Geostatistical Estimation Of Soil Aggregate Structure As A Composite Variable
Abstract
A. V. Zhukov, K. V. Andrusevich, K. V. Lapko, V. O. Sirotina
In article approaches for statistical estimation of composite variables are considered. The soil aggregate structure is described by indicators which concern a category composite variable, i.e. such which in the sum always compound the fixed number (in our case it is 100 %). Mathematical properties of composite variables is essential confine possibility of various types of mathematical actions, including statistical analysis, over the data on soil aggregate structure. For application of statistical and other mathematical methods of analysis of the data of aggregate structure this data should be preliminary transformed. The classical soil structure coefficient is closest on ideology to the transformed variables, but its mathematical form not to the full meets the requirements of the further statistical procedures as is somewhat arbitrary. In the literature there are various variants of bases of orthogonal log-transformation of the data, but there are no ecologically well-founded criteria for their choice. For a choice of the best basis of transformation we offer a method of comparison of transformation results with edaphic properties matrixes or matrixes of plants morphometry. The optimum decision represents such basis which gives the best correlation with matrix external in relation to a composite variable of properties. Ordinary and partial Mantel tests have allowed to establish that the variation of aggregation structure is at the bottom of variability morphometric indicators of corn from the sowings which are on given bedrock. In turn correlation of aggregation structure with other edaphic properties is a consequence of their co-ordinated variability owing to unity of soil as is natural-historical body.
<