J Seismol (2015) 19:69–82

DOI 10.1007/s10950-014-9451-2

ORIGINAL ARTICLE

Bayesian probabilities of earthquake occurrences

in Longmenshan fault system (China)
Ying Wang & Keyin Zhang & Qigang Gan & Wen Zhou &
Liang Xiong & Shihua Zhang & Chao Liu

Received: 3 March 2014 / Accepted: 21 July 2014 / Published online: 8 August 2014
# Springer Science+Business Media Dordrecht 2014

Abstract China has a long history of earthquake re- the empirical relationship of magnitude conversion for
cords, and the Longmenshan fault system (LFS) is a Ms and ML, the values of the magnitude of complete-
famous earthquake zone. We believed that the LFS ness Mc (3.5), and the Gutenberg–Richter b value be-
could be divided into three seismogenic zones (north, fore applying the Bayesian extreme-value distribution
central, and south zones) based on the geological struc- of earthquake occurrences method.
tures and the earthquake catalog. We applied the
Bayesian probability method using extreme-value dis- Keywords Seismic hazard assessment . Earthquake
tribution of earthquake occurrences to estimate the seis- occurrences . Bayesian probabilities . Magnitude
mic hazard in the LFS. The seismic moment, slip rate, conversion . Longmen Mountain . Western Sichuan
earthquake recurrence rate, and magnitude were consid-
ered as the basic parameters for computing the Bayesian
prior estimates of the seismicity. These estimates were
1 Introduction
then updated in terms of Bayes’ theorem and historical
estimates of seismicity in the LFS. Generally speaking,
Earthquakes are the most feared of all natural hazards;
the north zone seemingly is quite peaceful compared
they occur with no warning and can result in great
with the central and south zones. The central zone is the
destruction and loss of life (Parvez 2007).
most dangerous; however, the periodicity of earthquake
Longmenshan fault system (LFS) is about 500 km long
occurrences for M s = 8.0 is quite long (1,250 to
and 30–50 km wide (Li et al. 2010), which is a famous
5,000 years). The selection of upper bound probable
earthquake zone. The Wenchuan Ms 8.2 earthquake
magnitude influences the result, and the upper bound
(http://data.earthquake.cn/data/) caused enormous
magnitude of the south zone maybe 7.5. We obtained
economic losses, led to nearly 70,000 known fatalities
Y. Wang (*) : K. Zhang : Q. Gan : L. Xiong : S. Zhang
with thousands missing, and destroyed the homes of
Exploration & Production Research Institute, SWPB of more than 1.5 million people (Stone 2008). Another
Sinopec, Chengdu 610041 Sichuan, China strong earthquake (Ms = 7.0) hit Lushan County of
e-mail: wangying025@gmail.com Ya’an City on April 20, 2013. The earthquake killed
Y. Wang : W. Zhou
196 people, more than 10,000 people injured (Chen
State Key Laboratory of Reservoir Geology and Development et al. 2013).
Engineering, Chengdu University of Technology, The assessment of earthquake hazard is very impor-
Chengdu 610059 Sichuan, China tant, yet at present, there is a serious problem to estimate
C. Liu
the time, location, and magnitude of future events that
Jinma Oilfield Development Company, Liaohe Oilfield of cannot be predicted reliably (Parvez 2007) because of
CNPC, Panjin 124010 Liaoning, China the incomplete understanding of the geophysical
70 J Seismol (2015) 19:69–82

