You are on page 1of 15

Bulletin of the Seismological Society of America, Vol. 104, No. 2, pp. 840–854, April 2014, doi: 10.

1785/0120130041

Scaling Relationships of Source Parameters of M w 6.9–8.1 Earthquakes


in the Cocos–Rivera–North American Subduction Zone
by Alejandro Ramírez-Gaytán, Jorge Aguirre, Miguel A. Jaimes, and Víctor Huérfano

Abstract Seven slip models currently available from kinematic inversions, derived
from near-source strong-motion and teleseismic body waves in the 0–1.25 Hz fre-
quency range from Mexico’s subduction zone, are used to estimate source-scaling
relationships applicable to the region. Our results are compared with existing scaling
relations for subduction environments. The relationships for the rupture area of our
results are closer to those of Somerville et al. (2002) than to any other, but, like the
others, they have smaller areas than predicted by Somerville et al. (2002). Concerning
the combined area of asperities, Murotani et al. (2008) and our results predict smaller
areas than those obtained by Somerville et al. (2002). Concerning the area of largest
asperity, the relationships obtained in this study are slightly smaller than those de-
scribed by Somerville et al. (2002); this is a consistent result with the relationships
of total rupture area and combined area of asperities. In general, the error estimates for
the constrained equations derived in this study in all cases are smaller than those re-
lationships compared here. This might suggest that the expressions obtained in this
study could be appropriate for the simulations of strong ground motion for a specific
scenario of earthquake slip in the region. Also, these results could be an indication that
the relationships vary depending on a specific subduction tectonic region. On the other
hand, Aguirre and Irikura (2007) estimated the source area for 31 Mexican earth-
quakes using corner frequencies; these areas show close resemblance with those pre-
dicted by the relationships derived in this study. Based on these findings, an important
implication is that two different methodologies to determine the total area of asperities
based on either low- or high-frequency data generate similar results.

Online Material: Figures of fault models.

Introduction
Simulation of earthquake scenarios is a topic of rel- and Archuleta, 2003), in terms of rupture area to the seismic
evance in countries with a high rate of seismic activity, such moment for plate-boundary earthquakes (Kanamori and
as Mexico. A standard procedure to evaluate seismic hazard Anderson, 1975), and in terms of rupture area to the moment
is based on simulation of ground motions generated by active magnitude (Wells and Coppersmith, 1994). The influence of
faults. It has been demonstrated that asperities within the rup- slip heterogeneities in the prediction of strong ground motion
ture area control specific characteristics of ground motions within a subduction zone has been investigated by Somer-
(Miyake et al., 2004). Also, these relationships allow mod- ville et al. (2002), using worldwide events, and recently by
elers to understand the seismic source and generate response Murotani et al. (2008), who focused on Japanese events.
spectra, both useful pieces of information in structural engi- Blaser et al. (2010) and Strasser et al. (2010) derive source-
neering. Therefore, quantitative estimation of the size and scaling relations between rupture dimensions and moment
slip of the asperities is important in modeling the source magnitude for subduction-zone earthquakes. In general,
for predicting ground motions. these relationships establish equations useful in the genera-
Several methods are used to simulate strong ground tion of strong ground motions.
motions from potential seismic sources; one of them is the Another method used to simulate strong ground motions
so-called source-scaling relationships. Recent studies related from potential seismic sources is the empirical Green’s func-
the features of an earthquake source via the scaling of slip tion method (EGFM). This is a good alternative methodology
distributions (e.g., Somerville et al., 1999; Mai and Beroza, for simulating strong ground motion produced by large-
2000), slip complexity (e.g., Mai and Beroza, 2002; Lavallée magnitude earthquakes. This method requires a small-

840
Scaling Relationships of Source Parameters of Mw 6.9–8.1 Earthquakes 841

magnitude earthquake (element event) with a hypocenter subduction zone in Mexico. However, the accuracy of
close to the main earthquake (target event) that will be simu- source-scaling relationships depends strongly on the number
lated. The broadband modeling process developed and of kinematic inversions considered in the study. Somerville
detailed by Irikura (1986) produces in the end a source model et al. (1999) used 15 kinematic models from crustal earth-
in which synthetic strong ground motions share the best re- quakes, and Somerville et al. (2002) used 10 kinematic mod-
semblance with observed records of the main earthquake; it els from interplate subduction-zone earthquakes. Asano and
is from these final models that the source parameters are es- Iwata (2011) used five kinematic models from inland crustal
timated. There are two important characteristics of the strong earthquakes in Japan. In our study, despite the fact that in
ground motions generated by EGFM. First, the information of Mexico kinematic inversions are scarce, we utilized the data
crustal structure and site effects are included in the simula- provided for seven earthquakes in the Mexican subduction
tions. Because records of the element event used as a seed environment whose kinematic inversion is currently avail-
already include them, instrumentation and detailed studies to able. With these data and as a product of this work, we gen-
determine the crustal structure and site effects are not neces- erate new source-scaling relationships to simulate strong
sary. Second, EGFM modeling is possible in the 0.1–10 Hz ground motion applicable to Mexican subduction-zone mega-
frequency interval. This is relevant for civil engineering be- thrust earthquakes between the Cocos–Rivera and North
cause it is in this frequency range that many buildings, American plates.
bridges, and civil constructions have their dominant vibra- As a secondary task, we compare our results with the
tion frequency. scaling relationships for subduction earthquakes first pro-
Recently, some studies have been conducted in Mexico posed by Somerville et al. (2002) and with the other important
to simulate strong ground motions of important earthquakes recently developed scaling relations for subduction environ-
using EGFM (e.g., Aguirre, 2005; Garduño, 2006; Ramírez- ments: Murotani et al. (2008), Blaser et al. (2010), and
Gaytán et al., 2010). In these studies, source parameters that Strasser et al. (2010). Murotani et al. (2008) characterized
were estimated, such as the combined area of asperities and source rupture models with heterogeneous slip of plate boun-
area of largest asperities, are poorly matched with the relation- dary earthquakes in the Japan region. The slip models used are
ships proposed by Somerville et al. (2002). In their compari- based on 26 estimates of 11 plate boundary earthquakes con-
son, two important facts arise: (1) EGFM is a method capable structed by waveform inversions of strong ground motions,
of modeling strong-motion generation areas (SMGA), which teleseismic, geodetic, or tsunami data. The source-scaling re-
are defined as finite extent areas with relatively large slip lationships that specifically relate seismic moment to rupture
velocity within the total rupture area, and (2) the waveform area, average slip, and combined area of asperities derived by
inversions capture the presence of asperities (characterized Murotani et al. (2008) will be compared subsequently with
by heterogeneous slip distribution) based on low-frequency our results. Strasser et al. (2010) derived source-scaling rela-
(< 1 Hz) ground motions. Based on these descriptions, the tions between rupture dimensions and moment magnitude for
following questions arise: what do the two methods of study- subduction-zone earthquakes, distinguishing interface events
ing the source in different frequency bands have in common, from intraslab events. They use a database that includes 139
how much energy is radiated in the high-frequency band, and models corresponding to 95 interface events and 21 models
how can high frequencies be included in their approach. corresponding to 20 intraslab events. The relationship that
Miyake et al. (2003) showed that the SMGA generated from involves the rupture area versus moment magnitude for inter-
0.2–10 Hz frequency ranges coincide with the areas of the face events derived from Strasser et al. (2010) will be com-
asperities of heterogeneous slip distributions derived from pared with the equivalent relationship derived in this study.
low-frequency (< 1 Hz) waveform inversions for crustal Blaser et al. (2010) compiled a large database of source
earthquakes. In the same study, Miyake et al. (2003) showed parameter estimates of 283 earthquakes for which all focal
that SMGA contain both low- and high-frequency information mechanisms are represented, and specifically the focus is on
and that quantified SMGA have the ability to perform the subduction-zone events. The product of rupture length and
broadband ground-motion simulation. width for subduction-zone earthquakes and continental thrust
Based on the preceding discussion, we have compared events derived by Blaser et al. (2010) will be compared with
the scaling relationships between seismic moment and the the rupture area estimates in our study.
source parameters estimated using EGFM (combined area
of asperities, area of largest asperity) with those relationships Data
proposed by Somerville et al. (2002) for the studies conducted
on the important Mexican earthquakes mentioned here In this study, we have taken results of seven kinematic
(Aguirre, 2005; Garduño, 2006; Ramírez-Gaytán et al., 2010). inversions available from earthquakes in the Cocos–Rivera
The results of this comparison show a poor fit. These results and North American plates’ subduction environment in
might suggest the need to develop specific source-scaling Mexico (Table 1 and Fig. 1). These inversions consider results
relationships for Mexican subduction earthquakes. of thrust-faulting earthquakes with moment magnitudes from
In this study, we develop new scaling relationships for 6.9 to 8.1 that occurred within the last 26 years (1979–2003).
earthquakes occurring in the Cocos–Rivera–North American These results are composed of six kinematic inversions,
Table 1
842

