Reconstructing the Silk Road Network: Insights from Spatiotemporal Patterning of UNESCO World Heritage Sites

Abstract

Building on the observation of gaps in current research, this study provides a comprehensive analysis of the spatial patterns of heritage sites along the Silk Road, focusing on how historical trade routes shaped what are now recognized as heritage sites. Using data from the United Nations Educational, Scientific and Cultural Organization (UNESCO) World Heritage List, the research examines heritage sites across Eurasia and North Africa, with a specific emphasis on the Silk Road corridors. This study employs a spatiotemporal approach, categorizing sites into northern overland routes and southern maritime routes to highlight regional variations in network development. The key findings of this study reveal the significant influence of historical trade routes on the development of settlements, cities, and cultural landmarks along the Silk Road. These findings identify clear trends in the Silk Road network’s evolution over time, illustrating a shift in its spatial focus across different historical periods. Initially, the network was centered in the eastern Mediterranean during the Classical Period. In the medieval period, this focus expanded to include a dual core area in both the eastern Mediterranean and Central Asia. By the late Medieval period, the network had shifted again, with a new core emerging in Europe. This chronological and spatial analysis allows for a detailed examination of the Silk Road network’s heritage landscape evolution. The study underscores the interconnectedness of heritage sites across these regions, contributing to a deeper understanding of how landscape connectivity and trade network dynamics evolved over time. Furthermore, by identifying patterns of network development and shifts in centrality and density, this research offers valuable insights for the conservation and management of heritage landscapes. These findings are particularly crucial for preserving the historical and cultural integrity of Silk Road heritage sites.

1. Introduction

Systematic efforts are required for precise a definition and understanding of the vicissitudes of the Silk Road for researchers on this concept. The earliest definition of the Silk Road comes from the German geographer Ferdinand von Richthofen in 1877, although evidence shows that the concept was already emerging in the works of the German geographer Carl Ritter as early as 1838. It was first employed to refer to the ancient routes of transportation and trade between the three continents of the Eastern Hemisphere, particularly between Europe and Asia, and from the beginning, it included both the overland and maritime Silk Roads [1]. Richthofen’s description of the Silk Road marks the establishment of official interactions between China and the West in the 2nd century B.C.E., and his research made the concept widely known. Subsequent scholars further refined the description of the Silk Road, defining its eastern starting point in the middle and lower reaches of the Yellow River in China, while the western starting point extended from the eastern Mediterranean to Italy. They also proposed various perspectives on the trade patterns and historical significance of the Silk Road [2,3,4,5,6,7,8].

Based on the intuitive impressions formed by existing research on the Silk Road, the official description provided by the United Nations Educational, Scientific and Cultural Organization (UNESCO) offers a reference for the definition in this study. Here, the Silk Roads (sic) are described as “a diversity of routes and cargos” and “routes of dialogue”, acknowledging them as not only a trade network but also a network of culture, knowledge, and religion [9,10]. This provides a basis for defining the Silk Road beyond just trade. The Silk Road network (hereafter SRN) can be defined as the total network of routes connecting most cultural areas of the Old World, including the continents of Africa, Asia, and Europe. Precisely speaking, after the historically attested opening of the Silk Road by General Zhang Qian in the latter half of the 2nd century B.C.E., the SRN began to function as a system of contact, communication, and exchange across the three continents [11]. This network remained systematically integrated and consistent until Christopher Columbus’s first navigation to the Americas in 1492 [12]. Columbus’s navigation to the Americas initiated a new open system of contact among the five continents, and subsequent navigation of the Oceanic islands extended the reach of European culture into the Americas and Oceania. Since then, the route network has no longer been enclosed, thus incorporating a large portion of the global population and resources into the global route network (hereafter GRN) of communication and exchange. The marked difference in the range between the SRN and the GRN leads us to conclude that the study of the SRN as a synchronic system must end in the 15th century, as diachronically, this system evolved into a completely different GRN.

In fact, studies of the SRN based on documentary sources can be so precise that most of the connecting routes between major cities can be illustrated on maps. However, such studies face several issues. Firstly, the recorded distances in historical texts are not always accurate. The distance measurement units used by ancient civilizations often differ significantly from today’s precise units, and historical texts do not provide conversions between these ancient and modern units [13]. Furthermore, the locations mentioned in historical texts may differ from those of the same name today, due to differences in site selection [14]. Additionally, historical texts may mention many specific routes of interaction, but the extent to which these routes were used varies greatly [15]. Some routes were rarely used but received greater attention in SRN research due to their reference in texts [16]. In the absence of precise statistical data on ancient city populations and trade volumes, the weights of various routes and relevant nodes can only be estimated based on experience [17].

The challenges associated with documentary-based studies underscore the need for alternative methods to reflect the SRN more accurately. Recognizing the importance of comprehensive approaches to understanding the SRN, international organizations have taken significant steps to address these challenges. Recognizing the importance of diverse cultural exchanges along the Silk Roads, UNESCO in 1994 launched the “Global Strategy for a Representative, Balanced and Credible World Heritage List” [18]. This initiative aimed to ensure that the world’s cultural and natural diversity was fully represented. Central Asia, identified as one of the most under-represented regions, became a focal point for efforts to fill this gap in global heritage representation. UNESCO had earlier acknowledged the significance of the Silk Roads for intercultural exchange and the wealth of cultural heritage sites it encompassed. Consequently, UNESCO proposed a trans-national nomination project for the Silk Roads as part of their strategy, fostering international collaboration and addressing the lack of representation on the World Heritage List, such as the UNESCO/Japanese Funds-in-Trust project: Support for Silk Roads World Heritage Sites in Central Asia (Phase II) [19].

Given the existing issues in estimating the SRN based on documentary sources and these international efforts, a supplementary method is proposed: using the heritage site databases of the three continents to estimate and reconstruct SRN routes. The most concentrated and comprehensive information on World Heritage Sites (hereafter, WHSs) comes from the UNESCO’s World Heritage List. The downloadable WHS database provides the latitude and longitude information of heritage sites, and the UNESCO WHS List website provides detailed information on the establishment and abandonment dates of these sites [20,21,22]. By logging these heritage sites, this study has established a geographic information database reflecting the emergence, prosperity, and abandonment of heritage sites over different periods [23].

Using the geographic information database of heritage sites to reflect the evolution of the SRN is based on the following assumptions:

• The SRN, the system of interactions among Eurasia and Africa, is an enclosed and stable system.
• Heritage sites that once existed and prospered can reflect the prosperity of the aforementioned interaction system.
• The network density between heritage sites reflects the network density of the SRN.

Using heritage sites to reflect the SRN offers greater precision compared to documentary sources, as the geographic locations of the archeological sites and cultural landscapes represented by these heritage sites are accurate and do not require textual verification. The distances between heritage sites are precise. The issues of distance, location, and unit conversion that reduce precision in the textual verification of documentary sources do not arise when using heritage sites to reflect the SRN. Due to the lack of exact route markings, only straight-line measurements can be calculated. Thus, this study adopts spatial clustering methods.

