Submitted to Earth, Planets and Space 1

2

Did ionospheric anomalies precede the 2016 April Kumamoto earthquake 3

sequence? 4 5 1Kosuke Heki and 1,2Liming He 6

7

1. Dept. Earth Planet. Sci., Hokkaido University, Sapporo, Japan 8

2. Department of Geodesy and Geomatics, School of Resources and Civil Engineering, Northeastern 9

University, Shenyang, China 10

11

Abstract 12

High seismic intensities of the foreshock (Mw6.2) and the mainshock (Mw7.0) of the 2016 April 13

Kumamoto earthquake sequence brought huge damages. Here we examine if detectable ionospheric 14

anomalies preceded these inland shallow crustal earthquakes. We analyzed changes in ionospheric 15

total electron content (TEC) using Japanese dense network of Global Navigation Satellite System 16

(GNSS) receivers. However, we did not find anomalies of the kind which we often observe before 17

larger earthquakes. For comparison, we analyzed TEC before the Mw7.8 earthquake that occurred in 18

Ecuador, South America, on the next day of the Kumamoto mainshock. We found possible 19

preseismic TEC anomalies with amplitudes and precursor times consistent with past large 20

earthquakes. These results support the empirical relationship that sizes of the preseismic TEC 21

anomalies depend on earthquake Mw and background vertical TEC, but not on maximum seismic 22

intensities. We also found that a stationary linear positive TEC anomaly, with a shape similar to 23

medium-scale traveling ionospheric disturbance, emerged to the northwest of the epicenter 24

immediately before the Kumamoto mainshock. Further studies are needed to conclude its link to the 25

earthquake. 26

27

Keywords: 2016 Kumamoto earthquake, total electron content, ionosphere, MSTID, GNSS 28

29

1. Introduction 30

Ionospheric Total Electron Content (TEC) can be derived by comparing phases of two L band 31

microwave signals from Global Navigation Satellite System (GNSS) satellites, such as Global 32

Positioning System (GPS). Heki (2011) found ionospheric electron increase above the epicenter of 33

the 2011 Tohoku-oki earthquake (Mw 9.0), Japan, starting ~40 minutes prior to the earthquake using 34

data from GEONET (GNSS Earth Observation Network), a dense array of continuous GNSS stations 35

in Japan. This paper was followed by publications of critical papers (Kamogawa and Kakinami, 36

2013; Utada and Shimizu, 2014; Masci et al, 2015) and immediate rebuttals to them (Heki and 37

Enomoto, 2013; 2014; 2015). 38

Heki and Enomoto (2015) demonstrated that similar electron increases preceded eight past 39

earthquakes with Mw 8.2 or more. They also reported that the anomalies started about 20/40 minutes 40

before Mw 8/9 earthquakes, and that the changes in vertical TEC (VTEC) rate depended on Mw as 41

well as background absolute VTEC. Similar changes often occur due to geomagnetic activities, but 42

Heki and Enomoto (2015) demonstrated that they are not frequent enough to account for the 43

observed preseismic anomalies. The latest paper (He and Heki, 2016) revealed that both TEC 44

decrease and increase emerge simultaneously in different regions. Their 3-D structure showed that 45

positive and negative electron density anomalies line up from the epicenter along the geomagnetic 46

field. Such geometry resembles to the ionospheric response to positive surface charges as Kuo et al. 47

(2014) showed using numerical simulation. 48

The foreshock (Mw6.2) and the mainshock (Mw7.0) of the 2016 Kumamoto earthquake sequence 49

occurred on April 14 12:26 UT, and April 15 16:25 UT, with the temporal separation of ~28 hours. 50

Their shallow epicentral depths caused high seismic intensities (7 in the Japan Meteorology Agency 51

scale), and caused lots of damages to buildings and roads. Indeed, the peak seismic intensities of the 52

two earthquakes were comparable to the Mw9.0 2011 Tohoku-oki earthquake, the largest recorded 53

earthquake in Japan. According to Heki and Enomoto (2015), no preseismic TEC changes should 54

have preceded such earthquakes with Mw≤7. However, the series of study relied on cases of 55

interplate thrust earthquakes in deep-sea trenches, and its validity for shallow inland earthquakes 56

needs confirmation. Here we look for TEC anomalies immediately before these earthquakes using 57

