Abstract Interpretation of seismic reflectors is fundamental to seismic stratigraphy, yet manually delineating high-resolution stratigraphic horizons remains subjective and labor-intensive. We present Lateral Interpretation (LI), a new method for constructing high-resolution horizon geometries from 2D seismic profiles. LI consists of four sequential steps: (1) contour extraction, (2) reflector connection using a Traveling Salesman Problem (TSP) optimization framework, (3) automated onlap/downlap connection, and (4) lateral extension with vertical smoothing. We implemented the full workflow in an open-source software package equipped with a graphical user interface, enabling both fully automated processing and seamless integration with manual interpretation. To evaluate the performance of LI, we utilized the Dutch F3 block dataset, which contains the North Sea Groups. We assessed the lateral consistency of three representative horizons in a benchmark seismic dataset using cross-correlation across multiple adjacent inlines, obtaining an average R2 of 0.84. We further applied LI to distinguish three stratigraphic units within the upper Eridanos delta of the Dutch F3 block. The automatically derived horizons showed excellent agreement with manual interpretations, with correlation values ranging from 0.70 (Unit 1) to 0.99 (top of Unit 3). These results demonstrate that LI reliably reproduces manual-quality stratigraphic interpretations while offering a fast, objective, and reproducible alternative. The method holds strong potential for broader applications, including high-resolution seismic stratigraphy and inverse analyses of sea-level fluctuations.
Abstract Two major challenges we face with 4D seismic projects are: lengthy processing time, ranging from several weeks to several months for fast-track products and up to a year for full-track products to produce interpretable 4D seismic images, and the iterative time-consuming process of 4D interpretation and integration. In this paper, we address these challenges using 4D raw-data elastic full wavefield inversion (eFWI) for elastic property changes and 4D rock physics inversion for petrophysical property changes. Our approach, utilizing raw seismic data and High Performance Computing (HPC) with 4D eFWI, can produce high-resolution acoustic impedance and shear property changes within weeks of field data delivery. 4D rock physics inversion is then applied to estimate saturation and pressure changes in the reservoir, which are validated with well surveillance data and the estimated saturation/pressure changes from reservoir simulation models. We demonstrate these rapid turnaround 4D seismic workflows using a 4D OBN dataset from the Liza field in offshore Guyana, showcasing accelerated subsurface insights for improved reservoir management and opportunity generation.
Abstract Due to the considerable expense associated with large-scale 3D seismic acquisitions, operators often rely on both legacy and new 2D seismic datasets for regional analysis and preliminary assessments during emerging frontier explorations. In this paper, we introduce an innovative deep learning-assisted workflow designed to transform 2D seismic data into pseudo-3D seismic volumes. This approach offers a cost-effective solution for risk mitigation prior to investing in full 3D surveys and rejuvenating the value of existing legacy and regionally acquired 2D seismic data. By integrating deep learning techniques, deterministic algorithms, and domain-specific geological understanding, we achieve high-quality 2D-to-3D seismic image reconstruction via the proposed workflow even with ultra-sparse 2D lines (line spacing > 5 km), without assuming any specific 2D line organization geometries. It consistently outperforms traditional methods based on geophysical signal processing by better preserving the original 2D line characteristics, maintaining realistic seismic image styles, as well as reducing directional interpolation artifacts. Furthermore, the proposed technique is substantially more economical than conventional geophysical methods. We demonstrate the effectiveness and versatility of this workflow are through applications to legacy 2D seismic datasets from diverse regions worldwide.
