Landscapes of Movement Along the (Pre)Historical Libyan Sea: Keys for a Socio-Ecological History

2.2.1 Mutual Agency

HE renders a dialectical interrelationship between the human and non-human agents of the Earth’s biophysical system to explain how socio-natural systems operate (Marquardt, 2019). As stated by McGlade (1995), “structures are maintained only by constant energy/matter and entropy exchange, in which the action of positive feedbacks of environmental and cultural processes drive the system to new evolutionary states, which in turn provide the conditions for renewal of higher entropy production.” In other words, HE emphasises the interconnectedness between the human and non-human agents of socio-natural systems and how they mutually shape and influence one another through constant exchange and feedback loops.

Hence, it follows that analysing socio-natural systems requires envisioning these systems as composites, wherein each component can potentially interact with every other component. This has both epistemological and ontological implications. Following Ingerson (1994, p. 45), we should understand the culture/nature dichotomy as “quantitative variations along a single spectrum rather than as an either/or dichotomy [which, as a consequence,] requires us to disassociate qualities and characteristics” traditionally labelled either as “social” or “natural.” We find the concept of mutual agency quite suitable in this context. Mutual agency implies that humans and non-humans are deeply entangled through their agency. Indeed, as Bauer (2018; also, Bauer & Ellis, 2018) argued, it is through the recognition of this bidirectionality that we can correctly understand the ontological intertwinement between humans and non-humans. Moreover, such mutuality through bidirectional connections allows the system to change and, therefore, evolve. Thus, the synergy between humans and non-humans drives co-evolutional dynamics, which leads to arguing that history might be conceptualised as the coevolution of multiple species (Fitzhugh et al., 2018).

Understanding the synergy generated by the mutual agency of human and non-human agents is the task of HE (Maher & Harrison, 2014, pp. 2–3). To do so, HE incorporates an approach based on Complex Adaptive Sciences (CAS). The approach of CAS is not so much placed on the agents per se, but rather on the holistic picture derived from the agents’ interactions, changes, and patterns at multiple scales (Sinclair et al., 2018). These interactions and their governing mechanisms are thus of utmost interest; two other concepts in HE can be quite useful to describe them: non-linearity and heterarchy.

2.2.2 Non-Linearity

HE applies a reasoning based on non-linear dynamics of causality (McGlade, 1995). Complex dynamical systems, such as socio-cultural or biocultural ones, have the capacity to generate unpredictable, emergent behaviour via non-linear, especially chaotic, feedback. A direct consequence is that the systemic dynamics of causality hinge on an (initially) indeterminate number of agents who vary in number and effect depending on the moment, place, and scale of analysis. As such, “ecosystem structure is viewed as a series of weakly coupled sets within a hierarchy of process rates involving biotic interaction […] and abiotic factors […]. The non-linear couplings in these processes are further complicated by human action” (McGlade, 1995, p. 121). To understand these complex dynamics, it is essential to decompose the global picture into smaller parts, following a hierarchical analytical approach – as McGlade uses it within the context of hierarchy theory –, and gear them into a concrete, broad narrative.

This approach to causal explanations is fundamental to interpreting human–environment entanglements. Non-linearity accounts for the interlaced dynamics of complex systems, offering a more nuanced and comprehensive understanding of the interactions that occur within socio-natural systems (Maher & Harrison, 2014, p. 2; McGlade, 1995; Winterhalder, 1994). It can be said that non-linearity is one of the foundational governing principles driving the mutual agency of human and non-human agents.

2.2.3 Heterarchy

If non-linearity governs human–environment interrelationships, heterarchy is a type of organisation that defines the character of such interrelations for HE. Heterarchy defines a complex system that allows elements to be ranked or unranked (relative to other elements) in several ways depending on systemic requirements, creating a dynamic structure in which power and influence are distributed among various agents (Crumley, 1994b, p. 12). This is in contrast to hierarchies, where elements are ranked in a fixed order. In a heterarchy, the relative importance and behaviour of agents, whether human or non-human, are not constant but fluctuate depending on spatial and temporal contexts, leading to a flexible and adaptive structure of human–environment interrelationships.

As previously mentioned, non-linear bidirectional interrelationships involve multiple agents operating within a complex system across diverse spatial and temporal scales. The power that agents – be they human or not – can effect upon others is not static but variable; at times, they might be stimulators/dominant or respondents/dominated (Crumley, 2019, p. 293). Consequently, the relative behaviour and importance of agents are not static; their actions, interactions, and results fluctuate in response to changing circumstances, leading to a context-sensitive and adaptable system.

