You are on page 1of 27

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/227083601

Aftershock Statistics

Article  in  Pure and Applied Geophysics · June 2005


DOI: 10.1007/s00024-004-2661-8

CITATIONS READS
127 1,048

3 authors, including:

Robert Shcherbakov John B. Rundle


The University of Western Ontario University of California, Davis
97 PUBLICATIONS   2,327 CITATIONS    512 PUBLICATIONS   10,318 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Full pipeline simulation of earthquakes, tsunamis, and ionosphere View project

Virtual Quake View project

All content following this page was uploaded by Robert Shcherbakov on 10 March 2014.

The user has requested enhancement of the downloaded file.


Ó Birkhäuser Verlag, Basel, 2005
Pure appl. geophys. 162 (2005) 1051–1076
0033 – 4553/05/071051 – 26 Pure and Applied Geophysics
DOI 10.1007/s00024-004-2661-8

Aftershock Statistics
ROBERT SHCHERBAKOV1 ; DONALD L. TURCOTTE2 and JOHN B. RUNDLE1

Abstract—The statistical properties of aftershock sequences are associated with three empirical scaling
relations: (1) Gutenberg-Richter frequency-magnitude scaling, (2) Båth’s law for the magnitude of the
largest aftershock, and (3) the modified Omori’s law for the temporal decay of aftershocks. In this paper
these three laws are combined to give a relation for the aftershock decay rate that depends on only a few
parameters. This result is used to study the temporal properties of aftershock sequences of several large
California earthquakes. A review of different mechanisms and models of aftershocks are also given. The
scale invariance of the process of stress transfer caused by a main shock and the heterogeneous medium in
which aftershocks occur are responsible for the occurrence of scaling laws. We suggest that the observed
partitioning of energy could play a crucial role in explaining the physical origin of Båth’s law. We also
study the stress relaxation process in a simple model of damage mechanics and find that the rate of energy
release in this model is identical to the rate of aftershock occurrence described by the modified Omori’s law.

Key words: Earthquakes, aftershocks, damage mechanics, fracture, critical point, power-law scaling.

1. Introduction

Earthquakes are typically followed by aftershock sequences. The identification of


these sequences is usually a cumbersome problem which requires assumptions on the
spatial and temporal clustering of aftershocks (MOLCHAN and DMITRIEVA, 1992).
Another problem concerns the relationship between the magnitude of the main shock,
which initiated the aftershock sequence, and the magnitudes of the aftershocks. Can
certain aftershocks have magnitudes larger than the original main shock? If this is the
case then the main shock can be considered as a foreshock. This classification of all
earthquakes as foreshocks, main shocks, or aftershocks also raises questions
concerning the mechanisms responsible for each type of event. One point of view is
that the physics behind all earthquakes is the same and therefore this subdivision is not
fundamental. The alternative approach discriminates between these events by applying
different declustering algorithms in order to study their statistical and physical
properties based on their type. Aftershock sequences themselves have a complicated

1
Center for Computational Science and Engineering, University of California, Davis, CA 95616,
U.S.A. E-mails: rshcherbakov@ucdavis.edu, and jbrundle@ucdavis.edu
2
Department of Geology, University of California, Davis, CA 95616, U.S.A. E-mail: turcotte@
geology.ucdavis.edu
1052 R. Shcherbakov et al. Pure appl. geophys.,

hierarchical structure in which each aftershock can produce its own aftershock
sequence and so forth. Extensive reviews of the properties of aftershocks have been
given by UTSU et al. (1995), KISSLINGER (1996), UTSU (2002), and GROSS (2003).
A fundamental question in studies of aftershocks is: Why do they occur? The
simplified answer is that the main shock increases regional stresses resulting in the
subsequent aftershocks. In the simplest dislocation model of an earthquake, the
strain accumulation and release produces regions of stress increase (DAS and
SCHOLZ, 1981a,b). These stress ‘‘halos’’ certainly contribute to the occurrence of
aftershocks (TODA et al., 1998; GROSS and BURGMANN, 1998; RYBICKI, 1973;
MENDOZA and HARTZELL, 1988; KING et al., 1994; MARCELLINI, 1995; HARDEBECK
et al., 1998; STEIN, 1999). However, the actual stress fields prior to and after an
earthquake are undoubtedly very complex. Local stress concentrations resulting
from ‘‘asperities’’ and ‘‘barriers’’ also contribute to the generation of aftershocks in
the vicinity of the main shock rupture (SCHOLZ, 2002). In some cases a fault adjacent
to the main shock may be on the brink of failure. If this fault is ‘‘large,’’ a resultant
triggered earthquake may be larger than the initial main shock. In this case the
triggered earthquake is considered to be the main shock and the initial main shock
is defined to be a foreshock.
A second major question concerning aftershocks is: What is the physics of the
time delay associated with the occurrence of aftershocks? The stress transfer occurs
at the time of the main shock, but aftershocks continue for months and years after
the main shock. In this paper a model of damage mechanics will be used to provide
insights into some aspects of this time delay.
The sequence of foreshocks, main shock, and aftershocks occurs over a relatively
short timespan ss . For a moderate sized earthquake, m  6:0, this timespan is about a
year. The timespan for tectonic stress accumulation st is generally much longer, 102
to 104 years. We argue that it is reasonable to assume that ss  st and neglect
tectonic stress accumulation during an aftershock sequence. Thus, the primary cause
of aftershocks is the stress transfer to adjacent regions during the occurrence of a
main shock. It should be noted, however, that the seismic waves, changes in pore
fluid pressure associated with a main shock, may cause some reduction of the
strength (static friction) on adjacent faults and thus contribute to the occurrence of
aftershocks. Nevertheless, the role of aftershocks is to relax the excess stresses (above
the yield stress) generated by a main shock.
Despite considerable statistical variability associated with aftershock sequences, it
was observed that their behavior appears to satisfy several scaling laws. These are:
1 Aftershocks satisfy Gutenberg-Richter (GR) frequency-magnitude scaling just as
all earthquakes do. It is essential to note that the validity of the GR scaling is
related to a fractal scaling between the number of earthquakes and their rupture
areas. Thus, aftershock distributions are scale invariant in terms of the frequency,
rupture-area statistics.
Vol. 162, 2005 Aftershock Statistics 1053

2 Båth’s law states that the difference in magnitude between a main shock and
its largest aftershock is approximately a constant, independent of the main shock
magnitude. In this paper we will utilize an alternative form of Båth’s law
which considers the difference between the main shock magnitude and an inferred
‘‘largest’’ aftershock based on an extrapolation of the GR scaling. The constancy
of this difference is shown to be a consequence of the scale-invariant relation of
aftershock sequences relative to main shocks.
3 The modified Omori’s law gives the temporal decay in the rate of after-
shock occurrence, although alternative laws have been proposed (GROSS and KIS-
SLINGER, 1994). We show observational evidence for the validity of Omori’s law
and derive a generalized Omori’s law that includes both Båth’s law and GR
scaling. We also show that this generalized Omori’s law can be derived using a
simple model of damage mechanics.
The primary purpose of this paper is the study of the statistical properties of
aftershock sequences for several large California earthquakes. To accomplish this we
have incorporated the three scaling relations to give the generalized Omori’s law for
aftershock decay rates that depend on several parameters specific for each seismogenic
region. We have also shown that the characteristic time c, first introduced in the
modified Omori’s law, is no longer a constant but scales with the lower magnitude
cutoff and the main shock magnitude. We then show how a damage mechanics model
can reproduce time delays and stress relaxation similar to aftershock decay rates
expressed through the modified Omori’s law. Finally we present an extensive
discussion of different mechanisms and models used in studies of aftershocks.

2. Gutenberg-Richter Scaling

The frequency-magnitude statistics of earthquakes is well approximated by the


GR relation (GUTENBERG and RICHTER, 1954)
log10 N ð mÞ ¼ a  b m; ð2:1Þ
where N ð mÞ is the cumulative number of earthquakes in a specified region and time
window with magnitudes greater than m. The constant b or ‘‘b-value’’ varies from
region to region, but is generally in the range of 0:8 < b < 1:2 (FROHLICH and DAVIS,
1993). The constant a is a measure of the regional level of seismicity and gives the
logarithm of the number of earthquakes with magnitudes greater than zero. There are
a variety of measures for the magnitude, including local, body-wave, surface-wave,
and moment magnitude (LAY and WALLACE, 1995). In general, for small earthquakes
(m < 5:5), these different magnitude measures give approximately equivalent results.
The preferred quantification for large earthquakes is in terms of its moment. The
moment is then converted to a moment magnitude using an empirical scaling relation.
1054 R. Shcherbakov et al. Pure appl. geophys.,

Studies have shown that aftershocks also satisfy the GR relation (2.1) with
‘‘b-values’’ that are not statistically different from the frequency-magnitude distri-
bution of all earthquakes (KISSLINGER, 1996). It is important to note that GR scaling
is related to the power-law (fractal) scaling (TURCOTTE, 1997)

