The effects of soil behaviour models on the reliability of stability calculations in 2-D LEM model for dikes in the Dutch context

Abstract

Thousands of kilometres of dikes cover the Dutch coastline and river shores, in order to protect the densely populated (hinter)land from high tide and floods. As a consequence of economic and demographic development, but also climate change, the safety and design standards of a flood defence need to be updated every fifty years or so. Macro-instability, which is the loss of structural integrity due to high water, is one of the significant failure mechanisms a dike must be evaluated for.

For this research a case study is assessed by means of a 2-D LEM (Limit Equilibrium Methods) model (i.e. D-Stability): for sixteen schematised dike sections that are part of the river dike between Wijk bij Duurstede and Amerongen (WAM) this programme is used to determine a stability factor. The geotechnical parameters (i.e. soil strength and volumetric weight) that serve as input for these calculations, have already been established by local field work and laboratory experiments (i.e. triaxial compression tests and direct simple shear tests). As the soil reacts differently depending on the stress it is exposed to, and therefore the sustained axial strain and deformation, different standards can be assumed. For this study three soil behaviour models are considered, i.e. Mohr-Coulomb (low axial strain), peak strength (maximum soil strength), and critical state (high axial strain).

As could be expected, the relatively high soil strength parameters for Mohr-Coulomb and peak strength result in much higher stability factors (average of respectively 1.9 and 1.7), than the ones for the conservative critical state (average of 1.3). This correlation has been confirmed by all case study results, comprising three selected sliding surface shape models (i.e. Bishop, UpliftVan and Spencer). The primary aim of this research was however not only to analyse these three different soil behaviour models, but also to find out which one is able to produce the most reliable dike macro-stability results. By virtue of the Probabilistic toolkit for each dike section and stability factor, concomitant reliability indices were calculated, which are established on the standard deviation of the geotechnical parameters. The relation between the stability factor and the reliability index is presented for each soil behaviour model. Based on only the results of this study case, there is no reason to favour any soil behaviour model over critical state when assessing dike macro-stability: for a significant stability factor it produces both the highest reliability index, and a much smaller uncertainty range. Due to the limited amount of soil samples, and even more important, the lack of heterogeneity of the geometry, lithology and geotechnical parameters, additional research is recommended, in order to obtain a complete overview of the relationship between stability and reliability, and to reduce the uncertainty that comes along with each soil behaviour model.