The Driving Mechanisms of Traditional Villages’ Spatiotemporal Distribution in Fujian, China: Unraveling the Interplay of Economic, Demographic, Cultural, and Natural Factors

Abstract

Traditional villages (TVLGS) have significantly declined as a result of China’s fast urbanization, especially in Fujian Province, where efficient conservation efforts are hampered by a lack of thorough study. The geographical and temporal distribution features of Fujian’s traditional villages (FTVLGS) are investigated using ArcGIS 10.8 and GeoDa software. Additionally, it identifies 18 driving factors to investigate the primary influences and interaction mechanisms through a combination of Python 3.7 and GeoDa 1.16. The results show that: (1) FTVLGS are distributed both spatially and temporally in a pattern that is oriented from northeast to southwest to east. Over time, the distribution center of gravity moves from north to southeast, increasing directional tendencies and broadening the distribution area. (2) The impact of each driving factor on the spatial distribution of TVLGS varies, with the strongest influence being the interaction between average annual precipitation and the straight-line distance from provincial highways. The straight-line distance between TVLGS and provincial highways is found to be the most significant factor affecting their distribution. This study clarifies the intricate dynamics associated with the distribution of TVLGS and the factors that influence them, providing evidence-based recommendations for the future preservation and advancement of these TVLGS. It also aims to enhance the connectivity of developmental elements at a regional scale and to foster the advancement of global tourism within TVLGS.

1. Introduction

As vital custodians of China’s material and immaterial heritage, TVLGS exhibit significant polyvalent worth spanning historical, cultural, scientific, aesthetic, societal, and economic domains [1]. Studying and analyzing TVLGS not only supports the dynamic conservation of cultural heritage but also injects endogenous momentum into rural sustainable development by excavating their historical and cultural resources—an imperative amid global efforts to balance heritage preservation with socioeconomic progress. However, rapid urbanization and industrialization in recent decades have posed severe threats to TVLGS worldwide: rural hollowing and aging, agricultural land fragmentation, environmental pollution, and systemic risks to rural social structures have led to a sharp decline in the number of TVLGS, with some even facing extinction [2]. In China, the Ministry of Housing and Urban-Rural Development has jointly recognized 8155 “Chinese TVLGS” across six batches since 2012, and coordinated cultural protection measures have gradually fostered social consensus on TVLGS stewardship. Yet, these efforts still lack robust theoretical and empirical support from regional-scale systematic research, particularly for provinces with distinct geographical and cultural characteristics like Fujian.

1.1. Literature Review: Theoretical Progress, Methodological Limitations, and Knowledge Gaps

Current global and domestic research on TVLGS has formed an interdisciplinary landscape, but critical gaps remain in theoretical integration, methodological rigor, and regional targeting—limitations that this study aims to address.

From a theoretical perspective, early research primarily adopted a “morphological-static” paradigm, focusing on the description and classification of TVLGS spatial forms [3], social and cultural attributes [4], and ecological environments [5]. Domestic scholars initially concentrated on micro-level topics, such as TVLG morphological evolution [6], protection and development pathways [7], and human settlement optimization [8], constructing a foundational understanding of TVLGS’ characteristics but rarely linking these micro-observations to macro-regional dynamics. With the rise in rural tourism and climate change concerns, research perspectives expanded to include “evolutionary-dynamic” frameworks: scholars examined the adaptive evolution of TVLG human settlements under environmental change [9], assessed conservation challenges amid socioeconomic transitions [10], and built evaluation systems for TVLGS’ protection [11]. Additionally, the “spatial-synthetic” paradigm emerged, with studies exploring TVLGS’ spatial distribution [12], patterns [13,14], and differentiation [15,16]—yet these works often treated spatial patterns as independent outcomes rather than integrating them with historical, economic, or cultural processes, resulting in a fragmented theoretical understanding of TVLGS’ spatiotemporal dynamics.

Methodologically, technological advancements have driven progress but also revealed limitations. The application of Geographic Information Systems in rural spatial accessibility [17] and landscape gene identification [18], along with the use of machine learning for multi-source data processing [19], has provided new tools for TVLGS research. For example, Liu et al. [20] combed through Moran’s I lined ArcGIS, GeoDa, and Python to analyze factors influencing TVLGS’ spatial distribution in Henan, demonstrating the value of multi-tool synergy; Xia et al. [21] used ArcGIS and GeoDa to explore spatial autocorrelation between ecological land and thermal environments, offering insights into quantitative spatial analysis; and Chen et al. [22] integrated social media data and machine learning to link built environments with tourist sentiment, expanding data sources for TVLGS studies. However, three methodological shortcomings persist.

Scale imbalance: Most studies focus on either the national scale [23,24,25,26] or individual villages [27,28], with provincial-scale research—especially for coastal regions like Fujian—remaining underdeveloped.

Factor isolation: Existing works rarely integrate natural, economic, demographic, and cultural factors simultaneously. For instance, studies on FTVLGS mention topographic or hydrological influences [29,30,31] but fail to quantify how these interact with economic activities or cultural diffusion, leading to weak explanatory power for distribution patterns.

Methodological singularity: Few studies adopt a “multi-method collaboration” framework. While random forest or spatial regression models are increasingly used, they are often applied in isolation, unable to address high-dimensional variable redundancy or spatial autocorrelation—critical issues for disentangling the coupled driving mechanisms of TVLGS’ distribution.

Regionally, FTVLGS research suffers from a prominent “micro-focus, macro-absence” bias. Existing studies on Fujian primarily examine residential architecture [32], street space [33], or single-village cultural heritage [29], with scarce analysis of provincial-scale spatiotemporal patterns. Though some literature notes that FTVLGS’ layout is shaped by topography, river systems, and historical transportation [29,30,31], and identifies rivers as a key geographic factor [34]. These claims lack (1) systematic data integration; (2) quantitative validation of qualitative factors; and (3) exploration of factor interactions. Consequently, the current understanding of FTVLGS’ distribution remains fragmented, hindering the formulation of targeted, province-wide protection strategies.

1.2. Research Objectives and Core Question

Based on this, this study takes 1292 FTVLGS as the research subject, integrates multi-dimensional data including economic, demographic, natural, and cultural data, and uses tools such as ArcGIS 10.8, Python, and GeoDa to analyze the spatial distribution patterns of FTVLGS, as well as their spatial distribution characteristics, migration trajectories of the distribution center of gravity, and evolution laws of the distribution range from the pre-Tang Dynasty to the Qing Dynasty. It clarifies the distribution differences and geographical correlations across different historical periods, selects 18 driving factors from two dimensions—natural factors and human factors—and comprehensively constructs an information database of FTVLGS. Subsequently, by applying methods including random forest, spatial lag regression model, and geographical detector, this study identifies core driving factors and analyzes the interaction mechanism among multiple factors.

Centering on the above work, the core research question addressed in this study is as follows: what is the spatial survival law, and what are the key factors driving the formation, evolution, and preservation of FTVLGS, as well as the interaction mechanism between these factors (Figure 1)?

2. Materials and Methods

2.1. Study Area

Fujian Province (~124,000 km²) lies on China’s southeastern coast (23°30′–28°19′ N, 115°50′–120°47′ E) and has a mild subtropical climate, with annual average temperatures of 17–21 °C and annual rainfall of 1400–2000 mm.

