Oil spills threaten marine ecosystems, demanding swift detection and response. The northern entrance of the Suez Canal, a critical maritime route, is increasingly at risk of frequent oil spill incidents. This study employs the DeepLabv3 + deep learning model to automatically detect oil spills in the study area based on Sentinel-1 Synthetic Aperture Radar imagery provided by the European Space Agency. The model was trained separately on two datasets: the European Maritime Safety Agency CleanSeaNet (EMSA-CSN) dataset, comprising 1100 oil spill incidents, and a localized dataset containing 1500 oil spill incidents that occurred at the Egyptian territorial waters. A comparative analysis between the two models was conducted using 30 oil spill test cases located within the study area. The model trained on Egyptian data outperformed the EMSA-CSN-data- trained model, achieving a loss of 0.0516, an accuracy of 98.14%, a mean Intersection over Union (MIoU) of 0.7872, and a significantly higher ROC area of 0.91, compared to a loss of 0.1152, an accuracy of 96.45%, a MIoU of 0.7161, and a ROC area of 0.76 for the EMSA-CSN model. In addition, the area prediction analysis confirmed the superior performance of the Egyptian-data-trained model, which estimated a total affected area of 421.20 km2, closely aligning with the ground truth of 425.20 km2, whereas the EMSA-CSN-data-trained model underestimated oil spills of around 323.98 km2. These results highlight the benefits of region-specific training in improving segmentation quality and reducing errors. This study emphasizes the potential of AI-driven models for real-time oil spill monitoring, with applications in environmental protection and emergency response.
The northern entrance of the Suez Canal faces significant environmental and economic challenges due to its strategic location as a major maritime chokepoint with over 20,000 vessels passing through annually. Oil spills in this region can arise from tanker accidents, operational discharges, or vessel leaks, endangering marine ecosystems, coastal habitats, and local industries such as fisheries and tourism. These spills adversely affect marine life, coating organisms with a toxic layer that impairs their ability to breathe or feed. Spills may also contaminate water resources, posing risks to human health through exposure to toxic chemicals. Furthermore, oil spills near the Suez Canal can disrupt global shipping routes, leading to economic losses for both regional and international trade. Cleanup efforts are complicated by the area’s high traffic and sensitive ecological zones, highlighting the need for advanced detection technologies and coordinated response strategies to mitigate environmental damage. Real-time detection and accurate delineation of oil spill extent are critical for effective response and mitigation efforts.
Satellites offer an efficient and cost-effective method for daily scanning of large sea areas to detect potential oil pollution. Multiple remote sensing technologies have been examined for oil spill identification, each offering distinct advantages. Optical remote captures visible and near-infrared light reflected from the water surface, allowing for the identification of thin oil layers under clear conditions, but is hindered by cloud cover and atmospheric interference. Thermal infrared imaging detects temperature differences between oil and water, which is useful for identifying thicker oil spills, especially during cooler times or at night. Laser fluorescence uses laser-induced light to excite oil molecules, offering high specificity for oil detection, but is typically limited to smaller-scale or localized applications. Multi-angle Imaging SpectroRadiometer (MISR), with its ability to capture multi-angle images, helps in studying the surface roughness and composition of oil spills, providing valuable insights into their size and characteristics. Additionally, hyperspectral imaging offers a detailed spectral profile that enables precise characterization of oil spills based on their unique spectral signatures, such as those observed along the Nile River.
Operating in the microwave range, Synthetic Aperture Radar (SAR) is particularly effective for detecting oil spills regardless of weather or illumination conditions, as it captures variations in surface roughness caused by oil, making it ideal for large-scale and all-weather surveillance. Although SAR has limitations in precisely quantifying spill thickness, its reliability, wide-area coverage, and adaptability to adverse conditions make it the preferred choice for comprehensive and timely oil spill detection and mapping in this study. Its all-weather and day-night capabilities are indispensable in marine environments, where cloud cover and darkness frequently pose challenges.
