Latin Hypercube Sampling of Random Hydraulic-Conductivity Fields for Monte-Carlo Analysis of Groundwater Flow and Mass Transport

The accommodation of uncertainty in hydraulic conductivity, an issue that is at the root of the analysis of monitoring network design, normally requires the use of Monte-Carlo analysis in combination with groundwater flow and transport simulators. The algorithm used for the selection of realizations of the hydraulic conductivity field dictates, in large part, the amount of computational effort required in the analysis. Indeed, the amount of computational effort is a linear function of the number of realizations. Latin hypercube sampling is a strategy that can be shown to require fewer realizations to achieve a given level of accuracy than is required using either GSLIB or the turning bands methods. The method can be shown to be a cost-effective alternative to currently used approaches.