Metadata of the Book that will be visualized online

Book Title Synchronization and Triggering: from Fracture to Earthquake Processes

Book SubTitle Laboratory, Field Analysis and Theories
Copyright Year 2010
Copyright Holder Springer-Verlag Berlin Heidelberg
Corresponding Author Family Name Chelidze
Particle
Given Name T.
Suffix
Division
Organization Institute of Geophysics
Address 1 Alexidze str, 0171, Tbilisi, Georgia
Email tamaz.chelidze@gmail.com

Author Family Name Matcharashvili

Particle
Given Name T.
Suffix
Division
Organization Institute of Geophysics
Address 1 Alexidze str, 0171, Tbilisi, Georgia
Email matcharashvili@gtu.ge
Author Family Name Melikadze
Particle
Given Name G.
Suffix
Division
Organization Institute of Geophysics
Address 1 Alexidze str, 0171, Tbilisi, Georgia
Email
Chapter 17 1

Earthquakes’ Signatures in Dynamics of Water 2

Level Variations in Boreholes 3

T. Chelidze, T. Matcharashvili, and G. Melikadze 4

f
oo
It is known that variations of water level represent an integral response of aquifer 5
to different periodic as well as non-periodic forcing, including earthquake-related

Pr
strain generation in the earth crust. Quantitative analysis of impacts of separate
components in the observed integral dynamics remains one of the main geophysical
problems. It is especially important for non-periodic processes related to the earth-
6
7
8
9
quake generation, taking into account their possible prognostic value. We can for- 10
ted
mulate the problem as a nonlinear analysis of hydrological anomalies “triggered” by 11
both the earthquake preparation and post-seismic processes. 12
In the present study, the dynamical complexity of water level variations has been 13
analyzed. The dependence of dynamics on the presence of periodic components in 14
the data records (time series) under study was investigated. Modern tools of time 15
ec

series analysis such as complexity measure and singular value decomposition 16

technique have been used. Values of Lempel-Ziv complexity of water level records 17
before and after the Spitak and Racha earthquakes, both original and reconstructed 18
by singular value decomposition, were analyzed. The main purpose was to study 19
orr

dynamical response of water level variation to increased seismic activity around 20

boreholes. Spectral characteristics, Shannon entropy and mutual information of 21
water level variation time series were calculated. It is shown that most of boreholes 22
are responding to changes caused by seismic activity, but some are not. This can be 23
c

explained by the complexity of geological and stress field structures. Sensitive 24

boreholes reveal some general features, such as an increase of the order in water 25
Un

level variability in separate boreholes and a decrease of functional relationship 26

between water level variations in pairs of different boreholes before a strong 27
earthquake. 28

T. Chelidze (*), T. Matcharashvili, and G. Melikadze

Institute of Geophysics, 1 Alexidze str, 0171 Tbilisi, Georgia
e-mail: tamaz.chelidze@gmail.com; matcharashvili@gtu.ge

V. de Rubeis et al. (eds.), Synchronization and Triggering: from Fracture to Earthquake

Processes, Geoplanet: Earth and Planetary Sciences 1,
DOI 10.1007/978-3-642-12300-9_17, # Springer-Verlag Berlin Heidelberg 2010
 T. Chelidze et al.

29 17.1 Introduction

30 Generally speaking, the water level (WL) variation in deep boreholes is caused by a
31 number of different factors. One of the most important factors is a strain change in
32 the upper Earth crust. In fact, deep boreholes represent some kind of sensitive
33 volumetric strainmeters, where water level responds to the deformations of about
34 10-7 –10-8. Hence, it is obvious that the process of water level variations will reflect
35 also the integral response of aquifer to the earthquake-related strain redistribution in
36 the Earth crust (Kumpel, 1992; Gavrilenko, 2000). The network of water regime
37 boreholes existing in Georgia allows to create a spatial picture of the strain field

f
38 and observe its evolution in the time domain. The retrospective analysis of materi-

oo
39 als shows that a characteristic annual course of levels of underground waters is
40 disrupted in the period of strong earthquakes (Spitak 1988, Racha 1991, Java 1991,
41 and Barisakho 1992). The area of compression is characterized by underground
42 waters level increasing in comparison to the normal trend, and the area of dilatation
43
44
45
46
Pr

by its decrease (Melikadze, E. Ghlonti 2000). Therefore, investigation of water
level variations may provide additional understanding of the dynamics of processes
related to earthquake preparation in the earth crust (Manga and Wang, 2009).
Nevertheless, the problem of relationship between changes in dynamics of water
47 level variation and strong earthquake preparation still remains practically unsolved
ted
48 (King, 1999).
49 In the present study we have investigated the dynamics of WL variation in the
50 network of deep boreholes on the territory of Georgia. The aim of research was to
51 clarify the character of influence of seismic processes on dynamics of water level
variation. Taking into account practical problems and scientific discussions related
ec

