Pub Date : 2026-01-20DOI: 10.1016/j.jappgeo.2026.106124
Lei Qin , Hui Wang , Haifei Lin , Pengfei Liu , Shiyin Lv , Jiawei Li
The content of unfrozen water in frozen coal affects the permeability of coal at low temperature, and the study of the ice-water phase change during the freezing and thawing process of the coal body is the key to study the liquid nitrogen fracturing and seepage enhancement technology. In this paper, we take Hengyi bituminous coal as the research object, and study the pore structure evolution and unfrozen water distribution changes during the thawing process based on nuclear magnetic resonance technique for high water content and low water content coal samples at different freezing times. The results show that the water space ratio growth of coal samples during thawing can be divided into three stages; liquid nitrogen freeze-thaw coal sample can significantly promote the development of large pores and large pore throats, and the difference of initial water content only has a significant effect on the development of large pores and large pore throats. Pore diameter is positively correlated with pore ice melting-point, and in the frozen coal sample, the unfrozen water at the initial melting stage mainly exists in the small water space. Freezing process of low-temperature liquid nitrogen on the coal mass has been freeze-swelling and freeze-shrinking effect, different freezing time will affect the combined effect of freeze-swelling and freeze-shrinking effect, resulting in the variability of pore throat and pore space development under different freezing time.
{"title":"NMR study on the changes of water content characteristics and pore structure evolution during melting of coal frozen with liquid nitrogen","authors":"Lei Qin , Hui Wang , Haifei Lin , Pengfei Liu , Shiyin Lv , Jiawei Li","doi":"10.1016/j.jappgeo.2026.106124","DOIUrl":"10.1016/j.jappgeo.2026.106124","url":null,"abstract":"<div><div>The content of unfrozen water in frozen coal affects the permeability of coal at low temperature, and the study of the ice-water phase change during the freezing and thawing process of the coal body is the key to study the liquid nitrogen fracturing and seepage enhancement technology. In this paper, we take Hengyi bituminous coal as the research object, and study the pore structure evolution and unfrozen water distribution changes during the thawing process based on nuclear magnetic resonance technique for high water content and low water content coal samples at different freezing times. The results show that the water space ratio growth of coal samples during thawing can be divided into three stages; liquid nitrogen freeze-thaw coal sample can significantly promote the development of large pores and large pore throats, and the difference of initial water content only has a significant effect on the development of large pores and large pore throats. Pore diameter is positively correlated with pore ice melting-point, and in the frozen coal sample, the unfrozen water at the initial melting stage mainly exists in the small water space. Freezing process of low-temperature liquid nitrogen on the coal mass has been freeze-swelling and freeze-shrinking effect, different freezing time will affect the combined effect of freeze-swelling and freeze-shrinking effect, resulting in the variability of pore throat and pore space development under different freezing time.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"246 ","pages":"Article 106124"},"PeriodicalIF":2.1,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146038839","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 : 2026-01-20DOI: 10.1016/j.jappgeo.2026.106117
Ao Song , Aichun Liu , Zhixiang Li , Guanzhong Liu , Aipeng Guo , Junfeng Jiang
This study presents an innovative application of Distributed Acoustic Sensing (DAS) by repurposing urban sewer pipelines into a large-scale sensing network through the deployment of fiber-optic cables. This approach facilitates three major applications: subsurface imaging, pipeline blockage detection, and urban traffic monitoring. Using passive seismic interferometry on ambient noise signals acquired via the in-pipe fiber, we reconstructed high-resolution shear-wave velocity profiles of the shallow urban subsurface. Combined analysis of field data and numerical simulations identified characteristic patterns associated with pipeline blockages in cross-correlations (CCs), which were validated through closed-circuit television (CCTV) inspections. For traffic monitoring, vehicle-induced vibrations were processed using seismic attribute analysis and a U-Net convolutional neural network, enabling precise vehicle trajectory identification and speed estimation based on Hilbert instantaneous amplitude attributes. The results demonstrate that the proposed DAS-based method offers a non-invasive, cost-effective, and scalable solution for integrated urban monitoring, providing a sustainable alternative to traditional point-based sensing and enabling continuous, large-scale infrastructure assessment in densely populated areas.
