The subject matter of the article is the process of lossy compression of multilook Synthetic Aperture Radar (SAR) images corrupted by multiplicative, spatially correlated speckle noise, with a focus on operation in the neighborhood of the potential Optimal Operation Point (OOP). The goal of the article is to analyze the existence and properties of the OOP for SAR image compression using the Better Portable Graphics (BPG) coder, and to develop a practical method for achieving compression near this point. The tasks to be solved are: to verify the existence of the OOP for simulated Sentinel-1-like SAR images according to both traditional peak signal-to-noise ratio (PSNR) and visual quality (PSNR-HVS-M) metrics; to investigate the relationship between the compression control parameter (Q) and the resulting image quality and compression ratio (CR); and to propose and describe a practical iterative procedure for determining the Q parameter value corresponding to the OOP without requiring access to the noise-free reference image. The methods used are: simulation of SAR images with speckle relative variance equal to 0.05 using noise-free Sentinel-2 data as a reference; lossy compression using the BPG coder with parameter Q varying from 1 to 51; quantitative assessment using PSNR and PSNR-HVS-M metrics; calculation of compression ratio; analysis of rate-distortion curves between different image pairs; statistical estimation of equivalent noise variance for input PSNR prediction. The following results were obtained: It has been demonstrated that an OOP exists for the BPG coder when compressing multilook SAR images, confirmed by both PSNR and PSNR-HVS-M metrics. The OOP provides PSNR and PSNR-HVS-M values several dB higher compared to the uncompressed noisy image while achieving very high compression ratios (CR > 180). The OOP was found at high Q values (Q=48-49), where the coder aggressively suppresses noise but also introduces content distortions. A key practical result is the proposed method for determining Q at the OOP. Conclusions. The scientific novelty of the obtained results is as follows: For the first time, the existence of the OOP has been comprehensively demonstrated for the BPG coder applied to multilook SAR images with realistic speckle properties, considering not only the standard PSNR but also the visual quality metric PSNR-HVS-M, although the OOP is less pronounced for the latter; a method for practical OOP approximation has been developed, which operates without the need for the original noise-free (true) image, relying instead on an estimation of the speckle noise power from the available noisy data, making it applicable in real-world SAR image processing and transmission scenarios.

High-precision and efficient building extraction by fusing visible and synthetic aperture radar (SAR) imagery is critical for applications such as smart cities, disaster response, and UAV navigation. However, existing approaches often rely on complex multimodal feature extraction and deep fusion mechanisms, resulting in over-parameterized models and excessive computation, which makes it challenging to balance accuracy and efficiency. To address this issue, we propose a dual-branch lightweight architecture, DLiteNet, which functionally decouples the multimodal building extraction task into two sub-tasks: global context modeling and spatial detail capturing. Accordingly, we design a lightweight context branch and spatial branch to achieve an optimal trade-off between semantic accuracy and computational efficiency. The context branch jointly processes visible and SAR images, leveraging our proposed Multi-scale Context Attention Module (MCAM) to adaptively fuse multimodal contextual information, followed by a lightweight Short-Term Dense Atrous Concatenate (STDAC) module for extracting high-level semantics. The spatial branch focuses on capturing textures and edge structures from visible imagery and employs a Context-Detail Aggregation Module (CDAM) to fuse contextual priors and refine building contours. Experiments on the MSAW and DFC23 Track2 datasets demonstrate that DLiteNet achieves strong performance with only 5.6 M parameters and extremely low computational costs (51.7/5.8 GFLOPs), significantly outperforming state-of-the-art models such as CMGFNet (85.2 M, 490.9/150.3 GFLOPs) and MCANet (71.2 M, 874.5/375.9 GFLOPs). On the MSAW dataset, DLiteNet achieves the highest accuracy (83.6% IoU, 91.1% F1-score), exceeding the best MCANet baseline by 1.0% IoU and 0.6% F1-score. Furthermore, deployment tests on the Jetson Orin NX edge device show that DLiteNet achieves a low inference latency of 14.97 ms per frame under FP32 precision, highlighting its real-time capability and deployment potential in edge computing scenarios.

