Revealing Multi-Level Dynamic Spatial Community Patterns via Recursive Filtering of Travel-Distance Scales

Urban spatial structure is foundational for planning and management, yet its dynamic and multi-scale nature remains difficult to capture. Travel-flow networks derived from trajectory data provide a useful lens for revealing intra-urban interactions, but conventional community detection methods often overlook the scaling effect of travel distance and intra-day temporal dynamics. To address this limitation, this study proposes a novel framework to delineate urban structure at different scales, which identifies breakpoints in travel distances where interaction patterns shift. The method performs recursive community identification on the traffic flow, which is stratified by equal-distance intervals: it first detects communities using flows within a short-distance threshold, aggregates them into super-nodes, and then considers the short-distance flows and longer-distance flows between these communities and re-detects the communities at this level. When the community structure changes significantly, the flows from the previous distance range are excluded from further analysis, and the previous communities are treated as nodes for the next level, so that hierarchical organization is preserved and scale-specific interactions are isolated. Applying this method to Beijing taxi trajectory data across five time periods, a four-level spatial community structure is revealed: street-scale communities (short-distance flows), district-scale communities (medium-distance flows), urban–rural-scale communities (long-distance flows), and metropolitan-scale communities formed by ultra-long-distance flows. Significant intra-day variations are also identified, with morning-peak and night-time structures differing notably from other periods. The findings offer a multi-scale perspective on urban spatial organization and provide implications for traffic management and urban planning.