Velocity: One Number, Five Sciences
The sonic measures it over centimeters, semblance picks it over kilometers, rock physics explains it, tomography and FWI refine it, and pore-pressure prediction reads danger in its reversals. Velocity is the connective tissue of applied geophysics, and the same rock answers differently depending on who is asking. Follow one number through every discipline that claims it.
You can say what each velocity IS, moduli over density at the core, RMS and interval on the gather, anisotropic when the rock insists, convert between them without pretending they agree, pick velocities on semblance, read a CIG for model error, explain what tomography and FWI each add, hitchhike Vs and density when logs are missing, and read an overpressure warning in a velocity that turns back below its trend.
What velocity is
Vp is moduli over density in a trench coat: every velocity statement is secretly a statement about stiffness, and forgetting that is how velocities get treated as arbitrary knobs.
The borehole owns the only direct measurement: a transit time over a known span, at log resolution, in the rock itself; every surface velocity ultimately answers to it.
Running the machine yourself, K and G and density in, Vp and Vs out, is what makes velocity intuition quantitative instead of decorative.
The processor's velocity
The processor's velocities are travel-time bookkeeping, not rock properties; Dix turns RMS into interval, and confusing the family members is the classic way to be wrong by ten percent.
The semblance panel is where velocity becomes a decision: the pick flattens the gather, and everything downstream, stack, migration, amplitudes, inherits it.
When the rock is VTI the moveout velocity and the vertical velocity part company, and eta is the fee; ignore it and the far offsets and the depths are both quietly wrong.
Residual moveout across image gathers is data: tomography inverts it into a velocity update, turning the migration model from a guess into an iterated solution.
Migration spends velocity: flat common-image gathers are the receipt for a correct model, and curling gathers are the bill for a wrong one.
Full-waveform inversion stops picking events and fits the entire recorded wavefield; done well it returns velocity at a resolution tomography cannot touch, done carelessly it converges beautifully to the wrong earth.
What velocity carries
Half the wells are missing the shear or density log, and Castagna and Gardner ride along on Vp; knowing their error bars is what separates a pseudolog from a fabrication.
Effective stress sets velocity, so pore pressure is hiding in every velocity field; the rock that reads slow for its depth is the drilling hazard announcing itself years in advance.
A slow anomaly forges structure on everything beneath it; pushdown is the reminder that velocity errors do not stay local, they masquerade as geology.
Every velocity in the path converges here: depth conversion is where velocity uncertainty becomes structural uncertainty, in meters, on the map someone will drill.