The Loess Plateau is recognized as a typical climate-sensitive and ecologically vulnerable region in China. Understanding the spatiotemporal characteristics and potential driving factors of vegetation dynamics in different dry/wet climate zones within the Loess Plateau holds critical significance for the conservation and management of regional ecosystems. Based on the kernel normalized difference vegetation indices (kNDVIs) of the Loess Plateau from 2000 to 2022, this study investigated the spatiotemporal patterns of vegetation dynamics in different dry/wet climate zones within the Loess Plateau using the coefficient of variation and trend analysis. Employing the optimal parameter-based geographical detector model, this study accurately and scientifically identified the driving factors and ranges of vegetation dynamics under the spatial scale and zoning effect, effectively addressing the challenge of spatial heterogeneity. The results indicate that the average kNDVI of the Loess Plateau presented a spatial distribution pattern characterized by low values in the northwest and high values in the southeast. In terms of vegetation dynamics, 91.57% of the Loess Plateau showed an upward trend, with the semi-arid climate zone accounting for the highest proportion (60.41%). Different driving factors in the Loess Plateau corresponded to varying optimal dispersion methods and optimal interval breakpoints. Under the optimal zoning effect, low temperature and high rainfall were identified as the primary conditions for vegetation growth. The different ranges and types of driving factors exerted different effects on the spatial distribution of vegetation dynamics. The optimal parameter-based geographical detector model demonstrates that rainfall and land use type constituted the principal driving factors of the Loess Plateau, accounting for 65.45% of the total explanatory power. The q value (0.69) of the interaction between the two driving factors was higher than the q values of interactions between other factors. This study provides a comprehensive insight into the response mechanisms of vegetation dynamics under natural and human factors, thereby guiding the sustainable development of regional ecosystems.

Hyperspectral remote sensing （HRS） technology，with its high spectral resolution and extensive spectral coverage，demonstrates significant potential in geological prospecting. Focusing on the Qianhongquan gold deposit in the Beishan orogenic belt，Gansu Province，this study conducted altered mineral mapping and component analysis，using HRS data from the AHSI sensor on the ZY-1 02D satellite and the self-developed hyperspectral mineral mapping technique，GeoAHSI，revealing their spatial distribution characteristics. Besides，ground-based spectral measurements were conducted on typical profiles to validate the spectral data，thereby assessing the reliability of the hyperspectral mineral mapping results. The results indicate that the primary altered minerals in the Qianhongquan gold deposit and its surrounding rocks include sericites （low-aluminum，medium-aluminum，high-aluminum，and iron-rich muscovites），calcites，dolomites，epidotes，and chlorites. Their distribution is closely related to ductile shear zones，with the distribution of sericites，chlorites，and epidotes being particularly significant within these zones. This spatial correlation provides critical indicators for regional prospecting. Additionally，it was observed that the 2 200 nm absorption feature of sericites and the 2 250 nm absorption feature of chlorites exhibit marked enrichment in silicon （Si） and iron （Fe） around ore bodies，which is closely correlated to the chemical compositions of the minerals. By enhancing the identification of weak spectral features，this study successfully applied HRS technology to mineral identification and spatial distribution analysis. These findings provide a scientific basis for further exploration of the Qianhongquan gold deposit and offer valuable references and guidance for the application of HRS in similar deposits.

Given its all-day availability, all-weather adaptability, and high spatial resolution, the interferometric synthetic aperture radar (InSAR) technique has been widely applied in multiple fields, demonstrating strong adaptability and high practical value. Focusing on two typical landslide areas in Zhouqu County, Gansu Province, this study compared small baseline subset InSAR (SBAS-InSAR) monitoring results and Kalman filter prediction results with monitoring data from the global navigation satellite system (GNSS), confirming the reliability and accuracy of the SBAS-InSAR technique in monitoring landslide deformations. The results indicate that the SBAS-InSAR technique exhibited significant advantages in monitoring areas with deformations induced by geologic disasters, effectively overcoming the limitations of traditional monitoring means. This technique can provide critical technical support and scientific basis for early warning and management of geologic disasters in Zhouqu County and other areas prone to suffer these disasters.

Compared to natural surfaces, urban surfaces have more complex geometric structures, leading to significant impacts of the multiple scattering effect within pixels and the neighborhood effect on the inversion results of urban land surface temperature (LST). This study proposed a novel urban LST inversion algorithm that integrates machine learning and an enhanced temperature and emissivity separation (TES) method. Finally, the proposed algorithm was applied to China’s SDGSAT-1 thermal infrared data. The algorithm comprises three key steps: First, the inversion of urban canopy brightness temperature from SDGSAT-1 data was conducted using the eXtreme Gradient Boosting (XGBoost) algorithm. Second, an enhanced TES algorithm based on the sky view factor (SVF) was developed to account for urban geometry, enabling high-precision urban LST inversion. Third, the accuracy of the inversion algorithm was assessed and applied to the urban area of Beijing. The results demonstrate that inversion using an XGBoost algorithm and a split-window algorithm yielded root mean squared errors (RMSEs) of approximately 0.2 K and 1.2 K, respectively. The LST RMSEs with and without available water vapor data were determined at 0.36 K and 0.73 K, respectively; and the LSE RMSEs under three bands were 0.020/0.026, 0.018/0.023, and 0.020/0.023, respectively. The differences in the LST inversion results derived using the original and improved TES algorithm ranged from 0 to 1.86 K.