N ð AÞ / ADR =2 ; ð2:2Þ

where N ð AÞ is the number of earthquakes with rupture areas greater than A and
DR is the fractal dimension (DR ¼ 2b). The applicability of the fractal relation (2.2)
implies scale invariance.
Presumably aftershocks satisfy the GR scaling for the same reason that all
earthquakes do. However, no generally accepted theory is available for the
explanation of the scale-invariant nature of this distribution. Generally, two end-
member models have been proposed: 1) each fault has a GR distribution of
earthquakes and 2) there is a power-law frequency-area distribution of faults and
each fault has a reoccurrence of characteristic earthquakes. Observations generally
favor the second explanation (TURCOTTE, 1997). A cellular-automata model based
on the sandpile model (BAK et al., 1988) that utilizes a fractal distribution of box
sizes reproduces GR statistics for aftershocks (BARRIERE and TURCOTTE, 1991, 1994;
HENDERSON et al., 1994; HUANG et al., 1998).
Several examples of the validity of GR scaling for aftershocks are shown in Figure 1
where the cumulative numbers of aftershocks with magnitudes greater than m, N ð mÞ,
are given as a function of m for the m ¼ 7:3 June 28, 1992 Landers earthquake, the
m ¼ 6:7 January 17, 1994 Northridge earthquake, the m ¼ 7:1 October 16, 1999 Hector
Mine earthquake, and the m ¼ 6:5 December 22, 2003 San Simeon earthquake
(catalogs are provided by the Southern California Earthquake Center, SCSN catalog,
http://www.data.scec.org/, and the Northern California Earthquake Data Center,
NCSN catalog, http://quake.geo.berkeley.edu/ncedc/). All earthquakes in square
areas centered on the main shock epicenter for prescribed time periods were considered
to be aftershocks. For the Landers, Hector Mine, Northridge, and San Simeon
earthquakes the areas were 1:1  1:1 , 1:0  1:0 , 0:6  0:6 , and 0:9  0:9 ,
respectively. In each case the linear size of the box was approximately taken to be
about the length L of the aftershock zone which scales with the magnitude of the main
shock mms as L ¼ 0:02  100:5mms km (KAGAN, 2002). A time period of T ¼ 365 days
following the main shock was considered except for the San Simeon earthquake
where data were available for 100 days after the main shock. We have previously shown
that these distributions are not too sensitive to the areas and time windows used
(SHCHERBAKOV and TURCOTTE, 2004a). The least-squares fits of (1) to the data with
aftershocks greater than m  2:0 were performed. For the Landers earthquake we have
b ¼ 0:98  0:02 and a ¼ 6:08, for the Northridge earthquake we have b ¼ 0:91  0:02
and a ¼ 5:37, for the Hector Mine earthquake we have b ¼ 1:01  0:01 and a ¼ 5:81,
and for the San Simeon earthquake we have b ¼ 1:00  0:03 and a ¼ 5:40.
Vol. 162, 2005 Aftershock Statistics 1055

(a)

(b)

Figure 1
Cumulative numbers of aftershocks with magnitudes greater than m, N ð mÞ, are given as functions of m
for (a) the Landers and Northridge earthquakes, and (b) the Hector Mine and San Simeon earthquakes.
The solid straight lines are our best fits of the modified GR relation (4) to the data.

3. Modified Båth’s Law

Another important scaling law concerning aftershocks is Båth’s law. This law
states that it is a good approximation to assume that the difference in magnitude
between the main shock and its largest aftershock is a constant independent of the
magnitude of the main shock (BÅTH, 1965). That is
1056 R. Shcherbakov et al. Pure appl. geophys.,

Dm ¼ mms  mmax
as ð3:1Þ
with mms the magnitude of the main shock, mmax as the magnitude of the largest
aftershock, and Dm approximately a constant typically taken to be Dm  1:2.
A number of extensive studies of the statistical variability of Dm have been carried
out (VERE-JONES, 1969; KISSLINGER and JONES, 1991; TSAPANOS, 1990; FELZER et al.,
2002, 2003; CONSOLE et al., 2003; HELMSTETTER and SORNETTE, 2003b). Despite
progress in understanding the nature of Båth’s law, its validity still remains an open
question. VERE-JONES (1969) analyzed a simplified model where he assumed that
events in an aftershock sequence are drawn from the negative exponential
distribution and are distributed independently of each other. He showed that the
distribution of the difference between the largest and the second largest event is the
same negative exponential distribution. Using this result he obtained the value of
Dm ¼ 1=b ln 10. For the typical values of the parameter b  1:0 this gives Dm  0:43
which is smaller than the observed value. Recently, HELMSTETTER and SORNETTE
(2003b) pointed out that the selection process of aftershock sequence as a subset of
the whole seismic catalog plays a significant role in calculating Dm. They argue that
this difference is controlled not only by the magnitude scaling but also by the
aftershock productivity.
A modified version of Båth’s law has been proposed by SHCHERBAKOV and
TURCOTTE (2004a) and is based on an extrapolation of the GR statistics for
aftershocks. The magnitude of the ‘‘largest’’ aftershock consistent with GR scaling
for aftershocks is obtained by formally setting N ð mÞ ¼ 1 in (2.1) with the result
a ¼ bm? ; ð3:2Þ
where m? is the inferred magnitude of the ‘‘largest’’ aftershock for the given
aftershock sequence. In general, this extrapolated value will differ from the mean
value of the largest aftershock obtained by averaging over an ensemble of main
shock-aftershock sequences having the same main shock magnitude. If Båth’s law is
applicable to the inferred values of m? we can write
Dm? ¼ mms  m? ; ð3:3Þ
where Dm? is approximately a constant. Substitution of (3.2) and (3.3) into (2.1) gives
log10 N ð mÞ ¼ bðmms  Dm?  mÞ ð3:4Þ
with b, mms , and Dm? specified, the frequency-magnitude distribution of aftershocks
can be determined using (3.4).
To illustrate the application of both forms of Båth’s laws we consider the four
earthquakes studied in the previous section and shown in Figure 1. For the Landers
earthquake we have mms ¼ 7:3 and mmax as ¼ 6:3 so that Dm ¼ 1:0. From the fit to the
data using (3.2) (Fig 1a) with a ¼ 6:08, we obtain the magnitude of the inferred
‘‘largest’’ aftershock m? ¼ 6:2 so that Dm? ¼ 1:1. These results are given in Table 1.
Vol. 162, 2005 Aftershock Statistics 1057

Table 1
Summary of results

Main shock mms b Dm Dm? p R b cðm? Þ, sec

Landers 7.3 0:98  0:02 1.0 1:10  0:05 1:22  0:03 4:0  0:1 1.20 33  10
Northridge 6.7 0:91  0:02 0.8 0:75  0:05 1:18  0:02 3:4  0:3 1.06 23  10
Hector Mine 7.1 1:01  0:01 1.3 1:35  0:05 1:21  0:05 4:1  0:2 1.23 32  10
San Simeon 6.5 1:00  0:03 1.7 1:10  0:05 1:12  0:02 3:5  0:2 1.09 55  10

In addition to these four earthquakes SHCHERBAKOV and TURCOTTE (2004a) also


studied six other large earthquakes that occurred in California between 1987 and
2003 with magnitudes mms  5:5. The mean difference in magnitudes between these
ten main shocks and their largest detected aftershocks is 1:16 with a standard
deviation rDm ¼ 0:46 (Båth’s law). The estimated mean difference in magnitudes
between the main shocks and the inferred ‘‘largest’’ aftershocks was 1:11 with
rDm? ¼ 0:29 (modified Båth’s law). The scatter in the modified Båth’s law is
significantly less than the scatter in the original Båth’s law.
We believe that the modified approach provides a better test of the validity of
Båth’s law than the standard use of only the largest aftershock. In extrapolating the
GR scaling (2.1) a large number of earthquakes are used to give a statistical estimate
of the largest aftershock associated with a main shock. For the examples considered
we utilized sequences of 600 to 60,000 aftershocks.
SHCHERBAKOV and TURCOTTE (2004a) explained the general applicability of the
modified form of Båth’s law in terms of energy partitioning. The applicability of the law
implies that the average ratio of the total energy radiated in an aftershock sequence to
the energy radiated by the main shock is a constant. The energy E radiated in an
earthquake is related empirically to its moment magnitude m by (UTSU, 2002)
3
log10 EðmÞ ¼ m þ log10 ER ð3:5Þ
2
with ER ¼ 6:3  104 Joules. This relation can be used directly to relate the radiated
energy from the main shock Ems to the moment magnitude of the main shock mms
3
Ems ¼ ER 102mms ð3:6Þ
The total radiated energy in the aftershock sequence Eas is obtained by integrating
over the distribution of aftershocks. This can be written
Zm?  
dN
Eas ¼ EðmÞ  dm: ð3:7Þ
dm
1

Taking the derivative of (3.4) with respect to the aftershock magnitude m we have
?
dN ¼ bðln 10Þ10bðmms Dm mÞ dm: ð3:8Þ
1058 R. Shcherbakov et al. Pure appl. geophys.,

Substitution of (3.4) and (3.8) into (3.7) gives


Zm?
10ð2bÞm dm:
? 3
Eas ¼ bðln 10ÞER 10bðmms Dm Þ ð3:9Þ
1

Carrying out the integration using (3.3) we find


b 3 ?
Eas ¼ 3  ER 102ðmms Dm Þ : ð3:10Þ
2b
The ratio of the total radiated energy in aftershocks Eas to the radiated energy in the
main shock Ems is obtained by dividing (3.10) by (3.6) with the result
Eas b 3 ?
¼ 102Dm : ð3:11Þ
Ems 32  b
If we further assume that all earthquakes have the same seismic efficiency (ratio of
radiated energy to the total drop in stored elastic energy), then this ratio is also the
ratio of the drop in stored elastic energy due to the aftershocks to the drop in stored
elastic energy due to the main shock.
From (3.11) the fraction of the total energy associated with aftershocks is given by
3 ?
Eas 3
b
b
102Dm
¼ 2 b : ð3:12Þ
Ems þ Eas 1 þ 3 1032Dm?
b 2