GEONET data. We also examine, for comparison, the existence of preseismic TEC anomalies of the 58

2016 April Ecuador earthquake (Mw7.8), which occurred on the next day of the Kumamoto 59

mainshock. 60

61

2. Data and Results 62

2-1. Two Kumamoto earthquakes 63

We converted the raw data from ~1200 ground stations of GEONET to slant TEC (STEC). We 64

extracted two hour potions of the STEC time series including the two earthquakes. We selected data 65

with two GPS satellites closest to the local zenith in Kyushu. To highlight the difference from the 66

2011 Tohoku-oki earthquake case, we employed the same procedure as Heki (2011), i.e. we assumed 67

that the temporal change of the vertical TEC (VTEC) obeys a quadratic polynomial of time, and 68

defined the departure from the reference as the anomaly. In determining the reference curves, we 69

normally exclude a certain time window. Heki and Enomoto (2015) showed that preseismic 70

anomalies do not start earlier than 20 minutes before interplate events with Mw8 or less. Here we set 71

up somewhat longer time windows, ranging from -30 minutes to +20 minutes of the earthquakes. 72

The VTEC anomalies at three epochs, 20 minutes, 10 minutes, and immediately before earthquakes 73

are shown in Fig. 1. He and Heki (2016) showed that positive TEC anomalies appear at altitudes 74

~200 km in the area at the equator side (south in the northern hemisphere) of the epicenter, but we do 75

not see them in Fig.1. In spite of similar maximum seismic intensities, the Kumamoto earthquakes 76

do not show anomalies comparable to those before the 2011 Tohoku-oki earthquake. In Fig.1f, one 77

can recognize weak positive anomalies running NNW-SSE. This will be discussed later in Section 78

3.2. 79

80

81

Fig. 1. (a-c) VTEC anomalies at 20 minutes (a), 10 minutes (b) and immediately (c) before the 82

2016 Kumamoto foreshock. We used GPS satellites 1 (circle) and 11 (triangle). The black star 83

represents the epicenter. (d-f) VTEC anomalies before the Kumamoto mainshock drawn using 84

GPS satellites 19 (circle) and 6 (triangle). We used the same color scheme as the Fig. 3 in Heki 85

(2011) for the 2011 Tohoku-oki earthquake. 86

87

Fig. 2 shows the STEC time series at five stations in Kyushu for the two Kumamoto earthquakes. 88

There we selected stations with SIP located to the south of the epicenters. Although the reference 89

curves were estimated to fit the 50 minutes excluding the time interval shown with red bars at the top, 90

they overlap with the observed data throughout the interval, suggesting small anomalous behaviors 91

of the ionospheric electrons before and after the earthquakes. From Figs. 1 and 2, we conclude that 92

the 2016 April Kumamoto earthquake sequence were not associated with ionospheric TEC 93

anomalies like those shown for larger earthquakes in Heki and Enomoto (2015). We found a small 94

amplitude linear-shaped anomaly to the north of the epicenter, which will be discussed in Section 95

3.2. 96

97

98

99

Fig. 2. Slant TEC changes over two hour intervals including the foreshock (a) and the 100

mainshock (b) of the Kumamoto earthquakes, drawn using GPS satellites 1 and 19, respectively. 101

The maps show the epicenters (black stars), GNSS stations (red circles), and sub-ionospheric 102

point (SIP) trajectories calculated assuming thin ionosphere at altitude of 300 km. The observed 103

STEC changes were modeled assuming that VTEC changes obey quadratic functions of time. In 104

estimating the reference curves, we excluded the time shown with the red bars at the top, i.e. 105

from 30 minutes before earthquake to 20 minutes after earthquake. 106

107

2-2. The Ecuador earthquake 108

The Mw7.8 2016 Ecuador earthquake occurred at 23:58 UT, April 16, ~31.5 hours after the 109

Kumamoto mainshock, as an interplate thrust earthquake between the Nazca and the South American 110

