Vous êtes sur la page 1sur 6

Phvsics of the Earth and Planetary Interiors, 19 (1979) 331- 336

!'',!sevier Scientific Publishing Company, Amsterdam - Printed in the Netherlands

331

IMPROVED SIGNAL DISCRIMINATION IN TECTONOMAGNETISM: DISCOVERY OF A


VOLCANOMAGNETIC EFFECT AT KILAUEA, HAWAII
P.M. DAVIS
1

F.D. STACEY

C.J. ZABLOCKI

and J.V. OLSON

Department of Gmdesy and Geophysics, University of Cambridge, Cambridge (Great Britain)


2

Physics Department, University of Queensland. Brisbane (Australia)


3

U.S. Geological Surl'ey, Denrer, CO 80225 (U.S.A.)

Institute of Earth and Planetary Physics, University of Alberta, Edmonton, Alberta TfiG 2J I (Canada)

(Received June 19, 1978; revised and accepted December 13, 1978)

Davis, P.M., Stacey, F.D., Zablocki, C.J. and Olson, J.V., 1979. Improved signal discrimination in tectono
maf!nctism: discovery of a volcanomagnetic effect at Kilauea, Hawaii. Phys. Earth Planet. Inter., 19: 331- 336.
Cancellation of extraterrestrial magnetic disturbances by taking simple differences between total field read
ings at spaced stations is imperfect. It is shown that improvement is possible when three component observa
tory data are available from a single station in the general, but not necessarily immediate, vicinity of an array
of total field stations used in a tectono magnetic study. The local effects of a magnetic disturbance field depend
upon its orientation, so that local field differences are more effectively generated by certain orientations of
the disturbance field. The orientation of the disturbance field which correlates best with a local difference
field is determined by a least-squares method, so that the correlated vector signal can be routinely subtracted
from the difference field record. Application of the technique to daily averages of records from three synchro
nised proton magnetometers on Kilauea volcano reveals a 1 .5-nT change in the local field at the time of a
flank eruption in May, 1973. This effect was obscured by noise in the raw difference field data.

I. Introduction
In spite of the limitation imposed by the fact that
they measure only the total geomagnetic field strength,
proton-precession magnetometers are almost univers
ally favoured for tectonomagnetic studies, because
their absolute calibration and essential simplicity make
them ideal both for precisely-repeated surveys and
for continuously-recorded readings at spaced stations "
over long periods and wide areas. Magnetic distur
bances of extraterrestrial origin dominate the data
from individual stations but are largely eliminated by
taking simple differences between readings at
stations up to several tens of kilometres apart. How
ever crustal effects, which originate much closer to
the magnetometers, may affect them quite differently
and so become apparent on a difference record.

The simple difference method of 'noise' elimination is not completely effective because of local
variability in both magnetic and electromagnetic
induction. Several more elaborate differencing methods
have been considered in the hope of distinguishing
very small tectonomagnetic changes. Stacey and
Westcott (1965) examined second differences over a
linear array of three instruments, but found the
improvement over simple differences to be insignifi
cant. Rikitake (I 966) advocated the use of weighted
differences between total field readings of pairs of
adjacent stations, but we have applied this method to
two years of data from a magnetometer array on
Kilauea volcano, Hawaii, and found it to be ineffec
tive.
Beahn (1976) recognised that total field differences
were dependent on direction of the disturbing field

