Remote Sensing doi: 10.3390/rs18101638

Authors: Muhammad Atif Bilal Yongzhi Wang Kateryna Hlyniana Zubair Nabi

The Western Chagai Belt of Pakistan hosts major porphyry Cu-Au systems, but prospectivity mapping in this arc remains difficult because favorable lithology, intrusive bodies, fault corridors, hydrothermal alteration, and Cu geochemical anomalies are spatially heterogeneous across a structurally complex and arid terrain. These conditions create a scientific need for an integrated mapping framework that can combine remote sensing alteration evidence, geology, structure, and geochemistry within a unified and reproducible workflow. This study presents GeoHybridGNN, a hybrid deep learning framework for porphyry copper prospectivity mapping in the Western Chagai Belt. The framework integrates multi-source raster evidence, including remote sensing-derived spectral alteration indices, a Cu geochemical raster, and distance-to-fault information, with graph-based node representations that combine regular neighborhood adjacency on retained grid cells with node attributes derived from lithology and aligned geoscientific raster summaries. All predictors were harmonized to a common 30 m reference raster grid and evaluated using five-fold spatial block cross-validation to provide a more spatially realistic assessment than ordinary random splitting. The implemented model combines a CNN-based raster patch encoder with a GraphSAGE-based graph classifier. Raster patches extracted around graph nodes are encoded into 64-dimensional embeddings, and these embeddings are concatenated with node-level graph features before full-batch graph learning and prediction. Copper occurrences were used only for supervised label assignment and evaluation and were not used as predictive inputs. The results show that GeoHybridGNN produces spatially coherent prospectivity maps, stable fold-wise prediction patterns, and improved target delineation relative to the tested comparison models. Cu geochemical integration produces only a limited change in global discrimination but provides modest local target sharpening in selected zones. These results indicate that GeoHybridGNN can serve as an uncertainty-aware and geologically constrained decision support workflow for porphyry copper targeting. More broadly, the framework provides a transparent strategy for exploration screening in structurally complex and data-heterogeneous metallogenic belts where remote sensing, geological, structural, and geochemical evidence must be integrated consistently.

Remote Sensing doi: 10.3390/rs18101637

Authors: Hengshan Si Zhipeng Li Sen Lu Jinsong Zhang

Walnut leaf necrosis causes leaf desiccation and premature abscission, substantially reducing photosynthetic efficiency, impairing fruit development, and ultimately leading to yield loss and quality deterioration. In severe cases, it accelerates branch senescence or even whole-tree mortality, resulting in considerable economic damage to the walnut industry. Rapid and accurate monitoring of this disease is therefore essential for sustainable production. This study aimed to characterize the different stages of walnut leaf necrosis using spectral analysis and develop classification models for stage-specific identification. Leaf samples representing healthy leaves and the early, middle, and late stages of necrosis were analyzed for spectral responses. Sensitive bands were identified using the variable importance in projection (VIP), successive projections algorithm (SPA), and the combined VIP-SPA method, and corresponding vegetation indices were constructed. The selected features were incorporated into classification models based on random forest (RF), extreme gradient boosting (XGBoost), and convolutional neural networks (CNNs). Results revealed that the red-edge (640–700 nm) and near-infrared (720–1000 nm) regions were identified as key diagnostic spectral ranges. Among the vegetation indices evaluated, the Simple Ratio Index (SRI) calculated from reflectance at 705.7 nm and 707.1 nm, the Normalized Difference Index (NDI) using the same band pair, and the Difference Index (DI) derived from 417.1 nm and 638.7 nm emerged as the most sensitive indicators of disease severity. Classification accuracies for different necrosis stages reached 0.9583, 0.9583, and 0.9333, respectively. These findings demonstrate that the identified spectral bands and vegetation indices provide robust tools for monitoring the progression of walnut leaf necrosis.

Remote Sensing doi: 10.3390/rs18101636

Authors: Song Han Xing Han Yibo Xu Yongqin Tian Weidong Zhang Wenyi Zhao

Remote sensing image classification is one of the fundamental tasks in the field of remote sensing and plays a critical role in Earth observation applications. However, the inherent multi-scale characteristics of this task pose significant challenges to scene classification. To address these issues, we propose a novel framework that integrates the Contrastive Language–Image Pre-training (CLIP) model, multi-scale data, and efficient inference strategy. The proposed framework transfers general-purpose features learnt from natural images to remote sensing image classification. Specifically, this framework leverages the rich feature representations learnt by the CLIP model in the contrastive learning procedure and adopts it as the backbone network of the model to extract fine-grained and multi-scale features for remote sensing images. That is, the model can learn local fine-grained details but also encode global contextual information useful for the classification of visually similar scene categories. Afterwards, AdapterFormer module is inserted into the few selected layers of CLIP model, which can effectively enhance model performance and have low computational overhead. This helps efficient knowledge sharing and introduces new features at the model level. Furthermore, to alleviate possible performance deterioration brought about by multi-scale feature variation, a multi-scale training set is constructed at data level, providing complementary multi-scale information. Through the synergy of all these strategies above, the proposed method greatly improves the classification performance of multi-scale remote sensing images. Extensive experiments on the MEET dataset (it includes 80 fine categories and more than 800,000 samples) show that the proposed method greatly improves the performance. Compared with general-purpose classification networks and remote sensing-related models, the proposed method always gets state-of-the-art results.

Remote Sensing doi: 10.3390/rs18101635

Authors: Jiayong Yu Jing Zhang Jiangchao Mu Jiachun Guo Deliang Lv Xiaoxue Du Peng Lin

To improve the data quality of underwater point clouds acquired by UAV-borne LiDAR bathymetry, a scanline-based sliding window filtering method is proposed based on an analysis of scanline data characteristics. Scanline data of underwater point clouds are first extracted from raw point clouds, and the radius outlier removal algorithm is employed to eliminate outliers. Taking the acquisition time of scanline points as the X-axis and elevation as the Y-axis, a 3D problem is simplified into a 2D representation, and a sliding window is constructed along the scanline. Robust least-squares fitting is applied within the window. The median absolute deviation of the fitting residuals is adopted to calculate the terrain feature values for quantifying the terrain complexity, followed by an adaptive filtering threshold determination according to terrain feature values. Fine filtering of the individual scanlines is performed using a point-by-point sliding window. Experimental results demonstrate that the proposed method is adaptable to various terrain conditions, achieving a noise recall rate ≥ 96%, an overall filtering accuracy ≥99%, and an F1-score ≥ 0.9. Particularly, the precision rate in flat-water areas reached 97.37%. Overall, the proposed filtering method effectively separates noise points while preserving detailed terrain features and supports UAV-borne LiDAR bathymetry for mapping complex shallow-water regions.

Remote Sensing doi: 10.3390/rs18101634

Authors: Ning Guan Baojian Liu

DDM is the primary Level-1 observable of spaceborne Global Navigation Satellite System Reflectometry (GNSS-R). Over the past decade, the discretization strategy of Delay-Doppler Map (DDM) systems has been primarily optimized for ocean remote sensing. This study highlights the impact of discretization effects in DDM sampling on land applications. The discretization effect in the Doppler dimension is first evaluated by comparing simulated and observed DDM slices at the Doppler bin corresponding to the DDM peak. The results indicate that the noise in DDM observations can be approximated as additive thermal noise. Based on an ideal autocorrelation function template, a matched filtering analysis is then applied to estimate the optimized specular point delay and reconstruct the peak power. Using multi-constellation observations from Tianmu-1, the results show that the original DDM peak delay exhibits a systematic delay relative to the optimized specular point delay, with biases of approximately 0.02 chips for GPS and GLONASS, and 0.17 chips for BDS (BeiDou) and Galileo. For BOC(1,1) signals in BDS and Galileo, the reflectivity remains underestimated by ~1.4 dB even at a delay sampling interval of 1/8 chip. The results indicate that under coherent scattering conditions over land, direct use of the DDM peak leads to underestimation of reflectivity due to discretization. The correction proposed in this study reduces the relative differences in reflectivity observations among the four GNSS systems. This study suggests that peak under-sampling should be considered in GNSS-R applications, and higher delay sampling resolution is required for land observations.

Remote Sensing doi: 10.3390/rs18101633

Authors: Xiaoyu Ma Xin Su Yingshuang Li Yihong Yang

The Polarization and Directionality of the Earth’s Reflectances (POLDER)-3/GRASP (Generalized Retrieval of Aerosol and Surface Properties) aerosol products have been widely used in studies on radiative balance and climate change. However, the stability and consistency of the products have yet to be comprehensively evaluated, despite their critical importance for long-term studies. POLDER-3/GRASP products mainly consist of three variants: High-Precision (HP), Components, and Models. This study aims to evaluate the accuracy, stability, and consistency of these aerosol products at global and regional scales, and to provide usage recommendations. Compared with AERONET observations, the Components product shows the best performance for both aerosol optical depth (AOD) and Ångström Exponent (AE) retrievals, with Root Mean Square Error (RMSE) of 0.114 for AOD and 0.319 for AE. The Models AOD and HP AE also demonstrate relatively high validation accuracy, with RMSE of 0.138 for Models AOD and 0.366 for HP AE. Regionally, Components AOD and AE outperform those from the HP and Models products in 8 out of 10 regions. Stability evaluation shows that the stability metrics of the three AOD products range from 0.034 to 0.036 per decade, and none of them meet the Global Climate Observing System (GCOS) stability requirement (i.e., 0.02 per decade), which indicates that caution should be exercised when using POLDER-3/GRASP products for long-term analysis. In terms of consistency, Components AOD and Models AOD exhibit high agreement, while HP AOD is systematically higher than them. The AE retrieved by the three products shows considerable discrepancies, highlighting uncertainties in AE and spectral-AOD retrievals and pointing toward directions for future algorithmic improvements. In summary, considering global and regional accuracy, stability, and consistency, the Components AOD and AE products are generally recommended for use. For different regions, users can choose the appropriate product based on detailed validation and intercomparison results.

Remote Sensing doi: 10.3390/rs18101632

Authors: Jianfeng Liu Yibo Du Lifan Sun Xiaozheng Li Yanna Si Xiaoli Song Ruijuan Zheng

Remote sensing scene classification (RSSC) faces severe challenges from data scarcity and complex background clutter. To overcome these limitations, this paper draws inspiration from few-shot fine-grained image classification (FSFGIC) to filter noise and capture subtle details. However, existing methods often process global context and local features separately, which limits their ability to suppress background noise in complex scenes. Consequently, the Guided Relational Cross-Attention Dual-branch Network (GRCD-Net) is proposed. Its core Guided Relational Cross-Attention (GRC) block leverages global semantics to filter local background noise prior to bidirectional feature interaction. Additionally, Iterative Global Relation (IGR) and Patch-level Dual-Metric (PDM) modules are integrated to robustly refine global relations and capture local similarities. Extensive experiments demonstrate that GRCD-Net consistently outperforms baselines by 2–4% on standard FSFGIC benchmarks. Notably, on the challenging NWPU-RESISC45 RSSC dataset, it achieves an 81.39% one-shot accuracy and exceeds current state-of-the-art methods by 7.55%, validating its efficacy for complex Earth observation.

Remote Sensing doi: 10.3390/rs18101631

Authors: Huang Liu

Polarimetric Synthetic Aperture Radar (PolSAR) has significant advantages for land-cover classification for its all-weather, day-and-night, and multi-polarization observation capability. Traditional methods often exhibit limited classification accuracy in regions with strong noise and complex textures. Although deep learning methods can improve classification performance, they usually suffer from high model complexity, while lightweight models often show insufficient spatial consistency. To address these issues, this study proposes a PolSAR land-cover classification framework that integrates a Lightweight Dynamic Sequential Axial Network (LiteDSANet) with a polarization feature-guided Dense Conditional Random Field (PFG-DenseCRF). LiteDSANet is employed to generate the initial class probability map, and PFG-DenseCRF optimizes the classification results by introducing polarimetric features. Experiments were conducted on AIRSAR L-band and RADARSAT-2 C-band datasets from the San Francisco Bay and Flevoland regions, covering agricultural, urban, and natural land-cover scenes. The results show that the proposed method improves classification accuracy by 2.14~15.36% compared with other methods, while achieving a favorable balance between accuracy and computational efficiency. These results demonstrate the effectiveness of the proposed method for PolSAR land-cover classification in different regional environments.

Remote Sensing doi: 10.3390/rs18101630

Authors: Jifei Fang Dali Zhu

Cross-modal image–text retrieval enables searching and retrieving of semantically relevant data across heterogeneous modalities, acting as a pivotal technology for interpreting massive remote sensing (RS) data. Despite recent progress, most existing methods in remote sensing cross-modal image–text retrieval (RSCIR) rely on high-dimensional real-valued embeddings, which suffer from excessive storage overhead and slow retrieval speeds, severely limiting their scalability in real-world applications. Conversely, while hashing techniques offer efficiency, traditional methods often fail to preserve the fine-grained semantic consistency required for complex RS scenes, leading to significant performance degradation. To bridge this gap, this paper proposes a novel framework named ConHash (Cross-modal Contrastive Hashing), which transfers the discriminative power of pre-trained vision–language models into a compact binary Hamming space. Specifically, ConHash comprises three synergistic components: (1) a hash module designed to project continuous embeddings into a latent discrete space while reducing information loss; (2) a hash-aware contrastive constraint that enforces cross-modal alignment directly in the hash space; and (3) a collaborative hybrid optimization strategy that jointly constrains real-valued embeddings and hash representations. Extensive experiments on RSICD and RSITMD demonstrate that ConHash achieves a favorable balance between accuracy and efficiency. Using 512-bit hash codes with L1 quantization loss as the main configuration, ConHash achieves mR values of 21.69% and 35.79% on RSICD and RSITMD, respectively. It also provides up to 3.50× retrieval speedup and a 32× theoretical storage reduction compared with 512-dimensional float32 embeddings, making it suitable for scalable remote sensing retrieval applications.

Remote Sensing doi: 10.3390/rs18101629

Authors: Xiaofei Yang Yao Wei Jiarong Tan Shuqi Li Haojin Tang Waixi Liu

Deep learning methods have significantly improved hyperspectral image (HSI) classification by exploiting hierarchical feature learning to integrate spatial and spectral information, thus significantly improving classification accuracy. Nevertheless, current deep learning approaches (such as CNNs, Transformers and Mamba) still face three major challenges: inadequate mitigation of spectral redundancy, high computational costs associated with global modeling, and the loss of two-dimensional spatial structure during sequential processing. To address these issues, we propose LMamba, a task-oriented hybrid framework that combines multi-scale convolutional filtering with local-context-conditioned state space modeling for hyperspectral image classification. Rather than introducing a fundamentally new SSM formulation, LMamba focuses on adapting the input-dependent parameter projection of Mamba to HSI data by injecting local 2D neighborhood context into the generation of selective SSM parameters. This design enables the state space module to better preserve spatial continuity while maintaining linear-complexity sequence modeling. The framework consists of two core components. First, the Multi-scale Aggregation and Compression Block (MACB) employs parallel grouped convolutions with varying kernel sizes to capture spatial features at multiple scales while simultaneously reducing spectral redundancy through channel compression. Second, the Locally Guided 2D Scanning Mechanism replaces conventional unidirectional 1D scanning with a context-aware 2D scanning strategy, thereby preserving structural continuity and enhancing feature representation by integrating local neighborhood spatial information into state transitions. Validation on three prominent HSI datasets demonstrates that LMamba consistently outperforms state-of-the-art methods based on CNNs, Transformers, and SSMs as measured by overall accuracy (OA), average accuracy (AA), and the Kappa coefficient. In summary, LMamba provides an efficient and accurate HSI classification framework under the considered benchmark settings, and its compact complexity and low-sample robustness suggest potential usefulness for practical HSI analysis.

Remote Sensing doi: 10.3390/rs18101628

Authors: Lifeng Yu Yifan Liu Hongmin Gao

Hyperspectral anomaly detection (HAD) aims to identify rare targets without relying on prior target knowledge. However, background spectra in hyperspectral images often lie on highly complex and nonlinear manifolds, making accurate modeling challenging. Although models with strong nonlinear approximation capabilities, such as Kolmogorov–Arnold Networks (KANs), provide a promising solution for capturing such complexity, self-supervised reconstruction-based HAD methods still suffer from a fundamental issue known as anomaly leakage. When the model has high representation capacity, anomalous signatures tend to be partially reconstructed, which reduces residual contrast and degrades detection performance. To address this issue, we propose a Blind-Spot KAN-based background reconstruction network with prior purification (BKP-Net), which mitigates anomaly leakage from both data and model perspectives. Specifically, we first introduce a Background Prior Purification (BPP) module to construct a cleaner background prior. This module suppresses and replaces potential outlier pixels through spatial clustering and robust weighted mean estimation. We then design a Blind-Spot KAN-based Reconstruction backbone (BKCN) to model complex nonlinear background characteristics while preventing direct information flow from the center pixel, thereby reducing anomaly leakage during reconstruction. In addition, separable convolutions are employed to enhance spatial–spectral feature representation, followed by an attention-guided fusion mechanism to suppress cross-domain interference. Furthermore, a band-wise Guided Reconstruction Refinement (GRR) strategy is introduced in the detection phase to improve structural consistency between the reconstructed background and the original hyperspectral image, leading to more reliable anomaly discrimination. Experimental results on four hyperspectral datasets demonstrate that the proposed method achieves competitive performance compared with several representative traditional and deep learning-based detectors.

Remote Sensing doi: 10.3390/rs18101627

Authors: Wenqiang Zhou Shiwen Deng Shuying Zang Dianfan Guo

Accurate estimation of tree Above-Ground Biomass (AGB) is critical for the timely monitoring of forest dynamics. However, the scarcity of high-quality in situ measured data introduces considerable uncertainty into the precise spatial mapping of AGB. In this study, a novel method for AGB mapping is proposed to enhance the accuracy of AGB estimation based on the Semi-Supervised Ensemble Learning (SSEL) strategy. By expanding the sample set via an iterative self-training approach based on an Inverted Query-by-Committee (I-QBC) strategy, the model significantly enhances the accuracy of AGB estimation. Using Sentinel-2 data, the experimental results show that: (1) The I-QBC-driven SSEL model demonstrated significantly higher estimation accuracy for AGB compared to conventional tree-based ensemble models. Optimal stability (R2 = 0.80) and peak accuracy (R2 = 0.88) were achieved at sample increments of 20 and 30, respectively. (2) Among various feature types, Recursive Feature Elimination with Cross-Validation (RFECV) identified GNDVI, PSSRa, slope and texture correlation as the most critical predictors for AGB estimation in the study area. (3) The total AGB stock in the study area is estimated to range from 1.46 × 107 Mg to 1.71 × 107 Mg. The SSEL model provides a valuable reference for AGB estimation under sparse ground-truth sample conditions, while offering a novel approach for large-scale AGB mapping.

