Co-seismic ionospheric disturbances characteristics in different azimuths following the 2022 Mexico earthquake from GNSS observations

Abstract

Co-seismic ionospheric disturbances (CIDs) can help understand the coupled dynamics of earthquake-atmospheric coupling geophysical processes. On September 19, 2022, an earthquake occurred at UTC = 18:05:08 near the Pacific Coast of Mexico as the result of shallow thrust faulting with magnitude of 7.6 and depth of around 26.9 km. The epicenter of the earthquake is located at (18.455°N, 102.956°W). In this study, observation data from the global navigation satellite system (GNSS) are used to identify CIDs about 12 min after the earthquake occurred. The significant CIDs signals are observed with extending outward from the epicenter. The CID characteristics like amplitude, frequency, and waveform are also investigated and discussed. The waveform is a standard and inverse N-type, suggesting a connection to plate movement and geomagnetic field. In addition, the center frequency is within the range of acoustic wave frequency from 2 to 4 mHz. The propagation speed is approximately 0.81 km/s for PRN G18, 1.01 km/s for G23, and 1.16 km/s in the east and 1.44 km/s in the west for G32 between 18:00 and 19:00. For R21, the propagation speed is close to 1.06 km/s. It demonstrates that the main source of the CIDs is the acoustic wave. Also, it is discovered that the CID propagation velocity varies significantly depending on the azimuth. At the direction closed to the strike angle (287°), the maximum propagation speed is found. The rupture mainly occurred in the western region, and the rupture velocity is larger in the western region, which might cause the quicker CID in this direction.

Full text

1. Introduction
Earthquakes are common natural disasters (Jin & Park,2006). During the main shock of earthquakes, the
rupture and severe co-seismic vertical crust movements can excite acoustic resonance, and some of the
acoustic resonance can propagate upward into the ionosphere in the form of acoustic waves and induce
variations of the ionospheric total electron content (TEC), which is the so-called ionospheric disturbances
(Jin etal.,2011,2015). The ionospheric disturbances were related to the acoustic-gravity wave launched
by big earthquakes (Ronald,1966, 1967). The first ionosphere disturbance was detected by Leonard and
Barnes (1965) using the ionospheric vertical sounding following the great Alaska earthquake in 1964.
Davies and Baker(1965) found the frequency oscillations in radio signals following the same Alaska earth-
quake. Later more co-seismic ionospheric disturbances (CIDs) were reported with great attentions together
understanding the mechanism of earthquakes and crust vertical movement.
However, due to the limitation of the measurement instruments in the last few decades, it is difficult to
investigate the detailed characteristics of CID and coupling process between the ground motion and the
ionosphere (Okoh etal.,2018). Nowadays, dense Global Navigation Satellite System (GNSS) network ob-
servations can estimate the ionosphere and CID (Calais & Minster,1995; Jin & Su,2020; Jin etal.,2015)
with its strong imaging capability, high spatial resolution and sensibility for detecting Rayleigh wave in the
ionosphere (Ducic etal.,2003; Occhipinti etal.,2010). With the wide use of GNSS, the properties of CID
and the relationship among CID, earthquake and ionosphere will be better understood. By estimating the
ionospheric delays from Global Positioning System (GPS) (Jin etal.,2004), the TEC can be precisely calcu-
lated so that the seismic ionospheric anomaly related to the earthquake can be detected from the GPS-TEC
time series observation (Jin, Jin, & Kutoglu,2017; Jin etal.,2004). It will provide a chance to investigate the
complete process and the properties of the earthquake, after estimating the CID signal. Moreover, as the
Abstract Co-seismic ionospheric disturbance (CID) may provide insights on understanding the
coupled nature of earthquake–atmosphere geophysical processes. In this study, the CIDs following the
Mw 7.7 Jamaica earthquake on January 28, 2020 are detected about 12min after the main shock by the
dual-frequency Global Positioning System measurements. Significant CIDs at two azimuths are observed
from satellite PRN03, 04 and 26 with spreading out at 3.54, 3.51 and 3.48km/s respectively, which are
close to the spreading speeds of Rayleigh waves recorded by the seismographs. The significant CID
signals are found in south near-field area and southwest far-field area. Furthermore, CID characteristics
are analyzed in terms of amplitude, elevation and azimuth angle, waveform, and frequency domain.
Results show that CIDs are observed by PRN03, 04 and 26 at low elevation angles (<35°) in infrasonic
wave frequency domain and the average negative amplitudes of CIDs observed by PRN26 are larger than
−0.08 TECU, while the CID amplitudes observed by PRN03 and PRN04 are about −0.05 and −0.07 TECU,
respectively. Moreover, the azimuthal asymmetry of CID amplitude in SW and SE azimuths and different
initial polarities in disturbance signals are found and discussed from tectonic and nontectonic factors.
The relations among CID, Rayleigh wave and focal mechanism are interpreted. The upward propagating
secondary acoustic wave triggered by the seismic Rayleigh wave from earthquake is the main source
of CIDs. These results confirm that strike-slip earthquake can also generate pronounced co-seismic
ionospheric disturbances.
CHAI AND JIN
© 2021. American Geophysical Union.
All Rights Reserved.
Two-Azimuth Co-Seismic Ionospheric Disturbances
Following the 2020 Jamaica Earthquake From GPS
Observations
Yi Chai1,2 and Shuanggen Jin1
1Shanghai Astronomical Observatory, Chinese Academy of Sciences, Shanghai, China, 2University of Chinese Academy
of Sciences, Beijing, China
Key Points:
• Two-azimuth co-seismic ionospheric
disturbances (CIDs) are found
from Global Positioning System
observations following the 2020
Jamaica earthquake
• The CIDs are mainly triggered by
the upward propagating secondary
acoustic wave by the seismic
Rayleigh wave
• The co-seismic ionospheric
disturbances display different
polarity and amplitude
characteristics due to the fault
system
Correspondence to:
S. Jin,
sgjin@shao.ac.cn;
sg.jin@yahoo.com
Citation:
Chai, Y., & Jin, S. (2021). Two-azimuth
co-seismic ionospheric disturbances
following the 2020 Jamaica earthquake
from GPS observations. Journal of
Geophysical Research: Space Physics,
126, e2020JA028995. https://doi.
org/10.1029/2020JA028995
Received 2 DEC 2020
Accepted 25 AUG 2021
10.1029/2020JA028995
RESEARCH ARTICLE
1 of 13

