Accurate and timely rice mapping at regional scales is essential for global food security monitoring, agricultural policy development, and sustainable intensification efforts. Synthetic Aperture Radar (SAR) provides all-weather, day-and-night observation capabilities, making it particularly suitable for monitoring crop distribution. However, phenology-based methods often face challenges due to diverse cropping practices, and deep learning approaches typically require extensive, high-quality ground truth data that are labor-intensive and region-specific. To address these limitations, we propose an Automated Phenology-Guided Deep Learning framework for Rice Mapping (APDL-Rice), which integrates phenological knowledge with deep learning to eliminate the need for manual sample collection. Leveraging time-series SAR imagery, APDL-Rice automatically extracts phenological signals to generate reliable training samples. The framework was validated across four experimental sites in Guangdong and Heilongjiang, which have different rice cropping practices and agroecological conditions. Results demonstrate that APDL-Rice delivers consistently high performance, with classification accuracies ranging from 83.98% to 94.54%, often surpassing conventional sample-dependent methods. In addition, we investigate the effects of sample quantity and composition on model performance, providing practical insights for deep learning-based rice mapping in sample-constrained scenarios. These findings highlight the method’s strong adaptability across varying cropping systems and its potential for broader application under limited-sample conditions.

Accurate information on rice planting areas and spatial distribution is critical for agricultural management in China; however, mapping efforts in regions like Sichuan Province are severely constrained by persistent cloud cover and fragmented terrain. Existing phenology-based methods and coarse-resolution products often fail to provide precise paddy localizations or usable training labels in such complex environments. To address these limitations, this study proposes a robust framework integrating optical and Synthetic Aperture Radar (SAR) imagery. The methodology employs a phenology-driven strategy for rapid candidate area annotation, coupled with an asymmetric feature extraction mechanism that incorporates a Multi-Scale Semantic Enhancement Module (MSEM) and a Category-Balanced Feature Fusion Module (CBFM) to facilitate adaptive and effective cross-modal fusion. Applying this framework to the Tianfu New Area (2019–2023) yielded 10-meter resolution rice distribution maps with a rice Intersection over Union (IoU) of 83.31% and a statistical correlation ( R 2 ) of 0.972. These results demonstrate the framework’s capacity for selective multi-source fusion and cost-effective sampling, facilitating precise rice mapping in challenging agricultural landscapes.

Moon-based synthetic aperture radar (SAR) offers a promising platform for long-term monitoring with global coverage of Earth's vegetation tipping elements. However, wind-induced vegetation motion during aperture synthesis can degrade image stability and spatial resolution. To address this, we integrated multi-year wind speed records, a parameterized internal clutter motion model, and Moon-based SAR geometry to systematically analyze the dynamic scattering behavior across representative vegetation types under prevailing wind conditions. The results reveal two distinct degradation mechanisms: high-latitude sensitivity, driven by orbital geometry, and high-biomass severity, driven by vegetation structure. Crucially, we demonstrate that the unique ultra-high orbit enables Moon-based SAR to overcome these challenges, achieving sub-100 m resolution across all representative ecosystems by surpassing the traditional half-antenna length limit. Furthermore, the analysis indicates that C-band may be preferentially considered in system design, as it achieves a comparatively favorable trade-off between temporal coherence required for imaging stability and structural sensitivity across diverse vegetation conditions. These findings support the design and capability assessment of Moon-based SAR systems optimized for vegetation observation in the context of ecosystem transition monitoring.

Tropospheric delay remains a critical error source in time-series interferometric synthetic aperture radar (TS-InSAR), particularly in mountainous and plateau regions where seasonal stratification and stochastic turbulence coexist. Current correction methods based on global atmospheric models (GAM) often underestimate the turbulent delay and fail to effectively address its spatiotemporal variability. We propose a novel GAM-based tropospheric correction method termed GAM-STTD that simultaneously models stratified and turbulent delays. The method integrates (1) the Prophet forecasting model to capture atmospheric variability in the temporal domain and (2) an adaptive kernel density estimation (AKDE) strategy to optimize three-dimensional zenith total delay (3D-ZTD) sampling according to the terrain gradient. We incorporated GAM-STTD into the TS-InSAR framework and validated it using both simulated and RADARSAT-2 datasets over Lijiang Basin, China. The results showed that the GAM-STTD model overcomes the underestimation of stochastic turbulent delay observed in existing atmospheric models, with a mean bias of 0.29 cm relative to the ERA5 reference ZTD. The model reduced the average phase standard deviation (STD) across 125 interferograms from 2.88 rad to 2.51 rad. In addition, the GAM-STTD model reduces the turbulence-induced interferometric phase semi-variance from 2.38 rad² to 1.49 rad², which further improves the accuracy of the TS-InSAR deformation solution.

