ABSTRACT In the framework of site-specific seismic hazard assessment, the definition of reference motion is a crucial step. Reference motion is generally associated with hard-rock conditions, characterized by S-wave velocity (VS) exceeding 1500 m/s. However, ground motion recorded at sites with such conditions is underrepresented in existing strong-motion databases. Consequently, the validity domains of most empirical ground-motion models (GMMs) are not representative of reference hard-rock conditions. To address this limitation, we consider the empirical approach to retrieve and model reference ground motion for shallow crustal earthquakes in a seismically active region, proposed by Shible et al. (2023) using the Japanese Kiban–Kyoshin network data. Following this approach, we apply a deconvolution of site responses from strong-motion recordings to estimate ground motion at reference conditions in the Euro-Mediterranean region. The first step involves compiling a large database by merging the Engineering Strong-Motion Database (ESM database, Luzi et al., 2020) with additional data from the Greek, French, and Spanish networks. The second step involves estimating site response at network stations using both parametric and nonparametric generalized inversion techniques (GITs). A careful selection of reference hard-rock stations is made to constrain the inversions. The database signals are then deconvolved using the site terms obtained by GIT, allowing us to virtually bring the sites of the entire database to very high VS values, similar to outcropping bedrock conditions. GMMs are then determined for reference conditions, and the results are discussed. The derived GMMs show limited site-to-site variability (0.1–0.2 in natural log), indicating effective removal of site responses from data. In addition, an average amplification model, based on VS30, is proposed to complement the reference GMM prediction in cases where a site-specific soil response estimate is unavailable.

Estimating earthquake ground motions at reference bedrock is a major issue in site-specific seismic hazard assessment. Deriving or adjusting empirical ground motion models (GMMs) for reference bedrock is challenging and affected by large epistemic uncertainties. We propose a methodology to simulate region-specific reference bedrock time histories by combining spectral decompositions of ground motions with Empirical Green’s Functions (EGFs) simulation technique. First, we adopt the nonparametric spectral decomposition approach to separate the contribution of source, path, and site. We remove the average source and site effects from observed small-magnitude recordings in the target region through deconvolution in the Fourier domain. This way, the obtained deconvolved EGFs represent path term only. Then, we couple the EGFs with k− 2 kinematic rupture models for target scenario events. For each target magnitude, a set of rupture models following a ω-squared source spectrum are generated sampling the uncertainties in kinematic source parameters (e.g., slip distribution, rupture velocity, hypocentral location, stress drop, and rupture dimensions). The proposed approach is validated using recorded ground motions at reference sites from multiple earthquakes in Central Italy. The power of this approach lies in its ability to map the path-specific effects into the ground-motion field, providing 3-component time histories covering a wide frequency range, without the need for computationally expensive approaches to simulate 3D wave propagation. The region-specific, site-effects-free dataset produced by this approach can be used alone or in combination with existing empirical datasets to adjust existing GMMs, derive new GMMs, or select hazard-consistent time histories to be used in soil and structural response analyses.

