GANSim-3D for conditional geomodelling: theory and field application

Geomodelling of subsurface reservoirs is important for water resources, hydrocarbon exploitation, and Carbon Capture and Storage (CCS). Traditional geostatistics-based approaches cannot abstract complex geological patterns and are thus not able to simulate very realistic earth models. We present a Generative Adversarial Networks (GANs)-based 3D reservoir simulation framework, GANSim-3D, which can capture geological patterns and relationships between various conditioning data and earth models and is thus able to directly simulate multiple 3D realistic and conditional earth models of arbitrary sizes from given conditioning data. In GANSim-3D, the generator, designed to only include 3D convolutional layers, takes various 3D conditioning data and 3D random latent cubes (composed of random numbers) as inputs and produces a 3D earth model. Two types of losses, the original GANs loss and condition-based loss, are designed to train the generator progressively from shallow to deep layers to learn the geological patterns and relationships from coarse to fine resolutions. Conditioning data can include 3D sparse well facies data, 3D low-resolution probability maps, and global features like facies proportion, channel width, etc. Once trained on a training dataset where each training sample is a 3D cube of a small fixed size, the generator can be used for geomodelling of 3D reservoirs of large arbitrary sizes by directly extending the sizes of all inputs and the output of the generator proportionally. To illustrate how GANSim-3D is used for field geomodelling and also to verify GANSim-3D, a field karst cave reservoir in Tahe area of China is used as an example. The 3D well facies data and 3D probability map of caves obtained from geophysical interpretation are used as conditioning data. First, we create a training dataset consisting of facies models of 64×64×64 cells with a process-mimicking simulation method to integrate field geological patterns. The training well facies data and the training probability map data are produced from the training facies models. Then, the 3D generator is successfully trained and evaluated in two synthetic cases with various metrics. Next, we apply the pretrained generator for conditional geomodelling of two field cave reservoirs of Tahe area. The first reservoir is 800m×800m×64m and is divided into 64×64×64 cells, while the second is 4200m×3200m×96m and is divided into 336×256×96 cells. We fix the input well facies data and cave probability maps and randomly change the input latent cubes to allow the generator to produce multiple diverse cave reservoir realizations, which prove to be consistent with the geological patterns of real Tahe cave reservoir as well as the input conditioning data. The noise in the input probability map is suppressed by the generator. Once trained, the geomodelling process is quite fast: each realization with 336×256×96 cells takes 0.988 seconds using 1 GPU (V100). This study shows that GANSim-3D is robust for fast 3D conditional geomodelling of field reservoirs of arbitrary sizes.

3D GPR attributes to generate classified facies models: A case study from a dune island

Ground-penetrating radar (GPR) is a standard geophysical technique to image near-surface structures in sedimentary environments. In such environments, GPR data acquisition and processing are increasingly following 3D strategies. However, the processed GPR data volumes are typically still interpreted using selected 2D slices and manual concepts such as GPR facies analyses. In seismic volume interpretation, the application of (semi-)automated and reproducible approaches such as 3D attribute analyses as well as the production of attribute-based facies models are common practice today. In contrast, the field of 3D GPR attribute analyses and corresponding facies models is largely untapped. We develop and apply a workflow to produce 3D attribute-based GPR facies models comprising the dominant sedimentary reflection patterns in a GPR volume which images complex sandy structures on the dune island of Spiekeroog (Northern Germany). After presenting our field site and details regarding our data acquisition and processing, we calculate and filter 3D texture attributes to generate a database comprising the dominant texture features of our GPR data. Then, we perform a dimensionality reduction of this database to obtain meta texture attributes, which we analyze and integrate using composite imaging and (also considering additional geometric information) fuzzy c-means cluster analysis resulting in a classified GPR facies model. Considering our facies model and a corresponding GPR facies chart, we interpret our GPR data set in terms of near-surface sedimentary units, the corresponding depositional environments, and the recent formation history at our field site. Thus, we demonstrate the potential of the proposed workflow, which represents a novel and clear strategy to perform a more objective and consistent interpretation of 3D GPR data collected across different sedimentary environments.

Fractured reservoir distribution characterization using folding mechanism analysis and patterns recognition in the Tabnak hydrocarbon reservoir anticline

Naturally, fractured reservoirs play a considerable part in the study, production, and development of hydrocarbon fields because most hydrocarbon reservoirs in the Zagros Basin are naturally fractured. Production from those reservoirs is usually affected by the presence of a system of connected fractures. In this study, the Tabnak hydrocarbon field on the fold–thrust belt at the Zagros zone in the Persian plate has been analyzed by the facies models, folding mechanism analysis to identify fracture reservoir patterns. The results show a flexural fold with similarity in the folding mechanism and some open fracture potential made by limestone, shale, clay, and anhydrite in the study area's facies models. Consequently, the stress pattern and type of fracture issue on the fold's upper and lower layers will be similar. On the Tabnak anticline reservoir using image processing techniques in MATLAB R2019 software and kriging geostatistical methods, fracture surface patterns as a block model extended to the depth. Using the model results, fractures’ orientation distribution in adjacent wells 11, 14, and 15 is appropriate. The results also have similarities with the facies models, folding mechanism assessment, well test, and mud loss data analysis. These results can affect the development plans’ primary approach by drilling horizontal and sleep wells and hydrocarbon reservoir management strategies.

