Influencing Factors and Prediction Modelof Light-Duty Vehicle CO2 Emissionson Expressways in Mountainous Plateau Areas

We selected a 288 km expressway in the western Sichuan Plateau mountainous area to reveal the factors influencing CO 2 emissions from light-duty vehicles and establish a localized prediction model.Instantaneous CO 2 emission data (9,381 sets) were obtained through real-vehicle emission tests.The influential characteristics of CO 2 emission rates under different environmental characteristics, alignment conditions, and operational states were analyzed.CO 2 emission prediction models based on random forest (RF) and model-agnostic meta-learning (MAML) algorithms were constructed, compared, and analyzed.The findings indicated that: (1) Vehicle-specific power (VSP) was the most important factor determining the CO 2 emission rate.The feature importance of VSP in the upslope and downslope directions was 0.25 and 0.22, respectively; (2) The CO 2 emission rate distribution patterns were approximated by a Gamma distribution within different grade and angle change rate intervals.At grades < -1%, between -1 and 1%, and > 1%, CO 2 emission rates decreased, stabilized, and increased, respectively, with increasing angle change rate intervals; (3) The evaluation metrics for the MAML model outperformed those of the RF model, indicating higher adaptability to unknown tasks.