A Ground-Motion Prediction Model for 1 Shallow Crustal Earthquakes in Greece 1 2

6 Using a recently completed database of uniformly processed strong-motion 7 data recorded in Greece, we derive a ground-motion prediction model (GMM) for 8 horizontal-component peak ground velocity, peak ground acceleration, and 5%9 damped pseudo-acceleration response spectra at 105 periods ranging from 0.01 s 10 to 10 s. The equations were developed by modifying a global GMM to account 11 for more rapid attenuation and weaker magnitude scaling in the Greek ground 12 motions than in the global GMM. Our GMM is calibrated using the Greek data 13 for distances up to 300 km, magnitudes from 4.0 to 7.0, and time-averaged 30 m 14 shear-wave velocities from 150 to 1200 m/s. The GMM has important attributes 15 for hazard applications including magnitude scaling that extends the range of 16 applicability to M 8.0 and nonlinear site response. These features are possible 17 because they are well constrained by data in the global GMM from which our 18 model is derived. An interesting feature of the Greek data, also observed 19 previously in studies of mid-magnitude events (6.1-6.5), is that they are 20 substantially overpredicted by the global GMM, which may be a repeatable 21 regional feature, but may also be influenced by soil-structure interaction. This 22 bias is an important source of epistemic uncertainty that should be considered in 23 hazard analysis. 24 1 Peer-Review DISCLAIMER: This draft manuscript is distributed solely for purpose of scientific peer review. Its content is merely being considered for publication, so it is not to be disclosed or released by reviewers. Until the manuscript has been approved for publication by the U.S. Geological Survey (USGS), it does not represent any official USGS finding or policy. 2 C:\itsak_gmpe_project\paper2_analysis\Ground-Motion Prediction Model for Greece.v04.1.docx INTRODUCTION 25 As described in Margaris et al. (202x)—a companion paper to this article—the network 26 of digital strong-motion instruments in Greece has increased substantially since 2000, and 27 these instruments have provided a large number of well-recorded data. Moreover, the source, 28 path, and site metadata for new and older recordings has been substantially improved as part 29 of a long-term effort to raise the level of data quality and modeling in Greece to levels 30 typically applied in Next-Generation Attenuation projects, such as NGA-West2 for active 31 tectonic regions (Bozorgnia et al. 2014). 32 In this article, we use these data to derive a ground-motion prediction model (GMM) for 33 shallow (focal depth ≤ 30km) crustal earthquakes in Greece. Our objective is for the model to 34 be suitable for the conditions that control hazard at the long return periods used in 35 engineering design; disaggregation indicates these conditions to typically involve events with 36 magnitudes in the range of 6.5 to 7 and distances less than 20 km. Such conditions are 37 challenging for model development for two main reasons: (1) the upper portion of the 38 magnitude range is near the limit of empirical data sets for Greece (Margaris et al. 202x), and 39 indeed Europe (Akkar et al. 2014a; Bindi et al. 2019); and (2) the strong shaking that occurs 40 under these conditions produces nonlinear site response for soil sites, which may be difficult 41 to evaluate directly from data due to limited observations. As a result of these and other 42 factors, the models must be used for ranges of conditions that may be poorly represented in 43 empirical datasets. 44 To overcome these difficulties, our model development process modifies the Boore et al. 45 (2014) global GMM for intensity measures from earthquakes in active crustal regions; this 46 model is reasonably well constrained for the aforementioned hazard-controlling conditions, 47 for application in Greece. The modifications are targeted towards model attributes that are 48 known to be regionally variable, including the constant term, anelastic attenuation, and site 49 response. Similar approaches have been used previously for Italy, New Zealand, and Turkey 50 (Scasserra et al. 2009; Bradley 2013; Gülerce et al. 2016). This approach differs from those 51 used to develop currently available models, including models intended specifically for 52 Greece (Danciu and Tselentis 2007; Chousianitis et al, 2018) and others intended for Europe 53 (Akkar et al. 2014b; Bindi et al. 2014; Kotha et al. 2016; Kuehn and Scherbaum 2016; Kotha 54 et al. 2020). Those Greece/Europe models are derived from datasets restricted to those 55 3 C:\itsak_gmpe_project\paper2_analysis\Ground-Motion Prediction Model for Greece.v04.1.docx regions. Recalling the above-referenced challenges, the maximum magnitude in the datasets 56 is 7.1 and all of the models utilize linear site terms. 57 Because the model developed here is Greece-specific, we briefly review prior Greece58 specific models. The most recent previous equations for pseudo-spectral acceleration (PSA) 59 (Danciu and Tselentis, 2007) only used analog data recorded before 2000; the equations were 60 restricted to distances less than 136 km, the site conditions were parameterized by a few site 61 classes, and the equations were given for periods up to 4.0 s. A more recent study 62 (Chousianitis et al, 2018) used some data after 1999, but equations were not provided for 63 PSA. They provide equations for PGA and PGV, for distances up to 200 km and a site 64 parameterization similar to Danciu and Tselentis (2007). Both Danciu and Tselentis (2007) 65 and Chousianitis et al. (2018) use a linear magnitude dependence in their equations, whereas 66 most recent GMPEs find a nonlinear dependence, including those in this article, with the 67 scaling of ground motion being stronger for small magnitudes than large magnitudes for a 68 fixed distance. 69 The GMM proposed here for Greece is developed for the following horizontal70 component ground-motion intensity measures (GMIMs): pseudo-acceleration 5%-damped 71 response spectra (PSA) at 105 periods ranging from 0.01 s to 10 s, peak acceleration (PGA), 72 and peak velocity (PGV). The range of applicable moment magnitudes directly constrained 73 by data is 4.0 to 7.0, but given the “borrowing” of magnitude scaling from a global model, 74 the GMM can be applied (with additional uncertainty) up to M 8.0 events. The distance range 75 is 0 to 300 km, and the range of applicable site conditions (based on time-averaged 30 m 76 shear wave velocity, 30 S V ) is 150 to 1200 m/s. 77 Following this introduction, we first discuss the data used in the analysis. We then 78 provide the set of equations that define our GMM, and the derivation of the coefficients in 79 those equations is given next (the coefficients are provided in an electronic supplement). This 80 is followed by some comparisons of GMIMs from our GMM with those from Boore et al. 81 (2014) (hereafter BSSA14), and previously published GMMs that are specifically for Greece. 82 The article concludes with a Summary and Discussion section. 83