A detailed view of the injection-induced seismicity in a natural gas
reservoir in Zigong, southwestern Sichuan Basin, China
Xinglin Lei,1,2Shengli Ma,1Wenkang Chen,3Chunmei Pang,3Jie Zeng,3and Bing Jiang3
Received 28 December 2012; revised 25 July 2013; accepted 27 July 2013; published 20 August 2013.
 Seismicity ata gasreservoirlocated in the relatively stable Sichuan Basin, China,mirrors
the injection pressure of unwanted water, suggesting that the seismicity is injection induced.
Injection under high pressure on a routine basis began on 9 January 2009 and continued to
July 2011. During the injection period, over 120,000 m3of water was pumped under a
wellhead pressure of up to 6.2 MPa into the limestone formation of Permian 2.45 to 2.55 km
beneath the surface. The injection induced more than 7000 surface-recorded earthquakes,
including 2 M4+ (the largest one was ML4.4), 20 M3+, and more than 100 M2+ events. Data
observedbyanearby localseismicnetworkand fivetemporalstationsprovideadetailedview
of the spatiotemporal distribution of the induced earthquakes. Most events were limited to
In a map view, hypocenters are concentrated in a NNW extended ellipsoidal zone
approximately 6 km long and approximately 2 km wide centered approximately at the
injection well. Multisources of evidence such as the shear mechanism, pattern of hypocenter
region show that the induced earthquakes occurred as a result of lowering of the effective
normalstresson known or unknownpreexisting blind faults which are critically loaded under
the regional stress field. Epidemic-type aftershock sequence modeling results indicate that
injection inducing and earthquake triggering are both important during earlier periods of
injection, while later periods are dominated by forced (injection-induced) seismicity.
Citation: Lei, X., S. Ma, W. Chen, C. Pang, J. Zeng, and B. Jiang (2013), A detailed view of the injection-induced
seismicity in a natural gas reservoir in Zigong, southwestern Sichuan Basin, China, J. Geophys. Res. Solid Earth, 118,
 Significant seismicity induced or triggered in gas/oil
reservoirs by pore pressure increase due to fluid injection or
stress due to depletion is known from different regions [e.g.,
Kanamori and Hauksson, 1992; Dahm et al., 2007; Lei
et al., 2008]. The number of cases of such seismicity has
increased over the past decade as the amount of deep fluid
injection has risen. Although rare, with the recent increase in
fluid injections, noticeable and damaging earthquakes have
become more common and have attracted greater attention
[Kerr, 2012]. Several damaging earthquakes of moderate size
(approximately M4–5) have been documented [Healy et al.,
1968; Raleigh et al., 1976; Davis and Pennington, 1989;
Ahmad and Smith, 1988; Nicholson and Wesson, 1992; Ake
et al., 2005; Majer and Peterson, 2007; Lei et al., 2008;
Wang et al., 2012]. In areas of low tectonic seismic activity,
very shallow earthquakes of greater than M3 might be dama-
ging and thus have a very strong social impact. As a recent
example, an M3.4 induced earthquake in Basel, Switzerland,
rattled the local population and suspended the enhanced
geothermal system (EGS) project at that location [Kraft
et al., 2009; Goertz-Allmann et al., 2011].
 Global warming and clean energy requirements are
into deep crustal formations. Fracturing, which results in
seismicity, is an essential process in several industrial applica-
[e.g., Baisch et al., 2002] and fracking shale gas [Kerr, 2012].
In other applications, such as enhanced oil recovery [e.g.,
Grasso, 1992], disposal of fluid waste, and geological storage
of CO2, fracturing, although not required, is unavoidable. The
pressurized water forced into the formation could potentially
reactivate existing deep faults, triggering large earthquakes
[Lei et al., 2008; Giardini, 2009]. At the same time, injec-
tion-induced seismicity is useful for analyzing systematic
structures within hypocenter distributions [e.g., Fehler et al.,
1998; Nicholson and Wesson, 1992] and detecting crustal hy-
drological parameters [e.g., Shapiro et al., 1997, 1999;
Shapiro and Dinske, 2009].
 Some basic information, e.g., the extent to which seis-
micity is related to zones of significant fluid flow, the role of
1State Key Laboratory of Earthquake Dynamics, Institute of Geology,
China Earthquake Administration, Beijing, China.
2Geological Survey of Japan, AIST, Tsukuba, Japan.
3Zigong Earthquake Administration, Zigong, China.
Corresponding author: X. Lei, Geological Survey of Japan, AIST,
Higashi 1-1-1, Center 7, Tsukuba, Ibaraki, 305-8567, Japan. (xinglin-
©2013. American Geophysical Union. All Rights Reserved.
JOURNAL OF GEOPHYSICAL RESEARCH: SOLID EARTH, VOL. 118, 4296–4311, doi:10.1002/jgrb.50310, 2013
fracture distribution and mechanical coupling, the role of
aseismic rupture in enhancing permeability and triggering
tude of potential earthquakes, how long the injection-induced
effect remains after shutdown, how to predict the size of the
largest event that may be triggered, and whether events can
be prevented and/or mitigated, remains poorly understood. It
is expected that the maximum magnitude of induced earth-
quakes should be controllable to a level that is acceptable to
the local society, particularly in urban areas.
 Stable continental regions (SCRs), which have a very
low strain rate but potentially high stress, may be particularly
prone to triggering by fluid injection [Nicholson and
Wesson, 1992; Seeber et al., 2004; Lei et al., 2008]. At the
from injection-induced seismicity in SCRs are helpful for
First, the background seismicity is very low in SCRs, and thus
it is easy to correlate seismicity with injections. Second,
increasing pore pressure is the major force driving seismicity.
Third, the fluid diffusion pattern can be estimated to a certain
degree from history matching of injection data.
 The SichuanBasin,withthe exceptionof itsboundaries,
is relatively stable. However, in the southwestern of the basin,
there are a number of seismic clusters, which are important
because (1) they mirror fluid injections, (2) their source fault
can be compared to preexisting structure, thereby allowing
the role of fluid and structure in seismogenesis to be investi-
gated, and (3) they demonstrate how damaging small to mod-
erate earthquakes can be. Water injections through deep wells,
either for disposal of unwanted water or for dissolving mine
salt, started in the 1970s in gas/salt fields in the southwestern
region of the basin. Fractures of various sizes in tight sand-
stone and limestone/dolomite are important repositories of
natural gas in this region. The timing and locations of these
seismic clusters are strongly correlated with fluid injections,
indicating that these clusters were injection induced [Zhang
et al., 1993; Lei et al., 2008; Zhang et al., 2012]. In some
gas fields, injection has continued over 20 years. The
Rongchang gas field, located on the southeast edge of the
Sichuan Basin, China, is such a case. Since July 1988,
unwanted water has been intermittently injected to a depth of
approximately 3 km, and this injection has coincided with a
significant earthquake sequence, including two ML≥ 5 events,
end of 2006 (see Lei et al.  for details). The sequence is
probably the largest yet observed as a result of fluid injection.
The largest event (ML5.2) was a damaging earthquake and
caused an economic loss of approximately 63 million
Chinese yuan. Due to the limited number of available seismic
stations, a detailed understanding of the source process of the
induced earthquakes could not be obtained.
of injection-induced seismicity in a gas reservoir in Zigong
district to the west of the Rongchang gas field, focusing on
the spatiotemporal distribution of earthquakes and the connec-
local seismic network in the Zigong district, in addition to a
local temporal network of five stations [Long et al., 2010;
Zhang et al., 2012], allow us to obtain a detailed picture of
the spatiotemporal distribution of the induced earthquakes
and explore the connection between seismicity and formation
structures. Studies on such typical case of injection-induced
seismicity are of general significance due to the highly topical
problems of pressure-induced seismicity associated with
subsurface engineering issues worldwide in hydrocarbon
production, geothermal energy extraction, geological storage
of greenhouse gases, and production of shale gas by
hydrofracture and associated reinjection of produced water.