Remote Sensing doi: 10.3390/rs18101626

Authors: Miaomiao Zhang Quan Wang Wuxia Zhang Xiangpeng Chen Jiaxin Pan Huinan Guo

Remote sensing image super-resolution (RSISR) has gained significant attention in recent years due to its critical role in enhancing image analysis capabilities. While existing methods often focus on nonlocal feature extraction, they frequently overlook the importance of local information integration. Moreover, many methods reconstruct images by introducing more complex structures, which poses a challenge to resource-limited devices. To address these issues, we present a local–nonlocal dual-branch feature complementary fusion network (DFCFNet) featuring two key components: a lightweight dual-branch feature aggregation (DBFA) module and an Efficient Feed-Forward Network (EFFN). The DBFA employs a dual-branch structure comprising a Focused Local Feature Branch (FLFB) with novel Partial Convolution Channel Mixers for localized pattern modeling and a Non-Focal Exploration Branch (NFEB) utilizing global variance analysis for comprehensive feature extraction. This dual-branch design enables simultaneous capture of local and global contextual information. The EFFN is designed to further refine the features of the DBFA output in order to make full use of the detailed information of the image. Extensive experimental results show that the proposed DFCFNet reconstructs optimally on remote sensing datasets and is also optimal in terms of computational efficiency and network complexity. The framework’s versatility is further confirmed through successful adaptation to natural image SR tasks, showing consistent performance improvements across five standard datasets.

Remote Sensing doi: 10.3390/rs18101625

Authors: Jiazhen Dong Jun Hu Rong Gui Yibo Yuan Yuanjun Qin Zhiwei Mo

Digital terrain models (DTMs) are required in many hydrologic, geomorphic, and ecological applications, yet widely used global elevation products often retain above-ground elevation contributions, particularly from vegetation canopies. This study investigates whether useful bare-earth terrain can be reconstructed from DSM-derived information alone at inference time. Rather than regressing terrain elevation directly, the proposed framework predicts the residual DH=DSM−DTM and reconstructs the DTM by subtraction. The model uses Copernicus DEM GLO-30 as the input source and augments it with CSF-derived priors and DSM-derived terrain features, including slope, aspect encoding, curvature, and local relief. Unlike multi-source terrain correction products that rely on external auxiliary datasets, all inference-time inputs in the proposed framework are generated from the DSM itself. A residual U-Net is trained with a weighted Huber loss together with gradient-consistency and DTM-slope-consistency constraints. Experiments across multiple regions in the central and southeastern United States show that the proposed method outperforms the compared public DEM products and baseline methods under a unified evaluation protocol. Relative to FathomDEM, it reduces the mean absolute error from 1.0445 m to 0.8538 m and the root mean square error from 1.6969 m to 1.4697 m on the study region test split, while also improving NMAD, P99, and Recall@5m. Performance on the geographically separate Arkansas region is similar to that on the in-region test split. Remaining errors are concentrated mainly in extremely steep terrain, densely vegetated areas, and cases with large residual heights.

Remote Sensing doi: 10.3390/rs18101624

Authors: Yichang Wang Yanjun Wang Cheng Wang Andrei Materukhin Xuchao Tang

LiDAR point clouds provide accurate and direct representations of spatial locations and geometric structures of objects in 3D space, making them essential for applications such as target recognition in autonomous driving and 3D reconstruction in smart cities. However, large-scale point clouds pose challenges, including massive data volume, uneven density distribution, and complex object structures. Existing point-based and graph-based semantic segmentation networks often suffer from limitations such as loss of local contextual information, over-reliance on local graph construction, and insufficient modeling of relationships between neighboring points. To address these issues, we propose PG-Net, a novel network that integrates discrete point distribution features with local graph structural information. The framework includes: (1) a point branch equipped with a Local Adaptive Feature Augmentation (LAFA) module to extract efficient local features; (2) a graph branch featuring a Dynamic Graph Feature Aggregation (DGFA) module, which explicitly models relationships among points in local graphs and adaptively balances a point’s intrinsic features with its neighborhood context; and (3) fuses local features from both branches, allowing their complementary strengths to enhance feature representation, a process further promoted by a New Aggregation Loss Function. Experiments on the Toronto3D and S3DIS datasets show that PG-Net achieves overall accuracy (OA) of 97.69% and 89.87%, and mean Intersection-over-Union (mIoU) of 83.51% and 73.22%, respectively. Comparative and ablation studies against advanced methods such as RandLA-Net, BAAF-Net, and LACV-Net demonstrate the effectiveness and robustness of our approach. By jointly exploiting discrete point distribution and local graph structural relationships, PG-Net effectively leverages the complementary strengths of its dual-branch design, offering a reliable solution for efficient and accurate large-scale point cloud semantic segmentation.

Remote Sensing doi: 10.3390/rs18101623

Authors: Xueqi Wu Zhongzhen Sun Han Wu Kefeng Ji

In video synthetic aperture radar (Video SAR), target motion causes defocusing, making it impossible to determine the target’s real-time position using reflected echoes. However, the shadows formed by the target occluding ground reflections can accurately characterize the target’s real-time position. To address challenges such as varying shadow scales, low contrast with the moving background, and susceptibility to clutter interference, this paper proposes a shadow detection network called DSENet to enhance the detail and semantic features of shadows. First, to enhance shadow features and reduce sampling loss during backbone network feature extraction, we design a detailed information enhancement (DIE) module to achieve lossless downsampling and effectively preserve the detailed features of the shadowed target. Second, we propose a semantic spatial feature aggregation (SSFA) module to enhance global semantic space feature extraction, improve the contextual feature representation of the target’s shadow region, and provide robust semantic space prior information for the model. Finally, we designed a detailed semantic fusion (DSF) module to improve the neck network’s ability to fuse shadow details and semantic features in video SAR images, further enhancing the model’s localization performance for target shadow features and achieving accurate localization of moving targets in video SAR. Comparative and ablation experiments validate the effectiveness and superiority of the proposed method. Experimental results on the Sandia National Laboratories (SNL) public dataset demonstrate that DSENet is efficient and performs excellently, achieving a P of 92.4% and an F1 score of 83.1%.

Remote Sensing doi: 10.3390/rs18101622

Authors: Yijing Cao Yongqiang Zhang Yuyin Chen Xuanze Zhang Jing Tian Xuening Yang Qi Huang Jianzhong Su

Groundwater storage is vital for managing water resources, especially as global water scarcity intensifies. Estimating groundwater levels regionally is challenging due to natural heterogeneity. We employed a large groundwater observation sample, along with Global Land Data Assimilation System (GLDAS) and Gravity Recovery and Climate Experiments (GRACE) datasets, to develop a random forest model for predicting groundwater levels in China’s Yellow River Basin. The model showed robustness, achieving an R2 of 0.95 in calibration and an R2 of 0.91 ± 0.009 in 10-fold cross-validation with 100 repetitions. Temporal predictability was lower, with an R2 of 0.72 for April–May 2023; however, the temporal prediction is preliminary and limited by the short validation period (April–May 2023), which should be interpreted with caution. Spatial maps revealed significant seasonal declines in fall and winter, particularly in the middle and lower reaches. This study highlights the potential of machine learning with extensive observations to estimate regional groundwater levels and supports groundwater analysis with robust data.

Remote Sensing doi: 10.3390/rs18101617

Authors: Yi Zhang Dongyan Wang Hong Li

Elucidating the mechanisms by which cropland fragmentation impacts production and ecological functions is critical for ensuring food security and ecological sustainability. Using Jilin Province as a case study, this research develops a cropland fragmentation evaluation framework based on landscape pattern indices. A restricted cubic spline model is employed to quantify nonlinear relationships and identify critical thresholds between fragmentation and both production and ecological functions. Furthermore, the PLUS model is utilized to simulate land-use patterns for 2030 under three scenarios: natural development, cropland protection, and ecological protection. The primary findings are as follows: (1) From 2000 to 2023, cropland fragmentation displayed pronounced spatial heterogeneity. Fragmentation was consistently high in the eastern mountainous areas and showed significant spatial clustering; the central region maintained relatively contiguous cropland, while the western region exhibited marked spatial variability. (2) Cropland fragmentation exhibits a nonlinear negative correlation with production functions, wherein the marginal negative impact attenuates beyond a threshold of 0.340. Conversely, its association with ecological functions follows a U-shaped trajectory, with a critical inflection point at 0.363 marking a directional shift in the fragmentation–ecology nexus. (3) Based on these nonlinear thresholds, the study area was delineated into production-ecology synergy zones, dysfunctional sensitive zones, and ecosystem landscape trade-off zones. Specifically, the central agricultural core is characterized by functional synergy; the ecologically fragile western zone resides near the nadir of the U-shaped curve, rendering its balance between production and ecological functions highly vulnerable to shifts in development intensity; and the eastern ecological barrier zone manifests a distinct trade-off prioritizing ecological functions. (4) Multi-scenario simulations reveal that the natural development scenario exacerbates the expansion risk of dysfunctional sensitive zones. While the cropland protection scenario enhances production capacity, it concurrently introduces risks of ecological instability. Conversely, the ecological protection scenario effectively steers sensitive zones toward ecological recovery. Consequently, we propose a differentiated spatial regulation strategy: prioritizing land consolidation in the central region, integrating ecological restoration with capacity enhancement in the west, and sustaining ecological barriers in the east, thereby fostering sustainable regional development.

Remote Sensing doi: 10.3390/rs18101621

Authors: Hanhu Liu Xueliang Huang Wei Wang

This study uses Gonjo County, Chamdo City, Tibet, as the study area and addresses the challenges of lithological complexity and low efficiency of conventional geological surveys in the Qinghai–Tibet Plateau. This study applies the first systematic application of Chinese GF-5 AHSI data to conduct detailed lithological classification in a plateau environment. Three types of datasets were constructed, including the full-band (FB) dataset, shortwave infrared diagnostic bands (SWIR), and feature-selected bands (FS). Four classification models—Support Vector Machine (SVM), Long Short-Term Memory network (LSTM), Multi-Scale Convolutional Neural Network (MSCNN), and Spectral-Spatial Unified Network (SSUN)—were comparatively evaluated to systematically assess the performance of spectral feature selection and deep learning methods for hyperspectral lithological classification. The experimental results explicitly demonstrate the superiority of spectral-spatial feature extraction. Specifically, compared to the baseline Support Vector Machine (SVM) model, which achieved an overall accuracy of 74.67% and a kappa coefficient of 0.6952, the proposed SSUN model demonstrated an advantage, reaching an overall accuracy of 90.94% and a kappa coefficient of 0.8917. By jointly extracting spectral sequence features and spatial contextual information, SSUN effectively suppresses noise and enhances the spatial continuity of lithological boundaries. The results demonstrate the high practical applicability and spectral fidelity of GF-5 AHSI data for lithological identification in plateau stratigraphic environments. The shortwave infrared region is confirmed to be a critical spectral domain for lithological discrimination, and spectral-spatial deep learning models can maintain high classification accuracy after feature dimensionality reduction, achieving a balance between classification efficiency and accuracy. This study provides reliable methodological support for remote sensing lithological mapping and mineral resource exploration in complex plateau geological environments.

Remote Sensing doi: 10.3390/rs18101620

Authors: Yi Yang Rongling Ye Xuewei Yin Honglin Tian Zhuang Feng Yang Zhang Jin Yang Xiaochun Zhang Xin Dong Ryosuke Tajima

Maize is a globally important staple crop, and automated monitoring of germination and seedling emergence is essential for precision agriculture, enabling timely reseeding and reducing potential yield loss. To address this need, we propose Seedling-DETR, a transformer-based model for the real-time detection of emerged and missing maize seedlings using multispectral UAV imagery in an end-to-end manner. First, we construct a multispectral UAV dataset and ntroduce a dedicated annotation strategy in which missing seedlings were labeled individually rather than inferred indirectly. Then, we modify the feature fusion module of RT-DETR and develop a hybrid-scale feature fusion module to obtain richer and more expressive feature representations for missing seedling detection and improve the precision of missing seedling detection. Finally, we propose a channel fusion module to incorporate multispectral images into our model without requiring a dedicated multispectral backbone or additional pretraining, thereby improving model adaptability. The results show that, under a random train–test split (8:2), when using RGB images as input, our Seedling-DETR achieves a mean average precision (mAP) of 83.1% at an IoU threshold of 0.5, outperforming YOLOv11x and RT-DETR by 2.5% and 1.1%, respectively. The proposed method achieves an AP of 69.3% at an IoU threshold of 0.5 for missing seedling detection, which increases to 71.7% when multispectral inputs are incorporated. Similar performance trends are observed on an independent validation set collected on a different date. Although the model introduces moderate computational overhead (282 GFLOPs for RGB input and 418 GFLOPs for multispectral configuration, with 84.0 M and 85.1 M parameters, respectively), it can maintain efficient detection performance suitable for actual agricultural field deployment. The method is further validated at the field scale using orthomosaic-based analysis. Overall, this study provides an effective and scalable framework for the detection of emerged and missing maize seedlings under complex field conditions. The proposed framework supports accurate reseeding decisions, and contributes to automated maize production.

Remote Sensing doi: 10.3390/rs18101619

Authors: Jing Guo Ziti Jiao Lei Cui Zhilong Li Chenxia Wang Fangwen Yang Ge Gao Zheyou Tan Sizhe Chen Xin Dong

Understanding the variations in the bidirectional reflectance distribution function (BRDF) and albedo over snow surface under various conditions is important for interpreting the surface–atmosphere processes of the cryosphere, and the kernel-driven model is among the most popular methods to obtain this information for a comprehensive analysis. Recently, the RossThick-LiSparseReciprocal-Snow (RTLSRS) model was developed to better characterize the anisotropic reflectance of snow and shows strong potential for integration into operational remote sensing algorithms for snow BRDF/albedo retrieval. To comprehensively test the ability of the RTLSRS model to reproduce snow reflectance, the fitting accuracy to different multi-angular data derived from ground, tower, aircraft, and satellite platforms across the full optical wavelength range were demonstrated in this study. Special attention in this study was directed to analyzing the model performance under extreme illumination observation geometries, particularly with respect to the retrieval accuracy and stability under large Solar Zenith Angles (SZAs) and different Relative Azimuth Angles (RAAs). The model performance for silt-polluted snow surface with different concentrations is also assessed to provide necessary supplementation, relative to “pure” snow surface in the previous study. The main findings of this study are summarized as follows: (1) The RTLSRS model exhibits strong robustness under various SZAs; even when the SZA exceeds 80°, the model maintains high accuracy in BRDF reconstruction, with root mean square error (RMSE) values below 0.05. (2) The model also demonstrates satisfactory inversion capability when observations deviate from the principal plane (PP); the model can achieve fitting accuracy with R2 approaching 0.5 and RMSE below 0.05 for MODIS data. (3) In the spectral range below 1300 nm, the RTLSRS model effectively reconstructs the scattering characteristics of snow surfaces with light impurity levels (<20 g/0.5 m2). (4) The spectral shape of snow reflectance remains consistent across different view zenith angles (VZAs) in general. However, the variations caused by different SZAs can be as high as 38.49% and such SZA-induced difference can result in WSA estimation discrepancy of up to 63.43%. This comprehensive assessment further affirms and demonstrates the applicability of the RTLSRS model for the first time in fitting observations across different platforms with various optical wavelengths and geometries, and provides an improved understanding to analyze BRDF variations for the user community.

Remote Sensing doi: 10.3390/rs18101618

Authors: Yao Zhu

The Pamir–Tian Shan domain constitutes one of the most actively deforming intracontinental orogenic systems associated with continued India–Eurasia convergence. Characterizing present-day deformation in this region is fundamental to deciphering its geodynamic evolution and assessing seismic risk. Existing strain rate models based on GNSS measurements display noticeable discrepancies, largely attributable to variations in analytical strategies and uneven station distribution. In this study, we determine the present crustal strain and rotation fields across the Pamir–Tian Shan area using the most updated GNSS velocity solution referenced to stable Eurasia. To address the issues of inconsistent strain rate field results and lack of reliability verification in previous studies based on GNSS data, this paper computes the crustal strain rate field (principal strain rate, maximum shear strain rate, dilatation strain rate, and rotational strain rate) with a grid spacing of 0.75° × 0.75° in the study area, followed by numerical validation of the results’ reliability. The derived strain field is characterized by dominant NNW–SSE shortening throughout much of the orogenic system, with peak compressional strain rates (~1.0 × 10−7 yr−1) concentrated along the Pamir Frontal Thrust. By contrast, the interior of the Pamir Plateau exhibits clear EW extension, consistent with areas affected by normal-faulting earthquakes. High values of shear strain rates are primarily localized along major active fault systems, whereas negative dilatational components indicate overall contraction within the Tian Shan. The rotation-rate distribution reveals clockwise rotation of the Tarim Basin (approximately 0.6°/Myr) together with counterclockwise rotation affecting the Pamir and Tian Shan blocks, accommodated by prominent strike–slip fault networks. The close spatial agreement between the modeled strain patterns, active tectonic structures, and focal mechanism solutions supports the reliability of the inferred deformation field. The research results of this paper are of great scientific significance for in-depth study of the tectonic evolution and earthquake disaster assessment in the Pamir–Tian Shan region.

Remote Sensing doi: 10.3390/rs18101616

Authors: Caiyun Zhang Jing Guo Shuangcheng Jiang Lingling Li Miaofeng Yang

Nearshore mariculture is a major contributor to the supply of “blue food”; however, its rapid expansion in bay systems has intensified sea-space competition and environmental pressures, underscoring the need for accurate and long-term monitoring. This study used multitemporal Sentinel-2 imagery processed using Google Earth Engine (GEE) to develop an automated identification framework for raft and cage aquaculture along the coast of Fujian, China, from 2017 to 2024. Three widely used classifiers—U-Net, DeepLabV3+, and random forest (RF)—were comparatively evaluated. Of these methods, U-Net had the most stable overall performance under optically complex nearshore conditions and was, therefore, used for province-scale mapping. Based on the U-Net-derived maps, the spatiotemporal evolution of mariculture was quantified. The results showed that mariculture in Fujian exhibited a persistent bay-oriented, dual-core clustering pattern, with major hotspots concentrated in Ningde and Zhangzhou. In the 2024 winter–summer comparison, raft aquaculture displayed a clear seasonal contrast, characterized by expansion in winter and contraction in summer, whereas cage aquaculture showed relatively smaller seasonal variation. Interannually, the mariculture system shifted from a mixed cage–raft configuration toward the dominance of raft aquaculture, accompanied by a spatial redistribution of mapped aquaculture density from inner nearshore waters toward bay mouths and more open waters. Overall, in this study, we demonstrate the potential of deep learning-enabled Sentinel-2 remote sensing for monitoring nearshore mariculture structures and provide mode-specific observational evidence for marine spatial planning, environmental risk management, and sustainable mariculture development in nearshore waters and semi-enclosed bay systems.

Remote Sensing doi: 10.3390/rs18101614

Authors: Kangli Xie Jun Lin Hao Zhang Lanlan Fan Zunjian Bian Hua Li Yongming Du

Error in Table [...]

