Available online at www.sciencedirect.

com

ScienceDirect
Advances in Space Research 61 (2018) 248–263
www.elsevier.com/locate/asr

Multi precursors analysis associated with the powerful

Ecuador (MW = 7.8) earthquake of 16 April 2016 using Swarm
satellites data in conjunction with other multi-platform satellite
and ground data
Mehdi Akhoondzadeh a,⇑, Angelo De Santis b, Dedalo Marchetti b, Alessandro Piscini b,
Gianfranco Cianchini b
a
Remote Sensing Department, School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, North Amirabad Ave.,
Tehran, Iran
b
Istituto Nazionale di Geofisica e Vulcanologia, Roma, Italy

Received 2 June 2017; received in revised form 5 July 2017; accepted 7 July 2017
Available online 15 July 2017

Abstract

After DEMETER satellite mission (2004–2010), the launch of the Swarm satellites (Alpha (A), Bravo (B) and Charlie (C)) has created
a new opportunity in the study of earthquake ionospheric precursors. Nowadays, there is no doubt that multi precursors analysis is a
necessary phase to better understand the LAIC (Lithosphere Atmosphere Ionosphere Coupling) mechanism before large earthquakes. In
this study, using absolute scalar magnetometer, vector field magnetometer and electric field instrument on board Swarm satellites, GPS
(Global Positioning System) measurements, MODIS-Aqua satellite and ECMWF (European Centre for Medium-Range Weather Fore-
casts) data, the variations of the electron density and temperature, magnetic field, TEC (Total Electron Content), LST (Land Surface
Temperature), AOD (Aerosol Optical Depth) and SKT (SKin Temperature) have been surveyed to find the potential seismic anomalies
around the strong Ecuador (Mw = 7.8) earthquake of 16 April 2016. The four solar and geomagnetic indices: F10.7, Dst, Kp and ap were
investigated to distinguish whether the preliminary detected anomalies might be associated with the solar-geomagnetic activities instead
of the seismo-ionospheric anomalies. The Swarm satellites (A, B and C) data analysis indicate the anomalies in time series of electron
density variations on 7, 11 and 12 days before the event; the unusual variations in time series of electron temperature on 8 days preceding
the earthquake; the analysis of the magnetic field scalar and vectors data show the considerable anomalies 52, 48, 23, 16, 11, 9 and 7 days
before the main shock. A striking anomaly is detected in TEC variations on 1 day before earthquake at 9:00 UTC. The analysis of
MODIS-Aqua night-time images shows that LST increase unusually on 11 days prior to main shock. In addition, the AOD variations
obtained from MODIS measurements reach the maximum value on 10 days before the earthquake. The SKT around epicentral region
presents anomalous higher value about 40 days before the earthquake. It should be noted that the different lead times of the observed
anomalies could be acknowledged based on a reasonable LAIC earthquake mechanism. Our results emphasize that the Swarm satellites
measurements play an undeniable role in progress the studies of the ionospheric precursors.
Ó 2017 COSPAR. Published by Elsevier Ltd. All rights reserved.

Keywords: Swarm satellite; Earthquake precursor; Ecuador earthquake; LAIC mechanism

⇑ Corresponding author.
E-mail address: [email protected] (M. Akhoondzadeh).

http://dx.doi.org/10.1016/j.asr.2017.07.014
0273-1177/Ó 2017 COSPAR. Published by Elsevier Ltd. All rights reserved.
 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263 249

1. Introduction field (e.g. De Santis et al., 2017). The main mission sensors
are a couple of magnetometers, i.e. an Absolute Scalar
Earthquake is a dynamic phenomenon that can be antic- Magnetometer (ASM) providing, in nominal mode, mea-
ipated by some geophysical and geochemical variations surements of the field intensity, and a Vector Field Magne-
occurring in lithosphere, atmosphere and ionosphere dur- tometer (VFM) mounted halfway along the boom on an
ing its preparation phase. These seismic LAIC (Litho optical bench together with the Star Trackers, providing
spheric-Atmospheric-Ionospheric Coupling) perturbations field directions. Other sensors complete the payload of each
are widely interpreted as earthquake precursors. There satellite: in particular, two electric field and particle sensors
are many scientific evidences implying on observation of (Langmuir Probes), a GPS antenna and an accelerometer
seismic LAIC anomalies, which begins a few days before (Olsen et al., 2013). In contrast with the previous magnetic
the earthquake and stay until a few days after it (Parrot, satellites, Swarm measurements are made at different orbits
1995; Hayakawa and Molchanov, 2002; Pulinets and and altitudes. Therefore, these constellation and orbitogra-
Boyarchuk, 2004). Up to now, several earthquake mecha- phy could be a precise investigator of geomagnetic field to
nisms have been proposed to explain the occurrence of detect the anomalies likely connected to earthquakes in
the earthquake precursors, but this topic is still a challeng- their preparation phase. In this research, six months of
ing task (Tronin et al., 2002; Molchanov and Hayakawa, Swarm magnetic field and electron density satellite data
2008; Pulinets, 2009; Freund, 2011; Pulinets and were processed to detect the potential seismic anomalies
Ouzounov, 2011). Satellite measurements are up to date around the time and location of Ecuador earthquake.
and usually free access and also cover a wide area. There-
fore, they could be a supplementary data in addition to 2.2. GPS-TEC data
the ground stations measurements. Up to now, data
recorded by many different satellites (Alouette, GEOS 1 The Global Positioning System (GPS) has become a
and 2, Cosmos 1809, OGO 6, Aureol 3, IK Bulgaria very useful tool for ionospheric studies, especially for the
1300, IK 19 and 24, AE-C, ISIS 2, Topex-Poseidon, DE estimation of the ionospheric Total Electron Content
2, CHAMP, DEMETER, GPS Satellites, Mir, etc.) have (TEC). The TEC is the amount of free electrons along
been used for statistical analyses and understanding of the path of the electromagnetic wave between each satellite
the pre-seismic ionospheric variations. Since a precursor R satellite
and the receiver, given by, TEC ¼ receiver N  ds where N is
cannot appear alone before an impending earthquake,
the electron density. In this paper, the GIM (Global Iono-
multi-precursors analysis is usually taken into account in
spheric Map) data constructed into 5°  2.5° (Longitude,
some studies (e.g. Akhoondzadeh, 2011). In this paper we
Latitude) grid with a time resolution of 1 h were analysed.
will apply these concepts to the recent (16 April 2016)
GIM data are generated on a daily basis using data from
M7.8 Ecuador earthquake.
about 150 GPS sites of the IGS (International GNSS Ser-
vice) and other institutions. In this paper, TEC data based
2. Data
on the date and geographic location of the earthquake for
about six months around the location and date of the main
In this study, the variations of electron density/temper-
event were analysed.
ature and magnetic field intensity/vector are considered
using Swarm satellites measurements. In addition, GPS-
TEC satellites data, MODIS-Aqua satellite products 2.3. Modis-Aqua data
including the LST (Land Surface Temperature) and the
AOD (Aerosol Optical Depth) and ECMWF climatologi- The Moderate Resolution Imaging Spectroradiometer
cal data are investigated in search of possible complemen- (MODIS) on board the Terra satellite, was launched on
tary precursors. December 18, 1999 for global monitoring of the atmo-
sphere, terrestrial ecosystems, and oceans. On May 4,
2.1. Swarm satellites data 2002, a similar instrument was launched on the EOS-
Aqua satellite. MODIS, flying on these two satellites with
Swarm is an ESA (European Space Agency) satellite its 2330 km swath width provides almost complete dual
mission of three satellites to survey precisely the geomag- global daily coverage in 36 spectral bands between 0.415
netic signals from Earth’s system, as well as the ionosphere and 14.235 lm with spatial resolutions of 250 m (bands 1
and magnetosphere. The three Swarm satellites are named and 2), 500 m (bands 3, 4, 5, 6 and 7) and 1000 m (bands
A (Alpha), B (Bravo) and C (Charlie). A and C are flying 8–36).
almost side-by-side with longitude separation of 1.4° at In this work we considered two products of MODIS–
equator, an altitude close to 450 km, and inclination of Aqua and Terra satellites, i.e. LST and AOD data.
87.4°. The polar orbit altitude of the third satellite (B) is
close to 510 km. This specific constellation allows the scien- 2.3.1. LST data
tists to observe the small space-scale variations of the geo- In the following, LST (Land Surface Temperature)
magnetic field, particularly those linked to the lithospheric products of MODIS satellite images close to the studied
250 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263

