Author: Zoran Čučković
[ This is the author’s version of an article which will be published soon in The Journal of Archaeological Research and Theory ]
Abstract
This article is introducing a method for the analysis of landscape visual coherence. Inspired by landscape and architecture research, the landscape chambers method is based on quantitative analysis of visibility networks, modelled in digital environment. It relies on an algorithm for the detection of closely connected subgroups within an intervisibility network, in order to isolate visually distinct areas or landscape chambers.
This approach is applied to prehistoric landscapes in the Parisian Basin (2000 – 500 BC), where funerary monuments reveal complex relationships between past populations and their visual landscape. The analysis uncovered that these monuments were typically placed in visually exposed locations, with preference, in certain cases, for visually coherent landscape chambers. The proposed approach thus offered insights into the semiosis of the prehistoric landscape, i.e. the production of new meanings through visual discourse. More generally, the proposed method aims to provide a conceptual and methodological bridge between the study of physical and mental spaces.
Keywords : landscape archaeology, Bronze Age, Iron Age, spatial analysis, network analysis
Introduction
How does a landscape become meaningful? Even if it may seem intractable at first glance, this problem lies at the heart of landscape archaeology. The ways in which past or contemporary populations interact with their environment are necessarily channelled by their social representations and values. Conversely, landscape features are often construed as material evidence and justification of such representations; social exclusion, land ownership or group identity are quite frequently materialised or read into the landscape. In a nutshell, landscape is what meaning does to the environment.
A particular obstacle in dealing with human landscapes is posed by the common confrontational trope, opposing culture to nature. Scholars tend to focus on the environmental impact of past or present communities, or on their choices made in order to invest the land. The physical environment is thus seen as acted upon or thought about, it is adapted to and referred to, which are all variations to the ancient theme of the opposition between a living subject and an inert object. From that point of view, meaning-making in the context of landscape habitation may appear as label-making, i.e. one-way production of mental references.
The problem, however, is that landscapes do not come with handles ready to receive such labels. Topography in particular is ontologically ambiguous: when does a hill become a mountain, a stream a river? Indeed, works in linguistics and geography have demonstrated a dazzling diversity across cultures in terms of classification and terminology for landscape features (Basso, 1996; Turk et al., 2011). Therefore, human landscapes cannot be approached as a collection of objects, a candy box filled with features neatly separated from each other, but rather as a continuum of physical and perceptual variations. We need to evaluate the degree of meaningfulness of such variations, i.e. the likelihood that certain regularities would become recognised as specific landscape entities. It is only through such integrated approach, attentive to both the salience of topographic features and cultural practices of landscape inhabitation, that landscape archaeologists may address the semiotic space.
This paper is dealing with the problem of open spaces, their perceptual salience and cultural processes through which they become meaningful. It presents the landscape chambers method which is inspired by approaches in the analysis of architecture and human landscapes, namely space syntax and landscape rooms method (infra). Based on intervisibility network analysis, the landscape chambers method takes a formalised, quantitative approach; its aim is to break up the landscape in visually coherent zones and to measure the degree of such coherence. This approach is evaluated on the case of Bronze and Iron Age landscapes of the Parisian Basin (France; 2000-500 BCE). During that period, thousands of funerary monuments were constructed across the prehistoric countryside. These structures were built to be seen and by that token their distribution is indicative of the relationship of past populations with their visual landscape. What is more, by inscribing social memories and identities into the landscape, funerary monuments provide clues to practices of meaning-making within the social space. If the dead were placed to be seen in the landscape, we may ask: how did such semiosis alter the understanding of the space and which elements of the existing landscape were referenced in order to produce new meanings?
Archaeology and the cognitive landscape
A number of archaeologists have studied the role of human cognition in the shaping of past landscapes. Richard Bradley (1998) stressed the importance of visual settings for British prehistoric circular monuments, namely henges and stone circles. Some of these may have been deliberately placed at the centre of “landscape arenas” with clear views to an elevated, circular horizon. Others still may have emulated such experience with the help of raised circular banks (Bradley, 1998, p. 126). These hypotheses inspired the development of formal approaches in GIS digital environment, namely the analysis of visual horizons by Lake and Woodman (2003) and Lake and Ortega (2013) who examined some 300 Neolithic circular monuments. They formalise landscape arenas as topographic basins and measure visual properties of horizons that enclose the monuments (such as horizon distance, elevation or orientation). A series of complex methods and techniques for the characterisation of visual settings of prehistoric sites and monuments was developed by Marcos Llobera (1996, 2003, 2007). For instance, he considers that prehistoric linear ditches of Southern England may have been placed at visual thresholds between specific areas, namely on ridges, and by that virtue would have been used to mark territorial divisions (Llobera, 1996). An interesting and original attempt to evaluate the relationship between natural topography and cognitive landscape was made by Bernardini and Peeples (2015) for the ancestral Pueblo of the southwestern USA. Using visibility analysis, they isolated a set of particularly prominent mountain peaks and then modelled “sight communities” composed of local communities that were exposed to similar mountain views. The authors suggest that such shared views would have had an impact on cultural bonding and communal identity. An approach inspired by the architectural theory of space syntax was published by Criado Boado and Villoch Vázquez (2000), dealing with the prehistoric megalithic landscape of Galicia, north-western Spain. They attempted to characterise in detail the movement through a topographically well delimited area, a mountain plateau, and to analyse the sequence of views towards megalithic monuments along likely paths.
By no means exhaustive, this overview should suffice to substantiate several remarks. First, there is a strong and continuous interest in spatial perception and cognition within the domain of landscape archaeology. Landscape arenas, sight communities, but also territories under control are social and cognitive categories, they describe how people think about their environment. Second, the research in this domain appears as “top heavy”, driven by high level issues of social perception, identity, territoriality or symbolism. Multiples scales of analysis and levels of abstraction may thus be jumped-over, such as the general characterisation of the landscape in terms of its perceptual impact. In other words, methodologies and middle range theories are poorly developed in archaeological landscape research: how are we to formalise and evaluate social values attributed to various landscape units ? Finally, a closer look at archaeological literature also reveals a persistent inability to come to terms with space. Its frequent categorisations as “geometrical”, “cartesian”, or again “social” or “experiential” imply its somehow lesser or incomplete status, as if the space should pertain to something more concrete, as if it were a metaphor rather than physical reality.[1] This is perhaps related to the common object-oriented character of archaeological research, as opposed to space-oriented paradigm of human geography or urban research, but in any case, past landscapes cannot be successfully analysed without fully integrating concepts and methodologies of spatial science. Landscapes are composed of physical spaces, which can and should be studied in their own right.
The landscape chambers method, presented below, is inspired by architectural and landscape planning research, namely the space syntax and the landscape rooms approach. The fundamental postulate of the space syntax reads that the “space is lawful”, implying that the configuration of a specific space will strongly influence social life in terms of interaction and daily practice (Hillier, 2014, p. 19). Some spaces may hinder interaction and navigation, for instance when encumbered by a seemingly random mosaic of closely packed buildings, while others may channel human encounters and dictate social contact. This is clearly applicable to the open landscape, and indeed such experiential qualities were picked up by the concept of landscape arenas, among others.
The space syntax approach is particularly attentive to spatial integration, which denotes the connectivity across space. Well-integrated spaces are easily reachable and/or visually controllable from multiple directions (Hillier, 2004). The particularity of space syntax methodology is that such connectivity is evaluated on the basis of visual connections, considering that visual connection commonly implies physical connection, at least in a typical urban layout. Heavily used for the analysis of urban environments, space syntax has nevertheless seen few applications in landscape analysis. O’Sullivan and Turner (2001) proposed an approach based on visibility graphs which capture total visual connectivity within a specific area. Technically, this amounts to calculating visual links between all pixels of a typical digital elevation model (DEM) in raster format. Authors propose a number of metrics for the obtained visibility network, but only on theoretical basis. Among these, cohesive subgroups, composed of highly interlinked network clusters, are of particular interest for the landscape chambers method since, as the authors suggest, such visually integrated spaces may convey a “sense of enclosure” (O’Sullivan & Turner, 2001, p. 236). Turning to archaeology, Brughmans and Brandes (2017) presented a strong case for quantitative analysis of visibility networks. They focus on complex relations of intervisibility between archaeological sites, while envisaging a future use of space syntax methods as well.
