On Geoelectric Potential Variations Over a Planetary Se...地磁気観測所要領第29巻平成12年12月 Memoir冨oftheKakioka Magnetic Observatory柚1.29，December 2000 On Geoelectric

地磁気観測所要領第29巻平成12年12月Memoir冨oftheKakioka Magnetic Observatory柚1.29，December 2000

On Geoelectric Potential Variations Over a Planetary Sea.le

Ikuko Fujii (Kakioka Magnetic Observatory)，

Hisashi Utada (Earthquake Research Institute， The University of Tokyo)

November 6， 2000

Abstract

Plal1etary-scalc geoelectric potel1tial variations observed with four submarine cables in the Pacific

at periods from seconds to DC were stlldied. The planetary-scale voltages are produced by geo・

magnet.ic field fluctuations of the external origin， sea water mot.ions t.hrough a stcady geomagnet.ic

field， and secular variations of the geomagnetic field at the Earth's outer core. The externally alld

motionally induced fields were mainly examined， and the tidal component and secular variations in the cable voltages were briefly described.

The external1y induced field reflects electrical conduct.ivity distributioll beneath the sea floor. To

investigated spatially averaged conductivity distributions ofもhePhippine Sea Plate. we nsed volt.age

differences mearsured with the Guam-Ninomiya (GN) and Guam-Baler (GP) segmcnts of the TPC-

1 cable together with geomagnetic fields at. I<akioka， Guam， and Muntinlupa. The GN and GP

cables are about 2700km long in almost. nort.h-south al1d east.-west directions， respectivcly， al1d t.hree

geomagnetic observatories locat.e near ends of the two cables. The magnetotelluric (MT) responscs

of the GN and GP cable voltages to t.he geomagnetic variations spat.ially averaged over t.he cahle

distances were obtained at periods 50 seconds to 2 days. The GN alld GP responses show significallt

discrepancies at all periods， which lead to t.wo strikingly different opt.imum lD models. Howcver，

the thin sheet modeling and 2D自nitediπ'erence modeling revealed t.hat. (1) t.he t.hl'ee dimellsional

land-sea distribution alld topography of t.he area significantly disもortthe elect.ric field of the G P cable

and that. is the main cause of the discrepancies bet.ween the GN aud GP responscs， and (2) effect.s

of the subducting Pacific plate llnder the Philippine Sea Plat.e are seen at longer periods where t.he

t.hin sheet modeling was performed. As a result.， the Philippine Sea plate at t.he depth down t.o a few

hundred kilomet.ers is well approximat.ed by a stratified model which consist.s of a low conductivity

layer wit.h the thickness of 80km overlying a high conduct.ivit.y layer. If the low conductivit.y Jay(;~r

correspondsもothe lit.hosphere. it suggests that the Philippine Sea plate is rather t.hick.

The motionally induced filed was studicd using t.he voltage difference betweell Hanauma Bay.

Hawaii and Poillt Arena， California observed by the HAW-l cable with the aid of meteorological

data provided

2 I.Fujii and H.Utada

to t.emporal variations of the wind自eld.These results suggest that. t.he wind-driven flow bas large

scale components， and t.hat t.he planet.ary scale cable can monitor such large scale flows up to much

longer periods than other observations if the water currcnt 011 a large scale flows across t.he cable.

The t.ide and secular variations of a !oIet. of the GN， GP and HAW-l voltagesぉ wellas t.he vol t.age

of the Guam-Midway (GM) cable were estimated by a simple least. squares procedure. In the tidal

analyses， the HAW-l and GN voltages were found to contain larger signals of the oceanic origin

than t.he others. The DC and linear trend were carefully comput.ed from each voltage dat.a set..

Results of the GN and GM cables were lcss reliable than t.hose of the HAW-l and GP， because of the

power 8upply noise and short data lengt.h， respectively. The DC esもimat.eswith small error bars wel'e

obt.ained fl'om the HAW-l and GP cables. which were 0.2 '" 0.3 mV/km. The linear trends were of

an order of 1 x 10-4 mV / km cl吋 fol'every cable. Both the DC and trcnd showed observable amountl'

from an experimental view point.

1 Introduction

Wiもht.he decommissioning of trans-ocean t.elecommunicat.ions cables in 1990's. opportunities t.o re-use

t.hese cables fo1' geoscienti自cpu1'poses have been rapidly inc1'easing. In t.he Paci自c，a network of decom-

nlIssioned訓lbmarinecables has been organized in order to measm'e geoelect.1'ic pot.ent.ials OVf'.1' s(-'vel'al

t.housands kilomete1's. It aimsもoincreasf'. understanding of t.he planet.a.1'y-scale tellurI<' fields in t.he Padfic

region a.t periods from secondsもoDC and will provide a. unique oppo1'tunit.y to observe distribut.ion銭 of

large-scale telluric fields on t.he Ea.1'th 's sllrface. Seven cables are workillg as sensors of the t.elluric field

and severalmo1'e al'e goingもobe iuvolved at. presel1t.

Our knowledge onもhelarge-scale geoelect1'ic potent.ials has been considera.bly improved in l'ecent. yea1's

hy observational studies using ret.ired a.nd in-service cables and theoret.ical approach. Voltage observed by

using a. large-scale submarine ca.ble can be produced by three ca.uses of geophysicaI importance (Fig.l):

( 1) geomagnetic field fluct.uations dueもoもlIue・varyingelectric current. systems in t.he ionospher~ and

magnet.osphere， (2) motions of conducting sea watel' through t.he geomaglletic field， and (3) leakage of

t.ellul'ic currenもsassociated with t.he toroidal geomagnetic field at t.he co1'e-mant1e boundal'Y (MELONI rt

al.. 1983). Because the toroidal geomagnetic field itself is not obse1'vable on the Eart.h 's surface， t.lwre

is 110 way but. observing source 3 to detect it， as poilued out. by R.uNCORN (1954，HJ64)ト.Thesc 符町so川lIr吃αt'、S

have di日町'e引r泡湖e白n】ttもu山;I沿lmesca恥白:0ωmeach ot.“her. SOllrce 1 is dominallもaもperiods short.e1' t.ha.n a mont.h (e.g.， LANZEROTTI ctα1.， 1990)， while source 2 excluding t.he oceanic tides induces tlu空t.elluricCllrI'entぷ at.

periods longer t.han several days (e.g.， CHAVE et aL 1992a). Source 3 is expeded t.o be signifi(，Hnt. at

DC"'several decades (e.g.， YOI(OYAMA AND YUKUTAKE， 1991).

Magllit.ude of the tOL'Oidal geomagnet.ic. field is a considerably import.anもparameterもodetermine t.he

dynamo process in the outer core. Since t.he magnit.ude of t.he leaked COl'e field is too sma.ll to be measu1'ed

with short spanned electrodes， t.he Iarge-scale voltage measurement will provide a unique opport.unit.y t.o

obsel've it. (see review by LANZEROTTI et al.， 1993). Thc core恥ldscalc

011 Geoelect.ric Potential Variations Over a Planet.ary Scale

Period (day)

' Motionally Induced Field

Figure 1: Three main causes of large scale geoelectric potentials.

3

The mot.ionally induced field is not completely understood， especialIy on a planet.ary scale. It. is

essential to incre描 eknowledge on the motional illduced field because it is noise for the other purpos何 as

well as those provide information 00 large-scale oceanic flows. This phellomenon w部自rstrecognized by

FARADAY (1832) who attempted a water velocity observatioo in the River Thames. SANFORD (1971) and

CHAVE AND LUTHER (1990) developed the theory that the horizontal electric field 80t the oce8on bott.om

is proport.ional to the conductivity weighted depth-integrated flow. In the case of t.he cable volt.8oge，

horizontal integratioo over a cable length is added， and therefore the volt.age reflects the total alllount.

of water flux across the cable. The motional induced fields on the c80ble volt.8oges have been det.~ct.ed in

many experiments (e.g.， LONGUET-HIGGINS， 1949; STOMMEL， 1954; Cox et al.， 1964; TERAMOTO， HJil;

DUFFUS AND FOWLER， 1974; RICHARDS， 1977; MEDFORD et al.， 1981; MORJ， 1987; KAWATATE， ct al川

1991; TERAMOTO AND KOJJMA， 1994)， but old studies gave discoura.ging conclusions due to complexit.y of interpretation. On the other hand， measurements with short spanned cables in t.he St.rait.s of FIOl.ida

(SANFORD， 1982; LARSEN AND SANFORD， 1985; SPAIN AND SANFORD， 1987; LEAMAN et al. 1987;

LARSEN I 1992) and in Bering Strait (BLOOM， 1964) were two of successful examples to prove linea.r

relat.ionships between the cable volt.ages and independently observed t.ra.nsports. Situ8ot.ion is even WOl"se


in the case of a planetary-scale submarine cable， because independent tra.nsport measurements are ral'e

and the motiona助 inducedsignals are weak dueもospatial averaging (CHAVE et al.， 1992b). Numerical

simulation (e.g.， FLOSADOTTIR et al.， 1997a，b) will be one ofthe useful approaches to reduce uncert.ainties

in interpretation部 wellas longもel'mexperiments.

The externally induced field has been investigated fromωpect.s of the source process (e.g・.MEDFORD

et al.， 1981， 1989; LANZEROTTI et α1.， 1992; FUJII et al.， 1995) and the elec.trical conductivity distribut.ion

beneathもhesea floor (e.g.， DUFFUS AND FOWLER， 1974; FUJII et al.， 1993; LIZARRALDE et al.， 1995).

The electrical conductivity distribution is one of the most important parametersもodetermine the physical

conditions of the Earth's interior. Magnetotelluric (MT) and geomagnetic depth sounding (GDS) studies

revealed conductivity structures of land (e.g.， JONES AND HUTl'ON， 1979a，b; SCHULTZ AND LARSEN，

1990; OGAWA， 1987). Oceallic influence on the EM field of coおもalareas has been recognized recently

(e.g.， UTADA， 1987; MACKIE et al.， 1988) and knowledge of t恥 conductivit.ydistribution in the oceanic

area has been required. Howevel'， studies on the oceanic areas are limited (e.g.， FJLLOUX， 1967， 1980，

1982， 1983; YUKUTAKE et al.， 1983; LAW AND GREENHOUSE， 1981; SHIMAKAWA AND HONI<URA. lO91:

HEINSON et al.， 1993; TOH， 1993) mainly due to lack of stable geoelectromagnetic observatories on the sea

floor. LIZARRALDE et al. (1995) estimated t.he spatial averaged conducもivit.ydistribution in the nort.heast

Pacific down to the lower mantle usil1g cable volt.ages wit.h an aid of the GDS result.s at. H01101ulu. The

cables will be useful sensors of the regional telluric currents il1 t.he oceanic area o.ver a wide frequcl1cy

ra.nge， although the spatial variation of the geomagnetic自eldandもhedistortion of the telluric c.urrel1t

are uncertain factors.

Thus， it is meaningful to examine the large-scale cable voltages from various aspect札 Thevolt.age data

have been accumulating in the Pacific， andもhelongest dat.a seもh槌もen-yearduraもionfor now. Since it is

the日rsttime that such a long-term data set of planeもary-scalevoltages is available， it is timely to stu<.ly

the low frequency band carefully. Inもhisstudy， we examineもhevolもagesof four large-scale submariu('

cables in the Paci白cat the pcriod range frol11 seconds t.o DC. The ext

2 Observations

The cables used in this study are the Guam-Ninomiya(GN)， Guam-Baler(GP)， and Guam-Midway(GM)

segments of TPC・1and HAW-l (Hallauma Bay to Point Arit吋儲 shownin Figure 2. Spec.ificatioll of

these cable syst.ems and voltage measuremellt.s are described in this section.

2.1 Technical description

Each submarine cable mainly consi蜘 of co・似ialcopper cable， repeaters， equalizers， and elecもl'odesa.も

t.he two ends. The G N cable has a power supply system at the Guam end. The repeater and equalizer

are equipments to transfer signalsinもheiroriginal forms from one endもoもhe0もher，and the power supply

011 Geoelectric Potential Variations Over a Planctarv Scalc " 巳2

F'igure 2: Cable net¥'、!ork.The subl1larine cables l1sed in Lhis study are indicated by thick solid lines. TIII'

geomagnet.ic observatories (KA K， M UT， GAM， HON， I'RN， and日OU)are al50討hown

system is necessary to have the repeal，er and equalizer worked. The GN， GP and GM cables are grollllded

at. Guam， U.S.A. (13.soN. liI4.90E) with a telecommullical.iolls grolllld system which is shart>d by otlwr

a.ctive cables. The othel' ellds al'e gl'ollnded田 pa.ra.l，elyfrom oLher acLive ca.bles at Ninomiya.， .Ia.pan

(35.3'N， 139.3'E)， Baler. Philippille (15.iG'N， 121.56'E)， and Midwa.y， U.S.A. (28.2'N， lii.38'E) via

Wake. U.S.A. (19.2i'N， 166.65'E) ror the GN， GP， and GM cables， resperlively. The HJ¥W-I has 0¥¥'11

new grollnd system a.t Point Arena可 U.S.A. 1，0 avoid InAuence fl'oJll oLher active cabh:~s ， and is grounded

at. Hanauma Bay， U .S.A 、Ifitha l.clccollllllllllications ground systern which is not used by ot.her ca.blf's

Specifica.tion 01' the cabl四 isgiven in Table 1

111 a. large-scale voltage measuremenl， t.he cか axia.lcable works as the lead LO cOllnccL the cleclrod伺

with a. voltmeter as ShOWll in Figul'c :3. Volt.a.ge diflerences bet.weenもhecenter conductor Or t.hc ce:lble a.nd

a referellce ground direct.ly reHect. the geoelectrical potcntial in an 111lpowered Sta.LUS. Thc GP. GM. a.nd

IIA¥oV-l cables ha.ve been used in this way. On I.he other hand， in a powered st.at.lIs， primary IIlcasure is

t.he voltage or the power supply. Thc power reeding equipmcnt controls the supply currcnt cOllsl，a.IIL. As凡

Table 1: Specifica.t.ion of submarine cables

TPC-I HAW-l

GN GP GM(G-W) GM(W-M) norl.h

lellgth(km) 2700 2il6 2052 277U 3926

repeater 74 76 58 78 5i

equalizer 7 7 5 7

6 I.Fujii and II.ULada

Large-Scale Voltage Measurement

Cable Cross-Section

....

-聞・-掴・-園自・・4・・圃-・--・・h

Figure 3: La.rge-scale vollage meaSlll'CI1lCI1L with a cablc

result， the ioLal voltage which is t，he sum of th(' supply and exl，crnally induccd voltages is kepl ('0I151，ant..

Sillcc the supply volt，a.ge fludua.l，cs because or t.hc geoeleclric potenlial variations bebvecn the I，¥¥'o t"'lIds

of t，he cable， we can ext，ract. only a t.ime-varying parl of the geoelct.ric potentials. The GN cablc has

been used in the powered stat.¥Is. The supply voltage varia.tions of t.he GN cable aTC seriously al-rcd.ed by

f1ucluatIons of t.hc supply current due to Instahilit.y ol' the power supply equIpmcnt to keep t.he current

const.allt.. Thererorc it. is necesselry to Illonitol' both t.he voltage and t.he current.出ldt.o correcL l!w voltage

va.l'ialiolls by the current日lICtlla.tlOll

2.2 TPC-l

The t.hree cabl回 whichare gl'ollnded at GlIam have bccn IIscd by a group of Eart.hqllake R<.'scar('h

Insl，if.ute， the University of Tol、yo(e.g.， YUI<UTAI<8 AND I'IAMANO， 1990). 0111日rstmcasurcmcnt. was

madc on Dcccmbcr 17， 1991 using thc GN cablc. Voltagc difrcl'cnccs betwef'1l the ccnt.cr condudol'れnd

t，hc Guam grouncl w('re recordcd cvcry second for 8 hours. Sincc I，hat timc， the systelll has bCCII pow(>rcd

孔b0l11，4100V frol11 Gl1am by I，he DC constant cllrrent supply. ¥Ve have coniinllollsly observed thc voll.agc

and I.he current. of川町GNcable川 t.heGuam剖.al.Ionsillce 1992. A thermist.or 1.0 monit.or ihe t.cll1pera.turc

was al.t.ached to tJlC outside panel of I'.he DC powcr equipment ill Decembc/'， 1992. The ITlcasurelllcnj. of

t.he unpowered volta.ge of the GP cablc was a.dded at ihe Guam st.a.tion in Scpt.cmber， 1093 as well (¥s t.hc

111Ce.1SUI'Cment of the ret.urn curreni throllgh t.hc Eart.h at the Nillomiya station. The lInpowPrcd volt.a皆、

measurcment orthe GM cable start.ccl in Decel1lbcl'， 1994， alt.hough the grolllld on l!w Midway island was

cOlllplet.ed 011 March [" 1995

On Geoelectric Potential Variations Over a Planetary Scale 7

ω冨明、同

hω

ロωSFωhh問

aS国

Ninomiya Guam

Figure 4: Measurement system of the s叩 plyvoltage (V)， t.he c川 rentsat the Gllam (10) and Ninomiya

(ん)station， and t.he temperature at t.he outside of t.he DC power equipment. (T)

Figure 4 illllstrates t.he whole measurement syst.em working sincの September，1993. Thc PC at. the

Guam station handles measurements of the three kinds of volもages.the supply currenもofthe GN cable

andもhetemperature when it receives a trigger via the GN cable from the PC at the Ninomiya stat.ion.

The Ninomiya PC simultaneously measures the Cllrrent at the end of Ninomiya side， while it wait.s for

the data of the Guam stat.ion being sent. back via the cable. All data are recorded on a magnetか optica.I

disk at Ninomiya. The sampling inもervalis 1 minute during .January to May， 1992， 2 secωondゐisdurin

June， 1992 to September， 1993， 3 seconds from September司 1993to December， 1994. and 2 sfcomls

from December， 1994 to the present. We referred the inner clock of the PC for the first. 15 111οl1ths，

al1d the clock monotonically ga.ined a.gainst the Greenwich Time. The ma.ximum Jag was 2 minutcs.

Since the external clock with aut.omaもicsynchronization by broadcasting radio system started t.o work in

Sepもember，1993， the time has been adjusted to t.he standard time with an accuracy of a few milliseconds

except mismanagement.

2.3 HAW-l

The measurement of the unpowered voJta.ge using HAW-1 hぉ beencOllducted by a group of AT&T

Bell laboratories since April， 1990 (e.g.， LANZEROT1'I et al.， 1992). The HAW-1 syst.ems cOllsist.s of

t.wo parallel cables separated about. 100 km・.Voltage di町erellcesbetweell the center conduct.ors alld t.he

Hana.uma Bay grollnd have beell recorded every two secollds. ¥Ve used ollly t.he dat.a of t.he nort.hern

cable， becallse the southern one broke in April， 1991 a.nd the voltage suffers all offset. from t.he brea.1


(LANZEROTTI et al.， 1993).

3 Data

Volもagedaもaused inもhissもudyare 1 sec and 2 or 3 sec values ofもherespectively unpowered and powered

GN volもagesfrom December， 1991 to Augusも， 1994， 2 or 3 sec values of the GP cable from Sepもember，

1993 to August， 1994， 2 sec values of the GM cable from March to August， 1994， and 20 min v80lues of

the HAW-l c80ble from April， 1990 to February， 1994. We an80lyzed the GN 80nd GP data set t.o st.udy thc

externally induced effect， whileもheHAW-l d80ta set was used to examineもhemoもion8ollyinduced effect.

This is because (1)もheGN data set is noisy dueもopower supply in a low frequency balld， (2) the GP

d80ta set is gappy， and (3) the HAW-l data set isもhelOllgest. All d80ta seもswere used to obtain t.idal

signals and linear trends. Basic features ofもhevoltages will be described in next three sections.

The geomagnetic field data which were analyzed together withもheGN and GP data sets came

from Kakioka magnetic observatory of the Japan Meteorological Agellcy (KAK， 36.140N， 140.110E)，

from Guam observatory， Guam， U.S.A. (GAM， 13.60N. 144.90E) and Muntilllupa observatol'Y， LUZOll，

Philippines (MUT， 14.370N， 121.020E) which 80re maint80ined by the 2100MM magnetic observation grou(>

(YUMOTO et al.， 1992). Three component.s of the geomagnetic自eldin the geographic coordinate are

recorded every minute at KAK without data gaps. The GAM and MUT dat80 set.s consist， of 3 componcllt.s

of the geomagnet.ic field variaもionsin the geom8ognetic c.oordina.te recorded every second. They were

t.ransferred t.o the geographic coordinate using IGRF 90. The GAM's data contains 11 % of gaps aud t.he

gap r叫ioof MUT is more than 50%. KAK and GAM are used as reference for the GN c80ble wit.h westward

rotation about 130 along the cable， and GAM 80nd MUT， for the GP cable with eastward rotat.ion 8obout.

50 along the cable. Hourly values of the geomagnetic field at Honolulu (HON， 21.30 N， 202.00E)， FreSllO

(FRN， 37.10N， 240.30E)， Boulder (BOU， 40.10N， 254.80E) were provided by the US Geological Survey

and were referred to remove the externally induced voltage from the HAW-l dat.a set. The gap ratios

of the three observatory are 2.8%， 0.5%， and 0.5%， respectively. These geomagnetic observat.ories are

shown .in Figure 2.

All spectra in this study were computed using the multiple prol8ote windows (Slepian sequence) in-

troduced by THOMSON (1977). Applic8otion of t.hese windows c80n decrease spectr80lleakage much het.ter

もhana standard Hanning window. Furthermore， the degree ofも

3.1 GN

Charac.t.eristics of the voltages greatly depend on power mode. Figure 5 shows 8 how's of cable volt.ages

measured in both unpowered and powered statuses. The powered volt8oge contains much high frequency

lloise. The power spectra of the same d80ta shows th80t the lloise in the powered cable is mainly 80t

frequencies higherもhan4x 10-3 H Z (250 sec) (Fig.6). For this study， therefore， dat.a of the powered cable

will be used for frequencies lowerもhan4 X 10-3 H Z 80fter lowpass-filtering 80nd resampling.

9 On Geoelectric Potel1tial Variations Over a Planet.ary Scale

4

O

(〉)凶O〈'HJO〉

-4 20:00 1991.Dec.17 (UT) 14:00

4080

4076

(〉)凶O〈'HJ。〉

17:00 1993.J an.l (UT) 11:00 4072

Figure 5: 8-hour unpowered volもagesaも13:00-21:000nDecember 17， 1992 (top)、and8・hourpowered

voltages at 10:00-18:00 on January 1， 1993 (bottom).