earthquake epicenter have been analysed. Both the daytime gated from 1979 to 2015 in the interval from 1 March to
and night-time LST images provided by NASA (http://mo- 30 April, and compared with the same period of 2016, cov-
dis.gsfc.nasa.gov/data) were considered. These data are ering the one month and half preceding the 16 April 2016
generated on a daily basis at a temperature resolution of Ecuador earthquake. As regard the 2016, we prefer to
0:02  C. Each pixel of a LST image covers an area of use ECMWF analysis (i.e. operational archive) data, in
1  1 km2 on the ground. For each image, the mean of order to avoid possible smoothing effect caused by ERA-
LST values of 100  100 pixels area, centered on the Ecua- Interim re-analysis, which could confuse the potential pre-
dor earthquake epicenter, has been analysed. cursor thermal anomalies with the bias of the dataset.
For the present study, we used data for the region from
5°S to 5°N and from 75°W to 85°W, with 0.5° spatial res-
2.3.2. AOD data
olution. In order to distinguish possible earthquake-
Aerosol optical depth (AOD) is calculated by measuring
induced temperature enhancements from the normal
light absorption at specific wavelengths of the visible spec-
weather phenomenon, only local time night-time data were
trum. For the most widely used AOD data product, the
used (about 01:00, 06:00 UTC), because thermal anomalies
absorption measurement at 550 nm is the preferred wave-
related to normal weather variations are seldom at night-
length. AOD is a dimensionless quantity, expressing the
time, without solar radiation.
negative logarithm of the fraction of radiation (e.g., light)
that is not scattered or absorbed on a path. The AOD mea-
2.5. Geomagnetic indices
surements are taken by the MODIS sun-synchronous
instrument on board Terra and Aqua satellites every day.
It should be noted that the measured ionospheric
The satellites provide more continuous coverage nearer to
parameters using satellite data can be disturbed by external
the poles but there are more gaps in the satellite coverage
sources, such as the solar geomagnetic field and geomag-
closer to the equator. In this study, aerosols variations have
netic storms, especially in the equatorial and polar regions.
been analysed using one of the atmosphere daily global
Therefore, we discarded those anomalies in the ionospheric
1  1 products deduced from MODIS Terra and Aqua
parameters detected during the periods of solar-terrestrial
daily level-3 data. This product, which is named ‘‘Aerosol
disturbances. Therefore, to discriminate the seismo-
Optical Depth at 550 nm”, is available via: http://gdata1.
ionospheric perturbations from geomagnetic disturbances,
sci.gsfc.nasa.gov/daac-bin/G3/gui.cgi?instance_id=
we investigated the geomagnetic indices Dst, Kp, ap and
MODIS_DAILY_L3. In this paper, AOD data for six
F10.7, accessed through NOAA (http://spider.ngdc.noaa.-
months around the earthquake time and location has been
gov), during the periods of analyses. The Kp index records
processed.
the planetary activity on a worldwide scale while the Dst
index monitors the equatorial ring current variations
2.4. ECMWF climatological data (Mayaud, 1980). The F10.7 index shows a measure of dif-
fuse, non-radiative heating of the coronal plasma trapped
ECMWF aims at providing accurate medium-range glo- by magnetic fields over active regions, and is an excellent
bal weather forecasts out to 15 days and seasonal forecasts indicator of overall solar activity levels.
out to 12 months. ECMWF periodically uses its forecast
models and data assimilation systems to ‘re-analyse’ 3. Observations
archived observations, creating global data sets describing
the recent history of the atmosphere, land surface and On 16 April 2016 at 23:58:36 UTC (LT = UTC-5:00) a
oceans. ERA-Interim is a global atmospheric reanalysis strong earthquake of magnitude Mw = 7.8 occurred in
from 1979, continuously updated in real time. The data the coast of Ecuador approximately 27 km south-
assimilation is based on a 12-hourly four-dimensional vari- southeast of Muisne, in the province of Esmeraldas, at a
ational analysis with adaptive estimation of biases in satel- depth of 20.6 km. The epicenter is located where the Nazca
lite radiance data (Dee et al., 2011). Data sources of ERA- plate subducts eastward beneath the South America plate.
interim include new ERS (European Remote Sensing Satel- The earthquake occurred as a slip on the primary plate
lite) altimeter wave heights, EUMETSAT (European Orga- boundary interface, or megathrust (see https://earth-
nization for the Exploitation of Meteorological Satellites) quake.usgs.gov for more information). Table 1 summarize
reprocessed winds and clear-sky radiances, GOME (Global the characteristics of the main shock and of the major
Ozone Monitoring Experiment) ozone data from the aftershocks.
Rutherford Appleton Laboratory, and GPS radio occulta-
tion measurements by CHAMP (CHAllenging Minisatel- 3.1. Solar and geomagnetic indices during the period of
lite Payload), GRACE (Gravity Recovery and Climate analysis
Experiment) and COSMIC (Constellation Observing Sys-
tem for Meteorology, Ionosphere and Climate). Fig. 1(a) and (b) illustrates the variations of Kp and ap
ECMWF ERA-Interim Skin Temperature (SKT) 06:00 geomagnetic indices, respectively, during the period of 1st
UTC (i.e. local night-time) data have been here investi- November 2015 to 30 April 2016: an asterisk indicates
 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263 251

