GEOPHYSICAL RESEARCH LETTERS, VOL. 20, NO. 20, PAGES 2167-2170, OCTOBER 22, 1993
GPS MEASUREMENT
OF RELATIVE
ACCUMULATION
MOTION
OF THE COCOS AND CARIBBEAN
ACROSS
THE MIDDLE
AMERICA
PLATES
AND
STRAIN
TRENCH
Timothy H. Dixon
RosenstielSchoolof Marine and AtmosphericSciences
Universityof Miami
Abstract.Global PositioningSystem(GPS) measurements
in 1988 and 1991 on CocosIsland (Cocosplate), San Andres
Island (Caribbean plate), and Liberia (Caribbean plate,
mainlandCostaRica) providean estimateof relative motion
betweenthe CocosandCaribbeanplates. The datafor Cocos
definition [KornreichWolf et al., 1990; Kellogg and Dixon,
1990]. In the absenceof such networks, baseline-length
dependenterrors can be significant [Larson et al., 1991],
obscuringthe geologicsignalswe seekto measuregiven the
long distances,nearly 1000 km in the caseof the Cocos-San
and San Andres Islands, both located more than 400 km from
Andresbaseline.Freymuelleret al. [ 1993] reportresultsfrom
the Middle America Trench, define a velocity that is
theseexperimentsusing the "fiducial approach"in which a
equivalent within two standard errors (7 mm/yr rate, 5
subsetof the trackingsitesis fixed to predeterminedvalues.
degreesazimuth)to the NUVEL-1 plate motion model. The
One disadvantageof this approachis that systematicerrorsin
data for Liberia, 120 km from the trench, define a velocity
locationsof the fiducial sitespropagateinto satelliteorbit and
that is similar in azimuth but substantiallydifferent in rate
site positionestimates,leadingto systematicerrorsin the site
from NUVEL-1. The discrepancycan be explained with a
velocities of interest. This study uses the "no-fiducial"
simplemodelof elasticstrainaccumulation
with a subduction approach[Herring et al., 1991; Blewitt et al., 1992; Heflin et
zonethat is lockedto a relativelyshallow(20-Z_5
km) depth.
al., 1992] with the positionsof all sitesestimated,avoiding
systematicerror from fiducial site mislocation. Estimated
Introduction
positionsare subsequently
translated,rotatedandscaledinto a
stablereferenceframe (in this casethe Intemational Terrestrial
Global plate motion modelssuchas NUVEL-1 [DeMets et
al., 1990] predict relative plate motion in subductionzones
basedprimarily on closureconstraints
from the globalplate
circuit. Subduction zones provide no direct rate data to
constrainplate motionmodels,while the directionalconstraint
providedby trenchearthquakeslipvectorsis relativelyweak.
Slip vectorazimuthscan exhibita high degreeof scatter,and
may be systematicallybiased due to refraction of seismic
wavesby the cold slab or tectonicconsequences
of oblique
convergence[DeMets et al., 1990]. It is thereforeuseful to
testthe predictionsof globalmodelsfor subduction
zoneswith
spacegeodeticmeasurements.Complicatingfactorsinclude
precisionrequirements,
the greatdistances
necessary
to span
the deformingzone,andthe paucityof suitablegeodeticsites
on the subductingplate. Also, elasticstrainaccumulationand
releaseduring the earthquakecycle and permanentcrustal
deformationmany hundredsof kilometersfrom the trench
complicateinterpretation
of geodeticresults.
The CASA (Central and South America) experiments
investigatethese and related questionswith GPS geodetic
measurementsinvolving the Cocos, Nazca, Caribbean and
ReferenceFrame, ITRF-90) for comparisonof data acquired
at different epochs. Despite these differences,the results
reportedhereandby Freymuelleret al. [1993] arevery similar
and overlap within uncertainties(quoted here at twice the
within several hundred km of the Middle America Trench.
subduction zone even for much shorter time scales. While
standarderror).
Data were analyzed using the GIPSY software [Blewitt,
1989; Lichten, 1990] following analyticalconditionslistedin
Dixon et al. [ 1993] exceptthatpolarmotionwasestimatedand
the no fiducial approachwas used. Velocitiesand errorsare
based on weighted least squaresfits throughposition data,
with weightsfrom formal errorsscaledby day to day scatter.
Resultsare summarizedin Table 1 andFigures1 and 2.