Taking b ¼ 0:8, 1:0, and 1:2 this result is plotted in Figure 2. The applicability of the
modified form of Båth’s law requires that the ratio of radiated energy in aftershocks
to the radiated energy in the main shock be constant. This is consistent with the
generally accepted condition of self-similarity for earthquakes. Taking b ¼ 1:0 and
Dm ¼ 1:2 as preferred values it is found that 3% of the total energy is associated with
the aftershock sequence and 97% is associated with the main shock.
To test the validity of the derived ratio (3.12) SHCHERBAKOV and TURCOTTE
(2004a) estimated Eas =ðEms þ Eas Þ for the ten large California main shock-aftershock
sequences. In each case the time interval of 365 days was taken for each sequence
except for the San Simeon earthquake where it was 100 days. The values of Dm? were
calculated using equation (3.12) and ‘‘b-values’’ obtained from the GR scaling. The
results are plotted as open squares in Figure 2. For comparison, we also plot the
same estimated ratios Eas =ðEms þ Eas Þ against the values of Dm? obtained from the
extrapolation of GR scaling. They are shown as open circles in Figure 2. Both
approaches give approximately similar values of Dm? within statistical errors.

4. Generalized Omori’s Law

A third scaling relation, the modified form of Omori’s law (UTSU), 1961),
describes the temporal decay of aftershock activity and is given in the form (UTSU
et al., 1995; NANJO et al., 1998; SCHOLZ, 2002)
Vol. 162, 2005 Aftershock Statistics 1059

Figure 2
Dependence of the fraction of the total energy loss associated with aftershocks, Eas =ðEms þ Eas Þ on
the difference in magnitude between the main shock and the inferred ‘‘largest’’ aftershock Dm? from (3.12).
Symbols are the energy loss versus the estimated values of Dm? using GR scaling (open circles) and energy
partitioning (open squares) respectively for the ten large California earthquakes.

dN K
r ¼ ; ð4:1Þ
dt ðc þ tÞp
where r  dN =dt is the rate of occurrence of aftershocks with magnitudes greater
than m, t is the time that has elapsed since the main shock and K, p and c are
parameters. This law is a manifestation of temporal correlations in aftershock
sequences which can be viewed as complex relaxation processes occurring after main
shocks. In his original formulation OMORI (1894) introduced (4.1) with p ¼ 1, usually
quite a good approximation.
For convenience we rewrite the modified Omori’s law in the form
dN 1
rðt; mÞ  ¼ ; ð4:2Þ
dt s½1 þ t=cðmÞ p
where s and cðmÞ are characteristic times that specify the decay rate of aftershocks.
Comparing (4.1) and (4.2) we have
cðmÞp
K¼ : ð4:3Þ
s
1060 R. Shcherbakov et al. Pure appl. geophys.,

The analysis of earthquake data presented below suggests that the value of cðmÞ is
not a constant but scales with the lower magnitude cutoff m and the main shock
magnitude mms .
In order to relate the three aftershock scaling relations SHCHERBAKOV et al. (2004)
compared the total number of aftershocks given by the modified form of Omori’s law
(4.2) to the total number of aftershocks given by the modified GR relation (3.4). The
total number of aftershocks with magnitudes greater than m, N ð mÞ was obtained
by integrating (4.2) with the result
Z1 Z1
dN dt cðmÞ
N ð mÞ ¼ dt ¼ p ¼ : ð4:4Þ
dt s½1 þ t=cðmÞ sðp  1Þ
0 0

If the modified Omori’s law is assumed to be valid for prolonged times (no cutoff)
then the total number of aftershocks is finite only for p > 1 (KAGAN and KNOPOFF,
1981).
Equating (3.4) and (4.4) then gives
cðmÞ ?
s¼ 10bðmms Dm mÞ : ð4:5Þ
ðp  1Þ
Substitution of (4.5) into (4.2) gives
?
ðp  1Þ10bðmms Dm mÞ 1
rðt; mÞ ¼ : ð4:6Þ
cðmÞ ½1 þ t=cðmÞ p
This is the generalized Omori’s law for aftershocks and gives the rate of occurrence
of aftershocks with magnitudes greater than m as a function of time t. It includes the
‘‘b-value’’ from GR scaling (2.1), the Dm? from the modified form of Båth’s law, (3.2)
and (3.3), the magnitude of the main shock mms that initiated the sequence, the
characteristic time cðmÞ, and the exponent p from the modified Omori’s law (4.2).
This result is similar in form to that given by REASENBERG and JONES (1989, 1994)
in their Eq. (3). Based on studies of 62 California aftershock sequences (mms  5:0)
that occurred from 1933 to 1987, they found the following mean values with time
measured in days (‘‘generic California model’’): c ¼ 0:05 days, a0 ¼ 1:67, b ¼ 0:9,
p ¼ 1:08. We can compare these two approaches and deduce the mean value of Dm?
for the ‘‘generic California model’’ using the equation
1
Dm? ¼ ½log10 ðp  1Þ þ ðp  1Þ log10 c  a0 ¼ 0:52: ð4:7Þ
b
This is about one half the value of Dm ¼ 1:2 proposed by BÅTH (1965) and discussed
above based on actual values of the largest aftershocks.
Analyzing 27 aftershock sequences in Japan, YAMANAKA and SHIMAZAKI (1990)
proposed an alternative set of values: c ¼ 0:3 days, a0 ¼ 1:83, b ¼ 0:85, p ¼ 1:3.
From (4.7) the corresponding mean value is Dm? ¼ 1:35. This is close to the typical
Vol. 162, 2005 Aftershock Statistics 1061

value of Dm ¼ 1:2 for Båth’s law. FELZER et al. (2003) have also provided a similar
summary of aftershock statistics. These authors stack 62 California aftershock
sequences with mms  5:0 for the period 1975 to 2000. They scale with an equivalent
main shock magnitude mms ¼ 6:4 and m ¼ 4:8 and find p ¼ 1:08, c ¼ 0:014 days, and
K ¼ 0:116 days0:08 . With these values a0 ¼ 2:18 and from (4.7) we obtain
Dm? ¼ 0:98 again quite close to the typical value 1:2 for Båth’s law. These authors
also argue that the a0 -value given by REASENBERG and JONES (1989) is high leading to
the low value of Dm? .
To check the applicability of the proposed scaling law (4.6) SHCHERBAKOV et al.
(2004) studied the dependence of aftershock activity on time following the Landers,
Northridge, Hector Mine, and San Simeon earthquakes. The rates of occurrence of
aftershocks, rðt; mÞ in number per day, larger than a specified magnitude m are shown
in Figure 3 as a function of the time t since the main shock occurred. It is seen from
the figures that the ratios of rates for the equally spaced magnitude cutoffs, starting
at sufficiently long times (t
c) after the main shock, remain constant. This can be
expressed mathematically as follows
rðm1 Þ rðmn Þ
Rðt
c; dm Þ  ¼ ... ¼ ¼ const :; ð4:8Þ
rðm2 Þ rðmnþ1 Þ
where the positive difference in magnitude cutoffs dm is kept fixed
dm ¼ m2  m1 ¼ m3  m2 ¼ . . . ¼ mnþ1  mn . Using (4.6) it is possible to show that
in the limit of sufficiently long times, t
c, the ratio of cðmn Þ=cðmn þ dm Þ is
 1
cðmn Þ Rðt
c; dm Þ p1
¼ : ð4:9Þ
cðmn þ dm Þ 10bdm
This solution shows that the characteristic time cðmn Þ scales with the lower
magnitude cutoff mn . It remains constant only if Rðt
c; dm Þ ¼ 10bdm .
It is seen from (4.9) that the characteristic times cðmÞ decrease with increasing
magnitude cutoff which is also seen in the earthquake data shown in Figure 3. To
find cðmn Þ for each magnitude cutoff mn one has to measure the ratio of rates
Rðt
c; dm Þ from the seismic data. Then one has to fit the observed decay rates for a
given magnitude cutoff m0 using (4.6). Finally, the rest of the values of c’s for equally
spaced magnitude cutoffs can be found recursively using (4.9). In our studies the
value m0 ¼ 3:0 was used assuming the earthquake catalog is complete for this
magnitude cutoff even at short times after a main shock.
To find an explicit dependence of cðmÞ on the main shock magnitude mms and the
lower magnitude cutoff m, we propose the following scaling for R

Rðt
c; dm Þ ¼ 10bdm ; ð4:10Þ
where b is a constant. Substituting (4.10) into (4.9) and taking into account the fact
that m?  m ¼ ndm , we obtain scaling for cðmÞ
1062 R. Shcherbakov et al. Pure appl. geophys.,

(d)
(b)

