Regional Carbon-Peaking Pathways for Sustainable Transportation in China Based on an Integrated STIRPAT–LightGBM Model

Provincial differences in economic growth, energy use, and infrastructure make China’s transport-related emissions difficult to represent with a single national pathway. Using a 2006–2023 provincial panel for transportation, warehousing, and postal services, this study estimates emissions and evaluates six regional scenarios for 2024–2035. Predictors are selected through correlation screening and random forest within time-ordered training windows, after which an extended STIRPAT–Ridge model is combined with LightGBM. On the 2021–2023 holdout, the integrated model records a conditional MAPE of 4.56%, close to 4.82% for a four-factor accounting benchmark, while end-to-end MAPE rises to 12.11% when all drivers are projected. A driver-adjusted province fixed-effect and trend model records conditional and end-to-end MAPEs of 3.32% and 10.79%. Under the development–emission reduction coordination scenario, the eastern, western, central, and northeastern regions reach conditional maxima in 2030, 2031, 2032, and 2032, respectively. The scenario comparison shows how regional assumptions about energy intensity, economic activity, and transport structure shape the timing and level of future emissions.