“…Any permissible autocovariance model may be considered and, given that the random and deterministic components of the algorithm are separated in the spatial, rather than in the wavenumber, domain, local re-simulations of specific areas of the model grid are possible. The latter cannot be done with standard power spectral simulation techniques (e.g., Ikelle, 1993) and has led to the common application of FFT-MA for stochastic modeling and inversion (e.g., Le Ravalec-Dupin et al, 2004;Le Ravalec and Mouche, 2012;Liang and Marcotte, 2016;Yang and Zhu, 2017;Lauzon and Marcotte, 2019).…”