Journal of Global Positioning Systems (2004)
Vol. 3, No. 1-2: 218-225
Improved atmospheric modelling for large scale high-precision
positioning based on GNSS CORS networks in Australia
Craig Roberts[2], Kefei Zhang[1], Chris Rizos[2], Allison Kealy[3], Linlin Ge[2], Peter Ramm[4], Martin Hale[4],
Doug Kinlyside[5] and Paul Harcombe
[5]
School of Surveying and Spatial Information Systems, University of New South Wales
[1] School of Mathematical and Geospatial Sciences, RMIT University
[2] School of Surveying and Spatial Information Systems, UNSW
[3] Department of Geomatics, The University of Melbourne
[4] Spatial Information Infratructure, Dept of Sustainability and Environment, Victoria Government
[5] Department of Lands, NSW Government
Received: 15 Nov 2004 / Accepted: 3 Feb 2005
Abstract. This contribution describes a recent Australian
Research Council (ARC) project funded under the ARC-
Linkage Scheme. The research team comprises
researchers from RMIT University, UNSW, University of
Melbourne, Spatial Information Infrastructure and the
Department of Lands, NSW. The aim of the project is to
enhance the utility of continuously operating reference
station (CORS) networks in the states of Victoria and
New South Wales by developing improved atmospheric
correction models to support high accuracy, real-time
positioning even when the density of reference stations is
insufficient for standard operational GPS techniques such
as RTK (‘real-time kinematic’). Many applications of
Global Navigation Satellite System (GNSS) technology,
such as surveying, mapping and precise navigation,
require real-time positioning accuracies to centimetre
levels. To support these applications, many countries are
establishing dense CORS networks with stations,
positioned typically a few tens of kilometres apart.
However, for Australia with its large and sparsely
populated landmass, such dense networks cannot be
justified economically. This ARC project will investigate
enhancements of sparse networks to maintain similar
levels of accuracy as dense CORS networks. It will seek a
better understanding and modelling of atmospheric
conditions, currently a major limitation in the use of
sparse networks for high accuracy techniques. This paper
will describe the status of current developments in CORS
network infrastructure in Australia, namely GPSnet in
Victoria and SydNet in New South Wales. The major
research components of the project will be outlined and
the technical and practical challenges will be discussed,
including some methodologies that will be investigated.
Key words: CORS, GNSS, GPS, network RTK.
1. Introduction
The standard mode of high accuracy differential
positioning requires one Global Positioning System
(GPS) receiver to be located at a “reference station” with
known coordinates, while the second “user” receiver
simultaneously tracks the same satellite signals. When the
carrier phase measurements from the two receivers are
combined and processed, the mobile user’s receiver
coordinates are determined relative to the reference
receiver. This can be done in real-time, if the reference
receiver data is transmitted to the user’s receiver, even
while the receiver is moving. The ultimate
implementation of such a technique is known as “real-
time kinematic” (RTK), and is capable of cm-level
accuracy under certain constrained operational conditions
(Rizos, 2002a). Most precise land positioning
applications have required the user to set-up their own
reference station (and if real-time operations were
required, the data communication link between reference
and user receiver as well). An alternative to this “do-it-
yourself” approach is to take advantage of a CORS
network of receivers. This is more convenient for users,
as they are able to position using only a single receiver,
relying instead on a service provider to operate the
reference receivers for them. Users may have a variety of
accuracy requirements, so CORS networks must be able
to cater for the most demanding users, typically those

Roberts et al: Improved atmospheric modelling for large scale high-precision positioning 219
seeking cm-level accuracy. However, CORS networks
also permit innovative network-based techniques to be
used, and this is discussed later in this section.