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
2 of 13
short time for CID wave to the ionosphere (around 8min), CID will have potential in realizing near-real-
time earthquake monitoring and early tsunami warning (Astafyeva,2019; Komjathy etal.,2016; Rolland
etal.,2010).
Nowadays, a number of studies have been performed to study the ionospheric disturbance induced by
great earthquakes and provided a clear understanding of CIDs characteristics. For example, Afraimovich
etal. (2010) found the intense N-shaped shock-acoustic waves that propagated to the ionosphere and in-
duced the disturbance with a plane waveform following the 2008 Wenchuan earthquake. The CIDs in the
far-field following the 2008 Wenchuan earthquake were perpendicular to the direction of the fault rupture
(NE-SW) (Zhao & Hao,2015). The far-field CID following the 2015 Mw 7.8 Nepal earthquake was detected
by GPS-TEC observation up to 3,000km from the epicenter (H. Liu etal.,2021) as the acoustic gravity wave
induced by Rayleigh surface wave propagated into the ionosphere and caused the electronic and plasma
density oscillation (Sripathi etal.,2020). Also, after the 2011 Mw 9.0 Tohoku Earthquake in Japan, significant
ionospheric disturbance was investigated from nationwide GPS receiving networks and the disturbance was
confirmed with three different propagation velocities, which attributed to three different generation sources
(J. Y. Liu etal.,2011; Saito etal.,2011; Tsugawa etal.,2011). The phenomena of CIDs with different speeds
have been found for other earthquakes. For example, Astafyeva etal.(2009) found two-mode long-distance
CIDs following the great 1994 Kurile earthquake and the two-mode ionospheric disturbances with two dif-
ferent propagation velocities were detected and estimated following the 2005 Northern California offshore
earthquake (Jin,2018).
Whereas, there are still some problems and difficulties in investigating CID. For instance, the distinct TEC
anomaly can be detected by GPS measurement only for some earthquakes with large magnitudes (Mw>6.5)
(Perevalova etal.,2014), as the large vertical crustal displacement and magnitude have important influ-
ences on the amplitude of CID (Astafyeva et al., 2013). Uneven distribution of ground-based GPS net-
work makes the absence of ionospheric disturbance in some seismic regions. Besides, the directivity and
coupling mechanism between CIDs and earthquakes are still not clear. Because the coupling process of
earthquake-atmosphere-ionosphere and the characteristics of CID, such as amplitude, propagation speed
and directivity (Heki & Ping,2005), are controlled by earthquake parameters such as magnitude and focal
mechanism (Astafyeva etal.,2009; Heki etal.,2006), the pattern of rupture and ground deformation (Afrai-
movich etal.,2010), and nontectonic forcing mechanisms in terms of geomagnetic field (Heki & Ping,2005;
Rolland etal.,2011,2013), geometry of GPS line-of-sight signal (Afraimovich etal.,1998,2001; Astafyeva
etal., 2014) and ambient electron density gradient (Bagiya etal.,2019). Consequently, abundant investi-
gations of CIDs following different earthquake events are necessary to comprehend the seismic motion
process and coupling mechanism of earthquake-atmosphere-ionosphere.
In this paper, the CIDs are estimated and investigated following the Mw=7.7 Jamacia earthquake on Jan-
uary 28, 2020 from the dense GPS network observations. In Section2, data and methods are introduced,
results and discussion are presented in Section3, and finally conclusion is given in Section4.
2. Data and Method
2.1. 2020 Mw 7.7 Earthquake
The 2020 Mw 7.7 earthquake occurred in the Caribbean Sea to the northwest of Jamaica, with 10km in
depth at 19:10:22 (UTC), January 28, 2020, as the result of the strike-slip faulting on the plate boundary be-
tween the North America and Caribbean Plates. The epicenter (19.46°N, 78.79°W) was located at the plate
boundary and the fault plane struck along with the orientation of the plate boundary. The GPS observation
data with a sampling rate of 15s was obtained from dense University Navstar Consortium (UNAVCO) GPS
stations.
The distributions of 112 GPS stations and 13 seismographs are shown in Figure1 with the blue triangles
and red dots, respectively. The data of seismometers was provided by Incorporated Research Institutions
for Seismology (IRIS, http://ds.iris.edu/wilber3/find_stations/11176800). The red pentagram represents the
epicenter of the 2020 Mw 7.7 earthquake and the black line represents the fault plane near the epicenter. The
beach ball indicates the focal mechanism of the earthquake event at the upper-right corner of the figure.
Magnetic field (MF) parameters in height=350km involving inclination (I) and declination (D) are shown

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
3 of 13
in the white panel at lower-left quarter. The schematic view of magnetic
field plotted in arrows is shown in the white panel at the upper-left cor-
ner of Figure1a. The slip distribution map of the 2020 Mw 7.7 Jamaica
earthquake is shown in Figure1b. Related information (finite fault and
slip distribution) of this earthquake event is accessible from U.S. Geo-
logical survey (USGS). The slip distribution map indicates the location
and motion direction of fault plane in strike of 258° with arrows and the
horizontal co-seismic displacement in color.
2.2. Method
The CID can be extracted from GPS-TEC time series. When GPS satel-
lite signals propagate into the ionosphere, the ionospheric delay is relat-
ed to the GPS signal frequency and ionospheric TEC. In order to obtain
the ionospheric disturbances, the ionospheric TEC should be calculated
precisely from the dual-frequency GPS observation (f1=1,575.42MHz,
f2 = 1,227.60 MHz) by the following equation (Brunini & Azpilicue-
ta,2009; Jin, Jin, & Kutoglu,2017, Jin, Jin, & Li,2017; Jin etal.,2004):
STEC
STEC



 





ff
ff
LL Nb N b L
1
2
2
2
1
2
2
21 2 11 1 2 2 2
4 0 3.
 



