Study on Seasonal Variations of Plasma Bubble Occurrence over Hong Kong Area Using GNSS Observations

remote sensing
Letter
Study on Seasonal Variations of Plasma Bubble
Occurrence over Hong Kong Area Using
GNSS Observations
Long Tang 1,2 , Wu Chen 2, * , Osei-Poku Louis 2 and Mingli Chen 3
 1 School of Civil and Transportation Engineering, Guangdong University of Technology, Guangzhou 510006,
 China; ltang@gdut.edu.cn
 2 Department of Land Surveying and Geo-Informatics, Hong Kong Polytechnic University, Kowloon,
 Hong Kong; louis.oseipoku@connect.polyu.hk
 3 Department of Building Services Engineering, Hong Kong Polytechnic University, Kowloon, Hong Kong;
 mingli.chen@polyu.edu.hk
 * Correspondence: wu.chen@polyu.edu.hk
 



 Received: 6 July 2020; Accepted: 26 July 2020; Published: 28 July 2020 


 Abstract: In this study, the characteristics and causes of the seasonal variations in plasma bubble
 occurrence over the Hong Kong area were investigated using the local Global Navigation Satellite
 System (GNSS) network. Generally, the occurrences of plasma bubbles were larger in the two
 equinoxes than in the two solstices. Furthermore, two seasonal asymmetries in plasma bubble
 occurrence were observed: plasma bubble activity was more frequent in the spring equinox than
 in the autumn equinox (equinoctial asymmetry), and more frequent in the summer solstice than
 in the winter solstice (solstitial asymmetry). The equinoctial asymmetry could be explained using
 the Rayleigh–Taylor (R–T) instability mechanism, due to larger R–T growth rates in the spring
 equinox than in the autumn equinox. However, the R–T growth rate was smaller in the summer
 solstice than in the winter solstice, suggesting the R–T instability mechanism was inapplicable to the
 solstitial asymmetry. Our results showed there were more zonally propagating atmospheric gravity
 waves (GWs) induced by thunderstorm events over the Hong Kong area in the summer solstice than
 the winter solstice. So, the solstitial asymmetry could be attributed to the seeding mechanism of
 thunderstorm-driven atmospheric GWs.

 Keywords: GNSS; ionosphere; plasma bubble occurrence; seasonal variation

1. Introduction
 A plasma bubble is one kind of frequent ionospheric weather event in low-latitude areas.
The presence of a plasma bubble can cause severe effects on radio signals of communication and
navigation systems, such as the Global Navigation Satellite System (GNSS), when they travel through
the ionosphere (e.g., [1,2]). As a result, the study of climatology and the predictability of plasma bubbles
has attracted extensive focus from scientific communities in recent decades (e.g., [3–8]). Research shows
that plasma bubble occurrence presents seasonal variations and longitudinal variations (e.g., [7,8]).
Generally, plasma bubble activity is frequent during equinox periods in the Asian area and African
area, during the summer solstice in the Pacific area, and during the northern winter solstice in the
American-Atlantic area. In addition, asymmetric seasonal variations in plasma bubble occurrence in
different areas have also been reported (e.g., [8]).
 Two types of physical mechanisms have been employed to account for the seasonal variations
and longitudinal variations in plasma bubble occurrence. The first one is the Rayleigh–Taylor (R–T)
instability mechanism, which is widely applied to explain the formation of plasma bubbles [9].

Remote Sens. 2020, 12, 2423; doi:10.3390/rs12152423 www.mdpi.com/journal/remotesensing

Remote Sens. 2020, 12, 2423 2 of 10

The growth condition of R–T instability is considered to be the important factor that controls the
climatology of plasma bubble occurrence. Tsunoda [3] indicated that plasma bubble occurrence was
maximized when the geomagnetic field was parallel to the solar terminator. Kil et al. [10] and Sripathi
 Remote Sens. 2020, 12, x FOR PEER REVIEW 2 of 11
et al. [11] suggested the plasma density might exert an important effect on the seasonal variations in
plasma climatology
 bubble occurrence. Maruyama
 of plasma bubble et al.Tsunoda
 occurrence. [12] indicated thatthat
 [3] indicated meridional windoccurrence
 plasma bubble played awas dominant
role in the formation
 maximized when of the
 equinoctial
 geomagnetic asymmetry. The second
 field was parallel to the mechanism is the
 solar terminator. Kil atmospheric
 et al. [10] andgravity
wave (GW) Sripathi et al. [11]
 seeding suggested the
 mechanism, plasma
 which density might
 emphasizes theexert an important
 importance effect on
 of GWs to the
 the seasonal
 generation of
 variations in plasma bubble occurrence. Maruyama et al. [12] indicated that meridional wind
plasma bubbles [13,14]. Tsunoda [15] found plasma bubble occurrence was enhanced when a region
 played a dominant role in the formation of equinoctial asymmetry. The second mechanism is the
with frequent GW events was situated near the magnetic equator. Takahashi et al. [16] indicated
 atmospheric gravity wave (GW) seeding mechanism, which emphasizes the importance of GWs to
that medium-scale
 the generation traveling
 of plasma ionospheric disturbances
 bubbles [13,14]. Tsunoda [15] (MSTIDs) were likely
 found plasma bubbletooccurrence
 be a seedwas source for
plasma bubbles.
 enhanced when a region with frequent GW events was situated near the magnetic equator.
 Due Takahashi et al. [16] indicated
 to the longitudinal that medium-scale
 dependence, traveling ionospheric
 it was necessary to study the disturbances (MSTIDs) were
 seasonal variations in plasma
 likely to be ain
bubble occurrence seed source for
 different plasma
 areas bubbles.
 to fully understand its source mechanism. Here, we focused on the
 Due to the longitudinal dependence, it was necessary to study the seasonal variations in
Hong Kong area, which is situated near the magnetic equator. Ji et al. [17] and Kumar et al. [18] studied
 plasma bubble occurrence in different areas to fully understand its source mechanism. Here, we
the seasonal
 focusedvariations
 on the Hong in plasma bubble
 Kong area, occurrence
 which is situatedover
 nearthis
 the area using
 magnetic GNSSJitotal
 equator. et al.electron
 [17] andcontent
(TEC) dataKumarduring
 et al.2001–2012
 [18] studied and
 theobtained very meaningful
 seasonal variations in plasmaresults.
 bubble However,
 occurrence the
 overcause for using
 this area the seasonal
variationsGNSSin plasma bubblecontent
 total electron occurrence
 (TEC)was
 datararely
 duringinvolved
 2001–2012 in and
 theirobtained
 studies.very
 Thismeaningful
 study sought to address
 results.
this gapHowever,
 by looking theat cause for the seasonal
 the potential variations
 sources in plasmavariations
 of the seasonal bubble occurrence was bubble
 in plasma rarely involved
 occurrence.in
 their studies. This study sought to address this gap by looking at the potential sources of the
2. Data seasonal variations in plasma bubble occurrence.
 and Methods