Table 1
The characteristics of the Ecuador earthquake and the main aftershocks (reported by http://earthquake.usgs.gov/).
Date Time (UTC) Geographic Latitude, longitude Magnitude (MW) Focal depth (km)
16 April 2016 23:58:36 0.382°N, 79.922°W 7.8 20.6
20 April 2016 08:33:47 0.639°N, 80.210°W 6.2 14
20 April 2016 08:35:10 0.708°N, 80.035°W 6.0 10
22 April 2016 03:03:41 0.292°N, 80.504°W 6.0 10

Fig. 1. Results of analysis for the solar and geomagnetic indices from 01 November 2015 to 30 April 2016. (a), (b) and (c) are the variations of Kp, ap and
Dst geomagnetic indices, respectively. The Kp value is multiplied by 10. The earthquake time is indicated by an asterisk (a, b, c) or a vertical dotted line (d).
The x-axis represents the day relative to the earthquake day. The y-axis represents the time UTC (LT = UTC-5:00). (d) Variations of F10.7 solar index.
The unusual variations of indices in panels of a, b and c are depicted in red colors. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

the earthquake time. The X-axis represents the days rela- variations are seen on 6 March 2016 (41 days before the
tive to the Ecuador earthquake day, the Y-axis represents event) with the maximum value of 98 nT at 22:00 UTC.
the universal time coordinate. The high geomagnetic activ- The investigation of the variations of Dst geomagnetic
ities are clearly observed on 20 December 2015 (119 days index indicates the moderate and in some hours high geo-
before the event), when the Kp and ap values reach the magnetic activity in days close to earthquake date. The val-
maximum values of 6.7 and 111 at 18:00 UTC, respectively. ues of 60 and 59 nT on 8 (01:00 UTC) and 3 (06:00
The unusual variations of the Kp and ap indices are also UTC) days before the earthquake and also Dst value of
seen on 6 March 2016 (41 days before the event) between 55 nT on the earthquake day (21:00 UTC) are instances
19:00 and 24:00 UTC with the values of 6.3 and 94, respec- of the observed high geomagnetic activities around the
tively. These two indices acknowledge the fairly high geo- earthquake date.
magnetic activities around the earthquake date in some Fig. 1(d) shows the variations of solar radio flux (F10.7)
hours. The Kp and ap indices with the values of 5 and 48 during the period of 1st November 2015 to 30 April 2016.
on 12 April (4 days before the event) between 22:00 and The maximum value of 133.90 (SFU) on 23 December 2015
24:00 UTC can hide the likely seismic ionospheric anoma- (116 days before the event) is noticeable. Even the F10.7
lies at these hours. value gradually increases from about 16 days before the
Fig. 1(c) illustrates the variations of Dst geomagnetic earthquake and reaches the maximum value of 116.60
index during the period of 1st November 2015 to 30 April (SFU) on 11 April 2016 (5 days before the main shock).
2016. The abnormal Dst values are observed on 20 Decem- Therefore, we could conclude that the solar radio flux
ber 2015 (119 days before the event), when this parameter parameter probably did not affected the seismo-
exceeds the lower boundary value (i.e. 20 nT), reaching ionospheric anomalies detected around the earthquake
the lowest value of 155 nT at 23:00 UTC. Similar unusual date.
252 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263

