Available via license: CC BY 4.0
Content may be subject to copyright.
Predicting the Future Performance of the Planned Seismic Network in Mainland
China
Jiawei Li1, Arnaud Mignan1,2, Didier Sornette1,3, and Yu Feng1
1Institute of Risk Analysis, Prediction and Management (Risks-X), Academy for Advanced
Interdisciplinary Studies, Southern University of Science and Technology (SUSTech), Shenzhen,
China.
2Department of Earth and Space Sciences, Southern University of Science and Technology
(SUSTech), Shenzhen, China.
3Department of Management, Technology and Economics (D-MTEC), Swiss Federal Institute of
Technology in Zürich (ETH Zürich), Zürich, Switzerland
Corresponding author: Jiawei Li (lijw@cea-igp.ac.cn)
Key Points:
The spatial distribution of completeness magnitude for the new broadband seismic
network in China is predicted.
The completeness magnitude of China will soon decrease to 2.0, which means
approximately 3 times more earthquakes available per year.
The new seismic network will achieve the goal of 99% coverage for optimal earthquake
prediction research based on seismic precursor.
Abstract
The new broadband seismic network in China will increase the number of stations from
approximately 950 to 2000. The higher-resolution monitoring of the frequent smaller
earthquakes expected inside Mainland China can be quantified via the completeness magnitude
(
M
c) metric. Using the Bayesian Magnitude of Completeness (BMC) method, we generate the
spatial distribution of
M
cpredicted for the new network, based on the prior model calibrated on
the current earthquake catalog (2012 to 2021) and network configuration. If 99% of Mainland
China is at present covered down to
M
c= 2.7, this threshold will soon fall to
M
c= 2.0. This
means approximately 3 times more earthquakes available per year. Based on the observation that
seismic precursors are most likely to be observed at least at 3 units below the mainshock
magnitude, the new seismic network shall achieve the goal of almost total coverage for optimal
seismic-based earthquake prediction research.
Plain Language Summary
The China Earthquake Administration (CEA) has currently launched an ambitious nationwide
seismic network project, which will increase the number of stations from approximately 2,000 to
15,000 in total, from 950 to 2000 for the broadband seismic stations used to compile earthquake
catalog. The new network is planned to go online by the end of 2022. In half of Mainland China,
the inter-station distance will soon be smaller than 100 km, 23% be 50 km, and 3% be 25 km. Of
all possible ways to characterize the higher-resolution monitoring of the frequent smaller
earthquakes expected inside Mainland China, the completeness magnitude (Mc) remains one of
the most commonly used. The completeness magnitude Mcmetric is defined as the smallest
magnitude value above which the Gutenberg-Richter law is validated, or in other words, as the
threshold above which earthquakes in a given space-time volume are detected with a probability
tending to one. Using the prior model of the Bayesian Magnitude of Completeness (BMC)
method calibrated on the Chinese earthquake catalog from January 1, 2012 to July 11, 2021, we
predict the spatial distribution of
M
cfor the new network based on the planned network
configuration. If almost the entire Mainland China is at present covered down to
M
c= 2.7, this
threshold will fall to
M
c= 2.0 in the near future. This means approximately 3 times more
earthquakes will be recorded in the complete catalog available for statistical analysis per year
(for
a
= 6.76 and
b
= 0.85 in the Gutenberg-Richter law log10
N
=
a
-
b·M
c). Based on the
observation that abnormal seismicity as precursors are most likely to be observed at least at 3
units below the mainshock magnitude, and assuming earthquakes to be potentially damaging at
M
≥ 5, the new seismic network shall achieve the goal of 99% coverage for optimal seismic-
based earthquake prediction research.
1 Introduction
The densification of a seismic network improves earthquake detectability to smaller
magnitudes (Engdahl and Bondár, 2011). In turn, access to smaller earthquakes improves
statistical seismology studies by providing more data and thus better insights into the underlying
crustal processes. Interestingly, micro-seismicity has been shown to be critical for the
observation of seismic precursors prior to large and potentially damaging earthquakes (Jones et
al., 1982; Bernard et al., 1997; Wang et al., 2006; Ebel, 2008; Mignan, 2014; Brodsky, 2019;
Trugman and Ross, 2019). Mignan (2014) proposed the following empirical law based on a
meta-analysis of earthquake precursor studies:
mmin <M- 3.0 (1)
where mmin is the minimum magnitude threshold necessary to start observing seismic precursors
prior to a mainshock of magnitude M. The completeness magnitude Mcis the simplest, and one
of the best proxies to the detection capability of a seismic network. Earthquake data, considered
complete above this limit, follow the Gutenberg-Richter frequency-magnitude distribution. The
optimal use of data requires that mmin =Mc. As of 2011, the minimum Mcin Mainland China was
about 3.7 (Mignan et al., 2013).
