Jörg Kraft

Chemometric characterization of soil and plant pollution: Part 1: Multivariate data analysis and geostatistical determination of relationship and spatial structure of inorganic contaminants in soil

Geostatistical and multivariate methods of data analysis are used to describe patterns of soil pollution with inorganic contaminants in Celje County, Slovenia. Groups of contaminants and polluted sites were identified using cluster analysis and confirmed with multidimensional variance and discriminant analysis. Factor analysis yields an identification of not directly observable relationships between the contaminants. The spatial structure and distribution of contaminants were assessed by applying semivariogram analysis and kriging interpolation method. Zinc, Cd and Cu were identified as a pollutant emitted from the zinc smelter, Pb also from other sources, and Cr and Ni mostly from geological parent material.

Small-Scale variability of metals in soil and composite sampling

Soil pollution data is also strongly scattering at small scale. Sampling of composite samples, therefore, is recommended for pollution assessment. Different statistical methods are available to provide information about the accuracy of the sampling process. Autocorrelation and variogram analysis can be applied to investigate spatial relationships. Analysis of variance is a useful method for homogeneity testing. The main source of the total measurement uncertainty is the uncertainty arising from sampling. The sample mass required for analysis can also be estimated using an analysis of variance. The number of increments to be taken for a composite sample can be estimated by means of simple statistical formulae. Analytical results of composite samples obtained from different fusion procedures of increments can be compared by means of multiple mean comparison. The applicability of statistical methods and their advantages are demonstrated for a case study investigating metals in soil at a very small spatial scale. The paper describes important statistical tools for the quantitative assessment of the sampling process. Detailed results clearly depend on the purpose of sampling, the spatial scale of the object under investigation and the specific case study, and have to be determined for each particular case.

Examination and presentation of element distribution in soil

Goal, Scope and BackgroundThe distribution of elements cannot only scatter widely in investigation areas, but also to a small scale in investigation fields. Chemometric methods are useful tools to describe the spatial distribution of the elements and are suitable to characterize the inhomogeneity in the soil. This knowledge can also be beneficial, among other things, for the creation of problem-adapted sampling strategies. The element distribution at one sampling point, the so-called microinhomogeneity considerably affects the representativeness of pollution assessment for a whole investigation area.Methods (or Main Features)The case under investigation was a small area of 1 m2 uncultivated pasture, covered by grass and not specifically polluted. The distribution of 13 elements in the topsoil has been investigated. The samples were taken at the 25 nodes of a regular grid from the upper layer (sampling depth: 10 cm) which covered the tested area. After drying and sieving, the soil was digested by using aqua regia. The elements were determined by different techniques of atomic absorption spectroscopy: flame (Cu, Fe, K, Mn, Na, Zn), graphite furnace (Cd, Cr, Cu, Ni, Pb) and FI-hydride (As, Se). The local gradient method, multivariate statistics and mapping of the element distribution are used for quantitative assessment of the inhomogeneity of the element distribution in course of investigation.Results and DiscussionThe contents of the elements are measured in a small area of 1 m2, and the mean and some important parameters are determined. The contents are highly variable and scatter between the minimum and the maximum, with the standard deviation ranging between 11% and 51%. The observed concentrations were used in formulation of ‘local’ polynomial models which approximated element distributions inside the squares of the grid. Together, the local distribution formed a distribution map for the element over the 1 m2 area which was tested. Also, some global (mean, averaging) characteristics of distribution inhomogeneity were used. These values of global characteristics show no gradient type distribution of the elements over the whole tested area under investigation. Additional information about the inhomogeneity of the investigated area can be obtained by multivariate statistical methods (cluster analysis and principal components analysis) and some selected methods of data presentation (2D and 3D sequential diagrams).ConclusionThe advantages and disadvantages of the approach are discussed. The mapping and visualization of the element distribution together with the global characteristics of inhomogeneity is a useful and comfortable way of presenting and collecting data of environmental monitoring. The mapping appears to be a most impressive and user-friendly presentation of element distribution. The local inhomogeneity is more ‘intensive’ if more isolines cross a subsquare.Recommendation and OutlookThe investigation will be continued considering another case study. This particular case study is accentuated by a strong dustlike immisssion and subsequently characterized by a gradient of soil pollution.