52
53 to understanding of seismic processes, investigation of influence of seismic activity
54 on the dynamics of water level variation is important both from scientific and
55 practical points of view.
orr

56 17.2 Methods of analysis

c

57 As the water level variation in deep boreholes is caused by a number of endogenous

Un

58 and exogenous factors, we prepared a special programme for defining the tectonic
59 component; it calculates the theoretical signal composed of a sum of reduced values
60 of atmospheric pressure, tidal variations and precipitation. In order to extract the
61 geodynamical signal, correlation analysis between the real values of WL and the
62 theoretical signal has been made (Gavrilenko 2000, Melikadze 2000).The program
63 gives a possibility to calibrate values of deformation in 10-8 range by comparison of
64 geodynamic signals with tidal variations.
65 The WL variation data sets of deep boreholes in Georgia have been analyzed
66 by modern methods of nonlinear dynamics. The network of underground water
67 regime observations was set up in the Caucasus in 1985. Since that time, a network
17 Earthquakes’ Signatures in Dynamics of Water Level Variations in Boreholes

comprising 50 sub-artesian boreholes of different depths (ranging from 250 m down 68

to 3500 m) has been functioning in the Caucasus. The Georgian network consists of 69
10 boreholes and covers the entire territory of Georgia. The present study is based 70
on analysis of hourly water level variations time series of 6 boreholes, namely: Lisi 71
(44.45 N, 21.45 E), Borjomi (43.27 N, 41.52 E), Akhalkalaqi (43.34 N, 41.22 E), 72
Ajameti (42.49 N, 42.10 E), Marneuli (44.52 N, 41.26 E) and Kobuleti (41.48 N, 73
41.47 E) boreholes. Typical records of hourly water level variations are presented in 74
Fig. 17.1. Depending on the availability and quality of data, time series of different 75
length were analyzed. The longest one covers approximately two years (01.03.1990 76
through 29.02.1992) and the shortest one covers one month (11.10.1988 through 77
12.12.1988). 78

f
At first we calculated power spectrum as well as the power spectrum regression 79

oo
exponent (Allegrini, 1995; Feder, 1988) of water level time series. In order to 80
evaluate the character of probability distribution, the Shannon entropy S: 81

X
N
S¼

i¼1 Pr
Pi logðPi Þ;

where Pi is the probability of an event to occur within box i, was calculated using 82
different time-span sliding windows (Kantz, 1997; Schreiber, 2000). Then, in order 83
ted
to evaluate the strength of functional dependence between water level variations in 84
different boreholes we used a measure of statistical independence between two 85
variables, the averaged mutual information (Cover, 1991; Kantz, 1997): 86

X  
N
PðxðiÞ; xði þ TÞÞ
ec

IðTÞ ¼ PðxðiÞ; xði þ TÞÞ log2 ;

i¼1
PðxðiÞÞPðxði þ TÞÞ

where PðxðiÞÞ and P(x(i+T)) are, respectively, the probabilities of finding xðiÞ and x
orr

87
(i+T) measurements in time series, PðxðiÞ; xði þ TÞÞ is the joint probability of 88
finding measurements xðiÞ and xði þ TÞ in time series, and T is the time lag. 89
In the present study we analyzed the integral dynamics of water level variability. 90
Therefore we avoid any linear filtration or signal separation, only nonlinear noise 91
c

reduction procedure (Kantz, 1997; Hegger, 1999) after zero mean and unite vari- 92
ance normalization have been carried out.
Un

93

17.3 Results and discussion 94

As follows from our analysis, the response of WL variation of analyzed boreholes 95

to the increased regional seismic activity varies from case to case. Indeed, it is 96
shown in Fig. 17.1 that the water level change coinciding in time with strong Racha 97
earthquake is essential for Lisi borehole. At the same time, during the same period, 98
amplitudes of variation of water level in Akhalkalaki and Kobuleti boreholes are 99
 T. Chelidze et al.

