The combined use of metro and bus (MBCT) enhances public transport accessibility and reduces CO2 emissions. However, its emission reduction potential varies across trips. This study assesses trip-level CO2 emission reductions of MBCT using interpretable machine learning. Smart card and bus operation data are integrated to infer complete origins and destinations (ODs). CO2 emission savings are quantified by comparing MBCT with car travel for the same OD pairs. Various Gradient Boosting Machines (GBDT, XGBoost, LightGBM, CatBoost) model the nonlinear relationships between emission reduction and influencing factors, with methods of SHAP and PDP identifying key determinants. An empirical study in Shanghai reveals that MBCT reduces CO2 emissions by 169.5 g/km on average. The reduction rate depends on trip characteristics and the built environment, with metro travel proportion and the non-linear coefficients being the most crucial factors. These findings offer insights for enhancing metro-bus integration and fostering low-carbon urban transport toward carbon neutrality.
Stored only when both papers are in the corpus — 8 citations exist upstream.
Imported from OpenAlex (CC0) — nothing here is generated.