Field Data Application Research Articles

Some results of seismic travel-time reflection tomography study

Velocity model is essential for seismic data processing as it plays an important role in migration processes as well as time depth conversion. There are several techniques to reach that goal, among which tomographic inversion is an efficient one. As an upgrade version of handpicked velocity analysis, the tomography technique is based on the reflection ray tracing and conjugate gradient method to estimate an optimum velocity model and can create an initial high quality model for other intensive imaging and modelling module such as reverse-time migration (RTM) and full-waveform inversion (FWI). For the mentioned benefit, we develop a seismic travel-time reflection tomography (SeisT) module to study the accuracy of the approach along with building the technical capability in seismic processing. The accuracy of the module has been tested by both synthetic and real seismic field data; the efficiency and the accuracy of the model have been proven in terms of development method as well as field data application.

Receiver-extension strategy for time-domain full-waveform inversion using a relocalization approach

A receiver-extension strategy is presented as an alternative to recently promoted source-extension strategies, in the framework of high-resolution seismic imaging by full-waveform inversion (FWI). This receiver-extension strategy is directly applicable in time-domain FWI, and, unlike source-extension methods, it incurs negligible extra computational cost. After connections between different source-extension strategies are reviewed, the receiver-extension method is introduced and analyzed for single-arrival data. The method results in a misfit function convex with respect to the velocity model in this context. The method is then applied to three exploration-scale synthetic case studies representative of different geologic environments, based on the Marmousi model, the BP 2004 salt model, and the Valhall model. In all three cases, the receiver-extension strategy makes it possible to start FWI with crude initial models and to reconstruct meaningful subsurface velocity models. The good performance of the method, even considering the inaccurate amplitude prediction due to noise, imperfect modeling, and source wavelet estimation, bodes well for field data applications.

Amplitude-variation-with-offset inversion based on group sparse regularization

Seismic amplitude-variation-with-offset (AVO) inversion from prestack seismic data plays a significant role in estimating elastic parameters and characterizing reservoir properties. In general, sparse regularization is widely used to solve ill-posed inverse problems by reducing the solution space of subsurface parameters, which makes seismic AVO inversion more stable. However, the traditional sparse constraint inversion only focuses on the vector sparsity of reflectivity, instead of the structural sparse characteristics of the estimated parameters. Consequently, various elastic parameters demonstrate different formation structural features in the same location of stratum. In this study, we have developed a novel approach that combines the structural sparsity and the vector sparsity of the model reflectivity to establish the posterior probability density distribution and solve the objective function of the model parameters. Based on the relationship among multiple elastic parameters, we divide the model parameters to be inverted into several groups according to intrinsic structural sparse characteristics of elastic parameters. In this case, all of the model parameters at the same sampling point are classified into the identical group, which ensures that different estimated parameters indicate the same characteristic in terms of stratigraphic structure. From the perspective of Bayesian inference, we use the modified Cauchy probability density function (PDF) to characterize the group sparsity and describe the relationship among model parameters in the same group by Gaussian PDF. Furthermore, we estimate the optimum solution corresponding to the maximum a posteriori probability under Bayesian inference. Synthetic experiments on a Marmousi model prove that the estimated P-velocity, S-velocity, and density are consistent with those of the real models, and the application of field data confirms the availability and feasibility of group sparse inversion.

Simple design to estimate time-lapse microgravity response due to shallow subsurface density redistribution caused by land subsidence

A simple design for modeling shallow subsurface density redistribution due to land subsidence is designed to obtain the time-lapse microgravity response. The subsurface model at each point of gravity observation is represented by a rectangular prism. A numerical example of computational modeling is performed to estimate the effect of land subsidence to the data of a time-lapse microgravity. Simple numerical simulations with an initial model that have flat topography, homogeneous density, and homogeneous compaction thickness are carried out in variations of geological and hydrological information that are often found in a study area. Additional algorithms to accommodate information on topographic variations, density variations, and compaction thickness variations in the horizontal direction also shown with illustration. Field data application for this study utilize rough estimation of the geology and the land subsidence rate in Bandung Basin. The estimation results with numerical simulations give time-lapse microgravity anomaly 0.78 to 28.61 μGal/m and field data application give an anomaly up to 10 μGal.

