Assessment of ground motion variability and its effects on seismic hazard analysis: a case study for iceland

Probabilistic seismic hazard analysis (PSHA) generally relies on the basic assumption that ground motion prediction equations (GMPEs) developed for other similar tectonic regions can be adopted in the considered area. This implies that observed ground motion and its variability at considered sites could be modelled by the selected GMPEs. Until now ground-motion variability has been taken into account in PSHA by integrating over the standard deviation reported in GMPEs, which significantly affects estimated ground motions, especially at very low probabilities of exceedance. To provide insight on this issue, ground-motion variability in the South Iceland Seismic Zone (SISZ), where many ground-motion records are available, is assessed. Three statistical methods are applied to separate the aleatory variability into source (inter-event), site (inter-site) and residual (intra-event and intra-site) components. Furthermore, the current PSHA procedure that makes the ergodic assumption of equality between spatially and temporal variability is examined. In contrast to the ergodic assumption, several recent studies show that the observed ground-motion variability at an individual location is lower than that implied by the standard deviation of a GMPE. This could imply a mishandling of aleatory uncertainty in PSHA by ignoring spatial variability and by mixing aleatory and epistemic uncertainties in the computation of sigma. Station correction coefficients are introduced in order to capture site effects at different stations. The introduction of the non-ergodic assumption in PSHA leads to larger epistemic uncertainty, although this is not the same as traditional epistemic uncertainty modelled using different GMPEs. The epistemic uncertainty due to the site correction coefficients (i.e. mean residuals) could be better constrained for future events if more information regarding the characteristics of these seismic sources and path dependence could be obtained.     

Application-driven ground motion prediction equation for seismic hazard assessments in non-cratonic moderate-seismicity areas

We present a ground motion prediction equation (GMPE) for probabilistic seismic hazard assessments (PSHA) in low-to-moderate seismicity areas, such as Germany. Starting from the NGA-West2 flat-file (Ancheta et al. in Earthquake Spectra 30:989–1005, 2014), we develop a model tailored to the hazard application in terms of data selection and implemented functional form. In light of such hazard application, the GMPE is derived for hypocentral distance (along with the Joyner-Boore one), selecting recordings at sites with vs30 ≥ 360 m/s, distances within 300 km, and magnitudes in the range 3 to 8 (being 7.4 the maximum magnitude for the PSHA in the target area). Moreover, the complexity of the considered functional form is reflecting the availability of information in the target area. The median predictions are compared with those from the NGA-West2 models and with one recent European model, using the Sammon’s map constructed for different scenarios. Despite the simplification in the functional form, the assessed epistemic uncertainty in the GMPE median is of the order of those affecting the NGA-West2 models for the magnitude range of interest of the hazard application. On the other hand, the simplification of the functional form led to an increment of the apparent aleatory variability. In conclusion, the GMPE developed in this study is tailored to the needs for applications in low-to-moderate seismic areas and for short return periods (e.g., 475 years); its application in studies where the hazard is involving magnitudes above 7.4 and for long return periods is not advised.     

Ground Motion Prediction Equations in the San Jacinto Fault Zone: Significant Effects of Rupture Directivity and Fault Zone Amplification

We present a new set of Ground Motion Prediction Equations (GMPEs) for horizontal Peak Ground Acceleration, Peak Ground Velocity, and 5 % damped pseudo-spectral acceleration (PSA), developed for the San Jacinto Fault Zone (SJFZ) area. Besides using these equations to quantify seismic shaking in the area, the results allow us to examine the physics and local properties controlling the observed ground motions. The analyzed dataset includes ~30,000 observations from ~800 events spanning a magnitude range of 1.5 < M < 6.0 and recorded by up to 140 stations at epicentral distances ranging from essentially zero to 150 km. The local GMPE is developed for the SJFZ by applying classical regression techniques with predictive variables that include first distance and magnitude, and then site characteristics, rupture directivity, and fault zone amplification. The significance of these effects is determined by measuring the uncertainty-reduction of the GMPE due to each factor. The results show that, in contrast to many regional studies, traditional site characteristic has a relatively minor effect on peak amplitudes in our study area. However, rupture directivity is a significant factor controlling the amplitudes of ground motion even for small events. The dense seismic network and newly developed directivity tool enable us to extract efficiently directivity effects with statistical significance, using the ground-motion dataset during the regression analysis process. The obtained rupture directivities are consistent with the main focal mechanism orientations and surface trace orientations, known from other studies, and predictions for bimaterial ruptures in the trifurcation area of the SJFZ. Fault zone amplification is a second important factor, showing strong impact on the peak ground motion values, with increasing role for the lower frequency range (<10 Hz) examined in the 5 % damped PSA values. We also observe signatures of large amplitude-variances, which indicate additional source-related control on the distribution of amplitudes (besides rupture directivity) for aftershocks close in time and location to the M _L 5.1 earthquake of March 2013. Using the full set of records we present the most complete set of GMPEs for the SJFZ area, including a higher-amplitude prediction for regions in the direction of rupture.     