An Improved Method of Reservoir Facies Modeling Based on Generative Adversarial Networks

As the reservoir and its attribute distribution are obviously controlled by sedimentary facies, the facies modeling is one of the important bases for delineating the area of high-quality reservoir and characterizing the attribute parameter distribution. There are a large number of continental sedimentary reservoirs with strong heterogeneity in China, the geometry and distribution of various sedimentary microfacies are relatively complex. The traditional geostatistics methods which have shortage in characterization of the complex and non-stationary geological patterns, have limitation in facies modeling of continental sedimentary reservoirs. The generative adversarial network (GANs) is a recent state-of-the-art deep learning method, which has capabilities of pattern learning and generation, and is widely used in the domain of image generation. Because of the similarity in content and structure between facies models and specific images (such as fluvial facies and the images of modern rivers), and the various images generated by GANs are often more complex than reservoir facies models, GANs has potential to be used in reservoir facies modeling. Therefore, this paper proposes a reservoir facies modeling method based on GANs: (1) for unconditional modeling, select training images (TIs) based on priori geological knowledge, and use GANs to learn priori geological patterns in TIs, then generate the reservoir facies model by GANs; (2) for conditional modeling, a training method of “unconditional-conditional simulation cooperation” (UCSC) is used to realize the constraint of hard data while learning the priori geological patterns. Testing the method using both synthetic data and actual data from oil field, the results meet perfectly the priori geological patterns and honor the well point hard data, and show that this method can overcome the limitation that traditional geostatistics are difficult to deal with the complex non-stationary patterns and improve the conditional constraint effect of GANs based methods. Given its good performance in facies modeling, the method has a good prospect in practical application.

The turbidite-contourite-tidalite-baroclinite-hybridite problem: orthodoxy vs. empirical evidence behind the “Bouma Sequence”

The underpinning problems of deep-water facies still remain unresolved. (1) The Tb, Tc, and Td divisions of the turbidite facies model, with traction structures, are an integral part of the “Bouma Sequence” (Ta, Tb, Tc, Td, Te). However, deposits of thermohaline contour currents, wind-driven bottom currents, deep-marine tidal currents, and baroclinic currents (internal waves and tides) also develop discrete rippled units, mimicking Tc. (2) The application of “cut-out” logic of sequences, which was originally introduced for the “Bouma Sequence”, with sharp basal contacts and sandy divisions containing well-developed traction structures, to muddy contorts with gradational basal contacts and an absence of well-developed traction structures is incongruent. (3) The presence of five internal divisions and hiatus in the muddy contoured facies model is in dispute. (4) Intersection of along slope contour currents with down slope sediment-gravity flows, triggering hybrid flows, also develops traction structures. (5) The comparison of genuine hybrid flows with down slope flow transformation of gravity flows is inconsistent with etymology of the term “hybrid”. (6) A reexamination of the Annot Sandstone at the Peira Cava type locality in SE France fails to validate either the orthodoxy of five internal divisions of the “Bouma Sequence” or their origin by turbidity currents. For example, the “Ta” division is composed of amalgamated units with inverse grading and floating mudstone clasts, suggesting a mass-transport deposit (MTD). The “Tb” and “Tc” divisions are composed of double mud layers and sigmoidal cross bedding, respectively, which suggest a tidalite origin. (7) Although it was reasonable to introduce a simplistic “Bouma Sequence” in 1962, at a time of limited knowledge on deep-water processes, it is obsolete now in 2021 to apply this model to the rock record amid a wealth of new knowledge. (8) The disconnect between 12 observed, but questionable, modern turbidity currents and over 10,000 interpreted ancient turbidites defies the doctrine of uniformitarianism. This disconnect is attributed to routine application of genetic facies models, without a pragmatic interpretation of empirical data. (9) A suggested solution to these problems is to interpret traction structures in the sedimentary record pragmatically on the basis of empirical field and experimental evidence, without any built-in bias using facies models, such as the “Bouma Sequence”. (10) Until reliable criteria are developed to distinguish traction structures of each type of bottom currents based on uniformitarianism, a general term “BCRS” (i.e., bottom-current reworked sands) is appropriate for deposits of all four kinds of bottom currents.