{"title":"Triple-duty distributed acoustic sensing in urban environments: Concurrent subsurface imaging, pipeline diagnostics, and traffic surveillance","authors":"Ao Song , Aichun Liu , Zhixiang Li , Guanzhong Liu , Aipeng Guo , Junfeng Jiang","doi":"10.1016/j.jappgeo.2026.106117","DOIUrl":"10.1016/j.jappgeo.2026.106117","url":null,"abstract":"<div><div>This study presents an innovative application of Distributed Acoustic Sensing (DAS) by repurposing urban sewer pipelines into a large-scale sensing network through the deployment of fiber-optic cables. This approach facilitates three major applications: subsurface imaging, pipeline blockage detection, and urban traffic monitoring. Using passive seismic interferometry on ambient noise signals acquired via the in-pipe fiber, we reconstructed high-resolution shear-wave velocity profiles of the shallow urban subsurface. Combined analysis of field data and numerical simulations identified characteristic patterns associated with pipeline blockages in cross-correlations (CCs), which were validated through closed-circuit television (CCTV) inspections. For traffic monitoring, vehicle-induced vibrations were processed using seismic attribute analysis and a U-Net convolutional neural network, enabling precise vehicle trajectory identification and speed estimation based on Hilbert instantaneous amplitude attributes. The results demonstrate that the proposed DAS-based method offers a non-invasive, cost-effective, and scalable solution for integrated urban monitoring, providing a sustainable alternative to traditional point-based sensing and enabling continuous, large-scale infrastructure assessment in densely populated areas.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"246 ","pages":"Article 106117"},"PeriodicalIF":2.1,"publicationDate":"2026-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146038842","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 : 2026-01-19DOI: 10.1016/j.jappgeo.2026.106118
Maciej J. Mendecki , Rafał Warchulski , Mateusz Kicza
This study presents Electrical Resistivity Tomography (ERT) and Electromagnetic Induction (EMI) measurements conducted in two areas: the Kraków Bishops' Castle (area A1) and the Municipal Park (area A2) in Sławków. ERT data are displayed as cross-sections, while EMI data are mapped. A reference resistivity of 350 Ωm was established for natural geological substrates. Anomalies exceeding this threshold suggest anthropogenic origins, including remnants of the Bishops' Castle. In A1, ERT profiles ERT1–ERT3 revealed high-resistivity anomalies linked to rock fragments, possible tunnels, and castle walls; shallower ones (<2 m) were interpreted cautiously due to natural effects or artifacts. In A2, ERT4–ERT7 profiles indicated embankments, rock fragments, and inferred defensive structures. EMI confirmed anomalies: two subsurface features inside the castle near NE and SW walls (potential metallic objects or a well).
Extended verification analyzed ERT statistical analysis (RMS, χ2, residual analysis, observed vs. interpreted scatter plots), Depth of Investigation Index (DOI), and for EMI data analysis (spatial data analysis, variograms, EMI-derived resistivity, in-phase difference maps, and EMI data cross-validation), emphasizing careful interpretation under complex geological-anthropogenic conditions. The study refines archaeological geophysics practices, optimizing techniques for varied materials and site histories.
{"title":"Geophysical survey of the medieval Castle in Sławków, Poland: Insights from ERT and EM","authors":"Maciej J. Mendecki , Rafał Warchulski , Mateusz Kicza","doi":"10.1016/j.jappgeo.2026.106118","DOIUrl":"10.1016/j.jappgeo.2026.106118","url":null,"abstract":"<div><div>This study presents Electrical Resistivity Tomography (ERT) and Electromagnetic Induction (EMI) measurements conducted in two areas: the Kraków Bishops' Castle (area A1) and the Municipal Park (area A2) in Sławków. ERT data are displayed as cross-sections, while EMI data are mapped. A reference resistivity of 350 Ωm was established for natural geological substrates. Anomalies exceeding this threshold suggest anthropogenic origins, including remnants of the Bishops' Castle. In A1, ERT profiles ERT1–ERT3 revealed high-resistivity anomalies linked to rock fragments, possible tunnels, and castle walls; shallower ones (<2 m) were interpreted cautiously due to natural effects or artifacts. In A2, ERT4–ERT7 profiles indicated embankments, rock fragments, and inferred defensive structures. EMI confirmed anomalies: two subsurface features inside the castle near NE and SW walls (potential metallic objects or a well).</div><div>Extended verification analyzed ERT statistical analysis (RMS, χ<sup>2</sup>, residual analysis, observed vs. interpreted scatter plots), Depth of Investigation Index (DOI), and for EMI data analysis (spatial data analysis, variograms, EMI-derived resistivity, in-phase difference maps, and EMI data cross-validation), emphasizing careful interpretation under complex geological-anthropogenic conditions. The study refines archaeological geophysics practices, optimizing techniques for varied materials and site histories.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"246 ","pages":"Article 106118"},"PeriodicalIF":2.1,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146038394","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 : 2026-01-19DOI: 10.1016/j.jappgeo.2026.106107
Zixiang Zhou , Guochang Liu , Min Bai , Zhaoyang Ma , Zhiyong Wang , Yannan Wang
