Yousef Bozorgnia

This study evaluates the seismic performance implications of recently proposed near‐fault directivity factors for the California Department of Transportation (Caltrans) Seismic Design Criteria (SDC). Recent advances in probabilistic seismic hazard analysis have enabled site‐specific incorporation of near‐fault directivity into design spectra, including updates to the Caltrans SDC, yet their structural performance implications remain largely unexplored. To address this gap, nonlinear response history analyses of two typical long‐span ordinary bridges with single‐ and two‐column bents were conducted using detailed three‐dimensional bridge models subjected to bidirectional near‐fault ground motions. Ground motions were scaled to three hazard‐consistent target spectra: (1) a uniform hazard spectrum without directivity, (2) a site‐specific spectrum incorporating directivity models within the hazard integral, and (3) a spectrum based on recently proposed Caltrans directivity amplification factors. Motions were rotated across multiple incidence angles to evaluate direction‐dependent structural response, and analyses were conducted at two seismic hazard levels (1000‐ and 2475‐year return periods). Results indicate that directivity‐amplified spectra produce higher drift demands, higher collapse rates, and greater response variability compared to uniform hazard spectra without directivity, with differences becoming more pronounced at longer return periods. Comparisons between the new SDC factors and model‐based directivity spectra show similar overall trends but case‐dependent differences in deformation demands, reflecting the interaction between directivity representation and structural characteristics. Directional effects significantly influence structural response, with demand variations of up to 35% relative to median responses computed across multiple incidence angles. This study provides one of the first structural evaluations of recently proposed directivity‐based design spectra and underscores the need to explicitly incorporate directivity and ground‐motion incidence angle effects in performance‐based design and assessment of bridges near active faults.

Section 12.4.2 of the ASCE/SEI 7‐22 provisions offers two alternative methods for estimating the seismic load effects due to vertical ground motions: one based on horizontal spectral acceleration at a short horizontal period and the other uses the vertical spectral acceleration at a vertical structural period. This study examines similarities and discrepancies between the two methods for the contiguous U.S., covering Seismic Design Categories (SDCs) A through E, and vertical structural periods ranging from 0.01 to 10 s. Overall, 331,231 sites in the U.S. are analyzed in this study. A lognormal probabilistic model is developed to maintain approximate mathematical tractability of the distribution of the ratio of vertical seismic load effects computed by the two methods across the considered vertical period range for each SDC. Results indicate that, for sites located in the Western U.S. (WUS), and for vertical structural periods between 0.05 and 0.2 s, the approach based on vertical spectral acceleration generally results in higher vertical seismic load effects than the one based on short‐period horizontal spectral ordinates. For sites located in the Central and Eastern U.S. (CEUS), the short‐period horizontal spectral acceleration approach consistently yields comparable or higher estimates of the vertical seismic load effects than the one based on the vertical spectral accelerations.

This paper provides an overview of the Fault Displacement Hazard Initiative (FDHI). The FDHI is a community-based and multi-year research initiative with the goals of developing: (a) a comprehensive empirical database of surface fault displacement measurements and mapped ruptures, and (b) a suite of next-generation fault displacement predictive models. The database was developed by systematically collecting worldwide data from professional literature, followed by an extensive review and quality assurance process. The FDHI database contains 75 surface-rupturing earthquakes ranging from moment magnitude M 4.9 to 8.0 including data from strike-slip, reverse/reverse-oblique, and normal/normal-oblique faulting events. As part of the FDHI Program, employing the database, four new fault displacement models (FDMs) were developed. The FDMs predict principal or aggregate surface displacement, where aggregate is the combined displacement across principal and distributed ruptures. Two models were developed for all styles of faulting, while the other two are for reverse or strike-slip faulting. The new models include improved magnitude scaling and aleatory variability modeling. Compared with pre-existing models, average displacements in the new models are about 40% larger for M ~7, but smaller for lower and higher magnitudes. In addition, the new models predict significantly smaller displacement at low probability of exceedance. Besides the core FDHI tasks, in this article we also provide an overview of other parallel projects completed in coordination with the FDHI Program.