Remote Sensing doi: 10.3390/rs18101615

Authors: Xin Guo Shengpei Dai Hongxia Luo Wujun Lv Shanshan Jiang Yuhao Yang Yi Yang

Vegetation serves as a critical ecological barrier on tropical islands, but conventional assessments often emphasize greening magnitude while overlooking persistence, structural quality, and scale-dependent attribution. In this study, we reconstructed a high-precision fractional vegetation cover (FVC) dataset for Hainan Island, China, covering the period from 2000 to 2024 using Google Earth Engine (GEE). We then combined trend analysis, emerging hot spot analysis (EHSA), the coupling coordination degree model (CCDM), RESTREND, and Geodetector to examine vegetation change from complementary perspectives. The results show that FVC increased overall and gradually shifted toward a more stable state over time. EHSA further revealed a distinct core-periphery pattern, with persistent hot spots concentrated in the central mountainous region, persistent cold spots distributed along the coastal urban belt, and oscillating hot spots occurring within agricultural transition zones. Regarding quantity-structure coupling, FVC and the aggregation index (AI) generally improved together across the island; however, some agricultural ecotones exhibited weaker structural improvement despite increasing vegetation cover, suggesting potential risks of homogenization and structural simplification. In the broad attribution analysis, vegetation recovery was primarily associated with the combined influence of climatic and human-related improvement. In the factor-specific analysis, land cover and slope showed the strongest explanatory power, and their interactions with other variables further enhanced spatial differentiation. These results demonstrate that greening magnitude alone is insufficient for evaluating vegetation change on tropical islands. Structural coordination and scale-dependent attribution should also be considered when interpreting ecological improvement and related management implications.

Remote Sensing doi: 10.3390/rs18101613

Authors: Xuan Zhou Xuefeng Wei Zhi Qu Yusuke Sakai Hidetaka Kamigaito Taro Watanabe

Vision–language models (VLMs) hold considerable potential for interpreting large-scale remote sensing (RS) archives, which are critical for applications such as environmental monitoring, disaster response, and urban planning. However, general-purpose VLMs primarily target optical imagery and often underperform on RS tasks, while existing RS-specific VLMs still struggle with fine-grained understanding. To address these limitations, we propose GeoPilot, a tool-augmented multimodal assistant tailored for RS scenarios. GeoPilot interprets user instructions, autonomously determines whether to invoke external tools, and synthesizes their outputs to generate precise responses. A key capability of our approach is its ability to process both optical and Synthetic Aperture Radar (SAR) imagery, supporting representative tasks such as visual grounding, object detection, segmentation, and cross-domain reasoning. To support this setting, we construct a novel large-scale RS instruction dataset that jointly supports optical and SAR imagery together with explicit tool use reasoning traces, addressing the critical challenge of task-specific data scarcity. We also introduce GeoPilotBench, a benchmark for cross-domain, multi-task dialogue and tool-aware evaluation in RS, and use it to assess GeoPilot across representative tasks. Experimental results show that GeoPilot achieves strong task planning accuracy (92.6% overall planning accuracy) and competitive performance on VQA, SAR understanding, and referring object detection. End-to-end evaluation further confirms that GeoPilot’s learned tool policy introduces only limited overhead compared to standalone tool execution, demonstrating its practical value as a tool-augmented RS assistant.

Remote Sensing doi: 10.3390/rs18101612

Authors: Tianyi Wang Jian Chen Sheng Ma Weixu Yang Na Zhang Qiang Li Qiang Wu

Understanding the spatiotemporal evolution and driving mechanisms of soil salinization in the Yellow River Delta is a key research focus in the comprehensive utilization of saline–alkali land. Taking Zhanhua District as the study area, this study extracted soil salinization information using four remote sensing salinity index models (SDI1, SDI2, SDI3, SDI4). Model accuracy was evaluated, and the optimal model (SDI1, with an overall accuracy of 86.21%) was selected to analyze the spatiotemporal dynamics of soil salinization from 1993 to 2023. The XGBoost-SHAP framework was then applied to identify and interpret the driving factors of salinization. Furthermore, future soil salinization trends under climate change were projected based on four scenarios from the Sixth Coupled Model Intercomparison Project (CMIP6), including SSP1-2.6 (low forcing), SSP2-4.5 (medium forcing), SSP3-7.0 (medium-to-high-forcing), and SSP5-8.5 (high forcing). The results show the following: (1) Spatially, soil salinization in Zhanhua District exhibits a pattern of being “lighter in the south and heavier in the north.” Over the past 30 years, salinization has undergone a phased evolution characterized by a transition from “severe in the north and mild in the south” to “overall expansion” and finally to “improvement in the north and optimization in the south,” while the proportional structure of salinization severity levels has remained relatively stable. (2) Among the driving factors, evaporation is the dominant contributor (SHAP value = 0.3357), followed by precipitation (0.1732) and population density (0.1518). Soil moisture, land use, and temperature exert moderate influences, while nighttime light intensity, slope, and elevation contribute relatively less. Overall, soil salinization is jointly controlled by climatic factors and human–nature interactions. (3) Among the future climate scenarios, the SSP1-2.6 low-emission scenario exhibits the most pronounced mitigation trend, with a further reduction in salinization intensity projected by 2100. This study provides a scientific basis and data support for formulating soil salinization control and saline–alkali land management strategies in Zhanhua District and the Yellow River Delta.

Remote Sensing doi: 10.3390/rs18101611

Authors: Yongkang Yu Sijia Li Xingming Zheng Kai Li Jianhua Ren

Micro water bodies are essential to regional ecosystems but are difficult to extract from high-resolution remote sensing images due to fragmentation and building shadows. To address edge breakage and high false-alarm rates in existing semantic segmentation models, this study proposes MEF-TransUNet, an improved TransUNet-based model for fine micro water body extraction. The model integrates a multi-scale edge-guided attention module (MEGA), a high–low-frequency decomposition fusion module (HLFD), and a convolutional block attention module (CBAM). Specifically, MEGA extracts edge priors using a Laplacian pyramid to repair topological breaks in slender water bodies. HLFD uses frequency-domain decoupling to suppress high-frequency background noise and reduce confusion between water bodies and shadows. CBAM enhances channel and spatial feature attention. Experiments using PlanetScope images from the Songhuajiang River Basin in Daqing City of the Heilongjiang Province in China showed that MEF-TransUNet achieves 91.74% precision, a 90.07% F1-score, a recall of 90.22%, and a B-IoU of 43.88%. For the GID dataset, the model attains a precision of 91.85%, an F1-score of 91.48%, a recall of 92.01%, and a B-IoU of 55.42%. Its overall performance clearly outperforms DeepLabV3+, SegFormer, U-Net, AttenUNet, and UNet++, enabling accurate micro water body localization, high output purity, and reduced manual correction costs, thus supporting fine water resource management in complex surface environments.

Remote Sensing doi: 10.3390/rs18101610

Authors: Bowen Wang Liguo Liu Qingyi Zhang

Ship classification in synthetic aperture radar (SAR) imagery is essential for maritime surveillance but remains challenging due to limited resolution, insufficient textural details, and difficulties in effectively fusing multimodal information. Existing methods either rely on handcrafted features with limited adaptability or employ simplistic fusion strategies that fail to fully exploit the complementary guidance across modalities. To address these issues, we propose a multimodal fusion network based on a learnable feature preprocessing front-end (LFPF-MFN), which integrates polarimetric, textural, and geometric information in an end-to-end learnable manner. Specifically, LFPF-MFN introduces a learnable preprocessing front-end to embed scattering and enhanced textural features. Meanwhile, geometric information from the Automatic Identification System (AIS) is incorporated through textual embedding, and effective multimodal fusion is achieved via a bidirectional cross-attention mechanism. Extensive experiments on the OpenSARShip 2.0 dataset demonstrate that the proposed method achieves state-of-the-art performance in both three-class and six-class classification tasks, validating the effectiveness of each designed module and the superiority of the multimodal fusion strategy.

Remote Sensing doi: 10.3390/rs18101609

Authors: Thiago Lappicy Daniel Beltrão Luana Oliveira Sales Tati Almeida Guilherme Gomes Pessoa Saulo Souza Renato Prata de Moraes Frasson Rejane Ennes Cicerelli

High-quality water surface elevation (WSE) measurements are critical in hydrological applications, yet no systematic evaluation of the Surface Water and Ocean Topography (SWOT) mission exists for Brazil’s diverse lake systems, where satellite observations are essential given limited in situ monitoring. We evaluated WSE from SWOT over 132 Brazilian lakes, comparing LakeSP, Raster_250m, and Raster_100m products against field measurements over a 20-month period. The 68th percentile errors were under 29 cm for the full dataset, below 12 cm for Flag = 0, and below 21 cm for Flag = 1, indicating good agreement but also the presence of outliers and the need for data screening. A Random Forest analysis identified quality flags, lake geometry, and cross-track distance as key drivers of WSE precision. Flag = 0 is overly restrictive, retaining only 22% of observations, while Flag = 1 contains anomalous data. The SWOT Quality-Range Threshold for Lakes (SQRTL) filter combines Flag = 0 with cross-track constrained Flag = 1 observations. SQRTL more than triples data availability relative to Flag = 0, maintaining comparable precision (68th percentile below 16 cm) and reducing median revisit from 88–123 days to 16–18 days for raster products and from 25 to 14 days for LakeSP. These results provide the first large-scale SWOT WSE evaluation over Brazilian lakes and a transferable filtering framework applicable wherever SWOT and field observations overlap, with potential to extend monitoring to over 100,000 water bodies in the SWOT Prior Lake Database.

Remote Sensing doi: 10.3390/rs18101608

Authors: Pengchen Lei Xiaomeng Xin Xuena Qiu Wenli Huang Yang Wu Ye Deng

Remote sensing imagery is often degraded by cloud cover, causing severe information loss and hindering downstream Earth observation tasks. Although recent deep learning methods, including CNN- and Transformer-based models, have achieved promising progress in cloud removal, they mainly operate in the spatial domain and largely overlook the frequency-domain discrepancies introduced by clouds of different types and densities. This limitation restricts their ability to generalize across diverse cloud corruption scenarios. To address this issue, we propose a Frequency Interaction Cloud Removal Network (FI-CRNet), which introduces a novel Frequency-Aware Modulation (FAM) mechanism for high-fidelity cloud-free image reconstruction. The FAM module consists of two components. First, the Frequency Decomposition (FD) module explicitly separates input features into low-frequency cloud-affected components and high-frequency detail-rich components through spectral analysis, while aligning them with decoder features via cross-attention. Second, the Cross-Frequency Interaction (CFI) module adaptively integrates these components through a dual-gate weighting mechanism, including spatial and channel gates, to suppress cloud interference while enhancing structural and textural details. By jointly modeling frequency-domain cues and spatial features, FI-CRNet enables robust and adaptive reconstruction under diverse cloud conditions. Extensive experiments show that our method outperforms state-of-the-art techniques across diverse cloud scenarios.

Remote Sensing doi: 10.3390/rs18101603

Authors: Arthur Dias Freitas Daniel Constantino Zacharias Bruna Lüdtke Paim Agnès Borbon Adalgiza Fornaro

The Metropolitan Area of São Paulo (MASP), located in southeastern Brazil, faces significant air quality challenges due to its large vehicle fleet and complex fuel composition, including widespread ethanol use. Air pollution dynamics in this context are investigated, focusing on spatio-temporal variations in formaldehyde (HCHO) and nitrogen dioxide (NO2), and their role in ozone (O3) formation. High-resolution data from the TROPOspheric Monitoring Instrument (TROPOMI) on board the Sentinel-5 Precursor satellite are used to analyze HCHO and NO2 vertical column densities (VCDs) over a 5-year period (2019–2023). Results reveal high HCHO and NO2 VCDs over MASP, with spatial patterns related to land use and higher concentrations during the dry season, with HCHO mean VCD reaching 14.21 × 1015 molecules cm−2 and NO2 mean VCD reaching 8.91 × 1015 molecules cm−2. The Formaldehyde to Nitrogen dioxide Ratio (FNR) thresholds were derived based on observations from 24 CETESB surface O3 monitoring stations, providing region-specific constraints for O3 sensitivity classification in MASP, with lower and upper thresholds of 1.6 and 2.4. Based on these thresholds, the analysis indicates a predominance of VOC-sensitive conditions in the urban core, alongside transition and NOx-limited regimes in other areas.

Remote Sensing doi: 10.3390/rs18101605

Authors: Liu Guo Zhou Qiu Zhang Shan

Soil thermal conductivity (STC) is vital for environmental and engineering modeling, yet traditional unfrozen STC estimates often perform poorly under field conditions. This study develops an enhanced Johansen–Tarnawski model incorporating vegetation parameters (JT-V) and applies geospatial data for regional simulation. Residuals from mechanistic predictions were analyzed using Geodetector and Random Forest, revealing strong vegetation-type effects. Validation with 88 samples from 18 sites across five vegetation types showed the JT-V model significantly improved accuracy: R2 rose from 0.426 to 0.716, and RMSE decreased by 53%. The best performance occurred at the surface layer (RMSE = 0.074 W·m−1·K−1), with errors increasing with depth. Over 83% of sites achieved R2 > 0.7, and most linear regression slopes fell between 0.8 and 1.1. Applying JT-V to simulate thawing-season STC in Northeast China, it was found that lower values predominated in the Khingan Mountains and the Inner Mongolia Plateau, while higher values occurred across the Northeast Plain. Temporal dynamics exhibited three stages: stability (May–mid-July), rapid rise (mid-July–mid-August), and gradual decline (mid-August–September). The improved model advances regional land surface simulations and supports agricultural and engineering applications.

Remote Sensing doi: 10.3390/rs18101607

Authors: Jakub Miszczyszyn Piotr Wężyk Luiza Tymińska-Czabańska Jarosław Socha Marta Szostak

The aim of this study was to examine the potential of multispectral imaging derived from unmanned aerial vehicles (UAVs) for detecting the spread of mistletoe (Viscum album ssp. austriacum L.) in Scots pine stands and to assess the information potential of selected vegetation indices in mistletoe detection. UAV campaigns were performed in the Niepołomice Primeval Forest (Niepołomice Forest District, Regional Directorate of the Polish State Forests National Holding, Kraków, Poland). A fixed-wing UAV Trinity F90+ (Quantum Systems GmbH) equipped with a five-band multispectral MicaSense RedEdge-M camera and an RGB Sony UMC-R10C camera was employed. The number of trees infected by mistletoe, as well as the quantity and area of mistletoe biogroups, were derived based on the classification of true multispectral orthophotos using a support vector machine (SVM) classifier. The spectral information potential assessment identified NIR (B5) as the most important single spectral source of information, while the greatest information potential among vegetation indices was found in NormG, CIG, and GRVI. The mistletoe classification of the 22.5-ha compartment revealed 1735 mistletoe biogroups covering a total area of 489 m2, with 58.6% of the 2917 detected tree crowns identified as infected (Kappa = 0.74). The results confirm that UAV-based multispectral data, particularly when combined with green-sensitive vegetation indices, enable effective differentiation of mistletoe from host tree crowns. The integration of the near-infrared (NIR) band further enhanced classification performance. This study evaluates UAV-based multispectral and RGB imagery for detecting common mistletoe (Viscum album ssp. austriacum) in Scots pine stands. The information potential of 22 vegetation indices was assessed to identify the most effective spectral features for mistletoe classification.

Remote Sensing doi: 10.3390/rs18101606

Authors: Masoomeh Gomroki Amirreza Gomroki Robert H. Gulden Dilshan I. Benaragama Mahdi Hasanlou Nasem Badreldin Bahareh Kalantar Husam Al-Najjar

Change detection is a fundamental task in remote sensing with broad applications in urban monitoring, agriculture, watershed management, and land use and land cover analysis. In urban environments, accurate change detection is particularly critical for resource management, urban planning, and smart city development. Rapid urbanization has led to frequent and complex changes in buildings, which constitute key structural components of cities. Consequently, continuous and precise monitoring of building dynamics is essential for informed decision-making related to urban growth, environmental assessment, traffic management, and sustainable development. This paper presents a comprehensive review of two-dimensional (2D) and three-dimensional (3D) change detection methods applied to urban areas. Conventional and advanced approaches are systematically analyzed, and their strengths and limitations are critically discussed from a holistic perspective. Special emphasis is placed on recent learning-based techniques, which demonstrate enhanced robustness and accuracy in complex urban environments. Finally, current challenges and future research directions are identified to support the further development of effective 2D and 3D urban change detection methods.

Remote Sensing doi: 10.3390/rs18101604

Authors: Meizhen Xie Daosheng Chen Xiqing Sun Xiaoyan Cheng Huimin Wang Kehan Wang Weiqiang Li Hongwei Yu Jiahao Ma Xiaodong Yang

As an important component of forest ecosystem processes, leaf litterfall plays a key role in nutrient cycling and ecosystem functioning. However, monitoring litterfall dynamics in subtropical forests remains challenging due to complex community structures and asynchronous leaf phenology, which limit the applicability of remote sensing approaches developed for temperate forests. As a critical linkage between vegetation and soil carbon pools, leaf litterfall directly influences forest carbon sequestration by providing carbon inputs in the form of litter. Unlike the concentrated autumn leaf fall in temperate forests, subtropical forests exhibit complex community structures with concurrent leaf abscission and new leaf growth, limiting the applicability of temperate-focused remote sensing techniques. To address this, we collected annual leaf litterfall data from 18 plots in eastern China’s subtropical forests and integrated these with high-resolution Sentinel-2 imagery using supervised machine learning models to develop a novel monitoring method. Our results indicated that subtropical forests exhibited clear seasonal leaf litterfall peaks during spring, summer, and autumn. Sentinel-2 satellite imagery combined with supervised machine learning algorithms can effectively monitor forest leaf litterfall dynamics. Temporal models, which use multi-date monthly spectral differences (R2adj = 0.70, RMSE = 0.46, RPD = 1.86), significantly outperformed instantaneous models based on single-date canopy states (R2adj = 0.33, RMSE = 0.85, RPD = 1.24). Following variable selection, model performance improved, with R2 increasing by more than 2% in most models and the number of variables reduced by over 44%. Robustness analysis indicated that the model was spatially robust (no significant bias among sites), and despite seasonal intercept differences, the slopes were consistent, enabling reliable tracking of litterfall dynamics. Among the examined spectral indices and canopy characteristics, those reflecting canopy greenness, pigments, and structure contributed over 65%, with WV-VI, MCARI2, and LAI being most influential. Incorporating drought-sensitive water indices and soil exposure-related mineral indices further enhanced model performance. These indices may partially reflect drought stress or seasonal canopy opening. Our findings provide a new method for monitoring leaf litterfall dynamics in structurally complex subtropical forests and offer a critical theoretical basis for accurately assessing leaf fall dynamics. Our findings provide a novel and effective method for monitoring leaf litterfall dynamics in structurally complex subtropical forests, improving seasonal litterfall assessment and supporting vegetation monitoring, with potential implications for ecosystem- and carbon-related studies.

Remote Sensing doi: 10.3390/rs18101602

Authors: Yuanhui Zou Yundong Wu Jinhe Su Huilin Xu

Remote sensing provides vast unlabeled imagery at low cost, yet annotation remains expensive, making self-supervised learning (SSL) well suited to this domain. However, existing DINO-style SSL frameworks are not well suited to remote sensing imagery, where object extents vary substantially and standard multi-crop view generation often introduces cross-scale inconsistency. This issue is particularly severe for small objects and elongated structures, whose discriminative features can be lost under scale transformations. To address this limitation, we propose DINO-MS (DINO with multi-scale consistency), a scale-consistent SSL framework for remote sensing imagery. The key idea is to construct feature-aligned cross-scale local views and explicitly enforce prediction-level agreement among them. Specifically, DINO-MS first adopts a co-located multi-scale cropping strategy to sample local views from the same spatial location at different crop scales, and then introduces a local consistency loss that works jointly with the original DINO local-to-global objective. Extensive experiments on land-use classification and change detection benchmarks show that DINO-MS generally improves downstream transfer performance. Notably, on EuroSAT, it improves per-class accuracy from 80.60% to 87.80% for Highway and from 88.00% to 91.60% for River with DINO-MC, confirming its advantage for categories dominated by small objects.

