Seismic amplitude volumes continually grow in number, size and spatial resolution. Meanwhile, a reduced number of seismic interpreters are tasked with mapping an expanded list of geologic targets ranging from traditional hydrocarbon traps, through shallow drilling hazards, to reservoirs for CO2 sequestration. To aid in these efforts, modern seismic attributes convert the seismic wiggles into images that more closely resemble geology. These images enable geoscientists to directly apply their understanding of the depositional environment, tectonic history, and diagenetic alteration to construct interpretations that go well beyond the simple time-structure maps of the past.

Today’s seismic interpreters are generalists rather than specialists, with a responsibility to understand both geologic processes and engineering data, to coordinate with drilling teams and partners and to integrate all these tasks within the corporate project management framework. Most companies have outsourced seismic processing and imaging to service providers, resulting in limited hands-on learning experiences for their newer generation of employees. This limited knowledge on how the seismic data volumes are generated, coupled with the high quality of these data, can lead to a false sense of confidence in the data, leading to interpretation pitfalls.

In this presentation, I will focus on common pitfalls in seismic attribute analysis. False structure due to velocity pullup and pushdown and fault shadows on time-migrated data predate but are enhanced by modern seismic attributes. Acquisition footprint is actually enhanced by many attributes. Spectral decomposition measures apparent (vertical) tuning rather than the tuning perpendicular to a thin bed. Other pitfalls arise by choosing attribute parameters that inadvertently mix stacked stratigraphy or by choosing parameters that exacerbate noise in the data. A common pitfall is blindly trusting the default parameters of a given attribute or interpretation tool. Although they are not seismic processors, interpreters have rudimentary processing tools at their disposal that allow them to balance the spectrum and enhance edges, thereby enhancing both vertical and lateral resolution. A common pitfall here is to accept the nicer looking result without correlating with well log data or examining the part of the data that was rejected. Depth-migrated data introduces a new suite of attribute pitfalls and limitations. Depth migration is needed to image steep dips and areas with significant velocity heterogeneity in the overburden. Here, an attribute algorithm may not be able to differentiate steeply dipping fault plane reflections or steeply dipping migration artifacts from steeply dipping stratigraphic reflections. Whereas the dominant period in time-migrated data may vary by a factor of two from shallow to deep, in depth-migrated data the dominant wavelength may vary by a factor of eight, requiring more careful attribute computation strategies.

The most common pitfall is expecting an attribute to give you what you want rather than what it measures. I will illustrate these pitfalls through examples; where possible, I will suggest ways to avoid them or minimize their impact.