Seismic data denoising is a challenging task in complex noise environments, especially in unsupervised learning settings where labeled data is unavailable. Existing unsupervised learning methods, such as Deep Image Prior, effectively remove noise but still face issues related to network structure stability during training, which limits their accuracy. To further improve denoising performance, this paper proposes a patch selection-based dual attention deep learning model (PS-DADL) designed to suppress random and erratic noise in seismic data. First, we adopt a patch-based processing approach, selecting patches with high information content for training based on variance calculations, which improves the model's training efficiency. Then, we design a deep neural network that extracts features and recovers denoised signals through an encoder-decoder structure. Additionally, a dual attention module is introduced. This module adaptively aggregates dependencies within the data through spatial and channel attention mechanisms, enhancing feature representation and boosting the model's adaptability in complex noise environments. Experimental results show that PS-DADL, within the unsupervised learning framework, improves seismic data quality and demonstrates strong robustness, outperforming several baseline unsupervised learning methods.
{"title":"Patch selection-based dual attention unsupervised deep learning model for suppressing random and erratic noise in seismic data","authors":"Zixiang Zhou , Guochang Liu , Min Bai , Zhaoyang Ma , Zhiyong Wang , Yannan Wang","doi":"10.1016/j.jappgeo.2026.106107","DOIUrl":"10.1016/j.jappgeo.2026.106107","url":null,"abstract":"<div><div>Seismic data denoising is a challenging task in complex noise environments, especially in unsupervised learning settings where labeled data is unavailable. Existing unsupervised learning methods, such as Deep Image Prior, effectively remove noise but still face issues related to network structure stability during training, which limits their accuracy. To further improve denoising performance, this paper proposes a patch selection-based dual attention deep learning model (PS-DADL) designed to suppress random and erratic noise in seismic data. First, we adopt a patch-based processing approach, selecting patches with high information content for training based on variance calculations, which improves the model's training efficiency. Then, we design a deep neural network that extracts features and recovers denoised signals through an encoder-decoder structure. Additionally, a dual attention module is introduced. This module adaptively aggregates dependencies within the data through spatial and channel attention mechanisms, enhancing feature representation and boosting the model's adaptability in complex noise environments. Experimental results show that PS-DADL, within the unsupervised learning framework, improves seismic data quality and demonstrates strong robustness, outperforming several baseline unsupervised learning methods.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"246 ","pages":"Article 106107"},"PeriodicalIF":2.1,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146038393","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 : 2026-01-17DOI: 10.1016/j.jappgeo.2026.106111
Changxing Zhu, Duo Li, Dazhi Wu, Jiaxin Huo
To mitigate disasters such as sand inrush and water gushing potentially induced by sand layer geology in underground engineering, carbon fiber-reinforced grouting materials demonstrate promising potential for remediation. However, in practical engineering, static and dynamic loads often act on surrounding rock in combination, and existing research on this aspect remains limited. To address this, this study employs micro-nano carbon fibers to modify ultrafine cement-based grouting material. Laboratory grouting reinforcement tests were conducted on graded sand layers. Using a Split Hopkinson Pressure Bar (SHPB) equipped with an active confining pressure device, systematic dynamic compression tests were performed under various impact velocities (corresponding to strain rates of 49 to 92 s−1) and confining pressures (0 to 8 MPa). The results indicate that the peak stress of the specimens increases with the strain rate, exhibiting a significant strain rate effect. Under different impact velocities and confining pressures, the peak stress of fiber-containing specimens was significantly higher than that of plain specimens, confirming the enhancing and toughening effect of carbon fibers. When the confining pressure increased to 8 MPa, the peak stress of fiber-containing specimens was approximately 1.56 times higher than that under unconfined conditions, and the failure mode transitioned from tensile splitting to crushing failure. Microscopically, carbon fibers effectively inhibit crack propagation and enhance energy absorption capacity through “micro-reinforcement” and three-dimensional network bridging, with their primary failure modes being fiber debonding or fracture. The strain rate effect of the specimens originates from the combined action of microscopic damage evolution and inertial lateral confinement. The confining pressure enhancement mechanism primarily lies in suppressing brittle crack propagation, driving the material towards a triaxial stress state and inducing ductile hardening. This research reveals the dynamic response mechanism of carbon fiber-reinforced grouted bodies under coupled static-dynamic loading, providing a material basis and theoretical foundation for the reinforcement design in sand layer strata.
