Publication

Abstract : The research paper provides a comprehensive and effective comparison of various deep learning (DL) models applied to univariate data of carbon dioxide (CO2) concentrations spanning from 1870 to 2021, focusing on five European countries: the United Kingdom (UK), Germany, France, Ukraine, and Italy. The primary aim of the study is to contribute to a better understanding of climate change and to facilitate the development of policies and actions to mitigate CO2 emissions. Three DL models, namely long short-term memory (LSTM), convolutional neural network-long short-term memory (CNN-LSTM), and dynamic mode decomposition (DMD), are employed and thoroughly evaluated in this research. A novel aspect of this paper lies in the spatiotemporal modeling and forecasting capabilities, which enable simultaneous analysis for multiple countries using DMD. The results show that DMD outperforms other DL models, such as LSTM and CNN-LSTM, in accurately capturing data trends and forecasting future CO2 values with the highest precision. The findings presented in this research paper have significant implications for advancing our understanding of climate change dynamics and guiding more effective measures to tackle CO2 emissions.