The 1997–1998 Umbria-Marche sequence (central Italy): Source, path, and site effects estimated from strong motion data recorded in the epicentral area

The 1997–1998 Umbria-Marche sequence (central Italy): Source, path, and site effects estimated from strong motion data recorded in the epicentral area

Please download to get full document.

View again

of 17
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Publish on:

Views: 0 | Pages: 17

Extension: PDF | Download: 0

  The 1997--1998 Umbria-Marche sequence (central Italy):Source, path, and site effects estimated from strongmotion data recorded in the epicentral area D. Bindi Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Milano, Milan, Italy R. R. Castro Departamento de Sismologı´a, Divisio´n Ciencias de la Tierra, Centro de Investigacio´n Cientı´fica y Educacio´n Superior deEnsenada (CICESE), Ensenada, Mexico G. Franceschina, L. Luzi, and F. Pacor  Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Milano, Milan, ItalyReceived 20 October 2003; revised 20 February 2004; accepted 4 March 2004; published 28 April 2004. [ 1 ]  We apply a nonparametric spectral inversion scheme to calculate source spectra, S   wave attenuation, and site transfer functions from strong motion records in the epicentralregion of the 1997–1998 Umbria-Marche seismic sequence (central Italy). We userecords from moderate size earthquakes (4.6    M  l    5.9) to parameterize the spectralamplitude decay in the distance range from 5 to 40 km. We find that the average qualityfactor   Q  can be approximated by  Q (  f    ) = 49  f    0.9 , between 0.5 and 8 Hz, and thegeometrical spreading by r   0.9 . At high frequencies (  f    > 8 Hz) the dependence of   Q  onfrequency weakens, and it takes an approximate constant value of 318. We fit the sourcespectra to the  w -square model and calculate an average stress drop of (2 ± 1) MPa.The average value is consistent with the previous estimates from the weak events (1.4 <  M  l  < 4.5) of the Umbria-Marche seismic sequence. The most remarkable site effects arefound in correspondence of large sedimentary basins, filled by alternation of sandy-clayeydeposits. The estimated spectral parameters are used to simulate acceleration spectrarecorded during several earthquakes of the Umbria-Marche sequence. Both point sourceand finite fault effects are considered. Furthermore, attenuation relationships for peak ground velocity and ground acceleration are estimated using synthetic data, and comparedto existing relationships.  I   NDEX   T   ERMS  :  7203 Seismology: Body wave propagation; 7212Seismology: Earthquake ground motions and engineering; 7215 Seismology: Earthquake parameters;  K   EYWORDS  :  generalized inversion, strong motion, seismic attenuation, source parameters, Umbria-Marche Citation:  Bindi, D., R. R. Castro, G. Franceschina, L. Luzi, and F. Pacor (2004), The 1997–1998 Umbria-Marche sequence (centralItaly): Source, path, and site effects estimated from strong motion data recorded in the epicentral area,  J. Geophys. Res. ,  109 , B04312,doi:10.1029/2003JB002857. 1. Introduction [ 2 ] One of the main tasks of strong motion seismology isto provide reliable estimates of the seismic ground shakingto be used for seismic hazard assessment and earthquakeresistant design.[ 3 ] The advances in resolution and dynamic range of strong motion instruments achieved in the last twenty yearsmade the strong motion records the most suitable input datafor the design of earthquake resistant structures, allowingaccurate measures of ground motion parameters [ Trifunacand Todorovska , 2001]. The synthesis of strong groundmotion over the entire frequency range of engineeringinterest requires the knowledge of the rupture mechanismcontrolling the seismic energy radiated from the source, theeffect of wave propagation and the response of the near surface soil layers. Our study provides the estimates of source, attenuation, and site spectral parameters usingstrong motion data recorded in the epicentral region of the1997–1998 Umbria-Marche (central Italy) sequence. Thesespectral parameters can be used for predicting the effects of moderate earthquakes (  M     6.0) in this region.