Applied Computing and Geosciences最新文献_第10页

Semantic segmentation framework for atoll satellite imagery: An in-depth exploration using UNet variants and Segmentation Gym 环礁卫星图像的语义分割框架：使用UNet变体和分割体育馆的深入探索

IF 2.6 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Applied Computing and Geosciences

Pub Date : 2025-02-01 DOI: 10.1016/j.acags.2024.100217

Ray Wang , Tahiya Chowdhury , Alejandra C. Ortiz

This paper presents a framework for semantic segmentation of satellite imagery aimed at studying atoll morphometrics. Recent advances in deep neural networks for automated segmentation have been valuable across a variety of satellite and aerial imagery applications, such as land cover classification, mineral characterization, and disaster impact assessment. However, identifying an appropriate segmentation approach for geoscience research remains challenging, often relying on trial-and-error experimentation for data preparation, model selection, and validation. Building on prior efforts to create reproducible research pipelines for aerial image segmentation, we propose a systematic framework for custom segmentation model development using Segmentation Gym, a software tool designed for efficient model experimentation. Additionally, we evaluate state-of-the-art U-Net model variants to identify the most accurate and precise model for specific segmentation tasks. Using a dataset of 288 Landsat images of atolls as a case study, we conduct a detailed analysis of various annotation techniques, image types, and training methods, offering a structured framework for practitioners to design and explore segmentation models. Furthermore, we address dataset imbalance, a common challenge in geographical data, and discuss strategies to mitigate its impact on segmentation outcomes. Based on our findings, we provide recommendations for applying this framework to other geoscience research areas to address similar challenges.

针对环礁形态计量学的研究，提出了一种卫星图像语义分割框架。深度神经网络在自动分割方面的最新进展在各种卫星和航空图像应用中都很有价值，例如土地覆盖分类、矿物表征和灾害影响评估。然而，为地球科学研究确定合适的分割方法仍然具有挑战性，通常依赖于反复试验来进行数据准备、模型选择和验证。在先前为航空图像分割创建可重复研究管道的努力的基础上，我们提出了一个系统的框架，用于使用segmentation Gym开发自定义分割模型，这是一个专为高效模型实验而设计的软件工具。此外，我们评估了最先进的U-Net模型变体，以确定最准确和精确的模型，用于特定的分割任务。以288张环礁陆地卫星图像数据集为例，详细分析了各种标注技术、图像类型和训练方法，为从业者设计和探索分割模型提供了结构化框架。此外，我们解决了数据集不平衡，这是地理数据中的一个常见挑战，并讨论了减轻其对分割结果影响的策略。基于我们的发现，我们提供了将该框架应用于其他地球科学研究领域以解决类似挑战的建议。

{"title":"Semantic segmentation framework for atoll satellite imagery: An in-depth exploration using UNet variants and Segmentation Gym","authors":"Ray Wang , Tahiya Chowdhury , Alejandra C. Ortiz","doi":"10.1016/j.acags.2024.100217","DOIUrl":"10.1016/j.acags.2024.100217","url":null,"abstract":"<div><div>This paper presents a framework for semantic segmentation of satellite imagery aimed at studying atoll morphometrics. Recent advances in deep neural networks for automated segmentation have been valuable across a variety of satellite and aerial imagery applications, such as land cover classification, mineral characterization, and disaster impact assessment. However, identifying an appropriate segmentation approach for geoscience research remains challenging, often relying on trial-and-error experimentation for data preparation, model selection, and validation. Building on prior efforts to create reproducible research pipelines for aerial image segmentation, we propose a systematic framework for custom segmentation model development using Segmentation Gym, a software tool designed for efficient model experimentation. Additionally, we evaluate state-of-the-art U-Net model variants to identify the most accurate and precise model for specific segmentation tasks. Using a dataset of 288 Landsat images of atolls as a case study, we conduct a detailed analysis of various annotation techniques, image types, and training methods, offering a structured framework for practitioners to design and explore segmentation models. Furthermore, we address dataset imbalance, a common challenge in geographical data, and discuss strategies to mitigate its impact on segmentation outcomes. Based on our findings, we provide recommendations for applying this framework to other geoscience research areas to address similar challenges.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"25 ","pages":"Article 100217"},"PeriodicalIF":2.6,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143166135","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Predictive regressive models of recent marsh sediment thickness improve the quantification of coastal marsh sediment budgets 近期沼泽沉积物厚度的预测回归模型改进了沿海沼泽沉积物收支的量化

IF 2.6 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Applied Computing and Geosciences

Pub Date : 2025-02-01 DOI: 10.1016/j.acags.2024.100215

Christopher G. Smith , Julie Bernier , Alisha M. Ellis , Kathryn E.L. Smith

Coastal marsh wetlands experience variations in vertical gains and losses through time, which have allowed them to infill relict topography and record variations in drivers. The stratigraphic unit associated with the development of the marsh also reflects the long-term importance of key ecosystem services supplied by the marsh environment, including carbon storage and storm mitigation. Mapping these coastal wetland sediments and the marsh unit thickness is challenging as traditional coastal geophysical tools are not easily deployable (acoustic methods) or are unreliable in saline-soil environments (e.g., ground-penetrating radar), leaving core-based methods the most viable mapping method. In the present study, we utilized prior information on the geologic architecture of the region to select spatial and physical metrics that likely persisted throughout evolution of the marsh during the late Holocene. We then assessed the individual and collective power of these metrics to predict marsh thickness observed from cores. Employing regressive predictive models powered by these data, we improve the quantification of marsh thickness for a coastal fringing marsh within the Grand Bay estuary in Mississippi and Alabama (USA). The information gained from this approach yields improved estimates of the carbon stocks in this environment. Additionally, the stored sediment masses reflect the past, and potential future, persistence of the Grand Bay marsh under historical and present marsh-estuarine sediment exchange fluxes. Such improvements to both the sediment budget of recent marsh stratigraphic units and the spatial extent provide new resources for comparison with large-scale landscape models, the latter of which may be used, when validated, to predict future change and ecosystem transformations.

