GEOPHYSICAL RESEARCHLETTERS, VOL. 13• NO. 1, PAGES 149-152,
OBSERVATIONS OF LONG PERIOD
FEBRUARY 1986
EARTHQUAKES ACCOMPANYING HYDRAULIC FRACTURING
1
Dorthe
F.arth
Bame
and Space Sciences
and
Division,
Michael
Fehler
Los Alamos National
Laboratory
Abstract.
Waveforms of most seismic
events
accompanying hydraulic
fracturing
have been reported
to
contain
clear
P and S waves and have fault
plane
solutions
consistent
with
shear
displacement
Julian,
1985].
Since
hydraulic
fracturing
involves
the injection
of fluid
into
rock in much
the same Way that
movement
of magma beneath
a
volcano does, LP events occurring
during hydraulic
across a fault.
since classical
fracturinq
experiments
may indicate
the above mechanisms is occurring.
This observation
hydraulic
fracturing
is
surprising
theory
pre-
that
one of
dicts
the creation
of a tensile
opening
of a cavity
in response
to fluid
pressure.
Very small
long period
events,
similar
to long period
earth-
Method
quakes observed at volcanoes, were found to occur
during four hydraulic
fracturing
experiments carried out at Fenton Hill,
New Mexico.
Since the
To identify
potential
long period events, data
recorded by a seismic sensor located in wellbore
EE-1 during two hydraulic
fracturing
experiments
long period
earthquakes
occur in the same region
as the shear type events,
it is concluded that the
(termed 2018 and 2023)
Rock geothermal
energy
unusual
character
of the
long
period
earthquake
waveforms is due to a source effect
and not a path
effect.
The occurrence of long period earthquakes
during hydraulic
fracturing
could indicate
tensile
fracturing.
Many waveforms of these events are
identical,
which implies
that these events repre-
Mexico
sent
r•peated
proposed
activation
source
for
of
these
a
given
long
source.
period
Dry
New
were studied
in detail.
In addition,
we
briefly
looked at data from two other experiments,
called
2012 and 2032.
Figure 1 shows the injection
intervals
for the four hydraulic
fracturing
experiments,
the location
of the triaxial
geophone
package in well EE-1 and the geometry of the wells
A
events
conducted
at the Hot
site
at Fenton Hill,
is
F..E-2
and
F.E-3
experiments.
that
were
Preliminary
used
for
studies
the
fracture
of the data
the sudden opening of a channel that connects two
showed that very small signals were recorded that
cracks
were
filled
The sizes
with
fluid
each
crack.
spectral
to
be
different
of the two cracks
two or more peaks to
peaR: being associated
of
at
between
occur,
3
and
which causes
different
previously
from
reported
the
frequencies
crack
lengths
at
are
which
hydraulic
East
Geophone
out in low permeabeen observed
and
microseismic
experiments
[Pearson, 1981; House et al.,
frequencies
in
range
with
observations
of
ß
1985] with corner
80 - 800
Hz.
and a model of McGarr
turing,
is
mechanism,
occurring
For many years,
long period
volcanoes.
such
[Murphy
as
tensile
and Fehler,
volcanologists
E
oscillations
in
Shimozuru,
1961 ],
along
which
magma
al.,
1983;
Cbouet,
fluJd-fillmd
cracks
[Aki
at.
al.,
jacking)
1977,
-3900
-4100
magma chambers
(e.g.,
E-3
near
as
B•me,
1985;
fluid
conduits
[Farrick
at.
openings
Cbouet
-4300
PBR/Liner•(l
k
Exp.2012- 3130rn=/•x
-4500
of
-
- 1200
and
•
- 1000
J
I
-800
-600
Horizontal
1
Now at
Sandia
National
Number
, I
I
I
-400
-200
0
Distance
200
400
(m)
Figure
1. East-West
vertical
cross
section
map
showing
locations
of
hydraulic
fracturing
experiments
studied
and volume of fluid
injected
during
each experiment.
The location
of the three
component
borehole
seismometer
used
to
detect
Laboratory
This paper is not subject
to U.S. copyright.
Published
in 1986 by Americal
Geophysical
Union.