mechanism that generates earthquakes. However, earth- parameters for Bayesian extreme-value distribution of
quakes can be described by statistical summaries or by earthquake occurrence in the LFS. In the present study,
the probabilities of the occurrence of events of various we apply the Bayesian extreme-value distribution meth-
magnitudes in the region under investigation od to estimate the seismic hazard in the LFS. The
(Stavrakakis and Drakopoulos 1995). The Bayesian sta- seismic hazard is expressed in terms of the probability
tistics theory is useful for estimating earthquake param- of earthquake occurrence and the earthquake magnitude
eters because of two important properties (Pisarenko distribution. The geological, topographical, and seismic-
et al. 1996; Pisarenko and Lyubushin 1997; Lyubushin ity map of the LFS is shown in Fig. 1.
et al. 2002; Lyubushin and Parvez 2010; Yadav et al.
2013). First, Bayesian theory is a rigorous means of
combining prior information on seismicity, whether it
is judgmental, geological, or statistical, with historical 2 Method and data
observations of earthquake occurrences (Galanis et al.
2002). The second feature is that the method provides a 2.1 Bayesian extreme-value distribution
means of incorporating the statistical uncertainty asso-
ciated with the estimation of the parameters used to According to Campbell (1982, 1983), the Bayesian
quantify seismicity as well as the probabilistic uncer- probability that the largest earthquake magnitude Mmax
tainty associated with the inherent randomness of earth- will occur within a period of t years and will exceed
quake occurrence (Galanis et al. 2002). Campbell some specified magnitude m is given by
(1982; 1983) proposed a Bayesian extreme-value distri-
n″
t″
bution of earthquake occurrence to evaluate the seismic PðM max > m=t Þ ¼ 1− ð1Þ
t″ þ t ½1−F ðmÞ
hazard. A similar procedure was applied by Wang et al.
(2014a) for a probabilistic prediction of strong earth- where P(Mmax >m/t)is the Bayesian representation of
quakes in Shanxi rift system (China). the probability that Mmax, the largest earthquake ex-
The LFS can be divided into two or three sections on pected to occur within a period of t years, will
the basis of the geological structures (Jia et al. 2006; Li exceed some specified magnitude m. In this ex-
et al. 2008; Meng et al. 2013) and has a very compli- pression, n″ is the posterior (updated) Bayesian
cated geologic tectonic and evolutionary history. The estimate of the number of earthquakes having
LFS experienced at least two major periods of contrac- magnitudes greater than some lower threshold ml,
tion deformation in the Late Triassic and the Cenozoic t" is the posterior (updated) Bayesian estimate of
(Burchfiel et al. 1995; Jia et al. 2006). Tectonically, the the time period over which those earthquakes oc-
NE-trending LFS lies along the eastern boundary of the cur, and F(m) is the Bayesian distribution of the
Tibetan Plateau, which is bounded by the north China, earthquake magnitudes.
south China, and Qiangtang blocks in the tectonic col- The Bayesian distribution (Eq. 1) can be used
lage of China (Mattauer et al. 1985; Yin and Nie 1996). to combine the prior estimates of the seismicity
The LFS consists of a series of NW dipping thrust faults, with historical earthquake occurrences. Previous
most of which are associated with fault-related folds (Li researchers (Campbell 1982, 1983; Stavrakakis
et al. 2006b; Burchfiel et al. 2008; Royden et al. 2008). and Drakopoulos 1995; Pisarenko et al. 1996;
From NW to SE, these thrust faults correspond to the Pisarenko and Lyubushin 1997; Parvez 2007;
Maowen fault, the Beichuan fault, the Pengguan fault, Wang et al. 2014a) described the formulation of
and the Piedmont fault (Deng et al. 1994; Li et al. 1995). the Bayesian extreme-value distribution in detail,
Some researches (Li et al. 2006a; Yi et al. 2006; and we could get the parameters of seism-tectonic
Burchfiel et al. 2008; Zhang et al. 2009b; Zhang and (prior) and posterior (updated) evaluations of seis-
Li 2010; Xie et al. 2012) have reported the seismic micity by the formulations mentioned in those
activity of the LFS, only three M 6.0–M 6.5 were doc- papers.
umented in the central and southern of the LFS but no Generally speaking, the temporal occurrence of
earthquake larger than M 6.0 happened in the northern earthquakes may be represented by a Poisson process
of the LFS (Wen 1995) before 2008 Wenchuan Ms 8.2 if it is assumed that earthquakes are independent random
earthquake. However, there are few reports of events and no two events can occur at the same instant in
J Seismol (2015) 19:69–82 71

Fig. 1 Topographical and active structure map of the LFS for the (1994) and Li et al. (2008, 2014); the STRIM data set is provided
computation of the Bayesian probabilities of occurrence of earth- by Computer Network Information Center, Chinese Academy of
quakes with Ms ≥5.0. Structure information from Deng et al. Sciences (http://www.gscloud.cn/csearch.jsp)

time (Campbell 1982). Campbell (1982) proposed that account for the statistical uncertainty in the esti-
the distribution may be obtained by the expression: mation of ν, Eq. 2 is more accurately represented
by the Bayesian (compound) distribution
ðνt Þn e−νt (Benjamin 1968; Benjamin and Cornell 1970;
PðN ¼ njν; t Þ ¼ ð2Þ
n! Campbell 1982) and may be obtained by evaluat-
ing the integral equation:
where P(N=n|ν,t) is the probability that the num-
ber of earthquakes occurring within a specified
    
period of time t will be equal to n, given that  ∞ 
P N ¼ nt ¼ ∫0 P N ¼ nν; t f ″ðν Þdν ð3Þ
the mean rate of earthquake occurrences is ν. To
72 J Seismol (2015) 19:69–82

where f "(ν) represents the posterior probability den- differences between their magnitude and m l ,
sity function of ν, updated from the prior distribution of respectively.
ν by incorporating through Bayes’ theorem observations
on the occurrence of earthquakes (i.e., the no. of occur- 2.1.1 Seism-tectonic estimate of v and β
rences within a specified period of time).
By assuming that the earthquake occurrences are a In order to estimate v' (the prior mean of v) for the
Poisson process and that the uncertainty in ν may be number of earthquakes with magnitudes greater than
represented by a gamma distribution, Cornell (1972), ml, Campbell (1982) and Stavrakakis and Drakopoulos
Campbell (1977), and Mortgat and Shah (1979) have (1995) proposed that the following relationship could be
shown by application of Bayes’ theorem that f"(ν) may used:
be represented by another gamma distribution:
0
μAu C 2 −b b0 ðmu −ml Þ
″
f ðν Þ ¼ K 1 ν n″−1 −νt″
ð4Þ v0 ¼ 0 10 ð8Þ
e M 0 ðm u Þ b

where the normalizing constant K1 = t" n"/ Г(n") and where μ is the shear modulus, A is the total area of the
whereГ(n") represents the gamma function with fault plane, u is the slip rate, and M0(mu) is the seismic
parameter n″ (Campbell 1982). Based on Eq. 3, moment of the upper bound magnitude. The parameter
Campbell (1982) finally found that: b' represents the prior estimate of b from the relationship