When fusing inverse synthetic aperture radar (ISAR) images and high-resolution range profile (HRRP), the significant heterogeneity existing between the feature spaces of the two is not adequately considered, resulting in a low accuracy rate of space target recognition. A multi-modal fusion algorithm based on spatial attention and multi-scale temporal network is proposed in this paper. We carefully consider the data characteristics of HRRP and ISAR and design feature extraction networks, respectively. For HRRP, the local invariant features are extracted using dynamic convolution (DyConv), and the convolution depth is reduced. An improved multi-scale temporal convolution network is designed based on the temporal characteristics of HRRP to extract temporal features for target recognition. For ISAR images, an omnidirectional attention feature extraction module is designed to extract the deep semantic features of the images, and a noise reduction module with a spatial attention mechanism is designed before extracting the image features to reduce the background noise in the fused image. The superiority of the designed ISAR recognition network and HRRP recognition network for space target was verified through comparative and ablation experiments. The recognition rate for the target of the proposed algorithm is 98.41%.

Abstract. Publicly available datasets for oil spill detection are scarce, making it difficult to compare the performance of different detection algorithms. To address this, this paper introduces a comprehensive labeled dataset of oil slicks, look-alikes, and other remarkable oceanic phenomena, derived from Sentinel-1 Synthetic Aperture Radar (SAR) products in the Eastern Mediterranean Sea in 2019. The dataset contains 3225 oil objects across 1365 image patches, along with an additional 2290 image patches featuring look-alikes or other phenomena. Data are available at https://doi.org/10.1594/PANGAEA.980773 (Yang and Singha, 2025). This dataset enables researchers to evaluate their oil spill detection models and compare performance with other studies. To facilitate this, the performance of an oil spill detector from a previous study on the dataset is provided as a baseline. In addition, to help the researchers better understand what phenomena their object detector might be confusing with oil slicks, the image patches without oil objects were sorted into several subgroups. On the other hand, for researchers looking to apply object detection models to oil slick detection but lacking a starting dataset, this dataset can serve as a valuable training resource. Beyond dataset presentation, this paper also explains the formation of different oceanic phenomena and their SAR signatures, supported by examples and supplementary materials. These insights help researchers from various backgrounds, such as remote sensing, oceanography, and machine learning, better understand the sources of SAR signatures.

Vegetation vertical structure refers to the 3D distribution of vegetation aboveground biomass. Vegetation vertical structure of tropical forests influences other ecological and environmental variables that are essential for the functioning of the ecosystems. Integrating over 5.9 million Globel Ecosystem Dynamics Investigation (GEDI) LiDAR (Light Detection and Ranging) footprints, multispectral, and synthetic aperture radar (SAR) imagery, we built five national maps at 25 m resolution of five forest structural metrics for Colombia, South America, for the year 2020. We mapped canopy height, the height of half the cumulative returned energy from GEDI (RH50), total canopy cover, foliage height diversity, and total plant area index. The resulting maps tended to have the highest errors in the Amazon and Andean regions. Total cover had the highest relative error. Interrelationship curves between forest structural metrics of GEDI footprints are maintained across mapped metrics, indicating that the predictive models preserve structural relationships observed in GEDI data. Due to the medium-high spatial resolution and national coverage of the forest structural maps presented in this work, these maps will be useful for evaluating and mapping other ecological variables and conservation priorities in Colombia.