{"title":"Strain rate and confining pressure effects in micro-nano carbon fiber-reinforced grout: an SHPB impact study","authors":"Changxing Zhu, Duo Li, Dazhi Wu, Jiaxin Huo","doi":"10.1016/j.jappgeo.2026.106111","DOIUrl":"10.1016/j.jappgeo.2026.106111","url":null,"abstract":"<div><div>To mitigate disasters such as sand inrush and water gushing potentially induced by sand layer geology in underground engineering, carbon fiber-reinforced grouting materials demonstrate promising potential for remediation. However, in practical engineering, static and dynamic loads often act on surrounding rock in combination, and existing research on this aspect remains limited. To address this, this study employs micro-nano carbon fibers to modify ultrafine cement-based grouting material. Laboratory grouting reinforcement tests were conducted on graded sand layers. Using a Split Hopkinson Pressure Bar (SHPB) equipped with an active confining pressure device, systematic dynamic compression tests were performed under various impact velocities (corresponding to strain rates of 49 to 92 s<sup>−1</sup>) and confining pressures (0 to 8 MPa). The results indicate that the peak stress of the specimens increases with the strain rate, exhibiting a significant strain rate effect. Under different impact velocities and confining pressures, the peak stress of fiber-containing specimens was significantly higher than that of plain specimens, confirming the enhancing and toughening effect of carbon fibers. When the confining pressure increased to 8 MPa, the peak stress of fiber-containing specimens was approximately 1.56 times higher than that under unconfined conditions, and the failure mode transitioned from tensile splitting to crushing failure. Microscopically, carbon fibers effectively inhibit crack propagation and enhance energy absorption capacity through “micro-reinforcement” and three-dimensional network bridging, with their primary failure modes being fiber debonding or fracture. The strain rate effect of the specimens originates from the combined action of microscopic damage evolution and inertial lateral confinement. The confining pressure enhancement mechanism primarily lies in suppressing brittle crack propagation, driving the material towards a triaxial stress state and inducing ductile hardening. This research reveals the dynamic response mechanism of carbon fiber-reinforced grouted bodies under coupled static-dynamic loading, providing a material basis and theoretical foundation for the reinforcement design in sand layer strata.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"246 ","pages":"Article 106111"},"PeriodicalIF":2.1,"publicationDate":"2026-01-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146038837","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 : 2026-01-15DOI: 10.1016/j.jappgeo.2026.106112
Longxiang Han , Chengliang Wu , Huazhong Wang
Consistent-phase stacking of band-limited imaging wavelet from the same subsurface reflection point but different source-receiver pairs is a fundamental requirement for achieving high-fidelity and high-resolution seismic imaging. The final image is typically obtained by stacking common-image gathers (CIGs). However, inconsistent wavelet phases across different angles or offsets in CIGs can lead to destructive interference, waveform distortion, and amplitude loss, ultimately degrading image resolution. Most conventional phase correction methods assume a constant phase shift across all frequencies, which fails to account for frequency-dependent phase variations introduced by source signatures, absorption, and other real-field factors. Neglecting these variations can significantly degrade the fidelity and resolution of the final stacked image. To address this issue, we propose a statistical method for detecting and correcting frequency-dependent phase differences in CIGs. After flattening the CIGs, we perform multi-scale Gaussian filtering to divide the data into narrow frequency bands, effectively reducing noise and ensuring more stable phase estimation. Then, the phase differences between the original and a reference CIG—formed by averaging multiple traces within the effective illumination range—are estimated for each narrow frequency band using a particle swarm optimization (PSO) algorithm. Treating the measured phase shift in each band as corresponding to its center frequency, we employ spline interpolation to construct a smooth, continuous phase correction curve. This curve is then applied to correct the wavelet phase across the full bandwidth. Both synthetic and field data are used to demonstrate the effectiveness of the proposed method. Experimental results show that the method effectively corrects residual phase differences in CIGs, significantly enhancing the amplitude fidelity and resolution of the final seismic image.
