The sufficiency criterion has long been used to evaluate the effectiveness of a ground motion intensity measure (IM) in capturing the link between ground shaking and structural response. However, a typical sufficiency-based evaluation of an IM only tests for the possibility of linear dependency and the interaction among the upstream parameters is not considered. To address these and other limitations, two new IM evaluation methodologies are proposed. The first methodology considers the loss of statistical information when an IM is used to predict the engineering demand parameters (EDPs) without including the upstream parameters (i.e., earthquake magnitude, source-to-site distance and epsilon). The best IM is the one that minimizes the loss of predictive performance when it is the only model input relative to when it is used as a predictor together with the upstream parameters. To consider the possible interactive effects, a machine learning model is used when both the IM and upstream parameters are used as inputs. The second methodology uses a causal inference approach where the effect of the IM on the EDP distribution is quantified while considering the earthquake magnitude, source-to-site distance and epsilon as control variables. The double machine learning approach is implemented for this purpose. The two methodologies are applied to a set of five steel specifical moment resisting frames. The results show that the statistical loss-based and causal inferencing approaches produce results that are more conclusive than the sufficiency-based approach and more consistent with the physical laws that govern the IM-EDP relationship.

Databases of recorded motion are limited despite the increasing amount of data collected through strong motion instrumentation programs. Particular lack of data exists for large magnitude events and at close distances as well as on earthquakes in deep sedimentary basins. Additionally, databases of recorded motions are also limited in representation of energy at long periods due to the useable frequencies of recording instruments. This lack of data is currently partially addressed through assumption of ergodicity in development of empirical ground motion prediction equations (GMPEs). Nevertheless, challenges remain for calibration of empirical GMPEs as used in conventional approaches for probabilistic estimation of seismic hazard. At the same time, limited data on strong earthquakes and their effect on structures poses challenges for making reliable risk assessments particularly for tall buildings. For instance, while the collapse safety of tall buildings is likely controlled by large magnitude earthquakes with long du- rations and high long-period content, there are few available recorded ground motions to evaluate these issues. The influence of geologic basins on amplifying ground motion effects raises additional questions. Absent recorded motions from past large magnitude earthquakes, physics-based ground motion simulations provide a viable alternative due to the ability to consider extreme ground motions while being inherently site-specific and explicitly considering instances not well constrained by limited empirical data. This thesis focuses on utilization of physics-based simulated earthquake ground motions for performance assessment of tall buildings with three main goals: (1) developing confidence in the use of simulated ground motions through comparative assessments of recorded and simulated motions; (2) identifying important characteristics of extreme ground motions for col- lapse safety of tall buildings; (3) exploring areas where simulated ground motions provide significant advantages over recorded motions for performance-based engineering. To gain confidence in the use of simulated motions for full performance assessment of tall buildings, a 'similar intensity measure' validation study was performed. Structural responses to ground motions simulated with different methods on the Southern California Earthquake Center (SCEC) Broadband Platform (BBP) are contrasted to recorded motions from PEER NGA database with similar spectral shape and significant durations. Two tall buildings, a 20-story concrete frame and a 42-story concrete core wall building, are analyzed at increasing levels of ground motion intensity, up to structural collapse, to check for statistically significant differences between the responses to simulated and recorded motions. Considered demands include story drift ratios, floor accelerations and collapse