 The remainder of the present paper is organized as
follows. We first present a brief description of the geological
outline of the injection history and seismic monitoring in
section 3. Section 4 provides statistical features of the induced
seismicity and evidence that links injection, seismicity, and
and implications of seismic data is presented in section 5.
Finally, section 6 presents the general conclusions of the
 The Sichuan Basin in southwestern China (Figure 1) is
located within the northwest portion of the stable South
China block. To the west, the basin meets the active Tibetan
Plateau, which is characterized by complex Cenozoic struc-
tures marked by intense deformation and high levels of
seismic activity, including frequent large earthquakes [Shen
et al., 2005; Gan et al., 2007]. The Sichuan Basin has been
investigated extensively with respect to various aspects of
tectonics and petrogeology [cf. Ryder et al., 1991; e.g.,
Molnar and Tapponnier, 1975; Wang and Luo, 1989;
Korsch et al., 1991; Burchfiel et al., 1995; Wei et al., 2008;
Hubbard and Shaw, 2009; Xu et al., 2012]. The tectonic his-
tory of this region is a protracted, two-stage history: an exten-
sional early marine period and a later compressive terrestrial
period. The early marine environment originated on the
Neoproterozoic western passive margin of the South China
block, which collided with both the North China block and
Tibetan Plateau duringthe Upper TriassicIndosinian orogeny,
resulting in the closure of the Paleotethys. In map view, the
basin is characterized by three major structures: (1) a north-
western depression on the northwest side, (2) a northeast
trending central uplift, and (3) a southeastern fold belt on the
southeast side. In section view, the basin is characterized by
sediment layers overlying the Precambrian crystalline base-
ment. For oil and natural gas detection, shallow structures at
depths of up to approximately 6 km are well understood in
the gas/oil field based on geophysical prospecting and boring.
The upper layer is lightly deformed with flat folds, blind (few
exposed) reverse faults, and widely distributed detachments at
ment is characterized by strong metamorphic rock. The top
boundary of the basement is deeper in the northwest side and
shallower in southeast side, showing a gradual thinning from
approximately 13 to 7 km. The top boundary of the basement
is thought to be a detachment. The topmost layer of the base-
ment exhibits a high-velocity layer [Wang et al., 2012].
Some basement structures are of very large scale and affect
overlying structures. The most important of these structures
Geological Settings, Seismicity, and
LEI ET AL.: INJECTION-INDUCED SEISMICITY
is referred to as the Huayinshan basement fault. Although
there is a lack of strong evidence, basement faults are believed
to be active [e.g., Burchfiel et al., 1995; Hubbard and Shaw,
2009] based on the very low strain rate [Liu and Yang, 2005;
Gan et al., 2007] and lack of seismicity.
 The Sichuan Basin is a major petroleum producer [e.g.,
Wang and Luo, 1989] withan annual production of over 120 ×
than 20 commercial oil and gas fields have been discovered in
the Sichuan Basin. The southwestern Sichuan Basin is a major
area of gas reservoirs that has a production history of over than
30 years (Figure 2). Nonassociated gas reservoirs consist of
shallow marine limestone and dolomite of Carboniferous,
Permian, and lower and middle Triassic age with fracture and
vuggy porosity. Lower and middle Triassic evaporites, red
shale, and mudstone are excellent seals. Natural gas is trapped
in tightly compressed, faulted anticlines. In addition to gas
reservoirs, the southwestern Sichuan Basin is also known for
the production of mine salt by pumping water from salt forma-
tions. There are old wells that produce both gas and brine. As
an example, the Shenghai well in Zigong City has reached a
depth of more than 1 km in 1835 and is the first well deeper
than 1 km. This well is currently operational and produces
brine water and natural gas simultaneously.
 GPS data show that at present, the Sichuan Basin has a
very small strain rate [Gan et al., 2007]. Historically, the
Figure 1. Map view showing location and geographical and seismological features of the southwestern
Sichuan Basin and surrounding areas. A through K mark major seismic clusters (every cluster contains
one or more moderate earthquakes of approximately M4–5), which are located in gas reservoirs and/or salt
mines and are thought to be induced/triggered by fluid injection. The lower plot shows a cross section of
simplified geology, which is drawn by integrating information from several papers [Burchfiel et al.,
1995; Wei et al., 2008; Xu et al., 2012].
LEI ET AL.: INJECTION-INDUCED SEISMICITY
Sichuan Basin has exhibited low levels of naturally occurring
seismicity. However, as showninFigure1, there area number
of dense earthquake clusters. Major seismic activity in Zigong
was concentrated in several clusters and is thought to be
associated with either the production of salt water, natural
gas, or water injection. The timing and location of recent
seismic activity are strongly correlated with fluid injections
and thus suggest that these clusters were injection induced
[Zhang et al., 1993; Lei et al., 2008; Ruan et al., 2008; Long
et al., 2010; Zhang et al., 2012].
field, which is associated with the Niufudu-Putaisi anticline, is
located to the east of Zigong City, in the southwest part of the
central uplift structure of the Sichuan Basin (Figure 2).
Huangjiachang field is a small (<100 × 108m3) gas field.
a large (proven reserves of >300 × 108m3) gas field (Weiyuan
field) to the northwest and several small gas fields in Zigong
City and surrounding regions.
 In the region, a number of small-scale faults have been
mapped (Figure 2), although there is no geological evidence
 The Jia#33 well is a production well within the
Huangjiachang gas reservoir that is associated with the
Niufoudu-Putaisi anticline structure, which is a nose-like
anticline and has an axial direction of N80°E and deep flunks
Injection Operations and Seismicity
at the north, south, and west sides. Production at the Jia#33
well ended before 2007. Starting in 2007, Jia#33 was used
for the injection of unwanted water that was collected through
pipelines from nearby production wells. Water was injected
into the Maokou limestone formation (P1m) of late Permian
at a depth of approximately 2500 m. The formation contains
well-developed fractures and joints. Note that the underlying
lower Permian formation consists of dense limestone depos-
ited on an open marine platform, which was partially eroded
during the Dongwu orogeny at the end of the Early Permian.
formity with developed paleokarst which is covered by shale
layers [Hu et al., 2012]. The injection site is located in a tran-
sition zone of karst steep slope to karst platform. The well was
originally drilled to a depth of 2550 m from the surface, with
pipe casing to a depth of 2450 m and approximately 100 m
of open hole.
2.2.1.Nonpumping Disposal, 2007–2008
 The daily injection rate was lower than 300 m3up
until April 2008, and then increased to approximately
500 m3toward the end of 2008. During this period, fluid
was placed into the well under gravity flow. At the end of
2008, the total injected volume was approximately 150,000
m3. During the 2 year period, only a small number of earth-
quakes of magnitude <2.5 were recorded. Note that the
closest seismic station of the Zigong local network is located
16 km south of the injection well, and the detectability was
less than M0.5 at the injection site.
Figure 2. Close-up view of the center area in Figure 1 showing the distribution of faults, routinely
determined earthquake hypocenters, gas reservoirs, seismic stations (solid triangles: permanent, open
triangles: temporal), and other information. The present study focuses on cluster C, referred to as the
2009–2010 Zigong earthquake sequence, in Huangjiachang field associated with the Niufudu-Putaisi
anticline (G2) and the injection well Jia#33 in the center of the map. The large seismic cluster on the
right-hand side is associated with the injections in Rongchang gas field (see Lei et al.  for details).
The distribution ofgas fieldsis modified from Ma et al..TheGPS movement direction is also shown
[Gan et al., 2007].
LEI ET AL.: INJECTION-INDUCED SEISMICITY
 Since 10 January 2009, pumping under high pressure
was required for injection. Continuous injection was
performed almost every day. Each injection was followed
by a shut-in period, in which the wellhead was closed for a
number of hours, depending on the amount of unwanted wa-
ter that had been collected. Injection was performed continu-
ously for more than 24 h only a few times. Pumping injection
has led to significant seismicity [Long et al., 2010; Zhang
et al., 2012]. During the period of injection from 2009 to
2010, more than 5000 ML1+ earthquakes were recorded.
