Model-based Inversion Research Articles

Seismic Elastic Parameter Inversion via a FCRN and GRU Hybrid Network with Multi-Task Learning

Seismic elastic parameter inversion translates seismic data into subsurface structures and physical properties of formations. Traditional model-based inversion methods have limitations in retrieving complex geological structures. In recent years, deep learning methods have emerged as preferable alternatives. Nevertheless, inverting multiple elastic parameters using neural networks individually is computationally intensive and can lead to overfitting due to a shortage of labeled data in field applications. Multi-task learning can be employed to invert elastic parameters simultaneously. In this work, a hybrid network that leverages the fully convolutional residual network (FCRN) and the gated recurrent unit network (GRU) is designed for the simultaneous inversion of P-wave velocity and density from post-stack seismic data. The FCRN efficiently extracts local information from seismic data, while the GRU captures global dependency over time. To further improve the horizontal continuity and inversion stability, we use a multi-trace to single-trace (M2S) inversion strategy. Consequently, we name our proposed method the M2S multi-task FCRN and GRU hybrid network (M2S-MFCRGRU). Through anti-noise experiments and blind well tests, M2S-MFCRGRU exhibits superior anti-noise performance and generalization ability. Comprehensive experimental inversion results also showcase the excellent lateral continuity, vertical resolution, and stability of the M2S-MFCRGRU inversion results.

Improved lithospheric seismic velocity and density model of the Korean Peninsula from ambient seismic noise data using machine learning

To accurately estimate physical properties and tectonic behavior in the near-surface, a reliable seismic property model is needed for realistic seismic modeling and earthquake location estimation. Recording ambient seismic noise (ASN) data and producing interferometric reflection images traditionally provides subsurface structure observation without an active source. However, due to low signal-to-noise ratio (SNR) and vertical resolution, interpreting upper mantle structures and inverting seismic models from noise data is difficult. To address this, machine learning (ML) techniques are applied to enhance vertical resolution and interpret geologically meaningful boundaries. Spectral enhancement with convolutional U-net generates higher resolution data by preserving temporal continuity. Unsupervised ML interprets lithospheric boundaries more robustly and objectively than manual horizon picking, and model-based seismic inversion integrates improved seismic data with prior full-waveform inversion (FWI) models. ML-based results improve inverted models, displaying more detailed geological structures and seismic property changes, surpassing seismic data limitations.

Fishbone model-based inversion to estimate physical parameters of elastic structures under earthquake excitations

Fishbone model-based inversion to estimate physical parameters of elastic structures under earthquake excitations

Amplitude preservation and seismic inversion reliability post spectral shaping

Resolution enhancement is often applied to poststack seismic data to obtain higher frequency or wider frequency bandwidth. This enhancement could help seismic interpreters in resolving thinner layers that may not be visible on the original seismic data. The seismic inversion method is the other way to interpret seismic data. The inversion process produces an impedance volume as a layer property that can be used not only for interpretation but also for reservoir characterization. This research aims to study whether vertical resolution enhancement of seismic data preserves the amplitudes and how enhancement affects seismic inversion results. The experiments in this research are performed on synthetic and real data. The modeling is carried out by generating zero-offset synthetic traces using a convolutional model. Meanwhile, real data processing is carried out at the North Sea Volve Data Village at the Ty and Hugin formations. The resolution of the seismic trace is enhanced using a spectral division operator. The inversion techniques used in this research are model-based inversion and sparse spike inversion. The seismic trace after vertical resolution enhancement shows that the separation of events can be improved by this procedure. Inversion of the original and of the enhanced traces shows that the enhancement process still allows obtaining reliable inversion results both from model-based and sparse spike methods; however, there is no significant difference between the inverted impedance using the original trace and enhanced traces. This research concludes that the enhancement process keeps the seismic amplitude preserved for inversion, but it barely improves the inversion result.