Fault Parameters from the Heterogeneous Slip Models


Date Longitude; Depth Strike; Dip; Vr Sc Average Sl Le We
Reference* Event Name† (yyyy/mm/dd)‡ Latitude (°)† M 0 (N·m)† Mw † (km)† S (km2 )§ Rake (°)† (km=s)† (km=s)§ Slip (m)§ (km2 )§ (km)† (km)†

1 Petatlán 1979/03/14 −101.46; 17.46 1:37 × 1020 7.39 15.00 10 293; 14; 90 3.3 3800 0.288 3600 120 120
2 Playa Azul 1981/10/25 −102.24; 17.74 8:49 × 1019 7.25 15.00 2700 300; 14; 90 2.6 400 0.746 400 60 70
3 Michoacán 1985/09/19 −102.57; 18.18 1:15 × 1021 8.01 17.00 25020 300; 14; 72 2.6 5004 1.390 3127.5 180 139
4 Zihuatanejo 1985/09/21 −101.82; 17.6 1:53 × 1020 7.42 20.00 3500 300; 14; 100 2.6 1250 1.028 1250 90 90
5 San Marcos 1989/04/25 −99.12; 16.83‡ 2:4 × 1019 ‡ 6.90‡ 17.30‖ 2520‖ 276; 10; 66‡ – 324§ 1.265§ 324§ 60‖ 42‖
6 Manzanillo 1995/10/09 −104.58; 18.86 9:67 × 1020 7.96 16.55 17000 309; 14; 92 2.8 3400 1.355 2100 200 100
7 Tecomán 2003/01/22 −104.13; 18.71 2:30 × 1020 7.50 20.00 5950 300; 22; 93 3.5 1075 0.607 700 70 85

M 0 , seismic moment for slip subduction models; M w , moment magnitude; S, dimension of rupture area; V r , rupture velocity; Sc , estimated combined area of asperities; Sl , estimated area of
largest asperity; W e and Le , estimated effective width and length fault dimensions.
*1, Mendoza (1995); 2 and 4, Mendoza (1993); 3, Mendoza and Hartzell (1989); 5, Singh et al. (1989); 6, Mendoza and Hartzell (1999); 7, Yagi et al. (2004).

Information is from SRCMOD, Martin Mai’s Database of Finite-Source Rupture Models (see Data and Resources).

Information is from the Global Centroid Moment Tensor (CMT) project catalog (1976–2013).
§The source parameters determined in this study.

Information is from Singh et al. (1989).
A. Ramírez-Gaytán, J. Aguirre, M. A. Jaimes, and V. Huérfano
Scaling Relationships of Source Parameters of Mw 6.9–8.1 Earthquakes 843

tember 1999 (Castro and Ruiz-Cruz, 2005). However, these


studies were excluded from the analysis because the events
are not thrust-faulting but normal-faulting earthquakes. In
order to conduct an adequate comparison with relationships
proposed by Somerville et al. (2002) in this study, we take
only the seven reverse-slip subduction earthquakes available
at this time.

Methodology
The first step in processing the information was to esti-
mate the total rupture area and the number of asperities. The
rupture area dimensions of the inverted slip models are fre-
quently overestimated in order to accommodate the entire
fault rupture into a rectangular area. We reduced the dimen-
sions of the rectangular slips models in order to consider just
the area with significant slip contribution. We apply a standard
Figure 1. Distribution of Mexican earthquakes for which slip
models were available as of the present study. Events 1–7 are
criterion for trimming the edges, as proposed by Somerville
the events used in this study. Events 8 and 9 represent the events et al. (1999). The trimmed fault is then defined as the total
excluded from the present analysis. The color version of this figure rupture area. In our study and throughout the present paper,
is available only in the electronic edition. we follow the criteria of Somerville et al. (1999) to define this
and other source parameters: rupture area A, average slip D,
compiled primarily based on the SRCMOD database from
combined area of asperities Aa , area of largest asperities AL ,
Martin Mai and coworkers (see Data and Resources), from
and hypocentral distance to the center of the closest asperitiy
which subduction events have been extracted. Further, we
RA , among others. These were examined, quantified, and
added the study that describes the rupture process of the 25
scaled with respect to seismic moment M0 , using a regression
April 1989 M w 6.9 San Marcos event (Singh et al., 1989).
analysis method. Ⓔ The fault models for the earthquakes used
Table 2 presents the frequency range and type of data used
in this study and identification of asperities on them are avail-
on the kinematic inversion of earthquakes involved in this
able in the electronic supplement to this article.
study. With the exception of the study of the San Marcos earth-
quake (which used only strong ground motion records), slip
models were constructed by waveform inversions primarily Regression Analysis Method
from teleseismic data; there are 117 components from tele-
seismic data that represent 77.48% of data and 34 components The empirical relationships of source parameters ob-
from strong-motion data that represent 22.52% of data (Ta- tained by Somerville et al. (2002) are of the form
ble 2). In this study, we compare our results with those ob-
tained by Somerville et al. (2002). Table 3 presents the Y  α × Mβ0 ; 1
frequency range and type of data used for Somerville et al.’s
in which constants α and β are functions of the seismic mo-
kinematic inversion of earthquakes. As shown in Table 3, we
ment M0 in newton meters. This form is adopted in the
found that from the 10 slip models used by Somerville et al.
present paper, and constants α and β were determined by
(2002), and similarly to the tendency of data used in our study,
regression analysis of the obtained data (Table 4). For the
these were constructed primarily from teleseismic data. There
purpose of the regression analyses, it is useful to recast equa-
are 201 components from teleseismic data that represent
tion (1) to the form
84.10% of data and 38 components from strong-motion data
that represent 15.90% of data. The comparison between the
y  a  β × x; 2
percentages of teleseismic and strong ground motion data
used in both studies do not show significant differences. In in which y  lnY, a  lnα, and x  lnM 0 . The values
general in both studies, the quantity of teleseismic data is of a and slope β in equation (2) were determined by mini-
larger than strong-motion data. Also, Tables 2 and 3 show that, mizing the squared error between the available and computed
in both studies (with the exception of the study of the San Mar- seismic source parameters. Then α was back-calculated from
cos earthquake), the inversions are derived from ground- the relationship a  lnα. We use equation (3) to estimate
displacement waveforms in the 0–1.25 Hz frequency range. the standard error:
In addition to the seven studies shown in Table 1, there
s
Pn 
are two other studies available from Mexico: the kinematic
i1 yi − a  β xi 
2
inversions of the Michoacán earthquake of 11 January 1997 se  ; 3
(Santoyo et al., 2005) and the Oaxaca earthquake of 30 Sep- n − 2
844 A. Ramírez-Gaytán, J. Aguirre, M. A. Jaimes, and V. Huérfano

Table 2
Frequency Range and Type of Data Used in the Kinematic Inversion of Earthquakes Involved in This Study
Event Number Name of Earthquake Date (yyyy/mm/dd) References SGM* TS† Fmin (Hz) Fmax (Hz)

1 Petatlán 1979/03/14 Mendoza (1995) 0 15 0.01 0.2


2 Playa Azul 1981/10/25 Mendoza (1993) 0 15 0.00 0.2
3 Michoacán 1985/09/19 Mendoza and Hartzell (1989) 4 13 0.00 0.5
4 Zihuatanejo 1985/09/21 Mendoza (1993) 0 24 0.00 0.5
5 San Marcos 1989/04/25 Singh et al. (1989) 12 0 0.05 1.25
6 Manzanillo 1995/10/09 Mendoza and Hartzell (1999) 0 38 0.02 0.5
7 Tecomán 2003/01/22 Yagi et al. (2004) 18 12 0.01 0.5
Total 34 117 – –

All data, except that of the San Marcos earthquake (event 5) are provided from Martin Mai’s SRCMOD Database of Finite-Source
Rupture Models (see Data and Resources). Data for the San Marcos earthquake are from Singh et al. (1989).
*Number of strong ground motion component data used in the inversion.
†Number of teleseismic component data used in the inversion.