Determination of GMPE functional form for an active region with limited strong motion data: application to the Himalayan region

Advancement in the seismic networks results in formulation of different functional forms for developing any new ground motion prediction equation (GMPE) for a region. Till date, various guidelines and tools are available for selecting a suitable GMPE for any seismic study area. However, these methods are efficient in quantifying the GMPE but not for determining a proper functional form and capturing the epistemic uncertainty associated with selection of GMPE. In this study, the compatibility of the recent available functional forms for the active region is tested for distance and magnitude scaling. Analysis is carried out by determining the residuals using the recorded and the predicted spectral acceleration values at different periods. Mixed effect regressions are performed on the calculated residuals for determining the intra- and interevent residuals. Additionally, spatial correlation is used in mixed effect regression by changing its likelihood function. Distance scaling and magnitude scaling are respectively examined by studying the trends of intraevent residuals with distance and the trend of the event term with magnitude. Further, these trends are statistically studied for a respective functional form of a ground motion. Additionally, genetic algorithm and Monte Carlo method are used respectively for calculating the hinge point and standard error for magnitude and distance scaling for a newly determined functional form. The whole procedure is applied and tested for the available strong motion data for the Himalayan region. The functional form used for testing are five Himalayan GMPEs, five GMPEs developed under NGA-West 2 project, two from Pan-European, and one from Japan region. It is observed that bilinear functional form with magnitude and distance hinged at 6.5 M _w and 300 km respectively is suitable for the Himalayan region. Finally, a new regression coefficient for peak ground acceleration for a suitable functional form that governs the attenuation characteristic of the Himalayan region is derived.     

Ground-motion prediction equations for interface earthquakes of M7 to M9 based on empirical data from Japan

Ground motion prediction equations (GMPEs) have a major impact on seismic hazard estimates, because they control the predicted amplitudes of ground shaking. The prediction of ground-motion amplitudes due to mega-thrust earthquakes in subduction zones has been hampered by a paucity of empirical ground-motion data for the very large magnitudes (moment magnitude (M) $>$ 7) of most interest to hazard analysis. Recent data from Tohoku M9.0 2011 earthquake are important in this regard, as this is the largest well-recorded subduction event, and the only such event with sufficient data to enable a clear separation of the overall source, path and site effects. In this study, we use strong-ground-motion records from the M9 Tohoku event to derive an event-specific GMPE. We then extend this M9 GMPE to represent the shaking from other M $>$ 7 interface events in Japan by adjusting the source term. We focus on events in Japan to reduce ambiguity that results when combining data in different regions having different source, path and site effect attributes. Source levels (adjustment factors) for other Japanese events are determined as the average residuals of ground-motions with respect to the Tohoku GMPE, keeping all other coefficients fixed. The mean residuals (source terms) scale most steeply with magnitude at the lower frequencies; this is in accord with expectations based on overall source-scaling concepts. Interpolating source terms over the magnitude range of 7.0–9.0, we produce a GMPE for large interface events of M7–M9, for NEHRP B/C boundary site conditions (time-averaged shear-wave velocity of 760 m/s over the top 30 m) in both fore-arc and back-arc regions of Japan. We show how these equations may be adjusted to account for the deeper soil profiles (for the same value of $\hbox {V}_\mathrm{S30})$ in western North America. The proposed GMPE predicts lower motions at very long periods, higher motions at short periods, and similar motions at intermediate periods, relative to the simulation-based GMPE model of Atkinson and Macias (2009) for the Cascadia subduction zone.     