U npowered vo1tage

Without any special technique of data processing， the 8-hour segment of the unpowered volt.ages shows

high coherellcy with the horizontal geomagnet.ic field at KAK and GAM at. frequencies from i x 10-4 t.o

5x 10-3 H.= (Fig.7). This suggest.s that the ext.ernally illduced effect is dominant. at t.his frequency range

in the unpowered data set. A rapid decrease of coherency at lower frequencies is caused by cOlltaminat.ioll

of longer term signals (-DC)， and low coherency at higher frequencies is due t.o lack of sellsitivit.y of the

magnetometers in KAK and GAM. The coherency between the voltage and the perpendicular geomagnetic

field is also higb (2:: 0.9)， however slightly lower than one including the parallel geomagnetic自eld(Fig.7).

This is consistent with the expected relationship between the inducing geomagnetic and induced t.elluric

fields which are perpendicular with each other in la臼rallyhomogeneous half space. Inclusion of the

vertical component does not affectもhecoherency at this fl'equency range， which probably shows that the

vertical component is a linear response to t.he horizontal components of the geomagnetic field variations

(e.g.， RIKITAKE AND YOKOYAMA， 1955).

3.1.1

Powered data

One minute values of the powered voltage， currents aもGuamand Ninomiya， and temperat.ure at. Guam

were made by a following procedure. First， spikes defilled in a data adaptive way were removed. If t.he

3.1.2

I.Flljii and 1.1. Ulada 10

le-08

lc-1O

Ic-14

le-12

lc-16

Ic-18

(NZ両且曹三巴と〆})hU20&

k-20 Ie-05 !c+()1I Jc-Ol lc-03 Ie-U2

Frequcncy (Hz) 1<体4

1合igul'e6: Power specもra01" t，he unpowcl'ccl(solid line) and powcred(dashed lil川)volt，ages

0.6

O.H

0.4

0.2

つ一包句コ古川町

huz巴ω

占。U

]c+(K)

l

lじりlI c-D:¥ I c-02 F同 llucncy(1-17.)

lc.l)4 O lc-05

F'igurc i: Squared lIlult.iple coherclJcy bel.ween the t1upow<.'red voltage a.nd t. hl~ horiぉ011Lal geollla話Iwt.ic

field at I¥.a.kioka a.ncl Gua.m (solid line)， and partial cohcl'cllcy between the tI日powcl'eclvoltage a，nd t.he

east-west gcomagnct.ic自eld(dashcd lillc). The 05% conficlence limit. is indicatcd by a uash ancl dol.!'ccl

line.

first clcl'iva.t.ive or two adjaccnt， da.l.a is grcaLcr t.han 20 t.irnes or the Illean of the fil'st derIva.t.Ivf'S 1'01' a

3・da.ysegmcnIr， and if a sequencc 01' spiky dal.a， is short.er lhan 1.5 lillles of lhe resalllpling illlcl'val which

is one rninut.e， lhose dat.a. were lrcaled as dal，a gaps. The spikes were fOUlld aboul. 2 t.inlt'5 pcr day in t.he

powered vollage， a.llhough in olhcr cOlllponenls lhe spikcs wcrc allllost. none. Second， t，he lime ~日 "'語、、 erc

correded linearly by thc cubic splinc interpolation. Dat.a ga.ps shorter t.han 1.5 j.Imes or t.he 1'1、S<-llllpling

interval、、，1eresimult.aneou5ly filled by ihe intcrpola.tion. I:'inally， the da.1:a were lowpass filtercd wil.h t.he

ClIレo汀frequency5x 10-3 H z and wf:'rc r田 ell11pledInl，o 1 m.';11 tl.le. The cut.-ofr war; chosen 50 t.ha.l. t.hc powel

Oll Geoelect.ric Potential Variations Over a Planet.ary Scale 11

spect.ra slope of the powered volもageoverlaps tha.t ofもheunpowered vo1t.age in the overla.pped frequency

ba.nd using t.he first 3 m.onths of the whole data seも.30 min values were t.hell made ill t.he same procedure

as for 1 min values except the cut.-off frequency was 2.7 x 10-4 H =.

〉

4140 I ・・..，，--，-~

4110 P'{'¥.fV¥ fUIWA必爪仇州仰~叫ん .tI¥附吋'JV "'¥叫..ゴ

4080 ・・・・・・・・・ 2 ・・・o 10 20 30 40

day

υ

28 ・・・・・・・・~・

26 Fvv'"町、ハハ~戸~作(μV"V\久八~ .A ・~

24 ・・・o 10 20 30 40

day

367 ・・・・

364 ~ l¥.. -1

361 L--ムーL-..J・・・・・・・・o 10 20 30 40

day

〈

E

〈

E亡L。

司

L

n

y

ζ

U

守

f

ζ

U

ζ

U

司

3

弓

3

司

3

30 40

・・ E

一一一一，

20 day

nu

Figure 8: 45・daysplot of the powered voltage， the temperat.ure and t.he currents in Gllam a.nd Ninomiya

fromもopt.o bottom.

A major difference betweell the ullpowered a.nd powered volt.ages is noise level. The powerpd voll.a.ge，

t.he supply CUl'rent at Guam， and the t.emperature at t.he outside of the panel of t.he DC powpr eq1Jipment.

at Guam for 45 days look similar each other (Fig.8) suggest.ing significant influence of t.he supply currellt.

and temperature fluctuations. The problem in applirat.ion of t.he powered volt.age to geophysics is how

to reduce these noises. LARSEN (1991， 1992) observed these insuellce Oll t.he powered voltages of t.he Florida cables， and stat.istically removed them. The voltage caused by t.he current fluctuation corresponds

to Ohm 's }aw， that is， the product. of the current fluct.uatioll and circuit resistance if t.he whole cir<:・uit.

does not have capacit.allce 01' impedallce. The cause of the temperat.ure e宵ectwas cOllcluded that. t.he

resistance of a power separation filもerat one end fluct.uated with the temperature.

Our situation is slight.1y clifferent. from t.haもofthe Floricla cable. The current at Guam ext.remely

resembles to the temperature thel'c and does 110もcorrelatewit.h the CUl'rent. aもNinomiya(Fig.8). The

squared coherence betweenもhecurrent alld temperature aもGuamShOWll in Figure 9 sllggest.s t.hat. t.Jl('

current at Guam can be expressed as 'lowpassed' t.emperature. The relatiollship among the volt.age，

current and temperature at Guam was examined. At first， we examined the relationship betweell the

I.Fujii and H.Utada. 12

0.8

0.6

0.4

0.2

huzmw』

ωsou-Mggv悶

o le・07 0.001

Figure 9: Squared coherence beもweenthe current. and悦mperatureat. Guam.

0.0001 le・05Frequency (Hz)

le・06

current and旬mperature.

A linear relationship between t.he current and temperature is possible if the resister t.o monit.or the

supply current atもheGuam sもat.ion(Fig.4) has temperaもuredependence. The resist.or is a wire-wound

type whose temperature coefficient is expected to be much largerもha.nthaもofthe Ma.nganill resistα・of

the Nillomiya station. This mayexplain why the currenも叫 Ninomiya.is more stable.

The voltage difference 九betweentheもwoends ofもheresister Rg is givell by

、... ，， -E・・

，， ••

‘、

where 1 is the true supply curre肌 ¥Vea.ctually observeらinsteadof 1， and calcula.t.e 1 from 九usillga

constant resist.ance Rg. If the resistor has a linear one-dimensional temperature dependence， R.g can hc

written by

ら=RgI

(~)

whereα9 denotes. the temperature coefficient.， and tg is the time-varying parもofもhetemperature Tg

a.t the resistor. Following LARSEN (1991)， a.ll variables a.re assumedもobe divided inもot.he mean and

time-varying pa.rts as follows:

Rg = Rg(1 +αgtg(t))

(3) 1 = 1 + i(t)

(4) ら=ら+vg(t)

(5)

Wlien time-varying parts are much smaller thall means， t.he time-varying part of Equation 1 is writ.t.f'n

by using Equation 2-5，

九=九+tg(t)

(6)

(7)

ら=(1+αg九)i+αgItg

" i" "! g一応

where


ら=ら+ら(t) (8)

らisequivalent to the time-varying part of the supply current at Guam shown in Figure 8. Since we

observe the temperature 九=巧+tp(t) aもtheout.side of the panel of the DC power equipment. t.he

relationship between tp and tg is assumed as a linear convolution of tp similar to the expression by LARSEN

(1991 ):

t. = 10"" c.什)仲 (9)

Equation 9 determines the causal relationship;もhetemperature fluctuation outsideもhepallel causes

もhetemperature fluctuations of the inside. Not.e that thisぉsumptionis not necessary if tg is directly

measured. Cg is expected to be a s11100th function andもoapproach to 0 with illcreasing T， for illst.a.nce‘らis an exponential function in the diffusion process. Equation 9 leads an approximated form of Equation

6，

ig = (1 +仏山、，，nυ 官且.，

•••

、

Thus， the linear relationship between ig and tp is modeled in Equation 10. If the relationship of Equat.ion

10 is satisfied， ig contains i and tp

TheもransferfuncもionαgICgwas statistically determined in the time domain using 30 min values

of ig and tp・ SinceαgICg is supposed to be a smooth function， the least squares自t.tingwit.h ABIC

criterioll (AKAIKE， 1980) was applied. This means t.hat. the best. model was selected on t.he basis of

good fitness， simplicity， and smoothness. Figure 10 shows an obtained smooth transfer fu 11 C.もion.The

predic.ted currellt fluctuation using the transfer function of Figure 10 re加 onablyreproduces the observed

data (Fig.11) suggesting Equation 10 is a good approximation. The residual which is expect.edもobe

t.he real supply current multiplied by a unknown constant factor， clearly shows smaller suctuat.ions t.han

ら(Fig.ll)，however it. is not similar to the current at. Ninomiya shown in Figure 8. Rβasons why t.he

curren旬 st.ilIdiπers ill Guam and Ninomiya are unknown. One of possible c.auses would be current leakage

t.hrough the cable at deep water equipments in the circuit.

。

0.4

0.3ト。

U 司E 0.2

。

。。。。。。。。。。。

。

o 2 3 4 5 6 7 8 Time lag (hour)

Figure 10: Transfer function ofもhecurrent to the temperature in the time domain

14

2

2

I.Fujii and H.Utada

367 一一-.一一一「寸一丁一

364 1: r..r¥.. -1

361 ・・ a ・1

o 10 20 30 40 day

0 10 20 30 40 day

367 r ， --，

364ト 4

361 I

o 10 20 30 40

Figure 11: 45・daysplot of the observed current (top)， the predicted currenも(middle)，andもheresidual

(bottom). The duration of the data is the same部 Figure8.

Next， the noise of the powered voltage is estimated by using ig. The supply voltage九isgiven by

九=に附)1(t-T)dT +九 (11 )

where R represen旬 acircuit impedance， and 1 and Ve denote the current and t.he ext.ernally iuduced

voltage， respectively. Since v.， (，..，.， 4100V) isも00dangerousもomeasure directly， we divide the voltage int.o

abouも1000:1and calculate v.， using a constant ratio of the divider resistance. vs may contain temperat.ure

ftuctuations caused by the temperat.ure dependencies of the divider and circuit resist.allce， although they

canllot be sepa凶 edusing only Tp • Therefo民 lw鎚 Sl伽もit・utedforらtωoe慌xp戸re郎ssboωth自u旧1比ctu凶at.i悶。m∞nof も山hecurrent a剖n吋dt.旬.εemperature札， and R w 紛 s叩ubst“iもωuも旬edfおoran apparent i加111甲pedanceRc = Rc + η (1)

which reprωents a coefficient. including R. When time varying part.s are much smaller than means，' a.

time-varying part of V." namely V.，(t)， is rewritten as follows，

vs(t) =叫 (t)+にrc(T)ら(t一材+ve(t) (12)

The自rsもtermof the right-hand side contains effects ofもhecurrent and temperat.ure ftuctuations， and

is actually the same as what LARSEN (1991) estimated， although the meanings of factors are different..

The secondもermwhich w部 neglectedill LARSEN (1991) was required in our case. Complex nat.ure ofもhe

circuit impedance， for insもance，is included in thisもerm，unless they are nOlllinear. Equation 12 was evaluated in a robust method (CHAVE AND THOMSON， 1987; THOMSON AND CHAVE，

1991) USillg 2・yeardata seもofV." igl and t.he averaged geomaglletic field of KAK and GAM. The result.

indicatcs that Rc is expressed by all impedance like a RC circuit (Fig.12).

15 011 Geoelectric Pot.ential Variations Over a Planetary Scale

す

+

AT

・irAVA

TE@E--T

E.E

・-T

目。Ei

TUr

--Th田川広

ll川蹴叶l

16

10

6

14

12

8

(EZouC8ロSEEM-

le・03le・04Frequency (Hz)

4

2 le・06 le-02 le・05

40

?it--EEBB-

守

4laEE.EEaEBA

BEESE-E'EEl'I

ll-

。ftil』t

t

y'EE』EEBEE-EEEEEEBEE-&

Tliva--

τEEa@BEE--

Tls+lli

TAYA

TEE-E+llEi

--BEI'AVESE-A

YEE』E'aAVEEE'S』目&

20

。(ω巴凶ω司)ωおZ仏

-20

-40

le・02le・03le・04Frequency (Hz)

le-05 -60 le-06

Figure 12: Norm (top) and ph槌 e(bottom) of the circuit impedance.

Corrected vo1tage

We call vs-Rcig the corrected voltage， which gives a rough estimat.e ofthe voltage wit.hout the curreut. a.nd

temperature effects. Vs a.nd 19 for 23 rnonths are 4105.4 V and 363.2 mA， respectively， giving an est.imat.e

of Rc as 11.30 kn. 30 min values of the corrected voltage for about 1200 dαys are showl1 in Figure 13. The tidal beating is not dominant" which is different from other reports on the unpowered volt，a.ges. Tllf、

power spectra ofもhecorrected voltage for 23 dαys from December 27， 1993 increase monotollically with

decreasing frequency up to lx 10-5Hz and show a. moderate slope at lower frequencies (Fig.14). High

frequellcy Iloises may remain at highel' frequencies， because the slope changes at 3 x LO-3 H z. The tide， Sq

and their harmonics are clearly seell. The squared mu1t，iple coherency among the corrected voltag~ ， t.he

horizontal geomagneもicfields 抗 KAKand GAM， and ig shows that. the corrected volもageis dominat.ed

by the effects of the horizontal geomagnetic field and the current fluctuatiol1 (Fig.15). Furthermore t.he

3.1.3

I.Fujii and H.Utada 16

partial coherency in Figu1'e 15 inclicates that the geomagnet.ic field component pe1'pendicular to t.he cable

a宵ectsmuch the co1'1'ected voltage at. f1'equency highe1'もhan2.3 x 10-5 H z， then the geomagnet.ic field

tends to be weake1' at lowe1' frequellcies as t.he cu1'1'ent influence gets sも1'onge1'. This suggests t仙ha叫tst.able

esも“ima剖，tιt.ぬnof V叫ei白spossible a抗tlea槌s剖ta叫も highe1'f1'equencies whe1'e the noise level is low.

円幽?↑…ω

〉

Figu1'e 13: Corrected voltage fo1' 1200 days.

le・07

le-09

le-Il

le・JO

le・08

le-12

le・13

le-14

(HEN-(E、〉))』

go仏

le-15 le・07 le・02le・03le・05 le・04

Frequency (Hz) le・06

Figu1'e 14: Powe1' spectra of t.he co1'1'ect.ed voltage fo1' 23 days f1'om December 2i， 1993.

Iも isessentialもhaιnoisesa1'e p1'ecisely estimated in o1'derもoexpand analysis to lower frequencies. 1'00

Inもhissもudy，however， we did noもcompuもea time se1'ies Vs一I Rc(r)ら(t-r)dr because a suit.able ./-α3

esもimateof Rc was not obtained. Rc in Figure 12 is disturbed at frequencies of the t.ide， Sq and i旬

harmonics， and has f1'equency dependence aも su1'prisinglylow frequencies. These suggest that. a new

procedure is necessary to several noises and signals together. To be taken into accounもarethe oceanIc

and atmospheric tides， Sq with short wave lengths， the moもionallyinduced signals at low frequencies， and a smooth Rc as well asもheexもernallyinduced field. The basic idea of BAYTAP-G (ISHIGURO et al， 1984) and the remote reference met.hod (LARSEN， 1989， 1992) should be 1'efe1'1'ed.

Analysis of the powered data set. sugges旬もhatsignificant decl'ease of the noise in the powered volt.age

would be possible by (1) exchange of the monit.o1'i時1'esistorsfo1' t.hose wiもhsmall t.empe1'atu1'e depen-

dency， and (2) di1'ect. measurement.s of the tempe1'at.ure aももhemoniもo1'ingresisもo1'sand power supply

equipment. The 1'esisto1's that a1'e used in the telecommunIcations sys旬mhave several weak points fo1' a

17 On Geoelectric Potential Va.ria.tions Over a. PlaneLary Scale

0.8

0.6

0.2

0.4

つω』dヨ

σ同hυ£ω』ωz

。υ

O Ic-06 Ic-02

Figure 15: Mu1tip1e coherency beいveenLhe volta.gc， the current， and the horizonta.1 g巳orna.gnet.icfield at

Kakioka. and Guam (solid jine)， and parLial coherency of the pcrpendicular geoma品netIrf-i('ldれJaslled

line). The period of t.he data is the sa.me田 Figure14

lc-Q3 Ic-04 Frcqucncy (Hz)

Ic-05

detailed scicntific investiga.t.iol1， 50 thal it could be recolllmenc!ed 1.0 replace them wit.h Illore st.a.ble ones

when powered voltages are measured for thc scientific purposes

GP and GM

The original 2 01' 3 sec values of the unpowercc! volt.ages were processed ill the same p ro('ed \l l" ~ as thC'

powered voltage of the GN cable dcscribcd In Sediol1 3.1. The resampling interval is :30 scc， a.nd thcn

30 m.in， and the cut-o汀 frequenciesare ¥.ix ¥0-' and 2.8x ¥0-4/1ご， respertively.AII of :30 11/;1/ valucs

a.re shown in Figure 16. Unfortunat.ely， these are ga.ppy due to problerns of the da.t，a acqllisit.ioll systcll1

Therefore， the GP cable data were used in the analysis of the ext.ernally induced field， howcvcr I.he t1sC'

of t.he GM data set was limited only in the tidal and linear trend analyses

3.2

The power spectrllm level of t.he GP cable voltage varies with I.he local current. systcms in I.h(' iono-

sphcre and magnetosphere (FuJlI d.α1.， 1995). Fl1rlhermore I.hc vo1tage 101" 15 dαys shows good corre-

lation with the normal component.s of local geoma.gnetic fielcl varia.tion (GUM and MUT) as sho¥V1I in

Figure 1 I. The squared multiple coherence between t.hese observed variables keep high valllCS (........0.8) a.t

frequencies from 3.0x lQ-6 ........l.Ox 10-2/-{ z， ancl inclusion of the ot.her horizontal componcnt improvcd t.lw

coherency slightly a.t lower frequencies (Fig.l8). These suggest t.hal. t.lIC voltage is dOI川Ilat.edby the ex-

ternally induced field a1. the frequency band of Figure 18. Estimates of coherency rapidJy becall1C:川stable

at frequencies lo¥Vcr than 3.0x 10-6 H::， probably because t.he data segment contaIlls a. large gcorna..品目けIC

dist.urbance


~ 同〉

AU

11111Jm

rlij ヤ幽明』ω400

day

ハリ

T

l

n

U

司J-

〉

Time axis is in days frolll Figure 16: Unpowcred voltages of t.he GP (top) and GM (bottom) cables.

Sepもember1， 1993.

15

100 I ・ E

o Ft¥/'"-"'--'" γ -'V~ I It¥.v¥ 民八Aμh.M¥...〆¥..-r' v.ゅ"'¥.JへJ叫

-100ト y ，

o~r~内山内W'~f払川~ヘ

10

〉

-15 0

day 5

E圃ロ

15

100 IE~一-.---.

o f¥r""'-" '-"""-....---- """ 1.1I.AA 1ハcfrAlA川 N¥〆い〆いII/¥.JVV刈

-100ト Y' ，

AU day

5 。

トロ

15 10 5 。day

Figure 17: 15・da.yplot. oft.he GP voltage (top) a.nd the perpendicular geomagnet.ic fields at. Guam (middle)

and Muntinlupa (bott.om).

