Stochastic seismic inversion and Bayesian facies classification applied to porosity modeling and igneous rock identification

We apply stochastic seismic inversion and Bayesian facies classification for porosity modeling and igneous rock identification in the presalt interval of the Santos Basin. This integration of seismic and well-derived information enhances reservoir characterization. Stochastic inversion and Bayesian classification are powerful tools because they permit addressing the uncertainties in the model. We used the ES-MDA algorithm to achieve the realizations equivalent to the percentiles P10, P50, and P90 of acoustic impedance, a novel method for acoustic inversion in presalt. The facies were divided into five: reservoir 1,reservoir 2, tight carbonates, clayey rocks, and igneous rocks. To deal with the overlaps in acoustic impedance values of facies, we included geological information using a priori probability, indicating that structural highs are reservoir-dominated. To illustrate our approach, we conducted porosity modeling using facies-related rock-physics models for rock-physics inversion in an area with a well drilled in a coquina bank and evaluated the thickness and extension of an igneous intrusion near the carbonate-salt interface. The modeled porosity and the classified seismic facies are in good agreement with the ones observed in the wells. Notably, the coquinas bank presents an improvement in the porosity towards the top. The a priori probability model was crucial for limiting the clayey rocks to the structural lows. In Well B, the hit rate of the igneous rock in the three scenarios is higher than 60%, showing an excellent thickness-prediction capability.

Nonlinear inversion of electrical resistivity imaging using pruning Bayesian neural networks

Conventional artificial neural networks used to solve electrical resistivity imaging （ERI） inversion problem suffer from overfitting and local minima. To solve these problems, we propose to use a pruning Bayesian neural network （PBNN） nonlinear inversion method and a sample design method based on the K-medoids clustering algorithm. In the sample design method, the training samples of the neural network are designed according to the prior information provided by the K-medoids clustering results; thus, the training process of the neural network is well guided. The proposed PBNN, based on Bayesian regularization, is used to select the hidden layer structure by assessing the effect of each hidden neuron to the inversion results. Then, the hyperparameter αk, which is based on the generalized mean, is chosen to guide the pruning process according to the prior distribution of the training samples under the small-sample condition. The proposed algorithm is more efficient than other common adaptive regularization methods in geophysics. The inversion of synthetic data and field data suggests that the proposed method suppresses the noise in the neural network training stage and enhances the generalization. The inversion results with the proposed method are better than those of the BPNN, RBFNN, and RRBFNN inversion methods as well as the conventional least squares inversion.

Bayesian prestack seismic inversion with a self-adaptive Huber-Markov random-field edge protection scheme

Seismic inversion is a highly ill-posed problem, due to many factors such as the limited seismic frequency bandwidth and inappropriate forward modeling. To obtain a unique solution, some smoothing constraints, e.g., the Tikhonov regularization are usually applied. The Tikhonov method can maintain a global smooth solution, but cause a fuzzy structure edge. In this paper we use Huber-Markov random-field edge protection method in the procedure of inverting three parameters, P-velocity, S-velocity and density. The method can avoid blurring the structure edge and resist noise. For the parameter to be inverted, the Huber- Markov random-field constructs a neighborhood system, which further acts as the vertical and lateral constraints. We use a quadratic Huber edge penalty function within the layer to suppress noise and a linear one on the edges to avoid a fuzzy result. The effectiveness of our method is proved by inverting the synthetic data without and with noises. The relationship between the adopted constraints and the inversion results is analyzed as well.

Nonlinear joint PP-PS AVO inversion based on improved Bayesian inference and LSSVM

Multiwave seismic technology promotes the application of joint PP–PS amplitude versus offset (AVO) inversion;however conventional joint PP–PS AVO inversioan is linear based on approximations of the Zoeppritz equations for multiple iterations. Therefore the inversion results of P-wave, S-wave velocity and density exhibit low precision in the faroffset;thus, the joint PP–PS AVO inversion is nonlinear. Herein, we propose a nonlinear joint inversion method based on exact Zoeppritz equations that combines improved Bayesian inference and a least squares support vector machine (LSSVM) to solve the nonlinear inversion problem. The initial parameters of Bayesian inference are optimized via particle swarm optimization (PSO). In improved Bayesian inference, the optimal parameter of the LSSVM is obtained by maximizing the posterior probability of the hyperparameters, thus improving the learning and generalization abilities of LSSVM. Then, an optimal nonlinear LSSVM model that defi nes the relationship between seismic refl ection amplitude and elastic parameters is established to improve the precision of the joint PP–PS AVO inversion. Further, the nonlinear problem of joint inversion can be solved through a single training of the nonlinear inversion model. The results of the synthetic data suggest that the precision of the estimated parameters is higher than that obtained via Bayesian linear inversion with PP-wave data and via approximations of the Zoeppritz equations. In addition, results using synthetic data with added noise show that the proposed method has superior anti-noising properties. Real-world application shows the feasibility and superiority of the proposed method, as compared with Bayesian linear inversion.

