GEOPHYSICAL
RESEARCH
LETTERS,VOL. 17, NO. 3, PAGES219-222, MARCH1990
PRECISEREGIONALBASELINEESTIMATIONUSINGA PRIORIORBITALINFORMATION
Ulf I. Lindqwister,
Stephen
M. Lichten,
andGeoffrey
Blewitt
JetPropulsion
Laboratory,
California
Institute
ofTechnology
Abstract.A solutionusingGPS measurements
acquired
duringthe CASA UNO campaignhas resultedin 3-4 mm
horizontaldaily baselinerepeatabilityand 13 mm vertical
repeatability
for a 729km baseline,
locatedin NorthAmerica.
Theagreement
with VLBI is at the level of 10-20 mm for all
components.
The resultswereobtainedwith the GIPSY orbit
determination and baseline estimation software and are based
onfive single-daydata arcsspanningthe 20, 21, 25, 26, and
27 of January, 1988. The estimation strategy included
resolvingthe carrierphaseintegerambiguities(biasfixing),
utilizingan optimalset of fixed referencestations(fiducials),
andconstraining
GPS orbit parametersby applyinga priori
information(derivedfrom initial multi-daytrajectoryfits to
broadcast
ephemerisdata).A multi-day(January20-27) GPS
orbit and baseline solution has yielded similar 2-4 mm
constrained
orbitsolutionsagreebetterthantheunconstrained
solutions
with themulti-dayresultsandalsoshowimproved
baselinerepeatabilities.
The constrained
approach
hasbeen
widelyexploredin studiesby Beutlerandcoworkers[Beufier
et al., 1987],whereorbitconstraints
arenecessary
in placeof
lackingor weak fiducialinformation.In contrast,our solutions
are basedon the fiducial referencenetwork (collocatedwith
VLBI) andtheconstrained
orbitssimplyrepresent
a smallbut
significant enhancementof our solution strategy. The
followingsectiondiscusses
the constrained
orbit approach,
bias fixing, and the choice of fiducial sites. The above
strategywith weakorbitconstraints
hasyieldedfew mm-level
horizontaland 10-20 mm vertical daily repeatabilitiesfor
baselines,
rangingfrom245 to 729 km in length.In addition,
10-20 mm agreementwith VLBI was also obtainedfor most
baseline components.The subsequentsection discusses
simultaneous
multi-daydataarc solutions,the mostprecise
GPS orbit strategy.The results are consistentwith the
produces
solutions
whichapproach
thepreciseorbitsobtained constrainedsingle-daysolutions.The concludingsection
with unconstrained multi-day arc solutions. This work
summarizes
ourresultsanddiscusses
utilizingnominalorbits
suggests
thatthe generalavailabilityof nominalephemerides for regionalGPSestimation
strategies.
basedon many daysof global trackingdata would be very
valuableandpracticalfor regionalgeodesy,allowingfor high
Techniques
andResultsfor Single-DayEstimation
precisionwith single-daynetwork solutionswhen combined
with a continental scale fiducial network.
The single-dayarc resultsdiscussed
below were basedon
horizontal
daily repeatabilities
for the samebaseline,consistent
withtheconstrained
single-dayarc solutions.The application
of weakconstraints
to the orbitalstatefor single-daydataarcs
data from six stations in North America: Hat Creek Radio
Introduction
Few millimeter geodeticbaselineprecisionhas a broad
rangeof scientificapplications,includingmonitoringstrain
ratesbetweenand within tectonicplates,studyingvolcanic
uplift,anddetectingco- andpost-seismic
netdisplacements
in
earthquake
faultzones.The CASA UNO experiment
offereda
uniqueopportunityto test estimationtechniquesaimedat
achieving
suchhigh-precision
results,dueto thelargenumber
of co-monitoringstationswith uniformreceiverand antenna
t)T}es.