332
and that this dependence may itself be a function of
frequency, and he therefore carried out a spectral
analysis of his difference-field data. However he did
not make full use of the directional effect but correlated his difference fields with a particular vector
component of the disturbance field. Also, we note
that spectral analysis is a disadvantage if we are aiming to devise a method of data processing that can
give an immediate, on-line, in-field indication of a
tectonomagnetic effect. We present here an alternative approach that takes complete account of the
directionality but ignores the frequency effect.
There are three effects that may contribute to the
imperfect cancellation of difference fields:
(l) locally-variable electromagnetic induction in the
earth. due to contrasting conductivities. The component of a disturbance field parallel to the total
field vector, which is the only component affecting a
protun magnetometer. may be appreciably different at sites a few kilometres apart, the differences
being due to the induced fields in different conducting environments:
(2) variable induced magnetization, due to large
susceptibility contrasts: and
(3) total field differences arising from the vector
addition of uniform disturbance fields and nonparallel local total fields (Osgood, 1970).
The first of these effects is reduced by taking daily or
longer-term averages of difference-field readings,
because at sufficiently-long periods a moderatelyconducting body becomes transparent to an oscillatory magnetic field. However, even daily averages
are inadequate in removing the effect completely
and its frequency dependence makes the approach of
Beahn ( 1976) appropriate. Effects (2) and (3) arc nut
removed by period averaging (except in so far as the
disturbance fields themselves are averaged out) and
are independent of frequency because they are
Jirectly in phase with the disturbance fields. Spectral
analysis cannot contribute to the removal of these
effects. They are amenable to the simple vector-sub! ract ion procedure which is developed here. Tllis
procedure also reduces the 'noise' from effect (I), but
less perfectly than a predictive filter based on spectral
analysis. It also has the advantage that any shortterm piezomagnetic signals in the data are virtually
unaffected by the cleaning process.

2. The vector field differencing problem


A local rock magnetic susceptibility x will modify
a magnetic disturbance field !::J3 by an amount up to
E2max =

x!::JJ (S.l. units)

(I )

depending upon the geometry of the magnetic


materiaL the orientation of !::J3 and the position of the
recording station. The subscript 2 is used to identify
effect (2) above. For strongly-magnetic rock, as found
at Kilauea (Zablocki et a!., 197 4 ). x may be as high as
0.1 (0.1 /41T in e.m.u.), so that a disturbance field of
100 nT (typical of mid-latitude storm activity) could
be locally modified by as much as E2 max = 10 nT in
the worst case at a site immediately adjacent to the
rock. Local variations of 1 5 nT would be normal
in an area, such as Kilauea volcano. with highly
magnetic rock. In general the directions of the ambient
field. B. the externally-gencr:Jtcd disturbance, ~B, and
the locally-induced modification of the disturbance.
E 2 , arc all different. but for any site t 2 is uniquely
related to ll.B. Total-field instruments are sensitive
only to the component of E 2 parallel to 8, and our
hasic problem is to identify the orientation of 6.8
that gives the maximum difference between the
values of this component at two total-field sensors.
Then that particular component of any arbitrary ~B
can be used to correct ditlerence-field data.
On a scale of tens of kilometres, spatial variability of the geomagnetic field arises primarily from
rock magnetization, induced or remanent. The thin.l
of the effects listed above occurs when this variability appears as a difference in the orientation of B
at different total-field sites. Then a disturbance field
6.8 which is uniform over the area causes different
increments to appear in I B ~ at each of the sites. The
maximum angular det1ection of the field vector due
to magnetization AJ is

(2)
so that if M = 5 Am- 1 ( 5 X 10- 3 e.m.u.), a value
representative of the strongly-magnetized Kilauea
rocks (Dnell and Cox, 1965) and consistent with the
5
susceptibility assumed above, and B""' 4 X 10- T,
as in Hawaii, 0 max""' 0 .I 57 rad = 9.0. On this basis,
0
det1ections of the field by angles of the order of I
would he normal on Kilauea. The maximum difference

333

in total field increments, E 3 max at two sites where


the angle between 8 vectors is IJ max due to a uniform
disturbance .6.8, occurs when .6.8 is in the plane of the
8 vectors and perpendicular to their average direction.
Then
E3

tnax

;::o2t.f3sin(IJm 1 x/2)

(3)

