Pub Date : 2026-02-03DOI: 10.1016/j.jag.2026.105135
Yang Yang, Jiangong Xu, Yuchuan Bai, Liangyu Chen, Junli Li, Jun Pan, Mi Wang
Synthetic Aperture Radar (SAR) is crucial for Earth observation because it can acquire high-resolution images in all weather conditions. However, the presence of speckles—an inherent multiplicative noise caused by the coherent imaging process—severely degrades image quality and impairs the performance of subsequent interpretation tasks. To effectively capture both global contextual cues and fine-grained structural details in SAR image despeckling, we design a dual-branch Global-Local Collaborative Network (GLCNet) based on blind-spot convolution. GLCNet is trained in a self-supervised manner, requiring only original images for learning, making it well-suited for SAR data without ground truth. In the global branch, the SAR image is first decomposed into multiple frequency sub-bands through a Wavelet-Shuffle Downsampling (WSD), which decorrelates speckle components across scales and frequencies. A multi-scale blind-spot convolution is then applied to each sub-band in parallel, enabling the extraction of global textures without introducing speckle bias. In contrast, the local branch focuses on structure-aware restoration by jointly modeling frequency and spatial priors. By leveraging neighboring-pixel dependencies, this branch enhances local detail recovery and edge sharpness. Finally, an adaptive Detail-Guided Module (DGM) dynamically integrates complementary features from both branches, ensuring a harmonious balance between texture smoothness and structural fidelity. The proposed method is validated using various SAR sensors, including Sentinel-1, GF-3, TerraSAR-X, and Capella-X, demonstrating its superiority over traditional and deep learning approaches. Additionally, the application analysis confirms that the method enhances both the visual quality and analytical reliability of SAR images, making it a valuable preprocessing step for real-world scenarios. For reproducibility, our code and data are available at https://github.com/yangyang12318/LGCN.
{"title":"Self-supervised global − local collaborative network for real SAR despeckling","authors":"Yang Yang, Jiangong Xu, Yuchuan Bai, Liangyu Chen, Junli Li, Jun Pan, Mi Wang","doi":"10.1016/j.jag.2026.105135","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105135","url":null,"abstract":"Synthetic Aperture Radar (SAR) is crucial for Earth observation because it can acquire high-resolution images in all weather conditions. However, the presence of speckles—an inherent multiplicative noise caused by the coherent imaging process—severely degrades image quality and impairs the performance of subsequent interpretation tasks. To effectively capture both global contextual cues and fine-grained structural details in SAR image despeckling, we design a dual-branch Global-Local Collaborative Network (GLCNet) based on blind-spot convolution. GLCNet is trained in a self-supervised manner, requiring only original images for learning, making it well-suited for SAR data without ground truth. In the global branch, the SAR image is first decomposed into multiple frequency sub-bands through a Wavelet-Shuffle Downsampling (WSD), which decorrelates speckle components across scales and frequencies. A multi-scale blind-spot convolution is then applied to each sub-band in parallel, enabling the extraction of global textures without introducing speckle bias. In contrast, the local branch focuses on structure-aware restoration by jointly modeling frequency and spatial priors. By leveraging neighboring-pixel dependencies, this branch enhances local detail recovery and edge sharpness. Finally, an adaptive Detail-Guided Module (DGM) dynamically integrates complementary features from both branches, ensuring a harmonious balance between texture smoothness and structural fidelity. The proposed method is validated using various SAR sensors, including Sentinel-1, GF-3, TerraSAR-X, and Capella-X, demonstrating its superiority over traditional and deep learning approaches. Additionally, the application analysis confirms that the method enhances both the visual quality and analytical reliability of SAR images, making it a valuable preprocessing step for real-world scenarios. For reproducibility, our code and data are available at https://github.com/yangyang12318/LGCN.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"302 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-02-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146129286","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-03DOI: 10.1016/j.jag.2026.105140
Belen Franch, Italo Moletto-Lobos, Javier Tarín-Mestre, Lucio Mascolo, Eric Vermote, Natacha Kalecinski, Inbal Becker-Reshef, Alberto San-Bautista, Constanza Rubio, Sara San Francisco, Miguel Ángel Naranjo, Vanessa Paredes, David Nafria, Carlos Cantero-Martinez
Accurate and transferable crop monitoring from remote sensing remains challenging because vegetation signals are strongly affected by phenological asynchrony, climatic variability, and sensor-specific responses. Existing approaches rely on local calibrated relationships , limiting their effectiveness in data-sparse regions. This study investigates whether models calibrated on high-quality localized reference data can generalize to other regions by stabilizing sensor–biophysical relationships.
