Journal of Seismology 1: 39–45, 1997. 39c 1997 Kluwer Academic Publishers. Printed in Belgium.

Source parameters of the Pinotepa Nacional, Mexico, earthquake of 27March, 1996 (Mw = 5.4) estimated from near-field recordings of a singlestation

S. K. Singh, J. Pacheco, F. Courboulex & D. A. NoveloInstituto de Geofısica, Universidad Nacional Autonoma de Mexico, C.U., 04510 Mexico, D. F., Mexico

Received 5 November 1996; accepted in revised form 20 February 1997

Key words: near-source seismograms, earthquake source parameters, Mexican earthquakes

Abstract

We use near-field accelerograms recorded by the very broadband seismographic station of PNIG to locate thePinotepa Nacional earthquake of 27 March, 1996 (Mw = 5.4) and to determine its source parameters. The data fromPNIG on P and S arrival times, the azimuth of the arrival of P-wave, and the angle of incidence of the P-wave at thefree surface permit the determination of the location (16.365� N, 98.303� W, depth = 18 km) and the origin time(12:34:48.35) of the earthquake.

The displacement seismograms of the earthquake clearly show contribution from the near-field terms. Wecompute a suite of synthetic seismograms for local mechanisms in the vicinity of the mechanism reported by theU.S. Geological Survey (USGS) and compare them with the observed seismograms at PNIG. The point sourcewhose synthetics fit the observed records well has the following parameters: seismic moment, M0 = 1.2 � 1024

dyne-cm; source time function: a triangular pulse of 0.9 sec duration; fault plane: strike� = 291�, dip � = 10�, andrake � = 80�. The location and the source parameters obtained from the analysis of PNIG records differ significantlyfrom those reported by the USGS. This demonstrates again, what has been shown by some previous researchers,that high-quality recordings from a single near-field station can considerably improve the estimation of the sourceparameters of an earthquake.

The main event was preceded by a subevent which occurred �0.18 sec before and whose seismic moment was�1% of the main event. It is possible that even this subevent was preceded by a couple of smaller subevents. Thisearthquake supports the body of evidence showing that an earthquake begins with a sequence of smaller subevents,cascading in the occurrence of the main event.

Introduction

In 1991, Instituto de Geofısica (IGF) of UniversidadNacional Autonoma de Mexico (UNAM) began mod-ernizing its National Seismological Service (SSN) byinstalling a new network of very broadband seismo-graphs. The goals of this modernization are to provideprompt and reliable estimation of source parameters ofearthquakes occurring in Mexico, and to make high-quality data available to scientists for research. Thisnetwork, at present, consists of 14 stations sparse-ly distributed in the country (Figure 1). Each stationconsists of a STS-2 seismometer and a KinemetricsFBA-23 accelerometer connected to a 24-bit Quanter-

ra digitizer. Continuous velocity data, sampled at 1 Hzand 20 Hz, are saved in a buffer memory. For triggeredevents, both velocity and acceleration channels, sam-pled at 80 Hz, are saved. Six of the 14 stations are linkedto IGF through satellite (ZIIG, CAIG, PNIG, PLIG,HUIG and TUIG), and one is linked through a dedicat-ed ’phone line (CJIG). Six stations are still autonomous(MAIG, COIG, MOIG, YAIG, OXIG, LVIG), and areperiodically visited to collect data using a personalcomputer. The CUIG station is located in the UNAMcampus, at the bottom of a 20 m shaft, and is connectedto IGF through ethernet.

The progress towards the original goals has beenslow due to hardware, software and communication

Article: jose21 GSB: 704004 Pips nr 135447 BIO2KAP

*135447 jose21.tex; 25/06/1997; 11:58; v.7; p.1

40

Figure 1. The broadband seismic stations of Instituto de Geofısica (IGF), UNAM, currently in operation. Stations ZIIG, CAIG, PLIG,HUIG,and TUIG are connected to IGF through satellite; CJIG is connected through a dedicated phone line, and CUIG is accessed via ethernet. Allother stations are autonomous.

problems. The network is now producing regional andteleseismic data prolifically, but near-source record-ings of moderate and large earthquakes are infrequentdue to the sparse distribution of the stations. A recentexception is the earthquake of 27 March 1996 whichoccurred close to the coastal town of Pinotepa Nacionaland was recorded at PNIG. In this note, we presentthe source parameters of this earthquake based on thenear-field recording at PNIG. Our study follows that ofKanamori et al. (1990) who, among others, have shownthat a single three-component, near source, very broad-band recording can significantly improve the estimatedsource parameters of an earthquake.

