Statistical Analysis and Modeling of the CO2 Series Emitted by Thirty European Countries

Abstract

In recent decades, an increase in the earth’s atmospheric temperature has been noticed due to the augmentation of the volume of gases with the greenhouse effect (GHG) released into the atmosphere. To reduce this effect, the European Union’s directives indicate the action directions for reducing these emissions, among which carbon dioxide (CO2) recorded the highest amount. In this context, the article analyzes the CO2 series reported in 1990–2021 by 30 European countries. The Kruskal-Wallis test rejected the hypothesis that the series comes from the same underlying distribution. The Anderson-Darling test rejected the normality hypothesis for seven series out of thirty, and Sen’s procedure found a decreasing trend slope only for 17 series. ARIMA models have been built for all individual series. Grouping the series (by the k-means and hierarchical clustering) provided the base for building the Regional series (RegS), which describes the CO2 pollution evolution over Europe. The advantage of this approach is to provide the synthetic image of the regional evolution of the CO2 emission volume (mt), incorporating information from 30 series (one for each country) in only one—RegS. It is also shown that selecting the number of clusters involved in building RegS and assessing their stability is essential for the model’s goodness of fit.