noise-based ballistic wave passive seismic monitoring ... noise nased ballistic.pdf · 1...
TRANSCRIPT
Noise-based Ballistic Wave Passive Seismic Monitoring – Part 1: Body-waves
Journal: Geophysical Journal International
Manuscript ID GJI-18-1026.R1
Manuscript Type: Research Paper
Date Submitted by the Author: 17-Apr-2019
Complete List of Authors: Brenguier, Florent ; Univ. Grenoble Alpes, ISTerreCourbis, Roméo; Univ. Grenoble Alpes, ISTerre; SisprobeMordret, Aurélien; Massachusetts Institute of Technology, Department of Earth, Atmospheric and Planetary SciencesCampman, Xander; Shell International Exploration and Production B.V.Boué, Pierre; Univ. Grenoble Alpes, ISTerreChmiel, Małgorzata; Sisprobe; Univ. Grenoble Alpes, ISTerreTakano, Tomoya; Tohoku Univ., Solid Earth Physics LaboratoryLecocq, Thomas; Royal Observatory of Belgium, Seismology - GravimetryVan der Veen, Wim ; Nederlandse Aardolie Maatschappij BVPostif, Sophie; Shell International Exploration and Production B.V.Hollis, Daniel; Sisprobe
Keywords:
Volcano monitoring < VOLCANOLOGY, Interferometry < GEOPHYSICAL METHODS, Seismic noise < SEISMOLOGY, Earthquake interaction, forecasting, and prediction < SEISMOLOGY, Induced seismicity < SEISMOLOGY
Geophysical Journal International
Noise-based ballistic wave seismic monitoring
1
Noise-based Ballistic Wave Passive Seismic Monitoring – Part 1: Body-waves 1
2
F. Brenguier1, R. Courbis1,4, A. Mordret2, X. Campman3, P. Boué1, M. Chmiel4,1, T. 3
Takano5,1, T. Lecocq6, W. Van der Veen7, S. Postif3, and D. Hollis4 4
1Institut des Sciences de la Terre, Univ. Grenoble Alpes. 5
2Massachusetts Institute of Technology. 6
3Shell. 7
4Sisprobe. 8
5Solid Earth Physics Laboratory, Tohoku Univ. 9
6Royal Observatory of Belgium 10
7Nederlandse Aardolie Maatschappij. 11
12
Corresponding author: Florent Brenguier ([email protected]) 13
Université Grenoble Alpes 14
ISTerre 15
CS 40700 16
38058 GRENOBLE Cedex 9 17
FRANCE 18
Phone: +33 673 35 68 77 19
Page 1 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
2
Summary 20
Unveiling the mechanisms of earthquake and volcanic eruption preparation requires improving 21
our ability to monitor the rock mass response to transient stress perturbations at depth. The 22
standard passive monitoring seismic interferometry technique based on coda-waves is robust but 23
recovering accurate and properly localized P and S-velocity temporal anomalies at depth is 24
intrinsically limited by the complexity of scattered, diffracted waves. In order to mitigate this 25
limitation, we propose a complementary, novel, passive seismic monitoring approach based on 26
detecting weak temporal changes of velocities of ballistic waves recovered from seismic noise 27
correlations. This new technique requires dense arrays of seismic sensors in order to circumvent 28
the bias linked to the intrinsic high sensitivity of ballistic waves recovered from noise 29
correlations to changes in the noise source properties. In this work we use a dense network of 30
417 seismometers in the Groningen area of the Netherlands, one of Europe’s largest gas fields. 31
Over the course of 1 month our results show a 1.5 % apparent velocity increase of the P-wave 32
refracted at the basement of the 700 m thick sedimentary cover. We interpret this unexpected 33
high value of velocity increase for the refracted wave as being induced by the swings in 34
groundwater charge and discharge in a carbonate layer with water conductive fracture networks 35
at 700 m depth. We also observe a 0.25 % velocity decrease for the direct P-wave travelling in 36
the near-surface sediments but conclude that it might be partially biased by changes in time in 37
the noise source properties. The perspective of applying this new technique to detect localized 38
continuous variations of seismic velocity perturbations at a few kilometers depth paves the way 39
for improved in situ earthquake, volcano and producing reservoir monitoring. 40
41
Keywords: monitoring, seismic interferometry, earthquakes, volcanoes, producing reservoirs 42
Page 2 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
3
1 Introduction 43
Large earthquakes and volcanic eruptions result from long-lasting, steady, pressure 44
buildup on faults and magmatic reservoirs. However, the triggering mechanisms of impending 45
events are thought to be initiated by short time scale stress and pore pressure transients 46
associated with tectonic and volcanic interactions (e.g. Bouchon et al. 2011, Brenguier et al. 47
2014, Khoshmanesh & Shirzaei 2018) and possibly environmental perturbations (e.g. Johnson et 48
al. 2017). Anthropogenic activities such as hydrocarbon extraction, waste-water disposal, CO2 49
storage and geothermal production also induce fluid pore-pressure related deformation that can 50
lead to the triggering of induced seismicity (Talwani 2007, Ellsworth 2013, Chang and Segall 51
2016). Monitoring these stress and pore-pressure perturbations continuously in time with high 52
spatial accuracy at depth is thus critical to foresee forthcoming catastrophic tectonic and volcanic 53
events and to improve reservoir management. 54
55
Seismic velocities are sensitive to stress and pore-pressure perturbations. This has been 56
described in-situ for a wide range of processes like for example solid earth tides (Takano et al. 57
2014) and rainfall induced pore-pressure changes (Wang et al. 2017). Niu et al. (2008) used 58
active seismic source monitoring in boreholes at 1 km depth along the San Andreas Fault and 59
were able to observe the first clear preseismic seismic velocity-stress perturbations in the hours 60
preceding the occurrence of nearby earthquakes. Although convincing, these observations have 61
never been reproduced due to the costly and logistically complicated in-situ experiment. 62
63
Over the last decade, passive noise-based seismic monitoring proved success for 64
monitoring volcanoes (Sens‐Schönfelder et al. 2006, Brenguier et al. 2008a, Donaldson 2017), 65
Page 3 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
4
earthquakes (Brenguier et al. 2008b) and environmental/climate (Lecocq et al. 2017) changes. 66
Even though some attempts were made, no one succeeded in using passive seismic monitoring to 67
observe clear preseismic anomalies similar to those described by Niu et al. (2008). The standard 68
coda-based technique suffers from shortcomings that limit our ability to detect localized seismic 69
velocity perturbations at depth. This technique also referred to as coda-wave interferometry 70
(Poupinet et al. 1984) has the advantage of being very stable thanks to its low sensitivity to noise 71
source property changes (Colombi et al., 2014). It thus allows detecting weak temporal changes 72
of seismic velocities as small as 0.01 % such as those associated with solid Earth tides for 73
example (Mao et al. 2019). However, the counterpart of this high detection capability is that the 74
complexity of coda-wave propagation limits our ability to precisely characterize both the type (P 75
or S) of velocity change and accurate estimates of their spatial distribution at depth. 76
In this work, we propose a complementary monitoring approach that uses ballistic waves 77
reconstructed from noise correlations. This paper focuses on ballistic body-waves and the 78
companion paper Mordret et al. 2019 focuses on using ballistic surface-waves on the same 79
dataset with the same type of approach. Body-waves have a specific sensitivity to seismic 80
velocity changes at depth and are potentially less affected by near-surface environmental changes 81
than surface waves. By using ballistic waves instead of coda-waves, we can more easily model 82
the spatial sensitivity of the temporal change observations to local velocity perturbations at 83
depth. The drawback of using direct, ballistic body-waves instead of coda-waves is their strong 84
sensitivity to noise source temporal variations (Colombi et al. 2014). We use dense seismic 85
networks and azimuthal averaging to circumvent this issue but still need to carefully analyze the 86
stability of noise sources for such type of analysis. 87
Page 4 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
5
Different studies showed that body-wave extraction from noise correlations is possible at 88
various scales (Roux et al. 2005, Draganov at al. 2009, Poli et al. 2012). Nakata et al. (2015) 89
were able to implement the first passive 3-D P-wave velocity tomography from continuous 90
ground motion recorded on a dense array of more than 2500 seismic sensors installed at Long 91
Beach (California, USA). Recently Brenguier et al. (2016) and Nakata et al. (2016) proved the 92
temporal stability of direct virtual body-waves between dense arrays on Piton de la Fournaise 93
volcano thus opening the way for continuous, passive ballistic wave monitoring. 94
95
In this paper we describe the fundamental aspects of passive ballistic wave monitoring 96
using dense arrays and further illustrate its potential by applying it to a network of 417 seismic 97
stations in the Netherlands. We are able to measure temporal changes of apparent velocities from 98
both direct and refracted P-waves, and thus we are able to separate the response of the near 99
surface sediments and the basement located at 700 m depth. By providing direct observations of 100
the rock mass response to stress changes at depth, this new passive seismic approach paves the 101
way for improved in situ earthquake, volcano and producing reservoir monitoring. 102
103
2 Methods 104
105
The new approach is based on measuring temporal changes of apparent slowness of specific 106
ballistic waves that have been reconstructed from noise correlations using dense arrays of 107
seismic sensors (Boué et al., 2013b, Mordret et al. 2014, Nakata et al., 2015, Nakata et al., 2016). 108
The underlying requirement is that, as for Nakata et al. (2015), the high number of seismic 109
sensors (> 100) and thus of noise correlation receiver pairs allows for the reconstruction of a 110
virtual shot-gather of sufficiently high quality to be able to isolate and measure the apparent 111
Page 5 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
6
velocity of ballistic waves such as direct P or S-waves, refracted waves or surface waves with 112
clear mode separation (Mordret et al. 2019). 113
114
For this purpose of properly extracting high quality ballistic waves, we gather all possible 115