随着时间的推移，沿海沼泽湿地的垂直收益和损失会发生变化，这使它们能够填补遗留的地形并记录驱动因素的变化。与沼泽发展有关的地层单位也反映了沼泽环境提供的关键生态系统服务的长期重要性，包括碳储存和减缓风暴。绘制这些沿海湿地沉积物和沼泽单位厚度具有挑战性，因为传统的沿海地球物理工具不容易部署（声学方法），或者在盐碱地环境中不可靠（例如，探地雷达），因此基于岩心的方法是最可行的制图方法。在本研究中，我们利用该地区地质结构的先验信息来选择可能在全新世晚期沼泽进化过程中持续存在的空间和物理指标。然后，我们评估了这些指标的个人和集体能力，以预测从岩心观察到的沼泽厚度。利用这些数据支持的回归预测模型，我们改进了密西西比州和阿拉巴马州大湾河口沿海边缘沼泽厚度的量化。通过这种方法获得的信息可以改进对该环境中碳储量的估计。此外，储存的沉积物质量反映了历史和现在的沼泽-河口沉积物交换通量下大湾沼泽的过去和潜在的未来持续性。这种对近期沼泽地层单元的泥沙收支和空间范围的改进为与大尺度景观模型的比较提供了新的资源，后者在得到验证后可用于预测未来的变化和生态系统的转变。

{"title":"Predictive regressive models of recent marsh sediment thickness improve the quantification of coastal marsh sediment budgets","authors":"Christopher G. Smith , Julie Bernier , Alisha M. Ellis , Kathryn E.L. Smith","doi":"10.1016/j.acags.2024.100215","DOIUrl":"10.1016/j.acags.2024.100215","url":null,"abstract":"<div><div>Coastal marsh wetlands experience variations in vertical gains and losses through time, which have allowed them to infill relict topography and record variations in drivers. The stratigraphic unit associated with the development of the marsh also reflects the long-term importance of key ecosystem services supplied by the marsh environment, including carbon storage and storm mitigation. Mapping these coastal wetland sediments and the marsh unit thickness is challenging as traditional coastal geophysical tools are not easily deployable (acoustic methods) or are unreliable in saline-soil environments (e.g., ground-penetrating radar), leaving core-based methods the most viable mapping method. In the present study, we utilized prior information on the geologic architecture of the region to select spatial and physical metrics that likely persisted throughout evolution of the marsh during the late Holocene. We then assessed the individual and collective power of these metrics to predict marsh thickness observed from cores. Employing regressive predictive models powered by these data, we improve the quantification of marsh thickness for a coastal fringing marsh within the Grand Bay estuary in Mississippi and Alabama (USA). The information gained from this approach yields improved estimates of the carbon stocks in this environment. Additionally, the stored sediment masses reflect the past, and potential future, persistence of the Grand Bay marsh under historical and present marsh-estuarine sediment exchange fluxes. Such improvements to both the sediment budget of recent marsh stratigraphic units and the spatial extent provide new resources for comparison with large-scale landscape models, the latter of which may be used, when validated, to predict future change and ecosystem transformations.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"25 ","pages":"Article 100215"},"PeriodicalIF":2.6,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143166136","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Relationships between fault friction, slip time, and physical parameters explored by experiment-based friction model: A machine learning approach using recurrent neural networks (RNNs) 基于实验的摩擦模型探讨断层摩擦、滑动时间和物理参数之间的关系：一种使用循环神经网络（rnn）的机器学习方法

IF 2.6 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Applied Computing and Geosciences

Pub Date : 2025-02-01 DOI: 10.1016/j.acags.2025.100231

Tae-Hoon Uhmb , Yohei Hamada , Takehiro Hirose

Understanding the relationship between fault friction and physical parameters is crucial for comprehending earthquake physics. Despite various friction models developed to explain this relationship, representing the relationships in a friction model with greater detail remains a challenge due to intricate correlations, including the nonlinear interplay between physical parameters and friction. Here we develop new models to define the relationship between various physical parameters (slip velocity, axial displacement, temperature, rate of temperature, and rate of axial displacement), friction coefficient, and slip time. The models are established by utilizing Recurrent Neural Networks (RNNs) to analyze continuous data in high-velocity rotary shear experiments (HVR), as reported by previous work. The experiment has been conducted on diorite specimens at a slip velocity (0.004 m/s) in various normal stress (0.3–5.8 MPa). At this conditions, frictional heating occurs inevitably at the sliding surface, reaching temperature up to 68 °C. We first identified the optimal model by assessing its accuracy in relation to the time interval for defining friction. Following this, we explored the relationship between friction and physical parameters with varying slip time and conditions by analyzing the gradient importance of physical parameters within the identified model. Our results demonstrate that the importance of physical parameters continuously shifts over slip time and conditions, and temperature stands out as the most influential parameter affecting fault friction under slip conditions of this study accompanied by frictional heating. Our study demonstrates the potential of deep learning analysis in enhancing our understanding of complex frictional processes, contributing to the development of more refined friction models and improving predictive models for earthquake physics.

