Filling the Gaps—Computational Approaches to Incomplete Archaeological Networks

Journal of Archaeological Method and Theory https://doi.org/10.1007/s10816-024-09688-z (2025) 32:19 RESEARCH Filling the Gaps—Computational Approaches to Incomplete Archaeological Networks Deborah Priß1 · John Wainwright1 · Dan Lawrence2 · Laura Turnbull1 · Christina Prell3 · Christodoulos Karittevlis4 · Andreas A. Ioannides4 Accepted: 4 December 2024 © The Author(s) 2025 Abstract Networks are increasingly used to describe and analyse complex archaeological data in terms of nodes (archaeological sites or places) and edges (representing relationships or connections between each pair of nodes). Network analysis can then be applied to express local and global properties of the system, including structure (e.g. modularity) or connectivity. However, the usually high amount of missing data in archaeology and the uncertainty they cause make it difficult to obtain meaningful and robust results from the statistical methods utilised in the field of network analysis. Hence, we present in this paper manual and computational methods to (1) fill gaps in the settlement record and (2) reconstruct an ancient route system to retrieve a network that is as complete as possible. Our study focuses on the sites and routes, so-called hollow ways, in the Khabur Valley, Mesopotamia, during the Bronze and Iron Age as one of the most intensively surveyed areas worldwide. We were able to predict additional sites that were missing from the record as well as develop an innovative hybrid approach to complement the partly preserved hollow way system by integrating a manual and computational procedure. The set of methods we used can be adapted to significantly enhance the description of many other cases, and with appropriate extensions successfully tackle almost any archaeological region. Keywords Hollow ways · Mesopotamia · Archaeological networks · Algorithms · Computational archaeology * Deborah Priß 1 Department of Geography, Durham University, Durham, UK 2 Department of Archaeology, Durham University, Durham, UK 3 Department of Cultural Geography, Rijksuniversiteit Groningen, Groningen, Netherlands 4 AAI Scientific Cultural Services Ltd., Nicosia, Cyprus Vol.:(0123456789) 19 Page 2 of 33 D. Priß et al. Introduction Analysing and understanding the connectivity of human and non-human entities in socio-ecological systems can offer new perspectives on the functioning of human societies, present and past (Bell, 2020). Networks as a representation of connectivity and network analysis as a technique to quantify connectivity have become increasingly popular in archaeology to enhance our understanding of the archaeological record (Brughmans & Peeples, 2022; Brughmans et al., 2023). However, the potential for network analysis to be used in archaeology is often constrained by missing data in the archaeological record. Complete data sets can give unique insights into how and why people moved through the landscape, thereby revealing—or at least allowing interpretations of—the social processes hidden behind their physical remains. The overarching aim of this study is therefore to develop approaches to resolve the issue of missing archaeological data, i.e. fill the missing gaps, based on the physical network structure. The hollow ways in Northern Mesopotamia are one of the best-preserved route systems worldwide and formed more than 5000 years ago. The temporal scope of the study covers the periods from the Early Bronze Age to the Iron Age (c. 3000–600 BCE). Together with the settlements they connect, the hollow ways form a network that can be analysed with graph-theoretical tools or statistical models to improve our understanding of the past. In the Khabur Valley in Mesopotamia, a region with fairly homogenous elevation, a route network of so-called hollow ways represents the human movement of the Bronze and Iron Age (de Gruchy & Cunliffe, 2020; Ur, 2003, 2009; Wilkinson, 1993). Unlike the Chacoan, Inca or Roman roads, they not only provide information about trade and exchange but also about the daily and regular shortdistance movement of people (Ur, 2009; Wilkinson, 1993). The hollow ways, as mere depressions on the surface, are prone to attenuation and destruction through geomorphological processes and land-use changes. Although the Mesopotamian hollow ways are still one of the least fragmented ancient road networks globally, their visible remains are fragmented, posing limitations on the application of network-analytical methods. Previous attempts of route reconstruction in the Khabur Valley include both, network approaches (Menze & Ur, 2012; Palmisano & Altaweel, 2015) and optimal route models (de Gruchy, 2016). Menze and Ur (2012) converted the hollow way system, digitised from satellite imagery, into edges and use the sites as nodes. However, they do not provide a description about how they defined if a hollow way connects two sites or if and how they filled the gaps between the hollow ways. Palmisano and Altaweel (2015) developed a model to calculate traffic between sites in the Khabur Valley and weighted traffic that coincided with hollow ways higher. Although they were able to define factors that impact hollow way location, they did not attempt to fill the gaps between them. De Gruchy (2016) calculated optimal paths based on variables that are known to affect route choice and quantified their influence on route choice according to their overlap with hollow ways. This approach allowed the significance of distinct factors on Filling the Gaps Page 3 of 33 19 hollow way formation to be determined but did not involve reconstruction of the complete hollow way system. Reconstructing the complete network opens a new suite of analytical tools to allow for more advanced analysis of network properties and controls on network formation and is therefore not only a necessary first step in the analysis but a valuable addition to the archaeological toolbox. Road Reconstruction and Network Analysis in Archaeology Roads, tracks and paths between settlements are physical evidence of the repeated movement of people, resources and material culture, and are one type of network edges that can be represented as a graph. The ancient road infrastructure can give invaluable insights into the structure and functioning of the social, economic and political networks it is embedded in. Studies about ancient infrastructure include research on mediaeval Russian waterway networks (Pitts, 1965), the Roman roads in Britain (Dicks, 1972) and the Chacoan road network in the US Southwest (Ebert & Hitchcock, 1980). Later studies have focused on road networks where constructed features such as (partial) pavements, bridges and retaining walls are easily recognisable in the landscape, e.g. Roman roads in Europe (Dicks, 1972; Graham, 2006; Isaksen, 2008; Orengo & Livarda, 2016; Verhagen et al., 2019; Lewis, 2021) and the Inka roads in South America (Jenkins, 2001) as well as combining observed evidence of roads with (leas (...truncated)