Addressing the US Tropical Cyclone-Storm Surge risk using RAFT-DeepSurge, an advanced AI-based approach 

• DOI: 10.5194/egusphere-egu25-2013
• Publication Date: 2025-03-14T16:48:36Z
• Authors: Karthik Balaguru; Julian Rice; David Judi; Ning Sun; Brent Daniel
• Keywords: Storm Surge; Raft

Extreme climate events or tails of natural hazard distributions tend to be the most damaging in terms societal impacts. While traditional physics-based approaches are suitable for gaining mechanistic understanding and process-based studies, they may not adequate characterizing probabilistic risk from extreme events. Here, we demonstrate a ML/AI-based approach estimating tropical cyclones (TCs) associated coastal flooding. First, simulate nearly one million TCs current future climates using Risk Analysis Framework Tropical Cyclones (RAFT), hybrid model that combines physics with deep neural networks. Subsequently, apply RAFT-simulated TC tracks DeepSurge, an AI-based storm surge is trained on large number ADCIRC simulations North Atlantic. Our results suggest significant increase northern Gulf Coast, Florida some areas near mid-Atlantic. Also importantly, show ML/AI can leveraged effectively address potential issue ‘Grey Swan’ their