Pub Date : 2025-12-01DOI: 10.1016/j.acags.2025.100311
Jiyin Zhang, Wenjia Li, Xiang Que, Weilin Chen, Chenhao Li, Xiaogang Ma
With the recent integration of Large Language Models (LLMs) into geoscience applications, agentic LLM-driven workflows have emerged as an innovative approach to streamline automated data analysis processes. Advanced proprietary LLMs like ChatGPT demonstrate strong performance in customized workflows due to their substantial computational resources and extensive pretraining on diverse datasets. However, deploying such workflows with commercial LLMs can incur significant costs, especially in terms of token consumption, necessitating a shift toward open-source models. In this study, we fine-tuned an open-source LLM (Llama 3.1) to handle geoscience data analysis tasks, leveraging the self-instruct method to generate synthetic training datasets. The proposed pipeline for designing LLM-driven workflows and fine-tuning open-source models using synthetic datasets enables scalability, allowing the integration of additional LLM agents to accommodate more complex tasks. Furthermore, this workflow serves as a template for researchers in other domains to develop similar solutions tailored to their specific needs. Our experimental evaluation compares the performance of ChatGPT-4o with the fine-tuned Llama 3.1 in the context of the proposed geoscience data analysis workflow. Results demonstrate that the fine-tuned open-source model achieves performance comparable to proprietary models, extending the applicability of open LLMs to domain-specific agentic workflows in data analysis.
{"title":"Fine-tuning small and open LLMs to automate geoscience data analysis workflows: A scalable approach","authors":"Jiyin Zhang, Wenjia Li, Xiang Que, Weilin Chen, Chenhao Li, Xiaogang Ma","doi":"10.1016/j.acags.2025.100311","DOIUrl":"10.1016/j.acags.2025.100311","url":null,"abstract":"<div><div>With the recent integration of Large Language Models (LLMs) into geoscience applications, agentic LLM-driven workflows have emerged as an innovative approach to streamline automated data analysis processes. Advanced proprietary LLMs like ChatGPT demonstrate strong performance in customized workflows due to their substantial computational resources and extensive pretraining on diverse datasets. However, deploying such workflows with commercial LLMs can incur significant costs, especially in terms of token consumption, necessitating a shift toward open-source models. In this study, we fine-tuned an open-source LLM (Llama 3.1) to handle geoscience data analysis tasks, leveraging the self-instruct method to generate synthetic training datasets. The proposed pipeline for designing LLM-driven workflows and fine-tuning open-source models using synthetic datasets enables scalability, allowing the integration of additional LLM agents to accommodate more complex tasks. Furthermore, this workflow serves as a template for researchers in other domains to develop similar solutions tailored to their specific needs. Our experimental evaluation compares the performance of ChatGPT-4o with the fine-tuned Llama 3.1 in the context of the proposed geoscience data analysis workflow. Results demonstrate that the fine-tuned open-source model achieves performance comparable to proprietary models, extending the applicability of open LLMs to domain-specific agentic workflows in data analysis.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"28 ","pages":"Article 100311"},"PeriodicalIF":3.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145690944","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 : 2025-11-14DOI: 10.1016/j.acags.2025.100306
Monir Modjarrad , Hadi Mostafavi Amjad , Mahrokh G. Shayesteh , Amin Danandeh Hesar
This study demonstrates the evolution from traditional geoscientific methods to data-driven approaches for determining the tectonic setting of peridotites using their chemical composition. It introduces an application of Machine learning (ML) methods, including unsupervised learning and dimensionality reduction, to identify the tectonic setting of peridotite samples using their mineral chemistry. Integrating mineral chemistry analysis with ML overcomes inefficiencies in processing the vast datasets generated by electron probe micro-analyzers (EPMA).
