EAGE 2020 Annual Conference & Exhibition Online

Almost all forms of full waveform inversion use a source wavelet estimate: this wavelet can be extracted from recorded data during the pre-processing or sometimes a zero-phase band-limited synthetic wavelet can be used instead. Alternatively, the source information can be regarded as another unknown in the inverse problem and be updated within the inversion procedure itself.

The importance of reliable source wavelet information during the waveform inversion implementation has been widely implied or briefly mentioned by multiple authors, but little detailed study of the effects of poor wavelets has been presented in the literature.

It is the purpose of this study to examine if shape errors in the source wavelet manifest themselves in a significantly damaging way in the velocity model obtained using conventional time-domain acoustic waveform inversion, and if so by what magnitude.

In addition, although it is clear that a suboptimal source wavelet will degrade the velocity model, we will also assess what effect such degradation has on migrated image positions, and compare this depth error to the inherent uncertainty expected in seismic images. In other words, we will try to assess how much wavelet error is acceptable before the image is distorted beyond typically accepted positioning errors.

Least-square reverse time migration using plane-wave encoding has two problems: Encoding data will introduce crosstalk noise and the excessive number of plane-wave records will reduce the computational efficiency. In this paper, the Seislet transform, which is suitable for seismic data, is combined with the fractional order threshold function based on the Riemann-Liouville fractional integration theory. Then we apply it into the plane-wave least-square reverse time migration. Numerical tests on the complex model show that the plane-wave least-square reverse time migration based on Seislet fractional order threshold algorithm constraint can effectively suppress the crosstalk noise caused by multi-source data. Compared with the traditional method, this proposed method uses less number of plane-wave records to obtain the same imaging effect, and can improve the computational efficiency.

The aim of this study is to better understanding the microstructural and fluid transport properties of gas shale on small size of sample. We use our data to demonstrate of imaging datasets over nano-scales, the integration of 3D technique to identify microstructures properties and the behavior of gas shale which is necessary for modelling elastic behavior, gas storage, gas desorption and gas flow in gas shales. The 3D volumes of shale showed significant in nano-structure (pore size and volume distribution, mineral content, porosity, pore aspect ratios and surface area to volume ratios).

A well-established method for fluid characterization is to use regression on the critical parameters of the grouped components of an equation of state (EOS) to replicate the results of fluid experiments performed in the laboratory, mainly constant composition expansion (CCE), constant volume depletion (CVD), and differential liberation (DL). In the case of many mature reservoirs, however, proper fluid laboratory examination is not available. This paper proposes an alternative fluid characterization methodology based on the Engler distillation test (ASTM86). Its objective is to help engineers derive key fluid parameters such as formation volume factors and oil-gas ratios in the absence or limitation of PVT-cell experimental data, based only on the Engler distillation test (ASTM86) results and a fluid composition up to C5+.

The suggested methodology was applied on multiple highly heterogeneous fields located in the Dnieper-Donetsk Basin in Eastern Ukraine and proved to be useful for all the fields of varying fluid types ranging from lean gas with a condensate yield (presence of C5+ per cubic meter of gas) of 10 g/m3 to very rich retrograde gases of 500 g/m3.

Contamination of seismic data by background noise causes difficulties for the following inversion, imaging, stratigraphic interpretation, etc. Desert seismic records pose a particular problem because of the strong energy of desert random noise and its serious spectrum overlapping with effective signals. Thus, the robust principle component analysis (RPCA) is introduced into the denoising of desert seismic data. RPCA is a classical method of low-rank matrix recovery. By kernel norm optimization, it can decompose noisy data into optimal low-rank matrix (LM) and sparse matrix (SM) which contain most effective signals and noise, respectively. However, due to the low signal-to-noise ratio and serious spectrum overlapping of desert seismic record, there are still a lot of random noise in its optimal LM. Therefore, the convolutional neural network (CNN) is combined with RPCA to establish the optimal mapping relationship from noisy LM to desert random noise through the training of CNN, so as to accurately predict desert random noise from the LM of desert seismic record. Finally, the denoising result is obtained by subtracting the desert random noise predicted by CNN from the LM. Experiments show that this improved RPCA can suppress random noise and recover effective signal more effectively than the traditional methods.

The source function accuracy plays an important role in a successful full-waveform inversion (FWI) application. So we often need to estimate the source function before or within the inversion process. Source estimation requires additional computational cost, and an inaccurate source estimation can hamper the convergence of FWI. We develop a source-independent waveform inversion utilizing a recently introduced wavefield reconstruction based method we refer to as efficient wavefield inversion (EWI). In EWI, we essentially reconstruct the wavefield by fitting it to the observed data as well as a wave equation based on iterative Born scattering. However, a wrong source wavelet will induce errors in the reconstructed wavefield, which may lead to a divergence of this optimization problem. We use a convolution-based source-independent misfit function to replace the conventional data fitting term in EWI to formulate a source-independent EWI (SIEWI) objective function. In SIEWI, this new formulation is able to mitigate the cycle-skipping issue and the source wavelet uncertainty, simultaneously. We demonstrate those features on the Overthrust model.

