New method for ecological monitoring based on the method of self-organising mathematical models

In many situations it is necessary to generate a multidimensional mathematical modelling of the parameters or observations defined on an irregular grid of observation data. We have developed original algorithms that include the methods of self-organisation for this purpose. Unlike regression analysis, the method of self-organisation is based on the purposeful search for optimum model complexity. The optimum model is found by the well-directed exhaustive search within a set of the model-pretenders. The methods that we have developed, were used for analysing the consequences of the Chernobyl disaster. The three- and four-dimensional local polynomial models have been developed. This allowed us to calculate radionuclide distribution maps and carry on a number of prognostic calculations. The field of 137Cs distribution is characterised by a high determination level (D≈89.63%). This fact shows a “data consistency” and a low level of a randomness. The map of the 137Cs prognostic errors of each point allows conclusions to be made regarding whether the point is anomalous or informative. When considering the map of 90Sr distribution and the map of prognostic residuals, one can conclude that randomness in the 90Sr field is large (D≈59.88%). We have calculated a more correct map of the 90Sr distribution (D=71.66%) using the four-dimensional modelling where the 137Cs isotope distribution was introduced as a fourth variable.