On the use of non-Gaussian models for prediction of extreme pollution levels in environmental studies

Abstract

The aim of this work is to study the distribution of contamination at the territory on the data of snow samples analysis, in order to find an approach to forecasting of the extreme pollution levels. The hypothesis of normal distribution of the values of pollution index (the intensity of dust fallout on the territory, mg /m2/day) is not confirmed on the results of statistical analysis of the data for six different experimental sites (from 81 to 256 values of the index for each site). For the set of 243 values of the pollution index at the territory of a city, there is made an attempt of forecast of its possible extreme values not detected on the results of the snow sampling. For this, the linear dependence “pollution index - the number of points with the given pollution index” built in double logarithmic coordinates, is extrapolated into the area of high values of the pollution index.