Abstracts

Soil Ecosystems

Diversity of soils, as other natural objects, occurs not only at type or higher taxonomy levels. Large-scale and detail investigations of soils revealed that within uniform spatial units properties of soils are characterized by essential variability. Variability needs to consider soil properties as random values.

Probability distribution is full and universal characteristics of random value as any other probability characteristics may be obtained from it. Probability distribution is necessary for accurate and correct evaluation of soil property and its changes under nature and anthropogenic processes, for technique and land evaluation calculations, for mathematical modeling of soil processes using probabilistic approach (for example Monte-Karlo). Besides mathematical functions of probability distribution are needed for keeping information about soil variability in databases.

Information about types of statistical distributions of soil properties in science literature is limited and contradictory, but it refutes idea about universal “normality” of distributions. To make a priori theoretical suggestions about type of probability distribution is rather difficult, because the soil is open complex system and has hierarchical space structure, so to predict how the sum of different process results in this or that probability distribution is difficult. So investigation of probability distributions of soil properties were conducted on example of Kastanozems in southwest part of Kulunda steppe. These soils were formed from coarse-textured Quaternary alluvium. Input data were obtained after large-scale (1:25000) soil survey on big territory (16000km2). Data were grouped according their taxonomy type and texture. We determined probability distributions of properties in homogeneous groups of soil having different texture: sand, loamy sand, sandy clay loam, sandy clay. Numbers of observations in each group were large (n=50-600) that provide opportunity to carry out this statistical investigation.

Determinations of probability distribution functions were conducted by using method of “modeleteka”, that means choice of functions from large set of functions, which were organized on principle of completeness and minimum of abundance. Evaluations of parameters of distributions were determined by method of maximal probability using empirical data. Hypothesis of goodness were checked on six criteria’s (2, likelihood, Kolmogorov’s, Smirnov’s, и 2 Mizes’s criteria 2) for each distribution. On number of received values of statistics of these criteria’s the only distribution was selected, that bring maximal probability of similarity of empirical and theoretical distributions. We used program of statistical analysis developed at faculty of Applied Mathematics and Computer Sciences of Novosibirsk State Technical University.

Statistical analyses show that soil properties often have probability distributions from Johnson’s and exponential families, distributions of extreme values and normal and lognormal.

Note. Abstracts are published in author's edition

|Home Page| |English Part|

Go to Home