Pub Date : 2026-01-24DOI: 10.1016/j.jag.2026.105120
Huafei Yu, Min Yang, Xiang Lv, Tinghua Ai, Jingzhong Li
River network selection is essential for generating multiscale river network datasets used in earth observation (EO) applications, such as hydrological analysis and terrain modeling. However, how individual river features influence selection decisions, especially across different drainage patterns, remains unclear. To address this gap, this study introduces an explainable artificial intelligence framework that integrates eXtreme Gradient Boosting (XGBoost) with SHapley Additive exPlanations (SHAP) to uncover feature contributions for river network selection across different patterns. First, river strokes were constructed as analytical units and multiple geometric, topological, and hydrological features were extracted. Then, XGBoost-based selection models were trained for the parallel and rectangular drainage networks. Finally, the macro-, meso-, and micro-level SHAP analysis was conducted over the trained selection models. The experimental results reveal that features fall into three categories for the two types of patterns considered here: universal features (represented by upstream cumulative area, Horton code) that dominate selection regardless of pattern; pattern-sensitive features (represented by confluence angle, river proximity distance) whose influence varies with drainage patterns; and low-contribution features with negligible contribution. These findings explain why certain rivers are retained or removed across different drainage patterns, providing explainable insights to support automated, pattern-preserving generation of multiscale river networks.
{"title":"Integrating XGBoost and SHAP to uncover feature contributions for river network selection across different patterns","authors":"Huafei Yu, Min Yang, Xiang Lv, Tinghua Ai, Jingzhong Li","doi":"10.1016/j.jag.2026.105120","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105120","url":null,"abstract":"River network selection is essential for generating multiscale river network datasets used in earth observation (EO) applications, such as hydrological analysis and terrain modeling. However, how individual river features influence selection decisions, especially across different drainage patterns, remains unclear. To address this gap, this study introduces an explainable artificial intelligence framework that integrates eXtreme Gradient Boosting (XGBoost) with SHapley Additive exPlanations (SHAP) to uncover feature contributions for river network selection across different patterns. First, river strokes were constructed as analytical units and multiple geometric, topological, and hydrological features were extracted. Then, XGBoost-based selection models were trained for the parallel and rectangular drainage networks. Finally, the macro-, meso-, and micro-level SHAP analysis was conducted over the trained selection models. The experimental results reveal that features fall into three categories for the two types of patterns considered here: universal features (represented by upstream cumulative area, Horton code) that dominate selection regardless of pattern; pattern-sensitive features (represented by confluence angle, river proximity distance) whose influence varies with drainage patterns; and low-contribution features with negligible contribution. These findings explain why certain rivers are retained or removed across different drainage patterns, providing explainable insights to support automated, pattern-preserving generation of multiscale river networks.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"288 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146047825","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 : 2026-01-24DOI: 10.1016/j.jag.2026.105107
Adam M. Morley, Tamsin A. Mather, David M. Pyle, J-Michael Kendall
The remote detection of shallow, subsurface features can assist archaeological prospecting, environmental risk mitigation and national security. However, few subsurface detection studies have used geophysical methods to statistically validate the detection efficiency of broadband vegetation indices (VIs) in high-resolution multispectral satellite imagery. In this study, we use 37 broadband VIs on eight Maxar (now rebranded to Vantor) multispectral satellite images to remotely detect stressed meadow grass and anomalous soil characteristics over an Iron Age fogou (stone-walled underground passage) in Carn Euny, Cornwall, UK. Using Maxar’s high spatial resolution, the highest performing VIs are identified using a two-tier geophysical approach. First, we correlate the VIs with gravity and then, for a best performing image subset, we perform structural similarity index measurements (SSIMs) and 2D cross-correlations with ground penetrating radar (GPR) data. In doing so, we reprioritise the most suitable VIs for shallow, subsurface detection in temperate, grass covered environments. In summer months, the Iron Oxide index, Soil Salinity Index 7 (SI7) and the Structure Insensitive Pigment Index (SIPI) are most responsive across the fogou; all of which are algebraic manifestations of the Red/Blue reflectance ratio. By analysing their spectral profiles, gradient magnitudes, false colour composites (FCCs) and edge effects, we review the fogou’s effect on soil salinity, iron oxide concentration and chlorophyll production. Demonstrating the broader utility of Red/Blue reflectance ratios, we then present a multispectral image processing workflow which is calibrated to detect, at scale, shallow subsurface features in temperate, vegetated terrain.