Based on this assumption, this study first introduces the research methods and then demonstrate how heritage sites reflect the SRN. The network relationships between heritage sites indicate the network density of the SRN. The resulting network will exhibit varying levels of route usage. In areas with high network density between heritage sites, the density will be significantly higher than in areas with low density, forming a heterogeneous network in terms of usage levels. This situation contrasts with the uniform recording of multiple routes based on textual verification of documentary sources. Both methods have distinct focuses: faithfully maintaining textual records maximizes the representation of all possible routes in the SRN. All possible routes are uniformly recorded based on historical records. Routes formed between heritage sites give higher weight to those with higher network density, accurately reflecting the weights of various network parts and nodes, though influenced by existing heritage site registrations. This study integrates both methods to construct an evolution map of the SRN based on documented routes and heritage site weights, offering a comprehensive research framework. Finally, the study analyzes and discusses related factors influencing the changes in the SRN. Based on these, the study provides recommendations for the future preservation of Silk Road cultural heritage.

2. Materials and Methods

This study analyses the distribution of the UNESCO World Heritage Site Database heritage sites along the SRN. As a foundational step, our methodological framework, depicted in Figure 1, outlines the comprehensive approach taken from data collection through to spatial analysis. More specific, to achieve the aim and objectives of this research, which involves using heritage site databases from three continents to estimate and reconstruct SRN routes, three primary analytical methods are employed: spatial distribution analysis, proximity analysis, and Kernel Density Estimation. Firstly, spatial distribution analysis integrates multiple methods to examine the geographical patterns and spatial structures of heritage sites along the Silk Road. Spatial cluster analysis is employed to detect clusters of heritage sites, revealing areas of significant cultural and economic activity. This analysis is complemented by the chi-square test, which assesses the distribution of heritage sites along the northern and southern routes of the SRN, identifying periods of significant transformation in regional connectivity and land use. Additionally, the study utilizes spatial autocorrelation functions (Ripley’s K, L, and G) to quantify the degree of clustering or dispersion within the distribution of heritage sites. Ripley’s K function is particularly useful for identifying the extent of clustering compared to a random distribution, while Ripley’s L and G functions offer more detailed insights into the spatial processes influencing site distribution. Secondly, proximity analysis, primarily through the nearest neighbor distance (NND) method, provides critical insights into the spatial relationships between heritage sites over time. By calculating the distance between each site and its nearest neighbor, this method evaluates the concentration and dispersion of heritage sites along the SRN. The study also incorporates the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, which identifies clusters based on density and effectively manages noise and outliers. DBSCAN is particularly valuable in archeological research because it does not require a predetermined number of clusters, allowing for the identification of geographically significant differences between sites. Moreover, the silhouette coefficient is calculated to assess the quality of clusters identified by DBSCAN, providing a measure of cohesion within clusters and separation between different clusters. Finally, Kernel Density Estimation (KDE) is used to analyze the spatial distribution of heritage sites across various centuries, providing both statistical and visual representations of site density. KDE smooths spatial data to produce a continuous surface that highlights areas of high cultural and economic activity along the Silk Road. This method effectively demonstrates how the distribution of heritage sites evolved over time, reflecting changes in settlement patterns and land use. The KDE analysis is complemented by distance matrix construction and proximity analysis, which further explore the spatial relationships between sites. The results from KDE, combined with these other methods, provide a detailed understanding of the spatial dynamics and the role of the SRN in shaping connectivity and landscape evolution across Eurasia and North Africa.

2.1. Material Collection

The cultural heritage data used in this study were sourced from the UNESCO World Heritage site (www.unesco.org) official platform of the United Nations Educational, Scientific and Cultural Organization. The data and information provided are sourced from UNESCO and its partner organizations, and are subject to authoritative verification and approval, exerting a significant influence in the field of cultural heritage. Details such as the name of the heritage site, the country it is in, its coordinates, the region, and the dates of establishment and abandonment were obtained by accessing the descriptive pages of cultural heritage entries across Asia, Europe, and North Africa. The names of the heritage sites are derived from the entries themselves, while the country and coordinates are from the basic information section, and the region along with the dates are from the descriptive section of the entries.

The data are categorized chronologically, sorted by the date of establishment, and recorded in centuries from the time of their establishment to their abandonment. The temporal scope for the cultural heritage sites in this study is designated from the 2nd century B.C.E. to the 15th century. This timeframe was specifically chosen to align with research on the SRN, beginning with the historical mission of Zhang Qian to the Western Regions. Zhang Qian’s journey not only opened up routes to these regions but also marked the beginning of cultural and commercial exchanges between the East and the West, as detailed in the Records of the Grand Historian: Treatise on the Dayuan [24]. Over time, the prominence of the terrestrial Silk Road gradually diminished, supplanted by the maritime periods, particularly by the 15th century—a period notable for the Age of Discovery in Europe. During this era, European explorers, emboldened by advancements in Portuguese navigational techniques that refined traditional Mediterranean maritime skills to meet the demands of oceanic voyages, ventured into the New World, establishing new sea routes. This period is marked by the navigational achievements of explorers, most notably Christopher Columbus, whose voyages from 1492 to 1493 were meticulously planned and executed based on his exceptional navigational skills as documented in his logs [25], with the achievement of the discovery of maritime routes to the Americas. Italian navigators also played a significant role in these maritime explorations, though it was undoubtedly the golden age of Iberian power that first propelled European maritime expansion [26]. Concurrently, the overland Silk Road began to decline as these new sea routes emerged, allowing Europe to trade directly with Asia and the Americas via the newly discovered maritime routes, thereby reducing reliance on the traditional Silk Road, the nexus of the Old World [27]. The selected period from the 2nd century B.C.E. to the 15th century effectively reflects the changes in the Silk Road’s cultural heritage sites over its historical phases, thereby providing robust support for further analysis. Once the timeframe was established, the data were then systematically categorized by century.

In terms of data preprocessing for this study, cultural heritage information was entered in the sequence of heritage name, country, latitude, longitude, region, time of establishment, and time of abandonment. During the selection process, entries without specific establishment dates were removed, and smaller sub-entries under major entries were included to better highlight the Silk Road’s evolutionary process. For data transformation, the original geographical coordinates, originally in degrees, minutes, and seconds, were converted to decimal degrees format. This conversion facilitates the eventual integration of the data into illustrations produced by JupyterLab platform of Python (3.11.5) language and Rstudio platform of R (4.3.3) language. For heritage sites that remain to this day without a recorded date of abandonment, no abandonment time was entered; however, these sites are included in the dataset up to the 15th century, placed within the time frame from the 2nd century B.C.E. to the 15th century.

2.2. Research Methods

2.2.1. Spatial Distribution Analysis

In pursuit of clarifying the spatial clusters of SRN sites through 2nd century B.C.E. to the 15th century mentioned in the previous section, this study advances into the second pivotal phase: spatial cluster analysis. This analysis aims to unearth the underlying spatial structures and tendencies inherent within the geographic distribution of the sites. To achieve this objective, this study embarks on an application of spatial statistical techniques. To investigate the relationship between the establishment centuries of SRN sites and their geographical coordinates, this study implements the chi-square test, a statistical assay for its efficacy in discerning and comparing the distributions of categorical variables across distinct groups. The chi-square test is defined by the following formula:

$${\chi }^2=\sum {\displaystyle \frac{{\left(O_i-E_i\right)}^2}{E_i}}$$

(1)

where $O_i$ represents the observed frequency of heritage sites in each period, and $E_i$ is the expected frequency assuming no change over periods. This test enables us to quantitatively determine the homogeneity or heterogeneity in the geographical spread of sites through time. Meanwhile, this test serves to ascertain whether the distribution of heritage sites exhibits significant variation across several centuries, thereby permitting a quantified assessment of distributional homogeneity or disparity [28].

