Missing Sites, Shifting Centers? Evaluating Network Robustness in Early Mesopotamian Landscapes

This paper presents a proximity network-based analysis of settlement systems in southern Mesopotamia during the Early Dynastic (ca. 2900–2350 BCE) and Ur III (ca. 2112–2004 BCE) periods, conducted within the framework of the HyMes project (https://hymes.ch). The study evaluates the centrality of sites and assess the robustness of observed network structures despite incomplete archaeological data.
Using spatial datasets of known sites, networks are constructed to model inter-site connectivity based on geographic proximity. These networks are analyzed using standard centrality measures such as degree centrality, understood here as a proxy for interaction intensity within the settlement system. However, archaeological datasets are inherently fragmentary, and smaller sites are especially prone to underrepresentation due to survey bias and preservation conditions, raising concerns about the stability of inferred network hierarchies.
To address this, the study employs a sensitivity analysis that simulates data loss through the removal of small sites. Site selection is implemented within a Monte Carlo framework using weighted random sampling without replacement, where weights are derived from site size as a proxy for archaeological visibility, assigning higher probabilities of removal to smaller sites. This process is repeated across multiple iterations to generate a range of perturbed network configurations.
By comparing baseline and simulated networks, the analysis evaluates the stability of centrality measures and identifies which sites retain central positions despite the systematic loss of minor settlements. This approach enables a comparative assessment of robustness between the Early Dynastic and Ur III periods, contributing to methodological discussions on uncertainty in archaeological network analysis and offering new insights into the organization and resilience of early Mesopotamian settlement systems.