Continuous Injection Under High Pumping
These earthquakes are referred to collectively as the 2009–
2010 Zigong sequence.
 The earthquake occurrence rate increased rapidly
when the wellhead pressure exceeded 2 MPa at the end of
January 2009. A ML4.4 earthquake that occurred on 16
February 2009 marks the largest event of the earthquake
sequence. Another major ML4.2 earthquake occurred on 22
May. 2009. These large events were felt over a wide area,
and minor damage was reported at sites near the epicenter.
 The present study focuses on the period from 2009
 There is a local seismic network consisting of five
stations, covering the Zigong district. This network was
constructed in the 1970s motivated by the fact that felt earth-
quakes occurred at an abnormally high frequency in this area,
which was thoughttoberelativelystable. The Jia#33injection
wellislocated on the easternedge of thenetworkcoverage,16
km to the south of the nearest station. A temporary network of
five stations centered at the injection well with an interval dis-
tance of less than 10 km was installed in July 2009.
 Figure 3 shows examples of the three-component
velocity seismograms at selected stations for a very shallow
(~1 km) earthquake and a M2.5 earthquake having a focal
depth of approximately 3 km, which falls in the principal
depth range of the induced earthquakes. Most events exhibit
both up and down P first motions, clear S phases, and strong
surface waves, indicating a shear fracturing mechanism and a
very shallow focal depth.
 In order to determine the lower cutoff magnitude of the
earthquake catalogue for completeness, we examined the
magnitude-frequency distribution and scatterplots of magni-
tude against sequential number of events for two periods
before and after the installation of the temporal stations. As
shown in Figure 4, the lower cutoff magnitude without the
temporal stations is 1.0, which is 0.5 higher than with the
 Figure 5 shows the mean event rate, mean injection rate,
and event number as a function of the cumulative volume of
injected water. In total, the mean event rate and injection rate,
calculated for a time bin of 5 days, are fairly well correlated.
The injection history can be roughly divided into four major pe-
riods according to the changing pattern of the injection rate and
 Phase I. The daily injection rate exceeds 300 m3, and
the wellhead pressure rapidly jumped to approximately 2.8
MPa and then slowly increased to approximately 3 MPa.
 Phase II. The daily injection rate decreases gradually
to 150 m3, and the wellhead pressure increases to
approximately 4 MPa.
 Phase III. The daily injection rate increases gradually
to 300 m3, and the wellhead pressure increases to
approximately 5.5 MPa.
 Phase IV. The daily injection rate decreases gradually
to100m3,andthewellhead pressureremained intherangeof
5.5 to 6.2 MPa.
Temporal Evolution of the Induced Seismicity
Typical Phases of Injection and Seismicity
Figure 3. Examplesofthethree-componentvelocityrecords
of two typical events.Clear S phases and strong surface waves
indicate shear fracturing mechanism and show focal depth.
LEI ET AL.: INJECTION-INDUCED SEISMICITY
 A plot of the cumulative number of earthquakes with
respect to the cumulative injected volume demonstrates that
the productivity of induced earthquake in periods of increas-
ing injection rate (I and III) is clearly higher than that in pe-
riods of decreasing injection rate (II and IV) (Figure 6).
 In order to quantitatively investigate the occurrence of
the induced earthquakes, we examined temporal variations in
the event rate (n), cumulative event number (N), and seismic
b-value in the magnitude-frequency relation. As an integrated
set,theseparameterswerefound tobe useful as anindicator of
the critical point behavior of rock fracture in stressed rock
samples [Lei and Satoh, 2007] and are helpful for finding dif-
ferent phases ininjection-induced seismicity [Lei et al., 2008].
 Figure 6 shows the injection history and the major
statistical parameters, which were sequentially calculated for
consecutive groups of 100/200 events with a running step of
25/50 events. All parameters show clear variation patterns,
which, not surprisingly, coincide with the aforementioned
phases identified from injection data. Phase I shows a lower
b-value (~0.8) than in later phases. Phase II, which corre-
sponds to a period of decreasing injection rate and approxi-
mately constant injection pressure, shows lower event rates,
an increasing b-value (from 0.8 to 1.0). Phase III corresponds
creasing injection pressure, exhibiting an increasing event rate
and b-value (from 1.1 to 1.4). Phase IV shows a lower b-value
(~0.9) and a decreasing event rate and energy release rate.
 Note that the aforementioned long-term patterns fluc-
tuate with short-term variations. Interestingly, short-term
variations are also clearly correlated with short-term fluctua-
tions of injection pressure.
 Seismictriggeringplaysaveryimportantrole inearth-
quake occurrence. In the case of injection-induced earth-
quakes, fluid-inducing is always accompanied by stress
triggering. The epidemic-type aftershock sequence (ETAS)
model [Ogata, 1992], which incorporates the Omori law by
assuming that each earthquake has a magnitude-dependent
ability to trigger its own Omori-law-type aftershocks, is an
useful tool for finding changes in seismic patterns [Ogata,
1992, 2001; Helmstetter and Sornette, 2003] and extracting
a fluid signal from seismicity data [Hainzl and Ogata,
2005; Lei et al., 2008]. In the ETAS model, the total occur-
rence rate is described as the sum of the rate triggered by
all preceding earthquakes and a forcing rate λ0(t) that repre-
sents the background activity:
Epidemic-Type Aftershock Sequence Modeling
λ t ð Þ ¼ λ0t ð Þ þ ν t ð Þ;ν t ð Þ ¼
ðÞt ? tiþ c
whereMcisthe low cutoff magnitudeof the catalogueandα is
a constant that specifies the degree of magnitude dependence.
A constant forcing rate is normally assumed for seismicity at
regional scales. For the case of injection-induced seismicity,
Figure 4. Plot of magnitude-frequency distribution and the
seismic b-values estimated using the maximum likelihood
method (bml) for the 2009–2010 Zigong earthquake
sequence. The least squares fit (bls) is also made as a compar-
ison. Scatterplots of magnitude against sequential number of
events indicate thelower cutoff magnitudes usedinthisstudy
(1.0 and 0.5 for earlier and later periods, respectively) are
Figure 5. (left) Comparison of mean event rate and injection rate calculated for a time bin of 5 days and
(right) cumulative number of earthquakes plotted with respect to cumulative injected volume.
LEI ET AL.: INJECTION-INDUCED SEISMICITY
the background activity consists of tectonic background and
injection-induced activities. It has been found that the forcing
ratedepends oninjection factors andthusisnonstationary [Lei
etal., 2008].Forsucha case,a time-varying orpiecewise con-
stant forcing rate should be used to avoid underestimation of
the forcing rate and α value [Lombardi et al., 2010; Marsan
et al., 2013]. Model parameters were estimated by minimizing
the Akaike information criterion (AIC). According to our
experience, the ETAS model is very sensitive to c. In order
to ensure numerical stability, we perform a grid search on pa-
rameter c, while all other parameters are estimated using
Powell’s optimization method. Figure 7 shows the perfor-
mance of the proposed method. The AIC and other model
parameters are plotted with respect to c, which allows the ef-
fects of c on other parameters to be investigated. There is a
global minimum AIC.
 In order to estimate the time-dependent forcing rate
and other parameters, we applied a sophisticated algorithm,
inspired from Zhuang et al.  and Marsan et al.
. The method consists of
 1. initially assuming a constant forcing rate λ0
Figure6. (a) M-T and N-T plots of earthquakes occurring during theperiod between2000 and 2011.(b–f)
Close-up view of seismicity and injection data for the period between 2008 and 2011. The event rate (n),
energy release rate (e), cumulative event number (N), and seismic b-value in the magnitude-frequency re-
lation (b) were estimated for consecutive groups of 100 events with a running step of 25 events. Thick
dashed lines show smoothed results for consecutive groups of 200 events with a running step of 50 events.