In the workflow of Full-Waveform Inversion (FWI), we often tune the parameters of the inversion to help us avoid cycle skipping and obtain high resolution models. For example, typically start by using objective functions that avoid cycle skipping, and then later, we utilize the least squares misfit to admit high resolution information. Such hierarchical approaches are common in FWI, and they often depend on our manual intervention based on many factors, and of course, results depend on experience. However, with the large data size often involved in the inversion and the complexity of the process, making optimal choices is difficult even for an experienced practitioner. Thus, as an example, and within the framework of reinforcement learning, we utilize a deep-Q network (DQN) to learn an optimal policy to determine the proper timing to switch between different misfit functions. Specifically, we train the state-action value function (Q) to predict when to use the conventional L2-norm misfit function or the more advanced optimal-transport matching-filter (OTMF) misfit to mitigate the cycle-skipping and obtain high resolution, as well as improve convergence. We use a simple while demonstrative shifted-signal inversion examples to demonstrate the basic principles of the proposed method.

Regularization methods are widely used in impedance inversion problems to solve the ill-posed problem. At present, some common regularization methods are applied to the poststack seismic inversion problems by imposing isotropic smoothness on impedance. However, isotropic smoothness can blur small-scale geologic features and reduce the resolution of inversion results. To overcome this shortcoming, we have developed a new regularization that imposes smoothness along the orientations of geological structures. Such a structure-oriented regularization is often isotropic in most areas but can be anisotropic in areas where the structural features are anisotropic. Therefore, our method can preserve small-scale structural features and increase the accuracy and lateral continuity of inversion results. The inversion results of synthetic seismic data demonstrate that the proposed method can effectively improve the resolution and accuracy of the inversion results.

Locating microseismic events is an important procedure in oil and gas extraction. Time-reversal based methods generate source images, which can be used to locate the microseismic events with the help of proper source imaging conditions. However, such images are often contaminated with artifacts, including imaging artifacts, which hamper the proper identification and location of such events. We use the support vector machine (SVM) to develop a classification algorithm of microseismic events within a source image. Using certain features of the source image as input, this trained SVM system is able to distinguish whether each grid point in the source image corresponds to a proper event or an artifact. Applications on a homogeneous model and a Marmousi model show the effectiveness of the proposed method.

The standard acoustic approximation in transversely isotropic (TI) media results in an approximate P-wave phase velocity expression with a complicated square root term. Thus, the corresponding wavefields suffer from the unstable SV-wave artifacts. Many subsequent pure P-wave simulation methods were proposed to eliminate the SV-wave artifacts at the expense of decreasing the accuracy or increasing the computing cost. In this abstract, we propose an optimized acoustic approximation for P-wave in TI media to overcome the defect of the standard acoustic approximation. Since the P-wave propagation is weakly dependent on the vertical S-wave velocity, we construct a function of the vertical S-wave velocity squared to approximate the P-wave phase velocity. The corresponding expression is quite concise, without square roots and complicated fractions. Then, we derive a pure P-wave equation in tilted transversely isotropic (TTI) media. We compare and analyze the wavefields of several simulation methods in the homogeneous and complex TTI models. The wavefields of the proposed method only involve the P-wave propagation with high accuracy. And the computing cost is acceptable. So, this optimized acoustic approximation and the corresponding pure P-wave simulation method can be further used for the reverse time migration and full waveform inversion in TI media.

We have trained a neural network to estimate the near surface Vs profile directly from phase velocity vs. frequency panels. These panels are constructed from the raw shot gathers with all the surface and body waves present. As such, the method has the same goal as dispersion curve inversion. The same approach is applicable to estimation of Vp also. The method has been tested on the SEAM Arid model synthetic dataset and produces encouraging results. Generalization of the method to unseen data remains a challenge, but by brute force modelling and training progress can be made.

Traditional approaches of utilizing the dispersion curves in S-wave velocity reconstruction have many limitations, namely, the 1D layered model assumption and the automatic/manual picking of dispersion curves. On the other hand, conventional full-waveform inversion (FWI) can easily converge to one of the local minima when applied directly to complicated surface waves. Alternatively, a wave equation dispersion spectrum inversion can avoid these limitations, by inverting the slopes of arrivals at different frequencies. A local-similarity objective function is used to avoid possible cycle skipping. We apply the proposed method on the large-scale ambient-noise data recorded at a large-N array with over 3000 recorders. So we can estimate the shear-wave velocities down to 1.8 km depth. The main benefits of the proposed method are 1) it handles lateral variations; 2) it avoids picking dispersion curves; 3) it utilizes both the fundamental- and higher-modes of Rayleigh waves, and 4) it can be solved using gradient-based local optimizations. A good match between the observed and predicted dispersion spectra also leads to a reasonably good match between the observed and predicted waveforms, though the inversion does not aim to match the waveforms.

As the main storage space of shale reservoir, micro-nano pores have a great influence on the overall elastic property of shale. As a special type of organic mineral in shale, the state of kerogen in shale varies with maturity, meanwhile, kerogen is also the primary place for the growth of micro-nano pores. Conventional shale rock physics model cannot reflect the role of micro-nano pores, thus, we adopt a theory of micro-nano pore to describe its characteristics. Considering the micro-nanometer pores and the state of kerogen at different maturity, we establish a new rock physical model by applying the micro-nano pore model, anisotropic SCA-DEM model, anisotropic Eshelby-Cheng model and Brown-Korringa solid substitution equation. The sensitivity analysis show that the micro-nano pores have the greatest effect on the mature shale, while kerogen and clay have the least effect on the overmature shale.

