Sie sind auf Seite 1von 19

A NEW TROPOSPHERIC DELAY MODEL FOR

HIGH-ALTITUDE VEHICLE POSITIONING

Toshiaki Tsujii, Jinling Wang, Chris Rizos


School of Geomatic Engineering,
University of New South Wales,
Sydney, NSW 2052, Australia
Email: Tsujii@nal.go.jp

Masatoshi Harigae, Toshiharu Inagaki


Flight Systems Research Center,
National Aerospace Laboratory of Japan

ABSTRACT

GPS can play an important role in high-altitude vehicle positioning. In such GPS
applications the residual tropospheric delay in GPS measurements is a major error
source. Several strategies for estimating the residual zenith tropospheric delay have
been described in the scientific literature. In general, such models incorporate the time-
varying residual tropospheric errors within the stochastic models, which are then
estimated using a Kalman filter. However, these methods have not considered the
height difference between GPS receivers. In this paper, a new functional model for
tropospheric delay is proposed. The error of the meteorological data measured at a
ground reference site is estimated by Kalman filtering. Analytical results of a flight
experiment indicate that the proposed model is indeed superior to the standard models.
The new model can compensate for errors of meteorological data, and is feasible for use
in precise positioning of high-altitude receiver platforms.
1. INTRODUCTION
The residual tropospheric delay in GPS carrier phase observations is one of the most
significant error sources for precise kinematic (as well as static) applications. Many
investigators have proposed error mitigation techniques based on ground-based GPS
reference receiver networks (e.g. Han & Rizos, 1996; Gao et al., 1997; Wanninger,
1997; Zhang, 1999; to name just a few). These techniques appear to be well suited for
both static surveying and mobile vehicle positioning, on the ground. The typical
mitigation strategy is to estimate the effect of the residual delay on the user receiver by
interpolating the delay determined at a number of GPS reference receivers. However,
because the tropospheric delay depends on receiver altitude, this technique is effective
only when all receivers (including the user receiver) are located at a similar height. The
use of tropospheric delay models also has its problems, as on-the-fly ambiguity
resolution becomes critical in such circumstances. Therefore, these techniques may not
be feasible for precise positioning of high-altitude receiver platforms.

Japan has been conducting research on an airship system that will function as a
stratospheric platform (altitude of about 20km) for such purposes as environmental
monitoring, communications and broadcasting (Fig. 1). With respect to the first

Figure 1. Conceptual image of the stratospheric airship system.


application, if stratospheric airships are available in future for earth observation, they
will contribute significantly to improving forecasts and mitigating the effects of
disasters in the following ways:
- Intensive and continuous observation of specific areas.
- Acquisition of high accuracy data at the same point.
- Observation of objects from many viewpoints using multiple platforms.
The precise positioning of the airship(s) is one of the most important technical
challenges for such a project.

In the existing models for ground-based applications, the residuals of the relative zenith
tropospheric delay between receiver sites, or the scale factor for nominal values, is
estimated. However, estimating the residual of the relative zenith delay would not be
adequate for airborne applications because its partial derivative is not dependent on the
height difference between the reference and user antennas. The partial derivative for the
scale factor can depend on the height difference. However, adjusting the scale factor is
a computational technique to compensate the residual, but the physical meaning of the
estimate is ambiguous.

In this paper, a new functional model of tropospheric delay for use in high-altitude
vehicle positioning is proposed. The proposed functional model is explicitly dependent
on the height difference between observation sites, and one can select an appropriate
parameter to be estimated according to the kinds of meteorological sensors that are used,
and their accuracies.

