In remote sensing image oriented object detection tasks, weakly supervised learning methods based on horizontal bounding boxes have attracted much attention due to their lower annotation costs compared to fully supervised methods. However, remote sensing images, characterized by complex backgrounds, exhibit a wide range of target scales and diverse geometric characteristics across target categories. Existing methods exhibit inadequate exploitation of background and angular information under weak supervision, resulting in compromised perception of dense and high-aspect-ratio targets. Neglecting the imbalance in angle estimation samples further leads to excessively low detection accuracy for few-shot categories. To address the aforementioned issues, this paper proposes a Geometry-Aware Enhancement Network (WSOOD-GAEN) for weakly supervised oriented object detection tasks. First, in the backbone network stage, a channel-space deformable attention module (DAE-ResNet) was constructed. Through deformable sampling and screening of key regions, feature extraction has both morphological adaptability to complex shapes and semantic discriminability of key features in complex backgrounds. Secondly, in the feature pyramid stage, an Angle-Guided Feature Pyramid Network (AG-FPN) is proposed. This module dynamically applies rotation transformation to the sampling offsets of deformable convolutions, thereby enhancing the feature representation of objects with different orientations and scales. Furthermore, an adaptive geometric perception loss (AGL) was designed. Based on the geometric characteristics of different categories, it automatically learns differentiated rotation and flip consistency weights, thereby improving the prediction accuracy of small sample categories. Experiments on the DOTA-v1.0, HRSC, and RSAR datasets validate our approach. Specifically, under the AP75 evaluation metric, the proposed method outperforms existing weakly supervised methods by 1.51%, 9.86%, and 3.28%, respectively.

Effective crop monitoring during monsoon growing seasons in Central India faces challenges from persistent cloud cover that limits optical remote sensing during critical agricultural periods. This study presents the first attempt to develop a novel set of SAR-derived phenological metrics organized into five thematic categories for monsoon crop discrimination in smallholder agricultural systems. Five major monsoon crops (cotton, rice, maize, soybean, and urad) were analyzed across five different agroclimatic zones in Central India using Sentinel-1 data for the 2021 growing season. Phenological features were extracted from VV, VH polarizations, and their ratio, including seasonal extrema, threshold crossings, duration measures, curve shape descriptors, and area under the curve. Distinct crop-specific signatures were observed, with cotton showing extended phenology and cereal–legume crops displaying compressed, overlapping growth patterns. VV polarization achieved the highest statistical discrimination for intensity-based metrics, with 75% thresholds (VV_HP75V: F = 1287) providing higher separability than other thresholds by capturing near-peak biomass differences. VH performed best for duration and integration-based metrics, while VH/VV provided limited additional separability across metric types. For area-under-the-curve metrics, AUC25 outperformed AUC50 and AUC75 by capturing cumulative backscatter across the broader growing season while remaining robust to soil- and residue-dominated backscatter variability at sowing and harvest. Multiclass classification achieved 48.3% overall accuracy with systematic cereal–legume confusion, reflecting fundamental phenological convergence among monsoon-aligned crops. Cotton achieved the highest performance (F1: 0.79), with VH polarization dominating feature importance (65% of top 20 features). Binary classification revealed crop-specific discrimination patterns: cotton was best separated using VV intensity metrics, maize using the VH/VV ratio, and rice using timing-based features. Cross-district transferability showed the highest mean overall accuracy for rice (74%) and cotton (72%), while the remaining crops showed lower accuracy due to their phenological similarity. These findings highlight both the potential and limitations of SAR phenological metrics for monsoon crop discrimination, with effective results for structurally distinct crops but persistent cereal–legume confusion, requiring further investigation with multi-sensor approaches.

