Latest Issue
Online FirstList of IssuesDocument Types
Research Progress in urban thermal environment research towards public health assessment AI Introduction
Abstract：Against the backdrop of global climate change and rapid urbanization, the widespread replacement of natural land covers by impervious urban surfaces has substantially altered surface energy exchanges and intensified urban thermal stress. Urban heat environments have therefore become a critical challenge for sustainable urban development, with far-reaching implications for thermal comfort, heat exposure, and public health. Quantifying the spatiotemporal dynamics of urban thermal environments, identifying their health-related consequences, and developing effective mitigation and adaptation strategies are essential for advancing the United Nations Sustainable Development Goals, particularly those related to sustainable cities and communities, good health and well-being, and reduced inequalities. In recent years, research on urban heat environments has undergone a notable transition from physical monitoring of thermal conditions toward human-centered assessments of thermal perception, exposure inequality, adaptive capacity, and heat-related health risks. To clarify this evolving research paradigm, this paper reviews major advances in urban heat environment studies over the past two decades through bibliometric analysis and thematic synthesis. First, this review summarizes progress in urban heat environment monitoring based on in situ observations, satellite remote sensing, and numerical simulations. These approaches have improved the characterization of air temperature, land surface temperature, and urban microclimatic processes across multiple spatial and temporal scales. Nevertheless, existing monitoring frameworks remain constrained by limited spatial coverage, discontinuous temporal records, insufficient all-weather capability, and uncertainties in cross-scale integration. Second, this paper examines the growing emphasis on outdoor thermal comfort, urban heat exposure, and heat-related health risks. Research attention has increasingly shifted from the physical intensity of urban heat islands to the ways in which residents perceive, experience, and respond to thermal stress. However, fine-scale and dynamic assessments of heat exposure remain limited by the lack of globally consistent diurnal temperature data and short-term population mobility information, which constrains the evaluation of exposure inequality across different places, periods, and social groups. Finally, four future research directions are proposed. (1) integrated retrieval and reconstruction of multiple temperature parameters should be strengthened by combining ground observations, multisource satellite remote sensing, microwave data, data fusion, deep learning, and physically constrained modeling, thereby generating long-term, high-resolution, and all-weather temperature datasets for urban heat studies. (2) future research should further reveal the coupling relationships between urban heat exposure and urban morphology, ecological structure, infrastructure provision, population dynamics, socioeconomic conditions, and social development patterns, with particular attention to heat exposure inequality. (3) public health-oriented adaptation should be advanced through nature-based solutions, urban greening, blue-green infrastructure, vertical vegetation, landscape optimization, and climate-sensitive urban design, while fully considering exposure, vulnerability, and accessibility to cooling and medical resources. (4) under future warming and more frequent extreme heat events, interdisciplinary studies are needed to expand heat-health risk assessment across spatial and temporal scales, investigate the impacts of persistent heat on vulnerable populations, and support medical resource allocation, emergency response planning, and urban heat risk management.Keywords：Urban heat environment;thermal comfort;heat exposure;health risks of high temperatures;urban heatwaves;urban heat island9|0|0
<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-09-04Asymmetric dynamics between ratoon rice and double-season rice suitability under future climate scenarios in China AI Introduction
Abstract：Objective Double-season rice and ratoon rice represent two important rice-cropping systems in China's subtropical monsoon region. However, against the background of global warming, systematic insights into their relative climatic advantages remain limited. This study aims to clarify the adaptive characteristics of these two rice-cropping patterns under future climate change, thereby providing a scientific basis for maintaining regional and global food security.Method In this study, we employed climatic environmental variables derived from a combination of remote sensing satellite data and global weather station observations integrated with crop spatial occurrence records to evaluate the climatic suitability of double-season rice and ratoon rice. A set of 12 widely used species distribution models was tested, and high-performing models were selected on the basis of accuracy metrics. Ensemble predictions were then generated through weighted averaging using the true skill statistic (TSS), providing a robust assessment of current and future suitability patterns under multiple climate scenarios. Suitable areas were delineated using TSS-maximization thresholds, and within suitable areas, we further divided the suitability levels into three levels by equally partitioning the probability range. Comparative analyses of the suitability levels between the two rice systems allowed us to identify areas where ratoon rice displayed a higher suitability level than double-season rice did, thereby highlighting spatial and temporal asymmetries in the responses to climate change.Result The results indicate that under current climatic conditions, the suitable area of ratoon rice (137.91×10⁵ ha) is significantly broader than that of double-season rice (98.56×10⁵ ha). However, future climate change will drive an asymmetric spatiotemporal evolution of these two rice systems. In terms of the temporal dimension, the suitable area for double-season rice shows a continuous expansion trend under all the emission scenarios, with an approximate 27% increase by the end of this century under the SSP585 scenario. In contrast, the suitable area of ratoon rice initially expands but then decreases, with its suitable area projected to decrease by 10.29×105 ha by the end of this century under the SSP585 scenario compared with that during the current period. In terms of the spatial dimension, the area suitable for double-season rice exhibits unidirectional northward expansion with minimal shrinkage at low latitudes, whereas the area suitable for ratoon rice significantly decreases in low-latitude areas and expands in high-latitude regions and southwestern highlands. Despite this projected contraction, the advantageous area for ratoon rice remains significant. Even under the SSP585 scenario, by the end of the 21st century, the potential advantage area for ratoon rice is estimated to be 24.59×105 ha, far exceeding its current cultivation area of 12.4×105 ha.Conclusion By mitigating projection uncertainties through ensemble modeling and multiscenario analyses, this study provides a robust scientific basis for optimizing regional cropping layouts. Given its currently limited cultivation scale, ratoon rice has demonstrated expansion potential under both present and future climate scenarios. The asymmetric effects of climate change on double-season rice and ratoon rice underscore the necessity of developing targeted adaptation strategies to maintain food production stability under dynamic climatic conditions.Keywords：remote sensing application;climate change;suitability;ratoon rice;double-season rice;species distribution model10|0|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-09-04
