Abstract Within the regional EUREF Permanent
Network (EPN) all positioning is purely based on GPS.
This paper investigates, using the Bernese GNSS
analysis software, the influence of adding GLONASS
observations to the EPN processing using fixed orbits
from the International GNSS Service (IGS) as well as
from the CODE analysis centre. The GPS-only coor-
dinates and GPS + GLONASS coordinates will be
compared and the change in their repeatabilities will
be investigated. The influence of the used orbits will
also be outlined. The results show that a combined
GPS + GLONASS data analysis can be set up without
major efforts and that it will not degrade the positions
obtained within the EPN.
Keywords Glonass Æ EPN Æ Positioning
Introduction
The EUREF Permanent Network (EPN) consists of
190 permanent GPS stations from which 29 are also
tracking GLONASS satellites. The primary purpose of
the EPN is to maintain the European Terrestrial Ref-
erence System (ETRS89) and EUREF does this by
making available the tracking data of the EPN stations
and by generating weekly coordinate estimates for all
of them (Adam et al. 2002; Bruyninx 2004). Up to now,
all coordinate estimates have been based on only GPS
data and no GLONASS data is used. However with
• the growing number of commercially available
GPS + GLONASS receivers;
• the recent revitalization of GLONASS (with a
constellation of 18 satellites expected in 2007) and
• the availability of short latency combined IGS or-
bits for GLONASS and consistent GPS + GLON-
ASS CODE orbits
it has become worthwhile to assess the advantages
and disadvantages of adding GLONASS data to the
routine data analysis of the EPN network. For this
investigation, two networks of GNSS stations belong-
ing to the EPN have been analyzed: a first network
consisting of only GPS/GLONASS stations and a sec-
ond mixed network of GPS and GPS/GLONASS sta-
tions. For both networks, the station coordinates have
been estimated using GPS only observations and
GPS + GLONASS observations.
Presently, with a constellation of 29 active GPS
satellites and 13 GLONASS satellites, no significant
improvement can be expected from adding GLONASS
observations to GPS: the number of observations is
increased by a factor of 1.4, resulting in an expected
reduction of the formal errors byffiffiffiffiffiffiffiffiffi
1:45p
¼ 1:2: This
assumes that the number of parameters is not changed
by the introduction of the GLONASS data. As we will
see later, this is not the case.
In addition, the exercise to process observations
from a mixed GPS and GPS/GLONASS network in
which one of the two GNSS has an incomplete con-
stellation is a worthwhile exercise when preparing for
the future GALILEO system.
C. Bruyninx (&)Royal Observatory of Belgium, Av. Circulaire 3,1180 Brussels, Belgiume-mail: [email protected]
GPS Solut (2007) 11:97–106
DOI 10.1007/s10291-006-0041-9
123
ORIGINAL ARTICLE
Comparing GPS-only with GPS + GLONASS positioningin a regional permanent GNSS network
Carine Bruyninx
Received: 23 June 2006 / Accepted: 22 August 2006 / Published online: 19 September 2006� Springer-Verlag 2006
Set-up of the data processing
Networks analyzed
Two networks have been analyzed in order to assess
the influence of adding GLONASS data to the GPS-
only analysis. The first network consists of stations all
equipped with combined GPS + GLONASS equip-
ment. It contains the 25 GPS/GLONASS stations in-
cluded in the EPN on January 2006. As can be seen in
Fig. 1 (left side) most of the GPS/GLONASS stations
are located in central and northern Europe. The GPS/
GLONASS equipment used is given in Table 1.
The second network corresponds to the network
who’s GPS data are routinely analyzed by the Royal
Observatory of Belgium (ROB) and used to contribute
to the weekly coordinate solution of the EPN. The
network is a mixture of stations, mostly located in and
around Belgium, equipped with GPS-only and
GPS + GLONASS receivers; it is shown on the right
side of Fig. 1. We are using this network in order to
investigate whether the GPS-only ROB solution could
be replaced by a GPS + GLONASS solution without
altering the station coordinate time series and the
ROB contribution to the combined EPN solution.
