Simulate random surface by kriging but with no input points

I am looking to simulate 2-D random rough surfaces, however the height distribution of the surface topography is heavily skewed (i.e. not Gaussian). I believe this rules out using spectral analysis.

I'm therefore interested to know if I can use kriging/variogram analysis as an alternative to spectral analysis. Rather than interpolating a grid from known input points (as kriging is commonly used for), I want to specify a variogram model (including the variogram parameters) and create a random surface grid from this model, without input points.

Is this possible with any of the available matlab kriging functions (e.g. http://www.mathworks.com/matlabcentral/fileexchange/29025-ordinary-kriging or http://www.mathworks.com/matlabcentral/linkexchange/links/1876-mgstat-a-geostatistical-matlab-toolbox)? I am not familiar with this technique so any help would be appreciated.

Cheers, JK