Key points are not available for this paper at this time.
Understanding city hierarchy is crucial for assessing city development and informing policy-making. Over the past decades, most city hierarchies have been established through network-based methods with a single data type (i.e. attribute or flow data). However, it is gradually recognized that applying a single data type to understand city hierarchy is inadequate. Although recent studies have combined attribute data and flow data for further analysis, these attempts remain relatively simple. Graph convolutional networks (GCNs) are powerful models that can integrate multiple data types for graph-structured analysis. Motivated by this, we employ GCNs to explore city hierarchy through a case study of 296 Chinese cities based on social, economic, and environmental attributes and traffic flow data. It is found that (1) an east–west disparity exists in city hierarchy, with levels decreasing westward, (2) a polycentric hierarchical structure has formed, and high-tier cities are concentrated in major urban agglomerations (e.g. Yangtze River Delta), and (3) city linkages are highly concentrated among high-tier cities in the city network, while low-tier cities face marginalization. These findings indicate that the GCN-based analysis for city hierarchy is an effective way to provide valuable insights into the mechanisms behind city hierarchy formation.
Wu et al. (Sun,) studied this question.