Accurate estimation of global terrestrial evapotranspiration (ET) is fundamental for understanding the Earth’s water and energy cycles, yet existing multi-source ET products inevitably contain uncertainties that require spatially explicit characterization for optimal data merging or data assimilation. While Quadruple Collocation Analysis (QCA) offers a robust and reference-free approach to quantify uncertainties, its reliability in the ET discipline remains underexplored, and algorithmic non-convergence frequently results in substantial spatial data gaps. To address these limitations, this study evaluated the accuracy of the QCA method using validation errors derived from high-quality FLUXNET sites (N = 55). Moreover, we employed a Random Forest (RF) framework that is driven by 17 environmental variables to generate spatially seamless error maps for four mainstream ET products, i.e., ERA5, GLDAS, GLEAM, and MERRA2, from 2000 to 2020. Results demonstrate that QCA-based errors strongly correlated with ground-based errors as Pearson’s correlation coefficient was >0.3 for all four ET products. Furthermore, the RF model successfully reconstructed the spatial gaps in QCA errors, achieving an exceptionally low mean prediction error of approximately 0.03 mm/day. Based on these seamless maps, the global mean ET error is estimated at roughly 0.3 mm/day, with pronounced high-error clusters emerging in regions such as central Canada and northern Argentina driven by underlying land cover heterogeneity. Ultimately, this seamless gap-filling redefined the global map of product with the lowest estimated collocation error. ERA5 emerged as the superior choice across approximately 45% of the land surface (predominantly in the tropics and mid-to-high latitudes). Meanwhile, before algorithmic gap-filling, GLEAM was optimal across approximately 28% of the valid land pixels; after spatial gap-filling, it proved most effective across approximately 30% of the globe, particularly within arid deserts and glaciated regions. Our work provides useful geographic guidance for optimizing multi-source data merging and land data assimilation frameworks in future global hydrological studies.

Accurate and scalable wetland mapping remains challenging due to strong spatial heterogeneity and limited availability of reference data. Spatial transferability of classification algorithms offers a promising solution by enabling models trained in one region to be applied to other sites, but its effectiveness depends on the degree of domain shift, algorithm robustness, and data representation. In this study, we evaluate this ability for wetland mapping using multitemporal Sentinel-2 data across two wetland systems in France: the Camargue and the Étangs de la Champagne humide. Classification is performed for three main land-cover classes—open water, aquatic vegetation, and terrestrial vegetation—using one neural network (MLP), one deep-learning model (InceptionTime), and two machine-learning algorithms (Random Forest and XGBoost), and three feature configurations (spectral bands, spectral indices, and their combination). Results reveal that when models are trained on Camargue and applied to Champagne, the highest OA reaches 90% (using InceptionTime and XGBoost), when models are trained on Champagne and applied to Camargue, the highest OA reaches 84% (using InceptionTime and XGBoost), corresponding to a decrease of 6% in OA. Within the selected algorithms, InceptionTime and XGBoost achieve the highest OA in both transfer directions. Combining spectral bands and indices improves classification performance of InceptionTime and MLP by up to 8%, while XGBoost and RF perform better using band data (5% higher OA than the combination). Class-wise analysis highlights substantial differences in transferability. Terrestrial vegetation shows the highest and most stable performance across the tested configurations, with F1-scores up to 92%, followed by open water (F1 up to 88%), while aquatic vegetation remains the most challenging class to transfer, with F1-scores up to 85% depending on algorithm and configuration. Annual time series benefit aquatic vegetation, whereas shorter series covering only the vegetation growing season remain sufficient for more stable LC classes (terrestrial vegetation). InceptionTime and MLP show higher performance using annual time series, while RF and XGBoost perform better using short time series. Overall, these results highlight the potential of spatial transferability for wetland mapping within the context of the two studied sites, although further validation across a broader range of wetlands is required.