High quality GPS reference stations have been
established, in a sparse global network, since the late
1980s to support scientific applications such as
tectonic/earthquake hazard research, geodetic reference
system definition and maintenance, and atmospheric
studies. These stations are typically located hundreds, or
even thousands, of kilometres apart. Recently, networks
of CORS are being established to support demanding
navigation/positioning applications such as surveying and
mapping, engineering machinery guidance, precision
farming, harbour fleet management, rescue and
emergency services, and structural monitoring. By
improving the availability of reference station data for
users that demand high positioning accuracy (and
integrity), in real-time, the number and variety of
applications can grow rapidly. CORS networks are
therefore critical infrastructure enabling the basic utility
of high precision positioning to a diverse range of users.
Regional CORS networks are currently being established
in many countries as a foundation for the establishment of
a Spatial Data Infrastructure (Brown et al, 2002; Zhang et
al, 2001; Rizos & Han, 2003). The distribution and
density of a CORS network is constrained by the
establishment costs per reference station, the areal extent
to be serviced and the positioning accuracy requirements
of the most demanding users (it is assumed that if the
highest accuracy requirements are satisfied, all less
demanding requirements can be easily met). Ideally, a
dense network of reference receivers would be
established and able to satisfy cm-level user accuracy
requirements. Existing CORS networks in, for example,
Germany, UK and Japan are sufficiently dense to restrict
the maximum baseline length between a user and nearby
reference station to well under 40km, which is often
sufficient for cm-level accuracy techniques based on a
single reference station, using high quality (dual-
frequency) receivers that permit rapid “ambiguity
resolution” (AR). However, as the inter-receiver distance
increases, the residual atmospheric biases (due to
differential ionospheric and tropospheric delay of the
GPS satellite signals) in the double-differenced GPS
observables (the standard observation model used for
carrier phase-based positioning – Dedes & Rizos, 2002)
increases, making AR more difficult (and even
impossible using current rapid positioning techniques).
Hence this distance constraint for rapid AR makes
accurate positioning with respect to sparse CORS
networks problematic. This has profound ramifications in
Australia due to its large areal extent and relatively
sparse population.
The aim of this project is to enhance the utility of CORS
networks in Australia, by developing new GPS
atmospheric correction models to support cm-level, real-
time positioning even when the density of reference
stations is insufficient for standard operational
techniques.
2. Status of CORS networks in Australia
In the early 1990s, the Australian Regional GPS Network
was established to support geodetic applications, and now
consists of 16 reference stations on mainland Australia
(with an average spacing of one thousand kilometres).
A CORS network to support surveying, mapping and
high-end navigation users was first established in 1994
when Victoria’s GPSnet was deployed (Hale, 2000). It
has since grown to 19 reference stations with a regional
and rural network coverage and a denser Melbourne and
Environs network providing online GPS data access
(www.land.vic.gov.au/GPSnet). Users can combine
these files with GPS data collected across Victoria, for
post-processing to obtain cm-level accuracy position
results (see Figure 1). To date, this is the only state-wide
CORS network in Australia.
Figure 1 Current Status of Victoria’s Rural and Regional GPSnet
network infrastructure
CORS networks around the world will be progressively
upgraded to provide a real-time capability for high-end
users (Rizos, 2002b). This is also true of Victoria’s
GPSnet, with eight stations currently broadcasting
(single-reference) RTK corrections to local users.
However, the distance-dependent atmospheric biases
referred to earlier, restrict the length of GPS baselines,
thereby limiting the applications of RTK techniques to
within 10-15km of a broadcasting reference station. As
the average reference station spacing in GPSnet is of the
order of 100km (except in the Melbourne area where the
spacing decreases to about 50km), there remain large
areas of Victoria where RTK cannot be used (Roberts et
al, 2003).

220 Journal of Global Positioning Systems
Ibid (2003) proposes developing GPSnet from a
“passive” CORS network, that is where users must
download data from an archive site for post processing, to
an “active” network (data is broadcast to users) in a three
stage process. The first stage is already evolving and
entails simply broadcasting base station data via
UHF/VHF radio as standalone RTK units to local users
within a range of up to 20km.