a
–3

–3.5
Water level (meters)

–4 M 6.9 Racha

–4.5
M 6.2 Djava

–5

–5.5

f
–6

oo
0 2000 4000 6000 8000 10000 12000 14000 16000
Number of observation

b –0.1 Pr
M 6.9 Racha
Water level (meters)

M 6.2 Djava
–0.2
ted
–0.3

–0.4
ec

–0.5

0 2000 4000 6000 8000 10000 12000 14000 16000

Number of observation
orr

c
–0.4
M 6.9 Racha
Water level (meters)

–0.45
M 6.2 Djava
c

–0.5
–0.55
Un

–0.6
–0.65
–0.7
–0.75

0 2000 4000 6000 8000 10000 12000 14000 16000

Number of observation

Fig. 17.1 Water level variation at (a) Lisi, (b) Akhalkalaki and (c) Kobuleti boreholes for the time
period 01.03.1990
29.02.1992
17 Earthquakes’ Signatures in Dynamics of Water Level Variations in Boreholes

rather small or even practically unchanged comparing to the Lisi borehole. This can 100
be explained by the complex character of stress field in the region (Melikadze, E. 101
Ghlonti 2000). 102
Before the Spitak event, the infringement period of the background regime 103
appeared in separate wells at different times and with different intensities: in Lisi 104
from 25.11.88; in Borjomi from 28.11.88; in Lagodechi from 28.11.88. First of all, 105
the infringements are seen on boreholes located in northern part of the structure, in 106
the zone of strongest compression, and later in the south. As a rule, the character 107
of anomaly in the compression zone has a positive bay shape (Lisi, Borjomi, 108
Lagodekhi, etc.), and in the stretching zone - the sawtooth or negative bay (Marneuli, 109
etc.). The above-mentioned sequence of infringement occurrence can be caused 110

f
by directional development of deformation processes. 111

oo
The same is proved also by the materials of Racha earthquake (29.04.91) 112
obtained from the boreholes network. The anomalies in Hydro-geo-deformation 113
(GGF) field divide the Caucasus in two parts and have been precisely fixed before 114
the event. The division between the zones passed along a deep tectonic fault of sub-

Pr
meridian orientation. The zone of largest gradient specified the place of future
disaster, coinciding with the point of junction of the fault.
Deformation processes covered the whole observed territory of Caucasus.
115
116
117
118
Despite the complicated geological structure of the territory and different strain- 119
ted
sensitivity of boreholes, the general tendency of changes is noticed in the under- 120
ground waters regime. In boreholes located in the compression field, an abnormal 121
rise of water level is observed (in Akalkalaki from 01.02 by +10 cm; in Marneuli 122
from 15.04 by þ25 cm; in Lisi from 01.02.04 by +2 m; in Borjomi from 02.04 123
by +1 m; in Lagodekhi from 03.04 by +50 cm; in Chargali from 10.04 by +30 cm). 124
ec

The greatest deformations are in the Lisi borehole. In boreholes located in the 125
stretching area, the water level decreases: Kobuleti from 20.04 by -10 cm; Gali 126
from 10.04 by -0,5 meters; and Adjameti from 0.04 by -30 cm. 127
In the stress field, the transitive zone between deformations of different signs 128
orr

extends along the above-marked deep fault and crosses the epicentral zone, where 129
the dipole deformation structure is created. Its presence confirms the existing ideas 130
about the strain distribution in the source area of earthquakes (Fig 17.10). 131
In general, the water level variation dynamics before the Spitak earthquake 132
c

is also characterized by increased extent of regularity expressed by decreased 133

Shannon entropy value, though it is not so evident as for the Racha event; 134
Un

there are still local minima (Fig. 17.3). At the same time, contrary to the Racha 135
earthquake, here, after the strong event, the dynamics of water level variability 136
in most cases becomes much more disordered than before: the Shannon entropy 137
increases. 138
During the Spitak earthquake preparation period, within several months, a 139
strong stretching sub-latitude deformation was formed in the future epicenter 140
area, and a compression area appeared to the north of it. The expanding zone 141
of stretching, besides the territory of Armenia, covers a part of southern 142
Georgia (boreholes Marneuli and Akhalkalaki) and western Azerbaijan (borehole 143
Sheki). 144
 T. Chelidze et al.

