INVESTIGATION OF IONOSPHERIC PRECURSORS

OF EARTHQUAKES IN ROMANIA

USING THE ROMANIAN GNSS/GPS NETWORK

EDUARD ILIE NASTASE1, CHRISTINA OIKONOMOU2,

DRAGOS TOMA-DANILA1, HARIS HARALAMBOUS2,

ALEXANDRA MUNTEAN1, IREN ADELINA MOLDOVAN1

1National Institute for Earth Physics, PO BOX MG2, 077125, Magurele, Romania,

E-mails: eduard_nastase@infp.ro; toma@infp.ro;

muntean@infp.ro; irenutza_67@yahoo.com 2Frederick Research Center, Filokyprou St.7, Palouriotisa, Nicosia, 1036, Cyprus,

E-mails: res.ec@frederick.ac.cy; eng.hh@fit.ac.cy

Received August 31, 2015

We examine the lithosphere-atmosphere-ionosphere interaction with respect to

earthquake events using Total Electron Content (TEC) data deriving from the

Romanian permanent GPS network by applying three different techniques:

a) estimation of TEC deviations from the mean state, b) Cross-Correlation Analysis

and c) Spectral Analysis. The analysis concerns four seismic events that took place in

Romania with magnitude ranging from 5.2 to 6.0. The aim is to identify and study

possible ionospheric precursory phenomena linked to these seismic events.

Key words: Ionospheric earthquake precursors, Total Electron Content (TEC),

Romanian GNSS/GPS network.

1. INTRODUCTION

The last two decades a hard effort is put to analyze and possibly predict

seismic events through monitoring of ionosphere. That was possible mostly with

the aid of technological developments such as ground based techniques to study the

subionospheric propagation of VLF/LF radio waves [1] and the bottomside and the

middle ionosphere with ionosondes, satellite based instruments to carry out

investigations of the topside ionosphere. To complement these techniques, dense

networks of GNSS receivers enabled monitoring of the full extent of the

ionospheric plasma via Total Electron Content (TEC) and ionospheric tomography.

Lately, the link between earthquakes and preseismic ionospheric

perturbations has acquired significant attention. Possible earthquake precursors

have been identified including ground deformations, radon/helium emissions,

crustal stress, atmospheric thermal anomalies [2].

Rom. Journ. Phys., Vol. 61, No. 7–8, P. 1426–1436, 2016

2 Investigation of ionospheric precursors of earthquakes using GNSS/GPS technology 1427

Gas emissions occurring prior to an earthquake cause the ionisation of the neutral atmosphere above the epicenter and thus the generation of anomalous electric field that penetrates the ionosphere leading to large-scale positive and negative anomalies of electron concentration in the vicinity of the epicenter. This electric field is not restricted only to the epicenter, but covers an area that is a function of the earthquake magnitude named as earthquake preparation zone [3]. In addition, Atmospheric Gravity Waves (AGW) are generated by pre-seismic activity of emanating gases. The most widely known long wavelength perturbations are travelling ionospheric disturbances (TIDs) which are associated to AGW and also infrasonic waves propagating upwards, amplified by the exponential density decrease of the atmosphere [4]. It has been statistically proved that ionospheric precursors are observed between 12 days to a few hours prior to the earthquake and that earthquakes should exceed the magnitude of 5 in order to provoke ionospheric disturbances. The duration of a seismically induced ionospheric deviation is short about 4–6 hours being compared with perturbations caused by geomagnetic storms and anomalies [2].

The objective of this study is to investigate possible pre-earthquake ionospheric anomalies that occurred prior to four earthquakes in Romania with magnitude Mw ranging from 5.2 to 6.0, following a multi-technique approach, by using TEC data obtained from the ground-based Romanian GNSS/GPS receiver network.

2. DATA AND METHODOLOGY

The four seismic events selected in this study as well as their main characteristics are presented at Table 1. In order to detect possible ionospheric disturbances prior to earthquakes, TEC data were utilized as obtained from dual-frequency phase and code measurements made by the GPS receivers that are located in Romania and belong to the Romanian permanent GPS network which is operating since 2001. Figure 1 shows the map of the GPS network along with the earthquake epicenters and the preparation zone of each seismic event. In order to calculate the vertical TEC (vTEC) we have processed RINEX files from the selected GPS stations with a calibration algorithm [5]. This processing technique assumes ionospheric thin shell model (located at 350km of altitude) to obtain vTEC from slant total electron content (sTEC) at the Ionospheric Pierce Point (IPP). The elevation angle used here was ≥ 67°.

