File PDF .it

Condividi facilmente i tuoi documenti PDF con i tuoi contatti, il Web e i Social network.

Inviare un file File manager Cassetta degli attrezzi Ricerca PDF Assistenza Contattaci



3789 Ionospheric earthquake precursors .pdf



Nome del file originale: 3789-Ionospheric-earthquake-precursors.pdf
Titolo: CR 2015-006
Autore: D.A. Rhoades

Questo documento in formato PDF 1.6 è stato generato da Adobe Acrobat Pro 10.0.0, ed è stato inviato su file-pdf.it il 01/11/2016 alle 10:33, dall'indirizzo IP 93.44.x.x. La pagina di download del file è stata vista 1179 volte.
Dimensione del file: 2.2 MB (33 pagine).
Privacy: file pubblico




Scarica il file PDF









Anteprima del documento


Ionospheric Earthquake Precursors

D.A. Rhoades
R. Buxton

C. Mueller
M.C. Gerstenberger

GNS Science Consultancy Report 2015/06
April 2015

DISCLAIMER
This report has been prepared by the Institute of Geological and
Nuclear Sciences Limited (GNS Science) exclusively for and under
contract to EQC Research Foundation. Unless otherwise agreed in
writing by GNS Science, GNS Science accepts no responsibility for
any use of or reliance on any contents of this report by any person
other than EQC Research Foundation and shall not be liable to any
person other than EQC Research Foundation, on any ground, for
any loss, damage or expense arising from such use or reliance.
Use of Data:
Date that GNS Science can use associated data: 1 January 2015

BIBLIOGRAPHIC REFERENCE
Rhoades, D.A.; Mueller, C.; Buxton, R.; Gerstenberger, M.C. 2015.
Ionospheric Earthquake Precursors, GNS Science Consultancy
Report 2015/06. 26 p.

Project Number 410W1354

EQC Research Project BIE 10/591

CONTENTS
EXECUTIVE SUMMARY.......................................................................................................III
NON-TECHNICAL SUMMARY ............................................................................................ IV
1.0

INTRODUCTION ........................................................................................................1

2.0

COMPUTATION OF TIME-VARYING TEC FIELD FROM GPS DATA ....................... 3
2.1
2.2

3.0

VARIATION OF TEC OBSERVATIONS BEFORE SELECTED LARGE
EARTHQUAKES ........................................................................................................7
3.1
3.2
3.3
3.4
3.5

4.0

OPEN SOURCE SOFTWARE TOOLKIT ...................................................................3
COMPUTATIONAL PROCEDURE ...........................................................................3

SELECTED EARTHQUAKES .................................................................................7
MAXIMUM AND MEAN TEC VERSUS TIME ............................................................7
DEGREE OF ANOMALY SCALE........................................................................... 12
SPATIAL DISTRIBUTION OF TEC ANOMALIES ..................................................... 13
COMPARISON WITH ANOMALIES OBSERVED IN PREVIOUS STUDIES ..................... 16

INDEX OF LOCAL TEC ANOMALY ......................................................................... 18
4.1
4.2

DEFINITION OF INDEX .......................................................................................18
PLOTS OF LOCAL CORRELATION INDEX ............................................................. 20

5.0

CONCLUSION ..........................................................................................................24

6.0

ACKNOWLEDGEMENTS .........................................................................................25

7.0

REFERENCES .........................................................................................................25

FIGURES
Figure 1

Location of grid points at which TEC is estimated for each time of observation. .......................... 6

Figure 2

Lower: Maximum of TEC (y-axis) over the New Zealand region versus time in the period
2010/08/17 to 2010/09/10............................................................................................................. 8

Figure 3

Lower: Maximum of TEC (y-axis) over the New Zealand region versus time in the period
2011/02/05 to 2011/03/03............................................................................................................. 8

Figure 4

Lower: Maximum of TEC (y-axis) over the New Zealand region versus time in the period
2013/07/04 to 2013/07/23............................................................................................................. 9

Figure 5

Lower: Maximum of TEC (y-axis) over the New Zealand region versus time in the period
2013/07/30 to 2013/08/26............................................................................................................. 9

Figure 6

Lower: Mean of TEC (y-axis) over the New Zealand region versus time in the period
2010/08/17 to 2010/09/10........................................................................................................... 10

Figure 7

Lower: Mean of TEC (y-axis) over the New Zealand region versus time in the period
2011/02/05 to 2011/03/03........................................................................................................... 10

Figure 8

Lower: Mean of TEC (y-axis) over the New Zealand region versus time in the period
2013/07/04 to 2013/07/23........................................................................................................... 11

Figure 9

Lower: Maximum of TEC (y-axis) over the New Zealand region versus time in the period
2013/07/30 to 2013/08/26........................................................................................................... 11

GNS Science Consultancy Report 2015/06

i

EQC Research Project BIE 10/591

Figure 10

Spatial distribution of TEC over the New Zealand region at the time of the highest
anomaly of maximum TEC preceding the Darfield earthquake. ................................................. 13

Figure 11

Spatial distribution of TEC over the New Zealand region at the time of the highest
anomaly of mean TEC preceding the Darfield earthquake. ........................................................ 14

Figure 12

Spatial distribution of TEC over the New Zealand region at the time of the highest
anomaly of maximum and mean TEC preceding the Christchurch earthquake. ......................... 15

Figure 13

Spatial distribution of TEC over the New Zealand region at the time of the highest
anomaly of mean TEC preceding the Lake Grassmere earthquake. .......................................... 16

Figure 14

Plot of local correlation index versus time for the nearest grid point to each major
earthquake (a) Darfield, (b) Christchurch, (c) Seddon, and (d) Lake Grassmere. ...................... 21

Figure 15

Spatial variation of local correlation index on 18 February 2011, the time of greatest
anomaly in the index prior to the Christchurch earthquake......................................................... 22

TABLES
Table 1

Selected major earthquakes and associated time periods for TEC analysis. ............................... 7

Table 2

Lead time and degree of anomaly of highest value of maximum and mean TEC in the
lead up to each major earthquake. ............................................................................................. 12

