Preliminary engineering analysis of the August 24 th 2016 , M L 6 . 0 central Italy earthquake records

An earthquake of estimated local magnitude (ML) 6.0 struck central Italy on the 24th of August (01:36:32 UTC) in the vicinity of Accumoli (close to Rieti, central Italy) initiating a long-lasting seismic sequence that also featured events of larger magnitude within a few months. The earthquake caused widespread building damage and around three-hundred fatalities. Ground motion was recorded by hundreds of seismic stations. This work uses accelerometric records for a preliminary discussion, from the earthquake engineering perspective, of strong motion caused by the earthquake. Peak and integral ground motion intensity measures, are presented. The response spectra at some select stations are analysed with respect to the code-mandated design actions for various return periods at the recording sites. Hazard disaggregation for different return periods is discussed referring to the site of the epicentre of the earthquake. Finally, some preliminary considerations are made concerning the impact of rupture propagation on near-source ground motion; i.e., the records are scanned for traces of pulse-like forward-directivity effects.


I. INTRODUCTION
The national accelerometric network of Italy (RAN), operated by the governmental Dipartimento delle Protezione Civile (DPC), and the Italian seismic network (RSN), operated by the Istituto Nazionale di Geofisica e Vulcanologia (INGV) have made available, rapidly after the event, the records of the earthquake, with epicentre located in the vicinity of Accumoli, central Italy, that struck on Aug. 24 2016, at 1:36:32 AM -UTC.The moment magnitude (Mw) declared by INGV is 6.0, while other international institutions claim Mw 6.2.Corrected records and processing details are available on the Engineering Strong-Motion database website, while the uncorrected waveforms can be found on the RAN and RSN websites (see section VI).
The present short article deals with some aspects of recorded strong ground motion of earthquake engineering interest.First, shaking intensity parameters for some of the ground motions recorded nearest to the fault are presented.Then, response spectra for some of the stations closer to the source are compared to the code-mandated spectra.Finally, nearsource ground motions are examined for impulsive characteristics to be possibly attributed to rupture directivity.

II. GROUND MOTION INTENSITY MEASURES AND RESPONSE SPECTRA
Table 1 shows some peak and integral ground motion intensity measures (IMs) for the records within 30km from the source (i.e., Joyner-and-Boore distance, jb R ).More specifically, data reported in the table are: the ID of the station, the  To have a more complete picture of the characteristics of the recorded IMs, these have to be compared with ground motion prediction equations (GMPEs).One of such comparison can be found in ReLUIS-INGV Workgroup (2016a), that also includes the complete response spectra (elastic and inelastic) for the 2 IMs from AMT station are derived from the revised records provided by RAN website several weeks after the event.Moreover, at the time of the submission of this paper, recordings from AQA and NOR stations have been retracted pending possible revision by the DPC.AQA and NOR data analysed here are those available prior to said revision.
(0.3) Sa  records considered here, as well as for others more distant from the source.Another comprehensive comparison of observed ground motion peak parameters and predicted IMs is provided in Lanzano et al. (2016) with respect to two different GMPEs.In that paper, it is underlined that the GMPEs generally fit the observations for low spectral periods and short source-to-site distances and seem to underpredict the observed IMs at distances larger than 80 km.

III. THE RECORDS AND THE ITALIAN SEISMIC CODE
In Figure 1 the pseudo-acceleration response spectra associated to the horizontal ground motions, recorded by some of the stations with the smallest jb R distance, are compared with the elastic design spectra provided by the Italian building code, at the corresponding sites, for four different return periods (TR): 50, 475, 975 and 2475 years.Before proceeding any further, it is worthwhile recalling that NTC2008 spectra are a direct approximation of the uniform hazard spectra computed via the probabilistic seismic hazard analysis (PSHA) discussed in Stucchi et al. (2011).
The east-west component of AMT exceeds the 2475 years spectra in the 0s-0.4srange of periods, while at least one component of the same station exceeds the 475 years spectra for spectral periods up to 2.1s.NRC and FEMA (see Figure 2a, for the position with respect to the source) exceed the 475 years spectra in the range of periods 0s-0.3s and 0.35s-0.5srespectively.This applies to at least one of the two horizontal components (for NRC record, the exceedance happens also for 0.67s-0.88s).The NRC record also exceeds the 2475 years spectrum for periods between 0.13s and 0.28s.However, at all the stations for longer oscillation periods, and as soon as the distance increases, spectral ordinates become comparable with code-spectra corresponding to return periods of a few tens of years.
The maximum ratios of the peak of the pseudo-acceleration spectrum (5% damping) and the PGA are equal to 5.1 and 2.7, for NRC and AMT stations, respectively (these refer to the east-west and north-south directions without investigating other possible rotations).Notwithstanding the not completely intelligible (so far at least) differences between the two horizontal components of AMT, the shape and the amplitude of these spectra appear compatible with extensive damage in some villages, where the population of structures suffered significant damage or total collapse.
It should be also discussed that exceedance of code spectra close to the source of a strong earthquake does not directly imply inadequacy of PSHA at the basis of the code spectra (Iervolino, 2013).This is also because spectra from PSHA, are the results of an average of a series of scenarios considered possible (e.g., small and large source-to-site distances).Such an average may be exceeded close to the source of an earthquake, even if the corresponding scenario is included in the PSHA.

