ABSTRACT
In this paper two questions are addressed: (1) A stochastic finite element is applied to the Kriging problem (Journel 1977). In this context, the expectation of a random function Z(x) and its covariance (which does not lead to a negative kriging variance) are assumed, depending on x:
E(x)=m(x)
C(x,h)=E{(Z(x+h)Z(x)}-m(x+h)m(h)
(2) Using a stochastic finite element, the mean value is calculated taking into account different bases: polynomial, exponential, logarithmic etc. (Dhat, 1984). For the data distribution given at the nodes of a triangle, a Monte Carlo procedure is applied and an illustrative example shows the role of the base chosen on the estimation. Finally, several conclusions are given.