GNS Science Consultancy Report 2015/06

ii

EQC Research Project BIE 10/591

EXECUTIVE SUMMARY
Anecdotal studies suggest that ionospheric disturbances sometimes occur as short-term
precursors to earthquakes with a precursor time of several days. This study initiates an
examination of possible ionospheric earthquake precursors in New Zealand, with a view to
systematically testing their value for short-term earthquake forecasting.
GPSTk software produced by the University of Texas at Austin to compute Total Electron
Content (TEC) in the ionosphere from GPS network observations has been adapted to the
New Zealand region. It has been used to estimate the TEC in the ionosphere on a spatial
grid of locations surrounding the New Zealand region for selected periods preceding and
following four recent major earthquakes – the M 7.1 2010 Darfield earthquake, M6.3 2011
Christchurch earthquake, M 6.5 2013 Seddon earthquake and M 6.6 2013 Lake Grassmere
earthquake. The software gives snapshots of the spatial variation of TEC over New Zealand
at 30 second intervals.
The time series of maximum and mean regional TEC, in the lead-up to the four earthquakes,
was studied in a search for precursory anomalies. TEC has strong daily fluctuations, with
high values during daytime and near-zero values at night-time. Significant higher-thannormal TEC anomalies were found for the Darfield, Christchurch and Lake Grassmere
earthquakes with precursor times ranging from 7–11 days. Despite the similarities of the
precursor times, the spatial distributions of TEC at the times of these anomalies suggest that
they are unrelated to the earthquake concerned and more likely due to solar or geomagnetic
effects. They are different from proposed ionospheric precursors in the literature, in which the
anomalies are seen as local variations to the pattern of TEC variation close to the location of
forthcoming earthquakes.
A local correlation index was defined to identify TEC anomalies which are more likely to be
related to earthquake occurrence. This is a daily index that compares the variation of TEC
within a day at each point of the grid with the average daily variation at nearby surrounding
grid points. It is designed to distinguish local variations in TEC from the large scale variations
caused by solar and geomagnetic influences. Most values of the index are very close to one.
A value much less than one is anomalous.
The local correlation index was plotted for the same time periods before and after the four
selected earthquakes, for the grid point closest to each earthquake source. For the
Christchurch earthquake, the index showed an anomalously low value with a precursor time
of three days. For the other earthquakes no precursor anomaly was observed.
A short-term precursor can be quite informative for earthquake forecasting, even if it occurs
only for some of the major earthquakes. It is proposed that future studies should extend the
database of TEC observations to a much longer period. Then the local correlation could be
used as a systematic input for the formation of hybrid short-term earthquake forecasting
models. In that way its information value for earthquake forecasting, over and above what is
already known about how the likelihood of earthquake occurrence varies in time and space,
could be determined.

GNS Science Consultancy Report 2015/06

iii

EQC Research Project BIE 10/591

NON-TECHNICAL SUMMARY
Disturbances in the total electron content (TEC) of the ionosphere have been reported as
proposed short-term precursors to earthquakes in some regions, with a lead time time of
several days. In this project we computed TEC in the New Zealand region, with a view
towards testing the value of TEC anomalies for short term earthquake forecasting. We
adapted software produced by the University of Texas to compute TEC from GPS network
observations. We mapped TEC in the ionosphere on a spatial grid covering the whole of
New Zealand, at half-minute intervals, for several weeks before, and a few days after, four
recent major earthquakes – the 2010 Darfield, 2011 Christchurch, 2013 Seddon and 2013
Lake Grassmere earthquakes.
TEC has strong daily fluctuations, with high values during daytime and near-zero values at
night-time. Also, TEC averaged over the whole region was found to have anomalously high
spikes preceding three of the four earthquakes, with lead times ranging from 7–11 days.
However, the spatial distributions of TEC at the times of these spikes suggest that they are
probably due to solar or geomagnetic effects and not related to earthquake occurrence. They
are different from anomalies in the literature, which show local perturbations of the daily
fluctuations of TEC close to the location of forthcoming earthquakes.
A local correlation index was defined for each point in the spatial grid, to find TEC anomalies
that are more likely to be related to earthquakes. Most values of the index are very close to
one. A value much less than one is anomalous. The index showed an anomalously low value
near to the location of the Christchurch earthquake three days before its occurrence. The
lead time of three days is typical for proposed ionospheric earthquake precursors. The index
showed no local anomaly in the lead up to the Darfield, Seddon and Lake Grassmere
earthquakes.
A short-term precursor can be quite informative for earthquake forecasting, even if it occurs
only for some of the major earthquakes. Future studies should extend the database of TEC
observations to a much longer period. The local correlation index should then be used as
one possible systematic input to short-term earthquake forecasting models. Its information
value can best be measured in combination with other inputs, which include what we already
know about how the likelihood of earthquake occurrence varies in time and space.

GNS Science Consultancy Report 2015/06

iv

EQC Research Project BIE 10/591

1.0

INTRODUCTION

Anecdotal studies suggest the possible existence of a relation between Total Electron
Content (TEC) in the ionosphere and earthquake occurrence, with a lag time of about three
days between the ionospheric disturbance and the occurrence of a major earthquake. Our
aim is to assess the potential for TEC to contribute to short-term earthquake forecasting in
New Zealand, and to prepare a database derived from TEC measurements that could be
used as input to future regional earthquake likelihood models in the New Zealand
Earthquake Forecast Testing Centre.
Retrospective analyses have identified ionospheric anomalies preceding some well-known
earthquakes, including the 1999 M7.7 Chi-Chi and other earthquakes in Taiwan (Liu et al.
2000, 2001, 2002), the 2003 M6.6 San Simeon earthquake in California (Pulinets et al.,
2004), the 1995 M7.2 Kobe earthquake in Japan (Hayakawa, 2007), the 2006 Kithara
earthquake in Greece (Zakharenkova et al., 2007), the 2004 M>9 Sumatra and other
earthquakes in Indonesia (Saroso et al., 2008), and the 2008 M7.9 Wenchuan earthquake in
China (Zhou et al., 2009). We are not aware of any previous studies of ionospheric
precursors in relation to New Zealand earthquakes.
TEC can be derived from dual frequency GPS observations. The existence in New Zealand
of a substantial network of continuous GPS stations gives us the opportunity to track
changes in TEC with time and location over the New Zealand region with reasonable spatial
resolution, and hence to attempt to systematically relate such changes to earthquake
occurrence. Derived TEC measurements could add considerable value, at relatively low cost,
to the existing GeoNet database of GPS observations.
The objectives of this study are:
1.

Calculate TEC from New Zealand continuous GPS stations data, back-dating the
calculations as far as possible.

2.

Conduct a preliminary data analysis of variation in the temporal and spatial correlation
in the TEC series prior to selected New Zealand earthquakes.

3.

Compare the patterns revealed by the data analysis with those reported in similar
studies of other regions.

4.

Develop an index that reflects the strength of a localised TEC anomaly at a given
location and time. Such an index could be used as input to modify the earthquake
occurrence rate density in a regional earthquake likelihood model.