Complementing the chi-square analysis, this study invokes the use of spatial autocorrelation functions—namely Ripley’s K, L, and G. These functions are instrumental in dissecting the spatial clustering of sites, providing a spectrum of scales at which any agglomeration might manifest. Ripley’s K function is particularly adept at identifying the degree of clustering within a given radius compared to a random distribution. It is computed as

$$K\left(t\right)={\displaystyle \frac{1}{\lambda }}\sum _{i\ne j}I\left(d_{ij}\le t\right)$$

(2)

where λ is the intensity of points (number of points per unit area), d_ij is the distance between points i and j, and I is an indicator function that equals 1 if d_ij ≤ t and 0 otherwise. This function allows us to quantify the extent to which heritage sites are more clustered than would be expected in a random distribution.

In extension, Ripley’s L function, a transformation of the K function, affords an expected distance metric that is intuitively more interpretable, delineating clustering or dispersion phenomena in relation to complete spatial randomness. It can be used as

$$L\left(t\right)=\sqrt{{\displaystyle \frac{K\left(t\right)}{\pi }}}-t$$

(3)

The G function further enhances research analysis by offering an empirical distribution function of the distances from a randomly selected site to its nearest neighbor, thereby allowing an evaluation of the spatial processes influencing the point patterns [29,30]. It is given by

$$G\left(t\right)={\displaystyle \frac{1}{n}}\sum _{i=1}^n I\left(d_{i,\mathrm{n}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{t}}\le t\right)$$

(4)

where d_i,nearest is the distance from point i to its nearest neighbor. This metric provides insights into the intensity of clustering by looking at the proximity of each site to its closest neighbor.

2.2.2. Proximity Analysis

In the Proximity Analysis section, this study constructs a distance matrix for various sites to assess their spatial relationships and interactions over different historical periods. The nearest neighbor distance (NND) method, essential for evaluating the concentration of archeological sites, is widely employed in the statistical analysis of archeological and historical studies. This method calculates the distance between a site and its nearest neighbor to deduce the spatial distribution characteristics of these sites.

In applying this method to this research, the first step involves identifying archeological sites that existed from the 2nd century B.C.E through the 15th century. Subsequently, the NNDs for each site are calculated. Assuming each site is denoted as $i$, the distance d_i to its closest heritage site is computed using the following formula:

$$d_i=\underset{j\ne i}{min} d\left(s_i,s_j\right)$$

(5)

In this analysis, $d(s_i,s_j)$ represents the distance between heritage site $s_i$ and heritage site $s_j$, where $j\ne i$ to ensure that the site does not measure the distance to itself.

Subsequently, this study calculates the average nearest neighbor distance for all heritage sites across various centuries. The formula is as follows:

$$\overline{d}={\displaystyle \frac{1}{n}}\sum _{i=1}^n d_i$$

(6)

$n$ represents the total number of heritage sites within a given century, while $d_i$ refers to the nearest neighbor distance for the $i$th heritage site.

The spatial distribution characteristics across different historical periods can be revealed by calculating the average NND coefficient for each century. This approach enhances the understanding of the evolving spatial patterns over time and provides a comprehensive perspective on the spatial layout trends of each period. This exploration of both relative distances and statistical measures of spatial distribution enables a thorough understanding of how cultural and historical influences may have shaped the placement and significance of heritage sites.

This study utilized DBSCAN (Density-Based Spatial Clustering of Applications with Noise), a popular clustering algorithm for density-based grouping of points that can identify clusters of arbitrary shapes while managing noise and outliers [31]. DBSCAN’s primary parameters are ‘eps’, which defines the neighborhood radius, and ‘minPts’, which specifies the minimum number of neighboring points required for a point to be considered a core point.

The analysis commenced with the computation of the k-distance for each heritage site. This k-distance, defined as the Euclidean distance to a site’s k^th nearest neighbor, is crucial for ‘eps’ for the DBSCAN algorithm. The systematic calculation of these distances for each site within the dataset is performed as follows:

$$k-\mathrm{distance}\left(p\right)=\left|\left|p-N_k\left(p\right)\right|\right|$$

(7)

where $p$ denotes an individual heritage site, $N_k\left(p\right)$ signifies its k^th nearest neighbor, and $\left|\left|p-N_k\left(p\right)\right|\right|$ represents the Euclidean distance between them. The selection of a k-value, often predicated on the dataset’s intrinsic characteristics and the clustering objectives, leads to a collection of k-distances which, when ordered and visualized, elucidate a discernible ‘elbow’ or inflection point in their progression.

This study computed the k-distance to facilitate DBSCAN clustering. Unlike other clustering algorithms such as K-means, DBSCAN does not rely on a predetermined number of clusters but instead bases clustering on the density and proximity of points. This makes DBSCAN particularly sensitive to geographic differences between sites, a feature that is especially valuable in archeological research [31].

Subsequently, this study calculates the silhouette coefficient, a metric used to assess the quality of clustering. This coefficient integrates both cohesion—measured as the average distance of a point to others within the same cluster—and separation, or the average distance to points in the nearest different cluster. Each point is assigned a score ranging from −1 to 1. A silhouette coefficient close to 1 indicates well-defined clustering; a coefficient near −1 suggests poor clustering quality; and a score approaching 0 may indicate overlapping or randomly distributed clusters.

For each heritage site as one point $i$, the calculation of the silhouette coefficient includes two crucial distances: $a\left(i\right)$, the mean distance to other points within the same cluster, reflecting the cluster’s cohesion, and $b\left(i\right)$, the mean distance to points in the nearest different cluster, indicating the degree of separation. These distances are essential for determining the silhouette coefficient for each point, calculated as follows:

$$s\left(i\right)={\displaystyle \frac{b\left(i\right)-a\left(i\right)}{\max \left\{a\left(i\right),b\left(i\right)\right\}}}$$

(8)

After calculating the silhouette values for each point, it is necessary to compute the average values for each century. This allows the degree of clustering for different centuries to be presented and determines whether the results are similar across these periods.

2.2.3. Kernel Density Estimation

This study conducts a Kernel Density Estimation (KDE) to analyze the distribution of archeological sites across various centuries. KDE is a method extensively used in archeological research to examine spatial point distributions. Mathematically, KDE can be represented as follows:

$$\widehat{f}\left(x,y\right)={\displaystyle \frac{1}{n\cdot h^2}}\sum _{i=1}^n K\left({\displaystyle \frac{x-x_i}{h}},{\displaystyle \frac{y-y_i}{h}}\right)$$

(9)

$\left(x,y\right)$ represents the coordinates of the point on the evaluation grid. ($x_i$, $y_i$) represents the longitude and latitude coordinates of the $i$-th data point. n is the number of site data points. ℎ is the bandwidth. K is the Gaussian kernel function.

KDE could assess the density of heritage sites along the Silk Road, demonstrating KDE’s recognized effectiveness in spatial analysis. This study utilized the ‘scipy.stats’ module in Python 3.11.5, which automatically selects the bandwidth $h$ and applies kernel smoothing to the data points, generating a smoothed density estimate across the spatial region. Visualizing these density estimates on a map provides an intuitive understanding of the distribution density of heritage sites for each century.