{"title":"Residual phase correction for common imaging gathers and its application in fidelity high-resolution imaging","authors":"Longxiang Han , Chengliang Wu , Huazhong Wang","doi":"10.1016/j.jappgeo.2026.106112","DOIUrl":"10.1016/j.jappgeo.2026.106112","url":null,"abstract":"<div><div>Consistent-phase stacking of band-limited imaging wavelet from the same subsurface reflection point but different source-receiver pairs is a fundamental requirement for achieving high-fidelity and high-resolution seismic imaging. The final image is typically obtained by stacking common-image gathers (CIGs). However, inconsistent wavelet phases across different angles or offsets in CIGs can lead to destructive interference, waveform distortion, and amplitude loss, ultimately degrading image resolution. Most conventional phase correction methods assume a constant phase shift across all frequencies, which fails to account for frequency-dependent phase variations introduced by source signatures, absorption, and other real-field factors. Neglecting these variations can significantly degrade the fidelity and resolution of the final stacked image. To address this issue, we propose a statistical method for detecting and correcting frequency-dependent phase differences in CIGs. After flattening the CIGs, we perform multi-scale Gaussian filtering to divide the data into narrow frequency bands, effectively reducing noise and ensuring more stable phase estimation. Then, the phase differences between the original and a reference CIG—formed by averaging multiple traces within the effective illumination range—are estimated for each narrow frequency band using a particle swarm optimization (PSO) algorithm. Treating the measured phase shift in each band as corresponding to its center frequency, we employ spline interpolation to construct a smooth, continuous phase correction curve. This curve is then applied to correct the wavelet phase across the full bandwidth. Both synthetic and field data are used to demonstrate the effectiveness of the proposed method. Experimental results show that the method effectively corrects residual phase differences in CIGs, significantly enhancing the amplitude fidelity and resolution of the final seismic image.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"246 ","pages":"Article 106112"},"PeriodicalIF":2.1,"publicationDate":"2026-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146038841","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 : 2026-01-14DOI: 10.1016/j.jappgeo.2026.106115
Xiaoliang Xu , Yu He , Huifang Liu , Quan Shi , Xinlong Yao , Kaiyu Tang
Landslide disasters pose a serious threat to human life and property, and landslide susceptibility assessment (LSA) is a core technical approach for landslide risk prevention and control. Conventional LSA methods face challenges in efficiently extracting features and accurately classifying multi source data, and they often lack temporal responsiveness. This study proposes a multi model fusion LSA framework that integrates interferometric synthetic aperture radar (InSAR) data. The framework combines convolutional neural networks with tree based models, incorporates dynamic surface deformation data derived from InSAR inversion, and conducts joint modeling using 14 environmental factors covering topography, geology, hydrology, and human activities. In addition, the SHAP method is employed to provide an interpretable analysis of the model decision mechanisms. The results indicate that elevation, distance to rivers, slope, and surface deformation rate are the key driving factors for landslide occurrence. Among the six comparative models, the XGBoost–VGG fusion model achieves the best performance, with overall results significantly superior to other single and fusion models. Although incorporating the surface deformation factor slightly reduces the overall performance of the models, it substantially enhances their temporal responsiveness. The proposed timeliness oriented, model fusion based LSA approach provides scientific support for landslide risk assessment and demonstrates the practical engineering value of coupling model fusion techniques with dynamic surface deformation data in LSA applications.
