Thresholds and synergies: How urban spatial features collectively shape heavy-duty truck CO ₂ emissions

The carbon emissions produced by heavy-duty trucks (HDTs) are associated with complex interactions among various urban spatial features. However, nonlinear effects and interdependencies remain underexplored. This study integrates road network structure, facility density, and facility accessibility within a unified spatial grid framework across eight major inland cities in China. Using high-resolution GPS data and the COPERT model, we estimate HDT CO2 emissions and develop an interpretable stacking ensemble model. The SHAP-based analysis reveals three key insights. First, road network metrics (e.g., density, centrality) exhibit the strongest associations with emissions, but integrating multidimensional spatial features significantly enhances predictive accuracy and reveals synergistic effects. Second, nonlinear threshold patterns are identified: higher road density (>10.52 km/km2), industrial facility density (>5.43 units/km2), and strength centrality (>699.74) are positively associated with emissions, whereas higher residential and educational densities are negatively associated with emissions. Logistics hubs exhibit a U-shaped association pattern, with emissions peaking at 10.12 units/km2. Third, critical interactions are associated with changes in emission patterns: co-located logistics hubs and transport facilities correspond to higher emissions, while interactions between road network centrality and path length are associated with stronger positive associations with emissions. Our findings characterize spatial association patterns related to freight emissions and provide empirical references for threshold-sensitive and interaction-informed decarbonization strategies.