In the past ten years, the detection capability, or Mc, of the seismic network in Mainland
China has been improved with the upgrading of processing techniques (e.g., Wang et al., 2017),
while the basic network layout remains unchanged. The China Earthquake Administration (CEA)
has recently launched an ambitious nationwide network construction project that aims to extend
the scale and applications of the seismic network with a total investment of approximately 1.87
billion RMB (290 million USD;
https://www.cenc.ac.cn/cenc/zt/361404/361414/361563/index.html, last accessed: January 2022).
Through this project, the number of stations in the seismic network will increase from
approximately 2,000 to 15,000 in the future (Figure S1), with the data planned to be transmitted
in real-time. The customary assessment methods (e.g., Mignan and Woessner, 2012) to estimate
the Mcare hampered for this network at the planning, design and implementation stages by a lack
of observations.
The Bayesian Magnitude of Completeness (BMC) method proposed by Mignan et al.
(2011) provides a promising solution to address the above problem of lack of observations to
estimate Mcby using Bayesian inference relating the spatial density of stations in the network to
Mcvalues, weighted by the respective uncertainties of the observations and the prior model. This
method has been successfully applied in observed induced (Mignan, 2021) and natural seismicity,
e.g., Taiwan (Mignan et al., 2011), Mainland China (Mignan et al., 2013), Switzerland (Kraft et
al., 2013), Lesser Antilles arc (Vorobieva et al., 2013), California (Tormann et al., 2014), Greece
(Mignan and Chouliaras, 2014), Iceland (Panzera et al., 2017), South Africa (Brandt, 2019) and
Venezuela (https://www.statistics.gov.hk/wsc/CPS204-P47-S.pdf, last accessed: September
2021).
As an updating of the analysis applying BMC method to Mainland China by Mignan et al.
(2013), we begin this article with a evaluation of detectability capability based on the earthquake
catalog accumulated from 2012 to 2021 inside Mainland China. Once the BMC method is
calibrated for the existing seismic network, we then extend its application to predict the future
performance of the planned network. Finally, we quantify the potential improvement of
seismicity in a more complete catalog observed with the change in the network, and investigate
the significance for seismic-based earthquake prediction research based on the optimal use of
data, namely mmin =Mc. Our work will serve as an important reference to guide the design and
optimization of the planned seismic network upgrade in Mainland China.
2 Data
2.1 Earthquake catalog (2012-2021)
For our purpose, we define the inside of Mainland China as the region within 100 km
outside of the boundaries of Mainland China and Hainan Island (approximately 11.5 million
km2), where the China Earthquake Networks Center (CENC) produces a catalog based on the
same standard and workflow. Approximately 890,000 earthquakes, in which 18,000 with ML≥
3.0, that occurred inside Mainland China were reported in the catalog provided by the CENC
from January 1, 2012, to July 11, 2021. Among them, the largest and the most damaging
earthquake was the 2008 MS8.0 Wenchuan earthquake, which caused approximately 70,000
deaths and 20,000 missing people.
2.2 The seismic networks
The existing seismic network in Mainland China consists of a high-gain broadband
seismic network and a strong-motion network with a total of approximately 950 and 1,100
stations, respectively (Figure S1a). The new seismic network is being constructed by the
National System for Fast Report of Intensities and Earthquake Early Warning project of China.
This project is led by the CEA and was launched and implemented in 2015 and 2018,
respectively. The new network consists of a total of approximately 15,000 stations:
approximately 2,000 datum stations equipped with three-component broadband seismometers
and accelerometers, approximately 3,200 basic stations equipped with only three-component
accelerometers, and approximately 10,000 ordinary stations equipped with low-cost micro-
electro-mechanical system (MEMS) intensity sensors (Figure S1b). It is expected that, by end of
2022, the new network will officially be online (Wenhui Huang, Qiang Ma and Changsheng
Jiang, written comm., 2021). Only the data recorded by the broadband seismic stations in the
existing network and the datum stations in the planned network are used to compile catalogs.