response. These comparisons yield similar results in most cases but also reveal instances where certain simulated ground motions can result in biased responses. The source of bias is traced to differences in correlations of spectral values in some of the stochastic ground motion simulations. When the differences in correlations are removed, simulated and recorded motions yield comparable results. Moving beyond validation, the thesis also explored areas where the use of simulated motions provides advantages over approaches based on limited databases of recorded motions for performance-based engineering. One such area is seismic risk in deep sedimentary basins which is studied by examining collapse risk and drift demands of a 20-story archetype tall building utilizing ground motions at four sites in the Los Angeles basin. Seismic demands of the building are calculated form nonlinear structural analyses using large datasets (~500,000 ground motions per site) of unscaled, site-specific simulated seismograms. Seismic hazard and building performance from direct analysis of SCEC CyberShake motions are contrasted with values obtained based on 'conventional' approaches that rely on recorded motions coupled with probabilistic seismic hazard assessments. The analysis shows that, depending on the location of the site within the basin, the two approaches can yield drastically different results. For instance, at a deep basin site the CyberShake-based analysis yields around seven times larger mean annual frequency of collapse ( c) and significantly higher drift demands (e.g. drift demand of 1% is exceeded around three times more frequently) compared to the conventional approach. Both the hazard as well as the spectral shapes of the motions are shown to drive the differences in responses. Deaggregation of collapse risk is performed to identify the relative contributions of earthquake fault ruptures, linking building responses with specific seismograms and contrasting collapse risk with hazard. The effect of earthquake ground motions in deep sedimentary basins on structural collapse risk is further studied through the use of CyberShake earthquake simulations in the Los Angeles basin. Distinctive waveform characteristics of deep basin seismograms are used to classify the ground motions into several archetype groups, and the damaging influence of the basin effects are evaluated by comparing nonlinear structural responses under comparable basin and non-basin ground motions. The deep basin ground motions are observed to have larger durations and spectral intensities than non-basin motions for vibration periods longer than about 1.5 seconds, which can increase the relative structural collapse risk by up to 20 percent between ground motions with otherwise comparable spectral accelerations and significant durations. Two new metrics, termed sustained amplitude response spectra (RSx spectra) and significant duration spectra (Da spectra), are proposed to quantify period-dependent duration effects that are not otherwise captured by conventional ground motion intensity measures. The proposed sustained amplitude response spectra and significant duration spectra show promise for characterizing the damaging effects of long duration features of basin ground motions on buildings and other structures. The large database of CyberShake simulations is utilized to re-examine the relationships between engineering demand parameters and input ground motions on structural response. Focusing on collapse response, machine learning techniques are applied to results of about two million nonlinear time history analyses of an archetype 20-story tall building performed using CyberShake ground motions. The resulting feature selection (based on regularized logistic regression) generally confirms existing understanding of collapse predictors as gained from scaled recorded motions but also reveals the benefit of some novel intensity measures (IMs), in particular the RSx spectral features. In addition, the statistical interrogations of the large collection of hazard-consistent simulations demonstrate the utility of different IMs for collapse predictions in a way that is not possible with recorded motions. A small subset of robust IMs is identified and used in development of an efficient collapse classification algorithm, which is tested on benchmark results from other CyberShake sites. The classification algorithm yields promising results for application to regional risk assessment of building performance.

