One of the most efficient ways to produce unconditional simulations is with the spectral method using fast Fourier transform (FFT) [1]. But this approach is not applicable to arbitrary surfaces because no regular grid exists. However, points on the arbitrary surface can be generated randomly using uniform distribution to replace a regular grid. This paper will describe a nonstationary kernel convolution approach for data on arbitrary surfaces.
@article{arxiv.1509.01745,
title = {A Turning Band Approach to Kernel Convolution for Arbitrary Surfaces},
author = {Alexander Gribov},
journal= {arXiv preprint arXiv:1509.01745},
year = {2015}
}