We computed the inter-station distance from the average distance of a given site to its
closest fourth broadband and/or strong-motion stations (Kuyuk and Allen, 2013; Li et al., 2016;
Li et al., 2021) using a mesh of 0.5°×0.5° resolution. In 40% of Mainland China, the inter-station
distances of the existing network are less than 100 km, 7.8% are 50 km, and 0.4% are 25 km
(Figures S2a and S2c). In contrast, in half of Mainland China, the inter-station distance will soon
be smaller than 100 km, 23% be 50 km, and 3% be 25 km (Figure S2b and S2c). Although the
area with an inter-station distance smaller than 100 km to be covered by the new network is
almost the same as that covered by the old network, the areas with an inter-station distance
smaller than 50 km and 25 km are expanded by factors of 2 and 7, respectively. The density of
the new network is obviously improved. The new network will have an inter-station distance of
approximately 20-30 km in North China, the central China north-south seismic belt, the
southeast coast, the middle section of the Tianshan Mountains in Xinjiang, and approximately 10
km around large urban agglomerations, e.g., the center of Beijing Capital Circle, southeast
coastal city cluster, Lanzhou-Xi'an, Urumqi, Chengdu, and Kunming (Figure S2b). Overall, the
density of the new network in North China, the central China north-south seismic belt, the
southeast coast, the middle section of the Tianshan Mountains in Xinjiang will be comparable to
the most density seismic network worldwide, e.g., Japan, the U.S. west coast, and Taiwan (e.g.,
Kuyuk and Allen, 2013).
3 Monitoring capability estimated by Mc
3.1 Past network performance
Figure 1a shows Mcmaps for time periods from 2012 to 2021. We use the standard
frequency-magnitude distribution-based mapping method (Wiemer and Wyss, 2000) with grid
resolution of 0.5°×0.5° for each of them, and define a cylinder with a fixed radius of r= 50 km
and a minimum number of earthquakes Nmin = 50. The bulk Mcin each grid is computed based on
the events within a cylinder by using the non-parametric median-based analysis of the segment
slope technique (MBASS; Amorèse, 2007), and its deviation is obtained from 200 bootstrap
samples (Efron, 1979). Many gaps remain in grid points with low seismicity within the cylinder.
Simply increasing ris not recommended due to possible over-smoothing with obvious artifact
patterns (Mignan and Woessner, 2012). The BMC technique is based on the density of the
seismic network for determining r(Mignan et al., 2011). This method efficiently decreases the
number of gaps, but it can be used only where network information is available.
(a)
(b)
(c)
Figur e 1. (a) Observed
M
cmaps for time periods from 2012 to 2021. The median
M
cbased on
the
M
cspatial distribution is 1.8. The
M
cestimates are computed by using the median-based
analysis of the segment slope (MBASS; Amorèse, 2007) and based on the standard frequency-
magnitude distribution-based mapping method (Wiemer and Wyss, 2000). (b) Predicted Mcmap
based on prior information for the existing broadband seismic stations in Figure S1a. The
standard deviations σinside and outside Mainland China are 0.47 and 0.24, respectively. (c)
Posterior Mcmap estimated by the Bayesian magnitude of completeness (BMC) method for time
periods from 2012 to 2021. White and red lines show the Mc= 1.5 and Mcpost = 2.7 contours,
respectively.
The overall spatial Mcinside Mainland China in Figure 1a is basically the same as that
from October 1, 2008, to 31 August 2011 of Mignan et al. (2013), which estimated a median Mc
value of 1.6. The median Mcvalues based on its spatial distribution in Figure 1a is 1.8. Figure 1a
produced by the standard mapping method also exhibits many gaps in regions with low
seismicity where Mccannot be computed with confidence. The Mcvalues outside Mainland
China (e.g., Taiwan) are larger, because in this case, earthquakes were located by the CENC
using only the national broadband seismic network. The areas with Mcvalues smaller than 1.5
account for 17% of the total inside Mainland China. Areas with Mcvalues smaller than 1.5 are
sparse in the central China north-south seismic belt, northern part of Xinjiang, North China,
Southeast China, and Lhasa.
