Defense layout characteristics of Tang dynasty military sites in Aksu prefecture of China

Abstract

Tang dynasty military sites (TDMSs) in Aksu prefecture are significant witnesses to Xinjiang’s history, and an essential part of Xinjiang’s military and the Silk Road cultural heritage. However, their exploration and preservation currently require further enhancement. In this study, by establishing a database of 164 existing TDMSs, visibility analysis tool was applied to hypothesize the locations of potentially lost sites, which were then used as the supplementary and corroborating data. Combining network analysis method, qualitative and quantitative analyses were conducted on these sites. The results show: (1) TDMS in Aksu can be divided into three defense zones, each with different defense strategies. (2) The visibility network structure of each defense zone presented two types, namely simple linear network structure and complex large-scale structure. These findings reveal the overall defensive layout characteristics of the TDMS in Aksu, providing a positive basis for the protection of military heritage in Xinjiang.

Introduction

Xinjiang uygur autonomous region, which was historically known as the western regions (WR), is located in the heart of the Eurasian continent. As Fig. 1a displayed, Xinjiang was a key crossroads of the Silk Road (SR), linking the central plains of ancient China with Central Asia, Western Asia, and South Asia. In the third year of the Jianyuan era of the Han dynasty (138 BCE), Qian Zhang first made contact with the WR. Then in the third year of the Shenjue era (59 BCE), the central government of the Han dynasty established the WR Protectorate, officially incorporating the vast areas north and south of the Tianshan Mountains into the Han empire¹, and making them an inseparable part of China’s united multi-ethnic state. In the Tang dynasty, the central government established the Anxi protectorate (AP) and Beiting protectorate to govern the areas north and south of the Tianshan Mountains, further expanding the empire’s territory. At its peak, the Tang dynasty’s domain stretched from the Congling (old term for the Pamir Plateau) to Persia (old term for the Iran)². Afterward, the control of the central plains dynasties over the WR gradually weakened. It was not until the twenty-seventh year of the Qianlong reign (1762) that the government of the Qing dynasty, following the historical tradition of the Han and Tang dynasties in governing border regions through military administrations, reestablished direct control over the WR. In the tenth year of the Guangxu reign (1884), the region was renamed Xinjiang³. As Fig. 1b shown, Aksu prefecture historically known as Qiuci (龟兹) is located in the central-western part of the Xinjiang, with the Tianshan Mountains in the north and the Taklamakan Desert in the south. Since the Han dynasty, this region had been one of the key routes on the SR, falling under the jurisdiction of the Han dynasty’s WR protectorate. In the Tang dynasty, the southern Tianshan region was divided into four military garrisons known as the Anxi Four Towns, namely Qiuci, Yutian (于阗), Yanqi (焉耆)/Suiye (碎叶), and Shule (疏勒). These towns were governed by the AP. According to Houhanshu (后汉书), “If Qiuci is taken, then only one percent of the WR will remain unconquered,”¹ highlighting the significant role of Qiuci in the central plains dynasties’ control over the entire WR.

Fig. 1: Location map of the study area.

a Location of Xinjiang in China; b Location of Aksu prefecture in Xinjiang. Data Sources: Silk Road Historical Geography Information Open Platform (https://www.srhgis.com/homePage).

Specifically, from a transportation perspective, Qiuci was a crucial transportation hub in the ancient WR. The Tang dynasty was able to control the surrounding areas by holding Qiuci. From Qiuci to the east, past Yanqi (now known as Yanqi county), one could reach Xizhou (now known as Turpan city), which served as the gateway to the WR. To the west, one could reach Shule (now known as Kashgar prefecture) and onward to the Pamirs. To the south, it can follow the Hotan river and Kriya river to Yutian (now known as Hotan prefecture). To the north, through the Tianshan Mountain passes to reach Gongyue (now known as Yining county)and Suiye (now known as Tokmok, Kyrgyzstan), and further into Central Asia. From a resource perspective, Qiuci had vast oases and fertile land, making it suitable for large-scale military garrisons and farming. It also had abundant mineral and livestock resources, which could be used for forging weapons and raising warhorses. From a commercial perspective, Qiuci was a key stop on the Central Route of the Tang dynasty’s SR. There were many business travelers along the SR, with commerce and trade flourishing⁴. Thus, in order to manage this crucial trade route and military-political hub, the Tang government carried out garrison farming and border defense activities in Qiuci. A large number of cities, beacon towers were constructed, gradually developing a comprehensive military defense system^4[,5. Today, a lot of Tang dynasty city sites and beacon tower remains can still be found in Aksu. These Tang dynasty military sites (TDMSs) have become important carriers for witnessing and exploring the military and SR cultures of the Tang dynasty in the WR.

The investigation and compilation of military remains in Aksu began in the late Qing dynasty, especially during the second half of the 19th century to the early 20th century. During this period, many foreign explorers conducted a series of plundering archeological expeditions, marking the beginning of the investigation and study of Xinjiang’s ancient cultural heritage. These explorers, such as the British Hungarian Stein^6,7, the Germans A. Grunwedel and Le Coq⁸, the French Paul Pelliot⁹, and the Japanese Ōtani Kōzui¹⁰, carried out investigations of ancient sites within WR, including the Aksu region, providing the first important archeological data. These survey results have become key reference materials for later scholars studying the history and culture. In the first half of the 20th century, the work of the first modern Sino-Swiss scientific westnorth expedition hosted by Chinese scholars advanced the study of the history and geography of the WR^11,12. This work accumulated rich experience for subsequent archeological research and later preservation efforts in Xinjiang. After the establishment of the People’s Republic of China, the relevant archeological work made even more significant progress. In particular, several large-scale national cultural heritage surveys organized by the government have provided a detailed inventory of the region’s historical and cultural relics. These efforts resulted in a series of research achievements based on the findings^13,14. These achievements include the statistics, classification, distribution, and preservation status of historical and cultural heritage. In the past decade, based on earlier archeological survey results, scholars have further sorted out and studied remains at different regional scales. Key areas of research included the study of Tang dynasty military garrisons along the Tianshan Corridor, with Qiuci as the strategic hub^4,15,16,17; the classification of cultural remains in the Tarim Basin, focusing on the Aksu region^4,18; and the distribution of ancient cities in southern Xinjiang^19,20. These studies provided a more in-depth archeological analysis and organization of cultural remains of various periods and types, including the TDMS.

