Computers & Geosciences最新文献

{"title":"The multi-GPU Wetland DEM Ponding Model","authors":"Tonghe Liu ,&nbsp;Sean J. Trim ,&nbsp;Seok-Bum Ko ,&nbsp;Raymond J. Spiteri","doi":"10.1016/j.cageo.2025.105912","DOIUrl":"10.1016/j.cageo.2025.105912","url":null,"abstract":"<div><div>The Wetland DEM (Digital Elevation Model) Ponding Model (<span>WDPM</span>) is software that simulates how runoff water is distributed across the Canadian Prairies. Previous versions of the <span>WDPM</span> are able to run in parallel with a single CPU or GPU. Now that multi-device parallel computing has become an established method to increase computational throughput and efficiency, this study extends <span>WDPM</span> to a multi-GPU parallel algorithm with efficient data transmission methods via overlapping communication with computation. The new implementation is evaluated from several perspectives. First, the output summary and system are compared with the previous implementation to verify correctness and demonstrate convergence. Second, the multi-GPU code is profiled, showing that the algorithm carries out efficient data synchronization through optimized techniques. Finally, the new implementation was tested experimentally and showed improved performance and good scaling. Specifically, a speedup of 2.39 was achieved when using four GPUs compared to using one GPU.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"199 ","pages":"Article 105912"},"PeriodicalIF":4.2,"publicationDate":"2025-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143619687","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Quantitative bed-type classification for a global comparison of deep-water sedimentary systems","authors":"Soma Budai ,&nbsp;Luca Colombera ,&nbsp;Adam McArthur ,&nbsp;Marco Patacci","doi":"10.1016/j.cageo.2025.105917","DOIUrl":"10.1016/j.cageo.2025.105917","url":null,"abstract":"<div><div>Characterisation of deep-water successions is often undertaken at the scale of sedimentary beds. However, different studies often apply alternative bed-type classification schemes, rendering the quantitative comparison of bed properties of different deep-water systems difficult. In this study a quantitative approach to the development of a universal deep-water bed-type classification scheme is proposed based on the synthesis of a large sedimentological dataset, containing &gt;32,000 deep-water facies and &gt;10,000 beds accumulated in 27 turbidite-dominated systems. The classification scheme is applicable to discriminate and categorise lithological (sand, gravel) layers and is based on: (i) the proportion of, gravel, sand, sandy-mud and muddy-sand in the bed, (ii) the presence and nature of vertical sharp grain-size changes, and (iii) the presence and thickness ratio of laminated sedimentary facies. Comparing the bedding properties of channel-fills, terminal deposits (e.g. lobes or sheets) and levees showed that the three architectural-element types are characterised by differences in bed frequency and thickness, overlying mudstone proportions, vertical bed thickness trends, mud thickness and sand-gravel fraction values. Building on these recognised statistical differences an algorithm was developed that is capable of generating, in a stochastic manner, geologically realistic synthetic sedimentary logs depicting deep-water terminal-deposit, channel-fill and levee elements. The one-dimensional facies modelling is governed by a series of input parameters, including total number of beds, sand-gravel thickness, and sand-gravel fraction. The approach can be tailored to produce synthetic logs for specified types of depositional systems (e.g., categorised according to dominant grain size of deposits, age of deposition and global climate (icehouse vs. greenhouse conditions)). A large number of synthetic sedimentary logs can be generated, which can be utilised as training datasets in machine learning algorithms developed to aid subsurface interpretations of clastic sedimentary successions.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"199 ","pages":"Article 105917"},"PeriodicalIF":4.2,"publicationDate":"2025-03-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143619686","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The GEMMA (Geo-EnvironMental multivariate analysis) toolbox: A user-friendly software for multivariate analysis","authors":"Francesco Pilade ,&nbsp;Michele Licata ,&nbsp;Iuliana Vasiliev ,&nbsp;Giandomenico Fubelli ,&nbsp;Rocco Gennari","doi":"10.1016/j.cageo.2025.105914","DOIUrl":"10.1016/j.cageo.2025.105914","url":null,"abstract":"<div><div>Understanding the complex past and present environmental systems requires methodologies capable of analyzing large multi-parameter datasets. The intricate interrelationships within these heterogeneous data would not be effectively detectable using simply bivariate approaches (single parameters versus time/sequence) data series. The mono- or bivariate approaches often fall short, leading to over/underestimation of processes simultaneously affecting the environment. The Geo-EnvironMental Multivariate Analysis (GEMMA) toolbox addresses these challenges, offering a user-friendly software for multiparametric analyses of large and diverse datasets. In the context of increasing dataset complexity, the GEMMA toolbox employs advanced multivariate statistical methods to transcend traditional univariate analyses. It moves beyond compartmentalizing environmental systems, allowing for the analysis of diverse interrelations by setting an efficiency balance between simplicity and comprehensiveness. The GEMMA toolbox uses the programming language R and features a graphical user interface (GUI) to provide a user-friendly tool without requiring advanced programming skills. It allows to perform collinearity analysis (COA), cluster analysis (CLA), principal