Seismic Volume Research Articles

DELINEATION OF HYDROCARBON SATURATED RESERVOIR SAND USING INTEGRATED 3D PRE-STACK SEISMIC AND WELL LOG DATA IN BONGA – FIELD, CENTRAL SWAMP DEPOBELT, ONSHORE NIGER DELTA, NIGERIA

This study has successfully delineated the lateral continuity of hydrocarbon saturated sand reservoir in Bonga field, Niger Delta. 3D pre-stack seismic volume and well logs from two (2) exploratory wells were employed in the pre-stack seismic inversion analysis. The delineated BGA reservoir sand spans across the two (2) wells labelled Bonga-26 and Bonga-30. The reservoir depth ranges from 10490 ft to 10620 ft in Bonga-26 while the reservoir depth ranges from 10390 ft to 10490 ft in Bonga-30. The delineated reservoir is characterized by low gamma ray (< 75 API), water saturation, shale volume and high resistivity as deciphered in their respective well log curves signature. Rock attribute crossplot was carried out to discriminate between the formation fluid and lithology. The crossplot space of VP-VS ratio versus acoustic impedance (AI), discriminates the formation properties into lithology and fluid (gas and brine sand) based on clusters inferring the presence of each formation fluid properties. The inversion cross sections of P-impedance, S-impedance, density (ρ) and VP-VS ratio depicts the spread and lateral continuity of the reservoir sand across the well locations. The delineated zones reveal low P-impedance, density, VP-VS ratio and slight increase in S-impedance which further validate the presence of hydrocarbon in the field.

Quantification of modelled 4D response and viability of repeated seismic reservoir monitoring in J-Area Field, Central North Sea

4D reservoir monitoring has recently become a major tool used to manage the hydrocarbon production of reservoirs. When combined to production well data, high quality 4D seismic is very useful to address production changes in a reservoir over time. This becomes very challenging though, for most of the clastic reservoirs from the J-Area field, in Central North Sea. These reservoirs are frequently compartmentalized with complex faulting which can result in different initial fluid contacts and pressures across the same field. Full understanding of which faults are acting as hydrocarbon baffles or flow barriers would be very useful in optimizing drilling. This work aimed to determine whether 4D seismic techniques could realistically aid this understanding by quantifying the reservoir production effects relating to pore pressure and water saturation changes in J-Area Triassic using real and predicted well data. It further aimed at testing the viability of repeated seismic reservoir monitoring in this field using the normalized root mean square (NRMS) technique. The modelled 4D seismic response derived from synthetic seismic traces based on seismic volumes generated from the well data was used to design a new dedicated 4D survey. The results show that both pore pressure depletion and water saturation changes produce a significant 4D effects in the reservoir. Yet, pore pressure depletion become the major production effect in this field as the majority of 4D effects are due to pore pressure reduction through depletion of the reservoir. The study suggest that, for optimum reservoir monitoring, a baseline survey must be reprocessed in parallel with the monitor survey to reduce the NRMS noise or alternatively a dedicated repeat survey is acquired matching the design of the latest vintage of seismic.

The sedimentary architecture of hyperpycnites produced by transient turbulent flows in a shallow lacustrine environment

Hyperpycnal flows are river-derived extrabasinal turbidity currents transporting both sand and clay to lacustrine, coastal, shelf and deepwater sedimentary environments. Experimental research in the past twenty years has shown that the presence of clay in sediment-laden flows promotes a transitional behavior between fully turbulent flow and a quasi-laminar plug flow regime. However, to date, most work concerning the fluid dynamic interpretation of lacustrine hyperpycnal flows has been founded on concepts based on fully turbulent flows, and relatively little is known about the influence of clay on the rheological and turbulence characteristics of these flows. With the help of 3D seismic volume and wire-line logging data, the present study undertakes a description and interpretation of subsurface mixed sandstone–mudstone bedforms observed in cores from the Upper Cretaceous Heidimiao Sandstone, Nenjiang Formation, Songliao Basin, NE China. Six lithofacies and five lithofacies associations are recognized from the cores and interpreted as formed under both turbulent and transitional flows. Typical bedforms of transitional flows include large current ripples and low-amplitude bedwaves. The delta-fed shallow lacustrine hyperpycnal flow deposits are characterized by proximal sinuous channels (extending for c. 20 km) and lobe deposits (extending for >20 km), showing a fan-like geometry associated with distributary channel extension in the down-dip direction. Basal erosion in the proximal channels is produced by the initial turbulent and turbulence-enhanced transitional flows, with channel infill dominated by upper transitional plug flows or quasi-laminar plug flows. The lobe deposits possess a fan-like geometry with large aspect ratio, and were formed from turbulence-enhanced and lower transitional plug flows. The downdip transformation of sediment-laden flows in the lacustrine basin varies from an initial turbulent flow, via quasi laminar plug flows to upper and lower transitional plug flows, to turbulence enhanced transitional flows. These transitional flows are interpreted to have experienced gradual dilution and deceleration and eventually transformed to turbulent flows in their distal regions. A new model for delta-fed shallow lacustrine hyperpycnal flow deposits is presented incorporating decelerated transitional flows and flow transformations. This new model can aid understanding of the depositional processes of lacustrine transitional flows and the facies distribution of hyperpynal flows in both modern and ancient sediments.