Development of a National Integrated Agricultural Field Data Application Programming Interface (API) that Combines Field Polygons, Agricultural Field Pins, and Soil Map Data

Development of a National Integrated Agricultural Field Data Application Programming Interface (API) that Combines Field Polygons, Agricultural Field Pins, and Soil Map Data

Learning from unlabelled real seismic data: Fault detection based on transfer learning

ABSTRACTSignificant advances have been made towards fault detection using deep learning. However, the fault labelling of seismic data requires great human effort. The resulting small sample problem makes traditional deep learning methods difficult to achieve desired results. Existing research proposes to train a deep learning model with labelled synthetic seismic data to get good fault detection results. However, due to the complexity of the actual geological situation, there are inevitable differences between synthetic seismic data and real seismic data in many aspects such as seismic signal frequency, frequency of fault distribution and degree of noise disturbance, which lead to the fact that the deep learning model trained by synthetic seismic data is difficult to get good fault detection result in field data applications. We propose to use transfer learning to reduce the impact of data differences to solve this problem: part of the deep transfer learning model is used to learn fault‐related features. And the other part of the deep transfer learning model is used to mine common features between the real seismic data and the synthetic seismic data, which makes the deep transfer learning model more suitable for real seismic data. Compared with the latest research progress, our method can greatly improve the effect of fault detection without real data label, which can significantly save the cost of manual label processing.

A hybrid residual neural network–Monte Carlo approach to invert surface wave dispersion data

ABSTRACTSurface‐wave inversion is a non‐linear and ill‐conditioned problem usually solved through deterministic or global optimization approaches. Here, we present an alternative method based on machine learning. Under the assumption of a local one‐dimensional model, we train a residual neural network to predict the non‐linear mapping between the full dispersion image and the model space, parameterized in terms of shear wave velocity and layer thicknesses. On the one hand, compared to standard convolutional neural networks, the residual network prevents the vanishing gradient problem when training a deep network. On the other hand, the use of the full dispersion image avoids the time‐consuming and often ambiguous picking procedure and allows considering higher modes in the inversion framework. One key aspect of any machine learning inversion strategy is the definition of an appropriate training set. In this case, the models forming the training and validation examples are uniformly drawn from previously defined ranges that cover a wide range of possible near‐surface layered Vs models. The reflectivity method constitutes the forward modelling operator that converts the model parameters into the observed shot gathers. The inversion also includes a Monte Carlo simulation strategy that propagates onto the model space the uncertainties related to noise in the data and the modelling error introduced by the network approximation. We first discuss synthetic inversions to assess the applicability of the proposed method and to analyse the effect of erroneous model parameterizations. The inversion results are also benchmarked with those provided by a more standard approach in which the particle swarm optimization algorithm inverts the fundamental mode only. Then, we discuss a field data application. Our tests confirm that the residual neural network inversion provides accurate model estimations and reliable uncertainty appraisals. One of the main benefits of the proposed approach is that once the network is trained it provides the near‐surface shear wave velocity profile in near real‐time.

