A Comparison of the Efficiency and Robustness of Methods for Uncertainty Analysis. S. S. Isukapalli, A. Roy, and P. G. Georgopoulos, EOHSI, New Jersey
The computational costs associated with traditional Monte Carlo and Latin Hypercube Sampling methods for probabilistic risk assessment are often high. Here, two recent computationally efficient alternative methods are presented. These are the Stochastic Response Surface Method (SRSM) and the SRSM coupled with automatic differentiation method (SRSM-ADIFOR). These methods extend the classical response surface methodology to systems with stochastic inputs and outputs. This is accomplished by approximating both inputs and outputs of the uncertain system through stochastic series of "well behaved" standard random variables; the series expansions of the outputs contain unknown coefficients which are calculated by a method that uses the results of a limited number of model simulations. Here, the computational efficiency and robustness of these methods are compared with those of the Monte Carlo and LHS methods through case studies involving uncertainty propagation in complex, nonlinear models describing environmental and biological systems. The results indicate that the SRSM And SRSM-ADIFOR are computationally efficient that the Monte Carlo and the LHS methods, requiring an order of magnitude fewer simulations. Further, the Monte Carlo and the LHS appear to be inadequate in describing uncertainty propagation in complex, nonlinear models, whereas the SRSM and the SRSM-ADIFOR produce robust estimates with a limited set of simulations. Further, the SRSM and the SRSM-ADIFOR directly provide sensitivity information and insight into individual contributions of input uncertainties to the output uncertainties; producing such information would require a very large number of simulations using the traditional methods.
USEPA-NERL, USDOE (CRESP), and NJDEP.
Go to . . .
1999 SRA Table of Contents
1999 SRA Author Index
Main Abstracts Menu Page
RiskWorld Home Page