   Гðn þ n″Þ
t″ n″
t n log N ¼ a−bM ð9Þ


P N ¼ nn″; t″; t ¼ ð5Þ
n!Гðn″Þ t þ t″ t þ t″ And the coefficient C2 is defined in the expression

which is the Bayesian Poisson–gamma distribution logM 0 ¼ C 1 þ C 2 m ð10Þ

because the derivation of Eq. 5 was based on a Poisson
The seism-tectonic estimate of the magnitude–fre-
distribution of earthquake occurrences and a gamma
quency parameter β' is computed by Eq. 9 and the
distribution for ν. Equation 5 yields the probability of
relationship
n earthquakes occurring in t years taking in to account
the uncertainty concerning the Poisson distribution pa- β 0 ¼ b ln10 ð11Þ
rameter, that is the uncertainty about the rate of occur-
rence (Galanis et al. 2002).
Consistent with the treatment of ν, Campbell (1982) 2.1.2 Posterior (updated) estimates of v and β
accounted for the uncertainty in magnitude–frequency
parameter β and proposed the final expressions for the The updated estimates of v (v″), the mean rate of earth-
doubly truncated Bayesian exponential–gamma distri- quake occurrences having M>ml, and Vv (Vv "), the
bution of earthquake magnitude coefficient of variation of v, are defined by the following
expressions:
"
η″ #
  
 m″ n
00
F mml ; mu ¼ K″ 1− ð6Þ v 00 ¼ ð12Þ
m″ þ m−ml t 00

where 1
V v 00 ¼ pffiffiffiffiffi ð13Þ
"
η″ #−1 n00
m″
K″ ¼ 1− ð7Þ where
m″ þ mu −ml

2
00 v0
n ¼ n0 þ ð14Þ
In the above relationships, mu and ml are the upper σ0v
and lower boundary of the earthquake magnitude for the
region of interest, η" and m" represent updated Bayesian
v0
evaluation of the number of earthquake occurrences t 00 ¼ t 0 þ : ð15Þ
greater than the minimum value ml and the sum of the ð σ0 vÞ 2
J Seismol (2015) 19:69–82 73

In the above expressions, n0 is the number of earth- Wang et al. (2013) used earthquake catalog data (BC193
quakes observed within a time period of t0 years, t0 is the to AD2008) which got the spatial and temporal distri-
length of the historical record in years, σ'vis the prior bution characteristics of the b value in Sichuan and
estimate of the standard deviation of v, and they are Yunnan province.
defined by the expression In our case, for the period 21 April 1657 to 31
0 October 2013, the CENC catalog provides earth-
0 σ
Vv ¼ 0 ð16Þ quake locations, occurrence times, and some sur-
v face wave magnitudes Ms in the LFS, which we
or used to evaluate forecast models by applying the
Bayesian extreme-value distribution of earthquake
σ 0β
Vβ 0 ¼ : ð17Þ occurrence. At present, the empirical relationship
β0 Ms ←0.80 × ML+0.83 can be used for the magni-
The updated estimates of β (β"), the magnitude– tude conversion in China (Bormann et al. 2007).
frequency parameter, and Vβ (Vβ "), the coefficient of However, Ms ←1.01×ML−0.12 probably more ap-
the variation of β, are defined by propriate for the LFS as the mean difference and
the standard deviation of the estimated value and
η″ measured is smaller (Table 1). We obtained the
β″ ¼ ð18Þ
m″ empirical relationship Ms = 1.01ML − 0.12 by the
earthquakes which both have the Ms and ML for
the same event from 1970 to 2013 in our study
1 area (Fig. 2). So, we use our empirical relationship
V β ″ ¼ pffiffiffiffiffi ð19Þ
η″ for the magnitude conversion in the LFS.
Some researchers divided the LFS into six
where seismogenic zones based on the earthquake distribution
!2 density, earthquake intensity, and other earthquake pa-
β0
η ″ ¼ n0 þ 0 ð20Þ rameter (Yi et al. 2006). In fact, the LFS could be
σβ divided into three seismogenic zones (Fig. 1). We be-
lieve the three seismogenic zones can reflect the feature
of the geological structures and seismic activity of the
0
β LFS better, especially after the 2008 Wenchuan Ms 8.2
m″ ¼ n0 ðm̄ − ml Þ þ  2 ð21Þ and 2013 Lushan Ms 7.0 earthquake. The relationship
0
σβ between the earthquakes and all the seismogenic zones
mapped has polygonal shapes. Therefore, the area of
where m  is the mean magnitude of the historically each zone has been estimated using Surveyor’s formula
observed earthquakes and σ'βis the prior estimate of the giving vertices of the polygon (http://www.
standard deviation of β. After computing the seism- zhongguosou.com/zonghe/ce_hui_mian_ji_ji_suan.
tectonic, historical, and updated estimates of the seis- aspx).
micity parameters(v, β), the Bayesian probability repre- The Bayesian probabilities that a specific earthquake
sented by Eq. 1 can then be computed for the magnitude in the three zones will be exceeded in T=5,
seismogenic zone of interest. 20, and 100 years were computed, respectively.

