Modelling trace metal background to evaluate anthropogenic contamination in arable soils of south-western France

The trace metal (TM) content in arable soils has been monitored across a region of France characterised by a large proportion of calcareous soils. Within this particular geological context, the objectives were to first determine the natural levels of trace metals in the soils and secondly, to assess which sites were significantly contaminated. Because no universal contamination assessment method is currently available, four different methods were applied and compared in order to facilitate the best diagnosis of contamination. First, the TM geochemical background was determined by using basic descriptive statistics and linear regression models calculated with semi-conservative major elements as predictors. The natural concentrations of trace metals varied greatly due to the high soil heterogeneity encountered on the regional scale and were more-or-less accurately modelled according to the considered TM. Second, the basic descriptive statistics and the linear regression methods were then compared with the enrichment factor (EF) method and multivariate analysis (PCA), in order to evaluate whether the concentrations measured in soils were abnormally high or not. The advantages and disadvantages of each method were discussed and their results used to identify the most probable contamination cases, the influence of the soils characteristics, as well as the agricultural land cover. The basic descriptive method was good as a first and easy approach to describe the TM ambient concentrations, but may misinterpret the natural anomalies as contaminations. Based on geochemical associations, the linear regression method provided more realistic results even if the relationships between major and trace metals were not significant for the most mobile TM. The EF method was useful to identify high point source contaminations, but it was not suitable when considering a large dataset of low TM concentrations. Finally, the PCA method was a good preliminary tool for the description of the global TM concentrations in a studied area, but it could only give indication on the highest contaminated points. By comparing the results of the different methods in the studied region, we estimated that 24% of the arable soils were contaminated by at least one trace metal, mainly Cu in vineyards/orchards and Cd, Pb and/or Zn in grazing lands. In addition, the calcareous soils exhibited globally higher natural and anthropogenic TM concentrations than non-calcareous soils, probably because of t