了解断层摩擦与物理参数之间的关系对于理解地震物理是至关重要的。尽管开发了各种摩擦模型来解释这种关系，但由于复杂的相关性，包括物理参数和摩擦之间的非线性相互作用，更详细地表示摩擦模型中的关系仍然是一个挑战。在这里，我们开发了新的模型来定义各种物理参数（滑移速度、轴向位移、温度、温度速率和轴向位移速率）、摩擦系数和滑移时间之间的关系。该模型是利用递归神经网络（RNNs）对高速旋转剪切实验（HVR）中的连续数据进行分析而建立的。对闪长岩试样在不同的法向应力（0.3 ~ 5.8 MPa）下，以滑移速度（0.004 m/s）进行了试验。在这种情况下，在滑动表面不可避免地发生摩擦加热，温度可达68℃。我们首先通过评估其与定义摩擦的时间间隔的准确性来确定最佳模型。在此基础上，通过分析所识别模型中物理参数的梯度重要性，探讨了滑移时间和滑移条件不同时摩擦与物理参数之间的关系。我们的研究结果表明，物理参数的重要性随着滑动时间和条件的变化而不断变化，在本研究的滑移条件下，温度是影响断层摩擦的最重要参数，同时伴有摩擦加热。我们的研究证明了深度学习分析在增强我们对复杂摩擦过程的理解方面的潜力，有助于开发更精细的摩擦模型和改进地震物理预测模型。

{"title":"Relationships between fault friction, slip time, and physical parameters explored by experiment-based friction model: A machine learning approach using recurrent neural networks (RNNs)","authors":"Tae-Hoon Uhmb , Yohei Hamada , Takehiro Hirose","doi":"10.1016/j.acags.2025.100231","DOIUrl":"10.1016/j.acags.2025.100231","url":null,"abstract":"<div><div>Understanding the relationship between fault friction and physical parameters is crucial for comprehending earthquake physics. Despite various friction models developed to explain this relationship, representing the relationships in a friction model with greater detail remains a challenge due to intricate correlations, including the nonlinear interplay between physical parameters and friction. Here we develop new models to define the relationship between various physical parameters (slip velocity, axial displacement, temperature, rate of temperature, and rate of axial displacement), friction coefficient, and slip time. The models are established by utilizing Recurrent Neural Networks (RNNs) to analyze continuous data in high-velocity rotary shear experiments (HVR), as reported by previous work. The experiment has been conducted on diorite specimens at a slip velocity (0.004 m/s) in various normal stress (0.3–5.8 MPa). At this conditions, frictional heating occurs inevitably at the sliding surface, reaching temperature up to 68 °C. We first identified the optimal model by assessing its accuracy in relation to the time interval for defining friction. Following this, we explored the relationship between friction and physical parameters with varying slip time and conditions by analyzing the gradient importance of physical parameters within the identified model. Our results demonstrate that the importance of physical parameters continuously shifts over slip time and conditions, and temperature stands out as the most influential parameter affecting fault friction under slip conditions of this study accompanied by frictional heating. Our study demonstrates the potential of deep learning analysis in enhancing our understanding of complex frictional processes, contributing to the development of more refined friction models and improving predictive models for earthquake physics.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"25 ","pages":"Article 100231"},"PeriodicalIF":2.6,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143562771","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Geological object recognition in legacy maps through data augmentation and transfer learning techniques 利用数据增强和迁移学习技术在遗留地图中识别地质目标

IF 2.6 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Applied Computing and Geosciences

Pub Date : 2025-02-01 DOI: 10.1016/j.acags.2025.100233

Wenjia Li, Weilin Chen, Jiyin Zhang, Chenhao Li, Xiaogang Ma

Maps are crucial tools in geosciences, providing detailed representations of the spatial distribution and relationships among geological features. Accurate recognition and classification of geological objects within these maps are essential for applications in resource exploration, environmental management, and geological hazard assessment. Along the years, many legacy geological maps have been accumulated, and many of them are not in data formats ready for machines to read and analyze. The inherent diversity and complexity of geological features, combined with the labor-intensive process of manual annotation, pose significant challenges in the usage of those maps. This study addresses these challenges by proposing an innovative approach that leverages legend data for data augmentation and employs transfer learning techniques to improve the quality of object recognition. Legend data from geological maps offer standardized symbols and annotations. Using them to augment existing datasets increases the diversity and volume of training data, thereby enhances the model's ability to generalize across various geological contexts. A deep learning model called EfficientNet is then fine-tuned using the augmented dataset to recognize and classify geological features more accurately. The model's performance is evaluated based on accuracy, recall, and F1-score, with results showing significant improvements, particularly for datasets with texture-rich information. The proposed method demonstrates that the combination of data augmentation and transfer learning significantly enhances the accuracy and efficiency of geological object recognition. This approach not only reduces the manual effort needed for geological object recognition but also contributes to the advancement of geological mapping and analysis.

地图是地球科学的重要工具，提供了地质特征之间空间分布和关系的详细表示。在这些地图中准确识别和分类地质目标对于资源勘探、环境管理和地质灾害评估的应用至关重要。多年来，已经积累了许多遗留的地质图，其中许多地图的数据格式不适合机器读取和分析。地质特征固有的多样性和复杂性，加上人工标注的劳动密集型过程，对这些地图的使用构成了重大挑战。本研究通过提出一种创新的方法来解决这些挑战，该方法利用图例数据进行数据增强，并采用迁移学习技术来提高对象识别的质量。地质图中的图例数据提供了标准化的符号和注释。使用它们来扩充现有的数据集，增加了训练数据的多样性和数量，从而增强了模型在各种地质背景下的泛化能力。然后，使用增强的数据集对名为“效率网”的深度学习模型进行微调，以更准确地识别和分类地质特征。该模型的性能基于准确性、召回率和f1分数进行评估，结果显示出显著的改进，特别是对于具有丰富纹理信息的数据集。结果表明，将数据增强与迁移学习相结合，可以显著提高地质目标识别的精度和效率。该方法不仅减少了人工识别地质目标的工作量，而且有助于推进地质填图和分析。

