IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 38, NO.

1, JANUARY 2000 3

Dispersion Analysis of Crack-Waves in an Artificial

Subsurface Fracture Using Two Crack Models
Koji Nagano and Hiroaki Niitsuma

Abstract—We investigated crack-wave dispersions in an arti- We previously studied the propagation of crack-waves
ficial subsurface fracture both experimentally and numerically to characterize subsurface fractures [4]–[7]. We detected
using a wavelet analysis and two crack models. Crack-waves are crack-waves propagating along an artificial subsurface frac-
seismic modes that propagate along a fracture. The dispersion
characteristics of crack-waves depend on the geometry and ture, which was saturated with water at a depth of 370 m in
physical properties of a fracture. We measured crack-waves at an Higashi–Hachimantai Hot Dry Rock model field, Japan [4].
artificial subsurface fracture in Higashi–Hachimantai Hot Dry We estimated the dimensionless crack stiffness of the fracture
Rock model field, Japan. This subsurface fracture is at a depth of surfaces based on dispersions of the crack-waves [5], [8]. In
about 370 m. During a measurement, we injected water into the this analysis, we applied a pseudo-Wigner–Ville distribution
fracture and changed the interface conditions of the fracture. A
wavelet analysis provided the dispersion of the arrival times of (PWD) to estimate the dispersions of crack-waves. However,
crack-waves. The crack-waves showed positive velocity dispersion; it was difficult to estimate dispersions because of cross terms
i.e., low frequency components arrived later. As wellhead pressure in the PWD. Cross terms naturally occur in energy densities of
increased due to water injection, the dispersion characteristics the PWD. The dimensionless crack stiffness is calculated from
changed. A low-velocity-layer (LVL) model and a crack-stiffness three physical properties of a fracture (thickness of the fluid
model were examined to explain crack-wave dispersion. In the
LVL model, rock layers with a low velocity surround a fluid layer. layer, crack stiffness, and the shear modulus of the solid). We
There is no contact between the LVL’s. On the other hand, the could not estimate the primary physical properties of a fracture
crack-stiffness model considers crack stiffness due to contact in the crack-wave analysis, because the dimensionless crack
between asperities on fracture surfaces. The arrival-time curves stiffness is a theoretical parameter.
calculated by the crack-stiffness model showed a good fit to the Niitsuma and Saito [9] and Tanaka et al. [10] reported that
measured values. As wellhead pressure increased, crack stiffness
decreased and thickness of a fluid layer increased. In contrast, the a low-velocity zone could be found close to an artificial sub-
LVL model did not adequately duplicate the measured data. surface fracture that had been created in an intact rock layer by
Index Terms—Crack, crack-waves, dispersion, wavelet. hydraulic fracturing. The velocity of a compressional wave de-
creased in a path through the fracture when wellhead pressure
increased due to water injection into the fracture. There was no
I. INTRODUCTION natural crack in the rock layer before hydraulic fracturing. They
explained this decrease in compressional wave velocity in terms
A FLUID-FILLED crack and a fracture zone with low
velocities act as elastic interfaces for seismic waves. Such
interfaces trap seismic waves, and trapped seismic waves prop-
of the reopening of microcracks in the vicinity of the fracture.
This phenomenon indicates that a low-velocity zone is essential
agate along the interfaces. They are referred to as crack-waves, for subsurface fractures. Therefore, it is important to estimate
crack-interface waves, and fault zone-guided waves [1]–[3]. the low-velocity zone when subsurface fractures are analyzed.
In this paper, we refer to these seismic modes as crack-waves. The numerical analysis of crack-waves trapped in a single
The propagation characteristics of crack-waves are different fluid-filled crack has been reported by Chouet [1], Ferrazzini
from those of reflected or refracted seismic waves, which are and Aki [11], and Hayashi and Sato [8]. All of these authors ad-
used in conventional seismic surveys. Their waveforms (e.g., dressed very slow waves with positive dispersion. Chouet, and
velocity-frequency dispersion and amplitude-space distribu- Ferrazzini and Aki used a simple three-layer model for a sub-
tion) are strongly dependent on the crack’s geometry and its surface crack. A very thin fluid layer, compared with the wave-
physical properties. Therefore, measurement of crack-waves is length, was sandwiched between two solid half-spaces. Hayashi
an effective tool for characterizing a subsurface fracture. and Sato focused on the contact between asperities on fracture
surfaces [8]. The dimensionless crack stiffness was used to rep-
resent contact in their crack model.
Manuscript received July 1, 1998; revised March 24, 1999. This work was
supported by a Grant-in Aid for Encouragement of Young Scientists, Ministry In this study, we investigate the opening of a subsurface
of Education, Science, and Culture of Japan under Contracts 08751081 and fracture by comparing the dispersions of crack-waves mea-
10750667 and the Espec Foundation for Earth Environmental Research and sured in the field and numerical results calculated using two
Technologies (Charitable Trust), and was conducted as part of the MTC (More
Than Cloud) international collaborative project funded by NEDO, Japan. crack models. The crack-waves were measured at an artificial
K. Nagano is with the Department of Computer Science and Systems Engi- subsurface fracture at a depth of about 370 m. Two waveforms
neering, Muroran Institute of Technology, Muroran 050-8585, Japan (e-mail: of the crack-waves, which were recorded at the beginning of
[email protected]).
H. Niitsuma is with the Department of Geoscience and Technology, Graduate water injection and at the maximum wellhead pressure during
School of Engineering, Tohoku University, Sendai 980-8579, Japan (e-mail: ni- water injection, are analyzed. The wavelet transform (WT)
[email protected]). provides a time-frequency representation of the arrival time of
Publisher Item Identifier S 0196-2892(00)00001-2.