The first step, i.e., the generation of a database of TEC observations, is the major part of this
study, because it requires implementing the necessary software systems from scratch.
A key step in (2) and (4) is to discriminate ionospheric signals caused by solar and
geomagnetic disturbances from those caused by the preparation processes of earthquakes
(Pulinets and Liu, 2004; Afraimovich and Astafaya, 2008). The majority of ionospheric signals
are caused by solar or geomagnetic disturbances, and are therefore well correlated spatially
over wide regions. A minority of ionospheric signals are thought to originate from the ground.
Signals from the ground have only a local effect. Therefore, when a strong signal originates
from the ground, the signal in the region of the ionosphere above it should become decorrelated from the signal in neighbouring regions. Such de-correlation is the anomaly to look
for as a potential earthquake precursor.
GNS Science Consultancy Report 2015/06

1

EQC Research Project BIE 10/591

The study of ionospheric earthquake precursors (Liu et al., 2001, 2002) is a growing area of
research activity. It is a major branch of the burgeoning field of seismo-electromagnetics
(Uyeda et al., 2009). The whole field is somewhat controversial because of the anecdotal
nature of much of the research to date.
The TEC is one way of measuring ionospheric disturbance. It is derived from the propagation
delay experienced by GPS signals in the ionosphere, which is an important error source
affecting GPS positioning. It is calculated from the differential delay between signals on two
different carrier frequencies along the ray path between the ground receiver and the satellite,
with corrections applied for the changing ray path length as the satellite elevation angle
varies (Manucci et al., 1993; Calais and Minster, 1995; Liu et al., 2004; Surotkin et al., 2007;
Dautermann et al., 2007).
Physical explanations of coupling between seismicity and ionospheric disturbance have been
offered by Pulinets (2004). A variety of physical explanations have been given for ground
electromagnetic emissions, including stress changes in rocks through piezoelectric and
piezomagnetic effects (Bishop, 1981), induction effects due to motion of electric charges in
the geomagnetic field (Revel et al., 1999), the emission of radioactive gas or metallic ions, or
positive-hole defects in minerals (Freund, 2000). Models proposed to link electromagnetic
effects at ground level with ionospheric signals include direct coupling from the Earth’s
surface to the lower ionosphere through electric currents (Pulinets et al., 2000; Sorokin et al.,
2001) and the triggering of atmospheric gravity waves by gas releases or thermal anomalies
(Molchanov and Hayakawa, 2001; Pulinets, 2007).
Previous studies have tended to concentrate on finding anomalous ionospheric signals
before a few individual large earthquakes. This is a focus of the present study also, but it is
not the desired endpoint of our research programme into ionospheric precursors. Anecdotal
examples of ionospheric anomalies preceding large earthquakes are now sufficiently
numerous that additional anecdotes, positive or negative, can hardly affect the current status
of the science. Therefore (1) and (4) are seen as the most important objectives in this study.
The importance of objectives (2) and (3) is to give guidance on how the index in (4) should
be constructed.
The viability of empirical analyses of coupling between seismicity and TEC anomalies is
strongly affected by the spatial extent and length of time-period covered by the available TEC
series and the number of significant earthquakes in the region over the relevant time period.
The conclusions from a study based on only a few earthquakes are inevitably tentative,
whatever the outcome. The statistical confidence in the results grows as the database of
TEC measurements and earthquakes increases. This study initiates the setting up of a
database of TEC measurements and derived statistics, which, as it is augmented in time, will
eventually be capable of providing a clear result.
This study is therefore preparatory to future endeavours to make TEC in the ionosphere,
estimated from signals received at continuous New Zealand GPS stations, an authorised
database for models submitted to the New Zealand earthquake forecast testing centre. The
index developed in (4) could then be used to develop and test one or more short-term
earthquake forecasting models. By this means, the information value of ionospheric
precursors can be rigorously and transparently assessed (Schorlemmer et al., 2007).

GNS Science Consultancy Report 2015/06

2

EQC Research Project BIE 10/591

2.0

COMPUTATION OF TIME-VARYING TEC FIELD FROM GPS DATA

2.1

OPEN SOURCE SOFTWARE TOOLKIT

Total electron content has been computed using the GPSTk Open Source Toolkit for GPS
Processing,
Total
Electron
Content
Effects,
Measurements
and
Modelling
(http://www.gpstk.org, Gaussiran et al., 2004). The availability of this open source toolkit,
which is a product of the Space and Geophysics Laboratory within the Applied Research
Laboratories of the University of Texas at Austin, made the successful execution of this
project possible.
A considerable effort was first made to implement total electron content computations using
an alternative computing tool, made available to this project through our late colleague John
Beavan by Eric Calais. Dautermann, Calais et al. (2007) applied this method for a study in
southern California. Considerable technical difficulties arose in attempting to adapt this
software to New Zealand. A disadvantage of this method compared to the GPSTk toolkit is
that it provides estimates of total electron content for a single ground-based receiver station
by averaging over the satellites in its vicinity. In contrast, the GPSTk toolkit maps the TEC
over a whole region by interpolating between the points where all the signals between
ground-based receivers and satellites pass through the ionosphere. The latter approach is
much more suitable for our purposes. Once the availability of the GSPTk software suite
became known to us, the attempt to implement the method of Dautermann, Calais et al.
(2007) was abandoned.

2.2

COMPUTATIONAL PROCEDURE

This section describes how TEC is computed in GPSTk. The passages below are extracted,
with only minor modifications, from Gaussiran et al. (2004), where further details can also be
found.
The GPSTk software was developed at the Applied Research Laboratories of The University
of Texas at Austin for GPS research.
The GPSTk software suite includes an application that involves processing GPS data for
remote sensing of the ionosphere. It takes data from a large set of receivers that are
distributed within a compact geographical region and computes a contour map of the
ionospheric vertical TEC. This analysis adopts the common model assumption that the
ionosphere consists of a thin shell at a fixed height, usually 350 or 400 kilometres,
surrounding Earth. The two-dimensional (latitude, longitude) density of free electrons in this
shell is called the total electron content or TEC of the ionosphere. The TEC signal is
measured in TEC units (TECU) where one TECU = 1016 electrons/m2.
The GPS satellites broadcast a ranging signal to estimate the distance (or “range”) between
the satellite and receivers on the ground. Because the transmit time and the receive time are
different, it is impossible to measure the true range between the satellite and the receiver.
The apparent distance that is estimated is called the pseudorange, because it includes
unaccounted for receiver clock errors. The ranging signal is broadcast on dual frequency
bands from an altitude of 20,200 km, substantially above the ionosphere. Thus the GPS

GNS Science Consultancy Report 2015/06

3

EQC Research Project BIE 10/591

signals pass through the ionosphere, in general at an oblique angle. The electron content of
the ionosphere disperses and delays the signals. The delay, and the TEC along the line of
sight, or ‘slant TEC’, that produces it, is computed from the ranging signals as follows:

TEC slant =

( R1 − R2 )

α

TECUperM

(2.1)

where

α=

f1
− 1,
f2

(2.2)

and

TECUperM =

f 12
× 10 −16 .
20.48

(2.3)

Here pseudorange observations are denoted with symbol R and subscripts refer to GPS
frequencies. TECUPerM is a constant that converts metres to TECU. The constant α
depends only on the two GPS broadcast frequencies f1 and f2.
The slant TEC is related to the vertical TEC (VTEC) by an obliquity factor, computed from
simple geometric considerations.

