The impact of inter-country differences on transportation carbon emission transfer

Cross-border differences shape the generation and transfer of carbon emissions through international transport. This study develops a sector-specific accounting framework by integrating an environmentally extended multi-region input–output (EE-MRIO) model with interpretable machine learning. Our framework tracks and explains bilateral carbon transfers among 64 nations from 2000 to 2018. We first compute annual country-pair transfer matrices. Key drivers are then identified using the CatBoost algorithm, which employs Bayesian optimization for hyperparameter tuning after mutual-information screening and recursive feature elimination. Model explanations are derived from SHapley Additive exPlanations (SHAP) values. Subsequently, bilateral transfer pathways are classified via hierarchical clustering. Results indicate a steady increase and an eastward shift in transport-related carbon transfers. By 2018, China had become both the largest exporter and importer. Transport-specific indicators are dominant mechanisms: stronger liner-shipping connectivity, container port traffic, and international tourism amplify transfers, whereas rail development suppresses them. Disparities in manufacturing capacity and technological capability reduce outward transfers. This suggests that industrial upgrading and innovation help mitigate carbon spillovers. Clustering analysis reveals three recurrent pathway typologies: (1) a specialized transport–innovation pattern; (2) an investment–manufacturing intensive pattern; and (3) a background pattern with weak transport-specific drivers. Policy implications highlight the need for tailored interventions, such as establishing green corridors, fostering industrial collaboration for modal shift, and promoting system-wide digitalization. The integrated EE-MRIO and machine-learning approach offers a scalable and interpretable tool for sectoral mitigation analysis.