This paper provides a summary of initial research results investigating systematic site effects from the prediction residuals of empirical- and physics-based ground-motion models (GMMs) for small magnitude (i.e., 3.5 ≤ MW ≤ 5) active shallow crustal earthquakes in New Zealand (NZ). Advancing ground-motion predictability through physics-based GMMs is an iterative process and requires addressing fundamental questions like: Is there salient physics which has been overlooked? Which geographic regions have predictions that significantly deviate from observations and why? Which sites exhibit systematic prediction residuals and how can the attributes influencing them be identified? This preliminary study examines these questions by classifying 171 sites from the Canterbury and Wellington regions into four geomorphic categories: basin, basin-edge, hill, and valley, following the categorisation by Nweke et al. (2022). Trends in the site-to-site residuals for each geomorphic category indicate apparent differences between the four categories, with residuals for valley sites illustrating a clear dependence with the inferred fundamental site period. Computed residuals from both empirical- and physics-based GMMs also provided insight into the role of site-specific attributes vs. the different prediction methods, assisting to understand the salient causes of these residuals.
One of the current challenges in physics-based ground-motion simulations is to refine the modeling of local site effects. These effects require a finer spatial resolution in the material modeling than that generally considered in regional-scale simulations. Because of this, empirical amplification factors are typically applied to capture these unmodeled phenomena. The ergodic nature of this approach suggests that there is room for improvement. In this study, the predictive capability of simulations is evaluated using alternative methods for capturing local site effects. In addition to the conventional empirical approach, two methods are examined that allow for more site-specific information to be incorporated: the square-root impedance method and the 1D time domain site-response analysis. The three approaches are tested using 1000+ observed ground motions from 150+ small-magnitude events (3.5 ≤ Mw ≤ 5.0), recorded at 20 strong-motion stationsin the Canterbury, New Zealand, region. These 20 well-characterized sites represent a wide range of soil conditions, including stiff gravels with Vs30 values greater than 500 m/s, and sand and silt deposits with Vs30 valuesless than 200 m/s. Multiple intensity measures are computed and prediction residuals are partitioned using mixed-effects regression to rigorously assess the relative performance of the different approaches considered. The results indicate that the benefit of using more sophisticated methods is highly dependent on the characteristics of the site. Key site parameters and trends are identified and discussed in light of the assumptions and limitations of each approach.
The seismic performance of soil profiles with potentially liquefiable deposits is a complex phenomenon that requires a thorough understanding of the soil properties and ground motion characteristics. The limitations of simplified liquefaction assessment methods have prompted an increase in the use of non-linear dynamic analysis methods. Focusing on onedimensional site response of a soil column, this thesis validated a soil constitutive model using in-situ pore pressure measurements and then assessed the influence of input ground motion characteristics on soil column response using traditional and newly developed metrics. Pore pressure recordings during the Canterbury Earthquake Sequence (CES) in New Zealand were used to validate the PM4Sand constitutive model. Soil profile characterization was key to accurate prediction of excess pore pressure response and accounting for any densification during the CES. Response during multiple earthquakes was captured effectively and cross-layer interaction demonstrated the model capability to capture soil response at the system-level. Synthetic and observed ground motions from the Christchurch earthquake were applied to the validated soil column to quantify the performance of synthetic motions. New metrics were developed to facilitate a robust comparison to assess performance. The synthetic input motions demonstrated a slightly larger acceleration and excess pore pressure response compared to the observed input motions. The results suggest that the synthetic motions may accumulate higher excess pore pressure at a faster rate and with fewer number of cycles in the shear response. This research compares validated soil profile subject to spectrally-matched pulse and non-pulse motions, emphasizing the inclusion of pulse motions with distinctive characteristics in ground motion suites for non-linear dynamic analysis. However, spectral matching may lead to undesired alterations in pulse characteristics. Cumulative absolute velocity and significant duration significantly differed between these two groups compared to the other key characteristics and contributed considerably to the liquefaction response. Unlike the non-pulse motions, not all of the pulse motions triggered liquefaction, likely due to their shorter significant duration. Non-pulse motions developed a greater spatial extent of liquefaction triggering in the soil profile and extended to a greater depth