2.1. Data2. Data and Methods
 The2.1. Data data employed in this study were ionospheric TEC, which was extracted from GNSS
 primary
observationThe files. The 30
 primary datas sampling
 employed GNSS observation
 in this study files were
 were ionospheric acquired
 TEC, from
 which was the website
 extracted from of the
 GNSS observation files. The 30 s sampling GNSS observation files were
Hong Kong satellite reference network [19]. GNSS data during 2001–2012 were already processed byacquired from the website
 of the
Ji et al. [17] and Hong
 Kumar Kong satellite
 et al. reference
 [18], so network
 their results were[19]. GNSS referenced
 directly data duringin 2001–2012 were
 this study. already
 Here, GNSS data
 processed by Ji et al. [17] and Kumar et al. [18], so their results were directly referenced in this study.
during a four year period, from 2014 to 2017, were also processed to compare with thunderstorm
 Here, GNSS data during a four year period, from 2014 to 2017, were also processed to compare with
data (nothunderstorm
 data duringdata 2001–2012).
 (no data According to the F10.7
 during 2001–2012). index,tosolar
 According the activity decreased
 F10.7 index, year by year
 solar activity
during thisdecreased year by year during this period [20]. Figure 1 shows the GNSS station distributiononly
 period [20]. Figure 1 shows the GNSS station distribution in Hong Kong. Here, in three
stations,Hong
 HKNP, Kong. Here, only three stations, HKNP, HKKT, and HKOH, were adopted due to the shortstations.
 HKKT, and HKOH, were adopted due to the short distance between adjoining
In addition, the between
 distance adjoining stations.
 global ionospheric In addition,
 map (GIM) the global
 data and ionospheric
 thunderstorm map
 data (GIM)this
 during dataperiod
 and were
 thunderstorm data during this period were also applied in the analysis.
also applied in the analysis. The GIM files were downloaded from the Center for Orbit Determination The GIM files were
 downloaded from the Center for Orbit Determination in Europe (CODE) [21]; the data for
in Europe (CODE) [21]; the data for thunderstorm activity were obtained by a very low frequency
 thunderstorm activity were obtained by a very low frequency (VLF) detection network we
(VLF) detection network
 established we established in this area.
 in this area.

 36'

 30'

 HKKT
 24'

 22°N
 18.00'

 HKNP HKOH
 12'

 km
 0 4 8 12
 6'
 54' 114°E 6' 12' 18' 24'

 Figure 1. The Global Navigation Satellite System (GNSS) network in Hong Kong. The red points
 indicate the positions of stations. Three annotated stations, HKNP, HKKT, and HKOH, were applied in
 this study.

Remote Sens. 2020, 12, x FOR PEER REVIEW 3 of 11

 Figure 1. The Global Navigation Satellite System (GNSS) network in Hong Kong. The red points
 indicate
 Remote the 12,
 Sens. 2020, positions
 2423 of stations. Three annotated stations, HKNP, HKKT, and HKOH, were 3 of 10
 applied in this study.

 2.2.Methods
2.2. Methods

 ToTodetect
 detectplasma
 plasmabubbles,
 bubbles,ionospheric
 ionospheric TEC
 TEC waswas firstlyobtained
 firstly obtained from
 from GNSS
 GNSS observation
 observation files.
 files.
 Ionospheric TEC time series can be computed using the GNSS dual-frequency
Ionospheric TEC time series can be computed using the GNSS dual-frequency geometry-free geometry-free combined
 carrier phase
combined observations
 carrier for each pair
 phase observations of satellite
 for each pair ofand receiver
 satellite and[22]
 receiver [22]
 
 TEC== k· ( L −
 ∙ [( ) + ( − ) (1) (1)
 1 − L2 ) + (N1 − N2 )]
where = / 40.3( h  − )i, and are frequencies, and are carrier phase
 where k = f12and
observations, f22 / f12 − f22 are
 40.3and , f1carrier
 and f2phase
 are frequencies,
 ambiguities.L1 The
 and L 2 are carrier
 carrier phase phase observations,
 ambiguities were
 and N 1 and N 2 are carrier phase ambiguities. The carrier phase ambiguities
unknown but were constant. The unit of TEC is TECU (1 TECU = 10 /m ). The elevation mask angle
 16 2 were unknown but were
 constant.
was Thefor
 set as 20° unit
 eachof pair
 TECof is satellite
 TECU (1and TECU = 10 /m ). The elevation mask angle was set as 20◦ for
 receiver.
 16 2

 eachNext,
 pair ofwe
 satellite
 fit theand receiver.
 primitive TEC time series to eliminate the trend term and constant carrier
 Next, we fit the primitive
phase ambiguities. Here, the third-order TEC time series to eliminatesmoothing
 Savitzky–Golay the trend term
 filterand constant
 with movingcarrier phase
 window
 ambiguities.
length Here,
 of 2 hours wastheemployed
 third-order[23,24].
 Savitzky–Golay
 Then, thesmoothing
 detrendedfilter
 TECwith moving
 (dTEC) timewindow length
 series that of 2 h
 might
 was employed [23,24]. Then, the detrended
contain plasma bubble were computed as follows TEC (dTEC) time series that might contain plasma bubble
 were computed as follows
 
 dTEC== TEC−− TEC f (2) (2)

