CSChih-Hsuan Sung

Chih-Hsuan Sung

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Chih-Hsuan Sung has a total of 103 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Nonergodic Path Effects from 3D Ground-Motion Simulations of an M 8 Ryukyu Earthquake in Eastern Taiwan

    by Ming-Che Hsieh, Chih-Hsuan Sung, Yung-Ching Yang

    The Ryukyu subduction zone offshore eastern Taiwan possesses significant seismogenic potential, exemplified by the 1920 earthquake. However, the scarcity of near-field observations leaves the ground-motion characteristics of such mega-earthquakes poorly constrained, limiting the reliability of seismic hazard assessments. This study investigates long-period ground-motion characteristics of a potential M 8 Ryukyu earthquake using large-scale 3D ground-motion simulations. We simulated an M 8 scenario earthquake using characterized source models based on the “Recipe” procedure and conducted 3D finite-difference simulations for 192 source–velocity-model combinations, incorporating variations in hypocenter location, rupture velocity, asperity distribution, two tomographic velocity models, and topography. Varying the Gaussian source time function (STF) duration in postprocessing expands the dataset to 1728 scenarios. Synthetic waveforms were generated at 4950 virtual stations, producing more than 8.5 million synthetic ground-motion records, and analyzed using RotD50 spectral accelerations (SAs) at periods of 1.0, 3.0, and 5.0 s. Results indicate that rupture directivity and rupture velocity dominate the spatial variability of long-period ground motions, with parameter sensitivity exhibiting strong spatial nonuniformity across the study region. Calibration against the empirical CH20 ground-motion model (GMM) suggests Gaussian STFs of 2, 6, and 8 s provide the most consistent spectral representation for SA at 1.0, 3.0, and 5.0 s, respectively. The nonergodic path terms (δP2P) derived from two alternative velocity models, which capture the path effects of the deep 3D structure (VS≥2.4 km/s), reveal spatially consistent features, including a persistent amplification zone near Hualien linked to the Longitudinal Valley low-velocity structure, a deamplification belt along the Coastal Range and Central Range high-velocity material, and a frequency-dependent amplification in the Ilan Plain attributable to deeper crustal focusing or trapping rather than near-surface site effects. The workflow established here provides a transferable framework for refining GMMs and advancing nonergodic seismic hazard assessments in data-sparse subduction environments.

  • Open Access

    Simulation‐Based Nonlinear Site Amplification Models for Western United States

    by Okan Ilhan, Dylan Centella, Ahmad Sulaiman, Chih‐Hsuan Sung, Youssef M. A. Hashash, Norman A. Abrahamson

    This article presents one of the first nonlinear (NL) site amplification models relative to V S30 of 800 m/s for the Western United States (WUS) for both the smoothed effective amplitude spectrum (EAS), described as the orientation‐independent horizontal‐component Fourier amplitude spectrum (FAS) of ground acceleration, and the response spectra ( R S ). The amplification models were derived from a total of 286,254 1D site response analyses using randomizations of base‐case V S profiles, generated through a new slowness model developed using V S measurements at California, and 600 recorded ground‐motion pairs, scaled to match target spectra for 12 earthquake scenarios. The use of slowness model for generating V S randomizations yields a far broader suite of site conditions than from the preceding study employing only eight generic profiles. The simulation‐based EAS‐ and RS‐based NL amplification data exhibits more continuous behavior and alleviates the data sparsity in the low (r (peak ground acceleration at V S30 = 800 m/s) range as compared with prior work. Four NL mean amplification models with corresponding standard deviation relationships are proposed: two EAS‐based named as N1 EAS ( V S30 , EAS r at 5 Hz), and N2 EAS ( V S30 , frequency‐specific EAS r ), and two RS‐based models as N1 RS ( V S30 , PGA r ) and N2 RS ( V S30 , pseudo‐spectral acceleration at V S30 = 800 m/s (PSA r )). The N2 relationships exhibit lower standard deviation values of the residuals compared with their N1 counterparts. The compatibility between NL models’ predictions and observed NL amplification was assessed by incorporating NL relationships into an empirical ground motion model (GMM), resulting in within‐event residuals of GMM closer to zero relative to those of GMM without NL terms, particularly for sites with 200 m/s < V S30 < 500 m/s. The amplification models are applicable for sites with V S30 ≥ 200 m/s and motions with PGA r ≤ 1.0 g or EAS r   ≤ 0.3 g/s.