145 Within three months, the decrease of water level on the borehole Akhalkalaki
146 against the annual “background” reached 20 cm, that makes a half of the maximal
147 amplitude of seasonal fluctuations. On the Marneuli borehole, the decrease was
148 much larger (up to 50 cm), and almost the whole amplitude of seasonal fluctuation
149 was leveled by tensile deformation. On the Sheki borehole, the difference was up to
150 30 cm. In the compression zone, there is located the Borjomi borehole, where
151 during six months the difference of the level, in comparison to the average, makes
152 þ1.5 meters, and Lisi, where it is up to þ1 meter. The Lagodechi borehole changes
153 are of the same order. During some period of time, the intensity increases; at the
154 critical point of deformation, the lunar-solar tidal variations disappear in the
155 underground waters, and the impact of barometric pressure disappears too. At this

f
156 time, the strong endogenous factors which influence the water bearing horizon

oo
157 muffle the exogenous ones.
158 The appearance of anomalies before the Spitak event was distributed in time:
159 Lisi - from 25.11.88; Borjomi - from 28.11.88; Lagodechi - from 28.11.88; Sheki -
from 4.12.88; Novkhvani - from 4.12.88; Shemakcha - from 6.12.88. It is evident
160
161
162
163
Pr
that the anomaly first of all manifests itself in the boreholes located in the northern
part of the structure, in the strongest compression zone, and later on in the south. As
a rule, the anomaly in the compression zone has a positive bay shape (Lisi, Borjomi,
164 Lagodekhi, Novkhvani, Shemacha etc.) and in the stretching zone it looks like the
ted
165 sawtooth or negative deviation (Marneuli, Sheki etc.). Thus, there is some regular
166 pattern of the anomalous field evolution (Fig. 17.11).
167 All the changes described above are obviously related to the seismic activity,
168 because in most cases they concur with strong earthquakes. Such a relation is
169 evident from an analysis of the longest water level variability time series available
ec

170 for us. In Fig. 17.4, Shannon entropy for time series containing 16300 readings
171 (01.03.1990 through 29.02.1992) analyzed by two-week span sliding windows are
172 presented. It is shown that the dynamics of water level variation undergoes clear
173 changes both before and after strong earthquakes. As it is presented in Fig. 17.4 a,
orr

174 over about 9-5 month period of time before the strong Racha earthquake, water
175 level variation in Lisi borehole becomes strongly ordered. The Shannon entropy
176 value of water level variation noticeably decreases for this period clearly preceding
177 the strong earthquake. At the same time, immediately before the earthquake (over
c

178 30-15 days period) the water level variation became maximally disordered; the
179 entropy value has substantially increased.
Un

180 Generally similar is the situation for the long time series of Akhalkalaki borehole
181 (Fig. 17.4 b), where about 8 months before the Racha earthquake the water level
182 variation for short time period has minimal (but higher than for the Lisi borehole
183 and for a much shorter time period) value of Shannon entropy.
184 Immediately before the strong earthquake, the water level variation in Akhalk-
185 alaki borehole also became maximally disordered and is characterized by practi-
186 cally the same extent of Shannon entropy as for the Lisi borehole. At the same time,
187 the water level variation in Kobuleti borehole does not show the features observed
188 for the other two boreholes (Fig. 17.4 c), which can be explained by a relatively
189 large epicentral distance (King, 1999).
17 Earthquakes’ Signatures in Dynamics of Water Level Variations in Boreholes

Thus, in two out of the three long time series analyzed, the dynamics of water 190
level variation becomes more irregular two-three weeks before a strong earthquake. 191
At the same time, it is important to mention that some boreholes may by insensitive 192
(King, 1999) to changes caused by regional seismic activity (see, e.g., the results for 193
Kobuleti borehole in Fig. 17.4 c). 194
In spite of the observed differences in absolute values of amplitudes of water 195
level in different boreholes, the dynamics of variation still reveals some interesting 196
features which may be related to regional seismic activity. All the water level 197
variation time series analyzed are characterized by a broadband power spectrum. 198
By their exponents of power spectrum regression, the process of water level 199
variation in 20.5-day duration time series (600 data), generally can be attributed 200

f
to the coloured type of noise, both before and after strong earthquakes (see, e.g., 201

oo
Tables 17.1 and 17.2). The obtained values of power spectrum exponents are 202
typical for processes where the low frequency events (taking place on long time- 203
scales) dominate in the total variability compared to the high frequency component 204
(Pimm, 1988). In this respect, as it follows from Tables 17.1 and 17.2, there are no