TROPOSPHERC DELAY MODEL


The tropospheric propagation delay of GPS signals can be written in the general form as
(in metres):
d trop = d hZ mh + d wZ mw (1)

where
d hZ : hydrostatic zenith delay

d wZ : wet zenith delay


mh : hydrostatic mapping function

mw : water vapour mapping function

Niell’s mapping function is used in this work. In the case of the hydrostatic zenith delay,
the Saastamoinen model is used (Davis et al., 1985):
10−6 k1 Rd
d = ⋅P
Z
h (2)
gm
and the wet zenith delay is given by (Askne & Nordius, 1987):
10−6 (Tm k2′ + k3 )Rd e
dwZ = ⋅ (3)
gmλ′ − βRd T
where
k1 , k 2′ , k 3 : refractivity constants (Thayer, 1974)

Rd : gas constant for dry air

gm : gravity acceleration at the mass centre of a vertical column of the atmosphere

β : temperature lapse rate


λ′ : = λ + 1 ( λis water vapour lapse rate)
Tm : mean temperature of water vapour

= T (1 − βR d g m λ′)

P, T , e are pressure (in mbar), temperature (in degrees K), and water vapour pressure
(in mbar) measured at the observation site. Equation (3) agrees numerically with
Saastamoinen’s formula (Saastamoinen, 1973), by using β = 6.2, λ = 3 and the
refractivity constants used in Saastamoinen’s formula.

When airborne meteorological data are available, the tropospheric delay for the aircraft
can be calculated according to equations (2) and (3). Otherwise, the meteorological
data at the aircraft are estimated by introducing the temperature lapse rate and the water
vapour lapse rate (Askne & Nordius 1987; Collins et al., 1996).
g
 β(h − h0 )  R β
P = P0  1 −
d

 (4)
 T0 
λ ′g
e e 0  β (h − h0 )  R
−1

=
T T0 
1−  (5)
T0 
where P0 , T0 , e0 , h0 , g are pressure, temperature, water vapour pressure, height, and

gravity acceleration at the surface of the earth.

Equation (4) can be used below the tropopause ( h1 ≅ 11 km). In the lower stratosphere
( h ≅ 11-20 km), the temperature is identical to that of the tropopause (i.e. β =0), and
the pressure can be modelled as follows (U.S. Standard Atmosphere, 1976):
 g (h − h1 ) 
P = P1 ⋅ exp  −  (6)
 R d T1 

where P1 ,T1 are pressure and temperature at the bottom of lower stratosphere.
Combining equations (4) and (6), the pressure in the lower stratosphere with respect to
the surface meteorological data can be written as:
g
 β(h1 − h0 )  Rd β  g(h − h1) 
P = P0  1 −  ⋅ exp  −  (7)
   R {T − β (h − h ) }
T0  d 0 1 0 

The wet delay for a vehicle in the stratosphere will be negligible.

FUNCTIONAL MODELS
Based on the above discussion, the functional model of the double-differenced
tropospheric delay can be derived. The double-difference between the 1st and k-th
satellite, and between airborne and base receivers is given by:
∇∆d1trop
k
= dtrop,
1
a − d trop,a − d trop,b + d trop, b
k 1 k

= dtrop, a(ma − ma ) − dtrop, b (mb − mb )


Z 1 k Z 1 k
(8)
≅ (dtrop,
Z
a − d trop, b )(mb − m b )
Z 1 k

where the mapping function at the vehicle is approximated by that at the base (or
reference) receiver. In fact, Niell’s wet mapping function doesn’t depend on the
observation height. This expression can be used for the hydrostatic, wet, or total delay.
If this equation is applied to the total delay, the hydrostatic mapping function should be
used (Tralli & Lichten, 1990).
Previous studies have estimated the residual of the relative zenith delay or the scale
factor to the nominal values (Tralli & Lichten, 1990; Dodson et al., 1996; Collins et al.,
1996). The partial derivative for the residual of the relative zenith delay is the
differential mapping function, while that for the scale factor is the right hand side of
equation (8). Estimating the residual of the relative zenith delay would not be
appropriate for airborne applications because the characteristics of the residual
dependent on height difference are not taken account in the partial derivative. On the
other hand, although the partial derivative with respect to the scale factor depends on
the height difference, the physical meaning of the estimate is unclear.