Site specific prediction equations for peak acceleration of ground motion due to earthquakes induced by underground mining in Legnica-Głogów Copper District in Poland

Ground motion database from the region of ?elazny Most tailings pond, the largest in Europe ore-flotation waste repository, is used to identify ground motion prediction equations (GMPE-s) for peak horizontal and peak vertical acceleration. A GMPE model including both geometrical spreading and anelastic damping terms cannot be correctly identified and the model with only spreading term is accepted. The analysis of variance of this model’s residuals with station location as grouping variable indicates that station locations contribute significantly to the observed ground motion variability. Therefore, a site specific GMPE model with relative site amplifications is assessed. Despite short distances among stations, the amplification considerably vary from point to point, up to 1.8 times for the horizontal and 3.5 times for the vertical peak amplitude. The model including site effects enhances GMPE-s fit to observations, explains more than 60% dependent variables variability and correctly accounts for site effects.     

Pan-European ground-motion prediction equations for the average horizontal component of PGA,PGV, and 5 %-damped PSA at spectral periods up to 3.0 s using the RESORCE dataset

This article presents a set of Ground-Motion Prediction Equations (GMPEs) for Europe and the Middle East, derived from the RESORCE strong motion data bank, following a standard regression approach. The parametric GMPEs are derived for the peak ground acceleration, peak ground velocity, and 5 %-damped pseudo-absolute acceleration response spectra computed over 23 periods between 0.02 and 3 s, considering the average horizontal-component ground-motions. The GMPEs are valid for distances less than 300 km, hypocentral depth up to 35 km and over the magnitude range 4–7.6. Two metrics for the source-to-station distance (i.e. Joyner-Boore and hypocentral) are considered. The selected dataset is composed by 2,126 recordings (at a period of 0.1 s) related to 365 earthquakes, that includes strong-motion data from 697 stations.The EC8 soil classification (four classes from A to D) discriminates recording sites and four classes (normal, reverse, strike-slip, and unspecified) describe the style of faulting. A subset which contains only stations with measured Vs30 and earthquakes with specified focal mechanism (1,224 records from 345 stations and 255 earthquakes) is used to test of the accuracy of the median prediction and the variability associated to the broader data set. A random effect regression scheme is applied and bootstrap analyses are performed to estimate the 95 % confidence levels for the parameters. The total standard deviation sigma is decomposed into between-events and within-event components, and the site-to-site component is evaluated as well. The results show that the largest contribution to the total sigma is coming from the within-event component. When analyzing the residual distributions, no significant trends are observed that can be ascribed to the earthquake type (mainshock-aftershock classification) or to the non-linear site effects. The proposed GMPEs have lower median values than global models at short periods and large distances, while are consistent with global models at long periods $(\hbox {T} > 1)$  s. Consistency is found with two regional models developed for Turkey and Italy, as the considered dataset is dominated by waveforms recorded in these regions.     

Ground motion prediction equations for horizontal and vertical components of acceleration in Northern Iran

Recent studies have shown that the vertical component of ground motion can be quite destructive on a variety of structural systems. Development of response spectrum for design of buildings subjected to vertical component of earthquake needs ground motion prediction equations (GMPEs). The existing GMPEs for northern Iranian plateau are proposed for the horizontal component of earthquake, and there is not any specified GMPE for the vertical component of earthquake in this region. Determination of GMPEs is mostly based on regression analyses on earthquake parameters such as magnitude, site class, distance, and spectral amplitudes. In this study, 325 three-component records of 55 earthquakes with magnitude ranging from M _w 4.1 to M _w 7.3 are used for estimation on the regression coefficients. Records with distances less than 300 km are selected for analyses in the database. The regression analyses on earthquake parameters results in determination of GMPEs for peak ground acceleration and spectral acceleration for both horizontal and vertical components of the ground motion. The correlation between the models for vertical and horizontal GMPEs is studied in details. These models are later compared with some other available GMPEs. According to the result of this investigation, the proposed GMPEs are in agreement with the other relationships that were developed based on the local and regional data.     