3-D and 5-D reconstruction ofPreceiver functions via multichannel singular spectrum analysis

SUMMARYThe receiver function (RF) method is fundamental in assessing mantle seismic discontinuity depths and reflectivities. Most of the current approaches rely on phase equalization, though in many applications, high levels of incoherent noise, incomplete and irregular sampling customarily interfere with the analysis of weak secondary phases. In recent years, advancements in the field of multidimensional seismic data processing have triggered a shift in interest towards its application to RFs, specifically to single station gathers that depend on a single spatial dimension. Our work generalizes the application of singular spectrum analysis to RFs that rely on two and four spatial dimensions recorded by dense seismic arrays. We adopt a multidimensional signal processing approach known as multichannel singular spectrum analysis. We develop a strategy to assemble and enhance 3-D and 5-D seismic volumes via matrix rank reduction and a reinsertion algorithm to simultaneously suppress random noise, retrieve absent observations and boost identifiability of secondary conversions. We provide informative synthetic examples to gain insight into the effectiveness and limitations of our approach. In the real data example, we improve weak conversions from the mantle transition zone (MTZ) recorded by the USArray in the Yellowstone area. The reconstruction algorithm accurately recovers the timing and polarity of conversions associated with the 410-, 520- and 660-km seismic discontinuities. Our investigation shows that the simultaneous processing of several spatial variables expedites signal restoration, particularly in directions where large recording gaps exist due to a lack of earthquakes, which aids the mapping and identification of the MTZ interfaces. This study presents a theoretical/practical framework for the reconstruction of multidimensional RF data, and its full potential can be exploited with dense acquisition available from the emerging seismic nodal arrays to improve passive seismic imaging.

An integrated approach to distribute carbonate reservoir properties, using lithofacies and seismic attributes: a case study from SW of Iran

Complexity and heterogeneity of carbonate deposits make the understanding of the reservoir quality challenging. This study is a contribution to the understanding of the distribution of carbonate–evaporite reservoir property of the Permo–Triassic Kangan–Dalan succession in one of the giant gas fields in the southwest of Iran. To capture and describe the complexity and heterogeneity of the carbonate reservoir, we defined different rock properties approach to the management of quantitative data on carbonate reservoir which is based on well and seismic data with an integrated sedimentological investigation, petrophysical evaluation, and seismic analysis to determine the new reservoir lithofacies classification. A lithofacies-based reservoir characterization study was conducted carbonate reservoir to build a numerical 3-dimensional (3D) geological model including reservoir property prediction for reservoir flow simulation. To do this, full suite of logs from six wells, core from four wells in the studied interval and 3D seismic volume across the studied interval have been involved and investigated. Five lithofacies were identified from detailed analysis, and the relationship between lithofacies and reservoir property was considered and geostatistically propagated in the studied area. This study shows that an integrated study using seismic data and well logs can be used to basically divide the reservoir into relatively homogeneous zones based on core and petrophysical analysis, and then evaluate each zone to distinguish high-quality zones. This approach provides useful insight to Permo–Triassic carbonate reservoir and other similar reservoirs and can also be effectively used in areas which lack well or outcrop data.