In addition to the above investigations and studies, many scholars in the field of historiography have conducted extensive theoretical studies on the historical and social background of TDMS. These studies mainly focused on three aspects: the history of transportation in the WR, the history of military systems, and the history of trade and commerce along the SR. On the first aspect, the military remains in Aksu were closely related to the history of transportation in the WR. There were already detailed research findings regarding the distribution and routes of the Tang dynasty’s ancient roads within WR^21,22,23. Building upon this, scholars have pointed out that the Tang dynasty strengthened its control over the main transportation routes in the WR through the strategic placement of military facilities. It not only ensured smooth communication between the central plains and the WR, but also safeguarded the flow of the SR²⁴. At the same time, the defensive layout in Aksu during the Tang dynasty was closely aligned with the transportation routes. This integration of military facilities with key trade routes both enhanced the flexibility of military defense and ensured that the Tang dynasty could quickly mobilize forces in times of emergency²⁵. On the second aspect, the military system in the WR during the Tang dynasty has also been studied quite comprehensively. This kind of research primarily focused on the protectorate system, the military garrison system, and the Jimi prefecture system implemented in the Anxi region. These systems, as effective means of the control over border areas, formed the core of the military defense system in the WR. Among them, the protectorate system refered to institutions like the AP and Beiting protectorate established by the Tang dynasty in the WR, which were responsible for overseeing both military and administrative affairs. Under the leadership of the protectorates, the military garrison system provided direct military control over the WR²⁶, while the Jimi prefecture system governed the local ethnic groups by establishing prefectures and counties²⁷, with local tribal leaders acting as administrators on behalf of the central government. Through the “one region, two systems” approach, relying on military garrisons and a combination of military settlements and defense, the Tang dynasty not only maintained political stability in the WR, but also ensured military control over the various ethnic groups along the SR²⁸. In addition, the discovery and study of Tang dynasty military artifacts and documents confirmed the effective governance of the WR by the authorities^29,30. These findings provided empirical data for research on the military deployments of the Tang dynasty in the WR. On the third aspect, the Tang dynasty’s military deployment in Qiuci was closely linked to the economic activities along the SR. Scholars often emphasized that the four military garrisons of the Tang dynasty (known as Anxi Four Garrisons) were located at key strategic points on the SR, playing crucial roles in both trade and defense^31,32. Through the deployment of numerous city sites, beacon towers, and other military facilities, the Tang central government effectively controlled and guarded the SR trade routes connecting the central plains to Central Asia. Meanwhile, the prosperity and smooth flow of the SR ensured the transportation of military supplies, food, and horses, as well as the daily needs of soldiers and their families stationed in the region^33,34,35.

In addition to research on archeology and historiography, in recent years, some scholars have also approached the study from the perspective of heritage conservation. They have investigated the causes of damage or even the disappearance of some sites³⁶, and proposed corresponding protective measures^37,38. Overall, the foundational theoretical research on the Aksu region was already relatively detailed. However, due to a lack of historical records, incomplete remains, and archeological excavations have not yet been fully conducted, the military deployments in Aksu during the Tang dynasty, i.e., the defensive layout of military building is still unknown. The main gap lies in the focus on the spatial distribution of sites, but there is a lack of in-depth exploration of the internal defensive mechanisms of the site complexes. Additionally, existing research mainly relies on traditional archeological methods, with relatively little application of modern information technologies and methods. Currently, a growing number of academic research on similar military architectural sites have begun to integrate archeology with modern information technology. Among these, the research ideas based on GIS spatial analysis^{39,40,41,42,43} and network analysis^44,45,46,47 provide feasibility for revealing the social relationships between sites. For example, the visibility between military defense facilities is considered an important factor in explaining their site selection^48,49,50. The visibility network based on visibility analysis is believed to reflect various material and spiritual functions, such as resource acquisition, military operations, transportation choices, and religious worship, etc⁵¹. However, such methods are mainly used in American and European archeology at present, with fewer applications in Chinese archeology.

In response to the long history, numerous sites, broad distribution, incomplete remains, and complex relationships of the Aksu region’s sites, GIS-based analysis methods that integrate geographic and historical data can help explore the use of ancient military sites across multiple scales and dimensions. Therefore, this study draws on archeological GIS-based analysis research and practices from both domestic and international sources⁵². Based on Chinese ancient beacon system for information transmission^53,54, the visibility network was constructed through visibility analysis. Combined historical documents and existing research materials, visibility networks were interpreted and spatial defense relationships were in turn excavated. The results can provide new insights for reconstructing of the Tang dynasty’s military and administrative system in Xinjiang, as well as offer a feasible framework and comparative basis for the study and preservation of military site complexes in Central Asia.

This study consists of the following two main components.

• (1) Based on the ancient beacon system, ArcGIS software was utilized to reconstruct the visibility relationships of existing TDMSs in Aksu so as to preliminarily explore the military defensive strategies and layouts of Tang dynasty. Meanwhile, by analyzing the overlapping visible areas of adjacent sites, the locations of disappeared site points were then inferred and reconstructed, which were used as supplemental and supporting evidence for the existing sites. The overall characteristics of the military defensive layout in Aksu were further examined through integrating historical texts and existing research materials.

• (2) GIS-based visibility analysis results were input into Python to generate visual network diagrams. Combining various evaluation indexes of network analysis, the transmission paths of military information, the delineation of defensive zones, the network structural characteristics of each zone, and the defensive centers and information hubs in the network were examined in turn to reveal the complex military defensive strategies of the Tang dynasty in Aksu.

Methods

Data sources and database set-up

The data sources for the TDMS involved in this study were as follows:

• (1) The official survey listings from the Cultural Relics Bureau of Xinjiang, primarily including the Compilation of Results from the Third National Cultural Relics Survey of the Xinjiang (2011)¹³ and Xinjiang Great Wall Resources Survey Report (2014)¹⁴. These materials represented the most recent comprehensive survey reports on ancient city sites and Great Wall heritage in Xinjiang, serving as a crucial foundational data source for this study.

• (2) The archeological survey data of modern and contemporary sites in Xinjiang, mainly including the Marc Aurel Stein’s Archeological Atlas of the WR⁶, Archeological Atlas of the Asian Hinterland⁷, and Huang Wenbi’s Archeological Records of the Tarim Basin¹². These materials provided information on three beacon tower sites that have disappeared.

• (3) Modern and contemporary local histories of the Aksu prefecture in Xinjiang, which supplemented one beacon tower site that has disappeared.

• (4) Field survey. Based on the above materials, field investigations were conducted across the entire Xinjiang to further verify the site data.

