Hanfa Xing

Cloud occlusion severely restricts the utility of optical remote sensing imagery, particularly when thick clouds obscure the underlying surfaces. Inspired by advancements in diffusion models, this study proposes a CLoud-context- and scene fEAture-guided adaptive cloud Removal method via a residual Diffusion model (CLEAR-Diff), designed for effective single-temporal cloud removal. CLEAR-Diff utilizes cloud residuals as the reconstruction target within a diffusion framework to capture cloud-context characteristics accurately. A cloud-context prompt generator is designed to dynamically generate prompt information by learning the morphological features of cloud residuals, enabling adaptive processing of various cloud-context patterns, especially opaque clouds, and improving denoising performance. Additionally, a scene feature generation branch, built on a denoising network, extracts latent ground information beneath cloud-covered regions, providing reliable scene priors to enhance structural detail recovery during image reconstruction. Experimental results demonstrate that CLEAR-Diff significantly improves reconstruction quality in single-temporal cloud removal tasks, particularly under thick cloud conditions. The model excels in thin cloud scenarios and enhances structural detail recovery and fidelity in thick cloud regions. Compared with the state-of-the-art (SOTA) single-temporal approaches, CLEAR-Diff improves PSNR by 4.92 dB on RICE1 and 3.26 dB on RICE2, achieving 37.303 dB and 36.821 dB, respectively. The source code is available at https://github.com/oTvTog/CLEAR-Diff.

Place names—functioning as both locational identifiers and symbolic texts that encode power, economy, culture, and collective memory—exhibit discernible toponymic evolution patterns over time. Existing quantitative approaches to toponymic studies primarily depend on spatial distribution analysis based on attribute–value frameworks, disregarding the richer semantic and relational contexts underlying name changes. This study proposes a framework that integrates knowledge graphs and large language models (LLMs) to detect, classify, and interpret toponymic evolution patterns. Employing 143 Chinese historical cities as a case study, we construct a multilayered geographic knowledge graph capturing key events and figures, spatial changes, and local customs, thereby developing a knowledge graph–enhanced LLM for semantic extraction, graph retrieval, and pattern identification. The results reveal four distinct types of toponymic evolution: economic prosperity–driven (clustered in coastal and riverine regions and characterized by frequent new additions and abolitions), historical continuity–driven (concentrated in the political cores of successive dynasties and dominated by affiliation and administrative adjustments), natural resource–driven (clustered in regions with distinctive geographical features and characterized by renamings and new additions), and strategic hub–driven (found in vital transport corridors and military centers and characterized by abolitions and seat relocations). This study provides a scalable, data‐driven approach for elucidating the semantic logic underlying place name change and advances research in digital humanities, cultural geography, and spatial governance.

The spatial diffusion of culture describes how cultural elements spread and evolve across regions over time, shaped not only by spatial proximity but also by shared meanings and symbolic connections. Existing geographic studies, however, have largely emphasized spatial patterns while overlooking the role of cultural similarity in this diffusion. To address this gap, this study proposed a geographic knowledge graph (GeoKG) approach for identifying cultural diffusion patterns by integrating semantic similarity and geographic proximity. Based on the field model, we first constructed a large language model (LLM)‐enhanced GeoKG of cultural elements, from which cultural similarity that encodes various cultural entities (e.g., historical figures and events, architectural components and styles) and their relationships was quantified. This similarity served as a weight in generating a cultural density surface, enabling the identification of cultural hearths, regions, and diffusion trajectories using spatial analyses and costdistance calculation. Applying this framework to Buddhist architectural culture in Shanxi Province, the results revealed that Mount Wutai functions as the cultural hearth and the Taiyuan–Linfen corridor forms a first‐level cultural region. A bidirectional expansion pattern was also identified, characterized by northward diffusion toward Datong and southward diffusion along the Mount Wutai–Taiyuan–Linfen–Yuncheng corridor. These patterns may reflect the interaction between political centrality and socioeconomic connectivity in shaping cultural diffusion. This study developed a GeoKG‐based framework for modeling cultural diffusion, integrating semantic similarity and spatial proximity to offer a generalizable approach for analyzing geographically anchored cultural elements.

