Computers & Geosciences最新文献_第6页

PoreViT: Automated pore typing in carbonate rocks using vision transformers and neighborhood features PoreViT：利用视觉变压器和邻域特征对碳酸盐岩进行自动孔隙分型

IF 4.4 2区地球科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Geosciences

Pub Date : 2026-02-01 Epub Date: 2025-10-30 DOI: 10.1016/j.cageo.2025.106071

Yemna Qaiser , Mohammed Ishaq , Mohammed Yaqoob , Mohammed Yusuf Ansari , Isaac Sujay , Talha Khan , Harris Rabbani , Juan Carlos Laya , P.J. Moore , Thomas Daniel Seers

The classification of pores into their intrinsic depo-diagenetic or petrophysical morphotypes is a fundamental practice within carbonate petrography, providing linkage between pore-scale textures and their associated petrophysical signatures and/or paragenetic histories. Typically, pore classification is performed manually in a qualitative/semi-quantitative manner, which is hampered by inefficiency, subjectivity, and a lack of scalability. Though aimed at addressing the limitations of manual pore classification, efforts to automate petrographic pore-typing through artificial intelligence and computer vision techniques are limited by the inability of models to classify pores into genetic classes solely based upon simplistic size and shape features, which have been the focus of the existing literature. To address this nuanced classification problem, we present PoreViT: a Vision Transformer (ViT) model used to classify macropores observed in thin-sections into their respective Lucia classes (interparticle, touching vug, separate vug). The core novelty of PoreViT lies in its Feature Fusion block, which integrates ViT features, enhanced by a Global Token Addition layer, with spatial features extracted from a Convolutional Neural Network (CNN). Critically, our classifier leverages neighborhood information to provide the model with localized pore system topology, recognizing that pore types need to be identified not just by shape but also by their local spatial context. Trained and tested using 4115 labels obtained from 25 high-resolution thin-section scans, PoreViT provides an accurate, automated classification of carbonate macropores, achieving precision and recall values of 0.92 and 0.93 (macro-F1 0.92) corresponding to absolute improvements of ＋4.0% and ＋4.0%, and relative gains of ＋4.54% and ＋4.5%, respectively, over the best-performing CNN model (DenseNet121). The high throughput pore-textural classification capabilities demonstrated herein offer unprecedented opportunities in the integrated quantitative characterization of carbonates.

将孔隙划分为其内在的沉积成岩或岩石物理形态是碳酸盐岩岩石学的基本实践，它提供了孔隙尺度结构与其相关的岩石物理特征和/或共生历史之间的联系。通常，孔隙分类以定性/半定量的方式手动进行，效率低、主观性强、缺乏可扩展性。虽然旨在解决人工孔隙分类的局限性，但通过人工智能和计算机视觉技术自动化岩石学孔隙分型的努力受到模型无法仅根据简单的大小和形状特征将孔隙划分为遗传类的限制，这已经成为现有文献的重点。为了解决这个细致入微的分类问题，我们提出了PoreViT: Vision Transformer （ViT）模型，用于将薄切片中观察到的大孔隙分类为各自的Lucia类（颗粒间、接触空隙、分离空隙）。PoreViT的核心新颖之处在于其特征融合块，该块集成了由全局令牌添加层增强的ViT特征，以及从卷积神经网络（CNN）中提取的空间特征。关键的是，我们的分类器利用邻域信息为模型提供局部孔隙系统拓扑，认识到孔隙类型不仅需要通过形状来识别，还需要通过其局部空间环境来识别。使用从25个高分辨率薄层扫描中获得的4115个标签进行训练和测试，PoreViT提供了准确的、自动的碳酸盐大孔隙分类，与表现最好的CNN模型（DenseNet121）相比，精度和召回率分别达到0.92和0.93 (macro-F1 0.92)，分别提高了+4.0%和+4.0%，相对增益分别为+4.54%和+4.5%。本文所展示的高通量孔隙结构分类能力为碳酸盐的综合定量表征提供了前所未有的机会。

{"title":"PoreViT: Automated pore typing in carbonate rocks using vision transformers and neighborhood features","authors":"Yemna Qaiser , Mohammed Ishaq , Mohammed Yaqoob , Mohammed Yusuf Ansari , Isaac Sujay , Talha Khan , Harris Rabbani , Juan Carlos Laya , P.J. Moore , Thomas Daniel Seers","doi":"10.1016/j.cageo.2025.106071","DOIUrl":"10.1016/j.cageo.2025.106071","url":null,"abstract":"<div><div>The classification of pores into their intrinsic depo-diagenetic or petrophysical morphotypes is a fundamental practice within carbonate petrography, providing linkage between pore-scale textures and their associated petrophysical signatures and/or paragenetic histories. Typically, pore classification is performed manually in a qualitative/semi-quantitative manner, which is hampered by inefficiency, subjectivity, and a lack of scalability. Though aimed at addressing the limitations of manual pore classification, efforts to automate petrographic pore-typing through artificial intelligence and computer vision techniques are limited by the inability of models to classify pores into genetic classes solely based upon simplistic size and shape features, which have been the focus of the existing literature. To address this nuanced classification problem, we present PoreViT: a Vision Transformer (ViT) model used to classify macropores observed in thin-sections into their respective Lucia classes (interparticle, touching vug, separate vug). The core novelty of PoreViT lies in its Feature Fusion block, which integrates ViT features, enhanced by a Global Token Addition layer, with spatial features extracted from a Convolutional Neural Network (CNN). Critically, our classifier leverages neighborhood information to provide the model with localized pore system topology, recognizing that pore types need to be identified not just by shape but also by their local spatial context. Trained and tested using 4115 labels obtained from 25 high-resolution thin-section scans, PoreViT provides an accurate, automated classification of carbonate macropores, achieving precision and recall values of 0.92 and 0.93 (macro-F1 0.92) corresponding to absolute improvements of ＋4.0% and ＋4.0%, and relative gains of ＋4.54% and ＋4.5%, respectively, over the best-performing CNN model (DenseNet121). The high throughput pore-textural classification capabilities demonstrated herein offer unprecedented opportunities in the integrated quantitative characterization of carbonates.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"207 ","pages":"Article 106071"},"PeriodicalIF":4.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145467036","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