The East Java Basin is a productive Tertiary Basin that has been producing hydrocarbon. During the basin forming and development, East Java Basin has gone through three major geological events from Late Cretaceous until present day. North Madura Platform is located in the northern part of the East Java Basin, formed during the Paleogene divergence. During all basin development stages, North Madura Platform has been undisturbed by tectonic disturbances, allowing carbonate to grow and develop across the platform.

Identifying the varieties of the carbonate has been an integral part of the basin analysis in East Java Basin, as most of the proven reservoirs in North Madura Platform are found in Kujung carbonate. Various types of carbonate have been identified using the seismic data, including shelf edge carbonate, platform carbonate, and patch reefs.

Seismic FWI PSDM reprocessing in 2019 has shown tremendous improvement in resolving carbonate seismic facies in North Madura Platform. Detailed carbonate facies has been identified inside Kujung Carbonate, showing different facies and internal characteristics of the carbonates in different regions. This study will showcase the complexities identified in the carbonates that have been developed in North Madura Platform based on seismic facies characteristics from seismic FWI PSDM reprocessing.

The Marte field is located in the North East part of Block 31 offshore Angola in water depths up to 2km. The Marte reservoirs are made up of 3–5km wide lower Miocene deep water erosional turbidite slope channel complexes in a four-way asymmetrical anticline structure. Imaging on the Eastern flank of the structure is compromised due to an overlying shallow gas hydrate channel that has resulted in velocity model errors and absorption effects that have to date not been adequately modelled and compensated for.

In this case study we will show how an integrated workplan was put in place to resolve the imaging issues observed on the Marte field. This involved re-creating the issue with a synthetic model, using this model to design and then acquire a bespoke velocity survey and then using the resulting data to produce updated models of velocity, anisotropy and Q. These models were then used to generate improved images from a 4D monitor survey acquired over PSVM prior to the velocity survey. Finally, we will show the results of post imaging analysis performed to determine the most significant survey design factor that contributed to the improved models and images obtained.

Noise attenuation is a crucial and recurrent step in the seismic processing sequence. After noise attenuation, quality control (QC) is a mandatory process to ensure that the level of noise left in the data is acceptable and no signal leakage has occurred. This process is usually done by geophysicist and is time consuming and subjective. We train a U-Net convolutional neural network model to automatically perform the QC after swell noise attenuation and label the seismic samples as signal, noise or signal leakage. We show that the classification of the acquired seismic data after the swell noise attenuation with the trained model is very reliable and robust and model is able to detect both residual noise and signal leakage. We also propose a framework to use the classification result to steer the denoise process in an automated fashion. If the model detects residual noise or signal leakage during the denoise process, the selected parameters are automatically tuned to produce the best possible result for each seismic record. We demonstrate that the automated denoise process outperforms the fixed parameters denoise process.

We have implemented a deep learning based first break picker and trained it on various land seismic datasets and evaluated a number of neural network architectures. A deep network with U-net architecture, pre-trained on coarse scale input data provided the most accurate results. Because we use the full shot gather at various scales, the impact of noisy traces is reduced. The neural network corrects random mispicks. This suggest a practical application, namely to train, or re-train a pre-trained network via transfer learning, on a single dataset after conventional FB picking.

It has always been the focus of researchers to accurately identify gas hydrate location. Geophysical prospecting is a widely used method for gas hydrate exploration, which has high credibility, especially seismic exploration technology is most generally used. In our study, we analyze the different physical properties of gas hydrate and other minerals bearing in unconsolidated and high porosity marine sediments based on the effective medium theory. Thus, a new attribute is put forward to discriminate gas hydrate. The logging data at Dongsha area of South China Sea and Hydrate Ridge in Oregon continental margin are applied to validate this method. Our test results are basically in line with actual situation, which provides a new understanding in hydrate identification.

We propose a coherence calculation method based on cross-correlation in orthogonal directions. We first calculate a predicted seismic trace at the location of center trace in each direction using the inverse distance weighting interpolation algorithm, and based on the fact that the biggest difference shall be the difference between the predicted trace calculated in the direction parallel to the structural trend and that calculated in the direction perpendicular to the structural trend, and the corresponding cross-correlation value of the two traces shall be the smallest, we cross-correlate the predicted seismic traces in orthogonal directions and choose the minimum cross-correlation value as the final coherence attribute. Since every trace is weighted according to its relative distance from the centre trace during the prediction of centre trace, the final coherence result is mainly determined more by the most nearby traces of the centre trace than the distant seismic traces. Therefore, the positioning of structural boundaries are more accurate than other coherence methods. We demonstrate that structures detected by the proposed method are more accurate and much clearer than those detected by conventional C3 method via synthetic and field data sets. The new method may be a potentially tool for facilitating seismic interpretation.