0196–2892/00$10.00 © 2000 IEEE

4 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 38, NO. 1, JANUARY 2000

Fig. 1. Concept of crack-wave measurement. Fig. 3. Wellhead pressure and travel time of the compressional wave during
water injection.

into the artificial fracture after the hydraulic fracturing and

intersected the artificial fracture at a depth of 358.2 m. The
distance between the intersection points of wells F-1 and EE-4
was 6.7 m. The radius of the fracture was about 60 m [12]. A
transmissibility test showed that the fracture aperture was about
0.08 mm without pressurization and 0.2 mm at a wellhead
pressure of 3.0 MPa [13]. The transmissibility test also showed
that the reopening pressure of this fracture system was about
2.8 MPa at the wellhead. The velocities of compressional and
shear waves of intact rock in this field are 3100 m/s and 1860
m/s, respectively [10].
We carried out our measurements using this single artificial
fracture. A downhole air gun was used as a wave source at a
depth of 367.0 m in well F-1. The air gun was suspended two
meters from the intersection of the fracture so that we would
not damage the bore hole at the intersection. A hydrophone was
suspended at a depth of 358.2 m in well EE-4 (i.e., at the in-
Fig. 2. Higashi–Hachimantai HDR model field. Wells EE-4 and F-1 intersect
an artificial subsurface fracture at depths of 358.2 and 369.0 m, respectively. tersection with the fracture), to measure the crack-waves. This
hydrophone has a flat sensitivity in a frequency range of 60 Hz
to 10 kHz, and its sensitivity gradually decreases for frequency
crack-waves. The two crack models are the crack-stiffness
components below 60 Hz.
model, which was investigated by Hayashi and Sato [8], and
We pressurized the subsurface fracture to vary the interface
the low-velocity layer (LVL) model, which is a modified
conditions of the fracture. The wellheads of both wells were
three-layer model based on the experimental results of Niit-
closed with wireline lubricators. Wellhead pressure was mea-
suma and Saito [9]. We calculate the arrival time curves of
sured in well F-1. Since circulation loss and friction loss in the
crack-waves with the two crack models and fit these curves to
bore holes and this fracture system are small, the wellhead pres-
the spectral data in the time-frequency representation from the
sure in well EE-4 is almost equal to that in well F-1 [12]. Fig. 3
WT. We estimate the physical parameters of the crack models
shows the wellhead pressure during crack-wave measurement.
in the opening of a fracture based on crack-wave dispersions.
After we held the wellhead pressure at 3.0 MPa for 30 min,
we closed the wellheads completely. Wellhead pressure subse-
II. CRACK-WAVE MEASUREMENT IN HIGASHI–HACHIMANTAI
quently decreased naturally due to permeation into the rock.
HOT DRY ROCK (HDR) MODEL FIELD
Fig. 3 also shows the travel time of a compressional wave in
Crack-waves were measured at Higashi–Hachimantai Hot the crack-wave measurement. Since the air gun was two meters
Dry Rock (HDR) model field, Japan [12]. Fig. 1 illustrates above the intersection of the bore hole and the hydrophone was
the concept of a crack-wave measurement. Crack-waves can at another intersection, the path of the compressional wave was
be detected at the intersection of a bore hole and the fracture. close to the fracture. The velocity of the compressional wave
Fig. 2 depicts Higashi–Hachimantai HDR model field. An was calculated from the travel time and the distance, 6.7 m, be-
artificial subsurface fracture was created in intact, welded tuff tween the intersections of the bore holes and the fracture. The
at a depth of 369.0 m in well F-1 by hydraulic fracturing. velocity was 1900 m/s at the beginning of water injection and
During hydraulic fracturing, 40-mesh sand was injected as a 1700 m/s when the wellhead pressure was constant at 3.0 MPa
propping agent. Core samples of well F-1 showed no significant for about 30 min. These velocities are lower than that measured
joint or crack before the fracturing. Well EE-4 was drilled in other experiments in this field [10]. Another water-injection
NAGANO AND NIITSUMA: DISPERSION ANALYSIS OF CRACK-WAVES 5

