PRENLAB PROGRESS REPORT - 1.2 - Subproject 3
Year 1, yearly report, December 1996
Subproject 3:
MONITORING STRESS CHANGES BEFORE EARTHQUAKES USING SEISMIC SHEAR-WAVE
SPLITTING
Stuart Crampin scrampin@ed.ac.uk tel: +44 131 650 4908
fax: +44 131 668 3184
Helen J. Rowlands hjr@glg.ed.ac.uk tel: +44 131 650 8533
fax: +44 131 668 3184
Department of Geology and Geophysics, University of Edinburgh (UE), Grant
Institute, West Mains Road, Edinburgh EH9 3JW, Scotland UK
REPORT 1:
1 MILESTONES
As specified in the Work Program, the first year was to be spent on:
TASK 1: Identify optimal stations and search for precursors
The world-wide-web is being used to access seismic data from the SIL
seismic network. There are a variety of minor technical problems accessing
data, the solution of which involves interaction between UE and the
Icelandic Meteorological Office, which are progressively being resolved.
Seismicity maps demonstrate that there are a number of stations sited over
sufficient seismicity for analysis of shear-wave splitting to be viable.
Shear-wave splitting is observed but the splitting displays unusually large
time-delays which may be caused by high-temperatures and/or high pore-fluid
pressures both of which may well be present in the upper half of the
Icelandic crust.
The remarkable observations are that the first 100 days of the first SIL
seismic station we looked at, SAU, showed variations in shear-wave
splitting similar to those observed before earthquakes. Such behaviour can
be interpreted (and numerically modelled) as the effects of increasing
stress on the stress-aligned intergranular microcracks present in almost
all rocks. These variations at SAU were reported at the internal PRENLAB
meeting in Reykjavik, 10th September, 1996, but since this was the first
data analyzed, the results were not taken seriously. In fact, SAU is about
160 km WSW of the eruption site beneath the Vatnajökull ice cape. It is
suggested that the changes in shear-wave splitting at SAU were the result
of increasing pressure as magma was injected into the lower crust beneath
the eventual eruption site, see Figure 1. A Letter to Nature (Crampin,
Rowlands & Stefánsson, 1996) is being prepared, and a preliminary version
is attached to this report.
The observed variations at SAU are a demonstration that shear-wave
splitting is a sensitive monitor of current stress changes, and does not
depend on the complicated interactions in earthquake preparation zones. It
is very encouraging to see the effect of changes so early in the project.
(1) They confirm that shear-wave splitting has the potential for monitoring
the detailed stress behaviour in Iceland.
(2) They confirm that Iceland is an active natural laboratory for research
on earthquake source zones and volcanic manifestations.
(3) They suggest a number of new projects, listed below, to improve stress-
monitoring of the crust beneath Iceland.
FIGURE 1 CAPTION
Fig. 1(a). Schematic illustration of shear-wave splitting in stress-
aligned microcracks.
Preliminary analysis of shear-wave polarizations and time-delays from small
earthquakes beneath seismic station SAU, for 137 days from 1st May 1996.
Fig. 1(b). Equal-area polar map out to 35 degrees of shear-wave
polarizations, with rose diagram indicating the direction of maximum
horizontal stress (N44 degrees E).
Fig. 1(c). Equal-area polar map out to 35 degrees of normalized time-
delays between split shear waves.
Fig. 1(d). Variation with time of time-delays showing angular differences.
Open circles (and dashed line) are for ray paths at angles of 0 to 15
degrees to the crack plane, which are sensitive to crack density. Solid
circles (and solid line) are for ray paths at angles of 15 to 35 degrees
from the crack plane, which are sensitive to aspect-ratio. The arrow marks
the time of the eruption (30th September 1966) beneath the Vatnajökull ice
cap.
An increase in time-delays for ray paths 15 to 35 degrees from the crack
plane as indicated by the solid line is observationally and theoretically
diagnostic of an increase in the maximum horizontal stress acting on the
rock. The increase is interpreted as indicating increasing stress as magma
is injected into the crust beneath Vatnajökull.
2 PUBLICATIONS
2.1 Publications directly associated with project
Crampin, S., Rowlands, H. J. & Stefánsson, R., 1996. Monitoring stress
during magma injection, Nature, in preparation. (Attached in Section 4
REPORT PAPER)
Crampin, S. & Zatsepin, S. V., 1996, Forecasting earthquakes with APE, in
Seismology in Europe, European Seismological Commission, 318-323.
2.2 Associated Abstracts
1. Zatsepin, S. V. & Crampin, S., 1996. Modelling rockmass deformation
with APE: a mechanism for monitoring earthquake preparation zones with
seismic shear-waves (Abstract), XXV General Assembley, European
Seismological Commission, 9th Sept., 1996, Reykjavik, Paper E3.06.