{"title":"The yield strikes back: Enhancing the transferability of field scale wheat and barley yield models by leveraging Sentinel-1/2","authors":"Belen Franch, Italo Moletto-Lobos, Javier Tarín-Mestre, Lucio Mascolo, Eric Vermote, Natacha Kalecinski, Inbal Becker-Reshef, Alberto San-Bautista, Constanza Rubio, Sara San Francisco, Miguel Ángel Naranjo, Vanessa Paredes, David Nafria, Carlos Cantero-Martinez","doi":"10.1016/j.jag.2026.105140","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105140","url":null,"abstract":"Accurate and transferable crop monitoring from remote sensing remains challenging because vegetation signals are strongly affected by phenological asynchrony, climatic variability, and sensor-specific responses. Existing approaches rely on local calibrated relationships , limiting their effectiveness in data-sparse regions. This study investigates whether models calibrated on high-quality localized reference data can generalize to other regions by stabilizing sensor–biophysical relationships.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"26 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-02-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146129287","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-02DOI: 10.1016/j.jag.2026.105112
Surya Gupta, Simon Scheper, Christine Alewell
Soil bulk density (BD) is a key indicator of soil health and quality, influencing air and water fluxes in the soil, soil biology, plant growth, nutrient availability, and water retention. While BD is typically measured through field and lab methods, these are time-consuming and resource-intensive. Alternatively, researchers use pedotransfer functions and machine learning algorithms for BD prediction. Although several BD maps exist for Europe, Switzerland is often excluded due to its non-European Union member status, creating a data gap known as a “blank spot”. Additionally, existing Swiss BD maps have coarse spatial resolution (∼250 m). To address this, we used the national Swiss Soil Information System NABODAT dataset to produce high-resolution (30 m) BD maps at multiple depths (0, 30, 60, 100 cm) using a Quantile Random Forest algorithm. Using five-fold cross-validation, we obtained a concordance correlation coefficient (CCC) of 0.57 and an R2 of 0.42, while external validation resulted in a CCC of 0.39 and an R2 of 0.36. The maps revealed that croplands had the highest BD, followed by grasslands and forests. Regionally, the Central Plateau and Jura exhibited higher BD compared to the Alps. BD increased with depth, and key predictors were depth, elevation, and temperature. Although we initially expected surface reflectance to be a relevant predictor due to its link with organic carbon, it showed low importance in our model. These maps provide valuable insights for national-scale applications such as soil carbon stock estimation and compaction assessment.
{"title":"Mapping Swiss soil bulk density at 30 m Resolution: Insights from Machine Learning, environmental Covariates, and national data","authors":"Surya Gupta, Simon Scheper, Christine Alewell","doi":"10.1016/j.jag.2026.105112","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105112","url":null,"abstract":"Soil bulk density (BD) is a key indicator of soil health and quality, influencing air and water fluxes in the soil, soil biology, plant growth, nutrient availability, and water retention. While BD is typically measured through field and lab methods, these are time-consuming and resource-intensive. Alternatively, researchers use pedotransfer functions and machine learning algorithms for BD prediction. Although several BD maps exist for Europe, Switzerland is often excluded due to its non-European Union member status, creating a data gap known as a “blank spot”. Additionally, existing Swiss BD maps have coarse spatial resolution (∼250 m). To address this, we used the national Swiss Soil Information System NABODAT dataset to produce high-resolution (30 m) BD maps at multiple depths (0, 30, 60, 100 cm) using a Quantile Random Forest algorithm. Using five-fold cross-validation, we obtained a concordance correlation coefficient (CCC) of 0.57 and an R<ce:sup loc=\"post\">2</ce:sup> of 0.42, while external validation resulted in a CCC of 0.39 and an R<ce:sup loc=\"post\">2</ce:sup> of 0.36. The maps revealed that croplands had the highest BD, followed by grasslands and forests. Regionally, the Central Plateau and Jura exhibited higher BD compared to the Alps. BD increased with depth, and key predictors were depth, elevation, and temperature. Although we initially expected surface reflectance to be a relevant predictor due to its link with organic carbon, it showed low importance in our model. These maps provide valuable insights for national-scale applications such as soil carbon stock estimation and compaction assessment.