Traditional full-waveform inversion is a non-linear and ill-posed inversion problem. To reduce the non-linearity of it, joint migration inversion (joint migration inversion) was proposed as an alternative. Joint migration inversion tries to minimize the mismatch between measured and modelled reflection data. One key feature of joint migration inversion is its parameterization: two separate parameters, reflectivity (for the amplitudes of reflected events) and propagation velocity (for the phase effects). This separation helps to reduce the non-linearity of the inversion. During joint migration inversion, with the velocity being updated, the reflectors in the updated image are also shifting in depth accordingly, this phenomenon is called depth–velocity ambiguity. This interaction between the two parameters during inversion is desired to keep the image time consistent with the measured data but may lead to non-robustness of joint migration inversion due to the presence of local minima. Therefore, we propose a more robust joint migration inversion scheme, which parameterizes the models with vertical time, termed pseudo-time joint migration inversion. In pseudo-time, the updates of velocity will not result in the associated vertical location changes of reflectors in the estimated image. Instead, the reflectors are mainly getting more focused. One limitation is that the depth-pseudo-time conversion process assumes a simple linear relationship between depth and pseudo-time, which might cause some artefacts in the converted models when there exist strong lateral velocity variations. One subsequent round of depth joint migration inversion is recommended to resolve this issue. We demonstrate the effectiveness of our proposed method with a two-dimensional synthetic example in an extreme scenario, where the initial velocity model is homogeneous, a realistic offshore two-dimensional synthetic example, a two-dimensional field example from the Vøring basin in Norway and a simple three-dimensional synthetic example. In all examples, pseudo-time joint migration inversion manages to recover more reasonable updates in the inverted velocity and invert more focused reflectors in the inverted image, compared to depth joint migration inversion.

An important imaging challenge is creating reliable seismic images without internal multiple crosstalk, especially in cases with strong overburden reflectivity. Several data-driven methods have been proposed to attenuate the internal multiple crosstalk, for which fully sampled data in the source and receiver side are usually required. To overcome this acquisition constraint, model-driven full-wavefield migration (FWM) can automatically include internal multiples and only needs dense sampling in either the source or receiver side. In addition, FWM can correct for transmission effects at the reflecting interfaces. Although FWM has been shown to work effectively in compensating for transmission effects and suppressing internal multiple crosstalk compared to conventional least-squares primary wavefield migration (PWM), it tends to generate relatively weaker internal multiples during modeling. Therefore, some leaked internal multiple crosstalk can still be observed in the FWM image, which tends to blend in the background and can be misinterpreted as real geology. Thus, we adopted a novel framework using local primary-and-multiple orthogonalization (LPMO) on the FWM image as a postprocessing step for leaked internal multiple crosstalk estimation and attenuation. Due to their opposite correlation with the FWM image, a positive-only LPMO weight can be used to estimate the leaked internal multiple crosstalk, whereas a negative-only LPMO weight indicates the transmission effects that need to be retained. Application to North Sea field data validates the performance of the proposed framework for removing the weak but misleading leaked internal multiple crosstalk in the FWM image. Therefore, with this new framework, FWM can provide a reliable solution to the long-standing issue of imaging primaries and internal multiples automatically, with proper primary restoration.

Nowadays, to obtain a better understanding of dynamic time-lapse changes, frequent seismic monitoring is necessary, although it will generate a considerable cost increase. Therefore, low-cost frequent monitoring, e.g., sparse and/or nonrepeated surveys, is desired. The simultaneous inversion-based method allows the baseline and monitor parameters to communicate and compensate with each other during inversion via constraints and helps to reduce the artifacts caused by sparse acquisition. These features make it largely independent of the used low-cost acquisition geometry and suitable for inexpensive frequent monitoring surveys. Therefore, we have used this simultaneous inversion-based method as an effective time-lapse processing tool for data sets acquired from inexpensive, semi-continuous time-lapse monitoring surveys, which are based on the so-called instantaneous 4D (i4D) technology. We choose a specific simultaneous inversion method called simultaneous joint migration inversion (S-JMI), which combines a simultaneous processing strategy with the JMI method. In i4D technology, inexpensive localized/sparse surveys, called i4D surveys, are deployed frequently between the conventional full-field surveys. This technology can be treated as a special case of changing geometries during monitoring. In this case, the simultaneous strategy allows the information of the full-field survey to compensate for the insufficient illumination of the localized/sparse i4D surveys during processing. Furthermore, we apply constraints on the reflectivity and velocity differences between the baseline and monitor vintages along the calendar-time axis called calendar-time constraints. These constraints take advantage of the feature that time-lapse effects develop (semi-)continuously along the calendar-time axis, when the monitoring surveys are deployed (semi-)continuously over calendar time. Based on a complex synthetic example, we determined that S-JMI is a promising tool to process the data sets from the semi-continuous monitoring surveys based on i4D technology. Finally, we found that the calendar-time constraints significantly improve the quality of time-lapse effects.

