An Analysis of Integrating Conservatism Globally and Locally in the Environmental Evaluation Process. M. N. Gray, J. H. Samuelian, and W. R. Alsop, AMEC Earth & Environmental
Historically, the use of point estimates for calculations to describe a spatially and temporally changing set of conditions was required due to the inability to represent more closely a set of complex and changing parameter values. Probabilistic and numerical analyses have become prevalent as more sophisticated analytical tools and techniques have been developed. These tools and techniques allow the use of larger data sets in their entirety within series of linear and nonlinear equations. A point estimate calculation is usually a number of arbitrarily selected parameter values that are entered into an equation with one numerical result being generated. A probabilistic analysis can be designed to incorporate all available data. The distributions of input data are entered into an equation with the various combinations being used to generate results that incorporate the frequency of each value in the input distributions. The output results are retained as a distribution of values for subsequent numerical analysis. In many cases, oversight agencies have made decisions regarding the degree of conservatism in a point estimate for each of the inputs while not considering the effect of compounding a series of low probability events. Many times this individual input conservatism produces results that are impressive in their numerical value but do not provide the context that a probabilistic analysis does. A risk manager needs to know how close a point estimate is to the upper bound for a given set of circumstances in order to make an informed decision. A series of Ambient Water Quality Criterion calculations exemplifies the results of combining various point estimates and their placement in the corresponding probabilistic distribution. These juxtaposed results illustrate the effect of compounded independent individual conservative parameter value selection versus selection of a conservative value from the output distribution.
Go to . . .
2002 SRA Annual Meeting Table of Contents
2002 SRA Annual Meeting Author Index
Main Abstracts Menu Page
RiskWorld Home Page