Data

During the earthquake, the velocity traces at PNIGwere clipped soon after the arrival of S-wave (Figure 2,top). The acceleration traces of the event are shownin Figure 2 (bottom). The peak accelerations on Z,

EW, and NS components were 38, 40, and 63 gals,respectively, well below the full range of � 2 g of theaccelerometer. We compared the velocities obtainedby direct integration of the acceleration traces with thecorresponding traces from the velocity channels. Thetwo traces were identical until the saturation of thevelocity channel.

Location of the earthquake

Because of the sparse distribution of seismic stationsoperated by the SSN, the earthquake locations, in gen-eral, are in error by 20 km or so. For this reason weuse the PNIG recording alone to locate this earthquake.The observed (S-P) time and the azimuth (�s) to PNIGare 3.2 sec and 81.5�, respectively. (This refers to anepisode of large energy release, henceforth referred toas the main event. As we show later, the main event waspreceded by one or more subevents).To find a hypocen-ter consistent with the readings at PNIG, we rotated

jose21.tex; 25/06/1997; 11:58; v.7; p.2

41

Figure 2. Top: Velocity traces of 27 March, 1996 earthquake recorded at the near-source very broadband station of PNIG. Note the saturationof the traces soon after the arrival of the S-wave. Bottom: The corresponding acceleration traces.

the NS and EW components into radial and transversecomponents. The angle of incidence at the surface (i0),computed from the radial and theZ components of theP-wave, is 37� from the vertical. Because of decreas-

ing P-wave speed, �, towards the surface, the take-offangle at the source (ih) must be less than 143�. We takea layer (� = 5.0 km/sec) of thickness 1.5 km overlyinga half space (� = 6.2 km/sec) as an appropriate model

jose21.tex; 25/06/1997; 11:58; v.7; p.3

42

for the upper crust below PNIG. The value of � for thehalf space is taken from a refraction study in the region(Valdes et al., 1986). The thickness and the speed ofthe upper layer were chosen, through trial and error,to model the observed seismograms at PNIG (see lat-er discussion). For this crustal model, the epicentraldistance (�), the depth (H), and the takeoff angle (ih)required to match the (S-P) time of 3.2 sec and theangle of incidence at PNIG are 19.7 km, 18.0 km, and131.7�, respectively. Since the coordinates of PNIG are16.392�N and 98.127�W, we obtain the location of theearthquake as 16.365�N and 98.303�W. With respectto this location, the P-wave travel time to PNIG is4.38 sec. Subtracting this time from the P-wave arrivaltime at PNIG gives the origin time of the major ener-gy release during the earthquake as 12:34:48.35. Theerrors in the crustal model and in measuring angles�sand i0 are mapped in the location error of the earth-quake. The waveform modelling in next section sug-gests the horizontal and vertical errors in the location,using only the PNIG data, are probably� 2 km.

The moment tensor solution of the event wasreported by the U.S. Geological Survey (USGS) froman inversion of teleseismic body waves, using analgorithm developed by Sipkin (1982). The report-ed centroid location of the earthquake is 16.432�N,97.973�W, at a depth of 17 km, and the origin time is12:34:49.09. This USGS epicenter, which lies 36 kmtowards N78�W of the epicenter obtained here, is clear-ly inconsistent with the seismograms at PNIG. Themislocation of USGS epicenter follows the pattern pre-viously documented for Mexican events (Singh andLermo, 1985).

Source parameters of the earthquake

From the USGS moment tensor solution, the parame-ters of the best double couple source are: M0 = 2.7 �1024 dyne-cm; nodal plane 1: strike � = 279�, dip �= 29�, rake � = 80�; nodal plane 2: � = 111�, � =61�, � = 96�. In Figure 3a we show this focal mecha-nism and observed polarities at some of the broadbandstations. We note that for this mechanism the stationPNIG should be nodal. This is inconsistent with theobserved vertical seismogram which is clearly dilata-tional (Figure 2). As we show below, the syntheticsgenerated with the USGS mechanism also do not fitthe observed records.