HAW-l

The voltage measurement. wiもhthe HAW-l cable stal'ted at自I・st.among the observations with decommis-

sioned submarine cables in 90's. There are several repor胎 ont.he features of this voltage (e.g. LANZEROTTJ

et al.， 1992， 1993; CHAVE et α1.， 1992b; LIZARRALDE et al.， 1995). These reports revealed that. the HAW-

1 voltage is dominat.ed by t.he externally induced field up t.o periods longer tha.n in the case of point.

sensors at the ocean bot.t.om (e.g. Lu叩

3.3

19 On Geoelectric POもentialVariat.iolls Ovcr a. Planetary S((¥I('

0.8

0.6

0.4

0.2

hu戸

ω』ωz

。U司ω』四四

σ的

O Ic-06 Ic-Ol

Figure 18: Squarecl coherency betwecn the G P vott.ag:e alld t，hc perpendicular geomagl1eLic ficld汚，，1， (:，¥M

f.l.nd MUT (solid line)1 alld bet.ween t.he GP voltage atJd 1.¥¥'0 component.s of the ltori"oJltal geomagllf'tI('

日貯IdsaもC:AMand MlI'l' (dashed ¥ine)

le-02 Ic-04 Ic-03 rrcqucncy (Hz)

le-D5

ell'C relatively weak. 'Wc focus 011 the weak motionally induced field IIl t.his study. The characll'rIsl irs of

ihc externally induced fie¥d on thc HAW-I voltage is ilJ L1ZAR1b¥LDιdα1. (1995)

The vo¥tage data consist or 20 min medialJs rol' 1'125 days rl'OJll Apri¥ i， 1990 1，0 March 2. 199'1

(Fig.19(a)). The data are dominated by ¥arge periodic signals which correspond 1，0 t.hc l，idc alJd Sq河川l

Its harmonics. Since these have di町erent.featurcs fro川 ol，herincluced signals (c.f. EGHERT (;/ al、1992)."

robt叫 least.squares fiiting w拙 app¥iedto the vo¥l.a品eat. frequencies of 7 major oceanir: Lides (Q 1， 01. P卜

1¥ 1.， N2， M2， S2)， 1-8 cpd， and up t.o 8 addil，ional sidcbands 011 boLh sidcs of Sq and harJl10ni四 1110/11

values were made by ¥owpass-filt.cring aft.er t ¥w sillllsoida¥ pt>riodic signals and t.ile Illeall wcrf;' r('IIIO¥l('<1

and data gaps shortcr than 3 11011.J''<; wcre int.erpo¥atcd by t 1悶 cubicsplil】e.The hOllrly volt.agぐ ('0川川IlS

16 daLa gaps ，vhose durat.ion are 0.1 '"'-8.6 days ilS sho¥¥'n ill Figure J9(b)

Thc gaps of the voltage should be filled， becClusc our int.erest is a.t. t，he 10llgest， tcrll1 signals of t 11(" dat.<l

sct. The procedure usC'd in this study was originally rcportcd by EGsE則、(1992トandt.hcn was appliC'd

to the firsl. year dal.a set 01' ¥'IAW-I by Cl1AVr. cl. ，，1. (1992b) and L1ZARRALD1， ct ，，1. (1995). Firs1.，

t，he t.ransfer fundion of a ga.ppy da.ta set 1.0 a rcfcre!lcc daLa set. is est.imated In the fl'e(llIf.."llcy dornaill

by t.he robust meLhod. Sccond， the transfer funct.ion is Lransferred t.o a. finiLe illlpulse respollst.' III 1.1]('

I.illlc domain wiU】 ast.at.ist.ical crit.erion which requil朗 that.a transfer fUllct，ioll of :¥11 implll附 r四 ponsp

is smooth and fits well wit.h the t.ransfer funcLion of t.he dat.;;1 In the frequency domain. FinallY.1I1時Slllg

dal.a. are computed lIsing the impulse respollse and the rcference daもa.'1'he procedllre、¥'aS<lpplicd to l.IH、

geomaglletic field or FRN rderring io t.he BOll data， to the flON dat.a I'ererrillg 1，0 I.he fillcd FH，N ，¥..1."

and t.hen to t.he HAもV-IvoJLage rcルrring1，0 the filled HON data. Figllre 19(c) show何 Lhehourlj、alllf'sor

IIA'vV-J with t.he gaps fillcd. This procedure is reasollable because (1) amount. of lIlissing dal.a同 SlIwll

(2) dllraιiOIl of ea.('h gap Is Ilot too long， and (:J) the gappy and referencp dat.a conelatf.." eacll ol，h川川

lcast'. a.t period band covering Lhc ga.p dura.t.ioll

20 I.Fujii and H.Uもada

20

〉。-20 。200 400 600 800 1000 1200 1400

day

20

〉 O

圃 20。200 400 600 800 1000 1200 1400 day

20

〉 O

-20 。200 400 600 800 1000 1200 1400 day

3

〉 O

-3 。200 400 600 800 1000 1200 1400 day

Figure 19: 1400-day plot ofthe HAW-l voltage from Apl'il7， 1990 to March 2，1994. (a) 20 lllin medians唱

(b) 1 hour values wit.h the t.ides removed， (c) 1 hour values wit.h t.he gap filled， and (d) 12 hour values ar~

shown from topもobottom.

12 hour values were computed from the filled hourly values using a lowpass filもerwit.h t.he cut.oπ

frequency 7.7xl0-6Hz. The 12 hour values in Figure 19(d) show large variatiolls illduced hy f.he ge-

omagnetic st.orms. Indeed， the multiplc squared coherence between the voltage alld three componcnts

of the geomagneもicfield aもHONsuggesもsthat the externally induced自elddominates t.he volt.age at.

frequencies higher thall 3.7x 10-2cpd (Fig.20). As mentioned in CHAVE et al. (J 99~b) ， t.he externally

illduced field is seell in the cable voltage up to a lower frequency band than in the electric自elddaもaby a

point sensor at the oceall bottom， because the sma.ll spat.ial sca.le componellts of t.he 1110もionallyinduced

volもageare filtered out by averaging over the cable length. The power spectra of the voltage in Figure 21， however， tend to increase with decreasing frequency at a lower frequency band， which was not obst~rved

21 011 Geoelectric Potential Variatio11s Over a Planetary Scale

0.8

0.6

0.2

0.4

-HEgdmM的hυロω』ωzoυ

o O.脱却l 0.1

Figure 20: Squared coherency between the HAW-l voltage and three components of t.he geomagnetic

field at HON for 1326 dαys.

0.01 F問 quency(c凶)

0.001

0.1

0.01

(豆諸2(Eゼ〉EZgu邑凶』ωき色

O.∞1 O.∞01 0.1

Figure 21: Power spectra of the HAW-1 voltage for 1326 days.

0.01 Frequency (cpd)

O.∞l

il1 CHAVE et al. (1992b). This feature indicates that there are the motional1y induced signals at. a lower

frequency balld， and thaもseparationof the motional signals from the external ones are necessary in a

highcr frequency band.


4 Externally induced field

4.1 Introduction

The externally induced fields dominate t.he GN a.nd GP voltages at. periods short.er tha.n a few dανs as

shown in Sect.ion 3. Since t.he GN and GP a.re laid on the Philippine Sea plate ill roughJy llort，h-sout，h

and east.-west direct.ions， respectively， the spatially a.vera.ged elecもricalconductivit.・2・distriblltionof f，he

plate is possibly obtained by applying the MT method (CAGNIARD， 1953).

The conductivity distribution of the Philippine Sea pla.te is kllowll less t.ha.n that， of t，he Pa.cifk plat.C:'.

SeveraI experimellts with point， sensors were ca.rried out. in mal'gina.l areas. FILLOUX (1983) shldied t.I1(、

Mariana area and gave a lD model with a.n unusual thick resistive layer at the fOl'earc (Fig.4.1(a.)).

UTADA (1987) and TOH (1993) presented cross-sect.ion models of the [zu・BoninArc aud conclllded tha.t.

the Philippine Sea plate aももheage of about 30 1¥1α(KARIG， 1975) h回もhinllerlithosphere t.ha.n t，he

Pacific plate (Fig.23). UTADA (1987) also revealed a high conductivity materia.l subducting wit，h the

Pacific plate under the Honshu island， J apan，もhatis， the electric current. leakage a.ももheboundary

between the island and resistive lithosphere. TOH (1993) found a conductive part ju坑 beneaιht.1lt' island

arc except for the conductive pat，h a.long the Pa.cific plate and implied a. partial melting. SHIMAI¥A¥oVA

AND HONKURA (1991) reported the result， from the Ryukyu t，rough. On the 0もherha.nd， models of t.he

Pacific plate were welI developed. For il1stal1ce， lD models of the Pacific plate revealed t.hat thc t.hickness

of t.he lithosphere (resistive layer) getsもhickerwit，h incre硲 ingage as shown in Figlll'e 4.1(b) (FILLOUX，

1967. 1980， 1982; OLDENBERG， 1981; YUKUTAKE et al.， 1983; OLDENBERG ct al.， 1984; TARITS， 198()).

St，ructures of mid-oceanic ridges wel'e investigat.ed in several t，imes a.t t，he J ua.n de Fuca. Ridg~ ( LA W AND

GREENHOUSE， 1981; HEINSON et al.， 1993). MACKIE et al. (1988) fOllnd t.he lithosphere of an f.'xtremely

high resistivit，y-thickness product a.t the off..，hore of the North America.. The result， of LIZARRALDE el al.

(1995) s叩 portedexistence of that layer with an electric current. short and also revealed that， the upper

mantle beneathもheresistive lithosphere is more conductive than t.hat beneathもheNorth An】erica.

The spatially averaged model of the PhiIippine Sea. plat.e using the GN a.nd GP voltages a.nd t，he

geomaglletic fields at KAK， GAM， and MUT will be present.ed in

23 On Geoelect.ric Potential VariaもionsOver a Planet.ary Scale

101

O

100

200

OMA OMA ..

1135 MA --_._--700

800 10-4 1σ3 1σ2 1σ1 100

Conductivity (S/m) 101

O

300

500

400

600

ヘミ之さえ句。

r---Trough一一一o rea rc

100

200

600

300

400

500

ベミ之さ

bQ

700

800 1ぴ410-3 1σ2 10-1 100

Conductiviη(S/m)

Figure 22: lD conductivity profiles of the Philippine Sea Plate (le乱)and the Pacific plate (right.). The

profiles of the Mariana trough (solid line) and forearc (dashed line) in the left panel were reproduced

from FILLOUX (1983)， and those of the Pacifi(' rise (0 MA， solid line)， the Sea ftoor revisited (30 MA.

d鎚 hedline)， and the Sanriku offshore (135 MA， dott.ed and dashed line) were reproduced from FILLOUX

(1967， 1980， 1982) and YUKUTAKE et al. (1983).

Spatial and temporal variations of the geomagnetic field

Continuous data segments longer than 7 dαys were selected from the volt.ages and geomagnetic fields from

June，1992 to August， 1995 (Sept.ember， 1993 to August， 1995 for the GP cable) for the MT analysis.

Total data lengths used in the MT analysis are about 192 dαys for the GN cable and about 52 dαys

forもheGP cable dueもothe presence of the data gaps. The geomagnetic field variations of these data

segments were examined in this section.

The ratio of the geomagnetic fields was computed as follows:

4.2

‘ ... E'

、，‘，••

・・且4'E

‘、

、‘、‘..••.

，，，，

q-E--9auH

H町

H

J'tEEE¥

¥11E

，Ff

AZ

ム

ム

-

-zz

/，

tEE1¥

一一、、EE』，ノ

Hl H，~

11

Hl = (Hl， HIP and H2 = (Hl， HIP are恥 geomagneticfields at. GAM a.nd KAK for the GN cable， and at MUT and GAM for the GP cable.

Zムム andZIIII of the whole data sets were computed byもherobust procedure for the t.wo cases (Fig.24)・

The TBW w節目xedas 1 to avoid contamination of frequencies. Frequency dependencies are seen in


O

0.002 S/m 50

。。~ 0.002 S/m

150

nu

-nU

FhJ

AU

nu うん

-250 (km)

250 500

Figul'e 23: Cross section of the Izu-Bonin arc at. 300N. Reproduced from TOH (1993).

the amplitude rather than in the phase， alt.hough diπerencies are less than a fact.or of 2. The ratio is

roughly constanもatfrequencies higher than 7x 10-5 H z， and disturbed aもlowerfrequencies， especially at

frequellcies ofもheもides，Sq and its harmonics. The range of the amplitude ratio in the case of GAM and

KAK is 0.2.....1.7 for the whole frequencies and 0.3.....0.8 for frequencies except for the tide and SQ. 111 the

other case， it is 0.2.....1.5， and then 0.3.....1.5 aft.er the periodic frequencies are removed. This indicates that

the geomagnetic field gradually varies over the cable lengもhs(~ 3000km) and that a spatial 1ineality of

the geomagnetic variation is roughly satisfied excepもforthe particular periodic signals.

The transfer functions of the GN data set were limited at frequency higher thall 4x 10-6 H z dueもot.he

lengt.hs of the data segments. On t.he other hand， t.hose of the GP da.ta set. were obtained at. fl'equencies

down to 6x 10-6 H z. This difference between the GN and GP dat.a set comes from the data length alld

the locality of geomagnetic variations. Large geomagnetic disturbances were conもainedin four out of five

data segments for GAM and MUT. Local variaもionsof the geomagnetic field at MUT are possible because

the geomaglletic latitude of MUT is 30 which is within the latitude band of the equatorial elec.trojet while

that of GAM is 90 where is out of the laもitudeband of the equatrial electrojet.

Two data segments were selected for the two cable data sets in order to examine temporal variations of

the geomagnetic field. One segmenもconsist.sof magnetically quiet days， andもheother contains disturbcd

days (Fig.25 and 26). Zムム，ZIIII and squared ('.oherence of each segment are shown in Figures 27.....28.

25 On Geoelectric Pot.el1tial Variations Over a Planetary Scale

2

1.5

0.5

(冨〈O戸DE)Eoz

pe中endicul訂一一

parallel ........ ー・

2

1.5

0.5

(凶〈凶言〈

O)goz

le・5 le-4 le・3 le・2Frequency (Hz)

O. le-6

0 le-6 le・2

180 perpendicular .-.-ー

paralleト.，，-吋pe中endicul訂日ー『

parallelト Hー

90

。

(ωω昆ω司.2〈OPD2)ω詔昌弘

le・5 le・4 le・3Frequency (Hz)

J報掲.it'暗闘脚"

180

90

。

(ω巴凶ω司，凶〈凶、宮〈

O)ω話回向角田

-90 -90

le・2le・5 le-4 le-3 Frequency (Hz)

-180 le・6le・2le・5 le・4 le・3

Frequency (Hz)

-180 le-6

Figure 24: Norm (top) and phase (bottom) of the ratio bet.ween the geomagnetic field variat.ioll for tl開

GN (left) and GP (righも)cables.

The coherence of the disもurbeddata increases in the normal and parallel components for t.he t.wo cahles.

Theもemporaldifferences between GAM and KAK are withinもheerror bars except frequencies from

4xl0-4もolxl0-3Hz， and differences between GAM and MUT are largerもhant.hose between GAM and

KAK. These suggest that the geomagnetic fields of GAM and KAK vary in a simila.r way， and t.hat.もhe

geomagnetic field of eiもherGAM or MUT contains local variations.

1∞「・・・・・一一一..------.soトo 1r，.J肺~、A.I'Ý~倒神恥伊~þ山明〆個4

・soド' ・-・100~'--~--~~--~--~--~~

o s 也y

トE

トー一一~一…|ト-E

“)() ，....---，-一一 .-一一一・so ~今 ... ‘.皐.!.牟o除叫向叫Aah必拘山AAJ包Jvtj刷ι州帽V回栴ず

-50ト'-100 I ・・・.

o 5 dlly

ー-E

100 I ・・・・・50ト [，ーー・・、 4 ;¥ /'.ー・ぺh ・1

I. ...r.. .....-. L~ .1 l.. & [-.... .-:-ム~ !':..句、.__I 0 仇F'侃~~、が恥，勺伊野町、炉、\ 1;1'-"、.JP'-司、掛?¥'i

・sot= ・1∞

o S 10 day

ドロ

Figure 25: Normal (top) and parallel (bottom) componenもsof the geomagnetic field aもKAK(solid line)

and GAM (dashed line) for quiet (right) and disturbed (left) days.


乍 1伺「十十十十三]

=:ト十一一rl乍 Ir~十??!

宅 ltmr一一~γj

Figure 26: Normal (t.op) and pa1'allel (boもtom)componenもsof the geomagnetic field at GAM (solid line)

alld MUT (dashed line) fo1' quiet (right) and disturbed (left) days.

4.3 Analysis

The 1'elationships between the cable voltage a.nd horizontal geomagnetic field ma.y be writt.en in th~

frequency domain as follows: V(ω) 一了 =A(ω)Hム(ω)+ B(ω)HII(ω) ，

V(ω) _ ，4/，.¥ II 1..¥ I 0/..¥ II 1..¥ I Z(ω) 一了 =A(ω)Hょ(ω)+B(ω)HII(ω)+一了I(ω)•

(14)

(15)

whe1'e V，L，Hム，HIIare the volもage，もhelengもhof the cable， and t.he pe1'pendicular alld parallel geomag-

netic field， respectively， and A and B are t.he t1'ansfer funcもlons.ω，1and Z denot.e t.he a.ngular frequency，

the cu1'1'ent andもheimpedance of the circuit.. Equation 14 is for the case of au unpowered cable and

equat.ion 15 is fo1' a powe1'ed cable. The last term of equat.ion 15 is necessary in ol'de1' t.o r~ll1ove t.he

influellce of the CU1'l'ent fluct.uaもionsf1'om estimates of A and B. -Supposing V/L. Hょa.ndHII a1'e geoelc('-

t1'omagnetic :fields which a1'e averaged ove1'もhecable lengt.h， A a.nd B reflect an avera.ged magnetotelluric

impeda.nce which cont.ains info1'mation on t.he dist1'ibution of elect1'ical cOllductivity beneath the a1'ea.

The MT impedance is often t1'eated in the fo1'm of an appa1'ent resistivity and a phase. The a.ppa.l'ent.

resist.ivity pa fo1' A is as: dA一州一一a

ρ'

(1<5)

whe1'e μdenotes magnetic pe1'meability.

The impeda.nce， however， is diffe1'enもfromone in o1'dina.ry MT su1'veys， being t.he 1'atio of elect.ric

field at the bottom of a conducting sea and the magnetic field obse1'ved on t.he earth 's su1'face. Sincp. the

cable length is much larger thanもhesea depth， the paths between the sea surface and the sea bottom ca.n

be igno1'ed. That is， the cable volもageis treated as a value observed at the ocean bottom. We considf'!r

a simple siもuationas a first step and assume a one-dimensional st1'ati:fied eart.h whose top layer is t.he

ocean， and a geomagnetic field linearly va1'ying f1'om one end to the othe1'. In this case， the geomagnet.ic

field is actually equivalentもoa uniform one which is averaged between t.wo stations. One dimensionalit.y

is suggested by the fad. thaもt.hevoltage is highly coherellt with the pel'pelldicular geomagnet.ic field in

Sectioll 3.

The data from Novembe1'， 1993 to Ap1'il， 1994 were used fo1' the analysis at t.he high frequency band，

because the clock was accura除lyadjustedもothe Greenwich Time at this pe1'iod. The whole da.ta sets

described inもheprevious section was used fo1' t.he analysis at the low f1'equency balld where high aCCUl'acy

is not required.

27 On Geoelectric Potential Variations Over a PlaneもaryScale

2 2 qUlet・。

disturbed →-~

hAH

，J

偽μ市lt油w.

1.5

0.5

(凶〈凶、宮〈

O)E』

OZ

qUlet ら+ーー・

disturbed →ー

--F -a

dvp

Z--mw

時前四噌

空

--z--・均.‘ .••

1U似WHr

yITEaT&EEt

?I甲EATEE--ZEE-‘

官

E'AVa--TEABYEa

1.5

0.5

(凶〈凶泊者。

)Eoz

Ie・5 le・4 le・3Frequency (Hz)

O le・6

O le-6 le-2 le-2 Ie・5 le-4 le・3

Frequency (Hz)

180 qUlet・4

disturbed →ー

1I持・p・i，a"円

90

。

-90

-180 le-6

(ωω』凶

ω司，凶〈凶旨4〈O)ω沼SL

180 qUlet ~・一一・

disturbed →-<

胸a

・m

・-Lm za'

.由なT

ZO

舎・1A

主同TTE--h

h祁H仙町

I晶

i'e

T-zmT‘

90

。

弓

3戸M

Z

H

4+VJ

4

C

山

em

u

u

L

n

u

a

』

戸

』

-

F

E

4

E

a

一ζJ

出

e

-90

-180 le・6

(ωω』凶

ω唱.M〈U向、言〈O)ωω6Z角田

le・2

附

.Mdnt

uhM川

ιi

G4凶“げrllA

t、121'l

同情

Vl

bh仙川仰いl

い帥MWHUllφ'ill-

le・5 le・4 le・3Frequency (Hz)

0.8

0.6

0.2

0.4

hυロω』ωaouyw』

gdvω

le・2

0.8

0.6

0.2

0.4

hυ

ロ巴ωsou-vo』

gdvm

le・2le-5 le・4 le・3Frequency (Hz)

O le・6le・2le・5 le-4 le-3

F陀quency(Hz)

O le-6

Figure 27: Comparison of the geomagnetic responses in quiet (diamond) and dおもurbed(square) days.

Norm (top) and phase (middle) of the ratio between the geomagnetic field variation for the GN， and

squared coherency (bottom). Left and right panels are for the perpendicular and parallel components，

respectively.


2 2

qUlet .-・-・disturbed →ー

1.5 (冨〈OPDE)E』

OZ

quiet 0--+←4

1.5 (冨〈

OPD2)Ehoz

0.5 0.5

0 le・6

O le-6 le・2le・5 le・4 le・3

Frequency (Hz) le・2le・5 le-4 le・3

Frequency (Hz)

HTEA-EA

-A'A

• ，.vaEBEBEE--

• 180 180

90

。

-90

(g』凶

ω唱。苫〈

OPD2)ωaE

quíet 吋-~~.

disturbed --ー・

;ぬ;ゐi結22

90

。

-90

(ωω』凶

ω-M.2〈OPPE)ωお記

-180 le・6

-180 le-6 le-2 le・5 le・4 le・3

Frequency (Hz) le-2 le・5 le・4 le・3

Frequency (Hz)

qUlet -・---<

0.8

0.6

0.4

hυRukwsou-xw』

62vm

:~!rt附工0.8

0.6

0.4

hυggωaoU司

ω』62vm 0.2 0.2

le・2le・5 le・4 le・3Frequency (Hz)

0 le・6le・2le・5 le・4 le・3

Frequency (Hz)

O le・6

Figure 28: Comparison of the geomagnetic responses in quieも(diamond)and disturbed (square) days.

Norm (top) and phase (middle) ofもheratio betweenもhegeomagnetic field variation for the GP， and

squared coherency (bottom). Left and right panels are for the perpendicular and parallel compol1ent札

respectively.

29 On Geoelectric Potent.ial Variations Over a PlaneもarySca.le

The MT responses of the GN and GP cables wiもhjackknifed errors were calcul抗edusing a robust

procedure reported by CHAVE AND THOMSON (1989) and THOMSON AND CHAVE (1991) shown in Fig川 e

29. The TBW w槌 4for the GN spectra and a high frequency part of the GP spectra. The low frequellcy

part of the G P sper.tra w部 obta.inedwith the TBW 1 because of limited data. Several examples of the

resolution is indicated in the upper left panel. In the GN cable， the estimates from 10-3 to 5x 10-6 H z

were obtaiued from the powered data for 4，...，13 months， andもhosefrom 5x 10-4 to 10-3 H z were fromもhe

ullpowered voltage. Spectra which include Sq and its harmonics (24， 12， 8， 6， 4.8， 4 hOU1'S) and the four

major oceanic tides (01， Kl， M2， 82) withinもheirresolution ballds were removed. The longest. period of

もheresponse is abouも2dαysfor both cables， because of the currenもnoisealld limited data for the G N

and GP cables， respectively.

ar

--、事441T+11

、等、

¥

抽

可

¥

伊

晶

、

a咋..

