Computers & Geosciences最新文献

{"title":"Analysis of earthquake detection using deep learning: Evaluating reliability and uncertainty in prediction methods","authors":"Sebastián Gamboa-Chacón ,&nbsp;Esteban Meneses ,&nbsp;Esteban J. Chaves","doi":"10.1016/j.cageo.2025.105877","DOIUrl":"10.1016/j.cageo.2025.105877","url":null,"abstract":"<div><div>This study evaluates the performance and reliability of earthquake detection using the EQTransformer, a novel deep learning program that is widely used in seismological observatories and research for enhancing earthquake catalogs. We test the EQTransformer capabilities and uncertainties using seismic data from the Volcanological and Seismological Observatory of Costa Rica and compare two detection options: the simplified method (MseedPredictor) and the complex method (Predictor), the latter incorporating Monte Carlo Dropout, to assess their reproducibility and uncertainty in identifying seismic events. Our analysis focuses on 24 h-duration data that began on February 18, 2023, following a magnitude 5.5 mainshock. Notably, we observed that sequential experiments with identical data and parametrization yield different detections and a varying number of events as a function of time. The results demonstrate that the complex method, which leverages iterative dropout, consistently yields more reproducible and reliable detections than the simplified method, which shows greater variability and is more prone to false positives. This study highlights the critical importance of method selection in deep learning models for seismic event detection, emphasizing the need for rigorous evaluation of detection algorithms to ensure accurate and consistent earthquake catalogs and interpretations. Our findings provide valuable insights for the application of AI tools in seismology, particularly in enhancing the precision and reliability of seismic monitoring efforts.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"197 ","pages":"Article 105877"},"PeriodicalIF":4.2,"publicationDate":"2025-02-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143347037","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":"Virtual fieldwork in immersive environments using game engines","authors":"Armin Bernstetter ,&nbsp;Tom Kwasnitschka ,&nbsp;Jens Karstens ,&nbsp;Markus Schlüter ,&nbsp;Isabella Peters","doi":"10.1016/j.cageo.2025.105855","DOIUrl":"10.1016/j.cageo.2025.105855","url":null,"abstract":"<div><div>Fieldwork still is the first and foremost source of insight in many disciplines of the geosciences. Virtual fieldwork is an approach meant to enable scientists trained in fieldwork to apply these skills to a virtual representation of outcrops that are inaccessible to humans e.g. due to being located on the seafloor. For this purpose we develop a virtual fieldwork software in the game engine and 3D creation tool Unreal Engine. This software is developed specifically for a large, spatially immersive environment as well as virtual reality using head-mounted displays. It contains multiple options for quantitative measurements of visualized 3D model data. We visualize three distinct real-world datasets gathered by different photogrammetric and bathymetric methods as use cases and gather initial feedback from domain experts.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"196 ","pages":"Article 105855"},"PeriodicalIF":4.2,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143102334","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":"Improved modelling of fluid production and reinjection in geothermal reservoirs","authors":"Adrian Croucher,&nbsp;John O’Sullivan,&nbsp;Michael O’Sullivan","doi":"10.1016/j.cageo.2024.105822","DOIUrl":"10.1016/j.cageo.2024.105822","url":null,"abstract":"<div><div>Models of geothermal reservoirs used for power generation need to simulate the production of multiphase fluid via a complex network of sources, as well as the reinjection of some of this fluid back into the system. The use of standard numerical methods to model such systems with interacting sources typically gives poor non-linear solver convergence and impractical limitations on time-step sizes. However, these issues can be overcome by modifying the Jacobian matrix used in the non-linear equation solution, adding extra terms to represent the interactions between sources. This approach has been incorporated into the Waiwera parallel, open-source geothermal flow simulator. The improved performance offered by this method is demonstrated through test problems and a real geothermal reservoir model.