Improvement of Dredged Material Management Decisions by Uncertainty Analysis

d. Interval Analysis, Fuzzy Arithmetic, or Dependency Bounds Analysis

(Ferson and Burgman 1995; Ferson and Kuhn 1992). When insufficient

information is available to conduct probabilistic analysis, these types of

analysis might be appropriate. These methods explicitly acknowledge

uncertainty but do not require the analyst to define distributions to input

parameters.

e. Expert elicitation (Cooke 1991; National Council on Radiation

Protection and Measurements (NCRP) 1996; Graham, Hawkins, and

Roberts 1988; Evans et al. 1994) is also used in conjunction with these

methods, where quantitative or qualitative information needed to assess

risk is not available in the published literature.

f. Bayesian statistical procedures (Box and Tiao 1973; Gelman et al. 1995;

Morgan and Henrion 1990) are based on the Bayesian, or subjective,

view that probabilities can be estimated using scientific knowledge,

expert judgment, experience, and intuition combined with new data.

The preliminary ranking of uncertainty sources in this report relies primarily

on expert judgment. However, this ranking will be used to support

recommendations for future, more detailed assessments of uncertainty in dredged

material management. It is important to recognize that these methods can

introduce new sources of uncertainty. For example, probabilistic analysis

depends on the analyst identifying the appropriate shape of input distributions

where there may be few data to support this selection.

Improvement of Dredged Material Management

Decisions by Uncertainty Analysis

Evaluation of dredged material impacts on the environment depends on many

input parameters and models that are subject to a great deal of uncertainty. How

this uncertainty is treated quantitatively can lead to overestimates or

underestimates of risk. The tiered approach for dredged material disposal

decisions is designed to err on the side of conservatism to account for

uncertainty. However, the approach does not explicitly address uncertainty

quantitatively, resulting in predictions with unknown levels of conservatism built

into them. The degree of conservatism varies from assessment to assessment,

depending on the number of and mathematical relationship among input

parameters (Ferson and Long 1995). Uncertainties must be acknowledged and

quantified as much as possible to understand how confident one can be about a

disposal decision.

Distinguishing sources of uncertainty and variability can improve predictions

of potential adverse impacts. Uncertainty refers to a lack of knowledge, while

variability refers to natural heterogeneity. Nature is replete with variability. The

composition and abundance of populations can vary with climatic events or

seasonal changes. Individuals within populations might vary in their response to

chemical stressors in dredged material. For example, amphipods are often used as

test species in whole-sediment bioassays, but a very different result might be

Chapter 1 Introduction