Discussion
The velocityof SanAndresIsland(Caribbeanplate) with
respectto CocosIsland (Cocosplate) is 88+7 mm/yr at an
azimuth of 202+11 ø, identical within errors to the NUVEL-1
prediction. Since both sitesare well away from the neartrench deforming zone, we conclude that the NUVEL-1
SouthAmericanplates[KelloggandDixon, 1990]. I report model, despitelack of strongdata constraintsin subduction
hereresultsquantifyingconvergence
betweenthe Cocosand zonesanddespitebeingan averageoverseveralmillionyears,
Caribbeanplatesanddocumenting
elasticstrainaccumulation is an accuraterepresentation
of the far field motionacrossthis
Data AnalysisandResults
Major CASA GPS campaignswereconductedin 1988 and
1991. Eachcampaignincludedthe establishment
of a global
tracking network for satellite orbit and reference frame
Copyright1993by the AmericanGeophysicalUnion.
additional GPS measurementshere are important for
investigating
possiblebiasesin thedataor modelat thelevelof
a few per cent, the general agreementbetween GPS and
NUVEL-1 velocitiesis an importantresult.
In contrastthe velocityof Liberiaon thewestcoastof Costa
Rica relativeto CocosIslanddifferssignificantlyfrom the
NUVEL-1 prediction.While thevectorazimuthagreesat the
two standarderror level, the GPS rate (70+5 mrn/yr) is
significantlylessthan the NUVEL-1 rate (85+6 mrn/yr at
Liberia). There are four plausible explanationsfor the
discrepancy:1. theGPSresultis affectedby systematic
error;
Papernumber93GL02415
0094-8534/93/93GL-02415503.00
2167
2168
Dixon: GPS Measurement,Cocosand CaribbeanPlates
TABLE
lOO
1: Velocities Relative to Cocos Island 1
Cocos-Caribbean
(NUVEL-1)
Liberia
Albrook
San Andres
NUVEL
[•
Rate
70_+5
75+8
88_+7
203_+ 10
202_+ 11
80
89_+6
(mm/yr)
o
Azimuth
209+8
203_+3
o
(øEastof North)
60
o
1. NUVEL-1 vector (Caribbean-Cocos) at San Andres, from
DeMets et al. [1990] with 4% rate decrease for recent
ß•
/ /'":,.'
......
---0-25km
40
adjustmentto paleomagnetictime scale [DeMets, personal
communication]. Uncertainties are twice standarderror.
m
2. there is crustal shortening and permanent deformation
inland (northeast)of Liberia; 3; the GPS result is affectedby
coseismicoffsets; 4. elastic strain is accumulatingnear the
trench,to be releasedlaterin an earthquake.
I reject (for now) the first explanation,for severalreasons.
First, the analyticalapproachusedis designedto eliminatethe
largest known source of systematic error, namely the
systematic error associatedwith fiducial site mislocation.
Second,the result for the San Andres site is plausiblein the
context of other information such as the NUVEL-1
model,
suggestingthat systematicerrors are not dominating our
solutions. Third, the "no fiducial" techniquegives a general
indicationof data quality from differencesbetween"known"
tracking site coordinatesdefined by a given referenceframe
and estimated
coordinates
mm/yr
-•
1ON -
o
I
{ SAN
LIBERIA
• •
••
I
ANDRES
/
_//
/•
o•
L,
iberi•a•
, ,
0
1991
• •
400
600
Distance from Trench (km)
Fig. 2. GPS ratesand errors(2o) for sitesLiberia and San
Andres (Caribbean plate) relative to Cocos (Cocos plate).
Two data points at Liberia are same data analyzed with
fiducial [Freymuelleret al., 1993] andno fiducial(this study)
approaches.Curvesshowpredictedvelocity for elastichalf
spacemodelof strainaccumulationabovelockedsubduction
zone [Savage,1983] dippingat 15ø for threelockingdepths
(surface to 15, 25 and 35 km). Kinks show surfacelocation
abovetransitionfrom lockedbehaviorto aseismicslip.
•5•
•
•
II
•
• I
PANAMA
BLOCK
NAZCA
sitevelocityestimatesin the CASA regionby morethana few
mm/yr. Fourth, Freymuelleret al. (1993) report a result for
the Cocos-Liberia
80W
Fi•. 1. Sketchmap with plate or block bound•ies, predicted
motion (NUVEL-1, light lines, arrow) and observedmotion
(GPS, bold lines, arrow) o• selected sites (squares with
crosses)includin• San AndrosandLiberia (Caribbean
relative to Cocos(Cocosplate). Ellipsesat 2 standarde•ors.