TorchTEM3D: PyTorch-Driven forward modeling platform for fast 3D transient electromagnetic modeling and efficient sensitivity matrix calculation TorchTEM3D: pytorch驱动的正演建模平台，用于快速3D瞬变电磁建模和高效灵敏度矩阵计算

IF 4.4 2区地球科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Geosciences

Pub Date : 2026-02-01 Epub Date: 2025-10-04 DOI: 10.1016/j.cageo.2025.106063

Ziteng Li , Hai Li , Keying Li , Ahmed M. Beshr

The three-dimensional (3D) forward modeling of transient electromagnetic (TEM) data is often computationally demanding due to its high complexity and limited hardware acceleration, which also affects the efficiency of sensitivity matrix calculation. In recent years, deep learning frameworks, particularly PyTorch, have been widely used in various fields due to their high flexibility, parallel computing capabilities, and powerful automatic differentiation function. In this paper, we develop a time-domain finite-difference forward modeling platform for 3D TEM, named TorchTEM3D, based on the powerful parallel computing and GPU acceleration capabilities of PyTorch. By fully utilizing the automatic differentiation function of PyTorch, we achieve efficient and fast calculation of sensitivity matrix (the gradient of the electromagnetic response to the geoelectric model). Compared with existing open-source Python computing platforms such as SimPEG and custEM, our method improves computing speed by 15–60 times. Furthermore, high-precision sensitivity matrices can be obtained with a single forward modeling run.

瞬变电磁（TEM）数据的三维正演建模由于其高复杂性和有限的硬件加速，往往需要大量的计算量，这也影响了灵敏度矩阵计算的效率。近年来，深度学习框架，特别是PyTorch，由于其高灵活性、并行计算能力和强大的自动微分功能，在各个领域得到了广泛的应用。本文基于PyTorch强大的并行计算和GPU加速能力，开发了三维瞬变电磁法时域有限差分正演建模平台TorchTEM3D。充分利用PyTorch的自动微分功能，实现了灵敏度矩阵（电磁响应对地电模型的梯度）的高效快速计算。与现有的开源Python计算平台SimPEG和custEM相比，我们的方法将计算速度提高了15-60倍。此外，单次正演模拟可以获得高精度的灵敏度矩阵。

{"title":"TorchTEM3D: PyTorch-Driven forward modeling platform for fast 3D transient electromagnetic modeling and efficient sensitivity matrix calculation","authors":"Ziteng Li , Hai Li , Keying Li , Ahmed M. Beshr","doi":"10.1016/j.cageo.2025.106063","DOIUrl":"10.1016/j.cageo.2025.106063","url":null,"abstract":"<div><div>The three-dimensional (3D) forward modeling of transient electromagnetic (TEM) data is often computationally demanding due to its high complexity and limited hardware acceleration, which also affects the efficiency of sensitivity matrix calculation. In recent years, deep learning frameworks, particularly PyTorch, have been widely used in various fields due to their high flexibility, parallel computing capabilities, and powerful automatic differentiation function. In this paper, we develop a time-domain finite-difference forward modeling platform for 3D TEM, named TorchTEM3D, based on the powerful parallel computing and GPU acceleration capabilities of PyTorch. By fully utilizing the automatic differentiation function of PyTorch, we achieve efficient and fast calculation of sensitivity matrix (the gradient of the electromagnetic response to the geoelectric model). Compared with existing open-source Python computing platforms such as SimPEG and custEM, our method improves computing speed by 15–60 times. Furthermore, high-precision sensitivity matrices can be obtained with a single forward modeling run.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"207 ","pages":"Article 106063"},"PeriodicalIF":4.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145269658","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Dual watermarking algorithm for trajectory data based on vector decomposition 基于矢量分解的轨迹数据双水印算法

IF 4.4 2区地球科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Geosciences

Pub Date : 2026-02-01 Epub Date: 2025-10-27 DOI: 10.1016/j.cageo.2025.106072

Heyan Wang , Luanyun Hu , Yuchen Hu , Changqing Zhu , Wang Zhang , Na Ren

Trajectory data security demands simultaneous copyright protection and spatiotemporal integrity authentication, yet existing watermarking methods struggle with irreconcilable trade-offs between robustness, reversibility, and temporally-aware tamper detection. Traditional approaches often introduce irreversible geometric distortions under affine transformations and lack mechanisms to validate timestamp authenticity, leaving critical vulnerabilities in precision-sensitive applications. This paper proposes a dual watermarking framework integrating geometric-invariant reversible embeddings and semi-fragile temporal authentication. The method decomposes trajectory coordinates into affine-invariant α/β coefficients using vector decomposition, enabling high-capacity copyright watermark embedding resilient to geometric attacks. Temporal integrity verification is achieved through Pearson-correlation analysis of timestamp sequences, detecting malicious deletions with complete accuracy while tolerating legitimate temporal shifts. A bidirectional quantization modulation scheme guarantees lossless coordinate recovery, reducing restoration errors to sub-microradian precision by synchronizing geometric invariants with relative distance preservation. Comprehensive evaluations across 68,632 real-world trajectories demonstrate superior performance: the framework achieves normalized correlations above 0.94 against 45° rotation/200 % scaling attacks, 21 % higher robustness than state-of-the-art methods under compression, and full-error detection of timestamp tampering. By unifying geometric invariance, temporal causality verification, and reversible modulation within a single architecture, this work establishes a new paradigm for dual-function security in spatiotemporal data management, with direct applicability to GIS.

