Introduction
Archaeological research has long emphasised the central role of communication and mobility in shaping social and cultural interactions. Networks of paths, routes and landmarks structured the ways in which communities connected with each other, exchanged goods and cultivated traditions. However, despite the substantial body of evidence for mobility in European prehistory, it remains challenging to reconstruct the axes of mobility or communication that once linked settlements and regions. Physical traces of roads are rare and often ephemeral; they are also frequently covered by subsequent landscape transformations, such as peat accumulation or coastal changes.
One promising approach is to identify communication routes indirectly using visible monuments as potential markers. In northern Central Europe, for example, Bronze Age burial mounds are often arranged in linear patterns, leading to the assumption that these arrangements reflect prehistoric routes. However, the reconstructed routes are mostly based on subjective connections between the mounds, which raises questions about the reliability of these interpretations.
The Elbe-Weser-Triangle in north-western Germany was selected as the study area based on the available archaeological data, which has been thoroughly documented across the entire region, and its distinctive, highly dynamic landscape. Rather than developing new methods, the focus of this work was on applying existing methods to a geomorphological and archaeological region that has been underrepresented in studies to date. Thus, the primary research question explores how methods adapted to the landscape and archaeological context can be applied to reconstruct communication and mobility networks with prehistoric burial mounds as the central focus.
Working area
The north-western Elbe-Weser-Triangle, bordered by the Geeste river to the south and the Schwinge river to the east, was chosen as the study area. Considering the landscape of the southern and eastern North Sea coast, this region is exemplary and offers methodological advantages as a case study. The boundaries formed by the Elbe and Weser rivers to the east and west, and the North Sea to the north, minimise edge effects (cf. Mennenga 2016: 171). Much of the area belongs to the district and city of Cuxhaven, meaning the data collected should be relatively consistent and therefore comparable. Much of this area has also been systematically surveyed and almost completely prospected on the surface, resulting in the identification of numerous sites (information provided verbally by Dr A. Hüser, Archaeological Heritage Management Service of the District of Cuxhaven). However, when mapping sites in the northern Elbe-Weser-Triangle, significant regional differences in the level of detail of the available data in the state monument preservation database (ADABweb) become apparent. For instance, there is a significant difference between the areas north of the Geeste and Schwinge rivers and the regions further south. Further subdivision within the Bronze Age is rare and finds from both the Single Grave Culture (SGC) and the Bell Beaker Culture are not recorded. For the former, at least, it can be assumed that they are listed as Neolithic sites. Due to these obvious differences in data quality between the regions, the northern working area, delimited by Schwinge and Geeste, has been chosen to allow for better comparability. At the same time, there is a low density of finds in the vicinity of these rivers. This occurrence is most likely linked to the development of widespread wetlands throughout the region, a process that began during the Neolithic period (Behre 2005).
Archaeological background
The Funnel Beaker Culture (TRB) has been documented in the region since 3500 calBCE (Mennenga 2017: 93; 2019). It was the first Neolithic culture to be documented in the Elbe-Weser-Triangle, although the Swifterbant culture, which has been documented east of the Weser, may have been present in the area earlier (Anscher & Raemaekers 2022: 162; Mennenga 2022a). Between 2800 and 2700 calBCE (Mennenga 2017: 93), as in many areas of northern Germany, a cultural change can be detected, and the Young Neolithic period begins (Raemaekers & Bärenfänger 2013; Strahl 1990: 320). In the Elbe-Weser-Triangle, which is defined by the southern border extending between the Wümme lowlands and the Este valley, as well as the Verderner Platte and the Harburg hills, with an altitude limit of approximately 60 m above sea level (Hofmann 2008: 4–6), the supra-regional SGC and Bell Beaker cultures have been documented from this period (Strahl 1990: 320). However, there is no evidence of the Bell Beaker culture in the AdabWeb for the area north of the Geeste and Schwinge rivers. As demonstrated in the research by Strahl (1990: 320) and Müller et al. (2010), the Late Neolithic period from 2200 calBCE onwards is associated with the spread of Danish flint daggers. The End Neolithic–summarizing both the Young and Late Neolithic periods (Strahl 1990: 324), and the following Bronze Age, spanning from approximately 1700 to 700 BCE (Hofmann 2008: 18), is the subject of the following discussion.
