Multiple Seismic Attributes Research Articles

Research on the prediction method for fluvial-phase sandbody connectivity based on big data analysis--a case study of Bohai a oilfield

The connectivity of sandbodies is a key constraint to the exploration effectiveness of Bohai A Oilfield. Conventional connectivity studies often use methods such as seismic attribute fusion, while the development of contiguous composite sandbodies in this area makes it challenging to characterize connectivity changes with conventional seismic attributes. Aiming at the above problem in the Bohai A Oilfield, this study proposes a big data analysis method based on the Deep Forest algorithm to predict the sandbody connectivity. Firstly, by compiling the abundant exploration and development sandbodies data in the study area, typical sandbodies with reliable connectivity were selected. Then, sensitive seismic attribute were extracted to obtain training samples. Finally, based on the Deep Forest algorithm, mapping model between attribute combinations and sandbody connectivity was established through machine learning. This method achieves the first quantitative determination of the connectivity for continuous composite sandbodies in the Bohai Oilfield. Compared with conventional connectivity discrimination methods such as high-resolution processing and seismic attribute analysis, this method can combine the sandbody characteristics of the study area in the process of machine learning, and jointly judge connectivity by combining multiple seismic attributes. The study results show that this method has high accuracy and timeliness in predicting connectivity for continuous composite sandbodies. Applied to the Bohai A Oilfield, it successfully identified multiple sandbody connectivity relationships and provided strong support for the subsequent exploration potential assessment and well placement optimization. This method also provides a new idea and method for studying sandbody connectivity under similar complex geological conditions.

Integration of Geological Model and Numerical Simulation Technique to Characterize the Remaining Oil of Fractured Biogenic Limestone Reservoirs

AbstractFine geological modeling leads to accurate reservoirs numerical simulations. Fractured biogenic limestone has abundant storage spaces and flow paths to accumulate oil and gas. The complexity and diversity of fractured biogenic limestone also lead to challenges in accurately characterizing its pore volume and remaining oil. This investigation aimed to enhance the understanding of fractured biotite reservoir properties via geological modeling. Numerical simulations were used to characterize the remaining oil during the late stage of field development. Considering the differences in porosity and permeability between fractures and matrix, a facies-controlled stochastic modeling technique was used to establish a dual-porosity and dual-permeability (DPDP) model for numerical simulation. Core information, logging data, and multiple seismic attributes were combined to guide low-level sequence fault interpretation for tectonic refinement. Based on classified seismic inversion, sedimentary phases were reconstructed. A discrete fracture network (DFN) model was obtained based on fracture occurrences and density models. The optimized discrete adjoint (ODA) algorithm was utilized to calibrate model parameters. The findings revealed that dense tectonic fractures develop in thick biogenic limestone areas. Combined with advanced reservoir simulation technology, these findings suggest that areas of thicker biogenic limestone were consistent with areas of higher fracture matrix conductivity multipliers. The remaining oil distribution patterns were investigated, and to deploy new wells was guided. Therefore, it is essential to better understand the tectonic characteristics of fractured biogenic limestone reservoirs and their remaining oil distribution patterns by integrating multiple sources of information and mastering advanced reservoir simulation technology for oilfield development.

Deep-learning-based natural fracture identification method through seismic multi-attribute data: a case study from the Bozi-Dabei area of the Kuqa Basin, China