Abstract Carbon capture and storage (CCS) projects face significant subsurface uncertainty that affects regulatory compliance and economic viability. Under frameworks such as the U.S. class VI permitting process, operators must periodically reassess the area of review (AoR), i.e., the predicted extent of pressure and CO2 migration. Conventional monitoring methods, such as 4D seismic, help track plume migration but remain too costly and impractical for frequent surveillance. In this study, a predictive monitoring workflow combining agile, focused seismic techniques was introduced. This workflow was recently demonstrated as a fast, targeted CCS monitoring method at the 2024 CCUS Conference (Al Khatib, H., T. Roth, J. Grobys, and A. Szabados, Paper 3999159, Houston, TX, 11–13 March), featuring a novel ensemble-based clustering approach that groups reservoir models by their expected seismic response. The approach accelerates convergence toward the reservoir model whose pressure, CO2 saturation, and structural characteristics best explain the observed seismic signal early in the injection lifecycle. Applied to a CCS case with 91 stochastic realizations, the workflow integrates focused seismic measurements at critical divergence zones with iterative model discrimination. Convergence toward the model whose saturation best matched the observed seismic response reduced plume-extent uncertainty by up to 99% within 3 years of CO2 injection, shrinking the AoR envelope by 18% relative to the most expansive initial scenario. Injected-volume uncertainty similarly collapsed toward 51.93 Mt. These improvements deliver substantial economic benefits: securing USD 4.41 billion in 45Q tax credits, reducing monitoring, measurement, and verification costs by USD 36 million, and deferring well remediation investment with a net present value of USD 6.8 million. The findings confirm that focused seismic, combined with ensemble-based clustering, enables tightening of the predictive envelope early in the injection lifecycle. This approach is applicable to saline aquifers and depleted reservoirs, supporting adaptive field management and accelerating progress toward secure and economically viable carbon storage.
Abstract Seismic full-waveform inversion (FWI) is a powerful tool for monitoring subsurface changes during carbon capture and storage operations, but its ill-posed nature makes uncertainty quantification (UQ) essential for reliable interpretation. Sampling-based Bayesian methods, such as Markov chain Monte Carlo, provide rigorous UQ but are computationally demanding. In conventional FWI, elastic properties are assigned to densely discretized space-filling cells, resulting in a high-dimensional parameterization that makes large-scale elastic FWI computationally infeasible. To address this challenge, a rock-physics-guided parameter-reduction strategy that compactly represents the CO2 plume geometry using cubic splines controlled by a limited number of nodes was proposed. This parsimonious parameterization not only significantly reduced the number of model parameters and the forward simulations required for effective UQ using sampling methods but also showed potential to improve the practicality and efficiency of other types of UQ methods. Numerical experiments on a cross-well synthetic scenario and a field-scale synthetic case based on the Aquistore storage site in Saskatchewan, Canada, demonstrated that the method efficiently reconstructed the plume shape and its extent and that it converged to consistent posterior distributions across multiple Markov chains.
Abstract Managed aquifer recharge in urban basins increasingly relies on small-footprint stormwater control measures (SCMs), yet their actual infiltration performance and contribution to groundwater recharge remain poorly constrained after installation. This study evaluated two contrasting SCMs in the Los Angeles Basin — a deep drywell beneath a paved parking lot and a shallow bioswale within a narrow right-of-way — using controlled recharge experiments monitored with borehole-based, time-lapse electrical resistivity tomography (ERT). Instrumented boreholes were installed around each SCM to enable 3D (drywell) and 2D (bioswale) imaging, and ERT data were collected autonomously over several days, during which known volumes of water were applied to each system for calibration. Soil moisture was independently measured with neutron probes and in situ sensors to validate resistivity-derived wetting patterns. Results showed that the drywell rapidly conveyed water from shallow pretreatment features and deeper screened intervals into laterally extensive and stratigraphically controlled pathways, with preferential flow observed along higher-permeability layers. By contrast, infiltration in the bioswale was limited to the upper 2 m, even under elevated water levels exceeding typical storm conditions, indicating shallow storage and limited deep percolation. In both settings, changes in resistivity closely matched changes in measured soil moisture, confirming the ability of borehole ERT to resolve hydrologic processes in constrained built environments. The findings demonstrated that SCM infiltration performance is highly site specific, influenced by engineering design and subsurface heterogeneity, and that geophysical monitoring provides critical information for accurate recharge estimation, SCM efficacy and siting, maintenance planning, and long-term water-resource management in urban basins.
