Quantitative assessment of parameter sensitivity for SNESIM multiple-point geostatistics

In SNESIM (single normal equation simulation), the quality of modeling results is closely related to the parameter settings. To clarify the influence of parameters on the simulation results, quantitative metrics (including the porosity, pore throat distribution, variogram, and permeability) are presented and applied. Various SNESIM multiple-point statistics realizations with different simulation parameter settings are performed and compared against the TI (training image), and results show that (1) porosity shows no obvious sensitivity to the parameter settings in SNESIM simulation. (2) Circular search neighborhood, larger maximum number of conditioning data, and multi-grid number not too large or too small can ensure the reproduction quality of pore distribution. (3) Circular searching neighborhood, appropriate maximum number of conditioning data, minimum number of replicates not too small or too large, and reasonable number of multi-grids will benefit the reproduction quality of spatial variability. (4) Calculated permeability is sensitive to search radius settings, and show no sensitivity pattern to other parameter settings for the 2-d reproduction case in this paper. With constant Ry, the permeability in the x-direction of the simulation results show an overall climbing trend with increasing Rx, while the permeability in the y-direction tends to decrease. Similarly, when Rx is fixed, the permeability in the y-direction tends to increase with larger Ry, while the permeability in the x-direction shows a decreasing trend.