Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
DOI: 10.1007/s10950-006-9014-2
C Springer 2006
Hernando Tavera, Efraı́n Fernández, Isabel Bernal, Yanet Antayhua, Consuelo Agüero,
Henry Salas Simeón Rodrı́guez, Luis Vilcapoma, Yolanda Zamudio, David Portugal,
Adolfo Inza, Julia Carpio, Freddy Ccallo & Igor Valdivia
Centro Nacional de Datos Geofı́sicos – Sismologia, Instituto Geofı́sico del Perú, Calle Badajoz 169, Urb.
Mayorazgo IV Etapa, Ate, Lima, Perú
e-mail: hjtavera@geo.igp.gob.pe
Key words: Arequipa earthquake, focal mechanism, intensity, reverse faulting, subduction, Tsunami
Abstract
The western border of South America is one of the most important seismogenic regions in the world. In this
region the most damaging earthquake ever recorded occurred. In June 23rd, 2001, another very strong earthquake
(Mw = 8.1–8.2) occurred and produced death and damages in the whole southern region of Peru. This earthquake
was originated by a friction process between Nazca and South American plates and affected an area of about 300 km
×120 km defined by the distribution of more than 220 aftershocks recorded by a local seismic network that operated
20 days. The epicenter of the main shock was localized in the northwestern extremity of the aftershock area, which
suggests that the rupture propagated towards the SE direction. The modeling of P-wave for teleseismic distances
permitted to define a focal mechanism of reverse type with NW-SE oriented nodal planes and a possible fault plane
moving beneath almost horizontally in NE direction. The source time function (STF) suggests a complex process
of rupture during 85 sec with 2 successive sources. The second one of greater size, and located approximately
100–120 km toward the SE direction was estimated to have a rupture velocity of about 2 km/sec on a 28◦ -dipping
plane to the SE (N135◦ ). A second event happened 45 sec after the first one with an epicenter 130km farther to the
SE and a complex STF. This event and the second source of the main shock caused a Tsunami with waves from 7
to 8 meters that propagated almost orthogonally to the coast line, by affecting mainly the Camaná area.
Three of all the aftershocks presented magnitudes greater or equal to Mw = 6.6, two of them occurred in front of
the cities of Ilo and Mollendo (June 26th and July 7th) with focal mechanisms similar to the main seismic event.
The aftershock of July 5th shows a normal mechanism at a depth of 75 km, and is therefore most likely located
within the subducting Nazca plate and not in the coupling. The aftershocks of June 26th (Mw = 6.6) and July
5th (Mw = 6.6) show simple short duration STF. The aftershock of July 7th (Mw = 7.5) with 27-second duration
suggests a complex process of energy release with the possible occurrence of a secondary shock with lower focal
depth and focal mechanism of inverse type with a great lateral component. Simple and composed focal mechanisms
were elaborated for the aftershocks and all have similar characteristics to the main earthquake.
The earthquake of June 23rd caused major damages in the whole southern Peru. The damage in towns of Arequipa,
Moquegua allow to consider maximum intensities from 6 to 7 (MSK79). In Alto de la Alianza and Ciudad Nueva
zones from Tacna, the maximum intensity was of 7− (MSK79).
et al. (2002) and allow to localize the epicenter at a Figure 3 shows the spatial distribution, between the
50 km distance to the WSW from the one reported by coast line and the trench, of 220 aftershocks that de-
the Geological Survey Preliminary Report. The seis- fine a rectangular 300 × 120 km2 area. If the main
mic focus depth was estimated in 23 km. The Geo- earthquake epicenter location and the aftershocks area
physical Institute of Peru reported a ML(d) = 6.9 mag- geometry is considered, it is evident that the rup-
nitude obtained as from the earthquake record duration ture process was propagated on an heterogeneous ba-
in 3 stations located at 550 km medium distances and sis toward the SE direction, as from the locality of
then, such magnitude was re-calculated by using shal- Ocoña in the Department of Arequipa up to Ilo in
low waves in Ms = 7.8. Moquegua. In contrast to the NW extremity of the af-
Figure 2. Spatial distribution of the earthquakes with mb ≥ 6.0 occurred in the southern region of Peru and the North of Chile between years
1990 and 2000 (data from the NEIC). The broken lines circle the rupture areas of the earthquakes of 1868 and 1877. The black arrow indicates
the plates convergence direction and the gray one, the rupture length of the earthquake of 2001.