2.2 Earthquake catalog and seismogenic zones 2.3 Parameters

China has a long history of earthquake records. The The magnitude of completeness Mc is the magni-
historical earthquake data are very valuable. The China tude above which all events are reported. Mc was
earthquake catalog is produced by the China Earthquake estimated by fitting the Gutenberg–Richter (GR)
Networks Center (CENC). Chen et al. (2010) used model to the observed frequency–magnitude distri-
earthquake catalog data (1970 to 2008) to study charac- bution (FMD) (Mignan et al. 2013). The common-
teristics of earthquake frequency distribution in the LFS. ly used, so-called standard, FMD-based mapping
74 J Seismol (2015) 19:69–82

Table 1 The comparison of the two empirical relationship For the LFS, the slip rate was computed from the
Empirical relationship Mean difference Standard deviation relationship:
of the predicted of the predicted and
and measured measured
∑M 0
Bormann et al. 2007 0.115 0.251 u¼ ð22Þ
Ms ←0.80×ML+0.83 μt 0 A
This paper −0.005 0.226
Ms ←1.01×ML−0.12 where ∑M0 is the sum of the seismic moments
of the earthquakes that occurred in the past t0
Events number: 1,595; period: 1970 to 2013 years, A is the area of the fault slip, and μ is
the shear modulus (Brune 1968; Stavrakakis and
approach estimates Mc from the FMD generated Drakopoulos 1995). The seismic moments were
from events located in a cylindrical volume of computed using the relationship logM0 =16.1 +
fixed radius or from a fixed number of events 1.5M (Hanks and Kanamori 1979) for earthquakes
centered on each node of a spatial grid (Wiemer that occurred in the LFS. The shear modulus μ is
and Wyss 2000). This approach has been used in usually taken to be equal to 3 × 1011 dyn/cm 2,
numerous studies. Recent applications of the meth- which is commonly adopted for the Earth’s crust.
od have been carried out around the world (Nanjo The lower magnitude m l considered in this
et al. 2010; Mignan et al. 2011). The standard Mc study was Ms5.0, and the time data is 356 years.
mapping was also has been applied in mainland The assigned upper bound magnitude is the ob-
China (Su et al. 2003; Li et al. 2011). We used the served maximum magnitude in each zone during
mapping method of Wiemer and Wyss (2000) and the period 1657.3014 to 2013.7636. The historical-
applied the FMD to the LFS for the period ly observed maximum magnitude in each zone is
1630.0411 to 2013.8289. We consider events with also listed in Table 2.
the minimum magnitude earthquake that is Ms2.1
in the catalog (Fig. 3).
The b values of the Gutenberg–Richter formula log
N=a−bM were obtained by considering the events from 3 Results
1657.3014 to 2013.8289 with magnitude greater than or
equal to Ms 4.0 for Z1, Z2, and Z3 (Figs. 1 and 4). The b We assume three values for the coefficients of variation
value of the LFS is also given by considering all earth- Vv ' and Vβ ': 0.1, 0.25, and 1.0. The results of the
quakes in the catalog we used (Fig. 4). updating process on the estimation of the seismicity
parameters for the LFS are summarized in Table 3 (v'
is the seism-tectonic evaluation of the mean rate of
earthquake occurrences, β' is the seism-tectonic esti-
mate of the magnitude–frequency parameter,
v" and β"are the posterior (updated) evaluations of the
parameters, and V"v and V"β represent the coefficients of
variation of the mean rate of occurrence and magnitude–
frequency parameter, respectively).
The results are presented graphically in Figs 5,
6, and 7. The plots show the probability of occur-
rences of events of magnitude 5.0 and above for
the three values of V'v 0.1, 0.25, 1.0, and for T=5,
20, and 100. In all the figures, the probability is
decaying gradually, while the decay is sharp in the
vicinity of the maximum (upper bound) magnitude.
Fig. 2 Magnitude conversion for Ms and ML (earthquake catalog As the decrease of m u and increase of T, the
from the CENC, http://data.earthquake.cn/data/) probability of event occurrence is increase.
J Seismol (2015) 19:69–82 75

Fig. 3 Frequency magnitude distribution. a LFS and b LFS and adjacent region, the overall catalog plotted is both the cumulative (squares)
and noncumulative form (triangles)

Fig. 4 Magnitude–frequency relationship of the LFS and its subareas

76 J Seismol (2015) 19:69–82

Table 2 Parameters and summaries of the observed data set in the Bayesian estimate of seismicity for the LFS

Zone u(cm/yr) A(km2) Max. observed historical magnitude mu 

m n0 b

Z1 0.013 15,121.4518 05 August 2008—6.5 7.0, 7.5, 8.0, 8.5 5.3 32 0.92
Z2 1.687 14,058.0287 12 May 2008—8.2 7.0, 7.5, 8.0, 8.5 5.5 37 0.67
Z3 0.032 13,425.7389 20 April 2013—7.0 7.0, 7.5, 8.0, 8.5 5.6 11 0.63

4 Discussion and conclusions test show that there is an inverse relationship between
differential stress Δσ and b, and the b value can there-
In this study, the cutoff magnitude for the analysis is 5.0 fore be interpreted as a “stressmeter” in the Earth’s crust
(which is commonly taken as threshold of damaging (Schorlemmer et al. 2005). So, the b value can be used
earthquakes in seismic hazard studies) (Parvez 2007); to measure the regional stress (Wiemer and
however, b values are obtained using the events with Schorlemmer 2007).
magnitude greater than or equal to 4.0, taking into The slip rate is another important parameter, which
account the completeness of the catalog for the period plays a very important role in estimating seismotectonic
of 1657.3014–2013.7636 for the LFS. Results of rock mean rate of occurrence of earthquakes. Almost no

Table 3 Prior and posterior estimates of v and β parameters for Z1, Z2, and Z3 using different upper bound magnitude thresholds