Abstract To investigate gas migration along a near-surface fault and assess the performance of various seismic monitoring techniques, a controlled shallow CO2 release experiment was carried out at the CO2CRC Otway International Test Centre (Victoria, Australia). Sixteen tons of CO2 were injected at approximately 70 m depth into the Brumbys Fault. The monitoring program combined reverse 4D vertical seismic profiling using a downhole sparker source with a dense surface geophone array, cross-well seismic tomography with distributed acoustic sensors (DAS), passive borehole DAS data analysis of Rayleigh-wave amplitudes, and data from permanently deployed surface orbital vibrators. All methods detected the evolving plume consistently and provided complementary information on its geometry and migration. The results demonstrated that the integrated monitoring framework is capable of resolving small-scale CO2 movement and can be scaled up for continuous, cost-effective surveillance in larger storage projects.
Interpreting 4D seismic signals involves significant uncertainty, particularly in stiff and heterogeneous carbonate reservoirs, such as the Brazilian presalt. This study aimed to leverage seismic inversion to estimate changes in the elastic properties of rocks due to production activities and to evaluate the benefits of inversion techniques for 4D seismic interpreta- tion. The study focused on the productive Tupi Field in the Santos Basin, where high repeatability 4D seismic data were acquired in 2015 and 2017/2018 using ocean-bottom node technology. The constrained sparse-spike inversion method was used to generate 3D poststack inversions for both surveys, and 4D information was obtained through a straightforward subtraction from the impedance for both 3D results. Conse- quently, the inverted volumes produced delta impedance maps and sections, which were compared with amplitude difference and integrated with well-history data. The results demonstra- ted that the delta impedance maps effectively outlined the primary 4D anomalies associated with water-alternating-gas and injection wells. Compared to the amplitude map, the delta impedance provided a more precise delineation of the key anomalies, which were also in greater agreement with the well-history data and literature findings. These results suggest that integrating seismic inversion with well-history data significantly improved 4D seismic interpretability, offering valuable insights into reservoir dynamics. Additionally, the integration between the inversion outcomes and well-history data with geologic data helped identify potential barriers for fluid flow related to microporous zones within the upper section of the Barra Velha Formation. In conclusion, integrating seismic inversion with well history and other data sources enhances 4D seismic interpretation, providing valuable insights into reservoir dynamics and supporting more informed reservoir management decisions. This approach effectively addresses interpretational uncertainties, particularly in complex fields.
Abstract Conventional imaging methods like reverse-time migration impose assumptions, such as the Born approximation, that introduce strict data preprocessing requirements, which reduce their effectiveness in regions of strong impedance contrasts, complex structural geometries, and poor illumination. Multiparameter full-waveform inversion (MP-FWI) offers an alternative approach that enables the simultaneous estimation of many subsurface properties (e.g., VP and reflectivity) directly from raw seismic data. MP-FWI is a least-squares solution that uses the full wavefield, treating multiples as valuable signals to improve resolution and illumination. In recent years, MP-FWI approaches have assumed acoustic wave propagation to generate angle-dependent reflectivity for elastic amplitude-versus-angle (AVA) analysis, which enables P-impedance and VS/VP ratio estimation via an additional inversion step. Elastic MP-FWI offers the potential to skip this additional inversion step and determine these AVA attributes directly from the acquired data. In this article, two case studies from the Australian North West Shelf and the Gulf of Mexico are presented, demonstrating the fidelity of viscoelastic MP-FWI in deriving AVA attributes. The accuracy of these inverted models is evaluated against the conventional workflow and well data, and the role of quantitative interpretation expertise is discussed in the context of this novel paradigm.
Abstract We develop a low rank Fourier Neural Operator (LR-FNO) surrogate to address the dual challenge of accelerating large-scale 3D reservoir simulation while enabling efficient multi well placement optimization. By applying CP-decomposed spectral convolution kernels, LR-FNO reduces parameter count by several orders of magnitude and alleviates overfitting, while maintaining high predictive fidelity. Compared with the original FNO, LR-FNO lowers GPU memory usage by roughly two-thirds, allowing the use of larger spectral kernels and improving model expressiveness. Trained on thousands of OPM simulations with randomly sampled well locations, the surrogate predicts full 3D pressure and saturation fields across all timesteps in under one second. LR-FNO also exhibits improved generalization in limited-data regimes and accurately reproduces cumulative oil production. When coupled with a Bayesian optimizer, the surrogate completes 1,500 iterations of the well placement optimization aimed at maximizing oil production within a few hours, compared to approximately ten days using OPM, providing a practical, efficient, and reliable decision support workflow for complex reservoir systems.