Bayesian seismic multi-scale inversion in complex Laplace mixed domains

Seismic inversion performed in the time or frequency domain cannot always recover the long-wavelength background of subsurface parameters due to the lack of low-frequency seismic records. Since the low-frequency response becomes much richer in the Laplace mixed domains, one novel Bayesian impedance inversion approach in the complex Laplace mixed domains is established in this study to solve the model dependency problem. The derivation of a Laplace mixed-domain formula of the Robinson convolution is the first step in our work. With this formula, the Laplace seismic spectrum, the wavelet spectrum and time-domain reflectivity are joined together. Next, to improve inversion stability, the object inversion function accompanied by the initial constraint of the linear increment model is launched under a Bayesian framework. The likelihood function and prior probability distribution can be combined together by Bayesian formula to calculate the posterior probability distribution of subsurface parameters. By achieving the optimal solution corresponding to maximum posterior probability distribution, the low-frequency background of subsurface parameters can be obtained successfully. Then, with the regularization constraint of estimated low frequency in the Laplace mixed domains, multi-scale Bayesian inversion inthe pure frequency domain is exploited to obtain the absolute model parameters. The effectiveness, anti-noise capability and lateral continuity of Laplace mixed-domain inversion are illustrated by synthetic tests. Furthermore,one field case in the east of China is discussed carefully with different input frequency components and different inversion algorithms. This provides adequate proof to illustrate the reliability improvement in low-frequency estimation and resolution enhancement of subsurface parameters, in comparison with conventional Bayesian inversion in the frequency domain.

Pre-stack inversion for caved carbonate reservoir prediction:A case study from Tarim Basin,China

The major storage space types in the carbonate reservoir in the Ordovician in the TZ45 area are secondary dissolution caves.For the prediction of caved carbonate reservoir,post-stack methods are commonly used in the oilfield at present since pre-stack inversion is always limited by poor seismic data quality and insufficient logging data.In this paper,based on amplitude preserved seismic data processing and rock-physics analysis,pre-stack inversion is employed to predict the caved carbonate reservoir in TZ45 area by seriously controlling the quality of inversion procedures.These procedures mainly include angle-gather conversion,partial stack,wavelet estimation,low-frequency model building and inversion residual analysis.The amplitude-preserved data processing method can achieve high quality data based on the principle that they are very consistent with the synthetics.Besides,the foundation of pre-stack inversion and reservoir prediction criterion can be established by the connection between reservoir property and seismic reflection through rock-physics analysis.Finally,the inversion result is consistent with drilling wells in most cases.It is concluded that integrated with amplitude-preserved processing and rock-physics,pre-stack inversion can be effectively applied in the caved carbonate reservoir prediction.

Bayesian Markov chain Monte Carlo inversion for anisotropy of PP-and PS-wave in weakly anisotropic and heterogeneous media

A single set of vertically aligned cracks embedded in a purely isotropic background may be con- sidered as a long-wavelength effective transversely iso- tropy （HTI） medium with a horizontal symmetry axis. The crack-induced HTI anisotropy can be characterized by the weakly anisotropic parameters introduced by Thomsen. The seismic scattering theory can be utilized for the inversion for the anisotropic parameters in weakly aniso- tropic and heterogeneous HTI media. Based on the seismic scattering theory, we first derived the linearized PP- and PS-wave reflection coefficients in terms of P- and S-wave impedances, density as well as three anisotropic parameters in HTI media. Then, we proposed a novel Bayesian Mar- kov chain Monte Carlo inversion method of PP- and PS- wave for six elastic and anisotropic parameters directly. Tests on synthetic azimuthal seismic data contaminated by random errors demonstrated that this method appears more accurate, anti-noise and stable owing to the usage of the constrained PS-wave compared with the standards inver- sion scheme taking only the PP-wave into account.