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"6 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146129289","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Monitoring global vegetation gross primary productivity (GPP) at an hourly scale is critical for understanding terrestrial carbon dynamics, while recent global GPP products often suffer from limitations in their temporal resolutions. Here, a light use efficiency (LUE) model, integrated with FLUXNET and reanalysis datasets as meteorological inputs, was employed to obtain GPP at both 1-hourly and 6-hourly resolutions. We developed a downscaling algorithm that partitions 6-hourly GPP into 1-hourly estimates by weighting the 6-hourly values according to the hourly cosine of the solar zenith angle and applying linear regression. The algorithm was then applied to global 6-hourly reanalysis-driven GPP maps during 2001–2020. Using GPP simulated from 1-hourly meteorological inputs and eddy covariance (EC) GPP as references, the 6-hourly resolution GPP before and after downscaling were evaluated by mean-absolute-deviation (MAD) and nash–sutcliffe-efficiency (NSE). At 150 sites, results showed that the 6-hourly FLUXNET-driven and reanalysis-driven GPP after downscaling exhibited a significantly stronger correlation (MAD = 0.03 gCm−2h−1, NSE = 0.95) with corresponding 1-hourly estimates than the 6-hourly GPP estimates before downscaling (MAD = 0.06 gCm−2h−1, NSE = 0.83). Compared to EC GPP, the 6-hourly GPP estimates after downscaling also achieved notable improvements, with a MAD lower by 0.02 gCm−2h−1 and an NSE higher by 0.09. At the global scale, the mean annual bias in total GPP summed from 6-hourly reanalysis-driven maps, decreased from 4.14 gCyr−1 before downscaling to 0.53 gCyr−1 after downscaling over the period 2001–2020, as compared with corresponding 1-hourly GPP maps. At the hourly scale, the mean relative error between the 6-hourly and corresponding 1-hourly GPP maps decreased from 32.40% before downscaling to 18.49% after downscaling. This downscaling algorithm effectively reduces biases in global GPP estimates, which could offer valuable insights into carbon modeling at finer temporal resolutions.
{"title":"A downscaling algorithm for obtaining hourly gross primary productivity maps at the global scale","authors":"Yong Wang, Jiyan Wang, Wei Zhao, Yanqing Yang, Jiujiang Wu, Xiaobin Guan, Xinyao Xie","doi":"10.1016/j.jag.2025.105059","DOIUrl":"https://doi.org/10.1016/j.jag.2025.105059","url":null,"abstract":"Monitoring global vegetation gross primary productivity (GPP) at an hourly scale is critical for understanding terrestrial carbon dynamics, while recent global GPP products often suffer from limitations in their temporal resolutions. Here, a light use efficiency (LUE) model, integrated with FLUXNET and reanalysis datasets as meteorological inputs, was employed to obtain GPP at both 1-hourly and 6-hourly resolutions. We developed a downscaling algorithm that partitions 6-hourly GPP into 1-hourly estimates by weighting the 6-hourly values according to the hourly cosine of the solar zenith angle and applying linear regression. The algorithm was then applied to global 6-hourly reanalysis-driven GPP maps during 2001–2020. Using GPP simulated from 1-hourly meteorological inputs and eddy covariance (EC) GPP as references, the 6-hourly resolution GPP before and after downscaling were evaluated by mean-absolute-deviation (<ce:italic>MAD</ce:italic>) and nash–sutcliffe-efficiency (<ce:italic>NSE</ce:italic>). At 150 sites, results showed that the 6-hourly FLUXNET-driven and reanalysis-driven GPP after downscaling exhibited a significantly stronger correlation (<ce:italic>MAD</ce:italic> = 0.03 gCm<ce:sup loc=\"post\">−2</ce:sup>h<ce:sup loc=\"post\">−1</ce:sup>, <ce:italic>NSE</ce:italic> = 0.95) with corresponding 1-hourly estimates than the 6-hourly GPP estimates before downscaling (<ce:italic>MAD</ce:italic> = 0.06 gCm<ce:sup loc=\"post\">−2</ce:sup>h<ce:sup loc=\"post\">−1</ce:sup>, <ce:italic>NSE</ce:italic> = 0.83). Compared to EC GPP, the 6-hourly GPP estimates after downscaling also achieved notable improvements, with a <ce:italic>MAD</ce:italic> lower by 0.02 gCm<ce:sup loc=\"post\">−2</ce:sup>h<ce:sup loc=\"post\">−1</ce:sup> and an <ce:italic>NSE</ce:italic> higher by 0.09. At the global scale, the mean annual bias in total GPP summed from 6-hourly reanalysis-driven maps, decreased from 4.14 gCyr<ce:sup loc=\"post\">−1</ce:sup> before downscaling to 0.53 gCyr<ce:sup loc=\"post\">−1</ce:sup> after downscaling over the period 2001–2020, as compared with corresponding 1-hourly GPP maps. At the hourly scale, the mean relative error between the 6-hourly and corresponding 1-hourly GPP maps decreased from 32.40% before downscaling to 18.49% after downscaling. This downscaling algorithm effectively reduces biases in global GPP estimates, which could offer valuable insights into carbon modeling at finer temporal resolutions.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"38 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146129290","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-02DOI: 10.1016/j.jag.2026.105136
Lijing Lu, Zhou Huang, Yi Bao, Lin Wan, Zhihang Li
