Pedogeochemical mapping and background limit of trace elements in soils of Branicevo Province (Serbia)

摘要

Main soil properties, contents, distribution and background limits of hot acid extractable forms of trace elements: Cd, Cr, Cu, Ni, Hg, Pb, Zn have been analysed in Branicevo Province in Eastern Serbia. Contents of selected trace elements are relatively low and exploratory data analyses (EDA graphs) show that elements have asymmetrical distribution and great dispersion. This primarily applies to Cu, Cd, Hg, which at same time, have similar distribution patterns. Branicevo Province comprises 8 smaller districts, and is characterized by highly heterogenic natural conditions, which additionally complicate assessment of background limits. Different methods (classical [Mean+2Sdev], [Median+2MAD], box-plot and CDF) have been used for the evaluation of background limits, and for some methods besides natural, logtransformed values have also been used (Sdev - standard deviation, MAD - median of the absolute deviations from the data's median, CDF -cumulative probability plot). Values from [Median+2MAD] procedure have the lowest background limit, followed by the box-plot, while the classical method produces the highest background limits. Results for CDF, (where the background limits for all elements are approximately at cumulative 95%) are similar to those from logtransformed data of box-plot method (except for Cd and Ni). Thus, percentage of outliers in classical methods ranges from 1.1 to 4.2% for natural data (and 0.5-3.75 for logtransformed data), for box-plots method ranges from 0.8 to 11.0% (and 0.6-9.6%), and for [Median+ 2MAD] ranges from 3.4 to 22.0% (and from 1.4 to 15.8%). The smallest outliers percentage is for Cd, followed by Hg, than Cr and Ni, and it is higher for other elements. Outliers of Ni and Cr are most frequent in western disticts in Fluvisols of the Velika Morava River, Hg and Cd outliers are the commonest in the eastern district, while outliers of other elements are more evenly distributed. Since there is a statistically significant difference in distribution of trace elements per district, data have been arranged by homogeneous units, and then new, more realistic background limits have been determined.