Joint AVO inversion in the time and frequency domain with Bayesian interference

Amplitude variations with offset or incident angle （AVO/AVA） inversion are typically combined with statistical methods, such as Bayesian inference or deterministic inversion. We propose a joint elastic inversion method in the time and frequency domain based on Bayesian inversion theory to improve the resolution of the estimated P- and S-wave velocities and density. We initially construct the objective function using Bayesian inference by combining seismic data in the time and frequency domain. We use Cauchy and Gaussian probability distribution density functions to obtain the prior information for the model parameters and the likelihood function, respectively. We estimate the elastic parameters by solving the initial objective function with added model constraints to improve the inversion robustness. The results of the synthetic data suggest that the frequency spectra of the estimated parameters are wider than those obtained with conventional elastic inversion in the time domain. In addition, the proposed inversion approach offers stronger antinoising compared to the inversion approach in the frequency domain. Furthermore, results from synthetic examples with added Gaussian noise demonstrate the robustness of the proposed approach. From the real data, we infer that more model parameter details can be reproduced with the proposed joint elastic inversion.

Elastic modulus extraction based on generalized pre-stack PP–PS wave joint linear inversion

Joint PP–PS inversion offers better accuracy and resolution than conventional P-wave inversion. P-and S-wave elastic moduli determined through data inversions are key parameters for reservoir evaluation and fluid characterization. In this paper, starting with the exact Zoeppritz equation that relates P-and S-wave moduli, a coefficient that describes the reflections of P-and converted waves is established. This method effectively avoids error introduced by approximations or indirect calculations, thus improving the accuracy of the inversion results. Considering that the inversion problem is ill-posed and that the forward operator is nonlinear, prior constraints on the model parameters and modified low-frequency constraints are also introduced to the objective function to make the problem more tractable. This modified objective function is solved over many iterations to continuously optimize the background values of the velocity ratio, which increases the stability of the inversion process. Tests of various models show that the method effectively improves the accuracy and stability of extracting P and S-wave moduli from underdetermined data. This method can be applied to provide inferences for reservoir exploration and fluid extraction.

Impedance inversion of pre-stack seismic data in the depth domain

The extensive application of pre-stack depth migration has produced huge volumes of seismic data,which allows for the possibility of developing seismic inversions of reservoir properties from seismic data in the depth domain.It is difficult to estimate seismic wavelets directly from seismic data due to the nonstationarity of the data in the depth domain.We conduct a velocity transformation of seismic data to make the seismic data stationary and then apply the ridge regression method to estimate a constant seismic wavelet.The estimated constant seismic wavelet is constructed as a set of space-variant seismic wavelets dominated by velocities at different spatial locations.Incorporating the weighted superposition principle,a synthetic seismogram is generated by directly employing the space-variant seismic wavelets in the depth domain.An inversion workflow based on the model-driven method is developed in the depth domain by incorporating the nonlinear conjugate gradient algorithm,which avoids additional data conversions between the time and depth domains.The impedance inversions of the synthetic and field seismic data in the depth domain show good results,which demonstrates that seismic inversion in the depth domain is feasible.The approach provides an alternative for forward numerical analyses and elastic property inversions of depth-domain seismic data.It is advantageous for further studies concerning the stability,accuracy,and efficiency of seismic inversions in the depth domain.

Fransdimensional Bayesian inversion of timedomain airborne EM data

To reduce the dependence of EM inversion on the choice of initial model and to obtain the global minimum, we apply transdimensional Bayesian inversion to time-domain airborne electromagnetic data. The transdimensional Bayesian inversion uses the Monte Carlo method to search the model space and yields models that simultaneously satisfy the acceptance probability and data fitting requirements. Finally, we obtain the probability distribution and uncertainty of the model parameters as well as the maximum probability. Because it is difficult to know the height of the transmitting source during flight, we consider a fixed and a variable flight height. Furthermore, we introduce weights into the prior probability density function of the resistivity and adjust the constraint strength in the inversion model by changing the weighing coefficients. This effectively solves the problem of unsatisfactory inversion results in the middle high-resistivity layer. We validate the proposed method by inverting synthetic data with 3% Gaussian noise and field survey data.

