Seismic facies classification using machine learning tools

Graham Carter

Instructors: Satinder Chopra & Kurt Marfurt
Date: Nov. 7-8, 2019
Duration: 1 day
Members (early bird/price): CAD$ 800/1000 (plus GST)
Non-members (early bird/price): CAD$ 900/1100 (plus GST)
Location: TBD
Time: TBD


Satinder Chopra has 35 years of experience as a geophysicist specializing in processing, reprocessing, special processing and interactive interpretation of seismic data.  He has rich experience in processing various types of data like VSP, well log data, seismic data, etc, as well as excellent communication skills, as evidenced by the several presentations and talks delivered and books, reports, and papers written. He has been the 2010/11 CSEG Distinguished Lecturer, the 2011/12 AAPG/SEG Distinguished Lecturer and the 2014/15 EAGE e-Distinguished Lecturer. His research interests focus on techniques that are aimed at characterization of reservoirs.  He has published 8 books and more than 450 papers and abstracts and likes to make presentations at any beckoning opportunity. His work and presentations have won several awards, the most notable ones being the 2019 AAPG Distinguished Service Award, 2017 EAGE Honorary Membership Award, 2017 CSEG Symposium honouree, CSEG Honorary Membership (2014) and Meritorious Service (2005) Awards, 2014 APEGA Frank Spragins Award, the 2010 AAPG George Matson Award and the 2013 AAPG Jules Braunstein Award, SEG Best Poster Award (2007), CSEG Best Luncheon Talk award (2007) and several others. He is a member of SEG, CSEG, CSPG, EAGE, AAPG and APEGGA (Association of Professional Engineers, Geologists and Geophysicists of Alberta).


An ongoing challenge to seismic interpreters is to identify and extract heterogeneous seismic facies on data volumes that are continually increasing in size. Geometric, geomechanical, and spectral attributes help to extract key features but add to the number of data volumes to be examined. Common analysis tools include interactive co-rendering, crossplotting, and 3D visualization where we examine more than one attribute at a time, data reduction, where we mathematically reduce the number of data volumes to a more manageable subset, clustering, where the goal is to identify voxels that have similar expressions, and supervised classification, where the computer attempts to mimic the skills of an experienced interpreter. In this two-day course we compare several of the more well-established machine learning techniques: waveform classification, principal component analysis, k-means clustering, and supervised Bayesian classification. We also examine some less common clustering techniques applied to seismic attributes including independent component analysis, self-organizing mapping and generative topographic mapping.  We will demonstrate that the machine learning methods hold promise as each of them exhibits more vertical and spatial resolution than the waveform classification, or the supervised Bayesian classification. Discussion will also encompass the supervised and deep-learning applications.

A novel part of the course would be the practical component that includes the computation of seismic facies using many of the above-mentioned techniques. For this, seismic as well log data will be used from a few different and diverse basins around the world, and the course participants will be able to compute the different facies on them using the software designed and developed for the purpose at the Attribute Assisted Seismic Processing and Interpretation (AASPI) Consortium at the University of Oklahoma. This aspect would make the course more beneficial and the participants would take away the experience of not only generating but displaying the different seismic facies for their interpretation.