Seismic Soil Liquefaction Triggering Assessment of Gravelly Soil Case Histories

While the liquefaction susceptibility of fully saturated clean sands and their mixtures with low-plasticity fines has long been recognized in liquefaction engineering, gravelly soils were traditionally considered non-liquefiable due to their high pore pressure dissipation capacity, attributed to their larger grain sizes. However, with the increasing construction of gravelly earth fills, concerns about their liquefaction susceptibility have grown. Surprisingly, in these man-made earth fills, loose states were observed in some gravelly soil layers. The 2008 Wenchuan earthquake (M_w = 7.9) provided case histories revealing that natural gravelly soils can also exist in a loose state and, furthermore, they can liquefy. To assess this, probabilistic case history-based, liquefaction-triggering relationships have been developed. This paper discusses 65 available case histories of gravelly soil liquefaction, compiled from four earthquakes with magnitudes ranging from 6.5 to 9.2. It also explores the advantages of using alternative in situ testing methods such as the dynamic penetration test (DPT), large penetration test (LPT), Becker penetration test (BPT), and geophysical surveys like shear wave velocity (V_s) rather than relying solely on the standard penetration test (SPT) and cone penetration test (CPT). Simplified procedures are employed to assess the liquefaction-triggering potential of gravelly soil layers. The assessment process begins with identifying the susceptible critical layer, estimating in situ density, groundwater table depth, seismic intensity, and duration parameters, and considering grain size gradation characteristics, with an emphasis on addressing their uncertainties. The assessment results are presented in the cyclic stress ratio (CSR) vs. DPT blow count (N′₁₂₀) domain. Based on the developed case history database, probability-based liquefaction-triggering boundary curves have been established for engineering practice. Finally, these curves are compared with existing probabilistic models found in the literature.