Estimation of spatio-temporal correlations of prehistoric population and vegetation in North America

We discuss a simple methodology to enable a statistical comparison of human population with the vegetation of North America over the past 13000 years. Nonparametric kernel methods are applied for temporal and spatial smoothing of point data obtained from the Neotoma Paleoecology Database and the Canadian Archaeological Radiocarbon Database, which results in sequences of maps showing the development of population and different plant taxa during the Holocene. The estimation of smooth spatial and spatio-temporal cross-correlation functions is proposed in order to detect relationships between population and vegetation in fixed time intervals. Furthermore, the effects of varying environment on demographic changes as well as potential impacts of populations on plant taxa over time are analyzed. Pointwise confidence bands for cross-correlation functions are computed using nonparametric resampling methods to assess the significance of obtained results. Considering the example of oak, an interpretation of our results for eastern North America shows the value of this methodology.