Introduction
In recent years, seismo-ionospheric anomalies occurring before large
earthquakes have been widely investigated. The total electron content (TEC) derived from measurements
of local ground-based GPS receivers is used first by Liu et al. (2001) to
study the variations in ionospheric electron density before the 1999 M7.6
Chi-Chi earthquake and they found that the TEC decreases in the afternoon
period on day 1, 3 and 4 before this earthquake. This pre-earthquake
precursor is further confirmed by Liu et al. (2004) through a statistical
analysis of global ionosphere
map (GIM) TEC during the 20 M ≥ 6.0 earthquakes in Taiwan from
September 1999 to December 2002. After that, more similar studies were done by
using the GIM to study TEC anomalies before large earthquakes, and many
striking results have been found. For example, the statistical result of Liu
et al. (2009) indicates that the GIM TEC above the epicenter often
pronouncedly decreased on days 3–5 before 17 M ≥ 6.3 earthquakes
during the 10-year period from 1 May 1998 to 30 April 2008. Liu et al. (2010)
reported seismo-ionospheric precursors of anomalous decreases in TEC, which
appear 5 days prior to the 26 December 2004 M9.3 Sumatra–Andaman earthquake.
For the 2008 M8.0 Wenchuan earthquake, remarkable precursors are also
observed locally around both the epicenter and its conjugate point (Jhuang et
al., 2010; Kakinami et al., 2010; Liu et al., 2009; Yan et al., 2012; Zhao et
al., 2008). It is also found that the TEC over the epicenter significantly
enhances on day 1 before the 12 January 2010 M7 Haiti earthquake. The TECs
of the two midlatitude dense strips on 35∘ N/60∘ S and
those of the epicenter/conjugate point reach their maximum values on day 1
before the earthquake, while the northern crest of equatorial ionization
anomaly (EIA) moves poleward (Liu et al., 2011).
Meanwhile, many studies of earthquake ionospheric precursors have been
performed using the data from orbiting satellites. The French DEMETER
(Detection of Electro-Magnetic Emissions Transmitted from Earthquake Regions)
satellite has highly sensitive measurements over seismically active and quiet
regions, and provides another efficient and supplementary way to study
earthquake-related ionospheric variations. Using the DEMETER data, Zhang et
al. (2009a, b) reported some ionospheric anomalies possibly related to the
12 May 2008 M8.0 Wenchuan earthquake, such as the smallest value in O+
density and the ion temperature variations on 9 May above the epicenter, and
the ionosphere VLF electric field anomalies. The statistical results of
Akhoondzadeh et al. (2010) indicate that the positive and negative anomalies
in both of DEMETER and TEC data during days 1–5 before all studied
earthquakes during quiet geomagnetic conditions are highly regarded as
seismo-ionospheric precursors. Zhima et al. (2012a, b) show that possible
ionospheric electromagnetic perturbations occurred 4 days before the
earthquake in the ELF/VLF frequency range. Yan et al. (2014) show that among
37 earthquakes under study, positive electron density anomalies occur before
earthquakes, mostly within days 1–6 before the shocks. A statistical
analysis of the box-and-whisker method also shows that the nighttime Ne and
Ni (daytime Ti) over the epicenter significantly decrease (increase)
days 1–6 before the 2008 M8.0 Wenchuan earthquake (Liu et al., 2015).
The information of the main shock (M7.7, 17 July 2006, south of
Java). (a) The radius (ρ=100.43⋅M) of earthquake
preparation zone. (b) The location of the main shock (the red star)
and its preparation zone (the light-red circle). The two black curves show
two downward half-orbits (10835_0 and 10836_0) of DEMETER satellite in
the daytime on day 2 before the main shock (15 July 2006). (c) The
M ≥ 5.0 earthquakes occurred around the main shock from 2 June to
20 July 2006: (top) the magnitudes and (bottom) the distances from the main
shock, where the two red dots represent the main shock.
There are several simultaneous observations of seismo-ionospheric anomalies in
the GPS TEC and in situ electron density from orbiting satellites (Karia et
al., 2012; Ho et al., 2013; Liu et al., 2016), but there are very few cross-validation
analyses between TEC and plasma characteristics anomalies at almost the same
time before an impending earthquake. In the present paper, we, using the GPS
TEC and DEMETER plasma data, studied the ionospheric variations over the
epicenter of the 17 July 2006 M7.7 south of Java earthquake. The results
indicate that seismo-ionospheric anomalies in the GPS TEC and in situ plasma
density occur at almost the same time over the epicenter.
