A probabilistic approach to levee reliability based on sliding, backward erosion and overflowing mechanisms: Application to an inspired Canadian case study

Improving protection against fluvial floods requires a better estimation of levee failure. We developed an assessment method of levee failure probabilities for sliding, backward erosion, and overflowing each represented by fragility curves. We tested two approaches to aggregate those fragility curves into a global fragility curve respectively using: an enveloping curve and Monte-Carlo simulations. We implemented this approach to earthen levee reliability for several flood return periods to the Bow River in Calgary, Canada. We used limit equilibrium method to estimate the safety factor of the levee segment and Monte-Carlo simulations to estimate sliding probabilities. We used Terzaghi's critical hydraulic gradient to estimate backward erosion failure probabilities. The estimation of overflowing probabilities required expert judgment. We discussed how the choice of the hydraulic gradient area and the consideration of a steady state or transient model impact backward erosion failure probabilities. The results showed for our study case that, even though the transient model is a closer representation of reality, the levee saturation parameter has little impact on hydraulic gradient values, by extension, on sliding and backward erosion failure probabilities. The Monte-Carlo aggregated fragility curve is more realistic than the envelop curve of the failure mechanisms for an equivalent computation time.