2.2.4 Multiscalar: Long-Term, Large (Distributed) Scale

Historical ecologists have been vocal about the importance of using multiscalar archaeological data to gain a comprehensive multitemporal and multi-scalar understanding of human–environment interrelationships (Hambrecht et al., 2020; McGovern, Hambrecht, & Hicks, 2019; Ray, 2019). According to this approach, the landscape is the fundamental unit of analysis from which to develop a framework that includes distributed (local, regional, interregional, and supra-regional), longitudinal, and multitemporal (longue durée) sites. The coinage of the conceptual phrase “Distributed Observing Network of the Past” (DONOP) by some practitioners encompasses this perspective: “through the analysis of archaeological datasets, we have the potential to access long-term records of human interactions with natural systems at a wide variety of temporal and spatial scales and thus both reconstruct past environmental conditions and reveal the human dimensions of these processes” (Hambrecht et al., 2020).

Incorporating different spatial scales, from local to interregional, enhances the study of historical human–non-human interrelationships (Hambrecht et al., 2020). However, it is important to note that inclusion does not mean decontextualisation. Instead, a comparative and inclusive perspective that pays attention to each context is necessary to deploy coherent analyses. Multiscalar thus means to integrate different cartographies within the same narrative. The ultimate goal of such an approach is to provide “historical and diachronic knowledge of human–environmental relationships” and their impacts, which can inform present concerns (Ray, 2019).

3.1.1 Sources and Data Acquisition

Our study required sources and data, including legacy data, of different natures to cross-check and obtain further archaeo-historical and environmental information about this northeastern African region. Therefore, integrating published field data surveys, archaeological reports, maps, historical aerial photographs, and satellite imagery has proven valuable. These resources have been particularly useful in identifying scattered archaeological remains and traditional paths, which are typically considered proxies for inferring the human mobility of communities across landscapes. Moreover, environmental and geographical indicators can be reconstructed from these resources, thus allowing for a more holistic approach in line with our theoretical framework.

The identification of these archaeological remains in the specific case of Marmarica draws upon a variety of sources, most notably the aerial photographs taken by Squadron 113 of the Royal Air Force (RAF). The RAF flew over large stretches of the Eastern Marmarica coast using Hawker Hind Lighter Bomber biplanes from September 23 to November 1, 1938, to map through photographs this strategic region in the pre-war context of the Second World War (Ray & Nikolaus, 2022). These photographs provide a unique snapshot of the landscape and environment. Another crucial source of information is the cartography used, primarily based on a series of maps produced by the US Army Map Service that covers the entire study area and dates from the Second World War to the Cold War (Figure 2). Specifically, the P502 series (1954) of the US Army Map Service, 1:250,000 stands out. This latter series derived from the information provided by the Middle East Command maps produced in 1941–1942, which were based on the information collected in 1916 by the “Survey of Egypt” for the territories covering northeastern Africa (Vetter, Rieger, & Nicolay, 2014, p. 44).

Figure 2

Georeferenced historical topographic maps of Crete (Greece) and Marmarica (northeastern Libya/northwestern Egypt) (Basemap: © ESRI).

The information gleaned from these sources is of great relevance for two main reasons. First, it provides insights into both human and non-human agents, encompassing topographic and hydrographic features, as well as the exact locations of ancient settlements, burial sites, productive areas, traditional routes, and so on. Second, the collected information predates the large-scale agricultural, urban, and industrial developments that took place in this region from the second half of the twentieth century. As such, these data provide a valuable record of landscapes that no longer exist or that have undergone significant changes over time.

Two comprehensive datasets were developed to collect and process the archaeo-historical and environmental information within a GIS framework (ArcGIS 10.5). This was done in order to systematically record the archaeological sites and traditional paths that were identified. The first dataset integrates the information associated with each archaeological site, considering the variables listed in Table 1. The second dataset contains information related to historical routes, considering the variables listed in Table 2.

The subsequent step involved mapping the terrain surface using remote aerial survey techniques. GIS was used to georeference a variety of historical sources, including topographic maps, historical aerial imagery (as detailed above), field survey data, and published archaeological reports. Control points located on the surface were systematically used to select easily identifiable topographic features, which allowed precise linkages in terms of geospatial position. The study area was then divided into 10 × 10 km², within which a thorough remote aerial survey was conducted by crosschecking our sources and imagery. Thus, GIS was instrumental in allowing us to obtain precise information on the locations of archaeological sites and historical paths (Figures 3 and 4).

Figure 3

(Left) Topographic map from 1954, US Army Map Service (P502, Sheet NH 35-3 Matruh). (Right) Spatial distribution of digitised sites covering the Marsa Matruh area (Basemap: © ESRI).

Figure 4

Digitisation of historical routes and traditional paths in Marmarica. This process involved various historical sources and aerial imagery (Basemap © Microsoft).