The quantification of earthquake-generated site-specific ground motions is necessary for accurate and reliable seismic hazard assessment for critical structures. To this aim, empirical and numerical approaches are typically used in order to characterize the site response at the target site. This is not straightforward in low-seismicity regions and even more challenging when the target structure is located at great depth. This article, focusing on a practical application for an underground radioactive waste repository project to be located in the eastern part of the Paris Basin (France), presents and discusses the results and the faced challenges in the characterization of the site transfer function (TF) from surface to several depths as well as when assessing the applicability of the ergodic site terms from empirical ground motion models to the target site. We first present the data collected in the last years at the project site and the empirical TFs between the surface and two locations at depth (–490 m and –445 m). The empirical transfer function is robust in the frequency range 0.5–10 Hz and less reliable above 10 Hz due to the reduced number of usable recordings. A good consistency is found with the numerical transfer function based on a 1D soil model derived from local boreholes data. A transfer function based on noise recordings is also computed and compared to the earthquake-based one showing a good agreement of the horizontal components. These results clearly open several perspectives on the potential use of ambient noise data and numerical modelling in order to quantify the site response at any point of the underground project layout. Then, we investigate the site-specific high-frequency decay parameter $${\kappa }_{0}$$ using the available earthquake recordings at the surface. The estimates of $${\kappa }_{0}$$ obtained combining both acceleration- and displacement-based approaches are affected by large uncertainties owing mainly to the lack of data at short distances (< than 100 km). Nonetheless, a range of $${\kappa }_{0}$$ values having a central $${\kappa }_{0}$$ =0.042 s and lower and upper bounds of 0.02 s and 0.058 s is proposed and $$Vs-{\kappa }_{0}$$ adjustment factors are calculated using the IRVT approach (Al Atik 2014 BSSA 104:336–346 ) for few example cases. Such adjustment factors together with the empirical TFs and related uncertainties can be used to remove the ergodic assumption on the site term of GMMs in order to improve classical ergodic seismic hazard assessment for structures at surface and at depth.

Nepal is one of the most seismically active regions in the world, as highlighted by the recent devastating 2015, Mw~7.8 Gorkha earthquake, and a robust assessment of seismic hazard is paramount for the design of earthquake-resistant structures. In this study, we present a new probabilistic seismic hazard assessment (PSHA) for Nepal. We considered data and findings from recent scientific publications, which allowed us to develop a unified magnitude homogenized seismicity catalog and propose alternative seismic source characterization (SSC) models including up-to-date parameters of major thrust faults like main frontal thrust (MFT) and main boundary thrust (MBT), while also considering existing SSC models and various seismic hazard modeling strategies within a logic tree framework. The sensitivity analyses show the seismic hazard levels are generally higher for SSC models integrating the major thrust faults, followed by homogenous volume sources and smoothed seismicity approach. The seismic hazard maps covering the entirety of Nepal are presented as well as the uniform hazard spectra (UHS) for five selected locations (Kathmandu, Pokhara, Biratnagar, Nepalganj, and Dipayal) at return periods of 475- and 2475-years considering Vs,30 = 760 m/s. The results obtained are generally consistent with most recent studies. However, a notable variability in seismic hazard levels and several discrepancies with respect to the Nepal Building Building Code NBC105: 2020 and global hazard model, GEM are noted, and possible causes are discussed.

Please join us in thanking all those scientists and experts in the various fields represented in Journal of Seismology for devoting time and effort to review the papers that we have been sending them. The Editor-in-Chief and Publisher acknowledge the colleagues listed below for their excellent reviews of papers for which final decisions have beenmade during the period 1 January 2019 to 31 December 2019. J Seismol https://doi.org/10.1007/s10950-020-09903-w

Figure 20 of Drouet et al. (2020) shows the mean PGA at 475 years of return period instead of the median PGA at 475 years of return period as written in the figure caption.

This work proposes a new approach, based on Bayesian updating and extreme value statistics to determine the maximum magnitudes for truncated magnitude-frequency distributions such as the Gutenberg Richter model in the framework of Probabilistic Seismic Hazard Analyses. Only the maximum observed magnitude and the associated completeness period are required so that the approach is easy to implement and there is no need to determine and use the completeness periods for smaller events. The choice of maximum magnitudes can have a major impact on hazard curves when long return periods as required for safety analysis of nuclear power plants are considered. Here, not only a singular value but a probability distribution accounting for prior information, data and uncertainty is provided. Moreover, uncertainties related to magnitude frequency distributions, including the uncertainty related to the maximum observed magnitude are discussed and accounted for. The accuracy of the approach is validated based on simulated catalogues with various parameter values. Then the approach is applied to French data for a specific region characterized by high-seismic activity in order to determine the maximum magnitude distribution and to compare the results to other approaches.