3.2 Bayesian Magnitude of Completeness (BMC) application
The BMC method proposed by Mignan et al. (2011) merges the observed Mcwith prior
Bayesian information, which is deduced from the relationship between the density of the seismic
network and Mcobservations, weighted by their respective standard deviations. The posterior Mc
is obtained in a two-step procedure. First, a spatial resolution optimization is implemented.
However, we use the standard mapping method with fixed parameters to obtain the spatial
distribution of Mcto avoid the iterated use of the network information in observed Mcand prior
Bayesian information, at the cost of the non-optimization of the spatial resolution (in contrast to
previous BMC applications in other regions). Therefore, only the standard frequency-magnitude
distribution-based mapping method is adopted in this article to produce Figure 1a. In addition,
we still use the MBASS technique instead of the maximum curvature (MAXC) approach to
consider the possible Mcheterogeneities introduced by the aforementioned choice. The MAXC
approach is a standard method of estimating the observed Mcin the BMC method while resulting
in large uncertainty, e.g., σ= 0.56 in Mignan et al. (2013) for the Chinese catalog. The MBASS
technique is here shown to reduce those uncertainties in the context of this study. Second, a
Bayesian approach to merge the observed Mcobs and the prior Mcpred is completed using Gaussian
conjugates, giving the posterior Mcpost for each grid as
(2)
in which σand σ0are the standard deviations for Mcpred and Mcobs, respectively. The posterior
standard deviation σpost can be estimated as
(3)
For the prior model, Mcpred =f[d(4)], in which d(4) is the distance to the fourth closest
station that is used to fit the observed Mc. As a novelty, we compare three models: (1) Mcpred =
a
·
d(4) + b, with a degree-of-freedom (df) of 2, (2) Mcpred =a
·
d(4)c+b, with a df of 3, and (3)
Mcpred =a
·
log10
d
(4) +
b
, with a df of 2. The observed grid value Mcin every year from 2012 to
2021 is used to evaluate the alternative models. The four indicators are the sum of squares due to
error (SSE), the coefficient of determination (R-square), the root mean squared error (RMSE)
and the Akaike information criterion (AIC; Akaike, 1970). We found that the model Mcpred =
a
·
d(4)c+bperforms best in terms of both the goodness of fit (SSE = 1438, R-square = 0.0755,
and RMSE = 0.4695) and model evaluation (AIC = -9862.8) (Table S1). This model is exactly
the one used in the former BMC, e.g., for Taiwan (Mignan et al., 2011). Figure 1b shows the
predicted Mcmap based on a prior model for the existing broadband seismic stations. The
standard deviations σinside Mainland China are 0.47.
Figure 1c shows the spatial distribution of the posterior Mcestimated by the BMC
technique. The distribution of the posterior Mcis more heterogeneous than that of the prior Mc
since the latter is a continuous function of the density of stations. The minimum posterior Mc
value in Mainland China was about 2.7. The areas with posterior Mcvalues smaller than 1.5 and
2.0 account for 38% and 91% inside Mainland China, respectively, and are 8% lower for both of
them than for the prior Mcvalues, namely, 1.5 for 46% and 2.0 for 99% (Figure 2a). The area
differences larger than 5% at Mc= 1.4 and 1.5 (Figure 2a) can be explained by the relatively
large uncertainty (σ= 0.47) of the prior model compared with that for Switzerland (σ= 0.16),
Taiwan (σ= 0.18), South Africa (σ= 0.18), Venezuela (σ= 0.19), Greece (σ= 0.29), Iceland (σ=
0.3) and induced seismicity (σ= 0.37) (see matching references in the introduction). However,
our σis smaller than that computed by Mignan et al. (2013) for the Chinese catalog from 2008 to
2011, in which σ= 0.56.