Peridotites from the abyssal (oceanic lithosphere) and fore-arc (subduction) zones provide important insights into the composition and dynamics of the upper mantle, such as melt generation and mantle convection processes, interactions between tectonic plates, and oceanic crust generation. This study evaluates the efficacy of established mineral chemistry diagrams in discerning the tectonic settings of peridotite samples from ophiolitic fragments. Initial evaluations revealed that several conventional diagrams could classify abyssal and fore-arc peridotites, but significant misclassifications occurred, necessitating improved methods. We propose new diagrams utilizing spinel and olivine compositions against the pyroxenes' Al2O3 content, showing enhanced discrimination capabilities. Further, an ML approach was implemented, successfully categorizing abyssal and fore-arc settings with high accuracy (92–96 % based on a 20 % test subset) using olivine and spinel compositional data, outperforming traditional methods. This work highlights the value of ML workflows in supporting petrological analysis for accurate tectonic setting classification.
{"title":"Machine learning approaches in testing and enhancing tectonic setting discrimination of peridotites","authors":"Monir Modjarrad , Hadi Mostafavi Amjad , Mahrokh G. Shayesteh , Amin Danandeh Hesar","doi":"10.1016/j.acags.2025.100306","DOIUrl":"10.1016/j.acags.2025.100306","url":null,"abstract":"<div><div>This study demonstrates the evolution from traditional geoscientific methods to data-driven approaches for determining the tectonic setting of peridotites using their chemical composition. It introduces an application of Machine learning (ML) methods, including unsupervised learning and dimensionality reduction, to identify the tectonic setting of peridotite samples using their mineral chemistry. Integrating mineral chemistry analysis with ML overcomes inefficiencies in processing the vast datasets generated by electron probe micro-analyzers (EPMA).</div><div>Peridotites from the abyssal (oceanic lithosphere) and fore-arc (subduction) zones provide important insights into the composition and dynamics of the upper mantle, such as melt generation and mantle convection processes, interactions between tectonic plates, and oceanic crust generation. This study evaluates the efficacy of established mineral chemistry diagrams in discerning the tectonic settings of peridotite samples from ophiolitic fragments. Initial evaluations revealed that several conventional diagrams could classify abyssal and fore-arc peridotites, but significant misclassifications occurred, necessitating improved methods. We propose new diagrams utilizing spinel and olivine compositions against the pyroxenes' Al<sub>2</sub>O<sub>3</sub> content, showing enhanced discrimination capabilities. Further, an ML approach was implemented, successfully categorizing abyssal and fore-arc settings with high accuracy (92–96 % based on a 20 % test subset) using olivine and spinel compositional data, outperforming traditional methods. This work highlights the value of ML workflows in supporting petrological analysis for accurate tectonic setting classification.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"28 ","pages":"Article 100306"},"PeriodicalIF":3.2,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145575991","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}
Object detection in remote sensing imagery is a fundamental task in Earth observation applications. However, its accuracy faces significant challenges due to the influence of complex backgrounds and multi-scale object features. Despite advancements in the methods utilizing convolutional neural networks (CNNs) and self-attention, they encounter two fundamental challenges. CNNs, constrained by their local receptive fields, struggle to capture adequate global feature representations. Conversely, while self-attention excels at modeling long-range dependencies, its high computational complexity hinders practical application on high-resolution remote sensing imagery and can degrade the representation of local details. To resolve these challenges, this article proposes a novel detection model named MambaRetinaNet, which fuses multi-scale convolution with the Mamba architecture. We designed a Synergistic Perception Module (SPM) to efficiently model global information while enhancing local feature retention. Furthermore, we introduce an Asymmetric Feature Pyramid (MambaFPN), which optimizes the feature pyramid network through a differentiated processing strategy to achieve a balance between detection accuracy and computational efficiency. The experimental results indicate that MambaRetinaNet demonstrates significant advantages on four benchmark remote sensing datasets. Specifically, the mean Average Precision (mAP) on DOTA-v1.0, DOTA-v1.5, DOTA-v2.0, and DIOR-R are 80.03, 70.21, 57.17, and 71.50, respectively — representing an average improvement of approximately 11 over the baseline. While introducing a slight increase in the number of parameters, MambaRetinaNet nonetheless substantially reduces the computational cost, lowering the FLOPs by nearly three times compared to the baseline, and maintains the lowest FLOPs among competing methods. These results highlight the effectiveness and practical value of the proposed method in complex remote sensing scenarios.