3. Results

This section applies the above method and divides into three sections to show the findings. Before delving into the specific analysis, a statistical examination of the WHS data was conducted. Figure 2a presents the ranking of the top 20 countries or regions by the number of heritage sites. China ranks first with the most heritage sites, followed by Italy and Spain. Germany, France, and India also rank prominently. The number of heritage sites in these countries underscores their importance in global cultural heritage preservation. This figure indicates that Asian and European countries dominate in heritage site numbers, reflecting their rich historical and cultural legacy. Figure 2b illustrates changes in the number of heritage sites over various periods, with a polynomial trend line highlighting the overall trend. The horizontal axis represents centuries, from the 2nd century BCE to the 15th century CE, while the vertical axis shows the number of heritage sites. The number of heritage sites has gradually increased over time. The polynomial trend line further shows that this growth has become more pronounced in recent centuries, particularly after the 10th century CE. This suggests a continued increase in the number of cultural heritage sites over time.

3.1. Statistical Analysis

The chi-square test was employed to examine the association between the distribution of heritage sites along the terrestrial and maritime Silk Roads and their respective historical periods (shown in Figure 3). Using the average latitude of 33.4253° N, derived from cities on the UNESCO interactive map of the Silk Road, the SRN sites were divided into northern and southern regions. A contingency table was constructed to tally the number of sites across different centuries within these regions. The chi-square test yielded a value of 26.9515 with 16 degrees of freedom and a p-value of 0.04203. Since the p-value is below the significance level of 0.05, the results indicate a statistically significant association between the location of sites and their temporal classification. This suggests a notable, non-random pattern in the geographical placement of sites corresponding to their historical periods.

This does not, however, imply a causal relationship between the two variables. The following figures illustrate the spatial point patterns of a dataset as analyzed by Ripley’s K, L, and G functions within the context of their respective significance bounds. The analysis using Ripley’s K, L, and G functions reveals a clear pattern of non-randomness in the spatial distribution of the points studied. The consistent positioning of the observed functions above the theoretical lines and outside the significance envelopes for all three tests indicates a significant tendency towards clustering rather than randomness or regularity. These results should be interpreted as evidence of an underlying spatial process or interaction influencing the distribution of the points, which could be crucial for further research into the causative factors or mechanisms responsible for this pattern.

Figure 4 shows that the observed K function, represented by the solid black line, lies above the theoretical K function, indicated by the red dashed line, across the range of distances. This consistently higher value outside the grey confidence envelope suggests a significant clustering of points over the distance scale analyzed. This finding indicates that points are more densely located around each other than would be expected if they were randomly distributed, pointing to a potential underlying process that causes aggregation.

In Figure 5, the observed L-function, shown by the solid black line, is above the theoretical expectation (red dashed line) for all distances measured. Since the L-function is designed to have an expected value of zero under a random distribution, the observed positive values suggest clustering. The deviation of the L-function from zero beyond the grey significance bounds reinforces the evidence of spatial clustering at various scales.

Figure 6 depicts the G function. Here, the solid black line of the observed G function rises above the theoretical line (red dashed line), particularly at smaller distances, and remains outside the grey significance envelope. This pattern suggests that the nearest neighbor distances are generally shorter than expected under a random distribution, indicating a tendency toward clustering.

3.2. Spatial Clustering Analysis across Centuries

From the 2nd century B.C.E. to the 3rd century, the distribution of nearest neighbor coefficients (see Figure 7a) indicates that SRN sites were primarily concentrated in lower value ranges. The lighter color and smaller size of the bubbles indicate that the data points were closely spaced and relatively clustered. The silhouette scores (see Figure 7c) remained low and relatively stable at around 0.2, suggesting poor clustering performance with data points spatially dispersed or lacking significant aggregation. The clusters identified through the DBSCAN algorithm (see Figure 7b) show that most data points lie outside the red contour area, indicating they are relatively dispersed and likely classified as noise points or independent clusters. According to Figure 7d, from the 2nd to 1st century B.C.E., the clusters were relatively dispersed. This indicates that the sites did not form significant high-density aggregation areas, suggesting dispersed social activities or limited records. It also shows that during the 2nd and 3rd centuries CE, the clusters became more pronounced, indicating increased social activities or more the establishment of more heritage sites during this period.

From the 4th to the 8th century, the nearest neighbor coefficients gradually increased (see Figure 7a), with bubbles becoming darker and larger, indicating a more dispersed spatial distribution of data points. The silhouette scores (see Figure 7c) began to rise from the 6th century onwards, showing improved clustering performance and suggesting the emergence of spatial aggregation among sites. The DBSCAN-identified clusters (see Figure 7b) also became more distinct, with some data points located within high-density areas (inside the contour), while others remained more dispersed (outside the contour), indicating an increasing tendency toward clustering. According to Figure 7d, from the 4th to the 9th centuries, several clusters evenly appeared across these periods, indicating that sites had a certain degree of spatial aggregation.

From the 9th to the 13th century, the nearest neighbor coefficients significantly increased (see Figure 7a), especially during the 11th to 13th centuries, when some data points reached the highest values. The bubbles became darker and larger, indicating a highly dispersed spatial distribution of data points. The silhouette scores (see Figure 7c) rose sharply in the 10th century, reaching a peak, indicating significantly improved clustering performance. From the 11th century onwards, the silhouette scores remained high, close to 1.0, indicating very distinct clustering characteristics. The clusters identified through the DBSCAN algorithm (see Figure 7b) showed a marked increase in number and diversity, with several forming high-density areas, particularly during the 11th and 12th centuries. This indicates significant social, cultural, or environmental changes that led to concentrated site distributions. According to Figure 7d, from the 10th to the 15th century, the number and diversity of clusters significantly increased, indicating that the spatial distribution of sites became more complex and diverse. Particularly during the 11th and 12th centuries, several clusters occupied a large proportion, indicating significant social, cultural, or environmental changes that led to concentrated site distributions.

These findings illustrate the changes in the aggregation and dispersion of sites from the 2nd B.C.E. to the 13th century. During the early period (2nd B.C.E. to 9th century), site distribution was relatively dispersed with no significant high-density aggregation areas, possibly reflecting the dispersed nature of early societies or incomplete archeological records. During the middle period (10th to 12th century), site aggregation significantly increased, with multiple clusters forming high-density areas, possibly indicating a peak in social and cultural activities. During the late period (13th to 15th century), site distribution became complex and diverse again, with multiple clusters appearing at different times. This indicates diverse social activities and regional development, possibly reflecting significant social changes and environmental shifts.

This study presents intriguing findings where some areas exhibit low nearest neighbor coefficients but high clustering. This indicates that while the overall distribution may appear dispersed, certain localized areas have a dense concentration of heritage sites. The low nearest neighbor coefficients suggest that data points are closely spaced within these local regions, highlighting significant clustering at a smaller scale. Furthermore, high clustering observed at larger scales implies that within a broad area, several smaller regions have high point density, but the distances between these smaller regions are relatively large. This results in an overall low nearest neighbor coefficient. Archeological findings often concentrate in specific hotspot regions, likely centers of ancient civilization activities. Consequently, even though the distribution appears dispersed at a macro scale, the density of points within these hotspot regions is extremely high at a micro scale.