Settlements based on features from the Neolithic and Bronze Age in the Elbe-Weser-Triangle are very rare (Hofmann 2008: 51; Mennenga 2017; Mennenga et al. 2024; Nelson 2024). In many cases the function is reconstructed from surface finds, and detailed chronological classification is often impossible. Investigations by the local archaeological heritage agencies show that prehistoric multi-period sites are rare (verbal communication from D. Nösler, Archaeological Heritage Management Service of the District of Stade). This is very likely due to the significant changes that the landscape underwent throughout prehistory, which also led to changes in the usable and utilised space (Behre 2005). In micro-regions of the study area that have been examined in detail, it becomes clear how natural changes to the landscape necessitated the relocation of settlements (Mennenga et al. 2024). Human activity around settlements, such as deforestation and pastoral farming from the Neolithic to the Bronze Age (Wolters et al. 2024), can also cause significant redistribution of soil, leading to comprehensive changes in land use (e.g. Schwank 2017; Mennenga et al. 2024). Similar indications of significant landscape changes associated with changes in settlement and land use patterns have also been postulated for other regions (Haughton & Løvshal 2023: 1473). Therefore, it is assumed that settlements and activities were not in the same places throughout prehistory. Due to landscape changes and altered usage patterns, spatial relocation is very likely, as reflected in the archaeological data with few multi-period sites. Although prehistoric settlements and flat graves are rarely found, the landscape is characterised by above-ground monuments. Megalithic tombs and burial mounds are omnipresent. Megalithic tombs are primarily passage graves (Behrens & Mennenga 2025), and their construction boom can be traced back to around 3300 calBCE (Müller 2022: 259). Although 115 monuments of this type still exist in the study area (Behrens & Mennenga 2025), it is likely that many more have been destroyed, primarily due to modern land use (cf. Müller & Sjörgen 2022: 1215). People were buried in a wide variety of flat graves as well as in megalithic tombs. At the end of the TRB, cremations also appeared to the west of the Elbe (Kossian 2005). Flat graves are also known from the subsequent SGC, but are quite rare, possibly due to preservation conditions. In some cases, older graves were reused, with burials taking place in the chambers of megalithic tombs or burial mounds, or next to them (Behrens et al. 2022; Strahl 1990: 281–283). Stone cists were also present, but burial mounds with a diameter of 6–10 m and a height of about 1 m dominate the Elbe-Weser-Triangle, often occurring in groups (Strahl 1990: 280). Burial customs consisted almost exclusively of inhumation; cremations were extremely rare (Strahl 1990: 279). Burial mounds remained in use in the Early Bronze Age but were replaced by urn burials from the Middle Bronze Age onwards, some of which were embedded in existing mounds (Hofmann 2008: 51).
Like other regions of northern and northern Central Europe, the Elbe-Weser-Triangle is characterised by prehistoric burial mounds. Many of these monuments are still visible in the landscape today, despite often having been robbed and destroyed. However, it is difficult to estimate how many of these burial sites, which mainly date from the Young Neolithic and Bronze Age in the region, once existed.
Burial mounds as landmarks on prehistoric routes
In prehistoric times, when the exchange of physical objects and intangible goods could only take place through direct contact between individuals (Nakoinz 2012), physical connections played a decisive role in the exchange of knowledge, ideas, and resources (Løvshal 2013: 225). These connections could take the form of maritime routes, which have played an important role since at least the Bronze Age, as evidenced by iconography and trade relations (Segschneider 2007; Woltermann 2018; Holst & Rasmussen 2013: 109; Beuker et al. 2025). Other connections could take the form of terrestrial and limnic routes. Long-distance mobility also played an important role during the Bronze Age, as evidenced by finds in graves, deposits and the above-mentioned iconography (Løvshal 2013: 225–226). However, reconstructing these connections poses a major challenge as these types of sites are rarely preserved. Consolidated prehistoric roads are rarely documented in northern Central Europe, and wheel tracks are also rarely found; examples include Schneeverdingen in Lower Saxony, Oechlitz in Saxony-Anhalt (Burmeister 2018: 120) and Flintbek in Schleswig-Holstein (Mischka 2022: 103–108). Although cart tracks have been documented, it is generally assumed that transport, particularly for long-distance travel, relied on porters and pack animals rather than carts (Burmeister 2018: 133). This would presumably have required less extensive road construction. In addition to the limited evidence for terrestrial transport, the more frequently documented wooden trackways in bogs can also be considered evidence of prehistoric routes.
During the first comprehensive survey of burial mounds in the Elbe-Weser-Triangle, Aust (1956) noticed that some of them ran linearly through the landscape and seemed to form a network. Based on this observation, Aust (1956) hypothesised the existence of route axes along single burial mounds and fields of mounds. This hypothesis has not received significant criticism to date and has been corroborated by subsequent studies (Hesse 2010; Siegmüller 2015). In other regions of Bronze Age Northern Europe, it is also assumed that mounds, especially those arranged in a linear fashion, mark paths (Klassen 2014: 58–61), although this thesis has been criticised (Bourgeois 2013: 203). While the interpretation of these structures as bases for routes and communication axes is widely accepted in continental Northern Europe, it remains contested in the case of British monuments (Bell 2020: 164). This interpretation views the rows of mounds as territorial boundaries that delineate defined ritual spaces, thereby creating symbolic boundaries. Even in the case of this demarcating function, the mounds are understood to have had ‘channels of access’ and routes for ceremonial processions along their axes (Woodward & Woodward 1996). In the case of dispersed graves, these may also mark territorial expansion (Bourgeois 2013: 192). Many studies from other regions support the theory of paths or communication axes along linearly arranged mounds. There are various archaeological features that directly link burial mounds and paths. In Scandinavia, for instance, lines of graves have been found alongside watersheds, sunken roads and wheel tracks, as well as historical paths (Løvschal 2013: 227–228; Bang 2013). In parts of Denmark, it is thought that settlements and roads initially had a significant impact on the positioning of burial mounds. It was only with the construction of comprehensive structures from these mounds that they began to influence the location of routes (Johansen et al. 2004: 38). Thus, the construction of the mounds leads to the fixation of the paths in the landscape, and they can be seen as the accumulated spatial manifestation of a network of social interaction (Johansen et al. 2004: 38). Løvschal (2013: 232) suggests that paths or grave lines were created in their location for topographical as well as socio-ideological reasons – here, too, the fixation of mobility and the transformation of the landscape are postulated. Central junction points may indicate river crossings, for example, and can be places of intense interaction and social importance (Johansen et al. 2004: 38).