in (3), we can consider that the admissibility condition is satis-

fied [17]. In the WT calculation, we set .
We analyzed the WT of the crack-waves in a frequency range
between 2 and 256 Hz because of the frequency components
of the air gun. The waveforms and squared amplitudes of the
WT are presented in Fig. 5. Fig. 5(a) shows the crack-waves de-
tected at the beginning of water injection. The wellhead pressure
is 0.4 MPa. The crack-waves in Fig. 5(b) were detected when
the wellhead pressure was held constant at 3.0 MPa for about
30 min. The squared amplitude of the WT was normalized by
its maximum value at each frequency to enhance the contrast
of the variation in time for frequency components with small
amplitude. The air gun exploded seismic waves at a frequency
Fig. 4. Crack-waves during water injection in an artificial subsurface fracture. range below 300 Hz simultaneously [5]. Crack-waves are dis-
The crack-waves arrive at 40 ms. persive, since arrival time (which occurs at the onset of the rel-
ative maxima) is a function of frequency in Fig. 5. The lower
test was carried out about one hour before the crack-wave mea- frequency components of the crack-waves arrive later. The ve-
surement. These lower velocities indicate that the rock near the locity dispersion at a wellhead pressure of 0.4 MPa is weaker
fracture was undergoing relaxation after the previous water-in- than that at a wellhead pressure of 3.0 MPa.
jection.
Fig. 4 shows waveforms detected during water injection. The IV. CRACK MODELS
crack-waves convert to tube waves at the intersection with a A basic crack model for a fluid-filled crack is a three-layer
bore hole. Tube-wave analysis indicated that the waves with a model, as discussed by Ferrazzini and Aki [11], in which a thin
large amplitude at 40 ms in Fig. 4 are crack-waves [4]. As the fluid layer, compared with the wavelength of the crack-waves,
wellhead pressure was increased, the waveforms of the crack- is sandwiched between two solid half-spaces. There is no con-
waves changed. tact between the two solids. The three-layer model is not valid
for the analysis of crack-waves propagating along an artificial
III. WAVELET ANALYSIS OF CRACK-WAVES subsurface fracture. In particular, since a low-velocity zone was
We used the wavelet transform (WT) to obtain spectral data detected even in the vicinity of a fracture created in an intact
as a function of arrival time in a dispersion analysis of the crack- rock layer [9], [10], a low-velocity zone should be taken into
waves [14]–[16]. Time and frequency resolutions vary in the account in an analysis of subsurface fractures. Contact between
WT, while they are uniform in the traditional short-time Fourier asperities on fracture surfaces is also inherent in subsurface frac-
transform (STFT). This variable resolution is an advantage of ture.
the WT over the STFT. The WT is defined as We examined crack-wave dispersions using two crack
models: the LVL model and the crack-stiffness model. The
(1) two physical properties, a low-velocity zone and contact, are
independently introduced into a three-layer model. These crack
models, which are independent of each other, demonstrate the
where is the analyzing wavelet, is a scale parameter, effects of these physical properties of a fracture.
is a time-shift parameter, and * denotes the complex conjugate. Dispersion equations for the LVL and crack-stiffness models
The scale parameter is the reciprocal of the frequency. The are derived using common potential functions for the wave
analyzing wavelet requires the admissibility condition given by equation. We assume that layers lie in the -plane and that the
normal to the wavefront lies on the axis. Periodic solutions
of the wave equation may be found by combining with a
(2) compressional wave solution