[ 4 ] Several authors have dealt with source and attenuation parameters in the Umbria-Marche Area, herein after referredto as UMA [  Rovelli et al. , 1988;  Malagnini et al. , 2000;  Malagnini and Herrmann , 2000;  Castroet al. ,2002;  Bindi et al. ,2001].Unliketheexistingstudies,ourresearchisfocusedonthesimultaneousevaluationofsource,attenuationandsiteterms in the entire frequency range of engineering interest. Castro et al.  [2002] studied spatial and temporal character-isticsof seismic attenuation using  S  and coda waves recorded by the Marchesan Seismograph Network, which is located to JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109, B04312, doi:10.1029/2003JB002857, 2004 Copyright 2004 by the American Geophysical Union.0148-0227/04/2003JB002857$09.00 B04312  1 of 17  the east of the UMA. Further results on attenuation in thecentral Apennines, based on seismic data recorded in a widerange of magnitude (2    M  w    6), have been obtained by  Malagnini et al.  [2000] for frequencies up to 5 Hz. Then,  Malagnini and Herrmann  [2000] extended the validity of the previous results to 16 Hz, fitting a set of weak motions(  M  <5)recordedintheUMAatshortepicentraldistance(R<40 km). These last two relations are usually assumed as thereference for crustal attenuation in the central Apennines andthen we use them to discuss our results.[ 5 ] Source characteristics of the main events of Umbria-Marche earthquake sequence have also been widely inves-tigated. Different authors used stochastic simulationtechniques to compute acceleration time histories [  Berardiet al. , 2000;  Castro et al. , 2001]. They achieved a better fit with observations using a stress parameter in the range of 10to20MPa,whicharesimilartostressdropvaluesobtainedby  Rovelli et al.  [1988] for central and central-southern Apen-nines from strong motion data recorded in the magnituderange 4   M   7.  Bindi et al.  [2001] determined an averagestress drop equal to (3.8 ± 1.0) MPa for the UMA, using S   wave spectra of velocity recordings of about 500 weak earthquakes (  M   4.5) of the 1997 Umbria-Marche seismicsequence. Finally,  Pino et al.  [1999] studied the three mainshocks of the 1997 sequence applying the empirical Greenfunction method and estimated a stress drop of about 3 MPa.[ 6 ] This work represents the first attempt in the area toevaluate source, attenuation and site spectral parametersusing a large strong motion data set, not available before the1997–1998 Umbria-Marche seismic sequence. The afore-mentioned studies are in fact based either on weak motiondata or on analog strong motion records from different tectonic regions. A new calibration of spectral and attenu-ation parameters specific for UMA can allow a better definition of the seismic hazard in this region, which isone of the most risky in Italy.[ 7 ] Our study follows  Castro et al.  [2004], who applied a parametric inversion scheme to the UMA strong motiondata set, with the aim of evaluating the site response of 40 accelerometric stations. From the database analyzed by Castro et al.  [2004], we extract about 150 strong motionrecords from 25 stations operating in the area during the14 strongest events (4.6    M     5.9) of the 1997–1998sequence. Records from other two past earthquakes that occurred in the same area (  M   5.9 and  M   5.2) complete thedata set. We apply a generalized inversion technique toobtain source, attenuation and site functions [ Castro et al. ,1990]. The validity of the site transfer functions is verified by comparison with horizontal to vertical spectral ratios.The reliability of source spectra, anelastic attenuation term,and site transfer functions is then tested by simulating thestrong ground motions recorded during several events,using point source [  Boore , 1983, 2003] and finite fault models [  Mendez and Pacor  , 1994]. Finally, a comparisonwith the  Sabetta and Pugliese  [1987, 1996] attenuationrelationship is performed, simulating PGA and PGV valuesat fixed magnitudes and distances for soft site conditions. 2. Seismic Sequence and Geological Setting [ 8 ] The seismic sequence started on 4 September 1997with a  M  w  4.5 earthquake located at the boundary of theUmbria and Marche regions, close to the Colfiorito village.Figure 1 shows the distribution of the epicenters of thestrongest events of the sequence and Table 1 lists theepicentral parameters of the earthquakes analyzed.