LEI ET AL.: INJECTION-INDUCED SEISMICITY
 2. estimating the minimum AIC ETAS parameters
θ(K, α, c, p) knowing λ0(t);
 3. updating the estimate of the forcing rate based on
 4. repeating steps 2 and 3 until convergence of
 In the first step, we use the Gamma distribution of
waiting times between consecutive earthquakes:
f τ ð Þ ¼ Cτγ?1e?τ=β
where τ is the normalized interevent time that is obtained by
multiplying the interevent time △t with the earthquake rate.
Based on the assumption that the seismicity consists of a
Poissonian background activity and triggered aftershocks fol-
lowing the Omori law, Molchan  showed that the value
1/β is the maximum likelihood estimation of the fraction of
main shocks among all seismic events, i.e., the forcing rate.
ability ωithat the ith earthquake is a background earthquake, for
all i. This probability is defined as [Zhuang et al., 2002]
λ0ti ð Þ
λ0ti ð Þ þ ν ti ð Þ
 The forcing rate is then obtained by smoothing these
probabilities over time. It is convenient to compute the forc-
ing rate at tifor all i with a smoothing window of 2 × ne+1
earthquakes centered on ti:
λ0 ti ð Þ ¼
jp ¼ i ? ne; ja ¼ i þ ne
by minimizing following Akaike information criterion (AIC):
??? tjp?1þ tjp
ð Þ ¼ AIC ETAS
ðÞ þ 2N= 2neþ 1
where N is the number of earthquakes in the data set.
 The results of ETAS modeling for the 2009–2010
Zigong earthquake sequence are shown in Figure 8. The mini-
mum AIC is obtained at ne=6, corresponding to a smoothing
windows of 13 events. The obtained forcing rate and other
ETAS parameters are shown in Figure 9b. Figure 9a shows
the piecewise constant forcing rate λ0(t) estimated from fitting
of 50 events. In order to find possible changes on the Omori-
injection [Lei et al., 2008], we also estimated forcing rate λ0(t)
and other ETAS parameters for different periods (shown in
Figure 9c). In agreement with the multiphase features identified
also reveals some significant changes. In total, the obtained
ETASparametersexhibit amajor forcedcomponent,indicating
that more than 50% of the earthquakes were forced externally
by injection. The remainder represents Omori-type self-trig-
gered activity. During the first two phases, the total fraction of
forced activity is ~50%, which indicates injection-inducing
and natural seismic triggering are both significant. During
phasesIIIand IV, the forced activity takes more than 80%, thus
parameter α is 1.48 for phases I and II, and 1.51 and 1.01 for
phases III and IV, respectively. The Omori-type aftershock
decay parameter p is close to unit value, i.e., the global mean
value, for the first phase. For later phases, p shows increased
values which are 1.26 and 1.38 for phases III and IV, respec-
tively. The resulting small value of α indicates relatively weak
dependence on earthquake magnitude, so that the earthquake-
induced stresses should have relatively weak impact as
in the case of α=log(b), large and small earthquakes have the
same contribution in triggering aftershocks. These results
indicate that both pore-pressure diffusion and slip-induced
stress transfer have a role in earthquake occurrence.
 The routinely determined hypocenters of this sequence
are scattered within 10 km of the injection well. In order
to obtain a detailed understanding, we relocated earthquakes
having a magnitude greater than or equal to 0.75 in the
sequence with a velocity model obtained using the receive
function method of a nearby station reported by Wang
et al. .
 We applied the double-difference method [Waldhauser
and Ellsworth, 2000] to the relocation of earthquake hypocen-
ters. Catalogue phase data from both the local network and the
temporal stations were used. Limited by the small number and
poor positions of available stations, the hypocenter and the
focal depthof earthquakes that occurred beforethe installation
of the temporal stations could not be well determined. For
these earlier events, we used their nearest, well-located
neighbor event to approximate their hypocenters. The nearest
event isdefinedasthe event that has the shortest distance from
the target event in the travel time space recorded by the
Detailed View of Hypocenter Distribution
Relocation of Earthquake Hypocenters
Figure 7. Example of grid searching for the optimal param-
eter c in the ETAS model. Numbers in  indicate the mini-
mum and maximum values of the corresponding vertical
axis. Note that otherparameters suchas pand λ0arevery sen-
sitive to c for c values greater than 0.1, which may lead to nu-
merical divergence. A grid search algorithm is normally used
to a new data set to find a better guess and to constrain the
range of c for optimization algorithms.
LEI ET AL.: INJECTION-INDUCED SEISMICITY
Figure 8. Results of ETAS models of time-varying forcing rate λ0(τ) for the M≥1.0 earthquakes of the
Zigong 2009–2010 sequence. (a) Estimated forcing rate λ0(t) for three values of the smoothing parameter
neas labeled. (b) ETAS parameter K, α, p, and AIC, together with the total fraction of forcing seismicity
(background). The minimum AIC is obtained for ne=6.
Figure 9. ETAS results of the Zigong earthquake sequence. (a) The piecewise constant forcing rate λ0(t)
estimated from fitting only λ0for consecutive groups of 200 events with a running step of 50 events. (b)
Estimated forcing rate λ0(t) for the whole period (Figure 8, ne=6). (c) Estimated forcing rate λ0(t) for three
LEI ET AL.: INJECTION-INDUCED SEISMICITY
 In map view, the relocated hypocenters concentrated in a
wide centered approximately at the injection well. The
hypocenter distribution is consistent with the deep flunk at the
west side of the Niufudu-Putaisi nose-like anticline (Figures 10
and 11). A hypocenter density map (Figure 11) demonstrates
that the hypocenters are likely controlled by a set of preexisting
conjugate fractures. Such fractures are consistent with the anti-
cline structure and the regional stress field.
 In section views, more than 90% of hypocenters fall in
the depth range of 2.5 to 4 km, which corresponds to the
role in arresting fractures in the limestone. At the front of
hypocenters, seismic activity was probably bounded by dipping
at the northwest front, there are a number of very shallow and
Three-Dimensional Hypocenter Distribution and
relatively deep events, probably indicating a dip fault, which is
consistent with the anticline structure (Figure 12).
 We recall following two facts: (1) the top surface of the
paleokarst which is covered by shale layers [Hu et al., 2012;
Qin et al., 2012] and (2) the injection site is located in a transi-
tion zone of karst steep slope to karst platform. Thus, the reser-
voir might be very permeable. Interestingly, there are very
(Figure 12). This is consistent with the poroelastic model of
Segall  for earthquakes triggered by fluid extraction.
Earthquakes migrate outward from the reservoir boundaries
rather than the injection point.
Diffusion and Fault Reactivation
the pore pressure as a rough approximation of the injection.
Under the approximation of a point source of pore pressure
Hypocenter Migration: Evidence for Pore-Pressure
Huangjiachang gas field associated with the Niufudu-Putaisi anticline structure. AA′ and aa′ indicate two
perpendicular cross sections that are plotted in Figure 12.
Map view of the relocated hypocenter distribution. The enclosed area indicates the
conjugate fractures estimated from hypocenter data. Such fracture networks are consistent with and the
regional stress field, which demonstrates NW-SE to NWW-SEE compression.