 


ff
ff
PP d d p
1
2
2
2
1
2
2
212 12
4 0 3.
()

(1)
where STEC is the slant total electron content, L1 and L2 are the GPS
carrier phase measurements in frequency f1 and f2, P1 and P2 are the GPS
code measurements in frequency f1 and f2, λ1 and λ2 are the GPS signal
wavelength in frequency f1 and f2, N is the ambiguity, b is the instrument
biases for carrier phase, d1 and d2 are the differential code biases, and ε
is the residual. STEC represents the absolute magnitude of ionospheric
TEC. In order to get the relative variation of the ionospheric TEC and
estimate the characteristics of CID, the STEC along the GPS line of sight
(LOS) is converted to vertical TEC (VTEC) by the following mapping
function:

sin
VTEC STEC cos arcsin
Rz
RH




 





(2)
where H is the height of the ionosphere shell, where H is assumed at 350km of altitude. R is the Earth's
radius, and z is the elevation angle of the satellite. The cycle slip is the main error in obtaining high-pre-
cision TEC values from above method (Nguyen etal.,2019). Therefore, the second-order time-difference
phase ionospheric residual was used to eliminate cycle slip (Cai etal.,2013). Here the Butterworth filter of
a fourth-order zero-phase finite impulse was used to remove the background noise and obtain the filtered
TEC series. According to the Nyquist sampling theory, the Nyquist frequency is larger than 8mHz for GPS
observation as the sampling interval of GPS observation data is 15s. The 2mHz is the cutoff frequency of
the acoustic above the ionospheric height. The GPS-TEC time series obtained from station LMNL and sat-
ellite PRN26 with different passband frequencies are shown in Figure2. The distinct CID can be obtained
from the series with the 2–5mHz passband frequency filtering at about 12min after the occurrence of the
earthquake in the red line marked zone, so the fourth-order zero-phase Butterworth filter with passband
frequency of 2 and 5mHz was applied to obtain the CID time series.
Figure 1. The seismic information of the Mw 7.7 Jamaica earthquake
and distribution of Global Positioning System (GPS) dense network
stations and seismographs around the epicenter area and geomagnetic
information about the earthquake (a). The blue triangles and red dots
denote the location of GPS stations and seismographs, respectively. The
focal mechanism is obtained from U.S. Geological survey (USGS, https://
earthquake.usgs.gov/earthquakes/eventpage/pt20028001/moment-tensor)
(b). The slip distribution map was obtained from USGS (https://
earthquake.usgs.gov/earthquakes/eventpage/us60007idc/executive).

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
4 of 13
3. Results and Discussion
3.1. Co-Seismic Ionospheric Disturbances
The CIDs following the 2020 Jamaica earthquake were detected and es-
timated. Figure3 shows the TEC distribution maps during 19:10–19:30
UTC. The red pentagram represents the epicenter and the dots in color
correspond to the location of subionospheric points (SIPs). The color val-
ue shows the TEC variation amplitude of SIPs and the color bar indicates
the variation range of filtered TEC series (dTEC, in TECU, from −0.1 to
0.1 TECU). The 2020 Jamaica earthquake occurred at 19:10 UTC. Obvi-
ous ionospheric anomalies spreading out from the epicenter are first de-
tected at the south near-field region of epicenter (200–450km away from
the epicenter, Figure3b), after about 12min of the main shock. Most of
the TEC disturbances display in positive anomalies. The average ampli-
tude reaches to 0.06 TECU (1 TECU=1016 e/m2). Around 2min later
at 19:24 UTC, the TEC disturbances become stronger and the amplitude
reaches about −0.07 TECU at this time. It should be noted that the TEC
disturbances turn positive anomalies to negative anomalies in the near
field (Figure3c). Besides, in the southwest far field (marked by red cir-
cle), there are slight positive disturbances. The negative TEC anomalies
last for around 2min and turn back to positive anomalies in the south
near-field area at 19:26 UTC (Figure3d) and the southwest far-field TEC
disturbances turn slight positive anomalies to obvious negative anom-
alies. The average amplitude of near-field TEC disturbances is about
0.06 TECU, while the far-field TEC disturbances is about −0.07 TECU.
After the same time interval as the previous discussed TEC disturbance
(2min), the TEC anomalies can be detected only in the southeast region
(Figure3e). After 19:30 UTC, no obvious TEC disturbances can be detect-
ed in the seismic region.
3.2. CID in Different Azimuths
The propagation characteristics of CIDs were further analyzed in the pat-
tern, modes, generation mechanism and source of CIDs. In Section3.1,
the CIDs following the Jamacia earthquake have been detected though
GPS-TEC observation. In this section, the generation source and charac-
teristics of CID were estimated and discussed. The most pronounced ionospheric disturbances were detect-
ed by GPS satellites PRN03, PRN04 and PRN26. Figure4 shows the traveling-time diagrams of CIDs and SIP
tracks during 19:10–19:30 UTC. The SIP tracks diagrams show that SIP tracks with CID signals mainly cover
the southeast, south and southwest area (Figure3). The traveling-time diagrams demonstrate the linear
relationship between the CID travel time and distance from SIP to the epicenter. The horizontal and vertical
axes correspond to the UTC time and the distance between SIP and epicenter, respectively. The color of
curves indicates the amplitude of filtered TEC series. The color bar denotes the variation range of filtered
TEC (dTEC, in TECU, from −0.08 to 0.08 TECU). Significant traveling CID can be found through color var-
iance in traveling-time diagrams 250–500, 500–700 and 300–800km away from the epicenter, respectively.
Moreover, the CID captured by PRN26 has the largest TEC negative amplitude, larger than −0.08 TECU,
while the negative amplitude of CID detected by PRN03 and PRN04 only reaches to −0.05 TECU and −0.07
TECU, respectively. This indicates that these CIDs may have different amplitude characteristics. After per-
forming the linear fit, the propagation velocities of the CIDs captured by PRN03, 04 and 26 are about 3.54,
3.51 and 3.48km/s, respectively. The ionospheric disturbance generated by different sources can be distin-
guished through the velocities of their propagation. These velocities are larger than the sound speed at the
ionospheric altitude (∼1km/s) and close to the Rayleigh surface wave propagation speed along the ground
surface at 3,000–4,000m/s (Astafyeva etal.,2014). According to Jin(2018), these ionospheric disturbances
are probably the secondary acoustic wave generated by seismic Rayleigh waves with dynamic coupling.
Figure 2. Global Positioning System-total electron content (TEC) time
series obtained from station LMNL and satellite PRN26 with different filter
passband frequencies. (a) Filtered TEC series with 1–15mHz passband.
(b) Filtered TEC series with 2–5mHz passband. (c) Filtered TEC series
with 5–8mHz passband (d) Filtered TEC series with 8–15mHz passband.
The dashed black line represents the onset time of the 2020 Jamaica
earthquake.