{"title":"Landslide susceptibility assessment in a reservoir area using integrated models based on time-series InSAR","authors":"Xiaoliang Xu , Yu He , Huifang Liu , Quan Shi , Xinlong Yao , Kaiyu Tang","doi":"10.1016/j.jappgeo.2026.106115","DOIUrl":"10.1016/j.jappgeo.2026.106115","url":null,"abstract":"<div><div>Landslide disasters pose a serious threat to human life and property, and landslide susceptibility assessment (LSA) is a core technical approach for landslide risk prevention and control. Conventional LSA methods face challenges in efficiently extracting features and accurately classifying multi source data, and they often lack temporal responsiveness. This study proposes a multi model fusion LSA framework that integrates interferometric synthetic aperture radar (InSAR) data. The framework combines convolutional neural networks with tree based models, incorporates dynamic surface deformation data derived from InSAR inversion, and conducts joint modeling using 14 environmental factors covering topography, geology, hydrology, and human activities. In addition, the SHAP method is employed to provide an interpretable analysis of the model decision mechanisms. The results indicate that elevation, distance to rivers, slope, and surface deformation rate are the key driving factors for landslide occurrence. Among the six comparative models, the XGBoost–VGG fusion model achieves the best performance, with overall results significantly superior to other single and fusion models. Although incorporating the surface deformation factor slightly reduces the overall performance of the models, it substantially enhances their temporal responsiveness. The proposed timeliness oriented, model fusion based LSA approach provides scientific support for landslide risk assessment and demonstrates the practical engineering value of coupling model fusion techniques with dynamic surface deformation data in LSA applications.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"246 ","pages":"Article 106115"},"PeriodicalIF":2.1,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146038834","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 : 2026-01-14DOI: 10.1016/j.jappgeo.2026.106116
Zhijun Cheng , Xiang Wang , Guojun Mao , Weijian Mao , Shijun Cheng
Meta-learning-based physics-informed neural networks (Meta-PINN) show significant advantages in solving seismic wave equations across multi-velocity models, where model-agnostic meta-learning (MAML) algorithm is used to learn a shared initialization. The learned initialization helps physics-informed neural networks (PINNs) rapidly adapt to new seismic velocity models. However, the dual-loop optimization mechanism of MAML significantly increases meta-training cost. To address this issue, we introduce the idea of transfer learning into the meta-learning algorithm to reduce the computational burden during the meta-training stage. Specifically, we optimize the meta-model by performing fast gradient updates for a single velocity model on the support set, and then employing a parameter averaging strategy across multiple velocity models on the query set, and the resulting initialization is used for regular training of the new velocity model. Experimental results on diverse velocity models demonstrate that, compared to the conventional Meta-PINN, our method can provide a slightly faster convergence speed and, also, significantly reduce the meta-training time.
{"title":"Meta-transfer learning for efficient initialization of neural network wavefield solutions","authors":"Zhijun Cheng , Xiang Wang , Guojun Mao , Weijian Mao , Shijun Cheng","doi":"10.1016/j.jappgeo.2026.106116","DOIUrl":"10.1016/j.jappgeo.2026.106116","url":null,"abstract":"<div><div>Meta-learning-based physics-informed neural networks (Meta-PINN) show significant advantages in solving seismic wave equations across multi-velocity models, where model-agnostic meta-learning (MAML) algorithm is used to learn a shared initialization. The learned initialization helps physics-informed neural networks (PINNs) rapidly adapt to new seismic velocity models. However, the dual-loop optimization mechanism of MAML significantly increases meta-training cost. To address this issue, we introduce the idea of transfer learning into the meta-learning algorithm to reduce the computational burden during the meta-training stage. Specifically, we optimize the meta-model by performing fast gradient updates for a single velocity model on the support set, and then employing a parameter averaging strategy across multiple velocity models on the query set, and the resulting initialization is used for regular training of the new velocity model. Experimental results on diverse velocity models demonstrate that, compared to the conventional Meta-PINN, our method can provide a slightly faster convergence speed and, also, significantly reduce the meta-training time.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"246 ","pages":"Article 106116"},"PeriodicalIF":2.1,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146038835","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 : 2026-01-14DOI: 10.1016/j.jappgeo.2026.106096
Marieli Machado Zago , Maximilian Fries
Copper deposits are critical resources for modern industries, particularly in the transition toward clean energy technologies, electric vehicles, and digital infrastructure. In southern Brazil, the Lavras do Sul–Caçapava do Sul region represents a metallogenic province that has been extensively studied since the nineteenth century, hosting significant copper and gold occurrences. These deposits are commonly associated with volcanic rocks of the Hilário Formation, which play a central role in the regional mineralization processes. Although structural controls and hydrothermal alteration patterns have been previously documented, the three-dimensional geometry and connectivity of mineralized zones at depth remain insufficiently constrained. This study investigates the geophysical signature of copper mineralization within the Hilário Formation using 3D inversion of aeromagnetic data integrated with structural and geological information. Magnetic enhancement techniques such as the Tilt-angle derivative, Analytic Signal (AS), and Euler Deconvolution were applied to improve the detection of subsurface structures and magnetic sources. Additionally, Magnetization Vector Inversion (MVI) was employed to refine the delineation of magnetic bodies associated with mineralization. The integrated analysis revealed NE- and NW-trending fault systems as the dominant structural frameworks influencing copper mineralization. Magnetic lows near the surface, interpreted as hydrothermal alteration zones, were found overlying deeper magnetic highs related to magnetite-rich and potentially sulfide-bearing zones. The combined application of Euler Deconvolution and MVI produced consistent results that correlate well with known geological features, improving subsurface interpretation and reducing uncertainty in the modeling of mineralized bodies. Overall, the results demonstrate the effectiveness of integrating advanced geophysical techniques with geological and structural datasets for copper exploration. The proposed workflow enhances interpretive confidence, supports target delineation, and provides a robust framework for future exploration in the region.