The uncertainty of the aforementioned data need to be considered during its use. First, due to the limitations of archeological understanding, it is challenging to accurately determine the age of some sites that are poorly preserved and have yielded few artifacts. The official survey listings have tentatively classified such sites as suspected Tang dynasty based on their location and building form. During the analysis, if a high correlation was found between these sites and the Tang dynasty sites, they were therefore considered as a supplementary to the TDMS. The final number of extant military sites totals 164, including 82 Tang dynasty city sites (codes C1 ~ 82), 31 sites suspected Tang dynasty city sites according to official records (codes SC1 ~ 31), and 51 Tang dynasty beacon tower sites (codes B1 ~ B51). The latitude and longitude coordinates of these sites were input into ArcGIS 10.8 to create a dataset of site points in Aksu, as shown in Table 1 and Fig. 2. Second, as the Tang dynasty was over 1000 years ago, some sites have been lost due to natural erosion and human damage. In the analysis process, the locations of such disappeared sites was estimated individually. Eventually, 34 presumed military sites (codes PB1 ~ 34) were obtained and used as additional data for the study. The specific methodology and process for inferring location will be explained in detail in the “Visibility Analysis” subsection of the “Methods” section.

Fig. 2: Distribution map of sites in Aksu, including 164 extant military sites and 34 presumed military sites.

Map Source: Xinjiang Uygur Autonomous Region Platform for Common Geospatial Information Services, with review number 新S(2023)061 (https://xinjiang.tianditu.gov.cn/bzdt_code/bzdt.html).

Table 1 Sources of information for each site

The visibility analysis for this study required the following two types of data.

• (1) Digital Elevation Model (DEM). The SRTM GL1, released by NASA, is a DEM constructed using radar observations that penetrate through surface vegetation to reach the ground, with a precision of 30 meters. However, in terms of the detailed representation of surface morphology, a DEM with a resolution of 12.5 meters is closer to the actual terrain than one with a resolution of 30 meters. Therefore, in order to improve the accuracy of the visibility analysis results, this study selected the ALOS PALSAR DEM with a resolution of 12.5 m. It was an modified version of the SRTM GL1 enhanced with the use of the Radiometric Terrain Correction technique.

• (2) Restrictions on observation point height and observation distance. When conducting viewshed analysis in ArcGIS, the following two variables needed to be considered. First, the height of the observation point, which includes the observer’s height plus the height of the beacon tower or wall. Since beacon signals were mainly used to transmit information in the WR during the Tang dynasty was through beacon signals⁵⁴, beacon towers and city sites with raised platforms were the main carriers of information exchange. Fortresses, barrier garrisons, and other city sites would then take targeted military defensive measures upon observing military activity. Therefore, in the visibility analysis, it is necessary to separately consider the observation point heights for sites such as beacon towers, barrier garrisons, and city sites. According to the records in the Fengshi (烽式), “The beacon tower should be placed on high mountains with clear views in all directions, or in flat areas when no mountains are available, with a garrison fort built below. The height of the beacon tower is typically around five zhang (Tang dynasty 1 zhang (丈) = 3.07 m today)⁵⁵.” It can be inferred that the height of the beacon tower is approximately 15.35 m to enable long-distance observation, plus the average observer’s height (take 1.65 m⁵⁶). Thus, the observation point height of 17.00 m was used in this study for visibility analysis of beacon towers and city sites with raised platforms. For fortresses, barrier garrisons and city sites without raised platforms, the observation point height was determined by adding the height of the city walls to the observer’s height. According to existing research^57,58, the height of city walls during the Han and Tang dynasties was generally similar to the width of the wall base, typically ranging from 6.00 to 28.00 m. Therefore, this study used the city wall information from the national cultural relics census data as the basis for setting the observation heights for various types of sites. Second, the maximum radius of the viewshed calculation, which is the furthest distance the observer can see. Due to limitations such as the curvature of the Earth, the resolution of the human eye, and other factors, the observable distance is finite even under optimal weather conditions and there are no obstacles blocking the line of sight. Through calculations (First, under the influence of Earth’s curvature, the distance visible from a height of 17.00 m is approximately 14.730 km, which corresponds to 32.10 li (里) in the Tang dynasty. Second, within the human eye resolution range of 1/2000 to 1/5000, an object at a distance of 14.730 km must have a height between approximately 2.946 m and 7.365 m for it to be visible. It means that objects observed at 14.730 km from the observation point need to hSave a minimum height of 2.946 m, which is sufficient for beacon or smoke signals to be visible.), this distance closely matched with the mileage specified in the Tang dynasty beacon system as recorded in the Wujingzongyao (武经总要), which stated that “For border cities, a beacon should be placed every thirty li (Tang dynasty 1 li (里) = 459 m today)⁵⁵.” As a result, the calculated distance of 14.73 km was applied as the maximum observation radius for the visibility analysis in this study.

Analysis tools

Within the Aksu region, all military sites were represented as point features, with their geographic location, elevation information, and digital name information serving as the basic analysis parameters. After conducting the visibility analysis in ArcGIS, the results were imported into Python to further explore defensive relationship features using built-in analysis tools, including the following two steps: (1) visibility analysis. Using the visibility-line of sight tool, the visibility relationships between the various site points were determined. Meanwhile, the viewshed area of each site point was also obtained through visibility-viewshed tool. The overlapping area of the visible domains of two adjacent sites that cannot form a visual connection was the potential location range of the potential lost sites. The defense layout characteristics of the site complex were analyzed on the basis of existing sites, supplemented by presumed sites. (2) network analysis. In Python software, the numbers 0 to 153 was utilized to represent the existing sites, with 20 of the 164 existing sites having overlapping locations, while the numbers 154 to 187 were used for presumed sites. The visibility relationship between sites x and y was denoted by (x, y), and the visual network diagram was generated using the Kamada-Kawai (KK) algorithm. Combined with the evaluation metrics from the network analysis, the importance of the sites, local defensive relationships, and the overall network structure was further analyzed. The above methods and steps aimed to reveal the visibility relationships and defensive layout characteristics of TDMS in Aksu by incorporating ArcGIS, Python, and network analysis techniques.

Visibility analysis

In a military defense system, utilizing beacon signals to transmit military information is a crucial component. Beacon towers or lookout facilities within fortifications facilitate the observation of enemy movements and the reception and transmission of information by stationed soldiers. Analyzing the visibility quality of sites can determine whether the location of these sites were carefully planned^59,60. The visibility analysis methods employed in this study included two aspects. On the one hand, the line of sight analysis was used to determine whether there is visibility between observation points and target points. On the other hand, the viewshed analysis was applied to study the surface areas that can be observed from the observation points.

