Deep convolutional neural network with random field model for lake ice mapping from Sentinel-1 imagery

Sea ice is critical to map for safe and efficient maritime navigation, to mitigate ship trapping and capsizing. Sea ice is also vital to monitor to assess the state of the changing climate and a critical component in climate and weather models, reflecting sunlight towards space and acting as an insulating material between the ocean surface and the atmosphere.Professional sea ice analysts at national ice services map sea ice based on Synthetic Aperture Radar (SAR) images acquired by satellites, such as the Sentinel-1 (S1) satellite constellation. The ice analysts manually interpret the SAR images using their in-depth knowledge and experience to create sea ice charts with information on the sea ice conditions.A challenge for the S1 5.4 GHz SAR measurements is that the radar wave does not penetrate deep into the sea ice and is scattered/reflected by the surface. Therefore, the SAR images provide information primarily about the sea ice surface, useful for identifying and classify sea ice conditions. The charts describe, among others, the sea ice’s stage of development - the type of sea ice - an indicator of its thickness. The manual charting process apply sea ice classes, defined by the International Ice Charting Working Group (IICWG) on behalf of the World Meteorological Organization (WMO). Considerable uncertainties are associated with the ice classes that can vary from, e.g. 30-200cm or 70-120cm in thickness. Deep-learning models that produce stage-of-development information from S1 radar images exist but has the same inherent limitations of the sea ice charts in the model outputs.Current state-of-the-art sea ice thickness retrieval methods relies on altimeter satellites, such as the CryoSat-2 (CS2) satellite. The distance between the ocean and the sea ice is measured, known as the sea ice freeboard. For a Ku-band radar altimeter like CS2, it is assumed that the radar response penetrates the snow and returns from the sea ice surface. As the true penetration is unknown, and the radar wave propagation is delayed when the signal passes through snow, the measured quantity is known as the radar freeboard.The sea ice thickness can be estimated with an accuracy of 20-40% using the radar freeboard by calculating the sea ice's buoyancy based on snow and ice density estimates, and auxiliary snow depth information. However, CS2 only measures 1600m across the orbit and can thus only monitor sea ice thickness in the Arctic monthly - insufficient for many applications, such as maritime navigation, and leaves data record gaps. S1 SAR on the other hand, cover 400km in Extra Wide mode across the orbit with repeating coverage every week.Here, we present our preliminary results of circumventing the limitations of CS2 and S1 by training supervised deep-learning convolutional neural network (CNN) models to recognise sea ice textures in S1 SAR images and assign sea ice radar freeboard estimates acquired by CS2. This approach transfers information acquired by CS2 to S1, which we call Cryo2S1. A Cryo2S1 dataset is curated, containing several thousand collocated S1 SAR images and along-track CS2 measurements during 2020-2021.

Effective urban planning and management rely on accurate land cover mapping, which can be achieved through the combination of remote sensing data and machine learning algorithms. This study aimed to explore and demonstrate the potential benefits of integrating Sentinel-1 SAR and Sentinel-2 MSI satellite imagery for urban land cover classification in Gondar city, Ethiopia. Synthetic Aperture Radar (SAR) data from Sentinel-1A and Multispectral Instrument (MSI) data from Sentinel-2B for the year 2023 were utilized for this research work. Support Vector Machine (SVM) and Random Forest (RF) machine learning algorithms were utilized for the classification process. Google Earth Engine (GEE) was used for the processing, classification, and validation of the remote sensing data. The findings of the research provided valuable insights into the performance evaluation of the Support Vector Machine (SVM) and Random Forest (RF) algorithms for image classification using different datasets, namely Sentinel 2B Multispectral Instrument (MSI) and Sentinel 1A Synthetic Aperture Radar (SAR) data. When applied to the Sentinel 2B MSI dataset, both SVM and RF achieved an overall accuracy (OA) of 0.69, with a moderate level of agreement indicated by the Kappa score of 0.357. For the Sentinel 1A SAR data, SVM maintained the same OA of 0.69 but showed an improved Kappa score of 0.67, indicating its suitability for SAR image classification. In contrast, RF achieved a slightly lower OA of 0.66 with Sentinel 1A SAR data. However, when the datasets of Sentinel 2B MSI and Sentinel 1A SAR were combined, SVM achieved an impressive OA of 0.91 with a high Kappa score of 0.80, while RF achieved an OA of 0.81 with a Kappa score of 0.809. These findings highlight the potential of fusing satellite data from multiple sources to enhance the accuracy and effectiveness of image classification algorithms, making them valuable tools for various applications, including land use mapping and environmental monitoring.