This process also included the identification of topographic and hydrographic elements, for instance, integrating various digital elevation models (DEMs) and generating the hydrographic network (rivers, wadis, potentially floodable areas, etc.) from them. This highlights the importance of GIS in facilitating a comprehensive evaluation and systematisation of data related to not only archaeological sites and historic routes but also geographic and ecological factors (Figure 5). It additionally enables the analysis of multivariate data, thereby contributing to the reconstruction of how these communities (humans) moved across the landscape, while considering other (non-human) agents that influenced and were involved in these dynamics.

Figure 5

The project delineated the diverse ecological contexts of the Marmarica region, encompassing a range of landscapes defined by the coastal zone, Northern tableland, the Pre-Marmarican Plain, Marmarica Plateau, and desert margins (based on fieldwork conducted on the Eastern Marmarica Plateau in northwestern Egypt (Rieger, 2023)). These contrasting ecosystems contribute to the unique biodiversity and environmental significance of the region as a whole.

3.1.2 (Re)constructing Past Movements

Besides GIS, the specific application of spatial analysis and network research are also fundamental in achieving our aims. Accumulated cost surface analyses have proven to be highly useful, and among them, the use of FMN has yielded highly satisfactory results in several case studies for different contexts (Aceituno & Uriarte, 2019; Déderix, 2016; Fábrega-Álvarez, 2006, 2016; Llobera et al., 2011; Moreno-Navarro, 2022; Pažout, 2017; Stančo & Pažout, 2020; among others). Other methods for reconstructing and studying past movements, such as least cost path, have the limitation of starting from known sites as origins and destination points from which to establish routes (Fábrega-Álvarez & Parcero-Oubiña, 2007; Güimil-Fariña & Parcero-Oubiña, 2015; Maniére, Crépy, & Redon, 2020; Verhagen & Jeneson, 2012; Verhagen, Nuninger, & Groenhuijzen, 2019; among others). Therefore, given that the nature of archaeological data makes our sample always biased and unequal, we tend to overrepresent specific zones within the study area. By contrast, FMN can overcome the limitation of not having high reliability regarding all the archaeological sites in a given region or even not knowing them due to various issues such as lack of reliable dating. As discussed below, this method is based on the application of hydrological and cost–distance analysis tools, resulting in pathways or routes with potential flow accumulation, i.e. movement potential. Additionally, this method allows us to obtain accumulation values that are key to applying network approaches to our data and research aims.

Spatial analysis first requires identifying the factors influencing mobility and the temporal changes in landscapes. Our case encompasses a vast region located in northeastern Africa, predominantly characterised by flat terrain and arid and semi-arid conditions. Consequently, the factors affecting mobility for our study were (i) the physical and topographic characteristics of the terrain surface, such as elevation, slope, and the presence of valleys, plains, and plateaus; and (ii) the hydrographic network, which includes elements such as rivers, water catchment areas, and the presence of wadis (i.e. dry streams or channels with seasonal flood potential typical of arid and semi-arid regions).

After clearly defining the factors affecting mobility for our case, the next step was to establish precise mechanisms to quantify each one. The goal was to create a friction or cost surface map that represents the difficulty of moving across landscapes. In the case of Marmarica, due to its semi-arid and arid conditions, the presence of water catchment areas, wadis, and slopes are particularly relevant to consider (Table 3). Thus, each factor was quantified depending on the spatial context, for, as the principle of heterarchy states, not every agent is equally relevant (in this case, for human mobility); the significance varies according to the spatiality.

A comprehensive friction or cost surface map should not only represent the physical characteristics of the terrain but also embed a mobility function to simulate movement across different landscapes. This is critical for the model to reflect the complexities of historical mobility accurately. To this end, Naismith’s rule (1982) was applied as a foundational component of the mobility function. This function provides duration estimation by positing that a travel cost (i.e. time and effort) is expected to double for every 12% increase in slope. This rule was instrumental in the calculation of our final friction map, providing a nuanced view of how topography impacts mobility (Figure 6). Our choice is not arbitrary; it is grounded on the recognition that human movement is a key factor influencing mobility along transport routes. Indeed, it is highly likely that the use of packs or draught animals through these routes was accompanied by people on foot (Verhagen et al., 2019), which emphasises the importance of considering human mobility as a key in our analysis. The implications of this are twofold: not only does it acknowledge the role of humans as active agents shaping the landscape through mobility but it also asserts the critical nature of human mobility patterns in our spatial analysis.

Figure 6

Friction map generated by assigning a gradual cost to the hydrographic network depending on the ecological context and calculating the slope from a DEM. Values are represented using a colour categorisation scheme, with lighter shades indicating greater resistance to movement and darker shades indicating lower resistance. Dark-coloured circles indicate water catchment areas.