Applied methods and observations
TEC anomalies
As in Liu et al. (2004), the median and the interquartile range of data are
utilized to construct their upper and lower bound in order to separate
seismic anomalies from the background of normal variations. The upper and
lower bounds of the mentioned range can be calculated using Eqs. (1)–(4) (Akhoondzadeh et al., 2010):
xUB=M30+k⋅IQR,xLB=M30-k⋅IQR,Dx=(x-M30)/IQR,(xLB<x<xUB)p=±[Dx-k/k]⋅100%,
where x, xUB, xLB, M30, IQR,
Dx, k and p are the TEC value, upper bound, lower bound, 30-day
running median value, interquartile range, differential of TEC, threshold
value of the anomaly, and percentage of TEC change from the undisturbed state,
respectively. If the absolute value of Dx is larger than the k value
(Dx>k), the behavior of the relevant TEC value will be
regarded as anomalous. Considering the magnitude of the main shock, here we
set value k=2.0.
TEC anomaly analysis for the south of Java earthquake (17 July 2006)
from 2 June to 27 July 2006. The earthquake time is represented by a red
star. The x axis represents the day relative to the earthquake day. The
y axis represents the UT (LT = UT + 7 h). (a) Dst
geomagnetic index. (b) Kp geomagnetic index. (c) Solar
radio flux F10.7 index. (d) TEC anomaly detected under the following
conditions: Dst>-30 nT, Kp < 3, F10.7 < 100 sfu and
DTEC>2.0. Here
1 TECU = 1016 electrons m-2.
The GIM latitude–longitude–time (LLT) maps observed during the
period of 02:00–04:00, 04:00–06:00 and 12:00–14:00 UT on day 2 before the
17 July 2006 M7.7 south of Java earthquake. The GIM LLT maps during the
fixed period of 02:00–04:00 UT (left column), 04:00–06:00 UT (middle
column) and 12:00–14:00 UT (right column). Panels of row (a) are
the observed values on day 2 before the earthquake (15 July 2006), while row
(b) shows the median values of the period of days 1–30 before
(17 June to 16 July 2006) the earthquake. The red squares in rows (a, b) indicate the regions of interest around the earthquake, in range of
41.25∘ S–11.25∘ N latitude and 62.5–167.5∘ E
longitude. Panels of row (c) denote the extreme differences (DTEC>2.0) of the 30-day period that appeared on
15 July 2006. The color denotes the difference value of the TEC observed on
15 July 2006 from the relevant median value. The red dashed circles with the
radius ρ=2046.4 km represent the earthquake preparation area of the
lithosphere.
We first examine the variations in geomagnetic Dst and Kp index and solar
flux F10.7 index during the period from 2 June to 27 July 2006, i.e., day 45
before to day 10 after the 17 July 2006 M7.7 south of Java earthquake. As
seen from Fig. 2, geomagnetic and solar activities are relatively weak during
this period. Nevertheless, we still adopt a stringent condition
(Dst>-30 nT, Kp < 3 and F10.7 < 100 sfu) to separate
preseismic ionospheric phenomena from the ionospheric disturbances due to the
magnetospheric and solar activities. We compute the TEC around the epicenter
by using a linear interpolation of four data points nearest the epicenter
(9.28∘ S, 107.42∘ E). Since the resolutions of the GIM TEC
are 2.5∘ in latitude and 5∘ in longitude, we take the data
point centers within the ranges of 7.5–10∘ S and
105–110∘ E. Figure 2d illustrates the DTEC values from
2 June to 27 July 2006 close to the epicenter, which are computed according
to Eq. (3).
