Random forest models highlight early Homo sapiens habitats and their relationship to lithic assemblage composition

Eastern Africa plays an enduring role in understanding the evolution of modern humans in relation to palaeoenvironmental change. Numerous studies have implicated ecological and geographic factors in determining the spatiotemporal patterning of technological and cultural behaviour seen in the region during the Middle Stone Age (MSA), the technological phase associated with the emergence of Homo sapiens in Africa. We use an eco-cultural niche modelling approach to evaluate the importance of different environmental and geographical variables in determining where early humans could have lived, and the impact of habitat suitability on different aspects of the eastern African MSA lithic record. We apply random forests, a powerful and highly flexible machine-learning tool for niche modelling, in combination with palaeoclimatic simulations at high temporal resolution. Topographic factors, distance to water and net primary productivity are found to be the most important factors in predicting MSA site locations in eastern Africa, followed by a suite of precipitation and then temperature variables. We find that environmental suitability has a significant impact on overall assemblage composition, suggesting that populations occupying optimal and peripheral zones within the landscapes use distinct artefact types and reduction technologies. These include the use of burins, bifacially retouched tools and bipolar technology, which are more likely to occur in the most optimal areas. Core tools, Levallois point and blade technology, and denticulates are associated with more peripheral areas, and thus perhaps could be associated with adapations to riskier ecosystems.