MODELING THE Rn-222 AND Rn-220 PROGENY CONCENTRATIONS IN ATMOSPHERE USING MULTIPLE LINEAR REGRESSION WITH METEOROLOGICAL VARIABLES AS PREDICTORS

The present paper deals with the study of the multiple linear regression model for the estimation and prediction of the time series of radon and thoron progeny concentrations in atmosphere. The general purpose of multiple linear regression model is to find the linear relationship between a dependent (or explained) variable and several independent (or predictor) variables. Radon and thoron progeny data measured in Bacau station of the National Environmental Radioactivity Survey Network, coordinated by National Environmental Protection Agency, are modeled, at different time scales, by making use of the multiple linear regression with meteorological parameters as current and lagged independent variables. The daily average values of radon and thoron progeny concentrations for different months, as dependent variables, have been modeled as function of the following meteorological variables: air temperature, soil temperature, atmospheric pressure, relative humidity, precipitation and wind speed, as predictor variables. The collinearity and multicollinearity of independent variables have been analyzed. The estimation performances of the model have been checked by using the regression statistics: multiple correlation coefficient, coefficient of determination, F-test values, level of significance (p-value) and sum squared residuals. The prediction performances have been quantified by means of the sum squared residuals, correlation coefficient and the relative error specific to each time interval from prediction period. Also, there has been studied the prediction performances, considering different time periods on which the estimation regression has been built.