Shear-wave splitting inacriticalcrust III. Preliminary report of by elf15161


									                                      Journal of Applied Geophysics 54 (2003) 265 – 277

                     Shear-wave splitting in a critical crust:
            III. Preliminary report of multi-variable measurements
                               in active tectonics
                             Stuart Crampina,*, Sebastien Chastin, Yuan Gao b
               Shear-Wave Analysis Group, Department of Geology & Geophysics, Grant Institute, University of Edinburgh,
                                        West Mains Road, Edinburgh EH9 3JW, Scotland, UK


    This is a preliminary report on two sets of recent observations from a region of active tectonics that provide comparatively
direct evidence for the critical state of the fluid-saturated microcracked crust. The first data set from crosshole seismics in a
controlled source stress-monitoring site (SMS) shows that the crust of the Earth is highly compliant and responds to low-level
changes of tectonic stress at substantial distances. The second set of data from earthquake seismograms shows that the
                      ´ ´
seismically active Husavık – Flatey Fault plane is pervaded by critically high pore-fluid pressures, which cause 90j flips in the
polarisations of seismic shear waves. We suggest that both sets of observations confirm previous hypotheses for a compliant
crack-critical (CCC) crust. This is a new understanding of low-level pre-fracturing deformation that has fundamental
implications for a range of applications in solid earth geophysics. These applications range from monitoring hydrocarbon
production with time-lapse seismics to monitoring tectonic stress in in situ rock and stress-forecasting the times and magnitudes
of impending large earthquakes.
D 2003 Published by Elsevier B.V.

Keywords: 90j flips; Compliant crack-critical (CCC) crust; Crack-induced anisotropy; Shear-wave splitting; SMS; Stress-monitoring site

                                                                        1. Introduction

                                                                           ‘‘Make a better instrument or measure in a place
                                                                           where no one else has been and a great discovery
    * Corresponding author. Tel.: +44-131-650-4908; fax: +44-131-          will come your way.’’ Press (1979).
    E-mail addresses: (S. Crampin),
                                                                            This is the third International Workshop on Seismic (S. Chastin),, (Y. Gao).                                            Anisotropy (IWSA) where we have suggested that the
    URL:                    fluid-saturated grain-boundary cracks and low aspect-
      Also at Edinburgh Anisotropy Project, British Geological          ratio pores in the Earth’s crust are so closely spaced
Survey, Murchison House, West Mains Road, Edinburgh EH9 3LA,            that they form critical systems verging on (fracture)
Scotland, UK.
    b                                                                   criticality, breakdown, and deterministic chaos. We
      Also at Centre for Analysis and Prediction, China Seismo-
logical Bureau, Beijing 100036, China.                                  call this a compliant crack-critical (CCC) crust. Since

0926-9851/$ - see front matter D 2003 Published by Elsevier B.V.
266                           S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277

in situ rocks at depth are subject to high temperatures            stress on shear-wave splitting using small earthquakes
and pressures and are essentially inaccessible, proving            as the shear-wave source allowed the time and magni-
criticality in the response of in situ rock to small               tude of the impending earthquake to be estimated
changes of stress has proved difficult, and most of the            successfully. Note that magnitudes M refer to the
evidence is indirect. At the first of these three work-            Icelandic Bulletin magnitudes approximately equiva-
shops, 8IWSA, we reported (Crampin, 1998) the large                lent to mb.
range of phenomena (currently increased to more than                   Here, we present two sets of observations from
20) where the anisotropic poro-elasticity (APE) model              Northern Iceland. The detailed analyses will be
for the evolution of fluid-saturated microcracked rock             reported elsewhere. We present measurements of
matches, at least approximately, the behaviour of                  variations of seismic velocities at a stress-monitoring
various configurations of cracks, stress, and shear-               site (SMS) showing high sensitivity of P-, SV- and
wave splitting (Zatsepin and Crampin, 1997; Crampin                SH-wave velocities, and shear-wave splitting that
and Zatsepin, 1997; Crampin, 2000a). APE models                    correlate with several other geophysical observations
the evolution of fluid-saturated grain-boundary cracks             and with distant low-level seismicity. The velocities
and low aspect-ratio pores under changing conditions               were measured horizontally at approximately 500 m
where the driving mechanism is fluid migration by                  depth between two wells 315 m apart. Since the
flow or dispersion between neighbouring microcracks                direction was parallel to a major strike– slip fault,
at different orientations to the stress field. In general,         the sagittal plane was a symmetry plane so that shear
the detailed response of in situ rocks is so poorly                waves were split into SV- and SH-wave orientations.
quantified at depth that the match of APE modelling                    The second data set shows observations of 90j flips
cannot be adequately tested. Nevertheless, since the               in the polarisations of shear-wave splitting above small
underlying assumption of APE is that the fluid-satu-               earthquakes at seismic stations close to a major fault
rated cracks in in situ rock form a CCC system                     plane, which we suggest indicates the critical pressures
verging on fracture criticality and failure (Crampin               associated with all seismically active faults. These 90j
and Zatsepin, 1997), even the approximate match of                 flips have now been analysed and modelled (Crampin
APE to a large range of phenomena is confirmation,                 et al., 2002) and support the inferences in this paper.
albeit indirect, that the cracks in the crust are a CCC            Both sets of observations show effects that are critically
system.                                                            dependent on the existence of a CCC crust.
    Crampin and Chastin (2001) at 9IWSA reported the
successful modelling (in effect, prediction with hind-
sight) of the response of the reservoir to two CO2                 2. The stress-monitoring site experiment
injections by Angerer et al. (2000, 2002). One of the
pressures was high enough to cause 90j flips in shear-                 The European Commission-funded SMSITES
wave polarisations in the injected reservoir, where the            Project is developing a stress-monitoring site (Cram-
faster split shear wave flipped from approximately                                                                ´ ´
                                                                   pin, 2001) on the onshore extension of the Husavık–
parallel to approximately orthogonal to the direction              Flatey Fault (HFF), which is a transform fault in the
of maximum horizontal stress. In both injections, the                 ¨
                                                                   Tjornes Fracture Zone of the Mid-Atlantic Ridge in
match of modelled to observed shear-wave splitting for             Northern Iceland. The aim was to transmit shear
the evolution of the fluid-saturated cracks was almost             waves along specific stress-sensitive directions in
exact, and this study is the best calibration of in situ           order to try to identify the small changes in micro-
APE modelling to date. Again, the necessary underly-               crack geometry, which APE shows are the most
ing assumption is a CCC crust. Pressures that are high             immediate effects of accumulating stress, before fail-
enough to cause 90j flips will be referred to as critical          ure by fracturing occurs. We shall call such low-level
pressures.                                                         modifications pre-fracturing deformation. The range
    Crampin and Chastin (2001) also reported the first             of directions most sensitive to small changes of stress,
successful stress forecast of the time and magnitude of            known as Band-1, is the double-leafed solid angle
an M = 5 earthquake in SW Iceland by Crampin et al.                with ray paths 15– 45j on either side of the average
(1999). Monitoring the effects of increasing tectonic              crack plane (Crampin, 1999). It can show theoretically
S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277                                  267

