Accurate carbon emission prediction is critical for regional low-carbon transitions and the realization of China’s “dual carbon” goals. However, carbon emission systems exhibit significant complex nonlinear relationships and time-dependent characteristics, making it difficult for traditional statistical methods and conventional neural networks to fully capture their dynamic evolution. To address this challenge, this paper proposes a multivariate carbon emission prediction method that integrates a Temporal Dendritic Neural Model (TDNM) with a Dendritic Adaptive Learning (DAL) algorithm. Based on panel data from 54 prefecture-level cities in China spanning 1999 to 2023, this study systematically constructs a carbon emission driving factor system comprising 13 indicators across economic, social, and energy dimensions. Empirical results indicate that the proposed model achieves competitive predictive performance and improved stability compared with several conventional machine learning models and deep learning baselines. Furthermore, taking the carbon emission predictions of Suzhou City as a starting point, a detailed scenario analysis is conducted to clarify the evolution trends of future carbon emissions under different development pathways, providing methodological references for analyzing carbon emission evolution and low-carbon development pathways in industrial cities.
Zhang et al. (Tue,) studied this question.