First we tried to identify possible significant disturbances of TEC data prior to earthquakes using the statistical envelope method: Under the assumption of normal distribution with mean m and standard deviation σ of TECs the expected values of upper bound and lower bound of envelope are m ± 1.34σ. If the observed TEC falls out of either the associated lower or upper bounds of such an envelope then an abnormal signal is detected at the confidence level of about 82% [6].

1428 Eduard Ilie Nastase et al. 3

Table 1

List of the four seismic events in Romania studied here and their characteristics

Seismic

events

No.

Year Month Day

Latitude

North

(°)

Longitude

East

(°)

Magnitude

Mw

Depth

(Km)

Preparation

Area Radius

(Km)

1. 2014 11 22 45.87 27.15 5.7 39 282.488

2. 2013 10 6 45.67 26.58 5.2 135.1 172.1869

3. 2005 5 14 45.64 26.53 5.5 148.5 231.7395

4. 2004 10 27 45.84 26.63 6 105.4 380.1894

Fig. 1 – Map of Romania showing: a) the preparation areas and epicenters (starts) of the 4 examined

earthquakes, b) the Romanian GPS stations network. Pink colored boxes denoted GPS stations used

for the 5.2 Mw event, red colored boxes show stations used for the 5.5 Mw event, dark red colored

boxes correspond to the 5.7 event, and black colored boxes to the 6 Mw event. With grey colored

boxes are noted the stations within the actual GPS network while with white color boxes the stations

which are no longer in use are indicated. The GPS stations OROS and SOFI shown in the map belong

to the EUREF network.

4 Investigation of ionospheric precursors of earthquakes using GNSS/GPS technology 1429

Then, we applied the cross-correlation method on TEC values using one

measurement point located inside the earthquake preparation area and one or more

points located outside it. The preparation area is the area where the ionosphere

above it is affected by earthquake precursors and is defined as a circle with radius

ρ = 100.43Μ

km where M is the magnitude (Mw) of the earthquake [3]. Since

ionospheric variability induced by seismic activity is generally lower than those

variations related to geomagnetic storms, it is shadowed by storm-time variations.

In order to surpass this problem the two measurement points should be in the same

or very close geomagnetic latitude so as ionospheric variations registered at both

points to show similar behavior during quiet and disturbed geomagnetic conditions.

Therefore, their correlation coefficient will be high. On the opposite, during

seismic activity their correlation coefficient is expected to drop since the station

closer to epicenter (inside the preparation area) will be more sensitive to seismic

variations than the station outside the preparation area. In addition, the longitude of

the two sites should not be so different, as ionosphere is local time dependent.

Finally, we performed the Spectral Analysis using differential TEC data,

defined as the difference of sTEC measurement between two successive satellite

epochs. The period and amplitude of differential TEC fluctuations were estimated.

3. RESULTS AND DISCUSSION

In Fig. 2 (a, b, c and d) the diurnal TEC variations at the GPS stations during

the interval 30 days prior, during and up to 7 days after the earthquake are depicted

for all examined seismic events. In the same figures the geomagnetic conditions

during each studying period are described through the daily variations of the

geomagnetic index Dst which comprises a measure of the variation in the

geomagnetic field due to the equatorial ring current. A Dst index of –50 or deeper

indicates a geomagnetic storm-level disturbance.

As it can be detected from the TEC time series within 25 days before the

Mw = 5.7 crustal earthquake on 22 November 2014, positive anomalies occurred at

around 10 hours before the earthquake only at the GPS stations located inside the

preparation area and closer to the epicenter (VRAP, TINA, MARE, BICA, ROSU)

(Figs. 2d and 3). These anomalies have short duration of about 3 hours (7:00–10:00

UT) which is in consistence with [2] who state that pre-seismic variations are

comparatively short in time (around 4–6 hrs) relative to magnetic storm effects. On

the opposite, positive or negative anomalies of TEC variations were not found at

the GPS stations located outside the earthquake preparation zone (ROSP, BUZE)

during the same time. Therefore, this positive anomaly observed on 22 November

2014 can be possibly associated to the impending earthquake, taking also under

consideration that quiet geomagnetic conditions were prevailing during that day.