                                     that the time delays of shear waves in these directions
                                     respond to changes in crack aspect ratios, which are
                                     the most sensitive crack parameter to small changes of
                                     stress (Crampin, 1999). This is in contrast to Band-2,
                                     ray paths within 15j of the average crack plane, where
                                     shear-wave time delays are sensitive to (principally)
                                     changes in crack density. Changes in crack density
                                     generally occur only for comparatively large changes
                                     of stress (Crampin, 1999).
                                        The borehole source is the downhole orbital vibra-
                                     tor (DOV) of Geospace Engineering Resources Inter-
                                     national. The DOV sweeps an eccentric cam in
                                     clockwise and counterclockwise directions exerting
                                     a rotating radial force on the borehole wall (Daley and
                                     Cox, 2001). Signals from the DOV may be processed
                                     to yield shear-wave radiation equivalent to a point
                                     force in specified orientation relative to an on-board
                                     gyroscope. An earlier version of this source (the
                                     Conoco orbital vibrator or COV) produced signals
                                     that were analysed for shear-wave splitting in a
                                     reverse VSP (Liu et al., 1993).
                                        Fig. 1 shows geophysical observations in and
                                     around the SMSITES location. Fig. 2 shows shear-
                                     wave polarisations at seismic stations in Iceland
                                     1996– 2000. Fig. 3 shows shear-wave polarisations
                                     at several seismic stations around SMSITES for the
                                     year 2001 including polarisations at three new sta-
                                     tions, BRE, FLA, and HED.

                                     3. Observations of sensitivity

                                        As part of the setting-up procedure for SMSITES,
                                     we activated the source and receiver recording system
                                     by repeated sweeps of the DOV every 12 –20 s and
                                     stacking every 100 sweeps for 24 h a day for 13 days
                                     (11 –24 August 2001) with only minor interruptions.
                                     Both DOV and receivers were at about 500 m depth in
                                     wells 315 m apart. The DOV had a peak response at

                                     Fig. 1. Variations at the SMSITES SMS from 8 to 24 August 2001:
                                     (a) P-wave traveltimes in milliseconds; (b) traveltimes of SV-waves
                                     (green crosses) and SH-waves (blue crosses) in milliseconds; (c)
                                     time delay (SV – SH) in milliseconds; (f) GPS displacements around
                                       ´ ´
                                     Husavık in millimeters, North – South (blue circles) and East – West
                                     (red crosses); (e) pressure at 33 m depth in water well on Flatey
                                     Island in bars showing ocean tides and anomalous f 1 m drop in
                                     water level; (d) 12 hourly histogram of seismicity within 100 km of
                                                    ´ ´
                                     SMSITES, Husavık.
268                                S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277

Fig. 2. Equal-area rose diagrams (green petals) of shear-wave polarisations in the shear-wave window above small earthquakes recorded by the
seismic network in Iceland for 5 years (1996 – 2000) superimposed on equal-area polar projections out to 45j of the individual polarisations.
The white areas are ice caps (after Crampin et al., 2002).

f 250 Hz and is highly repeatable. The stacked                            days. The simple relaxation curves of both shear
signals were found to have stability in travel times                      waves suggests that shear waves are propagating
with a resolution of at least F 20 As. Well logs and                      along comparatively simple ray paths without too
cores indicate that the ray paths are near the top of a                   many complications. The difference between the two
200-m-thick layer of sandstone sandwiched between                         shear waves, the shear-wave splitting, also shows a
heavily fractured basalts.                                                10% variation over about 6 days.
    We had expected to see possible variations due to                        It was found that these phenomena coincide in time
source instabilities and possibly the effects of earth or                 with a series of some 106 small earthquakes (M V 2.8)
ocean tides. What we observed (Fig. 1a) was an                                       ´
                                                                          on the Grımsey Lineament transform fault some 70
abrupt 5-ms increase in P-wave travel times and then                      km NNW of SMSITES. The total energy released by
a linear decrease over 10 days and (Fig. 1b) classic S-                   these earthquakes is approximately equivalent to one
shaped relaxation curves in travel times of both                          M = 4 earthquake. This would be a comparatively
vertically and horizontally polarised shear waves with                    small earthquake with an expected fault slip of milli-
amplitudes of about 2 ms and durations of about 4                         meters on a fault plane possible 100 m in diameter.
                                   S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277                                      269

Fig. 3. Rose diagrams of shear-wave polarisations in the boxed area in the figure: years 1996 – 2000 (green petals, as in Fig. 2), and year 2001
(red petals). BRE, FLA, and HED are new seismic stations installed in January 2001 for the SMSITES Project. The SMSITES site is located
close to seismic station HED (after Crampin et al., 2002).

