Mapping cascades in urban lifeline networks

Ask a city's flood model what happens at +40 cm above forecast and it answers in hectares. Ask the emergency manager and they answer in services: which substation, which pump station, which hospital, in what order, with what slack. The distance between those two answers is where most territorial risk work currently dies.

Lifeline networks fail as a system because they are coupled at specific, physical points: the substation that feeds the pump hall, the pump hall that keeps the hospital basement dry, the road segment that is both evacuation route and utility access. These couplings are rarely in any single agency's dataset. Each operator maps their own network and treats the others as boundary conditions.

Modeling the cascade therefore starts as an information problem, not a simulation problem. The entities must be enumerated, their dependencies made explicit, and the whole treated as one interaction network, the same system-decomposition discipline industrial safety has used for decades, applied at territorial scale.

Once the network exists, propagation is tractable: an initiating hazard reaches its first targets through physical fluxes; service-loss fluxes carry it onward. The hard part is honesty about data. A cascade model that pretends to know feeder-level grid topology it doesn't have is worse than none. Assessment has to degrade gracefully with data quality, and say so.

We are building this discipline into Aegis with our government design partners, one territory at a time. If your agency holds the lifeline question nobody's model answers, we would like to hear it.