The correlation coefficient has inadequacies as a metric for comparison of seismic data to presumed truth in the form of synthetic seismograms or well logs. It is not a direct measure of the information content of a prediction, nor the shared information between the prediction and perfect data. All else being equal, the information content (useful or not) of seismic data decreases monotonically with decreasing bandwidth, but the correlation coefficient may vary differently. Quantities from information theory, such as cross entropy, variation of information, and channel capacity, are measures that can aid in the assessment of the information content of a prediction and are amenable for use as loss functions in machine learning and other applications. These ideas were demonstrated on real and synthetic data.
Abstract Monitoring CO₂ storage sites over decades requires cost-effective, reliable strategies to ensure containment and conformance. Conventional approaches such as repeated 3D/4D seismic surveys are expensive and logistically challenging, particularly offshore. The SPARSE project, developed under ACT4, introduces a sparse node-based multiphysics monitoring concept that integrates seismic, electromagnetics, gravity, and deformation measurements. This approach was designed to reduce survey frequency while maintaining sensitivity to pressure and saturation changes. Results from onshore field trials at Carbon Management Canada’s Newell County site and offshore feasibility studies based on the Smeaheia CO₂ storage model are presented. Onshore tests demonstrated the feasibility of using permanent seismic sources, distributed acoustic sensing, and electromagnetic configurations within a sparse monitoring framework. Offshore modeling indicated that seismic, electromagnetic, and gravity signals could detect plume arrival, monitor its development over time, and distinguish between different migration scenarios under realistic noise conditions. These findings supported the feasibility of sparse node multiphysics monitoring as a scalable and cost-efficient approach for long-term CO2 storage assurance, complementing conventional methods and informing future deployment.
Abstract Seismic-to-well tying remains a critical yet time-intensive component of seismic interpretation workflows. While automated approaches have reduced manual effort, many rely on dynamic stretch-and-squeeze adjustments that maximize trace similarity at the expense of distorting the time–depth relationship. In large multi-well projects, maintaining both efficiency and geophysical consistency remains challenging. This paper presents a physics-informed machine learning workflow that jointly estimates the seismic wavelet and an optimal bulk time shift while explicitly avoiding dynamic warping of the time–depth relationship. The model is trained on a large synthetic dataset constructed from randomly generated reflectivity series convolved with analytical wavelets of varying phase and bandwidth, with known bulk shifts applied. Realistic seismic variability is introduced through controlled augmentation, including additive noise, smooth low-frequency distortion, multiples, phase perturbations, and nonlinear time warping. The workflow is demonstrated on multiple wells from the publicly available F3 Netherlands offshore field dataset. Results show improved temporal alignment at zero lag, preservation of velocity plausibility, and rapid scalability. The method illustrates how machine learning can be integrated into interpretation workflows while maintaining physical realism.
Abstract Distance and Quadrant (DQ) trace attributes provide an integrated AVO-based framework for reservoir characterization, structural interpretation, and stratigraphic analysis. Although the DQ and ThetaPX attributes have been introduced in earlier work using synthetic traces, this paper presents the first comprehensive demonstration of the full ten-attribute DQ suite across four contrasting geologic and AVO settings: an onshore thin gas sand (Colony Formation, Alberta, Canada), an offshore CO2 storage reservoir (Sleipner field, Norway), a Class 4 AVO gas field in a faulted fluvial deltaic system (Kupe field, New Zealand), and a Class 1 mixed carbonate clastic multilithology play (Grayback SE field, Texas). A single DQ processing workflow generates all ten attribute volumes including DQ, ThetaPX, StickOgram, signed isochron, signed half isochron, Average DQ, Sum DQ, Median ThetaPX, and optimized near and far-minus-near stacks generated directly from migrated near- and far-angle stack volumes, without prior well calibration or wavelet estimation. The workflow embeds automated interpretation elements that organize amplitude and offset information into attribute volumes suitable for advanced seismic analysis and machine learning applications. These attributes support three-dimensional visualization, horizon tracking, fault interpretation, stratigraphic facies classification, well ties, layer thickness estimation, and AVO crossplot analysis. By demonstrating consistent attribute behavior across bright-spot, dim-spot, Class 3, Class 4, and CO2 saturation settings, this study establishes the DQ workflow as a broadly applicable preliminary screening tool prior to quantitative inversion or machine learning analysis. Results from this screening process can help focus and guide subsequent reservoir characterization efforts.
