Nonlinear and spatial effects of urban form on building electricity consumption in Seoul, South Korea: Urban morphometric and XGBoost approaches

초록

Urban form is a critical factor in determining building energy consumption. Many previous studies have depended on aggregated urban form indicators and linear models in estimating their effects on energy consumption. This study measured fine-grained urban form indicators considering multiple building and street features and identified the nonlinear and spatial effects of urban form indicators on building electricity use. Our spatial econometric analysis confirmed the overall associations between urban form and electricity consumption. eXtreme Gradient Boosting (XGBoost) and SHapley Additive exPlanations (SHAP) estimated nonlinear and spatial heterogeneous effects. The empirical models provide three key findings. First, higher residential densities, greater floor areas, and compact covered areas have the largest contributions to electricity consumption. Second, the urban form's effects are strongly nonlinear, with threshold behaviors frequently missed in conventional models. Third, the SHAP values' spatial distribution indicates strong local heterogeneity, highlighting the need for context-dependent planning. By integrating explainable machine-learning and spatial econometric approaches with advanced urban form measures, this study suggests insightful implications for stakeholders such as policymakers, urban planners, and architects aiming to maximize energy efficiency through urban form strategies.

키워드