[Paper Review] A Novel Hybrid Machine Learning Model for Rapid Assessment of Wave and Storm Surge Responses Over an Extended Coastal Region

Saeed Saviz Naeini, Reda Snaiki|arXiv (Cornell University)|Jan 11, 2024

Tropical and Extratropical Cyclones ResearchEarth and Planetary Sciences3 citations

TL;DR

This study presents a novel hybrid machine learning model that combines a deep autoencoder (DAE) for dimensionality reduction and a deep neural network (DNN) for mapping storm parameters to wave and storm surge responses, enabling rapid, accurate predictions over extended coastal regions. The model achieves high accuracy and computational efficiency, outperforming decoupled approaches using PCA or DAE alone, and is suitable for early warning systems and probabilistic risk assessment.

ABSTRACT

Storm surge and waves are responsible for a substantial portion of tropical and extratropical cyclones-related damages. While high-fidelity numerical models have significantly advanced the simulation accuracy of storm surge and waves, they are not practical to be employed for probabilistic analysis, risk assessment or rapid prediction due to their high computational demands. In this study, a novel hybrid model combining dimensionality reduction and data-driven techniques is developed for rapid assessment of waves and storm surge responses over an extended coastal region. Specifically, the hybrid model simultaneously identifies a low-dimensional representation of the high-dimensional spatial system based on a deep autoencoder (DAE) while mapping the storm parameters to the obtained low-dimensional latent space using a deep neural network (DNN). To train the hybrid model, a combined weighted loss function is designed to encourage a balance between DAE and DNN training and achieve the best accuracy. The performance of the hybrid model is evaluated through a case study using the synthetic data from the North Atlantic Comprehensive Coastal Study (NACCS) covering critical regions within New York and New Jersey. In addition, the proposed approach is compared with two decoupled models where the regression model is based on DNN and the reduction techniques are either principal component analysis (PCA) or DAE which are trained separately from the DNN model. High accuracy and computational efficiency are observed for the hybrid model which could be readily implemented as part of early warning systems or probabilistic risk assessment of waves and storm surge.

Motivation & Objective

• To address the computational infeasibility of high-fidelity numerical models for probabilistic risk assessment and real-time prediction of storm surge and wave impacts.
• To develop a data-driven framework that enables rapid assessment of wave and surge responses across extended coastal regions.
• To integrate dimensionality reduction with regression modeling in a joint training framework to improve prediction accuracy and efficiency.
• To evaluate the model’s performance against decoupled approaches using PCA and DAE in isolation from the regression model.
• To demonstrate the model’s applicability in operational early warning systems and coastal risk assessment.

Proposed method

• A deep autoencoder (DAE) is used to learn a low-dimensional latent representation of high-dimensional spatial wave and surge fields.
• A deep neural network (DNN) is trained to map storm parameters (e.g., wind speed, pressure, track) to the latent space learned by the DAE.
• A combined weighted loss function is designed to jointly optimize the DAE and DNN components, balancing reconstruction accuracy and prediction performance.
• The model is trained and validated using synthetic data generated from the North Atlantic Comprehensive Coastal Study (NACCS) for New York and New Jersey.
• Performance is benchmarked against two decoupled models: one using PCA + DNN and another using DAE + DNN, where the reduction and regression models are trained separately.
• The hybrid architecture enables end-to-end learning, preserving spatial coherence and improving generalization over extended coastal domains.

Experimental results

Research questions

• RQ1 Can a joint training framework of dimensionality reduction and regression improve prediction accuracy for wave and storm surge responses compared to decoupled models?
• RQ2 How does the performance of a DAE-based latent space compare to PCA in reducing spatial dimensionality for coastal hydrodynamic responses?
• RQ3 To what extent does the hybrid DAE-DNN model maintain spatial fidelity and prediction accuracy while reducing computational cost?
• RQ4 Can the proposed model be effectively deployed in real-time early warning systems or probabilistic risk assessments?
• RQ5 What is the impact of a combined weighted loss function on the convergence and accuracy of the hybrid model?

Key findings

• The hybrid DAE-DNN model achieves significantly higher prediction accuracy than decoupled models using PCA or DAE with DNN.
• The model reduces computational cost substantially while maintaining high spatial and temporal resolution in wave and surge predictions.
• The joint training with a weighted loss function improves generalization and prevents overfitting in the latent space representation.
• The model demonstrates robust performance across diverse storm scenarios in the NACCS synthetic dataset for New York and New Jersey.
• The hybrid approach preserves spatial coherence of storm surge and wave fields better than PCA-based alternatives.
• The model is computationally efficient enough to support real-time or probabilistic risk assessment applications in coastal hazard management.