1 Kashima Space Research Center
Communications Research Laboratory
893-1 Hirai, Kashima, Ibaraki 314-0012, Japan
2Communications Research Laboratory
4-2-1 Nukui-kita, Koganei, Tokyo 184-8795, Japan
Abstract: VLBI measurements using four fixed VLBI stations around the Tokyo
metropolitan area are producing continuous data of station positions and
baseline lengths. Accuracy of baseline length measurements is evaluated on the
basis of their repeatability in terms of root mean square variance in five
adjacent sessions. Five day continuous observation sessions demonstrate that
typical repeatability of about 1-2 mm in baseline length is achieved on our VLBI
network.
1. Introduction
We have been carrying out routine daily VLBI observations (6 hours per day) to
monitor crustal deformation around the Tokyo metropolitan area using four
stations: Kashima, Koganei, Miura and Tateyama (Figure 1). The project is named the Key
Stone Project (KSP) and is promoted by the Communications Research Laboratory.
Each VLBI station is equipped with the same VLBI facility, i.e., a parabolic
antenna with 11 m diameter and a highly automated data acquisition system
dedicated to KSP. The longest distance between KSP stations is about 135 km
(Kashima-Tateyama), so that the KSP network is very compact as a VLBI network.
The KSP started regular observations using Kashima and Koganei stations in
January, 1995. Later other stations joined and daily observations using all
four stations started in September, 1996. Since then the observation system
has experienced some refinements to improve its total performance.
In parallel with the daily observations using a conventional tape recording
method, we were establishing a real time correlation processing system using an
ATM (asynchronous transmitting mode) network which combines four stations with a
high speed digital link (maximum speed is 2.4 Gbps). Digitized signals observed
at each station are transmitted to Koganei, where a KSP correlator is located,
in real time through the ATM network. This real time processing has been used
in routine operations since June, 1997.
Figure 1. Key stone project VLBI network.
As VLBI/KSP enters a stable operation phase, we have begun an evaluation of
total system performance in terms of measurement accuracy. How accurately can
we measure baseline length among KSP stations using the current system, which
means both hardware and analysis software? To answer this inquiry, we conducted
continuous observation over 120 hours on the KSP network from July 28 to August
1, 1997. These observations were successfully finished and the formal error of
baseline length estimation for the last day is 0.7 mm, which is the champion
value of KSP at present time. In this paper, we make an evaluation of results
of the 120-hour-observation by comparing them with KSP results taken at other
times.
2. Observations
Continuous 120 hour observation was carried out from July 28 to August 1, 1997
and was divided into 5 sessions. Each session lasts for 24 hours and includes
about 600 scans of radio sources (quasars). Unmanned observations according to
the same observation schedule distributed from the KSP central station, Koganei,
were carried out at the all four stations. Observation status is always
monitored at Koganei automatically. An observation and its correlation
processing for 6 baselines is carried out simultaneously using data transmitted
from each station through an ATM network which connects all stations to the
Koganei central station.
Table 1 summarizes scheduled number of scans and valid scan number by session by
baseline. Valid scan number means the number used for a baseline analysis.
During the second session some problems occurred on the correlation processor,
and Miura station was down for about 5 hours on the fourth session. These
problems resulted in the low values of valid scans during the second and fourth
session compared with the other sessions.
A baseline analysis is made on the basis of each session.
Table 1. Summary of observations.