The successful implementation of FMN requires a series of sequential operations following the acquisition of the friction map or surface cost for the region. Initially, it is necessary to generate a mesh or grid of points, which can be achieved by creating a set of random points or an orthogonal layer of uniformly spaced points covering the entire study area (Figure 7). The generated mesh was then associated with a friction map, which was obtained by assigning cost values to each factor (Table 3). The following procedure entails executing a sequence of operations using GIS. The first step involved calculating the distance from each point in the mesh to each pixel with respect to the friction layer generated in the previous step. This is followed by the calculation of the flow direction using hydrological tools based on the cost distance layer calculated in the previous step. Finally, the flow accumulation at each point is computed based on the flow direction. All of these operations can be automated, leading to the generation of a raster layer for each point, in which the pixel value represents the accumulated flow towards that point. To generate the final network using FMN, the mean value of all layers was calculated, resulting in an integrated layer with the values of the accumulated flow, indicating potential circulation (Figure 8).

Figure 7

Orthogonal point layer covering the entire study area, with points spaced 10 km apart.

Figure 8

Preliminary results of the application of the FMN. Potential routes are depicted in graduated shades of black based on the accumulation of values obtained.

3.1.3 Exploring Network Dynamics

The aforementioned procedures provide an opportunity to identify pathways or routes with circulation potential. However, our research focused not only on the archaeological study of routes but also on representing and analysing interaction patterns and connectivity within and between the target regions. To achieve this, we employed network approaches to represent our archaeological data as network data, so that it can be analysed and explored. The reconstruction of possible pathways enhances our understanding of the topological structure of the routes within the region under study, while also obtaining accumulation values that allow measuring the strength of connections between different places based on their potential circulation. This is particularly significant in our study, which relies on spatial networks (Brughmans & Peeples, 2020), as connectivity is determined not by the number of connections between specific places but by the intensity of those connections.

Network science can inform us from additional angles of different patterns and explore the structure of our region’s relationships between places. However, prior to applying network science methods to our data, a sequence of preparatory steps must be carried out. First, it is necessary to review and correct topological errors that may have arisen during the network generated using FMN. These errors include fragmented segments and discontinuous intersections (Figure 8). This process is essential for obtaining a network that generates the least possible distortions when analysed. Second, our research necessitated the development of junction nodes that indicate the points at which each route segment or section ends (Figure 9a). This is because our goal was to study the network structure independently, treating each segment or section as an entity to be analysed.

Figure 9

(a) Topological network (in blue) of the Marmarica region obtained by processing results from the FMN approach and generating nodes at each intersection. The endpoint of each route segment is represented as a node (in red). (b) Table showing the different metrics calculated using statistical tools and integrated into network attributes.

The subsequent step involved integrating the corresponding values for each network segment. These values include the area, number of cells, sum, mean, standard deviation, range, and length for each segment (Figure 9b). The statistical analysis tools provided by GIS have allowed their calculation, which in turn allows us to integrate them as attributes associated with our network. In this way, we can correlate both, nodes and metrics, for each segment. The outcome is a layer representing our network for the study region and, most importantly, a network containing all the necessary information to carry out network analysis.

At this stage, we can delve into the exploration of our data using different techniques, simulations, and analyses through network visualisation software (Visone v. 2.24.1) to explore our networks from different configurations and structures (for various methods, see Brughmans et al., 2023; Brughmans & Peeples, 2023). For example, we can explore the role of nodes based on their positions and relationships with others within the network structure (Figure 10).

Figure 10

Network representation of the Marmarica region. Nodes (in blue) have been freed from their geographic position, allowing a visualisation based on the strength of their connections and the number of links within the network.

In our approach, the application of centrality measures has provided key insights into the structure and function of a network (Freeman, 1979). Centrality measures, such as “betweenness centrality,” were used to evaluate the influence of a node within the network structure (Figure 11a). This metric allows us to assess the relevance of a node in terms of its position based on the number of times this specific node acts as an intermediary between the nodes. Nodes with high betweenness centrality are considered key points because they are essential for the flow and connection between different areas of the network. Another important centrality measure is “closeness centrality,” which is used to quantify the proximity between nodes (Figure 11b). Nodes with high closeness centrality indicate how close they are to each other and are important for fast and effective communication across the network (Brughmans et al., 2023). Analysing various network metrics was therefore crucial, as they enhanced our understanding of the network dynamics.

Figure 11

Representation of the mobility network in the Marmarica region with nodes in their geographic position. (a) Nodes are gradually represented according to the betweenness centrality. Higher values are clearly observed along the coast and the central connection towards the inland oasis. (b) Nodes are gradually represented according to the closeness centrality. Higher values are again clearly observed along the coastal area.