(see, for example, Osgood, 1970). For a I 00-nT disturbance, as considered before, this gives E 3 max = 10
nT.
Thus, effects (2) and (3) are of comparable magnitudes, and which of them is more important in a
particular situation is a matter of geometry. In both
cases we need to seek the orientation of the disturbance field which causes greatest discrepancy between
difference-field readings. We have not distinguished
between these effects but have treated our data as
though we were dealing with only one effect by seeking the vector disturbance field which is most highly
correlated with total-field difference variations over
a two-year record. It is possible that there is also
some cancellation of any electromagnetic induction
effect [effect (I)] remaining in the 24-hour averaged
data which we used.
It is important to note that the differences between
total field readings change by less than I 0% of the
magnitude of the disturbing field. Thus corrections
to the difference field data can be applied using
vector field data with observational errors ten times
or more greater than those in the total-field data. In
particular, it is not necessary for the vector measurements to be made in precisely the same geographical location. The primary disturbance fields are
sufficiently uniform that vector field data from an observatory a few hundred kilometres from the total
field instruments appear to suffice. We have used
records from Honolulu Magnetic Observatory, 300
km from our total-field instruments at Kilauea
volcano on Hawaii Island.

record averages are


ll

1~

~-

1~

1~

1~

X=-LJX;
n i=I
ll

Y= -LJ Y;
n i=I
(4)
ll

Z = -LJZ;
n i=I
ll

D=~LJD;

i=I

and we are concerned with the deviations of the


period averages from the total record averages:

(5)
Z; =

Z;-

d; =D;- D
The values of X;,Y;, z; are the period averages of the
components of the disturbance field .6.8 and the d;
are period averages of E. We require the values of a, b,
and c that minimise the function

(6)
These are the coefficients of the least-squares-fitted
linear relationship between d; and (x;,Y;, z;). By putting d~/da = d~/db = d~/dc = 0 we obtain three
simultaneous equations for a, b, c:

a 6xf + b 6x;Y; + c 6x;z;

6x;d;

"~XiYi h~ ~}';'; ~y;d; r


+

Yi H

a6x;z;+h6y;z;+c6z[=6z;d;
i

3. The least-squares vector correlation analysis


Consider a series of n sets of period averages
X;, Y;, Z; of magnetic field components (north,
west, down) and corresponding period averages D; of
the difference in total field at two sites. The total

(7)

The direction cosines of the disturbance field which


is most effective in causing the total-field differences
are

c2)1121

a;=

aj(a2 + b2 +

(3 =

b/(a2 + b2 + c2YI2

r = cj(a2 + b2 + c2)112

(8)

334

In treating two years of records Crom three differentially co nnected proton magnetomete rs on
Kilauea volcano we have ado pt ed the following procedure :
(a) A data file of daily averages of each difference
record wa s prepared from th e raw data (values at 2minute int ervals). This gave the data set D;.
(b) Each difference record was de-trended to remove
the best-fitting linear time dependence and high-pa ss
filtered (b y con volution with a Butterworth filter
1
having a co rner frequency of 0.02 day- ). This yielded
the data set d; .
(c) The whole of each d; record was then used with
the corresponding (x;,.J';. z;) data fram Ho nolulu
Magnetic Observatory to obtain the coeffi cien ts (a,
b, c) by eqs. 7.
(d) Values of (ax;+ by;+ cz;) were then subtracted
fro m the unfil tered difference -field data, D;. We refer
to the resultin g data as 'clea ned' difference-fi eld
records. Note that trends and a non-zero average
remain in the "cleaned" D; data.