轨迹数据安全需要同时保护版权和时空完整性认证，然而现有的水印方法在鲁棒性、可逆性和时间感知篡改检测之间难以调和。传统的方法通常会在仿射变换下引入不可逆的几何扭曲，并且缺乏验证时间戳真实性的机制，这在精度敏感的应用中留下了严重的漏洞。提出了一种结合几何不变可逆嵌入和半脆弱时间认证的双水印框架。该方法利用矢量分解方法将轨迹坐标分解为仿射不变的α/β系数，使高容量版权水印嵌入能够抵御几何攻击。时间完整性验证是通过时间戳序列的皮尔逊相关分析来实现的，在允许合法时间偏移的同时，可以完全准确地检测恶意删除。双向量化调制方案保证了无损的坐标恢复，通过同步几何不变量和相对距离保持将恢复误差降低到亚微弧度精度。对68,632个真实世界轨迹的综合评估显示了卓越的性能：该框架在45°旋转/ 200%缩放攻击下实现了0.94以上的归一化相关性，比最先进的压缩方法高21%的鲁棒性，以及对时间戳篡改的全错误检测。通过在单一架构中统一几何不变性、时间因果验证和可逆调制，本工作建立了时空数据管理双功能安全的新范式，可直接适用于GIS。

{"title":"Dual watermarking algorithm for trajectory data based on vector decomposition","authors":"Heyan Wang , Luanyun Hu , Yuchen Hu , Changqing Zhu , Wang Zhang , Na Ren","doi":"10.1016/j.cageo.2025.106072","DOIUrl":"10.1016/j.cageo.2025.106072","url":null,"abstract":"<div><div>Trajectory data security demands simultaneous copyright protection and spatiotemporal integrity authentication, yet existing watermarking methods struggle with irreconcilable trade-offs between robustness, reversibility, and temporally-aware tamper detection. Traditional approaches often introduce irreversible geometric distortions under affine transformations and lack mechanisms to validate timestamp authenticity, leaving critical vulnerabilities in precision-sensitive applications. This paper proposes a dual watermarking framework integrating geometric-invariant reversible embeddings and semi-fragile temporal authentication. The method decomposes trajectory coordinates into affine-invariant α/β coefficients using vector decomposition, enabling high-capacity copyright watermark embedding resilient to geometric attacks. Temporal integrity verification is achieved through Pearson-correlation analysis of timestamp sequences, detecting malicious deletions with complete accuracy while tolerating legitimate temporal shifts. A bidirectional quantization modulation scheme guarantees lossless coordinate recovery, reducing restoration errors to sub-microradian precision by synchronizing geometric invariants with relative distance preservation. Comprehensive evaluations across 68,632 real-world trajectories demonstrate superior performance: the framework achieves normalized correlations above 0.94 against 45° rotation/200 % scaling attacks, 21 % higher robustness than state-of-the-art methods under compression, and full-error detection of timestamp tampering. By unifying geometric invariance, temporal causality verification, and reversible modulation within a single architecture, this work establishes a new paradigm for dual-function security in spatiotemporal data management, with direct applicability to GIS.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"207 ","pages":"Article 106072"},"PeriodicalIF":4.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145417519","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Implicit neural representations for 3D gravity inversion 三维重力反演的隐式神经表示

IF 4.4 2区地球科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Geosciences

Pub Date : 2026-02-01 Epub Date: 2025-12-02 DOI: 10.1016/j.cageo.2025.106082

Xiong Li , Jingtao Zhao , Shuai Zhou

Three-dimensional gravity inversion is vital for recovering subsurface density distribution, but resolving sparse geological anomalies with complex geometries from superimposed gravity signals remains difficult due to severe ill-posedness. Existing methods often face limitations in representational flexibility or rely heavily on predefined regularization strategies. We introduce the physics-driven Implicit Gravity Inversion (IGI) method to reconstruct the underground density distribution in geophysical exploration, which employs a neural network to implicitly represent the continuous density field directly from spatial coordinates. Critically, IGI is driven by physical laws embedded within the loss function, optimizing the neural network to predict a density field consistent with observed gravity anomalies, thus eliminating the need for extensive labeled training data typical of deep learning. Results show that IGI provides comparable anomaly distributions to those obtained using deep learning inversion. Ablation studies reveal that parameter coupling provides effective depth weighting, indicating that architectural choices control regularization strength within well-defined optimal ranges. Extensive experiments on synthetic benchmarks and the real-world San Nicolas deposit dataset demonstrate that IGI robustly recovers sparse and complex density models with high accuracy, effectively delineating intricate structures such as disconnected ore bodies and sharp contacts, surpassing traditional methods in these aspects.

三维重力反演对于恢复地下密度分布至关重要，但由于严重的病态性，从叠加重力信号中解决具有复杂几何形状的稀疏地质异常仍然很困难。现有的方法往往面临表示灵活性的限制，或者严重依赖于预定义的正则化策略。引入物理驱动的隐式重力反演（IGI）方法，利用神经网络直接从空间坐标隐式表示连续的密度场，重建地球物理勘探中的地下密度分布。关键是，IGI由嵌入损失函数中的物理定律驱动，优化神经网络以预测与观测到的重力异常一致的密度场，从而消除了对深度学习典型的大量标记训练数据的需求。结果表明，IGI提供的异常分布与使用深度学习反演获得的异常分布相当。消融研究表明，参数耦合提供了有效的深度加权，表明建筑选择将正则化强度控制在明确的最佳范围内。在合成基准和真实的San Nicolas矿床数据集上进行的大量实验表明，IGI能够以高精度鲁棒地恢复稀疏和复杂的密度模型，有效地描绘出分离矿体和尖锐接触等复杂结构，在这些方面超越了传统方法。