An additional inspiration for the approach presented here comes from the so-called landscape rooms method, developed by a group of Norwegian landscape planners and archaeologists (Gansum et al., 1997; Jerpåsen, 2009). This is a simpler, informal method devised to define visually coherent areas or landscape rooms. Such spaces are often delimited by higher terrain, for instance ridges enclosing a valley. Visual and physical boundaries may also appear at sharp plateau edges or shores of water bodies. The method has been used, among others, to study the landscape setting of prehistoric funerary mounds in Norway (Fry et al., 2004; Gansum et al., 1997). However, the landscape rooms approach has several shortcomings in terms of methodology. It is based on subjective evaluation of the landscape through field visits and analysis of topographic maps. GIS modules for visibility analysis may be used, but as an aid for such “soft phenomenology” (Jerpåsen, 2009, p. 135). For instance, this may permit to evaluate the visibility potential without the present-day vegetation cover. In any case, the lack of formal methods for measuring visual quality does not allow for quantitative comparisons between individual rooms or between different landscapes.
Landscape chambers methodology
Drawing on space syntax and landscape rooms analysis, the landscape chambers method was developed to provide researchers with a low-level method of general landscape characterisation, relieved of hypotheses on specific symbolic or cognitive charges. It is a decidedly spatial approach, focused on visual qualities of the space itself, rather than on visual impact of an object or visual experience of a person. The main goal of the method is to measure visual connectivity across space and to isolate areas of coherent visual structure. Considering that visibility is fundamental to navigation and social interaction in a given space, it can be argued that visually integrated areas would have a profound effect on human spatial cognition and social life (Penn, 2003). Such areas may be expected to be specifically designated by their inhabitants, for instance as a valley, a plateau etc. That being said, the existence of explicit cultural references is not implied by the landscape chambers approach; its purpose is not to reconstruct past or present mental maps, but rather to characterise the landscape in terms of visual salience and coherence.
Visually coherent zones are defined here as areas of high overlap between visual scenes. Such overlaps would indicate that, for a potential observer, there would be no abrupt change in the visual experience while moving through space. Likewise, if we imagine multiple observers spread across the landscape, we would expect a higher degree of consensus in terms of their sense of location. These overlaps can be represented as networks which connect all potential observers with their view targets (Čučković, 2023). A typical real-world visibility network will be composed of a multitude of dense intervisibility clusters, separated by visual barriers such as higher ground, architecture or vegetation (Figure 1). We should note that such clusters tend to map areas of high visual integration (as defined by space syntax approach; Hillier & Hanson, 1984, p. 108f; Koutsolampros et al., 2019).
In this work, visibility clusters or sub-groups are determined by the classic modularity index, as proposed by Newman (2006). The modularity index (Q) is calculated, not for the network itself, but rather for a given partition of the network into sub-groups (π):
where ls denotes the number of internal links and ds the number of connections for a given subgroup s. These values are evaluated against the total number of links in the network (L) (Fortunato & Barthélemy, 2007 eq. 1). Note that a typical visibility network will normally contain a number of links that cross cluster boundaries (Figure 1). These external links are crucial for the modularity calculation as they lower the ratio of internal links. The algorithmic solution for determining the modularity of a network is repeating this calculation hundreds or thousands of times, in order to find or to approximate a partition which yields the highest score. Finally, the resulting clusters are translated into landscape chambers using Vorony (Thiessen) tessellation, i.e. the chamber borders are traced at midpoint distance between observer points assigned to different clusters (Figure 1).
In order to render these calculations technically feasible, we need to implement a sampling strategy. Visibility networks can easily reach colossal sizes in terms of memory storage. For instance, a rather modest elevation model of 5 megapixels (2500 * 2000 pixels) can be expected to produce two and a half billion relations (2.5 * 109), given an average of 500 visible pixels per observer. Each of these relations has to be stored explicitly, as a pair of nodes. However, we can expect a high degree of overlap between visual scenes for closely neighbouring locations in a typical open landscape, i.e. devoid of architecture or other obstacles which may block the view at close distances. For such an open landscape, it is safe to sample only a fraction of locations (see: Dungan et al., 2018; Kim et al., 2004). Furthermore, for reasons of both technical and methodological simplicity, visibility networks are modelled here as eye-to-eye intervisibility, i.e. as reciprocal viewership. This is a simplification of real-world situations where visual connections may be unreturned; I may be visually exposed without having visual control over locations with potential observers, or conversely, I may control visually an area without being seen. Such level of detail would require the use of co-called directed network structure (connections are valid in one direction only) which is significantly more complex to process (Ahnert & Fink, 2008). As a side-note, the space syntax terminology uses the term visibility graph while referring almost exclusively to undirected intervisibility networks (Koutsolampros et al., 2019).
Founded on network analysis, the landscape chambers method nevertheless combines naturally with other, more common approaches in visibility modelling. This applies in particular to the cumulative and total viewshed. In these methods, areas of visual coverage are combined into a single model representing the frequency of viewshed overlaps (Figure 2). Most commonly, such frequency models are stored as pixel matrices or rasters. In the total viewshed approach, viewshed calculations are made for all data points, i.e. pixels if a raster digital elevation model (DEM) is used, while cumulative viewsheds are usually modelled for a more restricted selection of observation points (Brughmans et al., 2018; Llobera, 2003). It may be remarked that viewsheds and their overlaps can be deduced from the intervisibility graph: the frequency of views amounts to the number of visual connections per observer, i.e. to node degree in network science terminology. However, technical difficulties of visibility graph calculation render this approach laborious and imprecise, due to sampling and the assumption of visual reciprocity. Conversely, we may also be tempted to use viewsheds to delimit landscape chambers considering that such spaces can be defined as areas of strong overlaps between visual scenes (e.g. Llobera, 2003). However, viewsheds cannot be used to evaluate complex relationships between locations that do not share the same visual scenery, but still may belong to a visually coherent ensemble, for instance two ends of a mountain valley. Network analysis offers a richer, contextual approach, which can trace chains of such visual relations. The two methods should thus be regarded as complementary.
For the purposes of the following analysis, visibility calculations were carried out using the QGIS Visibility Analysis module, developed and maintained by the author (Čučković, 2016). In terms of data input, the module requires a digital elevation model (DEM), i.e. a digital representation of the local topography, and a set of observation and target points. In order to produce the intervisibility network required by our method, the observers act in their turn as view targets and only the reciprocated, eye-to-eye lines of sight are retained. The module also features a visibility index algorithm which can be regarded as a slight improvement of the total viewshed method. While the latter is often expressed as a simple frequency count, i.e. the number of views received by a sampled location, the visibility index calculates the view success of each observer, which amounts to the proportion of the observed surface within a specified view radius. The obtained values thus range from 0 to 1 (or 0 to 100%), where the maximum score indicates a completely unobstructed view. Such models are intuitively understandable, while conveying the comparable information as frequency counts. Considering the network analysis, the modularity segmentation was made using the implementation of the “Louvain method” by T. Aynaud (2018) in Python programming language.
Let us now consider higher level methods that will be used here to analyse the obtained visibility networks. When made over natural, undulating terrains, visibility networks are highly complex and their segmentation into clearly delimited clusters is all but trivial, as will become apparent in the following case study. In order to measure the visual coherence of such clusters, a visual quality index (VQI) was devised, based on three specific indices (Figure 3). The first one, connection density, is simply evaluated as the number of links per unit of surface, here a hectare. Only the internal links are taken into account, i.e. links between observer points located in the same group.
The ratio of internal and external links is particularly important for evaluating visual separation of individual chambers. This is calculated as the percentage of internal links within the total of links connecting to the chamber. Note that in this case external links count twice, for both chambers they connect to; from this perspective the total number of links (L) within the network thus amounts to
This implies that the ratio of external links will most likely be higher when calculated for individual chambers, in comparison to calculation for the overall network, where such link doubling is not required.
The internal structure of intervisibility clusters, which compose individual chambers, can be examined by applying the modularity method on each one of them. As explained above, the resulting score will inform us on the susceptibility of a network cluster to segmentation: how easy would it be to break apart? Highly modular clusters will thus appear as poor candidates for coherent landscape chambers, given their fragmentary internal structure.