'<

The records are of the total field at a base station


at Uwekahun a Vault, close to the Hawaiian Volcano
Ob se rvator y , and of the differences in total field
stren gt h between Uwekahuna and each of two slave
stations near Pauahi crater and at Puu Honuaula, 10
km and 43 km, respectively , from the vault along
Kilauea's East Rift Zone. The instrument at ion and
data collection are the sam e as those descri bed by
Davis et a!. ( 19 73), but the record used here is lon ger
than that used in the preliminary analysis , frl)lll which
it was inferred that no volc ano magnetic effect was ob servable .
Figure I. compares the 'raw' and 'cleaned' difference-field data from the Uwekahuna --Pauahi baseline. The values of (a, b, c) ob tained in thi s analysis
are (0.031 0.00 l R, - 0.024 0.00 I 0 , 0.027 0.0031 ),
which indi ca tes that, as suggested by the analysis in
section 2, about 3% of a disturbance field appears
on a t ypical difference- fi eld record. The o rientation
of the disturbance field which is ma ximall y correlated
with variations in the Uwekahuna Pauahi differen ce
record is about 40 away from the total-field orientation. Th e improvement in the cleaned reco rd in Fig. I

Cleaned d1fference record

~"'-t-~

oL-------------------------- -----------"}~.,....,............_

'<
0

"Raw" diff..-enco ..cord

~..,.,..~~~
,do

~o

2~0

~o

sOo

ttbo

1bo

Day Numbtf"

f- 1971-+Jo>------ 1972------ ---~-------1973 ----i

Fig. 1. A co mpariso n of the di ffere nce field Uw ekah una -~


Pauahi (daily averages) before and after "cleaning" . So m<?
high-frequency jitt er is removed by cleaning but trends appear
essentially un affect ed .

-A-Y.<!S."rl_l_l/WEK~A-HU-N-A. .' ,~"----.,----~


~
'=-:--~~~
~
:.1~-----r---~-----r-M
__

ili

4. Results from the Kilauea records

' .

~t....,..,--t~

Day Number

Fig. 2. A section of the Uwekahuna -- Pa uahi diffe re nce-field


reco rd spanning the flank eruption of May 5.19 73 . On this
sca le a very sma ll magnetic effec t is apparent in the cleaned
record but was no t obvious in th e raw data. A reco rd of tilt
at the Uwek ahuna vault is shown for co mparison .

'<.

UWEKAHUNA-PAUAHI
d1ffer.nce ~~~

~'.,JI\"'~. , ~"-f'"~ .

l:k:::===
100

' - 1971

lOO

200

1972

400

Day Nuntber

SOO

~
600

700

100

1973 - - - - - - - ;

Fig. 3. A comp arison of both diffe rence fields with tilt and
with seismic activity in the vicinity of Kilauea.

335

does not appear very remarkable, but in fact differencefield excursions due to magnetic storm activity
extending over several days are substantially reduced,
and it appears that the remaining trend in the data
is probably largely of tectonic origin. Figure 2 gives a
closer inspection of the cleaned record for 200 days
of 1973, which is compared with ground tilting, measured at the Uwekahuna Vault. This section of record
spans the major eruption of the period under examination, the flank eruption of May 5, 1973. A volcanomagnetic effect of about 1 .5 nT is apparent in the
cleaned data, but was not seen in the raw data.
In Figure 3 both difference-field records are plotted
and compared with local seismic energy release and
with tilt, both from Hawaiian Volcano Observatory
records.

5. Discussion
The method of 'cleaning' total-field difference
records by correlating total-field differences with
vector components of the field gives a noticeable
improvement in the record of a volcanomagnetic
effect on Kilauea. This is encouraging because the
Kilauea data do not appear to be particularly suitable
for this type of processing. The ocean is nowhere
more than about 30 km from the instrumental array
(much less in the case of the Puu Honuaula sensor)
and the diurnal variation in the difference field
records shows that they are strongly affected by currents induced in the sea. The data processing method
presented here is particularly well suited to the
removal of magnetic induction effects; the frequency
dependence of electromagnetic noise means that its
elimination will be less perfect. Thus we may expect
our method to be more effective when applied to data
from inland stations.
With regard to the volcano mechanism, the very
small magnitudes of the observed magnetic effects
require explanation. Much more striking transient
magnetic anomalies were observed by Johnston and
Stacey (1969a, b) to accompany (and precede)
eruptions of New Zealand volcanoes. This is presumed to indicate the presence of substantial stresses
and strong piezomagnetic effects. The pattern of eruptive activity of Kilauea volcano, as reported by
Kinoshita et al. (1969), has a characteristic sequence

