Abstract
It is of great significance to employ a statistical approach to the seepage problem with spatially high variability of soil properties. In our earlier work, it has been demonstrated that with the increasing standard deviation of permeability coefficient the mean seepage flow rate decreases and its covariance increases, and this is a result of spatial spanning of lower or higher permeability. Using the percolation theory, this paper attempts to reproduce such spatial spanning that forms "impervious wall" or "flow path". Spatially correlated random permeability fields are generated, and the finite element seepage analysis is carried out for each of these fields to examine percolation properties. As a result, it is shown that the indexes of percolation (i. e., threshold and threshold probability), defined in the percolation theory, could characterize the total flow rate in random permeability field with the same means, standard deviation and correlation distance.