6D phase-difference attributes for wide-azimuth seismic data interpretation

Phase information of seismic signals is sensitive to subsurface discontinuities. However, 1D phase attributes are not robust when dealing with noisy data. In addition, variations of seismic phase attributes with azimuth are seldom explored. To address these issues, we have developed 6D phase-difference attributes (PDAs) derived from azimuthal phase-frequency spectra. For the seismic volume of a certain azimuth and frequency, we first construct stacked phase traces at each common-depth point along a certain decomposed trending direction. Then, the 6D PDA is extracted by calculating the complex-valued covariance at a 6D phase space. The proposed method enables characterization of the subsurface discontinuities and indicates seismic anisotropy. Moreover, we provide one q-value attribute obtained by singular value decomposition to describe the variation intensity of PDA along different azimuths. Simulated and field wide-azimuth seismic data sets are used to illustrate the performance of the proposed 6D PDA and the derived q-value attribute. The results show that PDA at different frequencies can image various geologic features, including faults, fracture groups, and karst caves. Our field data example shows that PDA is able to discern the connectivity of karst caves using large-azimuth data. In addition, PDA along different azimuths and the q-value attribute provide a measurement of azimuthal variations, as well as the anisotropy. Our 6D PDA method can be used as a potential tool for wide-azimuth seismic data interpretation.

Modelling and mapping hydrocarbon saturated sand reservoir using Poisson’s impedance (PI) inversion: a case study of Bonna field, Niger Delta swamp depobelt, Nigeria

3D seismic volume and two well logs data labelled Bonna-6 and Bonna-8 were employed in the inversion process. The data set was simultaneously inverted to produce P- and S-impedances, density, VP − VS, and PI seismic attributes. An average “c” term value of 1.37 was obtained from the inverse of the slope of the crossplot of P-impedance versus S-impedance for Bonna-6 and Bonna-8 wells. This value was employed in the inversion process to generate the PI attribute, which aided in reducing the non-uniqueness inherent in discriminating the probable reservoir sands. Five seismic attributes slices were generated to ascertain the superiority of each attribute in delineating the probable reservoir sand. These attributes were: density, S-impedance, P-impedance, VP− VS ratio and PI. These attributes reveal low value of density (1.96 − 2.14 g/cc), P-impedance (1.8 × 104 − 2.1 × 104) ft/s*g/cc, S-impedance (9.2 × 103 − 1.1 × 104) ft/s*g/cc, VP − VS (1.65 − 1.72) and PI (4.9 × 103 − 5.1 × 104) ft/s*g/cc around the area inferred to be hydrocarbon saturated reservoir. Although the attributes considered reveals the same zone suspected to be probable hydrocarbon zone, PI gives a better discrimination when compared to other attributes. A distinctive spread and demarcation of the delineated hydrocarbon sand are observed in the PI attribute slice.

A combination of genetic inversion and seismic frequency attributes to delineate reservoir targets in offshore northern Orange Basin, South Africa

Abstract The reported occurrence of Albian- and Cenomanian-aged braided fluvio-deltaic channels in the Orange Basin, South Africa, opens a window of exploration activities to characterize these channels as they are renowned to form some of the world’s giant oil field. In this study, a seismic acoustic impedance inversion and seismic attributes (instantaneous frequency and iso-frequency) analysis is used to investigate potential Albian and Cenomanian fluvio-deltaic channels in offshore, northern Orange Basin. Reservoirs were mapped using a well and 3D seismic volume (8-bit) after initial dip-steering coherency filtering had been performed on the seismic volume to remove incoherent noise and improve data resolution. Model-based acoustic impedance inversion was applied on the seismic volume to delineate fluvio-deltaic channels in addition to using the RMS (root mean square) amplitude attribute. Iso-frequency using the cosine correlative transform (CCT) method was equally applied to delineate these channels. Instantaneous frequency attribute was analyzed for potential hydrocarbon-charged sediments. This was achieved by utilizing thirty-three seismic traces as an input in the Hilbert transform window, after which trace envelope and instantaneous phase were transformed into instantaneous frequency. Acoustic impedance inversion results reveal the presence of two channels within the Cenomanian sequence, which shows high porosity (∼40%) along its geometry. The CCT method shows that the 8 Hz frequency window resolved the presence of a channel within the Albian sequence. A meandering channel within the Albian sequence was equally delineated by the RMS, while the application of instantaneous frequency (IF) attribute indicates the presence of hydrocarbon-charged sediments of Cenomanian age in proximity to a listric normal fault because of the attenuation of frequency observed close to the fault. This study demonstrates a case study of the application of seismic impedance inversion and seismic attributes for the delineation of potential reservoirs and hydrocarbon-charged sediments in a basin.