Accurate estimation of forest above-ground biomass (AGB) is essential for quantifying forest carbon stocks and supporting regional carbon accounting. However, regional AGB mapping requires the integration of field observations with satellite data, and the associated scale transformation often causes the loss of spatial detail and reduced estimation consistency. To address this issue, this study proposes a forest canopy height-constrained area-to-area regression kriging (CCAM) method for upscaling UAV-derived AGB and generating a high-precision wall-to-wall AGB map for artificial forests in the sandy lands of northwest Liaoning Province, China. The framework integrates RFE-SVM-based feature selection, XGBoost-based UAV-AGB modeling, and CHM-constrained residual correction within a Regression-then-Kriging (R-K) strategy, while also evaluating the effects of moving-window size, scale transition, and the order of regression and kriging on upscaling performance. The results showed that the reconstructed UAV-AGB model achieved the highest accuracy, with R2 = 0.91 and rRMSE = 0.12, providing a reliable 0.1 m AGB baseline for subsequent upscaling. Among the tested moving-window sizes, the 7×7 window was identified as optimal. Under this setting, CCAM achieved R2 = 0.81 and rRMSE = 0.08, substantially outperforming direct GF-2-based estimation (R2 = 0.49, rRMSE = 0.24). The final 2 m regional AGB map further attained a validation accuracy of R2 = 0.79 and rRMSE = 0.18. These results demonstrate that CCAM can effectively preserve fine-scale UAV-derived biomass information during scale transformation and provide a reliable pathway for linking UAV and satellite observations in regional forest AGB mapping.

Atmospheric temperature and humidity profiles are essential for numerical weather prediction and severe weather monitoring. To effectively utilize data from the Geostationary Interferometric Infrared Sounder (GIIRS) onboard the FY-4 satellite, this study proposes a retrieval method based on a radial basis function (RBF) neural network, which integrates numerical model background profiles with GIIRS simulated radiance errors to construct a mapping from these two inputs to background profile errors. A channel selection strategy is developed using correlations between background errors and radiance errors to identify channels sensitive to temperature and humidity variations at different pressure levels. Experiments are conducted using data from land stations in Zhejiang Province, China, from August to December 2024, including 829 clear-sky and 2109 cloudy profiles. Under clear-sky conditions, the method reduces temperature and humidity root mean square error (RMSE) by approximately 39% and 22.3% compared to background profiles. Under cloudy conditions, despite severe radiation interference, RMSE reductions of 38.5% for temperature and 15.3% for humidity are achieved, with notable improvements below 900 hPa and above 750 hPa for humidity. Compared with the multilayer perceptron (MLP) method, RBF shows superior performance under all test conditions, especially in cloudy-sky humidity retrieval. The proposed approach provides an effective, physically constrained framework for operational GIIRS data application in temperature and humidity retrieval.

Municipal solid waste management remains a major environmental challenge across Africa, where rapid urbanisation has outpaced formal waste infrastructure and routine landfill monitoring is often absent. Rather than proposing a classification algorithm, this study investigates whether spaceborne hyperspectral imagery can reveal robust spectral fingerprints of landfill surfaces suitable for automated detection. Eight landfill sites across seven African countries were analysed using Hyperspectral Imager Suite (HISUI) data (400–2500 nm, 20 m resolution). A segment-based framework was applied after masking low signal-to-noise regions, combining brightness analysis, L2-normalised spectral shape comparison using Spectral Contrast Angle (SCA), and derivative spectroscopy across 109,275 pixels from six land-cover classes. Brightness-based discrimination exhibited strong inter-site variability, limiting its general applicability. In contrast, shape-based metrices revealed consistent separability between landfill-active surfaces and soil or urban classes in the shortwave infrared (SWIR), particularly within the 1538–1750 nm and 2075–2474 nm regions. Derivative analysis further identified stable extrema near approximately 1700 nm and 2200–2300 nm across all sites, indicating reproducible curvature-based fingerprints associated with exposed municipal solid waste. These results demonstrate that landfill surfaces exhibit intrinsic SWIR spectral characteristics that persist across diverse African environments. This study establishes the first multi-site hyperspectral library of African landfill surfaces, providing a physical basis for developing generalised landfill detection frameworks.