Johansen et al. (2004) considered the quantity and quality of finds along the axes and found that the mounds clearly reflected this spatial representation of social networks. The specific richness of finds along the reconstructed corridors around the main axes demonstrates their importance (Johansen et al. 2004: 50). In terms of a supra-regional interpretation related to the mobility that was deeply rooted in the culture and economy of the time, the concept of long-distance herding, with burial mounds serving as an indicator of general terrestrial movement patterns, is also considered in relation to the well-researched Danish region (Holst & Rasmussen 2013: 109). On a larger scale, the experience of walking (Bourgeois 2013: 203–204) and movement along an axis that does not necessarily equate to a road (Bourgeois 2013: 192) can carry further levels of social and cultural understanding (Løvschal 2013: 225). In summary, it can be said that there is a high probability of a connection between communication, mobility, and the paths that existed alongside burial mounds. These paths did not necessarily run directly alongside the mounds but could have been within sight of them. Thus, the mounds indicate corridors rather than defined paths (Klassen 2014: 58). The following text adheres to the definition in which the terms ‚paths‘ and ‚roads‘ are not used in the sense of clearly defined road or construction. Instead, roads and paths are understood as communication lines that such as as trails, tracks, and broad communication corridors (Løvschal 2013: 228). This also includes effects such as individual courses, dislocations and stagnant periods. The basis for the investigations is the connecting function between places and the physical movement along these objects. The aim is not to define clearly delineated paths, but rather to identify corridors through which exchange may have occurred.
As a connection between individual mounds can be assumed in the case of corridors, they can be considered interconnected elements, even if the individual mounds were not necessarily built at the same time. This raises the question of the continuity of linear mound arrangements, which can be considered in general terms through isolated case studies. In the Netherlands, for example, the linear arrangement of graves from the Neolithic period evolved into a more dispersed distribution during the Bronze Age. The existing structures continued to serve as reference points throughout this entire period (Bourgeois 2013). In Denmark, the relationship between younger and older mounds can also be demonstrated, suggesting that a joint consideration of the entire construction period is appropriate (Johansen et al. 2004: 37 & 41–42). It is hypothesised that the axes began forming in the Early Neolithic II period, gradually becoming denser, so the total number of mounds could be used to reconstruct paths (Klassen 2014: 61). In the working area, however, it is not possible to differentiate sufficiently between the periods in which the burial mounds were constructed, as the basis of the securely recovered archaeological material is insufficient. As well as comparing features from the construction periods of burial mounds, it is also possible to trace connections to older Middle Neolithic megalithic tombs in some regions (Klassen 2014: 58–59). Therefore, it can be assumed that the mounds along a corridor are related to each other, even if they were not necessarily build at the same time. All corridors or paths form the communication network, which, in its most comprehensive form, comprises the totality of all axes that have ever existed and those that can still be identified.
Data basis
Various data sources were used for the investigation. The analyses are based on information taken from the Lower Saxony Monument GIS (ADABWeb), containing 2,652 burial mounds (Figure 1). This database includes graves that still exist today, as well as those that have been destroyed based on older records. While the dating of most of the mounds is uncertain, it is reasonable to assume that very few date not from the period between the Young Neolithic and the Bronze Age up to the Iron Age. Of the 146 mounds that can be dated more precisely, only 12 date from the TRB period or are posterior to the Iron Age; these are excluded from the calculations. It should be noted, however, that there is temporal continuity within burial mound fields from the SGC to the Bronze Age (Strahl 1990: 312–314). The data used to compare the megalithic tombs is based on Behrens and Mennenga’s (2025) study. The natural environment, and above all the peatland coverage, are used to analyse the landscape context. Soil Science Map 50 and Original Peatland Distribution Map 50, both from the Lower Saxony State Office for Mining, Energy and Geology (LBEG), were used as a basis. The temporal depth of peatland development is decisive. Regarding peatlands, it should be noted that many of them began growing in the Mesolithic period and continued spreading until the Iron Age (see below). Additional Bronze Age and Neolithic sites were exported from AdabWeb for further analysis. Due to the involvement of different researchers and district archaeology departments in the study area, it cannot be determined whether definitions for findspot categories, such as settlements, were applied uniformly. As these more precise classifications play only a minor role, a detailed re-evaluation will not be undertaken at this point. Very few sites from the Young Neolithic period can be mapped in the study area, so they are not considered as a separate phase in the analysis.
Figure 1
Overview of the working area with all burial mounds of the Bronze Age and Single Grave Culture (landscape: Lower Saxony State Office for Mining, Energy and Geology).
A total of 2,572 Bronze Age burial mounds and single graves or undated sites of that kind are included in the study (Table 1). Assuming that there are paths alongside the mounds, a river that must be crossed in these cases can be considered a greater obstacle. Nevertheless, the defined study area can only be regarded as part of a much larger communication network. The following section, however, will take a closer look at this specific north-western area, which has a dynamic landscape consisting of Geest (see below), peatland and coastal marshland.
