sensors
Article
A Synchronous Magnitude Estimation with P-Wave Phases’
Detection Used in Earthquake Early Warning System
Dingwen Zhang 1 , Jihua Fu 1,2, * , Zhitao Li 1 , Linyue Wang 1 , Jiale Li 2 and Jianjun Wang 1
Citation: Zhang, D.; Fu, J.; Li, Z.;
Wang, L.; Li, J.; Wang, J. A
Synchronous Magnitude Estimation
with P-Wave Phases’ Detection Used
in Earthquake Early Warning System.
Sensors 2022, 22, 4534. https://
doi.org/10.3390/s22124534
Academic Editor: Filippo Ubertini
Received: 16 May 2022
Accepted: 14 June 2022
Published: 16 June 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Sensors 2022, 22, 4534. https://doi.org/10.3390/s22124534
1 National Institute of Natural Hazards, Ministry of Emergency Management of China, Beijing 100085, China;
zdw55922@163.com (D.Z.); zhitao.lee@163.com (Z.L.); 17860612108@163.com (L.W.);
wjj2855@vip.sina.com (J.W.)
2 Institute of Disaster Prevention, Sanhe 065201, China; jialelee666@163.com
* Correspondence: jihuafu@ninhm.ac.cn
Abstract: How to estimate an earthquake’s magnitude rapidly and accurately is a challenge for any
earthquake early warning system. In order to reach a balance between accuracy and timeliness,
a synchronous magnitude estimation method with P-wave phases’ detection is proposed. In this
method, the P-wave phases are detected by the changes of the signal-to-noise ratio (SNR) of the
seismic records, where the SNRs are calculated by the short-term power and long-term power ratio
(STP/LTP). Meanwhile, the variations of the SNR are applied to estimate the magnitude of the
earthquake. By the statistics of some earthquake cases, a synchronous magnitude estimation model
of the variation of the P-wave phases’ SNR, the earthquake magnitude, and the hypocentral distance
was built. Compared with some other magnitude estimation methods, the suggested method inherits
the robustness of the STP/LTP method and is more accurate and rapid than the peak displacement
(Pd ) method.
Keywords: earthquake early warning system; synchronous magnitude estimation; P-wave phases’
detection; STP/LTP; SNR
1. Introduction
A rapid and accurate earthquake warning system (EEW) can effectively reduce ca-
sualties and economic losses caused by earthquakes. The concept of an earthquake early
warning system was first proposed by Dr. Cooper of the United States in 1868 [1] and
has been built in many areas and achieved good results in earthquake prevention and
mitigation [2–6]. Where magnitude estimation plays a key role in earthquake early warning
system, the data used to estimate the magnitude are obtained by seismic sensors. The
MEMS accelerometer in the seismic sensor monitors the ground motion data in real time,
then calculates the parameters required for magnitude estimation through the correspond-
ing algorithm. On the premise of stable performance of the seismic sensors [7], it is a
challenge to improve the timeliness and accuracy of the magnitude estimation algorithm.
As two important parameters for estimating the magnitude of an earthquake, timeliness
and accuracy are always a pair of contradictions. A large number of research efforts have
been made to keep the balance between accuracy and timeliness. At present, there are
many methods for real-time magnitude estimation, mainly including periodic methods,
displacement amplitude methods, energy methods, and machine learning methods.
The τpmax method, as one of periodic methods, was first proposed by Nakamura, which
calculates the predominant period of ground motion according to real-time seismic velocity
records [1]. This method has certain limitations, and its accuracy and stability are easily
affected by sampling rate and data preprocessing. Selecting different filters or filtering
time windows will have a great impact on the calculation results of the characteristic
period. Allen and Kanamori improved the τp algorithm, and it was regarded that the
maximum value of the dominant period within a few seconds of the P waves phases’
https://www.mdpi.com/journal/sensors
Sensors 2022, 22, 4534 2 of 17

arrival is proportional to the magnitude. This method is applied to the Elarms system in
the United States and has achieved good results in earthquake early warning [3]. Then,
Kanamori proposed the τc method based on the predominant period τpmax method, which
improved the calculation method of characteristic period and was more stable and reliable
than the τpmax method [8]. However, this method shows a saturation phenomenon in
estimating strong earthquakes, and its accuracy is also affected by the length of time
window. Moreover, Wu et al., carried out studies on the basis of the τc method [9]. Yamada
et al., found that the τc method displays the saturation phenomenon for big earthquakes [10].
In the aspect of displacement amplitude method, Wu et al., proposed the ML10 method in
1998, which comprehensively calculates the estimation results of earthquake magnitude
by using the waveform records of all successive triggering stations within 10 s of the first
triggering station [11]. Unfortunately, this method is only suitable for the areas with high
network density. Wu and Kanamori found that there was a good correlation between the
displacement amplitude Pd and the peak velocity PGV of ground motion within 3 s after P
waves phase’s detection [12]. After that, Wu and Zhao proposed the empirical relationship
between Pd , hypocentral distance R and magnitude M, which is more stable than τp method
and τc method [13]. This method is more stable than the τp method and τc method, but there
are still some limitations. The Pd method also has the saturation phenomenon in the large
earthquake, and with the increase of time window, the critical ‘saturation’ magnitude will
increase accordingly. Lancieri et al., further studied Wu’s method with strong earthquake
records with hypocentral distance within 50 km [14]. Li et al., and Wang et al., conducted a
comparative analysis of the periodic methods and the displacement amplitude methods,
and established the corresponding regression model [15,16].
As for the energy methods, the cumulative absolute velocity (CAV) parameter is
used as one of important criteria for magnitude estimation in the EEW system in Istanbul,
Turkey [17]. The limitation of this method is that the parameter calculation needs a certain
time. In 2008, Yamada et al., proposed a new magnitude estimation parameter, intensity
magnitude M I . According to the recorded results of the station, the magnitude can be
calculated in real time and the results can be updated [10].
In recent years, with the rapid development of computer technology, machine learning
and other technologies have also been applied in magnitude estimation. For example,
Zhu et al., used a deep revolutionary neural network to estimate magnitude, and Mousavi
et al., and Hu et al., applied machine learning methods to achieve accurate magnitude
estimation [18–20]. The error of the machine learning model in this method is obviously
smaller than that of the traditional method, but its parameters need to be determined at the
3 s after the arrival of P waves, which is the limitation in the timeliness.
The magnitude estimation of the above-mentioned methods is carried out a few
seconds after the P waves phase’s detection in EEW system. Usually, the magnitude
estimation and the P waves phase’s detection are accordingly figured out. A method
for synchronous magnitude estimation with P waves’s phase detection is proposed in
this paper. In the proposed synchronous magnitude estimation method, one P waves
phase’s detection method based on signal-to-noise ratio (SNR) was modified to estimate
the magnitude right after the P phases detection [21]. The novel magnitude estimation
method is put forward according to the relationship among the instantaneous maximum
SNR, magnitude M, and hypocentral distance R. Based on the same algorithm model,
the magnitude estimation is synchronously executed with the P waves phase’s detection.
The proposed method can effectively improve the efficiency of magnitude estimation in
EEW system.

2. Materials and Methods

At present, many studies support the magnitude estimation based on a small amount
of initial information of earthquakes, and the τc method and Pd method are widely used.
There is a nucleation seismic phase in the initial fracture process of the fault, which directly
determines the final fracture morphology and the magnitude of the earthquake [22]. P
Sensors 2022, 22, 4534 3 of 17

waves and S waves produced by the earthquake carry a lot of seismic information, P waves
carry the information of fault sliding, and S waves carry the main energy information of
the earthquake. In this paper, it is assumed that there is a certain weak seismic energy in P
waves, and there is a certain attenuation relationship between the energy and hypocentral,
and the magnitude is estimated by using this initial energy information and hypocentral.
The relationship between the magnitude M, maximum displacement peak Pd , and
hypocentral distance R proposed by Wu et al., is shown in Equation (1).

