Harmonized local magnitude attenuation function for Europe using the European Integrated Data Archive (EIDA)

Abstract

We derive a harmonized local magnitude scale across Europe using data disseminated by network operators through the European Integrated Data Archive (EIDA). We first calibrate simultaneously a set of non-parametric attenuation functions regionalized by considering six different regions covering central and southern Europe, anchoring the models to the Richter's scale at 17 km. Uncertainties on the attenuation coefficients, station corrections and magnitude values are evaluated through bootstrap analysis. The obtained attenuation functions show significant differences among the regions, up to 0.4 m.u. at 400 km, being the attenuation of the Wood-Anderson amplitude stronger for regions in the Mediterranean area. The nonparametric attenuation functions capture the changes in the rate of attenuation with distance due to the effects of later arrivals generated by crustal heterogeneity. A second calibration is performed to derive a parametric attenuation model. We consider a piece-wise linear function to describe the attenuation with the logarithm of distance, introducing two breakpoint distances at 10 and 60 km. For distances above 10 km, we also consider the anelastic attenuation term. We apply a mixed effect regression with network-dependent random effects on the anelastic coefficients. The parametric analysis confirms the stronger attenuation for networks operating in the Mediterranean area, such as the Italian and Greek networks, with respect to networks located in continental Europe. The network-dependent random effects allow us to quantify the between-network variability for different networks operating in the same region or country. The observed between-network variability is within ±0.2 m.u., smaller than the variability among the six regions.

Spectral decomposition of the Engineering Strong Motion (ESM) flat file: regional attenuation, source scaling and Arias stress drop

“…Following previous study, such as the spectral decompositions performed for the Vrancea subduction (Oth et al 2008) or for Japan (Oth et al 2011), we introduced a regionalization of the attenuation, considering three different domains. The results confirm previous findings obtained for Europe by analyzing either response spectra (Kotha et al 2016) or high-pass filtered peak displacement for local magnitude calibration (Bindi et al 2019c): all studies found that the attenuation for the region including Italy is stronger than for the rest of Europe. Moreover, the non-parametric attenuation models capture the modulation of the attenuation with distance and frequency related to the contribution of later arrivals generated by crustal discontinuities, such as the effect of critical reflections from Moho previously highlighted for northern Italy (Bragato and Tento 2005) and northwestern Turkey (Baumbach et al 2003;Bindi et al 2006).…”

“…Polygons defining the three regions considered to derive the attenuation models: IT (cyan), GT (green) and RE (red). The epicenter (white circles) and station (blue triangle) locations are shown as well, considering only those with at least two (left) or five (right) recordings Previous studies highlighted significant variations in attenuation models derived across Europe (e.g., Kuehn and Scherbaum 2016;Kotha et al 2017;Bindi et al 2019c). Therefore, taking into consideration the distribution of data analyzed in this study, we estimate different attenuation spectral models A region (f;R n ) for three regions: the first one (hereinafter referred to as IT) is including mainly Italy; the second region (named GT) encompasses the Aegean and Anatolia regions; the third region (named RE) includes the rest of Europe, but with data mostly located in the Balkans and in Romania (Fig.…”

Data-driven and machine learning identification of seismic reference stations in Europe

“…The thresholds for event selection have been defined for finding the optimal compromise between a large number of events and a high SNR. Bindi et al (2019a) provide further information. In the end, 205 300 records from 12 721 earthquakes at 2271 sites (with co-located instruments at 541 of all 2271 sites) are analysed here.…”

“…Since the magnitude scale has originally been calibrated in California, any application for networks in regions with different attenuation characteristics requires a recalibration of the attenuation function. Since catalogues compiled using different calibration functions cannot easily be compared, Bindi et al (2019a) recently developed a first harmonized local magnitude scale for Europe. The magnitude scale has been calibrated following a non-parametric approach, whilst allowing possible regional variations in attenuation with distance.…”

“…The magnitude scale has been calibrated following a non-parametric approach, whilst allowing possible regional variations in attenuation with distance. In selecting the most suitable reference site condition, Bindi et al (2019a) calibrated their model by constraining the average over all station corrections in the Italian national seismic network (IV network) to zero.…”