Table 1
Chronological classification of the burial mounds listed in AdabWeb in the working area. Entries crossed out were not taken into account based on the archaeological dating.
| DATING | NUMBER OF MOUNDS |
|---|---|
| Unknown | 1910 |
| 4 | |
| 2 | |
| Iron Age | 5 |
| Bronze Age – Iron Age | 14 |
| Bronze Age | 85 |
| Bronze Age–Neolithic | 3 |
| Single Grave Culture | 23 |
| 6 | |
| Neolithic | 17 |
| Multi-period | 2 |
The landscape and its transformation
Significant transformations have taken place along the north-west German coastline since the onset of the Holocene. Prehistoric communities inhabiting the area were confronted with dynamic landscape changes linked to the North Sea that profoundly impacted their activity areas (Schneekloth 1970; Behre 2005, 2008; Bungenstock 2005; Behrens et al. 2022, 2024). The ancient moraine landscapes with its predominantly Pleistocene sandy soil (Geest) of north-western Germany formed the basis for landscape transformations since the beginning of the Holocene. Due to a rapid onset of the climatic optimum at the start of the Holocene, the glacial ice melted quickly. This resulted in a significant rise in the sea level (Scheder et al. 2022). Around 6000 BCE, the North Sea advanced up to the present-day coastline. Over the following millennia, the rate of sea level rise slowed until phases of transgression and regression began to alternate from 3000 BCE onwards (Behre 2008). The rise in sea level after the last glacial period led to confinement of groundwater within the coastal zones. This process resulted in the development of vast fens and thus breaking up the coastal moraine landscape (Behre 2005). However, the process of fen formation did not occur simultaneously; it depended on the local soil moisture regime (Petzelberger et al. 1999). As a result of particularly beneficial conditions during the Atlantic period a greater number of raised bogs developed (Behre 2005). In some places, these grew nearly directly on the Geest hilltops, forming raised bogs on the sand (‘wurzelechtes Hochmoor’). In other locations, a transition from fens to raised bogs can be observed (Petzelberger et al. 1999). The oldest C14 date for the onset of raised bog growth in the Elbe-Weser-Triangle was obtained at Hohe Moor. Raised bog formation began here around 7500 BCE. However, the average age of raised bog bases in the Elbe-Weser-Triangle is around 3500 calBCE (Gerdes et al. 2003). Over the following millennia, the raised bogs continued to grow and spread over vast areas. In the Elbe-Weser-Triangle, bog formation peaked between 4000 and 2000 BCE. The profound landscape changes resulting from this consequently took place during the Neolithic and Bronze Age. It was not until around 1000 BCE that the expansion of the bogs began to stagnate (Behre 2005). The rise in sea level and the subsequent expansion of lowland bogs and raised bogs did not merely alter the natural environment. The narrowing and fragmentation of the Geest areas gradually restricted potential settlement and activity space. Accordingly, the alteration of the landscape suggests a shift in the habitable area, leading to a reduction in living space as early as the Neolithic period (Behre 2005). The Elbe-Weser-Triangle was therefore characterised by extensive, difficult-to-traverse wetlands, tidal mudflats and large-scale peatlands at that time (Behre 2008). The general reconstruction of post-glacial landscape changes for the Elbe-Weser-Triangle is only fragmentary, with the exception of a few case studies (Behrens et al. 2024). Previous studies have shown that the expansion of peatland did not begin at the same time across the entire Elbe-Weser-Triangle (Gerdes et al. 2003). Therefore, bog growth in the Elbe-Weser-Triangle began as early as the Mesolithic period. The growth rate increased significantly during the Neolithic period. From the Late Bronze Age onwards, only minor changes in spatial extent have been documented (Gerdes et al. 2003; Behre 2005, 2008).
Methods
Early attempts to reconstruct prehistoric routes were primarily based on the visual interpretation of spatial patterns, particularly the distribution of burial mounds. These were considered to be prominent features of the terrain that served as reference points for early routes. This approach was first employed in the early 20th century and has continued to be used in the study area since the 1950s (Aust 1956), as well as in more recent research (Siegmüller 2015; Hesse 2010). However, quantitative approaches offer more advanced methodological possibilities for systematically modelling spatial connections. Unfortunately, many of these approaches require input data of a quality that is not available for the north-western German coastal area. There is no geodata on which reliable cost areas for the study area can be based. The same applies to other potential factors (see, for example, Faupel & Nakoinz 2018; Groenhuijzen & Verhagen 2017), such as costs related to soil or vegetation, which cannot be reconstructed due to the unavailability of sufficient data. The lack of data is primarily due to the dynamic landscape and the limited number of reconstructions of prehistoric natural areas, except in a few small regions. It can therefore be concluded that the recent landscape, which is covered by peatlands and marine sediments, cannot be used as a basis for modelling, due to the massive changes since prehistoric times. Although the general pattern of landscape development is well established, it is not possible to reconstruct detailed information on spatial changes to wetlands and the coastline across the entire study area. This means that key parameters of cost-based methods, such as the frequently used least-cost path (Herzog 2008, 2013; Verhagen et al. 2019), cannot be collected. Since it is not possible to determine other potential factors, such as soil conditions, vegetation density or hydrological conditions, with sufficient certainty for similar reasons, the present study focuses exclusively on methods based on the spatial location of the archaeological points, without including the landscape features in detail in the first step.
In general, little research has been conducted on calculating routes or communication corridors based on burial mounds. Nakoinz (2012) proposes a kernel density estimation (KDE)-based approach to visualise linear structures based on density distributions. The main advantage of this approach is that it prevents groups of burial mounds from appearing overrepresented while still allowing connections to emerge. This approach was initially tested in the study area. The bandwidth parameters were derived from the expected mean distance between the points (381 m) and the observed mean distance (107 m) and were increased step by step. To make linear structures visible, a distance ten times greater than the observed average is necessary, corresponding to Nakoinz’s (2012: 453) assumption. Nakoinz (2012) derives the sigma value of 1 km from the distance between Bronze Age settlements and graves, as defined by Ethelberg (2000). In the study area, it became apparent that the centres of density were almost exclusively located on local groups of burial mounds due to the coastal landscape and heterogeneous distribution. Linear structures could only be located at very large bandwidths, which then blurred into one another. While the KDE approach is unsuitable for the study area, it does highlight an important aspect: the necessity of grouping burial mound clusters into functional units.
