Statistická analýza dat z geologického průzkumu

Abstract

This study investigates the dynamic relationship between atmospheric carbon dioxide (CO2) concentrations and a set of meteorological variables using time series analysis. Hourly CO2 measurements and meteorological data (temperature, pressure, humidity, wind speed and direction, cloud cover, and precipitation) collected over a three-year period were aggregated to daily values and analyzed using transfer function modeling techniques. Two distinct methodological approaches were applied: a classical Box–Jenkins transfer function model with prewhitening, and a multi-input lagged regression model without prewhitening. Preprocessing techniques including Box–Cox transformation, harmonic regression, andseasonal smoothing were implemented to ensure stationarity and improve model performance. In the prewhitened approach, significant lags were first identified via cross-correlation analysis, and an AR(2) structure was incorporated into the regression residuals to address temporal autocorrelation. In contrast, the non-prewhitened model directly incorporated all lagged predictors within an ARIMA(0,0,1) framework. Comparative diagnostics revealed that the prewhitened model exhibited superior residual behavior, with reduced autocorrelation and improved interpretability of lagged effects. The findings emphasize the methodological importance of prewhitening in transfer function modeling, particularly for environmental time series characterized by complex internal structure. The results offer practical insight into how meteorological dynamics influence short-term variations in CO2 levels and provide a robust foundation for future modeling and forecasting efforts.

This study investigates the dynamic relationship between atmospheric carbon dioxide (CO2) concentrations and a set of meteorological variables using time series analysis. Hourly CO2 measurements and meteorological data (temperature, pressure, humidity, wind speed and direction, cloud cover, and precipitation) collected over a three-year period were aggregated to daily values and analyzed using transfer function modeling techniques. Two distinct methodological approaches were applied: a classical Box–Jenkins transfer function model with prewhitening, and a multi-input lagged regression model without prewhitening. Preprocessing techniques including Box–Cox transformation, harmonic regression, andseasonal smoothing were implemented to ensure stationarity and improve model performance. In the prewhitened approach, significant lags were first identified via cross-correlation analysis, and an AR(2) structure was incorporated into the regression residuals to address temporal autocorrelation. In contrast, the non-prewhitened model directly incorporated all lagged predictors within an ARIMA(0,0,1) framework. Comparative diagnostics revealed that the prewhitened model exhibited superior residual behavior, with reduced autocorrelation and improved interpretability of lagged effects. The findings emphasize the methodological importance of prewhitening in transfer function modeling, particularly for environmental time series characterized by complex internal structure. The results offer practical insight into how meteorological dynamics influence short-term variations in CO2 levels and provide a robust foundation for future modeling and forecasting efforts.