Hydrocarbon Reservoir Characterization Using Multi-point Stochastic Inversion Technique: A Case Study of Pennay Field

Various limitations mitigate our understanding of the intrinsic properties of subsurface formations during reservoir characterization. These challenges go beyond data set discrepancies, but the various analytical methods, workflow data reconciliation patterns, and scale disparities all affect the final outlook to be inadequate. A multi-point stochastic inversion technique was applied to characterize the three-dimensional seismic and a collection of Pennay Field borehole logs, Niger Delta, outside of well control. The characterization involved the structural interpretation of four horizons, namely the horizons 1–4. The identification of potential hydrocarbon zones involved identification of hydrocarbon-bearing sands, correlation of identified sands across wells, determination of petrophysical properties of identified hydrocarbon-bearing sand, and the integration of well information from seismic data. Potential hydrocarbon zones have been delineated, and the petrophysical properties have been predicted away from well control using the seismic volume data. The net-to-gross and net thicknesses in millisecond were computed over the interval and defined in terms of minimum, maximum, mean, standard deviation, and percentiles P1O, P25, P50, P75, and P90. The 90% net-to-gross values are above 0.6, while the 75% net-to-gross values are between 0.35 and 0.99 in most areas. Moreover, the 50% net-to-gross values range between 0.35 and 0.85 while 25% and 10% net-to-gross values are below 0.4 in most areas in the studied field. Attention should be given to regions with structural closure, higher net-to-gross, and subsequently higher net thickness, low volume of shale, low gamma-ray, low water saturation, and higher connectivity as predicted from the structural, petrophysics, stochastic, and cross-plot analyses for further exploration and exploitation activities.

Theory-guided data science-based reservoir prediction of a North Sea oil field

Data science-based methods, such as supervised neural networks, provide powerful techniques to predict reservoir properties from seismic and well data without the aid of a theoretical model. In these supervised learning approaches, the seismic-to-rock property relationship is learned from the data. One of the major factors limiting the success of these methods is whether there exists enough labeled data, sampled over the expected geology, to train the neural network adequately. To overcome these issues, this paper explores hybrid theory-guided data science (TGDS)-based methods. In particular, we build a two-component model in which the outputs of the theory-based component are the inputs in the data science component. First, we simulate many pseudowells based on the well statistics in the project area. The reservoir properties, such as porosity, saturation, mineralogy, and thickness, are all varied to create a well-sampled data set. Elastic and synthetic seismic data are then generated using rock physics and seismic theory. The resulting collection of pseudowell logs and synthetic seismic data, called the synthetic catalog, is used to train the neural network. The derived operator is then applied to the real seismic data to predict reservoir properties throughout the seismic volume. This TGDS method is applied to a North Sea data set to characterize a Paleocene oil sand reservoir. The TGDS results better characterize the geology and well control, including a blind well, compared to a solely theory-based approach (deterministic inversion) and a data science-based approach (neural network using only the original wells). These results suggest that theory and data science can complement each other to improve reservoir characterization predictions.

Choice of decomposition in multispectral coherence computation

Abstract The iconic coherence attribute generally computed on the full-bandwidth seismic data has been used for 25 years to delineate structural and stratigraphic subsurface geometric features. More recently, multispectral coherence has been introduced to not only better define the more prominent features, but also more subtle features that may be poorly imaged by traditional broadband coherence attribute displays. Multispectral coherence generation utilizes the spectrally decomposed seismic data in the form of complex voice components within the seismic bandwidth. There are various methods available for spectral decomposition of seismic data, including the continuous wavelet transform, complex matching pursuit and the maximum entropy methods. In addition, a nonlinear spectral probe algorithm is also available and could be utilized for multispectral coherence generation. With all these methods available with the seismic interpreter, a question that pops up is which of these methods could be utilized for a more effective generation of multispectral coherence attribute. We first discuss multispectral coherence itself and its generation and follow that with a description of the different spectral decomposition methods mentioned above. Next, we generate spectral voice components and spectral probe components on two seismic data volumes from northeast British Columbia, Canada, and put them through multispectral coherence generation. A comparative performance of the different methods is then evaluated in terms of the multispectral coherence displays. Finally, some conclusions are drawn from the complete exercise.

