Nat Hazards (2014) 71:1749–1769

DOI 10.1007/s11069-013-0973-7

ORIGINAL PAPER

Bayesian probabilities of earthquake occurrences

in Shanxi rift system (China)

Ying Wang • Qinglong Zhang • Chao Liu

Received: 1 July 2013 / Accepted: 26 November 2013 / Published online: 6 December 2013
Ó Springer Science+Business Media Dordrecht 2013

Abstract China has a long history of earthquake records. The Shanxi rift system (SRS) is
situated along the axial zone of the domal uplift of the Shanxi Highlands and is the
boundary between the Ordos block and the North China Plain block. Strong earthquakes in
the SRS have been recorded since the thirteenth century. In our work, we applied the
Bayesian probability method using extreme value distribution of earthquake occurrences to
estimate the seismic hazard in the SRS. The seismic moment, slip rate, earthquake
recurrence rate, and magnitude were considered as the basic parameters for computing the
Bayesian prior estimates of the seismicity. These estimates were then updated in terms of
Bayes’ theorem and historical estimates of seismicity in the SRS. The probability of
occurrence of Ms ¼ 5:0 for Z1, Z2, and Z3 is less than 0.3, 0.1, and 0.6, respectively (T = 5
years). The probability of the occurrence of M  8.0 is small for the whole SRS. The
selection of upper bound magnitude probably influences the result, and the upper bound
magnitude of Z1, Z2, and Z3 may be 7.5, 7.0, and 8.5, respectively. We obtained the values
of the magnitude of completeness Mc (3.2) and the Gutenberg–Richter b value before
applying the Bayesian extreme value distribution of earthquake occurrences method.

Keywords Central North China  Bayesian probabilities  Seismic hazard

assessment  Earthquake occurrences

Y. Wang (&)  Q. Zhang

School of Earth Sciences and Engineering, Nanjing University, 22 Hankou Road, Nanjing 210093,
Jiangsu, China
e-mail: wangying025@gmail.com

Y. Wang
Exploration and Production Research Institute, SWPB of Sinopec, Jitai Road, Chengdu 610041,
Sichuan, China

C. Liu
Jinma Oilfield Development Company, Liaohe Oilfield of CNPC, Panjin 124010, Liaoning, China

123
1750 Nat Hazards (2014) 71:1749–1769

1 Introduction

Earthquakes are the most feared of all natural hazards; they occur with no warning and
can result in great destruction and loss of life (Parvez 2007). The assessment of
earthquake hazard is very important, yet at present, there is serious problem to estimate
the time, location and magnitude of future events that cannot be predicted reliably
(Parvez 2007) because of the incomplete understanding of the geophysical mechanism
that generates earthquakes (Stavrakakis and Drakopoulos 1995). However, earthquakes
can be described by statistical summaries or by the probabilities of the occurrence of
events of various magnitudes in the region under investigation (Stavrakakis and
Drakopoulos 1995). The Shanxi rift system (SRS) in North China is an active tectonic
zone that has experienced many intense, higher frequent, severely destructive earth-
quakes (Deng et al. 1973, Xu and Deng 1992). For example, September 17, 1303, a
strong earthquake occurred in Hongtong, Linfen (Xu and Deng 1993). The earthquake
killed more than 200,000 people, and it is the earliest identified M8 earthquake in China
(Xu and Deng 1993). Another strong earthquake attacked Linfen (M = 7.6) on May 18,
1695 (Wu et al. 1998), the earthquake was documented by 125 counties, and tens of
thousands of people died (Wang et al. 1993). It is only 45 km away from the two
epicenters (Ma et al. 1993). The situation that two successive great earthquakes hap-
pened within such a short distance in 400 years is unique in the 2,000 years of earth-
quake history in North China (Ma et al. 1993).
In October 1989, the M6.1 Datong–Gaoyang strong earthquake swarm occurrence
interrupted the 174 years of inactivity that had prevailed since the 1815 M6.75 earth-
quake. Following the 1989 earthquake, a number of M5 earthquakes occurred in close
sequence, possibly marking the beginning of a new seismic activity phase for the SRS
(Xu and Deng 1992). Jin and Liu (1999) presented a method that determined the pro-
portion of large to small earthquakes based on a Bayesian criterion. However, there are
no reports of parameters for Bayesian extreme value distribution of earthquake occur-
rence in the SRS.
The Bayesian statistics theory is useful for estimating earthquake parameters because of
two important properties (Yadav et al. 2013). First, Bayesian theory is a rigorous means of
combining prior information on seismicity, whether it is judgmental, geological or sta-
tistical, with historical observations of earthquake occurrences (Galanis et al. 2002). Its
second feature provides a means of incorporating the statistical uncertainty associated with
the estimation of the parameters used to quantify seismicity as well as the probabilistic
uncertainty associated with the inherent randomness of earthquake occurrence (Galanis
et al. 2002).
Campbell (1982, 1983) proposed a Bayesian extreme value distribution of earthquake
occurrence to evaluate the seismic hazard along the San Jacinto fault. The most
important aspect of the method is the updating of current probabilities when new
information becomes available (Parvez 2007). A similar procedure was applied by
Stavrakakis and Drakopoulos (1995) for a probabilistic prediction of strong earthquakes
in Greece. In China, the seismicity in the Yunnan Province and along the Jianchuan-
Dali seismic belt was estimated based on the Bayesian probabilistic theory by Cheng
and Wang (1986).
In the present study, we apply the Bayesian extreme value distribution method to
estimate the seismic hazard in the SRS. The seismic hazard is expressed in terms of the
probability of earthquake occurrence and the earthquake magnitude distribution.