Table 3
Frequency Range and Type of Data Used for Kinematic Inversion of Earthquakes Described in Somerville et al. (2002)
Event Number Name Date (yyyy/mm/dd) References SGM* TS† Fmin (Hz) Fmax (Hz)

1‡ Kanto 1923/09/01 Wald and Somerville (1995) 0 6 0.00 0.1


2§ Tonankai 1944/12/07 Ichinose et al. (2003) 13 10 0.00 0.5
3‡,‖ Peru 1974/10/03 Hartzell and Langer (1993) 0 32 0.00 0.2
4‖ Peru 1974/10/09 Hartzell and Langer (1993) 0 38 – –
5‡ Petatlán 1979/03/14 Mendoza (1995) 0 15 0.01 0.2
6‡ Playa Azul 1981/10/25 Mendoza (1993) 0 15 0.00 0.2
7‡ Valparaiso 1985/03/03 Mendoza et al. (1994) 12 16 0.1 0.5
8‡ Michoacán 1985/09/19 Mendoza and Hartzell (1989) 4 13 0.00 0.5
9‡ Ziuathanejo 1985/09/21 Mendoza (1993) 0 24 0.00 0.5
10‡ Hokkaido 1993/07/12 Mendoza and Fukuyama (1996) 9 32 0.05 0.5
Total 38 201 – –

*Number of strong ground motion component data used in the inversion.



Number of teleseismic component data used in the inversion.

Information for events 1, 3, and 5–10 is provided from Martin Mai’s SRCMOD Database of Finite-Source Rupture Models (see Data and
Resources).
§
Information for event 2 (1944, Tonankai earthquake, Ms 8.01) is taken directly from Ichinose et al. (2003).

Information for event 3 (3 October 1974, Mw 8.0, Peru earthquake), is from Martin Mai’s SRCMOD Database of Finite-Source Rupture
Models (see Data and Resources). However, because information for its largest aftershock, event 4 (9 November 1974, M s 7.1) does not
appear in the database, the information is taken directly from Hartzell and Langer (1993)

in which yi  lnY i  is the observed value (with Y i as the does not vary as a function of rupture velocity. It will be
available data); a  βxi   lnα  β lnM oi  is the com- interesting to conduct future studies to compare aftershocks
puted value of the ith data point; and n is the total number and asperities defined by kinematic models for Mexican
of data points. This procedure leads to values of α and β for earthquakes. For the seven earthquakes studied here, it is dif-
equation (1) to represent the best fit, in the least-squared sense, ficult to estimate the area of the aftershock region because
of the available data. Thereby, we generated new source- there is not enough information available that indicates their
scaling relationships. After completing this process, the new locations. At minimum, this comparison will require accurate
relationships were compared with the worldwide large- location of aftershocks—information that, for some of the
magnitude (Mw 7.1–8.1) subduction earthquakes obtained earthquakes studied here, is not available.
by Somerville et al. (2002) and other published relationships.
In kinematic inversions, the area of asperities depends Results
on rupture velocity. The rupture velocities of the earthquakes By analyzing the properties of individual asperities, we
used here are listed in Table 1. These values show a narrow concluded that the number of asperities in the slip models
range from 2.6 to 3:5 km=s, with an average value of ranges from 1 to 4. A total of 13 asperities of the seven earth-
2:9 km=s. However, there is no evidence of the rupture ve- quakes were used here, with an average of 2.04 asperities.
locity’s dependence on seismic moment. On the other hand, The sum of the area for all asperities is only 19.57% of the
asperities of large earthquakes may be estimated within the overall rupture area. The largest asperity covers approxi-
first few days of aftershocks. The definition of aftershocks mately 13.9% of the overall rupture area.
Scaling Relationships of Source Parameters of Mw 6.9–8.1 Earthquakes 845

Table 4
Scaling Relationships Derived in This Study When Assuming Self-Similarity
Parameter Unit Results Analysis Type Best-Fit se *
−8
Rupture area (A) km 2
Unconstrained A  1:87 × 10 M0:570
0 0.568
Self-similar scaling with β  2=3 A  1:99 × 10−10 M 2=3
0 0.587
Somerville et al. (2002) A  2:41 × 10−10 M 2=3
0 0.630
Murotani et al. (2008) A  1:48 × 10−10 M 2=3
0 0.684

Strasser et al. (2010) log10 A  −3:476  0:952 × Mw 0.758
Blaser et al. (2010) log L  −2:69  0:64 × Mw ‡ 1.233
log W  −1:12  0:33 × M w ‡
Average slip (D) m Unconstrained D  8:45 × 10−3 M0:099
0 0.606
Self-similar scaling with β  1=3 D  1:43 × 10−7 M1=3
0 0.704
Somerville et al. (2002) D  1:14 × 10−7 M1=3
0 0.751
Murotani et al. (2008) D  1:48 × 10−7 M1=3
0 0.706
Combined area of asperities (Aa ) km 2
Unconstrained Aa  1:74 × 10−11 M0:684
0 0.669
Self-similar scaling with β  2=3 Aa  3:99 × 10−11 M2=3
0 0.667
Somerville et al. (2002) Aa  5:62 × 10−11 M2=3
0 0.839
Murotani et al. (2008) Aa  2:89 × 10−11 M2=3
0 0.708
Area of largest asperity (AL ) km 2
Unconstrained AL  7:78 × 10−9 M 0:550
0 0.652
Self-similar scaling with β  2=3 AL  3:25 × 10−11 M 2=3
0 0.676
Somerville et al. (2002) AL  4:12 × 10−11 M 2=3
0 0.732
Hypocentral distance to center km Unconstrained RA  2:74 × 10−3 M 0:177
0 0.870
of closest asperity (RA )
Self-similar scaling with β  1=3 RA  1:81 × 10−6 M 1=3
0 0.902
Somerville et al. (2002) RA  3:79 × 10−6 M 1=3
0 1.257
Hypocentral distance to center km Unconstrained Ra  2:41 × 10−5 M0:280
0 1.081
of largest asperity (Ra )
Self-similar scaling with β  1=3 Ra  2:00 × 10−6 M1=3
0 1.085
Overall slip duration (T s ) s Unconstrained T S  1:52 × 10−6 M0:309
0 1.161
Self-similar scaling with β  1=3 T S  4:80 × 10−7 M1=3
0 1.162
Somerville et al. (2002) T S  7:80 × 10−7 M1=3
0 1.296

*The standard error estimate for each regression, M 0 in newton meters.



The relationship in terms of moment magnitude obtained by assuming self-similar scaling. M w is an independent variable we have
converted to seismic moment by M w  2=3 log10 M 0  − 6:033 (M 0 is in newton meters; Hanks and Kanamori, 1979). Coefficient b is
forced to 1 to reflect direct proportionality between seismic moment and rupture area by assuming self-similar scaling (Strasser
et al., 2010).
‡Because Blaser et al. (2010) did not derive relationships for rupture area directly from the data, we simply multiply rupture length L and

width W, derived by them.

Figure 2 and Table 4 show the relationships between and


seismic moment and each seismic source parameter obtained D
in this study. For each parameter, the unconstrained equation  k2 : 5
L
(Fig. 2, black continuous lines) is shown first, followed by
the equation that imposes self-similarity (Fig. 2, gray con- So in equation (6), we rewrite the scalar seismic moment as
tinuous lines). The scaling self-similarity forces the coeffi-
cient β in equation (1) to reflect direct proportionality M0  μ × D
 × W × L  μ × k 1 × k 2 × L3 : 6
between seismic moment and source parameters as described
From this, M 0 ∝ L3 (Lay and Wallace, 1995), and, therefore,
after equation (6).
M01=3 ∝ L, as D is proportional to M 1=3
0 and the surface area
The scalar seismic moment is defined as the product of
that includes two dimensions is proportional to M 02=3 . This
the fault surface area, average displacement on the fault, and
condition is also known as the self-similar source model.
the rigidity of the rock. The assumption of constant stress
The self-similar model is convenient to use because it
drop leads to the static similarity condition, constant fault
provides a reasonably good description of the phenomena
aspect ratio, and constant strain (equations 4 and 5):
(Somerville et al., 1999). In this analysis of large-magnitude
Mexican subduction earthquakes, we found that scaling the
W fault parameters with the seismic moment fits reasonably
 k1 ; 4
L well with a self-similar scaling model. For the case of the
846 A. Ramírez-Gaytán, J. Aguirre, M. A. Jaimes, and V. Huérfano

