A method to generate a Gaussian and log normal random field is presented. The method is based on a matrix factrization technique of decomposition of a covariance matrix into its squre root matrix approximately by using the Chebyshev polynomials. When we generate a Gaussian or lognormal random field with Nx×Ny points covered by a mesh of a rectangular grid, the algorithm presented here requires only 3NxNy entries to be stored.