{"title":"Implicit neural representations for 3D gravity inversion","authors":"Xiong Li , Jingtao Zhao , Shuai Zhou","doi":"10.1016/j.cageo.2025.106082","DOIUrl":"10.1016/j.cageo.2025.106082","url":null,"abstract":"<div><div>Three-dimensional gravity inversion is vital for recovering subsurface density distribution, but resolving sparse geological anomalies with complex geometries from superimposed gravity signals remains difficult due to severe ill-posedness. Existing methods often face limitations in representational flexibility or rely heavily on predefined regularization strategies. We introduce the physics-driven Implicit Gravity Inversion (IGI) method to reconstruct the underground density distribution in geophysical exploration, which employs a neural network to implicitly represent the continuous density field directly from spatial coordinates. Critically, IGI is driven by physical laws embedded within the loss function, optimizing the neural network to predict a density field consistent with observed gravity anomalies, thus eliminating the need for extensive labeled training data typical of deep learning. Results show that IGI provides comparable anomaly distributions to those obtained using deep learning inversion. Ablation studies reveal that parameter coupling provides effective depth weighting, indicating that architectural choices control regularization strength within well-defined optimal ranges. Extensive experiments on synthetic benchmarks and the real-world San Nicolas deposit dataset demonstrate that IGI robustly recovers sparse and complex density models with high accuracy, effectively delineating intricate structures such as disconnected ore bodies and sharp contacts, surpassing traditional methods in these aspects.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"208 ","pages":"Article 106082"},"PeriodicalIF":4.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145694905","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Optimising the computational performance of high degree lithospheric field models 高阶岩石圈场模型的计算性能优化

IF 4.4 2区地球科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Geosciences

Pub Date : 2026-02-01 Epub Date: 2025-12-10 DOI: 10.1016/j.cageo.2025.106092

Michael Bareford , William Brown , Ciarán Beggan , Callum Watson , Mark Bull

The British Geological Survey (BGS) World Magnetic Anomaly Model (WMAM) code calculates spherical harmonic models of the natural magnetisation of the rocks of Earth’s crust. These models allow us to estimate the value of the full magnetic field vector at any location, based on scattered pointwise marine or aero-magnetic measurements of only the scalar magnetic field. Modelling the magnetic field in this way serves many important purposes, such as geological research, navigation and safe resource extraction. Global spherical harmonic models of degree and order 1440 (

\sim

28 km spatial resolution) have been successfully computed on the HPC facilities local to BGS, but such runs require nearly the full compute capacity for multiple days. Further, the available resolution of the scalar field measurements is too high to be fully exploited by the WMAM code, limiting models of the crustal magnetic field to a resolution of 28 km. To overcome these issues, we refactored the WMAM code such that models of spherical harmonic degree 1440 and 2000 (

\sim

20 km resolution) can be produced in hours rather than days. For example, a degree 2000 model was calculated using 64 HPE Cray EX nodes (8 192 cores) in 3 h and 44 mins. The resulting model power spectra and magnetic field maps showed excellent agreement with the existing degree 1440 model and the original input data. The performance of the WMAM code was further improved via offloading to GPU. We show the improvements due to GPU acceleration in terms of energy consumption as well as runtime. This fruitful collaboration between experts in the fields of Geoscience (BGS) and HPC (EPCC) has created the opportunity for the WMAM code to be used to gain new knowledge about crustal magnetic fields.

英国地质调查局（BGS）的世界磁异常模型（WMAM）代码计算了地壳岩石自然磁化的球谐模型。这些模型使我们能够估计在任何位置的全磁场矢量的值，基于散射的点对标量磁场的海洋或航空磁测量。以这种方式模拟磁场有许多重要用途，如地质研究、航海和安全资源开采。1440度和1440阶（~ 28 km空间分辨率）的全球球面调和模型已经在BGS本地的高性能计算设备上成功计算，但这样的运行几乎需要数天的全部计算能力。此外，标量场测量的可用分辨率太高，无法被WMAM代码充分利用，将地壳磁场模型限制在28公里的分辨率。为了克服这些问题，我们重构了WMAM代码，使球谐度1440和2000（~ 20公里分辨率）的模型可以在几小时内产生，而不是几天。例如，使用64个HPE Cray EX节点（8 192个内核）在3小时44分钟内计算了2000度模型。所得模型功率谱和磁场图与现有1440度模型和原始输入数据吻合良好。WMAM代码的性能通过卸载到GPU进一步提高。我们展示了由于GPU加速在能耗和运行时间方面的改进。地球科学（BGS）和HPC （EPCC）领域专家之间富有成效的合作为WMAM代码用于获取有关地壳磁场的新知识创造了机会。

{"title":"Optimising the computational performance of high degree lithospheric field models","authors":"Michael Bareford , William Brown , Ciarán Beggan , Callum Watson , Mark Bull","doi":"10.1016/j.cageo.2025.106092","DOIUrl":"10.1016/j.cageo.2025.106092","url":null,"abstract":"<div><div>The British Geological Survey (BGS) World Magnetic Anomaly Model (WMAM) code calculates spherical harmonic models of the natural magnetisation of the rocks of Earth’s crust. These models allow us to estimate the value of the full magnetic field vector at any location, based on scattered pointwise marine or aero-magnetic measurements of only the scalar magnetic field. Modelling the magnetic field in this way serves many important purposes, such as geological research, navigation and safe resource extraction. Global spherical harmonic models of degree and order 1440 (<span><math><mo>∼</mo></math></span>28 km spatial resolution) have been successfully computed on the HPC facilities local to BGS, but such runs require nearly the full compute capacity for multiple days. Further, the available resolution of the scalar field measurements is too high to be fully exploited by the WMAM code, limiting models of the crustal magnetic field to a resolution of 28 km. To overcome these issues, we refactored the WMAM code such that models of spherical harmonic degree 1440 and 2000 (<span><math><mo>∼</mo></math></span>20 km resolution) can be produced in hours rather than days. For example, a degree 2000 model was calculated using 64 HPE Cray EX nodes (8192 cores) in 3 h and 44 mins. The resulting model power spectra and magnetic field maps showed excellent agreement with the existing degree 1440 model and the original input data. The performance of the WMAM code was further improved via offloading to GPU. We show the improvements due to GPU acceleration in terms of energy consumption as well as runtime. This fruitful collaboration between experts in the fields of Geoscience (BGS) and HPC (EPCC) has created the opportunity for the WMAM code to be used to gain new knowledge about crustal magnetic fields.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"208 ","pages":"Article 106092"},"PeriodicalIF":4.4,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145791272","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Research on hyperspectral remote sensing alteration mineral mapping using an improved ViT model 基于改进ViT模型的高光谱遥感蚀变矿物填图研究

