Historical background

The tenth to thirteenth centuries AD were a dynamic period in the history of Central and East Asia. The zenith of nomadic power in Eurasia culminated in the early thirteenth century with the rise of the Mongol Empire, the largest land empire ever established (Biran Reference Biran2007; Fitzhugh et al. Reference Fitzhugh, Rossabi and Honeychurch2013). The two dynasties most often associated with construction of the medieval wall systems in question are the Khitan-Liao (907–1125) and the Jin (1115–1234), both of which were established by nomadic or semi-nomadic groups. By the end of the first millennium AD, the Khitan people had conquered large parts of Manchuria, eastern Inner Mongolia and the Mongolian Steppe to form the Khitan-Liao Empire (Figure 2A). The Liao Dynasty combined Chinese and nomadic elements and was a formidable enemy of the Northern Song (960–1127), the reigning Chinese dynasty to the south. During the early twelfth century, the Jurchen tribes of Manchuria—previously vassals of the Liao—united and quickly destroyed the Liao and pushed the Song southward. As the Jin Dynasty, they established rule over all of north China until, by the mid thirteenth century, it was conquered by the Mongol armies (Figure 2B).

Figure 2.

Maps of East Asia and Mongolia during the eleventh (A) and twelfth (B) centuries AD. The route of the northern wall line is marked in black on the two maps (figure by the authors).

Although the Liao and Jin Dynasties, along with the contemporaneous Northern and Southern Song dynasties, kept extensive historical records, none of those histories or related documents mention the construction of these wall systems. This is perhaps surprising given the massive undertaking that would have demanded a large workforce over a long period of time.

The Chinese historian Wang Guowei (1877–1927) suggested that the wall systems were built by the Jin Dynasty over a short period of time, in their attempt to repel the invasion of Chinggis Khan's Mongol armies (Wang Reference Wang1921). In Western scholarship, Lattimore (Reference Lattimore1963: 5) attributed the walls’ construction to the early Jin Dynasty, associating it with a change in administrative policy from the Liao Dynasty's active involvement in the Mongolian Steppe to the retreat eastwards and defensive policy adopted by the Jin. Both of these hypotheses, however, seem to refer mainly to the southern wall lines, which resemble the known historic borders of the Jin (compare Figures 1 & 2B). Indeed, the southern wall lines continue to be assigned to the Jin period (Jing Reference Jing2006; Franke Reference Franke, Franke and Twitchett2008: 250; Jing & Miao Reference Jing and Miao2008; Sun Reference Sun2010), although some scholars have recently challenged this, attributing at least the westernmost parts of the southern wall lines to the Western Xia Dynasty (1038–1227) (Figure 2; Kovalev & Erdenebaatar Reference Kovalev and Erdenebaatar2010).

Historical analysis suggests that the northern wall line, on which our research focuses, was located beyond the borders of the Jin Dynasty, but within the area controlled by the Liao (compare Figures 1 and 2A) (Tan Reference Tan1996; Franke Reference Franke, Franke and Twitchett2008: 236–37; Ma Reference Ma2013). Thus, our working hypothesis concerning this northernmost line (but not necessarily for other parts of the medieval wall system) is that it was planned and built by the Liao Dynasty in order to meet specific strategic objectives. The presence of the Liao in the region, from at least the eleventh century, is undisputable, and some records suggest the existence of fortifications in this region (Ma Reference Ma2013), although Liao official history (Liaoshi 遼史) does not mention the construction of a substantial wall in these northern territories.

Archaeological research on the medieval ‘Great Walls’

Surprisingly little archaeological work has concentrated on this medieval wall system, particularly the northern section. A survey of the different southern sub-lines in China has been conducted (Sun & Wang Reference Sun and Wang2008; Sun Reference Sun2010: 142–46), and the reconstructed wall line appears in the Chinese historical atlas (Tan Reference Tan1996). This survey work has documented the dimensions of the aboveground remains, tracing and mapping the visible wall lines. In some cases, the remains of auxiliary features, such as towers and camps, are also documented (e.g. Jing Reference Jing2006; Jing & Miao Reference Jing and Miao2008).