component analysis (PCA), detrended correspondence analysis (DCA), redundancy analysis (RDA), and integrates statistical analysis with time series (also from geological stratigraphic succession). At every step, the PCA, the DCA, and the RDA analysis are validated through Monte Carlo permutation tests, and results are automatically exported. Source-code is freely available (<span><span>https://github.com/NewGeoProjects/GEMMA_Toolbox</span><svg><path></path></svg></span>) to allow advanced R users to custom and further develop the software according to the open-source principles (<span><span>https://opensource.com/open-source-way</span><svg><path></path></svg></span>). Two case studies illustrate the flexibility and efficiency of this software in investigating datasets that differ greatly in data source and research purpose. The first explores the Miocene-Pliocene transition that occurred ∼5.33 million years ago in the Mediterranean Sea, tracking the environmental change at the end of the Messinian salinity crisis through alkenones molecular fossils record. The second case study investigates landslides induced by intense rainfall in north-western Italy, aiming to explore the impact of morphometric and lithological factors on landslide susceptibility.</div><div>The GEMMA toolbox is software designed to perform advanced multivariate statistical analyses on environmental datasets. It integrates data analysis algorithms, statistical validations methods, and dynamic workflow procedures into an easy-to-use GUI, proving GEMMA toolbox as an effective and flexible software solution for research.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"199 ","pages":"Article 105914"},"PeriodicalIF":4.2,"publicationDate":"2025-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143578467","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fault representation in structural modelling with implicit neural representations","authors":"Kaifeng Gao ,&nbsp;Florian Wellmann","doi":"10.1016/j.cageo.2025.105911","DOIUrl":"10.1016/j.cageo.2025.105911","url":null,"abstract":"<div><div>Implicit neural representations have been demonstrated to provide a flexible and scalable framework for computer graphics and three-dimensional modelling and, consequently, have found their way also into geological modelling. These networks are feature-based and resolution-independent, making them effective for modelling geological structures from scattered interface points, units, and structural orientations. Despite the promising characteristics of existing implicit neural representation approaches, modelling faults within implicit neural representations remains a significant challenge. In this work, we present a fault feature encoding approach to represent faults in implicit neural representations, where the discontinuous information is concatenated as additional features of observation points and query points for network input. We apply this methodology first to a synthetic model to evaluate its efficacy, and subsequently to a real-world dataset from a part of the Gullfaks field in the northern North Sea. The modelling results demonstrate the method’s capacity to generate a well-defined implicit scalar field while preserving sharp transitions at fault locations. Moreover, this work mentions the advantages of the presented approach over using Boolean operations and discontinuous activation functions. Furthermore, we discuss the potential opportunity to integrate prior domain knowledge and geophysics datasets into structural modelling by embedding them as model input features or incorporating them as constraints by loss functions.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"199 ","pages":"Article 105911"},"PeriodicalIF":4.2,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143534664","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Geological-knowledge-guided graph self-supervised pretraining framework for identifying mineralization-related geochemical anomalies","authors":"Zhiyi Chen,&nbsp;Renguang Zuo","doi":"10.1016/j.cageo.2025.105913","DOIUrl":"10.1016/j.cageo.2025.105913","url":null,"abstract":"<div><div>The identification of geochemical anomalies related to mineralization is crucial for predicting the presence of mineral resources. Graph neural networks are essential tools in identifying geochemical anomalies associated with mineralization owing to their ability to process spatially correlated data. However, the performance and generalizability of supervised learning models are often constrained by the limited availability of labeled data. As such, graph self-supervised learning (SSL) has received considerable attention because of its ability to strengthen representation learning by capturing the intrinsic structure and distribution of data, even in scenarios in which labeled data are scarce. The existing SSL methods often fail to effectively incorporate domain-specific knowledge during the learning process, restricting both the generalization and geological interpretability of SSL models. As such, a geological-knowledge-guided graph with self-supervised pretraining (GKGP) framework was proposed in this study. The GKGP framework substantially enhanced the SSL model's ability to capture important spatial relationships and geochemical features by embedding the power-law relationship between the spatial density of mineral deposits and their distance to ore-controlling factors into the pretraining phase of the SSL, which is combined with a transformer-based multi-head attention mechanism. A case study was conducted in the Suizao District, Hubei Province, China, to demonstrate that the proposed GKGP framework can be used to identify mineralization-related geochemical anomalies, even under high-intensity data perturbations, maintaining robust predictive performance. Additionally, the integration of prior geological knowledge notably increases the accuracy and interpretability of the SSL model, ensuring robustness under complex conditions. The proposed pretraining strategy provides reliable geochemical insights to guide future mineral exploration in the study area.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"199 ","pages":"Article 105913"},"PeriodicalIF":4.2,"publicationDate":"2025-02-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143528713","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}