123
Nat Hazards (2014) 71:1749–1769 1751

2 Outline of the Bayesian extreme value distribution model

According to Campbell (1982, 1983), the Bayesian probability that the largest earthquake
magnitude Mmax will occur within a period of t years and will exceed some specified
magnitude m is given by
 n00
t00
PðMmax [ m=tÞ ¼ 1  00 ð1Þ
t þ t½1  FðmÞ
where PðMmax [ m=tÞis the Bayesian representation of the probability that Mmax , the
largest earthquake expected to occur within a period of t years, will exceed some specified
magnitude m. In this expression, n00 is the posterior (updated) Bayesian estimate of the
number of earthquakes having magnitudes greater than some lower threshold ml , t00 is the
posterior (updated) Bayesian estimate of the time period over which those earthquakes
occur, and F ðmÞ is the Bayesian distribution of the earthquake magnitudes.
"  g00 #
00 m00
F ð mÞ ¼ K 1  ð2Þ
m00 þ m  ml

where
"  g00 #1
00 m00
K ¼ 1 00
ð3Þ
m þ mu  ml

In the above relationships, mu and ml are the upper and lower boundary of the earth-
quake magnitude for the region of interest, g00 is the posterior Bayesian estimate of the
number of earthquake occurrences having M [ ml , and m00 is the posterior Bayesian
estimate of the sum of the differences between the historically observed magnitudes and
ml .
The Bayesian distribution (Eq. (1)) can be used to combine the prior estimates of the
seismicity with historical earthquake occurrences.

2.1 Seismotectonic estimate of v and b

According to Shedlock et al. (1980) and Campbell (1982), the mean rate of occurrence of
earthquakes with magnitudes greater than m0 is given by
lAuðC2  b Þ h i1
0
0
v0 ¼ 0 M0 ðmu Þ10b ðmu m0 Þ  M0 ðm0 Þ ð4Þ
K0 b
where l is the shear modulus, A is the total area of the fault plane, u is the slip rate, M0 ðmu Þ
is the seismic moment of the upper bound magnitude, and M0 ðm0 Þ is the seismic moment
of m0. The parameter b0 represents the prior estimate of b from the relationship
log N ¼ a  bM ð5Þ
And the coefficient C2 is defined in the expression
logM0 ðmÞ ¼ C1 þ C2 m ð6Þ
0
In order to estimate v , the mean rate of occurrence of earthquakes greater than ml, we
must evaluate the expression

123
1752 Nat Hazards (2014) 71:1749–1769

v0 ¼ v00 ½1  FðmÞ ð7Þ

which, based on the doubly truncated distribution of magnitudes used to establish v00 ,
becomes

v0 ¼ v0 ð1  K0 f1  exp½b0 ðml  m0 ÞgÞ ð8Þ

The truncation factor K0 is given by Cornell and Vanmarcke (1969)

K0 ¼ f1  exp½b0 ðmu  m0 Þg1 ð9Þ

Since mu  m0 , the equations can be simplified considerably, resulting in the relationship,
0
lAu C2  b b0 ðmu ml Þ
v0 ¼ 10 ð10aÞ
M0 ðmu Þ b0
This relationship will be used to obtain the seismotectonic estimate of the mean rate of
earthquake occurrences having m [ ml . The seismotectonic estimate of the magnitude
frequency parameter b0 is computed by the relationships
b0 ¼ b ln 10 ð10bÞ

2.2 Posterior (updated) estimates of v and b

The updated estimates of v, the mean rate of earthquake occurrences having M [ ml, and
Vv, the coefficient of variation of v, are defined by the expressions
n00
v00 ¼ ð11Þ
t00
r00v 1
Vv00 ¼ ¼ pffiffiffiffiffi ð12Þ
v00 n00
where
v0
n00 ¼ n0 þ ð 0 Þ2 ð13Þ
rv
v0
t00 ¼ t0 þ ð14Þ
ðr0v Þ2
In the above expressions, t0 is the length of the historical record in years, v0 is the prior
mean estimate of v, r0v is the prior estimate of the standard deviation of v, and r is the
standard deviation defined by
0
rv
Vv0 ¼ ð15Þ
v0
or
r0b
Vb0 ¼ ð16Þ
b0
The updated estimates of b, the magnitude frequency parameter, and Vb , the coefficient
of the variation of b, are defined by