A wildland-urban interface (WUI) refers to the area where residential buildings meet or intermingle with natural vegetation such as forests. The delineation of the WUI plays an important role in fire risk management, forest resource development and utilization, climate change responses, and sustainable socio-economic development. Current methods for WUI delineation are primarily developed and refined based on the definition given in the Federal Register of the United States. Based on indicators such as building density, vegetation coverage, and the distance between buildings and vegetation, these methods can be categorized into three types: building density priority, fuel grade priority, and overlap between building-vegetation buffer zones. Initially, this study presented a summary and comparison of relevant literature on the three types of methods. Then, Wood Buffalo in Alberta, Canada, an area frequently affected by wildfires, was selected to compare the three methods using data on Canadian building footprints released by Microsoft, global land cover from GLC_FCS30-2020, and local historical fire points and fire scars. The results indicate that the building density priority method exhibited the highest coincidence rate with historical wildfire records. However, it overlooked low-density buildings that were also at risk of wildfire. The fuel grade priority method produced a larger delineation area, with a lower coincidence rate with historical wildfire records since it focused excessively on the vegetation around buildings while neglecting the buildings themselves. In contrast, overlap between building-vegetation buffer zones presented the lowest coincidence rate with historical wildfire records and the smallest delineation area. This occurred primarily due to the short distance setting of buffer zones. This study reveals the strengths and limitations of existing methods, contributing to more scientifically robust and rational WUI delineation in the future while also providing references for decision-making in fire risk management and emergency responses.

The analysis of patches showing changes in coastal areas of the Chinese mainland tends to encounter challenges such as low image resolution, long time intervals, and limited spatial coverage. This study aims to obtain high-frequency, accurate information on changes in coastal areas nationwide. This will facilitate the dynamic monitoring of marine resources and the implementation of relevant protection policies for coastal areas in China. To this end, using domestic high-resolution remote sensing data of 15 days (i.e., one cycle), as well as the iteratively reweighted multivariate alteration detection (IR-MAD) algorithm combined with visual interpretation, this study extracted patches reflecting 2020—2023 changes along the coasts of 11 provinces and cities in the Chinese mainland. Accordingly, this study analyzed their spatiotemporal characteristics, landscape patterns, and spatial correlation. The results indicate distinct directional changes in the patches. The patches reflecting changes from sea enclosure to reclamation exhibited the largest areas across various investigated areas. Except for Hainan Province, the area of this type of patches exceeded 1 000 km². The proportions of patches reflecting different types of changes gradually tended to be balanced. In the winter of 2022, the proportion of patches showing changes in the reclamation dropped below 50% for the first time. The aggregation degree of patches reflecting various types of changes showed increasing trends, suggesting that patches reflecting various changes will become more concentrated in the future. The centroids of these patches of various regions shifted in varying directions, and these patches exhibited significant spatial correlation within a 20 km range.

The current high-resolution remote sensing images involve complex scenes that are difficult to analyze. Meanwhile，owing to the diverse scenes，there is a lack of accurate reference obtained from the sample database. Therefore，this paper proposed a self-learning segmentation method for high-resolution remote sensing images，with reference to the visual dual-drive cognition mechanism. Based on the principle of visual perception，this method interpreted the typical ground objects in the scene through unsupervised adaptive analysis. In addition，it achieved self-learning identification of typical ground objects by integrating a neural network. Finally，the segmentation results were self-checked and corrected by combining unsupervised analysis and neural network learning. Using real high-resolution remote sensing image data containing complex ground scenes，the comparative experiments were conducted between the proposed method and two popular deep neural network segmentation methods：mask region-based convolutional neural network （Mask R-CNN） and scalable vision transformer （ScalableViT）. The results showed that the proposed method can maintain robust and reliable segmentation accuracy，and outperformed others in terms of ground object cognition，generalization performance，and anti-interference ability. As such，it proved to be a cost-effective and practical approach.

Using level-1 (L1) brightness temperature data from the Microwave Radiation Imager (MWRI) on board Fengyun-3D (FY-3D) satellite and the corresponding Level-2 (L2) precipitation products, this study established a precipitation rate inversion model for land surface heavy precipitation in Hunan Province based the polarization corrected temperature (PCT) and scatter index (SI). The proposed model was validated using individual examples. The results indicate that the precipitation rates retrieved from the L1 brightness temperature data of the FY-3D satellite were generally consistent with the results obtained from the L2 precipitation products. Compared to actual data, the retrieved precipitation rates were slightly higher in low precipitation areas but smaller in centers of high precipitation areas. The ascending orbit-based inversion model exhibited a correlation coefficient, mean absolute error (MAE), and root mean square error (RMSE) of 0.876 1, 0.771 1, and 1.151 4 mm/h, respectively. Conversely, the descending orbit-based inversion model presented a correlation coefficient, MAE, and RMSE of 0.911 3, 1.130 4, and 1.832 2 mm/h, respectively. The inversion results showed a larger precipitation distribution range than that of L2 products. Compared to the measurements at ground meteorological stations, the inversion model demonstrated higher accuracy than L2 products. This study successfully determined the distribution of land surface heavy precipitation in Hunan through inversion. The results of this study can provide a reference for investigating the relationship between microwave brightness temperature and precipitation and estimating land surface heavy precipitation.

