Analysis of meteorological data of the reanalysis model NCEP GFS for the atmosphere of Krasnoyarsk city

Abstract

Air pollution is an important problem for cities. Krasnoyarsk city is one of the dirtiest cities in the Russian Federation. Meteorological conditions have a significant impact on air pollution. In the present study, for constructing a regression model of forecasting periods of high levels of air pollution, the dimension of meteorological data of the global atmospheric model National Centers for Environmental Prediction Global Forecast System (NCEP GFS) was reduced. The meteorological data were collected between June 2019 and March 2022. To reduce the dimension of meteorological data were used correlation analysis and principal component analysis (PCA). These methods also resolved the problem of collinearity between independent variables. The same meteorological parameters of different vertical layers were reduced from 157 to 58 using correlation analysis. The principal component analysis made it possible to reduce the data dimension to 18 principal components that contain 90% of the total variance. The first 5 principal components contain 71% of the total variance. The principal components will be used to construct the principal components regression.