Assessing urban renewal opportunities by combining 3D building information and geographic big data

Abstract

The assessment of urban renewal (UR) potential aims to prioritize areas for UR, which are essential for sustainable urban revitalization. However, conventional data sources often fall short in encompassing diverse urban characteristics in the evaluation process, such as urban three-dimensional (3D) building information and the intensity of human activities. To address this gap, this study integrated 3D building data and geographic data to create a comprehensive set of 28 indicators spanning four dimensions: natural environmental conditions, land use, socio-economic factors, and building conditions. These indicators take into account vertical dimensions, dynamic aspects, and fine-scale details. Leveraging the existing UR practices as a positive example, we established an UR potential assessment model at street block scale using Presence and Background Learning combined with extreme gradient boosting algorithms (PBLXGBoost). Our findings revealed that the highest accuracy in evaluating industrial UR potential was achieved in Shenzhen (Fpb_avg = 0.80, RMSEavg = 0.21), followed by residential and commercial UR potential assessments. Conversely, other type UR exhibit lower accuracy. Street blocks with significant UR potential are predominantly located in Bao’an, Longgang, and Longhua Districts. Furthermore, employing the SHAP model to elucidate the evaluation results uncovered intricate hierarchical, positive-negative, and overlapping relationships among various factors and different UR types, where geographic big data and 3D building information showed strong correlations. The methodology proposed in this study enables objective and precise assessments of UR potential, offering valuable support for UR practice and sustainable urban development.

Keywords

• Urban renewal
• potential evaluation
• vertical dimension
• dynamic perspective
• fine scale
• Presence and Background Learning (PBL)