  • Open Access

    The Cascadia Region Earthquake Science Center (CRESCENT): Advancing Understanding of Cascadia’s Earthquake Hazards

    by Diego Melgar, Amanda M. Thomas, Valerie Sahakian, Pieter-Ewald Share, Andrew Meigs, Harold Tobin, Lydia Staisch, Timothy Melbourne, Jill Elizabeth, Rasheed Ajala, et al.

    The Cascadia subduction zone (CSZ) is capable of producing large megathrust earthquakes and tsunamis, posing substantial risks to the Pacific Northwest. This article examines the seismic characteristics of the CSZ, including its offshore coupling, sparse seismicity, and a complex paleoseismic record. Recent advances in geophysical methods and data collection have improved our understanding of subduction zone processes, but important uncertainties remain regarding rupture dynamics, ground shaking, and tsunami impacts. The Cascadia Region Earthquake Science Center (CRESCENT) has been established to address these challenges through coordinated research, education, and community engagement. By integrating scientific studies, fostering collaboration, and leveraging new technologies, CRESCENT seeks to improve hazard assessment and resilience in the region. This article outlines the primary challenges and opportunities for advancing subduction-zone science and supporting societal preparedness for future CSZ events.


  • Examining the Causes of Regional Differences in California Based on Ground Motions from Small to Moderate Earthquakes

    by Camilo I. Pinilla-Ramos, Norman A. Abrahamson, Savvas Marcou, Douglas S. Dreger, Alexandros Savvaidis, Gail M. Atkinson, Behzad Hassani, Chih-Hsuan Sung, Robert Youngs

    Residuals from ground-motion models (GMMs) for California show systematic differences in the long-period event terms for small-to-moderate magnitude earthquakes (Mw < 5) between Northern and Southern California. The spatial pattern of regional differences in event terms followed the boundaries of the seismic networks, suggesting a systematic difference in magnitudes for small-to-moderate earthquakes between the Northern and Southern California networks. Both networks employ the same workflow defined by the Advanced National Seismic System. Therefore, procedural differences in moment tensor inversion (MTI) are unlikely to be the cause of the observed regional differences in event terms. The two networks utilize different seismic-velocity models in the MTI, with lower velocities in coastal Northern California compared to Southern California. A physics-based analysis indicates that the regional amplification differences among the seismic-velocity models used in the MTI are consistent with the observed differences in the event terms between the two regions, providing evidence that the event terms in ergodic GMMs for California also capture regional wave-propagation differences, not just source differences. This affects the development of GMMs: a better methodology is needed to identify these regional wave propagation patterns from event-specific faulting processes, which are currently mapped in the event-terms. In addition, having several earthquakes in the datasets with magnitudes obtained without MTI is a potential source of regional differences in the event terms. Using earthquakes selected from the ongoing Next Generation Attenuation-West3 project, we conduct MTIs on 188 earthquakes, yielding 63 seismic moment estimates and an additional 43 values adopted directly from the Comprehensive Catalog. These 106 moment magnitude estimates can replace those from earthquakes in the dataset that lack a moment magnitude, reducing reliance on conversion models. We conclude this study by providing recommendations to improve the documentation of the MTI maintained by the seismic network operators.


  • New Approach to Nonergodic Path Effects in Ground-Motion Models

    by Maxime Lacour, Norman Abrahamson, Chih-Hsuan Sung

    A recent advance in modeling nonergodic path effects by Sung et al. (2023) includes a path term for the 3D velocity structure effects. For the spatial correlation of the path effects, the difference between two paths was parameterized by the vector sum of the distance between the two source locations and the distance between two site locations. This metric works well for very different paths with small correlations (0–0.1) and very similar paths with large correlations (0.9–1.0), but it overestimates the correlation for paths with intermediate correlations. We propose a new metric for measuring the difference between two ray paths based on the difference in site locations, azimuths, and rupture distances. This new metric leads to a correlation model that better captures the observed spatial correlation structure of the ground motions than the existing metric. The new metric has a better physical basis for parameterizing the difference in the path effects and improves the accuracy of the extrapolation of the model to scenarios outside of the range constrained by the data. In addition, we present an efficient numerical method (Scalable Kernel Interpolation for Product Kernels [SKIP]) that can be applied to large data sets (more than 100,000 ground motions), as is common in large simulation data sets. Using SKIP, the forward predictions of path effects for a grid of source locations for a given site can be computed with a calculation time of about a minute using traditional laptop computers with standard memory.