Real-world aerial image super-resolution (SR) remains particularly challenging because degradations in remote-sensing imagery involve random combinations of anisotropic blur, signal-dependent noise, and unknown downsampling kernels. Most existing SR methods either rely on simplified degradation assumptions or lack semantic perception of degradation, resulting in limited generalization to real-world conditions. To address these gaps, we propose a novel diffusion-based SR framework that integrates Multi-modal Large Language Models (MLLMs) and self-supervised contrastive learning for extracting degradation-insensitive representation. Specifically, we introduce a contrastive learning strategy into a ControlNet module, where the HR and LR counterparts of the same image are regarded as positive pairs, while representations from different images serve as negative pairs, enabling the network to learn degradation-insensitive structural features. To further enhance semantic awareness of degradation, an MLLM-generated change caption is incorporated into the diffusion process as textual guidance, allowing the model to explicitly perceive and reconstruct different degradation types. Moreover, a classifier-free guidance (CFG) distillation strategy compresses the original dual-branch diffusion model into a single lightweight network, substantially improving inference efficiency while maintaining high reconstruction fidelity. Extensive experiments conducted on various datasets have showcased the superior performance of our proposed model compared to existing state-of-the-art methods. Furthermore, our distillation algorithm achieves a twofold reduction in inference time compared to its non-distilled counterpart, making it more feasible for real-time and resource-limited applications.
{"title":"Multimodal large language models meet self-supervised diffusion for real-world aerial image super-resolution","authors":"Lijing Lu, Zhou Huang, Yi Bao, Lin Wan, Zhihang Li","doi":"10.1016/j.jag.2026.105136","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105136","url":null,"abstract":"Real-world aerial image super-resolution (SR) remains particularly challenging because degradations in remote-sensing imagery involve random combinations of anisotropic blur, signal-dependent noise, and unknown downsampling kernels. Most existing SR methods either rely on simplified degradation assumptions or lack semantic perception of degradation, resulting in limited generalization to real-world conditions. To address these gaps, we propose a novel diffusion-based SR framework that integrates Multi-modal Large Language Models (MLLMs) and self-supervised contrastive learning for extracting degradation-insensitive representation. Specifically, we introduce a contrastive learning strategy into a ControlNet module, where the HR and LR counterparts of the same image are regarded as positive pairs, while representations from different images serve as negative pairs, enabling the network to learn degradation-insensitive structural features. To further enhance semantic awareness of degradation, an MLLM-generated change caption is incorporated into the diffusion process as textual guidance, allowing the model to explicitly perceive and reconstruct different degradation types. Moreover, a classifier-free guidance (CFG) distillation strategy compresses the original dual-branch diffusion model into a single lightweight network, substantially improving inference efficiency while maintaining high reconstruction fidelity. Extensive experiments conducted on various datasets have showcased the superior performance of our proposed model compared to existing state-of-the-art methods. Furthermore, our distillation algorithm achieves a twofold reduction in inference time compared to its non-distilled counterpart, making it more feasible for real-time and resource-limited applications.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"47 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146129288","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01DOI: 10.1016/j.jag.2026.105139
Jin Zhang, Chong Huang, He Li, Shuxuan Wang, Qingsheng Liu, Xiaoya Tang, Chenchen Zhang, Fenzhen Su
Accurate mapping of salt pond systems is crucial for coastal resource management and ecological conservation. However, existing spectral indices for salt pond detection are mostly derived empirically, limiting their ability to differentiate pond subtypes or generalize across diverse coastal environments. This study proposes a big data-driven framework for constructing novel spectral indices that enhance the fine-scale classification of salt ponds in the Bohai Rim region of China. By systematically generating and screening over 390,000 candidate indices derived from Sentinel-2 bands, four new indices were identified to effectively discriminate salt pond subtypes: the Non-Water Body Detection Index (NWBDI), Crystallization Pond Index (CPI), Red-Green Concentration Pond Index (RGCPI), and Evaporation Pond Discrimination Index (EPDI). To rigorously evaluate their performance, three Random Forest configurations were compared, including a baseline model