 where 
where TEC fisisthe
 thefitted
 fittedTEC
 TECusing
 usingthetheSavitzky–Golay
 Savitzky–Golaysmoothing
 smoothingfilter.
 filter.To
 Todetect
 detecta aplasma
 plasmabubble,
 bubble,
a athreshold
 thresholdofof TEC depletion(minimum)
 TEC depletion (minimum)was wassetset to −3
 to −3 TECU;
 TECU; the adjacent
 the adjacent maximum
 maximum shouldshould be
 be smaller
smaller than half of the absolute value of TEC depletion. In addition, only the ones
 than half of the absolute value of TEC depletion. In addition, only the ones that were simultaneously that were
simultaneously
 observed by the observed by the three
 three employed GNSS employed
 stations,GNSS
 namely stations,
 HKNP,namely
 HKKT,HKNP, HKKT,
 and HKOH, andcounted
 were HKOH as ,
were counted
 effective plasma as bubbles.
 effective Figure
 plasma bubbles.anFigure
 2 presents example 2 presents
 of a plasmaan bubble
 example of a plasma
 detected bubble
 by station HKKT
detected by station HKKT and satellite PRN 23 on January 25, 2014. In this case,
 and satellite PRN 23 on January 25, 2014. In this case, the magnitude of TEC depletion was −10.3 the magnitude of
TEC depletion was −10.3 TECU, and the two adjacent
 TECU, and the two adjacent maximums were 4.2 and 4.5 TECU, respectively.maximums were 4.2 and 4.5 TECU,
respectively.
 (a) Primitive TEC and fitted TEC
 -40
 Primitive TEC
 -50 Fitted TEC
 TECU

 -60

 -70

 -80
 13 13.5 14 14.5 15 15.5 16 16.5 17
 UT
 (b) Detrended TEC
 10
 dTEC max1 max2
 5

 0
 TECU

 -5
 bubble
 -10
 min
 -15
 13 13.5 14 14.5 15 15.5 16 16.5 17
 UT

 Figure
 Figure AnAn
 2. 2. exampleofofa aplasma
 example plasmabubble
 bubbledetected
 detectedbyby station
 station HKKT
 HKKT andand satellite
 satellite PRN
 PRN23 23on
 onJanuary
 January25,
 25,2014.
 2014.(a)(a)Primitive
 Primitivetotal electron
 total electroncontent (TEC)
 content (TEC)time series
 time (blue
 series line)
 (blue andand
 line) fitted TECTEC
 fitted timetime
 series using
 series
 Savitzky–Golay smoothing filter (green line); (b) detrended TEC time series
 using Savitzky–Golay smoothing filter (green line); (b) detrended TEC time series obtained by obtained by subtracting
 between the
 subtracting primitive
 between theand fitted TEC
 primitive and time
 fittedseries
 TEC(red
 timeline). The
 series annotation
 (red line). The “min” represents
 annotation TEC
 “min”
 depletion (minimum) and the annotations “max1” and “max2” represent the adjacent two maximums.
 represents TEC depletion (minimum) and the annotations “max1” and “max2” represent the
 adjacent two maximums.
 After detecting a plasma bubble, we could calculate the seasonal occurrence. A year was divided
 into four seasons: the spring equinox (from February to April), the summer solstice (from May to July),
 the autumn equinox (from August to October), and the winter solstice (from November to January).
 The monthly plasma bubble occurrence was computed by dividing the number of days with plasma

Remote Sens. 2020, 12, 2423 4 of 10

bubbles in a month by the total number of days in the month. Then, the seasonal plasma bubble
occurrence was obtained by averaging the monthly plasma bubble occurrence of the three months
comprising a season.

3. Results
 Figure 3 presents the observed seasonal plasma bubble occurrence during each year from 2014 to
2017. Figure 4 presents average solar F10.7 indices for different seasons during each year from 2014 to
2017. Obviously, plasma bubble occurrence was maximum during the high solar activity year of 2014
and minimum during the low solar activity year of 2017, indicating that it was dependent on solar
activity. As seen in Figure 3, the occurrences of plasma bubbles were significantly larger in the two
equinoxes than in the two solstices in the high solar activity years of 2014 and 2015. In the low solar
activityRemote
 yearSens.
 of 2017,
 2020, 12,the difference
 x FOR in plasma bubble occurrence between the summer solstice5 and
 PEER REVIEW of 11 the
two equinoxes was not obvious.
 Remote Sens. 2020, 12, x FOR PEER REVIEW 5 of 11
 EPB Occurence
 100
 2014 2015
 EPB Occurence 2016 2017
 90
 100
 2014 2015 2016 2017
 80
 90

 70
 80

 60
 70

 60
 50
 %

 40
 50
 %

 30
 40

 20
 30

 10
 20

 100
 SE SS AE WS
 0 Season
 SE SS AE WS
 Season
 Figure 3. The seasonal plasma bubble occurrence during each year from 2014 to 2017. SE, SS, AE,
 and WSseasonal
 3. The represent the spring
 plasma bubble equinox, summer
 occurrence solstice,
 during eacheachautumn
 year equinox,
 fromfrom
 20142014 and SE,
 to 2017. winter AE,
 solstice, WS
 FigureFigure 3. The seasonal plasma bubble occurrence during year to 2017.SS,
 SE, SS, and
 AE,
 respectively.
 represent the spring equinox, summer solstice, autumn equinox, and winter solstice, respectively.
 and WS represent the spring equinox, summer solstice, autumn equinox, and winter solstice,
 respectively.
 F10.7 index
 200
 2014 F10.7 index
 2015 2016 2017
 180
 200
 2014 2015 2016 2017
 160
 180