Sa
, for two return periods (475 and 2475 years).The entire set of disaggregation distributions is not shown here for the sake of brevity (although it can be found in ReLUIS-INGV Working group, 2016a).Nevertheless, the relevant scenarios for the considered return periods and IMs are summarized here; they are the magnitude and distance intervals that have probability, of being causative for the exceedance of the corresponding IM, larger than 0.5.All of these scenarios are characterized by the same distance range between 0km -20km.Magnitude intervals for PGA are 5.3 -6.3 and 5.9 -6.8 for 475 years and 2475 years, respectively.For

 
1.0 Sa , and the same return periods, magnitude intervals are 6.0-7.0 and 6.3-7.0,respecitively.It may be concluded that, according to the hazard analysis the code is based on, exceedance of highfrequency spectral accelerations, corresponding to 475 years and 2475 years TR, is most likely caused by a close moderate-magnitude earthquake that is loosely compatible to what was observed.
Figure 1.Comparison between the observed ground motions and the elastic design spectra provided by NTC2008.

IV. ANALYSIS OF PULSE-LIKE DIRECTIVITY EFFECTS
Pulse-like near-source (NS) ground motions may be the result of rupture directivity.This can lead to a constructive wave interference effect, is manifested in the form of a double-sided velocity pulse that delivers most of the seismic energy early in the record (Somerville et al., 1997).Clues of impulsive features in near-source ground motions have been probably found in Italian seismic events of normal faulting style before (e.g., Chioccarelli and Iervolino, 2010).In this preliminary investigation for such pulse-like effects, the continuous wavelet transform algorithm suggested by Baker (2007) was implemented for all recordings (horizontal components) within a closest-to-rupture distance of 30km from the fault and for all orientations.
It should be noted that the adopted approach is purely phenomenological, extracting empirical evidence of impulsive characteristics directly from the recorded NS signals without attempting to assign a causal relation to specific effects falling under the banner of rupture directivity (i.e., rupture propagation, seismic source radiation pattern, motion polarization).Models regarding the phenomenon through the prism of the physics of finite-fault rupture are also available (e.g., Spudich and Chiou, 2008) but not followed in this preliminary analysis.
The surface projection of the fault rupture plane (Tinti et al., 2016) is shown in Figure 2a along with the equal-probability contours of the Iervolino and Cornell (2008) model for the probability of observing NS directivity pulses.It can be observed that, interestingly enough, some of the most prominent impulsive waveforms (NRC, NOR, FEMA, RM33) have been recorded at sites where the empiricallycalibrated model assigns low probability of pulse occurrence due to directivity.This is indicative of the fact that more research is required into the phenomenon for the case of normal faulting, but it should also be mentioned that there were hardly any accelerometric stations present in the area where pulse-like effects were most probable to be observed according to the model.
Out of all the records investigated, six ground motions exhibited impulsive characteristics, as expressed by a Pulse Indicator (PI) score in excess of 0.85 (see Baker, 2007).The AMT record revealed two distinct pulses.One with pulse period p T of 0.41s being predomi- nant in the fault-normal (FN) and another, longer duration pulse with p T = 0.98s in the fault-parallel (FP) direction (Figure 2b).The latter might be attributed to the breakage of a nearby asperity on the fault plane (Tinti et al., 2016).The FEMA record was also discovered to exhibit impulsive characteristics in both FN ( p T = 0.50s) and FP directions.The ground mo- tions recorded at the MNF station and RM33 were found to contain pulses in the FN direction ( T = 1.40s and 1.20s respectively) that also hinted at rupture directivity effects, despite the lower velocity amplitude due to the greater distance from the fault and consequent attenuation.This leaves the two ground motions recorded at Norcia, NRC and NOR, that were of particular interest.The NRC record was found to contain a 2.09s period pulse mostly towards orientations that lie between the FN and FP but somewhat more prevalent in the direction perpendicular to the strike.Interestingly, it is known that the NRC site sits upon deep soil deposits characterized by an inversion of velocity profile at a depth of more than 30m (see Bindi et al., 2011) and this cannot be disregarded when narrow-band characteristics are observed in the recordings.However, station NOR was also found to be pulse-like, with a 1.63s pulse in the FN direction (see Figure 2c).Furthermore, records obtained at the base/ground-level of instrumented, seismically monitored buildings distant up to 500m from the two accelerometric stations, contained velocity pulses almost identical in orientation and period to the NOR station.This consistency enhances the argument in favour of the presence of a directivity effect.
One prominent characteristic of directivity-induced velocity pulses is that they tend to dominate the spectral shape of the signal's pseudo-velocity spectrum.This is also confirmed in the case examined here; see Figures 2(d-e).Furthermore, pulse period p T (as de- fined via the wavelet transform, see Baker, 2007) is very well correlated with predominant ground motion period g T (vibration period where maximum pseudo-velocity occurs); also confirmed in this case.Note that the extracted pulses typically correspond to some local maximum also on the pseudo-acceleration spectrum ( Sa ), but do not necessarily account for the absolute maximum amplitude, which is usually determined by the higher-frequency content.
Finally, when the pulse periods extracted from the various NS sites examined are compared against the empirical regression model from Baltzopoulos et al. (2016), they are found in reasonable agreement with what is expected for a Mw 6.0 event ( p T geometric mean of 1.02s compared to predicted median of 1.29s).
Figure 2. Equal probability of pulse occurrence contours, according to the model of Iervolino and Cornell (2008), plotted against the surface projection of the rupture plane (a), velocity time-history and extracted pulse of AMT FN component (b) and NOR FN (c), pseudo-velocity spectra (PSV) of the extracted pulses plotted against PSV of the original signal for the FN components of AMT (d) and NOR (e), pseudo-acceleration spectra of the extracted pulses plotted against Sa of the original signal for the FN components of AMT (f) and NOR (g).

