Deciphering the spatiotemporal dynamics and driving mechanisms of carbon emissions in China's Greater Bay area: Insights from interpretable machine learning

Abstract

The Guangdong-Hong Kong-Macao Greater Bay Area (GBA) ranks among the four most prominent bay areas globally. Despite comprising less than 0.6% of China`s land area, the GBA generates over 10% of the nation`s economic output. Analyzing the spatiotemporal dynamics of carbon emission (CE) and their drivers is crucial for achieving “dual-carbon” targets and facilitating a green, low-carbon transition in global bay area economies. An XGBoost model, incorporating nighttime light data (NTL) , land surface temperature (LST) , population, and GDP, reconstructs energy-related CE evolution. Spatiotemporal CE dynamics within the GBA were elucidated through multi-dimensional spatial analysis. An XGBoost-SHAP model developed to identify key drivers, revealing nonlinear and interactive effects to inform optimization strategies. The findings show that ((1) Emissions exhibited a two-stage growth pattern: rapid expansion followed by stabilization. (2) Spatially, emissions shifted from a “three-core” model to a “multi-core” model. (3) Spatial agglomeration and heterogeneity coexisted, with a “high-center, low-periphery” distribution. (4) Municipal CE patterns demonstrated significant path dependence and lock-in, though this inertia gradually decreased, accompanied by increased spillover effects from high-emission areas. (5) Population, economic development, and urbanization were primary CE drivers, transitioning from scale-driven to quality-dominated growth. This study elucidates the complex dynamics of circular economy (CE) in rapidly developing regions, such as the GBA, supporting “dual-carbon” goals and sustainable development.