Application of an improved non-stationary random field model in the random seismic response analysis of a rebuilt landslide dam

Landslide-dammed blocks can be rebuilt into permanent water-retaining dams, whose dynamic responses are crucial for dam safety. However, available studies regarding the dynamic behaviours of landslide dams are deterministic, neglecting the spatial heterogeneity of depth-dependent soil properties. Based on random field (RF) theory, here, the significance of the probabilistic analysis of seismic responses is highlighted. The world’s first immediately rebuilt landside dam is explored, and a new and innovative non-stationary RF is proposed to properly characterize depth-dependent soil properties. Compared with conventional stationary and non-stationary RF simulations, simulations of the new non-stationary RF, for which the mean value increases with depth while the variance decreases, demonstrate the superiority and feasibility of the recommended method. The dynamic responses of landslide dams are jointly affected by the structural characteristics of the dam body and the non-stationary RF property of the maximum dynamic shear modulus. The means and variances of the horizontal peak acceleration and displacement increase with elevation, and those on the dam slope surface are generally greater than those in the dam interior at the same elevation. The proposed non-stationary RF can effectively decrease the variance and discreteness of dynamic responses, providing a more accurate probability analysis for dam security.