Tuning deep flooding risk with adaptive strategy: An application for NYC

Abstract

The impact of the storms may worsen in the coming decades due to the rapid development of the coastal zone in conjunction with sea-level rise and possibly increased storm activity due to climate change. Greater progress on coastal flood risk management is urgently needed. Previous studies proposed designs of dynamic seawalls (i.e., seawalls that can be heightened overtime to cope with the increasing effect of climate change), based on long-term climate model projections. However, significant uncertainties exist in long-term climate projections. Noticing that the climate condition can be observed over time, we develop a reinforcement-learning-based strategy of adaptive seawall design (i.e., the design is planned to be regularly updated based on observations), to cope with the deep uncertainty in climate change effects. We apply this method to New York City and estimate its optimal adaptive seawall design, based on climate projections of sea-level rise and storm surge flooding, building level exposure data, and estimated construction cost of the seawall. We show that the total lifetime cost (including the investment of the seawall and potential damage of the protected area) is significantly reduced (by 20% to 40%) when the dynamic, reinforced learning strategy is applied, compared to traditional design methods.