Estimation of Reservoir Porosity Using Seismic Post-Stack Inversion in Lower Indus Basin, Pakistan

Seismic post-stack inversion is one of the best techniques for effective reservoir characterization. This study intends to articulate the application of Model-Based Inversion (MBI) and Probabilistic Neural Networks (PNN) for the identification of reservoir properties i.e. porosity estimation. MBI technique is applied to observe the low impedance zone at the porous reservoir formation. PNN is a geostatistical technique that transforms the impedance volume into porosity volume. Inverted porosity is estimated to observe the spatial distribution of porosity in the Lower Goru sand reservoir beyond the well data control. The result of inverted porosity is compared with that of well-computed porosity. The estimated inverted porosity ranges from 13-13.5% which shows a correlation of 99.63% with the computed porosity of the Rehmat-02 well. The observed low impedance and high porosity cube at the targeted horizon suggest that it could be a probable potential sand channel. Furthermore, the results of seismic post-stack inversion and geostatistical analysis indicate a very good agreement with each other. Hence, the seismic post-stack inversion technique can effectively be applied to estimate the reservoir properties for further prospective zones identification, volumetric estimation and future exploration.

Semisupervised seismic impedance inversion with data augmentation and uncertainty analysis

Seismic inversion for impedance is an important task for reservoir characterization. Supervised-learning-based methods improve inversion accuracy for impedance. However, they require a large number of training labels. To reduce the reliance on training labels and further improve the accuracy of impedance inversion, we propose a new data augmentation strategy to simulate layer thickness variation during semisupervised learning which uses unlabeled seismic data properly. Using the Marmousi-II model, we first compare the impedance estimated using the open-loop (OL) method and that obtained using the physics-constrained closed-loop (PC-CL) method at different data augmentation rates to demonstrate the effectiveness of the proposed inversion method. In the case of applying the proposed method to a field data set, comparisons between the estimated impedance results using the OL and PC-CL methods and those obtained using the model-based inversion method are shown for verification. Detailed comparisons between the inversion results of the impedance and well-log data are displayed to emphasize the superiority of the proposed PC-CL method. The residuals between the observed field data and simulated seismic data generated using the estimated impedance are shown to further verify the accuracy of the estimated impedance of the proposed PC-CL method. Aside from the impedance inversion results, epistemic uncertainties are calculated for further performance evaluations of the impedance inversion.

Delineating tidal channel feature using integrated post-stack seismic inversion and spectral decomposition applications of the Upper Cretaceous reservoir Abu Roash C: A case study from Abu-Sennan oil field, Western Desert, Egypt

Delineating tidal channels that bear hydrocarbons in fluvial systems is crucial to minimize drilling risks, and raise the hydrocarbon potential. Therefore, we performed an integrated workflow that combines/compares the outcomes of a pair of post-stack inversion techniques (model-based inversion and colored inversion) and the spectral decomposition attribute to delineate (vertically and horizontally) the hydrocarbon-bearing tidal channel contained in the Upper Cretaceous Abu Roash (AR)/C reservoir in the South West Abu-Sennan (SWS) area in Abu-Gharadig basin in the Northern Egyptian Western Desert (NEWD). We used a dataset consisting of twenty seismic reflection profiles and borehole geophysical data of four wells. Data conditioning was performed in detail for the whole dataset that was used in this study. The application of the absolute impedance (model-based inversion) produced better results relative to the band-limited acoustic impendence (colored inversion). The model-based inverted sections gave an improved subsurface image of seismic data and delineated the reservoir properly. The low-frequency range detected through the use of seismic inversion techniques and the spectral decomposition attribute analysis showed a good fit with the low impedance values corresponding to hydrocarbons. The study results showed a correlation of roughly 0.99 exists between the generated synthetic impedance, original seismic, and the borehole geophysical data characterizing the tidal channel confined within the AR/C reservoir. The match between the post-stack inversion techniques and the spectral decomposition analysis results is considered a visual validation of the delineated tidal channel. The performed workflow can be applied and impact the understanding of comparable reservoirs with similar settings in near areas.

