Diagonal method

The spirit of this method is to give the same chance to models with a regular velocity increase and to models with sharp contrasts. The velocities are defined by a minimum value (first parameter, $V$ ). The other parameters are between 0 and $1$ . A total velocity variation is calculated from the second parameter $p_1$ , $\delta V=p_1*(V_{max}-V)$ where $V_{max}$ is the fixed maximum velocity (4000 m/s in this case). The third parameter is the intersection of the profile with the ascending diagonal of the rectangle defined by the $V$ and $V+\delta V$ of the top and the bottom layer, respectively. 0 means minimum velocity to be affected to the deepest layer. $1$ means maximum velocity to be affected to the highest layer. The already defined layer separates the stack into two sub-stacks that can be processed in the same way. The prior distribution for this method is given in figure B.6. The results are quite similar to the results of the preceding method, except for profiles with low velocity at depth where this method appears to be slightly more efficient. A symmetric distribution can also be generated by inverting the velocity of the first and last layer like in the above method.