UNCERTAINTY ASSOCIATED WITH SEMIVARIOGRAMS USED FOR SITE SIMULATION

Uncertainty is associated with interpretation of the subsurface, and s tochastic simulation techniques are incapable of accounting for all th e uncertainty if only a single deterministic semivariogram model is ut ilized. Jackknifing the data bounds the limits of model semivariograms but typically indicates that a large number of simulations must be co nducted to consider the full distribution of possible semivariograms. Latin-hypercube sampling, particularly when combined with expert opini on, reduces the number of simulations that must be created and evaluat ed. For small data sets, where there is significant uncertainty, this process provides for a more complete assessment of the potential varia bility of the subsurface and of flow paths for contaminants, given the available data. Such assessment can be used to guide the data collect ion program and the decision-making process.