Projecting Climate Dependent Coastal Flood Risk With a Hybrid Statistical Dynamical Model

Area of Science:

• Coastal geomorphology
• Climate change impacts
• Hydrodynamic modeling

Background:

• Traditional numerical models for coastal flooding are computationally intensive, limiting the simulation of diverse future climate scenarios.
• Accurate prediction of spatially varying flood surfaces is essential for understanding coastal hazards.

Purpose of the Study:

• To develop an efficient surrogate modeling framework for predicting nearshore and estuarine water levels under various offshore forcing conditions.
• To couple this framework with a climate emulator for downscaling hypothetical climate iterations.
• To assess present-day and future coastal flood risk, including event chronology and changes over time.

Main Methods:

• Development of a hybrid statistical-dynamical framework combining surrogate models (Gaussian process regression) with a climate emulator.
• Utilizing the Coastal Storm Modeling System (CoSMoS), including Delft3D and XBeach, as the dynamic simulator.
• Validation using in-situ tide gauge data and nearshore pressure sensor arrays.

Main Results:

• The framework efficiently predicts spatially varying nearshore and estuarine water levels for any combination of offshore forcing conditions.
• It assesses coastal flood risk, including the timing of flood and dune overtopping events.
• Reveals the influence of climate variability on flood resilience and the stochastic nature of future water levels.

Conclusions:

• The hybrid framework provides an efficient method for assessing current and future coastal flood risk.
• It aids in understanding the impact of climate variability on coastal resilience metrics.
• Enables detailed analysis of flood event characteristics and their evolution.