This poster presents a predictive modelling study of the Neolithic settlement system in the Boucle du Vaudreuil, a micro region of 1860 km² in the Seine Valley (Eure, France). The analysis focuses on the environmental context of the sites, particular topography, geology and soil capacities for agricultural production. We analyzed the environmental context of Neolithic excavated and surveyed sites in the area through principal component analysis (PCA) and cluster analysis. In this way, we established environmental ‘signatures’ of settlement locations that could be used as input for predictive modelling. The MaxEnt software was used to extrapolate and visualize the results. This program was developed by ecologists for modelling species distributions for presence-only species records. It estimates the relationship between species records and the environmental and/or spatial characteristics of the sites (Elith et al. 2011). Since the program is designed to work with insufficient samples it is very suitable for archaeological predictive modelling as well. MaxEnt provides probability map estimates by establishing the covariance between sites and environmental parameters. The poster presents a test-case study. We will also discuss the limits of this predictive modelling method, in particular the problem of model validation with non-random samples and the relevance of the parameters used for the analysis.

Reference:

Elith, J., Phillips, S.J., Hastie, T., Dudík, M., Chee, Y.E., Yates, C.J., 2011. A statistical explanation of MaxEnt for ecologists. Diversity and Distributions 17, 43-57.

Submission for CAA2012