2. Crampin, S. & Zatsepin, S. V., 1996. Forecasting earthquakes with APE
(Abstract), XXV General Assembley, European Seismological Commission, 9th
Sept., 1996, Reykjavik, Paper E3.07.
3. Crampin, S., 1996. Earthquake prediction and earthquake forecasting
(Abstract), Earthquake Research in Turkey: State of the Art, 30th Sept.-
5th Oct., 1996, Ankara.
4. Crampin, S. & Zatsepin, S. V., 1996. Opportunities for earthquake
forecasting (Abstract), Earthquake Research in Turkey: State of the Art,
30th Sept.-5th Oct., 1996, Ankara.
5. Crampin, S. & Zatsepin, S. V., 1996. The possibility of forecasting
earthquakes (Abstract), Assessment of schemes for earthquake prediction,
Geological Society, 7-8th Nov., 1996, London.
3 PROPOSED PROJECTS
ADVANCES IN SHEAR-WAVE SPLITTING: RELEVANT OPTIONS
3.1 Summary of objectives of Subproject 3:
- Systematic examination of shear-wave splitting in Iceland, particularly
around SISZ.
- Attempt to gain better understanding of the behaviour of shear-wave
splitting with respect to earthquake preparation zones.
- Attempt to monitor stress changes before earthquakes.
- Attempt to incorporate changes in shear-wave splitting into a wider
earthquake prediction scheme in Iceland.
- Attempt to automate measuring shear-wave splitting.
3.2 Extended objectives (consistent with PRENLAB Objectives, above)
(A) Since large earthquakes are not frequent, to get results quickly it is
necessary to calibrate shear-wave splitting with other possible variations
(see items marked A, below).
(B) Quantification is possible by collaboration with other investigations.
There are a number of possible options involving collaboration and
interaction with other PRENLAB contractors.
(C) Further developments of the methodology and discussions with other EC
Contractors during the European Seismological Commission in Reykjavik have
identified important additional areas where shear waves can monitor the
rockmass and lead to better understanding of the behaviour of stress
beneath Iceland.
(D) An essential requirement for several projects is additional SIL-type
stations within the shear-wave window of the same earthquakes. That is
clusters of at least three SIL-type station within 8 km, say, of each
other above swarms of small earthquakes.
(E) The recent behaviour at SAU confirms that shear-wave splitting is
sensitive to stress changes in the crust from whatever source, earthquakes
or eruptions, and suggests that analysis of shear-waves can monitor stress
changes beneath Iceland.
3.3 Specific New Sub-Projects (with objectives indicated)
1) Investigate why observations of time-delays between split shear waves in
Iceland appear to be approximately twice those observed elsewhere (almost
certainly due to high subsurface temperatures and/or high pore-fluid
pressures) (C,D,E).
2) Establish shear-wave splitting map of all seismic stations in the whole
of Iceland.
i) Search map for possible changes in shear-wave polarizations which would
monitor stress orientations that are thought to accompany strain waves
(C,D,E).
ii) Search map for possible orthogonal changes in shear-wave polarizations
which may indicate high fluid pressures (C,D,E).
iii) Search for other temporal changes in time-delays which may indicate
precursory sequences before earthquakes or eruptions, or may indicate
passage of strain waves (C,D,E).
3) Calibrate techniques and crustal behaviour where known changes occur.
i) Monitor and model extraction of high-temperature geothermal water at
pumping stations (A,B,D).
ii) Monitor and model extraction of low-temperature geothermal water at
pumping stations. Thought unlikely to show first-order effects, but a
necessary investigation (A,B,D).
iii) Monitor and model cold-water injection in a well near Akureyri with a
mini SIL network (A,B,D).
iv) Collaborate with Subproject 4 (borehole monitoring) by correlating
changes in boreholes with changes in shear-wave splitting at neighbouring
seismic stations (A,B,D).
v) Collaborate with Subproject 5A (SAR interferometry) in calibrating
subsidence near high-temperature extraction (A,B,D).
vi) Where appropriate digital three-component records exist, monitor, and
model the rock before and after the suggested dyke injections such as the
1987, M=5.8, Vatnafjöll earthquake, and other similar earthquakes (A,B).
vii) Monitor and model rock behaviour during continuing swarm activity for
example at Hengill and Torfajökull, believed to be caused by magma
injection and water cooling, respectively. Note that monitoring stress
before the Vatnajökull eruption demonstrates the potential value of such
investigations (A,B,D,E).