(c)
(a)
Vol. 162, 2005 Aftershock Statistics 1063

bb ?
cðmÞ ¼ cðm? Þ10p1ðmms Dm mÞ ; ð4:11Þ
where cðm? Þ is the characteristic time in (4.6) with m ¼ m? . This value introduces a lower
cutoff for c’s and signifies that the proposed scaling (4.6) is no longer valid for m > m? .
Equation (4.6) also gives an upper limit on the ratio cðmn Þ=cðmn þ dm Þ. The
definition of aftershock rates assumes that for all times the ratio of rates is
rðt; mn Þ=rðt; mn þ dm Þ > 1 for dm > 0. Particularly, at t ¼ 0 we have
rð0; mn Þ=rð0; mn þ dm Þ ¼ 10bdm cðmn þ dm Þ=cðmn Þ > 1 from which we obtain
cðmn Þ
< 10bdm : ð4:12Þ
cðmn þ dm Þ
For each of the four earthquakes our fit of the generalized Omori’s law (4.6) to
the data is given for different magnitude cutoffs m1 ¼ 1:5, m2 ¼ 2:0, m3 ¼ 2:5,
m4 ¼ 3:0, m5 ¼ 3:5, and m6 ¼ 4:0 with dm ¼ 0:5 (Fig. 3). For the Landers earthquake
we have taken b ¼ 0:98  0:02, mms ¼ 7:3, Dm? ¼ 1:1  0:05 from the analysis of
data given in Figure 1a. The exponent p ¼ 1:22  0:03 was used to fit the power-law
decay at times between 10 and 1000 days. The same time interval was used to
estimate the mean ratio of rates R ¼ 4:0  0:1. The value of the characteristic time
cðm ¼ 3:0Þ ¼ 0:7 days was estimated from the best fit of (4.6) to the decay rate. The
rest of the values of c’s for the other magnitude cutoffs were calculated using (4.9).
From (4.10) with dm ¼ 0:5 we find b ¼ 1:2 and from (4.11) we find cðm? Þ ¼ 33  10
sec. It is seen that the rate of occurrence of magnitude m1 ¼ 1:5 and above
aftershocks generally falls below the prediction, the rate of occurrence of magnitude
m2 ¼ 2:0 and above aftershocks falls below the prediction for the first 10 days,
and the rate of occurrence of magnitude m3 ¼ 2:5 and above aftershocks falls below
the prediction for the first three days. We attribute these discrepancies to the
undercounting of small aftershocks at short times. Evidence for the occurrence of
many small aftershocks is lost in the seismic codas of larger aftershocks. The same
analysis has been performed for three other earthquakes and the results are shown in
Figures 3b-3d and summarized in Table 1.
In general we find quite a good correlation between the aftershock data given in
Figure 2 and the prediction of the generalized Omori’s law given in (4.6). However, it
should be emphasized that in order to gain this agreement it was necessary to relax
the condition that the characteristic time c be a constant. We find that cðmÞ scales
both with the lower magnitude cutoff m and the main shock magnitude mms . This
b
Figure 3
The rates of occurrence of aftershocks with magnitudes greater than m in number per day, rðt; mÞ, are given
for (a) the Landers, (b) Northridge, (c) Hector Mine, and (d) San Simeon earthquakes for 1460 days (100
days for San Simeon earthquake) after a main shock. The same areas are used as in Figure 1. Lower
magnitude cutoffs were taken to be m ¼ 1:5, 2:0, 2:5, 3:0, 3:5, and 4:0. The value of cðm ¼ 3:0Þ (dm ¼ 0:5)
was obtained from the best fit (solid line) using (4.6). The rest of c’s were calculated using (4.9). Dashed
lines represent the predicted rates.
1064 R. Shcherbakov et al. Pure appl. geophys.,

dependence is required in order for the power-law scaling result given in (4.6) to
agree with the aftershock decay found for the four earthquakes we have considered.
It is based on two assumptions that aftershocks decay according to (4.2) and the
observed fact that the ratio of rates R for equally spaced magnitude cutoffs, starting
at sufficiently long times (t
c) after the main shock, remains constant. We will
show in the next section that the damage mechanics model for aftershocks could be
used to explain the observed dependence of c on m and mms .
Based on studies of the Landers and Hector Mine earthquakes WIEMER and
KATSUMATA (1999) and WIEMER et al. (2002) showed that there can be considerable
spatial variability of the constants a, b, and p for aftershock sequences. HOSONO and
YOSHIDA (2002) argue that large aftershocks can deviate significantly from the
modified Omori’s law. For the JMA magnitude 7.3 January 17, 1995 Kobe
earthquake they show that while aftershocks with m > 2:8 satisfy the modified
Omori’s law, aftershocks with m > 4:0 do not. For the JMA magnitude 7.3 October
6, 2000 Tottori earthquake the same analysis of aftershock data shows that while
aftershocks with m > 2:5 satisfy the modified Omori’s law, aftershocks with m > 3:3
do not. There will certainly be aftershock sequences that will be poorly approximated
by our scaling. This will usually be associated with anomalously large aftershocks.

5. Damage Analysis

We previously associated GR scaling and Båth’s law with scale invariance. We


now give an explanation for the applicability of the modified Omori’s law based on
damage mechanics. We then will discuss the relationships of this approach to other
explanations of the time delays associated with aftershocks. Damage mechanics is a
semi-empirical approach to the failure of brittle materials (KACHANOV, 1986; LEMAITRE
and CHABOCHE, 1990; KRAJCINOVIC, 1996; ALLIX and HILD, 2002; SHCHERBAKOV and
TURCOTTE, 2003) that has found wide usage in engineering applications. Damage
mechanics has also been applied to the brittle deformation of the Earth’s crust by a
number of authors (LYAKHOVSKY et al., 1997, 2001; AGNON and LYAKHOVSKY, 1995;
BEN-ZION and LYAKHOVSKY, 2002; TURCOTTE et al., 2003). In this approach a damage
variable a is introduced into the stress-strain relation according to
r ¼ E0 ð1  aÞ; ð5:1Þ
where E0 is the Young’s modulus of the undamaged material. The damage variable is
a measure of deviations from linear elasticity.
Based upon thermodynamic considerations LYAKHOVSKY et al. (1997) proposed
that the time evolution of the damage variable is related to strain by
da
¼ Cðn  n0 Þ; ð5:2Þ
dt
Vol. 162, 2005 Aftershock Statistics 1065

pffiffiffiffi
where n ¼ I1 = I2 with I1 ¼ k k and I2 ¼ ij ij the invariants of the strain tensor ij . If
n > n0 damage occurs and if n < n0 damage is healed.
In this paper we propose an alternative formulation for damage evolution in the
brittle solid. We introduce a yield stress ry and corresponding yield strain y . If the stress
is less than the yield stress, r ry , there is no damage and linear elasticity is applicable
r ¼ E0 : ð5:3Þ
If the stress is greater than the yield stress, r > ry , we introduce a damage variable a
according to
r  ry ¼ E0 ð1  aÞð  y Þ ð5:4Þ
with ry ¼ E0 y . In analogy to the approach taken by LYAKHOVSKY et al. (1997) we
assume that the time evolution of the damage variable is given by
 q  2
daðtÞ 1 rðtÞ ðtÞ
¼ 1 1 ; ð5:5Þ
dt td ry y

where td is a characteristic time scale for damage and q is a power to be determined from
experiments. The monotonic increase in the damage variable a given by (5.4) and (5.5)
represents the weakening of the brittle solid due to the nucleation and coalescence of
microcracks. If r < ry we assume that da=dt ¼ 0 and no additional damage occurs.
A particularly interesting set of experiments on brittle failure have been carried
out by GUARINO et al. (1998, 1999). These authors studied the failure of chipboard
panels under pressure. They carefully monitored the acoustic emissions associated
with microcracking prior to rupture. They found that these acoustic emission events
satisfied power-law frequency-magnitude scaling. They also found that the radiated
energy associated with these events increased as an inverse power-law of the time
remaining until failure. When the pressure was applied rapidly to a panel they
measured the time to failure as a function of the applied pressure. These data can be
fit by an inverse power of the excess pressure over a yield pressure (stress).
SHCHERBAKOV and TURCOTTE (2003) showed that both the power-law increase in
radiated energy and the power-law dependence of the time to failure on pressure
could be explained using the damage mechanics analysis.
For the chipboard panels considered by GUARINO et al. (1999), SHCHERBAKOV
and TURCOTTE (2003) found good agreement between the theory based on damage
mechanics and experiments taking q ¼ 2:25. GUARINO et al. (1998, 1999) found from
the distribution of locations of acoustic emission events that microcracks tend to
coalesce and interact prior to the catastrophic failure. These experimental results
have the remarkable feature of being consistently reproducible which is in contrast to
the studies of the brittle fracture of rock as given by MOGI (1962), LOCKNER et al.
(1992), LOCKNER and BEELER (2002), and others. We attribute this reproducibility to
a near uniform distribution of flaws introduced in the fabrication of chipboard
1066 R. Shcherbakov et al. Pure appl. geophys.,

