Rainfall-Induced Probability of Failure in Spatially Variable Soil Slopes and a Case Study of the Konkan Railway Slope Failure

Abstract

A novel framework was developed for the stability analysis of spatially variable slopes under rainfall by approximating the performance function in standard normal space by a second-order surface at the design point. Saturated hydraulic conductivity (ks) and normalized rainfall intensity (Q) were considered in the shear strength equation. The effects of drying and wetting soil water characteristic curves (SWCCs) were also considered. The most conservative model was proposed out of the five hydraulic conductivity models considered in the analysis. The concepts of spatial variability were then applied to the second-order reliability method (SORM). The random fields, gamma for stability number (c′/γH), lognormal for effective friction angle (ϕ′), nondimensional inverse of air entry values (αγH), and slope of SWCC (n), and inverse Gaussian rainfall intensity (Q) were generated for clayey slopes. The results revealed that the percentages difference (ε) in reliability index (βSORM) obtained using drying and wetting SWCCs are 19.39% and 37.17% for Q=0 and 0.5, respectively. The findings also illustrated a considerable shift in the critical slip surfaces (CSSs) due to the change in rainfall intensity. Practically, the vertical fluctuation scale was estimated more precisely than the horizontal one. The results show that the influence of vertical autocorrelation distance (ACD) on the reliability index is more significant when compared with horizontal ACD. However, it is noted that the influence of vertical ACD (δy) greater than 3.0 times the height of the slope on βSORM stabilizes. Also, the spatial variations in soil properties along the horizontal direction are relatively more uncertain due to the heterogeneity of soil deposits, sampling limitations, and difficulty of characterizing large-scale spatial variations. This uncertainty in the horizontal variability should be given due consideration in addition to vertical variability because it may have a greater impact on the estimation of the reliability index and overall behavior of the slopes. A parametric study was conducted to estimate the anticipated performance levels of rainfall-triggered slope failures. Additionally, a case study of slope failure triggered by rainfall in the Konkan region of Maharashtra in India is presented at the end. It is noted that the Konkan slope may not be stable when Q=0.83.