The double-differenced tropospheric delay can be written as a function explicitly


dependent on the height difference. The function of the double-differenced hydrostatic
delay for the vehicle in the troposphere is obtained by incorporating equations (2) and
(4) into equation (8):
 g

10 − 6 k1 Rd   β∆h  Rd β 
∇∆d h1k = P0  1 −  − 1 ( m1b − mbk ) (9)
gm  T0  
 

The simplest way to estimate the residual hydrostatic delay is to estimate the error of
surface pressure assuming that the surface temperature and the temperature lapse rate
are correct. In this case, the partial derivative is given by:

 g

∂∇∆d 1k −6
10 k1 Rd  
β∆h β R

 
d
h
=  1 −  − 1(m1b − mbk ) (10)
∂P0 gm  T0  
 

One possible option would be to estimate the temperature lapse rate by assuming that
the meteorological data were accurately measured. Hence, equation (10) becomes:
g
∂∇∆dh1k 10 −6 k1 P0  log(1 − β∆h /T0 ) ∆h  β∆h  Rd β 1
=  +  1 − (mb − mbk ) (11)
∂β β  β T0 − β∆h  T0 

The function of the double-differenced wet delay for the vehicle in the troposphere is
given by incorporating equations (3) and (5) into equation (8):
 λ′g
−1 
10 − 6 (Tm k 2′ + k3 ) Rd e0   β ∆ h  Rd β 
∇∆d 1wk = ⋅   1 −  − 1( m b1 − m bk ) (12)
g m λ ′ − β Rd T0   T0  
 

Again, the simplest method is to estimate the error of water vapour pressure measured at
the surface. Another option is to estimate the lapse rate by assuming that the
meteorological data are accurately measured. The partial derivatives can be formulated
in a similar way to equations (10) and (11).

ESTIMATION SCENARIOS
The method of estimation of the residual tropospheric delay, and the selection of the
parameter to be estimated, should be dependent on the circumstances of the experiments.
Are meteorological data available for both airborne and base receivers, or only for the
base receiver? What is the accuracy of the meteorological sensor(s)? Is the vehicle in
the troposphere or the stratosphere? Table 1 summarizes the suitable parameters to be
estimated for some of these cases.

Table 1. Summary of estimation parameters for various experimental scenarios.


Surface Airborne met. In troposphere In Stratosphere
met. data Data
Residual of relative Residual wet delay
YES YES
Case1 wet delay at the surface
λ( β fixed)
Case2 YES NO β ( λfixed) β
e0 ( β , λfixed)
NO NO P0 or e0 P0
Case3 roughly
( β , λfixed) ( β fixed)
measured

In Case 1, where accurate meteorological data for both airborne and base receivers are
available, the hydrostatic component of the tropospheric delay can be well computed by
the model and does not have to be estimated. Then the residual of the relative wet delay
would be the estimable parameter if the vehicle were in the troposphere. If the vehicle
were in the stratosphere, the residual wet delay at the surface would be estimated. In
these cases, the partial derivatives for the estimate do not depend on the height
difference. Therefore the estimate of the residual tropospheric delay would be carried
out as if the vehicle were on the ground. The meteorological data collected at the
aircraft/airship are useful for estimating the lapse rates and their temporal/regional
variation.

In Case 2, where only the ground meteorological data are available, the estimated
parameter would be the temperature lapse rate or the water vapour lapse rate, assuming
that the meteorological data were accurate and the delay were well modelled. When the
vehicle flies in the stratosphere, the temperature lapse rate is the suitable parameter to
be estimated because it is affected only by the hydrostatic delay. Even if the water
vapour pressure was measured at the surface, the wet delay may not be able to be
accurately computed using the models due to the inhomogeneous nature of the water
vapour. Therefore the error in the water vapour pressure would be a candidate as
estimable parameter if the vehicle were in the troposphere.

If roughly measured meteorological data or only 'standard model' data are available
(Case 3), the error in the meteorological data should be estimated with the lapse rates
fixed. (This paper demonstrates the effectiveness of the height dependent partial
derivatives used in Case 3.)

In all cases, the total (or wet) zenith delay for the ground receiver, determined from the
processing of continuous GPS network data, would be useful to improve the accuracy of
the tropospheric residual estimate. This is because the residual tropospheric delay for a
high-altitude vehicle can be attributed to the wet delay for the ground-based reference
receiver. This option will be further investigated in future work.