[ 9 ] On 26 September, at 0033 UT, a  M  1  5.6 earthquakeoccurred (event 290), followed soon after by a stronger event of   M  1  5.9 at 0940 UT that represents the main shock of theseismic sequence, and is located north of the previous one(event 286). During the following weeks the seismic activitywas very intense, with more than 2000 shocks and about 20 earthquakes exceeding magnitude 4. On 14 October, at 1523 UT, an earthquake of   M  1  5.6 (event 292) occurred tothe south of the main shock. After a decrease in the seismicactivity, other significant shocks (  M  1  > 4.5) struck the samearea in April–May 1998. Among these earthquakes, thestrongest one (  M  1  5.6, event 363) was located at theconsiderable depth of about 47 km [  Parolai et al. , 2001b].[ 10 ] The events can be associated with a NW-SE elon-gated fault zone with length of about 40 km [  Amato et al. ,1998]. The three largest shocks ruptured different segmentsof the fault zone (Figure 1) at shallow depth (5–7 km).These faults are characterized by moderate dip (average40   –60  ) and detach on a low-angle eastward deepeningnormal master fault, the Altotiberina fault, whose geometryhas been identified by the analysis of deep crust and seismicreflection profiles [  Boncio and Lavecchia , 2000].[ 11 ] Moment tensors show normal faulting mechanismswith NW-SE tension axes [  Ekstro¨m et al. , 1998], in agree- Figure 1.  The Umbria-Marche Area (UMA): strongmotion stations and location of the events of the Umbria-Marche sequence (triangles, analog station; squares, digitalstations; stars, epicenters of the three main shocks;numbers, earthquake codes as in Table 1). The  M   > 5earthquake focal mechanisms are plotted [  Morelli et al. ,2000]; the fault geometries (shaded rectangles) are takenfrom  Capuano et al.  [2000]; for event 290, L = 6 km; W =6 km; for event 286, L = 12 km; W = 7.5 km; for event 292, L = 7 km; W = 5 km. B04312  BINDI ET AL.: SPECTRAL PARAMETERS FROM STRONG MOTIONS2 of 17 B04312  ment with the aftershock depth distribution that defines aSW dipping plane.[ 12 ] The faulting mechanisms of the sequence fit wellwith the known seismotectonic regime of the central Apen-nines, a region characterized by frequent moderate magni-tude earthquakes generated by normal faulting.[ 13 ] According to the historical earthquake catalogue(CPTI Working Group, Catalogo Parametrico dei TerremotiItaliani, available at severalevents with I max  IX occurred in the recent centuries (1747,I max  = IX; 1751, I max  = X; 1703, I max  = X; 1730, I max  = IX;1832 I max  = VIII–IX; 1799 I max  = IX–X; 1873 I max  = IX).This sequence with I max  = IX–X is one of the strongest that occurred in this region.[ 14 ] From the geological point of view the area corre-sponds to the Umbria-Marche sedimentary sequence that consists of a multi-layered alternation of limestones, marlylimestones, marls, and flysch alternation [ Centamore et al. ,1979], which were first deformed by a compressive struc-tural phase, and subsequently dissected by Quaternarynormal faulting [ Calamita et al. , 1994]. The Quaternarynormal faults led to the formation of intramountain basins,filled with lacustrine deposits that can reach or exceed athickness of 200 m. Many of the analyzed accelerometricstations were installed on these sedimentary structures, peculiar to the central Apennines, (Figure 1, stations CLF,GBP, NRC). 3. Strong Motion Data Set [ 15 ] We use strong motion records from 25 stations(Table 2) located in and near the epicentral area of the1997–1998 Umbria-Marche seismic sequence.[ 16 ] The accelerometric sites are grouped into four classes(Table 2), accounting for both geotechnical features andgeomorphologic setting: A, lacustrine and alluvial depositswith thickness > 30 m; B, lacustrine and alluvial depositswith thickness < 30 m; C, shallow debris or colluvialdeposits (3–6 m) overlying rock; and D, rock.[ 17 ] Figure 1 shows the location of the accelerometricstations together with the distribution of the epicenters. Intotal we analyzed 149 horizontal component records from16 earthquakes with local magnitudes ranging between 4.6and 5.9 (Table 1) and hypocentral distances less than 70 km,although most of the events were recorded in the distancerange between 5 and 40 km, as shown in Figure 2.[ 18 ] Accelerometric recordings were mainly obtainedfrom the National Strong Motion Network (RAN) managed by National Seismic Survey (SSN) and SOGIN (formerlyItalian Electrical Company, ENEL), and from 10 mobilestations installed by SSN after the 26 September 1997earthquake (SSN-Monitoring System Group, 2002). The Table 1.  