Abstract：Objective Large-scale identification of idle rural residential land remains challenging in mountainous regions because of complex topographic conditions, fragmented spatial patterns, and heterogeneous land-use states. This study aims to develop a multi-source remote sensing framework for regional-scale identification of idle risk in rural residential land by integrating terrain-constrained spatial extraction with temporal indicators of human activity and vegetation dynamics.Method Zhangjiakou, Hebei Province, China, was selected as the study area. An improved U-Net semantic segmentation model was developed by incorporating a Digital Elevation Model (DEM) channel and an attention mechanism into the conventional U-Net architecture to enhance the extraction of rural residential land under complex terrain. Sentinel-2 imagery and DEM data were jointly incorporated to strengthen the representation of spectral, spatial, and topographic information. Urban built-up area boundaries were subsequently introduced to exclude urban pixels and construct a rural residential land mask. Multi-source temporal remote sensing features were then derived from SDGSAT-1 low-light nighttime imagery and multi-temporal Sentinel-2 imagery. Nighttime light statistics were used to characterize human activity intensity and temporal variability, whereas Normalized Difference Vegetation Index (NDVI) statistics were employed to characterize vegetation conditions and spatial heterogeneity. Correlation analysis was performed to reduce feature redundancy, and ADASYN combined with SMOTE-Tomek was employed to alleviate class imbalance. Random Forest, XGBoost, and Light Gradient Boosting Machine (LightGBM) were comparatively evaluated, and the optimal model was used to estimate the posterior probability of rural residential land being idle. The resulting probabilities were spatially smoothed and classified into five risk levels, followed by independent stratified validation.Result The improved U-Net achieved an Intersection over Union (IoU) of 0.8994 and an F1-score of 0.9470, while object-level evaluation yielded a patch integrity of 0.91, a relative area error of 0.11%, and an average boundary offset of 4.8 m. Among the machine learning models, LightGBM achieved the best overall performance, with an accuracy of 0.813, an area under the receiver operating characteristic curve (AUC) of 0.874, a recall of 0.867, a precision of 0.789, and an F1-score of 0.826. Feature importance analysis indicated that NDVI and nighttime light features provided complementary information for idle-risk identification. The integrated feature set achieved an AUC of 0.874, compared with 0.813 and 0.843 for nighttime light-only and vegetation-only features, respectively. Validation based on 500 independent stratified samples yielded an accuracy of 0.783, a precision of 0.745, a recall of 0.830, and an F1-score of 0.786. The proportion of samples exhibiting clear idle characteristics increased progressively from 18% in the low-risk class to 82% in the high-risk class. Global Moran’s I increased from 0.842 before spatial smoothing to 0.927 after smoothing, indicating significant positive spatial clustering of the estimated idle probabilities.Conclusion The proposed framework integrates terrain-constrained deep learning, multi-source temporal remote sensing, and machine learning to achieve regional-scale identification and probabilistic characterization of idle rural residential land. The synergistic use of nighttime light and vegetation features effectively captures complementary signals associated with the attenuation of human activity and the enhancement of vegetation cover. The posterior probability provides a continuous spatial representation of idle status and reveals distinct spatial differentiation across Zhangjiakou. Relatively high idle-risk areas are mainly concentrated in the Bashang Plateau and some mountainous areas, whereas areas with stronger urban influence and better transportation accessibility generally exhibit lower risk. The proposed framework provides an effective remote sensing-based approach for monitoring inefficient rural residential land and supporting spatially differentiated land-use management and rural land consolidation.Keywords：Rural residential land;Idle risk;DEM terrain constraints;SDGSAT-1 low-light attenuation;NDVI time-series vegetation enhancement;U-Net10|0|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-09-04Long-term water color evolution mechanisms of large shallow lakes under anthropogenic disturbance： A case study of Lake Chaohu AI Introduction
Abstract：Large shallow lakes are important for regional water supply and ecological security, and their long-term water quality dynamics are of great scientific and practical importance for watershed management and ecological protection. However, these lakes are highly sensitive to external disturbances and ecologically vulnerable. Under the combined effects of intensive human activities and continuous environmental management, the extent to which their water quality exhibits staged and nonlinear responses remains poorly understood. In this study, Lake Chaohu, the fifth-largest freshwater lake in China, was selected as a representative large shallow lake. The Forel-Ule Index (FUI), which provides an integrated measure of water optical properties and water quality status, was used to characterize long-term water quality changes. Based on MODIS long-term remote sensing data from 2000 to 2024, a high-accuracy improved water color inversion model was applied to generate a long-term water color series, with an overall validation mean relative error (MRE) of 1.38% and root mean square error (RMSE) of 3.32°. The spatiotemporal patterns of water color and their driving factors under long-term human disturbance were then systematically examined. The results showed that the multi-year mean FUI of Lake Chaohu ranged from 12 to 15, with a mean hue angle (α) of 211.67° ± 2.58°, indicating an overall green-to-yellow water color. Spatially, a clear west-to-east decreasing gradient was observed, with mean α values of 213.13° ± 2.22°, 211.67° ± 2.58°, and 209.79° ± 2.16° in the western, central, and eastern lake regions, respectively. This spatial pattern was largely consistent with the distribution of major inflowing rivers and their differences in pollution loads. At the seasonal scale, water color showed a distinct bimodal pattern, with the highest α value in autumn (214.23°), indicating the most turbid conditions, and the lowest value in summer (208.74°), indicating the clearest conditions. Seasonal variations in natural factors, particularly wind speed, precipitation, normalized difference vegetation index (NDVI), and temperature, were found to regulate these intra-annual changes. Over the long term, the annual mean α increased at a rate of 0.15° yr⁻¹ from 2000 to 2024, with the largest increase occurring in the eastern lake region (0.26° yr⁻¹). The Mann-Kendall and Pettitt tests jointly identified around 2014 as a key transition point. The decreasing trend during 2000–2014 (-0.13° yr⁻¹) changed to a significant increasing trend during 2014–2024 (0.44° yr⁻¹), indicating a clear shift from an earlier improvement to later deterioration in the water environment. SHAP-based attribution further showed that anthropogenic factors, including gross domestic product (GDP), impervious surface area, nighttime light, and population, contributed substantially more to long-term water color changes than natural climatic factors and were identified as the dominant drivers. Early watershed pollution control and ecological restoration measures effectively reduced environmental pressure and promoted water color improvement. However, as urbanization and socioeconomic activities continued to intensify, management gains gradually weakened, resulting in an imbalance in the lake's pressure-response relationship and a subsequent rebound in water color deterioration. Meanwhile, the Yangtze-to-Huaihe River Diversion Project and extreme climate events further intensified hydrodynamic disturbances and external pollutant inputs, increasing interannual variability and the risk of short-term water color deterioration. Overall, the results revealed distinct staged, nonlinear, and spatially heterogeneous patterns in the long-term water color evolution of large shallow lakes under sustained human disturbance. These findings provide a scientific basis for long-term water environmental management and adaptive regulation of Lake Chaohu and other large shallow lakes.Keywords：shallow lake;Chaohu Lake;water color index;long-term remote sensing;anthropogenic disturbance15|1|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-09-04Mamba-CSNet： Joint CNN and Mamba State-Space Modeling for Hyperspectral Image Compressive Sensing AI Introduction
Abstract：Objective With the rapid development of hyperspectral imaging technology, the high dimensionality of hyperspectral images poses severe challenges to data storage and transmission. Resource-constrained scenarios, such as spaceborne platforms, create an urgent demand for efficient onboard compression techniques. However, existing deep learning-based compressive sensing methods still struggle to achieve an effective balance between reconstruction accuracy and computational efficiency under low sampling rates. In particular, Transformer-based models relying on attention mechanisms incur computational and memory costs that grow quadratically with sequence length when processing hyperspectral data, leading to significant efficiency bottlenecks in practical applications. To address these issues, this study introduces the Mamba selective state-space mechanism with linear complexity to improve hyperspectral image reconstruction accuracy while maintaining low computational complexity. A quantization coding strategy is further incorporated to enhance the overall compression performance of the model.Method A hyperspectral image compressive sensing network, termed Mamba-CSNet, is proposed based on the Mamba selective state-space mechanism and a two-stage quantization coding strategy. The network consists of three key components. First, a low-complexity dual-branch Mamba encoding structure is designed to model long-range dependencies along the spectral and spatial dimensions, respectively, and cross-dimensional fusion is used to obtain joint feature representations. Second, an adaptive quantization and Brotli coding module is introduced to further exploit internal feature redundancy and reduce storage overhead. Finally, a CNN-Mamba feature