Z1 Z2 Z3

mu Parameters mu Parameters mu Parameters

7.0 v' 0.07 0.07 0.07 7.0 v' 4.84 4.84 4.84 7.0 v' 0.08 0.08 0.08
β' 2.12 2.12 2.12 β' 1.54 1.54 1.54 β' 1.45 1.45 1.45
'
Vv,Vβ ' 0.10 0.25 1.00 Vv' ,Vβ ' 0.10 0.25 1.00 Vv' ,Vβ ' 0.10 0.25 1.00
"
v" 0.07 0.08 0.09 v" 0.36 0.15 0.11 v 0.07 0.05 0.03
β" 2.32 2.80 3.28 β" 1.64 1.84 1.98 β" 1.47 1.53 1.65
Vv",Vβ" 0.09 0.14 0.17 Vv",V"β 0.09 0.14 0.16 V"v,V"β 0.09 0.19 0.29
7.5 v' 0.03 0.03 0.03 7.5 v' 1.86 1.86 1.86 7.5 v' 0.03 0.03 0.03
β' 2.12 2.12 2.12 β' 1.54 1.54 1.54 β' 1.45 1.45 1.45
Vv' ,Vβ ' 0.10 0.25 1.00 Vv' ,Vβ ' 0.10 0.25 1.00 V'v,Vβ ' 0.10 0.25 1.00
" "
v 0.04 0.06 0.09 v 0.33 0.15 0.11 v" 0.03 0.03 0.03
β" 2.32 2.80 3.28 β" 1.64 1.84 1.98 β" 1.47 1.53 1.65
V"v,V"β 0.09 0.14 0.17 V"v,V"β 0.09 0.14 0.16 V"v,V"β 0.09 0.19 0.29
' '
8.0 v 0.02 0.02 0.02 8.0 v 0.72 0.72 0.72 8.0 v' 0.01 0.01 0.01
β '
2.12 2.12 2.12 β '
1.54 1.54 1.54 β '
1.45 1.45 1.45
Vv' ,Vβ ' 0.10 0.25 1.00 V 'v,Vβ ' 0.10 0.25 1.00 Vv' ,Vβ ' 0.10 0.25 1.00
v" 0.02 0.04 0.08 v" 0.28 0.14 0.11 v" 0.01 0.01 0.03
β" 2.32 2.80 3.28 β" 1.64 1.84 1.98 β" 1.47 1.53 1.65
Vv",Vβ" 0.09 0.14 0.17 V"v,V"β 0.09 0.14 0.16 V"v,V"β 0.09 0.19 0.29
' '
8.5 v' 0.01 0.01 0.01 8.5 v 0.28 0.28 0.28 8.5 v 0.00 0.00 0.00
β' 2.12 2.12 2.12 β' 1.54 1.54 1.54 β' 1.45 1.45 1.45
Vv' ,Vβ ' 0.10 0.25 1.00 Vv' ,Vβ ' 0.10 0.25 1.00 Vv' ,Vβ ' 0.10 0.25 1.00
v" 0.01 0.02 0.07 v" 0.19 0.13 0.11 v" 0.00 0.01 0.02
β" 2.32 2.80 3.28 β" 1.64 1.84 1.98 β" 1.47 1.53 1.65
Vv",V"β 0.09 0.14 0.17 V"v,V"β 0.09 0.14 0.16 Vv",V"β 0.09 0.19 0.29
J Seismol (2015) 19:69–82 77

Fig. 5 Probabilities of occurrence of earthquakes for mu =7.0, 7.5, 8.0, and 8.5 with Ms ≥5.0 in Z1 for T=5, 20, and 100 years. The left
graphs correspond to T=5 years, the middle graphs correspond to T=20 years, and the right ones to T=100 years

relative movement between the Sichuan basin and the giving vertices of polygon is probably the maximum
LFS before the Wenchuan earthquake (Zhang et al. (Wang et al. 2014a). In addition, the boundary of the
2008). Therefore, the Wenchuan earthquake surprised seismic zone is unlikely to be straight lines, so it is only
seismologists for its intensity and suddenness (Zhang an approximation. Using the seism-geology method
et al. 2009a). In the present study, the slip rate is obtain- (dating of disconnected landscape), Xu et al. (2005,
ed by using Brune’s model, which is appropriate along 2008) got the slip rate of the LFS is 2–3 mm/year in
the major plate boundaries (Tselentis et al. 1988). the late Quaternary. And the slip rate is no more than
However, there are several methods to calculate the A. 3 mm/year in 10,000 years’ time scale (Zhang et al.
Obviously, different calculation methods result in dif- 2008). GPS measurements before the Wenchuan earth-
ferent results; the A computed by Surveyor’s formula quake suggest that slip rate across the entire LFS does
78 J Seismol (2015) 19:69–82

Fig. 6 Probabilities of occurrence of earthquakes for mu =7.0, 7.5, 8.0, and 8.5 with Ms ≥5.0 in Z2 for T=5, 20, and 100 years. The left
graphs correspond to T=5 years, the middle graphs correspond to T=20 years, and the right ones to T=100 years

not exceed about 2 mm/year (Zhang et al. 2008). We can respectively (T=5 years). The probability of the
see that the rates of slip obtained from the north and occurrence of Ms =8.0 for Z1, Z2, and Z3 is less
south zones (Table 2) are in approximate agreement than 0.00005, 0.0041, and 0.0007, respectively
with the rates obtained from other research methods. (T=5 years). Generally speaking, the north zone
However, the slip rate of the central zone is considerably is seemingly quite peaceful compared with the
larger than that of other research methods. That may be central and south zones. The central zone is the
mainly because of the 2008 Wenchuan M s 8.2 most dangerous; however, the periodicity is quite
earthquake. long (1,250 to 5,000 years). The study results of
The probability of occurrence of Ms =5.0 for Z1, other research methods show that the period of the
Z2, and Z3 is less than 0.36, 0.84, and 0.29, LFS ranged from 2,000 to 13,000 years (Li et al.
J Seismol (2015) 19:69–82 79

