Emergency responders in coastal cities are anticipated to provide effective evacuation of at-risk populations during the preparedness and response phases of coastal floods due to land-falling storms or cyclones. However, existing contingency plans primarily focus on the evacuation of the general public rather than special arrangement for elderly populations who constitute a large proportion of flood fatalities. Here we present a system-level methodology to elaborate citywide coastal flood evacuation plans for optimal deployment of shelters and effective transfer of elderly people with special needs. We conduct a comparative analysis between Shanghai and New York City, which are both among the most exposed coastal cities to storm-induced flooding but represent two distinct institutional systems of emergency operation. The results show marked disparities in evacuation patterns for elderly residents in the two cities. Storm flood evacuation is more challenging in Shanghai due to insufficient provision of shelter capacity (~230,000). Implementing risk-informed and strategic planning could not only meet the potentially huge demand of vulnerable elderly (~520,000) but also improve the overall efficiency of evacuee transfer by a factor of 3. Our work provides new insights into operational emergency evacuation decisions and informs flood management policy development for major coastal cities globally.

Sea level rise (SLR) and subsidence are expected to increase the risk of flooding and reliance on flood defenses for cities built on deltas. Here, we combine reliability analysis with hydrodynamic modeling to quantify the effect of projected relative SLR on dike failures and flood hazards for Shanghai, one of the most exposed delta cities. We find that flood inundation is likely to occur in low-lying and poorly protected periurban/rural areas of the city even under the present-day sea level. However, without adaptation measures, the risk increases by a factor of 3–160 across the densely populated floodplain under projected SLR to 2100. Impacts of frequent flood events are predicted to be more affected by SLR than those with longer return periods. Our results imply that including reliability-based dike failures in flood simulations enables more credible flood risk assessment for global delta cities where conventional methods have assumed either overtopping only or complete failure.