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
5 of 13
Figure5 shows maximum amplitudes of the TEC-CID time series obtained from PRN03, 04 and 26 during
19:10–19:30 UTC when the CIDs appeared. The horizontal and vertical axes correspond to the elevation
angle along the LOS between satellites and GPS stations and the ionospheric piercing point (IPP) epicentral
azimuth, respectively. The color in these dots indicates the value of the maximum amplitude. As is shown in
Figure5a, the dots with the maximum larger than 0.03 TECU are mainly at the elevation angles 20–30°. In
Figure5b, the dots are mainly at elevation angles 28–32°, while the dots gather in elevation angles 28–35° in
Figure5c. All these elevation angles are lower than 40°, which is favorable to detect the Rayleigh wave-in-
duced CID (Rolland etal., 2011). According to previous work (Afraimovich etal., 1998, 2001; Astafyeva
etal., 2014), lower elevation angle can enlarge the horizontal extent of the ionospheric disturbance. As
Rayleigh wave-induced CID waves propagate along nearly vertical direction into the ionosphere (Rolland
et al., 2011), so when the satellite-to-receiver LOS is perpendicular to the disturbance wave vector, the
Figure 3. Filtered total electron content (TEC) distribution maps extracted from Global Positioning System observation data during 19:10–19:30 UTC. The red
pentagram represents the epicenter and the color filled dots indicate the positions of subionospheric points. The color bar is the variation range of filtered TEC
(dTEC, in TECU).

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
6 of 13
Figure 4.

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
7 of 13
observed amplitude reaches its largest amount. On the other hand, the amplitude of the disturbance signal
is relevant to the satellite elevation angle (Heki etal.,2006).
Figure6 shows the azimuthal distribution of the average maximum amplitudes for all the filtered TEC time
series obtained from PRN03, 04 and 26 in every 15° azimuth bin during 19:10–19:30 UTC in the form of
polar diagram. The theta axis represents the epicenter azimuth (in degree). The radius axis stands for the
average maximum amplitudes of filtered GPS-TEC series (in TECU). The north direction is set as the 0°
azimuth. The maximums of filtered TEC less than 0.015 TECU are neglected. Obvious azimuthal asymme-
try in maximum of TEC series can be seen in this diagram. Specifically, the average maximum amplitudes
observed in southwest are equal to or greater than 0.06 TECU, which are larger than the average maximum
amplitudes in southeast region (lower than 0.04 TECU). The average maximum amplitudes observed in
southwest are close to the fault rupture region (azimuth angle 258°).
3.3. CID Waveform and Spectrum Signal
The spectrum analysis of the filtered GPS-TEC time series can provide more information about the CID.
Figure7 shows some cases of disturbance signal waveforms and seismic waveforms in different azimuthal
directions. Disturbance waveforms observed by PRN03, 04 and 26 are shown in Figures7a–7c, respectively.
The x-axis represents the UTC time. The dashed black lines in Figures7a–7c represent the onset time of the
earthquake and the name of selected station is located in the right side of each corresponding waveform.
The significant TEC disturbances can be distinguished from the waveforms after about 12min of the main
shock. As the distance between selected station and epicenter increases, the amplitude of waveforms be-
comes to decrease in Figure7c, and the appearance time of negative peak begins to delay in Figures7a–7c.
It is notable that, in Figure7c, the signals observed by station CN35 and SAN0 show a typical N-shape
waveform. However, as the distance from epicenter increases, the waveforms observed by far-field station
Figure 4. Traveling-time diagrams of co-seismic ionospheric disturbance (CID) and graphs of subionospheric point (SIP) tracks distribution through Global
Positioning System observation for PRN03, 04 and 26. (a) Traveling-time diagram of CID observed from PRN03 and SIP tracks for PRN03. (b) Traveling-time
diagram of CID observed from PRN04 and SIP tracks for PRN04. (c) Traveling-time diagram of CID observed from PRN26 and SIP tracks for PRN26. Dashed
black line represents the onset time of the 2020 Jamaica earthquake. The color bars indicate the value range of filtered total electron content (TEC) series. The
black diagonal line is used to linearly fit the propagation velocity of TEC disturbances. The red pentagram and the blue lines in SIP tracks distribution graphs
represent the epicenter and SIP tracks during 19:10–19:30 UTC, respectively.
Figure 5. (a) Distribution of ionospheric piercing point (IPP) epicentral azimuths and elevation angles of the
maximum in filtered total electron content (TEC) series obtained from PRN03. (b) Distribution of IPP epicentral
azimuths and elevation angles of the maximum in filtered TEC series obtained from PRN04. (c) Distribution of IPP
epicentral azimuths and elevation angles of the maximum in filtered TEC series obtained from PRN26. The dashed
black lines marked the elevation angles range of the maximum amplitude larger than 0.03 TECU. The maximums of
filtered TEC less than 0.01 TECU are neglected.

