MCvfit - Monte Carlo automatic velocity picks (fit)
MCvfit < in.H par= > out.H
input vscan(t,vel), can be n3>1
optional interval vel file ( = vint.H )
optional vel misfit error file ( = verr.H )
optional rms vel guess file ( = grms.H )
optional interval vel guess file ( = gint.H )
Vrms intercept at t0
Vrms "gradient"
Vrms "power" function.
nlayers per second. ~20 good.
max allowed MCvint contrast between 2 adjacent layers
Don`t allow trial MCvint(t) < v1bnd*Vint(t)
Don`t allow trial MCvint(t) > v2bnd*Vint(t)
Global min a value
Global min b value
Global min c value
global max MCvint value
near surface constant velocity
time thickness of near surface constant velocity layer
allows a dipping water bottom: tsurf(x) = tsurf + (x-x0)*dtsurfdx.
variance of random velocity perturbations.
number of random walks. A "random walk" starts with one initial model and tries many random steps away from that same initial model. The best model from one walk is used as the starting model for the next walk.
number of random steps per walk.
If fit is unchanged for nconv steps, go to a new walk. If the fit remains unchanged for nconv walks: stop.
If fit changes < toler, consider that no change. Note, a perfect fit is 1.0 (100%). The most critical parameters to try and optimize to get a better fit are: vmin*, nvdiv~10/20, dvzmax~0.6, sigma~dvzmax/2
Monte Carlo fit parameters...
This part randomly perturbs the starting Vint(t)
model corresponding
to the starting Vrms(t)
model found above.
These parameters refer to *interval* velocity, not Vrms!
seis/velan