Map view of the density distribution of earthquakes. The dashed lines illustrate a set of
LEI ET AL.: INJECTION-INDUCED SEISMICITY
perturbation in an infinite, hydraulically homogeneous, and
isotropic fluid-saturated medium, the triggering front r(t) has
the following form [Shapiro et al., 1997]:
r t ð Þ ¼
wheret isthe timefrom the injection start and D is the hydrau-
lic diffusivity. If the injection stops at time t0, then the earth-
quakes will gradually cease to occur. For times greater than
t0, a surface that describes the propagation of a maximal pore
pressure perturbation in the space can be defined. This surface
(also a sphere in homogeneous isotropic rocks) separates the
spatial domain that remains seismically active from the spatial
domain (around the injection point) that is already seismically
quiet. This surface was first described by Parotidis and
Shapiro  and was referred to as the back front of in-
duced seismicity [Shapiro and Dinske, 2009]:
rbf t ð Þ ¼
ln t= t ? t0
where d is 1, 2, or 3 and represents the dimension of the space
in which the pressure diffusion occurs. The back front is
another kinematic signature of pressure-diffusion-induced
tigraphy. The stratigraphy is drawn while referring to a nearby site at which the stratigraphy and deep struc-
tures are well constrained with boring data [cf. Ryder et al., 1991; from Korsch et al., 1991]. The depth
distribution of hypocenters indicates that more than 90% of earthquakes are located in a depth range be-
tween 2.5 and 4 km. Note that the injection interval is 2.45 to 2.55 km beneath the surface, which has an
elevation of 340 m.
Hypocenter depth distribution superimposed on simplified geological cross sections and stra-
Figure 13. EstimatesoftheoveralltriggeringfrontsfortheseismicityofwaterinjectionoftheJia#33well.The
wellhead pressure is shown for comparison. Hypocenter depths in phase I are poorly determined and so are plot-
roughly estimated from the shortest distance between every hypocenter to the reservoir boundaries.
LEI ET AL.: INJECTION-INDUCED SEISMICITY
 At first, we assume a single pore-pressure source
located in the injection well at the mean injection depth.
Figure 13a shows the corresponding r-t plot. Except for phase
I, in which the focal depth is poorly determined, most hypo-
centerslie withinthetriggeringfrontofD=0.1 m2/s. Theseis-
micity after shut down is too little (we recall that the threshold
magnitude of completeness of our data set is 0.5) for drawing
the back-front. As aforementioned the reservoir might be very
permeable and there are fewer earthquakes located within the
depleted reservoir volume. As shown in Figure 13b, by using
the shortest distance from each hypocenter to the reservoir
boundaries, the triggering front could be better tracked.
 There are several shut-in periods of relative longer
duration leading to significant pressure drops and decrease in
earthquake rate. Each shut-in was followed by a fast pressure
buildup when injection was restarted. In addition to the overall
sure changes. Figure 14 shows r-t plot after stacking over all
episodes. A hydraulic diffusivity on the order of ~5 m2/s is
estimated for such fast migration behaviors. Considering the
aforementioned connection between seismicity and formation
structures, the fast responses likely reflect pore-pressure diffu-
sion within the reservoir and along the surface of preexisting
faults, especially along the permeable intersections of faults.
Since the top surface of the Maokou formation at the injection
paleokarst platform [Sang et al., 2012], the well-developed fis-
sure-cavity system could play a role in the fast responses.
Furthermore, it seems intuitively sound that hydraulic diffusiv-
ity is larger once the fracturing and permeability enhancing
have taken place.
 In total, as seen from Figures 13 and 14, the triggering
fronts are poorly determined. The model for hydraulic
diffusion assumes (1) the medium is uniform (i.e., flow is
Fickian) and (2) no time-variation and longer-range pore pres-
sure transfer due to poroelasticity and deformation caused by
earthquake ruptures. The Zigong earthquake sequence covers
a spatial extent of 6 km and time span of 1.5 years. Both space
variation and time variation of hydraulic diffusivity are inter-
esting issues worth further studying. Unfortunately, our data
set, which was obtained from surface observations, is insuffi-
cient to make convincing analysis on these issues. We thus
calculate the evolution of mean triggering distance with
respect to time. It is a way relatively straightforward to test
the hypothesis of anomalous diffusion of seismicity [Huc
and Main, 2003]. We define all earthquakes of M≥1.0 as
“source” events, all earthquakes following a source event are
“triggered” events, and then calculate the histograms of
distances between each source-triggered pair of events. The
mean triggering distance is given by
0p r ð Þrdr
0p r ð Þdr
where p(r) is the probability density of triggering at a dis-
tance r±dr/2. For diffusion processes of temporally corre-
lated seismic activity, the mean distance takes the form of a
power law growth<r>~ tH. Many seismic systems show
anomalous diffusion with a growth exponent H<0.5 [e.g.,
Marsan and Lengline, 2008; Huc and Main, 2003; Marsan
et al., 2000], while a homogeneous normal fluid diffusion
process exhibits H=0.5.
 Wehave clearly proved that seismicity ismostly back-
ground (forced), it is thus important to examine fluid
diffusion embedded not only in forced seismicity but also
in triggered seismicity. For any earthquake taken as a
main shock, most of the subsequent earthquakes (typically
50–80% according to the ETAS analysis) are forced and thus
uncorrelated through cascade triggering chains. In order to
explore the (bare) contribution of directly triggered after
sequence, we calculate a mean triggering distance by
weighting each hypocenter distance rijwith the probabilities
Figure 14. Triggering fronts fitted to stacked hypocenters
following episodes of shut-ins. The obtained fast migration
patterns likely reflect fast pore pressure diffusion along
preexisting faults of enhanced permeability within or
connected with the depleted reservoir.
Figure 15. Mean hypocenter distance<r>between main
shocks and aftershocks versus time following the main shock
for bare (directly triggered) and background (forced)
aftershock sequences. The best power laws<r>~ tH
give H=0.25 (bare), and H=0.15 (t<10 min), and
H=0.06 (t>10 min) (background).
LEI ET AL.: INJECTION-INDUCED SEISMICITY
Ωijthat the ith event triggered the jth event. At the same time,
the contribution of background events can be estimated by
weighting rijwith 1?Ωij. Ωijcan be calculated from the
ETAS models. A similar approach has been proposed in pre-
vious studies [e.g., Helmstetter et al., 2003; Marsan and
Lengline, 2008]. Figure 15 shows evolution of the mean dis-
tance<r>between main shocks and aftershocks versus
time following themain shock forbare andbackground after-
shock sequences. It is not surprising to see power laws
growth of<r>in both of the bare and background after-
shock sequences. The best power laws<r>~ tH give
H=0.25 (t<7 min) (bare), and H=0.2 (t<15 min) and
H=0.06 (t>15 min) (background). Our results indicate that
fluid flow has a role in both forced background seismicity
and directly triggered seismicity.
 In general, there are a number of factors that may have
a role in the injection-induced seismicity. These factors can
be divided into four categories: (1) factors associated with
the in situ conditions for the formation into which fluids are
pumped; (2) tectonic settings such as loading rate and stress
regime; and (3) factors concerning injection, including injec-
tion rate, injection pressure, and the total amount of fluid
injected thus far; and (4) longer-range pore pressure transfer
due to geomechanical effects (e.g., poroelasticity or transfer
of effective stresses from earthquakes). It is important to be
able to assess the relative importance of these factors.
 The Zigong 2009–2010 sequence is one of the many
examples of injection-induced earthquakes in the southwest-
ern Sichuan Basin. In many instances, including this one, the
time-space correlation between seismicity and injection sug-
gests acausalconnection. Sincethepore-pressurerisingfrom
the injection is small compared to the tectonic stress, struc-
ture correlation is pertinent to seismogenesis in general.
Results based on ETAS model of time-varying background
rate show that the forced (by injection) seismicity takes about
50% in total at the earlier periods of the injection. Thus, the
remained seismically triggered activity is also significant, in-
dicating that the initial effective mean stress is low and most
optimally oriented fractures within and surrounding the res-
ervoir are critically stressed. A near-critical stress field is also
consistent with the ease with which seismicity can be in-
duced at all, and with the observed scale-invariance associ-
ated with the plethora of power laws that emerge in natural
as well as induced seismicity statistics [Main, 1996].