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
8 of 13
LMNL, PUMO, LEPA, PUJE, HUA2 and GRZA appear in the form of an
inverted N-shaped waveform (Astafyeva etal.,2013). In Figure7a, all the
waveforms have N type forms, which are different from the waveforms of
selected stations with PRN26. Besides, the inverted N-shaped waveform
and N-shape waveform both appear in the signal observed from PRN04.
With the same passband filtering, the seismic waveforms at 2–5mHz in
southeast and southwest direction from the vertical broadband high-gain
seismometers are shown in Figures7d and7e respectively. The amplitude
of waveforms represents the normalized vertical ground displacement.
The y-axis represents the distance between seismograph and epicenter
and the x-axis indicates UTC time. It is apparent that the vertical ground
displacements in SW direction are larger than SE direction. Through lin-
early fitting, the spread speed of seismic waves in southwest direction
is about 3.75km/s, which is close to the speed 3.76km/s in southeast
direction. These two propagation speeds are both in the velocity range of
Rayleigh surface wave and close to the propagation velocities of CID de-
scribed in Figure4. This confirms that the CIDs are the secondary acous-
tic waves generated by seismic Rayleigh waves with dynamic coupling.
Besides, it should be noticed that the seismic waves in SW show a nega-
tive polarity, which are consistent with the inverted N-shape waveforms
of Rayleigh wave-induced ionospheric disturbances observed by PRN26
in the southwest area. The same result can be obtained by comparing
Figures7a and7d.
Furthermore, Figure8 shows the spectrograms of the filtered GPS-TEC time series from selected stations
and satellites after using short-time Fourier transform (STFT) to convert TEC series from the time domain
to the frequency domain. The diagram order is station CN35 for PRN26, station HUA2 for PRN26, station
JME2 for PRN03 and station RDMS for PRN03 respectively. The left panel displays the filtered GPS-TEC
time series in blue line and distance changes in orange line, and the right panel represents the spectrogram
of corresponding GPS-TEC time series converting from STFT. The center frequencies of disturbance signals
for station CN35 and station GRZA are about 3.4 and 3mHz respectively, while frequencies of disturbance
signals for station JME2 and station RDMS are centered at about 3.3 and 3.1mHz. The center frequencies
for selected stations are all in the frequency range of infrasonic wave. Therefore, these CID signals have the
same frequency characteristic.
3.4. Discussion
The eruption of the 2020 Jamaica strike-slip earthquake excited seismic Rayleigh surface waves that in-
duced the upward secondary acoustic waves with dynamic coupling and caused TEC fluctuation in the ion-
osphere height. The distinct CIDs were captured by GPS observation (mainly PRN03, PRN04 and PRN26)
after about 12min of the main shock. This demonstrates that the strike-slip Jamaica earthquake can also
cause co-seismic ionospheric disturbance (Astafyeva etal., 2014; Cahyadi & Heki,2015) and the correla-
tion of directivity between ground vibrations and Rayleigh wave-induced CID has been confirmed (H. Liu
etal.,2017). The CID first appeared in the south area of the epicenter and propagated to the south far-field
outward from the epicenter, while no obvious CID can be detected in the north region of the epicenter. This
result is consistent with previous works, i.e., the 2003 Tokachi-Oki earthquake (Rolland etal.,2011), the
2008 Wenchuan earthquake (Zhao & Hao,2015) and the 2012 Haida Gwaii earthquake (Jin, Jin, & Li,2017).
Because, previous works have shown that the plasma waves cannot propagate perpendicularly to MF lines
and CID propagating to equatorward is at smaller angles with MF line and favorable to be detected for long
disturbance (Astafyeva etal.,2014; Bagiya etal.,2019; Heki & Ping,2005). The 2020 Jamacia earthquake
occurred in the northern hemisphere, and the detailed information about the geomagnetic field near the
epicenter area was obtained using the IGRF model from National Oceanic and Atmospheric Administration
(NOAA, https://www.ngdc.noaa.gov). The geomagnetic field has a westerly declination around 6.40°, and a
downward inclination around 47.65° at the ionosphere height of 350km. Therefore, the Rayleigh-induced
Figure 6. The polar diagram for the distribution of the average maximum
amplitude for all the filtered Global Positioning System-total electron
content (TEC) time series obtained from PRN03, 04 and 26 in every
15° azimuth bin during 19:10–19:30 UTC. The theta axis represents the
subionospheric point epicenter azimuth of the maximum TEC series (in
degree). The radius axis stands for the average maximum amplitudes of
filtered TEC series (in TECU). The north direction is set as the 0° azimuth.
The maximums of filtered TEC less than 0.015 TECU are neglected.

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
9 of 13
disturbance wave vector in the south area of epicenter propagates at small angles (less than 30°) to the MF
line and can be easily detected from GPS observation.
As is mentioned in Figure1b, the fault rupture was located at the epicentral azimuth 258° and the rupture
propagation direction was mainly along westward direction (Tira etal.,2020). Besides, the vertical ground
displacements shown in Figures7d and7e demonstrate that the southwest area has larger ground displace-
ments than southeast area. These facts corroborate the larger ground deformation and energy propagation
in the west and southwest of the epicenter. Due to the decreasing background air density, the amplitudes
of CID grow exponentially with the altitude to conserve energy when the acoustic waves propagate upward
into the atmosphere (Meng etal.,2019). Therefore, the CID with propagation along the southwest direc-
tion will have a more intense ionospheric response and larger amplitude after amplification. On the other
hand, the non-tectonic forcing mechanisms, such as satellite geometry, should be considered. However, the
elevation angles of satellite PRN03, 04 and 26 are all at low angle range, which may have less impact on the
amplitude of CID. Moreover, Zhao and Hao(2015) found two groups of CIDs with maximum amplitudes in
the direction of azimuth 150° and 135° following the 2008 Wenchuan earthquake, which are perpendicular
Figure 7. (a) Disturbance waveforms from observations of selected stations for PRN03. (b) Disturbance waveforms
from observation of selected stations for PRN04. (c) Disturbance waveforms from observation of selected stations for
PRN26. (d) Seismic waves in southeast direction. (e) Seismic waves in southwest direction.

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
10 of 13
Figure 8. The spectrograms of TEC disturbances series from selected stations and satellites. (a) Station CN35 for satellite PRN26. (b) Station GRZA for satellite
PRN26. (c) Station JME2 for satellite PRN03. (d) Station RDMS for satellite PRN03.