The GIM LLT maps observed during the global fixed period of
09:00–11:00, 11:00–13:00 and 19:00–21:00 LT on day 2 before the
17 July 2006 M7.7 south of Java earthquake. The GIM LLT maps during three
global fixed local times: (left column) 09:00–11:00 LT, (middle column)
11:00–13:00 LT and (right column) 19:00–21:00 LT, respectively. Panels of
row (a) show the observed values on day 2 before the earthquake
(15 July 2006), while those of (b) show the median values of the
period of days 1–30 before (17 June to 16 July 2006) the earthquake. The red
squares in rows (a, b) indicate the regions of interest around the
earthquake, in the range of 41.25∘S–11.25∘ N latitude and
62.5–167.5∘ E longitude. Panels of row (c) denote the
extreme differences (DTEC>2.0) of the 30-day
period, which appeared on 15 July 2006. The red dashed circles with the
radius ρ=2046.4 km represent the earthquake preparation area of the
lithosphere.
In our study, anomalous TEC time is searched only within the interval during
which DTEC>2.0, Dst>-30 nT,
Kp<3 and F10.7 < 100 sfu, and the TEC anomaly days are the
days during which the total anomalous TEC time exceed 6 h within the range
of 00:00–24:00 LT (local time). In Fig. 2d, the TEC anomaly is a function
of time (LT–UT) and number of days relative to the main shock onset (marked
with a red star). As shown in Fig. 2d, the day 2 before the earthquake is the
TEC anomaly day (15 July 2006). There are strong enhancement of TEC anomaly
during the periods of 02:00–04:00 UT (09:00–11:00 LT, +9.35 %
increases) and 04:00–06:00 UT (11:00–13:00 LT, +9.70 % increases),
and an anomalous reduction during 12:00–14:00 UT (19:00–21:00 LT,
-1.67 % decreases). This is consistent with the previous results that
the preseismic ionospheric anomalies can be positive as well as negative
(Akhoondzadeh et al., 2010; Pulinets and Davidenko, 2014; Pulinets et al.,
2003, 2015).
Spatial anomalies on latitude–longitude–time (LLT) maps
For the abovementioned three anomalous periods, we conduct a spatial
analysis to check if the GIM TECs simultaneously perturb in the earthquake
region. Each GIM map consists of 5040 (70×72) grid points, and covers
±87.5∘ N latitude and ±180∘ E longitude ranges with
spatial resolutions of 2.5∘ in latitude and 5∘ in longitude,
respectively.
Latitude–time–TEC (LTT) plots along the earthquake longitude
(107.42∘ E) extracted from the GIM during 9–20 July 2006. The solid
and open star symbols are the epicenter and corresponding conjugate point of
the south of Java earthquake, respectively. The three dashed white lines from
the bottom to top denote magnetic latitudes of the epicenter, magnetic
equator, and conjugate point, respectively. The two black dashed ellipses are
the regions of anomalous “asymmetry” during the period of 02:00–10:00 UT
(09:00–17:00 LT) on 16 and 17 July 2006.
Figure 3a shows the GIM TECs LLT map for each anomalous period. We compute
the median of the GIM TECs for each grid point in each anomalous period
during day 1–30 before the earthquake, i.e., 17 June to 16 July 2006
(Fig. 3b). Figure 3c shows the extreme differences (DTEC>2.0) between the observed GIM TEC and the associated 30-day median
at three anomalous periods that appeared on 15 July 2006, i.e.,
02:00–04:00 UT (left), 04:00–06:00 UT (middle) and 12:00–14:00 UT
(right). Here, the 30-day median represents the undisturbed background of GIM
TECs, while the positive (negative) difference represents the enhancement
(reduction) of the GIM TECs.
As shown in Fig. 3c, the ionospheric GIM TECs around the south of Java
earthquake epicenter marked by the red dashed circle drastically enhance by
∼ 0.81–42.14 % in the period of 02:00–04:00 UT
(09:00–11:00 LT) and ∼ 0.51–94.49 % in the period of
04:00–06:00 UT (11:00–13:00 LT) and decrease by
∼ 4.72–47.44 % in the period of 12:00–14:00 UT
(19:00–21:00 LT). The red circle with a radius of 2046.4 km represents the
earthquake preparation zone of the lithosphere. In order to exclude the local
time and/or EIA effects, the sequence of GIMs for three corresponding global
fixed local times is examined. As shown in Fig. 4, compared with the TEC
enhancements or reductions at three different universal times in Fig. 3, the
corresponding extreme enhancements or reductions in the GIM TECs at global
fixed local times are also mainly located around the forthcoming epicenter
and EIA region. Therefore, the spatial anomalies simultaneously and
prominently appear in the three anomalous periods on the TEC anomaly day
(15 July 2006) around the epicenter of the south of Java earthquake.