Another approach for reconstruction is adopted by Løvschal (2013: 233). Here, a method involving the connection of mounds is employed, based on network models. Such network models are usually based on the pairwise connection of nodes, via which a network is formed, for reconstruction in an archaeological context. Basic models are based solely on spatial distribution, with no further parameterisation. Examples include the Delaunay graph, the Gabriel graph, the relative neighbour graph and the minimal spanning tree. Other models parameterise assumptions to optimise networks and adapt them to specific questions (e.g. k-nearest-neighbour). The history of research and an overview of the methodology of network models have been presented in detail elsewhere (e.g. Brughmans & Peeples 2023; Peeples 2019; Groenhuijzen & Verhagen 2017; Verhagen et al. 2019; Jiménez-Badillo 2012; Nakoinz 2012). For the method comparison, a selection of fundamental, established models of spatial network analysis were used, considering the reduced data, which had been tested in the archaeological context mentioned above. First, Delaunay graphs were formed. In this triangular meshing, the condition is that the perimeters of the triangles do not contain any other points. In the Gabriel graph, no other point may lie within the area of an intersection circle formed by two points, and the points are then connected. In the relative neighbour method, a circle with a radius equal to the distance to a neighbouring point is drawn around each point. If no other points can be found within the overlapping circles of the node pairs, the pairs are connected. In the k-nearest neighbour case, k specifies the number of nearest neighbours to be connected; these neighbours are then connected according to their distance, and k must be defined beforehand. Finally, in the minimum spanning tree, all points with the shortest edge length are connected, and cycles are not permitted (Brughmans & Peeples 2023: 246–251). Due to the restrictions, the graphs become less dense, leading to different bases for interpretation. To select the most suitable model, the model expectation must therefore be estimated.
The model should reflect the expectations of a road or communication network from that period and environment. Where possible, all nodes in the network should be connected, as it cannot be assumed that larger independent units exist, which would imply the presence of multiple groups. Furthermore, the study area comprises large areas that were difficult to traverse in the Neolithic and Bronze Age due to features such as mudflats and wetlands (see Verhagen et al. 2019: 221). When reconstructing routes based on burial mounds, it should be noted that nodes in the network are not the starting or ending points of the entire route but rather bases along it. This has the advantage that, even without geoscience-based cost surfaces on a small scale, routes can be reconstructed more accurately as a connecting line between unknown starting and destination points. The aim should therefore be to gain an initial understanding of the corridors.
Data preparation
When calculating routes based on network models that take burial mounds into account, there is a tendency for groups of burial mounds to be overrepresented (Nakoinz 2012). Connecting points to their neighbours creates many short routes within these groups. This influences both the formation of the basic model and, to a large extent, centrality calculations. However, such a model does not necessarily correspond to reality; it is more likely that the paths led through or past the groups (Boroffka 2018: 14; Woltermann 2018: 221). In the archaeological context, it has been proven that individual mounds were connected to a central path via side paths, meaning there were no direct paths between mounds in such groups (see Nakoinz, 2012). From an archaeological perspective, groups of burial mounds can be considered focal points for connections, just as individual graves are in the landscape. As the burial mounds in the study area occur in spatial groups, they must be grouped before network analysis can be performed. As a general size estimate for such burial mound groups has not yet been obtained, the metric parameters will be considered out of the data. The observed mean distance (107.11 m) was used as a starting point. It was found that grouping the mounds together in a meaningful way was achieved by considering the range between the observed mean and half the observed mean distance between the mounds. At half the observed distance, local linear arrangements, which can be up to 2 km long, are sometimes retained as individual points. At the observed mean distance, these arrangements are also grouped together (Figure 2). Therefore, reducing the radius of the observed mean distance offers a good opportunity to group the graves. This reduces the total number of nodes in the network from 2,572 to 730, meaning that a spatial concentration of burial mounds can be assumed in more than half of the cases.
Figure 2
Comparison of the summary of the groups of grave mounds at individual bases, considering different distances (landscape: Lower Saxony State Office for Mining, Energy and Geology).
Models
To determine the most suitable network model for reconstructing the communication network in the Elbe-Weser-Triangle, the results (Figure 3) were compared with the defined expectations for the working area (see Brughmans & Peeples 2023: 255; Herzog 2013: 3). Delaunay and Gabriel generated a large number of edges between hills and groups, even over long distances, with many short connecting paths in proximity. These networks exaggerate the expected structure of prehistoric paths and were therefore unsuitable for answering the present question. K-nearest neighbour networks do not show any segmentation of different areas from k = 5 onwards, depicting an unrealistic mesh on a larger scale. Although the minimum spanning tree generates a fully connected network, its tree-like structure reduces the number of branches too much, preventing recursive networking. For this reason, it is also not preferable for interpretative purposes. Of the tested models, the Relative Neighbourhood Graph (RNG) most accurately represents the expected general impression (see Brughmans & Peeples 2023: 248). It does not depict disproportionately long connections, but it does depict small-scale features and large-scale axes. Due to these characteristics, the RNG is best suited to reconstructing potential communication corridors in the Elbe-Weser-Triangle and can serve as the basis for further analysis.
Figure 3
Comparison of different approaches to reconstructing a network based on burial mounds.
