Forecast-Informed Hysteretic Routing for Stable Urban Flood Networks

Urban flood route guidance must anticipate disruption without causing unnecessary route changes. We test whether short-horizon flood forecasts can determine when to replace a currently usable route. Predictive risk-score hysteretic routing (PRH) estimates next-hour segment threshold-crossing risk, generates risk-aware alternatives and retains the current recommendation until a forecast-based release condition is met. The method was developed on four simulated flood events and evaluated with frozen parameters on 38 held-out events in Norfolk, Virginia. PRH produced 8,285 next-hour path threshold crossings, 11.1 per cent fewer than 10 per cent bounded-detour retention, and 11,517 route switches, 23.4 per cent fewer than repeated shortest-path recomputation. Matched controls and ablations attributed this operating point to the interaction of risk-weighted candidates and retention. In a strict trigger-signal replacement, linear extrapolation reduced crossings further but required longer detours, showing that segment-level forecast accuracy alone does not determine route-control performance. The main contribution is a simulation-level decision framework for balancing threshold crossings, route switching and detour cost, rather than an operational route-safety claim.