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"196 ","pages":"Article 105822"},"PeriodicalIF":4.2,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143093106","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":"Knowledge-driven stochastic modeling of geological geometry features conditioned on drillholes and outcrop contacts","authors":"Xiaolong Wei ,&nbsp;Zhen Yin ,&nbsp;Wilson Bonner ,&nbsp;Jef Caers","doi":"10.1016/j.cageo.2024.105779","DOIUrl":"10.1016/j.cageo.2024.105779","url":null,"abstract":"<div><div>Geometry features in geosciences are crucial for understanding both near-surface and deeper Earth’s interior structures. Geometry features have traditionally been delineated through diverse data sources such as surface mapping, geophysics, and core samples. However, the quantitative integration of geological knowledge and insights with the geoscientific data remains insufficiently addressed in the model construction. We have formulated a novel framework that employs stochastic level set simulation to model subsurface geometric features. The uniqueness of our framework involves geological knowledge, represented by two-dimensional (2D) geological diagrams, in the numerical modeling using Procrustes analysis. We account for the geological diagrams’ variability and uncertainty through transformations such as rotation, scaling, and translation. By using the designed loss functions, the resulting models align with the established geological knowledge and are also conditioned on the lithology from drillhole and surface observations (i.e., outcrop contacts). We apply the methodology to a field study of a Cu-Ni-PGE (copper-nickel-platinum group element) prospect hosted in a mafic intrusion, the Crystal Lake Gabbro (CLG), in northwest Ontario. Our study focuses on a single segment of the y-shaped CLG. The numerical outcomes of this application demonstrate that the incorporation of expert knowledge, drillholes, and surface data yields models with reliable geological geometry features, particularly the distribution of bottom boundary for intrusion models which is highly associated with economic mineralization.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"196 ","pages":"Article 105779"},"PeriodicalIF":4.2,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143093168","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":"Quantitative lithology prediction from seismic data using deep learning","authors":"Wenliang Nie ,&nbsp;Jiayi Gu ,&nbsp;Bo Li ,&nbsp;Xiaotao Wen ,&nbsp;Xiangfei Nie","doi":"10.1016/j.cageo.2024.105821","DOIUrl":"10.1016/j.cageo.2024.105821","url":null,"abstract":"<div><div>Lithology prediction is essential for understanding subsurface structures and properties. Deep learning (DL) methods, which can capture the nonlinear relationship between lithology and seismic data, have gained significant attention as an effective tool in lithology prediction. However, these methods still face many challenges such as limited well-log data, class imbalances, and the need for robust predictive models. To address these issues, we propose an adaptive boosting-convolutional neural network (AdaBoost-CNN) framework integrated with improved inverse spectral decomposition (ISD) based on non-convex L_1-2 regularization. The ISD method generates high-resolution time-frequency (T-F) spectral maps from seismic data, which serve as inputs for the CNN. Furthermore, we introduce an enhanced sample weight adjustment strategy and a \"CNN transfer\" mechanism within the AdaBoost framework to address class imbalance and enhance training efficiency. The performance of AdaBoost–CNN was validated through field cases, and a comprehensive evaluation of the model parameters was conducted to understand their impact on performance. Field experiments demonstrated that the proposed method enhanced both the training efficiency and generalization ability of the models. Additionally, it effectively predicted lithology from seismic data and quantified lithology probabilities, thereby providing insights into the distribution of subsurface lithology.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"196 ","pages":"Article 105821"},"PeriodicalIF":4.2,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143093171","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":"Improving generalization through self-supervised learning using generative pre-training transformer for natural gas segmentation","authors":"Luiz Fernando Trindade Santos ,&nbsp;Marcelo Gattass ,&nbsp;Carlos Rodriguez ,&nbsp;Jan Hurtado ,&nbsp;Frederico Miranda ,&nbsp;Diogo Michelon ,&nbsp;Roberto Ribeiro","doi":"10.1016/j.cageo.2024.105809","DOIUrl":"10.1016/j.cageo.2024.105809","url":null,"abstract":"<div><div>Seismic reflection is an essential geophysical method for subsurface mapping in the oil and gas industry, used to deduce the position and extension of potential gas accumulations through the processing and interpretation of data. However, interpreting seismic reflection data presents inherent ambiguity, as similar signatures may arise from geological scenarios with distinct physical properties. Furthermore, seismic