Clearly, none of these measures would suffice on its own to fully describe landscape visual qualities, while others still may be taken into account. For instance, we may consider the coherence of the horizon line for each chamber (Lake & Woodman, 2003), or various measurements of network structure (Brughmans & Brandes, 2017; O’Sullivan & Turner, 2001). In principle however, the three chosen measures – density, transgression and modularity of intervisibility connections – describe both the internal structure of landscape chambers and their external relationships. They should permit, then, to evaluate the produced model, knowing that more sophisticated approaches may be required to address specific problems. The visual quality index (VQI) is then calculated through simple addition of normalised values of these measures (values scaled into 0 to 1 range). The modularity index values are reversed so that lower values would maximise the VQI score (Figure 3).
It should be noted that visual links on which the VQI relies are necessarily truncated on the edge of the analysed area. Devoid of potential observers, the space external to the area of analysis will thus appear as an artificial barrier, which means that landscape chambers situated along the analysis perimeter will appear as much better defined than in reality. More specifically, they will have significantly higher proportions of internal links. For these reasons, the edge effect needs to be carefully taken into account when designing the landscape chambers analysis, in principle by extending the area covered by the intervisibility network beyond the limits of the chosen study area.
Case study: funerary landscapes of the Seine-Yonne interfluve (France)
Introduction
During the Bronze and Iron Ages (2nd – 1st millennium BCE) the Seine-Yonne interfluve, in the Parisian Basin, was settled in loosely organised hamlets and individual farms (Mordant & Gouge, 1992; Peake et al., 2017). Houses and house compounds had rather short lifespans as their occupants preferred to relocate, perhaps following the life cycle of the household group (Gerritsen, 2008; Lafage et al., 2006). Larger settlements, sometimes fortified, became more frequent during the 1st millennium BCE, but their appearance did not change radically the overall dispersed, mobile settlement pattern (Peake, 2020). Burial grounds, on the other hand, were long-lasting features of the cultural landscapes, often used for centuries or even millennia. Alongside different types of flat graves, these sites also contained large numbers of ditched funerary monuments. Deliberately exposed to view, as we shall see later on, funerary structures acted as persistent landscape features, i.e. as landmarks. Two types are by far predominant: circular and square ditches, the former characteristic of the Bronze and the Early Iron Ages (2200-450 BCE), while the latter became more common from the 6th or 5th century BCE onwards (Baray et al., 1994; Chaume et al., 2007; Mordant & Nouvel, 2017). Their dimensions are similar, typically in the 10 – 20 meters range. A large number of these features must have been adorned with an earthen mound, a barrow, but the existence of these earthworks is difficult to confirm in the heavily ploughed countryside of the Parisian Basin. Palisades or similar timber constructions inserted into ditch infill also seem to be a relatively recurrent feature (Baray et al., 1994; Mordant & Mordant, 1970).
Ditched funerary monuments remain distinguishable by aerial surveys and have been recorded in thousands, especially on the chalky, calcareous substrate which is characteristic of large stretches of the Parisian basin. Widely scattered across the prehistoric countryside, prehistoric cemeteries are nevertheless visibly concentrated in river valleys, which must reflect the economic and political importance of the alluvial niche (Brun et al., 2005; Gouge et al., 1994). A number of authors relate burial places to prehistoric territorial organisation. The presence of sumptuous burials may thus indicate the proximity of political strongholds, for instance at junctions of river valleys (Mordant & Gouge, 1992). More generally, burial places may have substantiated land claims of local communities by demonstrating their historical attachment to the land (Chapman, 1995; Parker Pearson, 2003, p. 132ff). That being said, we also need to take into account the historicity of the funerary landscape, i.e. the long term accumulation of past actions and meanings which guided the use and maintenance of burial grounds. A number of studies have clearly demonstrated the lasting historical prestige of burial places during European later prehistory, in some cases maintained and reused over millennia (Q. Bourgeois, 2012; Bradley, 2002; Peake & Delattre, 2005). The funerary landscape thus carved its specific semiotic network, conjugating social memory and social space.
Our study area covers the undulating chalky plateau known as the Sénonais, situated between the Seine and Yonne rivers (Figure 4). The calcareous geology is favourable for aerial survey: several hundred ditched funerary monuments have been identified in the study area. Large majority of these was attributed to the Bronze and Iron Ages on the basis of characteristic layout shapes (Mordant & Nouvel, 2017). By far the most frequent types are circular and square shaped enclosures. In the neighbouring Seine and Yonne valleys the peak of construction of circular monuments dates to the Late Bronze Age and the beginning of the Iron Age (1350 – 600 BCE), knowing that circular funerary monuments were built throughout the entire span of the last two millennia BCE in the Parisian Basin (Peake et al., 2017; Roscio & Muller, 2012). Although only a handful of funerary sites were excavated on the Sénonais plateau, we can reasonably infer similar chronological patterns in the study area (Deffressigne & Villes, 1995; Filipiak et al., 2017). Square shaped ditched monuments, on the other hand, are generally more recent, proliferating from the 6th or 5th century BCE onwards (supra). That being said, much earlier examples are attested, among which we should note two Late Bronze Age rectangular monuments from the Sénonais plateau (radiocarbon dated between the 14th and 12th c. BCE: Filipiak et al., 2017). For the following study, 429 circular shaped monuments have been retained, as well as the two aforementioned rectangular ones, for a total of 431 structures (see Supplementary Information). We are thus focused on Late Bronze and Early Iron Age landscapes (1350 – 500 BCE approx.).
Regarding the environmental data required for visibility modelling, the digital terrain model (DTM) is certainly the most important. DTM produced by the French IGN institute was used, with the horizontal resolution of 25 meters (IGN, 2013).[2] The resolution of this dataset should be sufficient for large scale visibility analysis, with additional benefits of smooth representation of topography, as well as the absence of architecture and present vegetation. That being said, prehistoric vegetation cover was certainly an important environmental impediment to visual contact. Pollen samples, almost exclusively obtained from river valleys, indicate that open grassland was predominant throughout the later Prehistory, interrupted by stretches of oak groves. Humid zones in river valleys were covered with riparian forest to various extents (Leroyer et al., 2012). Unfortunately this information cannot be factored in visibility calculations due to the lack of spatial precision, all the more so for the Sénonais plateau which has seen little paleoenvironmental research so far. At best, the impact of vegetation cover can be considered on a qualitative level, e.g. as the existence of potential vegetation screens along watercourses and humid areas.
Visibility analysis
In order to understand the visual discourse of prehistoric funerary monuments, let us first examine their impact on the surrounding landscape. Did prehistoric builders deliberately choose visually exposed locations for the monuments? This can be verified through visibility index, by comparing the visual exposure of funerary monuments with the general visual exposure of the Sénonais landscape. For the following analysis, the observers were assigned the eye level of 1.6 m, and maximum view radius of 3.5 kilometres. Such short radius was chosen with regard to the size of the monuments, no more than 2 or 3 meters in height and 10 to 15 meters in diameter for typical cases (Rottier, 2009). These structures would have been difficult to perceive at longer distances, given the limits of human visual acuity and the outdoor setting (Ogburn, 2006; Shang & Bishop, 2000).[3]
.
The analysis revealed that the visual exposure of burial places is clearly higher than the average of the study area, with respective mean values of 0.34 vs 0.26 (Figure 6). In other words, the monuments are in general exposed to 8 % more potential observers than random locations.[4] Indeed, it can be readily observed that a number of burial places are situated on visually exposed ridges which dominate wider surroundings (Figure 7, right). Even in the flat alluvial plain, the search for visual exposure can be deduced from the preference for raised banks and gravel domes (Figure 7, left). This finding is consistent with results obtained for Bronze Age barrow landscapes elsewhere on the North-European lowlands, for instance in Belgium (De Reu et al., 2011) Netherlands (Q. Bourgeois, 2012) or Britain (Woodward & Woodward, 1996). That being said, we need to acknowledge the lack of vegetation cover in the datasets used. This is certainly an issue for river valleys, namely the Vanne valley, where the highly probable riverbank vegetation must have screened off a significant amount of views across the water. Nevertheless, even if somewhat overestimated, the results obtained by the visibility analysis of visual exposure should not be dismissed; prehistoric funerary monuments were clearly built to be seen and the obtained results do seem to reflect a genuine preference for exposed locations.
