Fig. 1. A schematic representation of how Monte Carlo uncertainty assessment would be accomplished in a decision support system by linking a simulation model with spatial analysis.

Fig. 2. Comparison of different levels of complexity of uncertainty assessment using two techniques: first-order Taylor series variance approximation (TS) and crude Monte Carlo technique (MC). Var(X) includes only variances, with no covariances, Cov(X) includes covariances between variables, and Cov(X, B) includes covariances between variables and calibrated coefficients. (Coefficient of variation is the standard deviation of the estimate expressed as a percent of the mean.)

Fig. 3. A representation of spatial uncertainty. (a) cumulative result from 500 realizations of a Monte Carlo GIS analysis. (b) The 90th, 95th, and 99th percentiles in (a) are shown in increasingly darker shades. Lower probability areas may be used as buffers or to provide information on the effects of compromise in the decision-making process.

Fig. 4. (a) Two alternatives with the same most likely predicted value (mean), but different variances (uncertainty). Alternative 1 with the smallest variance (uncertainty) would be preferable. (b) Alternative 3 has the same level of uncertainty (variance) as Alternative 1, but a larger mean value. Alternative 2 would then be preferable if there is concern of exceeding an environmental threshold.