Pub Date : 2025-09-01DOI: 10.1016/j.acags.2025.100287
Xin Liu , Xingyu Zhu , Xupeng He , Yuzhu Wang
This paper proposes an effective approach for automatically building the fault model based on the 3D seismic images via two steps of automatic seismic fault detection and fault surface construction. Automatic seismic fault detection is performed to automatically classify the seismic image into two phases of fault and background using a slightly revised deeplabv3_resnet50 architecture with pretrained parameters provided by PyTorch. The output of the automatic seismic fault detection is a binary image contains fault and background, where one fault may be separated into different fault segments, or several faults are connected with each other which need further distinguish. To reassemble these detected fault segments and construct the fault surface model, four steps are implemented including:1) a morphological workflow is used to separate all connected faults into separated fault segments; 2) the moving least square (MLS) method is used to fit each fault segments as a smooth, one-voxel thickness surface; 3) the weighted principle component analysis (WPCA) method is applied to calculate the normal vector of each surface voxel to judge whether two or more adjacent segments should be combined in one fault surface; 4) MLS method is applied again to fit all surface segments from one fault as an unique fault surface. The final output of the proposed method provides a fault model with well-defined, cleanly separated, labeled fault surfaces that is competent for structure modelling.
{"title":"Automatic seismic fault detection and surface construction","authors":"Xin Liu , Xingyu Zhu , Xupeng He , Yuzhu Wang","doi":"10.1016/j.acags.2025.100287","DOIUrl":"10.1016/j.acags.2025.100287","url":null,"abstract":"<div><div>This paper proposes an effective approach for automatically building the fault model based on the 3D seismic images via two steps of automatic seismic fault detection and fault surface construction. Automatic seismic fault detection is performed to automatically classify the seismic image into two phases of fault and background using a slightly revised deeplabv3_resnet50 architecture with pretrained parameters provided by PyTorch. The output of the automatic seismic fault detection is a binary image contains fault and background, where one fault may be separated into different fault segments, or several faults are connected with each other which need further distinguish. To reassemble these detected fault segments and construct the fault surface model, four steps are implemented including:1) a morphological workflow is used to separate all connected faults into separated fault segments; 2) the moving least square (MLS) method is used to fit each fault segments as a smooth, one-voxel thickness surface; 3) the weighted principle component analysis (WPCA) method is applied to calculate the normal vector of each surface voxel to judge whether two or more adjacent segments should be combined in one fault surface; 4) MLS method is applied again to fit all surface segments from one fault as an unique fault surface. The final output of the proposed method provides a fault model with well-defined, cleanly separated, labeled fault surfaces that is competent for structure modelling.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"27 ","pages":"Article 100287"},"PeriodicalIF":3.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145044135","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1016/j.acags.2025.100289
Birhan Getachew Tikuye, Ram Lakhan Ray
Forest ecosystems play a pivotal role in global carbon sequestration, serving as essential carbon sinks for climate change mitigation, while also providing a range of ecosystem services such as seed dispersal, pollination, pest control, and habitat provisioning. This study aimed to estimate aboveground biomass density (AGBD) using environmental covariates and a machine learning approach from the Global Ecosystem Dynamics Investigation Light Detection And Ranging (GEDI-LiDAR) in the Lower Brazos River Watershed, Texas, USA. Specifically, GEDI Level 4A data from the National Aeronautics and Space Administration (NASA) Oak Ridge National Laboratory Distributed Active Archive Center (ORNL DAAC) was integrated with Landsat-9 Operational Land Imagery (OLI) and Shuttle Radar Topographic Mission (SRTM) data to enhance predictive accuracy for AGBD. Spectral indices, including the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI), were derived from Landsat 9 to support AGBD prediction. Three machine learning models, such as Multivariate Adaptive Regression Splines (MARS), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost), were deployed, with performance assessed using the coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE). Among the models, XGBoost achieved the highest predictive accuracy (R2 = 0.43, RMSE = 31.03, MAE = 22.49). The modelling indicated that longitude, latitude, moisture stress indices (MSI), and digital elevation model (DEM) are among the critical predictors for AGBD. The mean AGBD across the watershed was estimated at 72.3 Mg ha-1, corresponding to a total biomass of approximately 66.6 million tons. Evergreen forests showed the highest AGBD values at 110 Mg ha-1, while cultivated lands averaged 33 Mg ha-1. These findings highlight the effectiveness of integrating environmental covariates with machine learning to estimate AGBD from GEDI LiDAR across diverse ecosystems. This approach provides a robust tool for advancing carbon management and climate change mitigation efforts, while also supporting data-driven conservation planning in both forested and agricultural landscapes.