interpretation is costly and labor-intensive, complicating the generation of extensive labeled datasets useful for conventional AI-assisted solutions. The emergence of self-supervised learning techniques has gained significant attention in the AI community, enabling the pre-training of deep neural networks without annotated data. This paper proposes a two-stage method for segmenting natural gas regions in seismic reflection images by using a generative pre-training transformer model for feature extraction and a recurrent neural network for per-seismic trace analysis. In the first stage, we pre-train the feature extractor in a self-supervised fashion under a seismic image reconstruction task. In the second stage, using the pre-trained feature extractor, we train the recurrent neural network for the segmentation of natural gas accumulations in 1D seismic image traces. We conducted quantitative and qualitative experiments on a private dataset to demonstrate how our self-supervised learning methodology helps achieve better gains in model generalization. The increase, which can reach up to 20%, in the F1-Score metric corroborates the latter. Thus, an improvement in generalization corresponds to an increase in success rates in drilling new wells.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"196 ","pages":"Article 105809"},"PeriodicalIF":4.2,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143093174","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 image inpainting of microresistivity logs: A DDPM-based pseudo-labeling approach with FPEM-GAN","authors":"Zhaoyan Zhong,&nbsp;Liguo Niu,&nbsp;Xintao Mu,&nbsp;Xin Wang","doi":"10.1016/j.cageo.2024.105812","DOIUrl":"10.1016/j.cageo.2024.105812","url":null,"abstract":"<div><div>In geophysical exploration, logging images are frequently incomplete due to the mismatch between the size of the logging instruments and that of the boreholes, which significantly impacts geological analysis. Existing methods, which rely on standard algorithms or unsupervised learning techniques, tend to be computationally intensive and time-consuming. In addition, they are difficult to inpaint regions with high-angle fractures or fine-grained textures. To address these challenges, we propose a deep learning approach for inpainting stratigraphic features. Our method utilizes pseudo-labeled training datasets to alleviate the issue of limited training labels, thereby reducing both computational cost and processing time. We introduce a Fusion-Perspective-Enhancement Module (FPEM) designed to accurately infer missing regions based on contextual guidance, thus enhancing the inpainting process for high-angle fractures. Furthermore, we present a novel discriminator known as SM-Unet, which improves fine-grained textures by adjusting the weight assigned to various regions through soft labeling during training. Our approach achieves a Peak Signal-to-Noise Ratio (PSNR) of 25.35 and a Structural Similarity Index (SSIM) of 0.901 on the logging image dataset. This performance surpasses that of state-of-the-art methods — particularly in managing high-angle fractures and fine-grained textures — while requiring less computational effort.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"196 ","pages":"Article 105812"},"PeriodicalIF":4.2,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143093176","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":"Imputation in well log data: A benchmark for machine learning methods","authors":"Pedro H.T. Gama ,&nbsp;Jackson Faria ,&nbsp;Jessica Sena ,&nbsp;Francisco Neves ,&nbsp;Vinícius R. Riffel ,&nbsp;Lucas Perez ,&nbsp;André Korenchendler ,&nbsp;Matheus C.A. Sobreira ,&nbsp;Alexei M.C. Machado","doi":"10.1016/j.cageo.2024.105789","DOIUrl":"10.1016/j.cageo.2024.105789","url":null,"abstract":"<div><div>Well log data are an important source of information about the geological patterns along the wellbore but may present missing values due to sensor failure, wellbore irregularities or the cost of acquisition. As a consequence, incomplete log sequences may impact the performance of machine learning (ML) models for classification or prediction. Although several approaches for this problem have been proposed in the literature, the lack of consistent evaluation protocols hinders the comparison of different solutions. This paper aims at bridging this gap by proposing a robust benchmark for comparing imputation ML methods. It contributes to establish a standardized experimental protocol that could be used by the petroleum industry in the development of new methodologies for this purpose. It differs from previous works that have been based on different datasets and metrics that prevent an unbiased comparison of results. Eight imputation methods were investigated: Autoencoders (AE), Bidirectional Recurrent Neural Network for Time