{"title":"MambaRetinaNet: Improving remote sensing object detection by fusing Mamba and multi-scale convolution","authors":"Junjie Chen , Jieru Wei , Gang Wu, Jichang Yang, Jiandong Shang, Hengliang Guo, Dujuan Zhang, Shengguang Zhu","doi":"10.1016/j.acags.2025.100305","DOIUrl":"10.1016/j.acags.2025.100305","url":null,"abstract":"<div><div>Object detection in remote sensing imagery is a fundamental task in Earth observation applications. However, its accuracy faces significant challenges due to the influence of complex backgrounds and multi-scale object features. Despite advancements in the methods utilizing convolutional neural networks (CNNs) and self-attention, they encounter two fundamental challenges. CNNs, constrained by their local receptive fields, struggle to capture adequate global feature representations. Conversely, while self-attention excels at modeling long-range dependencies, its high computational complexity hinders practical application on high-resolution remote sensing imagery and can degrade the representation of local details. To resolve these challenges, this article proposes a novel detection model named MambaRetinaNet, which fuses multi-scale convolution with the Mamba architecture. We designed a Synergistic Perception Module (SPM) to efficiently model global information while enhancing local feature retention. Furthermore, we introduce an Asymmetric Feature Pyramid (MambaFPN), which optimizes the feature pyramid network through a differentiated processing strategy to achieve a balance between detection accuracy and computational efficiency. The experimental results indicate that MambaRetinaNet demonstrates significant advantages on four benchmark remote sensing datasets. Specifically, the mean Average Precision (mAP) on DOTA-v1.0, DOTA-v1.5, DOTA-v2.0, and DIOR-R are 80.03, 70.21, 57.17, and 71.50, respectively — representing an average improvement of approximately 11 over the baseline. While introducing a slight increase in the number of parameters, MambaRetinaNet nonetheless substantially reduces the computational cost, lowering the FLOPs by nearly three times compared to the baseline, and maintains the lowest FLOPs among competing methods. These results highlight the effectiveness and practical value of the proposed method in complex remote sensing scenarios.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"28 ","pages":"Article 100305"},"PeriodicalIF":3.2,"publicationDate":"2025-11-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145525345","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 : 2025-11-04DOI: 10.1016/j.acags.2025.100303
Gilbert Jesse , Cyril D. Boateng , Jeffrey N.A. Aryee , Marian A. Osei , David D. Wemegah , Solomon S.R. Gidigasu , Akyana Britwum , Samuel K. Afful , Haoulata Touré , Vera Mensah , Prinsca Owusu-Afriyie
This study presents a comprehensive synthesis of machine learning (ML) techniques applied to groundwater level (GWL) prediction, focusing on model architectures, feature selection methods, hyperparameter tuning, optimization algorithms, and clustering techniques. A total of 223 peer-reviewed articles were systematically reviewed using the PRISMA framework to guide study identification, inclusion, and exclusion. Widely used models include artificial neural networks (ANN), support vector machines (SVM), long short-term memory networks (LSTM), and random forests (RF). More recent studies increasingly employ hybrid approaches that integrate wavelet transforms, signal decomposition, and optimization techniques such as particle swarm optimization (PSO), genetic algorithms (GA), and ant colony optimization (ACO). Transformer-based models have also begun to emerge as promising tools in this domain. A central focus of this review is feature selection, which remains one of the most underdeveloped areas in GWL modeling. Most studies rely on simple filter methods like autocorrelation and mutual information. While SHapley Additive exPlanations (SHAP) has gained some traction, more advanced techniques, such as recursive feature elimination (RFE), forward feature selection (FFS), factor analysis (FA), and self-organizing maps (SOM), are rarely used. Notably, no study systematically compared multiple feature selection strategies, limiting insights into their impact on model performance. Scientometric analysis shows that Iran, China, India, and the United States contribute the most impactful research. Despite strong predictive outcomes, trial-and-error remains the dominant approach to hyperparameter tuning. The review emphasizes the need for more systematic, interpretable, and generalizable ML approaches to support robust groundwater level (GWL) forecasting.