ABSTRACT Fractional vegetation cover (FVC) is a crucial biophysical indicator for monitoring vegetation abundance and distribution. Existing FVC estimation methods based on synthetic aperture radar (SAR) images often overlook the heterogeneity of scattering mechanisms across vegetation types, limiting accuracy in complex and topographically varied regions. To address this, this study proposes a vegetation-type-adaptive approach that integrates tailored feature selection with ensemble learning algorithms optimized by the dung beetle optimizer (DBO), based on dual-polarized SAR images. The study area is classified into cropland, woodland, and wetland using the European Space Agency’s (ESA’s) WorldCover 10-m product to address vegetation heterogeneity. Following this classification, 25 SAR features – including backscatter coefficients, polarimetric decomposition parameters, and radar vegetation indices – are extracted from Sentinel-1 SAR data over the Dongting Lake region, China. The minimum redundancy maximum relevance (mRMR) algorithm identifies optimal feature subsets for each vegetation type, effectively capturing unique scattering characteristics. Five ensemble models, adaptive boosting (Adaboost), categorical boosting (CatBoost), extreme gradient boosting (XGBoost), random forest (RF), and light gradient boosting machine (LightGBM), are trained separately for each vegetation type, with hyperparameters tuned via DBO. To mitigate terrain-induced layover effects, an adaptive OTSU-based thresholding strategy is applied. Experimental results demonstrate high-precision FVC estimation, with R 2 values of 0.8916, 0.8313, and 0.9303 for cropland, woodland, and wetland, respectively. Compared with conventional uniform models, this vegetation-type-specific approach significantly improves accuracy by addressing feature heterogeneity and geometric distortions. This method provides a robust and physically consistent solution for FVC mapping in heterogeneous-vegetated regions, significantly advancing SAR-based vegetation monitoring under complex terrain conditions.

Radar images are affected by speckle noise, which seriously affects its subsequent applications. Recently, the Convolutional Neural Networks (CNN) have been widely used in radar images despeckling and achieved remarkable results due to its powerful learning ability. However, most existing deep learning-based despeckling methods recover clear images in the spatial domain and rarely explore potential solutions in the frequency domain. This paper is the first attempt to combine spatial and frequency domain for radar image denoising, and proposes a spatial-frequency domain aggregation network for synthetic aperture radar (SAR) images despeckling, called as SFSARNet. SFSARNet contains three key components: a spatial-domain information branch, a frequency-domain information branch, and a dual-domain aggregation module. The spatial domain information branch mainly works in the spatial domain and is used to explore the local structural details. The frequency domain information branch works in the frequency domain for exploring global context information. The dual-domain aggregation module learns and integrates complementary information from both the spatial and frequency domains, enabling the network to benefit from both local and global context information. The proposed SFSARNet is validated on simulated and real radar image. Both quantitative and qualitative comparisons demonstrate that the proposed SFSARNet outperforms the existing mainstream methods.

Spaceborne optical sensors provide continuous Earth observation, but atmospheric interference still limits their practical reliability. On average, clouds cover 67% of the Earth's surface. This constant coverage degrades the data continuity needed for precision agriculture, disaster monitoring, and proactive Internet of Things (IoT) systems. Recent deep generative networks produce visually appealing cloud-free images. However, when faced with thick clouds ([Formula: see text] opacity), these models often hallucinate topologies. They synthesize statistical guesses instead of recovering the actual ground reflectance. For high-stakes telemetry, predictable failure is safer than an undetected hallucination. This paper introduces Thermo-Cloud Removal (Thermo-CR), a real-time cloud removal framework. It integrates Radiative Transfer inversion, weather-driven transmission estimates, geographic priors, and multi-scale fusion to restore optical imagery without requiring Synthetic Aperture Radar (SAR). Thermo-CR treats the cloudy atmosphere as a thermodynamic medium. By pulling live meteorological telemetry (Relative Humidity (RH) and Temperature (T)) through the Open-Meteo REST API, the system calculates optical depth and performs a deterministic inversion of the Radiative Transfer Model. Pure inverse models amplify noise under extreme occlusion ([Formula: see text]). To prevent this, we apply a Global Positioning System (GPS)-anchored multi-scale fusion with clear-sky temporal priors. We evaluated Thermo-CR on a synthetically occluded paired dataset covering varied topologies (Amazon, London, Seattle). The system degrades predictably under 90% cloud cover and avoids structural hallucination. It achieves an average Structural Similarity Index Measure (SSIM) of 0.9925 and a Peak Signal-to-Noise Ratio (PSNR) of 55.94 dB in under 13 milliseconds per frame, outperforming standard Dark Channel baselines.