3.3. Kernel Density Estimation Visualization

To better explore the relationships between heritage site locations, the results of the Kernel Density Estimation (KDE) analysis are presented below.

3.3.1. Heatmap Visualization and Interpretation

This section uses heatmap visualization to present the spatial clustering of heritage sites across different centuries. KDE heatmaps are employed to estimate the probability density function of a random variable in continuous space. In the context of geographic data, the heatmaps visualize the density of data points (such as settlements or events) across a geographic area, highlighting where these points are concentrated. The heatmaps use color gradients to represent point density across different regions of the map, with warmer colors (e.g., red) indicating higher densities (more data points in each area) and cooler colors (e.g., blue) indicating lower densities. KDE heatmaps provide an intuitive visual summary of data point clustering within a geographical area.

From analyzing the Kernel Density heatmap, a consistent trend is identified. In the time span from the 2nd century B.C.E. to the 3rd century (shown in Figure 8), the heatmap displays a core concentration situated in the eastern Mediterranean region, expanding continuously in both eastern and western directions. The influence extends eastward through the Iranian Plateau into the Central Asian Transoxiana region, linking with the Yellow River Basin in China, and westward along the Mediterranean to the western Mediterranean region. Examining the cultural landscape depicted by the heatmap, the Greek–Roman cultural sphere, focalized in the eastern Mediterranean, demonstrated robust external connections during this era, notably interacting with the cultural zones of the Mesopotamian region and Central Asia [33]. Furthermore, the cultural realm of the distant Chinese region, positioned in the east, was impacted by this cultural interchange centered on the eastern Mediterranean, marking the nascent stages of cross-cultural exchanges among major empires.

The Kernel Density heatmaps (shown in Figure 9) from the 4th to 8th centuries illustrate a transformative shift from a singular core initially centered on the eastern Mediterranean region to an extensive expansion in both easterly and westerly directions. This reconfiguration establishes a dual-core pattern encompassing the eastern Mediterranean–Mesopotamian region, and the Central Asian Transoxiana region, along with the Yellow River Basin in China emerging as a focal point, signifying a general shift towards the east. The former epitomizes the Near Eastern cultural realm, focusing on the eastern provinces of the Roman Empire and later the Byzantine Empire, situated amidst the Black Sea, Caspian Sea, Persian Gulf, Red Sea, and the Mediterranean—historically, a thriving hub and a key cultural center along the Silk Road network in classical antiquity [34]. The latter symbolizes the Central Asian cultural sphere, demonstrating remarkable prosperity and emphasizing the evolution of urban civilizations in the Central Asian Transoxiana region along the early medieval Silk Road, highlighting the crucial role of the Sogdian merchant network in facilitating East–West exchanges [35]. Notable bright areas in the heatmap extend eastward into the core of East Asia, establishing local hubs in the lower Yellow River region of China, showcasing the prosperity of Chinese culture and its interactions and amalgamations with other significant cultural spheres across the Eurasian continent [36].

The trend from the 9th to the 15th centuries indicates a notable decline in the Kernel Density heatmap of Central Asia and China (shown in Figure 10). This period can be segmented into two phases. Between the 9th and 11th centuries, the Silk Road continued its exchange patterns, showing cycles of prosperity and decline. The heat centers steadily shifted westward, with the focus in the Yellow River Basin of China weakening and shifting towards the southeast. Notably, during this time, the once significant heat centers in Syria and Iraq, along with the Central Asian Transoxiana region, experienced diminishing intensity and a westward shift, no longer maintaining distinct independent heat centers. Key historical events in Central Asia during this era included the Arab conquests and the unrest of the late Tang and Five Dynasties, which replaced eastern influences with those of the Arab Empire in the west [37]. In contrast, the Kernel Density heatmap from the 12th to the 15th century depicts a dual-center pattern with Europe and East Asia emerging as focal points on the Eurasian continent’s eastern and western sides. These centers operate largely independently, with Europe exhibiting a higher clustering level compared to East Asia. This era marked the rise of the Italian Renaissance and foreshadowed the Age of Discovery, originating from Western Europe. The economic prosperity of the Renaissance was fueled by the ascent of Italian city-states in the late 12th century, potentially planting the seeds of modern capitalist systems [38]. Conversely, the eastern and central regions of the Eurasian continent were besieged by conflicts linked to the Crusades, Mongol conquests, and the Timurid era wars [39]. Economic trade flourished primarily in East Asia and Southeast Asia, bolstering the prosperity of the East Asian cultural sphere (Confucian cultural circle). The disruption of the Silk Road trade severed connections running along the 40th parallel to Central Asia, hindering east–west interactions at the Tibetan Plateau along the 30th parallel. Trade from the Western regions to the expansive Indian trading zone had to navigate the low latitude waters of Southeast Asia to engage with East Asia, depicting a distinct trading pattern from earlier epochs [40].

With the discovery of the Americas in 1492, the conventional SRN eventually evolved from a closed system to a global exchange network. Europe assumed a central role in this emerging global structure. It is anticipated that extending the research methodology of this study to encompass investigations of the Americas and Oceania post-1492 would require evaluating heritage sites in these regions, contributing to conclusions that reinforce Europe’s pivotal position in global exchanges [41].

3.3.2. Scatter Plot Visualization and Interpretation

This section uses scatter plots to visualize changes in heritage sites, providing a clear representation of individual data points and their exact locations on the map. This method is typically used to study the relationships between variables or to understand the spatial distribution of individual observations. In a scatter plot, each data point is represented by a dot on the graph. The position of the dot corresponds to the exact coordinates of the observation. In this study, the red dots represent denser locations, while the blue dots indicate more dispersed area. This section provides a more detailed description of the SRN’s heritage site compared to previous section.

The density analysis map of World Heritage Sites along the SRN from the 2nd century B.C.E. to the 3rd century (shown in Figure 11) reveals a significant emergence of new heritage sites in the eastern Mediterranean region, including the Greek city-states of Asia Minor along the Aegean Coast and the Levant region on the eastern Mediterranean coast, compared to other regions. Additionally, there is a notable increase in the density of heritage sites eastward from the Central Asian Transoxiana region between the 2nd century B.C.E and the 3rd century, marking the eastern path of the ancient Silk Road extending to East Asia. The connections among high-density sites exhibit an east–west alignment, stretching from the Western Mediterranean to East Asia [42]. The 3rd to 4th centuries mark a transitional period [43]. During this time, the density of heritage sites in the Western Mediterranean declined, while it continued to grow in the Aegean Sea, the Levant, Iran, and Central Asia [44]. Consequently, the fluctuations in the number of heritage sites from the 4th to the 8th centuries indicate a new trend: a reduction in heritage sites in Western Europe, continuous growth in the eastern Mediterranean, and the emergence of new areas of heritage site growth in Central Asia and Western China.

