GEOPHYSICAL RESEARCH LETTERS, VOL. 23, NO. 17, PAGES 2275-2278, AUGUST 15, 1996

SeafloorMagnetotelluric Sounding Above Axial

Graham Heinson Schoolof Earth Sciences,Flinders University, Bedford Park, Australia

Steven Constable Instituteof Geophysicsand Planetary Physics, University of CaliforniaSan Diego, La Jolla

Antony White Schoolof Earth Sciences,Flinders University, Bedford Park, Australia

Abstract. Axial Seamountis a large,active, axis Instrumentation and Data locatedon the centralsegment of the Juande FucaRidge in Three setsof magnetotelluric(MT) instrumentswere de- the northeastPacific Ocean. Magnetotelluric (MT) datahave ployedfrom 3 ra July to the7 th September, 1994. The GPS- beencollected at threesites, approximately 4 km apartaround determineddeployment locations and depths(Plate 1) were the easternrim of the volcano,during a 65-day deployment. MT responses,in the bandwidth of 102- 105s, arealmost for Pele 45ø 55.47' N, 129ø 59.03' W (1550 m), Macques45 ø 57.48' N, 129ø 58.98' W (1520 m) andUlysses 45 ø 59.50' N, isotropic,with a weakly-definedprincipal direction of strike 129ø 59.81' W (1490 m). parallelto the maintopographic trends of Axial Seamount, Electric fields were recordedby the two-componentshort- and are relativelyflat over the wholebandwidth. Apparent resistivitiesare of the orderof 7 - 20 •m, andphases are as armed electrometerdescribed in Constableand Cox [1996], with the high-gainAC amplifiersreplaced by lower gain DC low as 30ø at the shortperiods. Diagonalterms of the MT amplifiers.The least-countwas 0.006/•V andthe samplerate tensorare an orderof magnitudesmaller than the off-diagonal 1 s. The magnetometers,developed at Flinders University terms,suggesting that three-dimensional effects on the data [White, 1979], areinstalled in standard17 inchglass instrument areminimal. Two-dimensionalinversions suggest that seafloor housings,which were swappedwith oneof two glassbuoyancy bathymetryand the distant coastlines have a surprisinglysmall floatsin the E-field package. The three-componentring-core effecton the MT response,and one-dimensional inversions fit the MT data to within the errorswith no serial correlationin the fluxgatesensors had a least-countof 0.1 nT and a samplerate set at 30 s. Additionally, a two-axis clinometerrecorded the residuals.A low crustalresistivity is themost robust part of the model,probably due to seawaterin fracturesand possibly due magnetometertilt with a precisionof 0.01ø every 78 minutes duringthe deployment. to a magmachamber. An electricalasthenosphere, although Three complete65 day recordswere obtainedfrom the elec- lesswell constrained,exists over a depthrange 30 - 60 km, trometers.Electrode potential drifts were rapidover the first 5 andthe resistivity of thisregion is compatiblewith about8% fraction of melt. to 10 days,of theorder of 20/•V/day, butsteadied to a relatively monotonicvariation of lessthan 5/•V/day for the remainder of the experiment. The magnetometerrecords ranged from Introduction 65 days at Pele, 10 days at Macquesand 4 days at Ulysses. Axial Seamount lies on the intersection of the Cobb- Tilt recordsindicated angles of magnetometerrepose of the EickelbergSeamount Chain andthe Juande FucaRidge, about orderof 1 - 5 ø in both x and y-orientations.Some ocean tidal 500 km west of the Oregon(USA) coast. The volcano(Plate variationsof the orderof 0.2ø were presentin the Ulyssestilt 1) is the shallowestpart of the Juande FucaRidge, indicating records,but Macquesand Pele showedlittle or no movement. a significantand robustsupply of magma [e.g. Delaney et al., For eachsite, time seriesdata were processedto obtainMT 1981], andis currentlyvolcanically and hydrothermally active, responsesusing the methodsand computer code of Egbertand with evidenceof recentlava flows,hydrothermal activity, and Booker [ 1986]. Data from the Victoria MagneticObservatory deformation[Embley et al., 1990; Fox, 1990]. Fowardmodel- were usedas a remotereference. Without the remotereference, ing and inversionof sea-surfacemagnetic anomaly data [Tivey uncorrelatednoise in thehorizontal magnetic field data reduces and Johnson,1990] indicatesa temperaturesof greaterthan the MT impedance,so that the apparentresistivity drops off 200øCat a depthof 700 m beneaththe ,suggesting the rapidly with the power spectraof the electricfield, although presenceof a magmachamber at3 kmdepth below the sum- phasesare less affected. MT response estimates were obtained mit,or a zoneof extensivesubsurface alteration. Hildebrand for sites Macques and Pele. The short duration of theUlysses et al. [ 1990] model sea-surfacegravity anomaliesand seafloor magnetometerrecord resulted in muchpoorer responses which gravitymeasurements to infer the existenceof a shallow,low- have not been used. densityregion. The low-densityregion could be melt, butother MT responseswere obtained in the bandwidth2x 102 to factors,such as porosityand mineralogy,are important. 6 x 104s; at shorter periods coherences dropped rapidly and at longerperiods source-field effects became dominant [Wanna- maker et al., 1989a]. The MT impedancetensor was thenro- Copyright1996 by theAmerican Geophysical Union. tatedto find the dominantstrike by the decompositionmethod of Lilley [1995]. Rotation angles were approximatelyfre- Papernumber 96GL01673 quencyindependent, and gave a predominantstrike of about 0094-8534/96/96 GL-01673505.00 20ø and 30ø westof geomagneticnorth for Pele and Macques