 In agreement with a previous study on injection-
induced earthquakes, the results for the present case indicate
that the ETAS model together with other statistic methods is
a promising approach in terms of identifying fluid signals in
seismicity patterns. The sequence demonstrates the following
features: (1) normal or relatively small b-value, depending
on local geomechanic conditions, (2) the sequence is more
swarm-like and is characterized by a smaller value α in the
ETAS model, (3) increasing fraction of forced seismicity with
increasing injection time, especially for the case of long-term
injection, and (4) the progressive increase in the maximum
magnitude in the very early stage. Similar features can be
found in natural seismicity driven by upward fluid flow from
a deep level as observed in geothermal active regions [Lei
et al., 2011].
 It is work addressing that the temporal change of b-
value demonstrates that the Zigong sequence is not “typical”
of injection-induced sequences. In many cases, a progressive
increase of the maximum magnitude was observed (such as
the case of the Rongchang gas field which is close to
Zigong [Lei et al., 2008]), the large magnitudes are often ob-
served at the middle to end of the sequence (and even after
shut down like the case of Basel [Goertz-Allmann et al.,
2011]). Lager injection volume results in a wider area of in-
creasing pore pressure, and thus, a rupture can propagate
over longer distances. This causes b-value to decay with time
as that observed in the Rongchang gas field [Lei et al., 2008].
Indeed, we observed decreasing b-value from phases III to
IV. However, increasing b-value from phases I toII is a ques-
tion under debate. Since the Zigong sequence is governed by
reactivation of preexisting fractures which are thought to be
critically stressed. Lower b-value at the beginning probably
indicates that rupture could be triggered on larger fractures
by comparatively smaller increase of pore pressure. It is an
interesting issue worth further study.
with the hydraulic diffusion. The triggering fronts shown in
Figures 13 and 14 are not clear and the mean distance growth
exponent is very small as compared to that expected for
homogeneous normal diffusion process. We recall that the
model used for hydraulic diffusion assumes the medium is
uniform. Longer-range pore pressure transfer due to
geomechanical effects (e.g., poroelasticity or transfer of ef-
fective stresses between earthquakes involved in the ETAS
model) is also ignored. Geological heterogeneity alone can
result in anomalous diffusion [Berkowitz and Scher, 1998]
and long-range poroelasticity can result in triggering outside
the traditional fluid flow front if the crust is previously near-
critically stressed [Maillot et al., 1999]. This effect has even
been seen in fluid flow rate correlations associated with
hydraulically reactive faults in oilfields [Main et al., 2006].
Natural seismicity triggered by stress transfer has also been
shown to have the properties of an anomalous diffusion
process [Huc and Main, 2003], and theETAS modeling done
here shows such triggering is in fact an important process
(especially during phases I and II of the injection) in
explaining the earthquake time series. Further works should
focus on in situ stress data and laboratory measurement of
rock friction properties to understand the criticality of stress
within and surrounding the reservoir.
Debate on b-Value and Hydraulic Diffusion
terchangeably. However, the distinction between the two
terms is worth mentioning. The term induced indicates a
causative activity that accounts for most of the stress change
or energy required to produce the earthquakes, whereas the
term triggered describes a process that accounts for only a
small fraction of the same stress change or energy [McGarr
and Simpson, 1997]. As such, we prefer to use the term trig-
gered if the external force is small compared to the stress
expected to cause seismogenic failure, or compared to the
stress drop of normal earthquakes, which is usually on the or-
der of 1 to 10 MPa. In the present case, the Coulomb failure
stress caused by the injection could be on the order of 1 MPa,
Induced Seismicity or Triggered Seismicity
LEI ET AL.: INJECTION-INDUCED SEISMICITY
which is comparable to the stress drop of earthquakes. Thus,
the 2009–2010 Zigong sequence falls into the induced
category. The study area is located in a relatively stable con-
tinental region, and the small value of strain rate (<1 mm/a)
predicts a reoccurrence time of moderate size on the order of
1000~ 10,000 years. A smaller pore-pressure increase can
fasten the failure by many years even cause a sudden failure.
However, in a depleted gas reservoir, the situation is rather
complex because gas extraction has reduced the pore pres-
sure to an unknown level. The fact that there is no significant
seismicity observed in 2008, during which water was place
into theformationunder gravity, likely indicates therecovery
process of pore pressure was reduced by gas extraction to the
Fractures, and Faults
 Multiple sources of evidence, such as the shear mecha-
nism, the pattern of hypocenter distribution, and small
elevated pore pressure as compared with the least principal
stress in the region, indicate that the induced seismicity
resulted from the reactivation of known or unknown
preexisting faults. In addition, the largest events occurred in
the very early stage (2 weeks delay) of injection under high
portant factor in the case. Injected fluids diffused outward
along preexisting faults, which are originally stressed, play a
role in weakening the faults and lead to reactivation.
 Our results indicate that lowering of the effective nor-
mal stress is the dominant mechanism inducing earthquakes
in the region, in agreement with the 9.1 km deep KTB site
[Baisch and Harjes, 2003]. This mechanism may give rise to
damaging earthquakes if a fault has a dimension on the order
of a few kilometers, which is sufficient for producing M4–5
earthquakes. Water can have a very large effect on rock
strength, especially at elevated temperatures. In addition,
long-range poroelasticity in a critically stressed crust and seis-
mic pumping could be alternative mechanisms causing the
strong channeling along faults invoked. Feedback between
pressure at the injection borehole and fault reactivation into
the dilatant regime has been observed for example in the prin-
cipal component maps of the statistical reservoir model of
Main et al.  and the physical reservoir model of
tion of constant rate had dropped abnormally after some large
earthquakes. Such kind of pressure drops are possible signs of
seismic pumping effect. Thus, the fault valve model, by which
a fault isreactivated by upwardfluidfroma deep level[Sibson
et al., 1975], can be used in our case. The widely developed
horizontals of evaporates act as a barrier. If a blind fault is
reactivated by such a mechanism, the maximum magnitude
of potential earthquakes would be determined by the size
and stress accumulated on the fault, rather than the amount
of water injected.
 Keeping in mind that the largest event of the 2009–
2010 Zigong sequence occurred at the very beginning of in-
jection under a relatively lower pressure, it is reasonable to
assume that the induced earthquake having the greatest mag-
nitude in the studding region is determined primarily by the
size distribution and position of pre-existing faults surround-
ing the injection well. Increasing injection time can increase
the possibility of reactivate larger faults at certain distances
Role of Preexisting Fissure-Cavity System,
from the injection well. No deep events were observed in
the present case. However, 40 km to the east, in the
Rongchang gas field, a number of deep events, including
some M5+ events, which were probably associated with the
basement faults underlying the gas field, were observed
[Lei et al., 2008]. Deep events of M5+ occurred after a long
injection period of 8 years. In the present case, the total injec-
tion period under high pressure was only 20 months. The
ETAS modeling results for the 2009–2010 Zigong sequence
are similar to those for earlier phases of injection-induced
seismicity in Rongchang reservoir. Thus, whether larger
earthquakes could be induced if the injection continued for
a longer time remains unknown. In Rongchang gas field,
there is a well-known large-scale basement fault underlying
the formation. Although no evidence of basement faults
was found beneath the Zigong area, we should note that a
fault of a few kilometers in length is sufficient to produce
M5 class earthquakes.
 There is strong evidence indicating the fast response
of seismicity to pressure changes. Shutdown of the injection
well in August 2010 caused a rapid cessation of seismicity,
and no significant earthquakes have occurred since. This
indicates that the induced seismicity in the region demon-
strates minor delayed activity and is thus easier to control.
 However, since major earthquakes may occur at the
very beginning of injection, it is difficult to make a probabil-
ity-based prediction for future earthquakes due to the lack of
forerunners. In such cases, prospecting for faults in the forma-
is necessarytoclarifythe size and position of any fault of,say,
aging earthquakes of moderate magnitude.
