hide
Free keywords:
Wien, Techn. Univ., Diss., 2016. - Keine Mehrfachexemplare zur Abgabe mehr vorhanden.
Abstract:
Very long baseline interferometry (VLBI) is one of the fundamental space geodetic techniques.
Important goals for the next generation of VLBI technology are continuous operations as well
as automated data processing. For this reason, it is necessary to introduce real time capable
parameter estimation algorithms, such as Kalman filters, to VLBI data analysis. In this study,
such a filter was implemented in the VLBI software VieVS@GFZ, and several aspects related to
VLBI data processing were investigated. Within the corresponding module VIE_KAL it is possible,
for example, to estimate all parameters important in VLBI analysis, adapt their stochastic models,
flexibly define the datum, integrate external data, as well as extract datum free normal equations.
The foci of the investigations were on the effects of the troposphere, the most important error
source in VLBI analysis, and on the determination of station positions, which are of great
importance in geodesy. For the stochastic model of the tropospheric delays, station- and timedependent
differences were considered. In comparisons with tropospheric parameters from
GNSS, water vapor radiometers and numerical weather models, the Kalman filter solution yielded
5 to 15% smaller differences than a least squares solution based on the same models and VLBI
data. Also in the case of estimated station coordinates, the Kalman filter solution exhibited better
baseline length and station coordinate repeatabilities. The application of station-based process
noise led to additional improvements. Furthermore, the Kalman filter was used to estimate subdaily
station coordinate variations caused by tidal and loading effects.
Finally, the findings were used to determine Kalman-filter-based global terrestrial reference
frames (TRFs). For the stochastic model of the coordinate variations of particular stations, loading
deformation time series were utilized. The non-deterministic approach of the Kalman filter
allowed the consideration of non-linear station movement, for example, due to irregular seasonal
effects or post-seismic deformations. In comparisons with a VLBI TRF solution from a classical
adjustment and ITRF2008, a good agreement in terms of transformation parameters and station
velocities was achieved. The findings from testing different options related to the parameterization
and to the stochastic model will help to improve future reference frames.