Seismic noise attenuation by signal reconstruction: an unsupervised machine learning approach

ABSTRACTRandom noise attenuation is an essential step in seismic data processing for improving seismic data quality and signal‐to‐noise ratio. We adopt an unsupervised machine learning approach to attenuate random noise via signal reconstruction strategy. This approach can be accomplished in the following steps: Firstly, we randomly mute a part of the input data of the neural network according to a certain percentage, and then the network outputs the reconstructed data influenced by this randomly mute. The objective function measures the distance between the input data and the reconstructed data. Secondly, we use the adaptive moment estimation algorithm to minimize the distance, and the network adjusts its internal parameters so that sparse representations can be captured by the multiple processing layers of the neural network. Finally, we take the same proportion of random mute on the raw seismic data which are fed to the trained neural network. Through this network, reconstruction of seismic data and attenuation of random noise are completed simultaneously. We use both synthetic and field data to testify the feasibility and applicability of the proposed method. Synthetic data experiment indicates that the proposed method achieves better denoised results than the conventional methods. Field data applications further demonstrate its superiority and practicality.

Pre-stack Q estimation based on inverse spectral decomposition

The quality factor can quantitatively measure the attenuation of seismic waves, and it can compensate for seismic wave amplitude attenuation and correct the phase distortion. The traditional Q extraction method uses the post-stack seismic data to extract the Q-factor. Due to the low-pass filtering effect of the stacked seismic data, the obtained quality factor may be inaccurate. A method of estimating the quality factor using pre-stack common midpoint (CMP) gather data is proposed in this paper. In this method, inversion spectrum decomposition (ISD) and shaping regularization are introduced to obtain the Q-factor. During the process, we obtained the average-Q through inversion and finally obtained the internal-Q by using the average-Q. The numerical simulation and field data application show that the Q obtained using this method can accurately identify the oil-bearing area. Then, the Q is applied to the inverse Q-filter and produces seismic data with a high resolution and a high signal-to-noise ratio.

The Slope-Attribute-Regularized High-Resolution Prestack Seismic Inversion

Prestack seismic inversion can be regarded as an optimization problem, which minimizes the error between the observed and synthetic data under the premise of certain geological/geophysical a priori information constraints. It has been proved to be a powerful approach for reconstructing the subsurface properties and building the elastic parameter models (e.g., P- and S-wave velocity, and density). With respect to the specific expressions of a priori information, the starting model and regularization are expected to be the most widely used and indispensable constraints to reconstruct structural features and subsurface properties. The conventional prestack inversion (trace-by-trace) methods perform well when the geological structure of the target area is not too complex. However, due to the lack of lateral constraint, such trace-independent methods are inevitably limited by their capability of characterization (including accuracy, resolution, and robustness) in the case of geologically complex structures, such as tilted stratum and steep faults. The geological structure-guided constraint, herein referred to as the seismic slope attribute, can be exploited as a lateral constraint integrated into the prestack inversion algorithm. In this work, the seismic slope attribute is introduced to the amplitude variation with offset/angle inversion from two aspects, i.e., starting model building and regularization penalty. Firstly, using the seismic slope attribute, instead of the traditional manual interpreted geological horizons, as a constraint, the well-log data are interpolated to build the initial model. The interpolation algorithm is formulated as solving the inverse problem by using the shaping regularization method rather than the kriging-based algorithm. Secondly, by rotating the coordinate system according to the seismic slope attribute, the directional total variation regularization is used as a constraint to improve the resolution (in both vertical and horizontal directions) and lateral continuity of the inversion results. Finally, the proposed methods are applied to synthetic and real seismic data. Synthetic tests and field data applications demonstrate that the proposed method is capable of revealing complex structural features and achieving stabilized inversion of multi-parameters with less uncertainty.

Investigation of rock properties distribution effect on pressure transient analysis of naturally fractured reservoirs

