Structure of the upper mantle under the Antarctic ... | Wiley Online Library

35 downloads 75 Views 564KB Size Report
Dec 1, 2001 - achieved is now within reach. In the last decade an increasing interest has been devoted to regional-scale surface wave studies of the southern ...
GEOPHYSICAL RESEARCH LETTERS, VOL. 28, NO. 23, PAGES 4395-4398, DECEMBER 1, 2001

Structure of the upper mantle under the Antarctic Plate from surface wave tomography S. Danesi

and A. Morelli

Istit,ut,o Nazionale di Geofisicae Vulcanologia,00143 Roma, Italy Abstract.

l/uan et al. [1999]invertedRayleighwave dispersion

We have analyzed surface wave data ibr 245 events oc{'•n'ringin Antarctica and surroundingoceans during yea.rs 1991-1999. We first apply dispersion analysis to t,he fimdamental mode of Rayleigh waves for peri()ds ranging from 20 to 150 sec. We then invert pathaverageddispersioncurvesto obtain 2D maps of group velocity for different periods, representedby splineson a,grid with 250 km spacing,with considerablymore detail than that achievedby global models. We finally use gr6up velocity maps in a nonlinear inversionfor the regional 3D v•v structure of the upper mantle. The resulting model showsthick continental roots under East Al•tarctica, reaching depths below 200 km; low velocities •mder oceanicridges,mostly confinedto the upper 150 kill; and a slow area under the Ross Sea volcanic (:entres,extending all the way down to 200 km, beyond which our data loose resolvingpower.

measurementsto study the structure of the Scotia Sea

and Bannister et al. [2000]invertedsurfacewavesto (:ohstrain shear wave velocity under the Transantarctic

Mountains. Recenthigh-detailglobalscalephase [EkstrSmet al., 1997;vanHeijst and Woodhouse, 1999]and /41'{)•p velo(:ity [Larsonand EkstrSm,2001]dispersion maps a,ls(• sh()w a marked difference between East and West Antarctica, but lack the ability to follow smaller s(:a, le,st,r•mtures and do not model the distribution of vs in depth.

We present results from measurementand inversion ()f a large dataset of fundamental mode Rayleigh wave group dispersioncurves. We derive new, more detailed group velocity maps, and a 3D model of vsv of the upper mantle under the Antarctic Plate. We describethe {lispersionanalysis,the linear tomographic inversionto {'{)mlmt, e group velocity maps, and the nonlinear inversi{m i'{)r shear wave velocity structure. Our goal being t() image the main geologicalfeatures of the region, we {lis(:ussthe principal features of our model with respect

Introduction

In seismol{)gical studies,the southernhemisphereof l-JmEarth has alwayssufferedfrom low data availability, a fa{'t which hampersour knowledgeof the structureof its lithosl)here and, more generally, of its upper mantl(•. This situation is howeverimproving, thanks to the re(:ent general advancesin seismographicstation coverage of the globe, and higher detail than previously

t(• ge()h)gica,1 expectations.

Data and Dispersion Analysis

We analyze intermediate and long-period minor-arc Ra.yleighwavetrains. From the NEIC Cataloguewe extract earthquakeswith latitudes southof 30øSand magnitude M > 5.0 occurring during the years 1991-1999. achieved is now within reach. In the last decade an increasinginterest has been devoted to regional-scale Altogether we collect about 850 paths running between surface wave studies of the southern hemisphere. Af- 245 el)icentres a,nd 19 seismographicstations. We acter the work by Roult et al. [1994], who computed {:el)t,pa.thsfor which a dispersioncurve can be reliably group velocity maps for the Antarctic, severalstudies pi(:kedin the whole period range (20-150 s). Figure 1 have investigated the structural characteristicsof the showst,he path density of our dataset. We, is()la.te and examine the fundamental mode of the •tppermantle through analysisof surfacewaveswith imR,a.yleigh wavetrain by the iterative application of multiprovedresolution. Among others, Debayleand Ldv•que

[1997]a.pplied waveforminversionto the study of the structure of the Indian Ocean, including the South In-

ple filtersin narrowfrequencybands [Dziewonski et al., 1969]and a phase-matched filter [Herrin and Goforth, 1977; Herrmann, 1988]. We assumethat the source