The Xiongba landslide group, characterized by continuous movement and distinct signs of reactivation, necessitates continuous monitoring to mitigate potential hazards. While Interferometric Synthetic Aperture Radar (InSAR) offers wide-area coverage and high precision, its limitation to line-of-sight (LOS) measurements and constrained temporal resolution hinders the accurate characterization of true kinematic directions and dynamic evolutionary patterns. To address these constraints, this study proposes a method that first projects LOS deformations onto the slope-aspect direction to reveal actual displacement vectors. Subsequently, an Unscented Kalman Filter (UKF) is employed to fuse ascending and descending Sentinel-1A time-series slope-aspect deformations, thereby enhancing temporal resolution and capturing the landslide's dynamic evolution. Monitoring results based on Sentinel-1A data from January 2018 to June 2024 indicate that the Xiongba, Sela, Maiba, and Gongba landslides remain in a state of continuous activity. Notably, the Xiongba landslide exhibits maximum LOS deformation rates of 94.78mm/yr (ascending) and 82.57mm/yr (descending). Quantitative validation demonstrates that the UKF method significantly improves monitoring accuracy, reducing the Root Mean Square Error (RMSE) by 32.1% compared to the traditional Kalman filter, while shortening the average observation interval to 7.18days. The evolutionary analysis reveals that the deformation is primarily driven by gravitational forces and river erosion traction, and triggered by seasonal rainfall events.

ABSTRACT Ship detection from synthetic aperture radar (SAR) imagery is essential for maritime monitoring and ocean surveillance. However, accurate detection remains challenging due to speckle noise, complex coastal environments and large variations in ship size and scattering characteristics. These factors often lead to incomplete feature representation and inaccurate localization in existing deep learning-based methods. To address these issues, we propose a dual-stream hierarchical feature fusion network (DHFFNet) for segmentation-driven SAR ship detection. The framework combines a Convolutional Neural Network primary encoder and a Swin Transformer parallel encoder to capture complementary local and global features. In addition, several modules are introduced to enhance feature representation, including a Spatial Enhancement Module (SEM), a Feature Detail Preservation Module (FDPM), a Channel Enhanced Attention Module (CEAM) and a Cross-resolution Pixel Attention (CPA) module for improved feature reconstruction. Extensive comparative experiments demonstrate that DHFFNet achieves superior performance and significantly improves detection accuracy on SAR ship imagery. The results indicate that DHFFNet effectively improves feature representation and detection accuracy in complex SAR scenes.

Eucalyptus supports Ethiopia's economy and its zero-carbon strategy, yet its rapid expansion in the highlands of Ethiopia creates ecological concerns. For better management, accurate mapping is needed, but it is challenged by cloud contamination, spectral mimicry, and the requirement for high-resolution commercial imagery. Therefore, this study used an integrated multi-sensor satellite data, such as Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 Multi-Spectral Imagery (MSI), for reliable and cost-effective mapping of Eucalyptus trees in Meket district, Ethiopia. Sentinel-2 data is endowed with multi-spectral bands that are sensitive to chlorophyll and leaf water content, whereas Sentinel-1 SAR also functions in all weather conditions and can capture moisture and structural information of trees. Eighteen features from spectral bands, radar backscatter, and vegetation indices were fused using a feature-level fusion strategy and classified using Random Forest (RF), Support Vector Machine (SVM), and Classification and Regression Trees (CART). Results show that using freely available satellite data, RF achieved the highest performance, with an Overall Accuracy (OA) of 90% and a kappa coefficient of 0.81 in detecting Eucalyptus trees. SVM also achieved nearly the same performance with a 1% difference from RF. The study concludes that using publicly available Sentinel-1/2 fusion data with an appropriate classifier provides cost-effective, reliable, and accurate results for mapping of Eucalyptus, and supports Ethiopia's zero-carbon strategy. It also helps policymakers and planners by providing geospatial technology-based land use planning for monitoring and sustainable management of Eucalyptus trees.