Topographically, Fujian is predominantly hilly, with over 90% of its land being elevated terrain. Two major mountain ranges cut diagonally across the province, intersecting a mosaic of river valleys and basins. From the western inland to the coast, the landscape transitions from hills and plateaus to flat coastal plains, with elevations generally decreasing from the northwest to the southeast.

Village distribution in Fujian is closely tied to its hilly terrain and complex river system (density: 0.1 km/km²). While Fujian has been part of China’s administrative system since the Han Dynasty, its core has long centered on the eastern coastal plains—an area with diverse ethnic communities and farming practices. Traditionally, villages followed the “stream in front, hill behind” layout, as surrounding mountains were critical for agricultural and pastoral activities. This design principle still shapes current village arrangements (Figure 2).

Unfortunately, FTVLGS’ landscapes and geomorphic systems face severe challenges. Regional economic imbalances, inadequate cultural preservation, and environmental degradation have caused significant consequences, including population decline, rural abandonment, and neglected farmland.

2.2. Data Sources and Processing

As of 2023, China has carried out in-depth and systematic initiatives for the preservation of its cultural heritage. Notably, the number of TVLGS inscribed on the national-level protection list has amounted to 8156, thereby forming the world’s largest concentration of protected traditional village sites. The country has made huge strides in safeguarding these cultural gems. For this paper, the list of 552 NDTVs (batches 1–6) was sourced directly from the Ministry of Housing and Construction’s official website. To supplement this core dataset, we further obtained lists of 917 provincial-level traditional villages (batches 1–4) and 74 villages with unique ethnic minority characteristics (batches 1–3) from the official website of the State Civil Affairs Commission.

To build a geospatial database of these FTVLGS, we obtained village latitude and longitude coordinates via the Baidu map and then imported that data into ArcGIS 10.8. The urban administrative divisions for the province came courtesy of the Fujian Provincial Geographic Information Public Service Platform. For environmental data, we turned to the National Earth System Science Data Center for average annual temperature and daily precipitation data at the county level. OpenStreetMap filled in the gaps on the province’s waterways and transportation networks.

The geospatial database construction for Fujian’s villages involved multiple authoritative data sources and analytical procedures. Village coordinates were georeferenced using Baidu Maps’API and subsequently integrated into ArcGIS 10.8 software for spatial analysis. Administrative boundary data at the urban jurisdictional level were sourced from the Fujian Provincial Geographic Information Public Service Platform. Environmental parameters, including mean annual temperature and precipitation metrics, were acquired from the National Earth System Science Data Center, with data aggregation conducted at the county administrative unit level. Hydrographic features and transportation infrastructure data were supplemented by OpenStreetMap’s collaborative mapping resources to ensure comprehensive network coverage across the provincial territory.

Fujian Province alone boasts 494 NDTVs, accounting for about 6% of the national total. As shown in Figure 3, these villages are scattered across Fujian’s ten municipal-level administrative districts, although their distribution is not exactly even.

2.3. Research Methodology Framework

To address the three core challenges in provincial-scale traditional village research—namely, the disconnect between macro patterns and micro factors, interference from spatial autocorrelation, and difficulties in quantifying the role of multi-factor interactions—this study develops a progressive methodological framework structured as the “preliminary groundwork-feature quantification-variable screening—intensity modeling-mechanism validation.” Within this framework, each step is sequentially linked: the output of the preceding step serves directly as the input for the subsequent step, thereby forming a complete logical closed loop.

2.3.1. Prefoundation: Multi-Source Data Integration and Geodatabase Construction

The collected data were standardized to form a structured geographic database comprising “1292 village samples + 18 indicators covering natural, economic, demographic, and cultural aspects,” providing foundational data for subsequent spatio-temporal feature analysis.

2.3.2. Quantification of Spatial and Temporal Distribution Characteristics

Through spatial measurement methods, abstract features such as “distribution pattern, agglomeration hotspots and dynamic evolution” of villages were transformed into quantitative indicators so as to clarify “which regions/time periods have a significant pattern of village distribution” and were then provided the “feature orientation” for the subsequent “core variable screening”, as well as the “feature orientation” for the subsequent “core variable screening” (e.g., prioritize factors related to “high agglomeration areas”).

Average Nearest Neighbor Index (NNI)

By analyzing the actual data alongside the nearest neighbor distances, we can assess the spatial distribution patterns of geographic features in the FTVLGS. These patterns may be categorized as clustered, uniform, or random.

$$R={\displaystyle \frac{r_1}{r_2}}=2\sqrt{D}$$

(1)

$$r_E={\displaystyle \frac{1}{2}}\sqrt{{\displaystyle \frac{n}{A}}}$$

This provides a quantitative basis for judging whether FTVLGS exhibit agglomeration characteristics.

Gini Coefficient

Computed Gini Coefficient (Python 3.7) revealed uneven village distribution, guiding the prioritization of variables correlated with high-concentration regions. The formula is shown below:

$$G=1-{\sum _{i=1}^n\left(pi\right)}^2$$

(2)

Moran’s I

Global Moran’s I identified inherent spatial dependence in the dataset [35], providing a basis for selecting spatial regression models in later steps (note: corrected notation: n = number of spatial units, and w_ij = spatial weight matrix).

$$I={\displaystyle \frac{n{\sum }_{i=1}^n{\sum }_{i=1}^n{w}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}{{\sum }_{j=1}^n{\sum }_{j=1}^n{w}_{ij}{\sum }_{i=1}^n{\left(x_i-\overline{x}\right)}^2}}$$

(3)

Kernel Density Estimation Method (KDE)

KDE is a non-parametric technique used to suss out the probability density function [36].

$${\mathrm{f}}_h\left(\mathrm{x}\right)={\displaystyle \frac{1}{n\mathrm{h}}}\sum _{i=1}^{\infty }k\left({\displaystyle \frac{x-x_i}{h}}\right)$$

(4)

Standard Deviation Ellipse

The evolving spatial footprint of FTVLGS can be effectively tracked by charting the movement of its average geographic center over time.

The formula is shown below:

$$C={\displaystyle \frac{1}{n}}\left\{\begin{array}{c}{\sum }_{i=1}^n{\overline{x_i}}^2{\sum }_{i=1}^n{\overline{x}}_i{\overline{y}}_i\\ {\sum }_{i=1}^n{\overline{x}}_i{\overline{y}}_i{\sum }_{i=1}^n{\overline{x_i}}^2\end{array}\right\},\left\{\begin{array}{c}\left(x_i-\overline{x^i}\right)\\ \left(y_i-\overline{y^i}\right)\end{array}\right.$$

(5)

KDE and standard deviation ellipse were implemented in ArcGIS to visualize agglomeration hotspots and shifts in the geographic center over time. This clarified the need to include historical context in driving factor analysis.

Buffer Zone Analysis

This study adheres to the concept of accessible tourism routes, concentrating on key river systems, national highways, and related transportation infrastructure. Following multiple iterations, we established a series of concentric buffer zones, spaced at 1000 m, 2000 m, 3000 m, and 4000 m intervals. We then layered an analysis of TVLGS onto these buffer zones, allowing us to quantify the number of villages encompassed within each zone.

2.3.3. Python: Core Driving Variable Screening

Extract key driving factors from 18 initial indicators to ensure modeling validity. Based on the quantified features (e.g., high-agglomeration hotspots and traffic correlation) from Section 2.3.2, a random forest (an ensemble learning algorithm [37] was implemented via Python’s Scikit-learn library. The model was trained with an 8:2 data split and 500 decision trees; mean square error (MSE) and R₂ verified its reliability.

$$\mathrm{F}\left(x\right)={\displaystyle \frac{1}{\mathrm{M}}}{\sum }_{m=1}^M{f}_m(x)$$

(6)

This step screened core variables and eliminated multi-collinearity, resolving the “multi-factor interference” challenge.