The remote sensing semantic segmentation task for offshore floating aquaculture (OFA) areas faces several challenges, including dense spatial distribution of aquaculture facilities, class imbalance, omission of small targets, and limited computational efficiency. To address these issues, this study constructs an NDVI-GB dataset based on the Optimum Index Factor (OIF) and proposes a Vision Transformer with Multi-scale Feature enhancement and Dynamic Attention fusion (ViT-MFDA). Based on the Swin Transformer framework, the model designs a Spatially Separable Self-Attention Vision Transformer (SSAVT) as the backbone network. By incorporating the Spatially Separable Self-Attention (SSSA) mechanism from Twins-SVT to optimize both local and global feature extraction, and integrating Conditional Positional Encoding (CPE) to enhance positional awareness, the model achieves efficient extraction of global context and local detailed information. Additionally, a Multi-scale Dynamic Weighted Attention Decoder (MDWAD) module is designed. This module integrates multi-scale contextual information through the Pyramid Pooling Module (PPM) and the Feature Pyramid Network (FPN), and introduces a dual channel-spatial attention mechanism to adaptively calibrate target features, which effectively mitigates false and missed detection in complex marine environments. Furthermore, ViT-MFDA employs a weighted joint loss function to alleviate class imbalance issues, improving the completeness and accuracy of segmentation results. Experimental results demonstrate that ViT-MFDA achieves mIoU, mF1-score, and mPrecision scores of 90.51%, 94.95%, and 95.34%, respectively, outperforming mainstream CNN and Transformer models in comprehensive performance, particularly excelling in dense target segmentation and model generalization. The method advances efficient large-scale aquaculture monitoring with real-world applicability.

Dynamic monitoring of cropland via remote sensing imagery is vital for safeguarding cropland resources and ensuring food security through accurate change detection (CD). However, convolutional neural network (CNN)-based methods, reliant on extensive convolution and pooling operations to extract deep cropland features, often accumulate irrelevant features and lose critical ones, degrading performance, especially in dense, continuous cropland regions. To address this, we first modified the recurrent residual convolutional neural network (R2U-Net) to improve the ability to extract cropland boundary, and developed a novel feature-difference-enhanced network (FDE-Net) to capture pixel-based cropland change information. Furthermore, we introduced a majority voting mechanism method to generate parcel-based cropland change maps from the cropland boundary and pixel-based CD results produced by the CNN. Using Sentinel-2 imagery as our data source, experiments revealed that R2U-Net achieved precision and F1-scores of 89.93% and 85.78%, respectively, for cropland boundary extraction. The FDE-Net yielded pixel-based cropland change detection results with precision, F1-scores and IOU of 68.09%,75.08% and 84.26%, respectively. Ultimately, we produced parcel-based cropland change maps for the Huicheng District through the majority voting mechanism, resulting in precision and F1-scores of 78.91% and 81.64%. Additionally, we extracted various types of cropland changes in the Huicheng District from 2019 to 2020, including cropland to cropland, cropland to urban, and cropland to other uses. Compared to the pixel-based method, the overall accuracy for these three change types was 84.86%, 81.86%, and 82.36%, respectively. Our methodology can obtain a more comprehensive and precise depiction of land use changes. Critically, this approach has surmounted challenges posed by fragmented or ambiguous change detection regions, ultimately elevating the reliability and fidelity of our findings.

Methane is a potent greenhouse gas that significantly drives climate change and affects human health and ecosystems by promoting surface ozone formation. However, the spatial continuity of current satellite-derived column-averaged dry-air mole fraction of methane (XCH₄) data remains limited, and the sparse distribution of ground-based observations further hinders our understanding of methane sources and trends. To address this, we developed a novel spatiotemporal Transformer model to fill gaps in satellite retrievals and correct biases, thereby generating a global, daily, gapless XCH₄ dataset at 0.1° × 0.1° (∼10 × 10 km near the equator) grid resolution from 2003 to 2020. This model integrates XCH₄ retrievals from the Scanning Imaging Absorption Spectrometer for Atmospheric Cartography (SCIAMACHY) and Greenhouse Gases Observing Satellite (GOSAT), together with XCH₄ reanalysis from the Copernicus Atmosphere Monitoring Service (CAMS), meteorological variables, and other auxiliary datasets. Our fused daily XCH₄ dataset yielded a mean coefficient of determination (R²) value of 0.93, a root-mean-square error (RMSE) of 13.10 ppb, and a bias of 7.45 ppb when independently compared with measurements from the Total Carbon Column Observing Network (TCCON). Furthermore, we performed bias correction on the reconstructed XCH₄ data using the TCCON network and supplementary variables to improve consistency across satellites and enhance accuracy, achieving a mean cross-validation R² (CV-R²) of 0.97, an RMSE of 7.51 ppb, and a bias of 0.51 ppb. We observed a global annual growth rate of 6.09 ppb/yr (p < 0.001) from 2003 to 2020, with the average annual trends in developed countries being much lower than those in developing countries, at 5.86 ppb/yr and 6.16 ppb/yr (p < 0.001), respectively. Bangladesh exhibited the highest annual average concentration and trend, with average values of 1850.21 ± 41.85 ppb and 8.07 ppb/yr (p < 0.001), respectively. This unique global gapless daily XCH₄ dataset enables the precise detection of local emissions and provides valuable support for informing methane emission management policies and enhancing our understanding of global methane sources and their spatiotemporal variations.