  • Ground-Motion Models Incorporating Nonergodic Effects from 3D Numerical Simulations in Japan

    by Chih-Hsuan Sung, Nobuyuki Morikawa, Asako Iwaki, Norman Abrahamson, Hiroe Miyake

    We analyzed scenario ground-motion simulations from the Miyagi, Kanto, Nobi, and Osaka regions to understand the spatial distribution of nonergodic site terms for ground motions of crustal events. We selected one of the Japan regional ground-motion models (GMMs) as the reference ergodic model and estimated adjustment terms based on the 3D simulation results to develop a nonergodic ground-motion model (NGMM). First, a Z2.1 basin depth term was constructed from the 3D simulations, denoted fbasin3d, and this term was applied to the Morikawa and Fujiwara (2013; hereafter, MF13v1) GMM. The basin terms from the simulations show stronger scaling with the basin depth compared to empirical ground-motion data from K-NET and KiK-net. Second, the spatial distribution of the nonergodic site term from the simulations, denoted δS2SNE, was estimated by applying the varying coefficient model within the region covered by the simulations. The net adjustment to the MF13v1 GMM is the sum of fbasin3d and δS2SNE. The nonergodic aleatory variability for the NGMM is reduced by 10%–20% compared to the ergodic GMM. The NGMM not only changes the average level of ground motion compared to the ergodic GMM but also modifies the spatial distribution of the nonergodic term and the aleatory variability. In all study regions, the nonergodic site terms derived from the 3D simulation are found to be within the epistemic uncertainty range of the site terms obtained from empirical data. This indicates that integrating simulation results into the GMM can potentially constrain the prediction in regions where empirical data are insufficient. In addition, lacking simulation results for a suite of alternative 3D velocity models to address epistemic uncertainty, we assume it is half the standard deviation of the nonergodic site terms in the region (0.5 ϕS2S).


  • Ground-Motion Modeling Using MyShake Smartphone Peak Acceleration Data

    by Savvas Marcou, Richard M. Allen, Norman A. Abrahamson, Chih-Hsuan Sung

    In the field of ground-motion modeling, the availability of densely sampled ground-motion data is becoming key to mapping repeatable source, path, and site effects to enable ground-motion models (GMMs) to more accurately predict shaking from future earthquakes. This is particularly important because the field is moving toward nonergodic GMMs with spatially variable coefficients. To achieve the level of sampling required, the addition of non-instrumental data collected at very high spatial resolution, like felt intensity data or smartphone data, could prove essential. The predictive power of this nontraditional data for free-field ground motion needs to be tested before these data are used. In this work, we present a new database of over 1600 ground-shaking waveforms collected between 2019 and 2023 by the MyShake smartphone app, which delivers earthquake early warning messages to users on the U.S. West Coast. We develop a GMM, MyShake GMM, for peak smartphone-recorded accelerations in 3≤M≤5.5 earthquakes recorded at short (<50 km) distances. We compare our model with free-field GMMs and show a similar geometric decay and a close match in predicted amplitudes for short-period spectral accelerations (SAs). We use residual correlation analysis to show that MyShake GMM residuals have a positive correlation with free-field residuals, with correlation coefficients of around 0.4 for peak ground acceleration, velocity, and short-period SA, similar to correlations previously reported between felt intensity and free-field data. This illustrates the potential that densely sampled smartphone ground-shaking data has in identifying repeatable free-field ground-motion effects for various ground-motion modeling applications. These could potentially include highly location-specific assessments of site response, ground-motion interpolation schemes like ShakeMap, or validating outputs from nonergodic, spatially variable coefficient GMMs.