{"title":"A systematic review of machine learning models for groundwater level prediction","authors":"Gilbert Jesse , Cyril D. Boateng , Jeffrey N.A. Aryee , Marian A. Osei , David D. Wemegah , Solomon S.R. Gidigasu , Akyana Britwum , Samuel K. Afful , Haoulata Touré , Vera Mensah , Prinsca Owusu-Afriyie","doi":"10.1016/j.acags.2025.100303","DOIUrl":"10.1016/j.acags.2025.100303","url":null,"abstract":"<div><div>This study presents a comprehensive synthesis of machine learning (ML) techniques applied to groundwater level (GWL) prediction, focusing on model architectures, feature selection methods, hyperparameter tuning, optimization algorithms, and clustering techniques. A total of 223 peer-reviewed articles were systematically reviewed using the PRISMA framework to guide study identification, inclusion, and exclusion. Widely used models include artificial neural networks (ANN), support vector machines (SVM), long short-term memory networks (LSTM), and random forests (RF). More recent studies increasingly employ hybrid approaches that integrate wavelet transforms, signal decomposition, and optimization techniques such as particle swarm optimization (PSO), genetic algorithms (GA), and ant colony optimization (ACO). Transformer-based models have also begun to emerge as promising tools in this domain. A central focus of this review is feature selection, which remains one of the most underdeveloped areas in GWL modeling. Most studies rely on simple filter methods like autocorrelation and mutual information. While SHapley Additive exPlanations (SHAP) has gained some traction, more advanced techniques, such as recursive feature elimination (RFE), forward feature selection (FFS), factor analysis (FA), and self-organizing maps (SOM), are rarely used. Notably, no study systematically compared multiple feature selection strategies, limiting insights into their impact on model performance. Scientometric analysis shows that Iran, China, India, and the United States contribute the most impactful research. Despite strong predictive outcomes, trial-and-error remains the dominant approach to hyperparameter tuning. The review emphasizes the need for more systematic, interpretable, and generalizable ML approaches to support robust groundwater level (GWL) forecasting.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"28 ","pages":"Article 100303"},"PeriodicalIF":3.2,"publicationDate":"2025-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145466504","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 : 2025-10-31DOI: 10.1016/j.acags.2025.100301
Mateus Roder , Clayton R. Pereira , João Paulo Papa , Altanir Flores de Mello Junior , Marcelo Fagundes de Rezende , Yaro Moisés Parizek Silva , Alexandre Vidal
This paper addresses the classification of pre-salt rock lithology using vision transformers, focusing on evaluating the learning and generalization capabilities of state-of-the-art pre-trained models under conditions of limited data and significant class imbalance. Our research investigates the performance of ten ViT architectures, fine-tuned on a private dataset of petrographic thin section images of carbonate rocks from two Brazilian oil wells. We also explore the impact of two different patch sizes (200 × 200 and 150 × 150 pixels) on model performance. The experimental results demonstrate that DaViT and MViTv2 achieve the highest accuracy and consistency, with DaViT surpassing 95% accuracy with the smaller patch size. In contrast, models like EfficientViT and FlexiViT exhibit lower performance and higher variability, indicating a need for further optimization. This study highlights the potential of vision transformers for geological applications and establishes new benchmarks for automated lithological classification, providing valuable insights for future research in enhancing the efficiency and accuracy of lithological studies.