 140
 160

 120
 140
 sfu

 120
 100
 sfu

 80
 100

 60
 80

 40
 60

 20
 40

 200
 SE SS AE WS
 0 Season
 SE SS AE WS
 4. The Seasonseasons during each year from 2014 to 2017.
 FigureFigure 4. average solar
 The average F10.7
 solar F10.7index
 index for different
 for different seasons during each year from 2014 to 2017. SE,
 SE, SS,SS,
 AE, and WS represent the spring equinox,
 AE, and WS represent the spring equinox, summer summer solstice,
 solstice, autumnautumn
 equinox, equinox,
 and winterand winter
 solstice,
 Figure 4. The average solar F10.7 index for different seasons during each year from 2014 to 2017. SE,
 solstice, respectively.
 respectively.
 SS, AE, and WS represent the spring equinox, summer solstice, autumn equinox, and winter solstice,
 respectively.
 4. Discussion
 4. Discussion
 The R–T instability mechanism is widely applied to explain the formation of plasma bubbles,
 which can be assessed using the R–T growth rate. The flux tube integrated R–T growth rate can be
 The R–T instability mechanism is widely applied to explain the formation of plasma bubbles,
 calculated using following formula [9]
 which can be assessed using the R–T growth rate. The flux tube integrated R–T growth rate can be

Remote Sens. 2020, 12, 2423 5 of 10

 It can be seen from Figure 3 that the plasma bubble occurrence was greater in the spring equinox
than in the autumn equinox, especially during the high solar activity years of 2014 and 2015, indicating
that the seasonal variations in plasma bubble occurrence presented equinoctial asymmetry. As for
the solstices, the plasma bubble activity was more frequent in the summer solstice than in the winter
solstice. In the winter solstice, there were no plasma bubbles during the low solar activity years of
2016 and 2017 (Figure 3). This indicated that solstitial asymmetry also existed in seasonal variations in
plasma bubble occurrence.
 The seasonal variation characteristics of plasma bubble occurrence during 2001–2012 over the
Hong Kong area were investigated by Ji et al. [17] and Kumar et al. [18]. According to their studies,
larger values of plasma bubble occurrence were observed during high solar activity years (maximum
in 2002). The plasma bubble activity was generally more frequent in the two equinoxes than in the two
solstices, except during 2007 and 2008 (low solar activity). In addition, the equinoctial asymmetry in
plasma bubble occurrence, namely, a larger value in the spring equinox, was also observed during
2002–2012. In 2001, a larger value of plasma bubble occurrence was observed in the autumn equinox.
As for the solstitial asymmetry, it did not attract their attention and was not involved in their studies.
Actually, the solstitial asymmetry in plasma bubble occurrence was also observable during 2001–2012
(Figure 16 in Ji et al. [17] and Figure 3 in Kumar et al. [18]). A larger value was generally observed in the
summer solstice; an exception was in 2004, when the larger values were observed in the winter solstice.
 The results observed during these 16 years (2001–2012 and 2014–2017) showed that the plasma
bubble occurrence over the Hong Kong area presented significant seasonal variations. Generally,
the occurrences of plasma bubbles were larger in the two equinoxes than in the two solstices.
Furthermore, plasma bubble activity was more frequent in the spring equinox than in the autumn
equinox and more frequent in the summer solstice than in the winter solstice.

4. Discussion
 The R–T instability mechanism is widely applied to explain the formation of plasma bubbles,
which can be assessed using the R–T growth rate. The flux tube integrated R–T growth rate can be
calculated using following formula [9]
  
 ΣFP 
 P ge  F
 γRT = F VP − UL − F K − RT (3)
 
 ΣP + ΣEP  ve f f 

where ΣEP and ΣFP are integrated Pedersen conductivities in the E region and F region, respectively;
ULP is the integrated neutral wind; VP is the integrated upward drift speed; ge and vFef f are the effective
gravity and effective ion-neural collision frequency, respectively; KF is the F region integrated electron
content gradient; and RT is the recombination rate.
 Wu [25] and Wu [26] calculated the R–T growth rate at 18:00 local time (LT) for the globe using
Thermosphere Ionosphere Electrodynamics General Circulation Model (TIEGCM) during the years
of 2003 (solar maximum), 2006, and 2009 (solar minimum). The results showed that the growth rate
was positively correlated to the solar activity, which was consistent with our observed variations in
plasma bubble occurrence. In addition, the R–T growth rate presented similar seasonal variations and
longitudinal variations during each year, although different in magnitude. Here, we took the R–T
growth rate during the year of 2006 as an example, shown in Figure 5.
 As seen in Figure 5, the R–T growth rate was high during the two equinoxes and low during the
two solstices over the Hong Kong area (about longitude 110–120◦ ). In addition, there were more days
that the growth rate was greater than 60*10−5 /s (light green area) during the spring equinox than the
autumn equinox. These can explain the characteristics of plasma bubble occurrence, namely that a
larger value was observed during the two equinoxes and it showed equinoctial asymmetry. During the
two equinoxes, the solar terminator aligned with the geomagnetic field, leading to enhanced plasma
bubble activity [3].

Wu [25] and Wu [26] calculated the R–T growth rate at 18:00 local time (LT) for the globe using
Thermosphere Ionosphere Electrodynamics General Circulation Model (TIEGCM) during the years
of 2003 (solar maximum), 2006, and 2009 (solar minimum). The results showed that the growth rate
was positively correlated to the solar activity, which was consistent with our observed variations in
plasma bubble occurrence. In addition, the R–T growth rate presented similar seasonal variations
Remote Sens. 2020, 12, 2423 6 of 10
and longitudinal variations during each year, although different in magnitude. Here, we took the
R–T growth rate during the year of 2006 as an example, shown in Figure 5.

 Figure
 Figure 5.
 5. Seasonal and longitudinal
 Seasonal and longitudinalvariations
 variationsofofthe
 theRayleigh–Taylor
 Rayleigh–Taylor instability
 instability growth
 growth raterate at 270
 at 270 km
 km during
 during the the
 yearyear of 2006
 of 2006 (Figure
 (Figure 2a in2aWu
 in Wu [25]).
 [25]).