(a) 100000 (b) 10

Rupture Area (km2)

Average Slip (m)


10000 1

1000 0.1

(c) 10000 (d) 10000


Combined Area of Asperities (km2)

Area of Largest Asperity (km2)


1000 1000

100 100

(e) 100 (f) 1000


Hypocentral Distance to Center of

Hypocentral Distance to Center of


Largest Asperities (km)
Closest Asperities (km)

100

10

10

1 1
1×1019 1×1020 1×1021 1×1022
(g) 100 Seismic Moment (N •m)
Overall Slip Duration (s)

10

0.1
1×1019 1×1020 1×1021 1×1022
Seismic Moment (N •m)

Figure 2. Regression results of the source-scaling relations of this study. In all cases (a–g), we show the relation between seismic moment
versus the following: (a) rupture area, (b) average slip, (c) combined area of asperities, (d) area of largest asperity, (e) hypocentral distance to
center of closest asperity, (f) hypocentral distance to center of largest asperity, and (g) overall slip duration. In all cases, filled circles indicate
the events, black continuous lines are the results of this study (unconstrained), gray lines are the fit when self-similar scaling is assumed, and
dashed lines are the 95% confidence intervals for the mean when self-similar scaling is assumed.

relationship of average slip versus seismic moment, the un- 1. unconstrained regression analysis to estimate α and β;
constrained relationship suggests a non-self-similar scaling. 2. self-similar scaling regressions to estimate α while the
For this case, the physical interpretation suggests the absence value of β is forced to be 2=3 (for rupture area, combined
of a scale factor in the average slip of Mexican subduction area of asperities, and area of largest asperity) or 1=3 (for
earthquakes. However, as can be appreciated from Figure 2b, average slip, hypocentral distance to the center of the
the data are dispersed and do not show a clear tendency. closest asperity, hypocentral distance to the center of the
Note that the standard deviation (Table 4) is smaller than that largest asperity, and overall slip duration);
of Somerville et al. (2002), even for the self-similar as- 3. self-similar scaling regression with values of α and β that
sumption. The results shown in Table 4 follow the regression were fixed according to Somerville et al. (2002) to deter-
analysis given by equation (1) for: mine the standard error se (values of α and β from
Scaling Relationships of Source Parameters of Mw 6.9–8.1 Earthquakes 847

Somerville et al. were computed to be compatible with D  1:43 × 10−5 M 1=3


0 : 10
units used in this work); and
4. standard errors were estimated for comparison with other From the self-similar scaling equations provided by
publications of source scaling for subduction zone earth- Somerville et al. (2002) and Murotani et al. (2008), se is
quakes (Murotani et al., 2008; Blaser et al., 2010; 0.751, and 0.706, respectively; these values are greater than
Strasser et al., 2010). se was determined by computing those found by the constrained equation that estimates the
the mean square error between the available data and average slip of large-magnitude Mexican subduction earth-
computed values from the relationships. Relationships quakes (se  0:704). These results show a close resemblance
expressed in terms of moment magnitude, as the indepen- between the relationship derived in this study and those pro-
dent variable, have been converted to seismic moment by posed by Murotani et al. (2008). The scaling law of com-
Mw  2=3 log10 M 0  − 6:033 (M 0 is in newton meters; bined area of asperities Aa (km2 ) and seismic moment M0
Hanks and Kanamori, 1979). (N·m) determined without constraining the slope is (Fig. 2c)
These regression analyses, which use the data from all Aa  1:74 × 10−11 M 0:684 : 11
0
earthquakes (Table 1), lead to the equations and data pro-
vided in Table 4. Figure 2 provides a visual representation By self-similar scaling of the slope to be 2=3, the relationship
for each equation obtained from each regression analysis: is
a dark continuous line shows the unconstrained relationship,
gray lines are the fit when self-similar scaling is assumed Aa  3:99 × 10−11 M2=3
0 : 12
together with the available data (filled circles), and dashed
lines correspond to the 95% confidence intervals for the As shown in Figure 2c, the self-similar scaling equation
mean when self-similar scaling is assumed for each seismic indicates that the estimated combined area of asperities of
source parameter. As stated previously, Figure 2 provides a large-magnitude Mexican subduction-zone earthquakes
graphical view of data scattering of the seismic source shows a clear dependence on seismic moment. The relation-
parameters relative to curves from regression analyses. As ships in equations (8), (10), and (12) provide smaller values
expected, self-similar scaling regression generally implies than the self-similar scaling equation provided by Somerville
large error of estimated se (Table 4), showing larger scatter et al. (2002). In this study, the area covered by asperities is
of the data about the best-fit curve. 19.4% of the total rupture area, which is smaller than the
25% proposed by Somerville et al. (2002). Table 4 shows
that, for the constrained equation provided by Somerville
Source Scaling et al. (2002) and Murotani et al. (2008), the se values are
0.839 and 0.708, respectively. These results are larger than
The relationship between rupture area A (km2 ) and seis-
those found by the constrained equation that estimates the
mic moment M0 (N·m) determined without constraining the
combined area of asperities of large-magnitude Mexican sub-
slope is (Fig. 2a)
duction earthquakes (se  0:667). The relationship between
A  1:87 × 10−8 M0:570
0 : 7 area of largest asperity AL (km2 ) and seismic moment M0
(N·m) determined without constraining the slope is (Fig. 2d)
By self-similar scaling of the slope to be 2=3, the relationship
is AL  7:78 × 10−9 M 0:55 13
0 :
A  1:99 × 10−10 M 02=3 : 8 By self-similar scaling of the slope to be 2=3, the relationship
is
From the self-similar scaling equations (see Fig. 3a and
Table 4) provided by Somerville et al. (2002), Murotani et al. AL  3:25 × 10−11 M 02=3 : 14
(2008), Strasser et al. (2010), and Blaser et al. (2010), se val-
ues are 0.630, 0.684, 0.758, and 1.233, respectively. In all As shown in Figure 2d, the comparison of both con-
four cases, these values are greater than estimated by the strained equations indicates that the estimated area of the
self-similar scaling equation with respect to the estimated largest asperity of large-magnitude Mexican subduction
rupture area of large-magnitude Mexican subduction earth- earthquakes is smaller than the constrained equation pro-
quakes (se  0:587). These results are examined in the Dis- vided by Somerville et al. (2002). The area of the largest
cussion section. The relationship between average slip D (m) asperity for the two relationships compared in this study
and seismic moment M0 (N·m) determined without con- shows a clear dependence on seismic moment. The average
straining the slope is (Fig. 2b) value for the area of largest asperity is 13.79% of the total
D  8:45 × 10−1 M0:099 : 9 rupture area versus the 17.5% proposed by Somerville et al.
0
(2002). Table 4 shows that, for the self-similar scaling equa-
By self-similar scaling of the slope to be 1=3, the relationship tion provided by Somerville et al. (2002), the error estimate
is (se  0:732) is greater than that found by the self-similar
848 A. Ramírez-Gaytán, J. Aguirre, M. A. Jaimes, and V. Huérfano

(a) 100000 10
(b)

Rupture Area (km2)

Average Slip (m)


10000

1
This study
1000 Somerville et al. (2002)
Murotani et al. (2008) This study
Strasser et al. (2010) Somerville et al. (2002)
Blaser et al. (2010) Murotani et al. (2008)
100 0.1

(c) 10000 10000


Combined Area of Asperities (km2)

(d)

Area of Largest Asperity (km2)


1000 1000

This study
Somerville et al. (2002)
Murotani et al. (2008)
100 100

(e) 100 1000

Hypocentral Distance to Center of


Hypocentral Distance to Center of

(f)

Largest Asperities (km)


Closest Asperities (km)

100

10

10

1 1
1×1019 1×1020 1×1021 1×1022
(g) 1000 Seismic Moment (N •m)
Overall Slip Duration (s)

100

10

0.1
1×1019 1×1020 1×1021 1×1022
Seismic Moment (N •m)

Figure 3. Comparison of scaling relations of this study with existing relations. In all cases (a–g) we show the relation between seismic
moments versus the following: (a) rupture area, (b) average slip, (c) combined area of asperities, (d) area of largest asperity, (e) hypocentral
distance to center of closest asperity, (f) hypocentral distance to center of largest asperity, and (g) overall slip duration. In all cases, gray lines
are the results of this study, dotted lines are from Somerville et al. (2002), dashed-dotted lines are from Murotani et al. (2008), and dotted
lines are from Strasser et al. (2010).