2.3.4. Quantifying the Driver Strength of Core Factors

Address spatial autocorrelation (detected in Section 2.3.2) and quantify the independent influence of each core variable. Traditional regression models fail to account for spatial dependence, so a spatial lag model (SLM) was constructed using GeoDa. The model adopted a Rook spatial weight matrix, with [38] the following.

Dependent variable: Village density (derived from KDE in Section 2.3.2).

Independent variables: A total of 18 variables (standardized via Z-score).

Formula:

$$Y=\rho WY+X\beta +\epsilon$$

(7)

2.3.5. Revealing the Multifactorial Coupling Driving Mechanism

Compensate for the limitation of single models (e.g., SLM) in identifying interactive effects, and finalize the driving mechanism. Use GeoDetector [39] in GeoDa to perform “single factor detection” and “interaction detection”.

Factor discretization: Natural breakpoint method.

Explanatory power quantification: q-value calculation is as follows:

$$q=1-{\displaystyle \frac{{\sum }_{h=1}^H{n}_h\cdot {\sigma }_{Y_h}^2}{N\cdot {\sigma }_Y^2}}$$

(8)

This step verified the explanatory power of core variables and classified their interaction types (e.g., synergy and non-linear enhancement), solving the “multi-factor coupling quantification” challenge.

3. Results

3.1. Spatial and Temporal Distribution Characteristics

3.1.1. Types of FTVLGS

The spatial distribution characteristics of FTVLGS were quantitatively analyzed using the nearest neighbor index method (Equation (1)) [37]. Statistical analysis revealed a determination coefficient of 0.85, accompanied by a statistically significant Z-score of −10.10 (p < 0.001). The observed mean nearest neighbor distance demonstrated substantial deviation from the theoretically expected distance under complete spatial randomness. These metrics collectively indicate a non-random spatial configuration, with the NNI value falling significantly below the uniform distribution threshold of 1. The analytical outcomes substantiate that FTVLGS in the study area exhibit a clustered spatial distribution pattern (Figure 4), reflecting distinct agglomeration characteristics in their geographical arrangement.

3.1.2. Spatial Layout

The Gini Coefficient (Equation (2)) was calculated using the distribution and data of TVLGS within the administrative divisions of each prefecture-level city in Fujian Province as the unit of municipal administrative divisions. The greater curvature of the Lorenz curve, which deviates from the line of uniform distribution, indicates an uneven distribution of villages within the city area. Ningde has 154 NDTVs, 143 PLTVs, and 31 characteristic villages of ethnic minorities, for a total of 328, followed by more than 160 TVLGS in Sanming, Longyan, and Nanping, and more than 100 TVLGS in Zhangzhou, Quanzhou, and Fuzhou, while there are 57 TVLGS in Putian, 14 in Pingtan, and just 3 in Xiamen (Figure 5).

This study investigated the spatial distribution patterns of TVLGS across 84 county-level units in Fujian Province, China, utilizing GeoDa software with a Rook contiguity-based spatial weight matrix. Spatial autocorrelation analysis revealed a statistically significant positive clustering pattern (Global Moran’s I = 0.20, p = 0.007, Z = 3.25) through 999 Monte Carlo permutation tests (p < 0.01). These results demonstrate pronounced spatial dependence among TVLGS, characterized by “high-high” or “low-low” agglomeration types at the provincial scale. Such spatial autocorrelation underscores the continuity and interdependence of regional cultural resource distribution, reflecting inherent geospatial linkages in heritage preservation dynamics.

Delving deeper with Moran’s I (Equation (4)) analysis, we found that Ningde and Nanping in the north and center of Fujian, Longyan in the west, and Sanming and Quanzhou in central Fujian, are hotspots—tight clusters exhibiting significant spatial connections among their TVLGS (Figure 6). On the other hand, Fuzhou’s Cangshan and Mawei Districts, along with Changtai County in Zhangzhou, form “cold spots”, showing both low density and weak connections between villages. Minqing County in Fuzhou bucks the trend, identified as a low–high outlier, suggesting that even with fewer villages, there is a noticeable spatial relationship at play. In short, it looks like urbanization is having an uneven effect on the spatial distribution of FTVLGS, leading to some imbalances.

The spatial distribution characteristics of FTVLGS were analyzed through KDE employing the Spatial Analyst module in ArcGIS 10.8. As illustrated in Figure 7, the analysis revealed marked spatial heterogeneity in village distribution patterns. The kernel density surface demonstrates significant clustering tendencies, with three principal concentration zones identified: northeastern, eastern, and southwestern sectors of the province. These core agglomerations contrast sharply with the fragmented distribution patterns observed in the remaining regions, exhibiting a dispersed spatial configuration with intermittent low-density clusters.

The most prominent high-density cluster (KDE = 2.86/km²) was identified in Ningde City, forming the primary distribution nucleus. Secondary clusters emerged along regional interfaces, particularly at the administrative boundaries between Longyan and Zhangzhou municipalities, within Fuzhou’s jurisdictional area, and at the transitional zone connecting Quanzhou and Sanming. This hierarchical density pattern suggests strong correlations between village preservation and regional geographical constraints, historical migration routes, and cultural diffusion processes. The spatial autocorrelation analysis (Moran’s I = 0.417, p < 0.01) further confirms the statistical significance of this clustered distribution pattern.

3.1.3. Diachronic Spatial Heterogeneity of FTVLGS: Distribution Dynamics Across Historical Periods

Despite historical discontinuities and wartime-induced archival fragmentation, this study systematically collated chronological records of TVLGS establishment in Fujian. Using a multi-stage analytical framework, temporal data were categorized into six phases: pre-Tang, Tang/Inter-Tang, Song, Yuan, Ming, and Qing. Via rigorous archival verification and geospatial cross-referencing, 645 TVLGS were chronologically authenticated, with founding epochs quantified as follows: pre-Tang (n = 14), Tang (n = 114), Song (n = 178), Yuan (n = 65), Ming (n = 169), and Qing (n = 73) (Equation (3)).

Settlement patterns evolved distinctly across periods. Pre-Tang villages showed early clustering in northern Fujian, reflecting initial Han migration. The Tang Dynasty saw expansion into central and eastern basins (Figure 8), linked to hydraulic infrastructure and the tea trade. Song-Era growth was sharp (+56.1% over Tang), extending southwest with agricultural reclamation. By the Yuan period, three settlement belts emerged: the northern cultural core, a central transitional zone, and a southwestern coastal interface.

Ming-Era villages, though fewer than in the Song, reached peak spatial complexity, with high-density clusters forming in mountainous nodes—suggesting defensive strategies against piracy. Qing distributions contracted, focusing on counties like Fuan, while central Fujian saw significant settlement loss (χ² = 7.89, p < 0.05), likely due to coastal evacuation policies.

Overall, the distribution of central villages is concentrated near the Wufeng and Daiyun mountain ranges in north-central Fujian. Beginning in the Yuan Dynasty, these central settlements aligned along the north–south mountain axes, exhibiting an overall trend of first developing in the north, then the south, and finally shifting eastward (Figure 9).