123
Nat Hazards (2014) 71:1749–1769 1753

g00
b00 ¼ ð17Þ
m00
r00b 1
Vb00 ¼ ¼ pffiffiffiffiffi00
00 ð18Þ
b g
where
!2
00 b0
g ¼ n0 þ ð19Þ
r0b

b0
m00 ¼ n0 ðm
  ml Þ þ ð20Þ
ðrb Þ2
0

After computing the seismotectonic, historical, and updated estimates of the seismicity
parameters ðv; bÞ, the Bayesian probability represented by Eq. (1) can then be computed
for the seismogenic zone of interest.

2.3 Application of the method in the Shanxi rift system (SRS)

Here, we apply the model described above to estimate the seismic hazard in the SRS. The
geological, topographical, and seismicity map of the SRS is shown in Fig. 1. The SRS is
situated along the axial zone of the domal uplift of the Shanxi Highlands (Xu et al. 1993).
Tectonically, the Shanxi Highlands are part of the North China Craton. The SRS is the
boundary between the Ordos block and the North China Plain active block (Zhang et al.
2003). The initial phase of the SRS rifting started in the Miocene at both the southern and
northern terminations of the rift, and the main phase of intense rifting and block faulting of
the whole rift system occurred during the Pliocene and the Quaternary (Xu et al. 1993).
China has a long history of earthquake records. The China earthquake catalogue is
produced by the China Earthquake Networks Center (CENC). Strong earthquake events
have been recorded in the SRS since the thirteenth century, and the historical earthquake
data are very valuable (Yi et al. 2004). Wen (2000, 2001) used these data to study strong
earthquake recurrence behavior and processes in the SRS.
For our study area, for the period 17 September 1303 to 28 June 2013, the CENC
catalog provides earthquake locations, occurrence times, and surface wave magnitudes Ms,
which we used to evaluate forecast models by applying the Bayesian extreme value dis-
tribution of earthquake occurrence. At present, the empirical relationship Ms 0:80 
Ml þ 0:83 can be used for the magnitude conversion in China (Bormann et al. 2007). The
entire region in the present study is divided into 3 seismogenic zones on the basis of
earthquake distribution density, earthquake intensity, and the relationship between the
earthquakes and the geological structures (Su and Wang 1992) (Fig. 1). All the seismo-
genic zones mapped have polygonal shapes. Therefore, the area of each zone has been
estimated using Surveyor’s formula giving vertices of the polygon (http://www.
zhongguosou.com/zonghe/ce_hui_mian_ji_ji_suan.aspx).
For the SRS, the slip rate was computed from the relationship
P
M0
u¼ ð21Þ
lt0 A
P
where M0 is the sum of the seismic moments of the earthquakes that occurred in the past
t0 years, A is the area of the fault slip, and l is the shear modulus (Brune 1968; Stavrakakis

123
1754 Nat Hazards (2014) 71:1749–1769

Fig. 1 Topographical and active structure map of the SRS, for computation of the Bayesian probabilities of
occurrence of earthquakes with Ms  5:0. Active structure information from Xu et al. (1993). The STRIM
data set is provided by International Scientific & Technical Data Mirror Site, Computer Network
Information Center, Chinese Academy of Sciences (http://datamirror.csdb.cn)

and Drakopoulos 1995). The shear modulus l is usually taken to be equal to

3  1011 dyn=cm2 , which is commonly adopted for the earth’s crust. The lower magnitude
ml considered in this study was Ms 5.0, and the time data is 710 years. The b values of the
Gutenberg–Richter formula logN ¼ a  bM were obtained by considering the events from
17 September 1303 to 28 June 2013 with magnitude greater than or equal to Ms 4:0 for Z1,
Z2, and Z3 (Fig. 2). The b value of the SRS is also given by considering all earthquakes in
the catalogue we used (Fig. 2). The magnitude of completeness Mc is the magnitude above
which all events are reported. Mc was estimated by fitting the Gutenberg–Richter (GR)
model to the observed frequency–magnitude distribution (FMD) (Mignan et al. 2013). The
commonly used, so-called standard, FMD-based mapping approach estimates Mc from the
FMD generated from events located in a cylindrical volume of fixed radius or from a fixed
number of events centered on each node of a spatial grid (Wiemer and Wyss 2000). This
approach has been used in numerous studies. Recent applications include Nanjo et al.
(2010) in Japan and Mignan et al. (2011) in Taiwan. In mainland China, standard Mc
mapping was applied in the Sichuan–Yunnan region for the period 1970–2001 by Su et al.
(2003) and in the Xinjiang region for the period 1970–2009 by Li et al. (2011). We used
the mapping method of Wiemer and Wyss (2000) and applied the FMD to the SRS for the
period 1303.7102–2013.4901. We consider events with depth \37 km, and the minimum
magnitude earthquake is Ms 2.2 in the catalogue (Fig. 3).