enhancement module, termed CMM, is constructed to leverage the complementarity between CNN and Mamba. This module collaboratively performs fine-grained local feature modeling and global context capture, thereby effectively improving the representation and reconstruction capability of compressed features.Result Comparative experiments are conducted on a hyperspectral dataset constructed in this study, containing 3,016 samples, and Mamba-CSNet is systematically compared with five representative hyperspectral image compression methods. The experimental results demonstrate that, under a 1% sampling rate, the proposed method achieves the best balance between reconstruction accuracy and computational efficiency. Compared with the current state-of-the-art method, Mamba-CSNet improves the PSNR by 0.872 dB, reaching 38.024 dB, while reducing the computational cost by 11.1%, with only 0.635 GFLOPs. Furthermore, benefiting from the proposed two-stage quantization coding strategy, the method exhibits excellent robustness in extremely bandwidth-limited scenarios and still achieves high-quality reconstruction with a PSNR of 37.058 dB at an ultra-low bitrate of 0.042 bits per pixel per band (bpppb).Conclusion By introducing the Mamba selective state-space mechanism and the two-stage quantization coding strategy, Mamba-CSNet effectively enhances the modeling of cross-band correlations in hyperspectral images and alleviates the limitations of existing learning-based models in terms of restricted reconstruction accuracy and high computational cost under low sampling rates. For practical applications involving sensors with varying numbers of spectral bands, future work will explore cross-band generalization mechanisms, with the aim of enabling unified compression and high-quality reconstruction of hyperspectral images with different band configurations using a single general-purpose model.Keywords：hyperspectral image;compressive sensing;deep learning;Mamba selective state-space mechanism;quantization coding;low bitrate;spaceborne platforms9|0|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-09-04Seasonal Dynamics-Aware Quantification of Traffic Sign Visibility and Simulation of Street Tree Pruning AI Introduction
Abstract：Objective Occlusion of traffic signs by street trees poses a threat to road traffic safety. However, existing studies mostly stop at static occlusion detection and lack a technical pathway that bridges occlusion quantification and pruning decision-making.Method This paper constructs a decision framework that connects seasonal occlusion dynamic quantification with multi-intensity pruning simulation and evaluation. An 8.0 km road section in Zhangjiang Hi-Tech Park, Shanghai, was selected as the study area, where 74 cantilever traffic signs and 1,524 street trees were analyzed. First, high-precision street tree models were reconstructed from the point clouds (mean error of 2.0 cm), and virtual leaves were incorporated to simulate seasonal variations. Second, a view frustum spatial analysis method was employed to quantify the visibility of traffic signs from the pavement in winter and summer. Finally, pruning simulations were conducted under two schemes with branch diameter thresholds of 5 cm and 10 cm, respectively, and the pruning effects were evaluated from three aspects: pruning cost, ecological loss, and landscape impact.Result The results show that the average visibility of traffic signs from the pavement is 91.15% in winter and decreases to 85.26% in summer, representing a reduction of 5.89%. Sign #7 exhibits the largest decrease of 41.45%, indicating the most significant seasonal impact. Signs with shorter cantilevers are more prone to occlusion. Under the 5 cm pruning scheme, branch volume loss is 6.20% and leaf loss is 20.70%; under the 10 cm scheme, branch volume loss increases to 19.20% and leaf loss to 39.21%, while the number of pruning points decreases from 558 to 219. Although high-intensity pruning reduces costs, it may cause severe structural damage to trees (e.g., a single tree near sign #43 experiences a branch loss of 57.69%, with some trees nearly reduced to trunks).Conclusion This study demonstrates that the proposed framework, by mapping occlusion quantification results into pruning threshold selection criteria and revealing the asymmetric cost relationship between the reduction of pruning points and the structural damage to trees, provides a scientific basis for the coordinated management of street trees that balances traffic safety, ecological conservation, and landscape preservation.Keywords：Mobile laser scanning point cloud;street tree;traffic sign;occlusion analysis;pruning simulation;traffic safety5|0|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-09-04Research and Remote Sensing Applications of Vegetation's Impact on Land Surface Energy Partitioning： Progress and Prospects AI Introduction
Abstract：Accelerated warming and intensified human activities have drastically altered global vegetation cover, profoundly affecting regional and global climate. Among these effects, vegetation change critically influences surface energy partitioning, which is the division of all-wave net radiation into latent, sensible, and soil heat fluxes. This partitioning directly regulates near-surface temperature, atmospheric stability, and boundary layer dynamics. Thus, clarifying vegetation's role in energy partitioning is essential for understanding the climatic impacts of land cover change and for improving climate models and land management. However, quantifying this response independently from background climate variability remains challenging. Nonlinear interactions, strong spatiotemporal heterogeneity, and limited long-term observational data hinder the separation of vegetation signals from climatic noise. Furthermore, the magnitude and even direction of vegetation-induced energy shifts depend on latitude, season, aridity, and ecosystem type, precluding simple generalizations. Focusing on vegetation–energy partitioning, this review systematically synthesizes studies over three decades, emphasizing remote sensing applications. Vegetation affects land surface energy partitioning via three interconnected pathways: (1) physiological traits (leaf area, stomatal conductance, root depth) regulating surface–atmosphere water and heat exchange; (2) modifying the surface radiation budget via albedo and emissivity changes; and (3) influencing the hydrological cycle through interception, transpiration, and soil moisture extraction, thereby altering latent and sensible heat fluxes. These processes are modulated by human activities (deforestation, irrigation, ecological restoration, urbanization) and natural disturbances (wildfires, extreme climate events). Energy flux interactions may amplify or offset one another; for example, increased latent heat typically suppresses sensible heat, cooling growing seasons, whereas reduced evaporation can worsen heatwaves in dry regions. Such complexity yields distinct climatic responses across regions and scales, demanding integrated assessments. Methodologically, numerical models, from land-surface schemes to Earth system models, still dominate, but observational evidence is gaining importance given model uncertainties in parameterization and forcing. Satellite remote sensing, with broad coverage, long time series, and multiple resolutions, overcomes the spatial sparsity of in situ networks. It supplies key variables (leaf area index, albedo, land surface temperature, evapotranspiration) as critical model inputs and constraints. Consequently, remote sensing has expanded vegetation–energy studies to global scales, enabling attribution of greening effects on energy balance and providing quantitative evidence for policy. Future research should prioritize several directions. First, enhancing the spatiotemporal resolution and accuracy of remote sensing products, through emerging techniques like remote sensing foundation models that merge deep learning with physical constraints, will better capture fine-scale heterogeneity and rapid land cover changes. Second, greater attention is needed for vegetation transition zones and complex terrains, including arid/semi-arid regions, urban areas, and mountains, where energy partitioning responses are especially sensitive and poorly understood. Third, incorporating coupled biogeochemical effects, particularly the links between carbon uptake, nutrient cycling, and energy exchange, will support more holistic assessments of vegetation–climate interactions. Fourth, fostering multi-platform, multi-sensor data fusion and developing robust causality attribution methods are critical for reducing uncertainties and guiding actionable strategies. Together, these advances will deepen mechanistic understanding of vegetation change and its regional climatic consequences from the energy partitioning perspective, ultimately contributing to sustainable ecosystem management and climate adaptation policies.Keywords：vegetation;surface energy partitioning;surface radiation budget;remote sensing;latent heat flux;sensible heat flux;land-atmosphere interactions;climate response12|0|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-09-04LSENet： Local Subdomain-aware and Edge-enhance Dynamic Fusion Network for Semantic Segmentation of Airborne LiDAR Point Cloud AI Introduction