Bayesian Rayleigh wave inversion with an unknown number of layers

Surface wave methods have received much attention due to their efficient, flexible and convenient characteristics. However, there are still critical issues regarding a key step in surface wave inversion. In most existing methods, the number of layers is assumed to be known prior to the process of inversion. However, improper assignment of this parameter leads to erroneous inversion results. A Bayesian nonparametric method for Rayleigh wave inversion is proposed herein to address this problem. In this method, each model class represents a particular number of layers with unknown S-wave velocity and thickness of each layer. As a result, determination of the number of layers is equivalent to selection of the most applicable model class. Regarding each model class, the optimization search of S-wave velocity and thickness of each layer is implemented by using a genetic algorithm. Then, each model class is assessed in view of its efficiency under the Bayesian framework and the most efficient class is selected. Simulated and actual examples verify that the proposed Bayesian nonparametric approach is reliable and efficient for Rayleigh wave inversion, especially for its capability to determine the number of layers.

Fast pre-stack multi-channel inversion constrained by seismic reflection features

Classical multi-channel technology can significantly reduce the pre-stack seismic inversion uncertainty, especially for complex geology such as high dipping structures. However, due to the consideration of complex structure or reflection features, the existing multi-channel inversion methods have to adopt the highly time-consuming strategy of arranging seismic data trace-by-trace, limiting its wide application in pre-stack inversion. A fast pre-stack multi-channel inversion constrained by seismic reflection features has been proposed to address this issue. The key to our method is to re-characterize the reflection features to directly constrain the pre-stack inversion through a Hadamard product operator without rearranging the seismic data. The seismic reflection features can reflect the distribution of the stratum reflection interface, and we obtained them from the post-stack profile by searching the shortest local Euclidean distance between adjacent seismic traces. Instead of directly constructing a large-size reflection features constraint operator advocated by the conventional methods, through decomposing the reflection features along the vertical and horizontal direction at a particular sampling point, we have constructed a computationally well-behaved constraint operator represented by the vertical and horizontal partial derivatives. Based on the Alternating Direction Method of Multipliers (ADMM) optimization, we have derived a fast algorithm for solving the objective function, including Hadamard product operators. Compared with the conventional reflection features constrained inversion, the proposed method is more efficient and accurate, proved on the Overthrust model and a field data set.

A Bayesian Inference Approach to Reduce Uncertainty in Magnetotelluric Inversion: A Synthetic Case Study

The deterministic geophysical inversion methods are dominant when inverting magnetotelluric data whereby its results largely depends on the assumed initial model and only a single representative solution is obtained. A common problem to this approach is that all inversion techniques suffer from non-uniqueness since all model solutions are subjected to errors, under-determination and uncertainty. A statistical approach in nature is a possible solution to this problem as it can provide extensive information about unknown parameters. In this paper, we developed a 1D Bayesian inversion code based Metropolis-Hastings algorithm whereby the uncertainty of the earth model parameters were quantified by examining the posterior model distribution. As a test, we applied the inversion algorithm to synthetic model data obtained from available literature based on a three layer model (K, H, A and Q). The frequency for the magnetotelluric impedance data was generated from 0.01 to 100 Hz. A 5% Gaussian noise was added at each frequency in order to simulate errors to the synthetic results. The developed algorithm has been successfully applied to all types of models and results obtained have demonstrated a good compatibility with the initial synthetic model data.

Pre-stack AVO inversion with adaptive edge preserving smooth filter regularization based on Aki-Richard approximation

With the development of exploration of oil and gas resources,the requirements for seismic inversion results are getting more accurate.In particular,it is hoped that the distribution patterns of oil and gas reservoirs can be finely characterized,and the seismic inversion results can clearly characterize the location of stratigraphic boundaries and meet the needs of accurate geological description.Specifically,for pre-stack AVO inversion,it is required to be able to distinguish smaller geological targets in the depth or time domain,and clearly depict the vertical boundaries of the geological objects.In response to the above requirements,we introduce the preprocessing regularization of the adaptive edge-preserving smooth filter into the pre-stack AVO elastic parameter inversion to clearly invert the position of layer boundary and improve the accuracy of the inversion results.

