Stochastic Signal Processing > Kriging
  • Step 1: An hybrid probability model.
    Bayesian kriging is the estimation operator associated to stationary and non-stationary probability models where m of x is controlled by an a priori distribution of possible values . The probability model then assumes modelling of m of x using a constant or a linear combination of external drifts, the coefficients of which being Gaussian random variables with known a priori mean and variance.
    In the case of a depth conversion of a single layer as shown in figure a, and in order to gain more control on the resulting velocity law parameters, a priori distributions of the external drifts coefficients may be added that are called Bayesian constraints of the drift as shown in figure b.
    Bayesian Kriging probability models stand between Simple kriging and Kriging with external drift models. They are useful for geoscience applications, as they take into account a priori knowledge on the range of acceptable kriging solutions.