Forced folding in the Kora Volcanic Complex, New Zealand: A case study with relevance to the production of hydrocarbons and geothermal energy

The intrusion of magma into relatively shallow strata often results in structural deformation. Forced folds are typically formed above intrusions so as to accommodate the addition of new volumes of magma into the sedimentary pile. This work uses a high-resolution 3D seismic volume from the offshore Taranaki Basin, New Zealand to investigate large forced folds formed in Upper Cretaceous-Paleocene sequences due to the intrusion of magmatic dikes and sills. These magmatic intrusions are related to the plumbing system of the Kora volcano. We apply a novel filtering technique to enhance the contrast between the dip-azimuth of seismic reflectors and that of any overprinting noise, preserving the lateral continuity of seismic events. Forced folds were generated by the intrusion of flat-laying saucer-shaped sills into shallow strata. These folds (Fold A and Fold B) have a domal geometry and a structural relief of 250−350 m, for a total area of 45 km2 and 35 km2, respectively. Out of the 18 interpreted sills, two sills (S1 and S4) contributed the most to the generation of the forced folds in the study area. The sill-fold relationships observed in this work suggest that sills with greater length-to-intrusion ratios deform the strata more effectively. However, it is also clear that forced folds are, in parts of the study area, larger than the intruded sills below. This hints at the presence of a larger volume of intruded magma than that imaged on seismic data where forced folds are significantly larger than the volume of the sills below. Such a character makes the establishment of fold-sill relationships crucial to the identification of magmatic bodies below forced folds - if one identifies a large forced fold on seismic profiles, but only a few small sills are imaged, then the mass of igneous rock below the fold is potentially larger than that imaged on seismic data. More importantly, forced folding in the Taranaki Basin is shown to have led to the development of four-way closures to form prolific traps for hydrocarbon and geothermal energy resources.

Evolutionary full-waveform inversion

SUMMARY We present a new approach to full-waveform inversion (FWI) that enables the assimilation of data sets that expand over time without the need to reinvert all data. This evolutionary inversion rests on a reinterpretation of stochastic Limited-memory Broyden–Fletcher–Goldfarb–Shanno (L-BFGS), which randomly exploits redundancies to achieve convergence without ever considering the data set as a whole. Specifically for seismological applications, we consider a dynamic mini-batch stochastic L-BFGS, where the size of mini-batches adapts to the number of sources needed to approximate the complete gradient. As an illustration we present an evolutionary FWI for upper-mantle structure beneath Africa. Starting from a 1-D model and data recorded until 1995, we sequentially add contemporary data into an ongoing inversion, showing how (i) new events can be added without compromising convergence, (ii) a consistent measure of misfit can be maintained and (iii) the model evolves over times as a function of data coverage. Though applied retrospectively in this example, our method constitutes a possible approach to the continuous assimilation of seismic data volumes that often tend to grow exponentially.

Assessment of updated low frequency model in delineating bypassed hydrocarbon reservoir: A 4D seismic study of ’X’ - field, offshore, Niger Delta, Nigeria

Aim/objectives: The aim of this research is, to use Time lapse (4D) seismic and investigate the influence of low frequency update in deterministic model-based seismic inversion employed in delineating a prospect saturated with bypassed hydrocarbon accumulation. Method: The dataset employed in this study incorporates 4D seismic volumes with fifteen (15) years production, interval between 2001 baseline and 2016 monitor seismic vintages. The inversion was carried out using full bandwidth of the updated low frequency and bandpass filtered low frequency approaches. The seismic vintages (baseline and monitor) were simultaneously inverted into acoustic impedance volumes for the two approaches. The formation fluid and lithology were discriminated through fluid replacement modelling (FRM) based on the colour separation between brine and gas saturation scenarios. Findings: The two inversion methods employed reveal six (6) zones suspected to be saturated with bypassed hydrocarbons. The delineated bypassed zones are masked in the full bandwidth approach,depicting the effect of the updated low frequency model. Meanwhile, the bandpass filtered approach result presents a better delineated bypassed reservoir as the zones are more pronounced when compared with the full bandwidth approach. Porosity estimate reveals that the bandpass filtered approach is characterized with excellent porosity in the suspected bypassed zones. The results equally gave more reliable and full delineated bypassed zones. Originality and novelty: The dataset employed in this study were obtained from a producing hydrocarbon field which, interest is to maximize the production of oil/gas. The study will bridge the inherent gab observed in using model-based seismic inversion approach to analyse and interpret seismic data in order to delineate hydrocarbon prospects. The research reveals that,the model-based seismic inversion method is still very effective in delineating hydrocarbon prospect when the updated low frequency is bandpass filtered to remove the model effect which influences the inverted acoustic impedance results. Keywords: Porosity; frequency; bypassed; reservoir and impedance