Pr
significant differences in spectral characteristics of water level variations during or
after strong earthquakes (excluding water level variation in Ajameti borehole for
rhe Spitak and Borjomi borehole for the Racha earthquakes).
205
206
207
208
Thus, spectral characteristics of water level variability do not react on the level 209
ted
of regional seismic activity which significantly increased for the analyzed time 210
period. 211
In spite of this integral insensitivity, we tried to clarify some details of the fine 212
dynamical structure of process of interest on the shorter time scales. For this 213
purpose, the Shannon entropy values of 10-day-span sliding windows of longer 214
ec

water level variation data sets (including time periods both before and after large 215
earthquakes) have been calculated. In Fig. 17.2, it is shown that the dynamics of 216
water level variability on 10-day time scale undergoes noticeable changes both 217
before and after the Racha earthquake. The same is true for the aftershock of the 218
orr

Table 17.1 Power spectrum regression exponents of water level variation before and after the t1:1
c

Spitak earthquake
Axalkalaki Borjomi Ajameti Lisi t1:2
Un

Before
1.59  0.08
1.69  0.09
2.130.02
1.74  0.08 t1:3

After
1.66  0.08
1.83  0.03
1.87  0.07
1.82  0.03 t1:4

Table 17.2 Power spectrum regression exponents of water level variation before and after the t2:1
Racha earthquake
Axalkalaki Borjomi Ajameti Lisi t2:2
Before
1.81 0.09
1.57 0.08
1.84  0.05
1.80  0.01 t2:3
After
1.86  0.07
1.79  0.05
1.93 0.08
1.82  0.03 t2:4
 T. Chelidze et al.

Fig. 17.2 Shannon entropy a

values of (a) Lisi, (b) 3
Lagodekhi, (c) Akhalkalaki,

Shannon entropy
2.5
and (d) Ajameti boreholes
water level hourly records 2 M 6.9 Racha
calculated for 240 data (10 1.5
M 6.2 Djava
day time span) sliding
window at 24 hour step. 1
Time of observation: 0.5
01.03.1991
30.06.1991
0
200 500 800 1100 1400 1700 2000 2300 2600 2900
Number of sliding window

f
b

oo
3
Shannon entropy

2.5

1.5

0.5
Pr M 6.9 Racha
M 6.2 Djava

0
ted
200 500 800 1100 1400 1700 2000 2300 2600 2900
Number of sliding window
c
3 M 6.9 Racha
Shannon entropy

2.5
ec

M 6.2 Djava
2

1.5

1
orr

0.5

0
200 500 800 1100 1400 1700 2000 2300 2600 2900
Number of sliding window
c

d 3
Un

2.5
Shannon entropy

M 6.9 Racha
2
M 6.2 Djava
1.5

0.5

0
200 500 800 1100 1400 1700 2000 2300 2600 2900
Number of sliding window
17 Earthquakes’ Signatures in Dynamics of Water Level Variations in Boreholes

Fig. 17.3 Shannon entropy a

values of (a) Lisi, (b) 3
Lagodekhi, (c) Borjomi, and

Shannon entropy
(d) Marneuli boreholes water 2.5
level hourly records M 6.9 Spitak
2
calculated for 240 data
(10 day time span) sliding 1.5
window at 1 hour step.
Time of observation: 1
11.10.1988
12.12.1988
0.5
200 300 400 500 600 700 800
b Number of sliding window

f
3

oo
Shannon entropy

2.5 M 6.9 Spitak

1.5

0.5
200
Pr
300 400 500 600 700 800
Number of sliding window
ted
c
3
Shannon entropy

2.5
M 6.9 Spitak
2
ec

1.5

0.5
orr

200 300 400 500 600 700 800

d Number of sliding window
3
Shannon entropy

2.5
c

2 M 6.9 Spitak
Un

1.5

0.5
200 300 400 500 600 700 800
Number of sliding window

Racha earthquake, M.6.2 Djava (15.06.1991) event. As shown in Fig. 17.2, in 219
almost all the cases, the Shannon entropy of water level variability approaches 220
their local extremes before the strong earthquake. 221
 T. Chelidze et al.