Table 1. Seismological stations used in this study: local and regional Table 2. Source velocity structures used in hypocenter determina-
ones tion and waveform inversion
and the third one in SE direction between the lo- the presence of an area that would have not expe-
calities of Camana and Ilo (GROUP3). From these rienced rupture during the earthquake could be ob-
groups, GROUP1 presents more aftershocks, being all served, and it could be considered as an asperity or
of them concentrated around the epicenter of the main if this not be the case, the plates displacement would
earthquake. In general, among the aftershocks groups, have been produced on an aseismic way in this area
(shadowed area in Figure 3). Both hypothesis are valid epicenter of the main earthquake, which suggests that
when considering that the earthquake of June 23 had the occurrence of small deformations could give rise
a complex rupture process over a very heterogeneous to other moderate-magnitude earthquakes after some
surface. time.
From all the aftershocks (Figure 3), three had high
magnitudes and were produced at different distances
from the main earthquake, and they caused secondary Focal mechanisms
damages in previously damaged housing, and also gen-
erated panic and desperation in the population that The focal mechanisms of the main earthquake and
thought a bigger disaster would take place. From these higher magnitude aftershocks were obtained by Tavera
aftershocks, the one taken place on June 26th, with a et al. (2002) as from the characteristics of the first P-
Mw = 6.0 magnitude had its epicenter at 36 km west- wave polarity, being all of them of inverse type with
ward with regard to the locality of Ilo, where V in- nodal planes oriented in NW-SE direction and an al-
tensities in the Modified Mercalli scale (MM) were most horizontal possible fault plane moving beneath
produced. The aftershock of July 5th with a Mw = 6.6 in NE directions. Solutions are coherent with regional
magnitude, had its epicenter at 79 km at NE direction deformation pattern as a result of the plates conver-
with regard to the main earthquake, by producing V gence. The aftershock of July 5th only corresponds to
(MM) intensities in the localities of Atico, Ocoña and a normal-type mechanism with the possible fault plane
Caraveli in Arequipa, and the aftershock of July 7th moving beneath with higher angle at a NE direction.
with a Mw = 7.5 magnitude took place at 54 km at This aftershock occurs at a higher depth (∼60 km) and
SW direction as from the locality of Mollendo, where the author states that it would have been produced due
it produced V (MM) intensities. The other aftershocks to the internal deformation of the ocean plate under the
had ML magnitudes, ranging from 2.4 to 4.8, and only continent.
some of them could be noticed at surface. In the case of the other aftershocks of earthquake
According to the vertical section (Figure 3), after- dated June 23rd, we have elaborated some composed
shocks are distributed in depth by following an ap- focal mechanisms by using the Brillinger logarithm
parent line with a 28◦ inclination up to around 60 km (Brillinger et al., 1980) logarithm and the polarity of
of depth, being these values coherent with the geom- the first P-wave impulse in the vertical component
etry of the subduction process in this region. A small (Figure 4). Based on the above-described regrouping
group of aftershocks are located in continent around the of the aftershocks, GROUP1 has been divided into 3
Figure 4. Simple and composed focal mechanisms elaborated as from the P-wave polarity for the three groups of aftershocks shown in Figure 3.
