Exploring the nonlinear impacts of the intra-urban built environment on carbon emissions using an interpretable ensemble machine learning framework

The urban built environment (UBE) substantially shapes carbon emissions trajectories, yet existing studies largely rely on fragmented dimension analyses, coarse administrative units, and homogeneous linear assumptions, leaving the comprehensively multidimensional, spatiotemporally heterogeneous, and nonlinear interplay between UBE and carbon emissions largely understudied, particularly from an intra-urban perspective reflecting real-world anthropogenic activities. Taking 286 Chinese prefecture-level cities from 2010 to 2024 as testbeds, this study architected an intra-urban UBE framework encompassing 5D (e.g. , density, diversity, design, destination accessibility, and distance to transit ) and developed an interpretable ensemble learning model to decouple the intricate association between intra-urban UBE and carbon emissions. The result reveals that intra-urban UBE collectively accounts for 61.51% of the impacts, surpassing the socioeconomic factors. Moreover, the results identify non-linear thresholds and interactive effects: density and diversity exhibit inverted U-shaped relationships, whereas urban compactness and urban green and blue spaces show consistent negative associations. Critically, active transport networks reduce emissions only when combined with public transit proximity; road density expansion cuts emissions only in compact cities; and polycentricity with compactness achieves emission reduction only within a moderate window. Temporally, the relative importance of greenness, active transport, diversity, and polycentricity has increased, while that of density has declined. Spatially, the effects of intra-urban UBE vary markedly across regions and cities. Polycentricity works best in emerging agglomerations, and transit proximity has reached diminishing returns in core megacities. The findings underscore that low-carbon urban policy must be spatiotemporally tailored with context-sensitive thresholds, positioning intra-urban UBE as a foundational climate mitigation instrument.