To improve the focal mechanism we used the fol-lowing steps. First, the acceleration traces were inte-

grated twice to obtain displacements. [The displace-ment components clearly show motion between P-and S-wave arrivals due to the near-field effect (Fig-ure 3)]. Then, synthetic seismograms were computedusing Bouchon’s (1982) algorithm, for a set of focalmechanisms varying around the mechanism reportedby the USGS. Initially, synthetics were computed fora half-space (� = 6.2 km/sec, Poisson solid) with a tri-angular source with a base, � , of 0.9 sec, in agreementwith the SH pulse width seen on the transverse com-ponent (Figure 3). While we could find a mechanismwhose synthetics would fit the P and SH pulses andthe near-field part of the displacement, it was not pos-sible to fit the SV pulse on the radial component and,to a lesser extent, on the Z component. This was dueto the arrival of SP phase for the half-space model andsuggested inclusion of a lower-velocity layer above thehalf space. After some tests, a layer of 1.5 km thick-ness with � = 5.0 km/sec (Poisson solid) was foundto yield reasonable synthetics. As shown in Figure 3b,the observed displacements are well fit by the syn-thetics. Based on the modelling of PNIG displacementrecords, the source parameters of the earthquake are:seismic moment M0 = 1.2� 1024 dyne-cm; nodal plane1: � = 291�, � = 10�, and � = 80� and nodal plane2: � = 121�, � = 80�, � = 92�; source duration � =0.9 sec. This focal mechanism, shown in Figure 3b,differs from that reported by the USGS in azimuth by12� and in dip by 19�. Figure 3a compares observeddisplacements with the synthetics generated using theUSGS focal mechanism. The fit is very poor to radialand Z components, especially to the near-field ramppart of the seismogram. To test whether there exists acrustal model which, in conjunction with USGS focalmechanism, would fit the observed seismograms, wegenerated synthetic seismograms for several reason-able crustal models. We found that the near-field rampis not sensitive to the crustal structure, suggesting thatthe USGS focal mechanism is inadequate to explainPNIG records.

For a circular rupture, the radius, r, of the fault canbe estimated by (Boatwright, 1980):

r = (v�1=2)=(1� vsin�=c) (1)

where �1=2 is the rise time of the far-field pulse (=0.45 sec for this event), v is the rupture speed, c isthe wave speed, and � is the take-off angle measuredfrom fault normal. For S-wave, c = � = 3.58 km/sec.Taking v = 0.8 � gives a fault radius r of 3.17 km. Fora circular fault the static stress drop, ��, is related to

jose21.tex; 25/06/1997; 11:58; v.7; p.4

43

Figure 3(a). Right: Best double couple focal mechanism of 27 March, 1996 earthquake reported by the U.S. Geological Survey (USGS),based on inversion of teleseismic body waves. The polarities at Mexican broadband station are shown (solid circle: compression, open circle:dilatation). In this mechanism PNIG, shown by open circle, is nodal. Left: Comparison of observed and synthetics displacements at PNIG. Thesynthetics are generated using the mechanism shown on the right, and a crustal model discussed in the text.

Figure 3(b). Right: The focal mechanism which fits the PNIG records. Left: comparison of observed displacement and synthetic displacement(computed using the mechanism shown on the right). Shaded area (on R and Z components) corresponds to S-wave from a small subevent whichpreceded the main event by 0.18 sec. This subevent is marked S on the P-wave, vertical, displacement seismogram shown in Figure 4.

M0 and r by (Keilis-Borok, 1959):

�� = (7=16)(M0)=r3 (2)

Equation 2 gives a �� of 17 bars for this event.

jose21.tex; 25/06/1997; 11:58; v.7; p.5

44

Figure 4. Top: Z component of the velocity trace. Roughly the firsttwo seconds of the record are shown.Bottom: Z component of the displacement (obtained by direct inte-gration of the velocity trace shown on the top). The vertical linesshow arrival of a subevent and the main event. S and M denote thesubevent and main event pulses, respectively. The areas under S andM are 5 � 10�4 cm-sec and 5.2� 10�2 cm-sec, respectively.