In order to gain insight into spatial and visual configurations of this funerary landscape, let us now turn to the landscape chambers method. The required intervisibility network was modelled for 25 000 randomly located observer points, with an average distance of 160 meters. The analysis perimeter was enlarged by approx. 5 km in relation to the study area, which should help to counter the edge effect (as discussed above, Figure 8). These points are connected with some 1.08 million reciprocal intervisibility links. The maximum sight distance was set to 3.5 km, in consistency with parameters for visibility index calculation. The observer height was set to rather low one metre. Models produced for standing observers turned out to be over-optimistic in terms of visual connections and were consequently of poor use for the segmentation into landscape chambers. The rather flat topography of our study area must have contributed to this problem. Perhaps more importantly, eye-to-eye intervisibility between standing observers (or seated, as calculated here) is more permissive in terms of visual connectivity than, for instance, eye-to-ground scenario (Brughmans et al., 2015, p. 74; Čučković, 2023). In any case, it should be borne in mind that landscape chambers method is not devised to reconstruct with fidelity a series of visual scenes, but rather to model the overall visual structure of the landscape.
We can see on the Figure 8 that a number of resulting chambers is well delimited by higher topography, for instance n° 2, 4 and 17. This is reassuring because such landscape units should be recognisable by their inhabitants, especially when centred on a water course. Nevertheless, a number of our chambers appear as problematic. For instance, the chamber n° 11 is stretching over the largest watercourse in the study area (Vanne river), as well as over its tributaries and a piece of rugged terrain to the north. This indeed seems an unlikely unit of the cognitive landscape. The intervisibility network of this chamber features a main thoroughfare along the river valley and a number of loosely attached branches which follow smaller streams (Figure 9).
Clearly, before proceeding further, we need to evaluate the visual coherence of individual landscape chambers. This can be achieved through the VQI (Visual Quality Index, supra). As already mentioned above, VQI scores are likely to be artificially increased in chambers situated along the border of the analysed area; therefore, the buffer zone between the study area and the analysis perimeter should not be taken into account.[5] Within the study area, two chambers stand out with excellent visual quality: n° 7 and 26 (Figure 10). The first one, situated in the northernmost corner of the study area, occupies a portion of the Seine alluvial plain. This flat area cannot be directly compared with the undulating Sénonais plateau, at least not through the VQI which tends to obtain very high scores in the alluvial zone along the northern border of the study area. The chamber 26, on the other hand, is particularly interesting for our analysis. The visual coherence of this landscape chamber is directly related to its morphology: a rounded depression 8 by 10 km large with a small stream running through its middle (Figure 11). The chamber limits consistently follow the rim of the depression. As we shall see later on, this chamber also accommodates an important cluster of funerary monuments.
A number of chambers obtained poor VQI scores. As expected, this is the case of n° 11, as well as of chambers that already by their irregular shape appear as problematic, such as n° 3. Other chambers may have a partially low score. For instance, the largest chamber, n° 5, has an above average ratio of internal links but also high modularity score. This implies that, even if relatively detached from its neighbours, the chamber n°5 still remains visually disconnected in its interior.
Let us now examine the distribution of prehistoric funerary monuments throughout landscape chambers (Figure 12). The highest densities of monuments can be found in chambers 11 and 26. The former is stretching over the Vanne river valley, an environmental niche which must have been attractive for prehistoric settlement (supra). However, we have seen that somewhat complex morphology of this zone was not translated into a coherent landscape chamber. A more interesting case is the chamber 26, occupied today by the village of Marigny-le-Châtel and hosting the highest density of recorded funerary monuments on the Sénonais plateau (0.83 per km²). As we have already seen, its simple morphology, consisting of a shallow, rounded depression, has been successfully captured by the graph segmentation routine (Figure 11). What is more, the VQI score of the chamber 26 is the second highest in the study area (2.09). Archaeological research in this specific area has been rather scarce, but in 2014 a rescue excavation ahead of a pipeline construction permitted to uncover a portion of an important Bronze Age cemetery, situated in the string of prehistoric sites traversing the chamber. Circular and quadrangular ditched enclosures were associated with cremation and inhumation burials dating between 1300 and 900 c. BCE (Figure 11). Settlement traces were also recorded nearby (Alcantara, 2015).
The correlation between the second highest VQI score and the density of prehistoric monuments in the Marigny-le-Châtel chamber is highly intriguing. Unlike the alluvial Vanne river valley, this area does not differ sensibly from the rest of the Sénonais plateau with its chalky substrate and sparsely distributed, small streams. No specific environmental advantage is discernible from geological data. However, the chamber 26 is characterised by particularly high visual integration, which makes it a good candidate for a “landscape arena” (see Introduction). Such space would have offered consistent visual experience for scattered or mobile observers. Indeed, the chamber 26 appears as a hotspot of visual exposure to funerary monuments (Figure 13 ). This cumulative viewshed model quantifies the frequencies of viewshed overlaps, which in this case mark the visual presence of individual monuments (with height estimated to 1 m and the observer’s eye height set to 1,6 m; the maximum view reach of 3.5 km is retained). Therefore, considering multiple lines of evidence, namely the high visual quality score and the density of prehistoric monuments, the Marigny-le-Châtel area appears as a potential landscape arena which provided a consistent and meaningful scenery for funerary monuments and related commemorative practices. Indeed, the monuments tend to gather in the centre of the chamber, in a string loosely following the watercourse, i.e. in the location which is pivotal in terms of both visual integration and visual salience.
Nevertheless, we can equally remark that other visual exposure hotspots on Figure 13 remain much smaller than modelled landscape chambers, and may even transgress their boundaries. Perhaps the choice of relatively short visual range, 3.5 km, contributed to this discrepancy, but we also need to remember that the success of the intervisibility segmentation routine was rather variable across the study area. The hotspot appearing on the southern part of the study area is clearly related to a stretch of the river Vanne valley, which indeed may have been a landscape arena comparable to the one in Marigny-le-Châtel chamber. However, our computational routine favoured significantly larger landscape units. A similar discrepancy between a large chamber and a localised visual exposure hotspot can be found in chamber 5. Therefore, our model may only partially reflect the spatial extent of potential landscape arenas. These tend to appear in low lying settings and are clearly related to local clusters of visually exposed funerary monuments.
We should thus verify whether the landscape arena scenario is indeed applicable to the rest of the prehistoric burial landscape. If this were the case, visual connections of burial mounds would tend to densify in low-lying chamber core areas, considering that these areas are visually well integrated. Indeed, we have seen that visual hotspots of the funerary landscape tend to occupy such chamber core areas. Following this insight, we can expect that views toward funerary monuments would cross chamber boundaries with less frequency than views toward randomly distributed targets. In order to verify this hypothesis, a separate network linking monuments to potential observers was created (Figure 14). Technically, this is a two-mode network, connecting two distinct categories of objects or beings (observers and monuments in our case; Borgatti, 2009; Filet & Rossi, 2023). Locations for observers were borrowed from the intervisibility network used for landscape chambers segmentation, and the same parameters were chosen as above, namely the monument height of one metre and the same observer eye level, while the maximum view reach was set to 3.5 kilometres. This approach ensures the direct comparability between the two datasets. Finally, unlike the previously discussed model, only full views of view targets were retained here, i.e. the eye-to-ground links.
Views between random observers | Views of monuments by random observers | |
Number of nodes | 15 751 random observers (in the study area perimeter) | 8 307 random observers with a view to a monument 431 monuments |
External links ratio | 5.6% | 7.1% |
Observers with one or more external links | 34% | 14% of observers |
Targets with one or more external links | not applicable (observers act as view targets) | 63% of monuments |
Rather surprisingly, the visibility network of funerary monuments contains a somewhat higher percentage of external links than the network between random observers (7.1% vs 5.6%; Table 1). This was calculated for visual links situated entirely within the study area. Of particular intertest is the markedly high transgression score of visual links towards monuments, 63% of structures were theoretically visible by at least one observer situated in another landscape chamber. Nevertheless, only 14% of our random observers would have been able to observe a funerary monument outside of their (digitally modelled) landscape chambers. Such marked discrepancy is in part related to the numerical disproportion where some eight thousand observers would have had only four hundred objects to look at. However, this cannot explain the entire range of difference as in the randomly generated network only 34% of locations maintain external connections. The search for visual exposure of funerary monuments should account for a non-negligible part of the connectivity pattern of prehistoric burial grounds.[6]
Funerary monuments thus do not tend to be visually isolated in their respective chambers, on the contrary, they have somewhat larger and boundary defiant visual reach in comparison to random observers. The model of landscape arena, even if clearly relevant for specific areas, such as the environs of Marigny-le-Châtel, still cannot be extended to the entirety of the prehistoric funerary landscape. If this were the case, we would expect lower than average scores of external links. Indeed, a number of burial grounds are placed on ridges through which run chamber boundaries and for this reason maintain visual connections with at least two adjacent landscape chambers (Figure 14, inset).
