Critical and residual states are key soil conditions relevant to slope stability. Evaluating the available shear strength in relation to these conditions is crucial for reliable stability analyses.
However, limited availability of direct measurements in the usual engineering practice makes such evaluation seldomly straightforward. The issue is sometimes alleviated by utilising empirical correlations with soil index properties, more easily determinable, yet much care is needed in applying generic correlations to specific materials, as the related uncertainties might override their predictive capability.
By analysing soil samples from the Dobkovicky landslide in Czech Republic, we discuss the extent to which some established correlations can be misleading, but also that relationships calibrated on site-specific data are not necessarily performing better. We then demonstrate that empirical models may fail to capture correlations if the sample set is rather homogeneous and noise from reasonable experimental uncertainties is introduced.
Finally, we provide some guidelines on the use of different soil strength parameters, laboratory tests, and empirical relationships for slope stability analyses in engineering practice.