1430 Eduard Ilie Nastase et al. 5

Figure 2c shows the diurnal TEC variations at the Romanian GPS stations

during the interval 23 days before and at the earthquake day on 6 October 2013. As

it can be seen, TEC values are reduced compared to the mean values three days

before the earthquake in all stations. This is attributed to the occurrence of a major

geomagnetic storm which commenced at 2 October 01:55 UT, had its initial and

main phase during the same day and its recovery phase during the 3-day period

prior to the earthquake (2–5 October). Since ionospheric variability induced by

seismic activity is generally lower than those variations related to geomagnetic

storms, it is most probable that any TEC anomaly related to the earthquake during

the storm period 2–5 October is shadowed by storm-time variations.

In Figure 2b the daily TEC variations within one month before the Mw = 5.5

deep earthquake on 14 May 2005, as well as the daily Dst geomagnetic index

variations are presented. As it can be observed from the Dst index values a major

geomagnetic storm occurred 6 days before the earthquake. The initial phase of the

storm started at 19:20 UT on 7 May and was followed by its main phase that took

place during 8 May and its recovery phase which lasted from 9 to 13 May, causing

long-lasting high TEC anomalies observed at all GPS stations. Thus, similar to the

previous seismic event no earthquake induced TEC anomaly could be detected.

Inspection of the TEC variations at the GPS stations during the interval

30 days prior to the deep Mw = 6 earthquake on 27 October 2004 reveals a high

positive TEC anomaly occurring on daily basis at the last 6 days before the

earthquake which lasts more than 6 hours (Fig. 2a). Though no geomagnetic storm

occurred during that period, a sequence of solar C- and M-flares was observed

(ftp://ftp.swpc.noaa.gov/pub/warehouse/2004/, NOAA, Space Weather Prediction

Center) which could be partially responsible for this daily increase of TEC values

[7]. Therefore, it was not possible to detect any earthquake induced TEC deviation.

Another possible cause for this could be that fact that this earthquake, though it is

of high magnitude, it has a large depth. According to [8] and [9] who have

statistically studied pre-earthquake ionospheric anomalies using different methods

the anomalies occurring before deep focus earthquakes have smaller intensity and

occurrence rates.

It has become obvious that TEC deviations from the mean state that could be

linked to earthquake were identified only for the crustal Mw = 5.7 earthquake on

22 November 2014 around 10 hours before the earthquake during also

geomagnetically quiet conditions, unlike the other seismic events which were

either deep (27 October 2004) or had low magnitude and their preparatory period

occurred during the development of major geomagnetic storms (6 October 2013, 14

May 2005). For this purpose and in order to obtain additional confidence and

confirmation about the association of the observed TEC anomalies on 22

November 2014 with the Mw = 5.7 seismic event we applied two additional

methods, the Cross-Correlation and the Spectral Analysis.

6 Investigation of ionospheric precursors of earthquakes using GNSS/GPS technology 1431

a.

b.

Fig. 2

1432 Eduard Ilie Nastase et al. 7

c.

d.

Fig. 2 (continued) – Diurnal vTEC variations (red) and corresponding upper and lower bounds (black)

fixed at m ± 1.34σ are depicted for selected GPS stations for 30 days before, during and up to 7 days

after the four examined seismic events. Pink shaded areas show the geomagnetically disturbed periods

and yellow vertical line shows the moment of earthquake.

8 Investigation of ionospheric precursors of earthquakes using GNSS/GPS technology 1433

Fig. 3 – Same as Fig. 2 but for 3 days before and during the earthquake on 22 Nov. 2014.