{"title":"Geological object recognition in legacy maps through data augmentation and transfer learning techniques","authors":"Wenjia Li, Weilin Chen, Jiyin Zhang, Chenhao Li, Xiaogang Ma","doi":"10.1016/j.acags.2025.100233","DOIUrl":"10.1016/j.acags.2025.100233","url":null,"abstract":"<div><div>Maps are crucial tools in geosciences, providing detailed representations of the spatial distribution and relationships among geological features. Accurate recognition and classification of geological objects within these maps are essential for applications in resource exploration, environmental management, and geological hazard assessment. Along the years, many legacy geological maps have been accumulated, and many of them are not in data formats ready for machines to read and analyze. The inherent diversity and complexity of geological features, combined with the labor-intensive process of manual annotation, pose significant challenges in the usage of those maps. This study addresses these challenges by proposing an innovative approach that leverages legend data for data augmentation and employs transfer learning techniques to improve the quality of object recognition. Legend data from geological maps offer standardized symbols and annotations. Using them to augment existing datasets increases the diversity and volume of training data, thereby enhances the model's ability to generalize across various geological contexts. A deep learning model called EfficientNet is then fine-tuned using the augmented dataset to recognize and classify geological features more accurately. The model's performance is evaluated based on accuracy, recall, and F1-score, with results showing significant improvements, particularly for datasets with texture-rich information. The proposed method demonstrates that the combination of data augmentation and transfer learning significantly enhances the accuracy and efficiency of geological object recognition. This approach not only reduces the manual effort needed for geological object recognition but also contributes to the advancement of geological mapping and analysis.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"25 ","pages":"Article 100233"},"PeriodicalIF":2.6,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143578333","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

X-ray Micro-CT based characterization of rock cuttings with deep learning 基于x射线微ct的岩屑深度学习表征

IF 2.6 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Applied Computing and Geosciences

Pub Date : 2025-02-01 DOI: 10.1016/j.acags.2025.100220

Nils Olsen , Yifeng Chen , Pascal Turberg , Alexandre Moreau , Alexandre Alahi

Rock cuttings from destructive boreholes are a common and cheaper source of drilling materials that can be used to determine underground geology compared to rock core samples. Classifying manually the series of cuttings can be a long and tedious process and can also be prone to subjectivity leading to errors. In this paper, a framework for the classification of multiple types of rock structures is introduced based on rock cutting images from X-ray micro-CT technology. The classification is performed using a simple yet effective deep learning model (a ResNet-18 architecture) to categorize five different lithologies: micritic limestone, bioclastic limestone, oolithic limestone, molassic sandstone and gneiss. The proposed network is trained on 2 datasets (laboratory and borehole) both containing the five lithologies and comprise over 10 000 images. The laboratory dataset consists of a well-controlled experiments with homogeneous samples and the borehole dataset with heterogeneous samples corresponding to a real case application. Among all the considered models, including ResNet-34, and SPP-CNN and human experts manual classification, ResNet-18 demonstrates superior performance across multiple evaluation metrics, including precision, recall, and F1-score. It is to our best knowledge, the first time a test comparing deep neural network and human performance is performed for this task. To optimize the performance of the proposed model, the transfer learning method is implemented. Furthermore, the experiments demonstrate that when employing transfer learning, the size of the dataset significantly impacts the performance of the model. In the studied design, the experimental results confirm that the proposed approach is a cost-effective and efficient method for automated rock cutting classification using the micro-CT technique, and it can be easily modified to adapt the rock cutting classification from various types and sources.

与岩心样本相比，来自破坏性钻孔的岩屑是一种常见且便宜的钻井材料来源，可用于确定地下地质情况。手工对一系列岩屑进行分类可能是一个漫长而繁琐的过程，也可能容易主观导致错误。本文介绍了一种基于x射线微ct技术岩石切割图像的多类型岩石结构分类框架。使用简单而有效的深度学习模型（ResNet-18架构）进行分类，对五种不同的岩性进行分类：泥晶灰岩、生物碎屑灰岩、鲕粒灰岩、摩尔系砂岩和片麻岩。所提出的网络在两个数据集（实验室和井眼）上进行训练，这两个数据集都包含五种岩性，包含超过10,000张图像。实验室数据集包括具有均匀样本的控制良好的实验数据和具有对应于实际案例应用的非均匀样本的井眼数据集。在所有考虑的模型中，包括ResNet-34、SPP-CNN和人类专家手动分类，ResNet-18在多个评估指标上表现出卓越的性能，包括精度、召回率和f1分数。据我们所知，这是第一次将深度神经网络和人类表现进行比较的测试。为了优化模型的性能，实现了迁移学习方法。此外，实验表明，当采用迁移学习时，数据集的大小显著影响模型的性能。在研究设计中，实验结果证实了该方法是一种经济高效的利用微ct技术进行岩石切割自动分类的方法，并且该方法易于修改以适应不同类型和来源的岩石切割分类。

{"title":"X-ray Micro-CT based characterization of rock cuttings with deep learning","authors":"Nils Olsen , Yifeng Chen , Pascal Turberg , Alexandre Moreau , Alexandre Alahi","doi":"10.1016/j.acags.2025.100220","DOIUrl":"10.1016/j.acags.2025.100220","url":null,"abstract":"<div><div>Rock cuttings from destructive boreholes are a common and cheaper source of drilling materials that can be used to determine underground geology compared to rock core samples. Classifying manually the series of cuttings can be a long and tedious process and can also be prone to subjectivity leading to errors. In this paper, a framework for the classification of multiple types of rock structures is introduced based on rock cutting images from X-ray micro-CT technology. The classification is performed using a simple yet effective deep learning model (a ResNet-18 architecture) to categorize five different lithologies: micritic limestone, bioclastic limestone, oolithic limestone, molassic sandstone and gneiss. The proposed network is trained on 2 datasets (laboratory and borehole) both containing the five lithologies and comprise over 10 000 images. The laboratory dataset consists of a well-controlled experiments with homogeneous samples and the borehole dataset with heterogeneous samples corresponding to a real case application. Among all the considered models, including ResNet-34, and SPP-CNN and human experts manual classification, ResNet-18 demonstrates superior performance across multiple evaluation metrics, including precision, recall, and F1-score. It is to our best knowledge, the first time a test comparing deep neural network and human performance is performed for this task. To optimize the performance of the proposed model, the transfer learning method is implemented. Furthermore, the experiments demonstrate that when employing transfer learning, the size of the dataset significantly impacts the performance of the model. In the studied design, the experimental results confirm that the proposed approach is a cost-effective and efficient method for automated rock cutting classification using the micro-CT technique, and it can be easily modified to adapt the rock cutting classification from various types and sources.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"25 ","pages":"Article 100220"},"PeriodicalIF":2.6,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143165478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Skillful prediction of Indian Ocean Dipole index using machine learning models 利用机器学习模型熟练预测印度洋偶极子指数

