Visualization facilitates uncertainty evaluation of multiple-point geostatistical stochastic simulation

Abstract

Stochastic simulation is an essential method for modeling complex geological structures in geosciences. Evaluating the uncertainty of the realizations of stochastic simulations can better describe real phenomena. However, uncertainty evaluation of stochastic simulation methods remains a challenge due to the limited data from geological surveys and the uncertainty in reliability estimation with stochastic simulation models. In addition, understanding the sensitivity of the parameters in stochastic simulation models is invaluable when exploring the parameters with a higher influence on the uncertainty associated with predictions generated from stochastic simulation. To facilitate uncertainty evaluation in stochastic simulation methods, we use the circular treemap as an interactive workflow to explore prediction uncertainty in and the parameter sensitivity of multiple-point geostatistical (MPS) stochastic simulation methods. In this work, we present a novel visualization framework for assessing the uncertainty in MPS stochastic simulation methods and exploring the parameter sensitivity of the MPS methods. We present a new indicator to integrate multiple metrics that characterize geospatial features and visualize these metrics to assist domain experts in making decisions. Parallel coordinates-scatter matrix plots and multi-dimensional scaling (MDS) plots are used to analyze the parametric sensitivity of MPS stochastic simulation methods. The realizations and parameters of two MPS stochastic simulation methods are used to test the applicability of the proposed visualization workflow and the visualization methods. The results demonstrate that our workflow and the visualization methods can assist experts in finding the model with less uncertainty and improve the efficiency of parameter adjustment using different MPS stochastic simulation methods.