Naturally fractured reservoirs (NFRs) are considered as one of the main resources of oil and gas around the world. The nature of fractures in this kind of reservoirs splits these reservoir rocks into two parts; matrix and fracture. In these two media, geometrical properties like matrix block size and also petrophysical properties such as porosity and permeability cannot considered to be constant and they usually contain heterogeneity in extent of the reservoir. In this study, a statistical distribution concept is used to be able to model variance of properties including matrix block size, matrix and fracture permeability, and matrix and fracture porosity as independent statistical variables. For the purpose of investigating effect of the five aforementioned properties distribution on pressure transient analysis (PTA) plots, five distinct cases of property distributions are considered with their corresponding precise statistical modeling applied on governing flow equation of a dual-porosity base case reservoir model. Abovementioned distributed parameters are included in semi-analytically derived dual porosity reservoir models and results are compared to their corresponding constant-value response. In each case the constant parameter is the median of corresponding statistical distribution. The statistical approach is validated using numerical modeling and field data applications are also presented. Different unimodal and bimodal matrix properties (block size, permeability, and porosity) distribution have slight to intense effects on PTA plots of an NFR. In matrix block size and porosity distribution cases, lower values, and in matrix permeability distribution case, higher values of the range dominate the pressure behavior of the fractured system. On the other hand, although fracture permeability and porosity distribution can shift PTA plots of the model but the effect is not significant and both fracture permeability and porosity distribution can be represented by constant representative values instead of a statistical model. It is worth mentioning that all matrix properties bimodal distribution has a common signature on PTA plots which is similar to the triple porosity reservoir response (i.e. the presence of double-bump descent on pressure derivative plot).

Visco-acoustic full waveform inversion using decoupled fractional Laplacian constant-Q wave equation and optimal transport-based misfit function

Full waveform inversion (FWI) has the potential to recover high-resolution physical parameters of the Earth from seismic data. Considering the attenuation effects of the subsurface, we develop a new visco-acoustic FWI method in the time domain with a known Q model. The seismic wavefield is modelled by solving a decoupled fractional Laplacian visco-acoustic wave equation, which can better describe amplitude attenuation and phase distortion during wave propagation. To weaken the initial model dependence of visco-acoustic FWI, we introduce the optimal transport-based misfit functional which can measure the data residual globally. Numerical examples on the 2D Marmousi model demonstrate the superiority of the proposed method compared with the L2 norm-based method which measures the misfit of observed and synthetic seismic data locally, generating cycle-skipping issue. The inversion results of seismic data without low-frequency components further demonstrate the validity and effectiveness of the proposed method, which is insensitive towards low-frequency components, and may achieve reasonable inversion results for field data application.

Nonstationary seismic inversion: joint estimation for acoustic impedance, attenuation factor and source wavelet

Seismic signal can be expressed by nonstationary convolution model (NCM) which integrates acoustic impedance (AI), attenuation factor (AF) and source wavelet (SW) into a single formula. Although it provides attractive potential to invert AI, AF and SW, simultaneously, effective joint inversion algorithm has not been developed because of the extreme instability of this nonlinear inverse problem. In this paper, we propose an alternating optimization scheme to achieve this nonlinear joint inversion. Our algorithm repeatedly alternates among three subproblems corresponding to AI, AF and SW recovery until changes in inverted models become smaller than the user-defined tolerances. Also, when we optimize one parameter, other two parameters are fixed. NCM is an explicit linear formula for AI; therefore, AI recovery is accomplished by linear inversion which is regularized by low-frequency model and isotropy total variation domain sparse constraints. However, NCM is a complicated nonlinear formula for AF. To facilitate the AF inversion, we propose a centroid frequency-based attenuation tomography method whose forward operator and observations are acquired from the time-varying wavelet amplitude spectra which is estimated according to Gabor domain factorization of NCM. SW is decoupled from NCM based on Toeplitz structure constraint, and we obtain an orthogonal wavelet transform domain sparse regularized SW inverse subproblem. Split Bregman technique is adopted to optimize AI and SW inverse subproblems. Numerical test and field data application confirm that the proposed nonstationary seismic inversion algorithm can stably generate accurate estimates of AI, AF and SW, simultaneously.