Application of multi-resolution techniques, intelligent classification and semantic data fusion for the identification of oil spills in SAR imagery.

Remote sensing for marine oil spill detection, mapping, and monitoring: A systematic review and bibliometric analysis.

The essential role of electricity supply for public and private services highlights the need to monitor the stability of power transmission networks during, or immediately after, hazardous events. In the aftermath of calamities, traditional field inspections may be impractical or unsafe, leaving operators without timely information on the condition of critical assets. In this paper, we present and discuss the performance of two automatic Artificial Intelligence (AI)-based models (Multi-Layer Perceptron (MLP) neural network architectures and Support Vector Machine (SVM) model) designed to automatically assess the status of high-voltage transmission towers and power lines through multi-temporal spaceborne Synthetic Aperture Radar (SAR) image analysis. Model development and testing rely on real COSMO-SkyMed Stripmap observations of damaged towers and power lines affected by documented hazardous events across Italy, complemented by simulated tower data generated with a physics-guided, signature-based SAR simulator designed to preserve the observed target-to-background contrast and spatial footprint patterns of real SAR tower signatures. Results indicate that the MLP, trained on either real or simulated data, achieved 100% Overall Accuracy (OA) with no observed false positives or false negatives within the considered visibility-screened real test set, while providing inference times on the order of tenths of milliseconds per target… Computational performance characteristics, operational advantages, and the potential pathway toward satellite on-board porting are discussed to enhance situational awareness and support the prioritisation of interventions during critical events.

The 2023 ML 4.4 Umbertide extensional earthquake in Central Italy provides a valuable opportunity to test the capability of Differential Interferometric Synthetic Aperture Radar (DInSAR) to detect surface deformation associated with small earthquakes (M < 5). In this work, we invert Sentinel-1 displacement maps from the European Plate Observing System (EPOS) platform to investigate the possible source. We perform a linear Bayesian static slip inversion of DInSAR line-of-sight data, constraining the fault geometry using the nodal planes and the new relocated mainshock. The results suggest the activation of the NE-dipping splay of the Alto Tiberina Fault, consistent with the relocated aftershock distribution. The deformation component, obtained by combining multiple line-of-sight displacement maps, corresponds to 2 cm of vertical and 1.5 cm of eastward horizontal deformation and further supports the NE-dipping solution. This work demonstrates the capability of an integrated InSAR-based approach to detect weak deformation and model frequent small-magnitude earthquakes in high seismogenic potential areas.

Oil spill detection on sea surface with dual-polarimetric SAR imagery integrating polarization features and oil-seawater boundary information.

This study presents a comprehensive assessment of land deformation patterns across a 22,043km² area of southeastern Louisiana and coastal Mississippi using Small Baseline Subset Interferometric Synthetic Aperture Radar (SBAS-InSAR) time series analysis from 2020 to 2024. The study processed 461 ascending-path interferometric Sentinel-1 SAR image pairs using the MintPy framework, implementing a minimum temporal coherence threshold and a land cover class-specific coherence thresholds filtering to optimize signal detection across diverse coastal environments. The analysis integrated National Land Cover Database (NLCD) 2021 data to characterize subsidence vulnerability patterns across different land cover types. The research results reveal extreme spatial heterogeneity in land surface deformation, with line-of-sight velocities ranging from - 43.71mm/year to + 30.23mm/year. Coherence-based filtering by land cover class (Developed Land, Pasture/Hay/Barren, Grassland/Shrub/Crops and Forest/Wetland/Marsh) produced LULC specific LOS velocity maps. Land cover class-specific coherence threshold filtering produced separate LOS velocity maps for Urban, Barren/Pasture, and Marsh/Agriculture land cover categories, enabling spatially refined deformation characterization across the delta. Results indicate that marsh and wetland areas experience the highest subsidence rates, particularly in organic-rich marshlands and low-elevation areas, while uplift signals are observed in southwestern coastal areas, related to sediment deposition processes. Validation against NOAA CORS GNSS stations (BVHS and GRIS) shows InSAR SBAS LOS velocities of -12.46mm/yr and - 11.82mm/yr respectively, compared to GNSS vertical-only rates of -2.5mm/yr to -5.5mm/yr. The integration of land cover data in this study proved crucial for understanding deformation mechanisms, as areas with rich organic soils and below-sea level elevations showed accelerate subsidence rates.