Data processing
When processing a regional network, such as the two
networks considered, precise a priori orbit information
in needed. The IGS makes separately combined orbits
available for GPS and GLONASS. The GLONASS
orbits are based on the computations from four anal-
ysis centres: BKG, CODE, ESA/ESOC, and IAC
(Weber and Fragner 2000; Gendt 2006). As we can see
in Fig. 2, recently the latency of the combined IGS
GLONASS orbits has been significantly reduced. Since
the analysis centres BKG and ESA use the final IGS
GPS orbits as input for the computation for their
GLONASS orbits, the combined IGS GLONASS or-
bits are mostly available with a delay of a few days with
respect to the final IGS GPS orbits. Following the
Fig. 1 Left GPS/GLONASSstations belonging to the EPNon 1 January 2006; rightmixed GPS and GPS/GLONASS network (GPS/GLONASS stations areindicated with station namelabel)
Table 1 GPS/GLONASS equipment used within the EPN(January 2006)
Station Receiver Antenna + radome
BISK ASHTECH Z18 ASH701946.2 NONEBOGI JPS E_GGD ASH701945C_M SNOWCAGZ JPS E_GGD JPSREGANT_DD_E NONECOMO TPS E_GGD TPSCR3_GGD CONEGOPE ASHTECH Z18 ASH701946.3 SNOWHELG JPS LEGACY ASH700936D_M SNOWHERT ASHTECH Z18 ASH701946.2 NONEHOE2 JPS LEGACY TPSCR3_GGD CONEJOZ2 ASHTECH Z18 ASH701941.B SNOWKIR0 JPS E_GGD AOAD/M_T OSODMAR6 JPS LEGACY AOAD/M_T OSODMARJ ASHTECH Z18 ASH701946.2 NONEONSA JPS E_GGD AOAD/M_B OSODPOUS TPS GB-1000 TPSCR3_GGD CONESASS JPS LEGACY TPSCR3_GGD CONESKE0 JPS E_GGD AOAD/M_T OSODSNEC ASHTECH Z18 ASH701946.2 NONESOFI TPS E_GGD AOAD/M_T NONESPT0 JPS LEGACY AOAD/M_T OSODVACO ASHTECH Z18 ASH701946.2 NONEVENE ASHTECH Z18 ASH701941.B NONEVILO JPS LEGACY AOAD/M_T OSODVISO JPS E_GGD AOAD/M_T OSODWARN JPS LEGACY TPSCR3_GGD CONEWTZR TPS E_GGD AOAD/M_T NONE
98 GPS Solut (2007) 11:97–106
123
Products web-page at the IGS Central Bureau, the
accuracy of the final IGS orbits is less than 5 cm for
GPS and less than 15 cm for GLONASS.
The CODE analysis centre is presently the only
analysis centre generating fully consistent GPS
+ GLONASS orbits, meaning that both orbits result
from the same simultaneous processing of GPS and
GLONASS observations. Figure 3 shows that the
GLONASS orbits from CODE agree with the IGS
orbits at the 5 cm level. Following work from Urschl,
based on an SLR validation, the CODE orbits have an
accuracy of 2.5 cm for GPS and 5 cm for GLONASS
(Urschl, in press). The IGS and CODE orbits are ex-
pressed in the IGb00 reference frame, which is the IGS
realization of the ITRF2000.
Neither IGS nor CODE make available GLONASS
satellite clock information. As we will show later, this
will cause no problem for the data analysis.
The GPS and GPS + GLONASS data analysis has
been done using the Bernese 5.0 software (Beutler et al.
2006) following the scheme displayed in Fig. 4. In a first
step, the ionospheric free P3 GPS code is analyzed to
compute the receiver clocks. This is done using only the
GPS data as no GLONASS satellite clocks are avail-
able. Then, baselines are formed between stations with
a maximum number of common dual frequency
GPS + GLONASS carrier phase observations. In the
first test only GPS observations are considered, later
both GPS and GLONASS observations are used. In
both cases, identically the same baselines are formed.
After the creation of the independent baselines, phase
double differences are created in order to perform
Fig. 2 Latency of precise IGS orbits and CODE orbits
Fig. 3 Agreement of CODEorbits with IGS GLONASSorbits (source http://www.gfz-potsdam.de). The SLR orbitsfrom MCC (Mission ControlCentre Moscow) do notcontribute to the IGScombined GLONASS orbits,but it is only shown ascomparison
Fig. 4 GNSS data processing
GPS Solut (2007) 11:97–106 99
123
baseline-by-baseline GPS ambiguity resolution using
the QIF (Quasi Ionospheric Free) method (Beutler
et al. 2006) supported by the CODE ionospheric model.