It thus becomes apparent that prehistoric burial grounds were distributed across a variety of topographic settings; while some are indeed visually integrated with their surroundings, many others are on hilltop or ridgetop locations, visually exposed to multiple landscape chambers. Hypothetically, such choices may indicate a preference for visual exposure over visual integration, i.e. less concern with the visual coherence of the surrounding landscape than with the reach of visual signal (up to 3.5 km in our approach). In any case, this result warns us that the search for visual integration within landscape arenas is only a part of a complex system of landscape preferences and choices.
Discussion
The visual analysis of the prehistoric Sénonais revealed several landscape patterns. Social groups and communities who constructed ditched funerary monuments clearly favoured visually exposed locations, as demonstrated by the visibility index analysis. On average, the area to which a monument is visually exposed tends to be 8 % larger than the general level of landscape visibility. Visually integrated zones also appear as particularly attractive, which was measured through the visual quality index (VQI). The landscape chamber of Marigny-le-Châtel in particular combines high VQI score and high density of prehistoric monuments. Finally, the analysis of visual connections of individual monuments demonstrated that the complexity of visual patterns may result from diverging or even mutually exclusive preferences. While the search for visual integration within tentative landscape arenas seems relevant for specific areas, in general burial grounds tend to visually communicate to disparate landscape settings, across landscape chamber boundaries.
These visual patterns of the prehistoric Sénonais should be understood in the context of contemporary settlement pattern. Unfortunately, the data on Bronze Age settlements remain particularly scarce for the study area, but we can still extrapolate several insights from the nearby Seine and Yonne river valleys, where archaeological research has been rather intense since the 1960s. There, burial grounds tend to grow and multiply in numbers in areas of high settlement density (Mordant & Gouge, 1992), as elsewhere in the Parisian Basin (Brun et al., 2005; Marcigny, 2012). We can thus assume a certain degree of spatial correlation between areas favoured by the living and those chosen for the dead in the Sénonais. This is not to say that prehistoric burials directly map inhabited areas; indeed, it has been stressed recently that specific locations in the Seine-Yonne confluence gather more burials than expected, perhaps because of associated prestige or other symbolic values (Auxerre-Guéron et al., 2022). For purposes of our analysis, the dispersed pattern of Bronze and Iron Age settlements seems particularly interesting as it may provide clues for the equally loose spatial distribution of burial grounds. We may perhaps envisage that burial places catered to social practices which favoured the general dispersion and mobility of the population in the local range. Indeed, the visual presence of funerary monuments must have been felt across the entirety of the Bronze Age landscape, as if they responded to a need to overlay the inhabited space with reminders to the deceased. Arguably, this may be related to the segmented social structure of Bronze Age societies and the need to link specific portions of the land to specific ancestors (Chapman, 1995; Parker Pearson, 2003, p. 132ff). In any case, such scattered patterns are characteristic for much of the Bronze Age funerary landscapes in Central and Northern Europe, from Britain (Woodward & Woodward, 1996), through Belgium and Netherlands (Q. Bourgeois, 2012) to Scandinavia (Johansen et al., 2004). In all these areas, Late Prehistoric settlements are markedly dispersed and most likely transient, shifting their location as one generation succeeded another (Gerritsen, 2003). Without delving further into this topic, we can assume that the complexity and more specifically the dispersion of visual discourse maintained by prehistoric funerary monuments of the Sénonais plateau would be related to such dispersed spatial practices. These could have included subsistence activities such as herding and agriculture, periodic settlement relocation or the competition for access to the land.
Returning to the landscape chambers method, it is fundamental to fully understand its conceptual underpinnings and its limits for a historical approach. Indeed, the historical existence and, by that token, the relevance of the model produced for the Sénonais landscape still remains unclear. What do landscape chambers represent? Our common intuition tends to evaluate the visual quality of a landscape in terms of impact and drama. We tend to think of wide views and prominent landmarks. Indeed, the concept of landscape arenas may fit into this line of thinking (see Introduction). Nevertheless, visual drama and visual impact do not figure among the criteria taken into account here. Landscape chambers of the unassuming, undulating Sénonais plateau have been modelled on the basis of visual integration only, i.e. network connectivity, and evaluated in terms of visual coherence (VQI). Even the most coherent chambers, such as the one of Marigny-le-Châtel, may not appear particularly picturesque or impressive in the rather ordinary, flatland landscape. Visual coherence should not be confused with visual impact.
Following the insights of the research on human spatial practice and experience, in particular the space syntax, we need to acknowledge the importance of visual experience beyond the immediate visual impact. The configuration of the visual field shapes what and whom we may see while moving through the space, which is clearly relevant for social interaction and spatial cognition. Even when subtle and mostly unrecognised, such landscape qualities may nevertheless exert a profound effect on social life (navigability, intelligibility of space, etc.: Hillier, 2004). Our method has been specifically devised to evaluate such landscape visual qualities regardless of their impact or indeed their acknowledgement by the inhabitants. In all likelihood many if not the most of our chambers were not recognised by past populations. Therefore, the landscape chambers method does not necessarily trace historical realities, such as mental geographies or political territories: it is the archaeologist’s task to find out when and by what means these landscape features entered the cultural repertoire, if they did. A range of scenarios may be envisaged, from vague, general visual experience unrelated to cultural patterns, to fully acknowledged places and topographic entities with clear markers of cultural significance.
The prehistoric Sénonais landscape is particularly interesting from this point of view. Lacking dramatic views and pronounced topography, it does not provide an easily recognisable structure that can be safely mapped to a theoretical cognitive scheme. While archaeologists working on landscape experience tend to focus on unambiguous landmarks and topographic features (mountain peaks, landscape arenas, etc.; see the Introduction), we are confronted here with the task of determining what cognitive units may have actually existed in the first place. This can only be achieved by establishing a correlation between a specific landscape character and specific cultural patterns. Indeed, this is the case with the Marigny-le-Châtel chamber, which hosts a high-density cluster of prehistoric monuments and displays specific visual qualities, namely high visual integration. If funerary monuments are understood as visual signals, this chamber offered a configuration which is particularly conducive to such visual discourse. Its high visual integration offered an arena where monuments could be juxtaposed to each other, while conveying a salient effect when seen at distance, en masse (provided an open, grassland landscape: see Introduction). Nevertheless, the existence of a cognitive unit referring to this small watershed basin cannot be taken for granted, considering its rather ordinary topography and the lack of remarkable topographic features. Rather than interacting with a clearly delimited and easily identifiable landscape unit, the inhabitants of the prehistoric Sénonais may have been, perhaps inadvertently, participating in creation and individuation of a cognitive spatial unit. The construction of funerary monuments can thus be seen not as a reply to a pre-existing landscape, but rather as a historical process through which the cognitive landscape was taking shape. Even if we assume that such a process lacked any specific plan or rationale, the long-term practice of funerary monument construction, shaped or canalised by visual properties of the space, would have stimulated the appearance of a meaningful landscape unit.
Let us finally examine several methodological issues which could be crucial for further development of the landscape chambers method. Perhaps the most problematic part of the approach presented here is the reliance on the VQI which can hardly account for all subtleties of human visual experience. A number of measurements may be included, for instance the analysis of visual horizon (Lake & Woodman, 2003), the vector field approach which measures the directionality of views (Llobera, 2003), as well as various indices of network structure within landscape chambers (Brughmans & Brandes, 2017). Furthermore, the formula used for the Sénonais plateau study may not be appropriate for other landscape configurations (mountains, littoral zone, etc.). Indeed, it tends to produce very high scores in flat areas, in comparison to lower scores obtained for the more pronounced topography of the Sénonais. Therefore, the VQI approach is by no means the general solution for landscape characterisation, but rather a heuristics which will most probably require further adjustments for different landscape contexts.