From the 4th to the 8th centuries, the analysis of the increase and decrease in the number of World Heritage Sites within the Silk Road network presents an intriguing scenario (shown in Figure 12). This period reveals noticeable transformations: a substantial decline in heritage site number is observed in the eastern Mediterranean, particularly around the Aegean Sea, North Africa, the Levant, and the Mesopotamian region. In contrast, there is a notable increase in heritage site number in the Central Asian Transoxiana region, East Asia, and Southeast Asia. The reasons for this change may lie in the fact that during this historical period, the fragmentation and fall of the Roman Empire led to prolonged periods of conflict and disunity in the Mediterranean region [45]. The collapse of the Western Roman Empire in 476 C.E. required the establishment of the Carolingian Empire by the Franks (771–814 C.E.) for stability to return to Western Europe. In the eastern Mediterranean, the Byzantine Empire engaged in protracted conflicts with the Sassanian Persians and the Arab Empire in the Levant and North Africa [46]. Climatically, the primary regions of the Silk Road network encountered cooling events during the 3rd to 6th centuries. A distinctive trend in climate change from lower to higher annual average temperature with notable fluctuations had a notable impact on North Africa, resulting in a continual decrease in heritage sites from the 4th to the 8th centuries [47,48]. Consequently, there was a reduction in the number of Roman heritage sites in the western region. Despite the upheavals in East Asia during the Six Dynasties, factors such as the introduction of Buddhism, the flourishing Sui and Tang dynasties, and the growth of thriving Sogdian merchant communities significantly bolstered trade and cultural exchange in the area [49]. The connections among high-density points formed a spindle-shaped pattern extending from northeast to southwest, with primary concentrations in the eastern Mediterranean and Central Asian Transoxiana region.

The 8th to 9th centuries signify yet another transitional period [50]. During this time, the increase in heritage sites in Central Asia and Western China stagnated, while the number of sites in Eastern China further declined. As a result, the SRN centered around the eastern Mediterranean, Iran, and Central Asia experienced an interruption in the trends of increasing and decreasing heritage sites observed from the 4th to the 8th centuries in various regions of the Silk Road. Meanwhile, the number of heritage sites in Western Europe, including present-day France, Germany, and Northern Italy, significantly increased. Consequently, the Silk Road underwent a historical decline from prosperity between the 9th and 15th centuries, a significant transformation that can be further divided into two stages: the 9th to 11th centuries and the 12th to 15th centuries.

From the 9th to the 15th century, there was a steady rise in the number of heritage sites rooted in medieval Europe (shown in Figure 13). The pace of growth accelerated notably, surpassing all other regions by the 12th century. Compared to Europe, the number of heritage sites in the Levant, Mesopotamia, Central Asia, and Western China did increase during this era, but the density gap widened significantly. The interaction patterns of the Silk Road network persisted between the 9th and 11th centuries, characterized by fluctuations. While the Silk Road’s path remains visible on maps, the eastern segment from Central Asia through Xinjiang to East Asia progressively diminishes between the 12th and 15th centuries until its eventual disappearance [7].

4. Discussion

4.1. Analysis of Assumptions in the SRN Study

The study on the SRN comprehensively addresses the hypotheses that frame its investigation into the historical development and dynamics of the SRN. Firstly, the SRN is confirmed to have functioned as an enclosed and stable system from its formation after General Zhang Qian’s visit to Central Asia in the 2nd century B.C.E. until the end of the 15th century, when the advent of Christopher Columbus’s navigation to the Americas in 1492 transformed it into the GRN. This transition integrated a significant portion of the global population and resources, marking the evolution from a closed to an open system. Secondly, the study utilizes heritage sites listed by UNESCO as precise indicators of the SRN’s prosperity and historical significance. By creating a geographic information database of these SRN sites, the study reconstructs the SRN routes with a higher degree of accuracy than documentary sources. This approach assumes that heritage sites, which once existed and flourished, effectively reflect the prosperity of the SRN’s interaction system. Lastly, the network density between these heritage sites is used to mirror the network density of the SRN itself. By employing spatial analysis, the study illustrates the distribution and evolution of heritage sites, revealing varying levels of route usage and significant historical interactions. This method highlights how different regions within the SRN experienced distinct impacts during periods of climate change, reinforcing the study’s assumptions. Overall, the research methodically validates the hypotheses that the SRN was an enclosed and stable system, that heritage sites reflect its prosperity, and that their network density accurately represents the SRN’s network density.

4.2. Analysis of Relevant Elements on SRN Heritage Sites

By comparing the evolution of the Silk Road network (SRN) across three major historical periods following its establishment in the 2nd century B.C.E., this study analyzes the intrinsic factors driving SRN changes.

4.2.1. Climate Change on the SRN

Based on the findings, significant spatial changes in many heritage sites are revealed. The heritage sites mentioned in this study can also be understood as traces of past human habitation, such as settlements. Climate change is a major natural factor influencing the evolution of the SRN. The alternation of warm and cold periods, driven by global climate fluctuations, has had a significant impact on SRN regions, particularly in the mid- to low-latitude areas of the Eurasian continent and the Mediterranean coast. Research indicates that climate fluctuations, particularly in precipitation and temperature, significantly influenced the rise and fall of civilizations in the arid regions of Central Asia. Warm and humid climates contributed to the rise of empires, while deteriorating climates led to the decline of civilizations and the abandonment of cities [36]. Moreover, fluctuations in precipitation levels also affected the rise and fall of trade routes and urban centers along the Silk Road in Central Asia. Wet periods promoted urban expansion, while dry periods hindered the development of these networks [51]. Further research reveals that long-term climate changes, such as mega-droughts, delayed or altered patterns of human settlement and cultural exchange along these routes [39]. However, the effects of climate on human activities may display a temporal lag. Furthermore, the effects of short-term climate changes on SRN alterations may not be immediately apparent. Therefore, the timing of climate change does not necessarily align with changes in the SRN. According to the three historical periods delineated in this study, the following general trends emerge [52]:

Classical Period Warming (2nd century B.C.E to 3rd century): During this period, the main SRN regions experienced higher average temperatures compared to the modern climate (represented by 1950). Until the 3rd century, the overall climate trend gradually shifted from warm to cold.

Post-Classical and Early Medieval Climate Fluctuation (4th to 10th century): After experiencing cooling during the 3rd to 4th centuries, the main SRN regions had average temperatures lower than those of modern climate. During these periods, the general trend was a temperature increase from a lower average, accompanied by significant fluctuations. This trend continued until a sudden rise in the latter half of the 8th century, which exceeded modern climate levels and persisted until the 10th century.

Late Medieval Cooling (11th to 15th century): This period was characterized by several major temperature drops (12th, 14th, and 15th centuries), leading to a slow cooling process. Temperatures declined from a peak around 1000 C.E. to a low around 1300 C.E., followed by a gradual increase, with more dramatic fluctuations compared to the first period.

This analysis highlights how climate change significantly influenced the development and shifts in the SRN, with different regions experiencing varied impacts during cooling and fluctuation periods.

Analysis using Kernel Density heatmaps reveals that during the warm period of the Classical Era, the SRN exhibited dominance in the eastern Mediterranean dominance and extended both east and west. This period was characterized by a high density of heritage sites centered in the eastern Mediterranean. During the Late Medieval Cooling, the SRN exhibited western dominance, with a high density of heritage sites in Europe. Conversely, during the climate fluctuation periods, the SRN showed a “dual-core” characteristic, with significant centers in the Near East and Transoxiana (Central Asia).

Considering the movement of the network’s central points, it is evident that during the first and third periods, the central points shifted westward, while during the second period, they shifted eastward. This indicates the substantial impact of climate change on different civilization regions within the Eurasian continent.