Spatial analysis indicated the most robust distribution during the Tang period (flattening rate lowest). Expansion peaked in the Yuan period (ellipse area 67,144.1 m²), with a wide but weakly directional spread. The Ming period saw contraction and weakened directionality, while the Qing period brought renewed expansion (ellipse area 63,501.14 m²) and stronger directional trends.

From the Qin and Han Dynasties through the Three Kingdoms period and the two Jin Dynasties, Fujian long remained a southeastern frontier, featuring vast yet sparsely inhabited land. It was not until the Yuan Dynasty that the wars in the Central Plains prompted a large-scale southward migration of Han Chinese into Fujian. The immigrants initially gathered in northern Fujian, utilizing soil and water resources to establish traditional agricultural areas, and then expanded along the Minjiang and Jinjiang River basins to central and southern Fujian, with agriculture always dominating the economy.

Since the late Tang Dynasty, the rise in the Maritime Silk Road promoted the development of Fujian’s marine economy. To the Song Dynasty, Quanzhou jumped to the first port in the East. The imperial court was to “open the sea rich country” as a policy to promote maritime trade. Commodity economy boomed in the Ming Dynasty—Fuzhou, Quanzhou, and Zhangzhou silk fabrics were exported overseas; Wuyi black tea was popular in Europe, Nanping, Ningde, and other traditional villages in northern Fujian, and the clustering effect is significant, with rapid economic growth.

During the Ming and Qing Dynasties, maritime trade saw a gradual transition from being government-dominated to privately operated. Concurrently, the coastal economy’s focal point moved southward, and villages spread toward the southeast. Nevertheless, the sea ban policy inflicted severe harm on the economy of coastal villages. Take the first year of the Kangxi reign in the Qing Dynasty as an instance: outside the relocated boundaries, merely 68 traditional villages persisted, making up 1.3% of the total. Among these, 45 villages had been founded prior to the Yuan Dynasty, while the Qing Dynasty saw the establishment of the smallest number, only 20 (Figure 10).

In terms of population, from the Southern Song Dynasty to the early Yuan Dynasty, Fujian’s population boomed, and the narrow land and dense population drove out-migration. During the Ming and Qing dynasties, war and famine forced large numbers of people to move to Taiwan or overseas to make a living, and out-migration became the mainstream of population movement.

Land control policies exerted notable influences as well. The coastal “land reclamation” policy caused certain villages close to mudflats (like those around Meizhou Island in Putian) to vanish because of alterations in land-use nature. Meanwhile, regulations prohibiting the reclamation of mountains and forests, such as the “Measures for the Management of Ecological Public Welfare Forests in Fujian Province”, resulted in the disappearance of some villages in western and northern Fujian. Moreover, these regulations have enabled villages in western and northern Fujian to retain the original layout of “backing mountains and facing water”, thus creating a “synergistic protection effect” through the combination of policy and natural geography.

To summarize, Fujian has historically experienced the transition from an inland agricultural economy to a marine economy, and the regional development has shown the process of expansion and contraction from land to sea. These factors have contributed to the migration of traditional villages to southern and southeastern Fujian, and the formation of agglomerations in northern Fujian, the Minjiang–Jinjiang river basin, and Longyan.

3.2. Geospatial Determinants of FTVLGS: A Multidimensional Drivers Analysis

3.2.1. Natural Factor

Per historical–geographical research and regional documentation [40], during the TVLGS’ establishment period (Tang–Qing), Fujian experienced no major geological disasters (e.g., large earthquakes, volcanic activity) that altered core topography (e.g., Jiufeng/Daiyun Mountain orientations, primary river basins). Major rivers (the Min River and the Jiulong River) maintained stable courses, and climate indicators (annual temperature and precipitation) showed only interannual fluctuations—not sustained extremes (e.g., prolonged droughts and regional cold spells). This millennial stability in terrain, hydrology, and climate enabled long-term constraints on TVLGS’ siting (e.g., “mountains-and-rivers” layout) and spatial distribution.

Elevation and Slope

Fujian has an elevation range of −29 to 2145 m, with higher elevations in the northwest, lower in the southeast, and a saddle-shaped cross-section. Key topographic elements include the following: eight major mountain ranges (Wuyi, Vulture Peak, Tai Mo Shan, Tortoise Mountains, Daiyun, and Boping Ridge), with Huanggang Mountain as the province’s highest peak; disconnected river valleys/basins between major ranges; narrow eastern coastal plains (Zhangzhou, Fuzhou, Xinghua, and Quanzhou); an extensive, meandering coastline with bays, peninsulas, and islands (Pingtan Island as the largest; Figure 11a).

Multivariate geospatial modeling identifies topography as a dominant constraint on 1292 studied villages: Elevation distribution: 270 villages (20.83%) < 200 m; 452 (35.01%) in 200–500 m hills; 570 (44.15%) in >500 m mountains. The province’s elevation span is −3 to 1455 m. Kernel density (paired with DEM data): Highest density clusters in upper river reaches and hill-plain transition zones; a secondary peak in hilly terrain. Overall trend: Village density declines with increasing elevation (Figure 11b).

Utilizing the Neighborhood Analysis tool within ArcGIS 10.8’s Spatial Analyst module (Equation (5)), this study computed mean slope gradients within 500 m radial buffers surrounding traditional village centroids (Figure 11b). Slope classification adhered to the Third National Land Survey technical protocols for Fujian Province, categorizing terrain into five gradient classes (

Table 1. Slope grade classification table.

Number	Slope Grade	Slope Range (°)
1	Grade 1 (Flat)	≤2
2	Grade 2 (Gentle)	2–6
3	Grade 3 (Moderate)	6–15
4	Grade 4 (Steep)	15–25
5	Grade 5 (Extreme)	≥25

):

As delineated in Figure 11d, the spatial distribution exhibits pronounced slope selectivity across 1296 analyzed villages:

• Low-gradient zones (Grades 1–2): 233 villages (18.04%), predominantly in coastal alluvial plains;
• Moderate slopes (Grade 3): 391 villages (30.31%), characteristic of foothill transitional ecotones;
• Steep terrain (Grade 4): 459 villages (35.58%), forming the primary settlement belt;
• Extreme gradients (Grade 5): 209 villages (16.20%), limited to constrained mountain niches.

Comparative analysis of mean buffer slope versus centroid slope revealed three distinct geomorphic adaptation strategies:

• Lowland nuclei (n = 481, 37.11%): Higher mean buffer slopes indicate villages centered in valley bottoms surrounded by steeper peripheries;
• Upland cores (n = 467, 36.03%): Lower buffer slopes suggest ridge-top settlements with descending slope gradients;
• Planar habitation (n = 344, 26.55%): Negligible slope differential denotes terrace-optimized flatland villages.

TVLGS in Fujian Province are predominantly situated in higher altitude regions, which are characterized by more abundant land resources, leading to a greater concentration of these villages in such areas. Conversely, the lowland areas, despite possessing greater population density and economic advancement, demonstrate a somewhat reduced quantity of TVLGS owing to their limited geographical expanse. This observation underscores the significant influence of modernization and urbanization on the preservation of TVLGS. Villages located at altitudes below 200 m are relatively scarce and are distributed in a sparse manner. Additionally, the presence of mountain ranges has facilitated the emergence of distinct, relatively autonomous economic circles within these TVLGS, resulting in the development of unique cultural and social characteristics.