 As
 As seen
 shown in in
 Figure
 Figure5, 4,
 thedifferences
 R–T growth rate was
 in solar highduring
 activity duringthe thetwo
 twoequinoxes
 equinoxeswas andnot lowobvious,
 during
the two solstices
suggesting that itover
 wasthenotHong Kong area
 the cause for the (about longitude
 equinoctial 110–120°). In
 asymmetry. Kiladdition,
 et al. [10] there were more
 suggested the
days
plasmathat the growth
 density rate was
 might exert greater than
 an important effect60*10
 on the/sseasonal
 −5 (light green area)induring
 variations plasmathe spring
 bubble equinox
 occurrence.
than theetautumn
Sripathi equinox. These
 al. [11] indicated can explain
 that electron density thewas characteristics
 important toofthe plasma bubble
 formation of occurrence,
 equinoctial
namely
asymmetry. that Using
 a largerGIM,value was observed
 we calculated during ionospheric
 the average the two equinoxes and it showed
 TEC (integrated electronequinoctial
 density) at
asymmetry.
18:00 LT overDuring the Hongthe Kong
 two equinoxes, the solar
 area for different terminator
 seasons duringaligned
 each year with
 fromthe2014
 geomagnetic
 to 2017, shown field,
leading
in Figure to6.enhanced
 As seen in plasma
 Figure bubble activity [3].
 6, ionospheric TEC during the spring equinox was significantly larger
thanAs shown
 during theinautumn
 Figure 4, differences
 equinox, in solar
 especially inactivity
 high solar during the two
 activity years.equinoxes
 This result wassuggested
 not obvious, the
suggesting
ionosphericthat TECit(orwas not the
 electron cause for
 density) wasthe equinoctial
 likely an important asymmetry.
 factor for Kilthe
 etequinoctial
 al. [10] suggested
 asymmetry. the
plasma
According density might(3),
 to Equation exert an important
 ionospheric TEC is not effect on theparameter
 an explicit seasonal to variations
 calculate the in R–T
 plasmagrowth bubble
 rate.
occurrence.
It could affect Sripathi
 the R–Tetgrowth
 al. [11]rateindicated that electron
 by the variation of the density was electric
 polarization important field.toWhen
 the formation
 the F-region of
equinoctial
TEC was large, asymmetry. Using GIM,
 short circuiting of thewe polarization
 calculated the average
 electric ionospheric
 field driven inTEC (integrated
 the F-region electron
 weakened,
density)
so that aat 18:00polarization
 larger LT over theelectric
 Hong Kong field was areagenerated
 for different [9].seasons
 Then, the during each year
 expression ΣFP /from
 (ΣFP +2014ΣEP ) to
 in
2017,
Equationshown in Figure
 (3) became large6. (getting
 As seenclose in Figure
 to unity),6, ionospheric
 leading to a TEC large during
 R–T growththe spring
 rate. equinox was
significantly
 We discussedlargerthethan during
 solstitial the autumn
 asymmetry equinox,
 in plasma especially
 bubble in high
 occurrence. As solar
 can beactivity
 seen from years. This
 Figure 5,
result suggested
R–T growth rate the
 wasionospheric
 almost lower TEC than electron/s density)
 (or 10*10 −5 during the wassummer
 likely an important
 solstice, whereasfactoritsfor the
 value
equinoctial
was about 20*10 −5 –30*10
 asymmetry. −5 /s during
 According to the
 Equation
 winter(3), ionospheric
 solstice over the TEC is not
 Hong Kongan explicit
 area. That parameter to
 is to say,
calculate the in
the variations R–Tthegrowth
 R–T growth rate.rate
 It during
 could affectthe twothe R–T growth
 solstices were notrate by thewith
 consistent variation of the
 that of plasma
polarization
bubble occurrence.electricInfield. When
 addition, the F-region
 ionospheric TEC TEC
 was also wasnotlarge,
 alwaysshort circuiting
 larger during of thethe
 summerpolarization
 solstice
electric fieldthe
than during driven
 winterinsolstice
 the F-region
 (seen in weakened,
 Figure 6). This so suggested
 that a largerthat,polarization
 apart from the electric
 R–T growthfield was
 rate,
generated [9]. Then, the expression Σ /(Σ + Σ ) in
other factors could also exert effects on the formation of solstitial asymmetry.Equation (3) became large (getting close to
unity), leadingshowed
 Research to a large R–T
 that growth rate.GWs (or MSTIDs) associated with the polarization electric
 atmospheric
fields also played an important role in plasma bubble formation [13–16]. To generate the polarization
electric fields, the phase front (perpendicular to propagation direction) of GWs should be approximately
aligned with the geomagnetic direction [27]. Therefore, zonally propagating GWs were necessary due
to the meridional magnetic field line. MSTIDs were signals of atmospheric GWs in the ionosphere,
which can be called as ionospheric GWs. Here, we investigated the ionospheric GWs as a proxy of
atmospheric GWs.

Remote Sens. 2020, 12, 2423 7 of 10
 Remote Sens. 2020, 12, x FOR PEER REVIEW 7 of 11

 Ionospheric TEC
 100
 2014 2015 2016 2017
 90

 80

 70

 60

 TECU 50

 40

 30

 20

 10

 0
 SE SS AE WS
 Season

 6. The6.average
 FigureFigure ionospheric
 The average ionosphericTECTECat at
 18:00
 18:00local
 localtime
 time (LT) for different
 (LT) for differentseasons
 seasons during
 during each
 each yearyear
 from 2014
 fromto2014
 2017.toSE, SS,SE,
 2017. AE,SS,
 andAE,WSand
 represent the spring
 WS represent equinox,
 the spring summer
 equinox, solstice,
 summer autumn
 solstice, equinox,
 autumn
 and winter solstice,
 equinox, respectively.
 and winter solstice, respectively.