IF 4.4 2区地球科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Geosciences

Pub Date : 2026-01-01 Epub Date: 2025-08-09 DOI: 10.1016/j.cageo.2025.106037

Xinhang Feng , Jiejun Huang , Ximing Chen , Han Zhou , Ming Zhang , Chuan Zhang , Fawang Ye

The distribution of altered minerals is a key indicator for finding strategic minerals such as uranium, cobalt, nickel, copper and zinc. In recent years, deep learning has shown outstanding advantages in the field of hyperspectral altered mineral mapping. However, constructing a large volume of high-quality training samples remains time-consuming and labor-intensive. Moreover, many models suffer from limited generalization capability—performing well on training data but exhibiting significant performance degradation on test datasets or in real-world applications. Therefore, a semi-automatic sample construction method was proposed. The sample construction involves three steps. Firstly, using mixed pixel decomposition to extract mineral abundance, then screening samples via mixed matching, and finally enhancing classification accuracy with spectral characteristic quantification. Experimental results show that the test accuracy of the dataset generated by the semi-automated method on the ViT model reached 92.81 %, which is close to that of manually labeled samples at 93.29 %. In terms of models, an improved Vision Transformer (ViT) model was proposed. The SpecPool-Transformer model (SPT) integrates the Grouped Spectral Embedding Module (GSE) and the Convolution-Pooling Module (CPM) to enhance the extraction of adjacent band features from the spectral curves. Additionally, the model's application to cross-source data was achieved through transfer learning. On the SASI dataset of the Baiyanghe uranium deposit, the overall accuracy (OA) and average accuracy (AA) of SpecPool-Transformer reached 96.76 % and 95.14 %, respectively, representing improvements of 3.95 % and 6.11 % over the original ViT model. In the generalization test, the proposed method achieved an OA of 86.10 % and an AA of 83.74 % on the SASI aerial dataset No.1007, outperforming the second-best model, LightGBM, by 20.22 % and 31.15 %, respectively. Field validation results further confirm the high reliability of the proposed model in large-scale alteration mineral mapping across data sources, making it suitable for rapid and extensive alteration mineral mapping applications.

蚀变矿物的分布是寻找诸如铀、钴、镍、铜和锌等战略矿物的关键指标。近年来，深度学习在高光谱蚀变矿物填图领域显示出突出的优势。然而，构建大量高质量的训练样本仍然是耗时和劳动密集型的。此外，许多模型的泛化能力有限——在训练数据上表现良好，但在测试数据集或实际应用中表现出明显的性能下降。为此，提出了一种半自动取样方法。示例构建包括三个步骤。首先利用混合像元分解提取矿物丰度，然后通过混合匹配筛选样品，最后利用光谱特征量化提高分类精度。实验结果表明，半自动化方法在ViT模型上生成的数据集的测试准确率达到92.81%，接近人工标记样本的93.29%。在模型方面，提出了一种改进的视觉变换器（ViT）模型。SpecPool-Transformer模型（SPT）集成了分组频谱嵌入模块（GSE）和卷积池化模块（CPM），增强了对光谱曲线邻带特征的提取。此外，通过迁移学习实现了模型对跨源数据的应用。在白洋河铀矿床SASI数据集上，SpecPool-Transformer模型的总体精度（OA）和平均精度（AA）分别达到96.76%和95.14%，比原ViT模型分别提高了3.95%和6.11%。在SASI航空数据集No.1007的泛化测试中，该方法的OA为86.10%，AA为83.74%，分别比第二优模型LightGBM高出20.22%和31.15%。现场验证结果进一步证实了该模型在跨数据源的大规模蚀变矿物填图中的高可靠性，适用于快速、广泛的蚀变矿物填图应用。

{"title":"Research on hyperspectral remote sensing alteration mineral mapping using an improved ViT model","authors":"Xinhang Feng , Jiejun Huang , Ximing Chen , Han Zhou , Ming Zhang , Chuan Zhang , Fawang Ye","doi":"10.1016/j.cageo.2025.106037","DOIUrl":"10.1016/j.cageo.2025.106037","url":null,"abstract":"<div><div>The distribution of altered minerals is a key indicator for finding strategic minerals such as uranium, cobalt, nickel, copper and zinc. In recent years, deep learning has shown outstanding advantages in the field of hyperspectral altered mineral mapping. However, constructing a large volume of high-quality training samples remains time-consuming and labor-intensive. Moreover, many models suffer from limited generalization capability—performing well on training data but exhibiting significant performance degradation on test datasets or in real-world applications. Therefore, a semi-automatic sample construction method was proposed. The sample construction involves three steps. Firstly, using mixed pixel decomposition to extract mineral abundance, then screening samples via mixed matching, and finally enhancing classification accuracy with spectral characteristic quantification. Experimental results show that the test accuracy of the dataset generated by the semi-automated method on the ViT model reached 92.81 %, which is close to that of manually labeled samples at 93.29 %. In terms of models, an improved Vision Transformer (ViT) model was proposed. The SpecPool-Transformer model (SPT) integrates the Grouped Spectral Embedding Module (GSE) and the Convolution-Pooling Module (CPM) to enhance the extraction of adjacent band features from the spectral curves. Additionally, the model's application to cross-source data was achieved through transfer learning. On the SASI dataset of the Baiyanghe uranium deposit, the overall accuracy (OA) and average accuracy (AA) of SpecPool-Transformer reached 96.76 % and 95.14 %, respectively, representing improvements of 3.95 % and 6.11 % over the original ViT model. In the generalization test, the proposed method achieved an OA of 86.10 % and an AA of 83.74 % on the SASI aerial dataset No.1007, outperforming the second-best model, LightGBM, by 20.22 % and 31.15 %, respectively. Field validation results further confirm the high reliability of the proposed model in large-scale alteration mineral mapping across data sources, making it suitable for rapid and extensive alteration mineral mapping applications.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"206 ","pages":"Article 106037"},"PeriodicalIF":4.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144865464","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Simulating major element diffusion in garnet using realistic 3D geometries 利用真实的三维几何图形模拟石榴石中主要元素的扩散