Stage two is part of a new Location Based Services
(LBS) Positioning System project currently underway
and being implemented by Spatial Information
Infrastructure of the Department of Sustainability and
Environment. Six well distributed base stations will
provide DGPS correction messages to users across
Victoria thereby providing better than 0.5m horizontal
positioning in real-time (see Figure 2). This service is
planned to go live wirelessly to GPSnet cooperative
hosts, contributors and partners in December 2004 and to
other users by June 2005.
Figure 2 Victoria’s GPSnet with LBS Positioning System stations
generating a networked, statewide DGPS correction message which is
then broadcast to users
The most challenging aspect of implementing such an
Internet-based DGPS system is connecting base stations
and the GPSnet Central Server Cluster in a manner that is
technically and economically sustainable. Access for
users will be via mobile Internet ie GPRS via
BluetoothTM to a GPS device. There will be up to 10
different types of user selectable solution types from
DGPS RTCM through to networked CMR (Compact
Measurement Record – Trimble Proprietary message
format) and FKP emulation (surface correction parameter
– FKP (translated from German)). The important aspect
of this service is that the user makes the selection
(Millner, 2004).
Ultimately by mid 2005 an estimated 21 GPSnet base
stations will able to be connected to the GPSnet Central
Server Cluster. Having stations connected to an active
network will ease the transition to stage three: cm-level
positioning across Victoria using full network RTK
correction generation.
Since the mid-1990s much research has been invested in
developing reference network techniques to map regional
atmospheric conditions (and other unmodelled double-
differenced biases). These network-based techniques
effectively extend the baseline length to support medium-
range RTK surveying, even up to 70km or more in mid-
latitude areas (Rizos & Han, 2003). The distribution of
GPSnet reference stations provides a suitable
infrastructure to investigate network-based RTK over a
significant regional extent.
Figure 3 Proposed coverage map for the SunPOZ network in SE
Queensland
First generation commercial network-based RTK systems
have been available over the last few years. A so-called
Virtual Reference Station (VRS) test-bed network has
been operating for several years in the Ipswich region,
south of Brisbane (see Figure 3). This network, known as
SunPOZ, uses Trimble’s proprietary VRS software (and
their hardware) to provide cm-level positioning in the
region. However, this CORS infrastructure is not
challenging the constraints that affect RTK, as the longest
distance between reference and user receivers is of the
order of only 30km (Higgins, 2002).
The latest CORS infrastructure in Australia is SydNet,
currently being deployed in the Sydney basin area (see
Figure 4). SydNet will consist of 10-15 reference stations

Roberts et al: Improved atmospheric modelling for large scale high-precision positioning 221
and will service the eastern seaboard from Wollongong to
Newcastle (Rizos et al, 2003). Like the SunPOZ VRS
network, SydNet will be a CORS network designed from
the very beginning to support real-time operations, in
contrast to GPSnet, which, due to its initial development
from 1994, is evolving toward real-time operations.
However, the high density of SydNet stations is only
intended to support research into new network-based GPS
positioning techniques. This is possible because of the
ability to “thin” the CORS network, and to use some
SydNet stations as simulated user receivers for check
purposes. Hence, although SydNet is an ideal test network
for the proposed research, it is not a blueprint for a state-
wide CORS network. The extension of the GPS network
across the state of New South Wales would be based on
receiver spacings similar to those of Victoria’s GPSnet
(or even greater).
Figure 4 Current SydNet configuration
SydNet will provide network generated RTK corrections
using the in-house software developed in cooperation
with Nanyang Technical University (NTU), Singapore
and currently in operation on the Singapore Integrated
Multiple Reference Station Network (SIMRSN) (Goh,
2002). Corrections will be archived on an Internet server
housed at the Network Control Centre (NCC) hosted by
ac3 (host company) at the Australian Technology Park,
Redfern. This is a secure site guaranteeing reliable
operations for users. The base stations are located at the
state government Rail Corporation owned train stations
and linked to the NCC via the existing optical fibre
network owned by the telecommunications provider
Argus. In the first instance, users will access the
correction messages from the Internet via a GPRS
connection using CDMA, 3G or W-LAN. Other methods
of broadcasting correction messages to users will be
investigated as SydNet evolves. An initial seven station
subset of SydNet will be launched at the GNSS2004
conference in Sydney.
