors–six clusters–multiple nodes” spatial structure. Specifically, two primary cultural corridors shape the overall distribution: the northwest-southeast oriented Yangbi River–Yuan River development corridor, and the north–south oriented Lijiang–Lanping cultural corridor. Within these corridors, six major high-density clustering areas have been identified: the Lijiang cultural core zone, the Heqing traditional settlement cluster, the Jianchuan historical village group, the Lanping ethnic cultural zone, the Erhai lakeside cultural circle, and the Weishan traditional village belt. Due to complex terrain conditions, some villages also demonstrate scattered spatial patterns. A total of 78 villages, accounting for 31.7% of the sample, exhibit significant spatial proximity clustering characteristics. This spatial pattern reflects the combined influence of both the regional geographic environment and historical-cultural processes (Figure 8).

(2)

Geomorphological Similarity Clustering Analysis

The geomorphological similarity clustering analysis indicates that traditional villages are strongly coupled with watershed structures, mountain orientations, and basin landforms. By integrating cluster and outlier detection with 3D terrain interpretation, four types of geomorphologically associated village clusters are identified. Based on the HH and LL clusters identified via Anselin Local Moran’s $I$, schematic ellipsoids were manually delineated to group statistically significant village clusters with high spatial proximity. These ellipsoids serve as a visual guide to the 16 geomorphological regions identified in

Table 5. Geomorphological Similarity Cluster Classification.

No.	Name
1	Shangri-la basin region
2	Lugu lake region
3	Tongdian river valley
4	Baishi river valley
5	Yunling mountain region
6	Jian lake region
7	Lijiang basin region
8	Heqing basin region
9	Chenghai lake region
10	Lancang river yunlong section
11	Bonan mountain region
12	Jizu mountain region
13	Yuan river valley
14	Midu basin region
15	Shuimu mountain region
16	Wuliang mountain region

. The first type is the watershed corridor type, distributed along the Tongdian, Baishi, Lancang, and Yuan river systems. These villages align linearly along river valleys, illustrating the spatial organizing effect of river systems. The second type is the lakeside type, found around Lugu Lake, Jian Lake, and Chenghai Lake, where villages are clustered along shorelines. These settings offer microclimatic regulation and habitable conditions that influence site selection. The third type comprises flat basin areas, including the Shangri-La, Lijiang, Heqing, and Midu basins, where village clusters are densely concentrated, highlighting the role of flat terrain in supporting agriculture, construction, and trade. The fourth type is the vertical mountainous type, located in mountainous areas such as Yunlong, Bonan, Jizu, Shuimu, and Wuliang mountains. These villages are often found on sun-facing slopes and reflect how vertical terrain influences architectural forms and village orientation (Figure 9, Table 5).

(3)

Cultural Integration Clustering analysis

In terms of cultural integration, villages located within a 5 km buffer zone along the Ancient Tea Horse Road demonstrate an intersection of ethnic, agrarian, and trade cultures. These villages exhibit unique cultural integration features, combining traditional livelihood practices with the historical legacy of transregional trade. The cultural influence zone is defined based on the spatial extent of the Ancient Tea Horse Road, identifying villages within its 5 km buffer as a specific cluster of trade-cultural legacy (Figure 10).

(4)

Architectural clustering analysis

The architectural clustering analysis reveals a strong correlation between architectural typologies and both environmental and cultural factors. Based on literature review and field data, traditional dwellings in Northwestern Yunnan are categorized into three major systems: the Log Crib System, the Qionglong System, and the Timber Frame System. The three architectural clusters correspond to the three primary residential systems (Log Crib, Qionglong, and Timber Frame) identified through literature review and field data, which serve as pre-defined categories for spatial density analysis. The Log Crib System is characterized by horizontally stacked timber structures and is the oldest and most widely used historical form. It remains prevalent in parts of Nujiang Prefecture and Lijiang (notably in Gongshan, Weixi, Lanping, and Ninglang), mainly among Nu and Lisu ethnic groups. However, its current distribution is relatively scattered. The Qionglong System, with its rammed earth or stone outer walls combined with layered timber frames, is primarily found in the Tibetan areas of Diqing. Its subtypes include flat-roofed Diaofang houses in Deqin and northern Shangri-La, and shingle-roofed houses in central Shangri-La. It is predominantly used by Tibetan populations, with limited presence in nearby multi-ethnic villages. The Timber Frame System includes both Chuandou and Tailiang structures, known for their spatial adaptability and widespread distribution, especially in Dali and Lijiang.