  • New Empirical Source-Scaling Laws for Crustal Earthquakes Incorporating Fault Dip and Seismogenic-Thickness Effects

    by Jyun-Yan Huang, Norman A. Abrahamson, Chih-Hsuan Sung, Shu-Hsien Chao

    New global source-scaling relations for the aspect ratio and rupture area for crustal earthquakes that include the width-limited effect and a possible free-surface effect are derived using a global dataset of finite-fault rupture models. In contrast to the commonly used scaling relations between moment magnitude (M), fault length (L), width (W), and area, we built self-consistent scaling relations by relating M to the aspect ratio (L/W) and to the fault area to model the change in the aspect ratio once the rupture width reaches the down-dip width limit of the fault. The width-limited effect of large-magnitude earthquakes depends on the fault dip and a regional term for the seismogenic thickness. The magnitude scaling of the aspect ratio includes a break in the magnitude scaling that is dip angle dependent. This dip angle-dependent magnitude scaling in the magnitude–area relation is modeled by a trilinear relation incorporating a dip-related transition range. The effect of the free surface was observed using a normalized depth term and parameterizing the source by the depth of the top of the fault rupture; it is more apparent in the area scaling relation. The scaling differences are related to the fault geometry, not to the rake angle, as commonly assumed. Finally, the corresponding L and W scaling relations obtained by converting the area and aspect ratio models to L and W models not only show good agreement with the previous regional scaling laws on average but also provide better fault-specific application due to the inclusion of a fault-specific dip angle and seismogenic thickness.


  • Nonergodic Ground-Motion Models for Subduction Zone and Crustal Earthquakes in Japan

    by Chih-Hsuan Sung, Hiroe Miyake, Norman Abrahamson, Nobuyuki Morikawa

    We investigate the nonergodic behavior of ground motions from subduction zone earthquakes and crustal earthquakes in Japan using the National Research Institute for Earth Science and Disaster Resilience strong‐motion flatfile together with various reference ergodic ground‐motion models (GMMs). For the nonergodic path effects, our nonergodic GMM has two path effects terms: a cell‐specific linear‐distance scaling, denoted as δP2PQ⁠, that mimics the effects of a 3D Q structure, and a source‐ and site‐specific term related to the effects of the 3D velocity structure, denoted δP2PV⁠. The resulting model provides spatially varying nonergodic source, site, and path terms. The GMM smoothly interpolates and extrapolates the nonergodic terms in space so that the GMM can be applied to any combination of source and site locations in the region. In regions where data are sparse, the median nonergodic terms will approach zero but with large epistemic uncertainty. Over the period range of 0–10 s, the ranges of the standard deviations of the nonergodic source, path, and site terms are 0.2–0.65, 0.25–0.40, and 0.15–0.40 natural log units, respectively. The fully nonergodic models lead to a 40%–55% decrease in the aleatory standard deviation compared to the reference ergodic GMMs. This large reduction in the aleatory standard deviation combined with the change in the median given by the nonergodic terms can exert a significant impact on the computed seismic hazard for the Japan region.


  • Methodology for Including Path Effects Due to 3D Velocity Structure in Nonergodic Ground-Motion Models

    by Chih-Hsuan Sung, Norman Abrahamson, Maxime Lacour

    A new approach is developed to incorporate the anisotropic path effects in ground motions due to the 3D velocity structure into nonergodic ground‐motion models (GMMs) using the varying coefficient model (VCM) in a two‐step process. The first step uses the VCM to estimate the spatially varying path term for each site separately with the spatial correlation based on the separation between earthquakes. The model for the correlation length varies as a function of rupture distance to reflect that, for a given separation between events, the path effects are more similar for longer path lengths. The second step uses the step‐1 VCM results as the input to estimate path terms for a specific source location for any site location. The final model is a smooth spatially varying nonergodic path term that can be applied to any source–site pair. An example application of the proposed method using a subset of 600 scenarios of the v15.4 CyberShake simulation for T = 3 s response spectral values shows that nonergodic path terms can be significant, ranging from −0.8 to 0.8 ln units, which corresponds to factors of 0.45–2.23 in the median ground motion compared to the ergodic approach. With nonergodic path terms, the single‐path aleatory variability for the nonergodic GMM is reduced to 0.34 ln units as compared to 0.52 ln units for single‐station sigma without systematic path effects. A comparison of probabilistic hazards shows that the proposed approach leads to a nonergodic GMM that can capture the path effects seen in the CyberShake simulations in terms of both the median and the aleatory standard deviation. The resulting nonergodic GMM extends the applicability of the 3D simulation results to site–source pairs in the region and not limited to the locations of the simulations, making the simulation results more practical to use in seismic hazard studies.