123
Nat Hazards (2014) 71:1749–1769 1755

Fig. 2 Magnitude–frequency relationship of the SRS and its subareas

The assigned upper bound magnitude is the observed maximum magnitude in each zone
during the period 1303.7102–2013.4901. The historically observed maximum magnitude in
each zone is also listed in Table 1.
The seismotectonic evaluation of the mean rate of earthquake occurrences v0 is per-
formed using Eq. (10a). The seismotectonic estimate of the magnitude frequency param-
eter b0 is obtained from the observed data and is computed using Eq. (10b). Then, the
posterior (updated) evaluations of the parameters v00 and b00 are calculated using Eqs. (11)
and (17). The output of the updating process for the estimation of the seismicity parameters
for the SRS is listed in Table 1.
The results of the updating process on the estimation of the seismicity parameters for
the SRS are summarized in Table 2.
We assume three values for the coefficients of variation Vv0 and Vb0 : 0.1, 0.25, and 1.0.
The Bayesian probabilities that a specific earthquake magnitude in the 3 zones will be
exceeded in t = 1, 5, 10, 20, 50, and 100 years were computed using Eq. (1), respectively.
The seismic moments were computed using the relationship lg M0 ¼ 16:1 þ 1:5M
(Hanks and Kanamori 1979) for earthquakes that occurred in the SRS. This is the first time
the Bayesian extreme value distribution of earthquake occurrence has been applied in the
SRS systematically. The results are presented graphically in Figs. 4, 5, 6, 7, 8, 9, 10, 11,
and 12. The plots show the probability of occurrences of events of magnitude 5 and above
0
for three values of Vv 0.1, 0.25, 1.0 and for t = 1, 5, 10, 20, 50, 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.

123
1756 Nat Hazards (2014) 71:1749–1769

Table 1 Parameters and summaries per zone of the observed data set in the Bayesian estimate of seismicity
for the SRS
Zone u (cm/year) A (km2) Max. observed mu 
m n0 b
historical magnitude

Z1 0.010 64,867.0868 1; 683:11:22  7:0 7:0; 7:5; 8:0; 8:5 5.7 32 0.54
Z2 0.004 20,543.5562 1; 618:05:20  6:5 7:0; 7:5; 8:0; 8:5 5.4 16 0.51
Z3 0.348 52,154.0101 1; 303:09:25  8:0 7:0; 7:5; 8:0; 8:5 5.7 31 0.35

Fig. 3 Frequency magnitude

distribution of the overall catalog
in the SRS. Plotted is both the
cumulative (squares) and non-
cumulative form (triangles)

2.4 Discussion and conclusions

The purpose of this study was to evaluate the seismic hazard in the SRS by means of the
Bayesian extreme value distribution of earthquake occurrence. We computed the magni-
tude of completeness (Mc ¼ 3:2) and the four b values of the SRS according to the
historical earthquake catalogue (1303.7102–2013.4901) along with other sources of
information such as the slip rate and the seismic moments of the events that occurred in the
SRS, and obtained the seismotectonic, historical, and updated estimates of the seismicity.
Many investigators have addressed the problem of assessing the seismic hazards in the
SRS. Based on the data of historical earthquakes and GPS observation, Yi et al. (2004)
estimated mean seismic moment rates and average recurrence intervals of strong earth-
quakes for Linfen basin (part of Z3), and its average recurrence interval for M = 7.5 is
estimated to be between 1,560 and 2,140 years. Our data show that the probability of
occurrence of Ms ¼ 7:5 for Z3 and mu ¼ 8:5 is 0.001–0.003. That may be because our
estimates area is larger.
In the present study, the cut-off magnitude for the analysis is 5.0 (which is commonly
taken as threshold of damaging earthquakes in seismic hazard studies) (Parvez 2007);
however, b values are obtained using the events with magnitude greater than or equal to
4.0, taking into account the completeness of the catalogue for the period of
1303.7102–2013.4901. The b values obtained for the 3 zones and the entire SRS region are
illustrated in Fig. 2. The other important parameter is the slip rate, which plays a very
important role in estimating seismotectonic mean rate of occurrence of earthquakes. In the
present study, the slip rate is obtained by using Brune’s model.