Hazard Assessment of Induced Seismicity
 The 2009–2010 Zigong earthquake sequence is closely
associated with the injection of unwanted water in the lime-
stone formation of Permian. A total of 130,000 m3of water
has been pressed into the formation during the period from
2009 to July 2010. During this period, more than 7000 earth-
quakes, among which 5000 events having magnitudes of
greater than or equal to 0.5 have been recorded by a nearby
local network and five temporary stations. Note that the injec-
tion at this site was started in 2007 but no pumping was
required until the end of 2008. During this early stage, only
a few earthquakes having magnitudes of less than 2.5
 ETAS modeling results indicate that injection induc-
ing and earthquake triggering are both important (50 to 50)
during earlier periods of injection, while later periods are
dominated by forced seismicity (~80%). These results agree
well with the injection-induced seismicity in the nearby
Rongchang gas field [Lei et al., 2008].
 The spatiotemporal distribution and other statistical
results indicate that the induced seismicity is characterized
by four typical phases, which reflect the patterns of the injec-
tion rate and wellhead pressure. The largest ML4.4 events
occurred when the wellhead pressure reached 0.9 MPa at
the very beginning of injection. Various factors, such as the
shear mechanism, the pattern of hypocenter distribution,
LEI ET AL.: INJECTION-INDUCED SEISMICITY
and the fractal dimensions, indicate that the induced seismic-
ity in the region resulted from the reactivation of preexisting
faults. Injected fluids diffuse outward along preexisting
faults, which were originally stressed, weakening the faults
and leading to their reactivation. The intersections of a set
of conjugate fractures are particularly suitable for fluid
flowing. Some relatively large dipped faults likely bound
the outward fluid flow and provide paths for upward leakage
and downward flow.
 In the region, the maximum size of the induced earth-
quake is most probably determined by the size distribution of
the positions of preexisting faults. Increasing the injection
time can increase the possibility of reactivating larger faults
at a certain distance from the injection well.
 There is strong evidence of the fast response of seis-
micity to pressure changes. The shutdown of the injection
well in August 2010 has rapidly decreased seismicity and
no significant earthquakes have occurred since its shutdown.
This also indicates the role of permeable fault surfaces.
 Acknowledgments. We thank the Associate Editor and two anony-
ent study was supported by State Key Laboratory of Earthquake Dynamics.
Ahmad, U. A., and J. A. Smith (1988), Earthquakes, injection wells, and the
Perry nucleation power plant, Cleveland, Ohio, Geology, 16, 739–742.
Ake, J., K. Mahrer, D. O’Connell, and L. Block (2005), Deep-injection and
closely monitored seismicity at Paradox Valley, Colorado, Bull. Seismol.
Soc. Am., 95, 664–683.
Baisch, S., and H.-P. Harjes (2003), A model for fluid-injection-induced
seismicity at the KTB, Germany, Geophys. J. Int., 152, 160–170.
Baisch, S., M. Bohnhoff, L. Ceranna, Y. Tu, and H.-P. Harjes (2002),
Probing the crust to 9 km depth: Fluid-injection experiments and induced
seismicity at the KTB superdeep drilling hole, Germany, Bull. Seismol.
Soc. Am., 92(6), 2369–2380.
in random fracture networks, Phys. Rev. E., 57(5), 858–5,869.
Burchfiel, B. C., Z. Chen, Y. Liu, and L. H. Royden (1995), Tectonics of the
Longmen Shan and adjacent regions, Int. Geol. Rev., 37(8), 661–735.
Dahm, T., F. Kruger, K. Stammler, K. Klinge, R. Kind, K. Wylegalla, and
J.-R. Grasso (2007), The 2004 Mw4.4 Rotenburg, Northern Germany,
earthquake and its possible relationship with gas recovery, Bull. Seismol.
Soc. Am., 97(3), 691–701, doi:10.1785/0120050149.
Cogdell oilfield of west Texas, Bull. Seismol. Soc. Am., 79, 1477–1495.
quakes induced during hydraulic fracturing, Tectonophysics, 289,
Gan, W., P. Zhang, Z.K. Shen, Z. Niu, M. Wang, Y. Wan, D. Zhou, and
J. Cheng (2007), Present-day crustal motion within the Tibetan Plateau in-
ferred from GPS measurements, J. Geophys. Res., 112, B08416,
Giardini, D. (2009), Geothermal quake risks must be faced, Nature, 462,
Goertz-Allmann, B.P., A. Goertz, and S. Wiemer (2011), Stress drop varia-
tions of induced earthquakes at the Basel geothermal site, Geophys. Res.
Lett., 38, L09308, doi:10.1029/2011GL047498.
Grasso, J. (1992), Mechanics of seismic instability induced by the recovery
of hydrocarbons, Pure Appl. Geophys., 139, 507–543.
Hainzl, S., and Y. Ogata (2005), Detecting fluid signals in seismicity data
through statistical earthquake modeling, J. Geophys. Res., 110, B05S07,
Healy, J. H., W. W. Rubey, D. T. Griggs, and C. B. Raleigh (1968), The
Denver earthquakes, Science, 161, 1301–1310.
Helmstetter, A., and D. Sornette (2003), Main shocks are aftershocks of con-
ditional foreshocks: How do foreshock statistical properties emerge from
aftershock laws, J. Geophys. Res., 108(B10), 2046, doi:10.1029/
Helmstetter A., D. Sornette, and J.-R. Grasso (2003), Main shocks are after-
shocks of conditional foreshocks: How do foreshock statistical properties
emerge from aftershock laws, J. Geophys. Res., 108(B1), 2046,
Hu, M., Z. Hu, G. Wei, W. Yang, and M. Liu (2012), Sequence lithofacies
paleogeography and reservoir prediction of the Maokou Formation in
Sichuan Basin, Petroleum Exploration and Development (In Chinese with
English abstract), 39(1), 45–55.
Hubbard,J., and J. H. Shaw(2009), Uplift of the LongmenShan and Tibetan
plateau, and the 2008 Wenchuan (M=7.9) earthquake, Nature, 458,
Huc, M., and I.G. Main (2003), Anomalous stress diffusion in earthquake
triggering: Correlation length, time dependence, and directionality,
J. Geophys. Res., 108(B7), 2324, doi:10.1029/2001JB001645.
Kanamori, H., and E. Hauksson (1992), A slow earthquake in the Santa
Maria basin, California, Bull. Seismol. Soc. Am., 82(5), 2087–2096.
Kerr, R. A. (2012), Learning how to not make your own earthquakes,
Science, 335, 1436–1437.
Korsch, R. J., H. Mai, Z. Sun, and J. D. Gorter (1991), The Sichuan Basin,
southwest China: A Late Proterozoic (Sinian) petroleum province,
Precambrian Res., 54, 45–63.
Kraft, T., P. M. Mai, S. Wiemer, N. Deichmann, J. Ripperger, P. Kästli,
C. Bachmann, D. Fäh, J. Wössner, and D. Giardini (2009), Enhanced geo-
thermal systems: Mitigating risk in urban areas, EOS, 90(32), 273–274.
failure inferred from pre-failure damage, Tectonophysics, 431, 97–111.
Lei, X.-L., C. Xie, and B. Fu (2011), Remotely triggered seismicity in
Yunnan, southwestern China, following the 2004 Mw9.3 Sumatra earth-
quake, J. Geophys. Res., 116, B08303, doi:10.1029/2011JB008245.
water injection at ~3 km depth within the Rongchang gas field, Chongqing,
China, J. Geophys. Res., 113, B10310, doi:10.1029/2008JB005604.
Liu, M, and Y. Yang (2005),Contrasting seismicitybetween the north China
and south China blocks: Kinematics and geodynamics, Geophys. Res.
Lett., 32, L12310, doi:10.1029/2005GL023048.
Lombardi, A. M., M. Cocco, and W. Marzocchi (2010), On the increase of
background seismicity rate during the 1997–1998 Umbria-Marche,
Central Italy, sequence: Apparent variation or fluid-driven triggering?,
Bull. Seismol. Soc. Am., 100(3), 1138–1152, doi:10.1785/0120090077.