scaling equation that estimates the area of largest asperities RA  1:81 × 10−6 M 1=3
0 : 16
of large-magnitude Mexican subduction earthquakes
(se  0:676). The scaling law that relates hypocentral distance
to the center of the closest asperity RA (km) and seismic mo- Table 4 shows that, for the self-similar scaling equation
ment M 0 (N·m) determined without constraining the slope is provided by Somerville et al. (2002), the error estimate
(Fig. 2e) (se  1:257) is greater than that found by the constrained
equation that estimates the hypocentral distance to the center
of the closest asperity for large-magnitude Mexican subduc-
RA  2:47 × 10−3 M 0:177
0 : 15 tion earthquakes (se  0:902). This can also be observed in
By self-similar scaling of the slope to be 1=3, the relationship is Figure 2e, in which the dashed lines corresponding to the
Scaling Relationships of Source Parameters of Mw 6.9–8.1 Earthquakes 849

10000
mean 95% confidence intervals when self-similar scaling Source area estimated by Aguirre & Irikura (2007)
Rupture area estimated by Somerville et al. (2002)
is assumed. This study: Rupture area
This study: Area of largest asperity
In this study, we compared the new relationships pro- 1000 This study: Combined area of asperities
posed in the present paper with the areas of 31 Mexican sub-
duction earthquakes estimated by Aguirre and Irikura (2007).
100

Area (km2)
One item to note is that these areas are generated by a process
that considered high frequencies, while our relationships that
involve the rupture area, area of largest asperity, and com-
10
bined area of asperities were determined primarily from
sources that consider low-frequency data. Aguirre and Iri-
kura (2007) used the flat level of acceleration spectra records 1
for high frequencies from 31 earthquakes recorded by the
Guerrero Mexico accelerographic array from 1985 to
1998. The 31 earthquakes belong to the subduction earth- 0.1
1×1015 1×1016 1×1017 1×1018 1×1019 1×1020 1×1021
quakes (those occurring on the plate interface). From the cor-
Seismic Moment (N •m)
ner frequency of the source spectra for the acceleration
records, the authors estimated the corresponding source area Figure 4. Relationships between seismic moment versus total
and compare their results with the area of the largest asperity rupture area, area of largest asperity, and combined area of asper-
provided by Somerville et al. (2002). In Figure 4, we show ities. The relationship between seismic moment and total rupture
our results for this comparison. area generated by Somerville et al. (2002) is represented by the dot-
ted line, the relationship between seismic moment and rupture area
obtained in this study is represented by the continuous line, the re-
Relationships between Stress Drop and Depth lationship between seismic moment and the area of largest asperity
obtained in this study is represented by the dashed-dotted line, the
Because the stress drop is a phenomenon related to the relationship between seismic moment and combined area of asper-
level of the seismic-wave radiation, which may vary as a ities obtained in this study is represented by the dashed line, and the
function of depth, we estimated the average stress drop of areas estimated by the corner frequencies of source spectra of ac-
individual asperities. For the seven earthquakes (kinematic celeration records estimated by Aguirre and Irikura (2007) are rep-
resented by gray circles.
studies) considered here, we estimate the average stress drop
of each one of the 13 asperities within the rupture area, as
well as the small stress drop of the surrounding area. The As mentioned in the Results section, for the relationship de-
static stress drop of each asperity is estimated from equa- rived in this study, the error estimate (se  0:587) is smaller
tion (17), following Madariaga (1979), when a circular than those relationships compared here, which are 0.630 for
source fault can be assumed (Boatwright, 1988). Somerville et al. (2002), 0.684 for Murotani et al. (2008),
 0.758 for Strasser et al. (2010), and 1.233 for Blaser et al.
7 M (2010).
Δσ a  × 2 0 ; 17
16 r ×R The results from this study show a close resemblance to
the relationships proposed by Somerville et al. (2002) and
in which Δσ a is the static stress drop on asperity in MPa, r is Murotani et al. (2008). The closest resemblance with Somer-
the equivalent asperity radius in kilometers, R is the equiv- ville et al. (2002) could be explained because four of the
alent radius of the source faults in kilometers, and M 0 is in source models, representing 40% of the data considered in
newton meters. The stress drop of the 13 asperities consid- the regression of Somerville et al. (2002), are derived from
ered here (Table 5) ranged from 0.504 to 3.284 MPa. There is the Mexican subduction zone; those models are considered
not a clear dependency of the stress drop of the asperities in our study and represent 57% of our data. Also, our results
with the depth in our dataset. highlight that there is a clear offset of a factor of ∼2 in rup-
ture area between our relationships and those obtained by
Discussion Blaser et al. (2010). A similar offset is obtained by Blaser
et al. (2010) when comparing their relationships that involve
Comparison with Recent Source-Scaling
rupture area with the result derived by Murotani et al. (2008).
Relationships
The difference between the results of this study and those
From analysis of Figure 3a–g, in which the results of this proposed by Blaser et al. (2010) could be due to two factors:
study are compared with four different source-scaling rela- (1) our study focuses only on Mexican events, whereas they
tionships for subduction events, it is apparent that our results used worldwide data, and (2) the method used in our study to
involving rupture area (Fig. 3a) show a similar tendency to estimate the rupture areas of the source is based only on slip
those obtained by Strasser et al. (2010), Blaser et al. (2010), distribution, unlike that of Strasser et al. (2010) and Blaser
and Murotani et al. (2008) in the sense that they generate et al. (2010), who used earthquakes from different tectonic
smaller areas than predicted by Somerville et al. (2002). environments in the world and some rupture areas that are
850 A. Ramírez-Gaytán, J. Aguirre, M. A. Jaimes, and V. Huérfano

Table 5
Stress Drop of All Asperities from Earthquakes Used in This Study and Depth Measured at the Center
of the Asperity
Event Number Name of Earthquake Date (yyyy/mm/dd) Number of Asperity Depth (km) Stress Drop (MPa)

1 Petatlán 1979/03/14 1 11.685 0.558


2 29.830 0.504
2 Playa Azul 1981/10/25 1 16.330 3.049
3 Michoacán 1985/09/19 1 15.247 1.577
2 18.610 1.442
4 Zihuatanejo 1985/09/21 1 20.005 3.284
5 San Marcos 1989/04/25 1 16.845 1.066
6 Manzanillo 1995/10/09 1 5.080 2.728
2 24.430 2.258
3 13.545 2.159
4 5.079 1.984
7 Tecomán 2003/01/22 1 18.125 2.971
2 29.365 3.046

based in the aftershock distribution. As explained by Blaser equation of large-magnitude Mexican subduction earth-
et al. (2010), this evaluation could involve subjective quakes is smaller than the self-similar scaling equation pro-
judgment and other sources of uncertainties. vided by Somerville et al. (2002). This can also be observed
The predicted areas generated by Murotani et al. (2008) in Figure 2a–g, with the dashed lines corresponding to the
that involve rupture area and combined area of asperities, like mean 95% confidence intervals when self-similar scaling
the corresponding relationships generated in this study, show is assumed. This result suggests that the expressions obtained
the same tendency as our results in the sense that they gener- in this study could be appropriate for simulation of strong
ated smaller areas than those obtained by Somerville et al. ground motions only in the case of the Mexican subduc-
(2002), as shown in Figure 3a,c. One reason for this behavior tion-zone environment, although it is necessary to take into
is that both relationships (Murotani et al., 2008; and the results account that the database used in this study is limited.
of this study) are derived from a similar process Irikura and Miyake (2011) define the rupture area and
(Somerville et al., 1999) to determine the area of asperities seismic moment as outer fault parameters and the slip hetero-
and rupture area, unlike the other relationships compared here geneities and combined area of asperities as inner fault
(Blaser et al., 2010; Strasser et al., 2010). Despite the fact that parameters. Figure 2a–d show that the scaling laws that in-
Murotani et al. (2008) and our results show a similar tendency, volve areas (inner and outer fault parameters) in general
the comparison highlights an important difference, because show a greater dependence with respect to seismic moment
our results generated larger areas than those proposed by than other relationships studied here (Fig. 2b–e). In the case
Murotani et al. (2008). There is not a clear explanation for of average slip, this dependence is less. However, Figure 2b
this behavior; however, it is important to note that both rela- shows that, with the exception of Petatlán and San Marcos
tionships are derived exclusively from specific subduction re-
earthquakes, the average slip of the remaining five earth-
gions (Japan and Mexico) and that one of the justifications of
quakes shows a dependence on seismic moment. The basic
these types of studies is to know whether or not this relation-
difference between constrained and unconstrained relation-
ship varies depending on a specific subduction tectonic
ships that relate the average slip versus seismic moment
region. Figure 3b shows that predicted average slip by Mur-
(equations 9 and 10; Fig. 2b) comes from the San Marcos
otani et al. (2008) is very similar to our results because both
earthquake event (25 April 1989). Interestingly, this event
lines are almost superimposed. Additionally, both lines show
only shows a different tendency in the slip versus seismic
the same tendency in the sense that both generated higher
moment, but not in the other relationships studied here.
values than those obtained by Somerville et al. (2002).
Particularly, relationships that relate the combined area of
asperities versus seismic moment (with both constrained
Comparison with Somerville et al. (2002) and unconstrained equations; equations 11 and 12) are very
As shown in Figure 3a–g from the source-scaling rela- similar. Additionally in this case, the point that represents the
tionships determined in this study, for most source parame- San Marcos event shows the best agreement with both ten-
ters studied here, our results present smaller values than dency lines (constrained and unconstrained). What can pro-
those obtained by Somerville et al. (2002), with the only ex- duce this behavior? Let us examine this in detail: (1) This
ception being the relationship that involves the average slip event is the only one that has not been taken from Martin
versus seismic moment. The se for each regression, summa- Mai’s SRCMOD Database of Finite-Source Rupture Models
rized in Table 4, suggests that in general, for all source (see Data and Resources). (2) In contrast to the other kinematic
parameters studied here, the se from the self-similar scaling inversions used in this study that mainly use teleseismic data,
Scaling Relationships of Source Parameters of Mw 6.9–8.1 Earthquakes 851