This is comparatively small-scale activity with a small                     3.1. Seismic effects
source zone, which conventional geophysics suggests
would have significant effects only in the immediate                       3.1.1. Variations of P-wave travel times
vicinity of the source. Thus, the sensitivity of the                          Fig. 1a shows variations in P-wave travel times.
seismic measurements to this minor stress release at                       Recording began on 11 August. The travel times are
considerable distance is remarkable and we believe                         scattered but suggest an increase in travel time of
has not previously been observed.                                          about 4 ms. Immediately following the highest value,
           ´ ´
   The Husavık – Flatey Fault has been subject to                          the travel times begin an almost linear decrease from
large earthquakes in the past, and there have been                         11 to 21 August, with possibly a small break of slope
several geophysical investigations by the Icelandic                        on August 16 coinciding with one of the gaps in
Meteorological Office and others, as well as by the                        recording. The two gaps in recording were when the
SMSITES project. This means that the area is com-                          DOV was thought (mistakenly) to be overheating and
paratively well instrumented. Fig. 1 shows seven                           the tool was allowed to cool. The amplitude of the
variations correlating with the distant low-level swarm                    decrease is about 5 ms over about 10 days. The
activity.                                                                  amplitudes and durations of the various variations
270                                   S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277

are listed in Table 1. Since the seismic measurements                      August during the S-shaped relaxation curves in Fig.
at SMSITES began on 11 August and the seismic                              1b. At the end of the increase, the difference levels off
activity on Grımsey Lineament began on 10 August,                          on 20 August to about 2.15 ms. The amplitude of this
there was no direct indication of velocity variations                      increase is about 0.2 ms over about 5 days (Fig. 1c).
before the seismic activity had started.                                      Note that this 10% change (Table 1) in the time
                                                                           delays between the split shear waves is the largest
3.1.2. Variations in SV-wave travel times                                  percentage seismic variation during the stress-induced
   The SV-wave travel times show irregularities dur-                       changes in Fig. 1. This is a further demonstration that
ing the first 3 days of recording (11– 13 August), but                     the time delay in shear-wave splitting is a highly
are constant during 14 August, and from 15 to 19                           sensitive parameter (Crampin, 1999).
August follow an S-shaped relaxation curve before
levelling off on 20 August with possibly some indi-                        3.1.5. Seismicity
cation of a further gradual decrease. The amplitude of                        The histogram of earthquakes in 100 km2 around
the S-shaped decrease is about 2 ms over about 4 days                      the SMSITES Site from 8 to 24 August 2001 shows a
(green crosses, Fig. 1b).                                                  2.5-day swarm of 106 earthquakes (10 –12 August)
                                                                           located on a 10-km segment of the Grımsey Linea-
3.1.3. Variations in SH-wave travel times                                  ment approximately 70 km NNW of SMSITES and 10
   The SH-wave travel times again show irregularities                                                       ´
                                                                           km NNE of the Island of Grımsey. The remaining
during the first 3 days of recording, although the initial                                                        ´
                                                                           seismicity is distributed along the Grımsey Lineament
variations are somewhat different in detail. They show                     with a few earthquakes on the HHF, but about 50% of
a similar S-shaped relaxation curve to that of the SV-                     the activity is still from the same 10 km segment of
waves, but are about 2 ms earlier. Similar to the SV-                             ´
                                                                           the Grımsey Lineament (Fig. 1f).
waves, the amplitude of the S-shaped decrease is about                        The largest earthquake is M = 2.8, and the total
2 ms over about 4 days (blue crosses, Fig. 1b).                            energy release of the initial burst of activity is ap-
                                                                           proximately equivalent to the energy released from
3.1.4. Variations in SV –SH anisotropy                                     one M = 4 event. Note that the area in Fig. 3, showing
    The difference between the travel times of SV and                      seismicity in the year 2001, currently has compara-
SH fluctuates during the first 3 days (when the SV-                        tively low-level seismic activity, typically of about
and SH-wave travel times themselves show irregular-                        three or four events per day, so that the first and last 2
ities), but shows an irregular increase from 15 to 19                      or 3 days in Fig. 1f are close to the background level.

Table 1
Summary of associated variations
Nature of phenomenon                                 Figure             Approximate size             Approximate         Seismic
                                                     number             or amplitude                 duration (days)     variation (%)
Linear decrease in P-wave traveltimes                1a                 5 ms                         10                   6
S-shaped decrease in SH-wave                         1b (blue)          2 ms                          4                   1
S-shaped decrease in SV-wave                         1b (green)         2 ms                           4                  1
Variations in SV – SH times                          1c                 0.2 ms                         6                 10
East – West GPS deformationa                         1d (red)           3 and 4 mm                    4 and 9             –
North – South GPS deformation                        1d (blue)          7 mm                         11                   –
Water-level decrease at Flatey                       1e                 1m                            5                   –
Seismicity: initial burst of activity                1f                 106 events, M V 2.8           2.5                 –
  on Grımsey Lineament 10 – 12
  August 2001, 10 km NNE of Grımsey   ´
          EW GPS deformation takes place in two phases (hence, two amplitudes and two durations).
                              S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277                      271

3.2. Changes in strain from Global Positioning                     negligible rainfall during the recording period in
System (GPS) displacements                                         August 2001 (Fig. 1e).