Evaluation of seismic attributes for reservoir characterization over Edi field, Niger delta, Nigeria using 3d seismic data

Seismic attributes were evaluated over Edi field, offshore Western Niger Delta, Nigeria, via 3D seismic data. Manual mappings of the horizons and faults on the in-lines and cross-lines of the seismic sections were done. Various attributes were calculated and out put on four horizons corresponding to the well markers at different formations within the well were identified. The four horizons identified, which includes: H1, H2, H3 and H4 were mapped and interpreted across the field. The operational agenda was thru picking given faults segments on the in–line of seismic volume. A total of five faults coded as F1, F2, F3, F4 and F5, F1 and F5 were the major fault and were observed as extending through the field. Structural and horizon mappings were used to generate time structure maps. The maps showed the various positions and orientations of the faults. Different attributes which include: root mean square amplitude, instantaneous phase, gradient magnitude and chaos were run on the 3D seismic data. The amplitude and incline magnitude maps indicate direct hydrocarbon on the horizon maps; this is very important in the drilling of wells because it shows areas where hydrocarbons are present in the subsurface. The seismic attributes revealed information, which was not readily apparent in the raw seismic data.  Â

Defining stratigraphic oil and gas plays by modifying structural plays: A case study from the Xihu Sag, east China Sea Shelf Basin

The study focuses on an area of relatively mature oil and gas exploration located in the western slope of the Xihu Sag, East China Sea Shelf Basin. Three stratigraphic oil and gas plays are introduced to replenish the shrinking prospect inventory in the area. The reservoirs of these plays are traditional pay intervals of the Eocene Pinghu Formation and Baoshi Formation composed of tide-dominated delta front, tidal flat and delta front sand bar facies. By investigating conventional seismic data, seismic inversion volumes, and well log data, multiple tidal-influenced delta front channels, beach sand packages have been mapped out in different reservoir sand intervals of the Pinghu Formation and Baoshi Formation. We recognize four types of stratigraphic terminations: channel flank pinch-outs, axial pinch-outs of sand-fills, channel mouth sand bar pinch-outs, and onlap pinch-outs for the Baoshi Formation. Both the depositional facies and relatively low sand-shale ratio of the reservoir formations favor the occurrence of stratigraphic terminations. The closure mechanism of all these plays are similar to the traditional structure plays in the area: fault blocks or fault-bounded rollover anticlines, except that one or more closing faults are replaced by these stratigraphic terminations. Several small-sized closures should exist but the relatively large ones are centered around a major intra-slope rise, raising the exploration potential in this maturely explored area. The new exploration drilling campaign has so far tested the play of delta front channel facies, with three out of four wells commercially successful. These play concepts open up a novel approach to identify new traps in the study area and many other rift basins with similar settings in eastern China.

Automated arrival-time picking using a pixel-level network

Arrival-time picking is an essential step in seismic processing and imaging. The explosion of seismic data volume requires automated arrival-time picking in a faster and more reliable way than existing methods. We have treated arrival-time picking as a binary image segmentation problem and used an improved pixel-wise convolutional network to pick arrival times automatically. Incorporating continuous spatial information in training enables us to preserve the arrival-time correlation between nearby traces, thus helping to reduce the risk of picking outliers that are common in a traditional trace-by-trace picking method. To train the network, we first convert seismic traces into gray-scale images. Image pixels before manually picked arrival times are labeled with zeros, and those after are tagged with ones. After training and validation, the network automatically learns representative features and generates a probability map to predict the arrival time. We apply the network to a field microseismic data set that was not used for training or validation to test the performance of the method. Then, we analyze the effects of training data volume and signal-to-noise ratio on our autopicking method. We also find the difference between 1D and 2D training data with borehole seismic data. Microseismic and borehole seismic data indicate the proposed network can improve efficiency and accuracy over traditional automated picking methods.

