Incorporating Uncertain Site Information into Groundwater Modeling

Proper site characterization is vital for cost effective design of a remediation system. Scarcity of data, spatially and temporally varying site conditions and cost associated with sampling, lead to decisions being made under uncertainty which may be categorized as stochastic, lexical or informal. Probability theory, developed for stochastic uncertainties, is not suitable for representing lexical and informal uncertainties. In this paper, a methodology is developed to represent different types of uncertain information and propagate the uncertainties through a groundwater flow model. The methodology includes using probability theory to represent random uncertainties and possibility theory for nonrandom uncertainties . A probability-possibility transformation is used to convert the probability density function to possibility distribution. Alpha cuts are used to convert the governing PDE to one with interval parameters and sparse grid interpolation is used to reduce the computational efforts.