In the heart of the landscape chambers method lies a community detection algorithm, based on Newman’s modularity (Newman, 2006). However, network segmentation into individual clusters is not a trivial problem. Clearly, certain clusters may be more tightly connected than others, they also may include a number of sub-clusters. This was demonstrated on the example of chamber n° 11 (Figure 9), which encompasses a number of sub-clusters that follow river and stream valleys. Therefore, the problem of network segmentation can be represented as the choice of a break-point within a continuous hierarchy of cohesive groups, morphing with each other as we move from local to global level. Certain software solutions permit to specify these breakpoints, and it remains to be seen whether such an approach would produce superior results (e.g. Networkx software: Hagberg, A. et al., 2008). Equally, more sophisticated algorithms that operate on directed networks, i.e. nonreciprocal viewership in our case, are also available and would merit an assessment (e.g. Csardi, G. & Nepusz, T., 2006). Furthermore, we should not forget that visibility networks are geographical in character and that a number of spatial parameters may be taken into account when evaluating the strength of a connection, such as distance or direction. This would call for highly sophisticated solutions which are not readily available in common software packages.
Conclusion
Fundamentally, landscape is a cognitive category, given that a piece of inhabited land cannot be isolated from representations, knowledge and memories through which it was approached and shaped. A path or a field boundary define not only our physical interaction with space, but also our mental representation of it, and may be placed or constructed accordingly. However, archaeological landscapes are all too often assimilated to an unproblematic collection of objects which exist by themselves and simply await to be perceived and cognised by an observer. Whether of cultural or natural origin, for instance a burial mound or a mountain peak, these objects tend to be considered as a priori knowable, because salient or otherwise self-evident. Nevertheless, much of the landscape is most commonly difficult to define and classify, especially its land forms: when should a prominence be considered a hill, a depression a valley? The problem resides in the degree of specificity or distinctiveness which clearly varies across the landscape.
The landscape chambers method was developed to tackle this issue, namely to characterise the landscape in terms of its visual integration. It is important to bear in mind that the purpose of the method is not to simply segment the landscape in visually distinct chambers, but rather to characterise such spaces, i.e. to measure the level of their visual distinctiveness and coherence. Therefore, the landscape chambers should not be considered as convenient spatial containers, but rather as areas of specific visual character, expressed in quantitative and/or qualitative terms. In our case, dealing with prehistoric burial places of the Parisian Basin, such character was evaluated though the Visual Quality Index (VQI), which quantifies visual connectivity within and across chamber boundaries. This approach revealed visually highly integrated landscape chambers, and provided valuable information on the relationship between such visual character and the development of prehistoric funerary landscapes. Of particular interest is the lack of visual drama or otherwise unique visual experience across the gently undulating calcareous plateau of the Sénonais; there is little ground to consider naturally determined, highly salient landscape features and spaces, but rather a continuum of possibilities or affordances for topographical meaning-making (sensu Gibson, 1986). The appearance of clusters of burial places in specific landscape chambers may thus testify to a historical process of individuation of meaningful landscape units. Unfortunately, the available archaeological data for the Sénonais plateau does not permit to trace the historical evolution of the burial landscape in much detail, although such line of inquiry would be highly relevant for the future study. Indeed, the increasing availability of radiocarbon dates for European prehistoric funerary monumets (3rd – 1st millenium BC) indicates very long sequences of both, the use of individual monuments (Q. P. J. Bourgeois & Fontijn, 2015; Last, 1998), and the evolution of the funerary landscape (Garwood, 2007; Peake & Delattre, 2005). Such memorial landscapes where one was frequently exposed to and reminded of the ancestors, or simply predecessors, evolved through long term negotiation of the landscape’s semiotic and memorial content.
On the conceptual level, the landscape chambers method is based on the idea of the rich space, i.e. a space replete with diverse configurations and connections that guide and inform human practice and cognition. Space syntax researchers, for instance, propose more than two dozen indices of connectivity, controllability or navigability of the architectural space, all based on visibility networks (Koutsolampros et al., 2019). Nevertheless, the space of much of archaeological landscape research remains rather poor, typically considered as an empty void between objects and often construed as an abstract, mathematical construct. Indeed, distance and surface are quite commonly the only variables used to characterise the archaeological space (e.g. territorial area, distance to resources, etc.). Human experience, however, is based on more complex insights and observations. The landscape chambers method is introducing one such indicator, visual connectivity, which is expressing the relationship of each location to the rest of the open space. It is also introducing, albeit in a rudimentary manner, the idea of hierarchical relationships between spaces which combine together as we move from local to global scale.
That being said, the assessment of landscape visual character is clearly an exceedingly complex issue and the approach provided here, based on three rather simple indices (density, border transgression and modularity of intervisibility sub-networks), cannot satisfy all needs for the analysis of archaeological spaces and landscapes. There is an infinitely rich array of such indices that remain to be tackled, in relation to an equally wide array of spatial practices (subsistence activities, trade and exchange, warfare and security, etc.). The prospects for the future research are wide and, it is hoped, encouraging for further development of the landscape chambers method.
Bibliography
Ahnert, S. E., & Fink, T. M. A. (2008). Clustering signatures classify directed networks. Physical Review E, 78(3), 036112. https://doi.org/10.1103/PhysRevE.78.036112
Alcantara, A. (2015). Marigny-le-Châtel (10) « Le Saussoir à Jollier ». Excavation report, Archéodunum SAS.
Auxerre-Guéron, F.-A., Dubuis, B., Gouge, P., Mordant, Cl., Nouvel, P., Peake, R., Riquier, V., & Roscio, M. (2022). Modèles d’occupation et nécropoles de l’âge du Bronze de la Haute Seine. Unpublished paper presented at « Bronze 2022 » APRAB conference: « Modèles d’occupation du sol à l’âge du Bronze en Europe », June 22, 2022 Metz (France).
Aynaud, Th. (2018). Louvain Community Detection. GitHub repository [Computer software]. https://github.com/taynaud/python-louvain
Baray, L., Deffressigne, S., Leroyer, C., & Villemeur, I. (1994). Nécropoles protohistoriques du Sénonais. Serbonnes/La Créole, Michery/La Longue Raie, Soucy/Mocques Bouteilles (Yonne). Maison des sciences de l’homme.
Basso, K. (1996). Wisdom Sits in Places. Landscape and Language among the Western Apache. University of New Mexico.
Bernardini, W., & Peeples, M. A. (2015). Sight Communities: The Social Significance of Shared Visual Landmarks. American Antiquity, 80(2), 215–235. https://doi.org/10.7183/0002-7316.80.2.215
Borgatti, S. P. (2009). Social Network Analysis, Two-Mode Concepts in. In R. A. Meyers (Ed.), Encyclopedia of Complexity and Systems Science (pp. 8279–8291). Springer. https://doi.org/10.1007/978-0-387-30440-3_491
Bourgeois, Q. (2012). Monuments on the Horizon: The Formation of the Barrow Landscape Throughout the 3rd and 2nd Millennium BC. Sidestone Press.
Bourgeois, Q. P. J., & Fontijn, D. R. (2015). The Tempo of Bronze Age Barrow Use: Modeling the Ebb and Flow in Monumental Funerary Landscapes. Radiocarbon, 57(1), 47-64. https://doi.org/10.2458/azu_rc.57.17925
Bradley, R. (1998). The Significance of Monuments: On the Shaping of Human Experience in Neolithic and Bronze Age Europe. Routledge.
Bradley, R. (2002). The Past in Prehistoric Societies. Routledge.
Brughmans, T., & Brandes, U. (2017). Visibility Network Patterns and Methods for Studying Visual Relational Phenomena in Archeology. Frontiers in Digital Humanities, 4. https://doi.org/10.3389/fdigh.2017.00017
Brughmans, T., Keay, S., & Earl, G. (2015). Understanding Inter-settlement Visibility in Iron Age and Roman Southern Spain with Exponential Random Graph Models for Visibility Networks. Journal of Archaeological Method and Theory, 22(1), 58-143. https://doi.org/10.1007/s10816-014-9231-x
Brughmans, T., van Garderen, M., & Gillings, M. (2018). Introducing visual neighbourhood configurations for total viewsheds. Journal of Archaeological Science, 96, 14–25. https://doi.org/10.1016/j.jas.2018.05.006
Brun, P., Cathelinais, C., Chatillon, S., Guichard, Y., Le Guen, P., & Néré, E. (2005). L’âge du Bronze dans la Vallée de l’Aisne. In J. Bourgeois & M. Talon (Eds.), Actes des congrès nationaux des sociétés historiques et scientifiques, 125″, Lille 2000 (pp. 189–208). Éditions du CTHS.
