Abstract
Status and Prospects for Combined GPS LOD and VLBI UT1 Measurements A Kalman filter was developed to combine VLBI estimates of UT1-TAI with biased length of day (LOD) estimates from GPS. The VLBI results are the analyses of the NASA Goddard Space Flight Center group from 24-hr multi-station observing sessions several times per week and the nearly daily 1-hr single-baseline sessions. Daily GPS LOD estimates from the International GNSS Service (IGS) are combined with the VLBI UT1-TAI by modeling the natural excitation of LOD as the integral of a white noise process (i.e., as a random walk) and the UT1 variations as the integration of LOD, similar to the method described by Morabito et al. (1988). To account for GPS technique errors, which express themselves mostly as temporally correlated biases in the LOD measurements, a Gauss-Markov model has been added to assimilate the IGS data, together with a fortnightly sinusoidal term to capture errors in the IGS treatments of tidal effects. Evaluated against independent atmospheric and oceanic axial angular momentum (AAM + OAM) excitations and compared to other UT1/LOD combinations, ours performs best overall in terms of lowest RMS residual and highest correlation with (AAM + OAM) over sliding intervals down to 3 d. The IERS 05C04 and Bulletin A combinations show strong high-frequency smoothing and other problems. Until modified, the JPL SPACE series suffered in the high frequencies from not including any GPS-based LODs. We find, surprisingly, that further improvements are possible in the Kalman filter combination by selective rejection of some VLBI data. The best combined results are obtained by excluding all the 1-hr single-baseline UT1 data as well as those 24-hr UT1 measurements with formal errors greater than 5 μs (about 18% of the multi-baseline sessions). A rescaling of the VLBI formal errors, rather than rejection, was not an effective strategy. These results suggest that the UT1 errors of the 1-hr and weaker 24-hr VLBI sessions are non-Gaussian and more heterogeneous than expected, possibly due to the diversity of observing geometries used, other neglected systematic effects, or to the much shorter observational averaging interval of the single-baseline sessions. UT1 prediction services could benefit from better handling of VLBI inputs together with proper assimilation of IGS LOD products, including using the Ultra-rapid series that is updated four times daily with 15 hr delay.
Highlights
Variations in the rotation rate of the Earth occur on all observable timescales and are caused by both external and internal processes
The International Earth Rotation and Reference Systems Service (IERS) and other groups provide regularized tabulations of UT1-UTC and length of day (LOD) variations derived from multi-technique combinations of Very Long Baseline Interferometry (VLBI) and satellite observations
Deficiencies in the Global Positioning System (GPS) observations of the Earth’s spin are presumably accommodated in the satellite orbit parameters to agree with the unbiased VLBI rotation rate. While this method might yield improved GPS orbits, that possibility was not examined and the utility for UT1/LOD combination was only tested by finding a reduction in RMS differences with the IERS C04 series from 15 to 11 μs
Summary
Variations in the rotation rate of the Earth occur on all observable timescales and are caused by both external and internal processes. The International Earth Rotation and Reference Systems Service (IERS) and other groups provide regularized tabulations of UT1-UTC and LOD variations derived from multi-technique combinations of VLBI and satellite observations. These products are required for a broad range. Deficiencies in the GPS observations of the Earth’s spin are presumably accommodated in the satellite orbit parameters to agree with the unbiased VLBI rotation rate While this method might yield improved GPS (inertial) orbits, that possibility was not examined and the utility for UT1/LOD combination was only tested by finding a reduction in RMS differences with the IERS C04 series from 15 to 11 μs. Previous results from this work have been reported previously (Senior et al, 2008a, 2008b)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have