{"title":"Geophysical validation of vegetation indices for subsurface detection: Evidence for the utility of red/blue reflectance ratios","authors":"Adam M. Morley, Tamsin A. Mather, David M. Pyle, J-Michael Kendall","doi":"10.1016/j.jag.2026.105107","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105107","url":null,"abstract":"The remote detection of shallow, subsurface features can assist archaeological prospecting, environmental risk mitigation and national security. However, few subsurface detection studies have used geophysical methods to statistically validate the detection efficiency of broadband vegetation indices (VIs) in high-resolution multispectral satellite imagery. In this study, we use 37 broadband VIs on eight Maxar (now rebranded to Vantor) multispectral satellite images to remotely detect stressed meadow grass and anomalous soil characteristics over an Iron Age fogou (stone-walled underground passage) in Carn Euny, Cornwall, UK. Using Maxar’s high spatial resolution, the highest performing VIs are identified using a two-tier geophysical approach. First, we correlate the VIs with gravity and then, for a best performing image subset, we perform structural similarity index measurements (SSIMs) and 2D cross-correlations with ground penetrating radar (GPR) data. In doing so, we reprioritise the most suitable VIs for shallow, subsurface detection in temperate, grass covered environments. In summer months, the Iron Oxide index, Soil Salinity Index 7 (SI7) and the Structure Insensitive Pigment Index (SIPI) are most responsive across the fogou; all of which are algebraic manifestations of the Red/Blue reflectance ratio. By analysing their spectral profiles, gradient magnitudes, false colour composites (FCCs) and edge effects, we review the fogou’s effect on soil salinity, iron oxide concentration and chlorophyll production. Demonstrating the broader utility of Red/Blue reflectance ratios, we then present a multispectral image processing workflow which is calibrated to detect, at scale, shallow subsurface features in temperate, vegetated terrain.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"178 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146047823","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}
Forests exhibit complex vertical structures, and understanding the vertical distribution of gross primary production (GPP) is critical for improving carbon cycle assessments. While previous studies have emphasized the contribution of understory, approaches such as eddy-covariance provide only site-level observations and are constrained by sparse spatial coverage and footprint limitations, leaving large-scale assessment and mapping unresolved. This study identified key parameters of GPP vertical heterogeneity and developed a remote sensing framework to quantify and map its distribution in forests. We integrated hyperspectral imagery and light detection and ranging (LiDAR) point clouds to retrieve physiological and biochemical parameters as well as the vertical distribution of canopy structure, which were used to drive GPP modeling. Based on the outputs of the SCOPE (Soil Canopy Observation of Photosynthesis and Energy Fluxes) model, we visualized the vertical profile of GPP and quantified the overstory and understory. Sensitivity and mechanistic analyses were performed using simulated data to investigate the effects of vegetation biophysical and biochemical parameters on vertical GPP. We further estimated and validated the vertical distribution of GPP across seven National Ecological Observatory Network (NEON) forests. The results showed that LAI (leaf area index) was the dominant driver of GPP in canopy layers (correlation coefficients = 0.74), with a stronger influence in the overstory than in the understory, and a trade-off relationship observed between layers. Vertical GPP heterogeneity was well captured, with understory contributions ranging from 5.9% to 35.8%, except in sparse-canopy subarctic forests. Model estimates agreed well with flux tower data (understory contribution: correlation coefficient = 0.941, R2 = 0.885; total GPP: R2 = 0.785). This study offers new insights into the role of understory vegetation in carbon cycling and informs vertical-dimension strategies for forest carbon management.
{"title":"Quantifying vertical heterogeneity of forest gross primary production using hyperspectral and LiDAR data","authors":"Zixi Shi, Shuo Shi, Peiqi Yang, Wei Gong, Chenxi Liu, Binhui Wang, Jiayun Niu, Minghui Wu","doi":"10.1016/j.jag.2026.105094","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105094","url":null,"abstract":"Forests exhibit complex vertical structures, and understanding the vertical distribution of gross primary production (GPP) is critical for improving carbon cycle assessments. While previous studies have emphasized the contribution of understory, approaches such as eddy-covariance provide only site-level observations and are constrained by sparse spatial coverage and footprint limitations, leaving large-scale assessment and mapping unresolved. This study identified key parameters of GPP vertical heterogeneity and developed a remote sensing framework to quantify and map its distribution in forests. We integrated hyperspectral imagery and light detection and ranging (LiDAR) point clouds to retrieve physiological and biochemical parameters as well as the vertical distribution of canopy structure, which were used to drive GPP modeling. Based on the outputs of the SCOPE (Soil Canopy Observation of Photosynthesis and Energy Fluxes) model, we visualized the vertical profile of GPP and quantified the overstory and understory. Sensitivity and mechanistic analyses were performed using simulated data to investigate the effects of vegetation biophysical and biochemical parameters on vertical GPP. We further estimated and validated the vertical distribution of GPP across seven National Ecological Observatory Network (NEON) forests. The results showed that LAI (leaf area index) was the dominant driver of GPP in canopy layers (correlation coefficients = 0.74), with a stronger influence in the overstory than in the understory, and a trade-off relationship observed between layers. Vertical GPP heterogeneity was well captured, with understory contributions ranging from 5.9% to 35.8%, except in sparse-canopy subarctic forests. Model estimates agreed well with flux tower data (understory contribution: correlation coefficient = 0.941, R<ce:sup loc=\"post\">2</ce:sup> = 0.885; total GPP: R<ce:sup loc=\"post\">2</ce:sup> = 0.785). This study offers new insights into the role of understory vegetation in carbon cycling and informs vertical-dimension strategies for forest carbon management.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"5 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146047826","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 : 2026-01-24DOI: 10.1016/j.jag.2026.105089