If the historical periods of the Eurasian continent are classified into the warm period, the cooling period, and the climate fluctuation period, the former corresponds to continuous temperature changes, while the latter corresponds to the fluctuation period. These periods impacted the main regions of the SRN, namely the eastern, central, and western parts of the mid- to low-latitude Eurasian continent and the Mediterranean coast. Notably, during the warm periods, both the eastern and western parts of the Eurasian continent experienced sustained development. However, Western Europe, being closer to the eastern Mediterranean, progressed at a faster rate, and the central points were westward [53]. During the prolonged cooling period, major civilization regions primarily faced significant challenges due to persistent cooling trends. In the western part of the Eurasian continent, characterized by maritime and Mediterranean climates, temperature drops could lead to reduced agricultural production, a shift towards more cold-resistant crops, and the exploitation of marine resources like fish [54]. In the temperate monsoon climate zone of the eastern Eurasian continent, the prolonged cooling period could cause severe cold, threatening agriculture and population, and the central points were also westward. This examination underscores how warm, cooling, and fluctuation periods had distinct and significant impacts on the historical development of the SRN and its regions.

Comparatively, the continental climate of central Eurasia was more severely affected by cooling periods, whether in the context of long-term or short-term cooling, potentially leading to interruptions in agricultural production or shifts to alternative livelihoods [48]. However, the inhabitants of this region were likely better adapted to cold climates than those in other climatic zones. During the cooling phases, they could change their livelihoods or migrate to cope with climate changes. Following the significant temperature drop at the end of the Classical Period, nomadic groups from the inland continental climate zones moved south, threatening the agricultural civilizations of East Asia and the Mediterranean world.

The Near Eastern civilization regions at the junction of Eurasia and Africa, and the Indian subcontinent, fall under subtropical desert climates and tropical monsoon climates, respectively. Cooling periods facilitated cultural development in these regions. Consequently, during the cooling period of the Classical Era, these areas were interconnected via the maritime Silk Road, becoming significant regions within the SRN.

During climate fluctuation periods, the greatest challenge faced by major civilizations was sudden cooling. There are records of abrupt cooling in the post-Classical and early Medieval periods, often linked to significant historical events. The severe impact of climate fluctuations during these periods frequently led to cross-civilization interactions. Compared to Western Europe and the Mediterranean, East Asia and Central Asia may have exhibited a greater adaptive capacity to sudden climate fluctuations. For example, during the severe cooling of the 3rd to 4th centuries, the Western Roman Empire completely collapsed, leading to significant decline in Western Europe and North Africa [55]. These areas gradually recovered with the arrival of another prolonged warm period in the 8th century. In contrast, while East Asia also experienced political crises due to the cooling in the 3rd to 4th centuries, the region’s greater longitudinal extent, crossing multiple climate zones, allowed for the preservation of governance and civilization in the southern regions, enabling rapid recovery in the next warm period [56].

During climate fluctuation periods, the short durations of each warming and cooling phase meant that East Asian states, benefiting from a larger north–south expanse, were not uniformly affected by climate changes across all zones [57]. The variability in climate impact allowed some regions within East Asia to adapt better to changes, leading to a more resilient response to these fluctuations [58]. Consequently, they exhibited greater adaptability to climatic variations than Western Europe and the Mediterranean. If the period surrounding the climate fluctuation period (the second period) was characterized by a gradual rise in average temperatures, the most significant difference between the two 800-year-apart continuous temperature-changing periods is the substantial climate fluctuations in the latter. The combined effects of warming, prolonged cooling, and significant climate fluctuations could lead to paradigm shifts in historical development. This period facilitated the transformation of the SRN into the GRN, establishing Europe’s central role within the SRN.

During the two continuous temperature changing periods, especially the late medieval period with significant climate fluctuations, Europe emerged as the SRN hub. This reflects the resilience of the maritime and Mediterranean climates in Western Eurasia to climate change. This resilience is due to several factors: First, Europe’s proximity to the North Atlantic Drift brings a stable, warm, and humid climate, making it milder than other regions at similar latitudes, thus maintaining stability during long lasting periods [59]. Second, the main characteristics of a maritime climate are warm winters and cool summers. The ocean’s high heat capacity moderate’s temperature changes, allowing Europe to remain stable during sudden cooling [60]. Lastly, the Mediterranean climate features wet winters and dry summers, providing the region with abundant agricultural and natural resources. These resources can be stored, helping society and the economy better withstand adverse climate impacts.

4.2.2. Historical Events and Heritage Site Dynamics on the SRN

Human activities had a profound impact on the changes in the SRN and heritage site locations [61]. Key driving factors included trade, the spread of religion, war, and political changes [62]. This study organizes these factors into a table and uses visualizations to display the timelines of different events (as seen in Figure 14), aiming to connect these with the developments of SRN heritage sites, particularly with the findings discussed in the previous section. Special attention is given to political events: the rise and fall of various civilizations and empires continually altered the main routes and nodes of the Silk Road. For example, the expansion of the Arabian Empire and the prosperity of the Tang Dynasty facilitated exchanges between Central Asia and China, while the unification under the Mongol Empire further strengthened East–West links [63]. These activities caused the rise and fall of cities and sites along the route, exemplified by the prosperity of Samarkand and Balkh during different periods [64].

From the 2nd century B.C.E. to the 3rd century, key historical events significantly influenced the establishment and consolidation of the SRN. Zhang Qian’s mission to the Western Regions (139 B.C.E.) marked Han China’s direct involvement in the Silk Road trade, promoting East–West exchange of goods and culture [65]. The Han Dynasty’s establishment of the Protectorate of the Western Regions (60 B.C.E.) further secured and facilitated Central Asian trade routes [65].

The Roman Republic’s expansion, through the Third Punic War (149–146 B.C.E.) and the conquest of Gaul (58–50 B.C.E.), established Roman dominance in the western Mediterranean [66,67]. This period also witnessed the spread of Buddhism into Central Asia and China, enhancing cultural exchanges along the Silk Road.

Heritage sites from this period are concentrated in the eastern Mediterranean, Central Asia, and China, reflecting vibrant trade and cultural interactions facilitated by these significant events.

From the 4th to the 8th century, SRN site density fluctuated due to climate changes and political upheavals. The decline of the Roman Empire and the rise of the Byzantine Empire marked significant shifts in the SRN’s western regions. The Justinian Plague (541 C.E.) and prolonged Byzantine–Sassanian conflicts disrupted stability, reducing heritage site density in the eastern Mediterranean [68].

In contrast, the stability of the Sassanian Empire in Iran and the flourishing Sogdian trade network in Central Asia contributed to the persistence and growth of heritage sites in these regions [69,70]. The spread of Buddhism continued to influence the establishment of religious sites along the SRN [71].

From the 9th to the 15th century, significant transformations in the SRN occurred due to the rise and fall of various empires and the impact of pandemics. Arab conquests and the establishment of trade routes under Islamic rule enhanced connectivity between the Mediterranean, Central Asia, and China [72]. The reunification of China by the Tang Dynasty and the establishment of the Silk Road as a major trade route promoted the growth of heritage sites in East Asia [73]. Interactions between the Byzantine Empire, Islamic Caliphates, and Tang Dynasty facilitated the exchange of goods, ideas, and culture, resulting in a high density of heritage sites in the Near East, Central Asia, and China.