Due to the complexity of synthetic aperture radar (SAR) imaging mechanisms, SAR ship detection faces challenges such as difficulty in sample annotation and the influence of complex backgrounds, leading to poor target readability and difficulties in feature representation. Compared to SAR images, optical remote sensing images offer advantages such as high resolution and intuitive visualization, making them complementary to SAR images. Based on this, this article leverages the more intuitive structural features of optical remote sensing images to guide SAR ship target feature learning and proposes the structure-guided domain-adaptive network (SGDANet) for few-shot SAR ship detection. This proposed network follows a convolutional neural networks (CNNs)-transformer architecture, modeling the structural and edge token features of optical remote sensing images and embedding them into the network. Additionally, a feature fusion mechanism based on the split-fuse-merge strategy and attention mechanism is designed to achieve improved domain adaptation performance during the adversarial learning stage. Experiments on three self-built ship datasets illustrate that SGDANet outperforms other related network models in both three-shot and five-shot scenarios. Notably, SGDANet also exhibits good performance in zero-shot SAR target detection tasks, indicating its strong generalization ability. Code is available at https://github.com/RSIP-NJUPT/SGDANet.

With the ongoing climate changes, the Arctic is experiencing large changes, which have the potential to negatively influence natural environments, human settlements, and cultural heritage. The permanently frozen ground (permafrost) and the seasonally thawing and freezing active layer is sensitive to changes in temperature and precipitation patterns. With permafrost degradation and active layer thickening, Arctic cultural heritage and modern buildings can be increasingly affected by permafrost-related hazards. Mitigating these hazards requires implementation of tools at the local to regional scale. This paper presents a spatial dataset of permafrost-related hazards in Svalbard, integrating three geospatial indicators documenting (1) geomorphology-based hazard susceptibility, (2) Interferometric Synthetic Aperture Radar (InSAR) ground movement, and (3) coastal erosion hazard susceptibility based on the distance to the coastlines. The resulting hazard indicator is combined with vulnerability indicators to provide risk estimates for cultural heritage sites and modern buildings in and around Longyearbyen and Ny-Ålesund. The dataset includes the products from all steps, from the initial spatial geodata to the hazard/vulnerability indicators and risk estimates.

Ground deformation monitoring is critical for safety and environmental management in modern mining. Active mining sites are highly exposed to terrain instabilities and subsidence, risking infrastructure integrity, disrupting operations, and posing hazards to communities. In this context, Differential Synthetic Aperture Radar Interferometry (DInSAR) techniques provide an effective and non-invasive tool capable of detecting millimetric surface displacements. This study implements the Small Baseline Subset (SBAS) technique through an open-source workflow based on the Python package hyp3_sbas, enabling semi-automated and reproducible interferometric processing by combining HyP3 with MintPy. The workflow is applied to the Björkdal gold mine (Sweden), a pilot site of the Horizon Europe XTRACT project focused on enhancing resilience in critical raw material supply chains. Integrating Sentinel-1 viewing geometries resolves the true vertical deformation field, yielding an overall mean velocity of −3.99 mm/year across the mining complex, with significant displacement rates concentrated below the 25th percentile (Q1) at −11.07 mm/year. Sector-specific analysis reveals localised subsidence accelerating over underground footprints and tailings storage facilities (mean velocities of −6.56 and −3.98 mm/year; Q1 thresholds near −13.00 mm/year), contrasting with the geomechanical stability observed at the open-pit area (mean: −0.45 mm/year). The proposed open-source framework shows strong potential for operational satellite-based monitoring, supporting predictive maintenance and early-warning strategies for risk management in mining environments while simplifying and standardising the interferometric processing workflow.

Forest dominant height is a fundamental structural attribute that reflects site conditions and forest growth potential. Here we present a nationwide 30 m resolution forest dominant height dataset for China (FDH-30C). The dataset is calibrated using 1,117 km² of high-density unmanned aerial vehicle (UAV) light detection and ranging (LiDAR) data distributed across all eight major vegetation divisions across China as reference data, representing diverse stand ages, structures, and species compositions. To produce spatially continuous estimates, the model used in this dataset integrates 30 geospatial predictors derived from multi-source remote sensing products, including climatic, edaphic, topographic, vegetation, and Synthetic Aperture Radar (SAR)-based variables. A two-stage hybrid modeling framework combines the UAV LiDAR reference data with these predictors to generate spatially coherent estimates while preserving local accuracy and reducing ecozone boundary effects. The resulting map provides a consistent national baseline for applications such as site-index mapping, growth-and-yield parameterization, biomass and carbon estimation, vertical structure analysis, and the evaluation of spaceborne LiDAR missions.