Evaluation of geochemical background levels around sulfide mines – A new statistical procedure with beanplots

Geochemical background levels and pre-mining conditions can be defined and quantified in several ways with different outcomes. Numerous authors have attacked the background assessment problem by generating various statistical procedures for estimating the level or threshold for concentrations assumed to represent the background. None of the methods have been generally accepted as a de facto standard. Some methods assume a normal distribution, leading to trimming or normalization procedures and some are based on more generally applicable non-parametric methods. The uncertainties always present in geochemical data are rarely reported with background levels. In this study the geochemical background is considered as the range of levels anchored at the time of sampling under pre-mining conditions. A statistical description of a distribution assumed to represent such a background is suggested including the non-parametric median, confidence interval of the median and a modified beanplot, which illustrate the variation and uncertainty of the data. The procedure is demonstrated on data from the vicinity of the Talvivaara black shale -hosted Ni–Cu–Co–Zn–Mn mine in eastern Finland, and the differences in levels between the mineralized and non-mineralized subareas in stream sediments and stream water are presented. The sampled materials are well-decomposed organic stream sediments (259 samples), poorly decomposed organic stream sediments (285 samples), minerogenic stream sediments (102 samples), and stream water (149 samples). The median values for Cu, Fe, Ni, Pb and Zn are significantly higher in mineralized subareas than in non-mineralized subareas in organic stream sediments and minerogenic stream sediments. In stream waters Cu, Ni and Zn behave similarly showing a significant difference. The results show that the average geochemical background in stream sediments and stream water can vary significantly across lithologies, and that medians with confidence intervals together with diagrams depicting the distributions in detail, such as beanplots, are useful tools for assessing and illustrating the background levels.