铜矿是现代工业的关键资源，尤其是在向清洁能源技术、电动汽车和数字基础设施转型的过程中。在巴西南部，lalavas do Sul - carapava do Sul地区是一个成矿省，自19世纪以来，人们对该地区进行了广泛的研究，发现了大量的铜和金矿床。这些矿床通常与Hilário组火山岩伴生，在区域成矿过程中起中心作用。尽管构造控制和热液蚀变模式已经被记录下来，但深部矿化带的三维几何形状和连通性仍然没有得到充分的限制。利用航磁数据三维反演，结合构造和地质信息，研究了Hilário组内铜成矿的地球物理特征。利用倾斜导数、解析信号（as）和欧拉反褶积等磁增强技术改进地下结构和磁源的探测。此外，利用磁化矢量反演（MVI）对矿化相关磁体进行了精细圈定。综合分析表明，NE向断裂和nw向断裂是影响铜矿化的主要构造格架。地表附近的磁低被解释为热液蚀变带，其上覆的磁高与富磁铁矿和潜在含硫化物带有关。欧拉反褶积和MVI的结合应用产生了与已知地质特征相关性良好的一致结果，提高了地下解释，减少了矿化体建模的不确定性。综上所述，研究结果证明了先进地球物理技术与地质构造数据相结合在铜矿勘查中的有效性。提出的工作流程提高了解释的可信度，支持目标描述，并为该地区未来的勘探提供了一个强大的框架。
{"title":"Geophysical insights into copper deposits at Mina Seival, Caçapava do Sul, Brazil: 3D magnetic inversions and euler deconvolution","authors":"Marieli Machado Zago , Maximilian Fries","doi":"10.1016/j.jappgeo.2026.106096","DOIUrl":"10.1016/j.jappgeo.2026.106096","url":null,"abstract":"<div><div>Copper deposits are critical resources for modern industries, particularly in the transition toward clean energy technologies, electric vehicles, and digital infrastructure. In southern Brazil, the Lavras do Sul–Caçapava do Sul region represents a metallogenic province that has been extensively studied since the nineteenth century, hosting significant copper and gold occurrences. These deposits are commonly associated with volcanic rocks of the Hilário Formation, which play a central role in the regional mineralization processes. Although structural controls and hydrothermal alteration patterns have been previously documented, the three-dimensional geometry and connectivity of mineralized zones at depth remain insufficiently constrained. This study investigates the geophysical signature of copper mineralization within the Hilário Formation using 3D inversion of aeromagnetic data integrated with structural and geological information. Magnetic enhancement techniques such as the Tilt-angle derivative, Analytic Signal (AS), and Euler Deconvolution were applied to improve the detection of subsurface structures and magnetic sources. Additionally, Magnetization Vector Inversion (MVI) was employed to refine the delineation of magnetic bodies associated with mineralization. The integrated analysis revealed NE- and NW-trending fault systems as the dominant structural frameworks influencing copper mineralization. Magnetic lows near the surface, interpreted as hydrothermal alteration zones, were found overlying deeper magnetic highs related to magnetite-rich and potentially sulfide-bearing zones. The combined application of Euler Deconvolution and MVI produced consistent results that correlate well with known geological features, improving subsurface interpretation and reducing uncertainty in the modeling of mineralized bodies. Overall, the results demonstrate the effectiveness of integrating advanced geophysical techniques with geological and structural datasets for copper exploration. The proposed workflow enhances interpretive confidence, supports target delineation, and provides a robust framework for future exploration in the region.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"246 ","pages":"Article 106096"},"PeriodicalIF":2.1,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146038392","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 : 2026-01-12DOI: 10.1016/j.jappgeo.2026.106095
Faxuan Wu, Yang Li, Zhenwu Fu, Bo Han, Yong Chen
Elastic full-waveform inversion (EFWI) can provide high-resolution subsurface structures and physical properties by iteratively matching observed and synthetic data. However, the success of EFWI relies on the availability of a good initial model and high signal-to-noise ratio observed data with sufficient low-frequency information, both of which are often challenging to obtain in practical applications. In addition, the coupling of different parameters degrades the inversion result. Recently, inversion methods based on physics-informed deep neural networks (DNN) have proven effective in mitigating the issue of multiple local minima caused by inaccurate initial models, missing low-frequency information, and noisy seismic data. However, existing DNN-based approaches commonly rely on fixed activation functions (e.g., rectified linear unit). In addition, their capacity to represent