IF 2.6 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Applied Computing and Geosciences

Pub Date : 2025-02-01 DOI: 10.1016/j.acags.2025.100228

J.V. Ratnam, Swadhin K. Behera, Masami Nonaka, Kalpesh R. Patil

In this study, we evaluated six machine learning models for their skill in predicting the Indian Ocean Dipole (IOD). The results based on the IOD index predictions at 1–8 month lead time indicate that the AdaBoost model with Multi-Layer Perceptron as the base estimator, AdaBoost(MLP), to perform better than the other five models in predicting the IOD index at all lead times. Interestingly, the IOD predictions of AdaBoost(MLP) had an anomaly correlation coefficient above 0.6 at almost all lead times. The results suggest that the AdaBoost(MLP) machine learning model to be a promising tool for predicting the IOD index with a long lead time of 8 months. Analysis revealed that the machine learning model predictions are aided by the signals from the Pacific region, owing to co-occurrences of some of the IODs with El Nino-Southern Oscillations.

在这项研究中，我们评估了六种机器学习模型在预测印度洋偶极子（IOD）方面的技能。基于提前期1-8个月的IOD指数预测结果表明，以多层感知器为基本估计器的AdaBoost模型（MLP）在预测所有提前期的IOD指数方面都优于其他5种模型。有趣的是，AdaBoost（MLP）的IOD预测在几乎所有提前期的异常相关系数都在0.6以上。结果表明，AdaBoost（MLP）机器学习模型是预测IOD指数的一个很有前途的工具，提前期为8个月。分析表明，由于一些iod与厄尔尼诺-南方涛动共同出现，机器学习模型的预测得到了来自太平洋地区的信号的辅助。

引用次数: 0

Streamlining geoscience data analysis with an LLM-driven workflow 通过llm驱动的工作流程简化地球科学数据分析

IF 2.6 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Applied Computing and Geosciences

Pub Date : 2025-02-01 DOI: 10.1016/j.acags.2024.100218

Jiyin Zhang, Cory Clairmont, Xiang Que, Wenjia Li, Weilin Chen, Chenhao Li, Xiaogang Ma

Large Language Models (LLMs) have made significant advancements in natural language processing and human-like response generation. However, training and fine-tuning an LLM to fit the strict requirements in the scope of academic research, such as geoscience, still requires significant computational resources and human expert alignment to ensure the quality and reliability of the generated content. The challenges highlight the need for a more flexible and reliable LLM workflow to meet domain-specific analysis needs. This study proposes an LLM-driven workflow that addresses the challenges of utilizing LLMs in geoscience data analysis. The work was built upon the open data API (application programming interface) of Mindat, one of the largest databases in mineralogy. We designed and developed an open-source LLM-driven workflow that processes natural language requests and automatically utilizes the Mindat API, mineral co-occurrence network analysis, and locality distribution heat map visualization to conduct geoscience data analysis tasks. Using prompt engineering techniques, we developed a supervisor-based agentic framework that enables LLM agents to not only interpret context information but also autonomously addressing complex geoscience analysis tasks, bridging the gap between automated workflows and human expertise. This agentic design emphasizes autonomy, allowing the workflow to adapt seamlessly to future advancements in LLM capabilities without requiring additional fine-tuning or domain-specific embedding. By providing the comprehensive context of the task in the workflow and the professional tool, we ensure the quality of LLM-generated content without the need to embed geoscience knowledge into LLMs through fine-tuning or human alignment. Our approach integrates LLMs into geoscience data analysis, addressing the need for specialized tools while reducing the learning curve through LLM-driven interactions between users and APIs. This streamlined workflow enhances the efficiency of exploratory data analysis, as demonstrated by the several use cases presented. In our future work we will explore the scalability of this workflow through the integration of additional agents and diverse geoscience data sources.

大型语言模型（llm）在自然语言处理和类人反应生成方面取得了重大进展。然而，培训和微调法学硕士以适应学术研究（如地球科学）范围内的严格要求，仍然需要大量的计算资源和人类专家校准，以确保生成内容的质量和可靠性。这些挑战突出了对更灵活和可靠的LLM工作流的需求，以满足特定领域的分析需求。本研究提出了一个法学硕士驱动的工作流程，解决了在地球科学数据分析中利用法学硕士的挑战。这项工作是建立在Mindat的开放数据API（应用程序编程接口）上的，Mindat是矿物学领域最大的数据库之一。我们设计并开发了一个开源的llm驱动的工作流程，它可以处理自然语言请求，并自动利用Mindat API、矿物共生网络分析和局部分布热图可视化来执行地球科学数据分析任务。利用即时工程技术，我们开发了一个基于监督的代理框架，使LLM代理不仅可以解释上下文信息，还可以自主处理复杂的地球科学分析任务，弥合自动化工作流程与人类专业知识之间的差距。这种代理设计强调自主性，允许工作流无缝地适应LLM功能的未来发展，而无需额外的微调或特定领域的嵌入。通过提供工作流程中任务的全面背景和专业工具，我们确保了法学硕士生成内容的质量，而无需通过微调或人工校准将地球科学知识嵌入法学硕士。我们的方法将llm集成到地球科学数据分析中，解决了对专业工具的需求，同时通过llm驱动的用户和api之间的交互减少了学习曲线。这个简化的工作流程提高了探索性数据分析的效率，正如所提供的几个用例所证明的那样。在未来的工作中，我们将通过集成其他代理和各种地球科学数据源来探索该工作流的可扩展性。