M = A × log( Pd ) + B × log( R) + C (1)

where M is the surface wave magnitude, A, B, and C are the fitting coefficients, Pd is the
peak amplitude of displacement in the first three seconds after the arrival of the P waves,
and R is the hypocentral distance.
The STP/LTP method provides a new method for P waves’ arrival extraction and
rapid calculation of instantaneous SNR [21]. Here, it can be considered to use the maximum
ratio PSNR of effective energy of instantaneous SNR to background noise as a new measure
to replace the Pd parameter to reconstruct this relationship model so as to realize the syn-
chronous estimation of earthquake magnitude. According to the literature and observation
records, the regression relationship among magnitude M, PSNR , and hypocentral distance
R is put forward, namely

M = A × log( PSNR ) + B × log( R) + C (2)

where M is the surface wave magnitude, A, B, and C are the fitting coefficients, PSNR is the
instantaneous maximum value of STP/LTP corresponding to the arrival of P waves, and R
is the hypocentral distance.
The STA/LTA method is often used as the criterion for P waves’ arrival. The long-time
window represents the energy of background noise and the short-time window represents
the energy of P waves’ effective signal. When the background noise is stable, namely when
the background noise conforms to the state of stationary random process, the effective signal
energy of P waves can be used as the representative of seismic energy. The Equation (3) is
the calculation equation of STA/LTA.

STAi ∑ij=i+1−S CFj /S

= i (3)
LTAi ∑k=i+1− L CFk /L

where STAi and LTAi are the mean value of characteristic function of short window and
long window respectively, CF is the characteristic function, and S is the length of the short
window and L is the length of the long window.
According to the internal relationship between signal-to-noise ratio (SNR) and STA/LTA,
Fu et al., proposed the automatic extraction method (STP/LTP) of seismic P waves’ arrival
time based on SNR. SNR is often used to represent the relative strength of effective signal
and background noise. The SNR calculation equation is shown in Equations (4) and (5).

SNR = 10lg( Pe /Pn ) (4)

dSNR = Pe /Pn (5)

where SNR is signal-to-noise ratio, Pe is the effective signal power, Pn is the background
noise power, and dSNR is the signal-to-noise ratio expressed linearly. Where Pe and Pn are
usually described by relative power, the calculation method is Equation (6):
N
P= ∑i=1 (xi − x)2 /N (6)

where xi (i = 1, 2, . . . , N) is the value of the observed signal and x is the mean value of the
observed signal.
Sensors 2022, 22, x FOR PEER REVIEW 4 of 1

where 𝑥𝑖 (𝑖 = 1,2, … , 𝑁) is the value of the observed signal and 𝑥̅ is the mean value of th
Sensors 2022, 22, 4534 observed signal. 4 of 17
In the STP/LTP method, in order to represent the energy of the ground vibration sig
nal effectively, the characteristic function is constructed from the perspective of relativ
power, asSTP/LTP
In the shown inmethod,
Equation in (7).
order to represent the energy of the ground vibration
signal effectively, the characteristic function is constructed from the perspective of relative
power, as shown in Equation (7). 𝐶𝐹 = (𝑥𝑖 − 𝑥̅ )2 (7
The STP/LTP equation is
CF = ( xi − x )2 (7)
2
𝑆𝑇𝑃𝑖 ∑𝑖𝑗=𝑖+1−𝑆(𝑥𝑗 − 𝑥̅𝑗 ) /𝑆 𝑃𝑆𝑖
The STP/LTP equation is = 𝑖 = (8
𝐿𝑇𝑃𝑖 ∑𝑘=𝑖+1−𝐿(𝑥𝑘 − 𝑥̅𝑘 )2 /𝐿 𝑃𝐿𝑖
i

2
When the STP/LTPSTP value ∑
exceeds j − x j /S
j=i +1−S axcertain Pi
threshold, the arrival of P waves
i
= i 2
= Si (8) can b
LTP ∑k=value x − x /L P
determined. At this time, the peak
i i +1− L ( of k)
k STP/LTP in PLwaves is 𝑃𝑆𝑁𝑅 , as shown in Equa
tion (9).
When the STP/LTP value exceeds a certain threshold, the arrival of P waves can be
determined. At this time, the peak value of STP/LTP in𝑆𝑇𝑃
P𝑖 waves is PSNR , as shown in
Equation (9). 𝑃𝑆𝑁𝑅 = max⁡ ( ) (9
  𝐿𝑇𝑃𝑖
STPi
PSNR =flowchart
Figure 1 shows the application max (9) appl
LTPof
i
the new method. In the practical
cation scenario, the sensor performs real-time monitoring, and the data is transmitted t
Figure 1 shows the application flowchart of the new method. In the practical appli-
the computer in real
cation scenario, the time
sensor to calculate
performs the monitoring,
real-time STP/LTP at andeachthetime
datainisthe observation
transmitted to signa
When
the the STP/LTP
computer exceeds
in real time the preset
to calculate trigger threshold
the STP/LTP at each timeatina the
certain time, that
observation means tha
signal.
the arrival
When of P waves
the STP/LTP exceedsis determined, which
the preset trigger is recorded
threshold as the time,
at a certain arrival time
that of Pthat
means waves, an
then
the arrival 𝑃𝑆𝑁𝑅
the of ofsensors-1730527-R2
P waves is determined, which P waves is determined.
is recorded as the arrivalFinally,
time of the 𝑃𝑆𝑁𝑅 and
P waves, and the hy
then the PSNR
pocentral of P waves
distance is determined.
obtained Finally,into
are substituted PSNRrelationship
the the and the hypocentral
model for distance
magnitude es
obtained
timation.are substituted into the relationship model for magnitude estimation.

Figure1.1.Application
Figure Application flowchart
flowchart ofnew
of the the magnitude
new magnitude estimation
estimation methodmethod
(PSNR ). (𝑃𝑆𝑁𝑅 ).
Sensors 2022, 22, x FOR PEER REVIEW 5 of 19

