Modelling Mobility of Hunter-Gatherer Populations: A Dynamic Simulation Approach Based on Cellular Automata.

Abstract

Understanding mobility of past hunter-gatherer populations requires dynamic approaches which incorporate uncertainty. Least cost models assume complete knowledge of the terrain on the part of the traveller, while ethnographic examples tend to be specific to the groups and territories studied. Most least cost models also assume that origin points, destination points, or both, are known in advance, limiting their utility for exploring movement potential in landscapes where evidence for occupation is scarce. This research addresses these limitations through an agent-based model of movement grounded in cellular automata (CA) theory, called DISPERSCA. Agents depart from a point, which may be specified or determined at random, and transit a fitness landscape for a fixed number of iterations according to decisions made within a defined area at each time step (a decision catchment), the CA neighbourhood. If the decision catchment is unknown multiple runs are made at different CA neighbourhood sizes and the results are compared. Neighbourhoods may be square or hexagonal, the former producing on average longer displacements, the latter ensuring that individual walks are of equal length in any direction. The model is demonstrated by application to Late Pleistocene Central Iberia, where confirmed archaeological sites are scarce. Some support can be advanced for the hypothesis that the Central Iberian mountains, probably combined with the Iberian System range, presented a significant barrier to hunter-gatherer groups. The model can be modified to account for agents' prior knowledge, or to include fitness variables unrelated to terrain cost, such as water, the presence of game animals or vegetation.