{"title":"Vision transformers as SOTA models for lithological classification of Brazilian pre-salt rocks","authors":"Mateus Roder , Clayton R. Pereira , João Paulo Papa , Altanir Flores de Mello Junior , Marcelo Fagundes de Rezende , Yaro Moisés Parizek Silva , Alexandre Vidal","doi":"10.1016/j.acags.2025.100301","DOIUrl":"10.1016/j.acags.2025.100301","url":null,"abstract":"<div><div>This paper addresses the classification of pre-salt rock lithology using vision transformers, focusing on evaluating the learning and generalization capabilities of state-of-the-art pre-trained models under conditions of limited data and significant class imbalance. Our research investigates the performance of ten ViT architectures, fine-tuned on a private dataset of petrographic thin section images of carbonate rocks from two Brazilian oil wells. We also explore the impact of two different patch sizes (200 × 200 and 150 × 150 pixels) on model performance. The experimental results demonstrate that DaViT and MViTv2 achieve the highest accuracy and consistency, with DaViT surpassing 95% accuracy with the smaller patch size. In contrast, models like EfficientViT and FlexiViT exhibit lower performance and higher variability, indicating a need for further optimization. This study highlights the potential of vision transformers for geological applications and establishes new benchmarks for automated lithological classification, providing valuable insights for future research in enhancing the efficiency and accuracy of lithological studies.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"28 ","pages":"Article 100301"},"PeriodicalIF":3.2,"publicationDate":"2025-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145466503","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}
Forecasting water availability is becoming increasingly vital due to rising human demands and changing climatic pressures have caused declines in groundwater levels across many regions. While numerous studies employ data-driven approaches to predict groundwater fluctuations using meteorological data and groundwater level observations, few incorporate measurements from the vadose zone into predictive models. This study proposes a novel method leveraging an advanced hydrogeological monitoring system with high spatio-temporal resolution to forecast groundwater levels in a shallow alluvial aquifer used for drinking purposes. The monitoring system comprises a thermo-pluviometric station and three probes that measure soil water content, electrical conductivity, and temperature at depths of 0.6, 0.9, and 1.7 meters, in addition to a piezometer with a permanent water-level sensor. Data was collected at 15 min intervals over two hydrological years and integrated as exogenous inputs to enhance model predictive performance. Statistical, machine learning and deep learning architectures were tested through ARIMA, SARIMAX, Prophet and NeuralProphet providing a comprehensive evaluation of different approaches. For a robust evaluation, a rolling K-fold cross-validation strategy was implemented and coupled with a grid search to fine-tune all the models. Evaluation metrics and correlation coefficients are employed to assess the predictive capabilities of each model. Our findings indicate that prediction accuracy improves across all models with increasing depth in the vadose zone, with machine learning and deep learning models showing the most significant improvements. Specifically, at 1.7 m depth, Prophet achieves a MAPE of 4.5%, and NeuralProphet achieves a MAPE of 4.1% compared to statistical models. This study has successfully highlighted the enhancement of AI-based models for estimating levels of groundwater incorporating subsurface information from the vadose zone at different depths and phreatic zones, alongside climatic variables.
{"title":"Groundwater level forecasting using data-driven models and vadose zone: A comparative analysis of ARIMA, SARIMAX, Prophet, and NeuralProphet","authors":"Alessandro Galdelli , Davide Fronzi , Gagan Narang , Adriano Mancini , Alberto Tazioli","doi":"10.1016/j.acags.2025.100304","DOIUrl":"10.1016/j.acags.2025.100304","url":null,"abstract":"<div><div>Forecasting water availability is becoming increasingly vital due to rising human demands and changing climatic pressures have caused declines in groundwater levels across many regions. While numerous studies employ data-driven approaches to predict groundwater fluctuations using meteorological data and groundwater level observations, few incorporate measurements from the vadose zone into predictive models. This study proposes a novel method leveraging an advanced hydrogeological monitoring system with high spatio-temporal resolution to forecast groundwater levels in a shallow alluvial aquifer used for drinking purposes. The monitoring system comprises a thermo-pluviometric station and three probes that measure soil water content, electrical conductivity, and temperature at depths of 0.6, 0.9, and 1.7 meters, in addition to a piezometer with a permanent