Prestack seismic inversion has emerged as a powerful technique for reconstructing parameters attribute to the subsurface properties and building the geophysical parameter models. However, the inversion algorithms always suffer from spatial blur and low resolution. Total variation (TV) regularization preserves the spatial variation boundary of data by highlighting the sparsity of the first-order difference, which is regarded as an important technical means for image restoration. However, when the data do not change along the spatial grid direction, TV regularization is prone to a staircase effect. In this article, a directional TV (DTV) method is proposed to conduct the prestack amplitude variation with offset/angle (AVO/AVA) inversion. The method consists of three essential steps: estimating the seismic slope attribute from the seismic data, introducing seismic slope attribute to the TV regularization to establish the objective function, and optimizing the objective function by the split-Bregman algorithm. Finally, the conventional and proposed methods are applied to the synthetic and the real seismic data. The comparison of different methods demonstrates that the proposed method is applicable to reveal the detailed subsurface models, alleviate the staircase effect or artifact substantially, and further upgrade the quality of prestack inversion results.

Accurate removal of surface-related multiples remains a challenge in shallow-water cases. One reason is that the success of surface-related multiple estimation (SRME)-related algorithms is sensitive to the quality of the near-offset reconstruction. When it comes to a larger missing gap and a shallower water bottom, the state-of-the-art near-offset gap construction method - the parabolic Radon transform - fails to provide reliable recovery of the shallow reflections due to the limited information from the data and highly curved events at near offsets with strong lateral amplitude variations. Therefore, we have developed a novel workflow that first deploys a deep-learning-based reconstruction of the shallow reflections and then uses the reconstructed data as the input for the subsequent surface multiple removal. In particular, we use a convolutional neural network architecture - U-net that was developed from convolutional autoencoders with extra direct skip connections between different levels of encoders and the corresponding decoders. Instead of using field data directly in network training, the training set is carefully synthesized based on the prior water-layer information of the field data; thus, a fully sampled field data set, which is difficult to obtain, is not needed for training in our workflow. An inversion-based approach - closed-loop SRME - is used for the surface multiple removal, in which the primaries are directly estimated via full-waveform inversion in a data-driven manner. Finally, the effectiveness of our workflow is determined based on 2D North Sea field data in a shallow-water scenario (92.5 m water depth) with a relatively large minimum offset (150 m).