Boreal lakes ecosystems can remain partially or completely covered by ice and snow during an important portion of the year. Alterations of lake and river ice properties can deteriorate the conditions of local ice roads, negatively influencing Nordic communities and economical activities. Monitoring of lake ice characteristics and dynamics is therefore crucial. In this study, Sentinel-1 Synthetic Aperture Radar (SAR) satellite imagery is exploited to evaluate changes in the ice regime over shallow and deep high-latitude lakes during the winters of 2018 and 2019. The methodology proposed, based on the combined analysis of SAR intensity and interferometric coherence maps, enables the extraction of important characteristics of ice dynamics. Overall, the lake ice thickness change derived from Differential Interferometric SAR (D-InSAR) increases with the lake depth. The D-InSAR-derived mean rate of growth, in general, agrees with the one estimated from records of in-situ ice thickness measurements. The methodology presented herein could be temporally extended to support the understanding of historical and current climate conditions. This could be done by using archived and newly available imagery to improve lake ice models.

Mapping water bodies with a high accuracy is necessary for water resource assessment, and mapping them rapidly is necessary for flood monitoring. Poyang Lake is the largest freshwater lake in China, and its wetland is one of the most important in the world. Poyang Lake is affected by floods from the Yangtze River basin every year, and the fluctuation of the water area and water level directly or indirectly affects the ecological environment of Poyang Lake. Synthetic Aperture Radar (SAR) is particularly suitable for large-scale water body mapping, as SAR allows data acquisition regardless of illumination and weather conditions. The two-satellite Sentinel-1 constellation, providing C-Band SAR data, passes over the Poyang Lake about five times a month. With its high temporal-spatial resolution, the Sentinel-1 SAR data can be used to accurately monitor the water body. After acquiring all the Sentinel-1 (1A and 1B) SAR data, to ensure the consistency of data processing, we propose the use of a Python and SeNtinel Application Platform (SNAP)-based engine (SARProcMod) to process the data and construct a Poyang Lake Sentinel-1 SAR dataset with a 10 m resolution. To extract water body information from Sentinel-1 SAR data, we propose an automatic classification engine based on a modified U-Net convolutional neural network (WaterUNet), which classifies all data using artificial sample datasets with a high validation accuracy. The results show that the maximum and minimum water areas in our study area were 2714.08 km2 on 20 July 2020, and 634.44 km2 on 4 January 2020. Compared to the water level data from the Poyang gauging station, the water area was highly correlated with the water level, with the correlation coefficient being up to 0.92 and the R2 from quadratic polynomial fitting up to 0.88; thus, the resulting relationship results can be used to estimate the water area or water level of Poyang Lake. According to the results, we can conclude that Sentinel-1 SAR and WaterUNet are very suitable for water body monitoring as well as emergency flood mapping.

Landfast ice polynyas are important features near many northern coastal communities, and their automated detection from synthetic aperture radar (SAR) imagery is positioned to support on-ice travel safety under changing Arctic sea ice conditions. This research leveraged original datasets of over 5,000 Sentinel-1 SAR observations of wintertime polynyas mapped near the Canadian communities of Sanikiluaq and Nain to investigate deep learning-based landfast ice polynya detection. The Faster-RCNN object detection network was optimized for polynya detection through several modifications to network design elements and training strategies. Resulting detection models generalized well between regions and accurately detected polynyas with local backscatter contrasts above 5 dB, e.g. achieving 90% target recall at 24% precision. Polynyas smaller than 500 meters and with local backscatter contrasts less than 3 dB, constituting approximately half of all observations, were frequently missed. Precision scores below 30% were consistently incurred in attempts to optimize recall. Results highlight challenges to the consistent performance of single-image detectors due to variable and frequently weak polynya signatures in dual-polarized Sentinel-1 SAR backscatter. Future investigations into multi-temporal, multi-frequency, and/or higher-resolution SAR imagery could further support the delivery of robust hazard detection systems relevant to community sea ice safety and monitoring.

