Quantifying the sensitivity of seismic facies classification to seismic attribute selection: An explainable machine learning study

During the past two decades, geoscientists have used machine learning to produce a more quantitative reservoir characterization and to discover hidden patterns in their data. However, as the complexity of these models increase, the sensitivity of their results to the choice of the input data becomes more challenging. Measuring how the model uses the input data to perform either a classification or regression task provides an understanding of the data-to-geology relationships which indicates how confident we are in the prediction. To provide such insight, the ML community has developed Local Interpretable Model-agnostic Explanations (LIME), and SHapley Additive exPlanations (SHAP) tools. In this study, we train a random forest architecture using a suite of seismic attributes as input to differentiate between mass transport deposits (MTDs), salt, and conformal siliciclastic sediments in a Gulf of Mexico dataset. We apply SHAP to understand how the model uses the input seismic attributes to identify target seismic facies and examine in what manner variations in the input such as adding band-limited random noise or applying a Kuwahara filter impact the models’ predictions. During our global analysis, we find that the attribute importance is dynamic, and changes based on the quality of the seismic attributes and the seismic facies analyzed. For our data volume and target facies, attributes measuring changes in dip and energy show the largest importance for all cases in our sensitivity analysis. We note that to discriminate between the seismic facies, the ML architecture learns a “set of rules” in multi-attribute space and that overlap between MTDs, salt, and conformal sediments might exist based on the seismic attribute analyzed. Finally, using SHAP at a voxel-scale, we understand why certain areas of interest were misclassified by the algorithm and perform an in-context interpretation to analyze how changes in the geology impact the model’s predictions.

Seismic Attribute-Aided Characterization of Margin Backstepping and Advance along Isolated Carbonate Sequences, Sirt Basin, Libya

Isolated carbonate platforms are common and contain significant hydrocarbon accumulations, particularly in the tectonically complex Sirt Basin in Libya. This study investigates the margin cyclicity of two carbonate stratigraphic sequences developed on an isolated carbonate platform in the NW Sirt Basin using 3-D post-stack seismic volume and wireline log data. The two sequences (sequences 4 and 5) are bounded by unconformity surfaces from the base and top. Seismic attributes show that each sequence displays a cycle of margin backstepping followed by margin advance for several hundred meters. This study concludes that the margin backstepping and advance are mainly influenced by sea-level changes. A rapid sea-level rise caused the backstepping, whereas slow sea-level rise caused the margin advance.

Use of Attributes on 3D Seismic Data for Studying Mishrif Formation in East Abu-Amoud Field, South-Eastern Iraq

The seismic method depends on the nature of the reflected waves from the interfaces between layers, which in turn depends on the density and velocity of the layer, and this is called acoustic impedance. The seismic sections of the East Abu-Amoud field that is located in Missan Province, south-eastern Iraq, were studied and interpreted for updating the structural picture of the major Mishrif Formation for the reservoir in the field. The Mishrif Formation is rich in petroleum in this area, with an area covering about 820 km2. The horizon was calibrated and defined on the seismic section with well logs data (well tops, check shot, sonic logs, and density logs) in the interpretation process to identify the upper and lower boundaries of the Formation. Seismic attributes were used to study the formation, including instantaneous phase attributes and relative acoustic impedance on time slice of 3D seismic data . Also, relative acoustic impedance was utilized to study the top of the Mishrif Formation. Based on these seismic attributes, karst features of the formation were identified. In addition, the nature of the lithology in the study area and the change in porosity were determined through the relative acoustic impedance The overlap of the top of the Mishrif Formation with the bottom of the Khasib Formation was determined because the Mishrif Formation is considered as an unconformity surface.

Application of Techniques of Natural Fracture Characterization for Appraisal of Tight Carbonate Reservoirs: A Case Study From Jurassic of Kuwait

The State of Kuwait is currently appraising and successfully developing the tight carbonates reservoirs of Jurassic age, which have very low matrix porosity and permeability. These reservoirs are affected by several tectonic events of faulting and folding, resulting in the development of interconnected natural fractures, which provide effective permeability to the reservoirs in form of production sweet spots. The objective of the study was to characterize the natural fractures and identify high permeability sweet spots as being appraisal drilling locations in a discovered field with tight carbonate reservoirs. An integrated approach was undertaken for building a discrete fracture network model by characterizing the developed faulting- and folding-related fractures and combining all subsurface data from multiple domains. The reservoir structure has a doubly plunging anticline at the field level that is affected by several strike-slip faults. The faulting-related fractures were characterized by generating multiple structural seismic attributes, highlighting subsurface discontinuities and fracture corridors. The folding-related fractures were modelled using structural restoration techniques by computing stresses resulting from the anticlinal folding. The fracture model was built in addition to the 3D matrix property model for this tight carbonate reservoir, resulting in a dual-porosity-permeability static model. Analogue data was used to compute fracture aperture and expected fracture porosity and permeability, to identify the sweet spots. Structural seismic attributes such as Ant Tracking and Consistent Dip were successful in highlighting and identifying the fault lineaments and fracture corridors. The seismic discontinuities were validated using the fractures interpreted in the image log data from the predrilled wells before being input into the fracture model. Paleo stresses, derived from structural restoration, were combined with the reservoir facies and geomechanical properties to gain important insight into predicting fractures developed due to folding. Several fracture aperture scenarios were run to capture the uncertainty associated with the computed fracture porosity and permeability. Based on the results, several sweet spots were identified, which were ranked based on their extent and connected volumes of the various permeability cases. Identifying these sweet spots helped make informed decisions regarding well planning and drilling sequence. High-inclination wells aligned parallel to the present-day maximum stress direction were proposed, which would cut across corridors of the predicted open fractures. Through this study, comprehensive fracture characterization and fracture permeability understanding of the tight carbonates in the field under study were successfully achieved. This workflow will be useful in exploratory or appraisal fields with tight carbonate reservoirs.

Biot's equations-based reservoir parameter inversion using deep neural networks

Reservoir parameter inversion from seismic data is an important issue in rock physics. The traditional optimisation-based inversion method requires high computational expense, and the process exhibits subjectivity due to the nonuniqueness of generated solutions. This study proposes a deep neural network (DNN)-based approach as a new means to analyse the sensitivity of seismic attributes to basic rock-physics parameters and then realise fast parameter inversion. First, synthetic data of inputs (reservoir properties) and outputs (seismic attributes) are generated using Biot's equations. Then, a forward DNN model is trained to carry out a sensitivity analysis. One can in turn investigate the influence of each rock-physics parameter on the seismic attributes calculated by Biot's equations, and the method can also be used to estimate and evaluate the accuracy of parameter inversion. Finally, DNNs are applied to parameter inversion. Different scenarios are designed to study the inversion accuracy of porosity, bulk and shear moduli of a rock matrix considering that the input quantities are different. It is found that the inversion of porosity is relatively easy and accurate, while more information is needed to make the inversion more accurate for bulk and shear moduli. From the presented results, the new approach makes it possible to realise accurate and pointwise inverse modelling with high efficiency for actual data interpretation and analysis.

A review of some amplitude-based seismic geometric attributes and their applications

Seismic interpreters frequently use seismic geometric attributes such as coherence, dip, curvature, and aberrancy for defining geological features, including faults, channels, angular unconformities, etc. Some of the commonly used coherence attributes, e. g. cross-correlation or energy ratio similarity are sensitive to only waveform shape changes, whereas the dip, curvature, aberrancy attributes are based on changes in reflector dips. There is another category of seismic attributes, which includes attributes that are sensitive to amplitude values. Root mean square amplitude is one of the better-known amplitude-based attributes, whereas coherent energy, Sobel-filter similarity, normalized amplitude gradients, and amplitude curvature are amongst lesser-known amplitude-based attributes. We compute not-so-common amplitude-based attributes on the Penobscot seismic survey from the Nova Scotia continental shelf consisting of the east coast of Canada, to bring out their interpretative value. We analyze seismic attributes at the level of the top of the Wyandot Formation that exhibits different geological features, including a synthetic transfer zone with two primary faults and several secondary faults, polygonal faults associated with differential compaction, as well as fixtures related to basement related faults. The application of the amplitude-based seismic attributes defines such features accurately. We take these applications forward by describing a situation where some geological features do not display any bending of reflectors, but only exhibit changes in amplitude. One of such examples is the Cretaceous Cree Sand channels, present in the same 3D seismic survey used for the previous applications. We compute amplitude curvature attributes and identify the channels, whereas these channels are not visible on the structural curvature display. In both the applications, we observe that appropriate corendering not-so-common amplitude based seismic attributes leads to convincing displays, that can be of immense aid in seismic interpretation and help define the different subsurface features with more clarity.