The conventional time-lapse processing workflow is usually sensitive to the non-repeatable uncertainties between different vintages caused by noise, acquisition designs and independent processing. Therefore, in order to reduce these non-repeatable uncertainties, all the datasets are usually acquired from well-sampled and well-repeated acquisition surveys, and the independent processing is always carefully tailored to maximally reduce the non-repeatable uncertainties during processing. Moreover, the conventional time-lapse analysis method, based on a time-shift map, is not always a good indicator of the actual velocity differences due to its local one-dimensional subsurface assumption. In order to relax these rigid requirements and have a better velocity change indicator, a robust high-resolution simultaneous joint migration inversion was proposed as an effective time-lapse tool for reservoir monitoring. The method combines a simultaneous data-processing strategy with the joint migration inversion method. Joint migration inversion is a full wavefield inversion method with a parameterization in terms of reflectivity and propagation velocity. The simultaneous strategy allows the baseline and monitor parameters to communicate and compensate with each other dynamically during inversion, thus, suppressing the non-repeatable uncertainties during the time-lapse processing. To investigate the feasibility of using high-resolution simultaneous joint migration inversion in practice, some numerical experiments are conducted to test the dependence of high-resolution simultaneous joint migration inversion on the quality of the time-lapse datasets including the following aspects: random noise; sparse surveys; ocean bottom node versus streamer (different types of monitoring surveys); non-repeated sources, including source positioning errors and non-repeated source wavelets; spatial weighting operators in the L2-norm penalty terms; and sensitivity to weak time-lapse effects. All the experiments are carried on the basis of a realistic synthetic time-lapse model based on the Grane field offshore Norway. These experiments show that high-resolution simultaneous joint migration inversion is robust to random noise, survey sparsity, survey non-repeatability, source positioning errors and source wavelet discrepancies. Furthermore, high-resolution simultaneous joint migration inversion remains effective when the spatial weighting operators in the L2-norm penalty terms are largely relaxed, and high-resolution simultaneous joint migration inversion is capable of detecting weak time-lapse changes (e.g. velocity changes down to (Formula presented.) 35 m/s).

Joint migration inversion is a recently proposed technology, accommodating velocity model building and seismic migration in one integrated process. Different from the widely accepted full waveform inversion technology, it uses imaging parameters, i.e. velocities and reflectivities of the subsurface, to parameterize its solution space. The unique feature of this new technology is its explicit capability to exploit multiples in its inversion scheme, which are treated as noise by most current technologies. In this paper, we comprehensively evaluate the state-of-the-art joint migration inversion technology from various angles: we first benchmark its performance, on both velocity model building and seismic imaging, against that of the well-accepted workflow comprising full waveform inversion and reverse-time migration using a fully controlled 2D realistic synthetic dataset. Next, we demonstrate its application on a 2D field dataset. Last, we use another 2D synthetic dataset to clearly illustrate the challenges the current joint migration inversion technology is facing. With this paper, we transparently reveal the pros of cons of the current joint migration inversion, and we will also point out the imminent research directions joint migration inversion technology should focus on in the next phase for it to be more widely accepted by the geophysics community.

Surface-related multiple elimination (SRME) is a solid and effective approach for primary estimation. However, due to the imperfections in data and method multiple energy leakage is commonly seen in the results of SRME-predicted primaries. Assuming that the primaries and multiples do not correlate locally in the time-space domain, we are able to extract the leaked multiples from the initially estimated primaries using multi-domain local primary-and-multiple orthogonalization. The proposed framework consists of two steps: an initial primary/multiple estimation step and a multiple-leakage extraction step. The initial step corresponds to SRME, which produces the initial estimated primary and multiple models. The second step is based on multi-domain local primary-and-multiple orthogonalization to retrieve the leaked multiples. Multi-domain indicates that we first extract the leaked multiples in shot domain, and then the residual can be further extracted in common-offset domain. Thus, we can obtain a better primary output which has much less leaked multiple energy. We demonstrate a good performance of our proposed framework on both synthetic and field data, where it repairs the leakage of standard global adaptive subtraction.

Microseismic methods are crucial for real-Time monitoring of the hydraulic fracturing dynamic status during the development of unconventional reservoirs. However, unlike the active-source seismic events, the microseismic events usually have low signal-To-noise ratio (SNR), which makes its data processing challenging. To overcome the noise issue of the weak microseismic events, we propose a new workflow for high-resolution microseismic event detection. For the preprocessing, fix-sized segmentation with a length of 2∗wavelength is used to divide the data into segments. Later on, 191 features have been extracted and used as the input data to train the support vector machine (SVM) model. These features include 63 1-D time/spectral-domain features, and 128 2-D texture features, which indicate the continuity, smoothness, and irregularity of the events/noise. The proposed feature extraction maximally exploits the limited information of each segment. Afterward, we use a combination of univariate feature selection and random-forest-based recursive feature elimination for feature selection to avoid overfitting. This feature selection strategy not only finds the best features, but also decides the optimal number of features that are needed for the best accuracy. Regarding the training process, SVM with a Gaussian kernel is used. In addition, a cross-validation (CV) process is implemented for automatic parameter setting. In the end, a group of synthetic and field microseismic data with different levels of complexity show that the proposed workflow is much more robust than the state-of-The-Art short-Term-Average over long-Term-Average ratio (STA/LTA) method and also performs better than the convolutional-neural-networks (CNN), for this case where the amount of training data sets is limited. A demo for the synthetic example is available: https://github.com/shanqu91/ML_event_detection_microseismic.

