Understanding the CO2 emissions and influencing factors of travelers' multiple modes can provide direction for energy conservation and emission reduction, which is of great significance for developing sustainable cities. Previous studies focused on the CO2 emissions of the transportation sector or individual modes. Which has overlooked the variations of emissions within the transport system. Hence, this study focuses on multiple modes (i.e., car, subway, bus, and bike) in the township in the Guangdong-Hong Kong-Macao Greater Bay Area. This study proposes a framework for exploring the spatial autocorrelation of urban transport emission structure based on ratios (i.e., CO2 emissions from each mode divided by total emissions) and key factors by combining spatial econometric model (i.e., Moran's I index and Spatial Error Model) and machine learning model (i.e., Random Forest and SHAP model). In addition, the spatial autocorrelation of ratios at different spatial scales is investigated. The results indicate the high spatial dependence in the ratios from each transport mode and Moran's I indices for four ratios are 0.883, 0.886, 0.706, and 0.776, respectively. In addition, subway and car ratios exhibit a negative spatial correlation (−0.798), and subway and bike show a positive correlation (0.570). Population density, road length, and land use diversity are the key drivers of CO2 emission ratios and have different effects on various transport modes. Furthermore, as the spatial scales expand from townships to distinct and city, the spatial autocorrelation of the ratios decreases. This study could provide policy implications for optimizing urban transport strategies and reducing CO2 emissions.
Read full abstract