Remote Sensing doi: 10.3390/rs18101601

Authors: Xiaofeng Li Xinxin Xie Yan Liu Yaqi Zhou Pablo Saavedra Garfias Yang Guo Jieying He

South China is characterized by abundant and complex precipitation, with frequent typhoons, heavy rainfall, and pronounced extreme events, making it an ideal region for precipitation microphysics research. This study uses rainfall observations from an OTT Parsivel2 (Parsivel) laser disdrometer and a Micro Rain Radar–2 (MRR–2) collected in Zhuhai during 2022–2023 to analyze the characteristics of stratiform rainfall (SR) and convective rainfall (CR). The results show that, although SR lasts longer, CR contributes much more to the total accumulated rainfall. In SR, samples with rain rate (RR) < 5 mm h−1 account for about 27% of occurrences and contribute less than 10% of total rainfall, whereas in CR, samples with RR > 8 mm h−1 represent only 7% of occurrences but contribute more than 45% of the accumulated rainfall. CR is characterized by a larger mass-weighted mean diameter (Dm), while SR shows a higher normalized intercept parameter (Nw). In SR, Dm increases with RR, whereas Nw changes little; in CR, both Dm and Nw increase with RR. Finally, by analyzing temporal/spatial collocated vertical rain profiles from MRR and Global Precipitation Measurement Dual-frequency Precipitation Radar (GPM DPR), the results show that CR exhibits larger RR, radar reflectivity and stronger vertical variability than SR, along with greater variations in Dm and log10(Nw). Ground-based MRR also provides an independent vertical reference for evaluating DPR-derived precipitation structure and interpreting the consistency and discrepancies between satellite and ground-based observations. Although the results are not conclusive due to a limited number of events, both instruments capture distinct microphysical characteristics in the analyzed SR and CR cases, despite differences in their retrieved vertical DSD structures.

Remote Sensing doi: 10.3390/rs18101599

Authors: Weichao Wu Shengyang Gu Yafei Wei Zhe Wang Yusong Qin Xiuqing Hu Yongmei Wang

We present the first results of polar mesospheric cloud (PMC) detection using ultraviolet observations from TROPOMI (TROPOspheric Monitoring Instrument). An improved retrieval algorithm, developed on the basis of the SBUV-type approach and adapted to TROPOMI UV1 (270–300 nm) measurements, combines spatial binning, iterative Rayleigh background modeling, and adaptive thresholding to extract PMC signals from the background atmosphere. The robustness of the TROPOMI retrievals is evaluated through multi-scale comparisons with PMC data from the Cloud Imaging and Particle Size experiment (CIPS) and the Ozone Mapping and Profiler Suite Nadir Profiler (OMPS-NP). Compared with CIPS, the two datasets show broadly consistent hemispheric-scale horizontal structures and a westward wave-like phase progression consistent with possible quasi-5-day planetary-wave modulation, despite local-time differences. Compared with OMPS-NP, residual albedo under matched spatiotemporal conditions shows strong agreement for bright PMCs, whereas differences in spatial resolution lead to discrepancies in the detection of faint clouds. Seasonal-scale comparisons of PMC occurrence frequency also show consistent variability among the datasets. These results demonstrate that TROPOMI can resolve PMC structures smaller than 250 km that are difficult to detect with current low-resolution instruments. TROPOMI therefore provides a bridge between long-term coarse-resolution records and high-resolution observations, offering valuable data for studies of mesospheric dynamics and climate change.

Remote Sensing doi: 10.3390/rs18101600

Authors: Teng Li Laide Guo Junchang Xin Hongfei Yu Bowen Li

Semantic segmentation of high-resolution remote sensing images constitutes an important foundation for urban mapping and land-cover interpretation. However, objects in remote sensing scenes usually exhibit large-scale variations, significant intra-class differences, and complex background interference. Due to these factors, existing methods for complex high-resolution scenes still suffer from insufficient global semantic modeling, boundary blurring, and small-object omission. To address the above challenges, this paper proposes a Global Category-Center Prior-Guided Spatial-Frequency Collaborative Network (GC2F-Net). Specifically, ResNet-50 is adopted as the encoder, and a Global Category-Center Module is utilized to generate a global category-center prior based on deep features, which is then combined with a Fourier Global Enhancement Module to enhance deep features in the frequency domain. During the decoding stage, a Local Category-Aware Frequency Attention Module is employed to progressively refine feature representations under the guidance of the global category-center prior, thereby achieving collaborative improvement in global semantic consistency and local detail recovery. Experimental results demonstrate that GC2F-Net achieves robust and competitive segmentation performance on multiple public remote sensing semantic segmentation datasets. The proposed method provides an effective spatial-frequency collaborative modeling paradigm for the semantic segmentation of high-resolution remote sensing images.

Remote Sensing doi: 10.3390/rs18101596

Authors: Kaiyuan Zheng Guojin He Ranyu Yin Guizhou Wang

Impervious surfaces are an important land surface indicator of urbanization level and human activity intensity, playing a crucial role in urban development monitoring and ecological environment assessment. However, in complex high-altitude regions such as the Qinghai–Xizang Plateau, the identification accuracy of existing medium-resolution impervious surface products remains limited at the regional scale due to complex land surface backgrounds, sparse distributions of impervious surfaces, and their generally small spatial extent. To address this challenge, this study proposes a Seed-Driven Grid Adaptation (SDGA) framework for large-scale impervious surface mapping over the Qinghai–Xizang Plateau. The proposed method uses the Google Satellite Embeddings (GSE) dataset as the primary input features and incorporates a 10 m impervious surface prior (P10) derived from a 2 m high-resolution impervious surface product to provide spatial constraints. Based on this prior information, a Prior-guided Hybrid Active Sampling (PHAS) strategy is developed to automatically construct high-value training samples through uncertainty-based positive sample mining and cluster-based negative sample mining. The framework first builds an initial seed knowledge base in the Lhasa seed area and subsequently performs local adaptive expansion within a 2° × 2° grid system, enabling automated impervious surface mapping across the Qinghai–Xizang Plateau. Experimental results show that, with only a small number of initial samples, the PHAS strategy significantly improves model performance, increasing the F1 score for impervious surface identification in the Lhasa seed area from 65.02% to 82.22%. During the grid-level adaptation stage, approximately 67% of the grids achieved improved accuracy, with an average F1 score increase of 0.1109 across the study area. Ultimately, the SDGA framework produced a 10 m resolution impervious surface product for the Qinghai–Xizang Plateau (SDGA-ISC10m), achieving an overall F1 score of 0.8223. Compared with seven existing medium-resolution impervious surface datasets, the proposed method demonstrates improved recognition performance under complex plateau environments, particularly in detecting sparsely distributed and small-scale impervious surfaces. The results indicate that integrating remote sensing embedding features with active learning strategies can effectively reduce the need for manual annotation and provide a new technical pathway for large-scale impervious surface mapping in complex regions.

Remote Sensing doi: 10.3390/rs18101598

Authors: Junjie Xiong Haopeng Zhang Zhiguo Jiang

Satellite video super-resolution (SVSR) aims to reconstruct high-resolution video frames from low-resolution satellite observations, providing enhanced visual details for remote sensing applications. Despite recent progress, existing methods still suffer from limited alignment accuracy under complex motion and insufficient feature aggregation across frames, which restricts reconstruction quality. To address these issues, we propose a lightweight SVSR framework that incorporates Kolmogorov–Arnold Networks (KAN) into both the alignment and fusion processes. Specifically, a KAN-based spatial attention module is introduced to enhance the first-order and second-order neighboring frames, improving the accuracy of frame alignment. In addition, a KAN-based channel attention mechanism is adopted to facilitate more effective multi-frame feature aggregation. Benefiting from these designs, the proposed framework achieves strong reconstruction capability while maintaining a lightweight model structure. Extensive experiments demonstrate that the proposed method achieves superior performance in terms of PSNR, SSIM, LPIPS, and tOF compared with existing approaches, verifying the effectiveness of integrating KAN into SVSR.

Remote Sensing doi: 10.3390/rs18101597

Authors: Pauline Mnev Kelsey E. Nyland Ryan N. Engstrom Alexander L. Kholodov

Interior Alaska’s rapidly thawing permafrost poses risks to environmental systems and infrastructure, challenging municipal planning. As part of a larger project examining frozen commons, McGrath, Alaska, officials and tribal council members requested a permafrost map. This paper presents ground thermal monitoring (October 2023 to March 2025) and imagery-derived land cover and permafrost/seasonal freezing maps developed after testing machine learning and contextual feature methods. Over the two years of observation, ground temperature warmed 0.26 °C year−1 at 1.5 m depth. A high-accuracy land cover classification was generated to project ground thermal conditions across the community. Several supervised machine learning algorithms were compared with and without contextual features on a Satellite Pour l’Observation de la Terre (SPOT) scene in ArcGIS Pro. Per-pixel classification performed better given the contiguous spectral features, and contextual features did not improve overall accuracy. Instead, a random forest classifier that yielded the highest overall accuracy was used to generate a 1.5 m resolution permafrost/seasonal freezing map. Maps and thermal data can inform community frozen commons decision-making, and methods can be repeated to monitor regional change. Discussion of results highlights potential permafrost mapping applications, particularly of Gabor and mean contextual features with object segmentation.

Remote Sensing doi: 10.3390/rs18101595

Authors: Alexei Kolgotin Detlef Müller

We developed a reference look-up table (RLUT) of particles of spheroidal shape. This RLUT will be used in our lidar-data inversion algorithm we have developed in the past 25 years for the retrieval of microphysical parameters of non-spherical particles from 3β + 2α + 3δ optical datasets measured with Raman/HSRL lidar. The optical datasets are described by particle backscatter coefficients (β) at three wavelengths λ = 355, 532, and 1064 nm, particle extinction coefficients (α) at two wavelengths λ = 355 and 532 nm, and particle linear depolarization ratios (PLDRs, δ) at three wavelengths λ = 355, 532, and 1064 nm. The RLUT contains 64,032 synthetic 3β + 2α + 3δ—datasets calculated on the basis of a light-scattering model of randomly oriented spheroids and spheroid particle size distributions described by different particle complex refractive indices (CRIs) and lognormal functions with different Gauss parameters such as mean radius (μ) and standard deviation (σ). We investigate major features of the RLUT such as information content encoded in the 3β + 2α + 3δ datasets, conditionality, determinacy and the sensitivity of the retrievals to the underlying measurement errors. We find that major features of the sphere and spheroid RLUTs are similar; however, extra information is encoded in the PLDRs. The PLDR spectrum on the domain λ ∈ [355; 1064] μm contains significant information about the size of spheroid particles. The analysis of the information content is more productive if we use the cross-polarized backscatter-related Ångström exponent (CrPBAE) at the wavelength pairs 355 and 532 nm [β˙⊥(355/532)] and the wavelength pairs 532 and 1064 nm [β˙⊥(532/1064)]. In particular, the cycloid-like behavior of the interdependency β˙⊥ (355/532) versus β˙⊥(532/1064), i.e., hysteresis, means that non-spherical particle size changes.

Remote Sensing doi: 10.3390/rs18101594

Authors: Nicola Furnitto Sabina I. G. Failla Giuseppe Sottosanti Marcella Avondo Matteo Bognanno Luisa Biondi Juan Miguel Ramírez-Cuesta

The accurate estimation of pasture above-ground biomass (AGB) is critical for optimizing stocking rates and ensuring the sustainable use of Mediterranean pastures. This study developed empirical models to estimate fresh (AGBfresh) and dry above-ground biomass (AGBdry) using multispectral imagery acquired by Unmanned Aerial Vehicles (UAVs) in a Hedysarum coronarium pasture in Sicily, Italy. Field biomass was destructively sampled simultaneously with UAV surveys in 28 georeferenced plots during pre- and post-grazing phases over the 2023–2024 and 2024–2025 seasons. Data were collected with a DJI Mavic 3 Multispectral (for the 2024 test) and a DJI Matrice 300 + Altum-PT (for the 2025 test) and radiometrically calibrated to surface reflectance. Because two different multispectral sensors were used across years, an inter-sensor harmonization step was applied before vegetation-index calculation. Thirty-three vegetation indices were extracted as mean values within circular buffers of 1 m radius, centered on each sample plot to accommodate GNSS/georeferencing uncertainty. For each vegetation index, linear and exponential models were calibrated using 66% of the dataset and validated on the remaining 33% to predict fresh and dry above-ground biomass, and model performance was assessed using R2 and RMSE. On the validation dataset, ARVI2 and EVI2 showed the highest explanatory power for AGBfresh (R2 = 0.89), with ARVI2 providing the lower RMSE (2047 g m−2). For AGBdry, visible-band indices such as NGRDI and GRVI were among the best performers, reaching R2 = 0.85 with RMSE = 1371 g m−2. Visible-band greenness indices were among the most competitive predictors, whereas several conventional NIR-based indices showed only moderate performance. Overall, this UAV-based multispectral approach represents a promising and interpretable tool for biomass estimation in heterogeneous Mediterranean pastures, although further validation across additional seasons and sites is required to strengthen its transferability.

Remote Sensing doi: 10.3390/rs18101593

Authors: Ali Taghimolla Ali Asgary Mahbod Aarabi

In 2023, Tropical Cyclone Freddy caused severe damage in southern Malawi, flooding much of the lowland area near Lake Chilwa and displacing many residents. This study evaluates long-term, region-specific mitigation strategies to lessen future risks, using a novel approach that combines drone and satellite data, building footprints, and 3D simulations to analyze how building elevation affects flood damage and assess Property-Level Flood Risk Adaptation measures. Results show a significant difference in ground elevation between affected and unaffected buildings, with damaged structures generally at lower levels. The 3D simulation confirmed a water-level rise of approximately 3.0 m caused by Freddy. Scenario analysis indicates that elevating buildings by 2.0, 2.5, and 3.0 m could reduce direct flood exposure and 64%, 76%, and 91% of damage, respectively. These insights can inform the development of targeted regional risk-mitigation strategies through Property-Level Flood Risk Adaptation in high-risk areas.

Remote Sensing doi: 10.3390/rs18101592

Authors: Qing Ma Dongpu Wu Yichen Zhang Jiquan Zhang Jinyuan Xu Yechi Yao

Building damage and collapse are the main sources of serious casualties and financial losses during earthquakes, which are among the most destructive natural disasters that endanger human life and property. Therefore, quick and precise post-earthquake building damage assessment is essential for risk assessment and emergency action. Convolutional neural networks (CNNs) primarily concentrate on local features and frequently ignore global contextual information within and across buildings, despite the fact that deep learning-based techniques allow automated damage identification. Transformer-based approaches, on the other hand, are good at capturing global dependencies, but their large memory and processing costs restrict their usefulness. As a result, existing networks still struggle to achieve an effective balance between accuracy and efficiency. To address this issue, this study proposes a lightweight and efficient network for post-earthquake building damage assessment. Specifically, we develop a two-stage method based on EfficientViM with an encoder–decoder architecture. In the encoder, Mamba is introduced to extract multi-scale change features with long-range dependencies, leveraging the state space model to preserve global modeling capability while significantly reducing computational complexity. In the decoder, two lightweight modules are designed to further enhance discriminative capability and computational efficiency. The network finally outputs building localization and pixel-level building damage, respectively. Experiments were conducted on four earthquake events from the BRIGHT dataset using a three-for-training and one-for-testing cross-event rotation evaluation strategy. The results demonstrate that LEViM-Net requires only 30.94 M parameters and 27.10 G FLOPs. In addition, for the Türkiye earthquake event, the proposed method achieves an F1 score of 80.49%, an overall accuracy (OA) of 88.17%, and a mean intersection over union (mIoU) of 49.73%. The proposed model enables efficient remote-sensing-based mapping of macroscopic and image-visible building damage, providing timely support for early-stage emergency response.

Remote Sensing doi: 10.3390/rs18101591

Authors: Yuyuan Zhang Xuechi Yang Shuai Yang Penglin Zhao Yuanye Cao Xiuguo Liu Liping Li Qihao Chen

A large landslide struck the Zhenba Dahekou area of Shaanxi Province, China, on 9 September 2021. To accurately extract landslide displacement in emergency situations, in this study, we explore the feasibility and effectiveness of using satellite images and post-failure emergency UAV images to investigate the large displacement of the landslide through the pixel offset tracking (POT) method and evaluate four different POT methods, including NCC operated in the spatial domain (NCC), the orientation correlation method in ImGRAFT software (ImGRAFT), the multi-pass method in GIV software (GIV), and the frequency method in COSI-Corr software (COSI-Corr). It is found that the Zhenba landslide has moved southwest by about 74.3 m~96.4 m with the sliding direction between 231°~258°. The southward displacement of the landslide gradually decreases from southeast to northwest, and the westward displacement on the west side is greater than that on the east side. The relative matching precision of the POT methods in stable areas reached 0.8 m, superimposed on an image registration RMSE of 1.2 m. Under the experimental conditions of this study, ImGRAFT demonstrated robust overall performance. In terms of matching ability, ImGRAFT and NCC outperform GIV and COSI-Corr. In terms of displacement gradients expression ability, ImGRAFT and COSI-Corr outperform NCC and GIV; in terms of matching efficiency, COSI-Corr, GIV, and ImGRAFT are far superior to NCC. This study expands the application of multi-source optical data to investigate landslides and provides suggestions for the displacement extraction of large-displacement landslides, which will be helpful for the emergency investigation and research of landslides in the future.

Remote Sensing doi: 10.3390/rs18101590

Authors: Hui Peng Wei Gao Zhifu Li Bobo Luo Qi Wang

This study presents the first systematic evaluation of the capability of the Surface Water and Ocean Topography (SWOT) satellite Level-2 High Rate Pixel Cloud (L2_HR_PIXC) product for retrieving topography in reservoir drawdown zones under varying terrain conditions in arid and semi-arid regions. Three representative reservoirs in southern Xinjiang, China—characterized by plain, canyon, and pocket-shaped canyon morphologies—were selected to establish a terrain-dependent validation framework. A novel multi-feature clustering strategy integrating elevation and radar backscatter coefficients was explored to reduce the misclassification of wet mudflats as water pixels in the PIXC product, aiming to improve DEM accuracy in reservoir drawdown zones. Based on this framework, multi-cycle SWOT-derived digital elevation models (DEMs) were generated and quantitatively evaluated against high-resolution unmanned aerial vehicle (UAV) Light Detection and Ranging (LiDAR) DEMs. Results demonstrate a strong terrain dependency in SWOT-derived elevation accuracy. In low-relief environments, sub-meter accuracy is achieved, with the root mean square error (RMSE) below 0.25 m, confirming the suitability of SWOT for high-precision monitoring. However, errors increase significantly in steep and complex terrains, reaching up to ±6 m, primarily due to interferometric decorrelation, geometric distortion, and slope-induced biases. Despite these limitations, multi-temporal observations exhibit generally similar spatial error patterns across terrains, indicating reasonable repeatability under the tested conditions. This study reveals the performance boundaries of SWOT-derived DEMs in dynamic land–water transition zones and provides a robust methodological framework for improving DEM extraction in similar environments. The findings contribute to advancing the application of SWOT data in hydrological monitoring and geomorphological analysis at regional scales.