123
 Table 2 Prior and posterior estimates of v and b parameters for Z1, Z2, and Z3 using different upper bound magnitude thresholds
Z1 Z2 Z3

mu Parameters mu Parameters mu Parameters

0 0
7.0 v 0.10 0.10 0.10 7.0 v 0.01 0.01 0.01 7.0 v0 2.26 2.26 2.26
0 0 0
b 1.25 1.25 1.25 b 1.18 1.18 1.18 b 0.81 0.81 0.81
0 0 0
Vv ; Vb0 0.1 0.25 1.0 Vv ; Vb0 0.1 0.25 1.0 Vv ; Vb0 0.1 0.25 1.0
v00 0.08 0.06 0.05 v00 0.01 0.02 0.02 v00 0.17 0.07 0.05
b00 1.30 1.41 1.49 b00 1.27 1.56 2.17 b00 0.90 1.13 1.37
Nat Hazards (2014) 71:1749–1769

Vv00 ; Vb00 0.09 0.14 0.17 Vv00 ; Vb00 0.09 0.18 0.24 Vv00 ; Vb00 0.09 0.15 0.18
0 0 0
7.5 v 0.03 0.03 0.03 7.5 v 0.00 0.00 0.00 7.5 v 0.60 0.60 0.60
0 0 0
b 1.25 1.25 1.25 b 1.18 1.18 1.18 b 0.81 0.81 0.81
0 0 0
Vv ; Vb0 0.1 0.25 1.0 Vv ; Vb0 0.1 0.25 1.0 Vv ; Vb0 0.1 0.25 1.0
v00 0.04 0.04 0.04 v00 0.00 0.01 0.02 v00 0.15 0.06 0.04
b00 1.30 1.41 1.49 b00 1.27 1.56 2.17 b00 0.90 1.13 1.37
Vv00 ; Vb00 0.09 0.14 0.17 Vv00 ; Vb00 0.09 0.18 0.24 Vv00 ; Vb00 0.09 0.15 0.18
0 0 0
8.0 v 0.01 0.01 0.01 8.0 v 0.00 0.00 0.00 8.0 v 0.16 0.16 0.16
0 0 0
b 1.25 1.25 1.25 b 1.18 1.18 1.18 b 0.81 0.81 0.81
0 0 0
Vv ; Vb0 0.1 0.25 1.0 Vv ; Vb0 0.1 0.25 1.0 Vv ; Vb0 0.1 0.25 1.0
00 00 00
v 0.01 0.02 0.04 v 0.00 0.00 0.01 v 0.10 0.06 0.04
b00 1.30 1.41 1.49 b00 1.27 1.56 2.17 b00 0.90 1.13 1.37
Vv00 ; Vb00 0.09 0.14 0.17 Vv00 ; Vb00 0.09 0.18 0.24 Vv00 ; Vb00 0.09 0.15 0.18
0 0 0
8.5 v 0.00 0.00 0.00 8.5 v 0.00 0.00 0.00 8.5 v 0.04 0.04 0.04
0 0 0
b 1.25 1.25 1.25 b 1.18 1.18 1.18 b 0.81 0.81 0.81
0 0 0
Vv ; Vb0 0.1 0.25 1.0 Vv ; Vb0 0.1 0.25 1.0 Vv ; Vb0 0.1 0.25 1.0
00 00 00
v 0.00 0.01 0.03 v 0.00 0.00 0.01 v 0.04 0.04 0.04
b00 1.30 1.41 1.49 b00 1.27 1.56 2.17 b00 0.90 1.13 1.37
Vv00 ; Vb00 0.09 0.14 0.17 Vv00 ; Vb00 0.09 0.18 0.24 Vv00 ; Vb00 0.09 0.15 0.18
1757

123
1758 Nat Hazards (2014) 71:1749–1769

Fig. 4 Probabilities of occurrence of earthquakes for mu = 7.0, 7.5, 8.0, and 8.5 with Ms  5:0 in Z1 for
t = 1 and 5 years. The left graph corresponds to t = 1 year and the right one to t = 5 years

In the present study, the slip rate is obtained by using Brune’s model, which is appropriate
along the major plate boundaries (Tselentis et al. 1988); however, there are several methods to
calculate the A. Brune (1968) computed the area of the fault slip of the Imperial Valley and used

123
Nat Hazards (2014) 71:1749–1769 1759

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 = 10 and 20 years. The left graph corresponds to t = 10 years and the right one to t = 20 years

the equation A ¼ Length  Depth, and he believed width and depth are both acceptable.
Stavrakakis and Campbell (1983) got A of the Jacinto fault zone using A ¼ L  W. The width,
measured along the dip of the fault plane, may be computed from the relationship, w ¼ sinH a,

123
1760 Nat Hazards (2014) 71:1749–1769

Fig. 6 Probabilities of occurrence of earthquakes for mu = 7.0, 7.5, 8.0, and 8.5 with Ms  5:0 in Z1 for
t = 50 and 100 years. The left graph corresponds to t = 50 years and the right one to t = 100 years

where a is the average dip of the fault plane as measured from the horizontal plane, and H is
the thickness of the seismogenic zone (Campbell 1983). Stavrakakis and Drakopoulos (1995)
used the equation A ¼ L  W; where L and W are length and width of the zone. Parvez (2007)

123
Nat Hazards (2014) 71:1749–1769 1761

Fig. 7 Probabilities of occurrence of earthquakes for mu = 7.0, 7.5, 8.0, and 8.5 with Ms  5:0 in Z2 for
t = 1 and 5 years. The left graph corresponds to t = 1 year and the right one to t = 5 years

obtained A of each study zone using Surveyor’s formula giving vertices of polygon. Obvi-
ously, different calculation methods result in different results; the A computed by Surveyor’s
formula giving vertices of polygon is probably the maximum.

