Beyond static census: Hourly mobile phone data reveals 2.6-fold underestimation of urban flood exposure

Accurate population exposure assessment is critical for urban flood disaster management, yet traditional approaches rely on static census data that fail to capture dynamic activity patterns. Using hourly mobile phone signaling data from four rainstorm events in Suqian, China, in 2023, we develop an analysis framework integrating baseline estimation, anomaly pattern recognition, and severity-graded dynamic exposure assessment. Hourly analysis reveals a W-shaped population response pattern during a prolonged daytime rainfall, with dual troughs aligned with morning and evening commutes. This demonstrates that disaster impacts couple with urban activity rhythms rather than depending solely on rainfall intensity. Population behavioral responses reveal resilience inequalities across functional zones. Industrial areas show high disturbance sensitivity, with 69–97.5% of grids exhibiting declining population responses, while residential zones demonstrate sheltering capacity, with up to 28.6% of grids showing ascending responses during disaster events. Critically, dynamic exposure assessment reveals that static census-based methods are estimated to underestimate exposure by 2.6-fold on average, with underestimation occurring in over 84% of industrial zones during daytime events. Our differentiated risk zoning framework identifies “High-High,” “Population-Driven,” and “Hazard-Driven” priority zones, enabling targeted resource allocation aligned with specific risk drivers. These findings suggest that hourly-resolution dynamic assessment can play a critical role in accurate flood exposure characterization and evidence-based emergency planning in rapidly urbanizing contexts.