Data for Model-based probabilistic inversion reveals influential assumptions underlying expert assessments of future sea-level rise

Data for Model-based probabilistic inversion reveals influential assumptions underlying expert assessments of future sea-level rise

Theoretical evaluation of utilizing rock physics inversion for hybrid models of gas-hydrate distribution in deep water sediments of Oman sea

Most of the hydrate assessment methods are developed based on relating the elastic properties to the hydrate saturation via establishment of robust and calibrated rock physics models and among them the most practical one is Effective Medium Theory (EMT) with four models of gas hydrate distributions. Generally, considering only one of these models in hydrate characterization could be an unrealistic assumption and therefore, rock physics inversion may cause erroneous assessments (over- or under-estimation).In this paper, hybrid models of hydrate distributions have been introduced for different combinations of two possible classical models by applying some modifications in the EMT workflow. Then, based on the modified EMT for hybrid models, a statistical rock physics inversion approach has been developed for estimation of gas hydrate saturations along with porosity, simultaneously. The result shows that theoretically it is possible to combine two models of distribution in both rock physics modeling and inversion processes. This developed approach was applied on a 2-D marine pre-stack time migrated (PSTM) seismic line passing through an anticlinal-ridge type structure in deep Oman sea sediment. For comparison, this approach was applied for single hydrate distribution as well. The results recommend to use the hybrid models of cementing and non-cementing distributions in the characterization to simultaneously consider all changes on the elastic properties of the host sediment imposed by hydrate presence. The results showed even in an un-explored area (without well information), this method can provide reliable robust result for porosity (independent of hydrate distribution assumption) and qualitative output for hydrate sweet-spots detection. Finally and after drilling the first exploration well, by having well log data in hydrate zone, along with possibility of validations and parameters calibrations for hybrid models, more reliable quantification could be achieved with this method; as the in-situ data increases the certainty of the determined elastic properties in model-based inversion process and also accuracy of defined mineralogical compositions.

Improved reservoir characterization by means of supervised machine learning and model-based seismic impedance inversion in the Penobscot field, Scotian Basin

The present research work attempted to delineate and characterize the reservoir facies from the Dawson Canyon Formation in the Penobscot field, Scotian Basin. An integrated study of instantaneous frequency, P-impedance, volume of clay and neutron-porosity attributes, and structural framework was done to unravel the Late Cretaceous depositional system and reservoir facies distribution patterns within the study area. Fault strikes were found in the EW and NEE-SWW directions indicating the dominant course of tectonic activities during the Late Cretaceous period in the region. P-impedance was estimated using model-based seismic inversion. Petrophysical properties such as the neutron porosity (NPHI) and volume of clay (VCL) were estimated using the multilayer perceptron neural network with high accuracy. Comparatively, a combination of low instantaneous frequency (15–30 Hz), moderate to high impedance (7000–9500 gm/cc∗m/s), low neutron porosity (27%–40%) and low volume of clay (40%–60%), suggests fair-to-good sandstone development in the Dawson Canyon Formation. After calibration with the well-log data, it is found that further lowering in these attribute responses signifies the clean sandstone facies possibly containing hydrocarbons. The present study suggests that the shale lithofacies dominates the Late Cretaceous deposition (Dawson Canyon Formation) in the Penobscot field, Scotian Basin. Major faults and overlying shale facies provide structural and stratigraphic seals and act as a suitable hydrocarbon entrapment mechanism in the Dawson Canyon Formation's reservoirs. The present research advocates the integrated analysis of multi-attributes estimated using different methods to minimize the risk involved in hydrocarbon exploration.

Model-constrained and data-driven double-supervision acoustic impedance inversion

Seismic impedance inversion is an important technique for structure identification and reservoir prediction. Model-based and data-driven impedance inversion are the commonly used inversion methods. In practice, the geophysical inversion problem is essentially an ill-posedness problem, which means that there are many solutions corresponding to the same seismic data. Therefore, regularization schemes, which can provide stable and unique inversion results to some extent, have been introduced into the objective function as constrain terms. Among them, given a low-frequency initial impedance model is the most commonly used regularization method, which can provide a smooth and stable solution. However, this model-based inversion method relies heavily on the initial model and the inversion result is band limited to the effective frequency bandwidth of seismic data, which cannot effectively improve the seismic vertical resolution and is difficult to be applied to complex structural regions. Therefore, we propose a data-driven approach for high-resolution impedance inversion based on the bidirectional long short-term memory recurrent neural network, which regards seismic data as time-series rather than image-like patches. Compared with the model-based inversion method, the data-driven approach provides higher resolution inversion results, which demonstrates the effectiveness of the data-driven method for recovering the high-frequency components. However, judging from the inversion results for characterization the spatial distribution of thin-layer sands, the accuracy of high-frequency components is difficult to guarantee. Therefore, we add the model constraint to the objective function to overcome the shortages of relying only on the data-driven schemes. First, constructing the supervisor1 based on the bidirectional long short-term memory recurrent neural network, which provides the predicted impedance with higher resolution. Then, convolution constraint as supervisor2 is introduced into the objective function to guarantee the reliability and accuracy of the inversion results, which makes the synthetic seismic data obtained from the inversion result consistent with the input data. Finally, we test the proposed scheme based on the synthetic and field seismic data. Compared to model-based and purely data-driven impedance inversion methods, the proposed approach provides more accurate and reliable inversion results while with higher vertical resolution and better spatial continuity. The inversion results accurately characterize the spatial distribution relationship of thin sands. The model tests demonstrate that the model-constrained and data-driven impedance inversion scheme can effectively improve the thin-layer structure characterization based on the seismic data. Moreover, tests on the oil field data indicate the practicality and adaptability of the proposed method.