Peng Wang, Hongwei Deng, Jielin Li, Zhen Jiang, Guanglin Tian, Yao Liu, Zheng Pan
Unstable slopes are highly susceptible to external triggers, representing a critical challenge for disaster risk management. However, few studies have systematically developed approaches for their regional-scale identification, and even fewer have explored their controlling factors in mining-affected regions with strong anthropogenic disturbances. To address this gap, this study investigates the composite factors influencing unstable slopes in the Datong Coalfield, China. Specifically, we used the Small Baseline Subset (SBAS) interferometric synthetic aperture radar (InSAR) technique to measure line-of-sight (LOS) surface deformation and subsequently identified unstable slopes as those with significant deformation velocities within a slope-unit framework. To examine the spatiotemporal effects of factors influencing slope instability, we conducted multivariate modeling at both long- and short-term scales. The results reveal that long-term modeling captures broadly consistent contributions of these factors across different periods, whereas short-term modeling highlights distinct seasonal associations with slope instability. Furthermore, cross-wavelet coherence analysis revealed that intermittently unstable slopes exhibit periodic lagged responses to dynamic environmental forcing, underscoring the significant role of environmental forcing in driving deformation of intermittently unstable slopes. Overall, this work provides new insights into the mechanisms of slope instability in mining-affected coalfields and proposes a transferable framework for regional-scale slope monitoring and risk management.
{"title":"Unraveling the controls of unstable slopes in mining-affected coalfields through InSAR observations and multivariate modeling","authors":"Peng Wang, Hongwei Deng, Jielin Li, Zhen Jiang, Guanglin Tian, Yao Liu, Zheng Pan","doi":"10.1016/j.jag.2026.105089","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105089","url":null,"abstract":"Unstable slopes are highly susceptible to external triggers, representing a critical challenge for disaster risk management. However, few studies have systematically developed approaches for their regional-scale identification, and even fewer have explored their controlling factors in mining-affected regions with strong anthropogenic disturbances. To address this gap, this study investigates the composite factors influencing unstable slopes in the Datong Coalfield, China. Specifically, we used the Small Baseline Subset (SBAS) interferometric synthetic aperture radar (InSAR) technique to measure line-of-sight (LOS) surface deformation and subsequently identified unstable slopes as those with significant deformation velocities within a slope-unit framework. To examine the spatiotemporal effects of factors influencing slope instability, we conducted multivariate modeling at both long- and short-term scales. The results reveal that long-term modeling captures broadly consistent contributions of these factors across different periods, whereas short-term modeling highlights distinct seasonal associations with slope instability. Furthermore, cross-wavelet coherence analysis revealed that intermittently unstable slopes exhibit periodic lagged responses to dynamic environmental forcing, underscoring the significant role of environmental forcing in driving deformation of intermittently unstable slopes. Overall, this work provides new insights into the mechanisms of slope instability in mining-affected coalfields and proposes a transferable framework for regional-scale slope monitoring and risk management.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"33 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146047868","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 : 2026-01-23DOI: 10.1016/j.jag.2026.105122
Teng Wang, Yunjia Wang, Feng Zhao, Sen Du, Nianbin Zhang, Kewei Zhang, Zhongwei Shen, José Fernández
Underground resource extraction redistributes subsurface stress, inducing overburden movement and ground deformation. These geomechanical changes can cause ground fissures, slope failures, and collapses, resulting in resource loss, economic damage, safety risks, and environmental degradation. Therefore, accurate monitoring of surface deformation and understanding of deformation propagation mechanisms are crucial for sustainable resource extraction and geohazard mitigation. To this end, a fusion framework is proposed that integrates a three-dimensional physical model with Interferometric Synthetic Aperture Radar (InSAR) technology to monitor surface deformation and investigate deformation propagation. A mining area in China was selected as a case study. Sentinel-1A images were processed using Stacking-InSAR and Small Baseline Subset InSAR (SBAS-InSAR) techniques to derive surface deformation. InSAR-derived deformation was validated against GNSS measurements, yielding mean absolute errors (MAE) of 35.7 mm for Stacking-InSAR and 6.5 mm for SBAS-InSAR. Although SBAS-InSAR exhibits higher precision in coherent areas, it provides missing/invalid measurements in the central high-intensity deformation zone (cumulative deformation > 300 mm), indicating limited applicability under strong mining-induced disturbances. Concurrently, a large-scale three-dimensional physical model was constructed, and overburden strain and surface deformation were measured using distributed optical fiber sensing (DOFS) and digital close-range industrial photogrammetry (DCRIP). After scaling to the prototype, the physical-model surface deformation agrees well with the Stacking-InSAR measurements (mean relative error 17%–18%), confirming the reliability of the constructed physical model. Moreover, a prediction model was developed using synergistic monitoring data, achieving a root mean square error (RMSE) of 52.2 mm (i.e., 6.35% of the maximum deformation of 822.4 mm) and a standard deviation (STD) of 51.8 mm (6.30%) relative to the Stacking-InSAR results. These results demonstrate that the proposed framework effectively bridges laboratory-scale observations and field-scale satellite measurements, thereby improving the understanding of underground-to-surface deformation propagation under mining-induced disturbances. Furthermore, this integration enables high-precision estimation of mining-induced deformation and supports coupled surface and underground deformation analysis.