Sensors 2022, 22, 4534 5 of 17

3. Performance Analysis and Results

3. Performance
Sichuan ProvinceAnalysis and Results
of China is located between the Indian Ocean Plate and the Eura-
sian Plate, where crustal movement
Sichuan Province of China is located is very frequent.
between theThere
Indianare many
Ocean seismic
Plate belts
and the in Si-
Eurasian
chuan Province. Under the action of multiple seismic belts, the underground
Plate, where crustal movement is very frequent. There are many seismic belts in Sichuan rock strata
are fracturedUnder
Province. and misplaced,
the actionresulting
of multiplein the occurrence
seismic belts,of large
the and small earthquakes
underground rock strata arein
the region.and
fractured Wenchuan
misplaced, M8.0 earthquake
resulting in theonoccurrence
12 May 2008 caused
of large anda large
smallnumber of casual-
earthquakes in the
ties and economic
region. Wenchuanlosses. M8.0 earthquake on 12 May 2008 caused a large number of casualties
andData for this
economic study were provided by the National Key Research and Development
losses.
Program DataofforChina (No. were
this study 2019YFC1509205),
provided by the Institute
National of Engineering
Key ResearchMechanics, China
and Development
Earthquake Administration and China Earthquake Networks
Program of China (No. 2019YFC1509205), Institute of Engineering Mechanics, China Center, National Earth-
quake Data Center
Earthquake (http://data.earthquake.cn,
Administration and China Earthquake accessed
Networks on 31 October
Center, 2021).Earthquake
National Figure 2
shows the locations
Data Center and distributions of the
(http://data.earthquake.cn, main epicenters
accessed on 31 October and stations. There2are
2021). Figure 75
shows
the locations
records and
of large, distributions
medium, of the
and small main epicenters
earthquakes for model andfitting
stations.andThere are 75 records
performance anal-
of large,
ysis, medium,
including Wenchuanand small
M8.0earthquakes
(12 May 2008), forJiuzhaigou
model fitting M7.0 and performance
(8 August 2017), analysis,
Luxian
including
M6.0 Wenchuan2021),
(16 September M8.0Xingwen
(12 May 2008),
M5.7 (16Jiuzhaigou
December M7.0 (8 August
2018), Changning2017),M4.7(17
Luxian M6.0No-
(16 September 2021), Xingwen M5.7 (16 December 2018),
vember 2021), Changning M4.6 (21 November 2021), Changning M4.0 (27 November Changning M4.7(17 November
2021),M3.6
2019), Changning
(17 March M4.62020),
(21 November
and M3.02021), Changning
(12 December M4.0
2020, 23(27 November
March 2021) in2019), M3.6
Sichuan
(17 MarchIn2020),
Province. addition,and eight
M3.0 earthquake
(12 December 2020,are
events 23selected
March 2021) in Sichuan
for model Province.
testing, includingIn
addition, M3.0
Gongxian eight(13 earthquake
March 2021),events are selected
Gongxian M3.4 (7forNovember
model testing,2020), including
Gongxian Gongxian
M3.8 (23
M3.02020),
June (13 March
Gongxian 2021), Gongxian
M4.0 M3.4
(June 12, (7 November
2020, 15 February2020),2021),Gongxian
GongxianM3.8 M4.1(23
(13June 2020),
Novem-
Gongxian M4.0 (June 12, 2020, 15 February 2021), Gongxian
ber 2020), Xingwen M5.7 (16 December 2018), Luxian M6.0 (16 September 2021). In these M4.1 (13 November 2020),
Xingwen
records, theM5.7 (16 December
verticle 2018), ground
(U−D) direction Luxian M6.0
motion(16data
September
has the2021).
stable In these records,
background the
noise,
(U −
which is convenient for seismic phase identification. In the process of selecting groundis
verticle D) direction ground motion data has the stable background noise, which
convenient
motion data,for
theseismic phase
principle of identification. In the process
station data selection of selecting
is as follows: ground
for the M ≤motion data,
5.0 earth-
the principle
quakes, of station
the station data data
withselection
hypocentralis asdistance
follows: less
for the
thanM50 ≤ km5.0 earthquakes, the the
are selected; for station
M
data with hypocentral distance less than 50 km are selected; for
> 5.0 earthquakes, station data with hypocentral distance less than 100 km are selected. the M > 5.0 earthquakes,
station data with hypocentral distance less than 100 km are selected.

Locations and distributions

Figure2.2.Locations
Figure distributionsofofthe
themain
mainepicenters and
epicenters stations.
and ForFor
stations. M ≤M5.0≤earthquakes,
the the 5.0 earth-
quakes, the hypocentral distance is within 50 km, and for the M > 5.0 earthquakes, the hypocentral
the hypocentral distance is within 50 km, and for the M > 5.0 earthquakes, the hypocentral distance is
distance is within
within 100 km. 100 km.
Sensors 2022, 22, 4534 6 of 17

Sensors 2022, 22, x FOR PEER REVIEW 6 of 19

3.1. Small Earthquake Data Analysis

3.1. Small Earthquake
In the sampleData the data of 3.0 ≤ M ≤ 4.7 small earthquakes were provided by
Analysis
data,
the In
National Keydata,
the sample Research and
the data of Development Program
3.0 ≤ 𝑀 ≤ 4.7 small of China
earthquakes (No.
were 2019YFC1509205).
provided by
TheNational
the data is monitored
Key Research byand
low-cost MEMSProgram
Development seismic sensors
of Chinanamed MEMS Network Strong
(No. 2019YFC1509205).
The data is
Motion monitored by(MNSMS),
Seismograph low-cost MEMSwhichseismic sensors named
is equipped with MEMS Network Strongthree-axis
a high-performance
Motion Seismograph
linear Class C MEMS (MNSMS), which and
accelerometer, is equipped with a high-performance
the photograph of the MNSMS isthree-axis
shown in Figure 3.
linear Class C MEMS
The MNSMS accelerometer,
is mainly composed and
ofthe photograph
some hardware of modules:
the MNSMSthe is shown
MEMSinaccelerometer
Fig-
ure 3. The TCP/IP
module, MNSMS module,
is mainlyPower
composed
overofEthernet
some hardware modules: and
(PoE) module, the MEMS accel- module
local storage
erometer module, TCP/IP module, Power over Ethernet (PoE) module, and local storage
(optional). The MNSMS can meet the needs of dense EEW in terms of noise, dynamic range,
module (optional). The MNSMS can meet the needs of dense EEW in terms of noise, dy-
useful resolution, reliability, and detecting capabilities with its high-performance [7]. The
namic range, useful resolution, reliability, and detecting capabilities with its high-perfor-
sampling
mance ratesampling
[7]. The of the MNSMS is MNSMS
rate of the 50 Hz. is 50 Hz.

The
Figure3.3.The
Figure photograph
photograph of MNSMS.
of the the MNSMS. Theintext
The text the in
redthe redislabel
label MEMS is Network
MEMS Network Strong Motion
Strong Mo-
tion Seismograph.
Seismograph. Thetext
The texton
onscreen
screen is
is the
the coordinate.
coordinate.

Figure
Figure 4 shows
4 shows thethe
records of verticle
records (U-D)(U-D)
of verticle direction of a M4.0
direction of earthquake event andevent and
a M4.0 earthquake
the corresponding results of STP/LTP, STP, LTP calculation.
the corresponding results of STP/LTP, STP, LTP calculation. In Figure 4c, In Figure 4c, it can be it
clearly
can be clearly
seen that when the P wave arrives, the STP/LTP value increases significantly. Soon, the
seen that when the P wave arrives, the STP/LTP value increases significantly. Soon, the
𝑃𝑆𝑁𝑅 value appears, which indicates that the ground motion information contains some
PSNR value appears, which indicates that the ground motion information contains some
energy at the initial stage of the P waves, and the energy is significantly more than the
energy ofat
energy thethe initial stage
background noise.of Itthe
canPbe waves, andthe
found that the𝑃energy is significantly more than the
𝑆𝑁𝑅 value corresponding to the
energy of the background noise. It can be found
S wave is smaller than the P wave’s, which is not that the energySNR that the P of S value
waves corresponding
is small, but to the
S wave is smaller than the P wave’s, which is not that the
the observation signal in the long-time window is mostly the energy of seismic waves. In energy of S waves is small, but
the observation
Figure 4e,f, when the signal in the
seismic long-time
signal does notwindow
arrive, theis mostly
relative the energy
power in theofshort-time
seismic waves. In
Figure 4e,f,
window is thewhen
relativethe seismic
power of thesignal does not
background arrive,
noise, whichtheis relative
close to 0,power
and theinstate
the isshort-time
very
windowstable.isWhen the seismic
the relative power waves arrive,
of the the relativenoise,
background powerwhich
in the is
short-time
close to 0, window
and the state is
isvery
the sum
stable. of the
When relative power ofwaves
the seismic the seismic
arrive,signal and the power
the relative relativeinpower of the back-window is
the short-time
ground
the sumnoise of thein relative
the short-time
powerwindow, but thesignal
of the seismic relativeandpower of the background
the relative power of the noise
background
in the short-time window is much smaller compared with the relative
noise in the short-time window, but the relative power of the background noise in the power of the seismic
signal. Therefore, in the process of an earthquake, the relative power of the short-time
short-time window is much smaller compared with the relative power of the seismic signal.
window can represent the energy of the ground motion signal. Finally, when the seismic
Therefore, in the process of an earthquake, the relative power of the short-time window can
waves are over, only background noise is left in the short-time window, and the relative
represent the energy of the ground motion signal. Finally, when the seismic waves are over,
power is back to about 0. In Figure 4g,h, when the seismic signal does not arrive, the rel-
onlypower
ative background noise is left
of the long-time windowin the is short-time
the relative window,
power of the andbackground
the relativenoise,powertheis back to
about 0. In Figure 4g,h, when the seismic signal does not arrive,
value is close to 0, and the state is also very stable. When the seismic waves arrive, the the relative power of the
long-time window is the relative power of the background
relative power of the long-time window is the sum of the relative power of the seismic noise, the value is close to 0,
and the
signal state
in the is also very
long-time window stable.
andWhen the seismic
the relative power waves arrive, the relative
of the background noise. In power
the of the
long-time
process of an window
earthquake,is the
thesum of the
relative relative
power of the power of the seismic
long window can alsosignal in the
represent thelong-time
energy of the ground motion signal. However, unlike the short-time
window and the relative power of the background noise. In the process of an earthquake, window, the length
of therelative
the long windowpowerisofvery thelong.
longAfterwindow averaging,
can also therepresent
relative power of the seismic
the energy signal motion
of the ground
signal. However, unlike the short-time window, the length of the long window is very long.
After averaging, the relative power of the seismic signal increases and decreases slowly.
Finally, when the seismic waves are over, only background noise is left in the long-time
window, and the relative power is back to about 0.
 Sensors 2022, 22, x FOR PEER REVIEW 7 of 19