Accurate delineation of glacier boundaries is essential for monitoring glacier changes. Debris-covered glaciers often resemble the textures of their surrounding terrain, making boundary extraction especially challenging. Interferometric coherence from Synthetic Aperture Radar (SAR) can help highlight these glaciers by reflecting surface changes. However, because optimal temporal baselines vary with glacier activity, this significantly limits the broader effectiveness of coherence for regional glacier boundary extraction. To address this, we introduce the Feature-based Amplitude Coherence Ratio (FACR), a new index that advances glacier mapping by framing temporal baseline selection as an optimizable feature. Unlike traditional approaches, FACR uses an F-test to automatically identify interferometric pairs that best separate glacier ice from stable terrain. We tested this method on glaciers in the Gyala Peri area of the southeastern Tibetan Plateau, using 14 ALOS−2 PALSAR images. The results show that FACR offers stronger robustness in capturing glacier surface information and achieves higher boundary accuracy (IoU: 0.68, Dice: 0.81) than other indices. FACR therefore provides a practical approach for regional glacier inventories and change detection.

Abstract. Accurate short-term wave forecasts are crucial for numerous maritime activities. Wind and surface currents, the primary forcings for spectral wave models, directly influence forecast accuracy. While remote sensing technologies like Satellite Synthetic Aperture Radar (SAR) and High Frequency Radar (HFR) provide high-resolution spatio-temporal data, their integration into operational ocean forecasting remains challenging. This contribution proposes a methodology for improving these operational forcings by correcting them with Artificial Neural Networks (ANNs). These ANNs leverage remote sensing data as targets, learning complex spatial patterns from the existing forcing fields used as predictors. The methodology has been tested at three pilot sites in the Iberian–Biscay–Ireland region: (i) Galicia, (ii) Tarragona and (iii) Gran Canaria. Using SAR as a reference, the ANN corrected winds present Root Mean Square Deviation (RMSD) reductions close to 35 % respect to ECMWF-IFS, and improvements close to 3 % for the scatter-index. Surface currents are also improved with ANNs, reaching speed and directional biases close to 2 cm s−1 and 6° and correlation close to 35 % and 50 %, respectively. Using these ANN forcings in a regional spectral wave model (Copernicus Marine IBI-WAV NRT) leads to improvements in the Wave Height (Hm0) bias and RMSD around 10 % and 5 % at the NE Atlantic. Mean wave period (Tm02) also improves, with reductions of 17 % and 5 % in bias and RMSD. Preliminary moderate improvements were also present in extreme events (e.g. storm Arwen at Galicia, November 2021), as the Hm0 was corrected close to 0.5 m and Tm02 by around 0.4 s. However, properly quantifying this impact requires further assessment.

CDANet: A context-detail aware network for marine oil spill detection in SAR imagery.

Improving soil moisture estimation in wet soils using L-band Synthetic Aperture Radar (SAR) through polarization and filtering optimization

Abstract Reliable rainfall estimation is essential for hydrological modeling, particularly as climate change intensifies rainfall extremes and challenges water resource management. In many intertropical basins, sparse observation networks limit quantitative precipitation estimation, making satellite precipitation products (SPPs) a key alternative. However, SPP performance varies geographically and must be assessed against high-resolution reference data. This study first evaluates five state-of-the-art SPPs at their native resolution (0.1°, 30 min) against high-resolution weather radar observations in French Guiana. Second, it introduces a correction framework that combines image classification with tailored bias adjustment. Rain fields are first grouped into clusters using k -means applied to their spatial features, distinguishing different rainfall structures. Within each cluster, a quantile matching by parts (QMP) correction scheme trained on weather radar data is applied. The correction scheme is first adjusted on a training dataset consisting of 4000 coincident radar and satellite rain maps. Then, it can be applied to other satellite rain maps and outside the radar coverage. The main findings are as follows: 1) Among the evaluated products, Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (GPM) (IMERG) Final shows the best performance, but IMERG Late is selected for application due to its shorter latency. 2) The cluster-specific QMP correction reduces SPP bias from −20% to +4%, improves the rainfall intensity distribution, and improves the spatial variability of the rain fields and the diurnal cycle. 3) In hydrological simulations of the Mana River basin, the corrected product improves the Kling–Gupta efficiency from 0.36 to 0.83 compared to benchmark simulations using radar data. Overall, this relatively simple and computationally light method improves satellite-based rainfall estimation for various potential applications in data-scarce tropical regions, offering a scalable solution as radar networks expand globally.