When processing GPS and GLONASS observations,
the double difference GLONASS ambiguities are esti-
mated but no attempt is made to fix them to their
integer value. Since double differences are used, there
is no need for a priori information on the satellite
clocks, which are eliminated by the double differencing.
After fixing the GPS ambiguities to their integer values,
the ionospheric free L3 double differences are formed
and the station coordinates are estimated together with
the troposphere. The troposphere is modelled as piece-
wise linear functions using the dry-Niell a priori
mapping function, and estimating each hour the tro-
posphere using the wet-Niell mapping function.
The GPS-only analysis is preformed using IGS final
orbits/clocks while the GPS + GLONASS analysis is
done using on one hand the final IGS GPS orbits/
clocks and GLONASS orbits (which we merged at the
SP3 level) and on the other hand the fully consistent
GNSS orbits from CODE (Ineichen et al. 2001).
For the first network, the data were analysed from 1
October 2005 to 28 February 2006. The second net-
work was analysed form 5 February until 1 April 2006.
For each day a set of coordinates was determined. The
datum of the coordinates was defined by applying three
translation conditions (minimal constraint) to the EPN
realization of the ITRF2000.
Results
GPS/GLONASS network
When processing the network of 25 GPS/GLONASS
stations, the additional GLONASS satellites increase
the number of observations by 47%. The associated
maximal reduction of the formal errors has a factor of
1.2. However, in our case, the introduction of the
GLONASS data also increases the number of param-
eters to be estimated considerably (by 47%). These
additional parameters are the GLONASS ambiguities.
Consequently no significant improvement in terms of
formal errors can be expected from adding GLONASS
data to GPS.
During the period that we analysed, several of the
GPS/GLONASS stations provided data of degraded
quality. The most striking example is the station SNEC
(Snezka, Czech Republic) whose coordinates wan-
dered away (see Fig. 5), especially in the height com-
ponent, because of a receiver malfunctioning. The
SNEC data have therefore been eliminated starting
from GPS week 1349 at the first symptoms of the
receiver error.
Starting 1 January 2006 the data from the ASH-
TECH Z-18 receivers at JOZ2 (Jozefoslaw, Poland)
and GOPE (Ondrejov, Czech Republic) became
unusable. After the midnight epoch these two receivers
started tracking all GLONASS satellites with a 1-s
delay causing the pseudoranges to be increased by
300,000 km. As a leap second was introduced at this
date, a link to this event was suspected. The other
ASHTECH Z-18 receivers in the EPN behaved nor-
mally. The problem at JOZ2 and GOPE was narrowed
down to the TEQC (GNSS Translating, Editing, and
Quality Checking) software (Estey and Meertens 1999)
used to convert the native data to the RINEX format.
The problem was solved by updating TEQC to its
latest version from 15 December 2005.
The data from the station SOFI (Sofia, Bulgaria)
had to be discarded from the processing because of a
lack of reliable data caused by a malfunction of the
station PC.
As can be seen in Fig. 6, in addition to the problems
mentioned above, the three Italian GPS/GLONASS
stations (CAGZ, COMO and VENE) are missing in
almost 20% of the final solutions. The data of these
stations are regularly missing at all the Data Centers
(without correlation between the missing days from the
Fig. 5 Estimated coordinatesfor the station SNEC
100 GPS Solut (2007) 11:97–106
123
different stations). Note that in Fig. 6 as well as the
following graphs, the stations are ordered according to
increasing latitude.
Following the scheme displayed in Fig. 4, daily
coordinates have been estimated for the remaining
stations in the network: first using only GPS data (and
final IGS orbits/clocks), and secondly using GPS as
well as GLONASS data. The last processing was done
once using IGS orbits and once using CODE orbits.
Figure 7 shows no significant differences in the
repeatabilities of the station coordinates depending on
the observations and orbits used. An exception is the
station SNEC with a significant degradation of the Up-
RMS caused by the introduction of GLONASS data.