Chapman, R. (1995). Ten Years After—Megaliths, Mortuary Practices, and the Territorial Model. In L. A. Beck (Ed.), Regional Approaches to Mortuary Analysis (pp. 29–51). Plenum Press. https://doi.org/10.1007/978-1-4899-1310-4_2
Chaume, B., Reinhard, W., & Wustrow, G. (2007). Les dépôts de l’enclos cultuel hallstattien de Vix ” les Herbues ” et la question des enceintes quadrangulaires. Bulletin de la Société préhistorique française, 104(2), 343–367. https://doi.org/10.3406/bspf.2007.13561
Criado Boado, F., & Villoch Vázquez, V. (2000). Monumentalizing Landscape: From Present Perception to the Past Meaning of Galician Megalithism (North-West Iberian Peninsula). European Journal of Archaeology, 3(2), 188–216. https://doi.org/10.1177/146195710000300203
Csardi, G. & Nepusz, T. (2006). The Igraph software package for complex network research. International Journal of Complex Systems, 1695(5), 1-9.
Čučković, Z. (2016). Advanced viewshed analysis: A Quantum GIS plug-in for the analysis of visual landscapes. The Journal of Open Source Software, 1(4), 32. https://doi.org/10.21105/joss.00032
Čučković, Z. (2023). Visibility Networks. In T. Brughmans, B. J. Mills, J. Munson, & M. A. Peeples (Eds.), The Oxford Handbook of Archaeological Network Research (1st ed., pp. 230–247). Oxford University Press. https://doi.org/10.1093/oxfordhb/9780198854265.013.13
De Reu, J., Bourgeois, J., De Smedt, P., Zwertvaegher, A., Antrop, M., Bats, M., De Maeyer, P., Finke, P., Van Meirvenne, M., Verniers, J., & Crombé, P. (2011). Measuring the relative topographic position of archaeological sites in the landscape, a case study on the Bronze Age barrows in northwest Belgium. Journal of Archaeological Science, 38(12), 3435–3446. https://doi.org/10.1016/j.jas.2011.08.005
Deffressigne, S., & Villes, A. (1995). Estissac (Aube). ‘La côte d’Ervaux’, sépulture à char. In Fastes des Celtes anciens. Catalogue d’exposition (26 mai—4 septembre 1995) (pp. 59–67). Musée de Troyes ; Musée de Nogent-sur-Seine.
Delattre, V., & Peake, R. (2013). Un millénaire de comportements funéraires dans la moyenne vallée de la Seine: Une meilleure approche du monde des vivants à l’âge du Bronze ? In J. Jaubert, N. Fourment, & P. Depaepe (Eds.), Transitions, ruptures et continuité en Préhistoire (Vol. 1). Actes du XXVIIe Congrès préhistorique de France, Bordeaux – Les Eyzies, 31 mai-5 juin 2010 (pp. 303–315). Société Préhistorique française.
Delor, J.-P. (Ed.). (2002). L’Yonne (Cartes archéologiques de la Gaule, 89). Académie des Inscriptions et Belles Lettres, Ministère de la Recherche, Ministère de la Culture et de la Communication, Conseil Général de l’Yonne, Maison des sciences de l’Homme.
Dungan, K. A., White, D., Déderix, S., Mills, B. J., & Safi, K. (2018). A total viewshed approach to local visibility in the Chaco World. Antiquity, 92(364), 905–921. https://doi.org/10.15184/aqy.2018.135
Filet, C., & Rossi, F. (2023). Network Methods and Properties. In T. Brughmans, B. J. Mills, J. Munson, & M. A. Peeples (Eds.), The Oxford Handbook of Archaeological Network Research (1st ed., pp. 15–33). Oxford University Press. https://doi.org/10.1093/oxfordhb/9780198854265.013.1
Filipiak, B. (2018). Grand Est, Aube, Marigny-le-Châtel ‘Chemin de la Pèze’, ‘Le Pont de Riom’. Une importante nécropole de l’étape initiale du Bronze final, des fosses de piègeage du Néolithique ancien et un habitat de l’époque médiévale. Excavation report, Inrap Grand-Est.
Filipiak, B., Peake, R., Le Goff, I., & Paresys, C. (2017). Marigny-le-Châtel «Chemin de la Pèze», «Pont de Riom« (Aube), «L’Epine Gérard». Une importante nécropole du Bronze final découverte sur la tracé du gazoduc «Arc de Dierrey». Bulletin APRAB, 15, 23–32.
Fortunato, S., & Barthélemy, M. (2007). Resolution limit in community detection. Proceedings of the National Academy of Sciences, 104(1), 36–41. https://doi.org/10.1073/pnas.0605965104
Fry, G. L. A., Skar, B., Jerpåsen, G., Bakkestuen, V., & Erikstad, L. (2004). Locating archaeological sites in the landscape: A hierarchical approach based on landscape indicators. Landscape and Urban Planning, 67(1–4), 97–107. https://doi.org/10.1016/S0169-2046(03)00031-8
Gansum, G. L. A., Jerpåsen, G., & Keller, Ch. (1997). Arkeologisk landskapsanalyse med visuelle metoder. Arkeologisk museum i Stavanger.
Garwood, P. (2007). Before the Hills in Order Stood: Chronology, time and history in the interpretation of Early Bronze Age round barrows. In J. Last (Ed.), Beyond The Grave. New Perspectives On Barrows (pp. 30–52). Oxbow Books.
Gerritsen, F. (2003). Local Identities: Landscape and Community in the Late Prehistoric Meuse-Demer-Scheldt Region. Amsterdam University Press. https://doi.org/10.5117/9789053565889
Gerritsen, F. (2008). Domestic Times: Houses and Temporalities in Late Prehistoric Europe. In A. Jones (Ed.), Prehistoric Europe. Theory and Practice (pp. 143–161). Wiley-Blackwell.
Gibson, J. J. (1986). The Ecological Approach to Visual Perception. Lawrence Erlbaum Associates.
Gillings, M., Hacıgüzeller, P., & Lock, G. (2020). Archaeology and spatial nalaysis. In P. Hacıgüzeller, M. Gillings, & G. Lock (Eds.), Archaeological spatial analysis: A methodological guide. (pp. 1–16). Routledge.
Gouge, P., Mordant, C., & Pihuit, P. (1994). Nécropoles de la Bassée. Age du Bronze. Centre départemental d’Archéologie de la Bassée.
Hagberg, A., Schult, D., & Swart, P. (2008). Exploring network structure, dynamics, and function using NetworkX. In Varoquaux, G., Vaught, T., & Millman, J. (Eds.), Proceedings of the 7th Python in Science Conference (pp. 11–15). http://conference.scipy.org.s3-website-us-east-1.amazonaws.com/proceedings/scipy2008/paper_2/
Hillier, B. (2004). Space is the machine. A configurational theory of architecture. Space Syntax Limited and UCL.
Hillier, B. (2014). Spatial analysis and cultural information: The need for theory as well as method in space syntax analysis. In E. Paliou, U. Lieberwirth, & S. Polla (Eds.), Spatial analysis and social spaces. Interdisciplinary approaches to the interpretation of prehistoric and historic built environments (pp. 19–48). De Gruyter.
Hillier, B., & Hanson, J. (1984). The Social Logic of Space. Cambridge University Press.
IGN. (2013). BD ALTI, Version 1. Institut national de l’information géographique et forestière. https://geoservices.ign.fr/sites/default/files/2021-07/SE_BDALTI.pdf
Jerpåsen, G. B. (2009). Application of Visual Archaeological Landscape Analysis: Some Results. Norwegian Archaeological Review, 42(2), 123–145. https://doi.org/10.1080/00293650903351052
Johansen, K. L., Laursen, S. T., & Holst, M. K. (2004). Spatial patterns of social organization in the Early Bronze Age of South Scandinavia. Journal of Anthropological Archaeology, 23(1), 33–55. https://doi.org/10.1016/j.jaa.2003.10.002
Kim, Y.-H., Rana, S., & Wise, S. (2004). Exploring multiple viewshed analysis using terrain features and optimisation techniques. Computers & Geosciences, 30(9–10), 1019–1032. https://doi.org/10.1016/j.cageo.2004.07.008
Koutsolampros, P., Sailer, K., Varoudis, T., & Haslem, R. (2019). Dissecting Visibility Graph Analysis: The metrics and their role in understanding workplace human behaviour. Proceedings of the 12th International Space Syntax Symposium (12 SSS). Space Syntax Limited. https://discovery.ucl.ac.uk/id/eprint/10073528/
Lafage, F., Auxiette, G., Brunet, P., Martial, E., Matterne, V., Praud, I., & Laplantine, N. (2006). Premières tentatives d’interprétation spatiale d’un site rural du Bronze final à Changis-sur-Marne (Seine-et-Marne). Bulletin de la Société préhistorique française, 103(2), 323–377. https://doi.org/10.3406/bspf.2006.13437
Lake, M., & Ortega, D. (2013). Compute-Intensive GIS Visibility Analysis of the Settings of Prehistoric Stone Circles. In A. Bevan & M. Lake (Eds.), Computational Approaches to Archaeological Spaces (pp. 213–242). Left Coast Press.
