Running the inversion algorithm

 

datag.first
Figure 10 The initial gradient of the stack semblance for the picked events tomographically back-projected into the model space by ${\bf G}^T$.

Figure 10 shows the initial gradient of the objective function with respect to the interval slowness model (taken with $\gamma=1$). This plot is the first iteration of a tomographic back-projection of the velocity errors. This gradient gives an idea of what parts of the model have errors and how changes in those parts of the model affect the stack semblance of the data.

 

deltaw.lastiter
Figure 11 Estimated change to the interval slowness model after 30 iterations of the velocity estimation algorithm. The inversion stopped when $\nabla_\gamma Q \approx 0$, that is, when all residual moveout had been explained.

The PARTAN algorithm was run for 30 iterations and stopped when the gradient of the objective function with respect to $\gamma$ was approximately zero. Figure 11 shows the cumulative change in the interval slowness model obtained.