Sensors 2022, 22, 4534 increases and decreases slowly. Finally, when the seismic waves are over, only back- 7 of 17
ground noise is left in the long-time window, and the relative power is back to about 0.

(a) (b)

(c) (d)

(e) (f)

(g) (h)

Figure 4. (a) The record of the U−D direction in the M4.0 earthquake. (b) The local enlargement
result of (a). (c) The STP/LTP result of the record. The length of the short-time window is 0.3 s. The
length of the long-time window is 3 s. (d) The local enlargement result of (c). (e) The STP result
of the record. (f) The local enlargement result of (e). (g) The LTP result of the record. (h) The local
enlargement result of (g).
 Figure 4. (a) The record of the U−D direction in the M4.0 earthquake. (b) The local enlargement
result of (a). (c) The STP/LTP result of the record. The length of the short-time window is 0.3 s. The
length of the long-time window is 3 s. (d) The local enlargement result of (c). (e) The STP result of
the record. (f) The local enlargement result of (e). (g) The LTP result of the record. (h) The local
Sensors 2022, 22, 4534 enlargement result of (g). 8 of 17

It is worth noting that when the noise is weak and stable, to a certain extent, although
long-time It is and
worth short-time
noting that windows
when the can represent
noise is weakthe and energy
stable,oftoseismic
a certain waves,
extent,our pur-
although
pose is to estimate the magnitude through the initial information
long-time and short-time windows can represent the energy of seismic waves, our purpose of P waves. If the short-
time
is towindow
estimateisthe justmagnitude
at the arrival time ofthe
through P waves, then the short-time
initial information of P waves.window
If therepresents
short-time
the relativeispower
window just at of theP arrival
waves,time while ofthe relativethen
P waves, power
the of P waves window
short-time in the long-time
represents win- the
dow is only
relative S/L after
power averaging.
of P waves, whileSince
the Srelative
is much less than
power of P L, the relative
waves power of the
in the long-time windowlong-is
time
onlywindow
S/L after is still considered
averaging. SinceasSthe powerless
is much of background noise atpower
than L, the relative the arrival
of thetime of P
long-time
waves. In this way, when the P waves arrive, it can not only effectively
window is still considered as the power of background noise at the arrival time of P waves. detect the P waves,
but
In also obtainwhen
this way, the relative
the P wavesenergy proportion
arrive, it can notat the
onlyinitial time ofdetect
effectively the P the
waves, and the
P waves, but
𝑃𝑆𝑁𝑅
alsovalue
obtainappears soon energy
the relative after the arrival of at
proportion thethe P waves, which
initial time is the
of the significance
P waves, and the of this
PSNR
method. Multiple
value appears experimental
soon results
after the arrival show
of the that thewhich
P waves, time isof the 𝑃𝑆𝑁𝑅 valueof
thesignificance is this
verymethod.
close
toMultiple
the timeexperimental
of the P waves’ results show
arrival, that the
which usually of the Pwithin
time occurs SNR value
2 s is very
after theclose to
arrival the
of time
the
P of the P and
waves, waves’ arrival, which
sometimes usually
even occurs occurs within 2with
simultaneously s after
thethetimearrival
of theofP the P waves,
waves’ arrival. and
sometimes even occurs simultaneously with the time of
Small earthquakes occur frequently in earthquake events. The signal’s energy ofthe P waves’ arrival.
Small earthquakes
small earthquakes is weak occur
andfrequently
the SNR isinlow, earthquake events. The
so the detection signal’s
of P waves’energy
arrivaloftime
small
earthquakes
faces is weak and
many difficulties. Thethe SNR ismethod
STP/LTP low, so the has detection of P waves’
strong robustness, canarrival
adapt timeto weakfaces
manygeneral
noise, difficulties.
noise,The andSTP/LTP
strong noise;method has strong robustness,
its misjudgment rate and missingcan adaptrate to
areweak
low. noise,
The
general
𝑃𝑆𝑁𝑅 method noise, and strong
follows noise; its misjudgment
the advantages of STP/LTP method rate andand missing rate are low.
can accurately The P
obtain the
SNR
method follows the advantages
𝑃𝑆𝑁𝑅 value in strong noise environment. of STP/LTP method and can accurately obtain the P SNR
valueForinthe strong
M < noise environment.
5.0 small earthquakes, the energy information carried by the P waves is
not very the Mand
Forstrong, < 5.0thesmall earthquakes,
information cannot thebeenergy information
monitored carriedfar
by the stations by away.
the P waves
Figureis
5 shows the relationship between the 𝑃𝑆𝑁𝑅 value and the hypocentral distance in aFigure
not very strong, and the information cannot be monitored by the stations far away. M4.0 5
shows the relationship
earthquake. It can be seen between
that when PSNRhypocentral
the the value and the hypocentral
distance reachesdistance
50 km, the in a𝑃M4.0
𝑆𝑁𝑅
earthquake. It can be seen that when the hypocentral
value has dropped to below 4, which makes it vulnerable to background noise. In distance reaches 50 km, the PSNR
addi-
value
tion, has dropped
Figure to below
5 also reflects the 4, which makes
attenuation it vulnerable
relationship to background
between 𝑃𝑆𝑁𝑅 value noise.
and In addition,
hypocen-
Figure 5 also reflects the attenuation relationship
tral distance. With the increase in hypocentral distance, the energy between P value and hypocentral
SNR of seismic waves de-
distance.
creases With thesoincrease
gradually, the 𝑃𝑆𝑁𝑅 in value
hypocentral
will also distance,
decrease.the energy of seismic waves decreases
gradually, so the PSNR value will also decrease.

10

6
Psnr

0
0 20 40 60
Hypocentral distance(km)
Figure 5. The relationship between and hypocentral distance in the M4.0 earthquake.
Figure 5. The relationship between and hypocentral distance in the M4.0 earthquake.
3.2. Large Earthquake Data Analysis
The large earthquake data (M = 6, 7, 8) were provided by the Institute of Engineering
Mechanics, China Earthquake Administration and China Earthquake Networks Center,
National Earthquake Data Center (http://data.earthquake.cn, accessed on 31 October 2021).
The sampling rate is 200 Hz. The rupture duration of small earthquakes is short, and P
waves and S waves are usually separated in time. Large earthquakes rupture for a long
time and P waves are continuously emitted during the rupture, which results in P waves
and S waves sometimes partly overlapping. It means that it is difficult to detect P waves.
Sensors 2022, 22, 4534 9 of 17

However, the large earthquake rupture growth is large enough, although the initial S waves
make some interference, and P waves can still be clearly detected [23].
Figure 6 shows the records of U−D direction of the M7.0 earthquake and the corre-
sponding results of STP/LTP, STP, LTP calculation. It can be seen that the waveform of
the large earthquake lasts for a long time. In Figure 6c,d, STP/LTP value will be at a high
level within a period of time after the arrival of the P waves; PSNR value is also determined
in a very short time. In Figure 6e, the relative power variation in short-time window and
long-time window of large earthquake is the same as that of small earthquake. Different
from small earthquakes, the relative power value of the seismic signal in short-time window
and long-time window of large earthquakes is higher than that of small earthquakes,
Sensors 2022, 22, x FOR PEER REVIEW and
10 of 19
the duration is longer than that of small earthquakes, which also reflects the characteristics
of large earthquakes.