Capturing and identifying field-based agricultural activities, such as the start, duration and end of irrigation, together with crop sowing/germination, growing period and time of harvest, offer informative metrics that can assist in precision agricultural activities in addition to broader water and food security monitoring efforts. While optically based band-ratios, such as the normalized difference vegetation index (NDVI) and normalized difference water index (NDWI), have been used as descriptors for monitoring crop dynamics, data are not always available due to the influence of clouds and other atmospheric effects on optical sensors. Satellite-based microwave systems, such as the synthetic aperture radar (SAR), offer an all-weather advantage in monitoring soil and crop conditions. In this paper, we leverage the relative strengths of both optical- and microwave-based approaches by combining high resolution Sentinel-1 SAR and Sentinel-2 optical imagery to monitor irrigation events and crop dynamics in a dryland agricultural landscape. A microwave backscatter model was used to analyze the responses of simulated backscatters to soil moisture, NDVI and NDWI (both are correlated with vegetation water content and can be regarded as vegetation descriptors), allowing an empirical relationship between these two platforms. A correlation analysis was also performed using Sentinel-1 SAR and Sentinel-2 optical data over crops of maize, alfalfa, carrot and Rhodes grass in Al Kharj farm of Saudi Arabia to identify an appropriate SAR-based vegetation descriptor. The results illustrate the relationship between SAR and both NDVI and NDWI and demonstrated the relationship between the cross-polarization ratio (VH/VV) and the two optical indices. We explore the capacity of this multi-platform and multi-sensor approach to inform on the spatio-temporal dynamics of a range of agricultural activities, which can be used to facilitate field-based management decisions.

Target Recognition Using Pre-Trained Convolutional Neural Networks and Transfer Learning

Reliable and accurate land use/land cover (LULC) map is a crucial data source for the understanding of coupled human–environment systems, monitoring changes, timely low-cost planning, and management of natural resources. Improvements in sensor technologies and machine learning capabilities have shifted the attention of remote sensing community to data complementarity through fusion of multi-sensor data for accurate feature extraction and mapping. Amalgamation of optical and synthetic aperture radar (SAR) images has shown promising advantages in enhancing the accuracy of extracting LULC as such method allows exploitation of information in sensors. This study investigated the potential of using freely available multisource Sentinel images to extract LULC maps in semi-arid environments through multi-scale geographic object-based image analysis (GEOBIA). A multi-scale classification framework that integrates GEOBIA, correlation-based feature selection (CFS), and random forest (RF)-supervised classification was adopted to extract LULC from assimilation of Sentinel multi-sensor products. First, Sentinel-1 and -2 images were pre-processed. Second, optimum multi-scale segmentation levels were selected using F-score segmentation quality measures. Third, 70 features of various spectral indices and derivatives and geometrical features from optical data and multiple ratios and textural features from dual-polarization SAR images were computed, and a CFS based on wrapper approach was used to select the most significant features at multi-scale levels. Finally, a single and multi-scale RF classifier was used to extract LULC classes using the most relevant features extracted from Sentinel SAR and optical images. Results of multi-scale image segmentation optimization showed that scale parameter (SP) values of 40, 60, and 150 were optimal for extraction of LULC classes. Results of feature selection showed that 22, 24, and 27 features were selected at scale SP values of 40, 60, and 150, respectively. Half of the features were common among the three scales. Single RF classification yielded overall accuracy (OA) values of 92.10%, 93%, and 91% and kappa coefficients of 0.901, 0.912, and 0.89 at scale values of 150, 60, and 40, respectively. Multiscale RF classification from scale values of 150 and 60 produced better LULC classification with OA 96.06% and kappa coefficient of 0.95 compared with other scale SP values. The integrated approach demonstrated an effective and promising method for high-quality LULC extraction from coupling optical and SAR images. Overall, multi-sensor Sentinel images along with the adopted approach feature a remarkable potential for improving LULC extraction and can effectively be used to update geographic information system layers for various applications.