Stars are •s>7.0 well located earthquakes(1950, •s=7.7;
1983, •s:7.2;
1991, •s:7.6).
CDT is Cordillera de
Talamanca, PN is Peninsulade Nicoya, PO is Peninsulade
Osa. Shaded area (west end, Panamablock) is diffuse shear
zone postulatedby Goeset al. [1993] to mare boundaryo•
Panamablock;lined •ea is co•espondin•location•rom Jacob
et al. [1991]. •odified •rom Goes et al. [19931.
baseline that is identical within errors to the
result reported here. Their result is based on three
experiments(1988, 1990, 1991) analyzedusing the fiducial
approach(we were unable to processthe 1990 data with the
no fiducial approachbecauseof the limited trackingnetwork
for that period). Without additionalobservationswe cannot
eliminate the possibility of a blunder or other local effect
specificto Liberia, but available evidencesuggeststhat the
velocityestimatereportedhereis realistic.
Significant crustal shorteningnortheastof Liberia can be
eliminatedon geologicgrounds.The discrepancy
(15 mm/yr),
if accommodatedby crustal shorteningover geologictime,
would resultin an activefold and thrustbelt with significant
topographicelevation, which is not observed. While such
deformationis clearly going on north of Panamaoffshore
alongthe North PanamaDeformedBelt (discussed
below) and
in southeastern
85W
200
for these sites based on the GPS
data. These differences(for 7 global sitesin 1988 and 12 in
1991) aretypicallylessthan30 millimeters,too smallto affect
'6 306090
20
Costa Rica in the Cordillera
de Talamanca
(Figure 1), it doesnot appearto be occurringin the "backarc"
regionbehindLiberia. Small amounts(a few mm/yr or less)
of suchdeformationcannotbe precluded,but the explanation
for the majorityof the discrepancy
mustbe soughtelsewhere.
The only significant earthquake between the GPS
measurementswas a Ms=7.0 earthquakein 1990, a shallow,
thrustingsubductionzone eventlocatednearthe trenchabout
100 km southof Liberia (M. Protti,personalcommunication).
Elastic dislocationmodels(Freymuelleret al., 1993) suggest
-20 mm of southward motion at Liberia, too small and in the
wrongdirectionto explainthe discrepancy.
Elasticstrainaccumulation
is a plausiblemechanism.Strain
accumulation models in many subductionzones satisfy a
varietyof geologic,seismicandgeodeticdata. Lockingat the
Dixon: GPS Measurement, Cocos and Caribbean Plates
thrustinterfacebetweensubductingand overridingplatesis
assumedto occur in the high strengthbrittle upper crust,
transferringstrain (and stress)to the overridingplate, to be
releasedat somefuturetime in an earthquake.I havetested
thishypothesis
with simpleforwardmodels,comparing
GPSbased
velocities
to
the
dislocation
model
for
strain
accumulation of Savage [1983]. The locked condition is
simulatedby addinga supplementary
solution(normalslip at
theplaterate)to steadystatesubduction
on thelockedportion
of thethrustzone,whicheverywhereelseslipsaseismically
at
the full plate rate. Surfacevelocitiesfor normal slip in an
elastic half spacewere obtainedfollowing Savage [1980,
Equation115]. Inputparameters
for themodelarethe far field
convergencerate, from NUVEL-1, the depth limit of the
lockedzone,variedbetween15 and35 km, andthe dip of the
locked zone. From the trench inland several tens of km,
seismicreflectionindicatesvery shallow(5ø-10ø) dipsfor the
decollementseparatingsubductingand overriding plates
[Shipley et al., 1982]. Within 40-50 km of the trench,
seismicitybeginsto definethe plateinterface,and showsthat
dipssteepento -10ø-25ø. This dip is maintainedto depthsof
50-60 km, then steepensfurther to -60 ø [Burbachet al.,
1984]. The intermediateportionof the plate interfaceis of
interesthere,consequently
a dip angleof 15øis employed.