ABSTRACT Production-induced earthquakes in the Groningen gas field caused damage to buildings and concerns for the population, the gas-field owner, and the local and national authorities and institutions. The largest event (ML=3.6) occurred in 2012 near Huizinge, and, despite the subsequent decision of the Dutch government to reduce the gas production in the following years, similar magnitude events occurred in 2018 and 2019 (ML=3.4). Thanks to the improvement of the local seismic networks in the last years, recent events provide a large number of recordings and an unprecedented opportunity to study the characteristics of induced earthquakes in the Groningen gas field and related ground motions. In this study, we exploit the S-wave Fourier amplitude spectra recorded by the 200 m depth borehole sensors of the G network from 2015 to 2019 to derive source and attenuation parameters for ML≥2 induced earthquakes. The borehole spectra are decomposed into source, attenuation, and site nonparametric functions, and parametric models are then adopted to determine moment magnitudes, corner frequencies, and stress drops of 21 events. Attenuation and source parameters are discussed and compared with previous estimates for the region. The impact of destructive interference of upgoing and downgoing waves at borehole depth on the derived parameters is also discussed and assessed to be minor. The analysis of the apparent source spectra reveals that several events show rupture directivity and provides clear observations of frequency-dependent directivity effects in induced earthquakes. The estimated rupture direction shows a good agreement with orientation of pre-existing faults within the reservoir. Our results confirm that rupture directivity is still an important factor for small-magnitude induced events, affecting the amplitude of recorded short-period response spectra and causing relevant spatial ground-motion variability.

We developed a ground-motion simulation code base on extended rupture modeling combined with the use of empirical Green's functions (EGFs), adapted for low-to-moderate seismicity regions (with a limited set of EGFs), and extended its range of applicability to the lowest source-to-site distances. This code is based on a kinematic source description of an extended fault and is designed to allow complex fault geometries and to generate a ground motion variability in agreement with that of the recorded databases. The code is developed to work with a sparse set of EGFs. Each available EGF is therefore used in several positions on the rupture area. To be used in positions different of their original position, we applied to the EGFs some adjustments. In addition to the classical adjustments (i.e. time delay correction, geometrical spreading correction and anelastic attenuation correction), we propose here a radiation pattern adjustment. The effectiveness of it is tested in a numerical application. We showed noticeable improvements at the lowest distances, and some limitations when approaching the nodal planes of the subevents the recording of which were used as EGFs. We took advantage of the development of this code, its ability to work with a sparse set of EGFs, its ability to take into account complex fault geometries and its ability to master the general variability, to perform a ground-motion simulation scenario on the Middle Durance Fault (MDF). We perform simulations for a hard rock site (V-S30 = 1800 m/s) and a sediment site (V-S30 = 440 m/s) of the CEA Nuclear Research Site of Cadarache (France), and compared the computed ground motion with several ground motion prediction equations (GMPEs). The GMPEs slightly underestimate the sediment site but strongly overestimate the ground motion amplitude on the hard rock site, even when using a specific correction factor which adapts GMPEs predictions from rock site to hard rock site. This general ascertainment confirms the need to continue efforts towards the establishment of consistent GMPEs applicable to hard-rock conditions.