high-frequency components – namely, fine-scale structural details – is inherently limited due to spectral bias. These limitations may, in turn, impede their broader applicability. To mitigate this issue, we propose a model reparameterized EFWI method based on a dual-channel convolutional neural network (CNN) and Kolmogorov–Arnold networks (KAN) to enhance the reconstruction of fine-scale structural details. Specifically, our network incorporates KAN into the U-Net architecture, where CNN and KAN operate in dual channels to efficiently capture nonlinear relationships in the data. The hybrid network maps an initial model to the subsurface parameter model, with the output of the network serving as input for partial differential equations (PDEs) to generate synthetic data. Various numerical examples are conducted to investigate the performance of the inversion method, including its ability to mitigate the parameter crosstalk issue, the effect of noise and missing low-frequency information, and the influence of different initial models and network inputs. The numerical results demonstrate that, by combining CNN’s fixed activation functions with KAN’s inherently learnable activations, our method – despite a modest increase in computational cost – outperforms both EFWI and CNN-based reparameterized EFWI in reconstruction accuracy and convergence efficiency.
{"title":"High-resolution elastic full-waveform inversion using dual-channel CNN and Kolmogorov–Arnold network","authors":"Faxuan Wu, Yang Li, Zhenwu Fu, Bo Han, Yong Chen","doi":"10.1016/j.jappgeo.2026.106095","DOIUrl":"10.1016/j.jappgeo.2026.106095","url":null,"abstract":"<div><div>Elastic full-waveform inversion (EFWI) can provide high-resolution subsurface structures and physical properties by iteratively matching observed and synthetic data. However, the success of EFWI relies on the availability of a good initial model and high signal-to-noise ratio observed data with sufficient low-frequency information, both of which are often challenging to obtain in practical applications. In addition, the coupling of different parameters degrades the inversion result. Recently, inversion methods based on physics-informed deep neural networks (DNN) have proven effective in mitigating the issue of multiple local minima caused by inaccurate initial models, missing low-frequency information, and noisy seismic data. However, existing DNN-based approaches commonly rely on fixed activation functions (e.g., rectified linear unit). In addition, their capacity to represent high-frequency components – namely, fine-scale structural details – is inherently limited due to spectral bias. These limitations may, in turn, impede their broader applicability. To mitigate this issue, we propose a model reparameterized EFWI method based on a dual-channel convolutional neural network (CNN) and Kolmogorov–Arnold networks (KAN) to enhance the reconstruction of fine-scale structural details. Specifically, our network incorporates KAN into the U-Net architecture, where CNN and KAN operate in dual channels to efficiently capture nonlinear relationships in the data. The hybrid network maps an initial model to the subsurface parameter model, with the output of the network serving as input for partial differential equations (PDEs) to generate synthetic data. Various numerical examples are conducted to investigate the performance of the inversion method, including its ability to mitigate the parameter crosstalk issue, the effect of noise and missing low-frequency information, and the influence of different initial models and network inputs. The numerical results demonstrate that, by combining CNN’s fixed activation functions with KAN’s inherently learnable activations, our method – despite a modest increase in computational cost – outperforms both EFWI and CNN-based reparameterized EFWI in reconstruction accuracy and convergence efficiency.</div></div>","PeriodicalId":54882,"journal":{"name":"Journal of Applied Geophysics","volume":"246 ","pages":"Article 106095"},"PeriodicalIF":2.1,"publicationDate":"2026-01-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145980131","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号