Scholarlink

<p>Transportation is a key sector driving the growth of carbon emissions in China, making its emission reduction pathways critical for achieving the “dual carbon” goals. However, existing studies predominantly rely on traditional statistical methods, which struggle to capture the complex nonlinear characteristics of carbon emissions and often lack systematic analysis of supply-side driving factors. To address this gap, this study adopts a supply-side perspective by improving the Logarithmic Mean Divisia Index (LMDI) method with a Cobb-Douglas (C-D) production function to decompose the influencing factors of CO<sub>2</sub> emission changes. Additionally, the Tapio decoupling model is employed to reveal the decoupling states between carbon emissions and eco-nomic development. The decomposition results from the CD-LMDI model are used as predictive indicators, and a hybrid Convolutional Neural Networks (CNN)-Gated Recur-rent Units (GRU)-Attention model is constructed to predict transportation carbon emissions. The results indicate that: (1) The CD-LMDI model quantitatively evaluates the effects of capital input, labor input, and technological progress on transportation carbon emissions. During 2000–2022, transport intensity and capital input were the primary inhibiting and promoting factors for emission growth, respectively. (2) The CNN-GRU-Attention model demonstrated smaller errors in MSE, MAE, and MAPE com-pared to the LSTM, CNN-RNN, and CNN-LSTM models, resulting in more accurate and reliable predictions. (3) Under the baseline scenario, optimistic scenario, and pessimistic scenario, the CO<sub>2</sub> emissions of China’s transportation industry will peak in 2040, 2035, and 2045, respectively, with peak carbon emissions of approximately 1.272&#xa0;billion tons, 1.176&#xa0;billion tons, and 1.374&#xa0;billion tons.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Decoupling effect and scenario prediction of carbon emission in transportation industry based on CD-LMDI and CNN-GRU-attention model

  • 期刊论文

  • 发表日期：

    2026年3月4日

查看全文
  • Xu Xizhen,
  • Liu Yuming,
  • Ou Guoliang

摘要

Transportation is a key sector driving the growth of carbon emissions in China, making its emission reduction pathways critical for achieving the “dual carbon” goals. However, existing studies predominantly rely on traditional statistical methods, which struggle to capture the complex nonlinear characteristics of carbon emissions and often lack systematic analysis of supply-side driving factors. To address this gap, this study adopts a supply-side perspective by improving the Logarithmic Mean Divisia Index (LMDI) method with a Cobb-Douglas (C-D) production function to decompose the influencing factors of CO2 emission changes. Additionally, the Tapio decoupling model is employed to reveal the decoupling states between carbon emissions and eco-nomic development. The decomposition results from the CD-LMDI model are used as predictive indicators, and a hybrid Convolutional Neural Networks (CNN)-Gated Recur-rent Units (GRU)-Attention model is constructed to predict transportation carbon emissions. The results indicate that: (1) The CD-LMDI model quantitatively evaluates the effects of capital input, labor input, and technological progress on transportation carbon emissions. During 2000–2022, transport intensity and capital input were the primary inhibiting and promoting factors for emission growth, respectively. (2) The CNN-GRU-Attention model demonstrated smaller errors in MSE, MAE, and MAPE com-pared to the LSTM, CNN-RNN, and CNN-LSTM models, resulting in more accurate and reliable predictions. (3) Under the baseline scenario, optimistic scenario, and pessimistic scenario, the CO2 emissions of China’s transportation industry will peak in 2040, 2035, and 2045, respectively, with peak carbon emissions of approximately 1.272 billion tons, 1.176 billion tons, and 1.374 billion tons.