The areas with black circles indicate compression, and the white ones, dilatation. P and T correspond to the pressure and tension axis.
groups (G11, G12 y G13) and for each of them, a com- tude aftershocks. The values for the seismic moment
posed focal mechanism has been elaborated, being all and radius fracture correspond to the average of values
of them of inverse type with nodal planes oriented to- obtained for the spectrum flat part and for the frequency
ward the same direction of the main earthquake, how- corner. For the main earthquake, a seismic moment of
ever groups G12 and G13 have a small strike-slip com- 1.2 ± 0.1 × 1021 Nm equivalent to a Mw = 8.2 magni-
ponent. In the case of GROUP2, the orientation of the tude is obtained; and for the aftershocks of June 26th, a
composed focal mechanism is similar to the one of the seismic moment of 1.8 ± 0.3 × 1019 Nm (Mw = 6.8);
main earthquake. In the case of GROUP3, 4 simple fo- for the aftershock of July 5th, a seismic moment of
cal mechanisms have been elaborated considering the 7.2 ± 0.4 × 1018 Nm (Mw = 6.5); and for the after-
aftershocks with more polarity data, being all solutions shock of July 7th, a seismic moment of 2.0 ± 0.3 × 1020
similar to the main earthquake one. In general the de- Nm (Mw = 7.5). The fracture radius for the main earth-
formation pattern over all the aftershocks area was ap- quake would be of the order of 100 km, lower to the
proximately the same one. one proposed as from the spatial distribution of the af-
tershocks, but could be equivalent to the dimensions
of the area affected only by the earthquake. The major
Spectral analysis magnitude aftershocks have 20, 18 and 50 km fracture
radius, respectively (Table 3).
Before the modeling of volume waves, the spectral
analysis of the main earthquake and of its three higher
magnitude aftershocks will be carried out by using for Modeling of P-wave
each of them, 6 P-waves records at teleseismic dis-
tances (IRIS network). In general, for a circular frac- The waveform and amplitude of P-wave were finally
ture model, the parameters of the earthquakes source used to estimate the source orientation, the focal depth,
could be estimated as from two of the characteristics the seismic moment and the time function of the seis-
of the amplitude spectrum, the frequency corner (fc) mic source when using the inversion method proposed
and its flat part (O ) The f c value is reversely propor- by Nabelek (1984). The P-wave recorded on broad-
tional to the source dimensions (r0 = f c−1 ); while, O band stations localized at distances between 30◦ and
is proportional to the scalar seismic moment (Hansk 90◦ were compared with waveforms of the synthetic
and Wyss, 1972). If a complete effort drop is consid- records generated by the combination of direct phases
ered, the seismic moment is defined as (Keilis-Borok, (P-wave) and reflected phases (pP and sP) that were
1959): originated in the focal area and propagated with a deter-
mined velocity in an homogeneous structure. The am-
Mo(P) = 4π.ρ.α 3 .0 (P).R/Rθφ (P) plitudes are adjusted considering the geometric spread-
ing factor and the attenuation (Futerman, 1962) with
and the source dimensions (Brune, 1970): r (P) = t ∗ = 1 second for the P waves. The fit between the
2.34α/2π f c (P) observed and synthetic traces is minimized with half
where, α is the wave velocity of the P-wave, ρ the mean square. The inversion allows the estimation of
medium density, O (P) the spectrum flat part, R the the source parameters (azimuth, dip and slip angle).
epicentral distance, Rθφ the P-wave radiation pattern, The depth and source time function are calculated with
and f c (P) the frequency corner. a series of triangles that overlap (Nabelek, 1984) and
The P-wave records of each station were corrected which number and duration are previously selected.
by their instrumental response, attenuation and radia- In order to estimate the source parameters, the
tion pattern by following the procedure described by process described by McCaffrey and Nabelek (1987),
Houston and Kanamori (1986). In order to extract the Molnar and Lyon-Caen (1989), Taymaz et al. (1991)
maximum information on the rupture process, the P- and Tavera and Buforn (2001) was followed. For elab-
wave records have not been filtered. Once the scalar orating the model, the parameters have been standard-
seismic moment be known, the Mw magnitude was es- ized to a 5000 amplitude and to a 40◦ epicenter dis-
timated with the following Mw = (3/2) log Mo −10.7 tance. The parameters that define the source orienta-
relation (Kanamori, 1977). tion from the focal mechanism of the main shock and
Figure 5 shows some examples of the displacement aftershocks of great magnitude were initially fixed by
amplitude for the main earthquake and major magni- using the P-wave polarity of Tavera et al. (2002), and
Figure 5. P-wave amplitude spectrums at teleseismic distances elaborated for the main earthquake and aftershocks of major magnitude. In
parenthesis, next to the station code, the epicentral distance and the azimuth in degrees are indicated. = flat part of the spectrum and f c,
frequency corner. Under each spectrum, the used seismic signal window of each spectrum.
