Soil stochastic parameter correlation impact in the piping erosion failure estimation of riverine flood defences

•The inclusion of correlation between grain size and hydraulic conductivity results in a more accurate piping failure estimation.•Clayton copula represents better the tail dependence between d70 and K.•An error factor of 1.84 was obtained when neglecting correlation in the present study. Piping erosion has been proved to be one of the failure mechanisms that contributes the most to the total probability of failure on the Dutch flood defence systems. The present study aimed to find the impact of correlation and tail dependence between soil parameters present in the Sellmeijer revised limit state equation for piping safety assessment, particularly between the grain size and hydraulic conductivity parameters. A copula based random sampling method was used as a tool to include this effect in the probabilistic estimation of this type of failure. The method was framed in a real case study for a flood defence along the Lek river, in The Netherlands. The results showed that inclusion of correlation between the two parameters reduces the variance of the limit state marginal distribution by almost 10% when compared to the uncorrelated case. This effect changes the tail values sampling frequency and therefore reduces the probability of failure by a factor of 1.7. The omission of correlation between the two parameters for safety assessment based on the Sellmeijer limit state function may result in over dimensioned structures.