1 3 Ahmed Abdel-Aziz M. Abdullah Sandikkaya Hamdy Abou-Elfath Lars Abrahamczyk Christoph Adam Jose Adam Naida Ademovic Rohit Adhikari Mahdi Adibi Hessam Afzali Rafael Aguilar Naveed Ahmad Aybige Akinci Sinan Akkar M. Shahria Alam Ayad Aldeka Alper Aldemir David Alexander Nicholas Alexander Gerardo Alguacil Cenk Alhan Muhammad Usman Ali Reza Allahvirdizadeh Joao Almeida Ahmet Can Altunişik Gonzalo Montalva Alvarado Claudio Amadio Giuseppina Amato Osama Amer Gabriele Ameri Sara Amoroso Hamidreza Anajafi Anthimos S. Anastasiadis Ioannis Anastasopoulos Nurdan Apaydin Georgios Apostolakis Amjad Aref Ersin Arel Juana Arias-Trujillo Danny Arroyo Carlos Arteta Aysegul Askan Panagiotis Asteris A. Athanatopoulou-Kyriakou Samuel Auclair Stefano Aversa Alberto Maria Avossa Özgür Avşar Konstantinos Bakalis Jack Baker Ihsan Engin Bal Georgios Baltzopoulos mehdi Banazadeh Arnab Banerjee Simone Barani Andre R. Barbosa Pierre-Yves Bard Selcuk Bas Alberto Basaglia Dhiman Basu Ramiro Bazaez Mohammad Hassan Baziar Céline Beauval Tracy Becker Farhad Behnamfar Andrea Belleri Amadeo Benavent-Climent Andrea Benedetti Rita Bento Anne Bergere Mehmet Berilgen Bjarni Bessason Michele Betti Katrin Beyer Kaiming Bi Huseyin Bilgin Dino Bindi Giovanni Biondi Anna Birely Gülüm Birgören-Tanırcan ACKNOWLEDGMENTS

Please join us in thanking all those scientists and experts in the various fields represented in Journal of Seismology for devoting time and effort to review the papers that we have been sending them. The editor and publisher acknowledge the colleagues listed below for their excellent reviews of papers for which final decisions have been made during the period January 2018 to December 2018. J Seismol (2019) 23:195–197 https://doi.org/10.1007/s10950-019-09815-4

The undergoing study aims at the elaboration of a probabilistic seismic hazard maps for metropolitan France taking into account the outcomes of recent research projects such as SIGMA (Research on SeIsmic Ground Motion Assessment, 2011 – 2016), which was devoted to improve knowledge on data, methods and tools to better quantify uncertainties in seismic hazard estimates. A new earthquake catalogue for France was developed including a revision of magnitude and depth for historical events and of location and magnitude for instrumental events. This catalogue is used as the backbone catalogue for the present study. Three area source models developed independently by 3 institutions are considered for which seismic activity is characterized using the Gutenberg-Richter model including an exploration of earthquake location and magnitude uncertainty. In addition, a zoneless approach is considered using spatially-adaptive kernel functions. The concept of large seismotectonic domains is introduced in order to constrain the estimation of the Gutenberg-Richter b-value and to derive maximum magnitude distributions using the Bayesian approach. The ground-motion model includes two GMPEs developed specifically for France as well as two additional models. Tests were carried on in order to investigate if the adopted ground-motion characterization logic tree captures epistemic uncertainty. Uncertainties related to the seismic activity (parameters a and b of the Gutenberg-Richter models), maximum magnitude, and hypocentral depth is also explored. The objectif of this study is to produce probabilistic hazard maps for a grid with a spatial sampling of 10 km for two return periods (475 and 2475 years) and three spectral periods (PGA, 0.2 and 1.0 seconds).

The accurate evaluation and appropriate treatment of uncertainties is of primary importance in modern probabilistic seismic hazard assessment (PSHA). One of the objectives of the SIGMA project was to establish a framework to improve knowledge and data on two target regions characterized by low-to-moderate seismic activity. In this paper, for South-Eastern France, we present the final PSHA performed within the SIGMA project. A new earthquake catalogue for France covering instrumental and historical periods was used for the calculation of the magnitude-frequency distributions. The hazard model incorporates area sources, smoothed seismicity and a 3D faults model. A set of recently developed ground motion prediction equations (GMPEs) from global and regional data, evaluated as adequately representing the ground motion characteristics in the region, was used to calculate the hazard. The magnitude-frequency distributions, maximum magnitude, faults slip rate and style-of-faulting are considered as additional source of epistemic uncertainties. The hazard results for generic rock condition (Vs30 = 800 m/s) are displayed for 20 sites in terms of uniform hazard spectra at two return periods (475 years and 10,000 years). The contributions of the epistemic uncertainties in the ground motion characterizations and in the seismic source characterization to the total hazard uncertainties are analyzed. Finally, we compare the results with existing models developed at national scale in the framework of the first generation of models supporting the Eurocode 8 enforcement, (MEDD 2002 and AFPS06) and at the European scale (within the SHARE project), highlighting significant discrepancies at short return periods.

