Sensitivity of time dependent loads for dike failure : revetment & geotechnical failure in the tidal rivers

Time aspects of hydraulic loads, such as storms, winds, and river floods, play a role in the failure of revetments and geotechnical aspects of flood defenses. For instance, a longer lasting load increases the damage and probability of failure. This study explores the impact of time aspects, including storm and flood durations and phase shifts on the probability of revetment and geotechnical failures. This information helps to decide if and how these time aspects should be included in future probabilistic models. The focus is on levees along the tidal-influenced sections of the Rhine and Meuse Rivers. Revetment mechanisms considered are grass on the outer slope (GEBU) and crest and inner slope (GEKB). Geotechnical failure is analyzed for backward erosion piping (STPH).
The sensitivity of revetment and geotechnical failure to time aspects is evaluated using importance factors (𝛼2-values) from a probabilistic model, which describe the relative contribution of each stochastic variable to the failure probability. This analysis is conducted for six locations along the tidal reaches for revetments and four locations with varying dike or subsoil properties for piping.
Sensitivity analyses are also performed to assess how the results depend on several assumptions.

Regarding revetments, previous studies indicated that including (deterministic) load evolutions, especially for GEKB, yields significant improvements in failure probabilities compared to the WBI-2017 method. This study introduces time aspects as stochastic variables and demonstrates that time aspects have similar importance factors compared to strength variables. Future research into strength parameters may result in smaller strength uncertainties. Furthermore, future studies into time aspects will most likely result in an increase of its importance since the effect on local peak water levels has been neglected for now. This indicates the relevance of time aspects. The use of characteristic values for time aspects often provides sufficiently accurate probability of failure estimates (up to a factor 3 overestimation compared to fully probabilistic), but the chosen value strongly depends on the location, orientation, and characteristics of the dike, such as crest height and lower limit of grass revetment. Based on the current results, a single characteristic value for time aspects for all situations cannot be determined. This could be overcome by broadening the set of cases for a better prediction of characteristic values or by introducing the time aspect as a stochastic parameter in the load model, increasing the number of hydraulic production simulations.
For piping, the study reveals that the time-dependent pipe growth model has a substantial impact on failure probabilities compared to the stationary WBI-2017 model (the Sellmeijer model). This impact is more significant than that of time variables in hydraulic loads. The importance of time variables increases towards the sea, particularly for cases with long seepage length or low permeability, where piping failure requires more time. Uncertainty in strength variables (permeability, grain size) is relatively large for piping, which results in a lower importance of time variables. A deterministic surge duration of 12 hours is deemed sufficiently conservative for piping, leading to an overestimation of failure probability depending on location and subsoil, up to a factor of 5.5.
Recommendations based on the results include implementing the phase difference between tide and surge as a stochastic variable with three discrete values. For revetment failure, treating surge and wind durations as stochastic variables is deemed most effective, considering that their impact strongly depends on dike characteristics. Other time variables (phase wind-surge, river discharge durations) can be treated as deterministic values, equal to the WBI2017 implementation. The study concludes with suggestions for further research, including additional in-depth analyses and improvements in computational methods.