GPSnet, SunPOZ and SydNet all provide a public utility
function supported by their respective state government
departments (Department of Sustainability and
Environment, Dept of Natural Resources, Mines &
Energy, Qld and Dept of Lands, NSW respectively).
However only SydNet (and to a lesser extent GPSnet) will
perform a dual function as a research test-bed and public
utility. SunPOZ is a proprietary system and relies on the
manufacturer (Trimble) for further advances and
enhancements.
3. Motivation and Aims for Research Project
CORS networks have already been established in many
countries, and these are likely to be upgraded to provide
cm-level positioning accuracy to high-end users, even in
real-time. One challenge is to make the spacing of
reference receivers as great as possible, without
sacrificing user accuracy and reliability. Most CORS
networks currently offering network-RTK services have
reference station spacings of a few tens of kilometres at
most. Internationally, research is being undertaken to
extend the station spacing, primarily through improved
atmospheric (ionospheric and tropospheric) bias
modelling. Achieving cm-level positioning accuracy over
distances of 70km or more has not yet been verified on an
operational basis (though university studies have
demonstrated promising results for short trials with
optimised research software). Using the commercially
available Trimble VRS system, Higgins (2002) and
Retscher (2002) have reported horizontal positioning
accuracy of the order of ±5cm, for baselines up to 35km
in length.
Other challenges include improving the stochastic models
for network-based positioning, extending the data
modelling algorithms to new satellite signals [i.e. “GPS
modernization” (McDonald, 2002); and GALILEO
(Galileo, 2003)], determining which is the “best”
correction message formats, and outputting the bias
modelling results in appropriate forms for non-
positioning users. Australia has already made significant
contributions to CORS research since the late-1990s (see
Rizos & Han, 2003, for an overview), and with access to
Victoria’s GPSnet and NSW’s SydNet further
improvements to real-time network positioning
techniques can be expected. The fundamental research
facilities to be used in this project represent a significant
infrastructure investment by the two government partners,
Department of Sustainability and Environment and Dept
of Lands, NSW. The research team has identified several
research areas, primarily concerned with improving or
extending current GPS atmospheric bias models to;

222 Journal of Global Positioning Systems
• Increase the inter-receiver distances for real-time
(or post-mission), rapid GPS carrier phase-based
positioning.
• Take advantage of the extra signals to be
transmitted by the “modernized” GPS
constellation (first satellites to be launched in
2005), and eventually the new GALILEO
system (planned deployment to commence later
in the decade).
• Permit more realistic network stochastic models
to be provided to users, hence improving the
quality or integrity information associated with
high accuracy GPS positioning results.
• Develop tropospheric signal bias “maps” as a
by-product of CORS data processing, so as to,
for example, correct differential Interferometric
Synthetic Aperture Radar (InSAR) results of
ground subsidence/deformation.
• Demonstrate the feasibility of a CORS network
to monitor the ionospheric and tropospheric
“weather” conditions, so contributing to
synoptic atmospheric studies.
4. Research Road Map
To achieve the aims of this project the following distinct
research tasks have been identified:
1. Spatio-temporal models for ionospheric and
tropospheric bias variability for incorporation
into “second generation” network-RTK systems.
2. Validation of the aforementioned models, and
techniques for interpolation of atmospheric
biases at a user’s location.
3. Stochastic models appropriate for network-based
positioning (for real-time and post-mission
modes of operation).
4. Extension of bias modelling for the additional
signal frequencies to be transmitted by future
Global Navigation Satellite Systems.
5. Investigations to determine the appropriate
message format for transmitting such model
information to users in real-time.
