PREDICTIVE DECONVOLUTION FROM THE POINT OF VIEW OF KRIGING

摘要

We review the geostatistical method of simple kriging and consider its application to the predictive deconvolution problem in seismology. We find that when kriging is applied to the ID time-series prediction problem, it can be used to obtain the usual Wiener-Levinson system of equations that are normally arrived at with the methods of stationary time-series analysis. The kriging weights can then be interpreted in terms of the prediction filter coefficients. This connection between predictive decon and geostatistics does not appear to be known. Perhaps a better understanding and exploitation of this type of connection can lead to productive synergies between signal processing and geostatistics.