Line of sight analysis

Line of sight analysis tool is a terrain analysis method based on DEM. By establishing a line of sight between observation points and target points, it can analyze the interaction between the path and terrain obstacles. After inputting the DEM of Aksu and the relative height of the observation points, it is possible to determine whether the target points are visible. The principle is to establish a line of sight path towards the target point P, starting from the observation point O. The slope of the line of sight from O to P is denoted as α_oi. The analysis determines whether there is any surface point J between O and P with a slope α_oj > α_oi. If such a point exists, the target point P is not visible. Otherwise, it is visible. Through simulating all site locations in Aksu, a through-view relationship network between the sites can be generated to reveal the spatial distribution and interactive relationships of ancient human military activities.

Viewshed analysis

Based on observation points and terrain elevation, viewshed analysis tool is developed to calculate the visibility of each surface point within the visible radius. The generated viewshed range is the set of surface points that are visualized with the observation point. The viewshed tool in ArcGIS allows for parameter settings according to specific needs. In detail, the DEM of Aksu was input in the “in raster” parameter, with frequency selected as the visibility analysis type. This type can output the visibility of each surface point in the region with respect to the observation points. Then, the elevation of the observation point was imported in the “observer elevation” parameter, and the vertical height of the observation point relative to the surface was entered in the “observer offset” parameter. The “outer radius” parameter defined the maximum distance of visibility, i.e., the maximum observation radius of the observation point, or 14.73 km as calculated in “Method” section”. Thus, the analysis result was the set of surface points visible from each observation point, within a circular area centered at each observation point with a radius of 14.73 km. It represented the visible range of the observation point.

The locations of the 34 presumed sites mentioned in “section 2.1.2” were derived through visibility analysis. The inference principle was as follows: Since the maximum observation distance of an observation site is 14.73 km, when the distance between two adjacent sites, A and B (a city site or a beacon tower), exceeds 14.73 km, these two will not have a direct line of sight. At this point, the visibility ranges of sites A and B can be overlaid. As illustrated in Fig. 3a, if there is an overlap in the visibility areas of the two observation sites, there is a possibility that a site C that has already disappeared may exist within it. As explained in Fig. 3b, when the distance between sites A and B exceeds 29.46 km, there may be two lost sites (denoted as sites D and E) between them. In this case, site D lies within the overlapping visibility area between A and E, while site E lies within the overlapping visibility area between B and D. Thus, a visible line of sight can be established between sites A and B. Using this method to compare the overlapping visibility areas of all adjacent existing site points, the approximate locations of potential lost sites can be inferred.

Fig. 3: Schematic diagram of inferring potential lost sites from the overlapping visibility areas of adjacent sites.

a The distance between sites A and B exceeds 14.73 km; (b) the distance between sites A and B exceeds 29.46 km.

Network analysis

Network analysis is a comprehensive method that integrates graph theory and optimization analysis, combining both empirical and mathematical characteristics^61,62. By treating individuals or entities in a system as nodes and their relationships as edges, a network is constructed. This approach is commonly used to find out the structure and relationships within complex systems. In the field of archeology, it provides a new approach to interpreting the relationships between sites and their environments, as well as between different sites^63,64. Combined with GIS-based visibility analysis, it is possible to reconstruct the social structure resulting from the interaction between ancient peoples and their environment. In this study, military site remains were represented as nodes in the network, while the edges connecting two nodes in the network represented the through-view relationship between the two sites.

KK layout conversion: generate network graph

KK algorithm is a graph layout method that uses a spring embedding approach to convert complex network layouts into a visual format. It not only facilitates the observation of the overall structure of the network as well as the connections and clustering between nodes, but also makes it easier to identify key nodes in the network, such as the center and intermediate nodes. For ancient military site visibility networks, using the KK layout can help to understand the relationships, importance, and overall structure between sites, providing a valuable enhancement to the presentation of defensive spatial networks.

Key evaluation merics for network analysis

Specific network evaluation metrics were also introduced in this study to described the attributes and structure of nodes and edges. For example, “centrality” can measure whether a node possesses certain important characteristics in the network⁶⁵. In order to state more succinctly, the remaining merics used, along with their corresponding formulas and principles, are listed in Table 2.

Table 2 Main evaluation metrics for network analysis

Results

Visibility relationship analysis of military sites in Aksu

Through-view relationships of existing sites

Figure 4 shows the through-view relationships of extant sites in Aksu. It should be noted that, based on this restoration result and combining the form and architectural features of each site, the 31 suspected Tang dynasity city sites in the cultural relics census data can be classified into 3 categories. First, sites that would result in communication disconnection, including SC1, SC4, SC5, SC10, SC11, SC13, SC15, SC17, SC18, and SC23. The visibility relationships of these sites with surrounding TDMSs indicated that their absence would lead to communication disruptions in that area. Therefore, they were used as nodes for subsequent network analysis in this study. Second, sites with similar structures and close visibility relationships, covering SC2, SC3, SC6, SC7, SC8, SC9, SC12, SC14, SC16, SC19, SC20, SC21, SC22, SC24, SC25, SC26, SC27, SC29, SC30, and SC31. Due to their highly similar structures and construction characteristics to TDMS, as well as their close visibility relationships with surrounding TDMSs, these were also included as supplementary nodes in the subsequent network analysis. Third, sites with unclear structural and construction characteristics as well as loose visibility relationships. There was only one such site (SC28) in Aksu, so it was removed in this study.

Fig. 4: Through-view relationships of existing sites (N = 164).

a Enlarged diagram of the through-view relationships of the western oasis sites; (b) Enlarged diagram of the through-view relationships of the eastern oasis sites. Map source: Xinjiang Uygur Autonomous Region Platform for Common Geospatial Information Services,Ðith review number 新S(2023)061 (https://xinjiang.tianditu.gov.cn/bzdt_code/bzdt.html).

Obviously, the through-view network of the study area can be divided into two parts, namely a dense region and a sparse region. In the southeastern oasis region (ancient Kuche Oasis, now the Kuche River-Weigan River Delta), the site distribution was dense and closely interconnected, forming a complex, intersecting communication system. In contrast, the site distribution in other regions was more sparse, with most sites isolated in the passes and lacking visibility connections with other sites. Only 5 sites form small local visibility networks (Zones A to E), with each network containing only 4 to 6 sites. The visibility network in the southeastern oasis region was relatively complete, with only two visibility breaks, i.e., between B2, B3, C50, and C1 to the east, and between C41 and SC25 to the south. But the lack of preserved sites in other regions made it difficult to form complete information transmission routes. Overlaying the Tang dynasty SR and ancient trade routes, it can be seen that the sites to the west of the Kuche Oasis were generally distributed along the SR, while the sites in the north were distributed follow the ancient paths across the Tianshan Mountains. It suggests a correlation between the distribution of sites and ancient transportation routes.