River Systems

As a fundamental ecological substrate, hydrological resources exert deterministic influences on anthropogenic spatial organization through their dual roles in subsistence support and cultural landscape formation [41,42]. Through ArcGIS 10.8-based geospatial analysis, we characterized river networks and delineated TVLGS distributions using 1 km BZ (Figure 12a). The 1 km riparian buffer zones contained 692 villages (67.05%), demonstrating strong hydrocentric settlement preferences. Within the 0–5 km range, a statistically significant inverse relationship (R² = 0.89, p < 0.01) was found between village density and the distance from rivers. This can be attributed to three synergistic mechanisms:

• Subsistence optimization: Direct water access for irrigation and domestic use;
• Transport facilitation: Riverine navigation supporting regional connectivity;
• Microclimate regulation: Riparian thermal buffering and humidity moderation.

Notably, 39 villages (3.78%) in the eastern coastal plains exceeded 5 km from perennial streams, their water needs met through alternative strategies:

• Coastal settlements (n = 27): Seawater utilization via evaporation ponds and filtration systems;
• Upland communities (n = 12): Groundwater exploitation through stepped well architectures.

This distribution dichotomy reflects Fujian’s dual geocultural adaptation models. Mountainous interiors exhibit tight hydro-topographic coupling, with 83.2% of villages clustered within 2 km of third-order streams or higher. Conversely, eastern coastal zones demonstrate maritime-oriented adaptations, where Ming-Era–Qing-Era mercantile expansion (1567–1644 CE) drove settlement nucleation around harbors, supported by hybrid water management systems integrating wells (mean depth: 8.3 m), rainwater catchment, and tidal aquaculture (Figure 12b).

The spatial model emphasizes the ability of civilization to transcend absolute environmental determinism through technological innovation in the hydrological context, especially in coastal areas where the dissemination of marine culture has made alternative water resource acquisition paradigms possible.

Temperature and Rainfall

The climatic environment is crucial in site selection for TVLGS, profoundly affecting human activities and settlement patterns. The planning and development of communities must take into account climatic considerations and conform to local climatic characteristics to improve living circumstances. Data reveals that in Fujian Province, average annual precipitation demonstrates a gradient decline from west to east, with the majority of precipitation concentrated in the northern and southwestern areas (Figure 13a). Additionally, the average annual temperature gradually decreases from southeast to north (Figure 13b).

In Fujian Province, TVLGS are predominantly situated within a mean annual temperature range of 10–20 °C, as extremes in temperature can adversely affect agricultural productivity (Figure 13c). Notably, 447 villages, representing 83%of the total, are located in areas where mean annual rainfall is below 1.9 kmL (Figure 13d). Conversely, the number of TVLGS in regions experiencing extreme precipitation and excessively high temperatures is relatively low. This observation implies that inadequate precipitation and elevated temperatures can result in water scarcity and unfavorable climatic conditions, which may severely impact agricultural output. Furthermore, excessive precipitation poses an increased risk of natural disasters, including food insecurity.

Ecological Environment

Vegetation, acting as an active participant in and responder to environmental changes, exerts a notable influence on the energy balance in both the biosphere and the Earth–atmosphere system [43]. NDVI serves as the premier indication of vegetation cover and growth status, as well as a proficient metric for monitoring the ecological environment [44]. NDVI values are elevated in the central and western portions of Fujian Province, whilst diminished values are noted in the eastern coastal areas (Figure 14). The growth of TVLGS in the eastern coastal region is impeded by both a suboptimal ecological environment and a more developed economy. Conversely, the central region, characterized by elevated NDVI values, hosts a significant concentration of TVLGS. TVLGS are few and scattered in the western region, despite the great natural conditions. This is attributed to a shortage of labor resulting from a low population density, as well as a reduced land utilization rate due to the region’s topographical features.

3.2.2. Population and Economy

The formation and distribution of ancient villages exhibit spatiotemporal coupling with population size. As human settlements, these villages lose their defining significance without inhabitants [45]. This is evidenced by five major waves of southward migration from the Central Plains between the Eastern Jin Dynasty and the late Qing Dynasty, which coincided with the golden age of ancient village development [46] (

Table 2. Table of population trends in Fujian through the ages (unit: 10,000 people) and Fujian’s population share of China’s total.

Time Period	Population (10,000 People)	Proportion of Fujian’s Population to the National Population (%)
Western Jin Taikang (280–289)	/	About 0.3
6th Year of Yuansang, Western Han Dynasty (111 BCE)	40	/
Tianbao Era, Tang Dynasty (742–756)	51	About 1
1st Year of Yuanfeng in the Northern Song Dynasty (1078)	/	About 6.1
3rd Year of Yuanfeng, Northern Song Dynasty (1080 CE)	204	/
Southern Song Heyuan (1195)	/	About 4.2
16th Year of Jiading, Southern Song Dynasty (1223 CE)	323	/
1st Year (1277–1289)	294	About 4.9
26th Year of Hongwu, Ming Dynasty (1393 CE)	392	/
6th Year of Wanli in the Ming Dynasty (1578)	/	About 2.9
18th year of Guangxu of the Qing Dynasty (1661 CE)	146	/
32nd Year of Qianlong in the Qing Dynasty (1767)	/	About 3.9
23rd Year of Guangxu in the Qing Dynasty corresponds to 1897 CE	2683	/
13th Year of Guangxu in the Qing Dynasty (1887)	/	About 6.6
36th Year of the Republic of China (1947 CE)	1106	/
1949	/	About 2.2
1986	/	About 2.6
1987 CE	2801	/

). Historically, Fujian’s population size steadily increased over time, correspondingly supporting a growing number of ancient villages. Consequently, the cumulative number of ancient villages across different periods exhibits a highly significant positive correlation with population density.

Population and Ethnicity

Fujian Province, according to the Seventh National Census, was home to 41.88 million people in 2023, reaching a population density of roughly 349 people per square kilometer—more than double the national average of 148. This density map paints a clear picture: geography dictates demographics. The province’s population is largely clustered along the southeastern coast, leaving the mountainous northwest and the Nanping area sparsely populated. The challenging terrain, with its high altitudes, winding roads, and remoteness from the sea, has historically hindered economic development in these inland areas, leading to significant population decline. With a whopping 53 TVLGS boasting populations under 160,000, and 94 with over 800,000 residents, it is clear that many smaller communities are found in places like Nanping and Longyan (Figure 15a,d). Spatial analysis indicates that traditional village populations are dwindling in these hinterland regions. Scarce human and financial resources in these areas translate to underdeveloped infrastructure and a lagging economy.

Conversely, rapid modernization and economic expansion have resulted in fewer TVLGS in the more densely populated areas. The 527 TVLGS, with their populations ranging between 400,000 and 800,000, suggest that a certain critical mass is beneficial for the preservation and prosperity of these communities. However, excessive population pressure will lead to increased environmental pressure and ultimately destroy the traditional appearance and cultural characteristics of the village.

The spatial distribution of FTVLGS exhibits an inverse correlation with regional population density. Of the 1291 TVLGS, 42 fall below the province’s average population density, representing a mere 3.25% of the provincial average. Nineteen of these are even below the national average. Unsurprisingly, the more urbanized areas, such as Fuzhou, Quanzhou, Xiamen, Zhangzhou, the northeastern reaches of Ningde, and the Xinluo District of Longyan, have higher population densities. Modernization and urbanization, in many places, have taken a heavy toll on these indigenous communities, sometimes pushing them to the settlements’ outskirts. In the sparsely populated locales such as Nanping, Youxi County in Sanming, Liancheng County in Longyan, and Pingnan County in Ningde, the scattered settlements, smaller populations, and challenging living conditions are reflected in the dwindling number of TVLGS (Figure 15b,e). To address this complex situation, a multifaceted approach is needed. Villages in remote, sparsely populated areas, those with moderate density, and those bordering urban centers all require distinct preservation strategies. Each faces a unique set of environmental and developmental challenges.