Fractures play a crucial role in tight sandstone gas reservoirs with low permeability and low effective porosity. If open, they not only significantly increase the permeability of the reservoir but also serve as channels connecting the storage space. Among numerous fracture identification methods, seismic data provide unique advantages for fracture identification owing to the provision of three-dimensional information between wells. How to accurately identify the development of fractures in geological bodies between wells using seismic data is a major challenge. In this study, a tight sandstone gas reservoir in the Kuqa Basin (China) was used as an example for identifying reservoir fractures using deep-learning-based method. First, a feasibility analysis is necessary. Intersection analysis between the fracture density and seismic attributes (the characteristics of frequency, amplitude, phase, and other aspects of seismic signals) indicates that there is a correlation between the two when the fracture density exceeds a certain degree. The development of fractures is closely related to the lithology and structure, indirectly affecting differences in seismic attributes. This indicates that the use of seismic attributes for fracture identification is feasible and reasonable. Subsequently, the effective fracture density data obtained from imaging logging were used as label data, and the optimized seismic attribute near the well data were used as feature data to construct a fracture identification sample dataset. Based on a feed-forward neural network algorithm combined with natural fracture density and effectiveness control factor constraints, a trained identification model was obtained. The identification model was applied to seismic multi-attribute data for the entire work area. Finally, the accuracy of the results from the training, testing, and validation datasets were used to determine the effectiveness of the method. The relationship between the fracture identification results and the location of the fractures in the target reservoir was used to determine the reasonableness of the results. The results indicate that there is a certain relationship between multiple seismic attributes and fracture development, which can be established using deep learning models. Furthermore, the deep-learning-based seismic data fracture identification method can effectively identify fractures in the three-dimensional space of reservoirs.

Ensemble of Neural Networks Utilizing Seismic Attributes for Rock‐Property Inversion With Uncertainty Estimation

AbstractSeismic inversion holds significant importance across various domains of geoscience and engineering, including the characterization of energy resource reservoirs, the assessment of polluted sites, and CO2 storage. It is a process of estimating rock properties from seismic data that is inherently uncertain, nonlinear, non‐unique, and highly challenging. Using multiple seismic attributes increases the size of the data, requiring considerable processing resources and time. However, deep learning can accurately fit quantities of nonlinear variables, making it an excellent method for predicting spatially distributed subsurface properties. We trained some multi‐output regression neural networks to carry out porosity inversion from seismic data. We initially computed a series of seismic attributes and generated the corresponding porosity using interpreted horizons, well logs, and seismic data. Subsequently, we proposed a technique to identify the most relevant seismic attributes for porosity inversion. Because our networks work as stochastic modeling entities, we created a weight‐averaging ensemble approach to build a strong model with the highest level of accuracy. We combined realizations from baseline entities, considering their respective performance levels. Using the statistics between these realizations and the robust model, we determined the degree of uncertainty associated with the outcome. We found an R2 of 0.993 and an MAE of 0.00112 in the F3 block offshore the Netherlands, proving the method's effectiveness. The mean porosity was 0.175193, compared to 0.175626 from a reference model, and the mean uncertainty was ±0.0008998.

MultiURNet for 3D seismic fault attributes fusion detection combined with PCA

Identifying faults and fractures in complex block reservoirs is critical for recovering residual oil-gas. It has been found that improving the identification accuracy of neural networks and enhancing their ability to learn multiple fault features are key to accurately identifying complex faults, including low-order and hidden faults. In this study, we developed a U-shaped residual network (URNet) that intelligently identifies seismic faults using principal component analysis (PCA). Numerous synthetic seismic records were generated using convolution method. Multiple seismic attributes were generated based on amplitude attributes and preferred attribute bodies. PCA is used to pre-process the features of multi-attribute bodies, downscale and preserve the original data information and reduce data complexity and storage space. A residual block was added to the deep coding and decoding network structure to improve the expression capability of the neural network while preventing gradient vanishing. Naming the network structure as MultiURNet and the network model was applied to the generated data and actual data, and the identification of the high-order and low-order faults were both significantly improved with the clarity of the fault lines. The results confirm the feasibility and efficiency of the proposed method, which has significant potential for future exploration in shallow and medium-shallow oil and gas fields.