Fig. 7 Probabilities of occurrence of earthquakes for mu =7.0, 7.5, 8.0, and 8.5 with Ms ≥5.0 in Z3 for T=5, 20, and 100 years. The left
graphs correspond to T=5 years, the middle graphs correspond to T=20 years, and the right ones to T=100 years

2006a; Densemore et al. 2007), which more or less Ms =5.0 for Z2 is 0.41–0.84, 0.87–1.00, and 1.00–1.00
agree with the estimates of this paper. for T=5, 20, and 100 years, respectively.
In Z1, probability of occurrence for Ms =5.0 of V'v = In Z3, from big to small, the order of probability of
0.1 is the smallest, V'v =1 is the biggest for mu =7.0, 7.5, occurrence for Ms =5.0 of V'v is 0.1, 0.25, and 1 for mu =
8.0, and 8.5 regardless of the T value. The probability of 7.0 and 1, 0.25, and 0.1 for mu =7.5, 8.0, and 8.5. The
occurrence of Ms =5.0 for Z1 is 0.06–0.36, 0.20–0.82, probability of occurrence of Ms =5.0 for Z3 is 0.02–
and 0.68–1.00 for T=5, 20, and 100 years, respectively. 0.29, 0.08–0.75, and 0.35–1.00 for T = 5, 20, and
In Z2, from big to small, the order of probability of 100 years, respectively.
occurrence for Ms =5.0 of V'v is 0.1, 0.25, and 1 for mu = The selection of upper bound probable magnitude
7.0, 7.5, 8.0, and 8.5. The probability of occurrence of influences the result (Figs 5, 6, and 7), and the upper
80 J Seismol (2015) 19:69–82

bound magnitude of Z3 maybe 7.5. For the probability Burchfiel BC, Royden LH, Van RD, Der Hilst BHH (2008) A
geological and geophysical context for the Wenchuan earth-
of occurrence of the same magnitude, earthquake is very
quake of 12 May 2008, Sichuan, People’s Republic of China.
close for the different V'v value besides considering the GSA Today 18(7):4–11
historically observed maximum magnitude in the zone. Campbell KW (1977) The use of seismotectonics in the Bayesian
As mentioned by other authors, the coefficient of the estimation of seismic risk, School of Engineering and
Applied Science, University of California, Los Angeles,
variation of seismic rate Vv is a very critical parameter of
Report UCLA-ENG-7744.
the Bayesian model (Campbell 1983; Stavrakakis and Campbell KW (1982) Bayesian-analysis of extreme earthquake
Drakopoulos 1995). This coefficient may bias the result occurrences. Part 1. Probabilistic hazard model. Bull Seismol
in favor of either the prior estimate or the historical data. Soc Am 72(5):1689–1705
Campbell KW (1983) Bayesian-analysis of extreme earthquake
We analyzed the results for three different values of Vv =
occurrences. Part 2. Application to the San Jacinto fault zone
0.1, 0.25,and 1.0. For Vv =1.0, the historical earthquake of southern California. Bull Seismol Soc Am 73(4):1099–
occurrences control the seismic hazard, whereas for Vv = 1115
0, the obtained probabilities are based only on the Chen XZ, Guo XY, Li YE (2010) Distribution characteristics of
the month—scale number for earthquakes occurred on the
seismotectonic estimates (Campbell 1983; Parvez
Longmenshan fault. Earthquake 30(2):20–28
2007). An important feature of the Bayesian distribution Chen YT, Yang ZX, Zhang Y, Liu C (2013) From Wenchuan
proposed in this study is its ability to combine prior earthquake to Lushan earthquake. Sci Sin Terrae 43(6):
estimates of seismicity with historical earthquake occur- 1064–1072
rences in computing seismic hazards (Campbell 1983). Cornell CA (1972) Bayesian statistical decision theory and
reliability-based design. In: Freudenthal AM (ed)
However, the sensitivity of the Bayesian extreme-value Proceedings of the International Conference on Structural
distribution to a certain number of parameters does not Safety and Reliability, April 9–11, 1969. Smithsonian
diminish (Parvez 2007). Information from different Institute, Washington
sources should be incorporated with the available his- Deng QD, Chen SF, Zhao XL (1994) Tectonics, scismisrry and
dynamics of Longmenshan mountains and its adjacent re-
torical seismicity data within the Bayesian statistics gions. Seismol Geol 16(4):389–403 (in Chinese with English
framework to assess the reliably of the seismic hazard abstract)
within a seismogenic zone (Stavrakakis and Densemore AL, Ellis M, Li Y (2007) Active tectonics of the
Drakopoulos 1995). To some extent, especially the Beichuan and Pengguan faults at the eastern margin of the
Tibetan Plateau. Tectonics 26:TC4005
structural geology, the structures usually record the Galanis OC, Tapanos TM, Papadopoulos GA, Kiratzi AA (2002)
complex tectonic history and influence the earthquake Bayesian extreme values distribution for seismicity parame-
distribution (cf. Wang et al. 2014b). ters assessment in South America. J Balk Geophys Soc 5:77–
86
Hanks TC, Kanamori H (1979) A moment magnitude scale (in
Acknowledgments This work is financially supported by the fault mechanics). J Geophys Res 84(B5):2348–2350
National Science and Technology Major Project of China Jia D, Wei GQ, Chen ZX, Li BL, Qing Z, Guang Y (2006)
(ID:2011ZX05002-004). We thank the anonymous reviewers for Longmen Shan fold-thrust belt and its relation to the western
the fruitful criticism of this study. Sichuan Basin in central China: new insights from hydrocar-
bon exploration. AAPG Bull 90(9):1425–1447
Li Y, Zeng YF, Yi HS (1995) Sedimentary and tectonic evolution
of the Longmenshan foreland basin, western Sichuan, China.
References Press of Chengdu University of Science and Technology,
Chengdu (in Chinese)
Li Y, Zhou RJ, Densemore AL, Ellis MA (2006a) Geodynamic
Benjamin JR (1968) Probabilistic models for seismic force design. processes of eastern margin of the Tibetan Plateau and its
J Struct Div ASCE 94(ST5):1175–1196 geological responses. Geological Publishing House, Beijing
Benjamin JR, Cornell CA (1970) Probability, statistics and deci- (in Chinese)
sion for civil engineers. McGraw-Hill, New York Li Y, Zhou RJ, Densmore AL, Ellis MA (2006b) Geomorphic
Bormann P, Liu R, Ren X, Gutdeutsch R, Kaiser D, Castellaro S evidence for the late Cenozoic strike-slipping and thrusting in
(2007) Chinese national network magnitudes, their relation to Longmen Mountain at the eastern margin of the Tibetan
NEIC Magnitudes, and recommendations for new IASPEI Plateau. Quat Sci 26(1):40–51 (in Chinese with English
magnitude standards. Bull Seismol Soc Am 97(1B):114–127 abstract)
Brune J (1968) Seismic moment, seismicity, and rate of slip along Li ZW, Liu SG, Chen HD, Liu S, Guo B, Tian XB (2008)
major fault zones. J Geophys Res 73:777–784 Structural segmentation and zonation and differential defor-
Burchfiel B, Zhiliang C, Yupinc L, Royden L (1995) Tectonics of mation across and along the Lomgmen thrust belt, west
the Longmen Shan and adjacent regions, central China. Int Sichuan, China. J Chengdu Univ Technol (Sci Technol)
Geol Rev 37(8):661–735 35(4):440–454 (in Chinese with English abstract)
J Seismol (2015) 19:69–82 81

