Numerical Modeling of Bank Instability by Seepage Erosion Undercutting of Layered Streambanks

Undercutting, primarily considered due to fluvial mechanisms, has been reported to have a major impact on slope failure. Predicting bank collapse specifically due to seepage erosion undercutting by particle mobilization on layered streambanks has not been fully studied or modeled, even though its role in streambank erosion may be important. The limitation originates from the limited field measurements or laboratory experiments as well as the unavailability of discrete element models that can effectively simulate seepage particle mobilization, undercutting, and the corresponding mass wasting. The objective of this research was to demonstrate a procedure for incorporating seepage undercutting into bank stability models and to investigate the role of seepage undercutting on bank instability. The question to be addressed is whether seepage particle mobilization can lead to distances of undercutting that are a significant cause of bank instability. A numerical finite-element model, SEEP/W, was used to model soil-water pressure variations during seepage observed in laboratory experiments with two-dimensional soil lysimeters. Flow parameters were calibrated using measured soil-water pressure and cumulative discharge. A general limit equilibrium bank stability model (SLOPE/W) was used to simulate bank stability with and without seepage erosion undercutting by comparing the computed factor of safety, $\mathit{Fs}$ , at different stages of the seepage erosion process with regard to input parameter uncertainty using Monte Carlo analysis. The percentage decrease in the mean $\mathit{Fs}$ ranged between 42 and 91% as the depth of undercutting increased, dependent upon the initial stability of the bank. A stable bank (i.e., $\mathit{Fs}>1$ ) can quickly become unstable (i.e., $\mathit{Fs}<1$ ) when seepage undercutting is considered. For stable banks, the probability of failure reached 100% when the depth of the undercutting reached approximately $30\text{to}50\mathrm{mm}$ under these experimental conditions. The results derived are specific to the streambank simulated but are expected to be comparable for similar layered streambank lithologies reported to occur in numerous geographical locations. This research also highlights the need to incorporate the dynamic process of seepage erosion undercutting into integrated subsurface flow and streambank stability models.