Asymmetry on latitude–time–TEC (LTT) plots
We apply another spatial analysis to further confirm the preseismic anomaly
related to the south of Java earthquake. We first extract the GIM TEC along
the epicenter longitude 107.42∘ E to produce LTT plots of the 12-day
period (day 8 before and day 3 after the earthquake) from 9 to 20 July 2006
(Fig. 5), and then compute the TEC around the epicenter longitude
107.42∘ E by using a linear interpolation of two data points from
the nearest longitudes (105 and 110∘ E).
It can be seen from Fig. 5 that the ionospheric GIM TEC between
10∘ S and 10∘ N along the epicenter longitude marked by the
black dashed ellipses significantly reduces on 16 and 17 July 2006. As a
result, an anomalous “asymmetric” structure is formed between the
epicenter-side and corresponding conjugate-side regions with respect to the
magnetic equator (∼ 10∘ N). In order to quantify this kind of
asymmetry, we introduce an asymmetry coefficient αm, which is
expressed in Eq. (5), to indicate the temporal variation in LTT asymmetry
with respect to the magnetic equator.
αm=(xN-xS)/[0.5⋅(xN+xS)],
where xN and xS are the symmetric TEC values in
the conjugate side (north) and epicenter side (south) of the magnetic equator,
respectively.
Figure 6a–d illustrate the distribution of asymmetry coefficient αm of LTTs in Fig. 5 in four continuous periods, i.e., 02:00–04:00,
04:00–06:00, 06:00–08:00 and 08:00–10:00 UT, which are before or at the
time of the earthquake (08:19 UT), during 9–20 July 2006. The asymmetry
coefficients αm at four central symmetric latitude pairs
(22.5∘ N–2.5∘ S, 20.0∘ N–0.0∘ S,
17.5–2.5∘ N and 15.0–5.0∘ N) with respect to the magnetic
equator increase simultaneously from the day 2 before the earthquake
(15 July 2006) and reach the peaks on the earthquake day and then return to
the normal levels gradually. As shown in Fig. 6, the “asymmetric” structure
is the most evident on the earthquake day, and the corresponding asymmetry
coefficients αm reach their
peaks and increase by 104.62, 398.41, 672.62 and 177.75 % from their
average levels, respectively. The anomalous “asymmetric” structure
occurring before the south of Java earthquake is distinctive and worthy of
further study.
The asymmetry of LTT plots. (a–d) The temporal
distribution of asymmetry coefficient αm of LTT plots in Fig. 5
during four continuous time periods, i.e., 02:00–04:00, 04:00–06:00 UT,
06:00–08:00 and 08:00–10:00 UT, from 9 to 20 July 2006. The green,
brown, blue and magenta dashed lines represent the distribution of
αm of LTT plots at four different symmetric latitude pairs with
respect to the magnetic equator, i.e., 22.5∘ N–2.5∘ S,
20.0∘ N–0.0∘ S, 17.5–2.5∘ N and
15.0–5.0∘ N, and the black solid line represents the
mean value of that of four different symmetric latitude pairs with respect to
the magnetic equator. The earthquake day and day 2 before the earthquake are represented
by red vertical solid and dashed lines, respectively.
Plasma parameter anomalies above the earthquake preparation zone
from 2 June to 27 July 2006. (a) Electron density and
(c) electron temperature derived from the measurements of the ISL in
the daytime. (b) Total ion density and
(d) ion temperature, (e) H+, (f) O+ and
(g) He+ density derived from the measurements of IAP in the
daytime. Since the DEMETER satellite sometimes did not pass over
the earthquake preparation zone in the daytime, there are some time gaps. The
black dashed lines indicate the median value (M56) and the two
black solid lines indicate the upper and lower bounds (M56±2.0⋅IQR). (h) The periods during which DEMETER satellite passed over
the earthquake preparation zone in the daytime (∼ 09:00–10:30 LT).
The x axis represents the day relative to the earthquake day. The
earthquake day and day 2 before the earthquake are represented by red vertical solid and
dashed lines, respectively.