(4)
where is a Fourier transform of the analyzing wavelet . The
analyzing wavelet in this paper is the modulated Gaussian and a shear wave solution
given by
(5)
(3) where is the angular frequency, is the wavenumber, and
and are unknown functions of and . Sub-
When the analyzing wavelet is the modulated Gaussian, uncer- script denotes the number of a layer.
tainty in the time-frequency domain is minimum. Therefore, the Higher order modes show more oscillation of pressure along
modulated Gaussian is efficient for representing a spectral com- the axis in a fluid layer. The aperture of a fluid layer is quite
ponent that varies with time. The modulated Gaussian does not narrow in this analysis. Therefore, higher order modes have
strictly satisfy the admissibility condition. However, if much less energy than the fundamental mode. The numerical
6 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 38, NO. 1, JANUARY 2000

(a) (b)
Fig. 5. Wavelet transforms of the crack-waves. The arrival time of each frequency component can be seen. (a) The crack-wave detected at a wellhead pressure
of 0.4 MPa. These data were recorded at the beginning of the injection. (b) The crack-wave detected when the wellhead pressure was held constant at 3.0 MPa for
1779 s (about 30 min). The aperture of the subsurface fracture increases when the wellhead pressure approaches 3.0 MPa.

analysis of Ferrazzini and Aki [11] showed that only the funda-
mental symmetric mode of crack-waves was strongly affected
by the interface conditions at fracture surfaces. Therefore, in this
paper, we analyzed the fundamental symmetric modes. Since
the velocities of fundamental symmetric modes of crack-waves
are lower than the fluid velocity, in the above potential functions

(6)

where is the phase velocity of crack-waves, and and

are the velocities of compressional and shear waves of the th Fig. 6. Low-velocity layer model for an artificial subsurface fracture.
layer.
matrix method by Haskell [18]. The top solid layer has a free
A. Low-Velocity Layer (LVL) Model surface on one side and the bottom solid layer is an infinite
Fig. 6 schematically presents the LVL model. Additional half-space. The axis is parallel to the layers in the direction of
solid layers are found on both sides of the fluid layer. The propagation. The origin of the axis is the top surface of each
velocities of these two solid layers are lower than those of the layer. The axis is positive in the direction of the medium.
top and bottom bedrock layers. The low-velocity layer (LVL) The layers are numbered starting from the top solid layer. The
represents a micro crack zone that is created by pressurization thickness of the th layer is , and the boundaries of the th
caused by water injection. We assume coordinates to use the layer are at and . The th layer is characterized by
NAGANO AND NIITSUMA: DISPERSION ANALYSIS OF CRACK-WAVES 7

Fig. 8. Crack-stiffness model for an artificial subsurface fracture.

(a)

(b)
Fig. 7. Effects of the thickness of the low-velocity layer on the phase velocity
of crack-waves: (a) the compressional wave velocity in the low-velocity layer is
1700 m/s and (b) the compressional wave velocity in the low-velocity layer is
1900 m/s. Fig. 9. Dispersion curves of crack-waves in the crack-stiffness model.

the compressional wave velocity , the shear wave velocity the dispersion equation that satisfies these boundary conditions
, and the density . All of the layers are homogeneous and using the matrix method [7], [18].
isotropic, and the fluid is inviscid and incompressible. Fig. 7 shows dispersion curves in the LVL model. When the
At the interface between a bedrock layer and an LVL, normal thickness of an LVL increases, velocities decrease. When the
and shear stresses are continuous, as are displacements in the thickness of an LVL exceeds a certain limit, velocities are al-
and axes. most similar in a high frequency range. The limit frequency
At the interface between a fluid layer and an LVL, the normal depends on the thickness of the LVL and the wavelength. For
stress and normal displacement must be continuous and shear example, in Fig. 7(a), there is little difference between the dis-
stress vanishes. Thus persion curves for the LVL thicknesses of 1.0 and 5.0 m at a
frequency above 30 Hz. It is difficult to measure crack-waves at
a very low frequency because of the limited frequency perfor-
at (7) mance of the wave source. Therefore, this convergence means
that there is an upper limit for estimating the thickness of an
LVL when we measure the velocities of crack-waves.
where and are the normal and shear stresses, respec-
tively, σ is the pressure of the fluid, is the displacement of B. Crack-Stiffness Model
the solid in the axis, and is the displacement of the fluid. Fig. 8 shows the crack-stiffness model, in which contact be-
Since the top surface in the first layer is free, both normal and tween asperities on fracture surfaces is taken into account [8].
shear stresses vanish at the top surface. There are no sources at Two solids are connected with springs. A fluid layer is between
infinity in the bottom layer, so that . We obtain the solids, and its thickness is . The origin of the axis is at
8 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 38, NO. 1, JANUARY 2000