3.2. Electron density and temperature satellite anomalies Median, and higher and lower bounds are drawn as blue
and green horizontal lines, respectively. In this figure, the
Dobrovolsky et al. (1979) showed that the radius of the days accompanied with high geomagnetic activities are
affected area by an impending earthquake can be estimated depicted with the M symbol. It should be noted that the
using the formula R ¼ 100:43M ; where R is the radius in kilo- values of the median and the allowable bounds were calcu-
metres of the earthquake preparation zone and M is its lated only using the quiet geomagnetic days. It is seen that
magnitude. Therefore, to detect the likely seismo- most of the detected anomalies are rejected and cannot be
ionospheric anomalies, we analysed the whole tracks of considered as seismo-ionospheric anomalies. It is seen that
satellites A, B and C crossing the Dobrovolsky’s circular the corrected curve exceeds the higher bound
region within R = 2259.4 km from the Ecuador earthquake (M þ 2  IQR) on 9 April 2016, 7 days prior to the earth-
epicenter during the period of 1st November 2015 to 30 quake with the value of 36.94% of the higher bound.
April 2016. (The second arrow in Fig. 2.) In order to do the detailed
In order to show that the likely observed anomalies analysis, the night-time tracks of satellite A on this date
found inside the Dobrovolsky’s area are not seen also out- were investigated. Fig. 3(a) shows the mentioned track
side it, the tracks are cut including a few degrees of latitude inside the Dobrovolsky’s area. The earthquake epicenter,
more. At the first step for each satellite, all tracks during the track and the Dobrovolsky’s area are shown as a red
the studied period were classified into the two groups asterisk, a red line and a green circle, respectively. The
including the daytime and the night-time. Then, the differ- track passed the Dobrovolsky’s area between the 20:36:20
ences DNe and DTe, between the measured electron density and 20:49:17 LT. The differences between the time-series
and temperature, and the correspondent values predicted of the measured electron density/temperature and a fitted
by IRI2016 (Bilitza, 2001) model were calculated. Then, polynomial of degree 12 along this track are shown in
after the median of the residuals for the electron density Fig. 3(b) and (c). The vertical axis represents the geomag-
and temperature parameters for each track was calculated, netic latitude. An unusual variation in the time series of
the time-series of the median values, for day and night dur- the residual curve of the electron density is clearly seen
ing the studied period, were constructed. At the last step, close to the earthquake location (the dotted red ellipse in
since the variations of the plasma parameters follow a non- Fig. 3(b)). Therefore, the detected anomaly on 9 April
linear behavior, a polynomial of degree 3 fitted the residual could be considered as a seismo-ionospheric anomaly. It
curve and the final corrected curve was obtained. Fig. 2 is also seen an anomaly on 4 April, 12 days before the
illustrates the satellite Swarm A night-time electron density earthquake (the first arrow in Fig. 2). This anomaly occurs
residual variations during the mentioned period. The after a day with high geomagnetic activity, but an abnor-
abscissa represents the days relative to the earthquake mal variation is seen in its track analysis. Therefore, it
day. The vertical dotted line shows the earthquake date. could be included as a potential seismic anomaly. The char-

Fig. 2. Results of Swarm electron density data analysis for the Ecuador earthquake (16 April 2016) from 1st November 2015 to 30 April 2016. The
earthquake day is represented as a vertical dotted line. The green horizontal lines indicate the upper and lower bounds (M  2:0  IQR). The blue
horizontal line indicates the median value (M). The values of the median and the allowable bounds were calculated only using the quiet geomagnetic days.
The x-axis represents the day relative to the earthquake day. The y-axis represents the Swarm satellite A night-time electron density median values for each
day. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263 253

Fig. 3. Results of Swarm A track analysis for the Ecuador earthquake (16 April 2016) on 09 April 2016. (a) The earthquake epicenter, the track and
Dobrovolsky’s area are shown as a red asterisk, a red line and a green circle, respectively. The track passed the Dobrovolsky’s area between the 20:36:20
and 20:49:17 LT. The horizontal and vertical axes represent the geographic longitude and latitude. (b) and (c) show the differences between the time-series
of the measured electron density and temperature and a fitted polynomial of degree 12 along this track. The vertical axis represents the geomagnetic
latitude. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

acteristics of the other detected anomalies in the electron Fig. 4). We made a confutation analysis for this observed
density and temperature variations for satellites A, B and anomaly. For example, this anomaly was also seen in con-
C are listed in Table 2. jugate geographic longitude. The careful investigation of
the solar geomagnetic parameters indicates that at the mea-
3.3. Magnetic field satellite anomalies surement time of this anomaly a geomagnetic storm is
beginning. The investigation of the satellite tracks of this
Absolute scalar magnetometer measures the strength of date does not show any unusual behavior. Therefore, this
the magnetic field to calibrate the vector fluxgate magne- observed anomaly could be considered as a level difference
tometer. Vector field magnetometer makes high precision between the IGRF and satellite measurements and could
measurements of the three components of the vector mag- not be considered as a seismo-ionospheric anomaly.
netic field. In order to detect the potential anomalies in the Another weak anomaly is also seen on 7 April, 9 days
variations of the magnetic field, at the first step, all tracks before the event. The track and confutation analysis
of each satellite were divided in two classes based on the declare that this anomaly is a potential earthquake anom-
local daytime and night-time. Then, for each sample of a aly. Fig. 5 shows the daytime corrected magnetic field vec-
track, the difference between the measured magnetic field tor X values measured using satellite A during the period of
value and the predicted one from the IGRF (International 1st November 2015 to 30 April 2016. The observed anoma-
Geomagnetic Reference Field) model was calculated. At lies are similar to the scalar magnetic field anomalies and
the next step, after the median of the residuals of magnetic the similar conclusions about considering them as seismic
values of the daytime and night-time tracks was obtained, anomalies were obtained. Fig. 6 illustrates the daytime cor-
we constructed the time-series of the magnetic field median rected magnetic field vector Y values measured using satel-
values during the period of 1st November 2015 to 30 April lite A during the period of 1st November 2015 to 30 April
2016. To eliminate the nonlinear variations a polynomial of 2016. The first anomaly (the first arrow in Fig. 6) happens
degree 3 fitted the time series and the residuals values were between the two high geomagnetic days. Also, it is detected
calculated. Fig. 4 illustrates the daytime corrected scalar 148 days before the event. Therefore, it is suspicious to
magnetic field values measured using satellite A during consider it as an anomaly due to the earthquake. The sec-
the studied period. A striking anomaly is clearly seen on ond anomaly (the second arrow in Fig. 6) is observed on 6
2 April, 14 days before the event (the first arrow in February, 70 days before the earthquake. We checked the
254 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263