Interferometric Synthetic Aperture Radar (InSAR) faces the challenges of the insufficient number of monitoring points and low monitoring accuracy when applied to complex environments with dense vegetation and large-gradient surface deformation in a mining area. To address these challenges, this study proposed an improved distributed scatterer InSAR (DS-InSAR) method assisted by stacking technology. This method identified statistically homogenous pixels using a confidence interval hypothesis test and achieved phase optimization utilizing a phase triangulation algorithm. Subsequently, the residual phases were derived by removing the linear deformation phases determined via stacking-based simulation. This step contributed to sparse deformation phase fringes, thereby enhancing the accuracy of spatiotemporal filtering and three-dimensional phase unwrapping within the subsequent DS-InSAR processing framework. Finally, the simulated phases were compensated, and thus complete deformation fields were determined. By processing Sentinel-1A SAR images covering the Xinjulong Coal Mine from October 2015 to March 2016, this study interpreted the time-series surface deformation characteristics in the mining area during this period. The findings revealed three significant deformation sites in the mining area, with a maximum cumulative radar line-of-sight (LOS) deformation of up to -313 mm. Compared to conventional small Baseline Subset (SBAS) InSAR, the proposed method yielded more uniformly distributed monitoring points via inversion, with a density approximately 12.9 times higher. The root mean squared error (RMSE) of the inversion was approximately 6.82 mm relative to leveling data, representing an accuracy improvement of about 3.0 mm compared to the SBAS results.

Scientifically and rationally demarcating urban and town areas is a fundamental task during China’s rapid urbanization stage. It serves as a critical basis for promoting the optimization and quality improvement of urban and rural spaces, scientifically coordinating urban and rural planning and construction management, and implementing territorial spatial planning. However, there is neither a unified concept nor a universal delimitation method for urban and town areas in China, hindering their planning, construction, development, and public management. Based on defining the relevant concepts of urban and town areas, this study proposed a people-centered method for determining town areas with no cities and counties set using geographic information system (GIS) technology, considering the characteristics and spatial relationships of land types. The data sources of this study include the results of the third national land resource survey, statistical bulletins, remote sensing image interpretation, and point-of-interest (POI) data. Finally, the proposed method was applied to demarcate the town areas in Xiapu County, Ningde City. The empirical study results demonstrate the effectiveness of the proposed method, which features a scientific and concise technology roadmap and strong operability. Therefore, the proposed method can provide a theoretical foundation for the rational territorial spatial planning.

Traditional methods for information extraction of small water bodies suffer from poor performance and low accuracy, failing to meet actual needs. Using the high-resolution images of the Erhai Lake basin from the Jilin-1 domestic satellite as the data source, this study proposed a deep learning-based semantic segmentation method using an improved DeepLabV3plus model. Replacing the ResNet-101 encoder with EfficientNet-B4, this study innovatively combined the BCE Loss and Dice Loss functions, identifying the optimal method for fine-scale information extraction of water bodies in the Erhai Lake Basin. The results indicate that compared to traditional methods, the improved DeepLabV3plus model performed better in the information extraction of water boundaries, enabling accurate identification of main water bodies, especially small streams. The improved DeepLabV3plus model exhibited higher precision (98.87%), recall (99.30%), and F1-Score (99.08%) than the normalized difference water index (NDWI) and object-oriented methods. Regarding comparison of details, the improved DeepLabV3plus model can effectively suppress the influence of building shadows, vegetation occlusion, and complex surface features, improving the information extraction effects of small water bodies and complex edge areas. In addition, ablation experiments show that the introduction of the combined loss functions and compound scaling strategy increased mIoU by 0.62% and 3.07%, respectively, significantly enhancing the model's segmentation accuracy and ability to extract multi-scale semantic information.