FLIGHT EXPERIMENT
Flight tests using the research aircraft (Dornier-228) of the National Aerospace
Laboratory (NAL) were conducted on 29 October 1999 in Hokkaido, Japan. A Trimble
4000SSI dual-frequency GPS receiver was installed in the aircraft. A reference receiver,
also a Trimble 4000SSI, was set up near the airfield. While the maximum distance from
the aircraft to the airfield was less than 20km, the aircraft flew as high as it could. As
shown in Figure 2, the aircraft flew up and down along a virtual cylinder in the air.
Positions are in a runway coordinate system whose origin is at the threshold of the
runway. The X-axis is along to the runway, and the Y-axis is perpendicular to the X-
axis in the horizontal plane. The integer ambiguities were determined prior to the flight,
and cross-checked with ambiguities determined by backward processing. All of the
analyses were carried out using a modified version of the kinematic GPS software
(KINGS), developed at NAL (Tsujii et al., 1998).

7000

6000

5000

4000
Z (m)

3000

2000

1000

0
15000

10000 10000
5000
5000
0

0 -5000
-10000
-5000 -15000
Y (m)
X (m)

Figure 2. Three-dimensional trajectory of the aircraft.

The height profile and the elevation angles of the observed GPS satellites are plotted in
Figure 3. The elevation mask angle was set at twelve degrees. As discussed in Collins
et al. (1996), only low elevation satellites have the observability for the estimation of
the tropospheric delay. However, the multipath effect is serious if the signals from very
low satellites are used. Unfortunately, the number of low elevation satellites observed
was not sufficient for this experiment. The aircraft height with observation of low
elevation satellites was up to 4.5 km, although the aircraft flew up to 6.5km altitude.
8000

6000

Height (m)
4000

2000

0
445000 446000 447000 448000 449000 450000 451000
GPS time (sec)

80
Elevation (deg)

60

40

20

445000 446000 447000 448000 449000 450000 451000


GPS time (sec)

Figure 3. Height profile of the aircraft (top) and elevation of GPS satellite (bottom).

The meteorological data (pressure, temperature, dew point) were measured at the
ground site and recorded every ten minutes, but no measurement was made at the
aircraft. The measured and interpolated values are shown in Figure 4. The surface dew
point was converted to water vapour pressure (Smith, 1966) and the wet zenith delay
was computed using equation (3). Since the recording rate was low, and low-cost

19
temp (deg)

18

17

16
445000 446000 447000 448000 449000 450000 451000

11
dew point (deg)

10

8
445000 446000 447000 448000 449000 450000 451000

1007
pressure (mb)

1006.5

1006

1005.5

1005
445000 446000 447000 448000 449000 450000 451000
GPS time (sec)

Figure 4. Measured and interpolated meteorological data during the experiment.


meteorological sensors were used, the results from this experiment are close to those
expected for Case 3 as indicated in Table 1.

RESULTS
In order to evaluate the performance of the tropospheric parameter estimate, the true
value of the tropospheric delay or the true position of the aircraft is necessary. Another
challenge for this flight experiment is that no meteorological data were measured at the
aircraft. It might be considered that the position estimate obtained by processing the
triple-differenced carrier phase measurements would be nearly correct because a large
part of the tropospheric delay would cancel by differentiating the data from successive
epochs. However, this expectation would be wrong for these kind of applications.
Figure 5 shows the differences of position estimated using triple-differenced (L1) carrier
phase data when the surface meteorological data are different. The dew point
temperatures used at the surface were 7.5, 8.5, 9.5, 10.5, and 11.5 degrees, while the
pressure and the temperature for all computations were the measured values as shown in
Figure 4. The trajectory with water vapour pressure 9.5 was selected as the reference
trajectory. It is clear that the height component accuracy varies with the height of

0.05
dx (m)

-0.05
445000 446000 447000 448000 449000 450000 451000

0.05
dy (m)

-0.05
445000 446000 447000 448000 449000 450000 451000

0.1

0.05
dz (m)

-0.05

-0.1
445000 446000 447000 448000 449000 450000 451000
GPS time (sec)

Figure 5. Position difference according to the various dew points (7.5, 8.5, 9.5, 10.5,
11.5 degs). The reference trajectory was computed with dew point 9.5.
aircraft, although the horizontal components are almost unchanged. This fact indicates
that the tropospheric delay changed drastically due to the ascending/descending motion
of aircraft so that the delay was not removed by time-differentiation. Therefore, even
the trajectory obtained from triple-differenced processing could not be used as the true
trajectory if accurate meteorological data were not applied.

