Peat is a highly compressible organic material with unique properties that differ from inorganic mineral soils, which poses a challenge in their constitutive modeling. The main specific challenge addressed in this dissertation include matching dynamic properties (i.e., modulus reduction and damping behavior). Constitutive models used in 1D site response typically use modulus reduction and damping curves as input parameters, and usually introduce a misfit of the desired behavior, particularly at high strains. This is problematic for peat because large strains are expected to develop during cyclic loading due to the peat softness.

Nonlinear one dimensional ground response models generally present a compromise between fitting the backbone curve or the hysteretic damping curve. Fitting the damping curve depends on unloading and reloading rules. Most of the models use Masing rules or extended Masing to correct the overdamping at high strains resulting from using Masing rules. Frequency dependent Rayleigh damping is used to introduce damping at low strains. I present a new formulation of unloading and reloading rules completely departing from Masing rules. The main idea is to rotate the axis of the stress strain curve and change the point of reference to calculate the stress at the next time step. The small strain damping is made hysteretic by increasing the initial departure tangent modulus when unloading, in a way consistent with what has been observed in laboratory tests. The unloading-reloading rule is implemented in a nonlinear code and is able to match any backbone and hysteretic damping without Rayleigh damping.

Dynamic curves are typically not used in 2D or 3D models because their inclusion in a plasticity framework is complicated due to their dependence on confining pressure, which can change during earthquake loading (e.g. when excess pore pressure develops under undrained loading). Hence, the damping behavior is not an input of current 3D constitutive models. In order to facilitate the inclusion of dynamic curves in constitutive models, I present a new concept that plots modulus reduction and damping curves against stress ratio instead of shear strain. This results in pressure-independent modulus reduction and damping curves for three empirical relationships commonly-used to derive modulus reduction and damping curves. This finding is useful for implementation in one-dimensional effective stress ground response analysis codes for undrained loading conditions, and in advanced plasticity models.

I then extend the developed unloading-reloading rule and include it in a 3D constitutive model that uses modulus reduction and damping curves that are plotted against stress ratio by using the aforementioned concept. The formulation of the model allows to match dynamic properties (i.e., modulus reduction and damping curves), in 1D and 2D site response. At large strains the strength is controlled by a bounding surface algorithm following the formulation from Dafalias and Manzari (2004). The volumetric response is controlled by a dilation surface that introduces plastic volumetric strains based on deviatoric plastic strains. Most of the input parameters are well-known engineering properties easily measured in laboratory tests. Default values are defined for the input parameters that are not easily measured. I present the implementation of the model in FLAC and some typical predictions of the model through simulations of cyclic triaxial and simple shear tests. Finally, I present the calibration of the model for Sherman Island peat based on laboratory tests, and the performance of the model in 1D site response simulations.

In Part I of this dissertation, equivalent static analysis (ESA) procedures for computing foundation demands during lateral spreading are applied to two parallel bridges that were damaged during the 2010 M 7.2 El Mayor-Cucapah earthquake in Baja California, Mexico. A railroad bridge span collapsed, whereas the adjacent highway bridge survived with one support pier near the river having modest flexural cracking of cover concrete. Cone penetration and geophysical test results are presented along with geotechnical and structural conditions evaluated from design documents. ESA using a beam-on-Winkler foundation model is found to accurately predict observed responses when liquefaction-compatible inertia demands are represented as spectral displacements that account for resistance from other bridge components. Pier columns for the surviving bridge effectively resisted lateral spreading demands in part because of restraint provided by the superstructure. Collapse of the surviving bridge is incorrectly predicted when inertial demands are computed for the individual bent in isolation from other components, and are represented by forces that do not consider global restraint.

In Part II, results of a parametric study of the influence of kinematic pile-soil interaction on foundation-input motions (FIM) are presented. One-dimensional nonlinear ground response analysis was used to define free-field motions, which were subsequently imposed on a beam-on-nonlinear-dynamic-Winkler-foundation pile model. The free-field ground surface motion (FFM) and top-of-pile “foundation-input motion” (FIM) computed from these results were then used to compute transfer functions and spectral ratios for use with the substructure method of seismic analysis. A total of 1,920 parametric combinations of different pile sizes, soil profiles, and ground motions were analyzed.

Results of the study show that significant reductions of the FFM occur for stiff piles in soft soil, which could result in a favorable reduction in design demands for short-period structures. Group effects considering spatially-variable (incoherent) ground motions are found to be modest over the footprint of a typical bridge bent, resulting in an additional reduction of FFM by 10 percent or less compared to an equivalent single pile. This study aims to overcome limitations of idealistic assumptions that have been employed in previous studies such as linear-elastic material behavior, drastically simplified stratigraphy, and harmonic oscillations in lieu of real ground motions. In order to capture the important influence of more realistic conditions such as material nonlinearity, subsurface heterogeneity, and variable frequency-content ground motions, a set of models for predicting transfer functions and spectral ratios has been developed through statistical regression of the results from this parametric study. These allow foundation engineers to predict kinematic pile-soil interaction effects without performing dynamic pile analyses.