Lake, M., & Woodman, P. E. (2003). Visibility Studies in Archaeology: A Review and Case Study. Environment and Planning B: Planning and Design, 30(5), 689-707. https://doi.org/10.1068/b29122
Last, J. (1998). Books of Life: Biography and Memory in a Bronze Age Barrow. Oxford Journal of Archaeology, 17(1), 43–53. https://doi.org/10.1111/1468-0092.00050
Leroyer, C., David, R., Mazier, F., Allenet de Ribemont, G., Lanos, Ph., & Dufresne, Ph. (2012). Environnement et anthropisation du milieu durant l’âge du Bronze dans le Bassin Parisien: L’apport de des données polliniques et de la modelisation du couvert végétal. In M. Mélin & C. Mougne (Eds.), Actes du Séminaire Archéologique de l’Ouest, 22 mars 2012, Université de Rennes 1 (pp. 7–26). Université de Rennes.
Llobera, M. (1996). Exploring the topography of mind: GIS, social space and archaeology. Antiquity, 70(269), 612–622. https://doi.org/10.1017/S0003598X00083745
Llobera, M. (2003). Extending GIS-based visual analysis: The concept of visualscapes. International Journal of Geographical Information Science, 17(1), 25–48. https://doi.org/10.1080/713811741
Llobera, M. (2007). Reconstructing visual landscapes. World Archaeology, 39(1), 51-69. https://doi.org/10.1080/00438240601136496
Llobera, M. (2012). Life on a Pixel: Challenges in the Development of Digital Methods Within an “Interpretive” Landscape Archaeology Framework. Journal of Archaeological Method and Theory, 19(4), 495-509. https://doi.org/10.1007/s10816-012-9139-2
Marcigny, C. (2012). Lieux funéraires, paysages et territoires de l’âge du Bronze en Normandie. In D. J. Bérenger, J. Bourgeois, M. Talon, & S. Wirth (Eds.), .) Gräberlandschaften der Bronzezeit / Paysages funéraires de l’âge du Bronze. Colloque international sur l’âge du Bronze, Herne, 15-18 octobre 2008 (pp. 595–617). Philip von Zabern.
Mordant, C., & Gouge, P. (1992). L’occupation du sol au Bronze final dans les vallées de l’Yonne et de la Seine. In C. Mordant & A. Richard (Eds.), Actes du Colloque international de Lons-le-Saunier, 15 -19 mai 1990 (pp. 133–164). Éditions du CTHS.
Mordant, C., & Mordant, D. (1970). Le site protohistorique des Gours-aux-Lions à Marolles-sur-Seine. Société préhistorique française.
Mordant, C., & Nouvel, P. (2017). Les nécropoles protohistoriques à enclos fossoyés de Bourgogne du nord et de Champagne méridionale: Apport de la prospection aérienne. In L. Baray (Ed.), De silex, de terre et de faïence, la mémoire du sol en Bourgogne. Hommage à Jean-Paul Delor (pp. 103–119). Editions Mergoil.
Newman, M. E. J. (2006). Modularity and community structure in networks. Proceedings of the National Academy of Sciences, 103(23), 8577–8582. https://doi.org/10.1073/pnas.0601602103
Ogburn, D. E. (2006). Assessing the level of visibility of cultural objects in past landscapes. Journal of Archaeological Science, 33(3), 405–413. https://doi.org/10.1016/j.jas.2005.08.005
O’Sullivan, D., & Turner, A. (2001). Visibility graphs and landscape visibility analysis. International Journal of Geographical Information Science, 15(3), 221–237. https://doi.org/10.1080/13658810151072859
Parker Pearson, M. (2003). The Archaeology of Death and Burial. Sutton Publishing.
Peake, R. (Ed.). (2020). Villiers-sur-Seine—Un habitat aristocratique du IXe siècle avant notre ère. INRAP / CNRS éditions.
Peake, R., & Delattre, V. (2005). L’apport des analyses 14 C à l’étude de la nécropole de l’âge du Bronze de“ La Croix de la Mission ” à Marolles-sur- Seine. Revue archéologique du Centre de la France, 44, 5–25.
Peake, R., Roscio, M., Issenmann, R., Riquier, V., & Simonin, D. (2017). Le Bronze moyen et le début du Bronze final dans le sud du Bassin parisien. In T. Lachenal, Cl. Mordant, T. Nicolas, & C. Vébér (Eds.), Actes du colloque APRAB « Bronze 2014 » (Strasbourg, 17-20 juin 2014) (pp. 195–216). Association pour la Valorisation de l’Archéologie du Grand Est.
Penn, A. (2003). Space Syntax And Spatial Cognition: Or Why the Axial Line? Environment and Behavior, 35(1), 30–65. https://doi.org/10.1177/0013916502238864
Roscio, M., & Muller, F. (2012). Les nécropoles de l’Yonne au Bronze Final: Bilan critique des connaissances. In D. J. Bérenger, J. Bourgeois, M. Talon, & S. Wirth (Eds.), Gräberlandschaften der Bronzezeit / Paysages funéraires de l’âge du Bronze. Colloque international sur l’âge du Bronze, Herne, 15-18 octobre 2008 (pp. 345–354). Philip von Zabern.
Rottier, S. (2009). Fonctionnement des tombes du début du Bronze final (XIVe – XIIe s. Av. J.C.) dans le sud-est du Bassin parisien (France). Bulletins et mémoires de la Société d’Anthropologie de Paris, 21(1–2), 19–46.
Shang, H., & Bishop, I. D. (2000). Visual thresholds for detection, recognition and visual impact in landscape settings. Journal of Environmental Psychology, 20, 125–140.
Tilley, Ch. (1994). A Phenomenology of Landscape Places, Paths and Monuments. Berg.
Turk, A. G., Mark, D. M., & Stea, D. (2011). Ethnophysiography. In D. M. Mark, A. G. Turk, N. Burenhult, & D. Stea (Eds.), Landscape in Language. Transdisciplinary perspectives (pp. 25–45). John Benjamins Publishing Company.
Vatan, A., Delattre, V., & Peake, R. (2001). Marolles-sur-Seine ‘Le Tureau des Gardes’ (Seine-et-Marne). Centre départemental d’archéologie de la Bassée, Service régional d’archéologie d’Île de France.
Woodward, A. B., & Woodward, P. J. (1996). The Topography of some Barrow Cemeteries in Bronze Age Wessex. Proceedings of the Prehistoric Society, 62, 275–291. https://doi.org/10.1017/S0079497X00002814
Notes
E.g. “…space cannot exist apart from the events and activities within which it is implicated” (Tilley, 1994, p. 10) ; ”What if the space of Mesolithic was different from the space of the Bronze Age…” (Gillings et al., 2020, p. 10). See also Llobera (2012) for a detailed discussion. ↑
Supplied in integer numeric format, which introduces a stepped effect, the DTM had to be smoothed using an average filter (which replaces original height values with the average values of their neighbourhood). ↑
Regarding the view direction, our analysis quantifies views received by each raster cell (as opposed to views realised from observer points). ↑
- Statistical t-test indicates highly significant difference between the two samples (p<0.0001 ; mean_1 = .34 ; StDev_1 = .11, mean_2 = .26, StDev_2 = .11, N > 100 for both samples). We can safely infer that monuments were more often placed on visually exposed locations than in a random choice scenario. ↑
More specifically, this concerns chambers 0, 1, 8, 13, 15, 16, 18, 20, 22, 25, the extent of which is predominantly situated outside of the study area perimeter. ↑
We should remark that due to relatively large number of observer points, the results of this analysis tend to converge with those obtained through visibility index calculation, namely in terms of visual connectivity of funerary monuments. Further statistical analysis is thus not deemed necessary. ↑