Paper
ons
Exp. 2032 - 21,200rn =
1984].
(LP) seismic
events
and tremor
LP events
have been interpreted
tensile
Topof Sanded-In
•: -3700
½•
•
have observed
[Crosson
and
oscillations
in
is
transported
and
/
Liner
frac-
being due to
1985]
Casing
Exp.
2018
-900rn
= i
,,
cumulative
seismic
moment of these
accounts only for a small portion
of
moment expected
for
the volume
of
other
Cernented-ln
-3500
Compared
" Exp.
2023
-150m
•
-3300
signals
occurring
are
shear
events
water injected during an experiment which suggests
some
.- CasingShoe
-3100
used to monitor the growth of the fracture system.
that
were
The small
in EE-1
fracturing
experiments
carried
bility
crystalline
rocks
have
[1976],
the
shear events
the seismic
that
Observation
•
Microseismic events induced during
othe'r
events
West
20m.
-2900
the
shear
-2700
estimated
Introduction
The majority
of
during
fracturing
the
by Pearson [1981].
appear in the spectra,
each
with
one physical
dimension
From
peaks
differ,
pressures.
events
51,6753
149
is
shown.
150
Bame and Fehler:
Long Period
Earthquakes
ing,
from
which
Hydrofracturing
have
corner
frequencies
in
the
range
of
80-800
Hz.
Both shear
and long
period
events
accompanying
hydraulic
fracturing
are very
small
j.n amplitude
and short
ip duration
when compared
to
their
tectonic
and
volcanic
counterparts
ß
Seismic events were found to occur sporadically
at
the beqinning
of an experiment
and the rate
at
150 ms
which
events
occurred
and ten per minute
increased
later
in
to
between
five
the ex3.•eriment.
The
first
events to occur in each experiment
were long
period.
After
approximately
two hours into
an
experiment,
most of the events
recorded
contained
clear p and S arrivals
and appeared to have a wide
150 ms
range
be
of
frequencies
shear
and,
hence,
are
considered
to
events.
Dany long period events with nearly
identical
waveforms
were observed
and these
were grouped
into earthquake families
1975].
In experiment
[Hamagucbi anQ Basegawa,
2018 there
were periods
during
which
certain
families
dominated.
Events
in other
families,
however,
occurred
indiscrimin-
150 ms
ately
throughout
the early
part
of the e•periment.
In
order
to
study
details
of
the
seismic
wa v•forms,
all
three
components
of
motion
vertical,
and both
horizontals
- of
a number
of
events
of
from
the
family
LP
were
digitized.
waveforms
were
Spectra
calculated
and
corrected
for
the
electronic
response
of
the
instrument
and recording
equipment.
Spectra from
events
in each family
were nearly
identical.
In
addition,
spectra
from some families
were simila•;
these
families
were
grouFed
together
and. are
called
a class
(see Table 1).
The grouping
of
150 ms
earthquake
800
each
entire
waveforms
made only
ms
for
the
into
purpose
families
of
this
and
classes
empirical
is
study.
Data
The long
were divided
period
events
from
experiment
2018
into
eight
families
and five
classes.
150 ms
125
150 ms
b
,o
-8.8
98
150 ms•
Figure
LP
2. Vertical
events.
component
The
first
waveforms
five
of
events
typical
are
from
e•periment
2018: (a) Family A, (b) Family P, (c)
Family C, (d) Family H, (e) Family T.
The rest of
the events are from experiment
2023: (f) Family A,
(g) Family
B, (h) Family D.
signals
were dominated
Hz) •nd generally
These signals
are
similarity
volcanoes
to
by low frequencies
lacked a second
called
long period
long
[Minikami,
period
siqnals
1974].
period" is used to differentiate
the shear events accompanying
(100-300
d
C
lOOO
FREQUENCY
Figure
Figure
3. Vertical
2:
(a)
lOO
lOOO
IHz)
component spectra
of
Family
A,
Spectral
events
Class
in
1,
(S) arrival.
due to their
experiment
experiment
2018;
2018;
(b)
(c)
Family
Family
C,
T,
Spectral
Spectral
Class
Class
3,
5,
observed
(d)
Family
A,
Spectral
Class
1,
experiment
2018;
"long
experiment
2023.
these events from
hydraulic
fractur-
to in text.
are spectra
The
lO
term
at
Dots denote the peaks referred
Lower spectrum in each a, b,
of noise preceeding
the event.
and e
Bame and Fehler:
TABLE
SPECTRAL
PROPERTIES
OF
1
VERTICAL
WAVF..FORMS OF LONG PF,RIOD
SPECTRAL
WAVEFORM
CLASS
F,xp.