In the framework of the SIGMA project, a study was launched to develop a parametric earthquake catalog for the historical period, covering the metropolitan territory and calibrated in Mw. A set of candidate calibration events was selected corresponding to earthquakes felt over a part of the French metropolitan territory, which are fairly well documented both in terms of macroseismic intensity distributions (SisFrance BRGM-EDF-IRSN) and magnitude estimates. The detailed analysis of the macroseismic data led us to retain only 30 events out of 65 with Mw ranging from 3.6 to 5.8. In order to supplement the dataset with data from larger magnitude events, Italian earthquakes were also considered (11 events posterior to 1900 with Mw ≥ 6.0 out of 15 in total), using both the DBMI11 macroseismic database (Locati et al. in Seismol Resour Lett 85(3):727–734, 2014) and the parametric information from the CPTI11 (Rovida et al. in CPTI11, la versione 2011 del Catalogo Parametrico dei Terremoti Italiani Istituto Nazionale di Geofisica et Vulcanologia, Milano, Bologna, 2011.  https://doi.org/10.6092/ingv.it-cpti11). To avoid introducing bias related to the differences in terms of intensity scales (MSK vs. MCS), only intensities smaller than or equal to VII were considered (Traversa et al. in On the use of cross-border macroseismic data to improve the estimation of past earthquakes seismological parameters, 2014). Mw and depth metadata were defined according to the Si-Hex catalogue (Cara et al. in Bull Soc Géol Fr 186:3–19, 2015.  https://doi.org/10.2113/qssqfbull.186.1.3), published information, and to the specific worked conducted within SIGMA related to early instrumental recordings (Benjumea et al. in Study of instrumented earthquakes that occurred during the first part of the 20th century (1905–1962), 2015). For the depth estimates, we also performed a macroseismic analysis to evaluate the range of plausible estimates and check the consistency of the solutions. Uncertainties on the metadata related to the calibration earthquakes were evaluated using the range of available alternative estimates. The intensity attenuation models were developed using a one-step maximum likelihood scheme. Several mathematical formulations and sub-datasets were considered to evaluate the robustness of the results (similarly to Baumont and Scotti in Accounting for data and modeling uncertainties in empirical macroseismic predictive equations (EMPEs). Towards “European” EMPEs based on SISFRANCE, DBMI, ECOS macroseismic database, 2008). In particular, as the region of interest may be characterized by significant laterally varying attenuation properties (Bakun and Scotti in Geophys J Int 164:596–610, 2006; Gasperini in Bull Seismol Soc Am 91:826–841, 2001), we introduced regional attenuation terms to account for this variability. Two zonation schemes were tested, one at the national scale (France/Italy), another at the regional scale based on the studies of Mayor et al. (Bull Earthq Eng, 2017.  https://doi.org/10.1007/s10518-017-0124-8) for France and Gasperini (2001) for Italy. Between and within event residuals were analyzed in detail to identify the best models, that is, the ones associated with the best misfit and most limited residual trends with intensity and distance. This analysis led us to select four sets of models for which no significant trend in the between- and within-event residuals is detected. These models are considered to be valid over a wide range of Mw covering ~ 3.5–7.0.