3.2.1. Variations in East –West GPS displacements
    GPS displacements from stations either side of the             4. Discussion of sensitivity to small changes of
HFF in an East –West direction show a change in                    stress
strain of about 4 mm in two phases (red crosses, Fig.
1d). Initially, a 4-day 3-mm pulse returning to the                   Fig. 1 displays what is, in effect, a time-lapse
initial level which is followed by a 9-day increase,               survey of eight variables. A summary of the amplitude
which then levels off to an offset of 4 mm. The initial            and duration of these various changes is listed in Table
rise and fall is associated with the larger North – South          1 together with the percentage change in seismic
strain, below. The 9-day increase in strain marks the              variations. There are several remarkable features.
return to the customary dextral movement of the HFF.               There is a wide variation in the duration of the
                                                                   changes. For example, the linear decrease in P-wave
3.2.2. Variations in North – South GPS displacements               travel times is over about 10 days, whereas the S-
   GPS displacements across the fault in a North –                 shaped decreases in both SH- and SV-waves are over
South direction from stations either side of the HFF               about 4 days. One initially might expect that seismic
show an abrupt 7-mm increase followed by an ap-                    P- and S-waves propagating along similar ray paths
proximately exponential decrease in strain over 10                 would respond to similar features of the rock mass.
days relaxing back to approximately zero displace-                 The different durations of their response clearly indi-
ment (blue circles, Fig. 1d).                                      cate that P- and S-waves respond to different phe-
                                                                   nomena and sample different features of rock
3.3. Changes in water level in well on the Island of               deformation. The whole range of eight different phe-
Flatey                                                             nomena is believed to be a unique data set, and their
                                                                   interpretation is likely to place constraints on the
3.3.1. Water pressure variations in well on Flatey                 interpretation of the response of in situ rock to small
   The pressure measurements are at about 33 m                     changes of stress in time-lapse studies.
depth in a water-filled well on the small island of                   The water level in the well on Flatey appears to
Flatey, close to the seismic station FLA, in Fig. 3, and           respond to variations in the tectonic regime. It is well
immediately above the seismicity of the HFF. Pressure              known that changes of water level in wells may be
monitors the water level above the sensor. Since 1 bar             associated with earthquakes (Roeloffs, 1988), al-
(0.1 MPa) is approximately equivalent to a pressure of             though the nature of the association is not fully
10 m of water, the f 0.1 bar decreasing pulse in                   understood. The changes may be precursory, co-seis-
pressure represents a f 1-m drop in water level. The               mic, or post-seismic, and the duration and polarity of
40-cm peak-to-peak sinusoids are tides. Since such                 change may vary widely and can occur at substantial
tides are visible on pressure measurements in water                distances from the seismic activity. Sometimes, the
wells in Iceland only when the wells are near the                  polarity can be associated with fault-plane compres-
coast, the effects appear to be due to oceanic tides.              sions and dilations, but in general, the effects are
The 1-m decrease starts with an initial increase                   thought to be local to the particular well and to be
approximately coinciding with the onset of seismic                 related to interactions of tectonic stress with local
activity in Fig. 1f. The offset decreases gradually over           faults or fractures at different orientations anywhere
4 days and, on the fifth day, rapidly returns to the               near the open section of the well. Although local
background level. Well pressures are recorded contin-              anomalies may have effects, the observations in Fig. 1
uously, and it is worth noting that this abrupt decrease           suggest that it is the whole rock mass that responds to
in pressure is the only significant pulse in 15 months             changes. Clearly, the rock mass responds to pre-
of records. Over the 15 months, there are broad                    seismic and post-seismic disturbances in ways which
increases in November 2000 and November 2001                       have previously not been identified. The 7-mm ex-
due presumably to autumn precipitation. There was                  tension in strain shown by the North –South GPS
272                          S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277

measurements, if interpreted as an increase in aspect             most interesting and potentially most informative
ratio of fluid-saturated microcracks over the 200-m-              feature of this data set, once the compliance of the
thick layer of permeable sandstone, is the correct                CCC crust has been accepted.
order of magnitude to account for the 1-m drop in                     The underlying assumption of APE is that fluid-
water level in the well on Flatey as the increase in              saturated cracks are so closely spaced that they are
microcrack capacity to absorb water (200 Â0.007 =                 critical systems with great sensitivity to small changes
1.4 m).                                                           and with potential for large-scale disruption at fracture
    Another remarkable feature is the near coincidence            criticality. These observations of sensitivity from Ice-
of the various interactions. The initial increase in              land are a direct confirmation that microcracked crust
pressure in water level at Flatey approximately coin-             is a critical system.
cides with the onset of seismicity on the Grımsey ´                   Note that by closely spaced cracks, we mean crack
Lineament some 50 km north of the well. This                      densities between about e c 0.015 and e c 0.045,
suggests that the rock mass responds almost immedi-               where e = Na3/v and N is the number of cracks of
ately to strain changes at f 50 km from Flatey. In                radius a in volume v (Crampin, 1994). This is a very
principle, the GPS measurements could determine                   narrow crack range and is equivalent to crack distri-
delays in response more exactly, but this has not yet             butions where each crack is approximately a crack
been done.                                                        diameter, and a crack radius, respectively, from eight
    A further remarkable feature of the variations in             other cracks in a uniform three-dimensional distribu-
Fig. 1 is the accuracy of the seismic travel-time                 tion of approximately similar-sized cracks. An image
measurements. Due to stacking and the highly repeat-              of such distributions can be found for example in
able DOV source, there are very well-observed var-                Crampin (1994, 1999) and elsewhere. Note also that
iations in seismic velocities at substantial distances            such effects are almost independent of porosity (Zat-
from comparatively small-scale stress release by an               sepin and Crampin, 1997). A similar range of implied
earthquake swarm. These show that the rock mass is                crack densities is found from shear-wave splitting in
extremely compliant and responds to very small                    30% porosity sandstones and in 1% porosity granites.
changes in conditions, even in a regime that is                   This suggests, and an albeit limited number of obser-
principally composed of crystalline basalts, which                vations tends to confirm, that progress towards frac-
might be thought to have minimal compliance. Such                 ture criticality is also largely independent of rock type
compliance is not expected in the brittle upper crust of          and porosity.
conventional geophysics, but such sensitivity is im-                  Thus, it appears that all fluid-saturated rocks, in at
plied by the anisotropic poro-elasticity (APE) model              least the upper half of the crust, contain a universal
of rock deformation. Consequently, the variations in              distribution of microcracks with a very limited range
Fig. 1 provide strong evidence confirming the APE                 of crack densities. Such remarkable universality is
mechanism of deformation in a CCC crust: that the                 characteristic of critical systems verging on criticality
immediate effect of low-level pre-fracturing deforma-             where the behaviour near (fracture) criticality is char-
tion is fluid migration by flow or dispersion along               acterised by the criticality rather than the physics of
pressure gradients between neighbouring grain-                    the subcritical medium (Bruce and Wallace, 1989).
boundary cracks and low aspect-ratio pores at differ-                 It is interesting to question why we see changes at
ent orientations to the stress field. This will particu-          70 km from a comparatively small earthquake, but do
larly effect shear-wave propagation and shear-wave                not see changes from much smaller earthquakes at a
splitting (Crampin, 1999). The particular sensitivity of          distance of only a few kilometers. The increase of
shear waves is supported by the classic S-shaped                  time delays before earthquakes and volcanic eruptions
relaxation curve which suggests that shear waves are              that allowed an earthquake to be stress-forecast are
displaying a fundamental property of the deformed                 thought to monitor the accumulation of stress (Cram-
rock mass: the poro-elastic response of fluid-saturated           pin, 1999; Volti and Crampin, 2003a,b). As they are
microcracks to small changes of stress (Zatsepin and              independent of the eventual source zone, they are not
Crampin, 1997). The interrelationship of P-waves and              earthquake precursors. Earthquakes are only symp-
shear waves is not yet understood and is probably the             toms of an abrupt release of energy. However, pre-
                              S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277                      273