A keyrequirement
for achievinghighbaselineprecision
forregionalnetworksis theaccuratedetermination
of theGPS
satelliteorbits. Currently, there are several successful
approaches
to preciseorbitdetermination
basedon thefiducial
approach,most notably: (1) multi-day arc simultaneous
adjustment
of all estimated
parameters
[DongandBock,1989;
LichtenandBertiger,1989],(2) multi-dayare adjustment
wheretrackingsitesarefirst usedto solvefor theGPSorbits,
followed
bya network
adjustment
of non-tracking
sites(i,e.,
thefixedorbitapproach)
[Lichten,
Bertiger,
andLindqwister,
1989],and(3) single-day
arcsimultaneous
adjustment
with
carderphaseambiguity
resolution
[Blewitt,!989;Dongand
Bock,1989].Moreover,
combinations
of thestrategies
canbe
used,
e.g.multi-daysimultaneous
solutions
withbiasfixing.
Thispaperinvestigates
thethirdstrategy
withtheaddition
of imposing
weaka prioriconstraints
on the GPSorbital
parameters.
The a priori information is obtainedfrom an
Observatory(CA), Fort Davis (TX), HaystackObservatory
(MA), OwensValley RadioObservatory
(CA), Mojave(:CA),
andMammothLakes(CA). Seethemapin thefu'starticleof
thisissue(andfigure 1) for geographical
locations.In orderto
establish the solutions in a well-defined reference frame and to
providegeometricala priori information,threeof the above
stationswere held fixed at their a priori VLBI locations,for
whichwe usedthe VLB! globalsolutionentitled:GLB223
(J.W. Ryan,C. Ma, andE. Himwich,GoddardSpaceFlight
CenterVLBI group,unpublished
results,1988). SeeDong
and Bock (1989), Lichten and Border(1987), and references
therein
forfurtherdiscussions
of thefiducialconcept.
Haystack
Hat
Creek
..•.....•-------•'-"•-••
•
ß• Mamm0•
....................
•,•o
•,;',•,,,,•
Mojave
Owens
Valley
Et. Davis
iterated,
3 weekfit to thebroadcast
ephemeris
data,andis
t,•mforeindependent
of the GPS solutionset.GPS orbits
fromweaklyconstrained
and unconstrained
solutions
are
Fig. 1. Two of •e •ee fiducial combina•onsexerted
Hat Creek-Ft. Davis-Haystack(solid lines), and Mojave-Ft.
Davis-Haystack(dott• SheS).The networkM• Hat Cr•k
compared
to multi-dayarcorbits.It is shownthattheweakly
pr•uc• 30-50%•t•r •ly b•elinem•a•h•fies.
(•right
O_ptim.•FiducialGeometry_
1990by the AmericanGeophysicalUnion.
'Paper number 90GL00414.
We havecompared
thedailyrepeatability
andthe a :•e•ent
•0094--8276/90/•-00414503
. 00
with VLB! for baselines estimated with different cmb•mations
219
220
Lindqwister
etal.:Precise
Baseline
Estimation
of fiducialsites.For example,utilizingHat Creek-Ft.Davis-
resolution
technique
isbased
onconstraining
thecarrier
phase
Haystack
asfiducials
ratherthanMojave-Ft.
Davis-Haystackobservables
utilizing
theprecise
Pcode
pseudorange
data
type
yields---30-50%improvement
in dailyrepeatabilities
forthe [Blewitt,1989].Theorbitalparameters
arecorrelated
withthe
Mojave-Hat Creek (729 kin) baseline.The two fiducial carrierphaseambiguities,
henceresolvingthe ambiguities
combinations
areillustratedin figure1. The formerfiducial produces
moreaccurate
GPSorbitestimates.
Biasfixingin
networkalsoproduces
similarimprovements
in agreement effect convertsthe carrier phaseobservableinto an ultrawithVLBI for thesamebaseline.