、臨唱

wfi

e

-有斗11i

HFilいい

le+2 ~

le+l h

le+O ~

EREFFIr-LEE

EEq--

--

e

e

---ae--且

le・3~

(EEzsbτ328出冨

Eaw内置〈

¥.....、.. Qf¥

t ~・

le+2

le+O

le-2

le・3

le+l

le-l

(ggashZ〉宮的冒

ω出冨

ω昆向日仏〈

le・4h le-4

le・5le-5

le・6le-5

le・6 le-l le・2le・4 le・3Frequency (Hz)

le・1le・2le・4 le-3 Frequency (Hz)

le・5

ーI~I L

匂

台、、・ι ・錨'

7 崎、モミ1、

180

120

60

。-60

(ωω』凶

ω司)ωおS仏

...

.... h晶画E

『司悟崎・・h ・L温..... -.

... 180

120

60

。-60

(ω巴凶ω唱)ωω雪山内田

-120 -120

le-I le・2le・4 le・3Frequency (Hz)

le-5 -180

le・6le-l le・2le-4 le・3Frequency (Hz)

le・5-180

le・6

Figure 29: Apparent resistivity (top) and phase (boもtom)of the GN(left) and GP(right) cables.

The apparenもresistivityand 胞 phasein Figure 29 give usもwoIlnplications. Firsも， we can focus

011 A as the MT response， because B is much smaller and less sもableもhanA. That ensures the lD

assumption. Second， a curve of the powered data smoothly connecもsto one of the unpowered data ill

the phase， however they are discontinuous for the apparent resistivity at a cross over point beもweent.ll(ヲ

powered and unpowered spectra is around 10-3 H z in the GN cable. This is perhaps because a spectrum

of the unpowered voltage at a lower frequency has a wide resolution band， and an illtegration of lOllg term signals cause a bias. Therefore，' we made a MT impedance curve of the A using the powered result.s

I.Fujii and ll.Utada 30

at frequency lower than 10-3 H z and the unpowered results at the higher frequencies. Smooth conllect.ioll

of the powered voltage to the unpowered one supports that t.he currenもnoiseis e町ectivelyremoved in

computing Equation 15.

Interpretation 4.4

lD illversion

The optimull1 conductivity struct.ures inferred from A of t，he GN and GP cable were obt.ained as ShOWll in

Figure 30 USillg a Monte-Carlo inversion dcveloped by UTADA (1987). We calculated 3・6layered lllodels

for several init.ial values for each la.yered model. The 3・layeredll10dels gave t，he minimum variauce for the

GN cable， although the 1'esistivit.y of the secolld layer saturated aもvalueshigher t.han about 4000 Qm.

The solid lille of Figure 30 showsもheoptirnum 3占Lye1'edmodel. We plot.ted 2.46x10-4S/m (4063 Qm) as

the second layer's resistivity， however lower conductivit.y values or higher resist.ivity values arc eqllivalent.

On the other hand， the variance between t.he observed and ll10deled respOllses of the GP cable decreased

with an increasing nUll1ber of layers. We determined the optimum model by calculating t.he ll1illimum

point of AIC (AKAIl也， 1973). This mcans t.hat t.he simplesもandbest.鮎 modelwas selected. The dashed

line of Figure 30 shows the opt.imum 5・layeredll10del thaもwasselected 加もhebest.. 1もisclear t.ha tもht'

MT 1'esponses of the optimum models 1'easonably fit the obse1'ved ones as shown ill Figure 31.

4.4.1

。 ...........，・・・..................................................

100

200

{EUCSHaω凸

3α} ............1

z ......................................................，

400 0.0001 10

Figure 30: One-dimensional conduct.ivity dist.1'ibuもionbetween Ninomiya and Guall1 (solid 1ine) and Baler

and Guam (dashed line).

0.01 0.1 Conductivity (S/m)

0.001

Roughly speaking， the conductivity moclels of t.hese t.wo cables have a commOll feat.ure that a high

conduct.ive layer (HCL) whose conductivity is about. 10-1S/m. follows a low conduct.ivity layer (LCL)

beneath the sea 8001'. This genera.1 feature has been found in a. number of studies onもhesubmal'Ine

conduct.ivity struct.u1'e by using poinもsensors.The fa.ct t.hatもhelal'ge-sc.ale avera.ged profile is similar to

loc.al Olles suggests t.hat. it is a common strucもureover the ocean.

The moclels of the GN and GP cables， however， show significa.nt differences. For instance， the HCL of

t.he GN cable locat.es at the depもhof 80 km with e1'ror bars frol11 59 t.o 110 km. This is about 1.5-2 times

31 On Geoelectric Potential Variations Over a Planet.ary Scale

GP .-可-rrTTn1"""一ーす τ rrTT宮町一

observation --→ optimum model .. ...

100

nd

-m出

mm

町

wm

N?ωm

G門

戸

。

1∞

世ー，β al.・l.

¥ 't.・ξ.

、宅bゆ令、'. 、. . . . 、.

a e

・-. L!...

le-2

tzトE

AU

0.1 ~

(EEZO)bτ志一む』冨ω局内芭〈

-----

a'-z

&E '

晶守

az

• 弘、・e. -v

• .•• ・¥.可、-

h町-e s-

h

，

.E・τ&v

・・10

0.1

(EEZO)kC一〉宮一8ZS旬怠〈

le-I le-4 le・3Frequency (Hz)

le-5 0.01

le-6

胴刷制副

le・1le-2 le-4 le・3Frl叫uency(Hz)

le・50.01

le-6

observation -....-. optimum model"…

..・・90 observation →ー

optimum model

90

三、・A

・、.、

¥

・、¥

量亭‘

、缶、‘

-丸子

60

30

。

(ωω』凶

ω唱)ω泊記

liih -、

町￥吋~ir.

W九九

60

30

。

(ωω』凶

ω唱)ωaa内向

le-I le・2le・4 le・3Frequency (Hz)

le・5

-30

-60 le・6le・1le・2le・4 Ie・3

F陀 quency(Hz)

..LU....-

Ie・5

-30

-60 Ie・6

Figure 31: Apparent. resistivity (top) and phase lag (bot.tom) at. t.he sea. surface calculat.ed from t.he ca.blぃ

voltage (diamond) and t.he optimum conduct.ivity model (solid lil1e) for the GN (le仇)and G P (right.)

cable. Error bar del10tes two standa.rd errors of t.he mean for each diagram.

deeper than the previous reports (UTADA， 1987; TOH， }993) al1d 部 deepas the Pacific plate at. t.he same

age (Fig.4.1(b)). On the other hand， the profile ofthe GP cable has an extremely thick LCL. It is IlIlU訓lal，

if t.he LCL corresponds to lithosphere as infel'red from previol1s works. A similar profile was obtaillf'd only

in the Mariana forearc from an ocean bottom EM observation by Filloux (1983) (Fig.4.1(a)). Differences

also exist in the shallower parts. The first layer of the GN model (3 km.， 2.8 S/m.) is approxima.t.ely

consist.el1t with the sea along the GN rOl1te while t.hat of t.he GP model is very resist.ive ("，，1 S'jm) a.nd

deep ("，，10 km) and accompanies a less conductive layer (0.3 S'jm.， 13 Æ~m.) beneat.h it..

A ql1estion arises about the causes of t.he differences. If they do not. reflect a. real diπerence ill tllt~

conductivity structure bet.ween the two regions， there may beもhreecandidates: (1) bathymet.ry， (2)

land-sea distribution， and (3) a subducting slab. The difference of sea depth affects the MT responses

observed at t.he sea surface. The GN cable rl1ns along the Izu-Bonil1-Mariana art' at t.he west.ern edgc of

the Philippil1e Sea plate and the bat.hymet.ry is approximat.ely uniform (勾 3k1n)， while t.he sea dept.h

alollg t.he GP cable is deeper (5......6 km.) and challges sharply at the Mariana trough， t.he Palcs V~ra ridgc

alld the Philippine trellch.

Enhallcemellt of the electric current at the boul1dary of sea and lalld due to sha.rp cont.rast of t.he


elect.rical conductivity between sea water (3......4 31m) and rock (:50.1 31m) is well known. Since t.he GN

and GP cables connect to Honshu island， Japan and Luzon island， Philippines， respect.ively， t.he cable volta.ges might be distorted at the shol'e region. The amoullもofthedisもortionfor the G P cable is expected

to be larger， beca.use the deepel' watel' depth would produc.e a larger disもortion.

The subducting Pacific plaもealong the Izu-Mariana subduction zOlle which is almost paral1el to t.he

GN cable could be another cause for the differences of t.he responses. The cold (resistive) slab could

increase the apparent resist.ivity of the GN cable especially at low frequencies， although it would not.

affect much the MT responses of the G P cable.

These three candidates were examined in order in the next two section. The effect.s of the bathymet.l'Y

alld land-sea distribution derive from a shallow part， and therefore we estimate them by a t.hin sheet

modeling (MAKIRDY et al.， 1985). A 2D modeling (UTADA， 1987) was applied to estimate the slab effect

because inhomogeneity in a deep part ca.n not be included into a thin sheet. modcling.

4.4.2 Thin sheet nlodeling

At low fl'equencies， three dimensional distribution of conductivity aもtheEarth 's surface can be approxi-

mated by a thin sheet with laterally het.erogeneous conduct.ance， I.e.， depth integrated conductivity. The

approach by MA KIRDY et al. (1985) bases on the thin sheeもapproximationof PRICE (1949) and treats

a bimodal induction with a boundary condiもionof the Neumann type and a 1 D sもratifiedEarth beneath

the sheet. This method can deal with the most complicated situationもhanother appl'oaches (e.g.， SASAJ，

1968; HONKURA， 1971; VESSEUR AND WEIDELT， 1977; DAWSON AND WEAVER， 1979)， therefore it is more appropriat.e fol' a complex land-sea distribution such部 thePhilippine Sea and its vicinit.y.

The thin sheet approximation hy PRICE (1949) is writ.ten in a lD case e¥S follows:

H(a， y， -0) -H(x， y， +0) =ァ(a，y)E(x， y， 0) (17)

where

T(川)=fσ川け (18)

where x， y and z denote a locat.ion in the Cartesian coordinates， H and E are the geomagnet.ic and

geoelect.ric fields， and T denotes the conductance. The Z direction is dowIlward positive， and -0 and +0

denote the bottom and top surfaces of the sheet. Equation 17 is an approximated form of the integrated

Ampere 's law from z = 0 to z = D. Since the thickness D is so small， the electric field was approximat.ed

as a vertically constant wit.hin D. This approximation is satisfied at. long periods where skill dept.h is

much larger t.han D， for inst.ance， periods longerもhanseveral decades of millutes.

The Philippine Sea is one of the deepest sea in t.he world. This fact requires about 10 km. as t.he

thin sheet thickness D. 1もisunclear whether the thin sheet approximation is satisfied at a period ha.nd

of the present concern. YAMAMOTO et al. (1989) proposed a correction to the approxima.tion. They

involved att.elluation of the elect.ric field in the sheet by usingもhelD t.heoretical electric field E1D (ιy，:)

as follows:

I'D H(x，y，-O) -H(x，y，+O) = Iσ(x， y， Z)E1D(X， y， z)(lz

rD _1_ ~~ ~.\T."I ，~~ _. _¥ElD(X， y， 0) = Iσ(x， y， z)E1D(a!， y， z)

ん E1D(X，y， 0)

33 On Geoelectric Potential VariaもionsOver a Planetary Scale

rD _1 _ _¥ ElD(X， y， z)

E1D(X， y， 0) Iσ(X，.y， z) ん EID(X，y， 0)

TeEID(X， y， 0)

一一

(19)

rD _， _ __ _¥ ElD(X，y，Z) TC(X， y) = Iσ(z，u z ) ' d z ん， -，E1D(X，y，0)

EID(X， y， 0) in Equation 19 is equivalent to E(x， y， 0) in Equation 17. In a two layered struct.ure， t.he electric and magnetic fields in the top layer whoseもhicknessis d are explicitly writ.ten as:

(20)

一一

where

(21) 4仰 -k1d{(1+与)出pko(d-け+(1ー与)ほp-w-z}}，~ ~o 応O

ikoA叫 -k1d{引与)e前向+(1-h)exp-ko{d-z}}・2wμ 応o 応。

一一EID(Z)

(22) 一一H1D(Z)

(23)

(24)

viwp石，.J[:;戸;

一一k。

where

一一

A is a scale constant，σ0， and σ1 are the electrical conductiviもyof t.he first layer， and that of the second layer， respectively.

The e宵ectof this correction is demonstrated in Figure 32. The ra.tios of the impedance of t.he thin

sheet approximation to th抗 ofthe1D theory were calculated at periods from 1 min t.o 10 dαys in t.he cases

of D = d = 0.1， 2， and 10 km. The conductivities were fixed 鎚 σ0=3 お 33 δ8/m a制削n吋dσ町1= 0，(ρ訓01"汗S'jρ111η』The thin sh陀eetimpedance E(z = 0的)/H(z= +0的)wit.hout も“hecorrection shows s討ign凶tif自icallt.dωif町f(。何灯m、刈T明rt可?明骨e引刊子汁nce何S

from E1D(やz= 0)/ H1D(Z = 0) at shorter periods both inもhenorm and ill t.he phase. Those differences

disappear when E1D(Z = 0)/ H(z = +0) with the correction is used. Underestimation in Equat.ion 17 is improved much even if t.he sea depth is as deep部 10km like the Mariana trench.

k1

O.lkm 0.1 km (correcled)一一

2km 2km (correcled)・

10km 10km (correcled)…・

. .

. . . . ，

. . . . . . . .

-・・・・・""'"ます~・-.-;..・2・・，，，，，，，，，，，，，，，，，，，，，，，，，，

10

。-10

-20

-30

(ωω勧告.0ヨωω号旦告

)ωSω沼田E

O.lkm 一一O.lkm (con倒的一一

2km 2km (co町田ted)・

10km IOkm (co口町led)…・・・

.....___..~ .. _:--;:._J ・・・・・・・・・・Tロー・・・・・・

. ' ・，

. . . . . . . . . . . . . .

. ， . . . . ，

. . . . . . ， . . . . . ，

0.5

{ロ-3ωωS2zg)O宮MM

-40

-50 0.01

o 0.01 1000

Figure 32: Norm (left) and phase (right) ofもheratio of the impedance (thin line)もothat with the

correction (thick line). The sea depもhis 0.1 km (solid line)， 2 km (dashed line)， and 10 km. (dotted line).

100 10 Period (hour)

0.1 ω∞ l∞ 10 Period (hour)

0.1


The thin sheet modeling with the correctiol1 was appliedもothe MT analysis of the G N a.nd (~p ca.blc、s.

H(x，払ー0)，H(x， y， +0)， and E(x， y， 0) were computed in the geographic X and Y polarizaf.ioll mod開

a.t periods 30 min， 1 11.，2 11.，3 h， and 5 h. The t.hickness of the thin sheet is 10 km. Thcミa.rea.COVNS

5000 bn x 5000 km (00 __450N， 1100 --1550E) using a. 72x72 grid for 30 rnin， while 6800 km x sHOO km

(_80 __500 N， 1020 --1600 E) using a 68 x 68 grid for the other periods. The conduct.ance dist.ribution of t.h台

t.hin sheet was comput.ed from TUG87 dat.a set， which consists of a 5'x!γreal topogl'aphy， aft.er averaging

over a. grid size as follows:

ア(x，y) =σwd(x， y) +σr(D -d(x， y)) (25)

whereσωand σ，. denote the elect.rical conductivities of sea watel' a.nd rock， respect.ively， and d(x， y)

denotes an averaged sea depth at the point (x， y).σω，σr， and D a.re 0.333 nm， 100 nm， and 10 km.

respect.ively， in our case. Figure 33 shows the conductance distribut.ion of t.he Philippine Sea elnd it.s

vicinity using a 68 x 68 grid.

Four 1 D models be附 aththe sheet. were exa.mined (Ta.ble 2). The models GN and GP we肘 made

from the lD opt.imllm models of t.hc GN and GP cables (Fig.30) aft.er開 movingthe first la.yer which

corresponds to the sea.. The model 1 represenもsa uniform upper mant.lc. A thin layer (2 krll， 100 nm，

・notshown in Tablc 2) was added following the thin sheet in all models in ordel' to avoid ohservillg an

enhancemellt of t.he electric current. due t.o a sharp conductiviもycontrast at. t.he bot.tom of the t.hin sheet..

which is not. ollr intention.

The 2D MT relat.ion on a grid may be writt.en as follows:

(:)=(;;:引)( ~~ ) (26)

The impeclance tensor Z was computecl along the cable after rotating E ancl H of t.he geogl'aphic coor-

dinate into our coorclinate sys旬m.Z is equivalent to one observed by a point sensor at. t.he sea surface if

the thin sheet assumption is sat.isfied. Figure 34 shows a pro自leof ZIIム alongthe G N ancl G P cables at

30 miηto 5 h in t.he case of Model G N. Znムbac;ic.allyvaries with topogl'aphy， t.herefore， t.he respons{' of

the GP cable is smaller thal1 one of the GN cable. Large enhancements of the amplit.ude occur arouncl

Honshu island (340 -- 350N) and LU;l，on island (1210

-- 1220E). The amplitude of the enhancement. is

larger for the GP cable t.han thaもforGN cable， ancl is larger at higher frequencies.

Equation 26 Is int.egrated along the cable for the cable voltage.

v = LB

EUds

Table 2: lD sもructuresbeneath the thin sheet

Model GN Model GP Modell Model2

d (km) p (nm) d (km) p (nm) d (km) p (nm) d (km) p (nm)

80 4000 1 :~ 3 600 100 10 LOO

6.5 300 2500 80 1000

50 500 6.!>

On Geoelectric Pot.ential Varia.t.ions Over a. Planelary Scale 35

120' 130' 140・ 150・ 160・

O'

50・50・

40' 40・

30' 30'

20・ 20'

10' 10・

O'

110・ 120・ 130' 140' 150・ 160・

圃Eコロ圃圃

loglQ CONDUCTANCE

Figure 3:3: Conductance dislribution of the area frol11 _80 lo 520N and 1020 1.0 J620E. ''1'1¥1:> ('0101' sca.le is

shown at. thc boltol11

，8

= I (ZIIょJJょ+ZIIIIHII)ds

l¥ and B are い¥'0end points of t.he cable. Equat.ion 27 i会approximatedas [0110¥¥'5

(27)

EII = ZIIL1h + ZIIIIJJII (28)

where

t||=j乞EII(i) (29)

)

)

2

7

(

よf

f

+

)

ー(

ムJ

J

(

EI-n4

一一i

J

I

(30)


0.5h _.. lh ....…-2h ...0・

3h" 5h 司申・・・

1000

1∞

10

(EEZO)b宮宮28M冨ω局内野凶〈

0.5h→ー

lh吋…・

2h ...0…・

3h 一件→

5hι ー

1000

100

10

(EEZO)bτ一怠ωω出富ω諸島内同〈

0.1 120

0.1 10 150 145 130 135 140

Longitude (degree) 125 40 35 20 25 30

Latitude (degree) 15

0.5h -+，一

Ih ....…・

2h 3h " 5h・4 …

。-20

-40 0.5h --一Ih一…….目.

2h ... 3h 苧咽一一一5h ...←.一….一.

180

160

140

-60

-80

-100

-120

-140

(8』凶ω唱)ωSE

120

80

60

40

1∞ (ω@』凶

ω司)ωω何回白色

150 145 130 135 140 Longitude (degree)

125

-160

-180 120 40 35 20 25 30

Latitude (degree) 15

20

0 10

Figure 34: Apparenもresistivity(top) and phase (bottom) along the GN (left) and GP (right.) ca.bles for

Model GN.

di|=j日11(1)+ Hn(η)) i denotes the grid number along the cable， and grids for i = 1 and i = n locate Guam and Ninomiya， and Baler， respectively. The ZIIょvaluesof Model GN and GP are compared with the observed ones in

Figure 35. The results of Model GP are greatly different from the observed responses， although t.he 1 D

optimum model of the GP cable reasonably reproduces the observed GP response (Fig.30). This suggest.s

もhatthe topography and land-sea distribution considerably affecもtheGP's observed response and cause a

nominal thick LCL in the 1D model. Onもheother hand，もheresul同 ofModel GN roughly reproduce hoth

the GN and GP observed responses. This indicates thaももheinfluence of the topography and land-sea

distribution on t.he GN response is relatively small andもhePhilippine Sea pla旬 isbasically represented

bya lD sもructurewith the real land-sea di抗ribution.

The ZII1. values which exclude EII di剖ortedby the land-sea distribution are also shown in Figure :35

(もhinline). Comparison of the so1id and dashed lines of Figure 35 ensures that the land-sea distribution

affects the GP response more than the GN ones， and also sugges胞もhaももheresistivity of t.he first. layer

should decl'ease， because t.he dashed line fi臼 betterもhanthe solid line at higher frequencies. Since a larg0

contrast of the conductivity tendsもoenhance a large currenもatthe boundary， the small enhancement. at. higher frequencies shown in Figure 35 suggests a smaller conductivity contrastもhanModel G N.

(31)

On Geoelectric Potential Variations Over a Planetary Scale

1∞

f H 争

ob!oervcd GN >-+-< simulaled GN --

simulaled GN (no dislonion)一一obllerved GP唱「

simulalcd GP -_. liimulaled GP (no dislOftion)一A

EEaovbZEe-E邑身

10

0.1 O.α1001 0.0001

F陀quency(Hz) O.∞1

180

150

observedGN→ー・liimulatcd GN -品imulalcdGN一一obllervc:d GP唱→

ぉimulalc:dGP一一持imulaledGP

120

AE凶豊富£

明

皐

w-

mw

TA--内?

?EAVEd 偽

W

AT

''ATlzgz

忍凶冨

育子ち芋CIICC!mぬ・・.~.;.-:，、

ω

30

0 O.似則)J O.似)()I

Frequency (Hz) 0.001

37

1∞

{Egash-写室EMESE〈

obほrvcdGN・←-simulaled GN --

simulatcd GN (110 di伽同ion)。除、c:rvedGP・骨← e

simulalc:d GP -.... simulalcd GP (110 dislortionト

10 f -H 09 Eロ伺~

町田・s・. -e -.

0.1 O.α>001 O.似)()I

F陪quc:ncy(Hz)O.側川

180 ま虫魚台b史官官需品町一ム.ー品・・~.;.';

ob崎町edGN・0 ・simulalαJGN-simulalcd GN obsc:n'c:dGP・，0-・

鈴imulatc:dGP ..“・・ぉimulalc:dGP

150

120

{BH岡崎芝忽暗記

如

'mmz

圃

I-iy

TE-z

・・備幽，

Y向

Z

TI--is'

怠官

句・... ーーl旬1........ - • • • 軍司哩=・==哩哩E抵血盟網" ・.

'・・ .