Multi-trace acoustic impedance inversion with multiplicative regularization

Acoustic impedance (AI) inversion is one of the important workflows in petroleum seismology. Many kinds of additive regularization techniques have been proposed to obtain impedance profiles with a relatively high resolution or to take the lateral information of the model into account. However, the determination of a proper regularization parameter is time-consuming. In this paper, we introduce the weighted L2-norm anisotropic total-variation multiplicative regularization to solve the multi-trace AI inversion problem. This regularization scheme has two main characteristics: adaptive adjustment of the regularization parameter, vertical resolution enhancement and anti-noise ability. Besides, space directional difference is naturally integrated to the regularization operator, by which the lateral connection is taken into consideration. To solve the large scale non-quadratic functional efficiently, the limited-memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) direction is used to update the acoustic impedance, and update step is calculated analytically in each iteration. Well log synthetic test demonstrates that the single trace multiplicative regularized AI inversion performs well. Further, a test on a classical wedge model with thin interbeds shows the superiority of the method. A complex two-dimensional test and field data application illustrate the effectiveness of the multi trace multiplicative regularized AI inversion scheme.

Joint impedance inversion and spectral decomposition for deepwater gas reservoir characterization: A case study in South China Sea

Gas reservoir characterization is one of the frontiers in seismic exploration. Acoustic impedance, one of the most effective seismic attributes, aims to describe the spatial distribution of rock properties. However, using acoustic impedance only is insufficient to describe gas-bearing layers accurately, in the case of rapid lithologic changes and complex geology in the deepwater area. The reflection seismograms indicate an absorption effect when seismic waves propagate through gas-bearing layers. The absorption effect can be used as an additional attribute to help gas reservoir characterization. Therefore, we have developed a new attribute for gas reservoir characterization in this study, which integrates the results of acoustic impedance and absorption coefficient. We estimate the acoustic impedance model by using poststack impedance inversion and then we calculate probability distribution functions. Functions are classified into gas-bearing and nongas layers. We discuss an absorption coefficient and obtain it from the spectrum gradient, in which the gradient is calculated by spectral decomposition using the matching pursuit method. We apply the new attribute to characterize the spatial distribution and thickness of deepwater gas reservoirs in the Pearl River Mouth Basin. Well-log and geologic information indicate that the study area has an enrichment of gas reservoirs. Field data application indicates the explicit distribution of the gas reservoir and in accordance with the well-log information, which indicates that our attribute can improve gas reservoir characterization.

Denoising Seismic Signal via Resampling Local Applicability Functions

We propose a novel seismic signal processing approach to efficiently and effectively attenuate seismic random noises. The proposed approach is a generalized seismic noise attenuation solution that can be applied to typical denoising operators. Our work has two main contributions. First, conventional filtering operators “regularize” the denoised results through the operator design. However, as seismic data have strong nonstationarity, it is inevitable to remove certain signal components. The resampling mechanism alleviates the signal loss. Second, the resampling operation does not require a lot of parameter tuning, which improves the denoising efficiency. Using the proposed approach, compared with existing denoising operators, the intrinsic seismic signal components are better recovered since random noise has been suppressed. Synthetic example and field data applications quantitatively and qualitatively demonstrate excellent performances of the proposed approach.

Data-driven pre-stack AVO inversion method based on fast orthogonal dictionary

Regularization techniques are commonly used to mitigate the ill-posed problems for pre-stack AVO (amplitude variation with offset) inversion. In recent years, a data-driven regularization method has successfully been applied to achieve desirable inversion accuracy. In contrast of assuming a certain distribution of the data, this method uses the dictionary learning algorithm to learn the structural characteristics of elastic parameters and then takes this information as a prior constraint in AVO inversion. However, in this method, KSVD (An Algorithm for Designing Overcomplete Dictionaries for Sparse Representation) algorithm is used, which is of low computational efficiency and involves many parameters. In this paper, we propose a novel data-driven pre-stack inversion approach based on the orthogonal dictionary (ORTD) learning method. Because the atoms are orthogonal, the efficiency of dictionary operation is nearly 100 times higher than that of KSVD redundant dictionary. Instead of regulating the sparsity, atomic size and atomic length for the KSVD method, the proposed method depends on two parameters: threshold parameters and dictionary atomic length. Moreover, the parameter manipulation has little impacts on the operation efficiency for this approach. This greatly improves the applicability of the data-driven inversion method. We use model tests and field data applications to verify the performance of the proposed method.

