FLOOD MAPPING IN PHRA NAKHON SI AYUTTHAYA, THAILAND, UTILIZING SENTINEL-1 SAR IMAGERY AND DEEP LEARNING APPROACHES

This study presents a deep learning (DL)-based flood detection framework for Phra Nakhon Si Ayutthaya Province, Thailand.The framework integrates U-Net architecture with three encoder variants-ResNet50, ResNet101, and ResNet152-using synthetic aperture radar (SAR) imagery from the Sentinel-1 satellite.Model performance is assessed through statistical metrics including accuracy, precision, recall, F1-score, Dice Loss, intersection over union (IoU), and computational time.The U-Net model with a ResNet101 encoder achieved the best performance, with an accuracy of 91.5%, F1score of 0.886, Dice Loss of 0.116, and IoU of 86.8%, requiring about 24 minutes of training.Despite longer training, the ResNet101-based U-Net substantially enhances flood detection accuracy, highlighting its value as a reliable tool for real-time monitoring and rapid response in flood-prone areas of Thailand.