Probabilitic Analysis of Reservoir Landslides Considering the Spatial Variation of Seepage Parameters under the Conditions of Rainstorm and Sudden Drop of Water Level

Traditional probabilitic analyses of landslides do not take into account the influence of the spatial variability of hydraulic conductivity of landslide mass. To characterize the effect of the spatial variability of the hydraulic conductivity of landslide mass, this paper proposes a back-propagation neural network-based method for slope reliability analysis involving spatially variable soil parameters. The Karhunen-Loève series expansion method is used to discretize the non-Gaussian random field of the saturated hydraulic conductivity of landslide mass. The back-propagation neural network is adopted to construct the surrogate model of the factor of safety of a spatially variable slope. The Baishuihe landslide is investigated as an example to estimate the landslide probability caused by the rainstorm and sudden drop of reservoir water level, respectively. The effectiveness of the proposed method is demonstrated through comparisons with other methods. The results indicate that the proposed method can not only effectively account for the influence of the spatial variability of the hydraulic conductivity of landslide mass on the landslide probability, but also achieve high computational efficiency for the probabilitic analysis of reservoir landslides. It can provide an effective and versatile tool for the landslide probability evaluation. In addition, when the spatial variability of soil hydraulic conductivity is considered, the Baishuihe landslide has a 19.5% probability of local failure under five consecutive days of rainstorm, while it has quite small occurrence possibility of local failure under the sudden drop of reservoir water level. © 2024 China University of Geosciences. All rights reserved.