Weak signal enhancement using adaptive local similarity and neighboring super-virtual trace for first arrival picking

Abstract Accurate traveltime of first arrivals is of great importance in investigating subsurface velocity information. A significant challenge preventing the picking of the first arrival, however, is that the recorded traces in complex mountain areas are often characterised by weak energy, strong noise and dramatic phase variation. The method of super-virtual refraction interferometry (SVI) is capable of retrieving and enhancing the weak first arrivals from those traces and attenuating the random noise. Unfortunately, the conventional SVI has equal-weighted stacking, and is susceptible to strong local noise. This paper introduces adaptive data-driven weights based on local similarity into SVI to solve this problem. Both near- and far-offset reference traces of high quality are technically selected for better preservation of useful information. Next, we develop some neighboring super-virtual traces in the stacking process for further enhancement of weak signals, which is a further extension and theoretically superior to conventional SVI in increasing the total stacking number. The successful applications of model and field data show the great advantages of our improved method. Compared with conventional SVI, our method has a better local noise suppression effect and stronger enhancement ability, especially at weak refractions. More importantly, it can provide a significant guarantee of higher quality data, thus distinctly achieving a more accurate and reliable traveltime in first arrival picking.

Frequency-dependent AVAZ for fractured reservoirs

Frequency-dependent anisotropy can be used for detection and evaluation of fractured reservoirs. In this study, three numerical models of fractured reservoirs are designed to study the frequency dependence of amplitude variation with azimuth (AVAZ). Here, two factors, which can make the AVAZ responses frequency dependent, are taken into consideration, the fluid flow and the tuning effect. Model I is used for the analysis of frequency-dependent AVAZ induced only by the fluid flow. Model II is used to study frequency-dependent AVAZ induced only by the tuning effect. For Model III, these two factors are in the presence at the same time. In this study, the fluid flow is simulated by Chapman’s multi-scale rock physics model, and the azimuthal seismic gathers are generated by propagating matrix method. Spectral decomposition is implemented based on the smoothed pseudo Wigner-Ville distribution, so that the frequency dependence for the AVAZ gathers can be discussed. It is demonstrated that both the fluid flow and the tuning effect can cause significant frequency-dependent AVAZ, and the frequency-dependent AVAZ is helpful for fluid identification and anisotropic thin bed detection. Besides, for both factors, the incident angle can influence the frequency-dependent AVAZ responses, accordingly azimuthal seismic gathers with large incident angles are necessary for field data application.

Improving the resolution of impedance inversion in karst systems by incorporating diffraction information: A case study of Tarim Basin, China

Different scales of voids and cavities in karst systems demonstrate considerable potential as exploration targets in the Tarim Basin, northwest China. Numerous diffraction events exist in the seismic data above the karst reservoir in this area because of the strong impedance contrast and irregular shape of voids. The conventional impedance inversion method using migrated data as an input cannot easily estimate the location of voids and the impedance inside the voids. In this case, an alternative approach to impedance inversion that considers the diffractive component of the total wavefield and uses the unmigrated data as an input should be used. The inversion consists of a least-squares minimization of the misfit function between the observed and modeled data. Forward modeling incorporates a combination of reflection and diffraction wavefield components. The adopted method is applied to physical modeling and field data recorded above the karst reservoir. This study’s physical modeling test simulates observed field data and is performed using a high-resolution and high-fidelity 3D modeling system. Inversion results obtained by the proposed and conventional methods are compared. Physical modeling and a field data application show that the adopted impedance inversion method improves the karst location estimation and the acoustic impedance within the voids.