VTEC = TEC slant × obliquity

(2.4)

obliquity = 1 − a 2

(2.5)

with

a=

R cos(e)
,
R+h

(2.6)

where e is the elevation of the satellite as seen from the receiver on the surface of the Earth,
R is the radius of the Earth and h is the height of the ionosphere.
The point at which the signal crosses the model ionospheric shell is called the ionospheric
pierce point or IPP. The IPP for a given satellite and receiver can be computed from the
receiver position and the observed elevation and azimuth angles of the satellite as seen at
the receiver, along with the height of the ionosphere. If the angle subtended at the centre of
the Earth by the receiver and the IPP is

p=

π
2

− e − sin −1 (a ) ,

(2.7)

then the latitude ϕIPP and longitude λIPP of the IPP are:

φ IPP = sin −1 (sin(φ ) cos( p) + cos(φ ) sin( p) cos(a) )

(2.8)

λ IPP = λ + sin −1 (sin) p) sin( a) / cos(φ ) ),

(2.9)

GNS Science Consultancy Report 2015/06

4

EQC Research Project BIE 10/591

where the receiver position is given by latitude and longitude (ϕ, λ ) and (e, a) are the
observed elevation and azimuth of the satellite. The radius of the IPP is the same as that of
the ionosphere, R+h.
The GPS measurement of ionospheric TEC is complicated by delays that exist in both the
satellite and the GPS receiver. These arise from differing path lengths for the two GPS signal
frequencies; they are essentially constant over short periods (days), but vary over longer
timescales. The sum of the two biases, satellite + receiver, may be estimated using GPS
data, provided the ionospheric TEC is known or modelled.
The procedure to account for these delays is first to estimate the ‘satellite plus receiver’
biases from the night-time portion of the data, when the ionospheric TEC is negligibly small.
Then the estimated biases are applied to the slant TEC measurements, from which the
vertical TEC is computed. This data, along with the estimated biases, are then used to fit a
simple model ionosphere at the points of a two dimensional grid at the ionospheric height
using a weighted least squares estimator.
The first step is pre-processing of the raw data. The GPSTk uses an extended version of the
Receiver Independent Exchange (RINEX) file format, which allows intermediate results, as
well as the raw data, to be stored within the RINEX format. The GPSTk application “ResCor”
uses dual frequency range data from the input RINEX file, obtained from GeoNet to compute
the slant ionospheric delay, the latitude and longitude of the IPP, and the elevation and
azimuth of the satellite. It stores the results in an ‘extended’ RINEX format file. The output file
includes, in the file header, a unique label for the receiver as well as the receiver position.
A GPSTk program called “IonoBias” simultaneously reads all the pre-processed files, limiting
consideration to data collected during the local night-time hours. IonoBias applies all the TEC
data to a least squares fit of both a simple model of the (night-time) ionospheric TEC and all
the biases:

VTEC = bias + obliquity × F (φ , λ ),

(2.10)

where the function F represents the model of the ionosphere. In what follows, wherever the
term “TEC” is used, it means “VTEC”.
The process of calculating TEC with the GPSTk software was adapted to the New Zealand
GPS data and automated by Christof Mueller. The automation system is implemented as a
module in the versatile python scripting language that runs GPSTk components in a batch
like process. The system runs efficiently through the steps required to create TEC maps for
any given time span. Currently it is feasible to generate TEC maps over the New Zealand
region at 30 second intervals over a two week period. Calculation times are of the order of
several tens of hours after optimization of the system through multi-threading (i.e., parallel
processing). The set of 2880 maps generated for a single day provide a detailed description
of the variation of estimated TEC over New Zealand during that day. At each time point the
TEC is estimated at 450 locations on an approximately rectangular grid, as shown in
Figure 1.

GNS Science Consultancy Report 2015/06

5

EQC Research Project BIE 10/591

Figure 1

Location of grid points at which TEC is estimated for each time of observation.

GNS Science Consultancy Report 2015/06

6

EQC Research Project BIE 10/591

3.0

VARIATION OF TEC OBSERVATIONS BEFORE SELECTED LARGE
EARTHQUAKES

3.1

SELECTED EARTHQUAKES

The computing time to execute TEC map calculations is such that it was not feasible with the
resources of this project to produce a continuous record at 30 second intervals for the whole
period for which GPS data are available. Accordingly, attention has been restricted to the
periods preceding four recent major earthquakes in New Zealand, as listed in Table 1.
Table 1

Selected major earthquakes and associated time periods for TEC analysis.

Earthquake
name

Time of
occurrence
(UT)

Location
(Lat., Long)

Depth
(km)

Magnitude
(ML)

Period of TEC
computations

Darfield

2010/09/03

-43.53,

11.0

7.1

2010/08/17–2010/09/13

16:35

172.17

2011/02/21

-43.58,

5.4

6.3

2011/02/05–2011/03/03

23:51

172.68

2013/07/21

-41.60,

12.9

6.5

2013/07/04–2013/07/31

05:09

174.33

2013/08/16

-41.73,

8.2

6.6

2013/07/30–2013/08/26

2:31

174.15

Christchurch

Seddon

Lake Grassmere

3.2

MAXIMUM AND MEAN TEC VERSUS TIME

Here we summarise the time variation of the spatial maximum of TEC over the New Zealand
region for the period preceding each of these four major earthquakes, starting two weeks
before (Figure 2–Figure 5). These plots show that there is a strong daily variation of TEC,
with daily maxima around noon and low values close to zero during the night. This pattern is
consistent with most of the TEC signal being of solar origin.
Figure 2–Figure 5 also show that in each case the highest value of the time series occurred
between 7 and 11 days before the major earthquake. In the case of the Lake Grassmere
earthquake (Figure 5), an equally high value occurred seven days after the earthquake.
Visually, the highest value appears most anomalous in the case of the Christchurch
earthquake (Figure 3) and least anomalous in the case of the Seddon earthquake (Figure 4).
Figure 6–Figure 9 show similar plots for the mean TEC over the New Zealand region. The
TEC anomaly is now visually more distinct prior to the Darfield earthquake (c.f. Figure 6 and
Figure 2), equally distinct prior to the Christchurch earthquake (c.f. Figure 7 and Figure 3),
slightly less indistinct prior to the Seddon earthquake (c.f. Figure 8 and Figure 4), and more
distinct prior to the Lake Grassmere earthquake (c.f. Figure 9 and Figure 5). In the latter case
the anomaly seven days after the earthquake is no longer comparable to the one 11 days
before it.

GNS Science Consultancy Report 2015/06

7

EQC Research Project BIE 10/591

Figure 2
Lower: Maximum of TEC (y-axis) over the New Zealand region versus time in the period
2010/08/17 to 2010/09/10. Upper: Magnitude versus time of occurrence for earthquakes in the New Zealand
region with magnitude M≥3.0, including the Darfield earthquake of 2010/09/03.

Figure 3
Lower: Maximum of TEC (y-axis) over the New Zealand region versus time in the period
2011/02/05 to 2011/03/03. Upper: Magnitude versus time of occurrence for earthquakes in the New Zealand
region with magnitude M≥3.0, including the Christchurch earthquake of 2011/02/21.

GNS Science Consultancy Report 2015/06

8

EQC Research Project BIE 10/591

Figure 4
Lower: Maximum of TEC (y-axis) over the New Zealand region versus time in the period
2013/07/04 to 2013/07/23. Upper: Magnitude versus time of occurrence for earthquakes in the New Zealand
region with magnitude M≥3.0, including the Seddon earthquake of 2013/07/21.

Figure 5
Lower: Maximum of TEC (y-axis) over the New Zealand region versus time in the period
2013/07/30 to 2013/08/26. Upper: Magnitude versus time of occurrence for earthquakes in the New Zealand
region with magnitude M≥3.0, including the Lake Grassmere earthquake of 2013/08/16.