panels that have a pre-existing state of ‘‘damage’’ (flaws). This is also the case for the
earth’s crust; the faults and joints that are main features of the seismogenic zones of
the crust constitute a pre-existing state of ‘‘damage.’’ The applicability of scaling laws
to aftershocks is direct evidence for the existence of this state.
To model the decay rate of aftershocks we use the damage mechanics approach.
Our working hypothesis is that stress transfer during a main shock increases the
stress r and strain  above the yield values ry and y in some regions adjacent to the
fault on which the main shock occurred. We assume the increases in stress and strain
are essentially instantaneous and follow linear elasticity as given in (5.3). As discussed
in the introduction, we believe it is a good approximation to neglect any increase in
regional stress due to tectonics during the aftershock sequence. We hypothesize that
the applied strain 0 remains constant and that the stress r relaxes to its yield value ry
due to damage (aftershocks). We further recognize that all aftershocks have secondary
aftershocks and so forth, however all aftershocks contribute to stress relaxation in a
self-similar way. All aftershocks, whether primary, secondary, etc., give a reduction in
regional stress as a function of the magnitudes of the aftershocks.
SHCHERBAKOV and TURCOTTE (2004b) considered a strain 0 > y applied
instantaneously at t ¼ 0 and held constant. The applicable equation for the rate of
increase of damage is obtained from (5.5) with the result
 q  2
da 1 rðtÞ 0
¼ 1 1 : ð5:6Þ
dt td ry y
From (5.1) the stress r is related to the damage variable a and the constantly applied
strain 0 by
rðtÞ  ry ¼ E0 ½1  aðtÞ ð0  y Þ: ð5:7Þ
Substitution of (5.7) into (5.6) using (y ¼ ry =E0 ) gives
 qþ2
da 1 0
¼ 1 ½1  aðtÞ q : ð5:8Þ
dt td y
Integrating with the initial condition að0Þ ¼ 0, we find
"  qþ2 #q1
1

t 0
aðtÞ ¼ 1  1 þ ðq  1Þ 1 : ð5:9Þ
td y

This result is valid as long as q > 1. The damage increases monotonically with time
and as t ! 1 the maximum damage is að1Þ ¼ 1. Using (5.9) with (5.1) one obtains
the stress relaxation in the material as a function of time t
 "  qþ2 #q1
1

rðtÞ 0 t 0
¼1þ  1 1 þ ðq  1Þ 1 : ð5:10Þ
ry y td y
Vol. 162, 2005 Aftershock Statistics 1067

At t ¼ 0 we have linear elasticity corresponding to a ¼ 0. In the limit t ! 1 the


stress relaxes to the yield stress rð1Þ ¼ ry below which no further damage can occur,
again as expected.
We recognize that the total energy radiated in aftershocks can be greater than the
elastic energy transferred to regions of higher stress by a main shock. This is because
an aftershock triggered by the increase in stress may have a stress drop that is
considerably larger than the transferred stress increase. The stress increase will often
trigger an earthquake on a fault that is on the brink of failure. An extreme example is
a main shock that is triggered by the stress transfer from a foreshock. However, we
argue that the radiated energy in aftershocks scales with the increase in elastic energy
due to stress transfer during a main shock.
In order to quantify the rate of aftershock occurrence we determine the rate of
energy release in the relaxation process considered above. The elastic energy density
(per unit mass) e0 in a sample after the instantaneous strain 0 has been applied is
1 
e0 ¼ E0 20  2y : ð5:11Þ
2
Since the strain is constant during the stress relaxation, no work is done on the
sample. We hypothesize that if the applied strain (stress) is instantaneously removed
during the relaxation process then the sample will return to a state of zero stress and
strain following a linear stress-strain path with slope E0 ð1  aÞ to stress ry and
following the path with slope E0 to zero stress. With this assumption the energy
density recovered during this stress relaxation is given by
1 h  2 i
e1 ¼ E0 20  2y  0  y aðtÞ : ð5:12Þ
2
We assume that the difference between the energy added e0 and the energy recovered
e1 is lost in aftershocks and find that this energy eas is given by
1  2
eas ¼ e0  e1 ¼ E0 0  y aðtÞ: ð5:13Þ
2
The rate of energy release is obtained by substituting (5.9) into (5.13) and taking the
time derivative with the result
 qþ4 "  qþ2  #q1
q

deas E0 2y 0 0 t
¼ 1 1 þ ðq  1Þ 1 : ð5:14Þ
dt 2td y y td

And we see that this has the same form as the modified Omori’s law given in (4.6).
The total energy of aftershocks east is obtained by substituting að1Þ ¼ 1 into
(5.13) with the result
1  2
east ¼ E0 0  y : ð5:15Þ
2
1068 R. Shcherbakov et al. Pure appl. geophys.,

In order to have the same power-law dependence in (4.6) and (5.14) we require
p
q¼ : ð5:16Þ
p1
It is also convenient to introduce a new characteristic time t according to
td
t ¼  qþ2 : ð5:17Þ
0
ðq  1Þ y 1

Substitution of (5.15) and (5.17) into (5.14) gives


1 deas p  1 1
¼ : ð5:18Þ
east dt t ð1 þ t=t Þp
Combining (3.4) and (4.6) we can write the modified Omori’s law in the form
1 dN p  1 1
¼ ; ð5:19Þ
Nt dt c ð1 þ t=cÞp
where Nt ¼ N ð> mÞ is the total number of aftershocks with magnitude greater than
m. With c ¼ t we find the fraction rate of energy lost e1 ast deas =dt given by damage
mechanics is equal to the rate of occurrence of aftershocks Nt1 ðdN =dtÞ given by the
modified Omori’s law.
As discussed in Section 4, REASENBERG and JONES (1989) and FELZER et al.
(2003) found that p ¼ 1:08 for 62 California aftershock sequences, from (5.16) this
corresponds to q ¼ 13:5. For 27 Japanese aftershock sequences YAMANAKA and
SHIMAZAKI (1990) found that p ¼ 1:3, from (5.16) this corresponds to q ¼ 4:3.
In our analysis of the four aftershock sequences given in the previous section we
found from (4.11) that c is a function of both the main shock magnitude mms and the
lower magnitude cutoff m. This dependence can be explained in terms of the result (5.17)
obtained above for the damage model. Assuming td to be a constant, a variation in 0 =y
leads to a variation in c ¼ t? . It is reasonable to suggest that there is a variation in the
excess strain 0 that is dependent on both mms and m. We can speculate here that on
average, smaller aftershocks predominantly occur in regions of minor excess strain.
One aspect of damage mechanics that has not been considered in this paper is
‘‘healing.’’ If a material ‘‘heals,’’ the damage and the damage variable decrease.
When studying material failure it is not necessary to consider healing, although any
steady-state deformation of a brittle material requires both the generation and
healing of damage.

6. Discussion

In the previous section we derived Omori’s law using damage mechanics.


Empirical time-dependent failure laws were developed independently for fiber bundles
Vol. 162, 2005 Aftershock Statistics 1069

(COLEMAN), 1958). The fiber-bundle models are applicable to the failure of composite
materials. KRAJCINOVIC (1996) and TURCOTTE et al. (2003) have shown that time-
dependent fiber-bundle models are equivalent to the standard damage mechanics
model.
Although damage mechanics and fiber bundle models are useful ways to quantify
the time delays associated with fracture and aftershocks, they do not provide a
physical explanation. The time delays associated with nucleation and growth of a
single crack are often attributed to stress corrosion. A number of authors have
applied this explanation to the time delays associated with aftershocks (DAS and
SCHOLZ, 1981b; YAMASHITA and KNOPOFF, 1987; MARCELLINI, 1997; MORENO et al.,
2001). But this association has not been quantified.
A related time delay is associated with friction. Laboratory studies of friction
between rock surfaces are empirically correlated with rate-and-state friction laws
(DIETERICH, 1978; RUINA, 1983). When the slip velocity is changed there is a time delay
associated with the establishment of a new steady-state coefficient of friction. The state
variable in the friction law has also been used to quantify the delay in rupture on a
frictional surface. DIETERICH (1994) has derived Omori’s law from the equations of
rate-and-state friction. GROSS and BURGMANN (1998) have applied this result to the
Loma Prieta earthquake and its application has been discussed by ZIV and RUBIN
(2003). There is clearly a close association between this work and the damage mechanics
approach. However, the frictional approach involves increases in applied stress
whereas we have argued that aftershocks are a process of stress relaxation.
A number of statistical physicists have related the time delays associated with
brittle fracture to the time delays associated with metastable phase changes. BUCHEL
and SETHNA (1997), ZAPPERI et al. (1997), and KUN and HERRMANN (1999) have
drawn an analogy between brittle fracture and a first-order phase transition.
SORNETTE and ANDERSEN (1998) and GLUZMAN and SORNETTE (2001) argue that
brittle rupture is analogous to a critical point phenomena. SELINGER et al. (1991),
RUNDLE et al. (1996, 1999, 2000), and ZAPPERI et al. (1999) have drawn an analogy
between brittle failure and spinodal nucleation.
RUNDLE et al. (1999) have associated aftershock sequences with the power-law
scaling in the vicinity of a spinodal. This concept provides an analogy between the
time delay associated with a phase change (i.e., nucleation in a superheated liquid)
and the time delay associated with the occurrence of aftershocks. In thermodynamic
equilibrium, a liquid will boil when heated to the boiling temperature. However, a
liquid can be superheated to temperatures above the boiling temperatures. The phase
change is then accomplished through the homogeneous and/or heterogeneous
nucleation of bubbles. The limit of superheating is the spinodal temperature. As this
temperature is approached self-similar homogeneous nucleation will occur. RUNDLE
et al. (1999) have applied the theory of this process to the aftershock problem and
found that they can recover the modified Omori’s law with p ¼ 1. However, other
values of the exponent p can be obtained using alternative nucleation laws.
1070 R. Shcherbakov et al. Pure appl. geophys.,