this event has been inverted using only strong ground motion (Ramírez-Gaytán et al., 2010). This fitting slowly improves
data (see Table 2). (3) The published article that studied the when an asperity approaches the hypocenter. In this case, the
rupture process of the San Marcos earthquake (Singh et al., location of the nearest asperity to the hypocenter is an im-
1989) reports only the slip distribution with arrows of different portant factor when dealing with strong ground motion sim-
sizes corresponding to the displacement in each subfault. Be- ulations. In this study, our result is 50.23% smaller than that
cause the discretization slip values were obtained from mea- proposed by Somerville et al. (2002) and in this particular
surements of the slip lines from the printed figures, this case may have a direct implication of bringing the asperity
could produce some uncertainties in our slip distribution data. closer to the hypocenter. This suggests that the nucleation
However, from Figure 2c, it is interesting to observe that the process of some earthquakes probably starts at the location
self-similarity assumption is almost the same as the uncon- of an asperity.
strained relation. One point that contributes to this behavior
is the well-confined area of asperities from the San Marcos
Comparison with Mexican Subduction Earthquakes
event. That means that even the absolute values of displace-
(High Frequencies)
ment assigned to each subfault could be inaccurate; the contrast
between them allows us to determine the combined area of Aguirre and Irikura (2007) conducted a study to estimate
asperities appropriately. the rupture areas of 31 earthquakes from the Mexican sub-
The scaling laws of seismic moment M0 versus total duction zone. The process considers the corner frequencies
area of rupture and average slip, Figure 2a and 2b, respec- of source spectra of acceleration records. In their study, they
tively, hold a relationship in the sense that the concept of the plot the source areas obtained with the process described
fault rupture area reveals the zone of the fault that radiates above and note that these areas fit well with the relationship
seismic energy, and this necessarily implies that this relation- proposed by Somerville et al. (2002) between the seismic
ship has a direct implication with average slip. Our results moment and the area of the largest asperity for subduction
show that only the case of the average slip relationship ob- earthquakes. They conclude that the estimated areas were
tained in this study (Fig. 2b) is larger than the relationship not the total rupture areas of earthquake. In Figure 4, we plot
proposed by Somerville et al. (2002). The relationship be- the same areas estimated by Aguirre and Irikura (2007) for
tween seismic moment versus total rupture area and seismic the 31 Mexican subduction-zone earthquakes, but now com-
moment versus average slip for large-magnitude Mexican pared with the results of the present study (rupture area, com-
subduction zone earthquakes (Figs. 2a and 2b) might con- bined area of asperities, area of largest asperities) and with
tribute to the explanation of small rupture areas obtained by the rupture area determined by Somerville et al. (2002).
Aguirre (2005), Garduño (2006), and Ramírez-Gaytán et al. The relationships between seismic moment and rupture
(2010). If seismic moment is the result of the product of area, area determined by Somerville et al. (2002) in Figure 4 are
average slip, and rock shear modulus, then a decrease in the represented with a dotted line, and the same relationship ob-
area of seismic source and combined area of asperities im- tained from the Mexican subduction earthquakes obtained in
plies a necessary increase of average slip in order to maintain this study is represented with a solid line (equation 8). In
a similar seismic moment. both cases, relationships lie considerably above the estimated
In this study, the area covered by asperities is 19.4% of areas of 31 earthquakes, confirming the conclusion obtained
the total rupture area, which is smaller than the 25% pro- by Aguirre and Irikura (2007) that the areas estimated for 31
posed by Somerville et al. (2002). The average value for earthquakes did not correspond to the total rupture areas.
the area of the largest asperity is 13.79% of the total rupture This could be explained because areas of high-frequency ra-
area, which is smaller than the 17.5% proposed by Somer- diation mainly correspond to regions adjacent to the large
ville et al. (2002). Both results are congruent with the two slip velocity area (Miyake et al., 2003). This means that the
relationships described above in the sense that a reduction in determined source area of these 31 earthquakes mainly re-
area implies a larger slip in order to maintain a similar seis- flects the area of asperities. Then, while using the rectangular
mic moment. Additionally, the comparison of the combined representation of the fault rupture from kinematic inversions
area of asperities versus area of largest asperity suggests that and source-scaling relationships, the asperities represent only
the largest asperity is responsible for releasing a large a fraction of the rupture area (asperities represent 22% of the
amount of energy during an earthquake. source area in the case of Somerville et al. [1999]). The
By comparing both constrained equations (see Fig. 2e) remaining 78% of the source area considered in this source-
of the estimated hypocentral distance with the closest asper- scaling relationship is not reflected by the source areas esti-
ity of large-magnitude Mexican subduction earthquakes, we mated using acceleration spectra.
observe that it is 50.23% smaller than the self-similar scaling The dashed-dotted line (equation 14) in Figure 4 repre-
equation provided by Somerville et al. (2002). During the sents the relationship of seismic moment versus area of largest
process of simulating strong ground motions, we found that asperity. This line gives a best-fit and is located inside a cloud
the location of an asperity, inside of a seismic source at large of gray circles that represents the estimated areas of 31
distance from the hypocenter, in most cases generates a poor earthquakes. However, 60% of these circles are located above
fitting between the synthetic and the observed motions the line. As mentioned previously, this can be explained
852 A. Ramírez-Gaytán, J. Aguirre, M. A. Jaimes, and V. Huérfano