Remote Sensing doi: 10.3390/rs18101582

Authors: Xinlan Xu Jiaqing Qiao Jialin Zhou Kuo Yuan Lei Feng

Fusion-based hyperspectral image super-resolution (HSI-SR) on diffusion models exhibits promising performance in generating high-quality, realistic features. However, existing methods are confronted with two limitations: (1) static conditional guidance is discordant with the dynamic denoising process, and (2) modality conflicts are inadequately addressed by concatenation. To address these challenges, we propose a novel Modulated Diffusion Framework with Spatial–Spectral Disentangled Guidance (SSDG). Specifically, it introduces a Dynamic Modulated Residual Network (DMRN), which leverages a time-aware mechanism to dynamically adjust conditional feature injection, ensuring adaptive guidance throughout all denoising stages. Furthermore, we design a training-free SSDG strategy to explicitly decouple spatial and spectral guidance during sampling, allowing for flexible control over the fusion process to mitigate modality conflicts. Extensive experiments on three public datasets demonstrate that the proposed method achieves state-of-the-art performance, exhibiting superior robustness, particularly in challenging noisy scenarios.

Remote Sensing doi: 10.3390/rs18101587

Authors: Chenxu Wang Man Peng Kaichang Di Yuke Kou Bin Xie

Lunar optical and synthetic aperture radar (SAR) imagery provide complementary information for characterizing the lunar surface. However, their joint use remains challenging because of substantial cross-modality differences and severe illumination constraints, particularly in polar regions. To address this challenge, we propose LS2ODiff (Lunar SAR-to-Optical Diffusion), a diffusion-based framework designed for SAR-to-optical image translation in lunar environments. LS2ODiff uses SAR observations as conditional guidance in the diffusion process and incorporates a partial-convolution strategy into the U-Net backbone to handle irregular invalid regions. In addition, self-attention modules are incorporated into the downsampling stages of the U-Net to model long-range spatial dependencies and enhance global structural consistency in complex lunar terrain. We further construct a dedicated paired dataset of the lunar south polar region by registering Chandrayaan-II DFSAR data with Lunar Reconnaissance Orbiter (LRO) Narrow-Angle Camera (NAC) imagery. Comparative experiments against Pix2Pix, CycleGAN, SynDiff, and ConDiff demonstrate that LS2ODiff achieves better visual fidelity and quantitative performance in terms of peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), Fréchet inception distance (FID), and learned perceptual image patch similarity (LPIPS). These results demonstrate the potential of diffusion models for high-fidelity lunar image translation, offering new opportunities for polar terrain interpretation and future exploration missions.

Remote Sensing doi: 10.3390/rs18101589

Authors: Hao He Shuyang Wang Lei Huang Xiaohu Fan Yongfei Li Dongfang Yang

Road extraction from high-resolution remote sensing imagery is fundamental to numerous practical applications, yet still faces notable challenges caused by label noise, particularly the underlabeling of rural roads within training datasets. End-to-end dense prediction networks deliver high efficiency and strong global context capture capability, yet they are highly vulnerable to such label noise. In contrast, patch-based methods achieve better robustness but sacrifice global reasoning ability and computational efficiency. This paper proposes a novel training strategy named Positive-guided Local Supervision (PLS), which integrates the strengths of the two aforementioned paradigms. PLS preserves the full end-to-end forward pass to leverage global context, while restricting loss computation to local patches centered on reliably annotated road pixels (positive samples) via a standard dense segmentation loss. By isolating the model from misleading gradients generated in underlabeled regions, PLS effectively mitigates the negative impact of underlabeling without compromising computational efficiency and prediction quality. We evaluate the proposed PLS on two datasets: the public DeepGlobe benchmark and a newly constructed challenging dataset, namely China Four Provinces (CH4P). CH4P includes 13,498 high-resolution images of rural China, which suffers from severe underlabeling inherited from public web maps. Extensive quantitative evaluations on DeepGlobe and the newly built CH4P dataset validate that our PLS strategy surpasses conventional end-to-end baselines and competitive state-of-the-art methods under both noisy original labels and manually refined annotations. On the refined DeepGlobe-mini-test and CH4P-mini-test subsets, PLS obtains prominent absolute IoU improvements of 0.127 and 0.104 over baseline models, respectively, showing distinct superiority in handling severe real-world underlabeling. Qualitative visualizations and cross-dataset generalization tests further demonstrate that PLS can effectively retrieve road segments omitted in raw annotations, delivers strong robustness against practical label noise, and introduces no extra computational burden in the inference stage.

Remote Sensing doi: 10.3390/rs18101588

Authors: Amir Hamedpour Ruth P. Tchana Wandji Bjarni D. Sigurdsson Asra Salimi Iolanda Filella Josep Peñuelas

Rising temperatures, driven by global climate change, are profoundly altering high-latitude ecosystems, influencing vegetation phenology and productivity. However, understanding the long-term, nuanced responses of these ecosystems remains a critical challenge. Soil warming experiments have served as useful tools for understanding these shifts. However, many of these studies have relied on a single measure, predominantly the Normalized Difference Vegetation (NDVI), measured at a single level of warming. This approach often fails to separate structural greening from underlying physiological responses. To address these gaps, this study provided a comprehensive snapshot assessment of growing season vegetation dynamics in a subarctic grassland ecosystem in Iceland that had been exposed to continuous geothermal soil warming for over 60 years. Using high-resolution multispectral drone imagery, twelve different vegetation indices (VIs) were derived to assess not only greenness but also physiological stress and photosynthetic efficiency across a range of mean annual soil temperatures (MATs). Using linear regression and redundancy analysis (RDA), the responses of these indices to warming and their relationships with other environmental drivers, such as standing biomass and plant nutrient concentrations (nitrogen and phosphorus), were analyzed. The results revealed significant positive linear relationships between most of the indices and MATs across the 5 to 11 °C range. This indicated that higher MATs led to increased biomass and structural growth, without revealing any significant thresholds or tipping points in vegetation response within the observed warming range. However, the Photochemical Reflectance (PRI) showed a significant negative relationship with warming, suggesting a decoupling between structural greening and photosynthetic light-use efficiency. Furthermore, RDA results indicated that, while most of the VIs were primarily driven by biomass, the decline in PRI was likely a compounding effect of physical canopy self-shading and plant phosphorus constraints. Ultimately, this study demonstrated that, while these subarctic grasslands exhibited local evidence of “Arctic greening” under further warming, multispectral drone remote sensing could detect underlying physiological adjustments and nutrient constraints that traditional greenness indices might overlook, providing a more nuanced understanding of ecosystem response.

Remote Sensing doi: 10.3390/rs18101583

Authors: Yuanfei Pan Jianhui Dong Yangdan Dong Minggao Tang Ran Tang Zhanxi Wei Xiao Wang Xinhao Yao

Accurate landslide susceptibility assessment remains challenging in mountainous regions with complex terrain, heterogeneous geology, and clustered landslide inventories. This study develops a slope–unit–based landslide susceptibility assessment framework for Tongren County, Qinghai Province, China, using a landslide inventory of 217 events, multi–source environmental data, Certainty Factor (CF)–based conditioning–factor analysis, and machine learning models. Eighteen conditioning factors derived from remote sensing, geological survey, and meteorological datasets were extracted at the slope–unit scale, and their collinearity was evaluated using Pearson’s correlation and the Variance Inflation Factor (VIF). Eight models—Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), AdaBoost, Decision Tree (DT), XGBoost, K–Nearest Neighbors (KNN), and Convolutional Neural Network (CNN)—were evaluated under a 70:30 train/test split. The results show clear performance differences among the tested models: SVM achieved the best overall balance between discrimination and landslide detection (AUC = 0.9489; recall = 0.879). The tested CNN baseline showed relatively weak performance under the current slope–unit–based tabular–data setting. Susceptibility zoning results showed that high– and very–high–susceptibility zones were mainly concentrated along the Longwu River and its tributaries, where middle–elevation dissected terrain, weak lithological materials, river–valley erosion, and human engineering activities spatially coincide. These results provide a practical basis for slope monitoring and land–use planning in Tongren County.

Remote Sensing doi: 10.3390/rs18101586

Authors: Aili Wang Xinyu Liu Haisong Chen

Hyperspectral image classification (HSIC) is a crucial task for remote sensing applications, requiring accurate pixel-level labeling while effectively capturing both spectral and spatial information. Traditional convolutional neural network architectures often struggle to balance local texture detail and global contextual consistency, and existing neural architecture search (NAS) methods rarely incorporate attention mechanisms, limiting their performance. To address these challenges, this study proposes a multi-scale Transformer-based NAS framework (TR-NAS) for fine-grained hyperspectral image classification. The framework combines local cube sampling, shallow and deep multi-scale convolutions, and a searchable Transformer module that adaptively selects global, local window, and multi-scale attention operators. Lightweight enhanced convolution operators, including dual-gated (DG-Conv) and mixed depthwise (MixConv) convolutions, are incorporated to improve spectral discrimination and scale robustness. Extensive experiments on the PU and Hanchuan datasets demonstrate that TR-NAS achieves superior classification accuracy, stability, and boundary consistency compared to traditional methods and existing NAS architectures, showing improved robustness to spectral similarity and spatial heterogeneity in complex remote sensing scenes.

Remote Sensing doi: 10.3390/rs18101585

Authors: Chuyang Liu Liangyi Wu Kai Liu Luyang Chen Xin Wei Xi Yang

High-resolution images of non-cooperative spacecraft are essential for on-board autonomous operations. Hardware bandwidth limits and continuously changing observation distances mean that a practical super-resolution (SR) system must handle arbitrary, non-integer magnification factors without retraining, a setting known as arbitrary-scale SR (ASSR). Recent 2D Gaussian splatting (2DGS) methods represent image content with explicit anisotropic Gaussian primitives and render at any continuous coordinate, offering substantially faster inference than implicit neural representation (INR) approaches. Yet spacecraft imagery presents a structural mismatch for uniform 2DGS regression: the target occupies a small, densely structured region within a vast, featureless deep-space background, so a network that minimizes average reconstruction loss inevitably over-invests capacity in the irrelevant background and smears the fine edges of antennas and solar panels. We propose S3R-GS, a saliency-guided framework that embeds semantic spatial priors into the 2DGS pipeline at three levels: an encoder-level module that suppresses background noise before it reaches the splatting stage; a discrete Gaussian routing mechanism that assigns each spatial location to a semantically appropriate kernel group and reformulates Gaussian modeling as semantic prototype selection; and a saliency-weighted training strategy that concentrates the optimization gradient on the spacecraft target. Experiments on the SPEED and SPEED+ benchmarks show that S3R-GS achieves strong PSNR performance, competitive SSIM, and improved perceptual quality across scale factors from ×2 to ×12; additional ablation, extreme-lighting, and efficiency analyses further support the robustness and practicality of the proposed design.

Remote Sensing doi: 10.3390/rs18101584

Authors: Hongzhu Wang Jiale Chen Wei Liang Guangli Xu

This study is conducted to investigate the deformation evolution and collapse mechanism of the Rongxing gypsum mine by integrating multi-source monitoring data, including synthetic aperture radar (SAR), global navigation satellite system (GNSS), and microseismic observations. Long-term surface deformation from 2015 to 2025 is retrieved using small baseline subset interferometric synthetic aperture radar (SBAS-InSAR), while GNSS data (2021–2022) are used to capture rapid ground displacement during the collapse event. Microseismic monitoring provides insights into the evolution of subsurface fracturing processes. The results show that the pre-collapse stage is characterized by continuous and spatially heterogeneous subsidence. Prior to the collapse, microseismic activity is observed to exhibit clear precursory signals, including an increase in event number, a decrease in b-value, and accelerated cumulative energy release, suggesting that the transition from distributed microcrack development to large-scale fracture coalescence is occurring. The b-value, derived from the Gutenberg–Richter frequency–magnitude relationship, describes the relative proportion of small to large seismic events and reflects variations in the statistical distribution of event magnitudes. During the collapse stage, abrupt, large-magnitude subsidence is observed by GNSS. After the collapse, the deformation is found to enter a long-term adjustment phase characterized by the coexistence of subsidence and uplift, indicating that stress redistribution within the overburden is occurring. Based on these observations, a conceptual model is proposed to describe the progressive failure mechanism of the goaf, with four stages: slow subsidence, accelerated deformation, collapse, and post-collapse adjustment. This study demonstrates the effectiveness of integrating SBAS-InSAR, GNSS, and microseismic monitoring for understanding the full lifecycle of goaf collapse. It provides valuable insights for early warning of mining-induced geohazards.

Remote Sensing doi: 10.3390/rs18101581

Authors: Anming Zhang Zheng Zhang Keli Shi Ping Tang

Recent advances in satellite missions, particularly the Landsat, Sentinel, and Gaofen series, have led to the rapid accumulation of high-quality remote sensing data with frequent revisits. As these data have become more widely available, Satellite Image Time Series (SITS) have become an important tool for monitoring Earth surface dynamics. SITS now supports a wide range of applications, including precision agriculture, Land Use/Cover Change (LULCC) monitoring, environmental management, and disaster response. This growth has also promoted the development of advanced SITS classification datasets. However, existing reviews have mainly focused on SITS classification algorithms or specific applications, while systematic comparisons of public SITS benchmark datasets remain limited. This lack of synthesis makes it difficult for researchers to navigate fragmented resources and select datasets that match specific scientific or operational tasks. To address this gap, this paper provides a comprehensive review and analysis of 29 publicly available medium-to-high-resolution SITS classification benchmark datasets released between 2017 and 2025. These datasets are intended for training, testing, and validating land-cover classification algorithms, rather than for direct use as operational map products. We conduct a detailed statistical and comparative analysis of these datasets, focusing on their key characteristics across spectral, temporal, and spatial dimensions, as well as their labeling systems. In addition, this review summarizes the SITS classification algorithms that have been developed and benchmarked using these datasets. Finally, we identify the main challenges in constructing and applying SITS classification datasets and discuss future research directions, particularly in data reconstruction, multimodal fusion, change analysis, and advanced model architectures. This survey provides the research community with a systematic overview of SITS classification benchmark datasets and aims to support continued progress in this rapidly developing field.

Remote Sensing doi: 10.3390/rs18101580

Authors: Jie Wang Runbin Hu Haiyang Zhang Yixuan Zhou

The eastern coast of China, characterized by a pronounced climatic gradient and diverse ecosystems, is an ideal region for exploring the spatiotemporal dynamics of carbon fluxes and their drivers. Based on observations from eight flux tower sites, together with meteorological, remote sensing, and ecohydrological variables from 2001 to 2022, this study developed Back Propagation (BP), Support Vector Regression (SVR), Extreme Gradient Boosting (XGBoost), and Random Forest (RF) models to estimate regional gross primary productivity (GPP), ecosystem respiration (ER), and net ecosystem productivity (NEP). Among them, RF performed best, achieving validation R2 values of 0.92, 0.84, and 0.83 for GPP, ER, and NEP, respectively, and was therefore selected for regional upscaling. The regional mean GPP, ER, and NEP were 1578.38, 1286.05, and 334.56 g C m−2 yr−1, respectively, indicating that the region functioned as a net carbon sink during the study period. GPP, ER, and NEP exhibited a clear spatial gradient, with higher values in the south and lower values in the north. Total regional NEP increased from 344.12 Tg C in 2001 to 517.73 Tg C in 2022, reflecting a continuous strengthening of terrestrial carbon sink strength. Forests contributed most to the regional carbon sink, while the ecosystem-level NEP contribution of croplands increased over time; by contrast, the total carbon sink of wetlands declined because of area loss. These results suggest that ecological restoration, vegetation greening, and land cover optimization jointly enhanced the carbon sink along the eastern coast of China. These findings have important implications for ecological management and green low-carbon development along the eastern coast of China.

Remote Sensing doi: 10.3390/rs18101579

Authors: Yudan Liu Yuxin Zhao Yan Yan Yan Shao Xinqi Qu Ling Wu

Accurate and timely information on forest disturbance drivers is important for sustainable forest management, global carbon cycle accounting, and climate change response. However, forest disturbance classification is difficult due to two major challenges: limited labeled samples and highly imbalanced disturbance class distribution. In this article, a new framework for multi-type forest disturbance classification based on collaborative semi-supervised learning and sample generation was proposed. First, forest disturbance is detected using long-term remote sensing time series data and disturbance detection algorithms. Spatiotemporal, spectral and terrain features of different disturbance types are extracted. On this basis, to address the problem of imbalanced and small-sample conditions, a collaborative classification strategy is developed. Based on a small number of labeled samples, Support Vector Machine (SVM) and Random Forest (RF) are used to build dual base classifiers. A confident learning (CL) framework is applied to select high-confidence pseudo-labeled samples from unlabeled data. Then, a latent diffusion model (LDM) is introduced to generate high-fidelity pseudo-samples. This increases the sample size and balances the class distribution. Based on the augmented dataset, the dual classifiers are iteratively optimized using a co-training strategy, which improves model generalization under complex conditions. The results show that the proposed framework could generate high-quality pseudo-samples and effectively reduce class imbalance. The overall accuracy (OA) of the proposed framework reaches 93.2%, which is 5.7% and 4.4% higher than single classifier baselines, respectively. After introducing the LDM-based balancing mechanism, performance is further improved by 1.8% compared with the pure semi-supervised framework. This study provides an efficient and reliable solution for large-scale forest ecosystem monitoring.

Remote Sensing doi: 10.3390/rs18101578

Authors: Haoliang Zhu Chunli Jiang Xiuli Zhu

Small-object detection in urban remote sensing imagery is essential for smart city applications, yet remains challenging due to limited target size, large scale variations, complex urban backgrounds, as well as the trade-off dilemma between detection accuracy and deployment-oriented efficiency. To address these issues, this paper proposes LMGANet, a parameter-efficient and real-time YOLOv11-based detector for urban remote sensing object detection. A C3K2-GDF module is introduced to enhance small-object representation through adaptive receptive-field adjustment and dynamic feature refinement. An Adaptive Multi-scale Feature Aggregation Network (AMFAN) is designed to strengthen cross-scale feature interaction and improve the fusion of spatial details and semantic information. In addition, a Lightweight Enhanced Shared (LES) detection head is developed to reduce parameter redundancy while preserving localization accuracy for small targets. Experiments on the VisDrone2019 and AI-TOD datasets show that LMGANet improves mAP50 by 4.8% and 3.2% over YOLOv11S, respectively, with only 3.63 M parameters and real-time inference capability. These results demonstrate that LMGANet achieves an effective balance among detection accuracy, parameter efficiency, and real-time inference performance for urban remote sensing applications.