Hypothesis-based through-view relationships: after adding presumed sites

Figure 5 illustrates the layout of site through-view relationships based on the assumptions, with Supplementary Information detailing the process for the 34 presumed sites. Guided by the transportation routes, the visibility range of each site was calculated. Through analyzing the overlapping visibility areas of adjacent sites, the locations of lost sites were identified, resulting in a more complete visibility network.

Fig. 5: Hypothesis-based site through-view relationships after supplementation of presumed sites (N = 198).

a Enlarged diagram of the through-view relationships of the western oasis sites; (b) enlarged diagram of the through-view relationships of the eastern oasis sites. Map source: Xinjiang Uygur Autonomous Region Platform for Common Geospatial Information Services, with review number 新S(2023)061 (https://xinjiang.tianditu.gov.cn/bzdt_code/bzdt.html).

First, the visibility reconstruction in the southeastern oasis region. Based on the through-view analysis results, it is clear that the southeastern oasis area was the military core area. As explained in Fig. 6a hypothesized point PB1 was added in the overlapping visibility area of SC22, C17, C41, and SC25 on the southern side of the oasis. Additionally, a hypothesized point PB2 was also added at the disconnected location between B3 and B7 on the northeastern side. This addition helped to establish connections between the previously disconnected sites with the main visibility network, including C41 and SC25 on the southern side, as well as C49 and the surrounding beacon towers and garrisons on the eastern side.

Fig. 6: Inferred locations of lost sites based on visibility zones.

a Lost site speculation between B2, B3, B7, and C1; b Lost site speculation between SC22, C17, SC25, and C41. Map Source: Xinjiang Uygur Autonomous Region Platform for Common Geospatial Information Services, with review number 新S(2023)061 (https://xinjiang.tianditu.gov.cn/bzdt_code/bzdt.html).

Second, visibility restoration in the northern Tianshan Mountains region. There is a small cluster of sites in each of the Baicheng Basin and the small oasis at the outlet of the Baozhadun ancient road. Of these, the distance between cluster C34 and C35 as well as cluster SC28 and SC31 to the terminal site C79 in the southeast oasis region was approximately 35 km. These two site clusters were most likely to establish a connection with the oasis region through C79. Therefore, an additional hypothetical point, PB5, was added in the basin to establish a connection between C32 and C72 with C33 to the east. At the same time, three hypothetical points, PB3, PB4, and PB34, were supplemented between C34, C35, and C79. Additionally, two hypothetical points, PB32 and PB33, were updated between SC28, SC31, and PB3 to establish a connection between these two small site clusters and C79, ensuring that military information from the C81 pass, the C32 pass, and within the basin can be promptly transmitted to the core area. The remaining two small site clusters were located at the mountain passes in the northern part of Aksu (Zones X and Y). Due to the distribution of Zones X and Y along the foothills, the single-point visualization area was greatly restricted by the mountainous areas and was too far away from the core area, making it insufficient to add speculative points. As a result, these two site clusters have not been connected to the core area through hypothesized lines of sight.

Third, the visibility restoration situation in the central-western region. To the west of the Kuche Oasis, there were several visibility gaps along the linear distribution of sites along the SR. With C21 as the central node, this area can be divided into four routes, namely lines B24–C21, C21–C26, C21–C71, and C21–B44. According to the principle of overlapping visual fields, the presumptive points on the B24–C21 route were PB5–PB12, on the C21–C26 route were PB13–PB14 and PB30–PB31, on the C21–C71 route were PB15–PB21, and on the C21–B44 route were PB22–PB28. Thus the complete military information transmission route was recovered.

As a complementary, a total of 19 disconnected locations existed in Aksu. Among them, ten had one putative point added, seven had two putative points added, and two had five putative points added. From Fig. 5, it can be observed that the primary communication routes extended predominantly in the east-west direction. The site distribution of oasis regions was notably dense, especially within the resource-rich Kuche Oasis in the southeast, where the number of sites accounted for 61.0% of the total in Aksu. The Kuche Oasis and the Aksu Oasis were connected by a linear strip of military installations along the Tianshan Mountains and desert interface, forming a corridor (B24–C21 line). This corridor linked the two major oases and branches into three lines at its western end at C21 (C21–C26 line, the C21–C71 line, and the C21–B44 line), which led to three directions. Two transmission routes C81–C79 and C72–C79 in the Baicheng Basin, north of the Kuche Oasis and at the southern foot of the Tianshan Mountains, intersected with the two eastern routes (B1–C2 line and B4–C2 line) at C2, where they were connected to the site complex of the core area in the Kuche Oasis. Overall, it can be believed that the addition of 34 inferred points has established connections between previously disconnected sites and the main visibility network, presenting a more complete military information transmission system.

Analysis of the visibility network structure of military sites in Aksu

Figures 7 and 8 demonstrate the new visual network layouts formed after the KK diagram transformation. With a clearer and more intuitive structure, where nodes with a high degree of relatedness were located at the center of the network, and nodes with a lower degree of relatedness were located at the edges of the network. In the network diagram based on the extant site (Fig. 7), the originally complex and intersecting through-view relationships in the southeastern oasis area were clearly shown after the graphic transformation. It revealed that there were three local visibility networks (Zones I to III), presenting a stable structural feature. In the complex network relationships of this region, occasional communication disruptions at individual sites did not affect the overall functioning of the communication system. In addition, each of the five groups of sites in the sparse zone (Zones A to E in Fig. 7) also formed stable small ring-like structures. These five structures were independent and did not establish a visibility connection with the core areas (Zones I to III). Remaining sites were distributed in a scattered manner. However, as presented in Fig. 8, the hypothesized network diagram connected the scattered groups of sites in the sparse areas based on the existing and presumed sites,creating a complete visibility relationship network. The visibility relationship network can be mainly divided into eight sections, including Zones I to VIII. Except for Zones I and II, the visibility relationships in the others were relatively simple, with the center node of each small ring structure also clearer. It indicates that these areas were primarily simple linear communications, with the local small circular structures serving as military-centered areas on each line of communication. Among them, Zone III was divided into three parts, with Zone III–1 being the core of the entire network structure and point C2 being the central site in this zone. In detail, Zone III–1 connected to Zone I and II via point B14, to Zone VI via point C79, to Zone III–3 via point C78, and to Zone III–2 via point B8. Meanwhile, Zone III–2 and III–3 were simple linear structures. Zones IV, IV–1 and IV–2 were both simple linear structures with no center node, while Zone V was a serial linear structure. Zones VI, VII and VIII were simple linear structures with some small ring-like structures locally.