2275 2276 HEINSON ET AL.' SEAFLOOR MAGNETOTELLURIC SOUNDING

Depth,m

_ _ MT Response Interpretation ' -- -1400 Heinson and Constable[1992] arguedthat for an electri- cally resistiveoceanic lithosphere underlying the whole ocean . , _ ,( -1600 basin,seafloor MT datawill be significantlydistorted from 1D due to the coast-effect. Tarits et al. [1993] pointedout that / / i,•-ß -1700in the presenceof leakagepaths to more conductiveparts of the mantle, the coast-effect distortion was much reduced, and ,,½oo' ' -,oo suggestedthat subductionzones may representsuch a leakage • • ' • -20• path, aswas thoughtto be the casein theEMSLAB experiment [Wannamaker et al., 1989b]. Here we make the casethat Axial Seamountmay alsoact as a leakagepath to the asthenosphere. • -22• The evidencefor a low resistivitypath to themantle is three- fold. Firstly, the observedisotropic MT responsewould not be 45o50,' % : ,• ' -23•predicted from the coast-effectalone. AlthoughHeinson and Constable [1992] showed that isotropic MT responsesmay I • •• .. , -24•occur in an oceanbasin with no leakagepaths to the deepman- 4 , / -26• tle, it requiredthe site to be approximatelyequidistant to two or more orthogonalcoasts. For Axial Seamount,this is not -27• the case as it is much closer to the coastline of western North 45*40' •- _x t • -28• America (450 km) thanany othercoast. Secondly,apparent re- 229* 50' •0 ø •' 230 ø 10' 23• 20' sistivities in the TE orientation are smaller than similar seafloor

Plate 1. Site location of Axial Seamount and the three de- MT responseson the [Wannamakeret al., ployed MT instruments 1989a] and elsewhere.Finally, we notethat relativelyflat MT responseover two decadesof period suggeststhat the resisi- tivity structurebeneath Axial Seamountis relatively uniform respectively. These orientationsare most sensitiveto local with depth.To testthese hypotheses, we haveinverted the data seafloorbathymetry changes and the trendof the coastline. usinglD and 2D techniques. Figure 1 showsthe MT responses,with error estimates,at A 2D inversion,using the Occam'salgorithm of deGroot- Pele. The most notablefeatures of theseplots are that the MT Hedlin and Constable [ 1990], was used to examine the effect data are almost isotropic, and apparentresistivities are low, of ridge topographyand the North American coastline. A of order10 f•m. Differencesin thealong-axis MT responsesprofile was chosenperpendicular to the coast which passed (TE mode) at Macques and Pele are generally within the er- throughsites of Macquesand Pele. Inverting single-siteMT ror estimates. Across-axis(TM mode) apparentresistivities responsesfor 2D structuremay seem somewhatambitious. are slightly larger at Macques,but phasesare similar. Diag- However, bathymetryand coastlines,coupled with onal elementsof the MT tensorsat both siteshave apparent resistivity,provides additional constraints on the modeling. resistivitieswhich are an order of magnitudesmaller than the off-diagonalelements, from which we concludethat 3D effects LogDm on the data are minor. 0.0•. i -- 1.6