GNS Science Consultancy Report 2015/06

9

EQC Research Project BIE 10/591

Figure 6
Lower: Mean of TEC (y-axis) over the New Zealand region versus time in the period 2010/08/17 to
2010/09/10. Upper: Magnitude versus time of occurrence for earthquakes in the New Zealand region with
magnitude M≥3.0, including the Darfield earthquake of 2010/09/03.

Figure 7
Lower: Mean of TEC (y-axis) over the New Zealand region versus time in the period 2011/02/05 to
2011/03/03. Upper: Magnitude versus time of occurrence for earthquakes in the New Zealand region with
magnitude M≥3.0 versus time, including the Christchurch earthquake of 2011/02/21.

GNS Science Consultancy Report 2015/06

10

EQC Research Project BIE 10/591

Figure 8
Lower: Mean of TEC (y-axis) over the New Zealand region versus time in the period 2013/07/04 to
2013/07/23. Upper: Magnitude versus time of occurrence for earthquakes in the New Zealand region with
magnitude M≥3.0 versus time, including the Seddon earthquake of 2013/07/21.

Figure 9
Lower: Maximum of TEC (y-axis) over the New Zealand region versus time in the period
2013/07/30 to 2013/08/26. Upper: Magnitude versus time of occurrence for earthquakes in the New Zealand
region with magnitude M≥3.0 versus time, including the Lake Grassmere earthquake of 2013/08/16.

GNS Science Consultancy Report 2015/06

11

EQC Research Project BIE 10/591

3.3

DEGREE OF ANOMALY SCALE

It is desirable to have an objective scale on which to measure the degree of anomaly of an
extreme observation. We propose the following.
For a time series, { x(i ), i = 1,  , n }, let x max denote the maximum value and xu and xl the
upper and lower quartiles, respectively. The degree of anomaly of x max relative to x is defined
as:

A( x max , x) =

( x max − xu )
.
( xu − xl )

(3.1)

In other words, the degree of anomaly of an extreme value is the difference between the
extreme and the upper quartile divided by the interquartile range. This non-parametric
definition of anomaly is analogous to the identification of outliers in box-plots (Tukey, 1977).
An outlier in a box plot is often taken as a value which is more than 1.5 times the interquartile
range above the upper quartile (or below the lower quartile). By this analogy, any extreme
value with a degree of anomaly greater than 1.5 would be regarded as an outlier of the
distribution and therefore anomalous.
The degree of anomaly of an extreme minimum value can be measured in an analogous
way, with the obvious adjustment to equation (3.1). An example of an extreme minimum
arises in Section 4.2.
The degree of anomaly is calculated for the time series of daily extremes in maximum and
mean TEC for the observation period before each major earthquake. Table 2 shows the lead
time between the highest value of TEC and the major earthquake and the degree of TEC
anomaly by this measure.
Table 2
Lead time and degree of anomaly of highest value of maximum and mean TEC in the lead up to
each major earthquake.

Earthquake

Lead time
(days)

Degree of anomaly
(Maximum TEC)

Degree of anomaly
(Mean TEC)

Darfield

11

1.63

2.24

Christchurch

7

4.10

3.75

Seddon

11

0.23

0.33

Lake Grassmere

11

0.99

4.19

According to the discussion above, the values in Table 2 indicate that the highest daily
maxima of maximum and mean TEC prior to the Darfield and Christchurch earthquakes are
anomalous by this measure. The highest daily maximum of mean TEC (but not maximum
TEC) prior to the Lake Grassmere earthquake is also anomalous. However, there is no
anomalously high daily maximum of TEC prior to the Seddon earthquake.

GNS Science Consultancy Report 2015/06

12

EQC Research Project BIE 10/591

3.4

SPATIAL DISTRIBUTION OF TEC ANOMALIES

We examine the spatial distribution of TEC at the times of the anomalies prior to the Darfield,
Christchurch and Lake Grassmere earthquakes.
Figure 10 and Figure 11 show the spatial distribution of TEC over the New Zealand region at
the times of the anomalous maximum and mean TEC, respectively, preceding the Darfield
earthquake. Figure 12 shows the spatial distribution of TEC at the time of the anomalous
maximum and mean TEC preceding the Christchurch earthquake. Figure 13 shows the
spatial distribution of TEC at the time of the anomalous mean TEC preceding the Lake
Grassmere earthquake. The times of these anomalies (U.T.) are indicated in the header of
the plots.

Figure 10
Spatial distribution of TEC over the New Zealand region at the time of the highest anomaly of
maximum TEC preceding the Darfield earthquake.

GNS Science Consultancy Report 2015/06

13

EQC Research Project BIE 10/591

In all cases the observed TEC distribution is not concentrated in the locality of the
earthquake. In fact, in all cases the highest values of TEC are far offshore in the north east or
north west of the mapped area. The lack of spatial correlation with the locations of
earthquakes downgrades the credibility of the observed anomalies as earthquake
precursors. More importantly, from the point-of-view of practical forecasting, it limits the
information value of TEC anomalies as predictors of earthquakes, since the anomalies
appear to contain no spatial information on the location of future earthquakes. Nevertheless,
an anomaly occurring with a lead time of one to two weeks before a high proportion of large
earthquakes, and rarely at other times, could still provide useful information gains.

Figure 11
Spatial distribution of TEC over the New Zealand region at the time of the highest anomaly of mean
TEC preceding the Darfield earthquake.

GNS Science Consultancy Report 2015/06

14

EQC Research Project BIE 10/591

This analysis does not provide any information on the occurrence of TEC anomalies at other
times. Therefore, based on this analysis, the usefulness of TEC anomalies as precursors of
large earthquakes cannot be ruled out. However, there is at first glance nothing in the timeof-day of the observed anomalies (noon to early afternoon local time), or in the spatial
distribution of TEC at the time of the anomalies, to suggest that the anomalies are related to
physical processes in the Earth’s crust close to the location of the subsequent major
earthquakes. To the contrary, the anomalies appear to be most likely of solar or geomagnetic
origin.

Figure 12
Spatial distribution of TEC over the New Zealand region at the time of the highest anomaly of
maximum and mean TEC preceding the Christchurch earthquake.

GNS Science Consultancy Report 2015/06

15

EQC Research Project BIE 10/591

3.5

COMPARISON WITH ANOMALIES OBSERVED IN PREVIOUS STUDIES

It is unlikely that the anomalies identified here are related to earthquake occurrence. These
anomalies are different from examples of TEC anomalies in the literature. In terms of
precursor time – the time between the observed anomaly and the subsequent major
earthquake – the times in the present study (7–11 days) are somewhat longer than those
typically reported in the literature, where the phrase “several days” is often used, or more
specifically “1–7 days” (Pulinets, Gaivoranska et al., 2004).

Figure 13
Spatial distribution of TEC over the New Zealand region at the time of the highest anomaly of mean
TEC preceding the Lake Grassmere earthquake.