cursory changes of stress are indicated by decreases in            direction of maximum horizontal tectonic stress. This
shear-wave splitting time delays in Band-1 of the                  direction varies from NE to SW in SW Iceland to
shear-wave window 4 days before the M = 5 earth-                   NNE to SSW in North – Central Iceland in the two
quake that was stress-forecast in SW Iceland (Volti                seismic regions where transform zones of the Mid-
and Crampin, 2003b). Similar precursory changes                    Atlantic Ridge run onshore. (The shear-wave window
have also been seen elsewhere (Crampin, 1999) and                  is the cone of arrivals, effectively 45j about the
are found in laboratory experiments (Gao and Cram-                 vertical, within which shear waves are not distorted
pin, 2003). This suggests that the behaviour of the                by S-to-P conversions at the free surface; Booth and
earthquake’s source zone is the driving mechanism for              Crampin, 1985.)
this precursory stress release. The actual earthquakes                 Fig. 2 shows rose diagrams superimposed on polar
themselves merely mark successive possibly minor                   plots out to 45j of the polarisations of faster split
releases of stress during the overall stress-release               shear wave for all earthquakes within the shear-wave
process. Thus, the effects of the individual 106 earth-            window of seismic stations in Iceland for earthquakes
quakes on the Grımsey Lineament, correlating with                  in the 5 years 1996 –2000. Fig. 3 shows similar plots
the observed changes, may be controlled by effects of              for stations in the boxed area in Fig. 2 with green
the overall source zone which has an equivalent                    petals for polarisations for earthquakes for the 5 years
energy release to one M = 4 earthquake at f 70 km                  1996– 2000, as in Fig. 2, and red petals for the year
distance. If we equate the effects of one unit decreases           2001. The red petals at the three new seismic stations
in magnitude (approximately equivalent to a factor of              (BRE, FLA, and HED), installed by the SMSITES
10 decreases in energy) with factors of 10 decreases in            Project in January 2001 close to the HFF, are approx-
distance, we would expect approximately similar                    imately orthogonal to the green petals. Note that the
effects for, say, M = 4 at 70 km, M = 3 at 7 km,                   red petals (2001) at stations SIG and LEI have
M = 2 at 700 m, M = 1 at 70 m, and M = 0 at 7 m.                   different normalisations to the green petals (1996 –
                                   ´ ´
Since small earthquakes near Husavık are likely to                 2000) and refer to very few earthquakes and are
have magnitudes M < 2 and are usually at several                   probably not significant.
kilometers depth, the effects are likely to be smaller                 APE modelling has shown that as pore-pressure
than those for an M = 4 at 70 km and appear to be                  increases, the polarisation of the faster split shear
below the observational limit.                                     waves flips from parallel to perpendicular to the
                                                                   maximum horizontal stress when the pore pressures
                                                                   approach the value of the maximum horizontal stress,
5. 90j flips in shear-wave polarisations                           at what we call critical pressures (Crampin et al.,
                                                                   2002). This is a result of the changes in the three-
    We now report the second data set from the                     dimensional distribution of crack aspect ratios as pore
SMSITES Project: records of shear-wave polarisa-                   pressure approaches values when the rock would
tions at three new seismic stations installed near the             hydraulically fracture. Angerer et al. (2000, 2002)
HFF. The polarisations of the faster split shear wave,             called such phenomena 90j flips. It is well known
propagating at less than 45j to the vertical in crack              that high pore-fluid pressures are needed to relieve
distributions at depth in the crust, are typically aligned         frictional stress on lithostatically clamped faults be-
parallel to the average strike of the distributions of             fore slippage and earthquakes can occur, and we
fluid-saturated microcracks. These are aligned perpen-             interpret the changes in shear-wave polarisations in
dicular to the direction of minimum compressive                    Fig. 3 as indicating 90j flips caused by critical
stress which, below a critical depth (usually between              pressures around the seismically active HFF fault
500 and 1000 m), is horizontal so that the cracks are              plane.
approximately vertical striking parallel to the direction              Such 90j flips have previously been observed in
of maximum horizontal stress. This means that the                  vertical seismic profiles in a critically pressurised
polarisations of shear-wave splitting observed in the              reservoir in the Caucasus Oil Field (Crampin et al.,
shear-wave windows of seismic stations throughout                  1996; Slater, 1997) and in reflection surveys (Angerer
Iceland (Fig. 2) are approximately parallel to the                 et al., 2000, 2002) of a critically pressurised CO2
274                           S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277