1434 Eduard Ilie Nastase et al. 9

First we have correlated TEC values recorded at the closest to the earthquake

station (VRAP) during the period 20–22 November 2014 with all the remaining

stations. The correlation coefficient was calculated for each of these three days for

four separate intervals: 00–24, 00–08, 09–16 and 17–24 hours (Table 2). As it can

be seen, though the correlation coefficient values at entire day level are high and do

not show disruption patterns, the analysis for the 8 hours intervals (especially for

the 17–24 h UT, which is during night time considering the local time) reveals low

correlations, mostly in the day before the earthquake (21 Nov. 2014). Since the

anomalies that we seek to find for a medium magnitude earthquake can't be

considerable and can be influenced by the diurnal variability, we aim to believe

that the night time low correlations that have been found could be somehow

connected to the seismic activity within the epicentral area. Indeed, correlations in

all stations except TINA station show in general a decrease of the correlation with

the distance from the epicenter. The fact that TINA station which is closest to

VRAP station shows the lowest correlation coefficient values instead of the highest

as it is expected is most probably due to technical problems at the station that time

that led to incorrect TEC values.

Table 2

Correlation coefficient values between VRAP station and remaining stations ordered by distance from

VRAP, during four time intervals: 00–24, 00–08, 09–16 and 17–24 hours for the 5.7 Mw earthquake

on 22 November 2014, 19:14:17 UT

GPS

STATIONS

24h correl. with

VRAP

0-8h correl. with

VRAP

9-16h correl. with

VRAP

17-24h correl. with

VRAP

20

Nov.

21

Nov.

22

Nov.

20

Nov.

21

Nov.

22

Nov.

20

Nov.

21

Nov.

22

Nov.

20

Nov.

21

Nov.

22

Nov.

TINA 0.997 0.995 0.997 0.998 0.997 0.998 0.995 0.995 0.998 0.691 0.360 0.942

ROSU 0.999 0.999 0.999 0.999 0.999 0.998 0.999 0.998 1.000 0.972 0.920 0.962

TUDO 0.992 0.975 0.999 0.992 0.997 0.999 0.972 0.898 0.999 0.887 0.854 0.946

BUCE 0.999 0.999 0.997 0.998 0.999 0.996 0.998 0.998 0.999 0.962 0.863 0.956

BICA 0.999 0.999 0.999 0.996 0.996 0.998 0.999 0.998 1.000 0.981 0.873 0.981

MARE 0.998 0.998 0.997 0.998 0.995 0.997 0.995 0.993 0.998 0.946 0.875 0.957

COTI 0.997 0.998 0.996 0.996 0.999 0.994 0.993 0.995 0.997 0.956 0.862 0.963

HIST 0.998 0.995 0.998 0.998 0.994 0.997 0.997 0.992 0.998 0.981 0.929 0.959

MALI 0.997 0.995 0.996 0.997 0.994 0.998 0.994 0.987 0.997 0.972 0.777 0.832

ROSP 0.997 0.996 0.997 0.991 0.994 0.993 0.997 0.992 0.996 0.956 0.919 0.967

BUZE 0.992 0.995 0.990 0.984 0.992 0.982 0.977 0.992 0.990 0.943 0.879 0.965

Then, the Spectral Analysis was performed using differential sTEC values

estimated for the propagation path between the close to the epicenter GPS receiver

10 Investigation of ionospheric precursors of earthquakes using GNSS/GPS technology 1435

stations BUCU and BACA and several satellites passing close and above the

earthquake preparation area during the day where large TEC disturbances have

been detected (22 Nov. 2014) (Figs. 2d and 3). In Fig. 4 high periodic sinusoidal

sTEC fluctuations can be identified at the same time that TEC anomalies on the 5.7

event have been noted (7–10 UT) for all satellites, having period of oscillations

around 10–20 minutes. These results confirm that the observed TEC disturbances

in the case of the 5.7 Mw seismic event can be produced by the penetration of

Atmosphere Gravity Waves (AGW) into the ionosphere as described by the

Lithosphere-Atmosphere-Ionosphere coupling theory. Other authors have also

observed preseismic VLF/LF radio signals with propagation path passing close

enough to the epicenters to have similar periods of about 10–20 min [10, 11].

Fig. 4 – Fluctuations of differential TEC obtained from measurements of 7satellites passing over the

area of interest at times of the highest TEC anomalies observed at 22 Nov. 2014 at GPS stations close

to epicenter. The power spectrum of the normalized amplitude is also shown. Map shows the number

and position of satellites IPP (blue asterisks), the position of the GPS receiver station BUCU (pink

triangle) and the earthquake epicenter (green asterisk).