Long Period Earthquakes from Hydrofracturing
events
look very similar
to those containing
the
high frequency component.
Preliminary
studies
of two other experiments,
COMPONENT
EARTBQUAKES
PEAK FREQUENCI FS
FAMILIES
151
( •z )
2012 and 2032,
(Figure
1) indicate
that
LP events
accompanying
these two experiments
are similar
to
those
seen in 2018 and 2023.
Experiment
2032
differs
from
the
it
place
in
took
other
three
a portion
experiments
of
the
in
that
had already been fractured.
The major difference
in the LP seismic signals
from this experiment
is
the absence of a high frequency component that was
2018
1
2
A
B
3
4
5
112,
115,
C,D,F,G
H
T
150,
185,
180,
268,
745
415
112, 185,
Many <400
110, 185,
268,
550
previously
superimposed
At
this
time,
long
hydraulic
A,F
2
3
B
C,D,E,I,J
the
low
frequency
Discussion
explain
1
on
signal.
750
Exp. 2023
no well
period
fracturing.
of
long
accepted
model
earthquakes
However,
period
exists
observed
we can
volcanic
The events in experiment
draw
upon
earthquakes
and
130, 800
models
130
110,130,180,850
volcanic tremor to guide our interpretation of the
data presented above. Since the waveforms of all
2023 were divided
and three classes.
into
Figure 2 shows
identical
waveforms have been reported
Figure
hydraulic
1 lists
the
families
that
make up
fracturing
are
very
small,
be
at Volcano
due to repeated activation
of one fault.
Since long period
earthquakes
that
Table
to
events
within
nearly
Usu in Japan [Takeo, 1983] and interpreted
vertical
component waveforms of representative
LP
earthquakes.
Spectra of some events •re shown in
3.
to
during
events
in a family
of earthquakes
appear
nearly
identical,
we must conclude
that
all
within
a family
occur at the same location
the Hot Dry Rock reservoir.
Events
with
nine families
that
reservoir
to be
accompany
we have
not
each spectral
class,
and the frequencies
at which
prominent peaks in the spectra
of the vertical
component waveforms occur.
Some of the characteristics
of the waveforms and their
spectra
will
now
detected
one event at enough stations
to allow us
to determine the location
of the event.
We can,
however, assume that at least
some of the events
•re located
very near to the point
of injection
of
be discussed.
water
into
the rock since
rock is very low and long
Experiment
2018
almost
shear
immediately
after
earthquakes
the permeability
period
earthquakes
injection
have
been
of the
begin
begins.
found
to
Since
occur
near
Most
LP events
from
Experiment
2018
are
characterized
by a waveform
that
is a superposition
of a high frequency
component and a lower
frequency
component as shown in Figure
2a.
The
almost monochromatic
high frequency
portion
of the
waveform is dominant.
As seen in Figure
3a, there
are two major peaks in the spectrum of the verti-
the injection
point,
and to completely
surround it
[Fouse et al.,
1985],
the path between many shear
events and the receiver
must be nearly
identical
to that for some, if not all,
of the long period
events.
We can thus rule out the path affect
as
being the dominant
cause of the unique
character
of long period earthquakes.
cal component.
Other examples of LP events for
F.xperiment 2018 are shown in Figure 2(b-e).
Some events
recorded
during
Experiment
2018
Since
the
long period
events
occur
immediately
after
fluid
injection
begins,
unreasonable
to attribute
their
source to
have
characteristics
similar
to
volcanic
tremor.
to
fluid
oscillations
in
a
fluid
flow
almost
it
is
be due
conduit
in
These events have emergent onsets and vary from
one to four seconds in length.
