It’s a small miracle that someone can look at a set of seismic lines, then produce an interpretation related to real-world geology.
And it’s no surprise at all that some of those interpretations are wildly wrong.
At AAPG’s upcoming Annual Convention and Exhibition in New Orleans, geoscientist Clare Bond will discuss the interpretation of seismic data, what builds confidence for interpreters – and what one technique produces the best results.
Bond, structural geologist for Midland Valley Exploration Ltd. in Glasgow, Scotland, will present “When There Isn’t a Right Answer – Dealing With the Uncertainty of Seismic Interpretation to Maximize Success.”
She and her colleagues have studied how geoscientists approach seismic information to develop geological models.
In particular, they were interested in the failed or inaccurate interpretation with a poor match to naturally evolved geology.
“You can kind of look at it with an expert eye and say, ‘That doesn’t work.’ We were interested in why people create cross-sections that don’t make any sense,” she said.
Geoscientists aren’t alone in struggling with incomplete information or flailing about because of unknown variables.
Bond cited a landmark paper by Nobel Prize laureate Daniel Kahneman and his frequent collaborator Amos Tversky, “Judgment Under Uncertainty: Heuristics and Biases,” published in the journal Science in 1974.
“It basically looked at how awful human beings are in making decisions when there is uncertainty,” she noted, “or at least risking.”
The Bearing of Bias
As part of this project, Bond gave a synthetic seismic section to various participants and asked for their interpretations. Participants in her studies have been drawn about 50-50 from industry and academia, and have ranged from university undergraduates to professionals with more than 15 years of experience, she said.
Out of 412 interpretations:
- Only 21 percent of the participants identified the correct tectonic setting of the original model.
- Only 23 percent were able to highlight the three main faults in the seismic image.
- Only 10 percent of the interpretations showed the correct inversion model for the section.
That experiment quantified the range of conceptual uncertainty for the seismic set and helped identify factors that influence how knowledge and concepts are applied to seismic interpretation.
By examining the results, Bond and her colleagues were able to classify distinct interpretational techniques, to determine what approaches worked best and to study how experience, bias and area of expertise affected the interpretation process.
Bond noted that one challenge for an interpreter is honoring the seismic data while devising a geometrically possible model – an interpretation relevant to and consistent with real-world geology.
“There’s always this interplay going on,” she said.
Bias can have a negative or positive effect. For instance, if an interpreter is biased toward a model involving basin subsidence and subsidence is a factor in the correct interpretation, the chance for a right answer increases.
But bias, area of expertise and prior experiences can lock an interpreter into a wrong or unhelpful approach.
To minimize that risk, Bond has suggested workflow strategies that broaden the range of concepts available to geoscientists as they interpret seismic data, followed by strategies to refine the models produced.
Specifically, her suggestions for interpretation include:
- Using one or more geoscientists with a range of prior knowledge.
- Exposing geoscientists to multiple concepts prior to interpretation.
- Removing regional and tectonic context.
- Encouraging multiple interpretations.
For assessing concepts and refining and risking models, she suggested:
- Using structural evolution and restoration techniques to determine model viability.
- Considering regional and tectonic context.
- Use of peer review and specialist technical assurance.
- Assessment of play impact.
Examples from the interpretation exercises can be found online in the Virtual Seismic Atlas(VSA).
VSA describes itself as an “open access community resource to share the geological interpretation of seismic data,” with high-resolution images that can be freely downloaded and explored.
Overseen jointly by the University of Aberdeen and the University of Leeds, the VSA project is supported by the Natural Environment Research Council, the Geological Society of London, the Petroleum Exploration Society of Great Britain and a number of energy companies.
Take Your Best Shot
Bond’s presentation at the New Orleans meeting will examine interpretation of seismic data sets by geoscientists at various career-stages. Her co-authors are Zoe Shipton, Euan Macrae and Chris Philo, all from the department of geographical and earth sciences at the University of Glasgow.
By studying the evolution of approaches to seismic interpretation along with the interpreters’ level of experience and expertise, Bond and her colleagues have been able to identify key skills and technical workflows.
That enabled the development of new training strategies, with the goal of helping geoscientists optimize their interpretational ability, at any career stage.
The overall study compared two experiments. In the first, geoscientists were asked to interpret a seismic section without any information about geological setting, stratigraphy or other context. In the second, participants were given a seismic section with five possible interpretations and asked to choose the best one, or to make their own.
Comparing those exercises provided information about how geoscientists evolve skills to deal with interpretational uncertainty and how their approaches change when similar problems are framed differently.
Macrae will follow that paper with the presentation “Uncertainty Analysis of Geological Interpretations,” with Bond and Shipton as co-authors.
His presentation will describe a statistical investigation of the factors that contribute to conceptual uncertainty in interpretation, called the Freyja Project.
“The original project was called Odin – we gave them the names of gods from Norse mythology,” Bond explained.
A questionnaire and seismic image will be handed out to members of the audience for the presentation, so everyone who wants to can participate in the project, she said.
What builds confidence for seismic interpreters?
The key is experience, Bond observed.
Right or wrong, geoscientists with more experience were more willing to identify and label features like fault lines in a seismic section.
And what one, specific approach produces the most accurate seismic interpretations?
Bond will discuss that finding, and others, during the April 12 afternoon session, “Seismic Interpretation of Faulted Reservoirs: How to Get the Right Answer the First Time.”