Assessment of the Earth orientation parameter accuracy from concurrent VLBI observations

Abstract

We have assessed accuracy of estimates of Earth orientation parameters (EOP) determined from several very long baseline interferometry (VLBI) observing programs that ran concurrently at different networks. We consider that the root mean square of differences in EOP estimates derived from concurrent observations is a reliable measure of accuracy. We confirmed that formal errors based on the assumption that the noise in observables is uncorrelated have a limited use. We found no evidence that advanced scheduling strategies with special considerations regarding the ability to better solve for atmospheric path in zenith direction applied for 1-hr single-baseline sessions have any measurable impact on the accuracy of EOP estimates. From this, we conclude that there is a certain limit in our ability to solve for the atmospheric path delay using microwave observations themselves and a scheduling strategy is not the factor that impairs accuracy of EOP determination. We determined that EOP errors vary with season, being smaller in winter and greater in summer. We got the quantitative estimate of the impact of unmodeled source structure on EOP estimates and we found that the seasonal extra variance is one order of magnitude greater than the impact of source structure. We have established that the EOP errors are scaled with an increase in duration of an observing session as a broken power law with the power of -0.3 at durations longer than 2-4 hours, which we explain as a manifestation of the presence of correlations in the atmospheric noise.

Figures (11)

• Figure 1: Station networks for different observing programs, color-coded by the number of experiments used in our analysis. Baseline from 1-hr experiments are explicitly depicted. Close telescopes (twins) are slightly offset in the maps.
• Figure 2: Illustration of reprocessing a given experiment four times using non-overlapping ranges of data shown with gray color and downweighting all other data.
• Figure 3: Differences in $E_3$ angle estimates between 22 single-baseline experiments at MgWs baseline versus estimates from concurrent ten-baseline 24-hr experiments. Left: experiment with code s22416 on December 13, 2023. Right: experiment with code s22423 on July 15, 2024.
• Figure 4: The normalized distributions of the $E_3$ angle differences derived from analysis of experiments at K2Ws (left) and KkWz baselines (right) divided by formal uncertainties. The blue smoothed lines show Gaussian distributions with the second moments 2.60 and 1.18 respectively.
• Figure 5: Differences in the $E_3$ angle estimates derived from processing data at single-baseline 1-hr experiments at KkWz baseline versus concurrent multi-baseline 24-hr r1 experiments as a function of time.