Abstract：Objective Airborne LiDAR point cloud semantic segmentation is fundamental to 3D scene understanding and has important applications in smart cities, urban modeling, infrastructure extraction, land surveying, and environmental monitoring. To address the insufficient coordination between local geometric representation and contextual modeling in existing methods, we propose LSENet (Local Subdomain-aware and Edge-enhanced Dynamic Fusion Network). LSENet extracts fine-grained geometric features through adaptive local subdomain partitioning, enhances information exchange via cross-subdomain interactions, and improves feature representation and boundary discrimination through an edge-enhanced dual-attention fusion module. Experiments on the Vaihingen 3D and DFC2019 datasets achieve Overall Accuracy (OA) values of 83.9% and 97.8%, and mean F1-scores of 70.5% and 89.1%, demonstrating its effectiveness and cross-scene applicability.MethodLSENet adopts a U-Net-style encoder–decoder architecture with three core modules. The Local Subdomain Space-aware feature Encoding Aggregation module (LSSEA) constructs an adaptive local coordinate system using PCA and encodes relative position, spatial distribution, orientation, and curvature information for fine-grained geometric representation. The Local Subdomain feature Enhancement and Interaction Attention module (LSEIA) enhances intra-subdomain features and establishes cross-subdomain interactions to improve local contextual modeling. The Edge-enhanced Grouped Multi-scale Dual-Attention dynamic fusion module (EGMDA) integrates edge-enhanced grouped channel attention and multi-scale spatial attention through dynamic fusion, strengthening boundary representation and classification accuracy.Result Comparative experiments were conducted on two widely used benchmark datasets, ISPRS Vaihingen 3D and DFC2019. On Vaihingen 3D, LSENet achieved an OA of 83.9% and a mean F1-score of 70.5%, obtaining the best OA among compared methods. Significant improvements were observed in challenging categories such as power lines, roofs, and trees, demonstrating the effectiveness of the proposed modules in preserving complex structures and clear boundaries. On DFC2019, LSENet achieved an OA of 97.8% and a mean F1-score of 89.1%, ranking first in OA and demonstrating strong cross-scene generalization. Visualization results further show that LSENet produces more coherent regions and clearer boundaries than the baseline.ConclusionLSENet effectively integrates local geometric awareness, cross-subdomain interaction, and edge-preserving dual attention with dynamic fusion to overcome limitations of existing point cloud segmentation networks. Results on two airborne LiDAR benchmarks validate its effectiveness, robustness, and generalization capability. Future work will focus on developing more discriminative geometric descriptors, lightweight dynamic inference strategies, and evaluating the proposed method on newly released large-scale and diverse airborne LiDAR datasets to improve its applicability in real-world scenarios.Keywords：airborne LiDAR point clouds;deep learning;semantic segmentation;local subdomain awareness;dynamic fusion;edge-enhance10|0|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-09-04Fine-Tuning and High-Frequency Enhanced Multi-Scale Segment Anything Model for Remote Sensing Image Semantic Segmentation AI Introduction
Abstract：The Segment Anything Model (SAM), as a representative vision foundation model, has demonstrated remarkable performance and strong generalization capability in various downstream computer vision tasks. However, directly applying SAM to remote sensing image semantic segmentation remains challenging due to the substantial domain gap between natural images and remote sensing scenes. Specifically, SAM is primarily trained on large-scale natural image datasets, making it difficult to effectively adapt to the complex characteristics of remote sensing images, such as significant scale variations, diverse spatial distributions, and heterogeneous land-cover patterns. Moreover, the image encoder of SAM is constructed based on the Vision Transformer (ViT) architecture, which mainly focuses on global contextual representation through self-attention mechanisms but has limited capability in capturing local textures and high-frequency spatial details. Consequently, important boundary information and fine-grained structural features of remote sensing objects may be weakened during feature extraction, resulting in blurred object boundaries and degraded segmentation accuracy. To address these limitations, this study proposes a Multi-Scale SAM with Fine-Tuning and High-Frequency Enhancement (MSHF-SAM) for high-resolution remote sensing image semantic segmentation. The proposed framework aims to improve the adaptability of SAM to remote sensing scenarios by enhancing high-frequency feature representation, optimizing model transfer strategies, and designing a more effective segmentation decoder. Specifically, in the encoder stage, a high-frequency enhancement module is introduced into the original SAM image encoder to strengthen the extraction and preservation of high-frequency spatial information, thereby improving the model’s ability to characterize fine-grained textures and object boundaries. Meanwhile, a parameter-efficient fine-tuning strategy is incorporated to alleviate the feature distribution discrepancy between the pretrained SAM model and remote sensing images, enabling effective domain adaptation with only a small number of trainable parameters. In the decoder stage, the original SAM mask decoder is replaced with a MACU-Net-based decoder to achieve more accurate multi-scale feature reconstruction. Furthermore, a lightweight boundary refinement branch is developed to enhance the representation of object contours and improve segmentation details, while maintaining computational efficiency. Through the collaborative optimization of the encoder and decoder, the proposed method can effectively integrate global semantic information with local spatial details, providing more precise segmentation results for complex remote sensing scenes. Extensive experiments are conducted on two widely used high-resolution remote sensing semantic segmentation datasets, namely LoveDA and ISPRS Vaihingen. The experimental results demonstrate that MSHF-SAM achieves superior performance compared with several state-of-the-art segmentation methods in terms of mean Intersection over Union (mIoU), F1-score, and other evaluation metrics. Compared with the classical UnetFormer model, the proposed method improves the mIoU by 4.33% on the LoveDA dataset and by 2.85% on the Vaihingen dataset. These results verify the effectiveness of the proposed high-frequency enhancement strategy, parameter-efficient fine-tuning scheme, and multi-scale decoding architecture. The proposed framework provides a promising solution for adapting large-scale vision foundation models to remote sensing semantic segmentation tasks and demonstrates the potential of foundation models in intelligent remote sensing image interpretation.Keywords：deep learning;Segment Anything Model(SAM);semantic segmentation;high-frequency enhancement;parameter-efficient fine-tuning;remote sensing images9|0|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-09-04a lightweight network for joint remote sensing image registration and change detection AI Introduction
Abstract：Image registration and change detection are two fundamental procedures for extracting and interpreting multitemporal remote sensing information. However, most deep learning approaches treat them as separate tasks, and existing joint frameworks either lack an explicit and interpretable registration mechanism or neglect the collaborative interaction between spatial and semantic difference features. This study aims to develop a lightweight unified network that can align misregistered bi-temporal optical remote sensing images and accurately identify real land-cover changes under geometric offsets and pseudo-changes induced by spatio-spectral variations. A lightweight joint registration and change detection network, named LJRCDNet, is proposed. MobileNetV3-Large is first adopted as a shared encoder to extract four-scale features for both registration and change detection. A spatial consistency module is then designed to perform semi-dense keypoint detection and local descriptor construction. Guided by self-supervised correspondences and knowledge distillation, this module estimates a homography-based spatial transformation and aligns multi-scale feature maps of the pre-event image with those of the post-event image. In the aligned feature space, a spatio-temporal difference collaboration module is introduced. It combines spatial cross-attention and channel cross-attention to model the coupling relationship between spatial difference features and semantic difference features, while depthwise separable convolution and efficient channel attention are used to control computational cost. Finally, multi-scale collaborative difference features are fused through a scale-adaptive perception module and upsampled to generate the final change