The analogy we have made between phase transitions and fracture also has a
thermodynamic basis. Thermal fluctuations are crucial in phase transitions of solids
and liquids. A fundamental question is whether temperature plays a significant role
in the damage of brittle materials. Some forms of ‘‘damage’’ are clearly thermally
activated. The deformation of solids by diffusion and dislocation creep is an example.
The ability of vacancies and dislocations to move through a crystal is governed by an
exponential dependence on absolute temperature with a well-defined activation
energy. The role of temperature in brittle fracture is unclear. GUARINO et al. (1998)
varied the temperature in their experiments on the fracture of chipboard and found
no effect. A systematic temperature dependence of rate and state friction was
documented by NAKATANI (2001). This has also been shown to be true for the
lifetime statistics of Kevlar fibers (WU et al., 1988).
Time delays associated with bubble nucleation in a superheated liquid are
explained in terms of thermal fluctuations. The fluctuations must become large
enough to overcome the stability associated with surface tension in a bubble. The
fundamental question in damage mechanics is the cause of the delay in the
occurrence of damage. This problem has been considered in some detail by
CILIBERTO et al. (2001) and SCORETTI et al. (2001). These authors attributed damage
to the ‘‘thermal’’ activation of microcracks. An effective ‘‘temperature’’ can be
defined in terms of the spatial disorder (heterogeneity) of the solid. The spatial
variability of stress in the solid is caused by the microcracking itself, not by thermal
fluctuations. This microcracking occurs on a wide range of scales.
An alternative explanation for aftershock delays is to associate them with pore
fluid diffusion (NUR and BOOKER, 1971; LI et al., 1987; BOSL and NUR, 2002).
Diffusion may certainly influence aftershocks but it is unlikely that it is the primary
explanation for time delays.
Statistical studies have also been carried out on the spatial distribution of
aftershocks. ROBERTSON et al. (1995) have applied the three-dimensional box-
counting method in order to determine whether the distribution is fractal. These
authors considered the aftershock sequence of the m ¼ 6:1 Joshua Tree earthquake
of April 23, 1992 (2600 events in a 20  20  19 km volume in 160 days) and the
aftershocks sequence of the m ¼ 6:2 Big Bear earthquake of June 28, 1992 (818 events
in a 20  20  17 km volume in 375 days). Using cubes with linear dimension
between 500 m and 20 km a reasonably good fractal correlation was obtained with
fractal dimension D  2:0. These authors pointed out that this is also the fractal
dimension of a three-dimensional percolation cluster. The suggestion is that the
connectivity of aftershocks is directly analogous to the connectivity of a three-
dimensional percolation cluster.
An alternative approach to the scaling behavior of aftershocks is the epidemic-
type aftershock sequence (ETAS) model (GUO and OGATA, 1997; HELMSTETTER and
SORNETTE, 2002a,b, 2003a; HELMSTETTER 2003; HELMSTETTER et al., 2003a,b). This
is a statistical branching model that includes as inputs the statistics of aftershocks
Vol. 162, 2005 Aftershock Statistics 1071

based on both the GR frequency-magnitude scaling and the modified Omori’s law.
Using this model HELMSTETTER and SORNETTE (2003b) were able to reproduce
Båth’s law to a reasonably good approximation over a range of the model
parameters. NARTEAU et al. (2002) have proposed a similar approach. We believe
that the statistical ETAS approach and the content of this paper are complimentary.
There are fundamental physical mechanisms such as those proposed in this paper but
there certainly is also a statistical component.
An attempt to introduce a universal scaling law for earthquakes has been
proposed by CHRISTENSEN et al. (2002) and BAK et al. (2002). These authors
estimated the probability distribution of the interoccurrence time intervals s between
earthquakes, PL;m ðsÞ, within an area of linear size L and cutoff magnitude m as scaling
parameters. For Southern California they found a universal scaling with
 
PL;m ðsÞ ¼ sp f sLdf 10bm : ð6:1Þ

They associated sp with Omori’s law (p  1:0), 10bm with GR scaling (b  1:0), and
Ldf with fractal spatial scaling (df  1:2) for the 2d location of earthquake epicenters.
Two distinct scaling regions were found, for short times these results correspond to
our generalized Omori’s law for aftershocks and for long times they are associated
with the uncorrelated regime of main shocks. To take into account the spatial
heterogeneity and nonstationarity of earthquake occurrence rates, CORRAL (2003) has
found that the fast decay for long times is not exponential, but another power law.

7. Conclusions

Aftershock satisfy three empirical scaling relations to a good approximation. The


frequency-magnitude statistics satisfy GR scaling in much the same way as for all
shocks. In both cases the scaling can be related to the frequency-size scaling of faults
and fault segments. Båth’s law states that the difference in magnitude between main
shocks and their largest aftershocks are constant to a good approximation. An
alternative form of Båth’s law is to extrapolate the GR statistics to deduce the
‘‘largest’’ aftershock. Studies of ten aftershock sequences in California show that the
modified Båth’s law has a factor of two less scatter but both give a magnitude
difference close to 1.2 as other compilations have shown. The applicability of Båth’s
law follows from a self-similar scaling between main shocks and their aftershock
sequences.
We have combined GR scaling (2.1) with Båth’s law (3.2) and (3.3) to give the
GR relation in terms of b, mms , and Dm (3.4). In (4.2) we have expressed the modified
Omori’s law in terms of two characteristic times, s and cðmÞ, and an exponent p. In
(4.5) the total number of aftershocks with magnitudes greater than m, N ð> mÞ,
obtained from the modified Omori’s law, is equated to the same quantity from GR
scaling in order to eliminate s. This allows the modified form of Omori’s law to be
1072 R. Shcherbakov et al. Pure appl. geophys.,

expressed in terms of b, mms , Dm, cðmÞ, and p (4.6). The validity of this law is tested in
Figure 3 for the Landers, Northridge, Hector Mine, and San Simeon earthquakes.
Analyses of aftershock decay rates of large California earthquakes suggests that the
characteristic time cðmÞ scales with the lower magnitude cutoff m and the main shock
magnitude mms . The value of cðmÞ increases when the cutoff m is decreased. This
suggests that the almost constant rate of seismic activity persists longer for smaller
magnitude cutoffs and then starts to decrease.
Problems of stress relaxation due to microcracking have been attacked in the
engineering literature using damage mechanics. We hypothesize that regions around
the main shock rupture have an increase in stress above the yield stress. The
aftershock sequence relaxes the excess stress to the yield stress. We model this
transient stress relaxation using damage mechanics. The resulting rate equation
(5.18) is identical to the modified Omori’s law given in (5.19). Clearly our basic
hypothesis is idealized. However, we think that it reproduces the scaling laws that we
have discussed. The complexities of the stress distributions before and after the main
shock produce deviations from the scaling relations but do not invalidate them. We
find that approximately 97% of the radiated energy is associated with the main shock
and this reflects the reduction in stored elastic energy during the primary rupture.
The remaining 3% of energy is radiated in the aftershock sequence and is associated
with the increase in stress during the primary rupture.

Acknowledgment

The authors express thanks to Yehuda Ben-Zion for many valuable and
stimulating discussions. This work has been supported by NASA/JPL Grant 1247848,
US DOE Grant DE-FG03-03ER15380, and NSF Grant ATM-0327571.

REFERENCES

AGNON, A. and LYAKHOVSKY, V., Damage distribution and localization during dyke intrusion. In The
Physics and Chemistry of Dykes (G. Baer and A. Heimann, eds.), The Physics and Chemistry of Dykes
(Balkema, Brookfield, 1995) pp. 65–78.
ALLIX, O. and HILD, F., eds., Continuum Damage Mechanics of Materials and Structures, 1st ed. (Elsevier,
2002).
BAK, P., CHRISTENSEN, K., DANON, L., and SCANLON, T. (2002), Unified Scaling Law for Earthquakes,
Phys. Rev. Lett. 88, 17.
BAK, P., TANG, C., and WIESENFELD, K. (1988), Self-organized Criticality, Phys. Rev. A 38, 1, 364–374.
BARRIERE, B. and TURCOTTE, D. L. (1991), A Scale-invariant Cellular-automata Model for Distributed
Seismicity, Geophys. Res. Lett. 18, 11, 2011–2014.
BARRIERE, B. and TURCOTTE, D. L. (1994), Seismicity and Self-organized Criticality, Phys. Rev. E 49, 2,
1151–1160.
BÅTH, M. (1965) Lateral Inhomogeneities in the Upper Mantle, Tectonophysics 2, 483–514.
BEN-ZION, Y. and LYAKHOVSKY, V. (2002) Accelerated Seismic Release and Related Aspects of Seismicity
Patterns on Earthquake Faults, Pure Appl. Geophys. 159, 10, 2385–2412.
Vol. 162, 2005 Aftershock Statistics 1073