Afterstudying
sixbaselines precisepseudorangedata type. Differencedcarrier biases
in California,rangingfrom 71-729 km and usingthree betweenall theCaliforniastationswereresolvedfor thefive
combinations
of fixed sites(interchanging
Hat Creekwith single-day
arcs.ForsixCaliforniabaselines,
rangingfrom71MojaveorOwensValley),wefoundthatthecombination
with 729 km, thedailyrepeatabilityandthe agreement
withVLBI
Hat Creek was superiorin all cases.As a rule the most improved.bya factorof •-2 in the eastandby 10-20%in the
favorablefiducialgeometry
is obtained
whenthemobilesites northand verticalbaselinecomponents
relativeto solutions
arein proximityto thereference
sitesandwhenthespatial withunresolved
carrierbiases.
In generalthelongerbaselines
extent of the fiducial network is maximized.
showedgreaterimprovementsincethey are moresensitive
to
DuringthethreeweekCASAUNO campaign,
Hat Creek GPS orbit errors.
was only availableduring20-22 and 25-27 of January.In
addition,only sporadicsamplesof datawereobtainedfrom The Constrained
Orbi....•
Approach
Mojaveon January22. In orderto obtainhigh-precision
solutionsa well balanceddatasetwith an optimalfiducial
in orderto achievepreciseorbitestimationGIPSY utilizes
geometrywas chosen,i.e., data from five single-dayarcs
(January20, 21, and 25-27) were usedwith Hat Creek,Ft.
Davis, and Haystackas fiducial locations.It is somewhat
forcemodelingwhichincludesthe ROCK IV solarradiation
pressure
model,a gravityfield modelto degreeandorder12in
sphericalharmonics,
andpointmasses
for the sun,moon,and
difficultto drawdefiniteconclusions
froma statistical
sample planets (See Lichten and Border, 1987, and references
providedby onlyfive stationdays.However,we findthatby
correlatingthe baselinerepeatability
resultswith a studyof
orbitaccuracies,
thecasefor usingweaka prioriconstraints
becomes
moreconvincing.
therein).
Theorbitintegration
isperformed
in thequasi-inertial
J2000reference
frame.Orbitestimation
proceeds
in twosteps,
wherein thefirststepa nominal
satellite
trajectory
isintegrated
based
onaniterated
least-squares
fit to a timeseries
of apriod
valuesof the satelliteposition,velocity, and solarradiation
pressure
parameters.
Broadcast
ephemeris
datasampled
over3
•TheStandard
Estimation
Strategy
weekswasusedin this study,but in the pastwe haveals0
usedothermulti-dayfitsto globaltrackingdata,forexample,
fromTI-4100receivers,
withapproximately
1.5meterpseudo- fromtheNSWC preciseephemeris
service[Swift,1985].The
rangeand1.0centimeter
carrierphasemeasurement
accuracy. secondstepis to adjustepochstatevaluesof satelliteorbital
Thedataerrorsareattributed
to multipath
andsystem
noise. parameters
usingGPS carrierphaseandpseudorange
data
The GPS measurementsfrom all sites above were obtained
The TI-4100
receivers track at most four satellites
types in a factorized Kalman filter. The GIPSY softwareis
simultaneously.
Theparameter
estimation
strategies
applied flexiblein choiceof arclength,useof a prioricovariances,
and
here used30 seconddata pointscompressed
to 6 minute stochastic
estirnation
of forceandnon-force
parameters.
normalpointsacquiredat the threefiducialsandat the three
mobilestations(OwensValley,Mojave,andMammoth).On
sitepressure
measurements
wereusedfor dry tropospheric
delaycalibration.The noisierwet tropospheric
delaywas
estimated
stochastically
in a factorized
Ka!manfilterusinga
randomwalk model,appliedhereto modelthetimevariations
in the watervaporinduceddelay.The randomwalkmodelis
characterized
by the rate of changeof the processnoise
covariance over time and was chosen to be 0.02 cm S-1/2
correspondingto about 6 cm delay variation in 24 hours
[Lichtenand Bertiger,1989]. The receiverandGPS satellite
clockswereestimated
aswhiteprocess
noise,i.e. uncorrelated
between
measurement
epochs.