Plasma variations
In this study, we have analyzed GIM TEC anomalies extracted from GPS
satellites during day 45 before to day 10 after the event over the earthquake
epicenter and its preparation zone. In order to confirm observed TEC
anomalies, a cross-validation analysis is conducted by using the DEMETER
data.
In a similar way to that described in Sect. 3.1, we use the data recorded by
ISL and IAP instruments on DEMETER to study the ionospheric parameters above
the preparation zone (radius ρ=2046.4 km) during day 45 before
(2 June 2006) to day 10 after (27 July 2006) the south of Java earthquake. As
the undisturbed reference background, we take the median value
(M56), the upper and lower bounds
(M56 ± 2.0 ⋅ IQR) of each ionospheric parameter
during the 56 days.
Figure 7 illustrates variations in different ionospheric parameters above the
south of Java earthquake preparation zone in the daytime from 2 June to
27 July 2006. Since the DEMETER satellite sometimes does not pass over the
earthquake preparation zone in the daytime, there are several time gaps. As
shown in Fig. 7a and b, both electron and total ion densities increase
significantly and reach their maximum values of the study period (exceeding the
upper bound by +19.52 and +13.80 %, respectively) on day 2 before the
earthquake (15 July 2006). The main component O+ density increases by
the order of 13.94 % (Fig. 7f) while H+ density simultaneously
reaches its minimum value (a 11.72 % decrease, Fig. 7e) on
15 July 2006. However, as shown in Fig. 7g, He+ density is relatively
stable. Furthermore, the ion temperature reduces oppositely by 1.59 % on
15 July 2006, on day 2 before the earthquake. The electron
temperature also decreases and reaches the minimum value in the daytime on
the same day. However, it does not drop below the lower bound.
Also, it should be noted that the plasma parameters reach the extremum values
on day 2 before the earthquake (15 July 2006), and in the daytime at about
10:30 LT (03:30 UT, Fig. 7h), when satellite is above the earthquake
preparation zone in two downward half-orbits (10835_0 and 10836_0,
Fig. 1b). Meanwhile, the DEMETER satellite also passed through the earthquake
preparation zone during the nighttime at about 22:30 LT, but we do not see
any similar variation before the occurrence of earthquake. This is also
consistent with the fact that the GIM TEC enhancements are observed at
the same daytime period, i.e., 09:00–13:00 LT (02:00–06:00 UT).
Discussion and conclusion
The simultaneous anomalies of GIM TEC derived from GPS data and plasma
parameters recorded by DEMETER clearly show the preseismic disturbed signals
on day 2 prior to the south of Java earthquake over the preparation zone of
the impending earthquake. The Dst, Kp and F10.7 indexes are rather small around
the earthquake time, which suggests that the anomalies around the epicenter
are not caused by the geomagnetic or solar activities within that period.
Furthermore, the rare preseismic quiescence (see Fig. 1c) provides a
favorable environment for earthquake precursor analysis.
The LLT maps in Fig. 3 show that the anomalies synchronously appear in the
earthquake preparation zone in the three universal/local periods, i.e.,
02:00–04:00 UT (09:00–11:00 LT), 04:00–06:00 UT (11:00–13:00 LT) and
12:00–14:00 UT (19:00–21:00 LT) on 15 July 2006, day 2 before the
earthquake. Taking into account the local time and/or EIA effects, the
geographic region and the signs of anomalies agree very well with those of
the same three TEC anomalies (Fig. 2d) on that day. In other words, the TEC
anomalies can be the positive enhancements (increase
∼ 0.81–42.14 % in the period of 02:00–04:00 UT and
∼ 0.51–94.49 % in the period of 04:00–06:00 UT) as well as the
negative reductions (decrease ∼ 4.72–47.44 % in the evening period
of 12:00–14:00 UT). This feature also coincides with the previous results
that TEC over the forthcoming epicenter region tends to significantly
increase or decrease several days before the earthquake occurrence
(Akhoondzadeh et al., 2010; Ho et al., 2013; Liu et al., 2004, 2011). It
should be mentioned that the spatial TEC anomalies are exactly observed
during the abovementioned three periods around the epicenter in the global
LLT maps, which allow us to exclude the possible induced effects from the
weak geomagnetic storm (Dst = ∼ -30 nT, from ∼ 18:00 UT
14 July to ∼ 06:00 UT 15 July) (Fig. 2a). Since another two observed
TEC anomalies only last for 2 h during 16:00–18:00 UT on 5 July 2006
(day 12 before the earthquake, increases) and 22:00–24:00 UT on 8 July 2006
(day 9 before the earthquake, decreases), there is little possibility that
they are related to the earthquake. Also, when we spatially check the LLT
maps during these two periods, we do not find any corresponding anomalies in
the preparation zone. Thus, the temporal and spatial coincidences strongly
suggest that the TEC anomaly on 15 July is the seismo-ionospheric precursor
of the south of Java earthquake.
