Neural network analysis of interdependences of the top-soil parameters

To test a possibility of reduction of soil parameters number used for description of total amount of soil organic matter, a neural network analysis of regional soil databases was carried out. It was shown, that two to three soil parameters are sufficient for the prediction of amount of soil organic matter. Herewith, to implement this prediction, a neural network can consist of less then four neurons. The obtained results indicate that it is possible to represent explicated dependencies in terms of relatively simple mathematical formulae. In turn, this gives promise to expect, if not a simplicity of a possible mathematical model of soil forming, a simple form of a dependence of steady states of the model on soil parameters considered in this work.