Table 2
The list of the detected anomalies. Day is relative to the earthquake day. Value is calculated by: p ¼ 100  ðjDxj  kÞ=k which indicates the percentage of
the parameter change from the undisturbed state; Dx ¼ ðx  MÞ=IQR ; x, M and IQR are parameter value, median and interquartile range, respectively. The
k value is proportional to the earthquake magnitude.
Parameter Day Time (LT) Value (%)
Electron density, Swarm A satellite 7 20:42 36.94
12 21:09 94.74
Electron density, Swarm B satellite 11 12:16 9.70
Electron density, Swarm C satellite 7 20:48 15.06
12 21:15 110.18
Electron temperature, Swarm C satellite 8 20:54 1.02
Magnetic field, Scalar, Swarm A satellite 9 08:51 1.04
Magnetic field, Vector X, Swarm A satellite 9 08:51 21.17
Magnetic field, Vector Y, Swarm A satellite 23 10:06 66.26
48 12:20 52.78
52 12:41 9.11
Magnetic field, Vector Z, Swarm A satellite 7 08:40 3
16 09:28 7.33
Magnetic field, Vector X, Swarm B satellite 9 12:06 37.17
Magnetic field, Vector Z, Swarm B satellite 11 12:16 8.19
Magnetic field, Scalar, Swarm C satellite 9 08:56 7.46
Magnetic field, Vector X, Swarm C satellite 9 08:56 30.43
Magnetic field, Vector Y, Swarm C satellite 23 10:11 100.40
48 12:26 66.50
Magnetic field, Vector Z, Swarm C satellite 7 08:46 1.52
16 09:34 1.65
25 10:22 44.67
43 11:59 8.92
TEC 1 04:00 14.29
LST 11 22:30 4.97
AOD 5 10:30 23.75
6 10:30 23.75
7 10:30 23.75
8 10:30 23.75
10 10:30 45.15
19 10:30 13.11

Fig. 4. Results of Swarm A scalar magnetic field data analysis for the Ecuador earthquake (16 April 2016) from 1st November 2015 to 30 April 2016. The
earthquake day is represented as vertical line. The green horizontal lines indicate the upper and lower bounds (M  1:5  IQR). The blue horizontal line
indicates the median value (M). The values of the median and the allowable bounds were calculated only using the quiet geomagnetic days. The x-axis
represents the day relative to the earthquake day. They-axis represents the daytime scalar magnetic field median values for each day. (For interpretation of
the references to colour in this figure legend, the reader is referred to the web version of this article.)

status of its track and found that this anomaly is also seen the confutation and track analysis indicate that the other
in other tracks outside the Dobrovolsky’s area. Therefore, detected anomalies (the third, fourth and fifth arrows in
probably this anomaly is not due to the earthquake. But, Fig. 6) could be included as potential seismo-ionospheric
 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263 255

Fig. 5. Results of Swarm A magnetic field vector X data analysis for the Ecuador earthquake (16 April 2016) from 1st November 2015 to 30 April 2016.
The earthquake day is represented as vertical line. The green horizontal lines indicate the upper and lower bounds (M  1:5  IQR). The blue horizontal
line indicates the median value (M). The values of the median and the allowable bounds were calculated only using the quiet geomagnetic days. The x-axis
represents the day relative to the earthquake day. The y-axis represents the daytime magnetic field vector X median values for each day. (For interpretation
of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 6. Results of Swarm A magnetic field vector Y data analysis for the Ecuador earthquake (16 April 2016) from 1st November 2015 to 30 April 2016.
The earthquake day is represented as vertical line. The green horizontal lines indicate the upper and lower bounds (M  1:5  IQR). The blue horizontal
line indicates the median value (M). The values of the median and the allowable bounds were calculated only using the quiet geomagnetic days. The x-axis
represents the day relative to the earthquake day. The y-axis represents the daytime magnetic field vector Y median values for each day. (For interpretation
of the references to colour in this figure legend, the reader is referred to the web version of this article.)

anomalies. Fig. 7(a) shows the recorded track close to the ation in the time series of the residual curve of the vector
Ecuador earthquake epicenter on 24 March, 23 days before Y of the magnetic field is clearly seen around the earth-
the earthquake. The earthquake epicenter, track and quake location (the dotted red ellipse in Fig. 7d). Fig. 8
Dobrovolsky’s area are shown as a red asterisk, a red line shows the daytime corrected magnetic field vector Z values
and a green circle, respectively. The track passed inside the measured using satellite A during the period of 1st Novem-
Dobrovolsky’s area between the 09:59:38 and 10:12:46 LT. ber 2015 to 30 April 2016. Three anomalies are observed on
The horizontal and vertical axes represent the geographic 18, 31 March and 9 April, i.e. 29, 16 and 7 days before the
longitude and latitude coordinates, respectively. The differ- earthquake, respectively. The confutation and track analy-
ences between the time-series of the first derivatives of the sis acknowledge that these detected anomalies could be
measured magnetic fields scalar and vectors (X, Y, Z) val- probably associated with the earthquake. For example
ues and a fitted polynomial of degree 12 along this track the red dotted ellipsoids in Fig. 9(d and e) shows the unu-
are shown in Fig. 7(b, c, d and e) respectively. The vertical sual variations in time series of the magnetic field vectors Y
axis represents the geomagnetic latitude. An unusual vari- and Z on 7 days before the earthquake. Since the satellites
256 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263

Fig. 7. Results of Swarm A track analysis for the Ecuador earthquake (16 April 2016) on 24 March 2016. (a) The earthquake epicenter, the track and
Dobrovolsky’s area are shown as a red asterisk, a red line and a green circle, respectively. The track passed the Dobrovolsky’s area between the 09:59:38
and 10:12:46 LT. The horizontal and vertical axes represent the geographic longitude and latitude. (b), (c), (d) and (e) The differences between the time-
series of the derivatives of the measured magnetic fields scalar and vectors (X, Y, Z) values, respectively, and a fitted polynomial of 12 degree along this
track. The vertical axis represents the geomagnetic latitude. (For interpretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)

Fig. 8. Results of Swarm A magnetic field vector Z data analysis for the Ecuador earthquake (16 April 2016) from 1st November 2015 to 30 April 2016.
The earthquake day is represented as vertical line. The green horizontal lines indicate the upper and lower bounds (M  1:5  IQR). The blue horizontal
line indicates the median value (M). The values of the median and the allowable bounds were calculated only using the quiet geomagnetic days. The x-axis
represents the day relative to the earthquake day. The y-axis represents the daytime magnetic field vector Z median values for each day. (For interpretation
of the references to colour in this figure legend, the reader is referred to the web version of this article.)