{"title":"A fusion framework of 3D physical model and InSAR monitoring for mining-induced deformation analysis","authors":"Teng Wang, Yunjia Wang, Feng Zhao, Sen Du, Nianbin Zhang, Kewei Zhang, Zhongwei Shen, José Fernández","doi":"10.1016/j.jag.2026.105122","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105122","url":null,"abstract":"Underground resource extraction redistributes subsurface stress, inducing overburden movement and ground deformation. These geomechanical changes can cause ground fissures, slope failures, and collapses, resulting in resource loss, economic damage, safety risks, and environmental degradation. Therefore, accurate monitoring of surface deformation and understanding of deformation propagation mechanisms are crucial for sustainable resource extraction and geohazard mitigation. To this end, a fusion framework is proposed that integrates a three-dimensional physical model with Interferometric Synthetic Aperture Radar (InSAR) technology to monitor surface deformation and investigate deformation propagation. A mining area in China was selected as a case study. Sentinel-1A images were processed using Stacking-InSAR and Small Baseline Subset InSAR (SBAS-InSAR) techniques to derive surface deformation. InSAR-derived deformation was validated against GNSS measurements, yielding mean absolute errors (MAE) of 35.7 mm for Stacking-InSAR and 6.5 mm for SBAS-InSAR. Although SBAS-InSAR exhibits higher precision in coherent areas, it provides missing/invalid measurements in the central high-intensity deformation zone (cumulative deformation <mml:math altimg=\"si1.svg\" display=\"inline\"><mml:mo>></mml:mo></mml:math> 300 mm), indicating limited applicability under strong mining-induced disturbances. Concurrently, a large-scale three-dimensional physical model was constructed, and overburden strain and surface deformation were measured using distributed optical fiber sensing (DOFS) and digital close-range industrial photogrammetry (DCRIP). After scaling to the prototype, the physical-model surface deformation agrees well with the Stacking-InSAR measurements (mean relative error 17%–18%), confirming the reliability of the constructed physical model. Moreover, a prediction model was developed using synergistic monitoring data, achieving a root mean square error (RMSE) of 52.2 mm (i.e., 6.35% of the maximum deformation of 822.4 mm) and a standard deviation (STD) of 51.8 mm (6.30%) relative to the Stacking-InSAR results. These results demonstrate that the proposed framework effectively bridges laboratory-scale observations and field-scale satellite measurements, thereby improving the understanding of underground-to-surface deformation propagation under mining-induced disturbances. Furthermore, this integration enables high-precision estimation of mining-induced deformation and supports coupled surface and underground deformation analysis.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"1 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146047824","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}
Ship detection in remote sensing imagery is essential for diverse maritime-related tasks, including ocean surveillance, fisheries management, and environmental assessment. In operational scenarios, optical imagery provides rich texture cues under clear conditions, whereas synthetic aperture radar (SAR) enables reliable observation in nighttime and cloudy weather. However, cross-domain ship detection across optical and SAR modalities is still challenging due to discrepancies in imaging mechanisms, speckle noise, and background clutter, particularly in near-shore scenarios with similar reflection characteristics, together with the arbitrariness of ship orientation. To address these issues, we propose RotCD-Ship, a rotated cross-domain ship detection framework that bridges the domain gap between optical and SAR images while enabling accurate detection of arbitrarily oriented ships. Specifically, a domain knowledge-guided semantic prompt (DKSP) strategy based on SAR physical priors is introduced to suppress background clutter such as ship wakes and coastal interference. To handle modal divergence, we design a progressive feature alignment scheme that combines multi-scale local feature alignment (MSL-align) and global feature alignment (GF-align), enabling transfer of both fine-grained textures and high-level semantics across domains. Furthermore, a coarse-to-fine rotated region of interest (CF-RRoI) generator is developed to enhance localization precision of strip-like ships in SAR images by progressively refining orientation-aware proposals. Extensive evaluations on five public ship detection datasets show that RotCD-Ship significantly outperforms state-of-the-art methods in both accuracy and robustness, achieving an average mAP improvement of 7.5% in the horizontal ship detection task and 5.5% in the oriented ship detection task compared to the best existing methods. In addition, large-scale tests on Gaofen-3 SAR images further verify the strong generalization in dense-ship and complex coastal environments, highlighting the practical applicability of our framework for all-weather maritime monitoring.