V. MODELING CRACK-WAVES
To estimate the physical parameters of these two crack
models from the observed crack-waves, we numerically
simulate the arrival times of crack-waves for the given crack
models. The contours in a scalogram represent the locations
of relative maxima of wave energy in the time and frequency
domains. A dispersive curve of the relative maxima is observed
in the initial motions of the crack-waves in Fig. 5. We fit the
arrival times of crack-waves to the dispersive relative maxima
in Fig. 5. The arrival times of wave energy are calculated from
group velocity and propagation distance. Group velocities are
calculated from the phase velocities obtained from dispersion
equations in the previous section. Because the hydrophone was
suspended at the depth of the intersection with the fracture and
the air gun was installed two meters above another intersection
(a) with the fracture, we neglect the distance between the air gun
and the intersection. The orientation of the artificial fracture,
which was estimated from tectonic stress measurement [19],
is almost the same as the inclination of the artificial fracture
in a core sample, which was obtained from well EE-4. Thus,
we assume that the artificial fracture is not curved between the
intersections of the bore holes. The propagation distance of
crack-waves is 6.7 m in calculations of the arrival time.

A. Arrival-Time Curves in the LVL Model

We examined the thicknesses of the fluid layer and the LVL
as parameters of the LVL model when calculated arrival times
were compared with the dispersion of measured crack-waves. To
simplify the calculation, the thicknesses of LVL’s are the same on
bothsidesofthefluidlayer.Wegavetypicalvaluesforthedensities
of rock and fluid and fluid velocity. The densities of the bedrock
(b) layerandtheLVLare2600kg/m3, andthedensityofthefluidlayer
is 1000 kg/m3. Fluid velocity is 1500 m/s. For other parameters of
Fig. 10. Arrival times calculated from group velocities with the LVL model
and a time-frequency representation of crack-waves at the beginning of water
the LVL model, data measured in the field are used. Velocities of
injection. (a) The fluid layer is 0.5 mm thick. (b) The fluid layer is 1.0 mm thick. compressional and shear waves in the bedrock layer are 3100 and
Compressional and shear wave velocities are 1900 and 1140 m/s, respectively. 1860 m/s, respectively. The velocity of a compressional wave in
The LVL is 0, 0.1, and 1.0 m thick in (a) and (b). the LVL is 1900 m/s at the beginning of water injection and 1700
m/s at the maximum wellhead pressure. We assume that Poisson’s
the center of the fluid layer. The boundary conditions at the in- ratio for the LVL is the same as that for the bedrock layer and is
terface between a fluid layer and a solid layer are constant during water injection.
We determined the thicknesses of the fluid layer and the LVL by
trialanderror. Fig. 10showssomecurvesforthecalculated arrival
times of crack-waves recorded at the beginning of water injection.
at (8)
The background noise level, which exists before the arrival of
crack-waves, can be recognized in a contour of the scalogram
when we compare the contour to a waveform (Fig. 5). The arrival
where and are the specific crack stiffnesses in directions of a waveoccurs at the onsetof a relativemaximum in ascalogram.
parallel and normal to the fluid layer, respectively. The sym- The measured crack-waves are less dispersive at a frequency
metric mode is a function of , and the antisymmetric mode above 30 Hz. On the other hand, calculated arrival-time curves
is a function of . A dispersion equation is obtained by substi- show strong dispersion at frequencies below 70 Hz. Therefore,
tuting stresses and displacements derived from (4) and (5) into calculated arrival-time curves are not similar to the dispersion
these boundary conditions [8]. of crack-waves measured at a frequency below 70 Hz. The best
Fig. 9 shows dispersion curves of crack-waves in the crack- combination in our trials was a fluid layer of 1.0 mm and an LVL
stiffness model. The thickness of the fluid layer is 0.2 mm, and of 0.1 m. With this combination, arrival-time curves fit in the
0, 45, 90, 225, 450 GPa/m for calculating the dispersion frequency range of 70–256 Hz. None of the arrival-time curves fit
curves in Fig. 9. Crack stiffness increases the phase velocities of the entire frequency range of 2–256 Hz in our trials.
crack-waves. Furthermore, the degree of dispersion decreases if Dispersion characteristics of crack-waves recorded at a well-
crack stiffness increases. head pressure of 3.0 MPa were also examined by fitting to ar-
NAGANO AND NIITSUMA: DISPERSION ANALYSIS OF CRACK-WAVES 9

