Latent space modeling of seismic data: An overview

Wallet, Bradley C.; Matos, Marcílio Castro de; Kwiatkowski, Jacek; Suarez, Yoscel

doi:10.1190/1.3272700

Cited by 34 publications

(13 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although the SOM has the same relatively high dimension of the input data, by construction SOM preserves the adjacent relationship among each SOM quantized vector (Matos et al, 2007). In this manner, SOM can be interpreted as a mapping of seismic attributes residing in r-dimension space onto a 1D, 2D, or 3D latent-space that preserves the original topological structure of the seismic amplitude data (Wallet et al, 2009). In this paper, we assume that the input variables to the SOM are the GLCM attributes, while the resulting 2D SOM is mapped against a 2D color bar.…”

Section: Self-organizing Mapsmentioning

confidence: 99%

“…Such texture attributes hold significant promise in quantifying geological features such as mass transport complexes, amalgamated channels, and dewatering features that exhibit a distinct lateral pattern beyond simple edges. Like seismic waveform classification (Coléou et al, 2003), and spectral components, GLCM attributes are amenable to subsequent clustering analysis using selforganizing and generative-topographic maps (Angelo et al, 2009;West et al, 2002;Gao, 2007;Wallet et al, 2009).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Integrated seismic texture segmentation and cluster analysis applied to channel delineation and chert reservoir characterization

et al. 2011

View full text Add to dashboard Cite

In recent years, 3D volumetric attributes have gained wide acceptance by seismic interpreters. The early introduction of the single-trace complex trace attribute was quickly followed by seismic sequence attribute mapping workflows. Three-dimensional geometric attributes such as coherence and curvature are also widely used. Most of these attributes correspond to very simple, easy-to-understand measures of a waveform or surface morphology. However, not all geologic features can be so easily quantified. For this reason, simple statistical measures of the seismic waveform such as rms amplitude and texture analysis techniques prove to be quite valuable in delineating more chaotic stratigraphy. In this paper, we coupled structure-oriented texture analysis based on the gray-level co-occurrence matrix with self-organizing maps clustering technology and applied it to classify seismic textures. By this way, we expect that our workflow should be more sensitive to lateral changes, rather than vertical changes, in reflectivity. We applied the methodology to a remote sensing image and to a 3D seismic survey acquired over Osage County, Oklahoma, USA. Our results indicate that our method can be used to delineate meandering channels as well as to characterize chert reservoirs.

show abstract

Section: Self-organizing Mapsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Integrated seismic texture segmentation and cluster analysis applied to channel delineation and chert reservoir characterization

et al. 2011

View full text Add to dashboard Cite

show abstract

“…Liu and Luo (2007) and Ni (2008) applied LLE and Isomap 1D seismic attribute optimization, respectively. While Wallet et al (2009) proposed using manifold learning to model the lower-dimension manifold characteristics of a latent space embedded in attribute space and they pointed out that manifold learning is extremely computationally demanding both in terms of processing and memory but they didn't put it into practice. The previous research only tried to apply manifold learning to single traces but not to 3D seismic attribute dimensionality reduction and didn't deeply discuss the selection of its key parameters (such as the number of neighbors).…”

Section: Introductionmentioning

confidence: 99%

Locally linear embedding-based seismic attribute extraction and applications

Liu

Zheng

et al. 2010

Appl. Geophys.

View full text Add to dashboard Cite

How to extract optimal composite attributes from a variety of conventional seismic attributes to detect reservoir features is a reservoir predication key, which is usually solved by reducing dimensionality. Principle component analysis (PCA) is the most widelyused linear dimensionality reduction method at present. However, the relationships between seismic attributes and reservoir features are non-linear, so seismic attribute dimensionality reduction based on linear transforms can't solve non-linear problems well, reducing reservoir prediction precision. As a new non-linear learning method, manifold learning supplies a new method for seismic attribute analysis. It can discover the intrinsic features and rules hidden in the data by computing low-dimensional, neighborhood-preserving embeddings of highdimensional inputs. In this paper, we try to extract seismic attributes using locally linear embedding (LLE), realizing inter-horizon attributes dimensionality reduction of 3D seismic data first and discuss the optimization of its key parameters. Combining model analysis and case studies, we compare the dimensionality reduction and clustering effects of LLE and PCA, both of which indicate that LLE can retain the intrinsic structure of the inputs. The composite attributes and clustering results based on LLE better characterize the distribution of sedimentary facies, reservoir, and even reservoir fluids.

show abstract

“…GTM finds a model based on a probability density function (PDF) that describes the distribution of the D-dimensional data in terms of smaller number of latent variables, or cluster nodes that approximate the vast majority of the probability mass of the data (Svensen, 1998). Wallet et al (2009) are probably the first to apply the GTM technique to seismic data, using a suite of phantom horizon slices through a seismic amplitude volume generating a "waveform classification." After iterative parameter estimation, the model fits the data, and we relate the points in the D-dimensional data space to grid points or nodes in the lower L-dimensional latent space.…”

Section: Introductionmentioning

confidence: 99%

Generative topographic mapping for seismic facies estimation of a carbonate wash, Veracruz Basin, southern Mexico

et al. 2014

View full text Add to dashboard Cite

Seismic facies estimation is a critical component in understanding the stratigraphy and lithology of hydrocarbon reservoirs. With the adoption of 3D technology and increasing survey size, manual techniques of facies classification have become increasingly time consuming. Besides, the numbers of seismic attributes have increased dramatically, providing increasingly accurate measurements of reflector morphology. However, these seismic attributes add multiple "dimensions" to the data greatly expanding the amount of data to be analyzed. Principal component analysis and self-organizing maps (SOMs) are popular techniques to reduce such dimensionality by projecting the data onto a lower order space in which clusters can be more readily identified and interpreted. After dimensional reduction, popular classification algorithms such as neural net, K-means, and Kohonen SOMs are routinely done for general well log prediction or analysis and seismic facies modeling. Although these clustering methods have been successful in many hydrocarbon exploration projects, they have some inherent limitations. We explored one of the recent techniques known as generative topographic mapping (GTM), which takes care of the shortcomings of Kohonen SOMs and helps in data classification. We applied GTM to perform multiattribute seismic facies classification of a carbonate conglomerate oil field in the Veracruz Basin of southern Mexico. The presence of conglomerate carbonates makes the reservoir units laterally and vertically highly heterogeneous, which are observed at well logs, core slabs, and thin section scales. We applied unsupervised GTM classification to determine the "natural" clusters in the data set. Finally, we introduced supervision into GTM and calculated the probability of occurrence of seismic facies seen at the wells over the reservoir units. In this manner, we were able to assign a level of confidence (or risk) to encountering facies that corresponded to good and poor production. 2 PEMEX Exploración y Producción, Veracruz-Llave, Mexico.

show abstract

Latent space modeling of seismic data: An overview

Cited by 34 publications

References 0 publications

Integrated seismic texture segmentation and cluster analysis applied to channel delineation and chert reservoir characterization

Integrated seismic texture segmentation and cluster analysis applied to channel delineation and chert reservoir characterization

Locally linear embedding-based seismic attribute extraction and applications

Generative topographic mapping for seismic facies estimation of a carbonate wash, Veracruz Basin, southern Mexico

Contact Info

Product

Resources

About