Full-visibility 3D imaging of oxygenation and blood flow by simultaneous multispectral photoacoustic fluctuation imaging (MS-PAFI) and ultrasound Doppler

We present a method and setup that provide complementary three-dimensional (3D) images of blood oxygenation (via quantitative photoacoustic imaging) and blood flow dynamics (via ultrasound Doppler). The proposed approach is label-free and exploits blood-induced fluctuations, and is implemented on a sparse array with only 256 elements, driven with a commercially available ultrasound electronics. We first implement 3D photoacoustic fluctuation imaging (PAFI) to image chicken embryo, and obtain full-visibility images of the vascular morphology. We obtain simultaneously 3D ultrasound power Doppler with a comparable image quality. We then introduce multispectral photoacoustic fluctuation imaging (MS-PAFI), and demonstrate that it can provide quantitative measurements of the absorbed optical energy density with full visibility and enhanced contrast, as compared to conventional delay-and-sum multispectral photoacoustic imaging. We finally showcase the synergy and complementarity between MS-PAFI, which provides 3D quantitative oxygenation (SO_2) imaging, and 3D ultrasound Doppler, which provides quantitative information on blood flow dynamics. MS-PAFI represents a promising alternative to model-based inversions with the advantage of resolving all the visibility artefacts without prior and regularization, by use of a straightforward processing scheme.

Simultaneous model-based inversion of pre-stack 3D seismic data targeting a deep geothermal reservoir, Northwest Hungary

The well-known, traditional way to extend P-wave acoustic impedance data between and beyond the well log locations is the post-stack inversion of seismic data usually available in the surroundings of the boreholes. A relatively new trend in the seismic exploration is based on the pre-stack inversion of the seismic CDP gathers providing both the P- and S-wave acoustic impedance sections (and volumes), as well as the estimated density data. This methodology is often called as simultaneous model-based inversion and can be utilized not only for hydrocarbon exploration, but it might also be a useful tool for the investigation of geothermal resources. In this study, we will compare the results of post-stack and pre-stack acoustic impedance inversions utilizing the same seismic volume. We will demonstrate and analyze the inverted attribute sections (and some of their derivatives) obtained by the pre-stack algorithm in detail. Finally, we will draw the conclusions about the lithological discrimination of the studied complex carbonate geothermal reservoir located in the pre-Cenozoic basement of the Little Hungarian Plain.

Accurate Detection and Localization of Water Pipe Leaks through Model-Based TDR Inversion

The problem of water scarcity affects many areas of the world due to water mismanagement and overconsumption and, more recently, to climate change. Monitoring the integrity of distribution systems is, therefore, increasingly important to avoid the waste of clean water. This paper presents a new signal processing technique for enhancing the performance of the methodology of leak detection in water distribution pipes based on time domain reflectometry (TDR). The new technique is based on a particular kind of TDR inversion (spatial TDR) based on a "gray-box" lumped parameter model of the system. The model does not include, e.g., radiative phenomena, non-TEM (transverse electromagnetic) modes etc. but is capable of reproducing accurately the complicated reflectograms obtained by a TDR leak detection system assuming a proper profile of capacitance per unit length along the sensing element. Even more importantly, the model is identified using only the reflectograms taken by the system with very little prior information about the system components. The developed technique is able to estimate with good accuracy, from reflectograms with unclear or ambiguous interpretation, the position and the extension of a region where water is located. The measurement is obtained without prior electromagnetic characterization of the TDR system components and without the need of modeling or quantifying a number of electromagnetic effects typical of on-site measurements.