Series Imputation, Last Observation Carry Forward (LOCF), Random Forests, Self Attention for Imputation of Time Series (SAITS), Transformers, UNet, and XGBoost. The Geolink, Taranaki and Teapot datasets were used to contemplate data from different locations, from which sequences of measurements were deleted and further imputed by the selected ML methods. The Mean Absolute Error, Mean Squared Error, Root Mean Squared Error, Pearson Correlation Coefficient and the Determination Coefficient were used for performance assessment in a set of 480 experiments. The results demonstrated that simple methods as the LOCF and the AE provided competitive imputation results, although the overall best model was SAITS. This reveals that self-attention models are a promising trend for imputation techniques. The choice for the LOCF, AE, SAITS, UNet, and XGBoost to compose the proposed benchmark was corroborated by subsequent statistical analyses, showing that it can be considered a compromise between simplicity, unbiasedness, variety and meaningfulness.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"196 ","pages":"Article 105789"},"PeriodicalIF":4.2,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143093387","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":"A continuous piecewise polynomial fitting algorithm for trend changing points detection of sea level","authors":"Mingyu Xiao,&nbsp;Taoyong Jin,&nbsp;Hao Ding","doi":"10.1016/j.cageo.2025.105876","DOIUrl":"10.1016/j.cageo.2025.105876","url":null,"abstract":"<div><div>Sea level data often contain trend changing points, simply called breakpoints. The commonly used overall linear fitting cannot fit such data well. This paper proposed a continuous piecewise polynomial fitting algorithm for detecting the breakpoints in sea level, considering its highly nonlinear characteristics and periodic signals. Additionally, a Monte Carlo-based confidence interval estimation method is presented. The reliability of the confidence interval estimation method and the stability of the proposed algorithm are validated. Comparing to the continuous piecewise linear fitting and other commonly used methods, the proposed algorithm not only obtains better fitting results and more accurate breakpoints, but also could simultaneously estimate signal periods. The algorithm is applied to several typical sea level datasets, yielding more precise estimates of breakpoints and corresponding piecewise trends. It is found that the global mean sea level caused by glacier mass loss transitioned from a linear rise to an accelerated rise trend around the year 1962 ± 1, with two approximately 52-year and 27-year periodic signals.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"196 ","pages":"Article 105876"},"PeriodicalIF":4.2,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143093409","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":"Marine gravity gradient model calculation based on wavelet numerical integration and CUDA parallel","authors":"Jingyu Bu ,&nbsp;Zhourun Ye ,&nbsp;Xinghui Liang ,&nbsp;Lintao Liu ,&nbsp;Jinzhao Liu ,&nbsp;Tianshuo Fu ,&nbsp;Chunju Zhang ,&nbsp;Yongchao Zhu","doi":"10.1016/j.cageo.2025.105852","DOIUrl":"10.1016/j.cageo.2025.105852","url":null,"abstract":"<div><div>Compared to traditional gravity observations, the disturbing gravity gradient captures more high-frequency information from the Earth's gravity field. Due to the difficulty in directly obtaining gravity gradient measurements, we can calculate the gravity gradient by the Stokes integral methodology. However, this computational process faces two main issues: 1) Obtaining an analytical expression from the integral of the original function is difficult, necessitating the use of numerical integration methods; and 2) Large volumes of data can lead to reduced computational speed. In our study, Chebyshev wavelet numerical integration is employed to improve the integration accuracy in calculating the disturbing gravity gradient. We also provide the derivation process for nodes and weights based on the Chebyshev wavelet integral. Concurrently, the Compute Unified Device Architecture (CUDA) enables parallel computing on the Graphics Processing Unit (GPU) to improve computational speed. We detail the application of these techniques and translate theoretical concepts into a practical computational program. Experiments conducted on marine area data illustrate that integrating Chebyshev wavelet numerical integration with CUDA parallel computing not only ensures precise calculations but also significantly boosts computational efficiency.</div></div>","PeriodicalId":55221,"journal":{"name":"Computers & Geosciences","volume":"196 ","pages":"Article 105852"},"PeriodicalIF":4.2,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143093098","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}