Storm Surge in Color: RGB-Encoded Physics-Aware Deep Learning for Storm Surge Forecasting

2025-06-26

AI Summary

Current storm surge forecasting models suffer from low spatial resolution, reliance on sparse in-situ observations, poor generalizability, and incompatibility with modern deep learning architectures. To address these limitations, we propose a physics-informed RGB image encoding scheme that maps unstructured storm surge fields onto three-channel images, integrating dynamic observed wind fields with static topographic and bathymetric data. Leveraging this representation, we develop an end-to-end spatiotemporal forecasting model based on ConvLSTM. Our approach is the first to bridge hydrodynamic data and CNN-based architectures via RGB encoding, significantly enhancing model interpretability and cross-regional generalizability. Evaluated on large-scale synthetic storm datasets for the Gulf of Mexico, the model achieves high-accuracy 48-hour forecasts along the Texas coast, demonstrating strong robustness and spatial scalability.

Technology Category

• Computer Vision: Representation Learning for Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web search models and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Storm surge forecasting plays a crucial role in coastal disaster preparedness, yet existing machine learning approaches often suffer from limited spatial resolution, reliance on coastal station data, and poor generalization. Moreover, many prior models operate directly on unstructured spatial data, making them incompatible with modern deep learning architectures. In this work, we introduce a novel approach that projects unstructured water elevation fields onto structured Red Green Blue (RGB)-encoded image representations, enabling the application of Convolutional Long Short Term Memory (ConvLSTM) networks for end-to-end spatiotemporal surge forecasting. Our model further integrates ground-truth wind fields as dynamic conditioning signals and topo-bathymetry as a static input, capturing physically meaningful drivers of surge evolution. Evaluated on a large-scale dataset of synthetic storms in the Gulf of Mexico, our method demonstrates robust 48-hour forecasting performance across multiple regions along the Texas coast and exhibits strong spatial extensibility to other coastal areas. By combining structured representation, physically grounded forcings, and scalable deep learning, this study advances the frontier of storm surge forecasting in usability, adaptability, and interpretability.

Problem

• Improve storm surge forecasting spatial resolution and generalization
• Enable deep learning on unstructured spatial surge data
• Integrate physical drivers like wind and topography for accurate prediction

Innovation

• RGB-encoded image representations for surge forecasting
• ConvLSTM networks for spatiotemporal forecasting
• Integrates wind fields and topo-bathymetry as inputs