{"title":"Streamlining geoscience data analysis with an LLM-driven workflow","authors":"Jiyin Zhang, Cory Clairmont, Xiang Que, Wenjia Li, Weilin Chen, Chenhao Li, Xiaogang Ma","doi":"10.1016/j.acags.2024.100218","DOIUrl":"10.1016/j.acags.2024.100218","url":null,"abstract":"<div><div>Large Language Models (LLMs) have made significant advancements in natural language processing and human-like response generation. However, training and fine-tuning an LLM to fit the strict requirements in the scope of academic research, such as geoscience, still requires significant computational resources and human expert alignment to ensure the quality and reliability of the generated content. The challenges highlight the need for a more flexible and reliable LLM workflow to meet domain-specific analysis needs. This study proposes an LLM-driven workflow that addresses the challenges of utilizing LLMs in geoscience data analysis. The work was built upon the open data API (application programming interface) of Mindat, one of the largest databases in mineralogy. We designed and developed an open-source LLM-driven workflow that processes natural language requests and automatically utilizes the Mindat API, mineral co-occurrence network analysis, and locality distribution heat map visualization to conduct geoscience data analysis tasks. Using prompt engineering techniques, we developed a supervisor-based agentic framework that enables LLM agents to not only interpret context information but also autonomously addressing complex geoscience analysis tasks, bridging the gap between automated workflows and human expertise. This agentic design emphasizes autonomy, allowing the workflow to adapt seamlessly to future advancements in LLM capabilities without requiring additional fine-tuning or domain-specific embedding. By providing the comprehensive context of the task in the workflow and the professional tool, we ensure the quality of LLM-generated content without the need to embed geoscience knowledge into LLMs through fine-tuning or human alignment. Our approach integrates LLMs into geoscience data analysis, addressing the need for specialized tools while reducing the learning curve through LLM-driven interactions between users and APIs. This streamlined workflow enhances the efficiency of exploratory data analysis, as demonstrated by the several use cases presented. In our future work we will explore the scalability of this workflow through the integration of additional agents and diverse geoscience data sources.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"25 ","pages":"Article 100218"},"PeriodicalIF":2.6,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143166138","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Integrating empirical analysis and deep learning for accurate monsoon prediction in Kerala, India 整合实证分析和深度学习，以准确预测印度喀拉拉邦的季风

IF 2.6 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Applied Computing and Geosciences

Pub Date : 2024-12-01 DOI: 10.1016/j.acags.2024.100211

Yajnaseni Dash, Ajith Abraham

Kerala, a coastal state in India characterized by its humid tropical monsoon climate, is profoundly influenced by the Western Ghats and the Arabian Sea. Kerala receives significant rainfall during both the southwest monsoon (June to September, JJAS) and the northeast monsoon (October to December, OND) seasons. Given the substantial impact of rainfall on the state's economy and livelihoods, accurate precipitation forecasting is of critical importance. Although Kerala's annual rainfall is approximately 2.5 times higher than the national average, the state frequently experiences water scarcity due to rapid runoff into the Arabian Sea. This study builds upon previous research concerning Kerala's rainfall patterns and introduces a novel approach to improving rainfall predictions. Usage of a hybrid model that integrates Empirical Mode Decomposition (EMD) with Detrended Fluctuation Analysis (DFA) and deep Long Short-Term Memory (LSTM) neural networks, demonstrates enhanced precision in forecasting. Thus, by integrating empirical data analysis with advanced deep learning techniques, this research offers a robust framework for predicting rainfall in Kerala, making a significant contribution to the field of climate informatics and providing practical benefits for the region's economy and environmental management.

喀拉拉邦是印度的一个沿海邦，以潮湿的热带季风气候为特征，深受西高止山脉和阿拉伯海的影响。喀拉拉邦在西南季风（6月至9月，JJAS）和东北季风（10月至12月，OND）季节都有大量降雨。鉴于降雨对该州经济和生计的重大影响，准确的降水预报至关重要。虽然喀拉拉邦的年降雨量大约是全国平均水平的2.5倍，但由于径流迅速流入阿拉伯海，该邦经常缺水。这项研究建立在先前关于喀拉拉邦降雨模式的研究基础上，并引入了一种改进降雨预测的新方法。结合经验模态分解（EMD）、去趋势波动分析（DFA）和深度长短期记忆（LSTM）神经网络的混合模型的使用，证明了预测精度的提高。因此，通过将经验数据分析与先进的深度学习技术相结合，本研究为预测喀拉拉邦的降雨提供了一个强大的框架，为气候信息学领域做出了重大贡献，并为该地区的经济和环境管理提供了实际效益。