water-level sensor. Data was collected at 15 min intervals over two hydrological years and integrated as exogenous inputs to enhance model predictive performance. Statistical, machine learning and deep learning architectures were tested through ARIMA, SARIMAX, Prophet and NeuralProphet providing a comprehensive evaluation of different approaches. For a robust evaluation, a rolling K-fold cross-validation strategy was implemented and coupled with a grid search to fine-tune all the models. Evaluation metrics and correlation coefficients are employed to assess the predictive capabilities of each model. Our findings indicate that prediction accuracy improves across all models with increasing depth in the vadose zone, with machine learning and deep learning models showing the most significant improvements. Specifically, at 1.7 m depth, Prophet achieves a MAPE of 4.5%, and NeuralProphet achieves a MAPE of 4.1% compared to statistical models. This study has successfully highlighted the enhancement of AI-based models for estimating levels of groundwater incorporating subsurface information from the vadose zone at different depths and phreatic zones, alongside climatic variables.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"28 ","pages":"Article 100304"},"PeriodicalIF":3.2,"publicationDate":"2025-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145467003","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 : 2025-10-28DOI: 10.1016/j.acags.2025.100300
Divyadeep Harbola, George Mathew
MapEX is an open-source Python toolkit for analysing multi-channel X-ray maps acquired through diverse analytical platforms, including electron probe microanalysis with wavelength-dispersive spectroscopy (EPMA-WDS), scanning and transmission electron microscopy with energy-dispersive spectroscopy (SEM/TEM-EDS), micro-X-ray fluorescence (μ-XRF), and synchrotron-based mapping techniques. The software reads native CSV or text file exports, records key acquisition metadata, and packages data in a HDF5 structure that supports fast access and fully reproducible workflows. Using a linear fit, a calibration panel implements region-of-interest regressions from map intensity to composition. It reports the fitted equation, coefficient of determination, and root-mean-square error, with pointwise inclusion or exclusion of standards. For phase classification, principal-component features are combined with unsupervised clustering methods to classify phases directly from elemental distributions; parameters can be tuned and results updated interactively. An interactive interface links elemental maps, correlation plots, and phase classification, with linked selection so that pixels chosen in plot space are highlighted across all images and vice versa. Line-profile tools extract compositional trends along user-defined paths, enabling targeted inspection of grain boundaries, reaction fronts, and alteration rims. By emphasising open formats, explicit assumptions, and pixel-level validation, MapEX offers a rigorous and transparent alternative to proprietary software and lowers barriers to routine X-ray map analysis in petrology and materials science.
{"title":"MapEX: A tool for X-ray map analysis","authors":"Divyadeep Harbola, George Mathew","doi":"10.1016/j.acags.2025.100300","DOIUrl":"10.1016/j.acags.2025.100300","url":null,"abstract":"<div><div>MapEX is an open-source Python toolkit for analysing multi-channel X-ray maps acquired through diverse analytical platforms, including electron probe microanalysis with wavelength-dispersive spectroscopy (EPMA-WDS), scanning and transmission electron microscopy with energy-dispersive spectroscopy (SEM/TEM-EDS), micro-X-ray fluorescence (μ-XRF), and synchrotron-based mapping techniques. The software reads native CSV or text file exports, records key acquisition metadata, and packages data in a HDF5 structure that supports fast access and fully reproducible workflows. Using a linear fit, a calibration panel implements region-of-interest regressions from map intensity to composition. It reports the fitted equation, coefficient of determination, and root-mean-square error, with pointwise inclusion or exclusion of standards. For phase classification, principal-component features are combined with unsupervised clustering methods to classify phases directly from elemental distributions; parameters can be tuned and results updated interactively. An interactive interface links elemental maps, correlation plots, and phase classification, with linked selection so that pixels chosen in plot space are highlighted across all images and vice versa. Line-profile tools extract compositional trends along user-defined paths, enabling targeted inspection of grain boundaries, reaction fronts, and alteration rims. By emphasising open formats, explicit assumptions, and pixel-level validation, MapEX offers a rigorous and transparent alternative to proprietary software and lowers barriers to routine X-ray map analysis in petrology and materials science.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"28 ","pages":"Article 100300"},"PeriodicalIF":3.2,"publicationDate":"2025-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145467001","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 : 2025-10-28DOI: 10.1016/j.acags.2025.100302