Earthquake Parameters a  EarthquakeDate( Latitude, deg Longitude, deg  M  1  H, km  M  o , N m D s ,MPa D s rms ,MPa115 19.09.79 21:35 42.730 12.960 5.9 6.0 7.0E + 17 1.94 2.20174 29.04.84 05:02 43.250 12.520 5.2 7.0 3.4E + 17 1.03 1.50290 26.09.97 00:33 43.021 12.888 5.6 3.8 4.0E + 17 2.28 2.42286 26.09.97 09:40 43.023 12.847 5.9 6.5 1.2E + 18 2.85 3.20350 03.10.97 08:55 43.034 12.842 5.4 5.7 8.6E + 16 2.58 2.83291 06.10.97 23:24 43.015 12.843 5.5 7.0 1.7E + 17 4.26 4.84352 07.10.97 01:24 43.010 12.783 4.6 10 2.3E + 15 2.77 2.46353 07.10.97 05:09 43.010 12.848 4.8 10 6.7E + 15 3.19 3.11355 12.10.97 11:08 42.911 12.950 5.3 2.2 7.8E + 16 0.90 1.01292 14.10.97 15:23 42.915 12.930 5.6 4.9 3.4E + 17 1.01 1.15358 16.10.97 12:00 43.034 12.890 4.6 2.0 3.9E + 15 1.29 1.31360 09.11.97 19:07 42.854 12.999 4.9 2.0 – – – 362 21.03.98 16:45 42.951 12.914 4.6 4.1 4.0E + 16 0.76 0.47363 26.03.98 16:261 43.191 12.886 5.6 47.0 1.2E + 17 1.15 2.12364 03.04.98 07:26 43.184 12.759 5.0 2.6 5.7E + 16 1.81 1.19365 05.04.98 15:52 43.190 12.773 4.7 5.4 1.9E + 16 1.36 1.42 a  Hypocentral coordinates, magnitude, and seismic moment are taken from the European strong motion database (N. Ambraseys et al., Internet-Site for European Strong-Motion Data, European Commission, Research-Directorate General, Environment and Climate Program, available at http://, 2002) and from  Cattaneo et al  . [2000];  Morelli et al  . [2000];  Zonno and Montaldo  [2000]. The stress drop  D s  values areestimated in this work. Table 2.  Characteristics of the Accelerometric Stations a  Code Name  I   Site  k  o , sANNI Annifo d C — ASSI Assisi d D 0.04BVG Bevagna a A 0.05CAG Cassignano a D — CESM Cesi Monte d C 0.03CESV Cesi Valle d A 0.06CLC Colfiorito-Casermette d A 0.04CLF Colfiorito a A — CSA Castelnuovo-Assisi a A 0.07CSC Cascia a D — CTR Borgo-Cerreto Torre d D 0.02FHC Forca Canapine a D 0.05GBB Gubbio a C — GBP Gubbio-Piana d A 0.06MNF Monte Fiegni a D — MTL Matelica a B 0.03 NCB Nocera Umbra-Biscontini d C 0.02 NCM Nocera Umbra-Salmata a B —  NCR2 Nocera Umbra 2 d C 0.02 NCR Nocera a C —  NOCE Nocera Umbra P.I. d B 0.02 NRC Norcia a A NRI Norcia-Zona Industriale a A 0.03SELE Sellano E d D 0.02SELW Sellano O d D 0.03 a   I   corresponds to instrument type (a is analog; d is digital). Sitecorresponds to: A, lacustrine and alluvial deposits with thickness >30 m; B,lacustrine and alluvial deposits with thickness <30 m; C, shallow debris or colluvial deposits (3–6 m) overlying rock; D, rock. Here  k  o  corresponds tothe high-frequency attenuation parameter, obtained from the inversion. B04312  BINDI ET AL.: SPECTRAL PARAMETERS FROM STRONG MOTIONS3 of 17 B04312  strong motion Records of Umbria-Marche Sequence, Sep-tember 1997–June 1998. CD-ROM). Additional recordscome from the permanent stations managed by ENEA(National Energy Agency). The permanent RAN network is mainly equipped with analog Kinemetrics SMA1, withnatural frequency of about 18 or 25 Hz and 0.5 or 1 g fullscale. The set of mobile instruments consisted of 2 analogrecorders of the same kind and 12 digital Kinemetrics Etnaor K2. All digital instruments have a natural frequency of 50 Hz and were set at 1 g full scale.[ 19 ] The accelerograms were baseline corrected and theeffect of instrument response removed. All the recordsanalyzed have a sampling rate of 200 samples per second.The analog records were filtered by visually selecting afrequency filtering interval; the average range used was0.5–25 Hz.[ 20 ] We calculated the Fourier amplitude spectra of theaccelerograms selecting time windows that start before the S   wave arrival and end when 80% of the total energy isreached. Analog data triggered by  S   wave pulse wereanalyzed selecting a time window starting from the first available sample.[ 21 ] The acceleration spectra were smoothed using avariable frequency band of ±25% of the central frequencyand vectorial addition of the two horizontal components was performed (  ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  EW  2 þ  NS  2 p   ).