map. Experiments are conducted on the SVCD, SYSU-CD and SECOND datasets under both co-registered and misregistered settings, and the proposed method is compared with thirteen representative change detection networks. In the co-registered setting, the model without the registration branch still achieves competitive performance with only 5.75 M parameters and 2.90 GFLOPs, indicating the effectiveness of the difference collaboration design. In the misregistered setting, the complete LJRCDNet obtains IoU scores of 79.62%, 65.09% and 52.24% on SVCD, SYSU-CD and SECOND, respectively, outperforming the second-best methods by absolute gains of 4.52%, 5.56% and 5.66%. Qualitative results show that the proposed method suppresses false alarms caused by geometric offsets, illumination changes, seasonal variations and land-object appearance differences, while preserving more complete boundaries of changed objects such as vehicles, impervious surfaces, cultivated land and buildings. Ablation experiments further verify that SCM and STDCM are both necessary: removing SCM reduces IoU by 1.87%, 3.03% and 2.16%, removing STDCM reduces IoU by 1.54%, 3.71% and 1.29%, and replacing SCM with SIFT also degrades performance. Matching experiments show that SCM achieves the best 3-pixel matching accuracy on all datasets. LJRCDNet integrates explicit spatial registration and spatio-temporal difference enhancement into an efficient end-to-end framework. The spatial consistency module provides reliably aligned features for subsequent change detection, and the spatio-temporal difference collaboration module enhances true spatio-spectral differences while suppressing pseudo-changes. The network achieves higher accuracy and better robustness than existing methods with relatively low complexity, making it suitable for large-scale remote sensing image processing and deployment on resource-limited devices. Future work will extend the unified framework to heterogeneous remote sensing imagery and strengthen the robustness of the matching module when large-area changes reduce the number of valid correspondences and degrade their spatial distribution.Keywords：remote sensing images;registration;change detection;Spatial consistency;spatio-temporal difference collaboration22|0|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-09-04Remote Sensing Mapping of Xinjiang Cotton Based on the Theory of Stable Classification with Limited Sample AI Introduction
Abstract：Objective Xinjiang is the largest cotton-producing region in China, and mapping cotton distribution in Xinjiang is of great significance for cotton production management. The accuracy of machine learning-based remote sensing mapping depends on the quantity and quality of samples. Due to the extensive cotton cultivation area in Xinjiang, sampling costs are high, making it difficult to meet the demands of high-precision dynamic mapping.Method This study, based on the theory of stable classification with limited samples and considering the high spatial heterogeneity of Xinjiang, constructs sample sets and designs sampling strategies by incorporating geographic spatial constraints for different cotton-growing regions (Northern Xinjiang, Southern Xinjiang, and Eastern Xinjiang), and explores the stability mechanism of cotton classification under limited sample conditions from two perspectives: minimum sample size and tolerance for erroneous samples.Result (1) In Northern Xinjiang, Southern Xinjiang, and Eastern Xinjiang, classification accuracy remains stable (with a decrease in accuracy of less than 1%) when using samples accounting for 16%, 24%, and 14% of the respective sample sets, or when the proportion of erroneous samples in the sample sets is controlled within 25%, 25%, and 10%, respectively; (2) Cotton mapping in Xinjiang based on the minimum sample size achieves satisfactory mapping results, with overall accuracies of 90.56%, 83.17%, and 94.57% for Northern Xinjiang, Southern Xinjiang, and Eastern Xinjiang, respectively. The estimated cotton area at the county scale shows a strong correlation with existing cotton distribution maps, with an R² of 0.99 and an RMSE of 5.38 kha; (3) Transferring the sample sets to 2018, a year when the interannual variation in cotton planting area was less than the tolerance for erroneous samples, also yields satisfactory mapping results, with overall accuracies of 95.37%, 88.85%, and 90.57% for Northern Xinjiang, Southern Xinjiang, and Eastern Xinjiang, respectively. The estimated cotton area at the county scale shows a strong correlation with existing cotton distribution maps, with an R² of 0.98 and an RMSE of 6.27 kha.Conclusion In response to the long-standing reliance on massive quantities of high-quality samples for cotton remote sensing mapping in Xinjiang, this study introduces the theory of stable classification with limited samples. Considering the pronounced spatial heterogeneity of Xinjiang, geographic spatial constraints were incorporated into sample set construction and sampling design across different cotton-growing regions. The stability mechanism of cotton classification under limited sample conditions was investigated from two perspectives: minimum sample size and tolerance for erroneous samples. According to the characteristics of cotton-growing regions in Xinjiang, additional geographic spatial constraints were imposed in Northern Xinjiang, Southern Xinjiang, and Eastern Xinjiang during sample construction and sampling design. A classification error difference of 1% was adopted as the criterion for stable classification. The minimum sample sizes required to achieve stable classification in Northern Xinjiang, Southern Xinjiang, and Eastern Xinjiang were 16%, 24%, and 14% of their respective sample sets, while the tolerances for erroneous samples were 25%, 25%, and 10%, respectively. Cotton mapping in Xinjiang based on the minimum sample sizes, as well as sample transfer mapping conducted within the tolerance for erroneous samples, both achieved satisfactory mapping performance (R² of 0.99 and 0.98, respectively), demonstrating strong stability and interannual applicability of the proposed approach. This study provides a new theoretical basis and technical pathway for establishing a high-accuracy and high-timeliness cotton remote sensing mapping system in Xinjiang, a region characterized by complex planting structures and high sample acquisition costs.Keywords：Stable classification theory with limited sample;Xinjiang cotton;remote sensing mapping;machine learning;limited sample;geographic spatial constraints;sample transfer249|26|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-08-17Rotation-Transform-Based Scattering Characteristics Extraction and Classification in Agricultural Areas Using Dual-Polarization SAR Data AI Introduction
Abstract：Objective In the field of polarimetric synthetic aperture radar (PolSAR) remote sensing, the capability to conduct all-weather, day-and-night observations makes it a pivotal tool for crop classification. Currently, establishing correlations between decomposition features and crop types via machine learning algorithms has become a mainstream approach. However, existing dual-polarimetric decomposition methodologies, such as Stokes or Cloude decompositions, are constrained by their reliance on predefined and fixed scattering models. These conventional approaches are limited to extracting discrete scattering features, which often fail to sufficiently characterize the geometric structures and scattering processes of complex agricultural vegetation, thereby hindering the achievement of higher classification accuracy.Method To effectively address these issues, the paper proposes a rotation transformation methodology specifically designed for dual-polarimetric data. By using the polarimetric scattering angle
Abstract：Vegetation phenology reflects the integrated responses of ecosystems to climatic conditions, land surface modifications, and anthropogenic disturbances, and serves as a critical indicator for assessing ecosystem feedbacks to urbanization and climate change. Rapid urban expansion has substantially altered surface thermal environments, vegetation structure, and ecological processes, leading to differentiated phenological responses among urban cores, towns, and surrounding rural areas. Although previous studies have documented urban–rural phenological contrasts, most analyses have relied on moderate-resolution remote sensing products, which are subject to mixed-pixel effects and limited capacity to resolve fine-scale urban heterogeneity. Systematic investigations based on high spatial resolution data across large spatial extents remain insufficient. This study aims to (1) extract high-resolution vegetation phenology metrics for major Chinese cities using Sentinel-2 imagery, (2) quantify phenological differences along the urban–town–rural gradient, and (3) examine how these differences vary across vegetation types, climatic zones, and city-size categories. This study aims to (1) extract high-resolution vegetation phenology metrics for major Chinese cities using Sentinel-2 imagery, (2) quantify phenological differences along the urban–town–rural gradient, and (3) examine how these differences vary across vegetation types, climatic zones, and