The paper present a semi-supervised machine learning workflow that integrates geochemical measurements, elastic logs, pre-stack seismic inversion parameters and non-seismic measurements to classify source rock maturity, and propagate the classes away from the wells in a controlled manner. Semi-supervised algorithms are able to discover spatial structures in high dimensional space by using the unlabeled data. This type of learning algorithms are useful in situations where data labels are limited. The algorithm spreads labels by constructing a similarity graph over the input items and minimizing a loss function with regularization properties making it robust to noise. Data analysis indicate that maturity is not only an attribute of the rock but it is also an attribute of the intrinsic properties of the location. Using location indicators, the algorithm was able to create a regional distribution of maturity.

Label and algorithm are two main factors that influence the effect of deep learning inversion (DLI). Most present researches focus on optimizing algorithms, and pay less attention to how the labels affect the inversion effect, which results in that the application effects of the same algorithm varying with the application regions. This article highlights the importance of label conditions on the effect of DLI. The comparison of inversion results performed with 4 different label sets indicates an ideal DL inversion requires a large number of high-quality and large-diversity labels.1) Label quality is the most important factor, for it directly determines the correctness of inversion results. 2) The increase in both quantity and diversity of the labels is very effective for improving the inversion result, and the diversity is relatively more critical. 3) The better the label structure matches the geological pattern, the better the inversion results are. The research demonstrates that forward modeling based on well interpolation model is an effective method for label augmentation. In order to take full advantage of deep learning, we should integrate it with classic geophysical methods and make them complement each other.

Accurate characterization of the near-surface velocity model is often a prerequisite for effective migration. Refraction tomography may fail to produce a satisfactory velocity model in the presence of velocity inversions and requires accurate traveltime picking. Full waveform inversion (FWI) can overcome these issues, but often requires a good initial model or the presence of sub-5Hz frequencies in the recorded data, and its 3D implementations can be computationally costly. To address these challenges, we propose a 1D version of Laplace-Fourier acoustic FWI, building on the relative insensitivity of Laplace-Fourier methods to the quality of the initial model. The proposed method approximates locally the 3D medium by an effective 1.5D medium and inverts independently for local 1D velocity profiles, that can be upscaled to a full 3D velocity volume. The independence of the 1D inversions and the cylindrical symmetry exhibited by 1.5D forward modelling can be taken advantage of to produce an efficient, highly parallelizable implementation. The feasibility of the method is studied using a synthetic example, with encouraging results.

As one of the most important procedures in shale gas exploration and development, sweet spots prediction is the preferred method for acquiring and maintaining high productivity and effective production of shale gas.

In this study, firstly, based on well log interpretation and rock physics analysis, the quantitative relationships between the elastic parameters and the evaluation parameters such as density and TOC, were established. Then, the planar distribution of some key reservoir parameters including TOC, high-quality reservoir thickness, brittleness and formation pressure were predicted through prestack inversion of all gathers. Finally, with the fuzzy comprehensive evaluation method, an evaluation scheme was proposed and the spatial distribution of sweet spots in the block was assessed. Based on this scheme and the regional distribution of sweet spots predicted by the seismic data, the Z block was divided into three classes. Class I and Class II as the sweet spots are mainly distributed in the east of the block, which are suggested to give priority to the development. This study can provide important guidance for the factory development of the block.

In presence of large elastic parameter variations, acoustic waveform inversion creates artefacts because the approach overfits the data with an inaccurate physical model. An elastic formulation is required to correctly account for the finite-frequency effects and notably the scattering that occurs inside the first Fresnel zone. Over the years, nodal acquisition has become more popular. In the waveform inversion context, it provides low-frequency, long-offset, and full-azimuth data that are very relevant to recover the low-to-mid wavenumber information of the earth parameters. Moreover, with offshore nodal acquisition, not only the pressure field is recorded but also the three particle-displacement or velocity fields. The noise does not have the same effects on the different recording component due to its directionality. Using a nodal acquisition for the Gulf of Mexico, we present the elastic inversion results obtained with the hydrophone and vertical geophone components. We discuss the relevance of considering the vertical geophone component in velocity model building.

In this study we provide a systematic workflow of fracture-vug extraction and reconstruction for electric well logging image preprocessing for the purpose of accurate quantitative assessment of fractures and vugs reservoirs. It includes first step of identification of fractures and their edges by incomplete path morphologic scheme, and the second step of pattern recognition of the electric well logging images through a correlation with a family of sinusoidal functions prepared for the picking up of fracture parameters for high dip fractures and the Hough transformation for low dip fractures. Then the statistical step has been applied to parameters extraction for the reservoir description. With the suitable combination of parameters can we derive complete fractures and vugs that may have admissible missing pixels by the incomplete path opening operation. The cluster of sinusoid can make up for the deficiency of Hough transform in the extraction of high dip angle fractures and incomplete fractures, and overcome the weak robustness of Hough transform to noise.