 FigureWe 7 presents
 discussed the the ionospheric detrended
 solstitial asymmetry TEC (dTEC)
 in plasma series extracted
 bubble occurrence. As canbybestation HKKT
 seen from
duringFigure 5, R–Tofgrowth
 the years 2014 andrate 2017
 was almost
 usinglower than 10*10
 the method /s during
 in −5Tang et al.the[24]
 summer
 (othersolstice,
 stationswhereas its
 had similar
 value
results). As was
 shownaboutin 20*10
 Figure−5–30*10−5/s during the winter solstice over the Hong Kong area. That is to say,
 7, the ionospheric dTEC series were detected during all months over the
Hong theKong variations in thewere
 area. There R–T moregrowth ratefeaturing
 days during the two solstices were
 high-amplitude not consistent
 ionospheric dTEC with thatduring
 series of
 plasma bubble occurrence. In addition, ionospheric TEC was also not always larger during the
the summer solstice than the winter solstice. We employed the dTEC series extracted by the three
 summer solstice than during the winter solstice (seen in Figure 6). This suggested that, apart from
stations, HKKT, HKNP, and HKOH, to calculate the propagation direction of GWs using the method
 the R–T growth rate, other factors could also exert effects on the formation of solstitial asymmetry.
in GarrisonResearch
 et al. [28]. The propagation
 showed that atmospheric directions
 GWs (orof ionospheric
 MSTIDs) GWswith
 associated at 17:30–18:30 LT for
 the polarization the two
 electric
 Remote Sens. 2020, 12, x FOR PEER REVIEW 8 of 11
solstices during the years of 2014 and 2017 are shown in Figure 8. This
 fields also played an important role in plasma bubble formation [13–16]. To generate the time interval of ionospheric GWs
was chosen
 had
 due toelectric
 polarization the fact
 frequent thunderstorm
 thatactivity
 fields, plasma
 the phase bubbles generally occurred
 frontMarch–October.
 during (perpendicular to after sunset
 propagation
 Using
 when
 direction)
 the processing method
 the
 of GWs enhanced
 should
 in Tang et
eastwardbe electric
 approximatelyfield destabilized
 aligned with the
 the ionosphere
 geomagnetic [17].
 direction As seen
 [27]. in Figure
 Therefore,
 al. [29], we obtained the average thunderstorm coverage area (grid counts) for different seasons 8,
 zonallythere were
 propagating more
 GWs days
 were
featuring necessary due to the meridional magnetic field
 −90 ◦line. ◦ during
 MSTIDs were signals of atmospheric GWs
 during each year from 2014 to 2017, shown in Figure 9. As seen in Figure 9, thunderstorm activity the
 azimuth of ionospheric GWs around or 90 the summer solstice than
 insolstice
winterduringthe ionosphere,
 theduring
 summer which
 both thecan
 solstice be significantly
 high
 was called activity
 solar as ionospheric
 year ofGWs.
 stronger Here,
 2014during
 than and the we investigated
 thelow solar
 winter the ionospheric
 activity
 solstice. year
 Tang et of
 al.2017.
 GWs
This meant as a
 [29] analyzed proxy
 that zonallyof atmospheric
 the relationship
 propagating GWs.
 between
 GWs thunderstorms
 were in existence and during
 the ionospheric
 the summer GWs solstice,
 over the leading
 Hong to
 Figure
 Kongplasma 7 presents
 area during the ionospheric
 theactivity.
 years of 2014–2017.detrended TEC (dTEC)
 Their research indicated series
 that extracted
 thunderstormby station
 eventsHKKT
 were
enhanced bubble
 during
 the mainthe years
 source ofof 2014especially
 GWs, and 2017during
 using the
 highmethod in Tangdays,
 thunderstorm et al.such
 [24] as
 (other stations solstice.
 the summer had similar
 results). As shown in Figure 7, the ionospheric dTEC series were detected during all months over
 (a)2014
 the Hong Kong area. 6There were more days featuring high-amplitude ionospheric dTEC series
 during the summer solstice than the winter solstice. We employed the dTEC series extracted by the
 3
 three stations, HKKT, HKNP, and HKOH, to calculate the propagation direction of GWs using the
 TECU

 method in Garrison et 0al. [28]. The propagation directions of ionospheric GWs at 17:30–18:30 LT for
 the two solstices during
 -3 the years of 2014 and 2017 are shown in Figure 8. This time interval of
 ionospheric GWs was chosen due to the fact that plasma bubbles generally occurred after sunset
 -6
 when the enhanced eastward electric2 field
 4 destabilized
 6 the
 8 ionosphere
 10 [17].
 12 As seen in Figure 8,
 Month
 there were more days featuring azimuth of ionospheric GWs around −90° or 90° during the summer
 (b)2017
 solstice than the winter6 solstice during both the high solar activity year of 2014 and the low solar
 activity year of 2017. This meant that zonally propagating GWs were in existence during the
 3
 summer solstice, leading to enhanced plasma bubble activity.
 TECU

 The above analysis 0 suggested the GW seeding mechanism was responsible for the formation of

 solstitial asymmetry in-3 plasma bubble occurrence over the Hong Kong area. The source of the GWs
 needs to be further explored. According to the local VLF detection network, the Hong Kong area
 -6
 2 4 6 8 10 12
 Month

 Figure 7. Ionospheric
 Figure detrended
 7. Ionospheric TEC TEC
 detrended (dTEC) series
 (dTEC) extracted
 series byby
 extracted GNSS
 GNSSstation
 stationHKKT
 HKKT during the years
 during the
 of 2014 and 2017. years of 2014 and 2017.

 (a)SS,2014 (b)WS,2014
 100 100

 50 50
 uth(°)

 uth(°)

 0 0

-3

 -6
 2 4 6 8 10 12
 Month

 Figure
Remote Sens. 2020, 7. Ionospheric detrended TEC (dTEC) series extracted by GNSS station HKKT during the
 12, 2423 8 of 10
 years of 2014 and 2017.

 (a)SS,2014 (b)WS,2014
 100 100

 50 50

 Azimuth(°)

 Azimuth(°)
 0 0

 -50 -50

 -100 -100
 0 20 40 60 80 0 20 40 60 80
 Days Days
 (c)SS,2017 (d)WS,2017
 100 100

 50 50
 Azimuth(°)

 Azimuth(°)
 0 0

 -50 -50

 -100 -100
 0 20 40 60 80 0 20 40 60 80
 Days Days

 8. Propagation
 FigureFigure direction
 8. Propagation of ionospheric
 direction of ionosphericgravity
 gravity waves (GWs)atat17:30–18:30
 waves (GWs) 17:30–18:30LT LT
 for for
 the the
 twotwo
 solstices during
 solstices the years
 during of 2014
 the years andand
 of 2014 2017. SSSSrepresents
 2017. represents the
 the summer solsticeand
 summer solstice andWS WS represents
 represents the the
 winterwinter solstice.
 solstice. Here,Here, the scope
 the scope of the
 of the azimuth
 azimuth is is−90 ◦ to
 −90° to 90◦.
 90°.