Theremaining
parameters
were
estimatedas constantparametersfor the single-dayarc
solutions.
Thea prioriparameter
sigmas
arelistedin Table1.
In orderto improvethe single-dayGPS orbit solutions,
carderphaseintegerbiaseswereresolved.The ambiguity
Table !. Constraints Used For Parameter Estimation
Parameter
A PrioriSigmas
Constrained
Satelliteposition
Satellitevelocity
Unconstrained
40 m
2x10-3m/s
20 km
20 m/s
25%
100%
25%
100%
Solar radiation
pressure
coefficients
X, Z
Y
Unconstrained
Parameters:Nonfiducialstations,whitenoise
clocks,zenithwettroposphere,
andcarderphasebiases.
Data Weights:
carrierphase:1.0cm,pseudorange:
!.5 m,
The multi-dayarc solutions(discussed
furtherin thenext
section), have resulted in meter-level GPS orbit accuracies
[LichtenandBertiger,1989].We havefoundthatournominfl
GPSorbitsbasedon 3 weekintegrated
fits to the broadcast
ephemeris,
whencompared
withthemulti-dayarcsolutions,
provedto be accurate
to ~ 20 meterswith accuracy
of about
! 0-7 cm/s2for nominalsolarradiationpressure
coefficients.
The broadcastephemerisis independentof the GPS
measurements
we usedin parameterestimation.In taking
advantage
of theinformation
in thenominalsatellite
trajectory,
we usedfairly conservative
a priori sigmas(Table 1)
constrain
single-day
orbitandbaseline
estimation.
Theapriori
sigmasof the satellitepositionswerechosento be 40 m in
eachcomponent
andthesigmasof the y-biassolarradiation
pressure
parmeterswere2x10-7crn/s2(corresponding
tothe
100%valueusedin thetable)[LichtenandBorder,1987].
The effectsof usinga prioriconstrained
GPSorbitsversus
unconstrained
orbitsfor single-day
arcsareillustrated
in fi•
2. The mappedestimates
from constrained
andunconstrmned
solutionshavebeendifferencedfrom the corresponding
estimates
obtained
in a multi-day
arecomputation,
whichhere
servesas the referenceset of orbits.The figure displays
mapped
differences
for thesingle-day
arc 12-hrdatainterv•
plusan additional12-hrprediction
intervalwhennodatawere
taken.The resultsshownin figure2 for GPS 8 were similar
for all sevensatellites:
thesolutionwitha prioriconstraints
shows
muchbetteragreement
withthemulti-dayarcreference
orbit.Notethedramatic
improvement
overtheorbitpredic
interval (12-24 hrs) for the constrainedorbit case,whichis
indicativeof improvedorbitaccuracy.
For down-track,
GPS8 formalerrorsrangedfrom0.5-2 m, 2.5-5m, and
m for the multi-dayarc,constrained
single-day
arc,and
unconstrained
single-day
arc,respectively.
Thelowernumber
in eachcaserepresents
thermsformalerroroverthe 12-hrda.ta
intervalandthehighernumberrepresents
theformalerrorov•
Lindqwister
et al.: PreciseBaselineEstimation
22 t
16-
........
,*....... UNCONSTRAINED
12
......................
RMS = 1 MM
........
RMS: 4 MM
ORBIT
m---CONSTRAINED
O'
.......................
RMS = 2 MM
........
RMS = 3 MM
•
, ......
•
4'
•
o
o
,
,
,,
,,,, • ..............
,,,,'-'
•
.. .... ,
.•.. ,
. .. ,
. ,., .... . .., .....
OWENS VALLEY-MOJAVE
MOJAVE-HATCREEK
4
-!0
20
22
24
(JA UA '88)
Fig. 3. Millimeter-level daily repeatabilitiesin horizontal
baselinecomponents
with GPS for theMojave-HatCreek(729
km) and the Owens Valley-Mojave (245 km) baselines.