{"title":"Integrating empirical analysis and deep learning for accurate monsoon prediction in Kerala, India","authors":"Yajnaseni Dash, Ajith Abraham","doi":"10.1016/j.acags.2024.100211","DOIUrl":"10.1016/j.acags.2024.100211","url":null,"abstract":"<div><div>Kerala, a coastal state in India characterized by its humid tropical monsoon climate, is profoundly influenced by the Western Ghats and the Arabian Sea. Kerala receives significant rainfall during both the southwest monsoon (June to September, JJAS) and the northeast monsoon (October to December, OND) seasons. Given the substantial impact of rainfall on the state's economy and livelihoods, accurate precipitation forecasting is of critical importance. Although Kerala's annual rainfall is approximately 2.5 times higher than the national average, the state frequently experiences water scarcity due to rapid runoff into the Arabian Sea. This study builds upon previous research concerning Kerala's rainfall patterns and introduces a novel approach to improving rainfall predictions. Usage of a hybrid model that integrates Empirical Mode Decomposition (EMD) with Detrended Fluctuation Analysis (DFA) and deep Long Short-Term Memory (LSTM) neural networks, demonstrates enhanced precision in forecasting. Thus, by integrating empirical data analysis with advanced deep learning techniques, this research offers a robust framework for predicting rainfall in Kerala, making a significant contribution to the field of climate informatics and providing practical benefits for the region's economy and environmental management.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"24 ","pages":"Article 100211"},"PeriodicalIF":2.6,"publicationDate":"2024-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143137714","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

CloudSense: A model for cloud type identification using machine learning from radar data CloudSense：一个利用雷达数据进行机器学习的云类型识别模型

IF 2.6 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Applied Computing and Geosciences

Pub Date : 2024-12-01 DOI: 10.1016/j.acags.2024.100209

Mehzooz Nizar , Jha K. Ambuj , Manmeet Singh , S.B. Vaisakh , G. Pandithurai

The knowledge of type of precipitating cloud is crucial for radar based quantitative estimates of precipitation. We propose a novel model called CloudSense which uses machine learning to accurately identify the type of precipitating clouds over the complex terrain locations in the Western Ghats (WG) of India. CloudSense uses vertical reflectivity profiles collected during July–August 2018 from an X-band radar to classify clouds into four categories namely stratiform, mixed stratiform-convective, convective and shallow clouds. The machine learning (ML) model used in CloudSense was trained using a dataset balanced by Synthetic Minority Oversampling Technique (SMOTE), with features selected based on physical characteristics relevant to different cloud types. Among various ML models evaluated Light Gradient Boosting Machine (LightGBM) demonstrate superior performance in classifying cloud types with a BAC (Balanced Accuracy) of 0.79 and F1-Score of 0.8. CloudSense generated results are also compared against conventional radar algorithms and we find that CloudSense performs better than radar algorithms. For 200 samples tested, the radar algorithm achieved a BAC of 0.69 and F1-Score of 0.68, whereas CloudSense achieved a BAC of 0.8 and F1-Score of 0.79. Our results show that ML based approach can provide more accurate cloud detection and classification which would be useful to improve precipitation estimates over the complex terrain of the WG.

降水云类型的知识对于基于雷达的降水定量估计至关重要。我们提出了一个名为CloudSense的新模型，该模型使用机器学习来准确识别印度西高止山脉（WG）复杂地形位置上的降水云类型。CloudSense使用2018年7月至8月从x波段雷达收集的垂直反射率剖面将云分为四类，即层状云、层-对流混合云、对流云和浅云。CloudSense中使用的机器学习（ML）模型使用由合成少数派过采样技术（SMOTE）平衡的数据集进行训练，并根据与不同云类型相关的物理特征选择特征。在评估的各种ML模型中，光梯度增强机（LightGBM）在云类型分类方面表现优异，BAC（平衡精度）为0.79，F1-Score为0.8。CloudSense生成的结果也与传统雷达算法进行了比较，我们发现CloudSense的性能优于雷达算法。在测试的200个样本中，雷达算法的BAC为0.69，F1-Score为0.68，而CloudSense的BAC为0.8，F1-Score为0.79。我们的研究结果表明，基于机器学习的方法可以提供更准确的云检测和分类，这将有助于改善WG复杂地形上的降水估计。

{"title":"CloudSense: A model for cloud type identification using machine learning from radar data","authors":"Mehzooz Nizar , Jha K. Ambuj , Manmeet Singh , S.B. Vaisakh , G. Pandithurai","doi":"10.1016/j.acags.2024.100209","DOIUrl":"10.1016/j.acags.2024.100209","url":null,"abstract":"<div><div>The knowledge of type of precipitating cloud is crucial for radar based quantitative estimates of precipitation. We propose a novel model called CloudSense which uses machine learning to accurately identify the type of precipitating clouds over the complex terrain locations in the Western Ghats (WG) of India. CloudSense uses vertical reflectivity profiles collected during July–August 2018 from an X-band radar to classify clouds into four categories namely stratiform, mixed stratiform-convective, convective and shallow clouds. The machine learning (ML) model used in CloudSense was trained using a dataset balanced by Synthetic Minority Oversampling Technique (SMOTE), with features selected based on physical characteristics relevant to different cloud types. Among various ML models evaluated Light Gradient Boosting Machine (LightGBM) demonstrate superior performance in classifying cloud types with a BAC (Balanced Accuracy) of 0.79 and F1-Score of 0.8. CloudSense generated results are also compared against conventional radar algorithms and we find that CloudSense performs better than radar algorithms. For 200 samples tested, the radar algorithm achieved a BAC of 0.69 and F1-Score of 0.68, whereas CloudSense achieved a BAC of 0.8 and F1-Score of 0.79. Our results show that ML based approach can provide more accurate cloud detection and classification which would be useful to improve precipitation estimates over the complex terrain of the WG.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"24 ","pages":"Article 100209"},"PeriodicalIF":2.6,"publicationDate":"2024-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142745714","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Enhancing prediction of fluid-saturated fracture characteristics using deep learning super resolution 利用深度学习超分辨率增强饱和流体裂缝特征预测

IF 2.6 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Applied Computing and Geosciences