The inspection of the coordinate time series of
SNEC (Fig. 8) shows that the degradation of the RMS
is the caused by a few outliers in the GPS + GLON-
ASS solution of GPS week 1345.
We have also drawn the coordinates time series of the
nearby station POUS, which, as can be seen in Fig. 9,
shows an offset in its height-component when the GPS-
only results are compared to the GPS + GLONASS
estimates. However, in these coordinate time series, no
special events are noted.
Figure 9 shows that adding GLONASS data to the
GPS-only analysis changes the coordinates up to
2.5 mm in the horizontal components. However, as can
be seen from Fig. 10, these differences are mainly due
to differences in the reference frame. After a Helmert
transformation, the horizontal differences are below
1.5 mm, with a general RMS of 0.4 mm. In the up-
component, the coordinate differences between GPS
and GPS + GLONASS are mostly below 4 mm, but
reach for one station (POUS) up to 6 mm.
The 3D RMS of the coordinate differences is
1.8 mm, which is reduced to 1.4 mm by the Helmert
transformation. In all cases, the GPS + GLONASS-
based coordinates obtained using IGS or CODE orbits,
agree at the 1-mm level.
The origin of the difference between the GPS-only
and GPS + GLONASS estimates for the up-compo-
nents of POUS (6 mm) is unclear.
As a by-product of our analysis tropospheric Zenith
Total Delays (ZTD) are estimated each hour. As can
be seen in Fig. 11, GPS + GLONASS underestimates,
for all stations except POUS (!), the ZTDs compared
to GPS only. It is clear that the station POUS is
showing an atypical response to the introduction of
Fig. 6 Percentage ofobservation days included foreach station in the dataprocessing
0
1
2
3
]m
m[
RMS East
0
1
2
3
0
1
2
3
0
1
2
3
]m
m[
RMS North
GPS onlyGPS+GLONASS using IGS orbitsGPS+GLONASS using CODE orbits
ZGA
C
EN
EV
OM
OC
OCA
V
RZTW
EPO
G
SUOP
KSIB
JR
AM
CEN
S
TR
EH
2Z
OJ
IGO
B
NRAW
GLE
H
SS
AS
2EOH
AS
NO
0SI
V
0T
PS
6RAM
0LI
V
0EKS
0RI
K
egarevA
01234567
]m
m[
RMS Up
01234567
Fig. 7 Coordinaterepeatabilities obtained usingGPS-only, GPS + GLONASSwith IGS orbits,GPS + GLONASS withCODE orbits
GPS Solut (2007) 11:97–106 101
123
GLONASS. Figures 12 and 13 show examples of the
ZTD behavior for the stations VACO and JOZ2.
Results for mixed network
In the network of mixed GPS and GPS/GLONASS
receivers, the introduction of the GLONASS data
increases the amount of used observations by 14%. A
similar increase is also noted in the number of esti-
mated parameters. As we can see in Fig. 14, as ex-
pected, the repeatabilities of the estimated coordinates
are independent of the introduction of the GLONASS
data (GPS/GLONASS stations are: HELG, HERT,
HOE2, KARL, ONSA, SPT0, WARN, WTZR).
In addition, Fig. 15 shows that no significant changes
in the coordinates can be seen. We can therefore con-
clude that, for this specific network, GLONASS data
can be introduced in the data analysis without any
problems. However, to avoid influencing the site
velocities, the introduction of GLONASS should be
done simultaneously with the introduction of the abso-
lute phase centre variations and the switch to ITRF2005.