ω

30

0 O.鋭胤)J O.α)()I

Frequency (Hz) 0.1)(11

Figure 35: Sirr叫 atedcable responses of Models G N (lefも)and GP (right) in comparison with t.he observed

ones. Apparent resistivit.y (top) and phase (bottom) are shown. Observed and simulated GN responses

are expressed by diamond and solid line， respectively， and those for GP are square and dashed line. Thick

a.nd Thin lines are simulated responses with and without the land-sea e宵ects，respectively.

In order to obtain a guide to the best model， we computed an extreme example (Model 1) WhOSf宅

街stlayer is thicker and less resistive than Model GN. Figure 36 compares the ZIIょofModel 1 wit.h the

ohserved ones. The results at higher frequencies shows better agreement while those at lower freqllencies

shows discrepancies. A thinner LCL than Model 1 is required to decrease the apparent resistivit.y at lower

frequencies. This support.s again that a GN type structure with the less resistive first layer is the best..

The result. of Model 2 shows the best agreement with the observed response among all four rnodels

(Fig.36). Model 2 is a mixture of Model GN and GP， that is， the conductive first layer， LCL， and HCL.

Discrepancies of the phase suggest.s thaもtheconductivity of the LCL will be slightly highert.han that. of

the Model 2. Theもhicknessof the LCL is consistent. with the 1D models of the Pacific plate rat.her than

t.he unusual model at the Mariana forearc.

The thin sheet mode1ing with the four 1D models clearly demonstrat.es that the structures beneath t.he

sea floor affect the electric field observed at the sea floor， even though the field is signi自cantly distorted

byもheconductivity contrasもbet.weensea water and rocks. This strongly suggests that the ocean bott.om

measurement is useful to reveal conductivity profiles inもheoceanic area， alld that the thin sheet modeling


mt 喝

岨 ~守勺?

-向刷且.

鍋

w

，E.Ei

備幽曹

100

10

(ggash忠之宮司自冨ω福島田〈

付置争時~、ξ下 ¥

f tt ，ぶご¥

100

10

(EEZO)bτznmg冨ω昆内向島〈

0.1 0.00001

0.1 0.00001 0.001 0.0001

Frequency (Hz) 0.001 0.0001

Frequency (Hz)

180 180

150 150

120 (ωω』凶

ω司)ωgz

120 (ωω』凶

ω司)ωω宮E

11 _ーーーーーT W ・H・匂・:;.~

CjI .......::・・企ー~、、ー民も電量 ...-...・り・z¥

a画均価 ¥.・ 06

-D 11._ ・

-0‘

智

EE

・E・--au・

AV

90

60 ー!を i

守口奇?去三ニム」....~.ø句冨~..å..~.:.:......土・・.

90

60

30 30

0.001 0.0001 Frequency (Hz)

0 0.00001 0.001 0.0001

Frequency (Hz)

0 0.00001

Figure 36: Simulated cable responses of Models 1 (lefiも)and 2 (right) in comparison with the observed

ones. Apparent resistivity (もop)and phase (botもom)are shown. Observed and simulated GN responses

are diamond and solid line， respectively， and those for GP are square and dashed line.

is one of t.he mosもeffectiveestimation me出odsof the distortionもoavoid misinterpret.ation.

2D modeling

The effecもsof deep structures such as a slab can emerge at low frequencies where the thin sheet modeling

could not be demonst.rated. The 2D modeling developed by UTADA (1987) and improved by lJVESHlMA

(1990) was appliedもoevaluate them.

The 2D modeling used in this study is an application ofもhefinite element method which divides the

whole modeling space into finite elements to solve a partial diπerential equation (Induct.ion equa.t.ion).

The triangular elements are used with consもantconductivity， and the electric field or t.he magnet.ic field is calc.ulat.ed at three nodes of each element in the cases of the E 01' H polarization， respectively. Then I

calculated fields aももhenodes are linearly int.erpolated at any other places. The boundary ('ondit.ions al'('

描 follows:(1) t.he source field is uniform at the top of the air layer， (2)もheinduced field is zero at t，he

bottom of the modelil1g space， and (3) the induced field is constant in the direction normal to t.he lateral

boundary.

4.4.3

0" G芝り PC'~~Sぶ部品出ca(l)39

日/ Izu-Bonin 0-1 • 3.3 0.03 0.03 一『闘Z咽

~ . ~ 0.01

0.01 四

a・0.01 E ~ ]

五二 100 4C-L d

Q)

ロ、鳴

0.1 [S/m} 0.1

200 。 500 1000

Distance[km]

日/ Mariana E

• • 3.3 0.03

b

0.01 ~ 50

E 」ピ'--'

五::; 100 4C・4L

ω Q

0.1 [S/m}

200 500 1500

Distance[km]

Figure 37: Cro時間ct.ionsof Izu-BonIn (top) a吋 Mariana(boUom) areas. The co吋 uctivityvaluE's aIド

5hO¥""11 in the pancls

40

W or

~ 50

E .::&:. ]

£CL 100 。。150

200 。

日/

04

~ 50

E .::&:. }

-E 100 4C・4L 。。

200 500

2DC~出s~Sèction (2)

0.03

0.01

0.1 [S/m}

• O:Ol fI，

0.1 [S/m}

Izu-Bonin

•

500

Distance[km]

Mariana 白酬 ix4wFeて可廿司副‘川3.3

一0.0'1 バ;'ti可能ι九

弘

1000

Distance[km]

znU1園 'om

1500

E

1000

E

Figurc 38: Cross sections of Izu-Bonin (top) and Ma.riana (bottom) areas in the case or no slab. The

conductivity values are shown in the panels

41 On Geoelectric Potential Variations Over a Planetary Scale

Two cross secもIonswere designed corresponding to the Izu・Boninand Mariana arcs (Fig.37) t.o repre-

sent t.he difference in the subduction angle between the regions; the Paci自cplate subducts with a steeper

angle in Mariana than in Izu-Bonin. The angles were determined from a distribution of hypocenters.

Bot.h cross sect.ions are simple and suitable for qualitative estimation. The real topography was referred

for making t.he shallower surface parts of the cross secもions.The results of theもhinsheeもmodelingand

YUKUTAKE et al. (1983) were taken into account in the cases of the Philippine Sea and Paci恥 pla.tes.

Since the modeling aims t.o evaluate the effect of the slab， the electric and magnetic fields were ca.lcula目もtc町:!(刈d l

a叫tthe surface and b。ωもttωomof the sea with and without the subducting slab below (Fi泡g

culat“ions were conducted in E polarization mode for the Izu-Bonin section， and in E and H polarization

modes for the Mariana section， because the GN is parallel to the strike of the Izu-Bonin-Mariana arc

but the GP is normal to the Mariana arc. The electric fields alongもhecables were integrated， and t.hen

divided by the mean magnetic field at the t.wo cable ends.

The apparent resistivity and phase of the GN and GP cables obtained with the 2D modeling were

shown in Figure 39. It was shown thaもtheslab could increase the apparent resistivity of the G N cable

at lower frequencies where the thin sheet modeling can not be carried out. On the ot.her hand， the GP

response does not suffer fromもheeπect of the slab. This indicates t.haもtheslah should be taken int・0

account in the case of a modeling ofもheasthenosphere by using volもagedata obtained by a submarine

cable nearly parallel to t.he subduct.ion zone.

Subduction一一-No Subduction

100

10

(EEZO)b宮古

22出冨ω局内皆同〈

Subduntion --No Subduction

100

(EEzsb宮宮

28出冨ω諸島色〈

le・05le-05 0.001 0.0001 Frequency (Hz)

Subduction -一一No Subduction

ζJ

Aυ

0

0

0

0

0

0

0

0

0

o

k

nヲ

00

守，

L

U

ζ

J

A

斗

司

3

司

L

'

且

0.001 0.0001 Frequency (Hz)

Subduntion --No Subduction ……・

(ωω勧告)ω詰記

む

λ

、

ζJnu

b』E

・E

・

-

.

9

0

0

0

0

0

0

0

0

o

k

nu，OO

守

'

ζ

U

4

J

a

斗

司

3

司4

・E

(ωω』凶

ω唱)ω凶伺』仏

0.001

Figure 39: Simulated responses of the GN (left) and GP (righも)cables. Apparent resistivit.y (top) and

phase (bottom) are shown.

O.∞01 Frequency (Hz)

0.001 0.0001 F陀quency(Hz)


4.5 Summary

Spatially averaged distribution of the electric conductivity of the Philippine Sea plate was est.imat.ed

fromもheGN and GP voltages. The MT method for t.he uniform source was applied toもhevolt.ages

and geomagneもicfield which were averaged over the cable lengths. This is allowed because the na.tural

geomagnet.ic field gradually varies het.ween the cable ends. The MT responses were calculated a.t periods

from 50 sεc to 2 dαys for the GN a.nd GP cables， and 1D optimum models were separately obtained. The resulもshowedsignificant di宵erencesbetween the models of the GN and GP ca.bles. The ca.use fOl' the

difference was investigated by using aもhinsheet modeling a.nd a 2D modeling. The land-sea distribut.ion

ofもheregion distorts the GP voltage， and affects the GN voltage less than the GP one. Influence of

t.he subducting Pacific plate wi11 be limiもedto the GN volもageonly aもfrequencieslower than about.

1 X 10-4 H z. As a resulも， the Philippine Sea pla.もewas approximated by a 1D structure which has a low

conductivit，y layer overla.ying a high conductivity layer atもhedepth of about 80 km.

5 Tide

The t.idal signals on the cable voltages have been focused many times because ofもheirsignificant ampli-

tudes (e.g.， LONGUET-HIGGINS， 1949; DUFFUS AND FOWLER， 1974). Those studies showed disc閃 pancics

between t.he theory and the observations. This is because the tide exists not only in the water ('urrent

but a.lso in the ionospheric and magnetospheric sigl1a.ls so that the ca.ble volta.ge cωont.吋t.凶'岨'，a叩1

eほxも旬erna.叫10ωri氾gi凶n錨 well a槌st仙ha抗も of t.h】eoceanic origil1. In other words， difficulty of t.he interpretatiol1 on t.he

tide lies in the separation of the two differenもsources.

A robust least. squares fitting was applied to the whole data sets of the four cables to extract the tidal

signals a.~ well 部 Sqand its harmonics. Since t.he two sources were not separated each other~ obt.a.ined

results were mixture of t.he ionospheric and oceanic tides. The amplitudes and phase of t.he t，idal signals

are listed in Ta.ble 3. Since the st.art.ing dates of the data set.s a.re different， comparison of t.he ph加 es

amollg the cables should be careful.

Table 3: Result. of the robust. least. squares fitting to the tidal periods. The amplitude (A)， phase (0)司

and error in % (e) are listed. The dat.a length andもhefirst day of the data are also shown.

GN GP G M HAW-1

Tide 1175.7 days (1992.JlID.1) 730 days (1993.Sep.1) 179 days (1995.Mar.6) 1395 days (1990.Apr.7)

A (V) 。(deg) c5 (%) A 。 6 A 。 6 A 9 6

Q1 0.0499 75.46 73.30 . . ー 0.1669 60.79 39.26 0.06713 12.57 18.22

01 0.2670 4.186 13.70 0.04190 -22.45 42.32 0.09588 -115.3 68.50 0.3781 130.9 3.2a5

P1 0.2967 -21.85 12.54 0.1403 -58.00 13.22 . . . 0.1898 -63.11 6.479

81 0.2199 -166.8 16.93 1.398 -151.2 1.326 1.845 30.44 11.04 0.6202 -156.7 1.983

K1 0.7077 -53.32 5.257 0.1536 63.19 12.08 . . . 0.2885 -111.6 4.263

N2 0.1820 79.76 20.10 0.03978 153.6 44.57 . . . 0.4559 -153.1 2.683

M2 0.9273 -4.132 3.945 0.2676 -50.79 6.626 0.4582 -121.4 14.21 2.108 -55.97 0.5801

82 0.8450 -64.89 4.409 0.9082 150.1 2.040 2.189 -20.82 4.947 1.984 -149.0 0.6198

1(2 0.1683 126.6 22.12 0.05519 50.25 33.56 0.2715 156.6 39.80 0.3160 -123.9 3.891

On Geoelect.ric Potential VariaもionsOver a Planetary Scale 43

The tidal amplitudes of all cables are 1 ~ 2V， which is about 10% of the voltage flucもuat.ions.This

shows that theもideand Sq are significant sources of the cable volもage.01 (25.81924 h.ours) is known as

a purely oceallic tide， and M2 (12.42059 hours) is mainly produced by the ocean. Onもheother hand， S2

(12.0 hours) and SI (24.0 hours) that are the major components of Sq contain large externally produccd

t.ides. In the HAW-l and GN cables，もheoceanic group has comparable amplitudes with the ext.ernal

group， while SI and S2 are much larger thall M2 and 01 in the GP and GM cables. This refleds a spatial

tendency of the tidal c.urrents in the Pacific， and indicates that the noise levels of the GP alld GM cables

are lower in the analyses of the exterllally induced and core fields.

6 Motionally induced field

6.1 Introduction

SANFORD (1971) and CHAVE AND LUTHER (1990) obtained an approximated relatiollship between t.hc

horizontal elec.tric自eldat the ocean bottom EIa and water velocity Vh at a low frequency band wherc t.he

self and mut.ual induction effects are negligible， as follows:

Eh =C凡と X {Vh}* ( :J2)

where Z-

JU

一

、リ

-z

LUγ-Ju

h-A7

V

一ιhy

Aり一

σ

何一J

Dτh

-1-

rJ一一一

• ‘‘.，， L

H

V

，，a、、

(33)

C，Fz，D， and σdenote a dimensionless scale factor，もhemean vertical geomagnetic field， t.he sea dept.h

and the sea water electric conductivity， respectively. They assumed a flat bottom and a horizontal flow

whose spatial scale is much larger than the sea depth. Equation 32 means thaもEIais proportional to t.he

conductivity weighもeddepth integrated flow (Vh)・.σisroughly depth independent (e.g.， CHAVE AND

LUTHER， 1990)， and therefore， (Vh.)'" reflects a depth independent (barotropic) flow. This relationship

was proved in experiments using point electric field sensors and rnoored currentmeters (Cox et al.， 1980;

LUTHER et al.， 1991). They showed tha.t the horizonもalelectric fields at the ocean bott.om were reasona.bly

consist.ent with the barotl'opic flows estimated by water velociもyrecords of the rnoorillgs. The barotropI<'

component.もendst.o be underest.imated in t.he mooring experiment.s due to difficulties in resolving a

vertical structure of the flow， and due to low sensitivit.y of the current. meter at the deep ocean. They

concluded thatもheocean bottom electrometer was proved to selectively extractもhebarotropic flow with

the sensitivity higher than the moored currentmeter at periods longer than a few dαys.

Equatioll 32 is laterally integrated in the case of the cable voltage. Two experiments in the St.rait.付

of Florida (SANFORD， 1982; LARSEN AND SANFORD， 1985; SPAIN AND SANFORD， 1987; LEAMAN ff (11.

1987; LARSEN， 1992) and in Bering Strait (BLOOM， 1964) showed that the short spanned cable voltage reflect.ed the water transport across the cable in comparisou with the water velocities observed along the

cable by the moorings. Old studies (e.g.， LONGUET-HIGGINS， 1949; STOMMEL， 1954; Cox ct al.， lD64;

DUFFUS AND FOWLER， 1974; RICHARDS， 1977) were not successful in this point maiuly because they aimed at the DC and tidal componenもswithout consideration of the significant influence of the corc and

externally illduced自elds，respectively， and because the approximation of Equation 32 is 1l0t. sat.is自edin

high frequency phenomena such槌 thetide whereもheselιand mutual induction e宵ectsare noもnegligiblc.

These successful experimellts， however， still leave open several problems on interpretat.ion of the ca-

ble voltage， for instance， effects of a large-scale electric current loop (SANFORD， 1971; LARSEN， 1992;

44 I.Fujii and H.Uもada

STEPHENSON AND BRVAN， 1992)， current short t.hrough conductive sediments (Cox， 1980) and topogra-phy (LARSEN， 1992). It is essential to establish a numerical simulation method so that uncel'ta.inties 011

the interpretation are resolved. The numerical simulation by FLOSADOTTIR et al. (1997a，b) is the first，

successful atもemptto compuもeEquation 32 over an open-ocean including general circulation models and

realistic topography. Their results ensured the lillear relationship between the cable voltagt~ alld 10ca.I

もransportobserved in the Straits of Florida， and suggested thaもvolもagesof a planetary scale submarille

cable will reflectもhewater transporもevenif the cable is laid in an area far from an illtense oceallic flow.

It is， however， still not dear whether such a large scale water transporもphenomenonill an open oceall

exists 01' how such a water transport is generaもed.

In this section， the motionally induced自eldis exa.mined using the nearly 4・yearcontinuous da.t.a of

the HAW-1 cable which extends 4000'km in the eastern North Pacific. We are motiva.ted by previolls

works listed above and availability of the long term dat.a sef，. We assumed that the motionally induced

voltage of the HAW-l cable is proportional to the water transport， and investigated features of the volt.age

focusing on the generation mechanism of subinertial oceanic flows in the region.

According to the kinetic energy map of the steady and time-varying barotropic flows by ¥¥TYRTI{!

et al. (1976) (Fig.40)， the vicinities of the in加 lseoceanic currents are characterized by both energetic

steady and time-varying components， while the other regions show relatively energetic variabilities but

much lower energies for t，he steady flows. Mesoscale eddies in the ocean were detec旬Uin 50、sもo70‘s

(e.g.， MASUZAWA， 1965; KOSHLYAKOV AND GRACEV， 1973; ROBINSON， 1976)， and import剖 lt，roll of

the mean current instability was recognized to produce a large energy level of the eddy variability in

t.he vicinity of a western boundary current (e.g.， WYRTKI et al.， 1976). However， causes of t，he eddy variabilities in t，he areas far from the western boundary currents were argued a matter of debate for

decades. For instance， WVRTKI et al. (1976) concluded th抗 theinsもabilityof the intense st.eady flow

also dominated the energy levels of the variabilities in distanもareas.Early theoretical models Suppol'ted

this conclusion by showing that wind-driven flow

P(い)= 'i a .._-，，j(k，ω) 1/，'-ωo十

(34)

011 Geoelecl.l'ic Potenlial Variations Over a. Planetary Scale 45

Kinetic Ener of the Mean Flow

Figure 40: I¥inetic energy of t.he st.eady日0¥V(¥01') and eddy (boLLom). R.eproduced froll】 ¥VYRTI¥I ('1川

( 19i6)

where 。OklWOτ一

ドー

(35)

P， k， /'， a.nd F are an oceanic response， a wave number ve叫or，a friction coeftic.icnl of j，he Ra.yleigh type

damping， and forcing， respectively. Equation 34 suggests tha.t the ocea.n responds to the atmcちpheric

forcillg depencling on frequencies， a.nd， waもenumbeJ's through a disp目別011relation 01' Equa.t.ion 3;' and

thc behavior can be c1assified with three regimes : (1) 0町一resonantregime (ω 》 ω。)， (2) resonant rcginw

(ωz ωoL and (3) Sverdl'up I'egillle (ω《 ω0).Ther田 ponse田 ba.rot.ropicin the on'-r田 onalltalld rE'S仙 aJ1t，

re巨imes，while the baroclinic respollse increases its illlportance inもheSverdrup regillle which is cxpf'('j.ed

t.o be at. periods longer t.han 100 dαys from the Illodel studies. In t.he ofr-resollant. regime (::; se、円 01

days)， only e、anescent，waves can eXJ引，and therefore the response is loca!ly limited. In t.he resonanl

46 I.Fujii and H .Utada

regime，onもhe0もherhand， free propagating Rρssby waves tendもobe activated and the response could be

nonlocal. Since the Sverdrup balance is expected in Sverdrup regime， the response becomes to be local agam.

It should be kepもinmind that factors which are not included in the stochastic models， for instance， t.he complexity ofもheもopographyor wind field， can easily change this simple overview. KOBLINSKY

(1990) presented the world wide distribution of the planetary component of the pot.ential vorticity f / H

from the real topography averaged over about 200 km. The ocean was divided into two regions， i.r..，

regions of large and small effective beta (三|て7(1/H) 1). The free wave band is expect.ed t.o be widcll in

the large effect.ive beもal'egion. CHAVE et al. (1991) examined the wind stress curl dat.a in the Nort.h

Pacific provided by the Fleet Numerical Oceanography Center (FNOC) and revealed that t.he curl is

considerably variable in space and time. The curl field also showed a spatial coherence at a frequency

over abollt 1000 km and many distant coherence maxima. This resulもimpliedth抗 simplifiedWilld fields

in t.he model studies may callse underestimate the atmospherically forced response.

DICKSON et al. (1982) found indirect evidence of the aもmosphericforcing for thc first. t.ime; they

observed seasonal fluctuations of deep kinetic eddy energy in the eastern North Atlantic. Then， direct.

evidences were reported in variolls parts of the oceans as a form of coherence between the oceanic and

at.mospheric variables. KOBLINSKY AND NIILER (1982)， NULER AND KOBLINSKY (1985) and I(OBLINSKY

et al. (1989) presented the local Sverdrup balance in the North Atlantic and in the North Pacific aもperiods

from 10もo100 dαys. The coherence， however， between water velocities by moorings and wind st.ress

curl was small and insignificant. This was known in the model stlldy by MULLER AND FRANKIGNOlJL

(1981) who gave almosもzerocoherence at many frequencies between the water velocit.y and curl above

a measurement site despiもea causal relationship. Furthermore， CUMMINS (1991) concluded that an oceanic response at a point is wave like at periods less than 100 dαys in the North Pacificもakingreal

topography into account.， and suggested thatもheSverdrup balance is observed ifもhevort.icity equatioll

term is averaged over several degrees.

This frusもrat.ingsituation was changed by BRINK (1989) who found significant non-local coherence


variabilities depending on measurement sites alld periods， which was interpreted as an influen('e of t.hl;'

local topography and complicated Willd field. The locations of the nonlocal maxima were consistent

with propagating directions of Rossby waves detected by array of instruments which showed int.ersit.e

consistency. They suggest.ed that their coherence was much largerもhanother experimental studies part.ly

because baroclinic noises were eπectively eliminated as gelleral characteristics of the bottom pressure e¥.l1d

(Vh)¥They also suggested 1.hat the frequency bandwidths wereも00broad in the ot.her st.udi何附lllt.i時

contamination of wavenumber through the dispersion relationship. On the other hand， LIPPEH.T ANU

MULLER (1995) proposed a different interpretation on significant nonlocal coherence maxima. They

computed coherence maps using t.he transfer functions of MULLER AND FRANKIGNOUL (1981) as a

function of a separation dist.ance and found thaもthenonlocal coherence is a common featul'e wit.hollt.

distant forcing in their statistically homogeneous model.