The East Qinling Mountains, located in the eastern Qinling orogen between the North China and Yangtze plates, boast the largest Mo-Au-W polymetallic metallogenic belt in China. Given that alteration played a key role in the mineralization process, its information extraction and distribution characteristics can provide critical insights for analyzing the mineralization mechanisms. To explore a more efficient method for extracting alteration information, this study investigated Dengfeng City in the East Qinling Mountains using data from the Sentinel-2A and Landsat-8 sensors. Data processing and analysis were conducted based on the Google Earth Engine (GEE) platform, and deep learning was applied to the extraction of alteration information. To improve the extraction efficiency, the information about vegetation, water bodies, and buildings was extracted first using the normalized difference vegetation index (NDVI), modified normalized difference water index (MNDWI), and normalized difference built-up index (NDBI), respectively. Subsequently, the interference information was masked by generating binary images using the threshold segmentation method. In combination with the spectral curves of typical hydroxyl minerals, the bands used to extract hydroxyl alteration information were determined. Then, the initial alteration information was extracted using the principal component analysis (PCA) method, and the pixels that overlapped spatially and exhibited concentrated information and high alteration levels were selected as labels to train the deep learning model. The potential information of remote sensing images was further extracted using the convolutional neural network (CNN) model that integrated multi-band data. Finally, in combination with the linear structure maps and mineralization anomalies of the target area, rock and soil samples were collected from the corresponding locations, and their main components were determined using X-ray fluorescence spectroscopy (XRF) and X-ray diffraction (XRD) analysis. In this manner, the reliability of the alteration information extracted was verified. The results indicate that compared to the PCA method alone, the CNN model can extract more comprehensive and clearer hydroxyl alteration information that was more easily graded. The samples collected at the field sampling points all contained minerals with hydroxyl alteration, such as muscovite, biotite, and chlorite. The laboratory XRF and XRD analysis results were consistent with the hydroxyl alteration information extracted using the CNN model. This verifies the reliability and efficiency of the interpretations of hydroxyl alteration information extracted using the deep learning-based CNN model. The results of this study can provide a theoretical and technical basis for remote sensing prospecting in the East Qinling Mountains.

Building extraction aims to separate building pixels from remote sensing images, which plays a crucial role in applications such as urban planning and urban dynamic monitoring. However, building extraction generally faces challenges, such as void, false positives, and false negatives. Given this, this paper proposed a densely nested network (DN-Net). The sub-networks in the DN-Net were integrated with the enhanced residual convolutional module (ERCM) to extract rough contours of buildings from remote sensing images. Furthermore, to accurately locate the buildings, a coordinate attention module (CAM) was incorporated, effectively avoiding false positives. To deal with the holes during building extraction, a cascade convolutional module (CCM) was used, allowing the extraction of richer details with convolution kernels of various sizes, thereby ensuring accurate building extraction. The DN-Net was tested with the WHU datasets to assess its accuracy. The results showed that the DN-Net exhibited an intersection over union (IoU) of 89.20% and a F1 score of 94.29% on the validation set and 89.85% and 94.65%, respectively, on the test set. The results confirm that the DN-Net can significantly improve the building extraction accuracy, with more complete and detailed boundaries of buildings being extracted, demonstrating an outstanding ability to extract buildings of varying sizes.

Land use serves as the primary cause of global environmental changes. Therefore, investigating its spatiotemporal changes and corresponding driving factors is significant for promoting the sustainable development of regional socioeconomics and ecosystems. Based on nine stages of remote sensing monitoring data on land use/land cover from 1980 to 2020, this study analyzed the spatiotemporal changes in land use types in the Golmud River basin. By combining the analysis of significant correlations, this study explored the major factors driving changes in land use within the basin. The results indicate that over the past 40 years, unused land and grassland proved to be dominant land use types in the Golmud River basin. The areas of cultivated lands, water bodies, and construction lands exhibited an increasing trend, while those of forest lands, grasslands, and unused lands trended downward. The period from 2015 to 2020 witnessed significant changes in both the areas and the dynamic degrees of various land use types within the basin. During this period, spatial changes in land use transition predominately occurred in the central and northern parts of the basin. Between 1980 and 2020, the unused land showed significant fragmentation. Human activities, particularly total population and regional gross domestic product, were identified as the main factors driving changes in the land use type within the basin.

Nonnegative matrix factorization (NMF) is commonly used in hyperspectral image (HSI) unmixing due to its high interpretability and computability. To effectively address HSI noise and improve unmixing efficiency, this study proposed a method for hyperspectral unmixing based on spectral-spatial total variation nonnegative matrix factorization (SSTVNMF) with feature space augmentation. First, the original data space was converted to the feature space through feature extraction, allowing the unmixing process to be performed in the feature space for enhanced unmixing efficiency. Second, to reduce the impact of noise, the spatial information was extracted using the bilateral filtering (BF) method for enhanced feature extraction, thereby ensuring the accuracy of extracted features. Third, to ensure the effectiveness of the unmixing method, total variation (TV) regularization that considers both spatial and spectral features was established based on the NMF method. The spatial TV promotes abundance smoothing by calculating the horizontal and vertical differences in abundance between neighboring pixels. Based on the minimum-volume TV, the spectral TV enhances endmember extraction by applying constraint forces between endmembers to minimize the volume. Finally, the proposed method was verified using the synthetic data from the USGS spectral library as simulated data and the Jasper Ridge, APEX, and Cuprite datasets as actual data. The experimental results demonstrate that the proposed method outperformed other improved NMF-based methods in terms of qualitative and quantitative assessments.