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
11 of 13
to the direction of the fault rupture. However, as the 2020 Jamacia earth-
quake has a different focal mechanism, the maximum amplitudes of
CIDs concentrate on the direction of azimuth 200–240°, which are close
to the direction of the fault rupture (258°, shown in Figure1b). For an-
other example, the 2012 Haida Gwaii earthquake, CID propagates toward
southeast direction, which is parallel to the fault rupture direction (Jin,
Jin, & Li,2017). The directivity relation between CID and fault rupture
may be determined by specific focal mechanism.
The disturbance signals display in the forms of inverted N-shape wave
and typical N-shape wave. Several CID signal waveforms show the same
polarity with generation source waves (Rayleigh waves). This demon-
strates that CIDs appear in different initial polarities may attribute to
different ground-motion patterns. Astafyeva and Heki(2009) suggested
that the waveform of disturbance signals repeat the initial ground crus-
tal motion and may contain the information about the focal mechanism
(Heki etal.,2006). The typical and inverse N-shape wave are caused by
mixed type of focal mechanism. This view can be supported from previ-
ous earthquake cases. For instance, Jin, Jin, and Li(2017) investigated the
CID following the 2012 Haida Gwaii earthquake through dual-frequency
GPS observation. The Rayleigh wave-induced CID showed a negative initial polarity, which was similar to
the polarity of waves recorded by the seismic station and bottom pressure records in the seafloor. Moreover,
Catherine etal.(2015) analyzed the ionospheric response following the 2012 Indian Ocean strike-slip earth-
quake using GPS-TEC measurements. Results suggest that TEC waveforms were mostly consistent with the
focal mechanism of the earthquake.
Besides, according to Sipkin(1994) and Kiratzi(2014), the focal mechanism illustrates the pattern of the
radiated seismic waves and can be determined by the first motion polarity of the body and surface wave.
Figure 9 represents the schematic diagram for focal mechanism. The specific information about focal
mechanism was obtained from USGS (https://earthquake.usgs.gov/earthquakes/eventpage/pt20028001/
moment-tensor). The P-axis, T-axis, fault plane and auxiliary plane are labeled in the diagram. The origin
represents the hypocenter and the theta axis shows the epicenter azimuths (in degree). It indicates that
during the slip, the southwest quadrant of the fault is a compression region while the southeast quadrant
can be considered as a dilatation or extension region. Thus, the appearance of inverted N-shaped waves in
the southwest area detected by PRN26 may attribute to the negative co-seismic vertical crustal movement,
and the typical N-shape waves detected in the near-field southeast area ascribes to the co-seismic vertical
ground uplift. The ground movement pattern matches the Rayleigh waves shown in Figures7d and7e.
However, Rolland etal.(2013) argued that the amplitude and waveform of TEC signals may be controlled
by other factors, such as geomagnetic field, geometry of the GPS line-of-sight signal and background ioni-
zation as well as geological structure (e.g., Tenzer etal.,2015). The dependence of waveform of near-field
CIDs on focal mechanisms can only be done in the regions where the geomagnetic field hardly distorts the
ionospheric response (Astafyeva & Heki,2009). However, different from Rolland etal.(2013), the inverted
N-shaped wave and N-shaped wave are both appear in the south area of epicenter, where the MF line has
moderate impact on the CID waves. As the absence of displacement and waveform data of seafloor in the
near-field area, further investigation is necessary to determine the formation of the inverted N-shaped wave.
4. Summary
The ionospheric responses to the 2020 Mw 7.7 Jamaica earthquake were studied and estimated by dense
dual-frequency GPS measurements. The CIDs are mainly captured by Satellite PRN03, PRN04 and PRN26
after about 12min of the main shock in the south near-field area, which is 300–500km away from the epi-
center and the southwest far-field area, which is 800km away from the epicenter. The propagation velocity
of CIDs observed from PRN03, PRN04 and PRN26 are 3.54km/s, 3.51km/s and 3.48km/s, respectively.
The variation amplitudes of the disturbances detected by PRN26 are larger than PRN03's and PRN04's. The
Figure 9. Schematic diagram for focal mechanism. Specific factors are
obtained from (https://earthquake.usgs.gov/earthquakes/eventpage/
pt20028001/moment-tensor).