(a) (b)

(c) (d)

(e) (f)
Figure 6.
Figure 6. (a)
(a) The
The record
record of
ofthe
theUU−D direction in
−D direction in the
the M7.0
M7.0 earthquake.
earthquake. (b)
(b) The
The local
local enlargement
enlargement
result of (a). (c) The STP/LTP result of the record. The length of the short-time window is 0.3 s. The
result of (a). (c) The STP/LTP result of the record. The length of the short-time window is 0.3 s. The
length of the long-time window is 3 s. (d) The local enlargement result of (c). (e) The STP result of
length of the long-time window is 3 s. (d) The local enlargement result of (c). (e) The STP result of the
the record. (f) The LTP result of the record.
record. (f) The LTP result of the record.
Compared with small earthquakes, the energy of large earthquakes is very strong,
and the duration of earthquakes is longer. According to the rule of small earthquakes, the
large earthquake response also conforms to the relationship of energy attenuation with
hypocentral distance. Figure 7 shows the relationship between the energy of the earth-
 Sensors 2022, 22, 4534 10 of 17

Compared with small earthquakes, the energy of large earthquakes is very strong, and
the duration of earthquakes is longer. According to the rule of small earthquakes, the large
earthquake response also conforms to the relationship of energy attenuation with hypocen-
tral distance. Figure 7 shows the relationship between the energy of the earthquake and
the hypocentral distance in the M7.0 earthquake. It can be seen that when the hypocentral
distance reaches 100 km, the PSNR value also drops to about 6, and when the distance is
larger, the PSNR value will also be submerged by noise. Different from small earthquakes,
Sensors 2022, 22, x FOR PEER REVIEW
for the same hypocentral distance, the PSNR of large earthquakes is significantly higher
than that of small earthquakes, and the attenuation of the PSNR in the large earthquakes
is slower, which indicates the characteristics of the large earthquake energy. In addition,
it should be noted that for the large earthquakes, the hypocentral distance of the stations
one
shouldhand,
not beittoo
is small.
difficult tothe
When distinguish between
station is close P waves
to the seismic and
center, S waves,
on the which
one hand, it affe
detection ofdistinguish
is difficult to P waves. between
On the Pother
waveshand,
and S due to which
waves, the overlap of detection
affects the P wavesofandP S wav
𝑃waves. On the
𝑆𝑁𝑅 value isother hand, due
too large, to the
which is overlap of P waves
not within and Sof
the range 𝑃𝑆𝑁𝑅theattenuation
waves, PSNR value isrelation
too large, which is not within the range of PSNR attenuation relationship.

10

6
Psnr

0
0 50 100
Hypocentral distance(km)
Figure 7. The relationship between PSNR and hypocentral distance in the M7.0 earthquake.
Figure 7. The relationship between 𝑃𝑆𝑁𝑅 and hypocentral distance in the M7.0 earthquake.
3.3. Comprehensive Analysis
For the earthquakes with different magnitude ranges, Figure 8 shows the distribution
of PSNR value on hypocentral distance R. The earthquakes ranging from M3.0 to M4.0, M4.5
to M6.0, and large earthquakes ranging from M7.0 to M8.0, are selected for comparative
analysis. It can be seen that for the same hypocentral distance, the PSNR value of large
earthquakes is significantly higher than that of small earthquakes. With the increase in
the magnitude, the energy increases, and the PSNR value will also increase. In the same
magnitude range, the PSNR value has also a certain attenuation trend with the hypocentral
distance. In each magnitude range, the PSNR values are distributed according to a certain
linear relationship. There are also some discrete points in the diagram, and even some
discrete points are located in other magnitude ranges, which is realistic and indicates the
complexity of the process of the seismic energy release.
 in the magnitude, the energy increases, and the 𝑃𝑆𝑁𝑅 value will also increase. In the same
magnitude range, the 𝑃𝑆𝑁𝑅 value has also a certain attenuation trend with the hypocentral
distance. In each magnitude range, the 𝑃𝑆𝑁𝑅 values are distributed according to a certain
linear relationship. There are also some discrete points in the diagram, and even some
Sensors 2022, 22, 4534 discrete points are located in other magnitude ranges, which is realistic and indicates11the
of 17
complexity of the process of the seismic energy release.

Attenuationrelationship
Figure8.8.Attenuation
Figure relationshipbetween PSNRand
between𝑃𝑆𝑁𝑅 andhypocentral
hypocentraldistance
distanceR.R.The
Thegreen
greenpoints
pointsinin
thefigure
the figureare the𝑃P
arethe SNR
𝑆𝑁𝑅 values
values of
of magnitude
magnitude M3.0
M3.0 toto M4.0;
M4.0; the
the blue
blue points
points are
are the
the 𝑃 PSNR
𝑆𝑁𝑅 values
values ofof
magnitude
magnitudeM4.5M4.5totoM6.0;
M6.0;the
theorange
orangepoints
pointsare the𝑃P
arethe 𝑆𝑁𝑅
SNR values
valuesof magnitude
of magnitude M7.0
M7.0 totoM8.0.
M8.0.

AAlarge
largenumber
numberofofstudies
studies have
have shown
shown that
that when
when thethe earthquake
earthquake energy
energy reaches
reaches a
a certain
certain degree,
degree, the magnitude
the magnitude saturation
saturation phenomenon
phenomenon will occur.
will occur. By comparing
By comparing the
the M7.0
M7.0 earthquake
earthquake and M8.0andearthquake
M8.0 earthquakein Figurein Figure
8, it is 8, it is found
found that 𝑃that PSNRnohas
𝑆𝑁𝑅 has
no stratifica-
stratification,
tion, which also indicates the magnitude saturation
which also indicates the magnitude saturation of 𝑃𝑆𝑁𝑅 . ButSNR of P . But the specific
the specific saturation saturation
magni-
magnitude needs the further study
tude needs the further study to determine. to determine.
InInorder
ordertotostudy
studythe thestability
stabilityand
andaccuracy
accuracyofofthe the𝑃PSNRmethod,method,the the𝑃 Pdmethod
methodisis
𝑆𝑁𝑅 𝑑
chosento
chosen to compare
compare withwith it.it.The
Thefitting
fittingmodels
models of of Pd method
thethe 𝑃𝑑 method and Pand
SNR 𝑃 method are shown
𝑆𝑁𝑅 method are
in Equations (10) and (11). Among them, the arrival time of
shown in Equations (10) and (11). Among them, the arrival time of P waves is extractedP waves is extracted by the
manual supervision method, the short window length is set as
by the manual supervision method, the short window length is set as 0.3 s, and the long0.3 s, and the long window
length is
window set asis3set
length s. In
as 3the Pd the
s. In method, the high
𝑃𝑑 method, thepass
highfilter
passfor integrating
filter displacement
for integrating displace-is
set as the second-order high pass Butterworth filter (the low-frequency
ment is set as the second-order high pass Butterworth filter (the low-frequency cut-off frequency
cut-off
is 0.025 Hz), since the
frequency is 0.025 Hz), since P d value determined by the first 3 s P waves
the 𝑃𝑑 value determined by the first 3 s P waves cannot cannot be usedbefor
M ≥ 6.7 earthquakes [13], and considering the accuracy of the algorithm, the time window
used for M ≥ 6.7 earthquakes [13], and considering the accuracy of the algorithm, the
for calculating Pd value is set as 4 s.
time window for calculating 𝑃𝑑 value is set as 4 s.
𝑀=
M 𝐴×
=A log( Pd𝑑)) +
× 𝑙𝑜𝑔(𝑃 +B𝐵×
× log ( R) +
𝑙𝑜𝑔(𝑅) +C𝐶 (10)
(10)