Chieh Tsai , Pei-Jun Lee , Shimaa Bergies , John Liobe , Vaidotas Barzdėnas
High-quality satellite imagery is critical in environmental monitoring, disaster response, and urban planning applications, where detailed and accurate images are essential for informed decision-making. However, images from small satellites often have low resolution, limiting their effectiveness in addressing precise analysis challenges. To overcome these limitations, this paper presents the Edge-Based Wavelet Transformer for Super-Resolution (EBWT-SR), an innovative technique designed to enhance satellite image resolution while optimizing computational efficiency. EBWT-SR combines Spatial-Wavelet Multi-Head Attention Mechanisms and a Multi-Modal Convolutional Shallow Feature Extractor within a Convolutional Transformer architecture, allowing for the refinement of object contours and textures. By incorporating edge-based wavelet transform convolutional layers and a specialized multi-modal loss function for fine-tuning, the developed EBWT-SR improves local feature representation without increasing computational complexity. The new model can improve the results by approximately 0.67 in Peak Signal-to-Noise Ratio (PSNR) and 0.63 in Perceptually Uniform Peak Signal-to-Noise Ratio (puPSNR) metrics, along with a 7.7 % reduction in Giga Floating-Point Operations Per Second (GFLOPS) compared to recent methods on the fine-grained satellite image dataset focused on ship classification and super-resolution tasks (FGCSR-42) dataset. highlighting its ability to enhance satellite image quality while significantly maintaining computational efficiency.
{"title":"Enhancing satellite image quality with the edge-based wavelet transformer for super-resolution","authors":"Chieh Tsai , Pei-Jun Lee , Shimaa Bergies , John Liobe , Vaidotas Barzdėnas","doi":"10.1016/j.acags.2025.100302","DOIUrl":"10.1016/j.acags.2025.100302","url":null,"abstract":"<div><div>High-quality satellite imagery is critical in environmental monitoring, disaster response, and urban planning applications, where detailed and accurate images are essential for informed decision-making. However, images from small satellites often have low resolution, limiting their effectiveness in addressing precise analysis challenges. To overcome these limitations, this paper presents the Edge-Based Wavelet Transformer for Super-Resolution (EBWT-SR), an innovative technique designed to enhance satellite image resolution while optimizing computational efficiency. EBWT-SR combines Spatial-Wavelet Multi-Head Attention Mechanisms and a Multi-Modal Convolutional Shallow Feature Extractor within a Convolutional Transformer architecture, allowing for the refinement of object contours and textures. By incorporating edge-based wavelet transform convolutional layers and a specialized multi-modal loss function for fine-tuning, the developed EBWT-SR improves local feature representation without increasing computational complexity. The new model can improve the results by approximately 0.67 in Peak Signal-to-Noise Ratio (PSNR) and 0.63 in Perceptually <strong>Uniform Peak Signal-to-Noise Ratio</strong> (puPSNR) metrics, along with a 7.7 % reduction in Giga Floating-Point Operations Per Second (GFLOPS) compared to recent methods on the fine-grained satellite image dataset focused on ship classification and super-resolution tasks (FGCSR-42) dataset. highlighting its ability to enhance satellite image quality while significantly maintaining computational efficiency.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"28 ","pages":"Article 100302"},"PeriodicalIF":3.2,"publicationDate":"2025-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145417072","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 : 2025-10-13DOI: 10.1016/j.acags.2025.100297
Alfredo Gimenez Zapiola, Andrea Boselli, Alessandra Menafoglio, Simone Vantini
This work concerns a detailed review of data analysis methods used for remotely sensed images of large areas of the Earth and of other solid astronomical objects. Focus is on the problem of inferring the materials that cover the surfaces captured by hyper-spectral images and estimating their abundances and spatial distributions within the region. Different hyper-spectral unmixing methods are reported as well as compared. The most important public data-sets in this setting, which are vastly used in the testing and validation of the former, are also systematically explored. Typically, a pixel-wise constrained regression is used assuming linear mixing. Yet, more recent methodologies go beyond such assumption and are thus analysed. Data-based testing of assumptions and uncertainty quantification are found to be scarce in the literature. Open problems are spotlighted and concrete recommendations for future research are provided.