Application of model-based acoustic impedance inversion in the prediction of thin interbedded coal seams: a case study in the Yuwang colliery, Yunnan Province

The Permian coal seams in eastern Yunnan and western Guizhou are thin, numerous, and staggered with other thin coal seams. Depicting the fine characteristics of coal reservoirs is pivotal for the safe and efficient exploitation of coal and coalbed methane (CBM), and is important for transparent mining. To improve the inverse resolution and accuracy of predicting reservoir thickness, this study used the model-based acoustic impedance (AI) inversion method that utilizes seismic and logging data. This method changes seismic data, reflecting stratigraphic interfaces, into AI data, reflecting lithologic structures. Moreover, it avoids the relevant assumptions of wavelets and reflection coefficients. Compared with other inversion methods, model-based AI inversion strengthens the description of thin reservoir horizontal and vertical changes. The results showed that comprehensively using intermediate-frequency seismic information and high-low-frequency logging data greatly broadens the seismic data frequency band and improves the dominant frequency of the reflected wave. Furthermore, the AI profile resolution and the prediction accuracy of the physical parameters for the target geological body can be improved. A cross-validation comparing the inverted thickness and measured thickness of borehole cores was applied to achieve fine prediction (error of appropriately 0.02–0.4 m), providing a basis for CBM development.

Uncertainty quantification of geologic model parameters in 3D gravity inversion by Hessian-informed Markov chain Monte Carlo

Geologic modeling has been widely adopted to investigate underground structures. However, modeling processes inevitably have uncertainties due to scarcity of data, measurement errors, and simplification of the modeling method. Recent developments in geomodeling methods have introduced a Bayesian framework to constrain the model uncertainties by considering the additional geophysical data in the modeling procedure. Markov chain Monte Carlo (MCMC) methods are normally used as tools to solve the Bayesian inference problem. To achieve a more efficient posterior exploration, advances in MCMC methods use derivative information. Hence, we introduce an approach to efficiently evaluate second-order derivatives in geologic modeling and adopt a Hessian-informed MCMC method, the generalized preconditioned Crank-Nicolson (gpCN), as a tool to solve the 3D model-based gravity Bayesian inversion problem. The result is compared with two other widely applied MCMC methods, random-walk Metropolis–Hastings and Hamiltonian Monte Carlo, on a synthetic geologic model and a realistic structural model of the Kevitsa deposit. Our experiment demonstrates that superior performance is achieved by the gpCN compared with the other two state-of-the-art sampling methods. This indicates the potential of the proposed method to be generalized to more complex models.

Improved Model-Based Forest Height Inversion Using Airborne L-Band Repeat-Pass Dual-Baseline Pol-InSAR Data

This paper proposes an improved model-based forest height inversion method for airborne L-band dual-baseline repeat-pass polarimetric synthetic aperture radar interferometry (PolInSAR) collections. A two-layer physical model with various volumetric scattering attenuation and dynamic motion properties is first designed based on the traditional Random Motion over Ground (RMoG) model. Related PolInSAR coherence functions with both volumetric and temporal decorrelations incorporated are derived, where the impacts of homogenous and heterogeneous attenuation and dynamic motion properties on the performance of forest height inversion were investigated by the Linear Volume Attenuation (LVA), Quadratic Volume Attenuation (QVA), Linear Volume Motion (LVM), and Quadratic Volume Motion (QVM) depictions in the volume layer. Dual-baseline PolInSAR data were acquired to increase the degree of freedom (DOF) of the coherence observations and thereby provide extra constraints on the forest parameters to address the underdetermined problem. The experiments were carried out on a boreal forest in Canada and a tropical one in Gabon, where physical models with LVA + QVM (RMSE: 3.56 m) and QVA + LVM (RMSE: 6.83 m) exhibited better performances on the forest height inversion over the boreal and tropical forest sites, respectively. To leverage the advantages of LVA, QVA, LVM, and QVM, a pixel-wise optimization strategy was used to obtain the best forest height inversion performance for the range of attenuation and motion profiles considered. This pixel-wise optimization surpasses the best-performing single model and achieves forest height inversion results with an RMSE of 3.21 m in the boreal forest site and an RMSE of 6.48 m in the tropical forest site.

