The construction of a velocity-depth model is an essential step for imaging complex structures.
One of the tecniques used to obtain this model is stereotomography, a tomographic method
based on the concept that locally coherent events provide information about velocity models.
Di erently from tradional tomography, stereotomography uses the horizontal components of
slowness as data for inversion. Moreover, it calculates, additionally to the velocity model,
a pair of ray segments that originates in the same point in depth. As any other inverse
method, stereotomography aims at nding the model parameters that best t the selected
data. The mathematical problem is non-linear, and it's solution is available through linear
iterations. The convergence of the algorithm depend strongly on the initial model. In the
original implementation of stereotomography, the initial velocity model should be given by
the user, while the initial ray parameters are computed in an homogeneous model. This
work extended the original algorithm by adding a new strategy for obtaining the initial ray
parameters, which are calculated through ray tracing in an arbitrary velocity model. With
this new source code, the in
uence of the initial model in stereotomography was evaluated.
The new strategy produced acceptable data mis ts in all tests. When using the models
resulting from time-to-depth convertion, the implemented tecnique produced better results
than the original one. This work also evaluated the in
uence of the data scaling on the
convergence of the algorithm. The appropriate choice of scales is essencial for obtaining
satisfatory data mis ts.
|
