Statistical Modeling of Average Daily Concentration of Pollutants in the Atmosphere over Moscow Megalopolis by the Multiple Regression Method

P. F. Demchenko, A. S. Ginzburg, G. G. Aleksandrov, A. I. Vereskov, G. I. Gorchakov, N. N. Zavalishin, P. V. Zakharova, E. A. Lezina, and N. I. Yudin

Presented are the results of the construction of statistical models of the time series of air pollutants (particulate matter with the size of less than 10 m (PM10), CO, and NO2) for the network of automatic stations of air pollution control over Moscow megalopolis. The multiple nonlinear regression of pollutant concentration to external factors (meteorological and other) and values of concentration on previous days are taken as the statistical model of the time series of the average daily concentration of a certain pollutant. The nonlinear nature of the models of the time series can be caused with the dependence of pollutant concentration on wind speed and with other factors. Nonlinear regression based on the relatively short learning samples was used for simulating the series of average daily concentra- tion of pollutants. The computations demonstrated that this gives much higher correlation between the computed and observed values of concentration and smaller standard deviation as compared with the model of inertial forecasting.

Joomla templates by a4joomla