Plates. The epicenter lies beneath the Pacific coast of northern Ecuador, at the depth ~20 km. This 111

earthquake does not have any causal relationship with the Kumamoto earthquakes. Nevertheless, it 112

would be meaningful to compare their preseismic ionospheric signals to understand the difference 113

coming from the two factors, Mw and background VTEC. We downloaded the GNSS raw data taken 114

at two stations in Ecuador, QUI4 (Quito) and RIOP (Riobamba), from the UNAVCO data archive 115

(www.unavco.org/data/). We then converted STEC to VTEC following He and Heki (2016). 116

Because of the small number of available GNSS stations, it is not appropriate to draw figures like 117

Fig.1. In Fig.3, we plot the VTEC time series. They show overall decrease in this period because the 118

time window includes the local sunset. We obtained the reference curves using polynomial of time 119

with degrees-8 excluding the time interval within ±20 minutes of the earthquake (red line at the top 120

of Fig.3). For a relatively small earthquake like this, it is difficult to constrain the positive bending of 121

VTEC using Akaike’s Information Criterion as done for larger earthquake in Heki and Enomoto 122

(2015). Nevertheless, the observed VTEC curves show clear positive anomalies starting 15-20 123

minutes before the earthquake, demonstrating clear differences from the Kumamoto cases (Fig.2). 124

125

126

127

Fig. 3. VTEC time series of three station-satellite pairs recording possible preseismic TEC 128

increases of the 2016 Ecuador earthquake. Red horizontal bars indicate the excluded period in 129

defining the reference curves. The map shows the SIP trajectories calculated assuming 300 km 130

for ionospheric height. Along the trajectories, we show hourly time marks (small circles), the 131

earthquake occurrence times (squares), and the precursor starting times (triangles). The yellow 132

star shows the epicenter. Because the Sat.7-Quito curve overlaps with the Sat.30-Quito curve, we 133

moved the former slightly upward for visual clarity. 134

135

3. Discussions and conclusion 136

3-1. Preseismic TEC anomalies 137

Fig. 4a shows that the geomagnetic activity was moderately high (Dst drops ~50 nT) at the 138

occurrence times of all the three earthquakes. This might be responsible for small scale undulations 139

as seen in the TEC time series in Fig. 2, but they did not cause serious departure from reference 140

curves within the studied time windows. The three earthquakes occurred at around 21:25, 01:27 and 141

18:58 in local times. Fig.4 compares the distribution of VTEC drawn using global ionospheric maps 142

(GIM) downloaded from University of Bern (ftp.unibe.ch/aiub/). We can see that VTEC above the 143

epicenter of the Ecuador earthquake were more than twice as high as the two Kumamoto earthquake 144

cases. 145

146

147 Fig. 4. Change of the Dst index (omniweb.gsfc.nasa.gov/form/dx1.html) over the 20 days 148

period including the three earthquakes (three red vertical lines) discussed here (a). Global 149

Ionospheric Map (GIM) at the time of the foreshock (b) and the mainshock (c) of the 2016 150

Kumamoto earthquake, and the 2016 Ecuador earthquake (c). The VTEC values at the epicenters 151

(white stars) are, ~15, ~10, and ~36 TECU, respectively. 152

153

Heki and Enomoto (2015) proposed the empirical relationship among the three quantities, 154

preseismic VTEC rate changes, earthquake Mw, and background VTEC. It suggests that a larger 155

VTEC rate change tends to occur before a larger earthquake and under a larger background VTEC. 156

Fig.5 is a leftward-expanded version of Fig. 4a in Heki and Enomoto (2015). It seems obvious that 157

preseismic TEC signals are not to be seen before the 2016 Kumamoto earthquakes, whose Mw are 158

only 6.2 and 7.0 (shown in red + in Fig.5). The lack of signals before these two earthquakes suggests 159

that even shallow inland crustal earthquakes do not cause preseismic ionospheric signals much larger 160

than expected from this diagram. 161

On the other hand, positive VTEC rate changes of a little less than 2 TECU/h is expected to occur 162

before the Mw7.8 Ecuador earthquake (see contours in Fig.5). The actually observed rate change is 163

1.75 TECU/h (VTEC: 26.6 TECU) from the RIOP-Sat.30 pair (red curve in Fig.3a), which is 164

consistent with this diagram. The anomaly started ~17 minutes before the earthquake, and this is also 165

consistent with past earthquakes, shown in Fig.5a of Heki and Enomoto (2015). 166