[ 22 ] For the few stations installed in intramountain basinsthe 80% of total energy criterion does not avoid thecontamination by surface-wave arrivals. In Figure 3 themain shock (event 286) accelerometric time historiesrecorded at two stations are shown. GBP belongs to siteclass A (lacustrine and alluvial deposits), whereas CTR toclass D (rock). The presence of surface waves determinesthe duration of the shaking at GBP and the windowsselected by applying the 80% of total energy criterioninclude also surface waves. The same effect is also observedat station CSA. 4. Nonparametric Inversion [ 23 ] We apply a nonparametric spectral inversion scheme[ Castro et al. , 1990] to estimate the source spectra, thespectral attenuation and the site transfer function in UMA.The dependence of the spectral amplitude  D (  f   ,  r  ) ondistance can be written as log  D f    ; r  ij    ¼ log  A f    ; r  ij    þ log ~ S  i  f   ð Þ ;  ð 1 Þ where  A (  f   ,  r  ij  ) describes the spectral attenuation along the path connecting the source  i -th to station  j- th, and  e S  i (  f   ) is ascalar that depends upon the size of the  i -th source. Thelinear system represented by equation (1) can be numeri-cally solved to determine the spectral attenuation withdistance at any given frequency  f   . A subsequent inversionallows the separation of the site amplification from thesource spectra [e.g.,  Andrews , 1986] by making log  R ij   f   ð Þ¼ log  Z   j   f   ð Þþ log S  i  f   ð Þ ;  ð 2 Þ where the residual  R ij  (  f   ) is obtained correcting the spectralamplitude  D (  f   ,  r  ij  ) for the effect of attenuation  A (  f   ,  r  ij  ).  Z    j  (  f   ) and  S  i (  f   ) represent the site and the source spectra,respectively.[ 24 ] To eliminate the linear dependence between sourceand site in equation (2), the transfer function relevant to a Figure 2.  Magnitude versus hypocentral distance for theselected recordings. Figure 3.  Accelerograms for event 286 at GBP (class A)and CTR (class D). Shaded areas represent the windowanalyzed in the present study. In each panel, from top to bottom the north-south, east-west and vertical componentsare shown. The hypocentral distance is also reported. B04312  BINDI ET AL.: SPECTRAL PARAMETERS FROM STRONG MOTIONS4 of 17 B04312  rock site (or the average of a set of rock sites) is usuallyconstrained to 1. Since amplification effects were detectedfor rock sites in UMA, due to topography or to the presence of weathered layers [ Castro et al. , 2004], we prefer to set the reference station function  Z  (  f   ) to an a priori known function of frequency. ASSI station isselected as reference site because many recordings areavailable in a wide magnitude and azimuth range. Thetransfer function of ASSI is estimated by computing thehorizontal to vertical spectral ratios (H/V). Figure 4 showsthe average ±1 standard deviation obtained considering50  S   wave windows from events having magnitudes in therange from 3 to 5.6. In the inversion process, ASSI isconstrained to have the transfer function estimated with theaverage H/V.[ 25 ] We perform independent inversions for 19 different frequencies between 0.4 and 18 Hz. The distance rangefrom 5 km to 61 km was discretized into 28 bins 2 kmwide. Systems (1) and (2) are solved in a least squaressense by applying the LSQR method [  Paige and Saunders ,1982].[ 26 ] Since distances below 5 km are not sampled, theattenuation function is constrained to be 1 at 5 km, in order to avoid one unresolved degree of freedom affecting thesystem (1). We also require the second derivative of   A (  f   ,  r  )to be small. 4.1. Spectral Attenuation [ 27 ] Figure 5a shows the spectral attenuation  A (  f   ,  r  ij  ) for asample of frequencies. The curves decrease with distance upto 40 km, then a change in slope occurs and the curves become flat. This feature might be due to the combinedeffect of reflected arrivals, scattered waves and site ampli-fication. To isolate the contribution due to geometricalspreading from the anelastic attenuation, we fit the follow- Figure 4.  Mean ±1 standard deviation of H/V ratiocomputed using 50 strong motions recorded at ASSI. Figure 5.  (a) Attenuation-distance curves obtained from the nonparametric inversion, computed at different frequencies; the geometrical spreading,  r   n , for different   n  is also plotted (dashed lines).(b) Geometrical spreading coefficient   n  and (c) quality factor   Q  versus frequency obtained fromequation (4); the quality factor is described by  Q (  f   ) = 49  f    0.9 for 0.4 <  f    8 Hz, and  Q (  f    ) = 318 for 8 <  f      18 Hz (solid line). B04312  BINDI ET AL.: SPECTRAL PARAMETERS FROM STRONG MOTIONS5 of 17 B04312
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks