The Multi-functional Satellite
Augmentation System (MSAS) is a
Japanese Satellite Based Augmentation
System (SBAS) [1] which provides
Differential GPS (DGPS) corrections
and is designed to supplement GPS by
reporting and monitoring the reliability
and accuracy of GPS signals in real
time. MSAS consists of two satellites,
(MTSAT-1R at 140E longitude and
MTSAT-2 at 145E longitude), four
Ground Monitor Stations (GMS), two
Master Control Stations (MCS), and two
Monitor and Ranging Stations [1]. The
MSAS architecture is shown in Figure 1.
Although it is designed for aircraft, the
signals also can be received and used on
land and sea. The primary MSAS coverage
area includes a wide signifi cant range
of latitude (25N to 45N in geographic
latitude; 15N to 35N in magnetic latitude)
[2]. Similar services are provided in US
by the Wide Area Augmentation System
(WAAS) and in Europe by the European
Geostationary Navigation Overlay Service
(EGNOS). Through the use of SBAS, an
individual GPS receiver is able to remove
correlated errors from its measurements
and thus, obtain much better accuracy.
The aim of this study is to evaluate MSAS
performance under severe ionospheric
conditions that were observed from 2001
to 2003, when the ionosphere was more
active because the 11-year solar cycle
was near its peak. Single-Frequency
PPP (SF-PPP) was used instead of using
MSAS correction data because MSAS was
not yet fi elded during the above period.
Another reason that Single-Frequency
PPP for was used for our evaluation was
that the performance of MSAS and SFPPP
was approximately same in terms
of theoretical error models. Many papers
have been published for both Single-
Frequency and Dual-Frequency PPP
algorithm, while several papers related
to pseudo-range based Single-Frequency
PPP have also been presented [3]. In
Section 2, the relationship between GNSS
and solar activity is briefl y described.
In Section 3, the outline of MSAS and
Single-Frequency PPP are described from
the view point of error models. In Section
4, the actual performance of MSAS and
SF-PPP were verifi ed using raw data
from 2008. From this data, it is found
that, as expected, MSAS and SF-PPP
performance were approximately the same.
SF-PPP positioning performance was then
investigated using raw data from 2001
to 2003, during the active solar period.
Figure 1. MSAS System Architecture
GNSS and solar activity
It is well-known that ionosphere is the
largest error source in positioning for
single-frequency GNSS users. Since solar
activity has a strong correlation with
ionosphere activity, positioning error can
increase signifi cantly at low latitudes
(and elsewhere, to a lesser degree) during periods of increased solar activity. During
these periods, ionospheric delays at the
L1 frequency are normally 5 to 15 meters
in the zenith direction, and the delay can
reach 100 meters at peak periods. Figure
2 shows the typical solar cycle given by
NASA (http://weather.msfc.nasa.gov/).
Major ionospheric effects correspond to
the rise and fall in the number of sunspots.
At present, we are in near solar minimum.
In this paper, MSAS performance during
solar maximum is estimated by using raw
data from the previous solar maximum.
Error models [4]
For standalone GPS point positioning after
the deactivation of Selective Availability
(SA) in 2000, the dominant error sources
consist of satellite orbits and clocks
and, ionosphere and troposphere delays,
resulting in 2serrors of about 10 meters
horizontally and 15 meters vertically.
Satellite orbits and clock corrections
are broadcast inside the GPS navigation
messages and are typically accurate to
1 - 3 meters. Ionospheric delay varies
with time and season up to 15 – 30
meters at zenith and is about 3-4 times
larger at low elevation angles. Using the
Klobuchar ionospheric model together
with the parameters included in the
navigation messages can remove about
50 % of the ionospheric delay. The other
part of the atmosphere, the troposphere,
causes a delay of about 2.5 meters at
zenith and can be as large as 20 meters
at a low elevation angle. However, this
can be reduced greatly by using one
of the various tropospheric models. In
the following sections, two approaches
to reduce these errors are shown. The
fi rst method is MSAS, and the second
method is Single-Frequency PPP. In this
paper, only pseudo-range measurements
and corrections are used for MSAS and
Single-Frequency PPP positioning.
