Abstract: The upper reaches of the Yangtze River serve as a critical area for implementing Chinas “dual carbon” strategy and constructing the ecological barrier of the Yangtze River Economic Belt, whose carbon emission governance in the transportation sector plays a significant role in achieving the objective of “synergistic enhancement of pollution reduction and carbon reduction.” Based on data from Chongqing, Guizhou, Sichuan, and Yunnan provinces/municipalities from 2000 to 2021, Lasso regression was employed to address multicollinearity of the STIRPAT model, a carbon emission prediction model integrating the grey wolf optimizer algorithm with random forest (GWO-RF) was constructed, and multi-scenario analysis was carried out. The research results indicate that under the baseline and low-carbon scenarios, carbon emissions from the transportation sector in this region are projected to peak in 2032, with peaks of 109.19 Mt CO2 and 104.08Mt CO2, respectively. Under the high-carbon scenario, the peak is delayed to 2034, with the peak value rising to 117.51 Mt CO2. From 2022 to 2040, carbon emissions from transportation sector in this region generally follow an evolutionary path of “initial growth followed by peaking, then gradual stabilization and decline”. This achievement can provide methodological references and decision-making basis for accurate prediction of carbon emissions and design of peak-reaching pathways in cross-regional transportation sectors.

Key words: traffic engineering; carbon emissions; machine learning; scenario analysis; grey wolf optimizer algorithm (GWO)

摘要： 长江上游地区作为我国 “双碳”战略实施与长江经济带生态屏障建设的关键区域，其交通运输业碳排放治理对实现“减污降碳协同增效”目标具有重要作用。基于2000—2021年重庆、 贵州、 四川、 云南这四省市的数据，运用Lasso回归对STIRPAT模型的多重共线性进行处理，构建融合灰狼优化算法与随机森林(GWO-RF)的碳排放预测模型, 并进行多情景分析。研究结果显示：在基准情景与低碳情景下，该地区交通运输业碳排放预计于2032年达峰，峰值分别为109.19、 104.08 Mt CO2；在高碳情景下，达峰时间将推迟至2034年，峰值上升至117.51 Mt CO2。2022—2040年间，该地区交通运输业碳排放总体呈现“先增长后达峰，随后逐步趋稳下降”的演变路径。该成果可为跨区域交通运输业碳排放精准预测与达峰路径设计提供方法参考与决策依据。

关键词: 交通工程； 碳排放； 机器学习； 情景分析； 灰狼优化算法(GWO)