According to the latest census figures, ethnic minorities constitute 8.9% of the Chinese population. As urbanization takes hold and socioeconomic prospects improve, it is natural that these groups tend to gravitate toward urban centers. Looking at the population distribution map for Fujian Province, it is clear that ethnic minorities are concentrated in the central and western parts of Sanming City, as well as in bustling areas like Xiamen, Quanzhou, the eastern reaches of Fuzhou, and the prosperous coastal zones of Ningde (Figure 15c,f). However, a concerning trend emerges: as the concentration of ethnic minorities increases, the number of FTVLGS decreases, suggesting these areas are not exactly conducive to the survival and prosperity of such villages.

Urbanization Rate

In addition to influencing the development and evolution of TVLGS, the spatial relationship between towns and TVLGS also influences the direction of TVLGS’ preservation and the reuse of cultural resources. Consequently, it is imperative to examine the attributes that distinguish TVLGS. In 2021, Fujian Province’s resident population’s urbanization rate was 69.7%, 5% higher than the national average. The “New Urbanization Plan for Fujian Province (2021–2035)” [47] states that by 2025, 71.5% of the province’s residents are expected to be urbanized. In regions of Fujian Province where the rate of urbanization is lower than the national average, there are 1045 TVLGS. Furthermore, 1114 TVLGS can be found in regions with lower rates of urbanization than the province as a whole (Figure 16). According to these statistics, TVLGS can be preserved and developed sustainably in county administrative regions due to their lower rate of urbanization.

Economic Development

As cities grow and people chase better economic prospects, preserving TVLGS is bound to become a real tug-of-war. This research digs into how GDP and GDP per capita play a role in where these villages are located in Fujian Province. Looking at Figure 17, statistical analysis reveals that areas with TVLGS tend to have a lower GDP intensity. In fact, the vast majority—a whopping 923 villages, or over two-thirds of the total—are found in areas where the GDP intensity is below 500. The way these villages are spread out across different administrative districts, each with its own level of economic development, really highlights the economic factor at play. For example, you will only find a couple of TVLGS in districts where the GRP per capita is on the lower side, say, around 51,200. This suggests that when the GRP per capita is down, so are the number of TVLGS, probably because tough economic times lead to folks moving away, making it hard for these villages to survive.

Transportation and Infrastructure

Highway and non-highway transportation routes serve not merely as functional corridors for point-to-point mobility, but they also function as critical “arteries” that underpin the operational dynamics of societies and facilitate the interaction and integration of diverse cultural systems [48]. Modern provincial highways, national highways, and expressways exhibit a high degree of spatial overlap with ancient post roads. This phenomenon indicates that the modern transportation system has inherited the functional role of historical roads, presenting strong spatial continuity between the ancient and modern transportation networks (Figure 18). Quantitative analysis of the spatial correlation between TVLGS and transportation facilities yields the following results.

For national highways, the number of TVLGS shows an overall decreasing trend with increasing distance from the highways; specifically, only 161 TVLGS are distributed within the 5 km buffer zone of national highways.

For provincial highways, 543 TVLGS (accounting for more than one-third of the total TVLGS) are located within the 5 km buffer zone (Figure 19).

For railway stations, no TVLGS are found within the 1 km buffer zone of the stations, while the number of TVLGS increases as the distance from the stations increases.

For expressway interchanges, the number of TVLGS within the 5 km buffer zone increases with distance from the interchanges; however, beyond the 10 km buffer zone, this number gradually decreases (Figure 20).

These findings align with TVLGS’ distribution along ancient post roads. Notably, an appropriate distance between TVLGS and modern transportation infrastructure supports protecting their original landscape, but also restricts tourism-centered economic growth to some extent.

Tourism Resource Intensity (TRI)

Tourism is a major economic driver that also boosts individual quality of life, quickly becoming one of the biggest industries worldwide [48]. Research on tourism’s influence on the layout of historic FTVLGS suggests these areas possess significant tourism development potential—a factor critical to both sustaining these settlements and preserving their unique heritage.

Per data published on 29 January 2024, Fujian Province had 494 A-class tourist attractions as of the end of 2023. Their point locations were extracted from Google Maps, and kernel density raster grids were generated via ArcGIS to quantify HCI around TVLGS. As shown in Figure 21, regions with thriving tourism sectors (e.g., Zhangzhou, Xiamen, and Putian) exhibit lower densities of TVLGS; conversely, areas with dense TVLGS only have moderate tourism development. In total, 1292 FTVLGS are situated in regions with moderate-to-high concentrations of tourism resources. This pattern indicates that mature tourism development may hinder traditional village preservation, while also implying substantial untapped potential within these villages.

According to official statistics from the Fujian Provincial Department of Culture and Tourism, the province had 494 A-grade cultural-tourism attractions as of December 2023. The geospatial coordinates of these attractions were georeferenced using Google Maps API v3.2 and subjected to KDE spatial interpolation via ArcGIS 10.8, generating a 100 m-resolution tourism resource intensity metric (Figure 21). Analytical results reveal a paradoxical inverse spatial correlation:

• High-intensity tourism zones (KDE > 1.54/km²) in Zhangzhou, Xiamen, and Putian exhibit depressed traditional village densities (12.7 villages/100 km², compared to the provincial average of 18.3 villages/100 km²).
• Moderate-intensity tourism clusters (KDE 0.82–1.54/km²) demonstrate optimal synergy, containing 86.3% of the total within their buffers—suggesting untapped culture–tourism complementarity.

This spatial disparity arises from two key mechanisms:

• Conservation pressures: Mature tourism development in coastal cities accelerates the fragmentation of vernacular landscapes.
• Resource underutilization: Inland villages retain more intact traditional architecture but lack essential tourism infrastructure.

3.2.3. History and Culture

Intangible Cultural Heritage (HCI) constitutes a critical vector of cultural memory, manifesting through performative traditions, artisanal craftsmanship, oral histories, and ecological knowledge systems that are intrinsically tied to place-specific material culture and historical landscapes [49]. As the custodians of 73% of China’s nationally recognized ICH elements, TVLGS serve as living repositories of these heritage assets, which are formally safeguarded through the National Cultural Relics Protection Unit framework encompassing architectural ensembles, archeological sites, funerary complexes, and lithic art sanctuaries.

Geospatial analysis of Fujian’s heritage landscape reveals 188 national-level HCI items and 182 nationally protected cultural relic units, with coordinates georeferenced using Baidu Maps API v3.0. KDE spatial interpolation (ArcGIS 10.8, Gaussian kernel, with a 500 m optimal bandwidth) quantified HCI across 1:50,000 topographic grids (Figure 22). The analysis demonstrates a significant negative spatial correlation (Pearson’s r = −0.67, p < 0.01) between HCI indices and traditional village density, particularly pronounced in coastal prefectures, where

• High-HCI zones (≥2.86 units/km²) contain merely 12.3% of villages;
• Moderate-HCI belts (1.24–2.85 units/km²) host 63.7% of settlements;
• Low-HCI areas (1.24 units/km²) account for 24.0% of villages.