Detecting oil spills in SAR images is based on its detection mechanism, which leverages the suppression of capillary and short-gravity capillary waves caused by oil spills. This suppression reduces the surface roughness of seawater, appearing as dark spots in SAR images, providing a distinct contrast to the surrounding ocean and enabling effective oil spill detection and monitoring. Effective detection typically requires a minimum wind speed of 2-3 m/s to ensure the oil film is visible; however, high wind speeds larger than 10 m/s can obscure the spill.
However, SAR-based methods face challenges in distinguishing oil spills from other phenomena that also produce dark spots in SAR imagery, such as wave shadows, algal blooms, and low-wind-speed regions behind land masses. These false positives present a significant obstacle to reliable oil spill detection. To overcome this limitation, multi-polarization SAR data offers a solution. By transmitting and receiving signals with varying polarimetric properties, multi-polarization SAR enhances the richness of scattering information, significantly improving the accuracy and reliability of oil spill detection. Consequently, identifying an oil spill in SAR images begins with detecting dark features, which can also include non-oil spill dark signatures resulting from meteorological or oceanographic conditions. The identification procedure typically involves isolating and contouring dark signatures through thresholding, extracting the key parameters related to their shape and radar backscattering contrast. Previous studies on oil spill detection near the entrance of the Suez Canal have primarily employed manual interpretation techniques, as summarized in Table 1.
This manual interpretation of SAR imagery is labour-intensive and susceptible to human error, underscoring the need for automated solutions. Marghany introduced the Quantum Immune Fast Spectral Clustering algorithm for the automatic detection of oil spills in quad-polarized RADARSAT-2 SAR data, leveraging quantum computing principles for feature clustering. This method enhances classification by optimizing feature clustering at a subatomic level, making it highly efficient in detecting subtle differences in oil spill characteristics. However, it relies on quad-polarized RADARSAT-2 data, which is not freely available.
Conversely, deep learning utilizes artificial neural networks to automatically learn patterns and features from large datasets, making it highly effective for complex image analysis tasks like oil spill detection in SAR imagery. By leveraging freely available Sentinel-1 data, the present study provides an accessible and cost-effective solution that ensures efficient and continuous monitoring without the constraints of proprietary data. This approach presents an advancement by utilizing deep learning for automatic oil spill detection in the region.
Numerous Deep learning models have emerged in recent years to address the challenge of oil spill detection from SAR imagery. Krestenitis, M. et al. discussed the use of deep convolutional neural networks (DCNNs) for efficient oil spill detection in SAR images, proposing a new publicly available dataset and demonstrating that DCNNs, particularly DeepLabv3 + , outperform other methods in accuracy and speed. Shaban et al. presented a two-stage deep-learning framework for detecting oil spills in SAR images, using a 23-layer CNN for classification and a five-stage U-Net for segmentation, achieving improved precision and dice scores. Singha et al. developed a fully automated approach for detecting oil spills using SAR images, combining classification tree analysis and fuzzy logic, with an accuracy of 85-93% compared to human analysts. Keramitsoglou et al. built a fully automated AI-based system for identifying oil spills in SAR images, which was developed and successfully tested on 35 images from the Aegean Sea, providing detailed outputs for decision-making.
Building on this body of research, the present study contributes to this theme by applying semantic segmentation on SAR images for oil spill detection at the northern entrance of the Suez Canal. Additionally, a new dataset was introduced for the scientific community comprising 1500 documented oil spill cases, enabling the model to learn from more localized data and improving its adaptability to regional environmental conditions. This approach enables real-time or near-real-time identification of oil spills, significantly reducing the time and resources needed compared to current manual monitoring methods. Additionally, this framework supports the Egyptian Environmental Affairs Agency (EEAA), which plays a crucial role in reporting and implementing oil spill response plans as part of Egypt’s national contingency strategy for combating marine pollution. This research also aligns with Egypt’s obligations under international conventions, such as MARPOL 73/78 and the Oil Pollution Preparedness, Response, and Co-operation (OPRC ’90) convention, which require effective measures for environmental monitoring and response efforts.