the hypocenter parameters reported by the IGP. After number of iterations, in which it could be seen that for
each inversion step, a visual check of the fit between all the data sets, it is possible to model the first arrival
the synthetic and observed records was carried out, in of the signal and pulses localized about 85 sec. In order
order to evaluate the uncertainty of each parameter un- to achieve a model for the whole signal, the possible
til the lower values for the RMS of the fit that should rupture propagation toward the SE was considered, as
correspond to the best solution be obtained. suggested by the spatial distribution of aftershocks. In
that sense, various inversion processes are carried out
Main shock by considering several values for the rupture propaga-
tion values, rupture plane inclination, rupture velocity
In order to model the amplitudes and waveforms of and focus depth. The results obtained for each param-
the main shock, 9 P-wave records were used. The final eter, based on the RMS solution are shown in Figure 8.
solution is presented in Figure 6, but before dealing According to the lower RMS, the rupture process starts
the results, it is necessary to describe the process that at a 29 km depth and is propagated at N135◦ direction
was followed. In general, when carrying out a compar- over a sloped surface in 28◦ NE and with a 2 km/second
ison between each main earthquake record based on velocity. The orientation and inclination of the rupture
the azimuthal distribution of the stations around the fo- plane is coherent with the geometry of the aftershocks
cal sphere, it could be observed that the waveforms are area and with the friction surface of the plates; however,
very complex and even more in the records of DBIC, the rupture velocity is very low, but this would explain
BOSA and ANMO stations. In order to propose vari- the great duration of the earthquake and its complexity.
ous possible solutions, one test with 6 data set shall be The synthetic data generated with these new param-
carried out, being the results shown in Figure 7. Line 3 eters allow modeling almost the whole P-waves record
of this figure shows the solution obtained after a certain (line 2, Figure 7), but it is still not possible to completely
Table 3. Source parameters of earthquake and great aftershock of June 23rd, 2001 obtained from hypocenter determination,
body wave inversion and spectral analysis.
model the records of the ANMO and CCM stations. In It was produced in a SE direction on a mean length
order to model their waveforms, the possible occur- of 100 km from the main shock. 45 sec after the start
rence of another earthquake showing the same focal of the main shock, a second one was produced along
mechanism trend shall be considered. Consequently, the same propagation trend of the rupture. Following
the inversion process was carried out by setting every the STF, this shock has generated three small ruptures
parameter of the first source and leaving free those cor- in a time period of approximately 40 sec. The seis-
responding to the second one. After various iterations, mic moment for the first shock is about 1.3 × 1021 Nm
stable solutions were reached with RMS = 17, and the (Mw = 8.1) and for the second one of about 2.8 × 1020
results appear on line 1, Figure 7. It can be seen that Nm (Mw = 7.6). The summary of results appears in
the fit between synthetic and observed records is much Table 3.
better. According to Figure 6, the focal mechanism of
both earthquakes are a reverse fault with nodal planes Aftershocks of June 26th
trending NW-SE with slightly different dip of the fault
plane. The focal depth of the first shock is 29 km and Figure 9 shows the solution for the aftershock of June
of the second one increases to 10 km. 26, using 10 P-waves records. The inversion process
These results suggest that the main shock has been has been initiated by setting the focal orientation ob-
generated by the occurrence of three ruptures that prop- tained by Tavera et al. (2002) and leaving free the focus
agated toward the SE (N135◦ ) on an 28◦ dipping plane depth and the temporary function of the STF seismic
and with a velocity of about 2 km/seg. This value for source. After various iterations, the best fit between
the rupture propagation velocity is close to the one pro- synthetic and observed is achieved with RMS = 14.
posed by Salden et al. (2004) based on seismic tomog- The solution corresponds to a focal mechanism with
raphy studies (1.6 km/second), and in general, are sim- a similar trend to the main shock with a plane fault dip-
ilar to the ones observed for the Tsunamis earthquakes. ping 21◦ to the east. The focal depth is fitted to 32km.