Geoacoustic Inversion for Bottom Parameters via Bayesian Theory in Deep Ocean

We develop a new approach to estimating bottom parameters based on the Bayesian theory in deep ocean. The solution in a Bayesian inversion is characterized by its posterior probability density （PPD）, which combines prior information about the model with information from an observed data set. Bottom parameters are sensitive to the transmission loss （TL） data in shadow zones of deep ocean. In this study, TLs of different frequencies from the South China Sea in the summer of 2014 are used as the observed data sets. The interpretation of the multidimensional PPD requires the calculation of its moments, such as the mean, covariance, and marginal distributions, which provide parameter estimates and uncertainties. Considering that the sensitivities of shallow- zone TLs vary for different frequencies of the bottom parameters in the deep ocean, this research obtains bottom parameters at varying frequencies. Then, the inversion results are compared with the sampling data and the correlations between bottom parameters are determined. Furthermore, we show the inversion results for multi- frequency combined inversion. The inversion results are verified by the experimental TLs and the numerical results, which are calculated using the inverted bottom parameters for different source depths and receiver depths at the corresponding frequency.

Simultaneous inversion of petrophysical parameters based on geostatistical a priori information

The high-resolution nonlinear simultaneous inversion of petrophysical parameters is based on Bayesian statistics and combines petrophysics with geostatistical a priori information. We used the fast Fourier transform–moving average（FFT–MA） and gradual deformation method（GDM） to obtain a reasonable variogram by using structural analysis and geostatistical a priori information of petrophysical parameters. Subsequently, we constructed the likelihood function according to the statistical petrophysical model. Finally, we used the Metropolis algorithm to sample the posteriori probability density and complete the inversion of the petrophysical parameters. We used the proposed method to process data from an oil fi eld in China and found good match between inversion and real data with high-resolution. In addition, the direct inversion of petrophysical parameters avoids the error accumulation and decreases the uncertainty, and increases the computational effi ciency.

Reservoir parameter inversion based on weighted statistics

Variation of reservoir physical properties can cause changes in its elastic parameters. However, this is not a simple linear relation. Furthermore, the lack of observations, data overlap, noise interference, and idealized models increases the uncertainties of the inversion result. Thus, we propose an inversion method that is different from traditional statistical rock physics modeling. First, we use deterministic and stochastic rock physics models considering the uncertainties of elastic parameters obtained by prestack seismic inversion and introduce weighting coefficients to establish a weighted statistical relation between reservoir and elastic parameters. Second, based on the weighted statistical relation, we use Markov chain Monte Carlo simulations to generate the random joint distribution space of reservoir and elastic parameters that serves as a sample solution space of an objective function. Finally, we propose a fast solution criterion to maximize the posterior probability density and obtain reservoir parameters. The method has high efficiency and application potential.

Generalized linear joint PP-PS inversion based on two constraints

Conventional joint PP-PS inversion is based on approximations of the Zoeppritz equations and assumes constant VP/VS;therefore,the inversion precision and stability cannot satisfy current exploration requirements.We propose a joint PP-PS inversion method based on the exact Zoeppritz equations that combines Bayesian statistics and generalized linear inversion.A forward model based on the exact Zoeppritz equations is built to minimize the error of the approximations in the large-angle data,the prior distribution of the model parameters is added as a regularization item to decrease the ill-posed nature of the inversion,low-frequency constraints are introduced to stabilize the low-frequency data and improve robustness,and a fast algorithm is used to solve the objective function while minimizing the computational load.The proposed method has superior antinoising properties and well reproduces real data.

Zoeppritz-based AVO inversion using an improved Markov chain Monte Carlo method

The conventional Markov chain Monte Carlo （MCMC） method is limited to the selected shape and size of proposal distribution and is not easy to start when the initial proposal distribution is far away from the target distribution. To overcome these drawbacks of the conventional MCMC method, two useful improvements in MCMC method, adaptive Metropolis （AM） algorithm and delayed rejection （DR） algorithm, are attempted to be combined. The AM algorithm aims at adapting the proposal distribution by using the generated estimators, and the DR algorithm aims at enhancing the efficiency of the improved MCMC method. Based on the improved MCMC method, a Bayesian amplitude versus offset （AVO） inversion method on the basis of the exact Zoeppritz equation has been developed, with which the P- and S-wave velocities and the density can be obtained directly, and the uncertainty of AVO inversion results has been estimated as well. The study based on the logging data and the seismic data demonstrates the feasibility and robustness of the method and shows that all three parameters are well retrieved. So the exact Zoeppritz-based nonlinear inversion method by using the improved MCMC is not only suitable for reservoirs with strong-contrast interfaces and longoffset ranges but also it is more stable, accurate and antinoise.