Resultsare plottedabouttheir meanvaluesand are basedon
single-dayconstrained
orbit solutions.
o
-8:
......
0
5
10
15
TIME PAST 2•JAN-1989 5:•
20
25
(hrs)
Rg. 2. GPS8 orbit•fference •tween thereference"•"
orbit detemined from a 1-week arc solution, and the
c•s•n•
•d uncons•ned s•g!e-day• orbitsdetemned
•om a 12-hrpassonJ•u• 25 only.•e we•y cons•ned
resultsa•ee much better wi• the referencemulti-day •c
estimates,
•
in •e 12-• data inte•al (0-12 •)
• pr•ction intent (12-24 hrs).
and •e 12-
the12-hrprediction
interval(12-hrinterpolated
intervalfor the
multi<layareorbitcase).Comparison
of thesenumbers
to the
down-track
valuesplottedatthetopof figure2 showsthatthe
The goalwasto exhibittheprecision
possible
withGPSunder
optimalconditions.Shownin figure3 are horizontaldaily
repeatabilities
for theOwensValley-Mojave(245 km) andfor
theMojave-HatCreek(729 km) baselines.
The horizontalrms
scatterfor the OwensValley-Mojavebaselineis 1-2 mm, and
for Mojave-HatCreekit is 3-4 min. The daily repeatability
in
the verticalcomponents
are 17 mm and 13 mm for the two
baselines. Additional
baselines studied included Owens
Valley-HatCreek(484 km), Hat Creek-Mammoth
(416 km),
andMojave-Mammoth
(313kin).Whenorbitsareconstrained
all baselines
producehorizontalrepeatability
of 2-6 mm and
verticalrepeatability
of 10-15mm.With orbitconsents much
smallerthan 40 m, systematicbiasesappearedin baseline
fc•rmalerrors mirror the actual trends in the dam.
in termsof arc lengthand commonview period,GPS 8
wasoneof the well-trackedsatellitesin this single-dayarc,
MOJAVE-HAT CREEK 729 KM
DAILY REPEATABILITY
whichincluded
dataonlyfromNorthAmericantrackingsites.
However,
theformaleiTors
for several
satellites
withpoor
visibilitystaylarge(above10 m) over mostof this24-hr
periodwhen the single-dayorbits are unconstrained.
Effectively,
only 3-4 satellites
play a significant
role in
baseline
parameter
estimation
for theunconstrained
case.By
weaklyconstraining
theorbitparameters,
however,
datafrom
all satellites,
including
the lessvisibleones,are used.The
covariances
stayfairlylow (severalmetersor better)for all the
satellites
overmuchof the 24-hr periodwith constrained
orbits.
Theunconstrained
solutions
degrade
towards
theends
of'thesatellite
passes
(asdotheformalerrors),hencetheuse
ofconstraints
effectively
increases
thelength
of theobserved
a.m.Thekeyto thedramaticimprovement
whenapplyinga
(HORIZONTAL)
EAST
NORTH
LENGTH
of Mojave-Hat
Creek(729km)rmsdaily
priori
constraints
is thattheorbitalcomponents
fordifferent Fig.4. Comparison
satellites
arehighlycorrelated
andwithoutsuchconstraints baseline repeatability for unconstrained(solid bars),
constrained
single-day(hatchedbars),and multi-day(open
would
bedominated
bylargedown-track
uncertainties.
yield baseline
In summary,
theestimation
strategy
usedfor single-day bars) solutions.The two latter approaches
of 2-4 mm(about3-5 partsin 109),il!ustmting
the
solutions
incl•aded:
fixingcarrierphase
biases,
selecting
a precision
of weaklyconstra'ming
thea prioriorbitcora.dance
is[rang
fiducia!
geometry,
constraining
aprioriGPSorbits,
and advantages
estimation
runswitha limitedregionalnetwod,
t.
