(View the PDF document here. Adobe Acrobat is
required:
)
Traditional velocity analysis, as done by most commercial software, follows these steps (SeismicStudio can be run in this mode if you'd like to try it):
The problem lies in step 2. Briefly stated, at any point where delay times vary over the area of the survey, step 2 becomes inaccurate.
Consider a source located at x=0 being shot into receivers extending in the positive x direction. The surface is perfectly flat. The refractor dips down to the right.
D(x) = delay time as a function on x
D(x) = D0 + M * x (if M > 0 then the refractor dips down to the right)
Pick(x) = T(x) = Source delay + offset term + detector delay = T0 + x / V + (T0 + M * x) = 2 * T0 + x * (M + 1 / V)
The measured velocity will be a function of the refractor velocity and the dip term:
1 / V(measured) = M + 1 / V
V(measured) = V / (1 + M * V)
Example: Let V=3000m/sec, and let M=15 millisecond per kilometer. The measured velocity will be 2870 m/s.
| The problem will be worse near the survey edges | |
| The greater the delay time variation, the worse the problem | |
| Steps 3 through 5 DO NOT "HEAL" the velocity! |
It seems as though very few people actually test their software on very simple, well-understood models. This is why we included in SeismicStudio the ability to create synthetic picks using imported geometry.
SeismicStudio contains a proprietary algorithm called Residual Velocity Correction (RVC). Based on extensive tests on synthetics and real data it can dramatically improve the velocity field.
(You don't need a SeismicStudio license)