with original spectral bands, an expert-index-enhanced model, and a model incorporating the proposed novel indices. Across the Bohai Rim region, the model with proposed new indices achieved an overall accuracy (OA) of 81.14%, demonstrating clear advantages over both original spectral bands (OA of 72.93%) and expert-index approaches (OA of 76.64%). The resulting fine-scale map revealed a total salt pond area of 2,214.33 km2 in 2020 across the Bohai Rim, consistent with national statistics. Beyond improving classification accuracy, the data-driven index discovery revealed physically meaningful spectral relationships linked to salinity gradients and hydro-biogeochemical properties, marking a methodological shift from expert-driven hypothesis testing to automated, data-driven feature discovery. This study demonstrates that the data-driven approach can provide a transferable solution for constructing task-specific spectral indices and advancing large-scale environmental monitoring.
{"title":"Big data-driven spectral index construction for fine-scale salt pond mapping","authors":"Jin Zhang, Chong Huang, He Li, Shuxuan Wang, Qingsheng Liu, Xiaoya Tang, Chenchen Zhang, Fenzhen Su","doi":"10.1016/j.jag.2026.105139","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105139","url":null,"abstract":"Accurate mapping of salt pond systems is crucial for coastal resource management and ecological conservation. However, existing spectral indices for salt pond detection are mostly derived empirically, limiting their ability to differentiate pond subtypes or generalize across diverse coastal environments. This study proposes a big data-driven framework for constructing novel spectral indices that enhance the fine-scale classification of salt ponds in the Bohai Rim region of China. By systematically generating and screening over 390,000 candidate indices derived from Sentinel-2 bands, four new indices were identified to effectively discriminate salt pond subtypes: the Non-Water Body Detection Index (NWBDI), Crystallization Pond Index (CPI), Red-Green Concentration Pond Index (RGCPI), and Evaporation Pond Discrimination Index (EPDI). To rigorously evaluate their performance, three Random Forest configurations were compared, including a baseline model with original spectral bands, an expert-index-enhanced model, and a model incorporating the proposed novel indices. Across the Bohai Rim region, the model with proposed new indices achieved an overall accuracy (OA) of 81.14%, demonstrating clear advantages over both original spectral bands (OA of 72.93%) and expert-index approaches (OA of 76.64%). The resulting fine-scale map revealed a total salt pond area of 2,214.33 km<ce:sup loc=\"post\">2</ce:sup> in 2020 across the Bohai Rim, consistent with national statistics. Beyond improving classification accuracy, the data-driven index discovery revealed physically meaningful spectral relationships linked to salinity gradients and hydro-biogeochemical properties, marking a methodological shift from expert-driven hypothesis testing to automated, data-driven feature discovery. This study demonstrates that the data-driven approach can provide a transferable solution for constructing task-specific spectral indices and advancing large-scale environmental monitoring.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"18 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146129291","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01DOI: 10.1016/j.jag.2026.105133
Luhan Wang, Pengfeng Xiao, Xueliang Zhang, Yina Song, Lei Guo, Di Wang, Yao Zhang
Detecting solid waste dumps (SWDs) is crucial for a sustainable environment and public health. Deep learning methods have demonstrated significant potential in detecting SWDs from high-spatial resolution remote sensing images (RSIs). However, training a SWD detection model necessitates annotated data that encompasses diverse geographical distributions and styles. Following the conventional approach of manually labeling data prior to model training is both costly and time-consuming. In this study, we propose a human-in-the-loop framework for the detection of SWDs. An initial model is trained using public datasets of Dumpsites, solid waste aerial detection (SWAD), and AerialWaste to identify potential samples from extensive unlabeled RSIs, which are then validated by experts for data expansion and model reinforcement. Specifically, multi-view inference is introduced to enhance the applicability of the model for real-world SWD detection tasks by integrating inference results from multiple style-transformed images. Moreover, we utilize adaptive thresholds that are dynamically calculated from inference results in each round to select potential SWDs, all while maintaining a low computational cost. With the proposed framework, we construct the LHRS-SWD dataset for SWD detection, derived from a random sampling of over 70 countries and encompassing 3,377 SWDs. The effectiveness of our framework is validated through experiments on LHRS-SWD, with a 20.26% improvement in AP50 versus initial iteration.
