Probabilistic analysis of backward erosion piping in an embankment dam considering the spatial variability of soil properties

The spatial variability of soil seepage parameters is an essential component of the uncertainty of earth dams. This paper presents a method for probabilistic backward erosion analysis that considers the spatial variability of seepage parameters in earth dams to assess the distribution of the Factor of Safety (FoS) and the failure probability ( P _f ). In this study, the Karhunen-Loeve method is used to generate Random Fields (RFs) of soil hydraulic properties, and Monte Carlo simulations (MCs) are used to generate realisations of these RFs, which are then introduced into the numerical model to assess the spatial variability of the seepage analysis results. The hydraulic gradient distribution is compared with the critical hydraulic gradient to assess the FoS and the P _f due to backward erosion at each point of the dam. This study also presents the effect of water level H , permeability anisotropy coefficient ξ , and RFs parameters (permeability covariance COV _Ks , correlation length L _h & L _v ) on the FoS and P _f of Backward Erosion Piping (BEP). The results illustrate that the failure probability P _f increases with H , COV _Ks and ξ . The failure probability P _f is not sensitive to the vertical correlation length, but may decrease with the horizontal correlation length. Additionally, the study discusses on the potential location of the initiation of internal erosion as well as the results obtained with criteria based on the local and global hydraulic gradient.