The survey further indicates that Fujian’s coastal areas exhibit both stronger economic development and a higher density of historical culture than inland regions, hinting that once a certain economic bar is cleared, local governments tend to prioritize historical cultural preservation.

3.2.4. Analysis of Spatial Heterogeneity

In order to further verify the intensity, spatial correlation, and multi-factor interaction effect of a single factor on village distribution, this part uses random forest, spatial lag regression, and geographic detector models to conduct cross-analysis to clarify the core driving mechanism.

Random Forest Model

This research framework incorporates a comprehensive set of independent variables categorized across five thematic domains to systematically examine their influence on the spatial distribution of TVLGS. The selected predictors include the following (

Table 3. Table of 18 driving factors for FTVLGS’ spatial distribution.

Category	Factor Code	Factor Name
Geophysical parameters	X1	Village elevation
	X2	Mean regional slope
	X3	Euclidean distance to water systems
	X4	Mean annual temperature
	X5	Average annual precipitation
Demographic and socioeconomic indicators	X6	Total population
	X7	Population density
	X8	Ethnic minority population density
	X9	Urbanization rate
	X10	Aggregate regional GDP
	X11	pcGRP
Transportation infrastructure accessibility	X12	Proximity to provincial highways
	X13	Euclidean distance to national highways
	X14	Euclidean distance to railway stations
	X15	Proximity to highway service areas
Cultural resource endowments	X16	TRI
	X17	HCI
Ecological characteristics	X18	NDVI

):

The kernel density estimation of TVLGS’ distribution (Y) was operationalized as the dependent variable in our spatial analysis model (Equation (6)). This multidimensional approach enables a robust examination of environmental, anthropogenic, and infrastructural determinants, shaping traditional settlement patterns.

After normalizing the dataset, a random forest model was created using the Sklearn library in Python. During the model development, 80% of the data was set aside for training, with the remaining 20% reserved for testing. The model yielded an MSE of 0.000 and an R² score of 0.92, indicating both high accuracy and robust predictive power. Notably, within the random forest framework, the straight-line distance to the provincial highway was identified as the most influential variable, followed closely by annual precipitation and the intensity of historical and cultural influences (Figure 23).

The Spatial Lag Regression Model

To address the inherent spatial autocorrelation of predictor variables, a significance diagnosis for variables was conducted to account for potential biases in their significance. The analysis included 18 standardized variables (normalized via z-score transformation for dimensional comparability). Spatial regression was implemented in GeoDa, using Simple Linear Regression (Equation (7)) as the primary method.

The model exhibited strong explanatory power (R² = 0.8218), meaning 82.18% of the variance in traditional village kernel density was explained by the combined predictors. This high R² confirms the model effectively captures spatial interdependencies between geographical, socioeconomic, and infrastructural drivers of TVLGS distribution.

As shown in Figure 24, 12 predictors had statistically significant associations with TVLGS layout.

Positive correlations (7 factors): Per capita GRP, historical-cultural resource intensity, proximity to railway stations, distance from provincial highways, regional elevation, mean regional slope, aggregate population size.

Inverse correlations (5 factors): Aggregate regional GRP, mean annual precipitation, distance from national highways/waterways, population density.

Notably, per capita GRP had the largest standardized coefficient (β), indicating it was the most influential factor shaping Fujian’s TVLGS distribution. This highlights the key role of economic gradients and resource accessibility in mediating cultural heritage village preservation and clustering.

Non-significant associations, among some variables, reflect the heterogeneity of multivariate spatial interactions governing TVLGS distribution. This uncertainty in predictor–outcome relationships requires context-sensitive research within a refined framework to disentangle localized interactions between anthropogenic and environmental drivers under the study area’s distinct socioecological conditions.

Earth Probe Results

• Single-factor testing

The spatial heterogeneity determinants governing FTVLGS were quantitatively examined through geographic detector analysis (Equation (8)), with variables hierarchically ranked by their explanatory capacity (q-statistic) as follows: proximity to provincial highways > population density > annual precipitation > aggregate regional GDP > mean regional slope > pcGRP > urbanization rate > TRI > total population > HCI > Euclidean distance to railway stations > ethnic minority population density > mean annual temperature > Euclidean distance to water systems > village elevation > Euclidean distance to national highways > NDVI > proximity to highway service areas.

Notably, NDVI and Euclidean distance to national highways failed to achieve statistical significance thresholds, whereas the remaining 16 variables demonstrated robust significance in shaping spatial patterns. The model identified four primary determinants with superior explanatory power: (1) provincial highway proximity; (2) aggregate regional GDP; (3) annual precipitation; (4) population density. Conversely, elevation, slope, NDVI, and selected infrastructural measures exhibited comparatively limited explanatory efficacy within the current analytical framework (Figure 25a). These findings underscore the predominance of anthropogenic–economic forces over biophysical factors in mediating the spatial organization of cultural heritage landscapes.

2.

Interaction detection

To evaluate the spatial differentiation capacity of key factors (and their logical relationships), GeoDa’s interaction detector analyzed 19 factors. The goal was to determine if two interacting factors influenced Fujian TVLGS’ (FTVLGS’) distribution independently or collaboratively by examining their combined effects.

Key interaction results (Figure 25b):

Top explanatory interaction: Mean annual precipitation + Euclidean distance to provincial highways, showing the highest explanatory power and a strong non-linear enhancement effect. This aligns with Section (“Precipitation affects agricultural production”) and Section (“Provincial roads as core transport corridors”), confirming natural–human factor synergy.

Other influential pairs: (1) Provincial highway proximity + per capita GRP; (2) provincial highway accessibility + ethnic minority population density; (3) highway distance + regional per capita economic output.

These findings confirm that Fujian’s TVLGS’ spatial organization stems from multidimensional synergies between demographic, economic, cultural, and biophysical determinants. Critical observations are as follows.

All interactions show non-linear enhancement, with joint explanatory power far exceeding individual factors.

Multifactorial dynamics rely on complex coupling (not simple addition), with transportation infrastructure proximity mediating socioeconomic and environmental variables’ spatial impacts.

This non-linear framework governs cultural heritage village distribution heterogeneity, emphasizing the need for integrated spatial governance that accounts for cross-dimensional factor interactions.

4. Discussion

The geographical layout and historical development of FTVLGS are shaped by the complex interactions of ecological, economic, demographic, and cultural factors. A comprehensive investigation into the spatiotemporal evolution of these settlements is of crucial significance for pushing research forward in this field. Identifying the core drivers of their formation is essential to understanding the intricate links between village layout, architectural traditions, and community adaptation—knowledge indispensable for safeguarding cultural heritage amid rapid urbanization and for integrating culture and tourism to drive high-quality tourism development.

4.1. Core Findings and Theoretical Dialog with Previous Studies

This study identifies distinct multi-core agglomeration patterns in FTVLGS, with three primary clusters in northern, eastern, and southwestern Fujian. This finding both aligns with and differs from regional studies elsewhere. Liu et al. noted a “strip-shaped” distribution of TVLGS in Henan Province along the Yellow River basin, driven by agricultural water resources. In contrast, Fujian’s multi-core pattern stems from its unique mountain–river system and historical migration routes. Xia et al. emphasized ecological land and thermal environments as key to village distribution in the central Shanxi urban agglomeration; while this study acknowledges ecological influences, it further highlights the interactive effects of transportation accessibility and precipitation—factors underemphasized in the Shanxi research.