Session (Date)
Baseline
Number of Scans
Scheduled
Valid for Analysis
Valid/Scheduled(%)
97JUL28XX (97/07/28) 01:10-24:51
KASHIMA-KOGANEI
593
487
82.1
KASHIMA-MIURA
593
550
92.7
KASHIMA-TATEYAMA
593
517
87.2
KOGANEI-MIURA
593
518
87.4
KOGANEI-TATEYAMA
593
510
86.0
MIURA-TATEYAMA
593
546
92.1
97JUL29XX (97/07/29) 01:13-24:51
KASHIMA-KOGANEI
598
390
65.2
KASHIMA-MIURA
598
386
64.5
KASHIMA-TATEYAMA
598
419
70.1
KOGANEI-MIURA
598
417
69.7
KOGANEI-TATEYAMA
598
395
66.1
MIURA-TATEYAMA
598
462
77.3
97JUL30XX (97/07/30) 01:10-24:51
KASHIMA-KOGANEI
598
507
84.8
KASHIMA-MIURA
598
496
82.9
KASHIMA-TATEYAMA
598
519
86.8
KOGANEI-MIURA
598
489
81.8
KOGANEI-TATEYAMA
598
507
84.8
MIURA-TATEYAMA
598
556
93.0
97JUL31XX (97/07/31) 01:13-24:54
KASHIMA-KOGANEI
599
483
80.6
KASHIMA-MIURA
599
446
74.5
KASHIMA-TATEYAMA
599
532
88.8
KOGANEI-MIURA
599
426
71.1
KOGANEI-TATEYAMA
599
504
84.1
MIURA-TATEYAMA
599
455
76.0
97AUG01XX (97/08/01) 01:17-24:52
KASHIMA-KOGANEI
597
526
88.1
KASHIMA-MIURA
597
568
95.1
KASHIMA-TATEYAMA
597
567
95.0
KOGANEI-MIURA
597
540
90.5
KOGANEI-TATEYAMA
597
526
88.1
MIURA-TATEYAMA
597
576
96.5
3. Method
We use baseline length analysis results to evaluate the accuracy of
measurements, because the estimation of baseline length is robust against the
model uncertainty such as earth rotation parameters. Root mean square variance
of continuous 5 samples of baseline lengths by baseline is compared with
adjacent 5 samples in other period. As for the formal error corresponding to 5
sessions, we take a simple average of that of each session concerned. KSP
observation participated by all four stations started in January, 1996.
Initially a video bandwidth of 2 MHz was used. After a system reliability
check, this was expanded to 8 MHz on June 1, 1997. To avoid any influence
that this may cause we limit the period to June 1, 1997 to August 13, 1997
for an evaluation study.
4. Results
Figure 2 presents scatter plots of rms variance of baseline lengths of the 5
sessions, with its formal error defined as a simple average of each session's
formal error. An open triangle in each panel represents a result for 5 sessions
from July 28 to August 1, 1997, i.e., the period of continuous observations. In
each panel rho means correlation coefficient and N is the number of
samples plotted in the figure. Weak correlation between rms variances and
formal errors can be seen. Open triangles, results of 120-hour-observations,
are mostly located at the lower-left edge of the population of samples. This
means that an improvement in both repeatability and formal error can be seen
for the 5 sessions when we compare them with other periods.
Figure 2. Repeatabilities and formal errors by baseline for the period between
June 1, 1997 to August 13, 1997. Open triangles represent data for
120-hour observations. Rho is a correlation coefficients between
parameters.
Table 2. Summary of Comparison.
Baseline
Repeatability (mm)
Mean Formal Error (mm)
for 1997/6/1 -1997/8/13
for 5 sessions
for 1997/6/1 -1997/8/13
for 5 sessions
KASHIMA-KOGANEI
4.3+/-1.5
1.6
2.2+/-0.7
1.2
KASHIMA-MIURA
3.7+/-1.4
2.5
2.1+/-0.5
1.4
KASHIMA-TATEYAMA
4.7+/-1.8
4.1
2.3+/-0.5
1.4
KOGANEI-MIURA
5.2+/-2.5
1.3
2.5+/-0.9
1.1
KOGANEI-TATEYAMA
6.2+/-2.6
1.6
2.7+/-1.0
1.1
MIURA-TATEYAMA
4.6+/-1.9
2.0
2.1+/-0.5
1.0
Table 2 summarizes comparison results between the entire period and 5 sessions
for repeatability and mean formal error.
Repeatability of the 5 sessions varies from 1.3 mm to 2.5 mm. However we can
see a clear improvement in repeatability with one exception in the case of the
Kashima-Tateyama baseline. These results demonstrate that 24-hour observation
in a day gives better repeatability than 6-hour a day observation.
5. Conclusion
We have monitored the deformation around Tokyo metropolitan area by using four
VLBI stations dedicated to this purpose. It is important to know the limitation
of accuracy achieved by the current measurement system for discriminating
anomalies in crustal deformation.
According to an experimental observation lasting over 120 hours which consists
of 5 sessions, the repeatability achieved by the current system is approaching 1
mm for baseline lengths.
We can see faint correlation between repeatability and formal error. This
suggests that it is difficult to improve the repeatability by improving only
the formal error, even though there exists a slight possibility. Improvement
of the formal error is mainly due to system hardware or observation
schedule. It is considered that a physical model, such as propagation delay
model for the atmosphere, is related to an improvement of the repeatability in 5
days. Thus an improvement in a physical model should be rather investigated.