123
1762 Nat Hazards (2014) 71:1749–1769

Fig. 8 Probabilities of occurrence of earthquakes for mu = 7.0, 7.5, 8.0, and 8.5 with Ms  5:0 in Z2 for
t = 10 and 20 years. The left graph corresponds to t = 10 years and the right one to t = 20 years

Given the potential of earthquake hazard in the Shanxi rift and adjacent regions,
several Chinese agencies have conducted a series of Global Positioning System (GPS)
field campaigns since 1992 (He et al. 2003). The earlier results indicated around

123
Nat Hazards (2014) 71:1749–1769 1763

Fig. 9 Probabilities of occurrence of earthquakes for mu = 7.0, 7.5, 8.0, and 8.5 with Ms  5:0 in Z2 for
t = 50 and 100 years. The left graph corresponds to t = 50 years and the right one to t = 100 years

0.10–0.19 cm/year slip rates for the SRS (Li et al. 2003; Yang et al. 2003). Recent
research suggests the slip rate of the SRS is 0.06 ± 0.05 cm/year (Wang et al. 2011). The
isotope analysis results indicated 0.03–0.09 cm/year slip rates in the Z1 and around

123
1764 Nat Hazards (2014) 71:1749–1769

Fig. 10 Probabilities of occurrence of earthquakes for mu = 7.0, 7.5, 8.0, and 8.5 with Ms  5:0 in Z3 for
t = 1 and 5 years. The left graph corresponds to t = 1 year and the right one to t = 5 years

0.5 cm/year in the Z3 (Xu and Ma 1992). We can see that the rates of slip obtained
(Table 1) are in approximate agreement with rates obtained from other research methods.
Some of these values are bigger than our data. That could be because some earthquakes

123
Nat Hazards (2014) 71:1749–1769 1765

Fig. 11 Probabilities of occurrence of earthquakes for mu = 7.0, 7.5, 8.0, and 8.5 with Ms  5:0 in Z3 for
t = 10 and 20 years. The left graph corresponds to t = 10 years and the right one to t = 20 years

123
1766 Nat Hazards (2014) 71:1749–1769

Fig. 12 Probabilities of occurrence of earthquakes for mu = 7.0, 7.5, 8.0, and 8.5 with Ms  5:0 in Z3 for
t = 50 and 100 years. The left graph correspond, to t = 50 years and the right one to t = 100 years

123
Nat Hazards (2014) 71:1749–1769 1767

(M \ 3.2) may not be recorded in the long historical process and if we adopt Campbell
1983’s method to calculate A, the u values should get bigger without regard to the
problems existing in the other methods.
The probability of occurrence of Ms ¼ 5:0 for Z1, Z2, and Z3 is less than 0.3, 0.1, and
0.6, respectively (T = 5 years). The probability of the occurrence of M  8:0 is small for the
whole SRS (although significantly higher magnitudes have been historically observed).
0
Z1, probability of occurrence of Vv0 ¼ 0:1 is bigger than that of Vv ¼ 0:25 and Vv0 ¼ 1:0
0
for mu = 7.0 regardless the t value; the value of Vv ¼ 0:1; 0:25; 1:0 is almost coincident for
mu = 7.5 regardless the t value
Z2, from big to small, the order of probability of occurrence for Vv0 is 1, 0.25, 0.1 for
mu = 7.5, 8.0, 8.5. The probability of occurrence is the lowest for the same magnitude
earthquake if other parameters are the same.
Z3, from big to small, the order of probability of occurrence for Vv0 is 0.1, 0.25, 1.0 for
mu = 7.0, 7.5, 8.0, 8.5 except when mu = 8.5, Ms ¼ 5:0 and t = 1, 5, 10, 20. The prob-
ability of occurrence of Ms ¼ 5:0 for Z3 is greater than 0.3, 0.6, 0.9 for t = 5, 10, 20, 50,
respectively.
The selection of upper bound magnitude probably influences the result (Figs. 4, 5, 6, 7,
8, 9, 10, 11, 12), and the upper bound magnitude of Z1, Z2, and Z3 may be 7.5, 7.0, and 8.5,
respectively. For the probability of occurrence of the same magnitude, earthquake is very
close for the different Vv0 value besides considering the historically observed maximum
magnitude in each zone.
As mentioned by other authors, the coefficient of the variation of seismic rate Vv is a
very critical parameter of the Bayesian model (Campbell 1983, Stavrakakis and Drako-
poulos 1995). This coefficient may bias the result in favor of either the prior estimate or the
historical data. For Vv ¼ 1:0, the historical earthquake occurrences control the seismic
hazard, whereas for Vv ¼ 0:0, the obtained probabilities are based only on the seismo-
tectonic estimates (Campbell 1983; Parvez 2007). Therefore, we analyzed the results for
three different values of Vv ¼ 0:1; 0:25; and 1:0.
An important feature of the Bayesian distribution proposed in this study is its ability to
combine prior estimates of seismicity with historical earthquake occurrences in computing
seismic hazards (Campbell 1983). However, the sensitivity of the Bayesian extreme value
distribution to a certain number of parameters does not diminish (Parvez 2007). Other
factors such as cumulative stress are not considered. Statistical estimates of seismicity
become unreliable when there are very few historical data on which to base them (Campbell
1983). Information from different sources should be incorporated with the available his-
torical seismicity data within the Bayesian statistics framework to assess the reliability of
the seismic hazard within a seismogenic zone (Stavrakakis and Drakopoulos 1995).