ning of water injection. Therefore, it is reasonable to assume

that the geometry of the LVL varied due to the increase in well-
head pressure.
Fig. 11 shows some arrival-time curves for crack-waves
recorded at a wellhead pressure of 3.0 MPa. These crack-waves
are more dispersive than those at a wellhead pressure of 0.4
MPa. Differences between the dispersions are observed at
frequencies below 70 Hz. Differences between the arrival-time
curves and the crack-waves measured at a wellhead pressure of
3.0 MPa are smaller than those at a wellhead pressure of 0.4
MPa. However, none of the arrival-time curves showed a good
fit over the entire frequency range of 2–256 Hz. When the fluid
layer is 2.0 mm and the LVL is 0.1 m, the arrival-time curve
shows a good fit with the onset of the relative maxima in our
(a) trials, even though the fit is not exact.
Differences between the arrival-time curves and the measured
crack-waves decreased when the wellhead pressure increased.
This means that the LVL model is more suitable for a highly
pressurized subsurface fracture rather than for a closed subsur-
face fracture. However, the estimated thicknesses of the fluid
layer and the LVL are not consistent with other data (i.e., the
transmissibility test and crack-wave measurement). When the
fluid layer is 2.0 mm and the LVL is 0.1 m, the arrival-time
curve showed a good fit to the crack-waves recorded at a well-
head pressure of 3.0 MPa in our trials. On the other hand, the
thickness of the fluid layer was estimated to be 0.2 mm even at
a wellhead pressure of 3.0 MPa in the transmissibility test [13].
The compressional wave velocity of rock is 3100 m/s in this
field and 1700 m/s in the vicinity of the fracture at a wellhead
pressure of 3.0 MPa. The air gun was suspended two meters
(b) above the intersection of a fracture, and the hydrophone was
installed at another intersection. The estimated thickness (0.1
m) of the LVL is too thin to decrease compressional wave ve-
locity in the vicinity of the fracture. If the thickness of the LVL
increases, the fluid layer should be thicker than 2.0 mm in the
LVL model. However, we cannot agree with a fluid layer thicker
than 2.0 mm, since a fluid layer of 2.0 mm is ten times as thick as
the result of the transmissibility test. Therefore, it is impossible
to simulate arrival-time curves based solely on the LVL model.
We should introduce mechanisms that increase the velocity of
crack-waves in the LVL model.

B. Arrival-Time Curves in the Crack-Stiffness Model

Specific crack stiffness in a direction normal to the fluid layer
and the thickness of the fluid layer were examined in the crack-
(c) stiffness model. The physical properties of the bedrock layers
Fig. 11. Arrival times calculated with the LVL model and a time-frequency and the fluid layer are the same as in the calculations for the
representation of crack-waves at a wellhead pressure of 3.0 MPa. (a)–(c) The LVL model. Fig. 12 shows arrival-time curves on a scalogram of
fluid layer is 0.5, 1.0, and 2.0 mm thick, respectively. Compressional and shear crack-waves recorded at a wellhead pressure of 0.4 MPa. Com-
wave velocities are 1700 and 1020 m/s, respectively. Dispersion curves are
calculated at LVL thicknesses of 0, 0.1, and 1.0 m in (a) and (b). binations of crack stiffness and fluid-layer thickness were found
by trial and error. Arrival-time curves for a high crack stiffness
show weak dispersion. They fit the relative maxima in the scalo-
rival-time curves calculated with the LVL model. This wellhead gram better than those in the LVL model. The best combination
pressure is higher than that in Fig. 10. A transmissibility test in our trials was a crack stiffness of 90 GPa/m and a fluid layer
showed that the aperture in the artificial subsurface fracture in- thickness of 0.2 mm.
creased at a wellhead pressure of 2.0 MPa [13]. The velocity of Calculated arrival-time curves and crack-waves, recorded at
a compressional wave of the LVL was 1700 m/s at this wellhead a wellhead pressure of 3.0 MPa, are compared in Fig. 13. The
pressure (Fig. 3). This velocity is lower than that at the begin- crack-waves at a wellhead pressure of 3.0 MPa were more
10 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 38, NO. 1, JANUARY 2000

(a) (a)

(b) (b)