GNS Science Consultancy Report 2015/06

16

EQC Research Project BIE 10/591

The anomalies recognised here are simple first-order effects, whereas in the literature a
detailed cross-correlation analysis of TEC values close to the earthquake location and those
farther away is used in order to reveal the anomalies. The type of anomaly typically observed
is an increase in the variability of the TEC signal close to the earthquake source compared to
the signals observed further away from the earthquake source; in other words, a breakdown
in the correlation between TEC signals situated inside and outside of the seismogenic zone.
Often these signals have to be extracted against a background of strong large scale temporal
and spatial variations of TEC.
Our analysis so far has not looked for signals of the type usually observed as precursory
anomalies, but only for large scale temporal and spatial variations of TEC. The earthquakerelated signals, if they exist, are to be found only by a much more careful analysis of the
temporal and spatial variability of TEC. The stronger the daily fluctuations in TEC are due to
causes of solar origin, the more subtle the earthquake-related variations will appear to be,
and the more sophisticated will be the analysis needed to extract them.

GNS Science Consultancy Report 2015/06

17

EQC Research Project BIE 10/591

4.0

INDEX OF LOCAL TEC ANOMALY

We wish to define a gridded variable (index) which could become an input to hybrid
earthquake likelihood forecasting models. Such a gridded variable would ideally be updated
once per day – the shortest time period currently used for updating earthquake forecasts in
the CSEP testing centres. In this section we propose an index which is a candidate to be
used for this purpose. However, it is beyond the resources of this study, and would require a
larger TEC data set than has been derived so far, to assess its worth for forecasting
earthquakes.
Earthquake related variations of TEC are assumed to occur on a spatial scale comparable to
the source region of a forthcoming earthquake. Normal fluctuations of TEC resulting from
solar or geomagnetic influences generally occur on a larger spatial scale. The index should
therefore be defined to distinguish smaller scale fluctuations of TEC from larger scale
fluctuations.
Pulinets et al. (2004) described a correlation technique for revealing ionospheric precursors
of earthquakes. It involves cross-correlating the intra-diurnal variations of TEC at two ground
stations – an “internal” station within the earthquake preparation area and an “external”
station outside the earthquake preparation area. The choice of the external ground station
has to be such that it is close to the internal station in geomagnetic latitude (latitude
measured in comparison the Earth’s magnetic poles), so that both stations would have
similar reactions to geomagnetic disturbances. Also the longitude of the two stations should
not differ too much because of the local time dependence of ionospheric reaction to
geomagnetic storm commencement. The terms “close to” and “too much” were not quantified
by Pulinets et al. (2004).
We adopt a similar approach here, but our locations of interest are not ground stations, but
rather points in a grid of TEC estimates in the ionosphere. We wish to define a correlation
index for each point in the grid. The index will have a daily value at each point of the grid and
will describe the correlation of TEC during the day at each point compared with that at
neighbouring points.

4.1

DEFINITION OF INDEX

Let {x(i, j , k ), i = 1,  , I , j = 1,  , J ; k = 1,  , K } denote the space-time TEC observations at a
grid of K locations over I days, at J evenly spaced times within each day. Suppose that gridpoints within a distance D of grid-point k are close enough to have similar reactions to
geomagnetic disturbances and similar ionospheric reaction to geomagnetic storm
commencement. The set of grid points within distance D of k, but not including k itself, is
denoted ND(k) and the number of grid points in this set is denoted nD(k). We define the local
correlation index by:
J

C (i, k ) =

∑ (x(i, j, k ) − x (i, k ))( y(i, j, k ) − y (i, k ))
j =1

GNS Science Consultancy Report 2015/06

( J − 1)σ x (i, k )σ y (i, k )

,

(4.1)

18

EQC Research Project BIE 10/591

where the daily mean of TEC at the kth grid location is
J

x (i, k ) = ∑
j =1

x(i, j , k )
,
J

(4.2)

the time-of-day mean of TEC in the neighbourhood of the kth grid location is

y (i, j , k ) =

x(i, j , l )
,
l∈N D ( k ) n D ( k )



(4.3)

the daily mean of TEC in the neighbourhood of the kth grid location is
J

y (i, k ) = ∑
j =1

y (i, j , k )
,
J

(4.4)

the daily variance of TEC at the kth grid location is
J

σ x (i, k ) = ∑
2

j =1

[x(i, j, k ) − x (i, k )]2 ,
J −1

(4.5)

and the daily variance of TEC in the neighbourhood of the kth grid location is
J

σ y (i, k ) 2 = ∑
j =1

[ y (i, j, k ) − y (i, k )]2 .
J −1

(4.6)

By definition, the correlation index takes values in the interval [-1,1]. We expect that the index
will nearly always be greater than zero, and that most values will be close 1, because of the
correlation induced by large scale solar and geomagnetic influences.
If the correlation index is very close to 1 for a given point in the spatial grid, there is no
deviation of the temporal pattern of TEC variation at the grid point from that in the wider area
surrounding it. In that case, it may be concluded that there are no local disturbances to the
large scale pattern of TEC created by solar and geomagnetic effects. If the correlation index
is appreciably less than 1, then there is evidence of a local disturbance to the TEC field
variation, and earthquake induced effects are a possible cause of this anomaly.
The correlation index, and indeed the TEC maps themselves, obviously depend to some
extent on the density of the GPS station network, the density of the satellite array, and the
grid density (Figure 1). These dependencies are not investigated here. However, it is notable
that the present network of continuous GPS stations in New Zealand is considerably denser
in the North Island than the South Island, and that the grid density in Figure 1 is sparse in
comparison to the GPS station densities even in the South Island. In some cases the nearest
grid point to an earthquake source may be tens of kilometres away. It is not known what
impact this may have on the ability of the index to detect precursory TEC anomalies, if they
exist.

GNS Science Consultancy Report 2015/06

19

EQC Research Project BIE 10/591

4.2

PLOTS OF LOCAL CORRELATION INDEX

Using a distance threshold of D = 250 km, we computed the local correlation index C (i, k ) for
the grid location nearest to the epicentre of each of the four major earthquakes (Table 1). For
this value of D, we have n D (k ) = 10 for most grid locations, except for those on the edge of
the grid (see Figure 1). For each major earthquake, we plotted the index as a function of time
for the grid point nearest to earthquake epicentre (Figure 14).
For the Darfield, Seddon and Lake Grassmere events there was no appreciable change in
the local correlation during the pre-earthquake and post-earthquake period. The correlation
remained close to one throughout the period plotted. For the Christchurch earthquake there
was a dip in the correlation starting 5 days before the earthquake. The correlation reached a
minimum value of 0.84 on 18 February 2011, three days before the Christchurch earthquake,
and recovered to a value close to one before the earthquake. There was a smaller dip in the
correlation five days after the earthquake.
The dip in the correlation before the Christchurch earthquake is similar to the proposed
ionospheric earthquake precursors reported in the literature. The timing of the dip fits well
within the range of one to seven days before the earthquake as reported by Pulinets et al.
(2004). The size of the dip is also within the range of examples of cross correlation of the
ionospheric F (upper) layer critical frequency foF2 between two stations – one inside the
earthquake preparation area and one outside. The critical frequency is the limiting frequency
below which a wave component is reflected by, and above which it penetrates through, an
ionospheric layer.
The background level of the correlation here is much more stable and consistently close to
one in this study than in the Pulinets et al. (2004). A difference between the two studies is
that here we are correlating 2880 observations at 30 second intervals within each daily
period, whereas Pulinets et al (2004) correlated only 24 hourly readings per day. Also, it is
possible that the interpolation and smoothing procedures that the GPSTk software uses to
compute the spatial grid of TEC values may increase the spatial correlations in the present
study.

