Sustainable Development Goal (SDG) 11 emphasizes sustainable urban development and enables cross-country and regional comparisons. However, conventional city-scale assessments often fail to capture fine-scale spatial heterogeneity within cities, particularly across Urban Functional Zones (UFZs). This study introduces an integrated framework that combines Very High Resolution (VHR) satellite imagery, geospatial data, and deep learning techniques to evaluate SDG11.2 (public transport accessibility) at the UFZ level. A customized deep learning framework, E-UFZ, was developed to accurately extract UFZs through three key procedures: (1) segmentation of UFZ units using the multi-resolution segmentation (MRS) method, (2) feature extraction via the CBAM-Deeplab, and (3) object-oriented classification using a Markov random field (MRF). Comparative validation against other methods demonstrated that E-UFZ achieved superior performance with an overall accuracy of 87.25%, effectively capturing fine-scale urban heterogeneity. UFZ-scale analysis across five Chinese cities-Shanghai, Suzhou, Hangzhou, Hefei, and Nanjing-reveals pronounced spatial disparities in SDG11.2 indicators. Residential (referring to ordinary residential communities), institutional, and commercial zones show higher accessibility than other UFZs. By establishing a refined paradigm for UFZ-scale urban transport sustainability assessment, this framework demonstrates strong potential for extension to other UFZ-related SDG indicators and adaptation to diverse urban contexts globally.
Yuan et al. (2026) studied this question.