ADAPTIVE STRATEGY FOR HURRICANE SURGE FLOODING UNDER UNCERTAIN CLIMATE: A CASE STUDY FOR NEW YORK CITY LOWER MANHATTAN TRIBUTARIES FOCUS AREA

Abstract

Rising sea-level and potential alteration of hurricane climatology due to climate change will escalate the risk for coastal flooding and exploit currently delicate coastal infrastructures. Existing codes and risk mitigations in response to the anticipated worsening of coastal floods have been outlining design criteria for certain risk aversion levels and particular sea-level projections. However, designing for certain assumptions and static conditions without quantifying the optimal level of protection will either expose vulnerable coastal communities to unanticipated extreme damages or spend astronomical budgets and overprotect the region. Previous studies have proposed the adaptation of dynamic floodwalls (i.e., floodwalls can be incremented periodically as anticipated risk of flooding increases) with a cost-benefit analysis to prepare for the future climate and extreme flood events while designing for the optimal protection level; and progress has been made on determining dynamic floodwall policies under deep uncertainty with climate condition observations and agreement on future adaptations. In this paper, we propose a novel Deep Learning (DL) scheme that dramatically reduces the order of complexity of algorithms from previous studies and achieves suboptimal results. We apply a wide range of design strategies including DL for floodwall protection level in the New York City lower Manhattan tributaries focus area and incorporate inundation modeling for floodwall overtopping damage estimation to compare with the recent New York and New Jersey Harbor and Tributaries Focus Area Feasibility Study (NYNJHATS) developed by the US Army Corps of Engineers (USACE), New York District. We show that DL is a suboptimal strategy, with significantly less computational cost, to Reinforcement Learning (RL) with its benefit of adaptation of a 71% to 163% (95% CI) improvement compared to the current strategy proposed in NYNJHATS under moderate emissions.