4) Develop techniques to make shear-wave splitting results available in
routine analysis of other seismic parameters involving shear waves, such as
fault-plane-mechanisms, earthquake locations, etc.
4 REPORT PAPER
Preliminary Version
MONITORING STRESS DURING MAGMA INJECTION
Stuart Crampin*, Helen J. Rowlands* and Ragnar Stefánsson^
*Department of Geology and Geophysics, University of Edinburgh, Grant
Institute, West Mains Road, Edinburgh EH9 3JW, Scotland UK.
^Icelandic Meteorological Office, Bustadavegur 9, 150 Reykjavik, Iceland
Observations of shear-wave splitting at a seismic station, about 160 km WSW
of the recent eruption beneath the Vatnajökull ice cap in Iceland, show
systematic changes in waveforms for several months before the eruption.
The changes show patterns of behaviour that suggest increasing stress
similar to patterns seen before the few larger earthquakes where
appropriate source-to-receiver geometry exists. Consequently, these
changes can be interpreted as monitoring increasing stress as magma is
injected into the base of the crust prior to the eruption.
FIG. 1 Schematic illustration of shear-wave splitting in stress-aligned
fluid-filled intergranular microcracks.
Shear-wave splitting, where the transverse vibrations of seismic shear-
waves split into two approximately orthogonal polarizations aligned with
the horizontal stress directions, is widely observed in almost all crustal
rocks (1,2) (Fig. 1). Such splitting, analogous to optical bi-refringence
in crystals, is diagnostic of some form of effective elastic anisotropy.
In the crust, the anisotropy is caused by stress-aligned fluid-filled
intergranular microcracks and intergranular pore-space that exist in almost
all rocks of all porosities (2,3). There are only a few well-understood
exceptions. Figure 2 shows the expected behaviour of crack distributions
for increases of crack density and increases of aspect-ratio (crack
"fatness"). Changes in crack density modify the size of the time-delay
between the split shear-waves without modifying the overall pattern
(compare Figs. 2a and 2b where the sections show different amplitudes),
whereas changes of aspect-ratio change the two-dimensional pattern of time-
delays without changing the maximum amplitude (compare the solid areas in
the +/-(15deg - 35deg) bands of directions in Figs. 2a and 2c).
FIG. 2 Equal-area polar projections about the vertical of the
characteristics of shear-wave splitting in distributions of parallel
vertical cracks striking east-west for three crack specifications in a
granite matrix (density = 2.65 gm/cm^3, Vp = 6.1, Vs = 3.5 km/sec): (a)
crack density e = 0.02, aspect-ratio g = 0.02; (b) e = 0.04, g = 0.02;
and (c) e = 0.02, g = 0.06. Left-hand roundels are the polarizations of
the faster split shear-waves; right-hand roundels are contoured time-
delays (with north-south sections) normalized to ms/km, where directions of
ray paths between +/-(15deg - 35deg) to the face of the cracks are solid.
Roundels show the shear-wave window marked at 35deg.
Changes in the pattern of time-delays have been observed in the +/-(15deg -
35deg) bands before a few large earthquakes, where suitable three-component
digital seismic networks record shear-waves from swarms of small
earthquakes near to the locations of larger earthquakes (4,5,6). This
behaviour was interpreted speculatively as crack aspect-ratios increasing
as stress accumulated before the earthquake, and decreasing to the initial
level when stress was relieved at the time of the earthquake (4,7). The
evolution of such microcracked rock under changing conditions can be
evaluated with an (a)nisotropic (p)oro-(e)lasticity (APE) model, where the
deformation is driven by microscale fluid-migration along pressure
gradients between neighbouring intergranular voids at different
orientations to the stress field (8,9). APE-modelling matches the behaviour
of shear-wave splitting in most circumstances (9,10,11,12) and appears to
be a good first-order numerical approximation to the equation of state for
microscale rock deformation. In particular, APE-modelling confirms that
increasing stress does increase crack aspect-ratios, and the observed
behaviour of shear waves before earthquakes can be modelled by modest
increases in the differential horizontal stress (11,12).
FIG. 3 Map showing locations of seismic station SAU, volcanoes Bárdarbunga
and Grímsvötn, and epicentres of earthquakes for 137 days from 1st May,
1996. Triangles are seismic stations, and ice fields are shaded.
As a part of a European Commission Project "Earthquake-prediction research
in a natural laboratory" (PRENLAB), shear-wave splitting is being
systematically monitored beneath seismic stations in Iceland. The first
station to be examined, SAU, (Figure 3) was located in the most seismically
active region of Iceland, the South Iceland Seismic Zone (SISZ), where
frequent small earthquakes in the shear-wave window beneath the station
provide sources of shear waves. (The shear-wave window is the cone of
directions at the surface with effective angle of incidence less than about
sin-1(Vs/Vp)~35deg, where Vp and Vs are the velocities of the P- and S-
wave, respectively. Shear-waves recorded within this window are not
distorted by free-surface interactions (13).)