utilizing
adatasetevenly
distributed
intimeforallsixstations.for short-arc
222
Lindqwister
et al.: PreciseBaselineEstimation
repeatabilities,
butlargera priorisigmas(50-80m), produced hundred
km in length,usingeitherconstrained
single-day
onlyslightlydegraded
resultsasusing40 m constraints.
solutions
ormulti-day
arccomputations,
despite
thehandica•
of a limited constellationof GPS satellitesandreceiverswith
The single-day
unconstrained
solutions
yieldrmserrorsof
3 mmin thehorizontal
and19mmin theverticalcomponents onlyfour satellitechannels.For the multi-dayarcstheorbital
for theOwensValley-Mojavebaseline
and4-7 mmhorizontal stateis betterdetermineddue to therepeatedrevolutionsof the
The useof a priori informationto initiallyconstrain
and22 mm verticalrepeatabilities
for theMojave-HatCreek satellites.
baseline.In generalthe daily repeatabilities
improveon the orbit covariancesconsistentlyimproves solutionswith
averageby a factor of 1.5, 1.3, and 1.2 for the east, north,
weak observability,suchas thosefrom single-dayarcs.The
andverticalcomponents
for the five baselineslistedabove, latterstrategyis valid providedthe appliedconstraints
havenot
when a priori orbital informationis used in parameter been determined from the same data set used for estimation.
estimation.
The assessment
of GPSaccuracy
is difficultdueto
Althoughthe nominalorbit usedfor this studywasbased
potential ground tie errors between GPS monumentsand
onbroadcasi
ephemeris
data,theconclusions
should
applym
VLBI antenna intersectionof axis. In addition, there is no
nominalorbitsproducedby reductionof severaldaysof GP$
VLBI solution for the site at Mammoth. For the constrained
datatakenfrom globaltrackingsites,e.g. Swift, 1985.Thisis
solutionthe agreementwith VLBI remainedbetween10-20
especiallypertinent,sincethe broadcastephemeriswill be
mm in all components
for the OwensValley-Mojaveandthe significantlydegradedby "selectiveavailability"in thenear
Mojave-Hat Creekbaselines.The OwensValley-Hat Creek future.Consequently,
a practicalscenariofor futureregional
baseline
agreedto5-10mmin theeastandverticalcomponents GPSanalysismayinvolvethetimelydissemination
of precise
anddifferedby 36 mmin thenorthcomponent.
Onaverage
the ephemeristablesby a dedicatedanalysisgroup,whichcould
agreementwith VLBI improvedrelativeto the unconstrained be usedin a similarfashionasthe broadcastephemerisdatafor
solutionby a factor of 1.7, !.1, and 1.7 for the east, north,
producinga nominal orbit. Subsequently,the orbit, with a
andverticalcomponents
for the threebaselines.
priori constraints,would be adjustedon a day by day basis
usingdatafrom a user'sregionalnetworkplusa tlu'eestation,
continentalscalefiduciaI network which encompasses
the
ResultsFrom Multi-dayArc Solutions
region.The considerablesavingsin computationaltime and
disk spacemakesthis option attractivefor userswho do not
The multi-day arc strategyfollowed the one applied
wish to be burdenedwith the taskof reducingmanydaysof
successfully
to otherexperimentsasdescribedin Lichtenand
datafrom a worldwidesetof GPS trackingstations.
Bertiger (1989). All the single-dayarc data and the sarne
fiducialswereusedfor themulti-dayarcsplusdatafromfour
Acknowledgments.The work describedin this letterwas
sites in South America and two sites in Europe. In the
carried out by the Jet Propulsion Laboratory, California
factorizedKalmanfilter stochastic
zenithtropospheric
delays Institute of Technology, under contract with the National
and stochastic GPS solar radiation coefficients were estimated.
Aeronautics
andSpaceAdministration.