{"title":"Bridging optical and SAR images via semantic prompt-guided progressive alignment for rotated cross-domain ship detection","authors":"Longli Ran, Jiaming Li, Haodong Wu, Anqi Wu, Yi He, Qingfeng Guan, Qiqi Zhu","doi":"10.1016/j.jag.2026.105119","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105119","url":null,"abstract":"Ship detection in remote sensing imagery is essential for diverse maritime-related tasks, including ocean surveillance, fisheries management, and environmental assessment. In operational scenarios, optical imagery provides rich texture cues under clear conditions, whereas synthetic aperture radar (SAR) enables reliable observation in nighttime and cloudy weather. However, cross-domain ship detection across optical and SAR modalities is still challenging due to discrepancies in imaging mechanisms, speckle noise, and background clutter, particularly in near-shore scenarios with similar reflection characteristics, together with the arbitrariness of ship orientation. To address these issues, we propose RotCD-Ship, a rotated cross-domain ship detection framework that bridges the domain gap between optical and SAR images while enabling accurate detection of arbitrarily oriented ships. Specifically, a domain knowledge-guided semantic prompt (DKSP) strategy based on SAR physical priors is introduced to suppress background clutter such as ship wakes and coastal interference. To handle modal divergence, we design a progressive feature alignment scheme that combines multi-scale local feature alignment (MSL-align) and global feature alignment (GF-align), enabling transfer of both fine-grained textures and high-level semantics across domains. Furthermore, a coarse-to-fine rotated region of interest (CF-RRoI) generator is developed to enhance localization precision of strip-like ships in SAR images by progressively refining orientation-aware proposals. Extensive evaluations on five public ship detection datasets show that RotCD-Ship significantly outperforms state-of-the-art methods in both accuracy and robustness, achieving an average mAP improvement of 7.5% in the horizontal ship detection task and 5.5% in the oriented ship detection task compared to the best existing methods. In addition, large-scale tests on Gaofen-3 SAR images further verify the strong generalization in dense-ship and complex coastal environments, highlighting the practical applicability of our framework for all-weather maritime monitoring.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"41 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146047870","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}
Leaf Water Content (LWC) is vital for assessing winter wheat growth. Due to spectral band sensitivity variations, improving spectral feature fitting through preprocessing and optimization is a challenging issue. This study systematically investigates wavelength-dependent LWC sensitivity under multiple preprocessing operators to identify spectral bands highly correlated with winter wheat LWC under different preprocessing methods and evaluate band replacement (representation selection at fixed wavelength) and fusion strategies (constrained union within the same feature-selection method with wavelength de-duplication) for optimizing feature combinations and enhancing model accuracy and robustness. Various feature extraction techniques were applied to identify LWC-correlated spectral bands. Bands replacement (producing “Replacement Features”) and constrained feature fusion (producing “Fusion Features”) were introduced for feature optimization. Multiple modeling methods were employed to assess LWC monitoring performance using initial, replaced, and fused feature sets under a consistent train–test split with training-set cross-validation. The combination of first derivative (FD) preprocessing, Iterative Variable Subset Optimization (IVSO) feature selection, and Backpropagation Neural Network (BPNN) modeling yielded the best monitoring results. Fusion Features selected by IVSO achieved the highest accuracy, evaluated by R2, RMSE, and Akaike information criterion (AIC) (R2Train = 0.960, RMSETrain = 0.016, AICTrain = -1010.061; R2Test = 0.956, RMSETest = 0.015, AICTest = -98.362). Both band replacement and fusion enhanced model robustness and addressed spectral sensitivity issues. This study demonstrated the importance of preprocessing, feature optimization, and modeling in improving LWC monitoring. The proposed multi-preprocessing representation replacement and constrained fusion framework improved LWC estimation accuracy, supporting precision agriculture for winter wheat.