Oil spills on the ocean from ships have a strong impact on the environment as well as on human societies. The MV X-Press Pearl container ship incident in 2021 became the worst marine environmental disaster in Sri Lanka due to its powerful impact. Oil leakage from the ship to the ocean continued for months. Synthetic Aperture Radar (SAR) imagery has the advantage of penetrating cloud cover, as well as day and night operating capabilities. Therefore, SAR imagery can be used to detect as well as observe the temporal changes of oil spills. The study in this article aimed to detect the oil spill caused by the MV X-Press Pearl container ship on Sri Lanka’s west coast from May to August 2021 using Sentinel-1 SAR data and measure the temporal changes of the oil slick. Three SAR imageries with Vertical-Vertical (VV) and Vertical-Horizontal (VH) polarization from June, July, and August of 2021 have been processed by applying the orbit file, running the multi-looking algorithm, and making the ellipsoid correction. The possible oil spills have been identified from each set of imagery as dark linear shapes with lengths of 3.1 ±0.012 km with an average reflectance of 15.6 decibels for June and 3.6 ±0.012 km with an average reflectance of 16.5 decibels for the other two months, and the shapes correspond with the oil slick classifications from other literature. The results indicate that the VV polarization is the most suitable to detect oil spills, and that the oil spill from the X-Press Pearl continued from June to the end of August 2021.

The classification of objects and man-made structures on the ocean surface using synthetic aperture radar (SAR) imagery finds an important use in monitoring for icebergs and sea ice. Convolutional Neural Network (CNN) method can be employed to effectively classify objects imaged through SAR data. In this paper, the CNN performance is evaluated when three different preprocessing procedures are applied to the SAR image data to prepare the CNN's input data. Segmentation or normalization algorithms are implemented in the three procedures. Experimental results demonstrate an improvement over unsegmented and un-normalized images. Performance metrics for all three methods are approximately 94%, indicating that while some image preprocessing is required to achieve higher performance, the CNN tested is robust to the noise present in the SAR images used.

Many research efforts have been devoted to applying machine learning (ML) algorithms to the task of Automatic Target Recognition (ATR). In the 90’s, ML techniques such as Neural Networks were less popular due to various technological barriers and applications. Computational resources were scarce and expensive. Today, computational resources are not as expensive as in the past; however, an abundance of sensors and business data need to be analyzed in real-time. High performance computing (HPC) enables ML-based decision making in real-time or near real-time. This research explores the application of deep learning algorithms, specifically convolutional neural networks, to the task of ATR in synthetic aperture radar (SAR) imagery. We developed a Convolution Neural Networks (CNN) architecture for achieving ATR in SAR imagery and found that classification accuracy levels of 99% can be achieved through the application of neural networks. We used graphics processing units (GPU) to accomplish the computational tasks.

In this article, a two-year time-series of multipolarization Sentinel-1 synthetic aperture radar (SAR) imagery is exploited to analyze the changes in the water-covered area of the Monte Cotugno (Italy) reservoir. A two-step processing chain, which includes water/land separation and waterline extraction, is proposed to accomplish this task. Experimental results, verified using independent in situ measurements, demonstrate that: first, Sentinel-1 time series can be successfully used to support the smart water management of reservoirs. In fact, the changes in the water-covered area inferred from the SAR time series agree with the seasonal behavior and they also fit anomalies; and second, multipolarization feature outperforms single-polarization ones in terms of accuracy of the extracted waterline profile.