Heavy precipitation in Northeast China is frequently modulated by the Northeast China Cold Vortex (NCCV), although the microphysical processes within squall lines under such conditions remain insufficiently understood. This study presents a comprehensive analysis of an NCCV-influenced squall line in Liaoning Province, utilizing coordinated S-band polarimetric radar and surface disdrometer observations. The raindrop size distribution (DSD) characteristics and three-dimensional microphysical structure are systematically examined for both convective and stratiform regimes. A comparative analysis of DSD and warm-rain microphysical mechanisms is also conducted with a Mei-yu event. Results show that convective rain in the NCCV squall line exhibits a continental-type DSD, characterized by fewer but larger raindrops compared to other heavy rainfalls in China. In contrast, the Mei-yu frontal convection under NCCV influence exhibits a transitional DSD pattern between the maritime and continental types, with raindrops smaller and denser than those in the NCCV squall line. Vertical structure of the mature squall line shows prominent differential reflectivity (ZDR) and specific differential phase (KDP) columns above the melting level within the convective region, indicating vigorous riming growth of graupel and hail driven by strong updrafts. Meanwhile, the stratiform region is characterized by ice crystals and aggregates, formed primarily through deposition and aggregation processes. The subsequent melting of ice-phase particles followed by collision–coalescence and evaporation-driven size sorting shapes the large but sparse raindrops in the NCCV squall line. Comparison with Mei-yu convection demonstrates that surface DSD is shaped by environmental conditions and vertical microphysics. The drier, more unstable environment in the NCCV squall line favors deep convection with active ice-phase processes, while the relatively moist and stable environment of the Mei-yu convection supports shallower convection dominated by warm-rain processes. Future multi-case studies with integrated observations are needed to quantify how environmental and aerosol conditions modulate these heavy precipitation processes.

Few-Shot Class-Incremental Learning (FSCIL) aims to sequentially learn new classes from very few labelled samples while preventing the forgetting of previously acquired knowledge, which has important practical value for remote sensing scene classification (RSSC). Recent studies have shown that applying a Vision Transformer (ViT) pre-trained on natural image datasets to FSCIL tasks can achieve significantly superior performance. Nevertheless, a substantial domain distribution gap exists between natural images and remote sensing images, which leads to severe performance degradation when such models are directly transferred to RSSC. To address the domain gap alongside FSCIL’s inherent stability–plasticity dilemma and overfitting under data scarcity, we propose a Dynamic Expansion Mixture-of-Experts with Pre-trained Vision Transformer (DEM-ViT) framework. Specifically, to alleviate the domain discrepancy, DEM-ViT incorporates an Adapter-Based Mixture-of-Experts (ABMoE) module, which captures the diverse visual patterns of remote sensing scenes through feature reconstruction in the representation space and collaborative learning among multiple experts. Furthermore, to address the stability–plasticity dilemma in FSCIL, we propose a Dynamic Expert Expansion (DEE) strategy, which progressively expands the model capacity along the incremental sessions. DEE provides sufficient space for learning new knowledge while mitigating the forgetting of previous knowledge. In addition, we propose a Semantic-Guided Feature Alignment (SGFA) method to reduce the risk of overfitting under data-scarce conditions. SGFA leverages textual information to construct robust text prototypes and uses them to calibrate the visual feature space. Extensive experiments across three benchmarks indicate that our framework exhibits highly competitive performance compared with state-of-the-art methods.

In Transformer-based image restoration models, the self-attention mechanism often introduces attention noise from irrelevant contextual feature, hindering the recovery of underlying clear content. Although many methods have been proposed to suppress attention noise, we note that most existing approaches are often developed for general vision tasks and fail to generalize across remote sensing image dehazing, where large-scale spatial structures pose additional challenges for attention modeling. How to effectively model scale-aware attention to suppress redundant activations becomes crucial for remote sensing image dehazing. In this paper, we propose a scale-adaptive differential Transformer (SDTformer), an architecture designed to suppress attention noise through a differential attention mechanism, thereby improving reconstruction fidelity. Specifically, the model incorporates a scale-adaptive differential self-attention module, which models contextual dependencies across different spatial scales and reduces redundant contextual interference by computing differential attention maps. Additionally, a dynamic differential feed-forward network is proposed to adaptively select informative spatial features, strengthening feature aggregation. To further enhance feature representation, a gated fusion module is introduced to aggregate multi-scale features generated by different encoder blocks, which facilitates the learning process of each decoder block and improves the final reconstruction performance. Extensive experimental results on the commonly used benchmarks show that our method achieves favorable performance against state-of-the-art approaches.