Kernel density analysis of the three systems reveals distinct spatial differentiation. While the Log Crib and Qionglong systems are generally dispersed, they show localized clustering patterns, as indicated by kernel density hotspots for each architectural type in the Tongdian River Valley and the Shangri-La Basin. In contrast, the Timber Frame System demonstrates more higher concentration spatial clustering, particularly in areas with favorable natural conditions, such as the Lijiang and Heqing basins, Jian Lake, Erhai Lake, and the Yuan River Valley (Figure 11).

3.2.2. Integrated Multifactor Clustering Analysis

By integrating the results of spatial proximity clustering, geomorphological similarity clustering, cultural integration clustering, and architectural similarity clustering, the comprehensive clustering characteristics of traditional villages in Northwestern Yunnan are identified. The analysis shows that 52 villages achieve the highest composite score of 4 and are categorized as comprehensive clusters, accounting for 21.14% of all villages. These villages are primarily concentrated in areas such as the Lijiang and Heqing basins, Jian Lake, and the Yuan River Valley. These regions generally benefit from favorable natural environments, higher levels of socio-economic development, and proximity to urban centers or major county towns.

A total of 51 villages fall into the category of specialized clusters, defined by composite scores of 3 or 2, representing 20.73% of the dataset. These villages are mainly located in areas such as the Baishi River Valley, Erhai Lake basin, and Midu basin. Compared with comprehensive clusters, these areas exhibit relatively less favorable environmental conditions, but still retain distinct features in one or more clustering dimensions.

Villages with scores of 1 or 0, classified as general clusters, total 143 in number and account for 58.13% of the total sample. These villages are widely dispersed across Northwestern Yunnan, including remote mountain areas such as the Three Rivers Parallel Region. The general clusters are characterized by their scattered nature and limited expression of composite clustering features (Figure 12,

Table 6. Classification Table of Integrated Clustering Results.

Type	Score	Quantity	Proportion	Proportion
General	0	65	26.42%	58.13%
	1	78	31.71%
Specialized	2	27	10.98%	20.73%
	3	24	9.76%
Comprehensive	4	52	21.14%	21.14%

).

The statistical results presented in Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9, Figure 10 and Figure 11 are derived from the dataset of 246 traditional villages described in Section 2.2.

4. Discussion

This study establishes a four-dimensional analytical framework encompassing spatial morphology, geomorphological features, cultural attributes, and architectural forms, and applies a combination of methods including the geographic concentration index, kernel density estimation, spatial autocorrelation analysis, and a hierarchical overlay model to systematically examine the spatial distribution patterns and clustering characteristics of traditional villages in Northwestern Yunnan.

From the perspective of spatial distribution, traditional villages in Northwestern Yunnan exhibit a distinct pattern characterized by “higher density in the north, lower in the south, clustering along rivers, and multi-core linkage.” Quantitative indicators confirm this trend: the Gini coefficient (26.4803) is significantly higher than the theoretical equilibrium value (15.3627), and the Lorenz curve shows a higher concentration concave form, indicating a highly uneven spatial distribution. The majority of villages (86.18%) are located within the mid-elevation range of 1500–2500 m, and 47% lie within 5 km of major water systems. This pattern reflects the traditional settlement wisdom of “proximity to water without closeness to flood risk.” Spatial autocorrelation analysis (Moran’s $I$ = 0.251) and nearest neighbor index ($R$ = 0.606) further validate the higher concentration clustering tendency of village distribution. These clustering patterns are quantitatively supported by spatial statistical indicators ($R$ < 1, Moran’s $I$ > 0, p < 0.05) and kernel density thresholds.