M𝑀==A𝐴× × log ( PSNR)) +

𝑙𝑜𝑔(𝑃 +𝐵 × 𝑙𝑜𝑔(𝑅)
B× log( R) +
+ 𝐶C (11)
(11)
𝑆𝑁𝑅
whereA,A,B,B,and
where andCCare
arethe
thecoefficients
coefficientstotobebefitted.
fitted.
The study of Zollo shows that the Pd method also has a saturation phenomenon [24].
When the time window length is 4 s, the saturation magnitude can reach M7.0. Therefore,
the M ≤ 7.0 earthquakes in the samples are compared and analyzed by Pd method and
PSNR method in this study, as shown in Table 1. The fitting equation by PSNR method is
M = −4.6912 + 4.2519logPSNR + 3.8137logR, and the fitting correlation coefficient reaches
0.8012, which conforms to the linear relationship. The fitting equation by Pd method is
M = −1.2729 + 1.3405logPd + 5.4202logR, and the fitting coefficient is 0.7552, which con-
forms to the linear relationship. As can be seen from Table 1, the R2 statistic of PSNR is
slightly larger than that of Pd , and the mean residual of the two methods are also roughly
the same, which represents the high accuracy and stability of this method. Figure 9 shows
the residual diagram of the two methods. From the discreteness of the residuals, the
residuals of the two methods are distributed around 0, with only a few outliers. It can be
seen that the fitting results of the two methods are good.
 𝑑
to the linear relationship. As can be seen from Table 1, the 𝑅2 statistic of 𝑃𝑆𝑁𝑅 is slightly
larger than that of 𝑃𝑑 , and the mean residual of the two methods are also roughly the
same, which represents the high accuracy and stability of this method. Figure 9 shows the
residual diagram of the two methods. From the discreteness of the residuals, the residuals
Sensors 2022, 22, 4534 12 of 17
of the two methods are distributed around 0, with only a few outliers. It can be seen that
the fitting results of the two methods are good.

Table Comparisons of P𝑃d𝑑 and

Table1.1.Comparisons 𝑃𝑆𝑁𝑅methods
andPSNR methodsin in
all all samples.
samples.

P𝑷d𝒅 𝑷𝑺𝑵𝑹
PSNR
𝑅R22 0.7552
0.7552 0.8012
0.8012
Variance of
Variance of error
error 0.4074
0.4074 0.3308
0.3308

(a) (b)
Figure
Figure9.9.(a,b)
(a,b) are
are residual graphsofofPd𝑃method
residualgraphs 𝑑 method PSNR𝑃𝑆𝑁𝑅
andand method,
method, respectively.
respectively. Thepoints
The blue blue points
are
are the residual points in the normal range, and the red points are the abnormal
the residual points in the normal range, and the red points are the abnormal points. points.

After
Afterthe
thefitting
fitting model is obtained,
obtained,thetheaccuracy
accuracyofofthe themodel
model is is tested
tested byby detecting
detecting
severalgroups
several groups ofof seismic
seismic events.
events. Table
Table22shows
showsthethetest
testresults.
results. NNis is
the number
the number of of
stations
stations
involvedin
involved ineach
each event.
event. 𝑀M𝑎a is
is the
theactual magnitude.M𝑀
actualmagnitude. e 𝑒is is
thethe
estimated
estimated magnitude.
magnitude. In In
each group of earthquake events,
each group of earthquake events, the 𝑃SNR the P value and hypocentral distance obtained by
𝑆𝑁𝑅 value and hypocentral distance obtained by
effectivestations
effective stations are
are brought
brought into
into the
themodel,
model,and
andthe
themagnitude
magnitude estimation
estimation results of each
results of each
station are obtained. Finally, the average value is the magnitude estimation result of the
station are obtained. Finally, the average value is the magnitude estimation result of the
earthquake event, which compensates for possible site effection. Each seismic event can
earthquake event, which compensates for possible site effection. Each seismic event can
roughly estimate the magnitude of the earthquake through only several stations, and these
roughly estimate the magnitude of the earthquake through only several stations, and
effective stations are close to the source, which also makes the magnitude quickly obtained
these effective
through stationswhen
close stations are close to the source,
the earthquake which also makes the magnitude quickly
occurs.
obtained through close stations when the earthquake occurs.
Table 2. Test results of the PSNR method.
Table 2. Test results of the 𝑃𝑆𝑁𝑅 method.
Event, Date N Ma Me
Event, Date N 𝑴𝒂 𝑴𝒆
Gongxian, 13 March 2021 2 3.0 3.4
Gongxian,
Gongxian,13 March 2021
7 November 2020 3 2 3.4 3.0 3.7 3.4
Gongxian, 7 November
Gongxian, 2020
23 June 2020 2 3 3.8 3.4 3.2 3.7
Gongxian,
Gongxian, 2312 June2020
June 2020 2 2 4.0 3.8 3.7 3.2
Gongxian, 15 February 2021 2 4.0 4.3
Gongxian, 12 June 2020
Gongxian, 13 November 2020 2
2 4.1
4.0 4.2
3.7
Gongxian,
Xingwen, 16February
15 December2021
2018 8 2 5.7 4.0 5.7 4.3
Luxian, 16 September 2021 2 6.0 5.5

Figure 10 shows the variation of the estimated magnitude Me with the number of
the stations N in the xingwen M5.7 event. With the increase in the N, the Me estimated
by PSNR is more and more equal to the actual magnitude Ma . In the EEW with the dense
sensor network, it is possible to obtain the more accurate magnitude by extracting the
characteristic parameter PSNR from the multiple sensors for estimation and averaging.
 Figure 10 shows the variation of the estimated magnitude 𝑀𝑒 with the number of
the stations N in the xingwen M5.7 event. With the increase in the N, the Me estimated by
𝑃𝑆𝑁𝑅 is more and more equal to the actual magnitude 𝑀𝑎 . In the EEW with the dense
sensor network, it is possible to obtain the more accurate magnitude by extracting the
Sensors 2022, 22, 4534
characteristic parameter 𝑃𝑆𝑁𝑅 from the multiple sensors for estimation and averaging.13 of 17