{lia,nridges;Vdovinet al. [1999]performedgroupvelocity t{mmgraphyof South America and surrounding group delay (:anbe neglected,as shownby Levshinet al. oceans;Danesiand Morelli [2000]calculatedgroupve- [1999]. For eachrecord,we determinea path-averaged

group dispersion curve, for periods between 20 and 150 e•,tal. [2001]alsoperformedgroupdispersion analysisto se('(mds. Then we invert such curves for retrieving 2D maps modelling lateral variation of group velocity at study upper mantle structure. On a more local scale, different periods. We follow the technique illustrated

locity maps for the whole Antarctic region; Ritzwoller

Copyright2001 by the AmericanGeophysical Union. Papernumber2001GL013431. 0094-8276/01/2001GL013431

$05.00

in DunesiandMorelli [2000]althoughin this studywe adopt a. finer model representation the grid of linear spline nodes is regularly spacedby 250 km, rather than 500 km, in an orthographic projection. We add a 4395

4396 DANESI AND MORELLI- STRUCTURE

OF THE UPPER MANTLE UNDER THE ANTARCTIC

PLATE

(lepth (lependenceof shear wave sensitivity kernels, can l•(• inverted to compute a 3D v$ model. We parametrize the 'v..•. mo(lel, laterally, usingthe samegrid of the 2D inversion,fi•r each node of which we have group velocities •.t the l•erio{ts considered. For each node of the grid, we Mfild a reihrencemodel constituted by the vsv pro-

file.fi'om rmlially anisotropicPREM [Dziewonskiand Anderson,1981] as we useRayleighdata and by ('r•tst,a.1st,ructure and velocity at the top of the mantle

fi'{mtCRUST2.0 [Bassinet al., 2000]. Radially,below the {:rustal layering given by CRUST2.0 and a mantle slice t,o rea,ch 50 km, we use layers 35 km thick down to 355 kin, and 50 km thick below. As vsv sensitivity k{;rnelsdepend on the velocity model itself, the inversi{m is n{mlinear. We fbllow the procedure describedby

T,,ro, ntola and Valette[1982]whichsolvesthe nonlinear lea.st,S(l•a.res 1)roblem with an iterative scheme.

180'

0

2

5

Figure 1. Logarithmof path densityin the region underexamination.Ray coverageis constantwith period.

(weak)gradientminimizationconstraint,by whichwe re(pfirethe velocityperturbationAU/U betweenadjacent nodes to vary by less than 3%. We estimate a

(la.tavarianceibr groupvelocityat eachperiod•

by

the mean square misfit remaining after the inversion. For central-periods(40-140 sec) the relative rms veloc-

ity misfit rangesbetween1.0% and 1.5% comparedto the referencevalue. Shortest and longest periods are w()rsefit (1.7% ibr 35 sec,2.0% for 30 sec,1.6%for 145 a.nd 150 sec). Variance reductionover the best-fitting regionaluniibrm model varieswith period, and is 57% (-)11average.

Two sample group velocity maps, for periods of 30

The res•fitingmodel is shownat different depths by the ma,ps in the bottom panel of Plate 1. Starting from the top of the •lI)per mantle down to at least 150 km (lel)t,h,intense negative anomaliesfollow oceanicridges em:ir{:lingthe continent. In particular, we see a strong signa,1at, shallow depths below the Southeast Indian R,idgecorrespondingto the fastest spreadingridge of t,he a,rea.. At, a depth of 50 km, slow signals are also tn'esentunder triple junctions and under the Scotia Sea r(',gion.Slow anomaliesassociatedto oceanicridgesprogressivelyfiatten out beyond 150 km and they vanishat about 180-200 km in depth. Large amounts of basalts a.nd volcanicrocksfound in West Antarctica originated the hypothesisof the presenceof a mantle plume head which activated the evolution of the region. The discussi(m is still open 'and it invokeseither the mantle plume hyl){•thesis,or a significant lower lithosphere stretchin/4 (•r a. Cenozoic crustal extension as possible causes {:ontributingto the trigger and evolution of the West

Anta,rcticR,ifi,System [Behrendt,2000].In our model, a.nextended 'areain West Antarctica is characterizedby a.ln'{m{1slowvelocity pattern below 150 km, beneath the rift systemand beneath active volcaniccentresin Marie

and 100 s, are shownin Plate I (top panel). Group Byrd and Victoria Land. The Ross Sea slow anomaly velocity ()f shorter period Rayleigh waves(up to 40-50 reachesbelow 200 km, persisting deeper in the mantle sec) is strongly affectedby crustal thickness:oceans, than the slow pattern underneath ridges. This pattern with a thin crust, are faster than average,whereasslow anomaliesare due to the presenceof a thick continental

crust. Longer period wavesare also affectedby upper mantle structure: the pattern of group velocity for 100 sec R,a,yleighwavesrevealsthe existenceof cold continental roots beneath

the East Antarctic

shield and hot

ma,ntle below oceanicridges. An anomalouslywarm upl)er mantle underneaththe RossEmbaymentand Marie Byrd Land volcanic centresalso showsup with negative velocities.