{"title":"Feature replacement and fusion enhance the accuracy of canopy spectral monitoring models for winter wheat leaf water content","authors":"Zhigang Wang, Sha Yang, Qing Liang, Xujing Yang, Meichen Feng, Xiaobin Yan, Xinkai Sun, Mingxing Qin, Chao Wang, Yu Zhao, Wude Yang, Lujie Xiao, Meijun Zhang, Xiaoyan Song, Yongkai Xie","doi":"10.1016/j.jag.2026.105110","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105110","url":null,"abstract":"Leaf Water Content (LWC) is vital for assessing winter wheat growth. Due to spectral band sensitivity variations, improving spectral feature fitting through preprocessing and optimization is a challenging issue. This study systematically investigates wavelength-dependent LWC sensitivity under multiple preprocessing operators to identify spectral bands highly correlated with winter wheat LWC under different preprocessing methods and evaluate band replacement (representation selection at fixed wavelength) and fusion strategies (constrained union within the same feature-selection method with wavelength de-duplication) for optimizing feature combinations and enhancing model accuracy and robustness. Various feature extraction techniques were applied to identify LWC-correlated spectral bands. Bands replacement (producing “Replacement Features”) and constrained feature fusion (producing “Fusion Features”) were introduced for feature optimization. Multiple modeling methods were employed to assess LWC monitoring performance using initial, replaced, and fused feature sets under a consistent train–test split with training-set cross-validation. The combination of first derivative (FD) preprocessing, Iterative Variable Subset Optimization (IVSO) feature selection, and Backpropagation Neural Network (BPNN) modeling yielded the best monitoring results. Fusion Features selected by IVSO achieved the highest accuracy, evaluated by R<ce:sup loc=\"post\">2</ce:sup>, RMSE, and Akaike information criterion (AIC) (R<ce:sup loc=\"post\">2</ce:sup><ce:inf loc=\"post\">Train</ce:inf> = 0.960, RMSE<ce:inf loc=\"post\">Train</ce:inf> = 0.016, AIC<ce:inf loc=\"post\">Train</ce:inf> = -1010.061; R<ce:sup loc=\"post\">2</ce:sup><ce:inf loc=\"post\">Test</ce:inf> = 0.956, RMSE<ce:inf loc=\"post\">Test</ce:inf> = 0.015, AIC<ce:inf loc=\"post\">Test</ce:inf> = -98.362). Both band replacement and fusion enhanced model robustness and addressed spectral sensitivity issues. This study demonstrated the importance of preprocessing, feature optimization, and modeling in improving LWC monitoring. The proposed multi-preprocessing representation replacement and constrained fusion framework improved LWC estimation accuracy, supporting precision agriculture for winter wheat.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"87 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146047828","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 : 2026-01-23DOI: 10.1016/j.jag.2026.105099
Wei Liu, Yuhang Zhong, Shida Zhao, Songling Luo, Yongtao Yu, Xiaomei Zhong, Weikai Tan, Haiyan Guan, Hongjie He, Jonathan Li
High-resolution digital elevation models (DEMs) are critical for applications such as environmental monitoring and urban planning, motivating the development of advanced DEM super-resolution (SR) techniques. While recent methods have shown promising results, effectively exploiting high-resolution remote sensing images (HRSIs) to guide DEM SR remains challenging, and progress has been hindered by the lack of large-scale, open-source benchmark datasets. We propose GSRMTL, a novel and parameter-efficient multi-task learning framework for HRSI-guided DEM SR. Given a low-resolution DEM and a paired HRSI, GSRMTL jointly performs DEM SR and semantic segmentation of the optical imagery, where segmentation acts as an auxiliary task to provide semantic priors for elevation reconstruction. To address the dataset bottleneck, we introduce GDEMSR, the first large-scale benchmark dataset specifically designed for HRSI-guided DEM SR. Extensive experiments on GDEMSR and the RGB-guided depth SR benchmark NYU-v2 demonstrate that GSRMTL consistently outperforms state-of-the-art methods while using significantly fewer parameters, highlighting its effectiveness and practical deployment potential.
