Multidimensional urban morphology and carbon intensity: Nonlinear drivers in dense cities

Urban morphology (UM) significantly influences carbon emissions (CEs), yet fine-scale mechanisms involving three-dimensional (3D) structure and natural substrates remain a "black box." Using refined block delineation with explainable machine learning (XGBoost-SHAP) to analyze 708 functionally distinct blocks in Wuhan, a high-density metropolis. Results indicate that natural geographic elements, 3D morphology, and human activity intensity outweigh traditional two-dimensional (2D) metrics as drivers. We identified three core mechanisms: (1) Geographic constraints dominate, with slope as the primary driver whose interaction with floor area ratio (FAR) reveals synergistic risks. (2) Nonlinear thresholds exist, exemplified by FAR's "step-like" emission surge after 1.0-1.5 and the sky view factor (SVF)'s pronounced L-shaped reduction. (3) Functional heterogeneity determines impact, evidenced by a hybrid morphology-emission mechanism in industrial blocks (IBs) and FAR's directional reversal in green space (GS). These findings demonstrate that low-carbon planning must shift toward multidimensional management informed by geographic constraints, nonlinear thresholds, and functional zoning.