city-size categories. Using Sentinel-2 imagery from the Copernicus program (2019-2024), Enhanced Vegetation Index (EVI) time series were constructed for 128 cities across China and their adjacent town and rural areas. A dynamic threshold approach was applied to extract the Start of Season (SOS) and End of Season (EOS) from annual EVI trajectories. To ensure biological plausibility, phenological metrics were constrained within reasonable day-of-year ranges. The derived phenological dates were validated against ground-based observations and compared with the moderate-resolution MCD12Q2 phenology product to assess absolute accuracy using MAE and RMSE metrics. Urban–town–rural differences were quantified for each city and subsequently analyzed across vegetation types, climatic zones, and city-size classes. Statistical comparisons and regression analyses were employed to evaluate spatial heterogeneity and scaling patterns. Validation results indicate that Sentinel-2-derived phenology exhibits substantially lower MAE and RMSE values than the MCD12Q2 product, demonstrating improved absolute accuracy under heterogeneous urban landscapes. Nationally, urban vegetation shows a consistent spring advancement and autumn delay relative to surrounding areas. On average, urban SOS occurs 1.26 days earlier than in towns and 1.49 days earlier than in rural areas, while urban EOS is delayed by 1.51 days and 1.25 days relative to towns and rural areas, respectively, indicating an extended growing season in urban environments. Phenological responses vary significantly among vegetation types. Forest ecosystems exhibit the strongest sensitivity to urbanization, showing the largest magnitude of spring advancement and autumn delay. In contrast, other vegetation types display comparatively moderate responses. Climatic background further modulates urbanization effects. The temperate climate zone shows the most pronounced urban-rural phenological contrasts, whereas subtropical and tropical zones exhibit weaker and less stable differences. City size also influences phenological patterns. The advancement of urban SOS generally intensifies with increasing city size, suggesting a scaling effect associated with enhanced urban heat island intensity. However, the delay of EOS is more evident in small and medium sized cities and may weaken or even reverse in megacities, possibly due to complex interactions among thermal stress, vegetation management, and land surface heterogeneity. Overall, this study provides high-resolution, large-scale evidence of differentiated vegetation phenological responses to urbanization in China. Urban expansion systematically modifies growing season dynamics, characterized by earlier spring onset and delayed autumn senescence, although the magnitude and direction of these effects depend on vegetation type, climatic background, and city size. By leveraging Sentinel-2 imagery and a dynamic threshold extraction framework, this research improves the quantitative reliability of urban phenology assessment compared with conventional moderate-resolution products. The findings enhance understanding of how urbanization reshapes ecosystem seasonal dynamics and contribute to clarifying the socio-ecological implications of phenological shifts. These results provide scientific support for sustainable urban planning, ecological infrastructure optimization, and climate adaptation strategies under continued urban expansion.Keywords：remote sensing;urbanization;vegetation phenology;sentinel-2;dynamic threshold method;gradient difference646|156|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-06-12
Abstract：Objective Accurate monitoring of forest resources at the individual tree level is fundamental for forest ecosystem management. Unmanned aerial vehicle (UAV) visible light (RGB) imagery provides a cost-effective and high-spatial-resolution data source for these wide-area monitoring tasks. High-spatial-resolution imagery comprehensively records the fine contours of trees and the background habitat of the forest. Utilizing panoptic segmentation technology for unified interpretation enables the synchronous extraction of all forest elements. Nevertheless, interpreting highly closed-canopy forest scenes remains a critical challenge. Traditional deep learning approaches often decouple semantic segmentation for background elements and instance segmentation for individual trees, leading to severe pixel-level classification conflicts and spatial topology inconsistencies. Furthermore, the limited spectral information in RGB imagery frequently causes severe spectral confusion among adjacent trees. To systematically address these challenges, this study proposes an end-to-end forest panoptic segmentation model named FSC-Mask2Former.Method The proposed FSC-Mask2Former builds upon the Mask2Former baseline by introducing two core architectural improvements tailored to the unstructured features of forests. First, a Frequency-domain Texture Awareness (FTA) module is incorporated into the feature extraction pathway to compensate for the loss of micro-texture details caused by spatial downsampling, essentially functioning as a learnable high-pass filter in the feature space to retain critical edge gradients. Second, an Instance-aware Query Contrastive (IQC) head is integrated at the output of the Transformer decoder to maximize the inter-class feature distance between spectrally similar tree species, imposing an anisotropic constraint on the feature distribution to enlarge decision boundaries and fundamentally suppress category assignment conflicts. To evaluate the model, a densely annotated dataset was constructed using UAV RGB imagery from Gaofeng Forest Farm in the Guangxi Zhuang Autonomous Region, supplemented by data from Genhe City in the Inner Mongolia Autonomous Region, Jixi County in Anhui Province, and Hengzhou City in the Guangxi Zhuang Autonomous Region to validate model transferability.Result Comprehensive experiments demonstrate that FSC-Mask2Former significantly outperforms existing mainstream networks. The model achieves an overall Panoptic Quality (PQ) of 57.0%, a substantial gain of 11.0 percentage points over the baseline. Most notably, the foreground Recognition Quality (RQ) reaches 56.0%, representing a 12.0 percentage point increase. Visualizations confirm that FSC-Mask2Former effectively separates touching instances in high-canopy-closure forest areas, precisely delineates boundaries for morphologically irregular canopies, and maintains the spatial coherence of background elements. Furthermore, multi-region experiments indicate robust generalization capabilities across different geographical and ecological conditions.Conclusion The proposed FSC-Mask2Former successfully overcomes the bottlenecks of spectral homogeneity and task separation in UAV-based forest interpretation. This research proves that accurate full-element forest mapping can be realized using universally accessible UAV RGB imagery, providing a practical, robust, and highly cost-effective technical paradigm for modern forest resource monitoring.Keywords：UAV remote sensing;Panoptic segmentation;Mask2Former;Frequency domain analysis;contrastive learning;Individual tree recognition487|156|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-06-12Cloud Detection of Simulated Geostationary Microwave Sounding Satellite （FengYun-4/GeoMWS） over Oceans AI Introduction
Abstract：Objective Leveraging the advantage of high-frequency observations, geostationary satellites effectively compensate for the limitations of polar-orbiting meteorological satellites in terms of long revisit cycles and insufficient timeliness in monitoring short-lived severe weather. Compared to infrared bands, microwave radiation is less affected by clouds and exhibits excellent penetration through non-precipitating clouds. Consequently, deploying microwave sounding instruments on geostationary orbits has become a crucial development direction in meteorological observation. The Fengyun-4 Geostationary Microwave Sounder (GeoMWS), currently under development in China, aims to achieve all-weather, all-day, and high-frequency three-dimensional observations of cloud and precipitation systems and their internal structures. This capability is of great significance for monitoring rapidly evolving weather systems and improving Numerical Weather Prediction (NWP). However, a prerequisite for satellite data assimilation (whether clear-sky or all-sky) is the accurate identification of clear-sky and cloudy pixels, along with the assignment of appropriate observation errors. Given that existing cloud detection algorithms are predominantly designed for polar-orbiting satellites and are difficult to directly adapt to the characteristics of geostationary microwave observations, this study aims to develop a specialized cloud detection method for GeoMWS to facilitate the effective assimilation of its data.Method This study proposes a fast and efficient cloud detection algorithm specifically tailored for the GeoMWS instrument. The method constructs a cloud detection index based on two window channels (Channel 2 and Channel 4). It fully accounts for the sensitivity differences of these channels across various latitudinal regions and introduces brightness temperature