noise correlations for all receiver pairs from a dense network into a single seismic panel 116
(propagation time versus virtual source-receiver offset) thus considering a 1D velocity model for 117
which the apparent slowness of a reconstructed ballistic wave measured at surface can be written 118
as: 119
120
𝑛 =1
𝑉=
∆𝑥𝑡
∆𝑥 121
122
where n is the apparent slowness and V the apparent velocity. We can estimate n as the slope of 123
apparent arrival times 𝑡 along distance 𝑥 under the assumption that the apparent velocity 𝑉 is 124
uniform along distance range ∆𝑥. We are now interested in measuring a temporal change in 125
apparent slowness ∆𝐶𝑡𝑛, the index 𝐶𝑡 being for Calendar time (Fig. 1a): 126
127
∆𝐶𝑡𝑛 = 𝑛𝐶𝑡2 − 𝑛𝐶𝑡1 = (∆𝑥𝑡
∆𝑥)
𝐶𝑡2− (
∆𝑥𝑡
∆𝑥)
𝐶𝑡1 128
129
The temporal change of apparent slowness can be measured directly as the difference of 130
slowness estimates (from a slant-stack or beam-forming analysis for example) at different 131
calendar times (de Cacqueray et al. 2016). Here we use an approach that estimates the temporal 132
Equation 1
Equation 2
Page 6 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
7
change of apparent slowness as the slope of the linear regression of the travel time shifts at 133
different calendar times for each distance offset along distance ∆𝑥 (Fig. 1b): 134
135
∆𝐶𝑡𝑛 =∆𝑥(∆𝐶𝑡𝑡)
∆𝑥 136
137
where ∆𝐶𝑡𝑡 are the measurements of travel time shifts at different offsets ∆𝑥 for two different 138
calendar times and for a specific windowed ballistic wave. This approach shows the advantage of 139
providing a direct estimate of the uncertainty of the estimated apparent slowness temporal 140
change by assessing how the measured travel time delays ∆𝐶𝑡𝑡 fit a linear regression model along 141
distance range ∆𝑥. 142
143
From equation 2, we can derive the relation linking the temporal change of apparent slowness 144
and apparent velocity: 145
∆𝐶𝑡𝑛 = −∆𝐶𝑡𝑉
𝑉2 146
147
By multiplying the above equation by V, the apparent uniform velocity of the studied ballistic 148
wave, this equation leads to (Fig 1c): 149
150
∆𝐶𝑡𝑛 × 𝑉 =∆𝑥(∆𝐶𝑡𝑡)
∆𝑥𝑡= −
∆𝐶𝑡𝑉
𝑉 151
This later equation shows that, for small velocity perturbations (<10%) we can estimate the 152
relative temporal change of apparent velocity, ∆𝐶𝑡𝑉
𝑉, of a given ballistic wave by estimating the 153
value of the slope of the linear regression of ∆𝐶𝑡𝑡 measurements along travel time 𝑡. This 154
Equation 3
Equation 4
Equation 5
Page 7 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
8
approach is sketched on Fig. 1. It applies to any kind of ballistic arrivals that can be clearly 155
identified and isolated on a virtual source gather section. The case of direct surface waves 156
requires mode separation and is challenging because of dispersion that leads to different 157
velocities for different periods at which ∆𝐶𝑡𝑡 values are measured (Mordret et al. 2019). As 158
suggested in Nakata et al. (2016), this approach can also be applied to noise-correlations between 159
two distant arrays in order to monitor diving body-waves probing magmatic reservoirs, seismic 160
faults or producing reservoirs at a few hundreds to a few kilometers depth. It is interesting to 161
note that in this situation of two distant arrays referred to as A and B, the estimates of temporal 162
changes of apparent velocities can be achieved using noise-correlations between arrays A and B 163
and B and A separately, thus providing two independent estimates of temporal velocity changes. 164
This method can also be applied using noise correlations between an individual seismic station 165
and a distant array. In this work we apply this approach to the monitoring of a direct and a 166
refracted wave using noise-correlations within a single array of about 8 km wide. 167
168
169
Page 8 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
9
170
Figure 1. Procedure for ballistic-wave apparent seismic velocity monitoring. a) propagation of a 171
direct ballistic wave. The dashed lines show the reference wave and the plain lines show the 172
wave affected by a velocity perturbation of -7 %. b) travel time shifts measurements and linear 173
regression along distance. c) conversion from distance to travel time by dividing distance by V, 174
the apparent velocity of the propagating wave. 175
3 Data 176
The Groningen gas field located in the northeast of the Netherlands is one of Europe’s 177
largest natural gas field. The reservoir located at 3 km depth is thought to be 40 by 50 km wide 178
and 250 m thick. Bourne et al. (2018) show that the gas production in this field led to a 15 MPa 179
average reservoir pore-pressure depletion since 1995 which is associated with seismicity rates 180
that increased as an exponential-like function. 181
182
Page 9 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
10
We use a network of 417 short period seismic stations deployed in the Groningen area of 183
the Netherlands (Fig. 2) from 11 February (day 42) to 12 March (day 71) 2017 over a time span 184
of 30 days. The network forms a grid array with aperture of the order of 8 km and an average 185
station distance of about 400 m. 186
187
188
189
Figure 2. a) Geometry of the 417 short-period stations used in this study. b) beamforming of the 190
30 days of continuous seismic data in the 1-4 seconds period range. 191
192
We computed an averaged seismic section of vertical to vertical noise cross-correlations. The 193
noise-correlations were stacked in time (over the 30 days of continuous data), in space (along 194
distance bins of 50 meters long) and for the causal and acausal parts following Boué et al. 195
(2013a) and Nakata et al. (2015). Figure 3 illustrates these binned data for two different 196
Page 10 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
11
frequency ranges (1-3 Hz and 3-12 Hz). The low frequency (1-3 Hz) section mainly shows the 197
propagation of low-velocity Rayleigh waves and also of ballistic P-waves at velocities between 198
1.5 and 3 km/s. The lower panel of Fig. 3 highlights the high frequency (3-12 Hz) P-waves 199
interpreted as (1) the direct, diving P-wave with velocity of ~1700 m/s (see model on the right 200
panel) contoured by a blue box and (2) a refracted wave at the 700 m interface with apparent 201
velocity of ~3300 m/s contoured by a red box. Thanks to the high stability in time of the useful 202
high-frequency ambient seismic noise, we are finally able to reconstruct repetitive in time 203
seismic sections from the correlations of daily records of ambient seismic noise leading us to 204
daily virtual shot seismic gathers. 205
206
207
Figure 3. Noise cross-correlation averaged binned section for frequency ranges 1- 3 Hz (a) and 208
3-12 Hz (b). The blue and red dashed boxes correspond to the selected windows used for the 209
analysis for the direct and refracted waves. The right panel (c) illustrates the average P-wave 210
velocity model of the area illustrating the velocities of the overburden (saturated sediments) 211
Page 11 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
12
around 1700 m/s and of the bedrock at ~700 m depth around 3000 m/s. It has been defined using 212
sonic logs from deep wells in the area (Kruiver et al. 2017). 213
214
4 Analysis and Results 215
In order to measure temporal changes, we further isolate the direct and refracted waves 216
by applying a tapered window (Fig. 4a,b). For the time shift measurements, ∆𝐶𝑡𝑡, we use a 217
cross-spectrum approach (Clarke et al. 2011) in the frequency range 3 to 8 Hz. We illustrate our 218
approach on Figure 4 by plotting travel time shifts for the direct and refracted waves between 219
references (stacks of days 42 to 50) and a current seismic sections (days 53 to 62 for the direct 220
and days 48 to 57 for the refracted waves). Even though the travel time shifts measurements 221
show large fluctuations likely associated with imperfect direct and refracted waves 222
reconstruction and noise source changes through time, they show a clear linear trend along 223
distance, especially for the refracted wave, indicating a clear change in apparent velocity 224
between these two time periods. By multiplying the slope of the ∆𝐶𝑡𝑡 over distance linear 225
regression by the apparent velocity of the direct (1700 m/s) and refracted (3300 m/s) waves (step 226
b) to c) on Fig. 1), we find velocity changes of -0.25 % and +1.5% for the direct and refracted 227
waves. 228
229
Page 12 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
13
230
231
Figure 4. Travel time shifts along distance plots. Left a), windowed reference direct waves 232
averaged for the time period (days 42 to 50). b) windowed current direct waves averaged for 233
days 53 to 62. c) travel time shift measurements, ∆𝐶𝑡𝑡, along distance between these two direct 234
waves. Right a), windowed reference refracted waves averaged for the time period (days 42 to 235
50). b) windowed current refracted waves averaged for days 48 to 57. c) travel time shift 236
measurements, ∆𝐶𝑡𝑡, along distance between these two refracted waves. 237
238
239
240
In order to gain insights on the temporal evolution of velocity changes, we further average the 241
daily seismic sections using a 10-days long, 1-day moving window. This leads us to 21, full 10-242
days averaged, daily seismic sections (from days 42 to 71). Following the method described 243
above, we measure velocity changes (∆𝐶𝑡𝑉
𝑉) between each 10-days averaged seismic section and 244
the reference section corresponding to a stack of days 42 to 50 for both the direct (Fig. 5a) and 245
refracted (Fig. 5b) waves. The apparent velocity change curves shown on Figure 5 indicate a 246
velocity decrease of maximum -0.25 % for the direct wave and a velocity increase of maximum 247
1.5 % for the refracted wave. The error bars correspond to the uncertainty of the linear 248
Page 13 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
14
regressions estimates of the travel time shifts, ∆𝐶𝑡𝑡, over travel time t (Fig. 1c) using a least-249
square approach following Brenguier et al. 2008a. 250
251
252
253 254
255
5 Discussions and conclusions 256
257
The most common source of error in passive, noise-based seismic monitoring results 258
from the non-stationarity of noise sources. Furthermore, ballistic waves are much more sensitive 259
to noise source variations than coda waves (Colombi et al. 2014). As a result, the main drawback 260
Figure 5. Apparent seismic
velocity temporal changes for both
the direct (top) and refracted
(bottom) waves.