5.85

5.8

5.75

Me
5.7

5.65

5.6

5.55
0 2 4 6 8 10
N

Figure 10. The variation of the Me with the number of the stations N in the Xingwen M5.7 event.
Figure 10. The variation of the 𝑀𝑒 with the number of the stations N in the Xingwen M5.7 event.
In terms of timeliness, the time of PSNR in all sample data are counted, as shown
inInFigure
terms11.of timeliness,
Each interval the time 𝑃𝑆𝑁𝑅 in in
of theofabscissa allthe
sample data are
histogram is counted, as shown
the time range afterinthe
Figure 11. Each interval of the abscissa in the histogram is the time
arrival of P waves. Most of the time corresponding to the PSNR ranges from 0–0.3 s, which range after the arrival
of Piswaves.
because Most
the of the time
length corresponding
of the to the 𝑃𝑆𝑁𝑅is ranges
short-time windows 0.3 s, andfromthere
0–0.3iss,nowhich is be- P
additional
cause the length of the short-time windows is 0.3 s, and there
waves’ power in the long-time windows. In this time range, the PSNR is most likely tois no additional P waves’
powerappear,in the long-time
which indicates windows.
the strong In timeliness
this time range,
of the Pthe 𝑃𝑆𝑁𝑅 is most likely to range
appear,
SNR method. A few times from
which indicates the strong timeliness
0.3–2 s. It can be determined that the occurrence of the 𝑃 𝑆𝑁𝑅 method. A few times range from
time of PSNR value is within 2 s after 0.3–2the
s. Itarrival
can be ofdetermined
P waves, but thatthe
the concrete
occurrence timetime of 𝑃𝑆𝑁𝑅
is still value is within
not certain. 2 s after
The process of the arrival
fault rupture
of Piswaves, but the concrete
very complex. The energy time ofisthe
still not earthquakes
small certain. The process
is small,ofand fault
therupture
duration is is
very
short.
complex.
The energyThe energy of theearthquakes
of the large small earthquakes
is strong,is small,
and theand the duration
duration is short.
of seismic Theisen-
waves long.
ergy of the large earthquakes is strong, and the duration of seismic
Even for the same earthquake, due to the inherent noise of the instrument and the local waves is long. Even
for geological
the same earthquake,
conditions, due to the inherenttime
the determination noise ofof the instrument
PSNR value is slightlyand the local geolog-
different. The PSNR
Sensors 2022, 22, x FOR PEER REVIEW 15 of 19
icalmethod
conditions, the determination
has better timeliness, which timecan 𝑃𝑆𝑁𝑅 value
ofgreatly is slightly
save costs and greatlydifferent.
reduce the 𝑃time
The 𝑆𝑁𝑅 of
method
magnitudehas better timeliness, which can greatly save costs and greatly reduce the time of
estimation.
magnitude estimation.

Figure11.
Figure Time-consumingdistribution
11.Time-consuming histogramofof 𝑃PSNR. .
distributionhistogram
𝑆𝑁𝑅

Figure 12 shows two examples of 𝑃𝑆𝑁𝑅 values in 0–0.3 s and 1.5–1.8 s after P waves’
arrival, the arrival time of P waves, 𝑃𝑆𝑁𝑅 and 𝑃𝑑 are also shown in the figure. In the first
example (a, b, c, d), after the arrival of P waves, the STP/LTP value immediately increased
Sensors 2022, 22, 4534 14 of 17

Figure 12 shows two examples of PSNR values in 0–0.3 s and 1.5–1.8 s after P waves’
arrival, the arrival time of P waves, PSNR and Pd are also shown in the figure. In the first
example (a, b, c, d), after the arrival of P waves, the STP/LTP value immediately increased
to the maximum, and the PSNR value quickly appeared. In the second example (e–h), the
STP/LTP value increases after the arrival of P waves, but it increases to the maximum
after a period of time. Although the occurrence time of PSNR is different between the two
examples, it is much earlier than the Pd method. With the comprehensive consideration
of timeliness and accuracy, the PSNR method has significant advantages. The magnitude
estimation can be carried out synchronously when P waves arrives, which greatly improves
Sensors 2022, 22, x FOR PEER REVIEW 16 of 19
the timeliness of earthquake early warning system. However, PSNR requires high accuracy
for P waves’ arrival extraction. If P waves’ seismic phase identification is wrong, this
method will also result in a false alarm.

(a) (b)

(c) (d)

(e) (f)

Figure 12. Cont.

Sensors 2022, 22, 4534 15 of 17

(e) (f)

(g) (h)
Figure 12. (a–d)
Figure 12. (a–d) are
arethe
theresults
resultsofofthe
the Timeliness
Timeliness comparison
comparison 𝑃𝑆𝑁𝑅
of PofSNR Pd in𝑃𝑑the
andand inM4.0
the M4.0 earth-
earthquake.
quake. (e–h) are the results of the Timeliness comparison of 𝑃
(e–h) are the results of the Timeliness comparison of PSNR and Pd𝑆𝑁𝑅 and 𝑃 in the M8.0 earthquake.
𝑑 earthquake. (b,d,f,h) are
in the M8.0
(b,d,f,h) are local enlargement results of figures (a,c,e,g), respectively. The green lines in all figures
local enlargement results of figures (a,c,e,g), respectively. The green lines in all figures represent the
represent the arrival time of the P wave. The red lines in (a,b,e,f) represent the time of the 𝑃𝑆𝑁𝑅 and
arrival time of the P wave. The red lines in (a,b,e,f) represent the time of the PSNR and Pd , respectively.
𝑃𝑑 , respectively. The blue lines in (c,d,g,h) represent the time of the 𝑃𝑆𝑁𝑅 and 𝑃𝑑 , respectively.
The blue lines in (c,d,g,h) represent the time of the PSNR and Pd , respectively.

After analyzing the accuracy and timeliness of PSNR algorithm, it can be seen that
PSNR can synchronize the extraction of P waves’ arrival time for magnitude estimation in
the initial short time of P waves. PSNR is suitable for the construction of earthquake early
warning system in terms of timeliness, accuracy, and robustness.

4. Discussion
The fast magnitude estimation method (PSNR ) based on SNR feature proposed in this
paper is a new method in the field of magnitude estimation. The significance of this study is
that the PSNR method can quickly estimate the magnitude in a short time; in particular, the
magnitude estimation time is controlled within two seconds after the arrival of the P waves,
which greatly meets the timeliness of the earthquake early warning system. Compared
with the existing magnitude estimation methods, the new method greatly reduces the time
used for magnitude estimation and improves the efficiency of earthquake early warning,
which is the main highlight of the PSNR method.
The PSNR method extends the advantages of the STP/LTP method and extends from
the time-of-arrival extraction of P waves to the magnitude estimation, so that the magnitude
estimation can be synchronized with the extraction of P waves, which is the cleverness of
the PSNR method.
With the development of sensor technology and computer technology, the network
of seismic monitoring stations will be more and more intensive. Under such technical
background, the efficiency of earthquake early warning will be significantly improved.
Fortunately, the calculation cost of PSNR method is low. After the P waves are extracted by
STP/LTP method, it is only necessary to find the PSNR value, without excessive calculation,
which can effectively reduce the pressure of the single chip microcomputer at the equipment
end. Based on the mode of cloud-edge collaboration, in the earthquake early warning
system, the edge equipment can directly record the PSNR value and transmit it to the
cloud server, which also greatly reduces the transmission pressure of the network and the
calculation pressure of the cloud server.
The PSNR method is more robust than the existing methods, which can be applied to
weak noise, general noise environment, and is not easily affected by instrument background
noise. The PSNR parameter performance is very stable in a weak noise environment.
Moreover, the PSNR method has the advantage of dimensionless, which can be widely used
in different types of station instruments and improving the efficiency of data processing.
PSNR can be obtained directly from the measuring units of the instruments used for different
specifications. In the station data of this study, the instrument specifications are not
Sensors 2022, 22, 4534 16 of 17

completely uniform, the use of the PSNR method greatly saves the cost of data processing
and avoid the calculation error caused by different units.
It should be noted that the geological structure and crustal movement are different
in different regions. All earthquake events in this paper are shallow earthquakes. When
the PSNR parameter is used to estimate the magnitude, local historical records should be
used for model training so as to improve the accuracy of the model and reduce the false
alarm rate.
In addition, the factors affecting the PSNR value cannot be ignored. The premise of
obtaining PSNR is to detect the P waves accurately. Many research methods have achieved
good results in P waves’ detection, such as the STP/LTP method [21], the AIC method [25],
and the deep learning method [26]. The length of the long-time window and short-time
window in sliding window also has a certain influence on PSNR value. If the length of the
short-time window is too short, there will be noise interference. If the length of the short
time window is too long, the seismic effective energy cannot be expressed in time. Finally,
the more stable the sensor performance for seismic monitoring is, the more accurate and
stable the PSNR value and the accuracy of the fitting model are.