Thus. the motionally induced field of the HAW-1 voltage is expected to reflect wind-drivell barotl'opic

flows in the easterll Paci自cat periods longer than several dαys， if it exists. Indecd， the first y<.~ar dat.a

set. showed signi自cantlocal and nonlocal coherence with the curl in the north eastern Pacific (仁~HAV~~ d

al.， 1992b). Since the 4-year data set is one of the longest duration among experimcllt.al st.udies 80 far

obtained， spatial and temporal feat.ures of at.mospheric forcing at. a wide period range wit.h fille resolllt.ion

call be obtained. Atmospheric forcing of small wave numbers is an int.eresting feature which is possibly

examined in t.his study， because small scale flows are filtered out in t.he cable voltage. It. is， how円'('1'，

di白cultto assess illterpreもationon nonlocal maxima with one sensor.

6.2 Analysis

6.2.1 Removal of noise

As described in Section 3.3， 12 hour values of t.he HAW-1 voltage (，ollt.ain the externally induced恥ld

前 wellωthe motionally one. An adaptive filter developed by WJDROW et al. (1975) wぉ applif'C(t.o

extract. the motionally illduced voltage. This method calculatesもime-varyingt.ransfe1' fUIlct.iolls of tll('

voltage to three components of the geomagnetic field at HON so thaもpredicteddat.a fi t well observf'd

ones. Parameters are flexibility of the transfer function and filter length. The flexibili1.y wa.<; scanned for

several decades in a filもり length， and the best model wぉ selected on t.he basis of good 自tnf'Ss.Thf'n

もhebest filt.er length was de1.ermined by comparing time series， power spectra， and coherence of result.s

fo1' di官erentfilter lengths. The AIC (AKAIKE， 1973) did not give a reasonable result， probably because

amplitudes of the residual were caused by shorter period signals such as 1 dαyandもheAIC didn 't. reflect

good fitness of longer period signals which is our interest. Before this filtering， meall and liJl(、artl'でnd

were removed from all data sets， alld the AR filもerwas applied to the geomagnetic自elda1. HON for

prewhitening. Use of the AR自ltergave a bett.er result in the adapもivefiltering， because the AR filt.er

reduces the power ofもhegeomagnetic field at lower frequencies so thaもpreventslower frequ町村ysignals

from being emphasized too much in the adaptive filtering.

Original and predicted voltages， and residual are shown in Figure 41. Lengもhsof the AR and ada.}パ，Ive

filもersare 1.5 and 22.5 dαys， respect.ively. Since t.he externally induced field is t.he major COmpOIl('"Ilt. in t.IH'"

predicted voltage， the motionally induced volta.ge is expected to he dominant component. in t.h('" l'esidua.1.

This is visually checked as follows. Squared coherence between the motionally induced volt.age and t.he

geomagnetic自eldat HON is much smaller than that without filtering at periods shorもert.han aboll 1. 27

dαys (Fig.42). Furthermore， the power spectra ofもhemotionally induced field monot.onically deCl'ease


〉剛山州山山州いい叫O 200 400 600 800 1000 1200 1400

day

〉 ~ ~LμいJJ.ムム4叫J.ムムμLL」叫j山山一iJIl一ムム一ムi凶ムムdムム叫I~ ιιJιん叫叫tlLい叫:dJ品>dJJ~…!ド"珂r"r円T"~'吋r.. -，咽π 司"JWT-I炉RJT>岬司".，.，.問宵γ，."，..川司η'1"'ry'阿町r-rrrn刊日苛叶，司冒一，一rTT"可..-..-.. 司~司開町…川.. 日，.T門叩，叶守 4

圃 30200 400 600 800 1000 1200 1400

day

〉

トー~ーFigure 41: 1400・dayplot of t.he HAW-l voltage from April 7， 1990 to March 2， 1994. (a) Origh凶 12

hour values， (b) externally induced field predicted by the adaptive filtering， and (c) residual are shown fromもopto bottom.

in comparison with that ofもheoriginal data at the frequency band where coherence changes (Fig.43).

The問 indicateもha.t.the filもerillgresulもshownin Figure 41 is reasona.ble， namely，もhegeomagneもicnoise is

effect.ively removed. Thus，もhemot.ionally induced voltage was obtained for 3.6 yeαrs from April 16， 19UO

to December 1， 1993. It. cannoもhedirectly translated to a tl'制lsportbecause there is no independent.

transpol't measurement along the cable t.o calibrate the dimensionless scale factor C in Equation 32.

The amplitude of the motionally induced field replotted in Figure 44 is rather small implying significant.

influence of t.he spatial averaging. Seasonal variaもionsare shown as well as longer t.erm Olles. Those

variations a.re listed ill Table 4. Va.riance is the largest ill winter (Ja札 toMar.)， which is consistent wit.h

a known fea.ture of the wind. Varia.nces also change greatly yearもoyear.

Table 4: Seasonal variance (V2) of the motionally induced voltage from 1990 to 1993

Seωon 1990 1991 1992 1993

Jan.-Mar. 0.03562609 0.07631434 0.07379515

Apr 0.02826856 0.04401183 0.03835214 0.06818821

Jul.-Sep. 0.02836272 0.04687581 0.02475272 0.02421333

Oct.ーDec. 0.02504055 0.05977371 0.04961853 0.03111951

40 On Geoeled.ric Potential VariaLiolls OVCI' a Planeta.ry Scale

0.8

'唱。、

0.6

0.4

0.2

官

ω』EdF的hυ=ωlMω

戸-。υ

0.1 0.01 Frequency (cpd)

0.001 0 0.0001

Figure 42: Squared cohercncc beいvcellthe original volta.ge and thrce COIl】poncntsor the geornagnel，ic neld

observecl at HON (solid line) ancl I，hal bel.ween thc rnotionally inclucccl field and t.he salllC geolllagll叫 IC

ficld (dashed line)

0.1

0.01

0.001

(玄uhNZ(E者〉E))25ω且凶包診。

ι


0.001 0.0001

0.0001

Power spcctra or t.he original volLage (solid line) and t.he Illot.ionally indurccl ficlrl (dashcd Pigure il3

line)


0.5

> 0

・0.5

91 92 93 94

Figure 44: Mot.ionally induced voltage versus the da民.The firsもdayis April 16， 1990 and the t.ime axis

is ticked every month.

6.2.2 Coherence Inap

Two-point squared coherellce between the volt.age and atmospheric varia.bles was ca1culat.ed in ord<:'I'

t.o obtain direct evidences of the atmospheric forcing. The wind velocity at 10 m hight. and surfa.re

air pressure were provided by the ElIropean Centre for Medium-Range Weather Forecasts (herea.fter

ECMWF). The ECMWF product.s consist ofωsimilated 12 hour values for 3933 sites spaced every 2.50

on the ent.ire Pacific. The duration of the data is exac.tly the same as the HAW-l voltage da.t.a. The

wind sもresswas computed frol11 the wind velocity， and then the vertical component of the wind st.ress curl was obtained at 3685 points frol11 t.he wind st.resses of 9 adjacent point.s using two dimensional finit.e

di町erenceformula in a spherical coordinate. Basic features of t.he ECMWF product has been ensurt'd

to be consist.ent with the FNOC product which LUTHER et al. (1990) a吋 CHAVEet (1，1. (1992a，b) lls(.d

(e.g.， CHAVE et al.， 1991). The spatial dist.ribution of the ECMWF product is more sparse t;hau t.ha.t. 01'

t.he FNOC product， and therefore the ECMWF wind field is expected to lose short.-scale componellt札

Squared coherence between t.he volもageand an atmospheric variable at each of 3933 gl'id points (:3s85

for the cUl'l) was comput.ed as a function of frequency. The atmospheric variables used hel'e were t.he

curl， meridional and zonal wind st.resses， surface pressure following BRINK (1989)， LUTHER et (1，1. (1990)，

and CHAVE et al. (1992a). Variables except. t.he curl do not have a direct forcing relationship wit.h t.he

voltage， however show significant. coherence. This is because the wind stress and surface pressure are

coherent with the curl and ('onsequent.ly can be substitutions for it. They act.ually played importallt. rolls

in CHAVE et al. (1992a) t.o ensure the forcing relatぬnbetween t.he curl and barotropic velocity， mainly

because they are expected to contain less noises than the curl which is obt.ained by a spatial derivat.iol1

of the wind.

The overlapped period bands were selected to observe transmission of the coherence pa“el'ns. Sixty five periods for 3.6 year data and eighteenおl'each year are list.ed with upper and lower limit.s in Tabl内

5 and 6， respectively. Since the t.ime bandwidt.h prodllct (TBW) was fixed as 4 for all cakulations， finer

resolutions are obt.ained at shorter periods in the time domain. Choice of TBW is important. to a.void

contamination of wave number， as a compromise bet.ween computati

Contour maps were obtained for the coherence (~ 0.4) bet.ween the voltage and an atmospheric

val'iable ovel' t.he Pacific basin for each period. The ('ollt.oul'ing interva.l was every 0.1. Figm・何 45( a.)""( d)

On Geoelect.ric Potential VariaもionsOver a Planetary Scale 51

Table 5: Calculated periods with both sides limits for the 3.6-year segment

Period (day) lower limit upper limit Period (day) lower limit upper limit

221.00 147.33 442.00 9.61 9.40 9.82

132.60 102.00 189.43 9.34 9.14 9.54

94.71 78.00 120.55 9.08 8.90 9.27

73.67 63.14 88.40 8.84 8.67 9.02

60.27 53.04 69.79 8.61 8.45 8.78

51.00 45.72 57.65 8.39 8.24 8.55

44.20 40.18 49.11 8.19 8.04 8.:H

39.00 35.84 42.77 7.99 7.85 8.1:l

34.89 32.34 37.89 7.80 7.66 7.!H

31.57 29.47 34.00 7.62 7.49 i.75

28.83 27.06 30.84 7.45 7.33 7.58

26.52 25.02 28.21 7.29 7.17 7.41

24.56 23.26 26.00 7.13 7.02 7.25

22.86 21.74 24.11 6.98 6.87 7.09

21.39 20.40 22.47 6.84 6.73 6.94

20.09 19.22 21.05 6.70 6.60 6.80

18.94 18.16 19.79 6.56 6.47 6.C>6

17.92 17.22 18.68 6.44 6.34 6.53

17.00 16.37 17.68 6.31 6.23 s.41

16.17 15.60 16.78 6.20 6.11 6.:l8

15.42 14.90 15.98 6.08 6.00 6.17

14.73 14.26 15.24 5.97 5.89 6.05

14.11 13.67 14.57 5.87 5.79 5.95

13.53 13.13 13.96 5.77 5.69 5.84

13.00 12.63 13.39 5.67 5.59 5.74

12.51 12.17 12.87 5.57 5.50 5.()4

12.05 11.73 12.39 5.48 5.41 5.!)5

11.63 11.33 11.95 5.39 5.3:3 5.46

11.24 10.96 11.53 5.30 5.24 !).37

10.87 10.61 11.14 5.22 5.16 5.28

10.52 10.28 10.78 5.14 5.08 5.20

10.20 9.97 10.44 5.06 5.00 !>.12

9.90 9.68 10.12

are coherence maps at the period of 32 dαys between the voltage and curl， zonal and meridional willd

sもresses，and surface pressure， respectively. Typical features of the maps at. all periods are seen in Figures 45: (1) Significant local and Ilonlocal coherellce patches are found between the voltage and all atmospherir

variables. (2) Coherellt patches in various sizes are scat.tered over the region forもhecurl， and t.hose in

large sizes are isolated for t.he pressure. The maps of the Willd stress are int.ermediate. And (3) peak

values are t.he highest ill maps of the curl. These features are b部 icallyconsistent wit.h t.hc r昔、sult.ぉof

52 I.Flljii and H. Utada

Table 6: Calculated periods with both sides limi旬 fort.he each year segment

Cen ter period (day) lower limiも upper limit

60.83 40.56 121.67

36.50 28.08 52.14

26.07 21.4'i 33.18

20.28 17.38 24.33

16.59 14.60 19.21

14.04 12.59 15.87

12.17 11.06 13.52

10.74 9.86 11.77

9.61 8.90 10.43

8.69 8.11 9.36

7.93 7.45 8.49

7.30 6.89 7.77

6.76 6.40 7.16

6.29 5.98 6.64

5.89 5.62 6.19

5.53 5.29 5.79

5.21 5.00 5.45

4.93 4.74 5.14

CHAVE et al. (1992a，b). These indicate thaもtheremay exist a large scale t.ransport of the 、，vind-dl'Iven

oceanic current t.hat is not diminished by the integration over the cable scale， and that the cable is

promised to work as a barot.ropic Cllrrent met.er as well as a point sensor.

Coherence patches spread over various part.s of t.he Pacific bac;in， even when micro pat.ches are l'引noved.

Coherence maxima are 0.5，.....0.8， alld diametel'S of the pat.ches are in the order of 1000 km. These are

comparable with the result.s of CHAVE et al. (1992a). Distances of the nonlocal maxima from the cabl仔

are ranged up to ，..... 10000 km， which is much larger t.han in other studies.

However， it should be noted tha目tsome of t.he llolllocal coherence may reflect teleconnection of t.lw ('url

自eldrather than the atmospheric forcing as described in CHAVE et al. (1992a). The curl field is kllOWll

t.o show significant spatial coherence with itself， that is， t.eleconnect.ion (e.g.， CHAVE et al.， 1991). This

makes t.he curl-voltage coherence indirectly illcrease. It is necessary to rule out the coherence due t.o t.he

teleconnection.

6.2.3 Selectioll

CHAVE et al. (1992a) proposed al1 empirical procedure to l'ule out the coherence patches which l'eflect.

t.he telecollnectioll and random noise. This procedure finds the mo叫 similarcurl to t.he volt.age dat.a

at. a period， il1 other words， finds the mosもstablephase relationship between the voltage and ('111'1. lt.

consists of three steps. First， coherence maxima are selected among patches which include Illore than a

few grid point.s (roughly 1000 km in scale). Figure 46 is a coherence map between t.he voltage and curl at.

On Geoelect.ric Pot.ent.日1Variat.ions Over a Planet.ary Scale 53

/

10. @)，

'0.

Figure 45: Coherence maps beいveen{，he voltage and (a) curl， (b) zonal wind stress， (c) meridional wind

sl，ress， iHld (d) surface pressure at. 32 days， clorkwisc frorn t.hc t.op left corner. Squared coherellcc 11101"<-'

t.hall 0.4 which is zem coherence at. 95 % significance level is cont.oured every 0.1

i dαys in which t.here arc 3 candidat.es markcd by solid circle， trie.lllgle， and CI'OSS. Second， illte rcoIH~ r f' n CC

beい¥'eena selecLed ClIrI and an oceanic variablc (eit.hcr of t.hc curl， moridional andぉol1alwind st.rcss， and

surface pressllrc) at. cach ofall grid point.s is comput.ed. Int.ercoherence maps arc macle fol' all at.l1lOsplterIc

variables (F'ig.47.........50). This means that， I，he volt.age in (a) of F'igures 47.........50 is substituted ror the curl at九

日 leciedmaximum point in (b).......(d) of F'igures 117.........50. Third， the intercoherence maps are compared wil.h

t，he original ones， alld a curl which can reprodtlce the original maps is determined. F'or example， maps of

candidate 1 (Fig.47( b)) and 2 (Fig.4 7( c)) reasonably reproduce the original volt.age-curl cohcrencc 11】汎p

54 l.Ii'ujii and j'I.Ut"da

"" o q

。も

O〈b匂

F'igure 46: Coherence maps bet，¥Vcen t，he voltage礼ndcl.lrl at 7 da.ys. Cobel'cnce maxil1la a.rt' mark円 l

by solid circle， t.riangle， and cross. Squared coherence Jl10re than 0.4 which is Zl'ro cohcrence at. !J!j %

significance ¥cvel is cont.oured evcry 0.1

(l"ig.4i(a))， however a map of candidate 3 (l"ig.47(d)) does not. Then， from Figures 48(α) -(c) it can be

concluded that the iniercoherence map of ca.ndida.t.e 1 is more similar to the original voltage-pressure map

than that of candidate 2. The iniercorrel叫 101111凶 psbet.ween t.he clIrl and wind sin'ss al50 SUPPOI't. t.his

tendcncy (Fig.40......50). Comparison 01' pressure maps is gencraJly more effect.ive th制 Olleof wind sl.ress

maps. Maps of candidat，e 3 <.10 110もrescmblc1.0 t.he originals in all atmospheric variablcs (Fig.47.......!)O)

Therefore， the curl of candi【lat.eI is considered to be the ll10st剖 milarto the voltage， and t.hf' pat.ぐ"問

、vhichcontains candidat.e 2 and 3 are ruled out. because thcir similarit.y to t.he voll，age is 111同 I.ableilnd

t.hey reflect t;he I，eleconnect.ioll. If a. linear causaJ relationship is sal，isfied beいveenthe volt.a呂(;'alld ('url，

On Geoelectric Potential Variations Over a Planeiary Scale

一『

'0

l:/。 /

! /' )/

r;r: ¥)iJ

Figure 47: Cornparison of the original coherence ancl intercoherence ma.ps， Clockwise fr0111 the t.O[l Icft

COl'l1cr， cohercnce between (a) the curl of all and voltage， and intercoherence Illaps beいvcent.he ('11 rl of

all and curl5 at (b)candidate 1 (501id circle)， (c) 2 (501id triangle) and (d) 3 (cro日)eU 7 days. Squared

cohercllce 11l0rc tha.n 0.4 which is zero cohercllce at 95 % significance level is contollred evcry 0.1

t.he curl can be a substitlliion of the voltage. On the other hand， il， is nol. necessa.rily t.rue if cohercllct'

beい.veenthc voltage and curl is produced by ieleconnection of the curl field

This procedure is subjective and there is 110 guarantee from a mathematical view point. Nevcrt.hclel;s，

it was successful in CHAVE el al. (L992a) who proved the consistency with the directio川 ofde1.cl'lnillf'd

maxima and wave vcctors detected from Illultiple coherent sensors. Since there is 110 other indC!pclldcnt

observation of a large-scale water transport along the HA ¥'V-1 cablc，もhispaLterl1 matching pro('edure

was applicd 1.0 t，he present ca.ble data set to separate the coherent. pat.ches whichぬreproduced by tlu;.


/

民、

Figure 48: Cornparisoll of the original cohercnce and inf，ercoherence maps. Clockwise frol11 t.hc tOJl Icrl

corllcr， coherence beiwecn (a) thc 判明sureof all and volLagc， and intercohel'cnce n同 pshct.w伺 nl.he

pressure of ，，11 alld cIlr1s al (b)calldidale 1 (solid circ1e)， (c) 2 (solid triallg1e) alld (【1):3 (cro田 ) 川 7days.

Squarccl coherence mol'c th引 10.4 which is zero coherencc at 95 % significance level is conもOlll'cdevcry

0.1

t.eleCOJlIlCclioll. 309 coherence maxInla of 35 periods were exarninecl for the 3.6 yeαl' segmcnt.， a.nd 14TCi

intcrcohercnce n、apswcrc produced In tote:¥1. 336 paもcheswCl'c.iuclged to be produced by thc telecollllect.ioll

of t.he curl rield， ancl 33 pat，ches rerna.inecl at 2L periods

6.3 Results

Significanl cohc!'ellce was originally observccl at. 35 periods inもheanalysis of the 3.6 yea.r segmcllt.， and

Lhen 21 pel'Iods ¥¥'el'e l'erna.Ined art，er Lhe selecLIon of Section 6.2.3. Thel'e Is 110 rcason for Lhe cable volt.aιρ

011 GeoelecLric Potcnt.ial Vari叫 ionsOver a Ph.¥1同同ryScale 57

:)I11 」晶

凡ij ¥仕 ?11で o，j¥メ〆。

/司J〆 /、

--"

Figure 49: Compa.rison of the origina.l coherencc and intercohel'ence maps. Clockwise from the t.op left

corner， coherence bet.weell (a) the zOllal wind strcss of aJl and voltage， and int.ercohercnce maps bet.ween

thc wnal wind 51.ress of all and CllrlS al (b)candidate 1 (solid circle)， (c) 2 (5olid triangle) and (d) 3

(cross)叫 7days. Squared cohercnce rnorc than 0.4 which is zerQ coherence at 95 % significan同 levelis

contoured every 0.1

LO be coherent wit.h the a.tmospheric variables unless t.he volt.age reflects the wind-driven barot.ropic flo¥¥

Thereforc， t.his Is an evidence of t.hc ability of the planet.ary-scale submarine cable as a barot.ropic ctllTenl

meter， and of the large-scale wind-driven barol，ropic now in the eastern North Pacifi('

The 35 periods where significant. coherence pat.ches were originally seen were divided into 6 period

bands: 5.1-5.6，5.7-9，9，11.2・14.7，17-20守 32-44，and 95-133 d"ys. Although CHAVE el ，，1. (1992b) reported

t.hat， 110 significant coherence appeared at. periods shorter t.han 20 dαys in the analysis of the first year dat.a

58 l.Flljii and lI.Utada

Figure 50: Comparison of the original coherence and intercoherence maps. Clockwise from the t.op left

corner， coherence beい，veen(a) the meridional wind stress of all and voltage， and inlercoherellce Illaps

between the meridional wind stre田 orall and curls叫 (b)candidate1 (solid circle)， (c) 2 (solicl triangle)

a.nd (d) 3 (cross) at 7 days. Squared coherence more than 0.4 which is zera coherence at 95 % significancc

level is contoured every 0.1

set， the procedure to remove the geomagnetic noise improved the data qua.lity and made the analY8is

at. shorter periods possible in this st，udy. Figure 5L shows t.he distribution o[ selected 21 periods in

comparison with the power spect.ra of the voltage. The diamonds indica.ting 21 periods are seen etl'olll1d

power spectral peaks. The period bands of the barotropic currents are well resolved in our case， and I.hey

are roughly consistent w仙 thoseobserved by BRIr口T附 (1989) (3.7下一8，8-23， and 23-76 dαys) and CIIAVE cl

α1. (1992a) (6-8， 10-14， and 20-50 dαys)， when differcnces or data lengths are considered. This suggesLs

that the wincl-clriven ba.rotropic flows have components of small wave numbers in comparison with the

cable length (abouも4000km)，、.vhichwas not well resolvecl by point observations

59 On Geoelect.l'ic Potential Variat.iolls Over a Planctary Scale

。。。0.1

0.01

0回 1

(Eu、FLEd〉E})己芯邑的

bB£

0.0∞1 0∞l 0.1

Figure 51: Frf'町叫【qucnc口，es、W‘ h】ere forc口叩11川ngrelat.i 。印IISare delected (dial1l、n、。IId)are supe町r口叩1口川111川1刊poscdon tl仙h】陀ep旧o、u、V町vでes叩pe目clraof the 仁印ablevol も凶age(solid li川n】e)

。01Frequcncy (c凶)

ab

'" O官庁行ι|

v 串 '/〆

〆

、、

Figure 52: Coherellce l1laps between the voltage and curl at 133 days (Ieft) and 95 days (riglll.). Solid 1".卜

angle indic-ales a selcded coherence maximum. Squared coherence Illore than 0.4) which is問 rocohcl'cnce

at 95 % signi自callcelevel， Is conloured every 0.1

Spatial patterns of coherence are considerably variable with perioc!s. The voltage-curl cohercncc lIlaps

which includeもhe33 selected coherencc patches at the 21 periods arc shown for each period band (Fig.52

57). The selected coherence maxIma were marked by solid triangles. A d叫山山on01' tl】e33 patches with

solid t.riangles is wide-spread even in a period band， except t.hat the triangles are around Lhe eqllator aL

the period band of 11.2-14 dαys. Especially in the shor回sttwo period bands (5.1-5.6 and 5.7-9.9 d削~s)，

the spatial coherence paLLerns quickly change witl】 periods，indicaLing Lhe complexity of Lhe wa.ve numbcr

sLrllct.ure in the wind-driven Aow field

60 I. Fujii and H. U tada

Figure 53: Coherence maps beもweenthe voltage and curl a.t 44 da.ys (left)， 35 days (middle)， and 32 days (right.). Solid triangle indicates a selected coherence maximum. Squared coherence more t.ha.n 0.4‘ whidl

is zero coherence aも95% significance level， is contoured every 0.1.