A key insight is that the interaction between “straight-line distance to provincial highways” and “average annual precipitation” is the most influential driver of village distribution, explainable via a multi-mechanism framework. Historically, Fujian’s mountainous terrain (over 90% hills/mountains) made provincial highways—often overlapping with ancient post roads—the primary arteries for trade and cultural exchange. Villages near these highways gained advantages in accessing agricultural product markets (e.g., tea and fruits) and importing essential goods, fostering their formation and prosperity. Additionally, Fujian’s subtropical climate (1400–2000 mm annual precipitation) links precipitation directly to agricultural productivity and water availability; however, excessive rainfall increases erosion and flood risks in mountainous areas. Thus, villages near provincial highways could access disaster relief more easily while benefiting from moderate precipitation, creating an optimal “accessibility-water balance” for settlement.

By comparison, Bi et al. identified river systems as dominant in Jiangnan’s village distribution, where flat terrain and dense water networks prioritize water transportation. In Fujian, fragmented topography limited river navigation, elevating provincial highways’ importance. This contrast underscores the need for region-specific driver analyses, as the relative significance of factors like transportation and hydrology varies with geographical context.

4.2. Universal Applicability and International Promotion Pathways

While this study’s methodology is applicable to other regions, Fujian’s specific cultural, historical, and topographic context requires adaptation when transferring findings internationally. For instance, in Southeast Asian mountainous regions (e.g., northern Vietnam and the Indonesian archipelago) with similar topography to Fujian, core drivers (transportation accessibility and climate) may be comparable—but the analytical framework needs adjustment. Rice cultivation dominates these areas, so paddy field distribution and irrigation system proximity should be added to the driving factor system, alongside highway distance and precipitation.

In contrast, European rural areas (e.g., Swiss/Austrian Alpine regions) with established heritage policies and tourism-driven economies may see stronger TRI-HCI interactions than Fujian. Here, the random forest model could be modified to weight cultural/tourism variables more heavily or add indicators (e.g., UNESCO heritage site proximity and tourist flow density) to enhance adaptability.

For arid/semi-arid regions (e.g., American Southwest, parts of Australia), water availability (e.g., groundwater depth and oasis distribution) replaces precipitation as the key climatic driver. “Distance to provincial highways” could be replaced with “distance to major water-transportation routes” (e.g., Arizona canals) to reflect unique resource constraints and transportation patterns.

4.3. Policy Implications, Implementation Challenges, and Success Measures

This study’s findings provide targeted policies for traditional village conservation and development, but implementation requires addressing region-specific challenges and leveraging successful practices.

4.3.1. Tiered Conservation Zoning and Implementation Pathways

The proposed tiered system (Priority Conservation Zones, Key Protective Precincts, and General Preservation Sectors) aligns with ICOMOS’ “cultural landscape protection” policies. Priority should be given to delineating “core protection units” along provincial highways in areas with 1400–1800 mm annual precipitation (e.g., high-density clusters in Ningde and Nanping), alongside improving flood control facilities and restoring traditional post roads. This approach protects villages and preserves the original “nature-humanities” synergy, grounding policies in empirical findings to avoid overgeneralization.

In Priority Conservation Zones, urgent measures (e.g., restricting new construction and restoring traditional architecture) are critical. A core challenge is balancing conservation and residents’ livelihoods, which can be mitigated by implementing traditional housing renovation subsidies (e.g., 50% cost coverage for Hakka villages in Longyan)

In Key Protective Precincts, controlled development should follow “adaptive reuse” principles (e.g., converting abandoned traditional houses to museums/homestays). Ensuring conversion authenticity is a challenge—mitigated via certification systems (e.g., “Traditional Village Renovation Authenticity Standards”) and community involvement in design. Ningde’s She ethnic minority villages demonstrated success here: community-led homestays raised tourist satisfaction by 30% while maintaining cultural integrity.

4.3.2. Coordinated Regional Development and Tourism Integration

The proposed polycentric governance model (optimizing tourism resources across administrative boundaries) can be supported by inter-municipal cooperation—e.g., establishing a “Fujian Traditional Village Tourism Alliance” to coordinate marketing, share tourist flows, and standardize services. In villages along provincial highways with medium per capita GRP (e.g., Longyan Hakka villages), integrating “traditional skills + short-distance tourism” (leveraging highway access to attract tourists while controlling development intensity) has already balanced protection and development.

Conflicting interests between administrative regions are a challenge—addressed via provincial incentives (e.g., additional tourism funds based on inter-regional cooperation levels). This approach succeeded in China’s Yangtze River Delta, where cross-provincial tourism alliances increased regional arrivals by 25%.

4.3.3. International Policy Transfer and Local Adaptation

For international policymakers, this study’s focus on integrating biophysical conservation, socio-cultural maintenance, and low-carbon development provides a holistic rural heritage framework. In resource-constrained developing countries (e.g., parts of Africa and Southeast Asia), Fujian’s “low-cost, community-led” strategies can be adapted—though weak institutional capacity and funding gaps must be addressed. International organizations could provide technical assistance and small grants; for example, a UNESCO-funded program in rural Nepal, modeled on Fujian’s community-led approach, reduced traditional house demolition by 40%.

4.4. Limitations and Future Research Directions

This study has three key limitations. First, it does not fully explore climate change’s long-term impacts on village distribution—incorporating climate projection data would improve model predictability. Second, some explanatory variables were quantified based on period-averaged data, which cannot fully capture the non-simultaneous timing of factor changes—for example, the Song Dynasty’s post-road expansion preceded population agglomeration by approximately 50 years, potentially leading to slight deviations in the temporal matching of variables. Third, due to the lack of systematic historical data, we omitted some potential driving variables, such as dynastic policies, large-scale wars, and ancient urban planning systems. These unincorporated factors may have indirectly influenced traditional village distribution and thus slightly weakened the explanatory power of the current model. Future research will integrate local gazetteers and archeological reports to supplement these variables and use event history analysis to address the temporal asynchrony of factors.

5. Conclusions

Current research on traditional village distribution often adopts fragmented approaches, focusing on single factors at limited scales. To address this, this study develops an integrated analytical framework combining GIS spatial modeling and Python-enabled statistical methods, forming a systematic protocol for disentangling complex human–environment interactions in cultural landscapes—progressing from individual factor assessment to interaction mechanism identification.

FTVLGS, with origins dating to the pre-Tang Dynasty, exhibit distinct spatiotemporal evolution: initial northern settlements formed a northeast–southwest axis, later expanding eastward to the coasts and outward from the inland highlands. Dynastic transitions drove three core changes: a southeastward shift in the distribution centroid, expanded distribution area, and strengthened directional expansion, reflecting societal adaptation to environmental, economic, and cultural shifts.

Geospatially, these villages cluster in three high-density cores, predominantly in elevated terrains with steep slopes, moderate temperatures, and precipitation—conditions optimal for traditional agriculture.

Through evaluating 18 variables via three methodologies, the interaction between “average annual precipitation” and “straight-line distance to provincial highways” emerges as the dominant driver, with the interaction of provincial highway accessibility and per capita GRP as a secondary driver. This highlights the non-linear, interdependent nature of factors shaping village distribution.

Theoretically, this study enriches rural settlement geography by providing a multi-factor interaction framework and region-specific insights, contrasting with other areas. Practically, it offers adaptable policy tools for traditional village protection and rural revitalization in Fujian and beyond. Future research should integrate micro-level data, climate change impacts, and longitudinal studies to deepen the understanding of village dynamics.