IF 4.4 2区地球科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Geosciences

Pub Date : 2026-01-01 Epub Date: 2025-08-08 DOI: 10.1016/j.cageo.2025.106023

Hugo Dominguez , Nathan Mäder , Pierre Lanari

Chemical diffusion of major elements in garnet is a common phenomenon in amphibolite to granulite facies metamorphic rocks. The study of this process has led to important constraints on the rate and timescale of metamorphism, for instance using geospeedometry and forward thermodynamic modelling. However, to date, most models have assumed spherical coordinates and simple geometries when modelling diffusion in garnet. In this study, we present a framework for running 3D multicomponent diffusion models from real grain geometries obtained by micro-computed tomography. We introduce an open-source code, DiffusionGarnet.jl, written for high performance in the Julia programming language. We demonstrate the high efficiency of the numerical solver, a stabilised explicit method, and its scalability using GPU acceleration. This approach is applied to two garnet grains with different characteristics, a euhedral well-shaped grain and a deformed sub-euhedral grain with a high connectivity to the matrix from core to rim. Starting from a similar initial composition and at constant conditions of 700 °C and 0.8 GPa for 10 Myr, the models show results with very different characteristics. The euhedral grain shows results similar to those predicted with a spherical assumption, largely preserving its original zoning. In contrast, the sub-euhedral grain shows significant re-equilibration, nearly erasing completely its initial zoning. This behaviour is caused by the high connectivity with the matrix. In addition to providing a robust solver for 3D diffusion modelling, these results demonstrate the role of grain geometry and matrix connectivity on intra-grain diffusion and highlight the power of 3D approaches to properly study the complexity of natural grains.

石榴石中主要元素的化学扩散是角闪岩-麻粒岩相变质岩中普遍存在的现象。对这一过程的研究导致了对变质作用速率和时间尺度的重要限制，例如使用地质测速法和正演热力学模型。然而，到目前为止，大多数模型在模拟石榴石中的扩散时都假设了球坐标和简单的几何形状。在这项研究中，我们提出了一个运行三维多组分扩散模型的框架，该模型是由微观计算机断层扫描获得的真实晶粒几何形状。我们介绍一个开源代码，DiffusionGarnet。jl，用Julia编程语言编写的高性能。我们证明了数值求解器的高效率，稳定的显式方法，以及使用GPU加速的可扩展性。该方法应用于两种具有不同特征的石榴石晶粒，一种是自面体的井形晶粒，另一种是变形的亚自面体晶粒，从岩心到边缘与基体的连通性较高。从相似的初始成分出发，在700°C和0.8 GPa / 10 Myr的恒定条件下，模型显示出非常不同的特征。自面形晶粒的结果与球形假设的预测结果相似，在很大程度上保留了其原有的分带。相比之下，亚自面体晶粒表现出明显的再平衡，几乎完全消除了其最初的分带。这种行为是由与矩阵的高连通性引起的。除了为三维扩散建模提供鲁棒解算器外，这些结果还证明了晶粒几何形状和矩阵连通性对晶粒内扩散的作用，并突出了3D方法在正确研究天然晶粒复杂性方面的作用。

{"title":"Simulating major element diffusion in garnet using realistic 3D geometries","authors":"Hugo Dominguez , Nathan Mäder , Pierre Lanari","doi":"10.1016/j.cageo.2025.106023","DOIUrl":"10.1016/j.cageo.2025.106023","url":null,"abstract":"<div><div>Chemical diffusion of major elements in garnet is a common phenomenon in amphibolite to granulite facies metamorphic rocks. The study of this process has led to important constraints on the rate and timescale of metamorphism, for instance using geospeedometry and forward thermodynamic modelling. However, to date, most models have assumed spherical coordinates and simple geometries when modelling diffusion in garnet. In this study, we present a framework for running 3D multicomponent diffusion models from real grain geometries obtained by micro-computed tomography. We introduce an open-source code, DiffusionGarnet.jl, written for high performance in the Julia programming language. We demonstrate the high efficiency of the numerical solver, a stabilised explicit method, and its scalability using GPU acceleration. This approach is applied to two garnet grains with different characteristics, a euhedral well-shaped grain and a deformed sub-euhedral grain with a high connectivity to the matrix from core to rim. Starting from a similar initial composition and at constant conditions of 700 °C and 0.8 GPa for 10 Myr, the models show results with very different characteristics. The euhedral grain shows results similar to those predicted with a spherical assumption, largely preserving its original zoning. In contrast, the sub-euhedral grain shows significant re-equilibration, nearly erasing completely its initial zoning. This behaviour is caused by the high connectivity with the matrix. In addition to providing a robust solver for 3D diffusion modelling, these results demonstrate the role of grain geometry and matrix connectivity on intra-grain diffusion and highlight the power of 3D approaches to properly study the complexity of natural grains.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"206 ","pages":"Article 106023"},"PeriodicalIF":4.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144829937","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Seismic random noise attenuation using structure-oriented 3D curvelet transform 面向结构的三维曲线变换地震随机噪声衰减