167 168

Fig. 5. Diagram showing the dependence of the preseismic VTEC rate change (size of the 169

circle) on the background VTEC (vertical axis) and the earthquake Mw (horizontal axis), similar 170

to Fig.4a of Heki and Enomoto (2015). Riobamba-Sat.30 pair was used in adding the 2016 171

Ecuador earthquake case (small red circle in the middle). Three black crosses indicate 172

earthquakes with no significant preseismic VTEC changes detected, already reported in Heki 173

(2011). For the two Kumamoto earthquakes, we did not find precursory VTEC changes, which 174

are shown with two additional crosses in red. The rate change (TECU/h) is modeled as 3.75 Mw 175

+ 0.11 VTEC −30.6, and the contour lines show the same rate changes of 2, 4, 6, 8, and 10 176

TECU/h. Dashed parts of the contours are not substantiated by data. 177

178

3-2. MSTID-like anomaly before the mainshock 179

TEC anomalies were not clear before the Kumamoto mainshock as seen in Fig.1d-f. However, by 180

isolating GPS satellite 6 data and replacing the color scheme, a weak (up to ~1 TECU) linear-shaped 181

positive TEC anomaly emerge to extend north-northwestward from the mainshock epicenter. The six 182

panels in Fig. 6a show snapshots taken every five minutes. This anomaly starts to grow ~25 minutes 183

before the earthquake, and decay in ~10 minutes after the earthquake. 184

Its NNW-SSE elongation and relatively short wavelength (~100 km) resemble to a typical 185

night-time Medium-Scale Traveling Ionospheric Disturbance (MSTID) that occur frequently during 186

summer nights (Otsuka et al., 2011). In fact, a typical night-time MSTID occurred about two days 187

before the mainshock (April 13) (Fig.6b). There is, however, a peculiar feature in the MSTID-like 188

anomaly on April 15, i.e. it is stationary in space. The difference is clear by comparing Fig.6a and 6b, 189

i.e. the one in Fig.6b propagates southwestward by ~200 m/s. The velocity of night-time MSTIDs in 190

Japan falls within 80-200 m/s (Otsuka et al., 2011), and the stagnant anomaly in Fig.6a is quite 191

exceptional for MSTID. 192

Both Night-time MSTID and preseismic TEC anomalies are considered to develop by electric 193

fields in the ionosphere. Night-time MSTID starts to develop due to polarization electric field and 194

grow with time through the Perkins instability, although it is not well understood what process 195

governs its southwestward propagation velocity (Garcia et al., 2000). On the other hand, He and 196

Heki (2016) suggested that the preseismic ionospheric anomalies could be a response to electric 197

fields made by positive charges on the surface above the epicenter. In the present case, the surface 198

electric charge before the Kumamoto mainshock may not be large enough to cause ionospheric 199

electron redistribution as seen before larger earthquakes by Heki and Enomoto (2015). However, the 200

charge might be large enough to become a seed of MSTID and make it stagnant above the epicenter. 201

Anyway, we need more observations to clarify the link. By the way, the largest “aftershock” of the 202

foreshock (Mw6.0) occurred at 15:03UT on the day of the foreshock, but we observed only typical 203

night-time MSTID propagating southwestward at that time. 204

205

3-3. Conclusions 206

We studied ionospheric anomalies immediately before the 2016 April Kumamoto earthquakes, 207

with seismic intensity comparable to the 2011 Tohoku-oki earthquake. The results can be concluded 208

as follows; 209

1) No preseismic TEC anomalies, similar to the 2011 Tohoku-oki earthquake case, were found for 210

the foreshock and the mainshock of the Kumamoto earthquake sequence. 211

2) Possible small preseismic TEC anomaly, consistent with Mw and background VTEC, was seen 212

before the Ecuador earthquake that occurred on the next day of the Kumamoto mainshock. 213