Page 14 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
15
of this ballistic wave based monitoring method is the potential error introduced by changes in 261
noise sources and great care has to be taken when interpreting the results. We thus need to 262
properly assess how noise source temporal azimuthal variations can hamper our results. To do so 263
we beamform the ambient seismic noise on a 10-days average basis in the period range 1 to 4 264
seconds (Fig. 2). Due to the minimum inter-station distance of 300 m we are not able to 265
beamform the noise in the frequency range of interest (3-8 Hz) and we thus make the assumption 266
that the 1-4 s beams are representative for the higher frequency noise characteristics. We found 267
that the noise source distribution is strongly anisotropic with a main spot coming from the North 268
Sea, north of our array (Fig. 2). Interestingly, the high frequency (3-8 Hz) noise correlations are 269
quite asymmetric proving that most of the high-frequency body-waves also come from the North 270
of our array pointing to shore break as a possible source of high-frequency noise. Our 271
assumption of comparing the high-frequency (3-8 Hz, likely shore break) to the low-frequency 272
(1-4 s, likely ocean microseismic) noise sources might not be fully relevant but still informative. 273
We observe that during the time span of our analysis, the noise source distribution in the 1-4 s 274
period range is mostly stable with small azimuthal variations of the centroid of the main spot by 275
less than 10°. Following the theoretical predictions of travel time errors of ballistic arrival 276
reconstructed from correlations of non-isotropically distributed noise sources from Weaver et al. 277
(2009) and the further applications of Froment et al. (2010) and Colombi et al. (2014), we assess 278
that, in case of a two sensors noise correlation, the error on the travel time shift measurements 279
(Fig. 5) would lead to a velocity change uncertainty of less than 0.5 %. In our case we 280
emphasize that our travel time shift measurements are obtained from azimuthally averaged noise 281
correlations from the dense 417 stations array. We are thus confident that the observed velocity 282
changes of +1.5 % for the refracted wave is mostly related to physical velocity changes. 283
Page 15 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
16
However, the velocity decrease of 0.25 % for the direct wave might thus be partly biased by 284
changes in the noise source properties. 285
In order to interpret the apparent velocity change of -0.25 % of the direct P-wave, we can 286
consider a simple 1D model of wave propagation in the sedimentary overburden and the ray 287
theory that predicts that the observed apparent velocity change on the direct P-wave corresponds 288
to a real P-wave velocity perturbation averaged over the first 200 m depth (maximum distance 289
between virtual source and stations of 2200 m). The only noticeable event during that time 290
period is an episode of rainfall that occurred on day 53 in the region (20 mm of cumulated water 291
height). The decrease of velocity for the direct wave could thus be related to a poro-elastic 292
process as described by Rivet et al. (2015) and Wang et al. (2017). We will address the study of 293
these shallow surface temporal velocity variations measurements in more details including the 294
analysis of surface waves in a companion paper. 295
Following again the ray theory, the increase of apparent velocity of 1.5 % for the 296
refracted wave can be directly attributed to a change of P-wave velocity of the carbonate bedrock 297
at 700 m depth. Considering this hypothesis of a bedrock velocity increase, one simple 298
interpretation could be the effect of loading from rainfall on the bedrock that would increase the 299
confining pressure and close cracks in the bedrock. However, it is unlikely that 2 cm of 300
additional water height on day 53, corresponding to an increase of loading of 0.2 kPa, leads to a 301
velocity increase of 1.6 % at 700 m depth. Indeed, following Yamamura et al. (2003), and their 302
observation of velocity-stress sensitivity of 10-7 Pa-1 the expected velocity change for a loading 303
of 0.2 kPa should be about 0.01 %. Moreover, we checked for INSAR and GPS observations. 304
These data don’t show any significant transient anomalies during the time span of our analysis. 305
Finally, we also preclude the potential effects of swings in gas production in the main reservoir 306
Page 16 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
17
at 3 km depth below our study area due to the large distance to the probed carbonate layer. These 307
induced pressure variations are of the order of less than 0.1 MPa locally and are thus too small to 308
potentially induce a 1.5 % velocity increase 2.3 km above in the carbonate layer. We also rule 309
out the effects of local induced earthquakes due to the low level of seismicity during the studied 310
time period. The most likely interpretation for this velocity increase relates to movements of 311
fluids in water conductive fracture networks within the carbonate layer at about 700 m depth. 312
This is discussed in a companion paper Mordret et al. 2019 that includes additional passive 313
monitoring observations using direct surface waves. 314
In conclusions, even though our results are hampered by high uncertainties and are 315
spanning a too short period to be interpreted properly, this new passive ballistic wave seismic 316
monitoring approach has the potential for revealing seismic wave velocity temporal variations at 317
localized areas at depth thus acting as a stress-strain probe. Even though the main drawback of 318
this technique is that it requires dense seismic networks, we believe that this technique together 319
with the recent step change in seismic instrumentation will lead to groundbreaking advances in 320
our understanding of natural and induced earthquakes, volcanic eruptions and will prove useful 321
for reservoir management. 322
323
Acknowledgments 324
We acknowledge two anonymous reviewers for their useful comments. This project received 325
funding from the Shell Game Changer project HiProbe. We also acknowledge support from the 326
French ANR grant T-ERC 2018, (FaultProbe), the European Research Council under grants no. 327
817803, FAULTSCAN and no. 742335, F-IMAGE and the European Union’s Horizon 2020 328
research and innovation program under grant agreement No 776622 (PACIFIC). AM 329
Page 17 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
18
acknowledges support from the National Science Foundation grants PLR-1643761. The data 330
were provided by NAM (Nederlandse Aardolie Maatschappij). We acknowledge M. Campillo, 331
N. Shapiro, G. Olivier, P. Roux, S. Garambois, R. Brossier and C. Voisin for useful discussions. 332
The authors thank Nederlandse Aardolie Maatschappij and Shell for permission to publish. 333
334
References 335
Bouchon, M., Karabulut, H., Aktar, M., Özalaybey, S., Schmittbuhl, J., & Bouin, M. P. (2011). 336
Extended nucleation of the 1999 Mw 7.6 Izmit earthquake. science, 331(6019), 877-880. 337
338
Boué, P., Roux, P., Campillo, M., & de Cacqueray, B. (2013a). Double beamforming processing 339
in a seismic prospecting context. Geophysics, 78(3), V101-V108. 340
341
Boué, P., Poli, P., Campillo, M., Pedersen, H., Briand, X., & Roux, P. (2013b). Teleseismic 342
correlations of ambient seismic noise for deep global imaging of the Earth. Geophysical Journal 343
International, 194(2), 844-848. 344
345
Bourne, S. J., Oates, S. J., & van Elk, J. (2018). The exponential rise of induced seismicity with 346
increasing stress levels in the Groningen gas field and its implications for controlling seismic 347
risk. Geophysical Journal International, 213(3), 1693-1700. 348
349
Brenguier, F., Shapiro, N. M., Campillo, M., Ferrazzini, V., Duputel, Z., Coutant, O., & 350
Nercessian, A. (2008a). Towards forecasting volcanic eruptions using seismic noise. Nature 351
Geoscience, 1(2), 126. 352
353
Brenguier, F., Campillo, M., Hadziioannou, C., Shapiro, N. M., Nadeau, R. M., & Larose, E. 354
(2008b). Postseismic relaxation along the San Andreas fault at Parkfield from continuous 355
seismological observations. Science, 321(5895), 1478-1481. 356
357
Brenguier, F., Campillo, M., Takeda, T., Aoki, Y., Shapiro, N. M., Briand, X., ... & Miyake, H. 358
(2014). Mapping pressurized volcanic fluids from induced crustal seismic velocity drops. 359
Science, 345(6192), 80-82. 360
361
Brenguier, F., Kowalski, P., Ackerley, N., Nakata, N., Boué, P., Campillo, M., ... & Roux, P. 362
(2016). Toward 4D Noise-Based Seismic Probing of Volcanoes: Perspectives from a Large-N 363
Experiment on Piton de la Fournaise Volcano. Seismological Research Letters, 87(1), 15-25. 364
365
de Cacqueray, B., Roux, P., & Campillo, M. (2016). Using slowness and azimuth fluctuations as 366
