Big data analytics for seismic fracture identification using amplitude-based statistics
- 11 Downloads
Present-day innovations in seismic acquisition tools and techniques have enabled the acquisition of detailed seismic datasets, which in many cases are extremely large (on the order of terabytes to petabytes). However, data analysis tools for extracting information on critical subsurface features such as fractures are still evolving. Traditional methods rely on time-consuming iterative workflows, which involve computing seismic attributes, de-noising, and expert interpretation. Additionally, with the increasingly widespread acquisition of time-lapse seismic surveys (4D), there is a heightened demand for reliable automated workflows to assist feature interpretation from seismic data. We present a novel data-driven tool for fast fracture identification in BIG post-stack seismic datasets, motivated by techniques developed for real-time face detection. The proposed algorithm computes spatio-temporal amplitude statistics using Haar-like bases, in order to characterize the seismic amplitude properties that correspond to fracture occurrence in a unit window or voxel. Under this approach, the amplitude data is decomposed into a collection of simple-to-calculate “mini-attributes,” which carry information on the amplitude gradient and curvature characteristics at varying locations and scales. These features then serve as inputs to a cascade of boosted classification tree models, which select and combine the most discriminative features to develop a probabilistic binary classification model. This overall approach helps to eliminate the computationally intensive and subjective use of ad hoc seismic attributes in existing approaches. We first demonstrate the viability of the proposed methodology for identifying discrete macro-fractures in a 2D synthetic seismic dataset. Next, we validate the approach using 3D post-stack seismic data from the Niobrara Shale interval within the Teapot Dome field. We show the applicability of the proposed framework for identifying sub-seismic fractures, by considering the amplitude profile adjacent to interpreted formation microimager (FMI) well log data. The upscaled spatial distribution of the predicted fractures shows agreement with existing geological studies and aligns with interpreted large-scale faults within the interval of interest.
KeywordsBIG data Fractured reservoirs Data mining
voting weight applied to weak classifier in AdaBoost model at iteration m
split nodes in classification tree model
minimum true positive ratio per stage in cascade training
real-valued AdaBoost classifier summation
maximum false positive ratio per stage in cascade training
AdaBoost classifier categorical prediction
weak classification model fitted at iteration m
number of terminal nodes in classification tree model
Haar-like feature geometry index
AdaBoost iteration index
total number of AdaBoost iterations
number of training windows
total number of Haar-like feature combinations
Haar-like feature score computed on window i based on unique feature combination j
height of Haar-like feature in time direction
number of windows used in cascaded classifier model evaluation
inline (or common depth point) index
width of Haar-like feature in inline (or common depth point) direction
length of Haar-like feature in crossline direction
true designation of window i
AdaBoost classification threshold
Unable to display preview. Download preview PDF.
The Niobrara fracture density log interpretations have been performed by Dr. Ahmed Ouenes (FracGeo) and his team. We are grateful for their support in making this data available for this study.
This research was supported by National Science Foundation Grant 1546553.
- 1.Aminzadeh, F.: A new concept for seismic anomaly detection. In: Offshore Technology Conference. Offshore Technology Conference, Houston (2005)Google Scholar
- 4.Chopra, S., Marfurt, K.J.: Volumetric curvature attributes for fault/fracture characterization. First Break 25(7), 35–46 (2007)Google Scholar
- 5.Cooper, S.P.: Deformation within a Basement-Cored Anticline: Teapot Dome, Wyoming. Master’s thesis, New Mexico Tech (2000)Google Scholar
- 8.Freund, Y., Schapire, R.: Experiments with a new boosting algorithm. In: International Conference on Machine Learning (1996)Google Scholar
- 11.Guitton, A., Wang, H., Trainor-Guitton, W.: Statistical imaging of faults in 3D seismic volumes using a machine learning approach. In: Annual International Meeting, SEG (2017)Google Scholar
- 13.Hastie, T., Tibshirani, R., Friedman, J.: The Elements of Statistical Learning. Data Mining, Inference, and Prediction. Springer Science & Business Media, New York (2013)Google Scholar
- 16.Ouenes, A.: Seismically driven characterization of unconventional shale plays. Canadian Soc. Explor. Geophys. Recorder 37, 22–28 (2012)Google Scholar
- 18.Schwartz, B.C.: Fracture Pattern Characterization of the Tensleep Formation, Teapot Dome, Wyoming. Master’s thesis, West Virginia University (2006)Google Scholar
- 19.SEG Wiki: Teapot dome 3D survey (2016)Google Scholar
- 21.Viola, P., Jones, M.: Rapid object detection using a boosted cascade of simple features. In: Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Mitsubishi Electric Research Laboratories, Cambridge (2001)Google Scholar
Open AccessSpringer Nature terms of reuse for archived author accepted manuscripts (AAMs) of subscription articles For articles published within the Springer Nature group of companies that have been archived into academic repositories such as institutional repositories, PubMed Central and its mirror sites, where a Springer Nature company holds copyright, or an exclusive license to publish, users may view, print, copy, download and text and data-mine the content, for the purposes of academic research, subject always to the full conditions of use. The conditions of use are not intended to override, should any national law grant further rights to any user.Conditions of use Articles, books and chapters published within the Springer Nature group of companies which are made available through academic repositories remain subject to copyright. The following restrictions on use of such articles, books and chapters apply: Academic research only 1. Archived content may only be used for academic research. Any content downloaded for text based experiments should be destroyed when the experiment is complete. Reuse must not be for Commercial Purposes 2. Archived content may not be used for purposes that are intended for or directed towards commercial advantage or monetary compensation by means of sale, resale, licence, loan, transfer or any other form of commercial exploitation ("Commercial Purposes"). Wholesale re-publishing is prohibited 3. Archived content may not be published verbatim in whole or in part, whether or not this is done for Commercial Purposes, either in print or online. 4. This restriction does not apply to reproducing normal quotations with an appropriate citation. In the case of text-mining, individual words, concepts and quotes up to 100 words per matching sentence may be used, whereas longer paragraphs of text and images cannot (without specific permission from Springer Nature). Moral rights 5. All use must be fully attributed. Attribution must take the form of a link - using the article DOI - to the published article on the journal's website. 6. All use must ensure that the authors' moral right to the integrity of their work is not compromised. Third party content 7. Where content in the document is identified as belonging to a third party, it is the obligation of the user to ensure that any use complies with copyright policies of the owner. Reuse at own risk 8. Any use of Springer Nature content is at your own risk and Springer Nature accepts no liability arising from such use.