Processing adequacy criterion

Is it proven to use the assumptions about homogeneous layer to solve the inverse kinematical problem for current layer? How can we increase the accuracy of processing and check our results?

Criterion value dimension is ms. Criterion calculated for every point of line as a difference measure of two solutions of one inverse problem. These solutions are so-called R-method and V-method. It is proved, that these solutions must be coincide to each other if and only if computations were made for homogeneous layer. Thus in this case time section was correctly separated to the homogeneous layers. If criterion value is higher than threshold value (as usually threshold equal to the sample rate) – we are probably deal with one of the situations, described below:

  • Heterogeneity of current layer. More detailed separation by layers is required. Two layers with high velocity contrast were probably considered to be a single layer. Therefore it is need to include new layer and resolve the inverse kinematical problem for it.
  • Inaccuracy of initial data. In this case you have to estimate arrival times and velocities for current layer more accurately.
  • Mistakes, which were collected on the previous stages of depth-velocity model building. You need to check your previous results and improve them.

Therefore criterion usage allows you to:

  • Control the adequacy and correspondence of depth-velocity model to the real data;
  • Check the initial data quality;
  • Find your mistakes from the previous processing stages;
  • Choose valid and correct processing parameters.
Examples of correct (on the left) and incorrect (on the right) depth velocity models, obtained for the same time section. Correct model is characterized by small (not higher than sampling rate) values of criterion. Incorrect model is characterized by high
Examples of correct (on the left) and incorrect (on the right) depth velocity models, obtained for the same time section. Correct model is characterized by small (not higher than sampling rate) values of criterion. Incorrect model is characterized by high