Pub Date : 2024-12-01 DOI: 10.1016/j.acags.2024.100208

Manju Pharkavi Murugesu , Vignesh Krishnan , Anthony R. Kovscek

Utilization of subsurface resources is essential to achieve energy sustainability including large-scale CO

_{2}

sequestration, H

_{2}

storage, geothermal energy extraction, and hydrocarbon recovery. In-situ visualization of fluid flow in geological media is essential to understand complex, coupled, physical and chemical processes underlying fluid injection, storage, extraction. X-ray Computed Tomography (CT) in the laboratory has proven beneficial to visualize changes in the flow field with rapid temporal resolution (10’s s) and moderate spatial resolution (100’s

μ m

). There is a trade-off between spatial and temporal resolution that limits accurate characterization of dynamics in rock features that are below spatial resolution of CT. While past literature has offered solutions to improve resolution of CT rock images, including deep learning-based algorithms, our study uniquely focuses on improving dynamic, partially and fully fluid-saturated geological images. Fluid-saturated CT images offer additional information, through augmented signals provided by the presence of fluid. Among challenges, CT images of geological media inherently possess limited information due to their single-channel gray-scale source. Additionally, fluid flows through partially saturated media frustrate existing super resolution techniques because unsaturated CT images are an inaccurate proxy for saturated dynamic rock images. The novelty of this work is the expansion of a generative adversarial network (GAN) for applications involving super resolution of partially saturated low resolution CT images using end-member, unsaturated high resolution

μ

CT images. To this end, we acquired multiscale low- and high-resolution CT rock images in unsaturated and saturated states. Among GAN and convolutional neural networks, GAN’s produce realistic high-resolution reconstructions of saturated geological media when trained using high-resolution, unsaturated images and lower resolution images in various saturation states. The model has direct usefulness for interpretation of real-time images.

地下资源的利用是实现能源可持续性的关键，包括大规模的二氧化碳封存、H2储存、地热能开采和碳氢化合物回收。地质介质中流体流动的现场可视化对于理解流体注入、储存和提取背后的复杂、耦合的物理和化学过程至关重要。在实验室中，x射线计算机断层扫描（CT）已被证明有利于可视化流场的变化，具有快速的时间分辨率（10 s）和中等的空间分辨率（100 μm）。在空间分辨率和时间分辨率之间存在权衡，这限制了对低于CT空间分辨率的岩石特征的准确动态表征。虽然过去的文献已经提供了提高CT岩石图像分辨率的解决方案，包括基于深度学习的算法，但我们的研究独特地专注于提高动态、部分和完全流体饱和的地质图像。流体饱和CT图像通过流体存在提供的增强信号提供额外的信息。在挑战中，地质介质的CT图像由于其单一通道灰度源而固有地具有有限的信息。此外，流体在部分饱和介质中的流动阻碍了现有的超分辨率技术，因为非饱和CT图像是饱和动态岩石图像的不准确代表。这项工作的新颖之处在于扩展了生成对抗网络（GAN），用于使用端元、不饱和高分辨率μCT图像来处理部分饱和低分辨率CT图像的超分辨率应用。为此，我们获得了非饱和和饱和状态下的多尺度低分辨率CT岩石图像。在GAN和卷积神经网络中，当使用各种饱和状态下的高分辨率图像、不饱和图像和低分辨率图像进行训练时，GAN可以产生逼真的高分辨率饱和地质介质重建。该模型对实时图像的判读具有直接的实用价值。

{"title":"Enhancing prediction of fluid-saturated fracture characteristics using deep learning super resolution","authors":"Manju Pharkavi Murugesu , Vignesh Krishnan , Anthony R. Kovscek","doi":"10.1016/j.acags.2024.100208","DOIUrl":"10.1016/j.acags.2024.100208","url":null,"abstract":"<div><div>Utilization of subsurface resources is essential to achieve energy sustainability including large-scale CO<span><math><msub><mrow></mrow><mrow><mn>2</mn></mrow></msub></math></span> sequestration, H<span><math><msub><mrow></mrow><mrow><mn>2</mn></mrow></msub></math></span> storage, geothermal energy extraction, and hydrocarbon recovery. In-situ visualization of fluid flow in geological media is essential to understand complex, coupled, physical and chemical processes underlying fluid injection, storage, extraction. X-ray Computed Tomography (CT) in the laboratory has proven beneficial to visualize changes in the flow field with rapid temporal resolution (10’s s) and moderate spatial resolution (100’s <span><math><mrow><mi>μ</mi><mi>m</mi></mrow></math></span>). There is a trade-off between spatial and temporal resolution that limits accurate characterization of dynamics in rock features that are below spatial resolution of CT. While past literature has offered solutions to improve resolution of CT rock images, including deep learning-based algorithms, our study uniquely focuses on improving dynamic, partially and fully fluid-saturated geological images. Fluid-saturated CT images offer additional information, through augmented signals provided by the presence of fluid. Among challenges, CT images of geological media inherently possess limited information due to their single-channel gray-scale source. Additionally, fluid flows through partially saturated media frustrate existing super resolution techniques because unsaturated CT images are an inaccurate proxy for saturated dynamic rock images. The novelty of this work is the expansion of a generative adversarial network (GAN) for applications involving super resolution of partially saturated low resolution CT images using end-member, unsaturated high resolution <span><math><mi>μ</mi></math></span>CT images. To this end, we acquired multiscale low- and high-resolution CT rock images in unsaturated and saturated states. Among GAN and convolutional neural networks, GAN’s produce realistic high-resolution reconstructions of saturated geological media when trained using high-resolution, unsaturated images and lower resolution images in various saturation states. The model has direct usefulness for interpretation of real-time images.</div></div>","PeriodicalId":33804,"journal":{"name":"Applied Computing and Geosciences","volume":"24 ","pages":"Article 100208"},"PeriodicalIF":2.6,"publicationDate":"2024-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143137712","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0