Shear velocity three-dimensional

model

The inibrmation carried by the tomographicgroup velocity maps at various periods, by virtue of different

is consistentwith the hypothesisthat a mantle plume may have been originated beneath the craton and deflected towards a region characterizedby a weakerand

thinnerlithosphere[Behrendt,2000]. The East Antarctica craton has deep, fast (cold) rt•ots evident in all maps and reaching the depth of 200 kin: a. very sharp border marks the boundary between East and West Antarctica and it correspondsto the Transantarctic Mountains. A uniform high velocity pattern describesa stable shield which is not strictly confined beneath the continent, but extends in depth towa.rds the SouthwestIndian Ridge (200km map, Plate

1). The maximum amplitude and depth is reachedby the sectorbetweenlongitude 120øE and 150øE, around XVilkes Land.

D:\NESI

AND 5I()RELLI:

STRUCTURE

OF THE UPPER

MANTLE

UNDER

THE

ANTARCTIC

PLATE

4397

'100

O'

6.0

3.0

1.5

0.0



-1.5

-3.0

180'

11t0'

Plate 1. Top panel:Rayleighgroupvelocitymapsa,tperiodsT=30 and T-100 seconds; colorshowspercentage (h,viati•mfi'(m•the medianregionalvalue. Bottompanel: horizontalsectionsat differentdepthsof the shear v•,l•'itv •m•dellfi•tted as percentagedeviationfi'om anisotropicPREM.

'l-'1•' h()riz(mt,al

{listrilmtion

of standard

errors reflects

t-l•(' {l•t,;• (lensitv 1)a, ttern shown in Figure 1. ¾•rti{';•1(listril•td(m (lel)en(ts on sensitivity with depth. At 5() ]th.un{:ertaintiesm'e everywhere below 1.6% with •'{'Sl>('{:tt(> PR,EM veh)cities. Errors are drasti-

[layleigliwax'(,gro•tpvelocityto shearvelocity [Danesi .,,•d Mo•'clli.2()()()].Sm'fa,ce wavesare very sensitiveto ,'•'•st;•l str•t('tan'½' 1)ut (l() n()t easily a.llowto invert for it, s,, w(, n•s(-r(,lv ()[tlt(,

()n •xternal information. Inadequacies

('r•st al n•()(lcl may be s()•n'ce of some contamina-

{'•11x" n'{l•{','{l at intermediate depths (they vary in the •',>• ;• t}•(, sh•lh)w(•st nmntle. The broad loxv velocity th(-,('raton and the high speedarea of range 0.6%-1.3% at a depth of 100 kin) but increase I)•( (('•'• l)(u•(-,;•.th ;,gain at 200 km always lower than 1.9% ow- [}(,11i•t•sha.•s('n S('&•(3()kin map, Plate 1) in fact hardly with t,h('•mifin'm 1)attern of tomographicdata. i•a t{• t.l•e {l(•m'ea, sed sensitivity of fimdamental mode ('()•')'•,1•)(, ..

DANESI

AND MORELLI:

STRUCTURE

OF THE UPPER

These features •night perhaps be ascribedto an inexact (lescription of the crust. Note that change in crustal

thicknessby 1 km reflectsinto a 5v$/v$ variationat 50 km (lepth by about 0.3%.

Surface waves are the only choice for imaging the •q•per rnantle under a region such as the Antarctic, where earthquakes are concentrated along the ring of mid ocean ridges, and seismographicstations are mainly to few oceanic

UNDER

THE ANTARCTIC

PLATE

4398

Del)ayle,E. and J.J. L6vSque,Upper mantle heterogeneities in the Indian Ocean from waveforminversion, Geophys. R.cs. Lctt., 2• (3), 245-248, 1997. Dziewonski,A.M., S. Blochand M. Landisman,A technique fi)r the analysisof transient seismicsignals,Bull. Seismol. Soc. Am., 59, 429-444, 1969.