5. Conclusions
Aiming at the timeliness and accuracy of a magnitude estimation in an earthquake
early warning system through analyzing the energy characteristics of the STP/LTP method
and combining the principle of SNR and the definition of relative power, a P waves arrival
fast magnitude estimation method (PSNR ) based on SNR is proposed in this paper. The
main conclusions are as follows:
(1) Both ghd PSNR method and Pd method have high stability and accuracy, and the PSNR
method has strong robustness.
(2) The highlight is that the PSNR method has good timeliness. The time corresponding
to PSNR value is usually within 2 s from the first arrival of P waves, and sometimes it
can even be synchronized to the first arrival of P waves, which can greatly save time
in the earthquake early warning magnitude estimation. PSNR is a good choice that
considers both timeliness and accuracy.
(3) The STP/LTP method based on SNR and relative power contains certain energy
information when seismic P waves arrive. There is a certain positive correlation
between PSNR value and magnitude, and there is an obvious attenuation relationship
between PSNR value and hypocentral distance R.
(4) Owing to the dimensionless characteristics, the PSNR method can be widely used in
different sensors and reduce costs in the processing of seismic records.
In all, PSNR method can meet the needs of the EEW in terms of timeliness and accuracy.

Author Contributions: Conceptualization, D.Z., J.F. and Z.L.; Methodology, D.Z., J.F., Z.L. and J.W.;
software, D.Z.; validation, L.W. and J.L.; formal analysis, L.W.; data curation, J.L.; writing—original
draft, D.Z.; writing—review and editing, J.F., Z.L., J.W. All authors have read and agreed to the
published version of the manuscript.
Funding: This research was supported by the National Key Research and Development Program
of China (No. 2019YFC1509205), the National Nature Science Foundation of China (No. 41874019),
and the research grant from the National Institute of Natural Hazards, Ministry of Emergency
Management of China (No. ZDJ2019-12).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data for this study were provided by the National Key Research
and Development Program of China (No. 2019YFC1509205), Institute of Engineering Mechanics,
China Earthquake Administration and China Earthquake Networks Center, National Earthquake
Data Center (http://data.earthquake.cn, accessed on 31 October 2021).
Conflicts of Interest: The authors declare no conflict of interest.
Sensors 2022, 22, 4534 17 of 17

References
1. Nakamura, Y. On the Urgent Earthquake Detection and Alarm System (UrEDAS). Proc. Ninth World Conf. Earthq. Eng. 1988, 2,
673–678.
2. Shigeki, H.; Hiroaki, N.; Kana, A.; Aya, K.; Yukio, F. An Automatic Processing System for Broadcasting Earthquake Alarms. Bull.
Seismol. Soc. Am. 2005, 95, 708–718.
3. Allen, R.M.; Kanamori, H. The Potential for Earthquake Early Warning in Southern California. Science 2003, 300, 786–789.
[CrossRef]
4. Espinosa-Aranda, J.M.; Rodríquez, F.H. 76—The Seismic Alert System of Mexico City. Int. Geophys. 2003, 81, 1253–1259.
5. Wu, Y.M.; Chung, J.K.; Shin, T.C.; Hsiao, N.C.; Tsai, Y.B.; Lee, H.K.; Teng, T.L. Development of an Integrated Earthquake Early
Warning System in Taiwan-Case for the Hualien Area Earthquakes. Terr. Atmos. Ocean. Sci. 1999, 10, 719–736. [CrossRef]
6. Erdik, M.; Fahjan, Y.; Ozel, O.; Alcik, H.; Mert, A.; Gul, M. Istanbul Earthquake Early Warning and Rapid Response System. Bull.
Earthq. Eng. 2003, 1, 157–163. [CrossRef]
7. Fu, J.; Li, Z.; Meng, H.; Wang, J.; Shan, X. Performance Evaluation of Low-Cost Seismic Sensors for Dense Earthquake Early
Warning: 2018–2019 Field Testing in Southwest China. Sensors 2019, 19, 1999. [CrossRef]
8. Kanamori, H. Real-Time seismology and earthquake damage mitigation. Annu. Rev. Earth Planet. Sci. 2005, 33, 195–214.
[CrossRef]
9. Wu, Y.M.; Kanamori, H.; Allen, R.M.; Hauksson, E. Determination of earthquake early warning parameters, τ c and Pd , for
southern California. Geophys. J. Int. 2007, 170, 711–717. [CrossRef]
10. Yamada, T.; Ide, S. Limitation of the Predominant-Period Estimator for Earthquake Early Warning and the Initial Rupture of
Earthquakes. Bull. Seismol. Soc. Am. 2008, 98, 2739–2745. [CrossRef]
11. Wu, Y.M.; Shin, T.C.; Tsai, Y.B. Quick and Reliable Determination of Magnitude for Seismic Early Warning. Bull. Seismol. Soc. Am.
1998, 88, 1254–1259. [CrossRef]
12. Wu, Y.M.; Kanamori, H. Experiment on an Onsite Early Warning Method for the Taiwan Early Warning System. Bull. Seismol. Soc.
Am. 2005, 95, 347–353. [CrossRef]
13. Wu, Y.M.; Zhao, L. Magnitude estimation using the first three seconds P-wave amplitude in earthquake early warning. Geophys.
Res. Lett. 2006, 33, 271–284. [CrossRef]
14. Lancieri, M.; Zollo, A. A Bayesian approach to the real-time estimation of magnitude from the early P and S wave displacement
peaks. J. Geophys. Res. Solid Earth 2008, 113, B12302. [CrossRef]
15. Li, H.; Zhang, J.; Chen, Y.; Yu, K.; Han, J.; Fan, H.; Hu, R. Research on earthquake early warning parameters and rapid magnitude
estimation based events in Yunnan region in 2014. World Inf. Earthq. Eng. 2018, 34, 123–131.
16. Wang, Y.; Li, S.; Song, J. Magnitude-scaling relationships based on initial P-wave information in the Xinjiang region, China. J.
Seismol. 2021, 25, 697–710. [CrossRef]
17. Alcik, H.; Ozel, O.; Apaydin, N.; Erdik, M. A study on warning algorithms for Istanbul earthquake early warning system. Geophys.
Res. Lett. 2009, 36, L00B05. [CrossRef]
18. Zhu, J.; Li, S.; Song, J.; Wang, Y. Magnitude Estimation for Earthquake Early Warning Using a Deep Convolutional Neural
Network. Front. Earth Sci. 2021, 9, 653226. [CrossRef]
19. Mousavi, M.S.; Beroza, G.C. A Machine-Learning Approach for Earthquake Magnitude Estimation. Geophys. Res. Lett. 2019,
41, e2019GL085976. [CrossRef]
20. Hu, A.D.; Zhang, H.M. Application of machine learning to magnitude estimation in earthquake emergency prediction system.
Chin. J. Geophys. 2020, 63, 2617–2626.
21. Fu, J.; Wang, X.; Li, Z.; Meng, H.; Wang, J.; Wang, W.; Tang, C. Automatic Phase-Picking Method for Detecting Earthquakes Based
on the Signal-to-Noise-Ratio Concept. Seismol. Res. Lett. 2019, 91, 334–342. [CrossRef]
22. Ellsworth, W.L.; Beroza, G.C. Seismic evidence for an earthquake nucleation phase. Science 1995, 268, 851–855. [CrossRef]
[PubMed]
23. Kodera, Y. Real-Time Detection of Rupture Development: Earthquake Early Warning Using P Waves From Growing Ruptures.
Geophys. Res. Lett. 2018, 45, 156–165. [CrossRef]
24. Zollo, A.; Lancieri, M.; Nielsen, S. Earthquake magnitude estimation from peak amplitudes of very early seismic signals on strong
motion records. Geophys. Res. Lett. 2006, 33, L23312. [CrossRef]
25. Akaike, H. Markovian representation of stochastic processes and its application to analysis of autoregressive moving average
processes. Ann. Inst. Stat. Math. 1974, 26, 363–387. [CrossRef]
26. Zhu, W.; Beroza, G.C. PhaseNet: A deep-neural-network-based seismic arrival-time picking method. Geophys. J. Int. 2018, 216,
261–273. [CrossRef]