Stochastic waves are simulated in a non-homogeneous field; layered med
ia with irregular interfaces. Observed waves are specified at one or m
ore points, and the proposed procedure simulates waves at arbitrary po
ints for which no motion has been proposed, using only information fro
m observed records. Stochastic waves are assumed to be composed of a d
eterministic component (trend wave) and a stochastic component (random
wave). We propose a simple trend model that uses the Fourier spectrum
of the observed wave. The kriging method is used for the optimum inte
rpolation of random waves. According to the conditional simulation, ra
ndom stochastic waves were generated on a non-homogeneous random field
. The simulated waves are coincident with known time histories at spec
ific points. To check the validity of the procedure developed, we calc
ulated the waves in layered media with irregular interfaces using the
discrete wave-number method and compared them to the waves simulated b
y our procedure. This procedure that includes the kriging technique pr
ovides an efficient means by which to simulate the stochastic waves of
a non-homogeneous random field.