Search for: All records

Physics-based ground motion simulations are a valuable tool for studying seismic sources with missing historical records, such as Cascadia Subduction Zone (CSZ) interface earthquakes. The last such event occurred in 1700 CE and is believed to be an M8-M9 rupture. The United States Geological Survey recently developed 30 physics-based simulations of a CSZ rupture to predict ground motions across the Pacific Northwest. Consideration of key modeling uncertainties across these simulations leads to estimates of ground motion intensity that vary by ~100% in some areas (e.g., Seattle). Paleoliquefaction, or soil liquefaction from past earthquakes, provides the best geologic evidence for constraining or "ground truthing" the intensity of past shaking, yet while paleoliquefaction has been documented throughout Cascadia, limited analyses have been performed to exploit this evidence. This study focuses on Kellogg Island, 2 mi south of Seattle, where liquefaction has been documented from several earthquakes, but not from the 1700 CE event. Therefore, using the CSZ simulations and in situ cone penetration test data, this study predicts the probability of surficial liquefaction manifestation at Kellogg Island during an M9 CSZ event. As part of this effort, velocity profiles are developed from multichannel analysis of surface waves, and non-linear site response analyses are used to propagate simulated motions to the surface. Results show a high probability of liquefaction near Kellogg Island for most simulations, whereas to date no evidence of 1700 CE liquefaction has been discovered at Kellogg Island, nor at any other location in the Puget Sound. The discrepancy between predictions and observations might indicate that the 1700 CE ground motions were less intense in Seattle than most predictions of M9 earthquakes indicate. Toward the goal of elucidating the expected impacts of future CSZ earthquakes, similar analyses are ongoing at additional sites across the region. 

Abstract The shear-wave velocity time averaged over the upper 30 m (VS30) is widely used as a proxy for site effects, forms the basis of seismic site class, and underpins site-amplification factors in empirical ground-motion models. Many earthquake simulations, therefore, require VS30. This presents a challenge at regional scale, given the infeasibility of subsurface testing over vast areas. Although various models for predicting VS30 have thus been proposed, the most popular U.S. national, or “background,” model is a regression equation based on just one variable. Given the growth of community data sets, satellite remote sensing, and algorithmic learning, more advanced and accurate solutions may be possible. Toward that end, we develop national VS30 models and maps using field data from over 7000 sites and machine learning (ML), wherein up to 17 geospatial parameters are used to predict subsurface conditions (i.e., VS30). Of the two models developed, that using geologic data performs marginally better, yet such data are not always available. Both models significantly outperform existing solutions in unbiased testing and are used to create new VS30 maps at ∼220 m resolution. These maps are updated in the vicinity of field measurements using regression kriging and cover the 50 U.S. states and Puerto Rico. Ultimately, and like any model, performance cannot be known where data is sparse. In this regard, alternative maps that use other models are proposed for steep slopes. More broadly, this study demonstrates the utility of ML for inferring below-ground conditions from geospatial data, a technique that could be applied to other data and objectives. 

The severity of surficial liquefaction manifestation was significantly over-predicted for a large subset of case histories from relatively recent earthquakes that impacted the Canterbury region of New Zealand. Such over-predicts generally occurred for profiles having predominantly high fines-content (FC), high-plasticity soil strata. Herein, the liquefaction case histories from the Canterbury earthquakes are used to investigate the performances of three different manifestation severity index (MSI) models. The prevalence of high FC, high-plasticity strata in a profile is quantified through the soil behavior type index averaged over the upper 10 m of a profile ( I_c10). It is shown that for each MSI model (1) the threshold MSI value distinguishing cases with and without manifestation increases as I_c10increases and (2) the ability of the MSI to segregate cases with and without manifestation decreases with increasing I_c10. Additionally, probabilistic models are proposed for evaluating the severity of surficial liquefaction manifestation as a function of MSI and I_c10. The approaches presented in this study allow better interpretations of predictions made by existing MSI models, although their efficacy decreases at sites with high I_c10. An improved MSI model is ultimately needed that better accounts for the effects of high-FC, high-plasticity soils more directly. 

While soil liquefaction is common in earthquakes, the case-history data required to train and test state-of-practice prediction models remains comparatively scarce, owing to the breadth and expense of data that comprise a single case history. The 2001 Nisqually, Washington, earthquake, for example, occurred in a metropolitan region and induced damaging liquefaction in the urban cores of Seattle and Olympia, yet case-history data have not previously been published. Accordingly, this article compiles 24 cone-penetration-test (CPT) case histories from free-field locations. The many methods used to obtain and process the data are detailed herein, as is the structure of the digital data set. The case histories are then analyzed by 18 existing liquefaction response models to determine whether any is better, and to compare model performance in Nisqually against global observations. While differences are measured, both between models and against prior global case histories, these differences are often statistically insignificant considering finite-sample uncertainty. This alludes to the general inappropriateness of championing models based on individual earthquakes or otherwise small data sets, and to the ongoing needs for additional case-history data and more rigorous adherence to best practices in model training and testing.