A and C fly close to each other, side-by-side with longitude detected in variations of magnetic field parameters mea-
separation of 1.4° at equator and an altitude close to sured by satellite C. Fig. 10 represents the daytime cor-
450 km, it is expected that the observed anomalies in differ- rected magnetic field vector Z values measured using
ent parameters of satellite A magnetic field would be satellite C during the period of 1st November 2015 to 30
 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263 257

Fig. 9. Results of Swarm A track analysis for the Ecuador earthquake (16 April 2016) on 09 April 2016. (a) The earthquake epicenter, the track and
Dobrovolsky’s area are shown as a red asterisk, a red line and a green circle, respectively. The track passed the Dobrovolsky’s area between the 08:34:37
and 08:45:14 LT. The horizontal and vertical axes represent the geographic longitude and latitude. (b), (c), (d) and (e) The differences between the time-
series of the derivatives of the measured magnetic fields scalar and vectors (X, Y, Z) values, respectively, and a fitted polynomial of 12 degree along this
track. The vertical axis represents the geomagnetic latitude. (For interpretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)

Fig. 10. Results of Swarm C magnetic field vector Z data analysis for the Ecuador earthquake (16 April 2016) from 1st November 2015 to 30 April 2016.
The earthquake day is represented as vertical line. The green horizontal lines indicate the upper and lower bounds (M  1:5  IQR). The blue horizontal
line indicates the median value (M). The values of the median and the allowable bounds were calculated only using the quiet geomagnetic days. The x-axis
represents the day relative to the earthquake day. The y-axis represents the daytime magnetic field vector Z median values for each day. (For interpretation
of the references to colour in this figure legend, the reader is referred to the web version of this article.)

April 2016. The C satellite detected anomalies that coincide method described by De Santis et al. (2017). This method
with those detected by satellite A. involves the analyses of each track separately. First, the
Swarm A satellite data (in particular Y magnetic com- data between 50 and +50 degrees of geomagnetic latitude
ponent of VFM) was also analysed according to the MASS are selected. Polar regions are excluded because they are
258 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263

influenced by external geomagnetic activity. The first differ- days. Fig. 12(b) illustrates the DTEC values estimated
ences in the 1 Hz data are calculated, and thus substantially according to the equation DTEC ¼ TECMedian
Interquartile
: Fig. 12(c)
this operation coincides with the first derivative of the Y shows the times when jDTECj > 1:5: To distinguish the
magnetic component. Next, the data is fitted by a spline likely seismo-ionospheric perturbations from the solar
and the residual is analysed by a 3° latitude moving win- and geomagnetic activities, three conditions including
dow with a 3/5 degree step. A window is defined as anoma- F10:7 < 110; 20 < Dst < 20 ðnTÞ; ap < 2:5 are joined to
lous if its standard deviation overpasses the standard jDTECj > 1:5 equation using AND operator (Fig. 12(d)).
deviation of the residual of the whole trace multiplied by As shown in Fig. 12(d) a striking TEC anomaly is seen
a threshold. For this analysis, a threshold of 3.25 is used. 1 day before the main shock at 09:00 UTC, when the
Fig. 11 represents the cumulative number of tracks that DTEC has an increase of the order of 14.29% from the nor-
contains anomaly windows 6 months before the earth- mal state.
quake and one month after the earthquake. Only the track
with quiet geomagnetic conditions are considered (|Dst|  3.5. LST anomalies
20 nT and ap  10 nT) and the analysis is extended during
all daytime. It is interesting to notice an acceleration that The NASA MODIS-LST products are available in two
occurs about 9 days before the earthquake. This behavior modes of daytime and night-time. In this study, the night-
is in agreement with the other analysis of Swarm data time LST products were used. Since the time series of LST
and supports it. variations follow from a strictly nonlinear behavior and
The characteristics of the other detected anomalies in seasonal effects, a third degree polynomial was fitted to
the variations of the scalar and vector magnetic data for the time series. Fig. 13 shows the differences between the
satellites A, B and C are listed in Table 2. time series of LST and the applied polynomial close to
the Ecuador earthquake epicenter during the period of
3.4. TEC anomalies 1st January to 30 April 2016. The horizontal axis repre-
sents the days relative to the earthquake day. The vertical
Fig. 12(a) shows the TEC variations from 1st November dotted line shows the earthquake date. Median, and higher
2015 to 30 April 2016. An asterisk represents the time of and lower bounds are drawn as green and red horizontal
the earthquake. A clear cluster of anomalies are seen on lines, respectively. The corrected LST value exceeds the
about 60 and 121 days before the earthquake. But higher bound (M þ 1:2  IQR) on 5 April 2016, 11 days
further analyses exclude these anomalies from the prior to the earthquake and reaches its maximum value
seismo-ionospheric perturbations, because the solar and of 4.97% of the higher bound (Fig. 13).
geomagnetic conditions are not quiet during the mentioned

Fig. 11. Results of Swarm A magnetic field vector Y automatic data analysis by MASS algorithm for the Ecuador earthquake (16 April 2016) from
180 days before to 30 days after the earthquake. Each black point represents an anomaly track. The cumulative number of anomaly tracks is represented.
The earthquake day is represented as a vertical dotted line. The red line is a sigmoidal fit that underlines a critical point about little more than 9 days before
the occurrence of the earthquake. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)
 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263 259

Fig. 12. Results of analysis for the Ecuador earthquake (16 April 2016) from 1st January to 30 April 2016. The earthquake time is indicated by an asterisk.
The x-axis represents the day relative to the earthquake day. The y-axis represents the time UTC (LT = UTC-5:00). (a) TEC variations, (b) DTEC
variations obtained by equation of DTEC ¼ TECMedian
Interquartile
, (c) detected TEC anomalies when jDTECj > 1:5 and (d) detected TEC anomalies when
F10:7 < 110; 20 < Dst < 20 ðnTÞ; ap < 2:5; and jDTECj > 1:5.