6. Tropospheric bias model products for non-
positioning applications, such as for
meteorological and remote sensing studies.
These research tasks are expanded as follows.
4.1 Models of Ionospheric and Tropospheric Bias
Variability
Amongst the issues to be investigated for research
activity into spatio-temporal models for ionospheric and
tropospheric signal delay are:
• The applicability of current global International
GPS Service (IGS) ionosphere and troposphere
models;
• The applicability of ionospheric models derived
from the U.S. Wide Area Augmentation System
(WAAS);
• The variability of ionospheric delay determined
using the dual-frequency instrumentation of
GPSnet; and
• The use of GPSnet and SydNet data to estimate
the variability of the tropospheric delay
(primarily wet component) from long time series
of GPS data.
The outcome will be improved atmospheric bias models
appropriate for network-based positioning, as well as
other non-positioning applications. To support this, the
Department of Sustainability and Environment,
Government of Victoria is currently making a significant
capital investment in infrastructure for a real-time,
network-positioning, wireless upgrade to GPSnet,
through which these models can be implemented. This
will have a critical impact on how state-wide CORS
networks can be designed and implemented, in Victoria,
NSW and other Australian states.
4.2 Validation of Atmospheric Models for Network-
RTK
Validation of the spatio-temporal models for ionospheric
and tropospheric signal delay developed in Sect 4.1, as
well as testing and developing interpolation algorithms
will encompass such activities as:
• The use of GPSnet to validate the ionospheric
models using data from extra dual-frequency
GPS receivers placed at interpolation (check)
points;
• The use of SydNet to validate the tropospheric
models using data from selected interpolation
(check) points (as the SydNet will be dense
enough to model horizontal tropospheric
inhomogenieties), and
• The investigation of appropriate interpolation
techniques (such as those identified by Dai et al,
2003).

Roberts et al: Improved atmospheric modelling for large scale high-precision positioning 223
The outcome will be validated atmospheric bias models
appropriate for network-based positioning, which can be
tested within network-RTK software such as that
currently being implemented in SydNet. (SydNet will use
the network-based positioning algorithms jointly
developed by UNSW and Nanyang Technological
University, and implemented in the Singapore Integrated
Multiple Reference Station Network – Chen et al, 2000;
Rizos, 2002b).
4.3 Stochastic Models for Network-Based Positioning
Research activity into stochastic modelling will
investigate:
• The impact of neglecting rigorously-derived
stochastic information on both network data
processing and user position determination;
• Appropriate simplified stochastic models; and
• The “packaging” of such stochastic information
into suitable network-RTK messages.
The outcome will be proposals for how such improved
stochastic modelling can be implemented within
operational network-RTK software, with initial tests to be
carried out on SydNet with the modified network-RTK
software described by Chen et al (2000). This work will
be based around some initial research in this area reported
by Musa et al (2003).
4.4 Atmospheric Bias Modelling for Next Generation
GNSS
This work will concentrate primarily on research activity
into ionospheric delay modelling for next generation
GNSS, and will investigate:
• The combinations of triple-frequency signals
from the modernized GPS’s L1, L2, L5, and
GALILEO’s multiple-frequencies that are best
suited for determining ionospheric bias
information, and
• The appropriate “transition” arrangements that
can be made to continue operations in current
(and upgraded) CORS networks as new satellites
are launched.
Much of this research will be undertaken using simulated
data generated using simple error models as implemented
in the Bernese geodetic software, Matlab-type tools as
well as via signal simulators. The outcome will be
proposals for how network-based positioning will
“evolve” as the new GNSS satellites are launched from
2005 through to the end of this decade.