Figure 54: Coherence maps between the voltage and curl at 19 days. Solid triangle indicaもesa select.ed

coherence maximum. Squared coherence more than 0.4， which is zero coherence aも95% significance level， is conもouredevery 0.1.

Temporal variations are shown as a form of power specもrain Figure 58 which compares the power

spectra for ea.ch year (a year from mid・April).For example， power spectrallevel at 30・40dαys increases

year by year， however that at 6・9dαys remains at almosもthesame level. Since the power spect.ral

peaks correspond to the significant. coherence patches for the 3.6-yeat' segment， temporal val'iatioJls of

the coherence bet.ween the oceanic and atmospheric variables are expected， Figures 59 are the coherency

maps between the voltage and curl at 39 dαys for 3.6 yeαrs and at 37 dαys for each year， The map of

the third year shows much larger coherent patches than those of the other years，ωexpected from t.he

diπerences of the power spectra. The map of the 3.6・yearsegment looks like an intermedia.te of three

yeαrs， however influence of the third year is greater than the other two years. In the period band of 6・9

On Geoelectric Pot，ential Variat.ions Over a Planetary Scale 61

Figure 55: Coherence maps beLween the voltage and curl at 13， 12，5， 12， and 11.6 c1a.ys c10ck、，Vlse白・0111

t.he top left corner. Solicl triangle indicates a selected coherence maximum. Squared coherence mOl'e t.h仙

0.4， which is zero coherence at 95 %剖gnificancelevel， is contoured every 0.1

dαys， active periods can be found for all segments of the dat.a (l'ig.60). These tenclencics are supported

by coherence l1laps between the voltage and other aLmospheric variables. Figurcs 61 and 62 show tl1 ....

voltage-pressure coherence l1laps for the same data segments拙F'igllres59 and 60， rcspectively

These temporal differences are partly due to variabilities of the wind field. Figllres 63 and 64 show

the distriblltion ofもhemean and variance of the curl for 3.6 yeα1"S and each year. Pat.l.erns of t.hc Ilwan

are sil1lilar to each 0もher，however those of the variance show difTerences with time. For installce， I，!le

contour line of lO-16g2/cm4s4 gets closer to the equator with time， while that of 1O-I!ig2/cm:1s4 sl.ays

at almost the same place throughout 3.6 yem's. Namely， the variability ofもhewind illcreased with timc

at mid laLitude. From a view of c1imatology， a year of 1992 w田 thebeginning of EI Niuo which ended in

1993

Further studies are necessaryもo50lve details of the observed phenomena. The cable neい，vorkwhich

continuously records the voltages in varIolls parts of the Pacific will contribute to improve the intel'prc-

t.atioll of the wind-driven barotropic current in the region

I.Fujii and j.I.Utada 62

、卜Ale¥

Fザゲ

ー

い搭

lr

.

0唱品!、14

ごUqq句

、門

ihAトぃ、尚一

'UMhレ、、占

Figure 56: Coherence maps bctwccn the vollagc and curl at 9.9， 8.6， 7.8， 7.6， 7，剖ld6.1 days c10ckw目e

[rol1l 1.lIe iop left COl'ner. Solid tri(1ngle Indicates a selected coherence maximum. Squared coherellce 1ll0rt'

t，han 0.4， which is zcro coherence叫 95% significa.nce level， Is coniollred every 0.1

Summary

The rnotiona.lly induced fielcl was 札叫 ractedfrorn lhe 3.6-year segment of the unpowered volia.ge obsPl"v引 l

witl】もhcHAW-l cable. Amplitudes 01" the extl'acted volもagewere l'at.her 5mall due to spat.ial a"eragIng

of t.hc motionally induced field. The extracted voltage was proved to relat.e to the atmosphcric vRriablcs

(wind stress curl， wind stress， and surface pressure) of the ECMWF produci by the cakulat.ion of cか

hercncc. Significant coherence between the voltage and atmospheric variabl四 wasobtainecl at pCl'iods

from 5-133 dαYS over the entire Pacific basin. This suggests that the voltage was induced by large-scale

wind-driven barotropic Aows in the region. Features of the wind-driven日owsare supposed t.o be complex，

because t.he coherence were both local and nonlocal， and coherence patterns were highly variable in space

and time. This may reflect the complexity of wind ficld variations

6.4

63 On Geoelectric PoLential Va.riatiolls Over a Planetary Scale

aJ戸、;'I:r:-~ ， .V ベ ¥s怜込f匂，¥ 0--ム-'<

f も! ぷ-E i 。1 ベ午r-可.

! ' 士i : 、ト，a 戸九、JJji -lι仁岬.‘b

， / 4乙E

Figure 57: Coherence maps beLween the ¥'oltage and curl at 5.6， 5.4， and 5.1 days fro111 left. 1.0 right

Solid t.ria.ngle indicat，es a selected coherence maximul11. Squared coherence more than 0.4， which is ZCI'O

coherence at. 95 % significance level， is cont.oured every 0.1

0，1

0.01

0.001

(司巴

υ¥NZ(EU宅〉

E))E百ω品目』

ωEO仏


0.001 0.0001

0.0001

I'igure 58: Power spectra of a year from April， 1990 (solid line)， from April， 1991 (doshed lineト川，d

April， 1992 (dotted line)

64 I.Fujii and fl.Utada

Figllre 59: Col附 encemaps betwecn the voltage and clIrI at. (a) 39 days for 3.6 yea民 at.37 days for (b) a

year from April， 1990， (c) a year from April， 1991， and (d) a year from 1992， c10ckwise from t.hc t.op lefも

comcr. Squa.red coherence morc than 0.4， which is問 rocohcrence at 95 % significance level， Is contourcd

every 0.1

On Geoeledric Potential Variations Over a Planetary Scalc 65

Figure 60: Coherence maps between the voltage and curl at (a) 7 days for 3.6 years， (b) 8 days for the

first. year， (c) 6.3 days for the second year， and (d) 7 days for the third year， c10ckwise from the top 10仇

corner. Squared coherence mQre than 0.4， which is zero coherence at 95 % significance level， is contourcd

every 0.1

66 l.Fujii and I-I.ULada

Figure 61: Coherence maps between the voltage and surface pressure at (a) 39 days for 3.6 years， and at.

37 days for (b) a year from April， 1990， (c) a year from April， 1991， (d) a year from April， 1992， clocklV目e

凸0111te 1.op left corner. Squared coherence more tha.n 0.4， which is zero coherence叫 95% significa.nce

level， Is cont.oured every 0.1

011 Geoeleclric Potcnlial Variations OVf'I'礼 PlanelarySca¥c

/ ズ

J

67

o

tτ〉

¥

、t-

U J ノ

Figllre 62: Coherence maps bclwccll the voltage and slIrface pressure at (a) 7 days ror :3.6 years、(h)ぷ

days for the first year， (c) 6.3 days for t.hc second )'eilr， and (d) 7 days for t，he t.hird ye川，clockwisf' rrom

t.he t.op le仇 corner.SCJlIared cohercllcc l1lore t，han 0.4， which is肘 1'0coherence eU O!) % sigllI日canCeI('vぺis cont.ollred cvery 0.1

68 L

F

、L

J

-I

〈j

rt

I

Figure 63: Mean or the curl for 3.6 years and each yca.r c10ck、visefroll1 t.he t.op ]("fj. conlf'r. COlllolll" lillf's

剖 edrawn every 2x lO-9g/cm2s'!. with dotted (ueg叫 ivevall附)ancl solid (posit.i刊 valucs) 1 i nC8

On Geoelect.ric P01.('nt ial Variai..iolls Over a. PIa.l1etar，Y Scale G!l

Pigllre 64: Variancc of t，he curl for :3.6 years and each ycar cJockwise rrom i..h(' top I('fl， <"orlll'r. IOglfl

(、ariance) 只 C0111011 rp(1 引 eryIOo-:?y:! jcm'ls'l

70 I.Fujii and H. Uもa.da

7 DC and linear trend

Det.ection of t.he electric field which is derived from the toroidal geomagnetic field a.t the out.er core is olle

of t.he most impol'ta.llt. topics in the cable works. There is no 0もherway to observe t.he core field on the

Eart.h's surface. Since RUNCORN (1954) and ROBERTS AND LOWES (1960) pointed out t.he possibilit.y of

the core field detect.ion， at.tention has been paidもothe DC component of the cable voltage (RUNCORN，

1964; DUFFUS AND FOWLER， 1974; LANZEROTTI et al.， 1985， 1992). RecentIy， UTADA AND HINATA

(1995) studied the AC component of the core field， and revealed t.hat the AC will have a compara.ble

amplitude with the DC. This is worもhconsidering because measurements of the A C componellt.s are

t.echnically much easier thanもhoseof the DC.

The meall a.nd linearもrendof t.he vo1tages of t.he foul' cables were estimaもedby a least squares fit.t.ing.

First.， the t.ida.l signals were 1'emoved fi仕1-om30 mηη~i灼ηn valu

2初omi灼11】 V刊a.lue筒sofもu山he陀eHAW-l voltage (いse関eSection 5幻).The corl'ected volta.ge was used fo1' the GN c.a.ble

instead of using the powered voltageもoreduceもhenoise by the supply cu1'1'enも. Secolld， t.he volta.ges were lowpassfiltered alld resampled into 12 h values. Since the cor1'ected voltage of the GN cahlc shows a

stepwise cha.nge at. a1'ound the 500t.h day， means we1'e sepa1'ately extracted from t.he dat.a set. before and

aft.er 500th day. The data ga.ps ofthe HAW-l voltage we1'e interpolatedぉ describedin Sect.ion 3， and the

exterllally illduced field obtained in Section 6 was removed f1'om the original12 h values of the HAW-I

cable. Finally， a linea.r regression estimation was conduct.edもot.he voltages ofもhefour ca.bles. That is，

a. model v( i) =αi + b was calcula.ted to fiも Wlもhthe d叫a'11{ i). Estimates of αand b wit.h fitt.illg e1'ro1'S

f， alld t.he va1'iance ofもhefitting fo1' each cable were listed in Table 7. The自ttingerrol's and vaJ'iallces

of the HAW-l and GP cables are small suggesting that the estimates are 1'eliable. The va.rianre ofもhe

HAW-l cable is abouもone-thirdof that of the GP cable， probably because t.he exte1'na.lly illduced fielcl

was efficient.ly removed. The extremely large variance of the GN cable indicates that the lloise of t.he

power supply st.ill 1'emaills. Estimates of the GM cable are instable dueもoa short data length.

Since the cable lengt.hs are about 3000" ， 5000 km， observed mean and trend per year are both a.bout.

several hundreds millivolts. These are observable amounts ac

Table 7: Meall and linea1'も1'end

Da.ta (day) T1'end (m V jkm day) DC (mVjkm) RMS

length ga.p a 6 b 6 (mV jkm)

HAW-l 1326 。-1.123x 10-4 0.029x 10-4 0.2320 0.0023 0.0580

GN 1163.5 471 -2.870 X 10-4 0.868x 10-4 ーー 1.009

GP 630 361 -1.804x 10-4 0.430x 10-4 0.3107 0.0161 0.1575

GM 156.5 124.5 3.939x 10-4 2.968x 10-4 -0.0205 0.0313 0.1638

On Geoelectric Potential VariatiollS Over a Planetary Srale 71

白'equenryband as studied in Section 6. III a lower frequellcy band. gelleration mechanism of a large-

sc.ale wat.er t.ransport is poorly ullderstood. Improvement of general circulatioll model and EM numerical

simulat.ion must be awaited.

8 Conclusion

The unpowered and powered voltages measured by four planetary scale submarine rables ill t.he Pa.cific

were examilled at periods from seconds to DC.

A spatial averaged conductivit.y dist.ribution of t.he Philippine Sea plate was estimated by using t.llf、

exもerllallyinduced field of the GN and GP cables. It revealed that the PhiIippine Sea plat.e approximat.ely

has a Olle dimellsional st.ructure belleath a three dimensionalland-sea distribution and that. the subduc.t.illg

Pacific plate should be considered if t.he asthenosphere beneath t.he Izu-Bonin-Mariana arぞ ismodelcd.

The mot.ionally induc.ed field w悩 obtainedfrom the unpowered voltage of the HAW-l cable for 3.C>

yeαrs. The amplitude of the voltage variation was at most 1 V， which is rather small t.han t.hose of tIHヲ

exもernaland tidal components. The extracもedfield showed sigllificanもcoherencehetween t.he at.・mospl】eri('

variables over the entire Pacific basin at periods from 5 to 133 dαys. This suggest.s that t.he wind-dl'iven

barotropic flows in the region have large-scale components which can be observed by t.he planct.ary-scalc.'

cable. Coherence pat.terlls were variable in space and time cOl'respondillg t.o variat.ions of t.he wind fi<'ld.

The t.idal compollent and secular variat.ions of the four cable dat.a set.s were comput.ed. The resultぷ

depended on the data set.s. The GP and GM voltages contailled smaller oceanic tides t.han t.llP ot.ht干1'8.

The DC and linear trend wit.h smallぞrrorbars were obtained in the HAW-l and GP data set.s. whilρ

t.he result.s of t.he GN and GM were less reliable beca.use ofもhepower supply noise and short. dat.a

lengt.h， respectively. Bot.h the DC and trend obtained in thisもhesisare large enough t.o measure frol1l an

experiment.al view point.

Acknowledgement

We t.hank KDD alld AT&T fo1' making cable observations possible， and thank L.J . Lanzerot.t.i and his

group at Bell laboratories for providing the HAW-l data and constructive comment.s. ¥Ve are also

sincerely grateful to A.D.Chave for his contribution t.o the motionally induction ana.lysis of t.his st.udy.

This work is based on the doctoral dissert.ation by I.Fujii in 1996.

References

AKAII(E， H.， Info1'mation theory and an extension of t.he maximum likelihood principle， 2nd Inter. Symp.

on Information. Th.eory， edited by B. N. Peもrovand F. Csaki， 267・281，1973

AKAIKE. H.， Likelihood and the Ba.yes procedure， in Ba:lJesia7l. Statistics. edited by J. M. Bernardo et. al.，

Universit.y Press， Valencia， 143-166， 1980

BLOOM， G. L.， Water t.ransport and temperature measurements in the eastern Bering St.rait.. 1053・lO58.

J. Geoph.ys. R.fS.， 69， :3335・3353，1964


BRINK， K.H.， EvidellCe fo1' wind-c1rivel】 cllr1'entfluctuations in the weste1'n No1'th At.Iantir， J. Gcophys.

Rcs.， 94， 2029・2044，1989

CAGNIARD， L.， Basic theo1'Y of the ma.gneto-telluric method of geophysica.l prospecting， Geoph.lJsiω‘45‘

1・16，1953

CHAVE， A. 0.， D. J. THOMSON alld M. E. ANDJ~R ， On the robust estimation of power spectra. coherenc(':o;.

and transfer functions， J. Geoph.ys. Res.， 92， 6:33-648， 1987

CHAVE， A. D. and D. J. THOMSON， Some comments on magnetotellu1'ic response function est.imat.ioll司 J.

Gcoph.ys. Res.， 94， 14215-14225， 1989

CHAVE， A. D. and D. S. LUTHER， Low frequency， motionally induced elect1'omagnet.ic自eldsill t.he ocean，

1: Theo1'Y， J. Geoph.ys. Res.， 95司 7185・7200，1990

CHAVE. A. D.， D. S. LUTHER ano .1. H. FILLOUX， Variability of the wind st.1'ess curl ovcr t.lH' nort.h

Pacific: Implications fo1' the oceanic 1'esponse， J. Geophys. Res.， 96， 18361-18379、1991

CIlAVE， A. D.， D. S. LUTHER and J. H. FILLOUX， The barot1'opic electromagnetic and pressure experi-

ment ]. Barotropic cur1'ent 1'esponse t.o at.mosphc1'ic forcing， J. Geoph.ys. Res.， 97， 9565・9593，1992a

CHAVl~ ， A. 0.， D. S. LUTHER， L. .1. LANZEROTTI and L. V. MEDF'ORD， Geoelectl'ic field meaSlll'Cments

on a 'plalleもal'ysca.le : oceanog1'a.phic and geophysical applications， Geoph.ys. Res. Letf.， 19， 141ト1414，

1992b

Cox， C. S.， Elect.1'omagnet.ic iuduct.ion in t.he oceallS an infel'ences on the const.it.ut.ion of thtヲ。ar(.h.(，'(0-

ph.y.s. Slt7・1'.，4，137・156、1980

COX， C. S.， T. TERAMOTO a.nd .J. H. FILLOUX， On cohel'ent eledl'ic and ma.gnetir fluct.uatiol1s in 1.h('

sea， in St-udies on Oceanograph.y， edit.ed by K. Yoshida， Tokyo， 449司 457句 1964

Cox， C. S.， J. H. FILLOUX， D. I. GOUGH， J. C. LARSEN， K. A. POEHLS， R. P. VON HBRZgN a.nd R・

¥VINTER， Atlant.ic lit.hosphere sounding， J. Geomag. Geoelectr.， 32， Suppl. 1， 8113-32， 1980

CUMMINS， P. F.， The bal'ot1'opic response of subpola.r NOl'th Pacific to stochastic wind forcing， .J. Oeo-

ph.ys. Res.， 96， 8869-8880‘1991

CUMMINS， P. F. and H. J. FREELAND， Observation a.nu ll10deling of wind driven cu1'rents in the nort.h(，3st.

Pacific， J. Ph.ys. Oceanogr.， 23， 488-502， 1U92

DAWSON， T. W. a.nd J. T. VVEAVER， Three-dimensiona.l induction in a non-unifo1'm thin she叶.at tllt'

おurfa.ceof a. uniformly conducting eart.h， Geophys. .J. R. astr. Soc.， 59， 445-462， 1979

On Geoelectric Potential Varia.t.ions Over a. Planetary Scale 73

DICKSON， R. R.， W. J. GOULD， P. A. GURBUTT and P. P. KILLWORTH， A seasonal signal in the ocean

currents to abyssal dept.hs， Nature， 295， 193・198，1982

DMITRIEV， V. 1. and M. N. BERDICHEVSKY， The fundamental model of magnet.ot.elluric sounding， Proc.

IEEE， 67， 1034・1044，1979

DUFFUS， H. J. and N. R. FOWLER， On planetary Voltages， ocean tides， and electrical conduct.ivity bdow

the Paci自c，Can. J. Earlh Sci.， 11， 873-892， 1974

EGBERT， G. D.， Noncausality of the discrete-time magnetotelluric impulse response句 Gcophysic.'i， 57，

1354-1358， 1992

EOBERT， G. D.， J. R. BOOKER， a.nd A. SCHULTZ， Very long period ma.gnetotelhll'ics at Tucsoll observa-

tory: Est.imat.ion of impedances， J. Geophys. Res.， 97， 15113・15128，1992

FARADAY， M.， Bakerian Lectllre : Experimental researches in electricity -second series， Phil. 1'rans. R.

Soc. Lond.， 122， 163-194， 1832

FILLOUX， J. H.， Oceanic electric current.s， geoma.gl1etic variations and the deep ele('trical condurt.ivit.y

st.rllctllre of ocean-cont.inent transition of central California， Ph.D Thesis， 166pp.、Univ.of Calif.， San

Diego， La J ol1a， 1967

FILLOUX， J. H.. Magnetotelluric soundings over the northeast町 paci自cmay rev~al spat.ial dependen("(~ of

dept.h and conductance of t.he asthenosphere司 EarthPla71et. Sci. Let.， 46、244-252，1980

FILLOUX， .J. H.， Magnetotelluric Experiment over the ROSE Area.， J. Geophys. Res.， 87， 8:364-8378， lH~1

FILLOUX司 J.H.， Seaftoor magnetotclluric soundings in the Mariana island arc area， in The Teぞt.onicand

Geologic Evollltion of SOlltheast Asian Seas and Islands : part. 2， Geophys. Alonogr. Slr.， 27， edit.ed hy

D. E. Hayes， pp.255・265，AGU， Wωhingt.on D心， 1983

FLOSADOTTIR， A. H.司 J.C. LAR.SEN and J. T. SMITH， Motional induction in North At.lant.ic circulation

model， J. Geoph.ys. Res.， 102， 10353-10372， 1997a.

FLOSADOTTIR， A. H.， J. C. L.ARSENβnd J. T. SMITH， Thc relation of seaftoor voltages t.o ocean t.ransportぉ

in North At.lant.ic cirClllation models: Model results and pract.ical considerations for t.ransport. monit.ol'-

ing， J. Phys. Oceall ogr・， 27，1547・1565，1997b

FRANKIGNOUL， C. a.nd P. MULLER， Quasi-geost.rophic response of a.n in自nite13-plane occa.n to St.OChflSt.ic

forcillg by t.hぞ atmosphere，J. Ph.ys. Oceanog1・.， 9，104・127，197~)

FUJII， 1.， H. UTADA， T. YUKUTAKE， Y. HAMANO， Y. TANAKム andK. YUMOTO， Conductivit.y dist.ribu-

74 I.Fujii and H.U凶da

tion below a sea自oorinferred from voltage measurements using a submarine cable (in Japanese)， Proc.