Synthetic aperture radar for a changing planet: A 25-year global synthesis in hazard assessment, urban development, and ecological applications

Applicability of synthetic aperture radar in remote sensing flood area calculation

ABSTRACT Optical images and synthetic aperture radar (SAR) data exhibit complementary advantages, offering rich spectral and spatial information. The fusion of these modalities to enhance the quality of remote sensing images has garnered increasing attention in recent years. However, fusing optical and SAR images remains challenging due to differences in imaging mechanisms, speckle noise in SAR data, and the difficulty in jointly preserving details, structure, and spectral fidelity. To address these challenges, this paper proposes a multi-directional alignment and dynamic context aggregation network (MDCA-Net) for optical and SAR image fusion, designed to effectively exploit the complementary features of both modalities to generate information-rich fused images. Specifically, the cross-modal multi-directional alignment (CMDA) module is designed to mitigate discrepancies caused by differing imaging mechanisms. To suppress speckle noise and enhance structural details in SAR images, SAR dynamic context detail enhancement (SDCE) module is developed. Furthermore, a globally shift-aware context aggregation (GSCA) module is designed to jointly preserve detail, structural coherence, and spectral fidelity in the fused image. Compared with seven representative fusion methods, MDCA-Net demonstrates superior visual quality and achieves more outstanding quantitative and qualitative evaluation results. In addition, MDCA-Net significantly improves classification accuracy across multiple land cover types, including wetland, built-up area, grassland, forest, and cropland. On the Hunan dataset, MDCA-Net achieves gains of 4.48%, 2.63%, and 4.81% in mean intersection over union (mIoU), mean producer’s accuracy (mPA), and mean precision (mPrecision), respectively.

In the recent era, our earth has become more prone for various disasters like Earthquake, Tsunami, Cyclones etc. Earthquakes are the most devastating natural disasters, causing significant loss of life, damage to infrastructure and long-term economic and social consequences. This study delves earthquake that struck Kahramanmaras, Turkey, on February 6, 2023, registering a magnitude of 7.8 and significantly impacted southeastern Turkey and northern Syria. The earthquake-affected zone, based on model estimates, spanned roughly 350,000 square kilometers, experiencing extensive destruction. Estimates indicate that up to 9.1 million people were directly impacted, with several million rendered homeless. In Turkey, nearly 46,000 people lost their lives, with the provinces of Hatay and Kahramanmaras being the hardest hit, reporting about 21,900 and 12,600 deaths respectively. Turkey’s Disaster and Emergency Management Presidency (AFAD) noted that approximately 280,000 buildings were either severely damaged or collapsed, while an additional 710,000 structures suffered major damage. In this continuation the focal point of this study is the application of remote sensing technologies, such as synthetic aperture radar (SAR) and optical satellite imagery to assess earthquake-induced changes in land surface deformation and infrastructure damage. The use of satellite data allows for rapid damage assessment, facilitating efficient disaster response and recovery efforts. This case study underscores the importance of integrating satellite-based remote sensing with ground-based data to provide a comprehensive assessment of earthquake impacts. The findings not only contribute to the understanding of seismic hazards in the region but also demonstrate the potential of Sentinel data, processed through SNAP, in improving disaster response strategies, hazard mapping and infrastructure planning in earthquake-prone areas like Kahramanmaras.

Landslides are among the most unpredictable and destructive sediment-related disasters, especially in mountainous regions with complex terrain and limited field accessibility. In 2004, a catastrophic landslide from the Mount Bawakaraeng Caldera delivered more than 100 million cubic meters (MCM) of sediment into the Bili-Bili Reservoir, filling its dead storage and threatening its long-term functionality. his study uses Sentinel-1A satellite imagery and Differential Interferometric Synthetic Aperture Radar (DInSAR) to perform a rapid, spatially driven assessment of landslide hazards in the Bili-Bili Reservoir Catchment Area. The results reveal surface deformation of up to ±1.55 meters, concentrated in upstream zones. High-risk areas span 71.00 km², with an estimated mobilizable volume of 110.04 MCM and a potential sediment yield of 27.14 MCM per year, nearly equal to the reservoir’s dead storage. To mitigate this threat, the study proposes an integrated mitigation framework. Structural interventions include rehabilitating existing sediment control systems and constructing new sabo dams. Non-structural strategies such as slope revegetation and bioengineering are also recommended. This study demonstrates how remote sensing can identify subtle ground deformation and provides actionable insights for safeguarding critical water infrastructure in sediment-prone tropical watersheds.