GNS Science Consultancy Report 2015/06

20

EQC Research Project BIE 10/591

Figure 14
Plot of local correlation index versus time for the nearest grid point to each major earthquake (a)
Darfield, (b) Christchurch, (c) Seddon, and (d) Lake Grassmere. The time of the major earthquake is shown by a
dotted line.
GNS Science Consultancy Report 2015/06

21

EQC Research Project BIE 10/591

Figure 15
Spatial variation of local correlation index on 18 February 2011, the time of greatest anomaly in the
index prior to the Christchurch earthquake.

The low value of the correlation index on 18 February 2011 constitutes an anomaly in the
time series, as defined in Section 3.3, with a degree of anomaly of 26.7 (c.f. the maximum
degree of anomaly of 4.19 in Table 2). However there are other similarly low and lower
correlations at other grid points on the same day. Figure 15 shows the spatial distribution of
the local correlation index on 18 February 2011. The lowest correlations are located in a
patch covering about five degrees of latitude and four degrees of longitude to the east of the
South Island. The western edge of this patch, which has two centres – about 150 km northeast and south-east of Christchurch – includes Christchurch itself. Another anomalously low
patch is located about 600 km from Christchurch off the west coast of the South Island. Both
of these patches are part of a wider region of low correlations bordering, on three sides, a
high-correlation area of about 13 degrees in latitude and five degrees of longitude. We have
no explanation for the spatial distribution of the anomaly.
GNS Science Consultancy Report 2015/06

22

EQC Research Project BIE 10/591

Of all the days for the correlation index was calculated, 18 February shows the largest area
of low correlation and the lowest values of the local correlation index.
During the whole time period considered in relation to the Seddon and Lake Grassmere
earthquakes – from 4 July 2013 to 26 August 2013 – no value of the local correlation index
lower than 0.99 was observed for any location.
It is impossible to establish the source of any given anomaly. The localness of the anomaly
on 18 February 2011 suggests that it is not of solar or geomagnetic origin. The precursor
time of three days and the concentration of spatial anomalies at locations not far from
Christchurch suggest that the anomaly could be associated with the preparation process of
the Christchurch earthquake. However, Kane (2011) reported a strong solar flare on
15 February 2011, followed by a coronal mass ejection which caused a weak geomagnetic
anomaly on 18 February 2011, coinciding approximately with the time of the TEC anomaly
near Christchurch. It is possible that the TEC anomaly is a local effect of the geomagnetic
anomaly. However, we note that the TEC anomaly began on 16 February and ended on
19 February, whereas the geomagnetic anomaly reported by Kane (2011) began early on
18 February and lasted less than 24 hours.
Perhaps further detailed analyses of the temporal and spatial geomagnetic and TEC
variation during the time between the solar flare and the time of the Christchurch earthquake
would provide further evidence as to whether or not the TEC anomaly is related to
earthquake occurrence. However, such an analysis would probably be time consuming and
the outcome, whatever it might be, would only be of anecdotal value. The over-riding goal of
the present study is to move the study of ionospheric earthquake precursors from the
anecdotal to the testing stage. Therefore, it is less important to analyse particular examples
of precursory anomalies in detail than it is to establish the overall information gain of all
observed anomalies in earthquake forecasting. The latter objective is best achieved by
extending the present database of TEC anomalies and using it as an input for fitting hybrid
earthquake forecasting models.

GNS Science Consultancy Report 2015/06

23

EQC Research Project BIE 10/591

5.0

CONCLUSION

The present pilot study has achieved most of its objectives, although the data set of TEC
observations covers a shorter time period overall than was expected at the outset. The
adaptation of the GPSTk software for computation of ionospheric TEC from GPS
observations over New Zealand has been accomplished and has been shown to work well.
Extension of the TEC data set to a much longer period requires accessing sufficient high
performance computing resources and automating running of the software. This is a problem
that will be overcome either as high performances computing resources now available are
exploited or more resources become available to us in the future.
The local correlation index defined here has been shown to be capable of detecting local
anomalies in TEC similar to those reported in the literature. It is clear that this type of TEC
anomaly does not occur consistently before medium-to-large earthquakes, because a shortterm precursory TEC anomaly is seen for only one of the four earthquakes studied. Also,
anomalies are occasionally seen at other times and locations apparently unrelated to
earthquake occurrence. Therefore, like all other precursors identified to date, TEC anomalies
are unlikely to be a “silver bullet” for earthquake prediction in the sense of Jordan (2006).
Nevertheless, if even a minor proportion of major earthquakes have short-term ionospheric
precursors, the information gain from including them in an earthquake forecasting system
could be considerable.
It is proposed therefore to continue this line of research in the direction originally
contemplated, subject to available resources and integrated with other efforts to improve
earthquake forecasting. This will involve extending the present database of atmospheric TEC
to a longer time period, and computing and storing the grid of local correlation index for each
day. These daily grids are potential inputs to hybrid earthquake forecasts, along with the
earthquake catalogue and other data possibly related to earthquake occurrence. The value
of TEC as an earthquake precursor will then be measured by the information gain when it is
incorporated in a hybrid earthquake forecasting model (Rhoades et al., 2014) which already
includes the best model derived from all other inputs.
The distance parameter D in the local correlation index was arbitrarily chosen to be 250 km
in this study. An investigation of the best value to use could be undertaken when a larger set
of precursory anomalies is available.

GNS Science Consultancy Report 2015/06

24

EQC Research Project BIE 10/591

6.0

ACKNOWLEDGEMENTS

This study was supported by the Earthquake Commission Research Foundation under grant
number BIE 10/591. The report has been internally reviewed by Annemarie Christophersen
and Sigrun Hreinsdottir.

