Large Volumes Of Seismic Data Research Articles

First-Break Picking Method Based on the Difference Between Multiwindow Energy Ratios

First-break picking is an important step during processing of both passive and active seismic data. Many automated algorithms have been developed to detect first-break points in large volumes of seismic data. However, it remains difficult to determine precise first-break points in seismograms with low signal-to-noise ratios. Therefore, we present a new approach based on the differences between multi-window energy ratios (DERs) that minimizes the effects of noise. First, the DER is defined and a thresholding method detecting first-break points using the DERs is proposed. Thresholding can be varied depending on the DER parameters, which ensures reliable results even if the parameters change. We use two types of seismic data to establish and verify the DER method: big data derived via global earthquake monitoring (STanford EArthquake Dataset) and ocean bottom cable (OBC) data acquired offshore of Pohang, Republic of Korea. We investigated the effects of parameter changes on the DER picking results. Good picking performance was verified under low signal-to-noise conditions and compared to conventional first-break picking methods. The DER accuracy was higher than that of conventional methods and outliers were rare.

Visual duplicates search of fracture zones of seismic databasesbased on the method of solving the ICP variational problemin closed form and inverted index

In this paper fast methods are proposed for search the fracture zones in seismic databases on two types of data: seismic section (two-dimensional data) and seismic cube (three-dimensional data). These methods are an integral part of the mapping technology for filtering channels and large volumes of seismic data and useful for automating interpretation of heterogeneous seismic data. The proposed methods for searching the similarity of fracture zones were investigated using the Open Seismic Repository reference dataset, which contains information about geological rocks in the area of the North Sea and compared with other known methods for solving this problem, the results were discussed in the article.

Machine learning for automatic slump identification from 3D seismic data at convergent plate margins

Plate subduction zones cause earthquakes and build mountain ranges due to plate collisions, which generate complex structures and induce the down-slope transport of large masses (i.e., slumps). Extensive 3D seismic data have been collected in plate subduction zones and can be used to investigate the slumps associated with earthquakes and hydrocarbon accumulations (i.e., hydrate and gas reservoirs). The development of artificial intelligence has provided new techniques (i.e., association rule learning, decision tree learning, and neural networks) for big-data analysis, such as interpreting large seismic data volumes. Here, we use a convolutional neural network (CNN) to automatically detect complex geological structures, such as slump units. We tested our method on the 3D seismic data acquired in the Nankai subduction zone. After manually defining slump units in several seismic profiles within the 3D data volume, we fed the information to the CNN, which accurately identified the spatial distribution of slump units. The CNN model was trained using real 3D seismic data until it achieved 90 % classification accuracy for slump units in the same region as the model was trained. We further applied our CNN model trained by the Nankai data to the 3D seismic data at another plate convergent margin (Sanriku-Oki in northeast Japan) and succeeded in identifying slump units in the Sanriku-Oki seismic volume. In addition to slump identification, our CNN predicted faults better than well-known methods based on seismic attributes. The slump units identified via CNN are distributed at fewer normal fault zone in the Nankai data. The high accuracy of these automatic interpretations shows that this approach can be applied to other forearc basins to investigate geological structures (e.g., slump and faults) at high spatial resolution.

Automatic Detection and Location of Microseismic Events from Sparse Network and Its Application to Post-mining Monitoring

In post-mining regions with seismic hazard, timely decision making for risk management faces the challenge of quick and reliable detection and location of seismic events. As a response to the increasing density of monitoring stations, generating large volumes of seismic data, automatic, full waveform-based methods have been developed in recent years in global seismology. Such methods often cannot be directly applied to post-mining monitoring with a limited station coverage, as it is the case when temporarily networks are installed as an emergency response. In this paper we propose a new methodology that bridges this gap and enables the application of a full waveform, backprojection based method (BackTrackBB) to data of sparse network. The methodology was successfully tested on an abandoned and flooded underground coalmine in South-eastern France. Steps preceding BackTrackBB application were implemented in order to remove the coherent noise that otherwise results in numerous false detections. First results indicate that seismic activity in the study area is controlled by water level variation within former room-and-pillar mine works and fault segments (re)activation below them.

Time-lapse seismic history matching with an iterative ensemble smoother and deep convolutional autoencoder

We have developed a time-lapse seismic history matching framework to assimilate production data and time-lapse seismic data for the prediction of static reservoir models. An iterative data assimilation method, the ensemble smoother with multiple data assimilation is adopted to iteratively update an ensemble of reservoir models until their predicted observations match the actual production and seismic measurements and to quantify the model uncertainty of the posterior reservoir models. To address computational and numerical challenges when applying ensemble-based optimization methods on large seismic data volumes, we develop a deep representation learning method, namely, the deep convolutional autoencoder. Such a method is used to reduce the data dimensionality by sparsely and approximately representing the seismic data with a set of hidden features to capture the nonlinear and spatial correlations in the data space. Instead of using the entire seismic data set, which would require an extremely large number of models, the ensemble of reservoir models is iteratively updated by conditioning the reservoir realizations on the production data and the low-dimensional hidden features extracted from the seismic measurements. We test our methodology on two synthetic data sets: a simplified 2D reservoir used for method validation and a 3D application with multiple channelized reservoirs. The results indicate that the deep convolutional autoencoder is extremely efficient in sparsely representing the seismic data and that the reservoir models can be accurately updated according to production data and the reparameterized time-lapse seismic data.