DFN modelling constrained by multiple seismic attributes using the steering pyramid technology

Fracture modelling is essential for understanding fluid flow in fractured hydrocarbon reservoirs, particularly in the phase of production; however, traditional discrete fracture network (DFN) modelling methods lack constraints that reflect characteristics of fracture development. Fractures or fracture networks exhibit a high degree of randomness; as such, it is difficult to model fracture characteristics. This paper proposes a new approach for DFN modelling constrained by seismic attributes. Firstly, the steerable pyramid method is adopted to improve seismic data resolution; secondly, multiple seismic attributes are extracted and combined into a composite attribute to characterize fracture spatial distribution; finally, a DFN modelling method is established by using the composite attribute as a location constraint. To verify the effectiveness of the approach, a case study is conducted in the Bonan Depression, in East China. The results show that, compared with the traditional DFN modelling methods, the DFN modelling with the location constraint create a more realistic fracture model which accurately reflects fracture distribution characteristics. The application demonstrates the potential of wide application prospects in fractured reservoirs.

An advanced ensemble modeling approach for predicting carbonate reservoir porosity from seismic attributes

This study uses a machine learning (ML) ensemble modeling approach to predict porosity from multiple seismic attributes in one of the most promising Main Dolomite hydrocarbon reservoirs in NW Poland. The presented workflow tests five different model types of varying complexity: K-nearest neighbors (KNN), random forests (RF), extreme gradient boosting (XGB), support vector machine (SVM), single layer neural network with multilayer perceptron (MLP). The selected models are additionally run with different configurations originating from the pre-processing stage, including Yeo–Johnson transformation (YJ) and principal component analysis (PCA). The race ANOVA method across resample data is used to tune the best hyperparameters for each model. The model candidates and the role of different pre-processors are evaluated based on standard ML metrics – coefficient of determination (R2), root mean squared error (RMSE), and mean absolute error (MAE). The model stacking is performed on five model candidates: two KNN, two XGB, and one SVM PCA with a marginal role. The results of the ensemble model showed superior accuracy over single learners, with all metrics (R2 0.890, RMSE 0.0252, MAE 0.168). It also turned out to be almost three times better than the neural net (NN) results obtained from commercial software on the same testing set (R2 0.318, RMSE 0.0628, MAE 0.0487). The spatial distribution of porosity from the ensemble model indicated areas of good reservoir properties that overlap with hydrocarbon production fields. This observation completes the evaluation of the ensemble technique results from model metrics. Overall, the proposed solution is a promising tool for better porosity prediction and understanding of heterogeneous carbonate reservoirs from multiple seismic attributes.

Regional stratigraphically discordant dolomite mapping based on a dolomitization mechanism on the Arabian carbonate shelf

Dolomite mapping, as a first step of carbonate diagenetic modeling, is critical to understanding dolomitization mechanisms and predicting reservoir quality. Stratigraphically discordant dolomite bodies within the Upper Jurassic intervals have long been studied on the Arabian shelf. Our dolomitization mechanisms find that faults/fractures serve as migration conduits and determine dolomite distribution by controlling fluid flow. We establish an innovative approach based on the diagenetic mechanisms and comprehensive characterization of fracture systems at multiple scales to delineate dolomite in the study area. The complex fracture networks in the subsurface are categorized into macroscale, mesoscale, and microscale fractures. Macroscale fractures, i.e., faults, are much greater than the seismic wavelength and are easily observed in seismic sections due to their obvious seismic discontinuities. Mesoscale fractures are slightly greater than or equal to the seismic wavelength and are recognized using seismic attributes. Microscale fractures are significantly smaller than the seismic wavelength and are observed mainly on core samples and thin sections. The dolomite mapping workflow includes three steps: (1) calibrate the seismic response of multiscale fractures with borehole image logs and core interpretations, (2) implement multiscale fracture characterization using multiple seismic attributes, and (3) interpret dolomite bodies based on fracture characterization and log identification. The mapping results show that massive dolomite bodies are heterogeneously developed and distributed mainly in the northeastern part of the study area, with a southward decrease in dolomite content. The application demonstrates that multiscale fracture systems play critical roles in massive dolomitization, and multiscale fracture prediction provides an innovative solution to delineate complex dolomitization systems and the combination of reservoir and seal in diagenetic traps.

