Investigating the spatiotemporal variability of mobility-based centrality in Seoul using mobile phone signaling data

Abstract

Urban centrality has often been examined using static models that overlook the dynamic rhythms of mobility. This study investigates the spatiotemporal variability of mobility-based centrality in Seoul by integrating PageRank centrality derived from mobile phone signaling data using a time-sliced Multiscale Geographically Weighted Regression (MGWR) framework. Origin-destination networks were constructed for four representative hours to capture weekday cycles and assess how demographic, socioeconomic, and land-use factors shape centrality across space and time. Because PageRank is normalized within each time slice, cross-time differences are interpreted as shifts in relative hierarchy rather than absolute trip volumes. The results revealed distinct temporal and spatial heterogeneities. PageRank showed centrality concentrated in the three business districts during the morning, diffusion at midday, expansion toward peripheral areas in the evening, and flattening late at night. MGWR highlighted variable-specific dynamics: employment density strongly influenced morning centrality but declined thereafter; household income showed context-dependent effects, including late-evening negative associations in peripheral districts; and single-person households exhibited spatially selective positive associations during daytime periods but became statistically undetectable late at night. The findings highlight the importance of when and where different attributes shape centrality, suggesting that mobility policies should be time-sensitive and spatially targeted, including efforts to mitigate peak-hour congestion, enhance off-peak accessibility, and adapt infrastructure to ongoing demographic changes.