Figure 3 depicts the UT1-UTC results for the two CONT campaigns. Shown are the ultra-rapid results (CONT-UR) together with the results derived from the INT sessions observed during the two CONTs, post-processing results with 1-h temporal resolution from the analysis of the complete CONT networks (CONT-1h), as well as the IVS quarterly solution with 24-h temporal resolution (IVS-Q). Results of three different INT analyses are presented, by the IVS analysis groups at the Bundesamt für Kartographie und Geodäsie (BKG), the Goddard Space Flight Center (GSFC), and the United States Naval Observatory (USNO). The ultra-rapid results presented were determined with a sliding-window approach using a window size of 6 h. The ultra-rapid results are displayed at the correct level, while all other data sets are offset by \(-5\) ms with respect to the corresponding previous data set, to improve visibility of the graphs.
For CONT11, there were only INT-1 sessions during the CONT, i.e., on the baseline Wettzell–Kokee. On the contrary, during CONT14 there were both INT-1 (baseline Wettzell–Kokee) during the weekdays, and INT-2 (baseline Wettzell–Tsukuba) during the weekends May 10/11 and May 17/18. Thus, for CONT14 there are no delay observations on the Onsala–Tsukuba baseline during these four INT-2 sessions. However, since the sliding window used was larger than the 1-h outage, there are nevertheless ultra-rapid UT1-UTC values reported during these gaps.
There is a large gap in the ultra-rapid results for CONT11 on September 21, because the Tsukuba telescope was stowed because of the Typhoon. Also the smaller gaps on September 25 and 26 that were caused by power outages at Onsala are visible.
In general, it can be noted that the different data sets show a very similar behavior and that the curves follow each other nicely. Referring the data sets to a common reference gives more insight into the details. As common reference, the IERS 08 C04 data were used and the corresponding residuals are presented in Fig. 4.
The UT1-UTC residuals presented in Fig. 4 clearly show that the ultra-rapid results derived from one baseline only have a larger variability than the results derived from analysis of the complete network data, both the post-processed results with 1-h temporal resolution and the IVS quarterly solution with 24-h temporal resolution. Also the INT results show a considerable scatter, and have of course a much worse temporal resolution as both the ultra-rapid results as well the 1-h post-processed network results. There are some small deviations of the ultra-rapid results probably caused by failure of the automated ambiguity resolution that is part of the automated processing. For the ultra-rapid CONT11 and CONT14 analysis, robust methods (Kareinen et al. 2016) were not part of the automated analysis chain yet.
Table 1 Information on the root-mean-square difference (rms), the bias (bias) and the standard deviation (\(\sigma \)) of the UT1-UTC results with respect to the IERS C04 series
Table 1 summarizes statistical information of these comparisons in terms of root-mean-square differences (rms), biases, and standard deviations. The IVS quarterly results agree in terms of standard deviation with the IERS 08 C04 on the level of 10-11 \(\upmu \)s and show biases of less than 5 \(\upmu \)s. The post-processed network results with 1-h temporal resolution agree slightly less good, in terms of standard deviation by a factor of 1.3–1.4 approximately. While there is no bias for CONT14, there is one of almost 12 \(\upmu \)s for CONT11. The INT results agree with IERS 08 C04 in terms of standard deviation during CONT11 on the level of 16–22 \(\upmu \)s, but for CONT14 on the level of 9–12 \(\upmu \)s. The biases for CONT11 are between 1 and 10 \(\upmu \)s while they are around 3–4 \(\upmu \)s for CONT14. The ultra-rapid results show both the largest biases and standard deviations on the order of 15–32 and 32–35 \(\upmu \)s, respectively. We can conclude from these comparisons that UT1-UTC derived from the analysis of networks with many globally well-distributed stations agree better with IERS 08 C04 than results derived on a single baseline. We also see that results derived from single baseline observations that were optimized for a single baseline agree better with IERS 08 C04 than results from single baseline analysis of observations that were not optimized for this purpose.
While the ultra-rapid results are about a factor of three worse than the results from the INT sessions, they are still better than UT1-UTC predictions. According to the IERS Bulletin-A, the prediction accuracy degrades as a function of time in days (D) since the last official IERS UT1-UTC value by
$$\begin{aligned} \sigma _\mathrm{UT1-UTC} = 250 \cdot D^{0.75}~~\left[ \upmu \mathrm{s}\right] . \end{aligned}$$
(1)
This means that after 6, 12, and 18 h, uncertainties on the order of 88, 149 and 202 \(\upmu \)s are to be expected for UT1-UTC predictions. The ultra-rapid results are better than this as confirmed by rms differences smaller than 50 \(\upmu \)s with respect to IERS 08 C04.