Meanwhile, the anomalous variations in the measured plasma parameters from
DEMETER satellite are observed in the daytime on the same day (15 July 2006,
Fig. 7). In comparison with GPS satellites, the DEMETER satellite can only
pass through the earthquake preparation zone at two local times: 10:30 and
22:30 LT. Thus, it provides us an opportunity to
confirm observed TEC anomalies by GPS satellites in the daytime, although the
penetration of anomalous electric field into the ionosphere in the nighttime
will be more efficient than that in the daytime (Pulinets and Boyarchhuk,
2004). A dominant increase in electron density by 19.52 %, which exceeds
the upper bound and reaches the maximum value, is observed at about 10:30 LT
on day 2 before the earthquake, which is consistent with the TEC enhancements
during 09:00–13:00 LT (02:00–06:00 UT). The total ion density also
increases to its maximum value within the same period. The main component
O+ density increases by 13.94 % while H+ density simultaneously
decreases by 11.72 % and He+ density remains relatively stable.
However, the decreases in electron and ion temperature are also observed
within the same period. It should be noted that almost all of the
abovementioned plasma parameters over the earthquake preparation zone on
15 July 2006 simultaneously yield their extremum values, but remain stable
during the remaining time, for the whole quiet geomagnetic period from day 45 before to day 10
after the earthquake. The results shown in Fig. 7 indicate that the plasma
variations on 15 July are very likely related to the south of Java
earthquake.
An anomalous “asymmetric” structure appears in the LTT plots (Fig. 5)
within the latitude range of ∼ 10∘ S–30∘ N during
the period of ∼ 02:00–10:00 UT (around the earthquake time 08:19 UT)
on 16 and 17 July 2006. However, located there is a “peanut-pod-like”
distribution whose waist line roughly overlaps with the magnetic equator
on other days. Moreover, as shown in Fig. 6, the asymmetry coefficients
αm at four central symmetric latitude pairs with respect to the
magnetic equator increase simultaneously from 15 July and reach their peaks
on 17 July 2006 (the earthquake day) and then return to the normal levels
gradually. The latitude range of the reduced TEC on the epicenter side in the
LTT plots, marked by the two black dashed ellipses in Fig. 5, are exactly
located over the earthquake preparation zone. Although the
perturbation process is not fully understood in great detail, the reductions in the GPS TEC
might be caused by the change of the electromagnetic environment (i.e.,
radiations, low-frequency electric fields and magnetic fields, etc.) around
the forthcoming epicenter area during the earthquake preparation period
(Freund, 2000; Bhattacharya et al., 2009; Liu et al., 2009). On the one hand, the
seismo-generated radiations might cause the plasma thermal expansions and
result in a large volume of the ionospheric electron density reduction and
outflow into the magnetosphere along the Earth's magnetic field line (Liu et
al., 2009). On the other hand, the disturbed electric field would also be
able to produce a plasma E×B drift (westward or
eastward), which in turn results in the extreme reductions in the GIM TEC
(Liu et al., 2009, 2010, 2011). Despite all that, this subject also needs to
be further investigated in the future.
In conclusion, the temporal and spatial anomalies of the GIM TEC and striking
plasma variations from the DEMETER over the epicenter indicate the existence
of a significant preseismic ionospheric precursor on day 2 before the
earthquake. In particular, the localization and synchronization of the long time
anomalies around the occurrence of earthquake suggest that these
perturbations on 15 July 2006 are highly related to the south of Java
earthquake. However, the perturbation process is still not fully
understood in great detail and needs further investigation in the future.