A modified method of multiple point geostatistics for spatial simulation of sedimentary facies for carbonate reservoirs

Sedimentary facies information plays an important role in elastic parameter modeling, seismic inversion, and other aspects of marine oil and gas exploration. Multiple-point geostatistics is an important method used to obtain accurate and reliable simulation sedimentary facies data; the filtersim algorithm. However, it is difficult to use the filtersim algorithm to extract the features in a training image by scanning indirect data, such as multiple seismic attributes, as several filters are required. Therefore, we constructed a new method by applying filters to directly scan the training image for a target, such as sedimentary facies, from which two types of prototypes of sedimentary facies and seismic attributes could be obtained. In addition, we modified the method of dividing the filter score and added the score range for the sedimentary facies boundary, which increased the accuracy of the pattern classification. We reduced the number of prototypes with multiple filters by dividing each sedimentary facies prototype by its boundary prototype. Subsequently, we eliminated some empty bins, and improved the efficiency of the simulation. We verified the accuracy of this simulation by comparing the distribution of sedimentary facies in well logs. The results showed that this simulation algorithm can effectively construct 2D or 3D spatial distributions of sedimentary facies in carbonate reservoir by adding the constraints of sedimentary facies and seismic attributes to the filtersim algorithm. The simulation results have valuable implications for situations in which marine oil and gas exploration is being conducted with limited information.

Seismic prediction of porosity in tight reservoirs based on transformer

Porosity is a crucial index in reservoir evaluation. In tight reservoirs, the porosity is low, resulting in weak seismic responses to changes in porosity. Moreover, the relationship between porosity and seismic response is complex, making accurate porosity inversion prediction challenging. This paper proposes a Transformer-based seismic multi-attribute inversion prediction method for tight reservoir porosity to address this issue. The proposed method takes multiple seismic attributes as input data and porosity as output data. The Transformer mapping transformation network consists of an encoder, a multi-head attention layer, and a decoder and is optimized for training with a gating mechanism and a variable selection module. Applying this method to actual data from a tight sandstone gas exploration area in the Sichuan Basin yielded a porosity prediction coincidence rate of 95% with the well data.

A novel method for seismic-attribute optimization driven by forward modeling and machine learning in prediction of fluvial reservoirs

The use of seismic attributes for hydrocarbon exploration is well established, and the integration of multiple attributes by supervised machine learning is increasingly being applied as an effective method for attribute optimization. However, this method relies upon a dense array of wells to be employed as a training dataset, which limits its application to fields with sparse boreholes, especially offshore. To address this limitation, this study proposes a novel method of seismic attribute integration driven by forward seismic modeling, enabling the integration of multiple attributes by supervised learning in fields with a limited number of wells. This proposed method consists of two main tasks: (i) establishing a forward geological (lithological) model on a well-correlation section by forward seismic modeling, based on seismic data and wells from the study area; (ii) fusing multiple seismic attributes by supervised machine learning, employing the forward geological model and its synthetic seismic reflection as the training dataset. This method is applied and tested on a real-world case study from the East China Sea region. In this case study, seismic surface attributes of root-mean-square amplitude, max peak amplitude and sweetness are selected and then integrated using the proposed method. The integrated seismic attribute shows significant advantages for the detection of channel belts. Notably, it results in markedly improved correlation between seismic attribute and sand thickness, with the correlation coefficient increasing from 0.586 to 0.849, compared to the original seismic attribute.