4. CONCLUSIONS

The analysis of four seismic events in Romania with magnitude Mw 5.2 to 6.0

by utilizing TEC data obtained from the Romanian permanent GPS station network

has shown that ionospheric precursory phenomena can be observed one day up to

few hours prior to the crustal Mw = 5.7 earthquake and to closest to the epicenter

1436 Eduard Ilie Nastase et al. 11

stations, whereas preseismic TEC anomalies were not identified in case of deep or

of low magnitude earthquakes and during the occurrence of major geomagnetic

storms. In addition, this study demonstrates that in order to increase the credibility

on the presence of ionospheric precursory phenomena associated with an

earthquake and to provide more safe conclusions, it is of high importance to

simultaneously apply different techniques such as the Cross-Correlation Analysis

and the Spectral Analysis. The Romanian GPS stations network has proven a

useful tool for the investigation of ionospheric precursory phenomena related to

earthquakes.

Acknowledgements. This paper was partially funded by grants of the Romanian National

Authority for Scientific Research: (i) Capacity Program, Module III – Projects supporting Romania's

participation in international research projects, Bilateral cooperation programs Romania – Cyprus,

2014–2015, project number 759/2014, and (ii) Nucleu Program PN 09 30/2009 and the project

Investigation of earthquake signatures on the ionosphere over Europe-ΔΙΑΚΡΑΤΙΚΕΣ/ΚΥ-

ΡΟΥ/0713/37which is co-financed by the Republic of Cyprus and the European Regional

Development Fund (through the ΔΕΣΜΗ 2009–2010 of the Cyprus Research Promotion Foundation).

REFERENCES

1. I.A. Moldovan, A.P. Constantin, P.F. Biagi, D. Toma Danila, A.S. Moldovan, P. Dolea, V.E.

Toader, T. Maggipinto. The Development of the Romanian VLF/LF Monitoring System as Part

of the International Network for Frontier Research on Earthquake Precursors (INFREP), Rom.

Journ. Phys., 60, 7–8, 1203–1217 (2015)

2. S.A. Pulinets and K.A. Boyarchuk. Ionospheric precursors of earthquakes, Springer, Berlin

(2004).

3. I.R. Dobrovolsky, S.I Zubkov and V.I. Myachkin. Estimation of the size of earthquake

preparation zones. Pageoph, 117, 1025–1044 (1979).

4. J. Artru, T. Farges and P. Lognonn’e. Acoustic waves generated from seismic surface waves:

propagation properties determined from doppler sounding observation and normal-modes

modeling. Geophys. J. Int., 158, 1067–1077 (2004).

5. L. Ciraolo. Evaluation of GPS L2-L1 biases and related daily TEC profiles, in: Proc. of the

GPS/Ionosphere Workshop, 90–97, Neustrelitz (1993).

6. P.I. Nenovski, M. Pezzopane, L. Ciraolo, M. Vellante, U. Villante and M. De Lauretis. Local

changes in the total electron content immediately before the 2009 Abruzzo earthquake. Adv. In

Space Res. 55, 243–258 (2015).

7. V. Yu, Yasyukevich, S.V. Voeykov, I.V. Zhivetiev and E.A. Kosogorov. Ionospheric Response

to Solar Flares of C and M Classes in January–February 2010. Cosmic Research, 51, 114–123

(2013).

8. Y. He, D. Yang, J. Qian and M. Parrot Response of the ionospheric electron density to different

types of seismic events. Nat. Hazards Earth Syst. Sci., 11, 2173–2180 (2011).

9. H. Le, J.Y. Liu and L. Liu A statistical analysis of ionospheric anomalies before 736 M6.0+

earthquakes during 2002–2010, J. Geophys. Res., 116, A02303 (2011).

10. M. Hayakawa. On the fluctuation spectra of seismo-electromagnetic phenomena. Nat. Hazards

Earth Syst. Sci., 11, 301–308 (2011).

11. A.A. Rozhnoi, M.S. Solovieva, P.F. Biagi, K. Schwingenschuh and M. Maekawa. Low

frequency signal spectrum analysis for strong earthquakes. Annals of geophysics, 55 (2012).