because the circles that represent the source areas of 31 ments. Concerning the relationships for the rupture area,
earthquakes primarily reflect the area of asperities, meaning our results are closer to those of Somerville et al. (2002) than
the total area of asperities. On the contrary, the trend line any other, but, like the others, they have smaller areas than
(dashed-dotted line) given here represents only the area of predicted by Somerville et al. (2002). This resemblance
the largest asperity. This relationship excludes the area of could be because 40% of the source data models considered
small asperities that contributes to the generation of high in Somerville et al. (2002) are derived from the Mexican sub-
frequencies. This could be the reason why the trend line duction zone and represent 57% of our data. In contrast, one
(dashed-dotted line) is located slightly below the center of the of the reasons for the difference with those relationships, pro-
clustered circles. The dashed line (equation 12) of Figure 4 posed by Strasser et al. (2010) and Blaser et al. (2010), could
represents the relationships of seismic moment versus com- be because our study focuses only on Mexican events,
bined area of asperities located near the center of the clustered whereas they use worldwide data. Another reason could be
circles, showing the best-fit of the four relationships com- because, for some earthquakes, the criteria to estimate the
pared here. As mentioned, this may explain why the circles rupture area used in their studies are based on the aftershock
representing the source areas of 31 earthquakes reflect the to- distribution. That is different than our study, which is based
tal area of asperities estimated through a process that consid- only on slip distribution. The criteria based on aftershock
ers high frequencies, and the trend line determined here in fact distribution could involve subjective judgment and other
represents the combined area of asperities generated through a sources of uncertainties (Blasser et al., 2010).
process that considers low frequencies. Because rupture areas from Murotani et al. (2008) are
This result is in accordance with that stated by Irikura derived from the same process applied in the present study
and Miyake (2011) in the sense that the flat level of the ac- (Somerville et al., 1999), their relationship also predicts
celeration spectra was theoretically related to the combined larger values than Strasser et al. (2010) and Blaser et al.
asperity areas. Based on these considerations, an important (2010), but smaller values than ours. There is not a clear ex-
implication is that two different methodologies to determine planation for this behavior; however, it is important to note
the source area provide similar results. Namely, these meth- that both relationships are derived exclusively from specific
odologies are (1) the source-scaling relationship to determine subduction regions (Japan and Mexico) and that one justifi-
the combined area of asperities from kinematic inversions cation of this type of study is to know whether or not this
(based on low frequencies) and (2) the area of combined relationship varies depending on a specific region.
asperities estimated by the corner frequencies (based on high Concerning the combined area of asperities, both Mur-
frequencies). otani et al. (2008) and our results predict smaller areas than
those obtained by Somerville et al. (2002). However, both
Stress Drop relationships highlight an important difference, because
our results generated larger areas than those proposed by
Strong dependence of stress drop when depth increases Murotani et al. (2008), which, as mentioned could be related
may indicate that the asperity areas could take smaller values to the specific subduction regions (Japan and Mexico).
than predicted by scaling relationships; that is, when the Concerning the area of largest asperity, the relationships
areas are located at larger depths within the rupture area. obtained in this study are slightly smaller than those de-
We calculate the stress drop of the 13 asperities considered scribed by Somerville et al. (2002), which is consistent with
here and found that they ranged from 0.504 to 3.284 MPa the relationships of total rupture area and combined area of
(Table 5). There is not a clear dependence of the stress drop asperities. This is important because they are involved in the
of those asperities with the depth in our dataset. Scholz source models used for strong ground motion simulation. In
(1982, 2002) stated that shear strength of the lithosphere in- the case of the average slip, our relationship is slightly larger
creases with depth. So, we expected that the deeper asperities than Somerville et al. (2002). That is an expected result be-
could have associated a larger stress drop. However, the data- cause a decrease in the area of the seismic source implies an
set used here, shows a null dependence of stress drop with increase of average slip in order to maintain the same seismic
the depth. Moya et al. (2000) estimated the stress drop for the moment.
aftershocks of the Kobe earthquake and compared the stress The error estimates for the constrained equations derived
drop with depth; their results show a similar trend with the in this study in all cases are smaller than those relationships
one obtained in this study, that is, although a slight indication compared here. This might suggest that the expressions ob-
of dependency can be recognized, in general the dependence tained in this study could be appropriate for the simulations
of stress drop with depth is very scattered. of strong ground motion for the specific scenario of earth-
quake slip on the Mexican subduction-zone megathrust be-
Conclusions tween the Cocos–Rivera and the North American plates. An
important implication of this result is that the source-scaling
We obtain source-scaling relationships for Mexican sub- relationships vary depending on specific regions. However,
duction events. These relations were compared with those the improvement of these relationships requires more infor-
proposed by four different authors for subduction environ- mation to become available.
Scaling Relationships of Source Parameters of Mw 6.9–8.1 Earthquakes 853

For the case of relationships that involve the seismic mo- Data and Resources
ment versus combined area of asperities, the small difference
between the results obtained by Somerville et al. (2002) and The slip models and slip distribution from the earth-
the result obtained in this study could only partially explain quakes used in this study are available from the SRCMOD
the result obtained by Aguirre (2005), Garduño (2006), and —Database of Finite-Source Rupture Models (http://equake
Ramírez-Gaytán et al. (2010). ‑rc.info/SRCMOD/searchmodels/allevents/; last accessed 30
From the comparison between the source areas of 31 July 2013). The Global Centroid Moment Tensor (CMT)
Mexican earthquakes estimated by Aguirre and Irikura project catalog (1976–2013) is available at www.globalcmt.
(2007) and the relationships generated in this study, the fol- org (last accessed 13 March 2010). The seven published slip
lowing are observed. (1) The relationships between seismic models of earthquakes used in this study are listed in the refer-
moment and rupture area determined by Somerville et al. ences. The ten published slip models of earthquakes used for
(2002) and the same relationship from Mexican subduction Somerville et al. (2002) are listed in the references. Maps were
earthquakes obtained in this study lie considerably above the created using ArcGIS version 9.3. Graphics and plots were
estimated areas of 31 earthquakes, confirming the conclusion created using MATLAB version 7.0.
obtained by Aguirre and Irikura (2007) in the sense that the
areas estimated for 31 earthquakes were not corresponding to Acknowledgments
the total rupture areas. (2) The relationships of seismic mo-
ment versus combined area of asperities generated in this The authors wish to thank Kojiro Irikura and one anonymous reviewer
whose comments helped to significantly improve the original manuscript.
study located near the center of the clustered circles obtained We thank Associate Editor Heather DeShon for her important comments that
by Aguirre and Irikura (2007) shows the best-fit of the four have led to significant improvements to this manuscript. This study was con-
relationships compared here. This result is in accordance siderably facilitated by the SRCMOD database compiled by Martin Mai and
with what is stated by Irikura and Miyake (2011) in the sense coworkers, who provided the finite fault-slip models from the Finite-Source
that acceleration level was theoretically related to the com- Rupture Model Database, and by the willingness of Carlos Mendoza, the
principal author of the models used in this study. We are grateful to Ramón
bined asperity areas. Based on these results, an important Zúñiga of the Instituto de Geociencias of the Universidad Nacional Autón-
implication is that two different methodologies to determine oma de México for providing additional data and Evan Hirakawa of Uni-
the total area of asperities show similar results. To reiterate, versity of California in San Diego/San Diego State University for
these methods are (1) the source-scaling relationship from valuable comments and review. This research was supported by the Consejo
kinematic inversions to determine the combined area of Nacional de Ciencia y Tecnología (CONACyT) under Grant Number
164501.
asperities (based on low-frequency data) and (2) the area of
Mexican earthquakes estimated by corner frequencies (based
on high-frequency data). Finally, the stress drop of asperities References
from our dataset does not show a clear dependence with
the depth. Aguirre, J. (2005). Escenarios sísmicos a través de la caracterización de las
fuentes sísmicas en México, Mem. XV Congreso Nacional de Ingen-
Although the seven slip models considered in this study iería Sísmica, Sociedad Mexicana de Ingeniería Sismica (Editor),
are the only available studies for the Mexican subduction México D.F. (available only in CD).
zone at this time, in the future it will be important to incor- Aguirre, J., and K. Irikura (2007). Source characterization of Mexican
porate new slip models in the simulation of strong ground subduction earthquakes from acceleration source spectra for the
prediction of strong ground motions, Bull. Seismol. Soc. Am. 97,
motions. A strong dependence of stress drop on increasing
1960–1969.
depth could imply that the asperity areas could take smaller Asano, K., and T. Iwata (2011). Characterization of stress drops on asperities
values than predicted by scaling relationships; that is, when estimated from the heterogeneous kinematic slip model for strong
the areas are located at larger depths within the rupture area. motion prediction for inland crustal earthquakes in Japan, Pure Appl.
The relationships between the stress drop and the depth of Geophys. 168, 105–116.
Blaser, L., F. Krüger, M. Ohrnberger, and F. Scherbaum (2010). Scaling re-
the asperity could then be used in the prediction of strong-
lations of earthquake source parameter estimates with special focus on
motion scenarios of future large magnitude earthquakes oc- subduction environment, Bull. Seismol. Soc. Am. 100, 2914–2926.
curring in the Mexican subduction zone. Boatwright, J. (1988). The seismic radiation from composite model of
The slip models considered in this study are the only faulting, Bull. Seismol. Soc. Am. 78, 489–598.
available studies for the Mexican subduction zone at this Castro, R., and E. Ruiz-Cruz (2005). Stochastic modeling of the September
time. In the near future, incorporation of new slip models in 1999 Mw 7.5 earthquake, Oaxaca, Mexico, Bull. Seismol. Soc. Am. 96,
2259–2271.
the simulation of strong ground motions will be an important Garduño, N. (2006). Procesos de la fuente del sismo del 15 de Julio de 1996
contribution. Also, a future task will be to determine how the usando el método de la función de Green empírica y algoritmos
relationships vary for specific regions. Further work in de- genéticos, Bachelor’s Thesis, Facultad de Ingeniería, Universidad
veloping these scaling relationships will make it possible to Nacional Autónoma de México, 11 pp. (in Spanish).
Hanks, T., and H. Kanamori (1979). Moment magnitude scale, J. Geophys.
more accurately predict and simulate strong ground motions
Res. 84, 2348–2350.
produced by large-magnitude earthquakes from specific Hartzell, S. H., and C. Langer (1993). Importance of model parameterization
seismogenic zones in close proximity to highly populated in finite-fault inversions: Application to the 1974 Mw 8.0 Peru
urban areas. earthquake, J. Geophys. Res. 98, 22,123–22,134.
854 A. Ramírez-Gaytán, J. Aguirre, M. A. Jaimes, and V. Huérfano