Fig. 7

Visibility network graphic of existing sites.

Fig. 8

Visibility network graphic after adding presumed sites.

Figures 9 and 10 show the network evaluation metrics for each site in the extant site network and the hypothesis-based network. Based on the comparison of the results of the two plots, the peaks of degree centrality both occurred in the intervals of points 36–54, 58–60, and 138–146, which coincidentally corresponded to the most densely distributed and complexly related portions of the sites in Zone I. It indicates that the addition of hypothesized sites did not significantly impact the overall network’s degree centrality. However, there was a noticeable difference in the betweenness centrality results. The peak values in Fig. 9b appeared at points 21, 57, 61, and 64–66. Among them, point 21 (B14) was a key node connecting Zone I with Zones III and IV. Points 57 (C10), 61 (SC18), and 64–66 (SC1, SC4, and B30) were intermediary points between Zones I and V. In contrast, the peak values in Fig. 10b were found at points 21, 35, 92–94, 155–156, and 158–162. Of the peak points that differ from the former, point 35 (B24) was the intermediary point between Zones I and V. Points 155–156 and 158–162 formed the traffic defense line in Zone V. In addition, the peak points 57, 61, 64–66 in Fig. 9b also had high betweenness values in Fig. 10b. It indicates that the overall connectivity of the network can be enhance by adding hypothesized points to the network, primarily by adding a few key intermediary points to the existing site network. Figures 9c and 10c display the closeness centrality of each site, which measured the efficiency of information transmission among nodes in a closed network. The peak values in the existing site network were observed in the five groups in the sparse areas (Zones I to V, while those were found in the areas that did not establish connections with the main network in the hypothetical network. It means that the information transmission efficiency was higher in the small ring structures. Additionally, it suggests that the hypothetical network had fewer independent small-ring structures, resulting in overall higher connectivity of the sites. The eigenvector centrality of each site was shown in Figs. 9d and 10d. The results after adding hypothetical sites were consistent with those from the existing site network. The peak of this metric was consistent with the peak area of degree centrality, but highlighted even more the importance of points 41–54, 58–60 69, and 138–143 in the network. These four groups of site points coincided with the densest site clusters within the southeastern oasis zone, i.e., Zone I.

Fig. 9: Evaluation metrics for the visibility network of existing sites.

a degree centrality; b betweeness centrality; c closeness centrality; d eigenvector centrality.

Fig. 10: Evaluation metrics for the hypothetical visibility network.

a degree centrality; b betweeness centrality; c closeness centrality; d eigenvector centrality.

Table 3 presents the network parameters for the existing site visibility network and the hypothetical visibility network. By evaluating the overall network, it can be observed that the network structure established under hypothetical conditions showed an increase in connectivity by 0.330, network efficiency by 0.017, network density by 0.022, and betweenness centrality by 0.053, compared to the network structure formed by existing sites. It illustrates that the former, as a visibility network, had a more advantageous structure. While the degree centrality value was 0.007 lower and the overall clustering coefficient value was 0.016 lower, indicating that the latter tended to be more reticulated and the former more dendritic. It also suggests that the former connected more nodes, forming longer communication lines. At the same time, its indicators of degree centrality and overall clustering coefficient were reduced due to the lengthening of the line.

Table 3 Comparison of overall visibility network evaluation before and after adding presumed sites

Discussion

Layout characteristics of the TDMS in Aksu reflected by the visibility relationship

Based on the results of the through-view relationship restoration of the existing sites, it was found that the site defense layouts in Aksu showed three types. First, a large number of sites were densely distributed within the southeastern oasis, and the complex intersectionality of the through-view relationship implies that the area was a military core. Second, the sites in the central and WR were distributed along the SR, with several interruptions in the line-of-sight connections between them. However, based on the direction of the ancient routes, it can be preliminarily concluded that these sites primarily served to guard the SR transportation routes. Third, a small number of sites were scattered along the southern foothills of the Tianshan Mountains in the northern region, where their function appears to have been to guard the mountain pass routes.

From the natural geographic features of Aksu and the relevant historical records of the Tang dynasty, it can be seen that the formation of these three types of spatial layouts was the result of adapting to local conditions. In other words, the Tang dynasty administrators made reasonable military deployment patterns based on the different natural environment conditions. First, the region with a dense distribution of sites in the southeastern part of Aksu corresponded to the ancient Kuche Oasis (Kuche River-Weigan River delta today). This area is one of the largest oases in Xinjiang and is also one of the lowest in elevation and the flattest in slope of Aksu. The oasis itself is not only flat and expansive with a dense network of water channels, but also fertile with abundant water and vegetation. As a result, during the Tang dynasty, this oasis area was extensively used for military garrisoning and farming to meet the needs of troop stationing and logistical supply for the entire Anxi region. Historical records provided relevant documentation. For example, the Tangliudian (唐六典) mentioned the situation of military farming in the Anxi region: “There are twenty garrisons in Anxi” and “each large garrison covering fifty qing (顷), and the smaller ones covering twenty qing” ⁶⁶. It implies that the establishment of large and small military farms based on geographical constraints aligned with the dense distribution of major sites and surrounding smaller sites within the southeastern oasis⁴. In addition, the situation of military garrisons in this region was described in the Jiutangshu (旧唐书), “There were farmlands to supply provisions, and herders to raise horses²,” aligning with the natural environment and site conditions of the eastern and western parts of the oasis. That is, the alluvial plain of the Weigan River basin in the western part was suitable for large-scale military farming. Under the Tang dynasty’s system of military garrisons, where soldiers and farmers were integrated, this area became the primary choice for the garrison of the Qiuci Army (the foremost of the four armies managed by the AP). In the eastern part of the oasis, near the Tangwang city (唐王城, C19), there were large areas of grass lakes, which were not only suitable for agricultural garrisoning but also ideal for pastoral farming. Combined with the ruins of the largest horse breeding site in the Qiuci region within the Tangwang city, it can be inferred that this area served as a large military horse breeding base during the Tang dynasty¹². It can be concluded that during the Tang dynasty, the site clusters in the southeastern oasis area formed two major garrison areas, i.e., the western area dominated by military garrison and farming, and the eastern area characterized by military herding and farming. At the same time, on the northern edge of the oasis, the highest military and administrative center (AP, site C2) of the Anxi region was established to oversee the various large military bases within the oasis. This layout pattern resulted in a dense concentration of military sites and a complex visibility network in the oasis, which was consistent with the visibility relationship reconstruction results for the southeastern oasis region shown in Fig. 4.