Simultaneous source technology can accelerate data acquisition and improve subsurface illumination. But those advantages are compromised due to dense interference. To address the intense interference in simultaneous source data, we have investigated a method based on a deep neural network. The designed architecture consists of convolutional and deconvolutional networks. The convolutional network can learn the local features of the training data set, and the deconvolutional network constructs the output using the extracted features to match the ground truth. Because the main computational cost results from the optimization of the network parameters, the trained network can separate simultaneous source data efficiently. Besides, with the given dithering code, we embed the trained network into an iterative framework that can further improve the deblending. A numerical test on synthetic data demonstrates that the iterative framework with the trained network can obtain comparable performance with high efficiency compared to the conventional method. Next, we test our method with two different trained networks (one is from a synthetic data set, and the other is from a field data set) on field data. The test results confirm the performance of our method.

Accurate multiple removal remains an important step in seismic data processing sequences. Most multiple removal methods, such as surface-related multiple elimination (SRME), consist of a multiple prediction step and an adaptive subtraction step. Due to imperfect circumstances (e.g., coarse data sampling) or built-in assumptions (e.g., 2D method versus 3D data), multiple leakage is commonly observed in the results. More aggressive adaptive multiple subtraction can reduce the leakage problem, for example, by using small local windows and a long filter length, but at the risk of severely damaging the primaries due to overfitting. In contrast, conservative adaptive subtraction with large or global windows and a short filter length can preserve most primary energy while tending to have more multiple leakage because of underfitting. Assuming that the primaries and multiples do not correlate locally in the time-space domain, our solution to this problem is to extract the leaked multiples from the initially estimated primaries using local primary-and-multiple orthogonalization (LPMO) rather than restoring the damaged primaries. Our framework consists of two steps: an initial primary estimation step and a multiple leakage extraction step. The initial step corresponds to conservative SRME (or an equivalent method) that produces the initially estimated primary and multiple models. The second step is based on LPMO to retrieve the leaked multiples from the estimated primaries via a time- and space-varying weight function that is estimated from the local correlation of predicted multiples and residual multiples in the estimated primaries with the help of shaping regularization. In this way, we can obtain a better primary model that has much less leaked multiple energy and less primary damage at the same time. We find good performance of our framework via two synthetic data examples and one field data example.

High-resolution velocity models of the subsurface in depth domain are increasingly obtained from seismic data by Full Waveform Inversion (FWI), which has become the leading-edge technology in this domain. FWI applications, however, often fail if severely band-limited seismic data lacks low-frequency information that determines the general trend of the subsurface velocity. The inversion then produces cycle skipping by adapting incompatible phases of the synthetic and measured data. In order to avoid this FWI failure, the general velocity trend may be obtained as starting models by alternative methods from severely band-limited seismic data. The Joint Migration Inversion (JMI) provides a better reconstruction of the velocity trend in the starting model, thus driving the FWI results much closer to the actual subsurface velocity. This is demonstrated for synthetic 2D seismic data from a realistic subsurface model obtained in marine gas-hydrate studies.

Simultaneous Joint Migration and Inversion (S-JMI) is an effective time-lapse tool for reservoir monitoring, which combines a joint time-lapse data processing strategy with the Joint Migration Inversion (JMI) method. In S-JMI, fine details are not expected in its inverted velocity model, as it only explains the propagation effects in the data, while the scattering effects are explained by the inverted reflectivity model. However, for time-lapse processing, high-resolution time-lapse velocity differences are usually a demand. Therefore, in order to get more localized time-lapse velocity differences, we propose a robust high-resolution S-JMI process by using the time-lapse reflectivity-difference as an extra constraint during S-JMI. This constraint makes a link between the reflectivity- and the velocity-difference by exploiting the relationship between them, which can also be explained as a constraint on density. In the end, with a highly realistic synthetic model based on the Grane field from offshore Norway, we demonstrate a more localized inverted time-lapse velocity difference using the proposed robust S-JMI, compared to the regular SJMI.