Building velocity models for depth imaging can be time-consuming and regularly requires manual intervention. The workflows tend to be ‘stop-go’ chained processes. An approach using pseudo-randomness can be used to build a velocity model, mitigating some of the inefficiency challenges associated with traditional velocity model building. Monte Carlo simulations use random sampling to resolve problems where the solution may be insufficiently defined. Using a Monte Carlo approach for velocity model building firstly requires an understanding of how the data quality impacts the model, prior to creating a population of models to invert. A statistical analysis loop of the inverted population of models, followed by numerous repeated cycles, enables a level of automation in velocity model building. The protracted chained approach of classical model building can be replaced by parallelized compute intensive methods to achieve an accurate velocity model in a reduced timeframe. In this abstract we demonstrate the benefits for both quality and time, of weight of statistics and automation when using a Monte Carlo simulation for velocity model building.

Seismic waves are affected by wavefront diffusion and medium absorption during propagation, resulting in high-frequency energy attenuation and phase distortion, which will seriously reduce the resolution of seismic data. With the improvement of high-resolution processing requirements for seismic data, it is very important to obtain high-quality seismic profile by compensation processing. Seismic data are non-stationary signals, neither the spectral analysis nor the simple waveform analysis can accurately describe the time variant characteristics of the seismic waves. In order to better describe the time-frequency characteristics of the seismic signal, the time-frequency analysis method is used to process the seismic signal. Therefore, the compensation of seismic signal in time-frequency domain is considered. Geophysicists implement attenuation compensation based on Gabor transform and continuous wavelet transform(CWT), respectively. We generalize and implement the attenuation compensation method based on Generalized S transform(GST). Subsequently, we analyse the accuracy of three kinds of time-frequency domain compensation by numerical simulation. It is concluded from the error analysis that the compensation based on GST has higher accuracy. The real data shows that the resolution are significantly improved and the fine structure of seismic section is highlighted. It greatly improves the accuracy of the compensation processing of the real data.

Cost-effective hydrocarbon production from unconventional reservoirs relies on multi-stage hydraulic fracturing operations. Suitable shale intervals are usually identified on the basis of: (i) organic-richness; and (ii) mechanical brittleness to promote the creation of a multidirectional fracture network, and therefore a larger stimulated reservoir volume. The in-situ stress state notoriously governs shale deformation and fracturing processes at depth. In the Goldwyer shale formation (Canning Basin, Western Australia), existing well logs from the Theia-1 well are used to build a 1-D mechanical earth model through a brittleness analysis and direct validation with existing laboratory triaxial rock mechanics tests. A progressive transition with depth from a strike-slip (shallower intervals) to a hybrid faulting regime (deeper intervals) is inferred in the Goldwyer formation. The changing stress regime correlates with the estimated variation in dynamic elastic brittleness with depth. This analysis suggests a power law relationship between static Young’s modulus and deformational brittleness (B3). This power law function is then used to create a continuous brittleness profile along the well, across the entire Goldwyer formation interval. Based on this study, the G-III unit (1500m to 1590m in the Theia-1 well) of the Goldwyer formation is the most prospective for hydraulic fracturing operations.

Common appraisal methods for oil and gas reservoir often begin with 3D seismic and exploration wells. These technologies provide spatial recognition along with focused stratigraphy and subsurface resources content. Even though models and simulations predicts a reservoir dynamic, measuring this key component in time complements spatial technologies while providing relevant information regarding field optimization. Further intents to go towards continuous monitoring have demonstrated the capability for seismic to detect reliable short-term calendar 4D effects that would be missed by conventional 4D seismic. These techniques have proven efficiency yet remain expensive; this paper presents a new light seismic asset monitoring solution.

An ultra-light continuous monitoring method has been developed to focus on a specific “spot” location defined by reservoir engineer studies. To illuminate a given spot, a seismic spread, composed of one receiver and one source position, is defined by analysis of existing 3D seismic data. This procedure allows for a very high temporal density monitoring tool targeted at a specific reservoir location and is economically attractive.

This production case study gives further understanding about an active gas storage dynamic showing encouraging results for such a light asset monitoring tool and paves the way for focused and continuous seismic monitoring.

We consider the orthorhombic (ORT) model to characterize the vertically fractured medium and analyse the plane wave reflection coefficients for the ORT layer embedded into the ORT space. We decompose the exact reflection coefficients into a series expansion wherein the successive series terms correspond to different orders of intrabed multiples. Under the weak-contrast assumption, we derive the first-order reflection coefficient approximations for the ORT layer. The approximations are tested numerically and can give insights into the reflection responses.

In 2018, Equinor released to the public an open subsurface dataset comprising all the data of the Volve Oil field from the central Norwegian North Sea. The dataset includes 4D seismic that we interpret, calibrate and quantify to better understand the distribution of water sweep in the reservoir.

Calibration of a dynamic Petro-Elastic Model (PEM) is a key part of the 4D interpretation. Saturation changes in the reservoir are determined to be the primary reason for a 4D signal with a maximum impedance change of 5% expected. The PEM is then combined with 4D forward modelling of well logs to establish links between the the character of the 4D signal and water sweep in the three reservoir zones at Volve. Discrepancies between production logging results and those which best match the 4D seismic response can be combined with the qualitative understanding of the seismic uncertainty to suggest improvements to the reservoir simulation model.

