Abstract Flood-induced risks increasingly disrupt interdependent infrastructure systems. However, how these risks propagate, cluster or diverge across interconnected systems is unclear. To bridge this gap, this study constructs a multi-tiered risk-weighted network to disentangle interactions and reveal risk communities. UK-wide flood events from 2003 to 2023 were simulated using CaMa-Flood. Using these simulations, a flood-specific risk resistance surface, incorporating device, spatial, and external disturbance dimensions, was used to construct an adjacency graph for 29,697 EV charging points, yielding a risk-weighted network with 192,862,650 annual edges. This integrated framework enables a unified examination of flood-driven interaction pathways across scales. We find: (1) Flood risk propagation is driven respectively by spatial amplification (63.8%) at the node level, hybrid natural–built coupling (45.3%, 49.6%) at the buffer scale, and impact-driven dynamics (49.0%, 23.6%, 27.0%) at the disturbance layer, representing the relative contributions of entropy-weighted factor groups that form each resistance layer; (2) Risk propagation is governed not by group size, but by the interplay among structural coherence, spatial buffering, and fragmentation constraints across scales; and (3) Risk propagation emerges in fragmented, outward-oriented communities, where spatial geographical interplay, rather than path length, shapes intergroup cascades. These findings highlight multi-scale structural mechanisms driving flood-induced cascade behaviour.
Wan et al. (Fri,) studied this question.