The multi-day arc solutionswere obtainedwith no a priori
constraintsappliedto the orbital parameters.Formal orbit
errorswere at themeter-levelor betterfor all components
for
References
the sevensatellitestrackedover muchof the week-longdata
arc. The estimatedMojave-Hat Creek baselineagreedwith
VLBi to 10 mm or better and showedrms daily scatterof
for theglobal
severalmillimeters in the horizontaland baselinelength Blewitt,G., Carderphaseambiguityresolution
positioning
systemappliedto geodeticbaselines
upto2•
components(see figure 4). The agreementwith VLBI is
km, J. GeophysicalRes., 94, 10187-10203,1989.
consistent with the sub meter formal errors in the GPS orbits.
Beutler, G., !. Bauersima,W. Gunner, M. Rothacher,andT.
The multi-day arc orbit solution served as a reference for
Schi!dnecht,
Evaluationof the !984 Alaskapositioning
comparisonwith varioussingle-dayarc solutionstrategies.
campaignwith the BeameseGPS software,J.G•ophysical
In figure 4, daily repeatabilityfor the Mojave-Hat Creek
Res., 92, 1295-1303, 1987.
baselineis shownfor the multi-day arc orbit caseandfor the
Dong,D., andY. Bock,GPS networkanalysiswith phase
single-day,constrainedand unconstrained
orbit cases.With
ambiguityresolutionappliedto crustaldeformation
stu•es
thetwo formerstrategies,
the horizontalandlengthprecisionis
in California,J. GeophysicalRes.,94, 3949-3966, 1989.
3-5 partsin 109(2-4mm)andfor theunconstrained
casethe Lichten,S.M., andW.!. Bertiger,Demonstration
of sub-meter
precision
is 6-9 partsin 109for thehorizontal
components
and
4 partsin ! 09forthelengthcomponent.
Formostbaselines
in
thisstudythe constrained
single-dayarc solutionsperformed
nearly as well as the multi-day arc approach,even though
Europeanand SouthAmericansiteswere not included.In the
vertical,both approaches
had daily repeatabilityof .• 13 mm
for Mojave-Hat Creek.Sincethe samefiducialswereused,all
three approachesproducedessentiallysimilar GPS-VLBI
agreement(10-20 ram). For the considerablyshorter245 km
Owens Valley-Mojave baseline, orbit error plays a less
importantrole in thebaselineerrorbudget,and thehorizontal
single-dayarc repeatabilities
were 1-2 mm or better(as shown
in figure 3), with no additionalimprovementwhenmulti-day
arcorbitswere used. Clearly, the single-dayconstrained
orbit
strategyfor the OwensValley-Mojavebaselinereducesorbit
GPS orbit determination
and 1.5 partsin 108threedimensional
baselineaccuracy,
..Bull.Geod.,63,167, 1989,
Lichten,S.M., W.I. Bertiger,andU.J. Lindqwister,
The
effectof fiducialnetworkstrategy
on high-accuracy
GPS
orbit and baselinedetermination,Proc. 5th Intl. Geode•
Symposium
onSatellitePositioning,
_I,pp.516-525,1989.
Lichten,S.M., andJ.S.Border,Strategies
for high-precision
global positioning system orbit determination, [
Geophysical.Res., 92, 12751-12762, 1987.
Swift,E., NSWC'sGPSorbit/clockdetermination
system,
Proc.FirstInt. Symp.on PrecisePositioning
withGPS,
ed. Goad,Vol. I, p. 51-62, Rockville,MD, May 1985.
errors to a level which is lower than errors from other sources,
U. Lindqwister,S. Lichten,andG. B!ewitt,MS: 238-624,
JetPropulsion
Laboratory,CaliforniaInstituteof Technology,
suchasmultipath,troposphere
mismodeling,
anddatanoise.
Pasadena,CA 91109.
Discussion and Conclusions
Our resultsindicatethatfew millimeter-levelprecisionmay
be achievablein horizontalcomponents
of baselines
up to 700
(ReceivedNovember8, 1989;
revisedJanuary30, 1990;
acceptedFebruary1, 1990)