Automatic Horizon Picking Using Multiple Seismic Attributes and Markov Decision Process

Picking the reflection horizon is an important step in velocity inversion and seismic interpretation. Manual picking is time-consuming and no longer suitable for current large-scale seismic data processing. Automatic algorithms using different seismic attributes such as instantaneous phase or dip attributes have been proposed. However, the computed attributes are usually inaccurate near discontinuities. The waveforms in the horizontal direction often change dramatically, which makes it difficult to track a horizon using the similarity of attributes. In this paper, we propose a novel method for automatic horizon picking using multiple seismic attributes and the Markov decision process (MDP). For the design of the MDP model, the decision time and state are defined as the horizontal and vertical spatial position on a seismic image, respectively. The reward function is defined in multi-dimensional feature attribute space. Multiple attributes can highlight different aspects of a seismic image and therefore overcome the limitations of the single-attribute MDP through the cross-constraint of multiple attributes. The optimal decision is made by searching the largest state value function in the reward function space. By considering cumulative reward, the lateral continuity of a seismic image can be effectively considered, and the impacts of abnormal waveform changes or bad traces in local areas for automatic horizon picking can be effectively avoided. An effective implementation scheme is designed for picking multiple reflection horizons. The proposed method has been successfully tested on both synthetic and field data.

Utilization of seismic attributes for structural pattern detection In Bualawn, Dor Mansour fields, Western Sirt Basin, Libya

An integrated approach to the study of fault patterns carried out in the complex geological structures of Bualawn, Dor Mansour fields by using multiple seismic attributes of 3-D seismic data. Each type of geological structure event usually generates a unique seismic signature that can be recognized and identified. This paper highlights the practical importance of analyze integrated attributes and interpret them within the context of an appropriate structural deformation. Many of discontinuity attributes such as variance, 3DEE (3D edge enhancement), ant tracking, chaos and structure smoothing have been selected to delineate fault styles and their displacement effectively, which cannot be fully delineated using seismic amplitude data. Conjugate growth faults in NW-SE and NE-SW direction, dipping toward SW with antithetic faults dipping in opposite direction to the growth faults and other minor faults. The faults were identified manually and then tested by discontinuity attributes. Fault system is represented by a fracture system comprising long en echelon seismic faults and several discontinuities. This system of en echelon faults was initiated by left-lateral wrenching movements. This wrenching system has been identified by negative and positive flower structure along the releasing and restraining bends. The current understanding that these faults are strike-slip faults despite the absence of extensive horizontal displacements along them as shown on different time slices. A periodic strike-slip movement along deep-seated basement faults have developed many structural features, such as horst-graben styles, which is the orientation of the transtensional movements linked to the Upper Cretaceous Succession.

Multiple seismic attributes fusion approach with support vector regression and forward simulation for sand body prediction and sedimentary facies interpretation — A case of the X gas field in Xihu Sag

Marine exploration and production play a vital role in petroleum industries. It is difficult to acquire sufficient well data in marine settings, so seismic data become the most important interpretation data in research. In general, the seismic data in marine exploration have low quality because of the deep depth from the surface. To partly address this, the support vector regression (SVR) algorithm is proposed to fuse multiple seismic attributes for sand thickness prediction. First, we use forward modeling to establish virtual wells for improving the training data set. Second, we select the optimal attributes by correlation analysis. Third, we apply the SVR algorithm to learn the relationship between seismic attributes and sand thickness. Fourth, we use the SVR model to predict the sand thickness between wells by calculating a fused attribute. The results indicate that the fusion attribute with SVR has a higher correlation coefficient with sand thickness than the original attributes by statistical method. The approach can be widely used for improving the seismic interpretation quality of the research area with few wells.

Spark map reduce based framework for seismic facies classification

Seismic facies analysis provides an efficient way to identify the structure and geology of reservoir units. In the past two decades, seismic attributes have been widely used to highlight the geological features of interest. The combined benefits of multiple seismic attributes have been explored to estimate seismic events accurately. However, with the increase in the number of seismic attributes and size of seismic data, the traditional methods of seismic facies interpretation become very difficult and time-consuming. To address this issue, several computer-assisted facies classification algorithms including k-Means, self-organizing map and generative topographic map were introduced to capture the spatial distribution of seismic facies. The automated facies classification methods accurately quantify the characteristics of geological facies. However, these methods are computationally very time-expensive. To resolve this problem, we provide Spark Map-Reduce based framework for unsupervised classification of seismic facies. The study aims to reduce the computational time complexity of various existing unsupervised facies classification techniques. Further, to demonstrate the applicability, the proposed parallel frameworks are applied to the seismic dataset of Dutch Sector (North Sea) and the classification results are compared in terms of run-time complexity and accuracy.

