A DATA-DRIVEN FRAMEWORK FOR SAR WAVE PARAMETER ESTIMATION DURING TROPICAL CYCLONES

Abstract

Tropical cyclones pose significant challenges for ocean wave monitoring due to their complex interactions between high wind speeds, heavy precipitation, and turbulent sea states. Traditional physics-based algorithms often struggle to accurately retrieve wave parameters such as significant wave height, wave period, and wave direction from Synthetic Aperture Radar (SAR) imagery under such extreme conditions. This study proposes a machine learning-based approach for reliable and accurate retrieval of wave parameters from SAR images during tropical cyclones. The algorithm leverages convolutional neural networks (CNNs) for automatic feature extraction from SAR imagery and integrates auxiliary meteorological data such as wind speed and cyclone intensity. By training the model on historical cyclone datasets with collocated buoy measurements and numerical wave model outputs, the proposed method demonstrates improved performance over conventional retrieval techniques.