Skew-Gaussian Random Fields

Skewness is often present in a wide range of spatial prediction problems, and modeling it in the spatial context remains a challenging problem. In this study a skew-Gaussian random field is considered. The skew-Gaussian random field is constructed by using the multivariate closed skew-normal distribution, which is a generalization of the traditional normal distribution. We present a Metropolis-Hastings algorithm for simulating realizations efficiently from the random field, and an algorithm for estimating parameters by maximum likelihood with a Monte Carlo approximation of the likelihood. We demonstrate and evaluate the algorithms on synthetic cases. The skewness in the skew-Gaussian random field is found to be strongly influenced by the spatial correlation in the field, and the parameter estimators appear as consistent with increasing size of the random field. Moreover, we use the closed skew-normal distribution in a multivariate random field predictive setting on real seismic data from the Sleipner field in the North Sea.