BOSL, W. J. and NUR, A. (2002), Aftershocks and Pore Fluid Diffusion Following the 1992 Landers
Earthquake, J. Geophys. Res. 107, B12 , Art. No. 2366.
BUCHEL, A. and SETHNA, J. P. (1997), Statistical Mechanics of Cracks: Fluctuations, Breakdown, and
Asymptotics of Elastic Theory, Phys. Rev. E 55, 6, 7669–7690.
CHRISTENSEN, K., DANON, L., SCANLON, T., and BAK, P. Unified Scaling Law for Earthquakes, Proc. Natl.
Acad. Sci. U.S.A. 99, 2509–2513.
CILIBERTO, S., GUARINO, A., and SCORRETTI, R. (2001), The Effect of Disorder on the Fracture Nucleation
Process, Physica D 158, 1–4, 83–104.
COLEMAN, B. D. (1958), Statistics and Time Dependence of Mechanical Breakdown in Fibers, J. Appl. Phys.
29, 968–983.
CONSOLE, R., LOMBARDI, A. M., MURRU, M., and RHOADES, D. (2003), Båth’s Law and the Self-similarity
of Earthquakes, J. Geophys. Res. 108, B2, Art. No. 2128.
CORRAL, Á. (2003) Local Distributions and Rate Fluctuations in a Unified Scaling Law for Earthquakes,
Phys. Rev. E 68, Art. No. 035102.
DAS, S. and SCHOLZ, C. H. (1981a), Off-fault Aftershock Clusters Caused by Shear-stress Increase, Bull.
Seismol. Soc. Am. 71, 5, 1669–1675.
DAS, S., and SCHOLZ, C. H. (1981b) Theory of Time-dependent Rupture in the Earth, J. Geophys. Res. 86,
NB7, 6039–6051.
DIETERICH, J. (1994), A Constitutive Law for Rate of Earthquake Production and its Application to
Earthquake Clustering, J. Geophys. Res. 99, B2, 2601–2618.
DIETERICH, J. H. (1978), Time-dependent Friction and Mechanics of Stick-Slip, Pure Appl. Geophys. 116,
4–5, 790–806.
FELZER, K. R., ABERCROMBIE, R. E., and EKSTRÖM, G. (2003), Secondary Aftershocks and their Importance
for Aftershock Forecasting, Bull. Seismol. Soc. Am. 93, 4, 1433–1448.
FELZER, K. R., BECKER, T. W., ABERCROMBIE, R. E., EKSTRÖM, G., and RICE, J. R. (2002), Triggering of
the 1999 MW 7.1 Hector Mine Earthquake by Aftershocks of the 1992 MW 7.3 Landers Earthquake,
J. Geophys. Res. 107, B9, Art. No. 2190.
FROHLICH, C., and DAVIS, S. D. (1993), Teleseismic b-values - or, much ado about 1.0, J. Geophys. Res. 98, B1,
631–644.
GLUZMAN, S., and SORNETTE, D. (2001), Self-consistent Theory of Rupture by Progressive Diffuse Damage,
Phys. Rev. E 6306, 6, Art. No. 066129.
GROSS, S. (2003) Failure-time Remapping in Compound Aftershock Sequences, Bull. Seismol. Soc. Am. 93, 4,
1449–1457.
GROSS, S. and BURGMANN, R. (1998) Rate and State of Background Stress Estimated from the Aftershocks
of the 1989 Loma Prieta, California, earthquake, J. Geophys. Res. 103, B3, 4915–4927.
GROSS, S.J. and KISSLINGER, C. (1994) Tests of Models of Aftershock Rate Decay, Bull. Seismol. Soc. Am.
84, 5, 1571–1579.
GUARINO, A., CILIBERTO, S., and GARCIMARTIN, A. (1999) Failure Time and Microcrack Nucleation,
Europhys. Lett. 47, 4, 456–461.
GUARINO, A., GARCIMARTIN, A., and CILIBERTO, S. (1998) An Experimental Test of the Critical Behaviour
of Fracture Precursors, Eur. Phys. J. B 6, 1, 13–24.
GUO, Z.Q. and OGATA, Y. (1997) Statistical Relations Between the Parameters of Aftershocks in Time,
Space, and Magnitude, J. Geophys. Res. 102, B2, 2857–2873.
GUTENBERG, B. and RICHTER, C.F., Seismicity of the Earth and Associated Phenomenon, 2nd ed. (Princeton
Univ. Press, Princeton, 1954).
HARDEBECK, J.L., NAZARETH, J.J., and HAUKSSON, E. (1998) The Static Stress Change Triggering Model:
Constraints from Two Southern California Aftershock Sequences, J. Geophys. Res. 103, B10, 24427–
24437.
HELMSTETTER, A. (2003) Is Earthquake Triggering Driven by Small Earthquakes? Phys. Rev. Lett. 91, 5,
Art. No. 058501.
HELMSTETTER, A. and SORNETTE, D. (2002b), Diffusion of Epicenters of Earthquake Aftershocks, Omori’s
Law, and Generalized Continuous-time Random Walk Models, Phys. Rev. E 66, 6, Art. No. 061104.
HELMSTETTER, A. and SORNETTE, D. (2002b), Subcritical and Supercritical Regimes in Epidemic Models of
Earthquake Aftershocks, J. Geophys. Res. 107, B10, Art. No. 2237.
1074 R. Shcherbakov et al. Pure appl. geophys.,

HELMSTETTER, A. and SORNETTE, D. (2003a), Predictability in the Epidemic-type Aftershock Sequence


Model of Interacting Triggered Seismicity, J. Geophys. Res. 108, B10, Art. No. 2482.
HELMSTETTER, A. and SORNETTE, D. (2003b), Båth’s Law Derived from the Gutenberg-Richter Law and
from Aftershock Properties, Geophys. Res. Lett. 30, 20, Art. No. 2069.
HELMSTETTER, A., SORNETTE, D., and GRASSO, J. R. (2003a), Mainshocks are Aftershocks of Conditional
Foreshocks: How do Foreshock Statistical Properties Emerge from Aftershock Laws, J. Geophys. Res.
108, B1, Art. No. 2046.
HELMSTETTER, A. S., OUILLON, G., and SORNETTE, D. (2003b), Are Aftershocks of Large Californian
Earthquakes Diffusing? J. Geophys. Res. 108, B10, Art. No. 2483.
HENDERSON, J. R., MAIN, I. G., MACLEAN, C., and NORMAN, M. G. (1994), A Fracture-mechanical
Cellular-automaton Model of Seismicity, Pure Appl. Geophys. 142, 3-4, 545–565.
HOSONO, K. and YOSHIDA, A. (2002), Do Large Aftershocks Decrease Similarly to Smaller Ones? Geophys.
Res. Lett. 29, 10.
HUANG, Y., SALEUR, H., SAMMIS, C., and SORNETTE, D. (1998), Precursors, Aftershocks, Criticality and
Self-organized Criticality, Europhys. Lett. 41, 1, 43–48.
KACHANOV, L. M., Introduction to Continuum Damage Mechanics (Martinus Nijhoff, Dordrecht, 1986).
KAGAN, Y. Y. (2002), Aftershock Zone Scaling, Bull. Seismol. Soc. Am. 92, 2, 641–655.
KAGAN, Y. Y. and KNOPOFF, L. (1981), Stochastic Synthesis of Earthquake Catalogs, J. Geophys. Res. 86,
2853–2862.
KING, G. C. P., STEIN, R. S., and LIN, J. (1994), Static Stress Changes and the Triggering of Earthquakes,
Bull. Seismol. Soc. Am. 84, 3, 935–953.
KISSLINGER, C., Aftershocks and fault-zone properties. In Advances in Geophysics, vol. 38 of Advances in
Geophysics (Academic Press, San Diego, 1996) pp. 1–36.
KISSLINGER, C. and JONES, L. M. (1991), Properties of Aftershock Sequences in Southern California,
J. Geophys. Res. 96, B7, 11947–11958.
KRAJCINOVIC, D., Damage Mechanics (Elsevier, Amsterdam, 1996).
KUN, F. and HERRMANN, H. J. (1991), Transition from Damage to Fragmentation in Collision of Solids,
Phys. Rev. E 59, 3, 2623–2632.
LAY, T. and WALLACE, T. C., Modern Global Seismology (Academic Press, San Diego, 1995).
LEMAITRE, J. and CHABOCHE, J.-L., Mechanics of Solid Materials (Cambridge University Press, Cambridge,
1990).
LI, V. C., SEALE, S. H., and CAO, T. Q. (1987), Postseismic Stress and Pore Pressure Readjustment and
Aftershock Distributions, Tectonophysics 144, 1–3, 37–54.
LOCKNER, D. A. and BEELER, N. M., Rock failure and earthquakes. In Inernational Handbook of Earthquake
and Engineering Seismology (W. H. K. Lee, H. Kanamori, P. C. Jennings, and C. Kisslinger, eds), vol.
Part A (Academic Press, Amsterdam, 2002) pp. 505–537.
LOCKNER, D. A., BYERLEE, J. D., KUKSENKO, J. D., PONOMAREV, V., and SIDORIN, A., Observations of
quasi-static fault growth from acoustic emissions. In Fault Mechanics and Transport Properties of Rocks
(Academic Press, London, 1992), pp. 3–31.
LYAKHOVSKY, V., BEN-ZION, Y., and AGNON, A. (2001) Earthquake Cycle, Fault Zones, and Seismicity
Patterns in a Theologically Layered Lithosphere, J. Geophys. Res. 106, B3, 4103–4120.
LYAKHOVSKY, V., BEN-ZION, Y., AGNON, A. (1997), Distributed Damage, Faulting, and Friction,
J. Geophys. Res. 102, B12, 27635–27649.
MARCELLINI, A. (1995), Arrhenius Behavior of Aftershock Sequences, J. Geophys. Res. 100, B4, 6463–6468.
MARCELLINI, A. (1997), Physical Model of Aftershock Temporal Behaviour, Tectonophysics 277, 1–3, 137–
146.
MENDOZA, C. and HARTZELL, S. H. (1998), Aftershock Patterns and Main Shock Faulting, Bull. Seismol.
Soc. Am. 78, 4, 1438–1449.
MOGI, K. (1962), Study of Elastic Shocks Caused by the Fracture of Hetergeneous Materials and its Relations
to Earthquake Phenomena, Bull. Earthquake Res. Inst. 40, 125–173.
MOLCHAN, G. M. and DMITRIEVA, O. E. (1992), Aftershock Identification: Methods and New Approaches,
Geophys. J. Int. 109, 3, 501–516.
MORENO, Y., CORREIG, A. M., GOMEZ, J. B., and PACHECO, A. F. (2001), A Model for Complex Aftershock
Sequences, J. Geophys. Res. 106, B4 , 6609–6619.
Vol. 162, 2005 Aftershock Statistics 1075