injection. These 90j flips have also been observed                 fault to recorders at the surface will not be critically
above small earthquakes immediately above the San                  pressurised and will display the conventional stress-
Andreas Fault in California by Liu et al. (1997), who              parallel shear-wave polarisations orthogonal to the
recognised the significance of 90j flips, and also by              90j flips. This means that the polarisations at the
Peacock et al. (1988) and Crampin et al. (1990, 1991),             surface will display the typical stress-parallel polar-
although at that time, the significance of the 90j flips           isations. However, the time delays will depend on the
had not been established.                                          relative proportion of the ratio critically to normal-
    The three stations, BRE, FLA, and HED in Fig. 3                pressurised segments of ray path. Following an earth-
showing 90j flips, close to the surface break of the               quake and slip on a critically pressurised fault, stress
HFF, were installed by the SMSITES Project in                      will be released, and the geometry of the triaxial stress
January 2001. The stations were sited near to the fault            and pore-fluid pressure will be modified. Consequent-
in anticipation of observing 90j flips in shear-wave               ly, the critical pressures will be redistributed so that
polarisations caused by high pressures prior to a larger           the proportion of the critically to lower pressurised
earthquake on the HFF. However, we now believe that                segments of the ray paths will be changed with
all earthquakes require critical pressures to allow                possibly seriously modified time delays after every
slippage on fault planes. The seismicity in Fig. 3                 earthquake. Thus, these repeated critically pressurised
shows that HFF has a high level of small-scale                     modifications will continue, and the earthquake
seismicity and suggests that high pore-fluid pressures             swarm or foreshock or aftershock will persist as long
are pervasive around probably all seismically active               as the critical-pressurised regions remain. When the
fault zones.                                                       critical fluid pressures disperse, along faults or frac-
                                                                   tures or by other mechanisms, the seismic activity will
                                                                   stop. APE can be used to model these effects and
6. Discussion of 90j flips                                         shows that varying proportions of critically high-
                                                                   pressurised rocks on seismically active fault planes
   The recognition of 90j flips in shear-wave polar-               cause 90j flips and varying time delays that can easily
isations in critically pressurised regions of seismically          explain the observed F 80% scatter.
active fault planes provides an explanation for the
large ( F 80%) scatter invariably observed in measure-
ments of time delays of shear-wave splitting above                 7. Conclusions
small earthquakes (for example, in Volti and Crampin,
2003a,b). Crampin et al. (2002) use APE to model                       The behaviour of both sets of SMSITES observa-
shear-wave polarisations in highly pressurised rocks.              tions can be described and modelled by APE: the
They show that 90j flips occur when pressures are                  observations of sensitivity by implication, and the 90j
sufficiently close to critical pressures. Further, it can          flips directly. A major assumption of the APE model
be argued that all seismically faults require critical             is that the crust is so densely permeated by fluid-
pressures to permit fault slip and earthquakes. Previ-             saturated stress-aligned microcracks that the cracks
ously, 90j flips have been observed directly at the                form critical systems. This means that the success of
surface only at two places on the San Andreas Fault in             APE in modelling a large range of phenomena in-
California (Liu et al., 1997; Peacock et al., 1988) and            cluding the sensitivity and the 90j flips in this paper
here on the HFF. The critical pressurised zone is likely           provides further direct confirmation that the crust of
to persist over much of the ray path to the surface only           the Earth is a CCC system (Crampin and Chastin,
on such major faults, where the flips in polarisations             2001).
are seen at the surface.                                               The important implications and applications of the
   In contrast on smaller faults, critical pressures are           CCC crust for academic and exploration seismics
likely to pervade only the region immediately around               have been discussed elsewhere (Crampin, 1998,
the fault plane which will not extend to the surface.              1999, 2000a,b; Crampin and Chastin, 2001) and will
Close to the fault plane, the shear waves will show                not be repeated here. This section will only refer to the
90j flips, but the remaining path length away from the             new observations of sensitivity and 90j flips.
                            S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277                      275

