Abstract Conventional subway analyses relying on static IC card data or limited surveys fail to capture the complex, dynamic interactions between passenger flows, land use patterns, and multimodal transportation choices in high-density urban environments. This study overcomes these methodological limitations by developing an innovative analytical framework that synergistically integrates large-scale mobile phone positioning data with traditional transport surveys and automated fare collection records. Using Guangzhou's extensive metro system (247 stations across 531 km) as a representative case study, we employ advanced data fusion techniques and machine learning algorithms to reconstruct complete travel chains and dynamically delineate station service areas with unprecedented spatial-temporal resolution. Our hybrid methodology combines multinomial logit modeling with random forest classification to systematically quantify subway competitiveness across different urban contexts, revealing three key findings: (1) distinct spatial thresholds for effective service areas (800m radius in central business districts vs. 2km in suburban corridors), (2) an inverted U-shaped relationship between parking supply-demand ratios and mode share with optimal balance at 1.75, and (3) significant equity disparities where low-income suburban commuters experience 2 times higher spatiotemporal costs than central city residents. The proposed framework provides urban planners with a robust, scalable tool for transit network optimization, offering particular value for rapidly urbanizing megacities in Asia and other developing regions. By effectively bridging cutting-edge big data analytics with established transportation modeling approaches, this research makes dual contributions: advancing theoretical understanding of fractal urban mobility patterns while delivering practical, data-driven strategies for sustainable transit development and equitable accessibility planning in high-density environments.
Li et al. (2025) studied this question.