Machine Learning for Property Extraction from Seismic Megamerges, Northern North Sea

3D seismic MegaMerges provide valuable information over large areas in mature petroleum provinces. While these surveys in combination with borehole data are useful for regional geological and geophysical analyses, they also offer an enormous opportunity for big-data analysis leading to new understandings. In this project, we combine seismic 3D-merges and well data in a geological context. The target is to predict petrophysical properties such as P-wave velocity (VP), Bulk Density (RHOB), Porosity (∅), and Clay Volume (Vclay) from well and seismic data using machine learning (ML) techniques. Our workflow comprises five steps. In the first step, seismic 3D surveys are conditioned, reprocessed, and finally merged. In the second step, a low-frequency Acoustic Impedance (ZP) model is generated using the well-log data over the whole MegaMerge before carrying out a model-based inversion to obtain a ZP cube. In the third step, Multi-linear and different neural network algorithms are used for the ML training and validation analysis. In step 4, the well-log trained ML model is applied to predict VP and RHOB from seismic. And in the final step 5, petrophysical relations and well data are used to predict Vclay and ∅ from the ML-derived properties. In the current example of the Northern North Sea (NNS) MegaMerge, this new approach shows excellent results, providing new insights into the basin geology, petroleum system, reservoir architecture and reservoir quality..

Prediction of Petrophysical Properties Through Comparative Post-Stack Inversion Techniques Using Advance Neural Networking

The sophisticated seismic inversion methodology develops the relationship between the interpreted seismic data and the elastic properties to discerning the desired facies in the reservoir characterization. The study is focused on the C-interval sand of the Lower Goru Formation by utilizing Post-stack Time Migration (PSTM) 3D seismic and borehole logs of six wells. The challenges in the facies discrimination arise due to various complexities, i.e., lithological variability and acquisition sensitivity of recording tools. Various quantitative interpretation techniques such as Band limited inversion, Model-Based inversion, and stochastic inversion are developed to handle the limitations effectively and precisely access the producing facies. A comparative analysis is performed for the heterogeneous sands by employing these QI techniques and evaluating their effectiveness. The inverted seismic attributes are further utilized in the petrophysical properties estimation through Probabilistic Neural Networking (PNN) algorithm to distinguish litho-facies by developing a petro-elastic relationship. Probabilistic Neural Networking works best for the heterogeneous reservoir’s petrophysical properties estimation, especially sand reservoirs with shales intercalation. The concluding results declare the efficiency of all applied techniques for potential sand zones, especially the stochastic inversion. The petrophysical properties volume of clay and effective porosity, estimated from the stochastic inversion attribute, matched with the results of the blind well, resolute precisely the reservoir potential, and delineated the depositional environment of the three segregated sand bodies.

Constrained non-linear AVO inversion based on the adjoint-state optimization

Pre-stack AVO inversion of seismic data is a modeling tool for estimating subsurface elastic properties. Our focus is on the model-based inversion method where then unknown variables are estimated by minimizing the misfit to the observed data. Standard approaches for non-linear AVO inversion are based on the gradient descent optimization algorithms that require the calculation of the gradient equations of the objective function. To improve the accuracy and efficiency of these methods, we developed a technique that uses an implementation of the adjoint-state-based gradient computation. The inversion algorithm relies on three basic modeling components consisting of a convolution-based forward model using a linearized approximation of the Zoeppritz equation, the definition of the objective function, and the adjoint-computed gradient. To achieve an accurate solution, we choose a second-order optimization algorithm known as the Limited memory-BFGS (L-BFGS) that implicitly approximates the inverse Hessian matrix. This approach is more efficient than traditional optimization methods. The main novelty of the proposed approach is the derivation of the adjoint-state equations for the gradient of the objective function. The application of the proposed method is demonstrated using 1D and 2D synthetic datasets based on data from the Edvard Grieg oil field. The seismic data for these applications is generated by using both convolutional modeling and finite difference methods. The results of the proposed method are accurate and the computational approach is efficient. The results show that the algorithm reliably retrieves the elastic variables, P- and S-wave velocities and density for both convolutional and finite difference models.