a
2.5

M 6.9 Racha
Shannon entropy 2

1.5

1 M 6.2 Djava

0.5

f
0

oo
0 2000 4000 6000 8000 10000 12000 14000 16000
Number of sliding window

b
2.5

2 Pr M 6.9 Racha
Shannon entropy

M 6.2 Djava

1.5
ted
1

0.5

0
ec

0 2000 4000 6000 8000 10000 12000 14000 16000

Number of sliding window

c
3
orr

M 6.9 Racha
2.5
Shannon entropy

M 6.2 Djava
2
c

1.5
Un

0.5

0
0 2000 4000 6000 8000 10000 12000 14000 16000
Number of sliding window

Fig. 17.4 Shannon entropy values of (a) Lisi, (b) Akhalkalaki and (c) Kobuleti boreholes water
level hourly records calculated for 360 data (two-week time span) sliding window at 24 hour step.
Time of observation: 01.03.1990
29.02.1992
17 Earthquakes’ Signatures in Dynamics of Water Level Variations in Boreholes

It is also important to say that on the analyzed time scale and for the tested period 222
of observation, the WL variation after the strong earthquake comes back to the 223
background level. 224
As there are important similarities in dynamical responses of water level varia- 225
tions for different boreholes to regional seismic activity, we have investigated the 226
strength of functional dependence between them. For this purpose, we calculated 227
averaged mutual information, the well known measure of statistical independence 228
between two variables (Cover, 1991; Kantz, 1997). The mutual information value is 229
the most suitable parameter for these purposes, becauseunlike the linear correlation 230
function, it takes into account nonlinear correlations too (Hegger, 1999). It is shown 231
in Fig. 17.5 that the strength of functional relationship between water level varia- 232

f
bility in pairs of boreholes before, during, as well as after strong earthquake 233

oo
undergoes noticeable changes. The water level variability in different boreholes is 234
maximally de-correlated over about two-to-three weeks period before the Racha 235
event (mutual information values have their minima). It is important that this strong 236
earthquake as well as its aftershock is preceded by brief sharp changes in extent of
interdependence of variability in boreholes.
Pr
In most cases, the above-mentioned sharp changes are characteristic for func-
tional interdependence between water level variations in boreholes during seismic
237
238
239
240
activity related to the Spitak earthquake (see Fig. 17.6). At the same time, the 241
ted
transition between functionally independent and dependent states are not as sharp 242
and clear as for the Racha earthquakes. It is interesting that after the Spitak 243
earthquake the dynamics of WL variation in different boreholes becomes more 244
functionally dependent than before (values of mutual information increase). 245
The above-mentioned features of dynamics of water level variations are espe- 246
ec

cially noticeable for the longest available time series. Indeed, as shown in 247
Fig. 17.7 a and b, the dynamics of water level variability in Lisi borehole reveals 248
a clear decrease in the extent of functional interdependence for several months 249
before the strong earthquake. At the same time, as it was said above, some bore- 250
orr

holes may be not sensitive to changes caused by seismic activity (King, 1999), e.g., 251
water level variability in Akhalkalaki and Kobuleti boreholes do not reveal changes 252
(Fig. 17.7 c), which can be explained by the character of regional stress field during 253
seismic activity. 254
c

Before and during Racha and Spitak earthquakes it became possible to monitor 255
spatial evolution of deformation processes and to define the anomalous zones 256
Un

relative to the background daily course of underground waters level (Figs. 17.10 257
and 17.11). This analysis bears information on the seismic event’s approaching 258
time. In the Racha earthquake, the compression zone anomalies are expressed by 259
the suppression of tide effects, first of all in the boreholes located in the zone of 260
the strongest gradient (Lisi - from 10.03. Chargali - from 05.04. Marneuli – from 261
08.04.) and further on boreholes located to the east, in the deformation zone with 262
lesser gradient (from 15.04), and later on - to the east on the territory of Azerbaijan 263
(Fig. 17.10). 264
The majority of data confirms the abnormal behavior of WL in boreholes on the 265
territory of the whole Caucasian region before and after strong earthquakes. The 266
 T. Chelidze et al.