Spatial pseudo-labeling for semi-supervised facies classification

For the Quantitative classification of facies is crucial to link seismic data with its corresponding lithology for the evaluation of reservoir properties. During the past decade, seismic volumes have increased to the degree that it is perplexing for the geoscientists and seismic interpreters to examine every single seismic section to carry out facies classification. This has triggered and inspired the need and development of machine learning approaches to efficiently predict seismic facies. However, classified data or labeled data (i.e., well log data) are limited by various restraints and needs huge amount of time and cost to obtain, whereas unlabeled data (seismic data) are in abundance and relatively easy to obtain. Geoscientists are tasked with interpreting large amounts of unlabeled data using very little labeled data. In this paper, we adopt a semi-supervised machine learning approach utilizing pseudo-labeling technique to predict seismic facies and to account for the scarcity of training data by taking advantage of unlabeled data. In each iteration, a small amount of unlabeled data is labeled in spatial manner and the most confident labels (labels that are within 95% confidence interval) are added back to the training data set, starting from labeling the unlabeled data adjacent to the well and spatially moving away in small steps (few traces in each step). The labels obtained from the unlabeled data near the well (pseudo-labels) are added to the labeled data to retrain the deep learning model (classifier) in each iteration. This addition of spatial pseudo-labels adds invariance and diversity to the classifier which in turn helps classifier capitalize on known input variance, which might act as a helpful prior knowledge for next prediction. In other words, the proposed workflow accounts for spatial change in lithology that occurs while moving away from the well. Using our proposed workflow, we observed that, the classification accuracy of a machine learning model trained on a small amount of training data can be enhanced considerably. Pseudo-labels also helped improve lateral continuity of facies. Furthermore, the proposed method outperformed conventional methods, achieving a test accuracy of 96.7% and logarithmic loss as low as 0.15 compared to 91% and 0.6, respectively for supervised baseline. Whereas, for training and validation; the training accuracy, validation accuracy and loss were 98.9%, 99.09% and 0.029 respectively.

A machine learning tool for interpretation of Mass Transport Deposits from seismic data

Machine learning is a tool that allows machines or intelligent systems to learn and get equipped to solve complex problems in predicting reliable outcome. The learning process consists of a set of computer algorithms that are employed to a small segment of data with a view to speed up realistic interpretation from entire data without extensive human intervention. Here we present an approach of supervised learning based on artificial neural network to automate the process of delineating structural distribution of Mass Transport Deposit (MTD) from 3D reflection seismic data. The responses, defined by a set of individual attributes, corresponding to the MTD, are computed from seismic volume and amalgamated them into a hybrid attribute. This generated new attribute, called as MTD Cube meta-attribute, does not only define the subsurface architecture of MTD distinctly but also reduces the human involvement thereby accelerating the process of interpretation. The system, after being fully trained, quality checked and validated, automatically delimits the structural geometry of MTDs within the Karewa prospect in northern Taranaki Basin off New Zealand, where MTDs are evidenced.

Analysis of Microseismicity and Reactivated Fault Size to Assess the Potential for Felt Events by CO2 Injection in the Illinois Basin

ABSTRACTThe results of monitoring of carbon dioxide (CO2) injection at the Illinois Basin—Decatur Project (IBDP) and the companion Illinois Industrial Carbon Capture and Sequestration Sources (IL-ICCS) project—have shown that reservoir response to fluid pressure changes can vary significantly at different injection locations within the same reservoir. Predrill reservoir characterization is important to identify potentially seismogenic faults. However, interpretations of newly reprocessed 3D seismic reflection data illustrate the challenges related to their identification in a region dominated by faulting with small vertical offsets. Faults interpreted in the 3D seismic volume range from ∼300 to 1200 m wide and are in the same size range as faults that could have been the source of historical events up to Mw 2.7 in central Illinois. The array of monitoring sensors that was installed for the IBDP continues to collect data, as injection operates in IL-ICCS, the second injection well. CO2 injection rates for the IL-ICCS well are on average 1.7 times the rates injected in the IBDP well, but a significantly reduced rate of induced seismicity is observed. This article presents results of passive seismic monitoring for the duration of the project to date, integrating active and passive seismic data to develop a new interpretation of the subsurface structure at the Decatur site that explicitly identifies pathways for fluid flow into the basement leading to induced seismicity, and provides a geological explanation for the sharp reduction of induced seismicity during injection at higher rates into the second well. The use of seismic moment to estimate the length of seismogenic slip planes in the local subsurface suggests that faults large enough to produce felt seismicity are unlikely to be present at or near the Decatur site.