FIG. 4 Three seconds of three-component seismograms of earthquake (96-07-
10 07-28-30.4) at SAU. Traces are, from the top, NS, EW, Vertical, and
horizontals rotated to faster (N225degE) and slower (N315degE) split shear-
waves directions showing a 0.09 second time difference.
Figure 4 shows an example of shear-wave splitting recorded at SAU. The
polarizations and time-delays between the split shear-waves are obtained by
rotating the horizontal components into the faster and slower polarization
directions.
FIG. 5 Variation of shear-wave splitting at SAU showing angular differences
for 137 days from 1st May 1996. Polar equal-area maps of (a)
polarizations, with rose diagram indicating average direction, and (b)
normalized time-delays. (c) Variation with time of time-delays normalized
to ms/km. Open circles (and dashed line) are time-delays for ray paths
within the bands with incidence +/-(0deg - 15deg) to the crack face (which
are sensitive to crack density), and solid circles (and solid line) are
time-delays for ray paths within the bands with incidence +/-(15deg -35deg)
to the crack face (which are sensitive to aspect-ratio). The arrow marks
the onset of the eruption on 30th September 1996.
Figure 5 shows the variation in time-delays at SAU for 137 days (the
currently available data) from 1st May to 14th September, 1996. The
polarizations of the faster split shear-wave are approximately parallel to
N44degE which fault-plane-mechanisms (14) show is the direction of maximum
differential horizontal stress in the SISZ. Such parallelism is typical of
the behaviour above small earthquakes which has been observed in many
places around the world. The analysis is preliminary and unconfirmed and
there is a large scatter, but the trend of the time-delays in Fig. 5c
matches exactly the pattern of behaviour expected for increasing stress.
Time-delays within +/- 15deg of the face of the average microcrack (open
circles) show a large scatter but no overall variation, suggesting that the
crack density does not significantly vary; whereas time-delays between +/-
(15deg - 35deg) to the average face (solid circles) increase substantially
over the 137 days suggesting a marked increase in aspect-ratio indicative
of increasing stress. (Note that scatter is inherent as the rectangular
bands are sampling a curvilinear surface.)
Such behaviour has been observed (and modelled) for the build-up of stress
before larger earthquakes (4,5,6,7). The first 100 days of the data were
shown at an internal PRENLAB meeting on the 10th September where it was
noted that the trend in the data might indicate an imminent larger
earthquake. However, it seemed unlikely that the first data analyzed at
the first station selected would display significant signals, and the trend
was not taken seriously.
The eruption began at about 22.00 on 30th September, 1996, marked by a
continuous tremor, typical of eruptions, recorded at a seismic station on
the Grímsvötn volcano. A public warning (15) of an eruption had been
issued three hours prior to the eruption, based on intense seismic activity
beginning on 29th September, 1996 around Bárdarbunga volcano some 20 km NNW
of the eruption site. An eruption on the northern flank of Grímsvötn was
confirmed by air on 1st October 1996.
The seismic activity near Bárdarbunga began with a magnitude M ~ 5
earthquake at 10.48 am on 29th September, 1996, which had a thrust
mechanism indicating NW-SE compressional axes in the direction of the
impending eruption. The earthquakes thereafter had strike-slip or thrust
faulting with the same direction of compression (16). Within 36 hours, the
seismicity migrated from Bárdarbunga towards the eruption site and
concentrated there until the eruption. The authors suggest that the
injection of magma into the uppermost crust started at the time of the
seismic activity 36 hours before the eruption. The increased compression
stress near Bárdarbunga was released by thrust faulting, which opened
pathways for magma to escape to the surface. The process of magma
injection in the ductile lower-crust began earlier and did not excite
earthquakes. Note that the effective trebling of time-delays in the +/-
(15deg - 35deg) band is comparatively fast from about 5 to 15 ms/km in
137 days from a probable minimum of about 2 ms/km. The increase appears
approximately linear, and extrapolation suggests that the increase began
(magma injection began) no more than about 200 days before the eruption.
This may be consistent with other seismic evidence. There was a burst of
seismic activity near Grímsvötn, 9-13th February, 1996, with short-term
swarm activity similar to typical volcanic tremors.
