Liquefaction assessment of cohesionless soils in the vicinity of large embankments

In a typical seismic dam safety evaluation, standard penetration, cone penetration, Becker penetration, or shear wave velocity (Vs) tests are often first conducted near the toe of an earth dam to infer if any liquefiable soil exists in the foundation of the dam footprint. In current practice, a level-ground condition is commonly assumed when normalizing penetration resistance and Vs, and may be assumed (particularly in preliminary assessments) in applying the cyclic stress method (with or without the Kα correction) to evaluate liquefaction. However, the presence of an earth dam, or any other large embankment or structure, significantly alters the normal and shear stresses in the foundation. This paper identifies and quantifies potential errors in ignoring altered stresses near heavy structures, and presents a methodology to incorporate these effects within the framework of the simplified procedure. Specifically, the effects of these altered stresses (in comparison to the level-ground assumption with and without Kα correction) on the: (1) normalization of field measurements such as penetration resistance and Vs; (2) cyclic stress ratio (CSR); (3) cyclic resistance ratio (CRR); and (4) factor of safety against liquefaction triggering (FSliq), are evaluated by considering static and dynamic analyses of a generic earthen embankment (60 m high) resting on a saturated, cohesionless foundation (30 m deep). Our analyses indicated that ignoring the presence of induced static shear stresses can result in potentially unconservative errors in overburden correction factors of 30% to 60% at shallow depth (although this error is greatly muted at depths exceeding about 15 m), while errors in CSR potentially can range from about 20% too conservative to 40% unconservative. Potential errors in CRR can approach 50% unconservative at shallow depths, but again, this error is muted at depths exceeding about 15 m. Combining these factors, potentially unconservative errors in computing FSliq could exceed 100% at shallow depths (less than 15 m to 20 m) while at greater depth (exceeding 20 m) errors approach 20% on the conservative side.