Part I of this dissertation describes a simple experimental technique that can be used to obtain the slopes of the plastic potential and yield functions during shear based on the deformation theory of plasticity. The method imposes small perturbations in the direction of the stress increment by closing the drainage valve, thereby abruptly switching from drained to undrained loading conditions during plastic loading. Elastoplastic moduli are obtained immediately before and after the perturbation from the measured deviatoric stress, mean effective stress, deviatoric strains, and volumetric strains for the stress paths immediately before and after closing the drain valve. During drained shear, samples were sheared while the mean effective stress was maintained constant. Combining tests performed at several confining stresses, the proposed method can map yield and plastic potential surfaces and predict their evolution for a wide range of stresses.

Part II of the dissertation focuses on providing insights into the effects of changes in clay mineralogy and pore-fluid chemistry on cyclic behavior of low-plasticity fine-grained soils. A series of cyclic and monotonic direct simple shear experiments was conducted on three low-plasticity fine-grained mixtures of non-plastic silt with either bentonite or kaolinite clay minerals blended with fresh deionized water or saline water. The clay fractions were adjusted to achieve a plasticity index of PI = 9 for all three mixtures to study differences or similarities in their behavior and to examine the effectiveness of index properties in identifying the observed differences in lab results.

Even though all three blends have the same plasticity index, significant differences in their cyclic response were observed. Results indicate that plasticity index is an insufficient indicator of the cyclic behavior of low-plasticity fine-grained soils, and corrections for pore fluid chemistry and clay minerology may be needed for future liquefaction susceptibility and cyclic softening assessment procedures. Site-specific cyclic and monotonic testing for important projects remains to be the recommended approach for properly characterizing the seismic behavior of such soils with the current level of understating of their complex behavior.

The collected experimental data, visualization and post-processing tools, as well as documented experimental procedures of Part I are curated and published for public access on http:www.DesignSafe-ci.org. The data from Part II is being curated and will be published online in the near future.

Soil-structure-interaction (SSI) plays an important role in seismically-induced ground failure; however, the contribution is poorly understood and rarely considered in geotechnical engineering practice. Semi-empirical methodologies used in practice to evaluate ground failure solely consider demand from vertically propagating shear waves in the free field. Mounting evidence suggests this approach may significantly underpredict induced demand in soils below shallow foundations by neglecting SSI-induced stress contributions. Observations made during post-earthquake reconnaissance suggest that soils below low- to mid-rise structures supported on shallow foundations may be more susceptible to ground failure than soils in the corresponding free field. Recent physical and numerical modeling studies substantiate this evidence by demonstrating that SSI-induced stresses contribute to ground failure potential and may exacerbate the consequences of ground failure. Studies to date have centered on liquefaction-based (e.g. sand-like) ground failure; complementary studies on cyclic softening (e.g. clay-like) ground failure are not available.

This research seeks to fill that knowledge gap by evaluating the SSI-induced seismic loading increment imparted by shallow foundations on fine-grained soils within the context of cyclic softening ground failure. The research is based on an analysis framework that utilizes elastic solutions to define SSI-induced stress demands beneath shallow foundations. The framework proposes a new demand parameter called the deviatoric strength ratio to quantify incremental stress demands under complex stress paths below footings. To substantiate the framework and provide empirical case history data on cyclic softening, the author performed a geotechnical centrifuge testing program at the Center for Geotechnical Modeling at the University of California Davis. The program consisted of two dynamic centrifuge tests incorporating structures founded on strip footings and bearing on fine grained soil. The structures incorporated a range of mass and stiffness properties at approximately the same bearing pressure to evaluate the influence of structure response on cyclic softening. Analysis results indicate a strong relationship between permanent foundation settlement or rotation and indices that represent spatially averaged cyclic strains in foundation soils when subjected to deviatoric stress demands from ground response and SSI.

Non-linear (NL) ground motion amplification functions have been developed for fine grained and highly organic soils using ground response analysis for two profiles in the Kushiro area of Hokkaido, Japan. The NL ground motion amplification functions will support a broader research effort to develop fragility functions for levees resting on organic soils in Hokkaido. The ground motion amplification functions were developed using one-dimensional (1D) NL ground response simulations in DEEPSOIL (Hashash et al. 2016) based on a suite of earthquake ground motions compiled by Baker (Baker et al. 2011). Corrections to the small-strain Darendeli (2001) modulus reduction curve for fine grained soil recommended by Yee et al. (2013) were used to realistically represent the large-strain portion of backbone curves by asymptotically approaching the shear strength at large strains. The GQ/H material model (Groholski, 2016) in DEEPSOIL was utilized to preserve the input undrained shear strength while still providing the ability to represent small-strain stiffness nonlinearity. The resulting ground motion amplification values, defined as the spectral acceleration of the surface motion divided by the spectral acceleration of the input outcrop motion, was regressed using the functional form for NL site ground motion amplification by Stewart et al. (2014).

Ground motion amplification functions for organic soils in the Sacramento / San Joaquin Delta were previously developed by Kishida, et al. (2006 and 2009a) using 1D equivalent-linear (EL) ground response analysis (GRA) that were subsequently regressed to derive ground motion amplification functions. These simulations utilized linear and nonlinear regression models for dynamic properties of highly organic soils through ongoing research on this topic that was ultimately concluded by Kishida et al. (2009b). However, it is not clear whether these ground motion amplification functions are applicable to the Hokkaido system due to differences in the Peat layers in Hokkaido compared with the Delta, nor is it clear the extent to which the equivalent linear assumption influenced the ground motion amplification functions. This thesis also presents a brief comparison between the NL ground motion amplification functions developed herein and the EL ground motion amplification functions for highly organic soils presented by Kishida et al. (2009a).