Long, F., F. Du, X. Ruan, Y. Deng, and T. Zhang (2010), Water injection
triggered earthquakes in the Zigong mineral wells in ETAS model,
Earthquake Research in China(In Chinese with English abstract), 26(2),
Ma, Y., S. Zhang, T. Guo, G. Zhu, X. Cai, and M. Li (2008), Petroleum ge-
ology of the Puguang sour gas field in the Sichuan Basin, SW China, Mar.
Pet. Geol., 25, 357–370.
Maillot, B., S. Nielsen, and I. Main(1999),Numericalsimulation ofseismic-
ity due to fluid injection in a brittle poroelastic medium, Geophys. J. Int.,
Main, I. (1996), Statistical physics, seismogenesis, and seismic hazard, Rev.
Geophys., 34, 433–462.
Main, I.G., L. Li, K.J. Heffer, O. Papasouliotis, and T. Leonard (2006).
Long-range, critical-point dynamics in oilfield flow rate data, Geophys.
Res. Lett., 33, L18308, doi:10.1029/2006GL027357.
Majer, E. L., and J. E. Peterson (2007), The impact of injection on seismicity
at The Geysers, California Geothermal Field, Int. J. Rock Mach. Min. Sci.,
Marsan, D., and O. Lengline (2008), Extending earthquakes’ reach through
cascading, Science, 319, 1076–1079, doi:10.1126/science.1148783.
Marsan, D., C. J. Bean, S. Steacy, and J. McCloskey (2000), Observation
of diffusion processes in earthquake populations and implications for
the predictability of seismicity systems, J. Geophys. Res., 105(B12),
Marsan, D., E. Prono, and A. Helmstetter (2013), Monitoring aseismic forc-
ing in fault zones using earthquake time series, Bull. Seismol. Soc. Am.,
103(1), 169–179, doi:10.1785/0120110304.
McGarr, A., and D. W. Simpson (1997), A broad look at induced and triggered
seismicity, “Rockbursts andseismicityin mines”,inProc. of4th Int. Symp.on
Rockbursts and Seismicity in Mines, edited by S. J. Gilbowicz and S. Lasocki,
pp. 385–396, A.A. Balkema, Rotterdam, Poland.
Molchan, G. (2005), Interevent time distribution in seismicity: A theoretical
approach, Pure Appl. Geophys., 162, 1135–1150, doi:10.1007/s00024-
Molnar, P., and P. Tapponnier (1975), Cenozoic tectonics of Asia: Effects of
a continental collision, Science, 189, 419–426.
Nicholson, C., and R. L. Wesson (1992), Triggered earthquakes and deep
well activities, Pure Appl. Geophys., 139, 561–578.
Ogata, Y. (1992), Detection of precursory relative quiescence before great
earthquakes through a statistical model, J. Geophys. Res., 97,
Ogata, Y. (2001), Increased probability of large earthquakes near aftershock
regions with relative quiescence, J. Geophys. Res., 106, 8729–8744.
LEI ET AL.: INJECTION-INDUCED SEISMICITY
Parotidis, M., and S.A. Shapiro (2004), A statistical model for the seismicity Download full-text
rate of fluid-injection-induced earthquakes, Geophys. Res. Lett., 31,
Ruan, X., W. Cheng, Y. Zhang, J. Li, and Y. Chen (2008), Research of the
earthquakes induced by water injection in salt mines in Changning,
Sichuan, Earthquake Research in China (in Chinese with English abstract),
Ryder, R.T., Rice D.D., Z. Sun, Y. Zhang, Y. Qiu, and Z. Guo (1991),
Petroleum geology of the Sichuan basin, China Report on U.S.
Geological Survey and Chinese Ministry of Geology and Mineral
Resources Field Investigations and Meeting, October 1991.
Sang, Q., J. Huang, C. Cheng, Y. Wei, Z. Lu, C. Wu, and Z. Peng (2012),
Research on the relation between the ancient karst landform and the develop-
gion, Carsologica Sinica (In Chinese with English abstract), 31(2), 212–219.
Seeber, L., J. G. Armbruster, and W.-Y. Kim (2004), A fluid-injection-triggered
earthquake sequence in Ashtabula, Ohio: Implications for seismogenesis in
stable continental regions, Bull. Seismol. Soc. Am., 94(1), 76–87.
Segall, P. (1989), Earthquakes triggered by fluid extraction, Geology, 17,
Shapiro, S. A.,and C. Dinske (2009), Fluid-induced seismicity: Pressuredif-
fusion and hydraulic fracturing, Geophys. Prospect., 57, 301–310,
ability from fluid-injection-induced seismic emission at the KTB site,
Geophys. J. Int., 131, F15–F18.
Shapiro,S., P. Sudigane, and J.-J.Royer (1999), Large-scale in situ permeability
tensor of rocks from induced microseismicity, Geophys. J. Int., 137, 207–213.
Shen, Z.-K., J. Lü, M. Wang, and R. Bürgmann (2005), Contemporary
crustal deformation around the southeast borderland of the Tibetan
Plateau, J. Geophys. Res., 110, doi:10.1029/2004JB003421.
Sibson, R. H., J. M. M. Moore, and A. H. Rankin (1975), Seismic pumping
—A hydrothermal fluid transport mechanism, J. Earth Sci., 131,
location algorithm: Method and application to the northern Hayward fault,
California, Bull. Seismol. Soc. Am., 90(6), 1353–1368.
Wang, J., and Z. Luo (1989), Formation and development of the Sichuan
Basin, in Sedimentary Basins of the World, 1: Chinese Sedimentary
Basins, edited by X. Zhu, pp. 147–152, Elsevier, Amsterdam.
Wang, X., S. Ma, X. Lei, X. Guo, Q. Wang, G. Yu, X. Gou, Y. Kuwahara,
K. Imanish, and X. Jiang (2012), Fine velocity structure and relocation
of the 2010 ML5.1 earthquake sequence in Rongchang gas field,
Seismology and Geology (In Chinese with English abstract), 34(2),
Wei, G., G. Chen, S. Du, L. Zhang, and W. Yang (2008), Petroleum systems
of the oldest gas field in China: Neoproterozoic gas pools in the Weiyuan
gas field, Sichuan Basin, Mar. Pet. Geol., 25, 371–386.
Xu, H., G. Wei, C. Jia, W. Yang, T. Zhou, W. Xie, C. Li, and B. Luo (2012),
Tectonic evolution of the Leshan-Longnusi paleo-uplift and its control on
gas accumulation in the Sinian strata, Petrol. Explor. Dev., 39(4), 436–446.
Zhang, B., R. Chen, H. Li, Y. Qi, G. Mao, P. Liu, and F. Li (1993),
Correlation between seismicity and water injection in Ziliujing anticline,
Acta Seismol. Sin., 5(2), 253–256.
Zhang, X., N. Koutsebaloulis, K. Heffer, I. Main, and L. Li (2007), Coupled
geomechanics-flow modelling at and below a critical stress state used to
investigate common statistical properties of field production data, in
Structurally Complex Reservoirs, Geol. Soc. London special publications,
edited by S. Jolley, D. Barr, J.J. Walsh, and R.J. Knipe, 292, 453–468,
Special Publications, London, U.K.
Zhang, Z., W. Cheng, M. Liang, F. Long, Y. Xu, W. Chen, and S. Wang
(2012), Seismicity and stress field characteristics of earthquakes
induced by water injection in Jia 33 well of Zigong City, Sichuan,
Chinese J. Geophys. (in Chinese), 55(5), 1635–1645, doi:10.6038/j.
Zhuang, J., Y. Ogata, and D. Vere-Jone (2002), Stochastic declustering of
space-time earthquake occurrences, Jou. Am. Statist.l Assoc., 97(458),
LEI ET AL.: INJECTION-INDUCED SEISMICITY