Remote Sensing doi: 10.3390/rs18101569

Authors: Ping Wang Jianli Ding Jinjie Wang Yitu Guo Fangqing Liu Shuang Zhao Haiyan Han Shiyi Yuan Wen Ma

Aerosol optical depth (AOD) and Ångström exponent (AE) are critical parameters for characterizing atmospheric aerosols, playing a pivotal role in atmospheric environmental monitoring and climate change studies. This study addressed the imperative need for a systematic evaluation of mainstream reanalysis products by conducting a comprehensive multi-scale assessment of the CAMS and MERRA-2 datasets (2003–2023), encompassing data quality verification, spatiotemporal pattern analysis, and trend evolution investigation. The following key findings emerge: (1) Both AOD data exhibited the best performance observed in low–mid latitudes. CAMS AOD (AODC) showed a slightly better correlation, while MERRA-2 AOD (AODM) demonstrated superior robustness. Both AE data performed similarly, and MERRA-2 AE (AEM) was superior. Both AE data performed better in low latitudes and near Europe. (2) CAMS and MERRA-2 showed good performance in annual and seasonal variations, with significant fluctuations and biases in the annual cycle. Both models achieved the highest AE performance in summer. MERRA-2 AOD demonstrated better hourly performance during daytime. The hourly stability of AE was slightly worse than AOD, with notably degraded performance during midday hours. (3) The distribution and trends of AOD over land showed spatial consistency. The distribution of AEM was generally lower than AEC’s. After ensemble empirical mode decomposition (EEMD), all datasets showed monotonically decreasing trends except for AEM. This study provides valuable insights into the strengths and limitations for CAMS and MERRA-2 and suggests possible areas for improvement in future data assimilation and parameterization.

Remote Sensing doi: 10.3390/rs18101577

Authors: Jinyu Xu Wenwei Liu Runze Tian Chengyou Wang Yuanbo Zhang

Balancing small object detection performance with model lightweighting remains a critical challenge in the remote sensing domain. To address the massive computational and parameter overhead of existing algorithms, we propose the hybrid scale dynamic detection transformer (HSD-DETR). This lightweight detector incorporates four core innovations to effectively enhance feature extraction for small objects. First, to reduce costs without compromising performance, we design a hybrid convolution and selective scanning fusion (HCSS-Fusion) module to reconstruct the backbone, combining local convolution with global linear scanning. Second, to preserve fine-grained information, we introduce a space-to-depth mixer (SPDMixer) to achieve pixel-level lossless downsampling. Third, to mitigate background interference and enhance small object representation, we develop a dynamic sparse adaptive intra-scale feature interaction (DSAIFI) module, employing a gating mechanism to dynamically select informative spatial tokens. Finally, to improve the localization precision for small objects, we propose the rational-focal minimum point distance intersection over union (RF-MPDIoU) loss, utilizing a non-linear mapping to dynamically modulate sample weights. Experimental results on public benchmarks confirmed that, compared to mainstream models, HSD-DETR achieves highly competitive accuracy while significantly reducing parameter scale and theoretical computational complexity. Ultimately, this research provides a lightweight and robust algorithmic solution for the field of remote sensing object detection.

Remote Sensing doi: 10.3390/rs18101576

Authors: Chanul Park Jeong-Hoon Park Seongwook Lee

Phase-modulated continuous-wave (PMCW) radar systems are gaining interest for autonomous sensing. However, high range resolution typically demands prohibitively high sampling rates and computational loads. To address this issue, we propose a novel nonlinear stepped-frequency PMCW (NSF-PMCW) radar system. The proposed NSF-PMCW radar system periodically transmits sequences whose carrier frequency varies nonlinearly over time, and the associated signal processing method synthesizes a wide effective bandwidth by processing and coherently summing these frequency-varying sequences. This approach successfully enhances the range resolution without increasing the bandwidth and sampling rate of the analog-to-digital converter. Furthermore, we propose an angle estimation algorithm that accounts for the time-varying frequency of sequences to improve the estimation accuracy. The simulation results show that the proposed system can achieve the range resolution of a 3 GHz PMCW radar system while using only 500 MHz of bandwidth with a root mean square error of 0.0081 m in range estimation and 0.1114∘ in angle estimation.

Remote Sensing doi: 10.3390/rs18101575

Authors: Ritik Pokharel Thanos Gentimis Manoch Kongchum Brenda Tubana Rejina Adhikari Tri Setiyono

Accurate in-season rice (Oryza sativa L.) yield prediction is crucial for improved nitrogen management and climate-smart decision making, yet rigorous comparative benchmarking of machine learning (ML) models using multi-temporal UAV spectral data with independent temporal validation remains limited. This study systematically evaluated four ML algorithms (Random Forest, XGBoost, Neural Network, and Linear Regression) and two Bayesian model averaging ensembles for rice yield prediction using UAV multispectral imagery. Field experiments spanning three growing seasons (2023–2025) at Louisiana State University comprised 9–10 varieties and six nitrogen rates (0–235 kg N ha−1; 576 plots). Vegetation indices and spectral bands from three growth stages were extracted as predictors. Models were compared using 300 random train–test iterations with systematic hyperparameter optimization, followed by independent validation on 2025 data. Among the individual models, XGBoost achieved the highest internal accuracy (R2 = 0.87, RMSE = 0.85 t ha−1), substantially outperforming Linear Regression (R2 = 0.66, RMSE = 1.32 t ha−1), while reduced BMA reached R2 = 0.89. XGBoost demonstrated robust temporal generalization (R2 = 0.62, NRMSE = 8.47%) despite environmental variation. The Enhanced Vegetation Index and Normalized Difference Red Edge at 90 days after planting (reproductive stage) were the most stable predictors across 300 iterations. Tree-based ML models substantially outperform traditional linear approaches, providing reliable pre-harvest yield forecasting for operational precision rice production.

Remote Sensing doi: 10.3390/rs18101574

Authors: Zanqiang Dong Zhaofeng Yang Wenbin Yu Hongjie Qian Yanfeng Fan Konglin Zhu Gaoping Liu

Short-term precipitation forecasting is essential for disaster prevention, urban management, and weather-sensitive decision making, yet radar-based nowcasting remains challenging because precipitation systems evolve nonlinearly and high-frequency echo structures are easily over-smoothed by deterministic sequence models. This paper proposes a ViT-modulated diffusion spatiotemporal prediction network (VSTPN) that cascades a spatiotemporal prediction module with a ViT-conditioned diffusion refinement module. The spatiotemporal module models the temporal evolution of radar echoes, whereas the ViT-Diffusion module uses global contextual features as conditional guidance during iterative denoising to refine spatial structures. Experiments on the HKO-7 benchmark show that VSTPN achieves lower MSE and higher SSIM than the tested baselines and improves CSI, HSS, and POD at the evaluated reflectivity thresholds. At the 40 dBZ threshold, the model improves CSI, HSS, and POD, while its FAR is slightly higher than that of ETCJ-PredNet, indicating a recall–false alarm trade-off for intense echoes. Additional post-hoc diagnostic analyses of relative gains, metric consistency, threshold sensitivity, and component effect sizes further support the stability of the reported improvements under the current experimental protocol. The results suggest that coupling spatiotemporal sequence modeling with diffusion-based radar echo refinement is a feasible direction for short-term precipitation forecasting; nevertheless, probabilistic uncertainty evaluation, multi-domain validation, and additional generative-quality metrics remain important directions for future work.

Remote Sensing doi: 10.3390/rs18101573

Authors: Chunhua Zhang Xiangkun Qi Hoi Shan Cheung Mingyang Zhang Yuemin Yue Kelin Wang

Accurate land use and land cover (LULC) data are essential for effective environmental management and reliable ecological modeling within complex landscapes such as the karst region of Southwest China. While new 10 m resolution global LULC products (i.e., ESA WorldCover, ESRI Land Cover, and annual mode composite of Dynamic World (DW)) offer unprecedented spatial detail, their reliability in heterogeneous karst remains poorly understood. We evaluated the accuracy and spatial consistency of these products for 2021 in the karst regions across five provinces in Southwest China using 1416 reference points collected through stratified random sampling. The ESA WorldCover dataset outperformed the others, achieving the highest overall accuracy (79.39 ± 2.19%). ESRI’s shrub metrics, however, reflect the structural absence of this class from its 2021 product rather than classification error. ESA’s superior performance in preserving fine-scale features is consistent with independent global assessments of both the 2020 and 2021 versions. This superior performance is attributed to its integration of Sentinel-1 SAR with optical data, a finer minimum mapping unit (100 m2), and expert-driven post-classification corrections. While all products successfully identified dominant classes like trees, substantial confusion emerged among spectrally similar classes such as shrubs, grass, and crops. A key finding was the strong effect of landscape heterogeneity on accuracy. Classification accuracy was 19.37% lower at patch edges (67.38%) compared to patch interiors (86.75%). Furthermore, edge reference points contribute disproportionately to total errors. Critically, none of the three products currently provide a sufficient basis for shrub-focused ecological monitoring in this region: ESA rarely detected shrub cover, DW mapped extensive but largely inaccurate shrub areas, and ESRI eliminated the shrub class from its 2021 product. These results show that while global 10 m products provide valuable information, careful product selection and regional validation remain essential for heterogeneous karst environments. Future improvements should integrate multi-source data (optical + synthetic aperture radar), apply topographic compensation for shadow effects, and develop region-specific approaches for mapping vegetation transitions.

Remote Sensing doi: 10.3390/rs18101572

Authors: Jinbo Wang Qiancheng Yu Ruiqing Zhang Nan Xiao

Remote sensing change detection (RSCD) is often disturbed by nuisance appearance variations, which can introduce pseudo-changes and degrade the reliability of predicted change masks. Robust change localization therefore requires that such spurious responses be suppressed while the structural integrity of change regions in complex, high-resolution scenes is maintained. We propose TriFusion-CD, a tri-branch framework that fuses complementary sources of information for reliable change localization. The first branch uses MobileSAM to provide global semantic guidance that promotes spatially coherent predictions. The second branch adopts the CLIP-ResNet50 image encoder with a change-aware enhancement module to extract detail-sensitive change features. The third branch performs frequency decomposition and interacts frequency features with CLIP text embeddings via cross-attention, producing a structural–semantic prior to suppress appearance-induced pseudo-changes. We further design a Semantic Attention Fusion Module (SAFM) to inject MobileSAM semantics into CLIP change features through cross-attention with learnable residual scaling. In addition, an Attention-Modulated Decoder (AMD) translates the fused guidance into multi-scale attention maps and performs progressive top-down refinement, extracting more spatially complete change regions. On the challenging SYSU-CD, JL1-CD, and CDD datasets, which exhibit diverse change patterns and frequent appearance-induced pseudo-changes, TriFusion-CD achieves 72.48% IoU/84.04% F1 on SYSU-CD, 66.04% IoU/79.54% F1 on JL1-CD, and 96.41% IoU/98.17% F1 on CDD, demonstrating strong performance.

Remote Sensing doi: 10.3390/rs18101571

Authors: Heng Wang Ruize Hong Yanxia Wu Dan Lin Liguo Wang

Masked auto-encoders (MAEs) have been extensively employed in the field of semi-supervised hyperspectral image classification (HSIC). However, the developed models encounter significant challenges in learning separable representations, as they do not sufficiently prioritize the reconstruction of the central pixel, which hinders their ability to learn separable representations. To address this limitation, we propose spectral–spatial MAE with central pixel reconstruction (SSMAE-CR), a novel self-supervised framework tailored for HSIC. To capture more comprehensive representations, SSMAE-CR employs a dual-branch architecture comprising a spectral MAE with central pixel reconstruction (SpecMAE-CR) and a spatial MAE with central pixel reconstruction (SpatMAE-CR). SpecMAE-CR highlights the significance of central pixel reconstruction by measuring the deviation between the central pixels of reconstructed and original samples. To preserve the holism of the learned latent representations, SpatMAE-CR maps the central pixels of the reconstructed samples back to their original counterparts through the introduction of an additional linear layer. Rigorous comparative experiments conducted on four publicly available datasets fully demonstrate that SSMAE-CR outperforms state-of-the-art methods. Furthermore, we validate the effectiveness of SSMAE-CR by evaluating the mean intra-class and inter-class distances of the learned representations. Experimental results demonstrate that prioritizing central pixel reconstruction yields a statistically significant increase in the mean inter-class distance, suggesting enhanced class separability in the representation space.

Remote Sensing doi: 10.3390/rs18101570

Authors: Yaoyue Gao Xinru Cheng Yimeng Li Dawei Xu Desheng Sun Yaoyi Hu

In all-weather Earth observation and complex unstructured environments, traditional single-modal remote sensing object detection often fails due to low illumination and strong background interference. While RGB–infrared fusion provides complementary information, existing methods are typically computationally intensive and struggle with dense small objects and modality discrepancies, limiting their deployment on resource-constrained platforms. To address these challenges, we propose ISDG-Net, a lightweight and efficient visible-infrared dual-modal object detection framework specifically tailored for edge deployment. ISDG-Net integrates four core components: (1) a channel-separated inverted bottleneck backbone (IBC-Conv) that reduces parameter redundancy while preserving modality-specific semantics; (2) a dynamic sparse attention module (DySparse) based on Bi-Level Routing Attention, enabling long-range dependency modeling with low computational cost; (3) an adaptive spatial fusion detection head (Detect-SASD) that aligns visible and infrared features at the pixel level to resolve semantic inconsistency and scale mismatch; and (4) a geometry-aware IoU selector (GIS) that mitigates over-suppression in crowded scenes by incorporating multi-dimensional geometric constraints into post-processing. Extensive experiments on the VEDAI, M3FD, and LLVIP datasets demonstrate the effectiveness and efficiency of ISDG-Net. It achieves 55.1% and 77.1% mAP@0.5 on VEDAI and M3FD, respectively, and 93.7% mAP@0.5 with 89.7% recall on LLVIP, while maintaining a compact model size of 4.2 M parameters and 11.3 GFLOPs. These results validate that accurate RGB–infrared detection is achievable under strict resource constraints, making ISDG-Net well-suited for deployment in edge-based remote sensing systems.

Remote Sensing doi: 10.3390/rs18101568

Authors: Fubin Zhang Zichi Zhang Feihu Zhang

Salient object detection in optical remote sensing images remains challenging due to complex backgrounds, blurred boundaries, small objects, unstable foreground–background contrast, and dense object distributions. Existing convolution-based methods are effective at modeling local structures, but they are limited in capturing long-range dependencies, whereas Transformer-based approaches usually incur substantial computational cost when handling high-resolution remote sensing imagery. To address these issues, this paper proposes EGMamba-Net, an edge-guided global–local collaborative network for salient object detection in optical remote sensing images. Specifically, a hybrid global–local backbone is first constructed to preserve shallow texture, edge, and geometric details while introducing Mamba-based global modeling in deeper stages for efficient long-range dependency representation. An Edge Prior Enhancement Module (EPEM) is then designed to explicitly extract boundary priors from shallow features and refine feature representations through edge-guided modulation. To alleviate the representation conflict between global semantics and local details, a Global–Local Interaction Module (GLIM) is further developed, where convolutional local modeling and Mamba-based global modeling interact through cross-gating for complementary feature learning. Moreover, a Region-Adaptive Routing Decoder (RARD) is introduced to dynamically assign different refinement paths according to regional saliency response, boundary intensity, and contextual complexity, thereby improving the recovery of small, low-contrast, and densely distributed objects. In addition, a Difficulty-Aware Joint Loss (DAJL) is designed to enhance optimization on boundary regions and hard samples, improving robustness under challenging conditions. Extensiveexperiments on ORSSD, EORSSD, and ORSI-4199 datasets demonstrate the superiority of the proposed method. In particular, on the more challenging EORSSD dataset, EGMamba-Net achieves 0.9389 S-measure, 0.8972 max F-measure, and 0.0066 MAE. Compared with the representative remote-sensing method DAF-Net, it improves S-measure and max F-measure by 0.0223 and 0.0358, respectively, indicating stronger capability in background suppression, structural preservation, and boundary recovery.

Remote Sensing doi: 10.3390/rs18101567

Authors: Yuan Dang Bo Yin Haipeng Cui Tao Bi Yiyun Guo

Short-term precipitation forecasting is an important research direction in meteorological studies, holding significant implications for disaster prevention and mitigation, urban flood drainage, and agricultural meteorological management. Existing deep learning models have achieved favourable results in modeling local features, yet they generally suffer from insufficient sensitivity to heavy precipitation areas, limitations in modeling temporal dependencies, and gradient instability issues. To address these limitations, we propose a novel spatiotemporal dual-branch neural network (ST-DualNet) for short-term precipitation forecasting based on radar echo maps. The network comprises a temporal branch (based on an enhanced ST-DConvLSTM) and a spatial branch (based on dilated convolutions and Transformer), respectively capturing the dynamic evolution and spatial structural features of precipitation. The two branches are integrated through the CBAM attention module and 3D convolution layer to achieve cross-branch feature fusion and prediction output. Experimental results demonstrate that ST-DualNet outperforms multiple mainstream models on the KNMI radar precipitation dataset, especially in heavy precipitation forecasting, providing an effective new framework for short-term precipitation forecasting.

Remote Sensing doi: 10.3390/rs18101566

Authors: Shuaijin Wu Yongjun Zheng Fenghui Li Zhaorong Zhuang

GNSS radio occultation (RO) measurements are essential to global Earth observation systems. The YUNYAO constellation offers dense global coverage, high vertical resolution, and all-sky capabilities. This study proposes an improved bending-angle observation operator and presents a comprehensive evaluation of YUNYAO’s bending-angle observations against those from COSMIC-2 for December 2024. Our results indicate that YUNYAO achieves near-global coverage and provides substantially more observations than COSMIC-2. Within their region of overlap, the bending-angle quality of YUNYAO is broadly comparable to that of COSMIC-2, though YUNYAO shows a more pronounced negative relative bias above 30 km. In the vertical profile obtained by horizontally averaging the relative bias, YUNYAO exhibits a smaller bias than COSMIC-2 below 30 km. For both RO constellations, bending angles retrieved from Galileo signals exhibit a smaller relative bias than those retrieved from GLONASS in the upper atmosphere. Finally, performance differences are also evident among YUNYAO’s individual receivers.

Remote Sensing doi: 10.3390/rs18101565

Authors: Yingman Guo Kaijun Yang Ruyi Feng Li Cao

The Dongting Lake Basin is a critical ecological zone in the middle reaches of the Yangtze River, playing a pivotal role in safeguarding regional ecological security and supporting socio-economic development. To investigate the spatiotemporal patterns and underlying drivers of water quality in Dongting Lake, this study developed a Spectral-Attention CNN (SA-CNN) inversion model integrated with the Efficient Channel Attention (ECA) mechanism, utilizing multi-source remote sensing data and convolutional neural networks. Results indicate that the proposed SA-CNN model significantly outperforms traditional machine learning approaches in predicting key water quality parameters, including total nitrogen (TN), total phosphorus (TP), ammonia nitrogen (NH3–N), and turbidity. Notably, the model achieved its highest predictive accuracy for TP, with an R2 value of 0.94. By incorporating spectral weight prior knowledge, the model was successfully transferred and trained on Landsat imagery. The validated model was subsequently applied to reconstruct and analyze the spatiotemporal trends from 2015 to 2025, revealing that water quality in Dongting Lake exhibits a fluctuating decline during winter months, while summer periods show an increasing trend in turbidity and TP concentrations. Further analysis suggests that water quality parameters are positively correlated with temperature and negatively correlated with precipitation, with anthropogenic activities also exerting a considerable influence.