Landslide disasters are frequent and widespread in southwestern China. The accurate identification and mapping of landslides using remote sensing imagery are of great significance for disaster prevention and mitigation. However，in complex environments，traditional remote sensing detection methods are often prone to misidentification due to background noise in the imagery. This paper proposed a dual-fusion landslide detection network （DLDNet） to improve landslide detection accuracy under challenging conditions. First，based on existing landslide samples，landslide simulation was conducted in complex environments using data augmentation techniques. Second，the ConvNeXt was adopted as the feature extraction backbone of DLDNet to capture more complex landslide features. Then，an attention module enhanced with deformable convolution was introduced to better focus on landslide-related information. Finally，a dual-fusion feature pyramid network （DFPN） was designed to thoroughly integrate feature information across different scales and receptive fields. The experimental results show that the proposed DLDNet achieved average precision （AP） scores of 56.9% for bounding box detection and 52.5% for segmentation，10.4 and 10.7 percentage points higher than those of the baseline model （Mask R-CNN）. Compared with other landslide detection models，the DLDNet demonstrates higher detection accuracy and a lower false alarm rate. The method，characterized by accurate landslide detection in complex environments，can support rapid landslide identification and emergency response.

The deep learning-based models currently available for detecting changes in remote sensing images face several challenges, including limited multi-granularity, poor classification performance of networks, high sensitivity to parameters, and great efforts in parameter adjustment. To address these challenges, this study proposed a deep forest-based model for detecting changes in remote sensing images. Initially, preliminary results were determined using a basic change detection method. Then, the results were optimized using the multi-granularity scanning characteristics and strong data classification of deep forest sub-networks. In this manner, the final change detection results were obtained. Verification experiments conducted on the LEVIR-CD and SYSU-CD datasets using various common change detection models indicated that the proposed deep forest-based model significantly outperformed other models in terms of precision, F1 score, and recall. Additionally, the proposed model exhibited strong adaptability on small datasets, as verified by loss function comparison, small-sample experiments, and ablation studies. This adaptability can reduce the complexity of parameter adjustment and address the issues that other deep learning sub-networks fail to be applicable to medium and small datasets.

At its 16th Plenary Session and Ministerial Summit, the Group on Earth Observations (GEO) proposed a new goal to build a “digital library for Earth observation applications”, highlighting the transition from “open data” to “open science”. It aims to achieve the management and sharing of knowledge resources, including data, algorithms, literature, and cases, thereby facilitating the comprehensive application and knowledge service provision of Earth observations in fields such as global change. Under this research background, this study systematically examined Earth observation data resources, including the conceptual system of Earth science variables, Earth observation satellites and payloads, observational and simulated data products, and open knowledge bases of academic literature. Based on the theories and techniques related to the Semantic Web and Knowledge Graph, this study established the Earth observation knowledge ontology with corresponding instances, involving Earth science variables, remote sensing satellites, observation payloads, observational and simulated datasets, journals, and academic literature. The knowledge representation results of this study will contribute to the representation, management, and integration of data and knowledge in the field of Earth observation applications. Moreover, they facilitate the discovery of potential associations between data and knowledge, enhancing the efficiency of scientific research and advancing scientific discovery.

The simple and efficient water index method has been widely used to monitor and identify surface water along with its spatiotemporal variations. However, with the application of narrow-band multispectral sensors, this method faces a challenge in selecting optimal bands with similar features when the data source changes during large-scale water monitoring. Guided by the normalized difference water index (NDWI) and the modified NDWI (MNDWI), and based on the Google Earth Engine (GEE) platform, this study constructed water indices using the green bands and eight red bands from the Sentienl-2A/B multispectral sensor data. Employing Otsu's method, this study identified and extracted water bodies in six quadrats measuring 90 km × 90 km across different temporal and spatial ranges in China. The results indicate that the optimal band combination for water body extraction varied across different times and locations. Compared to the eight water indices constructed from the Sentienl-2A/B multispectral sensor data, the water index based on the combination of B3 and B11 bands, combined with Otsu's method, achieved optimal water identification and extraction results. These results were observed in summer in the lake regions of the Northeast China Plain and mountains, the eastern plains, the Inner Mongolian Plateau, the Yunnan-Guizhou Plateau, Xinjiang, and the Qinghai-Tibet Plateau. In both spring and summer, the water index based on the combination of B3 and B11 bands exhibited an overall accuracy (OA) exceeding 90% and a Kappa coefficient above 0.9, indicating its applicability across different time periods. Overall, the results of this study provide a reference for the design and development of sensors targeting water extraction and monitoring, and for feature band selection in water monitoring and extraction applications based on narrow-band remote sensing data.