{"title":"DEM super-resolution guided by high-resolution remote sensing images using multitask learning","authors":"Wei Liu, Yuhang Zhong, Shida Zhao, Songling Luo, Yongtao Yu, Xiaomei Zhong, Weikai Tan, Haiyan Guan, Hongjie He, Jonathan Li","doi":"10.1016/j.jag.2026.105099","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105099","url":null,"abstract":"High-resolution digital elevation models (DEMs) are critical for applications such as environmental monitoring and urban planning, motivating the development of advanced DEM super-resolution (SR) techniques. While recent methods have shown promising results, effectively exploiting high-resolution remote sensing images (HRSIs) to guide DEM SR remains challenging, and progress has been hindered by the lack of large-scale, open-source benchmark datasets. We propose GSRMTL, a novel and parameter-efficient multi-task learning framework for HRSI-guided DEM SR. Given a low-resolution DEM and a paired HRSI, GSRMTL jointly performs DEM SR and semantic segmentation of the optical imagery, where segmentation acts as an auxiliary task to provide semantic priors for elevation reconstruction. To address the dataset bottleneck, we introduce GDEMSR, the first large-scale benchmark dataset specifically designed for HRSI-guided DEM SR. Extensive experiments on GDEMSR and the RGB-guided depth SR benchmark NYU-v2 demonstrate that GSRMTL consistently outperforms state-of-the-art methods while using significantly fewer parameters, highlighting its effectiveness and practical deployment potential.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"30 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146047873","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 : 2026-01-23DOI: 10.1016/j.jag.2025.105070
Jian Kang, Haiyan Guan, Dedong Zhang, Lingfei Ma, Lanying Wang, Yongtao Yu, Linlin Xu, Jonathan Li
Accurately and timely detecting road pavement damage helps monitor road deterioration extent, thereby guiding maintenance projects and ensuring traffic safety. Nevertheless, due to textural similarity and nested distribution between neighboring pavement damages, as well as the damages with the diversity sizes, irregular shapes, multiple categories, current methods have the limitation in the high-quality detection from road street-level images. To tackle these challenges, this paper develops a novel real-time anchor-free network with a one-stage processing architecture, named RPDNet, for precisely and accurately detecting pavement damages from streel-level road images. First, stacked with a layer-by-layer encoding structure boosted by a deformable fully-attentive module as the backbone extractor, the RPDNet can capture more fine-grained information and generate multiscale strong task-aware semantics, favoring significantly the discrimination noteworthy textural and geometric features. Then, by adopting a multi-level efficient aggregation neck, the RPDNet can promote informative spatial details and integrate the different-level damage encoding features, contributing to the light-weight and optimization of the whole architecture. Afterward, designed with a dual-large kernel module, embedded in a decoupled detection head with anchor-free guidance, the RPDNet can project the ranging dependency of salient and task-oriented pavement damage objects by adaptively aggregating information across large kernels in spatial-domain. Qualitative and quantitative evaluations confirmed that the RPDNet provided a promiseful solution for detecting pavement damages in industrial applications under complex street-level road conditions. Furthermore, comparative analysis with the latest anchor-based and anchor-free alternatives also proved the superiority and generalization of the RPDNet in pavement damage detection tasks. The assessment results displayed that the RPDNet obtained an average mAP@0.5, mAP@0.5:0.95, precision, and recall of 69.16%, 44.86%, 72.59%, and 60.41%, respectively, on two dataset. Additionally, we constructed a large-size multi-city road pavement damage image dataset to support urban road health monitoring.
{"title":"RPDNet: Street-level road pavement damage detection with a real-time anchor-free network","authors":"Jian Kang, Haiyan Guan, Dedong Zhang, Lingfei Ma, Lanying Wang, Yongtao Yu, Linlin Xu, Jonathan Li","doi":"10.1016/j.jag.2025.105070","DOIUrl":"https://doi.org/10.1016/j.jag.2025.105070","url":null,"abstract":"Accurately and timely detecting road pavement damage helps monitor road deterioration extent, thereby guiding maintenance projects and ensuring traffic safety. Nevertheless, due to textural similarity and nested distribution between neighboring pavement damages, as well as the damages with the diversity sizes, irregular shapes, multiple categories, current methods have the limitation in the high-quality detection from road street-level images. To tackle these challenges, this paper develops a novel real-time anchor-free network with a one-stage processing architecture, named RPDNet, for precisely and accurately detecting pavement damages from streel-level road images. First, stacked with a layer-by-layer encoding structure boosted by a deformable fully-attentive module as the backbone extractor, the RPDNet can capture more fine-grained information and generate multiscale strong task-aware semantics, favoring significantly the discrimination noteworthy textural and geometric features. Then, by adopting a multi-level efficient aggregation neck, the RPDNet can promote informative spatial details and integrate the different-level damage encoding features, contributing to the light-weight and optimization of the whole architecture. Afterward, designed with a dual-large