(c) (c)
Fig. 12. Arrival times calculated with the crack-stiffness model and a Fig. 13. Arrival times calculated with the crack-stiffness model and a
time-frequency representation of the crack-waves at the beginning of water time-frequency representation of crack-waves at a wellhead pressure of 3.0
injection. (a)–(c) The fluid layer is 0.2, 0.5, and 1.0 mm thick, respectively. The MPa. (a)–(c) The fluid layer is 0.2, 0.5, and 1.0 mm thick, respectively. The
arrival times of crack-waves were calculated at specific levels of crack stiffness. arrival times of crack-waves were calculated at specific levels of crack stiffness.

strongly dispersive than those at a wellhead pressure of 0.4 MPa.

the wellhead pressure increased, the thickness of the fluid layer
When the crack stiffness is 18 GPa/m and the thickness of the
increased from 0.2 to 0.5 mm, and the crack stiffness decreased
fluid layer is 0.5 mm, the arrival-time curve shows a good fit
from 90 to 18 GPa/m. In a simple model of a subsurface fracture
to the relative maxima in the scalogram for crack-waves at a
shown in Fig. 14, stiffness per unit area in the direction normal
wellhead pressure of 3.0 MPa in our trials.
to the interface is
The estimations of crack stiffness and the thickness of the
fluid layer are consistent with the opening of the fracture. When (9)
NAGANO AND NIITSUMA: DISPERSION ANALYSIS OF CRACK-WAVES 11

[5] K. Nagano and H. Niitsuma, “Crack stiffness from crack wave veloci-
ties,” Geophys. Res. Lett., vol. 23, pp. 689–692, Mar. 1996.
[6] K. Nagano, K. Sato, and H. Niitsuma, “Polarization of crack waves
along an artificial subsurface fracture,” Geophys. Res. Lett., vol. 23, pp.
2017–2020, Aug. 1996.
[7] K. Nagano, “Crack-wave dispersion at a fluid-filled fracture with low-
velocity layers,” Geophys. J. Int., vol. 134, pp. 903–710, Sept. 1998.
[8] K. Hayashi and K. Sato, “A theoretical study of AE traveling through
a fluid-filled crack with application to characterization of a geothermal
reservoir crack,” in Progress in Acoustic Emission VI, T. Kishi, K. Taka-
hashi, and M. Ohtsu, Eds. Tokyo, Japan: Japanese Soc. NDI, 1992, pp.
423–430.
[9] H. Niitsuma and H. Saito, “Characterization of a subsurface artificial
fracture by the triaxial shear shadow method,” Geophys. J. Int., vol. 107,
pp. 485–491, Nov. 1991.
[10] K. Tanaka, H. Moriya, H. Asanuma, and H. Niitsuma, “Detection
Fig. 14. Simple crack model.
of travel time delay caused by inflation of single artificial fracture,”
Geotherm. Sci. Technol., to be published.
where is the compressional stress normal to the interface, [11] V. Ferrazzini and K. Aki, “Slow waves trapped in a fluid-filled infinite
crack: Implication for volcanic tremor,” J. Geophys. Res., vol. 92, pp.
is the entire area, is the contact area, and is Young’s mod- 9215–9223, Aug. 1987.
ulus. The decrease in crack stiffness, which was observed in the [12] H. Niitsuma, “Fracture mechanics design and development of HDR
crack-wave measurement, can be explained by an increase in reservoirs-concept and results of the 0-project, Tohoku University,
Japan,” Int. J. Rock Mech. Min. Sci. Geomech. Abstr., vol. 26, no. 3/4,
the aperture of the fracture and by a decrease in the contact area pp. 169–175, 1989.
between asperities on crack surfaces, or both. [13] K. Hayashi and H. Abé, “Evaluation of hydraulic properties of the artifi-
cial subsurface system in Higashihachimantai geothermal model field,”
J. Geotherm. Res. Soc. Jpn., vol. 11, no. 3, pp. 203–215, 1989.
VI. CONCLUSIONS [14] F. Hlawatsh and G. F. Boudreaux-Bartels, “Linear and quadratic time-
We have investigated the dispersions of crack-waves using frequency signal representations,” IEEE Signal Processing Mag., vol. 9,
pp. 21–67, Apr. 1992.
two crack models. We analyzed crack-waves that were mea- [15] S. Qian and D. Chen, Joint Time-Frequency Analysis, Methods and Ap-
sured in an artificial subsurface fracture at a depth of about 370 plications. Englewood Cliffs, NJ: Prentice-Hall, 1996.
m. The WT of the crack-waves measured at an artificial sub- [16] L. J. Pyrak-Nolte and D. D. Nolte, “Wavelet analysis of velocity disper-
sion of elastic interface waves propagating along a fracture,” Geophys.
surface fracture showed positive dispersion. The crack-waves Res. Lett., vol. 22, pp. 1329–1332, June 1995.
were more strongly dispersive at a wellhead pressure of 3.0 MPa [17] R. Kronland, J. Morlet, and A. Grossmann, “Analysis of sound patterns
than at a wellhead pressure of 0.4 MPa. The LVL model and the through wavelet transforms,” Int. J. Pattern Recognit. Artif. Intell., vol.
1, no. 2, pp. 273–302, 1987.
crack-stiffness model were used to calculate arrival-time curves [18] N. A. Haskell, “The dispersion of surface waves on multilayered media,”
of crack-waves. The arrival-time curves calculated with the LVL Bull. Seismol. Soc. Amer., vol. 43, pp. 17–34, 1953.
model were late relative to the measured crack-waves. On the [19] K. Hayashi, T. Ito, and H. Abé, “In situ stress determination by hydraulic
fracturing—A method employing an artificial notch,” Int. J. Rock Mech.
other hand, the curves calculated with the crack-stiffness model Min. Sci. Geomech. Abstr., vol. 26, no. 3/4, pp. 197–202, 1989.
agreed with the dispersions of the crack-waves. We determined
the best combinations of the thickness of the fluid layer and the
crack stiffness at two levels of wellhead pressure (0.4 and 3.0
MPa). These estimates with the crack-stiffness model are con- Koji Nagano received the B.Eng., M.Eng., and the Dr.Eng. degrees in electrical
sistent with opening of the fracture. engineering from Tohoku University, Sendai, Japan, in 1983, 1986, and 1989,
respectively.
We noted that contact and the low-velocity zone are important From 1989 to 1991, he was a Fellow of the Japan Society for the Promotion
physical properties for estimating subsurface fractures. Contact of Science for Japanese Junior Scientists. He was a Research Associate at the
is expressed in the crack-stiffness model. On the other hand, Department of Resources Engineering, Tohoku University from 1991 to 1992,
and then joined the Department of Computer Science and Systems Engineering,
the LVL model represents a low-velocity zone. The results of Muroran Institute of Technology, Muroran, Japan, where he is currently a Re-
the crack-wave dispersion indicate that contact is more domi- search Associate. His research interests are seismic signal processing, cross-hole
nant than the low-velocity zone in physical properties that affect seismic measurement, and wave propagation for characterization of subsurface
crack.
crack-wave dispersion.

