Author(s): Yingqi Zhang

Linked Author(s):

Keywords: Random field; Latin hypercube sampling; Rank correlation

Abstract: A new approach is proposed to generate realizations for a random field which will be used as the input to a stochastic groundwater flow and transport model. To accurately evaluate hydraulic head and concentration variability, a large number of simulations have to be run with a large number of realizations, which can be a large computational burden. The proposed method combines a Latin Hypercube sampling technique with rank correlation, thereby, reducing the number of realizations needed to achieve a certain degree of accuracy. The Latin Hypercube Sampling technique is a stratified sampling method which takes more samples in the high distributed area and less samples in the low distributed area for a given probability distribution curve. At the same time we take into consideration the autocorrelation of the field. As a consequence of a comparison of the GSLIB algorithm and Turning Bands method, we can conclude that this method can be a cost-effective alternative to currently used methods.

DOI:

Year: 2001