new observables for four‐dimensional reservoir seismic monitoring. Geophysical Prospecting, 367
64(6), 1537-1555. 368
Page 18 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
19
369
Chang, K. W., & Segall, P. (2016). Injection‐induced seismicity on basement faults including 370
poroelastic stressing. Journal of Geophysical Research: Solid Earth, 121(4), 2708-2726. 371
372
Clarke, D., Zaccarelli, L., Shapiro, N. M., & Brenguier, F. (2011). Assessment of resolution and 373
accuracy of the Moving Window Cross Spectral technique for monitoring crustal temporal 374
variations using ambient seismic noise. Geophysical Journal International, 186(2), 867-882. 375
376
Colombi, A., Chaput, J., Brenguier, F., Hillers, G., Roux, P., & Campillo, M. (2014). On the 377
temporal stability of the coda of ambient noise correlations. Comptes Rendus Geoscience, 378
346(11), 307-316. 379
380
Donaldson, C., Caudron, C., Green, R. G., Thelen, W. A., & White, R. S. (2017). Relative 381
seismic velocity variations correlate with deformation at Kīlauea volcano. Science advances, 382
3(6), e1700219. 383
384
Draganov, D., X. Campman, J. Thorbecke, A. Verdel, and K. Wapenaar, 2009. Reflection 385
images from ambient seismic noise, Geophysics, 74 (5), A63–A67. 386
387
Ellsworth, W. L. (2013). Injection-induced earthquakes. Science, 341(6142), 1225942. 388
Johnson, C. W., Fu, Y., & Bürgmann, R. (2017). Seasonal water storage, stress modulation, and 389
California seismicity. Science, 356(6343), 1161-1164. 390
391
Froment, B., Campillo, M., Roux, P., Gouedard, P., Verdel, A., & Weaver, R. L. (2010). 392
Estimation of the effect of nonisotropically distributed energy on the apparent arrival time in 393
correlations. Geophysics, 75(5), SA85-SA93. 394
395
Khoshmanesh, M., & Shirzaei, M. (2018). Episodic creep events on the San Andreas Fault 396
caused by pore pressure variations. Nature Geoscience, 1. 397
398
Kruiver, P. P., van Dedem, E., Romijn, R., de Lange, G., Korff, M., Stafleu, J., ... & Doornhof, 399
D. (2017). An integrated shear-wave velocity model for the Groningen gas field, The 400
Netherlands. Bulletin of Earthquake Engineering, 15(9), 3555-3580. 401
402
Lecocq, T., Longuevergne, L., Pedersen, H. A., Brenguier, F., & Stammler, K. (2017). 403
Monitoring ground water storage at mesoscale using seismic noise: 30 years of continuous 404
observation and thermo-elastic and hydrological modeling. Scientific Reports, 7(1), 14241. 405
406
Mao, S., Campillo, M., van der Hilst, R. D., Brenguier, F., Stehly, L., & Hillers, G. (2019). High 407
temporal resolution monitoring of small variations in crustal strain by dense seismic arrays. 408
Geophysical Research Letters, 46(1), 128-137. 409
410
Mordret, A., Shapiro, N. M., & Singh, S. (2014). Seismic noise‐based time‐lapse monitoring of 411
the Valhall overburden. Geophysical Research Letters, 41(14), 4945-4952. 412
413
Page 19 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
20
Mordret, A., Courbis, R., Brenguier, F., Chmiel, M., Garambois, S., Mao, S., ... & Hollis, D. 414
(2019). Noise-based Ballistic Wave Passive Seismic Monitoring – Part 2: Surface Waves. 415
Geophysical Journal International, submitted. 416
417
Nakata, N., J. P. Chang, J. F. Lawrence, and P. Boué, (2015). Body-wave extraction and 418
tomography at Long Beach, California, with ambient-noise tomography, J. Geophys. Res., 120, 419
1159–1173, doi:10.1002/2015JB01870. 420
421
Nakata, N., Boué, P., Brenguier, F., Roux, P., Ferrazzini, V., & Campillo, M. (2016). Body‐and 422
surface‐wave reconstruction from seismic‐noise correlations between arrays at Piton de la 423
Fournaise volcano. Geophysical Research Letters, 43, 1047-1054. 424
425
Niu, F., Silver, P. G., Daley, T. M., Cheng, X., & Majer, E. L. (2008). Preseismic velocity 426
changes observed from active source monitoring at the Parkfield SAFOD drill site. Nature, 427
454(7201), 204. 428
429
Poli, P., M. Campillo, H. Pedersen, and L. W. Group, 2012. Body-wave imaging of Earth’s 430
mantle discontinuities from ambient seismic noise, Science, 338, 1063–1065. 431
432
Poupinet, G., Ellsworth, W. L., & Frechet, J. (1984). Monitoring velocity variations in the crust 433
using earthquake doublets: An application to the Calaveras Fault, California. Journal of 434
Geophysical Research: Solid Earth, 89(B7), 5719-5731. 435
436
Rivet, D., Brenguier, F., & Cappa, F. (2015). Improved detection of preeruptive seismic velocity 437
drops at the Piton de La Fournaise volcano. Geophysical Research Letters, 42(15), 6332-6339. 438
439
Romanet, P., Bhat, H. S., Jolivet, R., & Madariaga, R. (2018). Fast and slow slip events emerge 440
due to fault geometrical complexity. Geophysical Research Letters, 45, 4809–4819. 441
442
Rost, S., & Thomas, C. (2002). Array seismology: Methods and applications. Reviews of 443
geophysics, 40(3). 444
445
Roux, P., Sabra, K. G., Gerstoft, P., Kuperman, W. A., & Fehler, M. C. (2005). P‐waves from 446
cross‐correlation of seismic noise. Geophysical Research Letters, 32(19). 447
448
Sens‐Schönfelder, C., & Wegler, U. (2006). Passive image interferometry and seasonal 449
variations of seismic velocities at Merapi Volcano, Indonesia. Geophysical research letters, 450
33(21). 451
452
Shapiro, N. M., Campillo, M., Stehly, L., & Ritzwoller, M. H. (2005). High-resolution surface-453
wave tomography from ambient seismic noise. Science, 307(5715), 1615-1618. 454
455
Silverii, F., D'Agostino, N., Métois, M., Fiorillo, F., & Ventafridda, G. (2016). Transient 456
deformation of karst aquifers due to seasonal and multiyear groundwater variations observed by 457
GPS in southern Apennines (Italy). Journal of Geophysical Research: Solid Earth, 121(11), 458
8315-8337. 459
Page 20 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic wave seismic monitoring
21
460
Snieder, R., Grêt, A., Douma, H., & Scales, J. (2002). Coda wave interferometry for estimating 461
nonlinear behavior in seismic velocity. Science, 295(5563), 2253-2255. 462
463
Takano, T., Nishimura, T., Nakahara, H., Ohta, Y., & Tanaka, S. (2014). Seismic velocity 464
changes caused by the Earth tide: Ambient noise correlation analyses of small‐array data. 465
Geophysical Research Letters, 41(17), 6131-6136. 466
467
Talwani, P., Chen, L., & Gahalaut, K. (2007). Seismogenic permeability, ks. Journal of 468
Geophysical Research: Solid Earth, 112(B7). 469
470
Wang, Q. Y., Brenguier, F., Campillo, M., Lecointre, A., Takeda, T., & Aoki, Y. (2017). 471
Seasonal Crustal Seismic Velocity Changes Throughout Japan. Journal of Geophysical 472
Research: Solid Earth, 122(10), 7987-8002. 473
474
Weaver, R., Froment, B., & Campillo, M. (2009). On the correlation of non-isotropically 475
distributed ballistic scalar diffuse waves. The Journal of the Acoustical Society of America, 476
126(4), 1817-1826. 477
478
Yamamura, K., Sano, O., Utada, H., Takei, Y., Nakao, S., & Fukao, Y. (2003). Long‐term 479
observation of in situ seismic velocity and attenuation. Journal of Geophysical Research: Solid 480
Earth, 108(B6). 481
Page 21 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
1
Noise-based Ballistic Wave Passive Seismic Monitoring – Part 1: Body-waves 1
2
F. Brenguier1, R. Courbis1,4, A. Mordret2, X. Campman3, P. Boué1, M. Chmiel4,1, T. 3
Takano5,1, T. Lecocq6, W. Van der Veen7, S. Postif3, and D. Hollis4 4
1Institut des Sciences de la Terre, Univ. Grenoble Alpes. 5
2Massachusetts Institute of Technology. 6
3Shell. 7
4Sisprobe. 8
5Solid Earth Physics Laboratory, Tohoku Univ. 9
6Royal Observatory of Belgium 10
7Nederlandse Aardolie Maatschappij. 11
12
Corresponding author: Florent Brenguier ([email protected]) 13
Université Grenoble Alpes 14
ISTerre 15
CS 40700 16
38058 GRENOBLE Cedex 9 17
FRANCE 18
Phone: +33 673 35 68 77 19
Page 22 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
2
Summary 20
Unveiling the mechanisms of earthquake and volcanic eruption preparation requires improving 21
our ability to monitor the rock mass response to transient stress perturbations at depth. The 22
standard passive monitoring seismic interferometry technique based on coda-waves is robust 23
but recovering accurate and properly localized P and S-velocity temporal anomalies at 24
depth is intrinsically limited by the complexity of scattered, diffracted waves. In order to 25
mitigate this limitation, we propose a complementary, novel, passive seismic monitoring 26
approach based on detecting weak temporal changes of velocities of ballistic waves recovered 27
from seismic noise correlations. This new technique requires dense arrays of seismic sensors in 28
order to circumvent the bias linked to the intrinsic high sensitivity of ballistic waves recovered 29
from noise correlations to changes in the noise source properties. In this work we use a dense 30
network of 417 seismometers in the Groningen area of the Netherlands, one of Europe’s largest 31