Speckle noise inherent to synthetic aperture radar (SAR) imagery degrades image quality and complicates automated analysis in Earth observation applications. Quantitative assessment of despeckling results requires computing quality metrics against reference images, which are unavailable in operational SAR scenarios. This paper presents a method for a priori prediction of filtered Sentinel-1 SAR image quality metrics before applying speckle noise filters. Unlike existing approaches predicting relative quality improvement, the proposed method predicts absolute values of five metrics (PSNR, WSNR, SSIM, MS-SSIM, FSIM) for a specific filter, enabling direct comparison and rational filter selection. The methodology employs transfer learning of DenseNet-121 convolutional neural network, pre-trained on ImageNet, adapted for single-channel SAR inputs through architectural modifications including input layer transformation, pooling optimization, and regression head replacement. A novel synthetic data generation pipeline utilizes histogram matching of Sentinel-2 optical images with Sentinel-1 SAR references to create training samples preserving ground truth. Dynamic gamma-distributed speckle noise addition with variable ENL ∈ [2, 6] enhances data variability and model robustness. Experiments with six classical filters (Gamma MAP, Lee, Enhanced Lee, Frost, SRAD, Kuan) demonstrate high prediction accuracy across all filter-metric combinations. The coefficient of determination R² reaches 0.997 for best combinations and exceeds 0.97 for most of the 30 trained models. Mean absolute prediction errors remain below 0.29 dB for PSNR and 0.014 for SSIM across all tested configurations. The approach enables a priori quality prediction without reference images, allowing optimization of SAR processing workflows and resource planning before resource-intensive despeckling. Contributions of Authors: Conceptualization, Raed A. and Oleksii R.; methodology, Raed A.; formal analysis, Raed A.; investigation, Raed A. and Oleksii R.; data curation, Raed A. and Oleksii R.; writing—original draft preparation, Raed A.; writing—review and editing, Raed A. and Oleksii R.; visualization, Raed A. All authors have read and agreed to the published version of the manuscript. Funding: This research received no external funding. Data Availability Statement: The processed dataset is available in the Zenodo repository (https://zenodo.org/uploads/17253925). Python implementation code is available in the GitHub repository (https://github.com/rsenaikh/Predicting_Quality_after_Noise_Removal). Acknowledgments: The authors would like to express their sincere gratitude to the Copernicus Data Space Ecosystem for providing open access to Sentinel-1 and Sentinel-2 data. We are also grateful to reviewers and editors for their valuable comments, recommendations, and attention to the work. Conflicts of Interest: The authors declare no conflict of interest.

Robust unsupervised small area change detection from SAR imagery using deep learning

In this paper, we assess the flood mapping capabilities of the X-band Synthetic Aperture Radar (SAR) imagery acquired by the bistatic pair TanDEM-X/TerraSAR-X (TDX/TSX). The main objective is to investigate the added value of the bistatic TDX/TSX Interferometric Synthetic Aperture Radar (InSAR) coherence in addition to the SAR backscatter in the context of inundation mapping. As a classifier, we consider a Random Forest (RF) classification scheme using TDX/TSX SAR intensities and their bistatic InSAR coherence to extract the flood extent map. To evaluate the classification results and as no “ground truth” was available at the SAR data acquisition time, we set up a LISFLOOD-FP hydraulic model for simulating the temporal evolution of the flood water. The flood map simulated by the model shows good performances with an Overall Accuracy (OA) of 97.92 % and a Critical Success Index (CSI) of 94 . 01 % . The SAR-derived flood map is then compared to the LISFLOOD-FP extent map simulated at the SAR data acquisition time. As a test case, we consider the flooding event of the Richelieu River that occurred in the Montérégie region of Quebec (Canada) from April to June 2011. Experimental results highlight the potential of the bistatic InSAR coherence for more accurate flood mapping in a complex landscape with urban and vegetation areas. The classification results of the SAR-derived flood map with respect to the LISFLOOD-FP flood map reach an OA of 78.65 % and a Precision of 82.08 % when integrating the bistatic InSAR coherence. These classification OA and Precision values are 69.63 % and 64.52 % , respectively, using only the TDX/TSX SAR intensity.