The northern line of the wall has received far less research attention. Early descriptions of this wall and its associated features date back to eighteenth-century European travellers (see Lunkov et al. Reference Lunkov, Kharinskii, Kradin and Kovychev2011). More recently, Chinese archaeologists have carried out sporadic documentation on parts of the northern line, while Russian archaeologists have conducted more extensive field research on the part of the wall located in south-eastern Siberia and in Mongolia. In the 1950s, Kiselev (Reference Kiselev1958) dated the northern wall to the eleventh to twelfth centuries AD based on pottery collected from the surface of a circular structure located to the south of the wall. Later work includes the Soviet-Mongolian expedition in the late 1980s, which excavated sections of the wall in north-eastern Mongolia (Chichagov et al. Reference Chichagov, Chichagova, Cherkinskii and Avirmid1995). The Russian archaeologists Kirillov and Kovychev (Reference Kirillov and Kovychev2002) mapped and conducted test excavations on Siberian sections of the wall and associated forts (for a more recent detailed analysis of these Siberian sections, see Lunkov et al. (Reference Lunkov, Kharinskii, Kradin and Kovychev2011)). The consensus among Russian and Mongolian archaeologists is that the northern wall should date to the Liao period. Its function and its location within Liao territory, however, have so far not been addressed.

Current field research on the wall and associated structures in north-eastern Mongolia

At the outset of our project, we used high-resolution satellite imagery to identify systematically the line of the wall and any structures associated with it. This remote-sensing approach revealed that the structures along the wall are arranged in distinct clusters. We have traced the entire length of the wall, which, at 737km, corresponds well with previous estimates (Lunkov et al. Reference Lunkov, Kharinskii, Kradin and Kovychev2011; Ma Reference Ma2013). Along the wall line, we identified 72 individual structures arranged in 42 clusters (see Table S1 in the online supplementary material (OSM)). This is far more than the few structures described in even the most detailed published descriptions (e.g. Baasan Reference Baasan2006; Lunkov et al. Reference Lunkov, Kharinskii, Kradin and Kovychev2011; Ma Reference Ma2013). The clusters of structures are spread more or less evenly along the wall line, with distances between clusters varying from 8–29km (with one irregular 48km interval where the wall crosses the Argun River). We therefore suggest that travelling between most clusters on horseback, using ox-carts or even on foot would take no more than a few hours.

Fieldwork conducted in 2018 focused on exploring the wall line and a selection of structure clusters, located in Dornod Province. This collaborative project involved archaeologists from the Hebrew University, the Mongolian Academy of Sciences and Yale University, and consisted of three components:

1. 1. A systematic pedestrian survey of areas around a small section of the wall near cluster 27.

2. 2. A field survey, mapping and drone photography of 16 structures (in addition to the three structures of cluster 27).

3. 3. Vehicular reconnaissance and observations along the wall to its westernmost terminus.

The intensive survey was based on a well-established protocol previously used in north-east China (Shelach-Lavi et al. Reference Shelach-Lavi, Teng, Goldsmith, Wachtel, Ovadia, Wan and Marder2016). Two groups were each assigned a different survey tract marked on a high-resolution satellite image. As our predominant interest lies in non-sedentary societies and small communities in the vicinity of the wall, we anticipated that some sites under investigation would be small, with low densities of surface artefacts. During the survey, team members therefore walked slowly in a line, at 20m-interval spacing, carefully observing the ground for single artefacts. The basic unit of data recording was surface collection, established when the survey line encountered a threshold of three artefacts spread no more than 50m from each other (although when only a single artefact was found it was also recorded as a ‘findspot’). When artefacts were encountered by the survey group, the line halted to allow for intensive surface inspection. The boundaries of the artefact distribution (i.e. a collection unit) were marked on the printed satellite image and recorded with a GPS device. To ensure high resolution, we kept collection units small (0.67ha on average), and if artefacts continued to be found towards the periphery of an existing unit, we proceeded to open a new collection unit. This method is both efficient and accurate, as it does not depend on the subjective definition of a site, and the recording of the units is very accurate.

We intensively surveyed a total area of 16.8km² across two separate but nearby quadrants (Figure 3, left). The first quadrant consisted of a larger survey area around cluster 27, proceeding from that cluster towards the wall line. The second quadrant comprised a rectangular area along the wall line and farther to the east (Figure 3). In total, we made 36 collections of artefacts, and the accumulated size of the collection units was 24.4ha. The distribution of collection units and artefacts clearly shows that the only major concentration of artefacts is located around the structures comprising cluster 27. Here, we recovered large quantities of pot sherds and lithic artefacts (Figure 3); other collections usually contained one or two stray objects for which we recorded the findspot.

Figure 3.