1.5

Depth, km 1.4 okm 0.5

axial 1.31.2

1.1 5.2 'lml 7,5 km ' 1.0 12kin 1.0 23 km 46 'hm 96 km 0.90.8 203 kan

4.34 km 1.5 0.7

931 km 0.6 2003 km -100 -51 26 84 149 229 325 379 599 0.5 Distance km 1 10 2.0D_ I • .... I 0.4 0.9 1.0 l.rl 1.2 1.3 Resiqtivity,Ohm. m [logscale] RMS Misfit Plate 2 2D smooth Occam's inversion of both Pele TE and Plate 3. 1D Occaminversions, with first derivative roughness, TM modesfor a transectpassing through the coastlineand of theTE responsesat sitePele for RMS valuesvarying from Axial Seamount. The transect has a strike direction of 90 ø east 0.9 to 1.3. The dashedline showsthe preferredmisfit level, of geographicnorth. The RMS misfit is 1.0. the profileis shownin Figure2. HEINSON ET AL.: SEAFLOOR MAGNETOTELLURIC SOUNDING 2277

Althoughthere are many more model parameters than data, 0.0 partsof themodel that are unresolved are simply determined 0.2 by the smoothingpenalty. The 2D inversionof Pele MT data is shownin Plate 2, and 0.4 givesa fit to theTE andTM responsesto within the error estimates(RMS = 1.0) withno serialcorrelation of residu- Eo6 als. The fit to MacquesMT datais marginallyworse (RMS 2nd Derivative = 1.7),but the final model is essentiallythe same as for Pele. Roughness -] -: Theresistivity structure directly beneath Axial Seamount is constrainedprimarily by the TE component,and we find that - Dashed Line seafloortopography is sufficient to generatethe observed dif- 1st Derivakive ferencebetween the TE and TM modesfor both sites. 12 Roughness Toexamine the resistivity structure beneath Axial Seamount o 1.4 - Solid Line morefully, the TE responseswere inverted for 1Dstructure us- - r] ingthe D + algorithm[Parker and Whaler, 1981]. The utility 1.6 ofthe D + algorithmisthat it givesthe minimum possible RMS misfit and thereforecan be usedto test whetherdata are com- 1.8 patiblewith a 1Dstructure. For Pele, the RMS misfit to the TE datawas 0.8, and for Macques the RMS misfit was 1.0. In both 2.0 'l cases,no significant trends of serialcorrelation in the model 10 ¸ residualswere present. Such model fits suggest that these MT Resistivity responsesare compatible with 1D structure. Smoothresistivity models, using the 1D Occam's algorithm Figure 2. 1D Occaminversions of theTE responsesat site [Constableetal., 1987],which fits the TE responseatPele to Pele. The solid line showsthe first derivativeroughness and an RMS misfitsof 1.0 is shownin Figure2. Here we show the dashedline showsthe secondderivative roughness. The profileswhen penalising both the first and second derivative of RMS misfits are 1.0 for both inversions. theroughness. The coincidence of of thetwo models supports theinterpretation over the depth range shown. 1D inversion of MacquesTE data produces similar structure. The most striking Tectonic Interpretation featuresin Figure2 arethe low resistivity in thecrust and the dropin resistivity over depths of 30- 60kin. To examine the The modelsin Figure 2 suggestthat the electricalresis- resolutionof theinversion, we show,in Plate3, inversionsfor tivity of the crustbeneath the summitof Axial Seamountis differentlevels of RMS misit,ranging from RMS = 0.9 which anomalouslylow, approaching 1 •m. Althoughwe have little is the smallestobtainable, to RMS = 1.3 at which all structure resolutionof the crust,as skin depthsat the shortestperiods disappears.Plate 3 clearlyshows that the most persistent part will be at least5 km, the MT responsesare sensitive to crustal ofthe profile is the low resistivity atshallow depths, which is structures. Absolute resistivities cannot be determined, how- still identifiableto RMS = 1.2. However,the mid-mantlefall ever the crustalconductance (defined as the depth-integrated in resistivityis lesswell constrained by thedata. conductivity)is relaivelyrobust to theRMS misfitin Plate3, and can be estimated to be of order 1,000 S. This is consider- ablyhigher than expected for oceaniccrust away from a ridge or seamount.For example,the in situmeasurements of Becker 20 2D Occams Inversion 2D Occams Inversion et al. [ 1982] overthe top 2 kin, suggesta crustalconductance TE Mode TM Mode to be at aboutan order of magnitudesmaller. The mechanismproducing high crustal conductance is most • 10 probablydue to a fluid fraction,either seawater in fractures and/ormolten rock in a magmachamber. Assuminga fully interconnectedfluid fraction, which is equivalentto using Archies'Law with an exponentof 1.2 [Evans,1994], andfluid conductivityof 0.31 •m, the conductanceof a 6 km thick I i i i Iiiii i i i i iiiii 60 crustrequires a bulk porosityof 9 %. This is not unreason- ablefor hydrothermalfluids in highlyfractured basalt over the 50- top few hundredsof metres. The electricalresistivity of the hydrothermalfluids also decreases with temperature. The pres- ence of additionalmelt can neitherbe excludedor supported 30 as the thermal structureand porosityof Axial Seamountare unknown. 20- An increasein resistivitybelow the crust to 20 •m at 10- 10- 30 km depthis an importantfeature if it canbe constrainedby thedata. Sucha risein resistivitysuggests that significant melt o I I I IIIIII [ I • [][[ll I I I IllIll t , t lO z •0 3 104 •0 :t •0 3 104 fractionsdo not connectwith a deepermantle melting source. Period (s) Period (s) From geochemicaland petrologicalarguments, only a small Figure1. ObservedPele TE andTM modeMT responseswith amountof melting is expectedin the plagioclaseperidotite errorestimates, and the modeledresponses (solid lines) from stabilityfield at depthsless than 20 km below the seafloor the 2D section shown in Plate 2. The RMS misfit is 1.0. [Elthon, 1989]. High temperaturesat or slightlybelow the 2278 HEINSON ET AL.: SEAFLOOR MAGNETOTELLURIC SOUNDING