This dissertation focuses on developing models for ground motion intensity, to formulate improved design spectra for use in assessing the performance of buildings under earthquakes. Most seismic building codes and design guidelines are based on implicit performance goals that structures should achieve. Despite the significant uncertainty in future ground motion occurrence, building codes commonly check a structure's behavior under a single level of earthquake loading, quantified with a design spectrum. However, this explicit design check is often not defined with respect to the performance goals. The objective of this dissertation is to provide the link between the explicit design check and the implicit performance goals. Models for multivariate distributions of ground motion properties are refined (specifically, spectral accelerations at multiple periods and locations) and tractable methods to utilize those models to assess seismic reliability of systems are developed. Using structural reliability approaches, with environmental contours of spectral accelerations at multiple periods, a justification of the use of multiple conditional mean spectra for design checks is achieved. Performance assessment procedures for the response spectrum method as well as nonlinear response history analysis are proposed based on these conditional mean spectra. Finally, this dissertation provides an original spatial cross-correlation model for spectral accelerations at multiple periods, which allows one to conduct the design checks simultaneously for multiple structures in a region.

Earthquake ground motion records are used as inputs for seismic hazard analysis, development of ground motion prediction equations and nonlinear response history analysis of structures. Real records from past earthquake events have traditionally been recognized as the best representation of seismic input to these analysis. However, our current way of implementing recorded ground motions is poorly constrained and suffers from the paucity of certain condition ground motions, such as the one with short distance and large magnitude. Meanwhile, even though the scaled ground motion is capable of matching the target spectrum, the content of frequency domain and ground motion parameters become unrealistic. With the rapid growth of computational ability and efficiency of computers, simulated ground motion can be an alternative to provide detailed and accurate prediction of earthquake effect. At the same time, simulated ground motions can provide a better representation of the whole ground motion generation process, such as fault rupture, wave propagation phenomena, and site response characterization. Hence, the aforementioned disadvantage of recorded ground motion can be overcame.Despite ground motion simulations have existed for decades, and the design code, such as ASCE/SEI 7-10 (ASCE, 2010), allow use of simulated ground motions for engineering practice, engineers still worried about the stability in ground motion simulation process and similarity between response of engineered structures to similar simulated and recorded ground motions. In order to draw simulated ground motions into engineering applications and make them practical, this dissertation is making contribution to address this issue. Simulated ground motions have to be validated and compared with recorded ground motions to prove their equivalence in engineering applications.This dissertation proposes a simulation validation framework. First step: Identify ground motion waveform parameters that well correlate with response of Multi-Degree of Freedom (MDOF) buildings and bridges. Second step: Develop goodness-of-fit measures and error functions that can describe the difference between simulated and recorded ground motion waveform characteristics and their effect on MDOF systems. Third step: Device the required update to ground motion simulation methods through which better simulations are possible. Forth step: Assess the current state of simulated ground motions for engineering applications.In general, simulated ground motions are found to be an effective surrogate and replenishment of natural records in engineering applications. However, certain drawbacks are detected, 1) Simulated ground motions are likelihood to mismatch certain ground motion parameters, for example, Arias intensity, duration and so on; 2) Structural behavior resulting from recorded ground motions and simulated ground motions are different. The difference stems from the fact that simulated motions are mostly pulse like motions. Because the simulation methods are still developing, our intent is not ranking or classifying them, but rather to provide feedback to update ground motion simulation techniques such that future simulations are more representative of recorded motions.

The effect of the earthquake ground motion parameters on the probabilistic loss estimation of buildings is the major interest of this study. For the seismic performance assessment, real ground motion records from the past earthquakes are required. Estimation of repair costs in future earthquakes is the major component for seismic loss analysis. This study addresses the sensitivity of the statistical characteristics of ground motions contributing to the building loss. Among these characteristics are the ground-shaking intensity (Arias Intensity), duration, and frequency at the middle of strong-shaking phase of the ground motion. These parameters are vital in determining the seismic response of the building structure. A fine study on the sensitivity of the seismic response and corresponding loss of the building structure to ground motions model parameters is carried out using Performance-based Earth- quake Engineering and Performance Assessment Computational Tool, respectively. But due to the scarcity of moderate to large earthquakes, the real records fail to match the required characteristics of motions, as there are insufficient set of data available for analysis to be carried out. Even, the of technique scaling ground motions results in overall unrealistic properties. This has led to the simulation of ground motions which will provide the additional and hopefully accurate predicted information on characteristics of the moderate to large earthquakes. Hence, a fully non-stationary stochastic model for strong earthquake ground motion model is considered which employs the statistical characteristics (waveform parameters) as model parameters matched with those of identified for a large sample of recorded ground motions for specified earthquake and site characteristics, to deliver simulated ground motions to examine the building loss metrics, which depends on the uncertainties in various analysis process starting from obtaining Intensity Measure (IM), Demand parameters (EDPs) to the repair cost estimates. From the predictive equations, specified earthquake and site characteristics results in the model parameters.Further, the validity of simulated ground motion time series representing the real ground shaking during future earthquakes is a crucial step. This study employs the hybrid broad- band ground motion simulation applied simulations to validate against the real records. With the help of hybrid approach, making use of wave propagation phenomena and site response characterization, effort has been taken for validation of these simulated ground motions is conducted for the sensitivity of seismic response and loss for these simulated ground motions.

While site effects are accounted for in most modern U.S. seismic design codes for building structures, there exist no standardized procedures for the computationally efficient integration of nonlinear ground response analyses in broadband ground motion simulations. In turn, the lack of a unified methodology affects the prediction accuracy of site-specific ground motion intensity measures, the evaluation of site amplification factors when broadband simulations are used for the development of hybrid attenuation relations and the estimation of inelastic structural performance when strong motion records are used as input in aseismic structural design procedures. In this study, a set of criteria is established, which quantifies how strong nonlinear effects are anticipated to manifest at a site by investigating the empirical relation between nonlinear soil response, soil properties, and ground motion characteristics. More specifically, the modeling variability and parametric uncertainty of nonlinear soil response predictions are studied, along with the uncertainty propagation of site response analyses to the estimation of inelastic structural performance. Due to the scarcity of design level ground motion recording, the geotechnical information at 24 downhole arrays is used and the profiles are subjected to broadband ground motion synthetics.