gas fields. Over the course of 1 month our results show a 1.5 % apparent velocity increase of the 32
P-wave refracted at the basement of the 700 m thick sedimentary cover. We interpret this 33
unexpected high value of velocity increase for the refracted wave as being induced by the swings 34
in groundwater charge and discharge in a carbonate layer with water conductive fracture 35
networks at 700 m depth. We also observe a 0.25 % velocity decrease for the direct P-wave 36
travelling in the near-surface sediments but conclude that it might be partially biased by changes 37
in time in the noise source properties. The perspective of applying this new technique to detect 38
localized continuous variations of seismic velocity perturbations at a few kilometers depth paves 39
the way for improved in situ earthquake, volcano and producing reservoir monitoring. 40
41
Keywords: monitoring, seismic interferometry, earthquakes, volcanoes, producing reservoirs 42
Page 23 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
3
1 Introduction 43
Large earthquakes and volcanic eruptions result from long-lasting, steady, pressure 44
buildup on faults and magmatic reservoirs. However, the triggering mechanisms of impending 45
events are thought to be initiated by short time scale stress and pore pressure transients 46
associated with tectonic and volcanic interactions (e.g. Bouchon et al. 2011, Brenguier et al. 47
2014, Khoshmanesh & Shirzaei 2018) and possibly environmental perturbations (e.g. Johnson et 48
al. 2017). Anthropogenic activities such as hydrocarbon extraction, waste-water disposal, CO2 49
storage and geothermal production also induce fluid pore-pressure related deformation that can 50
lead to the triggering of induced seismicity (Talwani 2007, Ellsworth 2013, Chang and Segall 51
2016). Monitoring these stress and pore-pressure perturbations continuously in time with high 52
spatial accuracy at depth is thus critical to foresee forthcoming catastrophic tectonic and volcanic 53
events and to improve reservoir management. 54
55
Seismic velocities are sensitive to stress and pore-pressure perturbations. This has been 56
described in-situ for a wide range of processes like for example solid earth tides (Takano et al. 57
2014) and rainfall induced pore-pressure changes (Wang et al. 2017). Niu et al. (2008) used 58
active seismic source monitoring in boreholes at 1 km depth along the San Andreas Fault and 59
were able to observe the first clear preseismic seismic velocity-stress perturbations in the hours 60
preceding the occurrence of nearby earthquakes. Although convincing, these observations have 61
never been reproduced due to the costly and logistically complicated in-situ experiment. 62
63
Over the last decade, passive noise-based seismic monitoring proved success for 64
monitoring volcanoes (Sens‐Schönfelder et al. 2006, Brenguier et al. 2008a, Donaldson 2017), 65
Page 24 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
4
earthquakes (Brenguier et al. 2008b) and environmental/climate (Lecocq et al. 2017) changes. 66
Even though some attempts were made, no one succeeded in using passive seismic monitoring to 67
observe clear preseismic anomalies similar to those described by Niu et al. (2008). The standard 68
coda-based technique suffers from shortcomings that limit our ability to detect localized seismic 69
velocity perturbations at depth. This technique also referred to as coda-wave interferometry 70
(Poupinet et al. 1984) has the advantage of being very stable thanks to its low sensitivity to noise 71
source property changes (Colombi et al., 2014). It thus allows detecting weak temporal changes 72
of seismic velocities as small as 0.01 % such as those associated with solid Earth tides for 73
example (Mao et al. 2019). However, the counterpart of this high detection capability is that 74
the complexity of coda-wave propagation limits our ability to precisely characterize both 75
the type (P or S) of velocity change and accurate estimates of their spatial distribution at 76
depth. 77
In this work, we propose a complementary monitoring approach that uses ballistic waves 78
reconstructed from noise correlations. This paper focuses on ballistic body-waves and the 79
companion paper Mordret et al. 2019 focuses on using ballistic surface-waves on the same 80
dataset with the same type of approach. Body-waves have a specific sensitivity to seismic 81
velocity changes at depth and are potentially less affected by near-surface environmental changes 82
than surface waves. By using ballistic waves instead of coda-waves, we can more easily model 83
the spatial sensitivity of the temporal change observations to local velocity perturbations at 84
depth. The drawback of using direct, ballistic body-waves instead of coda-waves is their strong 85
sensitivity to noise source temporal variations (Colombi et al. 2014). We use dense seismic 86
networks and azimuthal averaging to circumvent this issue but still need to carefully analyze the 87
stability of noise sources for such type of analysis. 88
Page 25 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
5
Different studies showed that body-wave extraction from noise correlations is possible at 89
various scales (Roux et al. 2005, Draganov at al. 2009, Poli et al. 2012). Nakata et al. (2015) 90
were able to implement the first passive 3-D P-wave velocity tomography from continuous 91
ground motion recorded on a dense array of more than 2500 seismic sensors installed at Long 92
Beach (California, USA). Recently Brenguier et al. (2016) and Nakata et al. (2016) proved the 93
temporal stability of direct virtual body-waves between dense arrays on Piton de la Fournaise 94
volcano thus opening the way for continuous, passive ballistic wave monitoring. 95
96
In this paper we describe the fundamental aspects of passive ballistic wave monitoring 97
using dense arrays and further illustrate its potential by applying it to a network of 417 seismic 98
stations in the Netherlands. We are able to measure temporal changes of apparent velocities from 99
both direct and refracted P-waves, and thus we are able to separate the response of the near 100
surface sediments and the basement located at 700 m depth. By providing direct observations of 101
the rock mass response to stress changes at depth, this new passive seismic approach paves the 102
way for improved in situ earthquake, volcano and producing reservoir monitoring. 103
104
2 Methods 105
106
The new approach is based on measuring temporal changes of apparent slowness of specific 107
ballistic waves that have been reconstructed from noise correlations using dense arrays of 108
seismic sensors (Boué et al., 2013b, Mordret et al. 2014, Nakata et al., 2015, Nakata et al., 2016). 109
The underlying requirement is that, as for Nakata et al. (2015), the high number of seismic 110
sensors (> 100) and thus of noise correlation receiver pairs allows for the reconstruction of a 111
virtual shot-gather of sufficiently high quality to be able to isolate and measure the apparent 112
Page 26 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
6
velocity of ballistic waves such as direct P or S-waves, refracted waves or surface waves with 113
clear mode separation (Mordret et al. 2019). 114
115
For this purpose of properly extracting high quality ballistic waves, we gather all possible 116
noise correlations for all receiver pairs from a dense network into a single seismic panel 117
(propagation time versus virtual source-receiver offset) thus considering a 1D velocity model for 118
which the apparent slowness of a reconstructed ballistic wave measured at surface can be written 119
as: 120
121
𝑛 =1
𝑉=
∆𝑥𝑡
∆𝑥 122
123
where n is the apparent slowness and V the apparent velocity. We can estimate n as the slope of 124
apparent arrival times 𝑡 along distance 𝑥 under the assumption that the apparent velocity 𝑉 is 125
uniform along distance range ∆𝑥. We are now interested in measuring a temporal change in 126
apparent slowness ∆𝐶𝑡𝑛, the index 𝐶𝑡 being for Calendar time (Fig. 1a): 127
128
∆𝐶𝑡𝑛 = 𝑛𝐶𝑡2 − 𝑛𝐶𝑡1 = (∆𝑥𝑡
∆𝑥)
𝐶𝑡2− (
∆𝑥𝑡
∆𝑥)
𝐶𝑡1 129
130
The temporal change of apparent slowness can be measured directly as the difference of 131
slowness estimates (from a slant-stack or beam-forming analysis for example) at different 132
calendar times (de Cacqueray et al. 2016). Here we use an approach that estimates the temporal 133
Equation 1
Equation 2
Page 27 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
7
change of apparent slowness as the slope of the linear regression of the travel time shifts at 134
different calendar times for each distance offset along distance ∆𝑥 (Fig. 1b): 135
136
∆𝐶𝑡𝑛 =∆𝑥(∆𝐶𝑡𝑡)
∆𝑥 137
138
where ∆𝐶𝑡𝑡 are the measurements of travel time shifts at different offsets ∆𝑥 for two different 139
calendar times and for a specific windowed ballistic wave. This approach shows the advantage of 140
