Decoding CO₂ futures in representative Green Belt and road countries: Evidence from China, India, and Russia

Abstract

Rising global temperatures pose a critical challenge for Belt and Road Initiative (BRI) nations, which must sustain economic growth while transitioning to low-carbon development. However, existing models often oversimplify emissions dynamics and their key drivers. The study developed a Bagging GA-BiLSTM framework and evaluated its performance against three machine learning models (LSTM, Extreme Gradient Boosting, Random Forest) and a conventional STIRPAT-OLS model, enabling high-precision projection of sectoral emissions across multiple scenarios (SSP1, SSP2, SSP3) and variables for China, India, and Russia from 2023 to 2060. The Logarithmic Mean Divisia Index (LMDI) method was employed to assess the contribution of driving factors and decoupling patterns. The Bagging GA-BiLSTM achieves higher R² and lower mean squared errors than other models. Under SSP1, China's emissions peak at 12.81 ± 2 Gt in 2027 and decline to $2.87\pm 0.4$ Gt by 2060. Under SSP2 and SSP3, carbon peak emissions are delayed to 2029. India's emissions are projected to peak between 2050 and 2051, with substantial scenario-based variability. Russia exhibits a contraction-recovery-rebound trajectory, with emissions peaking at $2.38\pm 0.3$Gt in 2058 under SSP1. LMDI decomposition reveals that carbon emission changes in China and India are primarily driven by industrial structure and energy intensity, reflecting the characteristics of rapidly developing economies. In contrast, Russia exhibits a post-industrial emission pattern dominated by economic and urbanization factors. These findings quantitatively reinforce existing policy pathways, proving that differentiated strategies-such as infrastructure investment in low-income countries, efficiency improvements in middle-income economies, and institutional reforms in resource-rich nations-remain essential for effective low-carbon transitions across the BRI. This study provides updated and detailed scenario-based evidence that strengthens and deepens the foundation for these established strategies.