REFERENCES
[1] B. Chouet, “Dynamics of a fluid-driven crack in three dimensions by the Hiroaki Niitsuma received the Ph.D. degree from Tohoku University, Sendai,
finite difference method,” J. Geophys. Res., vol. 91, pp. 13 967–13 992, Japan, in 1975.
Dec. 1986. He has been a Professor with the Department of Geoscience and Technology,
[2] Y. G. Li, P. C. Leary, and K. Aki, “Observation and modeling of Faculty of Engineering, Graduate School of Tohoku University since 1988. His
fault-zone fracture seismic anisotropy—II: P-wave polarization anoma- recent research interests are in multicomponent seismic measurement and signal
lies,” Geophys. J. R. Astron. Soc., vol. 91, pp. 485–492, 1987. processing, bore hole measurement, and characterization of geothermal reser-
[3] M. Lou, J. A. Rial, and P. E. Malin, “Modeling fault-zone guided waves voir. He is a Representative of the International Collaboration Program “Es-
of microearthquakes in a geothermal reservoir,” Geophysics, vol. 62, pp. tablishment of new mapping/imaging technologies for advanced energy extrac-
1278–1284, July/Aug. 1997. tion from deep geothermal reservoirs (MTC project),” funded by NEDO and
[4] K. Nagano, H. Saito, and H. Niitsuma, “Guided waves trapped in an MESC. He was Chairman of the Subsurface Instrumentation Division of MMIJ
artificial subsurface fracture,” Geotherm. Sci. Technol., vol. 5, no. 1/2, from 1989 to 1995. He is now a Director of SEGJ, MMIJ, the SPWLA Japan
pp. 63–70, 1995. Chapter, and a Councilor of the Geothermal Research Society of Japan.