Model resultsare shownin Figure2. The GPS datacanbe
explained by elastic strain accumulation with the thrust
interface between the plates locked to shallow (<25 km)
depths.We cantesttheplausibilityof thismodelnotonlyby
its abilityto fit the sparseGPS data,butby theextentto which
the modeleddepthof the lockedzoneis consistentwith other
data,e.g., earthquakedepths. If we consideronly thrusting
earthquakes
thatrupturetheinterfacebetweensubducting
and
overridingplatesandnot deepereventswithin the slab,then
themaximumearthquake
depthshouldcorrespond
to thebase
of the lockedzonein themodel. Typicalseismogenic
depths
for large(Ms>7.0) interplatethrustingearthquakes
alongthe
Middle America Trench are less than 25 km [Singh and
Mortera, 1991;Burbachet al., 1984]in goodagreementwith
the model. A complicatingfactor for Costa Rica is that
shallowinterplatethrustearthquakes
arerare;the25 km depth
cutoffis betterdefinedonthebasisof earthquakes
to thenorth
near Nicaraguaand southernMexico [Singh and Mortera,
1991; Burbach et al., 1984]. While shallow thrusting
earthquakes
have occurredon the CostaRica segmentof the
trenchin the past,mostrecentlyin 1950 (Ms=7.7; Figure 1),
the depthsof theseeventsare poorly constrained.Smaller
thrustearthquakesoccurredin 1990just southof the Nicoya
region (Ms=7.0, depth 20 km)and in 1983 off the Osa
Peninsula(Ms=7.2, depth24 km) [Jacobet al. 1991] (Figure
1). In summary,available seismicdata suggestthat large
interplate thrusting earthquakesin this part of the trench
nucleateabove25 km, in agreementwith the strainmodel.
The Panama Block
The Panama block is defined mainly on the basis of the
Northern Panama Deformed Belt (NPDB), a series of active
shallow folds and thrust faults in the Caribbean basin offshore
northernPanama(Silver et al., 1990;Figure 1). GPS datafor
a singlesitein Panama(Albrook;Figure 1) showthatmotion
of the Panamablock relative to the Cocosplate is similar in
2169
directionto the Caribbeanplate but slower,probablydue to
compressionacrossthe NPDB (-13+10 mm/yr NNE-SSW
based on the difference
between the Albrook
and San Andres
velocity vectors). The April 1991 Ms=7.6 Costa Rica
earthquakeoccurredon a shallowsouthwestdippingthrust
fault associated with the onshore extension of the NPDB
in
Costa Rica (Figure 1) [Plafker and Ward, !992]. Could
deformationin thisbelt extendwest,affectingtheLiberiasite?
Aftershocks
for the 1991 event define a diffuse zone with
left-lateralstrikeslipmechanisms
trendingsouthwest
fromthe
main shock[Dziewonskiet al., 1992] andthereis geological
evidencefor a complexzoneof strikeslip faultingnear the
termination of the rupture zone along its northwestedge
[Ponceand Case, 1987]. These data led Goeset al. [1993] to
suggestthat the westernboundaryof the PanamaBlock is
configuredasshownin Figure1. Jacobet al. [1991] suggest
thatthisboundarytrendsmorewesterly(Figure1) throughthe
Valle Central of Costa Rica, on the basis of left lateral strike
slip focal mechanisms
for threeearthquakes
(M-7.0, 1924;
M=6.0, 1983; M=5.9, 1991).
In either case, Liberia is
locatedconsiderably
northof thepostulated
boundary(Figure
1). Thus deformation of the North Panama Deformed Belt
and motion
of the Panama
block
should not affect
our
interpretationof the Liberia GPS data.
Seismic Hazard
The resultsof this studysuggestthat at leastonesegmentof
the subduction
interface
between
the Cocos and Caribbean
platesis currentlylockedand accumulatingelasticstrain. The
Nicoya segmentof the Middle America Trench has broken
repeatedlywith large (M>7.0) earthquakes,includingevents
in 1827, 1853, 1863, 1900 (Ms-7.2), 1916 (Ms-7.4), 1939
(Ms=7.3) and 1950 (Ms=7.7) [Nishenko, 1991]. All of these
eventsarebelievedto be shallow(<25 km) interplatethrusting
earthquakes. While the last major earthquakethat fully
rupturedthis segmentof the trenchwas in 1950, more recent
events, including two Ms=7.0 eventsin 1978 and a Ms=7.0
event in 1990, have also occurrednear the Nicoya region
[Jacobet al., 1991]. Nishenko [1991] givesthe recurrence
intervalfor Ms_>7.0eventsas22+2 years,andliststhispartof
the Middle America Trench as a seismicgap with a 64%
chanceof rupturingin the period 1991-2001. The 1990 event
wasjust southof the Nicoya segment.