This article presents Bayesian interfrequency and interintensity empirical correlation coefficients of the horizontal Fourier effective amplitude spectrum (EAS) for frequencies ranging from 0.1 to 20 Hz. Interintensity correlations were calculated between the residuals of the EAS ordinates and the residuals of several peak and cumulative intensity measures, including 5%-damped pseudo-spectral acceleration (PSA), peak ground acceleration (PGA), peak ground velocity (PGV), cumulative absolute velocity (CAV), and Arias intensity (AI). The database used in the study consists of shallow crustal earthquakes in active tectonic regions. The correlation coefficients were calculated using bivariate Bayesian inference of the total residuals of ground-motion models developed by the authors. The component correlations were not found to be dependent on the magnitude of the event. Although the total correlations are magnitude-dependent because of the magnitude-dependent variance components used to calculate them, the differences are so small that the total large-magnitude correlations can be used for all magnitudes. The correlations are dependent on the component residual, the number of frequencies, and the type of ground-motion component. The interfrequency between-event correlations have the largest uncertainty and the weakest decrease with frequency away from the reference frequency than the other component correlations. The interfrequency correlations of EAS were found to be more peaked and to decrease faster away from the reference frequency than those of PSA. Nonetheless, EAS and PSA at the same frequency are strongly correlated with one another over a broad range of frequencies. The interintensity correlations between EAS and the other intensity measures are peaked in the mid-frequency range and do not show the same broadband correlation with EAS that PSA does.

The data sets, model formulations, and results from four new fault displacement models (FDMs) developed through the Fault Displacement Hazard Initiative (FDHI) Project are summarized and compared with each other and previously published FDMs. The models were developed using the new FDHI Database and provide predictions for principal or aggregate surface fault displacement, where aggregate is the combined displacement across principal and distributed ruptures. Different definitions of displacement are used among the models, and the differences should be considered when comparing model predictions or using multiple models in a logic tree. All new models are applicable between M 6.0 and 8.0, although some are also applicable to lower or higher magnitudes. Two models were developed for all styles of faulting, while the other two only apply to a single style. Quantitative comparisons are provided for a range of scenarios defined by style of faulting, magnitude, and normalized along-strike location. Average displacement predictions in the new models are within a factor of about 1.5 for most magnitudes and styles of faulting. Upper tail (i.e. 99th percentile) predictions in the new models are within a factor of about 2.5 in most cases. Compared to previously published models, average displacements in the new models are about 40% higher for M ∼7, whereas upper tail predictions are up to six times lower. Key features of the new models include the use of a large, high-quality empirical database and improved modeling of the magnitude scaling and aleatory variability. Together, these lead to upper tail predictions that are in reasonable agreement with empirical observations of maximum displacement for large magnitudes, which supports the use of the FDHI FDMs in probabilistic fault displacement hazard analysis (PFDHA) at long return periods.

It has become common in the development of ground‐motion models (GMMs), especially using mixed‐effects regression with crossed random effects, to calculate standard deviations, referred to as variance components, from sample statistics of the residuals (i.e. random effects and within‐group errors) rather than using the variance components reported by a mixed‐effects regression program, calculated using the same algorithms used in the mixed‐effects regression, or estimated using Bayesian inference. This practice leads to underestimating the standard deviations because it does not account for the uncertainty (i.e. standard errors) associated with fitting the random effects and within‐group errors during the regression analysis. In this study, we used a series of ground‐motion models for Fourier amplitude spectra developed using mixed‐effects regression of an Next Generation Attenuation (NGA)‐West2 database to show that residual‐based standard deviations can be significantly smaller than regression‐based standard deviations. These differences are exacerbated for those variance components with the fewest number of observations or when the residuals are partitioned (e.g. by earthquake magnitude). Using residual‐based variance components not only results in smaller standard deviations but can lead to biased inferences when attempting to compare the efficacy of a model based solely on a comparison of the values of the variance components or their related mean square errors. It can also lead to biased fixed‐effects coefficients if the variance components are derived from the residuals of a mixed‐effects regression rather than estimated during a regression. We show that variance components can be decomposed into various random‐effect grouping factors and partitioned into subsets of predictor variables, such as magnitude, from the total residuals of a non‐Bayesian mixed‐effects regression program using Bayesian inference. This process does not require the development of an entire GMM using Bayesian mixed‐effects regression.