{"title":"Human-in-the-loop based framework for solid waste dumps detection in remote sensing images","authors":"Luhan Wang, Pengfeng Xiao, Xueliang Zhang, Yina Song, Lei Guo, Di Wang, Yao Zhang","doi":"10.1016/j.jag.2026.105133","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105133","url":null,"abstract":"Detecting solid waste dumps (SWDs) is crucial for a sustainable environment and public health. Deep learning methods have demonstrated significant potential in detecting SWDs from high-spatial resolution remote sensing images (RSIs). However, training a SWD detection model necessitates annotated data that encompasses diverse geographical distributions and styles. Following the conventional approach of manually labeling data prior to model training is both costly and time-consuming. In this study, we propose a human-in-the-loop framework for the detection of SWDs. An initial model is trained using public datasets of Dumpsites, solid waste aerial detection (SWAD), and AerialWaste to identify potential samples from extensive unlabeled RSIs, which are then validated by experts for data expansion and model reinforcement. Specifically, multi-view inference is introduced to enhance the applicability of the model for real-world SWD detection tasks by integrating inference results from multiple style-transformed images. Moreover, we utilize adaptive thresholds that are dynamically calculated from inference results in each round to select potential SWDs, all while maintaining a low computational cost. With the proposed framework, we construct the LHRS-SWD dataset for SWD detection, derived from a random sampling of over 70 countries and encompassing 3,377 SWDs. The effectiveness of our framework is validated through experiments on LHRS-SWD, with a 20.26% improvement in AP50 versus initial iteration.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"89 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146129292","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Accurate Surface Soil Moisture (SSM) with high spatiotemporal resolution is essential for hydrological modeling and agricultural applications. The Cyclone Global Navigation Satellite System (CYGNSS) provides quasi-global SSM with relatively high temporal resolution, while its coarse spatial resolution limits regional usability. This study proposed a novel approach to generate daily continuous quasi-global SSM at 500-m resolution by integrating CYGNSS observations with optical imagery, without relying on in-situ SSM inputs. The method includes: (1) reconstruction of daily Surface Reflectivity (SR) using the Previously-Observed Behavior Interpolation (POBI) algorithm and retrieval of 9-km SSM via a modified Reflectivity-Vegetation-Roughness (R-V-R) model; (2) constructing a direct relationship between SSM and optical reflectance using the enhanced OPtical TRApezoid Model (OPTRAM) as the foundational basis for subsequent downscaling process; (3) downscale coarse-resolution CYGNSS SSM using the Bayesian algorithm without requiring any in-situ SSM observations as inputs. The framework is applied over August 2019–2022 and validated against in-situ SSM from 194 International Soil Moisture Network (ISMN) sites across diverse climate and land cover conditions. Results show improved spatial detail while maintaining accuracy, with a mean unbiased root-mean-square error (ubRMSE) of 0.061 cm3/cm3 and a positive GDOWN value of 0.017, indicating effective SSM downscaling. In addition, the proposed method outperforms both Linear and Random Forest models while maintaining robust performance. Overall, it offers a scalable solution for generating high-resolution, daily SSM products directly from satellite data.