7.1. Sensitivity to small changes of stress                      respectively). This confirms that shear-wave splitting
                                                                 and shear-wave anisotropy are sensitive diagnostics of
    Large earthquakes release substantial amounts of             the CCC crust, and that the CCC crust is not a stable
stress which has accumulated deep in the crust.                  phenomenon. APE implies that shear-wave splitting is
Previously, it was not known previously how the                  the most sensitive monitor of the temporally and
Earth stores such stress: how a sample of stressed               spatially varying geometry of the distribution of the
rock deep in an earthquake preparation zone in the               stress-aligned fluid-saturated distribution of grain-
crust differs from an unstressed sample. The answer              boundary cracks and pore throats. Roeloffs (1988)
appears to be in the pre-fracturing deformation of               cites eight examples of earthquakes of magnitudes
fluid-saturated grain-boundary cracks and pores as               between M = 5 and M = 6 in China and USA associ-
modelled by APE. APE shows that shear-wave split-                ated with precursory water-level variations at distan-
ting monitors the mechanism for storing and releasing            ces between 100 and 360 km from the eventual
stress and can identify the approach of fracture criti-          epicentre. This paper indicates that those earthquakes
cality and failure by fault slip and earthquakes.                would also be likely to have associated seismic and
    Note that the seismic effects are induced by small           GPS strain variations. In particular, the behaviour of
earthquakes, equivalent to one M = 4, at 70 km dis-              shear-wave splitting appears to be highly sensitive to
tance. These effects are clearly seen at several hun-            the detailed deformation of the CCC crust.
dred times the conventional source dimensions. These                 This ‘‘new geophysics’’ of the CCC crust has
displays exceptional sensitivity of the rock mass to             substantial implications for the whole behaviour of
small disturbances at large distances.                           the solid Earth. In situ rocks are compliant crack-
    The importance of these phenomena for the oil                critical and descend into deterministic chaos whenever
industry is that the effects are extremely sensitive to          fracture criticality approaches and rocks fracture,
small changes of stress and may be modified by                   fault, and earthquakes occur. In particular, the seismic
distant tectonic or other disturbances. This means that          observations reported here imply that current seismic
the possibly subtle effects of moving oil – water fluid          resolution is probably near its limits in resolution.
fronts in time-lapse seismics could be misinterpreted            Increased resolution and sensitivity is likely to mon-
or concealed.                                                    itor the compliance of the CCC crust, and the com-
                                                                 parative simplicity of the conventional brittle elastic
7.2. Significance of 90j flips                                   crust will be lost. This has important implications for
                                                                 hydrocarbon reservoir characterisation, hydrocarbon
   The demonstration in this paper and in Crampin et             recovery, and stress-forecasting earthquakes (Crampin,
al. (2002) that large seismically active faults are              2000b; Crampin and Chastin, 2001). Substantially
pervaded by (critically) high pore-fluid pressures               improved recovery and calculation (even prediction)
confirms that slippage at depth always requires faults           of the response of the reservoir to some recovery
or fractures to be pervaded by critically high pore-             operations may now be possible by being able to
fluid pressures. These lead to 90j flips in the imme-            understand and model the changes (Angerer et al.,
diate vicinity of the fault with the remaining ray path          2000, 2002).
reverting to typical alignments parallel to the maxi-                It is usually the aim of seismic surveys to
mum horizontal stress. This provides an explanation              maximise resolution and measurement accuracy. We
for the otherwise inexplicable F 80% scatter in time             have shown that resolution and accuracy in conven-
delays typically observed above small earthquakes                tional seismics is limited by system criticality. We
(Crampin et al., 2002).                                          suggest that, excluding shear-wave splitting, current
                                                                 techniques have probably reached the limit of con-
7.3. Overall implications                                        ventional interpretation. Unless surveys with similar
                                                                 recording geometry are repeated exactly, it is unlike-
   In Fig. 1 and Table 1 the percentage change in                ly that temporal changes due to critical systems will
shear-wave splitting time delays (10%) is greater than           be identified correctly, and that criticality changes
the change in P- and S-wave travel times (6% and 1%,             will be misinterpreted, possibly as spurious move-
276                               S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277