If it is assumed that the surface meteorological data shown in Figure 4 are sufficiently
accurate, there are two unknown parameters: the temperature lapse rate β and the water
vapour lapse rate λ. The temperature lapse rate was fixed at 6.5 K/km, as adopted in
the UNB4 model (Collins et al., 1996) and in the U.S. Standard Atmosphere (1976)
because it is rather more constant than the water vapour lapse rate, which is changeable
both in a temporal and regional sense. Next, the water vapour lapse rate is selected so
as to minimize the measurement residuals computed from processing (L1-carrier)
double-differences with known (fixed) ambiguities. The chosen value 1.8 is smaller
than that of the UNB3 model (2.8) and Saastamoinen’s (3.0). However, this is not
unrealistic considering that it may contain the error of surface meteorological data and
the temperature lapse rate. Then, the true trajectory is defined as the one computed
using (L1-carrier) double-differences with known ambiguities, surface meteorological
data, and ( β , λ) = (6.5, 1.8).

The results of estimating the residual tropospheric delay simulating the situation of Case
3 are discussed below.

Firstly, the nominal trajectory is computed by the least squares method using the L1-
carrier double-differences with known ambiguities, constant surface meteorological data,
and ( β , λ) = (6.5, 1.8). The residual tropospheric delay was not estimated at this time.
The surface temperature and pressure were fixed at 17.5 degrees and 1006 mbar
respectively, while the surface dew point is changed as a control parameter. Next, the
residual tropospheric delay is estimated using a Kalman filter, as well as the deviation
from nominal positions. The ambiguities were fixed to the pre-determined values. In
the following computations, the error of the water vapour pressure was estimated using
the proposed functional model. The dynamics of the error of water vapour pressure and
the deviation from the nominal trajectory were assumed to be a random walk.

Figure 6 shows the position errors with/without water vapour pressure estimated
compared against the 'true' trajectory. The surface dew point was set at 8.5 degrees,
which is a little lower than the measured values shown in Figure 4. These show the
results until the aircraft ascends to 4.5 km, since the only low elevation satellites are
useful for the estimation of residual tropospheric delay.

0.05
dx (m)

-0.05
445000 446000 447000
0.05
dy (m)

-0.05
445000 446000 447000
0.1
dz (m)

-0.1
445000 446000 447000
GPS time (sec)

Figure 6. Position errors with (bold line) and without water vapour pressure
estimate when the surface dew point was set at 8.5 degrees.

It can be seen that some components of the height error were recovered by estimating
the error of the water vapour pressure. There is no significant difference in horizontal
position errors between the cases with or without residual estimation. (The errors in
horizontal direction are not shown.)

Figure 7 shows the height error when the surface dew point was set at 7.5 degrees. If
the error of the water vapour pressure was not estimated, the height error reached 10cm,
according to the increasing height difference between receivers. On the other hand, a
large part of the height error was recovered by estimating the error of the water vapour
pressure.

0.1

0.05
dz (m)

-0.05

-0.1
445000 446000 447000 GPS Time (sec)

Figure 7. Height error with (bold line) and without water vapour pressure estimate
when the surface dew point was set at 7.5 degrees.

Figures 8 and 9 show the height errors when the dew point was set at 10.5 and 11.5
degrees, which are higher than the measured values. Although some of the height errors
were recovered, the efficiency of residual estimation was not sufficient, compared with
the previous two cases. This can be explained by considering the actual temporal
change of meteorological conditions. Since the dew point decreased when the aircraft
ascended, the difference between assumed dew points (10.5 and 11.5) and actual ones
increased. It would be difficult to estimate the residual tropospheric delay in such a
case, where only one low elevation satellite was observed.
0.1

0.05
dz (m)

-0.05

-0.1
445000 446000 447000 GPS Time (sec)

Figure 8. Height error with (bold line) and without water vapour pressure estimate
when the surface dew point was set at 10.5 degrees.