providing a direct estimate of the uncertainty of the estimated apparent slowness temporal 141
change by assessing how the measured travel time delays ∆𝐶𝑡𝑡 fit a linear regression model along 142
distance range ∆𝑥. 143
144
From equation 2, we can derive the relation linking the temporal change of apparent slowness 145
and apparent velocity: 146
∆𝐶𝑡𝑛 = −∆𝐶𝑡𝑉
𝑉2 147
148
By multiplying the above equation by V, the apparent uniform velocity of the studied ballistic 149
wave, this equation leads to (Fig 1c): 150
151
∆𝐶𝑡𝑛 × 𝑉 =∆𝑥(∆𝐶𝑡𝑡)
∆𝑥𝑡= −
∆𝐶𝑡𝑉
𝑉 152
This later equation shows that, for small velocity perturbations (<10%) we can estimate the 153
relative temporal change of apparent velocity, ∆𝐶𝑡𝑉
𝑉, of a given ballistic wave by estimating the 154
value of the slope of the linear regression of ∆𝐶𝑡𝑡 measurements along travel time 𝑡. This 155
Equation 3
Equation 4
Equation 5
Page 28 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
8
approach is sketched on Fig. 1. It applies to any kind of ballistic arrivals that can be clearly 156
identified and isolated on a virtual source gather section. The case of direct surface waves 157
requires mode separation and is challenging because of dispersion that leads to different 158
velocities for different periods at which ∆𝐶𝑡𝑡 values are measured (Mordret et al. 2019). As 159
suggested in Nakata et al. (2016), this approach can also be applied to noise-correlations between 160
two distant arrays in order to monitor diving body-waves probing magmatic reservoirs, seismic 161
faults or producing reservoirs at a few hundreds to a few kilometers depth. It is interesting to 162
note that in this situation of two distant arrays referred to as A and B, the estimates of temporal 163
changes of apparent velocities can be achieved using noise-correlations between arrays A and B 164
and B and A separately, thus providing two independent estimates of temporal velocity changes. 165
This method can also be applied using noise correlations between an individual seismic station 166
and a distant array. In this work we apply this approach to the monitoring of a direct and a 167
refracted wave using noise-correlations within a single array of about 8 km wide. 168
169
170
Page 29 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
9
171
Figure 1. Procedure for ballistic-wave apparent seismic velocity monitoring. a) propagation of a 172
direct ballistic wave. The dashed lines show the reference wave and the plain lines show the 173
wave affected by a velocity perturbation of -7 %. b) travel time shifts measurements and linear 174
regression along distance. c) conversion from distance to travel time by dividing distance by V, 175
the apparent velocity of the propagating wave. 176
3 Data 177
The Groningen gas field located in the northeast of the Netherlands is one of Europe’s 178
largest natural gas field. The reservoir located at 3 km depth is thought to be 40 by 50 km wide 179
and 250 m thick. Bourne et al. (2018) show that the gas production in this field led to a 15 MPa 180
average reservoir pore-pressure depletion since 1995 which is associated with seismicity rates 181
that increased as an exponential-like function. 182
183
Page 30 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
10
We use a network of 417 short period seismic stations deployed in the Groningen area of 184
the Netherlands (Fig. 2) from 11 February (day 42) to 12 March (day 71) 2017 over a time span 185
of 30 days. The network forms a grid array with aperture of the order of 8 km and an average 186
station distance of about 400 m. 187
188
189
190
Figure 2. a) Geometry of the 417 short-period stations used in this study. b) beamforming of the 191
30 days of continuous seismic data in the 1-4 seconds period range. 192
193
We computed an averaged seismic section of vertical to vertical noise cross-correlations. The 194
noise-correlations were stacked in time (over the 30 days of continuous data), in space (along 195
distance bins of 50 meters long) and for the causal and acausal parts following Boué et al. 196
(2013a) and Nakata et al. (2015). Figure 3 illustrates these binned data for two different 197
Page 31 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
11
frequency ranges (1-3 Hz and 3-12 Hz). The low frequency (1-3 Hz) section mainly shows the 198
propagation of low-velocity Rayleigh waves and also of ballistic P-waves at velocities between 199
1.5 and 3 km/s. The lower panel of Fig. 3 highlights the high frequency (3-12 Hz) P-waves 200
interpreted as (1) the direct, diving P-wave with velocity of ~1700 m/s (see model on the right 201
panel) contoured by a blue box and (2) a refracted wave at the 700 m interface with apparent 202
velocity of ~3300 m/s contoured by a red box. Thanks to the high stability in time of the useful 203
high-frequency ambient seismic noise, we are finally able to reconstruct repetitive in time 204
seismic sections from the correlations of daily records of ambient seismic noise leading us to 205
daily virtual shot seismic gathers. 206
207
208
Figure 3. Noise cross-correlation averaged binned section for frequency ranges 1- 3 Hz (a) and 209
3-12 Hz (b). The blue and red dashed boxes correspond to the selected windows used for the 210
analysis for the direct and refracted waves. The right panel (c) illustrates the average P-wave 211
velocity model of the area illustrating the velocities of the overburden (saturated sediments) 212
Page 32 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
12
around 1700 m/s and of the bedrock at ~700 m depth around 3000 m/s. It has been defined 213
using sonic logs from deep wells in the area (Kruiver et al. 2017). 214
215
4 Analysis and Results 216
In order to measure temporal changes, we further isolate the direct and refracted waves 217
by applying a tapered window (Fig. 4a,b). For the time shift measurements, ∆𝐶𝑡𝑡, we use a 218
cross-spectrum approach (Clarke et al. 2011) in the frequency range 3 to 8 Hz. We illustrate our 219
approach on Figure 4 by plotting travel time shifts for the direct and refracted waves between 220
references (stacks of days 42 to 50) and a current seismic sections (days 53 to 62 for the direct 221
and days 48 to 57 for the refracted waves). Even though the travel time shifts measurements 222
show large fluctuations likely associated with imperfect direct and refracted waves 223
reconstruction and noise source changes through time, they show a clear linear trend along 224
distance, especially for the refracted wave, indicating a clear change in apparent velocity 225
between these two time periods. By multiplying the slope of the ∆𝐶𝑡𝑡 over distance linear 226
regression by the apparent velocity of the direct (1700 m/s) and refracted (3300 m/s) waves (step 227
b) to c) on Fig. 1), we find velocity changes of -0.25 % and +1.5% for the direct and refracted 228
waves. 229
230
Page 33 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
13
231
Figure 4. Travel time shifts along distance plots. Left a), windowed reference direct waves 232
averaged for the time period (days 42 to 50). b) windowed current direct waves averaged for 233
days 53 to 62. c) travel time shift measurements, ∆𝐶𝑡𝑡, along distance between these two direct 234
waves. Right a), windowed reference refracted waves averaged for the time period (days 42 to 235
50). b) windowed current refracted waves averaged for days 48 to 57. c) travel time shift 236
measurements, ∆𝐶𝑡𝑡, along distance between these two refracted waves. 237
238
239
240
In order to gain insights on the temporal evolution of velocity changes, we further average the 241
daily seismic sections using a 10-days long, 1-day moving window. This leads us to 21, full 10-242
days averaged, daily seismic sections (from days 42 to 71). Following the method described 243
above, we measure velocity changes (∆𝐶𝑡𝑉
𝑉) between each 10-days averaged seismic section and 244
the reference section corresponding to a stack of days 42 to 50 for both the direct (Fig. 5a) and 245
refracted (Fig. 5b) waves. The apparent velocity change curves shown on Figure 5 indicate a 246
velocity decrease of maximum -0.25 % for the direct wave and a velocity increase of maximum 247
1.5 % for the refracted wave. The error bars correspond to the uncertainty of the linear 248
Page 34 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
14
regressions estimates of the travel time shifts, ∆𝐶𝑡𝑡, over travel time t (Fig. 1c) using a least-249
square approach following Brenguier et al. 2008a. 250
251
252
253 254
255
5 Discussions and conclusions 256
257
The most common source of error in passive, noise-based seismic monitoring results 258
from the non-stationarity of noise sources. Furthermore, ballistic waves are much more sensitive 259
to noise source variations than coda waves (Colombi et al. 2014). As a result, the main drawback 260
Figure 5. Apparent seismic
velocity temporal changes for both
the direct (top) and refracted
(bottom) waves.