Stochastic ground motion models to NGA-West2 and NGA-Sub databases using Bayesian neural network

Ground motion models (GMMs) are traditionally developed from a frequentist approach. The Bayesian framework has received recent attention in developing nonergodic models, measuring uncertainty, or updating the model with additional data. However, no neural networks are developed to date in this framework to predict ground motion parameters or spectra. Hence, the present work develops a probabilistic Bayesian neural network (PBNN) to next-generation attenuation – West2 and Subduction databases using variational inference with mean-field assumption. Network inputs are magnitude, rupture distance, hypocentral depth, shear wave velocity, style of faulting, and region flags; outputs are peak ground values and response spectra. Both models have two hidden layers with seven neurons in each hidden layer. The models are verified for potential overfit, and their performance is validated through the parametric study by varying inputs. The output of a deterministic model is a point estimate. Considering probabilistic layers in hidden and output layers enables the model to capture within-model epistemic uncertainty and aleatory variability. Obtained aleatory standard deviations are consistent with other models. Mean epistemic uncertainty and aleatory variability are in the range 0.07–0.10 and 0.62–0.78 (ln units) for NGA-West2 and 0.09–0.16 and 0.67–0.95 for NGA-Sub models, respectively. The correlation coefficients between recorded and overall mean predictions ranged from 0.94 to 0.97 for NGA-the West2 model and from 0.91 to 0.95 for the NGA-Sub models. Network performance for out-of-training inputs showed increased epistemic deviations with no effect on aleatory deviations.     

Empirical attenuation relationship for Arias Intensity

Arias Intensity is a ground motion parameter that captures the potential destructiveness of an earthquake as the integral of the square of the acceleration–time history. It correlates well with several commonly used demand measures of structural performance, liquefaction, and seismic slope stability. A new empirical relationship is developed to estimate Arias Intensity as a function of magnitude, distance, fault mechanism, and site category based on 1208 recorded ground motion data from 75 earthquakes in active plate‐margins. Its functional form is derived from the point‐source model, and the coefficients are determined through non‐linear regression analyses using a random‐effects model. The results show that for large magnitude earthquakes (M > 7) Arias Intensity was significantly overestimated by previous relationships while it was underestimated for smaller magnitude events (M ? 6). The average horizontal Arias Intensity is not significantly affected by forward rupture directivity in the near‐fault region. The aleatory variability associated with Arias Intensity is larger than that of most other ground motion parameters such as spectral acceleration. However, it may be useful in assessing the potential seismic performance of stiff engineering systems whose response is dominated by the short‐period characteristics of ground motions. Copyright © 2003 John Wiley & Sons, Ltd.     

CyberShake: A Physics-Based Seismic Hazard Model for Southern California

CyberShake, as part of the Southern California Earthquake Center??s (SCEC) Community Modeling Environment, is developing a methodology that explicitly incorporates deterministic source and wave propagation effects within seismic hazard calculations through the use of physics-based 3D ground motion simulations. To calculate a waveform-based seismic hazard estimate for a site of interest, we begin with Uniform California Earthquake Rupture Forecast, Version 2.0 (UCERF2.0) and identify all ruptures within 200?km of the site of interest. We convert the UCERF2.0 rupture definition into multiple rupture variations with differing hypocenter locations and slip distributions, resulting in about 415,000 rupture variations per site. Strain Green Tensors are calculated for the site of interest using the SCEC Community Velocity Model, Version 4 (CVM4), and then, using reciprocity, we calculate synthetic seismograms for each rupture variation. Peak intensity measures are then extracted from these synthetics and combined with the original rupture probabilities to produce probabilistic seismic hazard curves for the site. Being explicitly site-based, CyberShake directly samples the ground motion variability at that site over many earthquake cycles (i.e., rupture scenarios) and alleviates the need for the ergodic assumption that is implicitly included in traditional empirically based calculations. Thus far, we have simulated ruptures at over 200 sites in the Los Angeles region for ground shaking periods of 2?s and longer, providing the basis for the first generation CyberShake hazard maps. Our results indicate that the combination of rupture directivity and basin response effects can lead to an increase in the hazard level for some sites, relative to that given by a conventional Ground Motion Prediction Equation (GMPE). Additionally, and perhaps more importantly, we find that the physics-based hazard results are much more sensitive to the assumed magnitude-area relations and magnitude uncertainty estimates used in the definition of the ruptures than is found in the traditional GMPE approach. This reinforces the need for continued development of a better understanding of earthquake source characterization and the constitutive relations that govern the earthquake rupture process.     