Remote Sensing doi: 10.3390/rs18101564

Authors: Xinhua Tang Xiaoyu Fang Fei Huang

In urban environments with signal blockage and multipath effects, GNSS observation errors often exhibit temporal correlation. The Gaussian white noise assumption adopted in conventional tightly coupled Kalman filtering is prone to model mismatch under such conditions, which may lead to an underestimation of state uncertainty and consequently cause the protection level (PL) to fail to reliably bound the true positioning error. To address this issue, this paper proposes a tightly coupled GNSS/INS integrity monitoring method based on state augmentation and frequency-domain constrained parameter tuning. The method introduces first-order Gauss-Markov processes (GMP) to model major time-correlated error sources, including residual ephemeris and clock errors, residual tropospheric delay, and code multipath, by augmenting them into the filter state for joint estimation. The model parameters are further conservatively tuned based on power spectral density (PSD) envelope constraints to obtain more consistent covariance estimates. Based on this, the covariance output from the augmented filter is incorporated into the multiple hypothesis solution separation (MHSS) framework, enabling the protection level computation to better match the actual error statistics. Experimental results using vehicular field test data show that the proposed method effectively improves estimation consistency and significantly reduces the risk of PL underestimation in degraded environments. Furthermore, it achieves reliable bounding of horizontal positioning errors without noticeable degradation in positioning accuracy, while maintaining good system availability. These results demonstrate the effectiveness of covariance construction based on physical error modeling and PSD envelope constraints for integrity monitoring in complex environments.

Remote Sensing doi: 10.3390/rs18101563

Authors: Albert Reed Timothy Nagle-McNaughton Shiloh Elliott Jesse Mapel

Reconstructing 3D surfaces from electro-optical satellite imagery is an important capability for generating high-quality digital elevation models at scale. Recently, Gaussian splatting has emerged as a state-of-the-art technique for 3D reconstruction from satellite imagery. However, Gaussian splatting is optimized solely on RGB imagery, making it susceptible to errors when dealing with the radiometric inconsistencies and textureless regions common in satellite images. To address this, we propose a method for fusing Gaussian splatting with vision foundation models that is specifically tailored to satellite imagery. While recent work has explored fusing Gaussian splatting and vision foundation models, it has been studied only on terrestrial datasets, which, unlike multi-date satellite imagery, contain more constrained illumination at smaller scene scales. To account for these challenges, we introduce a method for computing multiscale satellite image embeddings along with a per-image feature alignment module. Benchmarked on the IARPA 2019 Challenge Dataset, our method reduces mean reconstruction error from 1.65 m to 1.57 m—a 5.2% relative improvement over previous methods. These results demonstrate that vision foundation models can enhance the geometric accuracy of satellite-based 3D reconstruction.

Remote Sensing doi: 10.3390/rs18101562

Authors: Ying Wang Zhongyuan Xie Huaiyong Shao Jichong Han Xiaofei Sun Long Ling Jiamei Long Ying Lin Liangliang Zhang

Cropland abandonment is increasing in the upper and middle Yangtze River Basin due to complex terrain, urbanization, and labor migration. This threatens regional food security. To address the challenge of monitoring abandonment in fragmented hilly areas, we developed a framework. We integrated machine learning with time-series analysis. We mapped cropland probability using multi-source remote sensing data, random forest, and kernel density estimation, then applied LandTrendr to detect land-use changes and track the spatiotemporal evolution of abandonment from 2000 to 2022. Next, we combined Geodetector and linear regression to identify driving factors. The results show that abandoned cropland exhibited an increasing trend from 2000 to 2010, with an average annual growth rate of 20.4%. From 2010 to 2013, the area of abandoned cropland declined rapidly, decreasing by 44.6%. Between 2013 and 2022, abandoned cropland decreased steadily, with an average annual reduction rate of 24.7%. Spatially, abandonment was clustered in the central mountains and southern hills. Key drivers included distance to towns (DtT), total grain output (GTO), and GDP. Our approach supports cropland management and rural revitalization in regions with complex terrain.

Remote Sensing doi: 10.3390/rs18101560

Authors: Runjie Wang Yuhang Liu Xianglei Liu Yahao Wu

Image-based PM2.5 estimation has emerged as a promising complementary approach to traditional physicochemical monitoring. However, achieving accurate predictions in severely polluted environments remains a critical challenge, as existing deep learning models tend to prioritize luminance variations induced by PM2.5 while neglecting the impact of complex atmospheric light interference, leading to substantial estimation errors. To address this issue, this paper proposes a novel luminance–spatial decoupling (LSD) module constructed based on L2–Lp Retinex theory and integrated into a VGG16 backbone. By establishing a prior knowledge module linking luminance to PM2.5, the proposed method achieves high-fidelity separation of atmospheric luminance (AL) and target luminance (TL) during feature extraction. TL represents the luminance variation induced by PM2.5 concentrations, whereas AL characterizes the luminance contribution arising from atmospheric light. Simulation experiments validate the reliability of the L2–Lp Retinex-based decomposition. Ablation studies reveal that the LSD module effectively mitigates haze interference in high-pollution conditions while minimizing influence on the backbone network in clear weather, thereby resolving the conflict between dehazing and feature extraction. Comparative experiments demonstrate that LSD-VGG16 significantly outperforms traditional methods and standard convolutional neural networks, achieving a minimum prediction error of 12.42 while exhibiting stronger stability against temporal variations. Furthermore, evaluation on the unseen RHID-AQI dataset without retraining confirms the model’s robust generalization capability under abrupt illumination fluctuations and diverse weather conditions.

Remote Sensing doi: 10.3390/rs18101558

Authors: Deyu Song Xiangyin Zhang Baichuan Wang Yalin Zhong Yuan Yao Kaiyu Qin

Unmanned aerial vehicles (UAVs) provide a highly flexible platform for synthetic aperture radar (SAR), enabling efficient, high-quality imaging in remote sensing applications. In realistic imaging missions, regions of interest (ROIs) usually have different sizes and spatial distributions. While deploying SAR-UAVs with heterogeneous flight and imaging capabilities can improve mission time efficiency, realizing this improvement depends critically on task assignment and path planning. In this paper, the joint task assignment and path planning problem for heterogeneous SAR-UAVs in multi-region imaging missions is addressed. First, flight and imaging models of SAR-UAVs are established, and a constrained optimization problem is formulated to minimize the mission completion time. Then, an improved clustering strategy based on area-density and cost prediction (ADCP) is proposed to align ROI-dependent imaging workloads with heterogeneous SAR-UAV capabilities, thereby leveraging capability advantages and reducing the mission completion time. Finally, a discrete secretary bird optimization algorithm (DSBOA) is developed to generate feasible, high-quality paths. To accelerate convergence, UAV paths are encoded as waypoint sequences, and a mutation-based operator is introduced to update the population. Extensive Monte Carlo simulations show that the proposed approach consistently outperforms the baselines in mission completion time, demonstrating its effectiveness in improving time efficiency for multi-region SAR imaging missions. Ablation experiments further confirm the independent contributions of the proposed ADCP method and DSBOA algorithm.

Remote Sensing doi: 10.3390/rs18101561

Authors: Yangyang Zhang Xu Han Jian Yang

The leaf area index (LAI) is a key parameter reflecting vegetation canopy structure and growth status. This study systematically compares the performance of spectral and texture features derived from Sentinel-2A imagery for LAI retrieval in winter wheat and maize. Multiple vegetation indices and gray-level co-occurrence matrix (GLCM) texture features were extracted, and three types of texture indices—Normalized Difference Texture Index (NDTI), Ratio Texture Index (RTI), and Difference Texture Index (DTI)—were constructed. Modeling was performed using Partial Least Squares Regression (PLSR) and Gaussian Process Regression (GPR). Results show that red-edge vegetation indices and mean texture features (e.g., NDVI_M) are robust predictors for both crops, with correlation coefficients reaching 0.87 for winter wheat and 0.83 for maize. Texture indices further enhance the representation of canopy structural information; the optimal NDTI achieved |R| > 0.88 for both crops, though the specific feature pairs were crop-specific. Using the proposed two-stage feature optimization strategy combined with GPR, the LAI estimation accuracy for winter wheat reached R2 = 0.87 with RMSE = 0.41 on an independent test set, while for maize the accuracy was R2 = 0.75 with RMSE = 0.38. The strategy significantly improved accuracy for winter wheat (uniform canopy) but yielded limited gains for maize (heterogeneous canopy), largely due to differences in canopy architecture. This study demonstrates that integrating multi-dimensional features with nonlinear modeling enhances LAI estimation accuracy. By providing a side-by-side comparative evaluation across two contrasting crop canopies, this study underscores the necessity of crop-adaptive feature selection and modeling strategies. The findings offer practical guidance rather than a universal model for large-scale crop monitoring in agricultural remote sensing.

Remote Sensing doi: 10.3390/rs18101557

Authors: Lingbo Meng Zhen Chen Kun Shang He Gu Yingjuan Wei

Synthetic Aperture Radar (SAR) is one of mainstream remote sensing techniques, offering all-weather, day-and-night operational capabilities. However, throughout the processes of signal transmission, propagation, and reception, it is difficult to ensure that the amplitude and phase of the SAR signal strictly follow a linear frequency modulation (LFM) characteristic. The resulting signal distortion often leads to main lobe broadening and sidelobe elevation, degrading the focusing performance of SAR images. Traditionally, this issue has been addressed primarily through SAR system internal calibration and pre-distortion compensation, which makes it challenging to maintain the signal in an ideal state over the long term. At the same time, many simplified SAR systems also lack an internal calibration design, such as low-cost UAV-borne SAR payloads. In this paper, we propose a novel signal distortion compensation method based on SAR image data. Without relying on SAR system calibration signals, this method estimates and compensates for signal distortion directly using SAR image data, thereby improving SAR image focusing performance, achieving a resolution closer to the theoretical bandwidth and lower sidelobe. The processing and analysis of both manned and unmanned airborne SAR image data and calibration signals demonstrate that the proposed method effectively compensates for signal distortion phases, achieving performance comparable to that of real-time calibration-signal-based methods.

Remote Sensing doi: 10.3390/rs18101556

Authors: Yu Cheng Xixiang Liu Shuai Chen Chuan Xu

Visual scene matching navigation (VSMN) for unmanned aerial vehicles (UAVs) boasts advantages such as high precision, high reliability, and autonomy. The biggest challenge lies in the tension between local fine-grained information and global semantics, as well as limited generalization ability in real-world environments. While existing Transformer-based cross-view geolocation methods enhance global context modeling capabilities, they still generally face issues such as high demands on training data and computational resources, insufficient fusion of local fine-grained information and global semantics, and real-time performance in real-world complex environment. To address these problems, we propose a scene matching and localization algorithm based on the Siamese-ViT. For feature extraction, we use the ViT model to extract global features and K-means clustering to aggregate local features. Combined with the global features extracted by the ViT, a robust local–global feature representation vector is generated. For feature matching, incremental principal component analysis (IPCA) is used to reduce the dimensionality of the high-dimensional feature space, and a KD-tree is constructed for fast feature retrieval to improve matching efficiency. We validated our algorithm on the University-1652 dataset and a dataset of real-world satellite-drone image pairs. The results show that our Siamese-ViT outperforms other models in both Recall and AP. We conduct flight experiments in real-world environments, capturing drone images of complex scenes, including farmland, urban buildings, and waterways. The results show that, at a flight altitude of 350 m, our algorithm achieves an average absolute value of 6.2063 m for latitude, 6.7552 m for longitude, and 10.1922 m for horizontal error. Therefore, our Siamese-ViT demonstrates ideal overall positioning accuracy.

Remote Sensing doi: 10.3390/rs18101559

Authors: Chuting Hu Size Dai Shifan Wu Qiaolin Ye He Yan

Accurate individual tree crown (ITC) segmentation from unmanned aerial vehicle (UAV) imagery is important for fine-scale forest inventory, plantation management, and ecological monitoring. However, delineating ITCs in dense plantation environments remains difficult because crowns are strongly adjacent, canopy structures are highly homogeneous, and crown boundaries are often blurred, making it hard for existing methods to preserve both regional integrity and boundary continuity. This study proposes the Perceptual Segment-Anything Model with Multi-head Cross-Parallel Attention (Per-SAM-MCPA), a lightweight and effective framework for fine-grained ITC segmentation in dense plantation scenes. Based on a compact ResNet-50 backbone, the framework integrates perceptual target-aware representation, multi-scale detail enhancement, global contextual modeling, and semantic-boundary collaborative refinement to improve crown discrimination and structural consistency. A perceptual relation module is used to strengthen pixel-level semantic dependency modeling, and a Multi-head Cross-Parallel Attention (MCPA) mechanism is designed to capture long-range contextual interactions through orthogonally decomposed spatial attention, improving global geometric consistency with limited computational overhead. A Composite Constraint Loss (CCL) that combines a weighted cross-entropy loss, a structural similarity loss, and a boundary term based on Hausdorff distance is introduced to jointly optimize region-level segmentation quality and boundary fidelity. Experiments on the Catalpa bungei UAV dataset show that the proposed method achieves an intersection over union (IoU) of 87.3% and an F1-score of 91.0%, outperforming representative baseline methods such as SAM and Mask R-CNN while maintaining an inference speed of 35.7 FPS on a single GPU. These results indicate that Per-SAM-MCPA offers an accurate, efficient, and practical solution for ITC segmentation in dense plantation environments.

Remote Sensing doi: 10.3390/rs18101555

Authors: Jianbin Huang Yuntao He Yinuo Zhang Xiaolu Li Lijun Xu

With the growing demand for precise absolute pose estimation of landers in lunar exploration missions, crater database-based navigation technology has become a core path to achieving this goal, but it faces challenges of low efficiency in large-scale data retrieval and insufficient matching robustness. To address these issues, a coarse-to-fine crater matching framework with database fast searching and robust triangle similarity matching is proposed. A Geo-KD search algorithm is designed to realize fast and accurate retrieval of craters within the field of view by combining Geohash and KD-tree. A robust triangle similarity matching algorithm is constructed through local neighborhood crater screening, triangle similarity matching, and mismatching elimination based on Random Sample Consensus (RANSAC) and Local Motion Consistency (LMC). Experiments show that the algorithm achieves an average retrieval time of 20 ms with an F1-score of 0.8 for the global lunar database with 1.29 million craters. It has an F1-score more than 0.746 and a single-frame matching time less than 1.005 s under lunar orbital phase, landing phase, and different camera pitch angles, outperforming other advanced algorithms and meeting on-orbit real-time requirements, providing reliable support for the absolute pose estimation of lunar probes.

Remote Sensing doi: 10.3390/rs18101552

Authors: Mengchun Qin Junzheng Liu Xinyu Liu Haijue Xu Yuchuan Bai

Braided rivers are difficult to monitor because of unstable mainstream migration, complex planform morphology, and intense channel adjustment. To address this challenge, this study develops an integrated remote-sensing framework that links cross-sensor surface-water extraction, geometry-reliable boundary reconstruction, and river-geometry metric derivation for channel dynamics monitoring. Using the braided reach of the Lower Yellow River (LYR) as the study area, the framework was applied to investigate abnormal channel dynamics during 1986–2025. Results show that the improved deep learning model achieved robust and consistent surface-water extraction across Landsat-8, Landsat-7, and Sentinel-2 imagery, while the boundary reconstruction procedure effectively reduced raster-induced jagged artefacts and improved the geometric reliability of extracted channel boundaries. Based on the reconstructed boundaries, water-surface width, river centerline, sinuosity, and the Deviation Degree from Regulated River Alignments were derived and used to identify abnormal channel-dynamics reaches. In the braided reach of the LYR, the results revealed clear spatial concentration, temporal intermittency, and an upstream shift in abnormal-reach occurrence after 2000. Overall, the proposed framework extends remote sensing from surface-water mapping to long-term, geometry-reliable monitoring of braided-river channel dynamics and provides practical support for potentially unstable reach screening and warning-oriented river management.

Remote Sensing doi: 10.3390/rs18101554

Authors: Yihan Wang Zhenhua Mu Hanyu Zhang Chuanming Song Xianghai Wang

In hyperspectral image (HSI) anomaly detection (AD) methods, detecting small targets or anomalies remains challenging. This difficulty arises because targets or anomalies may vary significantly in size, shape, and texture, causing them to be obscured by larger-scale background features. To address the above issue, this paper proposes an unsupervised multi-scale feature fusion network (UMF2Net) for HSI-AD. Firstly, central difference convolution analyzes the image at multiple scales to capture fine-to-coarse details and structural information. Additionally, three-dimensional (3D) convolution is employed to generate feature weights for the multi-scale features, assigning different weights to features with different contributions so that the model dynamically emphasizes features that have a greater impact on the AD results. Finally, by using the two proposed multi-scale feature fusion modules, the model effectively integrates features at different scales, thereby enhancing its ability to detect anomalies of varying sizes. Compared with several classical HSI-AD algorithms on real hyperspectral datasets from four scenarios, UMF2Net achieved competitive detection results, verifying the effectiveness of our algorithm.

Remote Sensing doi: 10.3390/rs18101553

Authors: Bing Zhang Yongjie Du Weidong Song Jichao Zhang Hongchang Sun Dongfeng Ren

Surface deformation is characterized by long-term accumulation and significant spatial differences, commonly inducing ground fissures and structural damage to roads and buildings, and in severe cases, causing collapses and other accidents that directly threaten human life. Reliable deformation monitoring and prediction are of great significance for early warning and infrastructure safety. Synthetic aperture radar interferometry (InSAR) technology can be used to acquire high-spatiotemporal-resolution temporal observations of surface deformation over a large area, but research on surface deformation prediction using InSAR temporal images is still relatively limited. Therefore, in this study, we used Sentinel-1 temporal imagery as the data foundation. Firstly, small baseline subset (SBAS)-InSAR inversion was used to obtain the line-of-sight-oriented cumulative deformation sequence and subsidence rate results. Based on this, the causal multi-trend fusion patch-to-point (CMTF-P2P) surface deformation prediction framework was developed. This framework effectively separates the long-term trends and short-term fluctuations in the deformation sequence through causal decomposition; introduces external drivers such as rainfall and event phase characteristics to enhance the temporal expression; and utilizes patch-to-point fusion of neighborhood spatial information to improve the spatial continuity and the ability to characterize local differences. Experimental results show that the method has an RMSE of 0.43 mm and an R2 of 0.92, outperforming time-series deformation prediction models such as LSTM and iTransformer. Compared with traditional models that learn overall change patterns from historical time series, this paper alleviates the confusion between trends and volatility using causal STL decomposition, making the model’s expression of long-term trends and seasonal and short-term fluctuations in volatility clearer; event phase encoding enhances the model’s ability to characterize sudden disturbances and phased responses to events; the patch-to-point structure incorporates spatiotemporal information within the neighborhood, enhancing the model’s ability to apply spatiotemporal information; multi-branch collaborative modeling enhances the model’s ability to characterize multi-scale temporal features; and incremental cumulative consistency constraints enhance the physical consistency of the prediction results.