mantlegeotherm of approximately1280øC will give rise to Delaney,J.R., H.P. Johnson,and J.L. Karsten,The Juande Fuca resistivitiesof theorder of 50 •m [Constableet al., 1992]. Ridge--propagating rift system:New tectonic,geo- We lack the dataquality and spatialdistribution to constrain chemicaland magneticdata, J. Geophys.Res., 86, 11,747- this layer moreprecisely. 11,750, 1981. The modelin Figure2 hasa fall in resistivityover a depth Egbert, G.D. and J.R. Booker, Robustestimation of geomag- netic transfer functions, Geophys. J. R. Astr. Soc., 87, intervalof 30 - 60 km in the mantle,which is weaklycon- 173-194, 1981. strainedby theMT responses.A resistivityof the orderof 5 - Embley, R.W., K.M. Murphy and C.G. Fox, High-resolution 10 •m suggestshigh temperatures and possibly a small melt studiesof the summitof Axial Volcano,J. Geophys.Res., fraction. Geochemicaland petrological studies roughly con- 95, 12,813-12,822, 1990. strainthe depthswhere most basaltic melt is generated.Small Elthon, D., Pressureof origin of primary mid-oceanridge amountsof meltingoccurs in thegarnet lherzolite stability field basalts,in Magmatismin the OceanBasins, edited by A.D. (65 - 80 km depth),but most melting occurs in thespinel peri- Saundersand M.J. Norry, GeologicalSoc. Spec. Pub.,pp. dotitestability field (20 - 65 km) [e.g.Elthon, 1989]. MORB 125-136, 1989. geochemicaland petrological studies suggest that an average Evans, R.L., Constraintson the large-scaleporosity and per- of 10 - 15% of the sub-ridgemantle melts, with a maximum meability structureof youngocean crust from velocity and resistivitydata, Geophys.J. Int., 119, 869-879, 1994. meltextraction of 20 - 30%. Someauthors [e.g. Sotin and Par- Fox, C.G., Evidenceof activeground deformation on the mid- mentier,1989] suggestthat the melt is rapidlydrained from the oceanridge: Axial Seamount,, April- mantleand less than 3% is presentat anytime. Assuminga June 1988, J. Geophys.Res., 95, 12,813-12,822, 1990. maximummelt percentageof 30%, and usingArchies' Law Heinson,G., and S. Constable,The electricalconductivity of withan exponent of 1.2,the resistivity could approach 1 •m, the oceanicupper mantle, Geophys.J. Int., 110, 159-179, but for 3% melt the correspondingresistivity is 20 •m. An 1992. 8% meltfraction gives resistivities of 15•m, compatiblewith Hildebrand, J.A., J.M. Stevenson, P.T.C. Hammer, M.A. Zum- the 1D inversions. berge,R.L. Parker, C.G. Fox and P.J.Meis, A seafloorand sea surfacegravity surveyof Axial Volcano, J. Geophys. Conclusions Res., 95, 12,751-12,763, 1990. Lilley, F.E.M., Strike direction: obtainedfrom basicmodels for Seafloor MT data collected on the summit of Axial 3D magnetotelluricdata, in Three-dimensionalelectromag- Seamountsuggest that the electricalresisitivity of the crust netics, editorsM. Oristaglio and B. Spies, Schlumberger- is low, and that there is an additional low resistivity layer at Doll Research,Ridgefield, Conn., USA,pp 359-369, 1995. 