{"title":"ScoreInver: 3D seismic impedance inversion based on scoring mechanism","authors":"Xinyuan Zhu ,&nbsp;Timing Li ,&nbsp;Kewen Li ,&nbsp;Guangyue Zhou ,&nbsp;Ruonan Yin","doi":"10.1016/j.cageo.2025.105896","DOIUrl":"10.1016/j.cageo.2025.105896","url":null,"abstract":"<div><div>In recent years, the introduction of deep learning has significantly advanced the field of seismic impedance inversion (SII). However, existing methods generally rely heavily on large volumes of expensive well logs, limiting their broader applicability, particularly in scenarios beyond mature or synthetic data. To reduce the dependency on well logs in deep learning-based SII research, this paper proposes a 3D data-driven SII approach based on the pseudo-labeling strategy in semi-supervised learning, termed the ScoreInver framework. The core of the ScoreInver framework lies in the design and training of a Scorer, which can precisely select high-quality pseudo-labels from seismic data, thereby enhancing data utilization and extracting geological information while minimizing the need for extensive well logs. This framework is highly versatile, capable of seamless integration into various semi-supervised learning architectures. Experimental results demonstrate that, when using only 9 well logs as training samples on synthetic data, the semi-supervised learning architectures based on the ScoreInver framework significantly outperforms traditional supervised learning methods, with improvements of 3.3% in Structural Similarity Index (SSIM) and a reduction of 29.1% in Mean Squared Error (MSE). Moreover, tests on field data reveal that the application of the ScoreInver framework yields more robust and reliable results, further validating its effectiveness and practicality in real-world exploration environments.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"198 ","pages":"Article 105896"},"PeriodicalIF":4.2,"publicationDate":"2025-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143429942","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}

{"title":"Hybrid Newton method for the acceleration of well event handling in the simulation of CO2 storage using supervised learning","authors":"Antoine Lechevallier ,&nbsp;Sylvain Desroziers ,&nbsp;Thibault Faney ,&nbsp;Eric Flauraud ,&nbsp;Frédéric Nataf","doi":"10.1016/j.cageo.2025.105872","DOIUrl":"10.1016/j.cageo.2025.105872","url":null,"abstract":"<div><div>Geological storage of CO₂ is an essential instrument for efficient Carbon Capture and Storage policies. Numerical simulations provide the solution to the multi-phase flow equations that model the behavior of the CO₂ injection site. However, numerical simulations of fluid flow in porous media are computationally demanding: it can take up to several hours on a HPC cluster in order to simulate one injection scenario for a large CO₂ reservoir if we want to accurately model the complex physical processes involved. This becomes a limiting issue when performing a large number of simulations, e.g. in the process of ‘history matching’. Well events, such as opening and closure, cause important numerical difficulties due to their instant impact on the pressure and saturation unknowns. This often forces a drastic reduction of the time step size to be able to solve the non-linear system of equations resulting from the discretization of the continuous mathematical model. However, these specific well events in a simulation have a relatively similar impact across space and time. We propose a proof of concept methodology to alleviate the impact of well events during the numerical simulation of CO₂ storage in the underground by using a machine-learning based non-linear preconditioning. We complement the standard fully implicit solver by predicting an initialization of Newton’s method directly in the domain of quadratic convergence using supervised learning. More specifically, we replace the initialization in pressure by a linear approximation obtained through an implicit solver and we use a Fourier Neural Operator (FNO) to predict the saturation initialization. Furthermore, we present an open-source Python framework for conducting reservoir simulations and integrating machine-learning models. We apply our methodology to two test cases derived from a realistic CO₂ storage in saline aquifer benchmark. We reduce the required number of Newton iterations to handle a well opening by 53% for the first test case, i.e required number of linear system to solve and by 38% for the second test case.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"197 ","pages":"Article 105872"},"PeriodicalIF":4.2,"publicationDate":"2025-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143387528","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}

