Seismic interpretation, from the ground up
Seismic data does not show the earth directly. It shows the earth convolved with a wavelet, and that wavelet sets which features can be resolved, which faults can be mapped, and whether a bright event means rock or interference.
Each module is built around one experiment. Move a slider, watch a bed or a fault drop below the resolution limit, and find where that limit sits.
Wavelength equals velocity over frequency. A 3000 m/s layer imaged at 30 Hz gives a 100 m wavelength. Nearly every limit on this site is some fraction of that one number.
Beds thinner than about a quarter of a wavelength stop showing a separate top and base. Twenty-five meters, in that example. The modules measure the limit directly rather than assuming it.
A thin bed can often be detected long after its thickness has stopped being measurable. Confusing the two is a common source of error in thickness maps.
The modules — in the order they build on each other
Build a column from the sea surface down to five kilometers. Compaction, an overpressured zone, depth into time, and what attenuation does to your bandwidth.
One column of rock. Set thickness, velocity and density, and follow it through impedance and reflection coefficients to a synthetic trace, with nothing hidden in between.
Thin a bed to nothing and watch the reflections merge, brighten, then fade, while the thickness picked off the section stops matching the bed.
Test out throw, frequency, and noise against your own eyes. Includes a quiz mode that hides the answer — and sometimes hides no fault at all.
One unmigrated trace samples a patch the size of several football fields, not a point. Migration collapses that patch. Watch two features merge as they close.
Velocity climbs, the earth attenuates the high frequencies, and λ/4 in meters grows with depth.
A channel that narrows and splits in two. Watch where it appears to end, and where the two distributaries stop looking like one.
A sand thinning updip onto a salt dome, turned into an amplitude extraction. A bright rim appears around the flank — and there is no hydrocarbon anywhere in the model.
Rotate the wavelet and watch the apparent bed boundary move. This is why well ties and picked horizons often disagree.
These modules are teaching models, built to make ideas visible. Every one of them is a deliberate simplification: they leave out noise, dip, anisotropy, multiples, velocity error and most of what a processing sequence actually does, and each module says in its own notes which of those it has set aside and why.
They are not processing or interpretation software, and the numbers they produce describe the model on the screen rather than any real survey. Use them to build intuition and to see why a limit exists — then check your own data, with your own velocities, before you act on anything.