Advanced fully site-specific seismic hazard assessm nt methods need two main elements. They need first an accurate estimation of the local amplification. The n, they need a reliable “reference” ground motion a t the base of the soil column. In order to make progress towar d site-specific applications in low-to-moderate sei smicity context, the CASHIMA (Cadarache seismic hazard inte grated multidisciplinary assessment) program coordinated numerous research actions since 2005. T his paper is a synthesis of the achieved works. As for the local amplification estimation, we conclude that bo h simulations and empirical measurements are manda tory: empirical measurements are required to “calibrate” th simulations. Conversely, simulations are mandat ory to extrapolate the amplification estimates beyond the seismic scenarios previously recorded by instrument ation. The definition of the hard-rock ground motion is a critical part of the fully site-specific seismic ha zard study. Indeed, the current Ground Motion Prediction Equati ons (GMPEs) are poorly constrained for hard-rock si tes. The standard procedure is to adjust the GMPEs from the host to the target site conditions by applying so-called VS-κ adjustments. In such approach, the most critical p arameter is the “target” site κ0. In low-seismicity areas, the collection of a sufficient number of records is dif ficult. Moreover, the κ0 physical bases are not fully understood. An alternative approach, that does not use κ0 and that involves a correction of the local amplif ication at each accelerometric site, was developed on a subset of t he Kik-net database. The results showed significant differences with respect to the standard host-to-ta rget approaches. In conclusion, as soon as these ne w GMPEs will be developed to allow the definition of consis tent hard-rock ground motion, including datasets sp ecific to European context, we estimate that fully site-speci fic seismic hazard study can be applied in low-to-m derate

Modern probabilistic seismic hazard assessment (PSHA) focuses on the separation and different treatment of epistemic and aleatory uncertainties. Recent site-specific PSHA studies have pointed out that, if the site response and its epistemic uncertainties can be appropriately considered by adjustments to median estimates from ground motion prediction equations (GMPEs), the aleatory variability (sigma) of the GMPEs can be replaced by the single-station sigma thus partially relaxing the ergodic assumption employed in the PSHA. The site-specific partially nonergodic approach, correctly applied, provides a more accurate representation of the seismic hazard at a specific site and a more rigorous treatment of uncertainties. This paper presents the strategy followed to apply this relatively recent approach to a critical infrastructure in Southern France located on hard-rock site conditions (Vs30 ≈ 2000 m/s). The target site conditions are defined in terms of shear-wave velocity (Vs) profiles and high-frequency attenuation (κ0) based on the results of site investigations and on the exploitation of earthquake records at seismic stations in the target site area. The host-to-target Vs-κ0 adjustment of median estimates for the selected GMPEs is performed by using the inverse random vibration theory approach (Al Atik et al. in Bull Seismol Soc Am 104:336–346, 2014) considering epistemic uncertainties in target Vs profile and κ0. The single-station sigma model is developed based on Rodriguez-Marek et al. (Bull Seismol Soc Am 104:1601–1619, 2013) due to the lack of local data. The results of the site-specific partially nonergodic PSHA are discussed by means of a sensitivity analysis and are compared to the results from standard ergodic PSHA. We found that, for the considered site, the site-specific approach provides a substantial reduction (up to 50%) of the uniform hazard spectra at 10,000-year return period compared to the ergodic approach.

Seismic analysis in the context of nuclear safety in France is currently guided by a pure deterministic approach based on Basic Safety Rule (Règle Fondamentale de Sûreté) RFS 2001-01 for seismic hazard assessment, and on the ASN/2/01 Guide that provides design rules for nuclear civil engineering structures. After the 2011 Tohohu earthquake, nuclear operators worldwide were asked to estimate the ability of their facilities to sustain extreme seismic loads. The French licensees then defined the 'hard core seismic levels', which are higher than those considered for design or re-assessment of the safety of a facility. These were initially established on a deterministic basis, and they have been finally justified through state-of-the-art probabilistic seismic hazard assessments. The appreciation and propagation of uncertainties when assessing seismic hazard in France have changed considerably over the past 15 years. This evolution provided the motivation for the present article, the objectives of which are threefold: (1) to provide a description of the current practices in France to assess seismic hazard in terms of nuclear safety; (2) to discuss and highlight the sources of uncertainties and their treatment; and (3) to use a specific case study to illustrate how extended source modeling can help to constrain the key assumptions or parameters that impact upon seismic hazard assessment. This article discusses in particular seismic source characterization, strong ground motion prediction, and maximal magnitude constraints, according to the practice of the French Atomic Energy Commission. Due to increases in strong motion databases in terms of the number and quality of the records in their metadata and the uncertainty characterization, several recently published empirical ground motion prediction models are eligible for seismic hazard assessment in France. We show that propagation of epistemic and aleatory uncertainties is feasible in a deterministic approach, as in a probabilistic way. Assessment of seismic hazard in France in the framework of the safety of nuclear facilities should consider these recent advances. In this sense, the opening of discussions with all of the stakeholders in France to update the reference documents (i.e., RFS 2001-01; ASN/2/01 Guide) appears appropriate in the short term.