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
12 of 13
average variation amplitude of the disturbances detected by PRN26 is larger than −0.08 TECU, while the
PRN03's reaches only to −0.05 TECU and PRN04's reaches to −0.07 TECU. The azimuthal asymmetry in
amplitude of filtered TEC series mainly attributes to the fault system and the spread distribution of earth-
quake energy. Besides, the center frequencies of the co-seismic ionospheric disturbances signals detected
by PRN26 are about 3.4 and 3.0mHz, while the disturbances signals detected by PRN03 are centered at 3.3
and 3.1mHz. These disturbance signals all belong to infrasonic waves. Furthermore, the ionospheric distur-
bances are the secondary acoustic waves in the infrasonic frequency range induced by the seismic Rayleigh
surface wave with dynamic coupling. The similar directivity and initial polarity between ground motion
and Rayleigh wave-induced CID will help comprehend the correlation and coupling mechanism of CID and
earthquake, but it should be further studied from more earthquake cases in the future.
Data Availability Statement
Great gratitude to UNAVCO (https://www.unavco.org/data/data.html) for providing the GPS observation
data, IRIS Data Management Center (http://ds.iris.edu/ds/) for seismograph data and NOAA (https://www.
ngdc.noaa.gov/geomag/calculators/magcalc.shtml#igrfwmm) for providing geomagnetic field information.
References
Afraimovich, E. L., Ding, F., Kiryushkin, V. V., Astafyeva, E. I., Jin, S., & Sankov, V. A. (2010). TEC response to the 2008 Wenchuan
Earthquake in comparison with other strong earthquakes. International Journal of Remote Sensing, 31(13), 3601–3613. https://doi.
org/10.1080/01431161003727747
Afraimovich, E. L., Palamartchouk, K. S., & Perevalova, N. P. (1998). GPS radio interferometry of traveling ionospheric disturbances. Jour-
nal of Atmospheric and Solar-Terrestrial Physics, 60(12), 1205–1223. https://doi.org/10.1016/j.asr.2016.11.001
Afraimovich, E. L., Perevalova, N. P., Plotnikov, A. V., & Uralov, A. M. (2001). The shock-acoustic waves generated by the earthquakes.
Annales Geophysicae, 19(4), 395–409. https://doi.org/10.5194/angeo-19-395-2001
Astafyeva, E. (2019). Ionospheric detection of natural hazards. Review of Geophysics, 57(4), 1265–1288. https://doi.org/10.1029/2019RG000668
Astafyeva, E., & Heki, K. (2009). Dependence of waveform of near-field coseismic ionospheric disturbances on focal mechanisms. Earth
Planets and Space, 61(7), 939–943. https://doi.org/10.1186/BF03353206
Astafyeva, E., Heki, K., Kiryushkin, V., Afraimovich, E., & Shalimov, S. (2009). Two-mode long-distance propagation of coseismic iono-
sphere disturbances. Journal of Geophysical Research, 114, A10307. https://doi.org/10.1029/2008JA013853
Astafyeva, E., Rolland, L. M., & Sladen, A. (2014). Strike-slip earthquakes can also be detected in the ionosphere. Earth and Planetary
Science Letters, 405, 180–193. https://doi.org/10.1016/j.epsl.2014.08.024
Astafyeva, E., Shalimov, S., Olshanskaya, E., & Lognonné, P. (2013). Ionospheric response to earthquakes of different magnitudes: Larger
quakes perturb the ionosphere stronger and longer. Geophysical Research Letters, 40, 1675–1681. https://doi.org/10.1002/grl.50398
Bagiya, M. S., Sunil, A. S., Rolland, L., Nayak, S., Ponraj, M., Thomas, D., & Ramesh, D. S. (2019). Mapping the impact of non-tecton-
ic forcing mechanisms on GNSS measured coseismic ionospheric perturbations. Scientific Reports, 9, 18640. https://doi.org/10.1038/
s41598-019-54354-0
Brunini, C., & Azpilicueta, F. J. (2009). Accuracy assessment of the GPS-based slant total electron content. Journal of Geodesy, 83(8),
773–785. https://doi.org/10.1007/s00190-008-0296-8
Cahyadi, M. N., & Heki, K. (2015). Coseismic ionospheric disturbance of the large strike-slip earthquakes in North Sumatra in 2012: Mw
dependence of the disturbance amplitudes. Geophysical Journal International, 200(1), 116–129. https://doi.org/10.1093/gji/ggu343
Cai, C., Liu, Z., Xia, P., & Dai, W. (2013). Cycle slip detection and repair for undifferenced GPS observations under high ionospheric activity.
GPS Solutions, 17, 247–260. https://doi.org/10.1007/s10291-012-0275-7
Calais, E., & Minster, J. B. (1995). GPS detection of ionospheric perturbations following the January 17, 1994, Northridge earthquake.
Geophysical Research Letters, 22(9), 1045–1048. https://doi.org/10.1029/95GL00168
Catherine, J. K., Vijayan, M. S. M., Syeda Rabiya, U. B., Shimna, K., Gahalaut, V. K., & Ramesh, D. S. (2015). Dichotomy in mode propaga-
tion of coseismic ionospheric disturbance: Observations from 11 April 2012 Indian Ocean earthquake. Journal of Geophysical Research,
120(5), 3854–3867. https://doi.org/10.1002/2014JA020621
Davies, K., D. M. Baker, (1965). Ionospheric effects observed around the time of the Alaskan earthquake of March 28, 1964. Journal of
Geophysical Research, 70(9), 2251–2253. https://doi.org/10.1029/JZ070i009p02251
Ducic, V., Artru, J., & Lognonné, P. (2003). Ionospheric remote sensing of the Denali Earthquake Rayleigh surface waves. Geophysical
Research Letters, 30(18), 1951. https://doi.org/10.1029/2003GL017812
Heki, K., Otsuka, Y., Choosakul, N., Hemmakorn, N., Komolmis, T., & Maruyama, T. (2006). Detection of ruptures of Andaman fault seg-
ments in the 2004 Great Sumatra Earthquake with coseismic ionospheric disturbances. Journal of Geophysical Research, 111. https://
doi.org/10.1029/2005JB004202
Heki, K., & Ping, J. S. (2005). Directivity and apparent velocity of the coseismic ionospheric disturbances observed with a dense GPS array.
Earth and Planetary Science Letters, 236, 845–855. https://doi.org/10.1016/j.epsl.2005.06.010
Jin, S. G. (2018). Two-mode ionospheric disturbances following the 2005 Northern California offshore earthquake from GPS measure-
ments. Journal of Geophysical Research, 123, 8587–8598. https://doi.org/10.1029/2017JA025001
Jin, S. G., Han, L., & Cho, J. (2011). Lower atmospheric anomalies following the 2008 Wenchuan Earthquake observed by GPS measure-
ments. Journal of Atmospheric and Solar-Terrestrial Physics, 73(7–8), 810–814. https://doi.org/10.1016/j.jastp.2011.01.023
Jin, S. G., Jin, R., & Kutoglu, H. (2017). Positive and negative ionospheric responses to the March 2015 geomagnetic storm from BDS ob-
servations. Journal of Geodesy, 91(6), 613–626. https://doi.org/10.1007/s00190-016-0988-4
Jin, S. G., Jin, R., & Li, D. (2017). GPS detection of ionospheric Rayleigh wave and its source following the 2012 Haida Gwaii earthquake.
Journal of Geophysical Research, 122(1), 1360–1372. https://doi.org/10.1002/2016JA023727
Acknowledgments
This work was supported by the
National Natural Science Foundation
of China (NSFC) Project (Grant No.
12073012) and National Natural Science
Foundation of China-German Science
Foundation (NSFC-DFG) Project (Grant
No. 41761134092).