Pansharpening of remote sensing images refers to the fusion of panchromatic images (PAN) and low-spatial-resolution hyperspectral (or multispectral) images (LR-HSI/LRMS) to produce high-spatial-resolution hyperspectral (or multispectral) images (HR-HSI/HRMS). Currently, deep learning-based pansharpening methods have increasingly matured. However, pansharpening still faces several challenges, including inadequate feature extraction, insufficient guidance for information fusion, and oversimplified single-stage architectures, resulting in HR-HSI imagery with compromised spatial and spectral fidelity. To address these issues, this paper proposed a two-stage pansharpening method for hyperspectral images based on feature enhancement and a Three-Stream Transformer architecture. In the first stage, preliminarily enhanced hyperspectral images (HSI) were generated using a feature enhancement module and a multi-scale fusion module. Specifically, the feature enhancement module strengthened spatial and spectral information across multiple scales, while the multi-scale fusion module integrated the enhanced HSI at different scales. In the second stage, the initially enhanced HSI, PAN, and images resulting from their fusion were treated as three separate feature streams using the self-attention mechanism of the Transformer. Then, these streams were transformed into the Q(Query), K(Key), and V(Value) matrices via linear layers, followed by multi-head attention computation, which effectively guides the extraction and fusion of spatial and spectral information. Furthermore, the enhanced HSI and an additional fusion module were leveraged to refine image quality, yielding HR-HSI results with richer spatial and spectral details. Validation experiments were conducted on three classic hyperspectral datasets. The results demonstrate that the proposed method outperforms both conventional and existing deep learning-based approaches in terms of quantitative evaluation metrics. Considering qualitative evaluation results, it also preserves spectral information of the HSI and spatial details of the PAN images, producing more realistic HR-HSI images.

Plastic greenhouses have gained extensive application in modern agriculture. This, however, gives rise to ecological issues. Remote sensing data enable effective extraction and identification of plastic greenhouses on a large scale. Existing studies largely focus on plastic greenhouse extraction using either classification or spectral indices methods. However, there exists a lack of the combination and comparative analysis of both methods. This study proposed a method for plastic greenhouse extraction that integrates multiple Sentinel-2 spectral indices and a one-class classification method (improved one-class random forest). Furthermore, this study extracted information on plastic greenhouses using an improved one-class random forest method, as well as six spectral indices of plastic greenhouses as classification features. The extraction results were then compared with those of the proposed method to demonstrate the effectiveness of the latter. The results indicate that the proposed method yielded an overall accuracy of above 97% across four seasons, with kappa coefficients exceeding 0.82 and F1 scores of over 0.84. These metrics all were better than those yielded using the six spectral indices. Furthermore, the proposed method exhibited differences in the overall accuracy, kappa coefficient, and F1 score across four seasons of less than 1%, under 0.1, and below 0.1 respectively. This suggests the high seasonal stability of the method, outperforming the extraction results obtained by using spectral indices alone. This study provides a method for accurately monitoring the spatial distribution of plastic greenhouses.

The positive effects generated by large-scale underground mining in China's western semi-arid region have attracted increasing attention in recent years. Accurately understanding and scientifically utilizing these positive effects in mines plays a significant role in saving ecological restoration costs for mines. To reveal the perturbation characteristics of subsidence basins on precipitable water vapor (PWV), this study investigated the Daliuta mine in the Shendong mining area. Based on the simulation results of the atmospheric radiative transfer model and the ground-based GPS real-time observation data, this study constructed a water vapor inversion model using Sentinel-2 satellite images, obtaining the near-surface PWV content from 2017 to 2021 in the Daliuta mine. Furthermore, this study analyzed the near-surface PWV distributions in the single-mining-face subsidence basin and the mining-face-group subsidence area. By deploying HOBO temperature and humidity sensors on site, this study comparatively analyzed the near-surface relative humidity inside and outside the subsidence basin. The results indicate that subsidence basins showed positive convergence effects on PWV. Specifically, the near-surface PWV in the single-mining-face subsidence basin decreased gradually from the inside to the outside of the basin. The near-surface PWV in the mining-face-group subsidence area was significantly improved after mining. The relative humidity was significantly higher inside the subsidence basin compared to the outside. The differences in relative humidity in the vertical direction from the surface were 14.52, 13.53, 12.43, 10.60, and 10.33 percentege point, respectively, indicating gradually weakening water vapor convergence effects in the subsidence basin with an increase in elevation. The water vapor convergence effects were significant at nighttime but nonsignificant at daytime. Finally, based on vegetation surveys and previous studies, this study proposed a conceptual model for water vapor convergence effects in subsidence basins to explain the mechanism governing water vapor convergence. Additionally, subsidence basins somewhat contribute to the benign cycle of ecosystems in semi-arid mining areas.