3.3 Bayesian Magnitude of Completeness (BMC) application
With the calibration of the BMC prior from the observed Mc, we can then predict the Mc
distribution for any seismic network configuration, assuming that the prior model is valid when
used in the same area (i.e., same seismic waveform attenuation laws). Figure 2b shows the
predicted Mcdistribution for the planned broadband seismic network. Figure 2a shows the
increase in the spatial distribution of the regions with Mcvalues smaller than 1.5. Quantitatively
compared with the posterior Mcof the existing network (Figure 1c), the area with Mc≤ 1.5
expands by a factor of 1.5 to reach 54%, and the Mc’s of almost the entire inside Mainland China
(99%) are less than 2.0. The monitoring capability in Xinjiang increases dramatically, followed
by the central China north-south seismic belt and North China. Southeast China shows the least
gain since the existing configuration is currently sufficient with only a small improvement of
monitoring capability needed.
(a)
(b)
Figure 2. (a) The binned distributions in prior (Figure 1b) and posterior (Figure 1c), and
predicted Mcvalues inside Mainland China. (b) Predicted Mcmap based on prior information for
the planned datum stations. The standard deviations σinside and outside Mainland China are
same as Figure 1b. White and red lines show the Mc= 1.5 and Mcpost = 2.0 contours, respectively.
4 Benefits from a more complete catalog
Better monitoring capability leads to the detection of more small earthquakes by offering
a higher resolution for active fault network structures (e.g., Ouillon and Sornette, 2011),
clarifying the relevant components of stress that would trigger subsequent earthquakes (e.g.,
Mancini et al., 2019), and helping to understand the earthquakes’ clustering behavior (e.g.,
Sornette and Werner, 2005a; 2005b) and the physics of the Earth’s crust (e.g., Mignan, 2011).
Indeed, according to current understanding, as a whole, small earthquakes contribute the most in
the triggering of other earthquakes, including medium size to large earthquakes (Helmstetter,
2003; Helmstetter et al., 2005; Nandan et al., 2021). It is therefore of the upmost importance to
record them, as much as possible (Ebel, 2008; Brodsky, 2019; Trugman and Ross, 2019).
The 99th percentiles of the posterior Mcfor the existing network (Figure 1c) and
prediction (prior) Mcfor the planned network (Figure 2b) are 2.7 and 2.0, respectively.
Statistically, 29,413 earthquakes with M≥ 2.7 were observed (on average, approximately 3,000
per year and 250 per month) from January 2012 to July 2021. According to the Gutenberg-
Richter law that fits the empirical magnitude distribution for magnitudes larger than 2.7 (a-value
of 6.76 and b-value of 0.85), approximately 500 earthquakes larger than 2.0 that are missed by
the existing network would be recorded in the planned network. These values suggest that the
available seismicity samples (approximately 115,800) will increase by a factor of at least 3. We
predict that the complete catalog of the planned broadband seismic network will record
approximately 12,100 and 1,000 earthquakes per year and month, respectively, which are both at
least a three-fold increase compared with the number of earthquakes recorded by the existing
network. The improvement would potentially strengthen both predictive skills of the statistical
and physical-based models because the sample number of earthquakes would be increased.
Figure 3. Demonstration of the decrease of mainshock magnitude Mfrom minimum foreshock
magnitude mmin = 2.7 for the existing network to mmin = 2.0 for the new network inside Mainland
China. A total of 77 data points as the background are from the meta-analysis of 37 published
studies by Mignan (2014), in which arguments are mainly based on heuristic, statistical or
physical considerations. Modified from Figure 2 of Mignan (2014).
Equation (1) that derives from a meta-analysis of 37 foreshock studies published from
1982 to 2013 shows that seismic precursors are most likely to be observed at least at 3 units
below the mainshock magnitude M(Figure 3). Figures 1c and 2b show that the minimum Mc
inside Mainland China is 2.7 for the existing network and 2.0 for the new network, respectively.
Considering the optimal use of data, namely mmin =Mc, we find that the magnitude Mof
mainshocks prior to which anomalous foreshock activity most likely to be found to emerge is
decreased from 5.7 for the existing network to 5.0 for the future network. Assuming that
earthquakes in Mainland China are potentially destructive for magnitude Mlarger than 5, the
future seismic network will achieve the goal of almost full coverage for optimal seismic
precursor-based earthquake prediction research.