The total duration of the whole process was about 85 The STF is simple with a duration of about 5 seconds
seconds and the third rupture was the most important. and a seismic moment of 1.19 × 1019 Nm, equivalent
Figure 6. Forms of the synthetic wave (synt) and observed ones (obs) corresponding to the main earthquake and obtained in the inversion
process according to the Nabelek methodology (1984). The amplitude of the records have been normalized at a 5000 gain and at a 40◦ distance.
The inversion window is stated with vertical lines over the records. The station code is in the left extremity of the record, being the epicentral
distance and azimuth in degrees under the code. The focal sphere corresponds to its projection in the lower hemisphere, being the P and T axis
represented by black and white circles. The STF is presented under the focal mechanism, as well as the scale of the records. In the upper part, the
focal mechanisms corresponding to the two seismic events (E1, E2) are shown as well as the solution according to Tavera et al. (2002) as from
the P wave polarity (TB). The figure of the upper left corner shows the epicentral location of the two seismic events associated to the rupture
process of the earthquake of June 23rd, 2001.
to a Mw = 6.6 magnitude. The summary of results ap- tween P and pP of about 18 sec equivalent to a bigger
pears in Table 3. focal depth of 60 km (Figure 10). The iterative process
begins considering the source orientation obtained by
Aftershock of July 5 Tavera et al. (2002) and leaving free the focal depth
and STF. After various iterations, the obtained solu-
In order to model the aftershock of July 5th, 6 records tions did not achieve a good fit between the synthetic
of P-wave were used and the results are shown in and observed records, mainly for the CMB and PAS
Figure 10. A preliminary analysis of those records al- stations; thus the slip angle was left free and proceeded
lows the observation of a difference in time arrivals be- new iterations. The final solution corresponds to a 0.23
Figure 7. Results obtained in the test carried out during the inversion process for the main earthquake. The solution of line 3 corresponds to the
synthetic-observed adjustment considering the orientation for the source according to Tavera et al. (2002); line 2 corresponds to the solution by
assuming the unilateral propagation of the rupture in SE direction and line 1, considering the occurrence of a second event and that corresponds
to the final solution for the earthquake of June 23rd, 2001. The arrows indicate the stretch of the record which synthetic-observed correlation
improves in each stage of the test.
Figure 8. Results of the test carried out for evaluating different values for the rupture propagation direction, plane slope and rupture velocity,
besides the depth of the focus corresponding to the earthquake of June 23rd, 2001.
RMS, and the first impulses of all the records could be phases PcP due to epicenter distances between 92◦ and
modeled, even the starts emerging from ESK, PAB and 94◦ . The solution corresponds to a normal type mecha-
VSL stations but not the following greater amplitude nism with nodal planes trending toward NW-SE, and a
pulses, those ones that could correspond to reflected possible fault plane, dipping to NE. This solution dif-
Figure 9. Synthetic (synt) and observed waves (obs) corresponding to the aftershock of June 26th, 2001 obtained during the inversion process.