{"title":"Estimating aboveground biomass using environmental covariates and a machine-learning approach in the Lower Brazos River Basin, Texas, USA","authors":"Birhan Getachew Tikuye, Ram Lakhan Ray","doi":"10.1016/j.acags.2025.100289","DOIUrl":"10.1016/j.acags.2025.100289","url":null,"abstract":"<div><div>Forest ecosystems play a pivotal role in global carbon sequestration, serving as essential carbon sinks for climate change mitigation, while also providing a range of ecosystem services such as seed dispersal, pollination, pest control, and habitat provisioning. This study aimed to estimate aboveground biomass density (AGBD) using environmental covariates and a machine learning approach from the Global Ecosystem Dynamics Investigation Light Detection And Ranging (GEDI-LiDAR) in the Lower Brazos River Watershed, Texas, USA. Specifically, GEDI Level 4A data from the National Aeronautics and Space Administration (NASA) Oak Ridge National Laboratory Distributed Active Archive Center (ORNL DAAC) was integrated with Landsat-9 Operational Land Imagery (OLI) and Shuttle Radar Topographic Mission (SRTM) data to enhance predictive accuracy for AGBD. Spectral indices, including the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI), were derived from Landsat 9 to support AGBD prediction. Three machine learning models, such as Multivariate Adaptive Regression Splines (MARS), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost), were deployed, with performance assessed using the coefficient of determination (R<sup>2</sup>), root mean square error (RMSE), and mean absolute error (MAE). Among the models, XGBoost achieved the highest predictive accuracy (R<sup>2</sup> = 0.43, RMSE = 31.03, MAE = 22.49). The modelling indicated that longitude, latitude, moisture stress indices (MSI), and digital elevation model (DEM) are among the critical predictors for AGBD. The mean AGBD across the watershed was estimated at 72.3 Mg ha<sup>-1</sup>, corresponding to a total biomass of approximately 66.6 million tons. Evergreen forests showed the highest AGBD values at 110 Mg ha<sup>-1</sup>, while cultivated lands averaged 33 Mg ha<sup>-1</sup>. These findings highlight the effectiveness of integrating environmental covariates with machine learning to estimate AGBD from GEDI LiDAR across diverse ecosystems. This approach provides a robust tool for advancing carbon management and climate change mitigation efforts, while also supporting data-driven conservation planning in both forested and agricultural landscapes.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"27 ","pages":"Article 100289"},"PeriodicalIF":3.2,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145104692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This study presents the first comprehensive evaluation in West Africa of four imputation methods, Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGB), and Ordinary Kriging (OK), applied to six core meteorological variables across Senegal over a ten-year period (2015–2024). By simulating realistic missing data scenarios informed by field conditions (e.g., power outages, observer absences, sensor failures), it establishes a robust benchmark for climate data reconstruction using machine learning in resource-constrained settings.
The findings highlight the clear superiority of ensemble learning approaches. XGB consistently outperformed all methods across variables and scenarios, achieving the highest average predictive accuracy with R2 values up to [95 % CI: 0.82–0.88], along with lower Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). RF yielded comparable performance, especially for maximum and minimum temperature (TMAX, TMIN), maintaining strong stability even at 20 % missingness. In contrast, DT performance declined sharply with increased data loss, and OK was constrained by the sparse spatial distribution of meteorological stations, notably impairing its ability to impute precipitation (PRCP) and wind speed (WDSP).
This work contributes a multivariable imputation framework specifically adapted to West African climatic and infrastructural realities. It also integrates block bootstrap methods to quantify uncertainty and derive 95 % confidence intervals for all error metrics. Results confirm that imputation effectiveness is highly variable-dependent: continuous and temporally autocorrelated variables (TMAX, TMIN, dew point temperature — DEWP) are well reconstructed, whereas discontinuous or noisy variables (WDSP and PRCP) remain challenging.
Although ensemble models offer clear advantages, their computational demands and need for hyperparameter tuning may limit real-time implementation in low-resource national meteorological services. Furthermore, the exclusion of satellite or reanalysis inputs may constrain model generalizability.
Ultimately, this study reinforces the role of advanced machine learning methods in improving climate data completeness and reliability in Africa. Although not a substitute for direct observations, imputation emerges as a critical complementary tool to support robust and resilient climate information systems essential for agriculture, public health, and disaster risk management under intensifying climate variability.
{"title":"Machine learning approaches for imputing missing meteorological data in Senegal","authors":"Mory Toure , Nana Ama Browne Klutse , Mamadou Adama Sarr , Md Abul Ehsan Bhuiyan , Annine Duclaire Kenne , Wassila Mamadou Thiaw , Daouda Badiane , Amadou Thierno Gaye , Ousmane Ndiaye , Cheikh Mbow","doi":"10.1016/j.acags.2025.100281","DOIUrl":"10.1016/j.acags.2025.100281","url":null,"abstract":"<div><div>This study presents the first comprehensive evaluation in West Africa of four imputation methods, Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGB), and Ordinary Kriging (OK), applied to six core meteorological variables across Senegal over a ten-year period (2015–2024). By simulating realistic missing data scenarios informed by field conditions (e.g., power outages, observer absences, sensor failures), it establishes a robust benchmark for climate data reconstruction using machine learning in resource-constrained settings.</div><div>The findings highlight the clear superiority of ensemble learning approaches. XGB consistently outperformed all methods across variables and scenarios, achieving the highest average predictive accuracy with R<sup>2</sup> values up to [95 % CI: 0.82–0.88], along with lower Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). RF yielded comparable performance, especially for maximum and minimum temperature (TMAX, TMIN), maintaining strong stability even at 20 % missingness. In contrast, DT performance declined sharply with increased data loss, and OK was constrained by the sparse spatial distribution of meteorological stations, notably impairing its ability to impute precipitation (PRCP) and wind speed (WDSP).</div><div>This work contributes a multivariable imputation framework specifically adapted to West African climatic and infrastructural realities. It also integrates block bootstrap methods to quantify uncertainty and derive 95 % confidence intervals for all error metrics. Results confirm that imputation effectiveness is highly variable-dependent: continuous and temporally autocorrelated variables (TMAX, TMIN, dew point temperature — DEWP) are well reconstructed, whereas discontinuous or noisy variables (WDSP and PRCP) remain challenging.</div><div>Although ensemble models offer clear advantages, their computational demands and need for hyperparameter tuning may limit real-time implementation in low-resource national meteorological services. Furthermore, the exclusion of satellite or reanalysis inputs may constrain model generalizability.</div><div>Ultimately, this study reinforces the role of advanced machine learning methods in improving climate data completeness and reliability in Africa. Although not a substitute for direct observations, imputation emerges as a critical complementary tool to support robust and resilient climate information systems essential for agriculture, public health, and disaster risk management under intensifying climate variability.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"27 ","pages":"Article 100281"},"PeriodicalIF":3.2,"publicationDate":"2025-08-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144885417","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-13DOI: 10.1016/j.acags.2025.100278
Yuyang Song , Lina Hao , Weile Li
Natural hazards such as landslides pose significant geological threats that can severely endanger the safety and property of residents in affected areas. Therefore, the prompt detection and accurate localisation of landslides are crucial. With the advancement of remote sensing technology and computational methods, artificial intelligence (AI)-based landslide detection techniques have emerged as effective solutions. Compared to traditional methods, these AI-driven approaches offer enhanced efficiency, accuracy and reliability, improving the speed and precision of landslide detection. They also provide valuable data for disaster prevention, mitigation and the assessment of landslide susceptibility and hazard levels. This study focuses on the western Sichuan region and constructs a historical landslide dataset using Google Earth imagery, which includes 4280 landslide samples (3424 for training and 856 for validation). To augment the dataset, 11 data augmentation techniques were applied, including copy–paste, random horizontal flipping, mosaic, random rotation, random hue, saturation and value transformation, affine transformation, random Gaussian noise, random scaling, random brightness and contrast adjustment, mixup and random cropping. These methods improve the diversity of landslide data, helping deep learning models capture more comprehensive global and local information during optimisation. This research utilises the YOLOv10-n object detection framework, enhanced with RepBlock from EfficientRep, FusedMBConv and MBConv techniques derived from EfficientNetV2, CSCGhostblockv2 from GhostNetv2, CReToNeXt from Damo-YOLO and CSCFocalNeXt. These innovations explore the impact of different backbone architectures on model performance. Additionally, the model incorporates four distinct attention mechanisms—convolutional block attention module (CBAM), global attention mechanism(GAM), sim attention module(SimAM) and selective kernel(SK) attention—to assess their influence on detection accuracy. The detection heads are optimised by substituting with three alternatives—DynamicHead, adaptive spatial feature fusion and real-time detection transformer—to enhance feature integration and investigate their effect on model performance. The results indicate that combining EfficientNetV2 with CBAM and v10Detect yields the highest performance. When applied to the historical landslide dataset from the western Sichuan region, the YOLO-EfficientNetV2 model achieves an average precision of 0.861 and an F1 score of 0.82, with a model size of 5.54 M. This model demonstrates superior capability in accurately identifying landslide locations, addressing the common challenge of balancing detection precision and speed in traditional object detection models, while also reducing parameter size and increasing detection speed.
{"title":"Landslide detection using deep learning on remotely sensed images","authors":"Yuyang Song , Lina Hao , Weile Li","doi":"10.1016/j.acags.2025.100278","DOIUrl":"10.1016/j.acags.2025.100278","url":null,"abstract":"<div><div>Natural hazards such as landslides pose significant geological threats that can severely endanger the safety and property of residents in affected areas. Therefore, the prompt detection and accurate localisation of landslides are crucial. With the advancement of remote sensing technology and computational methods, artificial intelligence (AI)-based landslide detection techniques have emerged as effective solutions. Compared to traditional methods, these AI-driven approaches offer enhanced efficiency, accuracy and reliability, improving the speed and precision of landslide detection. They also provide valuable data for disaster prevention, mitigation and the assessment of landslide susceptibility and hazard levels. This study focuses on the western Sichuan region and constructs a historical landslide dataset using Google Earth imagery, which includes 4280 landslide samples (3424 for training and 856 for validation). To augment the dataset, 11 data augmentation techniques were applied, including copy–paste, random horizontal flipping, mosaic, random rotation, random hue, saturation and value transformation, affine transformation, random Gaussian noise, random scaling, random brightness and contrast adjustment, mixup and random cropping. These methods improve the diversity of landslide data, helping deep learning models capture more comprehensive global and local information during optimisation. This research utilises the YOLOv10-n object detection framework, enhanced with RepBlock from EfficientRep, FusedMBConv and MBConv techniques derived from EfficientNetV2, CSCGhostblockv2 from GhostNetv2, CReToNeXt from Damo-YOLO and CSCFocalNeXt. These innovations explore the impact of different backbone architectures on model performance. Additionally, the model incorporates four distinct attention mechanisms—convolutional block attention module (CBAM), global attention mechanism(GAM), sim attention module(SimAM) and selective kernel(SK) attention—to assess their influence on detection accuracy. The detection heads are optimised by substituting with three alternatives—DynamicHead, adaptive spatial feature fusion and real-time detection transformer—to enhance feature integration and investigate their effect on model performance. The results indicate that combining EfficientNetV2 with CBAM and v10Detect yields the highest performance. When applied to the historical landslide dataset from the western Sichuan region, the YOLO-EfficientNetV2 model achieves an average precision of 0.861 and an F<sub>1</sub> score of 0.82, with a model size of 5.54 M. This model demonstrates superior capability in accurately identifying landslide locations, addressing the common challenge of balancing detection precision and speed in traditional object detection models, while also reducing parameter size and increasing detection speed.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"27 ","pages":"Article 100278"},"PeriodicalIF":3.2,"publicationDate":"2025-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144917695","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-13DOI: 10.1016/j.acags.2025.100282
Rached M. Rached , Hussain AlBahrani , Timothy E. Moellendick , J. Carlos Santamarina , Thomas Finkbeiner
A thorough understanding of stress distribution around wellbores is crucial for maintaining wellbore stability, especially in deep wells with complex trajectories and subsurface formations exhibiting coupled mechanical behaviors. This study introduces a new finite-element-based modular simulator designed to address a wide range of challenging drilling and boundary conditions, including the presence or absence of filter cake, high over-pressure, inhomogeneous and anisotropic formations, non-linear constitutive behavior, and deviated wells. The simulator uses finite element modeling to provide accurate stress predictions without the overly conservative assumptions common in existing commercial tools. Each module is pre-tested and validated against published analytical solutions and features a user-friendly interface with minimal input requirements, allowing for quick and robust simulations in both 2D and 3D configurations. The simulator can analyze various phenomena, including time-dependent pore pressure diffusion, temperature-induced stress variations, and the impact of heterogeneous formations and layering on stress concentrations. All pre-tested modules run in <60 s on a mid-range workstation while matching analytical solutions to within 0.2 %. We present several case studies that demonstrate the simulator's advantages over existing commercial tools, with all modules made openly available to facilitate broader application.
{"title":"Borehole integrity evaluation utilizing coupled hydraulic thermal and mechanical analyses in robust and pre-optimized finite element simulator","authors":"Rached M. Rached , Hussain AlBahrani , Timothy E. Moellendick , J. Carlos Santamarina , Thomas Finkbeiner","doi":"10.1016/j.acags.2025.100282","DOIUrl":"10.1016/j.acags.2025.100282","url":null,"abstract":"<div><div>A thorough understanding of stress distribution around wellbores is crucial for maintaining wellbore stability, especially in deep wells with complex trajectories and subsurface formations exhibiting coupled mechanical behaviors. This study introduces a new finite-element-based modular simulator designed to address a wide range of challenging drilling and boundary conditions, including the presence or absence of filter cake, high over-pressure, inhomogeneous and anisotropic formations, non-linear constitutive behavior, and deviated wells. The simulator uses finite element modeling to provide accurate stress predictions without the overly conservative assumptions common in existing commercial tools. Each module is pre-tested and validated against published analytical solutions and features a user-friendly interface with minimal input requirements, allowing for quick and robust simulations in both 2D and 3D configurations. The simulator can analyze various phenomena, including time-dependent pore pressure diffusion, temperature-induced stress variations, and the impact of heterogeneous formations and layering on stress concentrations. All pre-tested modules run in <60 s on a mid-range workstation while matching analytical solutions to within 0.2 %. We present several case studies that demonstrate the simulator's advantages over existing commercial tools, with all modules made openly available to facilitate broader application.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"27 ","pages":"Article 100282"},"PeriodicalIF":3.2,"publicationDate":"2025-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144865563","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-09DOI: 10.1016/j.acags.2025.100280
Lennart Siemann, Ramiro Relanez
The presented study investigates the prediction of buried paleo-channels for probabilistic ground modeling of offshore windfarm development areas using geostatistical methods. These channels, common in glaciogenic regions like the North Sea, can pose significant geohazards affecting turbine foundation stability. Conventional 2D seismic data interpretation provides the best estimate of the position but lacks probabilistic assessment, specifically at unexplored locations. Multiple-point statistics (MPS) and sequential indicator simulation (SIS) are applied to quantify the probability of channel features from seismic data, away from seismic lines. MPS utilizes training images to capture complex spatial structures, while SIS relies on variogram models for modeling spatial variability. Results demonstrate that denser seismic line spacing (150 m) yields higher accuracy compared to wider spacings (300 m and 600 m), underscoring the importance of data density in offshore subsurface site characterization. Additionally, the findings indicate that MPS provides lower errors, making it preferable for precise channel location prediction. The selected training image did not have a major impact on the outcome on the tested data. Conversely, SIS offers broader coverage of potential channel locations, which may be advantageous for further de-risking. This research contributes to more informed ground modeling by incorporating probabilistic approaches. Therefore, it supports in offshore wind farm site development by enhancing knowledge of the subsurface at an early stage of wind farm development to aid decisions in windfarm and further site investigation planning.
{"title":"Assessing paleo channel probability for offshore wind farm ground modeling - comparison of multiple-point statistics and sequential indicator simulation","authors":"Lennart Siemann, Ramiro Relanez","doi":"10.1016/j.acags.2025.100280","DOIUrl":"10.1016/j.acags.2025.100280","url":null,"abstract":"<div><div>The presented study investigates the prediction of buried paleo-channels for probabilistic ground modeling of offshore windfarm development areas using geostatistical methods. These channels, common in glaciogenic regions like the North Sea, can pose significant geohazards affecting turbine foundation stability. Conventional 2D seismic data interpretation provides the best estimate of the position but lacks probabilistic assessment, specifically at unexplored locations. Multiple-point statistics (MPS) and sequential indicator simulation (SIS) are applied to quantify the probability of channel features from seismic data, away from seismic lines. MPS utilizes training images to capture complex spatial structures, while SIS relies on variogram models for modeling spatial variability. Results demonstrate that denser seismic line spacing (150 m) yields higher accuracy compared to wider spacings (300 m and 600 m), underscoring the importance of data density in offshore subsurface site characterization. Additionally, the findings indicate that MPS provides lower errors, making it preferable for precise channel location prediction. The selected training image did not have a major impact on the outcome on the tested data. Conversely, SIS offers broader coverage of potential channel locations, which may be advantageous for further de-risking. This research contributes to more informed ground modeling by incorporating probabilistic approaches. Therefore, it supports in offshore wind farm site development by enhancing knowledge of the subsurface at an early stage of wind farm development to aid decisions in windfarm and further site investigation planning.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"27 ","pages":"Article 100280"},"PeriodicalIF":3.2,"publicationDate":"2025-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144827361","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-09DOI: 10.1016/j.acags.2025.100279
Glen T. Nwaila , Musa S.D. Manzi , Emmanuel John M. Carranza , Raymond J. Durrheim , Hartwig E. Frimmel
Potholes are local depression structures that disrupt stratigraphic continuity, such as in layered igneous intrusions. In the Bushveld Complex (South Africa), potholes range from a few to hundreds of meters in width, and may disrupt orebodies, cause ore loss and pose geotechnical challenges. However, potholes are of scientific value as they are proxies of magma chamber processes that are not directly observable. Unfortunately, it is seldom possible to map the full 3D geometry of potholes directly. Reflection seismics has the potential to map many potholes indirectly. However, the accurate segmentation of potholes in seismic data remains unresolved, particularly using geodata science-based methods. Here, we present a prototype segmentation framework that: (1) uses a physics-based, forward modelling method to synthesize 3D reflection seismic data and augments the training data; and (2) implements a standard deep learning, voxel classification-based pothole detection workflow using the data generated in step (1). Both components of the framework are general enough to permit further development, for example, as deep-learning architectures evolve or as the knowledge of potholes improve. We demonstrate that a self-reinforcing feedback loop of knowledge-driven data engineering and deep learning has the potential to overcome data quality issues in supervised tasks of seismic data analysis. We apply the trained model on augmented data to 3D seismic data acquired from a platinum group element Bushveld Complex orebody and demonstrate that automated pothole prediction is practical. Furthermore, physics-based data augmentation, as opposed to inferential types, provides a realistic path to recursive data augmentation that does not incur problems caused by the use of inferential data synthesis, such as model collapse.
{"title":"Pothole detection and segmentation in the Bushveld Complex using physics-based data augmentation and deep learning","authors":"Glen T. Nwaila , Musa S.D. Manzi , Emmanuel John M. Carranza , Raymond J. Durrheim , Hartwig E. Frimmel","doi":"10.1016/j.acags.2025.100279","DOIUrl":"10.1016/j.acags.2025.100279","url":null,"abstract":"<div><div>Potholes are local depression structures that disrupt stratigraphic continuity, such as in layered igneous intrusions. In the Bushveld Complex (South Africa), potholes range from a few to hundreds of meters in width, and may disrupt orebodies, cause ore loss and pose geotechnical challenges. However, potholes are of scientific value as they are proxies of magma chamber processes that are not directly observable. Unfortunately, it is seldom possible to map the full 3D geometry of potholes directly. Reflection seismics has the potential to map many potholes indirectly. However, the accurate segmentation of potholes in seismic data remains unresolved, particularly using geodata science-based methods. Here, we present a prototype segmentation framework that: (1) uses a physics-based, forward modelling method to synthesize 3D reflection seismic data and augments the training data; and (2) implements a standard deep learning, voxel classification-based pothole detection workflow using the data generated in step (1). Both components of the framework are general enough to permit further development, for example, as deep-learning architectures evolve or as the knowledge of potholes improve. We demonstrate that a self-reinforcing feedback loop of knowledge-driven data engineering and deep learning has the potential to overcome data quality issues in supervised tasks of seismic data analysis. We apply the trained model on augmented data to 3D seismic data acquired from a platinum group element Bushveld Complex orebody and demonstrate that automated pothole prediction is practical. Furthermore, physics-based data augmentation, as opposed to inferential types, provides a realistic path to recursive data augmentation that does not incur problems caused by the use of inferential data synthesis, such as model collapse.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"27 ","pages":"Article 100279"},"PeriodicalIF":3.2,"publicationDate":"2025-08-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144828175","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-08DOI: 10.1016/j.acags.2025.100274
Margaret A. Goldman , Graham W. Lederer , Joshua M. Rosera , Garth E. Graham , Asitang Mishra , Alice Yepremyan
The U.S. Geological Survey (USGS), Defense Advanced Projects Research Agency (DARPA), Jet Propulsion Laboratory (JPL), and MITRE ran a 12-week machine learning competition aimed at accelerating development of AI tools for critical mineral assessments. The Artificial Intelligence for Critical Mineral Assessment Competition solicited innovative solutions for two challenges: 1) automated georeferencing of historical geologic and topographic maps, and 2) automated feature extraction from historical maps. Competitors used a new dataset of historical map images to train, validate, and evaluate their models. Automated georeferencing pipelines attained a median root-mean square error of 1.1 km. Prompt-based extraction (i.e., with user input) of polygons, polylines, and points from geologic maps yielded median F1 scores of 0.77, 0.56, 0.35, respectively. Geologic maps pose numerous challenges for AI workflows because they vary significantly. However, despite its short duration, the competition yielded promising results that have since spurred further innovation in this area and led to the development of new AI tools to semi-automate key, time-consuming parts of the assessment workflow.
{"title":"Extracting data from maps: Lessons learned from the artificial intelligence for critical mineral assessment competition","authors":"Margaret A. Goldman , Graham W. Lederer , Joshua M. Rosera , Garth E. Graham , Asitang Mishra , Alice Yepremyan","doi":"10.1016/j.acags.2025.100274","DOIUrl":"10.1016/j.acags.2025.100274","url":null,"abstract":"<div><div>The U.S. Geological Survey (USGS), Defense Advanced Projects Research Agency (DARPA), Jet Propulsion Laboratory (JPL), and MITRE ran a 12-week machine learning competition aimed at accelerating development of AI tools for critical mineral assessments. The Artificial Intelligence for Critical Mineral Assessment Competition solicited innovative solutions for two challenges: 1) automated georeferencing of historical geologic and topographic maps, and 2) automated feature extraction from historical maps. Competitors used a new dataset of historical map images to train, validate, and evaluate their models. Automated georeferencing pipelines attained a median root-mean square error of 1.1 km. Prompt-based extraction (i.e., with user input) of polygons, polylines, and points from geologic maps yielded median F1 scores of 0.77, 0.56, 0.35, respectively. Geologic maps pose numerous challenges for AI workflows because they vary significantly. However, despite its short duration, the competition yielded promising results that have since spurred further innovation in this area and led to the development of new AI tools to semi-automate key, time-consuming parts of the assessment workflow.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"27 ","pages":"Article 100274"},"PeriodicalIF":3.2,"publicationDate":"2025-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144852248","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-07DOI: 10.1016/j.acags.2025.100276
Tanveer Alam Munshi, Khanum Popi, Labiba Nusrat Jahan, M. Farhad Howladar, Mahamudul Hashan
Measuring rock strength using an uniaxial testing machine is destructive and costly, requiring high-quality rock samples. This work suggests an alternate approach that makes use of machine learning techniques to predict uniaxial compressive strength (UCS). The input parameters for this investigation were derived from 180 datasets containing well log variables such as resistivity (RT), sonic travel time (DT), and gamma-ray (GR), as well as rock properties like density. All these datasets came from a shaly sand reservoir in the Bengal Basin. To forecast UCS, a number of methods were used, such as multilayer perceptron (MLP), random forest (RF), support vector machine (SVM), extreme gradient boosting (XGBoost), and multiple variable regression (MVR). Additionally, a hybrid stacking model that combines these algorithms was developed. Hyperparameter optimization was conducted using grid search and genetic algorithm. A notable contribution of this study lies in the application of both grid search and genetic algorithm (GA) for hyperparameter optimization, implemented across both individual base learners and the stacking ensemble model. Regression metrics including coefficient of determination (R2), mean absolute error (MAE), root mean square error (RMSE), maximum error (MaxE), and minimum error (MinE) were used to assess the effectiveness of the models. The proposed stacking model achieved a high testing R2 of 0.9762, outperforming individual models. The methodology provided in this paper can assist engineers and researchers in quickly and precisely determining the strength of reservoir rock by using a few log features, hence decreasing the reliance on labor-intensive and time-consuming laboratory work.