Ichinose, G. A., H. K. Thio, P. G. Somerville, T. Sato, and T. Ishii (2003). Santoyo, M. A., S. K. Singh, and T. Mikumo (2005). Source process and
Rupture process of the 1944 Tonankai earthquake (M s 8.1) from the stress change associated with the 11 January, 1997 (M w 7.1)
inversion of telesieismic and regional seismograms, J. Geophys. Res. Michoacán, Mexico, inslab earthquake, Geofis. Int. 44, 317–330.
108, 13-1–13-21. Scholz, C. H. (1982). Scaling laws for large earthquakes: Consequences for
Irikura, K. (1986). Prediction of strong accelerations motions using empiri- physical models, Bull. Seismol. Soc. Am. 72, 1–14.
cal Green’s function, 7th Japan Earthq. Eng. Symp., 10 December Scholz, C. H. (2002). The Mechanics of Earthquakes and Faulting, Second
1986, 151–156. Ed., Cambridge University Press, Cambridge, United Kingdom,
Irikura, K., and H. Miyake (2011). Recipe for predicting strong ground 463 pp.
motion from crustal earthquake scenarios, Pure Appl. Geophys. Singh, S. K., M. Ordaz, R. Quaas, and E. Mena (1989). Estudio preliminar
168, 85–104. de la fuente del temblor del 25 de abril de 1989 (Ms  6:9) a partir de
Kanamori, H., and D. L. Anderson (1975). Theoretical basis of some em- movimientos fuertes, Memorias del VIII Congreso Nacional de Ingen-
pirical relations in seismology, Bull. Seismol. Soc. Am. 65, 1073–1095. iería Sísmica, Acapulco, Vol. I, A-199–A-205.
Lavallée, D., and R. J. Archuleta (2003). Stochastic modeling of slip spatial Somerville, P., N. Collins, T. Sato, T. Ishii, K. Dan, and H. Fujiwara (2002).
complexities for the 1979 Imperial Valley, California, earthquake, Characterizing heterogeneous slip models for large subduction earth-
Geophys. Res. Lett. 30, no. 5, 1245, doi: 10.1029/2002GL015839. quakes for strong ground motion prediction, in Proc. of the 11th Symp.
Lay, T., and T. C. Wallace (1995). Modern Global Seismology, Academic of Earthq. Eng., Vol. 1, 163–166 (in Japanese).
Press, San Diego, California, 521 pp. Somerville, P., K. Irikura, R. Graves, S. Sawada, D. Wald, N. Abrahamson,
Madariaga, R. (1979). On the relation between seismic moment and stress Y. Iwasaki, T. Kagawa, N. Smith, and A. Kowada (1999). Character-
drop in the presence of stress and strength heterogeneity, J. Geophys. izing crustal earthquake slip models for the prediction of strong ground
Res. 84, 2243–2250. motion, Seismol. Res. Lett. 70, 59–80.
Mai, P. M., and G. C. Beroza (2000). Source scaling properties from finite- Strasser, F. O., M. C. Arango, and J. J. Bommer (2010). Scaling of the source
fault-rupture models, Bull. Seismol. Soc. Am. 90, 604–615. dimensions of interface and intraslab subduction-zone earthquakes
Mai, P. M., and G. C. Beroza (2002). A spatial random field model to char- with moment magnitude, Seismol. Res. Lett. 81, 941–950.
acterize complexity in earthquake slip, J. Geophys. Res. 107, no. B11, Wald, D. J., and P. Somerville (1995). Variable-slip rupture model of the
2308, doi: 10.1029/2001JB000588. Great 1923 Kanto, Japan, earthquake: Geodetic and body-waveform
Mendoza, C. (1993). Coseismic slip of two large-magnitude Mexican earth- analysis, Bull. Seismol. Soc. Am. 85, 156–177.
quakes from teleseismic body waveforms: Implications for asperity Wells, D. L., and K. J. Coppersmith (1994). New empirical relationships
interaction in the Michoacán plate boundary segment, J. Geophys. among magnitude, rupture length, rupture width, rupture area,
Res. 98, 8197–8210. and surface displacement, Bull. Seismol. Soc. Am. 84, 974–1002.
Mendoza, C. (1995). Finite fault analysis of the 1979 March 14, Petatlán, Yagi, Y., T. Mikumo, J. Pacheco, and G. Reyes (2004). Source rupture
Mexico earthquake using teleseismic P waveforms, Geophys. J. Int. process of Tecomán, Colima, México earthquake of 22 January,
121, 675–683. 2003, determined by joint inversion of teleseismic body-wave and
Mendoza, C., and E. Fukuyama (1996). Hokkaido-Nansei-Oki, Japan, near-source data, Bull. Seismol. Soc. Am. 94, 1795–1807.
earthquake: Coseismic slip pattern from strong motion and teleseismic
recordings, J. Geophys. Res 101, 791–801.
Mendoza, C., and S. H. Hartzell (1989). Slip distribution of the 19 Septem- Departamento de Ciencias Computaciónales
ber 1985 Michoacán, Mexico earthquake: Near-source and teleseismic Centro Universitario de Ciencias Exactas e Ingeniería (CUCEI) de la
constrains, Bull. Seismol. Soc. Am. 79, 655–669. Universidad de Guadalajara
Mendoza, C., and S. H. Hartzell (1999). Fault slip distribution of the 1995 Boulevard Marcelino García Barragán # 1421
Colima-Jalisco, Mexico, Earthquake, Bull. Seismol. Soc. Am. 89, CP. 44430, Guadalajara, Jalisco, México
1338–1344. l7m8s3r@gmail.com
Mendoza, C., S. H. Hartzell, and T. Monfret (1994). Wide-band analysis of (A.R.-G.)
the 3 March 1985 central Chile earthquake: Overall source process and
rupture history, Bull. Seismol. Soc. Am. 84, 269–283.
Miyake, H., T. Iwata, and K. Irikura (2003). Source characterization for Instituto de Ingeniería
broadband ground-motion simulation: Kinematic heterogeneous Universidad Nacional Autónoma de México
source model and strong motion generation area, Bull. Seismol. Ciudad Universitaria
Soc. Am. 93, 2531–2545. Coyoacán 04510, México DF, México
Miyake, H., T. Iwata, and K. Irikura (2004). Controlling factors of strong joagg@pumas.iingen.unam.mx
ground motion prediction for scenario earthquakes, 13th World mjaimest@iingen.unam.mx
Conference on Earthquake Engineering, Vancouver B.C., Canada, (J.A., M.A.J.)
Paper Number 2801, 10 pp.
Moya, A., J. Aguirre, and K. Irikura (2000). Inversion of source parameters
and site effects from strong ground motion records using genetic Puerto Rico Seismic Network
algorithms, Bull. Seismol. Soc. Am. 90, no. 4, 977–992. Department of Geology
Murotani, S., H. Miyake, and K. Koketsu (2008). Scaling of characterized University of Puerto Rico
slip models for plate-boundary earthquakes, Earth Planets Space 60, Residencie 2ª Mayagüez R. 9000, 00680
987–991. Victor@prsn.uprm.edu
Ramírez-Gaytán, A., J. Aguirre, and C. Huerta (2010). Tecomán earthquake: (V.H.)
Physical implications of seismic source modeling, applying the
empirical Green’s function method, and evidence of non-linear behav- Manuscript received 9 February 2013;
ior of ground, ISET J. Earthq. Technol. 47, no. 1, 1–23. Published Online 25 March 2014

You might also like