Others, see Figure 6.
fers partially from the one obtained by Tavera et al. sion process is initiated by setting the source orienta-
(2002), mainly in the orientation of the plane that dips tion obtained by Tavera et al. (2002) and leaving free
to the SW, and that allowed to improve the solution the focal depth and STF. According to Figure 11, the
obtained in this study. The focal depth is fitted to 75 wave shapes are complex and even more the ones corre-
km. The STF is calculated for 3 impulsions except for sponding to the records of the TUC, ANMO and HKT
the first main one with a 5-second duration and a seis- stations, being all of them located at the NW end of the
mic moment of 1.1 × 1019 Nm equivalent to Mw = 6.6 focal sphere. For a better analysis of the seismic signal,
magnitude. The summary of results appears in Table 3. five records have been selected for carrying out some
tests and the results obtained with the first inversions
Aftershock of July 7th appear in line 2, Figure 12. In such figure, it could be
seen that the records of HRV, BOSA and PLCA stations
In order to model the July 7th aftershock, 11 obser- are acceptably modeled, but not the records of the TUC
vation of P-waves (Figure 9) were used. The inver- and HKT stations, mainly after 10-record seconds.
Figure 10. Synthetic (synt) and observed waves (obs) corresponding to the aftershock of July 5th, 2001 obtained during the inversion process.
Others, see Figure 6.
In order to propose a model explaining all obser- shock, the focal depth is 18 km, and in the case of the
vations, we have processed the data set by assuming second one it is shallower, about 12 km. The STF of the
a possible occurrence of a second shock. The inver- first shock shows a 27-second duration and describes
sion process begins considering the parameters of the the occurrence of three quasi simultaneous ruptures
first event, and letting free those of the second one. with a seismic moment of 3.4 × 1020 Nm (Mw = 7.6).
After some iterations, the synthetic-observed correla- The second event shows one STF of similar form for
tion was improved showing an RMS = 27, as could be the first shock and with an approximate 25-second du-
seen on line 1, Figure 12. By following this solution, ration. The scalar seismic moment is 2.0 × 1020 Nm
the records of station TUC and HKT are better fitted. (Mw = 7.5). The summary of results appears in Table 3.
The final solution of Figure 11 suggests the occurrence
of two separated earthquakes in a time window of 5
sec. The first shock corresponds to a reverse type fo- Tsunami
cal mechanism with the same orientation than the main
shock, the second earthquake would present an thrust In June 23rd 2001, the earthquake produced a local
fault mechanism but with major strike-slip component Tsunami that affected only the town of Camana (Are-
and nodal planes oriented to NS and other to EW di- quipa region), localized 110km to the SE of the earth-
rection. During the inversion process, each inversion quake epicenter (Figure 13). The IGP sent a group
step was tried to be initiated by considering a source of professionals -geophysicians and geologists- to the
with the main shock orientation, but the solution was al- Tsunami-affected zone in order to carry out an eval-
ways stable with a lateral component. In the case of first uation of the damages and gather information on the
Figure 11. Synthetic (synt) and observed waves (obs) corresponding to the aftershock of July 7th, 2001 obtained during the inversion process.
Others, see Figure 6.
Tsunami characteristics. The gathered information and mana, the sea withdrew 200 meters and its return did
the obtained results indicate that 15 min after the main produce important damages with the Tsunami waves
earthquake took place, the sea withdrew between 100 that reached about 4 to 7–8 meters high between La
and 200 meters as from the locality of Chala up to Ilo. Chira and Las Cuevas beaches. The maximum height
In the localities of Chala, Quilca, Mollendo and Ilo, was observed in El Chiflon and in El Choro in the vicin-
the return of the sea middle level was slow and did not ity of the Camaná river mouth and the Tilimaco lake.
produce damages; while in the localities of Ocoña and In Camana beaches, the flooding occurred obliquely
Atico, the sea flooded the beaches between 30 and 50 to the coastline and extended along 23 km approxi-
meters through the mouth of the Ocoña river. In Ca- mately. Flood levels in perpendicular direction to the
Figure 12. Results of the test carried out for the inversion process of the aftershock of July 7th, 2001. The line 2 solution corresponds to the
synthetic-observed adjustment obtained considering the orientation for the source according to Tavera et al. (2002); and line 1, assuming the
occurrence of a second event and that corresponds to the final solution for such aftershock. The arrows indicate the stretch of the record which
synthetic-observed correlation improves in each test stage.