Wildfires are a critical global threat, necessitating advanced early detection and monitoring systems. This research introduces a novel multi-modal framework that integrates wide-area Synthetic Aperture Radar (SAR) for all-weather surveillance with high-resolution UAV-based optical and thermal imagery for precise analysis. The proposed hybrid learning framework utilizes FPANet, a Vision Transformer-based architecture that captures both local textures and global spatial dependencies to achieve robust segmentation from SAR data under cloudy or smoky conditions. For fine-grained analysis, the system employs DualSegFormer, a model designed for the synergistic multi-modal fusion of thermal and RGB UAV images, ensuring high-fidelity fire front delineation even when visibility is compromised. Additionally, a Vision-Language Model (VLM) is integrated to translate complex sensor data into actionable, human-readable insights for effective disaster response. Experimental results demonstrate a significant improvement over conventional methods: the SAR-based FPANet achieves an F1-score of 0.830 and an Intersection over Union (IoU) of 0.750, the UAV-based DualSegFormer attains a superior F1-score of 0.946, and the VLM component shows strong semantic alignment with a BERTScore of 0.953. These results confirm the ability of the proposed hybrid learning approach to provide more effective and reliable wildfire monitoring, thereby advancing ecosystem resilience and facilitating timely disaster response in alignment with international SDG initiatives.

The paper presents results of the research aimed at increasing the contribution of satellite remote sensing methods to addressing environmental protection issues in the Caspian Sea and identifying potential environmental risk zones. The study is based on data collected over a three-year period, from January 2022 to December 2024. The data source is high-resolution radar imagery obtained by Synthetic Aperture Radars (SAR) C-SAR onboard Sentinel-1A satellite, and the data in the visual (VIS) range of the electromagnetic spectrum provided by multispectral sensors MSI (Multispectral Instrument) on Sentinel-2A and Sentinel-2B satellites, as well as scanning radiometers OLI (Operational Land Imager) and OLI-2 on Landsat-8 and Landsat-9 satellites. The research presented here continues a series of studies we previously conducted in this region and reported in publications listed in the bibliography. Our primary focus was on the regions with the most severe surface oil pollution, identified in our previous research using long-term satellite data. These are the regions with natural hydrocarbon seepage from the seabed near Cape Sefid Rud and the Cheleken Peninsula; the oil production region of Oil Rocks; the South Caspian Basin, which features a cluster of mud volcanoes on the seabed; and the main shipping routes. We described the specific manifestations of each type of surface oil pollution. For each study region, we estimated the areas of the sea surface potentially exposed to oil pollution. Further, we performed a quantitative assessment of the interannual, seasonal, and spatial variability of different types of oil pollution on the Caspian Sea surface across various regions of interest. One of our findings is pronounced seasonal variability in the number of oil patches detected in VIS images of the regions of interest and in the size of individual oil slicks. The frequency and intensity of mud volcanism were also determined for 2022–2024. Special attention was paid to identifying anthropogenic pollution associated with the discharge of oil-containing waters from vessels. Based on satellite data collected from 2022 to 2024, we constructed a cumulative map of petroleum hydrocarbon pollution on the surface of the Caspian Sea, including pollution from oil-containing water spills from ships.