Second, the distribution of sites along the SR and the Tang dynasty’s official roads in the central and western parts of Aksu was relatively sparse. According to the previous studies^21,67, these roads were located at the junction of the southern foothills of the Tianshan Mountains and the oasis plains, which ensured the resupply of passing troops and avoided the impact of seasonal floods and riverbed changes in the oasis floodplains. Simultaneously, the control of roads was the foundation of the Tang dynasty’s central government in managing the WR. As documented in the Xintangshu (新唐书)², a large number of beacon towers and garrison forts were established along the transportation routes connecting the Qiuci garrison to other military towns. The AP was the central node on this route, from which five transportation defense lines extended outward, including the Yanqi-Qiuci road, the Qiuci-Gumo (姑墨) road, the Gumo-Suiye road, the Gumo-Shule road, and the Gumo-Yutian road (Fig. 4). Among them, C21 was an important transportation hub. The Qiuci-Gumo road branched into three after reaching C21, leading to the three major military garrisons of Suiye, Shule, and Yutian². Military remains along these routes were relatively few and poorly preserved, but generally linear in distribution. While ensuring the security and smoothness of the transportation corridors, they also formed a strong defensive barrier along the eastern and western borders of the AP. Under this defense layout, these transportation routes were not only a key economic trade route of the SR, but also an important military supply line for controlling the WR.

Third, in the northern part of Aksu, a small number of sites were scattered along the southern foothills of the Tianshan Mountains. Geographically, the towering and continuous Tianshan mountain range divided the WR into two parts, including the southern and northern areas. Qiuci was located adjacent to the southern foothills of the Tianshan Mountains, with valley formed by multiple rivers in the mountain as a natural channel connecting the ancient Qiuci with the nomadic tribes on the northern slopes of the Tianshan Mountains. In order to prevent tribes like the Western Turkic Khaganate from crossing Tianshan Mountains through these valleys and invading Qiuci, the Tang government built fortifications along the valleys and divided them into four roads for strategic defense.From east to west, they were the Gongyue (弓月) road, the Baozhadun (包扎墩) ancient road, the Xiata (夏塔) ancient road, and the Biedieli (别迭里) ancient road (Fig. 4). Among these, the Gongyue road was located to the northeast of the AP, connecting to the northern grasslands where nomadic tribes such as the Youludusi nomadic tribes. The Baozaduan ancient road lied to the northwest of the AP and served as the most direct route to the Ili River basin on the northern slopes of the Tianshan Mountains. The Xiata ancient road was located to the west of the Baozaduan ancient road, running south to the Baicheng basin in the Qiuci and north to the Tezhao basin, where nomadic tribes were active. The Biedieli ancient road was situated at the intersection of the Qiuci and Suiye military garrisons, and it crossed the Tianshan Mountains, providing a route to Suiye. These four ancient roads all had Tang dynasty military remains at the mountain passes along the southern foothills of the Tianshan Mountains, confirming the defensive layout pattern where the Tang dynasty strategically guarded the mountain passes in this region. Combined with the natural geographical barrier of the Tianshan Mountains, these military facilities formed a solid defense line along the northwestern border of Anxi region. The Baizadun and Xiata ancient roads both had small-scale military site groups located at the mountain passes on the southern foothills of Tianshan Mountains, as well as in the Bayicheng basin between the two roads (Zones A and B in Fig. 4, and Zone Y in Fig. 5). The through-view relationships of the existing sites indicate that these three groups of sites only had internal visibility connections and have not established connections with the AP. However, from a military defense perspective, enemy information from the mountain passes needed to be passed on to higher-ranking military centers in the rear. It can be inferred that there may be sites that no longer exist in this area, and their strategic importance should not be overlooked.

The speculation on the location of the lost sites was based on the theory of three types of defense layouts, which was obtained from historical data and through-view relationships. By incorporating quantitative analysis methods that considered the overlapping visibility ranges of adjacent sites, the accuracy of the speculations can be enhanced. Taking the speculation of lost sites between beacon towers B2, B3, and B7 as an example, the sites B2, B3, C49, and C50 on the right side of the disconnection position formed a small military defense area consisting of the Chian Garrison city (赤岸守捉城, C49) and surrounding beacon towers, as displayed in Fig. 6a. On the left side, it led to the AP through B7, B8, and B9. This route was the official road from the Yanqi military town to the AP², and there should not be any communication interruptions under normal circumstances. Since the left and right neighboring sites at the disconnection had exactly overlapping portions of their visibility ranges, a speculative point PB2 was added within the overlap zone to make this communication line connected. The connected line can more comprehensively present the military deployment and defensive layout of the Tang dynasty. Furthermore, this approach has been used to speculate on all possible locations of disconnections in the study area, which in turn resulted in a hypothesis-based network of through-view relationships (Fig. 5). Its military defense layout was also highly compatible with the historical records documented. Based on the layout of the five transportation defense lines reflected in the existing sites, the defensive orientation of each route was clearly presented, specifically the transportation defense lines from the AP to the other four major military towns. Concretely, C79 (Yanshui pass) was the first pass northwest of AP¹⁴. After crossing it, the line entered the Baicheng basin in the southern foothills of the Tianshan Mountains. There were also two defense lines in this direction. The C32–C79 (Baicheng basin defense line) started from the pass on the southern slope of Tianshan Mountains and ran along the Baicheng basin to C79, while the C81–C79 (Baozadun ancient road defense line) extended southward along the Baozadun ancient road on the southern slope of Tianshan Mountains to C79. These two defense lines guarded the mountain passes in the northwest direction of the AP and transmitted military information quickly to AP. In addition, there were also five military defense facilities located at the mountain pass of the Xiata ancient road (Zone Y in Fig. 5). Existing information indicates that this location was a defense deployment to guard the mountain pass⁴, and military intelligence from the pass should be relayed to a higher-level military center. However, due to insufficient supporting information, it is currently impossible to reconstruct the transmission route at this location.

