Study On The Stability Of Soil Slope Under Rainfall Infiltration-redistribution

Accurate prediction and the mitigation of landslide hazards are always one of the major strategies for governments around the world; and one of them, the studies of rainfall-induced landslides is the hot topic. Amount of research on the rainfall-induced landslides has been conducted, and substantive progress achieved. Unfortunately, the current studies are almost focused on analyzing the stability of the shallow landslide during the process of precipitation. And the corresponding research method is deterministic which the natural variability of soil parameters is not considered. Based on the landslide hazard analysis, the study of slope stability in the process of the rainfall infiltration-redistribution for the homogenous and two-layered soil slope is carried out using numerical analysis and physically based hydrologic model. Moreover, the effects of the variability of the saturated hydraulic conductivity and its trend component on the probability of failure are also investigated. The main contents and conclusions are lists as follows:1. A general conceptual model of rainfall infiltration-redistribution is presented using the classic Green-Ampt infiltration model for the homogenous and two-layered soil slope. The comparison results show that the redistribution of moisture content obtained from the model matches quite well with the numerical solution. The results demonstrate that:(1) previously infiltrated rainwater continues to move down under the action of its own gravity and matric suction in the slope depths. The wetting front also continues to move down as the rainwater redistributes, which results in the loss of matric suction in the original unsaturated soil. The validated rainwater redistribution model can explain the reasons for shallow landslides occurring after a rainfall event from the perspective of rainwater redistribution theoretically. And in the process of rainwater redistribution, the moisture content in the upper fine layer and the surface of the lower coarse layer decrease monotonously with time and eventually become flat.(2) the factors of safety of the slope in the redistribution process depend on the duration of rainwater redistribution and the antecedent accumulative infiltration, and decrease gently and tend toward stability when the accumulative infiltration increases to a certain value. It also illustrates that the antecedent accumulative infiltration is a leading cause that triggers landslides in the process of rainwater redistribution.2. The saturated hydraulic conductivity (Ks) appears to exhibit the strong variability and define percolation gradients inside the soil mantle. Therefore, the probabilistic framework incorporating the variability of the saturated hydraulic conductivity is presented based on the Monte Carlo method. Extend the classic Green-Ampt model to a new rainwater infiltration-redistribution model that can address the problems of two different infiltration boundaries caused by the variability of Ks and rainwater redistribution and establish a closed form of the limit state function combined with an infinite slope stability model. The following conclusions are drawn from the analyses:(1) a critical duration rainfall exists during the rainfall, where the probability of failure for the hypothetical unsaturated soil slope will increase sharply whether it is in the infiltration phase or the redistribution phase. Meanwhile, cases with a smaller coefficient of variation of Ks have a smaller probability of failure in the initial phase of rainfall infiltration. However, when the duration of rainfall lasts long enough, the cases with a smaller coefficient of variation of Ks have a larger probability of failure. More importantly, the most likely failure time of the slope is not affected by the coefficient of variation, but depends on the intensity of rainfall.(2) the probability of failure for the soil slope increases with the increasing rainfall intensity. The distribution zone of the failure time is relatively wide for the cases with the low rainfall intensity, which indicates that rainfall-induced landslides are more likely in the cases with lower rainfall intensity and longer duration of rainfall.(3) the duration of antecedent rainfall has a major influence on the failure probability of the slope during the period of rainwater redistribution and determines the corresponding lag times.(4) it may be favorable to avoid using mean safety factors to evaluate the stabilities for such rainfall-induced landslides because of its risk. 3. The local average method is adopted to simulate the corresponding one-dimension nonstationary random field of spatial varying permeability with FISH code on the platform of FLAC. The process of rainfall infiltration is calculated by the module of Two Phase Flow. With Monte Carlo Method as a template, combine the above seepage analysis with an infinite slope stability model considering the effects of positive pore pressure on the safety factor. Then the spatial variability of saturated hydraulic conductivity and its trend component how to influence the distribution of the critical failure surface and the failure probability under different duration of rainfall is investigated. The results show that:(1) the critical failure surface is not fixed and continuously changes as the rainfall progresses. The percentage of critical failure surfaces those are located at the impermeable layer decreases as the duration of rainfall increases. However, the critical failure surface must be located at the bottom of the slope when the rainfall lasts long enough.(2) the trend component of the saturated hydraulic conductivity (Ak) has an important effect on the distribution of the critical failure surface. More importantly, ignoring the effects of the trend component leads to less conservative estimates of the failure probability.(3) the proportion of critical failure surfaces that are located at the impermeable bedrock layer increases with an increase of Co V(KS)(the coefficient of variability) and lv (the vertical correlation length). However, if the duration of rainfall is sufficient, the proportion decreases with increasing of CoV(Ks) and lv.