Fig. 17.5 Mutual a

Information values calculated 0.9

Mutual Information
for 300 data length sliding 0.8
0.7 M 6.9 Racha
windows of water level time 0.6
series of pairs of boreholes: 0.5 M 6.2 Djava
(a) Lisi vs. Lagodekhi, 0.4
0.3
(b) Lisi vs. Ajameti, 0.2
(c) Lisi vs. Akhalkalaki, 0.1
(d) Akhalkalaki vs. 0
200 700 1200 1700 2200 2700
Ajameti, and (e) Ajameti Number of sliding window
vs. Lagodekhi. Time b 1.2

Mutual Information
of observation: M 6.9 Racha
1
01.03.1991
30.06.1991.

f
0.8
Sliding window at 1 hour step

oo
0.6
M 6.2 Djava
0.4
0.2
0
200 700 1200 1700 2200 2700
c
Pr
0.7
Number of sliding window
M 6.9 Racha
Mutual Information

0.6
0.5
0.4
ted
0.3
M 6.2 Djava
0.2
0.1
0
200 700 1200 1700 2200 2700
Number of sliding window
d
ec

0.7 M 6.9 Racha

Mutual Information

0.6
0.5 M 6.2 Djava
0.4
0.3
orr

0.2
0.1
0
200 700 1200 1700 2200 2700
Number of sliding window
e
c

0.8 M 6.9 Racha

Mutual Information

0.7
0.6
Un

0.5 M 6.2 Djava

0.4
0.3
0.2
0.1
0
200 700 1200 1700 2200 2700
Number of sliding window

267 example of regional effect is the identical behavior of boreholes Lisi and Esentuki,
268 where compression processes adjusted the exogenous effects two months before the
269 Racha event.
17 Earthquakes’ Signatures in Dynamics of Water Level Variations in Boreholes

Fig. 17.6 Mutual a

Information values calculated 1.2

Mutual Information
for 300 data length sliding 1
windows of water level time 0.8 M 6.9 Spitak
series of pairs of boreholes: 0.6
(a) Lisi vs. Marneuli,
0.4
(b) Lisi vs. Lagodekhi,
(c) Lisi vs. Borjomi, (c) 0.2
Lagodekhi vs. Marneuli, and 0
300 400 500 600 700 800
(d) Borjomi vs. Lagodekhi.
b Number of sliding window
Time of observation:
1.2
11.10.1988
12.12.1988.

Mutual Information
Sliding window at 1 hour step 1

f
0.8 M 6.9 Spitak

oo
0.6
0.4
0.2
0

c
Pr
300

1.2
400 500 600 700
Number of sliding window
800
Mutual Information

1
0.8
0.6 M 6.9 Spitak
ted
0.4
0.2
0
300 400 500 600 700 800
d Number of sliding window
ec

1.2
Mutual Information

1
0.8 M 6.9 Spitak
0.6
orr

0.4
0.2
0
300 400 500 600 700 800
Number of sliding window
e
c

1.2
Mutual Information

1
Un

0.8 M 6.9 Spitak

0.6
0.4
0.2
0
300 400 500 600 700 800
Number of sliding window

In a transitive zone (borehole Adjameti), the strong anomalies were not found. 270
The presence of complex mosaic structure of deformation is a possible explanation 271
of the fact that anomaly at the borehole Oni is fixed only one month before the 272
events - from 04.04. 273
 T. Chelidze et al.

a
0.6 M 6.9 Racha

Mutual Information 0.5

0.4
M 6.2 Djava
0.3

0.2

0.1

f
0

oo
0 2000 4000 6000 8000 10000 12000 14000 16000
Number of sliding window

b 0.7 M 6.9 Racha

0.6
Pr M 6.2 Djava
Mutual Information

0.5

0.4
ted
0.3

0.2

0.1

0
ec

0 2000 4000 6000 8000 10000 12000 14000 16000

Number of sliding window

c
0.6
orr

M 6.2 Djava
0.5 M 6.9 Racha
Mutual Information

0.4
c

0.3
Un

0.2

0.1

0
0 2000 4000 6000 8000 10000 12000 14000 16000
Number of sliding window

Fig. 17.7 Mutual Information values calculated for 300 data length sliding windows of water
level time series of pairs of boreholes: (a) Lisi vs. Akhalkalaki, (b) Lisi vs. Kobuleti, and (c)
Akhalkalaki vs. Kobuleti. Time of observation: 01.03.1990
29.02.1992. Sliding window at
1 hour step
17 Earthquakes’ Signatures in Dynamics of Water Level Variations in Boreholes

This figure will be printed in b/w

f
oo
Pr
Fig. 17.8 Graph of a tidal (the bottom line), atmospheric pressure (the top line) and the under-
ted
ground water level (the middle line) variations in time. Vertical lines show earthquakes occurred in
this period
ec

Coeff corr: top skv and top(tidal+atmos). Period in hours: 24

120
Coef. correlation
Covariance
orr

100

80

60
c

40
This figure will be printed in b/w
Un

20

–20

–40

–60
5200 5300 5400 5500 5600 5700 5800

Fig. 17.9 Change of values of correlation coefficient between values of a water level and the sum
of reduced values of atmospheric pressure and tidal variations
 T. Chelidze et al.