30 - 60 km depth. Electric currentsinduced in the sea are Parker, R.L. and K.A. Whaler, Numerical methods for estab- thereforeable to leak into the conductiveparts of the mantle. lishing solutionsto the inverseproblem of electromagnetic induction,J. Geophys.Res., 86, 9,574-9,584, 1981. Supportingevidence for this comesfrom almostisotropic MT Sotin, C. and E.M. Parmentier,Dynamical consequencesof responseswith minimal 2D or 3D effectsin spiteof the close compositionaland thermal density stratificationbeneath proximityof the coastline.The MT dataare fit to the observed spreadingcentres, Geophys. Res. Lett., 16, 835-838, 1989. errorsby 1D inversion,and bulk properties,such as the per- Tarits, P., Chave, A.D. and Schultz, A., Comment on 'The centageof eithermelt or hydrothermalfluids, can be estimated electricalconductivity of the oceanicupper mantle' by G. from invertedresistivities. Average crustal porosity in the up- Heinsonand S. Constable,Geophys. J. Int., 114, 711-716, per 6 km is about9%, and over a depthrange of 30 - 60 km 1993. resistivitiesmay be associatedwith a melt fractionof 8%. Tivey, M.A. and H.P. Johnson,The magneticstructure of Ax- ial Seamount,Juan de Fuca Ridge, J. Geophys. Res., 95, Acknowledgments. We thankthe captainand crew of theR/V 12,735-12,750, 1990. Wecoma. Tom Deaton, JacquesLemire, BrentonPerkins and Wannamaker,RE., et al., Magnetotelluricobservations across Bob Walker providedtechnical support, and financialsupport theJuan de Fucasubduction system in theEMSLAB project, came from the Australian Research Council, the National Sci- J. Geophys.Res., 94, 14,111-14,125, 1989a. ence Foundation, and a Japan Society for the Promotion of Wannamaker, P.E., J.R. Booker, A.G. Jones, A.D. Chave, J.H. ScienceFellowship. The CanadianGeological Survey kindly Filloux, H.S. Waft and L.K. Law, Resistivitycross section suppliedmagnetic observatory data. throughthe Juan de Fuca ridge subductionsystem and its tectonicimplications J. Geophys.Res., 94, 14,127-14,144, 1989b. References White, A., A seafloormagnetometer for the continentalshelf, Mar. Geophys.Res., 4, 105-114, 1979. Becker,K., et al., In situ electricalresistivity and bulk porosity of the oceanic crust, Costa Rica Rift, Nature, 300, 594-598, 1982. Graham Heinson, School of Earth Sciences, Flinders Uni- Constable, S.C., R.L. Parker and C.G. Constable, Occam's in- versity, Bedford Park, SA 5042, Australia. (e-mail: gra- version:a practicalalgorithm for generatingsmooth models ham.heinson@ flinders.edu.au) from electromagneticsounding data, Geophysics,52, 289- StevenConstable, University of CaliforniaSan Diego, Scripps 300, 1987. Institutionof Oceanography,Institute of Geophysicsand Plane- Constable, S., Shankland, T.J. and Duba, A., The electrical tary Physics0225, La Jolla, CA 92093-0225. (e-mail: sconsta- conductivityof an isotropicolivine mantle, J. Geophys. [email protected]) Res., 97, 3,397-3,404, 1992. Antony White, School of Earth Sciences, Flinders Uni- Constable, S., and C.S. Cox, Marine controlled-source elec- versity, Bedford Park, SA 5042, Australia. (e-mail: tromagneticsounding 2. The PEGASUS experiment,J. mgaw@ es. flinders.edu. au) Geophys.Res., 101, 5519-5530, 1996. deGroot-Hedlin, C. and S. Constable, Occam's inversion to generate smooth, two-dimensionalmodels from magne- (ReceivedJanuary 31, 1996;revised March 29, 1996;accepted totelluricdata, Geophysics,55, 1,613-1,624, 1990. May 17, 1996.)