5 Conclusions and discussion
The performance of the network will significantly improve by simply densifying the
existing layout. In half of Mainland China, the inter-station distance will soon be smaller than
100 km (and smaller than 50 km and 25 km for 23% and 3% of the mainland, respectively). The
area with an inter-station distance smaller than 50 km and 25 km that are to be covered by the
new network are expanded by factors 2 and 7, respectively. In this study, we quantified the
higher-resolution detection of the frequent smaller earthquakes inside Mainland China via the
completeness magnitude (Mc) metric for the existing and planned seismic networks. Using the
best performance prior model (σ= 0.47) of the Bayesian Magnitude of Completeness (BMC)
method calibrated on the Chinese catalog from January 1, 2012 to July 11, 2021, we predicted
the spatial distribution of Mcfor the future network based on its network configuration.
Compared with the posterior Mcof the existing network, the area with Mc≤ 1.5 that is covered
by the new network expands by a factor of 1.5 to reach 54%. If 95% of Mainland China is at
present covered down to Mc= 2.7, this value will soon fall to Mc= 2.0. We predicted that the
complete catalog of the planned network will record a factor of three more earthquakes per year
compared with that of the existing network (for a= 6.76 and b= 0.85 in the Gutenberg-Richter
law log10N=a-b
·
Mc). Taking the minimum Mcinside Mainland China as 2.0 for the future
network and assuming earthquakes to be potentially damaging at M ≥ 5, the new network shall
achieve the goal of almost total coverage for optimal seismic-based earthquake prediction
research in the entire Mainland China. Our investigation provides a useful reference for the real
functioning and further optimization of the new networks in Mainland China
The improvement in computing power makes the adoption of artificial intelligence (AI)
in the routine monitoring of seismic networks possible. Previous tests showed that more small
earthquakes by at least a factor of ten could be detected by AI compared with that produced by
the current processing techniques (e.g., Beroza et al., 2021). Therefore, we can expect that the
monitoring capabilities of the seismic network will usher a comprehensive and large
improvement in the near future with the revolution in seismic observational technology enabled
by AI and the further upgradation of the software in the near future.
Acknowledgments
The authors would like to thank Dr. Yang Zang, Dr. Zhengyi Yuan, Dr. Chen Yang, and Dr. Lei
Tian at the China Earthquake Networks Center; Dr. Ke Sun, and Dr. Zijian Cui at the Institute of
Earthquake Forecasting of China Earthquake Administration for providing the earthquake
catalog and information of stations. This project is supported by the National Natural Science
Foundation of China under Grant No. U2039202 and the Guangdong Basic and Applied Basic
Research Foundation under Grant No. 2020A1515110844.
Data Availability Statement
The earthquake catalog and information of the existing and the planned seismic networks were
provided by China Earthquake Networks Center (doi: 10.11998/SeisDmc/SN) through the
internal link (Earthquake Cataloging System at CEA: http://10.5.160.18/console/index.action).
The Bayesian magnitude of completeness method used in the study is from Mignan et al., (2011).
References
Akaike, H. (1970). Statistical predictor identification.
Ann. I. Stat. Math.
22, 203-217.
Amorèse, D. (2007). Applying a change-point detection method on frequency-magnitude
distributions.
Bull. Seismol. Soc. Am.
97, 1742-1749.
Bernard, P., Pinettes, P., Hatzidimitriou, P. M., Scordilis, E. M., Veis, G., & Milas, P. (1997).
From precursors to prediction: a few recent cases from Greece.
Geophys. J. Int.
131, 467-477.
Brandt, M. B. C. (2019). Performance of the South African National Seismograph Network from
October 2012 to February 2017: Spatially varying magnitude completeness.
S. Afr. J. Geol.
122,
57-68.
Brodsky, E. E. (2019). The importance of studying small earthquakes. Science 364, 736-737.
Ebel, J. E. (2008). The importance of small earthquakes. Seismol. Res. Lett. 79, 491-493.
Efron, B. (1979). 1977 rietz lecture, bootstrap methods-another look a the jackknife.
Ann. Statist.
7, 1-26.
Engdahl, E. R., & Bondár, I. (2011), Seismological Networks.
Encyclopedia of Solid Earth
Geophysics
(Gupta, H. K., ed.) 1324-1333.
Helmstetter, A. (2003). Is earthquake triggering driven by small earthquakes?.
Phys. Res. Lett.
91,
058501.
Helmstetter, A., Kagan, Y. Y., & Jackson, D. D. (2005). Importance of small earthquakes for
stress transfers and earthquake triggering. J. Geophys. Res. 110, B05S08.