Conclusion
The goal of this study was to investigate the advanta-
ges/disadvantages of analyzing combined GPS/
Fig. 8 Coordinate time seriesfor SNEC (top) and POUS(bottom) obtained using aGPS-only and aGPS + GLONASS analysisusing IGS orbits
102 GPS Solut (2007) 11:97–106
123
+
-3-2-10123
]m
m[
East
-3-2-10123
-3-2-10123
-3-2-10123
]m
m[
North
GPS versus GPS+GLONASS using IGS orbitsGPS versus GPS+GLONASS using CODE orbitsGPS+GLONASS using IGS or CODE orbits
ZGAC
EN
EV
OMOC
OCAV
RZTW
EPOG
SU
OP
KSIB
JR
AM
CENS
TR
EH
2ZOJ
IGOB
NRA
W
GLEH
SS
AS
2EO
H
AS
NO
0SI
V
0T
PS
6RA
M
0LIV
0E
KS
0RIK
-6-4-202468
]m
m[
Up
-6-4-202468
Fig. 9 Coordinate differencesbetween GPS-onlycoordinates and coordinatesobtained usingGPS + GLONASS data,computed respectively withIGS orbits and with CODEorbits
-3-2-10123
-3-2-10123
]m
m[
East
-3-2-10123
-3-2-10123
]m
m[
North
GPS versus GPS+GLONASS using IGS orbitsGPS versus GPS+GLONASS using CODE orbitsGPS+GLONASS using IGS or CODE orbits
ZGAC
ENEV
OMOC
OCAV
RZTW
EPOG
SUOP
KSIB
JRAM
CENS
TREH
2ZOJ
IGOB
NRAW
GLEH
SSAS
2EOH
ASNO
0SIV
0TPS
6RAM
0LIV
0EKS
0RIK
-6-4-202468
]m
m[
Up
-6-4-202468
Fig. 10 Residuals of7-parameter Helmerttransformation between GPS-only coordinates andcoordinates obtained usingGPS + GLONASS data,computed respectively withIGS orbits and with CODEorbits
-2,5-2
-1,5-1
-0,50
0,51
1,5
ZGAC
ENEV
OMOC
OCAV
RZTW
EPOG
SUOP
KSIB
JRAM
CENS
TREH
2ZOJ
IGOB
NRAW
GLEH
SSAS
2EOH
ASNO
0SIV
0TPS
6RAM
0LIV
0EKS
0RIK
]m
m[D
TZ
Fig. 11 Mean bias betweenZTDs from GPS andGPS + GLONASS
GPS Solut (2007) 11:97–106 103
123
2005.8 2005.9 2006 2006.1
Epoch
2050
2100
2150
2200
2250
2300
2350]
DT
Zm
m[
GPSGPS + GLONASS
2050
2100
2150
2200
2250
2300
2350
2005.94 2005.941 2005.942
Epoch
2160
2200
2240
2280
]D
TZ
mm[
Fig. 12 Comparison of ZTDs based on GPS-only and on GPS + GLONASS observations
2005.8 2005.9 2006
Epoch
2250
2300
2350
2400
2450
2500]
DT
Z m
m[
GPSGPS + GLONASS
2250
2300
2350
2400
2450
2500
2005.8 2005.9 2006
Epoch
-10
-5
0
5
10
]D
TZ
mm[
-10
-5
0
5
10
Fig. 13 Top ZTDs from GPSand GPS + GLONASS;bottomZTD(GPS + GLONASS) –ZTD(GPS)
0123456
0123456
]m
m[
RMS North
GPS onlyGPS+GLONASS, with IGS orbits
0
1
2
3
]m
m[
RMS East
0
1
2
3
MMIZ
LRAK
AZTW
RZTW
UDE
R
POLK
KS
UE
SURB
SREH
TREH
GSO
K
BBT
P
STOP
TR
SW
UBOH
ERAD
KR
OB
NRAW
GLEH
2EOH
PR
OM
DIM
S
PDUB
DLUS
AS
NO
EVNI
0TPS
SLS
O
eg
arevA
0
1
2
3
4
5
]m
m[
RMS Up
0
1
2
3
4
5
Fig. 14 Coordinaterepeatabilities obtained usingGPS-only andGPS + GLONASSobservations
104 GPS Solut (2007) 11:97–106
123
GLONASS data in a regional network of GPS and
GPS/GLONASS receivers.
For all tests, we used the Bernese 5.0 data analysis
software, which allows to process GPS-only or
GPS + GLONASS observations using identically the
same processing strategy (except for the ambiguity
resolution).
We have compared the GPS-only and GPS +
GLONASS coordinates obtained in the two networks:
• a regional network consisting of 25 GPS/GLON-
ASS stations (all GPS/GLONASS included in the
EPN at January 2006)
• a typical regional network of mixed GPS and GPS/
GLONASS stations (20 GPS and 8 GPS/GLON-
ASS stations).