Page 35 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
15
of this ballistic wave based monitoring method is the potential error introduced by changes in 261
noise sources and great care has to be taken when interpreting the results. We thus need to 262
properly assess how noise source temporal azimuthal variations can hamper our results. To do so 263
we beamform the ambient seismic noise on a 10-days average basis in the period range 1 to 4 264
seconds (Fig. 2). Due to the minimum inter-station distance of 300 m we are not able to 265
beamform the noise in the frequency range of interest (3-8 Hz) and we thus make the assumption 266
that the 1-4 s beams are representative for the higher frequency noise characteristics. We found 267
that the noise source distribution is strongly anisotropic with a main spot coming from the North 268
Sea, north of our array (Fig. 2). Interestingly, the high frequency (3-8 Hz) noise correlations 269
are quite asymmetric proving that most of the high-frequency body-waves also come from 270
the North of our array pointing to shore break as a possible source of high-frequency noise. 271
Our assumption of comparing the high-frequency (3-8 Hz, likely shore break) to the low-272
frequency (1-4 s, likely ocean microseismic) noise sources might not be fully relevant but 273
still informative. We observe that during the time span of our analysis, the noise source 274
distribution in the 1-4 s period range is mostly stable with small azimuthal variations of the 275
centroid of the main spot by less than 10°. Following the theoretical predictions of travel time 276
errors of ballistic arrival reconstructed from correlations of non-isotropically distributed noise 277
sources from Weaver et al. (2009) and the further applications of Froment et al. (2010) and 278
Colombi et al. (2014), we assess that, in case of a two sensors noise correlation, the error on the 279
travel time shift measurements (Fig. 5) would lead to a velocity change uncertainty of less than 280
0.5 %. In our case we emphasize that our travel time shift measurements are obtained from 281
azimuthally averaged noise correlations from the dense 417 stations array. We are thus confident 282
that the observed velocity changes of +1.5 % for the refracted wave is mostly related to physical 283
Page 36 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
16
velocity changes. However, the velocity decrease of 0.25 % for the direct wave might thus be 284
partly biased by changes in the noise source properties. 285
In order to interpret the apparent velocity change of -0.25 % of the direct P-wave, we can 286
consider a simple 1D model of wave propagation in the sedimentary overburden and the ray 287
theory that predicts that the observed apparent velocity change on the direct P-wave corresponds 288
to a real P-wave velocity perturbation averaged over the first 200 m depth (maximum distance 289
between virtual source and stations of 2200 m). The only noticeable event during that time 290
period is an episode of rainfall that occurred on day 53 in the region (20 mm of cumulated water 291
height). The decrease of velocity for the direct wave could thus be related to a poro-elastic 292
process as described by Rivet et al. (2015) and Wang et al. (2017). We will address the study of 293
these shallow surface temporal velocity variations measurements in more details including the 294
analysis of surface waves in a companion paper. 295
Following again the ray theory, the increase of apparent velocity of 1.5 % for the 296
refracted wave can be directly attributed to a change of P-wave velocity of the carbonate bedrock 297
at 700 m depth. Considering this hypothesis of a bedrock velocity increase, one simple 298
interpretation could be the effect of loading from rainfall on the bedrock that would increase the 299
confining pressure and close cracks in the bedrock. However, it is unlikely that 2 cm of 300
additional water height on day 53, corresponding to an increase of loading of 0.2 kPa, leads to a 301
velocity increase of 1.6 % at 700 m depth. Indeed, following Yamamura et al. (2003), and their 302
observation of velocity-stress sensitivity of 10-7 Pa-1 the expected velocity change for a loading 303
of 0.2 kPa should be about 0.01 %. Moreover, we checked for INSAR and GPS observations. 304
These data don’t show any significant transient anomalies during the time span of our analysis. 305
Finally, we also preclude the potential effects of swings in gas production in the main reservoir 306
Page 37 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
17
at 3 km depth below our study area due to the large distance to the probed carbonate layer. These 307
induced pressure variations are of the order of less than 0.1 MPa locally and are thus too small to 308
potentially induce a 1.5 % velocity increase 2.3 km above in the carbonate layer. We also rule 309
out the effects of local induced earthquakes due to the low level of seismicity during the studied 310
time period. The most likely interpretation for this velocity increase relates to movements of 311
fluids in water conductive fracture networks within the carbonate layer at about 700 m 312
depth. This is discussed in a companion paper Mordret et al. 2019 that includes additional 313
passive monitoring observations using direct surface waves. 314
In conclusions, even though our results are hampered by high uncertainties and are 315
spanning a too short period to be interpreted properly, this new passive ballistic wave seismic 316
monitoring approach has the potential for revealing seismic wave velocity temporal variations at 317
localized areas at depth thus acting as a stress-strain probe. Even though the main drawback of 318
this technique is that it requires dense seismic networks, we believe that this technique together 319
with the recent step change in seismic instrumentation will lead to groundbreaking advances in 320
our understanding of natural and induced earthquakes, volcanic eruptions and will prove useful 321
for reservoir management. 322
323
Acknowledgments 324
We acknowledge two anonymous reviewers for their useful comments. This project received 325
funding from the Shell Game Changer project HiProbe. We also acknowledge support from the 326
French ANR grant T-ERC 2018, (FaultProbe), the European Research Council under grants no. 327
817803, FAULTSCAN and no. 742335, F-IMAGE and the European Union’s Horizon 2020 328
research and innovation program under grant agreement No 776622 (PACIFIC). AM 329
Page 38 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
18
acknowledges support from the National Science Foundation grants PLR-1643761. The data 330
were provided by NAM (Nederlandse Aardolie Maatschappij). We acknowledge M. Campillo, 331
N. Shapiro, G. Olivier, P. Roux, S. Garambois, R. Brossier and C. Voisin for useful discussions. 332
The authors thank Nederlandse Aardolie Maatschappij and Shell for permission to publish. 333
334
References 335
Bouchon, M., Karabulut, H., Aktar, M., Özalaybey, S., Schmittbuhl, J., & Bouin, M. P. (2011). 336
Extended nucleation of the 1999 Mw 7.6 Izmit earthquake. science, 331(6019), 877-880. 337
338
Boué, P., Roux, P., Campillo, M., & de Cacqueray, B. (2013a). Double beamforming processing 339
in a seismic prospecting context. Geophysics, 78(3), V101-V108. 340
341
Boué, P., Poli, P., Campillo, M., Pedersen, H., Briand, X., & Roux, P. (2013b). Teleseismic 342
correlations of ambient seismic noise for deep global imaging of the Earth. Geophysical Journal 343
International, 194(2), 844-848. 344
345
Bourne, S. J., Oates, S. J., & van Elk, J. (2018). The exponential rise of induced seismicity with 346
increasing stress levels in the Groningen gas field and its implications for controlling seismic 347
risk. Geophysical Journal International, 213(3), 1693-1700. 348
349
Brenguier, F., Shapiro, N. M., Campillo, M., Ferrazzini, V., Duputel, Z., Coutant, O., & 350
Nercessian, A. (2008a). Towards forecasting volcanic eruptions using seismic noise. Nature 351