The current time-lapse practice is to exactly repeat well-sampled acquisition geometries to mitigate acquisition effects on the time-lapse differences. In order to relax the rigid requirements on acquisition effects, we propose simultaneous joint migration inversion as an effective time-lapse tool for reservoir monitoring, which combines a joint time-lapse data processing strategy with the joint migration inversion method. Joint migration inversion is a full-wavefield inversion method that explains the measured reflection data using a parameterization in terms of reflectivity and propagation velocity. Both the inversion process inside the imaging/inversion scheme and the extra illumination provided by including multiples in joint migration inversion makes the obtained velocity and reflectivity operator largely independent of the utilized acquisition geometry and, thereby, relaxes the strong requirement of non-repeatability during the monitoring. Because simultaneous joint migration inversion inverts for all datasets simultaneously and utilizes various constraints on the estimated reflectivities and velocity, the obtained time-lapse differences have much higher accuracy compared to inverting each dataset separately. It allows the baseline and monitor parameters to communicate with each other dynamically during inversion via a user-defined spatial weighting operator. In order to get more localized time-lapse velocity differences, we further extend the regular simultaneous joint migration inversion to a robust high-resolution simultaneous joint migration inversion process using the time-lapse reflectivity difference as an extra constraint for the velocity estimation during inversion. This constraint makes a link between the reflectivity- and the velocity difference by exploiting the relationship between them. We demonstrate the feasibility of the proposed method with a highly realistic synthetic model based on the Grane field offshore Norway and a time-lapse field dataset from the Troll Field.

Although joint migration inversion has been proposed for several years, a thorough derivation and description of the involved gradients was not published. In this paper, we derive the gradient of both the angle-independent reflectivity and the velocity in a framework of acoustic angle-independent joint migration inversion. With some further approximations taken, the conclusions shown in previous publications can also be reached from our new derivation.

Microseismic methods are crucial for real-timemonitoring of the hydraulic fracturing dynamic status during the development of unconventional reservoirs. However, unlike the active-source seismic events, the microseismic events usually have low signal-to-noise ratio (SNR), which makes its data processing challenging. To overcome the noise issue of the weak microseismic events, we propose a new workflow for high-resolution microseismic event detection. For the preprocessing, fix-sized segmentation with a length of 2.wavelength is used to divide the data into segments. Later on, 191 features have been extracted and used as the input data to train the support vector machine (SVM) model. These features include 63 1-D time/spectraldomain features, and 128 2-D texture features, which indicate the continuity, smoothness, and irregularity of the events/noise. The proposed feature extractionmaximally exploits the limited information of each segment. Afterward, we use a combination of univariate feature selection and random-forest-based recursive feature elimination for feature selection to avoid overfitting. This feature selection strategy not only finds the best features, but also decides the optimal number of features that are needed for the best accuracy. Regarding the training process, SVM with a Gaussian kernel is used. In addition, a cross-validation (CV) process is implemented for automatic parameter setting. In the end, a group of synthetic and field microseismic data with different levels of complexity show that the proposed workflow is much more robust than the state-of-the-art short-term-average over long-term-average ratio (STA/LTA) method and also performs better than the convolutional-neural-networks (CNN), for this case where the amount of training data sets is limited. A demo for the synthetic example is available: https://github.com/shanqu91/ML event detection microseismic.

Joint migration inversion (JMI) is a recently developed technology that aims at incorporating the seismic velocity model update and acoustic migration, including all multiple scatterings, into one closed-loop process. Full waveform inversion (FWI) is a commonly accepted technology for velocity model building, and reverse-time migration (RTM) is the main method adopted for depth imaging. In this paper we first use a 2D realistic deep water model to benchmark JMI against a workflow of FWI combined with RTM. With some insights on JMI gained from this comparison study, we point out that a good niche for JMI is to provide a high quality initial velocity model. We further demonstrate the performance of JMI on a 2D deep water field dataset with an initial velocity model derived from migration velocity analysis (MVA).