Statistical analyses for heavy metal contents in till and root samples in an area of southeastern Sweden

Abstract

The Geological Survey of Sweden (SGU) initiated a national mapping program in 1982 with three types of geochemical samples (bedrock, till and biogeochemical) and the objective to produce a detailed geochemical atlas of the entire country. This program is still ongoing. Problems concerning the calculation of mean values, outlier detection, relationships among elements, spatial distribution features and relationships among the different types of geochemical samples are therefore important to consider. To find an optimal way to solve these problems, the contents of seven heavy metal elements (Co, Cr, Cu, Ni, Pb, V and Zn) in 758 till samples and 851 root (with some moss) samples were extracted from the SGU database in a polygon of 75×75 km 2 in southeastern Sweden. Both univariate and multivariate analyses were carried out on the datasets. The combination of range method and principal component analysis (PCA) was satisfactorily utilized for the detection of outlying samples. The commonly-used mean calculation methods of median, geometrical mean and arithmetical mean are discussed and compared, and a new method for mean calculation named `robust-symmetric mean' (R-S mean) involving robust statistics and optimal data transformation is recommended. Results from multivariate analysis show that Pb has comparatively weak correlations with the other elements and Zn also possesses such a trend, which is caused by the differences of their contents in bedrocks and mineralization. Lead is enriched in the acid volcanic rocks, while the other metals are elevated in the basic rocks of the area. Lead mineralization in the acid volcanic rocks also causes elevated contents of Zn in the bedrock. Heavy metals in biogeochemical samples are more diversified than in tills, as they are more affected by external geochemical processes than in tills, also the relationships among elements in the biogeochemical samples are therefore more affected by these processes.