The changes in time-delays observed before earthquakes are believed to be
monitoring the accumulation of stress before a large earthquake can occur,
and since the strength of crustal rock is weak the stress accumulation
takes place over an enormous rock-volume, possibly millions, even many tens
of millions, of cubic kilometres. Consequently, the increasing stress
implied by the changes in shear-wave splitting at SAU may be attributed to
the increasing pressure of magma injection some 160 km from the seismic
station.
The analysis reported here is preliminary. Several features of the
observations require further investigation. The normalized time-delays of
up to 20 ms/km are more than twice those customarily observed in intact
rock elsewhere: a range of about 2 to 8 ms/km is typically observed below
1 km depth (2). Related phenomena which can produce large time-delays are
high temperatures and high pressures (4). In particular, high pore-fluid
pressures can lead to substantial increases in time-delays (17). In
Iceland, heat flow is large, and supercritical steam is sometimes exploited
in geothermal reservoirs, so that high pore-fluid pressures may also be
expected. Another, feature that requires investigation is the depth range
of the aligned microcracks. Shear-wave splitting above small earthquakes
does not provide information about the range of depths were the splitting
occurs. Typically aligned microcracks distributed uniformly along the path
between source and recorder provide the most consistent interpretations
(2). In Iceland, where rocks are much younger and hotter than is usual
elsewhere, the depth range of stress-aligned intergranular microcracks may
well be different and could be crucial for detailed interpretation of these
results.
Shear-wave splitting is known to be sensitive to changes in the
intergranular crack geometry (10,11), and microcrack geometry is known to
be sensitive to changes in fluid-rock conditions (2,8,9). The observations
reported here appear to confirm that changes of stress can be monitored by
analyzing changes in shear-wave splitting, and that in the case reported
here the increase of stress was caused by magma activity beneath
Vatnajökull. The response of shear-waves to magma activity began several
months before the eruption, and had the potential for indicating an
impending major tectonic event.
(1531 words, including first paragraph)
1. Crampin, S. Nature, 328, 491-496 (1987).
2. Crampin, S. Geophys. J. Int. 118, 428-438 (1994).
3. Fyfe, W. S., Price, N. J. & Thompson, A. B. Fluids in the Earth's
crust: Developments in Geochemistry 1, Elsevier Sci. Publ. Co. Inc.
(1978).
4. Crampin, S., Booth, D. C., Evans, R., Peacock, S. & Fletcher, J. B. J.
Geophys. Res. 95, 11,197-11,212 (1990);
5. Booth, D. C., Crampin, S., Lovell, J. H. & Chiu, J.-M. J. Geophys.
Res. 95, 11,151-11,174 (1990).
6. Liu, Y., Booth, D. C., Crampin, S., Evans, R. & Leary, P. Can. J.
Expl. Geophys. 29, 380-390 (1993).
7. Crampin, S., Booth, D. C., Evans, R., Peacock, S. & Fletcher, J. B. J.
Geophys. Res. 96, 6403-6414 (1991).
8. Zatsepin, S. V. & Crampin, S. 65th Ann. Int. SEG Meeting, Houston,
Expanded Abstracts, 918-921 (1995):
9. Zatsepin, S. V. & Crampin, S., Stress-induced coupling between
anisotropic permeability and shear-wave splitting. 58th Conf. EAGE,
Amsterdam, Extended Abstracts, C030 (1966).
10. Crampin, S. & Zatsepin, S. V. 65th Ann. Int. SEG Meeting, Houston,
Expanded Abstracts, 199-202 (1995).
11. Crampin, S. & Zatsepin, S. V. Geophys. J. Int. in press (1996).
12. Crampin, S. & Zatsepin, S. V. in Seismology in Europe, Europ. Seism.
Comm., 318-323 (1996).
13. Booth, D. C. & Crampin, S. Geophys. J. R. astr. Soc. 83, 31-45 (1985).
14. Stefánsson, R., Bödvarsson, R., Slunga, R., Einarsson, P.,
Jakobsdöttir, S., Bungum, H., Gregersen, S., Havskov, J., Hjelme, J. &
Korhonen, H. Bull. Seism. Soc., Am. 83, 696-716 (1993).
15. Einarsson, P., Brandsdóttir, B., Gudmundsson, M. T. & Björnsson, H. A
chronological account of the October eruption under the Vatnajökull ice
cap, Sci. Inst., Univ. Iceland, pp.9 (1996).
16. Stefánsson, R., Gudmundsson, G. B., Rögnvaldsson, S. T., Report on
Vatnajökull eruption, Icelandic Met. Off. (1996).
17. Crampin, S., Zatsepin, S. V., Slater, C. & Brodov, L. 58th Conf.
EAGE, Amsterdam, Extended Abstracts 1, X038 (1996).