 The above analysis suggested the GW seeding mechanism was responsible for the formation of
solstitial asymmetry in plasma bubble occurrence over the Hong Kong area. The source of the GWs
needs to be further explored. According to the local VLF detection network, the Hong Kong area had
frequent thunderstorm activity during March–October. Using the processing method in Tang et al. [29],
we obtained the average thunderstorm coverage area (grid counts) for different seasons during each
year from 2014 to 2017, shown in Figure 9. As seen in Figure 9, thunderstorm activity during the
summer solstice was significantly stronger than during the winter solstice. Tang et al. [29] analyzed
the relationship between thunderstorms and the ionospheric GWs over the Hong Kong area during
the years of 2014–2017. Their research indicated that thunderstorm events were the main source of
GWs, especially during
 Remote Sens. 2020, high
 12, x FOR thunderstorm
 PEER REVIEW days, such as the summer solstice. 9 of 11

 Thunderstorm area
 30
 2014 2015 2016 2017

 25

 20
 Counts

 15

 10

 5

 0
 SE SS AE WS
 Season

 9. The9.average
 FigureFigure thunderstorm
 The average thunderstormarea
 areafor
 fordifferent
 different seasons duringeach
 seasons during eachyear
 year from
 from 2014
 2014 to 2017.
 to 2017.
 SE, SS,SE,
 AE,SS,and
 AE, WS
 and represent the the
 WS represent spring equinox,
 spring equinox,summer
 summer solstice, autumnequinox,
 solstice, autumn equinox, andand winter
 winter
 solstice, respectively.
 solstice, respectively.

 According to the
 According toabove analysis,
 the above a conclusion
 analysis, a conclusioncould
 could be
 be drawn thatthe
 drawn that thesolstitial
 solstitial asymmetry
 asymmetry in in
plasmaplasma
 bubblebubble
 occurrence overover
 occurrence the Hong Kong
 the Hong area
 Kong could
 area bebe
 could attributed
 attributedtotothe
 theseeding mechanism of
 seeding mechanism
 of thunderstorm-driven GWs. When the growth rate of R–T instability was small, the GW seeding
 could play a dominant role in the process of plasma bubble generation. This could explain the
 reported exceptions during 2007 and 2008, in which plasma bubble occurrence was greater in the
 summer solstice than in the spring equinox over the Hong Kong area.

 5. Conclusions

Remote Sens. 2020, 12, 2423 9 of 10

thunderstorm-driven GWs. When the growth rate of R–T instability was small, the GW seeding could
play a dominant role in the process of plasma bubble generation. This could explain the reported
exceptions during 2007 and 2008, in which plasma bubble occurrence was greater in the summer
solstice than in the spring equinox over the Hong Kong area.

5. Conclusions
 In this study, the seasonal variations in plasma bubble occurrence over the Hong Kong area
were investigated using the local GNSS network. Furthermore, the causes of seasonal variation
characteristics were also carefully explored. The main results and conclusions are listed as follows:

1. The occurrences of plasma bubble were generally larger in the two equinoxes than in the two
 solstices. During the two equinoxes, the solar terminator aligned with the geomagnetic field,
 leading to enhanced plasma bubble activity.
2. Plasma bubble activity was more frequent in the spring equinox than in the autumn equinox
 (equinoctial asymmetry). The equinoctial asymmetry could be explained by the R–T instability
 mechanism, due to the larger R–T growth rate in the spring equinox than in the autumn equinox.
 Ionospheric TEC (or electron density) was likely an important factor in the equinoctial asymmetry.
3. The plasma bubble occurrence was greater in the summer solstice than in the winter solstice
 (solstitial asymmetry). The R–T instability mechanism was not suitable for the formation of
 solstitial asymmetry, due to the lower R–T growth rate in the summer solstice compared to that
 of the winter solstice. The solstitial asymmetry could be attributed to the seeding mechanism of
 thunderstorm-driven GWs.

 In other words, the seasonal variations in plasma bubble occurrence over the Hong Kong area
depended on combined effects of R–T instability and GW seeding. The reported exceptions during
2001 and 2004 need further research in the future work.

Author Contributions: L.T. contributed to data analysis and led manuscript writing. W.C. supervised this study.
O.-P.L. contributed to manuscript editing. M.C. contributed to data processing. All authors have read and agreed
to the published version of the manuscript.
Funding: This research was funded by the National Natural Science Foundation of China (Grant number 41804021,
41874031) and the Natural Science Foundation of Guangdong Province (Grant number 2017A030310242).
Acknowledgments: The authors thank the SMO in Hong Kong and the CODE for providing GNSS data and GIM
data. The authors also thank the anonymous reviewers for valuable and helpful comments.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Chen, W.; Gao, S.; Hu, C.; Chen, Y.; Ding, X. Effects of ionospheric disturbances on GPS observation in low
 latitude area. GPS Solut. 2008, 12, 33–41. [CrossRef]
2. Kelley, M.C.; Makela, J.J.; de La Beaujardière, O.; Retterer, J. Convective ionospheric storms: A review.
 Rev. Geophys. 2011, 49. [CrossRef]
3. Tsunoda, R.T. Control of the seasonal and longitudinal occurrence of equatorial scintillations by the
 longitudinal gradient in integrated E region Pedersen conductivity. J. Geophys. Res. Atmos. 1985, 90, 447–456.
 [CrossRef]
4. Aarons, J. The longitudinal morphology of equatorial F-layer irregularities relevant to their occurrence.
 Space Sci. Rev. 1993, 63, 209–243. [CrossRef]
5. Sobral, J.; Abdu, M.A.; Takahashi, H.; Taylor, M.J.; De Paula, E.R.; Zamlutti, C.J.; De Aquino, M.G.; Borba, G.L.
 Ionospheric plasma bubble climatology over Brazil based on 22 years (1977–1998) of 630 nm airglow
 observations. J. Atmos. Sol.-Terr. Phys. 2002, 64, 1517–1524. [CrossRef]
6. Burke, W.J.; Su, S.Y.; Gentile, L.C.; Huang, C.Y.; Valladares, C.E. Longitudinal variability of equatorial plasma
 bubbles observed by DMSP and ROCSAT-1. J. Geophys. Res. Space Phys. 2004, 109, 12301. [CrossRef]