On 6 February 2023, a devastating earthquake of magnitude 7.7 struck the southeastern part of Turkey, followed by multiple aftershocks. The sequence of earthquakes caused extensive damage in Southern Turkey and Northern Syria. This study evaluates the destructive potential of ground shaking resulting from the Pazarcik mainshock by developing spatial distribution of inelastic demand spectra across the region by using the recorded ground motions. These spectra provide an additional step toward more realistic estimates of the damage potential of ground shaking than the traditional elastic response spectra. Given the recorded ground motions, we also developed simulated ground‐motion time series at numerous un‐instrumented sites. We used these recorded and simulated motions to estimate ductility demands across the affected area. Constant‐ductility spectra were derived using recording stations within 100 km of the fault rupture. The results indicated that in the near‐fault area, and for structural periods of 0.5 and 1.0 s, the yield strength demand at a ductility level of 3 exceeded the levels specified in the local seismic design code for a 475‐year return period. Our findings demonstrate that with more refinement and efficiency of the ground‐motion simulation approach, the development of near‐real‐time spatial distribution of ductility demand is a promising approach for rapid seismic damage assessment of earthquakes.

This article presents the development of ground‐motion models (GMMs) of Fourier amplitude spectra for frequencies of 0.1–20 Hz and its potential extrapolation to 100 Hz at near‐source distances using the effective amplitude spectrum (EAS) ordinates developed for the NGA‐West2 project and the metadata and functional form from our previous NGA‐West2 GMMs. We developed the GMMs using three different approaches to study the impact of including random effects in the model development: (1) fixed‐effects regression (i.e. no random effects), (2) mixed‐effects regression with events as a random effect, and (3) mixed‐effects regression with both events and sites as random effects. Goodness‐of‐fit metrics show that the GMMs were improved with the addition of each random effect. We found the variance components other than the between‐site standard deviation to be magnitude‐dependent, which we estimated using Bayesian inference to incorporate uncertainty in the random‐effects and within‐group variability. As a result, our aleatory standard deviations are larger than residual‐based standard deviations that ignore uncertainty in these terms. The GMMs predict a slight spectral sag at intermediate frequencies and large magnitudes that becomes more pronounced with increasing magnitude, consistent with other empirical analyses and ground‐motion simulations available in the literature. We recommend use of the GMM that includes both event and site terms because of its appropriate modeling of repeatable effects. We present the other GMMs to demonstrate how each model is improved as random effects are added and to facilitate their comparison with other models that use a similar random‐effects structure.

Fault displacement models are a key component of probabilistic fault displacement hazard analysis (PFDHA), providing the displacement probability density function. We present a new fault displacement model developed using the comprehensive database compiled for the Fault Displacement Hazard Initiative (FDHI) project. The model predicts the total net displacement across multi-stranded surface ruptures as a function of moment magnitude and position along the rupture length. We provide separate models for strike-slip, reverse, and normal faulting. A bilinear magnitude scaling is used to capture the steeper magnitude scaling for smaller magnitudes observed in the data for strike-slip and normal faulting. The scaling for reverse faulting approaches log-linear but still uses a bilinear form. Our flexible location scaling functional form captures the asymmetric profile shapes common in the data for dip-slip events and more symmetrical elliptical profiles in the strike-slip data. We applied a Box-Cox transformation to the displacement dataset so that the data resemble a normal distribution, which allows the transformed displacement to be conveniently modeled as normally distributed. Between- and within-event aleatory variability are modeled separately. All model parameters are estimated using Bayesian regression, which allows within-model epistemic uncertainty to be quantified. Compared with earlier models, the new model produces less aleatory variability and smaller upper-tail displacements for magnitudes greater than about 6.5 for all styles of faulting and all locations along the rupture. Data dispersion at the rupture end-points in the strike-slip and reverse faulting resulted in large aleatory variability. Large aleatory variability for smaller magnitude strike-slip and normal events is driven by a lack of data at smaller magnitudes. The use of a large, high-quality database and advanced statistical modeling techniques in our new model provides improved predictions of fault displacement compared with earlier models.