This study investigates the applicability of simplified semi-empirical and advanced dynamic liquefaction assessment methods to reclaimed soils using a comprehensive set of data for two types of reclamations at the port of Wellington, New Zealand (CentrePort) that are challenging for liquefaction assessment: end-dumped gravelly fills and hydraulically placed silty-sandy fills. The gravelly fills are the primary focus of this study as they are comprised of complex gravel-sand-silt (G-S-S) mixtures that are not well-represented in current liquefaction databases used to develop existing semi-empirical methods for liquefaction assessment. As such, existing procedures to evaluate triggering and consequences of liquefaction may not be directly applicable or may need additional considerations when applied to the gravelly mixtures encountered at CentrePort. The hydraulic fills are also of significant interest as they relate to a range of issues in the simplified engineering assessment around effects of fines and their plasticity on the liquefaction resistance.

The detailed site characterization in the first part of this study, primarily based on CPT data and supplemented by borehole logs and index testing of borehole soil samples, show that the 10–22 m thick fill in the southern end of the port (i.e., the Thorndon reclamation) is composed of 60–80% fine- to-medium gravels and 20–40% finer sand-silt fractions. The proportion of sand and non-plastic silt in the G-S-S mixtures is sufficiently large for these finer fractions to govern the deformational behaviour and mechanical response of the matrix, as also suggested by the CPT characteristics (qc = 6–8 MPa; Ic = 1.9–2.3). The G-S-S fills in other parts of the port have similar CPT characteristics, however, the G-S-S layers are much thinner and contain several interbedded layers of non-liquefiable soils or soils with higher density. The hydraulic fill varies in thickness from 5 m to 10 m and is characterized by three distinct soil units. The sand-silt hydraulic fill (sands with 5–35% fines; qc ≈ 4.5 MPa; Ic ≈ 2.1) are most commonly encountered in the Log Yard, the silt-clay hydraulic fill (sands with 70–100% fines with mostly PI > 20; qc < 2 MPa; Ic > 3.0) is commonly encountered along Aotea Quay, and G- S-S mixtures (FC = 5–25%; GC = 15–55%; qc ≈ 5 MPa; Ic ≈ 2.0) are encountered in one location at Aotea Quay.

An existing CPT-based simplified liquefaction evaluation method is applied to the CentrePort profiles and compared against liquefaction performance observed in three recent earthquakes. There is generally good agreement in the triggering assessment for seismic demands well above or below the liquefaction triggering threshold for the fills, but overestimation for seismic demands close to the triggering threshold. While the triggering methods themselves are unable to clearly discern between different performances, key differences in the thicknesses and locations of liquefied fills are better indicators of the liquefaction-induced damage and potential for manifestation at the ground surface. Simplified estimates of settlements perform well for the cases where liquefaction triggering results are generally in agreement with observed damage, though with slight underestimation. Calculation of damage indices also matched reasonably well with the general trends in the severity of damage observed for all three earthquakes across most of CentrePort. However, the damage indices provide lesser degree of variation in the ground performance as compared to actual observations.

Two key issues in the simplified assessment of reclaimed fills are scrutinized. Firstly, uncertainty in the interpretation of material characterization on the triggering assessment are investigated through a sensitivity study. In particular, sensitivity of the computed response to uncertainties associated with effects of fines or complex soil composition on the liquefaction resistance (via FC or Ic parameters in the Boulanger and Idriss (2014) and Robertson and Wride

(1998b) methods, respectively) are scrutinized in detail. These two triggering methods can result in over 50% difference in estimates of liquefaction resistance in the CentrePort fills (characterized by Ic = 2.0–2.3) due to modelling uncertainty in the material characterization. Results show that well- defined critical layers have low penetration resistance, and that sensitivity of liquefaction resistance to material parameters is the smallest for such layers, hence the uncertainty in the cyclic demand tends to dominate. However, the triggering methods are much more sensitive to the material characterization in soil layers with much larger penetration resistance. The second issue investigated is the grain-size effects on penetration resistance and associated estimates of settlements via relative density relationships. Sand-based procedures for evaluation of liquefaction-induced settlement are found to be generally applicable to well-graded gravels that have a dominant silty sand fraction in the soil matrix, though they can significantly overestimate the relative density and consequently underestimate post- liquefaction settlement of gravelly soils in the case of dense fills. Results from investigation of both issues above imply that the overall sensitivity in the simplified assessment is density dependent.

Finally, a series of preliminary advanced dynamic effective stress analyses are performed for 13 representative profiles at CentrePort. Liquefaction resistance of advanced constitutive models are calibrated using CPT-based empirical relationships from simplified procedures, and hence the advanced numerical simulations of this study are consistent with the simplified liquefaction assessment in this regard. This facilitates comparative evaluation of the two approaches, and differences in performances reflect differences in modelling and analysis methodologies rather than in the liquefaction resistance characteristics. Overall, the dynamic effective stress analyses provide greater insight on the mechanisms of liquefaction, explains important dynamic interactions within the deposit, and explores the evolution of the system response for different seismic intensities. The analyses also explain some of the discrepancies between the observed performance and simplified methods. Comparison of damage indices calculated for the simplified methods with damage proxies from the dynamic simulations show reasonable agreement in the relative liquefaction performance for most sites across CentrePort, with some discrepancies identified.