This figure will be printed in b/w

Fig. 17.10 3-D and 2-D models of evolution of stress field during preparation of the Racha
earthquake of 29.04.1991 (red lines – main Caucasian faults)

f
oo

This figure will be printed in b/w

Pr
ted

Fig. 17.11 3-D and 2-D models of evolution of stress field during preparation of the Spirak
earthquake of08.12.1988 (red lines – main Caucasian faults)
ec

274 In the stretching zone, the deviation from the background is marked as a weak
orr

275 distortion of the tide effect (borehole Adjameti from 4.04; Gali from 03.04;
276 Sukhumi - 08.04). On boreholes Akalkalaki and Kobuleti, which are located far
277 from the epicenter, the amplitudes of barometric fluctuation and tide effects have
278 been gradually increasing since the beginning of March. According to the North-
Caucasian data, a similar anomaly was marked on the borehole Light-blue Lakes.
c

279
280 Such anomalies are caused by the amplification of aquifer’s reaction to exogenous
Un

281 processes and easing of horizontal stress due to endogenous processes.

282 17.4 Conclusions

283 The dynamics of water level variability is strongly sensitive to the borehole and
284 earthquake source location. Therefore, not all the boreholes react similarly to the
285 changes caused by tectonic/seismic activity in the analyzed region. In sensitive
286 boreholes, dynamical changes both before as well as after strong earthquakes are
17 Earthquakes’ Signatures in Dynamics of Water Level Variations in Boreholes

evident. When approaching time of earthquake occurrence, in most of the observed 287
cases, the water level variability becomes more and more regular comparing to the 288
preceding, seismically quiet time period. At the same time, the regularity in water 289
level variability is again distorted immediately prior to the earthquake occurrence. 290
In spite of observations of seismicity-related dynamical changes in water level 291
variability, there are essential qualitative and quantitative differences in the dynam- 292
ics of water level variability in separate boreholes, which supposedly are strongly 293
dependent on geological and strain field peculiarities of the considered area. 294

f
References 295

oo
Allegrini, P., Barbi, M., Grigolini, P., West, B.J. 1995. A Dynamical Mode the DNA Sequences. 296
Phys. Rev. E 52, 5281 297
Cover, T. M., Thomas, J. A., 1991. Elements of Information Theory. J.W.S, NY 280 pp. 298
Feder, J. 1988. Fractals Plenum Press, New York

Pr
Gavrilenko. P., G. Melikadze., Chelidze, T., Gibert, D., Kumsiashvili, G. 2000. Permanent water
level drop associated with Spitak earthquake: observations at Lisi borehole and modeling.
Geophys. J. Int., 143, 83-98.
Hegger, R., Kantz, H., Shreiber, T., 1999. Practical implementation of nonlinear time series
299
300
301
302
303
methods. The TISEAN package, Chaos. 9. P. 413-440. 304
ted
Kantz, H; Schreiber T., 1997. Nonlinear time series analysis. C. U. P.Cambridge 350 pp. 305
King, C. Y., Azuma, S., Igarashi, G., Ohno, M., Saito, H., Wakita, H. 1999. Earthquake – related 306
water-level changes at 16 closely clustered wells in Tono, central Japan, J. Geoph. Res. 104, 307
B6, 13073-13082 308
Kumpel, H. 1994. Evidence for self-similarity in the harmonic development of earth tides, in: 309
Kruhl, J.H. (Ed), Fractals and dynamic systems in geoscience, Springer, Berlin, 213-220.
ec

310
Manga, M and Wang C.-Y. 2009. Earthquake Hydrology. In: Earthquake Seismology, H. Kanamori 311
(Ed). Elsevier. P. 293-320. 312
Pimm, S., Redfearn, A. 1988. The variability of natural populations. Nature, 334, 613-614. 313
Schreiber, T. 2000. Measuring Information Transfer. Phys. Rev. Lett., 85:461 314
orr

Melikadze, G., Ghlonti, E. 2000. Some features of hydrogeodeformation field in Caucasus during 315
strong earthquakes preparation period. Journal of Georgian Geophysical Society, issue A. 316
Physics of Solid Earth, v. 5, 106-111. 317
c
Un