Kagan and Jackson[1991] suggestthat large subduction
zone earthquakesclusterin time ratherthan exhibit periodic
behavior. Ward [1991] suggeststhat both characteristic
(singlefault segment)and non-characteristic
(multiple fault
segment)rupturescould occurin the Middle America Trench
dependingon segmentinteraction. Thus while evidenceof
strain accumulation presentedhere confirms the obvious
seismichazard,by itself this is unlikelyto leadto an accurate
forecastof eitherthesizeor timingof a futureearthquake.
McNally and Minster [ 1981] notedthat seismicslip from
regulartrenchearthquakes
of the lastcenturyis muchlessthan
the total sliprequiredby globalplatemotionmodels.The data
presentedhereconstrainpossibleexplanations
for thisdeficit.
First, the GPS resultsrule out systematic
errorsin the plate
motion models. Second,if the deficit is causedby some
fractionof plate motionbeing accommodated
aseismically,
thenit is moredifficultto reconcilethedepthdatafor shallow
2170
Dixon: GPS Measurement,Cocosand CaribbeanPlates
thrusting earthquakesin the region with the locking depth
required by the GPS data, at least with the simple strain
accumulationmodelspresentedhere. While modelscan be
generatedthat match the sparseGPS data and allow some
aseismicslip (in Figure2, thesemodelswouldhavea velocity
stepat thetrench)therequireddepthlimit for partiallockingto
match the GPS data is deeperthan 25 km, i.e., deeperthan
observedinterplateseismicity.Alternately,aseismicslipis not
the explanationfor the slip deficit, implying that strainmay
accumulatethroughseveralseismiccycles,and characteristic
Jacob, K. H., J. Pacheco, and G. Santana, Costa Rica
earthquake
reconnaissance
report,Ch. 2, Seismology
and
Tectonics,EarthquakeSpectra,7, Supp.B, 15-33, 1991.
Kagan,Y. Y. andD. D. Jackson,Seismicgaphypothesis:
ten
yearsafter,J. Geophys.Res.,96, 21419-21431,1991.
Kellogg,J., andT. Dixon, CentralandSouthAmericaGPS
geodesy-CASA,Geophys.
Res.Lett., 17, 195-198,1990.
Kornreich Wolf, S., T. H. Dixon and J. Freymueller, The
effectof trackingnetworkconfiguration
on GPS baseline
estimatesfor the CASA Uno experiment,Geophys.Res.
Lett., ! 7, 647-650, 1990.
earthquakes
of the historicalrecordhavenot succeeded
in
releasingthis accumulatedlong term strain. If correct,this
Larson,K. M., F. Webb andD.C. Agnew,Applicationof the
explanation
requiresthattheslipdeficitwill be madeup in the
futureby earthquakeslargerthan characteristic
eventsof the
GPS to crustal deformation measurements 2. The influence
past. Stein et al. (1986) suggesteda similar explanationfor
the great 1960 Chile earthquake,since seismicslip in that
event was significantlylarger than that implied by simple
multiplicationof recurrence
intervalby platerate.
Res., 96, 16567-16584, 1991.
Acknowledgments.
The CASA experiments
representthe
effortsof hundreds
of people,butI wouldlike to mentionJim
Kellogg, Jeff Freymueller,Ruth Neilan, SteveFisher,Jim
Richardson,SusanKornreichWolf, David Townsend,Victor
Vega andPaulLundgren.I thankJeanneSauberandDon
Argusfor advi•ce
on subduction
zonemodels,andSethStein,
ChuckDeMets andMarino Prottifor commentson thepaper.
of errors in orbit determinationnetworks, J. Geophys.
Lichten,S. M., Estimation
andfilteringfor highprecision
GPS applications,
Man. Geod.,15, 159-176,1990.
McNally, K. C. and J. B. Minster,Nonuniformseismicslip
ratesalongtheMiddle AmericaTrench,J. Geophys.
Res.,
86, 4949-4959 1981.
Nishenko, S. P., Circum-Pacific seismic potential' 19891999, Pageoph,135, 169-259, 1991.
Plafker, G., and S. Ward, Backarcthrustfaulting and uplift
alongthe Caribbeanseacoastduringthe April 22, 1991
CostaRica earthquake,Tectonics,11,709-718, 1992.