{"title":"Linear filter theory for the forward Laplace transform and its use in calculating 1D EM responses","authors":"Niels Bøie Christensen","doi":"10.1016/j.cageo.2025.105865","DOIUrl":"10.1016/j.cageo.2025.105865","url":null,"abstract":"<div><div>The linear filter theory has previously been used for designing digital filters that allow Hankel and Fourier transforms to be calculated as discrete convolutions between sampled values of the kernel function and a set of filter coefficients. In this paper the linear filter theory is used to design filters for the forward Laplace transform that permit rapid and accurate calculations. Furthermore, it is shown that it is possible to estimate the computational errors. It is demonstrated that, in several cases, the Laplace transform developed in this paper can be used in the calculation of electromagnetic responses, traditionally calculated using Fast Hankel Transform filters. It is shown that for many instrument configurations, the Laplace transform approach is faster that the Fast Hankel Transform.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"198 ","pages":"Article 105865"},"PeriodicalIF":4.2,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143419744","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}

{"title":"Efficient computation on large regular grids of higher-order spatial statistics via fast Fourier transform","authors":"Dany Lauzon ,&nbsp;Sebastian Hörning","doi":"10.1016/j.cageo.2025.105878","DOIUrl":"10.1016/j.cageo.2025.105878","url":null,"abstract":"<div><div>The complex spatial structures of natural variables are often caused by geological, physicochemical, meteorological, and biological processes that have shaped the emergence of the fields. The typical prediction of the spatial distributions of these phenomena is based on second-order geostatistical models. However, this approach has limitations, given the high complexity, non-Gaussian distributions, and nonlinear spatial connectivity models inherent in geological systems. Recently, researchers have suggested using higher-order spatial statistics, based on two- and three-point spatial statistics, to better capture spatial phenomena. Nevertheless, applying these methods requires intense numerical calculations, particularly in the case of extensive geostatistical models, and becomes especially intricate when utilized for conditioning realizations, such as in inverse problems. Spatial asymmetries and higher-order spatial cumulants, as well as their generalizations, are important higher-order statistics for characterizing non-Gaussian features. In this study, we focus on third-order statistics derived from two- and three-point spatial statistics. A MATLAB program has been developed to compute efficiently these spatial statistics using the FFT algorithm. The overall approach of these programs draws inspiration from the method successfully used for the fast calculation of variograms and cross-covariances using FFT. We recall the methodology associated with the computation of direct- and cross-variograms using FFT, as well as transiograms for categorical data. Codes are created to process regular grid data, whether it is complete or incomplete. Post-processing tools have been added to help geomodelers visualize the results. Using the FFT method is faster and delivers the same results as conventional spatial methods for this type of data. These programs are particularly valuable tools for geostatistical modeling and estimation when higher-order statistics are present in the spatial structures of natural variables, providing an efficient solution to the computational challenges associated with such applications.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"198 ","pages":"Article 105878"},"PeriodicalIF":4.2,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143419293","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep learning contribution to the automatic picking of surface-wave dispersion for the characterization of railway earthworks","authors":"J. Cárdenas ,&nbsp;A. Burzawa ,&nbsp;N. Radic ,&nbsp;L. Bodet ,&nbsp;R. Vidal ,&nbsp;K. Diop ,&nbsp;M. Dangeard ,&nbsp;A. Dhemaied","doi":"10.1016/j.cageo.2025.105883","DOIUrl":"10.1016/j.cageo.2025.105883","url":null,"abstract":"<div><div>Railway Trackbed (RT), which collectively describes the subgrade structures that support rail tracks, is of great importance to the effective maintenance and rehabilitation of the rail network. Therefore, a comprehensive understanding of the mechanical condition of Railway Earthwork (RE) is necessary. The development of non-destructive and efficient methods for the characterization of REs is a priority. Previous studies have investigated the potential of surface waves for the characterization of RE. Preliminary results indicate that this approach is effective, particularly when using high yield acquisition setup such as landstreamer. However, these instruments generate an amount of data that necessitates optimized and automated processing. The potential of Deep Learning (DL) to automate the processing of surface wave data is being explored. In this study, the primary objective is to identify the energy maxima and propagation modes in dispersion images. A supervised convolutional neural network (CNN), designated as ‘U-Net’, was selected to perform segmentation tasks. This model, called ‘U2-pick’, integrates two U-net architectures: one for maxima identification and another for propagation mode identification. The training dataset was constructed using synthetic data that is representative of a French High-Speed-Line (HSL) RE structure. The preliminary outcomes on the synthetic datasets are encouraging, demonstrating accurate identification of energy maxima and mode classification. However, the predictions made on field datasets revealed that while the energy peaks were identified with a high degree of accuracy, the mode assignment proved to be less satisfactory, especially in the case of higher modes. Finally, a comparison of the accuracy and picking time was performed using more standard tools like maxima search, semi-automatic, and Machine Learning (ML) tools.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"198 ","pages":"Article 105883"},"PeriodicalIF":4.2,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143395140","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}