RNG model
The RNG network that has now been determined shows a possible connection between the hills (Figure 4). It reveals small-scale branched patterns in some areas where many knots remain after mounds have been grouped together. Further grouping could be carried out using reconstructed sightlines, but local topography and vegetation would play a decisive role here, and no detailed data is available. At the same time, some linear structures are not mapped as expected. In some cases, it shows that a different course would be more likely, based on the distribution of the mounds (Figure 5, 1). In another case, this causes the estimated axis to shift by a few hundred metres (Figure 5, 2). Furthermore, in some cases, detours are calculated for subjective assessment (Figure 5, 3). While such deviations may occur in detail, they can be considered minor in view of the entire working area and do not alter the overall course.
Figure 4
Relative neighbour network of the burial mounds in the working area.
Figure 5
Exemplary comparison of calculated and distribution based expected edges of the network.
Betweenness
When examining the network, it is important to focus less on individual connections and more on the overall impression, attempting to highlight the main axes. Assuming that the younger mounds relate to the older ones, this should indicate continuity of use. To determine these edges, betweenness is used as a centrality value. Betweenness shows the degree to which a node sits between lightly connected groups of other nodes in the network. However, this tends to indicate edges that connect subnetworks rather than important ones within the subnetwork (Verhagen et al. 2014), meaning that the main axes are more likely to be displayed. To determine the betweenness, the number of shortest paths between nodes running through edges was used to highlight the realistic frequency of the main axes. It is important to note that the nodes used in the calculation represent support points for the paths, rather than the starting and ending points of movement along the corridors. Due to the lack of comprehensive information on settlements that could be considered as such, only the known hills were used. The image (Figure 6) essentially shows a main axis running from east to west, which is located centrally within the image. From there, two main branches extend to the west and east, connecting the rest of the area to the centre.
Figure 6
Calculated betweenness of the network.
Network, archaeology and landscape
When the modelled network is superimposed on and compared with the landscape (Figure 7), a network emerges that branches out across the Geest areas and avoids the large wetland areas. This is to be expected, given the distribution of the sites. However, it is also notable that the edges run along the Geest ridges between the bogs. At the same time, areas where these peatlands are crossed can also be identified, often at narrow points. It is impossible to determine, however, whether these bogs existed at these narrow points during the Neolithic and Bronze Ages or if natural land bridges facilitated crossing. Taking betweenness into account also clarifies the connection between the identified central axes and the landscape. Possible edge effects must also be considered, although these could only occur at the southern edge of the working area. There are wetlands in the Geeste and Schwinge areas. Although it is uncertain whether these existed in the Bronze Age, the area is characterised by a lack of archaeological finds, particularly in the Geeste area. This would mean that the main access to the northern Elbe-Weser-Triangle would be between the rivers’ sources. For betweenness, this would mean that this region would become more important, but this should not greatly affect the overall trend in the working area. Examining betweenness in the working area reveals high values for connections traversing the entire area, from Heinbockel (Figure 7 No. 8) and Westenberg (Figure 7 No. 9) to Holte-Spangen (Figure 7 No. 1) and Debstedt (Figure 7 No. 5). A major node is located near Lintig (Figure 7 No. 6). These high betweenness values indicate that, once established, frequently used connections tend to persist for a long time. It can also be observed that the Bronze Age sites that are referred to as settlements (none of which are known from the SGC) line up along these main routes in the western working area, clustering in the regions around Wingst (Figure 7 No. 10) and Debstedt (Figure 7 No. 5). These regions have rather small-scale connections. At this point, the distribution of the sites fits the model very well. However, to examine the plausibility of the routes more closely, a comparison was made in the following between the sites without a grave context and the calculated edges (see also Herzog 2013; Verhagen et al. 2019: 232).
Figure 7
Network model and betweenness values in connection with modern natural spatial conditions (landscape: Lower Saxony State Office for Mining, Energy and Geology).
Plausibility
Comparison with trackways in the bog
To test the plausibility of the network, clear indicators of paths were used first, which are present in the form of trackways through the bog in the working area. Wooden trackways (Figure 8) can be used as an excellent indicator of physical connections. They are assumed to have been a response to the changing landscape, which was caused by the development of vast bogs and the resulting separation of the landscape (Boroffka 2018: 14; Mennenga et al. 2024). Trackways are the most common type of archaeological features in bogs in this area. Although there are around 550 known trackway sites in Lower Saxony, very few have been dated. In some cases, this has been achieved through finds; in a few instances, C14 or dendrochronological dating were carried out. However, such paths were constructed over a very long period, ranging from the Neolithic to the early modern period (Heumüller 2022: 196–197). There are 41 known trackway sites in the working area, 11 of which have been dated and are mostly prehistoric (see Table 2).
Figure 8
The calculated network and wooden or stone trackways in the bogs of the region.
Table 2
Dates of the trackways in the working area according to AbabWeb.
| DATING | NUMBER OF TRACKWAYS |
|---|---|
| unknown | 30 |
| Neolithic | 2 |
| Bronze Age | 3 |
| Iron Age | 3 |
| Roman Iron Age/Migration Period | 1 |
| Middle Ages | 1 |
| Modern Age | 1 |
When the known wooden or stone trackways in the study area are compared with the calculated paths, a correlation can often be seen; only in a few cases they deviate significantly from the network. One example is the moor path from Hymendorf FStNr. 13 (Figure 8 No. 4), which is assumed to be more recent as it is made of stones (cf. Heumüller 2022).