normalization to eliminate background noise caused by temperature variations. Furthermore, the optimal threshold for the cloud detection index is scientifically determined by analyzing the evolution characteristics of the index under different classification criteria.Result Evaluation results demonstrate that the proposed method achieves robust detection performance across different time periods. The Probability of Detection for clouds (POD_cld) and the Hit Rate (HR) both exceed 75%, indicating a high capability in identifying cloudy scenes. Meanwhile, the False Alarm Rate for clouds (FAR_cld) is effectively controlled below 20%. Compared to previous methods applied to this context (which maintain a detection rate of approximately 60%), the proposed method exhibits significant advantages in accuracy.Conclusion This study successfully develops a cloud detection algorithm optimized for GeoMWS. This method not only provides high-precision cloud detection information for data assimilation under both clear-sky and cloudy conditions but also offers reliable theoretical and methodological support for future operational applications. Particularly in critical areas such as typhoon monitoring and NWP data assimilation, it will significantly enhance the application effectiveness of geostationary microwave satellite data.Keywords：geostationary orbit;microwave sounding;geostationary microwave sounding;cloud detection over ocean;simulated data264|121|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-06-12
Abstract：The micro laser retroreflector array INRRI (INstrument for Landing-Roving Laser Retroreflector Investigations) is one of the international payloads onboard the Chang’e-6 mission. It receives laser signals emitted by the laser equipment onboard lunar orbiters and utilizes the characteristic of parallel reflection of incident laser on the reflective surface to achieve high-precision ranging. Through repeated observations and calculations, INRRI has become the very first absolute control point on the far side of the Moon, providing fundamental support for lunar geodesy, remote sensing mapping and positioning, high-precision orbit determination and navigation of lunar orbiters. This study presents mechanical structure and optical design of INRRI, and analyzes its key performance. In terms of design, a spherical dome structure and coating technology are adopted to enable INRRI to effectively receive the orbiter’s laser beam within a wider field of view. In terms of performance analysis, an effective reflection area model is established to analyze the reflecting signal intensity under different incident angles. The results show that the effective reflection area of a single corner cube reflector (CCR) reaches approximately 1 cm² under normal incidence, and INRRI can maintain stable effective reflection performance within a 60° half-aperture field of view. Furthermore, combined with the velocity aberration effect and far-field diffraction theory, the detectability of INRRI by the laser equipment onboard lunar orbiters is comprehensively evaluated. The dihedral angles of the CCR measured by a ZYGO interferometer are (0.37″, 0.64″, 0.37″), corresponding to a total beam deviation angle of approximately 3.01″. By establishing a far-field diffraction optical path and conducting simulation analysis, it is found that the far-field diffraction pattern exhibits spots with varying intensities distributed over an angular range of 40 μrad, thereby covering the calculated maximum velocity aberration offset angle of 11.01 μrad and satisfying the requirements of orbital observation and compensation. The observability of INRRI has been further confirmed by multiple successful detections using the Lunar Orbiter Laser Altimeter (LOLA) onboard the Lunar Reconnaissance Orbiter (LRO). This demonstrates the feasibility of employing micro laser retroreflector arrays as absolute control points on the lunar surface and provides a reference for the design optimization and future application of micro laser retroreflectors in subsequent deep-space exploration missions.Keywords：Laser retroreflector;laser ranging;effective reflection area;velocity aberration;far-field diffraction453|121|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-05-21Synergistic use of satellite and UAV imagery for continuous monitoring of field-scale paddy rice NDVI AI Introduction
Abstract：Objective The real-time and precise monitoring of crop growth status at the field scale is a critical component for achieving modern precision agriculture. Multi-source remote sensing technologies, including satellite platforms and unmanned aerial vehicles (UAVs), have emerged as effective non-destructive tools for this purpose. However, these data sources present a significant spatiotemporal trade-off, limiting their independent utility. Satellite remote sensing, while offering broad-area coverage, is often constrained by adverse weather conditions and insufficient spatial resolution to capture in-field variability. Conversely, UAV remote sensing provides exceptionally high spatial resolution but is hampered by limited battery endurance, making large-scale continuous (e.g., daily) monitoring challenging. Consequently, a single data source is inadequate for supporting the continuous monitoring of crop growth at the field scale. To address this critical gap, this study proposes a cross-platform spatiotemporal fusion method. The objective is to synergistically integrate satellite and UAV data, effectively leveraging their complementary temporal and spatial resolutions to generate a continuous, high-resolution dataset for precision agriculture.Method This research was based on a synergistic combination of multi-platform, multispectral remote sensing data, including high-resolution UAV imagery, Sentinel-2 data and PlanetScope SuperDove data. We developed an improved CACAO (Consistent Adjustment of the Climatology to Actual Observations) algorithm, adapting its core logic for cross-platform data fusion rather than its original climatological application. Two distinct data combination strategies were designed and tested: (1) a baseline “UAV+Sentinel-2” strategy and (2) an enhanced “SuperDove+Sentinel-2+UAV” strategy, which integrates high-frequency commercial satellite data. The CACAO framework was implemented using two distinct modes: a “forward prediction” (FP) mode, designed for near real-time applications, and a “backward updating” (BU) mode, which iteratively refines historical estimates as new data becomes available. The final output of the framework is a near real-time, daily 1-meter resolution normalized difference vegetation index (NDVI) time-series dataset. The accuracy of the fusion results was rigorously evaluated using two methods: (1) Leave-One-Out Cross-Validation (LOOCV), which assesses the model’s predictive power, and (2) a benchmark comparison against the established GLM-STF (Generalized Linear Model-based Spatiotemporal Fusion) algorithm.Result The prerequisite for data fusion was confirmed as NDVI data from the different platforms exhibited good consistency, with a strong correlation between Sentinel-2 and SuperDove (R = 0.97) and a reliable correlation was observed between UAV and satellite data (R > 0.75). In addition, the CACAO algorithm was proven to effectively reconstruct the phenological dynamics of the rice crop. A key finding was that the backward updating (BU) mode produced a significantly smoother and more robust NDVI time series than the forward prediction (FP) mode. Both CACAO-based data combination strategies achieved high overall accuracy (R > 0.94). Critically, the study demonstrated that introducing high-temporal-resolution SuperDove data during key phenological stages can substantially improve accuracy, with the correlation increasing from 0.51 to 0.67 in a specific validation case. Finally, in the comparative analysis, the CACAO algorithm demonstrated greater stability and slightly higher accuracy than the GLM-STF algorithm, particularly showing more robust performance across the entire growing season.Conclusion In conclusion, the cross-platform fusion framework proposed in this study, centered on the improved CACAO algorithm, is an effective and robust solution for generating continuous, high-precision (daily 1-meter) field-scale rice NDVI time series. This approach successfully overcomes the limitations of single-source data platforms. The framework provides strong technical support for the fine-grained monitoring of crop growth and the implementation of precision management strategies in modern agriculture.Keywords：PlanetScope;Sentinel-2;spatiotemporal data fusion;growth monitor;precision agriculture;field scale;phenological curve;near real-time monitoring1031|1090|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-03-17Unsupervised Automated Remote Sensing Monitoring of Crop Lodging for the Entire Pre- and Post-Disaster Process AI Introduction
