Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Stochastic simulation for imaging spatial uncertainty: Comparison and evaluation of available algorithms

Conference ·
OSTI ID:10185504
 [1];  [2]
  1. Nebraska Univ., Lincoln, NE (United States). Dept. of Biometry
  2. Sandia National Labs., Albuquerque, NM (United States)

Stochastic simulation has been suggested as a viable method for characterizing the uncertainty associated with the prediction of a nonlinear function of a spatially-varying parameter. Geostatistical simulation algorithms generate realizations of a random field with specified statistical and geostatistical properties. A nonlinear function is evaluated over each realization to obtain an uncertainty distribution of a system response that reflects the spatial variability and uncertainty in the parameter. Crucial management decisions, such as potential regulatory compliance of proposed nuclear waste facilities and optimal allocation of resources in environmental remediation, are based on the resulting system response uncertainty distribution. Many geostatistical simulation algorithms have been developed to generate the random fields, and each algorithm will produce fields with different statistical properties. These different properties will result in different distributions for system response, and potentially, different managerial decisions. The statistical properties of the resulting system response distributions are not completely understood, nor is the ability of the various algorithms to generate response distributions that adequately reflect the associated uncertainty. This paper reviews several of the algorithms available for generating random fields. Algorithms are compared in a designed experiment using seven exhaustive data sets with different statistical and geostatistical properties. For each exhaustive data set, a number of realizations are generated using each simulation algorithm. The realizations are used with each of several deterministic transfer functions to produce a cumulative uncertainty distribution function of a system response. The uncertainty distributions are then compared to the single value obtained from the corresponding exhaustive data set.

Research Organization:
Sandia National Labs., Albuquerque, NM (United States)
Sponsoring Organization:
USDOE, Washington, DC (United States)
DOE Contract Number:
AC04-76DP00789
OSTI ID:
10185504
Report Number(s):
SAND--93-0438C; CONF-9305268--1; ON: DE93040058
Country of Publication:
United States
Language:
English