Stochastic Signal Processing > Spatial Data Conditioning
  • Step 1: Spatial Analysis.
    The CDP gather displayed in figure a contains some dipping events crossing the flattened events. This "noise" may be linked to multiple reflections from the shallow subsurface. It is not related to AVO variations as it should be removed from the raw data before computing AVO attributes.
    Spatial analysis enables to identify the AVO noise and signal through their spatial covariances as it is illustrated in Figure b: The experimental variogram in the time and offset directions is modeled using 3 spatial structures, 1 accounting for the "AVO non constructive part" or noise, 2 for the "AVO constructive part" or signal. Modelling variogram is equivalent to modelling the frequency power spectrum in the frequency domain as shown in figure c.

  • Step 2: Factorial Kriging .
    Factorial Kriging is the spatial operator which best estimates the constructive part of each trace according to the modelling of the experimental spatial covariance in terms of signal and noise frequency content as shown in figure b.
    Figure c displays the results of the noise and signal estimated contents of raw gather data displayed in figure a.

  • Step 3: Validation.
    In order to validate the performance of the factorial kriging AVO intercept and slope attributes are computed on the raw data in figure a and b and on estimated AVO signal data after factorial kriging in figure c and d.
    It is clear that the values of AVO intercept and gradient values have not changed and that the fluctuations around the regression line have been reduced. This confirms that the removed estimated noise, although organised in space, does not contribute to AVO computation.