Full-waveform inversion (FWI) is a wave-equation-based inversion method to estimate the physical parameters of the geological structures by exploiting full-information of the seismograms. However, FWI is inherently an ill-posed problem that is sensitive to noise, especially to outliers in the dataset. Usually, this technique is formulated as a least-squares optimization problem that consists of to minimize the difference between the observed and the modelled seismic data (residuals). In this approach, the least-squares solution inversion problem determines the maximum likelihood for the residuals, where all residuals are assumed to follow a Gaussian distribution. However, the distribution of residuals is seldom Gaussian for non-linear problems. In this study, we propose an alternative objective function to mitigate FWI sensitivity to noise based on the q-Gaussian probability distribution. In contrast to Gaussian distribution, the q-Gaussian distribution has long-tails, being less sensitive to outliers. Application on acoustic synthetic noisy-data illustrates the performance between our proposal and FWI based on least-squares norm L2. In addition, we compare also with robust objective functions based on Huber criterion and least-absolute-values norm L1. Numerical experiments show that FWI based on the q-Gaussian probability distribution outperforms other approaches, especially in presence of outliers.

Conventional migrations are apt to structure imaging, but often incapable of generating true-amplitude subsurface image. We propose a true-amplitude Gaussian beam migration (GBM) method under the framework of seismic illumination compensation. A novel scheme based on the GBM is developed to estimate the point spread functions, with which the illumination compensation can be efficiently implemented in the local wave-number domain. The total computational cost of the proposed true-amplitude imaging includes one Born modelling process and two conventional GBM processes, which is more efficient than the true-amplitude imaging using least squares migration which requires multiple iterations. Numerical examples using synthetic data demonstrated the effectiveness and efficiency of the proposed method.

Low-salinity waterflooding (LSWF) is a promising IOR/EOR methodology which has been extensively investigated over the past ten years. However, mixing of the injected brine (low-salinity) with the in-situ high salinity brine (formation water) at the displacement front can decelerate oil recovery by LSWF. We propose that one promising approach to control in-situ mixing is to increase the viscosity of the low salinity injection brine by adding polymer which then improves the mobility ratio at the displacement front and subsequently suppresses dispersion of low salinity in high salinity. This, to our knowledge, has not been addressed before. To investigate systematically the mixing phenomenon in different salinity gradients and its sensitivity to HPAM polymer concentrations, sandpack experiments were designed and executed. Two sets of mixing experiments under single-phase condition were performed to capture the impact of absence or presence of polymer in the low-salinity brine. Our results show that polymer could be used as a mixing-control agent and the dispersivity of the system may be decreased to lower than 0.3 of its original value by adding 200 ppm of polymer.

Current full waveform inversion (FWI) does not make full use of all components data: using only the pressure component (by acoustic FWI) or velocity component (by elastic FWI) data. Based on the acoustic-elastic coupled equation (AECE), we propose a multiple component FWI method to get P-wave and S-wave velocities simultaneously. We use both pressure and velocity component data to form the objective function. Using the adjoint-state method, we deduce the adjoint equation of the AECE and the gradient formula in time domain. Numerical experiments show that results of our method using both pressure and velocity component data are the best, using only velocity component data are the second, and using only pressure component data are the worst. Meanwhile, FWI based on the AECE achieves better results than traditional elastic FWI. This indicates that this approach is an effective way for model building in OBC/OBS applications.

Solving the Helmholtz wave equation provides wavefield solutions that are dimensionally compressed, per frequency, compared to the time domain, which is useful for many applications, like full waveform inversion (FWI). However, the efficiency in attaining such wavefield solutions depends often on the size of the model, which tends to be large at high frequencies and for 3D problems. Thus, we use a recently introduced framework based on predicting such functional solutions through setting the underlying physical equation as a cost function to optimize a neural network for such a task. We specifically seek the solution of the functional scattered wavefield in the frequency domain through a neural network considering a simple homogeneous background model. Feeding the network a reasonable number random points from the model space will ultimately train a fully connected 8-layer deep neural network with each layer having a dimension of 20, to predict the scattered wavefield function. Initial tests on a two-box-shaped scatterer model with a source in the middle, as well as, a layered model with a source on the surface demonstrate the successful training of the NN for this application and provide us with a peek into the potential of such an approach.

The Utsira ocean-bottom node (OBN) survey was acquired during 2018/19 covering 1510 km2 over the Utsira high, which is ∼90 nautical miles west of Stavanger. As it was acquired using several asynchronous triple-source vessels, deblending was an essential step to recover signal at target depths. Deblending also removed very strong nearby seismic interference, by using the interfering survey’s shot times, with as many as 14 active sources firing.

Up-down deconvolution is a key step to address the challenges observed in the area as it achieves both free-surface multiple elimination and 3D signature deconvolution. A similar approach was developed to tackle the same issues for the down-going wavefield resulting in a more effective and efficient processing flow for imaging shallow targets than conventional techniques.

The long offsets and rich azimuthal sampling of the wavefield meant FWI could solve model building and imaging challenges over a range of depths. Such challenges included shallow channels and deeper cemented sand injectites.

This survey illustrates how advances in acquisition and processing technologies enable large-scale, high-density OBN surveys to be acquired and processed in accelerated time frames leading to new insights even in relatively well-explored areas.