In regions with sparse ground motion data, simulations provide alternative acceleration time series for evaluation of the dynamic response of a structure. Different ground motion simulation methods provide varying levels of goodness of fit between observed and synthetic data. Before using the seismologically acceptable synthetic records for engineering purposes, it is critical to investigate the efficiency of synthetics in predicting observed seismic responses of structures. For this purpose, in this study we present nonlinear time history analyses of multi-story reinforced concrete frames under observed and synthetic records of a particular earthquake. Synthetic records of 6 April 2009 L′Aquila (Italy) earthquake (Mw=6.3) are simulated using both the Hybrid Integral-Composite method and the Stochastic Finite-Fault method. Results of analyses from observed and the alternative synthetic records of this event are compared in terms of maximum displacement, acceleration and plastic beam rotation of each story level. Our results indicate that the match between the Fourier Amplitude Spectrum of the observed and synthetic records around the frequencies that correspond to the fundamental period of the structure (mainly within 0.2–1.2% of the fundamental period) governs the misfit between the observed and synthetic nonlinear responses. It is also shown that even for cases where nonlinear behavior is more likely, period-dependent SDOF indicators of goodness of fit between a particular observed and corresponding synthetic records represents the difference in MDOF behavior of frame structures due to these records. Finally, simulation of realistic amplitudes over the entire broadband frequency range of interest is found to be critical while using the synthetics for earthquake engineering purposes.

In low-to-moderate seismicity regions such as metropolitan France, characterized by limited strong-motion records in the magnitude-distance range of interest for seismic hazard assessment, the derivation of empirical ground motion prediction equations (GMPEs) is a major challenge. In this study, we take advantage of the RESORCE-2013 database (http://resorce-portal.eu/) that contains uniformly processed records for the Pan-European region including relevant number of French records. After discussing the metadata for French events and stations, we first derive a base-case GMPE that is used to investigate the within-event and between-event residuals. The short-period between-event residuals for French (and Swiss) events show larger variability with respect to larger magnitude events in other regions. We show that the between-event residuals are clearly correlated with the stress parameter and that such larger variability can be explained by accounting for stress-parameter scaling. We derive an empirical scaling of ground motion with stress parameter that is consistent across regions and with the scaling predicted by stochastic GMPEs. This suggests that the scaling of ground motion with stress parameter for a given magnitude is largely region independent whereas the absolute stress parameter values may vary regionally. Based on these results we propose to adopt the scaling model as a function of stress parameter and magnitude by Yenier and Atkinson (Bull Seismol Soc Am 105(4):1989–2009, 2015) by adapting the reference stress parameter to our target regions. By accounting for stress parameter scaling in the GMPE we reduce the between-event variability for French and Swiss small-magnitude events. Finally, we investigate the aleatory variability (σ) of the GMPE and its components (τ, ϕ, ϕss). We propose a heteroscedastic σ model to be used when the stress-parameter scaling is not considered in the GMPEs due to lack of information. If enough information on the stress-parameter is available the adjusted GMPE can be applied using a homoscedastic σ. Despite using small events, the ϕss for French stations is found to be consistent with other studies and confirms the stability of ϕss across different regions and datasets.