The detection of flooded areas using high-resolution synthetic aperture radar (SAR) imagery is a critical task with applications in crisis and disaster management, as well as environmental resource planning. However, the complex nature of SAR images presents a challenge that often leads to an overestimation of the flood extent. To address this issue, we propose a novel differential attention metric-based network (DAM-Net) in this study. The DAM-Net comprises two key components: a weight-sharing Siamese backbone to obtain multi-scale change features of multi-temporal images and tokens containing high-level semantic information of water-body changes, and a temporal differential fusion (TDF) module that integrates semantic tokens and change features to generate flood maps with reduced speckle noise. Specifically, the backbone is split into multiple stages. In each stage, we design three modules, namely, temporal-wise feature extraction (TWFE), cross-temporal change attention (CTCA), and temporal-aware change enhancement (TACE), to effectively extract the change features. In TACE of the last stage, we introduce a class token to record high-level semantic information of water-body changes via the attention mechanism. Another challenge faced by data-driven deep learning algorithms is the limited availability of flood detection datasets. To overcome this, we have created the S1GFloods open-source dataset, a global-scale high-resolution Sentinel-1 SAR image pairs dataset covering 46 global flood events between 2015 and 2022. The experiments on the S1GFloods dataset using the proposed DAM-Net showed top results compared to state-of-the-art methods in terms of overall accuracy, F1-score, and IoU, which reached 97.8%, 96.5%, and 93.2%, respectively. Our dataset and code will be available online at https://github.com/Tamer-Saleh/S1GFlood-Detection.
12 July 2025 in ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Urban flooding is a recurring and distressing issue with severe consequences, including the destruction of densely populated infrastructure and loss of life. Mapping inundated urban areas using synthetic aperture radar (SAR) data is crucial for local authorities to quickly assess risks and coordinate rescue efforts. However, due to the complexity of backscattering mechanisms, SAR-based urban floodwater mapping remains a challenge. In this work, we address this problem by introducing a novel algorithm, coherence-guided change transformer (CC-Former), for urban flood mapping that leverages the coherence of interferometric SAR (InSAR) with vision transformers. Specifically, CC-Former utilizes two Siamese weight-sharing encoders to extract multi-scale features from input InSAR coherence images and employs a decoder to generate final predictions. Additionally, we propose a coherence-based scaling (CoBS) module designed to focus on the acquired coherence features of urban flood classes and mitigate the imbalanced distribution of training classes. For qualitative and quantitative evaluation, the proposed CC-Former model was trained and validated using multi-temporal, dual-polarized Sentinel-1 SAR data to map the flood extent in Derna, Libya, following Tropical Storm Daniel in September 2023. Experimental results demonstrate that the proposed model outperforms state-of-the-art methods, achieving an F1 score of 89.4% and an IoU of 84.4% in both co- and cross-polarization, and an F1 score of 87.9% when integrating intensity and coherence. We conclude that the CC-Former model offers a promising solution for accurate and efficient urban flood mapping from InSAR coherence, with the potential for rapid generalization to other affected areas. As such, it can significantly aid disaster management efforts in vulnerable communities in near real-time.