NAKATANI, M. (2001), Conceptual and Physical Clarification of Rate and State Friction: Frictional Sliding as
a Thermally Activated Rheology, J. Geophys. Res. 106, B7, 13347–13380.
NANJO, K., NAGAHAMA, H., and SATOMURA, M. (1998), Rates of Aftershock Decay and the Fractal
Structure of Active Fault Systems, Tectonophysics 287, 1–4, 173–186.
NARTEAU, C., SHEBALIN, P., and HOLSCHNEIDER, M. (2002), Temporal Limits of the Power-Law Aftershock
Decay Rate, J. Geophys. Res. 107, B12.
NUR, A. and BOOKER, J. R. (1971), Aftershocks Caused by Pore Fluid Pressure? Science 175, 885–887.
OMORI, F. (1894), On After-shocks of Earthquakes, J. Coll. Sci. Imp. Univ. Tokyo 7, 113–200.
REASENBERG, P. A. and JONES, L. M. (1989), Earthquake Hazard after a Main Shock in California, Science
243, 4895, 1173–1176.
REASENBERG, P. A. and JONES, L. M. (1994), Earthquake Aftershocks - Update, Science 265, 5176, 1251–
1252.
ROBERTSON, M. C., SAMMIS, C. G., SAHIMI, M., and MARTIN, A. J. (1995), Fractal Analysis of three-
dimensional Spatial Distributions of Earthquakes with a Percolation Interpretation, J. Geophys. Res. 100,
B1, 609–620.
RUINA, A. (1983), Slip Instability and State Variable Friction Laws, J. Geophys. Res. 88, NB12, 359–370.
RUNDLE, J., KLEIN, W., TURCOTTE, D. L., and MALAMUD, B. D. (2000), Precursory Seismic Activation and
Critical-point Phenomena, Pure Appl. Geophys. 157, 11-12, 2165–2182.
RUNDLE, J. B., KLEIN, W., and GROSS, S. (1996), Dynamics of a Traveling Density Wave Model for
Earthquakes, Phys. Rev. Lett. 76, 22, 4285–4288.
RUNDLE, J. B., KLEIN, W., and GROSS, S. (1999) Physical Basis for Statistical Patterns in Complex
Earthquake Populations: Models, Predictions and Tests, Pure Appl. Geophys. 155, 2–4, 575–607.
RYBICKI, K. (1973), Analysis of Aftershocks on the Basis of Dislocation Theory, Phys. Earth. Planet. Int. 7,
409–422.
SCHOLZ, C. H. The Mechanics of Earthquakes and Faulting, 2nd ed. (Cambridge University Press,
Cambridge, 2002).
SCORRETTI, R., CILIBERTO, S., and GUARINO, A. (2001), Disorder Enhances the Effects of Thermal Noise in
the Fiber Bundle Model, Europhys. Lett. 55, 5, 626–632.
SELINGER, R. L. B., WANG, Z. G., GELBART, W. M., and BENSHAUL, A. (1991), Statistical-Thermodynamic
Approach to Fracture, Phys. Rev. A 43, 8, 4396–4400.
SHCHERBAKOV, R. and TURCOTTE, D. L. (2003), Damage and Self-similarity in Fracture, Theor. Appl. Frac.
Mech. 39, 3, 245–258.
SHCHERBAKOV, R. and TURCOTTE, D. L. (2004a) A Modified Form of Båth’s Law, Bull. Seismol Soc. Am. in
press.
SHCHERBAKOV, R. and TURCOTTE, D. L. (2004b), A Damage Mechanics Model for Aftershocks, Pure Appl.
Geophys. 161, 1–13.
SHCHERBAKOV, R., TURCOTTE, D. L., and RUNDLE J. B. (2004), A Generalized Omori’s Law for Earthquake
Aftershock Decay, Geophys. Res. Lett. 31, L11613.
SORNETTE, A. and SORNETTE, D. (1999), Renormalization of Earthquake Aftershocks, Geophys. Res. Lett.
26, 13, 1981–1984.
SORNETTE, D. and ANDERSEN, J. V. (1998), Scaling with Respect to Disorder in Time-to-Failure, Eur. Phys.
J. B 1, 3 , 353–357.
STEIN, R. S. (1999), The Role of Stress Transfer in Earthquake Occurrence, Nature 402, 6762, 605–609.
TODA, S., STEIN, R. S., REASENBERG, P. A., DIETERICH, J. H., and YOSHIDA, A. (1998), Stress Transferred
by the 1995 Mw = 6.9 Kobe, Japan, Shock: Effect on Aftershocks and Future Earthquake Probabilities,
J. Geophys. Res. 103, B10, 24543–24565.
TSAPANOS, T. M. (1990), Spatial-distribution of the Difference between the Magnitudes of the Main Shock
and the Largest Aftershock in the Circum-Pacific Belt, Bull. Seismol. Soc. Am. 80, 5, 1180–1189.
TURCOTTE, D. L., Fractals and Chaos in Geology and Geophysics, 2nd ed. (Cambridge University Press,
Cambridge, 1997).
TURCOTTE, D. L., NEWMAN, W. I., and SHCHERBAKOV, R. (2003), Micro-and Macroscopic Models of Rock
Fracture, Geophys. J. Int. 152, 3, 718–728.
UTSU, T. (1961), A Statistical Study on the Occurrence of Aftershocks, Geophys. Mag. 30, 521–605.
1076 R. Shcherbakov et al. Pure appl. geophys.,

UTSU, T. Relationship between magnitude scales. In International Handbook of Earthquake and Engineering
Seismology (W. H. K. Lee, H. Kanamori, P. C. Jennings, and C. Kisslinger, eds.) vol. Part A, (Academic
Press, Amsterdam, 2002) pp. 733–746.
UTSU, T., OGATA, Y., and MATSUURA, R. S. (1995), The Centenary of the Omori Formula for a Decay Law
of Aftershock Activity, J. Phys. Earth 43, 1, 1–33.
VERE-JONES, D. (1969), A Note on the Statistical Interpretation of Båth’s Law, Bull. Seismol. Soc. Am. 59,
1535–1541.
WIEMER, S., GERSTENBERGER, M., and HAUKSSON, E. (2002), Properties of the Aftershock Sequence of the
1999 Mw 7.1 Hector Mine Earthquake: Implications for Aftershock Hazard. Bull. Seismol. Soc. Am. 92, 4,
1227–1240.
WIEMER, S. and KATSUMATA, K. (1998), Spatial Variability of Seismicity Parameters in Aftershock Zones,
J. Geophys. Res. 104, B6, 13,135–13,151.
WU, H. F., PHOENIX, S. L., and SCHWARTZ, P. (1988), Temperature-dependence of Lifetime Statistics for
Single Kevlar 49 Filaments in Creep-rupture, J. Mater. Sci. 23, 5, 1851–1860.
YAMANAKA, Y. and SHIMAZAKI, K. (1990), Scaling Relationship between the Number of Aftershocks and the
Size of the Main Shock, J. Phys. Earth 38, 4, 305–324.
YAMASHITA, T. and KNOPOFF, L. (1987), Models of Aftershock Occurrence, Geophys. J. Royal Astro. Soc.
91, 1, 13–26.
ZAPPERI, S., RAY, P., STANLEY, H. E., and VESPIGNANI, A. (1997), First-order Transition in the Breakdown
of Disordered Media, Phys. Rev. Lett. 78, 8, 1408–1411.
ZAPPERI, S., RAY, P., STANLEY, H. E., and VESPIGNANI, A. (1997), Avalanches in Breakdown and Fracture
Processes, Phys. Rev. E 59, 5, 5049–5057.
ZIV, A. and RUBIN, A. M. (2003), Implications of Rate-and-state Friction for Properties of Aftershock
Sequence: Quasi-static Inherently Discrete Simulations, J. Geophys. Res. 108, B1, Art. No. 2054.

(Received December 2, 2003, accepted July 20, 2004)

To access this journal online:


http://www.birkhauser.ch

View publication stats

You might also like