0.1

0.05
dz (m)

-0.05

-0.1
445000 446000 447000 GPS Time (sec)

Figure 9. Height error with (bold line) and without water vapour pressure estimate
when the surface dew point was set at 11.5 degrees.
In addition, a first order Gauss-Markov process was assumed instead of the random
walk process for the dynamics of the error of the water vapour pressure. Results are
shown in Figures 10 and 11 when the surface dew point was set at 8.5 and 7.5 degrees.
The correlation time for the Gauss-Markov process was chosen to be 1000 seconds.
These plots demonstrate the effectiveness of the Gauss-Markov process as a stochastic
model. However, in order to establish the most suitable stochastic model for this
application, further investigations are necessary using a variety of data sets.

0.1

0.05
dz (m)

-0.05

-0.1
445000 446000 447000 GPS Time (sec)

Figure 10. Height error with (bold line) and without water vapour pressure estimate
using Gauss-Markov model when the surface dew point was set at 8.5
degrees.

0.1

0.05
dz (m)

-0.05

-0.1
445000 446000 447000 GPS Time (sec)

Figure 11. Height error with estimation of the water vapour pressure estimate using
Gauss-Markov model (bold line), and without estimation. The surface
dew point was set at 7.5 degrees.

Now, the results of estimating the residual tropospheric delay using the conventional
functional model are considered for a comparison. The residual of the relative wet
zenith delay was estimated, in which the partial derivative was not dependent on height
difference. The random walk process was assumed as the dynamics of the residual.
Figure 12 shows the height errors with or without estimating the residual tropospheric
delay. The meteorological data used were the same as in Figure 7. At the beginning of
aircraft ascent, the trend of the height error with residual delay estimation was opposite
to that without estimation. This means that the residual delay was over-estimated.

0.1

0.05
dz (m)

-0.05

-0.1
445000 446000 447000 GPS Time (sec)

Figure 12. Height error with estimation of the water vapour pressure estimate using
height-independent partial derivative (bold line), and without estimation.
The surface dew point was set at 7.5 degrees.

This result can be understood if the partial derivatives used for this estimation (Figure
13, top) are compared with the proposed partial derivatives (Figure 13, bottom). The
higher the aircraft flew, the larger the residual would be. Therefore, the use of
inadequately large partial derivatives at low height would cause this over-estimation of
the residual delay.

4
partial derivative

0
445000 446000 447000
GPS time (sec)

0.04
partial derivative

0.03

0.02

0.01

0
445000 446000 447000
GPS time (sec)

Figure 13. The conventional partial derivatives for estimation of residual


tropospheric delay (top) and the proposed height-dependent partial
derivatives (bottom).
CONCLUDING REMARKS
A new functional model of the residual tropospheric delay has been proposed in this
paper, which is explicitly height-dependent. The effectiveness of this new model for the
precise positioning of high-altitude vehicles has been demonstrated. Depending on the
types of meteorological sensors installed at the ground/airborne sites and their
accuracies, it would be possible to choose the appropriate physical parameter to be
estimated. This feature will be useful for a Japanese stratosphere airship as it is being
developed, and for which various types of sensors for earth environment research will
be installed.

This work treated only the short baseline case in order to focus on estimating the
residual tropospheric delay. In a further study, the possibility of estimating the
ambiguities as well as the residual tropospheric delay for high-altitude vehicle
positioning will be investigated. The use of GLONASS will be important for increasing
the number of low elevation satellites (Wang, 1998; Han et al. 1999; Tsujii et al., 2000).
For long baseline applications, the zenith tropospheric delay for ground receivers
determined from the processing of data from a continuously operating GPS network
would be useful.

ACKNOWLEDGMENTS

The authors would like to thank the staff of Taiki town office for their kind support
during the experiment. The first author would like to acknowledge support from the
Japan Science and Technology Corporation (JST) postdoctoral fellowship scheme.