The source complexity

So far we have interpreted the earthquake as a simplesource. A closer examination of Z component of thevelocity and displacement traces (Figure 4) during thefirst two seconds, shows that the main energy releaseduring the earthquake was preceded by at least one rel-atively large subevent. This subevent occurred 0.18 secbefore the main event. The correspondingS-wave fromthis subevent can be seen in Figure 3 (shaded on radialand vertical components), where it precedes the mainpulse by about 0.18 sec. Although two smaller eventscan be seen at the beginning of the seismogram in Fig-ure 4, these may be due to the noncausal nature of FIRfilters used in down sampling from 320 Hz to 80 Hz inthe Quanterra digitizer (see, e.g., Bouin et al., 1996).For this reason we ignore these smaller subevents.

We estimate the seismic moment of the subeventfrom the area under the P pulses of the subevent and themain event (marked S and M, respectively in Figure 4).Assuming the same focal mechanism for both eventswe can write:

(M0)M=(M0)S = AM=AS ; (3)

where subscripts M and S denote the main event and thesubevent, respectively, and A is the area of the pulse.From Figure 4, we have AM = 5.2� 10�2 cm-sec andAS = 5.0 � 10�4 cm-sec. Since (M0)M = 1.2 � 1024

dyne-cm, from Equation 3 we get (M0)S = 1.2 � 1022

dyne-cm (Mw = 4.1). For the subevent � 1=2 is 0.09 secwhich from Equations 1 and 2 gives r = 0.63 km and�� = 21 bars. Since the pulse width is not correctedfor attenuation, this estimate of the stress drop is lowerbound.

Complexity in earthquake rupture process has longbeen observed in teleseismic records (see, e.g., Wyssand Brune, 1967; Kanamori and Stewart, 1978 fororiginal works; and Singh and Mortera, 1991 for com-plexity during the Mexican earthquakes as seen onteleseismic records). Recently, several authors haveanalyzed near-source recordings to study the onset ofearthquakes (e.g., Iio, 1992; Wald et al., 1991; Aber-crombie and Mori, 1994; Ellsworth and Beroza, 1995;Anderson and Chen, 1995; Mori and Kanamori, 1996).The beginning of the Pinotepa earthquake is similarto the Landers earthquake of 28 June, 1992 (MW =7.3), analyzed by Abercrombie and Mori (1994), whoreported that two subevents preceded by about 3 secthe main energy release during the earthquake. Fur-thermore, they found that the subevents’ stress dropswere similar to the main shock stress drop and, in thissense, the subevents were ‘normal’ earthquakes. Thisis also true for the subevent of the Pinotepa earthquake.

Discussion and conclusion

Based almost exclusively on the near-field accelero-grams of the Pinotepa earthquake of 27 March, 1996,obtained at the very broadband station of PNIG, welocated the event and determined its source parame-ters. The only other data used in the analysis was afocal mechanism reported from the inversion of tele-seismic body waves. It was in the neighbourhood of thismechanism that we searched for a mechanism whosesynthetics fit well with the observed records. Evenif the teleseismic mechanism had not been available,we would have found an approximate starting focal

jose21.tex; 25/06/1997; 11:58; v.7; p.6

45

mechanism from the available first motion data, andthe expected mechanism in the region, from tecton-ic considerations. Alternatively, we could have com-puted the moment tensor solution of the earthquakesfrom the PNIG recordings alone, using an algorithmsuch as the one developed by Fan and Wallace (1991).PNIG recording reemphasizes the importance of high-quality, near-field records, even from a single sta-tion, in the reliable estimation of the source para-meters of an earthquake. Following is the summaryof the source parameters of the Pinotepa earthquake:

Location: 16.365�N, 98.303�W, depth = 18 km

Origin time: 12:34:48.35

Seismic moment: 1.2 � 1024 dyne-cm, MW

= 5.4; Source time function: isosceles trianglewith a base of 0.9 sec

Fault plane: strike � = 291�, dip � = 10�, andrake � = 80�

The above estimates refer to the main event of theearthquake. The records show that the main event waspreceded by one or more subevents. The largest ofthe subevents occurred 0.18 sec earlier, had a seismicmoment of 1.2 � 1022 dyne-cm (MW = 4.1), and astress drop of about 21 bars. This stress drop is aboutthe same as for the main event (17 bars), showing thatthe subevent was simply a smaller, ‘regular’ earthquakeand not a slow event. We conclude that the earthquakeof Pinotepa Nacional was complex: it began with therupture of one or more small subevents, which culmi-nated in the rupture of a larger area.