Fault enhancement using probabilistic neural networks and Laplacian of a Gaussian filter: A case study in the Great South Basin, New Zealand

Fault identification is critical in defining the structural framework for exploration and reservoir characterization studies. Interpreters routinely use edge-sensitive attributes to accelerate the manual picking process, in which the actual choice of a particular edge-sensitive attribute varies with the seismic data quality and with the reflectivity response of the faulted geologic formations. The cyan-magenta-yellow (CMY) color blending provides an effective way to combine the information content of two or three edge-sensitive attributes when more than one attribute is sensitive to faults. We have evaluated whether combining the information content of more than three attributes using probabilistic neural networks (PNNs) provides any additional uplift. We use training data consisting of manually picked faults on a coarse grid of 3D seismic lines, and then, we use an exhaustive search PNN to identify the optimal set of attributes to create a fault probability volume for a 3D survey acquired over the Great South Basin, New Zealand. We construct a suite of candidate attributes using our understanding of the attribute response to faults seen in the data and examples extracted from the published literature to use the list as the analyzed attributes. Using a subset of picked faults as training data, we evaluate which suite of attributes and hyperparameters exhibits the highest validation on the remaining training data. When used together, we find that volume aberrancy magnitude, gray-level cooccurrence matrix (GLCM) homogeneity, GLCM entropy, Sobel filter similarity, and envelope best predict the faults for this data set. The PNN supervised classification creates a seismic image volume that exhibits fault probabilities providing a simple combination of multiple seismic attributes. We also find that applying a directional Laplacian of a Gaussian and skeletonization filters to the PNN fault volumes provides a superior result to simple CMY blending techniques.

Enhanced interpretation of strike-slip faults using hybrid attributes: Advanced insights into fault geometry and relationship with hydrocarbon accumulation in Jurassic formations of the Junggar Basin

Strike-slip faults can play an important role in hydrocarbon accumulation. Therefore advanced studies on the interpretation of intra-basinal strike-slip faults are crucial. Hybrid attributes computed from conditioned 3D seismic data via artificial neural networks may enhance subsurface fault images. Jurassic formations within the C36 3D Prospect located in the Central Depression of the Junggar Basin are important to hydrocarbon accumulation. However, the geometry of strike-slip faults and their relationship with hydrocarbon accumulation are inadequately understood. The present study highlights these strike-slip faults using 3D seismic data from the C36 3D Prospect. Several conditioning approaches have enabled enhanced imaging of the strike-slip faults using computed hybrid attributes. They are based on implementing a dip-steering cube extracted from the original seismic data, along with analyses of multiple seismic attributes and a supervised neural network. Analysis of the hybrid attributes reveals the development of isolated, soft-linked, hard-linked, and coalesced faults along an approximately NE-SW trend. Various types of interactions were observed between the upper and lower fault systems. The different types of vertical linkages between the strike-slip faults provide hydrocarbon migration paths and thus, contribute differently to hydrocarbon migration. • Improved strike-slip fault imaging features from 3D seismic data using hybrid attributes are presented. • The geometry of strike-slip faults is analyzed according to hybrid attributes in the Jurassic formations. • The isolated, soft-linked, hard-linked, and coalesced fault tracesare illustrated, and a simple model is established. • The fault linkage types control the position and number of layers favorable for hydrocarbon accumulation.

A review of seismic attribute taxonomies, discussion of their historical use, and presentation of a seismic attribute communication framework using data analysis concepts

Beginning in the 1970s, seismic attributes have grown from a few simple measurements of wavelet amplitude, frequency, and phase to an expanded attribute toolbox that measures not only wavelet properties but also their context within the 3D seismic volume. When using multiple seismic attributes, the interpreter must understand not only each individual attribute but also the relationships between them. Researchers communicate these relationships via seismic attribute taxonomies, which group attributes by their signal property, mathematical formulation, or their interpretive value. The first attempts to organize seismic attributes began in the 1990s, and with new attributes and their increasing breadth of applications, continues to this day. Most scientific papers that use seismic attributes focus on a specific application, new algorithms, or a novel interpretation workflow, rather than how a specific attribute fits within the greater whole, leading to confusion for the less experienced interpreter. We have analyzed more than 2100 citing works, identified the 231 papers that discuss the taxonomies specifically, and found out how the authors use those citations. The result is a list of more than a dozen seismic attribute classification systems, which we reduce to a smaller subset by including only those that apply to general use. An optimal seismic attribute taxonomy should not only be useful to the interpretation community today, but it should also adapt to the ever-changing needs of the profession, including changes appropriate for their use in modern machine-learning algorithms. The adaptability of prior work to modern workflows remains a shortcoming. However, as we develop our work in two parts — the first covering the evolution of seismic attribute taxonomies and their use through time and the second proposing a new seismic attribute communication framework for the larger community — we link attributes together via data analysis principles and provide an extensible model as the profession and research expand.

A deep-learning method for latent space analysis of multiple seismic attributes

Seismic attributes are a well-established method for highlighting subtle features in seismic data to improve interpretability and suitability for quantitative analysis. Seismic attributes are an enabling technology in such areas as thin-bed analysis, geobody extraction, and seismic geomorphology. Seismic attributes are mathematical functions of the data that are designed to exploit geologic and/or geophysical principles to provide meaningful information about underlying processes. Seismic attributes often suffer from an “abundance of riches” because the high dimensionality of seismic attributes may cause great difficulty in accomplishing even simple tasks. Spectral decomposition, for instance, typically produces tens and sometimes hundreds of attributes. However, when it comes to visualization, for instance, we are limited to visualizing three or at most four attributes simultaneously. We have developed a deep-learning-based approach to latent space analysis. This method is superior to other methods in that it focuses upon capturing essential information rather than just focusing upon probability density functions or clusters. Our method provides a quantitative way to assess the fit of the latent space to the original data. We apply our method to a data set from Canterbury Basin, New Zealand. This data set contains a turbidite system, and it has been the subject of several other papers. We examine the goodness of fit of our model by comparing the input data to what can be reproduced, and we provide an interpretation based upon our method.

Porosity Prediction With Uncertainty Quantification From Multiple Seismic Attributes Using Random Forest

AbstractInferring porosity of subsurface from seismic data is of profound significance to many fields of Earth science and engineering applications, including but not limited to: hydrocarbon reservoir characterization, underground water flow modeling, geological CO2 storage, and geothermal energy exploitation. Traditional model‐driven approaches confront the problems of strong nonlinearity and geological heterogeneity, while machine learning is good at nonlinear mapping, providing higher efficiency and accuracy as well. We propose a Random Forest (RF) based method using multiple seismic attributes to predict the underground porosity distribution with uncertainty quantification. The standard deviation of base models' predictions is used to quantify the regression uncertainty of RF. The uncertainty can robustly indicate the prediction quality in numerous experiments, where low uncertainty corresponds to relatively precise prediction and high uncertainty gives a possibility of larger errors. Furthermore, we utilize the quantified uncertainty to improve the RF regression accuracy by correcting the originally predicted porosity according to the statistical relationship between the absolute error and the standard deviation. The application of the proposed method on seismic data shows its potential to characterize spatially varying reservoir parameters, and the quantified uncertainty profile offers insights into risk evaluation for hydrocarbon exploration and development.