Rapid Reconstruction of Storm Surge Footprints Using Principal Component Analysis and SFINCS Under Sea-Level Rise Scenarios

Storm surge, especially when combined with sea-level rise, is a key driver of coastal flood risk for communities and critical assets in tropical cyclone-prone regions. For applications such as insurance underwriting and portfolio risk assessment, probabilistic analyses based on natural catastrophe model require very large synthetic event sets (e.g., tens of thousands of storms), which in principle demand high-resolution simulations over extensive coastal domains. However, state-of-the-art hydrodynamic models are computationally expensive at fine spatial resolution and even reduced-physics models such as SFINCS (Super-Fast INundation of CoastS) are still computationally intensive when applied at high resolution to very large event sets. These computational constraints hinder timely and comprehensive assessments of storm surge hazards under current conditions and future sea-level-rise scenarios.

To address this challenge, a fast reconstruction framework that combines Principal Component Analysis (PCA) with SFINCS is developed to efficiently generate high-resolution storm surge footprints from a large event set. Using Busan, South Korea, as a case study, low-resolution simulations are conducted for a large set of storm surge events and maximum water level profiles at coastal stations are extracted. PCA is applied to these profiles to identify the dominant modes of variability. The resulting principal components form the basis for selecting a reduced subset of representative events via K-means clustering, which are then simulated at high resolution. Storm surge footprints for the full event set are then reconstructed through weighted interpolation in the space spanned by PCA components.

Retaining principal components that explain more than 90% of the variance, the proposed methodology achieves mean reconstruction errors below 0.25m and Jaccard indices above 0.8 across all deciles. Furthermore, applying the same PCA loading vectors to sea-level-rise scenarios yields mean reconstruction errors below 0.15 m and Jaccard indices above 0.95 across all deciles. Relative to benchmark simulations of the full event set, the approach substantially reduces computational time while preserving spatial accuracy, including elevated sea-level conditions. This PCA-SFINCS framework enhances efficient coastal hazard modeling, with potential extensions incorporating cyclonic parameters and optimization techniques. Beyond academic applications, it offers substantial value for industry-oriented risk management by enabling rapid and robust assessments of storm surge hazards under various sea levels under climate change scenarios. This framework supports systematic evaluation of risk variations and facilitates timely decision-making in coastal planning, insurance underwriting, and infrastructure resilience for industry-oriented stakeholders.