7.0

REFERENCES

Afraimovich, E.L. and Astafaya, E.I. 2008. TEC anomalies-Local TEC changes prior to earthquakes or
TEC response to solar and geomagnetic activity changes? Earth Planets Space, 60(9): 961–
966.
Calais, E. and Minster, J.B. 1995. GPS detection of ionospheric perturbations following the January
17, 1994, Northridge earthquake, Geophys. Res. Lett., 22: 1045–1048.
Chen, Y.I., Liu, J.Y., Tsai, Y.B. and Chen, C.S. 2004. Statistical tests for pre-earthquake ionospheric
anomaly, Terr. Atm. Ocean Sci., 15: 385–396.
Dautermann T., Calais E., Haase J. and Garrison J. 2007. Investigation of ionospheric electron
content variations before earthquakes in southern California, 2003–2004. Journal of
Geophysical Research, 112(B2): B02106, doi:10.1029/2006JB004447.
Gaussiran, T.L., Munton, T.D., Harris, B. and Tolman, B. 2004. An Open Source Toolkit for GPS
Processing, Total Electron Content Effects, Measurements and Modeling International Beacon
Satellite Symposium 2004 Trieste, Italy October, 2004.
Hayakawa, M. 2007. VLF/LF Radio Sounding of Ionospheric Perturbations Associated with
Earthquakes, Sensors 7: 1141–1158.
Jordan, T.H. 2006. Earthquake predictability, brick by brick. Seismological Research Letters, 77(1):
3–6.
Kane, R.P. 2011. Solar flare of 15 February 2011 and its geomagnetic effects, Indian Journal of Radio
& Space Physics, 40: 253–256.
Kim, V.P. and Hegai, V.V. 1997. On possible changes in the mid-latitude upper ionosphere before
strong earthquakes, J. Earthq. Predict. Res., 6: 275–280.
Liu, J.Y., Chen, Y.I., Pulinets, S.A., Tsai, Y.B. and Chuo, Y.J. 2000. Seismo-ionospheric signatures
prior to M = 6.0 Taiwan earthquakes, Geophysical Research Letters, 27: 3113–3116.
Liu, J.Y., Chen, Y.I. , Chuo, Y.J. and H.F. Tsai 2001. Variations of ionospheric total content during the
Chi-Chi earthquake. Geophys. Res. Lett., 28: 1381–1386.
Liu, J.Y., Chuo, Y.J., Pulinets, S.A., Tsai, H.F. and Zeng, X. 2002. A study on the TEC perturbations
prior to the Rei-Li, Chi-Chi and Chia-Yi earthquakes. In: Hayakawa M. and O. A. Molchanov
(Eds.), Seismo-Electromagnetics: Lithosphere-Atmosphere-Ionosphere Coupling, TERRAPUB,
Tokyo, 297-301p.
Liu, J.Y., Chuo, Y.J., Shan, S.J., Tsai, Y.B., Pulinets, S.A. and Yu, S.B. 2004. Pre-earthquake
ionospheric anomalies monitored by GPS TEC. Ann. Geophys., 22: 1585–1593.
Manucci, A., Wilson, B. and Edwards, C. 1993. A new method for monitoring the Earth’s ionospheric
total electron content using the GPS global network, in ION GPS-93, pp. 22– 24, Inst. of Navig.,
Fairfax, Va.

GNS Science Consultancy Report 2015/06

25

EQC Research Project BIE 10/591

Pulinets, S.A. 2004. Ionospheric precursors of earthquakes; recent advances in theory and practical
applications. TAO, 15(3), 413–435.
Pulinets, S. 2007. Natural radioactivity, earthquakes, and the ionosphere. EOS, 88: 217–218.
Pulinets, S.A. and Boyarchuk, K.A. 2004. Ionospheric Precursors of Earthquakes, Springer Verlag,
Berlin, Germany, 315pp.
Pulinets, S.A., Boyarchuk, K.A., Hegai, V.V., Kim, V.P. and Lomonosov, A.M. 2000. Quasielectrostatic
model of atmosphere-thermosphere-ionosphere coupling. Advances in Space Research, 26:
1209–1218.
Pulinets, S.A., Gaivoronska, T.B. and Ciraolo, L. 2004. Correlation analysis technique revealing
ionospheric precursors of earthquakes, Natural Hazards and Earth Science Systems, 4: 697–
702.
Pulinets, S.A., Legen’ka, A.D., Gaivoronskaya, T.V. and Depuev, V.Kh. 2003. Main phenomenological
features of ionospheric precursors of strong earthquakes, J. Atmos. Solar-Terr. Phys., 65:
1337–1347.
Pulinets, S.A. and Legen’ka, A.D. 2003. Spatial-temporal characteristics of the large scale
disturbances of electron concentration observed in the F-region of the ionosphere before strong
earthquakes, Cosm. Res., 41: 221–229.
Pulinets, S.A. and Liu, J.Y. 2004. Ionospheric variability unrelated to solar and geomagnetic activity,
Advances in Space Research, 34: 1926–1933.
Rhoades, D.A., Gerstenberger, M.C., Christophersen, A., Zechar, J.D., Schorlemmer, D., Werner, M.
and Jordan, T.H. 2014. Regional earthquake likelihood models: Information gains of
multiplicative hybrids. Bulletin of the Seismological Society of America, 104(6): 3072–3083.
Saroso, S., Liu, J.Y., Hattori K. and Chen, C.H. 2008. Ionospheric GPS TEC Anomalies and M >= 5.9
Earthquakes in Indonesia during 1993–2002, Terrestrial Atmospheric and Oceanic Sciences,
19(5): 481–488.
Schorlemmer, D., Gerstenberger, M.C., Wiemer, S., Jackson, D.D. and Rhoades, D.A. 2007.
Earthquake likelihood model testing. Seismological Research Letters, 78(1): 17–29.
Sorokin, V.M., Chmyrev, V.M. and Hayakawa, M. 2000. The formation of ionosphere-magnetosphere
ducts over the seismic zone. Planet. Space Sci., 48: 175–180.
Surotkin, V.A., Klimenko, V.V., and Koren’kov, Yu.N. 2007. Model calculations of the total electron
count for the GPS satellites system, International Journal of Geomagnetism and Aeronomy, 7,
GI1002, doi:10.1029/2005GI000106.
Tukey, J.W. 1977. Exploratory Data Analysis. Addison-Wesley. ISBN 0-201-07616-0. OCLC 3058187
Uyeda, S., Nagao, T. and Kamogawa, M. 2009. Short-term earthquake prediction: Current status of
seismo-electromagnetics, Tectonophysics, 470: 205–213.
Zakharenkova I.E., Shagimuratov I.I., and Krankowski, A. 2007. Features of the ionosphere behavior
before the Kythira 2006 earthquake, Acta Geophysica, 55(4): 524–534.
Zhou Y., Wu Y., Qiao X. and Zhang, X. 2009. Ionospheric anomalies detected by ground-based GPS
before the Mw7.9 Wenchuan earthquake of May 12, 2008, China Journal of Atmospheric and
Solar-Terrestrial Physics, 71(8–9): 959–966.

GNS Science Consultancy Report 2015/06

26

www.gns.cri.nz

Principal Location

Other Locations

1 Fairway Drive

Dunedin Research Centre

Wairakei Research Centre

National Isotope Centre

Avalon

764 Cumberland Street

114 Karetoto Road

30 Gracefield Road

PO Box 30368

Private Bag 1930

Wairakei

PO Box 31312

Lower Hutt

Dunedin

Private Bag 2000, Taupo

Lower Hutt

New Zealand

New Zealand

New Zealand

New Zealand

T +64-4-570 1444

T +64-3-477 4050

T +64-7-374 8211

T +64-4-570 1444

F +64-4-570 4600

F +64-3-477 5232

F +64-7-374 8199

F +64-4-570 4657


Documenti correlati


Documento PDF 3789 ionospheric earthquake precursors


Parole chiave correlate