How to link modern and ancient barrier island systems: Dimensional comparisons and updated sedimentary facies models

<p>Existing depositional and facies models for ancient barrier island systems are primarily based on modern observations. This approach overlooks processes tied to geologic time scales, such as multi-directional motion, erosion, and reworking, and their resulting expressions in preserved strata. We have investigated these and other challenges of linking modern and ancient barrier islands through outcrop studies and through data compilation from the rock record compared to modern barrier island dimensions. Results emphasize key depositional and preservation processes, and the dimensional differences between deposits formed over geologic versus modern time scales. For example, when comparing deposits from individual barrier islands, thickness measurement comparisons between modern and ancient examples do not vary systematically, suggesting that local accommodation and reworking dictate barrier island thickness preservation. A complementary outcrop study focusing on paralic strata from the Upper Cretaceous Straight Cliffs Formation in southern Utah (USA) is used to update models for barrier island motion and preservation to include geologic time-scale processes. Barrier island deposits are described using four facies associations (FA): backbarrier fill (FA1), lower and upper shoreface (FA2), proximal upper shoreface (FA3), and tidal channel facies (FA4). Three main architectural elements (barrier island shorefaces, shoreface-dominated inlet fill, and channel-dominated inlet fill) occur independently or in combination to create stacked barrier island deposits. Barrier island shorefaces record progradation, while shoreface-dominated inlet fill records lateral migration, and channel-dominated inlet fill records aggradation within the tidal inlet. Barrier islands are bound by lagoons or estuaries and are distinguished from other shoreface deposits by their internal facies and outcrop geometry, association with backbarrier facies, and position within transgressive successions. Tidal processes, in particular, tidal inlet migration and reworking of the upper shoreface, also distinguish barrier island successions. In sum, these datasets demonstrate that improved depositional and facies models must consider multidirectional island motion, ravinement, erosion, inlet migration, and reworking when describing processes and predicting barrier island dimensions.</p>

Subsurface heterogeneity patterns that emerge from interacting depositional processes

<p>Emergent structures define organizational patterns that spontaneously develop due to interactions between component properties or behaviors of complex dynamic systems, rather than being a simple compilation of the individual parts observed within the system at any one time. Traditional facies models used to predict subsurface lithic variations focus on defining the distribution of depositional environments on Earth’s surface and relating the hierarchy of preserved bedding units to different scales of surficial bedforms. It is increasingly recognized that such static models fail to predict the geometry and character of many types of preserved lithic bodies and discontinuity surfaces unless these observations are placed within the context of the overall evolving system. Numerical depositional process models are presented to show links between evolving depositional patterns and preserved facies patterns within different settings.</p><p>     Channel deposit internal variations tend not to be channel shaped, but rather sweet spots within the deposit resemble a string of beads, each formed as individual channel segments meander. Mouth bar deposits generally do not to have the circular to elliptical shape of a modern channel-mouth bedform, but rather tend to be more elongate fingers cut by a diachronous channel filled as river flows are choked off by loss of gradient during progradation. Although the final channel basal erosion surface appears continuous, timelines cross this surface along the length of the deposit. Deltaic shorelines that look identical at a given time preserve very different deposits when the feeding river avulses at different frequency, a condition that can change within an individual deposit formed alternately during periods of sea level rise and fall. Even major stratigraphic surfaces, like lowstand fluvial incision surfaces and wave-ravined falling stage and transgressive surfaces, are likely to gradually emerge from the migration of localized areas of erosion that were never as extensive at any one time as the preserved surface. Such surfaces may be regionally diachronous, with deposits of the same age locally preserved variably above and below the surface. Understanding emergent lithic bodies and internal heterogeneity patterns are fundamental to understanding how deposition is recorded in the rock record and for facies models used to predict how subsurface fluids move through shallow marine deposits.</p>

When Is a Barrier Island Not an Island? When It Is Preserved in the Rock Record

Existing barrier island facies models are largely based on modern observations. This approach highlights the heterogeneous and dynamic nature of barrier island systems, but it overlooks processes tied to geologic time scales, such as multi-directional motion, erosion, and reworking, and their expressions as preserved strata. Accordingly, this study uses characteristic outcrop expressions from paralic strata of the Upper Cretaceous Straight Cliffs Formation in southern Utah to update models for barrier island motion and preservation to include geologic time-scale processes. Results indicate that the key distinguishing facies and architectural elements of preserved barrier island systems have very little to do with “island” morphology as observed in modern systems. Four facies associations are used to describe and characterize these barrier island architectural elements. Barrier islands occur in association with backbarrier fill (FA1) and internally contain lower and upper shoreface (FA2), proximal upper shoreface (FA3), and tidal channel facies (FA4). Three main architectural elements (barrier island shorefaces, shoreface-dominated inlet fill, and channel-dominated inlet fill) occur independently or in combination to create stacked barrier island deposits. Barrier island shorefaces record progradation, while shoreface-dominated inlet fill records lateral migration, and channel-dominated inlet fill records aggradation within the tidal inlet. Barrier islands are bound by lagoons or estuaries and are distinguished from other shoreface deposits by their internal facies and outcrop geometry, association with backbarrier facies, and position within transgressive successions. Tidal processes, in particular, tidal inlet migration and reworking of the upper shoreface, also distinguish barrier island successions. In sum, this study expands barrier island facies models and provides new recognition criteria to account for the complex geometries of time-transgressive, preserved barrier island deposits.