MSAS [5]
In order to support a large service
area, MSAS provides vector correction
information, which consisting of separate
correction messages that include satellite
clock, satellite orbit, and ionospheric
propagation delay. The MSAS 250-bps message contains an 8-bits preamble,
a 6-bit message type ID, and a 24-bits
CRC parity check code. The remaining
212 bits are defi ned with respect to each
message type. When we look at the
messages to correct user errors models,
several message types are particularly
important. Message Types 2-5 contain
fast corrections to satellite clock that
are updated frequently. Message Type
25 contains long-term corrections to
satellite orbit and clock. Message Type
26 provides ionospheric corrections by
specifying the MSAS-estimated vertical
delay in meters at ionospheric grid points
(IGP) located every 5 degrees in latitude
and longitude. User receivers perform
spatial bilinear interpolation and verticalto-
slant conversion using algorithms
defi ned by the SBAS SARPs to obtain the
estimated line-of-sight delay at the nearby
user’s ionospheric pierce point (IPP) [6].
For SBAS, tropospheric correction is
not broadcast; hence they are computed
using a pre-defi ned model. Averaged
meteorological parameters are used for
this tropospheric correction, and a simple
mapping function is used to calculate
the line-of-sight tropospheric error.
Single-Frequency Precise Point
Positioning (PPP) [3]
Since 1994, The International GNSS
Service (IGS), has been providing several
types of precise satellite orbits and
clocks, including Ultra Rapid, Rapid, and
Final ephemerides. They are different in
latency and accuracy but have the same
sampling interval of 15 minutes. The
“Final” product was used in this paper
because it has the best performance. Its
accuracy is within 5cm and 0.1 ns in
both the satellite orbit and the satellite
clock. All these products are distributed
in SP3 format in which combined satellite
positions and clocks (http://igscb.jpl.nasa.
gov/components/prods.html). In order to
correct for antenna phase-center offset,
estimates of the offsets in each satellite
were obtained from the internet (ftp://
igscb.jpl.nasa.gov/pub/station/general/).
As noted earlier, the ionosphere is
the dominant source of error after the
elimination of S/A. Although dualfrequency
users can easily resolve this problem by directly estimating ionospheric
delay, ionospheric models still need to
be developed for single-frequency users.
Since 1996, Global Ionospheric maps
(GIMs) have been available from the
Center for Orbit Determination in Europe
(CODE). In GIMs, the total electron
content (TEC) of the ionosphere is given as a 2-dimensional map that is supplied
in IONosphere map EXchange (IONEX)
format. The resolution of the map is 5º in
longitude and 2.5º in latitude. Each daily
IONEX fi le contains 13 maps with 2-hour
time spacing (ftp://ftp.unibe.ch/aiub/
CODE/). From GIMs, only the vertical
TEC (VTEC) can be derived. Hence, a
mapping function is deployed to map
the vertical TEC value to the slant TEC
value. This technique is based on a singlelayer
model which models the ionosphere
as a shell of infi nitesimal thickness.
Although the accuracy of ionospheric
estimation strongly depends on the solar
activity, the standard accuracy is within a
decimeter under quiet solar conditions.
The Saastamoinen model was used to
estimate tropospheric delay in this paper
[4]. This model was derived using gas laws
and simplifying assumptions regarding
changes in pressure, temperature, and
humidity with altitude and estimates two
separate components of tropospheric delay.
The “dry delay” in the zenith direction
comprises most of the delay and can
be predicted with an accuracy of a
few millimeters from accurate surface
pressure measurements. The smaller “wet
delay” depends upon the distribution of
water vapor along the signal path and
is much more variable. Use of average
meteorological conditions rather than
actual measurements introduces additional
modeling errors in both dry and wet
delays, resulting in a total zenith delay
error of 5-10 cm. One of the mapping
functions was used to calculate the lineof-
sight tropospheric error [4]. In this
paper, L1-L2 biases are compensated
because satellite orbits and clocks
always refer to the ionosphere-free linear
combination between L1 and L2 codes.
Comparisons
Table 1 shows an error model comparison
between MSAS and Single-Frequency
PPP. There are few differences for
the error models regarding accuracy.
Figure 3 shows the expected degree
of error mitigation for both methods.
To generate Vertical TEC (VTEC)
corrections, MSAS uses 6 sites in Japan
and GIMs uses fewer. From a theoretical
point of view, MSAS performance
should be slightly better than Single-
Frequency PPP, but Single-Frequency
PPP can be used to estimate MSAS
performance with fairly good accuracy.
Testing and results
First, MSAS performance during the recent
solar-minimum period was evaluated.
Single-Frequency PPP performance was
also evaluated using the same period fro
comparison. Next, Single-Frequency PPP
performance was evaluated using raw-data
from the solar-maximum period. Finally,
dual-frequency-based ionospheric error
estimation was tested using the raw -data
from a known large solar-fl are event.
|