ABSTRACT Synthetic aperture radar (SAR) maritime target classification serves as a critical component in modern maritime surveillance. While deep learning networks, particularly convolutional neural networks (CNNs), have driven substantial progress in this domain, three key challenges constrain their performance and practical deployment: 1) In SAR maritime images, complex inshore backgrounds and speckle noise are prevalent. Targets such as ships span a wide range of scales due to different imaging resolutions and intrinsic size variability, exacerbating inter-class similarity and intra-class variability, 2) Labeled data for SAR maritime target classification are scarce, and sensor imaging modes differ markedly across platforms, and 3) Existing CNNs that fuse traditional hand-crafted features often explicitly treat hand-crafted feature extraction as a necessary component of the network and primarily focus on classification performance, overlooking the requirement to efficiently leverage their feature extraction capabilities in downstream tasks. To overcome these challenges, this article proposes a novel SAR maritime target classification network (MBLKNet) based on large kernel convolution and multi-task self-supervised learning. In MBLKNet, four improved designs for network structure are proposed to enhance classification accuracy: 1) macro design, 2) multi-branch large kernel convolution module (MBLKCM), 3) lightweight channel-interactive multi-layer perceptron (LCIMLP), and 4) micro design. In addition, a multi-resolution unlabeled SAR maritime target dataset (SL-SARShip) and a masked image modeling framework, HOGSparK, are proposed to enable the pre-training of MBLKNet under joint supervision of pixel and HOG features. Comparison results on OpenSARShip 2.0 and FUSAR-Ship with state-of-the-art networks, as well as experiments on SSDD for SAR downstream target detection and instance segmentation, demonstrate that the proposed MBLKNet achieves superior performance and strong feature extraction ability.

Abstract Hurbanovo, as the only station for continuous monitoring of gravity changes in Slovakia, is equipped with a relative spring gravimeter gPhoneX#108. The gravity station was established in 2019 and has been a part of the International Geodynamics and Earth Tide Service (IGETS) since 2021. In addition to gravity measurements, several complementary measurements at this station are carried out, including Global Navigation Satellite Systems (GNSS) for geodetic positioning co-located with Interferometric Synthetic Aperture Radar (InSAR) corner reflectors, seismometer for seismic monitoring, meteorological measurements including atmospheric pressure, temperature, and precipitation, as well as hydrological measurements of soil moisture and groundwater levels. Our contribution includes a fundamental statistical and correlation analysis of the temporal data collected through these diverse techniques and sensors, highlighting foundational insights obtained from the investigation. Results demonstrate the benefit of thermal insulation of the gPhoneX instrument but also reveal the increased anthropogenic noise at the Hurbanovo site coming most likely from the nearby traffic. Despite of the noisy environment gravimeter can capture significant precipitation events and larger groundwater variations.

China's expanded HIV testing program (since 2015) necessitates longitudinal evaluation of healthcare institution (HCI) performance, which provides critical evidence to enhance the precision and effectiveness of focused testing interventions, to optimize service delivery. We analyzed HIV surveillance data from 202 HCIs in Nanning, China (1989-2020), characterizing: 1) HCI infrastructure distribution, 2) newly reported HIV/AIDS cases and testing volumes, and 3) test positivity rates. Spatiotemporal trends were geo-visualized (ArcGIS 10.7). TOPSIS synthesized case reports and test positivity rates (2010-2020) into composite indices, with radar mapping identifying regional disparities. Reporting shifted from CDC-dominated systems (99.9%, 1989-2004) to hospital-led models (75.6%, 2015-2020). While annual testing volume and case reports increased significantly, test positivity rates declined. Distinct hospital-level stratification emerged: County/township HCIs served primarily local residents (96.2%) and older patients (≥ 50years; 65.7%). Municipal/provincial HCIs reported higher non-local residents (48.6%) and younger patients (15-49years; 53.4%). Geospatial diffusion showed progression from urban to rural areas. Regions with integrated AIDS treatment centers and robust primary care networks demonstrated enhanced case-finding performance. Tiered hospitals and primary care centers may provide complementary case-finding functions. Geographically stratified interventions, which leveraged primary networks for localized epidemics and referral hospitals for mobile populations, represent evidence-based strategies for regions with suboptimal performance.