Journal of Geophysical Research: Space Physics
CHAI AND JIN
10.1029/2020JA028995
13 of 13
Jin, S. G., Occhipinti, G., & Jin, R. (2015). GNSS ionospheric seismology: Recent observation evidences and characteristics. Earth-Science
Reviews, 147, 54–64. https://doi.org/10.1016/j.earscirev.2015.05.003
Jin, S. G., & Park, P. H. (2006). Strain accumulation in South Korea inferred from GPS measurements. Earth Planets and Space, 58(5),
529–534. https://doi.org/10.1186/BF03351950
Jin, S. G., & Su, K. (2020). PPP models and performances from single- to quad-frequency BDS observations. Satellite Navigation, 1(1), 16.
https://doi.org/10.1186/s43020-020-00014-y
Jin, S. G., Wang, J., Zhang, H., & Zhu, W. Y. (2004). Real-time monitoring and prediction of the total ionospheric electron content by means
of GPS observations. Chinese Journal of Astronomy and Astrophysics, 28(3), 331–337. https://doi.org/10.1016/j.chinastron.2004.07.008
Kiratzi, A. A. (2014). Mechanism of earthquake in Aegean. Encyclopedia of Earthquake Engineering, 4, 3966.
Komjathy, A., Yang, Y. M., Meng, X., Verkhoglyadova, O., Mannucci, A. J., & Langley, R. B. (2016). Review and perspectives: Understanding
natural-hazards-generated ionospheric perturbations using GPS measurements and coupled modeling. Radio Science, 51(7), 951–961.
https://doi.org/10.1002/2015RS005910
Leonard, R. S., & Barnes, R. A., JR (1965). Observation of ionospheric disturbances following the Alaska Earthquake. Journal of Geophys-
ical Research, 70(5), 1250–1253. https://doi.org/10.1029/JZ070i005p01250
Liu, H., Ding, F., Zhao, B., Li, J., Hu, L., Wan, W., & Ning, B. (2017). Ionospheric response following the Mw 7.8 Gorkha earthquake on 25
April 2015. Journal of Geophysical Research: Space Physics, 122, 6495–6507. https://doi.org/10.1002/2016JA023079
Liu, H., Zhang, K. K., Imtiaz, N., Song, Q., & Zhang, Y. (2021). Relating far-field coseismic ionospheric disturbances to geological struc-
tures. Journal of Geophysical Research: Space Physics, 126(7). https://doi.org/10.1029/2021JA029209
Liu, J. Y., Chen, C. H., Lin, C. H., Tsai, H.-F., Chen, C.-H., & Kamogawa, M. (2011). Ionospheric disturbances triggered by the 11 March
2011 M 9.0 Tohoku earthquake. Journal of Geophysical Research, 116, A06319. https://doi.org/10.1029/2011JA016761
Meng, X., Vergados, P., Komjathy, A., & Verkhoglyadova, O. (2019). Upper atmospheric responses to surface disturbances: An observation-
al perspective. Radio Science, 54(11), 1076–1098. https://doi.org/10.1029/2019RS006858
Nguyen, V. K., Rovira-Garcia, A., Juan, J. M., Sanz, J., González-Casado, G., La, T. V., & Ta, T. H. (2019). Measuring phase scintillation at
different frequencies with conventional GNSS receivers operating at 1 Hz. Journal of Geodesy, 93, 1985–2001. https://doi.org/10.1007/
s00190-019-01297-z
Occhipinti, G., Dorey, P., Farges, T., & Lognonné, P. (2010). Nostradamus: The radar that wanted to be a seismometer. Geophysical Research
Letters, 37, L18104. https://doi.org/10.1029/2010GL044009
Okoh, D., Onwuneme, S., Seemala, G., Jin, S. G., Rabiu, B., Nava, B., & Uwamahoro, J. (2018). Assessment of the NeQuick-2 and IRI-Plas
2017 models using global and long-term GNSS measurements. Journal of Atmospheric and Solar-Terrestrial Physics, 170, 1–10. https://
doi.org/10.1016/j.jastp.2018.02.006
Perevalova, N. P., Sankov, V. A., Astafyeva, E., & Zhupityaeva, A. S. (2014). Threshold magnitude for ionospheric response to earthquakes.
Journal of Atmospheric and Solar-Terrestrial Physics, 108, 77–90. https://doi.org/10.1016/j.jastp.2013.12.014
Rolland, L. M., Lognonné, P., & Munekane, H. (2011). Detection and modeling of Rayleigh wave induced patterns in the ionosphere. Jour-
nal of Geophysical Research: Space Physics, 116, 1–18. https://doi.org/10.1029/2010JA016060
Rolland, L. M., Occhipinti, G., Lognonné, P., & Loevenbruck, A. (2010). Ionospheric gravity waves detected offshore Hawaii after tsunamis.
Geophysical Research Letters, 37. https://doi.org/10.1029/2010GL044479
Rolland, L. M., Vergnolle, M., Nocquet, J. M., Sladen, A., Dessa, J. X., Tavakoli, F., etal. (2013). Discriminating the tectonic and non-tecton-
ic contributions in the ionospheric signature of the 2011, Mw 7.1, dip-slip van earthquake, eastern Turkey. Geophysical Research Letters,
40, 2518–2522. https://doi.org/10.1002/grl.50544
Ronald, V. R. (1966). Evidence of Long-Period Acoustic-Gravity Waves Launched into the F region by the Alaskan. Journal of Geophysical
Research, 71(1), 343–345. https://doi.org/10.1029/jz071i001p00343
Ronald, V. R. (1967). Acoustic-gravity waves in the upper atmosphere due to a nuclear detonation and an earthquake. Journal of Geophys-
ical Research, 72(5), 1599–1610. https://doi.org/10.1029/JZ072i005p01599
Saito, A., Tsugawa, T., Otsuka, Y., Nishioka, M., Iyemori, T., Matsumura, M., etal. (2011). Acoustic resonance and plasma depletion detect-
ed by GPS total electron content observation after the 2011 off the Pacific coast of Tohoku earthquake. Earth Planets and Space, 63(7),
863–867. https://doi.org/10.5047/eps.2011.06.034
Sipkin, S. A. (1994). Rapid determination of global moment-tensor solutions. Geophysical Research Letters, 21(16), 1667–1670. https://doi.
org/10.1029/94GL01429
Sripathi, S., Singh, R., Tiwari, P., & Kumar, M. R. (2020). On the co-seismic ionospheric disturbances (CIDs) in the rapid run ionosonde
observations over Allahabad following Mw 7.8 Nepal Earthquake on April 25, 2015. Journal of Geophysical Research, 125, 2169–2185.
https://doi.org/10.1029/2019JA027001
Tenzer, R., Chen, W., & Jin, S. G. (2015). Effect of the upper mantle density structure on the Moho geometry. Pure and Applied Geophysics,
172(6), 1563–1583. https://doi.org/10.1007/s00024-014-0960-2
Tira, T., Okuwaki, R., Yagi, Y., & Yamashita, S. (2020). Rupture process of the 2020 Caribbean Earthquake along the oriente trans-
form fault, involving supershear rupture and geometric complexity of fault. Geophysical Research Letters, 48(1). https://doi.
org/10.1029/2020GL090899
Tsugawa, T., Saito, A., Otsuka, Y., Nishioka, M., Maruyama, T., Shinagawa, H., etal (2011). Ionospheric disturbances detected by GPS total
electron content observation after the 2011 Tohoku earthquake. AGU Fall Meeting Abstracts.
Zhao, B., & Hao, Y. (2015). Ionospheric and geomagnetic disturbances caused by the 2008 Wenchuan earthquake: A revisit. Journal of
Geophysical Research: Space Physics, 120, 5758–5777. https://doi.org/10.1002/2015JA021035