4.5 Algorithms for Network-RTK Message Formats
This task will investigate how the network-derived
correction data will be “packaged” for users. There is
considerable debate concerning the RTCM-type messages
for network-RTK (Rizos & Han, 2003), and there are
several candidates. Amongst the formats to be
investigated are:
• The Virtual Reference Station (VRS)-based
approach of user-customised reference station
data (as implemented within Trimble’s VRS
system), that utilises the user’s location;
• The Virtual Reference Cell (VRC) approach that
utilises the user’s approximate location
(Retscher, 2002);
• The “Area Correction Parameter” (surface
correction parameter - FKP) approach sends out
the entire correction model (Wübbena & Bagge,
2002), permitting the user to then perform the
bias interpolation themselves; and
• Sending all raw data to the user, and allowing
them to select subsets of the data for integrity
checking.
The outcome will be recommendations to the “best”
approach, as well as a suite of algorithms from which the
industry-endorsed format can be implemented within the
network-RTK user software.
4.6 Tropospheric Models for Non-Positioning
Applications
Research activity into tropospheric model products will
investigate:
• Areal models of tropospheric delay correction
(for the wet and dry components) for differential
InSAR (DInSAR), and
• Time-sliced (synoptic) areal models for possible
assimilation into meteorological systems.
The outcome will be a methodology for mapping the
tropospheric delay from CORS networks, so that it can be
used to correct DInSAR-derived ground subsidence maps
described in Janssen et al (2004). Tests of the
tropospheric delay maps derived from GPSnet data on
DInSAR results for areas of the Gippsland that are
undergoing subsidence due to offshore oil extraction.
Research is currently underway in this area. The effect of
neglecting differential tropospheric signal delay in
DInSAR results is being assessed (Ge et al, 2001).

224 Journal of Global Positioning Systems
3. Concluding Remarks
CORS networks in other countries are sufficiently dense
to restrict the maximum baseline length between a user
and nearby reference station to well under 40km. Under
typical operational conditions this is usually adequate for
cm-level positioning accuracy. Due to Australia’s large
areal extent and relatively sparse population, the number
of reference stations required for a state-wide network
using existing techniques would therefore be
uneconomical. One option would be to, rather than offer a
state-wide service, provide the necessary station density
within regions with high population densities. However,
in Australia some of the most important users of satellite
positioning technology are in precision agriculture and
mining applications, and require infrastructure across
large rural areas. This research is therefore of national
significance as it will provide an infrastructure that
benefits activities for rural and regional communities. To
service these users adequately the state governments need
to implement the outcomes proposed in this research
project in order to capitalise on sparse CORS
infrastructure.
The primary outcome of the proposed research will be
enhanced CORS operational algorithms and procedures,
extending the capabilities of current commercially
available network-based positioning systems such as
VRS. Such a refinement will lead to “second generation”
CORS networks, with sparser spacing of reference
stations than current “first generation” VRS systems,
without sacrificing performance (accuracy, time-to-AR,
reliability, availability and integrity).
Acknowledgements: The authors would like to
gratefully acknowledge the research funds granted for
this project through the Australian Research Council
Linkage project LP0455170.
References
Brown, N., Kealy, A. & Williamson, I. (2002) Issues related to
GPS phase observations for improved quality estimation,
The Journal of Cartography, Australia, 31(2), 143-151.
Chen, X., Han, S., Rizos, C. & Goh, P.C. (2000) Improving
real-time positioning efficiency using the Singapore
Integrated Multiple Reference Station Network
(SIMRSN), 13th Int. Tech. Meeting of the Satellite
Division of the U.S. Inst. of Navigation, Salt Lake City,
Utah, 19-22 September, 9-18.
Dai, L., Han, S., Wang, J., Rizos, C. (2002) Comparison of
interpolation algorithms in network-based GPS
techniques, Journal of Navigation, 50(4), pp 277 – 293.
Dedes, G. & Rizos, C. (2002) GPS positioning models for
single point and baseline solutions, Chapter 9 in Manual
of Geospatial Science and Technology, J. Bossler, J.
Jenson, R. McMaster & C. Rizos (eds.), Taylor & Francis
Inc., ISBN 0-7484-0924-6, 114-126.