Acknowledgments This work is financially supported by the National Natural Science Foundation of
China (ID 40634021) and the National Science and Technology Major Project of China (ID 2008ZX05035-
006-001). Ph.D. candidate Hong Ye and Ph.D. candidate Cai Deng offered their help in completing the
manuscript. We are also indebted to the two anonymous reviewers, who provided valuable comments and
that significantly improved the original manuscript.

References

Bormann P, Liu RF, Ren X, Gutdeutsch R, Kaiser D, Castellaro S (2007) Chinese national network
magnitudes, their relation to NEIC magnitudes, and recommendations for new IASPEI magnitude
standards. Bull Seism Soc Am 97:114–127

123
1768 Nat Hazards (2014) 71:1749–1769

Brune JN (1968) Seismic moment, seismicity, and rate of slip along major fault zones. J Geophys Res
73:777–784
Campbell KW (1982) Bayesian-analysis of extreme earthquake occurrences. Part 1. Probabilistic hazard
model. Bull Seism Soc Am 72(5):1689–1705
Campbell KW (1983) Bayesian-analysis of extreme earthquake occurrences. Part 2. Application to the San
Jacinto fault zone of Southern California. Bull Seism Soc Am 73(4):1099–1115
Cheng DL, Wang ZZ (1986) Bayesian estimation of seismicity. J Seism Res 9(2):133–146 (in Chinese with
English abstract)
Cornell CA, Vanmarcke EH (1969) The major influences on seismic risk. In: Proceedings of the Fourth
World Conference on Earthquake Engineering, vol 1. Santiago, Chile, pp 69–83
Deng QD, Wang KL, Wang YP, Tang HJ, Wu YW, Ding ML (1973) On the tendency of seismicity and their
geological set up of the seismic belt of Shanxi Graben. Scientia Geol Sin 1:37–74 (in Chinese)
Galanis OC, Tapanos TM, Papadopoulos GA, Kiratzi AA (2002) Bayesian extreme values distribution for
seismicity parameters assessment in South America. J Balk Geophys Soc 5:77–86
Hanks TC, Kanamori H (1979) A moment magnitude scale (in fault mechanics). J Geophys Res
84(B5):2348–2350
He JK, Liu M, Li YX (2003) Is the Shanxi rift of northern China extending? Geophys Res Lett 30(23):SDE
1–SDE 7
Jin XS, Liu ZH (1999) A new method to determine the proportion of large to small earthquakes and its
application in earthquake prediction. J Seism Res 22(3):260–264 (in Chinese with English abstract)
Li YX, Yang GH, Li Z, Guo QL, Huang C, Zhu WY, Fu Y, Wang Q, Jiang ZS, Wang M (2003) Active
block movement and strain state of Chinese mainland. Sci china Ser D 33(Suppl):65–81 (in Chinese)
Li Z, Jiang C, Huang Y, Wang B (2011) Minimum magnitude of completeness for earthquake catalogue and
scientific layout of seismic stations in Xinjiang region. Acta Seism Sin 33:763–775 (in Chinese with
English abstract)
Ma ZJ, Deng QD, Liu GX, Zhao XP, Liu GD, Ma BL, Zhou KS, Jiang P, Yin ZQ, Gao ZH, Chen JY, Miao
LT (1993) Linfen region seismic study, earthquake pre-evaluation and prevention, Shanxi. In: Ma ZJ
(ed) Seismic study and disaster mitigation in Linfen, Shanxi. Seismol Press, Beijing, pp 1–11
Mignan A, Werner MJ, Wiemer S, Chen CC, Wu YM (2011) Bayesian estimation of the spatially varying
completeness magnitude of earthquake catalogs. Bull Seism Soc Am 101:1371–1385
Mignan A, Jiang C, Zechar JD, Wiemer S, Wu Z, Huang Z (2013) Completeness of the Mainland China
earthquake catalog and implications for the setup of the China Earthquake Forecast Testing Center.
Bull Seism Soc Am 103(2A):845–859
Nanjo KZ, Ishibe T, Tsuruoka H, Schorlemmer D, Ishigaki Y, Hirata N (2010) Analysis of the completeness
magnitude and seismic network coverage of Japan. Bull Seism Soc Am 100:3261–3268
Parvez IA (2007) On the Bayesian analysis of the earthquake hazard in the North-East Indian peninsula. Nat
Hazards 40(2):397–412
Shedlock KM, McGuire RK, Herd DG (1980) Earthquake recurrence in the San Francisco Bay region,
California, from Fault Slip and seismic moments. U.S. Geol Surv. Open-File Rept 80-999
Stavrakakis GN, Drakopoulos J (1995) Bayesian probabilities of earthquake occurrences in Greece and
surrounding areas. Pure Appl Geophys 144(2):307–319
Su YZ, Wang JY (1992) Seismic activity on the Fen–Wei seismic zone. Inland Earthq 6:143–150 (in
Chinese with English abstract)
Su YJ, Li YL, Li ZH, Yi GX, Liu LF (2003) Analysis of minimum completeness magnitude of earthquake
catalog in Sichuan–Yunnan region. J Seism Res 26(S1):10–16
Tselentis GA, Stavrakakis G, Makropoulos K, Latousakis J, Drakopoulos J (1988) Seismic moments of
earthquakes at the western Hellenic arc and their application to the seismic hazard of the area.
Tectonophysics 148:73–82
Wang TM, Zheng BH, Li XY, Wang YP (1993) Seismotectonic study of the 1695 Linfen 71/2 earthquake.
In: Ma ZJ (ed) Seismic study and disaster mitigation in Linfen, Shanxi. Seismol Press, Beijing,
pp 172–189
Wang H, Liu M, Cao JL, Shen XH, Zhang GM (2011) Slip rates and seismic moment deficits on major
active faults in mainland China. J Geophys Res 116(B02405):1–17
Wen XZ (2000) Influence of the 1303 great earthquake rupture, Shanxi on earthquake recurrence behavior
of its adjacent fault segments. Earthq Res China 16(01):25–30 (in Chinese with English abstract)
Wen XZ (2001) Earthquake behavior of variable rupture-scale on active faults and application of the
cascade-rupturing model. Acta Seism Sin 14(04):380–390 (in Chinese with English abstract)
Wiemer S, Wyss M (2000) Minimum magnitude of completeness in earthquake catalogs: examples from
Alaska, the western United States, and Japan. Bull Seism Soc Am 90(4):859–869