1993 CA symposI1tm， 79・92，1993

FUJII， 1.， L. J. LANZEROTTI， H. UTADA， H. KINOSHITA， J. KASAHARA， L. V. MEDFORD alld C.G.

MACLENNAN， Geoelectric power spectra over oceanic dis凶nces，Geoph.ys. Res. Lelt.， 22， 421-424， 1991)

HEINSON， G. S.， A. WHITE: L. K. LAW， Y. HAMANO， H. UTADA， T. YUKUTAKE， J. SEGAWA and H.

TOH， EMRIDGE: the electromagnetic investigation of the Juan de Fuca ridge， Mar. Geoph.ys. Rω.，15，

77-100， 1993

HONKURA， Y.， Geomagnetic variat.ion a.nomaly on Miyake-jima island， J. Geomag. Geoelectr.， 23， 307・

333， 1971

ISHIGURO， M.， T. SATO， Y. MAMURA and M. Ooe， Analysis of Earth tide data (iu Japa.nese)， B'llll. fnsl.

i¥f(lt.h. 81α1.， 32， 71-85、1984

JONES司 A.G. and R. HUTTON， A ll1111ti-statioll magnetotellurics study in southern Scotland -1. Field-

work， dat.a analysis and results， Geoph.ys. J. R. astr. Soc.， 56， 329・349，1979a

.JONES， A. G. and R. HUTTON， A multi-statiol1 magnetotellurics study in souもhernScotland・11.Montc-

Carlo illversion of the da.ta a.nd its geophysical and tecもonicimplications， Geoph.ys. J. R. astr. 8oc.， 56，

351-368， 1979b

KARIG， D. E.， Basill genesis in t.he Philippine Sea， in Initial Reports 0/ the Deep 8ea Drillin.q Projecf. 31，

edited by D. E. Karig， J. C. Ingle Jr.， et al.、 pp.8!>7-880，Washington (U .S. Government. Print.ing Office)，

1975

KAWATATE， K.， A. TASHIRO， M. ISHIBASlII， T. SHINOZAKI， T. NAGAHAMA， A. KANEKO， S. MIZUNO， J.

KOJIMA， T. AOI¥I， T. ISHIMOTO， B. H. CHOI， K. KIM， T. MITA and Y. OUCHI， A cross-spectl'a.l analysis

of small volta.ge variation in a. submarine cable between Hamada and Pusa.n wit.h speed va.ria.tion of the

Tsushima warm cllrrent， in Oceanography 0/ Asian Marginal Seas， .Elsetlier Oceallograph.y Se内es，54，

edit.ed by K. Ta.ka.no， pp.207-222， Elsevicr Sci. Pub. B. V.， Amsterdam， 1991

KOBLINSKY， C. J.， The globa.l distribution of //H and the barotropic response of the ocean， .J. Geoph，ys.

Res.， 95， 3213-3218， 1990

KOBLINSKY. C. J. a.nd P. P. NIILBH.， The rela.もionshipbetween deep ocean current.s alld winds ea.'5t of

Barbados. J. Phys. Ocea71ogr.， 12， 144・15:3， 1982

KOBLINSKY. C. J.、P.P. NIILER and ¥V. J. SCHrvJITZ司 Obsel'vat.iollsof windforced deep ocean current.s in

t.hf:、 Nort.hParifi(・../. GfOph.lP'， Rf..... 94. 1077:3-10790‘1989

|¥り<..;III，¥'¥I，()¥'¥|¥;llId、¥¥1.(;H，¥C'L¥'‘入I(刊 )-SC・al<'('III'I'f'lIt.ぉa.t.a hydl'ophysIca.l pol'ygon in t.he t.l'opI<'al


Atlantic， Deep-sea Res.， 20， 507-526， 1973

LANZEROTTI， L. V. MEDFORD， C. G. l¥fACLENNAN， D. J. THOMSON， A. MELONI and G. P. GREGORI

Me部 urementsof the large-scale direct-current Earth potential and possible implicatiolls for the geomag-

net.ic dynamo， Science， 229， 47-49， 1985

LANZEROTTI， L. J.， C. G. MACLENNAN and A. C. FRASER-SMITH， Background magnet.ic spect.ra: '" 10-5

to '" 105Hz， Geoph.ys. Res. Lelt.， 17， 1593・1596，1990

LANZEROTTI， L. J.， C. H. SAYRES， L. V. MEDFORD， J. S. KRAUS， and C. G. MACLENNAN， Earth pot.en-

tial over 4000km between Hawaii and California， Geophys. Res. Leti.， 19， 1177-1180， 1992

LANZEROTTI， L. J.， A. D. CHAVE， C. H. SAYRES， L. V. MEDFORD and C. G. MACLENNAN， Large-scale

f.'lectric日eldmeasurements on the Earth's surface: a review， J. Geophys. Res.， 98， 23525-2:3534， 199:3

LARSEN， Transfer functions: smooth robust est.imates by least-squares and remote reference met.hods

Geophgs. J. In1.， 99， 645・663，1989

LARSEN， J. C.， Transport measurements from in-service undersea旬lephonecables， IEEE J. Ocean.. En-

gin.， 16， 313， 1991

LARSEN， J. C.， Transport and hea.tflux of the Florida. current. at. 270 N derived from cross-st.ream volt.ages

and profiling data: t.heory and observa.tions， Phil. Trans. R. Soc. Lond.， 338， 169・236，1992

LARSEN， J. C. and T. B. SANFORD， Florida Current volume transports from voltage measurement.s， Sci-

ence， 227， 302-304， 1985

LAW， L. K. and J. P. GREENHOUSE， Geomagnetic variation sounding of the asthenosphere beneat.h t.he

Juan de Fuca Ridge， J. Geophys. Res.， 86， 967-978， 1981

LEAMAN， K. D.， R. L. MOLlNARI and P. S. VERTES， Structure alld variability of the Florida rurrent at.

270: Apri1 1982・July1984， J. Phys. Oceanogr.， 17， 566-583， 1987

LIPPERT， A. and P. MULLER， Direct atmospheric forcing of geostrophic eddies， part 11: Coherence maps，

J. Phgs. Oceanogr.， 25， 106・121，1995

LIZARRALDE， D.， A. D. CHAVE， G. HIRTH and A. SCHULTZ， Northeastern Paci自cmantle conduct.ivity

profile from long period magnetoもelluricsounding using Hawaii-to・Californiasubmarine cable dat.a， ./.

Geophys. R(s.， 100， 17837・17854，1995

LONGUET-HIGGINS， M. S.， The elect.ric and magnet.ic e汀ectsof t.idal st.reams， Alon. N 01.. R. .4 iーt，..が0('.

Geophy，c. Suppl.， 5， 285・307，1949


LUTHER， D. S.， A. D. CHAVE， J. H. FILLOUX and P. F. SPAIN， Evidence for local and nonlo('al barotl'opic

responses to atmospheric forcing during BEMPEX， Geophys. Res. Lett.， 11， 949・952，1990

Lu'I‘HER， D. S.， A. D. CHAVE and J. H. FILLOUX， Low frequency， motionally induced elect.romagnetic

自eldsill the ocean， 2: Electric field and Eulerian currenもcomparison，J. Geophys. Res.， 96， 12797・12814，

1991

MACI{lE， R. L.， B. R. BENNETT and T. R.. MADDEN， Long-period magnetotelluric measurements near

the central California: a land-Iocked view of the condllctivity structure lInder the Pacific Ocean. Gcophys.

J.， 95， 181・194，1988

MAKIRDY， D. McA.， J. T. ¥VEAVER and T. ¥V. DAWSON， Induction in a thin sheeもofvaria.ble condllc-

ta.nce at t.he sllrface of a strati自edearもh-2.Three-dimensional t.heory， Gcophys. J. R. astr. 8oc.， 80‘

177-194， 1985

MASUZAWA， J.， Meander of the Kuroshio: IJump of cold waもerin the off."hore of the southerll HOllShu，

J a.pan (in J apanese )， J( agaku， 35， 588・593，1965

MEDFORD， L. V.， A. MELONI， L. J. LANZEROTTI and G. P. GREGORI， Geomagnetic Indllction 011 a.

transat.la.ntic communications cable， Nature， 290， 392・393，1981

MEDFORD， L. J. LANZEROTTI， .J. S. KRAUS and C. G. MACLENNAN， Transat.lallt.ic earth pot.ent.ial va.ri-

at.ions durillg t.he March 1989 ma.gnetic 剖orl11， Geoph.ys. Res. Lett.， 16， 1145・1148，1989

MELONI， A.， L. J. LANZEROTTI and G. P. GREGORI， Induction of current.s in long submarine cahles by

llatural phenomena.， Rev. Geophys. Space Phys.， 21， 795-803， 1983

l¥10RI， T.， Variations in the geoelectric自eldwith relation to crustal condit.ions of the ea.rth， Geo1)h.lJsical

Afagazinc， 42， 41・104，もheJ apall Met.eorologica.l Agency， Tokyo， 1987

MULLER， P. a.lld C. FRANKIGNOUL， Direct. atmospheric forcing of geosもrophiceddies， .J. Phy.s. OCWllogl'.，

11， 287-:308. 1981

NIILER， P. P. and C. J. KOBLINSKY， A local， time-dependent. Sverdrup balance ill t.he east.ern Nort.h

Paci自cOcean， 8cience， 229， 754・756，1985

OGAWA， Y.， Two-dimensional resistivity modelillg based on regiona.l magnetotelluric survey ill the nort.h-

ern Tohoku district.， northwest Japan， J. Geomag. Gcoelectr.， 39， 349-366， 1987

OLDENBURG， D. W.， Conductivity structure of oceanic upper mantle benea.th the Pa.cific plate， Geophys.

J. R.αsfr. Soc.， 65， 359-394， 1981

OLDE~NBURG ， D. W.) K. P. WITTALL alld R. L. PARKER， Inversion of ocean bottom magnetot.elluric da.t.a.

On Geoelect.ric. Potential Variations Over a Planetary Scale 77

revisited， J. Geophys. Res.， 89， 1829-1833， 1984

PHILANDER. S. G. H.， Forced oceanic waves， Rell. Geoph.ys.， 16， 15・46，1978

PRICE， A. T.， The induct.ion of elec.tric current.s in non-uniform thin sheets and shells， Quart. .J. Mech.

Appl. Malh.， 2， 283-310， 1949

RICHARDS， M. L.， Tidal signals on long submarine cables， Ann. Geophys.， 33， 177・178，1977

RIKITAKE， T. and 1. YOKOYAMA， The anomalous behavior of geomagnetic variat.ions of short. period ill

J apan and its relation to the subterranean structure. The 6th report. (The results of further observations

and some considerations concerning the inftuences of the sea on geomagnet.ic variations)， B1I1I. Earfhq.

Res. Insl.， Univ. Tokyo， 33， 297-331， 1955

ROBERTS， P. H.， and F. J. LOWES， Earth currents of deep internal origin，よ Geoph.ys.Rc.s.， 66， 1243・

1254， 1961

ROBINSON， A. R.， Eddies and ocean circulation， Oceα7lUS， 19， no.3， 2-17， 1976

RUNCORN， S. K.， The eart.h's core， TransactioJl.s AGU， 35、49-63，1954

RUNCORN， S. K.， Measurements of planetary electric cllrrent.s， Nat包1'e，202， 10-13， 1964

SAMELSON， R. M.， St.ochastically forccd current ftuctuat.ions in vertical shear and over t.opography， J.

Geophys. R.， 94， 8207-8215， 1989

SAMELSON， R. M.， Evidence for wind-driven cunent日llct.uationsin the eastern North A t1antic， J. Gco-

phys. R.、95，11359-11368， 1990

SAMELSON， R.. 1¥1.， Correction "0 "Evidence for wind-driven current fluct.uat.ions in the <:'astern North

Atlantic" by R. M. SAMELSON， J. Geoph.ys. R.， 97， 821・822，1992

SAMELSON， R. M. and B. SHRAYER， Currellts forced by stoch部 ticwinds with meridionally-varying am-

plitude， J. Geoph.ys. R.， 96， 18425・18429，1991

SANFORD. T. B.， Motionally inducecl elect.ric and Maglletic fields in the sea， J. Geoph.ys. Res・， 76，3476・

3492， 1971

SANFORD， T. B.， Temperature t.ransport and motional induction in t.he Florida Current" .J. i¥lal'. R.，40.

suppl.， 621・6:39，1982

SASAI， Y.， Spatial dependellce of short.-period geomagnetic fluctuations on Oshima islalld (2)， 111t11.

Eal'thq. Res. /"51.， Unit1. Tokyo， 46， 907-926， 1968


SCHULTZ， A. and J. LARSEN， Onもheelecもricalconductivity of the mid-mantle: 11. delineatioIl of het引か

geneity by application of extrema.l inverse soluもions，Geophys. J. lnt.， 101， 565・580，1990

SHIMAKAWA， Y. and Y. HONKURA， Electrical conductivity sもruct.urebeneath the Ryukyu trench-arc sys-

t.em and its relation to the subduction of t.he Philippine sea plate， J. Geomag. Geoelectr.， 43. 1・20，1991

SPAIN， P. and T.B. SANFORD， Accurately monitoring the Florida Current with mot.ionally-induced volt.-

ages， J. !t1 ar. R.， 7， 843-870， 1987

STEPHENSON， D. and K. BRYAN， Large scale electric and magnetic fields generated by t.he Oreans， J.

Geophys. Res.， 97， 15467-15480， 1992

STOMMEL， H.， Exploratory measurement.s of electrical potential differences between widely spaced points

in the North Atlaut.ic Ocean， Arch. Met. Geophys. Bioklimaiol8er. A， 7， 292・304，1954

TARJTS， P.， Conductivity and fluids in the oceanic upper manもle，Phys. Earth Planet. lnteri.， 42， 215-

226， 1986

TERAMOTO， T.， Estimation of sea-bed conduct.ivity and it.s influence upon velocit.y measurements with

towed elecもrodes，J. Ocea.nogr. Soc. Ja.pan， 27， 7・18，1971

TERAMOTO， T. and H. KOJIMA， Eduction of hydrodynamical processes from cross-channel elect.ric-

pot.ential differences in the Kuroshio area aroundもheIzu ridge， J. Oceanogr.， 50， 339-363， 1994

TIIOMSON， D. J.， Spectrum estimation techniques for characterization and development of WT4 waveg-

uide， Bell Syst. Tech.よ， 56， pt.I， 1769・1815，pt.II， 1983-2005， 1977

THOMSON， D. J. and A. D. CHAVE， Jackknifed error estimates for spectra， coherences， and transfer func-

t.ions， in Advances in Specfru.m A nalysisαnd Array Processing， Vol. 1， edited by S.Haykin， pp.58・113，

Prentice Hall， 1991

TOH， H.， Electrical conductivity structure of the Izu・Boninarc revealed by seafloor elect.romagnet.ic ob-

servations， Ph.D Thesis， Univ. of Tokyo， l11pp.， 1993

UTADA， H.， A direct inversion method for twか dimensionalmodeling in t.he geomagnetic induction prob・

lem、Ph.DThesis， Univ. of Tokyo， 409pp.， 1987

UTADA， H. and K. HINATA， Imporもanceof electrical potential measurement by using the t.rallsoceanic

cable network， Ab.'jiracts ollUGG XXl General Assembly， A140， 1995

UYESHJMA， M.， Applicat.ion of Net.work MT methodもoもhestudy of electrical conductivit.y structure in

the central and eastern part of Hokkaido， Ph.D Thesis， Univ. of Tokyo， 332pp.， 1990

On Geoelectric Potential Variations Over a Planetary 5cale 79

VESSEUR， G. and P.羽'EIDELT，Bimodal electromagnetic induction in non-uniform thin sheets wit.h all

application to the northerll Pyrenean induction anomaly， Geophys. J. R. astr. Soc.， 51， 669-690， 1977

WIDROW， B.， J. R. GLOVER， Jr.， J. M. MCCOOL， J. KAUNITZ， C. 5. WILLIAMS， R. H. HEARN， .J. H.

Zeidler， E. DONG， Jr. and R. C. GOODLIN， Adaptive noise canceling: Principles and applicat.ions， Proc.

IEEE， 63， 1692-1716， 1975

WILLEBAND， J. 5.， G. H. PHILANDER and R. C. PACANOWSKI， The oceanic response {，O largc scale at.-

mospheric disturbances，よ Phys.Oceanogr.， 10， 411・429，1980

WYRTKI， K.， L. MAGAARD and J. HAGER， Eddy energy in t.he oceans，よ Geophys.Res.， 81、2641・264s，

1976

YAMAMOTO， T.， M. KUWASHIMA， M. HASHIMOTO， and N. KUMASAKA， Geomagnetic field measurement.s

at the ULF band in Izu-Oshima island (in Japanese)， Proc. 1989 CA sym.posium， 106・113，1989

YOI(OYAMA， Y. and T. YUKUl、AKE，5ixty year varia.t.ion in a. time series of t.he geoma.gnetic Ga.uss coef-

ficients between 1910 and 1983， J. Geom.ag. Geoelectr.， 43， 563-584， 1991

YUKUTAKE， T.， J. H. FILLOUX， J. SEGAWA， Y. HAMANO and H. UTADA， Preliminary Report. Oll a Ma民-

netotelluric Arra.y Study inもheNOl'thwest. Pacific， J. Geom.ag. Geoelectr.， 35， 575・587，1983

YUKUTAKE， T， and Y. HAMANO， Continuous measurements of the geomagnetic filed in the Pacific (in Ja.panese)， Kaiyo Monthly， 22，469・475，1990

YUMOTO， K.， Y. TANAKA， T. OGUTI， K. SHIOKAWA， Y. YOSHIMURA， A. ISONO， B. J. FRASER， F. W.

MENK， J. W. LYNN， M. SETO and 2100MM Magnetic Observation Group， Globally coordinated magnet.ic

observations along 2100magnetic meridian durillg STEP period : 1. Preliminary result.s of low-Iat.ituop

PC 3・s，J. Geom.α9・Geoelectr.，44， 261・276，1992

80 I.Fujii and ll.Ut.ada

地球規模の地電位差変動についての研究

藤井郁子・歌回久司

近年、国際電話通信に用いられてきた海底同軸ケーブルシステムが相次いで引退時期を

むかえたのに伴い、これらのケーブルを地球科学目的で利用することができるようになっ

た。すでに、太平洋では引退した七本のケーブルを用いた地電位差観測網が実現している。

この観測網により、大規模な地電位差変動の長期間にわたる観測が可能になった。

大規模な電位差の原因には、外部磁場変動による誘導、海流による誘導、地球中心核の

トロイダル磁場に伴う電場の漏れだし、の3種類が知られている。いずれも重要な情報を

含んでいるが、中でもトロイダル磁場に伴う電場はケーブルならではの観測量として注目

されている。

新しい観測網を活かすためには、まず一つ一つのケーブルでできることを検証すること

が不可欠である。本研究は，数千 km規模の電位差観測を行い，観測値の比較的周期の短

い成分に注目して、(1 )外部磁場変動による電場変動から二地点聞の海底の平均的な電

気伝導度構造を推定すること、 (2)海流による電場変動からケーブルを横切る海流の性

質を調べること、の二点を主な目的とする。短周期変動を詳しく調べることは微弱な中心

核からの電場を検出するうえで大きな助けになる。

これらの目的のため三本の太平洋横断ケーブルを使用した。そのうち、西太平洋にある

GN (グアム一二宮、全長 2700km)・GP(グアムーフィリピン、 2716km)の二本については、

1992年から継続して電位差観測を行った。 GNは給電されており他の二本に比べてデータ

の質が悪いが、観測項目を増やし解析を工夫することで研究に使用できることがわかった。

解析面では、 GN.GPに加えて、アメリカのグループと共同し HAW-l(ハワイーカルフォル

ニア、 4000km)を用いた研究にも参加した。

外部磁場による電位差変動の研究では、 GN・GPの電位差とケーブルの両端に近い3点

の地球磁場のデータに MT法を適用し、直交するこ方向についてフィリピン海プレートの

平均的な一次元電気伝導度モデルを求めた。 GNから推定されたモデルは海底下に 80kmの

低電気伝導度層を示しこれまでに知られているプレートモデルと似ているが、 GPから推

定されたモデルは低電気伝導度層の厚さが 300kmあり、低電気伝導度層とプレートのリソ

スフェアを対応させる考えからは受け入れ難い。 GNとGPのモデルに現われた著しい違い

は、 3次元的な海陸分布を取り入れた薄層近似モデリング(Makirdyet al.， 1985)と沈み

込むスラプの影響を考慮した2次元モデリング(Ut ada， 1987)で説明できることがわかった。

すなわち、表層の海陸分布が GPの電場をゆがめ一次元構造でみたときに低電気伝導度層

を厚くする効果を与えていること、 GNは海陸分布よりもスラブの沈み込みが影響してい

ること、が示唆された。結局、 GN・GP両方を満足するモデルとして、 GNで得られたタイ

プのモデルにスラブを加えたものが、フィリピン海プレートの電気伝導度分布として最適

であることがわかった。

011 Geoelect.rir Pot.ent.iol Vorio.t.ions Ovcr a Planetary Scalc

海流による電位差変動の研究では、ホノルルの磁場を参照して外部磁場による誘導分を

取り除いて、 3.6年分の海流による電位差変動を抽出することに成功した。次に、海流を

駆動する気象観測量と HAW-lの電位差の聞に高い相関が見られることを見つけた。周期5

-130日にわたり、ケーブルの電位差は ECMWFによる海面上の風、圧力と太平洋上の広い

範囲で最高 0.8の高い相関を示す。これまで数百 km以下のケーブルでは海水の流量と電

位差に比例関係があることが知られていたが、 4000kmのケーブルでも同様の現象がある

ことが示唆された。また、相関の高い地域・周期はデータの期間によって異なっており、

地球規模の気候変動を捉えている可能性がある。以上により、強い表面海流がない地域で

もケーブルによって海流による電位差変動が観測できることがわかった。海洋変動でこの

ような長期間の連続データが得られたのは非常に珍しく、海底ケーブルの今後の利用に新

しい可能性を示すことができた。

81

Documents

On Geoelectric Potential Variations Over a Planetary Se...地磁気観測所要領第29巻平成12年12月 Memoir冨oftheKakioka Magnetic Observatory柚1.29，December 2000 On Geoelectric