Investigating land-use-related carbon emissions (LCE) plays a vital role in achieving goals of peak carbon dioxide emissions and carbon neutrality (also referred to as the “dual carbon” goals). Research on the changes and prediction of LCE in Xiangxi Tujia and Miao Autonomous Prefecture (also referred to as the Xiangxi Prefecture) can provide a theoretical reference for the region to develop policies on the achievement of the “dual carbon” goals and for local balanced development and protection. Based on five sets of land use data from 2000 to 2020, this study analyzed the land use conditions and the spatiotemporal evolution of historical carbon emissions in Xiangxi Prefecture. The factors influencing LCE were determined using a decoupling model and a logarithmic mean Divisia index (LMDI) model. Furthermore, three land use scenarios were established: natural development, priority of cultivated land protection, and ecological protection priority. Using these scenarios, this study predicted the land use and carbon emissions in Xiangxi Prefecture in 2030. The results indicate that forest land represents the dominant land use type in Xiangxi Prefecture. Regarding land use transition, the period from 2000 to 2020 witnessed a significant increase in construction land, which encroached into substantial areas of forest land and cultivated land. Concurrently, water bodies and grassland decreased in area, being converted into forest land and cultivated land. From the perspective of carbon emissions, land use in the region exhibited a transformation from carbon sinks to carbon sources in general. During the 20-year span, the total LCE continued to increase. Construction land was identified as the primary land type as a carbon source, while forest land was the main land type as a carbon sink. Within the 20 years, carbon emissions decreased only in Huayuan County but increased in all other counties and cities. After 2010, the original regions with elevated carbon emissions showed a decrease in carbon emissions, while other regions witnessed growing carbon emissions to varying degrees. These regional changes in carbon emissions were largely attributed to the increased carbon emissions from construction land. Xiangxi Prefecture maintained a weak decoupling effect generally, with counties and cities fluctuating between weak decoupling and strong decoupling states. The economic output effect and energy efficiency effect served as the primary factors influencing carbon emissions. The overall land pattern remained relatively stable across the three scenarios. The carbon emissions of the three scenarios increased in the order of ecological protection priority, natural development, and priority of cultivated land protection. In the future, construction land will still represent the dominant factor causing overall changes in carbon emissions, while forest land will remain as the primary carbon sink.

This study aims at investigating variations in soil moisture and vegetation net primary productivity （NPP） in the Qingling River Irrigation Area，Yunnan （elevation 1 515~1 876 m），a typical subtropical alpine climate region. To this end，initially，this study recognized land surface temperature （LST） and normalized difference vegetation index （NDVI） as explanatory variables，leveraging remote sensing technology for rapid and long-term sequential monitoring. Subsequently，the SMAP L4 soil moisture product was downscaled to a 30 m spatial resolution using the random forest adaptive window regression algorithm. Then，the water stress parameter of the CASA model was modified using the land surface water index （LSWI），which integrated multi-source remote sensing data，such as surface reflectance，to estimate NPP. Following spatial resampling，a 30 m resolution NPP spatial distribution was achieved. Finally，multiple land cover scenarios，including forest land，paddy fields，and irrigated farmland，were established. The Pearson correlation coefficient was introduced for the quantitative evaluation of the spatial relationship between soil moisture and NPP in the study area. In terms of the spatial distribution of soil moisture，the study area exhibited higher values in the north and lower values in the south during summer，while lower values in the northwest and higher values in the southeast and south during winter. Compared to field measurements，the inverted NPP results showed a R²>0.7 and a RMSE<0.3. Both summer，winter，and annual average NPP values at the pixel level showed an increasing trend over time. Spatially，scenarios such as paddy fields and forested land presented correlation coefficients exceeding 0.5. Among these，forest land was least sensitive to water stress，while paddy fields and irrigated farmland were most affected. This study establishes a monitoring and feedback mechanism for the soil moisture-NPP balance from seasonal and spatial perspectives in the study area.

Coastlines serve as one of the most essential basic geographic elements. However，conventional methods generally face challenges in the accurate detection of their location，due to instantaneous remote sensing imaging and dynamic tidal phenomena. In response to this，this study developed a novel coastline extraction model that incorporates information on surface moisture content derived from long-time-series satellite remote sensing imagery. First，all available remote sensing images covering the study area during the target period were acquired to construct a high-quality remote sensing image stack. Second，the wetness components indicative of the surface moisture content were obtained using the tasseled cap transformation （TCT），from which a wetness index stack was constructed. Then，the wetness components were subjected to maximum value synthesis using the maximum spectral index composite （MSIC） algorithm，generating a maximum water surface composite image. Finally，the composite image was segmented using the OTSU algorithm to extract accurate coastline information. Validation experiments were conducted on Zhoushan Island using the Google Earth Engine （GEE） cloud computing platform and remote sensing imagery from the operational land imager （OLI） onboard the Landsat 8 satellite. The results indicate that the proposed model can precisely locate different types of coastlines with high spatial accuracy. Compared to visual interpretation，the model exhibited a mean deviation and a root mean square error （RMSE） of 3.42 m and 6.79 m，respectively，with 99.42% of validation points falling within one pixel width. This study provides an effective technical framework for high-accuracy coastline extraction，holding great significance for scientific management and sustainable development of coastal resources.

In optical remote sensing images with complex scenes and rich land cover information，the sea-land segmentation faces challenges such as low positioning accuracy and blurred edges. Therefore，this paper proposed a deep convolutional network model and a sea-land segmentation method that integrate contextual semantic information and edge features. First，the rich target semantic information was extracted from remote sensing images using the FusionNet semantic segmentation network module. Then，multi-scale and hierarchical contextual semantic features were extracted from the segmentation network using the enhanced atrous spatial pyramid pooling （ASPP） module and contextual attention module. Additionally，an edge extraction sub-network was built to extract multi-scale edge features. Finally，the semantic features and edge features were combined through a fusion module，thereby achieving accurate sea-land segmentation. This method was tested with two typical representative datasets. The results showed that this method achieved an overall prediction accuracy of 98.21%，an F1 score of 97.64%，and a boundary F1 score of 89.36%，all significantly outperforming other models. Particularly in complex backgrounds，this method can effectively improve the accuracy of segmentation and edge detection，demonstrating definite advantages in the segmentation of artificial coastlines and ports.