In some areas, however, the wooden trackways support in general the calculated edges, which span large areas of peatland. This can be seen northeast of Linting (Figure 8 No. 6) and south of Wingst (Figure 8 No. 10), where trackways lie on edges when crossing the bog. Linting has Bronze Age and Iron Age trackways, while Wingst’s date is unknown. The trackways can also be used to determine more precise routes in the terrain and possible unrecorded peatland crossings. For instance, at Abbenseth (undated–Figure 8 No. 7) and in the Langes Moor (Bronze Age and Iron Age–Figure 8 No. 11), ridges and peatland paths can be identified where the calculated network differs slightly from expectations due to additional turns. Southeast of Neuenwalde (Iron Age–Figure 8 No. 3), a relatively narrow crossing through peatland does not appear to have been recorded at all. Another example is trackways extending calculated axes without a specific destination, e.g. north of Bad Bederkesa (Figure 8 No. 13). Such paths are also found in other regions and are interpretated to have had a ritualistic purpose (Paulissen et al. 2022). No other trackways are known to cross bogs or connect Geest islands that are not located near the edges of the network. When comparing the bog paths with the edges of the network, the calculations are supported in many places by the archaeological evidence.
Comparison with other sites
Furthermore, comparisons can be made with other known sites to find distributions (Herzog 2013: 2–3; see also Woltermann 2018: 237; Johansen et al. 2004: 50). As suggested by Herzog (2013: 3), several runs were carried out to check whether the modelling and archaeological data match.
Since trackways only cover a small part of the network, sites dating from the time the graves were constructed should also be considered (cf. Herzog 2013: 2). Comparing the distances between the sites and the edges and randomly distributed points and edges should reveal whether the former are concentrated around the corridors. If sites without a grave context are significantly closer to potential paths than randomly distributed points, this would support the plausibility of the connections. However, to obtain a realistic picture, restrictions must be made due to the landscape. If random points were distributed throughout the entire working area, they would end up quite far from the edges because the distribution includes areas of mudflats and wetlands where there is little evidence of finds. In most cases, this is also a region that could never have been settled during the Neolithic and Bronze Age. This would mean that the proven sites are generally closer to the calculated edges than the randomly distributed sites. To overcome this issue, the area must be restricted for the creation of a comparable database. The part of the landscape with the best conditions for comparative observations is the Geest. Therefore, only sites on the Geest are considered (95%), and random points are only distributed here. The number of random points corresponds to the number of sites on the Geest. The Mann-Whitney U test (no normal distribution of data) is used to check whether the distances differ significantly. Randomisation and the test are repeated and checked for significance. Spatial continuity of the sites may lead to autocorrelation, but this is not expected between the three time periods examined: the Middle Neolithic, the time when the graves were constructed and the later phases, particularly for settlement sites.
For the Young Neolithic and Bronze Age periods, all AdabWeb sites that are not graves and are located on the Geest were mapped (Figure 9; n = 226). Firstly, the distance to the nearest edge in the network was determined. As can be seen, 95% of the locations are less than 1 km away (Figure 10). However, as this cannot be a reliable criterion in a dense network, a second step was taken to check whether the distances differ on average from those of randomly generated sites (see example of a run in Figure 11). Comparing the distances to the nearest edge shows that the sites are statistically significantly closer to the edges than the randomly distributed points. Examining the types of sites more closely also reveals that, apart from the cluster near Debstedt (Figure 9 No. 5), Bronze Age settlements are located very close to the edges (< 500 m). It can therefore be deduced that the calculated network provides a coherent representation of the communication and road network at the time of the burial mounds, despite the inability to determine the accuracy of the locations and the simultaneity of different edges.
Figure 9
The calculated network and the non-burial sites of the Bronze Age and Single Grave Culture.
Figure 10
Distances of the non-burial sites of the Bronze Age and Single Grave culture on the Geest to the nearest edge of the network.
Figure 11
Example of a comparison of the distances of random finds to those from the Bronze Age and Single Grave culture on the Geest.
Continuity
Funnel Beaker period
In southern regions of Lower Saxony as well as in the Netherlands, it has been suggested that road networks were built in relation to megalithic tombs (e.g. Bakker 2008). Due to their visibility above ground, these monumental structures provide a useful point of comparison for burial mounds. If these tombs can be linked to paths, it could suggest that corridors existed even before the first mounds were built. Archaeological evidence from numerous examples demonstrates that TRB burial sites in the region were used during the SGC period and the Bronze Age, too (Behrens et al. 2024: 27). The Megalithic tombs (Figure 12) are examined in the same way as SGC and Bronze Age sites. Like burial mounds, megalithic tombs are also found either in necropolises or standing alone in the landscape and are used as a basis for comparison (see Behrens & Mennenga 2025 for the calculation of necropolises). Here, too, the megalithic tombs on the Geest are located significant closer to the calculated routes than would be expected from a random distribution of points. Thus, it can be deduced that the monumental structures of the TRB partially describe the communication network characterised by burial mounds, at least in the study area. If the linearly arranged megalithic tombs are likewise interpreted as indicators of transport routes, their distribution suggests that some of these routes had probably already fallen out of use by the time the burial mounds were constructed. Most of the graves supporting these disused axes were found beneath layers of peat. From Neuenwalde to Wanna, the graves are aligned in the moor, alongside a wooden trackway of an unknown date. It seems that a connection shifted north towards Krempel and/or south here. There are also trackways here that suggest the bog areas were crossed. Finally, the Hainmühlen 1 site, located southeast of Lintig near one of the potential peatland crossings, should be highlighted as an exceptional site for the TRB. Repeated parallels to the Danish cult houses of this period suggest a similar interpretation of the findings (Mennenga 2022b).