From the 11th century onwards, significant transformations occurred in the SRN. The rise of the Mongol Empire under Genghis Khan and his successors created a vast, unified territory, enhancing the safety and efficiency of the Silk Road trade routes [74]. The Pax Mongolica (13th to 14th Century) increased commercial and cultural exchanges across Eurasia, reflected in the high density of heritage sites in Central Asia and China [75]. The decline of the Mongol Empire and the onset of the Black Death (1346–1353) caused trade disruptions and shifted the SRN center towards Western Europe [76]. The Renaissance in Italy and the Age of Discovery marked the decline of the overland Silk Road and the rise of maritime trade routes [77].

The evolution of the SRN’s sites is closely linked to significant human activities and specific historical events. From the expansion of empires and the spread of religions to climatic changes and pandemics, these factors shaped the spatial patterns and density of heritage sites along the SRN. Understanding these relationships provides valuable insights into the long-term impact of the Silk Road on cultural and economic exchanges between the East and the West [7].

4.3. Limitations

While this study provides a comprehensive analysis of the Silk Road network (SRN) using UNESCO World Heritage sites to reconstruct SRN routes, it has notable limitations. One significant limitation is the uneven distribution of heritage sites across different regions. Europe has a disproportionately high number of heritage sites compared to regions like Central Asia, the Near East, and East Asia. This imbalance could introduce bias, leading to an overrepresentation of European sites and potentially skewing the interpretation of the SRN’s historical significance and network density. Consequently, the heritage site density in Europe might not accurately reflect the actual historical network density of the SRN across all regions.

Despite valuable insights into the clustering and dispersion patterns of sites from the 2nd century B.C.E. to the 15th century, several limitations persist. The sensitivity of the DBSCAN algorithm to parameter settings can yield different clustering outcomes based on the chosen parameters. Interpreting silhouette scores is challenging, as low scores may reflect the intrinsic nature of spatial distributions rather than poor clustering quality. Furthermore, variations in data collection methods and preservation conditions may introduce biases, and the complexity of environmental and social factors is difficult to fully capture, leading to potential interpretative biases.

Future research might consider incorporating additional data sources or applying normalization techniques to address discrepancies, ensuring a more balanced and representative analysis of the SRN’s historical development and interactions. Additionally, future studies should integrate finer temporal and spatial data, ensure data quality and consistency, and employ multiple methods to validate results. This approach will enhance the reliability and accuracy of clustering analysis.

4.4. Recommendations for Enhanced Conservation and Management

This study analyzes SRN heritage distribution across centuries, revealing changes in clustering patterns over time. Using multiple methods, the research mapped the spatiotemporal evolution of settlements and identified areas of significant historical activity. The discussion links these findings to relevant historical events, hypothesizing reasons behind the observed changes in settlement patterns. These findings highlight the significant impact of historical trade routes on settlement development and land use patterns. Analysis of site clustering reveals that these clusters are not merely geographic concentrations but represent key nodes in the historical Silk Road network. The connections between these nodes form a crucial network structure for trade and cultural exchange. Based on the identified settlement clusters, recommendations are made to preserve and enhance the connectivity of cultural landscape heritage. The proposed conservation strategies focus not only on preserving individual sites but also on maintaining the integrity of the entire SRN [78]. Here are three recommendations:

• Prioritize Conservation Based on Spatial Clustering: Prioritizing conservation efforts using spatial clustering is vital for managing heritage landscapes along the SRN. This method preserves not only the physical integrity of sites but also the continuity of the heritage landscape and its historical narrative. Spatial clustering analysis allows for a transition from protecting isolated sites to adopting a more comprehensive strategy that safeguards entire clusters of heritage sites. By concentrating on clusters of significant cultural and economic activity, conservation efforts can more effectively preserve historically interconnected sites, thus maintaining the network’s overall integrity.
• Enhance and Protect Regional Connectivity: Regional connectivity is crucial to the formation and sustainability of the Silk Road network. Preserving connections between historical sites is not only essential for individual site protection but also for maintaining the coherence of the entire network. This helps in comprehensively understanding the cultural and historical significance of the heritage. To enhance protection, establishing buffer zones around high-density clusters may be necessary. These zones act as safeguards, preventing modern developments from disrupting the historical context and spatial relationships critical to the evolution of these heritage landscapes.
• Foster Transnational Collaboration: Beyond these protective measures, strategies to maintain and strengthen linkages between heritage sites are crucial. Notably, transnational linkages, such as the bid for the Routes Network of the Chang’an–Tianshan Corridor, highlight the importance of multinational cultural heritage and cooperation [79]. Preserving historic routes and preventing modern infrastructure from disrupting these connections is vital to maintaining the Silk Road network’s historical and cultural integrity. Such strategies are integral to a holistic approach to heritage conservation, ensuring that both the tangible and intangible aspects of the Silk Road’s cultural heritage are preserved for future generations.

5. Conclusions

This study provides a detailed analysis of the Silk Road network (SRN) by utilizing data from the UNESCO’s World Heritage Sites to reconstruct its historical routes and understand its evolution. The research demonstrates the SRN’s function as a closed and stable system from its establishment in the 2nd century BCE until the 15th century when it transformed into a global network. The use of heritage sites offers a precise reflection of the SRN’s prosperity and historical significance, and the network density between these sites effectively mirrors the network density of the SRN itself. The understanding of the SRN’s historical development and its role in facilitating cross-cultural exchanges is significantly enhanced by these findings.

The study utilizes spatial and statistical analyses, including Kernel Density heatmaps and spatial clustering, to draw its conclusions. The study reveals the spatial distribution of site clusters, offering an in-depth understanding of their key role in the Silk Road’s cultural network. The main findings validate previous assumptions and identify key factors driving changes in SRN heritage sites across different centuries, particularly climate change and human activities. It identifies trends in the evolution of the SRN over time, along with its changing spatial characteristics. These changes are characterized by a core area in the eastern Mediterranean during the Classical Period, a shift to a dual core in the eastern Mediterranean and Central Asia during the medieval period, and a subsequent transition to a European core in the late medieval period.

Preserving this network is crucial for maintaining its historical value; hence, the proposed conservation strategy emphasizes this aspect. Specifically, conservation efforts should focus on clusters of heritage sites with significant cultural and economic activity to preserve historically interconnected sites and maintain the integrity of the entire Silk Road network. Additionally, we should preserve and strengthen connections between heritage sites by implementing buffer zones around high-density clusters, safeguarding the network’s coherence, and preventing modern developments from disrupting historical routes and spatial relationships. Finally, we should encourage and support transnational linkages, such as multinational cultural heritage initiatives, to ensure the continued protection of the Silk Road’s historical and cultural integrity through cooperative efforts.

Overall, the practical value lies in providing a more accurate representation of the SRN compared to traditional documentary sources, thanks to the precise geographic and historical data of the heritage sites. The innovative insights include demonstrating how different regions within the SRN experienced distinct impacts during periods of climate change, and how the network density between heritage sites reflects the SRN’s overall network density. This research addresses the practical problem of reconstructing historical trade routes and has potential applications in historical geography, archeology, and cultural heritage preservation.