Ground motion prediction equation considering combined dataset of recorded and simulated ground motions

Himalayan region is one of the most active seismic regions in the world and many researchers have highlighted the possibility of great seismic event in the near future due to seismic gap. Seismic hazard analysis and microzonation of highly populated places in the region are mandatory in a regional scale. Region specific Ground Motion Predictive Equation (GMPE) is an important input in the seismic hazard analysis for macro- and micro-zonation studies. Few GMPEs developed in India are based on the recorded data and are applicable for a particular range of magnitudes and distances. This paper focuses on the development of a new GMPE for the Himalayan region considering both the recorded and simulated earthquakes of moment magnitude 5.3–8.7. The Finite Fault simulation model has been used for the ground motion simulation considering region specific seismotectonic parameters from the past earthquakes and source models. Simulated acceleration time histories and response spectra are compared with available records. In the absence of a large number of recorded data, simulations have been performed at unavailable locations by adopting Apparent Stations concept. Earthquakes recorded up to 2007 have been used for the development of new GMPE and earthquakes records after 2007 are used to validate new GMPE. Proposed GMPE matched very well with recorded data and also with other highly ranked GMPEs developed elsewhere and applicable for the region. Comparison of response spectra also have shown good agreement with recorded earthquake data. Quantitative analysis of residuals for the proposed GMPE and region specific GMPEs to predict Nepal–India 2011 earthquake of Mw of 5.7 records values shows that the proposed GMPE predicts Peak ground acceleration and spectral acceleration for entire distance and period range with lower percent residual when compared to exiting region specific GMPEs.     

Seismic hazard assessment of Western Coastal Province of Saudi Arabia:deterministic approach

Seismic hazard assessment is carried out by utilizing deterministic approach to evaluate the maximum expected earthquake ground motions along the Western Coastal Province of Saudi Arabia. The analysis is accomplished by incorporating seismotectonic source model,determination of earthquake magnitude(Mmax), set of appropriate ground motion predictive equations(GMPE), and logic tree sequence. The logic tree sequence is built up to assign weight to ground motion scaling relationships. Contour maps of ground acceleration are generated at different spectral periods. These maps show that the largest ground motion values are emerged in northern and southern regions of the western coastal province in Saudi Arabia in comparison with the central region.     

Criteria for Selecting and Adjusting Ground-Motion Models for Specific Target Regions: Application to Central Europe and Rock Sites

A vital component of any seismic hazard analysis is a model for predicting the expected distribution of ground motions at a site due to possible earthquake scenarios. The limited nature of the datasets from which such models are derived gives rise to epistemic uncertainty in both the median estimates and the associated aleatory variability of these predictive equations. In order to capture this epistemic uncertainty in a seismic hazard analysis, more than one ground-motion prediction equation must be used, and the tool that is currently employed to combine multiple models is the logic tree. Candidate ground-motion models for a logic tree should be selected in order to obtain the smallest possible suite of equations that can capture the expected range of possible ground motions in the target region. This is achieved by starting from a comprehensive list of available equations and then applying criteria for rejecting those considered inappropriate in terms of quality, derivation or applicability. Once the final list of candidate models is established, adjustments must be applied to achieve parameter compatibility. Additional adjustments can also be applied to remove the effect of systematic differences between host and target regions. These procedures are applied to select and adjust ground-motion models for the analysis of seismic hazard at rock sites in West Central Europe. This region is chosen for illustrative purposes particularly because it highlights the issue of using ground-motion models derived from small magnitude earthquakes in the analysis of hazard due to much larger events. Some of the pitfalls of extrapolating ground-motion models from small to large magnitude earthquakes in low seismicity regions are discussed for the selected target region.     

Basin effects on ground motion: the case of a high-resolution experiment in Cephalonia (Greece)

In this study data and results of a high-resolution experiment in Cephalonia (Greece) regarding empirical basin effects are presented. A total of 59 velocimeters and 17 accelerometers were deployed in the basin of Argostoli Cephalonia (Greece), for a period of 7 months (September 2011–April 2012). Due to high seismicity of the western Greece and surrounding area this array recorded thousands of local, regional and global events. Data used in this work come from a selection of 162 regional and local earthquakes, 3 km ≤ R ≤ 600 km, with magnitude range, 1.0 ≤ M ≤ 5.2. Based on high signal-to-noise ratio recordings and two selected reference stations, variation of several intensity measures (PGA, PGV, Arias Intensity, Cumulative Absolute Velocity), significant duration, HVSR and SSR of ground motion recordings on soil sites within the basin is carefully examined for a range of frequencies of engineering interest. Comparison of results with a detailed 2D geologic model shows a good consistency both in amplification and frequency domain. Influence of “reference” site on ground motion variation of soil sites is also discussed in light of our results. Finally, it is suggested that 2D or/and 3D theoretical modeling should be performed given the availability of geological and geophysical parameters to define a realistic model of the basin. Results of this study can undoubtedly serve in model validation and improvement of ground motion simulation tools.     