Li YQ, Jia D, Shaw J, Hubbard J, Lin AM, Wang MM, Luo L, Li Stone R (2008) An unpredictably violent fault. Science 320:1578–
HB, Wu L (2010) Structural interpretation of the coseismic 1580
faults of the Wenchuan earthquake: three-dimensional Su YJ, Li YL, Li ZH, Yi GX, Liu LF (2003) Analysis of minimum
modeling of the Longmen Shan fold‐and‐thrust belt. J completeness magnitude of earthquake catalog in Sichuan-
Geophys Res 115:B04317 Yunnan Region. J Seismol Res 26(S1):10–16 (in Chinese
Li Z.H, Jiang CS, Huang Y, W. BZ (2011) Minimum magnitude of with English abstract)
completeness for earthquake catalogue and scientific layout Tselentis GA, Stavrakakis G, Makropoulos K, Latousakis J,
of seismic stations in Xinjiang Region, Acta Seismol Sin Drakopoulos J (1988) Seismic moments of earthquakes at
33(6):763–775. (in Chinese with English abstract) the western Hellenic arc and their application to the seismic
Li YQ, Jia D, Wang MM, Shaw J, He JK, Lin AM, Lin X, Rao G hazard of the area. Tectonophysics 148:73–82
(2014) Structural geometry of the source region for the 2013 Wang H, Cao JL, Jing Y, Li Z (2013) Spatiotemporal pattern of b-
Mw 6.6 Lushan earthquake: implication for earthquake haz- value before major earthquakes in the Sichuan-Yunnan re-
ard assessment along the Longmen Shan. Earth Planet Sci gion. Seismol Geol 34(3):531–543 (in Chinese with English
Lett 390:275–286 abstract)
Lyubushin AA, Parvez IA (2010) Map of seismic hazard of India Wang Y, Zhang QL, Liu C (2014a) Bayesian probabilities of
using Bayesian approach. Nat Hazards 55(2):543–556 earthquake occurrences in Shanxi rift system (China). Nat
Lyubushin AA, Tsapanos TM, Pisarenko VF, Koravos GC (2002) Hazards 71(3):1749–1769
Seismic hazard for selected sites in Greece: a Bayesian esti- Wang Y, Zhang QL, Zhu WB, Wang LS, Xie GA, Liu C, Zou X
mates of seismic peak ground acceleration. Nat Hazards (2014b) Extension structural records in the Qinshui Basin
25(1):83–89 (north China) since the Late Mesozoic. Int J Earth Sci. doi:10.
Mattauer M, Matte PH, Malavieille J, Tapponnier P, Maluski H, 1007/s00531-014-1050-5
Xu Z, Lu Y, Tang Y, Lu Y, Tang Y (1985) Tectonics of the Wen XZ (1995) Quantitative assessments of earthquake
Qinling belt: build-up and evolution of eastern Asia. Nature potential on active faults. Seismological Press, Beijing
317:496–500 (in Chinese)
Meng W, Chen QC, Wu ML, Li GQ, Qin XH, Feng CJ (2013) Wiemer S, Schorlemmer D (2007) ALM: an asperity-based like-
Research on segmentation and characteristic of tectonic stress lihood model for California. Seismol Res Lett 78(1):134–140
field of Longmenshan fault zone. Prog Geophys 28(3):1150– Wiemer S, Wyss M (2000) Minimum magnitude of completeness
1160 (in Chinese with English abstract) in earthquake catalogs: examples from Alaska, the western
Mignan A, Werner MJ, Wiemer S, Chen CC, Wu YM (2011) United States, and Japan. Bull Seismol Soc Am 90(4):859–
Bayesian estimation of the spatially varying completeness 869
magnitude of earthquake catalogs. Bull Seismol Soc Am Xie ZJ, Lv YJ, Zhang LF (2012) Application of the quantitative
101(3):1371–1385 analysis of moderate-small earthquakes for the division of the
Mignan A, Jiang C, Zechar JD, Wiemer S, Wu Z, Huang Z (2013) seismic zones—taking Longmenshan Seismic Belt and its
Completeness of the mainland China earthquake catalog and adjacent region as an example. Northwest Seismol J 34(3):