Fast assimilation of frequently acquired 4D seismic data for reservoir history matching

The study in this paper proposed a new framework for history matching of frequently acquired 4D seismic data. To achieve this goal, the large volumes of seismic data from the many repeated 4D monitors are firstly condensed into a single attribute by directly correlating them to the reservoir production and injection performances. This ‘well2seis’ cross-correlation is achieved by defining a linear relationship between pressure and saturation-related 4D seismic responses and the cumulative changes of reservoir fluid volumes derived from wells. It is shown that such a cross-disciplinary attribute not only reduces the amount of 4D seismic data for history matching, but also enhances the seismic data reliability since the 4D seismic observations are conditioned by low-uncertainty production data from reservoir engineering domain. In the second step, Morris sensitivity analysis is adapted to fast diagnose the uncertainty reservoir model parameters that are sensitive to the well2seis attributes. To quantitatively assimilate the well2seis observations to calibrate these uncertainty parameters, we then proposed a well2seis objective function that quantifies the mismatch between the observed and model simulated well2seis attributes. Ensemble Smoother with Multiple Data Assimilation (ES-MDA) is performed at the end to iteratively assimilate the well2seis observations by minimizing the well2seis misfit. Application of the proposed workflow to a North Sea field case shows that, when history matching to the observed well2seis attribute that honours the information from seismic and reservoir engineering domains, it can significantly reduce the uncertainty of key reservoir parameters, hence improving the matching quality to both 4D seismic and production observations and enhancing the prediction reliability of the reservoir models. Compared to traditional history matching approaches that attempt to match individual seismic time-lapse attributes and production observations, this approach is observed to significantly boost the history matching efficiency by reducing the number of time-consuming iterations.

A scalable deep learning platform for identifying geologic features from seismic attributes

The modern requirement for analyzing and interpreting ever-larger volumes of seismic data to identify prospective hydrocarbon prospects within stringent time deadlines represents an ongoing challenge in petroleum exploration. To provide a computer-based aid in addressing this challenge, we have developed a “big data” platform to facilitate the work of geophysicists in interpreting and analyzing large volumes of seismic data with scalable performance. We have constructed this platform on a modern distributed-memory infrastructure, providing a customized seismic analytics software development toolkit, and a Web-based graphical workflow interface along with a remote 3D visualization capability. These support the management of seismic data volumes, attributes processing, seismic analytics model development, workflow execution, and 3D volume visualization on a scalable, distributed computing platform. Early experiences show that computationally demanding deep learning methods such as convolutional neural networks (CNN) provide improved results over traditional methods such as support vector machines (SVMs) and logistic regression for identifying geologic faults in 3D seismic volumes. Our experiments show encouraging accuracy in identifying faults by combining CNN and traditional machine learning models with a variety of seismic attributes, and the platform is able to deliver scalable performance.

Bayesian survey design to optimize resolution in waveform inversion

We describe a Bayesian methodology for designing seismic experiments that optimally maximize model-parameter resolution for imaging purposes. The proposed optimal experiment design algorithm finds the measurements that are likely to optimally reduce the expected uncertainty on the model parameters. This Bayesian [Formula: see text]-optimality-based algorithm minimizes the volume of the expected confidence ellipsoid and leads to the maximization of the expected resolution of the model parameters. Computational efficiency is achieved by a greedy algorithm in which the design is sequentially improved. In contrast to minimizing the uncertainty volume over the entire subsurface simultaneously, a refinement of the algorithm minimizes the marginal uncertainties in a region of interest. Minimizing marginal uncertainties simultaneously accounts for quantitative prior model uncertainties while honoring a qualitative focus on particular regions of interest. The benefits of the proposed method over traditional non-Bayesian ones are demonstrated with several geophysical examples. These include reducing large seismic data volumes for real-time imaging and solving the problem of designing seismic surveys that account for source bandwidth, signal-to-noise ratio, and attenuation.

Data mining for large scale 3D seismic data analysis

The automation of the analysis of large volumes of seismic data is a data mining problem, where a large database of 3D images is searched by content, for the identification of the regions that are of most interest to the oil industry. In this paper we perform this search using the 3D orientation histogram as a texture analysis tool to represent and identify regions within the data, which are compatible with a query texture.