Fig. 13. Results of LST data analysis for the Ecuador earthquake (16 April 2016) from 1st January to 30 April 2016. The earthquake day is represented as
a dotted vertical line. The red horizontal lines indicate the upper and lower bounds (M  2:0  IQR). The green horizontal line indicates the median value
(M). The x-axis represents the day relative to the earthquake day. The y-axis represents the mean of LST values of 100  100 pixels area centered on
earthquake epicenter for each day. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)

3.6. AOD anomalies increase in AOD is clearly observed 19 days before the
earthquake. Variations of AOD clearly exceed the upper
Fig. 14 shows a time series of area averaged AOD values bound (M þ 2  IQR) of the order of 13.11% (the first
centered on earthquake epicenter during the period of 1 arrow in Fig. 14). A striking AOD anomaly is observed
January to 30 Jun 2016. The vertical dotted line shows 10 days prior to the event, when the AOD value reaches
the earthquake date and the abscissa indicates the days the maximum value during the studied period with the
related to it. Median and bounds (upper and lower) are value of 45.15% of the higher bound (the second arrow
drawn as green and red horizontal lines, respectively. An in Fig. 14). Also, a group of AOD anomalies are detected
260 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263

Fig. 14. Results of AOD data analysis for the Ecuador earthquake (16 April 2016) from 1st January to 30 June 2016. The earthquake day is represented as
a dotted vertical line. The red horizontal lines indicate the upper and lower bounds (M  2:0  IQR). The green horizontal line indicates the median value
(M). The X-axis represents the day relative to the earthquake day. The Y-axis represents the AOD values of the nearest grid point to earthquake epicenter
for each day. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

between 5 and 8 days preceding the earthquake (the third respect to the historical mean, but with less statistical sig-
arrow in Fig. 14). The detected post seismic AOD anomaly nificance than 6 March.
could be related to the strong aftershocks (the forth arrow For a deeper analysis of the anomaly, the maps around
in Fig. 14). the anomaly days are extracted and showed in Fig. 15 from
(b) to (j). Each map is the difference between the repre-
sented day and the 28 March 2016. This latter day is
3.7. ECMWF SKT anomalies selected because it is very close to the historical mean.
These maps highlight the slowly (some days) formations
The period of two months was analysed in terms of SKT of a high temperature concentration blob close to the epi-
compared with the past thirty-seven year trend. The nov- center (slightly located in South-West). The increase of
elty of the method (called CAPRI for Climatological Anal- temperature reaches 4 K more than typical values. The
ysis for earthquake PRecursors Identification) stands in the blob quickly vanishes in about 1 day.
way the complete time series is reduced, where also the pos-
sible effect of global warming is properly removed (Piscini
et al., 2017). 4. Conclusions
Before being processed, the SKT data are spatially aver-
aged using only those over the land. The adopted algorithm Ecuador is close to the equator region and therefore it is
removes the long-term trend over the whole day by day difficult to detect the variations connected with the earth-
dataset. The advantage of analysing separately each day quakes, particularly for the equatorial and low latitudinal
is to take into account a possible different seasonal local ionosphere, which is highly dynamic and unpredictable.
effect over the years. Fig. 15 describes the output of the Fig. 16 shows the times which are perturbed by the solar
data analysis. The temperature represented by blue line in geomagnetic activities in pink color during the studied per-
Fig. 15(a) is the average of the SKT over all the years. In iod. Multi-precursors analysis could be an appropriate
addition, for each day the standard deviation r of the his- solution for these regions and the non-ionospheric precur-
torical values was calculated and represented by coloured sors such as LST, AOD, meteorological data and so on,
stripes (blue, green and yellow stripes represent 1.0, 1.5 can be considered. This figure also represents an overview
and 2.0 standard deviation, respectively). The 2016 SKT of the detected anomalies using the different satellite mea-
(red line) is then compared with the historical series. To surements. A clump of anomalies are seen between 7 and
make this comparison feasible, we impose the (operational 12 days before the earthquake. The thermal anomaly LST
archive) average value in the period analysed to coincide is detected about the beginning of these groups of anoma-
with the average of the historical (ERA-interim) time ser- lies. Therefore, it can acknowledge that LAIC mechanism
ies. More details of the SKT data analysis are reported in indicates that the seismic anomalies originate from the
Piscini et al. (2017). lithosphere and then they are transferred to the atmosphere
In the analysed period the SKT overpasses the mean and ionosphere (Pulinets and Boyarchuk, 2004). A simple
plus 2.0 standard deviation on 6 March 2016 only, with explanation could be that during the preparation phase
also the previous days with higher values of SKT with of an earthquake, the emitted gas molecules lead to
 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263 261

Fig. 15. Results of ECMWF data analysis for the Ecuador earthquake (16 April 2016) of skin temperature (SKT) at about 01:00 local time. (a) shows the
time series of SKT for the area around the epicenter for 2 month from 1 March to 30 April. Red line represents the SKT for 2016. Blue line represents the
historical mean of SKT in the same period from 1979 to 2015. The blue, green and yellow stripes represent 1.0, 1.5 and 2.0 standard deviation of the
historical values, respectively. The maps from (b) to (j) represent the SKT from 1 to 9 March 2016. The choice of a deeper analysis on these days is based
on the fact that on 6 March 2016 the SKT is greater of 0.45 K than the mean plus 2.0 standard deviation. (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.)

increase the amount of AOD. After the main shock and the pendent analyses: the results of the first are shown in
following aftershocks, the earthquake energy and also dust Figs. 2–10, those of the second in Fig. 11. Both analyses
may lead to increase the AOD parameter. The Swarm mag- show a convergence of the anomalies to the earthquake
netic satellite data are investigated by two completely inde- and a concentration about 9 days before the mainshock.
262 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263

Fig. 16. An overview of the detected anomalies in this study. The non-quiet times perturbed by geomagnetic activities are shown in pink colors. The
earthquake day is represented as a dotted vertical line. The x-axis represents the day relative to the earthquake day. The y-axis represents the local time.
The time of each anomaly is at the beginning of the text. EDA: Electron Density of satellite A, ETC: Electron Temperature of satellite C, MS: Magnetic
field Scalar, MVx: Magnetic Field Vector X, MVy: Magnetic Field Vector Y, MVz: Magnetic Field Vector Z, LST: Land Surface Temperature, AOD:
Aerosol Optical Depth, MASS: MASS algorithm, ECMWF: meteorological data. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