Survey results: left) quadrants surveyed, and the location of the collections and features; right) collection units made around cluster 27. The size of each circle represents the density of artefacts found in the collection; top right) distribution of potsherds; bottom right) distribution of stone artefacts. The area of each collection unit is marked with a black line and the density of artefacts found within them is indicated by the colour (figure by the authors).

The most common finds were hard-fired greyware sherds with punctate decorations. Mongolian archaeologists and others have identified these as Liao/Khitan ceramics (Figure 4; Makino Reference Makino2007). Our assemblages correspond well with those from previously published archaeological surveys and excavations along the wall in Mongolia and Russia (Kiselev Reference Kiselev1958; Baasan Reference Baasan2006; Lunkov et al. Reference Lunkov, Kharinskii, Kradin and Kovychev2011). Flint artefacts, including tools, flakes and cores, were also commonly encountered. The concentration of stone artefacts was especially dense at some collections near cluster 27, perhaps suggesting the location of Early to Middle Holocene sites (Janz et al. Reference Janz, Odsuren and Bukhchuluun2017).

Figure 4.

A typical collection of ceramic fragments found during our pedestrian survey from areas around cluster 27 (photograph by the authors).

In addition to the systematic survey, we also surveyed in and around the structures at clusters 1, 2, 7, 12, 22, 23, 24, 26 and 28. Finds at these clusters were very similar to those at cluster 27. It seems clear that these clusters were centres of human activity probably contemporaneous with the wall being in use. The rectangular structures, for instance, could have functioned as bases for people—soldiers, perhaps—who operated the wall system, and the circular structures may have been used as animal corrals. These hypotheses, however, require further research. The scarcity of finds in areas away from the structures, including near the wall line, suggests that these places were not occupied and were subject to minimal human activity.

A metal detector was used during our surveys to find metallic artefacts, particularly coins, that would allow us to date more accurately the use of the wall system. Although we recovered some small artefacts, such as nails and buttons, no coins were found. While the ceramics we recovered during the surveys are often dated to the Liao period (907–1125), they may well be from slightly later periods. Efforts to refine the ceramic typology using comparative techniques, as well as absolute dating of contexts associated with such ceramics in other parts of Mongolia, are underway (Makino Reference Makino2007: 32–37 & 71; Shiraishi & Tsogtbaatar Reference Shiraishi, Tsogtbaatar, Bemmann and Pohl2009; Park et al. Reference Park, Honeychurch and Chunag2019a).

Remains of the wall and accompanying structures

Although the wall is poorly preserved in some places, drone photography clearly shows that it is a continuous structure. The better-preserved sections survive to 1m above ground level. The current width of the wall is approximately 10m, some of which may be attributed to collapse and/or erosion of upper portions. This corresponds to Lunkov et al.'s (Reference Lunkov, Kharinskii, Kradin and Kovychev2011: 105) observations of the Russian section of the wall. Drone photography suggests the presence of a narrow ditch on the northern side of the wall, as evidenced by thicker vegetation compared to the southern side (Figure 5). In some places, especially on the westernmost parts of the wall, we identified structural openings (near cluster 35 for example) that are also visible on the satellite images. While it is possible that some of these openings are modern, we hypothesise that most represent ancient gates. This theory will be explored in future excavations.

Figure 5.

The wall line near cluster 12 (drone photograph by G. Shelach-Lavi).

We identified four major types of auxiliary structures through field survey and satellite imagery: large rectangular/square structures (n = 42) with an average enclosed area of 1625m²; small rectangular structures (n = 11) ranging between 265 and 578m² in area; mini rectangular structures (n = 6) with estimated areas between 10 and 50m²; and circular structures (n = 16), ranging in size from 4185–20106m² (Figure 6). Lunkov et al. (Reference Lunkov, Kharinskii, Kradin and Kovychev2011: 116) suggest that these structures usually appear in pairs. Our survey of all the structures, including those located on satellite images, however, shows that the large rectangular structures are focal features around which other features sometimes appear to cluster. These large rectangular structures can appear alone (n = 24) or together in a cluster with other types of structure (n = 18) (Table S1).

Figure 6.

Clusters 24 (left) and 23 (right) (drone photographs by N. Doron).

The large rectangular structures are made of earthen walls that survive to a greater height than the linear wall, in some places preserved to a height of 2m. In most cases, the layout is approximately square, with the corners usually oriented to the cardinal directions. The width of the walls at the corners of the best-preserved structures are wider, suggesting that they may have supported additional structures on top of them. Occasionally, an opening is present in a structure's wall that may have served as the entrance. Furthermore, we often identified depressions circumscribing the walls. These most likely represent a surrounding ditch from which soil was excavated during construction of the earthen-walled features (Figures 5–6).

