AASPI · University of Oklahoma
Coherence asks whether the reflector is broken. Curvature asks whether it is bent, and aberrancy whether the bend leans one way. Amplitude curvature asks the same question of how bright the reflector is rather than where it is. All of them are measured along the reflector, which means every one rests on an estimate of which way it is going at every sample in the volume — and inherits whatever that estimate got wrong.
These modules open that up. Build a structure, watch the attribute being computed on it, change the parameters that are usually left at their defaults, and see what happens to the answer. There is nothing to install.
What you need first. They are written for undergraduate geology and geophysics students. You should know what a seismic trace is, what a reflector is, and that a seismic section is recorded in two-way travel time rather than depth. Everything after that — dip, semblance, curvature, eigenvalues — is built up in order inside the modules.
Every geometric attribute is the output of an algorithm with parameters. Change the window, the aperture or the dip search and the answer changes — sometimes a little, sometimes completely.
Coherence is measured along dip. Curvature is the derivative of the dip field. Aberrancy is the derivative after that. An error in the dip estimate is passed on to all three.
Most attribute volumes are produced with whatever the software suggested. Those suggestions are reasonable starting points, not answers, and they are rarely revisited.
In the order they build on each other. Each one isolates a single attribute, or a single decision that goes into computing one. Work through them in sequence and you should be able to say what each attribute measures and why someone would run it, name the parameter that most changes the answer, read an attribute map without mistaking an artifact for geology, and give more than one explanation for any anomaly you see.
How dip is actually estimated: a brute-force search over candidate dips, scored by semblance. Watch the search run, then break it.
One dip component becomes two. Cut the same reflector along the inline and along the crossline and it shows two different dips, neither of which is the dip.
The same coherence calculation, run along dip and then along horizontal windows, on the same dipping data. One finds the fault; the other finds the bedding.
Semblance, variance, eigenstructure, energy ratio and Sobel computed side by side on one dataset, so the differences are visible rather than described.
The rate at which dip changes. Why bright curvature means flat rather than steep, why there are two principal curvatures, and why k_pos is not the same attribute as k₁.
Two principal curvatures recombined into how hard the reflector bends and what shape it bends into — dome, ridge, saddle, valley or bowl.
The second derivative of amplitude along the reflector rather than of the reflector itself. It finds channel edges and fan margins that have no structural expression at all.
The third derivative of structure. Three principal values instead of two, vectors instead of signed numbers, and the flexures left behind by faults too small to bend the reflector much.
These modules are teaching models, built to make ideas visible. The attribute algorithms are implemented as the literature describes them, but on small synthetic datasets and in their plain brute-force form, so the method can be read and watched rather than assumed. Each module lists what it has left out and why.
This is not attribute software. For production work on real volumes, use AASPI or your interpretation package. The numbers here describe the model on the screen, not your survey.
Built for teaching by Dr. Heather Bedle and Dr. April Moreno-Ward, School of Geosciences, University of Oklahoma, with the Attribute Assisted Seismic Processing and Interpretation (AASPI) consortium.
Companion to What Can You REALLY See in Seismic?, which covers resolution — how thin a bed, how small a fault, how narrow a feature. This set picks up where that one leaves off, and asks what the attributes computed from that data are actually measuring.
Everything runs in the browser. No installation and no account. Nothing you do inside a module — no slider, no click, no trace you generate — is transmitted anywhere. The module code makes no network requests of its own.
The one request each page does make is a page-view count, which records that the page was opened and nothing else. No cookie is set and no identifier is stored, and nothing about what you did on the page is sent. We keep that count for two reasons — so the modules people actually use are the ones that get improved, and so we can show the university that you all are using these, which is how they keep getting built.
Licensed CC BY-SA 4.0. Free to use, adapt and share, including in teaching, provided the source is credited and any adaptation is released under the same license. If you use it in a course or a talk, a credit line and a link back are all that is asked.