Dziewonski,A.M. and D.L. Anderson,Preliminary Reference Earth Model, Phys. Earth Planet. Inter., 25, 297-

Conclusions

lilnited

MANTLE

islands

or the coastline

of the

(:ontinent. Our study showsthat seismographicdata (:()verageibr the Antarctic region has resolution fair enough to tomographically i•nage •najor geologicfeat,re'es a.t a regional scale. There are however obvious possibilities ibr fi•ture improvement. With the available seismogrmns,depth resolutioncould be increased 1)ymodelling crustal thicknessand also waveswith different depth sensitivity such as Love wave data although their sensitivity to VSH imposes to invert for radial a.lfisotropy or surface wave overtones. An improvement of lateral resolution of tomographic modelsis lnore expensiveto obtain, as it passesthrough the installation of •nore seismographicstations. Very

few observatoriesexist in the continent, and significant ilnprovelnent may result from the installation of even few lnore permanent stations, and from long-term temp(•rary experiments with observatory-classinstrumentation.

356, 1981.

Ekstr6m G., J. Tromp and E.W.F.Larson, Measurements and global models of surfacewavespropagation, J. Geophys. R.cs., 102 B4, 8137-8157, 1997. Herrin, E. and T. Goforth, Phase-matchedfilter: application to the study of Rayleigh waves, Bull. Seismol. Soc. Am., , 6"/, 1259-1275, 1977.

Herrmann R.B., Computer Programs in Seisinology,St. Louis University, St. Louis, Missouri, 1988. Larson E.W.F. and G. EkstrSm, Global models of surface wavegroupvelocity,Pure Appl. Geophys.,158, 1377-1399, 2001.

Leeshin, A.L., M.H. Ritzwoller and J.S. Resovsky,Source effectson surfacewave group travel times and group velocity \naps, Phys. Earth Planet. Inter., , 115, 293-312, 1999.

Mooney, W.D., G.I. Laske and T.G. Masters, Crust5.1: A

globalcrustalmodel at 5ø x 5ø, J. Geophys.Res., 103 B1, 727-747, 1998.

Ritzwoller, M.H., N.M. Shapiro, A.L. Levshin, G.M. Leahy Crustal and upper mantle structure beneath Antarctica and surrounding oceans, J. Geophys.Res., in press. Iloult, G., D. Rouland and J.P. Montagner, Antarctica II: Upper-mantle structure from velocity and anisotropy, Phys. Earth Planet. Inter., 8•4, 33-57, 1994. Ta• antola, A. and B. Valett,e, Generalized Nonlinear Inverse Problems Solved Using the Least Square Criterion, Rev. G,ophys., 20 2, 219-232, 1982.

van Heijst, H.J. and J. Woodhouse,Global high-resolution

Acknowledgments.

We thank IRIS/DMC, G6oscope,

the Australian SeismologicalCenter AGSO and the Mednet Data Center ibr the availability of data, and R.B. Herrmann for •:omputer codes. This work was supported by the Italian Antarctic Program P NRA. References Bannister, S., R.K. Snieder and M.L. Passier, Shear-wave velocities

under the Transantarctic

Mountains

phase velocity distribution of overtone and fundamental-

ln()de surface waves determined by mode branch stripping, Gc.ophys. J. Int., 137, 601-620, 1999. V(lovin, ().Y, J.A. RiM, L.A. Levshin and M.H. Ritzwoller, Group velocity tomography of South America and the surroundingoceans, Geophys.J. Int., 136, 324-340, 1999.

\hmn, A., R. Cazzaro, G. Costa, M. Russi and G.F. Panza, S-wavevelocity models in the Scotia sea region, Antarcti('a, fi'om non-linear inversionof Rayleigh wave dispersion, Pure Appl. Geophys.,15•(1), 121-139, 1999.

and Terror

Riff, f¾omsurface waves inversion, Geophys.Res. Lett.,,

27 (2), 281-284, 2000. Bassin, C., Laske, G. and Masters, G, The Current Limits of Resolution for Surface Wave Tomography in North America, Eos Trans. A GU, 81 F897, 2000.

S. Danesi and A. Morelli, Istituto Nazionale di Geofisica e

Vnl('anologia, Via di Vigna Murata, 605 00143 Roma, Italy.

(e-mail: danesi{3ingv.it;[email protected])

http://mahi. ucsd.edu/Gabi/rem.html Behrendt J.C., Crustal and lithospheric structure of the ¾VestAntarctic Rift System from geophysicalinvestigations a review., Global Plan. Gh., œ3,25-44, 1999. Danesi, S. and A. Morelli, Group velocity of Rayleigh waves in the Antarctic Region, Phys. Earth Planet. Inter., 122, 1-2, 55-66, 2000.

(ReceivedMay 14, 2001; revisedAugust30, 2001; acceptedSeptember5, 2001.)