Acknowledgments

Stations CAIG, PLIG and TUIG were partiallyfinanced by the Japanese International CooperativeAgency (JICA), a European Union project (ContractCI1*-CT92-0025), and Pemex, respectively. Fundingfor all other VBB seismic stations came from the Min-istry of Interior, Mexico, through the National Cen-ter for Prevention of Disasters (CENAPRED). Prof.T. Mikumo kindly revised the manuscript. We thankJorge Estrada and Jesus Perez for their dedication inmaintaining the VBB network. It is due to their effortsthat the network functions as well as it does. Thisresearch was partly supported by DGAPA, UNAMprojects IN100795 and IN102494, and the EuropeanUnion (Contract CI1*-CT92-0025).

References

Abercrombie, R. and Mori, J., 1994, Local observations of the onsetof a large earthquake: 28 June 1992, Landers, California, Bull.Seism. Soc. Am. 83, 725–734.

Anderson, J. G. and Chen, Q., 1995, Beginnings of earthquakes inthe Mexican subduction zone on strong-motion accelerograms,Bull. Seism. Soc. Am. 85 , 1107–1115.

Boatwright, J., 1980, A spectral theory for circular seismic sources:simple estimates of source dimension, dynamic stress drops, andradiated energy, Bull. Seism. Soc. Am. 70, 1–28.

Bouchon, M., 1982, The complete synthetics of crustal seismicphases at regional distances, J. Geophys. Res. 87, 1735–1741.

Bouin, M.-P., Scherbaum, F., Singh, S. K. and Pacheco, J., 1996,Nucleation phases and numerical artefacts, Seism. Res. Lett.,(submitted).

Ellsworth, W. L. and Beroza, G. C., 1995, Seismic evidence for anearthquake nucleation phase, Science 268, 851–855.

Fan, G. and Wallace, T., 1991, The determination of source parame-ters for small earthquakes from a single, very broadband seismicstation, Geophys. Res. Lett. 18, 1385–1388.

Iio, Y., 1992. Slow initial phase of the P-wave velocity pulse gener-ated by microearthquakes, Geophys. Res. Lett. 19, 477–480.

Kanamori, H. and Stewart, G. S., 1978, Seismological aspects of theGuatemala earthquake of 4 February, 1976, J. Geophy. Res. 83,3427–3434.

Kanamori, H. Mori, J. and Heaton, T. H., 1990, The 3 December1988, Pasadena earthquake (ML = 4.9) recorded with very broad-band system in Pasadena, Bull. Seism. Soc. Am. 80, 483–487.

Keilis-Borok, V., 1959, On estimation of displacement in an earth-quake source and of source dimension, Ann. Geofis. (Rome) 12,205–214.

Mori, J. and Kanamori, H., 1996, Initial rupture of earthquakes inthe 1995 Ridgecrest, California, sequence, Geophys. Res. Lett.23, 2437–2440.

Singh, S. K. and Lermo, J., 1985, Mislocation of Mexican earth-quakes as reported in international bulletins, Geofis. Intern. 24,333–351.

Singh, S. K. and Mortera, F., 1991, Source time functions of largeMexican subduction earthquakes, morphology of the Benioffzone, age of the plate, and their tectonic implications, J. Geo-phys. Res. 96, 21487–21502.

Sipkin, S. A., 1982, Estimation of earthquake source parametersby the inversion of waveform data: synthetic waveforms, Phys.Earth Planet. Interiors 30, 242–259.

Valdes, C., Mooney, W. D., Singh, S. K., Meyer, R. P., Lomnitz,C., Luetgert, J. H., Helsley, C. E., Lewis, B. T. R. and Mena,M., 1986, Crustal structure of Oaxaca, Mexico, from seismicrefraction measurements, Bull. Seism. Soc. Am. 76, 547–563.

Wald, D. J., Helmberger, D. V. and Heaton, T. H., 1991, Rupturemodel of 1989 Loma Prieta earthquake from the inversion ofstrong-motion and broadband teleseismic data, Bull. Seism. Soc.Am. 81, 1540–1572.

Wyss, M. and Brune, J. N., 1967, The Alaska earthquake of 28March, 1964: a complex multiple rupture, Bull. Seism. Soc. Am.57, 1017–1023.

jose21.tex; 25/06/1997; 11:58; v.7; p.7