{"title":"Hyper-spectral Unmixing algorithms for remote compositional surface mapping: a review of the state of the art","authors":"Alfredo Gimenez Zapiola, Andrea Boselli, Alessandra Menafoglio, Simone Vantini","doi":"10.1016/j.acags.2025.100297","DOIUrl":"10.1016/j.acags.2025.100297","url":null,"abstract":"<div><div>This work concerns a detailed review of data analysis methods used for remotely sensed images of large areas of the Earth and of other solid astronomical objects. Focus is on the problem of inferring the materials that cover the surfaces captured by hyper-spectral images and estimating their abundances and spatial distributions within the region. Different hyper-spectral unmixing methods are reported as well as compared. The most important public data-sets in this setting, which are vastly used in the testing and validation of the former, are also systematically explored. Typically, a pixel-wise constrained regression is used assuming linear mixing. Yet, more recent methodologies go beyond such assumption and are thus analysed. Data-based testing of assumptions and uncertainty quantification are found to be scarce in the literature. Open problems are spotlighted and concrete recommendations for future research are provided.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"28 ","pages":"Article 100297"},"PeriodicalIF":3.2,"publicationDate":"2025-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145363624","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 : 2025-10-13DOI: 10.1016/j.acags.2025.100294
Mengzhen Hao , Wei Feng , Wenxing Bao , Xiaowu Zhang , Xuan Ma , Wenlong Wang
This study based on the Google Earth Engine (GEE) platform proposes a wetland classification method tailored to the diverse wetland types within Kenya, Africa, by leveraging multi-source feature extraction and integration. Firstly, a large-scale wetland sample collection strategy is developed by integrating existing land cover products and wetland-related datasets. The study places particular emphasis on the use of time-series Sentinel-2 imagery and Shuttle Radar Topography Mission (SRTM) data to design high-resolution texture feature extraction and spatiotemporal spectral information reconstruction techniques. The process yields four categories of multi-source feature sets, including spectral bands, spectral indices, topographic attributes, and texture features. Finally, a random forest algorithm is employed to perform remote sensing-based classification of wetland types across the study area. The experimental results obtained demonstrate that the integration of multi-source features has a significant effect on the enhancement of classification accuracy. In comparison with single-feature inputs, the optimal feature combination attains an overall accuracy of 82.65% and a Kappa coefficient of 77.51%. This study provides a reliable foundation for the scientific management and sustainable development of wetland ecosystems, as well as an efficient technical framework for large-scale wetland classification and mapping.
{"title":"Wetland classification based on the synergy of multi-source spatio-temporal spectral data — An example from Kenya","authors":"Mengzhen Hao , Wei Feng , Wenxing Bao , Xiaowu Zhang , Xuan Ma , Wenlong Wang","doi":"10.1016/j.acags.2025.100294","DOIUrl":"10.1016/j.acags.2025.100294","url":null,"abstract":"<div><div>This study based on the Google Earth Engine (GEE) platform proposes a wetland classification method tailored to the diverse wetland types within Kenya, Africa, by leveraging multi-source feature extraction and integration. Firstly, a large-scale wetland sample collection strategy is developed by integrating existing land cover products and wetland-related datasets. The study places particular emphasis on the use of time-series Sentinel-2 imagery and Shuttle Radar Topography Mission (SRTM) data to design high-resolution texture feature extraction and spatiotemporal spectral information reconstruction techniques. The process yields four categories of multi-source feature sets, including spectral bands, spectral indices, topographic attributes, and texture features. Finally, a random forest algorithm is employed to perform remote sensing-based classification of wetland types across the study area. The experimental results obtained demonstrate that the integration of multi-source features has a significant effect on the enhancement of classification accuracy. In comparison with single-feature inputs, the optimal feature combination attains an overall accuracy of 82.65% and a Kappa coefficient of 77.51%. This study provides a reliable foundation for the scientific management and sustainable development of wetland ecosystems, as well as an efficient technical framework for large-scale wetland classification and mapping.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"28 ","pages":"Article 100294"},"PeriodicalIF":3.2,"publicationDate":"2025-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145363623","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