Unveiling causal dynamics and forecasting of urban carbon emissions in major emitting economies through multisource interaction

Abstract

Mitigating city carbon emissions in major emitting economies is vital for sustainable development. This study exploits the complex interactions among major sources of emissions to address two research gaps in city carbon neutrality: causal discovery and forecasting in high-emission economies through developing a novel two-stage algorithm. In the initial phase, the algorithm intricately unravels the causal dynamics of emissions from thousands of urban sources. Next, it employs a deciphered causal matrix to forecast city emission trends, leveraging insights from the prior analysis. It outperformed other advanced algorithms in tests using daily emission data from China, the European Union, and the United States. Evaluations, including prediction errors, Taylor diagrams, Diebold-Mariano tests, and ablation experiments, validate its superior ability in identifying inter-city carbon emission interactions and more accurately making 1-step and 5-step advanced predictions. This research confirms that analyzing interactions among major urban emission sources can effectively reveal the causal patterns of carbon emissions among cities and predict future trends. It also finds that while the causal network traits of urban emissions in leading carbon-emitting economies show consistency, distinct differences exist. Leveraging these insights, this study offers tailored policy implications for collaborative carbon emission reduction in cities within these economies.