ment of fluids in time-lapse surveys of hydrocarbon                       voir. 70th Ann. Int. SEG Mtg., Calgary. Expanded Abstracts,
production for example. Accurate repetition of meas-                      vol. 2, pp. 1532 – 1535.
                                                                       Angerer, E., Crampin, S., Li, X.-Y., Davis, T.L., 2002. Processing,
urements of shear-wave splitting are likely to be the                     modelling, and predicting time-lapse effects of over-pressured
most diagnostic indicators of the presence of critical                    fluid-injection in a fractured reservoir. Geophys. J. Int. 149,
systems.                                                                  267 – 280.
   There are two principal conclusions.                                Booth, D.C., Crampin, S., 1985. Shear-wave polarizations on a
                                                                          curved wavefront at an isotropic free-surface. Geophys. J. R.
                                                                          Astron. Soc. 83, 31 – 45.
(1) Substantially improved resolution and improved                     Bruce, A., Wallace, D., 1989. Critical point phenomena: universal
    hydrocarbon recovery are unlikely to come from                        physics at large length scales. In: Davies, P. (Ed.), The New
    improving conventional techniques. Reservoirs                         Physics. Camb. Univ. Press, Cambridge, pp. 236 – 267.
    and crustal rocks are cracked, compliant, and                      Crampin, S., 1994. The fracture criticality of crustal rocks. Geo-
    critical, and the implications of the new geo-                        phys. J. Int. 118, 428 – 438.
                                                                       Crampin, S., 1998. Shear-wave splitting in a critical crust: the next
    physics of the crust must be accepted, and                            step. In: Rasolofosaon, P. (Ed.), Proc. 8th Int. Workshop on
    opportunities exploited, particularly by single- or                   Seismic Anisotropy, Boussens, 1998. Rev. Inst. Franc. Pet.,
    dual-well imaging techniques and using the                            vol. 53, pp. 749 – 763.
    calculability of anisotropic poro-elasticity.                      Crampin, S., 1999. Calculable fluid-rock interactions. J. Geol. Soc.
(2) The science and technology of stress-monitoring                       156, 501 – 514.
                                                                       Crampin, S., 2000a. The potential of shear-wave splitting in a
    sites for monitoring the stress buildup before large                  stress-sensitive compliant crust: a new understanding of
    earthquakes has been confirmed, indicating that                       pre-fracturing deformation from time-lapse studies. 70th
    appropriate stress-monitoring sites should be able                    Ann. Int. SEG Mtg., Calgary. Expanded Abstracts, vol. 2,
    to stress-forecast the times and magnitudes of                        pp. 1520 – 1523.
                                                                       Crampin, S., 2000b. Shear-wave splitting in a critical self-organized
    impending large earthquakes.
                                                                          crust: the New Geophysics. 70th Ann. Int. SEG Mtg., Calgary.
                                                                          Expanded Abstracts, vol. 2, pp. 1544 – 1547.
                                                                       Crampin, S., 2001. Developing stress-monitoring sites using cross-
Acknowledgements                                                          hole seismology to stress-forecast the times and magnitudes of
                                                                          future earthquakes. Tectonophysics 338, 233 – 245.
    This work was partially supported by the European                  Crampin, S., Chastin, S., 2001. Shear-wave splitting in a critical
                                                                          crust: II. Compliant, calculable, controllable fluid – rock interac-
Commission SMSITES Project, Contract Number
                                                                          tions. In: Ikelle, L.T., Gangi, T. (Eds.), Anisotropy 2000: Frac-
EVR1-CT1999-40002, and we thank Gilles Ollier                             tures Converted Waves and Case Studies. Proc. 9th Int.
of the Commission for his great help. YG was                              Workshop on Seismic Anisotropy, Cape Allen 2000. SEG Open
supported partly by China MOST under Contracts                            File Publication, vol. 6, pp. 21 – 48.
2001BA601B02 and NSFC Project 40274011, and                            Crampin, S., Zatsepin, S.V., 1997. Modelling the compliance of
                                                                          crustal rock: II. Response to temporal changes before earth-
partly by the UK Royal Society Fellowship Pro-
                                                                          quakes. Geophys. J. Int. 129, 495 – 506.
gramme. We thank Peter Leary of GERI and John                          Crampin, S., Booth, D.C., Evans, R., Peacock, S., Fletcher, J.B.,
Gregson of IMC for collaborating with the field work                      1990. Changes in shear wave splitting at Anza near the time of
                                  ´ ´
at the stress-monitoring site at Husavık, and Theodora                    the North Palm Springs Earthquake. J. Geophys. Res. 95,
Volti for interpreting many of the shear-wave seismo-                     11197 – 11212.
                                                                       Crampin, S., Booth, D.C., Evans, R., Peacock, S., Fletcher, J.B.,
grams. We particularly thank Hreinn Hjartarson of
                                                                          1991. Comment on ‘‘Quantitative measurements of shear wave
              ´ ´                          ´ ´
Orkuveita Husavıkur who made the Husavık bore-                            polarizations at the Anza Seismic Network, Southern California:
holes available to us and provided local logistics and                    implications for shear wave splitting and earthquake prediction’’
Kristjan Saemundsson of Orkustofnum who drew our                          by R.C. Aster, P.M. Shearer, J. Berger. J. Geophys. Res. 96,
attention to these particular boreholes.                                  6403 – 6414.
                                                                       Crampin, S., Zatsepin, S.V., Slater, C., Brodov, L.Y., 1996. Abnor-
                                                                          mal shear-wave polarizations as indicators of pressures and over
                                                                          pressures. 58th Conf., EAGE, Amsterdam. Extended Abstracts,
References                                                                vol. X038.
                                                                       Crampin, S., Volti, T., Stefansson, R., 1999. A successfully stress-
Angerer, E., Crampin, S., Li, X.-Y., Davis, T.L., 2000. Time-lapse        forecast earthquake. Geophys. J. Int. 138, F1 – F5.
  seismic changes in a CO2 injection process in a fractured reser-                                                                    ´
                                                                       Crampin, S., Volti, T., Chastin, S., Gudmundsson, A., Stefansson,
                                   S. Crampin et al. / Journal of Applied Geophysics 54 (2003) 265–277                                  277

   R., 2002. Indication of high pore-fluid pressures in a seismi-       Press, F., 1979. Response of Bowie Medallist. EOS 60 (30), 540.
   cally-active fault zone. J. Geophys. Int. 151, F1 – F5.              Roeloffs, E.A., 1988. Hydrological precursors to earthquakes: a
Daley, T.M., Cox, D., 2001. Orbital vibrator seismic source for            review. Pure Appl. Geophys. 126, 177 – 209.
   simultaneous P- and S-wave crosswell acquisition. Geophysics         Slater, C.P., 1997. Estimation and modelling of anisotropy in ver-
   66, 1471 – 1480.                                                        tical and walkaway seismic profiles at two North Caucasus oil
Gao, Y., Crampin, S., 2003. Temporal variation of shear-wave split-        fields. PhD dissertation, Univ. of Edinburgh, pp. 226.
   ting in field and laboratory in China. In: Gajewski, D. (Ed.),       Volti, T., Crampin, S., 2003a. A four-year study of shear-wave
   Proc. 10th Int. Workshop on Seismic Anisotropy, Tutzing, 2002.          splitting in Iceland: 1. Background and preliminary analysis.
   J. Appl. Geophys. This issue.                                           In new insights into structural interpretation and modelling.
Liu, E., Crampin, S., Queen, J.H., Rizer, W.D., 1993. Velocity and         In: Nieuwland, D.A. (Ed.), New insights into structural inter-
   attenuation anisotropy caused by microcracks and macrofrac-             pretation and modelling. Geol. Soc. Lond., Spec. Publ. 212,
   tures in multiazimuthal reverse VSPs. Can. J. Expl. Geophys.            117 – 133.
   29, 177 – 188.                                                       Volti, T., Crampin, S., 2003b. A four-year study of shear-wave
Liu, Y., Crampin, S., Main, I., 1997. Shear-wave anisotropy: spatial       splitting in Iceland: 2. Temporal changes before earthquakes
   and temporal variations in time delays at Parkfield, Central            and volcanic eruptions. In: Nieuwland, D.A. (Ed.), New insights
   California. Geophys. J. Int. 130, 771 – 785.                            into structural interpretation and modelling. Geol. Soc. Lond.,
Peacock, S., Crampin, S., Booth, D.C., Fletcher, J.B., 1988. Shear-        Spec. Publ. 212, 135 – 149.
   wave splitting in the Anza seismic gap, Southern California:         Zatsepin, S.V., Crampin, S., 1997. Modelling the compliance of
   temporal variations as possible precursors. J. Geophys. Res.            crustal rock: I. Response of shear-wave splitting to differential
   93, 3339 – 3356.                                                        stress. Geophys. J. Int. 129, 477 – 494.

To top