Factorial kriging for multiscale modelling
YZ Ma, J-J Royer, H Wang, Y Wang, T Zhang
This paper presents a matrix formulation of factorial kriging, and its
relationships with simple and ordinary kriging. Similar to other kriging
methods, factorial kriging can be applied to both stationary and intrinsic
stochastic processes, and is often used as a local operator. Therefore, the
concepts of intrinsic random function and local stationarity are first
briefly reviewed. Kriging is presented in a block matrix form in which
the kriging solution is useful not only for understanding the
relationships between simple and ordinary kriging methods, but also the
relationships between interpolative kriging and factorial kriging. When
used as a signal/noise-filtering method, factorial kriging is especially
useful for multiscale modelling. Examples for general signal analysis
and geophysical data signal filtering are given to illustrate the method..
Keywords: kriging in matrix form, locally stationary, block matrix, multiscale
modeling, relationship between simple and ordinary kriging, filtering.