Fluid mobility attributes extracted from poststack low-frequency seismic data have exhibited some potential for reservoir characterization and fluid identification. Compared to poststack seismic data, prestack seismic data contains more information about reservoirs and fluids. In order to extract fluid mobility information from prestack seismic data, we establish the relationship between fluid mobility and incidence angle based on frequency-dependent AVO analysis. We define the relationship as prestack fluid mobility which means that fluid mobility varies with angle/offset (FVA/FVO). By establishing models of reservoirs with different fluids, the corresponding prestack fluid mobility is estimated. The results show that prestack fluid mobility can distinguish the oil-bearing reservoir (class III AVO) from the water-bearing reservoir (class IV AVO). In order to estimate prestack fluid mobility, we further derive a Synchrosqueezed Generalized S-transform-based prestack fluid mobility calculation method. This method estimates the FVO by the difference in fluid mobility of the different partially stack gathers. The method is applied to the seismic data. Compared with poststack fluid mobility, FVO can better identify fluid properties and reduce the uncertainty of hydrocarbon-bearing reservoir prediction in the absence of wells.

We presented a multichannel blind acoustic impedance inversion method. The method is an extension of the principle of Euclid deconvolution, which can make full use of the information of multi-trace seismic data to obtain acoustic impedance without estimating seismic wavelet. Significantly, the 2D total variation (TV) constraint is added to the cost function to suppress the random noise and keep the boundary characteristics of the stratum. Also, to obtain the absolute impedance, the low-frequency information extracted from the well logs is used to supplement low-frequency component of the inversion result. Finally, to demonstrate the effectiveness of the proposed method, we apply the method to synthetic data and field data, and confirm that the proposed method can achieve credible acoustic impedance.

Deep convolutional neural networks (DCNNs) are growing in popularity in seismic data processing and inversion due to their achievements in signal and image processing. In this paper we explore the link between DCNN and seismic processing. We demonstrate the potential of the application of DCNNs to seismic processing by analysing its performance with data deblending as an example. We discuss challenges and issues to solve before deploying DCNNs to production, and suggest some directions of study.

Traditional model building techniques are challenged by the complexity of the wavefield in volcanic basin and finite frequency approach should be preferred. In view of the large velocity contrast observed at the top and base volcanic, elastic effects may occur. We apply FWI to build a velocity model for a narrow azimuth slanted streamer data set acquired in the North Atlantic margin. Both acoustic and elastic FWI are run to evaluate the severity of the elastic effects. The fit with the data after the inversion is good for both the acoustic and elastic inversions, but the acoustic velocity model shows fast velocity artifacts like the ones usually observed above salt interfaces. The migrated images and the pre-stack gather confirm that the elastic inversion result is superior to the initial model and the to the acoustic model obtained by inversion. Elastic FWI can reconstruct a complex velocity model with wavelength size variations of large magnitude and can potentially be the tool of choice for allowing imaging under volcanic layers.

Imaging the subsurface with land seismic data requires a high resolution model of the near-surface. Conventional methodologies produce inaccurate results in the presence of velocity inversions. Full waveform inversion (FWI) is able to reconstruct these velocity variations, but is computationally expensive and is ineffective in areas of poor seismic data quality. We obtained an accurate near-surface velocity model in a complex wadi structure previously studied. To achieve this, we employed the recently developed 1.5D version of Laplace-Fourier acoustic FWI coupled to an automatic method for near-surface analysis.

East Siberia, Russia is a vast, primarily underexplored area with prolific hydrocarbon potential. Some of the reasons for limited exploration can be attributed to its remoteness and to the extreme complexity of the surface and near surface conditions. While high density seismic acquisition techniques have been widely proven in open access areas where source and receiver points can be obtained or deployed in dense fashion, what can be done for remote, heavily forested areas to acquire high density surveys like East Siberia. In this paper we evaluate the planning and design work necessary for complex environments and propose acquisition techniques and survey designs that will make high trace density 3D acquisition possible amidst extreme surface and near surface complexities. We find that high data density survey techniques are now feasible using modern approaches for increasing channel count in the field (reduced bin spacing for improved spatial resolution) while increasing overall trace density albeit complex surface and near surface conditions.

Selecting a matched wavelet to the seismic wavelet is a key issue for characterizing time-frequency features of seismic data accurately. The three-parameter wavelet (TPW) can match different seismic wavelets by adjusting three parameters. However, it is difficult to select an appropriate TPW matched well with seismic wavelets in real applications. In this study, we propose a basic wavelet selection method by using the TPW and deep learning algorithm. The proposed workflow first builds a mapping relationship between seismic wavelet and seismic data by using the alternating iterative deep neural network (AIDNN). Based on this relationship, we then estimate the seismic wavelet. Based on the estimated seismic wavelet, we can finally obtain an analytical basic wavelet by matching the TPW to the extracted wavelet by solving an optimization problem. Note that we name the TPW with optimized parameters as the optimum TPW (OTPW), and its WT is OTPWT. To demonstrate the validity and effectiveness of the proposed workflow, we apply it to synthetic traces and field data. Both synthetic and field data examples illustrate that the OTPW characterizes time-frequency features of seismic data with high resolution and is with good anti-noise property.