{"title":"Stacking modeling with genetic algorithm-based hyperparameter tuning for uniaxial compressive strength prediction","authors":"Tanveer Alam Munshi, Khanum Popi, Labiba Nusrat Jahan, M. Farhad Howladar, Mahamudul Hashan","doi":"10.1016/j.acags.2025.100276","DOIUrl":"10.1016/j.acags.2025.100276","url":null,"abstract":"<div><div>Measuring rock strength using an uniaxial testing machine is destructive and costly, requiring high-quality rock samples. This work suggests an alternate approach that makes use of machine learning techniques to predict uniaxial compressive strength (UCS). The input parameters for this investigation were derived from 180 datasets containing well log variables such as resistivity (RT), sonic travel time (DT), and gamma-ray (GR), as well as rock properties like density. All these datasets came from a shaly sand reservoir in the Bengal Basin. To forecast UCS, a number of methods were used, such as multilayer perceptron (MLP), random forest (RF), support vector machine (SVM), extreme gradient boosting (XGBoost), and multiple variable regression (MVR). Additionally, a hybrid stacking model that combines these algorithms was developed. Hyperparameter optimization was conducted using grid search and genetic algorithm. A notable contribution of this study lies in the application of both grid search and genetic algorithm (GA) for hyperparameter optimization, implemented across both individual base learners and the stacking ensemble model. Regression metrics including coefficient of determination (R<sup>2</sup>), mean absolute error (MAE), root mean square error (RMSE), maximum error (MaxE), and minimum error (MinE) were used to assess the effectiveness of the models. The proposed stacking model achieved a high testing R<sup>2</sup> of 0.9762, outperforming individual models. The methodology provided in this paper can assist engineers and researchers in quickly and precisely determining the strength of reservoir rock by using a few log features, hence decreasing the reliance on labor-intensive and time-consuming laboratory work.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"27 ","pages":"Article 100276"},"PeriodicalIF":3.2,"publicationDate":"2025-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144809487","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-07DOI: 10.1016/j.acags.2025.100277
Rahul Kumar, Katsura Kobayashi, Christian Potiszil, Tak Kunihiro
Asteroidal materials contain organic matter (OM), which records a number of extraterrestrial environments and thus provides a record of Solar System processes. OM contain essential compounds for the origin of life. To understand the origin and evolution of OM, systematic identification and detailed observation using in-situ techniques is required. While both nm- and μm-sized OM were studied previously, only a small portion of a given sample surface was investigated in each study. Here, a novel workflow was developed and applied to identify and classify μm-sized OM on mm-sized asteroidal materials. The workflow involved image processing and machine learning, enabling a comprehensive and non-biased way of identifying, classifying, and measuring the properties of OM. We found that identifying OM is more accurate by classification with machine learning than by clustering. On the approach of classification with machine learning, five algorithms were tested. The random forest algorithm was selected as it scored the highest in 4 out of 5 accuracy parameters during evaluation. The workflow gave modal OM abundances that were consistent with those identified manually, demonstrating that the workflow can accurately identify 1-15 μm-sized OM. The size distribution of OM was modeled using the power-law distribution, giving slope α values that were consistent with fragmentation processes. The shape of the OM was quantified using circularity and solidity, giving a positive correlation and indicating these properties are closely related. Overall, the workflow enabled identification of many OM quickly and accurately and the obtainment of chemical and petrographic information. Such information can help the selection of OM for further in-situ techniques, and elucidate the origin and evolution of OM preserved in asteroidal materials.
{"title":"Development of a technique to identify μm-sized organic matter in asteroidal material: An approach using machine learning","authors":"Rahul Kumar, Katsura Kobayashi, Christian Potiszil, Tak Kunihiro","doi":"10.1016/j.acags.2025.100277","DOIUrl":"10.1016/j.acags.2025.100277","url":null,"abstract":"<div><div>Asteroidal materials contain organic matter (OM), which records a number of extraterrestrial environments and thus provides a record of Solar System processes. OM contain essential compounds for the origin of life. To understand the origin and evolution of OM, systematic identification and detailed observation using in-situ techniques is required. While both nm- and μm-sized OM were studied previously, only a small portion of a given sample surface was investigated in each study. Here, a novel workflow was developed and applied to identify and classify μm-sized OM on mm-sized asteroidal materials. The workflow involved image processing and machine learning, enabling a comprehensive and non-biased way of identifying, classifying, and measuring the properties of OM. We found that identifying OM is more accurate by classification with machine learning than by clustering. On the approach of classification with machine learning, five algorithms were tested. The random forest algorithm was selected as it scored the highest in 4 out of 5 accuracy parameters during evaluation. The workflow gave modal OM abundances that were consistent with those identified manually, demonstrating that the workflow can accurately identify 1-15 μm-sized OM. The size distribution of OM was modeled using the power-law distribution, giving slope α values that were consistent with fragmentation processes. The shape of the OM was quantified using circularity and solidity, giving a positive correlation and indicating these properties are closely related. Overall, the workflow enabled identification of many OM quickly and accurately and the obtainment of chemical and petrographic information. Such information can help the selection of OM for further in-situ techniques, and elucidate the origin and evolution of OM preserved in asteroidal materials.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"27 ","pages":"Article 100277"},"PeriodicalIF":3.2,"publicationDate":"2025-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144826846","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","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