of summit inflation, accompanied by summit eruptions, followed by deflation with flank eruptions.
A discussion of the eruptive activity during the
period of our records is given by Peterson et al. (1976).
The magnitude of the inflation indicates strains of
Jbout 10-4 , which, if elastic, should produce a striking piezomagnetic anomaly in the strongly-magnetic
rocks of Kilauea, (Davis, 1976). In response to the
earlier-reported absence of such effects (Davis et al.,
197 3), an inelastic model of Kilauea was presented
(Davis et al., 1974). Although a small piezomagnetic
effect has now been seen, it is much too small to be
compatible with an elastic model of the volcano. We
therefore confirm the earlier conclusion that Kilauea
volcano behaves inelastically.
The correlation between magnetic disturbance and
tilt apparent during the May, 1973 event is not always
observed. Davis et al. (1973) previously reported
coincidences of brief magnetic effects with irregularities in the tilt record, and concluded that the volcano
deformation was essentially inelastic but that transient
stresses and consequent magnetic anomalies accompanied local readjustments, so that the correlation
between a magnetic anomaly and tilt may be either
positive or negative and is not systematic. Our present
results are in accord with this conclusion.

References
Beahn, T.J., 1976. Geomagnetic field gradient measurements
and noise reduction techniques in Colorado. J. Geophys.
Res., 81: 6276-6280.
Davis, P.M., 1976. The computed piezomagnetic anomaly
field for Kilauea volcano, Hawaii. J. Geomagn. Geoelectr.,
28: 113-122.
Davis, P.M., Hastie, L.M. and Stacey, F.D., 1974. Stresses
within an active volcano - with particular reference to
Kilauea. Tectonophysics, 22: 355-362.
Davis, P.M., Jackson, D.B., Field, J. and Stacey, F.D., 1973.
Kilauea volcano, Hawaii: a search for the volcanomagnetic
effect. Science, 180: 73-74.
Doell, R.R. and Cox, A., 1965. Paleomagnetism of Hawaiian
lava flows. J. Geophys. Res., 70: 3377-3405.
Johnston, M.J.S. and Stacey, F.D., 1969a. Volcanomagnetic
effect observed on Mt. Ruapehu, New Zealand. J. Geophys. Res., 74: 6541-6544.
Johnston, M.J.S. and Stacey, F.D., 1969b. Transient magnetic anomalies accompanying volcanic eruptions in New
Zealand. Nature, 224: 1289-1290.
Kinoshita, W.T., Koyenagi, R.Y., Wright, T.L. and Fiske, R.S.,

336
1969. Kilauea volcano: the 1967---68 summit eruption.
Science, 166: 459-468.
Osgood, C., 1970. Design and use of a gradiometerconnected
rubidium magnetometer. Rev. Phys. Appl. 5: 113-118.
Peterson, D.W., Christiansen, R.L., Duffield, W.A., Holcomb,
R.T. and Tilling, R.K., 1976. Recent activity of Kilauea
volcano, Hawaii. Bull. Volcano!., 40: 646-656.
Rikitake, T., 1966. Elimination of nonlocal changes from
total intensity values of the geomagnetic field. Bull. Earth
quake Res. Inst. Tokyo Univ., 44: 1041-1070.

Stacey, F.D. and Westcott, P .. 1965. Scismomagnetic effect


limit ot obscrvability imposed by local variations in
geomagnetic disturbances. Nature, 206: 1209 --1211.
Zablocki, C.J., Tilling, R.I., Peterson, D.W., Christiansen,
R.L., Keller, G.V. and Murray. J .C., 1974. A deep research
drill hole at the summit of an active volcano, Kilauea,
Hawaii. Geophys. Res. Lett., I: 323-326.

Vous aimerez peut-être aussi