Comparison of Stochastic Simulation Algorithms in Mapping Spaces of Uncertainty of Non-linear Transfer Functions
Geostatistical simulations are routinely used to quantify the uncertainty in forecasts (responses) generated from any non-linear function of spatially varying parameters. The ability to map the uncertainty in these responses is critical. A comparison of sequential Gaussian simulation (SGS), sequential indicator simulation (SIS) and probability field simulation (PFS) is made in this study, using an exhaustive dataset sampled with a random stratified grid and three transfer functions, namely, minimum cost network flow, threshold proportion and geometric mean. The results show that SGS and SIS have comparable performance in terms of bias and precision while PFS performs less well in most cases. Increased data leads usually to better precision but not necessarily bias. The performance of the simulation methods in mapping spaces of uncertainty depends on the complexity of the transfer function, and that is not necessarily a well-understood aspect of the modelling process.
KeywordsTransfer Function Stochastic Simulation Simulation Algorithm Uncertainty Distribution Stochastic Simulation Algorithm
Unable to display preview. Download preview PDF.