精确的高时空分辨率地表土壤水分（SSM）是水文建模和农业应用的基础。气旋全球导航卫星系统（CYGNSS）提供了具有较高时间分辨率的准全球SSM，但其粗糙的空间分辨率限制了区域可用性。本研究提出了一种新的方法，通过将CYGNSS观测与光学图像相结合，在不依赖于原位SSM输入的情况下，生成500米分辨率的每日连续准全球SSM。该方法包括：(1)使用先前观测行为插值（POBI）算法重建日地表反射率（SR），并通过改进的反射率-植被-粗糙度（R-V-R）模型检索9 km SSM；(2)利用增强型光学梯形模型（OPTRAM）建立SSM与反射率之间的直接关系，为后续降尺度处理奠定基础；(3)采用贝叶斯算法，在不需要任何原位SSM观测作为输入的情况下，实现CYGNSS SSM的降尺度粗分辨率。该框架将于2019年8月至2022年8月期间应用，并根据不同气候和土地覆盖条件下194个国际土壤湿度网络（ISMN）站点的原位SSM进行验证。结果表明，在保持精度的同时，空间细节得到了改善，平均无偏均方根误差（ubRMSE）为0.061 cm3/cm3， GDOWN为0.017，表明SSM降尺度有效。此外，该方法在保持鲁棒性的同时，优于线性和随机森林模型。总的来说，它提供了一个可扩展的解决方案，可以直接从卫星数据生成高分辨率的每日SSM产品。
{"title":"A novel approach for Quasi-Global daily continuous surface soil moisture downscaling at 500-m resolution using CYGNSS observations","authors":"Jundong Wang, Ting Yang, Wanxue Zhu, Shiji Li, Zixuan Tang, Wei Wan, Zhigang Sun","doi":"10.1016/j.jag.2026.105137","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105137","url":null,"abstract":"Accurate Surface Soil Moisture (SSM) with high spatiotemporal resolution is essential for hydrological modeling and agricultural applications. The Cyclone Global Navigation Satellite System (CYGNSS) provides quasi-global SSM with relatively high temporal resolution, while its coarse spatial resolution limits regional usability. This study proposed a novel approach to generate daily continuous quasi-global SSM at 500-m resolution by integrating CYGNSS observations with optical imagery, without relying on in-situ SSM inputs. The method includes: (1) reconstruction of daily Surface Reflectivity (SR) using the Previously-Observed Behavior Interpolation (POBI) algorithm and retrieval of 9-km SSM via a modified Reflectivity-Vegetation-Roughness (R-V-R) model; (2) constructing a direct relationship between SSM and optical reflectance using the enhanced OPtical TRApezoid Model (OPTRAM) as the foundational basis for subsequent downscaling process; (3) downscale coarse-resolution CYGNSS SSM using the Bayesian algorithm without requiring any in-situ SSM observations as inputs. The framework is applied over August 2019–2022 and validated against in-situ SSM from 194 International Soil Moisture Network (ISMN) sites across diverse climate and land cover conditions. Results show improved spatial detail while maintaining accuracy, with a mean unbiased root-mean-square error (ubRMSE) of 0.061 cm<ce:sup loc=\"post\">3</ce:sup>/cm<ce:sup loc=\"post\">3</ce:sup> and a positive G<ce:inf loc=\"post\">DOWN</ce:inf> value of 0.017, indicating effective SSM downscaling. In addition, the proposed method outperforms both Linear and Random Forest models while maintaining robust performance. Overall, it offers a scalable solution for generating high-resolution, daily SSM products directly from satellite data.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"31 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146129294","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-30DOI: 10.1016/j.jag.2026.105127
Zhongzhen Sun, Xianghui Zhang, Xiangguang Leng, Xueqi Wu, Boli Xiong, Kefeng Ji, Gangyao Kuang
Multi-category synthetic aperture radar (SAR) ship detection is limited by heterogeneity in imaging mechanisms and severe class imbalance, yielding accurate localization but frequent misclassification. To address this issue, this paper proposes a Knowledge Fusion and Imbalance-Aware Network (KFIA-Net). Specifically, we first propose a Domain Knowledge Feature Extraction (DKFE) to extract and encode knowledge tokens from four priors. Second, a Knowledge Cross-Attention Fusion (KCAF) module is designed to perform interpretable and sparsely selectable channel modulation using cross-attention and FiLM decoding. Thirdly, we further design an Imbalance-Aware Loss Function (IALF) that combines prior calibration, minority category margin expansion, and knowledge-consistency weighting to reduce loss bias. Finally, systematic experiments and comparisons are conducted on three datasets: SRSDD-v1.0, FAIR-CSAR-v1.0, and NUDT-SARship-v1.0. Our KFIA-Net achieves mAP50 scores of 64.29%, 37.99%, and 78.26%, and mAP75 scores of 34.96%, 19.70%, and 66.36%, respectively. These results demonstrate knowledge injection simultaneously improves class accuracy and sustains robust localization. Furthermore, KFIA-Net requires only 11.47 M parameters and 66.79G FLOPs, achieving an inference speed of 47.21 FPS on a 1024 × 1024 input, achieving a good trade-off between accuracy and efficiency.