Abstract：Objective Crop lodging poses a significant threat to agricultural productivity and food security, yet existing monitoring approaches often suffer from low automation, insufficient integration of pre- and post-disaster data, and a lack of systematic spatiotemporal coordination. To overcome these limitations, we developed an automated framework, StandardCurve-iForest-RF, which aims to establish a resilient crop growth baseline, distinguish true lodging from noise, and enable precise spatiotemporal mapping for disaster management.Method The approach utilizes time-series Sentinel-2 satellite data to construct a Crop Growth Standard Curve (CGSC) for the target crop. The Soft Dynamic Time Warping (Soft-DTW) algorithm is employed to create this curve, which serves as a resilient reference model capable of accommodating inter-annual climatic variations. To detect lodging, the method calculates cumulative multi-feature anomaly scores by comparing post-disaster satellite observations against the pre-established standard curve. An Isolation Forest (iForest) algorithm is applied for initial anomaly detection across multiple spectral features. Subsequently, a spatiotemporal joint decision mechanism is implemented to refine the results, effectively suppressing false alarms caused by persistent cloud cover, cloud shadows, and other environmental noise. Finally, a Random Forest (RF) classifier is used to accurately map the spatial extent and precise boundaries of the lodged areas, completing the fully automated workflow from dynamic detection to precise mapping.Result The method was validated using a case study of a lodging event that occurred on September 15, 2020, in Bohetai Township, Zhaoyuan County, Daqing City, Heilongjiang Province. It successfully identified the lodging event, demonstrating its effectiveness in distinguishing actual crop damage from noise. The overall detection accuracy reached 80.36%, with a Kappa coefficient of 0.60, confirming a substantial agreement between the automated detection results and ground reference data. The results clearly showed that the integrated use of the standard curve, the anomaly scoring mechanism, and the spatiotemporal decision rules significantly enhanced the reliability of lodging identification and minimized false positives. The entire process, from data processing to the final generation of the lodging map, was executed automatically without manual intervention.Conclusion The StandardCurve-iForest-RF framework presents a significant advancement in automated crop disaster monitoring. Its core innovation lies in the construction of a resilient growth standard curve and a sophisticated spatiotemporal analysis pipeline that effectively differentiates true lodging from interference. The successful application in a real-world case study confirms the method's practical utility and accuracy. This framework provides a valuable tool for agricultural departments and emergency management agencies, enabling rapid assessment of crop damage extent and supporting timely disaster response and loss estimation. The methodology is adaptable and holds promise for application in other regions and for monitoring other types of abrupt agricultural disasters.Keywords：Crop Lodging Monitoring;Standard Growth Curve;time-series analysis;Spatiotemporal Joint Detection;Sentinel-2444|999|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-03-13Weakly Supervised Semantic Change Detection in Street View Imagery and Its Application in Urban Renewal Dynamics Mapping AI Introduction
Abstract：Objective Street view imagery (SVI) has emerged as an important geospatial big data source for perceiving the urban built environment. Accurately detecting facade-level changes and identifying their semantic categories is essential for monitoring urban renewal dynamics. However, existing change detection approaches struggle to separate temporal ownership of changed objects (change decomposition) and to directly provide semantic change information, leading to complex workflows and high data preparation costs. This study aims to develop a weakly supervised semantic change detection framework that integrates change decomposition and semantic labeling, and to apply it to dynamic mapping of urban renewal in Guangzhou, China.Method We propose Cross-C2PO, a novel dual-branch architecture designed to achieve end-to-end weakly supervised change decomposition. Unlike traditional single-branch models, Cross-C2PO introduces a cross-comparison mechanism to explicitly model asymmetric temporal differences, ensuring completeness and consistency regardless of input order. The model integrates a differentiable union operator to maintain consistency constraints during weak supervision and employs the proposed Cross-MTF feature fusion function to break commutativity for accurate temporal differentiation. Building on Cross-C2PO outputs, we design a semantic change detection workflow that leverages state-of-the-art segmentation models (e.g., DeepLabV3+) without requiring synthetic datasets. Finally, we introduce an urban renewal dynamic index to quantify facade-level changes and visualize renewal patterns across panoramic and directional views (front, back, left, right) for Guangzhou’s central districts (2013–2019), based on 11,431 pairs of Baidu street view panoramas.Result Traditional change detection methods fail to achieve end-to-end change decomposition and rely on complex multi-stage pipelines, limiting scalability and flexibility. In contrast, the proposed Cross-C2PO framework enables one-stage weakly supervised change decomposition and can seamlessly integrate with mainstream architectures, granting them both improved detection accuracy and the ability to perform temporal ownership splitting without additional labels. Experiments on multiple benchmark datasets demonstrate that our method consistently achieves state-of-the-art performance, outperforming existing approaches in both binary change detection and decomposition tasks. Ablation studies further validate the contribution of the cross-branch structure, Cross-MTF fusion, and the differentiable union operator. Applied to Guangzhou street view imagery, the workflow successfully produced urban renewal dynamic maps, revealing high-intensity updates clustered in Liwan and Baiyun industrial areas, while moderate changes dominate residential zones. Directional view analysis additionally highlights local disparities and micro-scale renewal patterns.Conclusion The proposed Cross-C2PO framework offers a simple yet effective solution for weakly supervised semantic change detection, enabling accurate change decomposition without additional synthetic labels. Combined with an interpretable urban renewal dynamic index, it provides a scalable and cost-effective approach for urban facade change analysis. This study bridges street view imagery and AI-based computer vision for urban analytics, offering new insights into spatiotemporal renewal dynamics. Future work will focus on optimizing computational efficiency and extending the method to multi-source data integration for large-scale applications.Keywords：urban renewal;street view imagery;semantic change detection;scene change detection;weak supervision;dynamic index654|871|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2026-03-06GF-1 Wide-Field Multi-temporal Water Body Extraction Dataset AI Introduction
Abstract：Objective This study aims to overcome the limitations of existing public water body datasets, such as single temporal resolution and low annotation accuracy. The objective is to construct a high-quality, multi-temporal lake extraction dataset based on the high-resolution wide-field multispectral imagery from the "GF-1" satellite, which offers improved temporal and spatial coverage of water bodies.Method To achieve this, three study areas with varying levels of dynamic change were selected: Poyang Lake (high dynamic change), Namtso Lake (moderate dynamic change), and Yangcheng Lake (low dynamic change). These areas were covered in four seasons of 2022. The GF-1 wide-field multispectral imagery underwent preprocessing, including radiometric correction, orthorectification, and quick atmospheric correction. For the annotation process, a hybrid strategy combining automated methods with manual visual interpretation was employed to ensure high annotation accuracy.Result The resulting dataset is characterized by multi-temporal data and high annotation accuracy, offering a significant improvement over existing datasets. The overall accuracy of the dataset for all three study areas and across all four seasons exceeded 94%.It provides reliable data for dynamic water body mapping and change monitoring across seasonal variations. Additionally, various water body extraction methods, including threshold segmentation, traditional machine learning algorithms, and deep learning techniques, were employed to validate the dataset’s practical utility. The results demonstrated that the dataset supports the effective training and evaluation of these methods.Conclusion The findings indicate that the constructed multi-temporal lake extraction dataset is highly reliable and can effectively support various water body extraction methods. It provides a robust data foundation for enhancing the performance of dynamic water body extraction algorithms, and contributes valuable data for research in dynamic water body monitoring and mapping using high-resolution remote sensing imagery.Keywords：Gaofen-1;Dynamic Water Body;water body extraction;dataset;feature extraction733|906|0<HTML><L-PDF><Enhanced-PDF><Meta-XML>Updated：2025-12-11