Abstract Elastic parameters, their combinations, and crossplots are frequently used for mapping subsurface fluids in exploration, appraisal, and development of hydrocarbons, and in other areas where monitoring produced and injected fluids is essential, such as time-lapse seismic or carbon capture storage. Seismic interpreters work with different seismic data sets, i.e., inversion products, amplitude variation with offset (AVO) volumes, or angle stacks. This study discusses the estimation of relative values (reflectivities) of common elastic properties from seismic data by weighting angle stacks or as chi-angle projections from AVO parameters. Furthermore, a review of lambda-rho and Poisson’s ratio multiplication as a fluid indicator is presented. The product approximates two other important fluid indicators: the moduli difference and the fluid-rho method. The study demonstrates that the fluid indicator proposed generates a nonlinear gain to boost fluid signal, which significantly enhances sensitivity for discrimination, and that it acts as a double-gate filter for fluid detection because both parameters must decrease simultaneously in the presence of hydrocarbons. A reparameterization of the AVO equation is proposed to extract reflectivities of lambda-rho, Poisson’s ratio, and density simultaneously from prestack data. This review and case study will help increase awareness of the importance of using elastic properties in seismic mapping of subsurface lithologies and fluids.
Abstract Reliable interpretation of time-lapse (4D) seismic data is essential for monitoring injected CO2 and ensuring safe long-term storage, yet observed differences between vintages are often dominated by acquisition nonrepeatability, near-surface variability, and imaging artifacts. An automated workflow that combines local orthogonalization weighting (LOW), a physics-guided method that isolates dynamic fluid-induced changes, with a compact 3D deep learning model (TLNet) that predicts CO2 plume probability from paired baseline–monitor volumes was developed. LOW suppresses repeatable geology and stabilizes the dynamic anomaly, enabling simple value and texture gates to produce volumetric plume labels without manual interpretation. TLNet learns from these seismic-driven labels, generating high-resolution plume probability volumes that highlight confinement-related seismic responses associated with thin layers, lateral variability, and intratier fingering. The workflow was applied to the Sleipner CO2 storage project, and the LOW component was further evaluated on the Cranfield and Duri fields to assess robustness across marine and land surveys, shallow and deep targets, and CO2 and thermal-enhanced oil recovery (EOR) settings.
Abstract Seismic data and elastic log conditioning are regularly carried out in preparation for seismic inversion. When the seismic quality is uncertain and the well data are sparse, poor or missing, an integrated and iterative workflow must be applied to provide a satisfactory level of confidence prior to inversion. A dataset from the Gulf of Thailand provided a challenging testing ground. The seismic data were old, noisy, and had limited offset. Only two wells lay within the seismic volume, one vertical with significant washouts and one highly deviated with limited log coverage. No shear velocity measurements were available in either well. Using inversion at the wells for guidance and offset wells for calibration, seismic data conditioning and well log synthesis and editing were iteratively adjusted to optimize the well tie. Significant increases in correlations between the processed well logs and inverted data were achieved whilst honoring elastic log trends and relationships observed in the offset wells. Based on these improvements combined with observations of the seismic data, inversion products and quick-look interpretation results in section and map view, confidence in the seismic data as fit for inversion and reservoir characterization was achieved.