V. CONCLUSIONS
This article provided a preliminary engineering-point-of-view analysis of the strong motion records obtained during the Amatrice earthquake that struck central Italy on 24 th of August 2016.An overview of the ground motion intensity parameters typically associated with structural response was given, identifying stations with highest IMs, with AMT exhibiting the highest horizontal PGA recorded in Italy so far.The comparison of some of the closest-to-rupture records' spectra and the code spectra showed cases of exceedance of the latter by the former at TR 475 years and even TR 2475 years.It was then briefly discussed how this is to be expected, even though to claim otherwise can be a common pitfall.Finally, it can be said that some indications of pulse-like NS motions were observed, which could be the result of rupture directivity.Points of interest that merit further investigation include the potential of site effects in the manifestation of some particular spectral shapes and the juxtaposition of inelastic response spectra with damages.

VI. POSTSCRIPT
This paper was prepared before the occurrence of other large earthquakes from same seismic sequence.Indeed, since the first submission of this article, two further strong earthquakes have struck the area, the 26/10/2016 Mw 5.9 Ussita event and the 30/10/2016 Mw 6.5 Norcia event.As such, the 24/08 Mw 6.0-6.2Amatrice earthquake is currently regarded as the initiating event of the long-lasting 2016 central Italy sequence, with the Norcia seismic event being typically nominated as the overall mainshock, so far.Herein the first event was discussed, while the reader is referred to ReLUIS-INGV Workgroup (2016b) for a more comprehensive analysis of the sequence as a whole.

VII. DATA AND SHARING RESOURCES
Records used herein were processed and provided to the authors by the ITACA-ESM Working Group of the Istituto Nazionale di Geofisica e Vulcanologia (INGV).They are also available at http://esm.mi.ingv.it/(last accessed 21/11/2016).
The unprocessed records can be accessed at http://ran.protezionecivile.it for the RAN network and at the European Integrated Data Archive (http://www.orfeus-eu.org/data/eida/) for the RSN.
The parameters of the finite-fault geometry used are available at http://esm.mi.ingv.itand are attributed to Tinti et al. (2016).
Accelerograms recorded at the monitored structures in Norcia are available from the Seismic Observatory of Structures of the National Civil Protection -www.mot1.it.

VIII. ACKNOWLEDGMENTS
The study presented in this paper was developed within the activities of ReLUIS (Rete dei Laboratori Universitari di Ingegneria Sismica).
Table 1.IMs of the recorded ground motions within 30 km from the source (distance in terms of jb R ).

jbR
distance, the soil class 1 according to the code (CS.LL.PP.2008, NTC2008 hereafter), the peak ground acceleration of the east-west horizontal component   E PGA , and of the north-south component   N PGA , the angle with respect to the north (positive clockwise) to which corresponds the maximum recorded