Figure 2e shows a
800 ms portion of one of these events.
The
the rock.
Since the granitic
rock into which
fluid is being injected has very low permeability
[Pearson, 1981], it is reasonable to conclude that
spectra
the
of
these
which portion
of
events
were
the
same no matter
the waveform was chosen for
analysis.
The spectra
contain
two large
about 110 and 185 Hz (Figure 3c).
There
bread peak close to 750 Fz.
Experiment
peaks at
is also a
2023
The LP events
that
Experiment
2023 were
occurred
most often
during
similar
to those that
oc-
curred
most often
during
that
the high
frequency
Experiment
component
2018,
was
except
not as
source
tensile
of
the
long
period
events
opening of a crack driven
is
the
by the fluid
pressure.
Most
investigators
also
conclude
that
unique
character
of the waveforms
recorded
volcanoes
during
long
period
earthquakes
the
at
and
tremor
is due to a source
effect,
not a path
effect
[Aki and Kayanagi,
1981; Fehler and Chouet,
1982].
Fehler
[1983]
has argued
that
tremor
is
composed of a sequence of long period earthquakes
that
occur
close
individual
events
enough
together
in
are indistinguishable.
time
that
Numerous
strong during Experiment 2023.
The spectra of the
two sets of events are also similar
except for a
slight
shift
in the frequencies at which peaks
attempts have been made to interpret
the peaks
observed in spectra of LP volcanic earthquakes and
tremor in terms of the geometry of the source
occur and the occurrence of
region.
both sides of the two
Some of the events
component altogether
frequency
component
secondary peaks on
major peaks of 2023 spectra.
lacked
the high
frequency
(Figure
2g) although
the low
of the waveforms
of these
A model that
is particularly
appropriate
for the present
situation
is that of A•i et.
[1977] in which injected
fluid
propagates
along a
tensile
crack
which extents
due to increase
in
fluid
pressure
in the crack.
Alternatively,
a
152
•ame and Fehler:
closed
channel
connecting
Long Period Earthquakes from Bydrofracturing
two open fluid
cracks may suddenly open, allowing
transported
between
different
pressures.
filled
magma ascent
to
J. Ge.o•hys. Res., 86, 7095-71 09.
fluid
cracks
containing
fluid
Once fluid
transport
be
at
has
occurred, the pressure in the high pressure crack
drops,
allowing
the
channel
to
close
until
pres-
sure builds up again as more fluid
is injected.
This process repeats itself
until
the pressure in
the
low pressure
the
channel
crack
increases
permanently
enough
open.
to
The size
keep
of
the
cracks on each side of the channel may be estimated from the spectral peak frequencies. Aki et.
al.,
[1977]
open
cracks
narrow
modeled a simple case in which two
of
equal
channel
spectral
mately
of
length
width
peak occurred
.88/2L
where
L,
are
AL.
separated
In
this
case,
at a frequency
8 is
the
S-wave
by
a
the
it
opens,
after
with
example,
of
therp
frequencies
for
events
will
given
studied
be two
by
with
spectral
.8B/L.
peak
130 and 800 Bz in a material
frequencies
with
gradients
in
fluid
fluid
pressure.
probably
pressure
they
Since
these
occur in direct
may be the
small
response to
most important
clue
J.
and D.A.
Geophys.
Bame.,
low frequency
Res.,
A spherical
volcanic
90,
source
earthquakes.
J. Geophys. Res., in press, 1985.
Fehl--er, M., ObserVations of volcanic tremor at Mr.
St.
Helens
3476-3485,
Fehler,
Volcano,
J.
network
long
under
Letters,
M.,
A.
Operation
on Mt.
of
originate
Ferrick,
Geophy.
s.
R.,
88,
1983.
M. and B. Chouet.,
Source
S-wave
Dynamics of an expanding
crack,
1985.
R.S.
model for
Research
For
velocity
3.5 km/sec,
the characteristic
lengths
would be 20m and 3m respectively.