It is interesting that the cumulative number of anomalous parameters, such as the magnetic field and the electron den-
tracks of satellite A has an unusual change about 9 days sity, which are a valuable resource to the study in the field
prior to the event and in this day the striking anomalies of earthquake precursors, and their integration would pos-
in all three Swarm satellites were detected. This suggests sibly offer a robust pre-earthquake warning system in
that some of the identified anomalies are probably induced future research studies.
by the lithospheric activity.
The observed anomalies in the measurements of Swarm Acknowledgements
satellite A sensors are usually seen in the correspondent
parameters measured by the Swarm satellite C, thanks to The authors would like to acknowledge the ESA for the
the little distance between the two satellites. The analysis Swarm data, NASA for the MODIS and TEC data and the
of the electron density data shows the anomaly only at NOAA for the geomagnetic indices. Part of this work has
night-time. This is consistent with the fact that the effi- been performed in the framework of LIMADOU-Science
ciency of the anomalous electric field penetration into the project, funded by the Italian Space Agency.
ionosphere at night is higher than in daytime (Pulinets
and Boyarchuk, 2004). Our results indicate that the vector
References
Y of the magnetic field has higher sensitivity to the seismo-
ionospheric anomalies in comparison to the other magnetic Akhoondzadeh, M., 2011. Comparative Study of the Earthquake Precur-
components, in accordance to crustal origin of these sors Obtained from Satellite Data (Ph.D. Thesis). University of
anomalies. There are a few detected anomalies in the mea- Tehran, Surveying and Geomatics Engineering Department, Remote
surements of the Swarm satellite B. It could be related to its Sensing Division.
Bilitza, D., 2001. International reference ionosphere 2000. Radio Sci. 36,
altitude, which is higher than the other two Swarm satel-
261–275.
lites. The difference of the number of anomalies detected De Santis, A., Balasis, G., Pavon-Carrasco, F.J., Cianchini, G., Mandea,
by the three satellites underlines the importance of having M., 2017. Potential earthquake precursory pattern from space: the
a large constellation of satellites to monitor the magnetic 2015 Nepal event as seen by magnetic Swarm satellites. Earth Planet.
field and the ionosphere in general. Our analysis shows that Sci. Lett. 461, 119–126.
Dee, D.P., Uppala, S.M., Simmons, A.J., Berrisford, P., Poli, P.,
the calculation of the differences between the values of the
Kobayashi, S., Andrae, U., Balmaseda, M.A., Balsamo, G., Bauer,
satellite magnetic field sensors and IGRF model is an P., Bechtold, P., Beljaars, A.C.M., van de Berg, L., Bidlot, J.,
appropriate method to eliminate the unwanted variations Bormann, N., Delsol, C., Dragani, R., Fuentes, M., Geer, A.J.,
due to the satellite orbitography and geographic location. Haimberger, L., Healy, S.B., Hersbach, H., Hólm, E.V., Isaksen, L.,
However, IRI model cannot precisely predicts the Kållberg, P., Köhler, M., Matricardi, M., McNally, A.P., Monge-
Sanz, B.M., Morcrette, J.-J., Park, B.-K., Peubey, C., de Rosnay, P.,
variations of the electron density and temperature, since
Tavolato, C., Thépaut, J.-N., Vitart, F., 2011. The ERA-Interim
these two parameters are highly variable based on the time, reanalysis: configuration and performance of the data assimilation
geographic location and magnetic activity. It should be system. Q.J.R. Meteorol. Soc. 137, 553–597. http://dx.doi.org/
noted that the Swarm satellites present different in-situ 10.1002/qj.828.
 M. Akhoondzadeh et al. / Advances in Space Research 61 (2018) 248–263 263

Dobrovolsky, I.R., Zubkov, S.I., Myachkin, V.I., 1979. Estimation of the size Parrot, M., 1995. Use of satellites to detect seismo-electromagnetic effects,
of earthquake preparation zones. Pure Appl. Geophys. 117, 1025–1044. main phenomenological features of ionospheric precursors of strong
Freund, F.T., 2011. Pre-earthquake signals: underlying physical processes. earthquakes. Adv. Space Res. 15 (11), 1337–1347.
J. Asian Earth Sci. 41, 383–400. Piscini, A., De Santis, A., Marchetti, D., Cianchini, G., 2017. A multi-
Hayakawa, M., Molchanov, O.A., 2002. Seismo-Electromagnetics: Litho- parametric climatological approach to study the 2016 Amatrice-Norcia
sphere-Atmosphere-Ionosphere Coupling. Terra Scientific Publishing (Central Italy) earthquake preparatory phase. Pure Appl. Geophys.
Co, Tokyo, p. 477. http://dx.doi.org/10.1007/s00024-017-1597-8, in press.
Mayaud, P.N., 1980. Derivation, Meaning and Use of Geomagnetic Pulinets, S.A., Boyarchuk, K.A., 2004. Ionospheric Precursors of Earth-
Indices. Geophy, 22. American Geo. Union, Washington, D.C. quakes. Springer, Berlin.
Molchanov, O.A., Hayakawa, M., 2008. Seismo Electromagnetics and Pulinets, S.A., 2009. Physical mechanism of the vertical electric field
Related Phenomena: History and Latest Results. TERRAPUB, generation over active tectonic faults. Adv. Space Res. 44, 767–773.
Tokyo, p. 189. Pulinets, S.A., Ouzounov, D., 2011. Lithosphere–Atmosphere–Ionosphere
Olsen, N., Friis-Christensen, E., Floberghagen, R., Alken, P., Beggan, C. Coupling (LAIC) model: an unified concept for earthquake precursors
D., Chulliat, A., Visser, P.N., 2013. The swarm satellite constellation validation. J. Asian Earth Sci. 41, 371–382.
application and research facility (SCARF) and swarm data products. Tronin, A., Hayakawa, M., Molchanov, O.A., 2002. Thermal IR satellite
Earth, Planets Space 64 (11), 1189–1200. http://dx.doi.org/10.5047/ data application for earthquake research in Japan and China. J.
eps.2013.07.001. Geodyn. 33, 519–534.