{"title":"KFIA-Net: a knowledge fusion and imbalance-aware network for multi-category SAR ship detection","authors":"Zhongzhen Sun, Xianghui Zhang, Xiangguang Leng, Xueqi Wu, Boli Xiong, Kefeng Ji, Gangyao Kuang","doi":"10.1016/j.jag.2026.105127","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105127","url":null,"abstract":"Multi-category synthetic aperture radar (SAR) ship detection is limited by heterogeneity in imaging mechanisms and severe class imbalance, yielding accurate localization but frequent misclassification. To address this issue, this paper proposes a Knowledge Fusion and Imbalance-Aware Network (KFIA-Net). Specifically, we first propose a Domain Knowledge Feature Extraction (DKFE) to extract and encode knowledge tokens from four priors. Second, a Knowledge Cross-Attention Fusion (KCAF) module is designed to perform interpretable and sparsely selectable channel modulation using cross-attention and FiLM decoding. Thirdly, we further design an Imbalance-Aware Loss Function (IALF) that combines prior calibration, minority category margin expansion, and knowledge-consistency weighting to reduce loss bias. Finally, systematic experiments and comparisons are conducted on three datasets: SRSDD-v1.0, FAIR-CSAR-v1.0, and NUDT-SARship-v1.0. Our KFIA-Net achieves mAP<ce:inf loc=\"post\">50</ce:inf> scores of 64.29%, 37.99%, and 78.26%, and mAP<ce:inf loc=\"post\">75</ce:inf> scores of 34.96%, 19.70%, and 66.36%, respectively. These results demonstrate knowledge injection simultaneously improves class accuracy and sustains robust localization. Furthermore, KFIA-Net requires only 11.47 M parameters and 66.79G FLOPs, achieving an inference speed of 47.21 FPS on a 1024 × 1024 input, achieving a good trade-off between accuracy and efficiency.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"236 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146129296","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Land subsidence is a growing geohazard that poses a significant threat to critical infrastructure, particularly in urban coastal cities. This study uses Interferometric Synthetic Aperture Radar (InSAR) data from 2016 to 2024 to estimate angular distortion rates to assess infrastructure damage risk in New York City. We applied a probabilistic framework to evaluate multiple “what-if” scenarios and project long-term risks, providing actionable insights for resilience and mitigation planning. Results reveal persistent subsidence in low-elevation and reclaimed zones (∼-5 mm/yr) with localized uplift (∼+1.5 mm/yr), affecting major airports, subway segments, and highways. Fifty-year projections indicate high angular distortion probabilities (0.6–0.8), with economic exposure estimated at ∼$8.20 billion for ∼ 6.1 km of subway lines and ∼$10.54 billion for ∼ 7.8 km of highways exceeding –2 mm/yr. Despite their limited spatial extent, these segments represent a disproportionately large share of total exposure. The findings emphasize the need for continuous monitoring, proactive mitigation, and targeted investment, highlighting the value of integrating geodetic data with probabilistic modeling to address subsidence and climate-related hazards.
{"title":"Probabilistic modeling of InSAR-derived land subsidence hazard in New York City for transportation infrastructure damage risk assessments","authors":"Ntambila Daud, Oluwaseyi Dasho, Manoochehr Shirzaei","doi":"10.1016/j.jag.2026.105118","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105118","url":null,"abstract":"Land subsidence is a growing geohazard that poses a significant threat to critical infrastructure, particularly in urban coastal cities. This study uses Interferometric Synthetic Aperture Radar (InSAR) data from 2016 to 2024 to estimate angular distortion rates to assess infrastructure damage risk in New York City. We applied a probabilistic framework to evaluate multiple “what-if” scenarios and project long-term risks, providing actionable insights for resilience and mitigation planning. Results reveal persistent subsidence in low-elevation and reclaimed zones (∼-5 mm/yr) with localized uplift (∼+1.5 mm/yr), affecting major airports, subway segments, and highways. Fifty-year projections indicate high angular distortion probabilities (0.6–0.8), with economic exposure estimated at ∼$8.20 billion for ∼ 6.1 km of subway lines and ∼$10.54 billion for ∼ 7.8 km of highways exceeding –2 mm/yr. Despite their limited spatial extent, these segments represent a disproportionately large share of total exposure. The findings emphasize the need for continuous monitoring, proactive mitigation, and targeted investment, highlighting the value of integrating geodetic data with probabilistic modeling to address subsidence and climate-related hazards.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"26 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146129299","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：