3) Existence and size of the TEC signals prior to these three earthquakes are consistent with Heki 214

and Enomoto (2015), the anomalies depend on Mw and background VTEC rather than seismic 215

intensity. 216

4) Stationary MSTID-like disturbance appeared shortly before the Kumamoto mainshock, but further 217

studies are needed to understand its link with the earthquake. 218

219

220

221

Fig. 6. Comparison of the development and movement of MSTID that appeared on April 15 222

shortly before the Kumamoto earthquake mainshock (a), and on April 13, two days before that 223

(b), shown by five-minute snap shots with GPS satellite 6 over 35 minutes interval. We drew 224

five gray lines with 100 km separation to visualize their propagation. Typical night-time 225

MSTID show southwestward movements like the example in (b). In (a), the positive crest of 226

MSTID is stagnant above the mainshock epicenter (black star). 227

228

Acknowledgements 229

We thank referees for constructive comments. The GEONET data were downloaded from the GSI 230

website (terras.gsi.go.jp). 231

232

Competing Interests 233

Here we declare that none of the authors do not have any competing interests in the manuscript. 234

235

Author Contributions 236

KH carried out the analysis of TEC before and after the Kumamoto and Ecuador earthquakes. LH 237

analyzed the night-time MSTID on non-earthquake days. Both read and approved the final 238

manuscript. 239

240

References 241

Garcia, F.J., M. C. Kelley, J. J. Makela, and C.–S. Huang (2000), Airglow observations of mesoscale 242

low-velocity traveling ionospheric disturbances at midlatitudes, J. Geophys. Res. Space Phys., 105, 243

18407-18415, doi:10.1029/1999JA000305. 244

He, L. and K. Heki (2016), Three-dimensional distribution of ionospheric anomalies prior to three 245

large earthquakes in Chile, Geophys. Res. Lett., doi:10.1002/2016GL069863, in press. 246

Heki, K. (2011), Ionospheric electron enhancement preceding the 2011 Tohoku-Oki earthquake, 247

Geophys. Res. Lett., 38, L17312, doi:10.1029/2011GL047908. 248

Heki, K., and Y. Enomoto (2013), Preseismic ionospheric electron enhancements revisited, J. 249

Geophys. Res. Space Phys., 118, 6618-6626, doi:10.1002/jgra.50578. 250

Heki, K. and Y. Enomoto (2014), Reply to comment by K. Heki and Y. Enomoto on "Preseismic 251

ionospheric electron enhancements revisited", J. Geophys. Res. Space Phys., 119, 252

doi:10.1002/2014JA020223. 253

Heki, K., and Y. Enomoto (2015), Mw dependence of the preseismic ionospheric electron 254

enhancements, J. Geophys. Res. Space Phys., 120, 7006-7020, doi:10.1002/2015JA021353 255

Kamogawa, M. and Y. Kakinami (2013), Is an ionospheric electron enhancement preceding the 2011 256

Tohoku-oki earthquake a precursor?, J. Geophys. Res. Space Phys., 118, 1-4, 257

doi:10.1002/jgra.50118. 258

Kuo, C. L., L. C. Lee, and J. D. Huba (2014), An improved coupling model for the 259

lithosphere-atmosphere-ionosphere system, J. Geophys. Res. Space Phys., 119, 3189-3205, 260

doi:10.1002/2013JA019392. 261

Masci, F., J. N. Thomas, F. Villani, J. A. Secan, and N. Rivera (2015), On the onset of ionospheric 262

precursors 40 min before strong earthquakes, J. Geophys. Res. Space Phys. 120, 1383-1393, 263

doi:10.1002/2014JA020822. 264

Otsuka, Y., N. Kotake, K. Shiokawa, T. Ogawa, T. Tsugawa, and A. Saito (2011), Statistical study of 265

Medium-Scale Traveling Ionospheric Disturbances observed with a GPS receiver network in 266

Japan, Aeronomy of the Earth’s Atmosphere and Ionosphere, IAGA Special Sopron Book Series, 267

Vol. 2, Part 3, 291–299, doi:10.1007/978-94-007-0326-1 21. 268

Utada, H., and H. Shimizu (2014), Comment on “Preseismic ionospheric electron enhancements 269

revisited” by K. Heki and Y. Enomoto, J. Geophys. Res. Space Phys., 119, 270

doi:10.1002/2014JA020044. 271