Recently, a major breakthrough in the exploration of volcanic gas reservoirs was first achieved in Sichuan Basin, which indicates the huge volcanic rock exploration potential in this area. However, the prediction of volcanic reservoirs is very challenging because of the strong heterogeneity, the vague interior reflection structure and the low exploration level with sparse wells in this area. To reduce the exploration risk, we develop an integrated prediction strategy for the gas-bearing volcanic reservoirs using the full stack seismic data by combining the Bayesian adaptive seismic inversion and the frequency-dependent fluid mobility attribute. In the seismic inversion, an automatically adjusted prior stabilizer is derived to balance between the vertical resolution and the inversion stability according to the noise level. In the gas detection, the fluid mobility attribute is calculated by the high precision matching pursuit algorithm to directly indicate the gas reservoirs. Application in a newly discovered volcanic gas production area in Sichuan Basin shows that the integrated prediction results of gas-bearing reservoir at the borehole-side traces match well with the log interpretation result, which demonstrates the effectiveness and feasibility of this integrated method.

Shear wave velocity is the basic data for reservoir prediction and fluid detection. Therefore, an effective method to accurately predict S-wave velocity is one of the important research re-quirements. In this paper, we propose a workflow for successfully predicting S-wave velocity in tight carbonate areas. In view of the general lack of mineral content and shear wave information in logging data, we carried out the inversion of the matrix modulus firstly by a self-adapting method, and then according to the characteristics of microcrack development in tight carbonate rocks, simplifying the pore type. Through the rock physics model, the shear wave velocity in this area is predicted successfully.

Intrinsic attenuation is an important parameter for reservoirs characterization due to its strong sensitivity to fluid saturation, fractures, and rock texture. Previous studies show that intrinsic attenuation in anisotropic rocks varies with propagation direction, and the attenuation anisotropy is sometimes more significant than the velocity anisotropy. The intrinsic anisotropic attenuation can be described by an attenuation matrix and its elements are the inverse of quality factors (Qij). Here, we focus on frequency-independent intrinsic attenuation and its anisotropy caused by stratified viscoelastic media. We define attenuation anisotropy parameters (AAPs) for transverse isotropic (TI) and orthorhombic attenuation. Based on Backus averaging theory, we derive a unified analytical expression of anisotropic attenuation for different anisotropic layered models. Analyzing the variation in layer-induced anisotropic attenuation assists in guiding the stratified viscoelastic reservoirs characterization.

In 1997 a 3D survey was acquired in the Golfo San Jorge Basin in the center of the Argentinean Patagonia. After thorough analysis it was possible to correlate the poor data areas with intrusive bodies that occur at shallow levels and appear as overlapping lenses of irregular shape. These lenses generate diffraction points that interact with the reflections in a chaotic manner. The zones of interest lay at a significant depth below these intrusives. A new 3D was recorded in the same location as the 1997 survey. Different techniques were used to accomplish these objectives including reprocessing, wave equation modelling, data simulation, increasing trace density and statistical diversity, midpoint scatter, receiver arrays and source arrays.

Field-wide seafloor subsidence measurements are a mature reservoir monitoring technology, utilized in ten hydrocarbon fields in Norway. These measurements provide lateral information on pore compaction and pressure depletion in the reservoir. The survey method uses water pressure measurements at the seafloor as a starting point and reaches accuracies of 2 – 5 mm. Field cases demonstrate that the lateral distribution of subsidence can be used to identify undrained compartments and to calibrate the geomechanical model, hence providing improved interpretation of seismic time-shifts in the overburden.

Shell manages the Ormen Lange field by utilizing a combination of technologies: time-lapse gravity is used to quantify mass changes in the reservoir, providing valuable constraints for dynamic reservoir modelling; seabed subsidence is used as an indirect measurement of compaction throughout the reservoir; and 4D seismic provides the vertical resolution required to interpret these datasets in three dimensions.

In this abstract we present a solution, based on temperature-stabilization, that eliminates temperature-induced effects in subsidence surveys. We show that it provides sub-centimeter accuracy at the Ormen Lange field, with a range of depths extending for almost one kilometer. Finally, we discuss how this data is used and combined with other data types to manage the reservoir.

Broadband high-density surveys have opened the way for the application of full-waveform inversion (FWI) on land. For these surveys, initial FWI results were promising. They demonstrated that land FWI can be an effective tool for velocity model building. However, despite these successes, difficulties remain associated with cycle skipping, and with convergence at high frequencies. Multi-dimensional Optimal Transport (OT) FWI has been shown to offer a solution for the inversion of low frequencies, with a better mitigation of the cycle skipping problem.

We present the results of a workflow, which updates with multi-dimensional OT-FWI to 16 Hz a velocity model for a land dataset from the Sultanate of Oman. The study is based on a modern full-azimuth long-offset dense broadband surface seismic survey. The geological context is a complex strike-slip fault system, causing sharp lateral velocity variations in the faulted area. Until recently, imaging beneath the faulted area was challenging due to the wavefield complexity induced by the lateral velocity variations. Previous attempts at building an accurate velocity model using ray-based tomography failed due to difficult horizon interpretation and residual move-out picking on migrated gathers. We show that using multi-dimensional OT-FWI leads to significant improvement concerning all these issues.