IF 4.4 2区地球科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Geosciences

Pub Date : 2026-01-01 Epub Date: 2025-08-09 DOI: 10.1016/j.cageo.2025.106020

Minggui Liang , Shaohuan Zu , Zifei Li , Wenlu Liu , Haojun Chen , Zhengyu Tan

Sparse constraints based on curvelet transform have been widely applied in seismic noise suppression. Traditional schemes usually adopt global or multi-scale thresholding to constrain the coefficients corresponding to noise, which may ignore the structural characteristics and result in signal distortion. It is usually challenging to balance noise suppression and signal preservation. To address this problem, we develop a structure-oriented 3D denoising method based on the 3D curvelet transform, which uses dip information to analyze the complexity of local features. The non-stationary thresholding scheme based on local complexity is implemented, providing strong constraints for low-complexity data to suppress noise and weak constraints for high-complexity data to protect signal. Furthermore, the coarse scale coefficients are insensitive to noise interference; only the fine-scales coefficients are constrained in our proposed scheme. Numerical tests on synthetic and field data demonstrate that the proposed method has better denoising performance for large dip feature data than the conventional global and multi-scale thresholding schemes.

基于曲线变换的稀疏约束在地震噪声抑制中得到了广泛的应用。传统的方案通常采用全局或多尺度阈值来约束噪声对应的系数，这可能会忽略结构特征，导致信号失真。在噪声抑制和信号保持之间取得平衡通常是一个挑战。为了解决这一问题，我们开发了一种基于三维曲线变换的面向结构的三维去噪方法，该方法利用倾角信息分析局部特征的复杂性。实现了基于局部复杂度的非平稳阈值方案，对低复杂度数据提供强约束抑制噪声，对高复杂度数据提供弱约束保护信号。此外，粗尺度系数对噪声干扰不敏感；在我们提出的方案中，只有精细尺度的系数受到约束。综合数据和现场数据的数值试验表明，该方法对大倾角特征数据的去噪性能优于传统的全局和多尺度阈值方法。

{"title":"Seismic random noise attenuation using structure-oriented 3D curvelet transform","authors":"Minggui Liang , Shaohuan Zu , Zifei Li , Wenlu Liu , Haojun Chen , Zhengyu Tan","doi":"10.1016/j.cageo.2025.106020","DOIUrl":"10.1016/j.cageo.2025.106020","url":null,"abstract":"<div><div>Sparse constraints based on curvelet transform have been widely applied in seismic noise suppression. Traditional schemes usually adopt global or multi-scale thresholding to constrain the coefficients corresponding to noise, which may ignore the structural characteristics and result in signal distortion. It is usually challenging to balance noise suppression and signal preservation. To address this problem, we develop a structure-oriented 3D denoising method based on the 3D curvelet transform, which uses dip information to analyze the complexity of local features. The non-stationary thresholding scheme based on local complexity is implemented, providing strong constraints for low-complexity data to suppress noise and weak constraints for high-complexity data to protect signal. Furthermore, the coarse scale coefficients are insensitive to noise interference; only the fine-scales coefficients are constrained in our proposed scheme. Numerical tests on synthetic and field data demonstrate that the proposed method has better denoising performance for large dip feature data than the conventional global and multi-scale thresholding schemes.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"206 ","pages":"Article 106020"},"PeriodicalIF":4.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144860331","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Large-scale 3-D magnetotelluric modeling in anisotropic media using extrapolation multigrid method on staggered grids 交错网格外推多网格法在各向异性介质中的大尺度三维大地电磁模拟

IF 4.4 2区地球科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Geosciences

Pub Date : 2026-01-01 Epub Date: 2025-08-18 DOI: 10.1016/j.cageo.2025.106019

Jinxuan Wang , Kejia Pan , Hongzhu Cai , Zhengguang Liu , Xu Han , Weiwei Ling

To improve the practicality and efficiency of 3D magnetotelluric (MT) data inversion, developing a 3D MT forward modeling algorithm with low computational cost in terms of time and memory is an important prerequisite. An extrapolation cascadic multigrid (EXCMG) method is developed on rectilinear grids to accelerate the solving process of large linear systems arising from the staggered-grid finite difference (SFD) discretization of Maxwell’s equations. Arbitrary anisotropic conductivity is considered, without adding extra unknowns to the SFD scheme. A new prolongation operator based on global extrapolation and mixed-order interpolation is developed to tackle the issue caused by non-nested unknown distribution. The divergence correction scheme for arbitrary anisotropy is employed to stabilize the smoothing process, especially for low-frequency cases. Several examples are tested to validate the accuracy and efficiency of the proposed algorithm, including synthetic models with anisotropy and topography, and the real-world Cascadia model. Results show that our EXCMG solver is more efficient than traditional iterative solvers (e.g., the preconditioned BiCGStab), the algebraic multigrid method and the geometric multigrid (GMG) method. The proposed method can efficiently solve large-scale problems with large grid stretching factors and arbitrary anisotropy, providing powerful engine for large-scale MT inversion.

为了提高大地电磁数据三维反演的实用性和效率，开发一种计算成本低、存储时间短的大地电磁三维正演算法是一个重要的前提。为了加速麦克斯韦方程组交错网格有限差分（SFD）离散引起的大型线性方程组的求解过程，在直线网格上提出了一种外推叶栅多重网格（EXCMG）方法。考虑任意各向异性电导率，而不向SFD方案添加额外的未知数。针对非嵌套未知分布的问题，提出了一种基于全局外推和混合阶内插的扩展算子。采用任意各向异性的散度校正方案来稳定平滑过程，特别是在低频情况下。通过对具有各向异性和地形的合成模型以及实际的Cascadia模型进行了测试，验证了该算法的准确性和效率。结果表明，我们的EXCMG求解器比传统的迭代求解器（如预处理的BiCGStab）、代数多重网格法和几何多重网格法（GMG）更有效。该方法可以有效地解决具有大网格拉伸因子和任意各向异性的大规模问题，为大规模大地电磁法反演提供了强大的引擎。