Jones, L. M., Wang, B. Q., Xu, S. X., & Fitch, T. J. (1982). The foreshock sequence of the
February 4, 1975, Haicheng earthquake (
M
= 7.3).
J. Geophys. Res.
87, 4575-4584.
Kraft, T., Mignan, A., & Giardini, D. (2013). Optimization of a large-scale microseismic
monitoring network in northern Switzerland.
Geophys. J. Int.
195, 474-490.
Kuyuk, H. S., & Allen, R. M. (2013). Optimal seismic network density for earthquake early
warning: A case study from California.
Seismol. Res. Lett.
84, 946-954.
Li, J. W., & Wu, Z. L. (2016). Controlling the ‘blind zone’ of an earthquake early warning
system (EEWS): A case study of the Beijing Capital Circle prototype EEWS (in Chinese with
English abstract).
Ear thq. Res. China
32, 584-594.
Li, J. W., Böse, M., Feng, Y., & Yang, C. (2021). Real-time characterization of finite rupture and
its implication for earthquake early warning: Application of FinDer to existing and planned
stations in Southwest China.
Front. Earth Sci.
9, 699560.
Mancini, S., Segou, M., Werner, M. J., & Cattania, C. (2019). Improving physics-based
aftershock forecasts during the 2016-2017 Central Italy Earthquake Cascade.
J. Geophys.
Res.
124, 8626-8643.
Mignan, A. (2011). Retrospective on the Accelerating Seismic Release (ASR) hypothesis:
Controversy and new horizons.
Tectonophys.
505, 1-16.
Mignan, A. (2014). The debate on the prognostic value of earthquake foreshocks: A meta-
analysis.
Sci. Rep.
4, 4099.
Mignan, A., Werner, M. J., Wiemer, S., Chen, C. C., & Wu, Y. M. (2011). Bayesian estimation
of the spatially varying completeness magnitude of earthquake catalogs.
Bull. Seismol. Soc.
Am.
101, 1371-1385.
Mignan, A., & Woessner, J. (2012). Estimating the magnitude of completeness for earthquake
catalogs.
Community Online Resource for Statistical Seismicity Analysis
. doi:10.5078/corssa-
00180805. Available at http://www.corssa.org (last access: November 11, 2021).
Mignan, A., Jiang, C. S., Zechar, J. D., Wiemer, S., Wu, Z. L., & Huang, Z. (2013).
Completeness of the Mainland China earthquake catalog and implications for the setup of the
China Earthquake Forecast Testing Center.
Bull. Seismol. Soc. Am.
103, 845-859.
Mignan, A., & Chouliaras, G. (2014). Fifty years of seismic network performance in Greece
(1964-2013): Spatiotemporal evolution of the completeness magnitude.
Seismol. Res. Lett.
85,
657-667.
Mignan, A. (2021). Induced seismicity completeness analysis for improved data mining.
Front.
Ear th Sci.
9, 635193. doi: 10.3389/feart.2021.635193.
Nandan, S., Ouillon, G., & Sornette, D. (2021). Triggering of large earthquakes is driven by their
twins. Preprint ETH Zurich: http://arxiv.org/abs/2104.04592.
Ouillon, G., & Sornette, D. (2011). Segmentation of fault networks determined from spatial
clustering of earthquakes.
J. Geophys. Res.
116, B02306.
Panzera, F., Mignan, A., & Vogfjörð, K. S. (2017). Spatiotemporal evolution of the
completeness magnitude of the Icelandic earthquake catalogue from 1991 to 2013.
J. Seismol.
21,
615-630.
Sornette, D., & Werner, M. J. (2005a). Constraints on the size of the smallest triggering
earthquake from the ETAS model, Baath's Law, and observed aftershock sequences. J. Geophys.
Res. 110, B08304.
Sornette, D., & Werner, M. J. (2005b). Apparent clustering and apparent background
earthquakes biased by undetected seismicity. J. Geophys. Res. 110, B09303.
Tormann, T., Wiemer, S., & Mignan, A. (2014). Systematic survey of high-resolution
b
value
imaging along Californian faults: Inference on asperities.
J. Geophys. Res.
119, 2029-2054.
Trugman, D. T., & Ross, Z. E. (2019). Pervasive foreshock activity across southern
California. Geophys. Res. Lett. 46, 8772-8781.