Through over 70 years of development, the World Meteorological Organization (WMO) has established a global data-sharing network that covers 193 members. This study analyzed the WMO’s meteorological data-sharing system from two aspects: system architecture and composition, and management norms and standards. The meteorological data-sharing system comprises the global observing system (GOS), the global telecommunication system (GTS), the WMO information system (WIS), and the global data-processing and forecasting system (GDPFS). Specifically, the GOS coordinates and schedules observational facilities from land and marine stations, aircraft, environmental satellites, and other platforms. The GTS manages the real-time collection and distribution of meteorological information. The WIS is responsible for discovering, accessing, and managing data and products. The GDPFS provides various climate forecasting data to users. By formulating a unified data policy and establishing this meteorological data-sharing system, the WMO has enabled the global sharing of Earth system science data in multiple fields, such as weather, climate, hydrology, atmospheric composition, cryosphere, and oceans. This study summarizes the achievements of WMO’s meteorological data-sharing system and its alignment with China’s relevant data strategy requirements. It assists in enhancing the understanding of international meteorological data-sharing activities and facilitating the construction of a similar multi-departmental comprehensive Earth observation data-sharing system in China.

The black soil region in Northeast China is a major grain-producing area in China. To ensure the sustainable development of agriculture in the black soil region, the data from the third national land resource survey, remote sensing data, and the digital elevation model (DEM) can be integrated to explore the ecological protection assessment methods for cultivated land. This study investigated Nenjiang City, Heilongjiang Province, from the location conditions of cultivated land and surrounding ecological land use. It constructed five indicators, including the forest health index, the proportion of ecological land surrounding cultivated land, the distance to the nearest forest, the slope, and the topographic position. Notably, an improved forest health index was designed based on the remote sensing ecological index to comprehensively assess the ecological protection of cultivated land in Nenjiang City. The results indicate that the cultivated land in Nenjiang City was dominated by medium-low and medium ecological protection grades, covering 34.21% and 45.28% of the cultivated land area, respectively. In contrast, the high-grade cultivated land accounted for merely 2.11%, indicating considerable potential for improving the ecological protection grade of cultivated land. Among individual indicators, the proportion of ecological land around cultivated land and the forest health index exhibited low values, serving as the primary factors leading to an overall slightly low geological protection grade in the study area.

Optical satellite remote sensing images of tropical and subtropical vegetation areas are often affected by cloud cover, leading to missing remote sensing information of surface features. Effectively detecting clouds, classifying clouds and objects, and extracting cloud cover information remain hot topics and challenges in remote sensing research. Many optical cameras in domestic satellites lack the short-wave and thermal infrared spectral bands, which are used in prevailing cloud detection algorithms, reducing the image data availability after cloud removal. Hence, this study suggested detecting the spatial distribution of cloud cover by utilizing only several spectral bands in the visible light - near-infrared range (400 nm to 1 000 nm). Based on the hyperspectral remote sensing images from the Zhuhai-1 satellite, this study constructed feature space scatter plots using spectral indices, including normalized difference vegetation index (NDVI) and normalized differential water index (NDWI), for cloud/object classification and detection. Moreover, this study extracted the cloud, water, and vegetation cover information from mixed pixels. The results demonstrate that compared to conventional cloud detection methods using spectral index thresholds, the cloud detection algorithm under the NDWI-NDVI feature space used in this study exhibited a superior cloud-water separation capability and simple operability. It can precisely describe the spatial distribution characteristics of cloud cover by suppressing the shadow effect on cloud cover. Overall, this study offers a novel technical approach for further developing cloud detection, cloud-water separation, and cloud cover information extraction algorithms for domestic optical satellite remote sensing data.

Resource development in mining areas alters land use patterns and causes ecological damage. This renders land use identification crucial to ecological restoration and management in mining areas. Although remote sensing imagery is widely used for land use classification, the use of a single data source has limitations in the classification for mining areas. Additionally, it is difficult for conventional machine learning algorithms to effectively perform the classification. To improve classification accuracy, this study investigated the eastern part of Dongsheng District, Ordos City as an example to conduct land use classification for mining areas using a convolutional neural network (CNN) combined with multi-source remote sensing data. First, a multi-source remote sensing time series feature set was developed using data from Sentinel-1/2, Luojia-1 01, and the NASA digital elevation model (DEM). Next, optimal features were selected using the Relief-F algorithm combined with a random forest algorithm. Finally, information on surface features was extracted using the ResNet50 CNN model. This facilitated land use classification in the mining area. The results show that the proposed method achieved an overall land use classification accuracy of 95.36% and a Kappa coefficient of 0.942 1, outperforming conventional methods such as the random forest approach. Furthermore, selecting optimal features using Relief-F combined with the random forest approach enhanced the classification accuracy of various classifiers. This study offers a methodological reference for land use classification of mining areas.