In terms of single-factor clustering, the spatial proximity cluster forms a “two-corridor, six-cluster, multi-node” structure, with 78 villages (31.7%) exhibiting clear spatial association. Geomorphological similarity clustering identifies four adaptive models: watershed corridor type, lakeside type, flat basin type, and vertical mountainous type. Cultural integration clustering, centered on the Ancient Tea Horse Road, reflects the overlap of trade and ethnic cultures. The architectural clustering reveals a high degree of coordination between three primary architectural systems—Log Crib, Qionglong, and Timber Frame—and local environmental and cultural conditions. To further clarify the relationships among the four analytical dimensions, a synthetic framework diagram is provided (Figure 13), summarizing the interactions between spatial proximity, geomorphological conditions, cultural integration, and architectural typologies. This visualization helps to better illustrate the multidimensional clustering process described above.

From the perspective of integrated clustering, traditional villages can be categorized into three types. First, comprehensive villages, which are predominantly found in regions such as the Lijiang and Heqing basins, Jian Lake, and the Yuan River valley—areas with superior natural and socio-economic conditions—exhibit strong coordination across all four dimensions. These villages should be prioritized for holistic, living heritage conservation, with emphasis on maintaining the integrity of spatial, cultural, and architectural systems, and promoting cultural-tourism integration and sustainable community development. Second, specialized villages, located in areas such as the Baishi River valley, Erhai Lake, and Midu basin, exhibit strengths in two to three core dimensions. For these villages, strategies should focus on enhancing their unique features (e.g., restoration of architectural complexes, revival of cultural scenes) while avoiding homogenized development. Third, general villages, which are scattered across remote mountain regions including the Three Rivers Parallel Area, show weak or singular feature expression. These villages require foundational conservation and targeted capacity-building, with a focus on infrastructure improvement, cultural element excavation, and adaptive strategies such as micro-renewal and ecological compensation to enhance their resilience (Figure 14).

This point is especially important because “traditional village” in the Chinese context is an institutional heritage category rather than a purely organic settlement type. The morphological, geographical, and cultural attributes identified in this study should therefore be understood as the observable characteristics of villages that have been recognized and preserved within this designation framework. In this sense, the present results reveal the spatial regularities of the recognized heritage landscape, rather than the total pattern of all rural settlements in Northwestern Yunnan. Nonetheless, the study has several limitations. First, because the analysis relies on officially recognized traditional villages, the observed spatial and cultural patterns may partly reflect the logic of heritage designation and data availability, rather than the complete distribution of all settlements in the region. Future research should incorporate nearby non-designated settlements as comparative or control samples to test the broader representativeness of the identified clustering patterns. Second, due to the vast and topographically complex nature of Northwestern Yunnan, data on cultural attributes and architectural forms rely primarily on literature and official listings, lacking large-scale field surveys and in-depth interviews. This may affect the precision of classification and limit the depth of cultural interpretation.

5. Conclusions

This study provides a systematic examination of the spatial distribution patterns and multidimensional clustering characteristics of traditional villages in Northwestern Yunnan. Using methods such as the geographic concentration index, spatial autocorrelation analysis, kernel density estimation, and hierarchical overlay modeling, the research offers a comprehensive quantitative and visual analysis across dimensions of spatial equity, clustering, environmental adaptability, and cultural–architectural features.

The core findings are as follows:

Traditional villages in Northwestern Yunnan exhibit a highly uneven and clustered spatial pattern, characterized by a “two corridors—six clusters—multiple nodes” structure.

Settlement distribution demonstrates strong coupling with mid-elevation plateau basins, river valley systems, and trade-cultural corridors shaped by the Ancient Tea Horse Road.

Multidimensional integration classifies villages into three typologies—comprehensive (21.14%), specialized (20.73%), and general (58.13%) clusters—reflecting different levels of coordination among spatial, environmental, cultural, and architectural dimensions.

Despite these limitations, this study provides a structured spatial interpretation of officially recognized traditional villages in Northwestern Yunnan and offers a replicable framework for identifying differentiated clustering types in mountainous multi-ethnic regions. The results may support more context-sensitive heritage conservation and rural management strategies by distinguishing comprehensive, specialized, and general village clusters according to their spatial, geomorphological, cultural, and architectural characteristics.The framework may also serve as a reference for prioritizing conservation resources and designing differentiated management pathways in ecologically fragile and culturally diverse rural areas.