REFERENCES
Askne, J. & H. Nordius (1987), Estimation of tropospheric delay for microwaves from
surface weather data. Radio Science, 22(3), 379-386.
Collins, J.P., R.B. Langley & J. LaMance (1996), Limiting factors in tropospheric
propagation delay error modelling for GPS airborne navigation. 52nd U.S. Institute
of Navigation Annual Meeting, Cambridge, MA, 19-21 June., 519-528.
Collins, J.P. & R.B. Langley (1997), Estimating the residual tropospheric delay for
airborne differential GPS positioning. 10th Int. Tech. Meeting of the Satellite
Division of the U.S. Inst. of Navigation, Kansas City, Missouri, 16-19 Sept., 1197-
1206.
Davis, J.L., T.A. Herring, I.I. Shapiro, A.E.E. Rogers & G. Elgered (1985), Geodesy by
radio interferometry: Effects of atmospheric modeling errors on estimates of
baseline length. Radio Science, 20(6), 1593-1607.
Dodson, A.H., P.J. Shardlow, L.C.M. Hubbard, G. Elgered & P.O.J. Jarlemark (1996),
Wet tropospheric effects on precise relative GPS height determination. Journal of
Geodesy, 70, 188-202.
Gao, Y., Z. Li & J. McLellan (1997), Carrier phase based regional area differential GPS
for decimeter-level positioning and navigation. 10th Int. Tech. Meeting of the
Satellite Division of the U.S. Inst. of Navigation, Kansas City, Missouri, 16-19 Sept.,
1305-1313.
Han, S. & C. Rizos (1996), GPS network design and error mitigation for real-time
continuous array monitoring systems. 9th Int. Tech. Meeting of the Satellite Division
of the U.S. Inst. of Navigation, Kansas City, Missouri, 17-20 Sept., 1827-1836.
Han, S., L. Dai & C. Rizos (1999), A new data processing strategy for combined
GPS/Glonass carrier phase-based positioning. 12th Int. Tech. Meeting of the
Satellite Division of the U.S. Inst. of Navigation, Nashville, Tennessee, 14-17 Sept.,
1619-1627.
Niell, A.E. (1996), Global mapping functions for the atmosphere delay at radio
wavelengths. Journal of Geophysical Research, 101(B2), 3227-3246.
NOAA, NASA, U.S. Air Force (1976), U. S. Standard Atmosphere, Washington, D.C.
Saastamoinen, J. (1973), Contributions to the theory of atmospheric refraction. In three
parts. Bulletin Geodesique, 105, 279-298; 106, 383-397; 107, 13-34.
Smith, W.L. (1966), Note on the relationship between total precipitable water and
surface dew point. Journal of Applied Meteorology, 5, 726-727.
Thayer, G.D. (1974), An improved equation for the radio refractive index of air. Radio
Science, 9(10), 803-807.
Tralli, D.M. & S.M. Lichten (1990), Stochastic estimation of tropospheric path delays in
Global Positioning System geodetic measurements. Bulletin Geodesique, 64, 127-
159.
Tsujii, T., M. Murata, M. Harigae, T. Ono & T. Inagaki (1998), Development of
Kinematic GPS Software, KINGS, and flight test evaluation. Technical Report of
National Aerospace Laboratory, Japan, TR-1357T.
Tsujii, T., M. Harigae, T. Inagaki & T. Kanai (2000), Flight tests of GPS/GLONASS
precise positioning versus dual frequency KGPS profile. Earth Planets Space,
52(10), 825-829.
Wang, J. (1998), Mathematical models for combined GPS and GLONASS positioning.
11th Int. Tech. Meeting of the Satellite Division of the U.S. Inst. of Navigation,
Nashville, Tennessee, 15-18 Sept., 1333-1344.
Wanninger, L. (1997), Real-time differential GPS error modeling in regional reference
station networks. IAG Symp. 118, Rio de Janeiro, Brazil, 86-92.
Zhang, J. (1999), Precise estimation of residual tropospheric delays in a spatial GPS
network. 12th Int. Tech. Meeting of the Satellite Division of the U.S. Inst. of
Navigation, Nashville, Tennessee, 14-17 Sept., 1391-1400.

Das könnte Ihnen auch gefallen