123
Nat Hazards (2014) 71:1749–1769 1769

Wu L, Qi SQ, Wang RP (1998) 1303 Hongtong M = 8 earthquake and 1695 Linfen M = 8 earthquake. In:
Guo JZ, Ma ZJ (eds) Great earthquake earthquake research in China. Press, Beijing, Seismol, pp 6–35
Xu XW, Deng QD (1992) Study on strong earthquake activity and risk areas in the Shanxi graben system.
Seism Geol 14(4):305–316 (in Chinese with English abstract)
Xu XW, Deng QD (1993) Seismotectonic of the 1303 Hongtong earthquake. In: Ma ZJ (ed) Seismic study
and disaster mitigation in Linfen, Shanxi. Seismol Press, Beijing, pp 149–158
Xu XW, Ma XY (1992) Geodynamics of the Shanxi rift system, China. Tectonophysics 208:325–340
Xu XW, Ma XY, Deng QD (1993) Neotectonic activity along the Shanxi rift system, China. Tectonophysics
219(4):305–325
Yadav R, Tsapanos TM, Bayrak Y, Koravos GC (2013) Probabilistic appraisal of earthquake hazard
parameters deduced from a Bayesian approach in the northwest frontier of the Himalayas. Pure Appl
Geophys 170(3):283–297
Yang GH, Han YP, Wang M (2003) Characteristics of the horizontal crustal movement of North China in
the last decade. Earthq Res China 19(4):324–333 (in Chinese with English abstract)
Yi GX, Wen XZ, Xu XW (2004) Average recurrence intervals of strong earthquakes and potential risky
segments along the Taiyuan–Linfen portion of the Shanxi graben system. Acta Seism Sin
26(4):387–395 (in Chinese with English abstract)
Zhang PZ, Deng QD, Zhang GM, Ma J, Gan WJ, Min W, Mao FY, Wang Q (2003) Strong earthquake
activity and active blocks in China mainland. Sci china Ser D 33(Suppl):12–20 (in Chinese)

123