Soil spatial patterns analysis at the ancient city of Ibida (Dobrogea, SE Romania), via portable X-ray fluorescence spectrometry and multivariate statistical methods

Abstract

In this study we investigated the potential usefulness of multi-elemental soil analysis, obtained by a portable XRF spectrometer, to recognize ancient anthropogenic features, compared to the natural context on the territory of Roman-Byzantine Ibida city (Dobrogea, SE Romania). Soil samples were collected from 16 soil profiles and 2 archaeological excavations, from each soil horizon and cultural layer. Pearson correlation matrix of topsoil elements showed significant correlations among the various geogenic elements such as Ti, Fe, Mn, Zr, Ba and Cr. In the case of P, a strong correlation was found with Ca, a lower correlation with K, and no significant correlation with the other elements. Also, Cu and Zn behave differently, showing no significant correlation with any of the elements. Principal components analysis (PCA) and agglomerative hierarchical clustering (AHC) were carried out in order to identify groups of inter-correlated variables and assess the degree of association among soil profiles. The correlation between soil elements and the principal components suggest the presence of three groups which were defined as the geogenic, archaeo-anthropogenic and agro-anthropogenic component. Spatial distribution of topsoil elements was mapped by geostatistical methods in order to see if their spatial patterns reveal the influence of human activities. Although the results show that certain elements such as Ca, P and K present higher concentrations in the Ibida fortress area, we found that the interpretation of multi-elemental composition of the topsoil only is not sufficient for an accurate identification of the ancient characteristics which can be linked to the Roman-Byzantine period. Hence, we wanted to know if the presence of archaeological sites could be revealed by higher variation degree of soil elemental data along the profile depth. Therefore, we tested the utility of a synthetic statistical index, namely the variation coefficient, which was computed for each soil profile using all the available data. The results showed that if the index was standardized according to profile depth, then it became a useful indicator for the presence of ancient human activities.