Brazilian presalt carbonates are notoriously heterogeneous and heavily overprinted by multiphase diagenesis, making reservoir-quality prediction away from well control particularly challenging. The spatially continuous constraints available for subsurface characterization are typically limited to inver- sion-derived geophysical attributes, most commonly acoustic impedance (IP) and, where prestack data quality permits, VP/VS. Using two wells from the Atapu field (Santos Basin) as a well-log-scale proxy for geophysical attributes available away from wells, this study evaluated the incremental informa- tion content of different attribute combinations for super- vised facies classification. Core, thin-section, and well-log data were integrated into three reservoir-quality-driven classes, and Random Forest models were trained using four feature scenarios: IP-only, IP + VP/VS, IP + deep resistivity (RESD), and full-log suite. Model performance was assessed through blind-well testing, and the attribute value was quantified using an information-theoretic framework based on entropy reduction. Adding VP/VS to IP yielded only modest infor- mation gain: relative to the IP-only baseline, facies entropy reduced moderately, whereas combining IP with RESD produced substantially larger gain, approaching the full-log-suite reference. The incremental value of VP/VS over IP was measurable but limited, less than that provided by attributes directly linked to pore connectivity. Quantifying information gain, rather than classification accuracy alone, provided a robust framework for evaluating geophysical attribute combina- tions away from well control. The workflow offers a practical template to assess the incremental value of elastic and nonelastic attributes for seismic-driven reservoir-characterization strategies in heterogeneous carbonate settings.
Selecting appropriate rock physics models in geologically complex settings with limited core or lab measurements poses a significant challenge for reliable calibration and property prediction. This is further hindered in reservoirs with multiple stacked pay zones and varying depositional environments, where separate rock physics models (RPMs) are typically required for different rock types to ensure accurate log prediction. This study presentend a rock physics-guided machine learn- ing workflow that integrated statistical learning with physi- cally constrained rock physics models to predict facies, elastic, and petrophysical properties across multiple wells simultane- ously. The method applied an expectation-maximization (EM) algorithm within a maximum-likelihood framework in which facies probabilities and RPM fitting parameters were jointly updated while remaining constrained within physically realistic bounds. This enabled automatic calibration of rock physics templates while simultaneously generating facies classifications and log predictions. The workflow was tested in two geologi- cally distinct settings: the Athabasca Oil Sands in northeastern Alberta and the Barnett interval of the Midland Basin. Using common well-log inputs (gamma ray, density, neutron porosity, resistivity, and elastic logs where available), the workflow calibrated rock physics models for multiple facies and gener- ated consistent elastic and petrophysical predictions across wells. Results from training and blind wells demonstrated that the method could reconstruct missing log suites, estimate elastic and petrophysical properties from limited inputs, and gener- ate per-facies depth trends consistent with regional geology. The workflow not only streamlined RPM calibration but also supported reservoir characterization workflows, such as log repair, well-to-seismic ties, and construction of low-frequency models for seismic inversion, thereby enabling faster and more reliable interpretation in data-scarce environments.
Abstract Geophysics technologies are critical tools for meeting regulatory requirements and operational demands of CO2 storage projects, specifically the indirect tracking of plumes for confidence in conformance and containment. The CCUS (carbon capture, utilization, and storage) industry is adapting existing technologies from the oil and gas industry and developing new technologies specifically for CO2 monitoring applications. Given the relatively low number of active, commercial CO2 storage projects globally, paired with the forecasted near-term increase in projects, funding for technology development may be challenging, albeit necessary, to procure. All sectors of industry (operators, vendors, and academia) must work toward the same goal of maturing technologies that will enable short-cycle-time, cost-effective, long-term CO2 plume monitoring sufficient to support reservoir management. An assessment of available indirect geophysics technologies and their technical maturity for CO2 plume monitoring in saline aquifers was performed to facilitate identification of less mature technologies for potential development and validation. In scope were remote sensing technologies that measure subsurface properties using the gravitational field, propagation of elastic waves (seismic), or propagation of electromagnetic energy. Technologies were identified and assessed based on internal knowledge, external literature, and interviews with internal subject matter experts. A technical readiness level (TRL) was assigned to each technology, then the technologies were high-graded based on relative cycle time, cost, coverage, and resolution. It was concluded that reducing costs and cycle time of active source seismic is a high priority, that multiphysics approaches that integrate multiple properties and geophysics technologies to characterize the subsurface should be matured, and that borehole potential field technologies need further demonstration to delineate possible detection limits.