Remote Sens. 2020, 12, 2423 10 of 10

7. Makela, J.J.; Ledvina, B.M.; Kelley, M.C.; Kintner, P.M. Analysis of the seasonal variations of equatorial
 plasma bubble occurrence observed from Haleakala, Hawaii. Ann. Geophys. 2004, 22, 3109–3121. [CrossRef]
8. Nishioka, M.; Saito, A.; Tsugawa, T. Occurrence characteristics of plasma bubble derived from global
 ground-based GPS receiver networks. J. Geophys. Res. Space Phys. 2008, 113. [CrossRef]
9. Sultan, P.J. Linear theory and modeling of the Rayleigh-Taylor instability leading to the occurrence of
 equatorial spreadF. J. Geophys. Res. Space Phys. 1996, 101, 26875–26891. [CrossRef]
10. Kil, H.; Demajistre, R.; Paxton, L.J. F-region plasma distribution seen from TIMED/GUVI and its relation to
 the equatorial spread F activity. Geophys. Res. Lett. 2004, 31, 179–211. [CrossRef]
11. Sripathi, S.; Kakad, B.; Bhattacharyya, A. Study of equinoctial asymmetry in the Equatorial Spread F (ESF)
 irregularities over Indian region using multi-instrument observations in the descending phase of solar cycle
 23. J. Geophys. Res. Space Phys. 2011, 116. [CrossRef]
12. Maruyama, T.; Saito, S.; Kawamura, M.; Nozaki, K.; Krall, J.; Huba, J.D. Equinoctial asymmetry of a
 low-latitude ionosphere-thermosphere system and equatorial irregularities: Evidence for meridional wind
 control. Ann. Geophys. 2009, 27, 2027–2034. [CrossRef]
13. Röttger, J. Equatorial spread-F by electric fields and atmospheric gravity waves generated by thunderstorms.
 J. Atmospheric Terr. Phys. 1981, 43, 453–462. [CrossRef]
14. McClure, J.P.; Singh, S.; Bamgboye, D.K.; Johnson, F.S.; Kil, H. Occurrence of equatorial F region irregularities:
 Evidence for topside seeding. J. Geophys. Res. 1998, 103, 29119–29135. [CrossRef]
15. Tsunoda, R.T. On equatorial spreadF: Establishing a seeding hypothesis. J. Geophys. Res. Space Phys. 2010, 115.
 [CrossRef]
16. Takahashi, H.; Wrasse, C.M.; Figueiredo, C.A.O.B.; Barros, D.; Abdu, M.A.; Otsuka, Y.; Shiokawa, K.
 Equatorial plasma bubble seeding by MSTIDs in the ionosphere. Prog. Earth Planet. Sci. 2018, 5, 32.
 [CrossRef]
17. Ji, S.; Chen, W.; Wang, Z.; Xu, Y.; Weng, D.; Wan, J.; Fan, Y.; Huang, B.; Fan, S.; Sun, G. A study of occurrence
 characteristics of plasma bubbles over Hong Kong area. Adv. Space Res. 2013, 52, 194–1958. [CrossRef]
18. Kumar, S.; Chen, W.; Liu, Z.; Ji, S. Effects of solar and geomagnetic activity on the occurrence of equatorial
 plasma bubbles over Hong Kong. J. Geophys. Res. Space Phys. 2016, 121, 9164–9178. [CrossRef]
19. Hong Kong Satellite Reference Network. Available online: Ftp://ftp.geodetic.gov.hk (accessed on 28 July 2020).
20. Space Weather Center of Japan. Available online: http://swc.nict.go.jp (accessed on 28 July 2020).
21. Center for Orbit Determination in Europe. Available online: http://ftp.aiub.unibe.ch/CODE (accessed on 28
 July 2020).
22. Tang, L.; Li, Z.; Zhou, B. Large-area tsunami signatures in ionosphere observed by GPS TEC after the 2011
 Tohoku earthquake. GPS Solut. 2018, 22, 93. [CrossRef]
23. Portillo, A.; Herraiz, M.; Radicella, S.M.; Ciraolo, L. Equatorial plasma bubbles studied using African slant
 total electron content observations. J. Atmos. Sol. Terr. Phys. 2008, 70, 907–917. [CrossRef]
24. Savitzky, A.; Golay, M.J.E. Smoothing and Differentiation of Data by Simplified Least Squares Procedures.
 Anal. Chem. 1964, 36, 1627–1639. [CrossRef]
25. Wu, Q. Longitudinal and seasonal variation of the equatorial flux tube integrated Rayleigh-Taylor instability
 growth rate. J. Geophys. Res. Space Phys. 2015, 120. [CrossRef]
26. Wu, Q. Solar effect on the Rayleigh-Taylor instability growth rate as simulated by the NCAR TIEGCM.
 J. Atmos. Sol.-Terr. Phys. 2017, 156, 97–102. [CrossRef]
27. Huang, C.-S.; Kelley, M.C. Nonlinear evolution of equatorial spread F: 1. On the role of plasma instabilities
 and spatial resonance associated with gravity wave seeding. J. Geophys. Res. Space Phys. 1996, 101, 283–292.
 [CrossRef]
28. Garrison, J.L.; Lee, S.-C.G.; Haase, J.S.; Calais, E. A method for detecting ionospheric disturbances and
 estimating their propagation speed and direction using a large GPS network. Radio Sci. 2007, 42. [CrossRef]
29. Tang, L.; Chen, W.; Chen, M.; Louis, O.P. Statistical Observation of Thunderstorm-Induced Ionospheric
 Gravity Waves above Low-Latitude Areas in the Northern Hemisphere. Remote Sens. 2019, 11, 2732.
 [CrossRef]

 © 2020 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 (http://creativecommons.org/licenses/by/4.0/).