kernel module, embedded in a decoupled detection head with anchor-free guidance, the RPDNet can project the ranging dependency of salient and task-oriented pavement damage objects by adaptively aggregating information across large kernels in spatial-domain. Qualitative and quantitative evaluations confirmed that the RPDNet provided a promiseful solution for detecting pavement damages in industrial applications under complex street-level road conditions. Furthermore, comparative analysis with the latest anchor-based and anchor-free alternatives also proved the superiority and generalization of the RPDNet in pavement damage detection tasks. The assessment results displayed that the RPDNet obtained an average <ce:italic>mAP@0.5</ce:italic>, <ce:italic>mAP@0.5:0.95</ce:italic>, <ce:italic>precision</ce:italic>, and <ce:italic>recall</ce:italic> of 69.16%, 44.86%, 72.59%, and 60.41%, respectively, on two dataset. Additionally, we constructed a large-size multi-city road pavement damage image dataset to support urban road health monitoring.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"288 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146047912","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 : 2026-01-23DOI: 10.1016/j.jag.2026.105101
Jianru Yang, Hao Zheng, Weiwei Sun, Yuekai Hu, Weiguo Zhang, Chunpeng Chen, Yunxuan Zhou, Heqin Cheng, Weiming Xie, Kai Tan
Salt-marsh Fairy circles (FC) are enigmatic, quasi-circular structures linked to interacting biogeophysical processes, yet they remain difficult to detect and quantify at scale from conventional RGB imagery. Limited labeled data, transient and variable FC appearance, and severe class-imbalance make single-model machine learning (ML) unreliable for quantitative monitoring. We propose a framework for automatic FC recognition and enumeration on 3-band imagery. A zero-shot foundation model (SAM) segments images into instance-level blocks. Novel distribution-pattern and geometric features, class-equalized losses, weighted resampling, and augmentation are applied within deep-learning (U-Net, Attention-U-Net, Swin-Unet) and ensemble-learning (Random Forest, XGBoost) models. The key innovation is an imbalance-aware Bayesian method that fuses pixel-wise probabilities across models; a counting algorithm then tallies FC instances. We evaluate eight pan-sharpened scenes covering four sites along China’s coast. No individual ML model or standard Bayesian fusion is fully satisfactory. The imbalance-aware Bayesian method improves over the best single model: tight scheme: κ rises from 0.69 to 0.76, F1-score from 70.9% to 75.8% (Class 1) and from 63.5% to 68.2% (Class 2), and AUC from 84.8% to 93.1% and from 78.5% to 84.8%; loose scheme: κ increases from 0.74 to 0.79, AUC from 85.1% to 90.3%, F1-score from 74.3% to 78.6%. The counting algorithm achieves RMSE 1.62 and MAPE 0.33% over 1,135 instances, outperforming DBSCAN. A 22-month case study on Chongming Island captures marsh expansion and dieback dynamics through shifts between FC classes. Our framework delivers reliable FC recognition and enumeration on a small dataset with severe class-imbalance, generalizing across salt-marsh types.
{"title":"Recognition of salt-marsh fairy circles in conventional optical satellite imagery: A generalizable framework with multiple machine learning models and imbalanced Bayesian probability updating","authors":"Jianru Yang, Hao Zheng, Weiwei Sun, Yuekai Hu, Weiguo Zhang, Chunpeng Chen, Yunxuan Zhou, Heqin Cheng, Weiming Xie, Kai Tan","doi":"10.1016/j.jag.2026.105101","DOIUrl":"https://doi.org/10.1016/j.jag.2026.105101","url":null,"abstract":"Salt-marsh Fairy circles (FC) are enigmatic, quasi-circular structures linked to interacting biogeophysical processes, yet they remain difficult to detect and quantify at scale from conventional RGB imagery. Limited labeled data, transient and variable FC appearance, and severe class-imbalance make single-model machine learning (ML) unreliable for quantitative monitoring. We propose a framework for automatic FC recognition and enumeration on 3-band imagery. A zero-shot foundation model (SAM) segments images into instance-level blocks. Novel distribution-pattern and geometric features, class-equalized losses, weighted resampling, and augmentation are applied within deep-learning (U-Net, Attention-U-Net, Swin-Unet) and ensemble-learning (Random Forest, XGBoost) models. The key innovation is an imbalance-aware Bayesian method that fuses pixel-wise probabilities across models; a counting algorithm then tallies FC instances. We evaluate eight pan-sharpened scenes covering four sites along China’s coast. No individual ML model or standard Bayesian fusion is fully satisfactory. The imbalance-aware Bayesian method improves over the best single model: tight scheme: κ rises from 0.69 to 0.76, F1-score from 70.9% to 75.8% (Class 1) and from 63.5% to 68.2% (Class 2), and AUC from 84.8% to 93.1% and from 78.5% to 84.8%; loose scheme: κ increases from 0.74 to 0.79, AUC from 85.1% to 90.3%, F1-score from 74.3% to 78.6%. The counting algorithm achieves RMSE 1.62 and MAPE 0.33% over 1,135 instances, outperforming DBSCAN. A 22-month case study on Chongming Island captures marsh expansion and dieback dynamics through shifts between FC classes. Our framework delivers reliable FC recognition and enumeration on a small dataset with severe class-imbalance, generalizing across salt-marsh types.","PeriodicalId":50341,"journal":{"name":"International Journal of Applied Earth Observation and Geoinformation","volume":"14 1","pages":""},"PeriodicalIF":7.5,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146047827","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：