Characteristics of the local military defense relations reflected in the visibility relationship

By analyzing the defensive relationships in each area based on the visibility network diagram and the results of various evaluation indicators, It was found that the characteristics and complexity of the local network structure were highly correlated with the military defense strategies in each area, which can be summarized into two types. First, the military defense arrangements at the mountain passes along the southern foothills of the Tianshan Mountains in the northern region, as well as those along various transportation routes in the central and WR, were relatively simple (Zones IV to VIII in Fig. 8). The main purpose of the defense arrangements in these areas was to control the mountain passes and transportation routes, ensuring smooth passage, and thereby maintaining control over the vast and sparsely populated regions of the WR². The visibility networks in these areas primarily exhibited linear structures, with some small ring structures in the linear networks indicating military centers along these transportation routes. Each of these small ring structures corresponded to significant city sites recorded in Tang dynasty historical sources. For example, within Zone VII, C28 corresponded to the Gudagan city (故达干城), while B50 was identified as the Jizhuo old relay station (济浊馆, also known as Heping old relay station)¹³. The intersection point C21, where the western route splited into three, was known as Bohuan garrison city. In addition, there was a relatively large city site, C25. Although its historical counterpart cannot be identified from existing records, it was presumed to be a city site of higher military rank based on the scale and shape and the layout of a circle of beacons surrounding it. Second, the visual network in the southeastern oasis area formed a complex, large-scale mesh structure, with intersecting visual relationships that enhance the overall stability of the region¹⁵. In case of an emergency, the beacon towers and garrison cities can respond quickly, allowing the stationed troops to make defensive arrangements based on the warning information. It confirmed the importance of this oasis area for military defense. In this complex large network, there were also multiple local military defense centers. For example, the middle Zone III–1 surrounded by Zones I, II, III, and IV in Fig. 8, was the core defense area of Qiuci in the Tang dynasty. The sites within this zone formed the first layer of the defense circle of the AP. Outward, it formed a second layer of defense circle with B14, B17, C80, B8, B13, C79, and C78. At the same time, B8, B14, C78, and C79 were intermediate nodes connecting the military base Zones I and II, as well as the forward defense Zones III and IV. In the small defense group of Zone III–2, C49 was the higher military ranked Chian garrison city, which was surrounded by multiple beacon towers for protection¹⁴. The above sites demonstrate that the method of inferring important ancient military centers through the structure of local centers in a network is feasible, and can be used in studies of defense layouts in other regions.

The role of each site in the visibility network can be evaluated by four centrality metrics, including degree centrality, betweenness centrality, closeness centrality, and eigenvector centrality⁴⁴. Sites with high degree centrality and eigenvector centrality had significant local influence and collaborative defensive roles in the network, and can be considered as regional defense centers⁴⁵. In both the extant site network and the hypothesis-based site network, such sites were concentrated in the southeastern oasis zone, as outlined in Figs. 7–10. The peak points of betweenness centrality, B14 (point 21) and B24 (point 35), served as crucial transit hubs connecting the AP with the garrison areas in the southern oasis and the three major military towns to the west, playing a key intermediary role in the transmission of military information. The high betweenness centrality of the nodes in Zone V (points 92–94 and 156–162) established on the basis of the hypothesis suggests that this section of the transportation and defense line was the key to connecting the core base of the AP with the other three major remote military towns. To ensure effective governance of the three major military towns in the west by AP and to maintain control over the middle road of SR, it is essential to ensure the smooth operation of the defense line in Zone V. A higher closeness centrality value of a site indicates that military information has the shortest overall path through this site to the entire network⁶³. In the hypothesized network, the number of such sites decreased. It is evident that after adding the inferred sites, the number of sites that cannot connect to the main visual network decreased. In summary, it can be seen that the addition of the presumed sites has increased the size of the network and further improved its structure. The status of key sites within the network has not been diminished. At the same time, according to the comparison results of the overall network evaluation (Table 3), the visual relationship based on the assumptions was primarily reflected in the overall enhancement of connectivity and closeness. The overall intermediary role of the site points has became more apparent, facilitating the transmission of information over longer routes. Moreover, the status of important central nodes remains unchanged. Whereas the hypothesis-based network had a small decrease in overall degree centrality and clustering coefficient compared to the extant site network. It indicates that the overall communication paths for information were longer and the overall efficiency was lower, which is consistent with the limitations of information transmission over longer distances in reality. Overall, the hypothesized visual relationship network can be served as a complement to the existing site network, providing a more complete representation of the Tang dynasty’s military defense layout in Aksu without altering the overall defense facts.

TDMS in Aksu represents the Tang dynasty’s governance achievements in the WR, and is an important part of China’s ancient military heritage. In order to raise awareness and promote preservation efforts, this study established a database of 164 existing sites in Aksu. Utilizing GIS-based analysis tools and network analysis methods, the military defense deployment of the Tang dynasty in the area was then explored. Additionally, the locations of potential lost sites were inferred to supplement and verify the defense layout characteristics. Despite some unavoidable errors caused by the disappearance of some sites and changes in the social environment, this study was still able to roughly reveal the characteristics of the Tang dynasty military defense layout in Aksu. The main findings are as follows.

• (1) TDMS in Aksu were centered around the AP (C2) and were divided into three major defense zones, namely the northern Tianshan Pass defense, the central-western ancient SR route defense, and the southeastern oasis garrison defense. Based on geographical conditions, differentiated defense strategies were implemented in these three defense zones. The northern defense zone took advantage of the natural geographic barrier of the Tianshan mountain range, using a small number of military structures to control various passes. In the central-western defense zone, military defense lines were established along the ancient SR to facilitate the control and protection of this important trade and transportation route. In the southeastern defense zone, the abundant natural resources of the Qiuci Oasis were fully utilized. Internally, it was divided into military garrisons, agricultural settlements, and pastoral zones, reflecting meticulous planning for both military defense and resource management.

• (2) Due to the different defense strategies, the visibility network structure in Aksu exhibited two types, including a simple linear network structure (in the northern and central-western defense zones) and a complex large-scale mesh structure (in the southeastern defense zone). The main functions of the former were front-line defense and transportation security, while the latter was a rear military garrison base. Additionally, under the overall defense layout, there were multiple local military bases centered around significant city sites, forming small ring-like structures in the linear networks and central areas in the large mesh-like structures. These various zones were connected by intermediary sites, ultimately forming a complete military communication network.

The defense system layout and defense relationship characteristics revealed by this study have significant academic value for further exploration of the military defense in the Tang dynasty’s WR. However, there are still limitations in the data sources for this study. First, visibility analysis tool was used to infer the potential locations of lost sites. While it can provide some insight into the overall defensive layout of the sites, more precise site information is still needed to refine the analysis of the local defense relationships. Second, the database contained 31 city sites that cannot be identified as belonging to the Tang dynasty. Existing official archeological survey information is suspected to be dated to the Tang dynasty through the location of the site, the shape of the building and the surface remains. Based on these, the visibility relationships between these sites and surrounding TDMSs were incorporated to infer that these sites date back to the Tang dynasty and then be used as supplementary data in the study. Future research will require more archeological excavations and in-depth studies, combined with a wider range of methods and technologies, to further validate and refine the conclusions, as well as to promote the development of comprehensive protection strategies.