We compared the GPS + GLONASS coordinates
obtained from the GPS/GLONASS network using on
one hand the IGS orbits and on the other hand the
CODE orbits. The CODE orbits are consistent GNSS
orbits, while the IGS computes separately its combined
GPS and its GLONASS orbits. The GPS-only coordi-
nates were computed using the IGS final orbits. A first
conclusion is that the GPS + GLONASS-based coor-
dinates obtained using either IGS or CODE orbits
agree in all three components at the 1-mm level after
applying a 7-parameter Helmert transformation.
From the two networks processed, we can see that
adding GLONASS data to the GPS data does not
significantly change the repeatabilities of any of the
station coordinates. For some stations, the repeatabil-
ities are slightly better using GPS-only, for others, the
repeatabilities improve when adding GLONASS.
In the GPS/GLONASS network, the differences
between the GPS-only coordinates and the GPS
+ GLONASS coordinates show that adding GLON-
ASS data can change the coordinates at the level of
1–2 mm in the horizontal components and between 2
and 6 mm for the vertical component. For the hori-
zontal components, the coordinate differences are
mainly caused by reference frame differences between
the two regional networks. For the vertical component,
one of the stations in the network shows an offset of
almost 6 mm when GPS-only coordinates are com-
pared to GPS + GLONASS coordinates. The cause of
this difference is not clear presently and will be subject
of further study.
In the mixed network, which corresponds to the
reality, all coordinate differences are below the 1 mm
level.
Acknowledgment The author wishes to thank DominiqueMesmaker for his help with the figures.
References
Adam J, Augath W, Boucher C, Bruyninx C, Caporali A, GublerE, Gurtner W, Habrich H, Harsson B, Hornik H, Ihde J,Kenyeres A, van der Marel H, Seeger H, Simek J, Stangl G,Torres J, Weber G (2002) Status of the European ReferenceFrame—EUREF. In: IAG Symposia 125:42–46
Beutler G, Bock H, Brockmann E, Dach R, Fridez P, GurtnerW, Habrich H, Hugentobler U, Ineichen D, Jaeggi A, Me-indl M, Mervart L, Rothacher M, Schaer S, Schmid R,Springer T, Steigenberger P, Svehla D, Thaller D, Urschl C,Weber R (2006) Bernese GPS software version 5.0 (draft)eds Urs Hugentobler, R. Dach, P. Fridez, M. Meindl, Univ.Bern, 464 pp
-1
-0.5
0
0.5
1
-1
-0.5
0
0.5
1
]m
m[
NorthGPS versus GPS+GLONASS
-1
-0.5
0
0.5
1
]m
m[
East
-1
-0.5
0
0.5
1
MMIZ
LRAK
AZTW
RZTW
UDE
R
POLK
KSU
E
SUR
B
SREH
TREH
GSO
K
BBT
P
STOP
TR
SW
UBO
H
ERAD
KROB
NRAW
GLE
H
2EO
H
PRO
M
DIMS
PDUB
DLU
S
ASNO
EVNI
0TP
S
SLSO
-2
-1
0
1
2
]m
m[Up
-2
-1
0
1
2
Fig. 15 Coordinatedifferences obtained usingGPS-only andGPS + GLONASSobservations
GPS Solut (2007) 11:97–106 105
123
Bruyninx C (2004) The EUREF Permanent Network a multi-disciplinary network serving surveyors as well as scientists.GeoInformatics 7:32–35
Estey LH, Meertens C (1999) TEQC: the multi-purpose toolkitfor GPS–GLONASS data. GPS Solut 3(1):42–49
Gendt G (2006) IGSREPORT-14104: Wk 1383 IGLOS Final Orbits.http://www.igscb.jpl.nasa.gov/mail/igsreport/2006/msg00751.html
Ineichen D, Springer T, Beutler G (2001) Combined processingof the IGS and IGEX network. J Geod 75:575–586
Urschl C, Beutler G, Gurtner W, Hugentobler U, Schaer S(in press) GPS/GLONASS orbit determination based oncombined microwave and SLR data analysis. In: Proceed-ings of IAG, IAPSO, IABO Joint Conference, DynamicPlanet 2005, 22–26 August 2005, Cairns, Australia
Weber R, Fragner E (2000) IGEX Analysis. In: IGS 1999Technical Reports, pp 273–278
106 GPS Solut (2007) 11:97–106
123