Ponce,D. A., and J. E. Case, Geophysicalinterpretationof
Costa Rica, in: Mineral Resources Assessment of the
References
Republicof CostaRica, U.S. Geol. Surv.Misc. Invest.
Blewitt,G., Carrier-phase
ambiguityresolution
for theGlobal
PositioningSystemappliedto geodeticbaselinesup to
2000 km, J. Geophys.Res.,94, 10187-10203,1989.
Blewitt, G. M. B. Heflin, F. H. Webb, U. J. Lindqwister,R.
P. Malla, Global coordinateswith centimeteraccuracyin
the InternationalTerrestrialReferenceFrame usingGPS,
Geophys.Res.Letters,19, 853-856, 1992.
Burbach, G. V. N., C. Frohlich, W. D. Penningtonand T.
Matumoto, Seismicityand tectonicsof the subducted
Cocosplate,J. Geophys.
Res.,89, 7719-7735,1984.
DeMets, C., R. G. Gordon, D. F. Argus, S. Stein, Current
platemotions,Geophys.
J. Int., 101,425-478, 1990.
Dixon, T. H., M. Bursik, S. KornreichWolf, M. Heflin, F.
Webb, F. Farina, and S. Robaudo, Constraints on
deformationof the resurgentdome,LongValley Caldera,
California from spacegeodesy,Am. Geophys.Union
Geodynamics
Series,SpaceGeodesyand Geodynamics
(In Press) 1993.
Dziewonski, A.M.,
G. Ekstrom, and M.P.
Salganik,
Centroid momenttensorsolutionsfor April-June, 1991,
Phys.EarthPlanet.Int., 71, 6-14, 1992.
Freymueller,
J. T., J. N. KelloggandV. Vega,Platemotions
in the North Andes,J. Geophys.Res., (In Press)1993.
Goes, S. D. B., A. Velasco, S. Schwartz, and T. Lay, The
Folio 1-1865, 8-17, 1987.
Savage,J. C., Dislocationsin Seismology,F. R. N. Nabarro,
ed., 251-339, 1980.
Savage,J. C., A dislocation
modelof strainaccumulation
and
releaseat a subductionzone, J. Geophys.Res., 88, 49844996, 1983.
Silver, E. A., D. L. Reed, J. E. Tagudin, and D. J. Heil,
Implicationsof the NorthandSouthPanamaThrustBelts
for the originof the Panamaorocline,Tectonics,9, 261281, 1990.
Singh,S. H. and F. Mortera, Sourcetime functionsof large
Mexican subductionearthquakes,morphology of the
Benioff zone, age of the plate and their tectonic
implications,J. Geophys.Res.,96, 21487-21 502, 1991.
Shipley,T. H., J. W. Ladd,R. T. Buffler,andJ. S. Watkins,
Tectonicprocesses
alongtheMiddle Americatrenchinner
slope, in J. K. Leggett, ed., Trench-ForearcGeology,
Geol. Soc.LondonSpec.Pub. 10, 95-106, 1982.
Stein, S., J. Engeln, C. DeMets, R. Gordon,D. Woods, P.
Lungren,D. Argus,C. Stein,D. Weins,The Nazca-South
Americaconvergence
rate andthe recurrenceof the great
1960 Chilean earthquake,Geophys.Res.Lett., 13,, 713716, 1986.
Ward, S. N., A syntheticseismicitymodel for the Middle
America Trench, J. Geophys.Res., 96, 21433-21442,
Ap•ril,1991,Valle de la Estrella,CostaRica(Mw=7.7)
earthquakeand its tectonicimplications:a broadband
seismicstudy,J. Geophys.Res.,(In Press)1993.
Heflin, M. W. Bertiger,G. Blewitt, A. Freedman,K. Hurst,
S. Lichten, U. LindqwiSter,Y. Vigue, F. Webb, T.
Yunck,J. Zumberge,GlobalGeodesyusingGPSwithout
fiducialsites,Geophys.Res.Lett., 19, 131-134,1992.
1991.
TimothyH. Dixon, Rosenstiel
Schoolof Marineand
Atmospheric
Sciences,
Universityof Miami
Miami, FL 33149
Herring,
T. A., D. DongandR. W. King,Sub-milliarCsecond
determination
of pole positionusingGlobalPositioning
Systemdata,Geophys.
Res.Lett.,18, 1893-1896,1991.
(ReceivedMay 26, 1993'
Accepted:June14, 1993.)