An Integrated Convolutional Neural Networks and Light Gradient Boosting Approach for Flood Classification Using Sentinel-1 SAR Satellite Imagery

od classification plays a crucial role in disaster mitigation, particularly in areas frequently affected by floods. This study proposes a novel model combining Convolutional Neural Networks (CNN) using ResNet-50 and Light Gradient Boosting Machine (LightGBM) for classifying flood and non-flood areas using Sentinel-1 SAR imagery. The dataset used consists of 21,016 images, evenly distributed between flood and non-flood classes, and processed through resizing, normalization, denoising, and augmentation. Feature extraction was conducted using the ResNet-50 architecture, which captured spatial and textural patterns efficiently, followed by LightGBM for classification. The proposed model achieved a high accuracy of 96%, with Precision, Recall, and F1-scores exceeding 95% for both classes. The evaluation metrics, including Precision-Recall Curve with an AUC of 0.9852 and a Confusion Matrix, confirmed the model's robustness and balance in classifying both categories. Additionally, comparisons with previous research, such as SAR-FloodNet, demonstrated the superiority of the proposed approach, achieving a 2% improvement in accuracy. Despite these results, limitations such as the exclusive use of Sentinel-1 data and the lack of validation across diverse environmental conditions remain. Future research should explore integrating multispectral Sentinel-2 data and testing on broader datasets to enhance scalability and reliability. The findings underscore the model's potential for real-world applications in flood monitoring and disaster management systems.