The
long
period earthquakes
can be interpreted
as being
more directly
associated with tensile
jacking and
probably
occur only in a region where there are
earthquakes
Crosson,
observations
of
ately
peaks
B. and B. Julian,
fluid-filled
11187-11198,
velocity
the channel closes almost immedi-
Hawaii,
Geo•hys. Res., 90, 1881-1893, 1985.
Chouet,
seismic
If
Kilauea,
Chouet, B., Excitation
of a buried magmatic pipe:
A seismic source model for volcanic
tremor,
J.
of approxi-
the rock.
mechanism under
St.
period
the
9,
seismic
1017-1020,
of
a digital
events
volcano,
that
Geophysical
1982.
Qamar and W.F.
mechanism
of
Helens Volcano and
St.
Lawrence.,
volcanic
tremor.
J.
Geophys, Res., 87, 8675-8683, 1982.
Hamaguchi,
H.,
and
A.
Hasegawa.,
Recurrent
occurrence
of
earthquakes
with
similar
waveforms and its related problems.
J. Seismo
Soc. Japan, 28, 153-159, 1975 (in Japanese).
House, L.,
H. Keppler
and H. Kaieda,
Seismic
studies
of
a
experiment,
105-110,
massive
Trans.
hydraulic
Geoth.
fracturing
Res.
Council,
regarding
tensile
fracturing
when high pressure
fluid
is injected
into rock.
At this time, our
model for these earthquakes
is necessarily
crude
MeGart, A., Seismic moments and volume changes,
Geophys. Res., 81, 1487-1494, 1976.
Minikami,
T., Seisinology of volcanoes in Japan,
because of
its
geometrical
simplicity.
In
addition,
tbp calculation
of far field
spectra
using a model that incorporates
the fluid motion
inside the tensile
fracture,
such as discussed by
Physical
Volcanology,
edited by Civetta,
L.,
and A. Rapulls, 1974.
Murphy, H. and M. Fehler,
Chouet
and Julian
estimates
(1985),
may yield
more reliable
hydraulic
of source size.
Acknowledgements.
with
F•von Stephani,
Everett
fracturing
analyzed•
the assistance
Horton
in
this
paper
of Bert Dennis,
and Jerry
Kolar.
We
Pearson,
C.,
Seismic observations
of
The
microseismicity
Hydraulic
J.
in
Elsevier,
New York,
P. GaspariDi, G. Luongo
granite
rock
Dry Rock geothermal reservoir,
Geophys. Union, 65, 1011, 1984.
The d•ta
were collected
9,
1985.
in
EOS Trans.
relationship
and high
stimulation
of
a Hot
Am.
between
pore pressure
experiments
during
in
low
benefited
greatly
from technical
discussions and
critical
reviews of the manuscript by Leigh Bouse,
Hugh Murphy, •ei Aki and Scott Phillips.
This
work was funded by the U.S. Deprtment of Energy
permeability
granite
rocks, J. Geophys. Res.•
86, 7855-7864, 1981.
Shimozuri, D., Volcanic micro-seisms - discussions
on the origin.
•ull,
Volcanol. Soc. Japan, 5,
and the governments
of the Federal
Germany and Japan.
We also greatfully
Republic
of
acknowledge
154-162,
Takeo,
M.,
Greg Nunz, who found the money in
tight budget to support this work.
an otherwise
1961.
Source
mechanisms
in
Japan,
earthquakes
and
implications.
Physics of
Planetary
Interiors,
Usu
their
the
Volcano,
tectonic
Earth and
37,
241-264,
1983.
Sandia
National
Laboratory,
References
D.
Aki,
K., M. Fehler and S. Das.
of volcanic
and their
Eruption,
259-287,
Source mechanism
tremor; Fluid-driven
application
to the
J. Volcan.
1977.
and
crack models
1963 Kilauea
Geothermal
Res.,
Bame,
MS5323A,
P.O. Box 5800, Albuquerque, NM 87185.
M. Febler, MSJ979, Los Alamos National
Laboratory,
Los Alamos, NM 87545.
2,
Aki, •. and R. Koyanagi. De.•p volcanic tremor and
(Received
November
12,
1985;
Accepted November27, 1985)