The interactions of carbon emission driving forces: Analysis based on interpretable machine learning

Abstract

Climate change caused by carbon emissions is a major threat to humanity. Understanding the drivers of carbon emissions has drawn much attention, but many studies overlook how these drivers interact. This study uses interpretable machine learning to explore the complex relationships between carbon emission intensity and its factors based on city-level data from China between 2002 and 2017. The findings show that as GDP per capita, population, population density, and knowledge complexity increase, their negative effects on carbon emission intensity weaken. In contrast, the positive effects of the secondary industry ratio and industrial land use grow stronger. FDI and innovation quantity show an inverted U-shaped relationship with carbon emission intensity. The study also finds significant interactions among these factors, such as between industrial land and population density, industrial structure and innovation, and economic development and population. SHapley Additive exPlanations (SHAP) identify economic development as the most important influencing factor, followed by population and industrial structure. These results provide valuable insights into controlling carbon emissions at different stages of urban growth and leveraging interactions among key drivers.