Remote Sensing doi: 10.3390/rs18101551

Authors: Baozhu Jia Zekun Guo Jin Xu Xinru Dong Lilin Chu Zheng Li Haixia Wang

X-band marine radar offers unique advantages for monitoring nearshore oil spills. However, oil films and sea clutter exhibit high pixel intensity overlap in radar images. Traditional threshold segmentation and machine learning methods have certain limitations in terms of feature extraction, Region of Interest (ROI) guidance, threshold optimization adaptability, and unsupervised capabilities. To address these issues, a method of oil film detection for ship radar based on multi-dimensional feature-guided extraction and hybrid optimization search is proposed. By combining Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering with multidimensional features, this method automatically extracts ROIs under unlabeled conditions, effectively suppressing sea clutter interference. Subsequently, an improved Beaver Behavior Optimizer (BBO) and simulated annealing (SA) hybrid algorithm (BBO-SA) is introduced within the ROIs, along with a designed adaptive temperature update strategy, to achieve coordinated optimization of global and local searches. The experimental results demonstrate that the method described in this paper performs exceptionally well across all evaluation metrics, confirming its accuracy and robustness in oil film detection. It provides a viable technical approach for emergency monitoring of nearshore oil spills.

Remote Sensing doi: 10.3390/rs18101550

Authors: Myrta Maria Macelloni Fabio Giulio Tonolo Vincenzo Di Pietra Umberto Morra di Cella Alberto Cina

Alpine glaciers are a vital resource for mountain regions. They provide water reserves, support energy production and tourism, and promote biodiversity. However, they are highly susceptible to climate change. In fact, they are recognised as being among the areas most affected by, and increasingly exposed to, natural hazards. The Rutor glacier in Aosta Valley, Italy, which has been the subject of repeated measurements since the 19th century and currently covers an area of around 8 km2, is undergoing significant and continuous retreat. It thus serves as an exemplary case study of the impact of climate change on the Italian Alps. This ongoing research has made it possible to conduct multi-temporal analysis of the glacier. Within this framework, Politecnico di Torino, in collaboration with ARPA Valle d’Aosta, has developed a multidisciplinary research approach focused on the characterisation of alpine environments. This study illustrates the geomatic workflows and derived geospatial products that can be used to carry out a 4D monitoring of the extent and volume of the Rutor Glacier and estimate its mass balance over the past six years. A specific focus of the study is the propagation of errors in multi-temporal analyses used to quantify glacier melt, with particular attention to the precision of input 3D geospatial data and to the Limit of Detection of elevation differences, ultimately enabling the estimation of the uncertainty associated with the derived quantities and their temporal trends. Finally, advantages and limitations in the multi-temporal and multi-sensor monitoring of glaciers are presented and discussed.

Remote Sensing doi: 10.3390/rs18101549

Authors: Yi Zhang Chengke Ren Jianyu Li Zhaojun Song

The Yellow River Delta, a young alluvial plain in China, is experiencing severe land subsidence that threatens its ecological security and sustainable development. However, the driving mechanisms of this subsidence exhibit strong spatial heterogeneity, which traditional global models fail to capture. This study integrates high-precision subsidence measurements from Sentinel-1A imagery and SBAS-InSAR technology (2017–2023) with multi-source environmental factors (topography, geology, land use, precipitation) to propose a Spatially Adaptive Ensemble Learning Model with feature selection (SA-GSE). The model concatenates predictions from base learners (CatBoost, XGBoost, Random Forest) with spatial features (e.g., distance to salt pans, local topographic variance) to form meta-features, which are then input into a multilayer perceptron meta-learner. Through 5-fold spatial cross-validation, SA-GSE learns spatially dynamic base-model weights, implicitly adapting to regional variations in subsidence drivers. The model achieves an R2 of 0.7810 and RMSE of 40.55 mm/yr on the test set, outperforming individual base models and ordinary stacking. Residual spatial autocorrelation is substantially reduced, with SA-GSE yielding the lowest Moran’s I (0.0334, p = 0.206) among all evaluated models, confirming effective capture of spatial heterogeneity. Driving force analysis reveals that distance to salt pans is the most important predictor (permutation importance: 0.4456), underscoring the dominant role of brine extraction-induced aquifer compaction. Lagged precipitation importance (0.3191) exceeds that of current precipitation (0.2453), indicating a recharge lag effect. SHAP interaction analysis uncovers a nonlinear “precipitation decoupling” mechanism in salt pan areas, where high precipitation paradoxically exacerbates subsidence. The resultant map of predicted subsidence rates highlights elevated rate zones in the northern salt pans and along the Guangli River. While the map does not represent a full risk assessment—as it does not include exposure or vulnerability—it provides a spatially explicit estimate of hazard likelihood. This ensemble framework yields novel perspectives on subsidence drivers in heterogeneous regions and can support land subsidence prevention and groundwater management planning.

Remote Sensing doi: 10.3390/rs18101547

Authors: Jianfeng Han Feijie Sun Zihan Xu Lili Song Jiandong Fang

In vision-based low-altitude unmanned aerial vehicle (UAV) remote sensing, detecting small targets accurately and maintaining stable tracking under fast-motion conditions remain significant challenges. Specifically, small-object detection suffers from low feature representation, while camera motion often induces tracking drift and identity switches. To address these issues, this paper proposes a novel small target detection and tracking algorithm named TCYOLO-SofByteTrack, which integrates an improved YOLOv11 with ByteTrack. The algorithm comprises two core innovative modules: First, the TCYOLO detector is designed by integrating the C3k2-TA feature enhancement module with triplet attention mechanism to achieve cross-dimensional interaction modeling, significantly improving small target feature representation capability and network contextual awareness. A Cross-Scale Feature Fusion Module for UAVs (CCFM-UAV) is constructed to provide precise detection support for small targets at different scales. Second, building upon the ByteTrack framework, the SofByteTrack tracker is designed, which introduces a sparse optical flow-based motion compensation strategy. This strategy estimates and compensates for image displacement caused by UAV motion in real time, ensuring the stability of target bounding boxes under fast-motion conditions, thereby effectively mitigating tracking drift and identity switches. Experimental results demonstrate that the TCYOLO detector achieves a 7.4% improvement in mAP for small target detection compared to the baseline YOLOv11 model. The complete TCYOLO-SofByteTrack tracking algorithm achieves a HOTA score of 45.3%, MOTA of 42.7%, and IDF1 of 57.8%, representing improvements of 4.5%, 5.9%, and 8.0%, respectively, over the baseline methods. Furthermore, the number of successfully tracked targets increased by 37.3%, while identity switches decreased by 23.4%. These results demonstrate the notable advantages of the proposed method in small target detection accuracy, tracking precision, and identity consistency. Its generalization capability is further validated on a custom highway inspection dataset. Moreover, deployment tests on an NVIDIA Jetson Orin NX platform show that, compared to YOLOv11n, the proposed algorithm achieves higher detection accuracy while still meeting real-time processing requirements, highlighting its practical applicability in resource-constrained scenarios.

Remote Sensing doi: 10.3390/rs18101548

Authors: Violetta Sokoła-Szewioła Paweł Sopata Dawid Mrocheń

This study investigates vertical surface displacements in an area previously impacted by extensive underground hard coal extraction, specifically focusing on the closed “Kazimierz-Juliusz” mine in the Upper Silesian Coal Basin (Poland). The cessation of mining operations and formal decommissioning do not necessarily signify the termination of ground instability; rather, the discontinuation of mine water pumping triggers a progressive groundwater rebound within the rock mass. This hydrogeological shift leads to a redistribution of stresses in the geological structure, inducing deformation processes that manifest as surface uplift. This research aims to characterize the temporal evolution and magnitude of post-closure surface elevation changes by integrating satellite radar interferometry with conventional geodetic surveys. The analysis, spanning a 28-month observation period, utilizes both Persistent Scatterer Interferometry (PSInSAR) and European Ground Motion Service (EGMS) data, complemented by precise geometric leveling. The results reveal a low-magnitude deformation process, with detected uplift rates reaching approximately 1 cm/year. The synergistic integration of InSAR-based monitoring and classical geodesy allowed for robust cross-validation, significantly enhancing the reliability of the findings both qualitatively and quantitatively.

Remote Sensing doi: 10.3390/rs18101546

Authors: Andrea Viñuales Nicolas Younes Mbam Itumo Marta Yebra Ignacio de la Calle Javier Madrigal

Given the increasing frequency, severity, and socioecological impacts of wildfires, there is an urgent need for robust frameworks to better characterize fire behavior and flammability patterns across ecosystems to support early warning, mitigation, and management strategies. However, flammability remains difficult to quantify and scale, as it involves multiple interacting components that are typically measured at the bench scale. This study aimed to establish empirical links between spectral information, plant traits, and flammability metrics, and to scale these relationships to satellite imagery to translate these metrics into a spatial context. We combined laboratory spectroscopy, plant trait measurements including leaf mass per area, carbon, and cellulose, and combustion experiments using a simple and reproducible burning device. In total, 84 samples were collected and analysed, allowing us to characterise how spectral signatures relate to vegetation traits and fire behaviour. Spectral indices were developed to estimate plant traits, which were subsequently used as predictors in flammability models. These models were then transferred to Environmental Mapping and Analysis Program (EnMAP) hyperspectral imagery to derive spatial estimates across eucalypt forests and grasslands of the Australian Capital Territory (ACT). Spectral information distinguished fuel types and captured variability of the plant traits, while these traits showed associations with combustion behaviour. Based on these links, the best-performing model predicted the rate of temperature increase, a combustibility metric, in eucalypt forests (R2 = 0.70; Root Mean Square Error = 32.48 °C/s). In contrast, grassland models showed limited predictive performance, likely due to weaker relationships between plant traits and flammability metrics. Overall, this study demonstrates a practical and scalable approach for deriving flammability maps from hyperspectral and in situ data, highlighting the potential of plant-trait-based remote sensing. The resulting maps should not be interpreted as standalone fire risk products, but rather as a characterization of the structural and biochemical drivers of flammability. The main constraint of this work is the limited sample size. Future research should expand spatial and temporal coverage to better capture vegetation variability and enable the inclusion of independent validation datasets. Exploring alternative combustion protocols and testing more advanced spectral modelling approaches for trait estimation would provide additional insights.

Remote Sensing doi: 10.3390/rs18101544

Authors: Yang Zhao Ruoxuan Li Xiangzhen Kong Cheng Cheng Yijian Chen Kangkang Zhuang Yinping Liu Qilin Zhang

Continuous monitoring and nowcasting of tornadic near-storm environments remain challenging, particularly in regions with limited ground-based weather radar coverage. High-spatiotemporal-resolution geostationary satellite remote sensing offers a valuable approach to track the evolution of severe convective storms. Combining 10 min cloud-top brightness temperature (TBB) data from the Himawari-8 satellite and ERA5 reanalysis, this study investigates the atmospheric environments of 177 documented tornadoes in China from 2016 to 2023. Tracking storm convective centers using TBB minima reveals clear regional differences in tornadogenesis paradigms. Southern China tornadoes exhibit a “dynamically driven” pattern within quasi-steady, warm, and moist environments. These environments feature low Lifted Condensation Levels (LCL; ~790 m) and weak Convective Inhibition (CIN). Intense low-level wind shear and storm-relative helicity (SRH) dominate the convective triggering. Northern China tornadoes follow a “coupled thermodynamic-kinematic” paradigm under relatively drier and cooler backgrounds. Their initiation relies on the rapid, synchronized accumulation of Mixed-Layer convective available potential energy (MLCAPE) and deep-layer SRH. Furthermore, intensity-based comparative analysis indicates that significant tornadoes (Enhanced Fujita [EF] scale, EF ≥ 2) are favored by higher MLCAPE, deep-layer shear, and lower LCLs compared to weak ones (EF ≤ 1). Himawari-8 TBB data capture a more rapid pre-storm convective cloud-top cooling for strong tornadoes, with medians reaching −73 °C. This study demonstrates that combining high-frequency satellite observations with reanalysis data provides quantitative precursor signals for regional severe tornado nowcasting.

Remote Sensing doi: 10.3390/rs18101542

Authors: Tao Li Jiang Li Hongzhe Jiang Lei Jiang Xiyun Jiao Yue Luo

Field water status is a critical variable for agricultural water management. In recent years, the development of space–air–ground multi-platform collaborative observation and data fusion technologies has provided new options for precision monitoring. However, challenges in applicability, robustness, and transferability persist. This study employs bibliometric analysis to systematically synthesize the literature, revealing that research has evolved from single-point observations to multi-platform synergy. Satellite, unmanned aerial vehicle (UAV), and ground-based monitoring are analyzed, as well as challenges in multi-source data fusion, including scale mismatch, error propagation, and uncertainty quantification. Finally, applicability and other barriers are evaluated across three typical agricultural scenarios: large-scale surface soil moisture monitoring, crop root zone soil moisture retrieval, and paddy field water depth estimation. The results indicate that space–air–ground collaborative observation constitutes a mature framework, with satellite and ground-based monitoring as core components and UAV technology as a supplement. However, scale transformation and error propagation mechanisms in multi-source data fusion remain unresolved. Currently available vertical water information is limited, and quantitative retrieval has yet to achieve the reliability required for operational applications. This limitation is particularly evident in paddy field water depth retrieval and root zone soil moisture retrieval. This review provides a theoretical reference for precision field water status monitoring and identifies future research priorities, including the integration of physical mechanisms with machine learning (ML) in multi-source data fusion, as well as error quantification and paddy field water depth retrieval.

Remote Sensing doi: 10.3390/rs18101545

Authors: Jinyong Chen Tingyu Zhu Gang Wang

There are differences in imaging characteristics between infrared and visible light images: visible light images can provide rich texture and color information, but imaging is limited in harsh weather conditions. Infrared images are based on the target’s thermal radiation characteristics and have the ability to resist environmental interference but lack details and background information. Effectively integrating the two can significantly enhance scene understanding ability and improve environmental perception and target recognition performance in applications such as intelligent driving. However, existing fusion methods still face challenges, especially in complex scenes where it is difficult to balance the full preservation of target information with the complete presentation of background details, often resulting in difficulties in extracting differentiated features from different modalities. This article proposes a target detection method based on the visible light infrared fusion diffusion model. This method introduces the Stable Diffusion architecture and designs a target perception spatial fusion weight module that can adaptively generate a spatial fusion weight map based on modal differences. By implementing a multi-stage dynamic fusion strategy, the fusion ratio is automatically adjusted at different diffusion stages. A full-step multi-step prediction mechanism is adopted to improve fusion quality and stability. Compared with existing methods, the method proposed in this article has significant advantages. Experiments on multiple publicly available datasets have shown that this method outperforms existing mainstream methods in key metrics such as Peak Signal to Noise Ratio(PSNR), Mean Square Error(MSE) , and ean Absolute Error(MAE) and also demonstrates good detection performance in downstream tasks for object detection.

Remote Sensing doi: 10.3390/rs18101543

Authors: Zhengshen Huang Weili Kou Chen Zheng Guangzhi Di Qixing Zhang Chenhao Ma

Forest fires cause severe ecological and economic losses, so timely and accurate detection becomes crucial for effective prevention and control. Edge devices with intelligent algorithms can detect forest fires in real time. Current deep learning algorithms can achieve high accuracy, but they are not suitable for edge devices because they require substantial computing resources. To address this issue, this study proposes a real-time lightweight forest fire detection network (RLFNet) improved from YOLOv11n, with three key enhancements to the backbone, neck, and head. (1) A Parallel Multi-Scale Extraction Block (PMEB) improves C3k2 with a dual-branch parallel strategy to enhance multi-scale feature extraction efficiency; (2) a Bidirectional Cross Fusion Module (BCFM) replaces simple Concat with a context-aware cross-gating mechanism to suppress background noise and reduce false alarms; and (3) a Faster Inference Detection Head (FIDH) leverages structural re-parameterization and group normalization to boost inference efficiency while reducing parameters. In addition, a Layer-Adaptive Magnitude-based Pruning (LAMP) strategy is applied to further improve model’s computational efficiency. Experimental results on the self-constructed Diverse Fire Scenario (DFS) dataset demonstrate that RLFNet reduces parameters and GFLOPs by 25.2% and 20.6%, boosts mAP50 by 5.3%, and achieves an inference speed of 225 FPS, attaining the best accuracy and speed among the compared models. Validation on a public remote sensing dataset further confirms its strong generalization. These results indicate that RLFNet provides a high efficiency and lightweight solution for edge devices to real-time detect forest fires.

Remote Sensing doi: 10.3390/rs18101541

Authors: Zongqing Cao Jianchao Yang Wang Sun Xingyu Lu Ke Tan Zheng Dai Wenchao Yu Hong Gu

Accurateestimation of clutter measurement density (CMD) is crucial for radar-based multi-target tracking (MTT), especially under spatially non-uniform and temporally varying environments. Existing methods, including finite mixture models, kernel density estimation, and normalizing flows, often require scene-specific tuning and exhibit limited generalization. To address these limitations, we propose HyperNCMD, a scene-adaptive CMD estimator that employs hypernetworks to dynamically generate the parameters of normalizing flows. To capture spatial variability, radar measurements are first embedded using Random Fourier Features (RFFs), and then processed by a spatio-temporal encoder that jointly models spatial structures and temporal clutter dynamics. The hypernetwork leverages the encoded embedding to adaptively produce flow parameters, enabling flexible CMD estimation across diverse environments. Lightweight data augmentation is further applied to make the estimator more robust across diverse environments, while a Feature-wise Linear Modulation (FiLM)-based fine-tuning scheme enhances test-time adaptation. Experiments on both synthetic and real radar datasets demonstrate that HyperNCMD achieves superior accuracy and robustness, achieving up to 10.5% reduction in per-point negative log-likelihood under dynamically varying conditions. These results highlight the potential of hypernetwork-driven CMD modeling for reliable radar perception in complex sensing environments.

Remote Sensing doi: 10.3390/rs18101540

Authors: Nikolaos L. Tsakiridis Uta Heiden Nikolaos Tziolas

Soil is the foundation of terrestrial life, underpinning ecosystem services, food production, and climate regulation. [...]

Remote Sensing doi: 10.3390/rs18101539

Authors: Wei Liang Hong Hua Qiling Sheng Yuebin Ding Lili Tu

Land surface temperature (LST) and its diurnal variation are critical for understanding the surface energy balance and water cycle processes. Traditional diurnal temperature cycle (DTC) models are widely used to reconstruct continuous temperature sequences from sparse satellite observations. However, these models rely on the idealized assumption of an isotropic surface and ignore the thermal radiation directionality caused by viewing geometry, which introduces substantial errors over heterogeneous surfaces. Thus, incorporating angular effects into DTC modeling has become an effective approach to improving LST simulation accuracy. This review traces the progress of DTC models from isotropic to anisotropic representations. First, we summarize the development and inherent limitations of conventional isotropic DTC models. Then, we synthesize representative angular-coupled models, ranging from early simple component-based models to recent kernel-driven coupling methods, and compare their physical assumptions, data requirements, parameter complexity, and applicable scenarios. Although these coupled models can significantly improve fitting accuracy over heterogeneous surfaces, they still face challenges. These include strict data requirements, limited all-weather applicability, a lack of nighttime angular correction, and incomplete validation systems. Future research can advance through multi-source data fusion, hybrid modeling strategies, and robust validation systems. These are key to generating high-precision, spatiotemporally consistent LST data.