The 11 small rectangular structures morphologically resemble their larger counterparts. Some of these small structures are located inside the larger ones, while others are located outside, but in close proximity to the larger structures. The six mini rectangular structures were identified on satellite imagery but were not visited on the ground. They are located very close to the wall, sometimes adjacent to what may be openings in the wall. Unlike the other structures that are located south of the wall line, five of the mini-structures are located on the northern side of the wall. Their intimate association with the wall may suggest that these mini rectangular structures functioned as small shelters for gatekeepers, but again, this hypothesis requires archaeological investigation.

Sixteen circular structures were also identified. Their walls are shorter than both those of the rectangular structures and the linear wall, in most cases rising to no more than 0.5m above ground level. Eight of the circular structures enclose a rectangular structure, with independent circular structures most often located on the southern side of a rectangular structure.

At some of the cluster locations, we identified the putative remains of other types of installations. In cluster 27, for example, we recorded an additional small circular structure adjacent to the large circular structure. To the west of the rectangular structure, we identified shallow walls that seem to belong to elongated features. It is unclear whether these additional walls are part of the same cluster and are contemporaneous with the rectangular and circular structures in cluster 27.

Analysis of the wall system

Analysis of the high-resolution satellite imagery indicates the pre-planned nature of the wall system. Not only are the clusters of structures located more or less equidistant along the wall, but also the wall's geographic location suggests that it was strategically placed. The wall stretches between two mountain ranges, the Da Xingan range in the east and the Khentii Mountains in the west, essentially blocking the wide lowland area between these two mountain ranges (Figure 7).

Figure 7.

Map of the wall line and a geographic cross-section of it. For a Google Maps version and additional information, see: tinyurl.com/y4hbydw3 (figure by the authors).

All the clusters of structures associated with the western wall section, from 1–21 (excluding two unverified examples), comprise one large rectangular feature with no additional structures. In contrast, almost all the clusters at the middle section of the wall, from 22–35, have both rectangular and circular structures, along with occasional additional structures. Clusters 36–42, associated with the section of the wall to the east of the Argun River, comprise predominantly single structures, although we documented some cases of combined rectangular and circular features within a single structure (Figure 8 & Table S1).

Figure 8.

Graph of the types and sizes of structures identified along the wall line (figure by the authors).

The location of the individual clusters vis-à-vis their immediate environment provides another important aspect of our geographic analysis of the wall system. If the clusters were part of a military defence system, we would expect them to be located in elevated places that would allow control over their surroundings and provide clear lines of sight towards the wall and adjacent clusters. To test this hypothesis, we created a cross section of the topography along a line that connected all clusters. The results did not confirm the hypothesis, but rather demonstrated a preference for locating the clusters at lower elevations (Figure 9). These data are also supported by a viewshed analysis of clusters 26–34. Using GIS tools, we compared the areas visible from clusters 26–34 to the areas visible from eight points located close to those clusters, but on higher ground (Figure 10). The difference is very clear, with each cluster on average ‘seeing’ 137 862 pixels of 30 × 30m, while each of the comparison points ‘sees’ an area almost ten times larger (1 259 007 pixels). Raising the height of the clusters’ viewpoints by 3m—assuming that people could have used taller buildings for observation—had a negligible effect on the results.

Figure 9.

Topographic cross-sections showing the location of clusters (numbers) in relation to the topography of their immediate environment; Y-axis represents elevation above sea level (figure by the authors).

Figure 10.

Viewshed analysis of: top) areas visible from clusters 26–34 (in parentheses are the number of 30 × 30m pixels visible from each cluster); bottom) areas visible from eight comparison points.

While the location of clusters at lower elevations could be associated with the position of local streams, they are more likely to have been associated with roads and paths that connected areas to the north and south of the wall. This notion is supported by the location of modern local pathways, many of which cross the wall at, or very near to, the location of major clusters. Although modern, these routes were not planned or built by the Mongolian government. Instead, they represent the natural development of travel routes taken by the local nomadic population. Thus, we argue that they represent ‘logical’ paths probably taken by nomadic groups since the medieval period. Similar research on the development of informal road systems in other parts of Mongolia supports this argument (Burentogtokh Reference Burentogtokh2017: 185–86).