Vorobieva, I., Narteau, C., Shebalin, P., Beauducel, F., Nercessian, A., Clouard, V., & Bouin, M.
P. (2013). Multiscale mapping of completeness magnitude of earthquake catalogs.
Bull. Seismol.
Soc. Am
. 103, 2188-2202.
Wang, K. L., Chen, Q. F., Sun, S. H., & Wang, A D. (2006). Predicting the 1975 Haicheng
earthquake.
Bull. Seismol. Soc. Am.
96, 757-795.
Wang, Y. W., Jiang, C. S., Liu, F., & Bi, J. M. (2017). Assessment of earthquake monitoring
capability and score of seismic station detection capability in China Seismic Network (2008-
2015) (in Chinese with English abstract).
Chinese J. Geophys.
60, 2767-2778.
Wiemer, S., & Wyss, M. (2000). Minimum magnitude of completeness in earthquake catalogs:
Examples from Alaska, the western United States, and Japan.
Bull. Seismol. Soc. Am.
90, 859-
869.
Supporting Information for
Predicting the Future Performance of the Planned Seismic Network in Mainland China
Jiawei Li1, Arnaud Mignan1,2, Didier Sornette1,3, and Yu Feng1
1Institute of Risk Analysis, Prediction and Management (Risks-X), Academy for Advanced
Interdisciplinary Studies, Southern University of Science and Technology (SUSTech), Shenzhen,
China.
2Department of Earth and Space Sciences, Southern University of Science and Technology
(SUSTech), Shenzhen, China.
3Department of Management, Technology and Economics (D-MTEC), Swiss Federal Institute of
Technology in Zürich (ETH Zürich), Zürich, Switzerland
Contents of this file
Figures S1 to Existing (as of 2018) and planned (as of 2022) stations in Mainland
China.
Figures S1 to Maps showing the inter-station distances for the existing and planned
networks in China.
Tables S1 to Characterization parameters of alternative models and the results of
their evaluation to observed
M
cvalues.
Introduction
Using the Bayesian Magnitude of Completeness (BMC) method, we generated the spatial
distribution of Mc predicted for the new network, based on the prior model calibrated on
the current earthquake catalog from January 1, 2012, to July 11, 2021, and network
configuration. We further quantify the potential improvement of seismicity in a more
complete catalog observed with the change in the network and investigate the significance
for seismic-based earthquake prediction research based on the optimal use of data. As an
important supplement to the text, this file contains plots of existing and planned stations
in Mainland China and maps showing the inter-station distances for the existing and
planned networks. Characterization parameters of alternative models and the results of
their evaluation to observed Mc values are also shown here.
(a)
(b)
Figure S1. (a) Existing (as of 2018) stations as part of the China Earthquake Networks
Center (CENC) and the China Strong Motion Network Center (CSMNC). (b) Planned (as of
2022) stations to be newly deployed or upgraded in the National System for Fast Report of
Intensities and Earthquake Early Warning of China. We define ‘inside’ in this study (black
line) as the region within 100 km outside of Mainland China’s boundaries.
(a)
(b)
(c)
Figure S2. Maps showing the inter-station distances for the (a) existing and (b) planned
networks. For any given site (assuming a grid of 0.5° × 0.5° resolution). The inter-station
distance is calculated as the average distance to the four closest stations (Kuyuk and
Allen, 2013; Li et al., 2016; Li et al., 2021). The histogram in (c) shows the binned
distribution of the inter-station distances inside Mainland China.
Table S1. Characterization parameters of alternative models and the results
of their evaluation to observed
M
cvalues.
Model
a
b
c
df
SSE
R-square
RMSE
AIC
y
=
a
·
d
(4) +
b
0.002
1.255
--
2
1448
0.0687
0.4712
-9819.6
y
=
a
·
d
(4)
c
+
b
0.128
0.767
0.365
3
1438
0.0755
0.4695
-9862.8
y
=
a
·log10
d
(4) +
b
0.602
0.261
--
2
1442
0.0731
0.4701
-9846.7
d
(4): distance to the closest fourth station; df: degree-of-freedom; SSE: sum of squares due to error;
R-square: coefficient of determination; RMSE: root mean square error; AIC: Akaike information
criterion.