implications for the wetup of the China earthquake forecast 277–283 (in Chinese with English abstract)
testing center. Bull Seismol Soc Am 103(2A):845–859 Xu XW, Zhang PZ, Wen XZ, Qin ZL, Chen GH, Zhu AL (2005)
Mortgat CP, Shah HC (1979) A Bayesian model for seismic Features of active tectonics and recurrence behaviors of
hazard mapping. Bull Seismol Soc Am 69:1237–1251 strong earthquakes in the western Sichuan Province and its
Nanjo KZ, Ishibe T, Tsuruoka H, Schorlemmer D, Ishigaki Y, adjacent regions. Seismol Geol 27(3):446–461
Hirata N (2010) Analysis of the completeness magnitude Xu XW, Wen XZ, Chen GH, Yu GH (2008) Discovery of the
and seismic network coverage of Japan. Bull Seismol Soc Longriba faults, eastern part of the Bayankela tectonic block
Am 100(6):3261–3268 and its geodynamic implications. Sci China Earth Sci 38(5):
Parvez IA (2007) On the Bayesian analysis of the earthquake 529–542 (in Chinese)
hazard in the north-east Indian Peninsula. Nat Hazards Yadav R, Tsapanos TM, Bayrak Y, Koravos GC (2013)
40(2):397–412 Probabilistic appraisal of earthquake hazard parameters de-
Pisarenko VF, Lyubushin AA (1997) Statistical estimation of duced from a Bayesian approach in the northwest frontier of
maximum peak ground acceleration at a given point of seis- the Himalayas. Pure Appl Geophys 170(3):283–297
mic region. J Seismol 1(4):395–405 Yi GX, Wen XZ, Wang SW, Long F, Fan J (2006) Study on fault
Pisarenko VF, Lyubushin AA, Lysenko VB, Golubeva TV (1996) sliding behaviors and strong-earthquake risk of the
Statistical estimation of seismic hazard parameters: maximal Longmenshan-Minshan fault zones from current seismicity
possible magnitude and related parameters. Bull Seismol Soc parameters. Earthq Res China 22(2):117–125 (in Chinese
Am 86(3):691–700 with English abstract)
Royden LH, Burchfiel BC, Van der Hilst RD (2008) The geolog- Yin A, Nie S (1996) A Phanerozoic palinspastic reconstruction of
ical evolution of the Tibetan Plateau. Science 321(5892): China and its neighboring regions. In: Yin A, Harrison M
1054–1058 (eds) The tectonic evolution of Asia. Cambridge University
Schorlemmer D, Wiemer S, Wyss M (2005) Variations in earth- Press, New York
quake—size distribution across different stress regimes. Zhang YQ, Li HL (2010) Late Quaternary active faulting along the
Nature 43:539–542 SW segment of the Longmenshan fault zone. Quat Sci 30(4):
Stavrakakis GN, Drakopoulos J (1995) Bayesian probabilities of 699–710 (in Chinese with English abstract)
earthquake occurrences in Greece and surrounding areas. Zhang PZ, Xu XW, Wen XZ, Ran YK (2008) Slip rates and
Pure Appl Geophys 144(2):307–319 recurrence intervals of the Longmen Shan active fault zone,
82 J Seismol (2015) 19:69–82

and tectonic implications for the mechanism of the May 12 Tectonic model of the great Wenchuan earthquake of
Wenchuan earthquake, 2008, Sichuan, China. Chin J May 12, 2008, Sichuan, China. Chin Sci Bull 54(7):
Geophys 51(4):1066–1073 (in Chinese with English 944–953 (in Chinese)
abstract) Zhang ZW, Cheng WZ, Ruan X, Wu P (2009b) Seismicity and
Zhang PZ, Wen XZ, Xu XW, Gan WJ, Wang M, Shen ZK, tectonic stress of the Longmenshan fault zone before 2008
Wang QL, Huang Y, Zheng Y, Li XJ, Zhang ZQ, Ma Wenchuan Ms 8.0 earthquake. Acta Seismol Sin 31(2):117–
SL, Ran YK, Liu QY, Ding ZF, Wu JP (2009a) 127+235 (in Chinese with English abstract)