Multiscale Extremity of Intense Rainstorms Induced by Spatiotemporal Variabilities: Inconsistencies and Constraints

Short-duration intense rainstorms drive flash floods in urban watersheds, yet their spatiotemporal variability (STV) complicates multiscale severity assessment. While equal exceedance probability across scales is commonly assumed, its validity and consequences in hazard assessment remain poorly evaluated. Here, we investigate the (in)consistency in multiscale rainstorm extremity and examine how it is moderated by STV factors, using radar-based storm tracking and multivariate frequency analysis of 41 Hong Kong watersheds. Results reveal significant inconsistencies across spatial (1–10 km) and temporal (6 min–daily) scales. The mean consistency index decreases with return period, constraining multiscale severities from reaching the same extremities, which potentially leads to discrepancies exceeding that between 2-year and 100-year events. Scale-dependent STV characteristics critically moderate such inconsistencies: larger-scale extremes originate from slower-moving rainstorms with lower spatial heterogeneity and spatiotemporal aggregation. Realistic hazard assessment thus requires joint probability frameworks incorporating STV interdependencies and multiscale constraints, rather than deterministic shortcuts.