Strong-motion observations from the OGS temporary seismic network during the 2012 Emilia sequence in northern Italy

Strong-motion data consisting of peak ground acceleration and velocity and 5 % damped response spectra are presented for 46 earthquakes of the Emilia seismic sequence which occurred in the Po Plain (northern Italy) in 2012. The data were recorded by the OGS temporary network installed close to the town of Ferrara following the main shock of May 20, 2012. Ground-motion peak parameters and spectral responses are compared with the ground-motion prediction equation (GMPE) of Bindi et al. (Bull Earthq Eng 9:1899–1920, 2011) for soft soils and reverse faults. Peak ground accelerations are in general in good agreement with those predicted by GMPE, while predicted peak ground velocities underestimate the observed data, especially for stronger events at more distant stations. The response spectra follow the trend in peak ground velocities, with observed values higher than predicted values at longer periods. This behavior has been interpreted as a site effect due to the deep soft alluvial cover of the Po Plain, which promotes ground motion characterized by a large low-frequency spectral content that is not yet well modeled by the Italian GMPE. A peculiar behavior was shown by the event occurring on June 6, 04:08:33 UTC, \(\hbox {M}=4.5\) , located at the eastern edge of the Po Plain, which produced peak ground accelerations exceeding three times the values estimated by attenuation laws. Such a great discrepancy could be related to post-critically reflected S-waves and multiples from the Moho (SmSM).     

A uniform approach to seismic site effect analysis in Bucharest, Romania

As a uniform approach to the assessment of ground motion variation within the Romanian capital Bucharest we analyze and compare strong motion records from analog recorders, weak motion data from a modern digital accelerometer network, and intensity observations of previous strong earthquakes. These different data sets allow to clearly characterize geographical trends in the distribution of ground shaking in the city for future earthquakes. Below 2 Hz the variability is small. Between 2 and 5 Hz, however, variations by a factor of 3–4 have to be expected. As the key source for the seismic hazard—the intermediate depth Vrancea earthquakes—remain at hypocentral distances in excess of 150 km from the city the ground motion variation must be predominantly attributed to site effects. This geometry of Vrancea sources to the site of Bucharest is ideal for the application of source-site separation techniques. However, despite this fact site effect amplification functions display a very large amount of aleatory uncertainty. In other words the standard source-site parameterization is too simple and we do not yet fully understand the cause and size of site effects.     

Understanding single-station ground motion variability and uncertainty (sigma): lessons learnt from EUROSEISTEST

Accelerometric data from the well-studied valley EUROSEISTEST are used to investigate ground motion uncertainty and variability. We define a simple local ground motion prediction equation (GMPE) and investigate changes in standard deviation (σ) and its components, the between-event variability (τ) and within-event variability (φ). Improving seismological metadata significantly reduces τ (30–50%), which in turn reduces the total σ. Improving site information reduces the systematic site-to-site variability, φ _S2S (20–30%), in turn reducing φ, and ultimately, σ. Our values of standard deviations are lower than global values from literature, and closer to path-specific than site-specific values. However, our data have insufficient azimuthal coverage for single-path analysis. Certain stations have higher ground-motion variability, possibly due to topography, basin edge or downgoing wave effects. Sensitivity checks show that 3 recordings per event is a sufficient data selection criterion, however, one of the dataset’s advantages is the large number of recordings per station (9–90) that yields good site term estimates. We examine uncertainty components binning our data with magnitude from 0.01 to 2 s; at smaller magnitudes, τ decreases and φ _SS increases, possibly due to κ and source-site trade-offs Finally, we investigate the alternative approach of computing φ _SS using existing GMPEs instead of creating an ad hoc local GMPE. This is important where data are insufficient to create one, or when site-specific PSHA is performed. We show that global GMPEs may still capture φ _SS , provided that: (1) the magnitude scaling errors are accommodated by the event terms; (2) there are no distance scaling errors (use of a regionally applicable model). Site terms (φ _S2S ) computed by different global GMPEs (using different site-proxies) vary significantly, especially for hard-rock sites. This indicates that GMPEs may be poorly constrained where they are sometimes most needed, i.e., for hard rock.