{"title":"Large-scale 3-D magnetotelluric modeling in anisotropic media using extrapolation multigrid method on staggered grids","authors":"Jinxuan Wang , Kejia Pan , Hongzhu Cai , Zhengguang Liu , Xu Han , Weiwei Ling","doi":"10.1016/j.cageo.2025.106019","DOIUrl":"10.1016/j.cageo.2025.106019","url":null,"abstract":"<div><div>To improve the practicality and efficiency of 3D magnetotelluric (MT) data inversion, developing a 3D MT forward modeling algorithm with low computational cost in terms of time and memory is an important prerequisite. An extrapolation cascadic multigrid (EXCMG) method is developed on rectilinear grids to accelerate the solving process of large linear systems arising from the staggered-grid finite difference (SFD) discretization of Maxwell’s equations. Arbitrary anisotropic conductivity is considered, without adding extra unknowns to the SFD scheme. A new prolongation operator based on global extrapolation and mixed-order interpolation is developed to tackle the issue caused by non-nested unknown distribution. The divergence correction scheme for arbitrary anisotropy is employed to stabilize the smoothing process, especially for low-frequency cases. Several examples are tested to validate the accuracy and efficiency of the proposed algorithm, including synthetic models with anisotropy and topography, and the real-world Cascadia model. Results show that our EXCMG solver is more efficient than traditional iterative solvers (e.g., the preconditioned BiCGStab), the algebraic multigrid method and the geometric multigrid (GMG) method. The proposed method can efficiently solve large-scale problems with large grid stretching factors and arbitrary anisotropy, providing powerful engine for large-scale MT inversion.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"206 ","pages":"Article 106019"},"PeriodicalIF":4.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144886346","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Controlled latent diffusion models for 3D porous media reconstruction 三维多孔介质重建的可控潜扩散模型

IF 4.4 2区地球科学 Q1 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Geosciences

Pub Date : 2026-01-01 Epub Date: 2025-08-26 DOI: 10.1016/j.cageo.2025.106038

Danilo Naiff , Bernardo P. Schaeffer , Gustavo Pires , Dragan Stojkovic , Thomas Rapstine , Fabio Ramos

Three-dimensional digital reconstruction of porous media presents a fundamental challenge in geosciences, requiring simultaneous resolution of fine-scale pore structures while capturing representative elementary volumes. This work introduces a computational framework that addresses this challenge through latent diffusion models operating within the Elucidated Diffusion Models (EDM) framework. The proposed approach reduces dimensionality via a custom Variational Autoencoder trained in binary geological volumes, improving efficiency and also enabling the generation of larger volumes than previously possible with diffusion models. A key innovation is the controlled unconditional sampling methodology, which enhances distribution coverage by first sampling target statistics from their empirical distributions, and then generating samples conditioned on these values. Extensive testing on four distinct rock types demonstrates that conditioning on porosity – a readily computable statistic – is sufficient to ensure a consistent representation of multiple complex properties, including permeability, two-point correlation functions, and pore size distributions. The framework achieves better generation quality than pixel-space diffusion while enabling significantly larger volume reconstruction (

25 6^{3}

voxels) with substantially reduced computational requirements, establishing a new state-of-the-art for digital rock physics applications.

多孔介质的三维数字重建是地球科学领域的一个基本挑战，它需要在捕获代表性基本体积的同时，对细尺度孔隙结构进行分辨率处理。这项工作引入了一个计算框架，通过在阐明扩散模型（EDM）框架内运行的潜在扩散模型来解决这一挑战。该方法通过自定义的变分自编码器（Variational Autoencoder）减少了二进制地质体积的维数，提高了效率，并且能够生成比以前使用扩散模型更大的体积。一个关键的创新是受控无条件抽样方法，它通过首先从其经验分布中抽样目标统计量，然后根据这些值生成样本来增强分布覆盖率。对四种不同岩石类型的广泛测试表明，孔隙度（一个易于计算的统计数据）的调节足以确保多种复杂属性的一致表示，包括渗透率、两点相关函数和孔隙大小分布。该框架实现了比像素空间扩散更好的生成质量，同时实现了更大的体积重建（2563体素），大大减少了计算需求，为数字岩石物理应用建立了新的技术水平。

{"title":"Controlled latent diffusion models for 3D porous media reconstruction","authors":"Danilo Naiff , Bernardo P. Schaeffer , Gustavo Pires , Dragan Stojkovic , Thomas Rapstine , Fabio Ramos","doi":"10.1016/j.cageo.2025.106038","DOIUrl":"10.1016/j.cageo.2025.106038","url":null,"abstract":"<div><div>Three-dimensional digital reconstruction of porous media presents a fundamental challenge in geosciences, requiring simultaneous resolution of fine-scale pore structures while capturing representative elementary volumes. This work introduces a computational framework that addresses this challenge through latent diffusion models operating within the Elucidated Diffusion Models (EDM) framework. The proposed approach reduces dimensionality via a custom Variational Autoencoder trained in binary geological volumes, improving efficiency and also enabling the generation of larger volumes than previously possible with diffusion models. A key innovation is the controlled unconditional sampling methodology, which enhances distribution coverage by first sampling target statistics from their empirical distributions, and then generating samples conditioned on these values. Extensive testing on four distinct rock types demonstrates that conditioning on porosity – a readily computable statistic – is sufficient to ensure a consistent representation of multiple complex properties, including permeability, two-point correlation functions, and pore size distributions. The framework achieves better generation quality than pixel-space diffusion while enabling significantly larger volume reconstruction (<span><math><mrow><mn>25</mn><msup><mrow><mn>6</mn></mrow><mrow><mn>3</mn></mrow></msup></mrow></math></span> voxels) with substantially reduced computational requirements, establishing a new state-of-the-art for digital rock physics applications.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"206 ","pages":"Article 106038"},"PeriodicalIF":4.4,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144907447","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0