Geoscience, 1(2), 126. 352
353
Brenguier, F., Campillo, M., Hadziioannou, C., Shapiro, N. M., Nadeau, R. M., & Larose, E. 354
(2008b). Postseismic relaxation along the San Andreas fault at Parkfield from continuous 355
seismological observations. Science, 321(5895), 1478-1481. 356
357
Brenguier, F., Campillo, M., Takeda, T., Aoki, Y., Shapiro, N. M., Briand, X., ... & Miyake, H. 358
(2014). Mapping pressurized volcanic fluids from induced crustal seismic velocity drops. 359
Science, 345(6192), 80-82. 360
361
Brenguier, F., Kowalski, P., Ackerley, N., Nakata, N., Boué, P., Campillo, M., ... & Roux, P. 362
(2016). Toward 4D Noise-Based Seismic Probing of Volcanoes: Perspectives from a Large-N 363
Experiment on Piton de la Fournaise Volcano. Seismological Research Letters, 87(1), 15-25. 364
365
de Cacqueray, B., Roux, P., & Campillo, M. (2016). Using slowness and azimuth fluctuations as 366
new observables for four‐dimensional reservoir seismic monitoring. Geophysical Prospecting, 367
64(6), 1537-1555. 368
Page 39 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
19
369
Chang, K. W., & Segall, P. (2016). Injection‐induced seismicity on basement faults including 370
poroelastic stressing. Journal of Geophysical Research: Solid Earth, 121(4), 2708-2726. 371
372
Clarke, D., Zaccarelli, L., Shapiro, N. M., & Brenguier, F. (2011). Assessment of resolution and 373
accuracy of the Moving Window Cross Spectral technique for monitoring crustal temporal 374
variations using ambient seismic noise. Geophysical Journal International, 186(2), 867-882. 375
376
Colombi, A., Chaput, J., Brenguier, F., Hillers, G., Roux, P., & Campillo, M. (2014). On the 377
temporal stability of the coda of ambient noise correlations. Comptes Rendus Geoscience, 378
346(11), 307-316. 379
380
Donaldson, C., Caudron, C., Green, R. G., Thelen, W. A., & White, R. S. (2017). Relative 381
seismic velocity variations correlate with deformation at Kīlauea volcano. Science advances, 382
3(6), e1700219. 383
384
Draganov, D., X. Campman, J. Thorbecke, A. Verdel, and K. Wapenaar, 2009. Reflection 385
images from ambient seismic noise, Geophysics, 74 (5), A63–A67. 386
387
Ellsworth, W. L. (2013). Injection-induced earthquakes. Science, 341(6142), 1225942. 388
Johnson, C. W., Fu, Y., & Bürgmann, R. (2017). Seasonal water storage, stress modulation, and 389
California seismicity. Science, 356(6343), 1161-1164. 390
391
Froment, B., Campillo, M., Roux, P., Gouedard, P., Verdel, A., & Weaver, R. L. (2010). 392
Estimation of the effect of nonisotropically distributed energy on the apparent arrival time in 393
correlations. Geophysics, 75(5), SA85-SA93. 394
395
Khoshmanesh, M., & Shirzaei, M. (2018). Episodic creep events on the San Andreas Fault 396
caused by pore pressure variations. Nature Geoscience, 1. 397
398
Kruiver, P. P., van Dedem, E., Romijn, R., de Lange, G., Korff, M., Stafleu, J., ... & Doornhof, 399
D. (2017). An integrated shear-wave velocity model for the Groningen gas field, The 400
Netherlands. Bulletin of Earthquake Engineering, 15(9), 3555-3580. 401
402
Lecocq, T., Longuevergne, L., Pedersen, H. A., Brenguier, F., & Stammler, K. (2017). 403
Monitoring ground water storage at mesoscale using seismic noise: 30 years of continuous 404
observation and thermo-elastic and hydrological modeling. Scientific Reports, 7(1), 14241. 405
406
Mao, S., Campillo, M., van der Hilst, R. D., Brenguier, F., Stehly, L., & Hillers, G. (2019). High 407
temporal resolution monitoring of small variations in crustal strain by dense seismic arrays. 408
Geophysical Research Letters, 46(1), 128-137. 409
410
Mordret, A., Shapiro, N. M., & Singh, S. (2014). Seismic noise‐based time‐lapse monitoring of 411
the Valhall overburden. Geophysical Research Letters, 41(14), 4945-4952. 412
413
Page 40 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
20
Mordret, A., Courbis, R., Brenguier, F., Chmiel, M., Garambois, S., Mao, S., ... & Hollis, D. 414
(2019). Noise-based Ballistic Wave Passive Seismic Monitoring – Part 2: Surface Waves. 415
Geophysical Journal International, submitted. 416
417
Nakata, N., J. P. Chang, J. F. Lawrence, and P. Boué, (2015). Body-wave extraction and 418
tomography at Long Beach, California, with ambient-noise tomography, J. Geophys. Res., 120, 419
1159–1173, doi:10.1002/2015JB01870. 420
421
Nakata, N., Boué, P., Brenguier, F., Roux, P., Ferrazzini, V., & Campillo, M. (2016). Body‐and 422
surface‐wave reconstruction from seismic‐noise correlations between arrays at Piton de la 423
Fournaise volcano. Geophysical Research Letters, 43, 1047-1054. 424
425
Niu, F., Silver, P. G., Daley, T. M., Cheng, X., & Majer, E. L. (2008). Preseismic velocity 426
changes observed from active source monitoring at the Parkfield SAFOD drill site. Nature, 427
454(7201), 204. 428
429
Poli, P., M. Campillo, H. Pedersen, and L. W. Group, 2012. Body-wave imaging of Earth’s 430
mantle discontinuities from ambient seismic noise, Science, 338, 1063–1065. 431
432
Poupinet, G., Ellsworth, W. L., & Frechet, J. (1984). Monitoring velocity variations in the crust 433
using earthquake doublets: An application to the Calaveras Fault, California. Journal of 434
Geophysical Research: Solid Earth, 89(B7), 5719-5731. 435
436
Rivet, D., Brenguier, F., & Cappa, F. (2015). Improved detection of preeruptive seismic velocity 437
drops at the Piton de La Fournaise volcano. Geophysical Research Letters, 42(15), 6332-6339. 438
439
Romanet, P., Bhat, H. S., Jolivet, R., & Madariaga, R. (2018). Fast and slow slip events emerge 440
due to fault geometrical complexity. Geophysical Research Letters, 45, 4809–4819. 441
442
Rost, S., & Thomas, C. (2002). Array seismology: Methods and applications. Reviews of 443
geophysics, 40(3). 444
445
Roux, P., Sabra, K. G., Gerstoft, P., Kuperman, W. A., & Fehler, M. C. (2005). P‐waves from 446
cross‐correlation of seismic noise. Geophysical Research Letters, 32(19). 447
448
Sens‐Schönfelder, C., & Wegler, U. (2006). Passive image interferometry and seasonal 449
variations of seismic velocities at Merapi Volcano, Indonesia. Geophysical research letters, 450
33(21). 451
452
Shapiro, N. M., Campillo, M., Stehly, L., & Ritzwoller, M. H. (2005). High-resolution surface-453
wave tomography from ambient seismic noise. Science, 307(5715), 1615-1618. 454
455
Silverii, F., D'Agostino, N., Métois, M., Fiorillo, F., & Ventafridda, G. (2016). Transient 456
deformation of karst aquifers due to seasonal and multiyear groundwater variations observed by 457
GPS in southern Apennines (Italy). Journal of Geophysical Research: Solid Earth, 121(11), 458
8315-8337. 459
Page 41 of 42 Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
Noise-based ballistic-wave seismic monitoring
21
460
Snieder, R., Grêt, A., Douma, H., & Scales, J. (2002). Coda wave interferometry for estimating 461
nonlinear behavior in seismic velocity. Science, 295(5563), 2253-2255. 462
463
Takano, T., Nishimura, T., Nakahara, H., Ohta, Y., & Tanaka, S. (2014). Seismic velocity 464
changes caused by the Earth tide: Ambient noise correlation analyses of small‐array data. 465
Geophysical Research Letters, 41(17), 6131-6136. 466
467
Talwani, P., Chen, L., & Gahalaut, K. (2007). Seismogenic permeability, ks. Journal of 468
Geophysical Research: Solid Earth, 112(B7). 469
470
Wang, Q. Y., Brenguier, F., Campillo, M., Lecointre, A., Takeda, T., & Aoki, Y. (2017). 471
Seasonal Crustal Seismic Velocity Changes Throughout Japan. Journal of Geophysical 472
Research: Solid Earth, 122(10), 7987-8002. 473
474
Weaver, R., Froment, B., & Campillo, M. (2009). On the correlation of non-isotropically 475
distributed ballistic scalar diffuse waves. The Journal of the Acoustical Society of America, 476
126(4), 1817-1826. 477
478
Yamamura, K., Sano, O., Utada, H., Takei, Y., Nakao, S., & Fukao, Y. (2003). Long‐term 479
observation of in situ seismic velocity and attenuation. Journal of Geophysical Research: Solid 480
Earth, 108(B6). 481
Page 42 of 42Geophysical Journal International
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960