Figure 12
The calculated network and the megalithic graves of the Funnel Beaker culture.
Iron Age to Middle Ages
Although burials in mounds continued in later periods, they accounted for only a small proportion of the total. Examining other sites from this period reveals that, during the Iron Age, settlements shifted from the centres of the Geest to its edges. This is particularly evident in the west of the Hohe Lieth. However, many of the inland sites from this period are still located close to the calculated network of burial mounds. With the Roman Iron Age, the use of the coastal marshlands began to increase. Except for the area around Lintig and the main routes leading north-west and south-west from there, few sites are found in the central Geest regions. This trend intensified during the Migration Period and the Middle Ages, resulting in the Geest losing its importance. Nevertheless, Aust (1956: 116–124) had already mapped medieval fortifications at some of the peatland sites through which the modelled paths run, suggesting that certain junctions were probably continuously important until then.
Discussion
Analyses of burial mounds from the Single Grave Culture and the Bronze Age on the southern North Sea coast — using the northern Elbe-Weser-Triangle as a case study — have demonstrated the feasibility of reconstructing prehistoric mobility and communication corridors within the study area. Although it is difficult to draw conclusions about the prehistoric landscape due to significant landscape changes since the last ice age and a lack of geoscientific spatial data, network models can be used as a suitable modelling approach. After grouping the burial mounds, these groups and the remaining individual mounds could be considered as potential route base points. The connection of these points was established after testing suitable methods using the relative neighbourhood network. This method avoids extreme distant edges while retaining both small-scale branches and supra-regional axes. This means that it best fulfils the expectations of a coherent prehistoric communication system. For the identified network, betweenness can be calculated and used to derive its significance, and thus possibly its duration of use or frequency in the landscape. In principle, it cannot be assumed that all connections were active simultaneously, even if they belonged to the same network. Landscape changes, such as the spread of wetlands and peatlands, would certainly have necessitated the expansion or relocation not only of activity areas but also of paths. It must, of course, also be noted critically that it is precisely the slightly elevated Geest areas that may have remained highly attractive over a long period of time. However, there are indications – at least in the case of the potential main axes – that in many instances these are to be found on the periphery of the Geest areas rather than at their centre or highest points. The construction of trackways can also be used to assess the plausibility of the network. In places where the network passes through peatlands, such paths have been verified by archaeology. However, when comparing the network with the distribution of features, it is clear that the network does not always accurately reflect the path axes. This shows that the calculated edges should not be considered exact paths, but rather corridors in which prehistoric activities took place. Nevertheless, the basic statement of connections and weighting can be considered a realistic representation. This is supported by a comparison with sites not related to graves. Here, too, there is a concentration of sites near the edges. This supports the argument that the calculated corridors are related to higher activity in their surroundings and are therefore entirely plausible. As well as being plausible at the time of the monuments’ primary construction, a comparison with sites of previous and subsequent land use can indicate the continuity of the communication corridors. A comparison with the distribution of megalithic tombs from the Middle Neolithic period reveals that these are also situated closer to these axes. However, other linear connections between these graves can also be identified that are not present in the calculated network. Assuming a connection between graves and communication structures here as well reveals a shift in some cases, possibly due to changes in the landscape, in addition to continuity across large areas of the network. This continuity is particularly notable given the significant cultural shifts between the TRB and the SGC. This is probably due to the strong presence of megalithic tombs in the prehistoric landscape, as evidenced by archaeological finds and features showing their integration into the subsequent SGC’s overall cultural picture. The increasing importance of roads and connections during this period may also be explained by a general increase in mobility (Haughton & Lövshal 2023: 1479). Following the Bronze Age, many connections lost their importance as settlements shifted mainly to the immediate coastal area. However, comparisons with younger sites suggest that the main east-west axes continued to exist. In addition to the aforementioned spatial resolution, the use of burial mounds for network analysis has limitations in terms of temporal resolution, a problem that, according to the current state of research, cannot be resolved. Accurate dating of the individual burial mounds and surrounding sites would greatly improve the model and its assessment. Similarly, sites hidden beneath layers of peat and coastal marshland (Behrens et al. 2022; Mennenga et al. in print) would further enrich the overall picture. Nevertheless, reconstructing the prehistoric landscape provides a good basis for optimising the models further, considering other calculation methods.
Despite methodological criticisms, the calculated corridors offer significant added value to archaeological research. In areas with narrow passages in wetlands in particular, further chronological classification of waterlogging can highlight its significance in terms of control and influence. This is especially pertinent when considering the exceptional sites already known to be located here, such as Hainmühlen 1 and later fortifications (Aust 1956). The less intensively researched or lesser-known sites of the SGC and the Bronze Age show great potential for improving our understanding of communication and continuity of use in the region. This approach is generally well suited to regions without topographical data, as it can be used to reconstruct communication corridors supported by monuments. The results are particularly meaningful at the level of large-scale axes as well as at bottlenecks and are validated by trackways and findspot patterns. However, limitations arise from the lack of a detailed chronology of the burial mounds, uncertainty regarding local landscape development and data-related gaps. However, it is precisely these limitations that highlight the value of the research. Rather than reconstructing a ‘road’, identifying corridors, nodes, and barriers allows us to re-examine mobility, land use, and possible lines of tradition from the Neolithic to the Bronze Age in the Elbe-Weser-Triangle.