Galileo (2003)
http://europa.eu.int/comm/dgs/energy_transport/galileo/ind
ex_en.htm
Ge, L., Chen, H.Y., Han, S. & Rizos, C. (2001) Integrated GPS
and Interferometric SAR techniques for highly dense
crustal deformation monitoring, 14th Int. Tech. Meeting
of the Satellite Division of the U.S. Inst. of Navigation,
Salt Lake City, Utah, 11-14 September, 2552-2563.
Goh, P.C. (2002) Development of an Integrated Multiple Base
Station Infrastructure to support Concurrent High
Precision Differential Positioning Applications, Nanyang
Technical University, Internal Report.
GPSnet (2004) http://land.vic.gov.au/GPSnet
Hale, M. (2000) GPSnet: The next decade, paper presented at
the Survey Conference of the Institution of Surveyors
Victoria and the Institution of Engineering Surveyors
Australia, Victoria Division, 27 May.
Higgins, M. (2002) Australia’s changing surveying
infrastructure from marks in the ground to virtual
reference stations, FIG XXII International Congress,
Washington, 19-26 April, TS5.6.
Janssen, V., Ge, L. & Rizos, C. (2004) Tropospheric
corrections to SAR interferometry from GPS
observations, GPS Solutions, 8(3), 140 – 151.
McDonald, K. (2002) The Modernization of GPS: Plans, new
capabilities and the future relationship to Galileo, Journal
of Global Positioning Systems, 1(1), 1-17.
Millner, J. (2004) GPSnet Location Based Services Positioning
System Project Manager, Personal communication
Musa, T.A., Wang, J., Rizos, C. & Satirapod, C. (2003)
Stochastic modelling for network-based GPS positioning,
6th Int. Symp. on Satellite Navigation Technology
Including Mobile Positioning & Location Services,
Melbourne, Australia, 22-25 July, CD-ROM proc., paper
41.
Retscher, G. (2002) Accuracy performance of virtual reference
station (VRS) networks, Journal of Global Positioning
Systems, 1(1), 40-47.
Rizos, C. (2002a) Making sense of the GPS techniques,
Chapter 11 in Manual of Geospatial Science and
Technology, J. Bossler, J. Jenson, R. McMaster & C.
Rizos (eds.), Taylor & Francis Inc., ISBN 0-7484-0924-6,
146-161.
Rizos, C. (2002b) Network RTK research and implementation:
A geodetic perspective, Journal of Global Positioning
Systems, 1(2), 144-150.
Rizos, C., Kinlyside, D., Yan, T., Omar, S. & Musa, T.A.
(2003) Implementing network RTK: The SydNet CORS
infrastructure, 6th Int. Symp. on Satellite Navigation
Technology Including Mobile Positioning & Location
Services, Melbourne, Australia, 22-25 July, CD-ROM
proc., paper 50.

Roberts et al: Improved atmospheric modelling for large scale high-precision positioning 225
Rizos, C. & Han, S. (2003) Reference station network based
RTK systems - Concepts & progress, Wuhan University
Journal of Nature Sciences, 8(2B), 566-574.
Roberts, C., Zhang, K., Hale, M. & Millner, J. (2003) Towards
real-time network-based RTK corrections for Victoria’s
GPSnet, 6th Int. Symp. on Satellite Navigation Technology
Including Mobile Positioning & Location Services,
Melbourne, Australia, 22-25 July.
Wübbena, G. & Bagge, A. (2002) RTCM Message Type 59–
FKP for transmission of FKP, Version 1.0, Geo++ White
Paper Nr. 2002.01, www.geopp.de/download/geopp-rtcm-
fkp59.pdf
Zhang, K., Talbot, N., Hale, M. & Millner, J. (2001) Victorian
high precision permanent GPS tracking network system,
14th Int. Tech. Meeting of the Satellite Division of the U.S.
Inst. of Navigation, Salt Lake City, Utah, 11-14 September,
3077-3085.